diff options
author | 2022-01-10 08:49:37 -0800 | |
---|---|---|
committer | 2022-01-10 08:49:37 -0800 | |
commit | 9b9e211360044c12d7738973c944d6f300134881 (patch) | |
tree | f994a379135e70397df3c653e6578ad7e5d295fe /tools/testing/selftests/arm64/abi | |
parent | Merge tag 'asm-generic-5.17' of git://git.kernel.org/pub/scm/linux/kernel/git... (diff) | |
parent | Merge branches 'for-next/misc', 'for-next/cache-ops-dzp', 'for-next/stacktrac... (diff) | |
download | linux-9b9e211360044c12d7738973c944d6f300134881.tar.gz linux-9b9e211360044c12d7738973c944d6f300134881.tar.bz2 linux-9b9e211360044c12d7738973c944d6f300134881.zip |
Merge tag 'arm64-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Pull arm64 updates from Catalin Marinas:
- KCSAN enabled for arm64.
- Additional kselftests to exercise the syscall ABI w.r.t. SVE/FPSIMD.
- Some more SVE clean-ups and refactoring in preparation for SME
support (scalable matrix extensions).
- BTI clean-ups (SYM_FUNC macros etc.)
- arm64 atomics clean-up and codegen improvements.
- HWCAPs for FEAT_AFP (alternate floating point behaviour) and
FEAT_RPRESS (increased precision of reciprocal estimate and
reciprocal square root estimate).
- Use SHA3 instructions to speed-up XOR.
- arm64 unwind code refactoring/unification.
- Avoid DC (data cache maintenance) instructions when DCZID_EL0.DZP ==
1 (potentially set by a hypervisor; user-space already does this).
- Perf updates for arm64: support for CI-700, HiSilicon PCIe PMU,
Marvell CN10K LLC-TAD PMU, miscellaneous clean-ups.
- Other fixes and clean-ups; highlights: fix the handling of erratum
1418040, correct the calculation of the nomap region boundaries,
introduce io_stop_wc() mapped to the new DGH instruction (data
gathering hint).
* tag 'arm64-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux: (81 commits)
arm64: Use correct method to calculate nomap region boundaries
arm64: Drop outdated links in comments
arm64: perf: Don't register user access sysctl handler multiple times
drivers: perf: marvell_cn10k: fix an IS_ERR() vs NULL check
perf/smmuv3: Fix unused variable warning when CONFIG_OF=n
arm64: errata: Fix exec handling in erratum 1418040 workaround
arm64: Unhash early pointer print plus improve comment
asm-generic: introduce io_stop_wc() and add implementation for ARM64
arm64: Ensure that the 'bti' macro is defined where linkage.h is included
arm64: remove __dma_*_area() aliases
docs/arm64: delete a space from tagged-address-abi
arm64: Enable KCSAN
kselftest/arm64: Add pidbench for floating point syscall cases
arm64/fp: Add comments documenting the usage of state restore functions
kselftest/arm64: Add a test program to exercise the syscall ABI
kselftest/arm64: Allow signal tests to trigger from a function
kselftest/arm64: Parameterise ptrace vector length information
arm64/sve: Minor clarification of ABI documentation
arm64/sve: Generalise vector length configuration prctl() for SME
arm64/sve: Make sysctl interface for SVE reusable by SME
...
