linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH v1] perf session: fix decompression of PERF_RECORD_COMPRESSED records
@ 2019-11-15  9:05 Alexey Budankov
  2019-11-15 15:11 ` Jiri Olsa
  0 siblings, 1 reply; 4+ messages in thread
From: Alexey Budankov @ 2019-11-15  9:05 UTC (permalink / raw)
  To: Arnaldo Carvalho de Melo
  Cc: Jiri Olsa, Namhyung Kim, Peter Zijlstra, Alexander Shishkin,
	Ingo Molnar, Andi Kleen, linux-kernel


Avoid termination of trace loading in case the last record in 
the decompressed buffer partly resides in the following 
mmaped PERF_RECORD_COMPRESSED record. In this case NULL value
returned by fetch_mmaped_event() means to proceed to the next 
mmaped record then decompress it and load compressed events. 

The issue can be reproduced like this:

  $ perf record -z -- some_long_running_workload
  $ perf report --stdio -vv
  decomp (B): 44519 to 163000
  decomp (B): 48119 to 174800
  decomp (B): 65527 to 131072
  fetch_mmaped_event: head=0x1ffe0 event->header_size=0x28, mmap_size=0x20000: fuzzed perf.data?
  Error:
  failed to process sample
  ...

Testing:
71: Zstd perf.data compression/decompression              : Ok

Fixes: 57fc032ad643 ("perf session: Avoid infinite loop when seeing invalid header.size")
Link: https://marc.info/?l=linux-kernel&m=156580812427554&w=2
Co-developed-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Alexey Budankov <alexey.budankov@linux.intel.com>
---
 tools/perf/util/session.c | 47 +++++++++++++++++++++++++++++++++--------------
 1 file changed, 33 insertions(+), 14 deletions(-)

diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c
index f07b8ecb91bc..3f6f812ec4ed 100644
--- a/tools/perf/util/session.c
+++ b/tools/perf/util/session.c
@@ -1957,9 +1957,31 @@ static int __perf_session__process_pipe_events(struct perf_session *session)
 	return err;
 }
 
+static union perf_event *
+prefetch_event(char *buf, u64 head, size_t mmap_size,
+	       bool needs_swap, union perf_event *ret);
+
 static union perf_event *
 fetch_mmaped_event(struct perf_session *session,
 		   u64 head, size_t mmap_size, char *buf)
+{
+	return prefetch_event(buf, head, mmap_size,
+			      session->header.needs_swap,
+			      ERR_PTR(-EINVAL));
+}
+
+static union perf_event *
+fetch_decomp_event(struct perf_session *session,
+		   u64 head, size_t mmap_size, char *buf)
+{
+	return prefetch_event(buf, head, mmap_size,
+			      session->header.needs_swap,
+			      NULL);
+}
+
+static union perf_event *
+prefetch_event(char *buf, u64 head, size_t mmap_size,
+	       bool needs_swap, union perf_event *ret)
 {
 	union perf_event *event;
 
@@ -1971,20 +1993,20 @@ fetch_mmaped_event(struct perf_session *session,
 		return NULL;
 
 	event = (union perf_event *)(buf + head);
+	if (needs_swap)
+		perf_event_header__bswap(&event->header);
 
-	if (session->header.needs_swap)
+	if (head + event->header.size <= mmap_size)
+		return event;
+
+	/* We're not fetching the event so swap back again */
+	if (needs_swap)
 		perf_event_header__bswap(&event->header);
 
-	if (head + event->header.size > mmap_size) {
-		/* We're not fetching the event so swap back again */
-		if (session->header.needs_swap)
-			perf_event_header__bswap(&event->header);
-		pr_debug("%s: head=%#" PRIx64 " event->header_size=%#x, mmap_size=%#zx: fuzzed perf.data?\n",
-			 __func__, head, event->header.size, mmap_size);
-		return ERR_PTR(-EINVAL);
-	}
+	pr_debug("%s: head=%#" PRIx64 " event->header_size=%#x, mmap_size=%#zx\n",
+		__func__, head, event->header.size, mmap_size);
 
-	return event;
+	return ret;
 }
 
 static int __perf_session__process_decomp_events(struct perf_session *session)
@@ -1997,10 +2019,7 @@ static int __perf_session__process_decomp_events(struct perf_session *session)
 		return 0;
 
 	while (decomp->head < decomp->size && !session_done()) {
-		union perf_event *event = fetch_mmaped_event(session, decomp->head, decomp->size, decomp->data);
-
-		if (IS_ERR(event))
-			return PTR_ERR(event);
+		union perf_event *event = fetch_decomp_event(session, decomp->head, decomp->size, decomp->data);
 
 		if (!event)
 			break;


^ permalink raw reply related	[flat|nested] 4+ messages in thread

* Re: [PATCH v1] perf session: fix decompression of PERF_RECORD_COMPRESSED records
  2019-11-15  9:05 [PATCH v1] perf session: fix decompression of PERF_RECORD_COMPRESSED records Alexey Budankov
@ 2019-11-15 15:11 ` Jiri Olsa
  2019-11-15 17:09   ` Alexey Budankov
  0 siblings, 1 reply; 4+ messages in thread
From: Jiri Olsa @ 2019-11-15 15:11 UTC (permalink / raw)
  To: Alexey Budankov
  Cc: Arnaldo Carvalho de Melo, Namhyung Kim, Peter Zijlstra,
	Alexander Shishkin, Ingo Molnar, Andi Kleen, linux-kernel

On Fri, Nov 15, 2019 at 12:05:14PM +0300, Alexey Budankov wrote:
> 
> Avoid termination of trace loading in case the last record in 
> the decompressed buffer partly resides in the following 
> mmaped PERF_RECORD_COMPRESSED record. In this case NULL value
> returned by fetch_mmaped_event() means to proceed to the next 
> mmaped record then decompress it and load compressed events. 
> 
> The issue can be reproduced like this:
> 
>   $ perf record -z -- some_long_running_workload
>   $ perf report --stdio -vv
>   decomp (B): 44519 to 163000
>   decomp (B): 48119 to 174800
>   decomp (B): 65527 to 131072
>   fetch_mmaped_event: head=0x1ffe0 event->header_size=0x28, mmap_size=0x20000: fuzzed perf.data?
>   Error:
>   failed to process sample
>   ...
> 
> Testing:
> 71: Zstd perf.data compression/decompression              : Ok
> 
> Fixes: 57fc032ad643 ("perf session: Avoid infinite loop when seeing invalid header.size")
> Link: https://marc.info/?l=linux-kernel&m=156580812427554&w=2
> Co-developed-by: Jiri Olsa <jolsa@kernel.org>
> Signed-off-by: Alexey Budankov <alexey.budankov@linux.intel.com>
> ---
>  tools/perf/util/session.c | 47 +++++++++++++++++++++++++++++++++--------------
>  1 file changed, 33 insertions(+), 14 deletions(-)
> 
> diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c
> index f07b8ecb91bc..3f6f812ec4ed 100644
> --- a/tools/perf/util/session.c
> +++ b/tools/perf/util/session.c
> @@ -1957,9 +1957,31 @@ static int __perf_session__process_pipe_events(struct perf_session *session)
>  	return err;
>  }
>  
> +static union perf_event *
> +prefetch_event(char *buf, u64 head, size_t mmap_size,
> +	       bool needs_swap, union perf_event *ret);

why not move prefetch_event definition in here?
I don't see any need for the static declaration..

> +
>  static union perf_event *
>  fetch_mmaped_event(struct perf_session *session,
>  		   u64 head, size_t mmap_size, char *buf)
> +{
> +	return prefetch_event(buf, head, mmap_size,
> +			      session->header.needs_swap,
> +			      ERR_PTR(-EINVAL));
> +}
> +
> +static union perf_event *
> +fetch_decomp_event(struct perf_session *session,
> +		   u64 head, size_t mmap_size, char *buf)
> +{

if this is decomp specific, it could take 'struct decomp*' as argument

> +	return prefetch_event(buf, head, mmap_size,
> +			      session->header.needs_swap,
> +			      NULL);
> +}
> +
> +static union perf_event *
> +prefetch_event(char *buf, u64 head, size_t mmap_size,
> +	       bool needs_swap, union perf_event *ret)
>  {

'error' might be more suitable then ret in here

thanks,
jirka


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH v1] perf session: fix decompression of PERF_RECORD_COMPRESSED records
  2019-11-15 15:11 ` Jiri Olsa
@ 2019-11-15 17:09   ` Alexey Budankov
  2019-11-18 10:46     ` Jiri Olsa
  0 siblings, 1 reply; 4+ messages in thread
From: Alexey Budankov @ 2019-11-15 17:09 UTC (permalink / raw)
  To: Jiri Olsa
  Cc: Arnaldo Carvalho de Melo, Namhyung Kim, Peter Zijlstra,
	Alexander Shishkin, Ingo Molnar, Andi Kleen, linux-kernel


On 15.11.2019 18:11, Jiri Olsa wrote:
> On Fri, Nov 15, 2019 at 12:05:14PM +0300, Alexey Budankov wrote:
<SNIP>
>> diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c
>> index f07b8ecb91bc..3f6f812ec4ed 100644
>> --- a/tools/perf/util/session.c
>> +++ b/tools/perf/util/session.c
>> @@ -1957,9 +1957,31 @@ static int __perf_session__process_pipe_events(struct perf_session *session)
>>  	return err;
>>  }
>>  
>> +static union perf_event *
>> +prefetch_event(char *buf, u64 head, size_t mmap_size,
>> +	       bool needs_swap, union perf_event *ret);
> 
> why not move prefetch_event definition in here?
> I don't see any need for the static declaration..

It is just for the sake of more readable patch formatting 
and, yes, could be avoided and replaced by the definition.

> 
>> +
>>  static union perf_event *
>>  fetch_mmaped_event(struct perf_session *session,
>>  		   u64 head, size_t mmap_size, char *buf)
>> +{
>> +	return prefetch_event(buf, head, mmap_size,
>> +			      session->header.needs_swap,
>> +			      ERR_PTR(-EINVAL));
>> +}
>> +
>> +static union perf_event *
>> +fetch_decomp_event(struct perf_session *session,
>> +		   u64 head, size_t mmap_size, char *buf)
>> +{
> 
> if this is decomp specific, it could take 'struct decomp*' as argument

Well, it makes sense. whole session object is not required here.
Just session->header.needs_swap could be passed as a param.
Shall we make it like this?

static union perf_event * 
fetch_decomp_event(u64 head, size_t mmap_size, char *buf, bool needs_swap)

> 
>> +	return prefetch_event(buf, head, mmap_size,
>> +			      session->header.needs_swap,
>> +			      NULL);
>> +}
>> +
>> +static union perf_event *
>> +prefetch_event(char *buf, u64 head, size_t mmap_size,
>> +	       bool needs_swap, union perf_event *ret)
>>  {
> 
> 'error' might be more suitable then ret in here

Accepted for v2.

~Alexey

> 
> thanks,
> jirka
> 
> 

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH v1] perf session: fix decompression of PERF_RECORD_COMPRESSED records
  2019-11-15 17:09   ` Alexey Budankov
@ 2019-11-18 10:46     ` Jiri Olsa
  0 siblings, 0 replies; 4+ messages in thread
From: Jiri Olsa @ 2019-11-18 10:46 UTC (permalink / raw)
  To: Alexey Budankov
  Cc: Arnaldo Carvalho de Melo, Namhyung Kim, Peter Zijlstra,
	Alexander Shishkin, Ingo Molnar, Andi Kleen, linux-kernel

On Fri, Nov 15, 2019 at 08:09:19PM +0300, Alexey Budankov wrote:
> 
> On 15.11.2019 18:11, Jiri Olsa wrote:
> > On Fri, Nov 15, 2019 at 12:05:14PM +0300, Alexey Budankov wrote:
> <SNIP>
> >> diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c
> >> index f07b8ecb91bc..3f6f812ec4ed 100644
> >> --- a/tools/perf/util/session.c
> >> +++ b/tools/perf/util/session.c
> >> @@ -1957,9 +1957,31 @@ static int __perf_session__process_pipe_events(struct perf_session *session)
> >>  	return err;
> >>  }
> >>  
> >> +static union perf_event *
> >> +prefetch_event(char *buf, u64 head, size_t mmap_size,
> >> +	       bool needs_swap, union perf_event *ret);
> > 
> > why not move prefetch_event definition in here?
> > I don't see any need for the static declaration..
> 
> It is just for the sake of more readable patch formatting 
> and, yes, could be avoided and replaced by the definition.

I think we're trying to avoid static declarations

> 
> > 
> >> +
> >>  static union perf_event *
> >>  fetch_mmaped_event(struct perf_session *session,
> >>  		   u64 head, size_t mmap_size, char *buf)
> >> +{
> >> +	return prefetch_event(buf, head, mmap_size,
> >> +			      session->header.needs_swap,
> >> +			      ERR_PTR(-EINVAL));
> >> +}
> >> +
> >> +static union perf_event *
> >> +fetch_decomp_event(struct perf_session *session,
> >> +		   u64 head, size_t mmap_size, char *buf)
> >> +{
> > 
> > if this is decomp specific, it could take 'struct decomp*' as argument
> 
> Well, it makes sense. whole session object is not required here.
> Just session->header.needs_swap could be passed as a param.
> Shall we make it like this?
> 
> static union perf_event * 
> fetch_decomp_event(u64 head, size_t mmap_size, char *buf, bool needs_swap)

ok, I just saw the call passing all the stuff from 'struct decomp'
so I thought it'd save some arguments if we pass the  struct itself
instead

jirka


^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2019-11-18 10:47 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-11-15  9:05 [PATCH v1] perf session: fix decompression of PERF_RECORD_COMPRESSED records Alexey Budankov
2019-11-15 15:11 ` Jiri Olsa
2019-11-15 17:09   ` Alexey Budankov
2019-11-18 10:46     ` Jiri Olsa

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).