perf record: Allow multiple recording time ranges
Adrian Hunter
reviewercover perf record: Allow multiple recording time ranges
Hi
This patch set extends perf record -D/--delay option to accept time ranges for when events are enabled, for instance:
perf record -e intel_pt// -D 10-20,30-40
to record 10ms to 20ms into the trace and 30ms to 40ms. Refer patch 5 for more details.
This ran into an issue with the handling of polling file descriptors. Essentially, calling perf_evlist__add_pollfd() would not actually result in the file descriptor being polled. Patches 1, 2 and 3 deal with that.
Patch 4 deals with the issue that when disabling events, we actually do not want to disable events collecting sideband information.
Patch 5, the last patch, actually makes the desired enhancement to perf record.
Adrian Hunter (5):
perf record: Fix way of handling non-perf-event pollfds perf record: Fix done_fd wakeup event
perf record: Change evlist->ctl_fd to use fdarray_flag__non_perf_event perf evlist: Add evlist__{en/dis}able_non_dummy()
perf record: Allow multiple recording time ranges
tools/lib/api/fd/array.h | 5 +- tools/perf/Documentation/perf-record.txt | 6 +- tools/perf/builtin-record.c | 117 ++++++++++++-- tools/perf/util/evlist.c | 270 +++++++++++++++++++++++++++++-- tools/perf/util/evlist.h | 12 +- 5 files changed, 374 insertions(+), 36 deletions(-)
Regards
Adrian
Adrian Hunter
reviewerPATCH 1/5 perf record: Fix way of handling non-perf-event pollfds
perf record __cmd_record() does not poll evlist pollfds. Instead it polls thread_data[0].pollfd. That happens whether or not threads are being used.
perf record duplicates evlist mmap pollfds as needed for separate threads. The non-perf-event represented by evlist->ctl_fd has to handled separately, which is done explicitly, duplicating it into the thread_data[0] pollfds. That approach neglects any other non-perf-event file descriptors. Currently there is also done_fd which needs the same handling.
Add a new generalized approach.
Add fdarray_flag__non_perf_event to identify the file descriptors that need the special handling. For those cases, also keep a mapping of the evlist pollfd index and thread pollfd index, so that the evlist revents can be updated.
Although this patch adds the new handling, it does not take it into use. There is no functional change, but it is the precursor to a fix, so is marked as a fix.
--- tools/lib/api/fd/array.h | 5 ++- tools/perf/builtin-record.c | 80 +++++++++++++++++++++++++++++++++++++ 2 files changed, 83 insertions(+), 2 deletions(-)
tools/lib/api/fd/array.h +3 -2
@@ -31,8 +31,9 @@ struct fdarray { };enum fdarray_flags { - fdarray_flag__default =0x00000000 , - fdarray_flag__nonfilterable =0x00000001 + fdarray_flag__default =0x00000000 , + fdarray_flag__nonfilterable =0x00000001 , + fdarray_flag__non_perf_event =0x00000002 , };void fdarray__init (struct fdarray *fda,int nr_autogrow);
tools/perf/builtin-record.c +80 -0
110 lines not on this page · raw message
Signature · 1 line
2.25.1
Re: [PATCH 1/5] perf record: Fix way of handling non-perf-event pollfds
Quoted 17 lines · Adrian Hunter
perf record
__cmd_record()does not poll evlist pollfds. Instead it polls thread_data[0].pollfd. That happens whether or not threads are being used.perf record duplicates evlist mmap pollfds as needed for separate threads. The non-perf-event represented by
evlist->ctl_fdhas to handled separately, which is done explicitly, duplicating it into the thread_data[0] pollfds. That approach neglects any other non-perf-event file descriptors. Currently there is alsodone_fdwhich needs the same handling.Add a new generalized approach.
Add
fdarray_flag__non_perf_eventto identify the file descriptors that need the special handling. For those cases, also keep a mapping of the evlist pollfd index and thread pollfd index, so that the evlist revents can be updated.
Quoted 6 lines
Although this patch adds the new handling, it does not take it into use. There is no functional change, but it is the precursor to a fix, so is marked as a fix.
Fixes 415ccb58f68a ("perf record: Introduce thread specific data array")Signed-off-by Adrian Hunter <[email protected]>
Thanks,
Ian
Quoted 143 lines
--- tools/lib/api/fd/array.h | 5 ++- tools/perf/builtin-record.c | 80 +++++++++++++++++++++++++++++++++++++ 2 files changed, 83 insertions(+), 2 deletions(-)tools/lib/api/fd/array.h +3
-2@@ -31,8 +31,9 @@ struct fdarray { };enum fdarray_flags { - fdarray_flag__default =0x00000000 , - fdarray_flag__nonfilterable =0x00000001 + fdarray_flag__default =0x00000000 , + fdarray_flag__nonfilterable =0x00000001 , + fdarray_flag__non_perf_event =0x00000002 , };void fdarray__init (struct fdarray *fda,int nr_autogrow);tools/perf/builtin-record.c +80
-1@@ -143,6 +143,11 @@ static const char *thread_spec_tags[THREAD_SPEC__MAX] = {"undefined" ,"cpu" ,"core" ,"package" ,"numa" ,"user" }; +struct pollfd_index_map { +int evlist_pollfd_index; +int thread_pollfd_index; +}; +struct record {struct perf_tool tool;struct record_opts opts;@@ -171,6 +176,9 @@ struct record {int nr_threads;struct thread_mask *thread_masks;struct record_thread *thread_data; +struct pollfd_index_map *index_map; +size_t index_map_sz; +size_t index_map_cnt; };static volatile int done;@@ -1074,6 +1082,70 @@ static void record__free_thread_data(struct record *rec)zfree (&rec->thread_data); } +static int record__map_thread_evlist_pollfd_indexes (struct record *rec, +int evlist_pollfd_index, +int thread_pollfd_index) +{ +size_t x = rec->index_map_cnt; + +if (realloc_array_as_needed (rec->index_map, rec->index_map_sz, x,NULL )) +return -ENOMEM ; + rec->index_map[x].evlist_pollfd_index = evlist_pollfd_index; + rec->index_map[x].thread_pollfd_index = thread_pollfd_index; + rec->index_map_cnt +=1 ; +return 0 ; +} + +static int record__update_evlist_pollfd_from_thread (struct record *rec, +struct evlist *evlist, +struct record_thread *thread_data) +{ +struct pollfd *e_entries = evlist->core.pollfd.entries; +struct pollfd *t_entries = thread_data->pollfd.entries; +int err =0 ; +size_t i; + +for (i =0 ; i < rec->index_map_cnt; i++) { +int e_pos = rec->index_map[i].evlist_pollfd_index; +int t_pos = rec->index_map[i].thread_pollfd_index; + +if (e_entries[e_pos].fd != t_entries[t_pos].fd || + e_entries[e_pos].events != t_entries[t_pos].events) { +pr_err ("Thread and evlist pollfd index mismatch\n" ); + err = -EINVAL ; +continue ; + } + e_entries[e_pos].revents = t_entries[t_pos].revents; + } +return err; +} + +static int record__dup_non_perf_events (struct record *rec, +struct evlist *evlist, +struct record_thread *thread_data) +{ +struct fdarray *fda = &evlist->core.pollfd; +int i, ret; + +for (i =0 ; i < fda->nr; i++) { +if (!(fda->priv[i].flags & fdarray_flag__non_perf_event)) +continue ; + ret =fdarray__dup_entry_from (&thread_data->pollfd, i, fda); +if (ret <0 ) { +pr_err ("Failed to duplicate descriptor in main thread pollfd\n" ); +return ret; + } +pr_debug2 ("thread_data[%p]: pollfd[%d] <- non_perf_event fd=%d\n" , + thread_data, ret, fda->entries[i].fd); + ret =record__map_thread_evlist_pollfd_indexes (rec, i, ret); +if (ret <0 ) { +pr_err ("Failed to map thread and evlist pollfd indexes\n" ); +return ret; + } + } +return 0 ; +} +static int record__alloc_thread_data (struct record *rec,struct evlist *evlist) {int t, ret;@@ -1121,6 +1193,11 @@ static int record__alloc_thread_data(struct record *rec, struct evlist *evlist) thread_data[t].pipes.msg[0]); }else { thread_data[t].tid =gettid (); + + ret =record__dup_non_perf_events (rec, evlist, &thread_data[t]); +if (ret <0 ) + goto out_free; +if (evlist->ctl_fd.pos ==-1 )continue ; ret = fdarray__dup_entry_from(&thread_data[t].pollfd, evlist->ctl_fd.pos,@@ -2530,6 +2607,9 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) record__thread_munmap_filtered, NULL) == 0) draining = true; + err =record__update_evlist_pollfd_from_thread (rec, rec->evlist, thread); +if (err) + goto out_child;evlist__ctlfd_update (rec->evlist, &thread->pollfd.entries[thread->ctlfd_pos]); } --2.25.1
Adrian Hunter
reviewerPATCH 2/5 perf record: Fix done_fd wakeup event
In reply to Adrian Hunterevlist__add_wakeup_eventfd() calls perf_evlist__add_pollfd() to add a non-perf-event to the evlist pollfds. Since commit 415ccb58f68a ("perf record: Introduce thread specific data array") that doesn't work because evlist pollfs is not polled and done_fd is not duplicated into thread-data.
