- 14 Dec, 2022 34 commits
-
-
Petar Gligoric authored
This patch adds the possibility to write the trace and the summary as csv files to a user specified file. A format as such simplifies further data processing. This is achieved by having ";" as separators instead of spaces and solely one header per file. Additional parameters are being considered, like in the normal usage of the script. Colors are turned off in the case of a csv output, thus the highlight option is also being ignored. Usage: Write standard task to csv file: $ perf script report tasks-analyzer --csv <file> write limited output to csv file in nanoseconds: $ perf script report tasks-analyzer --csv <file> --ns --limit-to-tasks 1337 Write summary to a csv file: $ perf script report tasks-analyzer --csv-summary <file> Write summary to csv file with additional schedule information: $ perf script report tasks-analyzer --csv-summary <file> --summary-extended Write both summary and standard task to a csv file: $ perf script report tasks-analyzer --csv --csv-summary The following examples illustrate what is possible with the CSV output. The first command sequence will record all scheduler switch events for 10 seconds, the task-analyzer calculates task information like runtimes as CSV. A small python snippet using pandas and matplotlib will visualize the most frequent task (e.g. kworker/1:1) runtimes - each runtime as a bar in a bar chart: $ perf record -e sched:sched_switch -a -- sleep 10 $ perf script report tasks-analyzer --ns --csv tasks.csv $ cat << EOF > /tmp/freq-comm-runtimes-bar.py import pandas as pd import matplotlib.pyplot as plt df = pd.read_csv("tasks.csv", sep=';') most_freq_comm = df["COMM"].value_counts().idxmax() most_freq_runtimes = df[df["COMM"]==most_freq_comm]["Runtime"] plt.title(f"Runtimes for Task {most_freq_comm} in Nanoseconds") plt.bar(range(len(most_freq_runtimes)), most_freq_runtimes) plt.show() $ python3 /tmp/freq-comm-runtimes-bar.py As a seconds example, the subsequent script generates a pie chart of all accumulated tasks runtimes for 10 seconds of system recordings: $ perf record -e sched:sched_switch -a -- sleep 10 $ perf script report tasks-analyzer --csv-summary task-summary.csv $ cat << EOF > /tmp/accumulated-task-pie.py import pandas as pd from matplotlib.pyplot import pie, axis, show df = pd.read_csv("task-summary.csv", sep=';') sums = df.groupby(df["Comm"])["Accumulated"].sum() axis("equal") pie(sums, labels=sums.index); show() EOF $ python3 /tmp/accumulated-task-pie.py A variety of other visualizations are possible in matplotlib and other environments. Of course, pandas, numpy and co. also allow easy statistical analysis of the data! Signed-off-by:Petar Gligoric <petar.gligoric@rohde-schwarz.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Ian Rogers <irogers@google.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: https://lore.kernel.org/r/20221206154406.41941-3-petar.gligor@gmail.comSigned-off-by:
Hagen Paul Pfeifer <hagen@jauu.net> Signed-off-by:
Arnaldo Carvalho de Melo <acme@redhat.com>
-
Hagen Paul Pfeifer authored
Introduce a new 'perf script' to analyze task scheduling behavior. During the task analysis, some data is always needed - which goes beyond the simple time of switching on and off a task (process/thread). This concerns for example the runtime of a process or the frequency with which the process was called. This script serves to simplify this recurring analyze process. It immediately provides the user with helpful task characteristic information about the tasks runtimes. Usage: Recorded can be in two ways: $ perf script record tasks-analyzer -- sleep 10 $ perf record -e sched:sched_switch -a -- sleep 10 The script can parse all perf.data files, most important: sched:sched_switch events are mandatory, other events will be ignored. Most simple report use case is to just call the script without arguments: $ perf script report tasks-analyzer Switched-In Switched-Out CPU PID TID Comm Runtime Time Out-In 15576.658891407 15576.659156086 4 2412 2428 gdbus 265 1949 15576.659111320 15576.659455410 0 2412 2412 gnome-shell 344 2267 15576.659491326 15576.659506173 2 74 74 kworker/2:1 15 13145 15576.659506173 15576.659825748 2 2858 2858 gnome-terminal- 320 63263 15576.659871270 15576.659902872 6 20932 20932 kworker/u16:0 32 2314582 15576.659909951 15576.659945501 3 27264 27264 sh 36 -1 15576.659853285 15576.659971052 7 27265 