From: Kan Liang <[email protected]>
------
Changes since V4:
- Added a bug fix patch which recalculate the 'size' in
perf_mmap__push()
- Added a cleanup patch for perf_mmap__push()
- Modified the patch which introduce perf_mmap__read_init()
- Added a patch to change return value logic for perf_mmap__read_init()
- Dropped the unnecessary refcnt check in perf_mmap__read_init()
The check will be done later in perf_mmap__read_event().
- Used warn_lost_event to replace !no_lost_event_warning
Changes since V3:
- Separated patches to add new interface perf_mmap__read_init and
apply to the perf_mmap__push()
- Corrected the comments of perf_mmap__read_done()
- Name the pointer parameter with 'p' postfix
- Add new rules to check per-event overwrite term in comments.
Do the check before perf_evlist__config()
- Add a new patch to disable/enable event lost warning in hists browser.
Changes since V2:
- Move duplicate 'map->prev' out of perf_mmap__read. Modify the
perf_mmap__read_event accordingly.
- Introduce new interface perf_mmap__read_init to calculate the ringbuffer
position
- Check perf_missing_features.write_backward
- Discard stale interfaces perf_mmap__read_backward and
perf_mmap__read_catchup
Changes since V1:
- New patches 4-6
- Support both overwrite mode and non-overwrite mode.
If kernel doesn't support default overwrite mode, fall back to
non-overwrite mode.
------
perf_top__mmap_read has severe performance issue in
Knights Landing/Mill, when monitoring in heavy load system. It costs
several minutes to finish, which is unacceptable.
Currently, perf top is non overwrite mode. For non overwrite mode, it
tries to read everything in the ringbuffer and doesn't pause the
ringbuffer. Once there are lots of samples delivered persistently,
the processing time could be very long. Also, the latest samples could
be lost when the ringbuffer is full.
It's better to change it to overwrite mode, which takes a snapshot for
the system by pausing the ringbuffer and could significantly reducing
the processing time (from several minutes to several seconds).
Also, the overwrite mode always keep the latest samples.
Patch 1-10: Introduce new interfaces for generic code to support
overwrite mode for one by one event read.
Discards stale interfaces.
The patches can be merged separately.
Patch 11-17: Add overwrite support to perf top.
Perf top should only support either overwrite or
non-overwrite mode.
Switch default mode to overwrite mode
If kernel doesn't support overwrite mode, fall back to
non-overwrite mode.
Kan Liang (17):
perf evlist: remove stale mmap read for backward
perf mmap: fix: recalculate size for overwrite mode.
perf mmap: cleanup perf_mmap__push()
perf mmap: introduce perf_mmap__read_init()
perf mmap: add new return value logic for perf_mmap__read_init()
perf mmap: discard 'prev' in perf_mmap__read()
perf mmap: introduce perf_mmap__read_done
perf mmap: introduce perf_mmap__read_event()
perf test: update mmap read functions for backward-ring-buffer test
perf mmap: discard legacy interface for mmap read
perf top: check per-event overwrite term
perf evsel: expose perf_missing_features.write_backward
perf top: add overwrite fall back
perf hists browser: add parameter to disable lost event warning
perf top: remove lost events checking
perf top: switch default mode to overwrite mode
perf top: check the latency of perf_top__mmap_read
tools/perf/builtin-c2c.c | 4 +-
tools/perf/builtin-report.c | 3 +-
tools/perf/builtin-top.c | 150 ++++++++++++++++++++++++++++++--
tools/perf/tests/backward-ring-buffer.c | 7 +-
tools/perf/ui/browsers/hists.c | 38 +++++---
tools/perf/ui/browsers/hists.h | 3 +-
tools/perf/util/evlist.c | 17 ----
tools/perf/util/evlist.h | 4 -
tools/perf/util/evsel.c | 5 ++
tools/perf/util/evsel.h | 2 +
tools/perf/util/hist.h | 6 +-
tools/perf/util/mmap.c | 141 +++++++++++++++++-------------
tools/perf/util/mmap.h | 10 ++-
13 files changed, 277 insertions(+), 113 deletions(-)
--
2.5.5
From: Kan Liang <[email protected]>
perf_evlist__mmap_read_catchup() and perf_evlist__mmap_read_backward()
are only for overwrite mode.
But they read the evlist->mmap buffer which is for non-overwrite mode.
It did not bring any serious problem yet, because there is no one use
it.
Remove the unused interfaces.
Acked-by: Wang Nan <[email protected]>
Signed-off-by: Kan Liang <[email protected]>
---
tools/perf/util/evlist.c | 17 -----------------
tools/perf/util/evlist.h | 4 ----
2 files changed, 21 deletions(-)
diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index f0a5e09..cf0fd35 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -714,28 +714,11 @@ union perf_event *perf_evlist__mmap_read_forward(struct perf_evlist *evlist, int
return perf_mmap__read_forward(md);
}
-union perf_event *perf_evlist__mmap_read_backward(struct perf_evlist *evlist, int idx)
-{
- struct perf_mmap *md = &evlist->mmap[idx];
-
- /*
- * No need to check messup for backward ring buffer:
- * We can always read arbitrary long data from a backward
- * ring buffer unless we forget to pause it before reading.
- */
- return perf_mmap__read_backward(md);
-}
-
union perf_event *perf_evlist__mmap_read(struct perf_evlist *evlist, int idx)
{
return perf_evlist__mmap_read_forward(evlist, idx);
}
-void perf_evlist__mmap_read_catchup(struct perf_evlist *evlist, int idx)
-{
- perf_mmap__read_catchup(&evlist->mmap[idx]);
-}
-
void perf_evlist__mmap_consume(struct perf_evlist *evlist, int idx)
{
perf_mmap__consume(&evlist->mmap[idx], false);
diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
index e7fbca6..bbf53b1 100644
--- a/tools/perf/util/evlist.h
+++ b/tools/perf/util/evlist.h
@@ -134,10 +134,6 @@ union perf_event *perf_evlist__mmap_read(struct perf_evlist *evlist, int idx);
union perf_event *perf_evlist__mmap_read_forward(struct perf_evlist *evlist,
int idx);
-union perf_event *perf_evlist__mmap_read_backward(struct perf_evlist *evlist,
- int idx);
-void perf_evlist__mmap_read_catchup(struct perf_evlist *evlist, int idx);
-
void perf_evlist__mmap_consume(struct perf_evlist *evlist, int idx);
int perf_evlist__open(struct perf_evlist *evlist);
--
2.5.5
From: Kan Liang <[email protected]>
In perf_mmap__push(), the 'size' need to be recalculated, otherwise the
invalid data might be pushed to the record in overwrite mode.
The issue is introduced by commit 7fb4b407a124 ("perf mmap: Don't
discard prev in backward mode").
When the ringbuffer is full in overwrite mode, backward_rb_find_range
will be called to recalculate the 'start' and 'end'. The 'size' needs to
be recalculated accordingly.
Unconditionally recalculate the 'size', not just for full ringbuffer in
overwrite mode. Because,
- There is no harmful to recalculate the 'size' for other cases.
- The code of calculating 'start' and 'end' will be factored out later.
The new function does not need to return 'size'.
Signed-off-by: Kan Liang <[email protected]>
---
tools/perf/util/mmap.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index 05076e6..97cf4fa 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -302,6 +302,8 @@ int perf_mmap__push(struct perf_mmap *md, bool overwrite,
return -1;
}
+ size = end - start;
+
if ((start & md->mask) + size != (end & md->mask)) {
buf = &data[start & md->mask];
size = md->mask + 1 - (start & md->mask);
--
2.5.5
From: Kan Liang <[email protected]>
The new function perf_mmap__read_init() is factored out from
perf_mmap__push().
It is to calculate the 'start' and 'end' of the available data in
ringbuffer.
No functional change.
Signed-off-by: Kan Liang <[email protected]>
---
tools/perf/util/mmap.c | 37 +++++++++++++++++++++++++++----------
tools/perf/util/mmap.h | 2 ++
2 files changed, 29 insertions(+), 10 deletions(-)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index fbbbe87..c19a4e6 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -267,24 +267,24 @@ static int overwrite_rb_find_range(void *buf, int mask, u64 head, u64 *start, u6
return -1;
}
-int perf_mmap__push(struct perf_mmap *md, bool overwrite,
- void *to, int push(void *to, void *buf, size_t size))
+/*
+ * Report the start and end of the available data in ringbuffer
+ */
+int perf_mmap__read_init(struct perf_mmap *md, bool overwrite,
+ u64 *startp, u64 *endp)
{
u64 head = perf_mmap__read_head(md);
u64 old = md->prev;
- u64 end, start;
unsigned char *data = md->base + page_size;
unsigned long size;
- void *buf;
- int rc = 0;
- start = overwrite ? head : old;
- end = overwrite ? old : head;
+ *startp = overwrite ? head : old;
+ *endp = overwrite ? old : head;
- if (start == end)
+ if (*startp == *endp)
return 0;
- size = end - start;
+ size = *endp - *startp;
if (size > (unsigned long)(md->mask) + 1) {
if (!overwrite) {
WARN_ONCE(1, "failed to keep up with mmap data. (warn only once)\n");
@@ -298,10 +298,27 @@ int perf_mmap__push(struct perf_mmap *md, bool overwrite,
* Backward ring buffer is full. We still have a chance to read
* most of data from it.
*/
- if (overwrite_rb_find_range(data, md->mask, head, &start, &end))
+ if (overwrite_rb_find_range(data, md->mask, head, startp, endp))
return -1;
}
+ return 1;
+}
+
+int perf_mmap__push(struct perf_mmap *md, bool overwrite,
+ void *to, int push(void *to, void *buf, size_t size))
+{
+ u64 head = perf_mmap__read_head(md);
+ u64 end, start;
+ unsigned char *data = md->base + page_size;
+ unsigned long size;
+ void *buf;
+ int rc = 0;
+
+ rc = perf_mmap__read_init(md, overwrite, &start, &end);
+ if (rc < 1)
+ return rc;
+
size = end - start;
if ((start & md->mask) + size != (end & md->mask)) {
diff --git a/tools/perf/util/mmap.h b/tools/perf/util/mmap.h
index e43d7b5..9ab2b48 100644
--- a/tools/perf/util/mmap.h
+++ b/tools/perf/util/mmap.h
@@ -94,4 +94,6 @@ int perf_mmap__push(struct perf_mmap *md, bool backward,
size_t perf_mmap__mmap_len(struct perf_mmap *map);
+int perf_mmap__read_init(struct perf_mmap *md, bool overwrite,
+ u64 *startp, u64 *endp);
#endif /*__PERF_MMAP_H */
--
2.5.5
From: Kan Liang <[email protected]>
The first assignment for 'start' and 'end' is redundant.
Signed-off-by: Kan Liang <[email protected]>
---
tools/perf/util/mmap.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index 97cf4fa..fbbbe87 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -272,7 +272,7 @@ int perf_mmap__push(struct perf_mmap *md, bool overwrite,
{
u64 head = perf_mmap__read_head(md);
u64 old = md->prev;
- u64 end = head, start = old;
+ u64 end, start;
unsigned char *data = md->base + page_size;
unsigned long size;
void *buf;
--
2.5.5
From: Kan Liang <[email protected]>
Switch to non-overwrite mode if kernel doesnot support overwrite
ringbuffer.
It's only effect when overwrite mode is supported.
No change to current behavior.
Signed-off-by: Kan Liang <[email protected]>
---
tools/perf/builtin-top.c | 36 ++++++++++++++++++++++++++++++++++++
1 file changed, 36 insertions(+)
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index c86eee4..ecba2cd 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -943,6 +943,27 @@ static int perf_top_overwrite_check(struct perf_top *top)
return 0;
}
+static int perf_top_overwrite_fallback(struct perf_top *top,
+ struct perf_evsel *evsel)
+{
+ struct record_opts *opts = &top->record_opts;
+ struct perf_evlist *evlist = top->evlist;
+ struct perf_evsel *counter;
+
+ if (!opts->overwrite)
+ return 0;
+
+ /* only fall back when first event fails */
+ if (evsel != perf_evlist__first(evlist))
+ return 0;
+
+ evlist__for_each_entry(evlist, counter)
+ counter->attr.write_backward = false;
+ opts->overwrite = false;
+ ui__warning("fall back to non-overwrite mode\n");
+ return 1;
+}
+
static int perf_top__start_counters(struct perf_top *top)
{
char msg[BUFSIZ];
@@ -967,6 +988,21 @@ static int perf_top__start_counters(struct perf_top *top)
try_again:
if (perf_evsel__open(counter, top->evlist->cpus,
top->evlist->threads) < 0) {
+
+ /*
+ * Specially handle overwrite fall back.
+ * Because perf top is the only tool which has
+ * overwrite mode by default, support
+ * both overwrite and non-overwrite mode, and
+ * require consistent mode for all events.
+ *
+ * May move it to generic code with more tools
+ * have similar attribute.
+ */
+ if (is_write_backward_fail() &&
+ perf_top_overwrite_fallback(top, counter))
+ goto try_again;
+
if (perf_evsel__fallback(counter, errno, msg, sizeof(msg))) {
if (verbose > 0)
ui__warning("%s\n", msg);
--
2.5.5
From: Kan Liang <[email protected]>
There would be some records lost in overwrite mode because of pausing
the ringbuffer. It has little impact for the accuracy of the snapshot
and could be tolerant for perf top.
The lost events checking is removed.
Signed-off-by: Kan Liang <[email protected]>
---
tools/perf/builtin-top.c | 8 +++++---
1 file changed, 5 insertions(+), 3 deletions(-)
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index c559107..1dd3b63 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -283,8 +283,9 @@ static void perf_top__print_sym_table(struct perf_top *top)
printf("%-*.*s\n", win_width, win_width, graph_dotted_line);
- if (hists->stats.nr_lost_warned !=
- hists->stats.nr_events[PERF_RECORD_LOST]) {
+ if (!top->record_opts.overwrite &&
+ (hists->stats.nr_lost_warned !=
+ hists->stats.nr_events[PERF_RECORD_LOST])) {
hists->stats.nr_lost_warned =
hists->stats.nr_events[PERF_RECORD_LOST];
color_fprintf(stdout, PERF_COLOR_RED,
@@ -611,7 +612,8 @@ static void *display_thread_tui(void *arg)
perf_evlist__tui_browse_hists(top->evlist, help, &hbt,
top->min_percent,
- &top->session->header.env, true);
+ &top->session->header.env,
+ !top->record_opts.overwrite);
done = 1;
return NULL;
--
2.5.5
From: Kan Liang <[email protected]>
Improve the readability by using meaningful enum (-EAGAIN, -EINVAL and
0) to replace the three returning states (0, -1 and 1).
Suggested-by: Wang Nan <[email protected]>
Signed-off-by: Kan Liang <[email protected]>
---
tools/perf/util/mmap.c | 12 ++++++------
1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index c19a4e6..38fa69d 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -282,7 +282,7 @@ int perf_mmap__read_init(struct perf_mmap *md, bool overwrite,
*endp = overwrite ? old : head;
if (*startp == *endp)
- return 0;
+ return -EAGAIN;
size = *endp - *startp;
if (size > (unsigned long)(md->mask) + 1) {
@@ -291,7 +291,7 @@ int perf_mmap__read_init(struct perf_mmap *md, bool overwrite,
md->prev = head;
perf_mmap__consume(md, overwrite);
- return 0;
+ return -EAGAIN;
}
/*
@@ -299,10 +299,10 @@ int perf_mmap__read_init(struct perf_mmap *md, bool overwrite,
* most of data from it.
*/
if (overwrite_rb_find_range(data, md->mask, head, startp, endp))
- return -1;
+ return -EINVAL;
}
- return 1;
+ return 0;
}
int perf_mmap__push(struct perf_mmap *md, bool overwrite,
@@ -316,8 +316,8 @@ int perf_mmap__push(struct perf_mmap *md, bool overwrite,
int rc = 0;
rc = perf_mmap__read_init(md, overwrite, &start, &end);
- if (rc < 1)
- return rc;
+ if (rc < 0)
+ return (rc == -EAGAIN) ? 0 : -1;
size = end - start;
--
2.5.5
From: Kan Liang <[email protected]>
The latency of perf_top__mmap_read should be lower than refresh time.
If not, give some hints to reduce the latency.
Signed-off-by: Kan Liang <[email protected]>
---
tools/perf/builtin-top.c | 9 +++++++++
1 file changed, 9 insertions(+)
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index 465a699..79ed39f 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -889,8 +889,10 @@ static void perf_top__mmap_read(struct perf_top *top)
{
bool overwrite = top->record_opts.overwrite;
struct perf_evlist *evlist = top->evlist;
+ unsigned long long start, end;
int i;
+ start = rdclock();
if (overwrite)
perf_evlist__toggle_bkw_mmap(evlist, BKW_MMAP_DATA_PENDING);
@@ -901,6 +903,13 @@ static void perf_top__mmap_read(struct perf_top *top)
perf_evlist__toggle_bkw_mmap(evlist, BKW_MMAP_EMPTY);
perf_evlist__toggle_bkw_mmap(evlist, BKW_MMAP_RUNNING);
}
+ end = rdclock();
+
+ if ((end - start) > (unsigned long long)top->delay_secs * NSEC_PER_SEC)
+ ui__warning("Too slow to read ring buffer.\n"
+ "Please try increasing the period (-c) or\n"
+ "decreasing the freq (-F) or\n"
+ "limiting the number of CPUs (-C)\n");
}
/*
--
2.5.5
From: Kan Liang <[email protected]>
Per-event overwrite term is not forbidden in perf top, which can bring
problems. Because perf top only support non-overwrite mode now.
Add new rules and check regarding to overwrite term for perf top.
- All events either have same per-event term or don't have per-event
mode setting. Otherwise, it will error out.
- Per-event overwrite term should be consistent as opts->overwrite.
If not, updating the opts->overwrite according to per-event term.
Make it possible to support either non-overwrite or overwrite mode.
The overwrite mode is forbidden now, which will be removed when the
overwrite mode is supported later.
Signed-off-by: Kan Liang <[email protected]>
---
tools/perf/builtin-top.c | 73 ++++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 73 insertions(+)
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index c6ccda5..c86eee4 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -881,6 +881,68 @@ static void perf_top__mmap_read(struct perf_top *top)
perf_top__mmap_read_idx(top, i);
}
+/*
+ * Check per-event overwrite term.
+ * perf top should support consistent term for all events.
+ * - All events don't have per-event term
+ * E.g. "cpu/cpu-cycles/,cpu/instructions/"
+ * Nothing change, return 0.
+ * - All events have same per-event term
+ * E.g. "cpu/cpu-cycles,no-overwrite/,cpu/instructions,no-overwrite/
+ * Using the per-event setting to replace the opts->overwrite if
+ * they are different, then return 0.
