All of lore.kernel.org
 help / color / mirror / Atom feed
From: Arnaldo Carvalho de Melo <acme@kernel.org>
To: Ingo Molnar <mingo@kernel.org>
Cc: linux-kernel@vger.kernel.org,
	Adrian Hunter <adrian.hunter@intel.com>,
	Jiri Olsa <jolsa@redhat.com>,
	Arnaldo Carvalho de Melo <acme@redhat.com>
Subject: [PATCH 09/22] perf tools: Add Intel PT support for decoding MTC packets
Date: Wed, 26 Aug 2015 12:57:59 -0300	[thread overview]
Message-ID: <1440604692-26918-10-git-send-email-acme@kernel.org> (raw)
In-Reply-To: <1440604692-26918-1-git-send-email-acme@kernel.org>

From: Adrian Hunter <adrian.hunter@intel.com>

MTC packets provide finer grain timestamp information than TSC packets.
MTC packets record time using the hardware crystal clock (CTC) which is
related to TSC packets using a TMA packet.

This patch just adds decoder support.

Support for a default value and validation of values is provided by a
later patch. Also documentation is updated in a separate patch.

For details refer to the June 2015 or later Intel 64 and IA-32
Architectures SDM Chapter 36 Intel Processor Trace.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1437150840-31811-21-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 .../perf/util/intel-pt-decoder/intel-pt-decoder.c  | 162 ++++++++++++++++++++-
 .../perf/util/intel-pt-decoder/intel-pt-decoder.h  |   1 +
 2 files changed, 159 insertions(+), 4 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 4a0e9fb1d173..f7119a11a4b6 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.c
@@ -85,7 +85,9 @@ struct intel_pt_decoder {
 	const unsigned char *buf;
 	size_t len;
 	bool return_compression;
+	bool mtc_insn;
 	bool pge;
+	bool have_tma;
 	uint64_t pos;
 	uint64_t last_ip;
 	uint64_t ip;
@@ -94,6 +96,15 @@ struct intel_pt_decoder {
 	uint64_t tsc_timestamp;
 	uint64_t ref_timestamp;
 	uint64_t ret_addr;
+	uint64_t ctc_timestamp;
+	uint64_t ctc_delta;
+	uint32_t last_mtc;
+	uint32_t tsc_ctc_ratio_n;
+	uint32_t tsc_ctc_ratio_d;
+	uint32_t tsc_ctc_mult;
+	uint32_t tsc_slip;
+	uint32_t ctc_rem_mask;
+	int mtc_shift;
 	struct intel_pt_stack stack;
 	enum intel_pt_pkt_state pkt_state;
 	struct intel_pt_pkt packet;
@@ -149,6 +160,13 @@ static void intel_pt_setup_period(struct intel_pt_decoder *decoder)
 	}
 }
 
+static uint64_t multdiv(uint64_t t, uint32_t n, uint32_t d)
+{
+	if (!d)
+		return 0;
+	return (t / d) * n + ((t % d) * n) / d;
+}
+
 struct intel_pt_decoder *intel_pt_decoder_new(struct intel_pt_params *params)
 {
 	struct intel_pt_decoder *decoder;
@@ -175,6 +193,39 @@ struct intel_pt_decoder *intel_pt_decoder_new(struct intel_pt_params *params)
 
 	intel_pt_setup_period(decoder);
 
