Bug 2391242

Created attachment 2104990 [details]
Journalctl logs from Firefox

After a longer observation, I can say, that although the rendering issue goes away after a while and "painting" with the mouse is not necessary anymore, the whole application is much slower than before.

I am using the following graphical card: c2:00.0 Display controller: Advanced Micro Devices, Inc. [AMD/ATI] Strix Halo [Radeon Graphics / Radeon 8050S Graphics / Radeon 8060S Graphics] (rev c4)

srp 27 11:17:07 framebox org.mozilla.firefox.desktop[3811]: [GFX1-]: Handling webrender error 2
srp 27 11:17:07 framebox org.mozilla.firefox.desktop[3811]: [GFX1-]: Fallback WR to SW-WR

Please report at https://bugzilla.mozilla.org/ and cc me there (component core / graphics).
Thanks.

I have opened it there. Thanks.

Proposed as a Blocker for 43-final by Fedora user lruzicka using the blocker tracking app because:

 https://fedoraproject.org/wiki/Fedora_43_Final_Release_Criteria#Default_application_functionality

(In reply to Lukas Ruzicka from comment #5)
> I have opened it there. Thanks.

Lukas, please add an upstream link.

I can't reproduce this bug with the following:

firefox-142.0-3.fc43.x86_64
mesa-dri-drivers-25.1.4-2.fc43.x86_64
gnome-shell-49~alpha.1-2.fc43.x86_64
mutter-49~alpha.1-3.fc43.x86_64

on Thinkpad T14s Gen6:
AMD Ryzen AI 5 PRO 340 w/ Radeon 840M
c4:00.0 Display controller [0380]: Advanced Micro Devices, Inc. [AMD/ATI] Krackan [Radeon 840M / 860M Graphics] [1002:1114] (rev d3)

Here's the upstream link:
https://bugzilla.mozilla.org/show_bug.cgi?id=1986254

Lukas isn't crazy, this has been hitting me constantly on my main system - using Rawhide Silverblue, but it's clearly the same thing. It seems for me to be triggered by certain *page content* - graphically heavy pages, maybe pages with video? It doesn't have anything to do with private mode, I don't think.

My system is also recent AMD:

c3:00.0 Display controller [0380]: Advanced Micro Devices, Inc. [AMD/ATI] Strix [Radeon 880M / 890M] [1002:150e] (rev c4)

Discussed at the 2025-09-01 (blocker / freeze exception) review meeting:

This is clearly too fuzzy to vote on right now. We need to punt it for further investigation to try to nail down the details.

https://meetbot-raw.fedoraproject.org//blocker-review_matrix_fedoraproject-org/2025-09-01/f43-blocker-review.2025-09-01-16.00.txt

Discussed at the 2025-09-08 (blocker / freeze exception) review meeting:

The causes and affected systems are still being narrowed down. We will hold off on any voting at this time.

https://meetbot-raw.fedoraproject.org//blocker-review_matrix_fedoraproject-org/2025-09-08/f43-blocker-review.2025-09-08-16.00.txt

I finally managed to reproduce this issue on a laptop with this GPU:
c4:00.0 Display controller [0380]: Advanced Micro Devices, Inc. [AMD/ATI] Krackan [Radeon 840M / 860M Graphics] [1002:1114] (rev d3)

It was also accompanied with "Detect DeviceReset" and then "Failed to compile vertex shader" error messages as in the upstream issue.


On my other laptop with an Intel GPU, I see no such behavior:
00:02.0 VGA compatible controller [0300]: Intel Corporation TigerLake-H GT1 [UHD Graphics] [8086:9a60] (rev 01)

Discussed at the 2025-09-15 (blocker / freeze exception) review meeting:

We agreed to postpone this one as the exact details of the problem are still not clear. We will continue working with upstream to clarify the issue for next week's meeting.

https://meetbot-raw.fedoraproject.org//blocker-review_matrix_fedoraproject-org/2025-09-15/f43-blocker-review.2025-09-15-16.00.txt

I just reproduced this on a desktop with Radeon 6700 XT dedicated GPU, using the F43 Beta Workstation Live image. In https://pagure.io/fedora-qa/blocker-review/issue/1896 we have another affected user with Radeon 6800 XT. So it looks quite possible that this affects all or a large majority of AMD GPU users (running the amdgpu kernel driver), with both integrated and dedicated cards.

