Instead of using zalloc(nr_entries * sizeof_entry) that is what calloc()
does.
In some places where linux/zalloc.h isn't needed, remove it, add when
needed and was getting it indirectly.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Perf stat is crashing on arm64 hosts with the following issue:
# make -C tools/perf DEBUG=1
# perf stat sleep 1
perf: util/evsel.c:2034: get_group_fd: Assertion `!(!leader->core.fd)' failed.
[1] 1220794 IOT instruction (core dumped) ./perf stat
The sorting function introduced by commit a745c0831c ("perf stat:
Sort default events/metrics") compares events based on their individual
properties. This can cause events from different groups to be
interleaved, resulting in group members appearing before their leaders
in the sorted evlist.
When the iterator opens events in list order, a group member may be
processed before its leader has been opened.
For example, CPU_CYCLES (idx=32) with leader STALL_SLOT_BACKEND (idx=37)
could be sorted before its leader, causing the crash when CPU_CYCLES
tries to get its group fd from the not-yet-opened leader.
Fix this by comparing events based on their leader's attributes instead
of their own attributes when the events are in different groups. This
ensures all members of a group share the same sort key as their leader,
keeping groups together and guaranteeing leaders are opened before their
members.
Fixes: a745c0831c ("perf stat: Sort default events/metrics")
Reported-by: Denis Yaroshevskiy <dyaroshev@meta.com>
Tested-by: Dmitry Ilvokhin <d@ilvokhin.com>
Tested-by: Ian Rogers <irogers@google.com>
Signed-off-by: Breno Leitao <leitao@debian.org>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
The index into the cpumap array and the number of entries within the
array can never be negative, so let's make them unsigned. This is
prompted by reports that gcc 13 with -O6 is giving a
alloc-size-larger-than errors. The change makes the cpumap changes and
then updates the declaration of index variables throughout perf and
libperf to be unsigned. The two things are hard to separate as
compiler warnings about mixing signed and unsigned types breaks the
build.
Reported-by: Chingbin Li <liqb365@163.com>
Closes: https://lore.kernel.org/lkml/20260212025127.841090-1-liqb365@163.com/
Tested-by: Chingbin Li <liqb365@163.com>
Signed-off-by: Ian Rogers <irogers@google.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Commit f5803651b4 ("perf stat: Choose the most disaggregate command
line option") changed aggregation option handling for `perf stat` but
not `perf stat report` leading to parse_cache_level being passed a
struct in the `perf stat` case but erroneously an aggr_mode enum value
for `perf stat report`. Change the `perf stat report` aggregation
handling to use the same opt_aggr_mode as `perf stat`. Also, just pass
the boolean for consistency with other boolean argument handling.
Fixes: f5803651b4 ("perf stat: Choose the most disaggregate command line option")
Signed-off-by: Ian Rogers <irogers@google.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
The evlist__for_each_cpu iterator will call sched_setaffitinity when
moving between CPUs to avoid IPIs.
If only 1 IPI is saved then this may be unprofitable as the delay to get
scheduled may be considerable.
This may be particularly true if reading an event group in `perf stat`
in interval mode.
Move the affinity handling completely into the iterator so that a single
evlist__use_affinity can determine whether CPU affinities will be used.
For `perf record` the change is minimal as the dummy event and the real
event will always make the use of affinities the thing to do.
In `perf stat`, tool events are ignored and affinities only used if >1
event on the same CPU occur.
Determining if affinities are useful is done by evlist__use_affinity
which tests per-event whether the event's PMU benefits from affinity use
- it is assumed only perf event using PMUs do.
Fix a bug where when there are no affinities that the CPU map iterator
may reference a CPU not present in the initial evsel. Fix by making the
iterator and non-iterator code common.
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Andres Freund <andres@anarazel.de>
Cc: Dapeng Mi <dapeng1.mi@linux.intel.com>
Cc: Dr. David Alan Gilbert <linux@treblig.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: James Clark <james.clark@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Falcon <thomas.falcon@intel.com>
Cc: Thomas Richter <tmricht@linux.ibm.com>
Cc: Yang Li <yang.lee@linux.alibaba.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
When interrupting perf stat in repeat mode with a signal the signal is
passed to the child process but the repeat doesn't terminate:
```
$ perf stat -v --null --repeat 10 sleep 1
Control descriptor is not initialized
[ perf stat: executing run #1 ... ]
[ perf stat: executing run #2 ... ]
^Csleep: Interrupt
[ perf stat: executing run #3 ... ]
[ perf stat: executing run #4 ... ]
[ perf stat: executing run #5 ... ]
[ perf stat: executing run #6 ... ]
[ perf stat: executing run #7 ... ]
[ perf stat: executing run #8 ... ]
[ perf stat: executing run #9 ... ]
[ perf stat: executing run #10 ... ]
Performance counter stats for 'sleep 1' (10 runs):
0.9500 +- 0.0512 seconds time elapsed ( +- 5.39% )
0.01user 0.02system 0:09.53elapsed 0%CPU (0avgtext+0avgdata 18940maxresident)k
29944inputs+0outputs (0major+2629minor)pagefaults 0swaps
```
Terminate the repeated run and give a reasonable exit value:
```
$ perf stat -v --null --repeat 10 sleep 1
Control descriptor is not initialized
[ perf stat: executing run #1 ... ]
[ perf stat: executing run #2 ... ]
[ perf stat: executing run #3 ... ]
^Csleep: Interrupt
Performance counter stats for 'sleep 1' (10 runs):
0.680 +- 0.321 seconds time elapsed ( +- 47.16% )
Command exited with non-zero status 130
0.00user 0.01system 0:02.05elapsed 0%CPU (0avgtext+0avgdata 70688maxresident)k
0inputs+0outputs (0major+5002minor)pagefaults 0swaps
```
Note, this also changes the exit value for non-repeat runs when
interrupted by a signal.
