kernel/tools/perf/util/session.c, branch linux-6.16.y

perf record: Add 8-byte aligned event type PERF_RECORD_COMPRESSED2

2025-05-16T20:31:40Z

The original PERF_RECORD_COMPRESS is not 8-byte aligned, which can cause asan runtime error: # Build with asan $ make -C tools/perf O=/tmp/perf DEBUG=1 EXTRA_CFLAGS="-O0 -g -fno-omit-frame-pointer -fsanitize=undefined" # Test success with many asan runtime errors: $ /tmp/perf/perf test "Zstd perf.data compression/decompression" -vv 83: Zstd perf.data compression/decompression: ... util/session.c:1959:13: runtime error: member access within misaligned address 0x7f69e3f99653 for type 'union perf_event', which requires 13 byte alignment 0x7f69e3f99653: note: pointer points here d0 3a 50 69 44 00 00 00 00 00 08 00 bb 07 00 00 00 00 00 00 44 00 00 00 00 00 00 00 ff 07 00 00 ^ util/session.c:2163:22: runtime error: member access within misaligned address 0x7f69e3f99653 for type 'union perf_event', which requires 8 byte alignment 0x7f69e3f99653: note: pointer points here d0 3a 50 69 44 00 00 00 00 00 08 00 bb 07 00 00 00 00 00 00 44 00 00 00 00 00 00 00 ff 07 00 00 ^ ... Since there is no way to align compressed data in zstd compression, this patch add a new event type `PERF_RECORD_COMPRESSED2`, which adds a field `data_size` to specify the actual compressed data size. The `header.size` contains the total record size, including the padding at the end to make it 8-byte aligned. Tested with `Zstd perf.data compression/decompression` Signed-off-by: Chun-Tse Shao Tested-by: Arnaldo Carvalho de Melo Cc: Adrian Hunter Cc: Alexander Shishkin Cc: Ben Gainey Cc: Christophe Leroy Cc: Ian Rogers Cc: Ingo Molnar Cc: James Clark Cc: Jiri Olsa Cc: Kan Liang Cc: Leo Yan Cc: Mark Rutland Cc: Namhyung Kim Cc: Nick Terrell Cc: Peter Zijlstra Link: https://lore.kernel.org/r/20250303183646.327510-1-ctshao@google.com Signed-off-by: Arnaldo Carvalho de Melo

perf session: Skip unsupported new event types

2025-04-25T15:32:01Z

`perf report` currently halts with an error when encountering unsupported new event types (`event.type >= PERF_RECORD_HEADER_MAX`). This patch modifies the behavior to skip these samples and continue processing the remaining events. Additionally, stops reporting if the new event size is not 8-byte aligned. Suggested-by: Arnaldo Carvalho de Melo Suggested-by: Namhyung Kim Reviewed-by: Ian Rogers Signed-off-by: Chun-Tse Shao Cc: Adrian Hunter Cc: Alexander Shishkin Cc: Ben Gainey Cc: Dmitriy Vyukov Cc: Ingo Molnar Cc: Jiri Olsa Cc: Kan Liang Cc: Mark Rutland Cc: Peter Zijlstra Link: https://lore.kernel.org/r/20250414173921.2905822-1-ctshao@google.com Signed-off-by: Arnaldo Carvalho de Melo

perf report: Add parallelism sort key

2025-02-18T06:00:50Z

Show parallelism level in profiles if requested by user. Signed-off-by: Dmitry Vyukov Reviewed-by: Andi Kleen Link: https://lore.kernel.org/r/7f7bb87cbaa51bf1fb008a0d68b687423ce4bad4.1739437531.git.dvyukov@google.com Signed-off-by: Namhyung Kim

perf sample: Make user_regs and intr_regs optional

2025-02-13T04:06:11Z

The struct dump_regs contains 512 bytes of cache_regs, meaning the two values in perf_sample contribute 1088 bytes of its total 1384 bytes size. Initializing this much memory has a cost reported by Tavian Barnes as about 2.5% when running `perf script --itrace=i0`: https://lore.kernel.org/lkml/d841b97b3ad2ca8bcab07e4293375fb7c32dfce7.1736618095.git.tavianator@tavianator.com/ Adrian Hunter replied that the zero initialization was necessary and couldn't simply be removed. This patch aims to strike a middle ground of still zeroing the perf_sample, but removing 79% of its size by make user_regs and intr_regs optional pointers to zalloc-ed memory. To support the allocation accessors are created for user_regs and intr_regs. To support correct cleanup perf_sample__init and perf_sample__exit functions are created and added throughout the code base. Signed-off-by: Ian Rogers Link: https://lore.kernel.org/r/20250113194345.1537821-1-irogers@google.com Signed-off-by: Namhyung Kim

