From 05169df5561363ff04ac04d6aad0be3b45c26ac1 Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Thu, 20 Aug 2015 11:26:45 +0300 Subject: [PATCH 1/9] perf script: Fix segfault using --show-mmap-events Patch "perf script: Don't assume evsel position of tracking events" changed 'perf script' to use 'perf_evlist__id2evsel()'. That results in a segfault if there is more than 1 event and there are synthesized mmap events e.g. $ perf record -e cycles,instructions -p$$ sleep 1 $ perf script --show-mmap-events Segmentation fault (core dumped) That happens because these synthesized events have an 'id' of zero which does not match any 'evsel'. Currently, these synthesized events use the sample type of the first evsel. Change 'perf_evlist__id2evsel()' to reflect that which also makes it consistent with 'perf_evlist__event2evsel()'. Signed-off-by: Adrian Hunter Fixes: 06b234ec26fd ("perf script: Don't assume evsel position of tracking events") Cc: Jiri Olsa Link: http://lkml.kernel.org/r/1440059205-1765-1-git-send-email-adrian.hunter@intel.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/evlist.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c index 373f65b02545..e9a5d432902c 100644 --- a/tools/perf/util/evlist.c +++ b/tools/perf/util/evlist.c @@ -573,7 +573,7 @@ struct perf_evsel *perf_evlist__id2evsel(struct perf_evlist *evlist, u64 id) { struct perf_sample_id *sid; - if (evlist->nr_entries == 1) + if (evlist->nr_entries == 1 || !id) return perf_evlist__first(evlist); sid = perf_evlist__id2sid(evlist, id); From 133de940435493266acd4100b0dae8ba7eaa7c71 Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Thu, 20 Aug 2015 13:07:40 +0300 Subject: [PATCH 2/9] perf tools: /proc/kcore requires CAP_SYS_RAWIO message too noisy The "/proc/kcore requires CAP_SYS_RAWIO" message comes up all the time for 'perf script' if vmlinux is not found and the user isn't root, even when the kernel is not being traced and even though the message is only really relevant for annotation. Change it to pr_debug and instead put a note in the message displayed if annotation is not possible. Also, the file being accessed might not be /proc/kcore. Tools can be directed to a different location using the --kallsyms option in which case kcore is expected to be in the same directory. Adjust the message so it is not misleading in that case. Signed-off-by: Adrian Hunter Cc: Jiri Olsa Cc: Li Zhang Cc: Sukadev Bhattiprolu Link: http://lkml.kernel.org/r/1440065260-8802-1-git-send-email-adrian.hunter@intel.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/annotate.c | 1 + tools/perf/util/symbol.c | 4 ++-- 2 files changed, 3 insertions(+), 2 deletions(-) diff --git a/tools/perf/util/annotate.c b/tools/perf/util/annotate.c index 8a18347709e1..d1eece70b84d 100644 --- a/tools/perf/util/annotate.c +++ b/tools/perf/util/annotate.c @@ -1126,6 +1126,7 @@ fallback: dso->annotate_warned = 1; pr_err("Can't annotate %s:\n\n" "No vmlinux file%s\nwas found in the path.\n\n" + "Note that annotation using /proc/kcore requires CAP_SYS_RAWIO capability.\n\n" "Please use:\n\n" " perf buildid-cache -vu vmlinux\n\n" "or:\n\n" diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c index 725640fd7cd8..42e98ab5a9bb 100644 --- a/tools/perf/util/symbol.c +++ b/tools/perf/util/symbol.c @@ -1138,8 +1138,8 @@ static int dso__load_kcore(struct dso *dso, struct map *map, fd = open(kcore_filename, O_RDONLY); if (fd < 0) { - pr_err("%s requires CAP_SYS_RAWIO capability to access.\n", - kcore_filename); + pr_debug("Failed to open %s. Note /proc/kcore requires CAP_SYS_RAWIO capability to access.\n", + kcore_filename); return -EINVAL; } From 81cd60cc29a9c3e92ad6ca167a4764b2c2d2fc04 Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Thu, 20 Aug 2015 11:51:32 +0300 Subject: [PATCH 3/9] perf tools: Fix Intel PT timestamp handling Events that don't sample the timestamp have a timestamp value of -1. Intel PT processing wasn't taking that into account. This is particularly noticeable with Intel BTS because timestamps are not requested by default. Then, if the conversion of -1 to TSC results in a small number, the processing is unaffected. However if the conversion results in a big number, then the data is processed prematurely before relevant sideband data like mmap events, which in turn results in samples with unknown dsos. Commiter note: Since BTS wasn't upstream, I split the patch to fold the BTS part with the patch introducing it, to avoid having this bug in the commit history. PT was already upstream, so this patch contains that part. Signed-off-by: Adrian Hunter Cc: Jiri Olsa Link: http://lkml.kernel.org/r/1440060692-5585-1-git-send-email-adrian.hunter@intel.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/intel-pt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/perf/util/intel-pt.c b/tools/perf/util/intel-pt.c index 2a4a4120473b..a5acd2fe2447 100644 --- a/tools/perf/util/intel-pt.c +++ b/tools/perf/util/intel-pt.c @@ -1450,7 +1450,7 @@ static int intel_pt_process_event(struct perf_session *session, return -EINVAL; } - if (sample->time) + if (sample->time && sample->time != (u64)-1) timestamp = perf_time_to_tsc(sample->time, &pt->tc); else timestamp = 0; From 6f56e9cf581c6cedcaea3eb69444b169867ccf3d Mon Sep 17 00:00:00 2001 From: Dean Nelson Date: Thu, 20 Aug 2015 11:16:32 -0400 Subject: [PATCH 4/9] tools lib traceevent: Add checks for returned EVENT_ERROR type Running the following perf-stat command on an arm64 system produces the following result... [root@aarch64 ~]# perf stat -e kmem:mm_page_alloc -a sleep 1 Warning: [kmem:mm_page_alloc] function sizeof not defined Warning: Error: expected type 4 but read 0 Segmentation fault [root@aarch64 ~]# The second warning was a result of the first warning not stopping processing after it detected the issue. That is, code that found the issue reported the first problem, but because it did not exit out of the functions smoothly, it caused the other warning to appear and not only that, it later caused the SIGSEGV. Signed-off-by: Dean Nelson Reviewed-by: Steven Rostedt Acked-by: Namhyung Kim Cc: Jiri Olsa Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/20150820151632.13927.13791.email-sent-by-dnelson@teal Signed-off-by: Arnaldo Carvalho de Melo --- tools/lib/traceevent/event-parse.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/tools/lib/traceevent/event-parse.c b/tools/lib/traceevent/event-parse.c index fcd8a9e3d2e1..5c1867a13ef2 100644 --- a/tools/lib/traceevent/event-parse.c +++ b/tools/lib/traceevent/event-parse.c @@ -1745,6 +1745,9 @@ process_cond(struct event_format *event, struct print_arg *top, char **tok) type = process_arg(event, left, &token); again: + if (type == EVENT_ERROR) + goto out_free; + /* Handle other operations in the arguments */ if (type == EVENT_OP && strcmp(token, ":") != 0) { type = process_op(event, left, &token); @@ -2004,6 +2007,12 @@ process_op(struct event_format *event, struct print_arg *arg, char **tok) goto out_warn_free; type = process_arg_token(event, right, tok, type); + if (type == EVENT_ERROR) { + free_arg(right); + /* token was freed in process_arg_token() via *tok */ + token = NULL; + goto out_free; + } if (right->type == PRINT_OP && get_op_prio(arg->op.op) < get_op_prio(right->op.op)) { From d0170af7004dce9cd90b749842c37e379476cbc8 Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Fri, 17 Jul 2015 19:33:43 +0300 Subject: [PATCH 5/9] perf tools: Add Intel BTS support Intel BTS support fits within the new auxtrace infrastructure. Recording is supporting by identifying the Intel BTS PMU, parsing options and setting up events. Decoding is supported by queuing up trace data by thread and then decoding synchronously delivering synthesized event samples into the session processing for tools to consume. Committer note: E.g: [root@felicio ~]# perf record --per-thread -e intel_bts// ls anaconda-ks.cfg apctest.output bin kernel-rt-3.10.0-298.rt56.171.el7.x86_64.rpm libexec lock_page.bpf.c perf.data perf.data.old [ perf record: Woken up 3 times to write data ] [ perf record: Captured and wrote 4.367 MB perf.data ] [root@felicio ~]# perf evlist -v intel_bts//: type: 6, size: 112, { sample_period, sample_freq }: 1, sample_type: IP|TID|IDENTIFIER, read_format: ID, disabled: 1, enable_on_exec: 1, sample_id_all: 1, exclude_guest: 1 dummy:u: type: 1, size: 112, config: 0x9, { sample_period, sample_freq }: 1, sample_type: IP|TID|IDENTIFIER, read_format: ID, disabled: 1, exclude_kernel: 1, exclude_hv: 1, mmap: 1, comm: 1, enable_on_exec: 1, task: 1, sample_id_all: 1, mmap2: 1, comm_exec: 1 [root@felicio ~]# perf script # the navigate in the pager to some interesting place: ls 1843 1 branches: ffffffff810a60cb flush_signal_handlers ([kernel.kallsyms]) => ffffffff8121a522 setup_new_exec ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8121a529 setup_new_exec ([kernel.kallsyms]) => ffffffff8122fa30 do_close_on_exec ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8122fa5d do_close_on_exec ([kernel.kallsyms]) => ffffffff81767ae0 _raw_spin_lock ([kernel.kallsyms]) ls 1843 1 branches: ffffffff81767af4 _raw_spin_lock ([kernel.kallsyms]) => ffffffff8122fa62 do_close_on_exec ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8122fa8e do_close_on_exec ([kernel.kallsyms]) => ffffffff8122faf0 do_close_on_exec ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8122faf7 do_close_on_exec ([kernel.kallsyms]) => ffffffff8122fa8b do_close_on_exec ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8122fa8e do_close_on_exec ([kernel.kallsyms]) => ffffffff8122faf0 do_close_on_exec ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8122faf7 do_close_on_exec ([kernel.kallsyms]) => ffffffff8122fa8b do_close_on_exec ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8122fa8e do_close_on_exec ([kernel.kallsyms]) => ffffffff8122faf0 do_close_on_exec ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8122faf7 do_close_on_exec ([kernel.kallsyms]) => ffffffff8122fa8b do_close_on_exec ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8122fa8e do_close_on_exec ([kernel.kallsyms]) => ffffffff8122faf0 do_close_on_exec ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8122faf7 do_close_on_exec ([kernel.kallsyms]) => ffffffff8122fa8b do_close_on_exec ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8122fa8e do_close_on_exec ([kernel.kallsyms]) => ffffffff8122faf0 do_close_on_exec ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8122faf7 do_close_on_exec ([kernel.kallsyms]) => ffffffff8122fa8b do_close_on_exec ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8122fa8e do_close_on_exec ([kernel.kallsyms]) => ffffffff8122faf0 do_close_on_exec ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8122faf7 do_close_on_exec ([kernel.kallsyms]) => ffffffff8122fa8b do_close_on_exec ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8122fac9 do_close_on_exec ([kernel.kallsyms]) => ffffffff8122fad2 do_close_on_exec ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8122fadd do_close_on_exec ([kernel.kallsyms]) => ffffffff8120fc80 filp_close ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8120fcaf filp_close ([kernel.kallsyms]) => ffffffff8120fcb6 filp_close ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8120fcc2 filp_close ([kernel.kallsyms]) => ffffffff812547f0 dnotify_flush ([kernel.kallsyms]) ls 1843 1 branches: ffffffff81254823 dnotify_flush ([kernel.kallsyms]) => ffffffff8120fcc7 filp_close ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8120fccd filp_close ([kernel.kallsyms]) => ffffffff81261790 locks_remove_posix ([kernel.kallsyms]) ls 1843 1 branches: ffffffff812617a3 locks_remove_posix ([kernel.kallsyms]) => ffffffff812617b9 locks_remove_posix ([kernel.kallsyms]) ls 1843 1 branches: ffffffff812617b9 locks_remove_posix ([kernel.kallsyms]) => ffffffff8120fcd2 filp_close ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8120fcd5 filp_close ([kernel.kallsyms]) => ffffffff812142c0 fput ([kernel.kallsyms]) ls 1843 1 branches: ffffffff812142d6 fput ([kernel.kallsyms]) => ffffffff812142df fput ([kernel.kallsyms]) ls 1843 1 branches: ffffffff8121430c fput ([kernel.kallsyms]) => ffffffff810b6580 task_work_add ([kernel.kallsyms]) ls 1843 1 branches: ffffffff810b65ad task_work_add ([kernel.kallsyms]) => ffffffff810b65b1 task_work_add ([kernel.kallsyms]) ls 1843 1 branches: ffffffff810b65c1 task_work_add ([kernel.kallsyms]) => ffffffff810bc710 kick_process ([kernel.kallsyms]) ls 1843 1 branches: ffffffff810bc725 kick_process ([kernel.kallsyms]) => ffffffff810bc742 kick_process ([kernel.kallsyms]) ls 1843 1 branches: ffffffff810bc742 kick_process ([kernel.kallsyms]) => ffffffff810b65c6 task_work_add ([kernel.kallsyms]) ls 1843 1 branches: ffffffff810b65c9 task_work_add ([kernel.kallsyms]) => ffffffff81214311 fput ([kernel.kallsyms]) Signed-off-by: Adrian Hunter Tested-by: Arnaldo Carvalho de Melo Cc: Jiri Olsa Link: http://lkml.kernel.org/r/1437150840-31811-9-git-send-email-adrian.hunter@intel.com [ Merged sample->time fix for bug found after first round of testing on slightly older kernel ] Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/Documentation/intel-bts.txt | 86 +++ tools/perf/arch/x86/util/Build | 1 + tools/perf/arch/x86/util/auxtrace.c | 49 +- tools/perf/arch/x86/util/intel-bts.c | 458 ++++++++++++ tools/perf/arch/x86/util/pmu.c | 3 + tools/perf/util/Build | 1 + tools/perf/util/auxtrace.c | 3 + tools/perf/util/auxtrace.h | 1 + tools/perf/util/intel-bts.c | 933 +++++++++++++++++++++++++ tools/perf/util/intel-bts.h | 43 ++ tools/perf/util/pmu.c | 4 - 11 files changed, 1576 insertions(+), 6 deletions(-) create mode 100644 tools/perf/Documentation/intel-bts.txt create mode 100644 tools/perf/arch/x86/util/intel-bts.c create mode 100644 tools/perf/util/intel-bts.c create mode 100644 tools/perf/util/intel-bts.h diff --git a/tools/perf/Documentation/intel-bts.txt b/tools/perf/Documentation/intel-bts.txt new file mode 100644 index 000000000000..8bdc93bd7fdb --- /dev/null +++ b/tools/perf/Documentation/intel-bts.txt @@ -0,0 +1,86 @@ +Intel Branch Trace Store +======================== + +Overview +======== + +Intel BTS could be regarded as a predecessor to Intel PT and has some +similarities because it can also identify every branch a program takes. A +notable difference is that Intel BTS has no timing information and as a +consequence the present implementation is limited to per-thread recording. + +While decoding Intel BTS does not require walking the object code, the object +code is still needed to pair up calls and returns correctly, consequently much +of the Intel PT documentation applies also to Intel BTS. Refer to the Intel PT +documentation and consider that the PMU 'intel_bts' can usually be used in +place of 'intel_pt' in the examples provided, with the proviso that per-thread +recording must also be stipulated i.e. the --per-thread option for +'perf record'. + + +perf record +=========== + +new event +--------- + +The Intel BTS kernel driver creates a new PMU for Intel BTS. The perf record +option is: + + -e intel_bts// + +Currently Intel BTS is limited to per-thread tracing so the --per-thread option +is also needed. + + +snapshot option +--------------- + +The snapshot option is the same as Intel PT (refer Intel PT documentation). + + +auxtrace mmap size option +----------------------- + +The mmap size option is the same as Intel PT (refer Intel PT documentation). + + +perf script +=========== + +By default, perf script will decode trace data found in the perf.data file. +This can be further controlled by option --itrace. The --itrace option is +the same as Intel PT (refer Intel PT documentation) except that neither +"instructions" events nor "transactions" events (and consequently call +chains) are supported. + +To disable trace decoding entirely, use the option --no-itrace. + + +dump option +----------- + +perf script has an option (-D) to "dump" the events i.e. display the binary +data. + +When -D is used, Intel BTS packets are displayed. + +To disable the display of Intel BTS packets, combine the -D option with +--no-itrace. + + +perf report +=========== + +By default, perf report will decode trace data found in the perf.data file. +This can be further controlled by new option --itrace exactly the same as +perf script. + + +perf inject +=========== + +perf inject also accepts the --itrace option in which case tracing data is +removed and replaced with the synthesized events. e.g. + + perf inject --itrace -i perf.data -o perf.data.new diff --git a/tools/perf/arch/x86/util/Build b/tools/perf/arch/x86/util/Build index a8be9f9d0462..2c55e1b336c5 100644 --- a/tools/perf/arch/x86/util/Build +++ b/tools/perf/arch/x86/util/Build @@ -10,3 +10,4 @@ libperf-$(CONFIG_LIBDW_DWARF_UNWIND) += unwind-libdw.o libperf-$(CONFIG_AUXTRACE) += auxtrace.o libperf-$(CONFIG_AUXTRACE) += intel-pt.o +libperf-$(CONFIG_AUXTRACE) += intel-bts.o diff --git a/tools/perf/arch/x86/util/auxtrace.c b/tools/perf/arch/x86/util/auxtrace.c index e7654b506312..7a7805583e3f 100644 --- a/tools/perf/arch/x86/util/auxtrace.c +++ b/tools/perf/arch/x86/util/auxtrace.c @@ -13,11 +13,56 @@ * */ +#include + #include "../../util/header.h" +#include "../../util/debug.h" +#include "../../util/pmu.h" #include "../../util/auxtrace.h" #include "../../util/intel-pt.h" +#include "../../util/intel-bts.h" +#include "../../util/evlist.h" -struct auxtrace_record *auxtrace_record__init(struct perf_evlist *evlist __maybe_unused, +static +struct auxtrace_record *auxtrace_record__init_intel(struct perf_evlist *evlist, + int *err) +{ + struct perf_pmu *intel_pt_pmu; + struct perf_pmu *intel_bts_pmu; + struct perf_evsel *evsel; + bool found_pt = false; + bool found_bts = false; + + intel_pt_pmu = perf_pmu__find(INTEL_PT_PMU_NAME); + intel_bts_pmu = perf_pmu__find(INTEL_BTS_PMU_NAME); + + if (evlist) { + evlist__for_each(evlist, evsel) { + if (intel_pt_pmu && + evsel->attr.type == intel_pt_pmu->type) + found_pt = true; + if (intel_bts_pmu && + evsel->attr.type == intel_bts_pmu->type) + found_bts = true; + } + } + + if (found_pt && found_bts) { + pr_err("intel_pt and intel_bts may not be used together\n"); + *err = -EINVAL; + return NULL; + } + + if (found_pt) + return intel_pt_recording_init(err); + + if (found_bts) + return intel_bts_recording_init(err); + + return NULL; +} + +struct auxtrace_record *auxtrace_record__init(struct perf_evlist *evlist, int *err) { char buffer[64]; @@ -32,7 +77,7 @@ struct auxtrace_record *auxtrace_record__init(struct perf_evlist *evlist __maybe } if (!strncmp(buffer, "GenuineIntel,", 13)) - return intel_pt_recording_init(err); + return auxtrace_record__init_intel(evlist, err); return NULL; } diff --git a/tools/perf/arch/x86/util/intel-bts.c b/tools/perf/arch/x86/util/intel-bts.c new file mode 100644 index 000000000000..9b94ce520917 --- /dev/null +++ b/tools/perf/arch/x86/util/intel-bts.c @@ -0,0 +1,458 @@ +/* + * intel-bts.c: Intel Processor Trace support + * Copyright (c) 2013-2015, Intel Corporation. + * + * This program is free software; you can redistribute it and/or modify it + * under the terms and conditions of the GNU General Public License, + * version 2, as published by the Free Software Foundation. + * + * This program is distributed in the hope it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + */ + +#include +#include +#include +#include + +#include "../../util/cpumap.h" +#include "../../util/evsel.h" +#include "../../util/evlist.h" +#include "../../util/session.h" +#include "../../util/util.h" +#include "../../util/pmu.h" +#include "../../util/debug.h" +#include "../../util/tsc.h" +#include "../../util/auxtrace.h" +#include "../../util/intel-bts.h" + +#define KiB(x) ((x) * 1024) +#define MiB(x) ((x) * 1024 * 1024) +#define KiB_MASK(x) (KiB(x) - 1) +#define MiB_MASK(x) (MiB(x) - 1) + +#define INTEL_BTS_DFLT_SAMPLE_SIZE KiB(4) + +#define INTEL_BTS_MAX_SAMPLE_SIZE KiB(60) + +struct intel_bts_snapshot_ref { + void *ref_buf; + size_t ref_offset; + bool wrapped; +}; + +struct intel_bts_recording { + struct auxtrace_record itr; + struct perf_pmu *intel_bts_pmu; + struct perf_evlist *evlist; + bool snapshot_mode; + size_t snapshot_size; + int snapshot_ref_cnt; + struct intel_bts_snapshot_ref *snapshot_refs; +}; + +struct branch { + u64 from; + u64 to; + u64 misc; +}; + +static size_t intel_bts_info_priv_size(struct auxtrace_record *itr __maybe_unused) +{ + return INTEL_BTS_AUXTRACE_PRIV_SIZE; +} + +static int intel_bts_info_fill(struct auxtrace_record *itr, + struct perf_session *session, + struct auxtrace_info_event *auxtrace_info, + size_t priv_size) +{ + struct intel_bts_recording *btsr = + container_of(itr, struct intel_bts_recording, itr); + struct perf_pmu *intel_bts_pmu = btsr->intel_bts_pmu; + struct perf_event_mmap_page *pc; + struct perf_tsc_conversion tc = { .time_mult = 0, }; + bool cap_user_time_zero = false; + int err; + + if (priv_size != INTEL_BTS_AUXTRACE_PRIV_SIZE) + return -EINVAL; + + if (!session->evlist->nr_mmaps) + return -EINVAL; + + pc = session->evlist->mmap[0].base; + if (pc) { + err = perf_read_tsc_conversion(pc, &tc); + if (err) { + if (err != -EOPNOTSUPP) + return err; + } else { + cap_user_time_zero = tc.time_mult != 0; + } + if (!cap_user_time_zero) + ui__warning("Intel BTS: TSC not available\n"); + } + + auxtrace_info->type = PERF_AUXTRACE_INTEL_BTS; + auxtrace_info->priv[INTEL_BTS_PMU_TYPE] = intel_bts_pmu->type; + auxtrace_info->priv[INTEL_BTS_TIME_SHIFT] = tc.time_shift; + auxtrace_info->priv[INTEL_BTS_TIME_MULT] = tc.time_mult; + auxtrace_info->priv[INTEL_BTS_TIME_ZERO] = tc.time_zero; + auxtrace_info->priv[INTEL_BTS_CAP_USER_TIME_ZERO] = cap_user_time_zero; + auxtrace_info->priv[INTEL_BTS_SNAPSHOT_MODE] = btsr->snapshot_mode; + + return 0; +} + +static int intel_bts_recording_options(struct auxtrace_record *itr, + struct perf_evlist *evlist, + struct record_opts *opts) +{ + struct intel_bts_recording *btsr = + container_of(itr, struct intel_bts_recording, itr); + struct perf_pmu *intel_bts_pmu = btsr->intel_bts_pmu; + struct perf_evsel *evsel, *intel_bts_evsel = NULL; + const struct cpu_map *cpus = evlist->cpus; + bool privileged = geteuid() == 0 || perf_event_paranoid() < 0; + + btsr->evlist = evlist; + btsr->snapshot_mode = opts->auxtrace_snapshot_mode; + + evlist__for_each(evlist, evsel) { + if (evsel->attr.type == intel_bts_pmu->type) { + if (intel_bts_evsel) { + pr_err("There may be only one " INTEL_BTS_PMU_NAME " event\n"); + return -EINVAL; + } + evsel->attr.freq = 0; + evsel->attr.sample_period = 1; + intel_bts_evsel = evsel; + opts->full_auxtrace = true; + } + } + + if (opts->auxtrace_snapshot_mode && !opts->full_auxtrace) { + pr_err("Snapshot mode (-S option) requires " INTEL_BTS_PMU_NAME " PMU event (-e " INTEL_BTS_PMU_NAME ")\n"); + return -EINVAL; + } + + if (!opts->full_auxtrace) + return 0; + + if (opts->full_auxtrace && !cpu_map__empty(cpus)) { + pr_err(INTEL_BTS_PMU_NAME " does not support per-cpu recording\n"); + return -EINVAL; + } + + /* Set default sizes for snapshot mode */ + if (opts->auxtrace_snapshot_mode) { + if (!opts->auxtrace_snapshot_size && !opts->auxtrace_mmap_pages) { + if (privileged) { + opts->auxtrace_mmap_pages = MiB(4) / page_size; + } else { + opts->auxtrace_mmap_pages = KiB(128) / page_size; + if (opts->mmap_pages == UINT_MAX) + opts->mmap_pages = KiB(256) / page_size; + } + } else if (!opts->auxtrace_mmap_pages && !privileged && + opts->mmap_pages == UINT_MAX) { + opts->mmap_pages = KiB(256) / page_size; + } + if (!opts->auxtrace_snapshot_size) + opts->auxtrace_snapshot_size = + opts->auxtrace_mmap_pages * (size_t)page_size; + if (!opts->auxtrace_mmap_pages) { + size_t sz = opts->auxtrace_snapshot_size; + + sz = round_up(sz, page_size) / page_size; + opts->auxtrace_mmap_pages = roundup_pow_of_two(sz); + } + if (opts->auxtrace_snapshot_size > + opts->auxtrace_mmap_pages * (size_t)page_size) { + pr_err("Snapshot size %zu must not be greater than AUX area tracing mmap size %zu\n", + opts->auxtrace_snapshot_size, + opts->auxtrace_mmap_pages * (size_t)page_size); + return -EINVAL; + } + if (!opts->auxtrace_snapshot_size || !opts->auxtrace_mmap_pages) { + pr_err("Failed to calculate default snapshot size and/or AUX area tracing mmap pages\n"); + return -EINVAL; + } + pr_debug2("Intel BTS snapshot size: %zu\n", + opts->auxtrace_snapshot_size); + } + + /* Set default sizes for full trace mode */ + if (opts->full_auxtrace && !opts->auxtrace_mmap_pages) { + if (privileged) { + opts->auxtrace_mmap_pages = MiB(4) / page_size; + } else { + opts->auxtrace_mmap_pages = KiB(128) / page_size; + if (opts->mmap_pages == UINT_MAX) + opts->mmap_pages = KiB(256) / page_size; + } + } + + /* Validate auxtrace_mmap_pages */ + if (opts->auxtrace_mmap_pages) { + size_t sz = opts->auxtrace_mmap_pages * (size_t)page_size; + size_t min_sz; + + if (opts->auxtrace_snapshot_mode) + min_sz = KiB(4); + else + min_sz = KiB(8); + + if (sz < min_sz || !is_power_of_2(sz)) { + pr_err("Invalid mmap size for Intel BTS: must be at least %zuKiB and a power of 2\n", + min_sz / 1024); + return -EINVAL; + } + } + + if (intel_bts_evsel) { + /* + * To obtain the auxtrace buffer file descriptor, the auxtrace event + * must come first. + */ + perf_evlist__to_front(evlist, intel_bts_evsel); + /* + * In the case of per-cpu mmaps, we need the CPU on the + * AUX event. + */ + if (!cpu_map__empty(cpus)) + perf_evsel__set_sample_bit(intel_bts_evsel, CPU); + } + + /* 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) + return err; + + tracking_evsel = perf_evlist__last(evlist); + + perf_evlist__set_tracking_event(evlist, tracking_evsel); + + tracking_evsel->attr.freq = 0; + tracking_evsel->attr.sample_period = 1; + } + + return 0; +} + +static int intel_bts_parse_snapshot_options(struct auxtrace_record *itr, + struct record_opts *opts, + const char *str) +{ + struct intel_bts_recording *btsr = + container_of(itr, struct intel_bts_recording, itr); + unsigned long long snapshot_size = 0; + char *endptr; + + if (str) { + snapshot_size = strtoull(str, &endptr, 0); + if (*endptr || snapshot_size > SIZE_MAX) + return -1; + } + + opts->auxtrace_snapshot_mode = true; + opts->auxtrace_snapshot_size = snapshot_size; + + btsr->snapshot_size = snapshot_size; + + return 0; +} + +static u64 intel_bts_reference(struct auxtrace_record *itr __maybe_unused) +{ + return rdtsc(); +} + +static int intel_bts_alloc_snapshot_refs(struct intel_bts_recording *btsr, + int idx) +{ + const size_t sz = sizeof(struct intel_bts_snapshot_ref); + int cnt = btsr->snapshot_ref_cnt, new_cnt = cnt * 2; + struct intel_bts_snapshot_ref *refs; + + if (!new_cnt) + new_cnt = 16; + + while (new_cnt <= idx) + new_cnt *= 2; + + refs = calloc(new_cnt, sz); + if (!refs) + return -ENOMEM; + + memcpy(refs, btsr->snapshot_refs, cnt * sz); + + btsr->snapshot_refs = refs; + btsr->snapshot_ref_cnt = new_cnt; + + return 0; +} + +static void intel_bts_free_snapshot_refs(struct intel_bts_recording *btsr) +{ + int i; + + for (i = 0; i < btsr->snapshot_ref_cnt; i++) + zfree(&btsr->snapshot_refs[i].ref_buf); + zfree(&btsr->snapshot_refs); +} + +static void intel_bts_recording_free(struct auxtrace_record *itr) +{ + struct intel_bts_recording *btsr = + container_of(itr, struct intel_bts_recording, itr); + + intel_bts_free_snapshot_refs(btsr); + free(btsr); +} + +static int intel_bts_snapshot_start(struct auxtrace_record *itr) +{ + struct intel_bts_recording *btsr = + container_of(itr, struct intel_bts_recording, itr); + struct perf_evsel *evsel; + + evlist__for_each(btsr->evlist, evsel) { + if (evsel->attr.type == btsr->intel_bts_pmu->type) + return perf_evlist__disable_event(btsr->evlist, evsel); + } + return -EINVAL; +} + +static int intel_bts_snapshot_finish(struct auxtrace_record *itr) +{ + struct intel_bts_recording *btsr = + container_of(itr, struct intel_bts_recording, itr); + struct perf_evsel *evsel; + + evlist__for_each(btsr->evlist, evsel) { + if (evsel->attr.type == btsr->intel_bts_pmu->type) + return perf_evlist__enable_event(btsr->evlist, evsel); + } + return -EINVAL; +} + +static bool intel_bts_first_wrap(u64 *data, size_t buf_size) +{ + int i, a, b; + + b = buf_size >> 3; + a = b - 512; + if (a < 0) + a = 0; + + for (i = a; i < b; i++) { + if (data[i]) + return true; + } + + return false; +} + +static int intel_bts_find_snapshot(struct auxtrace_record *itr, int idx, + struct auxtrace_mmap *mm, unsigned char *data, + u64 *head, u64 *old) +{ + struct intel_bts_recording *btsr = + container_of(itr, struct intel_bts_recording, itr); + bool wrapped; + int err; + + pr_debug3("%s: mmap index %d old head %zu new head %zu\n", + __func__, idx, (size_t)*old, (size_t)*head); + + if (idx >= btsr->snapshot_ref_cnt) { + err = intel_bts_alloc_snapshot_refs(btsr, idx); + if (err) + goto out_err; + } + + wrapped = btsr->snapshot_refs[idx].wrapped; + if (!wrapped && intel_bts_first_wrap((u64 *)data, mm->len)) { + btsr->snapshot_refs[idx].wrapped = true; + wrapped = true; + } + + /* + * In full trace mode 'head' continually increases. However in snapshot + * mode 'head' is an offset within the buffer. Here 'old' and 'head' + * are adjusted to match the full trace case which expects that 'old' is + * always less than 'head'. + */ + if (wrapped) { + *old = *head; + *head += mm->len; + } else { + if (mm->mask) + *old &= mm->mask; + else + *old %= mm->len; + if (*old > *head) + *head += mm->len; + } + + pr_debug3("%s: wrap-around %sdetected, adjusted old head %zu adjusted new head %zu\n", + __func__, wrapped ? "" : "not ", (size_t)*old, (size_t)*head); + + return 0; + +out_err: + pr_err("%s: failed, error %d\n", __func__, err); + return err; +} + +static int intel_bts_read_finish(struct auxtrace_record *itr, int idx) +{ + struct intel_bts_recording *btsr = + container_of(itr, struct intel_bts_recording, itr); + struct perf_evsel *evsel; + + evlist__for_each(btsr->evlist, evsel) { + if (evsel->attr.type == btsr->intel_bts_pmu->type) + return perf_evlist__enable_event_idx(btsr->evlist, + evsel, idx); + } + return -EINVAL; +} + +struct auxtrace_record *intel_bts_recording_init(int *err) +{ + struct perf_pmu *intel_bts_pmu = perf_pmu__find(INTEL_BTS_PMU_NAME); + struct intel_bts_recording *btsr; + + if (!intel_bts_pmu) + return NULL; + + btsr = zalloc(sizeof(struct intel_bts_recording)); + if (!btsr) { + *err = -ENOMEM; + return NULL; + } + + btsr->intel_bts_pmu = intel_bts_pmu; + btsr->itr.recording_options = intel_bts_recording_options; + btsr->itr.info_priv_size = intel_bts_info_priv_size; + btsr->itr.info_fill = intel_bts_info_fill; + btsr->itr.free = intel_bts_recording_free; + btsr->itr.snapshot_start = intel_bts_snapshot_start; + btsr->itr.snapshot_finish = intel_bts_snapshot_finish; + btsr->itr.find_snapshot = intel_bts_find_snapshot; + btsr->itr.parse_snapshot_options = intel_bts_parse_snapshot_options; + btsr->itr.reference = intel_bts_reference; + btsr->itr.read_finish = intel_bts_read_finish; + btsr->itr.alignment = sizeof(struct branch); + return &btsr->itr; +} diff --git a/tools/perf/arch/x86/util/pmu.c b/tools/perf/arch/x86/util/pmu.c