Patch "perf record: Fix way of handling non-perf-event pollfds" added a new approach that ensures file descriptors like done_fd are handled correctly by flagging them as fdarray_flag__non_perf_event.
Fix by flagging done_fd as fdarray_flag__non_perf_event.
Example:
Before:
$ sleep 3 & perf record -vv -p $! ... thread_data[0x55f44bd34140]: pollfd[0] <- event_fd=5 thread_data[0x55f44bd34140]: pollfd[1] <- event_fd=6 thread_data[0x55f44bd34140]: pollfd[2] <- event_fd=7 thread_data[0x55f44bd34140]: pollfd[3] <- event_fd=8 thread_data[0x55f44bd34140]: pollfd[4] <- event_fd=9 thread_data[0x55f44bd34140]: pollfd[5] <- event_fd=10 thread_data[0x55f44bd34140]: pollfd[6] <- event_fd=11 thread_data[0x55f44bd34140]: pollfd[7] <- event_fd=12 ...
After:
$ sleep 3 & perf record -vv -p $! ... thread_data[0x55a8ded89140]: pollfd[0] <- event_fd=5 thread_data[0x55a8ded89140]: pollfd[1] <- event_fd=6 thread_data[0x55a8ded89140]: pollfd[2] <- event_fd=7 thread_data[0x55a8ded89140]: pollfd[3] <- event_fd=8 thread_data[0x55a8ded89140]: pollfd[4] <- event_fd=9 thread_data[0x55a8ded89140]: pollfd[5] <- event_fd=10 thread_data[0x55a8ded89140]: pollfd[6] <- event_fd=11 thread_data[0x55a8ded89140]: pollfd[7] <- event_fd=12 thread_data[0x55a8ded89140]: pollfd[8] <- non_perf_event fd=4 ...
This patch depends on "perf record: Fix way of handling non-perf-event pollfds".
--- tools/perf/util/evlist.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
tools/perf/util/evlist.c +2 -1
@@ -608,7 +608,8 @@ int evlist__filter_pollfd(struct evlist *evlist, short revents_and_mask)int evlist__add_wakeup_eventfd (struct evlist *evlist,int fd) {return perf_evlist__add_pollfd (&evlist->core, fd,NULL ,POLLIN , - fdarray_flag__nonfilterable); + fdarray_flag__nonfilterable | + fdarray_flag__non_perf_event); }#endif
Signature · 1 line
2.25.1
Re: [PATCH 2/5] perf record: Fix done_fd wakeup event
Quoted 44 lines · Adrian Hunter
evlist__add_wakeup_eventfd()callsperf_evlist__add_pollfd()to add a non-perf-event to the evlist pollfds. Since commit415ccb58f68a("perf record: Introduce thread specific data array") that doesn't work because evlist pollfs is not polled anddone_fdis not duplicated into thread-data.Patch "perf record: Fix way of handling non-perf-event pollfds" added a new approach that ensures file descriptors like
done_fdare handled correctly by flagging them asfdarray_flag__non_perf_event.Fix by flagging
done_fdasfdarray_flag__non_perf_event.Example:
Before:
$ sleep 3 & perf record -vv -p $! ... thread_data[0x55f44bd34140]: pollfd[0] <- event_fd=5 thread_data[0x55f44bd34140]: pollfd[1] <- event_fd=6 thread_data[0x55f44bd34140]: pollfd[2] <- event_fd=7 thread_data[0x55f44bd34140]: pollfd[3] <- event_fd=8 thread_data[0x55f44bd34140]: pollfd[4] <- event_fd=9 thread_data[0x55f44bd34140]: pollfd[5] <- event_fd=10 thread_data[0x55f44bd34140]: pollfd[6] <- event_fd=11 thread_data[0x55f44bd34140]: pollfd[7] <- event_fd=12 ...After:
$ sleep 3 & perf record -vv -p $! ... thread_data[0x55a8ded89140]: pollfd[0] <- event_fd=5 thread_data[0x55a8ded89140]: pollfd[1] <- event_fd=6 thread_data[0x55a8ded89140]: pollfd[2] <- event_fd=7 thread_data[0x55a8ded89140]: pollfd[3] <- event_fd=8 thread_data[0x55a8ded89140]: pollfd[4] <- event_fd=9 thread_data[0x55a8ded89140]: pollfd[5] <- event_fd=10 thread_data[0x55a8ded89140]: pollfd[6] <- event_fd=11 thread_data[0x55a8ded89140]: pollfd[7] <- event_fd=12 thread_data[0x55a8ded89140]: pollfd[8] <- non_perf_event fd=4 ...
Quoted 5 lines
This patch depends on "perf record: Fix way of handling non-perf-event pollfds".
Fixes 415ccb58f68a ("perf record: Introduce thread specific data array")Signed-off-by Adrian Hunter <[email protected]>
Thanks,
Ian
Quoted 21 lines
--- tools/perf/util/evlist.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)tools/perf/util/evlist.c +2
-2@@ -608,7 +608,8 @@ int evlist__filter_pollfd(struct evlist *evlist, short revents_and_mask)int evlist__add_wakeup_eventfd (struct evlist *evlist,int fd) {return perf_evlist__add_pollfd (&evlist->core, fd,NULL ,POLLIN , - fdarray_flag__nonfilterable); + fdarray_flag__nonfilterable | + fdarray_flag__non_perf_event); }#endif --2.25.1
Adrian Hunter
reviewerPATCH 3/5 perf record: Change evlist->ctl_fd to use fdarray_flag__non_perf_event
In reply to Adrian HunterPatch "perf record: Fix way of handling non-perf-event pollfds" added a generic way to handle non-perf-event file descriptors like evlist->ctl_fd. Use it instead of handling evlist->ctl_fd separately.