+	decoder->mtc_shift = params->mtc_period;
+	decoder->ctc_rem_mask = (1 << decoder->mtc_shift) - 1;
+
+	decoder->tsc_ctc_ratio_n = params->tsc_ctc_ratio_n;
+	decoder->tsc_ctc_ratio_d = params->tsc_ctc_ratio_d;
+
+	if (!decoder->tsc_ctc_ratio_n)
+		decoder->tsc_ctc_ratio_d = 0;
+
+	if (decoder->tsc_ctc_ratio_d) {
+		if (!(decoder->tsc_ctc_ratio_n % decoder->tsc_ctc_ratio_d))
+			decoder->tsc_ctc_mult = decoder->tsc_ctc_ratio_n /
+						decoder->tsc_ctc_ratio_d;
+
+		/*
+		 * Allow for timestamps appearing to backwards because a TSC
+		 * packet has slipped past a MTC packet, so allow 2 MTC ticks
+		 * or ...
+		 */
+		decoder->tsc_slip = multdiv(2 << decoder->mtc_shift,
+					decoder->tsc_ctc_ratio_n,
+					decoder->tsc_ctc_ratio_d);
+	}
+	/* ... or 0x100 paranoia */
+	if (decoder->tsc_slip < 0x100)
+		decoder->tsc_slip = 0x100;
+
+	intel_pt_log("timestamp: mtc_shift %u\n", decoder->mtc_shift);
+	intel_pt_log("timestamp: tsc_ctc_ratio_n %u\n", decoder->tsc_ctc_ratio_n);
+	intel_pt_log("timestamp: tsc_ctc_ratio_d %u\n", decoder->tsc_ctc_ratio_d);
+	intel_pt_log("timestamp: tsc_ctc_mult %u\n", decoder->tsc_ctc_mult);
+	intel_pt_log("timestamp: tsc_slip %#x\n", decoder->tsc_slip);
+
 	return decoder;
 }
 
@@ -368,6 +419,7 @@ static inline void intel_pt_update_in_tx(struct intel_pt_decoder *decoder)
 static int intel_pt_bad_packet(struct intel_pt_decoder *decoder)
 {
 	intel_pt_clear_tx_flags(decoder);
+	decoder->have_tma = false;
 	decoder->pkt_len = 1;
 	decoder->pkt_step = 1;
 	intel_pt_decoder_log_packet(decoder);
@@ -400,6 +452,7 @@ static int intel_pt_get_data(struct intel_pt_decoder *decoder)
 		decoder->pkt_state = INTEL_PT_STATE_NO_PSB;
 		decoder->ref_timestamp = buffer.ref_timestamp;
 		decoder->timestamp = 0;
+		decoder->have_tma = false;
 		decoder->state.trace_nr = buffer.trace_nr;
 		intel_pt_log("Reference timestamp 0x%" PRIx64 "\n",
 			     decoder->ref_timestamp);
@@ -523,6 +576,7 @@ static uint64_t intel_pt_next_sample(struct intel_pt_decoder *decoder)
 	case INTEL_PT_PERIOD_TICKS:
 		return intel_pt_next_period(decoder);
 	case INTEL_PT_PERIOD_NONE:
+	case INTEL_PT_PERIOD_MTC:
 	default:
 		return 0;
 	}
@@ -542,6 +596,7 @@ static void intel_pt_sample_insn(struct intel_pt_decoder *decoder)
 		decoder->last_masked_timestamp = masked_timestamp;
 		break;
 	case INTEL_PT_PERIOD_NONE:
+	case INTEL_PT_PERIOD_MTC:
 	default:
 		break;
 	}
@@ -555,6 +610,9 @@ static int intel_pt_walk_insn(struct intel_pt_decoder *decoder,
 	uint64_t max_insn_cnt, insn_cnt = 0;
 	int err;
 
+	if (!decoder->mtc_insn)
+		decoder->mtc_insn = true;
+
 	max_insn_cnt = intel_pt_next_sample(decoder);
 
 	err = decoder->walk_insn(intel_pt_insn, &insn_cnt, &decoder->ip, ip,
@@ -861,6 +919,8 @@ static void intel_pt_calc_tsc_timestamp(struct intel_pt_decoder *decoder)
 {
 	uint64_t timestamp;
 
+	decoder->have_tma = false;
+
 	if (decoder->ref_timestamp) {
 		timestamp = decoder->packet.payload |
 			    (decoder->ref_timestamp & (0xffULL << 56));
@@ -878,17 +938,18 @@ static void intel_pt_calc_tsc_timestamp(struct intel_pt_decoder *decoder)
 	} else if (decoder->timestamp) {
 		timestamp = decoder->packet.payload |
 			    (decoder->timestamp & (0xffULL << 56));
+		decoder->tsc_timestamp = timestamp;
 		if (timestamp < decoder->timestamp &&
-		    decoder->timestamp - timestamp < 0x100) {
-			intel_pt_log_to("ERROR: Suppressing backwards timestamp",
+		    decoder->timestamp - timestamp < decoder->tsc_slip) {
+			intel_pt_log_to("Suppressing backwards timestamp",
 					timestamp);
 			timestamp = decoder->timestamp;
 		}
 		while (timestamp < decoder->timestamp) {
 			intel_pt_log_to("Wraparound timestamp", timestamp);
 			timestamp += (1ULL << 56);
+			decoder->tsc_timestamp = timestamp;
 		}
-		decoder->tsc_timestamp = timestamp;
 		decoder->timestamp = timestamp;
 		decoder->timestamp_insn_cnt = 0;
 	}
@@ -900,11 +961,73 @@ static int intel_pt_overflow(struct intel_pt_decoder *decoder)
 {
 	intel_pt_log("ERROR: Buffer overflow\n");
 	intel_pt_clear_tx_flags(decoder);
+	decoder->have_tma = false;
 	decoder->pkt_state = INTEL_PT_STATE_ERR_RESYNC;
 	decoder->overflow = true;
 	return -EOVERFLOW;
 }
 
+static void intel_pt_calc_tma(struct intel_pt_decoder *decoder)
+{
+	uint32_t ctc = decoder->packet.payload;
+	uint32_t fc = decoder->packet.count;
+	uint32_t ctc_rem = ctc & decoder->ctc_rem_mask;
+
+	if (!decoder->tsc_ctc_ratio_d)
+		return;
+
+	decoder->last_mtc = (ctc >> decoder->mtc_shift) & 0xff;
+	decoder->ctc_timestamp = decoder->tsc_timestamp - fc;
+	if (decoder->tsc_ctc_mult) {
+		decoder->ctc_timestamp -= ctc_rem * decoder->tsc_ctc_mult;
+	} else {
+		decoder->ctc_timestamp -= multdiv(ctc_rem,
+						  decoder->tsc_ctc_ratio_n,
+						  decoder->tsc_ctc_ratio_d);
+	}
+	decoder->ctc_delta = 0;
+	decoder->have_tma = true;
+	intel_pt_log("CTC timestamp " x64_fmt " last MTC %#x  CTC rem %#x\n",
+		     decoder->ctc_timestamp, decoder->last_mtc, ctc_rem);
+}
+
+static void intel_pt_calc_mtc_timestamp(struct intel_pt_decoder *decoder)
+{
+	uint64_t timestamp;
+	uint32_t mtc, mtc_delta;
+
+	if (!decoder->have_tma)
+		return;
+
+	mtc = decoder->packet.payload;
+
+	if (mtc > decoder->last_mtc)
+		mtc_delta = mtc - decoder->last_mtc;
+	else
+		mtc_delta = mtc + 256 - decoder->last_mtc;
+
+	decoder->ctc_delta += mtc_delta << decoder->mtc_shift;
+
+	if (decoder->tsc_ctc_mult) {
+		timestamp = decoder->ctc_timestamp +
+			    decoder->ctc_delta * decoder->tsc_ctc_mult;
+	} else {
+		timestamp = decoder->ctc_timestamp +
+			    multdiv(decoder->ctc_delta,
+				    decoder->tsc_ctc_ratio_n,
+				    decoder->tsc_ctc_ratio_d);
+	}
+
+	if (timestamp < decoder->timestamp)
+		intel_pt_log("Suppressing MTC timestamp " x64_fmt " less than current timestamp " x64_fmt "\n",
+			     timestamp, decoder->timestamp);
+	else
+		decoder->timestamp = timestamp;
+
+	decoder->timestamp_insn_cnt = 0;
+	decoder->last_mtc = mtc;
+}
+
 /* Walk PSB+ packets when already in sync. */
 static int intel_pt_walk_psbend(struct intel_pt_decoder *decoder)
 {
@@ -926,6 +1049,7 @@ static int intel_pt_walk_psbend(struct intel_pt_decoder *decoder)
 		case INTEL_PT_TRACESTOP:
 		case INTEL_PT_BAD:
 		case INTEL_PT_PSB:
+			decoder->have_tma = false;
 			intel_pt_log("ERROR: Unexpected packet\n");
 			return -EAGAIN;
 