27265 perf 118 5050741 [...] What is not shown here are the ASCII color sequences. For example, if the task consists of only one thread, the TID is grayed out. Runtime is the time the task was running on the CPU, Time Out-In is the time between the process being scheduled *out* and scheduled back *in*. So the last time span between two executions. If -1 is printed, then the task simply ran the first time in the measurements - a Out-In delta could not be calculated. In addition to the chronological representation, there is a summary on task level. This output can be additionally switched on via the --summary option and provides information such as max, min & average runtime per process. The maximum runtime is often important for debugging. The call looks like this: $ perf script report tasks-analyzer --summary Summary Task Information Runtime Information PID TID Comm Runs Accumulated Mean Median Min Max Max At 14 14 ksoftirqd/0 13 334 26 15 9 127 15571.621211956 15 15 rcu_preempt 133 1778 13 13 2 33 15572.581176024 16 16 migration/0 3 49 16 13 12 24 15571.608915425 20 20 migration/1 3 34 11 13 8 13 15571.639101555 25 25 migration/2 3 32 11 12 9 12 15575.639239896 [...] Besides these two options, there are a number of other options that change the output and behavior. This can be queried via --help. Options worth mentioning include: - filter-tasks - filter out unneeded tasks, --filter-task 1337,/sbin/init - highlight-tasks - more pleasant focusing, --highlight-tasks 1:red,mutt:yellow - extended-times - show combinations of elapsed times between schedule in/schedule out - summary-extended - summary with additional information, like maximum delta time statistics - rename-comms-by-tids - handy for inexpressive processnames like python, --rename 1337:my-python-app - ms - show timestamps in milliseconds, nanoseconds is also possible (--ns) - time-limit - limit the analyzer to a time range, --time-limit 15576.0:15576.1 Script is tested and prime time ready for python2 & python3: - make PYTHON=python3 prefix=/usr/local install - make PYTHON=python2 prefix=/usr/local install Signed-off-by: -
James Clark authored
Printing the info doesn't have any dependency on OpenCSD, and neither does recording Coresight data. Because it's sometimes useful to look at the info for debugging, it makes sense to be able to see it on the same platform that the recording was made on. So pull the auxtrace info printing parts into a new file that is always compiled into Perf. Signed-off-by:
James Clark <james.clark@arm.com> Cc: Al Grant <Al.Grant@arm.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: John Garry <john.g.garry@oracle.com> Cc: Leo Yan <leo.yan@linaro.org> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Mathieu Poirier <mathieu.poirier@linaro.org> Cc: Mike Leach <mike.leach@linaro.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Suzuki Poulouse <suzuki.poulose@arm.com> Cc: Will Deacon <will@kernel.org> Cc: coresight@lists.linaro.org Cc: linux-arm-kernel@lists.infradead.org Link: https://lore.kernel.org/r/20221212155513.2259623-6-james.clark@arm.comSigned-off-by:
-
James Clark authored
hdr is a copy of 3 values of ptr and doesn't need to be long lived. So just use ptr instead which means the malloc and the extra error path can be removed to simplify things. Signed-off-by:
-
James Clark authored
cs_etm__print_auxtrace_info() is called twice in case there is an error somewhere in cs_etm__process_auxtrace_info(), but all the info is already available at the beginning so just print it there instead. Also use u64 and the already cast ptr variable to make it more consistent with the rest of the etm code. Signed-off-by:
-
James Clark authored
These aren't used outside of cs-etm so don't need stubs. Leave cs_etm__process_auxtrace_info() which is used externally, and add an error message so that it's obvious to users why it causes errors. Signed-off-by:
-
James Clark authored
This is an error rather than just for the raw trace dump so always print it as an error. Also remove the duplicate header version check. Signed-off-by:
-
Namhyung Kim authored
Add test cases for the task and addr aggregation modes. $ sudo ./perf test -v contention 86: kernel lock contention analysis test : --- start --- test child forked, pid 680006 Testing perf lock record and perf lock contention Testing perf lock contention --use-bpf Testing perf lock record and perf lock contention at the same time Testing perf lock contention --threads Testing perf lock contention --lock-addr test child finished with 0 ---- end ---- kernel lock contention analysis test: Ok Signed-off-by:
Namhyung Kim <namhyung@kernel.org> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Blake Jones <blakejones@google.com> Cc: Ian Rogers <irogers@google.com> Cc: Ingo Molnar <mingo@kernel.org> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Song Liu <song@kernel.org> Cc: bpf@vger.kernel.org Link: https://lore.kernel.org/r/20221209190727.759804-5-namhyung@kernel.orgSigned-off-by:
-
Namhyung Kim authored
The -l/--lock-addr option is to implement per-lock-instance contention stat using LOCK_AGGR_ADDR. It displays lock address and optionally symbol name if exists. $ sudo ./perf lock con -abl sleep 1 contended total wait max wait avg wait address symbol 1 36.28 us 36.28 us 36.28 us ffff92615d6448b8 9 10.91 us 1.84 us 1.21 us ffffffffbaed50c0 rcu_state 1 10.49 us 10.49 us 10.49 us ffff9262ac4f0c80 8 4.68 us 1.67 us 585 ns ffffffffbae07a40 jiffies_lock 3 3.03 us 1.45 us 1.01 us ffff9262277861e0 1 924 ns 924 ns 924 ns ffff926095ba9d20 1 436 ns 436 ns 436 ns ffff9260bfda4f60 Signed-off-by: -
Namhyung Kim authored
The BPF didn't show the per-thread stat properly. Use task's thread id (PID) as a key instead of stack_id and add a task_data map to save task comm names. $ sudo ./perf lock con -abt -E 5 sleep 1 contended total wait max wait avg wait pid comm 1 740.66 ms 740.66 ms 740.66 ms 1950 nv_queue 3 305.50 ms 298.19 ms 101.83 ms 1884 nvidia-modeset/ 1 25.14 us 25.14 us 25.14 us 2725038 EventManager_De 12 23.09 us 9.30 us 1.92 us 0 swapper 1 20.18 us 20.18 us 20.18 us 2725033 EventManager_De Signed-off-by: -
Namhyung Kim authored
Accessing BPF maps should use the same data types. Add bpf_skel/lock_data.h to define the common data structures. No functional changes. Committer notes: Fixed contention_key.stack_id missing rename to contention_key.stack_or_task_id. Signed-off-by:
-
Khem Raj authored
Sometimes build systems may append options e.g. --sysroot etc. to CC variable especially in cross-compile environments like yocto project where CC varable is composed of cross-compiler name and some needed options for it to work in a relocatable environment. Therefore separate out the compiler name from rest of the options in CC, then add the options via second argument to Popen() API Reviewed-by:
Florian Fainelli <f.fainelli@gmail.com> Signed-off-by:
Khem Raj <raj.khem@gmail.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Fangrui Song <maskray@google.com> Cc: Florian Fainelli <f.fainelli@gmail.com> Cc: Ian Rogers <irogers@google.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: John Keeping <john@metanate.com> Cc: Leo Yan <leo.yan@linaro.org> Cc: Michael Petlan <mpetlan@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Nathan Chancellor <nathan@kernel.org> Cc: Nick Desaulniers <ndesaulniers@google.com> Cc: Sedat Dilek <sedat.dilek@gmail.com> Link: https://lore.kernel.org/r/20221205025534.150006-1-raj.khem@gmail.comSigned-off-by:
-
Namhyung Kim authored
In BTF, tracepoint definitions have the "btf_trace_" prefix. The off-cpu profiler needs to check the signature of the sched_switch event using that definition. But there's a typo (s/bpf/btf/) so it failed always. Fixes: b36888f7 ("perf record: Handle argument change in sched_switch") Signed-off-by:
-
Athira Rajeev authored
The event group test checks group creation for combinations of hw, sw and uncore PMU events. Some of the uncore pmus may require additional permission to access the counters. For example, in case of hv_24x7, partition need to have permissions to access hv_24x7 pmu counters. If not, event_open will fail. Hence add a sanity check to see if event_open succeeds before proceeding with the test. Fixes: 9d9b22be ("perf test: Add event group test for events in multiple PMUs") Signed-off-by:
Athira Jajeev <atrajeev@linux.vnet.ibm.com> Acked-by:
Ravi Bangoria <ravi.bangoria@amd.com> Cc: Disha Goel <disgoel@linux.ibm.com> Cc: Ian Rogers <irogers@google.com> Cc: James Clark <james.clark@arm.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kajol Jain <kjain@linux.ibm.com> Cc: linuxppc-dev@lists.ozlabs.org Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.com> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Nageswara R Sastry <rnsastry@linux.ibm.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Ravi Bangoria <ravi.bangoria@amd.com> Link: https://lore.kernel.org/r/20221207165815.774-1-atrajeev@linux.vnet.ibm.comSigned-off-by:
-
Arnaldo Carvalho de Melo authored
Some distros have older versions of libtraceevent where TEP_FIELD_IS_RELATIVE and its associated semantics are not present, so we need to check if the version has it, it was introduced in libtraceevent 1.5.0. Reported-by:
-
Ian Rogers authored
libtraceevent is now out-of-date and it is better to depend on the system version. Remove this code that is no longer depended upon by any builds. Committer notes: Removed the removed tools/lib/traceevent/ from tools/perf/MANIFEST, so that 'make perf-tar-src-pkg' works. Signed-off-by:
Ian Rogers <irogers@google.com> Acked-by:
Steven Rostedt (VMware) <rostedt@goodmis.org> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kan Liang <kan.liang@linux.intel.com> Cc: Leo Yan <leo.yan@linaro.org> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Masami Hiramatsu <mhiramat@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lore.kernel.org/lkml/20221130062935.2219247-5-irogers@google.comSigned-off-by:
-
Ian Rogers authored
Remove the LIBTRACEEVENT_DYNAMIC and LIBTRACEFS_DYNAMIC make command line variables. If libtraceevent isn't installed or NO_LIBTRACEEVENT=1 is passed to the build, don't compile in libtraceevent and libtracefs support. This also disables CONFIG_TRACE that controls "perf trace". CONFIG_LIBTRACEEVENT is used to control enablement in Build/Makefiles, HAVE_LIBTRACEEVENT is used in C code. Without HAVE_LIBTRACEEVENT tracepoints are disabled and as such the commands kmem, kwork, lock, sched and timechart are removed. The majority of commands continue to work including "perf test". Committer notes: Fixed up a tools/perf/util/Build reject and added: #include <traceevent/event-parse.h> to tools/perf/util/scripting-engines/trace-event-perl.c. Committer testing: $ rpm -qi libtraceevent-devel Name : libtraceevent-devel Version : 1.5.3 Release : 2.fc36 Architecture: x86_64 Install Date: Mon 25 Jul 2022 03:20:19 PM -03 Group : Unspecified Size : 27728 License : LGPLv2+ and GPLv2+ Signature : RSA/SHA256, Fri 15 Apr 2022 02:11:58 PM -03, Key ID 999f7cbf38ab71f4 Source RPM : libtraceevent-1.5.3-2.fc36.src.rpm Build Date : Fri 15 Apr 2022 10:57:01 AM -03 Build Host : buildvm-x86-05.iad2.fedoraproject.org Packager : Fedora Project Vendor : Fedora Project URL : https://git.kernel.org/pub/scm/libs/libtrace/libtraceevent.git/ Bug URL : https://bugz.fedoraproject.org/libtraceevent Summary : Development headers of libtraceevent Description : Development headers of libtraceevent-libs $ Default build: $ ldd ~/bin/perf | grep tracee libtraceevent.so.1 => /lib64/libtraceevent.so.1 (0x00007f1dcaf8f000) $ # perf trace -e sched:* --max-events 10 0.000 migration/0/17 sched:sched_migrate_task(comm: "", pid: 1603763 (perf), prio: 120, dest_cpu: 1) 0.005 migration/0/17 sched:sched_wake_idle_without_ipi(cpu: 1) 0.011 migration/0/17 sched:sched_switch(prev_comm: "", prev_pid: 17 (migration/0), prev_state: 1, next_comm: "", next_prio: 120) 1.173 :0/0 sched:sched_wakeup(comm: "", pid: 3138 (gnome-terminal-), prio: 120) 1.180 :0/0 sched:sched_switch(prev_comm: "", prev_prio: 120, next_comm: "", next_pid: 3138 (gnome-terminal-), next_prio: 120) 0.156 migration/1/21 sched:sched_migrate_task(comm: "", pid: 1603763 (perf), prio: 120, orig_cpu: 1, dest_cpu: 2) 0.160 migration/1/21 sched:sched_wake_idle_without_ipi(cpu: 2) 0.166 migration/1/21 sched:sched_switch(prev_comm: "", prev_pid: 21 (migration/1), prev_state: 1, next_comm: "", next_prio: 120) 1.183 :0/0 sched:sched_wakeup(comm: "", pid: 1602985 (kworker/u16:0-f), prio: 120, target_cpu: 1) 1.186 :0/0 sched:sched_switch(prev_comm: "", prev_prio: 120, next_comm: "", next_pid: 1602985 (kworker/u16:0-f), next_prio: 120) # Had to tweak tools/perf/util/setup.py to make sure the python binding shared object links with libtraceevent if -DHAVE_LIBTRACEEVENT is present in CFLAGS. Building with NO_LIBTRACEEVENT=1 uncovered some more build failures: - Make building of data-convert-bt.c to CONFIG_LIBTRACEEVENT=y - perf-$(CONFIG_LIBTRACEEVENT) += scripts/ - bpf_kwork.o needs also to be dependent on CONFIG_LIBTRACEEVENT=y - The python binding needed some fixups and util/trace-event.c can't be built and linked with the python binding shared object, so remove it in tools/perf/util/setup.py and exclude it