On mmap(), perf_events generates a RECORD_MMAP record and then checks
which events are interested in this record. There are currently 2 versions
of mmap records: RECORD_MMAP and RECORD_MMAP2. MMAP2 is larger. The event
configuration controls which version the user level tool accepts. If the
event->attr.mmap2=1 field then MMAP2 record is returned. The
perf_event_mmap_output() takes care of this. It checks attr->mmap2 and corrects
the record fields before putting it in the sampling buffer of the event.
At the end the function restores the modified MMAP record fields.
The problem is that the function restores the size but not the
type. Thus, if a subsequent event only accepts MMAP type, then it would
instead receive an MMAP2 record with a size of MMAP record.
This patch fixes the problem by restoring the record type on exit.
Signed-off-by: Stephane Eranian <[email protected]>
---
kernel/events/core.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 5f59d848171e..f170cd12beed 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -7188,6 +7188,7 @@ static void perf_event_mmap_output(struct perf_event *event,
struct perf_output_handle handle;
struct perf_sample_data sample;
int size = mmap_event->event_id.header.size;
+ u32 type = mmap_event->event_id.header.type;
int ret;
if (!perf_event_mmap_match(event, data))
@@ -7231,6 +7232,7 @@ static void perf_event_mmap_output(struct perf_event *event,
perf_output_end(&handle);
out:
mmap_event->event_id.header.size = size;
+ mmap_event->event_id.header.type = type;
}
static void perf_event_mmap_event(struct perf_mmap_event *mmap_event)
--
2.21.0.360.g471c308f928-goog
Em Thu, Mar 07, 2019 at 10:52:33AM -0800, Stephane Eranian escreveu:
> On mmap(), perf_events generates a RECORD_MMAP record and then checks
> which events are interested in this record. There are currently 2 versions
> of mmap records: RECORD_MMAP and RECORD_MMAP2. MMAP2 is larger. The event
> configuration controls which version the user level tool accepts. If the
> event->attr.mmap2=1 field then MMAP2 record is returned. The
> perf_event_mmap_output() takes care of this. It checks attr->mmap2 and corrects
> the record fields before putting it in the sampling buffer of the event.
> At the end the function restores the modified MMAP record fields.
>
> The problem is that the function restores the size but not the
> type. Thus, if a subsequent event only accepts MMAP type, then it would
> instead receive an MMAP2 record with a size of MMAP record.
>
> This patch fixes the problem by restoring the record type on exit.
Right, simple enough, bug fixed, Peter, I'm taking this one, ok?
- Arnaldo
> Signed-off-by: Stephane Eranian <[email protected]>
> ---
> kernel/events/core.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/kernel/events/core.c b/kernel/events/core.c
> index 5f59d848171e..f170cd12beed 100644
> --- a/kernel/events/core.c
> +++ b/kernel/events/core.c
> @@ -7188,6 +7188,7 @@ static void perf_event_mmap_output(struct perf_event *event,
> struct perf_output_handle handle;
> struct perf_sample_data sample;
> int size = mmap_event->event_id.header.size;
> + u32 type = mmap_event->event_id.header.type;
> int ret;
>
> if (!perf_event_mmap_match(event, data))
> @@ -7231,6 +7232,7 @@ static void perf_event_mmap_output(struct perf_event *event,
> perf_output_end(&handle);
> out:
> mmap_event->event_id.header.size = size;
> + mmap_event->event_id.header.type = type;
> }
>
> static void perf_event_mmap_event(struct perf_mmap_event *mmap_event)
> --
> 2.21.0.360.g471c308f928-goog
On Thu, Mar 07, 2019 at 04:03:30PM -0300, Arnaldo Carvalho de Melo wrote:
> Em Thu, Mar 07, 2019 at 10:52:33AM -0800, Stephane Eranian escreveu:
> > On mmap(), perf_events generates a RECORD_MMAP record and then checks
> > which events are interested in this record. There are currently 2 versions
> > of mmap records: RECORD_MMAP and RECORD_MMAP2. MMAP2 is larger. The event
> > configuration controls which version the user level tool accepts. If the
> > event->attr.mmap2=1 field then MMAP2 record is returned. The
> > perf_event_mmap_output() takes care of this. It checks attr->mmap2 and corrects
> > the record fields before putting it in the sampling buffer of the event.
> > At the end the function restores the modified MMAP record fields.
> >
> > The problem is that the function restores the size but not the
> > type. Thus, if a subsequent event only accepts MMAP type, then it would
> > instead receive an MMAP2 record with a size of MMAP record.
> >
> > This patch fixes the problem by restoring the record type on exit.
>
> Right, simple enough, bug fixed, Peter, I'm taking this one, ok?