+ * - Events have different per-event term
+ * E.g. "cpu/cpu-cycles,overwrite/,cpu/instructions,no-overwrite/"
+ * Return -1
+ * - Some of the event set per-event term, but some not.
+ * E.g. "cpu/cpu-cycles/,cpu/instructions,no-overwrite/"
+ * Return -1
+ */
+static int perf_top_overwrite_check(struct perf_top *top)
+{
+ struct record_opts *opts = &top->record_opts;
+ struct perf_evlist *evlist = top->evlist;
+ struct perf_evsel_config_term *term;
+ struct list_head *config_terms;
+ struct perf_evsel *evsel;
+ int set, overwrite = -1;
+
+ evlist__for_each_entry(evlist, evsel) {
+ set = -1;
+ config_terms = &evsel->config_terms;
+ list_for_each_entry(term, config_terms, list) {
+ if (term->type == PERF_EVSEL__CONFIG_TERM_OVERWRITE)
+ set = term->val.overwrite ? 1 : 0;
+ }
+
+ /* no term for current and previous event (likely) */
+ if ((overwrite < 0) && (set < 0))
+ continue;
+
+ /* has term for both current and previous event, compare */
+ if ((overwrite >= 0) && (set >= 0) && (overwrite != set))
+ return -1;
+
+ /* no term for current event but has term for previous one */
+ if ((overwrite >= 0) && (set < 0))
+ return -1;
+
+ /* has term for current event */
+ if ((overwrite < 0) && (set >= 0)) {
+ /* if it's first event, set overwrite */
+ if (evsel == perf_evlist__first(evlist))
+ overwrite = set;
+ else
+ return -1;
+ }
+ }
+
+ if ((overwrite >= 0) && (opts->overwrite != overwrite))
+ opts->overwrite = overwrite;
+
+ return 0;
+}
+
static int perf_top__start_counters(struct perf_top *top)
{
char msg[BUFSIZ];
@@ -888,6 +950,17 @@ static int perf_top__start_counters(struct perf_top *top)
struct perf_evlist *evlist = top->evlist;
struct record_opts *opts = &top->record_opts;
+ if (perf_top_overwrite_check(top)) {
+ ui__error("perf top only support consistent per-event "
+ "overwrite setting for all events\n");
+ goto out_err;
+ }
+
+ if (opts->overwrite) {
+ ui__error("not support overwrite mode yet\n");
+ goto out_err;
+ }
+
perf_evlist__config(evlist, opts, &callchain_param);
evlist__for_each_entry(evlist, counter) {
--
2.5.5
From: Kan Liang <[email protected]>
For overwrite mode, the ringbuffer will be paused. The event lost is
expected. It needs a way to notify the browser not print the warning.
It will be used later for perf top to disable lost event warning in
overwrite mode. There is no behavior change for now.
Signed-off-by: Kan Liang <[email protected]>
---
tools/perf/builtin-c2c.c | 4 ++--
tools/perf/builtin-report.c | 3 ++-
tools/perf/builtin-top.c | 2 +-
tools/perf/ui/browsers/hists.c | 38 +++++++++++++++++++++++++-------------
tools/perf/ui/browsers/hists.h | 3 ++-
tools/perf/util/hist.h | 6 ++++--
6 files changed, 36 insertions(+), 20 deletions(-)
diff --git a/tools/perf/builtin-c2c.c b/tools/perf/builtin-c2c.c
index c0debc3..054eb5e 100644
--- a/tools/perf/builtin-c2c.c
+++ b/tools/perf/builtin-c2c.c
@@ -2245,7 +2245,7 @@ static int perf_c2c__browse_cacheline(struct hist_entry *he)
c2c_browser__update_nr_entries(browser);
while (1) {
- key = hist_browser__run(browser, "? - help");
+ key = hist_browser__run(browser, "? - help", true);
switch (key) {
case 's':
@@ -2314,7 +2314,7 @@ static int perf_c2c__hists_browse(struct hists *hists)
c2c_browser__update_nr_entries(browser);
while (1) {
- key = hist_browser__run(browser, "? - help");
+ key = hist_browser__run(browser, "? - help", true);
switch (key) {
case 'q':
diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
index dd4df9a..9872d5b 100644
--- a/tools/perf/builtin-report.c
+++ b/tools/perf/builtin-report.c
@@ -527,7 +527,8 @@ static int report__browse_hists(struct report *rep)
case 1:
ret = perf_evlist__tui_browse_hists(evlist, help, NULL,
rep->min_percent,
- &session->header.env);
+ &session->header.env,
+ true);
/*
* Usually "ret" is the last pressed key, and we only
* care if the key notifies us to switch data file.
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index ecba2cd..c559107 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -611,7 +611,7 @@ static void *display_thread_tui(void *arg)
perf_evlist__tui_browse_hists(top->evlist, help, &hbt,
top->min_percent,
- &top->session->header.env);
+ &top->session->header.env, true);
done = 1;
return NULL;
diff --git a/tools/perf/ui/browsers/hists.c b/tools/perf/ui/browsers/hists.c
index 68146f4..6495ee5 100644
--- a/tools/perf/ui/browsers/hists.c
+++ b/tools/perf/ui/browsers/hists.c
@@ -608,7 +608,8 @@ static int hist_browser__title(struct hist_browser *browser, char *bf, size_t si
return browser->title ? browser->title(browser, bf, size) : 0;
}
-int hist_browser__run(struct hist_browser *browser, const char *help)
+int hist_browser__run(struct hist_browser *browser, const char *help,
+ bool warn_lost_event)
{
int key;
char title[160];
@@ -638,8 +639,9 @@ int hist_browser__run(struct hist_browser *browser, const char *help)
nr_entries = hist_browser__nr_entries(browser);
ui_browser__update_nr_entries(&browser->b, nr_entries);
- if (browser->hists->stats.nr_lost_warned !=
- browser->hists->stats.nr_events[PERF_RECORD_LOST]) {
+ if (warn_lost_event &&
+ (browser->hists->stats.nr_lost_warned !=
+ browser->hists->stats.nr_events[PERF_RECORD_LOST])) {
browser->hists->stats.nr_lost_warned =
browser->hists->stats.nr_events[PERF_RECORD_LOST];
ui_browser__warn_lost_events(&browser->b);
@@ -2763,7 +2765,8 @@ static int perf_evsel__hists_browse(struct perf_evsel *evsel, int nr_events,
bool left_exits,
struct hist_browser_timer *hbt,
float min_pcnt,
- struct perf_env *env)
+ struct perf_env *env,
+ bool warn_lost_event)
{
struct hists *hists = evsel__hists(evsel);
struct hist_browser *browser = perf_evsel_browser__new(evsel, hbt, env);
@@ -2844,7 +2847,8 @@ static int perf_evsel__hists_browse(struct perf_evsel *evsel, int nr_events,
nr_options = 0;
- key = hist_browser__run(browser, helpline);
+ key = hist_browser__run(browser, helpline,
+ warn_lost_event);
if (browser->he_selection != NULL) {
thread = hist_browser__selected_thread(browser);
@@ -3184,7 +3188,8 @@ static void perf_evsel_menu__write(struct ui_browser *browser,
static int perf_evsel_menu__run(struct perf_evsel_menu *menu,
int nr_events, const char *help,
- struct hist_browser_timer *hbt)
+ struct hist_browser_timer *hbt,
+ bool warn_lost_event)
{
struct perf_evlist *evlist = menu->b.priv;
struct perf_evsel *pos;
@@ -3203,7 +3208,9 @@ static int perf_evsel_menu__run(struct perf_evsel_menu *menu,
case K_TIMER:
hbt->timer(hbt->arg);
- if (!menu->lost_events_warned && menu->lost_events) {
+ if (!menu->lost_events_warned &&
+ menu->lost_events &&
+ warn_lost_event) {
ui_browser__warn_lost_events(&menu->b);
menu->lost_events_warned = true;
}
@@ -3224,7 +3231,8 @@ static int perf_evsel_menu__run(struct perf_evsel_menu *menu,
key = perf_evsel__hists_browse(pos, nr_events, help,
true, hbt,
menu->min_pcnt,
- menu->env);
+ menu->env,
+ warn_lost_event);
ui_browser__show_title(&menu->b, title);
switch (key) {
case K_TAB:
@@ -3282,7 +3290,8 @@ static int __perf_evlist__tui_browse_hists(struct perf_evlist *evlist,
int nr_entries, const char *help,
struct hist_browser_timer *hbt,
float min_pcnt,
- struct perf_env *env)
+ struct perf_env *env,
+ bool warn_lost_event)
{
struct perf_evsel *pos;
struct perf_evsel_menu menu = {
@@ -3309,13 +3318,15 @@ static int __perf_evlist__tui_browse_hists(struct perf_evlist *evlist,
menu.b.width = line_len;
}
- return perf_evsel_menu__run(&menu, nr_entries, help, hbt);
+ return perf_evsel_menu__run(&menu, nr_entries, help,
+ hbt, warn_lost_event);
}
int perf_evlist__tui_browse_hists(struct perf_evlist *evlist, const char *help,
struct hist_browser_timer *hbt,
float min_pcnt,
- struct perf_env *env)
+ struct perf_env *env,
+ bool warn_lost_event)
{
int nr_entries = evlist->nr_entries;
@@ -3325,7 +3336,7 @@ int perf_evlist__tui_browse_hists(struct perf_evlist *evlist, const char *help,
return perf_evsel__hists_browse(first, nr_entries, help,
false, hbt, min_pcnt,
- env);
+ env, warn_lost_event);
}
if (symbol_conf.event_group) {
@@ -3342,5 +3353,6 @@ int perf_evlist__tui_browse_hists(struct perf_evlist *evlist, const char *help,
}
return __perf_evlist__tui_browse_hists(evlist, nr_entries, help,
- hbt, min_pcnt, env);
+ hbt, min_pcnt, env,
+ warn_lost_event);
}
diff --git a/tools/perf/ui/browsers/hists.h b/tools/perf/ui/browsers/hists.h
index ba43177..9428bee 100644
--- a/tools/perf/ui/browsers/hists.h
+++ b/tools/perf/ui/browsers/hists.h
@@ -28,7 +28,8 @@ struct hist_browser {
struct hist_browser *hist_browser__new(struct hists *hists);
void hist_browser__delete(struct hist_browser *browser);
-int hist_browser__run(struct hist_browser *browser, const char *help);
+int hist_browser__run(struct hist_browser *browser, const char *help,
+ bool warn_lost_event);
void hist_browser__init(struct hist_browser *browser,
struct hists *hists);
#endif /* _PERF_UI_BROWSER_HISTS_H_ */
diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h
index f6630cb..02721b57 100644
--- a/tools/perf/util/hist.h
+++ b/tools/perf/util/hist.h
@@ -430,7 +430,8 @@ int hist_entry__tui_annotate(struct hist_entry *he, struct perf_evsel *evsel,
int perf_evlist__tui_browse_hists(struct perf_evlist *evlist, const char *help,
struct hist_browser_timer *hbt,
float min_pcnt,
- struct perf_env *env);
+ struct perf_env *env,
+ bool warn_lost_event);
int script_browse(const char *script_opt);
#else
static inline
@@ -438,7 +439,8 @@ int perf_evlist__tui_browse_hists(struct perf_evlist *evlist __maybe_unused,
const char *help __maybe_unused,
struct hist_browser_timer *hbt __maybe_unused,
float min_pcnt __maybe_unused,
- struct perf_env *env __maybe_unused)
+ struct perf_env *env __maybe_unused,
+ bool warn_lost_event __maybe_unused)
{
return 0;
}
--
2.5.5
From: Kan Liang <[email protected]>
perf_top__mmap_read has severe performance issue in
Knights Landing/Mill, when monitoring in heavy load system. It costs
several minutes to finish, which is unacceptable.
Currently, perf top is non overwrite mode. For non overwrite mode, it
tries to read everything in the ringbuffer and doesn't pause the
ringbuffer. Once there are lots of samples delivered persistently,
the processing time could be very long. Also, the latest samples could
be lost when the ringbuffer is full.
For overwrite mode, it takes a snapshot for the system by pausing the
ringbuffer, which could significantly reducing the processing time.
Also, the overwrite mode always keep the latest samples.
Considering the real time requirement for perf top, the overwrite mode
is more suitable for perf top.
Actually, perf top was overwrite mode. It is changed to non overwrite
mode since commit 93fc64f14472 ("perf top: Switch to non overwrite
mode"). It's better to change it back to overwrite mode by default.
For the kernel which doesn't support overwrite mode, it will fall back
to non overwrite mode.
There would be some records lost in overwrite mode because of pausing
the ringbuffer. It has little impact for the accuracy of the snapshot
and could be tolerant.
For overwrite mode, unconditionally wait 100 ms before each snapshot. It
also reduce the overhead caused by pausing ringbuffer, especially on
light load system.
Signed-off-by: Kan Liang <[email protected]>
---
tools/perf/builtin-top.c | 34 +++++++++++++++++++++++++---------
1 file changed, 25 insertions(+), 9 deletions(-)
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index 1dd3b63..465a699 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -809,15 +809,23 @@ static void perf_event__process_sample(struct perf_tool *tool,
static void perf_top__mmap_read_idx(struct perf_top *top, int idx)
{
+ struct record_opts *opts = &top->record_opts;
+ struct perf_evlist *evlist = top->evlist;
struct perf_sample sample;
struct perf_evsel *evsel;
+ struct perf_mmap *md;
struct perf_session *session = top->session;
union perf_event *event;
struct machine *machine;
+ u64 end, start;
int ret;
- while ((event = perf_evlist__mmap_read(top->evlist, idx)) != NULL) {
- ret = perf_evlist__parse_sample(top->evlist, event, &sample);
+ md = opts->overwrite ? &evlist->overwrite_mmap[idx] : &evlist->mmap[idx];
+ if (perf_mmap__read_init(md, opts->overwrite, &start, &end) < 0)
+ return;
+
+ while ((event = perf_mmap__read_event(md, opts->overwrite, &start, end)) != NULL) {
+ ret = perf_evlist__parse_sample(evlist, event, &sample);
if (ret) {
pr_err("Can't parse sample, err = %d\n", ret);
goto next_event;
@@ -871,16 +879,28 @@ static void perf_top__mmap_read_idx(struct perf_top *top, int idx)
} else
++session->evlist->stats.nr_unknown_events;
next_event:
- perf_evlist__mmap_consume(top->evlist, idx);
+ perf_mmap__consume(md, opts->overwrite);
}
+
+ perf_mmap__read_done(md);
}
static void perf_top__mmap_read(struct perf_top *top)
{
+ bool overwrite = top->record_opts.overwrite;
+ struct perf_evlist *evlist = top->evlist;
int i;
+ if (overwrite)
+ perf_evlist__toggle_bkw_mmap(evlist, BKW_MMAP_DATA_PENDING);
+
for (i = 0; i < top->evlist->nr_mmaps; i++)
perf_top__mmap_read_idx(top, i);
+
+ if (overwrite) {
+ perf_evlist__toggle_bkw_mmap(evlist, BKW_MMAP_EMPTY);
+ perf_evlist__toggle_bkw_mmap(evlist, BKW_MMAP_RUNNING);
+ }
}
/*
@@ -979,11 +999,6 @@ static int perf_top__start_counters(struct perf_top *top)
goto out_err;
}
- if (opts->overwrite) {
- ui__error("not support overwrite mode yet\n");
- goto out_err;
- }
-
perf_evlist__config(evlist, opts, &callchain_param);
evlist__for_each_entry(evlist, counter) {
@@ -1144,7 +1159,7 @@ static int __cmd_top(struct perf_top *top)
perf_top__mmap_read(top);
- if (hits == top->samples)
+ if (opts->overwrite || (hits == top->samples))
ret = perf_evlist__poll(top->evlist, 100);
if (resize) {
@@ -1238,6 +1253,7 @@ int cmd_top(int argc, const char **argv)
.uses_mmap = true,
},
.proc_map_timeout = 500,
+ .overwrite = 1,
},
.max_stack = sysctl_perf_event_max_stack,
.sym_pcnt_filter = 5,
--
2.5.5
From: Kan Liang <[email protected]>
Discards perf_mmap__read_backward() and perf_mmap__read_catchup(). No
tools use them.
There are tools still use perf_mmap__read_forward(). Keep it, but add
comments to point to the new interface for future use.
Signed-off-by: Kan Liang <[email protected]>
---
tools/perf/util/mmap.c | 50 ++++----------------------------------------------
tools/perf/util/mmap.h | 3 ---
2 files changed, 4 insertions(+), 49 deletions(-)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index f804926..91531a7 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -63,6 +63,10 @@ static union perf_event *perf_mmap__read(struct perf_mmap *map,
return event;
}
+/*
+ * legacy interface for mmap read.
+ * Don't use it. Use perf_mmap__read_event().
+ */
union perf_event *perf_mmap__read_forward(struct perf_mmap *map)
{
u64 head;
@@ -78,41 +82,6 @@ union perf_event *perf_mmap__read_forward(struct perf_mmap *map)
return perf_mmap__read(map, &map->prev, head);
}
-union perf_event *perf_mmap__read_backward(struct perf_mmap *map)
-{
- u64 head, end;
-
- /*
- * Check if event was unmapped due to a POLLHUP/POLLERR.
- */
- if (!refcount_read(&map->refcnt))
- return NULL;
-
- head = perf_mmap__read_head(map);
- if (!head)
- return NULL;
-
- /*
- * 'head' pointer starts from 0. Kernel minus sizeof(record) form
- * it each time when kernel writes to it, so in fact 'head' is
- * negative. 'end' pointer is made manually by adding the size of
- * the ring buffer to 'head' pointer, means the validate data can
- * read is the whole ring buffer. If 'end' is positive, the ring
- * buffer has not fully filled, so we must adjust 'end' to 0.
- *
- * However, since both 'head' and 'end' is unsigned, we can't
- * simply compare 'end' against 0. Here we compare '-head' and
- * the size of the ring buffer, where -head is the number of bytes
- * kernel write to the ring buffer.
- */
- if (-head < (u64)(map->mask + 1))
- end = 0;
- else
- end = head + map->mask + 1;
-
- return perf_mmap__read(map, &map->prev, end);
-}
-
/*
* Read event from ring buffer one by one.
* Return one event for each call.