@@ -937,6 +1061,7 @@ static int intel_pt_walk_psbend(struct intel_pt_decoder *decoder)
 			break;
 
 		case INTEL_PT_TMA:
+			intel_pt_calc_tma(decoder);
 			break;
 
 		case INTEL_PT_CBR:
@@ -961,6 +1086,9 @@ static int intel_pt_walk_psbend(struct intel_pt_decoder *decoder)
 			break;
 
 		case INTEL_PT_MTC:
+			intel_pt_calc_mtc_timestamp(decoder);
+			if (decoder->period_type == INTEL_PT_PERIOD_MTC)
+				decoder->state.type |= INTEL_PT_INSTRUCTION;
 			break;
 
 		case INTEL_PT_CYC:
@@ -1048,6 +1176,9 @@ static int intel_pt_walk_fup_tip(struct intel_pt_decoder *decoder)
 			break;
 
 		case INTEL_PT_MTC:
+			intel_pt_calc_mtc_timestamp(decoder);
+			if (decoder->period_type == INTEL_PT_PERIOD_MTC)
+				decoder->state.type |= INTEL_PT_INSTRUCTION;
 			break;
 
 		case INTEL_PT_CYC:
@@ -1159,13 +1290,31 @@ next:
 			break;
 
 		case INTEL_PT_MTC:
-			break;
+			intel_pt_calc_mtc_timestamp(decoder);
+			if (decoder->period_type != INTEL_PT_PERIOD_MTC)
+				break;
+			/*
+			 * Ensure that there has been an instruction since the
+			 * last MTC.
+			 */
+			if (!decoder->mtc_insn)
+				break;
+			decoder->mtc_insn = false;
+			/* Ensure that there is a timestamp */
+			if (!decoder->timestamp)
+				break;
+			decoder->state.type = INTEL_PT_INSTRUCTION;
+			decoder->state.from_ip = decoder->ip;
+			decoder->state.to_ip = 0;
+			decoder->mtc_insn = false;
+			return 0;
 
 		case INTEL_PT_TSC:
 			intel_pt_calc_tsc_timestamp(decoder);
 			break;
 
 		case INTEL_PT_TMA:
+			intel_pt_calc_tma(decoder);
 			break;
 
 		case INTEL_PT_CYC:
@@ -1237,6 +1386,7 @@ static int intel_pt_walk_psb(struct intel_pt_decoder *decoder)
 			break;
 
 		case INTEL_PT_MTC:
+			intel_pt_calc_mtc_timestamp(decoder);
 			break;
 
 		case INTEL_PT_TSC:
@@ -1244,6 +1394,7 @@ static int intel_pt_walk_psb(struct intel_pt_decoder *decoder)
 			break;
 
 		case INTEL_PT_TMA:
+			intel_pt_calc_tma(decoder);
 			break;
 
 		case INTEL_PT_CYC:
@@ -1267,6 +1418,7 @@ static int intel_pt_walk_psb(struct intel_pt_decoder *decoder)
 
 		case INTEL_PT_TRACESTOP:
 		case INTEL_PT_TNT:
+			decoder->have_tma = false;
 			intel_pt_log("ERROR: Unexpected packet\n");
 			if (decoder->ip)
 				decoder->pkt_state = INTEL_PT_STATE_ERR4;
@@ -1329,6 +1481,7 @@ static int intel_pt_walk_to_ip(struct intel_pt_decoder *decoder)
 			break;
 
 		case INTEL_PT_MTC:
+			intel_pt_calc_mtc_timestamp(decoder);
 			break;
 
 		case INTEL_PT_TSC:
@@ -1336,6 +1489,7 @@ static int intel_pt_walk_to_ip(struct intel_pt_decoder *decoder)
 			break;
 
 		case INTEL_PT_TMA:
+			intel_pt_calc_tma(decoder);
 			break;
 
 		case INTEL_PT_CYC:
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 56cc47baca11..02c38fec1c37 100644
--- a/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h
+++ b/tools/perf/util/intel-pt-decoder/intel-pt-decoder.h
@@ -36,6 +36,7 @@ enum intel_pt_period_type {
 	INTEL_PT_PERIOD_NONE,
 	INTEL_PT_PERIOD_INSTRUCTIONS,
 	INTEL_PT_PERIOD_TICKS,
+	INTEL_PT_PERIOD_MTC,
 };
 
 enum {
-- 
2.1.0


  parent reply	other threads:[~2015-08-26 15:59 UTC|newest]

