perf tools: Still more preparation for Intel PT
Adrian Hunter
reviewercover perf tools: Still more preparation for Intel PT
Hi
Here is the next selection of patches. That are based on top of 759e612bf96627b64fcafe4174b3f6f2dedf2c0d perf stat: Use strerror_r instead of strerror
They can also be found here:
http://git.infradead.org/users/ahunter/linux-perf.git/shortlog/refs/heads/for-acme
Adrian Hunter (23):
perf tools: Add a test for tracking with sched_switch
perf scripting: Add 'flush' callback to scripting API
perf tools: Rename machine__get_kernel_start_addr()
perf tools: Add machine__kernel_ip()
perf tools: Let a user specify a PMU event without any config terms perf tools: Let default config be defined for a PMU
perf tools: Add perf_pmu__scan_file()
perf tools: Add id index
perf pmu: Let pmu's with no events show up on perf list
perf session: Add perf_session__deliver_synth_event()
perf tools: Add a thread stack for synthesizing call chains
perf tools: Add facility to export data in database-friendly way
perf tools: Extend Python script interface to export data in a database-friendly way perf tools: Add Python script to export to postgresql
perf tools: Add branch type to db export
perf tools: Add branch_type and in_tx to Python export
perf tools: Enhance the thread stack to output call/return data
perf tools: Add call information to the database export API
perf tools: Add call information to Python export
perf tools: Defer export of comms that were not 'set'
perf tools: Add perf-with-kcore script
perf tools: Build programs to copy 32-bit compatibility VDSOs
perf tools: Add support for 32-bit compatibility VDSOs
Alexander Shishkin (1):
perf tools: Add feature checks to .gitignore
tools/perf/.gitignore | 2 + tools/perf/Makefile.perf | 48 +- tools/perf/builtin-inject.c | 1 + tools/perf/builtin-script.c | 13 + tools/perf/config/Makefile | 25 +- tools/perf/config/Makefile.arch | 8 + tools/perf/config/feature-checks/Makefile | 10 +- tools/perf/config/feature-checks/test-compile.c | 4 + tools/perf/perf-read-vdso.c | 34 + tools/perf/perf-with-kcore.sh | 259 ++++++++ .../scripts/python/bin/export-to-postgresql-record | 8 + .../scripts/python/bin/export-to-postgresql-report | 29 + tools/perf/scripts/python/export-to-postgresql.py | 444 +++++++++++++ tools/perf/tests/builtin-test.c | 4 + tools/perf/tests/pmu.c | 2 +- tools/perf/tests/switch-tracking.c | 572 +++++++++++++++++ tools/perf/tests/tests.h | 1 + tools/perf/util/comm.h | 1 + tools/perf/util/db-export.c | 426 +++++++++++++ tools/perf/util/db-export.h | 107 ++++ tools/perf/util/dso.h | 1 + tools/perf/util/event.c | 7 +- tools/perf/util/event.h | 41 ++ tools/perf/util/evlist.c | 26 +- tools/perf/util/evsel.h | 4 + tools/perf/util/find-vdso-map.c | 30 + tools/perf/util/machine.c | 31 +- tools/perf/util/machine.h | 18 + tools/perf/util/parse-events.c | 13 +- tools/perf/util/parse-events.y | 10 + tools/perf/util/pmu.c | 92 ++- tools/perf/util/pmu.h | 13 +- .../perf/util/scripting-engines/trace-event-perl.c | 6 + .../util/scripting-engines/trace-event-python.c | 399 +++++++++++- tools/perf/util/session.c | 136 ++++ tools/perf/util/session.h | 15 + tools/perf/util/symbol.h | 1 + tools/perf/util/thread-stack.c | 690 +++++++++++++++++++++ tools/perf/util/thread-stack.h | 79 +++ tools/perf/util/thread.c | 3 + tools/perf/util/thread.h | 4 + tools/perf/util/tool.h | 3 +- tools/perf/util/trace-event-scripting.c | 7 + tools/perf/util/trace-event.h | 1 + tools/perf/util/vdso.c | 207 ++++++- tools/perf/util/vdso.h | 4 +- 46 files changed, 3765 insertions(+), 74 deletions(-) create mode 100644 tools/perf/config/feature-checks/test-compile.c create mode 100644 tools/perf/perf-read-vdso.c create mode 100644 tools/perf/perf-with-kcore.sh create mode 100644 tools/perf/scripts/python/bin/export-to-postgresql-record create mode 100644 tools/perf/scripts/python/bin/export-to-postgresql-report create mode 100644 tools/perf/scripts/python/export-to-postgresql.py create mode 100644 tools/perf/tests/switch-tracking.c create mode 100644 tools/perf/util/db-export.c create mode 100644 tools/perf/util/db-export.h create mode 100644 tools/perf/util/find-vdso-map.c create mode 100644 tools/perf/util/thread-stack.c create mode 100644 tools/perf/util/thread-stack.h
Regards
Adrian
Adrian Hunter
reviewerPATCH 1/24 perf tools: Add a test for tracking with sched_switch
Add a test that checks that sched_switch events and tracking events can be recorded for a workload using the evsel->system_wide and evsel->tracking flags (respectively) with other events sometimes enabled or disabled.
--- tools/perf/Makefile.perf | 1 + tools/perf/tests/builtin-test.c | 4 + tools/perf/tests/switch-tracking.c | 572 +++++++++++++++++++++++++++++++++++++ tools/perf/tests/tests.h | 1 + 4 files changed, 578 insertions(+) create mode 100644 tools/perf/tests/switch-tracking.c
tools/perf/Makefile.perf +1 -0
@@ -425,6 +425,7 @@ endif endif LIB_OBJS += $(OUTPUT)tests/mmap-thread-lookup.o LIB_OBJS += $(OUTPUT)tests/thread-mg-share.o +LIB_OBJS += $(OUTPUT)tests/switch-tracking.o BUILTIN_OBJS += $(OUTPUT)builtin-annotate.o BUILTIN_OBJS += $(OUTPUT)builtin-bench.o
tools/perf/tests/builtin-test.c +4 -0
@@ -154,6 +154,10 @@ static struct test { .func = test__hists_cumulate, }, { + .desc = "Test tracking with sched_switch", + .func = test__switch_tracking, + }, + { .func = NULL, }, };
tools/perf/tests/switch-tracking.c new file +572 -0
572 lines not on this page · raw message
tools/perf/tests/tests.h +1 -0
@@ -48,6 +48,7 @@ int test__mmap_thread_lookup(void);int test__thread_mg_share (void );int test__hists_output (void );int test__hists_cumulate (void ); +int test__switch_tracking (void );#if defined(__x86_64__) || defined(__i386__) || defined(__arm__)#ifdef HAVE_DWARF_UNWIND_SUPPORT
Signature · 1 line
1.8.3.2
Arnaldo Carvalho de Melo
maintainerRe: [PATCH 01/24] perf tools: Add a test for tracking with sched_switch
Em Fri, Aug 15, 2014 at 10:08:36PM +0300, Adrian Hunter escreveu:
Quoted 4 lines
Add a test that checks that
sched_switchevents and tracking events can be recorded for a workload using theevsel->system_wideandevsel->trackingflags (respectively) with other events sometimes enabled or disabled.
Really nice exercise! Gives lots of evlist/evsel routines a good test, checking for expected behaviour, really good, thanks.
One thing I noticed was that it uses parse_events() for creating events, a perhaps simpler equivalent would be:
switch_evsel = perf_evsel__newtp("sched", "sched_switch");And then go on, like you did, configuring whatever attribute one wants to have set, like what you get from ":u", and the other things you touched, like:
switch_evsel->system_wide = true; switch_evsel->no_aux_samples = true; switch_evsel->immediate = true;
And when the evsel is all set up, if dealing with evlists, like in this case, just do a:
perf_evlist__add(evlist, switch_evsel);
Looks more natural than using parse_events(evlist, "sched:sched_switch");
To then infer that it must be the last entry at this point to retrieve it using:
switch_evsel = perf_evlist__last(evlist);
But, for the current architecture, that should be just a clarification, not a requirement for this specific test.
Applying!
- Arnaldo
Quoted 628 lines
Signed-off-by Adrian Hunter <[email protected]>--- tools/perf/Makefile.perf | 1 + tools/perf/tests/builtin-test.c | 4 + tools/perf/tests/switch-tracking.c | 572 +++++++++++++++++++++++++++++++++++++ tools/perf/tests/tests.h | 1 + 4 files changed, 578 insertions(+) create mode 100644 tools/perf/tests/switch-tracking.ctools/perf/Makefile.perf +1
-0@@ -425,6 +425,7 @@ endif endif LIB_OBJS += $(OUTPUT)tests/mmap-thread-lookup.o LIB_OBJS += $(OUTPUT)tests/thread-mg-share.o +LIB_OBJS += $(OUTPUT)tests/switch-tracking.o BUILTIN_OBJS += $(OUTPUT)builtin-annotate.o BUILTIN_OBJS += $(OUTPUT)builtin-bench.otools/perf/tests/builtin-test.c +4
-0@@ -154,6 +154,10 @@ static struct test { .func = test__hists_cumulate, }, { + .desc = "Test tracking with sched_switch", + .func = test__switch_tracking, + }, + { .func = NULL, }, };tools/perf/tests/switch-tracking.c new file +572
-0@@ -0,0 +1,572 @@ +#include <sys/time.h> +#include <sys/prctl.h> +#include <time.h> +#include <stdlib.h> + +#include "parse-events.h" +#include "evlist.h" +#include "evsel.h" +#include "thread_map.h" +#include "cpumap.h" +#include "tests.h" + +static int spin_sleep (void ) +{ +struct timeval start, now, diff, maxtime; +struct timespec ts; +int err, i; + + maxtime.tv_sec =0 ; + maxtime.tv_usec =50000 ; + + err =gettimeofday (&start,NULL ); +if (err) +return err; + +/* Spin for 50ms */ +while (1 ) { +for (i =0 ; i <1000 ; i++) +barrier (); + + err =gettimeofday (&now,NULL ); +if (err) +return err; + +timersub (&now, &start, &diff); + if (timercmp(&diff, &maxtime, > /* For checkpatch */)) +break ; + } + + ts.tv_nsec =50 *1000 *1000 ; + ts.tv_sec =0 ; + +/* Sleep for 50ms */ + err =nanosleep (&ts,NULL ); +if (err ==EINTR ) + err =0 ; + +return err; +} + +struct switch_tracking { +struct perf_evsel *switch_evsel; +struct perf_evsel *cycles_evsel; +pid_t *tids; +int nr_tids; +int comm_seen[4 ]; +int cycles_before_comm_1; +int cycles_between_comm_2_and_comm_3; +int cycles_after_comm_4; +}; + +static int check_comm (struct switch_tracking *switch_tracking, +union perf_event *event,const char *comm,int nr) +{ +if (event->header.type ==PERF_RECORD_COMM && + (pid_t )event->comm.pid ==getpid () && + (pid_t )event->comm.tid ==getpid () && +strcmp (event->comm.comm, comm) ==0 ) { +if (switch_tracking->comm_seen[nr]) { +pr_debug ("Duplicate comm event\n" ); +return -1 ; + } + switch_tracking->comm_seen[nr] =1 ; +pr_debug3 ("comm event: %s nr: %d\n" , event->comm.comm, nr); +return 1 ; + } +return 0 ; +} + +static int check_cpu (struct switch_tracking *switch_tracking,int cpu) +{ +int i, nr = cpu +1 ; + +if (cpu <0 ) +return -1 ; + +if (!switch_tracking->tids) { + switch_tracking->tids =calloc (nr,sizeof (pid_t)); +if (!switch_tracking->tids) +return -1 ; +for (i =0 ; i < nr; i++) + switch_tracking->tids[i] =-1 ; + switch_tracking->nr_tids = nr; +return 0 ; + } + +if (cpu >= switch_tracking->nr_tids) { +void *addr; + + addr =realloc (switch_tracking->tids, nr *sizeof (pid_t)); +if (!addr) +return -1 ; + switch_tracking->tids = addr; +for (i = switch_tracking->nr_tids; i < nr; i++) + switch_tracking->tids[i] =-1 ; + switch_tracking->nr_tids = nr; +return 0 ; + } + +return 0 ; +} + +static int process_sample_event (struct perf_evlist *evlist, +union perf_event *event, +struct switch_tracking *switch_tracking) +{ +struct perf_sample sample; +struct perf_evsel *evsel; +pid_t next_tid, prev_tid; +int cpu, err; + +if (perf_evlist__parse_sample (evlist, event, &sample)) { +pr_debug ("perf_evlist__parse_sample failed\n" ); +return -1 ; + } + + evsel =perf_evlist__id2evsel (evlist, sample.id); +if (evsel == switch_tracking->switch_evsel) { + next_tid =perf_evsel__intval (evsel, &sample,"next_pid" ); + prev_tid =perf_evsel__intval (evsel, &sample,"prev_pid" ); + cpu = sample.cpu; +pr_debug3 ("sched_switch: cpu: %d prev_tid %d next_tid %d\n" , + cpu, prev_tid, next_tid); + err =check_cpu (switch_tracking, cpu); +if (err) +return err; +/* +* Check for no missing sched_switch events i.e. that the +* evsel->system_wide flag has worked. +*/ +if (switch_tracking->tids[cpu] !=-1 && + switch_tracking->tids[cpu] != prev_tid) { +pr_debug ("Missing sched_switch events\n" ); +return -1 ; + } + switch_tracking->tids[cpu] = next_tid; + } + +if (evsel == switch_tracking->cycles_evsel) { +pr_debug3 ("cycles event\n" ); +if (!switch_tracking->comm_seen[0 ]) + switch_tracking->cycles_before_comm_1 =1 ; +if (switch_tracking->comm_seen[1 ] && + !switch_tracking->comm_seen[2 ]) + switch_tracking->cycles_between_comm_2_and_comm_3 =1 ; +if (switch_tracking->comm_seen[3 ]) + switch_tracking->cycles_after_comm_4 =1 ; + } + +return 0 ; +} + +static int process_event (struct perf_evlist *evlist,union perf_event *event, +struct switch_tracking *switch_tracking) +{ +if (event->header.type ==PERF_RECORD_SAMPLE ) +return process_sample_event (evlist, event, switch_tracking); + +if (event->header.type ==PERF_RECORD_COMM ) { +int err, done =0 ; + + err =check_comm (switch_tracking, event,"Test COMM 1" ,0 ); +if (err <0 ) +return -1 ; + done += err; + err =check_comm (switch_tracking, event,"Test COMM 2" ,1 ); +if (err <0 ) +return -1 ; + done += err; + err =check_comm (switch_tracking, event,"Test COMM 3" ,2 ); +if (err <0 ) +return -1 ; + done += err; + err =check_comm (switch_tracking, event,"Test COMM 4" ,3 ); +if (err <0 ) +return -1 ; + done += err; +if (done !=1 ) { +pr_debug ("Unexpected comm event\n" ); +return -1 ; + } + } + +return 0 ; +} + +struct event_node { +struct list_head list; +union perf_event *event; +u64 event_time; +}; + +static int add_event (struct perf_evlist *evlist,struct list_head *events, +union perf_event *event) +{ +struct perf_sample sample; +struct event_node *node; + + node =malloc (sizeof (struct event_node )); +if (!node) { +pr_debug ("malloc failed\n" ); +return -1 ; + } + node->event = event; +list_add (&node->list, events); + +if (perf_evlist__parse_sample (evlist, event, &sample)) { +pr_debug ("perf_evlist__parse_sample failed\n" ); +return -1 ; + } + +if (!sample.time) { +pr_debug ("event with no time\n" ); +return -1 ; + } + + node->event_time = sample.time; + +return 0 ; +} + +static void free_event_nodes (struct list_head *events) +{ +struct event_node *node; + +while (!list_empty (events)) { + node = list_entry(events->next, struct event_node, list); +list_del (&node->list); +free (node); + } +} + +static int compar (const void *a,const void *b) +{ +const struct event_node *nodea = a; +const struct event_node *nodeb = b; +s64 cmp = nodea->event_time - nodeb->event_time; + +return cmp; +} + +static int process_events (struct perf_evlist *evlist, +struct switch_tracking *switch_tracking) +{ +union perf_event *event; +unsigned pos, cnt =0 ; +LIST_HEAD (events); +struct event_node *events_array, *node; +int i, ret; + +for (i =0 ; i < evlist->nr_mmaps; i++) { +while ((event =perf_evlist__mmap_read (evlist, i)) !=NULL ) { + cnt +=1 ; + ret =add_event (evlist, &events, event); +perf_evlist__mmap_consume (evlist, i); +if (ret <0 ) + goto out_free_nodes; + } + } + + events_array =calloc (cnt,sizeof (struct event_node )); +if (!events_array) { +pr_debug ("calloc failed\n" ); + ret =-1 ; + goto out_free_nodes; + } + + pos =0 ; +list_for_each_entry (node, &events, list) + events_array[pos++] = *node; + +qsort (events_array, cnt,sizeof (struct event_node ), compar); + +for (pos =0 ; pos < cnt; pos++) { + ret =process_event (evlist, events_array[pos].event, + switch_tracking); +if (ret <0 ) + goto out_free; + } + + ret =0 ; +out_free: +pr_debug ("%u events recorded\n" , cnt); +free (events_array); +out_free_nodes: +free_event_nodes (&events); +return ret; +} + +/** +* test__switch_tracking - test using sched_switch and tracking events. +* +* This function implements a test that checks that sched_switch events and +* tracking events can be recorded for a workload (current process) using the +* evsel->system_wide and evsel->tracking flags (respectively) with other events +* sometimes enabled or disabled. +*/ +int test__switch_tracking (void ) +{ +const char *sched_switch ="sched:sched_switch" ; +struct switch_tracking switch_tracking = {0 }; +struct record_opts opts = { + .mmap_pages =UINT_MAX , + .user_freq =UINT_MAX , + .user_interval =ULLONG_MAX , + .freq =4000 , + .target = { + .uses_mmap = true, + }, + }; +struct thread_map *threads =NULL ; +struct cpu_map *cpus =NULL ; +struct perf_evlist *evlist =NULL ; +struct perf_evsel *evsel, *cpu_clocks_evsel, *cycles_evsel; +struct perf_evsel *switch_evsel, *tracking_evsel; +const char *comm; +int err =-1 ; + + threads =thread_map__new (-1 ,getpid (),UINT_MAX ); +if (!threads) { +pr_debug ("thread_map__new failed!\n" ); + goto out_err; + } + + cpus =cpu_map__new (NULL ); +if (!cpus) { +pr_debug ("cpu_map__new failed!\n" ); + goto out_err; + } + + evlist =perf_evlist__new (); +if (!evlist) { +pr_debug ("perf_evlist__new failed!\n" ); + goto out_err; + } + +perf_evlist__set_maps (evlist, cpus, threads); + +/* First event */ + err =parse_events (evlist,"cpu-clock:u" ); +if (err) { +pr_debug ("Failed to parse event dummy:u\n" ); + goto out_err; + } + + cpu_clocks_evsel =perf_evlist__last (evlist); + +/* Second event */ + err =parse_events (evlist,"cycles:u" ); +if (err) { +pr_debug ("Failed to parse event cycles:u\n" ); + goto out_err; + } + + cycles_evsel =perf_evlist__last (evlist); + +/* Third event */ +if (!perf_evlist__can_select_event (evlist, sched_switch)) { +fprintf (stderr," (no sched_switch)" ); + err =0 ; + goto out; + } + + err =parse_events (evlist, sched_switch); +if (err) { +pr_debug ("Failed to parse event %s\n" , sched_switch); + goto out_err; + } + + switch_evsel =perf_evlist__last (evlist); + +perf_evsel__set_sample_bit (switch_evsel,CPU ); +perf_evsel__set_sample_bit (switch_evsel,TIME ); + + switch_evsel->system_wide = true; + switch_evsel->no_aux_samples = true; + switch_evsel->immediate = true; + +/* Test moving an event to the front */ +if (cycles_evsel ==perf_evlist__first (evlist)) { +pr_debug ("cycles event already at front" ); + goto out_err; + } +perf_evlist__to_front (evlist, cycles_evsel); +if (cycles_evsel !=perf_evlist__first (evlist)) { +pr_debug ("Failed to move cycles event to front" ); + goto out_err; + } + +perf_evsel__set_sample_bit (cycles_evsel,CPU ); +perf_evsel__set_sample_bit (cycles_evsel,TIME ); + +/* Fourth event */ + err =parse_events (evlist,"dummy:u" ); +if (err) { +pr_debug ("Failed to parse event dummy:u\n" ); + goto out_err; + } + + tracking_evsel =perf_evlist__last (evlist); + +perf_evlist__set_tracking_event (evlist, tracking_evsel); + + tracking_evsel->attr.freq =0 ; + tracking_evsel->attr.sample_period =1 ; + +perf_evsel__set_sample_bit (tracking_evsel,TIME ); + +/* Config events */ +perf_evlist__config (evlist, &opts); + +/* Check moved event is still at the front */ +if (cycles_evsel !=perf_evlist__first (evlist)) { +pr_debug ("Front event no longer at front" ); + goto out_err; + } + +/* Check tracking event is tracking */ +if (!tracking_evsel->attr.mmap || !tracking_evsel->attr.comm) { +pr_debug ("Tracking event not tracking\n" ); + goto out_err; + } + +/* Check non-tracking events are not tracking */ +evlist__for_each (evlist, evsel) { +if (evsel != tracking_evsel) { +if (evsel->attr.mmap || evsel->attr.comm) { +pr_debug ("Non-tracking event is tracking\n" ); + goto out_err; + } + } + } + +if (perf_evlist__open (evlist) <0 ) { +fprintf (stderr," (not supported)" ); + err =0 ; + goto out; + } + + err =perf_evlist__mmap (evlist,UINT_MAX , false); +if (err) { +pr_debug ("perf_evlist__mmap failed!\n" ); + goto out_err; + } + +perf_evlist__enable (evlist); + + err =perf_evlist__disable_event (evlist, cpu_clocks_evsel); +if (err) { +pr_debug ("perf_evlist__disable_event failed!\n" ); + goto out_err; + } + + err =spin_sleep (); +if (err) { +pr_debug ("spin_sleep failed!\n" ); + goto out_err; + } + + comm ="Test COMM 1" ; + err =prctl (PR_SET_NAME , (unsigned long )comm,0 ,0 ,0 ); +if (err) { +pr_debug ("PR_SET_NAME failed!\n" ); + goto out_err; + } + + err =perf_evlist__disable_event (evlist, cycles_evsel); +if (err) { +pr_debug ("perf_evlist__disable_event failed!\n" ); + goto out_err; + } + + comm ="Test COMM 2" ; + err =prctl (PR_SET_NAME , (unsigned long )comm,0 ,0 ,0 ); +if (err) { +pr_debug ("PR_SET_NAME failed!\n" ); + goto out_err; + } + + err =spin_sleep (); +if (err) { +pr_debug ("spin_sleep failed!\n" ); + goto out_err; + } + + comm ="Test COMM 3" ; + err =prctl (PR_SET_NAME , (unsigned long )comm,0 ,0 ,0 ); +if (err) { +pr_debug ("PR_SET_NAME failed!\n" ); + goto out_err; + } + + err =perf_evlist__enable_event (evlist, cycles_evsel); +if (err) { +pr_debug ("perf_evlist__disable_event failed!\n" ); + goto out_err; + } + + comm ="Test COMM 4" ; + err =prctl (PR_SET_NAME , (unsigned long )comm,0 ,0 ,0 ); +if (err) { +pr_debug ("PR_SET_NAME failed!\n" ); + goto out_err; + } + + err =spin_sleep (); +if (err) { +pr_debug ("spin_sleep failed!\n" ); + goto out_err; + } + +perf_evlist__disable (evlist); + + switch_tracking.switch_evsel = switch_evsel; + switch_tracking.cycles_evsel = cycles_evsel; + + err =process_events (evlist, &switch_tracking); + +zfree (&switch_tracking.tids); + +if (err) + goto out_err; + +/* Check all 4 comm events were seen i.e. that evsel->tracking works */ +if (!switch_tracking.comm_seen[0 ] || !switch_tracking.comm_seen[1 ] || + !switch_tracking.comm_seen[2 ] || !switch_tracking.comm_seen[3 ]) { +pr_debug ("Missing comm events\n" ); + goto out_err; + } + +/* Check cycles event got enabled */ +if (!switch_tracking.cycles_before_comm_1) { +pr_debug ("Missing cycles events\n" ); + goto out_err; + } + +/* Check cycles event got disabled */ +if (switch_tracking.cycles_between_comm_2_and_comm_3) { +pr_debug ("cycles events even though event was disabled\n" ); + goto out_err; + } + +/* Check cycles event got enabled again */ +if (!switch_tracking.cycles_after_comm_4) { +pr_debug ("Missing cycles events\n" ); + goto out_err; + } +out: +if (evlist) { +perf_evlist__disable (evlist); +perf_evlist__delete (evlist); + }else { +cpu_map__delete (cpus); +thread_map__delete (threads); + } + +return err; + +out_err: + err =-1 ; + goto out; +}tools/perf/tests/tests.h +1
-0@@ -48,6 +48,7 @@ int test__mmap_thread_lookup(void);int test__thread_mg_share (void );int test__hists_output (void );int test__hists_cumulate (void ); +int test__switch_tracking (void );#if defined(__x86_64__) || defined(__i386__) || defined(__arm__)#ifdef HAVE_DWARF_UNWIND_SUPPORT Signature · 1 line
1.8.3.2
Adrian Hunter
reviewerRe: [PATCH 01/24] perf tools: Add a test for tracking with sched_switch
Quoted 29 lines · Arnaldo Carvalho de Melo
Em Fri, Aug 15, 2014 at 10:08:36PM +0300, Adrian Hunter escreveu:
Quoted 4 lines
Add a test that checks that
sched_switchevents and tracking events can be recorded for a workload using theevsel->system_wideandevsel->trackingflags (respectively) with other events sometimes enabled or disabled.Really nice exercise! Gives lots of evlist/evsel routines a good test, checking for expected behaviour, really good, thanks.
One thing I noticed was that it uses
parse_events()for creating events, a perhaps simpler equivalent would be:switch_evsel = perf_evsel__newtp("sched", "sched_switch");And then go on, like you did, configuring whatever attribute one wants to have set, like what you get from ":u", and the other things you touched, like:
switch_evsel->system_wide = true; switch_evsel->no_aux_samples = true; switch_evsel->immediate = true;And when the evsel is all set up, if dealing with evlists, like in this case, just do a:
perf_evlist__add(evlist, switch_evsel);Looks more natural than using
parse_events(evlist, "sched:sched_switch");
Quoted 7 lines
To then infer that it must be the last entry at this point to retrieve it using:
switch_evsel = perf_evlist__last(evlist);But, for the current architecture, that should be just a clarification, not a requirement for this specific test.
I will keep that in mind.
Quoted 1 line
Applying!
Thank you!
AFAICT the other 20 patches still apply cleanly to your tree, so I am not going to send any more for now.
Arnaldo Carvalho de Melo
maintainerRe: [PATCH 01/24] perf tools: Add a test for tracking with sched_switch
Em Fri, Aug 29, 2014 at 04:52:16PM +0300, Adrian Hunter escreveu:
Quoted 12 lines · Arnaldo Carvalho de Melo
Quoted 2 lines
But, for the current architecture, that should be just a clarification, not a requirement for this specific test.
I will keep that in mind.
Quoted 1 line
Applying!
Thank you!
AFAICT the other 20 patches still apply cleanly to your tree, so I am not going to send any more for now.
Ok, I'll try to process more of those, the db interface probably, after I try using it for some simple tests.
About the PMU ones, that I would like to hear from other people dealing with that area, like Jiri or even PeterZ, that IIRC discussed sysfs interface aspects in the past, Jiri, Peter?
- Arnaldo
Re: [PATCH 01/24] perf tools: Add a test for tracking with sched_switch
Quoted 14 lines · Arnaldo Carvalho de Melo
Em Fri, Aug 29, 2014 at 04:52:16PM +0300, Adrian Hunter escreveu:
Quoted 12 lines · Arnaldo Carvalho de Melo
Quoted 2 lines
But, for the current architecture, that should be just a clarification, not a requirement for this specific test.
I will keep that in mind.
Quoted 1 line
Applying!
Thank you!
AFAICT the other 20 patches still apply cleanly to your tree, so I am not going to send any more for now.
Quoted 6 lines
Ok, I'll try to process more of those, the db interface probably, after I try using it for some simple tests.
About the PMU ones, that I would like to hear from other people dealing with that area, like Jiri or even PeterZ, that IIRC discussed sysfs interface aspects in the past, Jiri, Peter?
I reviewed those, but I remember not getting an answer for my question about that
http://marc.info/?t=140533254600001&r=1&w=2
jirka
Adrian Hunter
reviewerRe: [PATCH 01/24] perf tools: Add a test for tracking with sched_switch
Quoted 18 lines · Jiri Olsa
Quoted 17 lines · Arnaldo Carvalho de Melo
Em Fri, Aug 29, 2014 at 04:52:16PM +0300, Adrian Hunter escreveu:
Quoted 12 lines · Arnaldo Carvalho de Melo
Quoted 2 lines
But, for the current architecture, that should be just a clarification, not a requirement for this specific test.
I will keep that in mind.
Quoted 1 line
Applying!
Thank you!
AFAICT the other 20 patches still apply cleanly to your tree, so I am not going to send any more for now.
Ok, I'll try to process more of those, the db interface probably, after I try using it for some simple tests.
Quoted 7 lines
Quoted 3 lines
About the PMU ones, that I would like to hear from other people dealing with that area, like Jiri or even PeterZ, that IIRC discussed sysfs interface aspects in the past, Jiri, Peter?
I reviewed those, but I remember not getting an answer for my question about that
http://marc.info/?t=140533254600001&r=1&w=2
I guess I didn't understand what you meant. I will reply to that thread.
Arnaldo Carvalho de Melo
maintainerRe: [PATCH 01/24] perf tools: Add a test for tracking with sched_switch
In reply to Adrian HunterEm Fri, Aug 15, 2014 at 10:08:36PM +0300, Adrian Hunter escreveu:
tools/perf/tests/switch-tracking.c line 312
Quoted 12 lines
+ *
test__switch_tracking- test usingsched_switchand tracking events.
+ *
+ * This function implements a test that checks thatsched_switchevents and
+ * tracking events can be recorded for a workload (current process) using the
+ *evsel->system_wideandevsel->trackingflags (respectively) with other events
+ * sometimes enabled or disabled.
+ */tools/perf/tests/switch-tracking.c new file +5
-0+int test__switch_tracking (void ) +{ +const char *sched_switch ="sched:sched_switch" ; +struct switch_tracking switch_tracking = {0 }; +struct record_opts opts = {
308 int test__switch_tracking(void)
309 {
310 const char *sched_switch = "sched:sched_switch";
311 struct switch_tracking switch_tracking = {0};
312 struct record_opts opts = {
313 .mmap_pages = UINT_MAX,
314 .user_freq = UINT_MAX,
315 .user_interval = ULLONG_MAX,
316 .freq = 4000,
317 .target = {
318 .uses_mmap = true,319 },
320 };
321 struct thread_map *threads = NULL;
Fails to build on f14, still building on other distros, will fix after lunch, no need to send any fix, just FYI:
CC /tmp/build/perf/arch/x86/util/header.o
cc1: warnings being treated as errorstests/switch-tracking.c: In function ‘test__switch_tracking’:tests/switch-tracking.c:311:9: error: missing initializer tests/switch-tracking.c:311:9: error: (near initialization for ‘switch_tracking.cycles_evsel’) make[1]: *** [/tmp/build/perf/tests/switch-tracking.o] Error 1
make[1]: *** Waiting for unfinished jobs....
make: *** [install] Error 2
make: Leaving directory `/home/acme/git/linux/tools/perf'
real 0m19.399s
user 0m53.823s
sys 0m7.346s
[acme@fedora14 linux]$ fg
-bash: fg: current: no such job
[acme@fedora14
tip-bot for Adrian Hunter
merge bot[tip:perf/core] perf tests: Add a test for tracking with sched_switch
In reply to Adrian Hunterperf tests: Add a test for tracking with sched_switch
Commit-ID: d44bc558297222d9a621fff2eca3f880e91c49f7 Gitweb: http://git.kernel.org/tip/d44bc558297222d9a621fff2eca3f880e91c49f7 Author: Adrian Hunter <[email protected]> AuthorDate: Fri, 15 Aug 2014 22:08:36 +0300 Committer: Arnaldo Carvalho de Melo <[email protected]> CommitDate: Fri, 22 Aug 2014 13:10:35 -0300
perf tests: Add a test for tracking with sched_switch
Add a test that checks that sched_switch events and tracking events can be recorded for a workload using the evsel->system_wide and evsel->tracking flags (respectively) with other events sometimes enabled or disabled.
--- tools/perf/Makefile.perf | 1 + tools/perf/tests/builtin-test.c | 4 + tools/perf/tests/switch-tracking.c | 572 +++++++++++++++++++++++++++++++++++++ tools/perf/tests/tests.h | 1 + 4 files changed, 578 insertions(+)
tools/perf/Makefile.perf +1 -0
@@ -425,6 +425,7 @@ endif endif LIB_OBJS += $(OUTPUT)tests/mmap-thread-lookup.o LIB_OBJS += $(OUTPUT)tests/thread-mg-share.o +LIB_OBJS += $(OUTPUT)tests/switch-tracking.o BUILTIN_OBJS += $(OUTPUT)builtin-annotate.o BUILTIN_OBJS += $(OUTPUT)builtin-bench.o
tools/perf/tests/builtin-test.c +4 -0
@@ -154,6 +154,10 @@ static struct test { .func = test__hists_cumulate, }, { + .desc = "Test tracking with sched_switch", + .func = test__switch_tracking, + }, + { .func = NULL, }, };
tools/perf/tests/switch-tracking.c new file +572 -0
572 lines not on this page · raw message
tools/perf/tests/tests.h +1 -0
@@ -48,6 +48,7 @@ int test__mmap_thread_lookup(void);int test__thread_mg_share (void );int test__hists_output (void );int test__hists_cumulate (void ); +int test__switch_tracking (void );#if defined(__x86_64__) || defined(__i386__) || defined(__arm__)#ifdef HAVE_DWARF_UNWIND_SUPPORT
Adrian Hunter
reviewerPATCH 2/24 perf scripting: Add 'flush' callback to scripting API
In reply to Adrian HunterIn order to defer some output via the scripting API, there needs to be a callback after session processing but before the session is deleted.
--- tools/perf/builtin-script.c | 13 +++++++++++++ tools/perf/util/scripting-engines/trace-event-perl.c | 6 ++++++ tools/perf/util/scripting-engines/trace-event-python.c | 6 ++++++ tools/perf/util/trace-event-scripting.c | 7 +++++++ tools/perf/util/trace-event.h | 1 + 5 files changed, 33 insertions(+)
tools/perf/builtin-script.c +13 -0
@@ -485,6 +485,11 @@ static int default_start_script(const char *script __maybe_unused,return 0 ; } +static int default_flush_script (void ) +{ +return 0 ; +} +static int default_stop_script (void ) {return 0 ;
@@ -498,6 +503,7 @@ static int default_generate_script(struct pevent *pevent __maybe_unused,static struct scripting_ops default_scripting_ops = { .start_script = default_start_script, + .flush_script = default_flush_script, .stop_script = default_stop_script, .process_event = process_event, .generate_script = default_generate_script,
@@ -513,6 +519,11 @@ static void setup_scripting(void) scripting_ops = &default_scripting_ops; } +static int flush_scripting (void ) +{ +return scripting_ops->flush_script (); +} +static int cleanup_scripting (void ) {pr_debug ("\nperf script stopped\n" );
@@ -1813,6 +1824,8 @@ int cmd_script(int argc, const char **argv, const char *prefix __maybe_unused) err =__cmd_script (&script); +flush_scripting (); + out_delete:perf_session__delete (session);
tools/perf/util/scripting-engines/trace-event-perl.c +6 -0
@@ -432,6 +432,11 @@ error:return err; } +static int perl_flush_script (void ) +{ +return 0 ; +} +/* * Stop trace script */
@@ -633,6 +638,7 @@ static int perl_generate_script(struct pevent *pevent, const char *outfile)struct scripting_ops perl_scripting_ops = { .name ="Perl" , .start_script = perl_start_script, + .flush_script = perl_flush_script, .stop_script = perl_stop_script, .process_event = perl_process_event, .generate_script = perl_generate_script,
tools/perf/util/scripting-engines/trace-event-python.c +6 -0
@@ -639,6 +639,11 @@ error:return err; } +static int python_flush_script (void ) +{ +return 0 ; +} +/* * Stop trace script */
@@ -823,6 +828,7 @@ static int python_generate_script(struct pevent *pevent, const char *outfile)struct scripting_ops python_scripting_ops = { .name ="Python" , .start_script = python_start_script, + .flush_script = python_flush_script, .stop_script = python_stop_script, .process_event = python_process_event, .generate_script = python_generate_script,
tools/perf/util/trace-event-scripting.c +7 -0
@@ -30,6 +30,11 @@struct scripting_context *scripting_context; +static int flush_script_unsupported (void ) +{ +return 0 ; +} +static int stop_script_unsupported (void ) {return 0 ;
@@ -74,6 +79,7 @@ static int python_generate_script_unsupported(struct pevent *peventstruct scripting_ops python_scripting_unsupported_ops = { .name ="Python" , .start_script = python_start_script_unsupported, + .flush_script = flush_script_unsupported, .stop_script = stop_script_unsupported, .process_event = process_event_unsupported, .generate_script = python_generate_script_unsupported,
@@ -137,6 +143,7 @@ static int perl_generate_script_unsupported(struct pevent *peventstruct scripting_ops perl_scripting_unsupported_ops = { .name ="Perl" , .start_script = perl_start_script_unsupported, + .flush_script = flush_script_unsupported, .stop_script = stop_script_unsupported, .process_event = process_event_unsupported, .generate_script = perl_generate_script_unsupported,
tools/perf/util/trace-event.h +1 -0
@@ -64,6 +64,7 @@ struct perf_session;struct scripting_ops {const char *name;int (*start_script) (const char *script,int argc,const char **argv); +int (*flush_script) (void );int (*stop_script) (void ); void (*process_event) (union perf_event *event, struct perf_sample *sample,
Signature · 1 line
1.8.3.2
tip-bot for Adrian Hunter
merge bot[tip:perf/core] perf scripting: Add 'flush' callback to scripting API
perf scripting: Add 'flush' callback to scripting API
Commit-ID: d445dd2a78eed884adf3b3426b078fe69d2516d8 Gitweb: http://git.kernel.org/tip/d445dd2a78eed884adf3b3426b078fe69d2516d8 Author: Adrian Hunter <[email protected]> AuthorDate: Fri, 15 Aug 2014 22:08:37 +0300 Committer: Arnaldo Carvalho de Melo <[email protected]> CommitDate: Fri, 22 Aug 2014 13:12:11 -0300
perf scripting: Add 'flush' callback to scripting API
In order to defer some output via the scripting API, there needs to be a callback after session processing but before the session is deleted.
--- tools/perf/builtin-script.c | 13 +++++++++++++ tools/perf/util/scripting-engines/trace-event-perl.c | 6 ++++++ tools/perf/util/scripting-engines/trace-event-python.c | 6 ++++++ tools/perf/util/trace-event-scripting.c | 7 +++++++ tools/perf/util/trace-event.h | 1 + 5 files changed, 33 insertions(+)
tools/perf/builtin-script.c +13 -0
@@ -485,6 +485,11 @@ static int default_start_script(const char *script __maybe_unused,return 0 ; } +static int default_flush_script (void ) +{ +return 0 ; +} +static int default_stop_script (void ) {return 0 ;
@@ -498,6 +503,7 @@ static int default_generate_script(struct pevent *pevent __maybe_unused,static struct scripting_ops default_scripting_ops = { .start_script = default_start_script, + .flush_script = default_flush_script, .stop_script = default_stop_script, .process_event = process_event, .generate_script = default_generate_script,
@@ -513,6 +519,11 @@ static void setup_scripting(void) scripting_ops = &default_scripting_ops; } +static int flush_scripting (void ) +{ +return scripting_ops->flush_script (); +} +static int cleanup_scripting (void ) {pr_debug ("\nperf script stopped\n" );
@@ -1813,6 +1824,8 @@ int cmd_script(int argc, const char **argv, const char *prefix __maybe_unused) err =__cmd_script (&script); +flush_scripting (); + out_delete:perf_session__delete (session);
tools/perf/util/scripting-engines/trace-event-perl.c +6 -0
@@ -432,6 +432,11 @@ error:return err; } +static int perl_flush_script (void ) +{ +return 0 ; +} +/* * Stop trace script */
@@ -633,6 +638,7 @@ static int perl_generate_script(struct pevent *pevent, const char *outfile)struct scripting_ops perl_scripting_ops = { .name ="Perl" , .start_script = perl_start_script, + .flush_script = perl_flush_script, .stop_script = perl_stop_script, .process_event = perl_process_event, .generate_script = perl_generate_script,
tools/perf/util/scripting-engines/trace-event-python.c +6 -0
@@ -639,6 +639,11 @@ error:return err; } +static int python_flush_script (void ) +{ +return 0 ; +} +/* * Stop trace script */
@@ -823,6 +828,7 @@ static int python_generate_script(struct pevent *pevent, const char *outfile)struct scripting_ops python_scripting_ops = { .name ="Python" , .start_script = python_start_script, + .flush_script = python_flush_script, .stop_script = python_stop_script, .process_event = python_process_event, .generate_script = python_generate_script,
tools/perf/util/trace-event-scripting.c +7 -0
@@ -30,6 +30,11 @@struct scripting_context *scripting_context; +static int flush_script_unsupported (void ) +{ +return 0 ; +} +static int stop_script_unsupported (void ) {return 0 ;
@@ -74,6 +79,7 @@ static int python_generate_script_unsupported(struct pevent *peventstruct scripting_ops python_scripting_unsupported_ops = { .name ="Python" , .start_script = python_start_script_unsupported, + .flush_script = flush_script_unsupported, .stop_script = stop_script_unsupported, .process_event = process_event_unsupported, .generate_script = python_generate_script_unsupported,
@@ -137,6 +143,7 @@ static int perl_generate_script_unsupported(struct pevent *peventstruct scripting_ops perl_scripting_unsupported_ops = { .name ="Perl" , .start_script = perl_start_script_unsupported, + .flush_script = flush_script_unsupported, .stop_script = stop_script_unsupported, .process_event = process_event_unsupported, .generate_script = perl_generate_script_unsupported,
tools/perf/util/trace-event.h +1 -0
@@ -64,6 +64,7 @@ struct perf_session;struct scripting_ops {const char *name;int (*start_script) (const char *script,int argc,const char **argv); +int (*flush_script) (void );int (*stop_script) (void ); void (*process_event) (union perf_event *event, struct perf_sample *sample,
Adrian Hunter
reviewerPATCH 3/24 perf tools: Rename machine__get_kernel_start_addr()
In reply to Adrian HunterRename machine__get_kernel_start_addr() tomachine__get_running_kernel_start() so that
a new function, with a similar name to the
original name, can be added that gets the
kernel start address from the kernel map.
--- tools/perf/util/machine.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
tools/perf/util/machine.c +4 -4
@@ -593,8 +593,8 @@ const char *ref_reloc_sym_names[] = {"_text", "_stext", NULL};* Returns the name of the start symbol in *symbol_name. Pass in NULL as * symbol_name if it's not that important. */ -static u64 machine__get_kernel_start_addr (struct machine *machine, -const char **symbol_name) +static u64 machine__get_running_kernel_start (struct machine *machine, +const char **symbol_name) {char filename[PATH_MAX ];int i;
@@ -621,7 +621,7 @@ static u64 machine__get_kernel_start_addr(struct machine *machine,int __machine__create_kernel_maps (struct machine *machine,struct dso *kernel) {enum map_type type; -u64 start =machine__get_kernel_start_addr (machine,NULL ); +u64 start =machine__get_running_kernel_start (machine,NULL );for (type =0 ; type <MAP__NR_TYPES ; ++type) {struct kmap *kmap;
@@ -940,7 +940,7 @@ int machine__create_kernel_maps(struct machine *machine) {struct dso *kernel =machine__get_kernel (machine);const char *name; -u64 addr =machine__get_kernel_start_addr (machine, &name); +u64 addr =machine__get_running_kernel_start (machine, &name);if (!addr)return -1 ;
Signature · 1 line
1.8.3.2
tip-bot for Adrian Hunter
merge bot[tip:perf/core] perf machine: Rename machine__get_kernel_start_addr() method
perf machine: Rename machine__get_kernel_start_addr() method
Commit-ID: 4b99375b38fa137f501cfa60b70e3f0a9da39c93 Gitweb: http://git.kernel.org/tip/4b99375b38fa137f501cfa60b70e3f0a9da39c93 Author: Adrian Hunter <[email protected]> AuthorDate: Fri, 15 Aug 2014 22:08:38 +0300 Committer: Arnaldo Carvalho de Melo <[email protected]> CommitDate: Fri, 22 Aug 2014 13:12:11 -0300
perf machine: Rename machine__get_kernel_start_addr() method
Rename machine__get_kernel_start_addr() to machine__get_running_kernel_start() so that a new function, with a similar name to the original name, can be added that gets the kernel start address from the kernel map.
--- tools/perf/util/machine.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
tools/perf/util/machine.c +4 -4
@@ -593,8 +593,8 @@ const char *ref_reloc_sym_names[] = {"_text", "_stext", NULL};* Returns the name of the start symbol in *symbol_name. Pass in NULL as * symbol_name if it's not that important. */ -static u64 machine__get_kernel_start_addr (struct machine *machine, -const char **symbol_name) +static u64 machine__get_running_kernel_start (struct machine *machine, +const char **symbol_name) {char filename[PATH_MAX ];int i;
@@ -621,7 +621,7 @@ static u64 machine__get_kernel_start_addr(struct machine *machine,int __machine__create_kernel_maps (struct machine *machine,struct dso *kernel) {enum map_type type; -u64 start =machine__get_kernel_start_addr (machine,NULL ); +u64 start =machine__get_running_kernel_start (machine,NULL );for (type =0 ; type <MAP__NR_TYPES ; ++type) {struct kmap *kmap;
@@ -940,7 +940,7 @@ int machine__create_kernel_maps(struct machine *machine) {struct dso *kernel =machine__get_kernel (machine);const char *name; -u64 addr =machine__get_kernel_start_addr (machine, &name); +u64 addr =machine__get_running_kernel_start (machine, &name);if (!addr)return -1 ;
Adrian Hunter
reviewerPATCH 4/24 perf tools: Add machine__kernel_ip()
In reply to Adrian HunterAdd a function to determine if an address is in
the kernel. This is based on the kernel functionkernel_ip().
--- tools/perf/util/event.c | 6 +++--- tools/perf/util/machine.c | 23 +++++++++++++++++++++++ tools/perf/util/machine.h | 17 +++++++++++++++++ 3 files changed, 43 insertions(+), 3 deletions(-)
tools/perf/util/event.c +3 -3
@@ -784,9 +784,9 @@ try_again:* "[vdso]" dso, but for now lets use the old trick of looking * in the whole kernel symbol list. */ -if ((long long )al->addr <0 && - cpumode ==PERF_RECORD_MISC_USER && - machine && mg != &machine->kmaps) { +if (cpumode ==PERF_RECORD_MISC_USER && machine && + mg != &machine->kmaps && +machine__kernel_ip (machine, al->addr)) { mg = &machine->kmaps; load_map = true; goto try_again;
tools/perf/util/machine.c +23 -0
@@ -32,6 +32,7 @@ int machine__init(struct machine *machine, const char *root_dir, pid_t pid) machine->symbol_filter =NULL ; machine->id_hdr_size =0 ; machine->comm_exec = false; + machine->kernel_start =0 ; machine->root_dir =strdup (root_dir);if (machine->root_dir ==NULL )
@@ -1559,3 +1560,25 @@ int machine__set_current_tid(struct machine *machine, int cpu, pid_t pid,return 0 ; } + +int machine__get_kernel_start (struct machine *machine) +{ +struct map *map =machine__kernel_map (machine,MAP__FUNCTION ); +int err =0 ; + +/* +* The only addresses above 2^63 are kernel addresses of a 64-bit +* kernel. Note that addresses are unsigned so that on a 32-bit system +* all addresses including kernel addresses are less than 2^32. In +* that case (32-bit system), if the kernel mapping is unknown, all +* addresses will be assumed to be in user space - see +* machine__kernel_ip(). +*/ + machine->kernel_start =1ULL <<63 ; +if (map) { + err =map__load (map, machine->symbol_filter); +if (map->start) + machine->kernel_start = map->start; + } +return err; +}
tools/perf/util/machine.h +17 -0
@@ -36,6 +36,7 @@ struct machine {struct list_head kernel_dsos;struct map_groups kmaps;struct map *vmlinux_maps[MAP__NR_TYPES ]; +u64 kernel_start;symbol_filter_t symbol_filter;pid_t *current_tid; };
@@ -46,6 +47,22 @@ struct map *machine__kernel_map(struct machine *machine, enum map_type type)return machine->vmlinux_maps[type]; } +int machine__get_kernel_start (struct machine *machine); + +static inline u64 machine__kernel_start (struct machine *machine) +{ +if (!machine->kernel_start) +machine__get_kernel_start (machine); +return machine->kernel_start; +} + +static inline bool machine__kernel_ip (struct machine *machine,u64 ip) +{ +u64 kernel_start =machine__kernel_start (machine); + +return ip >= kernel_start; +} +struct thread *machine__find_thread (struct machine *machine,pid_t pid,pid_t tid); struct comm *machine__thread_exec_comm(struct machine *machine,
Signature · 1 line
1.8.3.2
tip-bot for Adrian Hunter
merge bot[tip:perf/core] perf tools: Add machine__kernel_ip()
perf tools: Add machine__kernel_ip()
Commit-ID: fbe2af45f6bd27ee69fd775303c936c3af4a4807 Gitweb: http://git.kernel.org/tip/fbe2af45f6bd27ee69fd775303c936c3af4a4807 Author: Adrian Hunter <[email protected]> AuthorDate: Fri, 15 Aug 2014 22:08:39 +0300 Committer: Arnaldo Carvalho de Melo <[email protected]> CommitDate: Fri, 22 Aug 2014 13:12:12 -0300
perf tools: Add machine__kernel_ip()
Add a function to determine if an address is in the kernel. This is based on the kernel function kernel_ip().
--- tools/perf/util/event.c | 6 +++--- tools/perf/util/machine.c | 23 +++++++++++++++++++++++ tools/perf/util/machine.h | 17 +++++++++++++++++ 3 files changed, 43 insertions(+), 3 deletions(-)
tools/perf/util/event.c +3 -3
@@ -784,9 +784,9 @@ try_again:* "[vdso]" dso, but for now lets use the old trick of looking * in the whole kernel symbol list. */ -if ((long long )al->addr <0 && - cpumode ==PERF_RECORD_MISC_USER && - machine && mg != &machine->kmaps) { +if (cpumode ==PERF_RECORD_MISC_USER && machine && + mg != &machine->kmaps && +machine__kernel_ip (machine, al->addr)) { mg = &machine->kmaps; load_map = true; goto try_again;
tools/perf/util/machine.c +23 -0
@@ -32,6 +32,7 @@ int machine__init(struct machine *machine, const char *root_dir, pid_t pid) machine->symbol_filter =NULL ; machine->id_hdr_size =0 ; machine->comm_exec = false; + machine->kernel_start =0 ; machine->root_dir =strdup (root_dir);if (machine->root_dir ==NULL )
@@ -1559,3 +1560,25 @@ int machine__set_current_tid(struct machine *machine, int cpu, pid_t pid,return 0 ; } + +int machine__get_kernel_start (struct machine *machine) +{ +struct map *map =machine__kernel_map (machine,MAP__FUNCTION ); +int err =0 ; + +/* +* The only addresses above 2^63 are kernel addresses of a 64-bit +* kernel. Note that addresses are unsigned so that on a 32-bit system +* all addresses including kernel addresses are less than 2^32. In +* that case (32-bit system), if the kernel mapping is unknown, all +* addresses will be assumed to be in user space - see +* machine__kernel_ip(). +*/ + machine->kernel_start =1ULL <<63 ; +if (map) { + err =map__load (map, machine->symbol_filter); +if (map->start) + machine->kernel_start = map->start; + } +return err; +}
tools/perf/util/machine.h +17 -0
@@ -36,6 +36,7 @@ struct machine {struct list_head kernel_dsos;struct map_groups kmaps;struct map *vmlinux_maps[MAP__NR_TYPES ]; +u64 kernel_start;symbol_filter_t symbol_filter;pid_t *current_tid; };
@@ -46,6 +47,22 @@ struct map *machine__kernel_map(struct machine *machine, enum map_type type)return machine->vmlinux_maps[type]; } +int machine__get_kernel_start (struct machine *machine); + +static inline u64 machine__kernel_start (struct machine *machine) +{ +if (!machine->kernel_start) +machine__get_kernel_start (machine); +return machine->kernel_start; +} + +static inline bool machine__kernel_ip (struct machine *machine,u64 ip) +{ +u64 kernel_start =machine__kernel_start (machine); + +return ip >= kernel_start; +} +struct thread *machine__find_thread (struct machine *machine,pid_t pid,pid_t tid); struct comm *machine__thread_exec_comm(struct machine *machine,
Adrian Hunter
reviewerPATCH 5/24 perf tools: Let a user specify a PMU event without any config terms
In reply to Adrian HunterThis enables a PMU event to be specified in the form:
pmu//
which is effectively the same as:
pmu/config=0/
This patch is a precursor to defining
default config for a PMU.
--- tools/perf/util/parse-events.c | 6 ++++++ tools/perf/util/parse-events.y | 10 ++++++++++ 2 files changed, 16 insertions(+)
tools/perf/util/parse-events.c +6 -0
@@ -645,6 +645,12 @@ int parse_events_add_pmu(struct list_head *list, int *idx,memset (&attr,0 ,sizeof (attr)); +if (!head_config) { + attr.type = pmu->type; + evsel =__add_event (list, idx, &attr,NULL , pmu->cpus); +return evsel ?0 : -ENOMEM ; + } +if (perf_pmu__check_alias (pmu, head_config, &unit, &scale))return -EINVAL ;
tools/perf/util/parse-events.y +10 -0
@@ -210,6 +210,16 @@ PE_NAME '/' event_config '/' parse_events__free_terms($3); $$ = list; } +| +PE_NAME '/' '/' +{ + struct parse_events_evlist *data = _data; + struct list_head *list; + + ALLOC_LIST(list); + ABORT_ON(parse_events_add_pmu(list, &data->idx, $1, NULL)); + $$ = list; +} value_sym: PE_VALUE_SYM_HW
Signature · 1 line
1.8.3.2
tip-bot for Adrian Hunter
merge bot[tip:perf/core] perf tools: Let a user specify a PMU event without any config terms
perf tools: Let a user specify a PMU event without any config terms
Commit-ID: ad96227349901838e1a7f96f1dc22d96a97520c0 Gitweb: http://git.kernel.org/tip/ad96227349901838e1a7f96f1dc22d96a97520c0 Author: Adrian Hunter <[email protected]> AuthorDate: Fri, 15 Aug 2014 22:08:40 +0300 Committer: Arnaldo Carvalho de Melo <[email protected]> CommitDate: Wed, 17 Sep 2014 17:08:08 -0300
perf tools: Let a user specify a PMU event without any config terms
This enables a PMU event to be specified in the form:
pmu//
which is effectively the same as:
pmu/config=0/
This patch is a precursor to defining default config for a PMU.
Further explanation extracted from lkml thread:
Imagine that the 'tsc' term did not exist.
Intel PT trace data would not contain TSC packets, and the decoder would not know how to decode them.
Then imagine that a new version of the hardware adds 'tsc'.
It is such a useful feature that we want it by default, but older versions of the tools don't know how to decode it, so the kernel cannot turn it on by default.
It is similar to why the kernel does not select perf_event_attr.mmap2 by default.
The kernel doesn't know whether the tool supports it.
--- tools/perf/util/parse-events.c | 6 ++++++ tools/perf/util/parse-events.y | 10 ++++++++++ 2 files changed, 16 insertions(+)
tools/perf/util/parse-events.c +6 -0
@@ -645,6 +645,12 @@ int parse_events_add_pmu(struct list_head *list, int *idx,memset (&attr,0 ,sizeof (attr)); +if (!head_config) { + attr.type = pmu->type; + evsel =__add_event (list, idx, &attr,NULL , pmu->cpus); +return evsel ?0 : -ENOMEM ; + } +if (perf_pmu__check_alias (pmu, head_config, &unit, &scale))return -EINVAL ;
tools/perf/util/parse-events.y +10 -0
@@ -210,6 +210,16 @@ PE_NAME '/' event_config '/' parse_events__free_terms($3); $$ = list; } +| +PE_NAME '/' '/' +{ + struct parse_events_evlist *data = _data; + struct list_head *list; + + ALLOC_LIST(list); + ABORT_ON(parse_events_add_pmu(list, &data->idx, $1, NULL)); + $$ = list; +} value_sym: PE_VALUE_SYM_HW
Adrian Hunter
reviewerPATCH 6/24 perf tools: Let default config be defined for a PMU
In reply to Adrian HunterThis allows default config terms to be provided
for a PMU. So, for example, when the Intel PT
PMU is added, it will be possible to specify:
intel_pt//
which will be the same as:
intel_pt/tsc=1,noretcomp=0/
meaning that the trace should contain
TSC timestamps and perform 'return
compression'.
An important consideration of this
patch is that it must be possible
to overwrite the default values.
That has meant changing the logic
so that a zero value can replace
a non-zero value.
--- tools/perf/tests/pmu.c | 2 +- tools/perf/util/parse-events.c | 7 ++++++- tools/perf/util/pmu.c | 42 ++++++++++++++++++++++++++---------------- tools/perf/util/pmu.h | 9 ++++++++- 4 files changed, 41 insertions(+), 19 deletions(-)
tools/perf/tests/pmu.c +1 -1
@@ -152,7 +152,7 @@ int test__pmu(void)if (ret)break ; - ret =perf_pmu__config_terms (&formats, &attr, terms); + ret =perf_pmu__config_terms (&formats, &attr, terms, false);if (ret)break ;
tools/perf/util/parse-events.c +6 -1
@@ -643,7 +643,12 @@ int parse_events_add_pmu(struct list_head *list, int *idx,if (!pmu)return -EINVAL ; -memset (&attr,0 ,sizeof (attr)); +if (pmu->default_config) { +memcpy (&attr, pmu->default_config, +sizeof (struct perf_event_attr )); + }else { +memset (&attr,0 ,sizeof (attr)); + }if (!head_config) { attr.type = pmu->type;
tools/perf/util/pmu.c +26 -16
105 lines not on this page · raw message
tools/perf/util/pmu.h +8 -1
@@ -13,9 +13,12 @@ enum {#define PERF_PMU_FORMAT_BITS 64 +struct perf_event_attr ; +struct perf_pmu {char *name;__u32 type; +struct perf_event_attr *default_config;struct cpu_map *cpus;struct list_head format;/* HEAD struct perf_pmu_format -> list */ struct list_head aliases;/* HEAD struct perf_pmu_alias -> list */
@@ -27,7 +30,8 @@ int perf_pmu__config(struct perf_pmu *pmu, struct perf_event_attr *attr, struct list_head *head_terms);int perf_pmu__config_terms (struct list_head *formats,struct perf_event_attr *attr, -struct list_head *head_terms); +struct list_head *head_terms, +bool zero);int perf_pmu__check_alias (struct perf_pmu *pmu,struct list_head *head_terms,const char **unit,double *scale); struct list_head *perf_pmu__alias(struct perf_pmu *pmu,
@@ -46,4 +50,7 @@ void print_pmu_events(const char *event_glob, bool name_only);bool pmu_have_event (const char *pname,const char *name);int perf_pmu__test (void ); + +struct perf_event_attr *perf_pmu__get_default_config (struct perf_pmu *pmu); +#endif /* __PMU_H */
Signature · 1 line
1.8.3.2
Adrian Hunter
reviewerPATCH 7/24 perf tools: Add perf_pmu__scan_file()
In reply to Adrian HunterAdd a function to scan a sysfs file within the pmu device directory.
This will be used to read capability values from the PMU 'caps' subdirectory.
--- tools/perf/util/pmu.c | 37 +++++++++++++++++++++++++++++++++++++ tools/perf/util/pmu.h | 3 +++ 2 files changed, 40 insertions(+)
tools/perf/util/pmu.c +37 -0
@@ -3,6 +3,7 @@#include <unistd.h> #include <stdio.h> #include <stdbool.h> +#include <stdarg.h> #include <dirent.h> #include <api/fs/fs.h> #include <locale.h>
@@ -804,3 +805,39 @@ bool pmu_have_event(const char *pname, const char *name) }return false; } + +static FILE *perf_pmu__open_file (struct perf_pmu *pmu,const char *name) +{ +struct stat st; +char path[PATH_MAX ]; +const char *sysfs; + + sysfs =sysfs__mountpoint (); +if (!sysfs) +return NULL ; + +snprintf (path,PATH_MAX , +"%s" EVENT_SOURCE_DEVICE_PATH "%s/%s" , sysfs, pmu->name, name); + +if (stat (path, &st) <0 ) +return NULL ; + +return fopen (path,"r" ); +} + +int perf_pmu__scan_file (struct perf_pmu *pmu,const char *name,const char *fmt, + ...) +{ +va_list args; +FILE *file; +int ret =EOF ; + +va_start (args, fmt); + file =perf_pmu__open_file (pmu, name); +if (file) { + ret =vfscanf (file, fmt, args); +fclose (file); + } +va_end (args); +return ret; +}
tools/perf/util/pmu.h +3 -0
@@ -49,6 +49,9 @@ struct perf_pmu *perf_pmu__scan(struct perf_pmu *pmu);void print_pmu_events (const char *event_glob,bool name_only);bool pmu_have_event (const char *pname,const char *name); +int perf_pmu__scan_file (struct perf_pmu *pmu,const char *name,const char *fmt, + ...) __attribute__((format (scanf,3 ,4 ))); +int perf_pmu__test (void );struct perf_event_attr *perf_pmu__get_default_config (struct perf_pmu *pmu);
Signature · 1 line
1.8.3.2
Adrian Hunter
reviewerPATCH 8/24 perf tools: Add id index
In reply to Adrian HunterAdd an index of the event identifiers.
This is needed to queue Instruction
Trace samples according to the mmap
buffer from which they were recorded.
--- tools/perf/builtin-inject.c | 1 + tools/perf/util/event.c | 1 + tools/perf/util/event.h | 15 ++++++ tools/perf/util/evlist.c | 26 ++++++++-- tools/perf/util/evsel.h | 3 ++ tools/perf/util/session.c | 122 ++++++++++++++++++++++++++++++++++++++++++++ tools/perf/util/session.h | 10 ++++ tools/perf/util/tool.h | 3 +- 8 files changed, 177 insertions(+), 4 deletions(-)
tools/perf/builtin-inject.c +1 -0
@@ -410,6 +410,7 @@ int cmd_inject(int argc, const char **argv, const char *prefix __maybe_unused) .tracing_data = perf_event__repipe_op2_synth, .finished_round = perf_event__repipe_op2_synth, .build_id = perf_event__repipe_op2_synth, + .id_index = perf_event__repipe_op2_synth, }, .input_name = "-", .samples = LIST_HEAD_INIT(inject.samples),
tools/perf/util/event.c +1 -0
@@ -28,6 +28,7 @@ static const char *perf_event__names[] = { [PERF_RECORD_HEADER_TRACING_DATA] = "TRACING_DATA", [PERF_RECORD_HEADER_BUILD_ID] = "BUILD_ID", [PERF_RECORD_FINISHED_ROUND] = "FINISHED_ROUND", + [PERF_RECORD_ID_INDEX] = "ID_INDEX", };const char *perf_event__name (unsigned int id)
tools/perf/util/event.h +15 -0
@@ -187,6 +187,7 @@ enum perf_user_event_type { /* above any possible kernel type */PERF_RECORD_HEADER_TRACING_DATA =66 ,PERF_RECORD_HEADER_BUILD_ID =67 ,PERF_RECORD_FINISHED_ROUND =68 , +PERF_RECORD_ID_INDEX =69 , PERF_RECORD_HEADER_MAX };
@@ -213,6 +214,19 @@ struct tracing_data_event {u32 size; }; +struct id_index_entry { +u64 id; +u64 idx; +u64 cpu; +u64 tid; +}; + +struct id_index_event { +struct perf_event_header header; +u64 nr; +struct id_index_entry entries[0 ]; +}; +union perf_event {struct perf_event_header header;struct mmap_event mmap;
@@ -227,6 +241,7 @@ union perf_event {struct event_type_event event_type;struct tracing_data_event tracing_data;struct build_id_event build_id; +struct id_index_event id_index; };void perf_event__print_totals (void );
tools/perf/util/evlist.c +23 -3
@@ -489,6 +489,22 @@ static int perf_evlist__id_add_fd(struct perf_evlist *evlist,return 0 ; } +static void perf_evlist__set_sid_idx (struct perf_evlist *evlist, +struct perf_evsel *evsel,int idx,int cpu, +int thread) +{ +struct perf_sample_id *sid =SID (evsel, cpu, thread); + sid->idx = idx; +if (evlist->cpus && cpu >=0 ) + sid->cpu = evlist->cpus->map[cpu]; +else + sid->cpu =-1 ; +if (!evsel->system_wide && evlist->threads && thread >=0 ) + sid->tid = evlist->threads->map[thread]; +else + sid->tid =-1 ; +} +struct perf_sample_id *perf_evlist__id2sid (struct perf_evlist *evlist,u64 id) {struct hlist_head *head;
@@ -724,9 +740,13 @@ static int perf_evlist__mmap_per_evsel(struct perf_evlist *evlist, int idx,return -1 ; } -if ((evsel->attr.read_format &PERF_FORMAT_ID ) && -perf_evlist__id_add_fd (evlist, evsel, cpu, thread, fd) <0 ) -return -1 ; +if (evsel->attr.read_format &PERF_FORMAT_ID ) { +if (perf_evlist__id_add_fd (evlist, evsel, cpu, thread, + fd) <0 ) +return -1 ; +perf_evlist__set_sid_idx (evlist, evsel, idx, cpu, + thread); + } }return 0 ;
tools/perf/util/evsel.h +3 -0
@@ -38,6 +38,9 @@ struct perf_sample_id {struct hlist_node node;u64 id;struct perf_evsel *evsel; +int idx; +int cpu; +pid_t tid;/* Holds total ID period value for PERF_SAMPLE_READ processing. */ u64 period;
tools/perf/util/session.c +122 -0
149 lines not on this page · raw message
tools/perf/util/session.h +10 -0
@@ -127,4 +127,14 @@ int __perf_session__set_tracepoints_handlers(struct perf_session *session,extern volatile int session_done;#define session_done () (*(volatile int *)(&session_done)) + +int perf_event__process_id_index (struct perf_tool *tool, +union perf_event *event, +struct perf_session *session); + +int perf_event__synthesize_id_index (struct perf_tool *tool, +perf_event__handler_t process, +struct perf_evlist *evlist, +struct machine *machine); +#endif /* __PERF_SESSION_H */
tools/perf/util/tool.h +2 -1
@@ -39,7 +39,8 @@ struct perf_tool {event_attr_op attr;event_op2 tracing_data;event_op2 finished_round, - build_id; + build_id, + id_index;bool ordered_events;bool ordering_requires_timestamps; };
Signature · 1 line
1.8.3.2
Adrian Hunter
reviewerPATCH 9/24 perf pmu: Let pmu's with no events show up on perf list
In reply to Adrian Hunterperf list only lists PMUs with events. Add a
flag to cause a PMU to be also listed separately.
--- tools/perf/util/pmu.c | 13 +++++++++++-- tools/perf/util/pmu.h | 1 + 2 files changed, 12 insertions(+), 2 deletions(-)
tools/perf/util/pmu.c +11 -2
@@ -749,15 +749,18 @@ void print_pmu_events(const char *event_glob, bool name_only) pmu =NULL ; len =0 ; -while ((pmu =perf_pmu__scan (pmu)) !=NULL ) +while ((pmu =perf_pmu__scan (pmu)) !=NULL ) {list_for_each_entry (alias, &pmu->aliases, list) len++; +if (pmu->selectable) + len++; + } aliases =malloc (sizeof (char *) * len);if (!aliases)return ; pmu =NULL ; j =0 ; -while ((pmu =perf_pmu__scan (pmu)) !=NULL ) +while ((pmu =perf_pmu__scan (pmu)) !=NULL ) {list_for_each_entry (alias, &pmu->aliases, list) {char *name =format_alias (buf,sizeof (buf), pmu, alias);bool is_cpu = !strcmp (pmu->name,"cpu" );
@@ -774,6 +777,12 @@ void print_pmu_events(const char *event_glob, bool name_only) aliases[j] =strdup (aliases[j]); j++; } +if (pmu->selectable) { +scnprintf (buf,sizeof (buf),"%s//" , pmu->name); + aliases[j] =strdup (buf); + j++; + } + } len = j;qsort (aliases, len,sizeof (char *), cmp_string);for (j =0 ; j < len; j++) {
tools/perf/util/pmu.h +1 -0
@@ -18,6 +18,7 @@ struct perf_event_attr;struct perf_pmu {char *name;__u32 type; +bool selectable;struct perf_event_attr *default_config;struct cpu_map *cpus;struct list_head format;/* HEAD struct perf_pmu_format -> list */
Signature · 1 line
1.8.3.2
Adrian Hunter
reviewerPATCH 10/24 perf session: Add perf_session__deliver_synth_event()
In reply to Adrian HunterAdd a function to deliver synthesized events from
within a session.
--- tools/perf/util/session.c | 14 ++++++++++++++ tools/perf/util/session.h | 5 +++++ 2 files changed, 19 insertions(+)
tools/perf/util/session.c +14 -0
@@ -924,6 +924,20 @@ static s64 perf_session__process_user_event(struct perf_session *session, } } +int perf_session__deliver_synth_event (struct perf_session *session, +union perf_event *event, +struct perf_sample *sample, +struct perf_tool *tool) +{ +events_stats__inc (&session->stats, event->header.type); + +if (event->header.type >=PERF_RECORD_USER_TYPE_START ) +return perf_session__process_user_event (session, event, tool, +0 ); + +return perf_session__deliver_event (session, event, sample, tool,0 ); +} +static void event_swap (union perf_event *event,bool sample_id_all) {perf_event__swap_op swap;
tools/perf/util/session.h +5 -0
@@ -128,6 +128,11 @@ extern volatile int session_done;#define session_done () (*(volatile int *)(&session_done)) +int perf_session__deliver_synth_event (struct perf_session *session, +union perf_event *event, +struct perf_sample *sample, +struct perf_tool *tool); +int perf_event__process_id_index (struct perf_tool *tool,union perf_event *event,struct perf_session *session);
Signature · 1 line
1.8.3.2
Adrian Hunter
reviewerPATCH 11/24 perf tools: Add a thread stack for synthesizing call chains
In reply to Adrian HunterAdd a thread stack for synthesizing call chains from call and return events.
--- tools/perf/Makefile.perf | 2 + tools/perf/util/event.h | 26 +++++++ tools/perf/util/thread-stack.c | 151 +++++++++++++++++++++++++++++++++++++++++ tools/perf/util/thread-stack.h | 32 +++++++++ tools/perf/util/thread.c | 3 + tools/perf/util/thread.h | 3 + 6 files changed, 217 insertions(+) create mode 100644 tools/perf/util/thread-stack.c create mode 100644 tools/perf/util/thread-stack.h
tools/perf/Makefile.perf +2 -0
@@ -303,6 +303,7 @@ LIB_H += ui/util.h LIB_H += ui/ui.h LIB_H += util/data.h LIB_H += util/kvm-stat.h +LIB_H += util/thread-stack.h LIB_OBJS += $(OUTPUT)util/abspath.o LIB_OBJS += $(OUTPUT)util/alias.o
@@ -379,6 +380,7 @@ LIB_OBJS += $(OUTPUT)util/srcline.o LIB_OBJS += $(OUTPUT)util/data.o LIB_OBJS += $(OUTPUT)util/tsc.o LIB_OBJS += $(OUTPUT)util/cloexec.o +LIB_OBJS += $(OUTPUT)util/thread-stack.o LIB_OBJS += $(OUTPUT)ui/setup.o LIB_OBJS += $(OUTPUT)ui/helpline.o
tools/perf/util/event.h +26 -0
@@ -143,6 +143,32 @@ struct branch_stack {struct branch_entry entries[0 ]; }; +enum { +PERF_FLAG_BRANCH =1ULL <<0 , +PERF_FLAG_CALL =1ULL <<1 , +PERF_FLAG_RETURN =1ULL <<2 , +PERF_FLAG_CONDITIONAL =1ULL <<3 , +PERF_FLAG_SYSCALLRET =1ULL <<4 , +PERF_FLAG_ASYNC =1ULL <<5 , +PERF_FLAG_INTERRUPT =1ULL <<6 , +PERF_FLAG_TX_ABORT =1ULL <<7 , +PERF_FLAG_TRACE_BEGIN =1ULL <<8 , +PERF_FLAG_TRACE_END =1ULL <<9 , +PERF_FLAG_IN_TX =1ULL <<10 , +}; + +#define PERF_BRANCH_MASK (\ + PERF_FLAG_BRANCH |\ + PERF_FLAG_CALL |\ + PERF_FLAG_RETURN |\ + PERF_FLAG_CONDITIONAL |\ + PERF_FLAG_SYSCALLRET |\ + PERF_FLAG_ASYNC |\ + PERF_FLAG_INTERRUPT |\ + PERF_FLAG_TX_ABORT |\ + PERF_FLAG_TRACE_BEGIN |\ + PERF_FLAG_TRACE_END) +struct perf_sample {u64 ip;u32 pid, tid;
tools/perf/util/thread-stack.c new file +151 -0
151 lines not on this page · raw message
tools/perf/util/thread-stack.h new file +32 -0
@@ -0,0 +1,32 @@ +/* +* thread-stack.h: Synthesize a thread's stack using call / return events +* Copyright (c) 2014, Intel Corporation. +* +* This program is free software; you can redistribute it and/or modify it +* under the terms and conditions of the GNU General Public License, +* version 2, as published by the Free Software Foundation. +* +* This program is distributed in the hope it will be useful, but WITHOUT +* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or +* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for +* more details. +* +*/ + +#ifndef __PERF_THREAD_STACK_H +#define __PERF_THREAD_STACK_H + +#include <sys/types.h> + +#include <linux/types.h> + +struct thread ; +struct ip_callchain ; + +void thread_stack__event (struct thread *thread,u32 flags,u64 from_ip, +u64 to_ip,u16 insn_len,u64 trace_nr); +void thread_stack__sample (struct thread *thread,struct ip_callchain *chain, +size_t sz,u64 ip); +void thread_stack__free (struct thread *thread); + +#endif
tools/perf/util/thread.c +3 -0
@@ -4,6 +4,7 @@#include <string.h> #include "session.h" #include "thread.h" +#include "thread-stack.h" #include "util.h" #include "debug.h" #include "comm.h"
@@ -61,6 +62,8 @@ void thread__delete(struct thread *thread) {struct comm *comm, *tmp; +thread_stack__free (thread); +if (thread->mg) {map_groups__put (thread->mg); thread->mg =NULL ;
tools/perf/util/thread.h +3 -0
@@ -8,6 +8,8 @@#include "symbol.h" #include <strlist.h> +struct thread_stack ; + struct thread { union { struct rb_node rb_node;
@@ -25,6 +27,7 @@ struct thread {int comm_len;void *priv; +struct thread_stack *ts; };struct machine ;
Signature · 1 line
1.8.3.2
Adrian Hunter
reviewerPATCH 12/24 perf tools: Add facility to export data in database-friendly way
In reply to Adrian HunterThis patch introduces an abstraction for exporting sample data in a database-friendly way. The abstraction does not implement the actual output. A subsequent patch takes this facility into use for extending the script interface.
The abstraction is needed because static data like symbols, dsos, comms etc need to be exported only once. That means allocating them a unique identifier and recording it on each structure. The member 'db_id' is used for that. 'db_id' is just a 64-bit sequence number.
Exporting centres around the db_export__sample() function which exports the associated data structures if they have not yet been allocated a db_id.
--- tools/perf/Makefile.perf | 2 + tools/perf/util/comm.h | 1 + tools/perf/util/db-export.c | 268 ++++++++++++++++++++++++++++++++++++++++++++ tools/perf/util/db-export.h | 86 ++++++++++++++ tools/perf/util/dso.h | 1 + tools/perf/util/evsel.h | 1 + tools/perf/util/machine.h | 1 + tools/perf/util/symbol.h | 1 + tools/perf/util/thread.h | 1 + 9 files changed, 362 insertions(+) create mode 100644 tools/perf/util/db-export.c create mode 100644 tools/perf/util/db-export.h
tools/perf/Makefile.perf +2 -0
@@ -304,6 +304,7 @@ LIB_H += ui/ui.h LIB_H += util/data.h LIB_H += util/kvm-stat.h LIB_H += util/thread-stack.h +LIB_H += util/db-export.h LIB_OBJS += $(OUTPUT)util/abspath.o LIB_OBJS += $(OUTPUT)util/alias.o
@@ -381,6 +382,7 @@ LIB_OBJS += $(OUTPUT)util/data.o LIB_OBJS += $(OUTPUT)util/tsc.o LIB_OBJS += $(OUTPUT)util/cloexec.o LIB_OBJS += $(OUTPUT)util/thread-stack.o +LIB_OBJS += $(OUTPUT)util/db-export.o LIB_OBJS += $(OUTPUT)ui/setup.o LIB_OBJS += $(OUTPUT)ui/helpline.o
tools/perf/util/comm.h +1 -0
@@ -10,6 +10,7 @@ struct comm_str;struct comm {struct comm_str *comm_str;u64 start; +u64 db_id;struct list_head list;bool exec; };
tools/perf/util/db-export.c new file +268 -0
268 lines not on this page · raw message
tools/perf/util/db-export.h new file +86 -0
86 lines not on this page · raw message
tools/perf/util/dso.h +1 -0
@@ -101,6 +101,7 @@ struct dso {enum dso_swap_type needs_swap;enum dso_binary_type symtab_type;enum dso_binary_type binary_type; +u64 db_id; u8 adjust_symbols:1; u8 has_build_id:1; u8 has_srcline:1;
tools/perf/util/evsel.h +1 -0
@@ -96,6 +96,7 @@ struct perf_evsel {int sample_read;struct perf_evsel *leader;char *group_name; +u64 db_id; };union u64_swap {
tools/perf/util/machine.h +1 -0
@@ -39,6 +39,7 @@ struct machine {u64 kernel_start;symbol_filter_t symbol_filter;pid_t *current_tid; +u64 db_id; }; static inline
tools/perf/util/symbol.h +1 -0
@@ -73,6 +73,7 @@ struct symbol {struct rb_node rb_node;u64 start;u64 end; +u64 db_id;u16 namelen;u8 binding;bool ignore;
tools/perf/util/thread.h +1 -0
@@ -25,6 +25,7 @@ struct thread {bool dead;/* if set thread has exited */ struct list_head comm_list;int comm_len; +u64 db_id;void *priv;struct thread_stack *ts;
Signature · 1 line
1.8.3.2
Adrian Hunter
reviewerPATCH 13/24 perf tools: Extend Python script interface to export data in a database-friendly way
In reply to Adrian HunterUse the new db_export facility to export data in a database-friendly way.
A Python script selects the db_export mode by setting a global variable 'perf_db_export_mode' to True. The script then optionally implements functions to receive table rows. The functions are:
evsel_table
machine_table
thread_table
comm_table
dso_table
symbol_table
sample_table
An example script is provided in a subsequent patch.
--- .../util/scripting-engines/trace-event-python.c | 281 ++++++++++++++++++++- 1 file changed, 279 insertions(+), 2 deletions(-)
tools/perf/util/scripting-engines/trace-event-python.c +279 -2
@@ -24,6 +24,7 @@#include <stdio.h> #include <stdlib.h> #include <string.h> +#include <stdbool.h> #include <errno.h> #include "../../perf.h"
@@ -32,6 +33,9 @@#include "../util.h" #include "../event.h" #include "../thread.h" +#include "../comm.h" +#include "../machine.h" +#include "../db-export.h" #include "../trace-event.h" #include "../machine.h"
@@ -52,6 +56,21 @@ static int zero_flag_atom;static PyObject *main_module, *main_dict; +struct tables { +struct db_export dbe; +PyObject *evsel_handler; +PyObject *machine_handler; +PyObject *thread_handler; +PyObject *comm_handler; +PyObject *comm_thread_handler; +PyObject *dso_handler; +PyObject *symbol_handler; +PyObject *sample_handler; +bool db_export_mode; +}; + +static struct tables tables_global; + static void handler_call_die(const char *handler_name) NORETURN;static void handler_call_die (const char *handler_name) {
@@ -474,6 +493,210 @@ static void python_process_tracepoint(struct perf_sample *sample,Py_DECREF (t); } +static PyObject *tuple_new (unsigned int sz) +{ +PyObject *t; + + t =PyTuple_New (sz); +if (!t) +Py_FatalError ("couldn't create Python tuple" ); +return t; +} + +static int tuple_set_u64 (PyObject *t,unsigned int pos,u64 val) +{ +#if BITS_PER_LONG ==64 +return PyTuple_SetItem (t, pos,PyInt_FromLong (val)); +#endif +#if BITS_PER_LONG ==32 +return PyTuple_SetItem (t, pos,PyLong_FromLongLong (val)); +#endif +} + +static int tuple_set_s32 (PyObject *t,unsigned int pos,s32 val) +{ +return PyTuple_SetItem (t, pos,PyInt_FromLong (val)); +} + +static int tuple_set_string (PyObject *t,unsigned int pos,const char *s) +{ +return PyTuple_SetItem (t, pos,PyString_FromString (s)); +} + +static int python_export_evsel (struct db_export *dbe,struct perf_evsel *evsel) +{ + struct tables *tables = container_of(dbe, struct tables, dbe); +PyObject *t; + + t =tuple_new (2 ); + +tuple_set_u64 (t,0 , evsel->db_id); +tuple_set_string (t,1 ,perf_evsel__name (evsel)); + +call_object (tables->evsel_handler, t,"evsel_table" ); + +Py_DECREF (t); + +return 0 ; +} + +static int python_export_machine (struct db_export *dbe, +struct machine *machine) +{ + struct tables *tables = container_of(dbe, struct tables, dbe); +PyObject *t; + + t =tuple_new (3 ); + +tuple_set_u64 (t,0 , machine->db_id); +tuple_set_s32 (t,1 , machine->pid); +tuple_set_string (t,2 , machine->root_dir ? machine->root_dir :"" ); + +call_object (tables->machine_handler, t,"machine_table" ); + +Py_DECREF (t); + +return 0 ; +} + +static int python_export_thread (struct db_export *dbe,struct thread *thread, +u64 main_thread_db_id,struct machine *machine) +{ + struct tables *tables = container_of(dbe, struct tables, dbe); +PyObject *t; + + t =tuple_new (5 ); + +tuple_set_u64 (t,0 , thread->db_id); +tuple_set_u64 (t,1 , machine->db_id); +tuple_set_u64 (t,2 , main_thread_db_id); +tuple_set_s32 (t,3 , thread->pid_); +tuple_set_s32 (t,4 , thread->tid); + +call_object (tables->thread_handler, t,"thread_table" ); + +Py_DECREF (t); + +return 0 ; +} + +static int python_export_comm (struct db_export *dbe,struct comm *comm) +{ + struct tables *tables = container_of(dbe, struct tables, dbe); +PyObject *t; + + t =tuple_new (2 ); + +tuple_set_u64 (t,0 , comm->db_id); +tuple_set_string (t,1 ,comm__str (comm)); + +call_object (tables->comm_handler, t,"comm_table" ); + +Py_DECREF (t); + +return 0 ; +} + +static int python_export_comm_thread (struct db_export *dbe,u64 db_id, +struct comm *comm,struct thread *thread) +{ + struct tables *tables = container_of(dbe, struct tables, dbe); +PyObject *t; + + t =tuple_new (3 ); + +tuple_set_u64 (t,0 , db_id); +tuple_set_u64 (t,1 , comm->db_id); +tuple_set_u64 (t,2 , thread->db_id); + +call_object (tables->comm_thread_handler, t,"comm_thread_table" ); + +Py_DECREF (t); + +return 0 ; +} + +static int python_export_dso (struct db_export *dbe,struct dso *dso, +struct machine *machine) +{ + struct tables *tables = container_of(dbe, struct tables, dbe); +char sbuild_id[BUILD_ID_SIZE *2 +1 ]; +PyObject *t; + +build_id__sprintf (dso->build_id,sizeof (dso->build_id), sbuild_id); + + t =tuple_new (5 ); + +tuple_set_u64 (t,0 , dso->db_id); +tuple_set_u64 (t,1 , machine->db_id); +tuple_set_string (t,2 , dso->short_name); +tuple_set_string (t,3 , dso->long_name); +tuple_set_string (t,4 , sbuild_id); + +call_object (tables->dso_handler, t,"dso_table" ); + +Py_DECREF (t); + +return 0 ; +} + +static int python_export_symbol (struct db_export *dbe,struct symbol *sym, +struct dso *dso) +{ + struct tables *tables = container_of(dbe, struct tables, dbe); +PyObject *t; + + t =tuple_new (6 ); + +tuple_set_u64 (t,0 , sym->db_id); +tuple_set_u64 (t,1 , dso->db_id); +tuple_set_u64 (t,2 , sym->start); +tuple_set_u64 (t,3 , sym->end); +tuple_set_s32 (t,4 , sym->binding); +tuple_set_string (t,5 , sym->name); + +call_object (tables->symbol_handler, t,"symbol_table" ); + +Py_DECREF (t); + +return 0 ; +} + +static int python_export_sample (struct db_export *dbe, +struct export_sample *es) +{ + struct tables *tables = container_of(dbe, struct tables, dbe); +PyObject *t; + + t =tuple_new (19 ); + +tuple_set_u64 (t,0 , es->db_id); +tuple_set_u64 (t,1 , es->evsel->db_id); +tuple_set_u64 (t,2 , es->al->machine->db_id); +tuple_set_u64 (t,3 , es->thread->db_id); +tuple_set_u64 (t,4 , es->comm_db_id); +tuple_set_u64 (t,5 , es->dso_db_id); +tuple_set_u64 (t,6 , es->sym_db_id); +tuple_set_u64 (t,7 , es->offset); +tuple_set_u64 (t,8 , es->sample->ip); +tuple_set_u64 (t,9 , es->sample->time); +tuple_set_s32 (t,10 , es->sample->cpu); +tuple_set_u64 (t,11 , es->addr_dso_db_id); +tuple_set_u64 (t,12 , es->addr_sym_db_id); +tuple_set_u64 (t,13 , es->addr_offset); +tuple_set_u64 (t,14 , es->sample->addr); +tuple_set_u64 (t,15 , es->sample->period); +tuple_set_u64 (t,16 , es->sample->weight); +tuple_set_u64 (t,17 , es->sample->transaction); +tuple_set_u64 (t,18 , es->sample->data_src); + +call_object (tables->sample_handler, t,"sample_table" ); + +Py_DECREF (t); + +return 0 ; +} + static void python_process_general_event(struct perf_sample *sample, struct perf_evsel *evsel, struct thread *thread,
@@ -550,19 +773,25 @@ exit:Py_DECREF (t); } -static void python_process_event (union perf_event *event__maybe_unused , +static void python_process_event (union perf_event *event,struct perf_sample *sample,struct perf_evsel *evsel,struct thread *thread,struct addr_location *al) { +struct tables *tables = &tables_global; +switch (evsel->attr.type) {case PERF_TYPE_TRACEPOINT :python_process_tracepoint (sample, evsel, thread, al);break ;/* Reserve for future process_hw/sw/raw APIs */ default : -python_process_general_event (sample, evsel, thread, al); +if (tables->db_export_mode) +db_export__sample (&tables->dbe, event, sample, evsel, + thread, al); +else +python_process_general_event (sample, evsel, thread, al); } }
@@ -588,11 +817,53 @@ error:return -1 ; } +#define SET_TABLE_HANDLER_ (name, handler_name, table_name) do { \ + tables->handler_name = get_handler(#table_name); \ + if (tables->handler_name) \ + tables->dbe.export_ ## name = python_export_ ## name; \ +} while (0) + +#define SET_TABLE_HANDLER (name) \ + SET_TABLE_HANDLER_(name, name ## _handler, name ## _table) + +static void set_table_handlers (struct tables *tables) +{ +const char *perf_db_export_mode ="perf_db_export_mode" ; +PyObject *db_export_mode; +int ret; + +memset (tables,0 ,sizeof (struct tables )); +if (db_export__init (&tables->dbe)) +Py_FatalError ("failed to initialize export" ); + + db_export_mode =PyDict_GetItemString (main_dict, perf_db_export_mode); +if (!db_export_mode) +return ; + + ret =PyObject_IsTrue (db_export_mode); +if (ret ==-1 ) +handler_call_die (perf_db_export_mode); +if (!ret) +return ; + + tables->db_export_mode = true; + +SET_TABLE_HANDLER (evsel); +SET_TABLE_HANDLER (machine); +SET_TABLE_HANDLER (thread); +SET_TABLE_HANDLER (comm); +SET_TABLE_HANDLER (comm_thread); +SET_TABLE_HANDLER (dso); +SET_TABLE_HANDLER (symbol); +SET_TABLE_HANDLER (sample); +} +/* * Start trace script */ static int python_start_script (const char *script,int argc,const char **argv) { +struct tables *tables = &tables_global;const char **command_line;char buf[PATH_MAX ];int i, err =0 ;
@@ -631,6 +902,8 @@ static int python_start_script(const char *script, int argc, const char **argv)free (command_line); +set_table_handlers (tables); +return err; error:Py_Finalize ();
@@ -649,8 +922,12 @@ static int python_flush_script(void)*/ static int python_stop_script (void ) { +struct tables *tables = &tables_global; +try_call_object ("trace_end" ,NULL ); +db_export__exit (&tables->dbe); +Py_XDECREF (main_dict);Py_XDECREF (main_module);Py_Finalize ();
Signature · 1 line
1.8.3.2
Adrian Hunter
reviewerPATCH 14/24 perf tools: Add Python script to export to postgresql
In reply to Adrian HunterAdd a Python script to export to a postgresql database. The script requires the Python psycopg2 module. The caller of the script must be able to create postgresql databases.
The script takes the database name as a parameter. The database and database tables are created. Data is written to flat files which are then imported using SQL COPY FROM.
Example:
$ perf record ls ... $ perf script report export-to-postgresql lsdb 2014-02-14 10:55:38.631431 Creating database... 2014-02-14 10:55:39.291958 Writing to intermediate files... 2014-02-14 10:55:39.350280 Copying to database... 2014-02-14 10:55:39.358536 Removing intermediate files... 2014-02-14 10:55:39.358665 Adding primary keys 2014-02-14 10:55:39.658697 Adding foreign keys 2014-02-14 10:55:39.667412 Done $ psql lsdb lsdb=# select * from samples_view;
id | time | cpu | pid | tid | command | event | ip_hex | symbol | sym_offset | dso_short_name | to_ip_hex | to_symbol | to_sym_offset | to_dso_short_name
----+-----------------+-----+-------+-------+---------+--------+------------------+------------------------------+------------+-------------------+-----------+-----------+---------------+-------------------
1 | 592112202825015 | -1 | 17339 | 17339 | :17339 | cycles | ffffffff8104d24a | native_write_msr_safe | 10 | [kernel.kallsyms] | 0 | unknown | 0 | unknown 2 | 592112203258804 | -1 | 17339 | 17339 | :17339 | cycles | ffffffff8104d24a | native_write_msr_safe | 10 | [kernel.kallsyms] | 0 | unknown | 0 | unknown 3 | 592112203988119 | -1 | 17339 | 17339 | :17339 | cycles | ffffffff8104d24a | native_write_msr_safe | 10 | [kernel.kallsyms] | 0 | unknown | 0 | unknown
--- .../scripts/python/bin/export-to-postgresql-record | 8 + .../scripts/python/bin/export-to-postgresql-report | 24 ++ tools/perf/scripts/python/export-to-postgresql.py | 360 +++++++++++++++++++++ 3 files changed, 392 insertions(+) create mode 100644 tools/perf/scripts/python/bin/export-to-postgresql-record create mode 100644 tools/perf/scripts/python/bin/export-to-postgresql-report create mode 100644 tools/perf/scripts/python/export-to-postgresql.py
tools/perf/scripts/python/bin/export-to-postgresql-record new file +8 -0
@@ -0,0 +1,8 @@ +#!/bin/bash + +# +# export perf data to a postgresql database. Can cover +# perf ip samples (excluding the tracepoints). No special +# record requirements, just record what you want to export. +# +perf record $@
tools/perf/scripts/python/bin/export-to-postgresql-report new file +24 -0
@@ -0,0 +1,24 @@ +#!/bin/bash +# description: export perf data to a postgresql database +# args: [database name] [columns] +n_args=0 +for i in "$@" +do + if expr match "$i" "-" > /dev/null ; then + break + fi + n_args=$(( $n_args + 1 )) +done +if [ "$n_args" -gt 2 ] ; then + echo "usage: export-to-postgresql-report [database name] [columns]" + exit +fi +if [ "$n_args" -gt 1 ] ; then + dbname=$1 + columns=$2 + shift 2 +elif [ "$n_args" -gt 0 ] ; then + dbname=$1 + shift +fi +perf script $@ -s "$PERF_EXEC_PATH"/scripts/python/export-to-postgresql.py $dbname $columns
tools/perf/scripts/python/export-to-postgresql.py new file +360 -0
360 lines not on this page · raw message
Signature · 1 line
1.8.3.2
Adrian Hunter
reviewerPATCH 15/24 perf tools: Add branch type to db export
In reply to Adrian HunterAdd the ability to export branch types through the database export facility.
--- tools/perf/util/db-export.c | 48 +++++++++++++++++++++++++++++++++++++++++++++ tools/perf/util/db-export.h | 6 ++++++ 2 files changed, 54 insertions(+)
tools/perf/util/db-export.c +48 -0
@@ -205,6 +205,15 @@ static int db_ids_from_al(struct db_export *dbe, struct addr_location *al,return 0 ; } +int db_export__branch_type (struct db_export *dbe,u32 branch_type, +const char *name) +{ +if (dbe->export_branch_type) +return dbe->export_branch_type (dbe, branch_type, name); + +return 0 ; +} +int db_export__sample (struct db_export *dbe,union perf_event *event,struct perf_sample *sample,struct perf_evsel *evsel,struct thread *thread,struct addr_location *al)
@@ -266,3 +275,42 @@ int db_export__sample(struct db_export *dbe, union perf_event *event,return 0 ; } + +static struct { +u32 branch_type; +const char *name; +} branch_types[] = { + {0 ,"no branch" }, + {PERF_FLAG_BRANCH |PERF_FLAG_CALL ,"call" }, + {PERF_FLAG_BRANCH |PERF_FLAG_RETURN ,"return" }, + {PERF_FLAG_BRANCH |PERF_FLAG_CONDITIONAL ,"conditional jump" }, + {PERF_FLAG_BRANCH ,"unconditional jump" }, + {PERF_FLAG_BRANCH |PERF_FLAG_CALL |PERF_FLAG_INTERRUPT , +"software interrupt" }, + {PERF_FLAG_BRANCH |PERF_FLAG_RETURN |PERF_FLAG_INTERRUPT , +"return from interrupt" }, + {PERF_FLAG_BRANCH |PERF_FLAG_CALL |PERF_FLAG_SYSCALLRET , +"system call" }, + {PERF_FLAG_BRANCH |PERF_FLAG_RETURN |PERF_FLAG_SYSCALLRET , +"return from system call" }, + {PERF_FLAG_BRANCH |PERF_FLAG_ASYNC ,"asynchronous branch" }, + {PERF_FLAG_BRANCH |PERF_FLAG_CALL |PERF_FLAG_ASYNC | +PERF_FLAG_INTERRUPT ,"hardware interrupt" }, + {PERF_FLAG_BRANCH |PERF_FLAG_TX_ABORT ,"transaction abort" }, + {PERF_FLAG_BRANCH |PERF_FLAG_TRACE_BEGIN ,"trace begin" }, + {PERF_FLAG_BRANCH |PERF_FLAG_TRACE_END ,"trace end" }, + {0 ,NULL } +}; + +int db_export__branch_types (struct db_export *dbe) +{ +int i, err =0 ; + +for (i =0 ; branch_types[i].name ; i++) { + err =db_export__branch_type (dbe, branch_types[i].branch_type, + branch_types[i].name); +if (err) +break ; + } +return err; +}
tools/perf/util/db-export.h +6 -0
@@ -54,6 +54,8 @@ struct db_export { struct machine *machine);int (*export_symbol)(struct db_export *dbe,struct symbol *sym,struct dso *dso); +int (*export_branch_type)(struct db_export *dbe,u32 branch_type, +const char *name);int (*export_sample)(struct db_export *dbe,struct export_sample *es);u64 evsel_last_db_id;u64 machine_last_db_id;
@@ -79,8 +81,12 @@ int db_export__dso(struct db_export *dbe, struct dso *dso, struct machine *machine);int db_export__symbol (struct db_export *dbe,struct symbol *sym,struct dso *dso); +int db_export__branch_type (struct db_export *dbe,u32 branch_type, +const char *name);int db_export__sample (struct db_export *dbe,union perf_event *event,struct perf_sample *sample,struct perf_evsel *evsel,struct thread *thread,struct addr_location *al); +int db_export__branch_types (struct db_export *dbe); +#endif
Signature · 1 line
1.8.3.2
Adrian Hunter
reviewerPATCH 16/24 perf tools: Add branch_type and in_tx to Python export
In reply to Adrian HunterAdd branch_type and in_tx to Python db export and
the export-to-postgresql.py script.
--- tools/perf/scripts/python/export-to-postgresql.py | 32 ++++++++++++++++++---- .../util/scripting-engines/trace-event-python.c | 30 +++++++++++++++++++- 2 files changed, 55 insertions(+), 7 deletions(-)
tools/perf/scripts/python/export-to-postgresql.py +26 -6
@@ -123,6 +123,10 @@ do_query(query, 'CREATE TABLE symbols (' 'sym_end bigint,' 'binding integer,' 'name varchar(2048))') +do_query(query, 'CREATE TABLE branch_types (' + 'id integer NOT NULL,' + 'name varchar(80))') + if branches: do_query(query, 'CREATE TABLE samples (' 'id bigint NOT NULL,'
@@ -139,7 +143,9 @@ if branches: 'to_dso_id bigint,' 'to_symbol_id bigint,' 'to_sym_offset bigint,' - 'to_ip bigint)') + 'to_ip bigint,' + 'branch_type integer,' + 'in_tx boolean)') else: do_query(query, 'CREATE TABLE samples (' 'id bigint NOT NULL,'
@@ -160,7 +166,9 @@ else: 'period bigint,' 'weight bigint,' 'transaction bigint,' - 'data_src bigint)') + 'data_src bigint,' + 'branch_type integer,' + 'in_tx boolean)') do_query(query, 'CREATE VIEW samples_view AS ' 'SELECT '
@@ -178,7 +186,9 @@ do_query(query, 'CREATE VIEW samples_view AS ' 'to_hex(to_ip) AS to_ip_hex,' '(SELECT name FROM symbols WHERE id = to_symbol_id) AS to_symbol,' 'to_sym_offset,' - '(SELECT short_name FROM dsos WHERE id = to_dso_id) AS to_dso_short_name' + '(SELECT short_name FROM dsos WHERE id = to_dso_id) AS to_dso_short_name,' + '(SELECT name FROM branch_types WHERE id = branch_type) AS branch_type_name,' + 'in_tx' ' FROM samples')
@@ -234,6 +244,7 @@ comm_file = open_output_file("comm_table.bin") comm_thread_file = open_output_file("comm_thread_table.bin") dso_file = open_output_file("dso_table.bin") symbol_file = open_output_file("symbol_table.bin") +branch_type_file = open_output_file("branch_type_table.bin") sample_file = open_output_file("sample_table.bin") def trace_begin():
@@ -257,6 +268,7 @@ def trace_end(): copy_output_file(comm_thread_file, "comm_threads") copy_output_file(dso_file, "dsos") copy_output_file(symbol_file, "symbols") + copy_output_file(branch_type_file, "branch_types") copy_output_file(sample_file, "samples") print datetime.datetime.today(), "Removing intermediate files..."
@@ -267,6 +279,7 @@ def trace_end(): remove_output_file(comm_thread_file) remove_output_file(dso_file) remove_output_file(symbol_file) + remove_output_file(branch_type_file) remove_output_file(sample_file) os.rmdir(output_dir_name) print datetime.datetime.today(), "Adding primary keys"
@@ -277,6 +290,7 @@ def trace_end(): do_query(query, 'ALTER TABLE comm_threads ADD PRIMARY KEY (id)') do_query(query, 'ALTER TABLE dsos ADD PRIMARY KEY (id)') do_query(query, 'ALTER TABLE symbols ADD PRIMARY KEY (id)') + do_query(query, 'ALTER TABLE branch_types ADD PRIMARY KEY (id)') do_query(query, 'ALTER TABLE samples ADD PRIMARY KEY (id)') print datetime.datetime.today(), "Adding foreign keys"
@@ -352,9 +366,15 @@ def symbol_table(symbol_id, dso_id, sym_start, sym_end, binding, symbol_name, *x value = struct.pack(fmt, 6, 8, symbol_id, 8, dso_id, 8, sym_start, 8, sym_end, 4, binding, n, symbol_name) symbol_file.write(value) -def sample_table(sample_id, evsel_id, machine_id, thread_id, comm_id, dso_id, symbol_id, sym_offset, ip, time, cpu, to_dso_id, to_symbol_id, to_sym_offset, to_ip, period, weight, transaction, data_src, *x): +def branch_type_table(branch_type, name, *x): + n = len(name) + fmt = "!hiii" + str(n) + "s" + value = struct.pack(fmt, 2, 4, branch_type, n, name) + branch_type_file.write(value) + +def sample_table(sample_id, evsel_id, machine_id, thread_id, comm_id, dso_id, symbol_id, sym_offset, ip, time, cpu, to_dso_id, to_symbol_id, to_sym_offset, to_ip, period, weight, transaction, data_src, branch_type, in_tx, *x): if branches: - value = struct.pack("!hiqiqiqiqiqiqiqiqiqiqiiiqiqiqiq", 15, 8, sample_id, 8, evsel_id, 8, machine_id, 8, thread_id, 8, comm_id, 8, dso_id, 8, symbol_id, 8, sym_offset, 8, ip, 8, time, 4, cpu, 8, to_dso_id, 8, to_symbol_id, 8, to_sym_offset, 8, to_ip) + value = struct.pack("!hiqiqiqiqiqiqiqiqiqiqiiiqiqiqiqiiiB", 17, 8, sample_id, 8, evsel_id, 8, machine_id, 8, thread_id, 8, comm_id, 8, dso_id, 8, symbol_id, 8, sym_offset, 8, ip, 8, time, 4, cpu, 8, to_dso_id, 8, to_symbol_id, 8, to_sym_offset, 8, to_ip, 4, branch_type, 1, in_tx) else: - value = struct.pack("!hiqiqiqiqiqiqiqiqiqiqiiiqiqiqiqiqiqiqiq", 19, 8, sample_id, 8, evsel_id, 8, machine_id, 8, thread_id, 8, comm_id, 8, dso_id, 8, symbol_id, 8, sym_offset, 8, ip, 8, time, 4, cpu, 8, to_dso_id, 8, to_symbol_id, 8, to_sym_offset, 8, to_ip, 8, period, 8, weight, 8, transaction, 8, data_src) + value = struct.pack("!hiqiqiqiqiqiqiqiqiqiqiiiqiqiqiqiqiqiqiqiiiB", 21, 8, sample_id, 8, evsel_id, 8, machine_id, 8, thread_id, 8, comm_id, 8, dso_id, 8, symbol_id, 8, sym_offset, 8, ip, 8, time, 4, cpu, 8, to_dso_id, 8, to_symbol_id, 8, to_sym_offset, 8, to_ip, 8, period, 8, weight, 8, transaction, 8, data_src, 4, branch_type, 1, in_tx) sample_file.write(value)
tools/perf/util/scripting-engines/trace-event-python.c +29 -1
@@ -65,6 +65,7 @@ struct tables {PyObject *comm_thread_handler;PyObject *dso_handler;PyObject *symbol_handler; +PyObject *branch_type_handler;PyObject *sample_handler;bool db_export_mode; };
@@ -662,13 +663,31 @@ static int python_export_symbol(struct db_export *dbe, struct symbol *sym,return 0 ; } +static int python_export_branch_type (struct db_export *dbe,u32 branch_type, +const char *name) +{ + struct tables *tables = container_of(dbe, struct tables, dbe); +PyObject *t; + + t =tuple_new (2 ); + +tuple_set_s32 (t,0 , branch_type); +tuple_set_string (t,1 , name); + +call_object (tables->branch_type_handler, t,"branch_type_table" ); + +Py_DECREF (t); + +return 0 ; +} +static int python_export_sample (struct db_export *dbe,struct export_sample *es) { struct tables *tables = container_of(dbe, struct tables, dbe);PyObject *t; - t =tuple_new (19 ); + t =tuple_new (21 );tuple_set_u64 (t,0 , es->db_id);tuple_set_u64 (t,1 , es->evsel->db_id);
@@ -689,6 +708,8 @@ static int python_export_sample(struct db_export *dbe,tuple_set_u64 (t,16 , es->sample->weight);tuple_set_u64 (t,17 , es->sample->transaction);tuple_set_u64 (t,18 , es->sample->data_src); +tuple_set_s32 (t,19 , es->sample->flags &PERF_BRANCH_MASK ); +tuple_set_s32 (t,20 , !!(es->sample->flags &PERF_FLAG_IN_TX ));call_object (tables->sample_handler, t,"sample_table" );
@@ -855,6 +876,7 @@ static void set_table_handlers(struct tables *tables)SET_TABLE_HANDLER (comm_thread);SET_TABLE_HANDLER (dso);SET_TABLE_HANDLER (symbol); +SET_TABLE_HANDLER (branch_type);SET_TABLE_HANDLER (sample); }
@@ -904,6 +926,12 @@ static int python_start_script(const char *script, int argc, const char **argv)set_table_handlers (tables); +if (tables->db_export_mode) { + err =db_export__branch_types (&tables->dbe); +if (err) + goto error; + } +return err; error:Py_Finalize ();
Signature · 1 line
1.8.3.2
Adrian Hunter
reviewerPATCH 17/24 perf tools: Enhance the thread stack to output call/return data
In reply to Adrian HunterEnhance the thread stack to output detailed information about paired calls and returns.
--- tools/perf/util/thread-stack.c | 547 ++++++++++++++++++++++++++++++++++++++++- tools/perf/util/thread-stack.h | 47 ++++ 2 files changed, 590 insertions(+), 4 deletions(-)
tools/perf/util/thread-stack.c +543 -4
@@ -13,15 +13,48 @@* */ +#include <linux/rbtree.h> +#include <linux/list.h> #include "thread.h" #include "event.h" +#include "machine.h" #include "util.h" +#include "debug.h" +#include "symbol.h" +#include "comm.h" #include "thread-stack.h" -#define STACK_GROWTH 4096 +#define CALL_PATH_BLOCK_SHIFT 8 +#define CALL_PATH_BLOCK_SIZE (1 << CALL_PATH_BLOCK_SHIFT) +#define CALL_PATH_BLOCK_MASK (CALL_PATH_BLOCK_SIZE - 1) + +struct call_path_block { +struct call_path cp[CALL_PATH_BLOCK_SIZE ]; +struct list_head node; +}; + +struct call_path_root { +struct call_path call_path; +struct list_head blocks; +size_t next; +size_t sz; +}; + +struct call_return_processor { +struct call_path_root *cpr; +int (*process)(struct call_return *cr,void *data); +void *data; +}; + +#define STACK_GROWTH 2048struct thread_stack_entry {u64 ret_addr; +u64 timestamp; +u64 ref; +u64 branch_count; +struct call_path *cp; +bool no_call; };struct thread_stack {
@@ -29,6 +62,11 @@ struct thread_stack {size_t cnt;size_t sz;u64 trace_nr; +u64 branch_count; +u64 kernel_start; +u64 last_time; +struct call_return_processor *crp; +struct comm *comm; };static void thread_stack__grow (struct thread_stack *ts)
@@ -45,7 +83,8 @@ static void thread_stack__grow(struct thread_stack *ts) } } -static struct thread_stack *thread_stack__new (void ) +static struct thread_stack *thread_stack__new (struct thread *thread, +struct call_return_processor *crp) {struct thread_stack *ts;
@@ -59,6 +98,12 @@ static struct thread_stack *thread_stack__new(void)return NULL ; } +if (thread->mg && thread->mg->machine) + ts->kernel_start =machine__kernel_start (thread->mg->machine); +else + ts->kernel_start =1ULL <<63 ; + ts->crp = crp; +return ts; }
@@ -92,6 +137,64 @@ static void thread_stack__pop(struct thread_stack *ts, u64 ret_addr) } } +static bool thread_stack__in_kernel (struct thread_stack *ts) +{ +if (!ts->cnt) +return false; + +return ts->stack[ts->cnt -1 ].cp->in_kernel; +} + +static int thread_stack__call_return (struct thread *thread, +struct thread_stack *ts,size_t idx, +u64 timestamp,u64 ref,bool no_return) +{ +struct call_return_processor *crp = ts->crp; +struct thread_stack_entry *tse; +struct call_return cr = { + .thread = thread, + .comm = ts->comm, + .db_id =0 , + }; + + tse = &ts->stack[idx]; + cr.cp = tse->cp; + cr.call_time = tse->timestamp; + cr.return_time = timestamp; + cr.branch_count = ts->branch_count - tse->branch_count; + cr.call_ref = tse->ref; + cr.return_ref = ref; +if (tse->no_call) + cr.flags |=CALL_RETURN_NO_CALL ; +if (no_return) + cr.flags |=CALL_RETURN_NO_RETURN ; + +return crp->process (&cr, crp->data); +} + +static int thread_stack__flush (struct thread *thread,struct thread_stack *ts) +{ +struct call_return_processor *crp = ts->crp; +int err; + +if (!crp) { + ts->cnt =0 ; +return 0 ; + } + +while (ts->cnt) { + err =thread_stack__call_return (thread, ts, --ts->cnt, + ts->last_time,0 , true); +if (err) { +pr_err ("Error flushing thread stack!\n" ); + ts->cnt =0 ; +return err; + } + } + +return 0 ; +} +void thread_stack__event (struct thread *thread,u32 flags,u64 from_ip,u64 to_ip,u16 insn_len,u64 trace_nr) {
@@ -99,17 +202,22 @@ void thread_stack__event(struct thread *thread, u32 flags, u64 from_ip,return ;if (!thread->ts) { - thread->ts =thread_stack__new (); + thread->ts =thread_stack__new (thread,NULL );if (!thread->ts)return ; thread->ts->trace_nr = trace_nr; }if (trace_nr != thread->ts->trace_nr) { +if (thread->ts->trace_nr) +thread_stack__flush (thread, thread->ts); thread->ts->trace_nr = trace_nr; - thread->ts->cnt =0 ; } +/* Stop here if thread_stack__process() is in use */ +if (thread->ts->crp) +return ; +if (flags &PERF_FLAG_CALL ) {u64 ret_addr;
@@ -126,9 +234,22 @@ void thread_stack__event(struct thread *thread, u32 flags, u64 from_ip, } } +void thread_stack__set_trace_nr (struct thread *thread,u64 trace_nr) +{ +if (!thread || !thread->ts) +return ; + +if (trace_nr != thread->ts->trace_nr) { +if (thread->ts->trace_nr) +thread_stack__flush (thread, thread->ts); + thread->ts->trace_nr = trace_nr; + } +} +void thread_stack__free (struct thread *thread) {if (thread->ts) { +thread_stack__flush (thread, thread->ts);zfree (&thread->ts->stack);zfree (&thread->ts); }
@@ -149,3 +270,421 @@ void thread_stack__sample(struct thread *thread, struct ip_callchain *chain,for (i =1 ; i < chain->nr; i++) chain->ips[i] = thread->ts->stack[thread->ts->cnt - i].ret_addr; } + +static void call_path__init (struct call_path *cp,struct call_path *parent, +struct symbol *sym,u64 ip,bool in_kernel) +{ + cp->parent = parent; + cp->sym = sym; + cp->ip = sym ?0 : ip; + cp->db_id =0 ; + cp->in_kernel = in_kernel; +RB_CLEAR_NODE (&cp->rb_node); + cp->children =RB_ROOT ; +} + +static struct call_path_root *call_path_root__new (void ) +{ +struct call_path_root *cpr; + + cpr =zalloc (sizeof (struct call_path_root )); +if (!cpr) +return NULL ; +call_path__init (&cpr->call_path,NULL ,NULL ,0 , false); +INIT_LIST_HEAD (&cpr->blocks); +return cpr; +} + +static void call_path_root__free (struct call_path_root *cpr) +{ +struct call_path_block *pos, *n; + +list_for_each_entry_safe (pos, n, &cpr->blocks, node) { +list_del (&pos->node); +free (pos); + } +free (cpr); +} + +static struct call_path *call_path__new (struct call_path_root *cpr, +struct call_path *parent, +struct symbol *sym,u64 ip, +bool in_kernel) +{ +struct call_path_block *cpb; +struct call_path *cp; +size_t n; + +if (cpr->next < cpr->sz) { + cpb = list_last_entry(&cpr->blocks, struct call_path_block, + node); + }else { + cpb =zalloc (sizeof (struct call_path_block )); +if (!cpb) +return NULL ; +list_add_tail (&cpb->node, &cpr->blocks); + cpr->sz +=CALL_PATH_BLOCK_SIZE ; + } + + n = cpr->next++ &CALL_PATH_BLOCK_MASK ; + cp = &cpb->cp[n]; + +call_path__init (cp, parent, sym, ip, in_kernel); + +return cp; +} + +static struct call_path *call_path__findnew (struct call_path_root *cpr, +struct call_path *parent, +struct symbol *sym,u64 ip,u64 ks) +{ +struct rb_node **p; +struct rb_node *node_parent =NULL ; +struct call_path *cp; +bool in_kernel = ip >= ks; + +if (sym) + ip =0 ; + +if (!parent) +return call_path__new (cpr, parent, sym, ip, in_kernel); + + p = &parent->children.rb_node; +while (*p !=NULL ) { + node_parent = *p; + cp = rb_entry(node_parent, struct call_path, rb_node); + +if (cp->sym == sym && cp->ip == ip) +return cp; + +if (sym < cp->sym || (sym == cp->sym && ip < cp->ip)) + p = &(*p)->rb_left; +else + p = &(*p)->rb_right; + } + + cp =call_path__new (cpr, parent, sym, ip, in_kernel); +if (!cp) +return NULL ; + +rb_link_node (&cp->rb_node, node_parent, p); +rb_insert_color (&cp->rb_node, &parent->children); + +return cp; +} + +struct call_return_processor * +call_return_processor__new (int (*process)(struct call_return *cr,void *data), +void *data) +{ +struct call_return_processor *crp; + + crp =zalloc (sizeof (struct call_return_processor )); +if (!crp) +return NULL ; + crp->cpr =call_path_root__new (); +if (!crp->cpr) + goto out_free; + crp->process = process; + crp->data = data; +return crp; + +out_free: +free (crp); +return NULL ; +} + +void call_return_processor__free (struct call_return_processor *crp) +{ +if (crp) { +call_path_root__free (crp->cpr); +free (crp); + } +} + +static int thread_stack__push_cp (struct thread_stack *ts,u64 ret_addr, +u64 timestamp,u64 ref,struct call_path *cp, +bool no_call) +{ +struct thread_stack_entry *tse; + +if (ts->cnt == ts->sz) { +thread_stack__grow (ts); +if (ts->cnt == ts->sz) +return -ENOMEM ; + } + + tse = &ts->stack[ts->cnt++]; + tse->ret_addr = ret_addr; + tse->timestamp = timestamp; + tse->ref = ref; + tse->branch_count = ts->branch_count; + tse->cp = cp; + tse->no_call = no_call; + +return 0 ; +} + +static int thread_stack__pop_cp (struct thread *thread,struct thread_stack *ts, +u64 ret_addr,u64 timestamp,u64 ref, +struct symbol *sym) +{ +int err; + +if (!ts->cnt) +return 1 ; + +if (ts->cnt ==1 ) { +struct thread_stack_entry *tse = &ts->stack[0 ]; + +if (tse->cp->sym == sym) +return thread_stack__call_return (thread, ts, --ts->cnt, + timestamp, ref, false); + } + +if (ts->stack[ts->cnt -1 ].ret_addr == ret_addr) { +return thread_stack__call_return (thread, ts, --ts->cnt, + timestamp, ref, false); + }else { +size_t i = ts->cnt -1 ; + +while (i--) { +if (ts->stack[i].ret_addr != ret_addr) +continue ; + i +=1 ; +while (ts->cnt > i) { + err =thread_stack__call_return (thread, ts, + --ts->cnt, + timestamp, ref, + true); +if (err) +return err; + } +return thread_stack__call_return (thread, ts, --ts->cnt, + timestamp, ref, false); + } + } + +return 1 ; +} + +static int thread_stack__bottom (struct thread *thread,struct thread_stack *ts, +struct perf_sample *sample, +struct addr_location *from_al, +struct addr_location *to_al,u64 ref) +{ +struct call_path_root *cpr = ts->crp->cpr; +struct call_path *cp; +struct symbol *sym; +u64 ip; + +if (sample->ip) { + ip = sample->ip; + sym = from_al->sym; + }else if (sample->addr) { + ip = sample->addr; + sym = to_al->sym; + }else { +return 0 ; + } + + cp =call_path__findnew (cpr, &cpr->call_path, sym, ip, + ts->kernel_start); +if (!cp) +return -ENOMEM ; + +return thread_stack__push_cp (thread->ts, ip, sample->time, ref, cp, + true); +} + +static int thread_stack__no_call_return (struct thread *thread, +struct thread_stack *ts, +struct perf_sample *sample, +struct addr_location *from_al, +struct addr_location *to_al,u64 ref) +{ +struct call_path_root *cpr = ts->crp->cpr; +struct call_path *cp, *parent; +u64 ks = ts->kernel_start; +int err; + +if (sample->ip >= ks && sample->addr < ks) { +/* Return to userspace, so pop all kernel addresses */ +while (thread_stack__in_kernel (ts)) { + err =thread_stack__call_return (thread, ts, --ts->cnt, + sample->time, ref, + true); +if (err) +return err; + } + +/* If the stack is empty, push the userspace address */ +if (!ts->cnt) { + cp =call_path__findnew (cpr, &cpr->call_path, + to_al->sym, sample->addr, + ts->kernel_start); +if (!cp) +return -ENOMEM ; +return thread_stack__push_cp (ts,0 , sample->time, ref, + cp, true); + } + }else if (thread_stack__in_kernel (ts) && sample->ip < ks) { +/* Return to userspace, so pop all kernel addresses */ +while (thread_stack__in_kernel (ts)) { + err =thread_stack__call_return (thread, ts, --ts->cnt, + sample->time, ref, + true); +if (err) +return err; + } + } + +if (ts->cnt) + parent = ts->stack[ts->cnt -1 ].cp; +else + parent = &cpr->call_path; + +/* This 'return' had no 'call', so push and pop top of stack */ + cp =call_path__findnew (cpr, parent, from_al->sym, sample->ip, + ts->kernel_start); +if (!cp) +return -ENOMEM ; + + err =thread_stack__push_cp (ts, sample->addr, sample->time, ref, cp, + true); +if (err) +return err; + +return thread_stack__pop_cp (thread, ts, sample->addr, sample->time, ref, + to_al->sym); +} + +static int thread_stack__trace_begin (struct thread *thread, +struct thread_stack *ts,u64 timestamp, +u64 ref) +{ +struct thread_stack_entry *tse; +int err; + +if (!ts->cnt) +return 0 ; + +/* Pop trace end */ + tse = &ts->stack[ts->cnt -1 ]; +if (tse->cp->sym ==NULL && tse->cp->ip ==0 ) { + err =thread_stack__call_return (thread, ts, --ts->cnt, + timestamp, ref, false); +if (err) +return err; + } + +return 0 ; +} + +static int thread_stack__trace_end (struct thread_stack *ts, +struct perf_sample *sample,u64 ref) +{ +struct call_path_root *cpr = ts->crp->cpr; +struct call_path *cp; +u64 ret_addr; + +/* No point having 'trace end' on the bottom of the stack */ +if (!ts->cnt || (ts->cnt ==1 && ts->stack[0 ].ref == ref)) +return 0 ; + + cp =call_path__findnew (cpr, ts->stack[ts->cnt -1 ].cp,NULL ,0 , + ts->kernel_start); +if (!cp) +return -ENOMEM ; + + ret_addr = sample->ip + sample->insn_len; + +return thread_stack__push_cp (ts, ret_addr, sample->time, ref, cp, + false); +} + +int thread_stack__process (struct thread *thread,struct comm *comm, +struct perf_sample *sample, +struct addr_location *from_al, +struct addr_location *to_al,u64 ref, +struct call_return_processor *crp) +{ +struct thread_stack *ts = thread->ts; +int err =0 ; + +if (ts) { +if (!ts->crp) { +/* Supersede thread_stack__event() */ +thread_stack__free (thread); + thread->ts =thread_stack__new (thread, crp); +if (!thread->ts) +return -ENOMEM ; + ts = thread->ts; + ts->comm = comm; + } + }else { + thread->ts =thread_stack__new (thread, crp); +if (!thread->ts) +return -ENOMEM ; + ts = thread->ts; + ts->comm = comm; + } + +/* Flush stack on exec */ +if (ts->comm != comm && thread->pid_ == thread->tid) { + err =thread_stack__flush (thread, ts); +if (err) +return err; + ts->comm = comm; + } + +/* If the stack is empty, put the current symbol on the stack */ +if (!ts->cnt) { + err =thread_stack__bottom (thread, ts, sample, from_al, to_al, + ref); +if (err) +return err; + } + + ts->branch_count +=1 ; + ts->last_time = sample->time; + +if (sample->flags &PERF_FLAG_CALL ) { +struct call_path_root *cpr = ts->crp->cpr; +struct call_path *cp; +u64 ret_addr; + +if (!sample->ip || !sample->addr) +return 0 ; + + ret_addr = sample->ip + sample->insn_len; +if (ret_addr == sample->addr) +return 0 ;/* Zero-length calls are excluded */ + + cp =call_path__findnew (cpr, ts->stack[ts->cnt -1 ].cp, + to_al->sym, sample->addr, + ts->kernel_start); +if (!cp) +return -ENOMEM ; + err =thread_stack__push_cp (ts, ret_addr, sample->time, ref, + cp, false); + }else if (sample->flags &PERF_FLAG_RETURN ) { +if (!sample->ip || !sample->addr) +return 0 ; + + err =thread_stack__pop_cp (thread, ts, sample->addr, + sample->time, ref, from_al->sym); +if (err) { +if (err <0 ) +return err; + err =thread_stack__no_call_return (thread, ts, sample, + from_al, to_al, ref); + } + }else if (sample->flags &PERF_FLAG_TRACE_BEGIN ) { + err =thread_stack__trace_begin (thread, ts, sample->time, ref); + }else if (sample->flags &PERF_FLAG_TRACE_END ) { + err =thread_stack__trace_end (ts, sample, ref); + } + +return err; +}
tools/perf/util/thread-stack.h +47 -0
61 lines not on this page · raw message
Signature · 1 line
1.8.3.2
Adrian Hunter
reviewerPATCH 18/24 perf tools: Add call information to the database export API
In reply to Adrian HunterMake it possible for the database export API to use the enhanced thread stack and export detailed information about paired calls and returns.
--- tools/perf/util/db-export.c | 52 ++++++++++++++++++++++++++++++++++++++++++++- tools/perf/util/db-export.h | 12 +++++++++++ 2 files changed, 63 insertions(+), 1 deletion(-)
tools/perf/util/db-export.c +51 -1
@@ -21,6 +21,7 @@#include "comm.h" #include "symbol.h" #include "event.h" +#include "thread-stack.h" #include "db-export.h" int db_export__init (struct db_export *dbe)
@@ -29,8 +30,10 @@ int db_export__init(struct db_export *dbe)return 0 ; } -void db_export__exit (struct db_export *dbe__maybe_unused ) +void db_export__exit (struct db_export *dbe) { +call_return_processor__free (dbe->crp); + dbe->crp =NULL ; }int db_export__evsel (struct db_export *dbe,struct perf_evsel *evsel)
@@ -268,6 +271,13 @@ int db_export__sample(struct db_export *dbe, union perf_event *event, &es.addr_sym_db_id, &es.addr_offset);if (err)return err; +if (dbe->crp) { + err =thread_stack__process (thread, comm, sample, al, + &addr_al, es.db_id, + dbe->crp); +if (err) +return err; + } }if (dbe->export_sample)
@@ -314,3 +324,43 @@ int db_export__branch_types(struct db_export *dbe) }return err; } + +int db_export__call_path (struct db_export *dbe,struct call_path *cp) +{ +int err; + +if (cp->db_id) +return 0 ; + +if (cp->parent) { + err =db_export__call_path (dbe, cp->parent); +if (err) +return err; + } + + cp->db_id = ++dbe->call_path_last_db_id; + +if (dbe->export_call_path) +return dbe->export_call_path (dbe, cp); + +return 0 ; +} + +int db_export__call_return (struct db_export *dbe,struct call_return *cr) +{ +int err; + +if (cr->db_id) +return 0 ; + + err =db_export__call_path (dbe, cr->cp); +if (err) +return err; + + cr->db_id = ++dbe->call_return_last_db_id; + +if (dbe->export_call_return) +return dbe->export_call_return (dbe, cr); + +return 0 ; +}
tools/perf/util/db-export.h +12 -0
@@ -25,6 +25,9 @@ struct comm;struct dso ;struct perf_sample ;struct addr_location ; +struct call_return_processor ; +struct call_path ; +struct call_return ;struct export_sample {union perf_event *event;
@@ -57,6 +60,10 @@ struct db_export {int (*export_branch_type)(struct db_export *dbe,u32 branch_type,const char *name);int (*export_sample)(struct db_export *dbe,struct export_sample *es); +int (*export_call_path)(struct db_export *dbe,struct call_path *cp); +int (*export_call_return)(struct db_export *dbe, +struct call_return *cr); +struct call_return_processor *crp;u64 evsel_last_db_id;u64 machine_last_db_id;u64 thread_last_db_id;
@@ -65,6 +72,8 @@ struct db_export {u64 dso_last_db_id;u64 symbol_last_db_id;u64 sample_last_db_id; +u64 call_path_last_db_id; +u64 call_return_last_db_id; };int db_export__init (struct db_export *dbe);
@@ -89,4 +98,7 @@ int db_export__sample(struct db_export *dbe, union perf_event *event,int db_export__branch_types (struct db_export *dbe); +int db_export__call_path (struct db_export *dbe,struct call_path *cp); +int db_export__call_return (struct db_export *dbe,struct call_return *cr); +#endif
Signature · 1 line
1.8.3.2
Adrian Hunter
reviewerPATCH 19/24 perf tools: Add call information to Python export
In reply to Adrian HunterAdd the ability to export detailed information about paired calls and returns to Python db export and the export-to-postgresql.py script.
--- .../scripts/python/bin/export-to-postgresql-report | 15 ++-- tools/perf/scripts/python/export-to-postgresql.py | 66 ++++++++++++++++- .../util/scripting-engines/trace-event-python.c | 84 +++++++++++++++++++++- 3 files changed, 158 insertions(+), 7 deletions(-)
tools/perf/scripts/python/bin/export-to-postgresql-report +10 -5
@@ -1,6 +1,6 @@ #!/bin/bash # description: export perf data to a postgresql database -# args: [database name] [columns] +# args: [database name] [columns] [calls] n_args=0 for i in "$@" do
@@ -9,11 +9,16 @@ do fi n_args=$(( $n_args + 1 )) done -if [ "$n_args" -gt 2 ] ; then - echo "usage: export-to-postgresql-report [database name] [columns]" +if [ "$n_args" -gt 3 ] ; then + echo "usage: export-to-postgresql-report [database name] [columns] [calls]" exit fi -if [ "$n_args" -gt 1 ] ; then +if [ "$n_args" -gt 2 ] ; then + dbname=$1 + columns=$2 + calls=$3 + shift 3 +elif [ "$n_args" -gt 1 ] ; then dbname=$1 columns=$2 shift 2
@@ -21,4 +26,4 @@ elif [ "$n_args" -gt 0 ] ; then dbname=$1 shift fi -perf script $@ -s "$PERF_EXEC_PATH"/scripts/python/export-to-postgresql.py $dbname $columns +perf script $@ -s "$PERF_EXEC_PATH"/scripts/python/export-to-postgresql.py $dbname $columns $calls
tools/perf/scripts/python/export-to-postgresql.py +65 -1
126 lines not on this page · raw message
tools/perf/util/scripting-engines/trace-event-python.c +83 -1
120 lines not on this page · raw message
Signature · 1 line
1.8.3.2
Adrian Hunter
reviewerPATCH 20/24 perf tools: Defer export of comms that were not 'set'
In reply to Adrian HunterTracing for a workload begins before the comm event is seen, which results in the initial comm having a string of the form ":<pid>" (e.g. ":12345"). In order to export the correct string, defer the export until the new script 'flush' callback.
--- tools/perf/util/db-export.c | 62 +++++++++++++++++++++- tools/perf/util/db-export.h | 3 ++ .../util/scripting-engines/trace-event-python.c | 4 +- 3 files changed, 67 insertions(+), 2 deletions(-)
tools/perf/util/db-export.c +61 -1
@@ -21,17 +21,74 @@#include "comm.h" #include "symbol.h" #include "event.h" +#include "util.h" #include "thread-stack.h" #include "db-export.h" +struct deferred_export { +struct list_head node; +struct comm *comm; +}; + +static int db_export__deferred (struct db_export *dbe) +{ +struct deferred_export *de; +int err; + +while (!list_empty (&dbe->deferred)) { + de = list_entry(dbe->deferred.next, struct deferred_export, + node); + err = dbe->export_comm (dbe, de->comm); +list_del (&de->node); +free (de); +if (err) +return err; + } + +return 0 ; +} + +static void db_export__free_deferred (struct db_export *dbe) +{ +struct deferred_export *de; + +while (!list_empty (&dbe->deferred)) { + de = list_entry(dbe->deferred.next, struct deferred_export, + node); +list_del (&de->node); +free (de); + } +} + +static int db_export__defer_comm (struct db_export *dbe,struct comm *comm) +{ +struct deferred_export *de; + + de =zalloc (sizeof (struct deferred_export )); +if (!de) +return -ENOMEM ; + + de->comm = comm; +list_add_tail (&de->node, &dbe->deferred); + +return 0 ; +} +int db_export__init (struct db_export *dbe) {memset (dbe,0 ,sizeof (struct db_export )); +INIT_LIST_HEAD (&dbe->deferred);return 0 ; } +int db_export__flush (struct db_export *dbe) +{ +return db_export__deferred (dbe); +} +void db_export__exit (struct db_export *dbe) { +db_export__free_deferred (dbe);call_return_processor__free (dbe->crp); dbe->crp =NULL ; }
@@ -115,7 +172,10 @@ int db_export__comm(struct db_export *dbe, struct comm *comm, comm->db_id = ++dbe->comm_last_db_id;if (dbe->export_comm) { - err = dbe->export_comm (dbe, comm); +if (main_thread->comm_set) + err = dbe->export_comm (dbe, comm); +else + err =db_export__defer_comm (dbe, comm);if (err)return err; }
tools/perf/util/db-export.h +3 -0
@@ -17,6 +17,7 @@#define __PERF_DB_EXPORT_H#include <linux/types.h> +#include <linux/list.h> struct perf_evsel ;struct machine ;
@@ -74,9 +75,11 @@ struct db_export {u64 sample_last_db_id;u64 call_path_last_db_id;u64 call_return_last_db_id; +struct list_head deferred; };int db_export__init (struct db_export *dbe); +int db_export__flush (struct db_export *dbe);void db_export__exit (struct db_export *dbe);int db_export__evsel (struct db_export *dbe,struct perf_evsel *evsel);int db_export__machine (struct db_export *dbe,struct machine *machine);
tools/perf/util/scripting-engines/trace-event-python.c +3 -1
@@ -1024,7 +1024,9 @@ error:static int python_flush_script (void ) { -return 0 ; +struct tables *tables = &tables_global; + +return db_export__flush (&tables->dbe); }/*
Signature · 1 line
1.8.3.2
Adrian Hunter
reviewerPATCH 21/24 perf tools: Add perf-with-kcore script
In reply to Adrian HunterDecoding an Intel PT trace of the kernel requires an accurate kernel object image. This is provided by making a copy of kcore. However the copy needs to be made under the same conditions as the original recording, and then it needs to be associated with the perf.data file. The perf-with-kcore script does that.
The script also checks the permissions on the buildid cache and can be used to fix them. That is needed for distributions where root does not have a home directory and consequently writes to the same buildid cache as the user, resulting in cached files that the user does not have access to.
Example:
$ ./perf-with-kcore
Usage: perf-with-kcore <perf sub-command> <perf.data directory> [<sub-command options> [ -- <workload>]]
<perf sub-command> can be record, script, report or inject
or: perf-with-kcore fix_buildid_cache_permissions
$ ./perf-with-kcore record pt_uname -e intel_pt// -- uname
Recording
Using /home/ahunter/bin/perf
perf version 3.15.rc3.g4549ba
/home/ahunter/bin/perf record -o pt_uname/perf.data -e intel_pt// -- uname
Linux
[ perf record: Woken up 3 times to write data ]
[ perf record: Captured and wrote 0.023 MB pt_uname/perf.data ]
Copying kcore
[sudo] password for ahunter:
Done
$ tools/perf/perf-with-kcore.sh script pt_uname | head
Using /home/ahunter/bin/perf
perf version 3.15.rc3.g4549ba
/home/ahunter/bin/perf script -i pt_uname/perf.data --kallsyms=pt_uname/kcore_dir/kallsyms
swapper 0 [002] 161533.969666: sched:sched_switch: swapper/2:0 [120] R ==> perf:11316 [120]
:11315 11315 [003] 161533.969704: sched:sched_switch: perf:11315 [120] S ==> swapper/3:0 [120]
:11316 11316 [002] 161533.969783: sched:sched_switch: perf:11316 [120] R ==> migration/2:33 [0]
:33 33 [002] 161533.969791: sched:sched_switch: migration/2:33 [0] S ==> swapper/2:0 [120]
swapper 0 [003] 161533.969792: sched:sched_switch: swapper/3:0 [120] R ==> perf:11316 [120]
:11316 11316 [003] 161533.970062: branches: 0 [unknown] ([unknown]) => ffffffff810532fa native_write_msr_safe ([kernel.kallsyms])
:11316 11316 [003] 161533.970062: branches: ffffffff810532fd native_write_msr_safe ([kernel.kallsyms]) => ffffffff81035b31 pt_config_start ([kernel.kallsyms])--- tools/perf/.gitignore | 1 + tools/perf/Makefile.perf | 5 +- tools/perf/perf-with-kcore.sh | 259 ++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 264 insertions(+), 1 deletion(-) create mode 100644 tools/perf/perf-with-kcore.sh
tools/perf/.gitignore +1 -0
@@ -15,6 +15,7 @@ perf.data perf.data.old output.svg perf-archive +perf-with-kcore tags TAGS cscope*
tools/perf/Makefile.perf +4 -1
@@ -126,6 +126,7 @@ PYRF_OBJS = SCRIPT_SH = SCRIPT_SH += perf-archive.sh +SCRIPT_SH += perf-with-kcore.sh grep-libs = $(filter -l%,$(1)) strip-libs = $(filter-out -l%,$(1))
@@ -882,6 +883,8 @@ install-bin: all install-gtk $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)' $(call QUIET_INSTALL, perf-archive) \ $(INSTALL) $(OUTPUT)perf-archive -t '$(DESTDIR_SQ)$(perfexec_instdir_SQ)' + $(call QUIET_INSTALL, perf-with-kcore) \ + $(INSTALL) $(OUTPUT)perf-with-kcore -t '$(DESTDIR_SQ)$(perfexec_instdir_SQ)' ifndef NO_LIBPERL $(call QUIET_INSTALL, perl-scripts) \ $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)/scripts/perl/Perf-Trace-Util/lib/Perf/Trace'; \
@@ -927,7 +930,7 @@ config-clean: @$(MAKE) -C config/feature-checks clean >/dev/null clean: $(LIBTRACEEVENT)-clean $(LIBAPIKFS)-clean config-clean - $(call QUIET_CLEAN, core-objs) $(RM) $(LIB_OBJS) $(BUILTIN_OBJS) $(LIB_FILE) $(OUTPUT)perf-archive $(OUTPUT)perf.o $(LANG_BINDINGS) $(GTK_OBJS) + $(call QUIET_CLEAN, core-objs) $(RM) $(LIB_OBJS) $(BUILTIN_OBJS) $(LIB_FILE) $(OUTPUT)perf-archive $(OUTPUT)perf-with-kcore $(OUTPUT)perf.o $(LANG_BINDINGS) $(GTK_OBJS) $(call QUIET_CLEAN, core-progs) $(RM) $(ALL_PROGRAMS) perf $(call QUIET_CLEAN, core-gen) $(RM) *.spec *.pyc *.pyo */*.pyc */*.pyo $(OUTPUT)common-cmds.h TAGS tags cscope* $(OUTPUT)PERF-VERSION-FILE $(OUTPUT)PERF-CFLAGS $(OUTPUT)PERF-FEATURES $(OUTPUT)util/*-bison* $(OUTPUT)util/*-flex* $(QUIET_SUBDIR0)Documentation $(QUIET_SUBDIR1) clean
tools/perf/perf-with-kcore.sh new file +259 -0
259 lines not on this page · raw message
Signature · 1 line
1.8.3.2
Adrian Hunter
reviewerPATCH 22/24 perf tools: Build programs to copy 32-bit compatibility VDSOs
In reply to Adrian Hunterperf tools copy VDSO out of memory. However, on 64-bit machines there may be 32-bit compatibility VDOs also. To copy those requires separate 32-bit executables. This patch adds to the build additional programs perf-read-vdso32 and perf-read-vdsox32 for 32-bit and x32 respectively.
--- tools/perf/Makefile.perf | 38 ++++++++++++++++++++++--- tools/perf/config/Makefile | 25 +++++++++++++++- tools/perf/config/Makefile.arch | 8 ++++++ tools/perf/config/feature-checks/Makefile | 10 ++++++- tools/perf/config/feature-checks/test-compile.c | 4 +++ tools/perf/perf-read-vdso.c | 34 ++++++++++++++++++++++ tools/perf/util/find-vdso-map.c | 30 +++++++++++++++++++ tools/perf/util/vdso.c | 37 ++++-------------------- 8 files changed, 149 insertions(+), 37 deletions(-) create mode 100644 tools/perf/config/feature-checks/test-compile.c create mode 100644 tools/perf/perf-read-vdso.c create mode 100644 tools/perf/util/find-vdso-map.c
tools/perf/Makefile.perf +34 -4
@@ -60,6 +60,12 @@ include config/utilities.mak # # Define NO_LIBDW_DWARF_UNWIND if you do not want libdw support # for dwarf backtrace post unwind. +# +# Define NO_PERF_READ_VDSO32 if you do not want to build perf-read-vdso32 +# for reading the 32-bit compatibility VDSO in 64-bit mode +# +# Define NO_PERF_READ_VDSOX32 if you do not want to build perf-read-vdsox32 +# for reading the x32 mode 32-bit compatibility VDSO in 64-bit mode ifeq ($(srctree),) srctree := $(patsubst %/,%,$(dir $(shell pwd)))
@@ -171,11 +177,16 @@ $(OUTPUT)python/perf.so: $(PYTHON_EXT_SRCS) $(PYTHON_EXT_DEPS) SCRIPTS = $(patsubst %.sh,%,$(SCRIPT_SH)) -# -# Single 'perf' binary right now: -# PROGRAMS += $(OUTPUT)perf +ifndef NO_PERF_READ_VDSO32 +PROGRAMS += $(OUTPUT)perf-read-vdso32 +endif + +ifndef NO_PERF_READ_VDSOX32 +PROGRAMS += $(OUTPUT)perf-read-vdsox32 +endif + # what 'all' will build and 'install' will install, in perfexecdir ALL_PROGRAMS = $(PROGRAMS) $(SCRIPTS)
@@ -306,6 +317,7 @@ LIB_H += util/data.h LIB_H += util/kvm-stat.h LIB_H += util/thread-stack.h LIB_H += util/db-export.h +LIB_H += util/find-vdso-map.c LIB_OBJS += $(OUTPUT)util/abspath.o LIB_OBJS += $(OUTPUT)util/alias.o
@@ -735,6 +747,16 @@ $(OUTPUT)scripts/python/Perf-Trace-Util/Context.o: scripts/python/Perf-Trace-Uti $(OUTPUT)perf-%: %.o $(PERFLIBS) $(QUIET_LINK)$(CC) $(CFLAGS) -o $@ $(LDFLAGS) $(filter %.o,$^) $(LIBS) +ifndef NO_PERF_READ_VDSO32 +$(OUTPUT)perf-read-vdso32: perf-read-vdso.c util/find-vdso-map.c + $(QUIET_CC)$(CC) -m32 $(filter -static,$(LDFLAGS)) -Wall -Werror -o $@ perf-read-vdso.c +endif + +ifndef NO_PERF_READ_VDSOX32 +$(OUTPUT)perf-read-vdsox32: perf-read-vdso.c util/find-vdso-map.c + $(QUIET_CC)$(CC) -mx32 $(filter -static,$(LDFLAGS)) -Wall -Werror -o $@ perf-read-vdso.c +endif + $(LIB_OBJS) $(BUILTIN_OBJS): $(LIB_H) $(patsubst perf-%,%.o,$(PROGRAMS)): $(LIB_H) $(wildcard */*.h)
@@ -879,6 +901,14 @@ install-bin: all install-gtk $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(bindir_SQ)'; \ $(INSTALL) $(OUTPUT)perf '$(DESTDIR_SQ)$(bindir_SQ)'; \ $(LN) '$(DESTDIR_SQ)$(bindir_SQ)/perf' '$(DESTDIR_SQ)$(bindir_SQ)/trace' +ifndef NO_PERF_READ_VDSO32 + $(call QUIET_INSTALL, perf-read-vdso32) \ + $(INSTALL) $(OUTPUT)perf-read-vdso32 '$(DESTDIR_SQ)$(bindir_SQ)'; +endif +ifndef NO_PERF_READ_VDSOX32 + $(call QUIET_INSTALL, perf-read-vdsox32) \ + $(INSTALL) $(OUTPUT)perf-read-vdsox32 '$(DESTDIR_SQ)$(bindir_SQ)'; +endif $(call QUIET_INSTALL, libexec) \ $(INSTALL) -d -m 755 '$(DESTDIR_SQ)$(perfexec_instdir_SQ)' $(call QUIET_INSTALL, perf-archive) \
@@ -931,7 +961,7 @@ config-clean: clean: $(LIBTRACEEVENT)-clean $(LIBAPIKFS)-clean config-clean $(call QUIET_CLEAN, core-objs) $(RM) $(LIB_OBJS) $(BUILTIN_OBJS) $(LIB_FILE) $(OUTPUT)perf-archive $(OUTPUT)perf-with-kcore $(OUTPUT)perf.o $(LANG_BINDINGS) $(GTK_OBJS) - $(call QUIET_CLEAN, core-progs) $(RM) $(ALL_PROGRAMS) perf + $(call QUIET_CLEAN, core-progs) $(RM) $(ALL_PROGRAMS) perf perf-read-vdso32 perf-read-vdsox32 $(call QUIET_CLEAN, core-gen) $(RM) *.spec *.pyc *.pyo */*.pyc */*.pyo $(OUTPUT)common-cmds.h TAGS tags cscope* $(OUTPUT)PERF-VERSION-FILE $(OUTPUT)PERF-CFLAGS $(OUTPUT)PERF-FEATURES $(OUTPUT)util/*-bison* $(OUTPUT)util/*-flex* $(QUIET_SUBDIR0)Documentation $(QUIET_SUBDIR1) clean $(python-clean)
tools/perf/config/Makefile +24 -1
@@ -234,7 +234,9 @@ VF_FEATURE_TESTS = \ bionic \ liberty \ liberty-z \ - cplus-demangle + cplus-demangle \ + compile-32 \ + compile-x32 # Set FEATURE_CHECK_(C|LD)FLAGS-all for all CORE_FEATURE_TESTS features. # If in the future we need per-feature checks/flags for features not
@@ -620,6 +622,27 @@ ifdef HAVE_KVM_STAT_SUPPORT CFLAGS += -DHAVE_KVM_STAT_SUPPORT endif +ifeq (${IS_64_BIT}, 1) + ifndef NO_PERF_READ_VDSO32 + $(call feature_check,compile-32) + ifneq ($(feature-compile-32), 1) + NO_PERF_READ_VDSO32 := 1 + endif + endif + ifneq (${IS_X86_64}, 1) + NO_PERF_READ_VDSOX32 := 1 + endif + ifndef NO_PERF_READ_VDSOX32 + $(call feature_check,compile-x32) + ifneq ($(feature-compile-x32), 1) + NO_PERF_READ_VDSOX32 := 1 + endif + endif +else + NO_PERF_READ_VDSO32 := 1 + NO_PERF_READ_VDSOX32 := 1 +endif + # Among the variables below, these: # perfexecdir # template_dir
tools/perf/config/Makefile.arch +8 -0
@@ -21,3 +21,11 @@ ifeq ($(ARCH),x86_64) RAW_ARCH := x86_64 endif endif + +ifeq (${IS_X86_64}, 1) + IS_64_BIT := 1 +else ifeq ($(ARCH),x86) + IS_64_BIT := 0 +else + IS_64_BIT := $(shell echo __LP64__ | ${CC} ${CFLAGS} -E -x c - | tail -n 1) +endif
tools/perf/config/feature-checks/Makefile +9 -1
@@ -27,7 +27,9 @@ FILES= \ test-libunwind-debug-frame.bin \ test-stackprotector-all.bin \ test-timerfd.bin \ - test-libdw-dwarf-unwind.bin + test-libdw-dwarf-unwind.bin \ + test-compile-32.bin \ + test-compile-x32.bin CC := $(CROSS_COMPILE)gcc -MD PKG_CONFIG := $(CROSS_COMPILE)pkg-config
@@ -131,6 +133,12 @@ test-libdw-dwarf-unwind.bin: test-sync-compare-and-swap.bin: $(BUILD) -Werror +test-compile-32.bin: + $(CC) -m32 -o $(OUTPUT)$@ test-compile.c + +test-compile-x32.bin: + $(CC) -mx32 -o $(OUTPUT)$@ test-compile.c + -include *.d ###############################
tools/perf/config/feature-checks/test-compile.c new file +4 -0
@@ -0,0 +1,4 @@ +int main (void ) +{ +return 0 ; +}
tools/perf/perf-read-vdso.c new file +34 -0
@@ -0,0 +1,34 @@ +#include <stdio.h> +#include <string.h> + +#define VDSO__MAP_NAME "[vdso]" + +/* +* Include definition of find_vdso_map() also used in util/vdso.c for +* building perf. +*/ +#include "util/find-vdso-map.c" + +int main (void ) +{ +void *start, *end; +size_t size, written; + +if (find_vdso_map (&start, &end)) +return 1 ; + + size = end - start; + +while (size) { + written =fwrite (start,1 , size, stdout); +if (!written) +return 1 ; + start += written; + size -= written; + } + +if (fflush (stdout)) +return 1 ; + +return 0 ; +}
tools/perf/util/find-vdso-map.c new file +30 -0
@@ -0,0 +1,30 @@ +static int find_vdso_map (void **start,void **end) +{ +FILE *maps; +char line[128 ]; +int found =0 ; + + maps =fopen ("/proc/self/maps" ,"r" ); +if (!maps) { +fprintf (stderr,"vdso: cannot open maps\n" ); +return -1 ; + } + +while (!found &&fgets (line,sizeof (line), maps)) { +int m =-1 ; + +/* We care only about private r-x mappings. */ +if (2 !=sscanf (line,"%p-%p r-xp %*x %*x:%*x %*u %n" , + start, end, &m)) +continue ; +if (m <0 ) +continue ; + +if (!strncmp (&line[m],VDSO__MAP_NAME , +sizeof (VDSO__MAP_NAME ) -1 )) + found =1 ; + } + +fclose (maps); +return !found; +}
tools/perf/util/vdso.c +6 -31
@@ -15,6 +15,12 @@#include "linux/string.h" #include "debug.h" +/* +* Include definition of find_vdso_map() also used in perf-read-vdso.c for +* building perf-read-vdso32 and perf-read-vdsox32. +*/ +#include "find-vdso-map.c" +#define VDSO__TEMP_FILE_NAME "/tmp/perf-vdso.so-XXXXXX"struct vdso_file {
@@ -40,37 +46,6 @@ static struct vdso_info *vdso_info__new(void)return memdup (&vdso_info_init,sizeof (vdso_info_init)); } -static int find_vdso_map (void **start,void **end) -{ -FILE *maps; -char line[128 ]; -int found =0 ; - - maps =fopen ("/proc/self/maps" ,"r" ); -if (!maps) { -pr_err ("vdso: cannot open maps\n" ); -return -1 ; - } - -while (!found &&fgets (line,sizeof (line), maps)) { -int m =-1 ; - -/* We care only about private r-x mappings. */ -if (2 !=sscanf (line,"%p-%p r-xp %*x %*x:%*x %*u %n" , - start, end, &m)) -continue ; -if (m <0 ) -continue ; - -if (!strncmp (&line[m],VDSO__MAP_NAME , -sizeof (VDSO__MAP_NAME ) -1 )) - found =1 ; - } - -fclose (maps); -return !found; -} -static char *get_file (struct vdso_file *vdso_file) {char *vdso =NULL ;
Signature · 1 line
1.8.3.2
Arnaldo Carvalho de Melo
maintainerRe: [PATCH 22/24] perf tools: Build programs to copy 32-bit compatibility VDSOs
Em Fri, Aug 15, 2014 at 10:08:57PM +0300, Adrian Hunter escreveu:
Quoted 7 lines
perf tools copy VDSO out of memory. However, on 64-bit machines there may be 32-bit compatibility VDOs also. To copy those requires separate 32-bit executables. This patch adds to the build additional programs perf-read-vdso32 and perf-read-vdsox32 for 32-bit and x32 respectively.
Signed-off-by Adrian Hunter <[email protected]>
<SNIP>
Quoted 1 line
tools/perf/config/feature-checks/test-compile.c | 4 +++
<SNIP>
Quoted 1 line
8 files changed, 149 insertions(+), 37 deletions(-)
<SNIP>
Quoted 1 line
create mode 100644 tools/perf/config/feature-checks/test-compile.c
<SNIP>
In resolving conflicts that happen because I haven't merged some other files (thread_stack, etC) I noticed that this patch introduces something that is explicitely listed in a .gitignore file:
[acme@zoo linux]$ git show d316f16dddd8c7780c727bbd507338e4147c9280 commit d316f16dddd8c7780c727bbd507338e4147c9280
Author: Alexander Shishkin <[email protected]> Date: Fri Aug 15 22:08:59 2014 +0300
perf tools: Add feature checks to .gitignore
<SNIP>
+++ b/tools/perf/.gitignore
@@ -24,3 +24,4 @@ config.mak.autogen
*-flex.*
*.pyc
*.pyo
+config/feature-checks/test-*
[acme@zoo linux]$
So I think we should use some different name for this test-compile.c file?
I'll leave those two csets out of this pull req, so that you can rebase what is left on top of what is being merged and we can do one more progress step in having all this merged, ok?
I mean:
Quoted 6 lines
Quoted 6 lines
Up-to-date versions can be found from 15 August:
patchwork: https://patchwork.kernel.org/patch/4729131/ https://patchwork.kernel.org/patch/4729111/
- Arnaldo
Arnaldo Carvalho de Melo
maintainerRe: [PATCH 22/24] perf tools: Build programs to copy 32-bit compatibility VDSOs
Em Mon, Sep 15, 2014 at 12:47:52PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 40 lines
Em Fri, Aug 15, 2014 at 10:08:57PM +0300, Adrian Hunter escreveu:
Quoted 7 lines
perf tools copy VDSO out of memory. However, on 64-bit machines there may be 32-bit compatibility VDOs also. To copy those requires separate 32-bit executables. This patch adds to the build additional programs perf-read-vdso32 and perf-read-vdsox32 for 32-bit and x32 respectively.
Signed-off-by Adrian Hunter <[email protected]><SNIP>
Quoted 1 line
tools/perf/config/feature-checks/test-compile.c | 4 +++<SNIP>
Quoted 1 line
8 files changed, 149 insertions(+), 37 deletions(-)<SNIP>
Quoted 1 line
create mode 100644 tools/perf/config/feature-checks/test-compile.c<SNIP>
In resolving conflicts that happen because I haven't merged some other files (
thread_stack, etC) I noticed that this patch introduces something that is explicitely listed in a .gitignore file:[acme@zoo linux]$ git show
d316f16dddd8c7780c727bbd507338e4147c9280commitd316f16dddd8c7780c727bbd507338e4147c9280Author: Alexander Shishkin <[email protected]> Date: Fri Aug 15 22:08:59 2014 +0300perf tools: Add feature checks to .gitignore
<SNIP>
+++ b/tools/perf/.gitignore
(quoted diff) +1
-0@@ -24,3 +24,4 @@ config.mak.autogen *-flex.* *.pyc *.pyo +config/feature-checks/test-*[acme@zoo linux]$
Quoted 6 lines
So I think we should use some different name for this
test-compile.cfile?I'll leave those two csets out of this pull req, so that you can rebase what is left on top of what is being merged and we can do one more progress step in having all this merged, ok?
Since we can't ignore all test- prefixed files there, I'm removing the patch that added it to .gitignore as well:
commit d316f16dddd8c7780c727bbd507338e4147c9280
Author: Alexander Shishkin <[email protected]> Date: Fri Aug 15 22:08:59 2014 +0300
Aftewards I'll revisit this, because without this .gitignore patch I think your patch will be ok.
- Arnaldo
Quoted 11 lines
I mean:
Quoted 6 lines
Quoted 6 lines
Up-to-date versions can be found from 15 August:
patchwork: https://patchwork.kernel.org/patch/4729131/ https://patchwork.kernel.org/patch/4729111/- Arnaldo
Re: [PATCH 22/24] perf tools: Build programs to copy 32-bit compatibility VDSOs
Quoted 41 lines · Arnaldo Carvalho de Melo
Em Mon, Sep 15, 2014 at 12:47:52PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 40 lines
Em Fri, Aug 15, 2014 at 10:08:57PM +0300, Adrian Hunter escreveu:
Quoted 7 lines
perf tools copy VDSO out of memory. However, on 64-bit machines there may be 32-bit compatibility VDOs also. To copy those requires separate 32-bit executables. This patch adds to the build additional programs perf-read-vdso32 and perf-read-vdsox32 for 32-bit and x32 respectively.
Signed-off-by Adrian Hunter <[email protected]><SNIP>
Quoted 1 line
tools/perf/config/feature-checks/test-compile.c | 4 +++<SNIP>
Quoted 1 line
8 files changed, 149 insertions(+), 37 deletions(-)<SNIP>
Quoted 1 line
create mode 100644 tools/perf/config/feature-checks/test-compile.c<SNIP>
In resolving conflicts that happen because I haven't merged some other files (
thread_stack, etC) I noticed that this patch introduces something that is explicitely listed in a .gitignore file:[acme@zoo linux]$ git show
d316f16dddd8c7780c727bbd507338e4147c9280commitd316f16dddd8c7780c727bbd507338e4147c9280Author: Alexander Shishkin <[email protected]> Date: Fri Aug 15 22:08:59 2014 +0300perf tools: Add feature checks to .gitignore
<SNIP>
+++ b/tools/perf/.gitignore
(quoted diff) +1
-0@@ -24,3 +24,4 @@ config.mak.autogen *-flex.* *.pyc *.pyo +config/feature-checks/test-*[acme@zoo linux]$
Quoted 2 lines
Quoted 2 lines
So I think we should use some different name for this
test-compile.cfile?
The name is ok.
Quoted 8 lines
Quoted 4 lines
I'll leave those two csets out of this pull req, so that you can rebase what is left on top of what is being merged and we can do one more progress step in having all this merged, ok?
Since we can't ignore all test- prefixed files there, I'm removing the patch that added it to .gitignore as well:
Quoted 6 lines
commit
d316f16dddd8c7780c727bbd507338e4147c9280Author: Alexander Shishkin <[email protected]> Date: Fri Aug 15 22:08:59 2014 +0300Aftewards I'll revisit this, because without this .gitignore patch I think your patch will be ok.
The .gitignore patch is wrong. I picked it up because Alex had it in his tree but it must be very old because it was superseded by:
commit 56560ec692c142bb9ee404764e3b67999031ad19
Author: Chunwei Chen <[email protected]> Date: Sat Dec 21 13:48:11 2013 +0800
perf config: Ignore generated files in feature-checks
1. Rename the test-* binary files to test-*.bin for easier pattern matching as suggested by Ingo.
2. Ignore *.bin and *.d files.
Quoted 16 lines
- Arnaldo
Quoted 11 lines
I mean:
Quoted 6 lines
Quoted 6 lines
Up-to-date versions can be found from 15 August:
patchwork: https://patchwork.kernel.org/patch/4729131/ https://patchwork.kernel.org/patch/4729111/- Arnaldo
Arnaldo Carvalho de Melo
maintainerRe: [PATCH 22/24] perf tools: Build programs to copy 32-bit compatibility VDSOs
Em Tue, Sep 16, 2014 at 09:59:33AM +0300, Adrian Hunter escreveu:
Quoted 42 lines · Arnaldo Carvalho de Melo
Quoted 41 lines
Em Mon, Sep 15, 2014 at 12:47:52PM -0300, Arnaldo Carvalho de Melo escreveu:
Quoted 40 lines
Em Fri, Aug 15, 2014 at 10:08:57PM +0300, Adrian Hunter escreveu:
Quoted 7 lines
perf tools copy VDSO out of memory. However, on 64-bit machines there may be 32-bit compatibility VDOs also. To copy those requires separate 32-bit executables. This patch adds to the build additional programs perf-read-vdso32 and perf-read-vdsox32 for 32-bit and x32 respectively.
Signed-off-by Adrian Hunter <[email protected]><SNIP>
Quoted 1 line
tools/perf/config/feature-checks/test-compile.c | 4 +++<SNIP>
Quoted 1 line
8 files changed, 149 insertions(+), 37 deletions(-)<SNIP>
Quoted 1 line
create mode 100644 tools/perf/config/feature-checks/test-compile.c<SNIP>
In resolving conflicts that happen because I haven't merged some other files (
thread_stack, etC) I noticed that this patch introduces something that is explicitely listed in a .gitignore file:[acme@zoo linux]$ git show
d316f16dddd8c7780c727bbd507338e4147c9280commitd316f16dddd8c7780c727bbd507338e4147c9280Author: Alexander Shishkin <[email protected]> Date: Fri Aug 15 22:08:59 2014 +0300perf tools: Add feature checks to .gitignore
<SNIP>
+++ b/tools/perf/.gitignore
(quoted diff) +1
-0@@ -24,3 +24,4 @@ config.mak.autogen *-flex.* *.pyc *.pyo +config/feature-checks/test-*[acme@zoo linux]$
Quoted 4 lines
Quoted 2 lines
Quoted 2 lines
So I think we should use some different name for this
test-compile.cfile?The name is ok.
Ok
Quoted 28 lines
Quoted 8 lines
Quoted 4 lines
I'll leave those two csets out of this pull req, so that you can rebase what is left on top of what is being merged and we can do one more progress step in having all this merged, ok?
Since we can't ignore all test- prefixed files there, I'm removing the patch that added it to .gitignore as well:
Quoted 6 lines
commit
d316f16dddd8c7780c727bbd507338e4147c9280Author: Alexander Shishkin <[email protected]> Date: Fri Aug 15 22:08:59 2014 +0300Aftewards I'll revisit this, because without this .gitignore patch I think your patch will be ok.
The .gitignore patch is wrong. I picked it up because Alex had it in his tree but it must be very old because it was superseded by:
commit
56560ec692c142bb9ee404764e3b67999031ad19Author: Chunwei Chen <[email protected]> Date: Sat Dec 21 13:48:11 2013 +0800perf config: Ignore generated files in feature-checks
1. Rename the test-* binary files to test-*.bin for easier pattern matching as suggested by Ingo.
2. Ignore *.bin and *.d files.
Quoted 7 lines
Signed-off-by Chunwei Chen <[email protected]>Reviewed-by Ingo Molnar <[email protected]>Acked-by Jiri Olsa <[email protected]>Cc Ingo Molnar <[email protected]>Cc Jiri Olsa <[email protected]>Signed-off-by Arnaldo Carvalho de Melo <[email protected]>
Thanks, I'll fix this up.
- Arnaldo
Adrian Hunter
reviewerPATCH 23/24 perf tools: Add support for 32-bit compatibility VDSOs
In reply to Adrian Hunter'perf record' post-processes the event stream to create a list of build-ids for object files for which sample events have been recorded. That results in those object files being recorded in the build-id cache.
In the case of VDSO, perf tools reads it from memory and copies it into a temporary file, which as decribed above, gets added to the build-id cache.
Then when the perf.data file is processed by other tools, the build-id of VDSO is listed in the perf.data file and the VDSO can be read from the build-id cache. In that case the name of the map, the short name of the DSO, and the entry in the build-id cache are all "[vdso]".
However, in the 64-bit case, there also can be 32-bit compatibility VDSOs.
A previous patch added programs "perf-read-vdso32" and "perf read-vdsox32". This patch uses those programs to read the correct VDSO for a thread and create a temporary file just as for the 64-bit VDSO. The map name and the entry in the build-id cache are still "[vdso]" but the DSO short name becomes "[vdso32]" and "[vdsox32]" respectively.
--- tools/perf/util/vdso.c | 170 ++++++++++++++++++++++++++++++++++++++++++++++++- tools/perf/util/vdso.h | 4 +- 2 files changed, 172 insertions(+), 2 deletions(-)
tools/perf/util/vdso.c +169 -1
@@ -12,6 +12,7 @@#include "util.h" #include "symbol.h" #include "machine.h" +#include "thread.h" #include "linux/string.h" #include "debug.h"
@@ -28,10 +29,15 @@ struct vdso_file {bool error;char temp_file_name[sizeof (VDSO__TEMP_FILE_NAME )];const char *dso_name; +const char *read_prog; };struct vdso_info {struct vdso_file vdso; +#if BITS_PER_LONG ==64 +struct vdso_file vdso32; +struct vdso_file vdsox32; +#endif };static struct vdso_info *vdso_info__new (void )
@@ -41,6 +47,18 @@ static struct vdso_info *vdso_info__new(void) .temp_file_name = VDSO__TEMP_FILE_NAME, .dso_name = DSO__NAME_VDSO, }, +#if BITS_PER_LONG ==64 + .vdso32 = { + .temp_file_name = VDSO__TEMP_FILE_NAME, + .dso_name = DSO__NAME_VDSO32, + .read_prog = "perf-read-vdso32", + }, + .vdsox32 = { + .temp_file_name = VDSO__TEMP_FILE_NAME, + .dso_name = DSO__NAME_VDSOX32, + .read_prog = "perf-read-vdsox32", + }, +#endif };return memdup (&vdso_info_init,sizeof (vdso_info_init));
@@ -92,6 +110,12 @@ void vdso__exit(struct machine *machine)if (vdso_info->vdso.found)unlink (vdso_info->vdso.temp_file_name); +#if BITS_PER_LONG ==64 +if (vdso_info->vdso32.found) +unlink (vdso_info->vdso32.temp_file_name); +if (vdso_info->vdsox32.found) +unlink (vdso_info->vdsox32.temp_file_name); +#endif zfree (&machine->vdso_info); }
@@ -110,6 +134,143 @@ static struct dso *vdso__new(struct machine *machine, const char *short_name,return dso; } +#if BITS_PER_LONG ==64 + +static enum dso_type machine__thread_dso_type (struct machine *machine, +struct thread *thread) +{ +enum dso_type dso_type =DSO__TYPE_UNKNOWN ; +struct map *map; +struct dso *dso; + + map =map_groups__first (thread->mg,MAP__FUNCTION ); +for (; map ; map =map_groups__next (map)) { + dso = map->dso; +if (!dso || dso->long_name[0 ] !='/' ) +continue ; + dso_type =dso__type (dso, machine); +if (dso_type !=DSO__TYPE_UNKNOWN ) +break ; + } + +return dso_type; +} + +static int vdso__do_copy_compat (FILE *f,int fd) +{ +char buf[4096 ]; +size_t count; + +while (1 ) { + count =fread (buf,1 ,sizeof (buf), f); +if (ferror (f)) +return -errno; +if (feof (f)) +break ; +if (count &&writen (fd, buf, count) != (ssize_t )count) +return -errno; + } + +return 0 ; +} + +static int vdso__copy_compat (const char *prog,int fd) +{ +FILE *f; +int err; + + f =popen (prog,"r" ); +if (!f) +return -errno; + + err =vdso__do_copy_compat (f, fd); + +if (pclose (f) ==-1 ) +return -errno; + +return err; +} + +static int vdso__create_compat_file (const char *prog,char *temp_name) +{ +int fd, err; + + fd =mkstemp (temp_name); +if (fd <0 ) +return -errno; + + err =vdso__copy_compat (prog, fd); + +if (close (fd) ==-1 ) +return -errno; + +return err; +} + +static const char *vdso__get_compat_file (struct vdso_file *vdso_file) +{ +int err; + +if (vdso_file->found) +return vdso_file->temp_file_name; + +if (vdso_file->error) +return NULL ; + + err =vdso__create_compat_file (vdso_file->read_prog, + vdso_file->temp_file_name); +if (err) { +pr_err ("%s failed, error %d\n" , vdso_file->read_prog, err); + vdso_file->error = true; +return NULL ; + } + + vdso_file->found = true; + +return vdso_file->temp_file_name; +} + +static struct dso *vdso__findnew_compat (struct machine *machine, +struct vdso_file *vdso_file) +{ +const char *file_name; +struct dso *dso; + + dso =dsos__find (&machine->user_dsos, vdso_file->dso_name, true); +if (dso) +return dso; + + file_name =vdso__get_compat_file (vdso_file); +if (!file_name) +return NULL ; + +return vdso__new (machine, vdso_file->dso_name, file_name); +} + +static int vdso__dso_findnew_compat (struct machine *machine, +struct thread *thread, +struct vdso_info *vdso_info, +struct dso **dso) +{ +enum dso_type dso_type; + + dso_type =machine__thread_dso_type (machine, thread); +switch (dso_type) { +case DSO__TYPE_32BIT : + *dso =vdso__findnew_compat (machine, &vdso_info->vdso32); +return 1 ; +case DSO__TYPE_X32BIT : + *dso =vdso__findnew_compat (machine, &vdso_info->vdsox32); +return 1 ; +case DSO__TYPE_UNKNOWN : +case DSO__TYPE_64BIT : +default : +return 0 ; + } +} + +#endif +struct dso *vdso__dso_findnew (struct machine *machine,struct thread *thread__maybe_unused ) {
@@ -123,6 +284,11 @@ struct dso *vdso__dso_findnew(struct machine *machine,if (!vdso_info)return NULL ; +#if BITS_PER_LONG ==64 +if (vdso__dso_findnew_compat (machine, thread, vdso_info, &dso)) +return dso; +#endif + dso =dsos__find (&machine->user_dsos,DSO__NAME_VDSO , true);if (!dso) {char *file;
@@ -139,5 +305,7 @@ struct dso *vdso__dso_findnew(struct machine *machine,bool dso__is_vdso (struct dso *dso) { -return !strcmp (dso->short_name,DSO__NAME_VDSO ); +return !strcmp (dso->short_name,DSO__NAME_VDSO ) || + !strcmp (dso->short_name,DSO__NAME_VDSO32 ) || + !strcmp (dso->short_name,DSO__NAME_VDSOX32 ); }
tools/perf/util/vdso.h +3 -1
@@ -7,7 +7,9 @@#define VDSO__MAP_NAME "[vdso]" -#define DSO__NAME_VDSO "[vdso]" +#define DSO__NAME_VDSO "[vdso]" +#define DSO__NAME_VDSO32 "[vdso32]" +#define DSO__NAME_VDSOX32 "[vdsox32]"static inline bool is_vdso_map (const char *filename) {
Signature · 1 line
1.8.3.2
Adrian Hunter
reviewerPATCH 24/24 perf tools: Add feature checks to .gitignore
In reply to Adrian HunterFrom: Alexander Shishkin <[email protected]>
--- tools/perf/.gitignore | 1 + 1 file changed, 1 insertion(+)
tools/perf/.gitignore +1 -0
@@ -25,3 +25,4 @@ config.mak.autogen *-flex.* *.pyc *.pyo +config/feature-checks/test-*
Signature · 1 line
1.8.3.2