- 01 Sep, 2019 7 commits
-
-
Arnaldo Carvalho de Melo authored
We use refcount_t there, so we need that header or else it'll break when we remove dso.h, that is from where it is getting that definition now... Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: https://lkml.kernel.org/n/tip-5albrk0uve6x9cf6x3ebwpae@git.kernel.orgSigned-off-by:
Arnaldo Carvalho de Melo <acme@redhat.com>
-
Arnaldo Carvalho de Melo authored
No need to have it generally available in such a critical header as symbol.h. Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: https://lkml.kernel.org/n/tip-es1ufxv7bihiumytn5dm3drn@git.kernel.orgSigned-off-by:
-
Arnaldo Carvalho de Melo authored
Reducing the size of symbol.h by removing things that are better placed somewhere else. Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: https://lkml.kernel.org/n/tip-edenkmjt1oe5fks2s6umd30b@git.kernel.orgSigned-off-by:
-
Tzvetomir Stoyanov authored
To be compliant with XDG user directory layout, the user's plugin directory is changed from ~/.traceevent/plugins to ~/.local/lib/traceevent/plugins/ Suggested-by:
Patrick McLean <chutzpah@gentoo.org> Signed-off-by:
Tzvetomir Stoyanov <tstoyanov@vmware.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Patrick McLean <chutzpah@gentoo.org> Cc: linux-trace-devel@vger.kernel.org Link: https://lore.kernel.org/linux-trace-devel/20190313144206.41e75cf8@patrickm/ Link: http://lore.kernel.org/linux-trace-devel/20190801074959.22023-4-tz.stoyanov@gmail.com Link: http://lore.kernel.org/lkml/20190805204355.344622683@goodmis.orgSigned-off-by:
Steven Rostedt (VMware) <rostedt@goodmis.org> Signed-off-by:
-
Tzvetomir Stoyanov authored
The tep_register_trace_clock() API is used to instruct the traceevent library how to print the event time stamps. As event print interface if redesigned, this API is not needed any more. The new event print API is flexible and the user can specify how the time stamps are printed. Signed-off-by:
-
Tzvetomir Stoyanov authored
Libtraceevent APIs for printing various trace events information are complicated, there are complex extra parameters. To control the way event information is printed, the user should call a set of functions in a specific sequence. These APIs are reimplemented to provide a more simple interface for printing event information. Removed APIs: tep_print_event_task() tep_print_event_time() tep_print_event_data() tep_event_info() tep_is_latency_format() tep_set_latency_format() tep_data_latency_format() tep_set_print_raw() A new API for printing event information is introduced: void tep_print_event(struct tep_handle *tep, struct trace_seq *s, struct tep_record *record, const char *fmt, ...); where "fmt" is a printf-like format string, followed by the event fields to be printed. Supported fields: TEP_PRINT_PID, "%d" - event PID TEP_PRINT_CPU, "%d" - event CPU TEP_PRINT_COMM, "%s" - event command string TEP_PRINT_NAME, "%s" - event name TEP_PRINT_LATENCY, "%s" - event latency TEP_PRINT_TIME, %d - event time stamp. A divisor and precision can be specified as part of this format string: "%precision.divisord". Example: "%3.1000d" - divide the time by 1000 and print the first 3 digits before the dot. Thus, the time stamp "123456000" will be printed as "123.456" TEP_PRINT_INFO, "%s" - event information. TEP_PRINT_INFO_RAW, "%s" - event information, in raw format. Example: tep_print_event(tep, s, record, "%16s-%-5d [%03d] %s %6.1000d %s %s", TEP_PRINT_COMM, TEP_PRINT_PID, TEP_PRINT_CPU, TEP_PRINT_LATENCY, TEP_PRINT_TIME, TEP_PRINT_NAME, TEP_PRINT_INFO); Output: ls-11314 [005] d.h. 185207.366383 function __wake_up Signed-off-by:
-
Arnaldo Carvalho de Melo authored
bpf.h and build-id.h are not needed at all in event.h, remove them. And fixup the fallout of files that were getting needed stuff from this now pruned include. Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: https://lkml.kernel.org/n/tip-rdm3dgtlrndmmnlc4bafsg3b@git.kernel.orgSigned-off-by:
