diff options
author | 2005-07-27 11:44:44 -0700 | |
---|---|---|
committer | 2005-07-27 16:26:01 -0700 | |
commit | 51533b615e605d86154ec1b4e585c8ca1b0b15b7 (patch) | |
tree | 4a6d7d8494d2017632d83624fb71b36031e0e7e5 /arch/cris | |
parent | [PATCH] CRIS update: updates for 2.6.12 (diff) | |
download | linux-51533b615e605d86154ec1b4e585c8ca1b0b15b7.tar.gz linux-51533b615e605d86154ec1b4e585c8ca1b0b15b7.tar.bz2 linux-51533b615e605d86154ec1b4e585c8ca1b0b15b7.zip |
[PATCH] CRIS update: new subarchitecture v32
New CRIS sub architecture named v32.
From: Dave Jones <davej@redhat.com>
Fix swapped kmalloc args
Signed-off-by: Mikael Starvik <starvik@axis.com>
Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Diffstat (limited to 'arch/cris')
65 files changed, 20786 insertions, 0 deletions
diff --git a/arch/cris/arch-v32/Kconfig b/arch/cris/arch-v32/Kconfig new file mode 100644 index 000000000000..22f0ddc04c50 --- /dev/null +++ b/arch/cris/arch-v32/Kconfig @@ -0,0 +1,296 @@ +config ETRAX_DRAM_VIRTUAL_BASE + hex + depends on ETRAX_ARCH_V32 + default "c0000000" + +config ETRAX_LED1G + string "First green LED bit" + depends on ETRAX_ARCH_V32 + default "PA3" + help + Bit to use for the first green LED (network LED). + Most Axis products use bit A3 here. + +config ETRAX_LED1R + string "First red LED bit" + depends on ETRAX_ARCH_V32 + default "PA4" + help + Bit to use for the first red LED (network LED). + Most Axis products use bit A4 here. + +config ETRAX_LED2G + string "Second green LED bit" + depends on ETRAX_ARCH_V32 + default "PA5" + help + Bit to use for the first green LED (status LED). + Most Axis products use bit A5 here. + +config ETRAX_LED2R + string "Second red LED bit" + depends on ETRAX_ARCH_V32 + default "PA6" + help + Bit to use for the first red LED (network LED). + Most Axis products use bit A6 here. + +config ETRAX_LED3G + string "Third green LED bit" + depends on ETRAX_ARCH_V32 + default "PA7" + help + Bit to use for the first green LED (drive/power LED). + Most Axis products use bit A7 here. + +config ETRAX_LED3R + string "Third red LED bit" + depends on ETRAX_ARCH_V32 + default "PA7" + help + Bit to use for the first red LED (drive/power LED). + Most Axis products use bit A7 here. + +choice + prompt "Product debug-port" + depends on ETRAX_ARCH_V32 + default ETRAX_DEBUG_PORT0 + +config ETRAX_DEBUG_PORT0 + bool "Serial-0" + help + Choose a serial port for the ETRAX debug console. Default to + port 0. + +config ETRAX_DEBUG_PORT1 + bool "Serial-1" + help + Use serial port 1 for the console. + +config ETRAX_DEBUG_PORT2 + bool "Serial-2" + help + Use serial port 2 for the console. + +config ETRAX_DEBUG_PORT3 + bool "Serial-3" + help + Use serial port 3 for the console. + +config ETRAX_DEBUG_PORT_NULL + bool "disabled" + help + Disable serial-port debugging. + +endchoice + +choice + prompt "Kernel GDB port" + depends on ETRAX_KGDB + default ETRAX_KGDB_PORT0 + help + Choose a serial port for kernel debugging. NOTE: This port should + not be enabled under Drivers for built-in interfaces (as it has its + own initialization code) and should not be the same as the debug port. + +config ETRAX_KGDB_PORT0 + bool "Serial-0" + help + Use serial port 0 for kernel debugging. + +config ETRAX_KGDB_PORT1 + bool "Serial-1" + help + Use serial port 1 for kernel debugging. + +config ETRAX_KGDB_PORT2 + bool "Serial-2" + help + Use serial port 2 for kernel debugging. + +config ETRAX_KGDB_PORT3 + bool "Serial-3" + help + Use serial port 3 for kernel debugging. + +endchoice + +config ETRAX_MEM_GRP1_CONFIG + hex "MEM_GRP1_CONFIG" + depends on ETRAX_ARCH_V32 + default "4044a" + help + Waitstates for flash. The default value is suitable for the + standard flashes used in axis products (120 ns). + +config ETRAX_MEM_GRP2_CONFIG + hex "MEM_GRP2_CONFIG" + depends on ETRAX_ARCH_V32 + default "0" + help + Waitstates for SRAM. 0 is a good choice for most Axis products. + +config ETRAX_MEM_GRP3_CONFIG + hex "MEM_GRP3_CONFIG" + depends on ETRAX_ARCH_V32 + default "0" + help + Waitstates for CSP0-3. 0 is a good choice for most Axis products. + It may need to be changed if external devices such as extra + register-mapped LEDs are used. + +config ETRAX_MEM_GRP4_CONFIG + hex "MEM_GRP4_CONFIG" + depends on ETRAX_ARCH_V32 + default "0" + help + Waitstates for CSP4-6. 0 is a good choice for most Axis products. + +config ETRAX_SDRAM_GRP0_CONFIG + hex "SDRAM_GRP0_CONFIG" + depends on ETRAX_ARCH_V32 + default "336" + help + SDRAM configuration for group 0. The value depends on the + hardware configuration. The default value is suitable + for 32 MB organized as two 16 bits chips (e.g. Axis + part number 18550) connected as one 32 bit device (i.e. in + the same group). + +config ETRAX_SDRAM_GRP1_CONFIG + hex "SDRAM_GRP1_CONFIG" + depends on ETRAX_ARCH_V32 + default "0" + help + SDRAM configuration for group 1. The defult value is 0 + because group 1 is not used in the default configuration, + described in the help for SDRAM_GRP0_CONFIG. + +config ETRAX_SDRAM_TIMING + hex "SDRAM_TIMING" + depends on ETRAX_ARCH_V32 + default "104a" + help + SDRAM timing parameters. The default value is ok for + most hardwares but large SDRAMs may require a faster + refresh (a.k.a 8K refresh). The default value implies + 100MHz clock and SDR mode. + +config ETRAX_SDRAM_COMMAND + hex "SDRAM_COMMAND" + depends on ETRAX_ARCH_V32 + default "0" + help + SDRAM command. Should be 0 unless you really know what + you are doing (may be != 0 for unusual address line + mappings such as in a MCM).. + +config ETRAX_DEF_GIO_PA_OE + hex "GIO_PA_OE" + depends on ETRAX_ARCH_V32 + default "1c" + help + Configures the direction of general port A bits. 1 is out, 0 is in. + This is often totally different depending on the product used. + There are some guidelines though - if you know that only LED's are + connected to port PA, then they are usually connected to bits 2-4 + and you can therefore use 1c. On other boards which don't have the + LED's at the general ports, these bits are used for all kinds of + stuff. If you don't know what to use, it is always safe to put all + as inputs, although floating inputs isn't good. + +config ETRAX_DEF_GIO_PA_OUT + hex "GIO_PA_OUT" + depends on ETRAX_ARCH_V32 + default "00" + help + Configures the initial data for the general port A bits. Most + products should use 00 here. + +config ETRAX_DEF_GIO_PB_OE + hex "GIO_PB_OE" + depends on ETRAX_ARCH_V32 + default "00000" + help + Configures the direction of general port B bits. 1 is out, 0 is in. + This is often totally different depending on the product used. + There are some guidelines though - if you know that only LED's are + connected to port PA, then they are usually connected to bits 2-4 + and you can therefore use 1c. On other boards which don't have the + LED's at the general ports, these bits are used for all kinds of + stuff. If you don't know what to use, it is always safe to put all + as inputs, although floating inputs isn't good. + +config ETRAX_DEF_GIO_PB_OUT + hex "GIO_PB_OUT" + depends on ETRAX_ARCH_V32 + default "00000" + help + Configures the initial data for the general port B bits. Most + products should use 00000 here. + +config ETRAX_DEF_GIO_PC_OE + hex "GIO_PC_OE" + depends on ETRAX_ARCH_V32 + default "00000" + help + Configures the direction of general port C bits. 1 is out, 0 is in. + This is often totally different depending on the product used. + There are some guidelines though - if you know that only LED's are + connected to port PA, then they are usually connected to bits 2-4 + and you can therefore use 1c. On other boards which don't have the + LED's at the general ports, these bits are used for all kinds of + stuff. If you don't know what to use, it is always safe to put all + as inputs, although floating inputs isn't good. + +config ETRAX_DEF_GIO_PC_OUT + hex "GIO_PC_OUT" + depends on ETRAX_ARCH_V32 + default "00000" + help + Configures the initial data for the general port C bits. Most + products should use 00000 here. + +config ETRAX_DEF_GIO_PD_OE + hex "GIO_PD_OE" + depends on ETRAX_ARCH_V32 + default "00000" + help + Configures the direction of general port D bits. 1 is out, 0 is in. + This is often totally different depending on the product used. + There are some guidelines though - if you know that only LED's are + connected to port PA, then they are usually connected to bits 2-4 + and you can therefore use 1c. On other boards which don't have the + LED's at the general ports, these bits are used for all kinds of + stuff. If you don't know what to use, it is always safe to put all + as inputs, although floating inputs isn't good. + +config ETRAX_DEF_GIO_PD_OUT + hex "GIO_PD_OUT" + depends on ETRAX_ARCH_V32 + default "00000" + help + Configures the initial data for the general port D bits. Most + products should use 00000 here. + +config ETRAX_DEF_GIO_PE_OE + hex "GIO_PE_OE" + depends on ETRAX_ARCH_V32 + default "00000" + help + Configures the direction of general port E bits. 1 is out, 0 is in. + This is often totally different depending on the product used. + There are some guidelines though - if you know that only LED's are + connected to port PA, then they are usually connected to bits 2-4 + and you can therefore use 1c. On other boards which don't have the + LED's at the general ports, these bits are used for all kinds of + stuff. If you don't know what to use, it is always safe to put all + as inputs, although floating inputs isn't good. + +config ETRAX_DEF_GIO_PE_OUT + hex "GIO_PE_OUT" + depends on ETRAX_ARCH_V32 + default "00000" + help + Configures the initial data for the general port E bits. Most + products should use 00000 here. diff --git a/arch/cris/arch-v32/boot/Makefile b/arch/cris/arch-v32/boot/Makefile new file mode 100644 index 000000000000..26f293ab9617 --- /dev/null +++ b/arch/cris/arch-v32/boot/Makefile @@ -0,0 +1,14 @@ +# +# arch/cris/arch-v32/boot/Makefile +# +target = $(target_boot_dir) +src = $(src_boot_dir) + +zImage: compressed/vmlinuz + +compressed/vmlinuz: $(objtree)/vmlinux + @$(MAKE) -f $(src)/compressed/Makefile $(objtree)/vmlinuz + +clean: + rm -f zImage tools/build compressed/vmlinux.out + @$(MAKE) -f $(src)/compressed/Makefile clean diff --git a/arch/cris/arch-v32/boot/compressed/Makefile b/arch/cris/arch-v32/boot/compressed/Makefile new file mode 100644 index 000000000000..9f77eda914ba --- /dev/null +++ b/arch/cris/arch-v32/boot/compressed/Makefile @@ -0,0 +1,41 @@ +# +# lx25/arch/cris/arch-v32/boot/compressed/Makefile +# +# create a compressed vmlinux image from the original vmlinux files and romfs +# + +target = $(target_compressed_dir) +src = $(src_compressed_dir) + +CC = gcc-cris -mlinux -march=v32 -I $(TOPDIR)/include +CFLAGS = -O2 +LD = gcc-cris -mlinux -march=v32 -nostdlib +OBJCOPY = objcopy-cris +OBJCOPYFLAGS = -O binary --remove-section=.bss +OBJECTS = $(target)/head.o $(target)/misc.o + +# files to compress +SYSTEM = $(objtree)/vmlinux.bin + +all: vmlinuz + +$(target)/decompress.bin: $(OBJECTS) + $(LD) -T $(src)/decompress.ld -o $(target)/decompress.o $(OBJECTS) + $(OBJCOPY) $(OBJCOPYFLAGS) $(target)/decompress.o $(target)/decompress.bin + +$(objtree)/vmlinuz: $(target) piggy.img $(target)/decompress.bin + cat $(target)/decompress.bin piggy.img > $(objtree)/vmlinuz + rm -f piggy.img + cp $(objtree)/vmlinuz $(src) + +$(target)/head.o: $(src)/head.S + $(CC) -D__ASSEMBLY__ -c $< -o $@ + +# gzip the kernel image + +piggy.img: $(SYSTEM) + cat $(SYSTEM) | gzip -f -9 > piggy.img + +clean: + rm -f piggy.img $(objtree)/vmlinuz vmlinuz.o decompress.o decompress.bin $(OBJECTS) + diff --git a/arch/cris/arch-v32/boot/compressed/README b/arch/cris/arch-v32/boot/compressed/README new file mode 100644 index 000000000000..e33691d15c57 --- /dev/null +++ b/arch/cris/arch-v32/boot/compressed/README @@ -0,0 +1,25 @@ +Creation of the self-extracting compressed kernel image (vmlinuz) +----------------------------------------------------------------- +$Id: README,v 1.1 2003/08/21 09:37:03 johana Exp $ + +This can be slightly confusing because it's a process with many steps. + +The kernel object built by the arch/etrax100/Makefile, vmlinux, is split +by that makefile into text and data binary files, vmlinux.text and +vmlinux.data. + +Those files together with a ROM filesystem can be catted together and +burned into a flash or executed directly at the DRAM origin. + +They can also be catted together and compressed with gzip, which is what +happens in this makefile. Together they make up piggy.img. + +The decompressor is built into the file decompress.o. It is turned into +the binary file decompress.bin, which is catted together with piggy.img +into the file vmlinuz. It can be executed in an arbitrary place in flash. + +Be careful - it assumes some things about free locations in DRAM. It +assumes the DRAM starts at 0x40000000 and that it is at least 8 MB, +so it puts its code at 0x40700000, and initial stack at 0x40800000. + +-Bjorn diff --git a/arch/cris/arch-v32/boot/compressed/decompress.ld b/arch/cris/arch-v32/boot/compressed/decompress.ld new file mode 100644 index 000000000000..3c837feca3ac --- /dev/null +++ b/arch/cris/arch-v32/boot/compressed/decompress.ld @@ -0,0 +1,30 @@ +/*#OUTPUT_FORMAT(elf32-us-cris) */ +OUTPUT_ARCH (crisv32) + +MEMORY + { + dram : ORIGIN = 0x40700000, + LENGTH = 0x00100000 + } + +SECTIONS +{ + .text : + { + _stext = . ; + *(.text) + *(.rodata) + *(.rodata.*) + _etext = . ; + } > dram + .data : + { + *(.data) + _edata = . ; + } > dram + .bss : + { + *(.bss) + _end = ALIGN( 0x10 ) ; + } > dram +} diff --git a/arch/cris/arch-v32/boot/compressed/head.S b/arch/cris/arch-v32/boot/compressed/head.S new file mode 100644 index 000000000000..0c55b83b8287 --- /dev/null +++ b/arch/cris/arch-v32/boot/compressed/head.S @@ -0,0 +1,193 @@ +/* + * Code that sets up the DRAM registers, calls the + * decompressor to unpack the piggybacked kernel, and jumps. + * + * Copyright (C) 1999 - 2003, Axis Communications AB + */ + +#include <linux/config.h> +#define ASSEMBLER_MACROS_ONLY +#include <asm/arch/hwregs/asm/reg_map_asm.h> +#include <asm/arch/hwregs/asm/gio_defs_asm.h> +#include <asm/arch/hwregs/asm/config_defs_asm.h> + +#define RAM_INIT_MAGIC 0x56902387 +#define COMMAND_LINE_MAGIC 0x87109563 + + ;; Exported symbols + + .globl input_data + + .text +start: + di + + ;; Start clocks for used blocks. + move.d REG_ADDR(config, regi_config, rw_clk_ctrl), $r1 + move.d [$r1], $r0 + or.d REG_STATE(config, rw_clk_ctrl, cpu, yes) | \ + REG_STATE(config, rw_clk_ctrl, bif, yes) | \ + REG_STATE(config, rw_clk_ctrl, fix_io, yes), $r0 + move.d $r0, [$r1] + + ;; If booting from NAND flash we first have to copy some + ;; data from NAND flash to internal RAM to get the code + ;; that initializes the SDRAM. Lets copy 20 KB. This + ;; code executes at 0x38010000 if booting from NAND and + ;; we are guaranted that at least 0x200 bytes are good so + ;; lets start from there. The first 8192 bytes in the nand + ;; flash is spliced with zeroes and is thus 16384 bytes. + move.d 0x38010200, $r10 + move.d 0x14200, $r11 ; Start offset in NAND flash 0x10200 + 16384 + move.d 0x5000, $r12 ; Length of copy + + ;; Before this code the tools add a partitiontable so the PC + ;; has an offset from the linked address. +offset1: + lapcq ., $r13 ; get PC + add.d first_copy_complete-offset1, $r13 + +#include "../../lib/nand_init.S" + +first_copy_complete: + ;; Initialze the DRAM registers. + cmp.d RAM_INIT_MAGIC, $r8 ; Already initialized? + beq dram_init_finished + nop + +#include "../../lib/dram_init.S" + +dram_init_finished: + lapcq ., $r13 ; get PC + add.d second_copy_complete-dram_init_finished, $r13 + + move.d REG_ADDR(config, regi_config, r_bootsel), $r0 + move.d [$r0], $r0 + and.d REG_MASK(config, r_bootsel, boot_mode), $r0 + cmp.d REG_STATE(config, r_bootsel, boot_mode, nand), $r0 + bne second_copy_complete ; No NAND boot + nop + + ;; Copy 2MB from NAND flash to SDRAM (at 2-4MB into the SDRAM) + move.d 0x40204000, $r10 + move.d 0x8000, $r11 + move.d 0x200000, $r12 + ba copy_nand_to_ram + nop +second_copy_complete: + + ;; Initiate the PA port. + move.d CONFIG_ETRAX_DEF_GIO_PA_OUT, $r0 + move.d REG_ADDR(gio, regi_gio, rw_pa_dout), $r1 + move.d $r0, [$r1] + + move.d CONFIG_ETRAX_DEF_GIO_PA_OE, $r0 + move.d REG_ADDR(gio, regi_gio, rw_pa_oe), $r1 + move.d $r0, [$r1] + + ;; Setup the stack to a suitably high address. + ;; We assume 8 MB is the minimum DRAM and put + ;; the SP at the top for now. + + move.d 0x40800000, $sp + + ;; Figure out where the compressed piggyback image is + ;; in the flash (since we wont try to copy it to DRAM + ;; before unpacking). It is at _edata, but in flash. + ;; Use (_edata - herami) as offset to the current PC. + + move.d REG_ADDR(config, regi_config, r_bootsel), $r0 + move.d [$r0], $r0 + and.d REG_MASK(config, r_bootsel, boot_mode), $r0 + cmp.d REG_STATE(config, r_bootsel, boot_mode, nand), $r0 + beq hereami2 + nop +hereami: + lapcq ., $r5 ; get PC + and.d 0x7fffffff, $r5 ; strip any non-cache bit + move.d $r5, $r0 ; save for later - flash address of 'herami' + add.d _edata, $r5 + sub.d hereami, $r5 ; r5 = flash address of '_edata' + move.d hereami, $r1 ; destination + ba 2f + nop +hereami2: + lapcq ., $r5 ; get PC + and.d 0x00ffffff, $r5 ; strip any non-cache bit + move.d $r5, $r6 + or.d 0x40200000, $r6 + move.d $r6, $r0 ; save for later - flash address of 'herami' + add.d _edata, $r5 + sub.d hereami2, $r5 ; r5 = flash address of '_edata' + add.d 0x40200000, $r5 + move.d hereami2, $r1 ; destination +2: + ;; Copy text+data to DRAM + + move.d _edata, $r2 ; end destination +1: move.w [$r0+], $r3 + move.w $r3, [$r1+] + cmp.d $r2, $r1 + bcs 1b + nop + + move.d input_data, $r0 ; for the decompressor + move.d $r5, [$r0] ; for the decompressor + + ;; Clear the decompressors BSS (between _edata and _end) + + moveq 0, $r0 + move.d _edata, $r1 + move.d _end, $r2 +1: move.w $r0, [$r1+] + cmp.d $r2, $r1 + bcs 1b + nop + + ;; Save command line magic and address. + move.d _cmd_line_magic, $r12 + move.d $r10, [$r12] + move.d _cmd_line_addr, $r12 + move.d $r11, [$r12] + + ;; Do the decompression and save compressed size in _inptr + + jsr decompress_kernel + nop + + ;; Restore command line magic and address. + move.d _cmd_line_magic, $r10 + move.d [$r10], $r10 + move.d _cmd_line_addr, $r11 + move.d [$r11], $r11 + + ;; Put start address of root partition in r9 so the kernel can use it + ;; when mounting from flash + move.d input_data, $r0 + move.d [$r0], $r9 ; flash address of compressed kernel + move.d inptr, $r0 + add.d [$r0], $r9 ; size of compressed kernel + cmp.d 0x40200000, $r9 + blo enter_kernel + nop + sub.d 0x40200000, $r9 + add.d 0x4000, $r9 + +enter_kernel: + ;; Enter the decompressed kernel + move.d RAM_INIT_MAGIC, $r8 ; Tell kernel that DRAM is initialized + jump 0x40004000 ; kernel is linked to this address + nop + + .data + +input_data: + .dword 0 ; used by the decompressor +_cmd_line_magic: + .dword 0 +_cmd_line_addr: + .dword 0 +is_nand_boot: + .dword 0 + +#include "../../lib/hw_settings.S" diff --git a/arch/cris/arch-v32/boot/compressed/misc.c b/arch/cris/arch-v32/boot/compressed/misc.c new file mode 100644 index 000000000000..54644238ed59 --- /dev/null +++ b/arch/cris/arch-v32/boot/compressed/misc.c @@ -0,0 +1,318 @@ +/* + * misc.c + * + * $Id: misc.c,v 1.8 2005/04/24 18:34:29 starvik Exp $ + * + * This is a collection of several routines from gzip-1.0.3 + * adapted for Linux. + * + * malloc by Hannu Savolainen 1993 and Matthias Urlichs 1994 + * puts by Nick Holloway 1993, better puts by Martin Mares 1995 + * adoptation for Linux/CRIS Axis Communications AB, 1999 + * + */ + +/* where the piggybacked kernel image expects itself to live. + * it is the same address we use when we network load an uncompressed + * image into DRAM, and it is the address the kernel is linked to live + * at by vmlinux.lds.S + */ + +#define KERNEL_LOAD_ADR 0x40004000 + +#include <linux/config.h> + +#include <linux/types.h> +#include <asm/arch/hwregs/reg_rdwr.h> +#include <asm/arch/hwregs/reg_map.h> +#include <asm/arch/hwregs/ser_defs.h> + +/* + * gzip declarations + */ + +#define OF(args) args +#define STATIC static + +void* memset(void* s, int c, size_t n); +void* memcpy(void* __dest, __const void* __src, + size_t __n); + +#define memzero(s, n) memset ((s), 0, (n)) + + +typedef unsigned char uch; +typedef unsigned short ush; +typedef unsigned long ulg; + +#define WSIZE 0x8000 /* Window size must be at least 32k, */ + /* and a power of two */ + +static uch *inbuf; /* input buffer */ +static uch window[WSIZE]; /* Sliding window buffer */ + +unsigned inptr = 0; /* index of next byte to be processed in inbuf + * After decompression it will contain the + * compressed size, and head.S will read it. + */ + +static unsigned outcnt = 0; /* bytes in output buffer */ + +/* gzip flag byte */ +#define ASCII_FLAG 0x01 /* bit 0 set: file probably ascii text */ +#define CONTINUATION 0x02 /* bit 1 set: continuation of multi-part gzip file */ +#define EXTRA_FIELD 0x04 /* bit 2 set: extra field present */ +#define ORIG_NAME 0x08 /* bit 3 set: original file name present */ +#define COMMENT 0x10 /* bit 4 set: file comment present */ +#define ENCRYPTED 0x20 /* bit 5 set: file is encrypted */ +#define RESERVED 0xC0 /* bit 6,7: reserved */ + +#define get_byte() inbuf[inptr++] + +/* Diagnostic functions */ +#ifdef DEBUG +# define Assert(cond,msg) {if(!(cond)) error(msg);} +# define Trace(x) fprintf x +# define Tracev(x) {if (verbose) fprintf x ;} +# define Tracevv(x) {if (verbose>1) fprintf x ;} +# define Tracec(c,x) {if (verbose && (c)) fprintf x ;} +# define Tracecv(c,x) {if (verbose>1 && (c)) fprintf x ;} +#else +# define Assert(cond,msg) +# define Trace(x) +# define Tracev(x) +# define Tracevv(x) +# define Tracec(c,x) +# define Tracecv(c,x) +#endif + +static int fill_inbuf(void); +static void flush_window(void); +static void error(char *m); +static void gzip_mark(void **); +static void gzip_release(void **); + +extern char *input_data; /* lives in head.S */ + +static long bytes_out = 0; +static uch *output_data; +static unsigned long output_ptr = 0; + +static void *malloc(int size); +static void free(void *where); +static void error(char *m); +static void gzip_mark(void **); +static void gzip_release(void **); + +static void puts(const char *); + +/* the "heap" is put directly after the BSS ends, at end */ + +extern int _end; +static long free_mem_ptr = (long)&_end; + +#include "../../../../../lib/inflate.c" + +static void *malloc(int size) +{ + void *p; + + if (size <0) error("Malloc error"); + + free_mem_ptr = (free_mem_ptr + 3) & ~3; /* Align */ + + p = (void *)free_mem_ptr; + free_mem_ptr += size; + + return p; +} + +static void free(void *where) +{ /* Don't care */ +} + +static void gzip_mark(void **ptr) +{ + *ptr = (void *) free_mem_ptr; +} + +static void gzip_release(void **ptr) +{ + free_mem_ptr = (long) *ptr; +} + +/* decompressor info and error messages to serial console */ + +static inline void +serout(const char *s, reg_scope_instances regi_ser) +{ + reg_ser_rs_stat_din rs; + reg_ser_rw_dout dout = {.data = *s}; + + do { + rs = REG_RD(ser, regi_ser, rs_stat_din); + } + while (!rs.tr_rdy);/* Wait for tranceiver. */ + + REG_WR(ser, regi_ser, rw_dout, dout); +} + +static void +puts(const char *s) +{ +#ifndef CONFIG_ETRAX_DEBUG_PORT_NULL + while (*s) { +#ifdef CONFIG_ETRAX_DEBUG_PORT0 + serout(s, regi_ser0); +#endif +#ifdef CONFIG_ETRAX_DEBUG_PORT1 + serout(s, regi_ser1); +#endif +#ifdef CONFIG_ETRAX_DEBUG_PORT2 + serout(s, regi_ser2); +#endif +#ifdef CONFIG_ETRAX_DEBUG_PORT3 + serout(s, regi_ser3); +#endif + *s++; + } +/* CONFIG_ETRAX_DEBUG_PORT_NULL */ +#endif +} + +void* +memset(void* s, int c, size_t n) +{ + int i; + char *ss = (char*)s; + + for (i=0;i<n;i++) ss[i] = c; +} + +void* +memcpy(void* __dest, __const void* __src, + size_t __n) +{ + int i; + char *d = (char *)__dest, *s = (char *)__src; + + for (i=0;i<__n;i++) d[i] = s[i]; +} + +/* =========================================================================== + * Write the output window window[0..outcnt-1] and update crc and bytes_out. + * (Used for the decompressed data only.) + */ + +static void +flush_window() +{ + ulg c = crc; /* temporary variable */ + unsigned n; + uch *in, *out, ch; + + in = window; + out = &output_data[output_ptr]; + for (n = 0; n < outcnt; n++) { + ch = *out++ = *in++; + c = crc_32_tab[((int)c ^ ch) & 0xff] ^ (c >> 8); + } + crc = c; + bytes_out += (ulg)outcnt; + output_ptr += (ulg)outcnt; + outcnt = 0; +} + +static void +error(char *x) +{ + puts("\n\n"); + puts(x); + puts("\n\n -- System halted\n"); + + while(1); /* Halt */ +} + +void +setup_normal_output_buffer() +{ + output_data = (char *)KERNEL_LOAD_ADR; +} + +static inline void +serial_setup(reg_scope_instances regi_ser) +{ + reg_ser_rw_xoff xoff; + reg_ser_rw_tr_ctrl tr_ctrl; + reg_ser_rw_rec_ctrl rec_ctrl; + reg_ser_rw_tr_baud_div tr_baud; + reg_ser_rw_rec_baud_div rec_baud; + + /* Turn off XOFF. */ + xoff = REG_RD(ser, regi_ser, rw_xoff); + + xoff.chr = 0; + xoff.automatic = regk_ser_no; + + REG_WR(ser, regi_ser, rw_xoff, xoff); + + /* Set baudrate and stopbits. */ + tr_ctrl = REG_RD(ser, regi_ser, rw_tr_ctrl); + rec_ctrl = REG_RD(ser, regi_ser, rw_rec_ctrl); + tr_baud = REG_RD(ser, regi_ser, rw_tr_baud_div); + rec_baud = REG_RD(ser, regi_ser, rw_rec_baud_div); + + tr_ctrl.stop_bits = 1; /* 2 stop bits. */ + + /* + * The baudrate setup is a bit fishy, but in the end the tranceiver is + * set to 4800 and the receiver to 115200. The magic value is + * 29.493 MHz. + */ + tr_ctrl.base_freq = regk_ser_f29_493; + rec_ctrl.base_freq = regk_ser_f29_493; + tr_baud.div = (29493000 / 8) / 4800; + rec_baud.div = (29493000 / 8) / 115200; + + REG_WR(ser, regi_ser, rw_tr_ctrl, tr_ctrl); + REG_WR(ser, regi_ser, rw_tr_baud_div, tr_baud); + REG_WR(ser, regi_ser, rw_rec_ctrl, rec_ctrl); + REG_WR(ser, regi_ser, rw_rec_baud_div, rec_baud); +} + +void +decompress_kernel() +{ + char revision; + + /* input_data is set in head.S */ + inbuf = input_data; + +#ifdef CONFIG_ETRAX_DEBUG_PORT0 + serial_setup(regi_ser0); +#endif +#ifdef CONFIG_ETRAX_DEBUG_PORT1 + serial_setup(regi_ser1); +#endif +#ifdef CONFIG_ETRAX_DEBUG_PORT2 + serial_setup(regi_ser2); +#endif +#ifdef CONFIG_ETRAX_DEBUG_PORT3 + serial_setup(regi_ser3); +#endif + + setup_normal_output_buffer(); + + makecrc(); + + __asm__ volatile ("move $vr,%0" : "=rm" (revision)); + if (revision < 32) + { + puts("You need an ETRAX FS to run Linux 2.6/crisv32.\n"); + while(1); + } + + puts("Uncompressing Linux...\n"); + gunzip(); + puts("Done. Now booting the kernel.\n"); +} diff --git a/arch/cris/arch-v32/boot/rescue/Makefile b/arch/cris/arch-v32/boot/rescue/Makefile new file mode 100644 index 000000000000..f668a8198724 --- /dev/null +++ b/arch/cris/arch-v32/boot/rescue/Makefile @@ -0,0 +1,36 @@ +# +# Makefile for rescue code +# +target = $(target_rescue_dir) +src = $(src_rescue_dir) + +CC = gcc-cris -mlinux -march=v32 $(LINUXINCLUDE) +CFLAGS = -O2 +LD = gcc-cris -mlinux -march=v32 -nostdlib +OBJCOPY = objcopy-cris +OBJCOPYFLAGS = -O binary --remove-section=.bss + +all: $(target)/rescue.bin + +rescue: rescue.bin + # do nothing + +$(target)/rescue.bin: $(target) $(target)/head.o + $(LD) -T $(src)/rescue.ld -o $(target)/rescue.o $(target)/head.o + $(OBJCOPY) $(OBJCOPYFLAGS) $(target)/rescue.o $(target)/rescue.bin + cp -p $(target)/rescue.bin $(objtree) + +$(target): + mkdir -p $(target) + +$(target)/head.o: $(src)/head.S + $(CC) -D__ASSEMBLY__ -c $< -o $*.o + +clean: + rm -f $(target)/*.o $(target)/*.bin + +fastdep: + +modules: + +modules-install: diff --git a/arch/cris/arch-v32/boot/rescue/head.S b/arch/cris/arch-v32/boot/rescue/head.S new file mode 100644 index 000000000000..61ede5f30f99 --- /dev/null +++ b/arch/cris/arch-v32/boot/rescue/head.S @@ -0,0 +1,39 @@ +/* $Id: head.S,v 1.4 2004/11/01 16:10:28 starvik Exp $ + * + * This used to be the rescue code but now that is handled by the + * RedBoot based RFL instead. Nothing to see here, move along. + */ + +#include <linux/config.h> +#include <asm/arch/hwregs/reg_map_asm.h> +#include <asm/arch/hwregs/config_defs_asm.h> + + .text + + ;; Start clocks for used blocks. + move.d REG_ADDR(config, regi_config, rw_clk_ctrl), $r1 + move.d [$r1], $r0 + or.d REG_STATE(config, rw_clk_ctrl, cpu, yes) | \ + REG_STATE(config, rw_clk_ctrl, bif, yes) | \ + REG_STATE(config, rw_clk_ctrl, fix_io, yes), $r0 + move.d $r0, [$r1] + + ;; Copy 68KB NAND flash to Internal RAM (if NAND boot) + move.d 0x38004000, $r10 + move.d 0x8000, $r11 + move.d 0x11000, $r12 + move.d copy_complete, $r13 + and.d 0x000fffff, $r13 + or.d 0x38000000, $r13 + +#include "../../lib/nand_init.S" + + ;; No NAND found + move.d CONFIG_ETRAX_PTABLE_SECTOR, $r10 + jump $r10 ; Jump to decompresser + nop + +copy_complete: + move.d 0x38000000 + CONFIG_ETRAX_PTABLE_SECTOR, $r10 + jump $r10 ; Jump to decompresser + nop diff --git a/arch/cris/arch-v32/boot/rescue/rescue.ld b/arch/cris/arch-v32/boot/rescue/rescue.ld new file mode 100644 index 000000000000..42b11aa122b2 --- /dev/null +++ b/arch/cris/arch-v32/boot/rescue/rescue.ld @@ -0,0 +1,20 @@ +MEMORY + { + flash : ORIGIN = 0x00000000, + LENGTH = 0x00100000 + } + +SECTIONS +{ + .text : + { + stext = . ; + *(.text) + etext = . ; + } > flash + .data : + { + *(.data) + edata = . ; + } > flash +} diff --git a/arch/cris/arch-v32/drivers/Kconfig b/arch/cris/arch-v32/drivers/Kconfig new file mode 100644 index 000000000000..a33097f95362 --- /dev/null +++ b/arch/cris/arch-v32/drivers/Kconfig @@ -0,0 +1,625 @@ +config ETRAX_ETHERNET + bool "Ethernet support" + depends on ETRAX_ARCH_V32 + select NET_ETHERNET + help + This option enables the ETRAX FS built-in 10/100Mbit Ethernet + controller. + +config ETRAX_ETHERNET_HW_CSUM + bool "Hardware accelerated ethernet checksum and scatter/gather" + depends on ETRAX_ETHERNET + depends on ETRAX_STREAMCOPROC + default y + help + Hardware acceleration of checksumming and scatter/gather + +config ETRAX_ETHERNET_IFACE0 + depends on ETRAX_ETHERNET + bool "Enable network interface 0" + +config ETRAX_ETHERNET_IFACE1 + depends on ETRAX_ETHERNET + bool "Enable network interface 1 (uses DMA6 and DMA7)" + +choice + prompt "Network LED behavior" + depends on ETRAX_ETHERNET + default ETRAX_NETWORK_LED_ON_WHEN_ACTIVITY + +config ETRAX_NETWORK_LED_ON_WHEN_LINK + bool "LED_on_when_link" + help + Selecting LED_on_when_link will light the LED when there is a + connection and will flash off when there is activity. + + Selecting LED_on_when_activity will light the LED only when + there is activity. + + This setting will also affect the behaviour of other activity LEDs + e.g. Bluetooth. + +config ETRAX_NETWORK_LED_ON_WHEN_ACTIVITY + bool "LED_on_when_activity" + help + Selecting LED_on_when_link will light the LED when there is a + connection and will flash off when there is activity. + + Selecting LED_on_when_activity will light the LED only when + there is activity. + + This setting will also affect the behaviour of other activity LEDs + e.g. Bluetooth. + +endchoice + +config ETRAXFS_SERIAL + bool "Serial-port support" + depends on ETRAX_ARCH_V32 + help + Enables the ETRAX FS serial driver for ser0 (ttyS0) + You probably want this enabled. + +config ETRAX_SERIAL_PORT0 + bool "Serial port 0 enabled" + depends on ETRAXFS_SERIAL + help + Enables the ETRAX FS serial driver for ser0 (ttyS0) + Normally you want this on. You can control what DMA channels to use + if you do not need DMA to something else. + ser0 can use dma4 or dma6 for output and dma5 or dma7 for input. + +choice + prompt "Ser0 DMA in channel " + depends on ETRAX_SERIAL_PORT0 + default ETRAX_SERIAL_PORT0_NO_DMA_IN + help + What DMA channel to use for ser0. + + +config ETRAX_SERIAL_PORT0_NO_DMA_IN + bool "Ser0 uses no DMA for input" + help + Do not use DMA for ser0 input. + +config ETRAX_SERIAL_PORT0_DMA7_IN + bool "Ser0 uses DMA7 for input" + depends on ETRAX_SERIAL_PORT0 + help + Enables the DMA7 input channel for ser0 (ttyS0). + If you do not enable DMA, an interrupt for each character will be + used when receiveing data. + Normally you want to use DMA, unless you use the DMA channel for + something else. + +endchoice + +choice + prompt "Ser0 DMA out channel" + depends on ETRAX_SERIAL_PORT0 + default ETRAX_SERIAL_PORT0_NO_DMA_OUT + +config ETRAX_SERIAL_PORT0_NO_DMA_OUT + bool "Ser0 uses no DMA for output" + help + Do not use DMA for ser0 output. + +config ETRAX_SERIAL_PORT0_DMA6_OUT + bool "Ser0 uses DMA6 for output" + depends on ETRAX_SERIAL_PORT0 + help + Enables the DMA6 output channel for ser0 (ttyS0). + If you do not enable DMA, an interrupt for each character will be + used when transmitting data. + Normally you want to use DMA, unless you use the DMA channel for + something else. + +endchoice + +config ETRAX_SER0_DTR_BIT + string "Ser 0 DTR bit (empty = not used)" + depends on ETRAX_SERIAL_PORT0 + +config ETRAX_SER0_RI_BIT + string "Ser 0 RI bit (empty = not used)" + depends on ETRAX_SERIAL_PORT0 + +config ETRAX_SER0_DSR_BIT + string "Ser 0 DSR bit (empty = not used)" + depends on ETRAX_SERIAL_PORT0 + +config ETRAX_SER0_CD_BIT + string "Ser 0 CD bit (empty = not used)" + depends on ETRAX_SERIAL_PORT0 + +config ETRAX_SERIAL_PORT1 + bool "Serial port 1 enabled" + depends on ETRAXFS_SERIAL + help + Enables the ETRAX FS serial driver for ser1 (ttyS1). + +choice + prompt "Ser1 DMA in channel " + depends on ETRAX_SERIAL_PORT1 + default ETRAX_SERIAL_PORT1_NO_DMA_IN + help + What DMA channel to use for ser1. + + +config ETRAX_SERIAL_PORT1_NO_DMA_IN + bool "Ser1 uses no DMA for input" + help + Do not use DMA for ser1 input. + +config ETRAX_SERIAL_PORT1_DMA5_IN + bool "Ser1 uses DMA5 for input" + depends on ETRAX_SERIAL_PORT1 + help + Enables the DMA5 input channel for ser1 (ttyS1). + If you do not enable DMA, an interrupt for each character will be + used when receiveing data. + Normally you want this on, unless you use the DMA channel for + something else. + +endchoice + +choice + prompt "Ser1 DMA out channel " + depends on ETRAX_SERIAL_PORT1 + default ETRAX_SERIAL_PORT1_NO_DMA_OUT + help + What DMA channel to use for ser1. + +config ETRAX_SERIAL_PORT1_NO_DMA_OUT + bool "Ser1 uses no DMA for output" + help + Do not use DMA for ser1 output. + +config ETRAX_SERIAL_PORT1_DMA4_OUT + bool "Ser1 uses DMA4 for output" + depends on ETRAX_SERIAL_PORT1 + help + Enables the DMA4 output channel for ser1 (ttyS1). + If you do not enable DMA, an interrupt for each character will be + used when transmitting data. + Normally you want this on, unless you use the DMA channel for + something else. + +endchoice + +config ETRAX_SER1_DTR_BIT + string "Ser 1 DTR bit (empty = not used)" + depends on ETRAX_SERIAL_PORT1 + +config ETRAX_SER1_RI_BIT + string "Ser 1 RI bit (empty = not used)" + depends on ETRAX_SERIAL_PORT1 + +config ETRAX_SER1_DSR_BIT + string "Ser 1 DSR bit (empty = not used)" + depends on ETRAX_SERIAL_PORT1 + +config ETRAX_SER1_CD_BIT + string "Ser 1 CD bit (empty = not used)" + depends on ETRAX_SERIAL_PORT1 + +config ETRAX_SERIAL_PORT2 + bool "Serial port 2 enabled" + depends on ETRAXFS_SERIAL + help + Enables the ETRAX FS serial driver for ser2 (ttyS2). + +choice + prompt "Ser2 DMA in channel " + depends on ETRAX_SERIAL_PORT2 + default ETRAX_SERIAL_PORT2_NO_DMA_IN + help + What DMA channel to use for ser2. + + +config ETRAX_SERIAL_PORT2_NO_DMA_IN + bool "Ser2 uses no DMA for input" + help + Do not use DMA for ser2 input. + +config ETRAX_SERIAL_PORT2_DMA3_IN + bool "Ser2 uses DMA3 for input" + depends on ETRAX_SERIAL_PORT2 + help + Enables the DMA3 input channel for ser2 (ttyS2). + If you do not enable DMA, an interrupt for each character will be + used when receiveing data. + Normally you want to use DMA, unless you use the DMA channel for + something else. + +endchoice + +choice + prompt "Ser2 DMA out channel" + depends on ETRAX_SERIAL_PORT2 + default ETRAX_SERIAL_PORT2_NO_DMA_OUT + +config ETRAX_SERIAL_PORT2_NO_DMA_OUT + bool "Ser2 uses no DMA for output" + help + Do not use DMA for ser2 output. + +config ETRAX_SERIAL_PORT2_DMA2_OUT + bool "Ser2 uses DMA2 for output" + depends on ETRAX_SERIAL_PORT2 + help + Enables the DMA2 output channel for ser2 (ttyS2). + If you do not enable DMA, an interrupt for each character will be + used when transmitting data. + Normally you want to use DMA, unless you use the DMA channel for + something else. + +endchoice + +config ETRAX_SER2_DTR_BIT + string "Ser 2 DTR bit (empty = not used)" + depends on ETRAX_SERIAL_PORT2 + +config ETRAX_SER2_RI_BIT + string "Ser 2 RI bit (empty = not used)" + depends on ETRAX_SERIAL_PORT2 + +config ETRAX_SER2_DSR_BIT + string "Ser 2 DSR bit (empty = not used)" + depends on ETRAX_SERIAL_PORT2 + +config ETRAX_SER2_CD_BIT + string "Ser 2 CD bit (empty = not used)" + depends on ETRAX_SERIAL_PORT2 + +config ETRAX_SERIAL_PORT3 + bool "Serial port 3 enabled" + depends on ETRAXFS_SERIAL + help + Enables the ETRAX FS serial driver for ser3 (ttyS3). + +choice + prompt "Ser3 DMA in channel " + depends on ETRAX_SERIAL_PORT3 + default ETRAX_SERIAL_PORT3_NO_DMA_IN + help + What DMA channel to use for ser3. + + +config ETRAX_SERIAL_PORT3_NO_DMA_IN + bool "Ser3 uses no DMA for input" + help + Do not use DMA for ser3 input. + +config ETRAX_SERIAL_PORT3_DMA9_IN + bool "Ser3 uses DMA9 for input" + depends on ETRAX_SERIAL_PORT3 + help + Enables the DMA9 input channel for ser3 (ttyS3). + If you do not enable DMA, an interrupt for each character will be + used when receiveing data. + Normally you want to use DMA, unless you use the DMA channel for + something else. + +endchoice + +choice + prompt "Ser3 DMA out channel" + depends on ETRAX_SERIAL_PORT3 + default ETRAX_SERIAL_PORT3_NO_DMA_OUT + +config ETRAX_SERIAL_PORT3_NO_DMA_OUT + bool "Ser3 uses no DMA for output" + help + Do not use DMA for ser3 output. + +config ETRAX_SERIAL_PORT3_DMA8_OUT + bool "Ser3 uses DMA8 for output" + depends on ETRAX_SERIAL_PORT3 + help + Enables the DMA8 output channel for ser3 (ttyS3). + If you do not enable DMA, an interrupt for each character will be + used when transmitting data. + Normally you want to use DMA, unless you use the DMA channel for + something else. + +endchoice + +config ETRAX_SER3_DTR_BIT + string "Ser 3 DTR bit (empty = not used)" + depends on ETRAX_SERIAL_PORT3 + +config ETRAX_SER3_RI_BIT + string "Ser 3 RI bit (empty = not used)" + depends on ETRAX_SERIAL_PORT3 + +config ETRAX_SER3_DSR_BIT + string "Ser 3 DSR bit (empty = not used)" + depends on ETRAX_SERIAL_PORT3 + +config ETRAX_SER3_CD_BIT + string "Ser 3 CD bit (empty = not used)" + depends on ETRAX_SERIAL_PORT3 + +config ETRAX_RS485 + bool "RS-485 support" + depends on ETRAX_SERIAL + help + Enables support for RS-485 serial communication. For a primer on + RS-485, see <http://www.hw.cz/english/docs/rs485/rs485.html>. + +config ETRAX_RS485_DISABLE_RECEIVER + bool "Disable serial receiver" + depends on ETRAX_RS485 + help + It is necessary to disable the serial receiver to avoid serial + loopback. Not all products are able to do this in software only. + Axis 2400/2401 must disable receiver. + +config ETRAX_AXISFLASHMAP + bool "Axis flash-map support" + depends on ETRAX_ARCH_V32 + select MTD + select MTD_CFI + select MTD_CFI_AMDSTD + select MTD_OBSOLETE_CHIPS + select MTD_AMDSTD + select MTD_CHAR + select MTD_BLOCK + select MTD_PARTITIONS + select MTD_CONCAT + select MTD_COMPLEX_MAPPINGS + help + This option enables MTD mapping of flash devices. Needed to use + flash memories. If unsure, say Y. + +config ETRAX_SYNCHRONOUS_SERIAL + bool "Synchronous serial-port support" + depends on ETRAX_ARCH_V32 + help + Enables the ETRAX FS synchronous serial driver. + +config ETRAX_SYNCHRONOUS_SERIAL_PORT0 + bool "Synchronous serial port 0 enabled" + depends on ETRAX_SYNCHRONOUS_SERIAL + help + Enabled synchronous serial port 0. + +config ETRAX_SYNCHRONOUS_SERIAL0_DMA + bool "Enable DMA on synchronous serial port 0." + depends on ETRAX_SYNCHRONOUS_SERIAL_PORT0 + help + A synchronous serial port can run in manual or DMA mode. + Selecting this option will make it run in DMA mode. + +config ETRAX_SYNCHRONOUS_SERIAL_PORT1 + bool "Synchronous serial port 1 enabled" + depends on ETRAX_SYNCHRONOUS_SERIAL + help + Enabled synchronous serial port 1. + +config ETRAX_SYNCHRONOUS_SERIAL1_DMA + bool "Enable DMA on synchronous serial port 1." + depends on ETRAX_SYNCHRONOUS_SERIAL_PORT1 + help + A synchronous serial port can run in manual or DMA mode. + Selecting this option will make it run in DMA mode. + +config ETRAX_PTABLE_SECTOR + int "Byte-offset of partition table sector" + depends on ETRAX_AXISFLASHMAP + default "65536" + help + Byte-offset of the partition table in the first flash chip. + The default value is 64kB and should not be changed unless + you know exactly what you are doing. The only valid reason + for changing this is when the flash block size is bigger + than 64kB (e.g. when using two parallel 16 bit flashes). + +config ETRAX_NANDFLASH + bool "NAND flash support" + depends on ETRAX_ARCH_V32 + select MTD_NAND + select MTD_NAND_IDS + help + This option enables MTD mapping of NAND flash devices. Needed to use + NAND flash memories. If unsure, say Y. + +config ETRAX_I2C + bool "I2C driver" + depends on ETRAX_ARCH_V32 + help + This option enabled the I2C driver used by e.g. the RTC driver. + +config ETRAX_I2C_DATA_PORT + string "I2C data pin" + depends on ETRAX_I2C + help + The pin to use for I2C data. + +config ETRAX_I2C_CLK_PORT + string "I2C clock pin" + depends on ETRAX_I2C + help + The pin to use for I2C clock. + +config ETRAX_RTC + bool "Real Time Clock support" + depends on ETRAX_ARCH_V32 + help + Enabled RTC support. + +choice + prompt "RTC chip" + depends on ETRAX_RTC + default ETRAX_PCF8563 + +config ETRAX_PCF8563 + bool "PCF8563" + help + Philips PCF8563 RTC + +endchoice + +config ETRAX_GPIO + bool "GPIO support" + depends on ETRAX_ARCH_V32 + ---help--- + Enables the ETRAX general port device (major 120, minors 0-4). + You can use this driver to access the general port bits. It supports + these ioctl's: + #include <linux/etraxgpio.h> + fd = open("/dev/gpioa", O_RDWR); // or /dev/gpiob + ioctl(fd, _IO(ETRAXGPIO_IOCTYPE, IO_SETBITS), bits_to_set); + ioctl(fd, _IO(ETRAXGPIO_IOCTYPE, IO_CLRBITS), bits_to_clear); + err = ioctl(fd, _IO(ETRAXGPIO_IOCTYPE, IO_READ_INBITS), &val); + Remember that you need to setup the port directions appropriately in + the General configuration. + +config ETRAX_PA_BUTTON_BITMASK + hex "PA-buttons bitmask" + depends on ETRAX_GPIO + default "0x02" + help + This is a bitmask (8 bits) with information about what bits on PA + that are used for buttons. + Most products has a so called TEST button on PA1, if that is true + use 0x02 here. + Use 00 if there are no buttons on PA. + If the bitmask is <> 00 a button driver will be included in the gpio + driver. ETRAX general I/O support must be enabled. + +config ETRAX_PA_CHANGEABLE_DIR + hex "PA user changeable dir mask" + depends on ETRAX_GPIO + default "0x00" + help + This is a bitmask (8 bits) with information of what bits in PA that a + user can change direction on using ioctl's. + Bit set = changeable. + You probably want 0x00 here, but it depends on your hardware. + +config ETRAX_PA_CHANGEABLE_BITS + hex "PA user changeable bits mask" + depends on ETRAX_GPIO + default "0x00" + help + This is a bitmask (8 bits) with information of what bits in PA + that a user can change the value on using ioctl's. + Bit set = changeable. + +config ETRAX_PB_CHANGEABLE_DIR + hex "PB user changeable dir mask" + depends on ETRAX_GPIO + default "0x00000" + help + This is a bitmask (18 bits) with information of what bits in PB + that a user can change direction on using ioctl's. + Bit set = changeable. + You probably want 0x00000 here, but it depends on your hardware. + +config ETRAX_PB_CHANGEABLE_BITS + hex "PB user changeable bits mask" + depends on ETRAX_GPIO + default "0x00000" + help + This is a bitmask (18 bits) with information of what bits in PB + that a user can change the value on using ioctl's. + Bit set = changeable. + +config ETRAX_PC_CHANGEABLE_DIR + hex "PC user changeable dir mask" + depends on ETRAX_GPIO + default "0x00000" + help + This is a bitmask (18 bits) with information of what bits in PC + that a user can change direction on using ioctl's. + Bit set = changeable. + You probably want 0x00000 here, but it depends on your hardware. + +config ETRAX_PC_CHANGEABLE_BITS + hex "PC user changeable bits mask" + depends on ETRAX_GPIO + default "0x00000" + help + This is a bitmask (18 bits) with information of what bits in PC + that a user can change the value on using ioctl's. + Bit set = changeable. + +config ETRAX_PD_CHANGEABLE_DIR + hex "PD user changeable dir mask" + depends on ETRAX_GPIO + default "0x00000" + help + This is a bitmask (18 bits) with information of what bits in PD + that a user can change direction on using ioctl's. + Bit set = changeable. + You probably want 0x00000 here, but it depends on your hardware. + +config ETRAX_PD_CHANGEABLE_BITS + hex "PD user changeable bits mask" + depends on ETRAX_GPIO + default "0x00000" + help + This is a bitmask (18 bits) with information of what bits in PD + that a user can change the value on using ioctl's. + Bit set = changeable. + +config ETRAX_PE_CHANGEABLE_DIR + hex "PE user changeable dir mask" + depends on ETRAX_GPIO + default "0x00000" + help + This is a bitmask (18 bits) with information of what bits in PE + that a user can change direction on using ioctl's. + Bit set = changeable. + You probably want 0x00000 here, but it depends on your hardware. + +config ETRAX_PE_CHANGEABLE_BITS + hex "PE user changeable bits mask" + depends on ETRAX_GPIO + default "0x00000" + help + This is a bitmask (18 bits) with information of what bits in PE + that a user can change the value on using ioctl's. + Bit set = changeable. + +config ETRAX_IDE + bool "ATA/IDE support" + depends on ETRAX_ARCH_V32 + select IDE + select BLK_DEV_IDE + select BLK_DEV_IDEDISK + select BLK_DEV_IDECD + select BLK_DEV_IDEDMA + help + Enables the ETRAX IDE driver. + +config ETRAX_CARDBUS + bool "Cardbus support" + depends on ETRAX_ARCH_V32 + select PCCARD + select CARDBUS + select HOTPLUG + select PCCARD_NONSTATIC + help + Enabled the ETRAX Carbus driver. + +config PCI + bool + depends on ETRAX_CARDBUS + default y + +config ETRAX_IOP_FW_LOAD + tristate "IO-processor hotplug firmware loading support" + depends on ETRAX_ARCH_V32 + select FW_LOADER + help + Enables IO-processor hotplug firmware loading support. + +config ETRAX_STREAMCOPROC + tristate "Stream co-processor driver enabled" + depends on ETRAX_ARCH_V32 + help + This option enables a driver for the stream co-processor + for cryptographic operations. diff --git a/arch/cris/arch-v32/drivers/Makefile b/arch/cris/arch-v32/drivers/Makefile new file mode 100644 index 000000000000..a359cd20ae75 --- /dev/null +++ b/arch/cris/arch-v32/drivers/Makefile @@ -0,0 +1,13 @@ +# +# Makefile for Etrax-specific drivers +# + +obj-$(CONFIG_ETRAX_STREAMCOPROC) += cryptocop.o +obj-$(CONFIG_ETRAX_AXISFLASHMAP) += axisflashmap.o +obj-$(CONFIG_ETRAX_NANDFLASH) += nandflash.o +obj-$(CONFIG_ETRAX_GPIO) += gpio.o +obj-$(CONFIG_ETRAX_IOP_FW_LOAD) += iop_fw_load.o +obj-$(CONFIG_ETRAX_PCF8563) += pcf8563.o +obj-$(CONFIG_ETRAX_I2C) += i2c.o +obj-$(CONFIG_ETRAX_SYNCHRONOUS_SERIAL) += sync_serial.o +obj-$(CONFIG_PCI) += pci/ diff --git a/arch/cris/arch-v32/drivers/axisflashmap.c b/arch/cris/arch-v32/drivers/axisflashmap.c new file mode 100644 index 000000000000..78ed52b1cdac --- /dev/null +++ b/arch/cris/arch-v32/drivers/axisflashmap.c @@ -0,0 +1,455 @@ +/* + * Physical mapping layer for MTD using the Axis partitiontable format + * + * Copyright (c) 2001, 2002, 2003 Axis Communications AB + * + * This file is under the GPL. + * + * First partition is always sector 0 regardless of if we find a partitiontable + * or not. In the start of the next sector, there can be a partitiontable that + * tells us what other partitions to define. If there isn't, we use a default + * partition split defined below. + * + * Copy of os/lx25/arch/cris/arch-v10/drivers/axisflashmap.c 1.5 + * with minor changes. + * + */ + +#include <linux/module.h> +#include <linux/types.h> +#include <linux/kernel.h> +#include <linux/config.h> +#include <linux/init.h> + +#include <linux/mtd/concat.h> +#include <linux/mtd/map.h> +#include <linux/mtd/mtd.h> +#include <linux/mtd/mtdram.h> +#include <linux/mtd/partitions.h> + +#include <asm/arch/hwregs/config_defs.h> +#include <asm/axisflashmap.h> +#include <asm/mmu.h> + +#define MEM_CSE0_SIZE (0x04000000) +#define MEM_CSE1_SIZE (0x04000000) + +#define FLASH_UNCACHED_ADDR KSEG_E +#define FLASH_CACHED_ADDR KSEG_F + +#if CONFIG_ETRAX_FLASH_BUSWIDTH==1 +#define flash_data __u8 +#elif CONFIG_ETRAX_FLASH_BUSWIDTH==2 +#define flash_data __u16 +#elif CONFIG_ETRAX_FLASH_BUSWIDTH==4 +#define flash_data __u16 +#endif + +/* From head.S */ +extern unsigned long romfs_start, romfs_length, romfs_in_flash; + +/* The master mtd for the entire flash. */ +struct mtd_info* axisflash_mtd = NULL; + +/* Map driver functions. */ + +static map_word flash_read(struct map_info *map, unsigned long ofs) +{ + map_word tmp; + tmp.x[0] = *(flash_data *)(map->map_priv_1 + ofs); + return tmp; +} + +static void flash_copy_from(struct map_info *map, void *to, + unsigned long from, ssize_t len) +{ + memcpy(to, (void *)(map->map_priv_1 + from), len); +} + +static void flash_write(struct map_info *map, map_word d, unsigned long adr) +{ + *(flash_data *)(map->map_priv_1 + adr) = (flash_data)d.x[0]; +} + +/* + * The map for chip select e0. + * + * We run into tricky coherence situations if we mix cached with uncached + * accesses to we only use the uncached version here. + * + * The size field is the total size where the flash chips may be mapped on the + * chip select. MTD probes should find all devices there and it does not matter + * if there are unmapped gaps or aliases (mirrors of flash devices). The MTD + * probes will ignore them. + * + * The start address in map_priv_1 is in virtual memory so we cannot use + * MEM_CSE0_START but must rely on that FLASH_UNCACHED_ADDR is the start + * address of cse0. + */ +static struct map_info map_cse0 = { + .name = "cse0", + .size = MEM_CSE0_SIZE, + .bankwidth = CONFIG_ETRAX_FLASH_BUSWIDTH, + .read = flash_read, + .copy_from = flash_copy_from, + .write = flash_write, + .map_priv_1 = FLASH_UNCACHED_ADDR +}; + +/* + * The map for chip select e1. + * + * If there was a gap between cse0 and cse1, map_priv_1 would get the wrong + * address, but there isn't. + */ +static struct map_info map_cse1 = { + .name = "cse1", + .size = MEM_CSE1_SIZE, + .bankwidth = CONFIG_ETRAX_FLASH_BUSWIDTH, + .read = flash_read, + .copy_from = flash_copy_from, + .write = flash_write, + .map_priv_1 = FLASH_UNCACHED_ADDR + MEM_CSE0_SIZE +}; + +/* If no partition-table was found, we use this default-set. */ +#define MAX_PARTITIONS 7 +#define NUM_DEFAULT_PARTITIONS 3 + +/* + * Default flash size is 2MB. CONFIG_ETRAX_PTABLE_SECTOR is most likely the + * size of one flash block and "filesystem"-partition needs 5 blocks to be able + * to use JFFS. + */ +static struct mtd_partition axis_default_partitions[NUM_DEFAULT_PARTITIONS] = { + { + .name = "boot firmware", + .size = CONFIG_ETRAX_PTABLE_SECTOR, + .offset = 0 + }, + { + .name = "kernel", + .size = 0x200000 - (6 * CONFIG_ETRAX_PTABLE_SECTOR), + .offset = CONFIG_ETRAX_PTABLE_SECTOR + }, + { + .name = "filesystem", + .size = 5 * CONFIG_ETRAX_PTABLE_SECTOR, + .offset = 0x200000 - (5 * CONFIG_ETRAX_PTABLE_SECTOR) + } +}; + +/* Initialize the ones normally used. */ +static struct mtd_partition axis_partitions[MAX_PARTITIONS] = { + { + .name = "part0", + .size = CONFIG_ETRAX_PTABLE_SECTOR, + .offset = 0 + }, + { + .name = "part1", + .size = 0, + .offset = 0 + }, + { + .name = "part2", + .size = 0, + .offset = 0 + }, + { + .name = "part3", + .size = 0, + .offset = 0 + }, + { + .name = "part4", + .size = 0, + .offset = 0 + }, + { + .name = "part5", + .size = 0, + .offset = 0 + }, + { + .name = "part6", + .size = 0, + .offset = 0 + }, +}; + +/* + * Probe a chip select for AMD-compatible (JEDEC) or CFI-compatible flash + * chips in that order (because the amd_flash-driver is faster). + */ +static struct mtd_info *probe_cs(struct map_info *map_cs) +{ + struct mtd_info *mtd_cs = NULL; + + printk(KERN_INFO + "%s: Probing a 0x%08lx bytes large window at 0x%08lx.\n", + map_cs->name, map_cs->size, map_cs->map_priv_1); + +#ifdef CONFIG_MTD_AMDSTD + mtd_cs = do_map_probe("amd_flash", map_cs); +#endif +#ifdef CONFIG_MTD_CFI + if (!mtd_cs) { + mtd_cs = do_map_probe("cfi_probe", map_cs); + } +#endif + + return mtd_cs; +} + +/* + * Probe each chip select individually for flash chips. If there are chips on + * both cse0 and cse1, the mtd_info structs will be concatenated to one struct + * so that MTD partitions can cross chip boundries. + * + * The only known restriction to how you can mount your chips is that each + * chip select must hold similar flash chips. But you need external hardware + * to do that anyway and you can put totally different chips on cse0 and cse1 + * so it isn't really much of a restriction. + */ +extern struct mtd_info* __init crisv32_nand_flash_probe (void); +static struct mtd_info *flash_probe(void) +{ + struct mtd_info *mtd_cse0; + struct mtd_info *mtd_cse1; + struct mtd_info *mtd_nand = NULL; + struct mtd_info *mtd_total; + struct mtd_info *mtds[3]; + int count = 0; + + if ((mtd_cse0 = probe_cs(&map_cse0)) != NULL) + mtds[count++] = mtd_cse0; + if ((mtd_cse1 = probe_cs(&map_cse1)) != NULL) + mtds[count++] = mtd_cse1; + +#ifdef CONFIG_ETRAX_NANDFLASH + if ((mtd_nand = crisv32_nand_flash_probe()) != NULL) + mtds[count++] = mtd_nand; +#endif + + if (!mtd_cse0 && !mtd_cse1 && !mtd_nand) { + /* No chip found. */ + return NULL; + } + + if (count > 1) { +#ifdef CONFIG_MTD_CONCAT + /* Since the concatenation layer adds a small overhead we + * could try to figure out if the chips in cse0 and cse1 are + * identical and reprobe the whole cse0+cse1 window. But since + * flash chips are slow, the overhead is relatively small. + * So we use the MTD concatenation layer instead of further + * complicating the probing procedure. + */ + mtd_total = mtd_concat_create(mtds, + count, + "cse0+cse1+nand"); +#else + printk(KERN_ERR "%s and %s: Cannot concatenate due to kernel " + "(mis)configuration!\n", map_cse0.name, map_cse1.name); + mtd_toal = NULL; +#endif + if (!mtd_total) { + printk(KERN_ERR "%s and %s: Concatenation failed!\n", + map_cse0.name, map_cse1.name); + + /* The best we can do now is to only use what we found + * at cse0. + */ + mtd_total = mtd_cse0; + map_destroy(mtd_cse1); + } + } else { + mtd_total = mtd_cse0? mtd_cse0 : mtd_cse1 ? mtd_cse1 : mtd_nand; + + } + + return mtd_total; +} + +extern unsigned long crisv32_nand_boot; +extern unsigned long crisv32_nand_cramfs_offset; + +/* + * Probe the flash chip(s) and, if it succeeds, read the partition-table + * and register the partitions with MTD. + */ +static int __init init_axis_flash(void) +{ + struct mtd_info *mymtd; + int err = 0; + int pidx = 0; + struct partitiontable_head *ptable_head = NULL; + struct partitiontable_entry *ptable; + int use_default_ptable = 1; /* Until proven otherwise. */ + const char *pmsg = KERN_INFO " /dev/flash%d at 0x%08x, size 0x%08x\n"; + static char page[512]; + size_t len; + +#ifndef CONFIG_ETRAXFS_SIM + mymtd = flash_probe(); + mymtd->read(mymtd, CONFIG_ETRAX_PTABLE_SECTOR, 512, &len, page); + ptable_head = (struct partitiontable_head *)(page + PARTITION_TABLE_OFFSET); + + if (!mymtd) { + /* There's no reason to use this module if no flash chip can + * be identified. Make sure that's understood. + */ + printk(KERN_INFO "axisflashmap: Found no flash chip.\n"); + } else { + printk(KERN_INFO "%s: 0x%08x bytes of flash memory.\n", + mymtd->name, mymtd->size); + axisflash_mtd = mymtd; + } + + if (mymtd) { + mymtd->owner = THIS_MODULE; + } + pidx++; /* First partition is always set to the default. */ + + if (ptable_head && (ptable_head->magic == PARTITION_TABLE_MAGIC) + && (ptable_head->size < + (MAX_PARTITIONS * sizeof(struct partitiontable_entry) + + PARTITIONTABLE_END_MARKER_SIZE)) + && (*(unsigned long*)((void*)ptable_head + sizeof(*ptable_head) + + ptable_head->size - + PARTITIONTABLE_END_MARKER_SIZE) + == PARTITIONTABLE_END_MARKER)) { + /* Looks like a start, sane length and end of a + * partition table, lets check csum etc. + */ + int ptable_ok = 0; + struct partitiontable_entry *max_addr = + (struct partitiontable_entry *) + ((unsigned long)ptable_head + sizeof(*ptable_head) + + ptable_head->size); + unsigned long offset = CONFIG_ETRAX_PTABLE_SECTOR; + unsigned char *p; + unsigned long csum = 0; + + ptable = (struct partitiontable_entry *) + ((unsigned long)ptable_head + sizeof(*ptable_head)); + + /* Lets be PARANOID, and check the checksum. */ + p = (unsigned char*) ptable; + + while (p <= (unsigned char*)max_addr) { + csum += *p++; + csum += *p++; + csum += *p++; + csum += *p++; + } + ptable_ok = (csum == ptable_head->checksum); + + /* Read the entries and use/show the info. */ + printk(KERN_INFO " Found a%s partition table at 0x%p-0x%p.\n", + (ptable_ok ? " valid" : "n invalid"), ptable_head, + max_addr); + + /* We have found a working bootblock. Now read the + * partition table. Scan the table. It ends when + * there is 0xffffffff, that is, empty flash. + */ + while (ptable_ok + && ptable->offset != 0xffffffff + && ptable < max_addr + && pidx < MAX_PARTITIONS) { + + axis_partitions[pidx].offset = offset + ptable->offset + (crisv32_nand_boot ? 16384 : 0); + axis_partitions[pidx].size = ptable->size; + + printk(pmsg, pidx, axis_partitions[pidx].offset, + axis_partitions[pidx].size); + pidx++; + ptable++; + } + use_default_ptable = !ptable_ok; + } + + if (romfs_in_flash) { + /* Add an overlapping device for the root partition (romfs). */ + + axis_partitions[pidx].name = "romfs"; + if (crisv32_nand_boot) { + char* data = kmalloc(1024, GFP_KERNEL); + int len; + int offset = crisv32_nand_cramfs_offset & ~(1024-1); + char* tmp; + + mymtd->read(mymtd, offset, 1024, &len, data); + tmp = &data[crisv32_nand_cramfs_offset % 512]; + axis_partitions[pidx].size = *(unsigned*)(tmp + 4); + axis_partitions[pidx].offset = crisv32_nand_cramfs_offset; + kfree(data); + } else { + axis_partitions[pidx].size = romfs_length; + axis_partitions[pidx].offset = romfs_start - FLASH_CACHED_ADDR; + } + + axis_partitions[pidx].mask_flags |= MTD_WRITEABLE; + + printk(KERN_INFO + " Adding readonly flash partition for romfs image:\n"); + printk(pmsg, pidx, axis_partitions[pidx].offset, + axis_partitions[pidx].size); + pidx++; + } + + if (mymtd) { + if (use_default_ptable) { + printk(KERN_INFO " Using default partition table.\n"); + err = add_mtd_partitions(mymtd, axis_default_partitions, + NUM_DEFAULT_PARTITIONS); + } else { + err = add_mtd_partitions(mymtd, axis_partitions, pidx); + } + + if (err) { + panic("axisflashmap could not add MTD partitions!\n"); + } + } +/* CONFIG_EXTRAXFS_SIM */ +#endif + + if (!romfs_in_flash) { + /* Create an RAM device for the root partition (romfs). */ + +#if !defined(CONFIG_MTD_MTDRAM) || (CONFIG_MTDRAM_TOTAL_SIZE != 0) || (CONFIG_MTDRAM_ABS_POS != 0) + /* No use trying to boot this kernel from RAM. Panic! */ + printk(KERN_EMERG "axisflashmap: Cannot create an MTD RAM " + "device due to kernel (mis)configuration!\n"); + panic("This kernel cannot boot from RAM!\n"); +#else + struct mtd_info *mtd_ram; + + mtd_ram = (struct mtd_info *)kmalloc(sizeof(struct mtd_info), + GFP_KERNEL); + if (!mtd_ram) { + panic("axisflashmap couldn't allocate memory for " + "mtd_info!\n"); + } + + printk(KERN_INFO " Adding RAM partition for romfs image:\n"); + printk(pmsg, pidx, romfs_start, romfs_length); + + err = mtdram_init_device(mtd_ram, (void*)romfs_start, + romfs_length, "romfs"); + if (err) { + panic("axisflashmap could not initialize MTD RAM " + "device!\n"); + } +#endif + } + + return err; +} + +/* This adds the above to the kernels init-call chain. */ +module_init(init_axis_flash); + +EXPORT_SYMBOL(axisflash_mtd); diff --git a/arch/cris/arch-v32/drivers/cryptocop.c b/arch/cris/arch-v32/drivers/cryptocop.c new file mode 100644 index 000000000000..ca72076c630a --- /dev/null +++ b/arch/cris/arch-v32/drivers/cryptocop.c @@ -0,0 +1,3522 @@ +/* $Id: cryptocop.c,v 1.13 2005/04/21 17:27:55 henriken Exp $ + * + * Stream co-processor driver for the ETRAX FS + * + * Copyright (C) 2003-2005 Axis Communications AB + */ + +#include <linux/init.h> +#include <linux/sched.h> +#include <linux/module.h> +#include <linux/slab.h> +#include <linux/string.h> +#include <linux/fs.h> +#include <linux/mm.h> +#include <linux/spinlock.h> +#include <linux/stddef.h> + +#include <asm/uaccess.h> +#include <asm/io.h> +#include <asm/atomic.h> + +#include <linux/list.h> +#include <linux/interrupt.h> + +#include <asm/signal.h> +#include <asm/irq.h> + +#include <asm/arch/dma.h> +#include <asm/arch/hwregs/dma.h> +#include <asm/arch/hwregs/reg_map.h> +#include <asm/arch/hwregs/reg_rdwr.h> +#include <asm/arch/hwregs/intr_vect_defs.h> + +#include <asm/arch/hwregs/strcop.h> +#include <asm/arch/hwregs/strcop_defs.h> +#include <asm/arch/cryptocop.h> + + + +#define DESCR_ALLOC_PAD (31) + +struct cryptocop_dma_desc { + char *free_buf; /* If non-null will be kfreed in free_cdesc() */ + dma_descr_data *dma_descr; + + unsigned char dma_descr_buf[sizeof(dma_descr_data) + DESCR_ALLOC_PAD]; + + unsigned int from_pool:1; /* If 1 'allocated' from the descriptor pool. */ + struct cryptocop_dma_desc *next; +}; + + +struct cryptocop_int_operation{ + void *alloc_ptr; + cryptocop_session_id sid; + + dma_descr_context ctx_out; + dma_descr_context ctx_in; + + /* DMA descriptors allocated by driver. */ + struct cryptocop_dma_desc *cdesc_out; + struct cryptocop_dma_desc *cdesc_in; + + /* Strcop config to use. */ + cryptocop_3des_mode tdes_mode; + cryptocop_csum_type csum_mode; + + /* DMA descrs provided by consumer. */ + dma_descr_data *ddesc_out; + dma_descr_data *ddesc_in; +}; + + +struct cryptocop_tfrm_ctx { + cryptocop_tfrm_id tid; + unsigned int blocklength; + + unsigned int start_ix; + + struct cryptocop_tfrm_cfg *tcfg; + struct cryptocop_transform_ctx *tctx; + + unsigned char previous_src; + unsigned char current_src; + + /* Values to use in metadata out. */ + unsigned char hash_conf; + unsigned char hash_mode; + unsigned char ciph_conf; + unsigned char cbcmode; + unsigned char decrypt; + + unsigned int requires_padding:1; + unsigned int strict_block_length:1; + unsigned int active:1; + unsigned int done:1; + size_t consumed; + size_t produced; + + /* Pad (input) descriptors to put in the DMA out list when the transform + * output is put on the DMA in list. */ + struct cryptocop_dma_desc *pad_descs; + + struct cryptocop_tfrm_ctx *prev_src; + struct cryptocop_tfrm_ctx *curr_src; + + /* Mapping to HW. */ + unsigned char unit_no; +}; + + +struct cryptocop_private{ + cryptocop_session_id sid; + struct cryptocop_private *next; +}; + +/* Session list. */ + +struct cryptocop_transform_ctx{ + struct cryptocop_transform_init init; + unsigned char dec_key[CRYPTOCOP_MAX_KEY_LENGTH]; + unsigned int dec_key_set:1; + + struct cryptocop_transform_ctx *next; +}; + + +struct cryptocop_session{ + cryptocop_session_id sid; + + struct cryptocop_transform_ctx *tfrm_ctx; + + struct cryptocop_session *next; +}; + +/* Priority levels for jobs sent to the cryptocop. Checksum operations from + kernel have highest priority since TCPIP stack processing must not + be a bottleneck. */ +typedef enum { + cryptocop_prio_kernel_csum = 0, + cryptocop_prio_kernel = 1, + cryptocop_prio_user = 2, + cryptocop_prio_no_prios = 3 +} cryptocop_queue_priority; + +struct cryptocop_prio_queue{ + struct list_head jobs; + cryptocop_queue_priority prio; +}; + +struct cryptocop_prio_job{ + struct list_head node; + cryptocop_queue_priority prio; + + struct cryptocop_operation *oper; + struct cryptocop_int_operation *iop; +}; + +struct ioctl_job_cb_ctx { + unsigned int processed:1; +}; + + +static struct cryptocop_session *cryptocop_sessions = NULL; +spinlock_t cryptocop_sessions_lock; + +/* Next Session ID to assign. */ +static cryptocop_session_id next_sid = 1; + +/* Pad for checksum. */ +static const char csum_zero_pad[1] = {0x00}; + +/* Trash buffer for mem2mem operations. */ +#define MEM2MEM_DISCARD_BUF_LENGTH (512) +static unsigned char mem2mem_discard_buf[MEM2MEM_DISCARD_BUF_LENGTH]; + +/* Descriptor pool. */ +/* FIXME Tweak this value. */ +#define CRYPTOCOP_DESCRIPTOR_POOL_SIZE (100) +static struct cryptocop_dma_desc descr_pool[CRYPTOCOP_DESCRIPTOR_POOL_SIZE]; +static struct cryptocop_dma_desc *descr_pool_free_list; +static int descr_pool_no_free; +static spinlock_t descr_pool_lock; + +/* Lock to stop cryptocop to start processing of a new operation. The holder + of this lock MUST call cryptocop_start_job() after it is unlocked. */ +spinlock_t cryptocop_process_lock; + +static struct cryptocop_prio_queue cryptocop_job_queues[cryptocop_prio_no_prios]; +static spinlock_t cryptocop_job_queue_lock; +static struct cryptocop_prio_job *cryptocop_running_job = NULL; +static spinlock_t running_job_lock; + +/* The interrupt handler appends completed jobs to this list. The scehduled + * tasklet removes them upon sending the response to the crypto consumer. */ +static struct list_head cryptocop_completed_jobs; +static spinlock_t cryptocop_completed_jobs_lock; + +DECLARE_WAIT_QUEUE_HEAD(cryptocop_ioc_process_wq); + + +/** Local functions. **/ + +static int cryptocop_open(struct inode *, struct file *); + +static int cryptocop_release(struct inode *, struct file *); + +static int cryptocop_ioctl(struct inode *inode, struct file *file, + unsigned int cmd, unsigned long arg); + +static void cryptocop_start_job(void); + +static int cryptocop_job_queue_insert(cryptocop_queue_priority prio, struct cryptocop_operation *operation); +static int cryptocop_job_setup(struct cryptocop_prio_job **pj, struct cryptocop_operation *operation); + +static int cryptocop_job_queue_init(void); +static void cryptocop_job_queue_close(void); + +static int create_md5_pad(int alloc_flag, unsigned long long hashed_length, char **pad, size_t *pad_length); + +static int create_sha1_pad(int alloc_flag, unsigned long long hashed_length, char **pad, size_t *pad_length); + +static int transform_ok(struct cryptocop_transform_init *tinit); + +static struct cryptocop_session *get_session(cryptocop_session_id sid); + +static struct cryptocop_transform_ctx *get_transform_ctx(struct cryptocop_session *sess, cryptocop_tfrm_id tid); + +static void delete_internal_operation(struct cryptocop_int_operation *iop); + +static void get_aes_decrypt_key(unsigned char *dec_key, const unsigned char *key, unsigned int keylength); + +static int init_stream_coprocessor(void); + +static void __exit exit_stream_coprocessor(void); + +/*#define LDEBUG*/ +#ifdef LDEBUG +#define DEBUG(s) s +#define DEBUG_API(s) s +static void print_cryptocop_operation(struct cryptocop_operation *cop); +static void print_dma_descriptors(struct cryptocop_int_operation *iop); +static void print_strcop_crypto_op(struct strcop_crypto_op *cop); +static void print_lock_status(void); +static void print_user_dma_lists(struct cryptocop_dma_list_operation *dma_op); +#define assert(s) do{if (!(s)) panic(#s);} while(0); +#else +#define DEBUG(s) +#define DEBUG_API(s) +#define assert(s) +#endif + + +/* Transform constants. */ +#define DES_BLOCK_LENGTH (8) +#define AES_BLOCK_LENGTH (16) +#define MD5_BLOCK_LENGTH (64) +#define SHA1_BLOCK_LENGTH (64) +#define CSUM_BLOCK_LENGTH (2) +#define MD5_STATE_LENGTH (16) +#define SHA1_STATE_LENGTH (20) + +/* The device number. */ +#define CRYPTOCOP_MAJOR (254) +#define CRYPTOCOP_MINOR (0) + + + +struct file_operations cryptocop_fops = { + owner: THIS_MODULE, + open: cryptocop_open, + release: cryptocop_release, + ioctl: cryptocop_ioctl +}; + + +static void free_cdesc(struct cryptocop_dma_desc *cdesc) +{ + DEBUG(printk("free_cdesc: cdesc 0x%p, from_pool=%d\n", cdesc, cdesc->from_pool)); + if (cdesc->free_buf) kfree(cdesc->free_buf); + + if (cdesc->from_pool) { + unsigned long int flags; + spin_lock_irqsave(&descr_pool_lock, flags); + cdesc->next = descr_pool_free_list; + descr_pool_free_list = cdesc; + ++descr_pool_no_free; + spin_unlock_irqrestore(&descr_pool_lock, flags); + } else { + kfree(cdesc); + } +} + + +static struct cryptocop_dma_desc *alloc_cdesc(int alloc_flag) +{ + int use_pool = (alloc_flag & GFP_ATOMIC) ? 1 : 0; + struct cryptocop_dma_desc *cdesc; + + if (use_pool) { + unsigned long int flags; + spin_lock_irqsave(&descr_pool_lock, flags); + if (!descr_pool_free_list) { + spin_unlock_irqrestore(&descr_pool_lock, flags); + DEBUG_API(printk("alloc_cdesc: pool is empty\n")); + return NULL; + } + cdesc = descr_pool_free_list; + descr_pool_free_list = descr_pool_free_list->next; + --descr_pool_no_free; + spin_unlock_irqrestore(&descr_pool_lock, flags); + cdesc->from_pool = 1; + } else { + cdesc = kmalloc(sizeof(struct cryptocop_dma_desc), alloc_flag); + if (!cdesc) { + DEBUG_API(printk("alloc_cdesc: kmalloc\n")); + return NULL; + } + cdesc->from_pool = 0; + } + cdesc->dma_descr = (dma_descr_data*)(((unsigned long int)cdesc + offsetof(struct cryptocop_dma_desc, dma_descr_buf) + DESCR_ALLOC_PAD) & ~0x0000001F); + + cdesc->next = NULL; + + cdesc->free_buf = NULL; + cdesc->dma_descr->out_eop = 0; + cdesc->dma_descr->in_eop = 0; + cdesc->dma_descr->intr = 0; + cdesc->dma_descr->eol = 0; + cdesc->dma_descr->wait = 0; + cdesc->dma_descr->buf = NULL; + cdesc->dma_descr->after = NULL; + + DEBUG_API(printk("alloc_cdesc: return 0x%p, cdesc->dma_descr=0x%p, from_pool=%d\n", cdesc, cdesc->dma_descr, cdesc->from_pool)); + return cdesc; +} + + +static void setup_descr_chain(struct cryptocop_dma_desc *cd) +{ + DEBUG(printk("setup_descr_chain: entering\n")); + while (cd) { + if (cd->next) { + cd->dma_descr->next = (dma_descr_data*)virt_to_phys(cd->next->dma_descr); + } else { + cd->dma_descr->next = NULL; + } + cd = cd->next; + } + DEBUG(printk("setup_descr_chain: exit\n")); +} + + +/* Create a pad descriptor for the transform. + * Return -1 for error, 0 if pad created. */ +static int create_pad_descriptor(struct cryptocop_tfrm_ctx *tc, struct cryptocop_dma_desc **pad_desc, int alloc_flag) +{ + struct cryptocop_dma_desc *cdesc = NULL; + int error = 0; + struct strcop_meta_out mo = { + .ciphsel = src_none, + .hashsel = src_none, + .csumsel = src_none + }; + char *pad; + size_t plen; + + DEBUG(printk("create_pad_descriptor: start.\n")); + /* Setup pad descriptor. */ + + DEBUG(printk("create_pad_descriptor: setting up padding.