2001537 - [Linux Nightly] Constant process crashing on video start or stop, on amdgpu [@gbm_create_device]

Status: RESOLVED DUPLICATE of bug 1995035

(Core :: Widget: Gtk, defect, P3)

Description

[Tem PQD [:tempqd]]

Attachment: firefox-video-crashing.log

Steps to reproduce:

Latest Nightly (147.0a1 2025-11-20) on KDE git master (openSUSE Krypton latest snapshot) Wayland with Radeon 680m amdgpu.

KDE's crash handler DrKonqi recently updated to also show crashes from non-KDE programs, and on latest Nightly I now notice constant service crashes whenever video playback starts or stops. These coincide with video players buffering for a long time, or crashing to their error message screen.

A good reliable reproducer is the Twitch player, which seems to constantly start & stop streams and triggers the crash each time. Old Reddit embedded videos also seem to trigger it reliably. YouTube has similar severe issues as well but not as clear to me what's happening.

Actual results:

Backtrace of one crash attached. (The window does not close, I just get the DrKonqi popups shown in the screenshot constantly and video player crashes/errors)

Expected results:

Video starts quickly and stops without any crashing or error screens

Comment 1

[Tem PQD [:tempqd]]

Attachment: Screenshot of KDE git master DrKonqi's constant service crashes

Comment 2

[Tem PQD [:tempqd]]

Maybe regression from Bug 1998839?

My time is very limited atm and I suspect many will soon start encountering this, but I can try mozregression if needed.

Comment 3

[BugBot [:suhaib / :marco]]

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

Comment 4

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

Please attach your about:support page.
Thanks.

Comment 5

[Tem PQD [:tempqd]]

Attachment: about:support

Comment 6

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

The stat looks clean. Bug 1998839 is Windows only. AFAIK we use zero copy on AMD by default, you have it enabled (HW_DECODED_VIDEO_ZERO_COPY: default: available). You can try to flip media.ffmpeg.vaapi.force-surface-zero-copy to 0 at about:config but I doubt it's the problem here (but may be!).

#4  __fcntl64_nocancel_adjusted (fd=28, cmd=1027, arg=0xe) at ../sysdeps/unix/sysv/linux/fcntl_nocancel.c:63
#5  0x00007f20cbf0eb17 in __GI___libc_fcntl64 (fd=fd@entry=28, cmd=cmd@entry=1027)
    at ../sysdeps/unix/sysv/linux/fcntl64.c:51
#6  0x00007f206ec21d74 in os_same_file_description (fd1=28, fd2=fd2@entry=14) at ../src/util/os_file.c:248
#7  0x00007f206f13ae41 in are_file_descriptions_equal (fd1=<optimized out>, fd2=14)
    at ../src/gallium/winsys/amdgpu/drm/amdgpu_winsys.c:347
#8  amdgpu_winsys_create
    (fd=fd@entry=14, config=config@entry=0x7f20729b5970, screen_create=screen_create@entry=0x7f206f087590 <radeonsi_screen_create_impl>, is_virtio=is_virtio@entry=false)
    at ../src/gallium/winsys/amdgpu/drm/amdgpu_winsys.c:448
#9  0x00007f206f088768 in radeonsi_screen_create (fd=14, config=0x7f20729b5970)
    at ../src/gallium/drivers/radeonsi/si_pipe.c:1704
#10 0x00007f206e7666f6 in pipe_radeonsi_create_screen (fd=<optimized out>, config=<optimized out>)
    at ../src/gallium/auxiliary/target-helpers/drm_helper.h:189
#11 0x00007f206ef1e7a5 in pipe_loader_create_screen_vk
    (dev=0x7f20784264c0, sw_vk=sw_vk@entry=false, driver_name_is_inferred=driver_name_is_inferred@entry=false) at ../src/gallium/auxiliary/pipe-loader/pipe_loader.c:180
#12 0x00007f206ef1e7db in pipe_loader_create_screen
    (dev=<optimized out>, driver_name_is_inferred=driver_name_is_inferred@entry=false)
    at ../src/gallium/auxiliary/pipe-loader/pipe_loader.c:186
#13 0x00007f206e772b1d in dri2_init_screen
--Type <RET> for more, q to quit, c to continue without paging--c
    (screen=screen@entry=0x7f20cbc4c100, driver_name_is_inferred=driver_name_is_inferred@entry=false)
    at ../src/gallium/frontends/dri/dri2.c:1705
#14 0x00007f206e76a63e in driCreateNewScreen3
    (scrn=scrn@entry=0, fd=23, loader_extensions=0x7f207e7f1040 <gbm_dri_screen_extensions>, type=DRI_SCREEN_DRI3, driver_configs=0x7f20cbc3d590, driver_name_is_inferred=driver_name_is_inferred@entry=false, has_multibuffer=true, data=0x7f20cbc3d480) at ../src/gallium/frontends/dri/dri_util.c:130
#15 0x00007f207e7e7047 in dri_screen_create_for_driver
    (dri=dri@entry=0x7f20cbc3d480, driver_name=<optimized out>, driver_name_is_inferred=driver_name_is_inferred@entry=false) at ../src/gbm/backends/dri/gbm_dri.c:267
#16 0x00007f207e7e7af2 in dri_screen_create (dri=0x7f20cbc3d480) at ../src/gbm/backends/dri/gbm_dri.c:293
#17 dri_device_create (fd=23, gbm_backend_version=1) at ../src/gbm/backends/dri/gbm_dri.c:1231
#18 0x00007f207e7f4366 in backend_create_device (bd=0x7f20784c7060, fd=23) at ../src/gbm/main/backend.c:95
#19 load_backend_by_name
    (name=name@entry=0x7f207e7f5033 "dri", fd=fd@entry=23, warn_on_fail=warn_on_fail@entry=true)
    at ../src/gbm/main/backend.c:135
#20 0x00007f207e7f4455 in _gbm_create_device (fd=fd@entry=23) at ../src/gbm/main/backend.c:165
#21 0x00007f207e7f44c9 in gbm_create_device (fd=23) at ../src/gbm/main/gbm.c:138
#22 0x00007f2083c9dcc0 in mozilla::widget::DMABufDevice::GetDevice(mozilla::widget::DMABufDeviceLock*) ()
    at /home/tempqd/Downloads/firefox/libxul.so

The crash comes directly from AMD drivers when we're initializing DMABuf interface. I think you should report it at MESA directly:
https://gitlab.freedesktop.org/mesa/mesa/-/issues

Comment 7

[Tem PQD [:tempqd]]

Ah, I see! Disabling it with media.ffmpeg.vaapi.force-surface-zero-copy = 0 has no change unfortunately, still crashing the same it seems.

Will file mesa issue and report back here. Thanks!

Comment 8

[Tem PQD [:tempqd]]

Mesa report: https://gitlab.freedesktop.org/mesa/mesa/-/issues/14330

Comment 9

[Jan Alexander Steffens [:heftig]]

This is probably a duplicate of bug 1995035.

Comment 10

[Tem PQD [:tempqd]]

Can confirm fixed in latest Nightly after the patch to Bug 1995035. Thanks!

Let me know if any further investigation is needed, otherwise feel free to mark dupe & close.

Comment 11

[Tem PQD [:tempqd]]

Ironically, right after commenting this I got a "You have 216 unsent crash reports" popup for the first time, no need for those anymore 😄

Comment 12

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

Okay, Thanks.