I can reproduce this 80% reliably by uploading an image into the mastodon web interface.  As soon as the file selector comes up the firefox windows go blank.
Running under X/xfce

09:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Lexa PRO [Radeon 540/540X/550/550X / RX 540X/550/550X] (rev c7)

I can reproduce this too, like 90% of the time, on normal firefox usage (not private window). If I acces the https://get.webgl.org/ page, sometimes I see the cube, sometimes I don't. I already change the "performance" settings at about:preferences and reset again, and sometimes it helps, and sometimes it doesn't.

firefox-142.0.1-1.fc43.x86_64

Operating System: Fedora Linux 43
KDE Plasma Version: 6.4.4
KDE Frameworks Version: 6.17.0
Qt Version: 6.9.2
Kernel Version: 6.17.0-0.rc6.49.fc43.x86_64 (64-bit)
Graphics Platform: Wayland
Processors: 16 × AMD Ryzen 7 5700U with Radeon Graphics
Memory: 24 GiB of RAM (22.8 GiB usable)
Graphics Processor: AMD Radeon Graphics

I forgot to mention:
mesa-dri-drivers-25.2.2-6.fc43.x86_64

I think this bug should be moved to mesa, no?

(In reply to Jens Petersen from comment #19)
> I think this bug should be moved to mesa, no?

This bug was moved to Mesa a long time ago. See https://gitlab.freedesktop.org/mesa/mesa/-/issues/13877

I confirm the issue occurs on normal Firefox 142.0.1 with AMD Hawk Point:

## Hardware Information:
- **Hardware Model:**                              Dell Inc. Inspiron 14 7445 2-in-1
- **Memory:**                                      64.0 GiB
- **Processor:**                                   AMD Ryzen™ 7 8840HS w/ Radeon™ 780M Graphics × 16
- **Graphics:**                                    AMD Radeon™ 780M Graphics
- **Disk Capacity:**                               4.0 TB

## Software Information:
- **Firmware Version:**                            1.13.1
- **OS Name:**                                     Fedora Linux 43 (Design Suite Prerelease)
- **OS Type:**                                     64-bit
- **GNOME Version:**                               49
- **Windowing System:**                            Wayland
- **Kernel Version:**                              Linux 6.17.0-0.rc6.49.fc43.x86_64
- **Driver**                                       mesa-dri-drivers-25.2.2-6.fc43

(In reply to Adam Williamson from comment #10)
> 
> My system is also recent AMD:
> 
> c3:00.0 Display controller [0380]: Advanced Micro Devices, Inc. [AMD/ATI]
> Strix [Radeon 880M / 890M] [1002:150e] (rev c4)

Adam, I'm running here with your kernel scratch build https://gitlab.freedesktop.org/mesa/mesa/-/issues/13877#note_3105382 and so far it seems to me the bug is gone.

Kernel Version: 6.17.0-0.1awb.rc6.49.fc43.x86_64 (64-bit)
Graphics Platform: Wayland
Processors: 16 × AMD Ryzen 7 5700U with Radeon Graphics
Memory: 24 GiB of RAM (22.8 GiB usable)
Graphics Processor: AMD Radeon Graphics

05:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Lucienne (rev c1) (prog-if 00 [VGA controller])
	Subsystem: Acer Incorporated [ALI] Device 152b
	Flags: bus master, fast devsel, latency 0, IRQ 47, IOMMU group 7
	Memory at fce0000000 (64-bit, prefetchable) [size=256M]
	Memory at fcf0000000 (64-bit, prefetchable) [size=2M]
	I/O ports at 1000 [size=256]
	Memory at d0400000 (32-bit, non-prefetchable) [size=512K]
	Capabilities: <access denied>
	Kernel driver in use: amdgpu
	Kernel modules: amdgpu

Thanks Geraldo! Can other affected folks please try with the kernel above and see if they can still reproduce the issue? Note if you have Secure Boot enabled you'll have to disable it, obviously I cannot properly sign a scratch built kernel. You can check the .src.rpm to make sure I didn't put anything nasty in it.

