kernel/tools/perf/bench, branch linux-6.16.y

perf bench futex: Fix prctl include in musl libc

2025-06-17T21:29:42Z

Namhyung Kim reported: I've updated the perf-tools-next to v6.16-rc1 and found a build error like below on alpine linux 3.18. In file included from bench/futex.c:6: /usr/include/sys/prctl.h:88:8: error: redefinition of 'struct prctl_mm_map' 88 | struct prctl_mm_map { | ^~~~~~~~~~~~ In file included from bench/futex.c:5: /linux/tools/include/uapi/linux/prctl.h:134:8: note: originally defined here 134 | struct prctl_mm_map { | ^~~~~~~~~~~~ make[4]: *** [/linux/tools/build/Makefile.build:86: /build/bench/futex.o] Error 1 git bisect says it's the first commit introduced the failure. So both /usr/include/sys/prctl.h and /linux/tools/include/uapi/linux/prctl.h provide struct prctl_mm_map but their include guard must be different. /usr/include/sys/prctl.h provided by glibc contains the prctl() declaration. It includes also linux/prctl.h. The /usr/include/sys/prctl.h on alpine linux is different. This is probably coming from musl. It contains the PR_* definition and the prctl() declaration. So it clashes here because now the one struct is available twice. The man page for prctl(2) says: | #include /* Definition of PR_* constants */ | #include so musl doesn't follow this. So don't include linux/prctl.h explicitely and add some new defines needed if they aren't available. Acked-by: Sebastian Andrzej Siewior Reported-by: Namhyung Kim Closes: https://lore.kernel.org/r/20250611092542.F4ooE2FL@linutronix.de Link: https://www.openwall.com/lists/musl/2025/06/12/11 Signed-off-by: Arnaldo Carvalho de Melo

Merge tag 'perf-tools-for-v6.16-1-2025-06-03' of git://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools

2025-06-03T22:11:44Z

Pull perf tools updates from Arnaldo Carvalho de Melo: "perf report/top/annotate TUI: - Accept the left arrow key as a Zoom out if done on the first column - Show if source code toggle status in title, to help spotting bugs with the various disassemblers (capstone, llvm, objdump) - Provide feedback on unhandled hotkeys Build: - Better inform when certain features are not available with warnings in the build process and in 'perf version --build-options' or 'perf -vv' perf record: - Improve the --off-cpu code by synthesizing events for switch-out -> switch-in intervals using a BPF program. This can be fine tuned using a --off-cpu-thresh knob perf report: - Add 'tgid' sort key perf mem/c2c: - Add 'op', 'cache', 'snoop', 'dtlb' output fields - Add support for 'ldlat' on AMD IBS (Instruction Based Sampling) perf ftrace: - Use process/session specific trace settings instead of messing with the global ftrace knobs perf trace: - Implement syscall summary in BPF - Support --summary-mode=cgroup - Always print return value for syscalls returning a pid - The rseq and set_robust_list don't return a pid, just -errno perf lock contention: - Symbolize zone->lock using BTF - Add -J/--inject-delay option to estimate impact on application performance by optimization of kernel locking behavior perf stat: - Improve hybrid support for the NMI watchdog warning Symbol resolution: - Handle 'u' and 'l' symbols in /proc/kallsyms, resolving some Rust symbols - Improve Rust demangler Hardware tracing: Intel PT: - Fix PEBS-via-PT data_src - Do not default to recording all switch events - Fix pattern matching with python3 on the SQL viewer script arm64: - Fixups for the hip08 hha PMU Vendor events: - Update Intel events/metrics files for alderlake, alderlaken, arrowlake, bonnell, broadwell, broadwellde, broadwellx, cascadelakex, clearwaterforest, elkhartlake, emeraldrapids, grandridge, graniterapids, haswell, haswellx, icelake, icelakex, ivybridge, ivytown, jaketown, lunarlake, meteorlake, nehalemep, nehalemex, rocketlake, sandybridge, sapphirerapids, sierraforest, skylake, skylakex, snowridgex, tigerlake, westmereep-dp, westmereep-sp, westmereep-sx python support: - Add support for event counts in the python binding, add a counting.py example perf list: - Display the PMU name associated with a perf metric in JSON perf test: - Hybrid improvements for metric value validation test - Fix LBR test by ignoring idle task - Add AMD IBS sw filter ana d'ldlat' tests - Add 'perf trace --summary-mode=cgroup' test - Add tests for the various language symbol demanglers Miscellaneous: - Allow specifying the cpu an event will be tied using '-e event/cpu=N/' - Sync various headers with the kernel sources - Add annotations to use clang's -Wthread-safety and fix some problems it detected - Make dump_stack() use perf's symbol resolution to provide better backtraces - Intel TPEBS support cleanups and fixes. TPEBS stands for Timed PEBS (Precision Event-Based Sampling), that adds timing info, the retirement latency of instructions - Various memory allocation (some detected by ASAN) and reference counting fixes - Add a 8-byte aligned PERF_RECORD_COMPRESSED2 to replace PERF_RECORD_COMPRESSED - Skip unsupported event types in perf.data files, don't stop when finding one - Improve lookups using hashmaps and binary searches" * tag 'perf-tools-for-v6.16-1-2025-06-03' of git://git.kernel.org/pub/scm/linux/kernel/git/perf/perf-tools: (206 commits) perf callchain: Always populate the addr_location map when adding IP perf lock contention: Reject more than 10ms delays for safety perf trace: Set errpid to false for rseq and set_robust_list perf symbol: Move demangling code out of symbol-elf.c perf trace: Always print return value for syscalls returning a pid perf script: Print PERF_AUX_FLAG_COLLISION flag perf mem: Show absolute percent in mem_stat output perf mem: Display sort order only if it's available perf mem: Describe overhead calculation in brief perf record: Fix incorrect --user-regs comments Revert "perf thread: Ensure comm_lock held for comm_list" perf test trace_summary: Skip --bpf-summary tests if no libbpf perf test intel-pt: Skip jitdump test if no libelf perf intel-tpebs: Avoid race when evlist is being deleted perf test demangle-java: Don't segv if demangling fails perf symbol: Fix use-after-free in filename__read_build_id perf pmu: Avoid segv for missing name/alias_name in wildcarding perf machine: Factor creating a "live" machine out of dwarf-unwind perf test: Add AMD IBS sw filter test perf mem: Count L2 HITM for c2c statistic ...

tools headers: Synchronize prctl.h ABI header

2025-05-21T11:57:41Z

The prctl.h ABI header was slightly updated during the development of the interface. In particular the "immutable" parameter became a bit in the option argument. Synchronize prctl.h ABI header again and make use of the definition in the testsuite and "perf bench futex". Signed-off-by: Sebastian Andrzej Siewior Signed-off-by: Peter Zijlstra (Intel) Reviewed-by: André Almeida Link: https://lore.kernel.org/r/20250517151455.1065363-5-bigeasy@linutronix.de

tools/perf: Allow to select the number of hash buckets

2025-05-03T10:02:10Z

Add the -b/ --buckets argument to specify the number of hash buckets for the private futex hash. This is directly passed to prctl(PR_FUTEX_HASH, PR_FUTEX_HASH_SET_SLOTS, buckets, immutable) and must return without an error if specified. The `immutable' is 0 by default and can be set to 1 via the -I/ --immutable argument. The size of the private hash is verified with PR_FUTEX_HASH_GET_SLOTS. If PR_FUTEX_HASH_GET_SLOTS failed then it is assumed that an older kernel was used without the support and that the global hash is used. Signed-off-by: Sebastian Andrzej Siewior Signed-off-by: Peter Zijlstra (Intel) Link: https://lore.kernel.org/r/20250416162921.513656-20-bigeasy@linutronix.de

