1694792 - Visual corruption on Intel(R) HD Graphics 2500

Status: RESOLVED WORKSFORME

(Core :: Graphics: WebRender, defect)

Description

[Asif Youssuff]

Attachment: 1614195181608.webm

From: https://www.reddit.com/r/firefox/comments/lrko78/horrible_graphics_glitch_in_firefox_with_webrender/

See video and about:support.

Failure Log
(#0) Error: Updating unknown shared surface: 459561500681

Comment 1

[Asif Youssuff]

Attachment: about:support

Comment 2

[Sotaro Ikeda [:sotaro]]

:aosmond, can you comment to the bug?

Comment 3

[Robert Mader [:rmader]]

Asif, does this also happen with Wayland disabled? And with the X11/EGL backend (MOZ_X11_EGL=1)?

Comment 4

[Asif Youssuff]

Robert, the user has deleted their reddit account. I can't follow up here.

Comment 5

[Robert Mader [:rmader]]

Thanks. Closing then as we have no way to verify. Given that Webrender is enabled on Gnome by default, we should get new reports if this is still happening.

Comment 6

[2tt56k2hg]

Hello I am the one who made the reddit post. It also happens on xorg.

Comment 7

[Robert Mader [:rmader]]

(In reply to 2hgrb8kt4 from comment #6)

Hello I am the one who made the reddit post. It also happens on xorg.

Ah perfect (that you're still around, not the bug of course :) ).

Comment 8

[2tt56k2hg]

(In reply to Robert Mader [:rmader] from comment #5)

Thanks. Closing then as we have no way to verify. Given that Webrender is enabled on Gnome by default, we should get new reports if this is still happening.

I would ask that you don't close the bug, this makes firefox unusable for me (and presumably others). I closed the reddit accout because of a change in their tracking policy, but an happy to help here.

thanks
user

Comment 9

[2tt56k2hg]

(In reply to Robert Mader [:rmader] from comment #7)

(In reply to 2hgrb8kt4 from comment #6)

Hello I am the one who made the reddit post. It also happens on xorg.

Ah perfect (that you're still around, not the bug of course :) ).

Oh thank you, I missed this in my panic writing the other comment.

Comment 10

[Robert Mader [:rmader]]

Shot in the dark: does it still happen after bug 1694909 i.e. with latest nightly? If so we should put this hardware generation on a blocklist until it's fixed.

Comment 11

[2tt56k2hg]

It is still happening with nightly, yes. I have change to nightly because of this bug so will be about to post as soon as an update fixes it.

Comment 12

[Robert Mader [:rmader]]

Thanks!

Andrew, can we put Ivybridge on a blocklist for release for the time being? Maybe sandybridge as well if that makes sense.

Comment 13

[2tt56k2hg]

I should say that it only happens with Intel® HD Graphics 2500 graphics and not the Intel® HD Graphics 4000, that other Ivybridge cpus use. I know this as I also have a laptop with an i5-3320M that doesn't suffer from the same problem.

Comment 14

[Robert Mader [:rmader]]

Ah that's good to know. To me this sounds like a driver issue then - mind creating a mesa issue?

Comment 15

[2tt56k2hg]

I also thought this but hardware accelerated video works fine in firefox (as well as gnome web and celluloid, although this may be handled by the libva driver) this issue only affects image elements of web pages and the highlight text (as shown at the beginning of the video), not embedded images themselves (I'm sorry I don't know the correct technical terms). Is this consistent with a mesa driver issue?

Comment 16

[Robert Mader [:rmader]]

(In reply to 2hgrb8kt4 from comment #15)

Is this consistent with a mesa driver issue?

Yes pretty much. On Webrender all/most rendering is done via OpenGL so if there's visual corruption, then there's a high chance that the driver is doing something wrong. The other option would be that Firefox is doing something wrong - but all other drivers handle it somehow. Chances for that are - well not so high. Video decoding / libva is relatively unrelated here.