index fd11cc3ce780..79fe07158d00 100644 --- a/tools/perf/arch/x86/util/pmu.c +++ b/tools/perf/arch/x86/util/pmu.c @@ -3,6 +3,7 @@ #include #include "../../util/intel-pt.h" +#include "../../util/intel-bts.h" #include "../../util/pmu.h" struct perf_event_attr *perf_pmu__get_default_config(struct perf_pmu *pmu __maybe_unused) @@ -10,6 +11,8 @@ struct perf_event_attr *perf_pmu__get_default_config(struct perf_pmu *pmu __mayb #ifdef HAVE_AUXTRACE_SUPPORT if (!strcmp(pmu->name, INTEL_PT_PMU_NAME)) return intel_pt_pmu_default_config(pmu); + if (!strcmp(pmu->name, INTEL_BTS_PMU_NAME)) + pmu->selectable = true; #endif return NULL; } diff --git a/tools/perf/util/Build b/tools/perf/util/Build index c20473d1369e..e912856cc4e5 100644 --- a/tools/perf/util/Build +++ b/tools/perf/util/Build @@ -80,6 +80,7 @@ libperf-y += thread-stack.o libperf-$(CONFIG_AUXTRACE) += auxtrace.o libperf-$(CONFIG_AUXTRACE) += intel-pt-decoder/ libperf-$(CONFIG_AUXTRACE) += intel-pt.o +libperf-$(CONFIG_AUXTRACE) += intel-bts.o libperf-y += parse-branch-options.o libperf-$(CONFIG_LIBELF) += symbol-elf.o diff --git a/tools/perf/util/auxtrace.c b/tools/perf/util/auxtrace.c index 0f0b7e11e2d9..a980e7c50ee0 100644 --- a/tools/perf/util/auxtrace.c +++ b/tools/perf/util/auxtrace.c @@ -48,6 +48,7 @@ #include "parse-options.h" #include "intel-pt.h" +#include "intel-bts.h" int auxtrace_mmap__mmap(struct auxtrace_mmap *mm, struct auxtrace_mmap_params *mp, @@ -888,6 +889,8 @@ int perf_event__process_auxtrace_info(struct perf_tool *tool __maybe_unused, switch (type) { case PERF_AUXTRACE_INTEL_PT: return intel_pt_process_auxtrace_info(event, session); + case PERF_AUXTRACE_INTEL_BTS: + return intel_bts_process_auxtrace_info(event, session); case PERF_AUXTRACE_UNKNOWN: default: return -EINVAL; diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h index 7d12f33a3a06..bf72b77a588a 100644 --- a/tools/perf/util/auxtrace.h +++ b/tools/perf/util/auxtrace.h @@ -40,6 +40,7 @@ struct events_stats; enum auxtrace_type { PERF_AUXTRACE_UNKNOWN, PERF_AUXTRACE_INTEL_PT, + PERF_AUXTRACE_INTEL_BTS, }; enum itrace_period_type { diff --git a/tools/perf/util/intel-bts.c b/tools/perf/util/intel-bts.c new file mode 100644 index 000000000000..ea768625ab5b --- /dev/null +++ b/tools/perf/util/intel-bts.c @@ -0,0 +1,933 @@ +/* + * intel-bts.c: Intel Processor Trace support + * Copyright (c) 2013-2015, Intel Corporation. + * + * This program is free software; you can redistribute it and/or modify it + * under the terms and conditions of the GNU General Public License, + * version 2, as published by the Free Software Foundation. + * + * This program is distributed in the hope it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + */ + +#include +#include +#include +#include +#include +#include + +#include "cpumap.h" +#include "color.h" +#include "evsel.h" +#include "evlist.h" +#include "machine.h" +#include "session.h" +#include "util.h" +#include "thread.h" +#include "thread-stack.h" +#include "debug.h" +#include "tsc.h" +#include "auxtrace.h" +#include "intel-pt-decoder/intel-pt-insn-decoder.h" +#include "intel-bts.h" + +#define MAX_TIMESTAMP (~0ULL) + +#define INTEL_BTS_ERR_NOINSN 5 +#define INTEL_BTS_ERR_LOST 9 + +#if __BYTE_ORDER == __BIG_ENDIAN +#define le64_to_cpu bswap_64 +#else +#define le64_to_cpu +#endif + +struct intel_bts { + struct auxtrace auxtrace; + struct auxtrace_queues queues; + struct auxtrace_heap heap; + u32 auxtrace_type; + struct perf_session *session; + struct machine *machine; + bool sampling_mode; + bool snapshot_mode; + bool data_queued; + u32 pmu_type; + struct perf_tsc_conversion tc; + bool cap_user_time_zero; + struct itrace_synth_opts synth_opts; + bool sample_branches; + u32 branches_filter; + u64 branches_sample_type; + u64 branches_id; + size_t branches_event_size; + bool synth_needs_swap; +}; + +struct intel_bts_queue { + struct intel_bts *bts; + unsigned int queue_nr; + struct auxtrace_buffer *buffer; + bool on_heap; + bool done; + pid_t pid; + pid_t tid; + int cpu; + u64 time; + struct intel_pt_insn intel_pt_insn; + u32 sample_flags; +}; + +struct branch { + u64 from; + u64 to; + u64 misc; +}; + +static void intel_bts_dump(struct intel_bts *bts __maybe_unused, + unsigned char *buf, size_t len) +{ + struct branch *branch; + size_t i, pos = 0, br_sz = sizeof(struct branch), sz; + const char *color = PERF_COLOR_BLUE; + + color_fprintf(stdout, color, + ". ... Intel BTS data: size %zu bytes\n", + len); + + while (len) { + if (len >= br_sz) + sz = br_sz; + else + sz = len; + printf("."); + color_fprintf(stdout, color, " %08x: ", pos); + for (i = 0; i < sz; i++) + color_fprintf(stdout, color, " %02x", buf[i]); + for (; i < br_sz; i++) + color_fprintf(stdout, color, " "); + if (len >= br_sz) { + branch = (struct branch *)buf; + color_fprintf(stdout, color, " %"PRIx64" -> %"PRIx64" %s\n", + le64_to_cpu(branch->from), + le64_to_cpu(branch->to), + le64_to_cpu(branch->misc) & 0x10 ? + "pred" : "miss"); + } else { + color_fprintf(stdout, color, " Bad record!\n"); + } + pos += sz; + buf += sz; + len -= sz; + } +} + +static void intel_bts_dump_event(struct intel_bts *bts, unsigned char *buf, + size_t len) +{ + printf(".\n"); + intel_bts_dump(bts, buf, len); +} + +static int intel_bts_lost(struct intel_bts *bts, struct perf_sample *sample) +{ + union perf_event event; + int err; + + auxtrace_synth_error(&event.auxtrace_error, PERF_AUXTRACE_ERROR_ITRACE, + INTEL_BTS_ERR_LOST, sample->cpu, sample->pid, + sample->tid, 0, "Lost trace data"); + + err = perf_session__deliver_synth_event(bts->session, &event, NULL); + if (err) + pr_err("Intel BTS: failed to deliver error event, error %d\n", + err); + + return err; +} + +static struct intel_bts_queue *intel_bts_alloc_queue(struct intel_bts *bts, + unsigned int queue_nr) +{ + struct intel_bts_queue *btsq; + + btsq = zalloc(sizeof(struct intel_bts_queue)); + if (!btsq) + return NULL; + + btsq->bts = bts; + btsq->queue_nr = queue_nr; + btsq->pid = -1; + btsq->tid = -1; + btsq->cpu = -1; + + return btsq; +} + +static int intel_bts_setup_queue(struct intel_bts *bts, + struct auxtrace_queue *queue, + unsigned int queue_nr) +{ + struct intel_bts_queue *btsq = queue->priv; + + if (list_empty(&queue->head)) + return 0; + + if (!btsq) { + btsq = intel_bts_alloc_queue(bts, queue_nr); + if (!btsq) + return -ENOMEM; + queue->priv = btsq; + + if (queue->cpu != -1) + btsq->cpu = queue->cpu; + btsq->tid = queue->tid; + } + + if (bts->sampling_mode) + return 0; + + if (!btsq->on_heap && !btsq->buffer) { + int ret; + + btsq->buffer = auxtrace_buffer__next(queue, NULL); + if (!btsq->buffer) + return 0; + + ret = auxtrace_heap__add(&bts->heap, queue_nr, + btsq->buffer->reference); + if (ret) + return ret; + btsq->on_heap = true; + } + + return 0; +} + +static int intel_bts_setup_queues(struct intel_bts *bts) +{ + unsigned int i; + int ret; + + for (i = 0; i < bts->queues.nr_queues; i++) { + ret = intel_bts_setup_queue(bts, &bts->queues.queue_array[i], + i); + if (ret) + return ret; + } + return 0; +} + +static inline int intel_bts_update_queues(struct intel_bts *bts) +{ + if (bts->queues.new_data) { + bts->queues.new_data = false; + return intel_bts_setup_queues(bts); + } + return 0; +} + +static unsigned char *intel_bts_find_overlap(unsigned char *buf_a, size_t len_a, + unsigned char *buf_b, size_t len_b) +{ + size_t offs, len; + + if (len_a > len_b) + offs = len_a - len_b; + else + offs = 0; + + for (; offs < len_a; offs += sizeof(struct branch)) { + len = len_a - offs; + if (!memcmp(buf_a + offs, buf_b, len)) + return buf_b + len; + } + + return buf_b; +} + +static int intel_bts_do_fix_overlap(struct auxtrace_queue *queue, + struct auxtrace_buffer *b) +{ + struct auxtrace_buffer *a; + void *start; + + if (b->list.prev == &queue->head) + return 0; + a = list_entry(b->list.prev, struct auxtrace_buffer, list); + start = intel_bts_find_overlap(a->data, a->size, b->data, b->size); + if (!start) + return -EINVAL; + b->use_size = b->data + b->size - start; + b->use_data = start; + return 0; +} + +static int intel_bts_synth_branch_sample(struct intel_bts_queue *btsq, + struct branch *branch) +{ + int ret; + struct intel_bts *bts = btsq->bts; + union perf_event event; + struct perf_sample sample = { .ip = 0, }; + + event.sample.header.type = PERF_RECORD_SAMPLE; + event.sample.header.misc = PERF_RECORD_MISC_USER; + event.sample.header.size = sizeof(struct perf_event_header); + + sample.ip = le64_to_cpu(branch->from); + sample.pid = btsq->pid; + sample.tid = btsq->tid; + sample.addr = le64_to_cpu(branch->to); + sample.id = btsq->bts->branches_id; + sample.stream_id = btsq->bts->branches_id; + sample.period = 1; + sample.cpu = btsq->cpu; + sample.flags = btsq->sample_flags; + sample.insn_len = btsq->intel_pt_insn.length; + + if (bts->synth_opts.inject) { + event.sample.header.size = bts->branches_event_size; + ret = perf_event__synthesize_sample(&event, + bts->branches_sample_type, + 0, &sample, + bts->synth_needs_swap); + if (ret) + return ret; + } + + ret = perf_session__deliver_synth_event(bts->session, &event, &sample); + if (ret) + pr_err("Intel BTS: failed to deliver branch event, error %d\n", + ret); + + return ret; +} + +static int intel_bts_get_next_insn(struct intel_bts_queue *btsq, u64 ip) +{ + struct machine *machine = btsq->bts->machine; + struct thread *thread; + struct addr_location al; + unsigned char buf[1024]; + size_t bufsz; + ssize_t len; + int x86_64; + uint8_t cpumode; + int err = -1; + + bufsz = intel_pt_insn_max_size(); + + if (machine__kernel_ip(machine, ip)) + cpumode = PERF_RECORD_MISC_KERNEL; + else + cpumode = PERF_RECORD_MISC_USER; + + thread = machine__find_thread(machine, -1, btsq->tid); + if (!thread) + return -1; + + thread__find_addr_map(thread, cpumode, MAP__FUNCTION, ip, &al); + if (!al.map || !al.map->dso) + goto out_put; + + len = dso__data_read_addr(al.map->dso, al.map, machine, ip, buf, bufsz); + if (len <= 0) + goto out_put; + + /* Load maps to ensure dso->is_64_bit has been updated */ + map__load(al.map, machine->symbol_filter); + + x86_64 = al.map->dso->is_64_bit; + + if (intel_pt_get_insn(buf, len, x86_64, &btsq->intel_pt_insn)) + goto out_put; + + err = 0; +out_put: + thread__put(thread); + return err; +} + +static int intel_bts_synth_error(struct intel_bts *bts, int cpu, pid_t pid, + pid_t tid, u64 ip) +{ + union perf_event event; + int