perf bench evlist-open-close: Reduce scope of 2 variables

2025-04-25T15:32:13Z

Make 2 global variables local. Reduces ELF binary size by removing relocations. For a no flags build, the perf binary size is reduced by 4,144 bytes on x86-64. Reviewed-by: Arnaldo Carvalho de Melo Signed-off-by: Ian Rogers Tested-by: Arnaldo Carvalho de Melo Cc: Adrian Hunter Cc: Alexander Shishkin Cc: Dapeng Mi Cc: Dominique Martinet Cc: Dr. David Alan Gilbert Cc: Hao Ge Cc: Howard Chu Cc: Ingo Molnar Cc: James Clark Cc: Jiri Olsa Cc: Kan Liang Cc: Levi Yun Cc: Mark Rutland Cc: Namhyung Kim Cc: Peter Zijlstra Cc: Tengda Wu Cc: Thomas Richter Cc: Veronika Molnarova Cc: Weilin Wang Cc: Xu Yang Link: https://lore.kernel.org/r/20250410173631.1713627-3-irogers@google.com Signed-off-by: Arnaldo Carvalho de Melo

perf bench sched pipe: fix enforced blocking reads in worker_thread

2025-03-24T06:20:37Z

The function worker_thread() is programmed in a way that roughly doubles the number of expectable context switches, because it enforces blocking reads: Performance counter stats for 'perf bench sched pipe': 2,000,004 context-switches 11.859548321 seconds time elapsed 0.674871000 seconds user 8.076890000 seconds sys The result of this behavior is that the blocking reads by far dominate the performance analysis of 'perf bench sched pipe': Samples: 78K of event 'cycles:P', Event count (approx.): 27964965844 Overhead Command Shared Object Symbol 25.28% sched-pipe [kernel.kallsyms] [k] read_hpet 8.11% sched-pipe [kernel.kallsyms] [k] retbleed_untrain_ret 2.82% sched-pipe [kernel.kallsyms] [k] pipe_write From the code, it is unclear if that behavior is wanted but the log says that at least Ingo Molnar aims to mimic lmbench's lat_ctx, that doesn't handle the pipe ends that way (https://sourceforge.net/p/lmbench/code/HEAD/tree/trunk/lmbench2/src/lat_ctx.c) Fix worker_thread() by always first feeding the write ends of the pipes and then trying to read. This roughly halves the context switches and runtime of pure 'perf bench sched pipe': Performance counter stats for 'perf bench sched pipe': 1,005,770 context-switches 6.033448041 seconds time elapsed 0.423142000 seconds user 4.519829000 seconds sys And the blocking reads do no longer dominate the analysis at the above extreme: Samples: 40K of event 'cycles:P', Event count (approx.): 14309364879 Overhead Command Shared Object Symbol 12.20% sched-pipe [kernel.kallsyms] [k] read_hpet 9.23% sched-pipe [kernel.kallsyms] [k] retbleed_untrain_ret 3.68% sched-pipe [kernel.kallsyms] [k] pipe_write Signed-off-by: Dirk Gouders Acked-by: Ingo Molnar Link: https://lore.kernel.org/r/20250323140316.19027-2-dirk@gouders.net Signed-off-by: Namhyung Kim

perf bench: Fix perf bench syscall loop count

2025-03-05T17:19:23Z

Command 'perf bench syscall fork -l 100000' offers option -l to run for a specified number of iterations. However this option is not always observed. The number is silently limited to 10000 iterations as can be seen: Output before: # perf bench syscall fork -l 100000 # Running 'syscall/fork' benchmark: # Executed 10,000 fork() calls Total time: 23.388 [sec] 2338.809800 usecs/op 427 ops/sec # When explicitly specified with option -l or --loops, also observe higher number of iterations: Output after: # perf bench syscall fork -l 100000 # Running 'syscall/fork' benchmark: # Executed 100,000 fork() calls Total time: 716.982 [sec] 7169.829510 usecs/op 139 ops/sec # This patch fixes the issue for basic execve fork and getpgid. Fixes: ece7f7c0507c ("perf bench syscall: Add fork syscall benchmark") Signed-off-by: Thomas Richter Acked-by: Sumanth Korikkar Tested-by: Athira Rajeev Cc: Tiezhu Yang Link: https://lore.kernel.org/r/20250304092349.2618082-1-tmricht@linux.ibm.com Signed-off-by: Namhyung Kim

perf bench: Fix undefined behavior in cmpworker()