\n")); + cdesc = alloc_cdesc(alloc_flag); + if (!cdesc){ + DEBUG_API(printk("create_pad_descriptor: alloc pad desc\n")); + goto error_cleanup; + } + switch (tc->unit_no) { + case src_md5: + error = create_md5_pad(alloc_flag, tc->consumed, &pad, &plen); + if (error){ + DEBUG_API(printk("create_pad_descriptor: create_md5_pad_failed\n")); + goto error_cleanup; + } + cdesc->free_buf = pad; + mo.hashsel = src_dma; + mo.hashconf = tc->hash_conf; + mo.hashmode = tc->hash_mode; + break; + case src_sha1: + error = create_sha1_pad(alloc_flag, tc->consumed, &pad, &plen); + if (error){ + DEBUG_API(printk("create_pad_descriptor: create_sha1_pad_failed\n")); + goto error_cleanup; + } + cdesc->free_buf = pad; + mo.hashsel = src_dma; + mo.hashconf = tc->hash_conf; + mo.hashmode = tc->hash_mode; + break; + case src_csum: + if (tc->consumed % tc->blocklength){ + pad = (char*)csum_zero_pad; + plen = 1; + } else { + pad = (char*)cdesc; /* Use any pointer. */ + plen = 0; + } + mo.csumsel = src_dma; + break; + } + cdesc->dma_descr->wait = 1; + cdesc->dma_descr->out_eop = 1; /* Since this is a pad output is pushed. EOP is ok here since the padded unit is the only one active. */ + cdesc->dma_descr->buf = (char*)virt_to_phys((char*)pad); + cdesc->dma_descr->after = cdesc->dma_descr->buf + plen; + + cdesc->dma_descr->md = REG_TYPE_CONV(unsigned short int, struct strcop_meta_out, mo); + *pad_desc = cdesc; + + return 0; + + error_cleanup: + if (cdesc) free_cdesc(cdesc); + return -1; +} + + +static int setup_key_dl_desc(struct cryptocop_tfrm_ctx *tc, struct cryptocop_dma_desc **kd, int alloc_flag) +{ + struct cryptocop_dma_desc *key_desc = alloc_cdesc(alloc_flag); + struct strcop_meta_out mo = {0}; + + DEBUG(printk("setup_key_dl_desc\n")); + + if (!key_desc) { + DEBUG_API(printk("setup_key_dl_desc: failed descriptor allocation.\n")); + return -ENOMEM; + } + + /* Download key. */ + if ((tc->tctx->init.alg == cryptocop_alg_aes) && (tc->tcfg->flags & CRYPTOCOP_DECRYPT)) { + /* Precook the AES decrypt key. */ + if (!tc->tctx->dec_key_set){ + get_aes_decrypt_key(tc->tctx->dec_key, tc->tctx->init.key, tc->tctx->init.keylen); + tc->tctx->dec_key_set = 1; + } + key_desc->dma_descr->buf = (char*)virt_to_phys(tc->tctx->dec_key); + key_desc->dma_descr->after = key_desc->dma_descr->buf + tc->tctx->init.keylen/8; + } else { + key_desc->dma_descr->buf = (char*)virt_to_phys(tc->tctx->init.key); + key_desc->dma_descr->after = key_desc->dma_descr->buf + tc->tctx->init.keylen/8; + } + /* Setup metadata. */ + mo.dlkey = 1; + switch (tc->tctx->init.keylen) { + case 64: + mo.decrypt = 0; + mo.hashmode = 0; + break; + case 128: + mo.decrypt = 0; + mo.hashmode = 1; + break; + case 192: + mo.decrypt = 1; + mo.hashmode = 0; + break; + case 256: + mo.decrypt = 1; + mo.hashmode = 1; + break; + default: + break; + } + mo.ciphsel = mo.hashsel = mo.csumsel = src_none; + key_desc->dma_descr->md = REG_TYPE_CONV(unsigned short int, struct strcop_meta_out, mo); + + key_desc->dma_descr->out_eop = 1; + key_desc->dma_descr->wait = 1; + key_desc->dma_descr->intr = 0; + + *kd = key_desc; + return 0; +} + +static int setup_cipher_iv_desc(struct cryptocop_tfrm_ctx *tc, struct cryptocop_dma_desc **id, int alloc_flag) +{ + struct cryptocop_dma_desc *iv_desc = alloc_cdesc(alloc_flag); + struct strcop_meta_out mo = {0}; + + DEBUG(printk("setup_cipher_iv_desc\n")); + + if (!iv_desc) { + DEBUG_API(printk("setup_cipher_iv_desc: failed CBC IV descriptor allocation.\n")); + return -ENOMEM; + } + /* Download IV. */ + iv_desc->dma_descr->buf = (char*)virt_to_phys(tc->tcfg->iv); + iv_desc->dma_descr->after = iv_desc->dma_descr->buf + tc->blocklength; + + /* Setup metadata. */ + mo.hashsel = mo.csumsel = src_none; + mo.ciphsel = src_dma; + mo.ciphconf = tc->ciph_conf; + mo.cbcmode = tc->cbcmode; + + iv_desc->dma_descr->md = REG_TYPE_CONV(unsigned short int, struct strcop_meta_out, mo); + + iv_desc->dma_descr->out_eop = 0; + iv_desc->dma_descr->wait = 1; + iv_desc->dma_descr->intr = 0; + + *id = iv_desc; + return 0; +} + +/* Map the ouput length of the transform to operation output starting on the inject index. */ +static int create_input_descriptors(struct cryptocop_operation *operation, struct cryptocop_tfrm_ctx *tc, struct cryptocop_dma_desc **id, int alloc_flag) +{ + int err = 0; + struct cryptocop_dma_desc head = {0}; + struct cryptocop_dma_desc *outdesc = &head; + size_t iov_offset = 0; + size_t out_ix = 0; + int outiov_ix = 0; + struct strcop_meta_in mi = {0}; + + size_t out_length = tc->produced; + int rem_length; + int dlength; + + assert(out_length != 0); + if (((tc->produced + tc->tcfg->inject_ix) > operation->tfrm_op.outlen) || (tc->produced && (operation->tfrm_op.outlen == 0))) { + DEBUG_API(printk("create_input_descriptors: operation outdata too small\n")); + return -EINVAL; + } + /* Traverse the out iovec until the result inject index is reached. */ + while ((outiov_ix < operation->tfrm_op.outcount) && ((out_ix + operation->tfrm_op.outdata[outiov_ix].iov_len) <= tc->tcfg->inject_ix)){ + out_ix += operation->tfrm_op.outdata[outiov_ix].iov_len; + outiov_ix++; + } + if (outiov_ix >= operation->tfrm_op.outcount){ + DEBUG_API(printk("create_input_descriptors: operation outdata too small\n")); + return -EINVAL; + } + iov_offset = tc->tcfg->inject_ix - out_ix; + mi.dmasel = tc->unit_no; + + /* Setup the output descriptors. */ + while ((out_length > 0) && (outiov_ix < operation->tfrm_op.outcount)) { + outdesc->next = alloc_cdesc(alloc_flag); + if (!outdesc->next) { + DEBUG_API(printk("create_input_descriptors: alloc_cdesc\n")); + err = -ENOMEM; + goto error_cleanup; + } + outdesc = outdesc->next; + rem_length = operation->tfrm_op.outdata[outiov_ix].iov_len - iov_offset; + dlength = (out_length < rem_length) ? out_length : rem_length; + + DEBUG(printk("create_input_descriptors:\n" + "outiov_ix=%d, rem_length=%d, dlength=%d\n" + "iov_offset=%d, outdata[outiov_ix].iov_len=%d\n" + "outcount=%d, outiov_ix=%d\n", + outiov_ix, rem_length, dlength, iov_offset, operation->tfrm_op.outdata[outiov_ix].iov_len, operation->tfrm_op.outcount, outiov_ix)); + + outdesc->dma_descr->buf = (char*)virt_to_phys(operation->tfrm_op.outdata[outiov_ix].iov_base + iov_offset); + outdesc->dma_descr->after = outdesc->dma_descr->buf + dlength; + outdesc->dma_descr->md = REG_TYPE_CONV(unsigned short int, struct strcop_meta_in, mi); + + out_length -= dlength; + iov_offset += dlength; + if (iov_offset >= operation->tfrm_op.outdata[outiov_ix].iov_len) { + iov_offset = 0; + ++outiov_ix; + } + } + if (out_length > 0){ + DEBUG_API(printk("create_input_descriptors: not enough room for output, %d remained\n", out_length)); + err = -EINVAL; + goto error_cleanup; + } + /* Set sync in last descriptor. */ + mi.sync = 1; + outdesc->dma_descr->md = REG_TYPE_CONV(unsigned short int, struct strcop_meta_in, mi); + + *id = head.next; + return 0; + + error_cleanup: + while (head.next) { + outdesc = head.next->next; + free_cdesc(head.next); + head.next = outdesc; + } + return err; +} + + +static int create_output_descriptors(struct cryptocop_operation *operation, int *iniov_ix, int *iniov_offset, size_t desc_len, struct cryptocop_dma_desc **current_out_cdesc, struct strcop_meta_out *meta_out, int alloc_flag) +{ + while (desc_len != 0) { + struct cryptocop_dma_desc *cdesc; + int rem_length = operation->tfrm_op.indata[*iniov_ix].iov_len - *iniov_offset; + int dlength = (desc_len < rem_length) ? desc_len : rem_length; + + cdesc = alloc_cdesc(alloc_flag); + if (!cdesc) { + DEBUG_API(printk("create_output_descriptors: alloc_cdesc\n")); + return -ENOMEM; + } + (*current_out_cdesc)->next = cdesc; + (*current_out_cdesc) = cdesc; + + cdesc->free_buf = NULL; + + cdesc->dma_descr->buf = (char*)virt_to_phys(operation->tfrm_op.indata[*iniov_ix].iov_base + *iniov_offset); + cdesc->dma_descr->after = cdesc->dma_descr->buf + dlength; + + desc_len -= dlength; + *iniov_offset += dlength; + assert(desc_len >= 0); + if (*iniov_offset >= operation->tfrm_op.indata[*iniov_ix].iov_len) { + *iniov_offset = 0; + ++(*iniov_ix); + if (*iniov_ix > operation->tfrm_op.incount) { + DEBUG_API(printk("create_output_descriptors: not enough indata in operation.")); + return -EINVAL; + } + } + cdesc->dma_descr->md = REG_TYPE_CONV(unsigned short int, struct strcop_meta_out, (*meta_out)); + } /* while (desc_len != 0) */ + /* Last DMA descriptor gets a 'wait' bit to signal expected change in metadata. */ + (*current_out_cdesc)->dma_descr->wait = 1; /* This will set extraneous WAIT in some situations, e.g. when padding hashes and checksums. */ + + return 0; +} + + +static int append_input_descriptors(struct cryptocop_operation *operation, struct cryptocop_dma_desc **current_in_cdesc, struct cryptocop_dma_desc **current_out_cdesc, struct cryptocop_tfrm_ctx *tc, int alloc_flag) +{ + DEBUG(printk("append_input_descriptors, tc=0x%p, unit_no=%d\n", tc, tc->unit_no)); + if (tc->tcfg) { + int failed = 0; + struct cryptocop_dma_desc *idescs = NULL; + DEBUG(printk("append_input_descriptors: pushing output, consumed %d produced %d bytes.\n", tc->consumed, tc->produced)); + if (tc->pad_descs) { + DEBUG(printk("append_input_descriptors: append pad descriptors to DMA out list.\n")); + while (tc->pad_descs) { + DEBUG(printk("append descriptor 0x%p\n", tc->pad_descs)); + (*current_out_cdesc)->next = tc->pad_descs; + tc->pad_descs = tc->pad_descs->next; + (*current_out_cdesc) = (*current_out_cdesc)->next; + } + } + + /* Setup and append output descriptors to DMA in list. */ + if (tc->unit_no == src_dma){ + /* mem2mem. Setup DMA in descriptors to discard all input prior to the requested mem2mem data. */ + struct strcop_meta_in mi = {.sync = 0, .dmasel = src_dma}; + unsigned int start_ix = tc->start_ix; + while (start_ix){ + unsigned int desclen = start_ix < MEM2MEM_DISCARD_BUF_LENGTH ? start_ix : MEM2MEM_DISCARD_BUF_LENGTH; + (*current_in_cdesc)->next = alloc_cdesc(alloc_flag); + if (!(*current_in_cdesc)->next){ + DEBUG_API(printk("append_input_descriptors: alloc_cdesc mem2mem discard failed\n")); + return -ENOMEM; + } + (*current_in_cdesc) = (*current_in_cdesc)->next; + (*current_in_cdesc)->dma_descr->buf = (char*)virt_to_phys(mem2mem_discard_buf); + (*current_in_cdesc)->dma_descr->after = (*current_in_cdesc)->dma_descr->buf + desclen; + (*current_in_cdesc)->dma_descr->md = REG_TYPE_CONV(unsigned short int, struct strcop_meta_in, mi); + start_ix -= desclen; + } + mi.sync = 1; + (*current_in_cdesc)->dma_descr->md = REG_TYPE_CONV(unsigned short int, struct strcop_meta_in, mi); + } + + failed = create_input_descriptors(operation, tc, &idescs, alloc_flag); + if (failed){ + DEBUG_API(printk("append_input_descriptors: output descriptor setup failed\n")); + return failed; + } + DEBUG(printk("append_input_descriptors: append output descriptors to DMA in list.\n")); + while (idescs) { + DEBUG(printk("append descriptor 0x%p\n", idescs)); + (*current_in_cdesc)->next = idescs; + idescs = idescs->next; + (*current_in_cdesc) = (*current_in_cdesc)->next; + } + } + return 0; +} + + + +static int cryptocop_setup_dma_list(struct cryptocop_operation *operation, struct cryptocop_int_operation **int_op, int alloc_flag) +{ + struct cryptocop_session *sess; + struct cryptocop_transform_ctx *tctx; + + struct cryptocop_tfrm_ctx digest_ctx = { + .previous_src = src_none, + .current_src = src_none, + .start_ix = 0, + .requires_padding = 1, + .strict_block_length = 0, + .hash_conf = 0, + .hash_mode = 0, + .ciph_conf = 0, + .cbcmode = 0, + .decrypt = 0, + .consumed = 0, + .produced = 0, + .pad_descs = NULL, + .active = 0, + .done = 0, + .prev_src = NULL, + .curr_src = NULL, + .tcfg = NULL}; + struct cryptocop_tfrm_ctx cipher_ctx = { + .previous_src = src_none, + .current_src = src_none, + .start_ix = 0, + .requires_padding = 0, + .strict_block_length = 1, + .hash_conf = 0, + .hash_mode = 0, + .ciph_conf = 0, + .cbcmode = 0, + .decrypt = 0, + .consumed = 0, + .produced = 0, + .pad_descs = NULL, + .active = 0, + .done = 0, + .prev_src = NULL, + .curr_src = NULL, + .tcfg = NULL}; + struct cryptocop_tfrm_ctx csum_ctx = { + .previous_src = src_none, + .current_src = src_none, + .start_ix = 0, + .blocklength = 2, + .requires_padding = 1, + .strict_block_length = 0, + .hash_conf = 0, + .hash_mode = 0, + .ciph_conf = 0, + .cbcmode = 0, + .decrypt = 0, + .consumed = 0, + .produced = 0, + .pad_descs = NULL, + .active = 0, + .done = 0, + .tcfg = NULL, + .prev_src = NULL, + .curr_src = NULL, + .unit_no = src_csum}; + struct cryptocop_tfrm_cfg *tcfg = operation->tfrm_op.tfrm_cfg; + + unsigned int indata_ix = 0; + + /* iovec accounting. */ + int iniov_ix = 0; + int iniov_offset = 0; + + /* Operation descriptor cfg traversal pointer. */ + struct cryptocop_desc *odsc; + + int failed = 0; + /* List heads for allocated descriptors. */ + struct cryptocop_dma_desc out_cdesc_head = {0}; + struct cryptocop_dma_desc in_cdesc_head = {0}; + + struct cryptocop_dma_desc *current_out_cdesc = &out_cdesc_head; + struct cryptocop_dma_desc *current_in_cdesc = &in_cdesc_head; + + struct cryptocop_tfrm_ctx *output_tc = NULL; + void *iop_alloc_ptr; + + assert(operation != NULL); + assert(int_op != NULL); + + DEBUG(printk("cryptocop_setup_dma_list: start\n")); + DEBUG(print_cryptocop_operation(operation)); + + sess = get_session(operation->sid); + if (!sess) { + DEBUG_API(printk("cryptocop_setup_dma_list: no session found for operation.\n")); + failed = -EINVAL; + goto error_cleanup; + } + iop_alloc_ptr = kmalloc(DESCR_ALLOC_PAD + sizeof(struct cryptocop_int_operation), alloc_flag); + if (!iop_alloc_ptr) { + DEBUG_API(printk("cryptocop_setup_dma_list: kmalloc cryptocop_int_operation\n")); + failed = -ENOMEM; + goto error_cleanup; + } + (*int_op) = (struct cryptocop_int_operation*)(((unsigned long int)(iop_alloc_ptr + DESCR_ALLOC_PAD + offsetof(struct cryptocop_int_operation, ctx_out)) & ~0x0000001F) - offsetof(struct cryptocop_int_operation, ctx_out)); + DEBUG(memset((*int_op), 0xff, sizeof(struct cryptocop_int_operation))); + (*int_op)->alloc_ptr = iop_alloc_ptr; + DEBUG(printk("cryptocop_setup_dma_list: *int_op=0x%p, alloc_ptr=0x%p\n", *int_op, (*int_op)->alloc_ptr)); + + (*int_op)->sid = operation->sid; + (*int_op)->cdesc_out = NULL; + (*int_op)->cdesc_in = NULL; + (*int_op)->tdes_mode = cryptocop_3des_ede; + (*int_op)->csum_mode = cryptocop_csum_le; + (*int_op)->ddesc_out = NULL; + (*int_op)->ddesc_in = NULL; + + /* Scan operation->tfrm_op.tfrm_cfg for bad configuration and set up the local contexts. */ + if (!tcfg) { + DEBUG_API(printk("cryptocop_setup_dma_list: no configured transforms in operation.\n")); + failed = -EINVAL; + goto error_cleanup; + } + while (tcfg) { + tctx = get_transform_ctx(sess, tcfg->tid); + if (!tctx) { + DEBUG_API(printk("cryptocop_setup_dma_list: no transform id %d in session.\n", tcfg->tid)); + failed = -EINVAL; + goto error_cleanup; + } + if (tcfg->inject_ix > operation->tfrm_op.outlen){ + DEBUG_API(printk("cryptocop_setup_dma_list: transform id %d inject_ix (%d) > operation->tfrm_op.outlen(%d)", tcfg->tid, tcfg->inject_ix, operation->tfrm_op.outlen)); + failed = -EINVAL; + goto error_cleanup; + } + switch (tctx->init.alg){ + case cryptocop_alg_mem2mem: + if (cipher_ctx.tcfg != NULL){ + DEBUG_API(printk("cryptocop_setup_dma_list: multiple ciphers in operation.\n")); + failed = -EINVAL; + goto error_cleanup; + } + /* mem2mem is handled as a NULL cipher. */ + cipher_ctx.cbcmode = 0; + cipher_ctx.decrypt = 0; + cipher_ctx.blocklength = 1; + cipher_ctx.ciph_conf = 0; + cipher_ctx.unit_no = src_dma; + cipher_ctx.tcfg = tcfg; + cipher_ctx.tctx = tctx; + break; + case cryptocop_alg_des: + case cryptocop_alg_3des: + case cryptocop_alg_aes: + /* cipher */ + if (cipher_ctx.tcfg != NULL){ + DEBUG_API(printk("cryptocop_setup_dma_list: multiple ciphers in operation.\n")); + failed = -EINVAL; + goto error_cleanup; + } + cipher_ctx.tcfg = tcfg; + cipher_ctx.tctx = tctx; + if (cipher_ctx.tcfg->flags & CRYPTOCOP_DECRYPT){ + cipher_ctx.decrypt = 1; + } + switch (tctx->init.cipher_mode) { + case cryptocop_cipher_mode_ecb: + cipher_ctx.cbcmode = 0; + break; + case cryptocop_cipher_mode_cbc: + cipher_ctx.cbcmode = 1; + break; + default: + DEBUG_API(printk("cryptocop_setup_dma_list: cipher_ctx, bad cipher mode==%d\n", tctx->init.cipher_mode)); + failed = -EINVAL; + goto error_cleanup; + } + DEBUG(printk("cryptocop_setup_dma_list: cipher_ctx, set CBC mode==%d\n", cipher_ctx.cbcmode)); + switch (tctx->init.alg){ + case cryptocop_alg_des: + cipher_ctx.ciph_conf = 0; + cipher_ctx.unit_no = src_des; + cipher_ctx.blocklength = DES_BLOCK_LENGTH; + break; + case cryptocop_alg_3des: + cipher_ctx.ciph_conf = 1; + cipher_ctx.unit_no = src_des; + cipher_ctx.blocklength = DES_BLOCK_LENGTH; + break; + case cryptocop_alg_aes: + cipher_ctx.ciph_conf = 2; + cipher_ctx.unit_no = src_aes; + cipher_ctx.blocklength = AES_BLOCK_LENGTH; + break; + default: + panic("cryptocop_setup_dma_list: impossible algorithm %d\n", tctx->init.alg); + } + (*int_op)->tdes_mode = tctx->init.tdes_mode; + break; + case cryptocop_alg_md5: + case cryptocop_alg_sha1: + /* digest */ + if (digest_ctx.tcfg != NULL){ + DEBUG_API(printk("cryptocop_setup_dma_list: multiple digests in operation.\n")); + failed = -EINVAL; + goto error_cleanup; + } + digest_ctx.tcfg = tcfg; + digest_ctx.tctx = tctx; + digest_ctx.hash_mode = 0; /* Don't use explicit IV in this API. */ + switch (tctx->init.alg){ + case cryptocop_alg_md5: + digest_ctx.blocklength = MD5_BLOCK_LENGTH; + digest_ctx.unit_no = src_md5; + digest_ctx.hash_conf = 1; /* 1 => MD-5 */ + break; + case cryptocop_alg_sha1: + digest_ctx.blocklength = SHA1_BLOCK_LENGTH; + digest_ctx.unit_no = src_sha1; + digest_ctx.hash_conf = 0; /* 0 => SHA-1 */ + break; + default: + panic("cryptocop_setup_dma_list: impossible digest algorithm\n"); + } + break; + case cryptocop_alg_csum: + /* digest */ + if (csum_ctx.tcfg != NULL){ + DEBUG_API(printk("cryptocop_setup_dma_list: multiple checksums in operation.\n")); + failed = -EINVAL; + goto error_cleanup; + } + (*int_op)->csum_mode = tctx->init.csum_mode; + csum_ctx.tcfg = tcfg; + csum_ctx.tctx = tctx; + break; + default: + /* no algorithm. */ + DEBUG_API(printk("cryptocop_setup_dma_list: invalid algorithm %d specified in tfrm %d.\n", tctx->init.alg, tcfg->tid)); + failed = -EINVAL; + goto error_cleanup; + } + tcfg = tcfg->next; + } + /* Download key if a cipher is used. */ + if (cipher_ctx.tcfg && (cipher_ctx.tctx->init.alg != cryptocop_alg_mem2mem)){ + struct cryptocop_dma_desc *key_desc = NULL; + + failed = setup_key_dl_desc(&cipher_ctx, &key_desc, alloc_flag); + if (failed) { + DEBUG_API(printk("cryptocop_setup_dma_list: setup key dl\n")); + goto error_cleanup; + } + current_out_cdesc->next = key_desc; + current_out_cdesc = key_desc; + indata_ix += (unsigned int)(key_desc->dma_descr->after - key_desc->dma_descr->buf); + + /* Download explicit IV if a cipher is used and CBC mode and explicit IV selected. */ + if ((cipher_ctx.tctx->init.cipher_mode == cryptocop_cipher_mode_cbc) && (cipher_ctx.tcfg->flags & CRYPTOCOP_EXPLICIT_IV)) { + struct cryptocop_dma_desc *iv_desc = NULL; + + DEBUG(printk("cryptocop_setup_dma_list: setup cipher CBC IV descriptor.\n")); + + failed = setup_cipher_iv_desc(&cipher_ctx, &iv_desc, alloc_flag); + if (failed) { + DEBUG_API(printk("cryptocop_setup_dma_list: CBC IV descriptor.\n")); + goto error_cleanup; + } + current_out_cdesc->next = iv_desc; + current_out_cdesc = iv_desc; + indata_ix += (unsigned int)(iv_desc->dma_descr->after - iv_desc->dma_descr->buf); + } + } + + /* Process descriptors. */ + odsc = operation->tfrm_op.desc; + while (odsc) { + struct cryptocop_desc_cfg *dcfg = odsc->cfg; + struct strcop_meta_out meta_out = {0}; + size_t desc_len = odsc->length; + int active_count, eop_needed_count; + + output_tc = NULL; + + DEBUG(printk("cryptocop_setup_dma_list: parsing an operation descriptor\n")); + + while (dcfg) { + struct cryptocop_tfrm_ctx *tc = NULL; + + DEBUG(printk("cryptocop_setup_dma_list: parsing an operation descriptor configuration.\n")); + /* Get the local context for the transform and mark it as the output unit if it produces output. */ + if (digest_ctx.tcfg && (digest_ctx.tcfg->tid == dcfg->tid)){ + tc = &digest_ctx; + } else if (cipher_ctx.tcfg && (cipher_ctx.tcfg->tid == dcfg->tid)){ + tc = &cipher_ctx; + } else if (csum_ctx.tcfg && (csum_ctx.tcfg->tid == dcfg->tid)){ + tc = &csum_ctx; + } + if (!tc) { + DEBUG_API(printk("cryptocop_setup_dma_list: invalid transform %d specified in descriptor.\n", dcfg->tid)); + failed = -EINVAL; + goto error_cleanup; + } + if (tc->done) { + DEBUG_API(printk("cryptocop_setup_dma_list: completed transform %d reused.\n", dcfg->tid)); + failed = -EINVAL; + goto error_cleanup; + } + if (!tc->active) { + tc->start_ix = indata_ix; + tc->active = 1; + } + + tc->previous_src = tc->current_src; + tc->prev_src = tc->curr_src; + /* Map source unit id to DMA source config. */ + switch (dcfg->src){ + case cryptocop_source_dma: + tc->current_src = src_dma; + break; + case cryptocop_source_des: + tc->current_src = src_des; + break; + case cryptocop_source_3des: + tc->current_src = src_des; + break; + case cryptocop_source_aes: + tc->current_src = src_aes; + break; + case cryptocop_source_md5: + case cryptocop_source_sha1: + case cryptocop_source_csum: + case cryptocop_source_none: + default: + /* We do not allow using accumulating style units (SHA-1, MD5, checksum) as sources to other units. + */ + DEBUG_API(printk("cryptocop_setup_dma_list: bad unit source configured %d.\n", dcfg->src)); + failed = -EINVAL; + goto error_cleanup; + } + if (tc->current_src != src_dma) { + /* Find the unit we are sourcing from. */ + if (digest_ctx.unit_no == tc->current_src){ + tc->curr_src = &digest_ctx; + } else if (cipher_ctx.unit_no == tc->current_src){ + tc->curr_src = &cipher_ctx; + } else if (csum_ctx.unit_no == tc->current_src){ + tc->curr_src = &csum_ctx; + } + if ((tc->curr_src == tc) && (tc->unit_no != src_dma)){ + DEBUG_API(printk("cryptocop_setup_dma_list: unit %d configured to source from itself.\n", tc->unit_no)); + failed = -EINVAL; + goto error_cleanup; + } + } else { + tc->curr_src = NULL; + } + + /* Detect source switch. */ + DEBUG(printk("cryptocop_setup_dma_list: tc->active=%d tc->unit_no=%d tc->current_src=%d tc->previous_src=%d, tc->curr_src=0x%p, tc->prev_srv=0x%p\n", tc->active, tc->unit_no, tc->current_src, tc->previous_src, tc->curr_src, tc->prev_src)); + if (tc->active && (tc->current_src != tc->previous_src)) { + /* Only allow source switch when both the old source unit and the new one have + * no pending data to process (i.e. the consumed length must be a multiple of the + * transform blocklength). */ + /* Note: if the src == NULL we are actually sourcing from DMA out. */ + if (((tc->prev_src != NULL) && (tc->prev_src->consumed % tc->prev_src->blocklength)) || + ((tc->curr_src != NULL) && (tc->curr_src->consumed % tc->curr_src->blocklength))) + { + DEBUG_API(printk("cryptocop_setup_dma_list: can only disconnect from or connect to a unit on a multiple of the blocklength, old: cons=%d, prod=%d, block=%d, new: cons=%d prod=%d, block=%d.\n", tc->prev_src ? tc->prev_src->consumed : INT_MIN, tc->prev_src ? tc->prev_src->produced : INT_MIN, tc->prev_src ? tc->prev_src->blocklength : INT_MIN, tc->curr_src ? tc->curr_src->consumed : INT_MIN, tc->curr_src ? tc->curr_src->produced : INT_MIN, tc->curr_src ? tc->curr_src->blocklength : INT_MIN)); + failed = -EINVAL; + goto error_cleanup; + } + } + /* Detect unit deactivation. */ + if (dcfg->last) { + /* Length check of this is handled below. */ + tc->done = 1; + } + dcfg = dcfg->next; + } /* while (dcfg) */ + DEBUG(printk("cryptocop_setup_dma_list: parsing operation descriptor configuration complete.\n")); + + if (cipher_ctx.active && (cipher_ctx.curr_src != NULL) && !cipher_ctx.curr_src->active){ + DEBUG_API(printk("cryptocop_setup_dma_list: cipher source from inactive unit %d\n", cipher_ctx.curr_src->unit_no)); + failed = -EINVAL; + goto error_cleanup; + } + if (digest_ctx.active && (digest_ctx.curr_src != NULL) && !digest_ctx.curr_src->active){ + DEBUG_API(printk("cryptocop_setup_dma_list: digest source from inactive unit %d\n", digest_ctx.curr_src->unit_no)); + failed = -EINVAL; + goto error_cleanup; + } + if (csum_ctx.active && (csum_ctx.curr_src != NULL) && !csum_ctx.curr_src->active){ + DEBUG_API(printk("cryptocop_setup_dma_list: cipher source from inactive unit %d\n", csum_ctx.curr_src->unit_no)); + failed = -EINVAL; + goto error_cleanup; + } + + /* Update consumed and produced lengths. + + The consumed length accounting here is actually cheating. If a unit source from DMA (or any + other unit that process data in blocks of one octet) it is correct, but if it source from a + block processing unit, i.e. a cipher, it will be temporarily incorrect at some times. However + since it is only allowed--by the HW--to change source to or from a block processing unit at times where that + unit has processed an exact multiple of its block length the end result will be correct. + Beware that if the source change restriction change this code will need to be (much) reworked. + */ + DEBUG(printk("cryptocop_setup_dma_list: desc->length=%d, desc_len=%d.\n", odsc->length, desc_len)); + + if (csum_ctx.active) { + csum_ctx.consumed += desc_len; + if (csum_ctx.done) { + csum_ctx.produced = 2; + } + DEBUG(printk("cryptocop_setup_dma_list: csum_ctx producing: consumed=%d, produced=%d, blocklength=%d.\n", csum_ctx.consumed, csum_ctx.produced, csum_ctx.blocklength)); + } + if (digest_ctx.active) { + digest_ctx.consumed += desc_len; + if (digest_ctx.done) { + if (digest_ctx.unit_no == src_md5) { + digest_ctx.produced = MD5_STATE_LENGTH; + } else { + digest_ctx.produced = SHA1_STATE_LENGTH; + } + } + DEBUG(printk("cryptocop_setup_dma_list: digest_ctx producing: consumed=%d, produced=%d, blocklength=%d.\n", digest_ctx.consumed, digest_ctx.produced, digest_ctx.blocklength)); + } + if (cipher_ctx.active) { + /* Ciphers are allowed only to source from DMA out. That is filtered above. */ + assert(cipher_ctx.current_src == src_dma); + cipher_ctx.consumed += desc_len; + cipher_ctx.produced = cipher_ctx.blocklength * (cipher_ctx.consumed / cipher_ctx.blocklength); + if (cipher_ctx.cbcmode && !(cipher_ctx.tcfg->flags & CRYPTOCOP_EXPLICIT_IV) && cipher_ctx.produced){ + cipher_ctx.produced -= cipher_ctx.blocklength; /* Compensate for CBC iv. */ + } + DEBUG(printk("cryptocop_setup_dma_list: cipher_ctx producing: consumed=%d, produced=%d, blocklength=%d.\n", cipher_ctx.consumed, cipher_ctx.produced, cipher_ctx.blocklength)); + } + + /* Setup the DMA out descriptors. */ + /* Configure the metadata. */ + active_count = 0; + eop_needed_count = 0; + if (cipher_ctx.active) { + ++active_count; + if (cipher_ctx.unit_no == src_dma){ + /* mem2mem */ + meta_out.ciphsel = src_none; + } else { + meta_out.ciphsel = cipher_ctx.current_src; + } + meta_out.ciphconf = cipher_ctx.ciph_conf; + meta_out.cbcmode = cipher_ctx.cbcmode; + meta_out.decrypt = cipher_ctx.decrypt; + DEBUG(printk("set ciphsel=%d ciphconf=%d cbcmode=%d decrypt=%d\n", meta_out.ciphsel, meta_out.ciphconf, meta_out.cbcmode, meta_out.decrypt)); + if (cipher_ctx.done) ++eop_needed_count; + } else { + meta_out.ciphsel = src_none; + } + + if (digest_ctx.active) { + ++active_count; + meta_out.hashsel = digest_ctx.current_src; + meta_out.hashconf = digest_ctx.hash_conf; + meta_out.hashmode = 0; /* Explicit mode is not used here. */ + DEBUG(printk("set hashsel=%d hashconf=%d hashmode=%d\n", meta_out.hashsel, meta_out.hashconf, meta_out.hashmode)); + if (digest_ctx.done) { + assert(digest_ctx.pad_descs == NULL); + failed = create_pad_descriptor(&digest_ctx, &digest_ctx.pad_descs, alloc_flag); + if (failed) { + DEBUG_API(printk("cryptocop_setup_dma_list: failed digest pad creation.\n")); + goto error_cleanup; + } + } + } else { + meta_out.hashsel = src_none; + } + + if (csum_ctx.active) { + ++active_count; + meta_out.csumsel = csum_ctx.current_src; + if (csum_ctx.done) { + assert(csum_ctx.pad_descs == NULL); + failed = create_pad_descriptor(&csum_ctx, &csum_ctx.pad_descs, alloc_flag); + if (failed) { + DEBUG_API(printk("cryptocop_setup_dma_list: failed csum pad creation.\n")); + goto error_cleanup; + } + } + } else { + meta_out.csumsel = src_none; + } + DEBUG(printk("cryptocop_setup_dma_list: %d eop needed, %d active units\n", eop_needed_count, active_count)); + /* Setup DMA out descriptors for the indata. */ + failed = create_output_descriptors(operation, &iniov_ix, &iniov_offset, desc_len, ¤t_out_cdesc, &meta_out, alloc_flag); + if (failed) { + DEBUG_API(printk("cryptocop_setup_dma_list: create_output_descriptors %d\n", failed)); + goto error_cleanup; + } + /* Setup out EOP. If there are active units that are not done here they cannot get an EOP + * so we ust setup a zero length descriptor to DMA to signal EOP only to done units. + * If there is a pad descriptor EOP for the padded unit will be EOPed by it. + */ + assert(active_count >= eop_needed_count); + assert((eop_needed_count == 0) || (eop_needed_count == 1)); + if (eop_needed_count) { + /* This means that the bulk operation (cipeher/m2m) is terminated. */ + if (active_count > 1) { + /* Use zero length EOP descriptor. */ + struct cryptocop_dma_desc *ed = alloc_cdesc(alloc_flag); + struct strcop_meta_out ed_mo = {0}; + if (!ed) { + DEBUG_API(printk("cryptocop_setup_dma_list: alloc EOP descriptor for cipher\n")); + failed = -ENOMEM; + goto error_cleanup; + } + + assert(cipher_ctx.active && cipher_ctx.done); + + if (cipher_ctx.unit_no == src_dma){ + /* mem2mem */ + ed_mo.ciphsel = src_none; + } else { + ed_mo.ciphsel = cipher_ctx.current_src; + } + ed_mo.ciphconf = cipher_ctx.ciph_conf; + ed_mo.cbcmode = cipher_ctx.cbcmode; + ed_mo.decrypt = cipher_ctx.decrypt; + + ed->free_buf = NULL; + ed->dma_descr->wait = 1; + ed->dma_descr->out_eop = 1; + + ed->dma_descr->buf = (char*)virt_to_phys(&ed); /* Use any valid physical address for zero length descriptor. */ + ed->dma_descr->after = ed->dma_descr->buf; + ed->dma_descr->md = REG_TYPE_CONV(unsigned short int, struct strcop_meta_out, ed_mo); + current_out_cdesc->next = ed; + current_out_cdesc = ed; + } else { + /* Set EOP in the current out descriptor since the only active module is + * the one needing the EOP. */ + + current_out_cdesc->dma_descr->out_eop = 1; + } + } + + if (cipher_ctx.done && cipher_ctx.active) cipher_ctx.active = 0; + if (digest_ctx.done && digest_ctx.active) digest_ctx.active = 0; + if (csum_ctx.done && csum_ctx.active) csum_ctx.active = 0; + indata_ix += odsc->length; + odsc = odsc->next; + } /* while (odsc) */ /* Process descriptors. */ + DEBUG(printk("cryptocop_setup_dma_list: done parsing operation descriptors\n")); + if (cipher_ctx.tcfg && (cipher_ctx.active || !cipher_ctx.done)){ + DEBUG_API(printk("cryptocop_setup_dma_list: cipher operation not terminated.\n")); + failed = -EINVAL; + goto error_cleanup; + } + if (digest_ctx.tcfg && (digest_ctx.active || !digest_ctx.done)){ + DEBUG_API(printk("cryptocop_setup_dma_list: digest operation not terminated.\n")); + failed = -EINVAL; + goto error_cleanup; + } + if (csum_ctx.tcfg && (csum_ctx.active || !csum_ctx.done)){ + DEBUG_API(printk("cryptocop_setup_dma_list: csum operation not terminated.\n")); + failed = -EINVAL; + goto error_cleanup; + } + + failed = append_input_descriptors(operation, ¤t_in_cdesc, ¤t_out_cdesc, &cipher_ctx, alloc_flag); + if (failed){ + DEBUG_API(printk("cryptocop_setup_dma_list: append_input_descriptors cipher_ctx %d\n", failed)); + goto error_cleanup; + } + failed = append_input_descriptors(operation, ¤t_in_cdesc, ¤t_out_cdesc, &digest_ctx, alloc_flag); + if (failed){ + DEBUG_API(printk("cryptocop_setup_dma_list: append_input_descriptors cipher_ctx %d\n", failed)); + goto error_cleanup; + } + failed = append_input_descriptors(operation, ¤t_in_cdesc, ¤t_out_cdesc, &csum_ctx, alloc_flag); + if (failed){ + DEBUG_API(printk("cryptocop_setup_dma_list: append_input_descriptors cipher_ctx %d\n", failed)); + goto error_cleanup; + } + + DEBUG(printk("cryptocop_setup_dma_list: int_op=0x%p, *int_op=0x%p\n", int_op, *int_op)); + (*int_op)->cdesc_out = out_cdesc_head.next; + (*int_op)->cdesc_in = in_cdesc_head.next; + DEBUG(printk("cryptocop_setup_dma_list: out_cdesc_head=0x%p in_cdesc_head=0x%p\n", (*int_op)->cdesc_out, (*int_op)->cdesc_in)); + + setup_descr_chain(out_cdesc_head.next); + setup_descr_chain(in_cdesc_head.next); + + /* Last but not least: mark the last DMA in descriptor for a INTR and EOL and the the + * last DMA out descriptor for EOL. + */ + current_in_cdesc->dma_descr->intr = 1; + current_in_cdesc->dma_descr->eol = 1; + current_out_cdesc->dma_descr->eol = 1; + + /* Setup DMA contexts. */ + (*int_op)->ctx_out.next = NULL; + (*int_op)->ctx_out.eol = 1; + (*int_op)->ctx_out.intr = 0; + (*int_op)->ctx_out.store_mode = 0; + (*int_op)->ctx_out.en = 0; + (*int_op)->ctx_out.dis = 0; + (*int_op)->ctx_out.md0 = 0; + (*int_op)->ctx_out.md1 = 0; + (*int_op)->ctx_out.md2 = 0; + (*int_op)->ctx_out.md3 = 0; + (*int_op)->ctx_out.md4 = 0; + (*int_op)->ctx_out.saved_data = (dma_descr_data*)virt_to_phys((*int_op)->cdesc_out->dma_descr); + (*int_op)->ctx_out.saved_data_buf = (*int_op)->cdesc_out->dma_descr->buf; /* Already physical address. */ + + (*int_op)->ctx_in.next = NULL; + (*int_op)->ctx_in.eol = 1; + (*int_op)->ctx_in.intr = 0; + (*int_op)->ctx_in.store_mode = 0; + (*int_op)->ctx_in.en = 0; + (*int_op)->ctx_in.dis = 0; + (*int_op)->ctx_in.md0 = 0; + (*int_op)->ctx_in.md1 = 0; + (*int_op)->ctx_in.md2 = 0; + (*int_op)->ctx_in.md3 = 0; + (*int_op)->ctx_in.md4 = 0; + + (*int_op)->ctx_in.saved_data = (dma_descr_data*)virt_to_phys((*int_op)->cdesc_in->dma_descr); + (*int_op)->ctx_in.saved_data_buf = (*int_op)->cdesc_in->dma_descr->buf; /* Already physical address. */ + + DEBUG(printk("cryptocop_setup_dma_list: done\n")); + return 0; + +error_cleanup: + { + /* Free all allocated resources. */ + struct cryptocop_dma_desc *tmp_cdesc; + while (digest_ctx.pad_descs){ + tmp_cdesc = digest_ctx.pad_descs->next; + free_cdesc(digest_ctx.pad_descs); + digest_ctx.pad_descs = tmp_cdesc; + } + while (csum_ctx.pad_descs){ + tmp_cdesc = csum_ctx.pad_descs->next; + free_cdesc(csum_ctx.pad_descs); + csum_ctx.pad_descs = tmp_cdesc; + } + assert(cipher_ctx.pad_descs == NULL); /* The ciphers are never padded. */ + + if (*int_op != NULL) delete_internal_operation(*int_op); + } + DEBUG_API(printk("cryptocop_setup_dma_list: done with error %d\n", failed)); + return failed; +} + + +static void delete_internal_operation(struct cryptocop_int_operation *iop) +{ + void *ptr = iop->alloc_ptr; + struct cryptocop_dma_desc *cd = iop->cdesc_out; + struct cryptocop_dma_desc *next; + + DEBUG(printk("delete_internal_operation: iop=0x%p, alloc_ptr=0x%p\n", iop, ptr)); + + while (cd) { + next = cd->next; + free_cdesc(cd); + cd = next; + } + cd = iop->cdesc_in; + while (cd) { + next = cd->next; + free_cdesc(cd); + cd = next; + } + kfree(ptr); +} + +#define MD5_MIN_PAD_LENGTH (9) +#define MD5_PAD_LENGTH_FIELD_LENGTH (8) + +static int create_md5_pad(int alloc_flag, unsigned long long hashed_length, char **pad, size_t *pad_length) +{ + size_t padlen = MD5_BLOCK_LENGTH - (hashed_length % MD5_BLOCK_LENGTH); + unsigned char *p; + int i; + unsigned long long int bit_length = hashed_length << 3; + + if (padlen < MD5_MIN_PAD_LENGTH) padlen += MD5_BLOCK_LENGTH; + + p = kmalloc(padlen, alloc_flag); + if (!pad) return -ENOMEM; + + *p = 0x80; + memset(p+1, 0, padlen - 1); + + DEBUG(printk("create_md5_pad: hashed_length=%lld bits == %lld bytes\n", bit_length, hashed_length)); + + i = padlen - MD5_PAD_LENGTH_FIELD_LENGTH; + while (bit_length != 0){ + p[i++] = bit_length % 0x100; + bit_length >>= 8; + } + + *pad = (char*)p; + *pad_length = padlen; + + return 0; +} + +#define SHA1_MIN_PAD_LENGTH (9) +#define SHA1_PAD_LENGTH_FIELD_LENGTH (8) + +static int create_sha1_pad(int alloc_flag, unsigned long long hashed_length, char **pad, size_t *pad_length) +{ + size_t padlen = SHA1_BLOCK_LENGTH - (hashed_length % SHA1_BLOCK_LENGTH); + unsigned char *p; + int i; + unsigned long long int bit_length = hashed_length << 3; + + if (padlen < SHA1_MIN_PAD_LENGTH) padlen += SHA1_BLOCK_LENGTH; + + p = kmalloc(padlen, alloc_flag); + if (!pad) return -ENOMEM; + + *p = 0x80; + memset(p+1, 0, padlen - 1); + + DEBUG(printk("create_sha1_pad: hashed_length=%lld bits == %lld bytes\n", bit_length, hashed_length)); + + i = padlen - 1; + while (bit_length != 0){ + p[i--] = bit_length % 0x100; + bit_length >>= 8; + } + + *pad = (char*)p; + *pad_length = padlen; + + return 0; +} + + +static int transform_ok(struct cryptocop_transform_init *tinit) +{ + switch (tinit->alg){ + case cryptocop_alg_csum: + switch (tinit->csum_mode){ + case cryptocop_csum_le: + case cryptocop_csum_be: + break; + default: + DEBUG_API(printk("transform_ok: Bad mode set for csum transform\n")); + return -EINVAL; + } + case cryptocop_alg_mem2mem: + case cryptocop_alg_md5: + case cryptocop_alg_sha1: + if (tinit->keylen != 0) { + DEBUG_API(printk("transform_ok: non-zero keylength, %d, for a digest/csum algorithm\n", tinit->keylen)); + return -EINVAL; /* This check is a bit strict. */ + } + break; + case cryptocop_alg_des: + if (tinit->keylen != 64) { + DEBUG_API(printk("transform_ok: keylen %d invalid for DES\n", tinit->keylen)); + return -EINVAL; + } + break; + case cryptocop_alg_3des: + if (tinit->keylen != 192) { + DEBUG_API(printk("transform_ok: keylen %d invalid for 3DES\n", tinit->keylen)); + return -EINVAL; + } + break; + case cryptocop_alg_aes: + if (tinit->keylen != 128 && tinit->keylen != 192 && tinit->keylen != 256) { + DEBUG_API(printk("transform_ok: keylen %d invalid for AES\n", tinit->keylen)); + return -EINVAL; + } + break; + case cryptocop_no_alg: + default: + DEBUG_API(printk("transform_ok: no such algorithm %d\n", tinit->alg)); + return -EINVAL; + } + + switch (tinit->alg){ + case cryptocop_alg_des: + case cryptocop_alg_3des: + case cryptocop_alg_aes: + if (tinit->cipher_mode != cryptocop_cipher_mode_ecb && tinit->cipher_mode != cryptocop_cipher_mode_cbc) return -EINVAL; + default: + break; + } + return 0; +} + + +int cryptocop_new_session(cryptocop_session_id *sid, struct cryptocop_transform_init *tinit, int alloc_flag) +{ + struct cryptocop_session *sess; + struct cryptocop_transform_init *tfrm_in = tinit; + struct cryptocop_transform_init *tmp_in; + int no_tfrms = 0; + int i; + unsigned long int flags; + + init_stream_coprocessor(); /* For safety if we are called early */ + + while (tfrm_in){ + int err; + ++no_tfrms; + if ((err = transform_ok(tfrm_in))) { + DEBUG_API(printk("cryptocop_new_session, bad transform\n")); + return err; + } + tfrm_in = tfrm_in->next; + } + if (0 == no_tfrms) { + DEBUG_API(printk("cryptocop_new_session, no transforms specified\n")); + return -EINVAL; + } + + sess = kmalloc(sizeof(struct cryptocop_session), alloc_flag); + if (!sess){ + DEBUG_API(printk("cryptocop_new_session, kmalloc cryptocop_session\n")); + return -ENOMEM; + } + + sess->tfrm_ctx = kmalloc(no_tfrms * sizeof(struct cryptocop_transform_ctx), alloc_flag); + if (!sess->tfrm_ctx) { + DEBUG_API(printk("cryptocop_new_session, kmalloc cryptocop_transform_ctx\n")); + kfree(sess); + return -ENOMEM; + } + + tfrm_in = tinit; + for (i = 0; i < no_tfrms; i++){ + tmp_in = tfrm_in->next; + while (tmp_in){ + if (tmp_in->tid == tfrm_in->tid) { + DEBUG_API(printk("cryptocop_new_session, duplicate transform ids\n")); + kfree(sess->tfrm_ctx); + kfree(sess); + return -EINVAL; + } + tmp_in = tmp_in->next; + } + memcpy(&sess->tfrm_ctx[i].init, tfrm_in, sizeof(struct cryptocop_transform_init)); + sess->tfrm_ctx[i].dec_key_set = 0; + sess->tfrm_ctx[i].next = &sess->tfrm_ctx[i] + 1; + + tfrm_in = tfrm_in->next; + } + sess->tfrm_ctx[i-1].next = NULL; + + spin_lock_irqsave(&cryptocop_sessions_lock, flags); + sess->sid = next_sid; + next_sid++; + /* TODO If we are really paranoid we should do duplicate check to handle sid wraparound. + * OTOH 2^64 is a really large number of session. */ + if (next_sid == 0) next_sid = 1; + + /* Prepend to session list. */ + sess->next = cryptocop_sessions; + cryptocop_sessions = sess; + spin_unlock_irqrestore(&cryptocop_sessions_lock, flags); + *sid = sess->sid; + return 0; +} + + +int cryptocop_free_session(cryptocop_session_id sid) +{ + struct cryptocop_transform_ctx *tc; + struct cryptocop_session *sess = NULL; + struct cryptocop_session *psess = NULL; + unsigned long int flags; + int i; + LIST_HEAD(remove_list); + struct list_head *node, *tmp; + struct cryptocop_prio_job *pj; + + DEBUG(printk("cryptocop_free_session: sid=%lld\n", sid)); + + spin_lock_irqsave(&cryptocop_sessions_lock, flags); + sess = cryptocop_sessions; + while (sess && sess->sid != sid){ + psess = sess; + sess = sess->next; + } + if (sess){ + if (psess){ + psess->next = sess->next; + } else { + cryptocop_sessions = sess->next; + } + } + spin_unlock_irqrestore(&cryptocop_sessions_lock, flags); + + if (!sess) return -EINVAL; + + /* Remove queued jobs. */ + spin_lock_irqsave(&cryptocop_job_queue_lock, flags); + + for (i = 0; i < cryptocop_prio_no_prios; i++){ + if (!list_empty(&(cryptocop_job_queues[i].jobs))){ + list_for_each_safe(node, tmp, &(cryptocop_job_queues[i].jobs)) { + pj = list_entry(node, struct cryptocop_prio_job, node); + if (pj->oper->sid == sid) { + list_move_tail(node, &remove_list); + } + } + } + } + spin_unlock_irqrestore(&cryptocop_job_queue_lock, flags); + + list_for_each_safe(node, tmp, &remove_list) { + list_del(node); + pj = list_entry(node, struct cryptocop_prio_job, node); + pj->oper->operation_status = -EAGAIN; /* EAGAIN is not ideal for job/session terminated but it's the best choice I know of. */ + DEBUG(printk("cryptocop_free_session: pj=0x%p, pj->oper=0x%p, pj->iop=0x%p\n", pj, pj->oper, pj->iop)); + pj->oper->cb(pj->oper, pj->oper->cb_data); + delete_internal_operation(pj->iop); + kfree(pj); + } + + tc = sess->tfrm_ctx; + /* Erase keying data. */ + while (tc){ + DEBUG(printk("cryptocop_free_session: memset keys, tfrm id=%d\n", tc->init.tid)); + memset(tc->init.key, 0xff, CRYPTOCOP_MAX_KEY_LENGTH); + memset(tc->dec_key, 0xff, CRYPTOCOP_MAX_KEY_LENGTH); + tc = tc->next; + } + kfree(sess->tfrm_ctx); + kfree(sess); + + return 0; +} + +static struct cryptocop_session *get_session(cryptocop_session_id sid) +{ + struct cryptocop_session *sess; + unsigned long int flags; + + spin_lock_irqsave(&cryptocop_sessions_lock, flags); + sess = cryptocop_sessions; + while (sess && (sess->sid != sid)){ + sess = sess->next; + } + spin_unlock_irqrestore(&cryptocop_sessions_lock, flags); + + return sess; +} + +static struct cryptocop_transform_ctx *get_transform_ctx(struct cryptocop_session *sess, cryptocop_tfrm_id tid) +{ + struct cryptocop_transform_ctx *tc = sess->tfrm_ctx; + + DEBUG(printk("get_transform_ctx, sess=0x%p, tid=%d\n", sess, tid)); + assert(sess != NULL); + while (tc && tc->init.tid != tid){ + DEBUG(printk("tc=0x%p, tc->next=0x%p\n", tc, tc->next)); + tc = tc->next; + } + DEBUG(printk("get_transform_ctx, returning tc=0x%p\n", tc)); + return tc; +} + + + +/* The AES s-transform matrix (s-box). */ +static const u8 aes_sbox[256] = { + 99, 124, 119, 123, 242, 107, 111, 197, 48, 1, 103, 43, 254, 215, 171, 118, + 202, 130, 201, 125, 250, 89, 71, 240, 173, 212, 162, 175, 156, 164, 114, 192, + 183, 253, 147, 38, 54, 63, 247, 204, 52, 165, 229, 241, 113, 216, 49, 21, + 4, 199, 35, 195, 24, 150, 5, 154, 7, 18, 128, 226, 235, 39, 178, 117, + 9, 131, 44, 26, 27, 110, 90, 160, 82, 59, 214, 179, 41, 227, 47, 132, + 83, 209, 0, 237, 32, 252, 177, 91, 106, 203, 190, 57, 74, 76, 88, 207, + 208, 239, 170, 251, 67, 77, 51, 133, 69, 249, 2, 127, 80, 60, 159, 168, + 81, 163, 64, 143, 146, 157, 56, 245, 188, 182, 218, 33, 16, 255, 243, 210, + 205, 12, 19, 236, 95, 151, 68, 23, 196, 167, 126, 61, 100, 93, 25, 115, + 96, 129, 79, 220, 34, 42, 144, 136, 70, 238, 184, 20, 222, 94, 11, 219, + 224, 50, 58, 10, 73, 6, 36, 92, 194, 211, 172, 98, 145, 149, 228, 121, + 231, 200, 55, 109, 141, 213, 78, 169, 108, 86, 244, 234, 101, 122, 174, 8, + 186, 120, 37, 46, 28, 166, 180, 198, 232, 221, 116, 31, 75, 189, 139, 138, + 112, 62, 181, 102, 72, 3, 246, 14, 97, 53, 87, 185, 134, 193, 29, 158, + 225, 248, 152, 17, 105, 217, 142, 148, 155, 30, 135, 233, 206, 85, 40, 223, + 140, 161, 137, 13, 191, 230, 66, 104, 65, 153, 45, 15, 176, 84, 187, 22 +}; + +/* AES has a 32 bit word round constants for each round in the + * key schedule. round_constant[i] is really Rcon[i+1] in FIPS187. + */ +static u32 round_constant[11] = { + 0x01000000, 0x02000000, 0x04000000, 0x08000000, + 0x10000000, 0x20000000, 0x40000000, 0x80000000, + 0x1B000000, 0x36000000, 0x6C000000 +}; + +/* Apply the s-box to each of the four occtets in w. */ +static u32 aes_ks_subword(const u32 w) +{ + u8 bytes[4]; + + *(u32*)(&bytes[0]) = w; + bytes[0] = aes_sbox[bytes[0]]; + bytes[1] = aes_sbox[bytes[1]]; + bytes[2] = aes_sbox[bytes[2]]; + bytes[3] = aes_sbox[bytes[3]]; + return *(u32*)(&bytes[0]); +} + +/* The encrypt (forward) Rijndael key schedule algorithm pseudo code: + * (Note that AES words are 32 bit long) + * + * KeyExpansion(byte key[4*Nk], word w[Nb*(Nr+1)], Nk){ + * word temp + * i = 0 + * while (i < Nk) { + * w[i] = word(key[4*i, 4*i + 1, 4*i + 2, 4*i + 3]) + * i = i + 1 + * } + * i = Nk + * + * while (i < (Nb * (Nr + 1))) { + * temp = w[i - 1] + * if ((i mod Nk) == 0) { + * temp = SubWord(RotWord(temp)) xor Rcon[i/Nk] + * } + * else if ((Nk > 6) && ((i mod Nk) == 4)) { + * temp = SubWord(temp) + * } + * w[i] = w[i - Nk] xor temp + * } + * RotWord(t) does a 8 bit cyclic shift left on a 32 bit word. + * SubWord(t) applies the AES s-box individually to each octet + * in a 32 bit word. + * + * For AES Nk can have the values 4, 6, and 8 (corresponding to + * values for Nr of 10, 12, and 14). Nb is always 4. + * + * To construct w[i], w[i - 1] and w[i - Nk] must be + * available. Consequently we must keep a state of the last Nk words + * to be able to create the last round keys. + */ +static void get_aes_decrypt_key(unsigned char *dec_key, const unsigned char *key, unsigned int keylength) +{ + u32 temp; + u32 w_ring[8]; /* nk is max 8, use elements 0..(nk - 1) as a ringbuffer */ + u8 w_last_ix; + int i; + u8 nr, nk; + + switch (keylength){ + case 128: + nk = 4; + nr = 10; + break; + case 192: + nk = 6; + nr = 12; + break; + case 256: + nk = 8; + nr = 14; + break; + default: + panic("stream co-processor: bad aes key length in get_aes_decrypt_key\n"); + }; + + /* Need to do host byte order correction here since key is byte oriented and the + * kx algorithm is word (u32) oriented. */ + for (i = 0; i < nk; i+=1) { + w_ring[i] = be32_to_cpu(*(u32*)&key[4*i]); + } + + i = (int)nk; + w_last_ix = i - 1; + while (i < (4 * (nr + 2))) { + temp = w_ring[w_last_ix]; + if (!(i % nk)) { + /* RotWord(temp) */ + temp = (temp << 8) | (temp >> 24); + temp = aes_ks_subword(temp); + temp ^= round_constant[i/nk - 1]; + } else if ((nk > 6) && ((i % nk) == 4)) { + temp = aes_ks_subword(temp); + } + w_last_ix = (w_last_ix + 1) % nk; /* This is the same as (i-Nk) mod Nk */ + temp ^= w_ring[w_last_ix]; + w_ring[w_last_ix] = temp; + + /* We need the round keys for round Nr+1 and Nr+2 (round key + * Nr+2 is the round key beyond the last one used when + * encrypting). Rounds are numbered starting from 0, Nr=10 + * implies 11 rounds are used in encryption/decryption. + */ + if (i >= (4 * nr)) { + /* Need to do host byte order correction here, the key + * is byte oriented. */ + *(u32*)dec_key = cpu_to_be32(temp); + dec_key += 4; + } + ++i; + } +} + + +/**** Job/operation management. ****/ + +int cryptocop_job_queue_insert_csum(struct cryptocop_operation *operation) +{ + return cryptocop_job_queue_insert(cryptocop_prio_kernel_csum, operation); +} + +int cryptocop_job_queue_insert_crypto(struct cryptocop_operation *operation) +{ + return cryptocop_job_queue_insert(cryptocop_prio_kernel, operation); +} + +int cryptocop_job_queue_insert_user_job(struct cryptocop_operation *operation) +{ + return cryptocop_job_queue_insert(cryptocop_prio_user, operation); +} + +static int cryptocop_job_queue_insert(cryptocop_queue_priority prio, struct cryptocop_operation *operation) +{ + int ret; + struct cryptocop_prio_job *pj = NULL; + unsigned long int flags; + + DEBUG(printk("cryptocop_job_queue_insert(%d, 0x%p)\n", prio, operation)); + + if (!operation || !operation->cb){ + DEBUG_API(printk("cryptocop_job_queue_insert oper=0x%p, NULL operation or callback\n", operation)); + return -EINVAL; + } + + if ((ret = cryptocop_job_setup(&pj, operation)) != 0){ + DEBUG_API(printk("cryptocop_job_queue_insert: job setup failed\n")); + return ret; + } + assert(pj != NULL); + + spin_lock_irqsave(&cryptocop_job_queue_lock, flags); + list_add_tail(&pj->node, &cryptocop_job_queues[prio].jobs); + spin_unlock_irqrestore(&cryptocop_job_queue_lock, flags); + + /* Make sure a job is running */ + cryptocop_start_job(); + return 0; +} + +static void cryptocop_do_tasklet(unsigned long unused); +DECLARE_TASKLET (cryptocop_tasklet, cryptocop_do_tasklet, 0); + +static void cryptocop_do_tasklet(unsigned long unused) +{ + struct list_head *node; + struct cryptocop_prio_job *pj = NULL; + unsigned long flags; + + DEBUG(printk("cryptocop_do_tasklet: entering\n")); + + do { + spin_lock_irqsave(&cryptocop_completed_jobs_lock, flags); + if (!list_empty(&cryptocop_completed_jobs)){ + node = cryptocop_completed_jobs.next; + list_del(node); + pj = list_entry(node, struct cryptocop_prio_job, node); + } else { + pj = NULL; + } + spin_unlock_irqrestore(&cryptocop_completed_jobs_lock, flags); + if (pj) { + assert(pj->oper != NULL); + + /* Notify consumer of operation completeness. */ + DEBUG(printk("cryptocop_do_tasklet: callback 0x%p, data 0x%p\n", pj->oper->cb, pj->oper->cb_data)); + + pj->oper->operation_status = 0; /* Job is completed. */ + pj->oper->cb(pj->oper, pj->oper->cb_data); + delete_internal_operation(pj->iop); + kfree(pj); + } + } while (pj != NULL); + + DEBUG(printk("cryptocop_do_tasklet: exiting\n")); +} + +static irqreturn_t +dma_done_interrupt(int irq, void *dev_id, struct pt_regs * regs) +{ + struct cryptocop_prio_job *done_job; + reg_dma_rw_ack_intr ack_intr = { + .data = 1, + }; + + REG_WR (dma, regi_dma9, rw_ack_intr, ack_intr); + + DEBUG(printk("cryptocop DMA done\n")); + + spin_lock(&running_job_lock); + if (cryptocop_running_job == NULL){ + printk("stream co-processor got interrupt when not busy\n"); + spin_unlock(&running_job_lock); + return IRQ_HANDLED; + } + done_job = cryptocop_running_job; + cryptocop_running_job = NULL; + spin_unlock(&running_job_lock); + + /* Start processing a job. */ + if (!spin_trylock(&cryptocop_process_lock)){ + DEBUG(printk("cryptocop irq handler, not starting a job\n")); + } else { + cryptocop_start_job(); + spin_unlock(&cryptocop_process_lock); + } + + done_job->oper->operation_status = 0; /* Job is completed. */ + if (done_job->oper->fast_callback){ + /* This operation wants callback from interrupt. */ + done_job->oper->cb(done_job->oper, done_job->oper->cb_data); + delete_internal_operation(done_job->iop); + kfree(done_job); + } else { + spin_lock(&cryptocop_completed_jobs_lock); + list_add_tail(&(done_job->node), &cryptocop_completed_jobs); + spin_unlock(&cryptocop_completed_jobs_lock); + tasklet_schedule(&cryptocop_tasklet); + } + + DEBUG(printk("cryptocop leave irq handler\n")); + return IRQ_HANDLED; +} + + +/* Setup interrupts and DMA channels. */ +static int init_cryptocop(void) +{ + unsigned long flags; + reg_intr_vect_rw_mask intr_mask; + reg_dma_rw_cfg dma_cfg = {.en = 1}; + reg_dma_rw_intr_mask intr_mask_in = {.data = regk_dma_yes}; /* Only want descriptor interrupts from the DMA in channel. */ + reg_dma_rw_ack_intr ack_intr = {.data = 1,.in_eop = 1 }; + reg_strcop_rw_cfg strcop_cfg = { + .ipend = regk_strcop_little, + .td1 = regk_strcop_e, + .td2 = regk_strcop_d, + .td3 = regk_strcop_e, + .ignore_sync = 0, + .en = 1 + }; + + if (request_irq(DMA9_INTR_VECT, dma_done_interrupt, 0, "stream co-processor DMA", NULL)) panic("request_irq stream co-processor irq dma9"); + + (void)crisv32_request_dma(8, "strcop", DMA_PANIC_ON_ERROR, 0, dma_strp); + (void)crisv32_request_dma(9, "strcop", DMA_PANIC_ON_ERROR, 0, dma_strp); + + local_irq_save(flags); + + /* Reset and enable the cryptocop. */ + strcop_cfg.en = 0; + REG_WR(strcop, regi_strcop, rw_cfg, strcop_cfg); + strcop_cfg.en = 1; + REG_WR(strcop, regi_strcop, rw_cfg, strcop_cfg); + + /* Enable DMA9 interrupt */ + intr_mask = REG_RD(intr_vect, regi_irq, rw_mask); + intr_mask.dma9 = 1; + REG_WR(intr_vect, regi_irq, rw_mask, intr_mask); + + /* Enable DMAs. */ + REG_WR(dma, regi_dma9, rw_cfg, dma_cfg); /* input DMA */ + REG_WR(dma, regi_dma8, rw_cfg, dma_cfg); /* output DMA */ + + /* Set up wordsize = 4 for DMAs. */ + DMA_WR_CMD (regi_dma8, regk_dma_set_w_size4); + DMA_WR_CMD (regi_dma9, regk_dma_set_w_size4); + + /* Enable interrupts. */ + REG_WR(dma, regi_dma9, rw_intr_mask, intr_mask_in); + + /* Clear intr ack. */ + REG_WR(dma, regi_dma9, rw_ack_intr, ack_intr); + + local_irq_restore(flags); + + return 0; +} + +/* Free used cryptocop hw resources (interrupt and DMA channels). */ +static void release_cryptocop(void) +{ + unsigned long flags; + reg_intr_vect_rw_mask intr_mask; + reg_dma_rw_cfg dma_cfg = {.en = 0}; + reg_dma_rw_intr_mask intr_mask_in = {0}; + reg_dma_rw_ack_intr ack_intr = {.data = 1,.in_eop = 1 }; + + local_irq_save(flags); + + /* Clear intr ack. */ + REG_WR(dma, regi_dma9, rw_ack_intr, ack_intr); + + /* Disable DMA9 interrupt */ + intr_mask = REG_RD(intr_vect, regi_irq, rw_mask); + intr_mask.dma9 = 0; + REG_WR(intr_vect, regi_irq, rw_mask, intr_mask); + + /* Disable DMAs. */ + REG_WR(dma, regi_dma9, rw_cfg, dma_cfg); /* input DMA */ + REG_WR(dma, regi_dma8, rw_cfg, dma_cfg); /* output DMA */ + + /* Disable interrupts. */ + REG_WR(dma, regi_dma9, rw_intr_mask, intr_mask_in); + + local_irq_restore(flags); + + free_irq(DMA9_INTR_VECT, NULL); + + (void)crisv32_free_dma(8); + (void)crisv32_free_dma(9); +} + + +/* Init job queue. */ +static int cryptocop_job_queue_init(void) +{ + int i; + + INIT_LIST_HEAD(&cryptocop_completed_jobs); + + for (i = 0; i < cryptocop_prio_no_prios; i++){ + cryptocop_job_queues[i].prio = (cryptocop_queue_priority)i; + INIT_LIST_HEAD(&cryptocop_job_queues[i].jobs); + } + return 0; +} + + +static void cryptocop_job_queue_close(void) +{ + struct list_head *node, *tmp; + struct cryptocop_prio_job *pj = NULL; + unsigned long int process_flags, flags; + int i; + + /* FIXME: This is as yet untested code. */ + + /* Stop strcop from getting an operation to process while we are closing the + module. */ + spin_lock_irqsave(&cryptocop_process_lock, process_flags); + + /* Empty the job queue. */ + spin_lock_irqsave(&cryptocop_process_lock, process_flags); + for (i = 0; i < cryptocop_prio_no_prios; i++){ + if (!list_empty(&(cryptocop_job_queues[i].jobs))){ + list_for_each_safe(node, tmp, &(cryptocop_job_queues[i].jobs)) { + pj = list_entry(node, struct cryptocop_prio_job, node); + list_del(node); + + /* Call callback to notify consumer of job removal. */ + DEBUG(printk("cryptocop_job_queue_close: callback 0x%p, data 0x%p\n", pj->oper->cb, pj->oper->cb_data)); + pj->oper->operation_status = -EINTR; /* Job is terminated without completion. */ + pj->oper->cb(pj->oper, pj->oper->cb_data); + + delete_internal_operation(pj->iop); + kfree(pj); + } + } + } + spin_unlock_irqrestore(&cryptocop_process_lock, process_flags); + + /* Remove the running job, if any. */ + spin_lock_irqsave(&running_job_lock, flags); + if (cryptocop_running_job){ + reg_strcop_rw_cfg rw_cfg; + reg_dma_rw_cfg dma_out_cfg, dma_in_cfg; + + /* Stop DMA. */ + dma_out_cfg = REG_RD(dma, regi_dma8, rw_cfg); + dma_out_cfg.en = regk_dma_no; + REG_WR(dma, regi_dma8, rw_cfg, dma_out_cfg); + + dma_in_cfg = REG_RD(dma, regi_dma9, rw_cfg); + dma_in_cfg.en = regk_dma_no; + REG_WR(dma, regi_dma9, rw_cfg, dma_in_cfg); + + /* Disble the cryptocop. */ + rw_cfg = REG_RD(strcop, regi_strcop, rw_cfg); + rw_cfg.en = 0; + REG_WR(strcop, regi_strcop, rw_cfg, rw_cfg); + + pj = cryptocop_running_job; + cryptocop_running_job = NULL; + + /* Call callback to notify consumer of job removal. */ + DEBUG(printk("cryptocop_job_queue_close: callback 0x%p, data 0x%p\n", pj->oper->cb, pj->oper->cb_data)); + pj->oper->operation_status = -EINTR; /* Job is terminated without completion. */ + pj->oper->cb(pj->oper, pj->oper->cb_data); + + delete_internal_operation(pj->iop); + kfree(pj); + } + spin_unlock_irqrestore(&running_job_lock, flags); + + /* Remove completed jobs, if any. */ + spin_lock_irqsave(&cryptocop_completed_jobs_lock, flags); + + list_for_each_safe(node, tmp, &cryptocop_completed_jobs) { + pj = list_entry(node, struct cryptocop_prio_job, node); + list_del(node); + /* Call callback to notify consumer of job removal. */ + DEBUG(printk("cryptocop_job_queue_close: callback 0x%p, data 0x%p\n", pj->oper->cb, pj->oper->cb_data)); + pj->oper->operation_status = -EINTR; /* Job is terminated without completion. */ + pj->oper->cb(pj->oper, pj->oper->cb_data); + + delete_internal_operation(pj->iop); + kfree(pj); + } + spin_unlock_irqrestore(&cryptocop_completed_jobs_lock, flags); +} + + +static void cryptocop_start_job(void) +{ + int i; + struct cryptocop_prio_job *pj; + unsigned long int flags; + unsigned long int running_job_flags; + reg_strcop_rw_cfg rw_cfg = {.en = 1, .ignore_sync = 0}; + + DEBUG(printk("cryptocop_start_job: entering\n")); + + spin_lock_irqsave(&running_job_lock, running_job_flags); + if (cryptocop_running_job != NULL){ + /* Already running. */ + DEBUG(printk("cryptocop_start_job: already running, exit\n")); + spin_unlock_irqrestore(&running_job_lock, running_job_flags); + return; + } + spin_lock_irqsave(&cryptocop_job_queue_lock, flags); + + /* Check the queues in priority order. */ + for (i = cryptocop_prio_kernel_csum; (i < cryptocop_prio_no_prios) && list_empty(&cryptocop_job_queues[i].jobs); i++); + if (i == cryptocop_prio_no_prios) { + spin_unlock_irqrestore(&cryptocop_job_queue_lock, flags); + spin_unlock_irqrestore(&running_job_lock, running_job_flags); + DEBUG(printk("cryptocop_start_job: no jobs to run\n")); + return; /* No jobs to run */ + } + DEBUG(printk("starting job for prio %d\n", i)); + + /* TODO: Do not starve lower priority jobs. Let in a lower + * prio job for every N-th processed higher prio job or some + * other scheduling policy. This could reasonably be + * tweakable since the optimal balance would depend on the + * type of load on the system. */ + + /* Pull the DMA lists from the job and start the DMA client. */ + pj = list_entry(cryptocop_job_queues[i].jobs.next, struct cryptocop_prio_job, node); + list_del(&pj->node); + spin_unlock_irqrestore(&cryptocop_job_queue_lock, flags); + cryptocop_running_job = pj; + + /* Set config register (3DES and CSUM modes). */ + switch (pj->iop->tdes_mode){ + case cryptocop_3des_eee: + rw_cfg.td1 = regk_strcop_e; + rw_cfg.td2 = regk_strcop_e; + rw_cfg.td3 = regk_strcop_e; + break; + case cryptocop_3des_eed: + rw_cfg.td1 = regk_strcop_e; + rw_cfg.td2 = regk_strcop_e; + rw_cfg.td3 = regk_strcop_d; + break; + case cryptocop_3des_ede: + rw_cfg.td1 = regk_strcop_e; + rw_cfg.td2 = regk_strcop_d; + rw_cfg.td3 = regk_strcop_e; + break; + case cryptocop_3des_edd: + rw_cfg.td1 = regk_strcop_e; + rw_cfg.td2 = regk_strcop_d; + rw_cfg.td3 = regk_strcop_d; + break; + case cryptocop_3des_dee: + rw_cfg.td1 = regk_strcop_d; + rw_cfg.td2 = regk_strcop_e; + rw_cfg.td3 = regk_strcop_e; + break; + case cryptocop_3des_ded: + rw_cfg.td1 = regk_strcop_d; + rw_cfg.td2 = regk_strcop_e; + rw_cfg.td3 = regk_strcop_d; + break; + case cryptocop_3des_dde: + rw_cfg.td1 = regk_strcop_d; + rw_cfg.td2 = regk_strcop_d; + rw_cfg.td3 = regk_strcop_e; + break; + case cryptocop_3des_ddd: + rw_cfg.td1 = regk_strcop_d; + rw_cfg.td2 = regk_strcop_d; + rw_cfg.td3 = regk_strcop_d; + break; + default: + DEBUG(printk("cryptocop_setup_dma_list: bad 3DES mode\n")); + } + switch (pj->iop->csum_mode){ + case cryptocop_csum_le: + rw_cfg.ipend = regk_strcop_little; + break; + case cryptocop_csum_be: + rw_cfg.ipend = regk_strcop_big; + break; + default: + DEBUG(printk("cryptocop_setup_dma_list: bad checksum mode\n")); + } + REG_WR(strcop, regi_strcop, rw_cfg, rw_cfg); + + DEBUG(printk("cryptocop_start_job: starting DMA, new cryptocop_running_job=0x%p\n" + "ctx_in: 0x%p, phys: 0x%p\n" + "ctx_out: 0x%p, phys: 0x%p\n", + pj, + &pj->iop->ctx_in, (char*)virt_to_phys(&pj->iop->ctx_in), + &pj->iop->ctx_out, (char*)virt_to_phys(&pj->iop->ctx_out))); + + /* Start input DMA. */ + DMA_START_CONTEXT(regi_dma9, virt_to_phys(&pj->iop->ctx_in)); + + /* Start output DMA. */ + DMA_START_CONTEXT(regi_dma8, virt_to_phys(&pj->iop->ctx_out)); + + spin_unlock_irqrestore(&running_job_lock, running_job_flags); + DEBUG(printk("cryptocop_start_job: exiting\n")); +} + + +static int cryptocop_job_setup(struct cryptocop_prio_job **pj, struct cryptocop_operation *operation) +{ + int err; + int alloc_flag = operation->in_interrupt ? GFP_ATOMIC : GFP_KERNEL; + void *iop_alloc_ptr = NULL; + + *pj = kmalloc(sizeof (struct cryptocop_prio_job), alloc_flag); + if (!*pj) return -ENOMEM; + + DEBUG(printk("cryptocop_job_setup: operation=0x%p\n", operation)); + + (*pj)->oper = operation; + DEBUG(printk("cryptocop_job_setup, cb=0x%p cb_data=0x%p\n", (*pj)->oper->cb, (*pj)->oper->cb_data)); + + if (operation->use_dmalists) { + DEBUG(print_user_dma_lists(&operation->list_op)); + if (!operation->list_op.inlist || !operation->list_op.outlist || !operation->list_op.out_data_buf || !operation->list_op.in_data_buf){ + DEBUG_API(printk("cryptocop_job_setup: bad indata (use_dmalists)\n")); + kfree(*pj); + return -EINVAL; + } + iop_alloc_ptr = kmalloc(DESCR_ALLOC_PAD + sizeof(struct cryptocop_int_operation), alloc_flag); + if (!iop_alloc_ptr) { + DEBUG_API(printk("cryptocop_job_setup: kmalloc cryptocop_int_operation\n")); + kfree(*pj); + return -ENOMEM; + } + (*pj)->iop = (struct cryptocop_int_operation*)(((unsigned long int)(iop_alloc_ptr + DESCR_ALLOC_PAD + offsetof(struct cryptocop_int_operation, ctx_out)) & ~0x0000001F) - offsetof(struct cryptocop_int_operation, ctx_out)); + DEBUG(memset((*pj)->iop, 0xff, sizeof(struct cryptocop_int_operation))); + (*pj)->iop->alloc_ptr = iop_alloc_ptr; + (*pj)->iop->sid = operation->sid; + (*pj)->iop->cdesc_out = NULL; + (*pj)->iop->cdesc_in = NULL; + (*pj)->iop->tdes_mode = operation->list_op.tdes_mode; + (*pj)->iop->csum_mode = operation->list_op.csum_mode; + (*pj)->iop->ddesc_out = operation->list_op.outlist; + (*pj)->iop->ddesc_in = operation->list_op.inlist; + + /* Setup DMA contexts. */ + (*pj)->iop->ctx_out.next = NULL; + (*pj)->iop->ctx_out.eol = 1; + (*pj)->iop->ctx_out.saved_data = operation->list_op.outlist; + (*pj)->iop->ctx_out.saved_data_buf = operation->list_op.out_data_buf; + + (*pj)->iop->ctx_in.next = NULL; + (*pj)->iop->ctx_in.eol = 1; + (*pj)->iop->ctx_in.saved_data = operation->list_op.inlist; + (*pj)->iop->ctx_in.saved_data_buf = operation->list_op.in_data_buf; + } else { + if ((err = cryptocop_setup_dma_list(operation, &(*pj)->iop, alloc_flag))) { + DEBUG_API(printk("cryptocop_job_setup: cryptocop_setup_dma_list failed %d\n", err)); + kfree(*pj); + return err; + } + } + DEBUG(print_dma_descriptors((*pj)->iop)); + + DEBUG(printk("cryptocop_job_setup, DMA list setup successful\n")); + + return 0; +} + + +static int cryptocop_open(struct inode *inode, struct file *filp) +{ + int p = MINOR(inode->i_rdev); + + if (p != CRYPTOCOP_MINOR) return -EINVAL; + + filp->private_data = NULL; + return 0; +} + + +static int cryptocop_release(struct inode *inode, struct file *filp) +{ + struct cryptocop_private *dev = filp->private_data; + struct cryptocop_private *dev_next; + + while (dev){ + dev_next = dev->next; + if (dev->sid != CRYPTOCOP_SESSION_ID_NONE) { + (void)cryptocop_free_session(dev->sid); + } + kfree(dev); + dev = dev_next; + } + + return 0; +} + + +static int cryptocop_ioctl_close_session(struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg) +{ + struct cryptocop_private *dev = filp->private_data; + struct cryptocop_private *prev_dev = NULL; + struct strcop_session_op *sess_op = (struct strcop_session_op *)arg; + struct strcop_session_op sop; + int err; + + DEBUG(printk("cryptocop_ioctl_close_session\n")); + + if (!access_ok(VERIFY_READ, sess_op, sizeof(struct strcop_session_op))) + return -EFAULT; + err = copy_from_user(&sop, sess_op, sizeof(struct strcop_session_op)); + if (err) return -EFAULT; + + while (dev && (dev->sid != sop.ses_id)) { + prev_dev = dev; + dev = dev->next; + } + if (dev){ + if (prev_dev){ + prev_dev->next = dev->next; + } else { + filp->private_data = dev->next; + } + err = cryptocop_free_session(dev->sid); + if (err) return -EFAULT; + } else { + DEBUG_API(printk("cryptocop_ioctl_close_session: session %lld not found\n", sop.ses_id)); + return -EINVAL; + } + return 0; +} + + +static void ioctl_process_job_callback(struct cryptocop_operation *op, void*cb_data) +{ + struct ioctl_job_cb_ctx *jc = (struct ioctl_job_cb_ctx *)cb_data; + + DEBUG(printk("ioctl_process_job_callback: op=0x%p, cb_data=0x%p\n", op, cb_data)); + + jc->processed = 1; + wake_up(&cryptocop_ioc_process_wq); +} + + +#define CRYPTOCOP_IOCTL_CIPHER_TID (1) +#define CRYPTOCOP_IOCTL_DIGEST_TID (2) +#define CRYPTOCOP_IOCTL_CSUM_TID (3) + +static size_t first_cfg_change_ix(struct strcop_crypto_op *crp_op) +{ + size_t ch_ix = 0; + + if (crp_op->do_cipher) ch_ix = crp_op->cipher_start; + if (crp_op->do_digest && (crp_op->digest_start < ch_ix)) ch_ix = crp_op->digest_start; + if (crp_op->do_csum && (crp_op->csum_start < ch_ix)) ch_ix = crp_op->csum_start; + + DEBUG(printk("first_cfg_change_ix: ix=%d\n", ch_ix)); + return ch_ix; +} + + +static size_t next_cfg_change_ix(struct strcop_crypto_op *crp_op, size_t ix) +{ + size_t ch_ix = INT_MAX; + size_t tmp_ix = 0; + + if (crp_op->do_cipher && ((crp_op->cipher_start + crp_op->cipher_len) > ix)){ + if (crp_op->cipher_start > ix) { + ch_ix = crp_op->cipher_start; + } else { + ch_ix = crp_op->cipher_start + crp_op->cipher_len; + } + } + if (crp_op->do_digest && ((crp_op->digest_start + crp_op->digest_len) > ix)){ + if (crp_op->digest_start > ix) { + tmp_ix = crp_op->digest_start; + } else { + tmp_ix = crp_op->digest_start + crp_op->digest_len; + } + if (tmp_ix < ch_ix) ch_ix = tmp_ix; + } + if (crp_op->do_csum && ((crp_op->csum_start + crp_op->csum_len) > ix)){ + if (crp_op->csum_start > ix) { + tmp_ix = crp_op->csum_start; + } else { + tmp_ix = crp_op->csum_start + crp_op->csum_len; + } + if (tmp_ix < ch_ix) ch_ix = tmp_ix; + } + if (ch_ix == INT_MAX) ch_ix = ix; + DEBUG(printk("next_cfg_change_ix prev ix=%d, next ix=%d\n", ix, ch_ix)); + return ch_ix; +} + + +/* Map map_length bytes from the pages starting on *pageix and *pageoffset to iovecs starting on *iovix. + * Return -1 for ok, 0 for fail. */ +static int map_pages_to_iovec(struct iovec *iov, int iovlen, int *iovix, struct page **pages, int nopages, int *pageix, int *pageoffset, int map_length ) +{ + int tmplen; + + assert(iov != NULL); + assert(iovix != NULL); + assert(pages != NULL); + assert(pageix != NULL); + assert(pageoffset != NULL); + + DEBUG(printk("map_pages_to_iovec, map_length=%d, iovlen=%d, *iovix=%d, nopages=%d, *pageix=%d, *pageoffset=%d\n", map_length, iovlen, *iovix, nopages, *pageix, *pageoffset)); + + while (map_length > 0){ + DEBUG(printk("map_pages_to_iovec, map_length=%d, iovlen=%d, *iovix=%d, nopages=%d, *pageix=%d, *pageoffset=%d\n", map_length, iovlen, *iovix, nopages, *pageix, *pageoffset)); + if (*iovix >= iovlen){ + DEBUG_API(printk("map_page_to_iovec: *iovix=%d >= iovlen=%d\n", *iovix, iovlen)); + return 0; + } + if (*pageix >= nopages){ + DEBUG_API(printk("map_page_to_iovec: *pageix=%d >= nopages=%d\n", *pageix, nopages)); + return 0; + } + iov[*iovix].iov_base = (unsigned char*)page_address(pages[*pageix]) + *pageoffset; + tmplen = PAGE_SIZE - *pageoffset; + if (tmplen < map_length){ + (*pageoffset) = 0; + (*pageix)++; + } else { + tmplen = map_length; + (*pageoffset) += map_length; + } + DEBUG(printk("mapping %d bytes from page %d (or %d) to iovec %d\n", tmplen, *pageix, *pageix-1, *iovix)); + iov[*iovix].iov_len = tmplen; + map_length -= tmplen; + (*iovix)++; + } + DEBUG(printk("map_page_to_iovec, exit, *iovix=%d\n", *iovix)); + return -1; +} + + + +static int cryptocop_ioctl_process(struct inode *inode, struct file *filp, unsigned int cmd, unsigned long arg) +{ + int i; + struct cryptocop_private *dev = filp->private_data; + struct strcop_crypto_op *crp_oper = (struct strcop_crypto_op *)arg; + struct strcop_crypto_op oper = {0}; + int err = 0; + struct cryptocop_operation *cop = NULL; + + struct ioctl_job_cb_ctx *jc = NULL; + + struct page **inpages = NULL; + struct page **outpages = NULL; + int noinpages = 0; + int nooutpages = 0; + + struct cryptocop_desc descs[5]; /* Max 5 descriptors are needed, there are three transforms that + * can get connected/disconnected on different places in the indata. */ + struct cryptocop_desc_cfg dcfgs[5*3]; + int desc_ix = 0; + int dcfg_ix = 0; + struct cryptocop_tfrm_cfg ciph_tcfg = {0}; + struct cryptocop_tfrm_cfg digest_tcfg = {0}; + struct cryptocop_tfrm_cfg csum_tcfg = {0}; + + unsigned char *digest_result = NULL; + int digest_length = 0; + int cblocklen = 0; + unsigned char csum_result[CSUM_BLOCK_LENGTH]; + struct cryptocop_session *sess; + + int iovlen = 0; + int iovix = 0; + int pageix = 0; + int pageoffset = 0; + + size_t prev_ix = 0; + size_t next_ix; + + int cipher_active, digest_active, csum_active; + int end_digest, end_csum; + int digest_done = 0; + int cipher_done = 0; + int csum_done = 0; + + DEBUG(printk("cryptocop_ioctl_process\n")); + + if (!access_ok(VERIFY_WRITE, crp_oper, sizeof(struct strcop_crypto_op))){ + DEBUG_API(printk("cryptocop_ioctl_process: !access_ok crp_oper!\n")); + return -EFAULT; + } + if (copy_from_user(&oper, crp_oper, sizeof(struct strcop_crypto_op))) { + DEBUG_API(printk("cryptocop_ioctl_process: copy_from_user\n")); + return -EFAULT; + } + DEBUG(print_strcop_crypto_op(&oper)); + + while (dev && dev->sid != oper.ses_id) dev = dev->next; + if (!dev){ + DEBUG_API(printk("cryptocop_ioctl_process: session %lld not found\n", oper.ses_id)); + return -EINVAL; + } + + /* Check buffers. */ + if (((oper.indata + oper.inlen) < oper.indata) || ((oper.cipher_outdata + oper.cipher_outlen) < oper.cipher_outdata)){ + DEBUG_API(printk("cryptocop_ioctl_process: user buffers wrapped around, bad user!\n")); + return -EINVAL; + } + + if (!access_ok(VERIFY_WRITE, oper.cipher_outdata, oper.cipher_outlen)){ + DEBUG_API(printk("cryptocop_ioctl_process: !access_ok out data!\n")); + return -EFAULT; + } + if (!access_ok(VERIFY_READ, oper.indata, oper.inlen)){ + DEBUG_API(printk("cryptocop_ioctl_process: !access_ok in data!\n")); + return -EFAULT; + } + + cop = kmalloc(sizeof(struct cryptocop_operation), GFP_KERNEL); + if (!cop) { + DEBUG_API(printk("cryptocop_ioctl_process: kmalloc\n")); + return -ENOMEM; + } + jc = kmalloc(sizeof(struct ioctl_job_cb_ctx), GFP_KERNEL); + if (!jc) { + DEBUG_API(printk("cryptocop_ioctl_process: kmalloc\n")); + err = -ENOMEM; + goto error_cleanup; + } + jc->processed = 0; + + cop->cb_data = jc; + cop->cb = ioctl_process_job_callback; + cop->operation_status = 0; + cop->use_dmalists = 0; + cop->in_interrupt = 0; + cop->fast_callback = 0; + cop->tfrm_op.tfrm_cfg = NULL; + cop->tfrm_op.desc = NULL; + cop->tfrm_op.indata = NULL; + cop->tfrm_op.incount = 0; + cop->tfrm_op.inlen = 0; + cop->tfrm_op.outdata = NULL; + cop->tfrm_op.outcount = 0; + cop->tfrm_op.outlen = 0; + + sess = get_session(oper.ses_id); + if (!sess){ + DEBUG_API(printk("cryptocop_ioctl_process: bad session id.