err; + + auxtrace_synth_error(&event.auxtrace_error, PERF_AUXTRACE_ERROR_ITRACE, + INTEL_BTS_ERR_NOINSN, cpu, pid, tid, ip, + "Failed to get instruction"); + + err = perf_session__deliver_synth_event(bts->session, &event, NULL); + if (err) + pr_err("Intel BTS: failed to deliver error event, error %d\n", + err); + + return err; +} + +static int intel_bts_get_branch_type(struct intel_bts_queue *btsq, + struct branch *branch) +{ + int err; + + if (!branch->from) { + if (branch->to) + btsq->sample_flags = PERF_IP_FLAG_BRANCH | + PERF_IP_FLAG_TRACE_BEGIN; + else + btsq->sample_flags = 0; + btsq->intel_pt_insn.length = 0; + } else if (!branch->to) { + btsq->sample_flags = PERF_IP_FLAG_BRANCH | + PERF_IP_FLAG_TRACE_END; + btsq->intel_pt_insn.length = 0; + } else { + err = intel_bts_get_next_insn(btsq, branch->from); + if (err) { + btsq->sample_flags = 0; + btsq->intel_pt_insn.length = 0; + if (!btsq->bts->synth_opts.errors) + return 0; + err = intel_bts_synth_error(btsq->bts, btsq->cpu, + btsq->pid, btsq->tid, + branch->from); + return err; + } + btsq->sample_flags = intel_pt_insn_type(btsq->intel_pt_insn.op); + /* Check for an async branch into the kernel */ + if (!machine__kernel_ip(btsq->bts->machine, branch->from) && + machine__kernel_ip(btsq->bts->machine, branch->to) && + btsq->sample_flags != (PERF_IP_FLAG_BRANCH | + PERF_IP_FLAG_CALL | + PERF_IP_FLAG_SYSCALLRET)) + btsq->sample_flags = PERF_IP_FLAG_BRANCH | + PERF_IP_FLAG_CALL | + PERF_IP_FLAG_ASYNC | + PERF_IP_FLAG_INTERRUPT; + } + + return 0; +} + +static int intel_bts_process_buffer(struct intel_bts_queue *btsq, + struct auxtrace_buffer *buffer) +{ + struct branch *branch; + size_t sz, bsz = sizeof(struct branch); + u32 filter = btsq->bts->branches_filter; + int err = 0; + + if (buffer->use_data) { + sz = buffer->use_size; + branch = buffer->use_data; + } else { + sz = buffer->size; + branch = buffer->data; + } + + if (!btsq->bts->sample_branches) + return 0; + + for (; sz > bsz; branch += 1, sz -= bsz) { + if (!branch->from && !branch->to) + continue; + intel_bts_get_branch_type(btsq, branch); + if (filter && !(filter & btsq->sample_flags)) + continue; + err = intel_bts_synth_branch_sample(btsq, branch); + if (err) + break; + } + return err; +} + +static int intel_bts_process_queue(struct intel_bts_queue *btsq, u64 *timestamp) +{ + struct auxtrace_buffer *buffer = btsq->buffer, *old_buffer = buffer; + struct auxtrace_queue *queue; + struct thread *thread; + int err; + + if (btsq->done) + return 1; + + if (btsq->pid == -1) { + thread = machine__find_thread(btsq->bts->machine, -1, + btsq->tid); + if (thread) + btsq->pid = thread->pid_; + } else { + thread = machine__findnew_thread(btsq->bts->machine, btsq->pid, + btsq->tid); + } + + queue = &btsq->bts->queues.queue_array[btsq->queue_nr]; + + if (!buffer) + buffer = auxtrace_buffer__next(queue, NULL); + + if (!buffer) { + if (!btsq->bts->sampling_mode) + btsq->done = 1; + err = 1; + goto out_put; + } + + /* Currently there is no support for split buffers */ + if (buffer->consecutive) { + err = -EINVAL; + goto out_put; + } + + if (!buffer->data) { + int fd = perf_data_file__fd(btsq->bts->session->file); + + buffer->data = auxtrace_buffer__get_data(buffer, fd); + if (!buffer->data) { + err = -ENOMEM; + goto out_put; + } + } + + if (btsq->bts->snapshot_mode && !buffer->consecutive && + intel_bts_do_fix_overlap(queue, buffer)) { + err = -ENOMEM; + goto out_put; + } + + if (!btsq->bts->synth_opts.callchain && thread && + (!old_buffer || btsq->bts->sampling_mode || + (btsq->bts->snapshot_mode && !buffer->consecutive))) + thread_stack__set_trace_nr(thread, buffer->buffer_nr + 1); + + err = intel_bts_process_buffer(btsq, buffer); + + auxtrace_buffer__drop_data(buffer); + + btsq->buffer = auxtrace_buffer__next(queue, buffer); + if (btsq->buffer) { + if (timestamp) + *timestamp = btsq->buffer->reference; + } else { + if (!btsq->bts->sampling_mode) + btsq->done = 1; + } +out_put: + thread__put(thread); + return err; +} + +static int intel_bts_flush_queue(struct intel_bts_queue *btsq) +{ + u64 ts = 0; + int ret; + + while (1) { + ret = intel_bts_process_queue(btsq, &ts); + if (ret < 0) + return ret; + if (ret) + break; + } + return 0; +} + +static int intel_bts_process_tid_exit(struct intel_bts *bts, pid_t tid) +{ + struct auxtrace_queues *queues = &bts->queues; + unsigned int i; + + for (i = 0; i < queues->nr_queues; i++) { + struct auxtrace_queue *queue = &bts->queues.queue_array[i]; + struct intel_bts_queue *btsq = queue->priv; + + if (btsq && btsq->tid == tid) + return intel_bts_flush_queue(btsq); + } + return 0; +} + +static int intel_bts_process_queues(struct intel_bts *bts, u64 timestamp) +{ + while (1) { + unsigned int queue_nr; + struct auxtrace_queue *queue; + struct intel_bts_queue *btsq; + u64 ts = 0; + int ret; + + if (!bts->heap.heap_cnt) + return 0; + + if (bts->heap.heap_array[0].ordinal > timestamp) + return 0; + + queue_nr = bts->heap.heap_array[0].queue_nr; + queue = &bts->queues.queue_array[queue_nr]; + btsq = queue->priv; + + auxtrace_heap__pop(&bts->heap); + + ret = intel_bts_process_queue(btsq, &ts); + if (ret < 0) { + auxtrace_heap__add(&bts->heap, queue_nr, ts); + return ret; + } + + if (!ret) { + ret = auxtrace_heap__add(&bts->heap, queue_nr, ts); + if (ret < 0) + return ret; + } else { + btsq->on_heap = false; + } + } + + return 0; +} + +static int intel_bts_process_event(struct perf_session *session, + union perf_event *event, + struct perf_sample *sample, + struct perf_tool *tool) +{ + struct intel_bts *bts = container_of(session->auxtrace, struct intel_bts, + auxtrace); + u64 timestamp; + int err; + + if (dump_trace) + return 0; + + if (!tool->ordered_events) { + pr_err("Intel BTS requires ordered events\n"); + return -EINVAL; + } + + if (sample->time && sample->time != (u64)-1) + timestamp = perf_time_to_tsc(sample->time, &bts->tc); + else + timestamp = 0; + + err = intel_bts_update_queues(bts); + if (err) + return err; + + err = intel_bts_process_queues(bts, timestamp); + if (err) + return err; + if (event->header.type == PERF_RECORD_EXIT) { + err = intel_bts_process_tid_exit(bts, event->comm.tid); + if (err) + return err; + } + + if (event->header.type == PERF_RECORD_AUX && + (event->aux.flags & PERF_AUX_FLAG_TRUNCATED) && + bts->synth_opts.errors) + err = intel_bts_lost(bts, sample); + + return err; +} + +static int intel_bts_process_auxtrace_event(struct perf_session *session, + union perf_event *event, + struct perf_tool *tool __maybe_unused) +{ + struct intel_bts *bts = container_of(session->auxtrace, struct intel_bts, + auxtrace); + + if (bts->sampling_mode) + return 0; + + if (!bts->data_queued) { + struct auxtrace_buffer *buffer; + off_t data_offset; + int fd = perf_data_file__fd(session->file); + int err; + + if (perf_data_file__is_pipe(session->file)) { + data_offset = 0; + } else { + data_offset = lseek(fd, 0, SEEK_CUR); + if (data_offset == -1) + return -errno; + } + + err = auxtrace_queues__add_event(&bts->queues, session, event, + data_offset, &buffer); + if (err) + return err; + + /* Dump here now we have copied a piped trace out of the pipe */ + if (dump_trace) { + if (auxtrace_buffer__get_data(buffer, fd)) { + intel_bts_dump_event(bts, buffer->data, + buffer->size); + auxtrace_buffer__put_data(buffer); + } + } + } + + return 0; +} + +static int intel_bts_flush(struct perf_session *session __maybe_unused, + struct perf_tool *tool __maybe_unused) +{ + struct intel_bts *bts = container_of(session->auxtrace, struct intel_bts, + auxtrace); + int ret; + + if (dump_trace || bts->sampling_mode) + return 0; + + if (!tool->ordered_events) + return -EINVAL; + + ret = intel_bts_update_queues(bts); + if (ret < 0) + return ret; + + return intel_bts_process_queues(bts, MAX_TIMESTAMP); +} + +static void intel_bts_free_queue(void *priv) +{ + struct intel_bts_queue *btsq = priv; + + if (!btsq) + return; + free(btsq); +} + +static void intel_bts_free_events(struct perf_session *session) +{ + struct intel_bts *bts = container_of(session->auxtrace, struct intel_bts, + auxtrace); + struct auxtrace_queues *queues = &bts->queues; + unsigned int i; + + for (i = 0; i < queues->nr_queues; i++) { + intel_bts_free_queue(queues->queue_array[i].priv); + queues->queue_array[i].priv = NULL; + } + auxtrace_queues__free(queues); +} + +static void intel_bts_free(struct perf_session *session) +{ + struct intel_bts *bts = container_of(session->auxtrace, struct intel_bts, + auxtrace); + + auxtrace_heap__free(&bts->heap); + intel_bts_free_events(session); + session->auxtrace = NULL; + free(bts); +} + +struct intel_bts_synth { + struct perf_tool dummy_tool; + struct perf_session *session; +}; + +static int intel_bts_event_synth(struct perf_tool *tool, + union perf_event *event, + struct perf_sample *sample __maybe_unused, + struct machine *machine __maybe_unused) +{ + struct intel_bts_synth *intel_bts_synth = + container_of(tool, struct intel_bts_synth, dummy_tool); + + return perf_session__deliver_synth_event(intel_bts_synth->session, + event, NULL); +} + +static int intel_bts_synth_event(struct perf_session *session, + struct perf_event_attr *attr, u64 id) +{ + struct intel_bts_synth intel_bts_synth; + + memset(&intel_bts_synth, 0, sizeof(struct intel_bts_synth)); + intel_bts_synth.session = session; + + return perf_event__synthesize_attr(&intel_bts_synth.dummy_tool, attr, 1, + &id, intel_bts_event_synth); +} + +static int intel_bts_synth_events(struct intel_bts *bts, + struct perf_session *session) +{ + struct perf_evlist *evlist = session->evlist; + struct perf_evsel *evsel; + struct perf_event_attr attr; + bool found = false; + u64 id; + int err; + + evlist__for_each(evlist, evsel) { + if (evsel->attr.type == bts->pmu_type && evsel->ids) { + found = true; + break; + } + } + + if (!found) { + pr_debug("There are no selected events with Intel BTS data\n"); + return 0; + } + + memset(&attr, 0, sizeof(struct perf_event_attr)); + attr.size = sizeof(struct perf_event_attr); + attr.type = PERF_TYPE_HARDWARE; + attr.sample_type = evsel->attr.sample_type & PERF_SAMPLE_MASK; + attr.sample_type |= PERF_SAMPLE_IP | PERF_SAMPLE_TID | + PERF_SAMPLE_PERIOD; + attr.sample_type &= ~(u64)PERF_SAMPLE_TIME; + attr.sample_type &= ~(u64)PERF_SAMPLE_CPU; + attr.exclude_user = evsel->attr.exclude_user; + attr.exclude_kernel = evsel->attr.exclude_kernel; + attr.exclude_hv = evsel->attr.exclude_hv; + attr.exclude_host = evsel->attr.exclude_host; + attr.exclude_guest = evsel->attr.exclude_guest; + attr.sample_id_all = evsel->attr.sample_id_all; + attr.read_format = evsel->attr.read_format; + + id = evsel->id[0] + 1000000000; + if (!id) + id = 1; + + if (bts->synth_opts.branches) { + attr.config = PERF_COUNT_HW_BRANCH_INSTRUCTIONS; + attr.sample_period = 1; + attr.sample_type |= PERF_SAMPLE_ADDR; + pr_debug("Synthesizing 'branches' event with id %" PRIu64 " sample type %#" PRIx64 "\n", + id, (u64)attr.sample_type); + err = intel_bts_synth_event(session, &attr, id); + if (err) { + pr_err("%s: failed to synthesize 'branches' event type\n", + __func__); + return err; + } + bts->sample_branches = true; + bts->branches_sample_type = attr.sample_type; + bts->branches_id = id; + /* + * We only use sample types from PERF_SAMPLE_MASK so we can use + * __perf_evsel__sample_size() here. + */ + bts->branches_event_size = sizeof(struct sample_event) + + __perf_evsel__sample_size(attr.sample_type); + } + + bts->synth_needs_swap = evsel->needs_swap; + + return 0; +} + +static const char * const intel_bts_info_fmts[] = { + [INTEL_BTS_PMU_TYPE] = " PMU Type %"PRId64"\n", + [INTEL_BTS_TIME_SHIFT] = " Time Shift %"PRIu64"\n", + [INTEL_BTS_TIME_MULT] = " Time Muliplier %"PRIu64"\n", + [INTEL_BTS_TIME_ZERO] = " Time Zero %"PRIu64"\n", + [INTEL_BTS_CAP_USER_TIME_ZERO] = " Cap Time Zero %"PRId64"\n", + [INTEL_BTS_SNAPSHOT_MODE] = " Snapshot mode %"PRId64"\n", +}; + +static void intel_bts_print_info(u64 *arr, int start, int finish) +{ + int i; + + if (!dump_trace) + return; + + for (i = start; i <= finish; i++) + fprintf(stdout, intel_bts_info_fmts[i], arr[i]); +} + +u64 intel_bts_auxtrace_info_priv[INTEL_BTS_AUXTRACE_PRIV_SIZE]; + +int intel_bts_process_auxtrace_info(union perf_event *event, + struct perf_session *session) +{ + struct auxtrace_info_event *auxtrace_info = &event->auxtrace_info; + size_t min_sz = sizeof(u64) * INTEL_BTS_SNAPSHOT_MODE; + struct intel_bts *bts; + int err; + + if (auxtrace_info->header.size < sizeof(struct auxtrace_info_event) + + min_sz) + return -EINVAL; + + bts = zalloc(sizeof(struct intel_bts)); + if (!bts) + return -ENOMEM; + + err = auxtrace_queues__init(&bts->queues); + if (err) + goto err_free; + + bts->session = session; + bts->machine = &session->machines.host; /* No kvm support */ + bts->auxtrace_type = auxtrace_info->type; + bts->pmu_type = auxtrace_info->priv[INTEL_BTS_PMU_TYPE]; + bts->tc.time_shift = auxtrace_info->priv[INTEL_BTS_TIME_SHIFT]; + bts->tc.time_mult = auxtrace_info->priv[INTEL_BTS_TIME_MULT]; + bts->tc.time_zero = auxtrace_info->priv[INTEL_BTS_TIME_ZERO]; + bts->cap_user_time_zero = + auxtrace_info->priv[INTEL_BTS_CAP_USER_TIME_ZERO]; + bts->snapshot_mode = auxtrace_info->priv[INTEL_BTS_SNAPSHOT_MODE]; + + bts->sampling_mode = false; + + bts->auxtrace.process_event = intel_bts_process_event; + bts->auxtrace.process_auxtrace_event = intel_bts_process_auxtrace_event; + bts->auxtrace.flush_events = intel_bts_flush; + bts->auxtrace.free_events = intel_bts_free_events; + bts->auxtrace.free = intel_bts_free; + session->auxtrace = &bts->auxtrace; + + intel_bts_print_info(&auxtrace_info->priv[0], INTEL_BTS_PMU_TYPE, + INTEL_BTS_SNAPSHOT_MODE); + + if (dump_trace) + return 0; + + if (session->itrace_synth_opts && session->itrace_synth_opts->set) + bts->synth_opts = *session->itrace_synth_opts; + else + itrace_synth_opts__set_default(&bts->synth_opts); + + if (bts->synth_opts.calls) + bts->branches_filter |= PERF_IP_FLAG_CALL | PERF_IP_FLAG_ASYNC | + PERF_IP_FLAG_TRACE_END; + if (bts->synth_opts.returns) + bts->branches_filter |= PERF_IP_FLAG_RETURN | + PERF_IP_FLAG_TRACE_BEGIN; + + err = intel_bts_synth_events(bts, session); + if (err) + goto err_free_queues; + + err = auxtrace_queues__process_index(&bts->queues, session); + if (err) + goto err_free_queues; + + if (bts->queues.populated) + bts->data_queued = true; + + return 0; + +err_free_queues: + auxtrace_queues__free(&bts->queues); + session->auxtrace = NULL; +err_free: + free(bts); + return err; +} diff --git a/tools/perf/util/intel-bts.h b/tools/perf/util/intel-bts.h new file mode 100644 index 000000000000..ca65e21b3e83 --- /dev/null +++ b/tools/perf/util/intel-bts.h @@ -0,0 +1,43 @@ +/* + * intel-bts.h: Intel Processor Trace support + * Copyright (c) 2013-2014, Intel Corporation. + * + * This program is free software; you can redistribute it and/or modify it + * under the terms and conditions of the GNU General Public License, + * version 2, as published by the Free Software Foundation. + * + * This program is distributed in the hope it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for + * more details. + * + */ + +#ifndef INCLUDE__PERF_INTEL_BTS_H__ +#define INCLUDE__PERF_INTEL_BTS_H__ + +#define INTEL_BTS_PMU_NAME "intel_bts" + +enum { + INTEL_BTS_PMU_TYPE, + INTEL_BTS_TIME_SHIFT, + INTEL_BTS_TIME_MULT, + INTEL_BTS_TIME_ZERO, + INTEL_BTS_CAP_USER_TIME_ZERO, + INTEL_BTS_SNAPSHOT_MODE, + INTEL_BTS_AUXTRACE_PRIV_MAX, +}; + +#define INTEL_BTS_AUXTRACE_PRIV_SIZE (INTEL_BTS_AUXTRACE_PRIV_MAX * sizeof(u64)) + +struct auxtrace_record; +struct perf_tool; +union perf_event; +struct perf_session; + +struct auxtrace_record *intel_bts_recording_init(int *err); + +int intel_bts_process_auxtrace_info(union perf_event *event, + struct perf_session *session); + +#endif diff --git a/tools/perf/util/pmu.c b/tools/perf/util/pmu.c index 3c71138e7672..89c91a1a67e7 100644 --- a/tools/perf/util/pmu.c +++ b/tools/perf/util/pmu.c @@ -462,10 +462,6 @@ static struct perf_pmu *pmu_lookup(const char *name) LIST_HEAD(aliases); __u32 type; - /* No support for intel_bts so disallow it */ - if (!strcmp(name, "intel_bts")) - return NULL; - /* * The pmu data we store & need consists of the pmu * type value and format definitions. Load both right From 60b88d8743892218f82048a3df624f5fc5460843 Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Fri, 17 Jul 2015 19:33:44 +0300 Subject: [PATCH 6/9] perf tools: Put itrace options into an asciidoc include perf script, report and inject all have the same itrace options. Put them into an asciidoc include file. Signed-off-by: Adrian Hunter Cc: Jiri Olsa Link: http://lkml.kernel.org/r/1437150840-31811-10-git-send-email-adrian.hunter@intel.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/Documentation/itrace.txt | 22 ++++++++++++++++++++++ tools/perf/Documentation/perf-inject.txt | 23 +---------------------- tools/perf/Documentation/perf-report.txt | 23 +---------------------- tools/perf/Documentation/perf-script.txt | 23 +---------------------- 4 files changed, 25 insertions(+), 66 deletions(-) create mode 100644 tools/perf/Documentation/itrace.txt diff --git a/tools/perf/Documentation/itrace.txt b/tools/perf/Documentation/itrace.txt new file mode 100644 index 000000000000..2ff946677e3b --- /dev/null +++ b/tools/perf/Documentation/itrace.txt @@ -0,0 +1,22 @@ + i synthesize instructions events + b synthesize branches events + c synthesize branches events (calls only) + r synthesize branches events (returns only) + x synthesize transactions events + e synthesize error events + d create a debug log + g synthesize a call chain (use with i or x) + + The default is all events i.e. the same as --itrace=ibxe + + In addition, the period (default 100000) for instructions events + can be specified in units of: + + i instructions + t ticks + ms milliseconds + us microseconds + ns nanoseconds (default) + + Also the call chain size (default 16, max. 1024) for instructions or + transactions events can be specified. diff --git a/tools/perf/Documentation/perf-inject.txt b/tools/perf/Documentation/perf-inject.txt index b876ae312699..0c721c3e37e1 100644 --- a/tools/perf/Documentation/perf-inject.txt +++ b/tools/perf/Documentation/perf-inject.txt @@ -48,28 +48,7 @@ OPTIONS Decode Instruction Tracing data, replacing it with synthesized events. Options are: - i synthesize instructions events - b synthesize branches events - c synthesize branches events (calls only) - r synthesize branches events (returns only) - x synthesize transactions events - e synthesize error events - d create a debug log - g synthesize a call chain (use with i or x) - - The default is all events i.e. the same as --itrace=ibxe - - In addition, the period (default 100000) for instructions events - can be specified in units of: - - i instructions - t ticks - ms milliseconds - us microseconds - ns nanoseconds (default) - - Also the call chain size (default 16, max. 1024) for instructions or - transactions events can be specified. +include::itrace.txt[] SEE ALSO -------- diff --git a/tools/perf/Documentation/perf-report.txt b/tools/perf/Documentation/perf-report.txt index a18ba757a0ed..9c7981bfddad 100644 --- a/tools/perf/Documentation/perf-report.txt +++ b/tools/perf/Documentation/perf-report.txt @@ -331,28 +331,7 @@ OPTIONS --itrace:: Options for decoding instruction tracing data. The options are: - i synthesize instructions events - b synthesize branches events - c synthesize branches events (calls only) - r synthesize branches events (returns only) - x synthesize transactions events - e synthesize error events - d create a debug log - g synthesize a call chain (use with i or x) - - The default is all events i.e. the same as --itrace=ibxe - - In addition, the period (default 100000) for instructions events - can be specified in units of: - - i instructions - t ticks - ms milliseconds - us microseconds - ns nanoseconds (default) - - Also the call chain size (default 16, max. 1024) for instructions or - transactions events can be specified. +include::itrace.txt[] To disable decoding entirely, use --no-itrace. diff --git a/tools/perf/Documentation/perf-script.txt b/tools/perf/Documentation/perf-script.txt index 8e9be1f9c1dd..c0d24791a7f3 100644 --- a/tools/perf/Documentation/perf-script.txt +++ b/tools/perf/Documentation/perf-script.txt @@ -235,28 +235,7 @@ OPTIONS --itrace:: Options for decoding instruction tracing data. The options are: - i synthesize instructions events - b synthesize branches events - c synthesize branches events (calls only) - r synthesize branches events (returns only) - x synthesize transactions events - e synthesize error events - d create a debug log - g synthesize a call chain (use with i or x) - - The default is all events i.e. the same as --itrace=ibxe - - In addition, the period (default 100000) for instructions events - can be specified in units of: - - i instructions - t ticks - ms milliseconds - us microseconds - ns nanoseconds (default) - - Also the call chain size (default 16, max. 1024) for instructions or - transactions events can be specified. +include::itrace.txt[] To disable decoding entirely, use --no-itrace. From 4b715d24f4f14731c7b553cbb8604fe865cb8d3c Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Fri, 17 Jul 2015 19:33:45 +0300 Subject: [PATCH 7/9] perf tools: Add example call-graph script Add a script to produce a call-graph from data exported to a postgresql database and derived from a processor trace event like intel_pt or intel_bts. Refer to comments in the scripts call-graph-from-postgresql.py and export-to-postgresql.py for more details on how to set up the environment, install the required packages, etc. Committer note: From the scripts, for convenience while reading 'git log': An example of using this script with Intel PT: $ perf record -e intel_pt//u ls $ perf script -s ~/libexec/perf-core/scripts/python/export-to-postgresql.py pt_example branches calls 2015-05-29 12:49:23.464364 Creating database... 