[PATCH 1/1] perf,tools: error out unsupported group leader immediately for perf stat

[PATCH 1/1] perf,tools: error out unsupported group leader immediately for perf stat

[kan.liang]

From: Kan Liang <kan.liang@intel.com>

perf stat ignores the unsupported event and continue to count supported
event. But if the unsupported event is group leader, perf tool will
crash. After applying this patch, the unsupported group leader will
error out immediately.

Without this patch:
$ perf stat -x, -e '{node-prefetch-refs,cycles}' --  sleep 1
perf: util/evsel.c:1009: get_group_fd: Assertion `!(fd == -1)' failed.
Aborted (core dumped)

With this patch:
$ perf stat -x, -e '{node-prefetch-refs,cycles}' --  sleep 1
Error:
The node-prefetch-refs event is not supported.

Signed-off-by: Kan Liang <kan.liang@intel.com>
---
 tools/perf/builtin-stat.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index ff3d258..b24ecee 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -549,7 +549,10 @@ static int __run_perf_stat(int argc, const char **argv)
 					ui__warning("%s event is not supported by the kernel.\n",
 						    perf_evsel__name(counter));
 				counter->supported = false;
-				continue;
+
+				if ((counter->leader != counter) ||
+				    !(counter->leader->nr_members > 1))
+					continue;
 			}
 
 			perf_evsel__open_strerror(counter, &target,
-- 
1.8.3.1

Re: [PATCH 1/1] perf,tools: error out unsupported group leader immediately for perf stat

[Arnaldo Carvalho de Melo]

Em Thu, Jun 11, 2015 at 02:32:40AM -0400, kan.liang@intel.com escreveu:
> perf stat ignores the unsupported event and continue to count supported
> event. But if the unsupported event is group leader, perf tool will
> crash. After applying this patch, the unsupported group leader will
> error out immediately.
 
> Without this patch:
> $ perf stat -x, -e '{node-prefetch-refs,cycles}' --  sleep 1
> perf: util/evsel.c:1009: get_group_fd: Assertion `!(fd == -1)' failed.
> Aborted (core dumped)
 
> With this patch:
> $ perf stat -x, -e '{node-prefetch-refs,cycles}' --  sleep 1
> Error:
> The node-prefetch-refs event is not supported.

I get something different, maybe I am missing a patch?

With this patch:

  [acme@zoo linux]$ perf stat -x, -e '{node-prefetch-refs,cycles}' -- sleep 1
  Error:
  The sys_perf_event_open() syscall returned with 22 (Invalid argument)
  for event (node-prefetch-refs).
  /bin/dmesg may provide additional information.
  No CONFIG_PERF_EVENTS=y kernel support configured?
  [acme@zoo linux]$ 

Applying it anyway, as it avoids a core dump and the message, albeit not
optimal, points to a problem with that event.

Have you tried this with other tools, like perf record?

- Arnaldo

RE: [PATCH 1/1] perf,tools: error out unsupported group leader immediately for perf stat

[Liang, Kan]

> 
> Em Thu, Jun 11, 2015 at 02:32:40AM -0400, kan.liang@intel.com escreveu:
> > perf stat ignores the unsupported event and continue to count
> > supported event. But if the unsupported event is group leader, perf
> > tool will crash. After applying this patch, the unsupported group
> > leader will error out immediately.
> 
> > Without this patch:
> > $ perf stat -x, -e '{node-prefetch-refs,cycles}' --  sleep 1
> > perf: util/evsel.c:1009: get_group_fd: Assertion `!(fd == -1)' failed.
> > Aborted (core dumped)
> 
> > With this patch:
> > $ perf stat -x, -e '{node-prefetch-refs,cycles}' --  sleep 1
> > Error:
> > The node-prefetch-refs event is not supported.
> 
> I get something different, maybe I am missing a patch?
> 
> With this patch:
> 
>   [acme@zoo linux]$ perf stat -x, -e '{node-prefetch-refs,cycles}' -- sleep 1
>   Error:
>   The sys_perf_event_open() syscall returned with 22 (Invalid argument)
>   for event (node-prefetch-refs).
>   /bin/dmesg may provide additional information.
>   No CONFIG_PERF_EVENTS=y kernel support configured?
>   [acme@zoo linux]$
>
I don't have other patches for this issue.
It looks they have different error code, ENOENT VS. EINVAL. 
I have no idea why. 

