Crash due to fatal assert in winmm: cubeb_refill_stream/winmm_refill_stream

RESOLVED INCOMPLETE

Status

()

--
critical
RESOLVED INCOMPLETE
6 years ago
5 years ago

People

(Reporter: philipp, Assigned: kinetik)

Tracking

({crash, regression})

27 Branch
x86
Windows 7
crash, regression
Points:
---

Firefox Tracking Flags

(firefox26 unaffected, firefox27 affected, firefox28 affected, firefox29 affected)

Details

(crash signature)

(Reporter)

Description

6 years ago
a user in the support forum reported that firefox repeatedly crashes with the above signature whenever he receives an incoming facebook message, even in safemode. 
forum thread: https://support.mozilla.org/en-US/questions/937048
bug report: https://crash-stats.mozilla.com/report/index/e6a13958-d2f3-457b-a2aa-4d7f12120913

according to the crash stats this signature first appeared with firefox 15 and  some comments seem to be link it to playing media in the browser (maybe also a audio notification in facebook is triggering it in the reported case).
Severity: normal → critical
Crash Signature: [@ abort | abort | _wassert
Component: Untriaged → Video/Audio
Keywords: crash
Product: Firefox → Core
(Assignee)

Updated

6 years ago
Blocks: 623444
Status: UNCONFIRMED → NEW
Crash Signature: [@ abort | abort | _wassert → [@ abort | abort | _wassert | cubeb_refill_stream ]
Ever confirmed: true
Keywords: regression
Summary: crash @abort | abort | _wassert → Crash due to fatal assert in cubeb_refill_stream [@ abort | abort | _wassert | cubeb_refill_stream ]
(Assignee)

Comment 1

6 years ago
Some of the reports are hitting a similar assert in cubeb_get_next_buffer.

Updated

6 years ago
Crash Signature: [@ abort | abort | _wassert | cubeb_refill_stream ] → [@ abort | abort | _wassert]
Summary: Crash due to fatal assert in cubeb_refill_stream [@ abort | abort | _wassert | cubeb_refill_stream ] → Crash due to fatal assert in cubeb_refill_stream
(Assignee)

Updated

6 years ago
See Also: → bug 833161
Roc, can you have someone take a look a this? 

The [@ abort | abort | _wassert] signature has returned to #4 top crasher on Fx 27, which is about to go to beta.  Some recent comments point to Battlefield 4; some sort of interaction with "battlelog"
status-firefox27: --- → affected
tracking-firefox27: --- → ?
Flags: needinfo?(roc)
Version: 15 Branch → 27 Branch
Matthew Gregan is already on the bug.
Flags: needinfo?(roc)

Comment 4

5 years ago
roc, are you talking about bug 833161? That appears to be a linux-only bug, and this is Windows, so it's not clear to me that it is the same.

cpearce the assert location is well-known in this case: http://hg.mozilla.org/releases/mozilla-aurora/annotate/731470a7c4ca/media/libcubeb/src/cubeb_wasapi.cpp#l665

See e.g. the stack at https://crash-stats.mozilla.com/report/index/83dcae38-7539-41ea-a342-c10902131204 or https://crash-stats.mozilla.com/report/index/a4999c2b-0cca-4f0d-ab1f-7ffe02131205
Assignee: nobody → cpearce
So, this is not supposed to be reached, as per the comment, and MSDN. I can certainly write a patch if nobody has one.
(Assignee)

Comment 6

5 years ago
Crashes in cubeb_wasapi need a different bug, this one has always been about cubeb_winmm code.
tracking-firefox27: ? → +
Keywords: topcrash
(In reply to Paul Adenot (:padenot) from comment #5)
> So, this is not supposed to be reached, as per the comment, and MSDN. I can
> certainly write a patch if nobody has one.

You mean just take out the assertion? Yes please :-)
Well, MSDN has a long history of lying to us. Also, some driver also like to return funny stuff for whatever reason, and I'd rather use the software fallback we wrote than crashing on the assertion.