\n")); + kfree(cop); + kfree(jc); + return -EINVAL; + } + + if (oper.do_cipher) { + unsigned int cipher_outlen = 0; + struct cryptocop_transform_ctx *tc = get_transform_ctx(sess, CRYPTOCOP_IOCTL_CIPHER_TID); + if (!tc) { + DEBUG_API(printk("cryptocop_ioctl_process: no cipher transform in session.\n")); + err = -EINVAL; + goto error_cleanup; + } + ciph_tcfg.tid = CRYPTOCOP_IOCTL_CIPHER_TID; + ciph_tcfg.inject_ix = 0; + ciph_tcfg.flags = 0; + if ((oper.cipher_start < 0) || (oper.cipher_len <= 0) || (oper.cipher_start > oper.inlen) || ((oper.cipher_start + oper.cipher_len) > oper.inlen)){ + DEBUG_API(printk("cryptocop_ioctl_process: bad cipher length\n")); + kfree(cop); + kfree(jc); + return -EINVAL; + } + cblocklen = tc->init.alg == cryptocop_alg_aes ? AES_BLOCK_LENGTH : DES_BLOCK_LENGTH; + if (oper.cipher_len % cblocklen) { + kfree(cop); + kfree(jc); + DEBUG_API(printk("cryptocop_ioctl_process: cipher inlength not multiple of block length.\n")); + return -EINVAL; + } + cipher_outlen = oper.cipher_len; + if (tc->init.cipher_mode == cryptocop_cipher_mode_cbc){ + if (oper.cipher_explicit) { + ciph_tcfg.flags |= CRYPTOCOP_EXPLICIT_IV; + memcpy(ciph_tcfg.iv, oper.cipher_iv, cblocklen); + } else { + cipher_outlen = oper.cipher_len - cblocklen; + } + } else { + if (oper.cipher_explicit){ + kfree(cop); + kfree(jc); + DEBUG_API(printk("cryptocop_ioctl_process: explicit_iv when not CBC mode\n")); + return -EINVAL; + } + } + if (oper.cipher_outlen != cipher_outlen) { + kfree(cop); + kfree(jc); + DEBUG_API(printk("cryptocop_ioctl_process: cipher_outlen incorrect, should be %d not %d.\n", cipher_outlen, oper.cipher_outlen)); + return -EINVAL; + } + + if (oper.decrypt){ + ciph_tcfg.flags |= CRYPTOCOP_DECRYPT; + } else { + ciph_tcfg.flags |= CRYPTOCOP_ENCRYPT; + } + ciph_tcfg.next = cop->tfrm_op.tfrm_cfg; + cop->tfrm_op.tfrm_cfg = &ciph_tcfg; + } + if (oper.do_digest){ + struct cryptocop_transform_ctx *tc = get_transform_ctx(sess, CRYPTOCOP_IOCTL_DIGEST_TID); + if (!tc) { + DEBUG_API(printk("cryptocop_ioctl_process: no digest transform in session.\n")); + err = -EINVAL; + goto error_cleanup; + } + digest_length = tc->init.alg == cryptocop_alg_md5 ? 16 : 20; + digest_result = kmalloc(digest_length, GFP_KERNEL); + if (!digest_result) { + DEBUG_API(printk("cryptocop_ioctl_process: kmalloc digest_result\n")); + err = -EINVAL; + goto error_cleanup; + } + DEBUG(memset(digest_result, 0xff, digest_length)); + + digest_tcfg.tid = CRYPTOCOP_IOCTL_DIGEST_TID; + digest_tcfg.inject_ix = 0; + ciph_tcfg.inject_ix += digest_length; + if ((oper.digest_start < 0) || (oper.digest_len <= 0) || (oper.digest_start > oper.inlen) || ((oper.digest_start + oper.digest_len) > oper.inlen)){ + DEBUG_API(printk("cryptocop_ioctl_process: bad digest length\n")); + err = -EINVAL; + goto error_cleanup; + } + + digest_tcfg.next = cop->tfrm_op.tfrm_cfg; + cop->tfrm_op.tfrm_cfg = &digest_tcfg; + } + if (oper.do_csum){ + csum_tcfg.tid = CRYPTOCOP_IOCTL_CSUM_TID; + csum_tcfg.inject_ix = digest_length; + ciph_tcfg.inject_ix += 2; + + if ((oper.csum_start < 0) || (oper.csum_len <= 0) || (oper.csum_start > oper.inlen) || ((oper.csum_start + oper.csum_len) > oper.inlen)){ + DEBUG_API(printk("cryptocop_ioctl_process: bad csum length\n")); + kfree(cop); + kfree(jc); + return -EINVAL; + } + + csum_tcfg.next = cop->tfrm_op.tfrm_cfg; + cop->tfrm_op.tfrm_cfg = &csum_tcfg; + } + + prev_ix = first_cfg_change_ix(&oper); + if (prev_ix > oper.inlen) { + DEBUG_API(printk("cryptocop_ioctl_process: length mismatch\n")); + nooutpages = noinpages = 0; + err = -EINVAL; + goto error_cleanup; + } + DEBUG(printk("cryptocop_ioctl_process: inlen=%d, cipher_outlen=%d\n", oper.inlen, oper.cipher_outlen)); + + /* Map user pages for in and out data of the operation. */ + noinpages = (((unsigned long int)(oper.indata + prev_ix) & ~PAGE_MASK) + oper.inlen - 1 - prev_ix + ~PAGE_MASK) >> PAGE_SHIFT; + DEBUG(printk("cryptocop_ioctl_process: noinpages=%d\n", noinpages)); + inpages = kmalloc(noinpages * sizeof(struct page*), GFP_KERNEL); + if (!inpages){ + DEBUG_API(printk("cryptocop_ioctl_process: kmalloc inpages\n")); + nooutpages = noinpages = 0; + err = -ENOMEM; + goto error_cleanup; + } + if (oper.do_cipher){ + nooutpages = (((unsigned long int)oper.cipher_outdata & ~PAGE_MASK) + oper.cipher_outlen - 1 + ~PAGE_MASK) >> PAGE_SHIFT; + DEBUG(printk("cryptocop_ioctl_process: nooutpages=%d\n", nooutpages)); + outpages = kmalloc(nooutpages * sizeof(struct page*), GFP_KERNEL); + if (!outpages){ + DEBUG_API(printk("cryptocop_ioctl_process: kmalloc outpages\n")); + nooutpages = noinpages = 0; + err = -ENOMEM; + goto error_cleanup; + } + } + + /* Acquire the mm page semaphore. */ + down_read(¤t->mm->mmap_sem); + + err = get_user_pages(current, + current->mm, + (unsigned long int)(oper.indata + prev_ix), + noinpages, + 0, /* read access only for in data */ + 0, /* no force */ + inpages, + NULL); + + if (err < 0) { + up_read(¤t->mm->mmap_sem); + nooutpages = noinpages = 0; + DEBUG_API(printk("cryptocop_ioctl_process: get_user_pages indata\n")); + goto error_cleanup; + } + noinpages = err; + if (oper.do_cipher){ + err = get_user_pages(current, + current->mm, + (unsigned long int)oper.cipher_outdata, + nooutpages, + 1, /* write access for out data */ + 0, /* no force */ + outpages, + NULL); + up_read(¤t->mm->mmap_sem); + if (err < 0) { + nooutpages = 0; + DEBUG_API(printk("cryptocop_ioctl_process: get_user_pages outdata\n")); + goto error_cleanup; + } + nooutpages = err; + } else { + up_read(¤t->mm->mmap_sem); + } + + /* Add 6 to nooutpages to make room for possibly inserted buffers for storing digest and + * csum output and splits when units are (dis-)connected. */ + cop->tfrm_op.indata = kmalloc((noinpages) * sizeof(struct iovec), GFP_KERNEL); + cop->tfrm_op.outdata = kmalloc((6 + nooutpages) * sizeof(struct iovec), GFP_KERNEL); + if (!cop->tfrm_op.indata || !cop->tfrm_op.outdata) { + DEBUG_API(printk("cryptocop_ioctl_process: kmalloc iovecs\n")); + err = -ENOMEM; + goto error_cleanup; + } + + cop->tfrm_op.inlen = oper.inlen - prev_ix; + cop->tfrm_op.outlen = 0; + if (oper.do_cipher) cop->tfrm_op.outlen += oper.cipher_outlen; + if (oper.do_digest) cop->tfrm_op.outlen += digest_length; + if (oper.do_csum) cop->tfrm_op.outlen += 2; + + /* Setup the in iovecs. */ + cop->tfrm_op.incount = noinpages; + if (noinpages > 1){ + size_t tmplen = cop->tfrm_op.inlen; + + cop->tfrm_op.indata[0].iov_len = PAGE_SIZE - ((unsigned long int)(oper.indata + prev_ix) & ~PAGE_MASK); + cop->tfrm_op.indata[0].iov_base = (unsigned char*)page_address(inpages[0]) + ((unsigned long int)(oper.indata + prev_ix) & ~PAGE_MASK); + tmplen -= cop->tfrm_op.indata[0].iov_len; + for (i = 1; i<noinpages; i++){ + cop->tfrm_op.indata[i].iov_len = tmplen < PAGE_SIZE ? tmplen : PAGE_SIZE; + cop->tfrm_op.indata[i].iov_base = (unsigned char*)page_address(inpages[i]); + tmplen -= PAGE_SIZE; + } + } else { + cop->tfrm_op.indata[0].iov_len = oper.inlen - prev_ix; + cop->tfrm_op.indata[0].iov_base = (unsigned char*)page_address(inpages[0]) + ((unsigned long int)(oper.indata + prev_ix) & ~PAGE_MASK); + } + + iovlen = nooutpages + 6; + pageoffset = oper.do_cipher ? ((unsigned long int)oper.cipher_outdata & ~PAGE_MASK) : 0; + + next_ix = next_cfg_change_ix(&oper, prev_ix); + if (prev_ix == next_ix){ + DEBUG_API(printk("cryptocop_ioctl_process: length configuration broken.\n")); + err = -EINVAL; /* This should be impossible barring bugs. */ + goto error_cleanup; + } + while (prev_ix != next_ix){ + end_digest = end_csum = cipher_active = digest_active = csum_active = 0; + descs[desc_ix].cfg = NULL; + descs[desc_ix].length = next_ix - prev_ix; + + if (oper.do_cipher && (oper.cipher_start < next_ix) && (prev_ix < (oper.cipher_start + oper.cipher_len))) { + dcfgs[dcfg_ix].tid = CRYPTOCOP_IOCTL_CIPHER_TID; + dcfgs[dcfg_ix].src = cryptocop_source_dma; + cipher_active = 1; + + if (next_ix == (oper.cipher_start + oper.cipher_len)){ + cipher_done = 1; + dcfgs[dcfg_ix].last = 1; + } else { + dcfgs[dcfg_ix].last = 0; + } + dcfgs[dcfg_ix].next = descs[desc_ix].cfg; + descs[desc_ix].cfg = &dcfgs[dcfg_ix]; + ++dcfg_ix; + } + if (oper.do_digest && (oper.digest_start < next_ix) && (prev_ix < (oper.digest_start + oper.digest_len))) { + digest_active = 1; + dcfgs[dcfg_ix].tid = CRYPTOCOP_IOCTL_DIGEST_TID; + dcfgs[dcfg_ix].src = cryptocop_source_dma; + if (next_ix == (oper.digest_start + oper.digest_len)){ + assert(!digest_done); + digest_done = 1; + dcfgs[dcfg_ix].last = 1; + } else { + dcfgs[dcfg_ix].last = 0; + } + dcfgs[dcfg_ix].next = descs[desc_ix].cfg; + descs[desc_ix].cfg = &dcfgs[dcfg_ix]; + ++dcfg_ix; + } + if (oper.do_csum && (oper.csum_start < next_ix) && (prev_ix < (oper.csum_start + oper.csum_len))){ + csum_active = 1; + dcfgs[dcfg_ix].tid = CRYPTOCOP_IOCTL_CSUM_TID; + dcfgs[dcfg_ix].src = cryptocop_source_dma; + if (next_ix == (oper.csum_start + oper.csum_len)){ + csum_done = 1; + dcfgs[dcfg_ix].last = 1; + } else { + dcfgs[dcfg_ix].last = 0; + } + dcfgs[dcfg_ix].next = descs[desc_ix].cfg; + descs[desc_ix].cfg = &dcfgs[dcfg_ix]; + ++dcfg_ix; + } + if (!descs[desc_ix].cfg){ + DEBUG_API(printk("cryptocop_ioctl_process: data segment %d (%d to %d) had no active transforms\n", desc_ix, prev_ix, next_ix)); + err = -EINVAL; + goto error_cleanup; + } + descs[desc_ix].next = &(descs[desc_ix]) + 1; + ++desc_ix; + prev_ix = next_ix; + next_ix = next_cfg_change_ix(&oper, prev_ix); + } + if (desc_ix > 0){ + descs[desc_ix-1].next = NULL; + } else { + descs[0].next = NULL; + } + if (oper.do_digest) { + DEBUG(printk("cryptocop_ioctl_process: mapping %d byte digest output to iovec %d\n", digest_length, iovix)); + /* Add outdata iovec, length == <length of type of digest> */ + cop->tfrm_op.outdata[iovix].iov_base = digest_result; + cop->tfrm_op.outdata[iovix].iov_len = digest_length; + ++iovix; + } + if (oper.do_csum) { + /* Add outdata iovec, length == 2, the length of csum. */ + DEBUG(printk("cryptocop_ioctl_process: mapping 2 byte csum output to iovec %d\n", iovix)); + /* Add outdata iovec, length == <length of type of digest> */ + cop->tfrm_op.outdata[iovix].iov_base = csum_result; + cop->tfrm_op.outdata[iovix].iov_len = 2; + ++iovix; + } + if (oper.do_cipher) { + if (!map_pages_to_iovec(cop->tfrm_op.outdata, iovlen, &iovix, outpages, nooutpages, &pageix, &pageoffset, oper.cipher_outlen)){ + DEBUG_API(printk("cryptocop_ioctl_process: failed to map pages to iovec.\n")); + err = -ENOSYS; /* This should be impossible barring bugs. */ + goto error_cleanup; + } + } + DEBUG(printk("cryptocop_ioctl_process: setting cop->tfrm_op.outcount %d\n", iovix)); + cop->tfrm_op.outcount = iovix; + assert(iovix <= (nooutpages + 6)); + + cop->sid = oper.ses_id; + cop->tfrm_op.desc = &descs[0]; + + DEBUG(printk("cryptocop_ioctl_process: inserting job, cb_data=0x%p\n", cop->cb_data)); + + if ((err = cryptocop_job_queue_insert_user_job(cop)) != 0) { + DEBUG_API(printk("cryptocop_ioctl_process: insert job %d\n", err)); + err = -EINVAL; + goto error_cleanup; + } + + DEBUG(printk("cryptocop_ioctl_process: begin wait for result\n")); + + wait_event(cryptocop_ioc_process_wq, (jc->processed != 0)); + DEBUG(printk("cryptocop_ioctl_process: end wait for result\n")); + if (!jc->processed){ + printk(KERN_WARNING "cryptocop_ioctl_process: job not processed at completion\n"); + err = -EIO; + goto error_cleanup; + } + + /* Job process done. Cipher output should already be correct in job so no post processing of outdata. */ + DEBUG(printk("cryptocop_ioctl_process: operation_status = %d\n", cop->operation_status)); + if (cop->operation_status == 0){ + if (oper.do_digest){ + DEBUG(printk("cryptocop_ioctl_process: copy %d bytes digest to user\n", digest_length)); + err = copy_to_user((unsigned char*)crp_oper + offsetof(struct strcop_crypto_op, digest), digest_result, digest_length); + if (0 != err){ + DEBUG_API(printk("cryptocop_ioctl_process: copy_to_user, digest length %d, err %d\n", digest_length, err)); + err = -EFAULT; + goto error_cleanup; + } + } + if (oper.do_csum){ + DEBUG(printk("cryptocop_ioctl_process: copy 2 bytes checksum to user\n")); + err = copy_to_user((unsigned char*)crp_oper + offsetof(struct strcop_crypto_op, csum), csum_result, 2); + if (0 != err){ + DEBUG_API(printk("cryptocop_ioctl_process: copy_to_user, csum, err %d\n", err)); + err = -EFAULT; + goto error_cleanup; + } + } + err = 0; + } else { + DEBUG(printk("cryptocop_ioctl_process: returning err = operation_status = %d\n", cop->operation_status)); + err = cop->operation_status; + } + + error_cleanup: + /* Release page caches. */ + for (i = 0; i < noinpages; i++){ + put_page(inpages[i]); + } + for (i = 0; i < nooutpages; i++){ + int spdl_err; + /* Mark output pages dirty. */ + spdl_err = set_page_dirty_lock(outpages[i]); + DEBUG(if (spdl_err)printk("cryptocop_ioctl_process: set_page_dirty_lock returned %d\n", spdl_err)); + } + for (i = 0; i < nooutpages; i++){ + put_page(outpages[i]); + } + + if (digest_result) kfree(digest_result); + if (inpages) kfree(inpages); + if (outpages) kfree(outpages); + if (cop){ + if (cop->tfrm_op.indata) kfree(cop->tfrm_op.indata); + if (cop->tfrm_op.outdata) kfree(cop->tfrm_op.outdata); + kfree(cop); + } + if (jc) kfree(jc); + + DEBUG(print_lock_status()); + + return err; +} + + +static int cryptocop_ioctl_create_session(struct inode *inode, struct file *filp, unsigned int cmd, unsigned long arg) +{ + cryptocop_session_id sid; + int err; + struct cryptocop_private *dev; + struct strcop_session_op *sess_op = (struct strcop_session_op *)arg; + struct strcop_session_op sop; + struct cryptocop_transform_init *tis = NULL; + struct cryptocop_transform_init ti_cipher = {0}; + struct cryptocop_transform_init ti_digest = {0}; + struct cryptocop_transform_init ti_csum = {0}; + + if (!access_ok(VERIFY_WRITE, sess_op, sizeof(struct strcop_session_op))) + return -EFAULT; + err = copy_from_user(&sop, sess_op, sizeof(struct strcop_session_op)); + if (err) return -EFAULT; + if (sop.cipher != cryptocop_cipher_none) { + if (!access_ok(VERIFY_READ, sop.key, sop.keylen)) return -EFAULT; + } + DEBUG(printk("cryptocop_ioctl_create_session, sess_op:\n")); + + DEBUG(printk("\tcipher:%d\n" + "\tcipher_mode:%d\n" + "\tdigest:%d\n" + "\tcsum:%d\n", + (int)sop.cipher, + (int)sop.cmode, + (int)sop.digest, + (int)sop.csum)); + + if (sop.cipher != cryptocop_cipher_none){ + /* Init the cipher. */ + switch (sop.cipher){ + case cryptocop_cipher_des: + ti_cipher.alg = cryptocop_alg_des; + break; + case cryptocop_cipher_3des: + ti_cipher.alg = cryptocop_alg_3des; + break; + case cryptocop_cipher_aes: + ti_cipher.alg = cryptocop_alg_aes; + break; + default: + DEBUG_API(printk("create session, bad cipher algorithm %d\n", sop.cipher)); + return -EINVAL; + }; + DEBUG(printk("setting cipher transform %d\n", ti_cipher.alg)); + copy_from_user(ti_cipher.key, sop.key, sop.keylen/8); + ti_cipher.keylen = sop.keylen; + switch (sop.cmode){ + case cryptocop_cipher_mode_cbc: + case cryptocop_cipher_mode_ecb: + ti_cipher.cipher_mode = sop.cmode; + break; + default: + DEBUG_API(printk("create session, bad cipher mode %d\n", sop.cmode)); + return -EINVAL; + } + DEBUG(printk("cryptocop_ioctl_create_session: setting CBC mode %d\n", ti_cipher.cipher_mode)); + switch (sop.des3_mode){ + case cryptocop_3des_eee: + case cryptocop_3des_eed: + case cryptocop_3des_ede: + case cryptocop_3des_edd: + case cryptocop_3des_dee: + case cryptocop_3des_ded: + case cryptocop_3des_dde: + case cryptocop_3des_ddd: + ti_cipher.tdes_mode = sop.des3_mode; + break; + default: + DEBUG_API(printk("create session, bad 3DES mode %d\n", sop.des3_mode)); + return -EINVAL; + } + ti_cipher.tid = CRYPTOCOP_IOCTL_CIPHER_TID; + ti_cipher.next = tis; + tis = &ti_cipher; + } /* if (sop.cipher != cryptocop_cipher_none) */ + if (sop.digest != cryptocop_digest_none){ + DEBUG(printk("setting digest transform\n")); + switch (sop.digest){ + case cryptocop_digest_md5: + ti_digest.alg = cryptocop_alg_md5; + break; + case cryptocop_digest_sha1: + ti_digest.alg = cryptocop_alg_sha1; + break; + default: + DEBUG_API(printk("create session, bad digest algorithm %d\n", sop.digest)); + return -EINVAL; + } + ti_digest.tid = CRYPTOCOP_IOCTL_DIGEST_TID; + ti_digest.next = tis; + tis = &ti_digest; + } /* if (sop.digest != cryptocop_digest_none) */ + if (sop.csum != cryptocop_csum_none){ + DEBUG(printk("setting csum transform\n")); + switch (sop.csum){ + case cryptocop_csum_le: + case cryptocop_csum_be: + ti_csum.csum_mode = sop.csum; + break; + default: + DEBUG_API(printk("create session, bad checksum algorithm %d\n", sop.csum)); + return -EINVAL; + } + ti_csum.alg = cryptocop_alg_csum; + ti_csum.tid = CRYPTOCOP_IOCTL_CSUM_TID; + ti_csum.next = tis; + tis = &ti_csum; + } /* (sop.csum != cryptocop_csum_none) */ + dev = kmalloc(sizeof(struct cryptocop_private), GFP_KERNEL); + if (!dev){ + DEBUG_API(printk("create session, alloc dev\n")); + return -ENOMEM; + } + + err = cryptocop_new_session(&sid, tis, GFP_KERNEL); + DEBUG({ if (err) printk("create session, cryptocop_new_session %d\n", err);}); + + if (err) { + kfree(dev); + return err; + } + sess_op->ses_id = sid; + dev->sid = sid; + dev->next = filp->private_data; + filp->private_data = dev; + + return 0; +} + +static int cryptocop_ioctl(struct inode *inode, struct file *filp, unsigned int cmd, unsigned long arg) +{ + int err = 0; + if (_IOC_TYPE(cmd) != ETRAXCRYPTOCOP_IOCTYPE) { + DEBUG_API(printk("cryptocop_ioctl: wrong type\n")); + return -ENOTTY; + } + if (_IOC_NR(cmd) > CRYPTOCOP_IO_MAXNR){ + return -ENOTTY; + } + /* Access check of the argument. Some commands, e.g. create session and process op, + needs additional checks. Those are handled in the command handling functions. */ + if (_IOC_DIR(cmd) & _IOC_READ) + err = !access_ok(VERIFY_WRITE, (void *)arg, _IOC_SIZE(cmd)); + else if (_IOC_DIR(cmd) & _IOC_WRITE) + err = !access_ok(VERIFY_READ, (void *)arg, _IOC_SIZE(cmd)); + if (err) return -EFAULT; + + switch (cmd) { + case CRYPTOCOP_IO_CREATE_SESSION: + return cryptocop_ioctl_create_session(inode, filp, cmd, arg); + case CRYPTOCOP_IO_CLOSE_SESSION: + return cryptocop_ioctl_close_session(inode, filp, cmd, arg); + case CRYPTOCOP_IO_PROCESS_OP: + return cryptocop_ioctl_process(inode, filp, cmd, arg); + default: + DEBUG_API(printk("cryptocop_ioctl: unknown command\n")); + return -ENOTTY; + } + return 0; +} + + +#ifdef LDEBUG +static void print_dma_descriptors(struct cryptocop_int_operation *iop) +{ + struct cryptocop_dma_desc *cdesc_out = iop->cdesc_out; + struct cryptocop_dma_desc *cdesc_in = iop->cdesc_in; + int i; + + printk("print_dma_descriptors start\n"); + + printk("iop:\n"); + printk("\tsid: 0x%lld\n", iop->sid); + + printk("\tcdesc_out: 0x%p\n", iop->cdesc_out); + printk("\tcdesc_in: 0x%p\n", iop->cdesc_in); + printk("\tddesc_out: 0x%p\n", iop->ddesc_out); + printk("\tddesc_in: 0x%p\n", iop->ddesc_in); + + printk("\niop->ctx_out: 0x%p phys: 0x%p\n", &iop->ctx_out, (char*)virt_to_phys(&iop->ctx_out)); + printk("\tnext: 0x%p\n" + "\tsaved_data: 0x%p\n" + "\tsaved_data_buf: 0x%p\n", + iop->ctx_out.next, + iop->ctx_out.saved_data, + iop->ctx_out.saved_data_buf); + + printk("\niop->ctx_in: 0x%p phys: 0x%p\n", &iop->ctx_in, (char*)virt_to_phys(&iop->ctx_in)); + printk("\tnext: 0x%p\n" + "\tsaved_data: 0x%p\n" + "\tsaved_data_buf: 0x%p\n", + iop->ctx_in.next, + iop->ctx_in.saved_data, + iop->ctx_in.saved_data_buf); + + i = 0; + while (cdesc_out) { + dma_descr_data *td; + printk("cdesc_out %d, desc=0x%p\n", i, cdesc_out->dma_descr); + printk("\n\tvirt_to_phys(desc): 0x%p\n", (char*)virt_to_phys(cdesc_out->dma_descr)); + td = cdesc_out->dma_descr; + printk("\n\tbuf: 0x%p\n" + "\tafter: 0x%p\n" + "\tmd: 0x%04x\n" + "\tnext: 0x%p\n", + td->buf, + td->after, + td->md, + td->next); + printk("flags:\n" + "\twait:\t%d\n" + "\teol:\t%d\n" + "\touteop:\t%d\n" + "\tineop:\t%d\n" + "\tintr:\t%d\n", + td->wait, + td->eol, + td->out_eop, + td->in_eop, + td->intr); + cdesc_out = cdesc_out->next; + i++; + } + i = 0; + while (cdesc_in) { + dma_descr_data *td; + printk("cdesc_in %d, desc=0x%p\n", i, cdesc_in->dma_descr); + printk("\n\tvirt_to_phys(desc): 0x%p\n", (char*)virt_to_phys(cdesc_in->dma_descr)); + td = cdesc_in->dma_descr; + printk("\n\tbuf: 0x%p\n" + "\tafter: 0x%p\n" + "\tmd: 0x%04x\n" + "\tnext: 0x%p\n", + td->buf, + td->after, + td->md, + td->next); + printk("flags:\n" + "\twait:\t%d\n" + "\teol:\t%d\n" + "\touteop:\t%d\n" + "\tineop:\t%d\n" + "\tintr:\t%d\n", + td->wait, + td->eol, + td->out_eop, + td->in_eop, + td->intr); + cdesc_in = cdesc_in->next; + i++; + } + + printk("print_dma_descriptors end\n"); +} + + +static void print_strcop_crypto_op(struct strcop_crypto_op *cop) +{ + printk("print_strcop_crypto_op, 0x%p\n", cop); + + /* Indata. */ + printk("indata=0x%p\n" + "inlen=%d\n" + "do_cipher=%d\n" + "decrypt=%d\n" + "cipher_explicit=%d\n" + "cipher_start=%d\n" + "cipher_len=%d\n" + "outdata=0x%p\n" + "outlen=%d\n", + cop->indata, + cop->inlen, + cop->do_cipher, + cop->decrypt, + cop->cipher_explicit, + cop->cipher_start, + cop->cipher_len, + cop->cipher_outdata, + cop->cipher_outlen); + + printk("do_digest=%d\n" + "digest_start=%d\n" + "digest_len=%d\n", + cop->do_digest, + cop->digest_start, + cop->digest_len); + + printk("do_csum=%d\n" + "csum_start=%d\n" + "csum_len=%d\n", + cop->do_csum, + cop->csum_start, + cop->csum_len); +} + +static void print_cryptocop_operation(struct cryptocop_operation *cop) +{ + struct cryptocop_desc *d; + struct cryptocop_tfrm_cfg *tc; + struct cryptocop_desc_cfg *dc; + int i; + + printk("print_cryptocop_operation, cop=0x%p\n\n", cop); + printk("sid: %lld\n", cop->sid); + printk("operation_status=%d\n" + "use_dmalists=%d\n" + "in_interrupt=%d\n" + "fast_callback=%d\n", + cop->operation_status, + cop->use_dmalists, + cop->in_interrupt, + cop->fast_callback); + + if (cop->use_dmalists){ + print_user_dma_lists(&cop->list_op); + } else { + printk("cop->tfrm_op\n" + "tfrm_cfg=0x%p\n" + "desc=0x%p\n" + "indata=0x%p\n" + "incount=%d\n" + "inlen=%d\n" + "outdata=0x%p\n" + "outcount=%d\n" + "outlen=%d\n\n", + cop->tfrm_op.tfrm_cfg, + cop->tfrm_op.desc, + cop->tfrm_op.indata, + cop->tfrm_op.incount, + cop->tfrm_op.inlen, + cop->tfrm_op.outdata, + cop->tfrm_op.outcount, + cop->tfrm_op.outlen); + + tc = cop->tfrm_op.tfrm_cfg; + while (tc){ + printk("tfrm_cfg, 0x%p\n" + "tid=%d\n" + "flags=%d\n" + "inject_ix=%d\n" + "next=0x%p\n", + tc, + tc->tid, + tc->flags, + tc->inject_ix, + tc->next); + tc = tc->next; + } + d = cop->tfrm_op.desc; + while (d){ + printk("\n======================desc, 0x%p\n" + "length=%d\n" + "cfg=0x%p\n" + "next=0x%p\n", + d, + d->length, + d->cfg, + d->next); + dc = d->cfg; + while (dc){ + printk("=========desc_cfg, 0x%p\n" + "tid=%d\n" + "src=%d\n" + "last=%d\n" + "next=0x%p\n", + dc, + dc->tid, + dc->src, + dc->last, + dc->next); + dc = dc->next; + } + d = d->next; + } + printk("\n====iniov\n"); + for (i = 0; i < cop->tfrm_op.incount; i++){ + printk("indata[%d]\n" + "base=0x%p\n" + "len=%d\n", + i, + cop->tfrm_op.indata[i].iov_base, + cop->tfrm_op.indata[i].iov_len); + } + printk("\n====outiov\n"); + for (i = 0; i < cop->tfrm_op.outcount; i++){ + printk("outdata[%d]\n" + "base=0x%p\n" + "len=%d\n", + i, + cop->tfrm_op.outdata[i].iov_base, + cop->tfrm_op.outdata[i].iov_len); + } + } + printk("------------end print_cryptocop_operation\n"); +} + + +static void print_user_dma_lists(struct cryptocop_dma_list_operation *dma_op) +{ + dma_descr_data *dd; + int i; + + printk("print_user_dma_lists, dma_op=0x%p\n", dma_op); + + printk("out_data_buf = 0x%p, phys_to_virt(out_data_buf) = 0x%p\n", dma_op->out_data_buf, phys_to_virt((unsigned long int)dma_op->out_data_buf)); + printk("in_data_buf = 0x%p, phys_to_virt(in_data_buf) = 0x%p\n", dma_op->in_data_buf, phys_to_virt((unsigned long int)dma_op->in_data_buf)); + + printk("##############outlist\n"); + dd = phys_to_virt((unsigned long int)dma_op->outlist); + i = 0; + while (dd != NULL) { + printk("#%d phys_to_virt(desc) 0x%p\n", i, dd); + printk("\n\tbuf: 0x%p\n" + "\tafter: 0x%p\n" + "\tmd: 0x%04x\n" + "\tnext: 0x%p\n", + dd->buf, + dd->after, + dd->md, + dd->next); + printk("flags:\n" + "\twait:\t%d\n" + "\teol:\t%d\n" + "\touteop:\t%d\n" + "\tineop:\t%d\n" + "\tintr:\t%d\n", + dd->wait, + dd->eol, + dd->out_eop, + dd->in_eop, + dd->intr); + if (dd->eol) + dd = NULL; + else + dd = phys_to_virt((unsigned long int)dd->next); + ++i; + } + + printk("##############inlist\n"); + dd = phys_to_virt((unsigned long int)dma_op->inlist); + i = 0; + while (dd != NULL) { + printk("#%d phys_to_virt(desc) 0x%p\n", i, dd); + printk("\n\tbuf: 0x%p\n" + "\tafter: 0x%p\n" + "\tmd: 0x%04x\n" + "\tnext: 0x%p\n", + dd->buf, + dd->after, + dd->md, + dd->next); + printk("flags:\n" + "\twait:\t%d\n" + "\teol:\t%d\n" + "\touteop:\t%d\n" + "\tineop:\t%d\n" + "\tintr:\t%d\n", + dd->wait, + dd->eol, + dd->out_eop, + dd->in_eop, + dd->intr); + if (dd->eol) + dd = NULL; + else + dd = phys_to_virt((unsigned long int)dd->next); + ++i; + } +} + + +static void print_lock_status(void) +{ + printk("**********************print_lock_status\n"); + printk("cryptocop_completed_jobs_lock %d\n", spin_is_locked(&cryptocop_completed_jobs_lock)); + printk("cryptocop_job_queue_lock %d\n", spin_is_locked(&cryptocop_job_queue_lock)); + printk("descr_pool_lock %d\n", spin_is_locked(&descr_pool_lock)); + printk("cryptocop_sessions_lock %d\n", spin_is_locked(cryptocop_sessions_lock)); + printk("running_job_lock %d\n", spin_is_locked(running_job_lock)); + printk("cryptocop_process_lock %d\n", spin_is_locked(cryptocop_process_lock)); +} +#endif /* LDEBUG */ + + +static const char cryptocop_name[] = "ETRAX FS stream co-processor"; + +static int init_stream_coprocessor(void) +{ + int err; + int i; + static int initialized = 0; + + if (initialized) + return 0; + + initialized = 1; + + printk("ETRAX FS stream co-processor driver v0.01, (c) 2003 Axis Communications AB\n"); + + err = register_chrdev(CRYPTOCOP_MAJOR, cryptocop_name, &cryptocop_fops); + if (err < 0) { + printk(KERN_ERR "stream co-processor: could not get major number.\n"); + return err; + } + + err = init_cryptocop(); + if (err) { + (void)unregister_chrdev(CRYPTOCOP_MAJOR, cryptocop_name); + return err; + } + err = cryptocop_job_queue_init(); + if (err) { + release_cryptocop(); + (void)unregister_chrdev(CRYPTOCOP_MAJOR, cryptocop_name); + return err; + } + /* Init the descriptor pool. */ + for (i = 0; i < CRYPTOCOP_DESCRIPTOR_POOL_SIZE - 1; i++) { + descr_pool[i].from_pool = 1; + descr_pool[i].next = &descr_pool[i + 1]; + } + descr_pool[i].from_pool = 1; + descr_pool[i].next = NULL; + descr_pool_free_list = &descr_pool[0]; + descr_pool_no_free = CRYPTOCOP_DESCRIPTOR_POOL_SIZE; + + spin_lock_init(&cryptocop_completed_jobs_lock); + spin_lock_init(&cryptocop_job_queue_lock); + spin_lock_init(&descr_pool_lock); + spin_lock_init(&cryptocop_sessions_lock); + spin_lock_init(&running_job_lock); + spin_lock_init(&cryptocop_process_lock); + + cryptocop_sessions = NULL; + next_sid = 1; + + cryptocop_running_job = NULL; + + printk("stream co-processor: init done.\n"); + return 0; +} + +static void __exit exit_stream_coprocessor(void) +{ + release_cryptocop(); + cryptocop_job_queue_close(); +} + +module_init(init_stream_coprocessor); +module_exit(exit_stream_coprocessor); + diff --git a/arch/cris/arch-v32/drivers/gpio.c b/arch/cris/arch-v32/drivers/gpio.c new file mode 100644 index 000000000000..a551237dcb5e --- /dev/null +++ b/arch/cris/arch-v32/drivers/gpio.c @@ -0,0 +1,766 @@ +/* $Id: gpio.c,v 1.16 2005/06/19 17:06:49 starvik Exp $ + * + * ETRAX CRISv32 general port I/O device + * + * Copyright (c) 1999, 2000, 2001, 2002, 2003 Axis Communications AB + * + * Authors: Bjorn Wesen (initial version) + * Ola Knutsson (LED handling) + * Johan Adolfsson (read/set directions, write, port G, + * port to ETRAX FS. + * + * $Log: gpio.c,v $ + * Revision 1.16 2005/06/19 17:06:49 starvik + * Merge of Linux 2.6.12. + * + * Revision 1.15 2005/05/25 08:22:20 starvik + * Changed GPIO port order to fit packages/devices/axis-2.4. + * + * Revision 1.14 2005/04/24 18:35:08 starvik + * Updated with final register headers. + * + * Revision 1.13 2005/03/15 15:43:00 starvik + * dev_id needs to be supplied for shared IRQs. + * + * Revision 1.12 2005/03/10 17:12:00 starvik + * Protect alarm list with spinlock. + * + * Revision 1.11 2005/01/05 06:08:59 starvik + * No need to do local_irq_disable after local_irq_save. + * + * Revision 1.10 2004/11/19 08:38:31 starvik + * Removed old crap. + * + * Revision 1.9 2004/05/14 07:58:02 starvik + * Merge of changes from 2.4 + * + * Revision 1.8 2003/09/11 07:29:50 starvik + * Merge of Linux 2.6.0-test5 + * + * Revision 1.7 2003/07/10 13:25:46 starvik + * Compiles for 2.5.74 + * Lindented ethernet.c + * + * Revision 1.6 2003/07/04 08:27:46 starvik + * Merge of Linux 2.5.74 + * + * Revision 1.5 2003/06/10 08:26:37 johana + * Etrax -> ETRAX CRISv32 + * + * Revision 1.4 2003/06/05 14:22:48 johana + * Initialise some_alarms. + * + * Revision 1.3 2003/06/05 10:15:46 johana + * New INTR_VECT macros. + * Enable interrupts in global config. + * + * Revision 1.2 2003/06/03 15:52:50 johana + * Initial CRIS v32 version. + * + * Revision 1.1 2003/06/03 08:53:15 johana + * Copy of os/lx25/arch/cris/arch-v10/drivers/gpio.c version 1.7. + * + */ + +#include <linux/config.h> + +#include <linux/module.h> +#include <linux/sched.h> +#include <linux/slab.h> +#include <linux/ioport.h> +#include <linux/errno.h> +#include <linux/kernel.h> +#include <linux/fs.h> +#include <linux/string.h> +#include <linux/poll.h> +#include <linux/init.h> +#include <linux/interrupt.h> +#include <linux/spinlock.h> + +#include <asm/etraxgpio.h> +#include <asm/arch/hwregs/reg_map.h> +#include <asm/arch/hwregs/reg_rdwr.h> +#include <asm/arch/hwregs/gio_defs.h> +#include <asm/arch/hwregs/intr_vect_defs.h> +#include <asm/io.h> +#include <asm/system.h> +#include <asm/irq.h> + +/* The following gio ports on ETRAX FS is available: + * pa 8 bits, supports interrupts off, hi, low, set, posedge, negedge anyedge + * pb 18 bits + * pc 18 bits + * pd 18 bits + * pe 18 bits + * each port has a rw_px_dout, r_px_din and rw_px_oe register. + */ + +#define GPIO_MAJOR 120 /* experimental MAJOR number */ + +#define D(x) + +#if 0 +static int dp_cnt; +#define DP(x) do { dp_cnt++; if (dp_cnt % 1000 == 0) x; }while(0) +#else +#define DP(x) +#endif + +static char gpio_name[] = "etrax gpio"; + +#if 0 +static wait_queue_head_t *gpio_wq; +#endif + +static int gpio_ioctl(struct inode *inode, struct file *file, + unsigned int cmd, unsigned long arg); +static ssize_t gpio_write(struct file * file, const char * buf, size_t count, + loff_t *off); +static int gpio_open(struct inode *inode, struct file *filp); +static int gpio_release(struct inode *inode, struct file *filp); +static unsigned int gpio_poll(struct file *filp, struct poll_table_struct *wait); + +/* private data per open() of this driver */ + +struct gpio_private { + struct gpio_private *next; + /* The IO_CFG_WRITE_MODE_VALUE only support 8 bits: */ + unsigned char clk_mask; + unsigned char data_mask; + unsigned char write_msb; + unsigned char pad1; + /* These fields are generic */ + unsigned long highalarm, lowalarm; + wait_queue_head_t alarm_wq; + int minor; +}; + +/* linked list of alarms to check for */ + +static struct gpio_private *alarmlist = 0; + +static int gpio_some_alarms = 0; /* Set if someone uses alarm */ +static unsigned long gpio_pa_high_alarms = 0; +static unsigned long gpio_pa_low_alarms = 0; + +static DEFINE_SPINLOCK(alarm_lock); + +#define NUM_PORTS (GPIO_MINOR_LAST+1) +#define GIO_REG_RD_ADDR(reg) (volatile unsigned long*) (regi_gio + REG_RD_ADDR_gio_##reg ) +#define GIO_REG_WR_ADDR(reg) (volatile unsigned long*) (regi_gio + REG_RD_ADDR_gio_##reg ) +unsigned long led_dummy; + +static volatile unsigned long *data_out[NUM_PORTS] = { + GIO_REG_WR_ADDR(rw_pa_dout), + GIO_REG_WR_ADDR(rw_pb_dout), + &led_dummy, + GIO_REG_WR_ADDR(rw_pc_dout), + GIO_REG_WR_ADDR(rw_pd_dout), + GIO_REG_WR_ADDR(rw_pe_dout), +}; + +static volatile unsigned long *data_in[NUM_PORTS] = { + GIO_REG_RD_ADDR(r_pa_din), + GIO_REG_RD_ADDR(r_pb_din), + &led_dummy, + GIO_REG_RD_ADDR(r_pc_din), + GIO_REG_RD_ADDR(r_pd_din), + GIO_REG_RD_ADDR(r_pe_din), +}; + +static unsigned long changeable_dir[NUM_PORTS] = { + CONFIG_ETRAX_PA_CHANGEABLE_DIR, + CONFIG_ETRAX_PB_CHANGEABLE_DIR, + 0, + CONFIG_ETRAX_PC_CHANGEABLE_DIR, + CONFIG_ETRAX_PD_CHANGEABLE_DIR, + CONFIG_ETRAX_PE_CHANGEABLE_DIR, +}; + +static unsigned long changeable_bits[NUM_PORTS] = { + CONFIG_ETRAX_PA_CHANGEABLE_BITS, + CONFIG_ETRAX_PB_CHANGEABLE_BITS, + 0, + CONFIG_ETRAX_PC_CHANGEABLE_BITS, + CONFIG_ETRAX_PD_CHANGEABLE_BITS, + CONFIG_ETRAX_PE_CHANGEABLE_BITS, +}; + +static volatile unsigned long *dir_oe[NUM_PORTS] = { + GIO_REG_WR_ADDR(rw_pa_oe), + GIO_REG_WR_ADDR(rw_pb_oe), + &led_dummy, + GIO_REG_WR_ADDR(rw_pc_oe), + GIO_REG_WR_ADDR(rw_pd_oe), + GIO_REG_WR_ADDR(rw_pe_oe), +}; + + + +static unsigned int +gpio_poll(struct file *file, + poll_table *wait) +{ + unsigned int mask = 0; + struct gpio_private *priv = (struct gpio_private *)file->private_data; + unsigned long data; + poll_wait(file, &priv->alarm_wq, wait); + if (priv->minor == GPIO_MINOR_A) { + reg_gio_rw_intr_cfg intr_cfg; + unsigned long tmp; + unsigned long flags; + + local_irq_save(flags); + data = REG_TYPE_CONV(unsigned long, reg_gio_r_pa_din, REG_RD(gio, regi_gio, r_pa_din)); + /* PA has support for interrupt + * lets activate high for those low and with highalarm set + */ + intr_cfg = REG_RD(gio, regi_gio, rw_intr_cfg); + + tmp = ~data & priv->highalarm & 0xFF; + if (tmp & (1 << 0)) { + intr_cfg.pa0 = regk_gio_hi; + } + if (tmp & (1 << 1)) { + intr_cfg.pa1 = regk_gio_hi; + } + if (tmp & (1 << 2)) { + intr_cfg.pa2 = regk_gio_hi; + } + if (tmp & (1 << 3)) { + intr_cfg.pa3 = regk_gio_hi; + } + if (tmp & (1 << 4)) { + intr_cfg.pa4 = regk_gio_hi; + } + if (tmp & (1 << 5)) { + intr_cfg.pa5 = regk_gio_hi; + } + if (tmp & (1 << 6)) { + intr_cfg.pa6 = regk_gio_hi; + } + if (tmp & (1 << 7)) { + intr_cfg.pa7 = regk_gio_hi; + } + /* + * lets activate low for those high and with lowalarm set + */ + tmp = data & priv->lowalarm & 0xFF; + if (tmp & (1 << 0)) { + intr_cfg.pa0 = regk_gio_lo; + } + if (tmp & (1 << 1)) { + intr_cfg.pa1 = regk_gio_lo; + } + if (tmp & (1 << 2)) { + intr_cfg.pa2 = regk_gio_lo; + } + if (tmp & (1 << 3)) { + intr_cfg.pa3 = regk_gio_lo; + } + if (tmp & (1 << 4)) { + intr_cfg.pa4 = regk_gio_lo; + } + if (tmp & (1 << 5)) { + intr_cfg.pa5 = regk_gio_lo; + } + if (tmp & (1 << 6)) { + intr_cfg.pa6 = regk_gio_lo; + } + if (tmp & (1 << 7)) { + intr_cfg.pa7 = regk_gio_lo; + } + + REG_WR(gio, regi_gio, rw_intr_cfg, intr_cfg); + local_irq_restore(flags); + } else if (priv->minor <= GPIO_MINOR_E) + data = *data_in[priv->minor]; + else + return 0; + + if ((data & priv->highalarm) || + (~data & priv->lowalarm)) { + mask = POLLIN|POLLRDNORM; + } + + DP(printk("gpio_poll ready: mask 0x%08X\n", mask)); + return mask; +} + +int etrax_gpio_wake_up_check(void) +{ + struct gpio_private *priv = alarmlist; + unsigned long data = 0; + int ret = 0; + while (priv) { + data = *data_in[priv->minor]; + if ((data & priv->highalarm) || + (~data & priv->lowalarm)) { + DP(printk("etrax_gpio_wake_up_check %i\n",priv->minor)); + wake_up_interruptible(&priv->alarm_wq); + ret = 1; + } + priv = priv->next; + } + return ret; +} + +static irqreturn_t +gpio_poll_timer_interrupt(int irq, void *dev_id, struct pt_regs *regs) +{ + if (gpio_some_alarms) { + return IRQ_RETVAL(etrax_gpio_wake_up_check()); + } + return IRQ_NONE; +} + +static irqreturn_t +gpio_pa_interrupt(int irq, void *dev_id, struct pt_regs *regs) +{ + reg_gio_rw_intr_mask intr_mask; + reg_gio_r_masked_intr masked_intr; + reg_gio_rw_ack_intr ack_intr; + unsigned long tmp; + unsigned long tmp2; + + /* Find what PA interrupts are active */ + masked_intr = REG_RD(gio, regi_gio, r_masked_intr); + tmp = REG_TYPE_CONV(unsigned long, reg_gio_r_masked_intr, masked_intr); + + /* Find those that we have enabled */ + spin_lock(&alarm_lock); + tmp &= (gpio_pa_high_alarms | gpio_pa_low_alarms); + spin_unlock(&alarm_lock); + + /* Ack them */ + ack_intr = REG_TYPE_CONV(reg_gio_rw_ack_intr, unsigned long, tmp); + REG_WR(gio, regi_gio, rw_ack_intr, ack_intr); + + /* Disable those interrupts.. */ + intr_mask = REG_RD(gio, regi_gio, rw_intr_mask); + tmp2 = REG_TYPE_CONV(unsigned long, reg_gio_rw_intr_mask, intr_mask); + tmp2 &= ~tmp; + intr_mask = REG_TYPE_CONV(reg_gio_rw_intr_mask, unsigned long, tmp2); + REG_WR(gio, regi_gio, rw_intr_mask, intr_mask); + + if (gpio_some_alarms) { + return IRQ_RETVAL(etrax_gpio_wake_up_check()); + } + return IRQ_NONE; +} + + +static ssize_t gpio_write(struct file * file, const char * buf, size_t count, + loff_t *off) +{ + struct gpio_private *priv = (struct gpio_private *)file->private_data; + unsigned char data, clk_mask, data_mask, write_msb; + unsigned long flags; + unsigned long shadow; + volatile unsigned long *port; + ssize_t retval = count; + /* Only bits 0-7 may be used for write operations but allow all + devices except leds... */ + if (priv->minor == GPIO_MINOR_LEDS) { + return -EFAULT; + } + + if (!access_ok(VERIFY_READ, buf, count)) { + return -EFAULT; + } + clk_mask = priv->clk_mask; + data_mask = priv->data_mask; + /* It must have been configured using the IO_CFG_WRITE_MODE */ + /* Perhaps a better error code? */ + if (clk_mask == 0 || data_mask == 0) { + return -EPERM; + } + write_msb = priv->write_msb; + D(printk("gpio_write: %lu to data 0x%02X clk 0x%02X msb: %i\n",count, data_mask, clk_mask, write_msb)); + port = data_out[priv->minor]; + + while (count--) { + int i; + data = *buf++; + if (priv->write_msb) { + for (i = 7; i >= 0;i--) { + local_irq_save(flags); + shadow = *port; + *port = shadow &= ~clk_mask; + if (data & 1<<i) + *port = shadow |= data_mask; + else + *port = shadow &= ~data_mask; + /* For FPGA: min 5.0ns (DCC) before CCLK high */ + *port = shadow |= clk_mask; + local_irq_restore(flags); + } + } else { + for (i = 0; i <= 7;i++) { + local_irq_save(flags); + shadow = *port; + *port = shadow &= ~clk_mask; + if (data & 1<<i) + *port = shadow |= data_mask; + else + *port = shadow &= ~data_mask; + /* For FPGA: min 5.0ns (DCC) before CCLK high */ + *port = shadow |= clk_mask; + local_irq_restore(flags); + } + } + } + return retval; +} + + + +static int +gpio_open(struct inode *inode, struct file *filp) +{ + struct gpio_private *priv; + int p = MINOR(inode->i_rdev); + + if (p > GPIO_MINOR_LAST) + return -EINVAL; + + priv = (struct gpio_private *)kmalloc(sizeof(struct gpio_private), + GFP_KERNEL); + + if (!priv) + return -ENOMEM; + + priv->minor = p; + + /* initialize the io/alarm struct and link it into our alarmlist */ + + priv->next = alarmlist; + alarmlist = priv; + priv->clk_mask = 0; + priv->data_mask = 0; + priv->highalarm = 0; + priv->lowalarm = 0; + init_waitqueue_head(&priv->alarm_wq); + + filp->private_data = (void *)priv; + + return 0; +} + +static int +gpio_release(struct inode *inode, struct file *filp) +{ + struct gpio_private *p = alarmlist; + struct gpio_private *todel = (struct gpio_private *)filp->private_data; + /* local copies while updating them: */ + unsigned long a_high, a_low; + unsigned long some_alarms; + + /* unlink from alarmlist and free the private structure */ + + if (p == todel) { + alarmlist = todel->next; + } else { + while (p->next != todel) + p = p->next; + p->next = todel->next; + } + + kfree(todel); + /* Check if there are still any alarms set */ + p = alarmlist; + some_alarms = 0; + a_high = 0; + a_low = 0; + while (p) { + if (p->minor == GPIO_MINOR_A) { + a_high |= p->highalarm; + a_low |= p->lowalarm; + } + + if (p->highalarm | p->lowalarm) { + some_alarms = 1; + } + p = p->next; + } + + spin_lock(&alarm_lock); + gpio_some_alarms = some_alarms; + gpio_pa_high_alarms = a_high; + gpio_pa_low_alarms = a_low; + spin_unlock(&alarm_lock); + + return 0; +} + +/* Main device API. ioctl's to read/set/clear bits, as well as to + * set alarms to wait for using a subsequent select(). + */ + +unsigned long inline setget_input(struct gpio_private *priv, unsigned long arg) +{ + /* Set direction 0=unchanged 1=input, + * return mask with 1=input + */ + unsigned long flags; + unsigned long dir_shadow; + + local_irq_save(flags); + dir_shadow = *dir_oe[priv->minor]; + dir_shadow &= ~(arg & changeable_dir[priv->minor]); + *dir_oe[priv->minor] = dir_shadow; + local_irq_restore(flags); + + if (priv->minor == GPIO_MINOR_A) + dir_shadow ^= 0xFF; /* Only 8 bits */ + else + dir_shadow ^= 0x3FFFF; /* Only 18 bits */ + return dir_shadow; + +} /* setget_input */ + +unsigned long inline setget_output(struct gpio_private *priv, unsigned long arg) +{ + unsigned long flags; + unsigned long dir_shadow; + + local_irq_save(flags); + dir_shadow = *dir_oe[priv->minor]; + dir_shadow |= (arg & changeable_dir[priv->minor]); + *dir_oe[priv->minor] = dir_shadow; + local_irq_restore(flags); + return dir_shadow; +} /* setget_output */ + +static int +gpio_leds_ioctl(unsigned int cmd, unsigned long arg); + +static int +gpio_ioctl(struct inode *inode, struct file *file, + unsigned int cmd, unsigned long arg) +{ + unsigned long flags; + unsigned long val; + unsigned long shadow; + struct gpio_private *priv = (struct gpio_private *)file->private_data; + if (_IOC_TYPE(cmd) != ETRAXGPIO_IOCTYPE) { + return -EINVAL; + } + + switch (_IOC_NR(cmd)) { + case IO_READBITS: /* Use IO_READ_INBITS and IO_READ_OUTBITS instead */ + // read the port + return *data_in[priv->minor]; + break; + case IO_SETBITS: + local_irq_save(flags); + if (arg & 0x04) + printk("GPIO SET 2\n"); + // set changeable bits with a 1 in arg + shadow = *data_out[priv->minor]; + shadow |= (arg & changeable_bits[priv->minor]); + *data_out[priv->minor] = shadow; + local_irq_restore(flags); + break; + case IO_CLRBITS: + local_irq_save(flags); + if (arg & 0x04) + printk("GPIO CLR 2\n"); + // clear changeable bits with a 1 in arg + shadow = *data_out[priv->minor]; + shadow &= ~(arg & changeable_bits[priv->minor]); + *data_out[priv->minor] = shadow; + local_irq_restore(flags); + break; + case IO_HIGHALARM: + // set alarm when bits with 1 in arg go high + priv->highalarm |= arg; + spin_lock(&alarm_lock); + gpio_some_alarms = 1; + if (priv->minor == GPIO_MINOR_A) { + gpio_pa_high_alarms |= arg; + } + spin_unlock(&alarm_lock); + break; + case IO_LOWALARM: + // set alarm when bits with 1 in arg go low + priv->lowalarm |= arg; + spin_lock(&alarm_lock); + gpio_some_alarms = 1; + if (priv->minor == GPIO_MINOR_A) { + gpio_pa_low_alarms |= arg; + } + spin_unlock(&alarm_lock); + break; + case IO_CLRALARM: + // clear alarm for bits with 1 in arg + priv->highalarm &= ~arg; + priv->lowalarm &= ~arg; + spin_lock(&alarm_lock); + if (priv->minor == GPIO_MINOR_A) { + if (gpio_pa_high_alarms & arg || + gpio_pa_low_alarms & arg) { + /* Must update the gpio_pa_*alarms masks */ + } + } + spin_unlock(&alarm_lock); + break; + case IO_READDIR: /* Use IO_SETGET_INPUT/OUTPUT instead! */ + /* Read direction 0=input 1=output */ + return *dir_oe[priv->minor]; + case IO_SETINPUT: /* Use IO_SETGET_INPUT instead! */ + /* Set direction 0=unchanged 1=input, + * return mask with 1=input + */ + return setget_input(priv, arg); + break; + case IO_SETOUTPUT: /* Use IO_SETGET_OUTPUT instead! */ + /* Set direction 0=unchanged 1=output, + * return mask with 1=output + */ + return setget_output(priv, arg); + + case IO_CFG_WRITE_MODE: + { + unsigned long dir_shadow; + dir_shadow = *dir_oe[priv->minor]; + + priv->clk_mask = arg & 0xFF; + priv->data_mask = (arg >> 8) & 0xFF; + priv->write_msb = (arg >> 16) & 0x01; + /* Check if we're allowed to change the bits and + * the direction is correct + */ + if (!((priv->clk_mask & changeable_bits[priv->minor]) && + (priv->data_mask & changeable_bits[priv->minor]) && + (priv->clk_mask & dir_shadow) && + (priv->data_mask & dir_shadow))) + { + priv->clk_mask = 0; + priv->data_mask = 0; + return -EPERM; + } + break; + } + case IO_READ_INBITS: + /* *arg is result of reading the input pins */ + val = *data_in[priv->minor]; + if (copy_to_user((unsigned long*)arg, &val, sizeof(val))) + return -EFAULT; + return 0; + break; + case IO_READ_OUTBITS: + /* *arg is result of reading the output shadow */ + val = *data_out[priv->minor]; + if (copy_to_user((unsigned long*)arg, &val, sizeof(val))) + return -EFAULT; + break; + case IO_SETGET_INPUT: + /* bits set in *arg is set to input, + * *arg updated with current input pins. + */ + if (copy_from_user(&val, (unsigned long*)arg, sizeof(val))) + return -EFAULT; + val = setget_input(priv, val); + if (copy_to_user((unsigned long*)arg, &val, sizeof(val))) + return -EFAULT; + break; + case IO_SETGET_OUTPUT: + /* bits set in *arg is set to output, + * *arg updated with current output pins. + */ + if (copy_from_user(&val, (unsigned long*)arg, sizeof(val))) + return -EFAULT; + val = setget_output(priv, val); + if (copy_to_user((unsigned long*)arg, &val, sizeof(val))) + return -EFAULT; + break; + default: + if (priv->minor == GPIO_MINOR_LEDS) + return gpio_leds_ioctl(cmd, arg); + else + return -EINVAL; + } /* switch */ + + return 0; +} + +static int +gpio_leds_ioctl(unsigned int cmd, unsigned long arg) +{ + unsigned char green; + unsigned char red; + + switch (_IOC_NR(cmd)) { + case IO_LEDACTIVE_SET: + green = ((unsigned char) arg) & 1; + red = (((unsigned char) arg) >> 1) & 1; + LED_ACTIVE_SET_G(green); + LED_ACTIVE_SET_R(red); + break; + + default: + return -EINVAL; + } /* switch */ + + return 0; +} + +struct file_operations gpio_fops = { + .owner = THIS_MODULE, + .poll = gpio_poll, + .ioctl = gpio_ioctl, + .write = gpio_write, + .open = gpio_open, + .release = gpio_release, +}; + + +/* main driver initialization routine, called from mem.c */ + +static __init int +gpio_init(void) +{ + int res; + reg_intr_vect_rw_mask intr_mask; + + /* do the formalities */ + + res = register_chrdev(GPIO_MAJOR, gpio_name, &gpio_fops); + if (res < 0) { + printk(KERN_ERR "gpio: couldn't get a major number.\n"); + return res; + } + + /* Clear all leds */ + LED_NETWORK_SET(0); + LED_ACTIVE_SET(0); + LED_DISK_READ(0); + LED_DISK_WRITE(0); + + printk("ETRAX FS GPIO driver v2.5, (c) 2003-2005 Axis Communications AB\n"); + /* We call etrax_gpio_wake_up_check() from timer interrupt and + * from cpu_idle() in kernel/process.c + * The check in cpu_idle() reduces latency from ~15 ms to ~6 ms + * in some tests. + */ + if (request_irq(TIMER_INTR_VECT, gpio_poll_timer_interrupt, + SA_SHIRQ | SA_INTERRUPT,"gpio poll", &alarmlist)) { + printk("err: timer0 irq for gpio\n"); + } + if (request_irq(GEN_IO_INTR_VECT, gpio_pa_interrupt, + SA_SHIRQ | SA_INTERRUPT,"gpio PA", &alarmlist)) { + printk("err: PA irq for gpio\n"); + } + /* enable the gio and timer irq in global config */ + intr_mask = REG_RD(intr_vect, regi_irq, rw_mask); + intr_mask.timer = 1; + intr_mask.gen_io = 1; + REG_WR(intr_vect, regi_irq, rw_mask, intr_mask); + + return res; +} + +/* this makes sure that gpio_init is called during kernel boot */ + +module_init(gpio_init); diff --git a/arch/cris/arch-v32/drivers/i2c.c b/arch/cris/arch-v32/drivers/i2c.c new file mode 100644 index 000000000000..440c20a94963 --- /dev/null +++ b/arch/cris/arch-v32/drivers/i2c.c @@ -0,0 +1,611 @@ +/*!*************************************************************************** +*! +*! FILE NAME : i2c.c +*! +*! DESCRIPTION: implements an interface for IIC/I2C, both directly from other +*! kernel modules (i2c_writereg/readreg) and from userspace using +*! ioctl()'s +*! +*! Nov 30 1998 Torbjorn Eliasson Initial version. +*! Bjorn Wesen Elinux kernel version. +*! Jan 14 2000 Johan Adolfsson Fixed PB shadow register stuff - +*! don't use PB_I2C if DS1302 uses same bits, +*! use PB. +*| June 23 2003 Pieter Grimmerink Added 'i2c_sendnack'. i2c_readreg now +*| generates nack on last received byte, +*| instead of ack. +*| i2c_getack changed data level while clock +*| was high, causing DS75 to see a stop condition +*! +*! --------------------------------------------------------------------------- +*! +*! (C) Copyright 1999-2002 Axis Communications AB, LUND, SWEDEN +*! +*!***************************************************************************/ +/* $Id: i2c.c,v 1.2 2005/05/09 15:29:49 starvik Exp $ */ +/****************** INCLUDE FILES SECTION ***********************************/ + +#include <linux/module.h> +#include <linux/sched.h> +#include <linux/slab.h> +#include <linux/errno.h> +#include <linux/kernel.h> +#include <linux/fs.h> +#include <linux/string.h> +#include <linux/init.h> +#include <linux/config.h> + +#include <asm/etraxi2c.h> + +#include <asm/system.h> +#include <asm/io.h> +#include <asm/delay.h> + +#include "i2c.h" + +/****************** I2C DEFINITION SECTION *************************/ + +#define D(x) + +#define I2C_MAJOR 123 /* LOCAL/EXPERIMENTAL */ +static const char i2c_name[] = "i2c"; + +#define CLOCK_LOW_TIME 8 +#define CLOCK_HIGH_TIME 8 +#define START_CONDITION_HOLD_TIME 8 +#define STOP_CONDITION_HOLD_TIME 8 +#define ENABLE_OUTPUT 0x01 +#define ENABLE_INPUT 0x00 +#define I2C_CLOCK_HIGH 1 +#define I2C_CLOCK_LOW 0 +#define I2C_DATA_HIGH 1 +#define I2C_DATA_LOW 0 + +#define i2c_enable() +#define i2c_disable() + +/* enable or disable output-enable, to select output or input on the i2c bus */ + +#define i2c_dir_out() crisv32_io_set_dir(&cris_i2c_data, crisv32_io_dir_out) +#define i2c_dir_in() crisv32_io_set_dir(&cris_i2c_data, crisv32_io_dir_in) + +/* control the i2c clock and data signals */ + +#define i2c_clk(x) crisv32_io_set(&cris_i2c_clk, x) +#define i2c_data(x) crisv32_io_set(&cris_i2c_data, x) + +/* read a bit from the i2c interface */ + +#define i2c_getbit() crisv32_io_rd(&cris_i2c_data) + +#define i2c_delay(usecs) udelay(usecs) + +/****************** VARIABLE SECTION ************************************/ + +static struct crisv32_iopin cris_i2c_clk; +static struct crisv32_iopin cris_i2c_data; + +/****************** FUNCTION DEFINITION SECTION *************************/ + + +/* generate i2c start condition */ + +void +i2c_start(void) +{ + /* + * SCL=1 SDA=1 + */ + i2c_dir_out(); + i2c_delay(CLOCK_HIGH_TIME/6); + i2c_data(I2C_DATA_HIGH); + i2c_clk(I2C_CLOCK_HIGH); + i2c_delay(CLOCK_HIGH_TIME); + /* + * SCL=1 SDA=0 + */ + i2c_data(I2C_DATA_LOW); + i2c_delay(START_CONDITION_HOLD_TIME); + /* + * SCL=0 SDA=0 + */ + i2c_clk(I2C_CLOCK_LOW); + i2c_delay(CLOCK_LOW_TIME); +} + +/* generate i2c stop condition */ + +void +i2c_stop(void) +{ + i2c_dir_out(); + + /* + * SCL=0 SDA=0 + */ + i2c_clk(I2C_CLOCK_LOW); + i2c_data(I2C_DATA_LOW); + i2c_delay(CLOCK_LOW_TIME*2); + /* + * SCL=1 SDA=0 + */ + i2c_clk(I2C_CLOCK_HIGH); + i2c_delay(CLOCK_HIGH_TIME*2); + /* + * SCL=1 SDA=1 + */ + i2c_data(I2C_DATA_HIGH); + i2c_delay(STOP_CONDITION_HOLD_TIME); + + i2c_dir_in(); +} + +/* write a byte to the i2c interface */ + +void +i2c_outbyte(unsigned char x) +{ + int i; + + i2c_dir_out(); + + for (i = 0; i < 8; i++) { + if (x & 0x80) { + i2c_data(I2C_DATA_HIGH); + } else { + i2c_data(I2C_DATA_LOW); + } + + i2c_delay(CLOCK_LOW_TIME/2); + i2c_clk(I2C_CLOCK_HIGH); + i2c_delay(CLOCK_HIGH_TIME); + i2c_clk(I2C_CLOCK_LOW); + i2c_delay(CLOCK_LOW_TIME/2); + x <<= 1; + } + i2c_data(I2C_DATA_LOW); + i2c_delay(CLOCK_LOW_TIME/2); + + /* + * enable input + */ + i2c_dir_in(); +} + +/* read a byte from the i2c interface */ + +unsigned char +i2c_inbyte(void) +{ + unsigned char aBitByte = 0; + int i; + + /* Switch off I2C to get bit */ + i2c_disable(); + i2c_dir_in(); + i2c_delay(CLOCK_HIGH_TIME/2); + + /* Get bit */ + aBitByte |= i2c_getbit(); + + /* Enable I2C */ + i2c_enable(); + i2c_delay(CLOCK_LOW_TIME/2); + + for (i = 1; i < 8; i++) { + aBitByte <<= 1; + /* Clock pulse */ + i2c_clk(I2C_CLOCK_HIGH); + i2c_delay(CLOCK_HIGH_TIME); + i2c_clk(I2C_CLOCK_LOW); + i2c_delay(CLOCK_LOW_TIME); + + /* Switch off I2C to get bit */ + i2c_disable(); + i2c_dir_in(); + i2c_delay(CLOCK_HIGH_TIME/2); + + /* Get bit */ + aBitByte |= i2c_getbit(); + + /* Enable I2C */ + i2c_enable(); + i2c_delay(CLOCK_LOW_TIME/2); + } + i2c_clk(I2C_CLOCK_HIGH); + i2c_delay(CLOCK_HIGH_TIME); + + /* + * we leave the clock low, getbyte is usually followed + * by sendack/nack, they assume the clock to be low + */ + i2c_clk(I2C_CLOCK_LOW); + return aBitByte; +} + +/*#--------------------------------------------------------------------------- +*# +*# FUNCTION NAME: i2c_getack +*# +*# DESCRIPTION : checks if ack was received from ic2 +*# +*#--------------------------------------------------------------------------*/ + +int +i2c_getack(void) +{ + int ack = 1; + /* + * enable output + */ + i2c_dir_out(); + /* + * Release data bus by setting + * data high + */ + i2c_data(I2C_DATA_HIGH); + /* + * enable input + */ + i2c_dir_in(); + i2c_delay(CLOCK_HIGH_TIME/4); + /* + * generate ACK clock pulse + */ + i2c_clk(I2C_CLOCK_HIGH); + /* + * Use PORT PB instead of I2C + * for input. (I2C not working) + */ + i2c_clk(1); + i2c_data(1); + /* + * switch off I2C + */ + i2c_data(1); + i2c_disable(); + i2c_dir_in(); + /* + * now wait for ack + */ + i2c_delay(CLOCK_HIGH_TIME/2); + /* + * check for ack + */ + if(i2c_getbit()) + ack = 0; + i2c_delay(CLOCK_HIGH_TIME/2); + if(!ack){ + if(!i2c_getbit()) /* receiver pulld SDA low */ + ack = 1; + i2c_delay(CLOCK_HIGH_TIME/2); + } + + /* + * our clock is high now, make sure data is low + * before we enable our output. If we keep data high + * and enable output, we would generate a stop condition. + */ + i2c_data(I2C_DATA_LOW); + + /* + * end clock pulse + */ + i2c_enable(); + i2c_dir_out(); + i2c_clk(I2C_CLOCK_LOW); + i2c_delay(CLOCK_HIGH_TIME/4); + /* + * enable output + */ + i2c_dir_out(); + /* + * remove ACK clock pulse + */ + i2c_data(I2C_DATA_HIGH); + i2c_delay(CLOCK_LOW_TIME/2); + return ack; +} + +/*#--------------------------------------------------------------------------- +*# +*# FUNCTION NAME: I2C::sendAck +*# +*# DESCRIPTION : Send ACK on received data +*# +*#--------------------------------------------------------------------------*/ +void +i2c_sendack(void) +{ + /* + * enable output + */ + i2c_delay(CLOCK_LOW_TIME); + i2c_dir_out(); + /* + * set ack pulse high + */ + i2c_data(I2C_DATA_LOW); + /* + * generate clock pulse + */ + i2c_delay(CLOCK_HIGH_TIME/6); + i2c_clk(I2C_CLOCK_HIGH); + i2c_delay(CLOCK_HIGH_TIME); + i2c_clk(I2C_CLOCK_LOW); + i2c_delay(CLOCK_LOW_TIME/6); + /* + * reset data out + */ + i2c_data(I2C_DATA_HIGH); + i2c_delay(CLOCK_LOW_TIME); + + i2c_dir_in(); +} + +/*#--------------------------------------------------------------------------- +*# +*# FUNCTION NAME: i2c_sendnack +*# +*# DESCRIPTION : Sends NACK on received data +*# +*#--------------------------------------------------------------------------*/ +void +i2c_sendnack(void) +{ + /* + * enable output + */ + i2c_delay(CLOCK_LOW_TIME); + i2c_dir_out(); + /* + * set data high + */ + i2c_data(I2C_DATA_HIGH); + /* + * generate clock pulse + */ + i2c_delay(CLOCK_HIGH_TIME/6); + i2c_clk(I2C_CLOCK_HIGH); + i2c_delay(CLOCK_HIGH_TIME); + i2c_clk(I2C_CLOCK_LOW); + i2c_delay(CLOCK_LOW_TIME); + + i2c_dir_in(); +} + +/*#--------------------------------------------------------------------------- +*# +*# FUNCTION NAME: i2c_writereg +*# +*# DESCRIPTION : Writes a value to an I2C device +*# +*#--------------------------------------------------------------------------*/ +int +i2c_writereg(unsigned char theSlave, unsigned char theReg, + unsigned char theValue) +{ + int error, cntr = 3; + unsigned long flags; + + do { + error = 0; + /* + * we don't like to be interrupted + */ + local_irq_save(flags); + + i2c_start(); + /* + * send slave address + */ + i2c_outbyte((theSlave & 0xfe)); + /* + * wait for ack + */ + if(!i2c_getack()) + error = 1; + /* + * now select register + */ + i2c_dir_out(); + i2c_outbyte(theReg); + /* + * now it's time to wait for ack + */ + if(!i2c_getack()) + error |= 2; + /* + * send register register data + */ + i2c_outbyte(theValue); + /* + * now it's time to wait for ack + */ + if(!i2c_getack()) + error |= 4; + /* + * end byte stream + */ + i2c_stop(); + /* + * enable interrupt again + */ + local_irq_restore(flags); + + } while(error && cntr--); + + i2c_delay(CLOCK_LOW_TIME); + + return -error; +} + +/*#--------------------------------------------------------------------------- +*# +*# FUNCTION NAME: i2c_readreg +*# +*# DESCRIPTION : Reads a value from the decoder registers. +*# +*#--------------------------------------------------------------------------*/ +unsigned char +i2c_readreg(unsigned char theSlave, unsigned char theReg) +{ + unsigned char b = 0; + int error, cntr = 3; + unsigned long flags; + + do { + error = 0; + /* + * we don't like to be interrupted + */ + local_irq_save(flags); + /* + * generate start condition + */ + i2c_start(); + + /* + * send slave address + */ + i2c_outbyte((theSlave & 0xfe)); + /* + * wait for ack + */ + if(!i2c_getack()) + error = 1; + /* + * now select register + */ + i2c_dir_out(); + i2c_outbyte(theReg); + /* + * now it's time to wait for ack + */ + if(!i2c_getack()) + error = 1; + /* + * repeat start condition + */ + i2c_delay(CLOCK_LOW_TIME); + i2c_start(); + /* + * send slave address + */ + i2c_outbyte(theSlave | 0x01); + /* + * wait for ack + */ + if(!i2c_getack()) + error = 1; + /* + * fetch register + */ + b = i2c_inbyte(); + /* + * last received byte needs to be nacked + * instead of acked + */ + i2c_sendnack(); + /* + * end sequence + */ + i2c_stop(); + /* + * enable interrupt again + */ + local_irq_restore(flags); + + } while(error && cntr--); + + return b; +} + +static int +i2c_open(struct inode *inode, struct file *filp) +{ + return 0; +} + +static int +i2c_release(struct inode *inode, struct file *filp) +{ + return 0; +} + +/* Main device API. ioctl's to write or read to/from i2c registers. + */ + +static int +i2c_ioctl(struct inode *inode, struct file *file, + unsigned int cmd, unsigned long arg) +{ + if(_IOC_TYPE(cmd) != ETRAXI2C_IOCTYPE) { + return -EINVAL; + } + + switch (_IOC_NR(cmd)) { + case I2C_WRITEREG: + /* write to an i2c slave */ + D(printk("i2cw %d %d %d\n", + I2C_ARGSLAVE(arg), + I2C_ARGREG(arg), + I2C_ARGVALUE(arg))); + + return i2c_writereg(I2C_ARGSLAVE(arg), + I2C_ARGREG(arg), + I2C_ARGVALUE(arg)); + case I2C_READREG: + { + unsigned char val; + /* read from an i2c slave */ + D(printk("i2cr %d %d ", + I2C_ARGSLAVE(arg), + I2C_ARGREG(arg))); + val = i2c_readreg(I2C_ARGSLAVE(arg), I2C_ARGREG(arg)); + D(printk("= %d\n", val)); + return val; + } + default: + return -EINVAL; + + } + + return 0; +} + +static struct file_operations i2c_fops = { + owner: THIS_MODULE, + ioctl: i2c_ioctl, + open: i2c_open, + release: i2c_release, +}; + +int __init +i2c_init(void) +{ + int res; + + /* Setup and enable the Port B I2C interface */ + + crisv32_io_get_name(&cris_i2c_data, CONFIG_ETRAX_I2C_DATA_PORT); + crisv32_io_get_name(&cris_i2c_clk, CONFIG_ETRAX_I2C_CLK_PORT); + + /* register char device */ + + res = register_chrdev(I2C_MAJOR, i2c_name, &i2c_fops); + if(res < 0) { + printk(KERN_ERR "i2c: couldn't get a major number.\n"); + return res; + } + + printk(KERN_INFO "I2C driver v2.2, (c) 1999-2001 Axis Communications AB\n"); + + return 0; +} + +/* this makes sure that i2c_init is called during boot */ + +module_init(i2c_init); + +/****************** END OF FILE i2c.c ********************************/ diff --git a/arch/cris/arch-v32/drivers/i2c.h b/arch/cris/arch-v32/drivers/i2c.h new file mode 100644 index 000000000000..bfe1a13f9f35 --- /dev/null +++ b/arch/cris/arch-v32/drivers/i2c.h @@ -0,0 +1,15 @@ + +#include <linux/init.h> + +/* High level I2C actions */ +int __init i2c_init(void); +int i2c_writereg(unsigned char theSlave, unsigned char theReg, unsigned char theValue); +unsigned char i2c_readreg(unsigned char theSlave, unsigned char theReg); + +/* Low level I2C */ +void i2c_start(void); +void i2c_stop(void); +void i2c_outbyte(unsigned char x); +unsigned char i2c_inbyte(void); +int i2c_getack(void); +void i2c_sendack(void); diff --git a/arch/cris/arch-v32/drivers/iop_fw_load.c b/arch/cris/arch-v32/drivers/iop_fw_load.c new file mode 100644 index 000000000000..11f9895ded50 --- /dev/null +++ b/arch/cris/arch-v32/drivers/iop_fw_load.c @@ -0,0 +1,219 @@ +/* $Id: iop_fw_load.c,v 1.4 2005/04/07 09:27:46 larsv Exp $ + * + * Firmware loader for ETRAX FS IO-Processor + * + * Copyright (C) 2004 Axis Communications AB + */ + +#include <linux/module.h> +#include <linux/kernel.h> +#include <linux/init.h> +#include <linux/device.h> +#include <linux/firmware.h> + +#include <asm/arch/hwregs/reg_map.h> +#include <asm/arch/hwregs/iop/iop_reg_space.h> +#include <asm/arch/hwregs/iop/iop_mpu_macros.h> +#include <asm/arch/hwregs/iop/iop_mpu_defs.h> +#include <asm/arch/hwregs/iop/iop_spu_defs.h> +#include <asm/arch/hwregs/iop/iop_sw_cpu_defs.h> + +#define IOP_TIMEOUT 100 + +static struct device iop_spu_device[2] = { + { .bus_id = "iop-spu0", }, + { .bus_id = "iop-spu1", }, +}; + +static struct device iop_mpu_device = { + .bus_id = "iop-mpu", +}; + +static int wait_mpu_idle(void) +{ + reg_iop_mpu_r_stat mpu_stat; + unsigned int timeout = IOP_TIMEOUT; + + do { + mpu_stat = REG_RD(iop_mpu, regi_iop_mpu, r_stat); + } while (mpu_stat.instr_reg_busy == regk_iop_mpu_yes && --timeout > 0); + if (timeout == 0) { + printk(KERN_ERR "Timeout waiting for MPU to be idle\n"); + return -EBUSY; + } + return 0; +} + +int iop_fw_load_spu(const unsigned char *fw_name, unsigned int spu_inst) +{ + reg_iop_sw_cpu_rw_mc_ctrl mc_ctrl = { + .wr_spu0_mem = regk_iop_sw_cpu_no, + .wr_spu1_mem = regk_iop_sw_cpu_no, + .size = 4, + .cmd = regk_iop_sw_cpu_reg_copy, + .keep_owner = regk_iop_sw_cpu_yes + }; + reg_iop_spu_rw_ctrl spu_ctrl = { + .en = regk_iop_spu_no, + .fsm = regk_iop_spu_no, + }; + reg_iop_sw_cpu_r_mc_stat mc_stat; + const struct firmware *fw_entry; + u32 *data; + unsigned int timeout; + int retval, i; + + if (spu_inst > 1) + return -ENODEV; + + /* get firmware */ + retval = request_firmware(&fw_entry, + fw_name, + &iop_spu_device[spu_inst]); + if (retval != 0) + { + printk(KERN_ERR + "iop_load_spu: Failed to load firmware \"%s\"\n", + fw_name); + return retval; + } + data = (u32 *) fw_entry->data; + + /* acquire ownership of memory controller */ + switch (spu_inst) { + case 0: + mc_ctrl.wr_spu0_mem = regk_iop_sw_cpu_yes; + REG_WR(iop_spu, regi_iop_spu0, rw_ctrl, spu_ctrl); + break; + case 1: + mc_ctrl.wr_spu1_mem = regk_iop_sw_cpu_yes; + REG_WR(iop_spu, regi_iop_spu1, rw_ctrl, spu_ctrl); + break; + } + timeout = IOP_TIMEOUT; + do { + REG_WR(iop_sw_cpu, regi_iop_sw_cpu, rw_mc_ctrl, mc_ctrl); + mc_stat = REG_RD(iop_sw_cpu, regi_iop_sw_cpu, r_mc_stat); + } while (mc_stat.owned_by_cpu == regk_iop_sw_cpu_no && --timeout > 0); + if (timeout == 0) { + printk(KERN_ERR "Timeout waiting to acquire MC\n"); + retval = -EBUSY; + goto out; + } + + /* write to SPU memory */ + for (i = 0; i < (fw_entry->size/4); i++) { + switch (spu_inst) { + case 0: + REG_WR_INT(iop_spu, regi_iop_spu0, rw_seq_pc, (i*4)); + break; + case 1: + REG_WR_INT(iop_spu, regi_iop_spu1, rw_seq_pc, (i*4)); + break; + } + REG_WR_INT(iop_sw_cpu, regi_iop_sw_cpu, rw_mc_data, *data); + data++; + } + + /* release ownership of memory controller */ + (void) REG_RD(iop_sw_cpu, regi_iop_sw_cpu, rs_mc_data); + + out: + release_firmware(fw_entry); + return retval; +} + +int iop_fw_load_mpu(unsigned char *fw_name) +{ + const unsigned int start_addr = 0; + reg_iop_mpu_rw_ctrl mpu_ctrl; + const struct firmware *fw_entry; + u32 *data; + int retval, i; + + /* get firmware */ + retval = request_firmware(&fw_entry, fw_name, &iop_mpu_device); + if (retval != 0) + { + printk(KERN_ERR + "iop_load_spu: Failed to load firmware \"%s\"\n", + fw_name); + return retval; + } + data = (u32 *) fw_entry->data; + + /* disable MPU */ + mpu_ctrl.en = regk_iop_mpu_no; + REG_WR(iop_mpu, regi_iop_mpu, rw_ctrl, mpu_ctrl); + /* put start address in R0 */ + REG_WR_VECT(iop_mpu, regi_iop_mpu, rw_r, 0, start_addr); + /* write to memory by executing 'SWX i, 4, R0' for each word */ + if ((retval = wait_mpu_idle()) != 0) + goto out; + REG_WR(iop_mpu, regi_iop_mpu, rw_instr, MPU_SWX_IIR_INSTR(0, 4, 0)); + for (i = 0; i < (fw_entry->size / 4); i++) { + REG_WR_INT(iop_mpu, regi_iop_mpu, rw_immediate, *data); + if ((retval = wait_mpu_idle()) != 0) + goto out; + data++; + } + + out: + release_firmware(fw_entry); + return retval; +} + +int iop_start_mpu(unsigned int start_addr) +{ + reg_iop_mpu_rw_ctrl mpu_ctrl = { .en = regk_iop_mpu_yes }; + int retval; + + /* disable MPU */ + if ((retval = wait_mpu_idle()) != 0) + goto out; + REG_WR(iop_mpu, regi_iop_mpu, rw_instr, MPU_HALT()); + if ((retval = wait_mpu_idle()) != 0) + goto out; + /* set PC and wait for it to bite */ + if ((retval = wait_mpu_idle()) != 0) + goto out; + REG_WR_INT(iop_mpu, regi_iop_mpu, rw_instr, MPU_BA_I(start_addr)); + if ((retval = wait_mpu_idle()) != 0) + goto out; + /* make sure the MPU starts executing with interrupts disabled */ + REG_WR(iop_mpu, regi_iop_mpu, rw_instr, MPU_DI()); + if ((retval = wait_mpu_idle()) != 0) + goto out; + /* enable MPU */ + REG_WR(iop_mpu, regi_iop_mpu, rw_ctrl, mpu_ctrl); + out: + return retval; +} + +static int __init iop_fw_load_init(void) +{ + device_initialize(&iop_spu_device[0]); + kobject_set_name(&iop_spu_device[0].kobj, "iop-spu0"); + kobject_add(&iop_spu_device[0].kobj); + device_initialize(&iop_spu_device[1]); + kobject_set_name(&iop_spu_device[1].kobj, "iop-spu1"); + kobject_add(&iop_spu_device[1].kobj); + device_initialize(&iop_mpu_device); + kobject_set_name(&iop_mpu_device.kobj, "iop-mpu"); + kobject_add(&iop_mpu_device.kobj); + return 0; +} + +static void __exit iop_fw_load_exit(void) +{ +} + +module_init(iop_fw_load_init); +module_exit(iop_fw_load_exit); + +MODULE_DESCRIPTION("ETRAX FS IO-Processor Firmware Loader"); +MODULE_LICENSE("GPL"); + +EXPORT_SYMBOL(iop_fw_load_spu); +EXPORT_SYMBOL(iop_fw_load_mpu); +EXPORT_SYMBOL(iop_start_mpu); diff --git a/arch/cris/arch-v32/drivers/nandflash.c b/arch/cris/arch-v32/drivers/nandflash.c new file mode 100644 index 000000000000..fc2a619b035d --- /dev/null +++ b/arch/cris/arch-v32/drivers/nandflash.c @@ -0,0 +1,157 @@ +/* + * arch/cris/arch-v32/drivers/nandflash.c + * + * Copyright (c) 2004 + * + * Derived from drivers/mtd/nand/spia.c + * Copyright (C) 2000 Steven J. Hill (sjhill@realitydiluted.com) + * + * $Id: nandflash.c,v 1.3 2005/06/01 10:57:12 starvik Exp $ + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + * + */ + +#include <linux/version.h> +#include <linux/slab.h> +#include <linux/init.h> +#include <linux/module.h> +#include <linux/mtd/mtd.h> +#include <linux/mtd/nand.h> +#include <linux/mtd/partitions.h> +#include <asm/arch/memmap.h> +#include <asm/arch/hwregs/reg_map.h> +#include <asm/arch/hwregs/reg_rdwr.h> +#include <asm/arch/hwregs/gio_defs.h> +#include <asm/arch/hwregs/bif_core_defs.h> +#include <asm/io.h> + +#define CE_BIT 4 +#define CLE_BIT 5 +#define ALE_BIT 6 +#define BY_BIT 7 + +static struct mtd_info *crisv32_mtd = NULL; +/* + * hardware specific access to control-lines +*/ +static void crisv32_hwcontrol(struct mtd_info *mtd, int cmd) +{ + unsigned long flags; + reg_gio_rw_pa_dout dout = REG_RD(gio, regi_gio, rw_pa_dout); + + local_irq_save(flags); + switch(cmd){ + case NAND_CTL_SETCLE: + dout.data |= (1<<CLE_BIT); + break; + case NAND_CTL_CLRCLE: + dout.data &= ~(1<<CLE_BIT); + break; + case NAND_CTL_SETALE: + dout.data |= (1<<ALE_BIT); + break; + case NAND_CTL_CLRALE: + dout.data &= ~(1<<ALE_BIT); + break; + case NAND_CTL_SETNCE: + dout.data |= (1<<CE_BIT); + break; + case NAND_CTL_CLRNCE: + dout.data &= ~(1<<CE_BIT); + break; + } + REG_WR(gio, regi_gio, rw_pa_dout, dout); + local_irq_restore(flags); +} + +/* +* read device ready pin +*/ +int crisv32_device_ready(struct mtd_info *mtd) +{ + reg_gio_r_pa_din din = REG_RD(gio, regi_gio, r_pa_din); + return ((din.data & (1 << BY_BIT)) >> BY_BIT); +} + +/* + * Main initialization routine + */ +struct mtd_info* __init crisv32_nand_flash_probe (void) +{ + void __iomem *read_cs; + void __iomem *write_cs; + + reg_bif_core_rw_grp3_cfg bif_cfg = REG_RD(bif_core, regi_bif_core, rw_grp3_cfg); + reg_gio_rw_pa_oe pa_oe = REG_RD(gio, regi_gio, rw_pa_oe); + struct nand_chip *this; + int err = 0; + + /* Allocate memory for MTD device structure and private data */ + crisv32_mtd = kmalloc (sizeof(struct mtd_info) + sizeof (struct nand_chip), + GFP_KERNEL); + if (!crisv32_mtd) { + printk ("Unable to allocate CRISv32 NAND MTD device structure.