--- tools/perf/builtin-record.c | 15 +-------------- tools/perf/util/evlist.c | 19 ++----------------- tools/perf/util/evlist.h | 1 - 3 files changed, 3 insertions(+), 32 deletions(-)
tools/perf/builtin-record.c +1 -14
@@ -1198,18 +1198,7 @@ static int record__alloc_thread_data(struct record *rec, struct evlist *evlist)if (ret <0 ) goto out_free; -if (evlist->ctl_fd.pos ==-1 ) -continue ; - ret =fdarray__dup_entry_from (&thread_data[t].pollfd, evlist->ctl_fd.pos, - &evlist->core.pollfd); -if (ret <0 ) { -pr_err ("Failed to duplicate descriptor in main thread pollfd\n" ); - goto out_free; - } - thread_data[t].ctlfd_pos = ret; -pr_debug2 ("thread_data[%p]: pollfd[%d] <- ctl_fd=%d\n" , - thread_data, thread_data[t].ctlfd_pos, - evlist->core.pollfd.entries[evlist->ctl_fd.pos].fd); + thread_data[t].ctlfd_pos =-1 ;/* Not used */ } }
@@ -2610,8 +2599,6 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) err =record__update_evlist_pollfd_from_thread (rec, rec->evlist, thread);if (err) goto out_child; -evlist__ctlfd_update (rec->evlist, - &thread->pollfd.entries[thread->ctlfd_pos]); }if (evlist__ctlfd_process (rec->evlist, &cmd) >0 ) {
tools/perf/util/evlist.c +2 -17
@@ -1898,7 +1898,8 @@ int evlist__initialize_ctlfd(struct evlist *evlist, int fd, int ack) } evlist->ctl_fd.pos =perf_evlist__add_pollfd (&evlist->core, fd,NULL ,POLLIN , - fdarray_flag__nonfilterable); + fdarray_flag__nonfilterable | + fdarray_flag__non_perf_event);if (evlist->ctl_fd.pos <0 ) { evlist->ctl_fd.pos =-1 ;pr_err ("Failed to add ctl fd entry: %m\n" );
@@ -2148,22 +2149,6 @@ int evlist__ctlfd_process(struct evlist *evlist, enum evlist_ctl_cmd *cmd)return err; } -int evlist__ctlfd_update (struct evlist *evlist,struct pollfd *update) -{ -int ctlfd_pos = evlist->ctl_fd.pos; -struct pollfd *entries = evlist->core.pollfd.entries; - -if (!evlist__ctlfd_initialized (evlist)) -return 0 ; - -if (entries[ctlfd_pos].fd != update->fd || - entries[ctlfd_pos].events != update->events) -return -1 ; - - entries[ctlfd_pos].revents = update->revents; -return 0 ; -} -struct evsel *evlist__find_evsel (struct evlist *evlist,int idx) {struct evsel *evsel;
tools/perf/util/evlist.h +0 -1
@@ -418,7 +418,6 @@ void evlist__close_control(int ctl_fd, int ctl_fd_ack, bool *ctl_fd_close);int evlist__initialize_ctlfd (struct evlist *evlist,int ctl_fd,int ctl_fd_ack);int evlist__finalize_ctlfd (struct evlist *evlist);bool evlist__ctlfd_initialized (struct evlist *evlist); -int evlist__ctlfd_update (struct evlist *evlist,struct pollfd *update);int evlist__ctlfd_process (struct evlist *evlist,enum evlist_ctl_cmd *cmd);int evlist__ctlfd_ack (struct evlist *evlist);
Signature · 1 line
2.25.1
Re: [PATCH 3/5] perf record: Change evlist->ctl_fd to use fdarray_flag__non_perf_event
Quoted 6 lines · Adrian Hunter
Patch "perf record: Fix way of handling non-perf-event pollfds" added a generic way to handle non-perf-event file descriptors like
evlist->ctl_fd. Use it instead of handlingevlist->ctl_fdseparately.Signed-off-by Adrian Hunter <[email protected]>
Thanks,
Ian
Quoted 91 lines
--- tools/perf/builtin-record.c | 15 +-------------- tools/perf/util/evlist.c | 19 ++----------------- tools/perf/util/evlist.h | 1 - 3 files changed, 3 insertions(+), 32 deletions(-)tools/perf/builtin-record.c +1
-14@@ -1198,18 +1198,7 @@ static int record__alloc_thread_data(struct record *rec, struct evlist *evlist)if (ret <0 ) goto out_free; -if (evlist->ctl_fd.pos ==-1 ) -continue ; - ret =fdarray__dup_entry_from (&thread_data[t].pollfd, evlist->ctl_fd.pos, - &evlist->core.pollfd); -if (ret <0 ) { -pr_err ("Failed to duplicate descriptor in main thread pollfd\n" ); - goto out_free; - } - thread_data[t].ctlfd_pos = ret; -pr_debug2 ("thread_data[%p]: pollfd[%d] <- ctl_fd=%d\n" , - thread_data, thread_data[t].ctlfd_pos, - evlist->core.pollfd.entries[evlist->ctl_fd.pos].fd); + thread_data[t].ctlfd_pos =-1 ;/* Not used */ } }@@ -2610,8 +2599,6 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) err =record__update_evlist_pollfd_from_thread (rec, rec->evlist, thread);if (err) goto out_child; -evlist__ctlfd_update (rec->evlist, - &thread->pollfd.entries[thread->ctlfd_pos]); }if (evlist__ctlfd_process (rec->evlist, &cmd) >0 ) {tools/perf/util/evlist.c +2
-17@@ -1898,7 +1898,8 @@ int evlist__initialize_ctlfd(struct evlist *evlist, int fd, int ack) } evlist->ctl_fd.pos =perf_evlist__add_pollfd (&evlist->core, fd,NULL ,POLLIN , - fdarray_flag__nonfilterable); + fdarray_flag__nonfilterable | + fdarray_flag__non_perf_event);if (evlist->ctl_fd.pos <0 ) { evlist->ctl_fd.pos =-1 ;pr_err ("Failed to add ctl fd entry: %m\n" );@@ -2148,22 +2149,6 @@ int evlist__ctlfd_process(struct evlist *evlist, enum evlist_ctl_cmd *cmd)return err; } -int evlist__ctlfd_update (struct evlist *evlist,struct pollfd *update) -{ -int ctlfd_pos = evlist->ctl_fd.pos; -struct pollfd *entries = evlist->core.pollfd.entries; - -if (!evlist__ctlfd_initialized (evlist)) -return 0 ; - -if (entries[ctlfd_pos].fd != update->fd || - entries[ctlfd_pos].events != update->events) -return -1 ; - - entries[ctlfd_pos].revents = update->revents; -return 0 ; -} -struct evsel *evlist__find_evsel (struct evlist *evlist,int idx) {struct evsel *evsel;tools/perf/util/evlist.h +0
-2@@ -418,7 +418,6 @@ void evlist__close_control(int ctl_fd, int ctl_fd_ack, bool *ctl_fd_close);int evlist__initialize_ctlfd (struct evlist *evlist,int ctl_fd,int ctl_fd_ack);int evlist__finalize_ctlfd (struct evlist *evlist);bool evlist__ctlfd_initialized (struct evlist *evlist); -int evlist__ctlfd_update (struct evlist *evlist,struct pollfd *update);int evlist__ctlfd_process (struct evlist *evlist,enum evlist_ctl_cmd *cmd);int evlist__ctlfd_ack (struct evlist *evlist); --2.25.1
Adrian Hunter
reviewerPATCH 4/5 perf evlist: Add evlist__{en/dis}able_non_dummy()
In reply to Adrian HunterDummy events are used to provide sideband information like MMAP events that are always needed even when main events are disabled. Add functions that take that into account.
--- tools/perf/util/evlist.c | 30 ++++++++++++++++++++++++------ tools/perf/util/evlist.h | 2 ++ 2 files changed, 26 insertions(+), 6 deletions(-)
tools/perf/util/evlist.c +24 -6
@@ -480,7 +480,7 @@ static int evlist__is_enabled(struct evlist *evlist)return false; } -static void __evlist__disable (struct evlist *evlist,char *evsel_name) +static void __evlist__disable (struct evlist *evlist,char *evsel_name,bool excl_dummy) {struct evsel *pos;struct evlist_cpu_iterator evlist_cpu_itr;
@@ -502,6 +502,8 @@ static void __evlist__disable(struct evlist *evlist, char *evsel_name)continue ;if (pos->disabled || !evsel__is_group_leader (pos) || !pos->core.fd)continue ; +if (excl_dummy &&evsel__is_dummy_event (pos)) +continue ;if (pos->immediate) has_imm = true;if (pos->immediate != imm)
@@ -518,6 +520,8 @@ static void __evlist__disable(struct evlist *evlist, char *evsel_name)continue ;if (!evsel__is_group_leader (pos) || !pos->core.fd)continue ; +if (excl_dummy &&evsel__is_dummy_event (pos)) +continue ; pos->disabled = true; }
@@ -533,15 +537,20 @@ static void __evlist__disable(struct evlist *evlist, char *evsel_name)void evlist__disable (struct evlist *evlist) { -__evlist__disable (evlist,NULL ); +__evlist__disable (evlist,NULL , false); +} + +void evlist__disable_non_dummy (struct evlist *evlist) +{ +__evlist__disable (evlist,NULL , true); }void evlist__disable_evsel (struct evlist *evlist,char *evsel_name) { -__evlist__disable (evlist, evsel_name); +__evlist__disable (evlist, evsel_name, false); } -static void __evlist__enable (struct evlist *evlist,char *evsel_name) +static void __evlist__enable (struct evlist *evlist,char *evsel_name,bool excl_dummy) {struct evsel *pos;struct evlist_cpu_iterator evlist_cpu_itr;
@@ -560,6 +569,8 @@ static void __evlist__enable(struct evlist *evlist, char *evsel_name)continue ;if (!evsel__is_group_leader (pos) || !pos->core.fd)continue ; +if (excl_dummy &&evsel__is_dummy_event (pos)) +continue ;evsel__enable_cpu (pos, evlist_cpu_itr.cpu_map_idx); }affinity__cleanup (affinity);
@@ -568,6 +579,8 @@ static void __evlist__enable(struct evlist *evlist, char *evsel_name)continue ;if (!evsel__is_group_leader (pos) || !pos->core.fd)continue ; +if (excl_dummy &&evsel__is_dummy_event (pos)) +continue ; pos->disabled = false; }
@@ -581,12 +594,17 @@ static void __evlist__enable(struct evlist *evlist, char *evsel_name)void evlist__enable (struct evlist *evlist) { -__evlist__enable (evlist,NULL ); +__evlist__enable (evlist,NULL , false); +} + +void evlist__enable_non_dummy (struct evlist *evlist) +{ +__evlist__enable (evlist,NULL , true); }void evlist__enable_evsel (struct evlist *evlist,char *evsel_name) { -__evlist__enable (evlist, evsel_name); +__evlist__enable (evlist, evsel_name, false); }void evlist__toggle_enable (struct evlist *evlist)
tools/perf/util/evlist.h +2 -0
@@ -205,6 +205,8 @@ void evlist__enable(struct evlist *evlist);void evlist__toggle_enable (struct evlist *evlist);void evlist__disable_evsel (struct evlist *evlist,char *evsel_name);void evlist__enable_evsel (struct evlist *evlist,char *evsel_name); +void evlist__disable_non_dummy (struct evlist *evlist); +void evlist__enable_non_dummy (struct evlist *evlist);void evlist__set_selected (struct evlist *evlist,struct evsel *evsel);
Signature · 1 line
2.25.1
Re: [PATCH 4/5] perf evlist: Add evlist__{en/dis}able_non_dummy()
Quoted 6 lines · Adrian Hunter
Dummy events are used to provide sideband information like MMAP events that are always needed even when main events are disabled. Add functions that take that into account.