Updated

5 years ago
Assignee: cpearce → paul

Comment 9

5 years ago
(topcrash is deprecated)
Keywords: topcrash → topcrash-win
I think I took the bug by mistake.
Assignee: paul → kinetik
https://crash-stats.mozilla.com/report/index/6cf0606b-9f53-459e-acd1-3e1d32131219

From a friend: "Each time my match ends (Battlefield 4) I find that the browser has crashed". Note that Battlefield 4 uses browser plugins: ESN Sonar API & Battlelog Game Launcher.
status-firefox28: --- → affected
He's using this soundcard: "Creative Sound Blaster X-Fi Titanium Fatal1ty Champion", if that's relevant.
7 Day Ranking:
> Firefox 26: N/A
> Firefox 27: #23 @ 0.43% (-7 positions)
> Firefox 28: #17 @ 0.94% (new)
> Firefox 29: #27 @ 0.43% (-5 positions)

Given current rankings and the downward trend this is no longer a topcrash.
status-firefox26: --- → unaffected
status-firefox29: --- → affected
Keywords: topcrash-win
(Assignee)

Comment 14

5 years ago
gcp and ashughes, that's covered by bug 950026 for the WASAPI backend.  This bug is about the WinMM backend, which is much less commonly used now.
Summary: Crash due to fatal assert in cubeb_refill_stream → Crash due to fatal assert in winmm: cubeb_refill_stream/winmm_refill_stream
Is it possible to divert those away from here via changing the crash signature in the bug? If you hit the WASAPI case you'll get pointed to this bug.
(In reply to Gian-Carlo Pascutto (:gcp) from comment #15)
> Is it possible to divert those away from here via changing the crash
> signature in the bug? If you hit the WASAPI case you'll get pointed to this
> bug.

Yes, just make sure the Crash Signature field in each bug contains only the signatures(s) you want mapped to the bug.

Comment 17

5 years ago
I got a few crashes that looks related to this bug:
https://crash-stats.mozilla.com/report/index/bd4c58a9-bdd8-4415-aa68-3d02c2131225
https://crash-stats.mozilla.com/report/index/d8d4ef11-7453-48fc-8c6d-097442131225

The crash came in while using Outlook Web Mail interface, a popup came in from a Calendar event and it seems to have crashed while trying to play the sound clip.
NI on :kinetik to get some help with next steps here. Looks like this is no longer a top-crasher, but if its a reproducible crash regression, can we try fixing it ?
Flags: needinfo?(kinetik)
(Assignee)

Comment 19

5 years ago
(In reply to Minh Tran-Le from comment #17)
> I got a few crashes that looks related to this bug:
> https://crash-stats.mozilla.com/report/index/bd4c58a9-bdd8-4415-aa68-
> 3d02c2131225
> https://crash-stats.mozilla.com/report/index/d8d4ef11-7453-48fc-8c6d-
> 097442131225

These are bug 950026.

(In reply to bhavana bajaj [:bajaj] from comment #18)
> NI on :kinetik to get some help with next steps here. Looks like this is no
> longer a top-crasher, but if its a reproducible crash regression, can we try
> fixing it ?

I don't know how to reproduce it unfortunately.
Flags: needinfo?(kinetik)

Comment 20

5 years ago
(In reply to Gian-Carlo Pascutto (:gcp) from comment #12)
> He's using this soundcard: "Creative Sound Blaster X-Fi Titanium Fatal1ty
> Champion", if that's relevant.

I have the same Sound card, or almost, i dont think it's a "Champion", and i've been experiencing crashes randomly because of that bug, and i dont even play BF4, and never receive messages on Facebook.
abort | abort | _wassert has dropped to #46 on 27beta crash list. could the fix uplifted in bug 950026 be responsible?

Also, I was looking for some sort of audio driver correlation to attempt reproduding this crash. But the information isn't available, from what I can gather.
(In reply to [:tracy] Tracy Walker - QA Mentor from comment #21)
> abort | abort | _wassert has dropped to #46 on 27beta crash list. could the
> fix uplifted in bug 950026 be responsible?

Paul can you confirm if your patch may have helped here or any other input here ?
> 
> Also, I was looking for some sort of audio driver correlation to attempt
> reproduding this crash. But the information isn't available, from what I can
> gather.
Flags: needinfo?(padenot)
Bhavana, yes, it's plausible.
Flags: needinfo?(paul)

Updated

5 years ago
tracking-firefox27: + → ---
Clearing tracking given this is no longer a top-crash(main reason we tracked) and plausibly already fixed.