\n"); + err = -ENOMEM; + return NULL; + } + + read_cs = ioremap(MEM_CSP0_START | MEM_NON_CACHEABLE, 8192); + write_cs = ioremap(MEM_CSP1_START | MEM_NON_CACHEABLE, 8192); + + if (!read_cs || !write_cs) { + printk("CRISv32 NAND ioremap failed\n"); + err = -EIO; + goto out_mtd; + } + + /* Get pointer to private data */ + this = (struct nand_chip *) (&crisv32_mtd[1]); + + pa_oe.oe |= 1 << CE_BIT; + pa_oe.oe |= 1 << ALE_BIT; + pa_oe.oe |= 1 << CLE_BIT; + pa_oe.oe &= ~ (1 << BY_BIT); + REG_WR(gio, regi_gio, rw_pa_oe, pa_oe); + + bif_cfg.gated_csp0 = regk_bif_core_rd; + bif_cfg.gated_csp1 = regk_bif_core_wr; + REG_WR(bif_core, regi_bif_core, rw_grp3_cfg, bif_cfg); + + /* Initialize structures */ + memset((char *) crisv32_mtd, 0, sizeof(struct mtd_info)); + memset((char *) this, 0, sizeof(struct nand_chip)); + + /* Link the private data with the MTD structure */ + crisv32_mtd->priv = this; + + /* Set address of NAND IO lines */ + this->IO_ADDR_R = read_cs; + this->IO_ADDR_W = write_cs; + this->hwcontrol = crisv32_hwcontrol; + this->dev_ready = crisv32_device_ready; + /* 20 us command delay time */ + this->chip_delay = 20; + this->eccmode = NAND_ECC_SOFT; + + /* Enable the following for a flash based bad block table */ + this->options = NAND_USE_FLASH_BBT; + + /* Scan to find existance of the device */ + if (nand_scan (crisv32_mtd, 1)) { + err = -ENXIO; + goto out_ior; + } + + return crisv32_mtd; + +out_ior: + iounmap((void *)read_cs); + iounmap((void *)write_cs); +out_mtd: + kfree (crisv32_mtd); + return NULL; +} + diff --git a/arch/cris/arch-v32/drivers/pcf8563.c b/arch/cris/arch-v32/drivers/pcf8563.c new file mode 100644 index 000000000000..f894580b648b --- /dev/null +++ b/arch/cris/arch-v32/drivers/pcf8563.c @@ -0,0 +1,341 @@ +/* + * PCF8563 RTC + * + * From Phillips' datasheet: + * + * The PCF8563 is a CMOS real-time clock/calendar optimized for low power + * consumption. A programmable clock output, interupt output and voltage + * low detector are also provided. All address and data are transferred + * serially via two-line bidirectional I2C-bus. Maximum bus speed is + * 400 kbits/s. The built-in word address register is incremented + * automatically after each written or read byte. + * + * Copyright (c) 2002-2003, Axis Communications AB + * All rights reserved. + * + * Author: Tobias Anderberg <tobiasa@axis.com>. + * + */ + +#include <linux/config.h> +#include <linux/version.h> +#include <linux/module.h> +#include <linux/kernel.h> +#include <linux/types.h> +#include <linux/sched.h> +#include <linux/init.h> +#include <linux/fs.h> +#include <linux/ioctl.h> +#include <linux/delay.h> +#include <linux/bcd.h> + +#include <asm/uaccess.h> +#include <asm/system.h> +#include <asm/io.h> +#include <asm/rtc.h> + +#include "i2c.h" + +#define PCF8563_MAJOR 121 /* Local major number. */ +#define DEVICE_NAME "rtc" /* Name which is registered in /proc/devices. */ +#define PCF8563_NAME "PCF8563" +#define DRIVER_VERSION "$Revision: 1.1 $" + +/* Two simple wrapper macros, saves a few keystrokes. */ +#define rtc_read(x) i2c_readreg(RTC_I2C_READ, x) +#define rtc_write(x,y) i2c_writereg(RTC_I2C_WRITE, x, y) + +static const unsigned char days_in_month[] = + { 0, 31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31 }; + +int pcf8563_ioctl(struct inode *, struct file *, unsigned int, unsigned long); +int pcf8563_open(struct inode *, struct file *); +int pcf8563_release(struct inode *, struct file *); + +static struct file_operations pcf8563_fops = { + owner: THIS_MODULE, + ioctl: pcf8563_ioctl, + open: pcf8563_open, + release: pcf8563_release, +}; + +unsigned char +pcf8563_readreg(int reg) +{ + unsigned char res = rtc_read(reg); + + /* The PCF8563 does not return 0 for unimplemented bits */ + switch (reg) { + case RTC_SECONDS: + case RTC_MINUTES: + res &= 0x7F; + break; + case RTC_HOURS: + case RTC_DAY_OF_MONTH: + res &= 0x3F; + break; + case RTC_WEEKDAY: + res &= 0x07; + break; + case RTC_MONTH: + res &= 0x1F; + break; + case RTC_CONTROL1: + res &= 0xA8; + break; + case RTC_CONTROL2: + res &= 0x1F; + break; + case RTC_CLOCKOUT_FREQ: + case RTC_TIMER_CONTROL: + res &= 0x83; + break; + } + return res; +} + +void +pcf8563_writereg(int reg, unsigned char val) +{ +#ifdef CONFIG_ETRAX_RTC_READONLY + if (reg == RTC_CONTROL1 || (reg >= RTC_SECONDS && reg <= RTC_YEAR)) + return; +#endif + + rtc_write(reg, val); +} + +void +get_rtc_time(struct rtc_time *tm) +{ + tm->tm_sec = rtc_read(RTC_SECONDS); + tm->tm_min = rtc_read(RTC_MINUTES); + tm->tm_hour = rtc_read(RTC_HOURS); + tm->tm_mday = rtc_read(RTC_DAY_OF_MONTH); + tm->tm_wday = rtc_read(RTC_WEEKDAY); + tm->tm_mon = rtc_read(RTC_MONTH); + tm->tm_year = rtc_read(RTC_YEAR); + + if (tm->tm_sec & 0x80) + printk(KERN_WARNING "%s: RTC Voltage Low - reliable date/time " + "information is no longer guaranteed!\n", PCF8563_NAME); + + tm->tm_year = BCD_TO_BIN(tm->tm_year) + ((tm->tm_mon & 0x80) ? 100 : 0); + tm->tm_sec &= 0x7F; + tm->tm_min &= 0x7F; + tm->tm_hour &= 0x3F; + tm->tm_mday &= 0x3F; + tm->tm_wday &= 0x07; /* Not coded in BCD. */ + tm->tm_mon &= 0x1F; + + BCD_TO_BIN(tm->tm_sec); + BCD_TO_BIN(tm->tm_min); + BCD_TO_BIN(tm->tm_hour); + BCD_TO_BIN(tm->tm_mday); + BCD_TO_BIN(tm->tm_mon); + tm->tm_mon--; /* Month is 1..12 in RTC but 0..11 in linux */ +} + +int __init +pcf8563_init(void) +{ + /* Initiate the i2c protocol. */ + i2c_init(); + + /* + * First of all we need to reset the chip. This is done by + * clearing control1, control2 and clk freq and resetting + * all alarms. + */ + if (rtc_write(RTC_CONTROL1, 0x00) < 0) + goto err; + + if (rtc_write(RTC_CONTROL2, 0x00) < 0) + goto err; + + if (rtc_write(RTC_CLOCKOUT_FREQ, 0x00) < 0) + goto err; + + if (rtc_write(RTC_TIMER_CONTROL, 0x03) < 0) + goto err; + + /* Reset the alarms. */ + if (rtc_write(RTC_MINUTE_ALARM, 0x80) < 0) + goto err; + + if (rtc_write(RTC_HOUR_ALARM, 0x80) < 0) + goto err; + + if (rtc_write(RTC_DAY_ALARM, 0x80) < 0) + goto err; + + if (rtc_write(RTC_WEEKDAY_ALARM, 0x80) < 0) + goto err; + + if (register_chrdev(PCF8563_MAJOR, DEVICE_NAME, &pcf8563_fops) < 0) { + printk(KERN_INFO "%s: Unable to get major numer %d for RTC device.\n", + PCF8563_NAME, PCF8563_MAJOR); + return -1; + } + + printk(KERN_INFO "%s Real-Time Clock Driver, %s\n", PCF8563_NAME, DRIVER_VERSION); + + /* Check for low voltage, and warn about it.. */ + if (rtc_read(RTC_SECONDS) & 0x80) + printk(KERN_WARNING "%s: RTC Voltage Low - reliable date/time " + "information is no longer guaranteed!\n", PCF8563_NAME); + + return 0; + +err: + printk(KERN_INFO "%s: Error initializing chip.\n", PCF8563_NAME); + return -1; +} + +void __exit +pcf8563_exit(void) +{ + if (unregister_chrdev(PCF8563_MAJOR, DEVICE_NAME) < 0) { + printk(KERN_INFO "%s: Unable to unregister device.\n", PCF8563_NAME); + } +} + +/* + * ioctl calls for this driver. Why return -ENOTTY upon error? Because + * POSIX says so! + */ +int +pcf8563_ioctl(struct inode *inode, struct file *filp, unsigned int cmd, unsigned long arg) +{ + /* Some sanity checks. */ + if (_IOC_TYPE(cmd) != RTC_MAGIC) + return -ENOTTY; + + if (_IOC_NR(cmd) > RTC_MAX_IOCTL) + return -ENOTTY; + + switch (cmd) { + case RTC_RD_TIME: + { + struct rtc_time tm; + + memset(&tm, 0, sizeof (struct rtc_time)); + get_rtc_time(&tm); + + if (copy_to_user((struct rtc_time *) arg, &tm, sizeof tm)) { + return -EFAULT; + } + + return 0; + } + + case RTC_SET_TIME: + { +#ifdef CONFIG_ETRAX_RTC_READONLY + return -EPERM; +#else + int leap; + int year; + int century; + struct rtc_time tm; + + if (!capable(CAP_SYS_TIME)) + return -EPERM; + + if (copy_from_user(&tm, (struct rtc_time *) arg, sizeof tm)) + return -EFAULT; + + /* Convert from struct tm to struct rtc_time. */ + tm.tm_year += 1900; + tm.tm_mon += 1; + + /* + * Check if tm.tm_year is a leap year. A year is a leap + * year if it is divisible by 4 but not 100, except + * that years divisible by 400 _are_ leap years. + */ + year = tm.tm_year; + leap = (tm.tm_mon == 2) && ((year % 4 == 0 && year % 100 != 0) || year % 400 == 0); + + /* Perform some sanity checks. */ + if ((tm.tm_year < 1970) || + (tm.tm_mon > 12) || + (tm.tm_mday == 0) || + (tm.tm_mday > days_in_month[tm.tm_mon] + leap) || + (tm.tm_wday >= 7) || + (tm.tm_hour >= 24) || + (tm.tm_min >= 60) || + (tm.tm_sec >= 60)) + return -EINVAL; + + century = (tm.tm_year >= 2000) ? 0x80 : 0; + tm.tm_year = tm.tm_year % 100; + + BIN_TO_BCD(tm.tm_year); + BIN_TO_BCD(tm.tm_mday); + BIN_TO_BCD(tm.tm_hour); + BIN_TO_BCD(tm.tm_min); + BIN_TO_BCD(tm.tm_sec); + tm.tm_mon |= century; + + rtc_write(RTC_YEAR, tm.tm_year); + rtc_write(RTC_MONTH, tm.tm_mon); + rtc_write(RTC_WEEKDAY, tm.tm_wday); /* Not coded in BCD. */ + rtc_write(RTC_DAY_OF_MONTH, tm.tm_mday); + rtc_write(RTC_HOURS, tm.tm_hour); + rtc_write(RTC_MINUTES, tm.tm_min); + rtc_write(RTC_SECONDS, tm.tm_sec); + + return 0; +#endif /* !CONFIG_ETRAX_RTC_READONLY */ + } + + case RTC_VLOW_RD: + { + int vl_bit = 0; + + if (rtc_read(RTC_SECONDS) & 0x80) { + vl_bit = 1; + printk(KERN_WARNING "%s: RTC Voltage Low - reliable " + "date/time information is no longer guaranteed!\n", + PCF8563_NAME); + } + if (copy_to_user((int *) arg, &vl_bit, sizeof(int))) + return -EFAULT; + + return 0; + } + + case RTC_VLOW_SET: + { + /* Clear the VL bit in the seconds register */ + int ret = rtc_read(RTC_SECONDS); + + rtc_write(RTC_SECONDS, (ret & 0x7F)); + + return 0; + } + + default: + return -ENOTTY; + } + + return 0; +} + +int +pcf8563_open(struct inode *inode, struct file *filp) +{ + MOD_INC_USE_COUNT; + return 0; +} + +int +pcf8563_release(struct inode *inode, struct file *filp) +{ + MOD_DEC_USE_COUNT; + return 0; +} + +module_init(pcf8563_init); +module_exit(pcf8563_exit); diff --git a/arch/cris/arch-v32/drivers/pci/Makefile b/arch/cris/arch-v32/drivers/pci/Makefile new file mode 100644 index 000000000000..bff7482f2444 --- /dev/null +++ b/arch/cris/arch-v32/drivers/pci/Makefile @@ -0,0 +1,5 @@ +# +# Makefile for Etrax cardbus driver +# + +obj-$(CONFIG_ETRAX_CARDBUS) += bios.o dma.o diff --git a/arch/cris/arch-v32/drivers/pci/bios.c b/arch/cris/arch-v32/drivers/pci/bios.c new file mode 100644 index 000000000000..24bc149889b6 --- /dev/null +++ b/arch/cris/arch-v32/drivers/pci/bios.c @@ -0,0 +1,131 @@ +#include <linux/pci.h> +#include <linux/kernel.h> +#include <asm/arch/hwregs/intr_vect.h> + +void __devinit pcibios_fixup_bus(struct pci_bus *b) +{ +} + +char * __devinit pcibios_setup(char *str) +{ + return NULL; +} + +void pcibios_set_master(struct pci_dev *dev) +{ + u8 lat; + pci_read_config_byte(dev, PCI_LATENCY_TIMER, &lat); + printk(KERN_DEBUG "PCI: Setting latency timer of device %s to %d\n", pci_name(dev), lat); + pci_write_config_byte(dev, PCI_LATENCY_TIMER, lat); +} + +int pci_mmap_page_range(struct pci_dev *dev, struct vm_area_struct *vma, + enum pci_mmap_state mmap_state, int write_combine) +{ + unsigned long prot; + + /* Leave vm_pgoff as-is, the PCI space address is the physical + * address on this platform. + */ + vma->vm_flags |= (VM_SHM | VM_LOCKED | VM_IO); + + prot = pgprot_val(vma->vm_page_prot); + vma->vm_page_prot = __pgprot(prot); + + /* Write-combine setting is ignored, it is changed via the mtrr + * interfaces on this platform. + */ + if (remap_pfn_range(vma, vma->vm_start, vma->vm_pgoff, + vma->vm_end - vma->vm_start, + vma->vm_page_prot)) + return -EAGAIN; + + return 0; +} + +void +pcibios_align_resource(void *data, struct resource *res, + unsigned long size, unsigned long align) +{ + if (res->flags & IORESOURCE_IO) { + unsigned long start = res->start; + + if (start & 0x300) { + start = (start + 0x3ff) & ~0x3ff; + res->start = start; + } + } +} + +int pcibios_enable_resources(struct pci_dev *dev, int mask) +{ + u16 cmd, old_cmd; + int idx; + struct resource *r; + + pci_read_config_word(dev, PCI_COMMAND, &cmd); + old_cmd = cmd; + for(idx=0; idx<6; idx++) { + /* Only set up the requested stuff */ + if (!(mask & (1<<idx))) + continue; + + r = &dev->resource[idx]; + if (!r->start && r->end) { + printk(KERN_ERR "PCI: Device %s not available because of resource collisions\n", pci_name(dev)); + return -EINVAL; + } + if (r->flags & IORESOURCE_IO) + cmd |= PCI_COMMAND_IO; + if (r->flags & IORESOURCE_MEM) + cmd |= PCI_COMMAND_MEMORY; + } + if (dev->resource[PCI_ROM_RESOURCE].start) + cmd |= PCI_COMMAND_MEMORY; + if (cmd != old_cmd) { + printk("PCI: Enabling device %s (%04x -> %04x)\n", pci_name(dev), old_cmd, cmd); + pci_write_config_word(dev, PCI_COMMAND, cmd); + } + return 0; +} + +int pcibios_enable_irq(struct pci_dev *dev) +{ + dev->irq = EXT_INTR_VECT; + return 0; +} + +int pcibios_enable_device(struct pci_dev *dev, int mask) +{ + int err; + + if ((err = pcibios_enable_resources(dev, mask)) < 0) + return err; + + return pcibios_enable_irq(dev); +} + +int pcibios_assign_resources(void) +{ + struct pci_dev *dev = NULL; + int idx; + struct resource *r; + + while ((dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev)) != NULL) { + int class = dev->class >> 8; + + /* Don't touch classless devices and host bridges */ + if (!class || class == PCI_CLASS_BRIDGE_HOST) + continue; + + for(idx=0; idx<6; idx++) { + r = &dev->resource[idx]; + + if (!r->start && r->end) + pci_assign_resource(dev, idx); + } + } + return 0; +} + +EXPORT_SYMBOL(pcibios_assign_resources); diff --git a/arch/cris/arch-v32/drivers/pci/dma.c b/arch/cris/arch-v32/drivers/pci/dma.c new file mode 100644 index 000000000000..10329306d23c --- /dev/null +++ b/arch/cris/arch-v32/drivers/pci/dma.c @@ -0,0 +1,149 @@ +/* + * Dynamic DMA mapping support. + * + * On cris there is no hardware dynamic DMA address translation, + * so consistent alloc/free are merely page allocation/freeing. + * The rest of the dynamic DMA mapping interface is implemented + * in asm/pci.h. + * + * Borrowed from i386. + */ + +#include <linux/types.h> +#include <linux/mm.h> +#include <linux/string.h> +#include <linux/pci.h> +#include <asm/io.h> + +struct dma_coherent_mem { + void *virt_base; + u32 device_base; + int size; + int flags; + unsigned long *bitmap; +}; + +void *dma_alloc_coherent(struct device *dev, size_t size, + dma_addr_t *dma_handle, unsigned int __nocast gfp) +{ + void *ret; + struct dma_coherent_mem *mem = dev ? dev->dma_mem : NULL; + int order = get_order(size); + /* ignore region specifiers */ + gfp &= ~(__GFP_DMA | __GFP_HIGHMEM); + + if (mem) { + int page = bitmap_find_free_region(mem->bitmap, mem->size, + order); + if (page >= 0) { + *dma_handle = mem->device_base + (page << PAGE_SHIFT); + ret = mem->virt_base + (page << PAGE_SHIFT); + memset(ret, 0, size); + return ret; + } + if (mem->flags & DMA_MEMORY_EXCLUSIVE) + return NULL; + } + + if (dev == NULL || (dev->coherent_dma_mask < 0xffffffff)) + gfp |= GFP_DMA; + + ret = (void *)__get_free_pages(gfp, order); + + if (ret != NULL) { + memset(ret, 0, size); + *dma_handle = virt_to_phys(ret); + } + return ret; +} + +void dma_free_coherent(struct device *dev, size_t size, + void *vaddr, dma_addr_t dma_handle) +{ + struct dma_coherent_mem *mem = dev ? dev->dma_mem : NULL; + int order = get_order(size); + + if (mem && vaddr >= mem->virt_base && vaddr < (mem->virt_base + (mem->size << PAGE_SHIFT))) { + int page = (vaddr - mem->virt_base) >> PAGE_SHIFT; + + bitmap_release_region(mem->bitmap, page, order); + } else + free_pages((unsigned long)vaddr, order); +} + +int dma_declare_coherent_memory(struct device *dev, dma_addr_t bus_addr, + dma_addr_t device_addr, size_t size, int flags) +{ + void __iomem *mem_base; + int pages = size >> PAGE_SHIFT; + int bitmap_size = (pages + 31)/32; + + if ((flags & (DMA_MEMORY_MAP | DMA_MEMORY_IO)) == 0) + goto out; + if (!size) + goto out; + if (dev->dma_mem) + goto out; + + /* FIXME: this routine just ignores DMA_MEMORY_INCLUDES_CHILDREN */ + + mem_base = ioremap(bus_addr, size); + if (!mem_base) + goto out; + + dev->dma_mem = kmalloc(sizeof(struct dma_coherent_mem), GFP_KERNEL); + if (!dev->dma_mem) + goto out; + memset(dev->dma_mem, 0, sizeof(struct dma_coherent_mem)); + dev->dma_mem->bitmap = kmalloc(bitmap_size, GFP_KERNEL); + if (!dev->dma_mem->bitmap) + goto free1_out; + memset(dev->dma_mem->bitmap, 0, bitmap_size); + + dev->dma_mem->virt_base = mem_base; + dev->dma_mem->device_base = device_addr; + dev->dma_mem->size = pages; + dev->dma_mem->flags = flags; + + if (flags & DMA_MEMORY_MAP) + return DMA_MEMORY_MAP; + + return DMA_MEMORY_IO; + + free1_out: + kfree(dev->dma_mem->bitmap); + out: + return 0; +} +EXPORT_SYMBOL(dma_declare_coherent_memory); + +void dma_release_declared_memory(struct device *dev) +{ + struct dma_coherent_mem *mem = dev->dma_mem; + + if(!mem) + return; + dev->dma_mem = NULL; + iounmap(mem->virt_base); + kfree(mem->bitmap); + kfree(mem); +} +EXPORT_SYMBOL(dma_release_declared_memory); + +void *dma_mark_declared_memory_occupied(struct device *dev, + dma_addr_t device_addr, size_t size) +{ + struct dma_coherent_mem *mem = dev->dma_mem; + int pages = (size + (device_addr & ~PAGE_MASK) + PAGE_SIZE - 1) >> PAGE_SHIFT; + int pos, err; + + if (!mem) + return ERR_PTR(-EINVAL); + + pos = (device_addr - mem->device_base) >> PAGE_SHIFT; + err = bitmap_allocate_region(mem->bitmap, pos, get_order(pages)); + if (err != 0) + return ERR_PTR(err); + return mem->virt_base + (pos << PAGE_SHIFT); +} +EXPORT_SYMBOL(dma_mark_declared_memory_occupied); diff --git a/arch/cris/arch-v32/drivers/sync_serial.c b/arch/cris/arch-v32/drivers/sync_serial.c new file mode 100644 index 000000000000..c85a6df8558f --- /dev/null +++ b/arch/cris/arch-v32/drivers/sync_serial.c @@ -0,0 +1,1283 @@ +/* + * Simple synchronous serial port driver for ETRAX FS. + * + * Copyright (c) 2005 Axis Communications AB + * + * Author: Mikael Starvik + * + */ + +#include <linux/module.h> +#include <linux/kernel.h> +#include <linux/config.h> +#include <linux/types.h> +#include <linux/errno.h> +#include <linux/major.h> +#include <linux/sched.h> +#include <linux/slab.h> +#include <linux/interrupt.h> +#include <linux/poll.h> +#include <linux/init.h> +#include <linux/timer.h> +#include <linux/spinlock.h> + +#include <asm/io.h> +#include <asm/arch/dma.h> +#include <asm/arch/pinmux.h> +#include <asm/arch/hwregs/reg_rdwr.h> +#include <asm/arch/hwregs/sser_defs.h> +#include <asm/arch/hwregs/dma_defs.h> +#include <asm/arch/hwregs/dma.h> +#include <asm/arch/hwregs/intr_vect_defs.h> +#include <asm/arch/hwregs/intr_vect.h> +#include <asm/arch/hwregs/reg_map.h> +#include <asm/sync_serial.h> + +/* The receiver is a bit tricky beacuse of the continuous stream of data.*/ +/* */ +/* Three DMA descriptors are linked together. Each DMA descriptor is */ +/* responsible for port->bufchunk of a common buffer. */ +/* */ +/* +---------------------------------------------+ */ +/* | +----------+ +----------+ +----------+ | */ +/* +-> | Descr[0] |-->| Descr[1] |-->| Descr[2] |-+ */ +/* +----------+ +----------+ +----------+ */ +/* | | | */ +/* v v v */ +/* +-------------------------------------+ */ +/* | BUFFER | */ +/* +-------------------------------------+ */ +/* |<- data_avail ->| */ +/* readp writep */ +/* */ +/* If the application keeps up the pace readp will be right after writep.*/ +/* If the application can't keep the pace we have to throw away data. */ +/* The idea is that readp should be ready with the data pointed out by */ +/* Descr[i] when the DMA has filled in Descr[i+1]. */ +/* Otherwise we will discard */ +/* the rest of the data pointed out by Descr1 and set readp to the start */ +/* of Descr2 */ + +#define SYNC_SERIAL_MAJOR 125 + +/* IN_BUFFER_SIZE should be a multiple of 6 to make sure that 24 bit */ +/* words can be handled */ +#define IN_BUFFER_SIZE 12288 +#define IN_DESCR_SIZE 256 +#define NUM_IN_DESCR (IN_BUFFER_SIZE/IN_DESCR_SIZE) +#define OUT_BUFFER_SIZE 4096 + +#define DEFAULT_FRAME_RATE 0 +#define DEFAULT_WORD_RATE 7 + +/* NOTE: Enabling some debug will likely cause overrun or underrun, + * especially if manual mode is use. + */ +#define DEBUG(x) +#define DEBUGREAD(x) +#define DEBUGWRITE(x) +#define DEBUGPOLL(x) +#define DEBUGRXINT(x) +#define DEBUGTXINT(x) + +typedef struct sync_port +{ + reg_scope_instances regi_sser; + reg_scope_instances regi_dmain; + reg_scope_instances regi_dmaout; + + char started; /* 1 if port has been started */ + char port_nbr; /* Port 0 or 1 */ + char busy; /* 1 if port is busy */ + + char enabled; /* 1 if port is enabled */ + char use_dma; /* 1 if port uses dma */ + char tr_running; + + char init_irqs; + int output; + int input; + + volatile unsigned int out_count; /* Remaining bytes for current transfer */ + unsigned char* outp; /* Current position in out_buffer */ + volatile unsigned char* volatile readp; /* Next byte to be read by application */ + volatile unsigned char* volatile writep; /* Next byte to be written by etrax */ + unsigned int in_buffer_size; + unsigned int inbufchunk; + unsigned char out_buffer[OUT_BUFFER_SIZE] __attribute__ ((aligned(32))); + unsigned char in_buffer[IN_BUFFER_SIZE]__attribute__ ((aligned(32))); + unsigned char flip[IN_BUFFER_SIZE] __attribute__ ((aligned(32))); + struct dma_descr_data* next_rx_desc; + struct dma_descr_data* prev_rx_desc; + int full; + + dma_descr_data in_descr[NUM_IN_DESCR] __attribute__ ((__aligned__(16))); + dma_descr_context in_context __attribute__ ((__aligned__(32))); + dma_descr_data out_descr __attribute__ ((__aligned__(16))); + dma_descr_context out_context __attribute__ ((__aligned__(32))); + wait_queue_head_t out_wait_q; + wait_queue_head_t in_wait_q; + + spinlock_t lock; +} sync_port; + +static int etrax_sync_serial_init(void); +static void initialize_port(int portnbr); +static inline int sync_data_avail(struct sync_port *port); + +static int sync_serial_open(struct inode *, struct file*); +static int sync_serial_release(struct inode*, struct file*); +static unsigned int sync_serial_poll(struct file *filp, poll_table *wait); + +static int sync_serial_ioctl(struct inode*, struct file*, + unsigned int cmd, unsigned long arg); +static ssize_t sync_serial_write(struct file * file, const char * buf, + size_t count, loff_t *ppos); +static ssize_t sync_serial_read(struct file *file, char *buf, + size_t count, loff_t *ppos); + +#if (defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL_PORT0) && \ + defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL0_DMA)) || \ + (defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL_PORT1) && \ + defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL1_DMA)) +#define SYNC_SER_DMA +#endif + +static void send_word(sync_port* port); +static void start_dma(struct sync_port *port, const char* data, int count); +static void start_dma_in(sync_port* port); +#ifdef SYNC_SER_DMA +static irqreturn_t tr_interrupt(int irq, void *dev_id, struct pt_regs * regs); +static irqreturn_t rx_interrupt(int irq, void *dev_id, struct pt_regs * regs); +#endif + +#if (defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL_PORT0) && \ + !defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL0_DMA)) || \ + (defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL_PORT1) && \ + !defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL1_DMA)) +#define SYNC_SER_MANUAL +#endif +#ifdef SYNC_SER_MANUAL +static irqreturn_t manual_interrupt(int irq, void *dev_id, struct pt_regs * regs); +#endif + +/* The ports */ +static struct sync_port ports[]= +{ + { + .regi_sser = regi_sser0, + .regi_dmaout = regi_dma4, + .regi_dmain = regi_dma5, +#if defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL0_DMA) + .use_dma = 1, +#else + .use_dma = 0, +#endif + }, + { + .regi_sser = regi_sser1, + .regi_dmaout = regi_dma6, + .regi_dmain = regi_dma7, +#if defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL1_DMA) + .use_dma = 1, +#else + .use_dma = 0, +#endif + } +}; + +#define NUMBER_OF_PORTS (sizeof(ports)/sizeof(sync_port)) + +static struct file_operations sync_serial_fops = { + .owner = THIS_MODULE, + .write = sync_serial_write, + .read = sync_serial_read, + .poll = sync_serial_poll, + .ioctl = sync_serial_ioctl, + .open = sync_serial_open, + .release = sync_serial_release +}; + +static int __init etrax_sync_serial_init(void) +{ + ports[0].enabled = 0; + ports[1].enabled = 0; + + if (register_chrdev(SYNC_SERIAL_MAJOR,"sync serial", &sync_serial_fops) <0 ) + { + printk("unable to get major for synchronous serial port\n"); + return -EBUSY; + } + + /* Initialize Ports */ +#if defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL_PORT0) + if (crisv32_pinmux_alloc_fixed(pinmux_sser0)) + { + printk("Unable to allocate pins for syncrhronous serial port 0\n"); + return -EIO; + } + ports[0].enabled = 1; + initialize_port(0); +#endif + +#if defined(CONFIG_ETRAX_SYNCHRONOUS_SERIAL_PORT1) + if (crisv32_pinmux_alloc_fixed(pinmux_sser1)) + { + printk("Unable to allocate pins for syncrhronous serial port 0\n"); + return -EIO; + } + ports[1].enabled = 1; + initialize_port(1); +#endif + + printk("ETRAX FS synchronous serial port driver\n"); + return 0; +} + +static void __init initialize_port(int portnbr) +{ + struct sync_port* port = &ports[portnbr]; + reg_sser_rw_cfg cfg = {0}; + reg_sser_rw_frm_cfg frm_cfg = {0}; + reg_sser_rw_tr_cfg tr_cfg = {0}; + reg_sser_rw_rec_cfg rec_cfg = {0}; + + DEBUG(printk("Init sync serial port %d\n", portnbr)); + + port->port_nbr = portnbr; + port->init_irqs = 1; + + port->outp = port->out_buffer; + port->output = 1; + port->input = 0; + + port->readp = port->flip; + port->writep = port->flip; + port->in_buffer_size = IN_BUFFER_SIZE; + port->inbufchunk = IN_DESCR_SIZE; + port->next_rx_desc = &port->in_descr[0]; + port->prev_rx_desc = &port->in_descr[NUM_IN_DESCR-1]; + port->prev_rx_desc->eol = 1; + + init_waitqueue_head(&port->out_wait_q); + init_waitqueue_head(&port->in_wait_q); + + spin_lock_init(&port->lock); + + cfg.out_clk_src = regk_sser_intern_clk; + cfg.out_clk_pol = regk_sser_pos; + cfg.clk_od_mode = regk_sser_no; + cfg.clk_dir = regk_sser_out; + cfg.gate_clk = regk_sser_no; + cfg.base_freq = regk_sser_f29_493; + cfg.clk_div = 256; + REG_WR(sser, port->regi_sser, rw_cfg, cfg); + + frm_cfg.wordrate = DEFAULT_WORD_RATE; + frm_cfg.type = regk_sser_edge; + frm_cfg.frame_pin_dir = regk_sser_out; + frm_cfg.frame_pin_use = regk_sser_frm; + frm_cfg.status_pin_dir = regk_sser_in; + frm_cfg.status_pin_use = regk_sser_hold; + frm_cfg.out_on = regk_sser_tr; + frm_cfg.tr_delay = 1; + REG_WR(sser, port->regi_sser, rw_frm_cfg, frm_cfg); + + tr_cfg.urun_stop = regk_sser_no; + tr_cfg.sample_size = 7; + tr_cfg.sh_dir = regk_sser_msbfirst; + tr_cfg.use_dma = port->use_dma ? regk_sser_yes : regk_sser_no; + tr_cfg.rate_ctrl = regk_sser_bulk; + tr_cfg.data_pin_use = regk_sser_dout; + tr_cfg.bulk_wspace = 1; + REG_WR(sser, port->regi_sser, rw_tr_cfg, tr_cfg); + + rec_cfg.sample_size = 7; + rec_cfg.sh_dir = regk_sser_msbfirst; + rec_cfg.use_dma = port->use_dma ? regk_sser_yes : regk_sser_no; + rec_cfg.fifo_thr = regk_sser_inf; + REG_WR(sser, port->regi_sser, rw_rec_cfg, rec_cfg); +} + +static inline int sync_data_avail(struct sync_port *port) +{ + int avail; + unsigned char *start; + unsigned char *end; + + start = (unsigned char*)port->readp; /* cast away volatile */ + end = (unsigned char*)port->writep; /* cast away volatile */ + /* 0123456789 0123456789 + * ----- - ----- + * ^rp ^wp ^wp ^rp + */ + + if (end >= start) + avail = end - start; + else + avail = port->in_buffer_size - (start - end); + return avail; +} + +static inline int sync_data_avail_to_end(struct sync_port *port) +{ + int avail; + unsigned char *start; + unsigned char *end; + + start = (unsigned char*)port->readp; /* cast away volatile */ + end = (unsigned char*)port->writep; /* cast away volatile */ + /* 0123456789 0123456789 + * ----- ----- + * ^rp ^wp ^wp ^rp + */ + + if (end >= start) + avail = end - start; + else + avail = port->flip + port->in_buffer_size - start; + return avail; +} + +static int sync_serial_open(struct inode *inode, struct file *file) +{ + int dev = MINOR(inode->i_rdev); + sync_port* port; + reg_dma_rw_cfg cfg = {.en = regk_dma_yes}; + reg_dma_rw_intr_mask intr_mask = {.data = regk_dma_yes}; + + DEBUG(printk("Open sync serial port %d\n", dev)); + + if (dev < 0 || dev >= NUMBER_OF_PORTS || !ports[dev].enabled) + { + DEBUG(printk("Invalid minor %d\n", dev)); + return -ENODEV; + } + port = &ports[dev]; + /* Allow open this device twice (assuming one reader and one writer) */ + if (port->busy == 2) + { + DEBUG(printk("Device is busy.. \n")); + return -EBUSY; + } + if (port->init_irqs) { + if (port->use_dma) { + if (port == &ports[0]){ +#ifdef SYNC_SER_DMA + if(request_irq(DMA4_INTR_VECT, + tr_interrupt, + 0, + "synchronous serial 0 dma tr", + &ports[0])) { + printk(KERN_CRIT "Can't allocate sync serial port 0 IRQ"); + return -EBUSY; + } else if(request_irq(DMA5_INTR_VECT, + rx_interrupt, + 0, + "synchronous serial 1 dma rx", + &ports[0])) { + free_irq(DMA4_INTR_VECT, &port[0]); + printk(KERN_CRIT "Can't allocate sync serial port 0 IRQ"); + return -EBUSY; + } else if (crisv32_request_dma(SYNC_SER0_TX_DMA_NBR, + "synchronous serial 0 dma tr", + DMA_VERBOSE_ON_ERROR, + 0, + dma_sser0)) { + free_irq(DMA4_INTR_VECT, &port[0]); + free_irq(DMA5_INTR_VECT, &port[0]); + printk(KERN_CRIT "Can't allocate sync serial port 0 TX DMA channel"); + return -EBUSY; + } else if (crisv32_request_dma(SYNC_SER0_RX_DMA_NBR, + "synchronous serial 0 dma rec", + DMA_VERBOSE_ON_ERROR, + 0, + dma_sser0)) { + crisv32_free_dma(SYNC_SER0_TX_DMA_NBR); + free_irq(DMA4_INTR_VECT, &port[0]); + free_irq(DMA5_INTR_VECT, &port[0]); + printk(KERN_CRIT "Can't allocate sync serial port 1 RX DMA channel"); + return -EBUSY; + } +#endif + } + else if (port == &ports[1]){ +#ifdef SYNC_SER_DMA + if (request_irq(DMA6_INTR_VECT, + tr_interrupt, + 0, + "synchronous serial 1 dma tr", + &ports[1])) { + printk(KERN_CRIT "Can't allocate sync serial port 1 IRQ"); + return -EBUSY; + } else if (request_irq(DMA7_INTR_VECT, + rx_interrupt, + 0, + "synchronous serial 1 dma rx", + &ports[1])) { + free_irq(DMA6_INTR_VECT, &ports[1]); + printk(KERN_CRIT "Can't allocate sync serial port 3 IRQ"); + return -EBUSY; + } else if (crisv32_request_dma(SYNC_SER1_TX_DMA_NBR, + "synchronous serial 1 dma tr", + DMA_VERBOSE_ON_ERROR, + 0, + dma_sser1)) { + free_irq(21, &ports[1]); + free_irq(20, &ports[1]); + printk(KERN_CRIT "Can't allocate sync serial port 3 TX DMA channel"); + return -EBUSY; + } else if (crisv32_request_dma(SYNC_SER1_RX_DMA_NBR, + "synchronous serial 3 dma rec", + DMA_VERBOSE_ON_ERROR, + 0, + dma_sser1)) { + crisv32_free_dma(SYNC_SER1_TX_DMA_NBR); + free_irq(DMA6_INTR_VECT, &ports[1]); + free_irq(DMA7_INTR_VECT, &ports[1]); + printk(KERN_CRIT "Can't allocate sync serial port 3 RX DMA channel"); + return -EBUSY; + } +#endif + } + + /* Enable DMAs */ + REG_WR(dma, port->regi_dmain, rw_cfg, cfg); + REG_WR(dma, port->regi_dmaout, rw_cfg, cfg); + /* Enable DMA IRQs */ + REG_WR(dma, port->regi_dmain, rw_intr_mask, intr_mask); + REG_WR(dma, port->regi_dmaout, rw_intr_mask, intr_mask); + /* Set up wordsize = 2 for DMAs. */ + DMA_WR_CMD (port->regi_dmain, regk_dma_set_w_size1); + DMA_WR_CMD (port->regi_dmaout, regk_dma_set_w_size1); + + start_dma_in(port); + port->init_irqs = 0; + } else { /* !port->use_dma */ +#ifdef SYNC_SER_MANUAL + if (port == &ports[0]) { + if (request_irq(SSER0_INTR_VECT, + manual_interrupt, + 0, + "synchronous serial manual irq", + &ports[0])) { + printk("Can't allocate sync serial manual irq"); + return -EBUSY; + } + } else if (port == &ports[1]) { + if (request_irq(SSER1_INTR_VECT, + manual_interrupt, + 0, + "synchronous serial manual irq", + &ports[1])) { + printk(KERN_CRIT "Can't allocate sync serial manual irq"); + return -EBUSY; + } + } + port->init_irqs = 0; +#else + panic("sync_serial: Manual mode not supported.\n"); +#endif /* SYNC_SER_MANUAL */ + } + } /* port->init_irqs */ + + port->busy++; + return 0; +} + +static int sync_serial_release(struct inode *inode, struct file *file) +{ + int dev = MINOR(inode->i_rdev); + sync_port* port; + + if (dev < 0 || dev >= NUMBER_OF_PORTS || !ports[dev].enabled) + { + DEBUG(printk("Invalid minor %d\n", dev)); + return -ENODEV; + } + port = &ports[dev]; + if (port->busy) + port->busy--; + if (!port->busy) + /* XXX */ ; + return 0; +} + +static unsigned int sync_serial_poll(struct file *file, poll_table *wait) +{ + int dev = MINOR(file->f_dentry->d_inode->i_rdev); + unsigned int mask = 0; + sync_port* port; + DEBUGPOLL( static unsigned int prev_mask = 0; ); + + port = &ports[dev]; + poll_wait(file, &port->out_wait_q, wait); + poll_wait(file, &port->in_wait_q, wait); + /* Some room to write */ + if (port->out_count < OUT_BUFFER_SIZE) + mask |= POLLOUT | POLLWRNORM; + /* At least an inbufchunk of data */ + if (sync_data_avail(port) >= port->inbufchunk) + mask |= POLLIN | POLLRDNORM; + + DEBUGPOLL(if (mask != prev_mask) + printk("sync_serial_poll: mask 0x%08X %s %s\n", mask, + mask&POLLOUT?"POLLOUT":"", mask&POLLIN?"POLLIN":""); + prev_mask = mask; + ); + return mask; +} + +static int sync_serial_ioctl(struct inode *inode, struct file *file, + unsigned int cmd, unsigned long arg) +{ + int return_val = 0; + int dev = MINOR(file->f_dentry->d_inode->i_rdev); + sync_port* port; + reg_sser_rw_tr_cfg tr_cfg; + reg_sser_rw_rec_cfg rec_cfg; + reg_sser_rw_frm_cfg frm_cfg; + reg_sser_rw_cfg gen_cfg; + reg_sser_rw_intr_mask intr_mask; + + if (dev < 0 || dev >= NUMBER_OF_PORTS || !ports[dev].enabled) + { + DEBUG(printk("Invalid minor %d\n", dev)); + return -1; + } + port = &ports[dev]; + spin_lock_irq(&port->lock); + + tr_cfg = REG_RD(sser, port->regi_sser, rw_tr_cfg); + rec_cfg = REG_RD(sser, port->regi_sser, rw_rec_cfg); + frm_cfg = REG_RD(sser, port->regi_sser, rw_frm_cfg); + gen_cfg = REG_RD(sser, port->regi_sser, rw_cfg); + intr_mask = REG_RD(sser, port->regi_sser, rw_intr_mask); + + switch(cmd) + { + case SSP_SPEED: + if (GET_SPEED(arg) == CODEC) + { + gen_cfg.base_freq = regk_sser_f32; + /* FREQ = 0 => 4 MHz => clk_div = 7*/ + gen_cfg.clk_div = 6 + (1 << GET_FREQ(arg)); + } + else + { + gen_cfg.base_freq = regk_sser_f29_493; + switch (GET_SPEED(arg)) + { + case SSP150: + gen_cfg.clk_div = 29493000 / (150 * 8) - 1; + break; + case SSP300: + gen_cfg.clk_div = 29493000 / (300 * 8) - 1; + break; + case SSP600: + gen_cfg.clk_div = 29493000 / (600 * 8) - 1; + break; + case SSP1200: + gen_cfg.clk_div = 29493000 / (1200 * 8) - 1; + break; + case SSP2400: + gen_cfg.clk_div = 29493000 / (2400 * 8) - 1; + break; + case SSP4800: + gen_cfg.clk_div = 29493000 / (4800 * 8) - 1; + break; + case SSP9600: + gen_cfg.clk_div = 29493000 / (9600 * 8) - 1; + break; + case SSP19200: + gen_cfg.clk_div = 29493000 / (19200 * 8) - 1; + break; + case SSP28800: + gen_cfg.clk_div = 29493000 / (28800 * 8) - 1; + break; + case SSP57600: + gen_cfg.clk_div = 29493000 / (57600 * 8) - 1; + break; + case SSP115200: + gen_cfg.clk_div = 29493000 / (115200 * 8) - 1; + break; + case SSP230400: + gen_cfg.clk_div = 29493000 / (230400 * 8) - 1; + break; + case SSP460800: + gen_cfg.clk_div = 29493000 / (460800 * 8) - 1; + break; + case SSP921600: + gen_cfg.clk_div = 29493000 / (921600 * 8) - 1; + break; + case SSP3125000: + gen_cfg.base_freq = regk_sser_f100; + gen_cfg.clk_div = 100000000 / (3125000 * 8) - 1; + break; + + } + } + frm_cfg.wordrate = GET_WORD_RATE(arg); + + break; + case SSP_MODE: + switch(arg) + { + case MASTER_OUTPUT: + port->output = 1; + port->input = 0; + gen_cfg.clk_dir = regk_sser_out; + break; + case SLAVE_OUTPUT: + port->output = 1; + port->input = 0; + gen_cfg.clk_dir = regk_sser_in; + break; + case MASTER_INPUT: + port->output = 0; + port->input = 1; + gen_cfg.clk_dir = regk_sser_out; + break; + case SLAVE_INPUT: + port->output = 0; + port->input = 1; + gen_cfg.clk_dir = regk_sser_in; + break; + case MASTER_BIDIR: + port->output = 1; + port->input = 1; + gen_cfg.clk_dir = regk_sser_out; + break; + case SLAVE_BIDIR: + port->output = 1; + port->input = 1; + gen_cfg.clk_dir = regk_sser_in; + break; + default: + spin_unlock_irq(&port->lock); + return -EINVAL; + + } + if (!port->use_dma || (arg == MASTER_OUTPUT || arg == SLAVE_OUTPUT)) + intr_mask.rdav = regk_sser_yes; + break; + case SSP_FRAME_SYNC: + if (arg & NORMAL_SYNC) + frm_cfg.tr_delay = 1; + else if (arg & EARLY_SYNC) + frm_cfg.tr_delay = 0; + + tr_cfg.bulk_wspace = frm_cfg.tr_delay; + frm_cfg.early_wend = regk_sser_yes; + if (arg & BIT_SYNC) + frm_cfg.type = regk_sser_edge; + else if (arg & WORD_SYNC) + frm_cfg.type = regk_sser_level; + else if (arg & EXTENDED_SYNC) + frm_cfg.early_wend = regk_sser_no; + + if (arg & SYNC_ON) + frm_cfg.frame_pin_use = regk_sser_frm; + else if (arg & SYNC_OFF) + frm_cfg.frame_pin_use = regk_sser_gio0; + + if (arg & WORD_SIZE_8) + rec_cfg.sample_size = tr_cfg.sample_size = 7; + else if (arg & WORD_SIZE_12) + rec_cfg.sample_size = tr_cfg.sample_size = 11; + else if (arg & WORD_SIZE_16) + rec_cfg.sample_size = tr_cfg.sample_size = 15; + else if (arg & WORD_SIZE_24) + rec_cfg.sample_size = tr_cfg.sample_size = 23; + else if (arg & WORD_SIZE_32) + rec_cfg.sample_size = tr_cfg.sample_size = 31; + + if (arg & BIT_ORDER_MSB) + rec_cfg.sh_dir = tr_cfg.sh_dir = regk_sser_msbfirst; + else if (arg & BIT_ORDER_LSB) + rec_cfg.sh_dir = tr_cfg.sh_dir = regk_sser_lsbfirst; + + if (arg & FLOW_CONTROL_ENABLE) + rec_cfg.fifo_thr = regk_sser_thr16; + else if (arg & FLOW_CONTROL_DISABLE) + rec_cfg.fifo_thr = regk_sser_inf; + + if (arg & CLOCK_NOT_GATED) + gen_cfg.gate_clk = regk_sser_no; + else if (arg & CLOCK_GATED) + gen_cfg.gate_clk = regk_sser_yes; + + break; + case SSP_IPOLARITY: + /* NOTE!! negedge is considered NORMAL */ + if (arg & CLOCK_NORMAL) + rec_cfg.clk_pol = regk_sser_neg; + else if (arg & CLOCK_INVERT) + rec_cfg.clk_pol = regk_sser_pos; + + if (arg & FRAME_NORMAL) + frm_cfg.level = regk_sser_pos_hi; + else if (arg & FRAME_INVERT) + frm_cfg.level = regk_sser_neg_lo; + + if (arg & STATUS_NORMAL) + gen_cfg.hold_pol = regk_sser_pos; + else if (arg & STATUS_INVERT) + gen_cfg.hold_pol = regk_sser_neg; + break; + case SSP_OPOLARITY: + if (arg & CLOCK_NORMAL) + gen_cfg.out_clk_pol = regk_sser_neg; + else if (arg & CLOCK_INVERT) + gen_cfg.out_clk_pol = regk_sser_pos; + + if (arg & FRAME_NORMAL) + frm_cfg.level = regk_sser_pos_hi; + else if (arg & FRAME_INVERT) + frm_cfg.level = regk_sser_neg_lo; + + if (arg & STATUS_NORMAL) + gen_cfg.hold_pol = regk_sser_pos; + else if (arg & STATUS_INVERT) + gen_cfg.hold_pol = regk_sser_neg; + break; + case SSP_SPI: + rec_cfg.fifo_thr = regk_sser_inf; + rec_cfg.sh_dir = tr_cfg.sh_dir = regk_sser_msbfirst; + rec_cfg.sample_size = tr_cfg.sample_size = 7; + frm_cfg.frame_pin_use = regk_sser_frm; + frm_cfg.type = regk_sser_level; + frm_cfg.tr_delay = 1; + frm_cfg.level = regk_sser_neg_lo; + if (arg & SPI_SLAVE) + { + rec_cfg.clk_pol = regk_sser_neg; + gen_cfg.clk_dir = regk_sser_in; + port->input = 1; + port->output = 0; + } + else + { + gen_cfg.out_clk_pol = regk_sser_pos; + port->input = 0; + port->output = 1; + gen_cfg.clk_dir = regk_sser_out; + } + break; + case SSP_INBUFCHUNK: + break; + default: + return_val = -1; + } + + + if (port->started) + { + tr_cfg.tr_en = port->output; + rec_cfg.rec_en = port->input; + } + + REG_WR(sser, port->regi_sser, rw_tr_cfg, tr_cfg); + REG_WR(sser, port->regi_sser, rw_rec_cfg, rec_cfg); + REG_WR(sser, port->regi_sser, rw_frm_cfg, frm_cfg); + REG_WR(sser, port->regi_sser, rw_intr_mask, intr_mask); + REG_WR(sser, port->regi_sser, rw_cfg, gen_cfg); + + spin_unlock_irq(&port->lock); + return return_val; +} + +static ssize_t sync_serial_write(struct file * file, const char * buf, + size_t count, loff_t *ppos) +{ + int dev = MINOR(file->f_dentry->d_inode->i_rdev); + DECLARE_WAITQUEUE(wait, current); + sync_port *port; + unsigned long c, c1; + unsigned long free_outp; + unsigned long outp; + unsigned long out_buffer; + unsigned long flags; + + if (dev < 0 || dev >= NUMBER_OF_PORTS || !ports[dev].enabled) + { + DEBUG(printk("Invalid minor %d\n", dev)); + return -ENODEV; + } + port = &ports[dev]; + + DEBUGWRITE(printk("W d%d c %lu (%d/%d)\n", port->port_nbr, count, port->out_count, OUT_BUFFER_SIZE)); + /* Space to end of buffer */ + /* + * out_buffer <c1>012345<- c ->OUT_BUFFER_SIZE + * outp^ +out_count + ^free_outp + * out_buffer 45<- c ->0123OUT_BUFFER_SIZE + * +out_count outp^ + * free_outp + * + */ + + /* Read variables that may be updated by interrupts */ + spin_lock_irqsave(&port->lock, flags); + count = count > OUT_BUFFER_SIZE - port->out_count ? OUT_BUFFER_SIZE - port->out_count : count; + outp = (unsigned long)port->outp; + free_outp = outp + port->out_count; + spin_unlock_irqrestore(&port->lock, flags); + out_buffer = (unsigned long)port->out_buffer; + + /* Find out where and how much to write */ + if (free_outp >= out_buffer + OUT_BUFFER_SIZE) + free_outp -= OUT_BUFFER_SIZE; + if (free_outp >= outp) + c = out_buffer + OUT_BUFFER_SIZE - free_outp; + else + c = outp - free_outp; + if (c > count) + c = count; + +// DEBUGWRITE(printk("w op %08lX fop %08lX c %lu\n", outp, free_outp, c)); + if (copy_from_user((void*)free_outp, buf, c)) + return -EFAULT; + + if (c != count) { + buf += c; + c1 = count - c; + DEBUGWRITE(printk("w2 fi %lu c %lu c1 %lu\n", free_outp-out_buffer, c, c1)); + if (copy_from_user((void*)out_buffer, buf, c1)) + return -EFAULT; + } + spin_lock_irqsave(&port->lock, flags); + port->out_count += count; + spin_unlock_irqrestore(&port->lock, flags); + + /* Make sure transmitter/receiver is running */ + if (!port->started) + { + reg_sser_rw_cfg cfg = REG_RD(sser, port->regi_sser, rw_cfg); + reg_sser_rw_tr_cfg tr_cfg = REG_RD(sser, port->regi_sser, rw_tr_cfg); + reg_sser_rw_rec_cfg rec_cfg = REG_RD(sser, port->regi_sser, rw_rec_cfg); + cfg.en = regk_sser_yes; + tr_cfg.tr_en = port->output; + rec_cfg.rec_en = port->input; + REG_WR(sser, port->regi_sser, rw_cfg, cfg); + REG_WR(sser, port->regi_sser, rw_tr_cfg, tr_cfg); + REG_WR(sser, port->regi_sser, rw_rec_cfg, rec_cfg); + port->started = 1; + } + + if (file->f_flags & O_NONBLOCK) { + spin_lock_irqsave(&port->lock, flags); + if (!port->tr_running) { + if (!port->use_dma) { + reg_sser_rw_intr_mask intr_mask; + intr_mask = REG_RD(sser, port->regi_sser, rw_intr_mask); + /* Start sender by writing data */ + send_word(port); + /* and enable transmitter ready IRQ */ + intr_mask.trdy = 1; + REG_WR(sser, port->regi_sser, rw_intr_mask, intr_mask); + } else { + start_dma(port, (unsigned char* volatile )port->outp, c); + } + } + spin_unlock_irqrestore(&port->lock, flags); + DEBUGWRITE(printk("w d%d c %lu NB\n", + port->port_nbr, count)); + return count; + } + + /* Sleep until all sent */ + + add_wait_queue(&port->out_wait_q, &wait); + set_current_state(TASK_INTERRUPTIBLE); + spin_lock_irqsave(&port->lock, flags); + if (!port->tr_running) { + if (!port->use_dma) { + reg_sser_rw_intr_mask intr_mask; + intr_mask = REG_RD(sser, port->regi_sser, rw_intr_mask); + /* Start sender by writing data */ + send_word(port); + /* and enable transmitter ready IRQ */ + intr_mask.trdy = 1; + REG_WR(sser, port->regi_sser, rw_intr_mask, intr_mask); + } else { + start_dma(port, port->outp, c); + } + } + spin_unlock_irqrestore(&port->lock, flags); + schedule(); + set_current_state(TASK_RUNNING); + remove_wait_queue(&port->out_wait_q, &wait); + if (signal_pending(current)) + { + return -EINTR; + } + DEBUGWRITE(printk("w d%d c %lu\n", port->port_nbr, count)); + return count; +} + +static ssize_t sync_serial_read(struct file * file, char * buf, + size_t count, loff_t *ppos) +{ + int dev = MINOR(file->f_dentry->d_inode->i_rdev); + int avail; + sync_port *port; + unsigned char* start; + unsigned char* end; + unsigned long flags; + + if (dev < 0 || dev >= NUMBER_OF_PORTS || !ports[dev].enabled) + { + DEBUG(printk("Invalid minor %d\n", dev)); + return -ENODEV; + } + port = &ports[dev]; + + DEBUGREAD(printk("R%d c %d ri %lu wi %lu /%lu\n", dev, count, port->readp - port->flip, port->writep - port->flip, port->in_buffer_size)); + + if (!port->started) + { + reg_sser_rw_cfg cfg = REG_RD(sser, port->regi_sser, rw_cfg); + reg_sser_rw_tr_cfg tr_cfg = REG_RD(sser, port->regi_sser, rw_tr_cfg); + reg_sser_rw_rec_cfg rec_cfg = REG_RD(sser, port->regi_sser, rw_rec_cfg); + cfg.en = regk_sser_yes; + tr_cfg.tr_en = regk_sser_yes; + rec_cfg.rec_en = regk_sser_yes; + REG_WR(sser, port->regi_sser, rw_cfg, cfg); + REG_WR(sser, port->regi_sser, rw_tr_cfg, tr_cfg); + REG_WR(sser, port->regi_sser, rw_rec_cfg, rec_cfg); + port->started = 1; + } + + + /* Calculate number of available bytes */ + /* Save pointers to avoid that they are modified by interrupt */ + spin_lock_irqsave(&port->lock, flags); + start = (unsigned char*)port->readp; /* cast away volatile */ + end = (unsigned char*)port->writep; /* cast away volatile */ + spin_unlock_irqrestore(&port->lock, flags); + while ((start == end) && !port->full) /* No data */ + { + if (file->f_flags & O_NONBLOCK) + { + return -EAGAIN; + } + + interruptible_sleep_on(&port->in_wait_q); + if (signal_pending(current)) + { + return -EINTR; + } + spin_lock_irqsave(&port->lock, flags); + start = (unsigned char*)port->readp; /* cast away volatile */ + end = (unsigned char*)port->writep; /* cast away volatile */ + spin_unlock_irqrestore(&port->lock, flags); + } + + /* Lazy read, never return wrapped data. */ + if (port->full) + avail = port->in_buffer_size; + else if (end > start) + avail = end - start; + else + avail = port->flip + port->in_buffer_size - start; + + count = count > avail ? avail : count; + if (copy_to_user(buf, start, count)) + return -EFAULT; + /* Disable interrupts while updating readp */ + spin_lock_irqsave(&port->lock, flags); + port->readp += count; + if (port->readp >= port->flip + port->in_buffer_size) /* Wrap? */ + port->readp = port->flip; + port->full = 0; + spin_unlock_irqrestore(&port->lock, flags); + DEBUGREAD(printk("r %d\n", count)); + return count; +} + +static void send_word(sync_port* port) +{ + reg_sser_rw_tr_cfg tr_cfg = REG_RD(sser, port->regi_sser, rw_tr_cfg); + reg_sser_rw_tr_data tr_data = {0}; + + switch(tr_cfg.sample_size) + { + case 8: + port->out_count--; + tr_data.data = *port->outp++; + REG_WR(sser, port->regi_sser, rw_tr_data, tr_data); + if (port->outp >= port->out_buffer + OUT_BUFFER_SIZE) + port->outp = port->out_buffer; + break; + case 12: + { + int data = (*port->outp++) << 8; + data |= *port->outp++; + port->out_count-=2; + tr_data.data = data; + REG_WR(sser, port->regi_sser, rw_tr_data, tr_data); + if (port->outp >= port->out_buffer + OUT_BUFFER_SIZE) + port->outp = port->out_buffer; + } + break; + case 16: + port->out_count-=2; + tr_data.data = *(unsigned short *)port->outp; + REG_WR(sser, port->regi_sser, rw_tr_data, tr_data); + port->outp+=2; + if (port->outp >= port->out_buffer + OUT_BUFFER_SIZE) + port->outp = port->out_buffer; + break; + case 24: + port->out_count-=3; + tr_data.data = *(unsigned short *)port->outp; + REG_WR(sser, port->regi_sser, rw_tr_data, tr_data); + port->outp+=2; + tr_data.data = *port->outp++; + REG_WR(sser, port->regi_sser, rw_tr_data, tr_data); + if (port->outp >= port->out_buffer + OUT_BUFFER_SIZE) + port->outp = port->out_buffer; + break; + case 32: + port->out_count-=4; + tr_data.data = *(unsigned short *)port->outp; + REG_WR(sser, port->regi_sser, rw_tr_data, tr_data); + port->outp+=2; + tr_data.data = *(unsigned short *)port->outp; + REG_WR(sser, port->regi_sser, rw_tr_data, tr_data); + port->outp+=2; + if (port->outp >= port->out_buffer + OUT_BUFFER_SIZE) + port->outp = port->out_buffer; + break; + } +} + + +static void start_dma(struct sync_port* port, const char* data, int count) +{ + port->tr_running = 1; + port->out_descr.buf = (char*)virt_to_phys((char*)data); + port->out_descr.after = port->out_descr.buf + count; + port->out_descr.eol = port->out_descr.intr = 1; + + port->out_context.saved_data = (dma_descr_data*)virt_to_phys(&port->out_descr); + port->out_context.saved_data_buf = port->out_descr.buf; + + DMA_START_CONTEXT(port->regi_dmaout, virt_to_phys((char*)&port->out_context)); + DEBUGTXINT(printk("dma %08lX c %d\n", (unsigned long)data, count)); +} + +static void start_dma_in(sync_port* port) +{ + int i; + char* buf; + port->writep = port->flip; + + if (port->writep > port->flip + port->in_buffer_size) + { + panic("Offset too large in sync serial driver\n"); + return; + } + buf = (char*)virt_to_phys(port->in_buffer); + for (i = 0; i < NUM_IN_DESCR; i++) { + port->in_descr[i].buf = buf; + port->in_descr[i].after = buf + port->inbufchunk; + port->in_descr[i].intr = 1; + port->in_descr[i].next = (dma_descr_data*)virt_to_phys(&port->in_descr[i+1]); + port->in_descr[i].buf = buf; + buf += port->inbufchunk; + } + /* Link the last descriptor to the first */ + port->in_descr[i-1].next = (dma_descr_data*)virt_to_phys(&port->in_descr[0]); + port->in_descr[i-1].eol = regk_sser_yes; + port->next_rx_desc = &port->in_descr[0]; + port->prev_rx_desc = &port->in_descr[NUM_IN_DESCR - 1]; + port->in_context.saved_data = (dma_descr_data*)virt_to_phys(&port->in_descr[0]); + port->in_context.saved_data_buf = port->in_descr[0].buf; + DMA_START_CONTEXT(port->regi_dmain, virt_to_phys(&port->in_context)); +} + +#ifdef SYNC_SER_DMA +static irqreturn_t tr_interrupt(int irq, void *dev_id, struct pt_regs * regs) +{ + reg_dma_r_masked_intr masked; + reg_dma_rw_ack_intr ack_intr = {.data = regk_dma_yes}; + int i; + struct dma_descr_data *descr; + unsigned int sentl; + int found = 0; + + for (i = 0; i < NUMBER_OF_PORTS; i++) + { + sync_port *port = &ports[i]; + if (!port->enabled || !port->use_dma ) + continue; + + masked = REG_RD(dma, port->regi_dmaout, r_masked_intr); + + if (masked.data) /* IRQ active for the port? */ + { + found = 1; + /* Clear IRQ */ + REG_WR(dma, port->regi_dmaout, rw_ack_intr, ack_intr); + descr = &port->out_descr; + sentl = descr->after - descr->buf; + port->out_count -= sentl; + port->outp += sentl; + if (port->outp >= port->out_buffer + OUT_BUFFER_SIZE) + port->outp = port->out_buffer; + if (port->out_count) { + int c; + c = port->out_buffer + OUT_BUFFER_SIZE - port->outp; + if (c > port->out_count) + c = port->out_count; + DEBUGTXINT(printk("tx_int DMAWRITE %i %i\n", sentl, c)); + start_dma(port, port->outp, c); + } else { + DEBUGTXINT(printk("tx_int DMA stop %i\n", sentl)); + port->tr_running = 0; + } + wake_up_interruptible(&port->out_wait_q); /* wake up the waiting process */ + } + } + return IRQ_RETVAL(found); +} /* tr_interrupt */ + +static irqreturn_t rx_interrupt(int irq, void *dev_id, struct pt_regs * regs) +{ + reg_dma_r_masked_intr masked; + reg_dma_rw_ack_intr ack_intr = {.data = regk_dma_yes}; + + int i; + int found = 0; + + for (i = 0; i < NUMBER_OF_PORTS; i++) + { + sync_port *port = &ports[i]; + + if (!port->enabled || !port->use_dma ) + continue; + + masked = REG_RD(dma, port->regi_dmain, r_masked_intr); + + if (masked.data) /* Descriptor interrupt */ + { + found = 1; + while (REG_RD(dma, port->regi_dmain, rw_data) != + virt_to_phys(port->next_rx_desc)) { + + if (port->writep + port->inbufchunk > port->flip + port->in_buffer_size) { + int first_size = port->flip + port->in_buffer_size - port->writep; + memcpy((char*)port->writep, phys_to_virt((unsigned)port->next_rx_desc->buf), first_size); + memcpy(port->flip, phys_to_virt((unsigned)port->next_rx_desc->buf+first_size), port->inbufchunk - first_size); + port->writep = port->flip + port->inbufchunk - first_size; + } else { + memcpy((char*)port->writep, + phys_to_virt((unsigned)port->next_rx_desc->buf), + port->inbufchunk); + port->writep += port->inbufchunk; + if (port->writep >= port->flip + port->in_buffer_size) + port->writep = port->flip; + } + if (port->writep == port->readp) + { + port->full = 1; + } + + port->next_rx_desc->eol = 0; + port->prev_rx_desc->eol = 1; + port->prev_rx_desc = phys_to_virt((unsigned)port->next_rx_desc); + port->next_rx_desc = phys_to_virt((unsigned)port->next_rx_desc->next); + wake_up_interruptible(&port->in_wait_q); /* wake up the waiting process */ + DMA_CONTINUE(port->regi_dmain); + REG_WR(dma, port->regi_dmain, rw_ack_intr, ack_intr); + + } + } + } + return IRQ_RETVAL(found); +} /* rx_interrupt */ +#endif /* SYNC_SER_DMA */ + +#ifdef SYNC_SER_MANUAL +static irqreturn_t manual_interrupt(int irq, void *dev_id, struct pt_regs * regs) +{ + int i; + int found = 0; + reg_sser_r_masked_intr masked; + + for (i = 0; i < NUMBER_OF_PORTS; i++) + { + sync_port* port = &ports[i]; + + if (!port->enabled || port->use_dma) + { + continue; + } + + masked = REG_RD(sser, port->regi_sser, r_masked_intr); + if (masked.rdav) /* Data received? */ + { + reg_sser_rw_rec_cfg rec_cfg = REG_RD(sser, port->regi_sser, rw_rec_cfg); + reg_sser_r_rec_data data = REG_RD(sser, port->regi_sser, r_rec_data); + found = 1; + /* Read data */ + switch(rec_cfg.sample_size) + { + case 8: + *port->writep++ = data.data & 0xff; + break; + case 12: + *port->writep = (data.data & 0x0ff0) >> 4; + *(port->writep + 1) = data.data & 0x0f; + port->writep+=2; + break; + case 16: + *(unsigned short*)port->writep = data.data; + port->writep+=2; + break; + case 24: + *(unsigned int*)port->writep = data.data; + port->writep+=3; + break; + case 32: + *(unsigned int*)port->writep = data.data; + port->writep+=4; + break; + } + + if (port->writep >= port->flip + port->in_buffer_size) /* Wrap? */ + port->writep = port->flip; + if (port->writep == port->readp) { + /* receive buffer overrun, discard oldest data + */ + port->readp++; + if (port->readp >= port->flip + port->in_buffer_size) /* Wrap? */ + port->readp = port->flip; + } + if (sync_data_avail(port) >= port->inbufchunk) + wake_up_interruptible(&port->in_wait_q); /* Wake up application */ + } + + if (masked.trdy) /* Transmitter ready? */ + { + found = 1; + if (port->out_count > 0) /* More data to send */ + send_word(port); + else /* transmission finished */ + { + reg_sser_rw_intr_mask intr_mask; + intr_mask = REG_RD(sser, port->regi_sser, rw_intr_mask); + intr_mask.trdy = 0; + REG_WR(sser, port->regi_sser, rw_intr_mask, intr_mask); + wake_up_interruptible(&port->out_wait_q); /* Wake up application */ + } + } + } + return IRQ_RETVAL(found); +} +#endif + +module_init(etrax_sync_serial_init); diff --git a/arch/cris/arch-v32/kernel/Makefile b/arch/cris/arch-v32/kernel/Makefile new file mode 100644 index 000000000000..5d5b613cde8c --- /dev/null +++ b/arch/cris/arch-v32/kernel/Makefile @@ -0,0 +1,21 @@ +# $Id: Makefile,v 1.11 2004/12/17 10:16:13 starvik Exp $ +# +# Makefile for the linux kernel. +# + +extra-y := head.o + + +obj-y := entry.o traps.o irq.o debugport.o dma.o pinmux.o \ + process.o ptrace.o setup.o signal.o traps.o time.o \ + arbiter.o io.o + +obj-$(CONFIG_ETRAXFS_SIM) += vcs_hook.o + +obj-$(CONFIG_SMP) += smp.o +obj-$(CONFIG_ETRAX_KGDB) += kgdb.o kgdb_asm.o +obj-$(CONFIG_ETRAX_FAST_TIMER) += fasttimer.o +obj-$(CONFIG_MODULES) += crisksyms.o + +clean: + diff --git a/arch/cris/arch-v32/kernel/arbiter.c b/arch/cris/arch-v32/kernel/arbiter.c new file mode 100644 index 000000000000..3870d2fd5160 --- /dev/null +++ b/arch/cris/arch-v32/kernel/arbiter.c @@ -0,0 +1,297 @@ +/* + * Memory arbiter functions. Allocates bandwith through the + * arbiter and sets up arbiter breakpoints. + * + * The algorithm first assigns slots to the clients that has specified + * bandwith (e.g. ethernet) and then the remaining slots are divided + * on all the active clients. + * + * Copyright (c) 2004, 2005 Axis Communications AB. + */ + +#include <linux/config.h> +#include <asm/arch/hwregs/reg_map.h> +#include <asm/arch/hwregs/reg_rdwr.h> +#include <asm/arch/hwregs/marb_defs.h> +#include <asm/arch/arbiter.h> +#include <asm/arch/hwregs/intr_vect.h> +#include <linux/interrupt.h> +#include <linux/signal.h> +#include <linux/errno.h> +#include <linux/spinlock.h> +#include <asm/io.h> + +struct crisv32_watch_entry +{ + unsigned long instance; + watch_callback* cb; + unsigned long start; + unsigned long end; + int used; +}; + +#define NUMBER_OF_BP 4 +#define NBR_OF_CLIENTS 14 +#define NBR_OF_SLOTS 64 +#define SDRAM_BANDWIDTH 100000000 /* Some kind of expected value */ +#define INTMEM_BANDWIDTH 400000000 +#define NBR_OF_REGIONS 2 + +static struct crisv32_watch_entry watches[NUMBER_OF_BP] = +{ + {regi_marb_bp0}, + {regi_marb_bp1}, + {regi_marb_bp2}, + {regi_marb_bp3} +}; + +static int requested_slots[NBR_OF_REGIONS][NBR_OF_CLIENTS]; +static int active_clients[NBR_OF_REGIONS][NBR_OF_CLIENTS]; +static int max_bandwidth[NBR_OF_REGIONS] = {SDRAM_BANDWIDTH, INTMEM_BANDWIDTH}; + +DEFINE_SPINLOCK(arbiter_lock); + +static irqreturn_t +crisv32_arbiter_irq(int irq, void* dev_id, struct pt_regs* regs); + +static void crisv32_arbiter_config(int region) +{ + int slot; + int client; + int interval = 0; + int val[NBR_OF_SLOTS]; + + for (slot = 0; slot < NBR_OF_SLOTS; slot++) + val[slot] = NBR_OF_CLIENTS + 1; + + for (client = 0; client < NBR_OF_CLIENTS; client++) + { + int pos; + if (!requested_slots[region][client]) + continue; + interval = NBR_OF_SLOTS / requested_slots[region][client]; + pos = 0; + while (pos < NBR_OF_SLOTS) + { + if (val[pos] != NBR_OF_CLIENTS + 1) + pos++; + else + { + val[pos] = client; + pos += interval; + } + } + } + + client = 0; + for (slot = 0; slot < NBR_OF_SLOTS; slot++) + { + if (val[slot] == NBR_OF_CLIENTS + 1) + { + int first = client; + while(!active_clients[region][client]) { + client = (client + 1) % NBR_OF_CLIENTS; + if (client == first) + break; + } + val[slot] = client; + client = (client + 1) % NBR_OF_CLIENTS; + } + if (region == EXT_REGION) + REG_WR_INT_VECT(marb, regi_marb, rw_ext_slots, slot, val[slot]); + else if (region == INT_REGION) + REG_WR_INT_VECT(marb, regi_marb, rw_int_slots, slot, val[slot]); + } +} + +extern char _stext, _etext; + +static void crisv32_arbiter_init(void) +{ + static int initialized = 0; + + if (initialized) + return; + + initialized = 1; + + /* CPU caches are active. */ + active_clients[EXT_REGION][10] = active_clients[EXT_REGION][11] = 1; + crisv32_arbiter_config(EXT_REGION); + crisv32_arbiter_config(INT_REGION); + + if (request_irq(MEMARB_INTR_VECT, crisv32_arbiter_irq, SA_INTERRUPT, + "arbiter", NULL)) + printk(KERN_ERR "Couldn't allocate arbiter IRQ\n"); + +#ifndef CONFIG_ETRAX_KGDB + /* Global watch for writes to kernel text segment. */ + crisv32_arbiter_watch(virt_to_phys(&_stext), &_etext - &_stext, + arbiter_all_clients, arbiter_all_write, NULL); +#endif +} + + + +int crisv32_arbiter_allocate_bandwith(int client, int region, + unsigned long bandwidth) +{ + int i; + int total_assigned = 0; + int total_clients = 0; + int req; + + crisv32_arbiter_init(); + + for (i = 0; i < NBR_OF_CLIENTS; i++) + { + total_assigned += requested_slots[region][i]; + total_clients += active_clients[region][i]; + } + req = NBR_OF_SLOTS / (max_bandwidth[region] / bandwidth); + + if (total_assigned + total_clients + req + 1 > NBR_OF_SLOTS) + return -ENOMEM; + + active_clients[region][client] = 1; + requested_slots[region][client] = req; + crisv32_arbiter_config(region); + + return 0; +} + +int crisv32_arbiter_watch(unsigned long start, unsigned long size, + unsigned long clients, unsigned long accesses, + watch_callback* cb) +{ + int i; + + crisv32_arbiter_init(); + + if (start > 0x80000000) { + printk("Arbiter: %lX doesn't look like a physical address", start); + return -EFAULT; + } + + spin_lock(&arbiter_lock); + + for (i = 0; i < NUMBER_OF_BP; i++) { + if (!watches[i].used) { + reg_marb_rw_intr_mask intr_mask = REG_RD(marb, regi_marb, rw_intr_mask); + + watches[i].used = 1; + watches[i].start = start; + watches[i].end = start + size; + watches[i].cb = cb; + + REG_WR_INT(marb_bp, watches[i].instance, rw_first_addr, watches[i].start); + REG_WR_INT(marb_bp, watches[i].instance, rw_last_addr, watches[i].end); + REG_WR_INT(marb_bp, watches[i].instance, rw_op, accesses); + REG_WR_INT(marb_bp, watches[i].instance, rw_clients, clients); + + if (i == 0) + intr_mask.bp0 = regk_marb_yes; + else if (i == 1) + intr_mask.bp1 = regk_marb_yes; + else if (i == 2) + intr_mask.bp2 = regk_marb_yes; + else if (i == 3) + intr_mask.bp3 = regk_marb_yes; + + REG_WR(marb, regi_marb, rw_intr_mask, intr_mask); + spin_unlock(&arbiter_lock); + + return i; + } + } + spin_unlock(&arbiter_lock); + return -ENOMEM; +} + +int crisv32_arbiter_unwatch(int id) +{ + reg_marb_rw_intr_mask intr_mask = REG_RD(marb, regi_marb, rw_intr_mask); + + crisv32_arbiter_init(); + + spin_lock(&arbiter_lock); + + if ((id < 0) || (id >= NUMBER_OF_BP) || (!watches[id].used)) { + spin_unlock(&arbiter_lock); + return -EINVAL; + } + + memset(&watches[id], 0, sizeof(struct crisv32_watch_entry)); + + if (id == 0) + intr_mask.bp0 = regk_marb_no; + else if (id == 1) + intr_mask.bp2 = regk_marb_no; + else if (id == 2) + intr_mask.bp2 = regk_marb_no; + else if (id == 3) + intr_mask.bp3 = regk_marb_no; + + REG_WR(marb, regi_marb, rw_intr_mask, intr_mask); + + spin_unlock(&arbiter_lock); + return 0; +} + +extern void show_registers(struct pt_regs *regs); + +static irqreturn_t +crisv32_arbiter_irq(int irq, void* dev_id, struct pt_regs* regs) +{ + reg_marb_r_masked_intr masked_intr = REG_RD(marb, regi_marb, r_masked_intr); + reg_marb_bp_r_brk_clients r_clients; + reg_marb_bp_r_brk_addr r_addr; + reg_marb_bp_r_brk_op r_op; + reg_marb_bp_r_brk_first_client r_first; + reg_marb_bp_r_brk_size r_size; + reg_marb_bp_rw_ack ack = {0}; + reg_marb_rw_ack_intr ack_intr = {.bp0=1,.bp1=1,.bp2=1,.bp3=1}; + struct crisv32_watch_entry* watch; + + if (masked_intr.bp0) { + watch = &watches[0]; + ack_intr.bp0 = regk_marb_yes; + } else if (masked_intr.bp1) { + watch = &watches[1]; + ack_intr.bp1 = regk_marb_yes; + } else if (masked_intr.bp2) { + watch = &watches[2]; + ack_intr.bp2 = regk_marb_yes; + } else if (masked_intr.bp3) { + watch = &watches[3]; + ack_intr.bp3 = regk_marb_yes; + } else { + return IRQ_NONE; + } + + /* Retrieve all useful information and print it. */ + r_clients = REG_RD(marb_bp, watch->instance, r_brk_clients); + r_addr = REG_RD(marb_bp, watch->instance, r_brk_addr); + r_op = REG_RD(marb_bp, watch->instance, r_brk_op); + r_first = REG_RD(marb_bp, watch->instance, r_brk_first_client); + r_size = REG_RD(marb_bp, watch->instance, r_brk_size); + + printk("Arbiter IRQ\n"); + printk("Clients %X addr %X op %X first %X size %X\n", + REG_TYPE_CONV(int, reg_marb_bp_r_brk_clients, r_clients), + REG_TYPE_CONV(int, reg_marb_bp_r_brk_addr, r_addr), + REG_TYPE_CONV(int, reg_marb_bp_r_brk_op, r_op), + REG_TYPE_CONV(int, reg_marb_bp_r_brk_first_client, r_first), + REG_TYPE_CONV(int, reg_marb_bp_r_brk_size, r_size)); + + REG_WR(marb_bp, watch->instance, rw_ack, ack); + REG_WR(marb, regi_marb, rw_ack_intr, ack_intr); + + printk("IRQ occured at %lX\n", regs->erp); + + if (watch->cb) + watch->cb(); + + + return IRQ_HANDLED; +} diff --git a/arch/cris/arch-v32/kernel/asm-offsets.c b/arch/cris/arch-v32/kernel/asm-offsets.c new file mode 100644 index 000000000000..15b3d93a0496 --- /dev/null +++ b/arch/cris/arch-v32/kernel/asm-offsets.c @@ -0,0 +1,49 @@ +#include <linux/sched.h> +#include <asm/thread_info.h> + +/* + * Generate definitions needed by assembly language modules. + * This code generates raw asm output which is post-processed to extract + * and format the required data. + */ + +#define DEFINE(sym, val) \ + asm volatile("\n->" #sym " %0 " #val : : "i" (val)) + +#define BLANK() asm volatile("\n->" : : ) + +int main(void) +{ +#define ENTRY(entry) DEFINE(PT_ ## entry, offsetof(struct pt_regs, entry)) + ENTRY(orig_r10); + ENTRY(r13); + ENTRY(r12); + ENTRY(r11); + ENTRY(r10); + ENTRY(r9); + ENTRY(acr); + ENTRY(srs); + ENTRY(mof); + ENTRY(ccs); + ENTRY(srp); + BLANK(); +#undef ENTRY +#define ENTRY(entry) DEFINE(TI_ ## entry, offsetof(struct thread_info, entry)) + ENTRY(task); + ENTRY(flags); + ENTRY(preempt_count); + BLANK(); +#undef ENTRY +#define ENTRY(entry) DEFINE(THREAD_ ## entry, offsetof(struct thread_struct, entry)) + ENTRY(ksp); + ENTRY(usp); + ENTRY(ccs); + BLANK(); +#undef ENTRY +#define ENTRY(entry) DEFINE(TASK_ ## entry, offsetof(struct task_struct, entry)) + ENTRY(pid); + BLANK(); + DEFINE(LCLONE_VM, CLONE_VM); + DEFINE(LCLONE_UNTRACED, CLONE_UNTRACED); + return 0; +} diff --git a/arch/cris/arch-v32/kernel/crisksyms.c b/arch/cris/arch-v32/kernel/crisksyms.c new file mode 100644 index 000000000000..2c3bb9a0afe2 --- /dev/null +++ b/arch/cris/arch-v32/kernel/crisksyms.c @@ -0,0 +1,24 @@ +#include <linux/config.h> +#include <linux/module.h> +#include <linux/irq.h> +#include <asm/arch/dma.h> +#include <asm/arch/intmem.h> +#include <asm/arch/pinmux.h> + +/* Functions for allocating DMA channels */ +EXPORT_SYMBOL(crisv32_request_dma); +EXPORT_SYMBOL(crisv32_free_dma); + +/* Functions for handling internal RAM */ +EXPORT_SYMBOL(crisv32_intmem_alloc); +EXPORT_SYMBOL(crisv32_intmem_free); +EXPORT_SYMBOL(crisv32_intmem_phys_to_virt); +EXPORT_SYMBOL(crisv32_intmem_virt_to_phys); + +/* Functions for handling pinmux */ +EXPORT_SYMBOL(crisv32_pinmux_alloc); +EXPORT_SYMBOL(crisv32_pinmux_dealloc); + +/* Functions masking/unmasking interrupts */ +EXPORT_SYMBOL(mask_irq); +EXPORT_SYMBOL(unmask_irq); diff --git a/arch/cris/arch-v32/kernel/debugport.c b/arch/cris/arch-v32/kernel/debugport.c new file mode 100644 index 000000000000..ffc1ebf2dfee --- /dev/null +++ b/arch/cris/arch-v32/kernel/debugport.c @@ -0,0 +1,461 @@ +/* + * Copyright (C) 2003, Axis Communications AB. + */ + +#include <linux/config.h> +#include <linux/console.h> +#include <linux/init.h> +#include <linux/major.h> +#include <linux/delay.h> +#include <linux/tty.h> +#include <asm/system.h> +#include <asm/io.h> +#include <asm/arch/hwregs/ser_defs.h> +#include <asm/arch/hwregs/dma_defs.h> +#include <asm/arch/pinmux.h> + +#include <asm/irq.h> +#include <asm/arch/hwregs/intr_vect_defs.h> + +struct dbg_port +{ + unsigned char nbr; + unsigned long instance; + unsigned int started; + unsigned long baudrate; + unsigned char parity; + unsigned int bits; +}; + +struct dbg_port ports[] = +{ + { + 0, + regi_ser0, + 0, + 115200, + 'N', + 8 + }, + { + 1, + regi_ser1, + 0, + 115200, + 'N', + 8 + }, + { + 2, + regi_ser2, + 0, + 115200, + 'N', + 8 + }, + { + 3, + regi_ser3, + 0, + 115200, + 'N', + 8 + } +}; +static struct dbg_port *port = +#if defined(CONFIG_ETRAX_DEBUG_PORT0) +&ports[0]; +#elif defined(CONFIG_ETRAX_DEBUG_PORT1) +&ports[1]; +#elif defined(CONFIG_ETRAX_DEBUG_PORT2) +&ports[2]; +#elif defined(CONFIG_ETRAX_DEBUG_PORT3) +&ports[3]; +#else +NULL; +#endif + +#ifdef CONFIG_ETRAX_KGDB +static struct dbg_port *kgdb_port = +#if defined(CONFIG_ETRAX_KGDB_PORT0) +&ports[0]; +#elif defined(CONFIG_ETRAX_KGDB_PORT1) +&ports[1]; +#elif defined(CONFIG_ETRAX_KGDB_PORT2) +&ports[2]; +#elif defined(CONFIG_ETRAX_KGDB_PORT3) +&ports[3]; +#else +NULL; +#endif +#endif + +#ifdef CONFIG_ETRAXFS_SIM +extern void print_str( const char *str ); +static char buffer[1024]; +static char msg[] = "Debug: "; +static int buffer_pos = sizeof(msg) - 1; +#endif + +extern struct tty_driver *serial_driver; + +static void +start_port(struct dbg_port* p) +{ + if (!p) + return; + + if (p->started) + return; + p->started = 1; + + if (p->nbr == 1) + crisv32_pinmux_alloc_fixed(pinmux_ser1); + else if (p->nbr == 2) + crisv32_pinmux_alloc_fixed(pinmux_ser2); + else if (p->nbr == 3) + crisv32_pinmux_alloc_fixed(pinmux_ser3); + + /* Set up serial port registers */ + reg_ser_rw_tr_ctrl tr_ctrl = {0}; + reg_ser_rw_tr_dma_en tr_dma_en = {0}; + + reg_ser_rw_rec_ctrl rec_ctrl = {0}; + reg_ser_rw_tr_baud_div tr_baud_div = {0}; + reg_ser_rw_rec_baud_div rec_baud_div = {0}; + + tr_ctrl.base_freq = rec_ctrl.base_freq = regk_ser_f29_493; + tr_dma_en.en = rec_ctrl.dma_mode = regk_ser_no; + tr_baud_div.div = rec_baud_div.div = 29493000 / p->baudrate / 8; + tr_ctrl.en = rec_ctrl.en = 1; + + if (p->parity == 'O') + { + tr_ctrl.par_en = regk_ser_yes; + tr_ctrl.par = regk_ser_odd; + rec_ctrl.par_en = regk_ser_yes; + rec_ctrl.par = regk_ser_odd; + } + else if (p->parity == 'E') + { + tr_ctrl.par_en = regk_ser_yes; + tr_ctrl.par = regk_ser_even; + rec_ctrl.par_en = regk_ser_yes; + rec_ctrl.par = regk_ser_odd; + } + + if (p->bits == 7) + { + tr_ctrl.data_bits = regk_ser_bits7; + rec_ctrl.data_bits = regk_ser_bits7; + } + + REG_WR (ser, p->instance, rw_tr_baud_div, tr_baud_div); + REG_WR (ser, p->instance, rw_rec_baud_div, rec_baud_div); + REG_WR (ser, p->instance, rw_tr_dma_en, tr_dma_en); + REG_WR (ser, p->instance, rw_tr_ctrl, tr_ctrl); + REG_WR (ser, p->instance, rw_rec_ctrl, rec_ctrl); +} + +/* No debug */ +#ifdef CONFIG_ETRAX_DEBUG_PORT_NULL + +static void +console_write(struct console *co, const char *buf, unsigned int len) +{ + return; +} + +/* Target debug */ +#elif !defined(CONFIG_ETRAXFS_SIM) + +static void +console_write_direct(struct console *co, const char *buf, unsigned int len) +{ + int i; + reg_ser_r_stat_din stat; + reg_ser_rw_tr_dma_en tr_dma_en, old; + + /* Switch to manual mode */ + tr_dma_en = old = REG_RD (ser, port->instance, rw_tr_dma_en); + if (tr_dma_en.en == regk_ser_yes) { + tr_dma_en.en = regk_ser_no; + REG_WR(ser, port->instance, rw_tr_dma_en, tr_dma_en); + } + + /* Send data */ + for (i = 0; i < len; i++) { + /* LF -> CRLF */ + if (buf[i] == '\n') { + do { + stat = REG_RD (ser, port->instance, r_stat_din); + } while (!stat.tr_rdy); + REG_WR_INT (ser, port->instance, rw_dout, '\r'); + } + /* Wait until transmitter is ready and send.