Game: RADV Vega 9 loses GPU during intro cinematic on Linux #1737

Open
opened 2026-11-19 12:13:01 +00:00 by BeanBot · 5 comments
Member

Reported on a DM to the bot on Discord from kenokeefe

Longshot version and channel

The attached log identifies the build as v3.0.0-rc.8; the reporter stated v3.0.0. Exact channel/build provenance needs confirmation.

What happened

The game crashes with VK_ERROR_DEVICE_LOST near the end of the intro cinematic, after the title display. The reporter tried RX_VK_MESH_SHADERS=compute in the build folder's longshot_env.txt; it did not prevent the crash.

Where / frequency

End of the intro cinematic after title display. Frequency not specified.

GPU and driver

AMD Radeon Graphics (RADV RENOIR), identified in the log; reporter describes the GPU as Vega 9, Vulkan 1.3, Mesa 26. Exact Mesa/driver version needs confirmation.

CPU, RAM, OS

Ryzen 7 5800H, 8 cores / 16 threads; Ubuntu 24.04.5 LTS. RAM not specified.

Logs

Attached longshot_log.txt excerpt:

[hos host] longshot v3.0.0-rc.8 (source 5a1a671d)
[hos host] WARNING: /dev/udmabuf permission denied; GPU imports may be slower, especially on AMD/RADV. Ask your administrator for access via its device group or a device-specific udev rule (see README, On Linux). Do not run the game as root. (Permission denied)
[hos host] vk: device 0: AMD Radeon Graphics (RADV RENOIR), Vulkan 1.4, driver 0x6402008, type 1, id 1002:1638
[hos host] vk: no mesh shaders: task and mesh programs run as compute shaders (issue 640 fallback)
[hos Presentation Thread] vk: slow submit: 2116.1 ms: texture lookups 1.9 uploads 1.7 dispatch 2116.1 queue submit 12.6
radv/amdgpu: The CS has been cancelled because the context is lost. This context is guilty of a hard recovery.
[hos Presentation Thread] vk: batched submit 84897 failed (-4)
[hos Presentation Thread] vk: device lost: reading the fault info (RX_VK_CHECKPOINTS=1 for markers)
[hos Presentation Thread] vk: device fault:
[hos Presentation Thread] vk: FATAL: the GPU device was lost (VK_ERROR_DEVICE_LOST): the driver stopped or reset the GPU, and nothing sent to it runs any more (on Linux, dmesg or journalctl -k around this time names the reset)
vk: FATAL: the GPU device was lost (VK_ERROR_DEVICE_LOST): the driver stopped or reset the GPU, and nothing sent to it runs any more (on Linux, dmesg or journalctl -k around this time names the reset)

The complete log is attached to the report. Kernel/journal messages around the reset were not provided.

Still unknown

Exact release channel/build, exact Mesa driver version, RAM, how often it reproduces, kernel/journal messages around the GPU reset, and confirmation that the environment setting was in the active build's folder (the supplied path indicates /home/<user>/Games/Longshot/rc/v3.0.0/longshot_env.txt).

Reported on a DM to the bot on Discord from `kenokeefe` ### Longshot version and channel The attached log identifies the build as `v3.0.0-rc.8`; the reporter stated v3.0.0. Exact channel/build provenance needs confirmation. ### What happened The game crashes with `VK_ERROR_DEVICE_LOST` near the end of the intro cinematic, after the title display. The reporter tried `RX_VK_MESH_SHADERS=compute` in the build folder's `longshot_env.txt`; it did not prevent the crash. ### Where / frequency End of the intro cinematic after title display. Frequency not specified. ### GPU and driver AMD Radeon Graphics (RADV RENOIR), identified in the log; reporter describes the GPU as Vega 9, Vulkan 1.3, Mesa 26. Exact Mesa/driver version needs confirmation. ### CPU, RAM, OS Ryzen 7 5800H, 8 cores / 16 threads; Ubuntu 24.04.5 LTS. RAM not specified. ### Logs Attached `longshot_log.txt` excerpt: ```text [hos host] longshot v3.0.0-rc.8 (source 5a1a671d) [hos host] WARNING: /dev/udmabuf permission denied; GPU imports may be slower, especially on AMD/RADV. Ask your administrator for access via its device group or a device-specific udev rule (see README, On Linux). Do not run the game as root. (Permission denied) [hos host] vk: device 0: AMD Radeon Graphics (RADV RENOIR), Vulkan 1.4, driver 0x6402008, type 1, id 1002:1638 [hos host] vk: no mesh shaders: task and mesh programs run as compute shaders (issue 640 fallback) [hos Presentation Thread] vk: slow submit: 2116.1 ms: texture lookups 1.9 uploads 1.7 dispatch 2116.1 queue submit 12.6 radv/amdgpu: The CS has been cancelled because the context is lost. This context is guilty of a hard recovery. [hos Presentation Thread] vk: batched submit 84897 failed (-4) [hos Presentation Thread] vk: device lost: reading the fault info (RX_VK_CHECKPOINTS=1 for markers) [hos Presentation Thread] vk: device fault: [hos Presentation Thread] vk: FATAL: the GPU device was lost (VK_ERROR_DEVICE_LOST): the driver stopped or reset the GPU, and nothing sent to it runs any more (on Linux, dmesg or journalctl -k around this time names the reset) vk: FATAL: the GPU device was lost (VK_ERROR_DEVICE_LOST): the driver stopped or reset the GPU, and nothing sent to it runs any more (on Linux, dmesg or journalctl -k around this time names the reset) ``` The complete log is attached to the report. Kernel/journal messages around the reset were not provided. ### Still unknown Exact release channel/build, exact Mesa driver version, RAM, how often it reproduces, kernel/journal messages around the GPU reset, and confirmation that the environment setting was in the active build's folder (the supplied path indicates `/home/<user>/Games/Longshot/rc/v3.0.0/longshot_env.txt`).
Author
Member

Files from the Discord report:

Files from the Discord report: - [message.txt](https://git.magicbean.dev/attachments/2aa5c65b-5244-41d9-93f0-3fa9fa592447)
Author
Member

From a DM to the bot on Discord from kenokeefe:

Added Linux kernel messages around the GPU reset. The reporter requests that the computer hostname be anonymized; omitted the hostname noether.