Thread overview: 24+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-08-26 15:57 [GIT PULL 00/22] perf/core improvements and fixes Arnaldo Carvalho de Melo
2015-08-26 15:57 ` [PATCH 01/22] perf tools: Fix tarball build broken by pt/bts Arnaldo Carvalho de Melo
2015-08-26 15:57 ` [PATCH 02/22] perf annotate: Reset the dso find_symbol cache when removing symbols Arnaldo Carvalho de Melo
2015-08-26 15:57 ` [PATCH 03/22] perf ui tui progress: Implement the ui_progress_ops->finish() method Arnaldo Carvalho de Melo
2015-08-26 15:57 ` [PATCH 04/22] perf ordered_events: Clear the progress bar at the end of a flush Arnaldo Carvalho de Melo
2015-08-26 15:57 ` [PATCH 05/22] perf tools: Fix Intel PT 'instructions' sample period Arnaldo Carvalho de Melo
2015-08-26 15:57 ` [PATCH 06/22] perf tools: Add Intel PT support for PSB periods Arnaldo Carvalho de Melo
2015-08-26 15:57 ` [PATCH 07/22] perf tools: Add new Intel PT packet definitions Arnaldo Carvalho de Melo
2015-08-26 15:57 ` [PATCH 08/22] perf tools: Pass Intel PT information for decoding MTC and CYC Arnaldo Carvalho de Melo
2015-08-26 15:57 ` Arnaldo Carvalho de Melo [this message]
2015-08-26 15:58 ` [PATCH 10/22] perf tools: Add Intel PT support for using MTC packets Arnaldo Carvalho de Melo
2015-08-26 15:58 ` [PATCH 11/22] perf tools: Add Intel PT support for decoding CYC packets Arnaldo Carvalho de Melo
2015-08-26 15:58 ` [PATCH 12/22] perf tools: Add Intel PT support for using " Arnaldo Carvalho de Melo
2015-08-26 15:58 ` [PATCH 13/22] perf tools: Add Intel PT support for decoding TRACESTOP packets Arnaldo Carvalho de Melo
2015-08-26 15:58 ` [PATCH 14/22] perf tools: Update Intel PT documentation Arnaldo Carvalho de Melo
2015-08-26 15:58 ` [PATCH 15/22] perf probe: Prevent segfault when reading probe point with absolute address Arnaldo Carvalho de Melo
2015-08-26 15:58 ` [PATCH 16/22] perf tools: Remove export.h from MANIFEST Arnaldo Carvalho de Melo
2015-08-26 15:58 ` [PATCH 17/22] tools build: Allow duplicate objects in the object list Arnaldo Carvalho de Melo
2015-08-26 15:58 ` [PATCH 18/22] perf probe: Fix list result when symbol can't be found Arnaldo Carvalho de Melo
2015-08-26 15:58 ` [PATCH 19/22] perf probe: Fix list result when address is zero Arnaldo Carvalho de Melo
2015-08-26 15:58 ` [PATCH 20/22] perf probe: Fix error reported when offset without function Arnaldo Carvalho de Melo
2015-08-26 15:58 ` [PATCH 21/22] perf probe: Support probing at absolute address Arnaldo Carvalho de Melo
2015-08-26 15:58 ` [PATCH 22/22] tracing/uprobes: Do not print '0x (null)' when offset is 0 Arnaldo Carvalho de Melo
2015-08-28  6:24 ` [GIT PULL 00/22] perf/core improvements and fixes Ingo Molnar

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1440604692-26918-10-git-send-email-acme@kernel.org \
    --to=acme@kernel.org \
    --cc=acme@redhat.com \
    --cc=adrian.hunter@intel.com \
    --cc=jolsa@redhat.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mingo@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.