@@ -152,17 +121,6 @@ union perf_event *perf_mmap__read_event(struct perf_mmap *map,
return event;
}
-void perf_mmap__read_catchup(struct perf_mmap *map)
-{
- u64 head;
-
- if (!refcount_read(&map->refcnt))
- return;
-
- head = perf_mmap__read_head(map);
- map->prev = head;
-}
-
static bool perf_mmap__empty(struct perf_mmap *map)
{
return perf_mmap__read_head(map) == map->prev && !map->auxtrace_mmap.base;
diff --git a/tools/perf/util/mmap.h b/tools/perf/util/mmap.h
index 2871854..ec7d3a24 100644
--- a/tools/perf/util/mmap.h
+++ b/tools/perf/util/mmap.h
@@ -65,8 +65,6 @@ void perf_mmap__put(struct perf_mmap *map);
void perf_mmap__consume(struct perf_mmap *map, bool overwrite);
-void perf_mmap__read_catchup(struct perf_mmap *md);
-
static inline u64 perf_mmap__read_head(struct perf_mmap *mm)
{
struct perf_event_mmap_page *pc = mm->base;
@@ -87,7 +85,6 @@ static inline void perf_mmap__write_tail(struct perf_mmap *md, u64 tail)
}
union perf_event *perf_mmap__read_forward(struct perf_mmap *map);
-union perf_event *perf_mmap__read_backward(struct perf_mmap *map);
union perf_event *perf_mmap__read_event(struct perf_mmap *map,
bool overwrite,
--
2.5.5
From: Kan Liang <[email protected]>
Except perf record, other perf tools read events one by one from ring
buffer by perf_mmap__read_forward(). But it only supports non-overwrite
mode.
Introduce perf_mmap__read_event() to support both non-overwrite and
overwrite mode.
Usage:
perf_mmap__read_init()
while(event = perf_mmap__read_event()) {
//process the event
perf_mmap__consume()
}
perf_mmap__read_done()
It cannot use perf_mmap__read_backward(). Because it always read the
stale buffer which is already processed. Furthermore, the forward and
backward concepts have been removed. The perf_mmap__read_backward() will
be replaced and discarded later.
Signed-off-by: Kan Liang <[email protected]>
---
tools/perf/util/mmap.c | 39 +++++++++++++++++++++++++++++++++++++++
tools/perf/util/mmap.h | 4 ++++
2 files changed, 43 insertions(+)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index 4f59eae..f804926 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -113,6 +113,45 @@ union perf_event *perf_mmap__read_backward(struct perf_mmap *map)
return perf_mmap__read(map, &map->prev, end);
}
+/*
+ * Read event from ring buffer one by one.
+ * Return one event for each call.
+ *
+ * Usage:
+ * perf_mmap__read_init()
+ * while(event = perf_mmap__read_event()) {
+ * //process the event
+ * perf_mmap__consume()
+ * }
+ * perf_mmap__read_done()
+ */
+union perf_event *perf_mmap__read_event(struct perf_mmap *map,
+ bool overwrite,
+ u64 *startp, u64 end)
+{
+ union perf_event *event;
+
+ /*
+ * Check if event was unmapped due to a POLLHUP/POLLERR.
+ */
+ if (!refcount_read(&map->refcnt))
+ return NULL;
+
+ if (startp == NULL)
+ return NULL;
+
+ /* non-overwirte doesn't pause the ringbuffer */
+ if (!overwrite)
+ end = perf_mmap__read_head(map);
+
+ event = perf_mmap__read(map, startp, end);
+
+ if (!overwrite)
+ map->prev = *startp;
+
+ return event;
+}
+
void perf_mmap__read_catchup(struct perf_mmap *map)
{
u64 head;
diff --git a/tools/perf/util/mmap.h b/tools/perf/util/mmap.h
index 95549d4..2871854 100644
--- a/tools/perf/util/mmap.h
+++ b/tools/perf/util/mmap.h
@@ -89,6 +89,10 @@ static inline void perf_mmap__write_tail(struct perf_mmap *md, u64 tail)
union perf_event *perf_mmap__read_forward(struct perf_mmap *map);
union perf_event *perf_mmap__read_backward(struct perf_mmap *map);
+union perf_event *perf_mmap__read_event(struct perf_mmap *map,
+ bool overwrite,
+ u64 *startp, u64 end);
+
int perf_mmap__push(struct perf_mmap *md, bool backward,
void *to, int push(void *to, void *buf, size_t size));
--
2.5.5
From: Kan Liang <[email protected]>
The direction of overwrite mode is backward. The last perf_mmap__read()
will set tail to map->prev. Need to correct the map->prev to head which
is the end of next read.
It will be used later.
Signed-off-by: Kan Liang <[email protected]>
---
tools/perf/util/mmap.c | 11 +++++++++++
tools/perf/util/mmap.h | 1 +
2 files changed, 12 insertions(+)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index 125bfda..4f59eae 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -338,3 +338,14 @@ int perf_mmap__push(struct perf_mmap *md, bool overwrite,
out:
return rc;
}
+
+/*
+ * Mandatory for overwrite mode
+ * The direction of overwrite mode is backward.
+ * The last perf_mmap__read() will set tail to map->prev.
+ * Need to correct the map->prev to head which is the end of next read.
+ */
+void perf_mmap__read_done(struct perf_mmap *map)
+{
+ map->prev = perf_mmap__read_head(map);
+}
diff --git a/tools/perf/util/mmap.h b/tools/perf/util/mmap.h
index 9ab2b48..95549d4 100644
--- a/tools/perf/util/mmap.h
+++ b/tools/perf/util/mmap.h
@@ -96,4 +96,5 @@ size_t perf_mmap__mmap_len(struct perf_mmap *map);
int perf_mmap__read_init(struct perf_mmap *md, bool overwrite,
u64 *startp, u64 *endp);
+void perf_mmap__read_done(struct perf_mmap *map);
#endif /*__PERF_MMAP_H */
--
2.5.5
From: Kan Liang <[email protected]>
'start' and 'prev' are duplicate in perf_mmap__read()
Use 'map->prev' to replace 'start' in perf_mmap__read_*().
Suggested-by: Wang Nan <[email protected]>
Signed-off-by: Kan Liang <[email protected]>
---
tools/perf/util/mmap.c | 28 ++++++++++------------------
1 file changed, 10 insertions(+), 18 deletions(-)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index 38fa69d..125bfda 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -22,29 +22,27 @@ size_t perf_mmap__mmap_len(struct perf_mmap *map)
/* When check_messup is true, 'end' must points to a good entry */
static union perf_event *perf_mmap__read(struct perf_mmap *map,
- u64 start, u64 end, u64 *prev)
+ u64 *startp, u64 end)
{
unsigned char *data = map->base + page_size;
union perf_event *event = NULL;
- int diff = end - start;
+ int diff = end - *startp;
if (diff >= (int)sizeof(event->header)) {
size_t size;
- event = (union perf_event *)&data[start & map->mask];
+ event = (union perf_event *)&data[*startp & map->mask];
size = event->header.size;
- if (size < sizeof(event->header) || diff < (int)size) {
- event = NULL;
- goto broken_event;
- }
+ if (size < sizeof(event->header) || diff < (int)size)
+ return NULL;
/*
* Event straddles the mmap boundary -- header should always
* be inside due to u64 alignment of output.
*/
- if ((start & map->mask) + size != ((start + size) & map->mask)) {
- unsigned int offset = start;
+ if ((*startp & map->mask) + size != ((*startp + size) & map->mask)) {
+ unsigned int offset = *startp;
unsigned int len = min(sizeof(*event), size), cpy;
void *dst = map->event_copy;
@@ -59,20 +57,15 @@ static union perf_event *perf_mmap__read(struct perf_mmap *map,
event = (union perf_event *)map->event_copy;
}
- start += size;
+ *startp += size;
}
-broken_event:
- if (prev)
- *prev = start;
-
return event;
}
union perf_event *perf_mmap__read_forward(struct perf_mmap *map)
{
u64 head;
- u64 old = map->prev;
/*
* Check if event was unmapped due to a POLLHUP/POLLERR.
@@ -82,13 +75,12 @@ union perf_event *perf_mmap__read_forward(struct perf_mmap *map)
head = perf_mmap__read_head(map);
- return perf_mmap__read(map, old, head, &map->prev);
+ return perf_mmap__read(map, &map->prev, head);
}
union perf_event *perf_mmap__read_backward(struct perf_mmap *map)
{
u64 head, end;
- u64 start = map->prev;
/*
* Check if event was unmapped due to a POLLHUP/POLLERR.
@@ -118,7 +110,7 @@ union perf_event *perf_mmap__read_backward(struct perf_mmap *map)
else
end = head + map->mask + 1;
- return perf_mmap__read(map, start, end, &map->prev);
+ return perf_mmap__read(map, &map->prev, end);
}
void perf_mmap__read_catchup(struct perf_mmap *map)
--
2.5.5
From: Kan Liang <[email protected]>
perf top need it to handle overwrite fallback later.
Signed-off-by: Kan Liang <[email protected]>
---
tools/perf/util/evsel.c | 5 +++++
tools/perf/util/evsel.h | 2 ++
2 files changed, 7 insertions(+)
diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
index 4eea3b4..1dd3700 100644
--- a/tools/perf/util/evsel.c
+++ b/tools/perf/util/evsel.c
@@ -1670,6 +1670,11 @@ static bool ignore_missing_thread(struct perf_evsel *evsel,
return true;
}
+bool is_write_backward_fail(void)
+{
+ return perf_missing_features.write_backward;
+}
+
int perf_evsel__open(struct perf_evsel *evsel, struct cpu_map *cpus,
struct thread_map *threads)
{
diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h
index 846e416..1f46728 100644
--- a/tools/perf/util/evsel.h
+++ b/tools/perf/util/evsel.h
@@ -448,4 +448,6 @@ int perf_event_attr__fprintf(FILE *fp, struct perf_event_attr *attr,
struct perf_env *perf_evsel__env(struct perf_evsel *evsel);
+bool is_write_backward_fail(void);
+
#endif /* __PERF_EVSEL_H */
--
2.5.5
From: Kan Liang <[email protected]>
Use the new perf_mmap__read_* interfaces for overwrite ringbuffer test.
Signed-off-by: Kan Liang <[email protected]>
---
tools/perf/tests/backward-ring-buffer.c | 7 +++++--
1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/tools/perf/tests/backward-ring-buffer.c b/tools/perf/tests/backward-ring-buffer.c
index 4035d43..e0b1b41 100644
--- a/tools/perf/tests/backward-ring-buffer.c
+++ b/tools/perf/tests/backward-ring-buffer.c
@@ -31,10 +31,12 @@ static int count_samples(struct perf_evlist *evlist, int *sample_count,
int i;
for (i = 0; i < evlist->nr_mmaps; i++) {
+ struct perf_mmap *map = &evlist->overwrite_mmap[i];
union perf_event *event;
+ u64 start, end;
- perf_mmap__read_catchup(&evlist->overwrite_mmap[i]);
- while ((event = perf_mmap__read_backward(&evlist->overwrite_mmap[i])) != NULL) {
+ perf_mmap__read_init(map, true, &start, &end);
+ while ((event = perf_mmap__read_event(map, true, &start, end)) != NULL) {
const u32 type = event->header.type;
switch (type) {
@@ -49,6 +51,7 @@ static int count_samples(struct perf_evlist *evlist, int *sample_count,
return TEST_FAIL;
}
}
+ perf_mmap__read_done(map);
}
return TEST_OK;
}
--
2.5.5
On Thu, Jan 18, 2018 at 01:26:15PM -0800, [email protected] wrote:
> From: Kan Liang <[email protected]>
>
> ------
> Changes since V4:
> - Added a bug fix patch which recalculate the 'size' in
> perf_mmap__push()
> - Added a cleanup patch for perf_mmap__push()
> - Modified the patch which introduce perf_mmap__read_init()
> - Added a patch to change return value logic for perf_mmap__read_init()
> - Dropped the unnecessary refcnt check in perf_mmap__read_init()
> The check will be done later in perf_mmap__read_event().
> - Used warn_lost_event to replace !no_lost_event_warning
Acked-by: Jiri Olsa <[email protected]>
thanks,
jirka
Em Thu, Jan 18, 2018 at 01:26:17PM -0800, [email protected] escreveu:
> From: Kan Liang <[email protected]>
>
> In perf_mmap__push(), the 'size' need to be recalculated, otherwise the
> invalid data might be pushed to the record in overwrite mode.
>
> The issue is introduced by commit 7fb4b407a124 ("perf mmap: Don't
> discard prev in backward mode").
In these cases, please add the Fixes tag, like:
Fixes: 7fb4b407a124 ("perf mmap: Don't discard prev in backward mode")
I'm doing it this time.
Thanks,
- Arnaldo
> When the ringbuffer is full in overwrite mode, backward_rb_find_range
> will be called to recalculate the 'start' and 'end'. The 'size' needs to
> be recalculated accordingly.
>
> Unconditionally recalculate the 'size', not just for full ringbuffer in
> overwrite mode. Because,
> - There is no harmful to recalculate the 'size' for other cases.
> - The code of calculating 'start' and 'end' will be factored out later.
> The new function does not need to return 'size'.
>
> Signed-off-by: Kan Liang <[email protected]>
> ---
> tools/perf/util/mmap.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
> index 05076e6..97cf4fa 100644
> --- a/tools/perf/util/mmap.c
> +++ b/tools/perf/util/mmap.c
> @@ -302,6 +302,8 @@ int perf_mmap__push(struct perf_mmap *md, bool overwrite,
> return -1;
> }
>
> + size = end - start;
> +
> if ((start & md->mask) + size != (end & md->mask)) {
> buf = &data[start & md->mask];
> size = md->mask + 1 - (start & md->mask);
> --
> 2.5.5
> Em Thu, Jan 18, 2018 at 01:26:17PM -0800, [email protected] escreveu:
> > From: Kan Liang <[email protected]>
> >
> > In perf_mmap__push(), the 'size' need to be recalculated, otherwise
> > the invalid data might be pushed to the record in overwrite mode.
> >
> > The issue is introduced by commit 7fb4b407a124 ("perf mmap: Don't
> > discard prev in backward mode").
>
> In these cases, please add the Fixes tag, like:
>
> Fixes: 7fb4b407a124 ("perf mmap: Don't discard prev in backward mode")
>
> I'm doing it this time.
>
Thanks. I'll be careful next time.
Thanks,
Kan
> Thanks,
>
> - Arnaldo
>
> > When the ringbuffer is full in overwrite mode, backward_rb_find_range
> > will be called to recalculate the 'start' and 'end'. The 'size' needs
> > to be recalculated accordingly.
> >
> > Unconditionally recalculate the 'size', not just for full ringbuffer
> > in overwrite mode. Because,
> > - There is no harmful to recalculate the 'size' for other cases.
> > - The code of calculating 'start' and 'end' will be factored out later.
> > The new function does not need to return 'size'.
> >
> > Signed-off-by: Kan Liang <[email protected]>
> > ---
> > tools/perf/util/mmap.c | 2 ++
> > 1 file changed, 2 insertions(+)
> >
> > diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c index
> > 05076e6..97cf4fa 100644
> > --- a/tools/perf/util/mmap.c
> > +++ b/tools/perf/util/mmap.c
> > @@ -302,6 +302,8 @@ int perf_mmap__push(struct perf_mmap *md,
> bool overwrite,
> > return -1;
> > }
> >
> > + size = end - start;
> > +
> > if ((start & md->mask) + size != (end & md->mask)) {
> > buf = &data[start & md->mask];
> > size = md->mask + 1 - (start & md->mask);
> > --
> > 2.5.5
Em Thu, Jan 18, 2018 at 01:26:27PM -0800, [email protected] escreveu:
> From: Kan Liang <[email protected]>
>
> perf top need it to handle overwrite fallback later.
>
> Signed-off-by: Kan Liang <[email protected]>
> ---
> tools/perf/util/evsel.c | 5 +++++
> tools/perf/util/evsel.h | 2 ++
> 2 files changed, 7 insertions(+)
>
> diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c
> index 4eea3b4..1dd3700 100644
> --- a/tools/perf/util/evsel.c
> +++ b/tools/perf/util/evsel.c
> @@ -1670,6 +1670,11 @@ static bool ignore_missing_thread(struct perf_evsel *evsel,
> return true;
> }
>
> +bool is_write_backward_fail(void)
> +{
> + return perf_missing_features.write_backward;
> +}
> +
Humm, I think we better expose perf_missing_features, i.e. tools using
if (perf_missing_features.write_backward)
looks more clear than:
if (is_write_backward_fail())
This is minor, so I'm doing this myself, please holler if you disagree.
- Arnaldo
> int perf_evsel__open(struct perf_evsel *evsel, struct cpu_map *cpus,
> struct thread_map *threads)
> {
> diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h
> index 846e416..1f46728 100644
> --- a/tools/perf/util/evsel.h
> +++ b/tools/perf/util/evsel.h
> @@ -448,4 +448,6 @@ int perf_event_attr__fprintf(FILE *fp, struct perf_event_attr *attr,
>
> struct perf_env *perf_evsel__env(struct perf_evsel *evsel);
>
> +bool is_write_backward_fail(void);
> +
> #endif /* __PERF_EVSEL_H */
> --
> 2.5.5
> Em Thu, Jan 18, 2018 at 01:26:27PM -0800, [email protected] escreveu:
> > From: Kan Liang <[email protected]>
> >
> > perf top need it to handle overwrite fallback later.
> >
> > Signed-off-by: Kan Liang <[email protected]>
> > ---
> > tools/perf/util/evsel.c | 5 +++++
> > tools/perf/util/evsel.h | 2 ++
> > 2 files changed, 7 insertions(+)
> >
> > diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c index
> > 4eea3b4..1dd3700 100644
> > --- a/tools/perf/util/evsel.c
> > +++ b/tools/perf/util/evsel.c
> > @@ -1670,6 +1670,11 @@ static bool ignore_missing_thread(struct
> perf_evsel *evsel,
> > return true;
> > }
> >
> > +bool is_write_backward_fail(void)
> > +{
> > + return perf_missing_features.write_backward;
> > +}
> > +
>
> Humm, I think we better expose perf_missing_features, i.e. tools using
>
> if (perf_missing_features.write_backward)
>
It looks good.
Thanks a lot.
Kan
> looks more clear than:
>
> if (is_write_backward_fail())
>
> This is minor, so I'm doing this myself, please holler if you disagree.
>
> - Arnaldo
>
> > int perf_evsel__open(struct perf_evsel *evsel, struct cpu_map *cpus,
> > struct thread_map *threads)
> > {
> > diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h index
> > 846e416..1f46728 100644
> > --- a/tools/perf/util/evsel.h
> > +++ b/tools/perf/util/evsel.h
> > @@ -448,4 +448,6 @@ int perf_event_attr__fprintf(FILE *fp, struct
> > perf_event_attr *attr,
> >
> > struct perf_env *perf_evsel__env(struct perf_evsel *evsel);
> >
> > +bool is_write_backward_fail(void);
> > +
> > #endif /* __PERF_EVSEL_H */
> > --
> > 2.5.5
Commit-ID: ab463a252c4c790913c0ca0196e5dc7ea04b1f02
Gitweb: https://git.kernel.org/tip/ab463a252c4c790913c0ca0196e5dc7ea04b1f02
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:17 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Tue, 6 Feb 2018 10:11:49 -0300
perf mmap: Recalculate size for overwrite mode
In perf_mmap__push(), the 'size' need to be recalculated, otherwise the
invalid data might be pushed to the record in overwrite mode.