Signed-off-by Adrian Hunter <[email protected]>
Thanks,
Ian
Quoted 37 lines
--- tools/perf/util/evlist.c | 30 ++++++++++++++++++++++++------ tools/perf/util/evlist.h | 2 ++ 2 files changed, 26 insertions(+), 6 deletions(-)tools/perf/util/evlist.c +5
-1@@ -480,7 +480,7 @@ static int evlist__is_enabled(struct evlist *evlist)return false; } -static void __evlist__disable (struct evlist *evlist,char *evsel_name) +static void __evlist__disable (struct evlist *evlist,char *evsel_name,bool excl_dummy) {struct evsel *pos;struct evlist_cpu_iterator evlist_cpu_itr;@@ -502,6 +502,8 @@ static void __evlist__disable(struct evlist *evlist, char *evsel_name)continue ;if (pos->disabled || !evsel__is_group_leader (pos) || !pos->core.fd)continue ; +if (excl_dummy &&evsel__is_dummy_event (pos)) +continue ;if (pos->immediate) has_imm = true;if (pos->immediate != imm)@@ -518,6 +520,8 @@ static void __evlist__disable(struct evlist *evlist, char *evsel_name)continue ;if (!evsel__is_group_leader (pos) || !pos->core.fd)continue ; +if (excl_dummy &&evsel__is_dummy_event (pos)) +continue ; pos->disabled = true; }
tools/perf/util/evlist.c static void __evlist__disable(struct evlist *evlist, char *evsel_name) · line 545
Quoted 10 lines
tools/perf/util/evlist.c +6
-1@@ -533,15 +537,20 @@ static void __evlist__disable(struct evlist *evlist, char *evsel_name)void evlist__disable (struct evlist *evlist) { -__evlist__disable (evlist,NULL ); +__evlist__disable (evlist,NULL , false); +} + +void evlist__disable_non_dummy (struct evlist *evlist) +{ +__evlist__disable (evlist,NULL , true);
nit: I think it is better to prefer to name the argument when passing constants, it can avoid errors if later constants are introduced, etc. It can also be checked by compiler warnings. As this is limited to the same file it is not hugely critical.
Quoted 67 lines
tools/perf/util/evlist.c +13
-4}void evlist__disable_evsel (struct evlist *evlist,char *evsel_name) { -__evlist__disable (evlist, evsel_name); +__evlist__disable (evlist, evsel_name, false); } -static void __evlist__enable (struct evlist *evlist,char *evsel_name) +static void __evlist__enable (struct evlist *evlist,char *evsel_name,bool excl_dummy) {struct evsel *pos;struct evlist_cpu_iterator evlist_cpu_itr;@@ -560,6 +569,8 @@ static void __evlist__enable(struct evlist *evlist, char *evsel_name)continue ;if (!evsel__is_group_leader (pos) || !pos->core.fd)continue ; +if (excl_dummy &&evsel__is_dummy_event (pos)) +continue ;evsel__enable_cpu (pos, evlist_cpu_itr.cpu_map_idx); }affinity__cleanup (affinity);@@ -568,6 +579,8 @@ static void __evlist__enable(struct evlist *evlist, char *evsel_name)continue ;if (!evsel__is_group_leader (pos) || !pos->core.fd)continue ; +if (excl_dummy &&evsel__is_dummy_event (pos)) +continue ; pos->disabled = false; }@@ -581,12 +594,17 @@ static void __evlist__enable(struct evlist *evlist, char *evsel_name)void evlist__enable (struct evlist *evlist) { -__evlist__enable (evlist,NULL ); +__evlist__enable (evlist,NULL , false); +} + +void evlist__enable_non_dummy (struct evlist *evlist) +{ +__evlist__enable (evlist,NULL , true); }void evlist__enable_evsel (struct evlist *evlist,char *evsel_name) { -__evlist__enable (evlist, evsel_name); +__evlist__enable (evlist, evsel_name, false); }void evlist__toggle_enable (struct evlist *evlist)tools/perf/util/evlist.h +2
-1@@ -205,6 +205,8 @@ void evlist__enable(struct evlist *evlist);void evlist__toggle_enable (struct evlist *evlist);void evlist__disable_evsel (struct evlist *evlist,char *evsel_name);void evlist__enable_evsel (struct evlist *evlist,char *evsel_name); +void evlist__disable_non_dummy (struct evlist *evlist); +void evlist__enable_non_dummy (struct evlist *evlist);void evlist__set_selected (struct evlist *evlist,struct evsel *evsel); --2.25.1
Adrian Hunter
reviewerPATCH 5/5 perf record: Allow multiple recording time ranges
In reply to Adrian HunterAUX area traces can produce too much data to record successfully or analyze subsequently. Add another means to reduce data collection by allowing multiple recording time ranges.
This is useful, for instance, in cases where a workload produces predictably reproducible events in specific time ranges.
Today we only have perf record -D <msecs> to start at a specific region, or some complicated approach using snapshot mode and external scripts sending signals or using the fifos. But these approaches are difficult to set up compared with simply having perf do it.
Extend perf record option -D/--delay option to specifying relative time stamps for start stop controlled by perf with the right time offset, for instance:
perf record -e intel_pt// -D 10-20,30-40
to record 10ms to 20ms into the trace and 30ms to 40ms.
