1774150 - Firefox hangs at nsBaseWidget::CreateCompositorSession()

Status: RESOLVED WORKSFORME

(Core :: Graphics: WebRender, defect)

Description

[Marco Castelluccio [:marco]]

This is not so easy to debug.
Since this morning, Firefox is hanging at startup with one of my profiles (it is not hanging with another profile I have).
Using mozregression, I've noticed builds from a few days ago fail too, which rules out a Firefox change. Using very old builds, I have found the hang to start since https://hg.mozilla.org/integration/autoland/rev/96732a2f2846c158ac1c1dc9f1f2f9b1d676df83.

This means something changed on my system that caused Firefox to start failing, and only with this profile.
Here's the list of packages I upgraded yesterday:

2022-06-14 01:53:24 upgrade clang-15:amd64 1:15~++20220609101129+f8144700ebc9-1~exp1~20220609221215.550 1:15~++20220613111440+aebe24a856d2-1~exp1~20220613111528.557
2022-06-14 01:53:24 upgrade libclang-common-15-dev:amd64 1:15~++20220609101129+f8144700ebc9-1~exp1~20220609221215.550 1:15~++20220613111440+aebe24a856d2-1~exp1~20220613111528.557
2022-06-14 01:53:27 upgrade python3-lldb-15:amd64 1:15~++20220609101129+f8144700ebc9-1~exp1~20220609221215.550 1:15~++20220613111440+aebe24a856d2-1~exp1~20220613111528.557
2022-06-14 01:53:27 upgrade liblldb-15:amd64 1:15~++20220609101129+f8144700ebc9-1~exp1~20220609221215.550 1:15~++20220613111440+aebe24a856d2-1~exp1~20220613111528.557
2022-06-14 01:53:28 upgrade lldb-15:amd64 1:15~++20220609101129+f8144700ebc9-1~exp1~20220609221215.550 1:15~++20220613111440+aebe24a856d2-1~exp1~20220613111528.557
2022-06-14 01:53:28 upgrade llvm-15-dev:amd64 1:15~++20220609101129+f8144700ebc9-1~exp1~20220609221215.550 1:15~++20220613111440+aebe24a856d2-1~exp1~20220613111528.557
2022-06-14 01:53:37 upgrade libclang-cpp15:amd64 1:15~++20220609101129+f8144700ebc9-1~exp1~20220609221215.550 1:15~++20220613111440+aebe24a856d2-1~exp1~20220613111528.557
2022-06-14 01:53:39 upgrade libllvm15:amd64 1:15~++20220609101129+f8144700ebc9-1~exp1~20220609221215.550 1:15~++20220613111440+aebe24a856d2-1~exp1~20220613111528.557
2022-06-14 01:53:43 upgrade llvm-15:amd64 1:15~++20220609101129+f8144700ebc9-1~exp1~20220609221215.550 1:15~++20220613111440+aebe24a856d2-1~exp1~20220613111528.557
2022-06-14 01:53:45 upgrade llvm-15-runtime:amd64 1:15~++20220609101129+f8144700ebc9-1~exp1~20220609221215.550 1:15~++20220613111440+aebe24a856d2-1~exp1~20220613111528.557
2022-06-14 01:53:45 upgrade llvm-15-linker-tools:amd64 1:15~++20220609101129+f8144700ebc9-1~exp1~20220609221215.550 1:15~++20220613111440+aebe24a856d2-1~exp1~20220613111528.557
2022-06-14 01:53:46 upgrade llvm-15-tools:amd64 1:15~++20220609101129+f8144700ebc9-1~exp1~20220609221215.550 1:15~++20220613111440+aebe24a856d2-1~exp1~20220613111528.557
2022-06-14 01:53:46 upgrade libclang1-15:amd64 1:15~++20220609101129+f8144700ebc9-1~exp1~20220609221215.550 1:15~++20220613111440+aebe24a856d2-1~exp1~20220613111528.557
2022-06-14 01:53:47 upgrade libglib2.0-dev:amd64 2.72.1-1 2.72.2-2
2022-06-14 01:53:48 upgrade libglib2.0-dev-bin:amd64 2.72.1-1 2.72.2-2
2022-06-14 01:53:49 upgrade libglib2.0-bin:amd64 2.72.1-1 2.72.2-2
2022-06-14 01:53:49 upgrade libglib2.0-0:amd64 2.72.1-1 2.72.2-2
2022-06-14 01:53:49 upgrade libglib2.0-0:i386 2.72.1-1 2.72.2-2
2022-06-14 01:53:50 upgrade libglib2.0-data:all 2.72.1-1 2.72.2-2
2022-06-14 01:53:51 upgrade eog:amd64 42.1-1 42.2-1
2022-06-14 01:53:52 upgrade gdal-data:all 3.4.3+dfsg-1 3.5.0+dfsg-1
2022-06-14 01:53:52 upgrade gnome-maps:amd64 42.1-1 42.2-1
2022-06-14 01:53:53 upgrade libnftnl11:amd64 1.2.1-1 1.2.2-1
2022-06-14 01:53:53 upgrade libsdl2-2.0-0:amd64 2.0.22+dfsg-4 2.0.22+dfsg-5
2022-06-14 01:53:53 upgrade libsdl2-2.0-0:i386 2.0.22+dfsg-4 2.0.22+dfsg-5
2022-06-14 01:53:54 upgrade linux-compiler-gcc-11-x86:amd64 5.17.3-1 5.18.2-1
2022-06-14 01:54:00 upgrade linux-headers-amd64:amd64 5.17.3-1 5.18.2-1
2022-06-14 01:54:10 upgrade linux-image-amd64:amd64 5.17.3-1 5.18.2-1
2022-06-14 01:54:10 upgrade linux-libc-dev:amd64 5.17.3-1 5.18.2-1
2022-06-14 01:54:11 upgrade network-manager-gnome:amd64 1.26.0-1 1.28.0-1

