- 30 Jun, 2015 2 commits
-
-
Peter Zijlstra authored
Commit 1b7b938f ("perf/x86/intel: Fix PMI handling for Intel PT") conditionally increments active_events in x86_add_exclusive() but unconditionally decrements in x86_del_exclusive(). These extra decrements can lead to the situation where active_events is zero and thus the PMI handler is 'disabled' while we have active events on the PMU generating PMIs. This leads to a truckload of: Uhhuh. NMI received for unknown reason 21 on CPU 28. Do you have a strange power saving mode enabled? Dazed and confused, but trying to continue messages and generally messes up perf. Remove the condition on the increment, double increment balanced by a double decrement is perfectly fine. Restructure the code a little bit to make the unconditional inc a bit more natural. Signed-off-by:
Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Borislav Petkov <bp@alien8.de> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: alexander.shishkin@linux.intel.com Cc: brgerst@gmail.com Cc: dvlasenk@redhat.com Cc: luto@amacapital.net Cc: oleg@redhat.com Fixes: 1b7b938f ("perf/x86/intel: Fix PMI handling for Intel PT") Link: http://lkml.kernel.org/r/20150624144750.GJ18673@twins.programming.kicks-ass.netSigned-off-by:
Ingo Molnar <mingo@kernel.org>
-
Ingo Molnar authored
Merge tag 'perf-core-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/urgent Pull perf/core improvements and fixes from Arnaldo Carvalho de Melo: User visible changes: - Validate syscall list passed via -e argument to 'perf trace'. (Arnaldo Carvalho de Melo) - Introduce 'perf stat --per-thread'. (Jiri Olsa) - Check access permission for --kallsyms and --vmlinux. (Li Zhang) Infrastructure changes: - Move stuff out of 'perf stat' and into the lib for further use. (Jiri Olsa) Signed-off-by:
Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by:
-
- 26 Jun, 2015 26 commits
-
-
Li Zhang authored
There 2 problems when reading symbols files: * It doesn't report any errors even if when users specify symbol files which don't exist with --kallsyms or --vmlinux. The result just shows the address without symbols, which is not what is expected. So it's better to report errors and exit the program. * When using command perf report --kallsyms=/proc/kallsyms with a non-root user, symbols are resolved. Then select one symbol and annotate it, it reports the error as the following: Can't annotate __clear_user: No vmlinux file with build id xxx was found. The problem is caused by reading /proc/kcore without access permission. /proc/kcore requires CAP_SYS_RAWIO capability to access, so it needs to change access permission to allow a specific user to read /proc/kcore or use root to execute the perf command. This patch is to report errors when symbol files specified by users don't exist. And check access permission of /proc/kcore when reading it. Signed-off-by:
Li Zhang <zhlcindy@linux.vnet.ibm.com> Cc: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com> Link: http://lkml.kernel.org/r/1434704253-2632-1-git-send-email-zhlcindy@linux.vnet.ibm.comSigned-off-by:
-
Jiri Olsa authored
Currently all the -p option PID arguments tasks values get aggregated and printed as single values. Adding --per-tasks option to print values per task. $ perf stat -e cycles,instructions --per-thread -p 30190,30242 ^C Performance counter stats for process id '30190,30242': cat-30190 0 cycles yes-30242 3,842,525,421 cycles cat-30190 0 instructions yes-30242 10,370,817,010 instructions 1.143155657 seconds time elapsed Also works under interval mode: $ perf stat -e cycles,instructions --per-thread -p 30190,30242 -I 1000 # time comm-pid counts unit events 1.000073435 cat-30190 89,058 cycles 1.000073435 yes-30242 3,360,786,902 cycles (100.00%) 1.000073435 cat-30190 14,066 instructions 1.000073435 yes-30242 9,069,937,462 instructions 2.000204830 cat-30190 0 cycles 2.000204830 yes-30242 3,351,667,626 cycles 2.000204830 cat-30190 0 instructions 2.000204830 yes-30242 9,045,796,885 instructions ^C 2.771286639 cat-30190 0 cycles 2.771286639 yes-30242 2,593,884,166 cycles 2.771286639 cat-30190 0 instructions 2.771286639 yes-30242 7,001,171,191 instructions It works only with -t and -p options, otherwise following error is printed: $ perf stat -e cycles --per-thread -I 1000 ls The --per-thread option is only available when monitoring via -p -t options. -p, --pid <pid> stat events on existing process id -t, --tid <tid> stat events on existing thread id Signed-off-by:Jiri Olsa <jolsa@kernel.org> Tested-by:
-
Jiri Olsa authored
Centralize counters print code into single print_counters function. Signed-off-by:
-
Jiri Olsa authored
The init_stats function is meant to init 'struct stats'. Reported-by:
David Ahern <dsahern@gmail.com> Signed-off-by:
-
Jiri Olsa authored
It suits better, because the function also reads counter's data. Also the 'print_interval' name will be used in following generalization of counters display. Signed-off-by:
-
Jiri Olsa authored
It's no longer used, the stat command uses perf_evsel__read now. Signed-off-by:
-
Jiri Olsa authored
Moving perf_stat initialization counter process code, to make the reading path free of processing logic. Signed-off-by:
-
Jiri Olsa authored
Moving zero_per_pkg into counter process code, to make the reading path free of processing logic. Signed-off-by:
-
Jiri Olsa authored
Separating counters reading and processing so we could use the processing part in following patches. Using simple reading via perf_evsel__read function to read counters now, because part of the processing was in the read_cb callback. Signed-off-by:
-
Jiri Olsa authored
Moving read counters logic into single read_counters function, which will be called for both interval and overall processing legs. The reason is to split reading and processing (following patches) counters code, so we could read counters from other sources (like perf.data) and process them in the same way as 'perf stat' command does. Signed-off-by:
-
Jiri Olsa authored
Adding simple read function that reads/store data into given struct perf_counts_values *count object. Signed-off-by:
-
Jiri Olsa authored
Move all stat allocation logic related to stat object under single function. This way we can use it separately for stat object out of evlist object. Signed-off-by:
-
Jiri Olsa authored
Moving perf_evlist__(alloc|free|reset)_stats into stat object, so it could be used in following patches. Signed-off-by:
-
Jiri Olsa authored
Moving perf_evsel__(alloc|free)_prev_raw_counts into stat object, so it could be used in following patches. Signed-off-by:
-
Jiri Olsa authored
Moving perf_evsel__(alloc|free|reset)_stat_priv into stat object, so it could be used outside stat command in following patches. Signed-off-by:
-
Jiri Olsa authored
To fit in with the rest of the helpers (alloc and free). Signed-off-by:
-
Jiri Olsa authored
Renaming 'struct xyarray *cpu' pointer to more fitting/generic values, because now we store both cpu and thread values. Signed-off-by:
-
Jiri Olsa authored
Now that we have space for thread dimension counts, let's store it. Signed-off-by:
-
Jiri Olsa authored
Switching single dimensional array of 'struct perf_counts_values' with xyarray object, so we could store thread dimension counts. Signed-off-by:
-
Jiri Olsa authored
Introducing perf_counts function, that returns 'struct perf_counts_values' pointer for given cpu. Also moving perf_counts* structures into stat.h. Signed-off-by:
-
Jiri Olsa authored
Adding thread_map object tests for comm name values. Signed-off-by:
-
Jiri Olsa authored
Adding support to hold comm name together with pids in 'struct thread_map'. It will be useful for --per-thread option to display task pid together with task name. Adding thread_map__read_comms function that reads/set comm string for the 'struct thread_map'. Getting the task name from /proc/$pid/comm. Signed-off-by:
-
Jiri Olsa authored
We need to reset newly allocated 'struct thread_map_data' entries, because we will introduce new comm memeber, which will get set later or not at all. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
The 'trace' tool was accepting any names passed and just looking if syscalls returned via the raw_syscalls:* tracepoints were in that list, leading to it accepting perf events and then never finding any, as those are not valid syscall names, confusing users. Fix it by checking each entry in the list using audit_name_to_syscall, telling the user which entries are invalid and suggesting where to look for valid syscall names. E.g: [root@zoo ~]# trace -e open,foo,bar,close,baz Error: Invalid syscall bar, baz, foo Hint: try 'perf list syscalls:sys_enter_*' Hint: and: 'man syscalls' [root@zoo ~]# Reported-by:
Flavio Leitner <fbl@redhat.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Borislav Petkov <bp@suse.de> Cc: David Ahern <dsahern@gmail.com> Cc: Don Zickus <dzickus@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Link: http://lkml.kernel.org/n/tip-4g1i3m1z6fzsrznn2umi02wa@git.kernel.orgSigned-off-by:
-
Ingo Molnar authored
