Bug 28652

Created attachment 12527 [details]
Screenshot of web page crashed

Can't confirm this behaviour. Did you try Firefox Safe mode? https://support.mozilla.org/en-US/kb/troubleshoot-firefox-crashes-closing-or-quitting?

CC: (none) => sturm-fr

The crashes also occur on safe mode. But I think it might be my system that is in a bad state. In Chromium I cannot see youtube videos (don't play) and some pages look weird. I feel most like reinstalling and retest.

I've reinstalled mageia and switched to Cauldron. Still same problem

URL: https://www.washingtonpost.com/ https://danskebank.dk/privat https://www.dr.dk/ => https://www.washingtonpost.com/ https://danskebank.dk/privat https://www.dr.dk/ youtube.com

> I've reinstalled mageia and switched to Cauldron. Still same problem

Did you try if 78.9.0esr from mga8 testing work for you before switching to Cauldron?

FWIW no problem here on *mga7*-64 Firefox 78.9.0esr from testing

CC: (none) => fri

Hi

after upgrading to latest cauldron, I confirm I see the same problem ; some sites crash and firefox display "the tab has crashed"

this is using firefox-78.9.0-1.mga9 64 bit rpm.

I didn't have such issue with the firefox version that ships in mga8

CC: (none) => npomarede

I didn't check 78.9.0esr in mga8 testing before switching to Cauldron. Sorry.

I'm seeing the same issue.  It seems to be limited to pages with media like audio (deezer.com) or video.

- tried Safe Mode.  Tab crashes
- "Refreshed" Firefox.  Tab crashes
- created a new profile (firefox -P).  Tab crashes
- enabled webrender in the config.  Still crashes.

(Aside: Chromium is crashing for me as well, but I don't think it's the same issue)

Thank you all for your contributions.

Assigning this globally as Firefox is done by various people.

Assignee: bugsquad => pkg-bugs
Summary: Firefox crashes on certain webpages => Firefox 78.9.0-1.mga9 crashes on certain webpages (multiple reports)

While trying to produce a crash report I discovered that if I disable multi-process Firefox (Electrolysis E10S) with an environment variable, the tabs no longer crash:


~/tmp$ cat firefox-single.sh 
#!/bin/sh

export MOZ_FORCE_DISABLE_E10S=1
cd /usr/lib64/firefox
./firefox


I realize this is a workaround and not the solution.

I see this too.  If you run FF from the command line, you get 

###!!! [Parent][MessageChannel] Error: (msgtype=0x5B001C,name=PHttpChannel::Msg_DeleteSelf) Channel error: cannot send/recv


###!!! [Parent][MessageChannel] Error: (msgtype=0x5B001C,name=PHttpChannel::Msg_DeleteSelf) Channel error: cannot send/recv


###!!! [Parent][MessageChannel] Error: (msgtype=0x5B001C,name=PHttpChannel::Msg_DeleteSelf) Channel error: cannot send/recv

when the tab crashes.

CC: (none) => ftg

And I confirm that the workaround works.

*** Bug 28721 has been marked as a duplicate of this bug. ***

CC: (none) => omeritzicschwartz

Hi

additional infos : I noticed that each time the tab crashes, I get a message in syslog.

Apr  6 22:05:09 muon kernel: [ 1787.911805] MediaDe~hine #1[94962]: segfault at 0 ip 00007f2b43d7d4b9 sp 00007f2b24448920 error 4 in libmozsandbox.so[7f2b43d73000+16000]
Apr  6 22:05:09 muon kernel: [ 1787.911825] Code: 24 88 01 00 00 31 c0 4c 8b 47 20 48 8b 47 28 83 fe 9c 0f 84 c1 00 00 00 f6 c4 10 0f 85 88 00 00 00 4d 85 d2 0f 84 b8 00 00 00 <41> 80 3b 2f 0f 85 bd 00 00 00 89 c2 81 e2 ff f6 ff ff 0f 85 20 01

Maybe this can help narrowing the issue

Also, there's a report for icecat here https://bugzilla.redhat.com/show_bug.cgi?id=1945720 that really looks similar to this issue.

Maybe they will have a fix that works here too.

I receive the exact same sequence of numbers in the log ("Code: 24 88 ...").  The workaround in the icecat bug works for me as well (no tabs crashing).

MOZ_DISABLE_CONTENT_SANDBOX=1 firefox

I've spent the morning digging into this and have gotten coredumps and such, and submitted an upstream bug report: https://bugzilla.mozilla.org/show_bug.cgi?id=1704374

CC: (none) => ngompa13

Maybe graphics/processor related?  FWIW, on AMD Ryzen 5 3400G with Radeon Vega Graphics, no GPU.
[rolf@x570i immigration]$ rpm -qa | grep firefox
firefox-78.9.0-1.mga8
firefox-en_GB-78.9.0-1.mga8
firefox-en_US-78.9.0-1.mga8
[rolf@x570i immigration]$

I've got all the suspect urls from this BR I can glean open at once in separate tabs:

https://danskebank.dk/privat
https://www.washingtonpost.com/?reload=true&_=1618160795614
https://www.dr.dk/  ---from Hans's comment #0

https://www.deezer.com/us/  ---from Curtis's comment #8

https://test.webrtc.org/
https://www.youtube.com/watch?v=AgmMuHRMRgM  ---from Neal's upstream bug report linked in comment #16

At each site, I only (seemingly) enabled/acknowledged cookies (no sprechen Deutsche), and, at https://test.webrtc.org/ I disabled the web cam since I don't have one and microphone, that got taken out with that module.

