perf/urgent fixes:
. Fix parsing with no sample_id_all bit set, this regression prevents perf from reading old perf.data files generated in systems where perf_event_attr.sample_id_all isn't available, from Adrian Hunter. . Add signal checking to the inner 'perf trace' event processing loop, allowing faster response to control+C. . Fix formatting of long symbol names removing the hardcoding of a buffer size used to format histogram entries, which was truncating the lines. . Separate progress bar update when processing events, reducing potentially big overhead in not needed TUI progress bar screen updates, from Jiri Olsa. . Fix 'perf trace' build in architectures where MAP_32BIT is not defined, from Kyle McMartin. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.14 (GNU/Linux) iQIcBAABAgAGBQJSKN3vAAoJENZQFvNTUqpAHRIQAJnS3nUqxwsoaqKB2IJZjhTK eZhplHS3kxoMuP0ZNIR6cDT/gHfILak2jiTYLpuDpp8MfXBVDKUfV73UwKunIO7w CYD+WstJw6zH2HhXkL1a+Tk6R06G+fls6tErFlg4J4wuFo3wWTwUYfWi7QYyUdf6 FaU8YF+xQT08XR/f1dBxv6fK/2ibblBxuh1puhbnoJd2VCRtaeIO+7XbOEGNnDqN rkLgSlBxzEfIyHvm4yo1TsKfZMKH9PrJZ3OY6/Ga6MAlC1uZ7b0eHrNpd5SeWjKL z+QPpoFKdW+7hDHEffml/3+xyhHOGc3Dr1BfEEn0ue30GfeLV1mRFu9qofN9agJO 5HuimdrA49BTsx3bjuiaQLy4vfUVSxPOmgAUKVXMRjf8WVXlsy5ygBWeX5/DP4NK zMexGE+IKVFv33cJFVp6HBwmLB68q0YOZ9SHCTXtKQKvIFzC+W0JQ3wHX+7YeI7f u02zezxdmdcH0H+QVztwM1k9rnkwR02lN17SOcRNj0NNgyzb4UD2Xqijn/PJVN/k SnikNdi4w3v/I02NjAun7KBdzlZlglsoU96FrdifwXrj6q8ZR2yjx/v8cUfSeebu GeYUteR4Me00p1s7FGyTd9yVAuJ2/mOZzD4BFz8Q0LxdCBaMf2aMCA9jUqmQBsz4 m6/bpt2YdC8t+KzDEHDW =akcD -----END PGP SIGNATURE----- Merge tag 'perf-urgent-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/urgent Pull perf/urgent fixes from Arnaldo Carvalho de Melo: * Fix parsing with no sample_id_all bit set, this regression prevents perf from reading old perf.data files generated in systems where perf_event_attr.sample_id_all isn't available, from Adrian Hunter. * Add signal checking to the inner 'perf trace' event processing loop, allowing faster response to control+C. * Fix formatting of long symbol names removing the hardcoding of a buffer size used to format histogram entries, which was truncating the lines. * Separate progress bar update when processing events, reducing potentially big overhead in not needed TUI progress bar screen updates, from Jiri Olsa. * Fix 'perf trace' build in architectures where MAP_32BIT is not defined, from Kyle McMartin. Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: Ingo Molnar <mingo@kernel.org>
This commit is contained in:
commit
72f4a11d2f
|
@ -394,6 +394,8 @@ ifeq ($(ARCH),x86)
|
||||||
LIB_OBJS += $(OUTPUT)tests/perf-time-to-tsc.o
|
LIB_OBJS += $(OUTPUT)tests/perf-time-to-tsc.o
|
||||||
endif
|
endif
|
||||||
LIB_OBJS += $(OUTPUT)tests/code-reading.o
|
LIB_OBJS += $(OUTPUT)tests/code-reading.o
|
||||||
|
LIB_OBJS += $(OUTPUT)tests/sample-parsing.o
|
||||||
|
LIB_OBJS += $(OUTPUT)tests/parse-no-sample-id-all.o
|
||||||
|
|
||||||
BUILTIN_OBJS += $(OUTPUT)builtin-annotate.o
|
BUILTIN_OBJS += $(OUTPUT)builtin-annotate.o
|
||||||
BUILTIN_OBJS += $(OUTPUT)builtin-bench.o
|
BUILTIN_OBJS += $(OUTPUT)builtin-bench.o
|
||||||
|
@ -439,7 +441,6 @@ PERFLIBS = $(LIB_FILE) $(LIBLK) $(LIBTRACEEVENT)
|
||||||
ifneq ($(OUTPUT),)
|
ifneq ($(OUTPUT),)
|
||||||
CFLAGS += -I$(OUTPUT)
|
CFLAGS += -I$(OUTPUT)
|
||||||
endif
|
endif
|
||||||
LIB_OBJS += $(OUTPUT)tests/sample-parsing.o
|
|
||||||
|
|
||||||
ifdef NO_LIBELF
|
ifdef NO_LIBELF
|
||||||
EXTLIBS := $(filter-out -lelf,$(EXTLIBS))
|
EXTLIBS := $(filter-out -lelf,$(EXTLIBS))
|
||||||
|
|
|
@ -100,7 +100,9 @@ static size_t syscall_arg__scnprintf_mmap_flags(char *bf, size_t size,
|
||||||
|
|
||||||
P_MMAP_FLAG(SHARED);
|
P_MMAP_FLAG(SHARED);
|
||||||
P_MMAP_FLAG(PRIVATE);
|
P_MMAP_FLAG(PRIVATE);
|
||||||
|
#ifdef MAP_32BIT
|
||||||
P_MMAP_FLAG(32BIT);
|
P_MMAP_FLAG(32BIT);
|
||||||
|
#endif
|
||||||
P_MMAP_FLAG(ANONYMOUS);
|
P_MMAP_FLAG(ANONYMOUS);
|
||||||
P_MMAP_FLAG(DENYWRITE);
|
P_MMAP_FLAG(DENYWRITE);
|
||||||
P_MMAP_FLAG(EXECUTABLE);
|
P_MMAP_FLAG(EXECUTABLE);
|
||||||
|
@ -994,6 +996,9 @@ again:
|
||||||
|
|
||||||
handler = evsel->handler.func;
|
handler = evsel->handler.func;
|
||||||
handler(trace, evsel, &sample);
|
handler(trace, evsel, &sample);
|
||||||
|
|
||||||
|
if (done)
|
||||||
|
goto out_unmap_evlist;
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|
|
@ -111,6 +111,10 @@ static struct test {
|
||||||
.desc = "Test using a dummy software event to keep tracking",
|
.desc = "Test using a dummy software event to keep tracking",
|
||||||
.func = test__keep_tracking,
|
.func = test__keep_tracking,
|
||||||
},
|
},
|
||||||
|
{
|
||||||
|
.desc = "Test parsing with no sample_id_all bit set",
|
||||||
|
.func = test__parse_no_sample_id_all,
|
||||||
|
},
|
||||||
{
|
{
|
||||||
.func = NULL,
|
.func = NULL,
|
||||||
},
|
},
|
||||||
|
|
|
@ -0,0 +1,108 @@
|
||||||
|
#include <sys/types.h>
|
||||||
|
#include <stddef.h>
|
||||||
|
|
||||||
|
#include "tests.h"
|
||||||
|
|
||||||
|
#include "event.h"
|
||||||
|
#include "evlist.h"
|
||||||
|
#include "header.h"
|
||||||
|
#include "util.h"
|
||||||
|
|
||||||
|
static int process_event(struct perf_evlist **pevlist, union perf_event *event)
|
||||||
|
{
|
||||||
|
struct perf_sample sample;
|
||||||
|
|
||||||
|
if (event->header.type == PERF_RECORD_HEADER_ATTR) {
|
||||||
|
if (perf_event__process_attr(NULL, event, pevlist)) {
|
||||||
|
pr_debug("perf_event__process_attr failed\n");
|
||||||
|
return -1;
|
||||||
|
}
|
||||||
|
return 0;
|
||||||
|
}
|
||||||
|
|
||||||
|
if (event->header.type >= PERF_RECORD_USER_TYPE_START)
|
||||||
|
return -1;
|
||||||
|
|
||||||
|
if (!*pevlist)
|
||||||
|
return -1;
|
||||||
|
|
||||||
|
if (perf_evlist__parse_sample(*pevlist, event, &sample)) {
|
||||||
|
pr_debug("perf_evlist__parse_sample failed\n");
|
||||||
|
return -1;
|
||||||
|
}
|
||||||
|
|
||||||
|
return 0;
|
||||||
|
}
|
||||||
|
|
||||||
|
static int process_events(union perf_event **events, size_t count)
|
||||||
|
{
|
||||||
|
struct perf_evlist *evlist = NULL;
|
||||||
|
int err = 0;
|
||||||
|
size_t i;
|
||||||
|
|
||||||
|
for (i = 0; i < count && !err; i++)
|
||||||
|
err = process_event(&evlist, events[i]);
|
||||||
|
|
||||||
|
if (evlist)
|
||||||
|
perf_evlist__delete(evlist);
|
||||||
|
|
||||||
|
return err;
|
||||||
|
}
|
||||||
|
|
||||||
|
struct test_attr_event {
|
||||||
|
struct attr_event attr;
|
||||||
|
u64 id;
|
||||||
|
};
|
||||||
|
|
||||||
|
/**
|
||||||
|
* test__parse_no_sample_id_all - test parsing with no sample_id_all bit set.
|
||||||
|
*
|
||||||
|
* This function tests parsing data produced on kernel's that do not support the
|
||||||
|
* sample_id_all bit. Without the sample_id_all bit, non-sample events (such as
|
||||||
|
* mmap events) do not have an id sample appended, and consequently logic
|
||||||
|
* designed to determine the id will not work. That case happens when there is
|
||||||
|
* more than one selected event, so this test processes three events: 2
|
||||||
|
* attributes representing the selected events and one mmap event.
|
||||||
|
*
|
||||||
|
* Return: %0 on success, %-1 if the test fails.
|
||||||
|
*/
|
||||||
|
int test__parse_no_sample_id_all(void)
|
||||||
|
{
|
||||||
|
int err;
|
||||||
|
|
||||||
|
struct test_attr_event event1 = {
|
||||||
|
.attr = {
|
||||||
|
.header = {
|
||||||
|
.type = PERF_RECORD_HEADER_ATTR,
|
||||||
|
.size = sizeof(struct test_attr_event),
|
||||||
|
},
|
||||||
|
},
|
||||||
|
.id = 1,
|
||||||
|
};
|
||||||
|
struct test_attr_event event2 = {
|
||||||
|
.attr = {
|
||||||
|
.header = {
|
||||||
|
.type = PERF_RECORD_HEADER_ATTR,
|
||||||
|
.size = sizeof(struct test_attr_event),
|
||||||
|
},
|
||||||
|
},
|
||||||
|
.id = 2,
|
||||||
|
};
|
||||||
|
struct mmap_event event3 = {
|
||||||
|
.header = {
|
||||||
|
.type = PERF_RECORD_MMAP,
|
||||||
|
.size = sizeof(struct mmap_event),
|
||||||
|
},
|
||||||
|
};
|
||||||
|
union perf_event *events[] = {
|
||||||
|
(union perf_event *)&event1,
|
||||||
|
(union perf_event *)&event2,
|
||||||
|
(union perf_event *)&event3,
|
||||||
|
};
|
||||||
|
|
||||||
|
err = process_events(events, ARRAY_SIZE(events));
|
||||||
|
if (err)
|
||||||
|
return -1;
|
||||||
|
|
||||||
|
return 0;
|
||||||
|
}
|
|
@ -39,5 +39,6 @@ int test__perf_time_to_tsc(void);
|
||||||
int test__code_reading(void);
|
int test__code_reading(void);
|
||||||
int test__sample_parsing(void);
|
int test__sample_parsing(void);
|
||||||
int test__keep_tracking(void);
|
int test__keep_tracking(void);
|
||||||
|
int test__parse_no_sample_id_all(void);
|
||||||
|
|
||||||
#endif /* TESTS_H */
|
#endif /* TESTS_H */
|
||||||
|
|
|
@ -350,9 +350,9 @@ static int hist_entry__period_snprintf(struct perf_hpp *hpp,
|
||||||
}
|
}
|
||||||
|
|
||||||
static int hist_entry__fprintf(struct hist_entry *he, size_t size,
|
static int hist_entry__fprintf(struct hist_entry *he, size_t size,
|
||||||
struct hists *hists, FILE *fp)
|
struct hists *hists,
|
||||||
|
char *bf, size_t bfsz, FILE *fp)
|
||||||
{
|
{
|
||||||
char bf[512];
|
|
||||||
int ret;
|
int ret;
|
||||||
struct perf_hpp hpp = {
|
struct perf_hpp hpp = {
|
||||||
.buf = bf,
|
.buf = bf,
|
||||||
|
@ -360,8 +360,8 @@ static int hist_entry__fprintf(struct hist_entry *he, size_t size,
|
||||||
};
|
};
|
||||||
bool color = !symbol_conf.field_sep;
|
bool color = !symbol_conf.field_sep;
|
||||||
|
|
||||||
if (size == 0 || size > sizeof(bf))
|
if (size == 0 || size > bfsz)
|
||||||
size = hpp.size = sizeof(bf);
|
size = hpp.size = bfsz;
|
||||||
|
|
||||||
ret = hist_entry__period_snprintf(&hpp, he, color);
|
ret = hist_entry__period_snprintf(&hpp, he, color);
|
||||||
hist_entry__sort_snprintf(he, bf + ret, size - ret, hists);
|
hist_entry__sort_snprintf(he, bf + ret, size - ret, hists);
|
||||||
|
@ -392,6 +392,8 @@ size_t hists__fprintf(struct hists *hists, bool show_header, int max_rows,
|
||||||
.ptr = hists_to_evsel(hists),
|
.ptr = hists_to_evsel(hists),
|
||||||
};
|
};
|
||||||
bool first = true;
|
bool first = true;
|
||||||
|
size_t linesz;
|
||||||
|
char *line = NULL;
|
||||||
|
|
||||||
init_rem_hits();
|
init_rem_hits();
|
||||||
|
|
||||||
|
@ -479,6 +481,13 @@ size_t hists__fprintf(struct hists *hists, bool show_header, int max_rows,
|
||||||
goto out;
|
goto out;
|
||||||
|
|
||||||
print_entries:
|
print_entries:
|
||||||
|
linesz = hists__sort_list_width(hists) + 3 + 1;
|
||||||
|
line = malloc(linesz);
|
||||||
|
if (line == NULL) {
|
||||||
|
ret = -1;
|
||||||
|
goto out;
|
||||||
|
}
|
||||||
|
|
||||||
for (nd = rb_first(&hists->entries); nd; nd = rb_next(nd)) {
|
for (nd = rb_first(&hists->entries); nd; nd = rb_next(nd)) {
|
||||||
struct hist_entry *h = rb_entry(nd, struct hist_entry, rb_node);
|
struct hist_entry *h = rb_entry(nd, struct hist_entry, rb_node);
|
||||||
float percent = h->stat.period * 100.0 /
|
float percent = h->stat.period * 100.0 /
|
||||||
|
@ -490,10 +499,10 @@ print_entries:
|
||||||
if (percent < min_pcnt)
|
if (percent < min_pcnt)
|
||||||
continue;
|
continue;
|
||||||
|
|
||||||
ret += hist_entry__fprintf(h, max_cols, hists, fp);
|
ret += hist_entry__fprintf(h, max_cols, hists, line, linesz, fp);
|
||||||
|
|
||||||
if (max_rows && ++nr_rows >= max_rows)
|
if (max_rows && ++nr_rows >= max_rows)
|
||||||
goto out;
|
break;
|
||||||
|
|
||||||
if (h->ms.map == NULL && verbose > 1) {
|
if (h->ms.map == NULL && verbose > 1) {
|
||||||
__map_groups__fprintf_maps(&h->thread->mg,
|
__map_groups__fprintf_maps(&h->thread->mg,
|
||||||
|
@ -501,6 +510,8 @@ print_entries:
|
||||||
fprintf(fp, "%.10s end\n", graph_dotted_line);
|
fprintf(fp, "%.10s end\n", graph_dotted_line);
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
|
free(line);
|
||||||
out:
|
out:
|
||||||
free(rem_sq_bracket);
|
free(rem_sq_bracket);
|
||||||
|
|
||||||
|
|
|
@ -446,20 +446,25 @@ static int perf_evlist__event2id(struct perf_evlist *evlist,
|
||||||
static struct perf_evsel *perf_evlist__event2evsel(struct perf_evlist *evlist,
|
static struct perf_evsel *perf_evlist__event2evsel(struct perf_evlist *evlist,
|
||||||
union perf_event *event)
|
union perf_event *event)
|
||||||
{
|
{
|
||||||
|
struct perf_evsel *first = perf_evlist__first(evlist);
|
||||||
struct hlist_head *head;
|
struct hlist_head *head;
|
||||||
struct perf_sample_id *sid;
|
struct perf_sample_id *sid;
|
||||||
int hash;
|
int hash;
|
||||||
u64 id;
|
u64 id;
|
||||||
|
|
||||||
if (evlist->nr_entries == 1)
|
if (evlist->nr_entries == 1)
|
||||||
return perf_evlist__first(evlist);
|
return first;
|
||||||
|
|
||||||
|
if (!first->attr.sample_id_all &&
|
||||||
|
event->header.type != PERF_RECORD_SAMPLE)
|
||||||
|
return first;
|
||||||
|
|
||||||
if (perf_evlist__event2id(evlist, event, &id))
|
if (perf_evlist__event2id(evlist, event, &id))
|
||||||
return NULL;
|
return NULL;
|
||||||
|
|
||||||
/* Synthesized events have an id of zero */
|
/* Synthesized events have an id of zero */
|
||||||
if (!id)
|
if (!id)
|
||||||
return perf_evlist__first(evlist);
|
return first;
|
||||||
|
|
||||||
hash = hash_64(id, PERF_EVLIST__HLIST_BITS);
|
hash = hash_64(id, PERF_EVLIST__HLIST_BITS);
|
||||||
head = &evlist->heads[hash];
|
head = &evlist->heads[hash];
|
||||||
|
|
|
@ -504,6 +504,7 @@ static int flush_sample_queue(struct perf_session *s,
|
||||||
u64 limit = os->next_flush;
|
u64 limit = os->next_flush;
|
||||||
u64 last_ts = os->last_sample ? os->last_sample->timestamp : 0ULL;
|
u64 last_ts = os->last_sample ? os->last_sample->timestamp : 0ULL;
|
||||||
unsigned idx = 0, progress_next = os->nr_samples / 16;
|
unsigned idx = 0, progress_next = os->nr_samples / 16;
|
||||||
|
bool show_progress = limit == ULLONG_MAX;
|
||||||
int ret;
|
int ret;
|
||||||
|
|
||||||
if (!tool->ordered_samples || !limit)
|
if (!tool->ordered_samples || !limit)
|
||||||
|
@ -526,7 +527,7 @@ static int flush_sample_queue(struct perf_session *s,
|
||||||
os->last_flush = iter->timestamp;
|
os->last_flush = iter->timestamp;
|
||||||
list_del(&iter->list);
|
list_del(&iter->list);
|
||||||
list_add(&iter->list, &os->sample_cache);
|
list_add(&iter->list, &os->sample_cache);
|
||||||
if (++idx >= progress_next) {
|
if (show_progress && (++idx >= progress_next)) {
|
||||||
progress_next += os->nr_samples / 16;
|
progress_next += os->nr_samples / 16;
|
||||||
ui_progress__update(idx, os->nr_samples,
|
ui_progress__update(idx, os->nr_samples,
|
||||||
"Processing time ordered events...");
|
"Processing time ordered events...");
|
||||||
|
|
Loading…
Reference in New Issue