Comment 25

5 years ago
Issue is confirmed NOT resolved with Creative Soundblaster and Realtek hardware when vendor specific enhancements are enabled.

Crash can be reproduced 100% until "Disable [Product name] Enhancements" is ticked in Speaker device properties

I expect the enhancement modules provided with Recon and Z(x/R) series soundblaster cards will exhibit the same issue.

However, whilst i can not trigger the crash with facebook message tones or youtube, i can instantly triggr it using vine and tumblr players.

https://crash-stats.mozilla.com/report/index/84243c37-eb98-44ea-9278-097f32140208
https://crash-stats.mozilla.com/report/index/da35a63d-c06c-4206-91a5-be4b12140208
https://crash-stats.mozilla.com/report/index/d83e9d96-e411-43c5-916d-d87d32140208
https://crash-stats.mozilla.com/report/index/03601244-ce1c-4f0e-ad45-ee1f22140208
https://crash-stats.mozilla.com/report/index/5b76fb4f-4ed0-4827-a934-005412140208
https://crash-stats.mozilla.com/report/index/ac8e8f41-020c-49db-8182-b2c6e2140208
https://crash-stats.mozilla.com/report/index/157893f1-dcef-4c93-aedb-31bf32140208
https://crash-stats.mozilla.com/report/index/6810e793-a3b8-4095-9678-fd0e22140208

FTR, i have disabled all creative enhancements inside the creative software, it is not till "Disable Soundblaster Enhancements" is checked that the aforementioned video content players back without issue.
(Assignee)

Comment 26

5 years ago
(In reply to Danial Horton from comment #25)
> Issue is confirmed NOT resolved with Creative Soundblaster and Realtek
> hardware when vendor specific enhancements are enabled.
> 
> Crash can be reproduced 100% until "Disable [Product name] Enhancements" is
> ticked in Speaker device properties

Thanks for the report.  There are all crashes in the WASAPI code, this bug is about the WinMM code, so it's not this bug.  Probably related to bug 950026, which you also posted in.

Comment 27

5 years ago
(In reply to Danial Horton from comment #25)
> Issue is confirmed NOT resolved with Creative Soundblaster and Realtek
> hardware when vendor specific enhancements are enabled.
> 
> Crash can be reproduced 100% until "Disable [Product name] Enhancements" is
> ticked in Speaker device properties
> 
> I expect the enhancement modules provided with Recon and Z(x/R) series
> soundblaster cards will exhibit the same issue.
> 
> However, whilst i can not trigger the crash with facebook message tones or
> youtube, i can instantly triggr it using vine and tumblr players.
> 
> https://crash-stats.mozilla.com/report/index/84243c37-eb98-44ea-9278-
> 097f32140208
> https://crash-stats.mozilla.com/report/index/da35a63d-c06c-4206-91a5-
> be4b12140208
> https://crash-stats.mozilla.com/report/index/d83e9d96-e411-43c5-916d-
> d87d32140208
> https://crash-stats.mozilla.com/report/index/03601244-ce1c-4f0e-ad45-
> ee1f22140208
> https://crash-stats.mozilla.com/report/index/5b76fb4f-4ed0-4827-a934-
> 005412140208
> https://crash-stats.mozilla.com/report/index/ac8e8f41-020c-49db-8182-
> b2c6e2140208
> https://crash-stats.mozilla.com/report/index/157893f1-dcef-4c93-aedb-
> 31bf32140208
> https://crash-stats.mozilla.com/report/index/6810e793-a3b8-4095-9678-
> fd0e22140208
> 
> FTR, i have disabled all creative enhancements inside the creative software,
> it is not till "Disable Soundblaster Enhancements" is checked that the
> aforementioned video content players back without issue.

This means these 2 reports are related:
https://crash-stats.mozilla.com/report/index/86e4d36e-8add-41c6-8cab-bdfd52140208 (release)
https://crash-stats.mozilla.com/report/index/10682e1e-301f-4295-b674-307742140208 (nightly x64)

To trigger the crash with youtube, you have to activate html5 and try to watch an html5 compatible video, else, it will be displayed using flash.