Long story short, no tab crash in that firefox instance for maybe 25 minutes.

Thanks.

CC: (none) => rolfpedersen

(In reply to Rolf Pedersen from comment #17)
>
> At each site, I only (seemingly) enabled/acknowledged cookies (no sprechen
> Deutsche), and, at https://test.webrtc.org/ I disabled the web cam since I
> don't have one and microphone, that got taken out with that module.
> 

Not enabling the webcam avoids the crash.

(In reply to Neal Gompa from comment #18)
> (In reply to Rolf Pedersen from comment #17)
> >
> > At each site, I only (seemingly) enabled/acknowledged cookies (no sprechen
> > Deutsche), and, at https://test.webrtc.org/ I disabled the web cam since I
> > don't have one and microphone, that got taken out with that module.
> > 
> 
> Not enabling the webcam avoids the crash.

Err, I meant to ask that as a question, not a statement.

Created attachment 12613 [details]
webrtc test result

The observation I attempted to convey was that, with all those triggering urls open in firefox at one time, there was no tab crash in the 20 minutes or so that I tested.

I don't have a webcam connected (not, at least, since the support group planners for our annual Christmas party shot down my idea to hold a clothing-optional zoom meeting).  I have a mic, which webrtc detected, but the site proposed my hauppauge atsc tuner for the web cam.  I didn't want to try with that.  Web cam and mic come together as a pair in that one module, so I disabled the module.

I don't much use firefox, only when seamonkey lacks some feature to work on some website, and, such as that is, I have never seen a tab crash.  Let me load all those urls in firefox, again, this time connecting a usb webcam for webrtc...

I loaded all the urls, ran the webrtc test (results in attached png), played a succession of sample deezer clips, and played the entire 14:16 of some self-important condescending youtube puke, then on into autoplay.  

So, 20 minutes later, no tab crash.

same as comment #15, I now run with MOZ_DISABLE_CONTENT_SANDBOX=1 as a temporary workaround and there was no crash at all since several days.

I just submitted a potential fix for this based on reports from Fedora on IceCat to Cauldron. Let's see how it goes with firefox-78.10.0-2.mga9.

just made some tests with some of the sites that crashed before for me and everything works so fat, no more "tab crash".
thanks

I don't see any crashes either. I have surfed for hours now. No crashes.

At this point, I think the package in testing is ready to be submitted as an update.

Suggested advisory:
========================

Updated firefox packages fix various bugs:

This update fixes crashes in the process sandbox caused by miscompilation with Rust 1.48 or newer.

References:
https://bugs.mageia.org/28652
https://bugzilla.redhat.com/1945720
https://bugzilla.mozilla.org/1704374
https://bugzilla.mozilla.org/1680166
========================

Updated packages in core/updates_testing:
========================
firefox-78.10.0-1.2.mga8
firefox-devel-78.10.0-1.2.mga8


Source RPMs: 
firefox-78.10.0-1.2.mga8.src.rpm

Assignee: pkg-bugs => qa-bugs

I have also been using it for a while without problems.
Mga8-64, Intel i7, nvidia-current, plasma

Hi all, 

This new compilation have the bug 28359 still, and I can't open the Bankinter web login (https://empresas.bankinter.com/secure/es).

Greetings!!

CC: (none) => joselp

(In reply to Neal Gompa from comment #25)
> At this point, I think the package in testing is ready to be submitted as an
> update.
> 
> Suggested advisory:
> ========================
> 
> Updated firefox packages fix various bugs:
> 
> This update fixes crashes in the process sandbox caused by miscompilation
> with Rust 1.48 or newer.
> 
> References:
> https://bugs.mageia.org/28652
> https://bugzilla.redhat.com/1945720
> https://bugzilla.mozilla.org/1704374
> https://bugzilla.mozilla.org/1680166
> ========================
> 
> Updated packages in core/updates_testing:
> ========================
> firefox-78.10.0-1.2.mga8
> firefox-devel-78.10.0-1.2.mga8
> 
> 
> Source RPMs: 
> firefox-78.10.0-1.2.mga8.src.rpm

Fixed in Cauldron.
Switching this to a real Mageia 8 bug.

CC: (none) => ouaurelien
Version: Cauldron => 8
Summary: Firefox 78.9.0-1.mga9 crashes on certain webpages (multiple reports) => Firefox 78.9.0-1.mga8 crashes on certain webpages (multiple reports)
Source RPM: firefox-78.9.0-1.mga9.src.rpm => firefox-78.10.0-1.1.mga8.src.rpm

Remark, as this is still firefox 78.10, there is no
update to firefox-l10n-78.10.0-1.mga8.src.rpm

MGA8-64-OK

Validating.
Advisory committed to SVN.

An update for this issue has been pushed to the Mageia Updates repository.

https://advisories.mageia.org/MGAA-2021-0118.html

Status: NEW => RESOLVED
Resolution: (none) => FIXED