2025-01-18T18:14:36Z

The comparison function cmpworker() violates the C standard's requirements for qsort() comparison functions, which mandate symmetry and transitivity: Symmetry: If x < y, then y > x. Transitivity: If x < y and y < z, then x < z. In its current implementation, cmpworker() incorrectly returns 0 when w1->tid < w2->tid, which breaks both symmetry and transitivity. This violation causes undefined behavior, potentially leading to issues such as memory corruption in glibc [1]. Fix the issue by returning -1 when w1->tid < w2->tid, ensuring compliance with the C standard and preventing undefined behavior. Link: https://www.qualys.com/2024/01/30/qsort.txt [1] Fixes: 121dd9ea0116 ("perf bench: Add epoll parallel epoll_wait benchmark") Cc: stable@vger.kernel.org Signed-off-by: Kuan-Wei Chiu Reviewed-by: James Clark Link: https://lore.kernel.org/r/20250116110842.4087530-1-visitorckw@gmail.com Signed-off-by: Namhyung Kim

perf bench: Remove reference to cmd_inject

2024-12-18T19:24:33Z

Avoid `perf bench internals inject-build-id` referencing the cmd_inject sub-command that requires perf-bench to backward reference internals of builtins. Replace the reference to cmd_inject with a call to main. To avoid python.c needing to link with something providing main, drop the libperf-bench library from the python shared object. Signed-off-by: Ian Rogers Tested-by: Arnaldo Carvalho de Melo Cc: Adrian Hunter Cc: Alexander Shishkin Cc: Andi Kleen Cc: Athira Rajeev Cc: Colin Ian King Cc: Dapeng Mi Cc: Howard Chu Cc: Ilya Leoshkevich Cc: Ingo Molnar Cc: James Clark Cc: Jiri Olsa Cc: Josh Poimboeuf Cc: Kan Liang Cc: Mark Rutland Cc: Michael Petlan Cc: Namhyung Kim Cc: Peter Zijlstra Cc: Thomas Richter Cc: Veronika Molnarova Cc: Weilin Wang Link: https://lore.kernel.org/r/20241119011644.971342-17-irogers@google.com Signed-off-by: Arnaldo Carvalho de Melo

perf header: Move is_cpu_online to numa bench

2024-11-16T19:36:47Z

The helper function is only used in the NUMA benchmark as typically online CPUs are determined through perf_cpu_map__new_online_cpus(). Reduce the scope of the function for now. Reviewed-by: James Clark Signed-off-by: Ian Rogers Tested-by: Xu Yang Cc: Adrian Hunter Cc: Albert Ou Cc: Alexander Shishkin Cc: Alexandre Ghiti Cc: Athira Rajeev Cc: Ben Zong-You Xie Cc: Benjamin Gray Cc: Bibo Mao Cc: Clément Le Goffic Cc: Dima Kogan Cc: Dr. David Alan Gilbert Cc: Huacai Chen Cc: Ingo Molnar Cc: Jiri Olsa Cc: John Garry Cc: Kan Liang Cc: Leo Yan Cc: Mark Rutland Cc: Masami Hiramatsu Cc: Mike Leach Cc: Namhyung Kim Cc: Palmer Dabbelt Cc: Paul Walmsley Cc: Peter Zijlstra Cc: Ravi Bangoria Cc: Sandipan Das Cc: Will Deacon Cc: Yicong Yang Cc: linux-arm-kernel@lists.infradead.org Cc: linux-riscv@lists.infradead.org Link: https://lore.kernel.org/r/20241107162035.52206-3-irogers@google.com Signed-off-by: Arnaldo Carvalho de Melo