Diffstat (limited to 'tools/testing/selftests/arm64/abi')
-rw-r--r-- | tools/testing/selftests/arm64/abi/.gitignore | 1 | ||||
-rw-r--r-- | tools/testing/selftests/arm64/abi/Makefile | 8 | ||||
-rw-r--r-- | tools/testing/selftests/arm64/abi/syscall-abi-asm.S | 240 | ||||
-rw-r--r-- | tools/testing/selftests/arm64/abi/syscall-abi.c | 318 |
4 files changed, 567 insertions, 0 deletions
diff --git a/tools/testing/selftests/arm64/abi/.gitignore b/tools/testing/selftests/arm64/abi/.gitignore new file mode 100644 index 000000000000..b79cf5814c23 --- /dev/null +++ b/tools/testing/selftests/arm64/abi/.gitignore @@ -0,0 +1 @@ +syscall-abi diff --git a/tools/testing/selftests/arm64/abi/Makefile b/tools/testing/selftests/arm64/abi/Makefile new file mode 100644 index 000000000000..96eba974ac8d --- /dev/null +++ b/tools/testing/selftests/arm64/abi/Makefile @@ -0,0 +1,8 @@ +# SPDX-License-Identifier: GPL-2.0 +# Copyright (C) 2021 ARM Limited + +TEST_GEN_PROGS := syscall-abi + +include ../../lib.mk + +$(OUTPUT)/syscall-abi: syscall-abi.c syscall-abi-asm.S diff --git a/tools/testing/selftests/arm64/abi/syscall-abi-asm.S b/tools/testing/selftests/arm64/abi/syscall-abi-asm.S new file mode 100644 index 000000000000..983467cfcee0 --- /dev/null +++ b/tools/testing/selftests/arm64/abi/syscall-abi-asm.S @@ -0,0 +1,240 @@ +// SPDX-License-Identifier: GPL-2.0-only +// Copyright (C) 2021 ARM Limited. +// +// Assembly portion of the syscall ABI test + +// +// Load values from memory into registers, invoke a syscall and save the +// register values back to memory for later checking. The syscall to be +// invoked is configured in x8 of the input GPR data. +// +// x0: SVE VL, 0 for FP only +// +// GPRs: gpr_in, gpr_out +// FPRs: fpr_in, fpr_out +// Zn: z_in, z_out +// Pn: p_in, p_out +// FFR: ffr_in, ffr_out + +.arch_extension sve + +.globl do_syscall +do_syscall: + // Store callee saved registers x19-x29 (80 bytes) plus x0 and x1 + stp x29, x30, [sp, #-112]! + mov x29, sp + stp x0, x1, [sp, #16] + stp x19, x20, [sp, #32] + stp x21, x22, [sp, #48] + stp x23, x24, [sp, #64] + stp x25, x26, [sp, #80] + stp x27, x28, [sp, #96] + + // Load GPRs x8-x28, and save our SP/FP for later comparison + ldr x2, =gpr_in + add x2, x2, #64 + ldp x8, x9, [x2], #16 + ldp x10, x11, [x2], #16 + ldp x12, x13, [x2], #16 + ldp x14, x15, [x2], #16 + ldp x16, x17, [x2], #16 + ldp x18, x19, [x2], #16 + ldp x20, x21, [x2], #16 + ldp x22, x23, [x2], #16 + ldp x24, x25, [x2], #16 + ldp x26, x27, [x2], #16 + ldr x28, [x2], #8 + str x29, [x2], #8 // FP + str x30, [x2], #8 // LR + + // Load FPRs if we're not doing SVE + cbnz x0, 1f + ldr x2, =fpr_in + ldp q0, q1, [x2] + ldp q2, q3, [x2, #16 * 2] + ldp q4, q5, [x2, #16 * 4] + ldp q6, q7, [x2, #16 * 6] + ldp q8, q9, [x2, #16 * 8] + ldp q10, q11, [x2, #16 * 10] + ldp q12, q13, [x2, #16 * 12] + ldp q14, q15, [x2, #16 * 14] + ldp q16, q17, [x2, #16 * 16] + ldp q18, q19, [x2, #16 * 18] + ldp q20, q21, [x2, #16 * 20] + ldp q22, q23, [x2, #16 * 22] + ldp q24, q25, [x2, #16 * 24] + ldp q26, q27, [x2, #16 * 26] + ldp q28, q29, [x2, #16 * 28] + ldp q30, q31, [x2, #16 * 30] +1: + + // Load the SVE registers if we're doing SVE + cbz x0, 1f + + ldr x2, =z_in + ldr z0, [x2, #0, MUL VL] + ldr z1, [x2, #1, MUL VL] + ldr z2, [x2, #2, MUL VL] + ldr z3, [x2, #3, MUL VL] + ldr z4, [x2, #4, MUL VL] + ldr z5, [x2, #5, MUL VL] + ldr z6, [x2, #6, MUL VL] + ldr z7, [x2, #7, MUL VL] + ldr z8, [x2, #8, MUL VL] + ldr z9, [x2, #9, MUL VL] + ldr z10, [x2, #10, MUL VL] + ldr z11, [x2, #11, MUL VL] + ldr z12, [x2, #12, MUL VL] + ldr z13, [x2, #13, MUL VL] + ldr z14, [x2, #14, MUL VL] + ldr z15, [x2, #15, MUL VL] + ldr z16, [x2, #16, MUL VL] + ldr z17, [x2, #17, MUL VL] + ldr z18, [x2, #18, MUL VL] + ldr z19, [x2, #19, MUL VL] + ldr z20, [x2, #20, MUL VL] + ldr z21, [x2, #21, MUL VL] + ldr z22, [x2, #22, MUL VL] + ldr z23, [x2, #23, MUL VL] + ldr z24, [x2, #24, MUL VL] + ldr z25, [x2, #25, MUL VL] + ldr z26, [x2, #26, MUL VL] + ldr z27, [x2, #27, MUL VL] + ldr z28, [x2, #28, MUL VL] + ldr z29, [x2, #29, MUL VL] + ldr z30, [x2, #30, MUL VL] + ldr z31, [x2, #31, MUL VL] + + ldr x2, =ffr_in + ldr p0, [x2, #0] + wrffr p0.b + + ldr x2, =p_in + ldr p0, [x2, #0, MUL VL] + ldr p1, [x2, #1, MUL VL] + ldr p2, [x2, #2, MUL VL] + ldr p3, [x2, #3, MUL VL] + ldr p4, [x2, #4, MUL VL] + ldr p5, [x2, #5, MUL VL] + ldr p6, [x2, #6, MUL VL] + ldr p7, [x2, #7, MUL VL] + ldr p8, [x2, #8, MUL VL] + ldr p9, [x2, #9, MUL VL] + ldr p10, [x2, #10, MUL VL] + ldr p11, [x2, #11, MUL VL] + ldr p12, [x2, #12, MUL VL] + ldr p13, [x2, #13, MUL VL] + ldr p14, [x2, #14, MUL VL] + ldr p15, [x2, #15, MUL VL] +1: + + // Do the syscall + svc #0 + + // Save GPRs x8-x30 + ldr x2, =gpr_out + add x2, x2, #64 + stp x8, x9, [x2], #16 + stp x10, x11, [x2], #16 + stp x12, x13, [x2], #16 + stp x14, x15, [x2], #16 + stp x16, x17, [x2], #16 + stp x18, x19, [x2], #16 + stp x20, x21, [x2], #16 + stp x22, x23, [x2], #16 + stp x24, x25, [x2], #16 + stp x26, x27, [x2], #16 + stp x28, x29, [x2], #16 + str x30, [x2] + + // Restore x0 and x1 for feature checks + ldp x0, x1, [sp, #16] + + // Save FPSIMD state + ldr x2, =fpr_out + stp q0, q1, [x2] + stp q2, q3, [x2, #16 * 2] + stp q4, q5, [x2, #16 * 4] + stp q6, q7, [x2, #16 * 6] + stp q8, q9, [x2, #16 * 8] + stp q10, q11, [x2, #16 * 10] + stp q12, q13, [x2, #16 * 12] + stp q14, q15, [x2, #16 * 14] + stp q16, q17, [x2, #16 * 16] + stp q18, q19, [x2, #16 * 18] + stp q20, q21, [x2, #16 * 20] + stp q22, q23, [x2, #16 * 22] + stp q24, q25, [x2, #16 * 24] + stp q26, q27, [x2, #16 * 26] + stp q28, q29, [x2, #16 * 28] + stp q30, q31, [x2, #16 * 30] + + // Save the SVE state if we have some + cbz x0, 1f + + ldr x2, =z_out + str z0, [x2, #0, MUL VL] + str z1, [x2, #1, MUL VL] + str z2, [x2, #2, MUL VL] + str z3, [x2, #3, MUL VL] + str z4, [x2, #4, MUL VL] + str z5, [x2, #5, MUL VL] + str z6, [x2, #6, MUL VL] + str