perf_events: add support for per-cpu per-cgroup monitoring (v4)
RFC PATCH 2/2 perf_events: add support for per-cpu per-cgroup monitoring (v4)
This perf tool patch adds the ability to filter monitoring based on container groups (cgroups) for both perf stat and perf record. The cgroup to monitor are passed via a new -G option followed by a list of cgroups.
The cgroup filesystem has to be mounted. The tool will find it automatically, open the right file and pass the descriptor to perf_events.
In this fourth version, we fix the case where no event is specified yet -G is used. In that case, the cgroup constraint is applied to all events.
Example:
$ perf stat -B -a -e cycles:u,cycles:u,cycles:u -G test1,,test2 -- sleep 1 Performance counter stats for 'sleep 1':
2,368,667,414 cycles test1
2,369,661,459 cycles
<not counted> cycles test21.001856890 seconds time elapsed
Signed-off-by Stephane Eranian <[email protected]>
---
tools/perf/Documentation/perf-record.txt +6 -0
@@ -116,6 +116,12 @@ Do not update the builid cache. This saves some overhead in situations where the information in the perf.data file (which includes buildids) is sufficient. +-G name:: +--cgroup name:: +monitor only in the container called "name". This option is available only in per-cpu +mode. The cgroup filesystem must be mounted. All threads belonging to container "name" +are monitored when they run on the monitored CPUs. + SEE ALSO -------- linkperf:perf-stat[1], linkperf:perf-list[1]
tools/perf/Documentation/perf-stat.txt +5 -0
@@ -53,6 +53,11 @@ comma-sperated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2 In per-thread mode, this option is ignored. The -a option is still necessary to activate system-wide monitoring. Default is to count on all CPUs. +-G name:: +--cgroup name:: +monitor only in the container called "name". This option is available only in per-cpu +mode. The cgroup filesystem must be mounted. All threads belonging to container "name" +are monitored when they run on the monitored CPUs. EXAMPLES --------
tools/perf/Makefile +2 -0
@@ -417,6 +417,7 @@ LIB_H += util/probe-finder.h LIB_H += util/probe-event.h LIB_H += util/pstack.h LIB_H += util/cpumap.h +LIB_H += util/cgroup.h LIB_OBJS += $(OUTPUT)util/abspath.o LIB_OBJS += $(OUTPUT)util/alias.o
@@ -464,6 +465,7 @@ LIB_OBJS += $(OUTPUT)util/hist.o LIB_OBJS += $(OUTPUT)util/probe-event.o LIB_OBJS += $(OUTPUT)util/util.o LIB_OBJS += $(OUTPUT)util/cpumap.o +LIB_OBJS += $(OUTPUT)util/cgroup.o BUILTIN_OBJS += $(OUTPUT)builtin-annotate.o
tools/perf/builtin-record.c +23 -1
@@ -22,6 +22,7 @@#include "util/session.h" #include "util/symbol.h" #include "util/cpumap.h" +#include "util/cgroup.h" #include <unistd.h> #include <sched.h>
@@ -229,6 +230,8 @@ static void create_counter(int counter, int cpu)char *filter = filters[counter];struct perf_event_attr *attr = attrs + counter;struct perf_header_attr *h_attr; +unsigned long flags =0 ; +int pid;int track = !counter;/* only the first counter needs these */ int thread_index;int ret;
@@ -286,6 +289,9 @@ static void create_counter(int counter, int cpu) attr->sample_type |=PERF_SAMPLE_CPU ; } +if (cgroups[counter]) + flags =PERF_FLAG_PID_CGROUP ; + attr->mmap = track; attr->comm = track; attr->inherit = !no_inherit;
@@ -296,8 +302,13 @@ static void create_counter(int counter, int cpu)for (thread_index =0 ; thread_index < thread_num; thread_index++) { try_again: +if (cgroups[counter]) + pid = cgroups_fd[counter]; +else + pid = all_tids[thread_index]; + fd[nr_cpu][counter][thread_index] =sys_perf_event_open (attr, - all_tids[thread_index], cpu, group_fd,0 ); + pid, cpu, group_fd, flags);if (fd[nr_cpu][counter][thread_index] <0 ) {int err = errno;
@@ -828,6 +839,9 @@ static const struct option options[] = { "don't sample"),OPT_BOOLEAN ('N' ,"no-buildid-cache" , &no_buildid,"do not update the buildid cache" ), +OPT_CALLBACK ('G' ,"cgroup" ,NULL ,"name" , +"monitor in cgroup name only" , + parse_cgroups), OPT_END() };
@@ -851,6 +865,9 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) write_mode =WRITE_FORCE ; } +if (nr_cgroups && !system_wide) +usage_with_options (record_usage, options); +symbol__init ();if (no_buildid)disable_buildid_cache ();
@@ -861,6 +878,9 @@ int cmd_record(int argc, const char **argv, const char *prefix __used) attrs[0 ].config =PERF_COUNT_HW_CPU_CYCLES ; } +if (open_cgroups ()) +usage_with_options (record_usage, options); +if (target_pid !=-1 ) { target_tid = target_pid; thread_num =find_all_tid (target_pid, &all_tids);
@@ -870,6 +890,7 @@ int cmd_record(int argc, const char **argv, const char *prefix __used)usage_with_options (record_usage, options); } }else { + err = -ENOMEM ; all_tids=malloc (sizeof (pid_t));if (!all_tids) goto out_symbol_exit;
@@ -921,5 +942,6 @@ out_free_fd: all_tids =NULL ; out_symbol_exit:symbol__exit (); +close_cgroups ();return err; }
tools/perf/builtin-stat.c +38 -3
107 lines not on this page · raw message