from the list of dependencies in the python/perf.so Makefile.perf target. Building without libtraceevent-devel installed uncovered more build failures: - The python binding tools/perf/util/python.c was assuming that traceevent/parse-events.h was always available, which was the case when we defaulted to using the in-kernel tools/lib/traceevent/ files, now we need to enclose it under ifdef HAVE_LIBTRACEEVENT, just like the other parts of it that deal with tracepoints. - We have to ifdef the rules in the Build files with CONFIG_LIBTRACEEVENT=y to build builtin-trace.c and tools/perf/trace/beauty/ as we only ifdef setting CONFIG_TRACE=y when setting NO_LIBTRACEEVENT=1 in the make command line, not when we don't detect libtraceevent-devel installed in the system. Simplification here to avoid these two ways of disabling builtin-trace.c and not having CONFIG_TRACE=y when libtraceevent-devel isn't installed is the clean way. From Athira: <quote> tools/perf/arch/powerpc/util/Build -perf-y += kvm-stat.o +perf-$(CONFIG_LIBTRACEEVENT) += kvm-stat.o </quote> Then, ditto for arm64 and s390, detected by container cross build tests. - s/390 uses test__checkevent_tracepoint() that is now only available if HAVE_LIBTRACEEVENT is defined, enclose the callsite with ifder HAVE_LIBTRACEEVENT. Also from Athira: <quote> With this change, I could successfully compile in these environment: - Without libtraceevent-devel installed - With libtraceevent-devel installed - With “make NO_LIBTRACEEVENT=1” </quote> Then, finally rename CONFIG_TRACEEVENT to CONFIG_LIBTRACEEVENT for consistency with other libraries detected in tools/perf/. Signed-off-by:
-
Ian Rogers authored
Currently the 'MetricExpr' json value is passed from the json file to the pmu-events.c. This change introduces an expression tree that is parsed into. The parsing is done largely by using operator overloading and python's 'eval' function. Two advantages in doing this are: 1) Broken metrics fail at compile time rather than relying on `perf test` to detect. `perf test` remains relevant for checking event encoding and actual metric use. 2) The conversion to a string from the tree can minimize the metric's string size, for example, preferring 1e6 over 1000000, avoiding multiplication by 1 and removing unnecessary whitespace. On x86 this reduces the string size by 2,930bytes (0.07%). In future changes it would be possible to programmatically generate the json expressions (a single line of text and so a pain to write manually) for an architecture using the expression tree. This could avoid copy-pasting metrics for all architecture variants. v4. Doesn't simplify "0*SLOTS" to 0, as the pattern is used to fix Intel metrics with topdown events. v3. Avoids generic types on standard types like set that aren't supported until Python 3.9, fixing an issue with Python 3.6 reported-by John Garry. v3 also fixes minor pylint issues and adds a call to Simplify on the read expression tree. v2. Improvements to type information. Committer notes: Added one-line fixer from Ian, see first Link: tag below. Signed-off-by:John Garry <john.g.garry@oracle.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Cc: Sumanth Korikkar <sumanthk@linux.ibm.com> Cc: Thomas Richter <tmricht@linux.ibm.com> Link: https://lore.kernel.org/r/CAP-5=fWa=zNK_ecpWGoGggHCQx7z-oW0eGMQf19Maywg0QK=4g@mail.gmail.com Link: https://lore.kernel.org/r/20221207055908.1385448-1-irogers@google.comSigned-off-by:
-
Namhyung Kim authored
perf stat: Update event skip condition for system-wide per-thread mode and merged uncore and hybrid events In print_counter_aggrdata(), it skips some events that has no aggregate count. It's actually for system-wide per-thread mode and merged uncore and hybrid events. Let's update the condition to check them explicitly. Fixes: 91f85f98 ("perf stat: Display event stats using aggr counts") Reported-by:
-
Ian Rogers authored
If LIBTRACEEVENT_DYNAMIC is enabled then avoid the install step for the plugins. If disabled correct DESTDIR so that the plugins are installed under <lib>/traceevent/plugins. Fixes: ef019df0 ("perf build: Install libtraceevent locally when building") Reported-by:
Alexander Gordeev <agordeev@linux.ibm.com> Signed-off-by:
-
Arnaldo Carvalho de Melo authored
