mirror_ubuntu-kernels/tools/perf
Namhyung Kim b5711042a1 perf lock contention: Use per-cpu array map for spinlocks
Currently lock contention timestamp is maintained in a hash map keyed by
pid.  That means it needs to get and release a map element (which is
proctected by spinlock!) on each contention begin and end pair.  This
can impact on performance if there are a lot of contention (usually from
spinlocks).

It used to go with task local storage but it had an issue on memory
allocation in some critical paths.  Although it's addressed in recent
kernels IIUC, the tool should support old kernels too.  So it cannot
simply switch to the task local storage at least for now.

As spinlocks create lots of contention and they disabled preemption
during the spinning, it can use per-cpu array to keep the timestamp to
avoid overhead in hashmap update and delete.

In contention_begin, it's easy to check the lock types since it can see
the flags.  But contention_end cannot see it.  So let's try to per-cpu
array first (unconditionally) if it has an active element (lock != 0).
Then it should be used and per-task tstamp map should not be used until
the per-cpu array element is cleared which means nested spinlock
contention (if any) was finished and it nows see (the outer) lock.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Ian Rogers <irogers@google.com>
Cc: Hao Luo <haoluo@google.com>
Cc: Song Liu <song@kernel.org>
Cc: bpf@vger.kernel.org
Link: https://lore.kernel.org/r/20231020204741.1869520-3-namhyung@kernel.org
2023-10-25 10:02:55 -07:00
..
arch tools/perf/arch/powerpc: Fix the CPU ID const char* value by adding 0x prefix 2023-10-17 12:40:51 -07:00
bench perf bench sched pipe: Add -G/--cgroups option 2023-10-25 10:02:10 -07:00
dlfilters perf dlfilter: Add al_cleanup() 2023-08-15 16:41:49 -03:00
Documentation perf bench sched pipe: Add -G/--cgroups option 2023-10-25 10:02:10 -07:00
include/perf perf dlfilter: Add al_cleanup() 2023-08-15 16:41:49 -03:00
jvmti
pmu-events perf vendor events: Update PMC used in PM_RUN_INST_CMPL event for power10 platform 2023-10-19 23:35:20 -07:00
python
scripts perf scripts python gecko: Launch the profiler UI on the default browser with the appropriate URL 2023-08-24 14:41:49 -03:00
tests perf test: Skip CoreSight tests if cs_etm// event is not available 2023-10-25 10:02:01 -07:00
trace perf tools changes for v6.6: 2023-09-09 20:06:17 -07:00
ui perf hists browser: Avoid potential NULL dereference 2023-10-12 10:01:56 -07:00
util perf lock contention: Use per-cpu array map for spinlocks 2023-10-25 10:02:55 -07:00
.gitignore
Build
builtin-annotate.c perf addr_location: Add init/exit/copy functions 2023-06-12 15:57:53 -03:00
builtin-bench.c perf tools changes for v6.6: 2023-09-09 20:06:17 -07:00
builtin-buildid-cache.c perf buildid-cache: Fix use of uninitialized value 2023-10-12 10:01:56 -07:00
builtin-buildid-list.c perf util: Move input_name to util 2023-04-10 19:21:31 -03:00
builtin-c2c.c perf callchain: Use pthread keys for tls callchain_cursor 2023-06-12 15:57:54 -03:00
builtin-config.c perf path: Make mkpath thread safe, remove 16384 bytes from .bss 2023-05-28 10:24:14 -03:00
builtin-daemon.c perf tools: Add missing else to cmd_daemon subcommand condition 2023-06-27 12:09:44 -07:00
builtin-data.c perf util: Move input_name to util 2023-04-10 19:21:31 -03:00
builtin-diff.c perf diff: Replaces some ',' as separator with the more usual ';' 2023-07-20 11:43:45 -03:00
builtin-evlist.c perf util: Move input_name to util 2023-04-10 19:21:31 -03:00
builtin-ftrace.c perf tools fixes for v6.4: 2nd batch 2023-05-31 15:31:56 -03:00
builtin-help.c perf path: Make mkpath thread safe, remove 16384 bytes from .bss 2023-05-28 10:24:14 -03:00
builtin-inject.c perf inject: Lazily allocate guest_event event_buf 2023-06-12 18:18:14 -03:00
builtin-kallsyms.c perf symbol: Remove now unused symbol_conf.sort_by_name 2023-06-23 21:48:49 -07:00
builtin-kmem.c perf callchain: Use pthread keys for tls callchain_cursor 2023-06-12 15:57:54 -03:00
builtin-kvm.c perf evsel: Introduce evsel__name_is() method to check if the evsel name is equal to a given string 2023-04-24 14:28:11 -03:00
builtin-kwork.c perf kwork: Fix spelling mistake "Captuer" -> "Capture" 2023-09-26 21:47:11 -07:00
builtin-list.c perf list: Don't print Unit for "default_core" 2023-08-31 16:24:47 -03:00
builtin-lock.c perf lock: Fix a memory leak on an error path 2023-10-12 10:01:57 -07:00
builtin-mem.c perf addr_location: Add init/exit/copy functions 2023-06-12 15:57:53 -03:00
builtin-probe.c perf probe: Dynamically allocate params memory 2023-05-28 10:24:02 -03:00
builtin-record.c perf record: Track sideband events for all CPUs when tracing selected CPUs 2023-09-12 17:31:43 -03:00
builtin-report.c perf symbol: Remove now unused symbol_conf.sort_by_name 2023-06-23 21:48:49 -07:00
builtin-sched.c perf sched: Avoid large stack allocations 2023-06-12 18:18:14 -03:00
builtin-script.c perf: script: fix missing ',' for fields option 2023-10-17 12:40:51 -07:00
builtin-stat.c perf stat: Fix aggr mode initialization 2023-09-11 10:26:35 -03:00
builtin-timechart.c perf addr_location: Add init/exit/copy functions 2023-06-12 15:57:53 -03:00
builtin-top.c perf top: Don't pass an ERR_PTR() directly to perf_session__delete() 2023-08-17 14:13:04 -03:00
builtin-trace.c perf trace: Use heuristic when deciding if a syscall tracepoint "const char *" field is really a string 2023-08-18 16:33:28 -03:00
builtin-version.c perf version: Add status of bpf skeletons 2023-09-18 16:46:15 -07:00
builtin.h perf usage: Move usage strings 2023-04-10 19:20:53 -03:00
check-headers.sh perf tools: Add get_unaligned_leNN() 2023-10-17 12:40:02 -07:00
command-list.txt
CREDITS
design.txt
Makefile
Makefile.config perf build: Default BUILD_BPF_SKEL, warn/disable for missing deps 2023-09-18 16:46:26 -07:00
Makefile.perf perf build: Add missing comment about NO_LIBTRACEEVENT=1 2023-10-12 10:01:55 -07:00
MANIFEST
perf-archive.sh
perf-completion.sh perf shell completion: Support completion of metrics/metricgroups 2023-09-11 10:26:36 -03:00
perf-iostat.sh
perf-read-vdso.c
perf-sys.h
perf.c perf parse-events: Remove BPF event support 2023-08-15 16:41:48 -03:00
perf.h perf util: Move perf_guest/host declarations 2023-04-10 19:22:05 -03:00