From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752346AbbH1GiX (ORCPT ); Fri, 28 Aug 2015 02:38:23 -0400 Received: from terminus.zytor.com ([198.137.202.10]:54645 "EHLO terminus.zytor.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752310AbbH1GiU (ORCPT ); Fri, 28 Aug 2015 02:38:20 -0400 Date: Thu, 27 Aug 2015 23:38:08 -0700 From: tip-bot for Adrian Hunter Message-ID: Cc: mingo@kernel.org, hpa@zytor.com, adrian.hunter@intel.com, linux-kernel@vger.kernel.org, jolsa@redhat.com, acme@redhat.com, tglx@linutronix.de Reply-To: hpa@zytor.com, mingo@kernel.org, tglx@linutronix.de, acme@redhat.com, jolsa@redhat.com, linux-kernel@vger.kernel.org, adrian.hunter@intel.com In-Reply-To: <1437150840-31811-18-git-send-email-adrian.hunter@intel.com> References: <1437150840-31811-18-git-send-email-adrian.hunter@intel.com> To: linux-tip-commits@vger.kernel.org Subject: [tip:perf/core] perf tools: Add Intel PT support for PSB periods Git-Commit-ID: bc9b6bf07c8b3f4e85509f9b3a552c86e567b4ae X-Mailer: tip-git-log-daemon Robot-ID: Robot-Unsubscribe: Contact to get blacklisted from these emails MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset=UTF-8 Content-Disposition: inline Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Commit-ID: bc9b6bf07c8b3f4e85509f9b3a552c86e567b4ae Gitweb: http://git.kernel.org/tip/bc9b6bf07c8b3f4e85509f9b3a552c86e567b4ae Author: Adrian Hunter AuthorDate: Fri, 17 Jul 2015 19:33:52 +0300 Committer: Arnaldo Carvalho de Melo CommitDate: Mon, 24 Aug 2015 17:45:08 -0300 perf tools: Add Intel PT support for PSB periods The PSB packet is a synchronization packet that provides a starting point for decoding or recovery from errors. This patch adds support for a new Intel PT feature that allows the frequency of PSB packets to be specified. Support for this feature is indicated by /sys/bus/event_source/devices/intel_pt/caps/psb_cyc which contains "1" if the feature is supported and "0" otherwise. The PSB period can be specified as a PMU config term e.g. perf record -e intel_pt/psb_period=2/u sleep 1 The default value is 3 or the nearest lower value that is supported. 0 is always supported. Valid values are given by: /sys/bus/event_source/devices/intel_pt/caps/psb_periods which contains a hexadecimal value, the bits of which represent valid values e.g. bit 2 set means value 2 is valid. The value is converted to the approximate number of trace bytes between PSB packets as: 2 ^ (value + 11) e.g. value 3 means 16KiB bytes between PSBs If an invalid value is entered, the error message will give a list of valid values e.g. $ perf record -e intel_pt/psb_period=15/u uname Invalid psb_period for intel_pt. Valid values are: 0-5 tools/perf/Documentation/intel-pt.txt is updated in a later patch as there are a number of new features being added. For more information about PSB periods refer to the Intel 64 and IA-32 Architectures SDM Chapter 36 Intel Processor Trace from June 2015 or later. Signed-off-by: Adrian Hunter Cc: Jiri Olsa Link: http://lkml.kernel.org/r/1437150840-31811-18-git-send-email-adrian.hunter@intel.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/arch/x86/util/intel-pt.c | 217 ++++++++++++++++++++++++++++++++++-- 1 file changed, 210 insertions(+), 7 deletions(-) diff --git a/tools/perf/arch/x86/util/intel-pt.c b/tools/perf/arch/x86/util/intel-pt.c index da7d2c1..145975b 100644 --- a/tools/perf/arch/x86/util/intel-pt.c +++ b/tools/perf/arch/x86/util/intel-pt.c @@ -99,17 +99,121 @@ static int intel_pt_parse_terms(struct list_head *formats, const char *str, return intel_pt_parse_terms_with_default(formats, str, config); } -static size_t intel_pt_psb_period(struct perf_pmu *intel_pt_pmu __maybe_unused, - struct perf_evlist *evlist __maybe_unused) +static u64 intel_pt_masked_bits(u64 mask, u64 bits) { - return 256; + const u64 top_bit = 1ULL << 63; + u64 res = 0; + int i; + + for (i = 0; i < 64; i++) { + if (mask & top_bit) { + res <<= 1; + if (bits & top_bit) + res |= 1; + } + mask <<= 1; + bits <<= 1; + } + + return res; +} + +static int intel_pt_read_config(struct perf_pmu *intel_pt_pmu, const char *str, + struct perf_evlist *evlist, u64 *res) +{ + struct perf_evsel *evsel; + u64 mask; + + *res = 0; + + mask = perf_pmu__format_bits(&intel_pt_pmu->format, str); + if (!mask) + return -EINVAL; + + evlist__for_each(evlist, evsel) { + if (evsel->attr.type == intel_pt_pmu->type) { + *res = intel_pt_masked_bits(mask, evsel->attr.config); + return 0; + } + } + + return -EINVAL; +} + +static size_t intel_pt_psb_period(struct perf_pmu *intel_pt_pmu, + struct perf_evlist *evlist) +{ + u64 val; + int err, topa_multiple_entries; + size_t psb_period; + + if (perf_pmu__scan_file(intel_pt_pmu, "caps/topa_multiple_entries", + "%d", &topa_multiple_entries) != 