Example:
The example workload is:
$ cat repeat-usleep.c
int usleep(useconds_t usec);
int usage(int ret, const char *msg)
{
if (msg)
fprintf(stderr, "%s\n", msg);
fprintf(stderr, "Usage is: repeat-usleep <microseconds>\n");
return ret;}
int main(int argc, char *argv[])
{
unsigned long usecs;
char *end_ptr;
if (argc != 2)
return usage(1, "Error: Wrong number of arguments!");
errno = 0;
usecs = strtoul(argv[1], &end_ptr, 0);
if (errno || *end_ptr || usecs > UINT_MAX)
return usage(1, "Error: Invalid argument!");
while (1) {
int ret = usleep(usecs);
if (ret & errno != EINTR)
return usage(1, "Error: usleep() failed!");
}
return 0;}
$ perf record -e intel_pt//u --delay 10-20,40-70,110-160 -- ./repeat-usleep 500 Events disabled Events enabled Events disabled Events enabled Events disabled Events enabled Events disabled [ perf record: Woken up 5 times to write data ] [ perf record: Captured and wrote 0.204 MB perf.data ] Terminated
A dlfilter is used to determine continuous data collection (timestamps less than 1ms apart):
$ cat dlfilter-show-delays.c
static __u64 start_time;
static __u64 last_time;
int start(void **data, void *ctx)
{
printf("%-17s\t%-9s\t%-6s\n", " Time", " Duration", " Delay");
return 0;}
int filter_event_early(void *data, const struct perf_dlfilter_sample *sample, void *ctx)
{
__u64 delta;
if (!sample->time)
return 1;
if (!last_time)
goto out;
delta = sample->time - last_time;
if (delta < 1000000)
goto out2;;
printf("%17.9f\t%9.1f\t%6.1f\n", start_time / 1000000000.0, (last_time - start_time) / 1000000.0, delta / 1000000.0);
out:
start_time = sample->time;
out2:
last_time = sample->time;
return 1;}
int stop(void *data, void *ctx)
{
printf("%17.9f\t%9.1f\n", start_time / 1000000000.0, (last_time - start_time) / 1000000.0);
return 0;}
The result shows the times roughly match the --delay option:
$ perf script --itrace=qb --dlfilter dlfilter-show-delays.so Time Duration Delay 39215.302317300 9.7 20.5 39215.332480217 30.4 40.9 39215.403837717 49.8
--- tools/perf/Documentation/perf-record.txt | 6 +- tools/perf/builtin-record.c | 24 ++- tools/perf/util/evlist.c | 234 +++++++++++++++++++++++ tools/perf/util/evlist.h | 9 + 4 files changed, 269 insertions(+), 4 deletions(-)
tools/perf/Documentation/perf-record.txt +4 -2
@@ -430,8 +430,10 @@ if combined with -a or -C options. -D:: --delay=:: After starting the program, wait msecs before measuring (-1: start with events -disabled). This is useful to filter out the startup phase of the program, which -is often very different. +disabled), or enable events only for specified ranges of msecs (e.g. +-D 10-20,30-40 means wait 10 msecs, enable for 10 msecs, wait 10 msecs, enable +for 10 msecs, then stop). Note, delaying enabling of events is useful to filter +out the startup phase of the program, which is often very different. -I:: --intr-regs::
tools/perf/builtin-record.c +22 -2
@@ -2498,6 +2498,10 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) } } + err =event_enable_timer__start (rec->evlist->eet); +if (err) + goto out_child; +trigger_ready (&auxtrace_snapshot_trigger);trigger_ready (&switch_output_trigger);perf_hooks__invoke_record_start ();
@@ -2621,6 +2625,14 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) } } + err =event_enable_timer__process (rec->evlist->eet); +if (err <0 ) + goto out_child; +if (err) { + err =0 ; + done =1 ; + } +/* * When perf is starting the traced process, at the end events * die with the process and we wait for that. Thus no need to
@@ -2842,6 +2854,12 @@ static int perf_record_config(const char *var, const char *value, void *cb)return 0 ; } +static int record__parse_event_enable_time (const struct option *opt,const char *str,int unset) +{ +struct record *rec = (struct record *)opt->value; + +return evlist__parse_event_enable_time (rec->evlist, &rec->opts, str, unset); +}static int record__parse_affinity (const struct option *opt,const char *str,int unset) {
@@ -3303,8 +3321,10 @@ static struct option __record_options[] = {OPT_CALLBACK ('G' ,"cgroup" , &record.evlist,"name" ,"monitor event in cgroup name only" , parse_cgroups), -OPT_INTEGER ('D' ,"delay" , &record.opts.initial_delay, -"ms to wait before starting measurement after program start (-1: start with events disabled)" ), +OPT_CALLBACK ('D' ,"delay" , &record,"ms" , +"ms to wait before starting measurement after program start (-1: start with events disabled), " +"or ranges of time to enable events e.g. '-D 10-20,30-40'" , + record__parse_event_enable_time),OPT_BOOLEAN (0 ,"kcore" , &record.opts.kcore,"copy /proc/kcore" ),OPT_STRING ('u' ,"uid" , &record.opts.target.uid_str,"user" , "user to profile"),
tools/perf/util/evlist.c +234 -0
264 lines not on this page · raw message
tools/perf/util/evlist.h +9 -0
@@ -48,6 +48,8 @@ enum bkw_mmap_state { BKW_MMAP_EMPTY, }; +struct event_enable_timer ; +struct evlist {struct perf_evlist core;bool enabled;
@@ -79,6 +81,7 @@ struct evlist {int ack;/* ack file descriptor for control commands */ int pos;/* index at evlist core object to check signals */ } ctl_fd; +struct event_enable_timer *eet; };struct evsel_str_handler {
@@ -426,6 +429,12 @@ int evlist__ctlfd_ack(struct evlist *evlist);#define EVLIST_ENABLED_MSG "Events enabled\n"#define EVLIST_DISABLED_MSG "Events disabled\n" +int evlist__parse_event_enable_time (struct evlist *evlist,struct record_opts *opts, +const char *str,int unset); +int event_enable_timer__start (struct event_enable_timer *eet); +void event_enable_timer__exit (struct event_enable_timer **ep); +int event_enable_timer__process (struct event_enable_timer *eet); +struct evsel *evlist__find_evsel (struct evlist *evlist,int idx);int evlist__scnprintf_evsels (struct evlist *evlist,size_t size,char *bf);
Signature · 1 line
2.25.1
Ian Rogers
reviewerQuoted 13 lines · Adrian Hunter
AUX area traces can produce too much data to record successfully or analyze subsequently. Add another means to reduce data collection by allowing multiple recording time ranges.
This is useful, for instance, in cases where a workload produces predictably reproducible events in specific time ranges.
Today we only have perf record -D <msecs> to start at a specific region, or some complicated approach using snapshot mode and external scripts sending signals or using the fifos. But these approaches are difficult to set up compared with simply having perf do it.
Quoted 7 lines
Extend perf record option -D/--delay option to specifying relative time stamps for start stop controlled by perf with the right time offset, for instance:
perf record -e intel_pt// -D 10-20,30-40to record 10ms to 20ms into the trace and 30ms to 40ms.
Could you add a test like this in:
https://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git/tree/tools/perf/tests/shell/record.sh?h=perf/core
If for no other reason than code coverage. It would also be interesting to see what happens if floating point values are passed, ranges are incomplete, etc. As the functionality is generic the event could be cycles or instructions rather than intel_pt.
Thanks,
Ian
Quoted 490 lines
Example:
The example workload is:
$ cat repeat-usleep.cint
usleep(useconds_t usec);int
usage(int ret, const char *msg){ if (msg) fprintf(stderr, "%s\n", msg); fprintf(stderr, "Usage is: repeat-usleep <microseconds>\n"); return ret;}
int
main(int argc, char *argv[]){ unsigned long usecs; char *end_ptr; if (argc != 2) return usage(1, "Error: Wrong number of arguments!"); errno = 0; usecs = strtoul(argv[1], &end_ptr, 0); if (errno || *end_ptr || usecs > UINT_MAX) return usage(1, "Error: Invalid argument!"); while (1) { int ret = usleep(usecs); if (ret & errno != EINTR) return usage(1, "Error: usleep() failed!"); } return 0;}
$ perf record -e intel_pt//u --delay 10-20,40-70,110-160 -- ./repeat-usleep 500 Events disabled Events enabled Events disabled Events enabled Events disabled Events enabled Events disabled [ perf record: Woken up 5 times to write data ] [ perf record: Captured and wrote 0.204 MB perf.data ] TerminatedA dlfilter is used to determine continuous data collection (timestamps less than 1ms apart):
$ cat dlfilter-show-delays.cstatic
__u64start_time;
static__u64last_time;int
start(void **data, void *ctx){ printf("%-17s\t%-9s\t%-6s\n", " Time", " Duration", " Delay"); return 0;}
int
filter_event_early(void *data, const struct perf_dlfilter_sample *sample, void *ctx){ __u64 delta; if (!sample->time) return 1; if (!last_time) goto out; delta = sample->time - last_time; if (delta < 1000000) goto out2;; printf("%17.9f\t%9.1f\t%6.1f\n", start_time / 1000000000.0, (last_time - start_time) / 1000000.0, delta / 1000000.0); out: start_time = sample->time; out2: last_time = sample->time; return 1;}
int
stop(void *data, void *ctx){ printf("%17.9f\t%9.1f\n", start_time / 1000000000.0, (last_time - start_time) / 1000000.0); return 0;}
The result shows the times roughly match the --delay option:
$ perf script --itrace=qb --dlfilter dlfilter-show-delays.so Time Duration Delay 39215.302317300 9.7 20.5 39215.332480217 30.4 40.9 39215.403837717 49.8Signed-off-by Adrian Hunter <[email protected]>--- tools/perf/Documentation/perf-record.txt | 6 +- tools/perf/builtin-record.c | 24 ++- tools/perf/util/evlist.c | 234 +++++++++++++++++++++++ tools/perf/util/evlist.h | 9 + 4 files changed, 269 insertions(+), 4 deletions(-)tools/perf/Documentation/perf-record.txt +4