perf cpumap: Reduce transitive dependencies on libperf MAX_NR_CPUS

2024-12-09T20:52:41Z

libperf exposes MAX_NR_CPUS via tools/lib/perf/include/internal/cpumap.h which is internal. The preferred dependency should be the definition in tools/perf/perf.h. Add the includes of perf.h so that MAX_NR_CPUS can be hidden in libperf. Reviewed-by: Leo Yan Signed-off-by: Ian Rogers Cc: Adrian Hunter Cc: Alexander Shishkin Cc: Ben Gainey Cc: Ingo Molnar Cc: James Clark Cc: Jiri Olsa Cc: Kan Liang Cc: Kyle Meyer Cc: Mark Rutland Cc: Namhyung Kim Cc: Peter Zijlstra Link: https://lore.kernel.org/r/20241206044035.1062032-3-irogers@google.com Signed-off-by: Arnaldo Carvalho de Melo

tools/perf: Correctly calculate sample period for inherited SAMPLE_READ values

2024-10-02T21:58:03Z

Sample period calculation in deliver_sample_value is updated to calculate the per-thread period delta for events that are inherit + PERF_SAMPLE_READ. When the sampling event has this configuration, the read_format.id is used with the tid from the sample to lookup the storage of the previously accumulated counter total before calculating the delta. All existing valid configurations where read_format.value represents some global value continue to use just the read_format.id to locate the storage of the previously accumulated total. perf_sample_id is modified to support tracking per-thread values, along with the existing global per-id values. In the per-thread case, values are stored in a hash by tid within the perf_sample_id, and are dynamically allocated as the number is not known ahead of time. Signed-off-by: Ben Gainey Cc: james.clark@arm.com Link: https://lore.kernel.org/r/20241001121505.1009685-2-ben.gainey@arm.com Signed-off-by: Namhyung Kim

perf env: Find correct branch counter info on hybrid

2024-09-11T16:08:46Z

No event is printed in the "Branch Counter" column on hybrid machines. For example, $ perf record -e "{cpu_core/branch-instructions/pp,cpu_core/branches/}:S" -j any,counter $ perf report --total-cycles # Branch counter abbr list: # cpu_core/branch-instructions/pp = A # cpu_core/branches/ = B # '-' No event occurs # '+' Event occurrences may be lost due to branch counter saturated # # Sampled Cycles% Sampled Cycles Avg Cycles% Avg Cycles Branch Counter # ............... .............. ........... .......... .............. 44.54% 727.1K 0.00% 1 |+ |+ | 36.31% 592.7K 0.00% 2 |+ |+ | 17.83% 291.1K 0.00% 1 |+ |+ | The branch counter information (br_cntr_width and br_cntr_nr) in the perf_env is retrieved from the CPU_PMU_CAPS. However, the CPU_PMU_CAPS is not available on hybrid machines. Without the width information, the number of occurrences of an event cannot be calculated. For a hybrid machine, the caps information should be retrieved from the PMU_CAPS, and stored in the perf_env->pmu_caps. Add a perf_env__find_br_cntr_info() to return the correct branch counter information from the corresponding fields. Committer notes: While testing I couldn't s ee those "Branch counter" columns enabled by pressing 'B' on the TUI, after reporting it to the list Kan explained the situation: For a hybrid client, the "Branch Counter" feature is only supported starting from the just released Lunar Lake. Perf falls back to only "ANY" on your Raptor Lake. The "The branch counter is not available" message is expected. Here is the 'perf evlist' result from my Lunar Lake machine, # perf evlist -v cpu_core/branch-instructions/pp: type: 4 (cpu_core), size: 136, config: 0xc4 (branch-instructions), { sample_period, sample_freq }: 4000, sample_type: IP|TID|TIME|READ|PERIOD|BRANCH_STACK|IDENTIFIER, read_format: ID|GROUP|LOST, disabled: 1, freq: 1, enable_on_exec: 1, precise_ip: 2, sample_id_all: 1, exclude_guest: 1, branch_sample_type: ANY|COUNTERS # Fixes: 6f9d8d1de2c61288 ("perf script: Add branch counters") Reviewed-by: Ian Rogers Signed-off-by: Kan Liang Cc: Adrian Hunter Cc: Jiri Olsa Cc: Kan Liang Cc: Namhyung Kim Link: https://lore.kernel.org/r/20240909184201.553519-1-kan.liang@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo

perf header: Remove repipe option

2024-08-30T12:24:27Z

No longer used by `perf inject` the repipe_fd is always -1 and repipe is always false. Remove the options and associated code knowing the constant values of the removed variables. Signed-off-by: Ian Rogers Cc: Adrian Hunter Cc: Alexander Shishkin Cc: Ingo Molnar Cc: James Clark Cc: Jiri Olsa Cc: Kan Liang Cc: Mark Rutland Cc: Namhyung Kim Cc: Nick Terrell Cc: Peter Zijlstra Cc: Yanteng Si Cc: Yicong Yang Link: https://lore.kernel.org/r/20240829150154.37929-8-irogers@google.com Signed-off-by: Arnaldo Carvalho de Melo