z7, [x2, #7, MUL VL] + str z8, [x2, #8, MUL VL] + str z9, [x2, #9, MUL VL] + str z10, [x2, #10, MUL VL] + str z11, [x2, #11, MUL VL] + str z12, [x2, #12, MUL VL] + str z13, [x2, #13, MUL VL] + str z14, [x2, #14, MUL VL] + str z15, [x2, #15, MUL VL] + str z16, [x2, #16, MUL VL] + str z17, [x2, #17, MUL VL] + str z18, [x2, #18, MUL VL] + str z19, [x2, #19, MUL VL] + str z20, [x2, #20, MUL VL] + str z21, [x2, #21, MUL VL] + str z22, [x2, #22, MUL VL] + str z23, [x2, #23, MUL VL] + str z24, [x2, #24, MUL VL] + str z25, [x2, #25, MUL VL] + str z26, [x2, #26, MUL VL] + str z27, [x2, #27, MUL VL] + str z28, [x2, #28, MUL VL] + str z29, [x2, #29, MUL VL] + str z30, [x2, #30, MUL VL] + str z31, [x2, #31, MUL VL] + + ldr x2, =p_out + str p0, [x2, #0, MUL VL] + str p1, [x2, #1, MUL VL] + str p2, [x2, #2, MUL VL] + str p3, [x2, #3, MUL VL] + str p4, [x2, #4, MUL VL] + str p5, [x2, #5, MUL VL] + str p6, [x2, #6, MUL VL] + str p7, [x2, #7, MUL VL] + str p8, [x2, #8, MUL VL] + str p9, [x2, #9, MUL VL] + str p10, [x2, #10, MUL VL] + str p11, [x2, #11, MUL VL] + str p12, [x2, #12, MUL VL] + str p13, [x2, #13, MUL VL] + str p14, [x2, #14, MUL VL] + str p15, [x2, #15, MUL VL] + + ldr x2, =ffr_out + rdffr p0.b + str p0, [x2, #0] +1: + + // Restore callee saved registers x19-x30 + ldp x19, x20, [sp, #32] + ldp x21, x22, [sp, #48] + ldp x23, x24, [sp, #64] + ldp x25, x26, [sp, #80] + ldp x27, x28, [sp, #96] + ldp x29, x30, [sp], #112 + + ret diff --git a/tools/testing/selftests/arm64/abi/syscall-abi.c b/tools/testing/selftests/arm64/abi/syscall-abi.c new file mode 100644 index 000000000000..d8eeeafb50dc --- /dev/null +++ b/tools/testing/selftests/arm64/abi/syscall-abi.c @@ -0,0 +1,318 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Copyright (C) 2021 ARM Limited. + */ + +#include <errno.h> +#include <stdbool.h> +#include <stddef.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <unistd.h> +#include <sys/auxv.h> +#include <sys/prctl.h> +#include <asm/hwcap.h> +#include <asm/sigcontext.h> +#include <asm/unistd.h> + +#include "../../kselftest.h" + +#define ARRAY_SIZE(a) (sizeof(a) / sizeof(a[0])) +#define NUM_VL ((SVE_VQ_MAX - SVE_VQ_MIN) + 1) + +extern void do_syscall(int sve_vl); + +static void fill_random(void *buf, size_t size) +{ + int i; + uint32_t *lbuf = buf; + + /* random() returns a 32 bit number regardless of the size of long */ + for (i = 0; i < size / sizeof(uint32_t); i++) + lbuf[i] = random(); +} + +/* + * We also repeat the test for several syscalls to try to expose different + * behaviour. + */ +static struct syscall_cfg { + int syscall_nr; + const char *name; +} syscalls[] = { + { __NR_getpid, "getpid()" }, + { __NR_sched_yield, "sched_yield()" }, +}; + +#define NUM_GPR 31 +uint64_t gpr_in[NUM_GPR]; +uint64_t gpr_out[NUM_GPR]; + +static void setup_gpr(struct syscall_cfg *cfg, int sve_vl) +{ + fill_random(gpr_in, sizeof(gpr_in)); + gpr_in[8] = cfg->syscall_nr; + memset(gpr_out, 0, sizeof(gpr_out)); +} + +static int check_gpr(struct syscall_cfg *cfg, int sve_vl) +{ + int errors = 0; + int i; + + /* + * GPR x0-x7 may be clobbered, and all others should be preserved. + */ + for (i = 9; i < ARRAY_SIZE(gpr_in); i++) { + if (gpr_in[i] != gpr_out[i]) { + ksft_print_msg("%s SVE VL %d mismatch in GPR %d: %llx != %llx\n", + cfg->name, sve_vl, i, + gpr_in[i], gpr_out[i]); + errors++; + } + } + + return errors; +} + +#define NUM_FPR 32 +uint64_t fpr_in[NUM_FPR * 2]; +uint64_t fpr_out[NUM_FPR * 2]; + +static void setup_fpr(struct syscall_cfg *cfg, int sve_vl) +{ + fill_random(fpr_in, sizeof(fpr_in)); + memset(fpr_out, 0, sizeof(fpr_out)); +} + +static int check_fpr(struct syscall_cfg *cfg, int sve_vl) +{ + int errors = 0; + int i; + + if (!sve_vl) { + for (i = 0; i < ARRAY_SIZE(fpr_in); i++) { + if (fpr_in[i] != fpr_out[i]) { + ksft_print_msg("%s Q%d/%d mismatch %llx != %llx\n", + cfg->name, + i / 2, i % 2, + fpr_in[i], fpr_out[i]); + errors++; + } + } + } + + return errors; +} + +static uint8_t z_zero[__SVE_ZREG_SIZE(SVE_VQ_MAX)]; +uint8_t z_in[SVE_NUM_PREGS * __SVE_ZREG_SIZE(SVE_VQ_MAX)]; +uint8_t z_out[SVE_NUM_PREGS * __SVE_ZREG_SIZE(SVE_VQ_MAX)]; + +static void setup_z(struct syscall_cfg *cfg, int sve_vl) +{ + fill_random(z_in, sizeof(z_in)); + fill_random(z_out, sizeof(z_out)); +} + +static int check_z(struct syscall_cfg *cfg, int sve_vl) +{ + size_t reg_size = sve_vl; + int errors = 0; + int i; + + if (!sve_vl) + return 0; + + /* + * After a syscall the low 128 bits of the Z registers should + * be preserved and the rest be zeroed or preserved. + */ + for (i = 0; i < SVE_NUM_ZREGS; i++) { + void *in = &z_in[reg_size * i]; + void *out = &z_out[reg_size * i]; + + if (memcmp(in, out, SVE_VQ_BYTES) != 0) { + ksft_print_msg("%s SVE VL %d Z%d low 128 bits changed\n", + cfg->name, sve_vl, i); + errors++; + } + } + + return errors; +} + +uint8_t p_in[SVE_NUM_PREGS * __SVE_PREG_SIZE(SVE_VQ_MAX)]; +uint8_t p_out[SVE_NUM_PREGS * __SVE_PREG_SIZE(SVE_VQ_MAX)]; + +static void setup_p(struct syscall_cfg *cfg, int sve_vl) +{ + fill_random(p_in, sizeof(p_in)); + fill_random(p_out, sizeof(p_out)); +} + +static int check_p(struct syscall_cfg *cfg, int sve_vl) +{ + size_t reg_size = sve_vq_from_vl(sve_vl) * 2; /* 1 bit per VL byte */ + + int errors = 0; + int i; + + if (!sve_vl) + return 0; + + /* After a syscall the P registers should be preserved or zeroed */ + for (i = 0; i < SVE_NUM_PREGS * reg_size; i++) + if (p_out[i] && (p_in[i] != p_out[i])) + errors++; + if (errors) + ksft_print_msg("%s SVE VL %d predicate registers non-zero\n", + cfg->name, sve_vl); + + return errors; +} + +uint8_t ffr_in[__SVE_PREG_SIZE(SVE_VQ_MAX)]; +uint8_t ffr_out[__SVE_PREG_SIZE(SVE_VQ_MAX)]; + +static void setup_ffr(struct syscall_cfg *cfg, int sve_vl) +{ + /* + * It is only valid to set a contiguous set of bits starting + * at 0. For now since we're expecting this to be cleared by + * a syscall just set all bits. + */ + memset(ffr_in, 0xff, sizeof(ffr_in)); + fill_random(ffr_out, sizeof(ffr_out)); +} + +static int check_ffr(struct syscall_cfg *cfg, int sve_vl) +{ + size_t reg_size = sve_vq_from_vl(sve_vl) * 2; /* 1 bit per VL byte */ + int errors = 0; + int i; + + if (!sve_vl) + return 0; + + /* After a syscall the P registers should be preserved or zeroed */ + for (i = 0; i < reg_size; i++) + if (ffr_out[i] && (ffr_in[i] != ffr_out[i])) + errors++; + if (errors) + ksft_print_msg("%s SVE VL %d FFR non-zero\n", + cfg->name, sve_vl); + + return errors; +} + +typedef void (*setup_fn)(struct syscall_cfg *cfg, int sve_vl); +typedef int (*check_fn)(struct syscall_cfg *cfg, int sve_vl); + +/* + * Each set of registers has a setup function which is called before + * the syscall to fill values in a global variable for loading by the + * test code and a check function which validates that the results are + * as expected. Vector lengths are passed everywhere, a vector length + * of 0 should be treated as do not test. + */ +static struct { + setup_fn setup; + check_fn check; +} regset[] = { + { setup_gpr, check_gpr }, + { setup_fpr, check_fpr }, + { setup_z, check_z }, + { setup_p, check_p }, + { setup_ffr, check_ffr }, +}; + +static bool do_test(struct syscall_cfg *cfg, int sve_vl) +{ + int errors = 0; + int i; + + for (i = 0; i < ARRAY_SIZE(regset); i++) + regset[i].setup(cfg, sve_vl); + + do_syscall(sve_vl); + + for (i = 0; i < ARRAY_SIZE(regset); i++) + errors += regset[i].check(cfg, sve_vl); + + return errors == 0; +} + +static void test_one_syscall(struct syscall_cfg *cfg) +{ + int sve_vq, sve_vl; + + /* FPSIMD only case */ + ksft_test_result(do_test(cfg, 0), + "%s FPSIMD\n", cfg->name); + + if (!(getauxval(AT_HWCAP) & HWCAP_SVE)) + return; + + for (sve_vq = SVE_VQ_MAX; sve_vq > 0; --sve_vq) { + sve_vl = prctl(PR_SVE_SET_VL, sve_vq * 16); + if (sve_vl == -1) + ksft_exit_fail_msg("PR_SVE_SET_VL failed: %s (%d)\n", + strerror(errno), errno); + + sve_vl &= PR_SVE_VL_LEN_MASK; + + if (sve_vq != sve_vq_from_vl(sve_vl)) + sve_vq = sve_vq_from_vl(sve_vl); + + ksft_test_result(do_test(cfg, sve_vl), + "%s SVE VL %d\n", cfg->name, sve_vl); + } +} + +int sve_count_vls(void) +{ + unsigned int vq; + int vl_count = 0; + int vl; + + if (!(getauxval(AT_HWCAP) & HWCAP_SVE)) + return 0; + + /* + * Enumerate up to SVE_VQ_MAX vector lengths + */ + for (vq = SVE_VQ_MAX; vq > 0; --vq) { + vl = prctl(PR_SVE_SET_VL, vq * 16); + if (vl == -1) + ksft_exit_fail_msg("PR_SVE_SET_VL failed: %s (%d)\n", + strerror(errno), errno); + + vl &= PR_SVE_VL_LEN_MASK; + + if (vq != sve_vq_from_vl(vl)) + vq = sve_vq_from_vl(vl); + + vl_count++; + } + + return vl_count; +} + +int main(void) +{ + int i; + + srandom(getpid()); + + ksft_print_header(); + ksft_set_plan(ARRAY_SIZE(syscalls) * (sve_count_vls() + 1)); + + for (i = 0; i < ARRAY_SIZE(syscalls); i++) + test_one_syscall(&syscalls[i]); + + ksft_print_cnts(); + + return 0; +} |