2015-05-29 12:49:26.281717 Writing to intermediate files... 2015-05-29 12:49:27.190383 Copying to database... 2015-05-29 12:49:28.140451 Removing intermediate files... 2015-05-29 12:49:28.147451 Adding primary keys 2015-05-29 12:49:28.655683 Adding foreign keys 2015-05-29 12:49:29.365350 Done $ python tools/perf/scripts/python/call-graph-from-postgresql.py pt_example # The result is a GUI window with a tree representing a context-sensitive # call-graph. Expanding a couple of levels of the tree and adjusting column # widths to suit will display something like: Call Graph: pt_example Call Path |Object |Count|Time(ns)|Time(%)|Branch Count|Branch Count(%) v- ls v- 2638:2638 v- _start ld-2.19.so 1 10074071 100.0 211135 100.0 |- unknown unknown 1 13198 0.1 1 0.0 >- _dl_start ld-2.19.so 1 1400980 13.9 19637 9.3 >- _d_linit_internal ld-2.19.so 1 448152 4.4 11094 5.3 v-__libc_start_main@plt ls 1 8211741 81.5 180397 85.4 >- _dl_fixup ld-2.19.so 1 7607 0.1 108 0.1 >- __cxa_atexit libc-2.19.so 1 11737 0.1 10 0.0 >- __libc_csu_init ls 1 10354 0.1 10 0.0 |- _setjmp libc-2.19.so 1 0 0.0 4 0.0 v- main ls 1 8182043 99.6 180254 99.9 Signed-off-by: Adrian Hunter Tested-by: Arnaldo Carvalho de Melo Cc: Jiri Olsa Link: http://lkml.kernel.org/r/1437150840-31811-11-git-send-email-adrian.hunter@intel.com [ Added 'python-pyside qt-postgresql' to the yum cmdline installing required packages ] Signed-off-by: Arnaldo Carvalho de Melo --- .../python/call-graph-from-postgresql.py | 327 ++++++++++++++++++ .../scripts/python/export-to-postgresql.py | 47 +++ 2 files changed, 374 insertions(+) create mode 100644 tools/perf/scripts/python/call-graph-from-postgresql.py diff --git a/tools/perf/scripts/python/call-graph-from-postgresql.py b/tools/perf/scripts/python/call-graph-from-postgresql.py new file mode 100644 index 000000000000..e78fdc2a5a9d --- /dev/null +++ b/tools/perf/scripts/python/call-graph-from-postgresql.py @@ -0,0 +1,327 @@ +#!/usr/bin/python2 +# call-graph-from-postgresql.py: create call-graph from postgresql database +# Copyright (c) 2014, Intel Corporation. +# +# This program is free software; you can redistribute it and/or modify it +# under the terms and conditions of the GNU General Public License, +# version 2, as published by the Free Software Foundation. +# +# This program is distributed in the hope it will be useful, but WITHOUT +# ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or +# FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for +# more details. + +# To use this script you will need to have exported data using the +# export-to-postgresql.py script. Refer to that script for details. +# +# Following on from the example in the export-to-postgresql.py script, a +# call-graph can be displayed for the pt_example database like this: +# +# python tools/perf/scripts/python/call-graph-from-postgresql.py pt_example +# +# Note this script supports connecting to remote databases by setting hostname, +# port, username, password, and dbname e.g. +# +# python tools/perf/scripts/python/call-graph-from-postgresql.py "hostname=myhost username=myuser password=mypassword dbname=pt_example" +# +# The result is a GUI window with a tree representing a context-sensitive +# call-graph. Expanding a couple of levels of the tree and adjusting column +# widths to suit will display something like: +# +# Call Graph: pt_example +# Call Path Object Count Time(ns) Time(%) Branch Count Branch Count(%) +# v- ls +# v- 2638:2638 +# v- _start ld-2.19.so 1 10074071 100.0 211135 100.0 +# |- unknown unknown 1 13198 0.1 1 0.0 +# >- _dl_start ld-2.19.so 1 1400980 13.9 19637 9.3 +# >- _d_linit_internal ld-2.19.so 1 448152 4.4 11094 5.3 +# v-__libc_start_main@plt ls 1 8211741 81.5 180397 85.4 +# >- _dl_fixup ld-2.19.so 1 7607 0.1 108 0.1 +# >- __cxa_atexit libc-2.19.so 1 11737 0.1 10 0.0 +# >- __libc_csu_init ls 1 10354 0.1 10 0.0 +# |- _setjmp libc-2.19.so 1 0 0.0 4 0.0 +# v- main ls 1 8182043 99.6 180254 99.9 +# +# Points to note: +# The top level is a command name (comm) +# The next level is a thread (pid:tid) +# Subsequent levels are functions +# 'Count' is the number of calls +# 'Time' is the elapsed time until the function returns +# Percentages are relative to the level above +# 'Branch Count' is the total number of branches for that function and all +# functions that it calls + +import sys +from PySide.QtCore import * +from PySide.QtGui import * +from PySide.QtSql import * +from decimal import * + +class TreeItem(): + + def __init__(self, db, row, parent_item): + self.db = db + self.row = row + self.parent_item = parent_item + self.query_done = False; + self.child_count = 0 + self.child_items = [] + self.data = ["", "", "", "", "", "", ""] + self.comm_id = 0 + self.thread_id = 0 + self.call_path_id = 1 + self.branch_count = 0 + self.time = 0 + if not parent_item: + self.setUpRoot() + + def setUpRoot(self): + self.query_done = True + query = QSqlQuery(self.db) + ret = query.exec_('SELECT id, comm FROM comms') + if not ret: + raise Exception("Query failed: " + query.lastError().text()) + while query.next(): + if not query.value(0): + continue + child_item = TreeItem(self.db, self.child_count, self) + self.child_items.append(child_item) + self.child_count += 1 + child_item.setUpLevel1(query.value(0), query.value(1)) + + def setUpLevel1(self, comm_id, comm): + self.query_done = True; + self.comm_id = comm_id + self.data[0] = comm + self.child_items = [] + self.child_count = 0 + query = QSqlQuery(self.db) + ret = query.exec_('SELECT thread_id, ( SELECT pid FROM threads WHERE id = thread_id ), ( SELECT tid FROM threads WHERE id = thread_id ) FROM comm_threads WHERE comm_id = ' + str(comm_id)) + if not ret: + raise Exception("Query failed: " + query.lastError().text()) + while query.next(): + child_item = TreeItem(self.db, self.child_count, self) + self.child_items.append(child_item) + self.child_count += 1 + child_item.setUpLevel2(comm_id, query.value(0), query.value(1), query.value(2)) + + def setUpLevel2(self, comm_id, thread_id, pid, tid): + self.comm_id = comm_id + self.thread_id = thread_id + self.data[0] = str(pid) + ":" + str(tid) + + def getChildItem(self, row): + return self.child_items[row] + + def getParentItem(self): + return self.parent_item + + def getRow(self): + return self.row + + def timePercent(self, b): + if not self.time: + return "0.0" + x = (b * Decimal(100)) / self.time + return str(x.quantize(Decimal('.1'), rounding=ROUND_HALF_UP)) + + def branchPercent(self, b): + if not self.branch_count: + return "0.0" + x = (b * Decimal(100)) / self.branch_count + return str(x.quantize(Decimal('.1'), rounding=ROUND_HALF_UP)) + + def addChild(self, call_path_id, name, dso, count, time, branch_count): + child_item = TreeItem(self.db, self.child_count, self) + child_item.comm_id = self.comm_id + child_item.thread_id = self.thread_id + child_item.call_path_id = call_path_id + child_item.branch_count = branch_count + child_item.time = time + child_item.data[0] = name + if dso == "[kernel.kallsyms]": + dso = "[kernel]" + child_item.data[1] = dso + child_item.data[2] = str(count) + child_item.data[3] = str(time) + child_item.data[4] = self.timePercent(time) + child_item.data[5] = str(branch_count) + child_item.data[6] = self.branchPercent(branch_count) + self.child_items.append(child_item) + self.child_count += 1 + + def selectCalls(self): + self.query_done = True; + query = QSqlQuery(self.db) + ret = query.exec_('SELECT id, call_path_id, branch_count, call_time, return_time, ' + '( SELECT name FROM symbols WHERE id = ( SELECT symbol_id FROM call_paths WHERE id = call_path_id ) ), ' + '( SELECT short_name FROM dsos WHERE id = ( SELECT dso_id FROM symbols WHERE id = ( SELECT symbol_id FROM call_paths WHERE id = call_path_id ) ) ), ' + '( SELECT ip FROM call_paths where id = call_path_id ) ' + 'FROM calls WHERE parent_call_path_id = ' + str(self.call_path_id) + ' AND comm_id = ' + str(self.comm_id) + ' AND thread_id = ' + str(self.thread_id) + + 'ORDER BY call_path_id') + if not ret: + raise Exception("Query failed: " + query.lastError().text()) + last_call_path_id = 0 + name = "" + dso = "" + count = 0 + branch_count = 0 + total_branch_count = 0 + time = 0 + total_time = 0 + while query.next(): + if query.value(1) == last_call_path_id: + count += 1 + branch_count += query.value(2) + time += query.value(4) - query.value(3) + else: + if count: + self.addChild(last_call_path_id, name, dso, count, time, branch_count) + last_call_path_id = query.value(1) + name = query.value(5) + dso = query.value(6) + count = 1 + total_branch_count += branch_count + total_time += time + branch_count = query.value(2) + time = query.value(4) - query.value(3) + if count: + self.addChild(last_call_path_id, name, dso, count, time, branch_count) + total_branch_count += branch_count + total_time += time + # Top level does not have time or branch count, so fix that here + if total_branch_count > self.branch_count: + self.branch_count = total_branch_count + if self.branch_count: + for child_item in self.child_items: + child_item.data[6] = self.branchPercent(child_item.branch_count) + if total_time > self.time: + self.time = total_time + if self.time: + for child_item in self.child_items: + child_item.data[4] = self.timePercent(child_item.time) + + def childCount(self): + if not self.query_done: + self.selectCalls() + return self.child_count + + def columnCount(self): + return 7 + + def columnHeader(self, column): + headers = ["Call Path", "Object", "Count ", "Time (ns) ", "Time (%) ", "Branch Count ", "Branch Count (%) "] + return headers[column] + + def getData(self, column): + return self.data[column] + +class TreeModel(QAbstractItemModel): + + def __init__(self, db, parent=None): + super(TreeModel, self).