*/ + do { + stat = REG_RD (ser, port->instance, r_stat_din); + } while (!stat.tr_rdy); + REG_WR_INT (ser, port->instance, rw_dout, buf[i]); + } + + /* Restore mode */ + if (tr_dma_en.en != old.en) + REG_WR(ser, port->instance, rw_tr_dma_en, old); +} + +static void +console_write(struct console *co, const char *buf, unsigned int len) +{ + if (!port) + return; + console_write_direct(co, buf, len); +} + + + +#else + +/* VCS debug */ + +static void +console_write(struct console *co, const char *buf, unsigned int len) +{ + char* pos; + pos = memchr(buf, '\n', len); + if (pos) { + int l = ++pos - buf; + memcpy(buffer + buffer_pos, buf, l); + memcpy(buffer, msg, sizeof(msg) - 1); + buffer[buffer_pos + l] = '\0'; + print_str(buffer); + buffer_pos = sizeof(msg) - 1; + if (pos - buf != len) { + memcpy(buffer + buffer_pos, pos, len - l); + buffer_pos += len - l; + } + } else { + memcpy(buffer + buffer_pos, buf, len); + buffer_pos += len; + } +} + +#endif + +int raw_printk(const char *fmt, ...) +{ + static char buf[1024]; + int printed_len; + va_list args; + va_start(args, fmt); + printed_len = vsnprintf(buf, sizeof(buf), fmt, args); + va_end(args); + console_write(NULL, buf, strlen(buf)); + return printed_len; +} + +void +stupid_debug(char* buf) +{ + console_write(NULL, buf, strlen(buf)); +} + +#ifdef CONFIG_ETRAX_KGDB +/* Use polling to get a single character from the kernel debug port */ +int +getDebugChar(void) +{ + reg_ser_rs_status_data stat; + reg_ser_rw_ack_intr ack_intr = { 0 }; + + do { + stat = REG_RD(ser, kgdb_instance, rs_status_data); + } while (!stat.data_avail); + + /* Ack the data_avail interrupt. */ + ack_intr.data_avail = 1; + REG_WR(ser, kgdb_instance, rw_ack_intr, ack_intr); + + return stat.data; +} + +/* Use polling to put a single character to the kernel debug port */ +void +putDebugChar(int val) +{ + reg_ser_r_status_data stat; + do { + stat = REG_RD (ser, kgdb_instance, r_status_data); + } while (!stat.tr_ready); + REG_WR (ser, kgdb_instance, rw_data_out, REG_TYPE_CONV(reg_ser_rw_data_out, int, val)); +} +#endif /* CONFIG_ETRAX_KGDB */ + +static int __init +console_setup(struct console *co, char *options) +{ + char* s; + + if (options) { + port = &ports[co->index]; + port->baudrate = 115200; + port->parity = 'N'; + port->bits = 8; + port->baudrate = simple_strtoul(options, NULL, 10); + s = options; + while(*s >= '0' && *s <= '9') + s++; + if (*s) port->parity = *s++; + if (*s) port->bits = *s++ - '0'; + port->started = 0; + start_port(port); + } + return 0; +} + +/* This is a dummy serial device that throws away anything written to it. + * This is used when no debug output is wanted. + */ +static struct tty_driver dummy_driver; + +static int dummy_open(struct tty_struct *tty, struct file * filp) +{ + return 0; +} + +static void dummy_close(struct tty_struct *tty, struct file * filp) +{ +} + +static int dummy_write(struct tty_struct * tty, + const unsigned char *buf, int count) +{ + return count; +} + +static int +dummy_write_room(struct tty_struct *tty) +{ + return 8192; +} + +void __init +init_dummy_console(void) +{ + memset(&dummy_driver, 0, sizeof(struct tty_driver)); + dummy_driver.driver_name = "serial"; + dummy_driver.name = "ttyS"; + dummy_driver.major = TTY_MAJOR; + dummy_driver.minor_start = 68; + dummy_driver.num = 1; /* etrax100 has 4 serial ports */ + dummy_driver.type = TTY_DRIVER_TYPE_SERIAL; + dummy_driver.subtype = SERIAL_TYPE_NORMAL; + dummy_driver.init_termios = tty_std_termios; + dummy_driver.init_termios.c_cflag = + B115200 | CS8 | CREAD | HUPCL | CLOCAL; /* is normally B9600 default... */ + dummy_driver.flags = TTY_DRIVER_REAL_RAW | TTY_DRIVER_NO_DEVFS; + + dummy_driver.open = dummy_open; + dummy_driver.close = dummy_close; + dummy_driver.write = dummy_write; + dummy_driver.write_room = dummy_write_room; + if (tty_register_driver(&dummy_driver)) + panic("Couldn't register dummy serial driver\n"); +} + +static struct tty_driver* +crisv32_console_device(struct console* co, int *index) +{ + if (port) + *index = port->nbr; + return port ? serial_driver : &dummy_driver; +} + +static struct console sercons = { + name : "ttyS", + write: console_write, + read : NULL, + device : crisv32_console_device, + unblank : NULL, + setup : console_setup, + flags : CON_PRINTBUFFER, + index : -1, + cflag : 0, + next : NULL +}; +static struct console sercons0 = { + name : "ttyS", + write: console_write, + read : NULL, + device : crisv32_console_device, + unblank : NULL, + setup : console_setup, + flags : CON_PRINTBUFFER, + index : 0, + cflag : 0, + next : NULL +}; + +static struct console sercons1 = { + name : "ttyS", + write: console_write, + read : NULL, + device : crisv32_console_device, + unblank : NULL, + setup : console_setup, + flags : CON_PRINTBUFFER, + index : 1, + cflag : 0, + next : NULL +}; +static struct console sercons2 = { + name : "ttyS", + write: console_write, + read : NULL, + device : crisv32_console_device, + unblank : NULL, + setup : console_setup, + flags : CON_PRINTBUFFER, + index : 2, + cflag : 0, + next : NULL +}; +static struct console sercons3 = { + name : "ttyS", + write: console_write, + read : NULL, + device : crisv32_console_device, + unblank : NULL, + setup : console_setup, + flags : CON_PRINTBUFFER, + index : 3, + cflag : 0, + next : NULL +}; + +/* Register console for printk's, etc. */ +int __init +init_etrax_debug(void) +{ + static int first = 1; + + if (!first) { + unregister_console(&sercons); + register_console(&sercons0); + register_console(&sercons1); + register_console(&sercons2); + register_console(&sercons3); + init_dummy_console(); + return 0; + } + first = 0; + register_console(&sercons); + start_port(port); + +#ifdef CONFIG_ETRAX_KGDB + start_port(kgdb_port); +#endif /* CONFIG_ETRAX_KGDB */ + return 0; +} + +__initcall(init_etrax_debug); diff --git a/arch/cris/arch-v32/kernel/dma.c b/arch/cris/arch-v32/kernel/dma.c new file mode 100644 index 000000000000..b92e85799b44 --- /dev/null +++ b/arch/cris/arch-v32/kernel/dma.c @@ -0,0 +1,224 @@ +/* Wrapper for DMA channel allocator that starts clocks etc */ + +#include <linux/kernel.h> +#include <linux/spinlock.h> +#include <asm/dma.h> +#include <asm/arch/hwregs/reg_map.h> +#include <asm/arch/hwregs/reg_rdwr.h> +#include <asm/arch/hwregs/marb_defs.h> +#include <asm/arch/hwregs/config_defs.h> +#include <asm/arch/hwregs/strmux_defs.h> +#include <linux/errno.h> +#include <asm/system.h> +#include <asm/arch/arbiter.h> + +static char used_dma_channels[MAX_DMA_CHANNELS]; +static const char * used_dma_channels_users[MAX_DMA_CHANNELS]; + +static DEFINE_SPINLOCK(dma_lock); + +int crisv32_request_dma(unsigned int dmanr, const char * device_id, + unsigned options, unsigned int bandwidth, + enum dma_owner owner) +{ + unsigned long flags; + reg_config_rw_clk_ctrl clk_ctrl; + reg_strmux_rw_cfg strmux_cfg; + + if (crisv32_arbiter_allocate_bandwith(dmanr, + options & DMA_INT_MEM ? INT_REGION : EXT_REGION, + bandwidth)) + return -ENOMEM; + + spin_lock_irqsave(&dma_lock, flags); + + if (used_dma_channels[dmanr]) { + spin_unlock_irqrestore(&dma_lock, flags); + if (options & DMA_VERBOSE_ON_ERROR) { + printk("Failed to request DMA %i for %s, already allocated by %s\n", dmanr, device_id, used_dma_channels_users[dmanr]); + } + if (options & DMA_PANIC_ON_ERROR) + panic("request_dma error!"); + return -EBUSY; + } + clk_ctrl = REG_RD(config, regi_config, rw_clk_ctrl); + strmux_cfg = REG_RD(strmux, regi_strmux, rw_cfg); + + switch(dmanr) + { + case 0: + case 1: + clk_ctrl.dma01_eth0 = 1; + break; + case 2: + case 3: + clk_ctrl.dma23 = 1; + break; + case 4: + case 5: + clk_ctrl.dma45 = 1; + break; + case 6: + case 7: + clk_ctrl.dma67 = 1; + break; + case 8: + case 9: + clk_ctrl.dma89_strcop = 1; + break; +#if MAX_DMA_CHANNELS-1 != 9 +#error Check dma.c +#endif + default: + spin_unlock_irqrestore(&dma_lock, flags); + if (options & DMA_VERBOSE_ON_ERROR) { + printk("Failed to request DMA %i for %s, only 0-%i valid)\n", dmanr, device_id, MAX_DMA_CHANNELS-1); + } + + if (options & DMA_PANIC_ON_ERROR) + panic("request_dma error!"); + return -EINVAL; + } + + switch(owner) + { + case dma_eth0: + if (dmanr == 0) + strmux_cfg.dma0 = regk_strmux_eth0; + else if (dmanr == 1) + strmux_cfg.dma1 = regk_strmux_eth0; + else + panic("Invalid DMA channel for eth0\n"); + break; + case dma_eth1: + if (dmanr == 6) + strmux_cfg.dma6 = regk_strmux_eth1; + else if (dmanr == 7) + strmux_cfg.dma7 = regk_strmux_eth1; + else + panic("Invalid DMA channel for eth1\n"); + break; + case dma_iop0: + if (dmanr == 2) + strmux_cfg.dma2 = regk_strmux_iop0; + else if (dmanr == 3) + strmux_cfg.dma3 = regk_strmux_iop0; + else + panic("Invalid DMA channel for iop0\n"); + break; + case dma_iop1: + if (dmanr == 4) + strmux_cfg.dma4 = regk_strmux_iop1; + else if (dmanr == 5) + strmux_cfg.dma5 = regk_strmux_iop1; + else + panic("Invalid DMA channel for iop1\n"); + break; + case dma_ser0: + if (dmanr == 6) + strmux_cfg.dma6 = regk_strmux_ser0; + else if (dmanr == 7) + strmux_cfg.dma7 = regk_strmux_ser0; + else + panic("Invalid DMA channel for ser0\n"); + break; + case dma_ser1: + if (dmanr == 4) + strmux_cfg.dma4 = regk_strmux_ser1; + else if (dmanr == 5) + strmux_cfg.dma5 = regk_strmux_ser1; + else + panic("Invalid DMA channel for ser1\n"); + break; + case dma_ser2: + if (dmanr == 2) + strmux_cfg.dma2 = regk_strmux_ser2; + else if (dmanr == 3) + strmux_cfg.dma3 = regk_strmux_ser2; + else + panic("Invalid DMA channel for ser2\n"); + break; + case dma_ser3: + if (dmanr == 8) + strmux_cfg.dma8 = regk_strmux_ser3; + else if (dmanr == 9) + strmux_cfg.dma9 = regk_strmux_ser3; + else + panic("Invalid DMA channel for ser3\n"); + break; + case dma_sser0: + if (dmanr == 4) + strmux_cfg.dma4 = regk_strmux_sser0; + else if (dmanr == 5) + strmux_cfg.dma5 = regk_strmux_sser0; + else + panic("Invalid DMA channel for sser0\n"); + break; + case dma_sser1: + if (dmanr == 6) + strmux_cfg.dma6 = regk_strmux_sser1; + else if (dmanr == 7) + strmux_cfg.dma7 = regk_strmux_sser1; + else + panic("Invalid DMA channel for sser1\n"); + break; + case dma_ata: + if (dmanr == 2) + strmux_cfg.dma2 = regk_strmux_ata; + else if (dmanr == 3) + strmux_cfg.dma3 = regk_strmux_ata; + else + panic("Invalid DMA channel for ata\n"); + break; + case dma_strp: + if (dmanr == 8) + strmux_cfg.dma8 = regk_strmux_strcop; + else if (dmanr == 9) + strmux_cfg.dma9 = regk_strmux_strcop; + else + panic("Invalid DMA channel for strp\n"); + break; + case dma_ext0: + if (dmanr == 6) + strmux_cfg.dma6 = regk_strmux_ext0; + else + panic("Invalid DMA channel for ext0\n"); + break; + case dma_ext1: + if (dmanr == 7) + strmux_cfg.dma7 = regk_strmux_ext1; + else + panic("Invalid DMA channel for ext1\n"); + break; + case dma_ext2: + if (dmanr == 2) + strmux_cfg.dma2 = regk_strmux_ext2; + else if (dmanr == 8) + strmux_cfg.dma8 = regk_strmux_ext2; + else + panic("Invalid DMA channel for ext2\n"); + break; + case dma_ext3: + if (dmanr == 3) + strmux_cfg.dma3 = regk_strmux_ext3; + else if (dmanr == 9) + strmux_cfg.dma9 = regk_strmux_ext2; + else + panic("Invalid DMA channel for ext2\n"); + break; + } + + used_dma_channels[dmanr] = 1; + used_dma_channels_users[dmanr] = device_id; + REG_WR(config, regi_config, rw_clk_ctrl, clk_ctrl); + REG_WR(strmux, regi_strmux, rw_cfg, strmux_cfg); + spin_unlock_irqrestore(&dma_lock,flags); + return 0; +} + +void crisv32_free_dma(unsigned int dmanr) +{ + spin_lock(&dma_lock); + used_dma_channels[dmanr] = 0; + spin_unlock(&dma_lock); +} diff --git a/arch/cris/arch-v32/kernel/entry.S b/arch/cris/arch-v32/kernel/entry.S new file mode 100644 index 000000000000..a8ed55e5b403 --- /dev/null +++ b/arch/cris/arch-v32/kernel/entry.S @@ -0,0 +1,820 @@ +/* + * Copyright (C) 2000-2003 Axis Communications AB + * + * Authors: Bjorn Wesen (bjornw@axis.com) + * Tobias Anderberg (tobiasa@axis.com), CRISv32 port. + * + * Code for the system-call and fault low-level handling routines. + * + * NOTE: This code handles signal-recognition, which happens every time + * after a timer-interrupt and after each system call. + * + * Stack layout in 'ret_from_system_call': + * ptrace needs to have all regs on the stack. + * if the order here is changed, it needs to be + * updated in fork.c:copy_process, signal.c:do_signal, + * ptrace.c and ptrace.h + * + */ + +#include <linux/config.h> +#include <linux/linkage.h> +#include <linux/sys.h> +#include <asm/unistd.h> +#include <asm/errno.h> +#include <asm/thread_info.h> +#include <asm/arch/offset.h> + +#include <asm/arch/hwregs/asm/reg_map_asm.h> +#include <asm/arch/hwregs/asm/intr_vect_defs_asm.h> + + ;; Exported functions. + .globl system_call + .globl ret_from_intr + .globl ret_from_fork + .globl resume + .globl multiple_interrupt + .globl nmi_interrupt + .globl spurious_interrupt + .globl do_sigtrap + .globl gdb_handle_exception + .globl sys_call_table + + ; Check if preemptive kernel scheduling should be done. +#ifdef CONFIG_PREEMPT +_resume_kernel: + di + ; Load current task struct. + movs.w -8192, $r0 ; THREAD_SIZE = 8192 + and.d $sp, $r0 + + addoq +TI_preempt_count, $r0, $acr + move.d [$acr], $r10 ; Preemption disabled? + bne _Rexit + nop + +_need_resched: + addoq +TI_flags, $r0, $acr + move.d [$acr], $r10 + btstq TIF_NEED_RESCHED, $r10 ; Check if need_resched is set. + bpl _Rexit + nop + + ; Do preemptive kernel scheduling. + jsr preempt_schedule_irq + nop + + ; Load new task struct. + movs.w -8192, $r0 ; THREAD_SIZE = 8192. + and.d $sp, $r0 + + ; One more time with new task. + ba _need_resched + nop +#else +#define _resume_kernel _Rexit +#endif + + ; Called at exit from fork. schedule_tail must be called to drop + ; spinlock if CONFIG_PREEMPT. +ret_from_fork: + jsr schedule_tail + nop + ba ret_from_sys_call + nop + +ret_from_intr: + ;; Check for resched if preemptive kernel, or if we're going back to + ;; user-mode. This test matches the user_regs(regs) macro. Don't simply + ;; test CCS since that doesn't necessarily reflect what mode we'll + ;; return into. + addoq +PT_ccs, $sp, $acr + move.d [$acr], $r0 + btstq 16, $r0 ; User-mode flag. + bpl _resume_kernel + + ; Note that di below is in delay slot. + +_resume_userspace: + di ; So need_resched and sigpending don't change. + + movs.w -8192, $r0 ; THREAD_SIZE == 8192 + and.d $sp, $r0 + + addoq +TI_flags, $r0, $acr ; current->work + move.d [$acr], $r10 + and.d _TIF_WORK_MASK, $r10 ; Work to be done on return? + bne _work_pending + nop + ba _Rexit + nop + + ;; The system_call is called by a BREAK instruction, which looks pretty + ;; much like any other exception. + ;; + ;; System calls can't be made from interrupts but we still stack ERP + ;; to have a complete stack frame. + ;; + ;; In r9 we have the wanted syscall number. Arguments come in r10,r11,r12, + ;; r13,mof,srp + ;; + ;; This function looks on the _surface_ like spaghetti programming, but it's + ;; really designed so that the fast-path does not force cache-loading of + ;; non-used instructions. Only the non-common cases cause the outlined code + ;; to run.. + +system_call: + ;; Stack-frame similar to the irq heads, which is reversed in + ;; ret_from_sys_call. + subq 12, $sp ; Skip EXS, EDA. + move $erp, [$sp] + subq 4, $sp + move $srp, [$sp] + subq 4, $sp + move $ccs, [$sp] + subq 4, $sp + ei ; Allow IRQs while handling system call + move $spc, [$sp] + subq 4, $sp + move $mof, [$sp] + subq 4, $sp + move $srs, [$sp] + subq 4, $sp + move.d $acr, [$sp] + subq 14*4, $sp ; Make room for R0-R13. + movem $r13, [$sp] ; Push R0-R13 + subq 4, $sp + move.d $r10, [$sp] ; Push orig_r10. + +; Set S-bit when kernel debugging to keep hardware breakpoints active. +#ifdef CONFIG_ETRAX_KGDB + move $ccs, $r0 + or.d (1<<9), $r0 + move $r0, $ccs +#endif + + movs.w -ENOSYS, $r0 + addoq +PT_r10, $sp, $acr + move.d $r0, [$acr] + + ;; Check if this process is syscall-traced. + movs.w -8192, $r0 ; THREAD_SIZE == 8192 + and.d $sp, $r0 + + addoq +TI_flags, $r0, $acr + move.d [$acr], $r0 + btstq TIF_SYSCALL_TRACE, $r0 + bmi _syscall_trace_entry + nop + +_syscall_traced: + ;; Check for sanity in the requested syscall number. + cmpu.w NR_syscalls, $r9 + bhs ret_from_sys_call + lslq 2, $r9 ; Multiply by 4, in the delay slot. + + ;; The location on the stack for the register structure is passed as a + ;; seventh argument. Some system calls need this. + move.d $sp, $r0 + subq 4, $sp + move.d $r0, [$sp] + + ;; The registers carrying parameters (R10-R13) are intact. The optional + ;; fifth and sixth parameters is in MOF and SRP respectivly. Put them + ;; back on the stack. + subq 4, $sp + move $srp, [$sp] + subq 4, $sp + move $mof, [$sp] + + ;; Actually to the system call. + addo.d +sys_call_table, $r9, $acr + move.d [$acr], $acr + jsr $acr + nop + + addq 3*4, $sp ; Pop the mof, srp and regs parameters. + addoq +PT_r10, $sp, $acr + move.d $r10, [$acr] ; Save the return value. + + moveq 1, $r9 ; "Parameter" to ret_from_sys_call to + ; show it was a sys call. + + ;; Fall through into ret_from_sys_call to return. + +ret_from_sys_call: + ;; R9 is a parameter: + ;; >= 1 from syscall + ;; 0 from irq + + ;; Get the current task-struct pointer. + movs.w -8192, $r0 ; THREAD_SIZE == 8192 + and.d $sp, $r0 + + di ; Make sure need_resched and sigpending don't change. + + addoq +TI_flags, $r0, $acr + move.d [$acr], $r1 + and.d _TIF_ALLWORK_MASK, $r1 + bne _syscall_exit_work + nop + +_Rexit: + ;; This epilogue MUST match the prologues in multiple_interrupt, irq.h + ;; and ptregs.h. + addq 4, $sp ; Skip orig_r10. + movem [$sp+], $r13 ; Registers R0-R13. + move.d [$sp+], $acr + move [$sp], $srs + addq 4, $sp + move [$sp+], $mof + move [$sp+], $spc + move [$sp+], $ccs + move [$sp+], $srp + move [$sp+], $erp + addq 8, $sp ; Skip EXS, EDA. + jump $erp + rfe ; Restore condition code stack in delay-slot. + + ;; We get here after doing a syscall if extra work might need to be done + ;; perform syscall exit tracing if needed. + +_syscall_exit_work: + ;; R0 contains current at this point and irq's are disabled. + + addoq +TI_flags, $r0, $acr + move.d [$acr], $r1 + btstq TIF_SYSCALL_TRACE, $r1 + bpl _work_pending + nop + ei + move.d $r9, $r1 ; Preserve R9. + jsr do_syscall_trace + nop + move.d $r1, $r9 + ba _resume_userspace + nop + +_work_pending: + addoq +TI_flags, $r0, $acr + move.d [$acr], $r10 + btstq TIF_NEED_RESCHED, $r10 ; Need resched? + bpl _work_notifysig ; No, must be signal/notify. + nop + +_work_resched: + move.d $r9, $r1 ; Preserve R9. + jsr schedule + nop + move.d $r1, $r9 + di + + addoq +TI_flags, $r0, $acr + move.d [$acr], $r1 + and.d _TIF_WORK_MASK, $r1 ; Ignore sycall trace counter. + beq _Rexit + nop + btstq TIF_NEED_RESCHED, $r1 + bmi _work_resched ; current->work.need_resched. + nop + +_work_notifysig: + ;; Deal with pending signals and notify-resume requests. + + addoq +TI_flags, $r0, $acr + move.d [$acr], $r13 ; The thread_info_flags parameter. + move.d $r9, $r10 ; do_notify_resume syscall/irq param. + moveq 0, $r11 ; oldset param - 0 in this case. + move.d $sp, $r12 ; The regs param. + jsr do_notify_resume + nop + + ba _Rexit + nop + + ;; We get here as a sidetrack when we've entered a syscall with the + ;; trace-bit set. We need to call do_syscall_trace and then continue + ;; with the call. + +_syscall_trace_entry: + ;; PT_r10 in the frame contains -ENOSYS as required, at this point. + + jsr do_syscall_trace + nop + + ;; Now re-enter the syscall code to do the syscall itself. We need to + ;; restore R9 here to contain the wanted syscall, and the other + ;; parameter-bearing registers. + addoq +PT_r9, $sp, $acr + move.d [$acr], $r9 + addoq +PT_orig_r10, $sp, $acr + move.d [$acr], $r10 ; PT_r10 is already -ENOSYS. + addoq +PT_r11, $sp, $acr + move.d [$acr], $r11 + addoq +PT_r12, $sp, $acr + move.d [$acr], $r12 + addoq +PT_r13, $sp, $acr + move.d [$acr], $r13 + addoq +PT_mof, $sp, $acr + move [$acr], $mof + addoq +PT_srp, $sp, $acr + move [$acr], $srp + + ba _syscall_traced + nop + + ;; Resume performs the actual task-switching, by switching stack + ;; pointers. Input arguments are: + ;; + ;; R10 = prev + ;; R11 = next + ;; R12 = thread offset in task struct. + ;; + ;; Returns old current in R10. + +resume: + subq 4, $sp + move $srp, [$sp] ; Keep old/new PC on the stack. + add.d $r12, $r10 ; R10 = current tasks tss. + addoq +THREAD_ccs, $r10, $acr + move $ccs, [$acr] ; Save IRQ enable state. + di + + addoq +THREAD_usp, $r10, $acr + move $usp, [$acr] ; Save user-mode stackpointer. + + ;; See copy_thread for the reason why register R9 is saved. + subq 10*4, $sp + movem $r9, [$sp] ; Save non-scratch registers and R9. + + addoq +THREAD_ksp, $r10, $acr + move.d $sp, [$acr] ; Save kernel SP for old task. + + move.d $sp, $r10 ; Return last running task in R10. + and.d -8192, $r10 ; Get thread_info from stackpointer. + addoq +TI_task, $r10, $acr + move.d [$acr], $r10 ; Get task. + add.d $r12, $r11 ; Find the new tasks tss. + addoq +THREAD_ksp, $r11, $acr + move.d [$acr], $sp ; Switch to new stackframe. + movem [$sp+], $r9 ; Restore non-scratch registers and R9. + + addoq +THREAD_usp, $r11, $acr + move [$acr], $usp ; Restore user-mode stackpointer. + + addoq +THREAD_ccs, $r11, $acr + move [$acr], $ccs ; Restore IRQ enable status. + move.d [$sp+], $acr + jump $acr ; Restore PC. + nop + +nmi_interrupt: + +;; If we receive a watchdog interrupt while it is not expected, then set +;; up a canonical frame and dump register contents before dying. + + ;; This prologue MUST match the one in irq.h and the struct in ptregs.h! + subq 12, $sp ; Skip EXS, EDA. + move $nrp, [$sp] + subq 4, $sp + move $srp, [$sp] + subq 4, $sp + move $ccs, [$sp] + subq 4, $sp + move $spc, [$sp] + subq 4, $sp + move $mof, [$sp] + subq 4, $sp + move $srs, [$sp] + subq 4, $sp + move.d $acr, [$sp] + subq 14*4, $sp ; Make room for R0-R13. + movem $r13, [$sp] ; Push R0-R13. + subq 4, $sp + move.d $r10, [$sp] ; Push orig_r10. + move.d REG_ADDR(intr_vect, regi_irq, r_nmi), $r0 + move.d [$r0], $r0 + btstq REG_BIT(intr_vect, r_nmi, watchdog), $r0 + bpl 1f + nop + jsr handle_watchdog_bite ; In time.c. + move.d $sp, $r10 ; Pointer to registers +1: btstq REG_BIT(intr_vect, r_nmi, ext), $r0 + bpl 1f + nop + jsr handle_nmi + move.d $sp, $r10 ; Pointer to registers +1: addq 4, $sp ; Skip orig_r10 + movem [$sp+], $r13 + move.d [$sp+], $acr + move [$sp], $srs + addq 4, $sp + move [$sp+], $mof + move [$sp+], $spc + move [$sp+], $ccs + move [$sp+], $srp + move [$sp+], $nrp + addq 8, $sp ; Skip EXS, EDA. + jump $nrp + rfn + + .comm cause_of_death, 4 ;; Don't declare this anywhere. + +spurious_interrupt: + di + jump hard_reset_now + nop + + ;; This handles the case when multiple interrupts arrive at the same + ;; time. Jump to the first set interrupt bit in a priotiry fashion. The + ;; hardware will call the unserved interrupts after the handler + ;; finishes. +multiple_interrupt: + ;; This prologue MUST match the one in irq.h and the struct in ptregs.h! + subq 12, $sp ; Skip EXS, EDA. + move $erp, [$sp] + subq 4, $sp + move $srp, [$sp] + subq 4, $sp + move $ccs, [$sp] + subq 4, $sp + move $spc, [$sp] + subq 4, $sp + move $mof, [$sp] + subq 4, $sp + move $srs, [$sp] + subq 4, $sp + move.d $acr, [$sp] + subq 14*4, $sp ; Make room for R0-R13. + movem $r13, [$sp] ; Push R0-R13. + subq 4, $sp + move.d $r10, [$sp] ; Push orig_r10. + +; Set S-bit when kernel debugging to keep hardware breakpoints active. +#ifdef CONFIG_ETRAX_KGDB + move $ccs, $r0 + or.d (1<<9), $r0 + move $r0, $ccs +#endif + + jsr crisv32_do_multiple + move.d $sp, $r10 + jump ret_from_intr + nop + +do_sigtrap: + ;; Sigtraps the process that executed the BREAK instruction. Creates a + ;; frame that Rexit expects. + subq 4, $sp + move $eda, [$sp] + subq 4, $sp + move $exs, [$sp] + subq 4, $sp + move $erp, [$sp] + subq 4, $sp + move $srp, [$sp] + subq 4, $sp + move $ccs, [$sp] + subq 4, $sp + move $spc, [$sp] + subq 4, $sp + move $mof, [$sp] + subq 4, $sp + move $srs, [$sp] + subq 4, $sp + move.d $acr, [$sp] + di ; Need to disable irq's at this point. + subq 14*4, $sp ; Make room for r0-r13. + movem $r13, [$sp] ; Push the r0-r13 registers. + subq 4, $sp + move.d $r10, [$sp] ; Push orig_r10. + + movs.w -8192, $r9 ; THREAD_SIZE == 8192 + and.d $sp, $r9 + + ;; thread_info as first parameter + move.d $r9, $r10 + moveq 5, $r11 ; SIGTRAP as second argument. + jsr ugdb_trap_user + nop + jump ret_from_intr ; Use the return routine for interrupts. + nop + +gdb_handle_exception: + subq 4, $sp + move.d $r0, [$sp] +#ifdef CONFIG_ETRAX_KGDB + move $ccs, $r0 ; U-flag not affected by previous insns. + btstq 16, $r0 ; Test the U-flag. + bmi _ugdb_handle_exception ; Go to user mode debugging. + nop ; Empty delay-slot (cannot pop R0 here). + ba kgdb_handle_exception ; Go to kernel debugging. + move.d [$sp+], $r0 ; Restore R0 in delay slot. +#endif + +_ugdb_handle_exception: + ba do_sigtrap ; SIGTRAP the offending process. + move.d [$sp+], $r0 ; Restore R0 in delay slot. + + .data + + .section .rodata,"a" +sys_call_table: + .long sys_restart_syscall ; 0 - old "setup()" system call, used + ; for restarting. + .long sys_exit + .long sys_fork + .long sys_read + .long sys_write + .long sys_open /* 5 */ + .long sys_close + .long sys_waitpid + .long sys_creat + .long sys_link + .long sys_unlink /* 10 */ + .long sys_execve + .long sys_chdir + .long sys_time + .long sys_mknod + .long sys_chmod /* 15 */ + .long sys_lchown16 + .long sys_ni_syscall /* old break syscall holder */ + .long sys_stat + .long sys_lseek + .long sys_getpid /* 20 */ + .long sys_mount + .long sys_oldumount + .long sys_setuid16 + .long sys_getuid16 + .long sys_stime /* 25 */ + .long sys_ptrace + .long sys_alarm + .long sys_fstat + .long sys_pause + .long sys_utime /* 30 */ + .long sys_ni_syscall /* old stty syscall holder */ + .long sys_ni_syscall /* old gtty syscall holder */ + .long sys_access + .long sys_nice + .long sys_ni_syscall /* 35 old ftime syscall holder */ + .long sys_sync + .long sys_kill + .long sys_rename + .long sys_mkdir + .long sys_rmdir /* 40 */ + .long sys_dup + .long sys_pipe + .long sys_times + .long sys_ni_syscall /* old prof syscall holder */ + .long sys_brk /* 45 */ + .long sys_setgid16 + .long sys_getgid16 + .long sys_signal + .long sys_geteuid16 + .long sys_getegid16 /* 50 */ + .long sys_acct + .long sys_umount /* recycled never used phys( */ + .long sys_ni_syscall /* old lock syscall holder */ + .long sys_ioctl + .long sys_fcntl /* 55 */ + .long sys_ni_syscall /* old mpx syscall holder */ + .long sys_setpgid + .long sys_ni_syscall /* old ulimit syscall holder */ + .long sys_ni_syscall /* old sys_olduname holder */ + .long sys_umask /* 60 */ + .long sys_chroot + .long sys_ustat + .long sys_dup2 + .long sys_getppid + .long sys_getpgrp /* 65 */ + .long sys_setsid + .long sys_sigaction + .long sys_sgetmask + .long sys_ssetmask + .long sys_setreuid16 /* 70 */ + .long sys_setregid16 + .long sys_sigsuspend + .long sys_sigpending + .long sys_sethostname + .long sys_setrlimit /* 75 */ + .long sys_old_getrlimit + .long sys_getrusage + .long sys_gettimeofday + .long sys_settimeofday + .long sys_getgroups16 /* 80 */ + .long sys_setgroups16 + .long sys_select /* was old_select in Linux/E100 */ + .long sys_symlink + .long sys_lstat + .long sys_readlink /* 85 */ + .long sys_uselib + .long sys_swapon + .long sys_reboot + .long old_readdir + .long old_mmap /* 90 */ + .long sys_munmap + .long sys_truncate + .long sys_ftruncate + .long sys_fchmod + .long sys_fchown16 /* 95 */ + .long sys_getpriority + .long sys_setpriority + .long sys_ni_syscall /* old profil syscall holder */ + .long sys_statfs + .long sys_fstatfs /* 100 */ + .long sys_ni_syscall /* sys_ioperm in i386 */ + .long sys_socketcall + .long sys_syslog + .long sys_setitimer + .long sys_getitimer /* 105 */ + .long sys_newstat + .long sys_newlstat + .long sys_newfstat + .long sys_ni_syscall /* old sys_uname holder */ + .long sys_ni_syscall /* sys_iopl in i386 */ + .long sys_vhangup + .long sys_ni_syscall /* old "idle" system call */ + .long sys_ni_syscall /* vm86old in i386 */ + .long sys_wait4 + .long sys_swapoff /* 115 */ + .long sys_sysinfo + .long sys_ipc + .long sys_fsync + .long sys_sigreturn + .long sys_clone /* 120 */ + .long sys_setdomainname + .long sys_newuname + .long sys_ni_syscall /* sys_modify_ldt */ + .long sys_adjtimex + .long sys_mprotect /* 125 */ + .long sys_sigprocmask + .long sys_ni_syscall /* old "create_module" */ + .long sys_init_module + .long sys_delete_module + .long sys_ni_syscall /* 130: old "get_kernel_syms" */ + .long sys_quotactl + .long sys_getpgid + .long sys_fchdir + .long sys_bdflush + .long sys_sysfs /* 135 */ + .long sys_personality + .long sys_ni_syscall /* for afs_syscall */ + .long sys_setfsuid16 + .long sys_setfsgid16 + .long sys_llseek /* 140 */ + .long sys_getdents + .long sys_select + .long sys_flock + .long sys_msync + .long sys_readv /* 145 */ + .long sys_writev + .long sys_getsid + .long sys_fdatasync + .long sys_sysctl + .long sys_mlock /* 150 */ + .long sys_munlock + .long sys_mlockall + .long sys_munlockall + .long sys_sched_setparam + .long sys_sched_getparam /* 155 */ + .long sys_sched_setscheduler + .long sys_sched_getscheduler + .long sys_sched_yield + .long sys_sched_get_priority_max + .long sys_sched_get_priority_min /* 160 */ + .long sys_sched_rr_get_interval + .long sys_nanosleep + .long sys_mremap + .long sys_setresuid16 + .long sys_getresuid16 /* 165 */ + .long sys_ni_syscall /* sys_vm86 */ + .long sys_ni_syscall /* Old sys_query_module */ + .long sys_poll + .long sys_nfsservctl + .long sys_setresgid16 /* 170 */ + .long sys_getresgid16 + .long sys_prctl + .long sys_rt_sigreturn + .long sys_rt_sigaction + .long sys_rt_sigprocmask /* 175 */ + .long sys_rt_sigpending + .long sys_rt_sigtimedwait + .long sys_rt_sigqueueinfo + .long sys_rt_sigsuspend + .long sys_pread64 /* 180 */ + .long sys_pwrite64 + .long sys_chown16 + .long sys_getcwd + .long sys_capget + .long sys_capset /* 185 */ + .long sys_sigaltstack + .long sys_sendfile + .long sys_ni_syscall /* streams1 */ + .long sys_ni_syscall /* streams2 */ + .long sys_vfork /* 190 */ + .long sys_getrlimit + .long sys_mmap2 + .long sys_truncate64 + .long sys_ftruncate64 + .long sys_stat64 /* 195 */ + .long sys_lstat64 + .long sys_fstat64 + .long sys_lchown + .long sys_getuid + .long sys_getgid /* 200 */ + .long sys_geteuid + .long sys_getegid + .long sys_setreuid + .long sys_setregid + .long sys_getgroups /* 205 */ + .long sys_setgroups + .long sys_fchown + .long sys_setresuid + .long sys_getresuid + .long sys_setresgid /* 210 */ + .long sys_getresgid + .long sys_chown + .long sys_setuid + .long sys_setgid + .long sys_setfsuid /* 215 */ + .long sys_setfsgid + .long sys_pivot_root + .long sys_mincore + .long sys_madvise + .long sys_getdents64 /* 220 */ + .long sys_fcntl64 + .long sys_ni_syscall /* reserved for TUX */ + .long sys_ni_syscall + .long sys_gettid + .long sys_readahead /* 225 */ + .long sys_setxattr + .long sys_lsetxattr + .long sys_fsetxattr + .long sys_getxattr + .long sys_lgetxattr /* 230 */ + .long sys_fgetxattr + .long sys_listxattr + .long sys_llistxattr + .long sys_flistxattr + .long sys_removexattr /* 235 */ + .long sys_lremovexattr + .long sys_fremovexattr + .long sys_tkill + .long sys_sendfile64 + .long sys_futex /* 240 */ + .long sys_sched_setaffinity + .long sys_sched_getaffinity + .long sys_ni_syscall /* sys_set_thread_area */ + .long sys_ni_syscall /* sys_get_thread_area */ + .long sys_io_setup /* 245 */ + .long sys_io_destroy + .long sys_io_getevents + .long sys_io_submit + .long sys_io_cancel + .long sys_fadvise64 /* 250 */ + .long sys_ni_syscall + .long sys_exit_group + .long sys_lookup_dcookie + .long sys_epoll_create + .long sys_epoll_ctl /* 255 */ + .long sys_epoll_wait + .long sys_remap_file_pages + .long sys_set_tid_address + .long sys_timer_create + .long sys_timer_settime /* 260 */ + .long sys_timer_gettime + .long sys_timer_getoverrun + .long sys_timer_delete + .long sys_clock_settime + .long sys_clock_gettime /* 265 */ + .long sys_clock_getres + .long sys_clock_nanosleep + .long sys_statfs64 + .long sys_fstatfs64 + .long sys_tgkill /* 270 */ + .long sys_utimes + .long sys_fadvise64_64 + .long sys_ni_syscall /* sys_vserver */ + .long sys_ni_syscall /* sys_mbind */ + .long sys_ni_syscall /* 275 sys_get_mempolicy */ + .long sys_ni_syscall /* sys_set_mempolicy */ + .long sys_mq_open + .long sys_mq_unlink + .long sys_mq_timedsend + .long sys_mq_timedreceive /* 280 */ + .long sys_mq_notify + .long sys_mq_getsetattr + .long sys_ni_syscall /* reserved for kexec */ + .long sys_waitid + + /* + * NOTE!! This doesn't have to be exact - we just have + * to make sure we have _enough_ of the "sys_ni_syscall" + * entries. Don't panic if you notice that this hasn't + * been shrunk every time we add a new system call. + */ + + .rept NR_syscalls - (.-sys_call_table) / 4 + .long sys_ni_syscall + .endr + diff --git a/arch/cris/arch-v32/kernel/fasttimer.c b/arch/cris/arch-v32/kernel/fasttimer.c new file mode 100644 index 000000000000..ea2b4a97c8c7 --- /dev/null +++ b/arch/cris/arch-v32/kernel/fasttimer.c @@ -0,0 +1,996 @@ +/* $Id: fasttimer.c,v 1.11 2005/01/04 11:15:46 starvik Exp $ + * linux/arch/cris/kernel/fasttimer.c + * + * Fast timers for ETRAX FS + * This may be useful in other OS than Linux so use 2 space indentation... + * + * $Log: fasttimer.c,v $ + * Revision 1.11 2005/01/04 11:15:46 starvik + * Don't share timer IRQ. + * + * Revision 1.10 2004/12/07 09:19:38 starvik + * Corrected includes. + * Use correct interrupt macros. + * + * Revision 1.9 2004/05/14 10:18:58 starvik + * Export fast_timer_list + * + * Revision 1.8 2004/05/14 07:58:03 starvik + * Merge of changes from 2.4 + * + * Revision 1.7 2003/07/10 12:06:14 starvik + * Return IRQ_NONE if irq wasn't handled + * + * Revision 1.6 2003/07/04 08:27:49 starvik + * Merge of Linux 2.5.74 + * + * Revision 1.5 2003/06/05 10:16:22 johana + * New INTR_VECT macros. + * + * Revision 1.4 2003/06/03 08:49:45 johana + * Fixed typo. + * + * Revision 1.3 2003/06/02 12:51:27 johana + * Now compiles. + * Commented some include files that probably can be removed. + * + * Revision 1.2 2003/06/02 12:09:41 johana + * Ported to ETRAX FS using the trig interrupt instead of timer1. + * + * Revision 1.3 2002/12/12 08:26:32 starvik + * Don't use C-comments inside CVS comments + * + * Revision 1.2 2002/12/11 15:42:02 starvik + * Extracted v10 (ETRAX 100LX) specific stuff from arch/cris/kernel/ + * + * Revision 1.1 2002/11/18 07:58:06 starvik + * Fast timers (from Linux 2.4) + * + * Revision 1.5 2002/10/15 06:21:39 starvik + * Added call to init_waitqueue_head + * + * Revision 1.4 2002/05/28 17:47:59 johana + * Added del_fast_timer() + * + * Revision 1.3 2002/05/28 16:16:07 johana + * Handle empty fast_timer_list + * + * Revision 1.2 2002/05/27 15:38:42 johana + * Made it compile without warnings on Linux 2.4. + * (includes, wait_queue, PROC_FS and snprintf) + * + * Revision 1.1 2002/05/27 15:32:25 johana + * arch/etrax100/kernel/fasttimer.c v1.8 from the elinux tree. + * + * Revision 1.8 2001/11/27 13:50:40 pkj + * Disable interrupts while stopping the timer and while modifying the + * list of active timers in timer1_handler() as it may be interrupted + * by other interrupts (e.g., the serial interrupt) which may add fast + * timers. + * + * Revision 1.7 2001/11/22 11:50:32 pkj + * * Only store information about the last 16 timers. + * * proc_fasttimer_read() now uses an allocated buffer, since it + * requires more space than just a page even for only writing the + * last 16 timers. The buffer is only allocated on request, so + * unless /proc/fasttimer is read, it is never allocated. + * * Renamed fast_timer_started to fast_timers_started to match + * fast_timers_added and fast_timers_expired. + * * Some clean-up. + * + * Revision 1.6 2000/12/13 14:02:08 johana + * Removed volatile for fast_timer_list + * + * Revision 1.5 2000/12/13 13:55:35 johana + * Added DEBUG_LOG, added som cli() and cleanup + * + * Revision 1.4 2000/12/05 13:48:50 johana + * Added range check when writing proc file, modified timer int handling + * + * Revision 1.3 2000/11/23 10:10:20 johana + * More debug/logging possibilities. + * Moved GET_JIFFIES_USEC() to timex.h and time.c + * + * Revision 1.2 2000/11/01 13:41:04 johana + * Clean up and bugfixes. + * Created new do_gettimeofday_fast() that gets a timeval struct + * with time based on jiffies and *R_TIMER0_DATA, uses a table + * for fast conversion of timer value to microseconds. + * (Much faster the standard do_gettimeofday() and we don't really + * wan't to use the true time - we wan't the "uptime" so timers don't screw up + * when we change the time. + * TODO: Add efficient support for continuous timers as well. + * + * Revision 1.1 2000/10/26 15:49:16 johana + * Added fasttimer, highresolution timers. + * + * Copyright (C) 2000,2001 2002, 2003 Axis Communications AB, Lund, Sweden + */ + +#include <linux/errno.h> +#include <linux/sched.h> +#include <linux/kernel.h> +#include <linux/param.h> +#include <linux/string.h> +#include <linux/vmalloc.h> +#include <linux/interrupt.h> +#include <linux/time.h> +#include <linux/delay.h> + +#include <asm/irq.h> +#include <asm/system.h> + +#include <linux/config.h> +#include <linux/version.h> + +#include <asm/arch/hwregs/reg_map.h> +#include <asm/arch/hwregs/reg_rdwr.h> +#include <asm/arch/hwregs/timer_defs.h> +#include <asm/fasttimer.h> +#include <linux/proc_fs.h> + +/* + * timer0 is running at 100MHz and generating jiffies timer ticks + * at 100 or 1000 HZ. + * fasttimer gives an API that gives timers that expire "between" the jiffies + * giving microsecond resolution (10 ns). + * fasttimer uses reg_timer_rw_trig register to get interrupt when + * r_time reaches a certain value. + */ + + +#define DEBUG_LOG_INCLUDED +#define FAST_TIMER_LOG +//#define FAST_TIMER_TEST + +#define FAST_TIMER_SANITY_CHECKS + +#ifdef FAST_TIMER_SANITY_CHECKS +#define SANITYCHECK(x) x +static int sanity_failed = 0; +#else +#define SANITYCHECK(x) +#endif + +#define D1(x) +#define D2(x) +#define DP(x) + +#define __INLINE__ inline + +static int fast_timer_running = 0; +static int fast_timers_added = 0; +static int fast_timers_started = 0; +static int fast_timers_expired = 0; +static int fast_timers_deleted = 0; +static int fast_timer_is_init = 0; +static int fast_timer_ints = 0; + +struct fast_timer *fast_timer_list = NULL; + +#ifdef DEBUG_LOG_INCLUDED +#define DEBUG_LOG_MAX 128 +static const char * debug_log_string[DEBUG_LOG_MAX]; +static unsigned long debug_log_value[DEBUG_LOG_MAX]; +static int debug_log_cnt = 0; +static int debug_log_cnt_wrapped = 0; + +#define DEBUG_LOG(string, value) \ +{ \ + unsigned long log_flags; \ + local_irq_save(log_flags); \ + debug_log_string[debug_log_cnt] = (string); \ + debug_log_value[debug_log_cnt] = (unsigned long)(value); \ + if (++debug_log_cnt >= DEBUG_LOG_MAX) \ + { \ + debug_log_cnt = debug_log_cnt % DEBUG_LOG_MAX; \ + debug_log_cnt_wrapped = 1; \ + } \ + local_irq_restore(log_flags); \ +} +#else +#define DEBUG_LOG(string, value) +#endif + + +#define NUM_TIMER_STATS 16 +#ifdef FAST_TIMER_LOG +struct fast_timer timer_added_log[NUM_TIMER_STATS]; +struct fast_timer timer_started_log[NUM_TIMER_STATS]; +struct fast_timer timer_expired_log[NUM_TIMER_STATS]; +#endif + +int timer_div_settings[NUM_TIMER_STATS]; +int timer_delay_settings[NUM_TIMER_STATS]; + + +static void +timer_trig_handler(void); + + + +/* Not true gettimeofday, only checks the jiffies (uptime) + useconds */ +void __INLINE__ do_gettimeofday_fast(struct timeval *tv) +{ + unsigned long sec = jiffies; + unsigned long usec = GET_JIFFIES_USEC(); + + usec += (sec % HZ) * (1000000 / HZ); + sec = sec / HZ; + + if (usec > 1000000) + { + usec -= 1000000; + sec++; + } + tv->tv_sec = sec; + tv->tv_usec = usec; +} + +int __INLINE__ timeval_cmp(struct timeval *t0, struct timeval *t1) +{ + if (t0->tv_sec < t1->tv_sec) + { + return -1; + } + else if (t0->tv_sec > t1->tv_sec) + { + return 1; + } + if (t0->tv_usec < t1->tv_usec) + { + return -1; + } + else if (t0->tv_usec > t1->tv_usec) + { + return 1; + } + return 0; +} + +/* Called with ints off */ +void __INLINE__ start_timer_trig(unsigned long delay_us) +{ + reg_timer_rw_ack_intr ack_intr = { 0 }; + reg_timer_rw_intr_mask intr_mask; + reg_timer_rw_trig trig; + reg_timer_rw_trig_cfg trig_cfg = { 0 }; + reg_timer_r_time r_time; + + r_time = REG_RD(timer, regi_timer, r_time); + + D1(printk("start_timer_trig : %d us freq: %i div: %i\n", + delay_us, freq_index, div)); + /* Clear trig irq */ + intr_mask = REG_RD(timer, regi_timer, rw_intr_mask); + intr_mask.trig = 0; + REG_WR(timer, regi_timer, rw_intr_mask, intr_mask); + + /* Set timer values */ + /* r_time is 100MHz (10 ns resolution) */ + trig = r_time + delay_us*(1000/10); + + timer_div_settings[fast_timers_started % NUM_TIMER_STATS] = trig; + timer_delay_settings[fast_timers_started % NUM_TIMER_STATS] = delay_us; + + /* Ack interrupt */ + ack_intr.trig = 1; + REG_WR(timer, regi_timer, rw_ack_intr, ack_intr); + + /* Start timer */ + REG_WR(timer, regi_timer, rw_trig, trig); + trig_cfg.tmr = regk_timer_time; + REG_WR(timer, regi_timer, rw_trig_cfg, trig_cfg); + + /* Check if we have already passed the trig time */ + r_time = REG_RD(timer, regi_timer, r_time); + if (r_time < trig) { + /* No, Enable trig irq */ + intr_mask = REG_RD(timer, regi_timer, rw_intr_mask); + intr_mask.trig = 1; + REG_WR(timer, regi_timer, rw_intr_mask, intr_mask); + fast_timers_started++; + fast_timer_running = 1; + } + else + { + /* We have passed the time, disable trig point, ack intr */ + trig_cfg.tmr = regk_timer_off; + REG_WR(timer, regi_timer, rw_trig_cfg, trig_cfg); + REG_WR(timer, regi_timer, rw_ack_intr, ack_intr); + /* call the int routine directly */ + timer_trig_handler(); + } + +} + +/* In version 1.4 this function takes 27 - 50 us */ +void start_one_shot_timer(struct fast_timer *t, + fast_timer_function_type *function, + unsigned long data, + unsigned long delay_us, + const char *name) +{ + unsigned long flags; + struct fast_timer *tmp; + + D1(printk("sft %s %d us\n", name, delay_us)); + + local_irq_save(flags); + + do_gettimeofday_fast(&t->tv_set); + tmp = fast_timer_list; + + SANITYCHECK({ /* Check so this is not in the list already... */ + while (tmp != NULL) + { + if (tmp == t) + { + printk("timer name: %s data: 0x%08lX already in list!\n", name, data); + sanity_failed++; + return; + } + else + { + tmp = tmp->next; + } + } + tmp = fast_timer_list; + }); + + t->delay_us = delay_us; + t->function = function; + t->data = data; + t->name = name; + + t->tv_expires.tv_usec = t->tv_set.tv_usec + delay_us % 1000000; + t->tv_expires.tv_sec = t->tv_set.tv_sec + delay_us / 1000000; + if (t->tv_expires.tv_usec > 1000000) + { + t->tv_expires.tv_usec -= 1000000; + t->tv_expires.tv_sec++; + } +#ifdef FAST_TIMER_LOG + timer_added_log[fast_timers_added % NUM_TIMER_STATS] = *t; +#endif + fast_timers_added++; + + /* Check if this should timeout before anything else */ + if (tmp == NULL || timeval_cmp(&t->tv_expires, &tmp->tv_expires) < 0) + { + /* Put first in list and modify the timer value */ + t->prev = NULL; + t->next = fast_timer_list; + if (fast_timer_list) + { + fast_timer_list->prev = t; + } + fast_timer_list = t; +#ifdef FAST_TIMER_LOG + timer_started_log[fast_timers_started % NUM_TIMER_STATS] = *t; +#endif + start_timer_trig(delay_us); + } else { + /* Put in correct place in list */ + while (tmp->next && + timeval_cmp(&t->tv_expires, &tmp->next->tv_expires) > 0) + { + tmp = tmp->next; + } + /* Insert t after tmp */ + t->prev = tmp; + t->next = tmp->next; + if (tmp->next) + { + tmp->next->prev = t; + } + tmp->next = t; + } + + D2(printk("start_one_shot_timer: %d us done\n", delay_us)); + + local_irq_restore(flags); +} /* start_one_shot_timer */ + +static inline int fast_timer_pending (const struct fast_timer * t) +{ + return (t->next != NULL) || (t->prev != NULL) || (t == fast_timer_list); +} + +static inline int detach_fast_timer (struct fast_timer *t) +{ + struct fast_timer *next, *prev; + if (!fast_timer_pending(t)) + return 0; + next = t->next; + prev = t->prev; + if (next) + next->prev = prev; + if (prev) + prev->next = next; + else + fast_timer_list = next; + fast_timers_deleted++; + return 1; +} + +int del_fast_timer(struct fast_timer * t) +{ + unsigned long flags; + int ret; + + local_irq_save(flags); + ret = detach_fast_timer(t); + t->next = t->prev = NULL; + local_irq_restore(flags); + return ret; +} /* del_fast_timer */ + + +/* Interrupt routines or functions called in interrupt context */ + +/* Timer interrupt handler for trig interrupts */ + +static irqreturn_t +timer_trig_interrupt(int irq, void *dev_id, struct pt_regs *regs) +{ + reg_timer_r_masked_intr masked_intr; + + /* Check if the timer interrupt is for us (a trig int) */ + masked_intr = REG_RD(timer, regi_timer, r_masked_intr); + if (!masked_intr.trig) + return IRQ_NONE; + timer_trig_handler(); + return IRQ_HANDLED; +} + +static void timer_trig_handler(void) +{ + reg_timer_rw_ack_intr ack_intr = { 0 }; + reg_timer_rw_intr_mask intr_mask; + reg_timer_rw_trig_cfg trig_cfg = { 0 }; + struct fast_timer *t; + unsigned long flags; + + local_irq_save(flags); + + /* Clear timer trig interrupt */ + intr_mask = REG_RD(timer, regi_timer, rw_intr_mask); + intr_mask.trig = 0; + REG_WR(timer, regi_timer, rw_intr_mask, intr_mask); + + /* First stop timer, then ack interrupt */ + /* Stop timer */ + trig_cfg.tmr = regk_timer_off; + REG_WR(timer, regi_timer, rw_trig_cfg, trig_cfg); + + /* Ack interrupt */ + ack_intr.trig = 1; + REG_WR(timer, regi_timer, rw_ack_intr, ack_intr); + + fast_timer_running = 0; + fast_timer_ints++; + + local_irq_restore(flags); + + t = fast_timer_list; + while (t) + { + struct timeval tv; + + /* Has it really expired? */ + do_gettimeofday_fast(&tv); + D1(printk("t: %is %06ius\n", tv.tv_sec, tv.tv_usec)); + + if (timeval_cmp(&t->tv_expires, &tv) <= 0) + { + /* Yes it has expired */ +#ifdef FAST_TIMER_LOG + timer_expired_log[fast_timers_expired % NUM_TIMER_STATS] = *t; +#endif + fast_timers_expired++; + + /* Remove this timer before call, since it may reuse the timer */ + local_irq_save(flags); + if (t->prev) + { + t->prev->next = t->next; + } + else + { + fast_timer_list = t->next; + } + if (t->next) + { + t->next->prev = t->prev; + } + t->prev = NULL; + t->next = NULL; + local_irq_restore(flags); + + if (t->function != NULL) + { + t->function(t->data); + } + else + { + DEBUG_LOG("!trimertrig %i function==NULL!\n", fast_timer_ints); + } + } + else + { + /* Timer is to early, let's set it again using the normal routines */ + D1(printk(".\n")); + } + + local_irq_save(flags); + if ((t = fast_timer_list) != NULL) + { + /* Start next timer.. */ + long us; + struct timeval tv; + + do_gettimeofday_fast(&tv); + us = ((t->tv_expires.tv_sec - tv.tv_sec) * 1000000 + + t->tv_expires.tv_usec - tv.tv_usec); + if (us > 0) + { + if (!fast_timer_running) + { +#ifdef FAST_TIMER_LOG + timer_started_log[fast_timers_started % NUM_TIMER_STATS] = *t; +#endif + start_timer_trig(us); + } + local_irq_restore(flags); + break; + } + else + { + /* Timer already expired, let's handle it better late than never. + * The normal loop handles it + */ + D1(printk("e! %d\n", us)); + } + } + local_irq_restore(flags); + } + + if (!t) + { + D1(printk("ttrig stop!\n")); + } +} + +static void wake_up_func(unsigned long data) +{ +#ifdef DECLARE_WAITQUEUE + wait_queue_head_t *sleep_wait_p = (wait_queue_head_t*)data; +#else + struct wait_queue **sleep_wait_p = (struct wait_queue **)data; +#endif + wake_up(sleep_wait_p); +} + + +/* Useful API */ + +void schedule_usleep(unsigned long us) +{ + struct fast_timer t; +#ifdef DECLARE_WAITQUEUE + wait_queue_head_t sleep_wait; + init_waitqueue_head(&sleep_wait); + { + DECLARE_WAITQUEUE(wait, current); +#else + struct wait_queue *sleep_wait = NULL; + struct wait_queue wait = { current, NULL }; +#endif + + D1(printk("schedule_usleep(%d)\n", us)); + add_wait_queue(&sleep_wait, &wait); + set_current_state(TASK_INTERRUPTIBLE); + start_one_shot_timer(&t, wake_up_func, (unsigned long)&sleep_wait, us, + "usleep"); + schedule(); + set_current_state(TASK_RUNNING); + remove_wait_queue(&sleep_wait, &wait); + D1(printk("done schedule_usleep(%d)\n", us)); +#ifdef DECLARE_WAITQUEUE + } +#endif +} + +#ifdef CONFIG_PROC_FS +static int proc_fasttimer_read(char *buf, char **start, off_t offset, int len +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,2,0) + ,int *eof, void *data_unused +#else + ,int unused +#endif + ); +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,2,0) +static struct proc_dir_entry *fasttimer_proc_entry; +#else +static struct proc_dir_entry fasttimer_proc_entry = +{ + 0, 9, "fasttimer", + S_IFREG | S_IRUGO, 1, 0, 0, + 0, NULL /* ops -- default to array */, + &proc_fasttimer_read /* get_info */, +}; +#endif +#endif /* CONFIG_PROC_FS */ + +#ifdef CONFIG_PROC_FS + +/* This value is very much based on testing */ +#define BIG_BUF_SIZE (500 + NUM_TIMER_STATS * 300) + +static int proc_fasttimer_read(char *buf, char **start, off_t offset, int len +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,2,0) + ,int *eof, void *data_unused +#else + ,int unused +#endif + ) +{ + unsigned long flags; + int i = 0; + int num_to_show; + struct timeval tv; + struct fast_timer *t, *nextt; + static char *bigbuf = NULL; + static unsigned long used; + + if (!bigbuf && !(bigbuf = vmalloc(BIG_BUF_SIZE))) + { + used = 0; + bigbuf[0] = '\0'; + return 0; + } + + if (!offset || !used) + { + do_gettimeofday_fast(&tv); + + used = 0; + used += sprintf(bigbuf + used, "Fast timers added: %i\n", + fast_timers_added); + used += sprintf(bigbuf + used, "Fast timers started: %i\n", + fast_timers_started); + used += sprintf(bigbuf + used, "Fast timer interrupts: %i\n", + fast_timer_ints); + used += sprintf(bigbuf + used, "Fast timers expired: %i\n", + fast_timers_expired); + used += sprintf(bigbuf + used, "Fast timers deleted: %i\n", + fast_timers_deleted); + used += sprintf(bigbuf + used, "Fast timer running: %s\n", + fast_timer_running ? "yes" : "no"); + used += sprintf(bigbuf + used, "Current time: %lu.%06lu\n", + (unsigned long)tv.tv_sec, + (unsigned long)tv.tv_usec); +#ifdef FAST_TIMER_SANITY_CHECKS + used += sprintf(bigbuf + used, "Sanity failed: %i\n", + sanity_failed); +#endif + used += sprintf(bigbuf + used, "\n"); + +#ifdef DEBUG_LOG_INCLUDED + { + int end_i = debug_log_cnt; + i = 0; + + if (debug_log_cnt_wrapped) + { + i = debug_log_cnt; + } + + while ((i != end_i || (debug_log_cnt_wrapped && !used)) && + used+100 < BIG_BUF_SIZE) + { + used += sprintf(bigbuf + used, debug_log_string[i], + debug_log_value[i]); + i = (i+1) % DEBUG_LOG_MAX; + } + } + used += sprintf(bigbuf + used, "\n"); +#endif + + num_to_show = (fast_timers_started < NUM_TIMER_STATS ? fast_timers_started: + NUM_TIMER_STATS); + used += sprintf(bigbuf + used, "Timers started: %i\n", fast_timers_started); + for (i = 0; i < num_to_show && (used+100 < BIG_BUF_SIZE) ; i++) + { + int cur = (fast_timers_started - i - 1) % NUM_TIMER_STATS; + +#if 1 //ndef FAST_TIMER_LOG + used += sprintf(bigbuf + used, "div: %i delay: %i" + "\n", + timer_div_settings[cur], + timer_delay_settings[cur] + ); +#endif +#ifdef FAST_TIMER_LOG + t = &timer_started_log[cur]; + used += sprintf(bigbuf + used, "%-14s s: %6lu.%06lu e: %6lu.%06lu " + "d: %6li us data: 0x%08lX" + "\n", + t->name, + (unsigned long)t->tv_set.tv_sec, + (unsigned long)t->tv_set.tv_usec, + (unsigned long)t->tv_expires.tv_sec, + (unsigned long)t->tv_expires.tv_usec, + t->delay_us, + t->data + ); +#endif + } + used += sprintf(bigbuf + used, "\n"); + +#ifdef FAST_TIMER_LOG + num_to_show = (fast_timers_added < NUM_TIMER_STATS ? fast_timers_added: + NUM_TIMER_STATS); + used += sprintf(bigbuf + used, "Timers added: %i\n", fast_timers_added); + for (i = 0; i < num_to_show && (used+100 < BIG_BUF_SIZE); i++) + { + t = &timer_added_log[(fast_timers_added - i - 1) % NUM_TIMER_STATS]; + used += sprintf(bigbuf + used, "%-14s s: %6lu.%06lu e: %6lu.%06lu " + "d: %6li us data: 0x%08lX" + "\n", + t->name, + (unsigned long)t->tv_set.tv_sec, + (unsigned long)t->tv_set.tv_usec, + (unsigned long)t->tv_expires.tv_sec, + (unsigned long)t->tv_expires.tv_usec, + t->delay_us, + t->data + ); + } + used += sprintf(bigbuf + used, "\n"); + + num_to_show = (fast_timers_expired < NUM_TIMER_STATS ? fast_timers_expired: + NUM_TIMER_STATS); + used += sprintf(bigbuf + used, "Timers expired: %i\n", fast_timers_expired); + for (i = 0; i < num_to_show && (used+100 < BIG_BUF_SIZE); i++) + { + t = &timer_expired_log[(fast_timers_expired - i - 1) % NUM_TIMER_STATS]; + used += sprintf(bigbuf + used, "%-14s s: %6lu.%06lu e: %6lu.%06lu " + "d: %6li us data: 0x%08lX" + "\n", + t->name, + (unsigned long)t->tv_set.tv_sec, + (unsigned long)t->tv_set.tv_usec, + (unsigned long)t->tv_expires.tv_sec, + (unsigned long)t->tv_expires.tv_usec, + t->delay_us, + t->data + ); + } + used += sprintf(bigbuf + used, "\n"); +#endif + + used += sprintf(bigbuf + used, "Active timers:\n"); + local_irq_save(flags); + local_irq_save(flags); + t = fast_timer_list; + while (t != NULL && (used+100 < BIG_BUF_SIZE)) + { + nextt = t->next; + local_irq_restore(flags); + used += sprintf(bigbuf + used, "%-14s s: %6lu.%06lu e: %6lu.%06lu " + "d: %6li us data: 0x%08lX" +/* " func: 0x%08lX" */ + "\n", + t->name, + (unsigned long)t->tv_set.tv_sec, + (unsigned long)t->tv_set.tv_usec, + (unsigned long)t->tv_expires.tv_sec, + (unsigned long)t->tv_expires.tv_usec, + t->delay_us, + t->data +/* , t->function */ + ); + local_irq_disable(); + if (t->next != nextt) + { + printk("timer removed!\n"); + } + t = nextt; + } + local_irq_restore(flags); + } + + if (used - offset < len) + { + len = used - offset; + } + + memcpy(buf, bigbuf + offset, len); + *start = buf; +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,2,0) + *eof = 1; +#endif + + return len; +} +#endif /* PROC_FS */ + +#ifdef FAST_TIMER_TEST +static volatile unsigned long i = 0; +static volatile int num_test_timeout = 0; +static struct fast_timer tr[10]; +static int exp_num[10]; + +static struct timeval tv_exp[100]; + +static void test_timeout(unsigned long data) +{ + do_gettimeofday_fast(&tv_exp[data]); + exp_num[data] = num_test_timeout; + + num_test_timeout++; +} + +static void test_timeout1(unsigned long data) +{ + do_gettimeofday_fast(&tv_exp[data]); + exp_num[data] = num_test_timeout; + if (data < 7) + { + start_one_shot_timer(&tr[i], test_timeout1, i, 1000, "timeout1"); + i++; + } + num_test_timeout++; +} + +DP( +static char buf0[2000]; +static char buf1[2000]; +static char buf2[2000]; +static char buf3[2000]; +static char buf4[2000]; +); + +static char buf5[6000]; +static int j_u[1000]; + +static void fast_timer_test(void) +{ + int prev_num; + int j; + + struct timeval tv, tv0, tv1, tv2; + + printk("fast_timer_test() start\n"); + do_gettimeofday_fast(&tv); + + for (j = 0; j < 1000; j++) + { + j_u[j] = GET_JIFFIES_USEC(); + } + for (j = 0; j < 100; j++) + { + do_gettimeofday_fast(&tv_exp[j]); + } + printk("fast_timer_test() %is %06i\n", tv.tv_sec, tv.tv_usec); + + for (j = 0; j < 1000; j++) + { + printk("%i %i %i %i %i\n",j_u[j], j_u[j+1], j_u[j+2], j_u[j+3], j_u[j+4]); + j += 4; + } + for (j = 0; j < 100; j++) + { + printk("%i.%i %i.%i %i.%i %i.%i %i.%i\n", + tv_exp[j].tv_sec,tv_exp[j].tv_usec, + tv_exp[j+1].tv_sec,tv_exp[j+1].tv_usec, + tv_exp[j+2].tv_sec,tv_exp[j+2].tv_usec, + tv_exp[j+3].tv_sec,tv_exp[j+3].tv_usec, + tv_exp[j+4].tv_sec,tv_exp[j+4].tv_usec); + j += 4; + } + do_gettimeofday_fast(&tv0); + start_one_shot_timer(&tr[i], test_timeout, i, 50000, "test0"); + DP(proc_fasttimer_read(buf0, NULL, 0, 0, 0)); + i++; + start_one_shot_timer(&tr[i], test_timeout, i, 70000, "test1"); + DP(proc_fasttimer_read(buf1, NULL, 0, 0, 0)); + i++; + start_one_shot_timer(&tr[i], test_timeout, i, 40000, "test2"); + DP(proc_fasttimer_read(buf2, NULL, 0, 0, 0)); + i++; + start_one_shot_timer(&tr[i], test_timeout, i, 60000, "test3"); + DP(proc_fasttimer_read(buf3, NULL, 0, 0, 0)); + i++; + start_one_shot_timer(&tr[i], test_timeout1, i, 55000, "test4xx"); + DP(proc_fasttimer_read(buf4, NULL, 0, 0, 0)); + i++; + do_gettimeofday_fast(&tv1); + + proc_fasttimer_read(buf5, NULL, 0, 0, 0); + + prev_num = num_test_timeout; + while (num_test_timeout < i) + { + if (num_test_timeout != prev_num) + { + prev_num = num_test_timeout; + } + } + do_gettimeofday_fast(&tv2); + printk("Timers started %is %06i\n", tv0.tv_sec, tv0.tv_usec); + printk("Timers started at %is %06i\n", tv1.tv_sec, tv1.tv_usec); + printk("Timers done %is %06i\n", tv2.tv_sec, tv2.tv_usec); + DP(printk("buf0:\n"); + printk(buf0); + printk("buf1:\n"); + printk(buf1); + printk("buf2:\n"); + printk(buf2); + printk("buf3:\n"); + printk(buf3); + printk("buf4:\n"); + printk(buf4); + ); + printk("buf5:\n"); + printk(buf5); + + printk("timers set:\n"); + for(j = 0; j<i; j++) + { + struct fast_timer *t = &tr[j]; + printk("%-10s set: %6is %06ius exp: %6is %06ius " + "data: 0x%08X func: 0x%08X\n", + t->name, + t->tv_set.tv_sec, + t->tv_set.tv_usec, + t->tv_expires.tv_sec, + t->tv_expires.tv_usec, + t->data, + t->function + ); + + printk(" del: %6ius did exp: %6is %06ius as #%i error: %6li\n", + t->delay_us, + tv_exp[j].tv_sec, + tv_exp[j].tv_usec, + exp_num[j], + (tv_exp[j].tv_sec - t->tv_expires.tv_sec)*1000000 + tv_exp[j].tv_usec - t->tv_expires.tv_usec); + } + proc_fasttimer_read(buf5, NULL, 0, 0, 0); + printk("buf5 after all done:\n"); + printk(buf5); + printk("fast_timer_test() done\n"); +} +#endif + + +void fast_timer_init(void) +{ + /* For some reason, request_irq() hangs when called froom time_init() */ + if (!fast_timer_is_init) + { + printk("fast_timer_init()\n"); + +#ifdef CONFIG_PROC_FS +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,2,0) + if ((fasttimer_proc_entry = create_proc_entry( "fasttimer", 0, 0 ))) + fasttimer_proc_entry->read_proc = proc_fasttimer_read; +#else + proc_register_dynamic(&proc_root, &fasttimer_proc_entry); +#endif +#endif /* PROC_FS */ + if(request_irq(TIMER_INTR_VECT, timer_trig_interrupt, SA_INTERRUPT, + "fast timer int", NULL)) + { + printk("err: timer1 irq\n"); + } + fast_timer_is_init = 1; +#ifdef FAST_TIMER_TEST + printk("do test\n"); + fast_timer_test(); +#endif + } +} diff --git a/arch/cris/arch-v32/kernel/head.S b/arch/cris/arch-v32/kernel/head.S new file mode 100644 index 000000000000..3cfe57dc391d --- /dev/null +++ b/arch/cris/arch-v32/kernel/head.S @@ -0,0 +1,448 @@ +/* + * CRISv32 kernel startup code. + * + * Copyright (C) 2003, Axis Communications AB + */ + +#include <linux/config.h> + +#define ASSEMBLER_MACROS_ONLY + +/* + * The macros found in mmu_defs_asm.h uses the ## concatenation operator, so + * -traditional must not be used when assembling this file. + */ +#include <asm/arch/hwregs/reg_rdwr.h> +#include <asm/arch/hwregs/asm/mmu_defs_asm.h> +#include <asm/arch/hwregs/asm/reg_map_asm.h> +#include <asm/arch/hwregs/asm/config_defs_asm.h> +#include <asm/arch/hwregs/asm/bif_core_defs_asm.h> + +#define CRAMFS_MAGIC 0x28cd3d45 +#define RAM_INIT_MAGIC 0x56902387 +#define COMMAND_LINE_MAGIC 0x87109563 + + ;; NOTE: R8 and R9 carry information from the decompressor (if the + ;; kernel was compressed). They must not be used in the code below + ;; until they are read! + + ;; Exported symbols. + .global etrax_irv + .global romfs_start + .global romfs_length + .global romfs_in_flash + .global swapper_pg_dir + .global crisv32_nand_boot + .global crisv32_nand_cramfs_offset + + ;; Dummy section to make it bootable with current VCS simulator +#ifdef CONFIG_ETRAXFS_SIM + .section ".boot", "ax" + ba tstart + nop +#endif + + .text +tstart: + ;; This is the entry point of the kernel. The CPU is currently in + ;; supervisor mode. + ;; + ;; 0x00000000 if flash. + ;; 0x40004000 if DRAM. + ;; + di + + ;; Start clocks for used blocks. + move.d REG_ADDR(config, regi_config, rw_clk_ctrl), $r1 + move.d [$r1], $r0 + or.d REG_STATE(config, rw_clk_ctrl, cpu, yes) | \ + REG_STATE(config, rw_clk_ctrl, bif, yes) | \ + REG_STATE(config, rw_clk_ctrl, fix_io, yes), $r0 + move.d $r0, [$r1] + + ;; Set up waitstates etc + move.d REG_ADDR(bif_core, regi_bif_core, rw_grp1_cfg), $r0 + move.d CONFIG_ETRAX_MEM_GRP1_CONFIG, $r1 + move.d $r1, [$r0] + move.d REG_ADDR(bif_core, regi_bif_core, rw_grp2_cfg), $r0 + move.d CONFIG_ETRAX_MEM_GRP2_CONFIG, $r1 + move.d $r1, [$r0] + move.d REG_ADDR(bif_core, regi_bif_core, rw_grp3_cfg), $r0 + move.d CONFIG_ETRAX_MEM_GRP3_CONFIG, $r1 + move.d $r1, [$r0] + move.d REG_ADDR(bif_core, regi_bif_core, rw_grp4_cfg), $r0 + move.d CONFIG_ETRAX_MEM_GRP4_CONFIG, $r1 + move.d $r1, [$r0] + +#ifdef CONFIG_ETRAXFS_SIM + ;; Set up minimal flash waitstates + move.d 0, $r10 + move.d REG_ADDR(bif_core, regi_bif_core, rw_grp1_cfg), $r11 + move.d $r10, [$r11] +#endif + + ;; Setup and enable the MMU. Use same configuration for both the data + ;; and the instruction MMU. + ;; + ;; Note; 3 cycles is needed for a bank-select to take effect. Further; + ;; bank 1 is the instruction MMU, bank 2 is the data MMU. +#ifndef CONFIG_ETRAXFS_SIM + move.d REG_FIELD(mmu, rw_mm_kbase_hi, base_e, 8) \ + | REG_FIELD(mmu, rw_mm_kbase_hi, base_c, 4) \ + | REG_FIELD(mmu, rw_mm_kbase_hi, base_b, 0xb), $r0 +#else + ;; Map the virtual DRAM to the RW eprom area at address 0. + ;; Also map 0xa for the hook calls, + move.d REG_FIELD(mmu, rw_mm_kbase_hi, base_e, 8) \ + | REG_FIELD(mmu, rw_mm_kbase_hi, base_c, 0) \ + | REG_FIELD(mmu, rw_mm_kbase_hi, base_b, 0xb) \ + | REG_FIELD(mmu, rw_mm_kbase_hi, base_a, 0xa), $r0 +#endif + + ;; Temporary map of 0x40 -> 0x40 and 0x00 -> 0x00. + move.d REG_FIELD(mmu, rw_mm_kbase_lo, base_4, 4) \ + | REG_FIELD(mmu, rw_mm_kbase_lo, base_0, 0), $r1 + + ;; Enable certain page protections and setup linear mapping + ;; for f,e,c,b,4,0. +#ifndef CONFIG_ETRAXFS_SIM + move.d REG_STATE(mmu, rw_mm_cfg, we, on) \ + | REG_STATE(mmu, rw_mm_cfg, acc, on) \ + | REG_STATE(mmu, rw_mm_cfg, ex, on) \ + | REG_STATE(mmu, rw_mm_cfg, inv, on) \ + | REG_STATE(mmu, rw_mm_cfg, seg_f, linear) \ + | REG_STATE(mmu, rw_mm_cfg, seg_e, linear) \ + | REG_STATE(mmu, rw_mm_cfg, seg_d, page) \ + | REG_STATE(mmu, rw_mm_cfg, seg_c, linear) \ + | REG_STATE(mmu, rw_mm_cfg, seg_b, linear) \ + | REG_STATE(mmu, rw_mm_cfg, seg_a, page) \ + | REG_STATE(mmu, rw_mm_cfg, seg_9, page) \ + | REG_STATE(mmu, rw_mm_cfg, seg_8, page) \ + | REG_STATE(mmu, rw_mm_cfg, seg_7, page) \ + | REG_STATE(mmu, rw_mm_cfg, seg_6, page) \ + | REG_STATE(mmu, rw_mm_cfg, seg_5, page) \ + | REG_STATE(mmu, rw_mm_cfg, seg_4, linear) \ + | REG_STATE(mmu, rw_mm_cfg, seg_3, page) \ + | REG_STATE(mmu, rw_mm_cfg, seg_2, page) \ + | REG_STATE(mmu, rw_mm_cfg, seg_1, page) \ + | REG_STATE(mmu, rw_mm_cfg, seg_0, linear), $r2 +#else + move.d REG_STATE(mmu, rw_mm_cfg, we, on) \ + | REG_STATE(mmu, rw_mm_cfg, acc, on) \ + | REG_STATE(mmu, rw_mm_cfg, ex, on) \ + | REG_STATE(mmu, rw_mm_cfg, inv, on) \ + | REG_STATE(mmu, rw_mm_cfg, seg_f, linear) \ + | REG_STATE(mmu, rw_mm_cfg, seg_e, linear) \ + | REG_STATE(mmu, rw_mm_cfg, seg_d, page) \ + | REG_STATE(mmu, rw_mm_cfg, seg_c, linear) \ + | REG_STATE(mmu, rw_mm_cfg, seg_b, linear) \ + | REG_STATE(mmu, rw_mm_cfg, seg_a, linear) \ + | REG_STATE(mmu, rw_mm_cfg, seg_9, page) \ + | REG_STATE(mmu, rw_mm_cfg, seg_8, page) \ + | REG_STATE(mmu, rw_mm_cfg, seg_7, page) \ + | REG_STATE(mmu, rw_mm_cfg, seg_6, page) \ + | REG_STATE(mmu, rw_mm_cfg, seg_5, page) \ + | REG_STATE(mmu, rw_mm_cfg, seg_4, linear) \ + | REG_STATE(mmu, rw_mm_cfg, seg_3, page) \ + | REG_STATE(mmu, rw_mm_cfg, seg_2, page) \ + | REG_STATE(mmu, rw_mm_cfg, seg_1, page) \ + | REG_STATE(mmu, rw_mm_cfg, seg_0, linear), $r2 +#endif + + ;; Update instruction MMU. + move 1, $srs + nop + nop + nop + move $r0, $s2 ; kbase_hi. + move $r1, $s1 ; kbase_lo. + move $r2, $s0 ; mm_cfg, virtual memory configuration. + + ;; Update data MMU. + move 2, $srs + nop + nop + nop + move $r0, $s2 ; kbase_hi. + move $r1, $s1 ; kbase_lo + move $r2, $s0 ; mm_cfg, virtual memory configuration. + + ;; Enable data and instruction MMU. + move 0, $srs + moveq 0xf, $r0 ; IMMU, DMMU, DCache, Icache on + nop + nop + nop + move $r0, $s0 + nop + nop + nop + +#ifdef CONFIG_SMP + ;; Read CPU ID + move 0, $srs + nop + nop + nop + move $s10, $r0 + cmpq 0, $r0 + beq master_cpu + nop +slave_cpu: + ; A slave waits for cpu_now_booting to be equal to CPU ID. + move.d cpu_now_booting, $r1 +slave_wait: + cmp.d [$r1], $r0 + bne slave_wait + nop + ; Time to boot-up. Get stack location provided by master CPU. + move.d smp_init_current_idle_thread, $r1 + move.d [$r1], $sp + add.d 8192, $sp + move.d ebp_start, $r0 ; Defined in linker-script. + move $r0, $ebp + jsr smp_callin + nop +master_cpu: +#endif +#ifndef CONFIG_ETRAXFS_SIM + ;; Check if starting from DRAM or flash. + lapcq ., $r0 + and.d 0x7fffffff, $r0 ; Mask off the non-cache bit. + cmp.d 0x10000, $r0 ; Arbitrary, something above this code. + blo _inflash0 + nop +#endif + + jump _inram ; Jump to cached RAM. + nop + + ;; Jumpgate. +_inflash0: + jump _inflash + nop + + ;; Put the following in a section so that storage for it can be + ;; reclaimed after init is finished. + .section ".init.text", "ax" + +_inflash: + + ;; Initialize DRAM. + cmp.d RAM_INIT_MAGIC, $r8 ; Already initialized? + beq _dram_initialized + nop + +#include "../lib/dram_init.S" + +_dram_initialized: + ;; Copy the text and data section to DRAM. This depends on that the + ;; variables used below are correctly set up by the linker script. + ;; The calculated value stored in R4 is used below. + moveq 0, $r0 ; Source. + move.d text_start, $r1 ; Destination. + move.d __vmlinux_end, $r2 + move.d $r2, $r4 + sub.d $r1, $r4 +1: move.w [$r0+], $r3 + move.w $r3, [$r1+] + cmp.d $r2, $r1 + blo 1b + nop + + ;; Keep CRAMFS in flash. + moveq 0, $r0 + move.d romfs_length, $r1 + move.d $r0, [$r1] + move.d [$r4], $r0 ; cramfs_super.magic + cmp.d CRAMFS_MAGIC, $r0 + bne 1f + nop + + addoq +4, $r4, $acr + move.d [$acr], $r0 + move.d romfs_length, $r1 + move.d $r0, [$r1] + add.d 0xf0000000, $r4 ; Add cached flash start in virtual memory. + move.d romfs_start, $r1 + move.d $r4, [$r1] +1: moveq 1, $r0 + move.d romfs_in_flash, $r1 + move.d $r0, [$r1] + + jump _start_it ; Jump to cached code. + nop + +_inram: + ;; Check if booting from NAND flash (in that case we just remember the offset + ;; into the flash where cramfs should be). + move.d REG_ADDR(config, regi_config, r_bootsel), $r0 + move.d [$r0], $r0 + and.d REG_MASK(config, r_bootsel, boot_mode), $r0 + cmp.d REG_STATE(config, r_bootsel, boot_mode, nand), $r0 + bne move_cramfs + moveq 1,$r0 + move.d crisv32_nand_boot, $r1 + move.d $r0, [$r1] + move.d crisv32_nand_cramfs_offset, $r1 + move.d $r9, [$r1] + moveq 1, $r0 + move.d romfs_in_flash, $r1 + move.d $r0, [$r1] + jump _start_it + nop + +move_cramfs: + ;; Move the cramfs after BSS. + moveq 0, $r0 + move.d romfs_length, $r1 + move.d $r0, [$r1] + +#ifndef CONFIG_ETRAXFS_SIM + ;; The kernel could have been unpacked to DRAM by the loader, but + ;; the cramfs image could still be inte the flash immediately + ;; following the compressed kernel image. The loaded passes the address + ;; of the bute succeeding the last compressed byte in the flash in + ;; register R9 when starting the kernel. + cmp.d 0x0ffffff8, $r9 + bhs _no_romfs_in_flash ; R9 points outside the flash area. + nop +#else + ba _no_romfs_in_flash + nop +#endif + move.d [$r9], $r0 ; cramfs_super.magic + cmp.d CRAMFS_MAGIC, $r0 + bne _no_romfs_in_flash + nop + + addoq +4, $r9, $acr + move.d [$acr], $r0 + move.d romfs_length, $r1 + move.d $r0, [$r1] + add.d 0xf0000000, $r9 ; Add cached flash start in virtual memory. + move.d romfs_start, $r1 + move.d $r9, [$r1] + moveq 1, $r0 + move.d romfs_in_flash, $r1 + move.d $r0, [$r1] + + jump _start_it ; Jump to cached code. + nop + +_no_romfs_in_flash: + ;; Look for cramfs. +#ifndef CONFIG_ETRAXFS_SIM + move.d __vmlinux_end, $r0 +#else + move.d __end, $r0 +#endif + move.d [$r0], $r1 + cmp.d CRAMFS_MAGIC, $r1 + bne 2f + nop + + addoq +4, $r0, $acr + move.d [$acr], $r2 + move.d _end, $r1 + move.d romfs_start, $r3 + move.d $r1, [$r3] + move.d romfs_length, $r3 + move.d $r2, [$r3] + +#ifndef CONFIG_ETRAXFS_SIM + add.d $r2, $r0 + add.d $r2, $r1 + + lsrq 1, $r2 ; Size is in bytes, we copy words. + addq 1, $r2 +1: + move.w [$r0], $r3 + move.w $r3, [$r1] + subq 2, $r0 + subq 2, $r1 + subq 1, $r2 + bne 1b + nop +#endif + +2: + moveq 0, $r0 + move.d romfs_in_flash, $r1 + move.d $r0, [$r1] + + jump _start_it ; Jump to cached code. + nop + +_start_it: + + ;; Check if kernel command line is supplied + cmp.d COMMAND_LINE_MAGIC, $r10 + bne no_command_line + nop + + move.d 256, $r13 + move.d cris_command_line, $r10 + or.d 0x80000000, $r11 ; Make it virtual +1: + move.b [$r11+], $r12 + move.b $r12, [$r10+] + subq 1, $r13 + bne 1b + nop + +no_command_line: + + ;; The kernel stack contains a task structure for each task. This + ;; the initial kernel stack is in the same page as the init_task, + ;; but starts at the top of the page, i.e. + 8192 bytes. + move.d init_thread_union + 8192, $sp + move.d ebp_start, $r0 ; Defined in linker-script. + move $r0, $ebp + move.d etrax_irv, $r1 ; Set the exception base register and pointer. + move.d $r0, [$r1] + +#ifndef CONFIG_ETRAXFS_SIM + ;; Clear the BSS region from _bss_start to _end. + move.d __bss_start, $r0 + move.d _end, $r1 +1: clear.d [$r0+] + cmp.d $r1, $r0 + blo 1b + nop +#endif + +#ifdef CONFIG_ETRAXFS_SIM + /* Set the watchdog timeout to something big. Will be removed when */ + /* watchdog can be disabled with command line option */ + move.d 0x7fffffff, $r10 + jsr CPU_WATCHDOG_TIMEOUT + nop +#endif + + ; Initialize registers to increase determinism + move.d __bss_start, $r0 + movem [$r0], $r13 + + jump start_kernel ; Jump to start_kernel() in init/main.c. + nop + + .data +etrax_irv: + .dword 0 +romfs_start: + .dword 0 +romfs_length: + .dword 0 +romfs_in_flash: + .dword 0 +crisv32_nand_boot: + .dword 0 +crisv32_nand_cramfs_offset: + .dword 0 + +swapper_pg_dir = 0xc0002000 + + .section ".init.data", "aw" + +#include "../lib/hw_settings.S" diff --git a/arch/cris/arch-v32/kernel/io.c b/arch/cris/arch-v32/kernel/io.c new file mode 100644 index 000000000000..6bc9f263c3d6 --- /dev/null +++ b/arch/cris/arch-v32/kernel/io.c @@ -0,0 +1,154 @@ +/* + * Helper functions for I/O pins. + * + * Copyright (c) 2004 Axis Communications AB. + */ + +#include <linux/config.h> +#include <linux/types.h> +#include <linux/errno.h> +#include <linux/init.h> +#include <linux/string.h> +#include <linux/ctype.h> +#include <linux/kernel.h> +#include <linux/module.h> +#include <asm/io.h> +#include <asm/arch/pinmux.h> +#include <asm/arch/hwregs/gio_defs.h> + +struct crisv32_ioport crisv32_ioports[] = +{ + { + (unsigned long*)REG_ADDR(gio, regi_gio, rw_pa_oe), + (unsigned long*)REG_ADDR(gio, regi_gio, rw_pa_dout), + (unsigned long*)REG_ADDR(gio, regi_gio, r_pa_din), + 8 + }, + { + (unsigned long*)REG_ADDR(gio, regi_gio, rw_pb_oe), + (unsigned long*)REG_ADDR(gio, regi_gio, rw_pb_dout), + (unsigned long*)REG_ADDR(gio, regi_gio, r_pb_din), + 18 + }, + { + (unsigned long*)REG_ADDR(gio, regi_gio, rw_pc_oe), + (unsigned long*)REG_ADDR(gio, regi_gio, rw_pc_dout), + (unsigned long*)REG_ADDR(gio, regi_gio, r_pc_din), + 18 + }, + { + (unsigned long*)REG_ADDR(gio, regi_gio, rw_pd_oe), + (unsigned long*)REG_ADDR(gio, regi_gio, rw_pd_dout), + (unsigned long*)REG_ADDR(gio, regi_gio, r_pd_din), + 18 + }, + { + (unsigned long*)REG_ADDR(gio, regi_gio, rw_pe_oe), + (unsigned long*)REG_ADDR(gio, regi_gio, rw_pe_dout), + (unsigned long*)REG_ADDR(gio, regi_gio, r_pe_din), + 18 + } +}; + +#define NBR_OF_PORTS sizeof(crisv32_ioports)/sizeof(struct crisv32_ioport) + +struct crisv32_iopin crisv32_led1_green; +struct crisv32_iopin crisv32_led1_red; +struct crisv32_iopin crisv32_led2_green; +struct crisv32_iopin crisv32_led2_red; +struct crisv32_iopin crisv32_led3_green; +struct crisv32_iopin crisv32_led3_red; + +/* Dummy port used when green LED and red LED is on the same bit */ +static unsigned long io_dummy; +static struct crisv32_ioport dummy_port = +{ + &io_dummy, + &io_dummy, + &io_dummy, + 18 +}; +static struct crisv32_iopin dummy_led = +{ + &dummy_port, + 0 +}; + +static int __init crisv32_io_init(void) +{ + int ret = 0; + /* Initialize LEDs */ + ret += crisv32_io_get_name(&crisv32_led1_green, CONFIG_ETRAX_LED1G); + ret += crisv32_io_get_name(&crisv32_led1_red, CONFIG_ETRAX_LED1R); + ret += crisv32_io_get_name(&crisv32_led2_green, CONFIG_ETRAX_LED2G); + ret += crisv32_io_get_name(&crisv32_led2_red, CONFIG_ETRAX_LED2R); + ret += crisv32_io_get_name(&crisv32_led3_green, CONFIG_ETRAX_LED3G); + ret += crisv32_io_get_name(&crisv32_led3_red, CONFIG_ETRAX_LED3R); + crisv32_io_set_dir(&crisv32_led1_green, crisv32_io_dir_out); + crisv32_io_set_dir(&crisv32_led1_red, crisv32_io_dir_out); + crisv32_io_set_dir(&crisv32_led2_green, crisv32_io_dir_out); + crisv32_io_set_dir(&crisv32_led2_red, crisv32_io_dir_out); + crisv32_io_set_dir(&crisv32_led3_green, crisv32_io_dir_out); + crisv32_io_set_dir(&crisv32_led3_red, crisv32_io_dir_out); + + if (!strcmp(CONFIG_ETRAX_LED1G, CONFIG_ETRAX_LED1R)) + crisv32_led1_red = dummy_led; + if (!strcmp(CONFIG_ETRAX_LED2G, CONFIG_ETRAX_LED2R)) + crisv32_led2_red = dummy_led; + + return ret; +} + +__initcall(crisv32_io_init); + +int crisv32_io_get(struct crisv32_iopin* iopin, + unsigned int port, unsigned int pin) +{ + if (port > NBR_OF_PORTS) + return -EINVAL; + if (port > crisv32_ioports[port].pin_count) + return -EINVAL; + + iopin->bit = 1 << pin; + iopin->port = &crisv32_ioports[port]; + + if (crisv32_pinmux_alloc(port, pin, pin, pinmux_gpio)) + return -EIO; + + return 0; +} + +int crisv32_io_get_name(struct crisv32_iopin* iopin, + char* name) +{ + int port; + int pin; + + if (toupper(*name) == 'P') + name++; + + if (toupper(*name) < 'A' || toupper(*name) > 'E') + return -EINVAL; + + port = toupper(*name) - 'A'; + name++; + pin = simple_strtoul(name, NULL, 10); + + if (pin < 0 || pin > crisv32_ioports[port].pin_count) + return -EINVAL; + + iopin->bit = 1 << pin; + iopin->port = &crisv32_ioports[port]; + + if (crisv32_pinmux_alloc(port, pin, pin, pinmux_gpio)) + return -EIO; + + return 0; +} + +#ifdef CONFIG_PCI +/* PCI I/O access stuff */ +struct cris_io_operations* cris_iops = NULL; +EXPORT_SYMBOL(cris_iops); +#endif + diff --git a/arch/cris/arch-v32/kernel/irq.c b/arch/cris/arch-v32/kernel/irq.c new file mode 100644 index 000000000000..c78cc2685133 --- /dev/null +++ b/arch/cris/arch-v32/kernel/irq.c @@ -0,0 +1,413 @@ +/* + * Copyright (C) 2003, Axis Communications AB. + */ + +#include <asm/irq.h> +#include <linux/irq.h> +#include <linux/interrupt.h> +#include <linux/smp.h> +#include <linux/config.h> +#include <linux/kernel.h> +#include <linux/errno.h> +#include <linux/init.h> +#include <linux/profile.h> +#include <linux/proc_fs.h> +#include <linux/seq_file.h> +#include <linux/threads.h> +#include <linux/spinlock.h> +#include <linux/kernel_stat.h> +#include <asm/arch/hwregs/reg_map.h> +#include <asm/arch/hwregs/reg_rdwr.h> +#include <asm/arch/hwregs/intr_vect.h> +#include <asm/arch/hwregs/intr_vect_defs.h> + +#define CPU_FIXED -1 + +/* IRQ masks (refer to comment for crisv32_do_multiple) */ +#define TIMER_MASK (1 << (TIMER_INTR_VECT - FIRST_IRQ)) +#ifdef CONFIG_ETRAX_KGDB +#if defined(CONFIG_ETRAX_KGDB_PORT0) +#define IGNOREMASK (1 << (SER0_INTR_VECT - FIRST_IRQ)) +#elif defined(CONFIG_ETRAX_KGDB_PORT1) +#define IGNOREMASK (1 << (SER1_INTR_VECT - FIRST_IRQ)) +#elif defined(CONFIG_ETRAX_KGB_PORT2) +#define IGNOREMASK (1 << (SER2_INTR_VECT - FIRST_IRQ)) +#elif defined(CONFIG_ETRAX_KGDB_PORT3) +#define IGNOREMASK (1 << (SER3_INTR_VECT - FIRST_IRQ)) +#endif +#endif + +DEFINE_SPINLOCK(irq_lock); + +struct cris_irq_allocation +{ + int cpu; /* The CPU to which the IRQ is currently allocated. */ + cpumask_t mask; /* The CPUs to which the IRQ may be allocated. */ +}; + +struct cris_irq_allocation irq_allocations[NR_IRQS] = + {[0 ... NR_IRQS - 1] = {0, CPU_MASK_ALL}}; + +static unsigned long irq_regs[NR_CPUS] = +{ + regi_irq, +#ifdef CONFIG_SMP + regi_irq2, +#endif +}; + +unsigned long cpu_irq_counters[NR_CPUS]; +unsigned long irq_counters[NR_REAL_IRQS]; + +/* From irq.c. */ +extern void weird_irq(void); + +/* From entry.S. */ +extern void system_call(void); +extern void nmi_interrupt(void); +extern void multiple_interrupt(void); +extern void gdb_handle_exception(void); +extern void i_mmu_refill(void); +extern void i_mmu_invalid(void); +extern void i_mmu_access(void); +extern void i_mmu_execute(void); +extern void d_mmu_refill(void); +extern void d_mmu_invalid(void); +extern void d_mmu_access(void); +extern void d_mmu_write(void); + +/* From kgdb.c. */ +extern void kgdb_init(void); +extern void breakpoint(void); + +/* + * Build the IRQ handler stubs using macros from irq.h. First argument is the + * IRQ number, the second argument is the corresponding bit in + * intr_rw_vect_mask found in asm/arch/hwregs/intr_vect_defs.h. + */ +BUILD_IRQ(0x31, (1 << 0)) /* memarb */ +BUILD_IRQ(0x32, (1 << 1)) /* gen_io */ +BUILD_IRQ(0x33, (1 << 2)) /* iop0 */ +BUILD_IRQ(0x34, (1 << 3)) /* iop1 */ +BUILD_IRQ(0x35, (1 << 4)) /* iop2 */ +BUILD_IRQ(0x36, (1 << 5)) /* iop3 */ +BUILD_IRQ(0x37, (1 << 6)) /* dma0 */ +BUILD_IRQ(0x38, (1 << 7)) /* dma1 */ +BUILD_IRQ(0x39, (1 << 8)) /* dma2 */ +BUILD_IRQ(0x3a, (1 << 9)) /* dma3 */ +BUILD_IRQ(0x3b, (1 << 10)) /* dma4 */ +BUILD_IRQ(0x3c, (1 << 11)) /* dma5 */ +BUILD_IRQ(0x3d, (1 << 12)) /* dma6 */ +BUILD_IRQ(0x3e, (1 << 13)) /* dma7 */ +BUILD_IRQ(0x3f, (1 << 14)) /* dma8 */ +BUILD_IRQ(0x40, (1 << 15)) /* dma9 */ +BUILD_IRQ(0x41, (1 << 16)) /* ata */ +BUILD_IRQ(0x42, (1 << 17)) /* sser0 */ +BUILD_IRQ(0x43, (1 << 18)) /* sser1 */ +BUILD_IRQ(0x44, (1 << 19)) /* ser0 */ +BUILD_IRQ(0x45, (1 << 20)) /* ser1 */ +BUILD_IRQ(0x46, (1 << 21)) /* ser2 */ +BUILD_IRQ(0x47, (1 << 22)) /* ser3 */ +BUILD_IRQ(0x48, (1 << 23)) +BUILD_IRQ(0x49, (1 << 24)) /* eth0 */ +BUILD_IRQ(0x4a, (1 << 25)) /* eth1 */ +BUILD_TIMER_IRQ(0x4b, (1 << 26))/* timer */ +BUILD_IRQ(0x4c, (1 << 27)) /* bif_arb */ +BUILD_IRQ(0x4d, (1 << 28)) /* bif_dma */ +BUILD_IRQ(0x4e, (1 << 29)) /* ext */ +BUILD_IRQ(0x4f, (1 << 29)) /* ipi */ + +/* Pointers to the low-level handlers. */ +static void (*interrupt[NR_IRQS])(void) = { + IRQ0x31_interrupt, IRQ0x32_interrupt, IRQ0x33_interrupt, + IRQ0x34_interrupt, IRQ0x35_interrupt, IRQ0x36_interrupt, + IRQ0x37_interrupt, IRQ0x38_interrupt, IRQ0x39_interrupt, + IRQ0x3a_interrupt, IRQ0x3b_interrupt, IRQ0x3c_interrupt, + IRQ0x3d_interrupt, IRQ0x3e_interrupt, IRQ0x3f_interrupt, + IRQ0x40_interrupt, IRQ0x41_interrupt, IRQ0x42_interrupt, + IRQ0x43_interrupt, IRQ0x44_interrupt, IRQ0x45_interrupt, + IRQ0x46_interrupt, IRQ0x47_interrupt, IRQ0x48_interrupt, + IRQ0x49_interrupt, IRQ0x4a_interrupt, IRQ0x4b_interrupt, + IRQ0x4c_interrupt, IRQ0x4d_interrupt, IRQ0x4e_interrupt, + IRQ0x4f_interrupt +}; + +void +block_irq(int irq, int cpu) +{ + int intr_mask; + unsigned long flags; + + spin_lock_irqsave(&irq_lock, flags); + intr_mask = REG_RD_INT(intr_vect, irq_regs[cpu], rw_mask); + + /* Remember; 1 let thru, 0 block. */ + intr_mask &= ~(1 << (irq - FIRST_IRQ)); + + REG_WR_INT(intr_vect, irq_regs[cpu], rw_mask, intr_mask); + spin_unlock_irqrestore(&irq_lock, flags); +} + +void +unblock_irq(int irq, int cpu) +{ + int intr_mask; + unsigned long flags; + + spin_lock_irqsave(&irq_lock, flags); + intr_mask = REG_RD_INT(intr_vect, irq_regs[cpu], rw_mask); + + /* Remember; 1 let thru, 0 block. */ + intr_mask |= (1 << (irq - FIRST_IRQ)); + + REG_WR_INT(intr_vect, irq_regs[cpu], rw_mask, intr_mask); + spin_unlock_irqrestore(&irq_lock, flags); +} + +/* Find out which CPU the irq should be allocated to. */ +static int irq_cpu(int irq) +{ + int cpu; + unsigned long flags; + + spin_lock_irqsave(&irq_lock, flags); + cpu = irq_allocations[irq - FIRST_IRQ].cpu; + + /* Fixed interrupts stay on the local CPU. */ + if (cpu == CPU_FIXED) + { + spin_unlock_irqrestore(&irq_lock, flags); + return smp_processor_id(); + } + + + /* Let the interrupt stay if possible */ + if (cpu_isset(cpu, irq_allocations[irq - FIRST_IRQ].mask)) + goto out; + + /* IRQ must be moved to another CPU. */ + cpu = first_cpu(irq_allocations[irq - FIRST_IRQ].mask); + irq_allocations[irq - FIRST_IRQ].cpu = cpu; +out: + spin_unlock_irqrestore(&irq_lock, flags); + return cpu; +} + +void +mask_irq(int irq) +{ + int cpu; + + for (cpu = 0; cpu < NR_CPUS; cpu++) + block_irq(irq, cpu); +} + +void +unmask_irq(int irq) +{ + unblock_irq(irq, irq_cpu(irq)); +} + + +static unsigned int startup_crisv32_irq(unsigned int irq) +{ + unmask_irq(irq); + return 0; +} + +static void shutdown_crisv32_irq(unsigned int irq) +{ + mask_irq(irq); +} + +static void enable_crisv32_irq(unsigned int irq) +{ + unmask_irq(irq); +} + +static void disable_crisv32_irq(unsigned int irq) +{ + mask_irq(irq); +} + +static void ack_crisv32_irq(unsigned int irq) +{ +} + +static void end_crisv32_irq(unsigned int irq) +{ +} + +void set_affinity_crisv32_irq(unsigned int irq, cpumask_t dest) +{ + unsigned long flags; + spin_lock_irqsave(&irq_lock, flags); + irq_allocations[irq - FIRST_IRQ].mask = dest; + spin_unlock_irqrestore(&irq_lock, flags); +} + +static struct hw_interrupt_type crisv32_irq_type = { + .typename = "CRISv32", + .startup = startup_crisv32_irq, + .shutdown = shutdown_crisv32_irq, + .enable = enable_crisv32_irq, + .disable = disable_crisv32_irq, + .ack = ack_crisv32_irq, + .end = end_crisv32_irq, + .set_affinity = set_affinity_crisv32_irq +}; + +void +set_exception_vector(int n, irqvectptr addr) +{ + etrax_irv->v[n] = (irqvectptr) addr; +} + +extern void do_IRQ(int irq, struct pt_regs * regs); + +void +crisv32_do_IRQ(int irq, int block, struct pt_regs* regs) +{ + /* Interrupts that may not be moved to another CPU and + * are SA_INTERRUPT may skip blocking. This is currently + * only valid for the timer IRQ and the IPI and is used + * for the timer interrupt to avoid watchdog starvation. + */ + if (!block) { + do_IRQ(irq, regs); + return; + } + + block_irq(irq, smp_processor_id()); + do_IRQ(irq, regs); + + unblock_irq(irq, irq_cpu(irq)); +} + +/* If multiple interrupts occur simultaneously we get a multiple + * interrupt from the CPU and software has to sort out which + * interrupts that happened. There are two special cases here: + * + * 1. Timer interrupts may never be blocked because of the + * watchdog (refer to comment in include/asr/arch/irq.h) + * 2. GDB serial port IRQs are unhandled here and will be handled + * as a single IRQ when it strikes again because the GDB + * stubb wants to save the registers in its own fashion. + */ +void +crisv32_do_multiple(struct pt_regs* regs) +{ + int cpu; + int mask; + int masked; + int bit; + + cpu = smp_processor_id(); + + /* An extra irq_enter here to prevent softIRQs to run after + * each do_IRQ. This will decrease the interrupt latency. + */ + irq_enter(); + + /* Get which IRQs that happend. */ + masked = REG_RD_INT(intr_vect, irq_regs[cpu], r_masked_vect); + + /* Calculate new IRQ mask with these IRQs disabled. */ + mask = REG_RD_INT(intr_vect, irq_regs[cpu], rw_mask); + mask &= ~masked; + + /* Timer IRQ is never masked */ + if (masked & TIMER_MASK) + mask |= TIMER_MASK; + + /* Block all the IRQs */ + REG_WR_INT(intr_vect, irq_regs[cpu], rw_mask, mask); + + /* Check for timer IRQ and handle it special. */ + if (masked & TIMER_MASK) { + masked &= ~TIMER_MASK; + do_IRQ(TIMER_INTR_VECT, regs); + } + +#ifdef IGNORE_MASK + /* Remove IRQs that can't be handled as multiple. */ + masked &= ~IGNORE_MASK; +#endif + + /* Handle the rest of the IRQs. */ + for (bit = 0; bit < 32; bit++) + { + if (masked & (1 << bit)) + do_IRQ(bit + FIRST_IRQ, regs); + } + + /* Unblock all the IRQs. */ + mask = REG_RD_INT(intr_vect, irq_regs[cpu], rw_mask); + mask |= masked; + REG_WR_INT(intr_vect, irq_regs[cpu], rw_mask, mask); + + /* This irq_exit() will trigger the soft IRQs. */ + irq_exit(); +} + +/* + * This is called by start_kernel. It fixes the IRQ masks and setup the + * interrupt vector table to point to bad_interrupt pointers. + */ +void __init +init_IRQ(void) +{ + int i; + int j; + reg_intr_vect_rw_mask vect_mask = {0}; + + /* Clear all interrupts masks. */ + REG_WR(intr_vect, regi_irq, rw_mask, vect_mask); + + for (i = 0; i < 256; i++) + etrax_irv->v[i] = weird_irq; + + /* Point all IRQ's to bad handlers. */ + for (i = FIRST_IRQ, j = 0; j < NR_IRQS; i++, j++) { + irq_desc[j].handler = &crisv32_irq_type; + set_exception_vector(i, interrupt[j]); + } + + /* Mark Timer and IPI IRQs as CPU local */ + irq_allocations[TIMER_INTR_VECT - FIRST_IRQ].cpu = CPU_FIXED; + irq_desc[TIMER_INTR_VECT].status |= IRQ_PER_CPU; + irq_allocations[IPI_INTR_VECT - FIRST_IRQ].cpu = CPU_FIXED; + irq_desc[IPI_INTR_VECT].status |= IRQ_PER_CPU; + + set_exception_vector(0x00, nmi_interrupt); + set_exception_vector(0x30, multiple_interrupt); + + /* Set up handler for various MMU bus faults. */ + set_exception_vector(0x04, i_mmu_refill); + set_exception_vector(0x05, i_mmu_invalid); + set_exception_vector(0x06, i_mmu_access); + set_exception_vector(0x07, i_mmu_execute); + set_exception_vector(0x08, d_mmu_refill); + set_exception_vector(0x09, d_mmu_invalid); + set_exception_vector(0x0a, d_mmu_access); + set_exception_vector(0x0b, d_mmu_write); + + /* The system-call trap is reached by "break 13". */ + set_exception_vector(0x1d, system_call); + + /* Exception handlers for debugging, both user-mode and kernel-mode. */ + + /* Break 8. */ + set_exception_vector(0x18, gdb_handle_exception); + /* Hardware single step. */ + set_exception_vector(0x3, gdb_handle_exception); + /* Hardware breakpoint. */ + set_exception_vector(0xc, gdb_handle_exception); + +#ifdef CONFIG_ETRAX_KGDB + kgdb_init(); + /* Everything is set up; now trap the kernel. */ + breakpoint(); +#endif +} + diff --git a/arch/cris/arch-v32/kernel/kgdb.c b/arch/cris/arch-v32/kernel/kgdb.c new file mode 100644 index 000000000000..480e56348be2 --- /dev/null +++ b/arch/cris/arch-v32/kernel/kgdb.c @@ -0,0 +1,1660 @@ +/* + * arch/cris/arch-v32/kernel/kgdb.c + * + * CRIS v32 version by Orjan Friberg, Axis Communications AB. + * + * S390 version + * Copyright (C) 1999 IBM Deutschland Entwicklung GmbH, IBM Corporation + * Author(s): Denis Joseph Barrow (djbarrow@de.ibm.com,barrow_dj@yahoo.com), + * + * Originally written by Glenn Engel, Lake Stevens Instrument Division + * + * Contributed by HP Systems + * + * Modified for SPARC by Stu Grossman, Cygnus Support. + * + * Modified for Linux/MIPS (and MIPS in general) by Andreas Busse + * Send complaints, suggestions etc. to <andy@waldorf-gmbh.de> + * + * Copyright (C) 1995 Andreas Busse + */ + +/* FIXME: Check the documentation. */ + +/* + * kgdb usage notes: + * ----------------- + * + * If you select CONFIG_ETRAX_KGDB in the configuration, the kernel will be + * built with different gcc flags: "-g" is added to get debug infos, and + * "-fomit-frame-pointer" is omitted to make debugging easier. Since the + * resulting kernel will be quite big (approx. > 7 MB), it will be stripped + * before compresion. Such a kernel will behave just as usually, except if + * given a "debug=<device>" command line option. (Only serial devices are + * allowed for <device>, i.e. no printers or the like; possible values are + * machine depedend and are the same as for the usual debug device, the one + * for logging kernel messages.) If that option is given and the device can be + * initialized, the kernel will connect to the remote gdb in trap_init(). The + * serial parameters are fixed to 8N1 and 115200 bps, for easyness of + * implementation. + * + * To start a debugging session, start that gdb with the debugging kernel + * image (the one with the symbols, vmlinux.debug) named on the command line. + * This file will be used by gdb to get symbol and debugging infos about the + * kernel. Next, select remote debug mode by + * target remote <device> + * where <device> is the name of the serial device over which the debugged + * machine is connected. Maybe you have to adjust the baud rate by + * set remotebaud <rate> + * or also other parameters with stty: + * shell stty ... </dev/... + * If the kernel to debug has already booted, it waited for gdb and now + * connects, and you'll see a breakpoint being reported. If the kernel isn't + * running yet, start it now. The order of gdb and the kernel doesn't matter. + * Another thing worth knowing about in the getting-started phase is how to + * debug the remote protocol itself. This is activated with + * set remotedebug 1 + * gdb will then print out each packet sent or received. You'll also get some + * messages about the gdb stub on the console of the debugged machine. + * + * If all that works, you can use lots of the usual debugging techniques on + * the kernel, e.g. inspecting and changing variables/memory, setting + * breakpoints, single stepping and so on. It's also possible to interrupt the + * debugged kernel by pressing C-c in gdb. Have fun! :-) + * + * The gdb stub is entered (and thus the remote gdb gets control) in the + * following situations: + * + * - If breakpoint() is called. This is just after kgdb initialization, or if + * a breakpoint() call has been put somewhere into the kernel source. + * (Breakpoints can of course also be set the usual way in gdb.) + * In eLinux, we call breakpoint() in init/main.c after IRQ initialization. + * + * - If there is a kernel exception, i.e. bad_super_trap() or die_if_kernel() + * are entered. All the CPU exceptions are mapped to (more or less..., see + * the hard_trap_info array below) appropriate signal, which are reported + * to gdb. die_if_kernel() is usually called after some kind of access + * error and thus is reported as SIGSEGV. + * + * - When panic() is called. This is reported as SIGABRT. + * + * - If C-c is received over the serial line, which is treated as + * SIGINT. + * + * Of course, all these signals are just faked for gdb, since there is no + * signal concept as such for the kernel. It also isn't possible --obviously-- + * to set signal handlers from inside gdb, or restart the kernel with a + * signal. + * + * Current limitations: + * + * - While the kernel is stopped, interrupts are disabled for safety reasons + * (i.e., variables not changing magically or the like). But this also + * means that the clock isn't running anymore, and that interrupts from the + * hardware may get lost/not be served in time. This can cause some device + * errors... + * + * - When single-stepping, only one instruction of the current thread is + * executed, but interrupts are allowed for that time and will be serviced + * if pending. Be prepared for that. + * + * - All debugging happens in kernel virtual address space. There's no way to + * access physical memory not mapped in kernel space, or to access user + * space. A way to work around this is using get_user_long & Co. in gdb + * expressions, but only for the current process. + * + * - Interrupting the kernel only works if interrupts are currently allowed, + * and the interrupt of the serial line isn't blocked by some other means + * (IPL too high, disabled, ...) + * + * - The gdb stub is currently not reentrant, i.e. errors that happen therein + * (e.g. accessing invalid memory) may not be caught correctly. This could + * be removed in future by introducing a stack of struct registers. + * + */ + +/* + * To enable debugger support, two things need to happen. One, a + * call to kgdb_init() is necessary in order to allow any breakpoints + * or error conditions to be properly intercepted and reported to gdb. + * Two, a breakpoint needs to be generated to begin communication. This + * is most easily accomplished by a call to breakpoint(). + * + * The following gdb commands are supported: + * + * command function Return value + * + * g return the value of the CPU registers hex data or ENN + * G set the value of the CPU registers OK or ENN + * + * mAA..AA,LLLL Read LLLL bytes at address AA..AA hex data or ENN + * MAA..AA,LLLL: Write LLLL bytes at address AA.AA OK or ENN + * + * c Resume at current address SNN ( signal NN) + * cAA..AA Continue at address AA..AA SNN + * + * s Step one instruction SNN + * sAA..AA Step one instruction from AA..AA SNN + * + * k kill + * + * ? What was the last sigval ? SNN (signal NN) + * + * bBB..BB Set baud rate to BB..BB OK or BNN, then sets + * baud rate + * + * All commands and responses are sent with a packet which includes a + * checksum. A packet consists of + * + * $<packet info>#<checksum>. + * + * where + * <packet info> :: <characters representing the command or response> + * <checksum> :: < two hex digits computed as modulo 256 sum of <packetinfo>> + * + * When a packet is received, it is first acknowledged with either '+' or '-'. + * '+' indicates a successful transfer. '-' indicates a failed transfer. + * + * Example: + * + * Host: Reply: + * $m0,10#2a +$00010203040506070809101112131415#42 + * + */ + + +#include <linux/string.h> +#include <linux/signal.h> +#include <linux/kernel.h> +#include <linux/delay.h> +#include <linux/linkage.h> +#include <linux/reboot.h> + +#include <asm/setup.h> +#include <asm/ptrace.h> + +#include <asm/irq.h> +#include <asm/arch/hwregs/reg_map.h> +#include <asm/arch/hwregs/reg_rdwr.h> +#include <asm/arch/hwregs/intr_vect_defs.h> +#include <asm/arch/hwregs/ser_defs.h> + +/* From entry.S. */ +extern void gdb_handle_exception(void); +/* From kgdb_asm.S. */ +extern void kgdb_handle_exception(void); + +static int kgdb_started = 0; + +/********************************* Register image ****************************/ + +typedef +struct register_image +{ + /* Offset */ + unsigned int r0; /* 0x00 */ + unsigned int r1; /* 0x04 */ + unsigned int r2; /* 0x08 */ + unsigned int r3; /* 0x0C */ + unsigned int r4; /* 0x10 */ + unsigned int r5; /* 0x14 */ + unsigned int r6; /* 0x18 */ + unsigned int r7; /* 0x1C */ + unsigned int r8; /* 0x20; Frame pointer (if any) */ + unsigned int r9; /* 0x24 */ + unsigned int r10; /* 0x28 */ + unsigned int r11; /* 0x2C */ + unsigned int r12; /* 0x30 */ + unsigned int r13; /* 0x34 */ + unsigned int sp; /* 0x38; R14, Stack pointer */ + unsigned int acr; /* 0x3C; R15, Address calculation register. */ + + unsigned char bz; /* 0x40; P0, 8-bit zero register */ + unsigned char vr; /* 0x41; P1, Version register (8-bit) */ + unsigned int pid; /* 0x42; P2, Process ID */ + unsigned char srs; /* 0x46; P3, Support register select (8-bit) */ + unsigned short wz; /* 0x47; P4, 16-bit zero register */ + unsigned int exs; /* 0x49; P5, Exception status */ + unsigned int eda; /* 0x4D; P6, Exception data address */ + unsigned int mof; /* 0x51; P7, Multiply overflow register */ + unsigned int dz; /* 0x55; P8, 32-bit zero register */ + unsigned int ebp; /* 0x59; P9, Exception base pointer */ + unsigned int erp; /* 0x5D; P10, Exception return pointer. Contains the PC we are interested in. */ + unsigned int srp; /* 0x61; P11, Subroutine return pointer */ + unsigned int nrp; /* 0x65; P12, NMI return pointer */ + unsigned int ccs; /* 0x69; P13, Condition code stack */ + unsigned int usp; /* 0x6D; P14, User mode stack pointer */ + unsigned int spc; /* 0x71; P15, Single step PC */ + unsigned int pc; /* 0x75; Pseudo register (for the most part set to ERP). */ + +} registers; + +typedef +struct bp_register_image +{ + /* Support register bank 0. */ + unsigned int s0_0; + unsigned int s1_0; + unsigned int s2_0; + unsigned int s3_0; + unsigned int s4_0; + unsigned int s5_0; + unsigned int s6_0; + unsigned int s7_0; + unsigned int s8_0; + unsigned int s9_0; + unsigned int s10_0; + unsigned int s11_0; + unsigned int s12_0; + unsigned int s13_0; + unsigned int s14_0; + unsigned int s15_0; + + /* Support register bank 1. */ + unsigned int s0_1; + unsigned int s1_1; + unsigned int s2_1; + unsigned int s3_1; + unsigned int s4_1; + unsigned int s5_1; + unsigned int s6_1; + unsigned int s7_1; + unsigned int s8_1; + unsigned int s9_1; + unsigned int s10_1; + unsigned int s11_1; + unsigned int s12_1; + unsigned int s13_1; + unsigned int s14_1; + unsigned int s15_1; + + /* Support register bank 2. */ + unsigned int s0_2; + unsigned int s1_2; + unsigned int s2_2; + unsigned int s3_2; + unsigned int s4_2; + unsigned int s5_2; + unsigned int s6_2; + unsigned int s7_2; + unsigned int s8_2; + unsigned int s9_2; + unsigned int s10_2; + unsigned int s11_2; + unsigned int s12_2; + unsigned int s13_2; + unsigned int s14_2; + unsigned int s15_2; + + /* Support register bank 3. */ + unsigned int s0_3; /* BP_CTRL */ + unsigned int s1_3; /* BP_I0_START */ + unsigned int s2_3; /* BP_I0_END */ + unsigned int s3_3; /* BP_D0_START */ + unsigned int s4_3; /* BP_D0_END */ + unsigned int s5_3; /* BP_D1_START */ + unsigned int s6_3; /* BP_D1_END */ + unsigned int s7_3; /* BP_D2_START */ + unsigned int s8_3; /* BP_D2_END */ + unsigned int s9_3; /* BP_D3_START */ + unsigned int s10_3; /* BP_D3_END */ + unsigned int s11_3; /* BP_D4_START */ + unsigned int s12_3; /* BP_D4_END */ + unsigned int s13_3; /* BP_D5_START */ + unsigned int s14_3; /* BP_D5_END */ + unsigned int s15_3; /* BP_RESERVED */ + +} support_registers; + +enum register_name +{ + R0, R1, R2, R3, + R4, R5, R6, R7, + R8, R9, R10, R11, + R12, R13, SP, ACR, + + BZ, VR, PID, SRS, + WZ, EXS, EDA, MOF, + DZ, EBP, ERP, SRP, + NRP, CCS, USP, SPC, + PC, + + S0, S1, S2, S3, + S4, S5, S6, S7, + S8, S9, S10, S11, + S12, S13, S14, S15 + +}; + +/* The register sizes of the registers in register_name. An unimplemented register + is designated by size 0 in this array. */ +static int register_size[] = +{ + 4, 4, 4, 4, + 4, 4, 4, 4, + 4, 4, 4, 4, + 4, 4, 4, 4, + + 1, 1, 4, 1, + 2, 4, 4, 4, + 4, 4, 4, 4, + 4, 4, 4, 4, + + 4, + + 4, 4, 4, 4, + 4, 4, 4, 4, + 4, 4, 4, 4, + 4, 4, 4 + +}; + +/* Contains the register image of the kernel. + (Global so that they can be reached from assembler code.) */ +registers reg; +support_registers sreg; + +/************** Prototypes for local library functions ***********************/ + +/* Copy of strcpy from libc. */ +static char *gdb_cris_strcpy(char *s1, const char *s2); + +/* Copy of strlen from libc. */ +static int gdb_cris_strlen(const char *s); + +/* Copy of memchr from libc. */ +static void *gdb_cris_memchr(const void *s, int c, int n); + +/* Copy of strtol from libc. Does only support base 16. */ +static int gdb_cris_strtol(const char *s, char **endptr, int base); + +/********************** Prototypes for local functions. **********************/ + +/* Write a value to a specified register regno in the register image + of the current thread. */ +static int write_register(int regno, char *val); + +/* Read a value from a specified register in the register image. Returns the + status of the read operation. The register value is returned in valptr. */ +static int read_register(char regno, unsigned int *valptr); + +/* Serial port, reads one character. ETRAX 100 specific. from debugport.c */ +int getDebugChar(void); + +#ifdef CONFIG_ETRAXFS_SIM +int getDebugChar(void) +{ + return socketread(); +} +#endif + +/* Serial port, writes one character. ETRAX 100 specific. from debugport.c */ +void putDebugChar(int val); + +#ifdef CONFIG_ETRAXFS_SIM +void putDebugChar(int val) +{ + socketwrite((char *)&val, 1); +} +#endif + +/* Returns the character equivalent of a nibble, bit 7, 6, 5, and 4 of a byte, + represented by int x. */ +static char highhex(int x); + +/* Returns the character equivalent of a nibble, bit 3, 2, 1, and 0 of a byte, + represented by int x. */ +static char lowhex(int x); + +/* Returns the integer equivalent of a hexadecimal character. */ +static int hex(char ch); + +/* Convert the memory, pointed to by mem into hexadecimal representation. + Put the result in buf, and return a pointer to the last character + in buf (null). */ +static char *mem2hex(char *buf, unsigned char *mem, int count); + +/* Convert the array, in hexadecimal representation, pointed to by buf into + binary representation. Put the result in mem, and return a pointer to + the character after the last byte written. */ +static unsigned char *hex2mem(unsigned char *mem, char *buf, int count); + +/* Put the content of the array, in binary representation, pointed to by buf + into memory pointed to by mem, and return a pointer to + the character after the last byte written. */ +static unsigned char *bin2mem(unsigned char *mem, unsigned char *buf, int count); + +/* Await the sequence $<data>#<checksum> and store <data> in the array buffer + returned. */ +static void getpacket(char *buffer); + +/* Send $<data>#<checksum> from the <data> in the array buffer. */ +static void putpacket(char *buffer); + +/* Build and send a response packet in order to inform the host the + stub is stopped. */ +static void stub_is_stopped(int sigval); + +/* All expected commands are sent from remote.c. Send a response according + to the description in remote.c. Not static since it needs to be reached + from assembler code. */ +void handle_exception(int sigval); + +/* Performs a complete re-start from scratch. ETRAX specific. */ +static void kill_restart(void); + +/******************** Prototypes for global functions. ***********************/ + +/* The string str is prepended with the GDB printout token and sent. */ +void putDebugString(const unsigned char *str, int len); + +/* A static breakpoint to be used at startup. */ +void breakpoint(void); + +/* Avoid warning as the internal_stack is not used in the C-code. */ +#define USEDVAR(name) { if (name) { ; } } +#define USEDFUN(name) { void (*pf)(void) = (void *)name; USEDVAR(pf) } + +/********************************** Packet I/O ******************************/ +/* BUFMAX defines the maximum number of characters in + inbound/outbound buffers */ +/* FIXME: How do we know it's enough? */ +#define BUFMAX 512 + +/* Run-length encoding maximum length. Send 64 at most. */ +#define RUNLENMAX 64 + +/* Definition of all valid hexadecimal characters */ +static const char hexchars[] = "0123456789abcdef"; + +/* The inbound/outbound buffers used in packet I/O */ +static char input_buffer[BUFMAX]; +static char output_buffer[BUFMAX]; + +/* Error and warning messages. */ +enum error_type +{ + SUCCESS, E01, E02, E03, E04, E05, E06, +}; + +static char *error_message[] = +{ + "", + "E01 Set current or general thread - H[c,g] - internal error.", + "E02 Change register content - P - cannot change read-only register.", + "E03 Thread is not alive.", /* T, not used. */ + "E04 The command is not supported - [s,C,S,!,R,d,r] - internal error.", + "E05 Change register content - P - the register is not implemented..", + "E06 Change memory content - M - internal error.", +}; + +/********************************** Breakpoint *******************************/ +/* Use an internal stack in the breakpoint and interrupt response routines. + FIXME: How do we know the size of this stack is enough? + Global so it can be reached from assembler code. */ +#define INTERNAL_STACK_SIZE 1024 +char internal_stack[INTERNAL_STACK_SIZE]; + +/* Due to the breakpoint return pointer, a state variable is needed to keep + track of whether it is a static (compiled) or dynamic (gdb-invoked) + breakpoint to be handled. A static breakpoint uses the content of register + ERP as it is whereas a dynamic breakpoint requires subtraction with 2 + in order to execute the instruction. The first breakpoint is static; all + following are assumed to be dynamic. */ +static int dynamic_bp = 0; + +/********************************* String library ****************************/ +/* Single-step over library functions creates trap loops. */ + +/* Copy char s2[] to s1[]. */ +static char* +gdb_cris_strcpy(char *s1, const char *s2) +{ + char *s = s1; + + for (s = s1; (*s++ = *s2++) != '\0'; ) + ; + return s1; +} + +/* Find length of s[]. */ +static int +gdb_cris_strlen(const char *s) +{ + const char *sc; + + for (sc = s; *sc != '\0'; sc++) + ; + return (sc - s); +} + +/* Find first occurrence of c in s[n]. */ +static void* +gdb_cris_memchr(const void *s, int c, int n) +{ + const unsigned char uc = c; + const unsigned char *su; + + for (su = s; 0 < n; ++su, --n) + if (*su == uc) + return (void *)su; + return NULL; +} +/******************************* Standard library ****************************/ +/* Single-step over library functions creates trap loops. */ +/* Convert string to long. */ +static int +gdb_cris_strtol(const char *s, char **endptr, int base) +{ + char *s1; + char *sd; + int x = 0; + + for (s1 = (char*)s; (sd = gdb_cris_memchr(hexchars, *s1, base)) != NULL; ++s1) + x = x * base + (sd - hexchars); + + if (endptr) { + /* Unconverted suffix is stored in endptr unless endptr is NULL. */ + *endptr = s1; + } + + return x; +} + +/********************************* Register image ****************************/ + +/* Write a value to a specified register in the register image of the current + thread. Returns status code SUCCESS, E02 or E05. */ +static int +write_register(int regno, char *val) +{ + int status = SUCCESS; + + if (regno >= R0 && regno <= ACR) { + /* Consecutive 32-bit registers. */ + hex2mem((unsigned char *)®.r0 + (regno - R0) * sizeof(unsigned int), + val, sizeof(unsigned int)); + + } else if (regno == BZ || regno == VR || regno == WZ || regno == DZ) { + /* Read-only registers. */ + status = E02; + + } else if (regno == PID) { + /* 32-bit register. (Even though we already checked SRS and WZ, we cannot + combine this with the EXS - SPC write since SRS and WZ have different size.) */ + hex2mem((unsigned char *)®.pid, val, sizeof(unsigned int)); + + } else if (regno == SRS) { + /* 8-bit register. */ + hex2mem((unsigned char *)®.srs, val, sizeof(unsigned char)); + + } else if (regno >= EXS && regno <= SPC) { + /* Consecutive 32-bit registers. */ + hex2mem((unsigned char *)®.exs + (regno - EXS) * sizeof(unsigned int), + val, sizeof(unsigned int)); + + } else if (regno == PC) { + /* Pseudo-register. Treat as read-only. */ + status = E02; + + } else if (regno >= S0 && regno <= S15) { + /* 32-bit registers. */ + hex2mem((unsigned char *)&sreg.s0_0 + (reg.srs * 16 * sizeof(unsigned int)) + (regno - S0) * sizeof(unsigned int), val, sizeof(unsigned int)); + } else { + /* Non-existing register. */ + status = E05; + } + return status; +} + +/* Read a value from a specified register in the register image. Returns the + value in the register or -1 for non-implemented registers. */ +static int +read_register(char regno, unsigned int *valptr) +{ + int status = SUCCESS; + + /* We read the zero registers from the register struct (instead of just returning 0) + to catch errors. */ + + if (regno >= R0 && regno <= ACR) { + /* Consecutive 32-bit registers. */ + *valptr = *(unsigned int *)((char *)®.r0 + (regno - R0) * sizeof(unsigned int)); + + } else if (regno == BZ || regno == VR) { + /* Consecutive 8-bit registers. */ + *valptr = (unsigned int)(*(unsigned char *) + ((char *)®.bz + (regno - BZ) * sizeof(char))); + + } else if (regno == PID) { + /* 32-bit register. */ + *valptr = *(unsigned int *)((char *)®.pid); + + } else if (regno == SRS) { + /* 8-bit register. */ + *valptr = (unsigned int)(*(unsigned char *)((char *)®.srs)); + + } else if (regno == WZ) { + /* 16-bit register. */ + *valptr = (unsigned int)(*(unsigned short *)(char *)®.wz); + + } else if (regno >= EXS && regno <= PC) { + /* Consecutive 32-bit registers. */ + *valptr = *(unsigned int *)((char *)®.exs + (regno - EXS) * sizeof(unsigned int)); + + } else if (regno >= S0 && regno <= S15) { + /* Consecutive 32-bit registers, located elsewhere. */ + *valptr = *(unsigned int *)((char *)&sreg.s0_0 + (reg.srs * 16 * sizeof(unsigned int)) + (regno - S0) * sizeof(unsigned int)); + + } else { + /* Non-existing register. */ + status = E05; + } + return status; + +} + +/********************************** Packet I/O ******************************/ +/* Returns the character equivalent of a nibble, bit 7, 6, 5, and 4 of a byte, + represented by int x. */ +static inline char +highhex(int x) +{ + return hexchars[(x >> 4) & 0xf]; +} + +/* Returns the character equivalent of a nibble, bit 3, 2, 1, and 0 of a byte, + represented by int x. */ +static inline char +lowhex(int x) +{ + return hexchars[x & 0xf]; +} + +/* Returns the integer equivalent of a hexadecimal character. */ +static int +hex(char ch) +{ + if ((ch >= 'a') && (ch <= 'f')) + return (ch - 'a' + 10); + if ((ch >= '0') && (ch <= '9')) + return (ch - '0'); + if ((ch >= 'A') && (ch <= 'F')) + return (ch - 'A' + 10); + return -1; +} + +/* Convert the memory, pointed to by mem into hexadecimal representation. + Put the result in buf, and return a pointer to the last character + in buf (null). */ + +static char * +mem2hex(char *buf, unsigned char *mem, int count) +{ + int i; + int ch; + + if (mem == NULL) { + /* Invalid address, caught by 'm' packet handler. */ + for (i = 0; i < count; i++) { + *buf++ = '0'; + *buf++ = '0'; + } + } else { + /* Valid mem address. */ + for (i = 0; i < count; i++) { + ch = *mem++; + *buf++ = highhex (ch); + *buf++ = lowhex (ch); + } + } + /* Terminate properly. */ + *buf = '\0'; + return buf; +} + +/* Same as mem2hex, but puts it in network byte order. */ +static char * +mem2hex_nbo(char *buf, unsigned char *mem, int count) +{ + int i; + int ch; + + mem += count - 1; + for (i = 0; i < count; i++) { + ch = *mem--; + *buf++ = highhex (ch); + *buf++ = lowhex (ch); + } + + /* Terminate properly. */ + *buf = '\0'; + return buf; +} + +/* Convert the array, in hexadecimal representation, pointed to by buf into + binary representation. Put the result in mem, and return a pointer to + the character after the last byte written. */ +static unsigned char* +hex2mem(unsigned char *mem, char *buf, int count) +{ + int i; + unsigned char ch; + for (i = 0; i < count; i++) { + ch = hex (*buf++) << 4; + ch = ch + hex (*buf++); + *mem++ = ch; + } + return mem; +} + +/* Put the content of the array, in binary representation, pointed to by buf + into memory pointed to by mem, and return a pointer to the character after + the last byte written. + Gdb will escape $, #, and the escape char (0x7d). */ +static unsigned char* +bin2mem(unsigned char *mem, unsigned char *buf, int count) +{ + int i; + unsigned char *next; + for (i = 0; i < count; i++) { + /* Check for any escaped characters. Be paranoid and + only unescape chars that should be escaped. */ + if (*buf == 0x7d) { + next = buf + 1; + if (*next == 0x3 || *next == 0x4 || *next == 0x5D) { + /* #, $, ESC */ + buf++; + *buf += 0x20; + } + } + *mem++ = *buf++; + } + return mem; +} + +/* Await the sequence $<data>#<checksum> and store <data> in the array buffer + returned. */ +static void +getpacket(char *buffer) +{ + unsigned char checksum; + unsigned char xmitcsum; + int i; + int count; + char ch; + + do { + while((ch = getDebugChar ()) != '$') + /* Wait for the start character $ and ignore all other characters */; + checksum = 0; + xmitcsum = -1; + count = 0; + /* Read until a # or the end of the buffer is reached */ + while (count < BUFMAX) { + ch = getDebugChar(); + if (ch == '#') + break; + checksum = checksum + ch; + buffer[count] = ch; + count = count + 1; + } + + if (count >= BUFMAX) + continue; + + buffer[count] = 0; + + if (ch == '#') { + xmitcsum = hex(getDebugChar()) << 4; + xmitcsum += hex(getDebugChar()); + if (checksum != xmitcsum) { + /* Wrong checksum */ + putDebugChar('-'); + } else { + /* Correct checksum */ + putDebugChar('+'); + /* If sequence characters are received, reply with them */ + if (buffer[2] == ':') { + putDebugChar(buffer[0]); + putDebugChar(buffer[1]); + /* Remove the sequence characters from the buffer */ + count = gdb_cris_strlen(buffer); + for (i = 3; i <= count; i++) + buffer[i - 3] = buffer[i]; + } + } + } + } while (checksum != xmitcsum); +} + +/* Send $<data>#<checksum> from the <data> in the array buffer. */ + +static void +putpacket(char *buffer) +{ + int checksum; + int runlen; + int encode; + + do { + char *src = buffer; + putDebugChar('$'); + checksum = 0; + while (*src) { + /* Do run length encoding */ + putDebugChar(*src); + checksum += *src; + runlen = 0; + while (runlen < RUNLENMAX && *src == src[runlen]) { + runlen++; + } + if (runlen > 3) { + /* Got a useful amount */ + putDebugChar ('*'); + checksum += '*'; + encode = runlen + ' ' - 4; + putDebugChar(encode); + checksum += encode; + src += runlen; + } else { + src++; + } + } + putDebugChar('#'); + putDebugChar(highhex (checksum)); + putDebugChar(lowhex (checksum)); + } while(kgdb_started && (getDebugChar() != '+')); +} + +/* The string str is prepended with the GDB printout token and sent. Required + in traditional implementations. */ +void +putDebugString(const unsigned char *str, int len) +{ + /* Move SPC forward if we are single-stepping. */ + asm("spchere:"); + asm("move $spc, $r10"); + asm("cmp.d spchere, $r10"); + asm("bne nosstep"); + asm("nop"); + asm("move.d spccont, $r10"); + asm("move $r10, $spc"); + asm("nosstep:"); + + output_buffer[0] = 'O'; + mem2hex(&output_buffer[1], (unsigned char *)str, len); + putpacket(output_buffer); + + asm("spccont:"); +} + +/********************************** Handle exceptions ************************/ +/* Build and send a response packet in order to inform the host the + stub is stopped. TAAn...:r...;n...:r...;n...:r...; + AA = signal number + n... = register number (hex) + r... = register contents + n... = `thread' + r... = thread process ID. This is a hex integer. + n... = other string not starting with valid hex digit. + gdb should ignore this n,r pair and go on to the next. + This way we can extend the protocol. */ +static void +stub_is_stopped(int sigval) +{ + char *ptr = output_buffer; + unsigned int reg_cont; + + /* Send trap type (converted to signal) */ + + *ptr++ = 'T'; + *ptr++ = highhex(sigval); + *ptr++ = lowhex(sigval); + + if (((reg.exs & 0xff00) >> 8) == 0xc) { + + /* Some kind of hardware watchpoint triggered. Find which one + and determine its type (read/write/access). */ + int S, bp, trig_bits = 0, rw_bits = 0; + int trig_mask = 0; + unsigned int *bp_d_regs = &sreg.s3_3; + /* In a lot of cases, the stopped data address will simply be EDA. + In some cases, we adjust it to match the watched data range. + (We don't want to change the actual EDA though). */ + unsigned int stopped_data_address; + /* The S field of EXS. */ + S = (reg.exs & 0xffff0000) >> 16; + + if (S & 1) { + /* Instruction watchpoint. */ + /* FIXME: Check against, and possibly adjust reported EDA. */ + } else { + /* Data watchpoint. Find the one that triggered. */ + for (bp = 0; bp < 6; bp++) { + + /* Dx_RD, Dx_WR in the S field of EXS for this BP. */ + int bitpos_trig = 1 + bp * 2; + /* Dx_BPRD, Dx_BPWR in BP_CTRL for this BP. */ + int bitpos_config = 2 + bp * 4; + + /* Get read/write trig bits for this BP. */ + trig_bits = (S & (3 << bitpos_trig)) >> bitpos_trig; + + /* Read/write config bits for this BP. */ + rw_bits = (sreg.s0_3 & (3 << bitpos_config)) >> bitpos_config; + if (trig_bits) { + /* Sanity check: the BP shouldn't trigger for accesses + that it isn't configured for. */ + if ((rw_bits == 0x1 && trig_bits != 0x1) || + (rw_bits == 0x2 && trig_bits != 0x2)) + panic("Invalid r/w trigging for this BP"); + + /* Mark this BP as trigged for future reference. */ + trig_mask |= (1 << bp); + + if (reg.eda >= bp_d_regs[bp * 2] && + reg.eda <= bp_d_regs[bp * 2 + 1]) { + /* EDA withing range for this BP; it must be the one + we're looking for. */ + stopped_data_address = reg.eda; + break; + } + } + } + if (bp < 6) { + /* Found a trigged BP with EDA within its configured data range. */ + } else if (trig_mask) { + /* Something triggered, but EDA doesn't match any BP's range. */ + for (bp = 0; bp < 6; bp++) { + /* Dx_BPRD, Dx_BPWR in BP_CTRL for this BP. */ + int bitpos_config = 2 + bp * 4; + + /* Read/write config bits for this BP (needed later). */ + rw_bits = (sreg.s0_3 & (3 << bitpos_config)) >> bitpos_config; + + if (trig_mask & (1 << bp)) { + /* EDA within 31 bytes of the configured start address? */ + if (reg.eda + 31 >= bp_d_regs[bp * 2]) { + /* Changing the reported address to match + the start address of the first applicable BP. */ + stopped_data_address = bp_d_regs[bp * 2]; + break; + } else { + /* We continue since we might find another useful BP. */ + printk("EDA doesn't match trigged BP's range"); + } + } + } + } + + /* No match yet? */ + BUG_ON(bp >= 6); + /* Note that we report the type according to what the BP is configured + for (otherwise we'd never report an 'awatch'), not according to how + it trigged. We did check that the trigged bits match what the BP is + configured for though. */ + if (rw_bits == 0x1) { + /* read */ + strncpy(ptr, "rwatch", 6); + ptr += 6; + } else if (rw_bits == 0x2) { + /* write */ + strncpy(ptr, "watch", 5); + ptr += 5; + } else if (rw_bits == 0x3) { + /* access */ + strncpy(ptr, "awatch", 6); + ptr += 6; + } else { + panic("Invalid r/w bits for this BP."); + } + + *ptr++ = ':'; + /* Note that we don't read_register(EDA, ...) */ + ptr = mem2hex_nbo(ptr, (unsigned char *)&stopped_data_address, register_size[EDA]); + *ptr++ = ';'; + } + } + /* Only send PC, frame and stack pointer. */ + read_register(PC, ®_cont); + *ptr++ = highhex(PC); + *ptr++ = lowhex(PC); + *ptr++ = ':'; + ptr = mem2hex(ptr, (unsigned char *)®_cont, register_size[PC]); + *ptr++ = ';'; + + read_register(R8, ®_cont); + *ptr++ = highhex(R8); + *ptr++ = lowhex(R8); + *ptr++ = ':'; + ptr = mem2hex(ptr, (unsigned char *)®_cont, register_size[R8]); + *ptr++ = ';'; + + read_register(SP, ®_cont); + *ptr++ = highhex(SP); + *ptr++ = lowhex(SP); + *ptr++ = ':'; + ptr = mem2hex(ptr, (unsigned char *)®_cont, register_size[SP]); + *ptr++ = ';'; + + /* Send ERP as well; this will save us an entire register fetch in some cases. */ + read_register(ERP, ®_cont); + *ptr++ = highhex(ERP); + *ptr++ = lowhex(ERP); + *ptr++ = ':'; + ptr = mem2hex(ptr, (unsigned char *)®_cont, register_size[ERP]); + *ptr++ = ';'; + + /* null-terminate and send it off */ + *ptr = 0; + putpacket(output_buffer); +} + +/* Returns the size of an instruction that has a delay slot. */ + +int insn_size(unsigned long pc) +{ + unsigned short opcode = *(unsigned short *)pc; + int size = 0; + + switch ((opcode & 0x0f00) >> 8) { + case 0x0: + case 0x9: + case 0xb: + size = 2; + break; + case 0xe: + case 0xf: + size = 6; + break; + case 0xd: + /* Could be 4 or 6; check more bits. */ + if ((opcode & 0xff) == 0xff) + size = 4; + else + size = 6; + break; + default: + panic("Couldn't find size of opcode 0x%x at 0x%lx\n", opcode, pc); + } + + return size; +} + +void register_fixup(int sigval) +{ + /* Compensate for ACR push at the beginning of exception handler. */ + reg.sp += 4; + + /* Standard case. */ + reg.pc = reg.erp; + if (reg.erp & 0x1) { + /* Delay slot bit set. Report as stopped on proper instruction. */ + if (reg.spc) { + /* Rely on SPC if set. */ + reg.pc = reg.spc; + } else { + /* Calculate the PC from the size of the instruction + that the delay slot we're in belongs to. */ + reg.pc += insn_size(reg.erp & ~1) - 1 ; + } + } + + if ((reg.exs & 0x3) == 0x0) { + /* Bits 1 - 0 indicate the type of memory operation performed + by the interrupted instruction. 0 means no memory operation, + and EDA is undefined in that case. We zero it to avoid confusion. */ + reg.eda = 0; + } + + if (sigval == SIGTRAP) { + /* Break 8, single step or hardware breakpoint exception. */ + + /* Check IDX field of EXS. */ + if (((reg.exs & 0xff00) >> 8) == 0x18) { + + /* Break 8. */ + + /* Static (compiled) breakpoints must return to the next instruction + in order to avoid infinite loops (default value of ERP). Dynamic + (gdb-invoked) must subtract the size of the break instruction from + the ERP so that the instruction that was originally in the break + instruction's place will be run when we return from the exception. */ + if (!dynamic_bp) { + /* Assuming that all breakpoints are dynamic from now on. */ + dynamic_bp = 1; + } else { + + /* Only if not in a delay slot. */ + if (!(reg.erp & 0x1)) { + reg.erp -= 2; + reg.pc -= 2; + } + } + + } else if (((reg.exs & 0xff00) >> 8) == 0x3) { + /* Single step. */ + /* Don't fiddle with S1. */ + + } else if (((reg.exs & 0xff00) >> 8) == 0xc) { + + /* Hardware watchpoint exception. */ + + /* SPC has been updated so that we will get a single step exception + when we return, but we don't want that. */ + reg.spc = 0; + + /* Don't fiddle with S1. */ + } + + } else if (sigval == SIGINT) { + /* Nothing special. */ + } +} + +static void insert_watchpoint(char type, int addr, int len) +{ + /* Breakpoint/watchpoint types (GDB terminology): + 0 = memory breakpoint for instructions + (not supported; done via memory write instead) + 1 = hardware breakpoint for instructions (supported) + 2 = write watchpoint (supported) + 3 = read watchpoint (supported) + 4 = access watchpoint (supported) */ + + if (type < '1' || type > '4') { + output_buffer[0] = 0; + return; + } + + /* Read watchpoints are set as access watchpoints, because of GDB's + inability to deal with pure read watchpoints. */ + if (type == '3') + type = '4'; + + if (type == '1') { + /* Hardware (instruction) breakpoint. */ + /* Bit 0 in BP_CTRL holds the configuration for I0. */ + if (sreg.s0_3 & 0x1) { + /* Already in use. */ + gdb_cris_strcpy(output_buffer, error_message[E04]); + return; + } + /* Configure. */ + sreg.s1_3 = addr; + sreg.s2_3 = (addr + len - 1); + sreg.s0_3 |= 1; + } else { + int bp; + unsigned int *bp_d_regs = &sreg.s3_3; + + /* The watchpoint allocation scheme is the simplest possible. + For example, if a region is watched for read and + a write watch is requested, a new watchpoint will + be used. Also, if a watch for a region that is already + covered by one or more existing watchpoints, a new + watchpoint will be used. */ + + /* First, find a free data watchpoint. */ + for (bp = 0; bp < 6; bp++) { + /* Each data watchpoint's control registers occupy 2 bits + (hence the 3), starting at bit 2 for D0 (hence the 2) + with 4 bits between for each watchpoint (yes, the 4). */ + if (!(sreg.s0_3 & (0x3 << (2 + (bp * 4))))) { + break; + } + } + + if (bp > 5) { + /* We're out of watchpoints. */ + gdb_cris_strcpy(output_buffer, error_message[E04]); + return; + } + + /* Configure the control register first. */ + if (type == '3' || type == '4') { + /* Trigger on read. */ + sreg.s0_3 |= (1 << (2 + bp * 4)); + } + if (type == '2' || type == '4') { + /* Trigger on write. */ + sreg.s0_3 |= (2 << (2 + bp * 4)); + } + + /* Ugly pointer arithmetics to configure the watched range. */ + bp_d_regs[bp * 2] = addr; + bp_d_regs[bp * 2 + 1] = (addr + len - 1); + } + + /* Set the S1 flag to enable watchpoints. */ + reg.ccs |= (1 << (S_CCS_BITNR + CCS_SHIFT)); + gdb_cris_strcpy(output_buffer, "OK"); +} + +static void remove_watchpoint(char type, int addr, int len) +{ + /* Breakpoint/watchpoint types: + 0 = memory breakpoint for instructions + (not supported; done via memory write instead) + 1 = hardware breakpoint for instructions (supported) + 2 = write watchpoint (supported) + 3 = read watchpoint (supported) + 4 = access watchpoint (supported) */ + if (type < '1' || type > '4') { + output_buffer[0] = 0; + return; + } + + /* Read watchpoints are set as access watchpoints, because of GDB's + inability to deal with pure read watchpoints. */ + if (type == '3') + type = '4'; + + if (type == '1') { + /* Hardware breakpoint. */ + /* Bit 0 in BP_CTRL holds the configuration for I0. */ + if (!(sreg.s0_3 & 0x1)) { + /* Not in use. */ + gdb_cris_strcpy(output_buffer, error_message[E04]); + return; + } + /* Deconfigure. */ + sreg.s1_3 = 0; + sreg.s2_3 = 0; + sreg.s0_3 &= ~1; + } else { + int bp; + unsigned int *bp_d_regs = &sreg.s3_3; + /* Try to find a watchpoint that is configured for the + specified range, then check that read/write also matches. */ + + /* Ugly pointer arithmetic, since I cannot rely on a + single switch (addr) as there may be several watchpoints with + the same start address for example. */ + + for (bp = 0; bp < 6; bp++) { + if (bp_d_regs[bp * 2] == addr && + bp_d_regs[bp * 2 + 1] == (addr + len - 1)) { + /* Matching range. */ + int bitpos = 2 + bp * 4; + int rw_bits; + + /* Read/write bits for this BP. */ + rw_bits = (sreg.s0_3 & (0x3 << bitpos)) >> bitpos; + + if ((type == '3' && rw_bits == 0x1) || + (type == '2' && rw_bits == 0x2) || + (type == '4' && rw_bits == 0x3)) { + /* Read/write matched. */ + break; + } + } + } + + if (bp > 5) { + /* No watchpoint matched. */ + gdb_cris_strcpy(output_buffer, error_message[E04]); + return; + } + + /* Found a matching watchpoint. Now, deconfigure it by + both disabling read/write in bp_ctrl and zeroing its + start/end addresses. */ + sreg.s0_3 &= ~(3 << (2 + (bp * 4))); + bp_d_regs[bp * 2] = 0; + bp_d_regs[bp * 2 + 1] = 0; + } + + /* Note that we don't clear the S1 flag here. It's done when continuing. */ + gdb_cris_strcpy(output_buffer, "OK"); +} + + + +/* All expected commands are sent from remote.c. Send a response according + to the description in remote.c. */ +void +handle_exception(int sigval) +{ + /* Avoid warning of not used. */ + + USEDFUN(handle_exception); + USEDVAR(internal_stack[0]); + + register_fixup(sigval); + + /* Send response. */ + stub_is_stopped(sigval); + + for (;;) { + output_buffer[0] = '\0'; + getpacket(input_buffer); + switch (input_buffer[0]) { + case 'g': + /* Read registers: g + Success: Each byte of register data is described by two hex digits. + Registers are in the internal order for GDB, and the bytes + in a register are in the same order the machine uses. + Failure: void. */ + { + char *buf; + /* General and special registers. */ + buf = mem2hex(output_buffer, (char *)®, sizeof(registers)); + /* Support registers. */ + /* -1 because of the null termination that mem2hex adds. */ + mem2hex(buf, + (char *)&sreg + (reg.srs * 16 * sizeof(unsigned int)), + 16 * sizeof(unsigned int)); + break; + } + case 'G': + /* Write registers. GXX..XX + Each byte of register data is described by two hex digits. + Success: OK + Failure: void. */ + /* General and special registers. */ + hex2mem((char *)®, &input_buffer[1], sizeof(registers)); + /* Support registers. */ + hex2mem((char *)&sreg + (reg.srs * 16 * sizeof(unsigned int)), + &input_buffer[1] + sizeof(registers), + 16 * sizeof(unsigned int)); + gdb_cris_strcpy(output_buffer, "OK"); + break; + + case 'P': + /* Write register. Pn...=r... + Write register n..., hex value without 0x, with value r..., + which contains a hex value without 0x and two hex digits + for each byte in the register (target byte order). P1f=11223344 means + set register 31 to 44332211. + Success: OK + Failure: E02, E05 */ + { + char *suffix; + int regno = gdb_cris_strtol(&input_buffer[1], &suffix, 16); + int status; + + status = write_register(regno, suffix+1); + + switch (status) { + case E02: + /* Do not support read-only registers. */ + gdb_cris_strcpy(output_buffer, error_message[E02]); + break; + case E05: + /* Do not support non-existing registers. */ + gdb_cris_strcpy(output_buffer, error_message[E05]); + break; + default: + /* Valid register number. */ + gdb_cris_strcpy(output_buffer, "OK"); + break; + } + } + break; + + case 'm': + /* Read from memory. mAA..AA,LLLL + AA..AA is the address and LLLL is the length. + Success: XX..XX is the memory content. Can be fewer bytes than + requested if only part of the data may be read. m6000120a,6c means + retrieve 108 byte from base address 6000120a. + Failure: void. */ + { + char *suffix; + unsigned char *addr = (unsigned char *)gdb_cris_strtol(&input_buffer[1], + &suffix, 16); + int len = gdb_cris_strtol(suffix+1, 0, 16); + + /* Bogus read (i.e. outside the kernel's + segment)? . */ + if (!((unsigned int)addr >= 0xc0000000 && + (unsigned int)addr < 0xd0000000)) + addr = NULL; + + mem2hex(output_buffer, addr, len); + } + break; + + case 'X': + /* Write to memory. XAA..AA,LLLL:XX..XX + AA..AA is the start address, LLLL is the number of bytes, and + XX..XX is the binary data. + Success: OK + Failure: void. */ + case 'M': + /* Write to memory. MAA..AA,LLLL:XX..XX + AA..AA is the start address, LLLL is the number of bytes, and + XX..XX is the hexadecimal data. + Success: OK + Failure: void. */ + { + char *lenptr; + char *dataptr; + unsigned char *addr = (unsigned char *)gdb_cris_strtol(&input_buffer[1], + &lenptr, 16); + int len = gdb_cris_strtol(lenptr+1, &dataptr, 16); + if (*lenptr == ',' && *dataptr == ':') { + if (input_buffer[0] == 'M') { + hex2mem(addr, dataptr + 1, len); + } else /* X */ { + bin2mem(addr, dataptr + 1, len); + } + gdb_cris_strcpy(output_buffer, "OK"); + } + else { + gdb_cris_strcpy(output_buffer, error_message[E06]); + } + } + break; + + case 'c': + /* Continue execution. cAA..AA + AA..AA is the address where execution is resumed. If AA..AA is + omitted, resume at the present address. + Success: return to the executing thread. + Failure: will never know. */ + + if (input_buffer[1] != '\0') { + /* FIXME: Doesn't handle address argument. */ + gdb_cris_strcpy(output_buffer, error_message[E04]); + break; + } + + /* Before continuing, make sure everything is set up correctly. */ + + /* Set the SPC to some unlikely value. */ + reg.spc = 0; + /* Set the S1 flag to 0 unless some watchpoint is enabled (since setting + S1 to 0 would also disable watchpoints). (Note that bits 26-31 in BP_CTRL + are reserved, so don't check against those). */ + if ((sreg.s0_3 & 0x3fff) == 0) { + reg.ccs &= ~(1 << (S_CCS_BITNR + CCS_SHIFT)); + } + + return; + + case 's': + /* Step. sAA..AA + AA..AA is the address where execution is resumed. If AA..AA is + omitted, resume at the present address. Success: return to the + executing thread. Failure: will never know. */ + + if (input_buffer[1] != '\0') { + /* FIXME: Doesn't handle address argument. */ + gdb_cris_strcpy(output_buffer, error_message[E04]); + break; + } + + /* Set the SPC to PC, which is where we'll return + (deduced previously). */ + reg.spc = reg.pc; + + /* Set the S1 (first stacked, not current) flag, which will + kick into action when we rfe. */ + reg.ccs |= (1 << (S_CCS_BITNR + CCS_SHIFT)); + return; + + case 'Z': + + /* Insert breakpoint or watchpoint, Ztype,addr,length. + Remote protocol says: A remote target shall return an empty string + for an unrecognized breakpoint or watchpoint packet type. */ + { + char *lenptr; + char *dataptr; + int addr = gdb_cris_strtol(&input_buffer[3], &lenptr, 16); + int len = gdb_cris_strtol(lenptr + 1, &dataptr, 16); + char type = input_buffer[1]; + + insert_watchpoint(type, addr, len); + break; + } + + case 'z': + /* Remove breakpoint or watchpoint, Ztype,addr,length. + Remote protocol says: A remote target shall return an empty string + for an unrecognized breakpoint or watchpoint packet type. */ + { + char *lenptr; + char *dataptr; + int addr = gdb_cris_strtol(&input_buffer[3], &lenptr, 16); + int len = gdb_cris_strtol(lenptr + 1, &dataptr, 16); + char type = input_buffer[1]; + + remove_watchpoint(type, addr, len); + break; + } + + + case '?': + /* The last signal which caused a stop. ? + Success: SAA, where AA is the signal number. + Failure: void. */ + output_buffer[0] = 'S'; + output_buffer[1] = highhex(sigval); + output_buffer[2] = lowhex(sigval); + output_buffer[3] = 0; + break; + + case 'D': + /* Detach from host. D + Success: OK, and return to the executing thread. + Failure: will never know */ + putpacket("OK"); + return; + + case 'k': + case 'r': + /* kill request or reset request. + Success: restart of target. + Failure: will never know. */ + kill_restart(); + break; + + case 'C': + case 'S': + case '!': + case 'R': + case 'd': + /* Continue with signal sig. Csig;AA..AA + Step with signal sig. Ssig;AA..AA + Use the extended remote protocol. ! + Restart the target system. R0 + Toggle debug flag. d + Search backwards. tAA:PP,MM + Not supported: E04 */ + + /* FIXME: What's the difference between not supported + and ignored (below)? */ + gdb_cris_strcpy(output_buffer, error_message[E04]); + break; + + default: + /* The stub should ignore other request and send an empty + response ($#<checksum>). This way we can extend the protocol and GDB + can tell whether the stub it is talking to uses the old or the new. */ + output_buffer[0] = 0; + break; + } + putpacket(output_buffer); + } +} + +void +kgdb_init(void) +{ + reg_intr_vect_rw_mask intr_mask; + reg_ser_rw_intr_mask ser_intr_mask; + + /* Configure the kgdb serial port. */ +#if defined(CONFIG_ETRAX_KGDB_PORT0) + /* Note: no shortcut registered (not handled by multiple_interrupt). + See entry.S. */ + set_exception_vector(SER0_INTR_VECT, kgdb_handle_exception); + /* Enable the ser irq in the global config. */ + intr_mask = REG_RD(intr_vect, regi_irq, rw_mask); + intr_mask.ser0 = 1; + REG_WR(intr_vect, regi_irq, rw_mask, intr_mask); + + ser_intr_mask = REG_RD(ser, regi_ser0, rw_intr_mask); + ser_intr_mask.data_avail = regk_ser_yes; + REG_WR(ser, regi_ser0, rw_intr_mask, ser_intr_mask); +#elif defined(CONFIG_ETRAX_KGDB_PORT1) + /* Note: no shortcut registered (not handled by multiple_interrupt). + See entry.S. */ + set_exception_vector(SER1_INTR_VECT, kgdb_handle_exception); + /* Enable the ser irq in the global config. */ + intr_mask = REG_RD(intr_vect, regi_irq, rw_mask); + intr_mask.ser1 = 1; + REG_WR(intr_vect, regi_irq, rw_mask, intr_mask); + + ser_intr_mask = REG_RD(ser, regi_ser1, rw_intr_mask); + ser_intr_mask.data_avail = regk_ser_yes; + REG_WR(ser, regi_ser1, rw_intr_mask, ser_intr_mask); +#elif defined(CONFIG_ETRAX_KGDB_PORT2) + /* Note: no shortcut registered (not handled by multiple_interrupt). + See entry.S. */ + set_exception_vector(SER2_INTR_VECT, kgdb_handle_exception); + /* Enable the ser irq in the global config. */ + intr_mask = REG_RD(intr_vect, regi_irq, rw_mask); + intr_mask.ser2 = 1; + REG_WR(intr_vect, regi_irq, rw_mask, intr_mask); + + ser_intr_mask = REG_RD(ser, regi_ser2, rw_intr_mask); + ser_intr_mask.data_avail = regk_ser_yes; + REG_WR(ser, regi_ser2, rw_intr_mask, ser_intr_mask); +#elif defined(CONFIG_ETRAX_KGDB_PORT3) + /* Note: no shortcut registered (not handled by multiple_interrupt). + See entry.S. */ + set_exception_vector(SER3_INTR_VECT, kgdb_handle_exception); + /* Enable the ser irq in the global config. */ + intr_mask = REG_RD(intr_vect, regi_irq, rw_mask); + intr_mask.ser3 = 1; + REG_WR(intr_vect, regi_irq, rw_mask, intr_mask); + + ser_intr_mask = REG_RD(ser, regi_ser3, rw_intr_mask); + ser_intr_mask.data_avail = regk_ser_yes; + REG_WR(ser, regi_ser3, rw_intr_mask, ser_intr_mask); +#endif + +} +/* Performs a complete re-start from scratch. */ +static void +kill_restart(void) +{ + machine_restart(""); +} + +/* Use this static breakpoint in the start-up only. */ + +void +breakpoint(void) +{ + kgdb_started = 1; + dynamic_bp = 0; /* This is a static, not a dynamic breakpoint. */ + __asm__ volatile ("break 8"); /* Jump to kgdb_handle_breakpoint. */ +} + +/****************************** End of file **********************************/ diff --git a/arch/cris/arch-v32/kernel/kgdb_asm.S b/arch/cris/arch-v32/kernel/kgdb_asm.S new file mode 100644 index 000000000000..b350dd279ed2 --- /dev/null +++ b/arch/cris/arch-v32/kernel/kgdb_asm.S @@ -0,0 +1,552 @@ +/* + * Copyright (C) 2004 Axis Communications AB + * + * Code for handling break 8, hardware breakpoint, single step, and serial + * port exceptions for kernel debugging purposes. + */ + +#include <linux/config.h> +#include <asm/arch/hwregs/intr_vect.h> + + ;; Exported functions. + .globl kgdb_handle_exception + +kgdb_handle_exception: + +;; Create a register image of the caller. +;; +;; First of all, save the ACR on the stack since we need it for address calculations. +;; We put it into the register struct later. + + subq 4, $sp + move.d $acr, [$sp] + +;; Now we are free to use ACR all we want. +;; If we were running this handler with interrupts on, we would have to be careful +;; to save and restore CCS manually, but since we aren't we treat it like every other +;; register. + + move.d reg, $acr + move.d $r0, [$acr] ; Save R0 (start of register struct) + addq 4, $acr + move.d $r1, [$acr] ; Save R1 + addq 4, $acr + move.d $r2, [$acr] ; Save R2 + addq 4, $acr + move.d $r3, [$acr] ; Save R3 + addq 4, $acr + move.d $r4, [$acr] ; Save R4 + addq 4, $acr + move.d $r5, [$acr] ; Save R5 + addq 4, $acr + move.d $r6, [$acr] ; Save R6 + addq 4, $acr + move.d $r7, [$acr] ; Save R7 + addq 4, $acr + move.d $r8, [$acr] ; Save R8 + addq 4, $acr + move.d $r9, [$acr] ; Save R9 + addq 4, $acr + move.d $r10, [$acr] ; Save R10 + addq 4, $acr + move.d $r11, [$acr] ; Save R11 + addq 4, $acr + move.d $r12, [$acr] ; Save R12 + addq 4, $acr + move.d $r13, [$acr] ; Save R13 + addq 4, $acr + move.d $sp, [$acr] ; Save SP (R14) + addq 4, $acr + + ;; The ACR register is already saved on the stack, so pop it from there. + move.d [$sp],$r0 + move.d $r0, [$acr] + addq 4, $acr + + move $bz, [$acr] + addq 1, $acr + move $vr, [$acr] + addq 1, $acr + move $pid, [$acr] + addq 4, $acr + move $srs, [$acr] + addq 1, $acr + move $wz, [$acr] + addq 2, $acr + move $exs, [$acr] + addq 4, $acr + move $eda, [$acr] + addq 4, $acr + move $mof, [$acr] + addq 4, $acr + move $dz, [$acr] + addq 4, $acr + move $ebp, [$acr] + addq 4, $acr + move $erp, [$acr] + addq 4, $acr + move $srp, [$acr] + addq 4, $acr + move $nrp, [$acr] + addq 4, $acr + move $ccs, [$acr] + addq 4, $acr + move $usp, [$acr] + addq 4, $acr + move $spc, [$acr] + addq 4, $acr + +;; Skip the pseudo-PC. + addq 4, $acr + +;; Save the support registers in bank 0 - 3. + clear.d $r1 ; Bank counter + move.d sreg, $acr + +;; Bank 0 + move $r1, $srs + nop + nop + nop + move $s0, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s1, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s2, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s3, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s4, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s5, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s6, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s7, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s8, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s9, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s10, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s11, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s12, $r0 + move.d $r0, [$acr] + addq 4, $acr + + ;; Nothing in S13 - S15, bank 0 + clear.d [$acr] + addq 4, $acr + clear.d [$acr] + addq 4, $acr + clear.d [$acr] + addq 4, $acr + +;; Bank 1 and bank 2 have the same layout, hence the loop. + addq 1, $r1 +1: + move $r1, $srs + nop + nop + nop + move $s0, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s1, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s2, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s3, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s4, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s5, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s6, $r0 + move.d $r0, [$acr] + addq 4, $acr + + ;; Nothing in S7 - S15, bank 1 and 2 + clear.d [$acr] + addq 4, $acr + clear.d [$acr] + addq 4, $acr + clear.d [$acr] + addq 4, $acr + clear.d [$acr] + addq 4, $acr + clear.d [$acr] + addq 4, $acr + clear.d [$acr] + addq 4, $acr + clear.d [$acr] + addq 4, $acr + clear.d [$acr] + addq 4, $acr + clear.d [$acr] + addq 4, $acr + + addq 1, $r1 + cmpq 3, $r1 + bne 1b + nop + +;; Bank 3 + move $r1, $srs + nop + nop + nop + move $s0, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s1, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s2, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s3, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s4, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s5, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s6, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s7, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s8, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s9, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s10, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s11, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s12, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s13, $r0 + move.d $r0, [$acr] + addq 4, $acr + move $s14, $r0 + move.d $r0, [$acr] + addq 4, $acr +;; Nothing in S15, bank 3 + clear.d [$acr] + addq 4, $acr + +;; Check what got us here: get IDX field of EXS. + move $exs, $r10 + and.d 0xff00, $r10 + lsrq 8, $r10 +#if defined(CONFIG_ETRAX_KGDB_PORT0) + cmp.d SER0_INTR_VECT, $r10 ; IRQ for serial port 0 + beq sigint + nop +#elif defined(CONFIG_ETRAX_KGDB_PORT1) + cmp.d SER1_INTR_VECT, $r10 ; IRQ for serial port 1 + beq sigint + nop +#elif defined(CONFIG_ETRAX_KGDB_PORT2) + cmp.d SER2_INTR_VECT, $r10 ; IRQ for serial port 2 + beq sigint + nop +#elif defined(CONFIG_ETRAX_KGDB_PORT3) + cmp.d SER3_INTR_VECT, $r10 ; IRQ for serial port 3 + beq sigint + nop +#endif +;; Multiple interrupt must be due to serial break. + cmp.d 0x30, $r10 ; Multiple interrupt + beq sigint + nop +;; Neither of those? Then it's a sigtrap. + ba handle_comm + moveq 5, $r10 ; Set SIGTRAP (delay slot) + +sigint: + ;; Serial interrupt; get character + jsr getDebugChar + nop ; Delay slot + cmp.b 3, $r10 ; \003 (Ctrl-C)? + bne return ; No, get out of here + nop + moveq 2, $r10 ; Set SIGINT + +;; +;; Handle the communication +;; +handle_comm: + move.d internal_stack+1020, $sp ; Use the internal stack which grows upwards + jsr handle_exception ; Interactive routine + nop + +;; +;; Return to the caller +;; +return: + +;; First of all, write the support registers. + clear.d $r1 ; Bank counter + move.d sreg, $acr + +;; Bank 0 + move $r1, $srs + nop + nop + nop + move.d [$acr], $r0 + move $r0, $s0 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s1 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s2 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s3 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s4 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s5 + addq 4, $acr + +;; Nothing in S6 - S7, bank 0. + addq 4, $acr + addq 4, $acr + + move.d [$acr], $r0 + move $r0, $s8 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s9 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s10 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s11 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s12 + addq 4, $acr + +;; Nothing in S13 - S15, bank 0 + addq 4, $acr + addq 4, $acr + addq 4, $acr + +;; Bank 1 and bank 2 have the same layout, hence the loop. + addq 1, $r1 +2: + move $r1, $srs + nop + nop + nop + move.d [$acr], $r0 + move $r0, $s0 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s1 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s2 + addq 4, $acr + +;; S3 (MM_CAUSE) is read-only. + addq 4, $acr + + move.d [$acr], $r0 + move $r0, $s4 + addq 4, $acr + +;; FIXME: Actually write S5/S6? (Affects MM_CAUSE.) + addq 4, $acr + addq 4, $acr + +;; Nothing in S7 - S15, bank 1 and 2 + addq 4, $acr + addq 4, $acr + addq 4, $acr + addq 4, $acr + addq 4, $acr + addq 4, $acr + addq 4, $acr + addq 4, $acr + addq 4, $acr + + addq 1, $r1 + cmpq 3, $r1 + bne 2b + nop + +;; Bank 3 + move $r1, $srs + nop + nop + nop + move.d [$acr], $r0 + move $r0, $s0 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s1 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s2 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s3 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s4 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s5 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s6 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s7 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s8 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s9 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s10 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s11 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s12 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s13 + addq 4, $acr + move.d [$acr], $r0 + move $r0, $s14 + addq 4, $acr + +;; Nothing in S15, bank 3 + addq 4, $acr + +;; Now, move on to the regular register restoration process. + + move.d reg, $acr ; Reset ACR to point at the beginning of the register image + move.d [$acr], $r0 ; Restore R0 + addq 4, $acr + move.d [$acr], $r1 ; Restore R1 + addq 4, $acr + move.d [$acr], $r2 ; Restore R2 + addq 4, $acr + move.d [$acr], $r3 ; Restore R3 + addq 4, $acr + move.d [$acr], $r4 ; Restore R4 + addq 4, $acr + move.d [$acr], $r5 ; Restore R5 + addq 4, $acr + move.d [$acr], $r6 ; Restore R6 + addq 4, $acr + move.d [$acr], $r7 ; Restore R7 + addq 4, $acr + move.d [$acr], $r8 ; Restore R8 + addq 4, $acr + move.d [$acr], $r9 ; Restore R9 + addq 4, $acr + move.d [$acr], $r10 ; Restore R10 + addq 4, $acr + move.d [$acr], $r11 ; Restore R11 + addq 4, $acr + move.d [$acr], $r12 ; Restore R12 + addq 4, $acr + move.d [$acr], $r13 ; Restore R13 + +;; +;; We restore all registers, even though some of them probably haven't changed. +;; + + addq 4, $acr + move.d [$acr], $sp ; Restore SP (R14) + + ;; ACR cannot be restored just yet. + addq 8, $acr + + ;; Skip BZ, VR. + addq 2, $acr + + move [$acr], $pid ; Restore PID + addq 4, $acr + move [$acr], $srs ; Restore SRS + nop + nop + nop + addq 1, $acr + + ;; Skip WZ. + addq 2, $acr + + move [$acr], $exs ; Restore EXS. + addq 4, $acr + move [$acr], $eda ; Restore EDA. + addq 4, $acr + move [$acr], $mof ; Restore MOF. + + ;; Skip DZ. + addq 8, $acr + + move [$acr], $ebp ; Restore EBP. + addq 4, $acr + move [$acr], $erp ; Restore ERP. + addq 4, $acr + move [$acr], $srp ; Restore SRP. + addq 4, $acr + move [$acr], $nrp ; Restore NRP. + addq 4, $acr + move [$acr], $ccs ; Restore CCS like an ordinary register. + addq 4, $acr + move [$acr], $usp ; Restore USP + addq 4, $acr + move [$acr], $spc ; Restore SPC + ; No restoration of pseudo-PC of course. + + move.d reg, $acr ; Reset ACR to point at the beginning of the register image + add.d 15*4, $acr + move.d [$acr], $acr ; Finally, restore ACR. + rete ; Same as jump ERP + rfe ; Shifts CCS diff --git a/arch/cris/arch-v32/kernel/pinmux.c b/arch/cris/arch-v32/kernel/pinmux.c new file mode 100644 index 000000000000..a2b8aa37c1bf --- /dev/null +++ b/arch/cris/arch-v32/kernel/pinmux.c @@ -0,0 +1,229 @@ +/* + * Allocator for I/O pins. All pins are allocated to GPIO at bootup. + * Unassigned pins and GPIO pins can be allocated to a fixed interface + * or the I/O processor instead. + * + * Copyright (c) 2004 Axis Communications AB. + */ + +#include <linux/init.h> +#include <linux/errno.h> +#include <linux/kernel.h> +#include <linux/string.h> +#include <linux/spinlock.h> +#include <asm/arch/hwregs/reg_map.h> +#include <asm/arch/hwregs/reg_rdwr.h> +#include <asm/arch/pinmux.h> +#include <asm/arch/hwregs/pinmux_defs.h> + +#undef DEBUG + +#define PORT_PINS 18 +#define PORTS 4 + +static char pins[PORTS][PORT_PINS]; +static DEFINE_SPINLOCK(pinmux_lock); + +static void crisv32_pinmux_set(int port); + +int +crisv32_pinmux_init(void) +{ + static int initialized = 0; + + if (!initialized) { + reg_pinmux_rw_pa pa = REG_RD(pinmux, regi_pinmux, rw_pa); + initialized = 1; + pa.pa0 = pa.pa1 = pa.pa2 = pa.pa3 = + pa.pa4 = pa.pa5 = pa.pa6 = pa.pa7 = regk_pinmux_yes; + REG_WR(pinmux, regi_pinmux, rw_pa, pa); + crisv32_pinmux_alloc(PORT_B, 0, PORT_PINS - 1, pinmux_gpio); + crisv32_pinmux_alloc(PORT_C, 0, PORT_PINS - 1, pinmux_gpio); + crisv32_pinmux_alloc(PORT_D, 0, PORT_PINS - 1, pinmux_gpio); + crisv32_pinmux_alloc(PORT_E, 0, PORT_PINS - 1, pinmux_gpio); + } + + return 0; +} + +int +crisv32_pinmux_alloc(int port, int first_pin, int last_pin, enum pin_mode mode) +{ + int i; + unsigned long flags; + + crisv32_pinmux_init(); + + if (port > PORTS) + return -EINVAL; + + spin_lock_irqsave(&pinmux_lock, flags); + + for (i = first_pin; i <= last_pin; i++) + { + if ((pins[port][i] != pinmux_none) && (pins[port][i] != pinmux_gpio) && + (pins[port][i] != mode)) + { + spin_unlock_irqrestore(&pinmux_lock, flags); +#ifdef DEBUG + panic("Pinmux alloc failed!