Use session to cache and order events from all CPUs so we get
correct ordering of comms/mmaps and faults.
Signed-off-by: Stanislav Fomichev <stfomic...@yandex-team.ru>
---
tools/perf/builtin-trace.c | 117 ++++++++++++++++++---------------------------
1 file changed, 46 insertions(+), 71 deletions(-)
diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c
index 3cce721f8b06..e163998547ef 100644
--- a/tools/perf/builtin-trace.c
+++ b/tools/perf/builtin-trace.c
@@ -20,6 +20,11 @@
#include <sys/mman.h>
#include <linux/futex.h>
+/* keeping the max events to a modest level to keep
+ * the processing of samples per mmap smooth.
+ */
+#define PERF_TRACE__MAX_EVENTS_PER_MMAP 25
+
/* For older distros: */
#ifndef MAP_STACK
# define MAP_STACK 0x20000
@@ -2001,10 +2006,12 @@ static int perf_evlist__add_pgfault(struct perf_evlist
*evlist,
static int trace__run(struct trace *trace, int argc, const char **argv)
{
+ struct perf_data_file file = {
+ .mode = PERF_DATA_MODE_WRITE,
+ };
+ struct perf_session *session;
struct perf_evlist *evlist = perf_evlist__new();
- struct perf_evsel *evsel;
- int err = -1, i;
- unsigned long before;
+ int err = -1;
const bool forks = argc > 0;
trace->live = true;
@@ -2049,6 +2056,18 @@ static int trace__run(struct trace *trace, int argc,
const char **argv)
perf_evlist__config(evlist, &trace->opts);
+ session = perf_session__new(&file, false, &trace->tool);
+ if (session == NULL) {
+ err = -ENOMEM;
+ goto out_delete_evlist;
+ }
+ session->evlist = evlist;
+ perf_session__set_id_hdr_size(session);
+ machine__synthesize_threads(&session->machines.host,
+ &trace->opts.target,
+ evlist->threads, false);
+ trace->host = &session->machines.host;
+
signal(SIGCHLD, sig_handler);
signal(SIGINT, sig_handler);
@@ -2068,7 +2087,7 @@ static int trace__run(struct trace *trace, int argc,
const char **argv)
err = perf_evlist__mmap(evlist, trace->opts.mmap_pages, false);
if (err < 0) {
fprintf(trace->output, "Couldn't mmap the events: %s\n",
strerror(errno));
- goto out_delete_evlist;
+ goto out_delete_session;
}
perf_evlist__enable(evlist);
@@ -2077,67 +2096,19 @@ static int trace__run(struct trace *trace, int argc,
const char **argv)
perf_evlist__start_workload(evlist);
trace->multiple_threads = evlist->threads->map[0] == -1 ||
evlist->threads->nr > 1;
-again:
- before = trace->nr_events;
-
- for (i = 0; i < evlist->nr_mmaps; i++) {
- union perf_event *event;
-
- while ((event = perf_evlist__mmap_read(evlist, i)) != NULL) {
- const u32 type = event->header.type;
- tracepoint_handler handler;
- struct perf_sample sample;
-
- ++trace->nr_events;
- err = perf_evlist__parse_sample(evlist, event, &sample);
- if (err) {
- fprintf(trace->output, "Can't parse sample, err
= %d, skipping...\n", err);
- goto next_event;
- }
-
- if (!trace->full_time && trace->base_time == 0)
- trace->base_time = sample.time;
-
- if (type != PERF_RECORD_SAMPLE) {
- trace__process_event(trace, trace->host, event,
&sample);
- continue;
- }
-
- evsel = perf_evlist__id2evsel(evlist, sample.id);
- if (evsel == NULL) {
- fprintf(trace->output, "Unknown tp ID %" PRIu64
", skipping...\n", sample.id);
- goto next_event;
- }
-
- if (evsel->attr.type == PERF_TYPE_TRACEPOINT &&
- sample.raw_data == NULL) {
- fprintf(trace->output, "%s sample with no
payload for tid: %d, cpu %d, raw_size=%d, skipping...\n",
- perf_evsel__name(evsel), sample.tid,
- sample.cpu, sample.raw_size);
- goto next_event;
- }
-
- handler = evsel->handler;
- handler(trace, evsel, event, &sample);
-next_event:
- perf_evlist__mmap_consume(evlist, i);
-
- if (interrupted)
- goto out_disable;
- }
- }
+ while (!done) {
+ int rc;
- if (trace->nr_events == before) {
- int timeout = done ? 100 : -1;
+ rc = perf_session__mmap_read(&trace->tool, session, evlist,
+ PERF_TRACE__MAX_EVENTS_PER_MMAP);
+ if (rc < 0)
+ break;
- if (poll(evlist->pollfd, evlist->nr_fds, timeout) > 0)
- goto again;
- } else {
- goto again;
+ if (!rc && !done)
+ err = poll(evlist->pollfd, evlist->nr_fds, 100);
}
-out_disable:
perf_evlist__disable(evlist);
if (!err) {
@@ -2153,6 +2124,8 @@ out_disable:
}
}
+out_delete_session:
+ perf_session__delete(session);
out_delete_evlist:
perf_evlist__delete(evlist);
out:
@@ -2187,17 +2160,6 @@ static int trace__replay(struct trace *trace)
struct perf_evsel *evsel;
int err = -1;
- trace->tool.sample = trace__process_sample;
- trace->tool.mmap = perf_event__process_mmap;
- trace->tool.mmap2 = perf_event__process_mmap2;
- trace->tool.comm = perf_event__process_comm;
- trace->tool.exit = perf_event__process_exit;
- trace->tool.fork = perf_event__process_fork;
- trace->tool.attr = perf_event__process_attr;
- trace->tool.tracing_data = perf_event__process_tracing_data;
- trace->tool.build_id = perf_event__process_build_id;
-
- trace->tool.ordered_samples = true;
trace->tool.ordering_requires_timestamps = true;
/* add tid to output */
@@ -2547,6 +2509,19 @@ int cmd_trace(int argc, const char **argv, const char
*prefix __maybe_unused)
if (!argc && target__none(&trace.opts.target))
trace.opts.target.system_wide = true;
+ trace.tool.sample = trace__process_sample;
+ trace.tool.mmap = perf_event__process_mmap;
+ trace.tool.mmap2 = perf_event__process_mmap2;
+ trace.tool.comm = perf_event__process_comm;
+ trace.tool.exit = perf_event__process_exit;
+ trace.tool.fork = perf_event__process_fork;
+ trace.tool.attr = perf_event__process_attr;
+ trace.tool.tracing_data = perf_event__process_tracing_data;
+ trace.tool.build_id = perf_event__process_build_id;
+
+ trace.tool.ordered_samples = true;
+ perf_tool__fill_defaults(&trace.tool);
+
if (input_name)
err = trace__replay(&trace);
else
--
1.8.3.2
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