The issue is introduced by commit 7fb4b407a124 ("perf mmap: Don't
discard prev in backward mode").
When the ring buffer is full in overwrite mode, backward_rb_find_range()
will be called to recalculate the 'start' and 'end'. The 'size' needs to
be recalculated accordingly.
Unconditionally recalculate the 'size', not just for full ring buffer in
overwrite mode. Because:
- There is no harmful to recalculate the 'size' for other cases.
- The code of calculating 'start' and 'end' will be factored out later.
The new function does not need to return 'size'.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Fixes: 7fb4b407a124 ("perf mmap: Don't discard prev in backward mode")
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/util/mmap.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index 05076e6..97cf4fa 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -302,6 +302,8 @@ int perf_mmap__push(struct perf_mmap *md, bool overwrite,
return -1;
}
+ size = end - start;
+
if ((start & md->mask) + size != (end & md->mask)) {
buf = &data[start & md->mask];
size = md->mask + 1 - (start & md->mask);
Commit-ID: 0ecec70a6aae6fc128549b28b33da7f7f82b45bd
Gitweb: https://git.kernel.org/tip/0ecec70a6aae6fc128549b28b33da7f7f82b45bd
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:18 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Tue, 6 Feb 2018 10:11:49 -0300
perf mmap: Cleanup perf_mmap__push()
The first assignment for 'start' and 'end' is redundant.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/util/mmap.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index 97cf4fa..fbbbe87 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -272,7 +272,7 @@ int perf_mmap__push(struct perf_mmap *md, bool overwrite,
{
u64 head = perf_mmap__read_head(md);
u64 old = md->prev;
- u64 end = head, start = old;
+ u64 end, start;
unsigned char *data = md->base + page_size;
unsigned long size;
void *buf;
Commit-ID: d3035da2c854545f915ff5bcb54572860fe18ee7
Gitweb: https://git.kernel.org/tip/d3035da2c854545f915ff5bcb54572860fe18ee7
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:16 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Tue, 6 Feb 2018 10:11:49 -0300
perf evlist: Remove stale mmap read for backward
perf_evlist__mmap_read_catchup() and perf_evlist__mmap_read_backward()
are only for overwrite mode.
But they read the evlist->mmap buffer which is for non-overwrite mode.
It did not bring any serious problem yet, because there is no one use
it.
Remove the unused interfaces.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Acked-by: Wang Nan <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/util/evlist.c | 17 -----------------
tools/perf/util/evlist.h | 4 ----
2 files changed, 21 deletions(-)
diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index ac35cd2..e5fc14e 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -715,28 +715,11 @@ union perf_event *perf_evlist__mmap_read_forward(struct perf_evlist *evlist, int
return perf_mmap__read_forward(md);
}
-union perf_event *perf_evlist__mmap_read_backward(struct perf_evlist *evlist, int idx)
-{
- struct perf_mmap *md = &evlist->mmap[idx];
-
- /*
- * No need to check messup for backward ring buffer:
- * We can always read arbitrary long data from a backward
- * ring buffer unless we forget to pause it before reading.
- */
- return perf_mmap__read_backward(md);
-}
-
union perf_event *perf_evlist__mmap_read(struct perf_evlist *evlist, int idx)
{
return perf_evlist__mmap_read_forward(evlist, idx);
}
-void perf_evlist__mmap_read_catchup(struct perf_evlist *evlist, int idx)
-{
- perf_mmap__read_catchup(&evlist->mmap[idx]);
-}
-
void perf_evlist__mmap_consume(struct perf_evlist *evlist, int idx)
{
perf_mmap__consume(&evlist->mmap[idx], false);
diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
index 75f8e0a..336b838 100644
--- a/tools/perf/util/evlist.h
+++ b/tools/perf/util/evlist.h
@@ -133,10 +133,6 @@ union perf_event *perf_evlist__mmap_read(struct perf_evlist *evlist, int idx);
union perf_event *perf_evlist__mmap_read_forward(struct perf_evlist *evlist,
int idx);
-union perf_event *perf_evlist__mmap_read_backward(struct perf_evlist *evlist,
- int idx);
-void perf_evlist__mmap_read_catchup(struct perf_evlist *evlist, int idx);
-
void perf_evlist__mmap_consume(struct perf_evlist *evlist, int idx);
int perf_evlist__open(struct perf_evlist *evlist);
Commit-ID: e0c75323e9f7002d84abbf68650277a6d971b3c0
Gitweb: https://git.kernel.org/tip/e0c75323e9f7002d84abbf68650277a6d971b3c0
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:20 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Tue, 6 Feb 2018 10:11:49 -0300
perf mmap: Add new return value logic for perf_mmap__read_init()
Improve the readability by using meaningful enum (-EAGAIN, -EINVAL and
0) to replace the three returning states (0, -1 and 1).
Suggested-by: Wang Nan <[email protected]>
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/util/mmap.c | 12 ++++++------
1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index c19a4e6..38fa69d 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -282,7 +282,7 @@ int perf_mmap__read_init(struct perf_mmap *md, bool overwrite,
*endp = overwrite ? old : head;
if (*startp == *endp)
- return 0;
+ return -EAGAIN;
size = *endp - *startp;
if (size > (unsigned long)(md->mask) + 1) {
@@ -291,7 +291,7 @@ int perf_mmap__read_init(struct perf_mmap *md, bool overwrite,
md->prev = head;
perf_mmap__consume(md, overwrite);
- return 0;
+ return -EAGAIN;
}
/*
@@ -299,10 +299,10 @@ int perf_mmap__read_init(struct perf_mmap *md, bool overwrite,
* most of data from it.
*/
if (overwrite_rb_find_range(data, md->mask, head, startp, endp))
- return -1;
+ return -EINVAL;
}
- return 1;
+ return 0;
}
int perf_mmap__push(struct perf_mmap *md, bool overwrite,
@@ -316,8 +316,8 @@ int perf_mmap__push(struct perf_mmap *md, bool overwrite,
int rc = 0;
rc = perf_mmap__read_init(md, overwrite, &start, &end);
- if (rc < 1)
- return rc;
+ if (rc < 0)
+ return (rc == -EAGAIN) ? 0 : -1;
size = end - start;
Commit-ID: a41cf9d483fbf0a97e8b131b6a006f558fdd502d
Gitweb: https://git.kernel.org/tip/a41cf9d483fbf0a97e8b131b6a006f558fdd502d
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:21 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Tue, 6 Feb 2018 10:11:50 -0300
perf mmap: Discard 'prev' in perf_mmap__read()
The 'start' and 'prev' variables are duplicates in perf_mmap__read().
Use 'map->prev' to replace 'start' in perf_mmap__read_*().
Suggested-by: Wang Nan <[email protected]>
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/util/mmap.c | 28 ++++++++++------------------
1 file changed, 10 insertions(+), 18 deletions(-)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index 38fa69d..125bfda 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -22,29 +22,27 @@ size_t perf_mmap__mmap_len(struct perf_mmap *map)
/* When check_messup is true, 'end' must points to a good entry */
static union perf_event *perf_mmap__read(struct perf_mmap *map,
- u64 start, u64 end, u64 *prev)
+ u64 *startp, u64 end)
{
unsigned char *data = map->base + page_size;
union perf_event *event = NULL;
- int diff = end - start;
+ int diff = end - *startp;
if (diff >= (int)sizeof(event->header)) {
size_t size;
- event = (union perf_event *)&data[start & map->mask];
+ event = (union perf_event *)&data[*startp & map->mask];
size = event->header.size;
- if (size < sizeof(event->header) || diff < (int)size) {
- event = NULL;
- goto broken_event;
- }
+ if (size < sizeof(event->header) || diff < (int)size)
+ return NULL;
/*
* Event straddles the mmap boundary -- header should always
* be inside due to u64 alignment of output.
*/
- if ((start & map->mask) + size != ((start + size) & map->mask)) {
- unsigned int offset = start;
+ if ((*startp & map->mask) + size != ((*startp + size) & map->mask)) {
+ unsigned int offset = *startp;
unsigned int len = min(sizeof(*event), size), cpy;
void *dst = map->event_copy;
@@ -59,20 +57,15 @@ static union perf_event *perf_mmap__read(struct perf_mmap *map,
event = (union perf_event *)map->event_copy;
}
- start += size;
+ *startp += size;
}
-broken_event:
- if (prev)
- *prev = start;
-
return event;
}
union perf_event *perf_mmap__read_forward(struct perf_mmap *map)
{
u64 head;
- u64 old = map->prev;
/*
* Check if event was unmapped due to a POLLHUP/POLLERR.
@@ -82,13 +75,12 @@ union perf_event *perf_mmap__read_forward(struct perf_mmap *map)
head = perf_mmap__read_head(map);
- return perf_mmap__read(map, old, head, &map->prev);
+ return perf_mmap__read(map, &map->prev, head);
}
union perf_event *perf_mmap__read_backward(struct perf_mmap *map)
{
u64 head, end;
- u64 start = map->prev;
/*
* Check if event was unmapped due to a POLLHUP/POLLERR.
@@ -118,7 +110,7 @@ union perf_event *perf_mmap__read_backward(struct perf_mmap *map)
else
end = head + map->mask + 1;
- return perf_mmap__read(map, start, end, &map->prev);
+ return perf_mmap__read(map, &map->prev, end);
}
void perf_mmap__read_catchup(struct perf_mmap *map)
Commit-ID: 701b3ef37c0a4a5168bc07cfd48b35727b5a922c
Gitweb: https://git.kernel.org/tip/701b3ef37c0a4a5168bc07cfd48b35727b5a922c
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:19 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Tue, 6 Feb 2018 10:11:49 -0300
perf mmap: Introduce perf_mmap__read_init()
The new function perf_mmap__read_init() is factored out from
perf_mmap__push().
It is to calculate the 'start' and 'end' of the available data in
ringbuffer.
No functional change.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/util/mmap.c | 37 +++++++++++++++++++++++++++----------
tools/perf/util/mmap.h | 2 ++
2 files changed, 29 insertions(+), 10 deletions(-)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index fbbbe87..c19a4e6 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -267,24 +267,24 @@ static int overwrite_rb_find_range(void *buf, int mask, u64 head, u64 *start, u6
return -1;
}
-int perf_mmap__push(struct perf_mmap *md, bool overwrite,
- void *to, int push(void *to, void *buf, size_t size))
+/*
+ * Report the start and end of the available data in ringbuffer
+ */
+int perf_mmap__read_init(struct perf_mmap *md, bool overwrite,
+ u64 *startp, u64 *endp)
{
u64 head = perf_mmap__read_head(md);
u64 old = md->prev;
- u64 end, start;
unsigned char *data = md->base + page_size;
unsigned long size;
- void *buf;
- int rc = 0;
- start = overwrite ? head : old;
- end = overwrite ? old : head;
+ *startp = overwrite ? head : old;
+ *endp = overwrite ? old : head;
- if (start == end)
+ if (*startp == *endp)
return 0;
- size = end - start;
+ size = *endp - *startp;
if (size > (unsigned long)(md->mask) + 1) {
if (!overwrite) {
WARN_ONCE(1, "failed to keep up with mmap data. (warn only once)\n");
@@ -298,10 +298,27 @@ int perf_mmap__push(struct perf_mmap *md, bool overwrite,
* Backward ring buffer is full. We still have a chance to read
* most of data from it.
*/
- if (overwrite_rb_find_range(data, md->mask, head, &start, &end))
+ if (overwrite_rb_find_range(data, md->mask, head, startp, endp))
return -1;
}
+ return 1;
+}
+
+int perf_mmap__push(struct perf_mmap *md, bool overwrite,
+ void *to, int push(void *to, void *buf, size_t size))
+{
+ u64 head = perf_mmap__read_head(md);
+ u64 end, start;
+ unsigned char *data = md->base + page_size;
+ unsigned long size;
+ void *buf;
+ int rc = 0;
+
+ rc = perf_mmap__read_init(md, overwrite, &start, &end);
+ if (rc < 1)
+ return rc;
+
size = end - start;
if ((start & md->mask) + size != (end & md->mask)) {
diff --git a/tools/perf/util/mmap.h b/tools/perf/util/mmap.h
index e43d7b5..9ab2b48 100644
--- a/tools/perf/util/mmap.h
+++ b/tools/perf/util/mmap.h
@@ -94,4 +94,6 @@ int perf_mmap__push(struct perf_mmap *md, bool backward,
size_t perf_mmap__mmap_len(struct perf_mmap *map);
+int perf_mmap__read_init(struct perf_mmap *md, bool overwrite,
+ u64 *startp, u64 *endp);
#endif /*__PERF_MMAP_H */
Commit-ID: f4075778f7c9f743109e59311b1bdd60f914804e
Gitweb: https://git.kernel.org/tip/f4075778f7c9f743109e59311b1bdd60f914804e
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:22 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Tue, 6 Feb 2018 10:11:50 -0300
perf mmap: Introduce perf_mmap__read_done()
The direction of overwrite mode is backward. The last perf_mmap__read()
will set tail to map->prev. Need to correct the map->prev to head which
is the end of next read.
It will be used later.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/util/mmap.c | 11 +++++++++++
tools/perf/util/mmap.h | 1 +
2 files changed, 12 insertions(+)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index 125bfda..4f59eae 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -338,3 +338,14 @@ int perf_mmap__push(struct perf_mmap *md, bool overwrite,
out:
return rc;
}
+
+/*
+ * Mandatory for overwrite mode
+ * The direction of overwrite mode is backward.
+ * The last perf_mmap__read() will set tail to map->prev.
+ * Need to correct the map->prev to head which is the end of next read.
+ */
+void perf_mmap__read_done(struct perf_mmap *map)
+{
+ map->prev = perf_mmap__read_head(map);
+}
diff --git a/tools/perf/util/mmap.h b/tools/perf/util/mmap.h
index 9ab2b48..95549d4 100644
--- a/tools/perf/util/mmap.h
+++ b/tools/perf/util/mmap.h
@@ -96,4 +96,5 @@ size_t perf_mmap__mmap_len(struct perf_mmap *map);
int perf_mmap__read_init(struct perf_mmap *md, bool overwrite,
u64 *startp, u64 *endp);
+void perf_mmap__read_done(struct perf_mmap *map);
#endif /*__PERF_MMAP_H */
Commit-ID: 7257bb0efc6ab4b3a16e1c5154f70d30551e3a7b
Gitweb: https://git.kernel.org/tip/7257bb0efc6ab4b3a16e1c5154f70d30551e3a7b
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:23 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Tue, 6 Feb 2018 10:11:50 -0300
perf mmap: Introduce perf_mmap__read_event()
Except for 'perf record', the other perf tools read events one by one
from the ring buffer using perf_mmap__read_forward(). But it only
supports non-overwrite mode.
Introduce perf_mmap__read_event() to support both non-overwrite and
overwrite mode.
Usage:
perf_mmap__read_init()
while(event = perf_mmap__read_event()) {
//process the event
perf_mmap__consume()
}
perf_mmap__read_done()
It cannot use perf_mmap__read_backward(). Because it always reads the
stale buffer which is already processed. Furthermore, the forward and
backward concepts have been removed. The perf_mmap__read_backward() will
be replaced and discarded later.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/util/mmap.c | 39 +++++++++++++++++++++++++++++++++++++++
tools/perf/util/mmap.h | 4 ++++
2 files changed, 43 insertions(+)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index 4f59eae..f804926 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -113,6 +113,45 @@ union perf_event *perf_mmap__read_backward(struct perf_mmap *map)
return perf_mmap__read(map, &map->prev, end);
}
+/*
+ * Read event from ring buffer one by one.
+ * Return one event for each call.
+ *
+ * Usage:
+ * perf_mmap__read_init()
+ * while(event = perf_mmap__read_event()) {
+ * //process the event
+ * perf_mmap__consume()
+ * }
+ * perf_mmap__read_done()
+ */
+union perf_event *perf_mmap__read_event(struct perf_mmap *map,
+ bool overwrite,
+ u64 *startp, u64 end)
+{
+ union perf_event *event;
+
+ /*
+ * Check if event was unmapped due to a POLLHUP/POLLERR.
+ */
+ if (!refcount_read(&map->refcnt))
+ return NULL;
+
+ if (startp == NULL)
+ return NULL;
+
+ /* non-overwirte doesn't pause the ringbuffer */
+ if (!overwrite)
+ end = perf_mmap__read_head(map);
+
+ event = perf_mmap__read(map, startp, end);
+
+ if (!overwrite)
+ map->prev = *startp;
+
+ return event;
+}
+
void perf_mmap__read_catchup(struct perf_mmap *map)
{
u64 head;
diff --git a/tools/perf/util/mmap.h b/tools/perf/util/mmap.h
index 95549d4..2871854 100644
--- a/tools/perf/util/mmap.h
+++ b/tools/perf/util/mmap.h
@@ -89,6 +89,10 @@ static inline void perf_mmap__write_tail(struct perf_mmap *md, u64 tail)
union perf_event *perf_mmap__read_forward(struct perf_mmap *map);
union perf_event *perf_mmap__read_backward(struct perf_mmap *map);
+union perf_event *perf_mmap__read_event(struct perf_mmap *map,
+ bool overwrite,
+ u64 *startp, u64 end);
+
int perf_mmap__push(struct perf_mmap *md, bool backward,
void *to, int push(void *to, void *buf, size_t size));
Commit-ID: fe00d73152ed18080ea895bc84c495d3bca71019
Gitweb: https://git.kernel.org/tip/fe00d73152ed18080ea895bc84c495d3bca71019
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:24 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Tue, 6 Feb 2018 10:11:50 -0300
perf test: Update mmap read functions for backward-ring-buffer test
Use the new perf_mmap__read_* interfaces for overwrite ringbuffer test.
