1878754 - rendering bugs on several pages when moving the mouse pointer over links/page

Status: NEW

(Core :: Graphics: WebRender, defect, P1)

Description

[jbin]

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:122.0) Gecko/20100101 Firefox/122.0

Steps to reproduce:

Just browse the page, scroll or move the mouse pointer over some points/links of the page. It can happen on several websites and pages. Some times you don't have to do nothing to it happen.

I also tried this on beta and nightly versions and I got the same problem.
This bug doesn't happen on any Linux distribution.

System info;
11th Gen Intel(R) Core(TM) i5-1135G7 @ 2.40GHz
2x8GB 2400MHz 19CL
Intel(R) Iris(R) Xe Graphics
SSD01: PNY CS900 250GB SSD (SATA3) (Linux)
SSD02: WDC WDS480G2G0C-00AJM0 (NVME-PCIEX) (Windows)
Microsoft Windows 10 Home Single Language
10.0.19045 N/A compilação 19045
SAMSUNG ELECTRONICS CO., LTD. - Samsung Book 550XDA
Windows Virtual Memory is disabled
Hibernation is disabled ("powercfg -h off")

Actual results:

It causes some kind of rendering issues on the page. I don't know how to tell exactly, but I can show it.
See the images in the links below.
https://i.imgur.com/QLcpc7D.jpeg
https://i.imgur.com/PXADlP5.jpeg
https://i.imgur.com/wU0kTQS.jpeg

Expected results:

Nothing.

Comment 1

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

The Bugbug bot thinks this bug should belong to the 'Core::Graphics: WebRender' component, and is moving the bug to that component. Please correct in case you think the bot is wrong.

Comment 2

[Daniel Holbert [:dholbert]]

Attachment: reporter's screenshot 1

Comment 3

[Daniel Holbert [:dholbert]]

Attachment: reporter's screenshot 2

Comment 4

[Daniel Holbert [:dholbert]]

Attachment: reporter's screenshot 3

Comment 5

[Daniel Holbert [:dholbert]]

Thanks for the bug report.

Could you visit about:support, and copy the contents of the "Graphics" section (you might have to scroll a bit to get the whole section) into a textfile included as a file-attachment here, using the "Attach New File" button?

Comment 6

[jbin]

Attachment: TXT containing data from: about:support->Graphics

Comment 7

[jbin]

Attachment: reporter's screenshot 4 - bug happening on about:support

Comment 8

[jbin]

Comment on attachment 9378441 [details]
reporter's screenshot 4 - bug happening on about:support

It happened when I was selecting the table content.

Comment 9

[jbin]

I tested some older versions (117.0, 116.0 and 115.0) and this bug doesn't happen on these versions.

Comment 10

[Daniel Holbert [:dholbert]]

That's useful to know, thanks.

If you're able to determine whether a build is good/bad pretty quickly, maybe you could try bisecting Firefox Nightly builds using https://mozilla.github.io/mozregression/ to find out the first-bad build? (At the end of the nightly-bisecting process, it should give you a URL near the bottom of its output-log with a commit range of potentially-guilty commits, which is super handy to be able to isolate the issue.)

Comment 11

[jbin]

Attachment: mozregression output-log

Comment 12

[jbin]

I'm not sure if I did it right, but I attached the text log.

I think you want this URL: Differential Revision: https://phabricator.services.mozilla.com/D185878

The last thing that appeared in the "Build Info"; (maybe useful)
pushlog_url: https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=6d77e3e6758ed20403fb693b35651d0f549d18ff&tochange=a5bcd3edffd78b69932e1ea7c5dda0894107edaa
repo_name: autoland

Comment 13

[Daniel Holbert [:dholbert]]

Thanks! Indeed, it looks like this is a regression from that patch for bug 1848066

The last thing that appeared in the "Build Info"; (maybe useful)
pushlog_url: https://hg.mozilla.org/integration/autoland/pushloghtml?fromchange=6d77e3e6758ed20403fb693b35651d0f549d18ff&tochange=a5bcd3edffd78b69932e1ea7c5dda0894107edaa

Thanks -- yeah, it looks like that's a push range from just before bisection finished. In your attached log, it shows the bisecting got a bit further, with commit bdeed127 being judged as 'good' which only leaves one commit in the regression range, the one from bug 1848066.

gw, could you take a look here?

Comment 14

[Glenn Watson [:gw]]

We've seen this previously, it's a driver bug - the earlier Xe drivers had this problem. From the about:support, the driver version is from 3-22-2021, which is quite old. jbin, would it be possible to try update your graphics driver?

Comment 15

[Daniel Holbert [:dholbert]]

(assuming that's the explanation: maybe we should be block-listing certain old known-buggy Intel Xe driver-versions, so that we fall back to software webrender or whatever?)

