perf intel-pt: Add support for emulated ptwrite

ptwrite is an Intel x86 instruction that writes arbitrary values into an
Intel PT trace. It is not supported on all hardware, so provide an
alternative that makes use of TNT packets to convey the payload data.
TNT packets encode Taken/Not-taken conditional branch information, so
taking branches based on the payload value will encode the value into
the TNT packet. Refer to the changes to the documentation file
perf-intel-pt.txt in this patch for an example.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Link: https://lore.kernel.org/r/20220509152400.376613-2-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

4 files changed, 100 insertions, 2 deletions

diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
index e1d8f7504cbe..0ac860c8dd2b 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
@@ -137,6 +137,7 @@ struct intel_pt_decoder {
 	bool in_psb;
 	bool hop;
 	bool leap;
+	bool emulated_ptwrite;
 	bool vm_time_correlation;
 	bool vm_tm_corr_dry_run;
 	bool vm_tm_corr_reliable;
@@ -481,6 +482,8 @@ static int intel_pt_ext_err(int code)
 		return INTEL_PT_ERR_LOST;
 	case -ELOOP:
 		return INTEL_PT_ERR_NELOOP;
+	case -ECONNRESET:
+		return INTEL_PT_ERR_EPTW;
 	default:
 		return INTEL_PT_ERR_UNK;
 	}
@@ -497,6 +500,7 @@ static const char *intel_pt_err_msgs[] = {
 	[INTEL_PT_ERR_LOST]   = "Lost trace data",
 	[INTEL_PT_ERR_UNK]    = "Unknown error!",
 	[INTEL_PT_ERR_NELOOP] = "Never-ending loop (refer perf config intel-pt.max-loops)",
+	[INTEL_PT_ERR_EPTW]   = "Broken emulated ptwrite",
 };
 
 int intel_pt__strerror(int code, char *buf, size_t buflen)
@@ -1535,17 +1539,108 @@ static int intel_pt_walk_tip(struct intel_pt_decoder *decoder)
 	return intel_pt_bug(decoder);
 }
 
+struct eptw_data {
+	int bit_countdown;
+	uint64_t payload;
+};
+
+static int intel_pt_eptw_lookahead_cb(struct intel_pt_pkt_info *pkt_info)
+{
+	struct eptw_data *data = pkt_info->data;
+	int nr_bits;
+
+	switch (pkt_info->packet.type) {
+	case INTEL_PT_PAD:
+	case INTEL_PT_MNT:
+	case INTEL_PT_MODE_EXEC:
+	case INTEL_PT_MODE_TSX:
+	case INTEL_PT_MTC:
+	case INTEL_PT_FUP:
+	case INTEL_PT_CYC:
+	case INTEL_PT_CBR:
+	case INTEL_PT_TSC:
+	case INTEL_PT_TMA:
+	case INTEL_PT_PIP:
+	case INTEL_PT_VMCS:
+	case INTEL_PT_PSB:
+	case INTEL_PT_PSBEND:
+	case INTEL_PT_PTWRITE:
+	case INTEL_PT_PTWRITE_IP:
+	case INTEL_PT_EXSTOP:
+	case INTEL_PT_EXSTOP_IP:
+	case INTEL_PT_MWAIT:
+	case INTEL_PT_PWRE:
+	case INTEL_PT_PWRX:
+	case INTEL_PT_BBP:
+	case INTEL_PT_BIP:
+	case INTEL_PT_BEP:
+	case INTEL_PT_BEP_IP:
+	case INTEL_PT_CFE:
+	case INTEL_PT_CFE_IP:
+	case INTEL_PT_EVD:
+		break;
+
+	case INTEL_PT_TNT:
+		nr_bits = data->bit_countdown;
+		if (nr_bits > pkt_info->packet.count)
+			nr_bits = pkt_info->packet.count;
+		data->payload <<= nr_bits;
+		data->payload |= pkt_info->packet.payload >> (64 - nr_bits);
+		data->bit_countdown -= nr_bits;
+		return !data->bit_countdown;
+
+	case INTEL_PT_TIP_PGE:
+	case INTEL_PT_TIP_PGD:
+	case INTEL_PT_TIP:
+	case INTEL_PT_BAD:
+	case INTEL_PT_OVF:
+	case INTEL_PT_TRACESTOP:
+	default:
+		return 1;
+	}
+
+	return 0;
+}
+
+static int intel_pt_emulated_ptwrite(struct intel_pt_decoder *decoder)
+{
+	int n = 64 - decoder->tnt.count;
+	struct eptw_data data = {
+		.bit_countdown = n,
+		.payload = decoder->tnt.payload >> n,
+	};
+
+	decoder->emulated_ptwrite = false;
+	intel_pt_log("Emulated ptwrite detected\n");
+
+	intel_pt_pkt_lookahead(decoder, intel_pt_eptw_lookahead_cb, &data);
+	if (data.bit_countdown)
+		return -ECONNRESET;
+
+	decoder->state.type = INTEL_PT_PTW;
+	decoder->state.from_ip = decoder->ip;
+	decoder->state.to_ip = 0;
+	decoder->state.ptw_payload = data.payload;
+	return 0;
+}
+
 static int intel_pt_walk_tnt(struct intel_pt_decoder *decoder)
 {
 	struct intel_pt_insn intel_pt_insn;
 	int err;
 
 	while (1) {
+		if (decoder->emulated_ptwrite)
+			return intel_pt_emulated_ptwrite(decoder);
 		err = intel_pt_walk_insn(decoder, &intel_pt_insn, 0);
-		if (err == INTEL_PT_RETURN)
+		if (err == INTEL_PT_RETURN) {
+			decoder->emulated_ptwrite = intel_pt_insn.emulated_ptwrite;
 			return 0;
-		if (err)
+		}
+		if (err) {
+			decoder->emulated_ptwrite = false;
 			return err;
+		}
 
 		if (intel_pt_insn.op == INTEL_PT_OP_RET) {
 			if (!decoder->return_compression) {
diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h
index efb2cb3ae0ca..c773028df80e 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h
@@ -58,6 +58,7 @@ enum {
 	INTEL_PT_ERR_LOST,
 	INTEL_PT_ERR_UNK,
 	INTEL_PT_ERR_NELOOP,
+	INTEL_PT_ERR_EPTW,
 	INTEL_PT_ERR_MAX,
 };
 
diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-insn-decoder.c b/tools/perf/util/intel-pt-decoder/intel-pt-insn-decoder.c
index 9d5e65cec89b..1376077183f7 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-insn-decoder.c
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-insn-decoder.c
@@ -32,6 +32,7 @@ static void intel_pt_insn_decoder(struct insn *insn,
 	int ext;
 
 	intel_pt_insn->rel = 0;
+	intel_pt_insn->emulated_ptwrite = false;
 
 	if (insn_is_avx(insn)) {
 		intel_pt_insn->op = INTEL_PT_OP_OTHER;
diff --git a/tools/perf/util/intel-pt-decoder/intel-pt-insn-decoder.h b/tools/perf/util/intel-pt-decoder/intel-pt-insn-decoder.h
index c2861cfdd768..e3338b56a75f 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-insn-decoder.h
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-insn-decoder.h
@@ -37,6 +37,7 @@ enum intel_pt_insn_branch {
 struct intel_pt_insn {
 	enum intel_pt_insn_op		op;
 	enum intel_pt_insn_branch	branch;
+	bool				emulated_ptwrite;
 	int				length;
 	int32_t				rel;
 	unsigned char			buf[INTEL_PT_INSN_BUF_SZ];