Commiter notes:
Testing:
[root@seventh ~]# perf test -v backward
48: Read backward ring buffer :
--- start ---
test child forked, pid 8309
Using CPUID GenuineIntel-6-9E
mmap size 1052672B
mmap size 8192B
Finished reading overwrite ring buffer: rewind
test child finished with 0
---- end ----
Read backward ring buffer: Ok
[root@seventh ~]#
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/tests/backward-ring-buffer.c | 7 +++++--
1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/tools/perf/tests/backward-ring-buffer.c b/tools/perf/tests/backward-ring-buffer.c
index 4035d43..e0b1b41 100644
--- a/tools/perf/tests/backward-ring-buffer.c
+++ b/tools/perf/tests/backward-ring-buffer.c
@@ -31,10 +31,12 @@ static int count_samples(struct perf_evlist *evlist, int *sample_count,
int i;
for (i = 0; i < evlist->nr_mmaps; i++) {
+ struct perf_mmap *map = &evlist->overwrite_mmap[i];
union perf_event *event;
+ u64 start, end;
- perf_mmap__read_catchup(&evlist->overwrite_mmap[i]);
- while ((event = perf_mmap__read_backward(&evlist->overwrite_mmap[i])) != NULL) {
+ perf_mmap__read_init(map, true, &start, &end);
+ while ((event = perf_mmap__read_event(map, true, &start, end)) != NULL) {
const u32 type = event->header.type;
switch (type) {
@@ -49,6 +51,7 @@ static int count_samples(struct perf_evlist *evlist, int *sample_count,
return TEST_FAIL;
}
}
+ perf_mmap__read_done(map);
}
return TEST_OK;
}
Commit-ID: e6b33238e03466ca9b096368226f29fb4aec627a
Gitweb: https://git.kernel.org/tip/e6b33238e03466ca9b096368226f29fb4aec627a
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:26 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Tue, 6 Feb 2018 10:11:50 -0300
perf top: Check per-event overwrite term
Per-event overwrite term is not forbidden in 'perf top', which can bring
problems. Because 'perf top' only support non-overwrite mode now.
Add new rules and check regarding to overwrite term for 'perf top'.
- All events either have same per-event term or don't have per-event
mode setting. Otherwise, it will error out.
- Per-event overwrite term should be consistent as opts->overwrite.
If not, updating the opts->overwrite according to per-event term.
Make it possible to support either non-overwrite or overwrite mode.
The overwrite mode is forbidden now, which will be removed when the
overwrite mode is supported later.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
[ Renamed perf_top_overwrite_check to perf_top__overwrite_check, to follow existing convention ]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/builtin-top.c | 73 ++++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 73 insertions(+)
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index c6ccda5..1778379 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -881,6 +881,68 @@ static void perf_top__mmap_read(struct perf_top *top)
perf_top__mmap_read_idx(top, i);
}
+/*
+ * Check per-event overwrite term.
+ * perf top should support consistent term for all events.
+ * - All events don't have per-event term
+ * E.g. "cpu/cpu-cycles/,cpu/instructions/"
+ * Nothing change, return 0.
+ * - All events have same per-event term
+ * E.g. "cpu/cpu-cycles,no-overwrite/,cpu/instructions,no-overwrite/
+ * Using the per-event setting to replace the opts->overwrite if
+ * they are different, then return 0.
+ * - Events have different per-event term
+ * E.g. "cpu/cpu-cycles,overwrite/,cpu/instructions,no-overwrite/"
+ * Return -1
+ * - Some of the event set per-event term, but some not.
+ * E.g. "cpu/cpu-cycles/,cpu/instructions,no-overwrite/"
+ * Return -1
+ */
+static int perf_top__overwrite_check(struct perf_top *top)
+{
+ struct record_opts *opts = &top->record_opts;
+ struct perf_evlist *evlist = top->evlist;
+ struct perf_evsel_config_term *term;
+ struct list_head *config_terms;
+ struct perf_evsel *evsel;
+ int set, overwrite = -1;
+
+ evlist__for_each_entry(evlist, evsel) {
+ set = -1;
+ config_terms = &evsel->config_terms;
+ list_for_each_entry(term, config_terms, list) {
+ if (term->type == PERF_EVSEL__CONFIG_TERM_OVERWRITE)
+ set = term->val.overwrite ? 1 : 0;
+ }
+
+ /* no term for current and previous event (likely) */
+ if ((overwrite < 0) && (set < 0))
+ continue;
+
+ /* has term for both current and previous event, compare */
+ if ((overwrite >= 0) && (set >= 0) && (overwrite != set))
+ return -1;
+
+ /* no term for current event but has term for previous one */
+ if ((overwrite >= 0) && (set < 0))
+ return -1;
+
+ /* has term for current event */
+ if ((overwrite < 0) && (set >= 0)) {
+ /* if it's first event, set overwrite */
+ if (evsel == perf_evlist__first(evlist))
+ overwrite = set;
+ else
+ return -1;
+ }
+ }
+
+ if ((overwrite >= 0) && (opts->overwrite != overwrite))
+ opts->overwrite = overwrite;
+
+ return 0;
+}
+
static int perf_top__start_counters(struct perf_top *top)
{
char msg[BUFSIZ];
@@ -888,6 +950,17 @@ static int perf_top__start_counters(struct perf_top *top)
struct perf_evlist *evlist = top->evlist;
struct record_opts *opts = &top->record_opts;
+ if (perf_top__overwrite_check(top)) {
+ ui__error("perf top only support consistent per-event "
+ "overwrite setting for all events\n");
+ goto out_err;
+ }
+
+ if (opts->overwrite) {
+ ui__error("not support overwrite mode yet\n");
+ goto out_err;
+ }
+
perf_evlist__config(evlist, opts, &callchain_param);
evlist__for_each_entry(evlist, counter) {
Commit-ID: b47160428c794efb32aa6ad6aac323ad79b1e834
Gitweb: https://git.kernel.org/tip/b47160428c794efb32aa6ad6aac323ad79b1e834
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:25 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Tue, 6 Feb 2018 10:11:50 -0300
perf mmap: Discard legacy interface for mmap read
Discards perf_mmap__read_backward() and perf_mmap__read_catchup(). No
tools use them.
There are tools still use perf_mmap__read_forward(). Keep it, but add
comments to point to the new interface for future use.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/util/mmap.c | 50 ++++----------------------------------------------
tools/perf/util/mmap.h | 3 ---
2 files changed, 4 insertions(+), 49 deletions(-)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index f804926..91531a7 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -63,6 +63,10 @@ static union perf_event *perf_mmap__read(struct perf_mmap *map,
return event;
}
+/*
+ * legacy interface for mmap read.
+ * Don't use it. Use perf_mmap__read_event().
+ */
union perf_event *perf_mmap__read_forward(struct perf_mmap *map)
{
u64 head;
@@ -78,41 +82,6 @@ union perf_event *perf_mmap__read_forward(struct perf_mmap *map)
return perf_mmap__read(map, &map->prev, head);
}
-union perf_event *perf_mmap__read_backward(struct perf_mmap *map)
-{
- u64 head, end;
-
- /*
- * Check if event was unmapped due to a POLLHUP/POLLERR.
- */
- if (!refcount_read(&map->refcnt))
- return NULL;
-
- head = perf_mmap__read_head(map);
- if (!head)
- return NULL;
-
- /*
- * 'head' pointer starts from 0. Kernel minus sizeof(record) form
- * it each time when kernel writes to it, so in fact 'head' is
- * negative. 'end' pointer is made manually by adding the size of
- * the ring buffer to 'head' pointer, means the validate data can
- * read is the whole ring buffer. If 'end' is positive, the ring
- * buffer has not fully filled, so we must adjust 'end' to 0.
- *
- * However, since both 'head' and 'end' is unsigned, we can't
- * simply compare 'end' against 0. Here we compare '-head' and
- * the size of the ring buffer, where -head is the number of bytes
- * kernel write to the ring buffer.
- */
- if (-head < (u64)(map->mask + 1))
- end = 0;
- else
- end = head + map->mask + 1;
-
- return perf_mmap__read(map, &map->prev, end);
-}
-
/*
* Read event from ring buffer one by one.
* Return one event for each call.
@@ -152,17 +121,6 @@ union perf_event *perf_mmap__read_event(struct perf_mmap *map,
return event;
}
-void perf_mmap__read_catchup(struct perf_mmap *map)
-{
- u64 head;
-
- if (!refcount_read(&map->refcnt))
- return;
-
- head = perf_mmap__read_head(map);
- map->prev = head;
-}
-
static bool perf_mmap__empty(struct perf_mmap *map)
{
return perf_mmap__read_head(map) == map->prev && !map->auxtrace_mmap.base;
diff --git a/tools/perf/util/mmap.h b/tools/perf/util/mmap.h
index 2871854..ec7d3a24 100644
--- a/tools/perf/util/mmap.h
+++ b/tools/perf/util/mmap.h
@@ -65,8 +65,6 @@ void perf_mmap__put(struct perf_mmap *map);
void perf_mmap__consume(struct perf_mmap *map, bool overwrite);
-void perf_mmap__read_catchup(struct perf_mmap *md);
-
static inline u64 perf_mmap__read_head(struct perf_mmap *mm)
{
struct perf_event_mmap_page *pc = mm->base;
@@ -87,7 +85,6 @@ static inline void perf_mmap__write_tail(struct perf_mmap *md, u64 tail)
}
union perf_event *perf_mmap__read_forward(struct perf_mmap *map);
-union perf_event *perf_mmap__read_backward(struct perf_mmap *map);
union perf_event *perf_mmap__read_event(struct perf_mmap *map,
bool overwrite,
Commit-ID: c14dec366b440fe46f35f99215434f3f9277e19b
Gitweb: https://git.kernel.org/tip/c14dec366b440fe46f35f99215434f3f9277e19b
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:30 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Tue, 6 Feb 2018 10:11:51 -0300
perf top: Remove lost events checking
There would be some records lost in overwrite mode because of pausing
the ringbuffer. It has little impact for the accuracy of the snapshot
and could be tolerated by 'perf top'.
Remove the lost events checking.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/builtin-top.c | 8 +++++---
1 file changed, 5 insertions(+), 3 deletions(-)
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index 7def861..5965306 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -283,8 +283,9 @@ static void perf_top__print_sym_table(struct perf_top *top)
printf("%-*.*s\n", win_width, win_width, graph_dotted_line);
- if (hists->stats.nr_lost_warned !=
- hists->stats.nr_events[PERF_RECORD_LOST]) {
+ if (!top->record_opts.overwrite &&
+ (hists->stats.nr_lost_warned !=
+ hists->stats.nr_events[PERF_RECORD_LOST])) {
hists->stats.nr_lost_warned =
hists->stats.nr_events[PERF_RECORD_LOST];
color_fprintf(stdout, PERF_COLOR_RED,
@@ -611,7 +612,8 @@ static void *display_thread_tui(void *arg)
perf_evlist__tui_browse_hists(top->evlist, help, &hbt,
top->min_percent,
- &top->session->header.env, true);
+ &top->session->header.env,
+ !top->record_opts.overwrite);
done = 1;
return NULL;
Commit-ID: 6a669091c0d93cc1ac4a0de7af722d1359bfe98f
Gitweb: https://git.kernel.org/tip/6a669091c0d93cc1ac4a0de7af722d1359bfe98f
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:28 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Tue, 6 Feb 2018 10:11:51 -0300
perf top: Add overwrite fall back
Switch to non-overwrite mode if kernel doesnot support overwrite
ringbuffer.
It's only effect when overwrite mode is supported. No change to current
behavior.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
[ Use perf_missing_features.write_backward instead of the non merged is_write_backward_fail() ]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/builtin-top.c | 36 ++++++++++++++++++++++++++++++++++++
1 file changed, 36 insertions(+)
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index 1778379..ee4bba1 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -943,6 +943,27 @@ static int perf_top__overwrite_check(struct perf_top *top)
return 0;
}
+static int perf_top_overwrite_fallback(struct perf_top *top,
+ struct perf_evsel *evsel)
+{
+ struct record_opts *opts = &top->record_opts;
+ struct perf_evlist *evlist = top->evlist;
+ struct perf_evsel *counter;
+
+ if (!opts->overwrite)
+ return 0;
+
+ /* only fall back when first event fails */
+ if (evsel != perf_evlist__first(evlist))
+ return 0;
+
+ evlist__for_each_entry(evlist, counter)
+ counter->attr.write_backward = false;
+ opts->overwrite = false;
+ ui__warning("fall back to non-overwrite mode\n");
+ return 1;
+}
+
static int perf_top__start_counters(struct perf_top *top)
{
char msg[BUFSIZ];
@@ -967,6 +988,21 @@ static int perf_top__start_counters(struct perf_top *top)
try_again:
if (perf_evsel__open(counter, top->evlist->cpus,
top->evlist->threads) < 0) {
+
+ /*
+ * Specially handle overwrite fall back.
+ * Because perf top is the only tool which has
+ * overwrite mode by default, support
+ * both overwrite and non-overwrite mode, and
+ * require consistent mode for all events.
+ *
+ * May move it to generic code with more tools
+ * have similar attribute.
+ */
+ if (perf_missing_features.write_backward &&
+ perf_top_overwrite_fallback(top, counter))
+ goto try_again;
+
if (perf_evsel__fallback(counter, errno, msg, sizeof(msg))) {
if (verbose > 0)
ui__warning("%s\n", msg);
Commit-ID: 73508fab8d06316e4dbd6cccfc9fcb21dde677b7
Gitweb: https://git.kernel.org/tip/73508fab8d06316e4dbd6cccfc9fcb21dde677b7
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:31 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Tue, 6 Feb 2018 10:11:51 -0300
perf top: Switch default mode to overwrite mode
perf_top__mmap_read() has a severe performance issue in the Knights
Landing/Mill platform, when monitoring heavy load systems. It costs
several minutes to finish, which is unacceptable.
Currently, 'perf top' uses the non overwrite mode. For non overwrite
mode, it tries to read everything in the ringbuffer and doesn't pause
it. Once there are lots of samples delivered persistently, the
processing time could be very long. Also, the latest samples could be
lost when the ringbuffer is full.
For overwrite mode, it takes a snapshot for the system by pausing the
ringbuffer, which could significantly reduce the processing time. Also,
the overwrite mode always keep the latest samples. Considering the real
time requirement for 'perf top', the overwrite mode is more suitable for
it.
Actually, 'perf top' was overwrite mode. It is changed to non overwrite
mode since commit 93fc64f14472 ("perf top: Switch to non overwrite
mode"). It's better to change it back to overwrite mode by default.
For the kernel which doesn't support overwrite mode, it will fall back
to non overwrite mode.
There would be some records lost in overwrite mode because of pausing
the ringbuffer. It has little impact for the accuracy of the snapshot
and can be tolerated.
For overwrite mode, unconditionally wait 100 ms before each snapshot. It
also reduces the overhead caused by pausing ringbuffer, especially on
light load system.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/builtin-top.c | 34 +++++++++++++++++++++++++---------
1 file changed, 25 insertions(+), 9 deletions(-)
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index 5965306..2b4914f 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -809,15 +809,23 @@ static void perf_event__process_sample(struct perf_tool *tool,
static void perf_top__mmap_read_idx(struct perf_top *top, int idx)
{
+ struct record_opts *opts = &top->record_opts;
+ struct perf_evlist *evlist = top->evlist;
struct perf_sample sample;
struct perf_evsel *evsel;
+ struct perf_mmap *md;
struct perf_session *session = top->session;
union perf_event *event;
struct machine *machine;
+ u64 end, start;
int ret;
- while ((event = perf_evlist__mmap_read(top->evlist, idx)) != NULL) {
- ret = perf_evlist__parse_sample(top->evlist, event, &sample);
+ md = opts->overwrite ? &evlist->overwrite_mmap[idx] : &evlist->mmap[idx];
+ if (perf_mmap__read_init(md, opts->overwrite, &start, &end) < 0)
+ return;
+
+ while ((event = perf_mmap__read_event(md, opts->overwrite, &start, end)) != NULL) {
+ ret = perf_evlist__parse_sample(evlist, event, &sample);
if (ret) {
pr_err("Can't parse sample, err = %d\n", ret);
goto next_event;
@@ -871,16 +879,28 @@ static void perf_top__mmap_read_idx(struct perf_top *top, int idx)
} else
++session->evlist->stats.nr_unknown_events;
next_event:
- perf_evlist__mmap_consume(top->evlist, idx);
+ perf_mmap__consume(md, opts->overwrite);
}
+
+ perf_mmap__read_done(md);
}
static void perf_top__mmap_read(struct perf_top *top)
{
+ bool overwrite = top->record_opts.overwrite;
+ struct perf_evlist *evlist = top->evlist;
int i;
+ if (overwrite)
+ perf_evlist__toggle_bkw_mmap(evlist, BKW_MMAP_DATA_PENDING);
+
for (i = 0; i < top->evlist->nr_mmaps; i++)
perf_top__mmap_read_idx(top, i);
+
+ if (overwrite) {
+ perf_evlist__toggle_bkw_mmap(evlist, BKW_MMAP_EMPTY);
+ perf_evlist__toggle_bkw_mmap(evlist, BKW_MMAP_RUNNING);
+ }
}
/*
@@ -979,11 +999,6 @@ static int perf_top__start_counters(struct perf_top *top)
goto out_err;
}
- if (opts->overwrite) {
- ui__error("not support overwrite mode yet\n");
- goto out_err;
- }
-
perf_evlist__config(evlist, opts, &callchain_param);
evlist__for_each_entry(evlist, counter) {
@@ -1144,7 +1159,7 @@ static int __cmd_top(struct perf_top *top)
perf_top__mmap_read(top);
- if (hits == top->samples)
+ if (opts->overwrite || (hits == top->samples))
ret = perf_evlist__poll(top->evlist, 100);
if (resize) {
@@ -1238,6 +1253,7 @@ int cmd_top(int argc, const char **argv)
.uses_mmap = true,
},
.proc_map_timeout = 500,
+ .overwrite = 1,
},
.max_stack = sysctl_perf_event_max_stack,
.sym_pcnt_filter = 5,
Commit-ID: 1db6c3a2144eaf0d8aaeecf46df73106834bad75
Gitweb: https://git.kernel.org/tip/1db6c3a2144eaf0d8aaeecf46df73106834bad75
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:32 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Tue, 6 Feb 2018 10:11:51 -0300
perf top: Check the latency of perf_top__mmap_read()
The latency of perf_top__mmap_read() should be lower than refresh time.
If not, give some hints to reduce the latency.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/builtin-top.c | 9 +++++++++
1 file changed, 9 insertions(+)
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index 2b4914f..b7c823b 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -889,8 +889,10 @@ static void perf_top__mmap_read(struct perf_top *top)
{
bool overwrite = top->record_opts.overwrite;
struct perf_evlist *evlist = top->evlist;
+ unsigned long long start, end;
int i;
+ start = rdclock();
if (overwrite)
perf_evlist__toggle_bkw_mmap(evlist, BKW_MMAP_DATA_PENDING);
@@ -901,6 +903,13 @@ static void perf_top__mmap_read(struct perf_top *top)
perf_evlist__toggle_bkw_mmap(evlist, BKW_MMAP_EMPTY);
perf_evlist__toggle_bkw_mmap(evlist, BKW_MMAP_RUNNING);
}
+ end = rdclock();
+
+ if ((end - start) > (unsigned long long)top->delay_secs * NSEC_PER_SEC)
+ ui__warning("Too slow to read ring buffer.\n"
+ "Please try increasing the period (-c) or\n"
+ "decreasing the freq (-F) or\n"
+ "limiting the number of CPUs (-C)\n");
}
/*
Commit-ID: 6aac72d3bd37408bc5ae2434c12fc574404291f4
Gitweb: https://git.kernel.org/tip/6aac72d3bd37408bc5ae2434c12fc574404291f4
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:29 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Tue, 6 Feb 2018 10:11:51 -0300
perf hists browser: Add parameter to disable lost event warning
For overwrite mode, the ringbuffer will be paused. The event lost is
expected. It needs a way to notify the browser not print the warning.