Acked-by: Peter Zijlstra (Intel) <[email protected]>
Em Thu, Mar 07, 2019 at 08:09:39PM +0100, Peter Zijlstra escreveu:
> On Thu, Mar 07, 2019 at 04:03:30PM -0300, Arnaldo Carvalho de Melo wrote:
> > Em Thu, Mar 07, 2019 at 10:52:33AM -0800, Stephane Eranian escreveu:
> > > On mmap(), perf_events generates a RECORD_MMAP record and then checks
> > > which events are interested in this record. There are currently 2 versions
> > > of mmap records: RECORD_MMAP and RECORD_MMAP2. MMAP2 is larger. The event
> > > configuration controls which version the user level tool accepts. If the
> > > event->attr.mmap2=1 field then MMAP2 record is returned. The
> > > perf_event_mmap_output() takes care of this. It checks attr->mmap2 and corrects
> > > the record fields before putting it in the sampling buffer of the event.
> > > At the end the function restores the modified MMAP record fields.
> > > The problem is that the function restores the size but not the
> > > type. Thus, if a subsequent event only accepts MMAP type, then it would
> > > instead receive an MMAP2 record with a size of MMAP record.
> > > This patch fixes the problem by restoring the record type on exit.
> > Right, simple enough, bug fixed, Peter, I'm taking this one, ok?
> Acked-by: Peter Zijlstra (Intel) <[email protected]>
BTW, added:
Fixes: 13d7a2410fa6 ("perf: Add attr->mmap2 attribute to an event")
For the stable guys to pick this up. Please try to add the fixes tag in
the future.
Thanks,
- Arnaldo
On Thu, Mar 7, 2019 at 11:26 AM Arnaldo Carvalho de Melo
<[email protected]> wrote:
>
> Em Thu, Mar 07, 2019 at 08:09:39PM +0100, Peter Zijlstra escreveu:
> > On Thu, Mar 07, 2019 at 04:03:30PM -0300, Arnaldo Carvalho de Melo wrote:
> > > Em Thu, Mar 07, 2019 at 10:52:33AM -0800, Stephane Eranian escreveu:
> > > > On mmap(), perf_events generates a RECORD_MMAP record and then checks
> > > > which events are interested in this record. There are currently 2 versions
> > > > of mmap records: RECORD_MMAP and RECORD_MMAP2. MMAP2 is larger. The event
> > > > configuration controls which version the user level tool accepts. If the
> > > > event->attr.mmap2=1 field then MMAP2 record is returned. The
> > > > perf_event_mmap_output() takes care of this. It checks attr->mmap2 and corrects
> > > > the record fields before putting it in the sampling buffer of the event.
> > > > At the end the function restores the modified MMAP record fields.
>
> > > > The problem is that the function restores the size but not the
> > > > type. Thus, if a subsequent event only accepts MMAP type, then it would
> > > > instead receive an MMAP2 record with a size of MMAP record.
>
> > > > This patch fixes the problem by restoring the record type on exit.
>
> > > Right, simple enough, bug fixed, Peter, I'm taking this one, ok?
>
> > Acked-by: Peter Zijlstra (Intel) <[email protected]>
>
> BTW, added:
>
> Fixes: 13d7a2410fa6 ("perf: Add attr->mmap2 attribute to an event")
>
> For the stable guys to pick this up. Please try to add the fixes tag in
> the future.
Ok, thanks.
>
>
> Thanks,
>
> - Arnaldo
Commit-ID: d9c1bb2f6a2157b38e8eb63af437cb22701d31ee
Gitweb: https://git.kernel.org/tip/d9c1bb2f6a2157b38e8eb63af437cb22701d31ee
Author: Stephane Eranian <[email protected]>
AuthorDate: Thu, 7 Mar 2019 10:52:33 -0800
Committer: Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Mon, 11 Mar 2019 11:56:02 -0300
perf/core: Restore mmap record type correctly
On mmap(), perf_events generates a RECORD_MMAP record and then checks
which events are interested in this record. There are currently 2
versions of mmap records: RECORD_MMAP and RECORD_MMAP2. MMAP2 is larger.
The event configuration controls which version the user level tool
accepts.
If the event->attr.mmap2=1 field then MMAP2 record is returned. The
perf_event_mmap_output() takes care of this. It checks attr->mmap2 and
corrects the record fields before putting it in the sampling buffer of
the event. At the end the function restores the modified MMAP record
fields.
The problem is that the function restores the size but not the type.
Thus, if a subsequent event only accepts MMAP type, then it would
instead receive an MMAP2 record with a size of MMAP record.
This patch fixes the problem by restoring the record type on exit.
Signed-off-by: Stephane Eranian <[email protected]>
Acked-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Fixes: 13d7a2410fa6 ("perf: Add attr->mmap2 attribute to an event")
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
---
kernel/events/core.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 6fb27b564730..514b8e014a2d 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -7189,6 +7189,7 @@ static void perf_event_mmap_output(struct perf_event *event,
struct perf_output_handle handle;
struct perf_sample_data sample;
int size = mmap_event->event_id.header.size;
+ u32 type = mmap_event->event_id.header.type;
int ret;
if (!perf_event_mmap_match(event, data))
@@ -7232,6 +7233,7 @@ static void perf_event_mmap_output(struct perf_event *event,
perf_output_end(&handle);
out:
mmap_event->event_id.header.size = size;
+ mmap_event->event_id.header.type = type;
}
static void perf_event_mmap_event(struct perf_mmap_event *mmap_event)