Direct link: https://koji.fedoraproject.org/koji/taskinfo?taskID=137288826

Adam, that's looking pretty good - been using it for a few hours now and not had any problems:

dg@dalek:~$ uname -a
Linux dalek 6.17.0-0.1awb.rc6.49.fc43.x86_64 #1 SMP PREEMPT_DYNAMIC Fri Sep 19 22:37:44 UTC 2025 x86_64 GNU/Linux
dg@dalek:~$ inxi -G
Graphics:
  Device-1: Advanced Micro Devices [AMD/ATI] Lexa PRO [Radeon
    540/540X/550/550X / RX 540X/550/550X] driver: amdgpu v: kernel
  Display: x11 server: X.Org v: 21.1.18 with: Xwayland v: 24.1.8 driver: X:
    loaded: amdgpu unloaded: modesetting dri: radeonsi gpu: amdgpu resolution:
    1: 1920x1080~60Hz 2: 3840x2160~60Hz
  API: EGL v: 1.5 drivers: kms_swrast,radeonsi,swrast
    platforms: gbm,x11,surfaceless,device
  API: OpenGL v: 4.6 compat-v: 4.5 vendor: amd mesa v: 25.2.2 renderer: AMD
    Radeon RX 550 / 550 Series (radeonsi polaris12 ACO DRM 3.64
    6.17.0-0.1awb.rc6.49.fc43.x86_64)
  API: Vulkan v: 1.4.321 drivers: radv,llvmpipe surfaces: N/A
  Info: Tools: api: clinfo, eglinfo, glxinfo, vulkaninfo de: kscreen-console,
    kscreen-doctor, xfce4-display-settings gpu: radeontop,umr wl: wayland-info
    x11: xdriinfo, xdpyinfo, xprop, xrandr

+6 in https://pagure.io/fedora-qa/blocker-review/issue/1896 , marking accepted. Latest news is in https://gitlab.freedesktop.org/drm/amd/-/issues/4568#note_3106986 - in a shocking turn of events this has been kicked right back up the stack to firefox (doing something it shouldn't be doing with OpenGL in forked processes, apparently).

This is accepted under the 'basic functionality' criterion, as Firefox is borderline unusable on all or at least most adapters which use amdgpu, which is a pretty wide range.

There's now an upstream fix for glycin - https://gitlab.gnome.org/GNOME/glycin/-/commit/8af36048dbdda27a05b87b2fc896c05161d21f64 - though I'm not sure if this is more of a case of 'workaround Firefox's bad behaviour', i.e. whether we should re-assign this to glycin or still expect Firefox to change something here.

There's also a Firefox fix (I think) - https://bugzilla.mozilla.org/show_bug.cgi?id=1990162 . Here is a scratch build with that patch backported, for testing: https://koji.fedoraproject.org/koji/taskinfo?taskID=137452120

The scratch build seems resolving the issue on Ryzen 7 8840HS powering Dell Inspiron 7445 convertible. 
The painting effect and the slowness are gone as tested with the latest kernel 6.17.0-0.rc7.56.fc43.x86_64.
Running from terminal showed this message:

> [GFX1-]: RenderCompositorSWGL failed mapping default framebuffer, no dt

Ideally, that warning should get resolved but it seems harmless for now in practice.

FEDORA-2025-edd4efffa4 (firefox-143.0.1-3.fc43) has been submitted as an update to Fedora 43.
https://bodhi.fedoraproject.org/updates/FEDORA-2025-edd4efffa4

FEDORA-2025-edd4efffa4 (firefox-143.0.1-3.fc43) has been pushed to the Fedora 43 stable repository.
If problem still persists, please make note of it in this bug report.

With the latest updates, I confirm that I have not seen any issue for the entire day of using it.