It is used in bpf_lock_contention.c and builtin-lock.c will be made CONFIG_LIBTRACEEVENT=y conditional, so move it to machine.c, that is always available. This makes those 4 global variables for sched and lock text start and end to move to 'struct machine' too, as conceivably we can have that info for several machine instances, say some 'perf diff' like tool. Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Ian Rogers <irogers@google.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Nick Desaulniers <ndesaulniers@google.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Cc: bpf@vger.kernel.org Link: http://lore.kernel.org/lkml/Signed-off-by:
-
Ravi Bangoria authored
Multiple events in a group can belong to one or more PMUs, however there are some limitations. One of the limitations is that perf doesn't allow creating a group of events from different hw PMUs. Write a simple test to create various combinations of hw, sw and uncore PMU events and verify group creation succeeds or fails as expected. Signed-off-by:
Kan Liang <kan.liang@linux.intel.com> Acked-by:
Madhavan Srinivasan <maddy@linux.ibm.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Ananth Narayan <ananth.narayan@amd.com> Cc: Athira Jajeev <atrajeev@linux.vnet.ibm.com> Cc: Carsten Haitzler <carsten.haitzler@arm.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Kajol Jain <kjain@linux.ibm.com> Cc: Leo Yan <leo.yan@linaro.org> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Sandipan Das <sandipan.das@amd.com> Cc: Santosh Shukla <santosh.shukla@amd.com> Cc: Thomas Richter <tmricht@linux.ibm.com> Link: https://lore.kernel.org/r/20221206043237.12159-3-ravi.bangoria@amd.comSigned-off-by:
-
Ravi Bangoria authored
The 'pmus' list variable is defined as static variable under pmu.c file. Introduce a new pmus.c file and migrate this variable to it. Also make it non static so that it can be accessed from outside. Suggested-by:
-
Ian Rogers authored
Avoid libtraceevent dependency for tep_is_bigendian or trace-event.h dependency for bigendian. Add a new host_is_bigendian to util.h, using the compiler defined __BYTE_ORDER__ when available. Committer notes: Added: #else /* !__BYTE_ORDER__ */ On that nested #ifdef block, as per Namhyung's suggestion. Signed-off-by:
-
Ian Rogers authored
Remove git reference by changing GIT_COMPAT_UTIL_H to __PERF_UTIL_H. Signed-off-by:
-
James Clark authored
In this context, 'os' is already a pointer so the extra dereference isn't required. This fixes the following test failure on aarch64: $ ./perf test "json output" -vvv 92: perf stat JSON output linter : --- start --- Checking json output: no args Test failed for input: ... Fatal error: glibc detected an invalid stdio handle ---- end ---- perf stat JSON output linter: FAILED! Fixes: e7f4da31 ("perf stat: Pass struct outstate to printout()") Signed-off-by:
-
Namhyung Kim authored
When a metric produces more than one values, it missed to print the opening bracket. Fixes: ab6baaae ("perf stat: Fix JSON output in metric-only mode") Reported-by:
Weilin Wang <weilin.wang@intel.com> Signed-off-by:
-
Ian Rogers authored
Compute the headers to be installed from their source headers and make each have its own build target to install it. Using dependencies avoids headers being reinstalled and getting a new timestamp which then causes files that depend on the header to be rebuilt. Signed-off-by:
-
Ian Rogers authored
Compute the headers to be installed from their source headers and make each have its own build target to install it. Using dependencies avoids headers being reinstalled and getting a new timestamp which then causes files that depend on the header to be rebuilt. Signed-off-by:
-
Ian Rogers authored
Compute the headers to be installed from their source headers and make each have its own build target to install it. Using dependencies avoids headers being reinstalled and getting a new timestamp which then causes files that depend on the header to be rebuilt. Signed-off-by:
-
Ian Rogers authored
Compute the headers to be installed from their source headers and make each have its own build target to install it. Using dependencies avoids headers being reinstalled and getting a new timestamp which then causes files that depend on the header to be rebuilt. Signed-off-by:
-
Athira Rajeev authored
