lkml.org 
[lkml]   [2013]   [Jul]   [17]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH V6 06/12] perf tools: remove unnecessary callchain validation
On 16/07/13 15:05, Jiri Olsa wrote:
> On Tue, Jul 16, 2013 at 09:38:12AM +0300, Adrian Hunter wrote:
>
> SNIP
>
>> }
>> }
>>
>> -static int perf_session__preprocess_sample(struct perf_session *session,
>> - union perf_event *event, struct perf_sample *sample)
>> -{
>> - if (event->header.type != PERF_RECORD_SAMPLE ||
>> - !sample->callchain)
>> - return 0;
>> -
>> - if (!ip_callchain__valid(sample->callchain, event)) {
>> - pr_debug("call-chain problem with event, skipping it.\n");
>> - ++session->stats.nr_invalid_chains;
>> - session->stats.total_invalid_chains += sample->period;
>
> How about the '*invalid_chains' stats here? I dont see
> it incremented in the parsing routine.
>
> Also the current behaviour is to increments stats for invalid
> callchains, but dont fail. With your changes we fail during the
> parsing.

It would fail during parsing sometimes anyway. The code was:

if (type & PERF_SAMPLE_CALLCHAIN) {
if (sample_overlap(event, array, sizeof(data->callchain->nr)))
return -EFAULT;

data->callchain = (struct ip_callchain *)array;

if (sample_overlap(event, array, data->callchain->nr))
return -EFAULT;

array += 1 + data->callchain->nr;
}

But sample overlap did not handle size being effectively negative
i.e. 'offset + size' overflows

static bool sample_overlap(const union perf_event *event,
const void *offset, u64 size)
{
const void *base = event;

if (offset + size > base + event->header.size)
return true;

return false;
}


>
> On the other hand.. maybe we should fail ;-) I think that
> invalid callchain data is serious enough to be overlooked
> by not seeing the nr_invalid_chains got incremented.
>
> let's see other comments and then silently push it :-)
>
> jirka
>
>



\
 
 \ /
  Last update: 2013-07-17 10:42    [W:0.045 / U:2.344 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site