Merge tag 'perf-core-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/urgent Pull perf/core improvements and refactorings from Arnaldo Carvalho de Melo: Infrastructure changes: - Reference count the cpu_map and thread_map classes. (Jiri Olsa) - Set evsel->{cpus,threads} from the evlist, if not set, allowing the generalization of some 'perf stat' functions that previously were accessing private static evlist variable. (Jiri Olsa) - Delete an unnecessary check before the calling free_event_desc() (Markus Elfring) Signed-off-by:
-
Ingo Molnar authored
Merge tag 'perf-urgent-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/urgent Pull perf/urgent fix from Arnaldo Carvalho de Melo: - Fix to show proper convergence stats in 'perf bench numa' (Srikar Dronamraju) Signed-off-by:
-
- 25 Jun, 2015 9 commits
-
-
Jiri Olsa authored
Because we now propagate all evlist's cpu_maps and thread_map objects through all evsels, the perf_evsel__(nr_)cpus no longer need to be specific to stat object and check evlist and target objects. Signed-off-by:
-
Jiri Olsa authored
Propagate evlist's thread_map object through all the evsel objects. It'll be handy to access evsel's threads directly in following patches. The reason is there's no link from evsel to evlist which hold threads map now and evlist is not always available. Signed-off-by:
-
Jiri Olsa authored
Propagate evlist's cpu_map object through all the evsel objects, while keeping already configured evsel->cpus. It'll be handy to access evsel's cpus directly in following patches. Signed-off-by:
-
Jiri Olsa authored
Adding reference counting for thread_map object, so it could be easily shared among other objects. Using thread_map__put instead thread_map__delete and making thread_map__delete static. Signed-off-by:
-
Jiri Olsa authored
Adding refference counting for cpu_map object, so it could be easily shared among other objects. Using cpu_map__put instead cpu_map__delete and making cpu_map__delete static. Signed-off-by:
-
Markus Elfring authored
The free_event_desc() function tests whether its argument is NULL and then returns immediately. Thus the test around the call is not needed. This issue was detected by using the Coccinelle software. Signed-off-by:
Markus Elfring <elfring@users.sourceforge.net> Cc: Julia Lawall <julia.lawall@lip6.fr> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: kernel-janitors@vger.kernel.org Link: http://lkml.kernel.org/r/558C2ABA.3000603@users.sourceforge.netSigned-off-by:
-
Arnaldo Carvalho de Melo authored
Signed-off-by:
-
Srikar Dronamraju authored
With commit: e1e455f4 (perf tools: Work around lack of sched_getcpu in glibc < 2.6), perf_bench numa mem with -c or -m option is not able to correctly calculate convergence. With the above commit, sched_getcpu always seems to return -1. The intention of commit e1e455f4 was to add a sched_getcpu in glibc < 2.6. Hence keep the sched_getcpu definition under an ifdef. This regression happened occurred between v4.0 and v4.1 Signed-off-by:
Srikar Dronamraju <srikar@linux.vnet.ibm.com> Acked-by:
-
Ingo Molnar authored
Merge tag 'perf-core-for-mingo' 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: User visible changes: - Move toggling event logic from 'perf top' and into hists browser, allowing freeze/unfreeze with event lists with more than one entry (Namhyung Kim) - Add missing newlines when dumping PERF_RECORD_FINISHED_ROUND and showing the Aggregated stats in 'perf report -D' (Adrian Hunter) Infrastructure changes: - Allow auxtrace data alignment (Adrian Hunter) - Allow events with dot (Andi Kleen) - Fix failure to 'perf probe' events on arm (He Kuang) - Add testing for Makefile.perf (Jiri Olsa) - Add test for make install with prefix (Jiri Olsa) - Fix single target build dependency check (Jiri Olsa) - Access thread_map entries via accessors, prep patch to hold more info per entry, for ongoing 'perf stat --per-thread' work (Jiri Olsa) - Use __weak definition from compiler.h (Sukadev Bhattiprolu) - Split perf_pmu__new_alias() (Sukadev Bhattiprolu) Signed-off-by:
-
- 23 Jun, 2015 3 commits
-
-
Adrian Hunter authored
Allow auxtrace data to be a multiple of something other than page size. That is needed for BTS where the buffer contains 24-byte records. Signed-off-by:
Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@redhat.com> Link: http://lkml.kernel.org/r/1432906425-9911-11-git-send-email-adrian.hunter@intel.comSigned-off-by:
-
Jiri Olsa authored
We need to store command names with the pid. Changing map entries to be a struct holding pid. Process name is coming in shortly. Signed-off-by:
-
Jiri Olsa authored
Instead provide a method to set the array entries, and another to access the contents. Signed-off-by:
-