In 'perf stat' with CSV output option, number of fields in metrics output is not matching with number of fields in other event output lines. Sample output below after applying patch to fix printing os->prefix. # ./perf stat -x, --per-socket -a -C 1 ls S0,1,82.11,msec,cpu-clock,82111626,100.00,1.000,CPUs utilized S0,1,2,,context-switches,82109314,100.00,24.358,/sec ------ ====> S0,1,,,,,,,1.71,stalled cycles per insn The above command line uses field separator as "," via "-x," option and per-socket option displays socket value as first field. But here the last line for "stalled cycles per insn" has more separators. Each csv output line is expected to have 8 field separators (for the 9 fields), where as last line has 9 "," in the result. Patch fixes this issue. The counter stats are displayed by function "perf_stat__print_shadow_stats" in code "util/stat-shadow.c". While printing the stats info for "stalled cycles per insn", function "new_line_csv" is used as new_line callback. The fields printed in each line contains: "Socket_id,aggr nr,Avg,unit,event_name,run,enable_percent,ratio,unit" The metric output prints Socket_id, aggr nr, ratio and unit. It has to skip through remaining five fields ie, Avg,unit,event_name,run,enable_percent. The csv line callback uses "os->nfields" to know the number of fields to skip to match with other lines. Currently it is set as: os.nfields = 3 + aggr_fields[config->aggr_mode] + (counter->cgrp ? 1 : 0); But in case of aggregation modes, csv_sep already gets printed along with each field (Function "aggr_printout" in util/stat-display.c). So aggr_fields can be removed from nfields. And fixed number of fields to skip has to be "4". This is to skip fields for: "avg, unit, event name, run, enable_percent" This needs 4 csv separators. Patch removes aggr_fields and uses 4 as fixed number of os->nfields to skip. After the patch: # ./perf stat -x, --per-socket -a -C 1 ls S0,1,79.08,msec,cpu-clock,79085956,100.00,1.000,CPUs utilized S0,1,7,,context-switches,79084176,100.00,88.514,/sec ------ ====> S0,1,,,,,,0.81,stalled cycles per insn Fixes: 92a61f64 ("perf stat: Implement CSV metrics output") Reported-by:
Disha Goel <disgoel@linux.vnet.ibm.com> Reviewed-by:
Kajol Jain <kjain@linux.ibm.com> Signed-off-by:
-
Anshuman Khandual authored
This adds all remaining branch filters i.e "no_cycles", "no_flags" and "hw_index". While here, also updates the documentation. Signed-off-by:
Anshuman Khandual <anshuman.khandual@arm.com> Cc: James Clark <james.clark@arm.com> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lore.kernel.org/lkml/20221205064443.533587-1-anshuman.khandual@arm.comSigned-off-by:
-
Ian Rogers authored
We need to check if we have a OS prefix, otherwise we stumble on a metric segv that I'm now seeing in Arnaldo's tree: $ gdb --args perf stat -M Backend true ... Performance counter stats for 'true': 4,712,355 TOPDOWN.SLOTS # 17.3 % tma_core_bound Program received signal SIGSEGV, Segmentation fault. __strlen_evex () at ../sysdeps/x86_64/multiarch/strlen-evex.S:77 77 ../sysdeps/x86_64/multiarch/strlen-evex.S: No such file or directory. (gdb) bt #0 __strlen_evex () at ../sysdeps/x86_64/multiarch/strlen-evex.S:77 #1 0x00007ffff74749a5 in __GI__IO_fputs (str=0x0, fp=0x7ffff75f5680 <_IO_2_1_stderr_>) #2 0x0000555555779f28 in do_new_line_std (config=0x555555e077c0 <stat_config>, os=0x7fffffffbf10) at util/stat-display.c:356 #3 0x000055555577a081 in print_metric_std (config=0x555555e077c0 <stat_config>, ctx=0x7fffffffbf10, color=0x0, fmt=0x5555558b77b5 "%8.1f", unit=0x7fffffffbb10 "% tma_memory_bound", val=13.165355724442199) at util/stat-display.c:380 #4 0x00005555557768b6 in generic_metric (config=0x555555e077c0 <stat_config>, metric_expr=0x55555593d5b7 "((CYCLE_ACTIVITY.STALLS_MEM_ANY + EXE_ACTIVITY.BOUND_ON_STORES) / (CYCLE_ACTIVITY.STALLS_TOTAL + (EXE_ACTIVITY.1_PORTS_UTIL + tma_retiring * EXE_ACTIVITY.2_PORTS_UTIL) + EXE_ACTIVITY.BOUND_ON_STORES))"..., metric_events=0x555555f334e0, metric_refs=0x555555ec81d0, name=0x555555f32e80 "TOPDOWN.SLOTS", metric_name=0x555555f26c80 "tma_memory_bound", metric_unit=0x55555593d5b1 "100%", runtime=0, map_idx=0, out=0x7fffffffbd90, st=0x555555e9e620 <rt_stat>) at util/stat-shadow.c:934 #5 0x0000555555778cac in perf_stat__print_shadow_stats (config=0x555555e077c0 <stat_config>, evsel=0x555555f289d0, avg=4712355, map_idx=0, out=0x7fffffffbd90, metric_events=0x555555e078e8 <stat_config+296>, st=0x555555e9e620 <rt_stat>) at util/stat-shadow.c:1329 #6 0x000055555577b6a0 in printout (config=0x555555e077c0 <stat_config>, os=0x7fffffffbf10, uval=4712355, run=325322, ena=325322, noise=4712355, map_idx=0) at util/stat-display.c:741 #7 0x000055555577bc74 