Comment 28

5 years ago
I haven't managed to reproduce a crash using the youtube html5 player

That said, I have not come across any mono aac youtube video's so that could make up the difference - All Vine uploads are mono aac.

Bug 970147 is tracking this crash in handle_channel_layout

Updated

5 years ago
Duplicate of this bug: 970519

Comment 30

5 years ago
First of all, being a newbie with reporting bugs here, I opened bug# 970159 only to find out this one here. Please find comments inline.

(In reply to [:tracy] Tracy Walker - QA Mentor from comment #2)
> Some recent comments point to
> Battlefield 4; some sort of interaction with "battlelog"

(In reply to Gian-Carlo Pascutto (:gcp) from comment #11)
> https://crash-stats.mozilla.com/report/index/6cf0606b-9f53-459e-acd1-
> 3e1d32131219
> 
> From a friend: "Each time my match ends (Battlefield 4) I find that the
> browser has crashed". Note that Battlefield 4 uses browser plugins: ESN
> Sonar API & Battlelog Game Launcher.

I experience the exact same issue. Details:
* Windows 8 64-bit Greek
* Battlefield 4 playing 64-bit
* had no issue at all up to and including FF 26
* issue appeared at 27
* FF 27 crash is 100% reproducible; just wait for the game round to end. 

(In reply to Gian-Carlo Pascutto (:gcp) from comment #12)
> He's using this soundcard: "Creative Sound Blaster X-Fi Titanium Fatal1ty
> Champion", if that's relevant.

Found this comment rather related. An exact same card was installed, but was deinstalled since there were some issues with its Windows 8 driver. A Soundblaster Z took its place. But, like I said, had no issues at all till FF 26.

I elected to mark bug# 970159 as a duplicate of this one. If you need more crash reports, I can easily produce them. FYI, some crash dumps:

* https://crash-stats.mozilla.com/report/index/bp-2c543455-47a1-4e4c-afc8-a32012140210
* https://crash-stats.mozilla.com/report/index/bp-1055813e-df07-4c68-868c-e21492140210
* https://crash-stats.mozilla.com/report/index/bp-f4993f10-20fa-4fc9-9fc8-099902140210
* https://crash-stats.mozilla.com/report/index/bp-1f0e0c4a-757a-4764-8f27-aa8ea2140210

Comment 31

5 years ago
Michail, as pointed out in your bug, 

Bug 970147 is tracking this crash in handle_channel_layout

those crash reports are 100% the same as the ones i provided.

Comment 32

5 years ago
But thank you, you have confirmed that the Soundblaster Enhancements is also producing the same results on Sounbblaster Z(x/xr) and Recon cards.

Comment 33

5 years ago
(In reply to Danial Horton from comment #32)
> But thank you, you have confirmed that the Soundblaster Enhancements is also
> producing the same results on Sounbblaster Z(x/xr) and Recon cards.

You are welcome, and apologies to the thread posters here for posting in the wrong report.
So it seems like *this* crash was resolved in the common case but there are still scenarios where it occurs (as pointed out by Danial and Michail). Is there more we can do in *this* bug report or should work be branched off to their reports? If more work needs to happen, do devs need more info to proceed?
(Assignee)

Comment 35

5 years ago
(In reply to Anthony Hughes, QA Mentor (:ashughes) from comment #34)
> So it seems like *this* crash was resolved in the common case but there are
> still scenarios where it occurs (as pointed out by Danial and Michail). Is
> there more we can do in *this* bug report or should work be branched off to
> their reports? If more work needs to happen, do devs need more info to
> proceed?

See comment 6, comment 14, comment 26, etc..  This bug is tracking a assert failure in a different audio backend.
> See comment 6, comment 14, comment 26, etc..  This bug is tracking a assert
> failure in a different audio backend.