It will be used later for perf top to disable lost event warning in
overwrite mode. There is no behavior change for now.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/builtin-c2c.c | 4 ++--
tools/perf/builtin-report.c | 3 ++-
tools/perf/builtin-top.c | 2 +-
tools/perf/ui/browsers/hists.c | 38 +++++++++++++++++++++++++-------------
tools/perf/ui/browsers/hists.h | 3 ++-
tools/perf/util/hist.h | 6 ++++--
6 files changed, 36 insertions(+), 20 deletions(-)
diff --git a/tools/perf/builtin-c2c.c b/tools/perf/builtin-c2c.c
index c0815a3..539c3d4 100644
--- a/tools/perf/builtin-c2c.c
+++ b/tools/perf/builtin-c2c.c
@@ -2245,7 +2245,7 @@ static int perf_c2c__browse_cacheline(struct hist_entry *he)
c2c_browser__update_nr_entries(browser);
while (1) {
- key = hist_browser__run(browser, "? - help");
+ key = hist_browser__run(browser, "? - help", true);
switch (key) {
case 's':
@@ -2314,7 +2314,7 @@ static int perf_c2c__hists_browse(struct hists *hists)
c2c_browser__update_nr_entries(browser);
while (1) {
- key = hist_browser__run(browser, "? - help");
+ key = hist_browser__run(browser, "? - help", true);
switch (key) {
case 'q':
diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
index 42a52dc..4ad5dc6 100644
--- a/tools/perf/builtin-report.c
+++ b/tools/perf/builtin-report.c
@@ -530,7 +530,8 @@ static int report__browse_hists(struct report *rep)
case 1:
ret = perf_evlist__tui_browse_hists(evlist, help, NULL,
rep->min_percent,
- &session->header.env);
+ &session->header.env,
+ true);
/*
* Usually "ret" is the last pressed key, and we only
* care if the key notifies us to switch data file.
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index ee4bba1..7def861 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -611,7 +611,7 @@ static void *display_thread_tui(void *arg)
perf_evlist__tui_browse_hists(top->evlist, help, &hbt,
top->min_percent,
- &top->session->header.env);
+ &top->session->header.env, true);
done = 1;
return NULL;
diff --git a/tools/perf/ui/browsers/hists.c b/tools/perf/ui/browsers/hists.c
index 68146f4..6495ee5 100644
--- a/tools/perf/ui/browsers/hists.c
+++ b/tools/perf/ui/browsers/hists.c
@@ -608,7 +608,8 @@ static int hist_browser__title(struct hist_browser *browser, char *bf, size_t si
return browser->title ? browser->title(browser, bf, size) : 0;
}
-int hist_browser__run(struct hist_browser *browser, const char *help)
+int hist_browser__run(struct hist_browser *browser, const char *help,
+ bool warn_lost_event)
{
int key;
char title[160];
@@ -638,8 +639,9 @@ int hist_browser__run(struct hist_browser *browser, const char *help)
nr_entries = hist_browser__nr_entries(browser);
ui_browser__update_nr_entries(&browser->b, nr_entries);
- if (browser->hists->stats.nr_lost_warned !=
- browser->hists->stats.nr_events[PERF_RECORD_LOST]) {
+ if (warn_lost_event &&
+ (browser->hists->stats.nr_lost_warned !=
+ browser->hists->stats.nr_events[PERF_RECORD_LOST])) {
browser->hists->stats.nr_lost_warned =
browser->hists->stats.nr_events[PERF_RECORD_LOST];
ui_browser__warn_lost_events(&browser->b);
@@ -2763,7 +2765,8 @@ static int perf_evsel__hists_browse(struct perf_evsel *evsel, int nr_events,
bool left_exits,
struct hist_browser_timer *hbt,
float min_pcnt,
- struct perf_env *env)
+ struct perf_env *env,
+ bool warn_lost_event)
{
struct hists *hists = evsel__hists(evsel);
struct hist_browser *browser = perf_evsel_browser__new(evsel, hbt, env);
@@ -2844,7 +2847,8 @@ static int perf_evsel__hists_browse(struct perf_evsel *evsel, int nr_events,
nr_options = 0;
- key = hist_browser__run(browser, helpline);
+ key = hist_browser__run(browser, helpline,
+ warn_lost_event);
if (browser->he_selection != NULL) {
thread = hist_browser__selected_thread(browser);
@@ -3184,7 +3188,8 @@ static void perf_evsel_menu__write(struct ui_browser *browser,
static int perf_evsel_menu__run(struct perf_evsel_menu *menu,
int nr_events, const char *help,
- struct hist_browser_timer *hbt)
+ struct hist_browser_timer *hbt,
+ bool warn_lost_event)
{
struct perf_evlist *evlist = menu->b.priv;
struct perf_evsel *pos;
@@ -3203,7 +3208,9 @@ static int perf_evsel_menu__run(struct perf_evsel_menu *menu,
case K_TIMER:
hbt->timer(hbt->arg);
- if (!menu->lost_events_warned && menu->lost_events) {
+ if (!menu->lost_events_warned &&
+ menu->lost_events &&
+ warn_lost_event) {
ui_browser__warn_lost_events(&menu->b);
menu->lost_events_warned = true;
}
@@ -3224,7 +3231,8 @@ browse_hists:
key = perf_evsel__hists_browse(pos, nr_events, help,
true, hbt,
menu->min_pcnt,
- menu->env);
+ menu->env,
+ warn_lost_event);
ui_browser__show_title(&menu->b, title);
switch (key) {
case K_TAB:
@@ -3282,7 +3290,8 @@ static int __perf_evlist__tui_browse_hists(struct perf_evlist *evlist,
int nr_entries, const char *help,
struct hist_browser_timer *hbt,
float min_pcnt,
- struct perf_env *env)
+ struct perf_env *env,
+ bool warn_lost_event)
{
struct perf_evsel *pos;
struct perf_evsel_menu menu = {
@@ -3309,13 +3318,15 @@ static int __perf_evlist__tui_browse_hists(struct perf_evlist *evlist,
menu.b.width = line_len;
}
- return perf_evsel_menu__run(&menu, nr_entries, help, hbt);
+ return perf_evsel_menu__run(&menu, nr_entries, help,
+ hbt, warn_lost_event);
}
int perf_evlist__tui_browse_hists(struct perf_evlist *evlist, const char *help,
struct hist_browser_timer *hbt,
float min_pcnt,
- struct perf_env *env)
+ struct perf_env *env,
+ bool warn_lost_event)
{
int nr_entries = evlist->nr_entries;
@@ -3325,7 +3336,7 @@ single_entry:
return perf_evsel__hists_browse(first, nr_entries, help,
false, hbt, min_pcnt,
- env);
+ env, warn_lost_event);
}
if (symbol_conf.event_group) {
@@ -3342,5 +3353,6 @@ single_entry:
}
return __perf_evlist__tui_browse_hists(evlist, nr_entries, help,
- hbt, min_pcnt, env);
+ hbt, min_pcnt, env,
+ warn_lost_event);
}
diff --git a/tools/perf/ui/browsers/hists.h b/tools/perf/ui/browsers/hists.h
index ba43177..9428bee 100644
--- a/tools/perf/ui/browsers/hists.h
+++ b/tools/perf/ui/browsers/hists.h
@@ -28,7 +28,8 @@ struct hist_browser {
struct hist_browser *hist_browser__new(struct hists *hists);
void hist_browser__delete(struct hist_browser *browser);
-int hist_browser__run(struct hist_browser *browser, const char *help);
+int hist_browser__run(struct hist_browser *browser, const char *help,
+ bool warn_lost_event);
void hist_browser__init(struct hist_browser *browser,
struct hists *hists);
#endif /* _PERF_UI_BROWSER_HISTS_H_ */
diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h
index f6630cb..02721b57 100644
--- a/tools/perf/util/hist.h
+++ b/tools/perf/util/hist.h
@@ -430,7 +430,8 @@ int hist_entry__tui_annotate(struct hist_entry *he, struct perf_evsel *evsel,
int perf_evlist__tui_browse_hists(struct perf_evlist *evlist, const char *help,
struct hist_browser_timer *hbt,
float min_pcnt,
- struct perf_env *env);
+ struct perf_env *env,
+ bool warn_lost_event);
int script_browse(const char *script_opt);
#else
static inline
@@ -438,7 +439,8 @@ int perf_evlist__tui_browse_hists(struct perf_evlist *evlist __maybe_unused,
const char *help __maybe_unused,
struct hist_browser_timer *hbt __maybe_unused,
float min_pcnt __maybe_unused,
- struct perf_env *env __maybe_unused)
+ struct perf_env *env __maybe_unused,
+ bool warn_lost_event __maybe_unused)
{
return 0;
}
Commit-ID: 600a7cfe88de2c6e44e23d61dd721b996b790eb2
Gitweb: https://git.kernel.org/tip/600a7cfe88de2c6e44e23d61dd721b996b790eb2
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:24 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 15 Feb 2018 09:54:08 -0300
perf test: Update mmap read functions for backward-ring-buffer test
Use the new perf_mmap__read_* interfaces for overwrite ringbuffer test.
Commiter notes:
Testing:
[root@seventh ~]# perf test -v backward
48: Read backward ring buffer :
--- start ---
test child forked, pid 8309
Using CPUID GenuineIntel-6-9E
mmap size 1052672B
mmap size 8192B
Finished reading overwrite ring buffer: rewind
test child finished with 0
---- end ----
Read backward ring buffer: Ok
[root@seventh ~]#
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/tests/backward-ring-buffer.c | 7 +++++--
1 file changed, 5 insertions(+), 2 deletions(-)
diff --git a/tools/perf/tests/backward-ring-buffer.c b/tools/perf/tests/backward-ring-buffer.c
index 4035d43..e0b1b41 100644
--- a/tools/perf/tests/backward-ring-buffer.c
+++ b/tools/perf/tests/backward-ring-buffer.c
@@ -31,10 +31,12 @@ static int count_samples(struct perf_evlist *evlist, int *sample_count,
int i;
for (i = 0; i < evlist->nr_mmaps; i++) {
+ struct perf_mmap *map = &evlist->overwrite_mmap[i];
union perf_event *event;
+ u64 start, end;
- perf_mmap__read_catchup(&evlist->overwrite_mmap[i]);
- while ((event = perf_mmap__read_backward(&evlist->overwrite_mmap[i])) != NULL) {
+ perf_mmap__read_init(map, true, &start, &end);
+ while ((event = perf_mmap__read_event(map, true, &start, end)) != NULL) {
const u32 type = event->header.type;
switch (type) {
@@ -49,6 +51,7 @@ static int count_samples(struct perf_evlist *evlist, int *sample_count,
return TEST_FAIL;
}
}
+ perf_mmap__read_done(map);
}
return TEST_OK;
}
Commit-ID: 3effc2f165a842d640873e29d4c5cc1650143aef
Gitweb: https://git.kernel.org/tip/3effc2f165a842d640873e29d4c5cc1650143aef
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:25 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 15 Feb 2018 09:54:17 -0300
perf mmap: Discard legacy interface for mmap read
Discards perf_mmap__read_backward() and perf_mmap__read_catchup(). No
tools use them.
There are tools still use perf_mmap__read_forward(). Keep it, but add
comments to point to the new interface for future use.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/util/mmap.c | 50 ++++----------------------------------------------
tools/perf/util/mmap.h | 3 ---
2 files changed, 4 insertions(+), 49 deletions(-)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index f804926..91531a7 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -63,6 +63,10 @@ static union perf_event *perf_mmap__read(struct perf_mmap *map,
return event;
}
+/*
+ * legacy interface for mmap read.
+ * Don't use it. Use perf_mmap__read_event().
+ */
union perf_event *perf_mmap__read_forward(struct perf_mmap *map)
{
u64 head;
@@ -78,41 +82,6 @@ union perf_event *perf_mmap__read_forward(struct perf_mmap *map)
return perf_mmap__read(map, &map->prev, head);
}
-union perf_event *perf_mmap__read_backward(struct perf_mmap *map)
-{
- u64 head, end;
-
- /*
- * Check if event was unmapped due to a POLLHUP/POLLERR.
- */
- if (!refcount_read(&map->refcnt))
- return NULL;
-
- head = perf_mmap__read_head(map);
- if (!head)
- return NULL;
-
- /*
- * 'head' pointer starts from 0. Kernel minus sizeof(record) form
- * it each time when kernel writes to it, so in fact 'head' is
- * negative. 'end' pointer is made manually by adding the size of
- * the ring buffer to 'head' pointer, means the validate data can
- * read is the whole ring buffer. If 'end' is positive, the ring
- * buffer has not fully filled, so we must adjust 'end' to 0.
- *
- * However, since both 'head' and 'end' is unsigned, we can't
- * simply compare 'end' against 0. Here we compare '-head' and
- * the size of the ring buffer, where -head is the number of bytes
- * kernel write to the ring buffer.
- */
- if (-head < (u64)(map->mask + 1))
- end = 0;
- else
- end = head + map->mask + 1;
-
- return perf_mmap__read(map, &map->prev, end);
-}
-
/*
* Read event from ring buffer one by one.
* Return one event for each call.
@@ -152,17 +121,6 @@ union perf_event *perf_mmap__read_event(struct perf_mmap *map,
return event;
}
-void perf_mmap__read_catchup(struct perf_mmap *map)
-{
- u64 head;
-
- if (!refcount_read(&map->refcnt))
- return;
-
- head = perf_mmap__read_head(map);
- map->prev = head;
-}
-
static bool perf_mmap__empty(struct perf_mmap *map)
{
return perf_mmap__read_head(map) == map->prev && !map->auxtrace_mmap.base;
diff --git a/tools/perf/util/mmap.h b/tools/perf/util/mmap.h
index 2871854..ec7d3a24 100644
--- a/tools/perf/util/mmap.h
+++ b/tools/perf/util/mmap.h
@@ -65,8 +65,6 @@ void perf_mmap__put(struct perf_mmap *map);
void perf_mmap__consume(struct perf_mmap *map, bool overwrite);
-void perf_mmap__read_catchup(struct perf_mmap *md);
-
static inline u64 perf_mmap__read_head(struct perf_mmap *mm)
{
struct perf_event_mmap_page *pc = mm->base;
@@ -87,7 +85,6 @@ static inline void perf_mmap__write_tail(struct perf_mmap *md, u64 tail)
}
union perf_event *perf_mmap__read_forward(struct perf_mmap *map);
-union perf_event *perf_mmap__read_backward(struct perf_mmap *map);
union perf_event *perf_mmap__read_event(struct perf_mmap *map,
bool overwrite,
Commit-ID: 6888ff66c44ffa3077ed69e978902d0ff4b84ae1
Gitweb: https://git.kernel.org/tip/6888ff66c44ffa3077ed69e978902d0ff4b84ae1
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:16 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 15 Feb 2018 09:50:53 -0300
perf evlist: Remove stale mmap read for backward
perf_evlist__mmap_read_catchup() and perf_evlist__mmap_read_backward()
are only for overwrite mode.
But they read the evlist->mmap buffer which is for non-overwrite mode.
It did not bring any serious problem yet, because there is no one use
it.
Remove the unused interfaces.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Acked-by: Wang Nan <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/util/evlist.c | 17 -----------------
tools/perf/util/evlist.h | 4 ----
2 files changed, 21 deletions(-)
diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index ac35cd2..e5fc14e 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -715,28 +715,11 @@ union perf_event *perf_evlist__mmap_read_forward(struct perf_evlist *evlist, int
return perf_mmap__read_forward(md);
}
-union perf_event *perf_evlist__mmap_read_backward(struct perf_evlist *evlist, int idx)
-{
- struct perf_mmap *md = &evlist->mmap[idx];
-
- /*
- * No need to check messup for backward ring buffer:
- * We can always read arbitrary long data from a backward
- * ring buffer unless we forget to pause it before reading.
- */
- return perf_mmap__read_backward(md);
-}
-
union perf_event *perf_evlist__mmap_read(struct perf_evlist *evlist, int idx)
{
return perf_evlist__mmap_read_forward(evlist, idx);
}
-void perf_evlist__mmap_read_catchup(struct perf_evlist *evlist, int idx)
-{
- perf_mmap__read_catchup(&evlist->mmap[idx]);
-}
-
void perf_evlist__mmap_consume(struct perf_evlist *evlist, int idx)
{
perf_mmap__consume(&evlist->mmap[idx], false);
diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
index 75f8e0a..336b838 100644
--- a/tools/perf/util/evlist.h
+++ b/tools/perf/util/evlist.h
@@ -133,10 +133,6 @@ union perf_event *perf_evlist__mmap_read(struct perf_evlist *evlist, int idx);
union perf_event *perf_evlist__mmap_read_forward(struct perf_evlist *evlist,
int idx);
-union perf_event *perf_evlist__mmap_read_backward(struct perf_evlist *evlist,
- int idx);
-void perf_evlist__mmap_read_catchup(struct perf_evlist *evlist, int idx);
-
void perf_evlist__mmap_consume(struct perf_evlist *evlist, int idx);
int perf_evlist__open(struct perf_evlist *evlist);
Commit-ID: dc6c35c679e96987dc83a003f30bc2cc33c84c00
Gitweb: https://git.kernel.org/tip/dc6c35c679e96987dc83a003f30bc2cc33c84c00
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:17 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 15 Feb 2018 09:51:57 -0300
perf mmap: Recalculate size for overwrite mode
In perf_mmap__push(), the 'size' need to be recalculated, otherwise the
invalid data might be pushed to the record in overwrite mode.
The issue is introduced by commit 7fb4b407a124 ("perf mmap: Don't
discard prev in backward mode").
When the ring buffer is full in overwrite mode, backward_rb_find_range()
will be called to recalculate the 'start' and 'end'. The 'size' needs to
be recalculated accordingly.
Unconditionally recalculate the 'size', not just for full ring buffer in
overwrite mode. Because:
- There is no harmful to recalculate the 'size' for other cases.
- The code of calculating 'start' and 'end' will be factored out later.