in print_counter_aggrdata (config=0x555555e077c0 <stat_config>, counter=0x555555f289d0, s=0, os=0x7fffffffbf10) at util/stat-display.c:838 #8 0x000055555577c1d8 in print_counter (config=0x555555e077c0 <stat_config>, counter=0x555555f289d0, os=0x7fffffffbf10) at util/stat-display.c:957 #9 0x000055555577dba0 in evlist__print_counters (evlist=0x555555ec3610, config=0x555555e077c0 <stat_config>, _target=0x555555e01c80 <target>, ts=0x0, argc=1, argv=0x7fffffffe450) at util/stat-display.c:1413 #10 0x00005555555fc821 in print_counters (ts=0x0, argc=1, argv=0x7fffffffe450) at builtin-stat.c:1040 #11 0x000055555560091a in cmd_stat (argc=1, argv=0x7fffffffe450) at builtin-stat.c:2665 #12 0x00005555556b1eea in run_builtin (p=0x555555e11f70 <commands+336>, argc=4, argv=0x7fffffffe450) at perf.c:322 #13 0x00005555556b2181 in handle_internal_command (argc=4, argv=0x7fffffffe450) at perf.c:376 #14 0x00005555556b22d7 in run_argv (argcp=0x7fffffffe27c, argv=0x7fffffffe270) at perf.c:420 #15 0x00005555556b26ef in main (argc=4, argv=0x7fffffffe450) at perf.c:550 (gdb) Fixes: f123b2d8 ("perf stat: Remove prefix argument in print_metric_headers()") Signed-off-by:
-
- 05 Dec, 2022 4 commits
-
-
Namhyung Kim authored
This reverts commit c4b41b83. As Ian said, the "cpu-count" is not appropriate for uncore events, also it caused a perf test failure. Signed-off-by:
-
Hans-Peter Nilsson authored
When using "sort -nu", arm64 syscalls were lost. That is, the io_setup syscall (number 0) and all but one (typically ftruncate; 64) of the syscalls that are defined symbolically (like "#define __NR_ftruncate __NR3264_ftruncate") at the point where "sort" is applied. This creation-of-syscalls.c-scheme is, judging from comments, copy-pasted from powerpc, and worked there because at the time, its tools/arch/powerpc/include/uapi/asm/unistd.h had *literals*, like "#define __NR_ftruncate 93". With sort being numeric and the non-numeric key effectively evaluating to 0, the sort option "-u" means these "duplicates" are removed. There's no need to remove syscall lines with duplicate numbers for arm64 because there are none, so let's fix that by just losing the "-u". Having the table numerically sorted on syscall-number for the rest of the syscalls looks nice, so keep the "-n". Reviewed-by:
Leo Yan <leo.yan@linaro.org> Signed-off-by:
Hans-Peter Nilsson <hp@axis.com> Tested-by:
Arnd Bergmann <arnd@arndb.de> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: John Garry <john.garry@huawei.com> Cc: Kim Phillips <kim.phillips@arm.com> Cc: Leo Yan <leo.yan@linaro.org> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Mathieu Poirier <mathieu.poirier@linaro.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Will Deacon <will@kernel.org> Cc: linux-arm-kernel@lists.infradead.org Link: https://lore.kernel.org/r/20201228023941.E0DE2203B5@pchp3.se.axis.comSigned-off-by:
-
James Clark authored
Commit 93315e46 ("perf/core: Add speculation info to branch entries") added a new field in between type and new_type. Perf has its own copy of this struct so update it to match the kernel side. This doesn't currently cause any issues because new_type is only used by the Arm BRBE driver which isn't merged yet. Committer notes: Is this really an ABI? How are we supposed to deal with old perf.data files with new tools and vice versa? :-\ Fixes: 93315e46 ("perf/core: Add speculation info to branch entries") Reviewed-by:
-
Sean Christopherson authored
Use the dedicated non-atomic helpers for {clear,set}_bit() and their test variants, i.e. the double-underscore versions. Depsite being defined in atomic.h, and despite the kernel versions being atomic in the kernel, tools' {clear,set}_bit() helpers aren't actually atomic. Move to the double-underscore versions so that the versions that are expected to be atomic (for kernel developers) can be made atomic without affecting users that don't want atomic operations. No functional change intended. Signed-off-by:Sean Christopherson <seanjc@google.com> Acked-by:
-
- 24 Nov, 2022 2 commits
-
-
Arnaldo Carvalho de Melo authored
To pick up fixes. Signed-off-by: -
Ian Rogers authored
Missed previously, add libpfm support for 'perf list' callbacks and thereby JSON support. Committer notes: Add __maybe_unused to the args of the new print_libpfm_events() in the else HAVE_LIBPFM block. Fixes: e42b0ee61282a2f9 ("perf list: Add JSON output option") Signed-off-by:
-