-2@@ -430,8 +430,10 @@ if combined with -a or -C options. -D:: --delay=:: After starting the program, wait msecs before measuring (-1: start with events -disabled). This is useful to filter out the startup phase of the program, which -is often very different. +disabled), or enable events only for specified ranges of msecs (e.g. +-D 10-20,30-40 means wait 10 msecs, enable for 10 msecs, wait 10 msecs, enable +for 10 msecs, then stop). Note, delaying enabling of events is useful to filter +out the startup phase of the program, which is often very different. -I:: --intr-regs::tools/perf/builtin-record.c +22
-2@@ -2498,6 +2498,10 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) } } + err =event_enable_timer__start (rec->evlist->eet); +if (err) + goto out_child; +trigger_ready (&auxtrace_snapshot_trigger);trigger_ready (&switch_output_trigger);perf_hooks__invoke_record_start ();@@ -2621,6 +2625,14 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) } } + err =event_enable_timer__process (rec->evlist->eet); +if (err <0 ) + goto out_child; +if (err) { + err =0 ; + done =1 ; + } +/* * When perf is starting the traced process, at the end events * die with the process and we wait for that. Thus no need to @@ -2842,6 +2854,12 @@ static int perf_record_config(const char *var, const char *value, void *cb)return 0 ; } +static int record__parse_event_enable_time (const struct option *opt,const char *str,int unset) +{ +struct record *rec = (struct record *)opt->value; + +return evlist__parse_event_enable_time (rec->evlist, &rec->opts, str, unset); +}static int record__parse_affinity (const struct option *opt,const char *str,int unset) {@@ -3303,8 +3321,10 @@ static struct option __record_options[] = {OPT_CALLBACK ('G' ,"cgroup" , &record.evlist,"name" ,"monitor event in cgroup name only" , parse_cgroups), -OPT_INTEGER ('D' ,"delay" , &record.opts.initial_delay, -"ms to wait before starting measurement after program start (-1: start with events disabled)" ), +OPT_CALLBACK ('D' ,"delay" , &record,"ms" , +"ms to wait before starting measurement after program start (-1: start with events disabled), " +"or ranges of time to enable events e.g. '-D 10-20,30-40'" , + record__parse_event_enable_time),OPT_BOOLEAN (0 ,"kcore" , &record.opts.kcore,"copy /proc/kcore" ),OPT_STRING ('u' ,"uid" , &record.opts.target.uid_str,"user" , "user to profile"),tools/perf/util/evlist.c +234
-0@@ -15,6 +15,7 @@#include "target.h" #include "evlist.h" #include "evsel.h" +#include "record.h" #include "debug.h" #include "units.h" #include "bpf_counter.h" @@ -40,12 +41,14 @@#include <sys/ioctl.h> #include <sys/mman.h> #include <sys/prctl.h> +#include <sys/timerfd.h> #include <linux/bitops.h> #include <linux/hash.h> #include <linux/log2.h> #include <linux/err.h> #include <linux/string.h> +#include <linux/time64.h> #include <linux/zalloc.h> #include <perf/evlist.h> #include <perf/evsel.h> @@ -147,6 +150,7 @@ static void evlist__purge(struct evlist *evlist)void evlist__exit (struct evlist *evlist) { +event_enable_timer__exit (&evlist->eet);zfree (&evlist->mmap);zfree (&evlist->overwrite_mmap);perf_evlist__exit (&evlist->core);@@ -2167,6 +2171,236 @@ int evlist__ctlfd_process(struct evlist *evlist, enum evlist_ctl_cmd *cmd)return err; } +/** +* struct event_enable_time - perf record -D/--delay single time range. +* @start: start of time range to enable events in milliseconds +* @end: end of time range to enable events in milliseconds +* +* N.B. this structure is also accessed as an array of int. +*/ +struct event_enable_time { +int start; +int end; +}; + +static int parse_event_enable_time (const char *str,struct event_enable_time *range,bool first) +{ +const char *fmt = first ?"%u - %u %n" :" , %u - %u %n" ; +int ret, start, end, n; + + ret =sscanf (str, fmt, &start, &end, &n); +if (ret !=2 || end <= start) +return -EINVAL ; +if (range) { + range->start = start; + range->end = end; + } +return n; +} + +static ssize_t parse_event_enable_times (const char *str,struct event_enable_time *range) +{ +int incr = !!range; +bool first = true; +ssize_t ret, cnt; + +for (cnt =0 ; *str; cnt++) { + ret =parse_event_enable_time (str, range, first); +if (ret <0 ) +return ret; +/* Check no overlap */ +if (!first && range && range->start <= range[-1 ].end) +return -EINVAL ; + str += ret; + range += incr; + first = false; + } +return cnt; +} + +/** +* struct event_enable_timer - control structure for perf record -D/--delay. +* @evlist: event list +* @times: time ranges that events are enabled (N.B. this is also accessed as an +* array of int) +* @times_cnt: number of time ranges +* @timerfd: timer file descriptor +* @pollfd_pos: position in @evlist array of file descriptors to poll (fdarray) +* @times_step: current position in (int *)@times)[], +* refer event_enable_timer__process() +* +* Note, this structure is only used when there are time ranges, not when there +* is only an initial delay. +*/ +struct event_enable_timer { +struct evlist *evlist; +struct event_enable_time *times; +size_t times_cnt; +int timerfd; +int pollfd_pos; +size_t times_step; +}; + +static int str_to_delay (const char *str) +{ +char *endptr; +long d; + + d =strtol (str, &endptr,10 ); +if (*endptr || d >INT_MAX || d <-1 ) +return 0 ; +return d; +} + +int evlist__parse_event_enable_time (struct evlist *evlist,struct record_opts *opts, +const char *str,int unset) +{ +enum fdarray_flags flags = fdarray_flag__nonfilterable | fdarray_flag__non_perf_event; +struct event_enable_timer *eet; +ssize_t times_cnt; +ssize_t ret; +int err; + +if (unset) +return 0 ; + + opts->initial_delay =str_to_delay (str); +if (opts->initial_delay) +return 0 ; + + ret =parse_event_enable_times (str,NULL ); +if (ret <0 ) +return ret; + + times_cnt = ret; +if (times_cnt ==0 ) +return -EINVAL ; + + eet =zalloc (sizeof (*eet)); +if (!eet) +return -ENOMEM ; + + eet->times =calloc (times_cnt,sizeof (*eet->times)); +if (!eet->times) { + err = -ENOMEM ; + goto free_eet; + } + +if (parse_event_enable_times (str, eet->times) != times_cnt) { + err = -EINVAL ; + goto free_eet_times; + } + + eet->times_cnt = times_cnt; + + eet->timerfd =timerfd_create (CLOCK_MONOTONIC ,TFD_CLOEXEC ); +if (eet->timerfd ==-1 ) { + err = -errno; +pr_err ("timerfd_create failed: %s\n" ,strerror (errno)); + goto free_eet_times; + } + + eet->pollfd_pos =perf_evlist__add_pollfd (&evlist->core, eet->timerfd,NULL ,POLLIN , flags); +if (eet->pollfd_pos <0 ) { + err = eet->pollfd_pos; + goto close_timerfd; + } + + eet->evlist = evlist; + evlist->eet = eet; + opts->initial_delay = eet->times[0 ].start; + +return 0 ; + +close_timerfd: +close (eet->timerfd); +free_eet_times: +free (eet->times); +free_eet: +free (eet); +return err; +} + +static int event_enable_timer__set_timer (struct event_enable_timer *eet,int ms) +{ +struct itimerspec its = { + .it_value.tv_sec = ms /MSEC_PER_SEC , + .it_value.tv_nsec = (ms %MSEC_PER_SEC ) *NSEC_PER_MSEC , + }; +int err =0 ; + +if (timerfd_settime (eet->timerfd,0 , &its,NULL ) <0 ) { + err = -errno; +pr_err ("timerfd_settime failed: %s\n" ,strerror (errno)); + } +return err; +} + +int event_enable_timer__start (struct event_enable_timer *eet) +{ +int ms; + +if (!eet) +return 0 ; + + ms = eet->times[0 ].end - eet->times[0 ].start; + eet->times_step =1 ; + +return event_enable_timer__set_timer (eet, ms); +} + +int event_enable_timer__process (struct event_enable_timer *eet) +{ +struct pollfd *entries; +short revents; + +if (!eet) +return 0 ; + + entries = eet->evlist->core.pollfd.entries; + revents = entries[eet->pollfd_pos].revents; + entries[eet->pollfd_pos].revents =0 ; + +if (revents &POLLIN ) { +size_t step = eet->times_step; +size_t pos = step /2 ; + +if (step &1 ) { +evlist__disable_non_dummy (eet->evlist); +pr_info (EVLIST_DISABLED_MSG ); +if (pos >= eet->times_cnt -1 ) { +/* Disarm timer */ +event_enable_timer__set_timer (eet,0 ); +return 1 ;/* Stop */ + } + }else { +evlist__enable_non_dummy (eet->evlist); +pr_info (EVLIST_ENABLED_MSG ); + } + + step +=1 ; + pos = step /2 ; + +if (pos < eet->times_cnt) { +int *times = (int *)eet->times;/* Accessing 'times' as array of int */ +int ms = times[step] - times[step -1 ]; + + eet->times_step = step; +return event_enable_timer__set_timer (eet, ms); + } + } + +return 0 ; +} + +void event_enable_timer__exit (struct event_enable_timer **ep) +{ +if (!ep || !*ep) +return ; +free ((*ep)->times); +zfree (ep); +} +struct evsel *evlist__find_evsel (struct evlist *evlist,int idx) {struct evsel *evsel;tools/perf/util/evlist.h +9
-1@@ -48,6 +48,8 @@ enum bkw_mmap_state { BKW_MMAP_EMPTY, }; +struct event_enable_timer ; +struct evlist {struct perf_evlist core;bool enabled;@@ -79,6 +81,7 @@ struct evlist {int ack;/* ack file descriptor for control commands */ int pos;/* index at evlist core object to check signals */ } ctl_fd; +struct event_enable_timer *eet; };struct evsel_str_handler {@@ -426,6 +429,12 @@ int evlist__ctlfd_ack(struct evlist *evlist);#define EVLIST_ENABLED_MSG "Events enabled\n"#define EVLIST_DISABLED_MSG "Events disabled\n" +int evlist__parse_event_enable_time (struct evlist *evlist,struct record_opts *opts, +const char *str,int unset); +int event_enable_timer__start (struct event_enable_timer *eet); +void event_enable_timer__exit (struct event_enable_timer **ep); +int event_enable_timer__process (struct event_enable_timer *eet); +struct evsel *evlist__find_evsel (struct evlist *evlist,int idx);int evlist__scnprintf_evsels (struct evlist *evlist,size_t size,char *bf); --2.25.1
Adrian Hunter
reviewerQuoted 20 lines · Ian Rogers
Quoted 19 lines · Adrian Hunter
AUX area traces can produce too much data to record successfully or analyze subsequently. Add another means to reduce data collection by allowing multiple recording time ranges.