-
- 31 Aug, 2019 2 commits
-
-
Arnaldo Carvalho de Melo authored
And fixup the fallout of c files not building due to now missing headers. Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: https://lkml.kernel.org/n/tip-sw8k3kpla98pr3rqypbjk9hf@git.kernel.orgSigned-off-by:
-
Arnaldo Carvalho de Melo authored
All we need there is a forward declaration for 'union perf_event', so remove it from there and add missing header directives in places using things from this indirect include. Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: https://lkml.kernel.org/n/tip-7ftk0ztstqub1tirjj8o8xbl@git.kernel.orgSigned-off-by:
-
- 29 Aug, 2019 31 commits
-
-
Arnaldo Carvalho de Melo authored
And fix the fallout, adding it to places that must have it since they use its definitions. Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: https://lkml.kernel.org/n/tip-1s3jel4i26chq2g0lydoz7i3@git.kernel.orgSigned-off-by:
-
Arnaldo Carvalho de Melo authored
With the movement of lots of stuff out of perf.h to other headers we ended up not needing it in lots of places, remove it from those places. Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: https://lkml.kernel.org/n/tip-c718m0sxxwp73lp9d8vpihb4@git.kernel.orgSigned-off-by:
-
Arnaldo Carvalho de Melo authored
Its not needed there, add it to the places that need it and were getting it via those headers. Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: https://lkml.kernel.org/n/tip-5yulx1u16vyd0zmrbg1tjhju@git.kernel.orgSigned-off-by:
-
Arnaldo Carvalho de Melo authored
Seems to be a better place for this function to live, further shrinking the hodge-podge that perf.h was. Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: https://lkml.kernel.org/n/tip-0zzt1u9rpyjukdy1ccr2u5r9@git.kernel.orgSigned-off-by:
-
Arnaldo Carvalho de Melo authored
And remove unneeded include directives from perf-sys.h to prune the header dependency tree. Fixup the fallout in places where definitions were being used without the needed include directives that were being satisfied because they were in perf-sys.h. Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: https://lkml.kernel.org/n/tip-7b1zvugiwak4ibfa3j6ott7f@git.kernel.orgSigned-off-by:
-
Arnaldo Carvalho de Melo authored
To reduce perf-sys.h and eventually nuke it. Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Link: https://lkml.kernel.org/n/tip-ars2j5m3if3gypsvkbbijucq@git.kernel.orgSigned-off-by:
-
Arnaldo Carvalho de Melo authored
Remove traceevent/event-parse.h and traceevent/trace-seq.h from places where it is not needed. Should avoid rebuilding those files when these traceevent headers get changed. Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Steven Rostedt (VMware) <rostedt@goodmis.org> Cc: Tzvetomir Stoyanov <tstoyanov@vmware.com> Link: https://lkml.kernel.org/n/tip-26hn75jn9rdealn4uqtzend6@git.kernel.orgSigned-off-by:
-
Kyle Meyer authored
Display a warning when attempting to profile more than MAX_NR_CPU CPUs. This patch should not change any behavior. Signed-off-by:
Kyle Meyer <kyle.meyer@hpe.com> Reviewed-by:
Jiri Olsa <jolsa@redhat.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russ Anderson <russ.anderson@hpe.com> Link: http://lore.kernel.org/lkml/20190827214352.94272-8-meyerk@stormcage.eag.rdlabs.hpecorp.netSigned-off-by:
-
Kyle Meyer authored
The function cpu__max_cpu() returns the possible number of CPUs as defined in the sysfs and can be used as an alternative for MAX_NR_CPUS in write_cache. MAX_CACHES is replaced by cpu__max_cpu() * MAX_CACHE_LVL. Signed-off-by:
-
Kyle Meyer authored
nr_cpus, the number of CPUs online during a record session bound by MAX_NR_CPUS, can be used as a dynamic alternative for MAX_NR_CPUS in __machine__synthesize_threads and machine__set_current_tid. Signed-off-by:
-
Kyle Meyer authored
nr_cpus, the number of CPUs online during a record session bound by MAX_NR_CPUS, can be used as a dynamic alternative for MAX_NR_CPUS in perf_session__cpu_bitmap. Signed-off-by:
-
Kyle Meyer authored
The function cpu__max_cpu() returns the possible number of CPUs as defined in the sysfs and can be used as an alternative for MAX_NR_CPUS in zero_per_pkg() and check_per_pkg(). Signed-off-by:
-
Kyle Meyer authored
'nr_cpus', the number of CPUs online during a record session bound by MAX_NR_CPUS, can be used as a dynamic alternative for MAX_NR_CPUS in svg_build_topology_map(). The value of nr_cpus can be passed into str_to_bitmap(), scan_core_topology(), and svg_build_topology_map() to replace MAX_NR_CPUS as well. Signed-off-by:
-
Kyle Meyer authored
Exchange the parameters of svg_build_topology_map() with 'struct perf_env *env' and adjust the function accordingly. This patch should not change any behavior, it is merely refactoring for the following patch. Committer notes: No need to include env.h from svghelper.h, all it needs is a forward declaration for 'struct perf_env', so move the include directive to svghelper.c, where it is really needed. Signed-off-by:
-
Jiri Olsa authored
There's wrong bitmap considered when checking for cpu count of specific node. We do the needed computation for 'set' variable, but at the end we use the 'c2c_he->cpuset' weight, which shows misleading numbers. Fixes: 1e181b92 ("perf c2c report: Add 'node' sort key") Reported-by:
Joe Mario <jmario@redhat.com> Signed-off-by:
Jiri Olsa <jolsa@kernel.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Michael Petlan <mpetlan@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lore.kernel.org/lkml/20190820140219.28338-1-jolsa@kernel.orgSigned-off-by:
-
Ingo Molnar authored
Merge tag 'perf-core-for-mingo-5.4-20190829' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/core Pull perf/core improvements and fixes from Arnaldo Carvalho de Melo: perf top: Namhyung Kim: - Decay all events in the evlist, we were decaying just the first event in a group. - Fix linking of histograms in different evsels in a event group with more than two events. With the two fixes above a command line such as: # perf top -e '{cycles,instructions,cache-misses,cache-references} Should work as expected, with four columns and with all of them being decayed over time, i.e. less weight is given for older samples. perf record: Arnaldo Carvalho de Melo: - Fix collection of build-ids when using setns() to get into namespaces, which had been broken with the introduction of the extra thread to react to PERF_RECORD_BPF_EVENT, i.e. to collect extra info for BPF programs. We need to unshare(CLONE_FS) in that thread so that the main one can do the setns(CLONE_NEWNS) when collectingthe build-ids. Without that symbol resolution gets more difficult and potentially misresolves symbols. core: Igor Lubashev: - Further alignment in permission checking via capabilities to how the kernel checks what tooling tries to do. PowerPC: Naveen N. Rao: - Sync powerpc syscall.tbl, so that 'perf trace' gets the definitions for recent syscalls. libperf: Jiri Olsa: - Move the rest of the PERF_RECORD_ metadata struct definitions so that we can use 'union perf_event'. libtraceevent: Steven Rostedt (VMware): - Do not free tep->cmdlines in add_new_comm() on failure. - Remove unneeded qsort and uses memmove instead Signed-off-by:
Ingo Molnar <mingo@kernel.org>
-
Steven Rostedt (VMware) authored
While reading a trace data file that had 100,000s of tasks, the process took an extremely long time. I profiled it down to add_new_comm(), which was doing a qsort() call on an array that was pretty much already sorted (all but the last element. qsort() isn't very efficient when dealing with mostly sorted arrays, and this definitely showed its issues. When adding a new task to the task list, instead of using qsort(), do another bsearch() with a function that will find the element before where the new task will be inserted in. Then simply shift the rest of the array, and insert the task where it belongs. Fixes: f7d82350 ("tools/events: Add files to create libtraceevent.a") Signed-off-by:
-
Steven Rostedt (VMware) authored
If the re-allocation of tep->cmdlines succeeds, then the previous allocation of tep->cmdlines will be freed. If we later fail in add_new_comm(), we must not free cmdlines, and also should assign tep->cmdlines to the new allocation. Otherwise when freeing tep, the tep->cmdlines will be pointing to garbage. Fixes: a6d2a61a ("tools lib traceevent: Remove some die() calls") Signed-off-by:
-
Arnaldo Carvalho de Melo authored
When we started using a thread to catch the PERF_RECORD_BPF_EVENT meta data events to then ask the kernel for further info (BTF, etc) for BPF programs shortly after they get loaded, we forgot to use unshare(CLONE_FS) as was done in: 868a8329 ("perf top: Support lookup of symbols in other mount namespaces.") Do it so that we can enter the namespaces to read the build-ids at the end of a 'perf record' session for the DSOs that had hits. Before: Starting a 'stress-ng --cpus 8' inside a container and then, outside the container running: # perf record -a --namespaces sleep 5 # perf buildid-list | grep stress-ng # We would end up with a 'perf.data' file that had no entry in its build-id table for the /usr/bin/stress-ng binary inside the container that got tons of PERF_RECORD_SAMPLEs. After: # perf buildid-list | grep stress-ng f2ed02c68341183a124b9b0f6e2e6c493c465b29 /usr/bin/stress-ng # Then its just a matter of making sure that that binary debuginfo package gets available in a place that 'perf report' will look at build-id keyed ELF files, which, in my case, on a f30 notebook, was a matter of installing the debuginfo file for the distro used in the container, fedora 31: # rpm -ivh http://fedora.c3sl.ufpr.br/linux/development/31/Everything/x86_64/debug/tree/Packages/s/stress-ng-debuginfo-0.07.29-10.fc31.x86_64.rpm Then, because perf currently looks for those debuginfo files (richer ELF symtab) inside that namespace (look at the setns calls): openat(AT_FDCWD, "/proc/self/ns/mnt", O_RDONLY) = 137 openat(AT_FDCWD, "/proc/13169/ns/mnt", O_RDONLY) = 139 setns(139, CLONE_NEWNS) = 0 stat("/usr/bin/stress-ng", {st_mode=S_IFREG|0755, st_size=3065416, ...}) = 0 openat(AT_FDCWD, "/usr/bin/stress-ng", O_RDONLY) = 140 fcntl(140, F_GETFD) = 0 fstat(140, {st_mode=S_IFREG|0755, st_size=3065416, ...}) = 0 mmap(NULL, 3065416, PROT_READ, MAP_PRIVATE, 140, 0) = 0x7ff2fdc5b000 munmap(0x7ff2fdc5b000, 3065416) = 0 close(140) = 0 stat("stress-ng-0.07.29-10.fc31.x86_64.debug", 0x7fff45d71260) = -1 ENOENT (No such file or directory) stat("/usr/bin/stress-ng-0.07.29-10.fc31.x86_64.debug", 0x7fff45d71260) = -1 ENOENT (No such file or directory) stat("/usr/bin/.debug/stress-ng-0.07.29-10.fc31.x86_64.debug", 0x7fff45d71260) = -1 ENOENT (No such file or directory) stat("/usr/lib/debug/usr/bin/stress-ng-0.07.29-10.fc31.x86_64.debug", 0x7fff45d71260) = -1 ENOENT (No such file or directory) stat("/root/.debug/.build-id/f2/ed02c68341183a124b9b0f6e2e6c493c465b29", 0x7fff45d711e0) = -1 ENOENT (No such file or directory) To only then go back to the "host" namespace to look just in the users's ~/.debug cache: setns(137, CLONE_NEWNS) = 0 chdir("/root") = 0 close(137) = 0 close(139) = 0 stat("/root/.debug/.build-id/f2/ed02c68341183a124b9b0f6e2e6c493c465b29/elf", 0x7fff45d732e0) = -1 ENOENT (No such file or directory) It continues to fail to resolve symbols: # perf report | grep stress-ng | head -5 9.50% stress-ng-cpu stress-ng [.] 0x0000000000021ac1 8.58% stress-ng-cpu stress-ng [.] 0x0000000000021ab4 8.51% stress-ng-cpu stress-ng [.] 0x0000000000021489 7.17% stress-ng-cpu stress-ng [.] 0x00000000000219b6 3.93% stress-ng-cpu stress-ng [.] 0x0000000000021478 # To overcome that we use: # perf buildid-cache -v --add /usr/lib/debug/usr/bin/stress-ng-0.07.29-10.fc31.x86_64.debug Adding f2ed02c68341183a124b9b0f6e2e6c493c465b29 /usr/lib/debug/usr/bin/stress-ng-0.07.29-10.fc31.x86_64.debug: Ok # # ls -la /root/.debug/.build-id/f2/ed02c68341183a124b9b0f6e2e6c493c465b29/elf -rw-r--r--. 