Comment 16

[Glenn Watson [:gw]]

That's a good idea - I think at the time we last saw it that driver wasn't so old, but now it's probably reasonable to block that version.

Comment 17

[jbin]

It's working now. I'm sorry for the time waste.
I thought Windows 10/11 would keep the drivers updated automatically, because it installs the Samsung Book software and keeps my BIOS updated automatically.

Comment 18

[Daniel Holbert [:dholbert]]

(In reply to jbin from comment #17)

It's working now.

That's good to hear, thanks for confirming!

(In reply to Glenn Watson [:gw] from comment #16)

That's a good idea - I think at the time we last saw it that driver wasn't so old, but now it's probably reasonable to block that version.

Maybe worth morphing this bug to do that, if/when you (or someone else who maintains that blocklist) has cycles?

Comment 19

[Glenn Watson [:gw]]

Yup, agreed. Andrew, is adding driver blocks something you're still able to do, or suggestions on who knows that code well?

Comment 20

[Andrew Osmond [:aosmond] (he/him)]

I can do this.

Comment 21

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

Set release status flags based on info from the regressing bug 1848066

Comment 22

[Bob Hood [:bhood]]

Andrew, would you mind applying Priority/Severity settings to this?

Comment 24

[Ryan VanderMeulen [:RyanVM]]

Is this just waiting on a blocklist update? Anything we can do to move this along in time for 125 going to RC next week?

Comment 25

[Andrew Osmond [:aosmond] (he/him)]

So we can block Tigerlake for some drivers, but since this seems to have been fixed by a driver update, we don't normally block everything older:
https://github.com/jrmuizel/gpu-db/blob/1aa20825576f83329d3ef39717bbf84d1db4484b/intel.json#L450

What driver version did things start working for you? Can you attach an updated about:support now that things are working? Thanks!

Comment 26

[jbin]

I don't know what is the first good version. I was using a really old version when I made this report, and now I'm using the latest version of the driver.
I'm not able to find out what is the first good version of that driver, because Intel's website doesn't allow me to download much older versions of their drivers.

Comment 27

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Bug 1878754 - Block Windows Intel TigerLake drivers older than 30.*.

Comment 28

[Jeff Muizelaar [:jrmuizel]]

We need to know a more precise range for the broken driver before we can block this.

Comment 29

[Jeff Muizelaar [:jrmuizel]]

I'll try to get some similar hardware in the Toronto office.

Comment 30

[Jeff Muizelaar [:jrmuizel]]

I ordered a Samsung Galaxy Book Pro 13.3" FHD i5 1135G7. I should get it between Wed, Apr 24–Tue, Apr 30

Comment 31

[Jeff Muizelaar [:jrmuizel]]

I have the computer now.

Comment 32

[Jeff Muizelaar [:jrmuizel]]

I can reproduce this the 27.20.100.9415 driver installed

Comment 33

[Jeff Muizelaar [:jrmuizel]]

I can reproduce with 27.20.100.9271 as well

Comment 34

[Jeff Muizelaar [:jrmuizel]]

But not with 27.20.100.8439 on a different gen12 laptop (0x9a78)

Comment 35

[Donal Meehan [:dmeehan]]

:jrmuizel any updates on next steps here?
We could take it in dot release for 126 is you know what to block, or at least 127

Comment 36

[Bob Hood [:bhood]]

Can we let Andrew block those two identified driver versions?

Comment 37

[Donal Meehan [:dmeehan]]

:bhood are there any expectations around getting a patch in time for Fx127?
If not even in time for the Fx126 planned dot release on 2024-05-28

Comment 38

[Jim Mathies [:jimm]]

Andrew is busy, I doubt he'll be able to get to this. Can someone else pick this up?

Comment 39

[Donal Meehan [:dmeehan]]

Setting 126 to fix-optional in case this gets a patch in time for the planned dot release.
Could someone pick this up at least in time for Fx127?

Comment 40

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

As triage owner, can you help find someone to work on this ? Thanks!

Comment 41

[Glenn Watson [:gw]]

Bob, is there someone who can work on this?

Comment 42

[Brad Werth [:bradwerth]]

I'll try to reproduce this on my comparable system, and figure out which driver versions are susceptible.

Comment 43

[Brad Werth [:bradwerth]]

I can't find a download for older Intel graphics drivers for my Icelake system. The oldest I can find are the currently-supported 30.x versions. Not sure how to proceed on this to find the version number between 27 and 30 that is our target cutoff version.

Jeff do you have access to a driver archive?

Comment 44

[Jeff Muizelaar [:jrmuizel]]

https://wiki.mozilla.org/QA/Platform/Graphics/Inventory has some suggestions