Reported-by: Ingo Molnar <mingo@kernel.org>
Closes: https://lore.kernel.org/lkml/aS5wjmbAM9ka3M2g@gmail.com/
Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Thomas Richter <tmricht@linux.ibm.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Events may fail to open as no supported CPUs were specified on the
command line. In this case a confusing "error" message of "success"
can be reported. Let's skip the error in that case.
Before:
```
$ perf stat -C2048 -e cycles -- true
WARNING: A requested CPU in '2048' is not supported by PMU 'cpu' (CPUs 0-7) for event 'cycles'
Error:
No supported events found.
The sys_perf_event_open() syscall returned with 0 (Success) for event (cpu/unknown-hardware/).
"dmesg | grep -i perf" may provide additional information.
```
After:
```
$ perf stat -C2048 -e cycles -- true
WARNING: A requested CPU in '2048' is not supported by PMU 'cpu' (CPUs 0-7) for event 'cycles'
Error:
No supported events found.
```
Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: Thomas Richter <tmricht@linux.ibm.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
When reading a metric like memory bandwidth on multiple sockets, the
additional sockets will be on CPUS > 0. Because of the affinity
reading, the counters are read on CPU 0 along with the time, then the
later sockets are read. This can lead to the later sockets having a
bandwidth larger than is possible for the period of time. To avoid
this move the reading of tool events to occur after all other events
are read.
Signed-off-by: Ian Rogers <irogers@google.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
walltime_nsecs_stats is no longer used for counter values, move into
that stat_config where it controls certain things like noise
measurement.
Signed-off-by: Ian Rogers <irogers@google.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
The ru_stats are used to capture user and system time stats when a
process exits. These are then applied to user and system time tool
events if their reads fail due to the process terminating. Reduce the
scope now the metric code no longer reads these values.
Signed-off-by: Ian Rogers <irogers@google.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
When running in interval mode every third count of a time event isn't
showing properly:
```
$ perf stat -e duration_time -a -I 1000
1.001082862 1,002,290,425 duration_time
2.004264262 1,003,183,516 duration_time
3.007381401 <not counted> duration_time
4.011160141 1,003,705,631 duration_time
5.014515385 1,003,290,110 duration_time
6.018539680 <not counted> duration_time
7.022065321 1,003,591,720 duration_time
```
The regression came in with a different fix, found through bisection,
commit 68cb156743 ("perf tool_pmu: Fix aggregation on
duration_time"). The issue is caused by the enabled and running time
of the event matching the old_count's and creating a delta of 0, which
is indicative of an error.
Fixes: 68cb156743 ("perf tool_pmu: Fix aggregation on duration_time")
Signed-off-by: Ian Rogers <irogers@google.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Some Default group metrics require their events showing for
consistency with perf's previous behavior. Add a flag to indicate when
this is the case and use it in stat-display.
As events are coming from Default metrics remove that default hardware
and software events from perf stat.
Following this change the default perf stat output on an alderlake looks like:
```
$ perf stat -a -- sleep 1
Performance counter stats for 'system wide':
20,550 context-switches # nan cs/sec cs_per_second
TopdownL1 (cpu_core) # 9.0 % tma_bad_speculation
# 28.1 % tma_frontend_bound
TopdownL1 (cpu_core) # 29.2 % tma_backend_bound
# 33.7 % tma_retiring
6,685 page-faults # nan faults/sec page_faults_per_second
790,091,064 cpu_atom/cpu-cycles/
# nan GHz cycles_frequency (49.83%)
2,563,918,366 cpu_core/cpu-cycles/
# nan GHz cycles_frequency
# 12.3 % tma_bad_speculation
# 14.5 % tma_retiring (50.20%)
# 33.8 % tma_frontend_bound (50.24%)
76,390,322 cpu_atom/branches/ # nan M/sec branch_frequency (60.20%)
1,015,173,047 cpu_core/branches/ # nan M/sec branch_frequency
1,325 cpu-migrations # nan migrations/sec migrations_per_second
# 39.3 % tma_backend_bound (60.17%)
0.00 msec cpu-clock # 0.000 CPUs utilized
# 0.0 CPUs CPUs_utilized
554,347,072 cpu_atom/instructions/ # 0.64 insn per cycle
# 0.6 instructions insn_per_cycle (60.14%)
5,228,931,991 cpu_core/instructions/ # 2.04 insn per cycle
# 2.0 instructions insn_per_cycle
4,308,874 cpu_atom/branch-misses/ # 5.65% of all branches
# 5.6 % branch_miss_rate (49.76%)
9,890,606 cpu_core/branch-misses/ # 0.97% of all branches
# 1.0 % branch_miss_rate
1.005477803 seconds time elapsed
```
Signed-off-by: Ian Rogers <irogers@google.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Getting context for what a tool is doing, such as the perf_inject
instance, using container_of the tool is a common pattern in the
code. This isn't possible event_op2, event_op3 and event_op4 callbacks
as the tool isn't passed. Add the argument and then fix function
signatures to match. As tools maybe reading a tool from somewhere
else, change that code to use the passed in tool.