Comment 17

[2tt56k2hg]

I see. I have download and am trying to use mozregression to confirm. If I can't find a working version of firefox I will try running a distro with old mesa to check that too.

thanks again

Comment 18

[2tt56k2hg]

Ok, I mozregressed all the way back to 2019 and the problem was still present; its not a firefox issue. I thought if it was a driver problem I would have noticed elsewhere on my system and not just firefox. I will try what I said with testing and older driver, but have no clue how to file a bug for mesa.

Thanks for your help Robert and Asif
user

Comment 19

[Robert Mader [:rmader]]

Older mesa may help, but may as well not. Webrender is a quite demanding engine and likely to expose bugs. See e.g. https://gitlab.freedesktop.org/mesa/mesa/-/issues/4227 for a similar recent bug.

Thanks for your help Robert and Asif

Thank you!

Comment 20

[Jim Blandy :jimb]

If this is a driver issue, then it sounds like this bug can be closed. If I've misunderstood, please re-open and explain what we should look into.

Comment 21

[Robert Mader [:rmader]]

Jim, if this bug is present for many users, shouldn't we keep it open and take some kind of action? E.g. either investigating and opening a mesa issue or at least blocklisting affected devices from WR?

Comment 22

[Jim Blandy :jimb]

That's a good point. We should make a decision about a blacklist entry before closing this.

Comment 23

[Jeff Muizelaar [:jrmuizel]]

2hgrb8kt4, are you certain the other Ivybridge systems that don't reproduce this problem are running the same kernel and mesa versions?

Comment 24

[2tt56k2hg]

for the Graphics 4000 laptop the issue isn't present on current arch linux (which I take to have newest stable kernel/driver) and is still present for Graphics 2500 on fedora rawhide (which would get any new driver fix soonest and is on the 5.12 kernel). I was just a few weeks ago running both machines on the same arch kernel/driver and it was the same.

Comment 25

[mehrmor]

I believe I have basically the same issue as the reporter. Here's a short clip:
https://1drv.ms/u/s!AlqFeqoW1KpnhOJQXv_ho1GNt_8J0Q?e=g9aIMi

I didn't experience the issue using the nightly version: 88.0a1.

Running version 86.0.1 on the latest Ubuntu LTS, with integrated AMD GPU. Mesa version is 3.0 Mesa 20.0.8.
The nightly's mesa version is: 3.0 Mesa 20.2.6

So how come does this work on the nightly version even though the underlying driver is the same?

Comment 26

[Jeff Muizelaar [:jrmuizel]]

This may have been fixed in Nightly. If this is true it would be valuable if you could find the change that fixed it. This should be doable with https://mozilla.github.io/mozregression/

Comment 27

[Robert Mader [:rmader]]

(In reply to mehrmor from comment #25)

I believe I have basically the same issue as the reporter. Here's a short clip:
https://1drv.ms/u/s!AlqFeqoW1KpnhOJQXv_ho1GNt_8J0Q?e=g9aIMi

I didn't experience the issue using the nightly version: 88.0a1.

Running version 86.0.1 on the latest Ubuntu LTS, with integrated AMD GPU. Mesa version is 3.0 Mesa 20.0.8.
The nightly's mesa version is: 3.0 Mesa 20.2.6

So how come does this work on the nightly version even though the underlying driver is the same?

As you're on AMD you might see bug 1679681 which is fixed in nightly (unintentionally fixed by bug 1694909).

Comment 28

[mehrmor]

I can confirm the bug can't be reproduced on the current beta version : 88.0b4

Comment 30

[2tt56k2hg]

well i've have changed from using an nvidia graphics adapter beck to the in build graphics and it seems to be fix for me too. im not sure what changed however?

Comment 31

[Robert Mader [:rmader]]

(In reply to 2tt56k2hg from comment #30)

and it seems to be fix for me too.

Closing then. Please reopen if you see a similar issue again.