\n"); +#endif + return -EPERM; + } + } + + for (i = first_pin; i <= last_pin; i++) + pins[port][i] = mode; + + crisv32_pinmux_set(port); + + spin_unlock_irqrestore(&pinmux_lock, flags); + + return 0; +} + +int +crisv32_pinmux_alloc_fixed(enum fixed_function function) +{ + int ret = -EINVAL; + char saved[sizeof pins]; + unsigned long flags; + + spin_lock_irqsave(&pinmux_lock, flags); + + /* Save internal data for recovery */ + memcpy(saved, pins, sizeof pins); + + reg_pinmux_rw_hwprot hwprot = REG_RD(pinmux, regi_pinmux, rw_hwprot); + + switch(function) + { + case pinmux_ser1: + ret = crisv32_pinmux_alloc(PORT_C, 4, 7, pinmux_fixed); + hwprot.ser1 = regk_pinmux_yes; + break; + case pinmux_ser2: + ret = crisv32_pinmux_alloc(PORT_C, 8, 11, pinmux_fixed); + hwprot.ser2 = regk_pinmux_yes; + break; + case pinmux_ser3: + ret = crisv32_pinmux_alloc(PORT_C, 12, 15, pinmux_fixed); + hwprot.ser3 = regk_pinmux_yes; + break; + case pinmux_sser0: + ret = crisv32_pinmux_alloc(PORT_C, 0, 3, pinmux_fixed); + ret |= crisv32_pinmux_alloc(PORT_C, 16, 16, pinmux_fixed); + hwprot.sser0 = regk_pinmux_yes; + break; + case pinmux_sser1: + ret = crisv32_pinmux_alloc(PORT_D, 0, 4, pinmux_fixed); + hwprot.sser1 = regk_pinmux_yes; + break; + case pinmux_ata0: + ret = crisv32_pinmux_alloc(PORT_D, 5, 7, pinmux_fixed); + ret |= crisv32_pinmux_alloc(PORT_D, 15, 17, pinmux_fixed); + hwprot.ata0 = regk_pinmux_yes; + break; + case pinmux_ata1: + ret = crisv32_pinmux_alloc(PORT_D, 0, 4, pinmux_fixed); + ret |= crisv32_pinmux_alloc(PORT_E, 17, 17, pinmux_fixed); + hwprot.ata1 = regk_pinmux_yes; + break; + case pinmux_ata2: + ret = crisv32_pinmux_alloc(PORT_C, 11, 15, pinmux_fixed); + ret |= crisv32_pinmux_alloc(PORT_E, 3, 3, pinmux_fixed); + hwprot.ata2 = regk_pinmux_yes; + break; + case pinmux_ata3: + ret = crisv32_pinmux_alloc(PORT_C, 8, 10, pinmux_fixed); + ret |= crisv32_pinmux_alloc(PORT_C, 0, 2, pinmux_fixed); + hwprot.ata2 = regk_pinmux_yes; + break; + case pinmux_ata: + ret = crisv32_pinmux_alloc(PORT_B, 0, 15, pinmux_fixed); + ret |= crisv32_pinmux_alloc(PORT_D, 8, 15, pinmux_fixed); + hwprot.ata = regk_pinmux_yes; + break; + case pinmux_eth1: + ret = crisv32_pinmux_alloc(PORT_E, 0, 17, pinmux_fixed); + hwprot.eth1 = regk_pinmux_yes; + hwprot.eth1_mgm = regk_pinmux_yes; + break; + case pinmux_timer: + ret = crisv32_pinmux_alloc(PORT_C, 16, 16, pinmux_fixed); + hwprot.timer = regk_pinmux_yes; + spin_unlock_irqrestore(&pinmux_lock, flags); + return ret; + } + + if (!ret) + REG_WR(pinmux, regi_pinmux, rw_hwprot, hwprot); + else + memcpy(pins, saved, sizeof pins); + + spin_unlock_irqrestore(&pinmux_lock, flags); + + return ret; +} + +void +crisv32_pinmux_set(int port) +{ + int i; + int gpio_val = 0; + int iop_val = 0; + + for (i = 0; i < PORT_PINS; i++) + { + if (pins[port][i] == pinmux_gpio) + gpio_val |= (1 << i); + else if (pins[port][i] == pinmux_iop) + iop_val |= (1 << i); + } + + REG_WRITE(int, regi_pinmux + REG_RD_ADDR_pinmux_rw_pb_gio + 8*port, gpio_val); + REG_WRITE(int, regi_pinmux + REG_RD_ADDR_pinmux_rw_pb_iop + 8*port, iop_val); + +#ifdef DEBUG + crisv32_pinmux_dump(); +#endif +} + +int +crisv32_pinmux_dealloc(int port, int first_pin, int last_pin) +{ + int i; + unsigned long flags; + + crisv32_pinmux_init(); + + if (port > PORTS) + return -EINVAL; + + spin_lock_irqsave(&pinmux_lock, flags); + + for (i = first_pin; i <= last_pin; i++) + pins[port][i] = pinmux_none; + + crisv32_pinmux_set(port); + spin_unlock_irqrestore(&pinmux_lock, flags); + + return 0; +} + +void +crisv32_pinmux_dump(void) +{ + int i, j; + + crisv32_pinmux_init(); + + for (i = 0; i < PORTS; i++) + { + printk("Port %c\n", 'B'+i); + for (j = 0; j < PORT_PINS; j++) + printk(" Pin %d = %d\n", j, pins[i][j]); + } +} + +__initcall(crisv32_pinmux_init); diff --git a/arch/cris/arch-v32/kernel/process.c b/arch/cris/arch-v32/kernel/process.c new file mode 100644 index 000000000000..882be42114f7 --- /dev/null +++ b/arch/cris/arch-v32/kernel/process.c @@ -0,0 +1,270 @@ +/* + * Copyright (C) 2000-2003 Axis Communications AB + * + * Authors: Bjorn Wesen (bjornw@axis.com) + * Mikael Starvik (starvik@axis.com) + * Tobias Anderberg (tobiasa@axis.com), CRISv32 port. + * + * This file handles the architecture-dependent parts of process handling.. + */ + +#include <linux/config.h> +#include <linux/sched.h> +#include <linux/err.h> +#include <linux/fs.h> +#include <linux/slab.h> +#include <asm/arch/hwregs/reg_rdwr.h> +#include <asm/arch/hwregs/reg_map.h> +#include <asm/arch/hwregs/timer_defs.h> +#include <asm/arch/hwregs/intr_vect_defs.h> + +extern void stop_watchdog(void); + +#ifdef CONFIG_ETRAX_GPIO +extern void etrax_gpio_wake_up_check(void); /* Defined in drivers/gpio.c. */ +#endif + +extern int cris_hlt_counter; + +/* We use this if we don't have any better idle routine. */ +void default_idle(void) +{ + local_irq_disable(); + if (!need_resched() && !cris_hlt_counter) { + /* Halt until exception. */ + __asm__ volatile("ei \n\t" + "halt "); + } + local_irq_enable(); +} + +/* + * Free current thread data structures etc.. + */ + +extern void deconfigure_bp(long pid); +void exit_thread(void) +{ + deconfigure_bp(current->pid); +} + +/* + * If the watchdog is enabled, disable interrupts and enter an infinite loop. + * The watchdog will reset the CPU after 0.1s. If the watchdog isn't enabled + * then enable it and wait. + */ +extern void arch_enable_nmi(void); + +void +hard_reset_now(void) +{ + /* + * Don't declare this variable elsewhere. We don't want any other + * code to know about it than the watchdog handler in entry.S and + * this code, implementing hard reset through the watchdog. + */ +#if defined(CONFIG_ETRAX_WATCHDOG) + extern int cause_of_death; +#endif + + printk("*** HARD RESET ***\n"); + local_irq_disable(); + +#if defined(CONFIG_ETRAX_WATCHDOG) + cause_of_death = 0xbedead; +#else +{ + reg_timer_rw_wd_ctrl wd_ctrl = {0}; + + stop_watchdog(); + + wd_ctrl.key = 16; /* Arbitrary key. */ + wd_ctrl.cnt = 1; /* Minimum time. */ + wd_ctrl.cmd = regk_timer_start; + + arch_enable_nmi(); + REG_WR(timer, regi_timer, rw_wd_ctrl, wd_ctrl); +} +#endif + + while (1) + ; /* Wait for reset. */ +} + +/* + * Return saved PC of a blocked thread. + */ +unsigned long thread_saved_pc(struct task_struct *t) +{ + return (unsigned long)user_regs(t->thread_info)->erp; +} + +static void +kernel_thread_helper(void* dummy, int (*fn)(void *), void * arg) +{ + fn(arg); + do_exit(-1); /* Should never be called, return bad exit value. */ +} + +/* Create a kernel thread. */ +int +kernel_thread(int (*fn)(void *), void * arg, unsigned long flags) +{ + struct pt_regs regs; + + memset(®s, 0, sizeof(regs)); + + /* Don't use r10 since that is set to 0 in copy_thread. */ + regs.r11 = (unsigned long) fn; + regs.r12 = (unsigned long) arg; + regs.erp = (unsigned long) kernel_thread_helper; + regs.ccs = 1 << (I_CCS_BITNR + CCS_SHIFT); + + /* Create the new process. */ + return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, ®s, 0, NULL, NULL); +} + +/* + * Setup the child's kernel stack with a pt_regs and call switch_stack() on it. + * It will be unnested during _resume and _ret_from_sys_call when the new thread + * is scheduled. + * + * Also setup the thread switching structure which is used to keep + * thread-specific data during _resumes. + */ + +extern asmlinkage void ret_from_fork(void); + +int +copy_thread(int nr, unsigned long clone_flags, unsigned long usp, + unsigned long unused, + struct task_struct *p, struct pt_regs *regs) +{ + struct pt_regs *childregs; + struct switch_stack *swstack; + + /* + * Put the pt_regs structure at the end of the new kernel stack page and + * fix it up. Note: the task_struct doubles as the kernel stack for the + * task. + */ + childregs = user_regs(p->thread_info); + *childregs = *regs; /* Struct copy of pt_regs. */ + p->set_child_tid = p->clear_child_tid = NULL; + childregs->r10 = 0; /* Child returns 0 after a fork/clone. */ + + /* Set a new TLS ? + * The TLS is in $mof beacuse it is the 5th argument to sys_clone. + */ + if (p->mm && (clone_flags & CLONE_SETTLS)) { + p->thread_info->tls = regs->mof; + } + + /* Put the switch stack right below the pt_regs. */ + swstack = ((struct switch_stack *) childregs) - 1; + + /* Paramater to ret_from_sys_call. 0 is don't restart the syscall. */ + swstack->r9 = 0; + + /* + * We want to return into ret_from_sys_call after the _resume. + * ret_from_fork will call ret_from_sys_call. + */ + swstack->return_ip = (unsigned long) ret_from_fork; + + /* Fix the user-mode and kernel-mode stackpointer. */ + p->thread.usp = usp; + p->thread.ksp = (unsigned long) swstack; + + return 0; +} + +/* + * Be aware of the "magic" 7th argument in the four system-calls below. + * They need the latest stackframe, which is put as the 7th argument by + * entry.S. The previous arguments are dummies or actually used, but need + * to be defined to reach the 7th argument. + * + * N.B.: Another method to get the stackframe is to use current_regs(). But + * it returns the latest stack-frame stacked when going from _user mode_ and + * some of these (at least sys_clone) are called from kernel-mode sometimes + * (for example during kernel_thread, above) and thus cannot use it. Thus, + * to be sure not to get any surprises, we use the method for the other calls + * as well. + */ +asmlinkage int +sys_fork(long r10, long r11, long r12, long r13, long mof, long srp, + struct pt_regs *regs) +{ + return do_fork(SIGCHLD, rdusp(), regs, 0, NULL, NULL); +} + +/* FIXME: Is parent_tid/child_tid really third/fourth argument? Update lib? */ +asmlinkage int +sys_clone(unsigned long newusp, unsigned long flags, int *parent_tid, int *child_tid, + unsigned long tls, long srp, struct pt_regs *regs) +{ + if (!newusp) + newusp = rdusp(); + + return do_fork(flags, newusp, regs, 0, parent_tid, child_tid); +} + +/* + * vfork is a system call in i386 because of register-pressure - maybe + * we can remove it and handle it in libc but we put it here until then. + */ +asmlinkage int +sys_vfork(long r10, long r11, long r12, long r13, long mof, long srp, + struct pt_regs *regs) +{ + return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, rdusp(), regs, 0, NULL, NULL); +} + +/* sys_execve() executes a new program. */ +asmlinkage int +sys_execve(const char *fname, char **argv, char **envp, long r13, long mof, long srp, + struct pt_regs *regs) +{ + int error; + char *filename; + + filename = getname(fname); + error = PTR_ERR(filename); + + if (IS_ERR(filename)) + goto out; + + error = do_execve(filename, argv, envp, regs); + putname(filename); + out: + return error; +} + +unsigned long +get_wchan(struct task_struct *p) +{ + /* TODO */ + return 0; +} +#undef last_sched +#undef first_sched + +void show_regs(struct pt_regs * regs) +{ + unsigned long usp = rdusp(); + printk("ERP: %08lx SRP: %08lx CCS: %08lx USP: %08lx MOF: %08lx\n", + regs->erp, regs->srp, regs->ccs, usp, regs->mof); + + printk(" r0: %08lx r1: %08lx r2: %08lx r3: %08lx\n", + regs->r0, regs->r1, regs->r2, regs->r3); + + printk(" r4: %08lx r5: %08lx r6: %08lx r7: %08lx\n", + regs->r4, regs->r5, regs->r6, regs->r7); + + printk(" r8: %08lx r9: %08lx r10: %08lx r11: %08lx\n", + regs->r8, regs->r9, regs->r10, regs->r11); + + printk("r12: %08lx r13: %08lx oR10: %08lx\n", + regs->r12, regs->r13, regs->orig_r10); +} diff --git a/arch/cris/arch-v32/kernel/ptrace.c b/arch/cris/arch-v32/kernel/ptrace.c new file mode 100644 index 000000000000..208489da2a87 --- /dev/null +++ b/arch/cris/arch-v32/kernel/ptrace.c @@ -0,0 +1,597 @@ +/* + * Copyright (C) 2000-2003, Axis Communications AB. + */ + +#include <linux/kernel.h> +#include <linux/sched.h> +#include <linux/mm.h> +#include <linux/smp.h> +#include <linux/smp_lock.h> +#include <linux/errno.h> +#include <linux/ptrace.h> +#include <linux/user.h> +#include <linux/signal.h> +#include <linux/security.h> + +#include <asm/uaccess.h> +#include <asm/page.h> +#include <asm/pgtable.h> +#include <asm/system.h> +#include <asm/processor.h> +#include <asm/arch/hwregs/supp_reg.h> + +/* + * Determines which bits in CCS the user has access to. + * 1 = access, 0 = no access. + */ +#define CCS_MASK 0x00087c00 /* SXNZVC */ + +#define SBIT_USER (1 << (S_CCS_BITNR + CCS_SHIFT)) + +static int put_debugreg(long pid, unsigned int regno, long data); +static long get_debugreg(long pid, unsigned int regno); +static unsigned long get_pseudo_pc(struct task_struct *child); +void deconfigure_bp(long pid); + +extern unsigned long cris_signal_return_page; + +/* + * Get contents of register REGNO in task TASK. + */ +long get_reg(struct task_struct *task, unsigned int regno) +{ + /* USP is a special case, it's not in the pt_regs struct but + * in the tasks thread struct + */ + unsigned long ret; + + if (regno <= PT_EDA) + ret = ((unsigned long *)user_regs(task->thread_info))[regno]; + else if (regno == PT_USP) + ret = task->thread.usp; + else if (regno == PT_PPC) + ret = get_pseudo_pc(task); + else if (regno <= PT_MAX) + ret = get_debugreg(task->pid, regno); + else + ret = 0; + + return ret; +} + +/* + * Write contents of register REGNO in task TASK. + */ +int put_reg(struct task_struct *task, unsigned int regno, unsigned long data) +{ + if (regno <= PT_EDA) + ((unsigned long *)user_regs(task->thread_info))[regno] = data; + else if (regno == PT_USP) + task->thread.usp = data; + else if (regno == PT_PPC) { + /* Write pseudo-PC to ERP only if changed. */ + if (data != get_pseudo_pc(task)) + ((unsigned long *)user_regs(task->thread_info))[PT_ERP] = data; + } else if (regno <= PT_MAX) + return put_debugreg(task->pid, regno, data); + else + return -1; + return 0; +} + +/* + * Called by kernel/ptrace.c when detaching. + * + * Make sure the single step bit is not set. + */ +void +ptrace_disable(struct task_struct *child) +{ + unsigned long tmp; + + /* Deconfigure SPC and S-bit. */ + tmp = get_reg(child, PT_CCS) & ~SBIT_USER; + put_reg(child, PT_CCS, tmp); + put_reg(child, PT_SPC, 0); + + /* Deconfigure any watchpoints associated with the child. */ + deconfigure_bp(child->pid); +} + + +asmlinkage int +sys_ptrace(long request, long pid, long addr, long data) +{ + struct task_struct *child; + int ret; + unsigned long __user *datap = (unsigned long __user *)data; + + lock_kernel(); + ret = -EPERM; + + if (request == PTRACE_TRACEME) { + /* are we already being traced? */ + if (current->ptrace & PT_PTRACED) + goto out; + ret = security_ptrace(current->parent, current); + if (ret) + goto out; + /* set the ptrace bit in the process flags. */ + current->ptrace |= PT_PTRACED; + ret = 0; + goto out; + } + + ret = -ESRCH; + read_lock(&tasklist_lock); + child = find_task_by_pid(pid); + + if (child) + get_task_struct(child); + + read_unlock(&tasklist_lock); + + if (!child) + goto out; + + ret = -EPERM; + + if (pid == 1) /* Leave the init process alone! */ + goto out_tsk; + + if (request == PTRACE_ATTACH) { + ret = ptrace_attach(child); + goto out_tsk; + } + + ret = ptrace_check_attach(child, request == PTRACE_KILL); + if (ret < 0) + goto out_tsk; + + switch (request) { + /* Read word at location address. */ + case PTRACE_PEEKTEXT: + case PTRACE_PEEKDATA: { + unsigned long tmp; + int copied; + + ret = -EIO; + + /* The signal trampoline page is outside the normal user-addressable + * space but still accessible. This is hack to make it possible to + * access the signal handler code in GDB. + */ + if ((addr & PAGE_MASK) == cris_signal_return_page) { + /* The trampoline page is globally mapped, no page table to traverse.*/ + tmp = *(unsigned long*)addr; + } else { + copied = access_process_vm(child, addr, &tmp, sizeof(tmp), 0); + + if (copied != sizeof(tmp)) + break; + } + + ret = put_user(tmp,datap); + break; + } + + /* Read the word at location address in the USER area. */ + case PTRACE_PEEKUSR: { + unsigned long tmp; + + ret = -EIO; + if ((addr & 3) || addr < 0 || addr > PT_MAX << 2) + break; + + tmp = get_reg(child, addr >> 2); + ret = put_user(tmp, datap); + break; + } + + /* Write the word at location address. */ + case PTRACE_POKETEXT: + case PTRACE_POKEDATA: + ret = 0; + + if (access_process_vm(child, addr, &data, sizeof(data), 1) == sizeof(data)) + break; + + ret = -EIO; + break; + + /* Write the word at location address in the USER area. */ + case PTRACE_POKEUSR: + ret = -EIO; + if ((addr & 3) || addr < 0 || addr > PT_MAX << 2) + break; + + addr >>= 2; + + if (addr == PT_CCS) { + /* don't allow the tracing process to change stuff like + * interrupt enable, kernel/user bit, dma enables etc. + */ + data &= CCS_MASK; + data |= get_reg(child, PT_CCS) & ~CCS_MASK; + } + if (put_reg(child, addr, data)) + break; + ret = 0; + break; + + case PTRACE_SYSCALL: + case PTRACE_CONT: + ret = -EIO; + + if (!valid_signal(data)) + break; + + /* Continue means no single-step. */ + put_reg(child, PT_SPC, 0); + + if (!get_debugreg(child->pid, PT_BP_CTRL)) { + unsigned long tmp; + /* If no h/w bp configured, disable S bit. */ + tmp = get_reg(child, PT_CCS) & ~SBIT_USER; + put_reg(child, PT_CCS, tmp); + } + + if (request == PTRACE_SYSCALL) { + set_tsk_thread_flag(child, TIF_SYSCALL_TRACE); + } + else { + clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE); + } + + child->exit_code = data; + + /* TODO: make sure any pending breakpoint is killed */ + wake_up_process(child); + ret = 0; + + break; + + /* Make the child exit by sending it a sigkill. */ + case PTRACE_KILL: + ret = 0; + + if (child->exit_state == EXIT_ZOMBIE) + break; + + child->exit_code = SIGKILL; + + /* Deconfigure single-step and h/w bp. */ + ptrace_disable(child); + + /* TODO: make sure any pending breakpoint is killed */ + wake_up_process(child); + break; + + /* Set the trap flag. */ + case PTRACE_SINGLESTEP: { + unsigned long tmp; + ret = -EIO; + + /* Set up SPC if not set already (in which case we have + no other choice but to trust it). */ + if (!get_reg(child, PT_SPC)) { + /* In case we're stopped in a delay slot. */ + tmp = get_reg(child, PT_ERP) & ~1; + put_reg(child, PT_SPC, tmp); + } + tmp = get_reg(child, PT_CCS) | SBIT_USER; + put_reg(child, PT_CCS, tmp); + + if (!valid_signal(data)) + break; + + clear_tsk_thread_flag(child, TIF_SYSCALL_TRACE); + + /* TODO: set some clever breakpoint mechanism... */ + + child->exit_code = data; + wake_up_process(child); + ret = 0; + break; + + } + case PTRACE_DETACH: + ret = ptrace_detach(child, data); + break; + + /* Get all GP registers from the child. */ + case PTRACE_GETREGS: { + int i; + unsigned long tmp; + + for (i = 0; i <= PT_MAX; i++) { + tmp = get_reg(child, i); + + if (put_user(tmp, datap)) { + ret = -EFAULT; + goto out_tsk; + } + + datap++; + } + + ret = 0; + break; + } + + /* Set all GP registers in the child. */ + case PTRACE_SETREGS: { + int i; + unsigned long tmp; + + for (i = 0; i <= PT_MAX; i++) { + if (get_user(tmp, datap)) { + ret = -EFAULT; + goto out_tsk; + } + + if (i == PT_CCS) { + tmp &= CCS_MASK; + tmp |= get_reg(child, PT_CCS) & ~CCS_MASK; + } + + put_reg(child, i, tmp); + datap++; + } + + ret = 0; + break; + } + + default: + ret = ptrace_request(child, request, addr, data); + break; + } +out_tsk: + put_task_struct(child); +out: + unlock_kernel(); + return ret; +} + +void do_syscall_trace(void) +{ + if (!test_thread_flag(TIF_SYSCALL_TRACE)) + return; + + if (!(current->ptrace & PT_PTRACED)) + return; + + /* the 0x80 provides a way for the tracing parent to distinguish + between a syscall stop and SIGTRAP delivery */ + ptrace_notify(SIGTRAP | ((current->ptrace & PT_TRACESYSGOOD) + ? 0x80 : 0)); + + /* + * This isn't the same as continuing with a signal, but it will do for + * normal use. + */ + if (current->exit_code) { + send_sig(current->exit_code, current, 1); + current->exit_code = 0; + } +} + +/* Returns the size of an instruction that has a delay slot. */ + +static int insn_size(struct task_struct *child, unsigned long pc) +{ + unsigned long opcode; + int copied; + int opsize = 0; + + /* Read the opcode at pc (do what PTRACE_PEEKTEXT would do). */ + copied = access_process_vm(child, pc, &opcode, sizeof(opcode), 0); + if (copied != sizeof(opcode)) + return 0; + + switch ((opcode & 0x0f00) >> 8) { + case 0x0: + case 0x9: + case 0xb: + opsize = 2; + break; + case 0xe: + case 0xf: + opsize = 6; + break; + case 0xd: + /* Could be 4 or 6; check more bits. */ + if ((opcode & 0xff) == 0xff) + opsize = 4; + else + opsize = 6; + break; + default: + panic("ERROR: Couldn't find size of opcode 0x%lx at 0x%lx\n", + opcode, pc); + } + + return opsize; +} + +static unsigned long get_pseudo_pc(struct task_struct *child) +{ + /* Default value for PC is ERP. */ + unsigned long pc = get_reg(child, PT_ERP); + + if (pc & 0x1) { + unsigned long spc = get_reg(child, PT_SPC); + /* Delay slot bit set. Report as stopped on proper + instruction. */ + if (spc) { + /* Rely on SPC if set. FIXME: We might want to check + that EXS indicates we stopped due to a single-step + exception. */ + pc = spc; + } else { + /* Calculate the PC from the size of the instruction + that the delay slot we're in belongs to. */ + pc += insn_size(child, pc & ~1) - 1; + } + } + return pc; +} + +static long bp_owner = 0; + +/* Reachable from exit_thread in signal.c, so not static. */ +void deconfigure_bp(long pid) +{ + int bp; + + /* Only deconfigure if the pid is the owner. */ + if (bp_owner != pid) + return; + + for (bp = 0; bp < 6; bp++) { + unsigned long tmp; + /* Deconfigure start and end address (also gets rid of ownership). */ + put_debugreg(pid, PT_BP + 3 + (bp * 2), 0); + put_debugreg(pid, PT_BP + 4 + (bp * 2), 0); + + /* Deconfigure relevant bits in control register. */ + tmp = get_debugreg(pid, PT_BP_CTRL) & ~(3 << (2 + (bp * 4))); + put_debugreg(pid, PT_BP_CTRL, tmp); + } + /* No owner now. */ + bp_owner = 0; +} + +static int put_debugreg(long pid, unsigned int regno, long data) +{ + int ret = 0; + register int old_srs; + +#ifdef CONFIG_ETRAX_KGDB + /* Ignore write, but pretend it was ok if value is 0 + (we don't want POKEUSR/SETREGS failing unnessecarily). */ + return (data == 0) ? ret : -1; +#endif + + /* Simple owner management. */ + if (!bp_owner) + bp_owner = pid; + else if (bp_owner != pid) { + /* Ignore write, but pretend it was ok if value is 0 + (we don't want POKEUSR/SETREGS failing unnessecarily). */ + return (data == 0) ? ret : -1; + } + + /* Remember old SRS. */ + SPEC_REG_RD(SPEC_REG_SRS, old_srs); + /* Switch to BP bank. */ + SUPP_BANK_SEL(BANK_BP); + + switch (regno - PT_BP) { + case 0: + SUPP_REG_WR(0, data); break; + case 1: + case 2: + if (data) + ret = -1; + break; + case 3: + SUPP_REG_WR(3, data); break; + case 4: + SUPP_REG_WR(4, data); break; + case 5: + SUPP_REG_WR(5, data); break; + case 6: + SUPP_REG_WR(6, data); break; + case 7: + SUPP_REG_WR(7, data); break; + case 8: + SUPP_REG_WR(8, data); break; + case 9: + SUPP_REG_WR(9, data); break; + case 10: + SUPP_REG_WR(10, data); break; + case 11: + SUPP_REG_WR(11, data); break; + case 12: + SUPP_REG_WR(12, data); break; + case 13: + SUPP_REG_WR(13, data); break; + case 14: + SUPP_REG_WR(14, data); break; + default: + ret = -1; + break; + } + + /* Restore SRS. */ + SPEC_REG_WR(SPEC_REG_SRS, old_srs); + /* Just for show. */ + NOP(); + NOP(); + NOP(); + + return ret; +} + +static long get_debugreg(long pid, unsigned int regno) +{ + register int old_srs; + register long data; + + if (pid != bp_owner) { + return 0; + } + + /* Remember old SRS. */ + SPEC_REG_RD(SPEC_REG_SRS, old_srs); + /* Switch to BP bank. */ + SUPP_BANK_SEL(BANK_BP); + + switch (regno - PT_BP) { + case 0: + SUPP_REG_RD(0, data); break; + case 1: + case 2: + /* error return value? */ + data = 0; + break; + case 3: + SUPP_REG_RD(3, data); break; + case 4: + SUPP_REG_RD(4, data); break; + case 5: + SUPP_REG_RD(5, data); break; + case 6: + SUPP_REG_RD(6, data); break; + case 7: + SUPP_REG_RD(7, data); break; + case 8: + SUPP_REG_RD(8, data); break; + case 9: + SUPP_REG_RD(9, data); break; + case 10: + SUPP_REG_RD(10, data); break; + case 11: + SUPP_REG_RD(11, data); break; + case 12: + SUPP_REG_RD(12, data); break; + case 13: + SUPP_REG_RD(13, data); break; + case 14: + SUPP_REG_RD(14, data); break; + default: + /* error return value? */ + data = 0; + } + + /* Restore SRS. */ + SPEC_REG_WR(SPEC_REG_SRS, old_srs); + /* Just for show. */ + NOP(); + NOP(); + NOP(); + + return data; +} diff --git a/arch/cris/arch-v32/kernel/setup.c b/arch/cris/arch-v32/kernel/setup.c new file mode 100644 index 000000000000..b17a39a2e164 --- /dev/null +++ b/arch/cris/arch-v32/kernel/setup.c @@ -0,0 +1,118 @@ +/* + * Display CPU info in /proc/cpuinfo. + * + * Copyright (C) 2003, Axis Communications AB. + */ + +#include <linux/config.h> +#include <linux/seq_file.h> +#include <linux/proc_fs.h> +#include <linux/delay.h> +#include <linux/param.h> + +#ifdef CONFIG_PROC_FS + +#define HAS_FPU 0x0001 +#define HAS_MMU 0x0002 +#define HAS_ETHERNET100 0x0004 +#define HAS_TOKENRING 0x0008 +#define HAS_SCSI 0x0010 +#define HAS_ATA 0x0020 +#define HAS_USB 0x0040 +#define HAS_IRQ_BUG 0x0080 +#define HAS_MMU_BUG 0x0100 + +struct cpu_info { + char *cpu_model; + unsigned short rev; + unsigned short cache_size; + unsigned short flags; +}; + +/* Some of these model are here for historical reasons only. */ +static struct cpu_info cpinfo[] = { + {"ETRAX 1", 0, 0, 0}, + {"ETRAX 2", 1, 0, 0}, + {"ETRAX 3", 2, 0, 0}, + {"ETRAX 4", 3, 0, 0}, + {"Simulator", 7, 8, HAS_ETHERNET100 | HAS_SCSI | HAS_ATA}, + {"ETRAX 100", 8, 8, HAS_ETHERNET100 | HAS_SCSI | HAS_ATA | HAS_IRQ_BUG}, + {"ETRAX 100", 9, 8, HAS_ETHERNET100 | HAS_SCSI | HAS_ATA}, + + {"ETRAX 100LX", 10, 8, HAS_ETHERNET100 | HAS_SCSI | HAS_ATA | HAS_USB + | HAS_MMU | HAS_MMU_BUG}, + + {"ETRAX 100LX v2", 11, 8, HAS_ETHERNET100 | HAS_SCSI | HAS_ATA | HAS_USB + | HAS_MMU}, + + {"ETRAX FS", 32, 32, HAS_ETHERNET100 | HAS_ATA | HAS_MMU}, + + {"Unknown", 0, 0, 0} +}; + +int +show_cpuinfo(struct seq_file *m, void *v) +{ + int i; + int cpu = (int)v - 1; + int entries; + unsigned long revision; + struct cpu_info *info; + + entries = sizeof cpinfo / sizeof(struct cpu_info); + info = &cpinfo[entries - 1]; + +#ifdef CONFIG_SMP + if (!cpu_online(cpu)) + return 0; +#endif + + revision = rdvr(); + + for (i = 0; i < entries; i++) { + if (cpinfo[i].rev == revision) { + info = &cpinfo[i]; + break; + } + } + + return seq_printf(m, + "processor\t: %d\n" + "cpu\t\t: CRIS\n" + "cpu revision\t: %lu\n" + "cpu model\t: %s\n" + "cache size\t: %d KB\n" + "fpu\t\t: %s\n" + "mmu\t\t: %s\n" + "mmu DMA bug\t: %s\n" + "ethernet\t: %s Mbps\n" + "token ring\t: %s\n" + "scsi\t\t: %s\n" + "ata\t\t: %s\n" + "usb\t\t: %s\n" + "bogomips\t: %lu.%02lu\n\n", + + cpu, + revision, + info->cpu_model, + info->cache_size, + info->flags & HAS_FPU ? "yes" : "no", + info->flags & HAS_MMU ? "yes" : "no", + info->flags & HAS_MMU_BUG ? "yes" : "no", + info->flags & HAS_ETHERNET100 ? "10/100" : "10", + info->flags & HAS_TOKENRING ? "4/16 Mbps" : "no", + info->flags & HAS_SCSI ? "yes" : "no", + info->flags & HAS_ATA ? "yes" : "no", + info->flags & HAS_USB ? "yes" : "no", + (loops_per_jiffy * HZ + 500) / 500000, + ((loops_per_jiffy * HZ + 500) / 5000) % 100); +} + +#endif /* CONFIG_PROC_FS */ + +void +show_etrax_copyright(void) +{ + printk(KERN_INFO + "Linux/CRISv32 port on ETRAX FS (C) 2003, 2004 Axis Communications AB\n"); +} diff --git a/arch/cris/arch-v32/kernel/signal.c b/arch/cris/arch-v32/kernel/signal.c new file mode 100644 index 000000000000..fb4c79d5b76b --- /dev/null +++ b/arch/cris/arch-v32/kernel/signal.c @@ -0,0 +1,708 @@ +/* + * Copyright (C) 2003, Axis Communications AB. + */ + +#include <linux/sched.h> +#include <linux/mm.h> +#include <linux/kernel.h> +#include <linux/signal.h> +#include <linux/errno.h> +#include <linux/wait.h> +#include <linux/ptrace.h> +#include <linux/unistd.h> +#include <linux/stddef.h> +#include <linux/syscalls.h> +#include <linux/vmalloc.h> + +#include <asm/io.h> +#include <asm/processor.h> +#include <asm/ucontext.h> +#include <asm/uaccess.h> +#include <asm/arch/ptrace.h> +#include <asm/arch/hwregs/cpu_vect.h> + +extern unsigned long cris_signal_return_page; + +/* Flag to check if a signal is blockable. */ +#define _BLOCKABLE (~(sigmask(SIGKILL) | sigmask(SIGSTOP))) + +/* + * A syscall in CRIS is really a "break 13" instruction, which is 2 + * bytes. The registers is manipulated so upon return the instruction + * will be executed again. + * + * This relies on that PC points to the instruction after the break call. + */ +#define RESTART_CRIS_SYS(regs) regs->r10 = regs->orig_r10; regs->erp -= 2; + +/* Signal frames. */ +struct signal_frame { + struct sigcontext sc; + unsigned long extramask[_NSIG_WORDS - 1]; + unsigned char retcode[8]; /* Trampoline code. */ +}; + +struct rt_signal_frame { + struct siginfo *pinfo; + void *puc; + struct siginfo info; + struct ucontext uc; + unsigned char retcode[8]; /* Trampoline code. */ +}; + +int do_signal(int restart, sigset_t *oldset, struct pt_regs *regs); +void keep_debug_flags(unsigned long oldccs, unsigned long oldspc, + struct pt_regs *regs); +/* + * Swap in the new signal mask, and wait for a signal. Define some + * dummy arguments to be able to reach the regs argument. + */ +int +sys_sigsuspend(old_sigset_t mask, long r11, long r12, long r13, long mof, + long srp, struct pt_regs *regs) +{ + sigset_t saveset; + + mask &= _BLOCKABLE; + + spin_lock_irq(¤t->sighand->siglock); + + saveset = current->blocked; + + siginitset(¤t->blocked, mask); + + recalc_sigpending(); + spin_unlock_irq(¤t->sighand->siglock); + + regs->r10 = -EINTR; + + while (1) { + current->state = TASK_INTERRUPTIBLE; + schedule(); + + if (do_signal(0, &saveset, regs)) { + /* + * This point is reached twice: once to call + * the signal handler, then again to return + * from the sigsuspend system call. When + * calling the signal handler, R10 hold the + * signal number as set by do_signal(). The + * sigsuspend call will always return with + * the restored value above; -EINTR. + */ + return regs->r10; + } + } +} + +/* Define some dummy arguments to be able to reach the regs argument. */ +int +sys_rt_sigsuspend(sigset_t *unewset, size_t sigsetsize, long r12, long r13, + long mof, long srp, struct pt_regs *regs) +{ + sigset_t saveset; + sigset_t newset; + + if (sigsetsize != sizeof(sigset_t)) + return -EINVAL; + + if (copy_from_user(&newset, unewset, sizeof(newset))) + return -EFAULT; + + sigdelsetmask(&newset, ~_BLOCKABLE); + spin_lock_irq(¤t->sighand->siglock); + + saveset = current->blocked; + current->blocked = newset; + + recalc_sigpending(); + spin_unlock_irq(¤t->sighand->siglock); + + regs->r10 = -EINTR; + + while (1) { + current->state = TASK_INTERRUPTIBLE; + schedule(); + + if (do_signal(0, &saveset, regs)) { + /* See comment in function above. */ + return regs->r10; + } + } +} + +int +sys_sigaction(int signal, const struct old_sigaction *act, + struct old_sigaction *oact) +{ + int retval; + struct k_sigaction newk; + struct k_sigaction oldk; + + if (act) { + old_sigset_t mask; + + if (!access_ok(VERIFY_READ, act, sizeof(*act)) || + __get_user(newk.sa.sa_handler, &act->sa_handler) || + __get_user(newk.sa.sa_restorer, &act->sa_restorer)) + return -EFAULT; + + __get_user(newk.sa.sa_flags, &act->sa_flags); + __get_user(mask, &act->sa_mask); + siginitset(&newk.sa.sa_mask, mask); + } + + retval = do_sigaction(signal, act ? &newk : NULL, oact ? &oldk : NULL); + + if (!retval && oact) { + if (!access_ok(VERIFY_WRITE, oact, sizeof(*oact)) || + __put_user(oldk.sa.sa_handler, &oact->sa_handler) || + __put_user(oldk.sa.sa_restorer, &oact->sa_restorer)) + return -EFAULT; + + __put_user(oldk.sa.sa_flags, &oact->sa_flags); + __put_user(oldk.sa.sa_mask.sig[0], &oact->sa_mask); + } + + return retval; +} + +int +sys_sigaltstack(const stack_t __user *uss, stack_t __user *uoss) +{ + return do_sigaltstack(uss, uoss, rdusp()); +} + +static int +restore_sigcontext(struct pt_regs *regs, struct sigcontext __user *sc) +{ + unsigned int err = 0; + unsigned long old_usp; + + /* Always make any pending restarted system calls return -EINTR */ + current_thread_info()->restart_block.fn = do_no_restart_syscall; + + /* + * Restore the registers from &sc->regs. sc is already checked + * for VERIFY_READ since the signal_frame was previously + * checked in sys_sigreturn(). + */ + if (__copy_from_user(regs, sc, sizeof(struct pt_regs))) + goto badframe; + + /* Make that the user-mode flag is set. */ + regs->ccs |= (1 << (U_CCS_BITNR + CCS_SHIFT)); + + /* Restore the old USP. */ + err |= __get_user(old_usp, &sc->usp); + wrusp(old_usp); + + return err; + +badframe: + return 1; +} + +/* Define some dummy arguments to be able to reach the regs argument. */ +asmlinkage int +sys_sigreturn(long r10, long r11, long r12, long r13, long mof, long srp, + struct pt_regs *regs) +{ + sigset_t set; + struct signal_frame __user *frame; + unsigned long oldspc = regs->spc; + unsigned long oldccs = regs->ccs; + + frame = (struct signal_frame *) rdusp(); + + /* + * Since the signal is stacked on a dword boundary, the frame + * should be dword aligned here as well. It it's not, then the + * user is trying some funny business. + */ + if (((long)frame) & 3) + goto badframe; + + if (!access_ok(VERIFY_READ, frame, sizeof(*frame))) + goto badframe; + + if (__get_user(set.sig[0], &frame->sc.oldmask) || + (_NSIG_WORDS > 1 && __copy_from_user(&set.sig[1], + frame->extramask, + sizeof(frame->extramask)))) + goto badframe; + + sigdelsetmask(&set, ~_BLOCKABLE); + spin_lock_irq(¤t->sighand->siglock); + + current->blocked = set; + + recalc_sigpending(); + spin_unlock_irq(¤t->sighand->siglock); + + if (restore_sigcontext(regs, &frame->sc)) + goto badframe; + + keep_debug_flags(oldccs, oldspc, regs); + + return regs->r10; + +badframe: + force_sig(SIGSEGV, current); + return 0; +} + +/* Define some dummy variables to be able to reach the regs argument. */ +asmlinkage int +sys_rt_sigreturn(long r10, long r11, long r12, long r13, long mof, long srp, + struct pt_regs *regs) +{ + sigset_t set; + struct rt_signal_frame __user *frame; + unsigned long oldspc = regs->spc; + unsigned long oldccs = regs->ccs; + + frame = (struct rt_signal_frame *) rdusp(); + + /* + * Since the signal is stacked on a dword boundary, the frame + * should be dword aligned here as well. It it's not, then the + * user is trying some funny business. + */ + if (((long)frame) & 3) + goto badframe; + + if (!access_ok(VERIFY_READ, frame, sizeof(*frame))) + goto badframe; + + if (__copy_from_user(&set, &frame->uc.uc_sigmask, sizeof(set))) + goto badframe; + + sigdelsetmask(&set, ~_BLOCKABLE); + spin_lock_irq(¤t->sighand->siglock); + + current->blocked = set; + + recalc_sigpending(); + spin_unlock_irq(¤t->sighand->siglock); + + if (restore_sigcontext(regs, &frame->uc.uc_mcontext)) + goto badframe; + + if (do_sigaltstack(&frame->uc.uc_stack, NULL, rdusp()) == -EFAULT) + goto badframe; + + keep_debug_flags(oldccs, oldspc, regs); + + return regs->r10; + +badframe: + force_sig(SIGSEGV, current); + return 0; +} + +/* Setup a signal frame. */ +static int +setup_sigcontext(struct sigcontext __user *sc, struct pt_regs *regs, + unsigned long mask) +{ + int err; + unsigned long usp; + + err = 0; + usp = rdusp(); + + /* + * Copy the registers. They are located first in sc, so it's + * possible to use sc directly. + */ + err |= __copy_to_user(sc, regs, sizeof(struct pt_regs)); + + err |= __put_user(mask, &sc->oldmask); + err |= __put_user(usp, &sc->usp); + + return err; +} + +/* Figure out where to put the new signal frame - usually on the stack. */ +static inline void __user * +get_sigframe(struct k_sigaction *ka, struct pt_regs * regs, size_t frame_size) +{ + unsigned long sp; + + sp = rdusp(); + + /* This is the X/Open sanctioned signal stack switching. */ + if (ka->sa.sa_flags & SA_ONSTACK) { + if (!on_sig_stack(sp)) + sp = current->sas_ss_sp + current->sas_ss_size; + } + + /* Make sure the frame is dword-aligned. */ + sp &= ~3; + + return (void __user *)(sp - frame_size); +} + +/* Grab and setup a signal frame. + * + * Basically a lot of state-info is stacked, and arranged for the + * user-mode program to return to the kernel using either a trampiline + * which performs the syscall sigreturn(), or a provided user-mode + * trampoline. + */ +static void +setup_frame(int sig, struct k_sigaction *ka, sigset_t *set, + struct pt_regs * regs) +{ + int err; + unsigned long return_ip; + struct signal_frame __user *frame; + + err = 0; + frame = get_sigframe(ka, regs, sizeof(*frame)); + + if (!access_ok(VERIFY_WRITE, frame, sizeof(*frame))) + goto give_sigsegv; + + err |= setup_sigcontext(&frame->sc, regs, set->sig[0]); + + if (err) + goto give_sigsegv; + + if (_NSIG_WORDS > 1) { + err |= __copy_to_user(frame->extramask, &set->sig[1], + sizeof(frame->extramask)); + } + + if (err) + goto give_sigsegv; + + /* + * Set up to return from user-space. If provided, use a stub + * already located in user-space. + */ + if (ka->sa.sa_flags & SA_RESTORER) { + return_ip = (unsigned long)ka->sa.sa_restorer; + } else { + /* Trampoline - the desired return ip is in the signal return page. */ + return_ip = cris_signal_return_page; + + /* + * This is movu.w __NR_sigreturn, r9; break 13; + * + * WE DO NOT USE IT ANY MORE! It's only left here for historical + * reasons and because gdb uses it as a signature to notice + * signal handler stack frames. + */ + err |= __put_user(0x9c5f, (short __user*)(frame->retcode+0)); + err |= __put_user(__NR_sigreturn, (short __user*)(frame->retcode+2)); + err |= __put_user(0xe93d, (short __user*)(frame->retcode+4)); + } + + if (err) + goto give_sigsegv; + + /* + * Set up registers for signal handler. + * + * Where the code enters now. + * Where the code enter later. + * First argument, signo. + */ + regs->erp = (unsigned long) ka->sa.sa_handler; + regs->srp = return_ip; + regs->r10 = sig; + + /* Actually move the USP to reflect the stacked frame. */ + wrusp((unsigned long)frame); + + return; + +give_sigsegv: + if (sig == SIGSEGV) + ka->sa.sa_handler = SIG_DFL; + + force_sig(SIGSEGV, current); +} + +static void +setup_rt_frame(int sig, struct k_sigaction *ka, siginfo_t *info, + sigset_t *set, struct pt_regs * regs) +{ + int err; + unsigned long return_ip; + struct rt_signal_frame __user *frame; + + err = 0; + frame = get_sigframe(ka, regs, sizeof(*frame)); + + if (!access_ok(VERIFY_WRITE, frame, sizeof(*frame))) + goto give_sigsegv; + + /* TODO: what is the current->exec_domain stuff and invmap ? */ + + err |= __put_user(&frame->info, &frame->pinfo); + err |= __put_user(&frame->uc, &frame->puc); + err |= copy_siginfo_to_user(&frame->info, info); + + if (err) + goto give_sigsegv; + + /* Clear all the bits of the ucontext we don't use. */ + err |= __clear_user(&frame->uc, offsetof(struct ucontext, uc_mcontext)); + err |= setup_sigcontext(&frame->uc.uc_mcontext, regs, set->sig[0]); + err |= __copy_to_user(&frame->uc.uc_sigmask, set, sizeof(*set)); + + if (err) + goto give_sigsegv; + + /* + * Set up to return from user-space. If provided, use a stub + * already located in user-space. + */ + if (ka->sa.sa_flags & SA_RESTORER) { + return_ip = (unsigned long) ka->sa.sa_restorer; + } else { + /* Trampoline - the desired return ip is in the signal return page. */ + return_ip = cris_signal_return_page + 6; + + /* + * This is movu.w __NR_rt_sigreturn, r9; break 13; + * + * WE DO NOT USE IT ANY MORE! It's only left here for historical + * reasons and because gdb uses it as a signature to notice + * signal handler stack frames. + */ + err |= __put_user(0x9c5f, (short __user*)(frame->retcode+0)); + + err |= __put_user(__NR_rt_sigreturn, + (short __user*)(frame->retcode+2)); + + err |= __put_user(0xe93d, (short __user*)(frame->retcode+4)); + } + + if (err) + goto give_sigsegv; + + /* + * Set up registers for signal handler. + * + * Where the code enters now. + * Where the code enters later. + * First argument is signo. + * Second argument is (siginfo_t *). + * Third argument is unused. + */ + regs->erp = (unsigned long) ka->sa.sa_handler; + regs->srp = return_ip; + regs->r10 = sig; + regs->r11 = (unsigned long) &frame->info; + regs->r12 = 0; + + /* Actually move the usp to reflect the stacked frame. */ + wrusp((unsigned long)frame); + + return; + +give_sigsegv: + if (sig == SIGSEGV) + ka->sa.sa_handler = SIG_DFL; + + force_sig(SIGSEGV, current); +} + +/* Invoke a singal handler to, well, handle the signal. */ +extern inline void +handle_signal(int canrestart, unsigned long sig, + siginfo_t *info, struct k_sigaction *ka, + sigset_t *oldset, struct pt_regs * regs) +{ + /* Check if this got called from a system call. */ + if (canrestart) { + /* If so, check system call restarting. */ + switch (regs->r10) { + case -ERESTART_RESTARTBLOCK: + case -ERESTARTNOHAND: + /* + * This means that the syscall should + * only be restarted if there was no + * handler for the signal, and since + * this point isn't reached unless + * there is a handler, there's no need + * to restart. + */ + regs->r10 = -EINTR; + break; + + case -ERESTARTSYS: + /* + * This means restart the syscall if + * there is no handler, or the handler + * was registered with SA_RESTART. + */ + if (!(ka->sa.sa_flags & SA_RESTART)) { + regs->r10 = -EINTR; + break; + } + + /* Fall through. */ + + case -ERESTARTNOINTR: + /* + * This means that the syscall should + * be called again after the signal + * handler returns. + */ + RESTART_CRIS_SYS(regs); + break; + } + } + + /* Set up the stack frame. */ + if (ka->sa.sa_flags & SA_SIGINFO) + setup_rt_frame(sig, ka, info, oldset, regs); + else + setup_frame(sig, ka, oldset, regs); + + if (ka->sa.sa_flags & SA_ONESHOT) + ka->sa.sa_handler = SIG_DFL; + + if (!(ka->sa.sa_flags & SA_NODEFER)) { + spin_lock_irq(¤t->sighand->siglock); + sigorsets(¤t->blocked,¤t->blocked,&ka->sa.sa_mask); + sigaddset(¤t->blocked,sig); + recalc_sigpending(); + spin_unlock_irq(¤t->sighand->siglock); + } +} + +/* + * Note that 'init' is a special process: it doesn't get signals it doesn't + * want to handle. Thus you cannot kill init even with a SIGKILL even by + * mistake. + * + * Also note that the regs structure given here as an argument, is the latest + * pushed pt_regs. It may or may not be the same as the first pushed registers + * when the initial usermode->kernelmode transition took place. Therefore + * we can use user_mode(regs) to see if we came directly from kernel or user + * mode below. + */ +int +do_signal(int canrestart, sigset_t *oldset, struct pt_regs *regs) +{ + int signr; + siginfo_t info; + struct k_sigaction ka; + + /* + * The common case should go fast, which is why this point is + * reached from kernel-mode. If that's the case, just return + * without doing anything. + */ + if (!user_mode(regs)) + return 1; + + if (!oldset) + oldset = ¤t->blocked; + + signr = get_signal_to_deliver(&info, &ka, regs, NULL); + + if (signr > 0) { + /* Deliver the signal. */ + handle_signal(canrestart, signr, &info, &ka, oldset, regs); + return 1; + } + + /* Got here from a system call? */ + if (canrestart) { + /* Restart the system call - no handlers present. */ + if (regs->r10 == -ERESTARTNOHAND || + regs->r10 == -ERESTARTSYS || + regs->r10 == -ERESTARTNOINTR) { + RESTART_CRIS_SYS(regs); + } + + if (regs->r10 == -ERESTART_RESTARTBLOCK){ + regs->r10 = __NR_restart_syscall; + regs->erp -= 2; + } + } + + return 0; +} + +asmlinkage void +ugdb_trap_user(struct thread_info *ti, int sig) +{ + if (((user_regs(ti)->exs & 0xff00) >> 8) != SINGLE_STEP_INTR_VECT) { + /* Zero single-step PC if the reason we stopped wasn't a single + step exception. This is to avoid relying on it when it isn't + reliable. */ + user_regs(ti)->spc = 0; + } + /* FIXME: Filter out false h/w breakpoint hits (i.e. EDA + not withing any configured h/w breakpoint range). Synchronize with + what already exists for kernel debugging. */ + if (((user_regs(ti)->exs & 0xff00) >> 8) == BREAK_8_INTR_VECT) { + /* Break 8: subtract 2 from ERP unless in a delay slot. */ + if (!(user_regs(ti)->erp & 0x1)) + user_regs(ti)->erp -= 2; + } + sys_kill(ti->task->pid, sig); +} + +void +keep_debug_flags(unsigned long oldccs, unsigned long oldspc, + struct pt_regs *regs) +{ + if (oldccs & (1 << Q_CCS_BITNR)) { + /* Pending single step due to single-stepping the break 13 + in the signal trampoline: keep the Q flag. */ + regs->ccs |= (1 << Q_CCS_BITNR); + /* S flag should be set - complain if it's not. */ + if (!(oldccs & (1 << (S_CCS_BITNR + CCS_SHIFT)))) { + printk("Q flag but no S flag?"); + } + regs->ccs |= (1 << (S_CCS_BITNR + CCS_SHIFT)); + /* Assume the SPC is valid and interesting. */ + regs->spc = oldspc; + + } else if (oldccs & (1 << (S_CCS_BITNR + CCS_SHIFT))) { + /* If a h/w bp was set in the signal handler we need + to keep the S flag. */ + regs->ccs |= (1 << (S_CCS_BITNR + CCS_SHIFT)); + /* Don't keep the old SPC though; if we got here due to + a single-step, the Q flag should have been set. */ + } else if (regs->spc) { + /* If we were single-stepping *before* the signal was taken, + we don't want to restore that state now, because GDB will + have forgotten all about it. */ + regs->spc = 0; + regs->ccs &= ~(1 << (S_CCS_BITNR + CCS_SHIFT)); + } +} + +/* Set up the trampolines on the signal return page. */ +int __init +cris_init_signal(void) +{ + u16* data = (u16*)kmalloc(PAGE_SIZE, GFP_KERNEL); + + /* This is movu.w __NR_sigreturn, r9; break 13; */ + data[0] = 0x9c5f; + data[1] = __NR_sigreturn; + data[2] = 0xe93d; + /* This is movu.w __NR_rt_sigreturn, r9; break 13; */ + data[3] = 0x9c5f; + data[4] = __NR_rt_sigreturn; + data[5] = 0xe93d; + + /* Map to userspace with appropriate permissions (no write access...) */ + cris_signal_return_page = (unsigned long) + __ioremap_prot(virt_to_phys(data), PAGE_SIZE, PAGE_SIGNAL_TRAMPOLINE); + + return 0; +} + +__initcall(cris_init_signal); diff --git a/arch/cris/arch-v32/kernel/smp.c b/arch/cris/arch-v32/kernel/smp.c new file mode 100644 index 000000000000..2c5cae04a95c --- /dev/null +++ b/arch/cris/arch-v32/kernel/smp.c @@ -0,0 +1,348 @@ +#include <asm/delay.h> +#include <asm/arch/irq.h> +#include <asm/arch/hwregs/intr_vect.h> +#include <asm/arch/hwregs/intr_vect_defs.h> +#include <asm/tlbflush.h> +#include <asm/mmu_context.h> +#include <asm/arch/hwregs/mmu_defs_asm.h> +#include <asm/arch/hwregs/supp_reg.h> +#include <asm/atomic.h> + +#include <linux/err.h> +#include <linux/init.h> +#include <linux/timex.h> +#include <linux/sched.h> +#include <linux/kernel.h> +#include <linux/cpumask.h> +#include <linux/interrupt.h> + +#define IPI_SCHEDULE 1 +#define IPI_CALL 2 +#define IPI_FLUSH_TLB 4 + +#define FLUSH_ALL (void*)0xffffffff + +/* Vector of locks used for various atomic operations */ +spinlock_t cris_atomic_locks[] = { [0 ... LOCK_COUNT - 1] = SPIN_LOCK_UNLOCKED}; + +/* CPU masks */ +cpumask_t cpu_online_map = CPU_MASK_NONE; +cpumask_t phys_cpu_present_map = CPU_MASK_NONE; + +/* Variables used during SMP boot */ +volatile int cpu_now_booting = 0; +volatile struct thread_info *smp_init_current_idle_thread; + +/* Variables used during IPI */ +static DEFINE_SPINLOCK(call_lock); +static DEFINE_SPINLOCK(tlbstate_lock); + +struct call_data_struct { + void (*func) (void *info); + void *info; + int wait; +}; + +static struct call_data_struct * call_data; + +static struct mm_struct* flush_mm; +static struct vm_area_struct* flush_vma; +static unsigned long flush_addr; + +extern int setup_irq(int, struct irqaction *); + +/* Mode registers */ +static unsigned long irq_regs[NR_CPUS] = +{ + regi_irq, + regi_irq2 +}; + +static irqreturn_t crisv32_ipi_interrupt(int irq, void *dev_id, struct pt_regs *regs); +static int send_ipi(int vector, int wait, cpumask_t cpu_mask); +static struct irqaction irq_ipi = { crisv32_ipi_interrupt, SA_INTERRUPT, + CPU_MASK_NONE, "ipi", NULL, NULL}; + +extern void cris_mmu_init(void); +extern void cris_timer_init(void); + +/* SMP initialization */ +void __init smp_prepare_cpus(unsigned int max_cpus) +{ + int i; + + /* From now on we can expect IPIs so set them up */ + setup_irq(IPI_INTR_VECT, &irq_ipi); + + /* Mark all possible CPUs as present */ + for (i = 0; i < max_cpus; i++) + cpu_set(i, phys_cpu_present_map); +} + +void __devinit smp_prepare_boot_cpu(void) +{ + /* PGD pointer has moved after per_cpu initialization so + * update the MMU. + */ + pgd_t **pgd; + pgd = (pgd_t**)&per_cpu(current_pgd, smp_processor_id()); + + SUPP_BANK_SEL(1); + SUPP_REG_WR(RW_MM_TLB_PGD, pgd); + SUPP_BANK_SEL(2); + SUPP_REG_WR(RW_MM_TLB_PGD, pgd); + + cpu_set(0, cpu_online_map); + cpu_set(0, phys_cpu_present_map); +} + +void __init smp_cpus_done(unsigned int max_cpus) +{ +} + +/* Bring one cpu online.*/ +static int __init +smp_boot_one_cpu(int cpuid) +{ + unsigned timeout; + struct task_struct *idle; + + idle = fork_idle(cpuid); + if (IS_ERR(idle)) + panic("SMP: fork failed for CPU:%d", cpuid); + + idle->thread_info->cpu = cpuid; + + /* Information to the CPU that is about to boot */ + smp_init_current_idle_thread = idle->thread_info; + cpu_now_booting = cpuid; + + /* Wait for CPU to come online */ + for (timeout = 0; timeout < 10000; timeout++) { + if(cpu_online(cpuid)) { + cpu_now_booting = 0; + smp_init_current_idle_thread = NULL; + return 0; /* CPU online */ + } + udelay(100); + barrier(); + } + + put_task_struct(idle); + idle = NULL; + + printk(KERN_CRIT "SMP: CPU:%d is stuck.\n", cpuid); + return -1; +} + +/* Secondary CPUs starts uing C here. Here we need to setup CPU + * specific stuff such as the local timer and the MMU. */ +void __init smp_callin(void) +{ + extern void cpu_idle(void); + + int cpu = cpu_now_booting; + reg_intr_vect_rw_mask vect_mask = {0}; + + /* Initialise the idle task for this CPU */ + atomic_inc(&init_mm.mm_count); + current->active_mm = &init_mm; + + /* Set up MMU */ + cris_mmu_init(); + __flush_tlb_all(); + + /* Setup local timer. */ + cris_timer_init(); + + /* Enable IRQ and idle */ + REG_WR(intr_vect, irq_regs[cpu], rw_mask, vect_mask); + unmask_irq(IPI_INTR_VECT); + unmask_irq(TIMER_INTR_VECT); + local_irq_enable(); + + cpu_set(cpu, cpu_online_map); + cpu_idle(); +} + +/* Stop execution on this CPU.*/ +void stop_this_cpu(void* dummy) +{ + local_irq_disable(); + asm volatile("halt"); +} + +/* Other calls */ +void smp_send_stop(void) +{ + smp_call_function(stop_this_cpu, NULL, 1, 0); +} + +int setup_profiling_timer(unsigned int multiplier) +{ + return -EINVAL; +} + + +/* cache_decay_ticks is used by the scheduler to decide if a process + * is "hot" on one CPU. A higher value means a higher penalty to move + * a process to another CPU. Our cache is rather small so we report + * 1 tick. + */ +unsigned long cache_decay_ticks = 1; + +int __devinit __cpu_up(unsigned int cpu) +{ + smp_boot_one_cpu(cpu); + return cpu_online(cpu) ? 0 : -ENOSYS; +} + +void smp_send_reschedule(int cpu) +{ + cpumask_t cpu_mask = CPU_MASK_NONE; + cpu_set(cpu, cpu_mask); + send_ipi(IPI_SCHEDULE, 0, cpu_mask); +} + +/* TLB flushing + * + * Flush needs to be done on the local CPU and on any other CPU that + * may have the same mapping. The mm->cpu_vm_mask is used to keep track + * of which CPUs that a specific process has been executed on. + */ +void flush_tlb_common(struct mm_struct* mm, struct vm_area_struct* vma, unsigned long addr) +{ + unsigned long flags; + cpumask_t cpu_mask; + + spin_lock_irqsave(&tlbstate_lock, flags); + cpu_mask = (mm == FLUSH_ALL ? CPU_MASK_ALL : mm->cpu_vm_mask); + cpu_clear(smp_processor_id(), cpu_mask); + flush_mm = mm; + flush_vma = vma; + flush_addr = addr; + send_ipi(IPI_FLUSH_TLB, 1, cpu_mask); + spin_unlock_irqrestore(&tlbstate_lock, flags); +} + +void flush_tlb_all(void) +{ + __flush_tlb_all(); + flush_tlb_common(FLUSH_ALL, FLUSH_ALL, 0); +} + +void flush_tlb_mm(struct mm_struct *mm) +{ + __flush_tlb_mm(mm); + flush_tlb_common(mm, FLUSH_ALL, 0); + /* No more mappings in other CPUs */ + cpus_clear(mm->cpu_vm_mask); + cpu_set(smp_processor_id(), mm->cpu_vm_mask); +} + +void flush_tlb_page(struct vm_area_struct *vma, + unsigned long addr) +{ + __flush_tlb_page(vma, addr); + flush_tlb_common(vma->vm_mm, vma, addr); +} + +/* Inter processor interrupts + * + * The IPIs are used for: + * * Force a schedule on a CPU + * * FLush TLB on other CPUs + * * Call a function on other CPUs + */ + +int send_ipi(int vector, int wait, cpumask_t cpu_mask) +{ + int i = 0; + reg_intr_vect_rw_ipi ipi = REG_RD(intr_vect, irq_regs[i], rw_ipi); + int ret = 0; + + /* Calculate CPUs to send to. */ + cpus_and(cpu_mask, cpu_mask, cpu_online_map); + + /* Send the IPI. */ + for_each_cpu_mask(i, cpu_mask) + { + ipi.vector |= vector; + REG_WR(intr_vect, irq_regs[i], rw_ipi, ipi); + } + + /* Wait for IPI to finish on other CPUS */ + if (wait) { + for_each_cpu_mask(i, cpu_mask) { + int j; + for (j = 0 ; j < 1000; j++) { + ipi = REG_RD(intr_vect, irq_regs[i], rw_ipi); + if (!ipi.vector) + break; + udelay(100); + } + + /* Timeout? */ + if (ipi.vector) { + printk("SMP call timeout from %d to %d\n", smp_processor_id(), i); + ret = -ETIMEDOUT; + dump_stack(); + } + } + } + return ret; +} + +/* + * You must not call this function with disabled interrupts or from a + * hardware interrupt handler or from a bottom half handler. + */ +int smp_call_function(void (*func)(void *info), void *info, + int nonatomic, int wait) +{ + cpumask_t cpu_mask = CPU_MASK_ALL; + struct call_data_struct data; + int ret; + + cpu_clear(smp_processor_id(), cpu_mask); + + WARN_ON(irqs_disabled()); + + data.func = func; + data.info = info; + data.wait = wait; + + spin_lock(&call_lock); + call_data = &data; + ret = send_ipi(IPI_CALL, wait, cpu_mask); + spin_unlock(&call_lock); + + return ret; +} + +irqreturn_t crisv32_ipi_interrupt(int irq, void *dev_id, struct pt_regs *regs) +{ + void (*func) (void *info) = call_data->func; + void *info = call_data->info; + reg_intr_vect_rw_ipi ipi; + + ipi = REG_RD(intr_vect, irq_regs[smp_processor_id()], rw_ipi); + + if (ipi.vector & IPI_CALL) { + func(info); + } + if (ipi.vector & IPI_FLUSH_TLB) { + if (flush_mm == FLUSH_ALL) + __flush_tlb_all(); + else if (flush_vma == FLUSH_ALL) + __flush_tlb_mm(flush_mm); + else + __flush_tlb_page(flush_vma, flush_addr); + } + + ipi.vector = 0; + REG_WR(intr_vect, irq_regs[smp_processor_id()], rw_ipi, ipi); + + return IRQ_HANDLED; +} + diff --git a/arch/cris/arch-v32/kernel/time.c b/arch/cris/arch-v32/kernel/time.c new file mode 100644 index 000000000000..d48e397f5fa4 --- /dev/null +++ b/arch/cris/arch-v32/kernel/time.c @@ -0,0 +1,341 @@ +/* $Id: time.c,v 1.19 2005/04/29 05:40:09 starvik Exp $ + * + * linux/arch/cris/arch-v32/kernel/time.c + * + * Copyright (C) 2003 Axis Communications AB + * + */ + +#include <linux/config.h> +#include <linux/timex.h> +#include <linux/time.h> +#include <linux/jiffies.h> +#include <linux/interrupt.h> +#include <linux/swap.h> +#include <linux/sched.h> +#include <linux/init.h> +#include <linux/threads.h> +#include <asm/types.h> +#include <asm/signal.h> +#include <asm/io.h> +#include <asm/delay.h> +#include <asm/rtc.h> +#include <asm/irq.h> + +#include <asm/arch/hwregs/reg_map.h> +#include <asm/arch/hwregs/reg_rdwr.h> +#include <asm/arch/hwregs/timer_defs.h> +#include <asm/arch/hwregs/intr_vect_defs.h> + +/* Watchdog defines */ +#define ETRAX_WD_KEY_MASK 0x7F /* key is 7 bit */ +#define ETRAX_WD_HZ 763 /* watchdog counts at 763 Hz */ +#define ETRAX_WD_CNT ((2*ETRAX_WD_HZ)/HZ + 1) /* Number of 763 counts before watchdog bites */ + +unsigned long timer_regs[NR_CPUS] = +{ + regi_timer, +#ifdef CONFIG_SMP + regi_timer2 +#endif +}; + +extern void update_xtime_from_cmos(void); +extern int set_rtc_mmss(unsigned long nowtime); +extern int setup_irq(int, struct irqaction *); +extern int have_rtc; + +unsigned long get_ns_in_jiffie(void) +{ + reg_timer_r_tmr0_data data; + unsigned long ns; + + data = REG_RD(timer, regi_timer, r_tmr0_data); + ns = (TIMER0_DIV - data) * 10; + return ns; +} + +unsigned long do_slow_gettimeoffset(void) +{ + unsigned long count; + unsigned long usec_count = 0; + + static unsigned long count_p = TIMER0_DIV;/* for the first call after boot */ + static unsigned long jiffies_p = 0; + + /* + * cache volatile jiffies temporarily; we have IRQs turned off. + */ + unsigned long jiffies_t; + + /* The timer interrupt comes from Etrax timer 0. In order to get + * better precision, we check the current value. It might have + * underflowed already though. + */ + + count = REG_RD(timer, regi_timer, r_tmr0_data); + jiffies_t = jiffies; + + /* + * avoiding timer inconsistencies (they are rare, but they happen)... + * there are one problem that must be avoided here: + * 1. the timer counter underflows + */ + if( jiffies_t == jiffies_p ) { + if( count > count_p ) { + /* Timer wrapped, use new count and prescale + * increase the time corresponding to one jiffie + */ + usec_count = 1000000/HZ; + } + } else + jiffies_p = jiffies_t; + count_p = count; + /* Convert timer value to usec */ + /* 100 MHz timer, divide by 100 to get usec */ + usec_count += (TIMER0_DIV - count) / 100; + return usec_count; +} + +/* From timer MDS describing the hardware watchdog: + * 4.3.1 Watchdog Operation + * The watchdog timer is an 8-bit timer with a configurable start value. + * Once started the whatchdog counts downwards with a frequency of 763 Hz + * (100/131072 MHz). When the watchdog counts down to 1, it generates an + * NMI (Non Maskable Interrupt), and when it counts down to 0, it resets the + * chip. + */ +/* This gives us 1.3 ms to do something useful when the NMI comes */ + +/* right now, starting the watchdog is the same as resetting it */ +#define start_watchdog reset_watchdog + +#if defined(CONFIG_ETRAX_WATCHDOG) +static short int watchdog_key = 42; /* arbitrary 7 bit number */ +#endif + +/* number of pages to consider "out of memory". it is normal that the memory + * is used though, so put this really low. + */ + +#define WATCHDOG_MIN_FREE_PAGES 8 + +void +reset_watchdog(void) +{ +#if defined(CONFIG_ETRAX_WATCHDOG) + reg_timer_rw_wd_ctrl wd_ctrl = { 0 }; + + /* only keep watchdog happy as long as we have memory left! */ + if(nr_free_pages() > WATCHDOG_MIN_FREE_PAGES) { + /* reset the watchdog with the inverse of the old key */ + watchdog_key ^= ETRAX_WD_KEY_MASK; /* invert key, which is 7 bits */ + wd_ctrl.cnt = ETRAX_WD_CNT; + wd_ctrl.cmd = regk_timer_start; + wd_ctrl.key = watchdog_key; + REG_WR(timer, regi_timer, rw_wd_ctrl, wd_ctrl); + } +#endif +} + +/* stop the watchdog - we still need the correct key */ + +void +stop_watchdog(void) +{ +#if defined(CONFIG_ETRAX_WATCHDOG) + reg_timer_rw_wd_ctrl wd_ctrl = { 0 }; + watchdog_key ^= ETRAX_WD_KEY_MASK; /* invert key, which is 7 bits */ + wd_ctrl.cnt = ETRAX_WD_CNT; + wd_ctrl.cmd = regk_timer_stop; + wd_ctrl.key = watchdog_key; + REG_WR(timer, regi_timer, rw_wd_ctrl, wd_ctrl); +#endif +} + +extern void show_registers(struct pt_regs *regs); + +void +handle_watchdog_bite(struct pt_regs* regs) +{ +#if defined(CONFIG_ETRAX_WATCHDOG) + extern int cause_of_death; + + raw_printk("Watchdog bite\n"); + + /* Check if forced restart or unexpected watchdog */ + if (cause_of_death == 0xbedead) { + while(1); + } + + /* Unexpected watchdog, stop the watchdog and dump registers*/ + stop_watchdog(); + raw_printk("Oops: bitten by watchdog\n"); + show_registers(regs); +#ifndef CONFIG_ETRAX_WATCHDOG_NICE_DOGGY + reset_watchdog(); +#endif + while(1) /* nothing */; +#endif +} + +/* last time the cmos clock got updated */ +static long last_rtc_update = 0; + +/* + * timer_interrupt() needs to keep up the real-time clock, + * as well as call the "do_timer()" routine every clocktick + */ + +//static unsigned short myjiff; /* used by our debug routine print_timestamp */ + +extern void cris_do_profile(struct pt_regs *regs); + +static inline irqreturn_t +timer_interrupt(int irq, void *dev_id, struct pt_regs *regs) +{ + int cpu = smp_processor_id(); + reg_timer_r_masked_intr masked_intr; + reg_timer_rw_ack_intr ack_intr = { 0 }; + + /* Check if the timer interrupt is for us (a tmr0 int) */ + masked_intr = REG_RD(timer, timer_regs[cpu], r_masked_intr); + if (!masked_intr.tmr0) + return IRQ_NONE; + + /* acknowledge the timer irq */ + ack_intr.tmr0 = 1; + REG_WR(timer, timer_regs[cpu], rw_ack_intr, ack_intr); + + /* reset watchdog otherwise it resets us! */ + reset_watchdog(); + + /* Update statistics. */ + update_process_times(user_mode(regs)); + + cris_do_profile(regs); /* Save profiling information */ + + /* The master CPU is responsible for the time keeping. */ + if (cpu != 0) + return IRQ_HANDLED; + + /* call the real timer interrupt handler */ + do_timer(regs); + + /* + * If we have an externally synchronized Linux clock, then update + * CMOS clock accordingly every ~11 minutes. Set_rtc_mmss() has to be + * called as close as possible to 500 ms before the new second starts. + * + * The division here is not time critical since it will run once in + * 11 minutes + */ + if ((time_status & STA_UNSYNC) == 0 && + xtime.tv_sec > last_rtc_update + 660 && + (xtime.tv_nsec / 1000) >= 500000 - (tick_nsec / 1000) / 2 && + (xtime.tv_nsec / 1000) <= 500000 + (tick_nsec / 1000) / 2) { + if (set_rtc_mmss(xtime.tv_sec) == 0) + last_rtc_update = xtime.tv_sec; + else + last_rtc_update = xtime.tv_sec - 600; /* do it again in 60 s */ + } + return IRQ_HANDLED; +} + +/* timer is SA_SHIRQ so drivers can add stuff to the timer irq chain + * it needs to be SA_INTERRUPT to make the jiffies update work properly + */ + +static struct irqaction irq_timer = { timer_interrupt, SA_SHIRQ | SA_INTERRUPT, + CPU_MASK_NONE, "timer", NULL, NULL}; + +void __init +cris_timer_init(void) +{ + int cpu = smp_processor_id(); + reg_timer_rw_tmr0_ctrl tmr0_ctrl = { 0 }; + reg_timer_rw_tmr0_div tmr0_div = TIMER0_DIV; + reg_timer_rw_intr_mask timer_intr_mask; + + /* Setup the etrax timers + * Base frequency is 100MHz, divider 1000000 -> 100 HZ + * We use timer0, so timer1 is free. + * The trig timer is used by the fasttimer API if enabled. + */ + + tmr0_ctrl.op = regk_timer_ld; + tmr0_ctrl.freq = regk_timer_f100; + REG_WR(timer, timer_regs[cpu], rw_tmr0_div, tmr0_div); + REG_WR(timer, timer_regs[cpu], rw_tmr0_ctrl, tmr0_ctrl); /* Load */ + tmr0_ctrl.op = regk_timer_run; + REG_WR(timer, timer_regs[cpu], rw_tmr0_ctrl, tmr0_ctrl); /* Start */ + + /* enable the timer irq */ + timer_intr_mask = REG_RD(timer, timer_regs[cpu], rw_intr_mask); + timer_intr_mask.tmr0 = 1; + REG_WR(timer, timer_regs[cpu], rw_intr_mask, timer_intr_mask); +} + +void __init +time_init(void) +{ + reg_intr_vect_rw_mask intr_mask; + + /* probe for the RTC and read it if it exists + * Before the RTC can be probed the loops_per_usec variable needs + * to be initialized to make usleep work. A better value for + * loops_per_usec is calculated by the kernel later once the + * clock has started. + */ + loops_per_usec = 50; + + if(RTC_INIT() < 0) { + /* no RTC, start at 1980 */ + xtime.tv_sec = 0; + xtime.tv_nsec = 0; + have_rtc = 0; + } else { + /* get the current time */ + have_rtc = 1; + update_xtime_from_cmos(); + } + + /* + * Initialize wall_to_monotonic such that adding it to xtime will yield zero, the + * tv_nsec field must be normalized (i.e., 0 <= nsec < NSEC_PER_SEC). + */ + set_normalized_timespec(&wall_to_monotonic, -xtime.tv_sec, -xtime.tv_nsec); + + /* Start CPU local timer */ + cris_timer_init(); + + /* enable the timer irq in global config */ + intr_mask = REG_RD(intr_vect, regi_irq, rw_mask); + intr_mask.timer = 1; + REG_WR(intr_vect, regi_irq, rw_mask, intr_mask); + + /* now actually register the timer irq handler that calls timer_interrupt() */ + + setup_irq(TIMER_INTR_VECT, &irq_timer); + + /* enable watchdog if we should use one */ + +#if defined(CONFIG_ETRAX_WATCHDOG) + printk("Enabling watchdog...\n"); + start_watchdog(); + + /* If we use the hardware watchdog, we want to trap it as an NMI + and dump registers before it resets us. For this to happen, we + must set the "m" NMI enable flag (which once set, is unset only + when an NMI is taken). + + The same goes for the external NMI, but that doesn't have any + driver or infrastructure support yet. */ + { + unsigned long flags; + local_save_flags(flags); + flags |= (1<<30); /* NMI M flag is at bit 30 */ + local_irq_restore(flags); + } +#endif +} diff --git a/arch/cris/arch-v32/kernel/traps.c b/arch/cris/arch-v32/kernel/traps.c new file mode 100644 index 000000000000..6e3787045560 --- /dev/null +++ b/arch/cris/arch-v32/kernel/traps.c @@ -0,0 +1,160 @@ +/* + * Copyright (C) 2003, Axis Communications AB. + */ + +#include <linux/config.h> +#include <linux/ptrace.h> +#include <asm/uaccess.h> + +#include <asm/arch/hwregs/supp_reg.h> + +extern void reset_watchdog(void); +extern void stop_watchdog(void); + +extern int raw_printk(const char *fmt, ...); + +void +show_registers(struct pt_regs *regs) +{ + /* + * It's possible to use either the USP register or current->thread.usp. + * USP might not correspond to the current proccess for all cases this + * function is called, and current->thread.usp isn't up to date for the + * current proccess. Experience shows that using USP is the way to go. + */ + unsigned long usp; + unsigned long d_mmu_cause; + unsigned long i_mmu_cause; + + usp = rdusp(); + + raw_printk("CPU: %d\n", smp_processor_id()); + + raw_printk("ERP: %08lx SRP: %08lx CCS: %08lx USP: %08lx MOF: %08lx\n", + regs->erp, regs->srp, regs->ccs, usp, regs->mof); + + raw_printk(" r0: %08lx r1: %08lx r2: %08lx r3: %08lx\n", + regs->r0, regs->r1, regs->r2, regs->r3); + + raw_printk(" r4: %08lx r5: %08lx r6: %08lx r7: %08lx\n", + regs->r4, regs->r5, regs->r6, regs->r7); + + raw_printk(" r8: %08lx r9: %08lx r10: %08lx r11: %08lx\n", + regs->r8, regs->r9, regs->r10, regs->r11); + + raw_printk("r12: %08lx r13: %08lx oR10: %08lx acr: %08lx\n", + regs->r12, regs->r13, regs->orig_r10, regs->acr); + + raw_printk("sp: %08lx\n", regs); + + SUPP_BANK_SEL(BANK_IM); + SUPP_REG_RD(RW_MM_CAUSE, i_mmu_cause); + + SUPP_BANK_SEL(BANK_DM); + SUPP_REG_RD(RW_MM_CAUSE, d_mmu_cause); + + raw_printk(" Data MMU Cause: %08lx\n", d_mmu_cause); + raw_printk("Instruction MMU Cause: %08lx\n", i_mmu_cause); + + raw_printk("Process %s (pid: %d, stackpage: %08lx)\n", + current->comm, current->pid, (unsigned long) current); + + /* Show additional info if in kernel-mode. */ + if (!user_mode(regs)) { + int i; + unsigned char c; + + show_stack(NULL, (unsigned long *) usp); + + /* + * If the previous stack-dump wasn't a kernel one, dump the + * kernel stack now. + */ + if (usp != 0) + show_stack(NULL, NULL); + + raw_printk("\nCode: "); + + if (regs->erp < PAGE_OFFSET) + goto bad_value; + + /* + * Quite often the value at regs->erp doesn't point to the + * interesting instruction, which often is the previous + * instruction. So dump at an offset large enough that the + * instruction decoding should be in sync at the interesting + * point, but small enough to fit on a row. The regs->erp + * location is pointed out in a ksymoops-friendly way by + * wrapping the byte for that address in parenthesis. + */ + for (i = -12; i < 12; i++) { + if (__get_user(c, &((unsigned char *) regs->erp)[i])) { +bad_value: + raw_printk(" Bad IP value."); + break; + } + + if (i == 0) + raw_printk("(%02x) ", c); + else + raw_printk("%02x ", c); + } + + raw_printk("\n"); + } +} + +/* + * This gets called from entry.S when the watchdog has bitten. Show something + * similiar to an Oops dump, and if the kernel if configured to be a nice doggy; + * halt instead of reboot. + */ +void +watchdog_bite_hook(struct pt_regs *regs) +{ +#ifdef CONFIG_ETRAX_WATCHDOG_NICE_DOGGY + local_irq_disable(); + stop_watchdog(); + show_registers(regs); + + while (1) + ; /* Do nothing. */ +#else + show_registers(regs); +#endif +} + +/* This is normally the Oops function. */ +void +die_if_kernel(const char *str, struct pt_regs *regs, long err) +{ + if (user_mode(regs)) + return; + +#ifdef CONFIG_ETRAX_WATCHDOG_NICE_DOGGY + /* + * This printout might take too long and could trigger + * the watchdog normally. If NICE_DOGGY is set, simply + * stop the watchdog during the printout. + */ + stop_watchdog(); +#endif + + raw_printk("%s: %04lx\n", str, err & 0xffff); + + show_registers(regs); + +#ifdef CONFIG_ETRAX_WATCHDOG_NICE_DOGGY + reset_watchdog(); +#endif + + do_exit(SIGSEGV); +} + +void arch_enable_nmi(void) +{ + unsigned long flags; + local_save_flags(flags); + flags |= (1<<30); /* NMI M flag is at bit 30 */ + local_irq_restore(flags); +} diff --git a/arch/cris/arch-v32/kernel/vcs_hook.c b/arch/cris/arch-v32/kernel/vcs_hook.c new file mode 100644 index 000000000000..64d71c54c22c --- /dev/null +++ b/arch/cris/arch-v32/kernel/vcs_hook.c @@ -0,0 +1,96 @@ +// $Id: vcs_hook.c,v 1.2 2003/08/12 12:01:06 starvik Exp $ +// +// Call simulator hook. This is the part running in the +// simulated program. +// + +#include "vcs_hook.h" +#include <stdarg.h> +#include <asm/arch-v32/hwregs/reg_map.h> +#include <asm/arch-v32/hwregs/intr_vect_defs.h> + +#define HOOK_TRIG_ADDR 0xb7000000 /* hook cvlog model reg address */ +#define HOOK_MEM_BASE_ADDR 0xa0000000 /* csp4 (shared mem) base addr */ + +#define HOOK_DATA(offset) ((unsigned*) HOOK_MEM_BASE_ADDR)[offset] +#define VHOOK_DATA(offset) ((volatile unsigned*) HOOK_MEM_BASE_ADDR)[offset] +#define HOOK_TRIG(funcid) do { *((unsigned *) HOOK_TRIG_ADDR) = funcid; } while(0) +#define HOOK_DATA_BYTE(offset) ((unsigned char*) HOOK_MEM_BASE_ADDR)[offset] + + +// ------------------------------------------------------------------ hook_call +int hook_call( unsigned id, unsigned pcnt, ...) { + va_list ap; + unsigned i; + unsigned ret; +#ifdef USING_SOS + PREEMPT_OFF_SAVE(); +#endif + + // pass parameters + HOOK_DATA(0) = id; + + /* Have to make hook_print_str a special case since we call with a + parameter of byte type. Should perhaps be a separate + hook_call. */ + + if (id == hook_print_str) { + int i; + char *str; + + HOOK_DATA(1) = pcnt; + + va_start(ap, pcnt); + str = (char*)va_arg(ap,unsigned); + + for (i=0; i!