Signed-off-by: Ian Rogers <irogers@google.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Commit b8308511f6 bumped the max events to 1024 but this results in
BPF verifier issues if the number of command line events is too
large. Workaround this by:
1) moving the constants to a header file to share between BPF and perf
C code,
2) testing that the maximum number of events doesn't cause BPF
verifier issues in debug builds,
3) lower the max events from 1024 to 128,
4) in perf stat, if there are more events than the BPF counters can
support then disable BPF counter usage.
The rodata setup is factored into its own function to avoid
duplicating it in the testing code.
Signed-off-by: Ian Rogers <irogers@google.com>
Fixes: b8308511f6 ("perf stat bperf cgroup: Increase MAX_EVENTS from 32 to 1024")
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Without a PMU perf matches an event against any PMU with the
event. Unfortunately some PMU drivers advertise a "cycles" event which
is typically just a core event. To make perf's behavior consistent,
just look up default events with their designated PMU types.
Signed-off-by: Ian Rogers <irogers@google.com>
Tested-by: James Clark <james.clark@linaro.org>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
The bperf BPF counter code doesn't handle "any"(-1) CPU events, always
wanting to aggregate a count against a CPU, which avoids the need for
atomics so let's not change that. Force evsels used for BPF counters
to require a CPU when not in system-wide mode so that the "any"(-1)
value isn't used during map propagation and evsel's CPU map matches
that of the PMU.
Fixes: b91917c0c6 ("perf bpf_counter: Fix handling of cpumap fixing hybrid")
Signed-off-by: Ian Rogers <irogers@google.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
If an event shows as "<not supported>" in perf stat output, in verbose
mode add the strerror output to help diagnose the issue.
Consider:
```
$ perf stat -e cycles,data_read,instructions true
Performance counter stats for 'true':
357,457 cycles:u
<not supported> MiB data_read:u
156,182 instructions:u # 0.44 insn per cycle
0.001250315 seconds time elapsed
0.001283000 seconds user
0.000000000 seconds sys
```
To understand why the data_read uncore event failed you can run it
again with -v option. This change adds detailed message about the
error and suggestion how to fix it potentially.
Warning:
data_read:u event is not supported by the kernel.
Invalid event (data_read:u) in per-thread mode, enable system wide with '-a'.
Signed-off-by: Ian Rogers <irogers@google.com>
[ simplified the commit message ]
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
For the sake of Intel topdown events commit 9eac5612da ("perf
stat: Don't skip failing group events") changed 'perf stat' error
handling making it so that more errors were fatal and didn't report
"<not supported>" events. The change outside of topdown events was
unintentional.
The notion of "fatal" error handling was introduced in commit
e0e6a6ca3a ("perf stat: Factor out open error handling") and
refined in commits like commit cb5ef60067 ("perf stat: Error out
unsupported group leader immediately") to be an approach for avoiding
later assertion failures in the code base.
This change fixes those issues and removes the notion of a fatal error
on an event. If all events fail to open then a fatal error occurs with
the previous fatal error message. This seems to best match the notion of
supported events and allowing some errors not to stop 'perf stat', while
allowing the truly fatal no event case to terminate the tool early.
The evsel->errored flag is only used in the stat code but always just
meaning !evsel->supported although there is a comment about it being
sticky. Force all evsels to be supported in evsel__init and then clear
this when evsel__open fails. When an event is tried the supported is
set to true again. This simplifies the notion of whether an evsel is
broken.
In the get_group_fd code, fail to get a group fd when the evsel isn't
supported. If the leader isn't supported then it is also expected that
there is no group_fd as the leader will have been skipped. Therefore
change the BUG_ON test to be on supported rather than skippable. This
corrects the assertion errors that were the reason for the previous
fatal error handling.
Fixes: 9eac5612da ("perf stat: Don't skip failing group events")
Reviewed-by: James Clark <james.clark@linaro.org>
Signed-off-by: Ian Rogers <irogers@google.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Chun-Tse Shao <ctshao@google.com>
Cc: Dapeng Mi <dapeng1.mi@linux.intel.com>
Cc: Howard Chu <howardchu95@gmail.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Falcon <thomas.falcon@intel.com>
Link: https://lore.kernel.org/r/20251002220727.1889799-2-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