Thanks Matthew. So is there anything more you need to investigate a solution to these winmm crashes?
(Assignee)

Comment 37

5 years ago
Yes, unfortunately I don't have way to reproduce the bug and there's not enough information in the initial crash reports to fix the bug.
All the reports linked in this bug seem to correlate to wasapi, which you say is not *this* bug. If there's a report I'm missing please point it out. We'll need to correlate those crashes to some hardware/software configs if we have any hope of attempting to reproduce this.
(Assignee)

Comment 39

5 years ago
The initial crash report in the opening comment (which has timed out) is all we have.  This code only runs on WinXP as of Firefox 25 or so (possibly a bit later, the WASAPI backend was disabled in a beta if I remember correctly), so the number of machines running this code is significantly lower than when the bug was filed.  Also, it's possible that the bug only manifests on Windows Vista up and thus would no longer be triggered (no reason to believe this, just mentioning as a possibility).  I can't see any recent stacks with "winmm" in them that correlate to this bug.

Since we have so little to go on and to avoid continuing confusion with WASAPI bugs, maybe we're better off closing this as WFM/INCOMPLETE for now and opening a new bug if it turns up again.
(In reply to Matthew Gregan [:kinetik] from comment #39)
> maybe we're better off closing this as WFM/INCOMPLETE for now
> and opening a new bug if it turns up again.

I tend to agree and would opt for INCOMPLETE given the circumstances. Release Management, what do you think?
Flags: needinfo?(release-mgmt)

Comment 41

5 years ago
Waaait a second, i am still crashing, more precisely, using 4chanx, when i activate the *beep* sound when someone posts an answer on the thread. I have the sound card mentionned on this page.

https://crash-stats.mozilla.com/report/index/732e8473-8886-4048-9209-cb7552140209
Hi Rag,

Your crash report correlates to WASAPI which is not what *this* bug report is about. This one is specifically out WINMM. Matthew can point you to the correct bug.

Comment 44

5 years ago
Are these two also related?
Both were present when "Disable Soundblaster Enhancements" were not ticked.
(With Creative Sound Blaster X-Fi)

(In reply to Albirew from comment #27)
> https://crash-stats.mozilla.com/report/index/86e4d36e-8add-41c6-8cab-
> bdfd52140208 (release)
> https://crash-stats.mozilla.com/report/index/10682e1e-301f-4295-b674-
> 307742140208 (nightly x64)

Comment 45

5 years ago
1st is bug 970147

second, im not sure. doesn't correspond to this or bug 970147

Comment 46

5 years ago
second is from x64 nightly but have same effect / resolution than the other (so Creative Sound Blaster related)
(Assignee)

Comment 47

5 years ago
(In reply to Albirew from comment #44)
> Are these two also related?

You'd have to be running WinXP to hit the problem this bug is about, as all newer versions of Windows will use the WASAPI backend.  Those two crashes are in WASAPI and are likely both bug 970147 (the first certainly is, the second doesn't have symbols resolved but looks the same).  Please try the latest nightly build and see if it still crashes, if not bug 970147 fixed the crash but you now might see bug 971693 (no sound for those videos) instead.

Comment 48

5 years ago
(In reply to Matthew Gregan [:kinetik] from comment #47)
> You'd have to be running WinXP to hit the problem this bug is about, as all
> newer versions of Windows will use the WASAPI backend.  Those two crashes
> are in WASAPI and are likely both bug 970147 (the first certainly is, the
> second doesn't have symbols resolved but looks the same).  Please try the
> latest nightly build and see if it still crashes, if not bug 970147 fixed
> the crash but you now might see bug 971693 (no sound for those videos)
> instead.

Still crashing with latest nightly:
https://crash-stats.mozilla.com/report/index/c91cd530-f8d9-4277-bd3f-c75282140212
bit it might be bug 958080 instead.
(Assignee)

Comment 49

5 years ago
Yes, it is bug 958080.
Looks like this is an INCOMPLETE, yes.
Flags: needinfo?(release-mgmt)
(Assignee)

Comment 51

5 years ago
Closing as INCOMPLETE as we don't have enough information to fix the bug and most Windows systems (Vista upwards) are now using a different audio backend (WASAPI).
Status: NEW → RESOLVED
Last Resolved: 5 years ago
Resolution: --- → INCOMPLETE
You need to log in before you can comment on or make changes to this bug.