This is useful, for instance, in cases where a workload produces predictably reproducible events in specific time ranges.
Today we only have perf record -D <msecs> to start at a specific region, or some complicated approach using snapshot mode and external scripts sending signals or using the fifos. But these approaches are difficult to set up compared with simply having perf do it.
Extend perf record option -D/--delay option to specifying relative time stamps for start stop controlled by perf with the right time offset, for instance:
perf record -e intel_pt// -D 10-20,30-40
Quoted 8 lines
Quoted 1 line
to record 10ms to 20ms into the trace and 30ms to 40ms.
Could you add a test like this in:
https://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git/tree/tools/perf/tests/shell/record.sh?h=perf/core
If for no other reason than code coverage. It would also be interesting to see what happens if floating point values are passed, ranges are incomplete, etc. As the functionality is generic the event could be cycles or instructions rather thanintel_pt.
Just checking arguments does not add much value. Really it needs to get information about what perf is doing. A simple way would be to add debug messages specifically formatted to give that information. Then a test could capture them and check they match what is expected.
For example, to test the -D option, print debug messages with timestamps when events are enabled / disabled, when the workload starts etc.
Maybe want a new pr_... for such messages like pr_test() that print at, say, verbose level 3 or 4 (or 5)? Or a new "verbose" option just for test messages?
Quoted 494 lines
Thanks,
IanQuoted 490 lines
Example:
The example workload is:
$ cat repeat-usleep.cint
usleep(useconds_t usec);int
usage(int ret, const char *msg){ if (msg) fprintf(stderr, "%s\n", msg); fprintf(stderr, "Usage is: repeat-usleep <microseconds>\n"); return ret;}
int
main(int argc, char *argv[]){ unsigned long usecs; char *end_ptr; if (argc != 2) return usage(1, "Error: Wrong number of arguments!"); errno = 0; usecs = strtoul(argv[1], &end_ptr, 0); if (errno || *end_ptr || usecs > UINT_MAX) return usage(1, "Error: Invalid argument!"); while (1) { int ret = usleep(usecs); if (ret & errno != EINTR) return usage(1, "Error: usleep() failed!"); } return 0;}
$ perf record -e intel_pt//u --delay 10-20,40-70,110-160 -- ./repeat-usleep 500 Events disabled Events enabled Events disabled Events enabled Events disabled Events enabled Events disabled [ perf record: Woken up 5 times to write data ] [ perf record: Captured and wrote 0.204 MB perf.data ] TerminatedA dlfilter is used to determine continuous data collection (timestamps less than 1ms apart):
$ cat dlfilter-show-delays.cstatic
__u64start_time;
static__u64last_time;int
start(void **data, void *ctx){ printf("%-17s\t%-9s\t%-6s\n", " Time", " Duration", " Delay"); return 0;}
int
filter_event_early(void *data, const struct perf_dlfilter_sample *sample, void *ctx){ __u64 delta; if (!sample->time) return 1; if (!last_time) goto out; delta = sample->time - last_time; if (delta < 1000000) goto out2;; printf("%17.9f\t%9.1f\t%6.1f\n", start_time / 1000000000.0, (last_time - start_time) / 1000000.0, delta / 1000000.0); out: start_time = sample->time; out2: last_time = sample->time; return 1;}
int
stop(void *data, void *ctx){ printf("%17.9f\t%9.1f\n", start_time / 1000000000.0, (last_time - start_time) / 1000000.0); return 0;}
The result shows the times roughly match the --delay option:
$ perf script --itrace=qb --dlfilter dlfilter-show-delays.so Time Duration Delay 39215.302317300 9.7 20.5 39215.332480217 30.4 40.9 39215.403837717 49.8Signed-off-by Adrian Hunter <[email protected]>--- tools/perf/Documentation/perf-record.txt | 6 +- tools/perf/builtin-record.c | 24 ++- tools/perf/util/evlist.c | 234 +++++++++++++++++++++++ tools/perf/util/evlist.h | 9 + 4 files changed, 269 insertions(+), 4 deletions(-)tools/perf/Documentation/perf-record.txt +4
-2@@ -430,8 +430,10 @@ if combined with -a or -C options. -D:: --delay=:: After starting the program, wait msecs before measuring (-1: start with events -disabled). This is useful to filter out the startup phase of the program, which -is often very different. +disabled), or enable events only for specified ranges of msecs (e.g. +-D 10-20,30-40 means wait 10 msecs, enable for 10 msecs, wait 10 msecs, enable +for 10 msecs, then stop). Note, delaying enabling of events is useful to filter +out the startup phase of the program, which is often very different. -I:: --intr-regs::tools/perf/builtin-record.c +22
-2@@ -2498,6 +2498,10 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) } } + err =event_enable_timer__start (rec->evlist->eet); +if (err) + goto out_child; +trigger_ready (&auxtrace_snapshot_trigger);trigger_ready (&switch_output_trigger);perf_hooks__invoke_record_start ();@@ -2621,6 +2625,14 @@ static int __cmd_record(struct record *rec, int argc, const char **argv) } } + err =event_enable_timer__process (rec->evlist->eet); +if (err <0 ) + goto out_child; +if (err) { + err =0 ; + done =1 ; + } +/* * When perf is starting the traced process, at the end events * die with the process and we wait for that. Thus no need to @@ -2842,6 +2854,12 @@ static int perf_record_config(const char *var, const char *value, void *cb)return 0 ; } +static int record__parse_event_enable_time (const struct option *opt,const char *str,int unset) +{ +struct record *rec = (struct record *)opt->value; + +return evlist__parse_event_enable_time (rec->evlist, &rec->opts, str, unset); +}static int record__parse_affinity (const struct option *opt,const char *str,int unset) {@@ -3303,8 +3321,10 @@ static struct option __record_options[] = {OPT_CALLBACK ('G' ,"cgroup" , &record.evlist,"name" ,"monitor event in cgroup name only" , parse_cgroups), -OPT_INTEGER ('D' ,"delay" , &record.opts.initial_delay, -"ms to wait before starting measurement after program start (-1: start with events disabled)" ), +OPT_CALLBACK ('D' ,"delay" , &record,"ms" , +"ms to wait before starting measurement after program start (-1: start with events disabled), " +"or ranges of time to enable events e.g. '-D 10-20,30-40'" , + record__parse_event_enable_time),OPT_BOOLEAN (0 ,"kcore" , &record.opts.kcore,"copy /proc/kcore" ),OPT_STRING ('u' ,"uid" , &record.opts.target.uid_str,"user" , "user to profile"),tools/perf/util/evlist.c +234