> Applying it anyway, as it avoids a core dump and the message, albeit not
> optimal, points to a problem with that event.
> 
> Have you tried this with other tools, like perf record?

Yes. Perf record and perf top  will error out immediately if any unsupported event found.
So they don't have such issue.

Thanks,
Kan

> 
> - Arnaldo

Re: [PATCH 1/1] perf,tools: error out unsupported group leader immediately for perf stat

[Arnaldo Carvalho de Melo]

Em Thu, Jun 11, 2015 at 02:26:47PM +0000, Liang, Kan escreveu:
> > Em Thu, Jun 11, 2015 at 02:32:40AM -0400, kan.liang@intel.com escreveu:
> > > With this patch:
> > > $ perf stat -x, -e '{node-prefetch-refs,cycles}' --  sleep 1
> > > Error:
> > > The node-prefetch-refs event is not supported.

> > I get something different, maybe I am missing a patch?

> > With this patch:

> >   [acme@zoo linux]$ perf stat -x, -e '{node-prefetch-refs,cycles}' -- sleep 1
> >   Error:
> >   The sys_perf_event_open() syscall returned with 22 (Invalid argument)
> >   for event (node-prefetch-refs).
> >   /bin/dmesg may provide additional information.
> >   No CONFIG_PERF_EVENTS=y kernel support configured?

> I don't have other patches for this issue.
> It looks they have different error code, ENOENT VS. EINVAL. 
> I have no idea why. 

Well, here I have:

[acme@zoo ~]$ uname -a
Linux zoo.ghostprotocols.net 3.19.7-200.fc21.x86_64 #1 SMP Thu May 7
22:00:21 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux

I am rebooting now, due to an unrelated reason, to 4.1.0-rc5+, lets see
if this explains this discrepancy...
 
> > Applying it anyway, as it avoids a core dump and the message, albeit not
> > optimal, points to a problem with that event.

> > Have you tried this with other tools, like perf record?
 
> Yes. Perf record and perf top  will error out immediately if any unsupported event found.
> So they don't have such issue.

Ok, need to revisit the event setup code again and remove the code
duplication once and for all :-\

- Arnaldo

[tip:perf/core] perf stat: Error out unsupported group leader immediately

[tip-bot for Kan Liang]

Commit-ID:  cb5ef60067c11cc8887122f6f168c21941c5d624
Gitweb:     http://git.kernel.org/tip/cb5ef60067c11cc8887122f6f168c21941c5d624
Author:     Kan Liang <kan.liang@intel.com>
AuthorDate: Thu, 11 Jun 2015 02:32:40 -0400
Committer:  Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Thu, 11 Jun 2015 11:07:40 -0300

perf stat: Error out unsupported group leader immediately

perf stat ignores the unsupported event and continue to count supported
event. But if the unsupported event is group leader, perf tool will
crash. After applying this patch, the unsupported group leader will
error out immediately.

Without this patch:

  $ perf stat -x, -e '{node-prefetch-refs,cycles}' --  sleep 1
  perf: util/evsel.c:1009: get_group_fd: Assertion `!(fd == -1)' failed.
  Aborted (core dumped)

With this patch:

  $ perf stat -x, -e '{node-prefetch-refs,cycles}' --  sleep 1
  Error:
  The node-prefetch-refs event is not supported.

Commiter note: Here I got a different output, but no core dump:

  [acme@zoo linux]$ perf stat -x, -e '{node-prefetch-refs,cycles}' -- sleep 1
  Error:
  The sys_perf_event_open() syscall returned with 22 (Invalid argument)
  for event (node-prefetch-refs).
  /bin/dmesg may provide additional information.
  No CONFIG_PERF_EVENTS=y kernel support configured?

Signed-off-by: Kan Liang <kan.liang@intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Andi Kleen <andi@firstfloor.org>
Link: http://lkml.kernel.org/r/1434004360-8570-1-git-send-email-kan.liang@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/builtin-stat.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index ff3d258..b24ecee 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -549,7 +549,10 @@ static int __run_perf_stat(int argc, const char **argv)
 					ui__warning("%s event is not supported by the kernel.\n",
 						    perf_evsel__name(counter));
 				counter->supported = false;
-				continue;
+
+				if ((counter->leader != counter) ||
+				    !(counter->leader->nr_members > 1))
+					continue;
 			}
 
 			perf_evsel__open_strerror(counter, &target,