Which makes me think this is due to the recent kernel upgrade.
It must also be due to something specific in my profile, given that I can't reproduce in a clean profile or other "dirty" profiles I have.

Comment 1

[Marco Castelluccio [:marco]]

I tried to reboot with the older kernel, 5.17.3-1, and Firefox is not crashing.

Note: it's actually a hang, which forces me to kill Firefox.

Comment 2

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

Set release status flags based on info from the regressing bug 1749174

Comment 3

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

:emilio, since you are the author of the regressor, bug 1749174, could you take a look?
For more information, please visit auto_nag documentation.

Comment 4

[Emilio Cobos Álvarez [:emilio]]

Do you have a stack trace or something that could help diagnose this?

Comment 5

[Robert Mader [:rmader]]

And can you attach your about:support?

Comment 6

[Marco Castelluccio [:marco]]

Here's a forced crash report: https://crash-stats.mozilla.org/report/index/a97bbeaf-01cf-4ba1-9a94-136380220615.

Comment 7

[Marco Castelluccio [:marco]]

Attachment: about_support.txt

And here are the contents of my about:support.

Comment 8

[Emilio Cobos Álvarez [:emilio]]

Interesting, that's exactly the same crash stack as bug 1772691.

Comment 9

[Robert Mader [:rmader]]

Fedora 36 + testing just updated to 5.18 (5.18.4-201.fc36.x86_64) as well but I can't reproduce this hang.

Comment 10

[Martin Stránský [:stransky] (ni? me)]

This is a hang in NVIDIA binary drivers - see Thread 26, Name: Renderer.

This is long term issue that all calls to Renderer thread are blocking so a bug/delay in drivers causes whole Firefox hang - it's the nsBaseWidget::CreateCompositorSession() or FlushRendering and similar.

Would be great to have it non-blocking and fallback to SW rendering when there's no response from GL after some time.

Comment 11

[Marco Castelluccio [:marco]]

I can no longer reproduce this, probably because I upgraded to 5.18.5-1.

Comment 12

[Marco Castelluccio [:marco]]

The previous version was 5.18.2-1.

Comment 13

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

Set release status flags based on info from the regressing bug 1749174