1) + topa_multiple_entries = 0; + + /* + * Use caps/topa_multiple_entries to indicate early hardware that had + * extra frequent PSBs. + */ + if (!topa_multiple_entries) { + psb_period = 256; + goto out; + } + + err = intel_pt_read_config(intel_pt_pmu, "psb_period", evlist, &val); + if (err) + val = 0; + + psb_period = 1 << (val + 11); +out: + pr_debug2("%s psb_period %zu\n", intel_pt_pmu->name, psb_period); + return psb_period; +} + +static int intel_pt_pick_bit(int bits, int target) +{ + int pos, pick = -1; + + for (pos = 0; bits; bits >>= 1, pos++) { + if (bits & 1) { + if (pos <= target || pick < 0) + pick = pos; + if (pos >= target) + break; + } + } + + return pick; } static u64 intel_pt_default_config(struct perf_pmu *intel_pt_pmu) { + char buf[256]; + int psb_cyc, psb_periods, psb_period; + int pos = 0; u64 config; - intel_pt_parse_terms(&intel_pt_pmu->format, "tsc", &config); + pos += scnprintf(buf + pos, sizeof(buf) - pos, "tsc"); + + if (perf_pmu__scan_file(intel_pt_pmu, "caps/psb_cyc", "%d", + &psb_cyc) != 1) + psb_cyc = 1; + + if (psb_cyc) { + if (perf_pmu__scan_file(intel_pt_pmu, "caps/psb_periods", "%x", + &psb_periods) != 1) + psb_periods = 0; + if (psb_periods) { + psb_period = intel_pt_pick_bit(psb_periods, 3); + pos += scnprintf(buf + pos, sizeof(buf) - pos, + ",psb_period=%d", psb_period); + } + } + + pr_debug2("%s default config: %s\n", intel_pt_pmu->name, buf); + + intel_pt_parse_terms(&intel_pt_pmu->format, buf, &config); + return config; } @@ -239,6 +343,103 @@ static int intel_pt_track_switches(struct perf_evlist *evlist) return 0; } +static void intel_pt_valid_str(char *str, size_t len, u64 valid) +{ + unsigned int val, last = 0, state = 1; + int p = 0; + + str[0] = '\0'; + + for (val = 0; val <= 64; val++, valid >>= 1) { + if (valid & 1) { + last = val; + switch (state) { + case 0: + p += scnprintf(str + p, len - p, ","); + /* Fall through */ + case 1: + p += scnprintf(str + p, len - p, "%u", val); + state = 2; + break; + case 2: + state = 3; + break; + case 3: + state = 4; + break; + default: + break; + } + } else { + switch (state) { + case 3: + p += scnprintf(str + p, len - p, ",%u", last); + state = 0; + break; + case 4: + p += scnprintf(str + p, len - p, "-%u", last); + state = 0; + break; + default: + break; + } + if (state != 1) + state = 0; + } + } +} + +static int intel_pt_val_config_term(struct perf_pmu *intel_pt_pmu, + const char *caps, const char *name, + const char *supported, u64 config) +{ + char valid_str[256]; + unsigned int shift; + unsigned long long valid; + u64 bits; + int ok; + + if (perf_pmu__scan_file(intel_pt_pmu, caps, "%llx", &valid) != 1) + valid = 0; + + if (supported && + perf_pmu__scan_file(intel_pt_pmu, supported, "%d", &ok) == 1 && !ok) + valid = 0; + + valid |= 1; + + bits = perf_pmu__format_bits(&intel_pt_pmu->format, name); + + config &= bits; + + for (shift = 0; bits && !(bits & 1); shift++) + bits >>= 1; + + config >>= shift; + + if (config > 63) + goto out_err; + + if (valid & (1 << config)) + return 0; +out_err: + intel_pt_valid_str(valid_str, sizeof(valid_str), valid); + pr_err("Invalid %s for %s. Valid values are: %s\n", + name, INTEL_PT_PMU_NAME, valid_str); + return -EINVAL; +} + +static int intel_pt_validate_config(struct perf_pmu *intel_pt_pmu, + struct perf_evsel *evsel) +{ + if (!evsel) + return 0; + + return intel_pt_val_config_term(intel_pt_pmu, "caps/psb_periods", + "psb_period", "caps/psb_cyc", + evsel->attr.config); +} + static int intel_pt_recording_options(struct auxtrace_record *itr, struct perf_evlist *evlist, struct record_opts *opts) @@ -251,6 +452,7 @@ static int intel_pt_recording_options(struct auxtrace_record *itr, const struct cpu_map *cpus = evlist->cpus; bool privileged = geteuid() == 0 || perf_event_paranoid() < 0; u64 tsc_bit; + int err; ptr->evlist = evlist; ptr->snapshot_mode = opts->auxtrace_snapshot_mode; @@ -281,6 +483,10 @@ static int intel_pt_recording_options(struct auxtrace_record *itr, if (!opts->full_auxtrace) return 0; + err = intel_pt_validate_config(intel_pt_pmu, intel_pt_evsel); + if (err) + return err; + /* Set default sizes for snapshot mode */ if (opts->auxtrace_snapshot_mode) { size_t psb_period = intel_pt_psb_period(intel_pt_pmu, evlist); @@ -366,8 +572,6 @@ static int intel_pt_recording_options(struct auxtrace_record *itr, * threads. */ if (have_timing_info && !cpu_map__empty(cpus)) { - int err; - err = intel_pt_track_switches(evlist); if (err == -EPERM) pr_debug2("Unable to select sched:sched_switch\n"); @@ -394,7 +598,6 @@ static int intel_pt_recording_options(struct auxtrace_record *itr, /* Add dummy event to keep tracking */ if (opts->full_auxtrace) { struct perf_evsel *tracking_evsel; - int err; err = parse_events(evlist, "dummy:u", NULL); if (err)