__init__(parent) + self.db = db + self.root = TreeItem(db, 0, None) + + def columnCount(self, parent): + return self.root.columnCount() + + def rowCount(self, parent): + if parent.isValid(): + parent_item = parent.internalPointer() + else: + parent_item = self.root + return parent_item.childCount() + + def headerData(self, section, orientation, role): + if role == Qt.TextAlignmentRole: + if section > 1: + return Qt.AlignRight + if role != Qt.DisplayRole: + return None + if orientation != Qt.Horizontal: + return None + return self.root.columnHeader(section) + + def parent(self, child): + child_item = child.internalPointer() + if child_item is self.root: + return QModelIndex() + parent_item = child_item.getParentItem() + return self.createIndex(parent_item.getRow(), 0, parent_item) + + def index(self, row, column, parent): + if parent.isValid(): + parent_item = parent.internalPointer() + else: + parent_item = self.root + child_item = parent_item.getChildItem(row) + return self.createIndex(row, column, child_item) + + def data(self, index, role): + if role == Qt.TextAlignmentRole: + if index.column() > 1: + return Qt.AlignRight + if role != Qt.DisplayRole: + return None + index_item = index.internalPointer() + return index_item.getData(index.column()) + +class MainWindow(QMainWindow): + + def __init__(self, db, dbname, parent=None): + super(MainWindow, self).__init__(parent) + + self.setObjectName("MainWindow") + self.setWindowTitle("Call Graph: " + dbname) + self.move(100, 100) + self.resize(800, 600) + style = self.style() + icon = style.standardIcon(QStyle.SP_MessageBoxInformation) + self.setWindowIcon(icon); + + self.model = TreeModel(db) + + self.view = QTreeView() + self.view.setModel(self.model) + + self.setCentralWidget(self.view) + +if __name__ == '__main__': + if (len(sys.argv) < 2): + print >> sys.stderr, "Usage is: call-graph-from-postgresql.py " + raise Exception("Too few arguments") + + dbname = sys.argv[1] + + db = QSqlDatabase.addDatabase('QPSQL') + + opts = dbname.split() + for opt in opts: + if '=' in opt: + opt = opt.split('=') + if opt[0] == 'hostname': + db.setHostName(opt[1]) + elif opt[0] == 'port': + db.setPort(int(opt[1])) + elif opt[0] == 'username': + db.setUserName(opt[1]) + elif opt[0] == 'password': + db.setPassword(opt[1]) + elif opt[0] == 'dbname': + dbname = opt[1] + else: + dbname = opt + + db.setDatabaseName(dbname) + if not db.open(): + raise Exception("Failed to open database " + dbname + " error: " + db.lastError().text()) + + app = QApplication(sys.argv) + window = MainWindow(db, dbname) + window.show() + err = app.exec_() + db.close() + sys.exit(err) diff --git a/tools/perf/scripts/python/export-to-postgresql.py b/tools/perf/scripts/python/export-to-postgresql.py index 4cdafd880074..84a32037a80f 100644 --- a/tools/perf/scripts/python/export-to-postgresql.py +++ b/tools/perf/scripts/python/export-to-postgresql.py @@ -15,6 +15,53 @@ import sys import struct import datetime +# To use this script you will need to have installed package python-pyside which +# provides LGPL-licensed Python bindings for Qt. You will also need the package +# libqt4-sql-psql for Qt postgresql support. +# +# The script assumes postgresql is running on the local machine and that the +# user has postgresql permissions to create databases. Examples of installing +# postgresql and adding such a user are: +# +# fedora: +# +# $ sudo yum install postgresql postgresql-server python-pyside qt-postgresql +# $ sudo su - postgres -c initdb +# $ sudo service postgresql start +# $ sudo su - postgres +# $ createuser +# Shall the new role be a superuser? (y/n) y +# +# ubuntu: +# +# $ sudo apt-get install postgresql +# $ sudo su - postgres +# $ createuser +# Shall the new role be a superuser? (y/n) y +# +# An example of using this script with Intel PT: +# +# $ perf record -e intel_pt//u ls +# $ perf script -s ~/libexec/perf-core/scripts/python/export-to-postgresql.py pt_example branches calls +# 2015-05-29 12:49:23.464364 Creating database... +# 2015-05-29 12:49:26.281717 Writing to intermediate files... +# 2015-05-29 12:49:27.190383 Copying to database... +# 2015-05-29 12:49:28.140451 Removing intermediate files... +# 2015-05-29 12:49:28.147451 Adding primary keys +# 2015-05-29 12:49:28.655683 Adding foreign keys +# 2015-05-29 12:49:29.365350 Done +# +# To browse the database, psql can be used e.g. +# +# $ psql pt_example +# pt_example=# select * from samples_view where id < 100; +# pt_example=# \d+ +# pt_example=# \d+ samples_view +# pt_example=# \q +# +# An example of using the database is provided by the script +# call-graph-from-postgresql.py. Refer to that script for details. + from PySide.QtSql import * # Need to access PostgreSQL C library directly to use COPY FROM STDIN From 66671d001aeb525b9101e0ccb9062627539de555 Mon Sep 17 00:00:00 2001 From: Arnaldo Carvalho de Melo Date: Tue, 18 Aug 2015 15:19:50 -0300 Subject: [PATCH 8/9] perf tools: Initialize reference counts in map__clone() Map clone was written before we introduced reference counts for maps and dsos, so all that was needed was just a copy and then we would insert it into the new map_groups instance. Fix it by, after copying, initializing the map->refcnt, grabbing a struct dso refcount and resetting pointers that may be used to determine if a map, when deleted, is in a rb_tree. Cc: Adrian Hunter Cc: Borislav Petkov Cc: David Ahern Cc: Frederic Weisbecker Cc: Jiri Olsa Cc: Namhyung Kim Cc: Stephane Eranian Link: http://lkml.kernel.org/n/tip-pd4mr80o5b9gvk50iineacec@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/map.c | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-) diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c index ce37e95bc513..b1c475d9b240 100644 --- a/tools/perf/util/map.c +++ b/tools/perf/util/map.c @@ -348,9 +348,18 @@ struct symbol *map__find_symbol_by_name(struct map *map, const char *name, return dso__find_symbol_by_name(map->dso, map->type, name); } -struct map *map__clone(struct map *map) +struct map *map__clone(struct map *from) { - return memdup(map, sizeof(*map)); + struct map *map = memdup(from, sizeof(*map)); + + if (map != NULL) { + atomic_set(&map->refcnt, 1); + RB_CLEAR_NODE(&map->rb_node); + dso__get(map->dso); + map->groups = NULL; + } + + return map; } int map__overlap(struct map *l, struct map *r) From 1c0bd0e891aaed0219010bfe79b32e1b0b82d662 Mon Sep 17 00:00:00 2001 From: Wang Nan Date: Fri, 21 Aug 2015 10:09:02 +0000 Subject: [PATCH 9/9] perf probe: Try to use symbol table if searching debug info failed A problem can occur in a statically linked perf when vmlinux can be found: # perf probe --add sys_epoll_pwait probe-definition(0): sys_epoll_pwait symbol:sys_epoll_pwait file:(null) line:0 offset:0 return:0 lazy:(null) 0 arguments Looking at the vmlinux_path (7 entries long) Using /lib/modules/4.2.0-rc1+/build/vmlinux for symbols Open Debuginfo file: /lib/modules/4.2.0-rc1+/build/vmlinux Try to find probe point from debuginfo. Symbol sys_epoll_pwait address found : ffffffff8122bd40 Matched function: SyS_epoll_pwait Failed to get call frame on 0xffffffff8122bd40 An error occurred in debuginfo analysis (-2). Error: Failed to add events. Reason: No such file or directory (Code: -2) The reason is caused by libdw that, if libdw is statically linked, it can't load libebl_{arch}.so reliable. In this case it is still possible to get the address from /proc/kalksyms. However, perf tries that only when libdw returns -EBADF. This patch gives it another chance to utilize symbol table, even if libdw returns an error code other than -EBADF. After applying this patch: # perf probe -nv --add sys_epoll_pwait probe-definition(0): sys_epoll_pwait symbol:sys_epoll_pwait file:(null) line:0 offset:0 return:0 lazy:(null) 0 arguments Looking at the vmlinux_path (7 entries long) Using /lib/modules/4.2.0-rc1+/build/vmlinux for symbols Open Debuginfo file: /lib/modules/4.2.0-rc1+/build/vmlinux Try to find probe point from debuginfo. Symbol sys_epoll_pwait address found : ffffffff8122bd40 Matched function: SyS_epoll_pwait Failed to get call frame on 0xffffffff8122bd40 An error occurred in debuginfo analysis (-2). Trying to use symbols. Opening /sys/kernel/debug/tracing/kprobe_events write=1 Added new event: Writing event: p:probe/sys_epoll_pwait _text+2276672 probe:sys_epoll_pwait (on sys_epoll_pwait) You can now use it in all perf tools, such as: perf record -e probe:sys_epoll_pwait -aR sleep 1 Although libdw returns an error (Failed to get call frame), perf tries symbol table and finally gets correct address. Signed-off-by: Wang Nan Cc: Alexei Starovoitov Cc: Brendan Gregg Cc: Daniel Borkmann Cc: David Ahern Cc: He Kuang Cc: Jiri Olsa Cc: Kaixu Xia Cc: Masami Hiramatsu Cc: Namhyung Kim Cc: Peter Zijlstra Cc: Zefan Li Cc: pi3orama@163.com Link: http://lkml.kernel.org/r/1440151770-129878-2-git-send-email-wangnan0@huawei.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/util/probe-event.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/tools/perf/util/probe-event.c b/tools/perf/util/probe-event.c index fe4941a94a25..f07374bc9c5a 100644 --- a/tools/perf/util/probe-event.c +++ b/tools/perf/util/probe-event.c @@ -705,9 +705,10 @@ static int try_to_find_probe_trace_events(struct perf_probe_event *pev, } /* Error path : ntevs < 0 */ pr_debug("An error occurred in debuginfo analysis (%d).\n", ntevs); - if (ntevs == -EBADF) { - pr_warning("Warning: No dwarf info found in the vmlinux - " - "please rebuild kernel with CONFIG_DEBUG_INFO=y.\n"); + if (ntevs < 0) { + if (ntevs == -EBADF) + pr_warning("Warning: No dwarf info found in the vmlinux - " + "please rebuild kernel with CONFIG_DEBUG_INFO=y.\n"); if (!need_dwarf) { pr_debug("Trying to use symbols.\n"); return 0;