perf inject: Overhaul handling of pipe files

2024-08-30T12:23:51Z

Previously inject->is_pipe was set if the input or output were a pipe. Determining the input was a pipe had to be done prior to starting the session and opening the file. This was done by comparing the input file name with '-' but it fails if the pipe file is written to disk. Opening a pipe file from disk will correctly set perf_data.is_pipe, but this is too late for 'perf inject' and results in a broken file. A workaround is 'cat pipe_perf|perf inject -i - ...'. This change removes inject->is_pipe and changes the dependent conditions to use the is_pipe flag on the input (inject->session->data) and output files (inject->output). This ensures the is_pipe condition reflects things like the header being read. The change removes the use of perf file header repiping, that is writing the file header out while reading it in. The case of input pipe and output file cannot repipe as the attributes for the file are unknown. To resolve this, write the file header when writing to disk and as the attributes may be unknown, write them after the data. Update sessions repipe variable to be trace_event_repipe as those are the only events now impacted by it. Update __perf_session__new as the repipe_fd no longer needs passing. Fully removing repipe from session header reading will be done in a later change. Committer testing: root@number:~# perf record -e syscalls:sys_enter_*sleep/max-stack=4/ -o - sleep 0.01 | perf report -i - # To display the perf.data header info, please use --header/--header-only options. # [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.050 MB - ] # # Total Lost Samples: 0 # # Samples: 1 of event 'syscalls:sys_enter_clock_nanosleep' # Event count (approx.): 1 # # Overhead Command Shared Object Symbol # ........ ....... ............. ............................... # 100.00% sleep libc.so.6 [.] clock_nanosleep@GLIBC_2.2.5 | ---__libc_start_main@@GLIBC_2.34 __libc_start_call_main 0x562fc2560a9f clock_nanosleep@GLIBC_2.2.5 # # (Tip: Create an archive with symtabs to analyse on other machine: perf archive) # root@number:~# perf record -e syscalls:sys_enter_*sleep/max-stack=4/ -o - sleep 0.01 > pipe.data [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.050 MB - ] root@number:~# perf report --stdio -i pipe.data # To display the perf.data header info, please use --header/--header-only options. # # # Total Lost Samples: 0 # # Samples: 1 of event 'syscalls:sys_enter_clock_nanosleep' # Event count (approx.): 1 # # Overhead Command Shared Object Symbol # ........ ....... ............. ............................... # 100.00% sleep libc.so.6 [.] clock_nanosleep@GLIBC_2.2.5 | ---__libc_start_main@@GLIBC_2.34 __libc_start_call_main 0x55f775975a9f clock_nanosleep@GLIBC_2.2.5 # # (Tip: To set sampling period of individual events use perf record -e cpu/cpu-cycles,period=100001/,cpu/branches,period=10001/ ...) # root@number:~# Signed-off-by: Ian Rogers Tested-by: Arnaldo Carvalho de Melo Cc: Adrian Hunter Cc: Alexander Shishkin Cc: Ingo Molnar Cc: James Clark Cc: Jiri Olsa Cc: Kan Liang Cc: Mark Rutland Cc: Namhyung Kim Cc: Nick Terrell Cc: Peter Zijlstra Cc: Yanteng Si Cc: Yicong Yang Link: https://lore.kernel.org/r/20240829150154.37929-7-irogers@google.com Signed-off-by: Arnaldo Carvalho de Melo

perf tools: Print lost samples due to BPF filter

2024-08-28T21:07:20Z

Print the actual dropped sample count in the event stat. $ sudo perf record -o- -e cycles --filter 'period < 10000' \ -e instructions --filter 'ip > 0x8000000000000000' perf test -w noploop | \ perf report --stat -i- [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.058 MB - ] Aggregated stats: TOTAL events: 469 MMAP events: 268 (57.1%) COMM events: 2 ( 0.4%) EXIT events: 1 ( 0.2%) SAMPLE events: 16 ( 3.4%) MMAP2 events: 22 ( 4.7%) LOST_SAMPLES events: 2 ( 0.4%) KSYMBOL events: 89 (19.0%) BPF_EVENT events: 39 ( 8.3%) ATTR events: 2 ( 0.4%) FINISHED_ROUND events: 1 ( 0.2%) ID_INDEX events: 1 ( 0.2%) THREAD_MAP events: 1 ( 0.2%) CPU_MAP events: 1 ( 0.2%) EVENT_UPDATE events: 2 ( 0.4%) TIME_CONV events: 1 ( 0.2%) FEATURE events: 20 ( 4.3%) FINISHED_INIT events: 1 ( 0.2%) cycles stats: SAMPLE events: 2 LOST_SAMPLES (BPF) events: 4010 instructions stats: SAMPLE events: 14 LOST_SAMPLES (BPF) events: 3990 Signed-off-by: Namhyung Kim Cc: Adrian Hunter Cc: Ian Rogers Cc: Ingo Molnar Cc: Jiri Olsa Cc: KP Singh Cc: Kan Liang Cc: Peter Zijlstra Cc: Song Liu Link: https://lore.kernel.org/r/20240820154504.128923-2-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo