1264969 - crash in av_buffer_unref

Status: RESOLVED WORKSFORME

(Core :: Audio/Video: Playback, defect, P3)

Description

[Paul Silaghi, QA [:pauly]]

This bug was filed from the Socorro interface and is 
report bp-39916cd7-8d40-4967-ba24-b3f202160415.
=============================================================

[Note]:
- This issue is intermittent

[Affected versions]:
- 46b11

[Affected platforms]:
- Ubuntu 12.04 x86

[Steps to reproduce]:
1. Open facebook.com, yahoo.com, about:preferences
2. Switch tabs until the rendering issue in bug 1194358 shows up
3. Switch tabs a few more times

[Expected result]:
- no crash

[Actual result]:
- crash

[Regression range]:
- reproduced on 46b10
- not reproduced on 45

Comment 1

[Paul Silaghi, QA [:pauly]]

Couldn't reproduce the crash with layers.offmainthreadcomposition.enabled = FALSE

Comment 2

[Liz Henry (:lizzard) (relman/hg->git project)]

This was uncovered in testing for beta 10. I don't see any obvious crashes with this signature in crash-stats so I am not too concerned about this for 46 release.

Comment 3

[u279076]

This is a bug in Media not Graphics.

Paul, can you please try reproducing this with the open source Nouveau driver? We need to rule it out as a bug in NVIDIA's proprietary Linux driver. It'd also be useful to know if you can reproduce this in any other pre-release Firefox version and on any other machines. Thanks.

Note that we have two prior instances of this crash:
bp-e18d3331-8a3e-47c7-a080-d0e0b2160402 (46.0b4 on Windows on April 2nd)
bp-59e89632-b6ca-4db1-8c7a-524822160302 (46.0a2 on Linux on March 22nd)

We have no crashes reported in all history prior to March 22nd. All other crashes are from Paul.

Comment 4

[u279076]

(In reply to Anthony Hughes (:ashughes) [GFX][QA][Mentor] from comment #3)
> Paul, can you please try reproducing this with the open source Nouveau
> driver? We need to rule it out as a bug in NVIDIA's proprietary Linux
> driver. It'd also be useful to know if you can reproduce this in any other
> pre-release Firefox version and on any other machines. Thanks.

Sorry Paul, forgot the need-info flag. Please get this info as soon as possible.

Comment 5

[Paul Silaghi, QA [:pauly]]

(In reply to Anthony Hughes (:ashughes) [GFX][QA][Mentor] from comment #3)
> Paul, can you please try reproducing this with the open source Nouveau
> driver? We need to rule it out as a bug in NVIDIA's proprietary Linux
> driver. 
I'm not sure how to do that. In System/Additional Drivers I see "No proprietary drivers are in use on this system". But if I run "sudo lshw -c video" in the terminal, I see: "configuration: driver=nvidia latency=0". I also tried the instructions on http://askubuntu.com/questions/12937/remove-nvidia-driver-and-go-back-to-nouveau, but still no change.

> It'd also be useful to know if you can reproduce this in any other
> pre-release Firefox version and on any other machines. Thanks.
I can reproduce this on FX 46b1.
I can reproduce this only on a single machine with NVIDIA Geforce 620 GPU, but I can reproduce bug 	1194358 on multiple machines with NVIDIA GPU.

Comment 6

[u279076]

(In reply to Paul Silaghi, QA [:pauly] from comment #5)
> I'm not sure how to do that. In System/Additional Drivers I see "No
> proprietary drivers are in use on this system". 

I thought you were using the proprietary driver based on the metadata in your crash report. Please attach a copy of about:support to this bug report so we can confirm your system details.

> 
> > It'd also be useful to know if you can reproduce this in any other
> > pre-release Firefox version and on any other machines. Thanks.
> I can reproduce this on FX 46b1.

Sorry, what I meant was can you reproduce this on Firefox 47 or 48. 

> I can reproduce this only on a single machine with NVIDIA Geforce 620 GPU

Is this a discrete card? If so, can you move it to another machine and test?

Comment 7

[Paul Silaghi, QA [:pauly]]

I'll get back with the details as soon as we're done with the 46 RC and ESRs builds.

Comment 8

[Jean-Yves Avenard [:jya]]

I don't see any references to gfx nor a link with the nvidia or nouveau driver at all. How did the mention of those come about?

That's a very bizarre crash report. Is this with an official build ?

The crash occurs when attempting to register the VP8 decoder, and is done via the macro REGISTER_DECODER

which is:
#define REGISTER_DECODER(X, x)                                          \
    {                                                                   \
        extern AVCodec ff_##x##_decoder;                                \
        if (CONFIG_##X##_DECODER)                                       \
            avcodec_register(&ff_##x##_decoder);                        \
    }

so this call avcodec_register with &ff_vp8_decoder (ff_vp8_decoder is a structure)

which should normally call ff_vp8_decode_init but here it calls ff_vp8_decode_free

It sounds like the compilation was done with a header from a different version of ffmpeg.

Where does this build come from ?

Comment 9

[Jean-Yves Avenard [:jya]]

the alternative of course is that there's a memory corruption happening somewhere

Comment 10

[Paul Silaghi, QA [:pauly]]

(In reply to Anthony Hughes (:ashughes) [GFX][QA][Mentor] from comment #6)
> Please attach a copy of about:support to this bug report
> so we can confirm your system details.
attached
 
> Sorry, what I meant was can you reproduce this on Firefox 47 or 48. 
Just reproduced on 49.0a1 (2016-04-27).

> Is this a discrete card? If so, can you move it to another machine and test?
Yes, it's a dedicated card. I'm going to need IT Dep's approval for that, so let's see if this is really necessary.


(In reply to Jean-Yves Avenard [:jya] from comment #8)
> I don't see any references to gfx nor a link with the nvidia or nouveau
> driver at all. How did the mention of those come about?
I can reproduce this only on a single machine with NVIDIA Geforce 620 GPU. Also, bug 1194358 is NVIDIA specific too.

> That's a very bizarre crash report. Is this with an official build ?
Yes, the official release 46.

Comment 11

[Paul Silaghi, QA [:pauly]]

Attachment: about:support.txt

Comment 12

[u279076]

(In reply to Paul Silaghi, QA [:pauly] from comment #11)
> Created attachment 8746027 [details]
> about:support.txt

This shows "Driver Version: 4.3.0 NVIDIA 319.32" which would seem to indicate you're using the proprietary driver, not the open source driver. Can you please double check?

(In reply to Paul Silaghi, QA [:pauly] from comment #10)
> > Is this a discrete card? If so, can you move it to another machine and test?
> Yes, it's a dedicated card. I'm going to need IT Dep's approval for that, so
> let's see if this is really necessary.

It is necessary to eliminate your system as a variable.

Comment 13

[Paul Silaghi, QA [:pauly]]

I switched the video cards with another machine. I couldn't reproduced the issue on the second machine (having now the NVIDIA Geforce 620 GPU). Instead, I reproduced the issue on the old system having now an AMD Radeon R7 200 GPU. I guess this makes the crash to be system specific, and unrelated to bug 1194358.
https://crash-stats.mozilla.com/report/index/4b6c5eb2-d2cf-4c7a-93c8-a5e442160506

Comment 14

[Paul Silaghi, QA [:pauly]]

Also couldn't reproduced the crash on another machine with Ubuntu 12.04. All the tests were made on FF 46.0.1.

Comment 15

[Anthony Jones (:ajones, :kentuckyfriedtakahe, :k17e)]

Mass change P2 -> P3

Comment 16

[BugBot [:suhaib / :marco/ :calixte]]

Since the crash volume is low (less than 5 per week), the severity is downgraded to S3. Feel free to change it back if you think the bug is still critical.

For more information, please visit auto_nag documentation.

Comment 17

[BugBot [:suhaib / :marco/ :calixte]]

Closing because no crashes reported for 12 weeks.