-0@@ -15,6 +15,7 @@#include "target.h" #include "evlist.h" #include "evsel.h" +#include "record.h" #include "debug.h" #include "units.h" #include "bpf_counter.h" @@ -40,12 +41,14 @@#include <sys/ioctl.h> #include <sys/mman.h> #include <sys/prctl.h> +#include <sys/timerfd.h> #include <linux/bitops.h> #include <linux/hash.h> #include <linux/log2.h> #include <linux/err.h> #include <linux/string.h> +#include <linux/time64.h> #include <linux/zalloc.h> #include <perf/evlist.h> #include <perf/evsel.h> @@ -147,6 +150,7 @@ static void evlist__purge(struct evlist *evlist)void evlist__exit (struct evlist *evlist) { +event_enable_timer__exit (&evlist->eet);zfree (&evlist->mmap);zfree (&evlist->overwrite_mmap);perf_evlist__exit (&evlist->core);@@ -2167,6 +2171,236 @@ int evlist__ctlfd_process(struct evlist *evlist, enum evlist_ctl_cmd *cmd)return err; } +/** +* struct event_enable_time - perf record -D/--delay single time range. +* @start: start of time range to enable events in milliseconds +* @end: end of time range to enable events in milliseconds +* +* N.B. this structure is also accessed as an array of int. +*/ +struct event_enable_time { +int start; +int end; +}; + +static int parse_event_enable_time (const char *str,struct event_enable_time *range,bool first) +{ +const char *fmt = first ?"%u - %u %n" :" , %u - %u %n" ; +int ret, start, end, n; + + ret =sscanf (str, fmt, &start, &end, &n); +if (ret !=2 || end <= start) +return -EINVAL ; +if (range) { + range->start = start; + range->end = end; + } +return n; +} + +static ssize_t parse_event_enable_times (const char *str,struct event_enable_time *range) +{ +int incr = !!range; +bool first = true; +ssize_t ret, cnt; + +for (cnt =0 ; *str; cnt++) { + ret =parse_event_enable_time (str, range, first); +if (ret <0 ) +return ret; +/* Check no overlap */ +if (!first && range && range->start <= range[-1 ].end) +return -EINVAL ; + str += ret; + range += incr; + first = false; + } +return cnt; +} + +/** +* struct event_enable_timer - control structure for perf record -D/--delay. +* @evlist: event list +* @times: time ranges that events are enabled (N.B. this is also accessed as an +* array of int) +* @times_cnt: number of time ranges +* @timerfd: timer file descriptor +* @pollfd_pos: position in @evlist array of file descriptors to poll (fdarray) +* @times_step: current position in (int *)@times)[], +* refer event_enable_timer__process() +* +* Note, this structure is only used when there are time ranges, not when there +* is only an initial delay. +*/ +struct event_enable_timer { +struct evlist *evlist; +struct event_enable_time *times; +size_t times_cnt; +int timerfd; +int pollfd_pos; +size_t times_step; +}; + +static int str_to_delay (const char *str) +{ +char *endptr; +long d; + + d =strtol (str, &endptr,10 ); +if (*endptr || d >INT_MAX || d <-1 ) +return 0 ; +return d; +} + +int evlist__parse_event_enable_time (struct evlist *evlist,struct record_opts *opts, +const char *str,int unset) +{ +enum fdarray_flags flags = fdarray_flag__nonfilterable | fdarray_flag__non_perf_event; +struct event_enable_timer *eet; +ssize_t times_cnt; +ssize_t ret; +int err; + +if (unset) +return 0 ; + + opts->initial_delay =str_to_delay (str); +if (opts->initial_delay) +return 0 ; + + ret =parse_event_enable_times (str,NULL ); +if (ret <0 ) +return ret; + + times_cnt = ret; +if (times_cnt ==0 ) +return -EINVAL ; + + eet =zalloc (sizeof (*eet)); +if (!eet) +return -ENOMEM ; + + eet->times =calloc (times_cnt,sizeof (*eet->times)); +if (!eet->times) { + err = -ENOMEM ; + goto free_eet; + } + +if (parse_event_enable_times (str, eet->times) != times_cnt) { + err = -EINVAL ; + goto free_eet_times; + } + + eet->times_cnt = times_cnt; + + eet->timerfd =timerfd_create (CLOCK_MONOTONIC ,TFD_CLOEXEC ); +if (eet->timerfd ==-1 ) { + err = -errno; +pr_err ("timerfd_create failed: %s\n" ,strerror (errno)); + goto free_eet_times; + } + + eet->pollfd_pos =perf_evlist__add_pollfd (&evlist->core, eet->timerfd,NULL ,POLLIN , flags); +if (eet->pollfd_pos <0 ) { + err = eet->pollfd_pos; + goto close_timerfd; + } + + eet->evlist = evlist; + evlist->eet = eet; + opts->initial_delay = eet->times[0 ].start; + +return 0 ; + +close_timerfd: +close (eet->timerfd); +free_eet_times: +free (eet->times); +free_eet: +free (eet); +return err; +} + +static int event_enable_timer__set_timer (struct event_enable_timer *eet,int ms) +{ +struct itimerspec its = { + .it_value.tv_sec = ms /MSEC_PER_SEC , + .it_value.tv_nsec = (ms %MSEC_PER_SEC ) *NSEC_PER_MSEC , + }; +int err =0 ; + +if (timerfd_settime (eet->timerfd,0 , &its,NULL ) <0 ) { + err = -errno; +pr_err ("timerfd_settime failed: %s\n" ,strerror (errno)); + } +return err; +} + +int event_enable_timer__start (struct event_enable_timer *eet) +{ +int ms; + +if (!eet) +return 0 ; + + ms = eet->times[0 ].end - eet->times[0 ].start; + eet->times_step =1 ; + +return event_enable_timer__set_timer (eet, ms); +} + +int event_enable_timer__process (struct event_enable_timer *eet) +{ +struct pollfd *entries; +short revents; + +if (!eet) +return 0 ; + + entries = eet->evlist->core.pollfd.entries; + revents = entries[eet->pollfd_pos].revents; + entries[eet->pollfd_pos].revents =0 ; + +if (revents &POLLIN ) { +size_t step = eet->times_step; +size_t pos = step /2 ; + +if (step &1 ) { +evlist__disable_non_dummy (eet->evlist); +pr_info (EVLIST_DISABLED_MSG ); +if (pos >= eet->times_cnt -1 ) { +/* Disarm timer */ +event_enable_timer__set_timer (eet,0 ); +return 1 ;/* Stop */ + } + }else { +evlist__enable_non_dummy (eet->evlist); +pr_info (EVLIST_ENABLED_MSG ); + } + + step +=1 ; + pos = step /2 ; + +if (pos < eet->times_cnt) { +int *times = (int *)eet->times;/* Accessing 'times' as array of int */ +int ms = times[step] - times[step -1 ]; + + eet->times_step = step; +return event_enable_timer__set_timer (eet, ms); + } + } + +return 0 ; +} + +void event_enable_timer__exit (struct event_enable_timer **ep) +{ +if (!ep || !*ep) +return ; +free ((*ep)->times); +zfree (ep); +} +struct evsel *evlist__find_evsel (struct evlist *evlist,int idx) {struct evsel *evsel;tools/perf/util/evlist.h +9
-1@@ -48,6 +48,8 @@ enum bkw_mmap_state { BKW_MMAP_EMPTY, }; +struct event_enable_timer ; +struct evlist {struct perf_evlist core;bool enabled;@@ -79,6 +81,7 @@ struct evlist {int ack;/* ack file descriptor for control commands */ int pos;/* index at evlist core object to check signals */ } ctl_fd; +struct event_enable_timer *eet; };struct evsel_str_handler {@@ -426,6 +429,12 @@ int evlist__ctlfd_ack(struct evlist *evlist);#define EVLIST_ENABLED_MSG "Events enabled\n"#define EVLIST_DISABLED_MSG "Events disabled\n" +int evlist__parse_event_enable_time (struct evlist *evlist,struct record_opts *opts, +const char *str,int unset); +int event_enable_timer__start (struct event_enable_timer *eet); +void event_enable_timer__exit (struct event_enable_timer **ep); +int event_enable_timer__process (struct event_enable_timer *eet); +struct evsel *evlist__find_evsel (struct evlist *evlist,int idx);int evlist__scnprintf_evsels (struct evlist *evlist,size_t size,char *bf); --2.25.1
Quoted 9 lines · Adrian Hunter
Hi
This patch set extends perf record -D/--delay option to accept time ranges for when events are enabled, for instance:
perf record -e intel_pt// -D 10-20,30-40to record 10ms to 20ms into the trace and 30ms to 40ms. Refer patch 5 for more details.
Great! I just needed that for something.
-Andi