3 root root 2401184 Jul 27 07:03 /root/.debug/.build-id/f2/ed02c68341183a124b9b0f6e2e6c493c465b29/elf # file /root/.debug/.build-id/f2/ed02c68341183a124b9b0f6e2e6c493c465b29/elf /root/.debug/.build-id/f2/ed02c68341183a124b9b0f6e2e6c493c465b29/elf: ELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked, interpreter \004, BuildID[sha1]=f2ed02c68341183a124b9b0f6e2e6c493c465b29, for GNU/Linux 3.2.0, with debug_info, not stripped, too many notes (256) # Now it finally works: # perf report | grep stress-ng | head -5 23.59% stress-ng-cpu stress-ng [.] ackermann 23.33% stress-ng-cpu stress-ng [.] is_prime 17.36% stress-ng-cpu stress-ng [.] stress_cpu_sieve 6.08% stress-ng-cpu stress-ng [.] stress_cpu_correlate 3.55% stress-ng-cpu stress-ng [.] queens_try # I'll make sure that it looks for the build-id keyed files in both the "host" namespace (the namespace the user running 'perf record' was a the time of the recording) and in the container namespace, as it shouldn't matter where a content based key lookup finds the ELF file to use in resolving symbols, etc. Reported-by:
Karl Rister <krister@redhat.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Alexei Starovoitov <ast@kernel.org> Cc: Brendan Gregg <brendan.d.gregg@gmail.com> Cc: Daniel Borkmann <daniel@iogearbox.net> Cc: Krister Johansen <kjlx@templeofstupid.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Song Liu <songliubraving@fb.com> Cc: Stanislav Fomichev <sdf@google.com> Cc: Thomas-Mich Richter <tmricht@linux.vnet.ibm.com> Fixes: 657ee553 ("perf evlist: Introduce side band thread") Link: https://lkml.kernel.org/n/tip-g79k0jz41adiaeuqud742t2l@git.kernel.orgSigned-off-by:
-
Jiri Olsa authored
So it's available for libperf's users. Signed-off-by:
-
Jiri Olsa authored
Even more, to have a "perf_record_" prefix, so that they match the PERF_RECORD_ enum they map to. Signed-off-by:
-
Jiri Olsa authored
So it's available for libperf's users. Signed-off-by:
-
Jiri Olsa authored
Move the PERF_RECORD_COMPRESSED event definition to libperf's event.h. In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used events to their generic '__u*' versions. Signed-off-by:
-
Jiri Olsa authored
Move the PERF_RECORD_HEADER_FEATURE event definition to libperf's event.h. In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used events to their generic '__u*' versions. Signed-off-by:
-
Jiri Olsa authored
Move the PERF_RECORD_TIME_CONV event definition to libperf's event.h. In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used events to their generic '__u*' versions. Signed-off-by:
-
Jiri Olsa authored
Move the PERF_RECORD_STAT_ROUND event definition to libperf's event.h. In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used events to their generic '__u*' versions. Signed-off-by:
-
Jiri Olsa authored
Move the PERF_RECORD_STAT event definition to libperf's event.h. In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used events to their generic '__u*' versions. Signed-off-by:
-
Jiri Olsa authored
Move the PERF_RECORD_STAT_CONFIG event definition to libperf's event.h. In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used events to their generic '__u*' versions. Signed-off-by:
-
Jiri Olsa authored
Move the PERF_RECORD_THREAD_MAP event definition to libperf's event.h. In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used events to their generic '__u*' versions. Signed-off-by:
-
Jiri Olsa authored
Move the PERF_RECORD_SWITCH event definition to libperf's event.h. In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used events to their generic '__u*' versions. Signed-off-by:
-
Jiri Olsa authored
Move the PERF_RECORD_ITRACE_START event definition to libperf's event.h. In order to keep libperf simple, we switch 'u64/u32/u16/u8' types used events to their generic '__u*' versions. Signed-off-by:
-