The new function does not need to return 'size'.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Fixes: 7fb4b407a124 ("perf mmap: Don't discard prev in backward mode")
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/util/mmap.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index 05076e6..97cf4fa 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -302,6 +302,8 @@ int perf_mmap__push(struct perf_mmap *md, bool overwrite,
return -1;
}
+ size = end - start;
+
if ((start & md->mask) + size != (end & md->mask)) {
buf = &data[start & md->mask];
size = md->mask + 1 - (start & md->mask);
Commit-ID: 63878a53cedc3df31bd4ba8740a49fa0fc116ac6
Gitweb: https://git.kernel.org/tip/63878a53cedc3df31bd4ba8740a49fa0fc116ac6
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:26 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 15 Feb 2018 09:54:42 -0300
perf top: Check per-event overwrite term
Per-event overwrite term is not forbidden in 'perf top', which can bring
problems. Because 'perf top' only support non-overwrite mode now.
Add new rules and check regarding to overwrite term for 'perf top'.
- All events either have same per-event term or don't have per-event
mode setting. Otherwise, it will error out.
- Per-event overwrite term should be consistent as opts->overwrite.
If not, updating the opts->overwrite according to per-event term.
Make it possible to support either non-overwrite or overwrite mode.
The overwrite mode is forbidden now, which will be removed when the
overwrite mode is supported later.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
[ Renamed perf_top_overwrite_check to perf_top__overwrite_check, to follow existing convention ]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/builtin-top.c | 73 ++++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 73 insertions(+)
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index c6ccda5..1778379 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -881,6 +881,68 @@ static void perf_top__mmap_read(struct perf_top *top)
perf_top__mmap_read_idx(top, i);
}
+/*
+ * Check per-event overwrite term.
+ * perf top should support consistent term for all events.
+ * - All events don't have per-event term
+ * E.g. "cpu/cpu-cycles/,cpu/instructions/"
+ * Nothing change, return 0.
+ * - All events have same per-event term
+ * E.g. "cpu/cpu-cycles,no-overwrite/,cpu/instructions,no-overwrite/
+ * Using the per-event setting to replace the opts->overwrite if
+ * they are different, then return 0.
+ * - Events have different per-event term
+ * E.g. "cpu/cpu-cycles,overwrite/,cpu/instructions,no-overwrite/"
+ * Return -1
+ * - Some of the event set per-event term, but some not.
+ * E.g. "cpu/cpu-cycles/,cpu/instructions,no-overwrite/"
+ * Return -1
+ */
+static int perf_top__overwrite_check(struct perf_top *top)
+{
+ struct record_opts *opts = &top->record_opts;
+ struct perf_evlist *evlist = top->evlist;
+ struct perf_evsel_config_term *term;
+ struct list_head *config_terms;
+ struct perf_evsel *evsel;
+ int set, overwrite = -1;
+
+ evlist__for_each_entry(evlist, evsel) {
+ set = -1;
+ config_terms = &evsel->config_terms;
+ list_for_each_entry(term, config_terms, list) {
+ if (term->type == PERF_EVSEL__CONFIG_TERM_OVERWRITE)
+ set = term->val.overwrite ? 1 : 0;
+ }
+
+ /* no term for current and previous event (likely) */
+ if ((overwrite < 0) && (set < 0))
+ continue;
+
+ /* has term for both current and previous event, compare */
+ if ((overwrite >= 0) && (set >= 0) && (overwrite != set))
+ return -1;
+
+ /* no term for current event but has term for previous one */
+ if ((overwrite >= 0) && (set < 0))
+ return -1;
+
+ /* has term for current event */
+ if ((overwrite < 0) && (set >= 0)) {
+ /* if it's first event, set overwrite */
+ if (evsel == perf_evlist__first(evlist))
+ overwrite = set;
+ else
+ return -1;
+ }
+ }
+
+ if ((overwrite >= 0) && (opts->overwrite != overwrite))
+ opts->overwrite = overwrite;
+
+ return 0;
+}
+
static int perf_top__start_counters(struct perf_top *top)
{
char msg[BUFSIZ];
@@ -888,6 +950,17 @@ static int perf_top__start_counters(struct perf_top *top)
struct perf_evlist *evlist = top->evlist;
struct record_opts *opts = &top->record_opts;
+ if (perf_top__overwrite_check(top)) {
+ ui__error("perf top only support consistent per-event "
+ "overwrite setting for all events\n");
+ goto out_err;
+ }
+
+ if (opts->overwrite) {
+ ui__error("not support overwrite mode yet\n");
+ goto out_err;
+ }
+
perf_evlist__config(evlist, opts, &callchain_param);
evlist__for_each_entry(evlist, counter) {
Commit-ID: 189f2cc91f9f2efef5d5f4dde43684c01b5f6f2f
Gitweb: https://git.kernel.org/tip/189f2cc91f9f2efef5d5f4dde43684c01b5f6f2f
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:20 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 15 Feb 2018 09:52:49 -0300
perf mmap: Add new return value logic for perf_mmap__read_init()
Improve the readability by using meaningful enum (-EAGAIN, -EINVAL and
0) to replace the three returning states (0, -1 and 1).
Suggested-by: Wang Nan <[email protected]>
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/util/mmap.c | 12 ++++++------
1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index c19a4e6..38fa69d 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -282,7 +282,7 @@ int perf_mmap__read_init(struct perf_mmap *md, bool overwrite,
*endp = overwrite ? old : head;
if (*startp == *endp)
- return 0;
+ return -EAGAIN;
size = *endp - *startp;
if (size > (unsigned long)(md->mask) + 1) {
@@ -291,7 +291,7 @@ int perf_mmap__read_init(struct perf_mmap *md, bool overwrite,
md->prev = head;
perf_mmap__consume(md, overwrite);
- return 0;
+ return -EAGAIN;
}
/*
@@ -299,10 +299,10 @@ int perf_mmap__read_init(struct perf_mmap *md, bool overwrite,
* most of data from it.
*/
if (overwrite_rb_find_range(data, md->mask, head, startp, endp))
- return -1;
+ return -EINVAL;
}
- return 1;
+ return 0;
}
int perf_mmap__push(struct perf_mmap *md, bool overwrite,
@@ -316,8 +316,8 @@ int perf_mmap__push(struct perf_mmap *md, bool overwrite,
int rc = 0;
rc = perf_mmap__read_init(md, overwrite, &start, &end);
- if (rc < 1)
- return rc;
+ if (rc < 0)
+ return (rc == -EAGAIN) ? 0 : -1;
size = end - start;
Commit-ID: f92c8cbe597a5a2ccec702dff824f3fe0f3623eb
Gitweb: https://git.kernel.org/tip/f92c8cbe597a5a2ccec702dff824f3fe0f3623eb
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:18 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 15 Feb 2018 09:52:05 -0300
perf mmap: Cleanup perf_mmap__push()
The first assignment for 'start' and 'end' is redundant.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/util/mmap.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index 97cf4fa..fbbbe87 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -272,7 +272,7 @@ int perf_mmap__push(struct perf_mmap *md, bool overwrite,
{
u64 head = perf_mmap__read_head(md);
u64 old = md->prev;
- u64 end = head, start = old;
+ u64 end, start;
unsigned char *data = md->base + page_size;
unsigned long size;
void *buf;
Commit-ID: ebebbf082357f86cc84a4d46ce897a5750e41b7a
Gitweb: https://git.kernel.org/tip/ebebbf082357f86cc84a4d46ce897a5750e41b7a
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:31 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 15 Feb 2018 09:56:54 -0300
perf top: Switch default mode to overwrite mode
perf_top__mmap_read() has a severe performance issue in the Knights
Landing/Mill platform, when monitoring heavy load systems. It costs
several minutes to finish, which is unacceptable.
Currently, 'perf top' uses the non overwrite mode. For non overwrite
mode, it tries to read everything in the ringbuffer and doesn't pause
it. Once there are lots of samples delivered persistently, the
processing time could be very long. Also, the latest samples could be
lost when the ringbuffer is full.
For overwrite mode, it takes a snapshot for the system by pausing the
ringbuffer, which could significantly reduce the processing time. Also,
the overwrite mode always keep the latest samples. Considering the real
time requirement for 'perf top', the overwrite mode is more suitable for
it.
Actually, 'perf top' was overwrite mode. It is changed to non overwrite
mode since commit 93fc64f14472 ("perf top: Switch to non overwrite
mode"). It's better to change it back to overwrite mode by default.
For the kernel which doesn't support overwrite mode, it will fall back
to non overwrite mode.
There would be some records lost in overwrite mode because of pausing
the ringbuffer. It has little impact for the accuracy of the snapshot
and can be tolerated.
For overwrite mode, unconditionally wait 100 ms before each snapshot. It
also reduces the overhead caused by pausing ringbuffer, especially on
light load system.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/builtin-top.c | 34 +++++++++++++++++++++++++---------
1 file changed, 25 insertions(+), 9 deletions(-)
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index 5965306..2b4914f 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -809,15 +809,23 @@ static void perf_event__process_sample(struct perf_tool *tool,
static void perf_top__mmap_read_idx(struct perf_top *top, int idx)
{
+ struct record_opts *opts = &top->record_opts;
+ struct perf_evlist *evlist = top->evlist;
struct perf_sample sample;
struct perf_evsel *evsel;
+ struct perf_mmap *md;
struct perf_session *session = top->session;
union perf_event *event;
struct machine *machine;
+ u64 end, start;
int ret;
- while ((event = perf_evlist__mmap_read(top->evlist, idx)) != NULL) {
- ret = perf_evlist__parse_sample(top->evlist, event, &sample);
+ md = opts->overwrite ? &evlist->overwrite_mmap[idx] : &evlist->mmap[idx];
+ if (perf_mmap__read_init(md, opts->overwrite, &start, &end) < 0)
+ return;
+
+ while ((event = perf_mmap__read_event(md, opts->overwrite, &start, end)) != NULL) {
+ ret = perf_evlist__parse_sample(evlist, event, &sample);
if (ret) {
pr_err("Can't parse sample, err = %d\n", ret);
goto next_event;
@@ -871,16 +879,28 @@ static void perf_top__mmap_read_idx(struct perf_top *top, int idx)
} else
++session->evlist->stats.nr_unknown_events;
next_event:
- perf_evlist__mmap_consume(top->evlist, idx);
+ perf_mmap__consume(md, opts->overwrite);
}
+
+ perf_mmap__read_done(md);
}
static void perf_top__mmap_read(struct perf_top *top)
{
+ bool overwrite = top->record_opts.overwrite;
+ struct perf_evlist *evlist = top->evlist;
int i;
+ if (overwrite)
+ perf_evlist__toggle_bkw_mmap(evlist, BKW_MMAP_DATA_PENDING);
+
for (i = 0; i < top->evlist->nr_mmaps; i++)
perf_top__mmap_read_idx(top, i);
+
+ if (overwrite) {
+ perf_evlist__toggle_bkw_mmap(evlist, BKW_MMAP_EMPTY);
+ perf_evlist__toggle_bkw_mmap(evlist, BKW_MMAP_RUNNING);
+ }
}
/*
@@ -979,11 +999,6 @@ static int perf_top__start_counters(struct perf_top *top)
goto out_err;
}
- if (opts->overwrite) {
- ui__error("not support overwrite mode yet\n");
- goto out_err;
- }
-
perf_evlist__config(evlist, opts, &callchain_param);
evlist__for_each_entry(evlist, counter) {
@@ -1144,7 +1159,7 @@ static int __cmd_top(struct perf_top *top)
perf_top__mmap_read(top);
- if (hits == top->samples)
+ if (opts->overwrite || (hits == top->samples))
ret = perf_evlist__poll(top->evlist, 100);
if (resize) {
@@ -1238,6 +1253,7 @@ int cmd_top(int argc, const char **argv)
.uses_mmap = true,
},
.proc_map_timeout = 500,
+ .overwrite = 1,
},
.max_stack = sysctl_perf_event_max_stack,
.sym_pcnt_filter = 5,
Commit-ID: 204721d7eabe6ee98aafce791ce3efdbc4715834
Gitweb: https://git.kernel.org/tip/204721d7eabe6ee98aafce791ce3efdbc4715834
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:28 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 15 Feb 2018 09:56:14 -0300
perf top: Add overwrite fall back
Switch to non-overwrite mode if kernel doesnot support overwrite
ringbuffer.
It's only effect when overwrite mode is supported. No change to current
behavior.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
[ Use perf_missing_features.write_backward instead of the non merged is_write_backward_fail() ]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/builtin-top.c | 36 ++++++++++++++++++++++++++++++++++++
1 file changed, 36 insertions(+)
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index 1778379..ee4bba1 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -943,6 +943,27 @@ static int perf_top__overwrite_check(struct perf_top *top)
return 0;
}
+static int perf_top_overwrite_fallback(struct perf_top *top,
+ struct perf_evsel *evsel)
+{
+ struct record_opts *opts = &top->record_opts;
+ struct perf_evlist *evlist = top->evlist;
+ struct perf_evsel *counter;
+
+ if (!opts->overwrite)
+ return 0;
+
+ /* only fall back when first event fails */
+ if (evsel != perf_evlist__first(evlist))
+ return 0;
+
+ evlist__for_each_entry(evlist, counter)
+ counter->attr.write_backward = false;
+ opts->overwrite = false;
+ ui__warning("fall back to non-overwrite mode\n");
+ return 1;
+}
+
static int perf_top__start_counters(struct perf_top *top)
{
char msg[BUFSIZ];
@@ -967,6 +988,21 @@ static int perf_top__start_counters(struct perf_top *top)
try_again:
if (perf_evsel__open(counter, top->evlist->cpus,
top->evlist->threads) < 0) {
+
+ /*
+ * Specially handle overwrite fall back.
+ * Because perf top is the only tool which has
+ * overwrite mode by default, support
+ * both overwrite and non-overwrite mode, and
+ * require consistent mode for all events.
+ *
+ * May move it to generic code with more tools
+ * have similar attribute.
+ */
+ if (perf_missing_features.write_backward &&
+ perf_top_overwrite_fallback(top, counter))
+ goto try_again;
+
if (perf_evsel__fallback(counter, errno, msg, sizeof(msg))) {
if (verbose > 0)
ui__warning("%s\n", msg);
Commit-ID: 8cc42de736b617827a4e7664fb8d7a325bc125bc
Gitweb: https://git.kernel.org/tip/8cc42de736b617827a4e7664fb8d7a325bc125bc
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:32 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 15 Feb 2018 09:57:06 -0300
perf top: Check the latency of perf_top__mmap_read()
The latency of perf_top__mmap_read() should be lower than refresh time.
If not, give some hints to reduce the latency.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/builtin-top.c | 9 +++++++++
1 file changed, 9 insertions(+)
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index 2b4914f..b7c823b 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -889,8 +889,10 @@ static void perf_top__mmap_read(struct perf_top *top)
{
bool overwrite = top->record_opts.overwrite;
struct perf_evlist *evlist = top->evlist;
+ unsigned long long start, end;
int i;
+ start = rdclock();
if (overwrite)
perf_evlist__toggle_bkw_mmap(evlist, BKW_MMAP_DATA_PENDING);
@@ -901,6 +903,13 @@ static void perf_top__mmap_read(struct perf_top *top)
perf_evlist__toggle_bkw_mmap(evlist, BKW_MMAP_EMPTY);
perf_evlist__toggle_bkw_mmap(evlist, BKW_MMAP_RUNNING);
}
+ end = rdclock();
+
+ if ((end - start) > (unsigned long long)top->delay_secs * NSEC_PER_SEC)
+ ui__warning("Too slow to read ring buffer.\n"
+ "Please try increasing the period (-c) or\n"
+ "decreasing the freq (-F) or\n"
+ "limiting the number of CPUs (-C)\n");
}
/*
Commit-ID: 7bb45972952db9298fe5cc440160dcad1a66bfbc
Gitweb: https://git.kernel.org/tip/7bb45972952db9298fe5cc440160dcad1a66bfbc
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:23 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 15 Feb 2018 09:53:40 -0300
perf mmap: Introduce perf_mmap__read_event()
Except for 'perf record', the other perf tools read events one by one
from the ring buffer using perf_mmap__read_forward(). But it only
supports non-overwrite mode.
Introduce perf_mmap__read_event() to support both non-overwrite and
overwrite mode.
Usage:
perf_mmap__read_init()
while(event = perf_mmap__read_event()) {
//process the event
perf_mmap__consume()
}
perf_mmap__read_done()
It cannot use perf_mmap__read_backward(). Because it always reads the
stale buffer which is already processed. Furthermore, the forward and
backward concepts have been removed. The perf_mmap__read_backward() will
be replaced and discarded later.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/util/mmap.c | 39 +++++++++++++++++++++++++++++++++++++++
tools/perf/util/mmap.h | 4 ++++
2 files changed, 43 insertions(+)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index 4f59eae..f804926 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -113,6 +113,45 @@ union perf_event *perf_mmap__read_backward(struct perf_mmap *map)
return perf_mmap__read(map, &map->prev, end);
}
+/*
+ * Read event from ring buffer one by one.
+ * Return one event for each call.
+ *
+ * Usage:
+ * perf_mmap__read_init()
+ * while(event = perf_mmap__read_event()) {
+ * //process the event
+ * perf_mmap__consume()
+ * }
+ * perf_mmap__read_done()
+ */
+union perf_event *perf_mmap__read_event(struct perf_mmap *map,
+ bool overwrite,
+ u64 *startp, u64 end)
+{
+ union perf_event *event;
+
+ /*
+ * Check if event was unmapped due to a POLLHUP/POLLERR.
+ */
+ if (!refcount_read(&map->refcnt))
+ return NULL;
+
+ if (startp == NULL)
+ return NULL;
+
+ /* non-overwirte doesn't pause the ringbuffer */
+ if (!overwrite)
+ end = perf_mmap__read_head(map);
+
+ event = perf_mmap__read(map, startp, end);
+
+ if (!overwrite)
+ map->prev = *startp;
+
+ return event;
+}
+
void perf_mmap__read_catchup(struct perf_mmap *map)
{
u64 head;
diff --git a/tools/perf/util/mmap.h b/tools/perf/util/mmap.h
index 95549d4..2871854 100644
--- a/tools/perf/util/mmap.h
+++ b/tools/perf/util/mmap.h
@@ -89,6 +89,10 @@ static inline void perf_mmap__write_tail(struct perf_mmap *md, u64 tail)
union perf_event *perf_mmap__read_forward(struct perf_mmap *map);
union perf_event *perf_mmap__read_backward(struct perf_mmap *map);
+union perf_event *perf_mmap__read_event(struct perf_mmap *map,
+ bool overwrite,
+ u64 *startp, u64 end);
+
int perf_mmap__push(struct perf_mmap *md, bool backward,
void *to, int push(void *to, void *buf, size_t size));
Commit-ID: 06cc1a470ab237b991901729b125404c164f3660
Gitweb: https://git.kernel.org/tip/06cc1a470ab237b991901729b125404c164f3660
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:29 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 15 Feb 2018 09:56:26 -0300
perf hists browser: Add parameter to disable lost event warning
For overwrite mode, the ringbuffer will be paused. The event lost is
expected. It needs a way to notify the browser not print the warning.