=pcnt; i++) { + HOOK_DATA_BYTE(8+i) = str[i]; + } + HOOK_DATA_BYTE(8+i) = 0; /* null byte */ + } + else { + va_start(ap, pcnt); + for( i = 1; i <= pcnt; i++ ) HOOK_DATA(i) = va_arg(ap,unsigned); + va_end(ap); + } + + // read from mem to make sure data has propagated to memory before trigging + *((volatile unsigned*) HOOK_MEM_BASE_ADDR); + + // trigger hook + HOOK_TRIG(id); + + // wait for call to finish + while( VHOOK_DATA(0) > 0 ) {} + + // extract return value + + ret = VHOOK_DATA(1); + +#ifdef USING_SOS + PREEMPT_RESTORE(); +#endif + return ret; +} + +unsigned +hook_buf(unsigned i) +{ + return (HOOK_DATA(i)); +} + +void print_str( const char *str ) { + int i; + for (i=1; str[i]; i++); /* find null at end of string */ + hook_call(hook_print_str, i, str); +} + +// --------------------------------------------------------------- CPU_KICK_DOG +void CPU_KICK_DOG(void) { + (void) hook_call( hook_kick_dog, 0 ); +} + +// ------------------------------------------------------- CPU_WATCHDOG_TIMEOUT +void CPU_WATCHDOG_TIMEOUT( unsigned t ) { + (void) hook_call( hook_dog_timeout, 1, t ); +} diff --git a/arch/cris/arch-v32/kernel/vcs_hook.h b/arch/cris/arch-v32/kernel/vcs_hook.h new file mode 100644 index 000000000000..7d73709e3cc6 --- /dev/null +++ b/arch/cris/arch-v32/kernel/vcs_hook.h @@ -0,0 +1,42 @@ +// $Id: vcs_hook.h,v 1.1 2003/08/12 12:01:06 starvik Exp $ +// +// Call simulator hook functions + +#ifndef HOOK_H +#define HOOK_H + +int hook_call( unsigned id, unsigned pcnt, ...); + +enum hook_ids { + hook_debug_on = 1, + hook_debug_off, + hook_stop_sim_ok, + hook_stop_sim_fail, + hook_alloc_shared, + hook_ptr_shared, + hook_free_shared, + hook_file2shared, + hook_cmp_shared, + hook_print_params, + hook_sim_time, + hook_stop_sim, + hook_kick_dog, + hook_dog_timeout, + hook_rand, + hook_srand, + hook_rand_range, + hook_print_str, + hook_print_hex, + hook_cmp_offset_shared, + hook_fill_random_shared, + hook_alloc_random_data, + hook_calloc_random_data, + hook_print_int, + hook_print_uint, + hook_fputc, + hook_init_fd, + hook_sbrk + +}; + +#endif diff --git a/arch/cris/arch-v32/lib/Makefile b/arch/cris/arch-v32/lib/Makefile new file mode 100644 index 000000000000..05b3ec6978d6 --- /dev/null +++ b/arch/cris/arch-v32/lib/Makefile @@ -0,0 +1,6 @@ +# +# Makefile for Etrax-specific library files.. +# + +lib-y = checksum.o checksumcopy.o string.o usercopy.o memset.o csumcpfruser.o spinlock.o + diff --git a/arch/cris/arch-v32/lib/checksum.S b/arch/cris/arch-v32/lib/checksum.S new file mode 100644 index 000000000000..32e66181b826 --- /dev/null +++ b/arch/cris/arch-v32/lib/checksum.S @@ -0,0 +1,111 @@ +/* + * A fast checksum routine using movem + * Copyright (c) 1998-2001, 2003 Axis Communications AB + * + * csum_partial(const unsigned char * buff, int len, unsigned int sum) + */ + + .globl csum_partial +csum_partial: + + ;; r10 - src + ;; r11 - length + ;; r12 - checksum + + ;; check for breakeven length between movem and normal word looping versions + ;; we also do _NOT_ want to compute a checksum over more than the + ;; actual length when length < 40 + + cmpu.w 80,$r11 + blo _word_loop + nop + + ;; need to save the registers we use below in the movem loop + ;; this overhead is why we have a check above for breakeven length + ;; only r0 - r8 have to be saved, the other ones are clobber-able + ;; according to the ABI + + subq 9*4,$sp + subq 10*4,$r11 ; update length for the first loop + movem $r8,[$sp] + + ;; do a movem checksum + +_mloop: movem [$r10+],$r9 ; read 10 longwords + + ;; perform dword checksumming on the 10 longwords + + add.d $r0,$r12 + addc $r1,$r12 + addc $r2,$r12 + addc $r3,$r12 + addc $r4,$r12 + addc $r5,$r12 + addc $r6,$r12 + addc $r7,$r12 + addc $r8,$r12 + addc $r9,$r12 + + ;; fold the carry into the checksum, to avoid having to loop the carry + ;; back into the top + + addc 0,$r12 + addc 0,$r12 ; do it again, since we might have generated a carry + + subq 10*4,$r11 + bge _mloop + nop + + addq 10*4,$r11 ; compensate for last loop underflowing length + + movem [$sp+],$r8 ; restore regs + +_word_loop: + ;; only fold if there is anything to fold. + + cmpq 0,$r12 + beq _no_fold + + ;; fold 32-bit checksum into a 16-bit checksum, to avoid carries below. + ;; r9 and r13 can be used as temporaries. + + moveq -1,$r9 ; put 0xffff in r9, faster than move.d 0xffff,r9 + lsrq 16,$r9 + + move.d $r12,$r13 + lsrq 16,$r13 ; r13 = checksum >> 16 + and.d $r9,$r12 ; checksum = checksum & 0xffff + add.d $r13,$r12 ; checksum += r13 + move.d $r12,$r13 ; do the same again, maybe we got a carry last add + lsrq 16,$r13 + and.d $r9,$r12 + add.d $r13,$r12 + +_no_fold: + cmpq 2,$r11 + blt _no_words + nop + + ;; checksum the rest of the words + + subq 2,$r11 + +_wloop: subq 2,$r11 + bge _wloop + addu.w [$r10+],$r12 + + addq 2,$r11 + +_no_words: + ;; see if we have one odd byte more + cmpq 1,$r11 + beq _do_byte + nop + ret + move.d $r12,$r10 + +_do_byte: + ;; copy and checksum the last byte + addu.b [$r10],$r12 + ret + move.d $r12,$r10 diff --git a/arch/cris/arch-v32/lib/checksumcopy.S b/arch/cris/arch-v32/lib/checksumcopy.S new file mode 100644 index 000000000000..9303ccbadc6d --- /dev/null +++ b/arch/cris/arch-v32/lib/checksumcopy.S @@ -0,0 +1,120 @@ +/* + * A fast checksum+copy routine using movem + * Copyright (c) 1998, 2001, 2003 Axis Communications AB + * + * Authors: Bjorn Wesen + * + * csum_partial_copy_nocheck(const char *src, char *dst, + * int len, unsigned int sum) + */ + + .globl csum_partial_copy_nocheck +csum_partial_copy_nocheck: + + ;; r10 - src + ;; r11 - dst + ;; r12 - length + ;; r13 - checksum + + ;; check for breakeven length between movem and normal word looping versions + ;; we also do _NOT_ want to compute a checksum over more than the + ;; actual length when length < 40 + + cmpu.w 80,$r12 + blo _word_loop + nop + + ;; need to save the registers we use below in the movem loop + ;; this overhead is why we have a check above for breakeven length + ;; only r0 - r8 have to be saved, the other ones are clobber-able + ;; according to the ABI + + subq 9*4,$sp + subq 10*4,$r12 ; update length for the first loop + movem $r8,[$sp] + + ;; do a movem copy and checksum + +1: ;; A failing userspace access (the read) will have this as PC. +_mloop: movem [$r10+],$r9 ; read 10 longwords + movem $r9,[$r11+] ; write 10 longwords + + ;; perform dword checksumming on the 10 longwords + + add.d $r0,$r13 + addc $r1,$r13 + addc $r2,$r13 + addc $r3,$r13 + addc $r4,$r13 + addc $r5,$r13 + addc $r6,$r13 + addc $r7,$r13 + addc $r8,$r13 + addc $r9,$r13 + + ;; fold the carry into the checksum, to avoid having to loop the carry + ;; back into the top + + addc 0,$r13 + addc 0,$r13 ; do it again, since we might have generated a carry + + subq 10*4,$r12 + bge _mloop + nop + + addq 10*4,$r12 ; compensate for last loop underflowing length + + movem [$sp+],$r8 ; restore regs + +_word_loop: + ;; only fold if there is anything to fold. + + cmpq 0,$r13 + beq _no_fold + + ;; fold 32-bit checksum into a 16-bit checksum, to avoid carries below + ;; r9 can be used as temporary. + + move.d $r13,$r9 + lsrq 16,$r9 ; r0 = checksum >> 16 + and.d 0xffff,$r13 ; checksum = checksum & 0xffff + add.d $r9,$r13 ; checksum += r0 + move.d $r13,$r9 ; do the same again, maybe we got a carry last add + lsrq 16,$r9 + and.d 0xffff,$r13 + add.d $r9,$r13 + +_no_fold: + cmpq 2,$r12 + blt _no_words + nop + + ;; copy and checksum the rest of the words + + subq 2,$r12 + +2: ;; A failing userspace access for the read below will have this as PC. +_wloop: move.w [$r10+],$r9 + addu.w $r9,$r13 + subq 2,$r12 + bge _wloop + move.w $r9,[$r11+] + + addq 2,$r12 + +_no_words: + ;; see if we have one odd byte more + cmpq 1,$r12 + beq _do_byte + nop + ret + move.d $r13,$r10 + +_do_byte: + ;; copy and checksum the last byte +3: ;; A failing userspace access for the read below will have this as PC. + move.b [$r10],$r9 + addu.b $r9,$r13 + move.b $r9,[$r11] + ret + move.d $r13,$r10 diff --git a/arch/cris/arch-v32/lib/csumcpfruser.S b/arch/cris/arch-v32/lib/csumcpfruser.S new file mode 100644 index 000000000000..600ec16b9f28 --- /dev/null +++ b/arch/cris/arch-v32/lib/csumcpfruser.S @@ -0,0 +1,69 @@ +/* + * Add-on to transform csum_partial_copy_nocheck in checksumcopy.S into + * csum_partial_copy_from_user by adding exception records. + * + * Copyright (C) 2001, 2003 Axis Communications AB. + * + * Author: Hans-Peter Nilsson. + */ + +#include <asm/errno.h> + +/* Same function body, but a different name. If we just added exception + records to _csum_partial_copy_nocheck and made it generic, we wouldn't + know a user fault from a kernel fault and we would have overhead in + each kernel caller for the error-pointer argument. + + unsigned int csum_partial_copy_from_user + (const char *src, char *dst, int len, unsigned int sum, int *errptr); + + Note that the errptr argument is only set if we encounter an error. + It is conveniently located on the stack, so the normal function body + does not have to handle it. */ + +#define csum_partial_copy_nocheck csum_partial_copy_from_user + +/* There are local labels numbered 1, 2 and 3 present to mark the + different from-user accesses. */ +#include "checksumcopy.S" + + .section .fixup,"ax" + +;; Here from the movem loop; restore stack. +4: + movem [$sp+],$r8 +;; r12 is already decremented. Add back chunk_size-2. + addq 40-2,$r12 + +;; Here from the word loop; r12 is off by 2; add it back. +5: + addq 2,$r12 + +;; Here from a failing single byte. +6: + +;; Signal in *errptr that we had a failing access. + move.d [$sp],$acr + moveq -EFAULT,$r9 + subq 4,$sp + move.d $r9,[$acr] + +;; Clear the rest of the destination area using memset. Preserve the +;; checksum for the readable bytes. + move.d $r13,[$sp] + subq 4,$sp + move.d $r11,$r10 + move $srp,[$sp] + jsr memset + clear.d $r11 + + move [$sp+],$srp + ret + move.d [$sp+],$r10 + + .previous + .section __ex_table,"a" + .dword 1b,4b + .dword 2b,5b + .dword 3b,6b + .previous diff --git a/arch/cris/arch-v32/lib/dram_init.S b/arch/cris/arch-v32/lib/dram_init.S new file mode 100644 index 000000000000..47b6cf5f4afd --- /dev/null +++ b/arch/cris/arch-v32/lib/dram_init.S @@ -0,0 +1,120 @@ +/* $Id: dram_init.S,v 1.4 2005/04/24 18:48:32 starvik Exp $ + * + * DRAM/SDRAM initialization - alter with care + * This file is intended to be included from other assembler files + * + * Note: This file may not modify r8 or r9 because they are used to + * carry information from the decompresser to the kernel + * + * Copyright (C) 2000-2003 Axis Communications AB + * + * Authors: Mikael Starvik (starvik@axis.com) + */ + +/* Just to be certain the config file is included, we include it here + * explicitely instead of depending on it being included in the file that + * uses this code. + */ + +#include <linux/config.h> +#include <asm/arch/hwregs/asm/reg_map_asm.h> +#include <asm/arch/hwregs/asm/bif_core_defs_asm.h> + + ;; WARNING! The registers r8 and r9 are used as parameters carrying + ;; information from the decompressor (if the kernel was compressed). + ;; They should not be used in the code below. + + ; Refer to BIF MDS for a description of SDRAM initialization + + ; Bank configuration + move.d REG_ADDR(bif_core, regi_bif_core, rw_sdram_cfg_grp0), $r0 + move.d CONFIG_ETRAX_SDRAM_GRP0_CONFIG, $r1 + move.d $r1, [$r0] + move.d REG_ADDR(bif_core, regi_bif_core, rw_sdram_cfg_grp1), $r0 + move.d CONFIG_ETRAX_SDRAM_GRP1_CONFIG, $r1 + move.d $r1, [$r0] + + ; Calculate value of mrs_data + ; CAS latency = 2 && bus_width = 32 => 0x40 + ; CAS latency = 3 && bus_width = 32 => 0x60 + ; CAS latency = 2 && bus_width = 16 => 0x20 + ; CAS latency = 3 && bus_width = 16 => 0x30 + + ; Check if value is already supplied in kernel config + move.d CONFIG_ETRAX_SDRAM_COMMAND, $r2 + bne _set_timing + nop + + move.d 0x40, $r4 ; Assume 32 bits and CAS latency = 2 + move.d CONFIG_ETRAX_SDRAM_TIMING, $r1 + and.d 0x07, $r1 ; Get CAS latency + cmpq 2, $r1 ; CL = 2 ? + beq _bw_check + nop + move.d 0x60, $r4 + +_bw_check: + ; Assume that group 0 width is equal to group 1. This assumption + ; is wrong for a group 1 only hardware (such as the grand old + ; StorPoint+). + move.d CONFIG_ETRAX_SDRAM_GRP0_CONFIG, $r1 + and.d 0x200, $r1 ; DRAM width is bit 9 + beq _set_timing + lslq 2, $r4 ; mrs_data starts at bit 2 + lsrq 1, $r4 ; 16 bits. Shift down value. + + ; Set timing parameters (refresh off to avoid Guinness TR 83) +_set_timing: + move.d CONFIG_ETRAX_SDRAM_TIMING, $r1 + and.d ~(3 << reg_bif_core_rw_sdram_timing___ref___lsb), $r1 + move.d REG_ADDR(bif_core, regi_bif_core, rw_sdram_timing), $r0 + move.d $r1, [$r0] + + ; Issue NOP command + move.d REG_ADDR(bif_core, regi_bif_core, rw_sdram_cmd), $r5 + moveq regk_bif_core_nop, $r1 + move.d $r1, [$r5] + + ; Wait 200us + move.d 10000, $r2 +1: bne 1b + subq 1, $r2 + + ; Issue initialization command sequence + move.d _sdram_commands_start, $r2 + and.d 0x000fffff, $r2 ; Make sure commands are read from flash + move.d _sdram_commands_end, $r3 + and.d 0x000fffff, $r3 +1: clear.d $r6 + move.b [$r2+], $r6 ; Load command + or.d $r4, $r6 ; Add calculated mrs + move.d $r6, [$r5] ; Write rw_sdram_cmd + ; Wait 80 ns between each command + move.d 4000, $r7 +2: bne 2b + subq 1, $r7 + cmp.d $r2, $r3 ; Last command? + bne 1b + nop + + ; Start refresh + move.d CONFIG_ETRAX_SDRAM_TIMING, $r1 + move.d REG_ADDR(bif_core, regi_bif_core, rw_sdram_timing), $r0 + move.d $r1, [$r0] + + ; Initialization finished + ba _sdram_commands_end + nop + +_sdram_commands_start: + .byte regk_bif_core_pre ; Precharge + .byte regk_bif_core_ref ; refresh + .byte regk_bif_core_ref ; refresh + .byte regk_bif_core_ref ; refresh + .byte regk_bif_core_ref ; refresh + .byte regk_bif_core_ref ; refresh + .byte regk_bif_core_ref ; refresh + .byte regk_bif_core_ref ; refresh + .byte regk_bif_core_ref ; refresh + .byte regk_bif_core_mrs ; mrs +_sdram_commands_end: diff --git a/arch/cris/arch-v32/lib/hw_settings.S b/arch/cris/arch-v32/lib/hw_settings.S new file mode 100644 index 000000000000..5182e8c2cff2 --- /dev/null +++ b/arch/cris/arch-v32/lib/hw_settings.S @@ -0,0 +1,73 @@ +/* + * $Id: hw_settings.S,v 1.3 2005/04/24 18:36:57 starvik Exp $ + * + * This table is used by some tools to extract hardware parameters. + * The table should be included in the kernel and the decompressor. + * Don't forget to update the tools if you change this table. + * + * Copyright (C) 2001 Axis Communications AB + * + * Authors: Mikael Starvik (starvik@axis.com) + */ + +#include <linux/config.h> +#include <asm/arch/hwregs/asm/reg_map_asm.h> +#include <asm/arch/hwregs/asm/bif_core_defs_asm.h> +#include <asm/arch/hwregs/asm/gio_defs_asm.h> + + .ascii "HW_PARAM_MAGIC" ; Magic number + .dword 0xc0004000 ; Kernel start address + + ; Debug port +#ifdef CONFIG_ETRAX_DEBUG_PORT0 + .dword 0 +#elif defined(CONFIG_ETRAX_DEBUG_PORT1) + .dword 1 +#elif defined(CONFIG_ETRAX_DEBUG_PORT2) + .dword 2 +#elif defined(CONFIG_ETRAX_DEBUG_PORT3) + .dword 3 +#else + .dword 4 ; No debug +#endif + + ; Register values + .dword REG_ADDR(bif_core, regi_bif_core, rw_grp1_cfg) + .dword CONFIG_ETRAX_MEM_GRP1_CONFIG + .dword REG_ADDR(bif_core, regi_bif_core, rw_grp2_cfg) + .dword CONFIG_ETRAX_MEM_GRP2_CONFIG + .dword REG_ADDR(bif_core, regi_bif_core, rw_grp3_cfg) + .dword CONFIG_ETRAX_MEM_GRP3_CONFIG + .dword REG_ADDR(bif_core, regi_bif_core, rw_grp4_cfg) + .dword CONFIG_ETRAX_MEM_GRP4_CONFIG + .dword REG_ADDR(bif_core, regi_bif_core, rw_sdram_cfg_grp0) + .dword CONFIG_ETRAX_SDRAM_GRP0_CONFIG + .dword REG_ADDR(bif_core, regi_bif_core, rw_sdram_cfg_grp1) + .dword CONFIG_ETRAX_SDRAM_GRP1_CONFIG + .dword REG_ADDR(bif_core, regi_bif_core, rw_sdram_timing) + .dword CONFIG_ETRAX_SDRAM_TIMING + .dword REG_ADDR(bif_core, regi_bif_core, rw_sdram_cmd) + .dword CONFIG_ETRAX_SDRAM_COMMAND + + .dword REG_ADDR(gio, regi_gio, rw_pa_dout) + .dword CONFIG_ETRAX_DEF_GIO_PA_OUT + .dword REG_ADDR(gio, regi_gio, rw_pa_oe) + .dword CONFIG_ETRAX_DEF_GIO_PA_OE + .dword REG_ADDR(gio, regi_gio, rw_pb_dout) + .dword CONFIG_ETRAX_DEF_GIO_PB_OUT + .dword REG_ADDR(gio, regi_gio, rw_pb_oe) + .dword CONFIG_ETRAX_DEF_GIO_PB_OE + .dword REG_ADDR(gio, regi_gio, rw_pc_dout) + .dword CONFIG_ETRAX_DEF_GIO_PC_OUT + .dword REG_ADDR(gio, regi_gio, rw_pc_oe) + .dword CONFIG_ETRAX_DEF_GIO_PC_OE + .dword REG_ADDR(gio, regi_gio, rw_pd_dout) + .dword CONFIG_ETRAX_DEF_GIO_PD_OUT + .dword REG_ADDR(gio, regi_gio, rw_pd_oe) + .dword CONFIG_ETRAX_DEF_GIO_PD_OE + .dword REG_ADDR(gio, regi_gio, rw_pe_dout) + .dword CONFIG_ETRAX_DEF_GIO_PE_OUT + .dword REG_ADDR(gio, regi_gio, rw_pe_oe) + .dword CONFIG_ETRAX_DEF_GIO_PE_OE + + .dword 0 ; No more register values diff --git a/arch/cris/arch-v32/lib/memset.c b/arch/cris/arch-v32/lib/memset.c new file mode 100644 index 000000000000..ffca1214674e --- /dev/null +++ b/arch/cris/arch-v32/lib/memset.c @@ -0,0 +1,253 @@ +/*#************************************************************************#*/ +/*#-------------------------------------------------------------------------*/ +/*# */ +/*# FUNCTION NAME: memset() */ +/*# */ +/*# PARAMETERS: void* dst; Destination address. */ +/*# int c; Value of byte to write. */ +/*# int len; Number of bytes to write. */ +/*# */ +/*# RETURNS: dst. */ +/*# */ +/*# DESCRIPTION: Sets the memory dst of length len bytes to c, as standard. */ +/*# Framework taken from memcpy. This routine is */ +/*# very sensitive to compiler changes in register allocation. */ +/*# Should really be rewritten to avoid this problem. */ +/*# */ +/*#-------------------------------------------------------------------------*/ +/*# */ +/*# HISTORY */ +/*# */ +/*# DATE NAME CHANGES */ +/*# ---- ---- ------- */ +/*# 990713 HP Tired of watching this function (or */ +/*# really, the nonoptimized generic */ +/*# implementation) take up 90% of simulator */ +/*# output. Measurements needed. */ +/*# */ +/*#-------------------------------------------------------------------------*/ + +#include <linux/types.h> + +/* No, there's no macro saying 12*4, since it is "hard" to get it into + the asm in a good way. Thus better to expose the problem everywhere. + */ + +/* Assuming 1 cycle per dword written or read (ok, not really true), and + one per instruction, then 43+3*(n/48-1) <= 24+24*(n/48-1) + so n >= 45.7; n >= 0.9; we win on the first full 48-byte block to set. */ + +#define ZERO_BLOCK_SIZE (1*12*4) + +void *memset(void *pdst, + int c, + size_t plen) +{ + /* Ok. Now we want the parameters put in special registers. + Make sure the compiler is able to make something useful of this. */ + + register char *return_dst __asm__ ("r10") = pdst; + register int n __asm__ ("r12") = plen; + register int lc __asm__ ("r11") = c; + + /* Most apps use memset sanely. Only those memsetting about 3..4 + bytes or less get penalized compared to the generic implementation + - and that's not really sane use. */ + + /* Ugh. This is fragile at best. Check with newer GCC releases, if + they compile cascaded "x |= x << 8" sanely! */ + __asm__("movu.b %0,$r13 \n\ + lslq 8,$r13 \n\ + move.b %0,$r13 \n\ + move.d $r13,%0 \n\ + lslq 16,$r13 \n\ + or.d $r13,%0" + : "=r" (lc) : "0" (lc) : "r13"); + + { + register char *dst __asm__ ("r13") = pdst; + + /* This is NONPORTABLE, but since this whole routine is */ + /* grossly nonportable that doesn't matter. */ + + if (((unsigned long) pdst & 3) != 0 + /* Oops! n=0 must be a legal call, regardless of alignment. */ + && n >= 3) + { + if ((unsigned long)dst & 1) + { + *dst = (char) lc; + n--; + dst++; + } + + if ((unsigned long)dst & 2) + { + *(short *)dst = lc; + n -= 2; + dst += 2; + } + } + + /* Now the fun part. For the threshold value of this, check the equation + above. */ + /* Decide which copying method to use. */ + if (n >= ZERO_BLOCK_SIZE) + { + /* For large copies we use 'movem' */ + + /* It is not optimal to tell the compiler about clobbering any + registers; that will move the saving/restoring of those registers + to the function prologue/epilogue, and make non-movem sizes + suboptimal. + + This method is not foolproof; it assumes that the "asm reg" + declarations at the beginning of the function really are used + here (beware: they may be moved to temporary registers). + This way, we do not have to save/move the registers around into + temporaries; we can safely use them straight away. + + If you want to check that the allocation was right; then + check the equalities in the first comment. It should say + "r13=r13, r12=r12, r11=r11" */ + __asm__ volatile (" \n\ + ;; Check that the register asm declaration got right. \n\ + ;; The GCC manual says it will work, but there *has* been bugs. \n\ + .ifnc %0-%1-%4,$r13-$r12-$r11 \n\ + .err \n\ + .endif \n\ + \n\ + ;; Save the registers we'll clobber in the movem process \n\ + ;; on the stack. Don't mention them to gcc, it will only be \n\ + ;; upset. \n\ + subq 11*4,$sp \n\ + movem $r10,[$sp] \n\ + \n\ + move.d $r11,$r0 \n\ + move.d $r11,$r1 \n\ + move.d $r11,$r2 \n\ + move.d $r11,$r3 \n\ + move.d $r11,$r4 \n\ + move.d $r11,$r5 \n\ + move.d $r11,$r6 \n\ + move.d $r11,$r7 \n\ + move.d $r11,$r8 \n\ + move.d $r11,$r9 \n\ + move.d $r11,$r10 \n\ + \n\ + ;; Now we've got this: \n\ + ;; r13 - dst \n\ + ;; r12 - n \n\ + \n\ + ;; Update n for the first loop \n\ + subq 12*4,$r12 \n\ +0: \n\ + subq 12*4,$r12 \n\ + bge 0b \n\ + movem $r11,[$r13+] \n\ + \n\ + addq 12*4,$r12 ;; compensate for last loop underflowing n \n\ + \n\ + ;; Restore registers from stack \n\ + movem [$sp+],$r10" + + /* Outputs */ : "=r" (dst), "=r" (n) + /* Inputs */ : "0" (dst), "1" (n), "r" (lc)); + } + + /* Either we directly starts copying, using dword copying + in a loop, or we copy as much as possible with 'movem' + and then the last block (<44 bytes) is copied here. + This will work since 'movem' will have updated src,dst,n. */ + + while ( n >= 16 ) + { + *((long*)dst)++ = lc; + *((long*)dst)++ = lc; + *((long*)dst)++ = lc; + *((long*)dst)++ = lc; + n -= 16; + } + + /* A switch() is definitely the fastest although it takes a LOT of code. + * Particularly if you inline code this. + */ + switch (n) + { + case 0: + break; + case 1: + *(char*)dst = (char) lc; + break; + case 2: + *(short*)dst = (short) lc; + break; + case 3: + *((short*)dst)++ = (short) lc; + *(char*)dst = (char) lc; + break; + case 4: + *((long*)dst)++ = lc; + break; + case 5: + *((long*)dst)++ = lc; + *(char*)dst = (char) lc; + break; + case 6: + *((long*)dst)++ = lc; + *(short*)dst = (short) lc; + break; + case 7: + *((long*)dst)++ = lc; + *((short*)dst)++ = (short) lc; + *(char*)dst = (char) lc; + break; + case 8: + *((long*)dst)++ = lc; + *((long*)dst)++ = lc; + break; + case 9: + *((long*)dst)++ = lc; + *((long*)dst)++ = lc; + *(char*)dst = (char) lc; + break; + case 10: + *((long*)dst)++ = lc; + *((long*)dst)++ = lc; + *(short*)dst = (short) lc; + break; + case 11: + *((long*)dst)++ = lc; + *((long*)dst)++ = lc; + *((short*)dst)++ = (short) lc; + *(char*)dst = (char) lc; + break; + case 12: + *((long*)dst)++ = lc; + *((long*)dst)++ = lc; + *((long*)dst)++ = lc; + break; + case 13: + *((long*)dst)++ = lc; + *((long*)dst)++ = lc; + *((long*)dst)++ = lc; + *(char*)dst = (char) lc; + break; + case 14: + *((long*)dst)++ = lc; + *((long*)dst)++ = lc; + *((long*)dst)++ = lc; + *(short*)dst = (short) lc; + break; + case 15: + *((long*)dst)++ = lc; + *((long*)dst)++ = lc; + *((long*)dst)++ = lc; + *((short*)dst)++ = (short) lc; + *(char*)dst = (char) lc; + break; + } + } + + return return_dst; /* destination pointer. */ +} /* memset() */ diff --git a/arch/cris/arch-v32/lib/nand_init.S b/arch/cris/arch-v32/lib/nand_init.S new file mode 100644 index 000000000000..aba5c751c282 --- /dev/null +++ b/arch/cris/arch-v32/lib/nand_init.S @@ -0,0 +1,179 @@ +##============================================================================= +## +## nand_init.S +## +## The bootrom copies data from the NAND flash to the internal RAM but +## due to a bug/feature we can only trust the 256 first bytes. So this +## code copies more data from NAND flash to internal RAM. Obvioulsy this +## code must fit in the first 256 bytes so alter with care. +## +## Some notes about the bug/feature for future reference: +## The bootrom copies the first 127 KB from NAND flash to internal +## memory. The problem is that it does a bytewise copy. NAND flashes +## does autoincrement on the address so for a 16-bite device each +## read/write increases the address by two. So the copy loop in the +## bootrom will discard every second byte. This is solved by inserting +## zeroes in every second byte in the first erase block. +## +## The bootrom also incorrectly assumes that it can read the flash +## linear with only one read command but the flash will actually +## switch between normal area and spare area if you do that so we +## can't trust more than the first 256 bytes. +## +##============================================================================= + +#include <asm/arch/hwregs/asm/reg_map_asm.h> +#include <asm/arch/hwregs/asm/gio_defs_asm.h> +#include <asm/arch/hwregs/asm/pinmux_defs_asm.h> +#include <asm/arch/hwregs/asm/bif_core_defs_asm.h> +#include <asm/arch/hwregs/asm/config_defs_asm.h> +#include <linux/config.h> + +;; There are 8-bit NAND flashes and 16-bit NAND flashes. +;; We need to treat them slightly different. +#if CONFIG_ETRAX_FLASH_BUSWIDTH==2 +#define PAGE_SIZE 256 +#else +#error 2 +#define PAGE_SIZE 512 +#endif +#define ERASE_BLOCK 16384 + +;; GPIO pins connected to NAND flash +#define CE 4 +#define CLE 5 +#define ALE 6 +#define BY 7 + +;; Address space for NAND flash +#define NAND_RD_ADDR 0x90000000 +#define NAND_WR_ADDR 0x94000000 + +#define READ_CMD 0x00 + +;; Readability macros +#define CSP_MASK \ + REG_MASK(bif_core, rw_grp3_cfg, gated_csp0) | \ + REG_MASK(bif_core, rw_grp3_cfg, gated_csp1) +#define CSP_VAL \ + REG_STATE(bif_core, rw_grp3_cfg, gated_csp0, rd) | \ + REG_STATE(bif_core, rw_grp3_cfg, gated_csp1, wr) + +;;---------------------------------------------------------------------------- +;; Macros to set/clear GPIO bits + +.macro SET x + or.b (1<<\x),$r9 + move.d $r9, [$r2] +.endm + +.macro CLR x + and.b ~(1<<\x),$r9 + move.d $r9, [$r2] +.endm + +;;---------------------------------------------------------------------------- + +nand_boot: + ;; Check if nand boot was selected + move.d REG_ADDR(config, regi_config, r_bootsel), $r0 + move.d [$r0], $r0 + and.d REG_MASK(config, r_bootsel, boot_mode), $r0 + cmp.d REG_STATE(config, r_bootsel, boot_mode, nand), $r0 + bne normal_boot ; No NAND boot + nop + +copy_nand_to_ram: + ;; copy_nand_to_ram + ;; Arguments + ;; r10 - destination + ;; r11 - source offset + ;; r12 - size + ;; r13 - Address to jump to after completion + ;; Note : r10-r12 are clobbered on return + ;; Registers used: + ;; r0 - NAND_RD_ADDR + ;; r1 - NAND_WR_ADDR + ;; r2 - reg_gio_rw_pa_dout + ;; r3 - reg_gio_r_pa_din + ;; r4 - tmp + ;; r5 - byte counter within a page + ;; r6 - reg_pinmux_rw_pa + ;; r7 - reg_gio_rw_pa_oe + ;; r8 - reg_bif_core_rw_grp3_cfg + ;; r9 - reg_gio_rw_pa_dout shadow + move.d 0x90000000, $r0 + move.d 0x94000000, $r1 + move.d REG_ADDR(gio, regi_gio, rw_pa_dout), $r2 + move.d REG_ADDR(gio, regi_gio, r_pa_din), $r3 + move.d REG_ADDR(pinmux, regi_pinmux, rw_pa), $r6 + move.d REG_ADDR(gio, regi_gio, rw_pa_oe), $r7 + move.d REG_ADDR(bif_core, regi_bif_core, rw_grp3_cfg), $r8 + +#if CONFIG_ETRAX_FLASH_BUSWIDTH==2 + lsrq 1, $r11 +#endif + ;; Set up GPIO + move.d [$r2], $r9 + move.d [$r7], $r4 + or.b (1<<ALE) | (1 << CLE) | (1<<CE), $r4 + move.d $r4, [$r7] + + ;; Set up bif + move.d [$r8], $r4 + and.d CSP_MASK, $r4 + or.d CSP_VAL, $r4 + move.d $r4, [$r8] + +1: ;; Copy one page + CLR CE + SET CLE + moveq READ_CMD, $r4 + move.b $r4, [$r1] + moveq 20, $r4 +2: bne 2b + subq 1, $r4 + CLR CLE + SET ALE + clear.w [$r1] ; Column address = 0 + move.d $r11, $r4 + lsrq 8, $r4 + move.b $r4, [$r1] ; Row address + lsrq 8, $r4 + move.b $r4, [$r1] ; Row adddress + moveq 20, $r4 +2: bne 2b + subq 1, $r4 + CLR ALE +2: move.d [$r3], $r4 + and.d 1 << BY, $r4 + beq 2b + movu.w PAGE_SIZE, $r5 +2: ; Copy one byte/word +#if CONFIG_ETRAX_FLASH_BUSWIDTH==2 + move.w [$r0], $r4 +#else + move.b [$r0], $r4 +#endif + subq 1, $r5 + bne 2b +#if CONFIG_ETRAX_FLASH_BUSWIDTH==2 + move.w $r4, [$r10+] + subu.w PAGE_SIZE*2, $r12 +#else + move.b $r4, [$r10+] + subu.w PAGE_SIZE, $r12 +#endif + bpl 1b + addu.w PAGE_SIZE, $r11 + + ;; End of copy + jump $r13 + nop + + ;; This will warn if the code above is too large. If you consider + ;; to remove this you don't understand the bug/feature. + .org 256 + .org ERASE_BLOCK + +normal_boot: diff --git a/arch/cris/arch-v32/lib/spinlock.S b/arch/cris/arch-v32/lib/spinlock.S new file mode 100644 index 000000000000..2437ae7f6ed2 --- /dev/null +++ b/arch/cris/arch-v32/lib/spinlock.S @@ -0,0 +1,33 @@ +;; Core of the spinlock implementation +;; +;; Copyright (C) 2004 Axis Communications AB. +;; +;; Author: Mikael Starvik + + + .global cris_spin_lock + .global cris_spin_trylock + + .text + +cris_spin_lock: + clearf p +1: test.d [$r10] + beq 1b + clearf p + ax + clear.d [$r10] + bcs 1b + clearf p + ret + nop + +cris_spin_trylock: + clearf p +1: move.d [$r10], $r11 + ax + clear.d [$r10] + bcs 1b + clearf p + ret + move.d $r11,$r10 diff --git a/arch/cris/arch-v32/lib/string.c b/arch/cris/arch-v32/lib/string.c new file mode 100644 index 000000000000..98e282ac824a --- /dev/null +++ b/arch/cris/arch-v32/lib/string.c @@ -0,0 +1,219 @@ +/*#************************************************************************#*/ +/*#-------------------------------------------------------------------------*/ +/*# */ +/*# FUNCTION NAME: memcpy() */ +/*# */ +/*# PARAMETERS: void* dst; Destination address. */ +/*# void* src; Source address. */ +/*# int len; Number of bytes to copy. */ +/*# */ +/*# RETURNS: dst. */ +/*# */ +/*# DESCRIPTION: Copies len bytes of memory from src to dst. No guarantees */ +/*# about copying of overlapping memory areas. This routine is */ +/*# very sensitive to compiler changes in register allocation. */ +/*# Should really be rewritten to avoid this problem. */ +/*# */ +/*#-------------------------------------------------------------------------*/ +/*# */ +/*# HISTORY */ +/*# */ +/*# DATE NAME CHANGES */ +/*# ---- ---- ------- */ +/*# 941007 Kenny R Creation */ +/*# 941011 Kenny R Lots of optimizations and inlining. */ +/*# 941129 Ulf A Adapted for use in libc. */ +/*# 950216 HP N==0 forgotten if non-aligned src/dst. */ +/*# Added some optimizations. */ +/*# 001025 HP Make src and dst char *. Align dst to */ +/*# dword, not just word-if-both-src-and-dst- */ +/*# are-misaligned. */ +/*# */ +/*#-------------------------------------------------------------------------*/ + +#include <linux/types.h> + +void *memcpy(void *pdst, + const void *psrc, + size_t pn) +{ + /* Ok. Now we want the parameters put in special registers. + Make sure the compiler is able to make something useful of this. + As it is now: r10 -> r13; r11 -> r11 (nop); r12 -> r12 (nop). + + If gcc was allright, it really would need no temporaries, and no + stack space to save stuff on. */ + + register void *return_dst __asm__ ("r10") = pdst; + register char *dst __asm__ ("r13") = pdst; + register const char *src __asm__ ("r11") = psrc; + register int n __asm__ ("r12") = pn; + + + /* When src is aligned but not dst, this makes a few extra needless + cycles. I believe it would take as many to check that the + re-alignment was unnecessary. */ + if (((unsigned long) dst & 3) != 0 + /* Don't align if we wouldn't copy more than a few bytes; so we + don't have to check further for overflows. */ + && n >= 3) + { + if ((unsigned long) dst & 1) + { + n--; + *(char*)dst = *(char*)src; + src++; + dst++; + } + + if ((unsigned long) dst & 2) + { + n -= 2; + *(short*)dst = *(short*)src; + src += 2; + dst += 2; + } + } + + /* Decide which copying method to use. Movem is dirt cheap, so the + overheap is low enough to always use the minimum block size as the + threshold. */ + if (n >= 44) + { + /* For large copies we use 'movem' */ + + /* It is not optimal to tell the compiler about clobbering any + registers; that will move the saving/restoring of those registers + to the function prologue/epilogue, and make non-movem sizes + suboptimal. */ + __asm__ volatile (" \n\ + ;; Check that the register asm declaration got right. \n\ + ;; The GCC manual explicitly says TRT will happen. \n\ + .ifnc %0-%1-%2,$r13-$r11-$r12 \n\ + .err \n\ + .endif \n\ + \n\ + ;; Save the registers we'll use in the movem process \n\ + \n\ + ;; on the stack. \n\ + subq 11*4,$sp \n\ + movem $r10,[$sp] \n\ + \n\ + ;; Now we've got this: \n\ + ;; r11 - src \n\ + ;; r13 - dst \n\ + ;; r12 - n \n\ + \n\ + ;; Update n for the first loop \n\ + subq 44,$r12 \n\ +0: \n\ + movem [$r11+],$r10 \n\ + subq 44,$r12 \n\ + bge 0b \n\ + movem $r10,[$r13+] \n\ + \n\ + addq 44,$r12 ;; compensate for last loop underflowing n \n\ + \n\ + ;; Restore registers from stack \n\ + movem [$sp+],$r10" + + /* Outputs */ : "=r" (dst), "=r" (src), "=r" (n) + /* Inputs */ : "0" (dst), "1" (src), "2" (n)); + + } + + /* Either we directly starts copying, using dword copying + in a loop, or we copy as much as possible with 'movem' + and then the last block (<44 bytes) is copied here. + This will work since 'movem' will have updated src,dst,n. */ + + while ( n >= 16 ) + { + *((long*)dst)++ = *((long*)src)++; + *((long*)dst)++ = *((long*)src)++; + *((long*)dst)++ = *((long*)src)++; + *((long*)dst)++ = *((long*)src)++; + n -= 16; + } + + /* A switch() is definitely the fastest although it takes a LOT of code. + * Particularly if you inline code this. + */ + switch (n) + { + case 0: + break; + case 1: + *(char*)dst = *(char*)src; + break; + case 2: + *(short*)dst = *(short*)src; + break; + case 3: + *((short*)dst)++ = *((short*)src)++; + *(char*)dst = *(char*)src; + break; + case 4: + *((long*)dst)++ = *((long*)src)++; + break; + case 5: + *((long*)dst)++ = *((long*)src)++; + *(char*)dst = *(char*)src; + break; + case 6: + *((long*)dst)++ = *((long*)src)++; + *(short*)dst = *(short*)src; + break; + case 7: + *((long*)dst)++ = *((long*)src)++; + *((short*)dst)++ = *((short*)src)++; + *(char*)dst = *(char*)src; + break; + case 8: + *((long*)dst)++ = *((long*)src)++; + *((long*)dst)++ = *((long*)src)++; + break; + case 9: + *((long*)dst)++ = *((long*)src)++; + *((long*)dst)++ = *((long*)src)++; + *(char*)dst = *(char*)src; + break; + case 10: + *((long*)dst)++ = *((long*)src)++; + *((long*)dst)++ = *((long*)src)++; + *(short*)dst = *(short*)src; + break; + case 11: + *((long*)dst)++ = *((long*)src)++; + *((long*)dst)++ = *((long*)src)++; + *((short*)dst)++ = *((short*)src)++; + *(char*)dst = *(char*)src; + break; + case 12: + *((long*)dst)++ = *((long*)src)++; + *((long*)dst)++ = *((long*)src)++; + *((long*)dst)++ = *((long*)src)++; + break; + case 13: + *((long*)dst)++ = *((long*)src)++; + *((long*)dst)++ = *((long*)src)++; + *((long*)dst)++ = *((long*)src)++; + *(char*)dst = *(char*)src; + break; + case 14: + *((long*)dst)++ = *((long*)src)++; + *((long*)dst)++ = *((long*)src)++; + *((long*)dst)++ = *((long*)src)++; + *(short*)dst = *(short*)src; + break; + case 15: + *((long*)dst)++ = *((long*)src)++; + *((long*)dst)++ = *((long*)src)++; + *((long*)dst)++ = *((long*)src)++; + *((short*)dst)++ = *((short*)src)++; + *(char*)dst = *(char*)src; + break; + } + + return return_dst; /* destination pointer. */ +} /* memcpy() */ diff --git a/arch/cris/arch-v32/lib/usercopy.c b/arch/cris/arch-v32/lib/usercopy.c new file mode 100644 index 000000000000..f0b08460c1be --- /dev/null +++ b/arch/cris/arch-v32/lib/usercopy.c @@ -0,0 +1,470 @@ +/* + * User address space access functions. + * The non-inlined parts of asm-cris/uaccess.h are here. + * + * Copyright (C) 2000, 2003 Axis Communications AB. + * + * Written by Hans-Peter Nilsson. + * Pieces used from memcpy, originally by Kenny Ranerup long time ago. + */ + +#include <asm/uaccess.h> + +/* Asm:s have been tweaked (within the domain of correctness) to give + satisfactory results for "gcc version 3.2.1 Axis release R53/1.53-v32". + + Check regularly... + + Note that for CRISv32, the PC saved at a bus-fault is the address + *at* the faulting instruction, with a special case for instructions + in delay slots: then it's the address of the branch. Note also that + in contrast to v10, a postincrement in the instruction is *not* + performed at a bus-fault; the register is seen having the original + value in fault handlers. */ + + +/* Copy to userspace. This is based on the memcpy used for + kernel-to-kernel copying; see "string.c". */ + +unsigned long +__copy_user (void __user *pdst, const void *psrc, unsigned long pn) +{ + /* We want the parameters put in special registers. + Make sure the compiler is able to make something useful of this. + As it is now: r10 -> r13; r11 -> r11 (nop); r12 -> r12 (nop). + + FIXME: Comment for old gcc version. Check. + If gcc was allright, it really would need no temporaries, and no + stack space to save stuff on. */ + + register char *dst __asm__ ("r13") = pdst; + register const char *src __asm__ ("r11") = psrc; + register int n __asm__ ("r12") = pn; + register int retn __asm__ ("r10") = 0; + + + /* When src is aligned but not dst, this makes a few extra needless + cycles. I believe it would take as many to check that the + re-alignment was unnecessary. */ + if (((unsigned long) dst & 3) != 0 + /* Don't align if we wouldn't copy more than a few bytes; so we + don't have to check further for overflows. */ + && n >= 3) + { + if ((unsigned long) dst & 1) + { + __asm_copy_to_user_1 (dst, src, retn); + n--; + } + + if ((unsigned long) dst & 2) + { + __asm_copy_to_user_2 (dst, src, retn); + n -= 2; + } + } + + /* Movem is dirt cheap. The overheap is low enough to always use the + minimum possible block size as the threshold. */ + if (n >= 44) + { + /* For large copies we use 'movem'. */ + + /* It is not optimal to tell the compiler about clobbering any + registers; that will move the saving/restoring of those registers + to the function prologue/epilogue, and make non-movem sizes + suboptimal. */ + __asm__ volatile ("\ + ;; Check that the register asm declaration got right. \n\ + ;; The GCC manual explicitly says TRT will happen. \n\ + .ifnc %0%1%2%3,$r13$r11$r12$r10 \n\ + .err \n\ + .endif \n\ + \n\ + ;; Save the registers we'll use in the movem process \n\ + ;; on the stack. \n\ + subq 11*4,$sp \n\ + movem $r10,[$sp] \n\ + \n\ + ;; Now we've got this: \n\ + ;; r11 - src \n\ + ;; r13 - dst \n\ + ;; r12 - n \n\ + \n\ + ;; Update n for the first loop \n\ + subq 44,$r12 \n\ +0: \n\ + movem [$r11+],$r10 \n\ + subq 44,$r12 \n\ +1: bge 0b \n\ + movem $r10,[$r13+] \n\ +3: \n\ + addq 44,$r12 ;; compensate for last loop underflowing n \n\ + \n\ + ;; Restore registers from stack \n\ + movem [$sp+],$r10 \n\ +2: \n\ + .section .fixup,\"ax\" \n\ +4: \n\ +; When failing on any of the 1..44 bytes in a chunk, we adjust back the \n\ +; source pointer and just drop through to the by-16 and by-4 loops to \n\ +; get the correct number of failing bytes. This necessarily means a \n\ +; few extra exceptions, but invalid user pointers shouldn't happen in \n\ +; time-critical code anyway. \n\ + jump 3b \n\ + subq 44,$r11 \n\ + \n\ + .previous \n\ + .section __ex_table,\"a\" \n\ + .dword 1b,4b \n\ + .previous" + + /* Outputs */ : "=r" (dst), "=r" (src), "=r" (n), "=r" (retn) + /* Inputs */ : "0" (dst), "1" (src), "2" (n), "3" (retn)); + + } + + while (n >= 16) + { + __asm_copy_to_user_16 (dst, src, retn); + n -= 16; + } + + /* Having a separate by-four loops cuts down on cache footprint. + FIXME: Test with and without; increasing switch to be 0..15. */ + while (n >= 4) + { + __asm_copy_to_user_4 (dst, src, retn); + n -= 4; + } + + switch (n) + { + case 0: + break; + case 1: + __asm_copy_to_user_1 (dst, src, retn); + break; + case 2: + __asm_copy_to_user_2 (dst, src, retn); + break; + case 3: + __asm_copy_to_user_3 (dst, src, retn); + break; + } + + return retn; +} + +/* Copy from user to kernel, zeroing the bytes that were inaccessible in + userland. The return-value is the number of bytes that were + inaccessible. */ + +unsigned long +__copy_user_zeroing (void __user *pdst, const void *psrc, unsigned long pn) +{ + /* We want the parameters put in special registers. + Make sure the compiler is able to make something useful of this. + As it is now: r10 -> r13; r11 -> r11 (nop); r12 -> r12 (nop). + + FIXME: Comment for old gcc version. Check. + If gcc was allright, it really would need no temporaries, and no + stack space to save stuff on. */ + + register char *dst __asm__ ("r13") = pdst; + register const char *src __asm__ ("r11") = psrc; + register int n __asm__ ("r12") = pn; + register int retn __asm__ ("r10") = 0; + + /* The best reason to align src is that we then know that a read-fault + was for aligned bytes; there's no 1..3 remaining good bytes to + pickle. */ + if (((unsigned long) src & 3) != 0) + { + if (((unsigned long) src & 1) && n != 0) + { + __asm_copy_from_user_1 (dst, src, retn); + n--; + } + + if (((unsigned long) src & 2) && n >= 2) + { + __asm_copy_from_user_2 (dst, src, retn); + n -= 2; + } + + /* We only need one check after the unalignment-adjustments, because + if both adjustments were done, either both or neither reference + had an exception. */ + if (retn != 0) + goto copy_exception_bytes; + } + + /* Movem is dirt cheap. The overheap is low enough to always use the + minimum possible block size as the threshold. */ + if (n >= 44) + { + /* It is not optimal to tell the compiler about clobbering any + registers; that will move the saving/restoring of those registers + to the function prologue/epilogue, and make non-movem sizes + suboptimal. */ + __asm__ volatile ("\ + .ifnc %0%1%2%3,$r13$r11$r12$r10 \n\ + .err \n\ + .endif \n\ + \n\ + ;; Save the registers we'll use in the movem process \n\ + ;; on the stack. \n\ + subq 11*4,$sp \n\ + movem $r10,[$sp] \n\ + \n\ + ;; Now we've got this: \n\ + ;; r11 - src \n\ + ;; r13 - dst \n\ + ;; r12 - n \n\ + \n\ + ;; Update n for the first loop \n\ + subq 44,$r12 \n\ +0: \n\ + movem [$r11+],$r10 \n\ + \n\ + subq 44,$r12 \n\ + bge 0b \n\ + movem $r10,[$r13+] \n\ + \n\ +4: \n\ + addq 44,$r12 ;; compensate for last loop underflowing n \n\ + \n\ + ;; Restore registers from stack \n\ + movem [$sp+],$r10 \n\ + .section .fixup,\"ax\" \n\ + \n\ +;; Do not jump back into the loop if we fail. For some uses, we get a \n\ +;; page fault somewhere on the line. Without checking for page limits, \n\ +;; we don't know where, but we need to copy accurately and keep an \n\ +;; accurate count; not just clear the whole line. To do that, we fall \n\ +;; down in the code below, proceeding with smaller amounts. It should \n\ +;; be kept in mind that we have to cater to code like what at one time \n\ +;; was in fs/super.c: \n\ +;; i = size - copy_from_user((void *)page, data, size); \n\ +;; which would cause repeated faults while clearing the remainder of \n\ +;; the SIZE bytes at PAGE after the first fault. \n\ +;; A caveat here is that we must not fall through from a failing page \n\ +;; to a valid page. \n\ + \n\ +3: \n\ + jump 4b ;; Fall through, pretending the fault didn't happen. \n\ + nop \n\ + \n\ + .previous \n\ + .section __ex_table,\"a\" \n\ + .dword 0b,3b \n\ + .previous" + + /* Outputs */ : "=r" (dst), "=r" (src), "=r" (n), "=r" (retn) + /* Inputs */ : "0" (dst), "1" (src), "2" (n), "3" (retn)); + } + + /* Either we directly start copying here, using dword copying in a loop, + or we copy as much as possible with 'movem' and then the last block + (<44 bytes) is copied here. This will work since 'movem' will have + updated src, dst and n. (Except with failing src.) + + Since we want to keep src accurate, we can't use + __asm_copy_from_user_N with N != (1, 2, 4); it updates dst and + retn, but not src (by design; it's value is ignored elsewhere). */ + + while (n >= 4) + { + __asm_copy_from_user_4 (dst, src, retn); + n -= 4; + + if (retn) + goto copy_exception_bytes; + } + + /* If we get here, there were no memory read faults. */ + switch (n) + { + /* These copies are at least "naturally aligned" (so we don't have + to check each byte), due to the src alignment code before the + movem loop. The *_3 case *will* get the correct count for retn. */ + case 0: + /* This case deliberately left in (if you have doubts check the + generated assembly code). */ + break; + case 1: + __asm_copy_from_user_1 (dst, src, retn); + break; + case 2: + __asm_copy_from_user_2 (dst, src, retn); + break; + case 3: + __asm_copy_from_user_3 (dst, src, retn); + break; + } + + /* If we get here, retn correctly reflects the number of failing + bytes. */ + return retn; + +copy_exception_bytes: + /* We already have "retn" bytes cleared, and need to clear the + remaining "n" bytes. A non-optimized simple byte-for-byte in-line + memset is preferred here, since this isn't speed-critical code and + we'd rather have this a leaf-function than calling memset. */ + { + char *endp; + for (endp = dst + n; dst < endp; dst++) + *dst = 0; + } + + return retn + n; +} + +/* Zero userspace. */ + +unsigned long +__do_clear_user (void __user *pto, unsigned long pn) +{ + /* We want the parameters put in special registers. + Make sure the compiler is able to make something useful of this. + As it is now: r10 -> r13; r11 -> r11 (nop); r12 -> r12 (nop). + + FIXME: Comment for old gcc version. Check. + If gcc was allright, it really would need no temporaries, and no + stack space to save stuff on. */ + + register char *dst __asm__ ("r13") = pto; + register int n __asm__ ("r12") = pn; + register int retn __asm__ ("r10") = 0; + + + if (((unsigned long) dst & 3) != 0 + /* Don't align if we wouldn't copy more than a few bytes. */ + && n >= 3) + { + if ((unsigned long) dst & 1) + { + __asm_clear_1 (dst, retn); + n--; + } + + if ((unsigned long) dst & 2) + { + __asm_clear_2 (dst, retn); + n -= 2; + } + } + + /* Decide which copying method to use. + FIXME: This number is from the "ordinary" kernel memset. */ + if (n >= 48) + { + /* For large clears we use 'movem' */ + + /* It is not optimal to tell the compiler about clobbering any + call-saved registers; that will move the saving/restoring of + those registers to the function prologue/epilogue, and make + non-movem sizes suboptimal. + + This method is not foolproof; it assumes that the "asm reg" + declarations at the beginning of the function really are used + here (beware: they may be moved to temporary registers). + This way, we do not have to save/move the registers around into + temporaries; we can safely use them straight away. + + If you want to check that the allocation was right; then + check the equalities in the first comment. It should say + something like "r13=r13, r11=r11, r12=r12". */ + __asm__ volatile ("\ + .ifnc %0%1%2,$r13$r12$r10 \n\ + .err \n\ + .endif \n\ + \n\ + ;; Save the registers we'll clobber in the movem process \n\ + ;; on the stack. Don't mention them to gcc, it will only be \n\ + ;; upset. \n\ + subq 11*4,$sp \n\ + movem $r10,[$sp] \n\ + \n\ + clear.d $r0 \n\ + clear.d $r1 \n\ + clear.d $r2 \n\ + clear.d $r3 \n\ + clear.d $r4 \n\ + clear.d $r5 \n\ + clear.d $r6 \n\ + clear.d $r7 \n\ + clear.d $r8 \n\ + clear.d $r9 \n\ + clear.d $r10 \n\ + clear.d $r11 \n\ + \n\ + ;; Now we've got this: \n\ + ;; r13 - dst \n\ + ;; r12 - n \n\ + \n\ + ;; Update n for the first loop \n\ + subq 12*4,$r12 \n\ +0: \n\ + subq 12*4,$r12 \n\ +1: \n\ + bge 0b \n\ + movem $r11,[$r13+] \n\ + \n\ + addq 12*4,$r12 ;; compensate for last loop underflowing n \n\ + \n\ + ;; Restore registers from stack \n\ + movem [$sp+],$r10 \n\ +2: \n\ + .section .fixup,\"ax\" \n\ +3: \n\ + movem [$sp],$r10 \n\ + addq 12*4,$r10 \n\ + addq 12*4,$r13 \n\ + movem $r10,[$sp] \n\ + jump 0b \n\ + clear.d $r10 \n\ + \n\ + .previous \n\ + .section __ex_table,\"a\" \n\ + .dword 1b,3b \n\ + .previous" + + /* Outputs */ : "=r" (dst), "=r" (n), "=r" (retn) + /* Inputs */ : "0" (dst), "1" (n), "2" (retn) + /* Clobber */ : "r11"); + } + + while (n >= 16) + { + __asm_clear_16 (dst, retn); + n -= 16; + } + + /* Having a separate by-four loops cuts down on cache footprint. + FIXME: Test with and without; increasing switch to be 0..15. */ + while (n >= 4) + { + __asm_clear_4 (dst, retn); + n -= 4; + } + + switch (n) + { + case 0: + break; + case 1: + __asm_clear_1 (dst, retn); + break; + case 2: + __asm_clear_2 (dst, retn); + break; + case 3: + __asm_clear_3 (dst, retn); + break; + } + + return retn; +} diff --git a/arch/cris/arch-v32/mm/Makefile b/arch/cris/arch-v32/mm/Makefile new file mode 100644 index 000000000000..9146f88484b1 --- /dev/null +++ b/arch/cris/arch-v32/mm/Makefile @@ -0,0 +1,3 @@ +# Makefile for the Linux/cris parts of the memory manager. + +obj-y := mmu.o init.o tlb.o intmem.o diff --git a/arch/cris/arch-v32/mm/init.c b/arch/cris/arch-v32/mm/init.c new file mode 100644 index 000000000000..f2fba27d822c --- /dev/null +++ b/arch/cris/arch-v32/mm/init.c @@ -0,0 +1,174 @@ +/* + * Set up paging and the MMU. + * + * Copyright (C) 2000-2003, Axis Communications AB. + * + * Authors: Bjorn Wesen <bjornw@axis.com> + * Tobias Anderberg <tobiasa@axis.com>, CRISv32 port. + */ +#include <linux/config.h> +#include <linux/mmzone.h> +#include <linux/init.h> +#include <linux/bootmem.h> +#include <linux/mm.h> +#include <linux/config.h> +#include <asm/pgtable.h> +#include <asm/page.h> +#include <asm/types.h> +#include <asm/mmu.h> +#include <asm/io.h> +#include <asm/mmu_context.h> +#include <asm/arch/hwregs/asm/mmu_defs_asm.h> +#include <asm/arch/hwregs/supp_reg.h> + +extern void tlb_init(void); + +/* + * The kernel is already mapped with linear mapping at kseg_c so there's no + * need to map it with a page table. However, head.S also temporarily mapped it + * at kseg_4 thus the ksegs are set up again. Also clear the TLB and do various + * other paging stuff. + */ +void __init +cris_mmu_init(void) +{ + unsigned long mmu_config; + unsigned long mmu_kbase_hi; + unsigned long mmu_kbase_lo; + unsigned short mmu_page_id; + + /* + * Make sure the current pgd table points to something sane, even if it + * is most probably not used until the next switch_mm. + */ + per_cpu(current_pgd, smp_processor_id()) = init_mm.pgd; + +#ifdef CONFIG_SMP + { + pgd_t **pgd; + pgd = (pgd_t**)&per_cpu(current_pgd, smp_processor_id()); + SUPP_BANK_SEL(1); + SUPP_REG_WR(RW_MM_TLB_PGD, pgd); + SUPP_BANK_SEL(2); + SUPP_REG_WR(RW_MM_TLB_PGD, pgd); + } +#endif + + /* Initialise the TLB. Function found in tlb.c. */ + tlb_init(); + + /* Enable exceptions and initialize the kernel segments. */ + mmu_config = ( REG_STATE(mmu, rw_mm_cfg, we, on) | + REG_STATE(mmu, rw_mm_cfg, acc, on) | + REG_STATE(mmu, rw_mm_cfg, ex, on) | + REG_STATE(mmu, rw_mm_cfg, inv, on) | + REG_STATE(mmu, rw_mm_cfg, seg_f, linear) | + REG_STATE(mmu, rw_mm_cfg, seg_e, linear) | + REG_STATE(mmu, rw_mm_cfg, seg_d, page) | + REG_STATE(mmu, rw_mm_cfg, seg_c, linear) | + REG_STATE(mmu, rw_mm_cfg, seg_b, linear) | +#ifndef CONFIG_ETRAXFS_SIM + REG_STATE(mmu, rw_mm_cfg, seg_a, page) | +#else + REG_STATE(mmu, rw_mm_cfg, seg_a, linear) | +#endif + REG_STATE(mmu, rw_mm_cfg, seg_9, page) | + REG_STATE(mmu, rw_mm_cfg, seg_8, page) | + REG_STATE(mmu, rw_mm_cfg, seg_7, page) | + REG_STATE(mmu, rw_mm_cfg, seg_6, page) | + REG_STATE(mmu, rw_mm_cfg, seg_5, page) | + REG_STATE(mmu, rw_mm_cfg, seg_4, page) | + REG_STATE(mmu, rw_mm_cfg, seg_3, page) | + REG_STATE(mmu, rw_mm_cfg, seg_2, page) | + REG_STATE(mmu, rw_mm_cfg, seg_1, page) | + REG_STATE(mmu, rw_mm_cfg, seg_0, page)); + + mmu_kbase_hi = ( REG_FIELD(mmu, rw_mm_kbase_hi, base_f, 0x0) | + REG_FIELD(mmu, rw_mm_kbase_hi, base_e, 0x8) | + REG_FIELD(mmu, rw_mm_kbase_hi, base_d, 0x0) | +#ifndef CONFIG_ETRAXFS_SIM + REG_FIELD(mmu, rw_mm_kbase_hi, base_c, 0x4) | +#else + REG_FIELD(mmu, rw_mm_kbase_hi, base_c, 0x0) | +#endif + REG_FIELD(mmu, rw_mm_kbase_hi, base_b, 0xb) | +#ifndef CONFIG_ETRAXFS_SIM + REG_FIELD(mmu, rw_mm_kbase_hi, base_a, 0x0) | +#else + REG_FIELD(mmu, rw_mm_kbase_hi, base_a, 0xa) | +#endif + REG_FIELD(mmu, rw_mm_kbase_hi, base_9, 0x0) | + REG_FIELD(mmu, rw_mm_kbase_hi, base_8, 0x0)); + + mmu_kbase_lo = ( REG_FIELD(mmu, rw_mm_kbase_lo, base_7, 0x0) | + REG_FIELD(mmu, rw_mm_kbase_lo, base_6, 0x0) | + REG_FIELD(mmu, rw_mm_kbase_lo, base_5, 0x0) | + REG_FIELD(mmu, rw_mm_kbase_lo, base_4, 0x0) | + REG_FIELD(mmu, rw_mm_kbase_lo, base_3, 0x0) | + REG_FIELD(mmu, rw_mm_kbase_lo, base_2, 0x0) | + REG_FIELD(mmu, rw_mm_kbase_lo, base_1, 0x0) | + REG_FIELD(mmu, rw_mm_kbase_lo, base_0, 0x0)); + + mmu_page_id = REG_FIELD(mmu, rw_mm_tlb_hi, pid, 0); + + /* Update the instruction MMU. */ + SUPP_BANK_SEL(BANK_IM); + SUPP_REG_WR(RW_MM_CFG, mmu_config); + SUPP_REG_WR(RW_MM_KBASE_HI, mmu_kbase_hi); + SUPP_REG_WR(RW_MM_KBASE_LO, mmu_kbase_lo); + SUPP_REG_WR(RW_MM_TLB_HI, mmu_page_id); + + /* Update the data MMU. */ + SUPP_BANK_SEL(BANK_DM); + SUPP_REG_WR(RW_MM_CFG, mmu_config); + SUPP_REG_WR(RW_MM_KBASE_HI, mmu_kbase_hi); + SUPP_REG_WR(RW_MM_KBASE_LO, mmu_kbase_lo); + SUPP_REG_WR(RW_MM_TLB_HI, mmu_page_id); + + SPEC_REG_WR(SPEC_REG_PID, 0); + + /* + * The MMU has been enabled ever since head.S but just to make it + * totally obvious enable it here as well. + */ + SUPP_BANK_SEL(BANK_GC); + SUPP_REG_WR(RW_GC_CFG, 0xf); /* IMMU, DMMU, ICache, DCache on */ +} + +void __init +paging_init(void) +{ + int i; + unsigned long zones_size[MAX_NR_ZONES]; + + printk("Setting up paging and the MMU.\n"); + + /* Clear out the init_mm.pgd that will contain the kernel's mappings. */ + for(i = 0; i < PTRS_PER_PGD; i++) + swapper_pg_dir[i] = __pgd(0); + + cris_mmu_init(); + + /* + * Initialize the bad page table and bad page to point to a couple of + * allocated pages. + */ + empty_zero_page = (unsigned long) alloc_bootmem_pages(PAGE_SIZE); + memset((void *) empty_zero_page, 0, PAGE_SIZE); + + /* All pages are DMA'able in Etrax, so put all in the DMA'able zone. */ + zones_size[0] = ((unsigned long) high_memory - PAGE_OFFSET) >> PAGE_SHIFT; + + for (i = 1; i < MAX_NR_ZONES; i++) + zones_size[i] = 0; + + /* + * Use free_area_init_node instead of free_area_init, because it is + * designed for systems where the DRAM starts at an address + * substantially higher than 0, like us (we start at PAGE_OFFSET). This + * saves space in the mem_map page array. + */ + free_area_init_node(0, &contig_page_data, zones_size, PAGE_OFFSET >> PAGE_SHIFT, 0); + + mem_map = contig_page_data.node_mem_map; +} diff --git a/arch/cris/arch-v32/mm/intmem.c b/arch/cris/arch-v32/mm/intmem.c new file mode 100644 index 000000000000..41ee7f7997fd --- /dev/null +++ b/arch/cris/arch-v32/mm/intmem.c @@ -0,0 +1,139 @@ +/* + * Simple allocator for internal RAM in ETRAX FS + * + * Copyright (c) 2004 Axis Communications AB. + */ + +#include <linux/list.h> +#include <linux/slab.h> +#include <asm/io.h> +#include <asm/arch/memmap.h> + +#define STATUS_FREE 0 +#define STATUS_ALLOCATED 1 + +struct intmem_allocation { + struct list_head entry; + unsigned int size; + unsigned offset; + char status; +}; + + +static struct list_head intmem_allocations; +static void* intmem_virtual; + +static void crisv32_intmem_init(void) +{ + static int initiated = 0; + if (!initiated) { + struct intmem_allocation* alloc = + (struct intmem_allocation*)kmalloc(sizeof *alloc, GFP_KERNEL); + INIT_LIST_HEAD(&intmem_allocations); + intmem_virtual = ioremap(MEM_INTMEM_START, MEM_INTMEM_SIZE); + initiated = 1; + alloc->size = MEM_INTMEM_SIZE; + alloc->offset = 0; + alloc->status = STATUS_FREE; + list_add_tail(&alloc->entry, &intmem_allocations); + } +} + +void* crisv32_intmem_alloc(unsigned size, unsigned align) +{ + struct intmem_allocation* allocation; + struct intmem_allocation* tmp; + void* ret = NULL; + + preempt_disable(); + crisv32_intmem_init(); + + list_for_each_entry_safe(allocation, tmp, &intmem_allocations, entry) { + int alignment = allocation->offset % align; + alignment = alignment ? align - alignment : alignment; + + if (allocation->status == STATUS_FREE && + allocation->size >= size + alignment) { + if (allocation->size > size + alignment) { + struct intmem_allocation* alloc = + (struct intmem_allocation*) + kmalloc(sizeof *alloc, GFP_ATOMIC); + alloc->status = STATUS_FREE; + alloc->size = allocation->size - size - alignment; + alloc->offset = allocation->offset + size; + list_add(&alloc->entry, &allocation->entry); + + if (alignment) { + struct intmem_allocation* tmp; + tmp = (struct intmem_allocation*) + kmalloc(sizeof *tmp, GFP_ATOMIC); + tmp->offset = allocation->offset; + tmp->size = alignment; + tmp->status = STATUS_FREE; + allocation->offset += alignment; + list_add_tail(&tmp->entry, &allocation->entry); + } + } + allocation->status = STATUS_ALLOCATED; + allocation->size = size; + ret = (void*)((int)intmem_virtual + allocation->offset); + } + } + preempt_enable(); + return ret; +} + +void crisv32_intmem_free(void* addr) +{ + struct intmem_allocation* allocation; + struct intmem_allocation* tmp; + + if (addr == NULL) + return; + + preempt_disable(); + crisv32_intmem_init(); + + list_for_each_entry_safe(allocation, tmp, &intmem_allocations, entry) { + if (allocation->offset == (int)(addr - intmem_virtual)) { + struct intmem_allocation* prev = + list_entry(allocation->entry.prev, + struct intmem_allocation, entry); + struct intmem_allocation* next = + list_entry(allocation->entry.next, + struct intmem_allocation, entry); + + allocation->status = STATUS_FREE; + /* Join with prev and/or next if also free */ + if (prev->status == STATUS_FREE) { + prev->size += allocation->size; + list_del(&allocation->entry); + kfree(allocation); + allocation = prev; + } + if (next->status == STATUS_FREE) { + allocation->size += next->size; + list_del(&next->entry); + kfree(next); + } + preempt_enable(); + return; + } + } + preempt_enable(); +} + +void* crisv32_intmem_phys_to_virt(unsigned long addr) +{ + return (void*)(addr - MEM_INTMEM_START+ + (unsigned long)intmem_virtual); +} + +unsigned long crisv32_intmem_virt_to_phys(void* addr) +{ + return (unsigned long)((unsigned long )addr - + (unsigned long)intmem_virtual + MEM_INTMEM_START); +} + + + diff --git a/arch/cris/arch-v32/mm/mmu.S b/arch/cris/arch-v32/mm/mmu.S new file mode 100644 index 000000000000..27b70e5006af --- /dev/null +++ b/arch/cris/arch-v32/mm/mmu.S @@ -0,0 +1,141 @@ +/* + * Copyright (C) 2003 Axis Communications AB + * + * Authors: Mikael Starvik (starvik@axis.com) + * + * Code for the fault low-level handling routines. + * + */ + +#include <asm/page.h> +#include <asm/pgtable.h> + +; Save all register. Must save in same order as struct pt_regs. +.macro SAVE_ALL + subq 12, $sp + move $erp, [$sp] + subq 4, $sp + move $srp, [$sp] + subq 4, $sp + move $ccs, [$sp] + subq 4, $sp + move $spc, [$sp] + subq 4, $sp + move $mof, [$sp] + subq 4, $sp + move $srs, [$sp] + subq 4, $sp + move.d $acr, [$sp] + subq 14*4, $sp + movem $r13, [$sp] + subq 4, $sp + move.d $r10, [$sp] +.endm + +; Bus fault handler. Extracts relevant information and calls mm subsystem +; to handle the fault. +.macro MMU_BUS_FAULT_HANDLER handler, mmu, we, ex + .globl \handler +\handler: + SAVE_ALL + move \mmu, $srs ; Select MMU support register bank + move.d $sp, $r11 ; regs + moveq 1, $r12 ; protection fault + moveq \we, $r13 ; write exception? + orq \ex << 1, $r13 ; execute? + move $s3, $r10 ; rw_mm_cause + and.d ~8191, $r10 ; Get faulting page start address + + jsr do_page_fault + nop + ba ret_from_intr + nop +.endm + +; Refill handler. Three cases may occur: +; 1. PMD and PTE exists in mm subsystem but not in TLB +; 2. PMD exists but not PTE +; 3. PMD doesn't exist +; The code below handles case 1 and calls the mm subsystem for case 2 and 3. +; Do not touch this code without very good reasons and extensive testing. +; Note that the code is optimized to minimize stalls (makes the code harder +; to read). +; +; Each page is 8 KB. Each PMD holds 8192/4 PTEs (each PTE is 4 bytes) so each +; PMD holds 16 MB of virtual memory. +; Bits 0-12 : Offset within a page +; Bits 13-23 : PTE offset within a PMD +; Bits 24-31 : PMD offset within the PGD + +.macro MMU_REFILL_HANDLER handler, mmu + .globl \handler +\handler: + subq 4, $sp +; (The pipeline stalls for one cycle; $sp used as address in the next cycle.) + move $srs, [$sp] + subq 4, $sp + move \mmu, $srs ; Select MMU support register bank + move.d $acr, [$sp] + subq 4, $sp + move.d $r0, [$sp] +#ifdef CONFIG_SMP + move $s7, $acr ; PGD +#else + move.d per_cpu__current_pgd, $acr ; PGD +#endif + ; Look up PMD in PGD + move $s3, $r0 ; rw_mm_cause + lsrq 24, $r0 ; Get PMD index into PGD (bit 24-31) + move.d [$acr], $acr ; PGD for the current process + addi $r0.d, $acr, $acr + move $s3, $r0 ; rw_mm_cause + move.d [$acr], $acr ; Get PMD + beq 1f + ; Look up PTE in PMD + lsrq PAGE_SHIFT, $r0 + and.w PAGE_MASK, $acr ; Remove PMD flags + and.d 0x7ff, $r0 ; Get PTE index into PMD (bit 13-23) + addi $r0.d, $acr, $acr + move.d [$acr], $acr ; Get PTE + beq 2f + move.d [$sp+], $r0 ; Pop r0 in delayslot + ; Store in TLB + move $acr, $s5 + ; Return + move.d [$sp+], $acr + move [$sp], $srs + addq 4, $sp + rete + rfe +1: ; PMD missing, let the mm subsystem fix it up. + move.d [$sp+], $r0 ; Pop r0 +2: ; PTE missing, let the mm subsystem fix it up. + move.d [$sp+], $acr + move [$sp], $srs + addq 4, $sp + SAVE_ALL + move \mmu, $srs + move.d $sp, $r11 ; regs + clear.d $r12 ; Not a protection fault + move.w PAGE_MASK, $acr + move $s3, $r10 ; rw_mm_cause + btstq 9, $r10 ; Check if write access + smi $r13 + and.w PAGE_MASK, $r10 ; Get VPN (virtual address) + jsr do_page_fault + and.w $acr, $r10 + ; Return + ba ret_from_intr + nop +.endm + + ; This is the MMU bus fault handlers. + +MMU_REFILL_HANDLER i_mmu_refill, 1 +MMU_BUS_FAULT_HANDLER i_mmu_invalid, 1, 0, 0 +MMU_BUS_FAULT_HANDLER i_mmu_access, 1, 0, 0 +MMU_BUS_FAULT_HANDLER i_mmu_execute, 1, 0, 1 +MMU_REFILL_HANDLER d_mmu_refill, 2 +MMU_BUS_FAULT_HANDLER d_mmu_invalid, 2, 0, 0 +MMU_BUS_FAULT_HANDLER d_mmu_access, 2, 0, 0 +MMU_BUS_FAULT_HANDLER d_mmu_write, 2, 1, 0 diff --git a/arch/cris/arch-v32/mm/tlb.c b/arch/cris/arch-v32/mm/tlb.c new file mode 100644 index 000000000000..8233406798d3 --- /dev/null +++ b/arch/cris/arch-v32/mm/tlb.c @@ -0,0 +1,208 @@ +/* + * Low level TLB handling. + * + * Copyright (C) 2000-2003, Axis Communications AB. + * + * Authors: Bjorn Wesen <bjornw@axis.com> + * Tobias Anderberg <tobiasa@axis.com>, CRISv32 port. + */ + +#include <asm/tlb.h> +#include <asm/mmu_context.h> +#include <asm/arch/hwregs/asm/mmu_defs_asm.h> +#include <asm/arch/hwregs/supp_reg.h> + +#define UPDATE_TLB_SEL_IDX(val) \ +do { \ + unsigned long tlb_sel; \ + \ + tlb_sel = REG_FIELD(mmu, rw_mm_tlb_sel, idx, val); \ + SUPP_REG_WR(RW_MM_TLB_SEL, tlb_sel); \ +} while(0) + +#define UPDATE_TLB_HILO(tlb_hi, tlb_lo) \ +do { \ + SUPP_REG_WR(RW_MM_TLB_HI, tlb_hi); \ + SUPP_REG_WR(RW_MM_TLB_LO, tlb_lo); \ +} while(0) + +/* + * The TLB can host up to 256 different mm contexts at the same time. The running + * context is found in the PID register. Each TLB entry contains a page_id that + * has to match the PID register to give a hit. page_id_map keeps track of which + * mm's is assigned to which page_id's, making sure it's known when to + * invalidate TLB entries. + * + * The last page_id is never running, it is used as an invalid page_id so that + * it's possible to make TLB entries that will nerver match. + * + * Note; the flushes needs to be atomic otherwise an interrupt hander that uses + * vmalloc'ed memory might cause a TLB load in the middle of a flush. + */ + +/* Flush all TLB entries. */ +void +__flush_tlb_all(void) +{ + int i; + int mmu; + unsigned long flags; + unsigned long mmu_tlb_hi; + unsigned long mmu_tlb_sel; + + /* + * Mask with 0xf so similar TLB entries aren't written in the same 4-way + * entry group. + */ + local_save_flags(flags); + local_irq_disable(); + + for (mmu = 1; mmu <= 2; mmu++) { + SUPP_BANK_SEL(mmu); /* Select the MMU */ + for (i = 0; i < NUM_TLB_ENTRIES; i++) { + /* Store invalid entry */ + mmu_tlb_sel = REG_FIELD(mmu, rw_mm_tlb_sel, idx, i); + + mmu_tlb_hi = (REG_FIELD(mmu, rw_mm_tlb_hi, pid, INVALID_PAGEID) + | REG_FIELD(mmu, rw_mm_tlb_hi, vpn, i & 0xf)); + + SUPP_REG_WR(RW_MM_TLB_SEL, mmu_tlb_sel); + SUPP_REG_WR(RW_MM_TLB_HI, mmu_tlb_hi); + SUPP_REG_WR(RW_MM_TLB_LO, 0); + } + } + + local_irq_restore(flags); +} + +/* Flush an entire user address space. */ +void +__flush_tlb_mm(struct mm_struct *mm) +{ + int i; + int mmu; + unsigned long flags; + unsigned long page_id; + unsigned long tlb_hi; + unsigned long mmu_tlb_hi; + + page_id = mm->context.page_id; + + if (page_id == NO_CONTEXT) + return; + + /* Mark the TLB entries that match the page_id as invalid. */ + local_save_flags(flags); + local_irq_disable(); + + for (mmu = 1; mmu <= 2; mmu++) { + SUPP_BANK_SEL(mmu); + for (i = 0; i < NUM_TLB_ENTRIES; i++) { + UPDATE_TLB_SEL_IDX(i); + + /* Get the page_id */ + SUPP_REG_RD(RW_MM_TLB_HI, tlb_hi); + + /* Check if the page_id match. */ + if ((tlb_hi & 0xff) == page_id) { + mmu_tlb_hi = (REG_FIELD(mmu, rw_mm_tlb_hi, pid, + INVALID_PAGEID) + | REG_FIELD(mmu, rw_mm_tlb_hi, vpn, + i & 0xf)); + + UPDATE_TLB_HILO(mmu_tlb_hi, 0); + } + } + } + + local_irq_restore(flags); +} + +/* Invalidate a single page. */ +void +__flush_tlb_page(struct vm_area_struct *vma, unsigned long addr) +{ + int i; + int mmu; + unsigned long page_id; + unsigned long flags; + unsigned long tlb_hi; + unsigned long mmu_tlb_hi; + + page_id = vma->vm_mm->context.page_id; + + if (page_id == NO_CONTEXT) + return; + + addr &= PAGE_MASK; + + /* + * Invalidate those TLB entries that match both the mm context and the + * requested virtual address. + */ + local_save_flags(flags); + local_irq_disable(); + + for (mmu = 1; mmu <= 2; mmu++) { + SUPP_BANK_SEL(mmu); + for (i = 0; i < NUM_TLB_ENTRIES; i++) { + UPDATE_TLB_SEL_IDX(i); + SUPP_REG_RD(RW_MM_TLB_HI, tlb_hi); + + /* Check if page_id and address matches */ + if (((tlb_hi & 0xff) == page_id) && + ((tlb_hi & PAGE_MASK) == addr)) { + mmu_tlb_hi = REG_FIELD(mmu, rw_mm_tlb_hi, pid, + INVALID_PAGEID) | addr; + + UPDATE_TLB_HILO(mmu_tlb_hi, 0); + } + } + } + + local_irq_restore(flags); +} + +/* + * Initialize the context related info for a new mm_struct + * instance. + */ + +int +init_new_context(struct task_struct *tsk, struct mm_struct *mm) +{ + mm->context.page_id = NO_CONTEXT; + return 0; +} + +/* Called in schedule() just before actually doing the switch_to. */ +void +switch_mm(struct mm_struct *prev, struct mm_struct *next, + struct task_struct *tsk) +{ + int cpu = smp_processor_id(); + + /* Make sure there is a MMU context. */ + spin_lock(&next->page_table_lock); + get_mmu_context(next); + cpu_set(cpu, next->cpu_vm_mask); + spin_unlock(&next->page_table_lock); + + /* + * Remember the pgd for the fault handlers. Keep a seperate copy of it + * because current and active_mm might be invalid at points where + * there's still a need to derefer the pgd. + */ + per_cpu(current_pgd, cpu) = next->pgd; + + /* Switch context in the MMU. */ + if (tsk && tsk->thread_info) + { + SPEC_REG_WR(SPEC_REG_PID, next->context.page_id | tsk->thread_info->tls); + } + else + { + SPEC_REG_WR(SPEC_REG_PID, next->context.page_id); + } +} + diff --git a/arch/cris/arch-v32/output_arch.ld b/arch/cris/arch-v32/output_arch.ld new file mode 100644 index 000000000000..d60a57db0ec2 --- /dev/null +++ b/arch/cris/arch-v32/output_arch.ld @@ -0,0 +1,2 @@ +/* At the time of this writing, there's no equivalent ld option. */ +OUTPUT_ARCH (crisv32) diff --git a/arch/cris/arch-v32/vmlinux.lds.S b/arch/cris/arch-v32/vmlinux.lds.S new file mode 100644 index 000000000000..adb94605d92a --- /dev/null +++ b/arch/cris/arch-v32/vmlinux.lds.S @@ -0,0 +1,134 @@ +/* ld script to make the Linux/CRIS kernel + * Authors: Bjorn Wesen (bjornw@axis.com) + * + * It is VERY DANGEROUS to fiddle around with the symbols in this + * script. It is for example quite vital that all generated sections + * that are used are actually named here, otherwise the linker will + * put them at the end, where the init stuff is which is FREED after + * the kernel has booted. + */ + +#include <linux/config.h> +#include <asm-generic/vmlinux.lds.h> + +jiffies = jiffies_64; +SECTIONS +{ + . = DRAM_VIRTUAL_BASE; + dram_start = .; + ebp_start = .; + + /* The boot section is only necessary until the VCS top level testbench */ + /* includes both flash and DRAM. */ + .boot : { *(.boot) } + + . = DRAM_VIRTUAL_BASE + 0x4000; /* See head.S and pages reserved at the start. */ + + _text = .; /* Text and read-only data. */ + text_start = .; /* Lots of aliases. */ + _stext = .; + __stext = .; + .text : { + *(.text) + SCHED_TEXT + LOCK_TEXT + *(.fixup) + *(.text.__*) + } + + _etext = . ; /* End of text section. */ + __etext = .; + + . = ALIGN(4); /* Exception table. */ + __start___ex_table = .; + __ex_table : { *(__ex_table) } + __stop___ex_table = .; + + RODATA + + . = ALIGN (4); + ___data_start = . ; + __Sdata = . ; + .data : { /* Data */ + *(.data) + } + __edata = . ; /* End of data section. */ + _edata = . ; + + . = ALIGN(8192); /* init_task and stack, must be aligned. */ + .data.init_task : { *(.data.init_task) } + + . = ALIGN(8192); /* Init code and data. */ + __init_begin = .; + .init.text : { + _sinittext = .; + *(.init.text) + _einittext = .; + } + .init.data : { *(.init.data) } + . = ALIGN(16); + __setup_start = .; + .init.setup : { *(.init.setup) } + __setup_end = .; + __start___param = .; + __param : { *(__param) } + __stop___param = .; + .initcall.init : { + __initcall_start = .; + *(.initcall1.init); + *(.initcall2.init); + *(.initcall3.init); + *(.initcall4.init); + *(.initcall5.init); + *(.initcall6.init); + *(.initcall7.init); + __initcall_end = .; + } + + .con_initcall.init : { + __con_initcall_start = .; + *(.con_initcall.init) + __con_initcall_end = .; + } + SECURITY_INIT + + __per_cpu_start = .; + .data.percpu : { *(.data.percpu) } + __per_cpu_end = .; + + .init.ramfs : { + __initramfs_start = .; + *(.init.ramfs) + __initramfs_end = .; + /* + * We fill to the next page, so we can discard all init + * pages without needing to consider what payload might be + * appended to the kernel image. + */ + FILL (0); + . = ALIGN (8192); + } + + __vmlinux_end = .; /* Last address of the physical file. */ + __init_end = .; + + __data_end = . ; /* Move to _edata? */ + __bss_start = .; /* BSS. */ + .bss : { + *(COMMON) + *(.bss) + } + + . = ALIGN (0x20); + _end = .; + __end = .; + + /* Sections to be discarded */ + /DISCARD/ : { + *(.text.exit) + *(.data.exit) + *(.exitcall.exit) + } + + dram_end = dram_start + CONFIG_ETRAX_DRAM_SIZE*1024*1024; +} |