It will be used later for perf top to disable lost event warning in
overwrite mode. There is no behavior change for now.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/builtin-c2c.c | 4 ++--
tools/perf/builtin-report.c | 3 ++-
tools/perf/builtin-top.c | 2 +-
tools/perf/ui/browsers/hists.c | 38 +++++++++++++++++++++++++-------------
tools/perf/ui/browsers/hists.h | 3 ++-
tools/perf/util/hist.h | 6 ++++--
6 files changed, 36 insertions(+), 20 deletions(-)
diff --git a/tools/perf/builtin-c2c.c b/tools/perf/builtin-c2c.c
index c0815a3..539c3d4 100644
--- a/tools/perf/builtin-c2c.c
+++ b/tools/perf/builtin-c2c.c
@@ -2245,7 +2245,7 @@ static int perf_c2c__browse_cacheline(struct hist_entry *he)
c2c_browser__update_nr_entries(browser);
while (1) {
- key = hist_browser__run(browser, "? - help");
+ key = hist_browser__run(browser, "? - help", true);
switch (key) {
case 's':
@@ -2314,7 +2314,7 @@ static int perf_c2c__hists_browse(struct hists *hists)
c2c_browser__update_nr_entries(browser);
while (1) {
- key = hist_browser__run(browser, "? - help");
+ key = hist_browser__run(browser, "? - help", true);
switch (key) {
case 'q':
diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
index 42a52dc..4ad5dc6 100644
--- a/tools/perf/builtin-report.c
+++ b/tools/perf/builtin-report.c
@@ -530,7 +530,8 @@ static int report__browse_hists(struct report *rep)
case 1:
ret = perf_evlist__tui_browse_hists(evlist, help, NULL,
rep->min_percent,
- &session->header.env);
+ &session->header.env,
+ true);
/*
* Usually "ret" is the last pressed key, and we only
* care if the key notifies us to switch data file.
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index ee4bba1..7def861 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -611,7 +611,7 @@ static void *display_thread_tui(void *arg)
perf_evlist__tui_browse_hists(top->evlist, help, &hbt,
top->min_percent,
- &top->session->header.env);
+ &top->session->header.env, true);
done = 1;
return NULL;
diff --git a/tools/perf/ui/browsers/hists.c b/tools/perf/ui/browsers/hists.c
index 68146f4..6495ee5 100644
--- a/tools/perf/ui/browsers/hists.c
+++ b/tools/perf/ui/browsers/hists.c
@@ -608,7 +608,8 @@ static int hist_browser__title(struct hist_browser *browser, char *bf, size_t si
return browser->title ? browser->title(browser, bf, size) : 0;
}
-int hist_browser__run(struct hist_browser *browser, const char *help)
+int hist_browser__run(struct hist_browser *browser, const char *help,
+ bool warn_lost_event)
{
int key;
char title[160];
@@ -638,8 +639,9 @@ int hist_browser__run(struct hist_browser *browser, const char *help)
nr_entries = hist_browser__nr_entries(browser);
ui_browser__update_nr_entries(&browser->b, nr_entries);
- if (browser->hists->stats.nr_lost_warned !=
- browser->hists->stats.nr_events[PERF_RECORD_LOST]) {
+ if (warn_lost_event &&
+ (browser->hists->stats.nr_lost_warned !=
+ browser->hists->stats.nr_events[PERF_RECORD_LOST])) {
browser->hists->stats.nr_lost_warned =
browser->hists->stats.nr_events[PERF_RECORD_LOST];
ui_browser__warn_lost_events(&browser->b);
@@ -2763,7 +2765,8 @@ static int perf_evsel__hists_browse(struct perf_evsel *evsel, int nr_events,
bool left_exits,
struct hist_browser_timer *hbt,
float min_pcnt,
- struct perf_env *env)
+ struct perf_env *env,
+ bool warn_lost_event)
{
struct hists *hists = evsel__hists(evsel);
struct hist_browser *browser = perf_evsel_browser__new(evsel, hbt, env);
@@ -2844,7 +2847,8 @@ static int perf_evsel__hists_browse(struct perf_evsel *evsel, int nr_events,
nr_options = 0;
- key = hist_browser__run(browser, helpline);
+ key = hist_browser__run(browser, helpline,
+ warn_lost_event);
if (browser->he_selection != NULL) {
thread = hist_browser__selected_thread(browser);
@@ -3184,7 +3188,8 @@ static void perf_evsel_menu__write(struct ui_browser *browser,
static int perf_evsel_menu__run(struct perf_evsel_menu *menu,
int nr_events, const char *help,
- struct hist_browser_timer *hbt)
+ struct hist_browser_timer *hbt,
+ bool warn_lost_event)
{
struct perf_evlist *evlist = menu->b.priv;
struct perf_evsel *pos;
@@ -3203,7 +3208,9 @@ static int perf_evsel_menu__run(struct perf_evsel_menu *menu,
case K_TIMER:
hbt->timer(hbt->arg);
- if (!menu->lost_events_warned && menu->lost_events) {
+ if (!menu->lost_events_warned &&
+ menu->lost_events &&
+ warn_lost_event) {
ui_browser__warn_lost_events(&menu->b);
menu->lost_events_warned = true;
}
@@ -3224,7 +3231,8 @@ browse_hists:
key = perf_evsel__hists_browse(pos, nr_events, help,
true, hbt,
menu->min_pcnt,
- menu->env);
+ menu->env,
+ warn_lost_event);
ui_browser__show_title(&menu->b, title);
switch (key) {
case K_TAB:
@@ -3282,7 +3290,8 @@ static int __perf_evlist__tui_browse_hists(struct perf_evlist *evlist,
int nr_entries, const char *help,
struct hist_browser_timer *hbt,
float min_pcnt,
- struct perf_env *env)
+ struct perf_env *env,
+ bool warn_lost_event)
{
struct perf_evsel *pos;
struct perf_evsel_menu menu = {
@@ -3309,13 +3318,15 @@ static int __perf_evlist__tui_browse_hists(struct perf_evlist *evlist,
menu.b.width = line_len;
}
- return perf_evsel_menu__run(&menu, nr_entries, help, hbt);
+ return perf_evsel_menu__run(&menu, nr_entries, help,
+ hbt, warn_lost_event);
}
int perf_evlist__tui_browse_hists(struct perf_evlist *evlist, const char *help,
struct hist_browser_timer *hbt,
float min_pcnt,
- struct perf_env *env)
+ struct perf_env *env,
+ bool warn_lost_event)
{
int nr_entries = evlist->nr_entries;
@@ -3325,7 +3336,7 @@ single_entry:
return perf_evsel__hists_browse(first, nr_entries, help,
false, hbt, min_pcnt,
- env);
+ env, warn_lost_event);
}
if (symbol_conf.event_group) {
@@ -3342,5 +3353,6 @@ single_entry:
}
return __perf_evlist__tui_browse_hists(evlist, nr_entries, help,
- hbt, min_pcnt, env);
+ hbt, min_pcnt, env,
+ warn_lost_event);
}
diff --git a/tools/perf/ui/browsers/hists.h b/tools/perf/ui/browsers/hists.h
index ba43177..9428bee 100644
--- a/tools/perf/ui/browsers/hists.h
+++ b/tools/perf/ui/browsers/hists.h
@@ -28,7 +28,8 @@ struct hist_browser {
struct hist_browser *hist_browser__new(struct hists *hists);
void hist_browser__delete(struct hist_browser *browser);
-int hist_browser__run(struct hist_browser *browser, const char *help);
+int hist_browser__run(struct hist_browser *browser, const char *help,
+ bool warn_lost_event);
void hist_browser__init(struct hist_browser *browser,
struct hists *hists);
#endif /* _PERF_UI_BROWSER_HISTS_H_ */
diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h
index f6630cb..02721b57 100644
--- a/tools/perf/util/hist.h
+++ b/tools/perf/util/hist.h
@@ -430,7 +430,8 @@ int hist_entry__tui_annotate(struct hist_entry *he, struct perf_evsel *evsel,
int perf_evlist__tui_browse_hists(struct perf_evlist *evlist, const char *help,
struct hist_browser_timer *hbt,
float min_pcnt,
- struct perf_env *env);
+ struct perf_env *env,
+ bool warn_lost_event);
int script_browse(const char *script_opt);
#else
static inline
@@ -438,7 +439,8 @@ int perf_evlist__tui_browse_hists(struct perf_evlist *evlist __maybe_unused,
const char *help __maybe_unused,
struct hist_browser_timer *hbt __maybe_unused,
float min_pcnt __maybe_unused,
- struct perf_env *env __maybe_unused)
+ struct perf_env *env __maybe_unused,
+ bool warn_lost_event __maybe_unused)
{
return 0;
}
Commit-ID: ee023de05f35484691f7d9e5c1f92195ac4d64d2
Gitweb: https://git.kernel.org/tip/ee023de05f35484691f7d9e5c1f92195ac4d64d2
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:22 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 15 Feb 2018 09:53:15 -0300
perf mmap: Introduce perf_mmap__read_done()
The direction of overwrite mode is backward. The last perf_mmap__read()
will set tail to map->prev. Need to correct the map->prev to head which
is the end of next read.
It will be used later.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/util/mmap.c | 11 +++++++++++
tools/perf/util/mmap.h | 1 +
2 files changed, 12 insertions(+)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index 125bfda..4f59eae 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -338,3 +338,14 @@ int perf_mmap__push(struct perf_mmap *md, bool overwrite,
out:
return rc;
}
+
+/*
+ * Mandatory for overwrite mode
+ * The direction of overwrite mode is backward.
+ * The last perf_mmap__read() will set tail to map->prev.
+ * Need to correct the map->prev to head which is the end of next read.
+ */
+void perf_mmap__read_done(struct perf_mmap *map)
+{
+ map->prev = perf_mmap__read_head(map);
+}
diff --git a/tools/perf/util/mmap.h b/tools/perf/util/mmap.h
index 9ab2b48..95549d4 100644
--- a/tools/perf/util/mmap.h
+++ b/tools/perf/util/mmap.h
@@ -96,4 +96,5 @@ size_t perf_mmap__mmap_len(struct perf_mmap *map);
int perf_mmap__read_init(struct perf_mmap *md, bool overwrite,
u64 *startp, u64 *endp);
+void perf_mmap__read_done(struct perf_mmap *map);
#endif /*__PERF_MMAP_H */
Commit-ID: 8872481bd04850b19e053dc579de5a11b83b16fc
Gitweb: https://git.kernel.org/tip/8872481bd04850b19e053dc579de5a11b83b16fc
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:19 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 15 Feb 2018 09:52:22 -0300
perf mmap: Introduce perf_mmap__read_init()
The new function perf_mmap__read_init() is factored out from
perf_mmap__push().
It is to calculate the 'start' and 'end' of the available data in
ringbuffer.
No functional change.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/util/mmap.c | 37 +++++++++++++++++++++++++++----------
tools/perf/util/mmap.h | 2 ++
2 files changed, 29 insertions(+), 10 deletions(-)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index fbbbe87..c19a4e6 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -267,24 +267,24 @@ static int overwrite_rb_find_range(void *buf, int mask, u64 head, u64 *start, u6
return -1;
}
-int perf_mmap__push(struct perf_mmap *md, bool overwrite,
- void *to, int push(void *to, void *buf, size_t size))
+/*
+ * Report the start and end of the available data in ringbuffer
+ */
+int perf_mmap__read_init(struct perf_mmap *md, bool overwrite,
+ u64 *startp, u64 *endp)
{
u64 head = perf_mmap__read_head(md);
u64 old = md->prev;
- u64 end, start;
unsigned char *data = md->base + page_size;
unsigned long size;
- void *buf;
- int rc = 0;
- start = overwrite ? head : old;
- end = overwrite ? old : head;
+ *startp = overwrite ? head : old;
+ *endp = overwrite ? old : head;
- if (start == end)
+ if (*startp == *endp)
return 0;
- size = end - start;
+ size = *endp - *startp;
if (size > (unsigned long)(md->mask) + 1) {
if (!overwrite) {
WARN_ONCE(1, "failed to keep up with mmap data. (warn only once)\n");
@@ -298,10 +298,27 @@ int perf_mmap__push(struct perf_mmap *md, bool overwrite,
* Backward ring buffer is full. We still have a chance to read
* most of data from it.
*/
- if (overwrite_rb_find_range(data, md->mask, head, &start, &end))
+ if (overwrite_rb_find_range(data, md->mask, head, startp, endp))
return -1;
}
+ return 1;
+}
+
+int perf_mmap__push(struct perf_mmap *md, bool overwrite,
+ void *to, int push(void *to, void *buf, size_t size))
+{
+ u64 head = perf_mmap__read_head(md);
+ u64 end, start;
+ unsigned char *data = md->base + page_size;
+ unsigned long size;
+ void *buf;
+ int rc = 0;
+
+ rc = perf_mmap__read_init(md, overwrite, &start, &end);
+ if (rc < 1)
+ return rc;
+
size = end - start;
if ((start & md->mask) + size != (end & md->mask)) {
diff --git a/tools/perf/util/mmap.h b/tools/perf/util/mmap.h
index e43d7b5..9ab2b48 100644
--- a/tools/perf/util/mmap.h
+++ b/tools/perf/util/mmap.h
@@ -94,4 +94,6 @@ int perf_mmap__push(struct perf_mmap *md, bool backward,
size_t perf_mmap__mmap_len(struct perf_mmap *map);
+int perf_mmap__read_init(struct perf_mmap *md, bool overwrite,
+ u64 *startp, u64 *endp);
#endif /*__PERF_MMAP_H */
Commit-ID: b4b036b4c76341a5034e872aca3727c4988a7304
Gitweb: https://git.kernel.org/tip/b4b036b4c76341a5034e872aca3727c4988a7304
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:21 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 15 Feb 2018 09:53:06 -0300
perf mmap: Discard 'prev' in perf_mmap__read()
The 'start' and 'prev' variables are duplicates in perf_mmap__read().
Use 'map->prev' to replace 'start' in perf_mmap__read_*().
Suggested-by: Wang Nan <[email protected]>
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/util/mmap.c | 28 ++++++++++------------------
1 file changed, 10 insertions(+), 18 deletions(-)
diff --git a/tools/perf/util/mmap.c b/tools/perf/util/mmap.c
index 38fa69d..125bfda 100644
--- a/tools/perf/util/mmap.c
+++ b/tools/perf/util/mmap.c
@@ -22,29 +22,27 @@ size_t perf_mmap__mmap_len(struct perf_mmap *map)
/* When check_messup is true, 'end' must points to a good entry */
static union perf_event *perf_mmap__read(struct perf_mmap *map,
- u64 start, u64 end, u64 *prev)
+ u64 *startp, u64 end)
{
unsigned char *data = map->base + page_size;
union perf_event *event = NULL;
- int diff = end - start;
+ int diff = end - *startp;
if (diff >= (int)sizeof(event->header)) {
size_t size;
- event = (union perf_event *)&data[start & map->mask];
+ event = (union perf_event *)&data[*startp & map->mask];
size = event->header.size;
- if (size < sizeof(event->header) || diff < (int)size) {
- event = NULL;
- goto broken_event;
- }
+ if (size < sizeof(event->header) || diff < (int)size)
+ return NULL;
/*
* Event straddles the mmap boundary -- header should always
* be inside due to u64 alignment of output.
*/
- if ((start & map->mask) + size != ((start + size) & map->mask)) {
- unsigned int offset = start;
+ if ((*startp & map->mask) + size != ((*startp + size) & map->mask)) {
+ unsigned int offset = *startp;
unsigned int len = min(sizeof(*event), size), cpy;
void *dst = map->event_copy;
@@ -59,20 +57,15 @@ static union perf_event *perf_mmap__read(struct perf_mmap *map,
event = (union perf_event *)map->event_copy;
}
- start += size;
+ *startp += size;
}
-broken_event:
- if (prev)
- *prev = start;
-
return event;
}
union perf_event *perf_mmap__read_forward(struct perf_mmap *map)
{
u64 head;
- u64 old = map->prev;
/*
* Check if event was unmapped due to a POLLHUP/POLLERR.
@@ -82,13 +75,12 @@ union perf_event *perf_mmap__read_forward(struct perf_mmap *map)
head = perf_mmap__read_head(map);
- return perf_mmap__read(map, old, head, &map->prev);
+ return perf_mmap__read(map, &map->prev, head);
}
union perf_event *perf_mmap__read_backward(struct perf_mmap *map)
{
u64 head, end;
- u64 start = map->prev;
/*
* Check if event was unmapped due to a POLLHUP/POLLERR.
@@ -118,7 +110,7 @@ union perf_event *perf_mmap__read_backward(struct perf_mmap *map)
else
end = head + map->mask + 1;
- return perf_mmap__read(map, start, end, &map->prev);
+ return perf_mmap__read(map, &map->prev, end);
}
void perf_mmap__read_catchup(struct perf_mmap *map)
Commit-ID: a1ff5b05e988ca3620027148cd61013408ea4194
Gitweb: https://git.kernel.org/tip/a1ff5b05e988ca3620027148cd61013408ea4194
Author: Kan Liang <[email protected]>
AuthorDate: Thu, 18 Jan 2018 13:26:30 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Thu, 15 Feb 2018 09:56:43 -0300
perf top: Remove lost events checking
There would be some records lost in overwrite mode because of pausing
the ringbuffer. It has little impact for the accuracy of the snapshot
and could be tolerated by 'perf top'.
Remove the lost events checking.
Signed-off-by: Kan Liang <[email protected]>
Acked-by: Jiri Olsa <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jin Yao <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Wang Nan <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
tools/perf/builtin-top.c | 8 +++++---
1 file changed, 5 insertions(+), 3 deletions(-)
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index 7def861..5965306 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -283,8 +283,9 @@ static void perf_top__print_sym_table(struct perf_top *top)
printf("%-*.*s\n", win_width, win_width, graph_dotted_line);
- if (hists->stats.nr_lost_warned !=
- hists->stats.nr_events[PERF_RECORD_LOST]) {
+ if (!top->record_opts.overwrite &&
+ (hists->stats.nr_lost_warned !=
+ hists->stats.nr_events[PERF_RECORD_LOST])) {
hists->stats.nr_lost_warned =
hists->stats.nr_events[PERF_RECORD_LOST];
color_fprintf(stdout, PERF_COLOR_RED,
@@ -611,7 +612,8 @@ static void *display_thread_tui(void *arg)
perf_evlist__tui_browse_hists(top->evlist, help, &hbt,
top->min_percent,
- &top->session->header.env, true);
+ &top->session->header.env,
+ !top->record_opts.overwrite);
done = 1;
return NULL;