Nov 19 12:04:28 amdgpu 0000:04:00.0: Dumping IP State
Nov 19 12:04:28 amdgpu 0000:04:00.0: Dumping IP State Completed
Nov 19 12:04:28 amdgpu 0000:04:00.0: [drm] AMDGPU device coredump file has been created
Nov 19 12:04:28 amdgpu 0000:04:00.0: [drm] Check your /sys/class/drm/card1/device/devcoredump/data
Nov 19 12:04:28 amdgpu 0000:04:00.0: ring gfx timeout, signaled seq=1591941, emitted seq=1591944
Nov 19 12:04:28 amdgpu 0000:04:00.0:  Process longshot pid 22872 thread Presentation Th pid 22884
Nov 19 12:04:28 amdgpu 0000:04:00.0: Starting gfx ring reset
Nov 19 12:04:28 amdgpu 0000:04:00.0: Ring gfx reset succeeded
Nov 19 12:04:28 amdgpu 0000:04:00.0: [drm] device wedged, but recovered through reset
From a DM to the bot on Discord from `kenokeefe`: Added Linux kernel messages around the GPU reset. The reporter requests that the computer hostname be anonymized; omitted the hostname `noether`. ```text Nov 19 12:04:28 amdgpu 0000:04:00.0: Dumping IP State Nov 19 12:04:28 amdgpu 0000:04:00.0: Dumping IP State Completed Nov 19 12:04:28 amdgpu 0000:04:00.0: [drm] AMDGPU device coredump file has been created Nov 19 12:04:28 amdgpu 0000:04:00.0: [drm] Check your /sys/class/drm/card1/device/devcoredump/data Nov 19 12:04:28 amdgpu 0000:04:00.0: ring gfx timeout, signaled seq=1591941, emitted seq=1591944 Nov 19 12:04:28 amdgpu 0000:04:00.0: Process longshot pid 22872 thread Presentation Th pid 22884 Nov 19 12:04:28 amdgpu 0000:04:00.0: Starting gfx ring reset Nov 19 12:04:28 amdgpu 0000:04:00.0: Ring gfx reset succeeded Nov 19 12:04:28 amdgpu 0000:04:00.0: [drm] device wedged, but recovered through reset ```
Author
Member

From a DM to the bot on Discord from kenokeefe:

The reporter's added kernel messages confirm an AMDGPU graphics-ring timeout and successful ring reset while Longshot's Presentation Thread was active. Hostname omitted from the issue comment.

From a DM to the bot on Discord from `kenokeefe`: The reporter's added kernel messages confirm an AMDGPU graphics-ring timeout and successful ring reset while Longshot's Presentation Thread was active. Hostname omitted from the issue comment.
Author
Member

From a DM to the bot on Discord from kenokeefe:

The reporter reran with RX_VK_CHECKOINTS=1 and provided checkpoint output. Note the variable as typed is CHECKOINTS (not CHECKPOINTS); no confirmation was provided whether that is a typo or the setting was recognized. Output:

[hos Presentation Thread] vk: batched submit 113543 failed (-4)
[hos Presentation Thread] vk: device lost: reading the buffer markers
[hos Presentation Thread] vk: device lost: 116756 markers set, 64511 reached; the last whose earlier work all finished: 115904
[hos Presentation Thread] vk:   finished before: marker 115898 (upload of texture 00000010ce9cea00, frame 26081)
[hos Presentation Thread] vk:   finished before: marker 115899 (upload of texture 00000010ce9cea00, frame 26081)
[hos Presentation Thread] vk:   finished before: marker 115900 (upload of texture 00000010ce9cea00, frame 26081)
[hos Presentation Thread] vk:   finished before: marker 115901 (dispatch of program d5ae882ba38e7737, frame 26081)
[hos Presentation Thread] vk:   finished before: marker 115902 (draw of program d5ae882ba38e7737, frame 26081)
[hos Presentation Thread] vk:   finished before: marker 115903 (upload of texture 00000010ce4e5a00, frame 26081)
[hos Presentation Thread] vk:   RUNNING after: marker 115904 (upload of texture 00000010ce4e5a00, frame 26081)
[hos Presentation Thread] vk:   running: marker 115905 (dispatch of program d5ae882ba38e7737, frame 26081)
[hos Presentation Thread] vk:   running: marker 115906 (draw of program d5ae882ba38e7737, frame 26081)
[hos Presentation Thread] vk:   running: marker 115907 (dispatch of program 2bcc4174a654992f, frame 26081)
[hos Presentation Thread] vk:   running: marker 115908 (draw of program 2bcc4174a654992f, frame 26081)
[hos Presentation Thread] vk:   running: marker 115909 (dispatch of program 2bcc4174a654992f, frame 26081)
[hos Presentation Thread] vk:   running: marker 115910 (draw of program 2bcc4174a654992f, frame 26081)
[hos Presentation Thread] vk:   running: marker 115911 (dispatch of program 2bcc4174a654992f, frame 26081)
[hos Presentation Thread] vk:   running: marker 115912 (draw of program 2bcc4174a654992f, frame 26081)
From a DM to the bot on Discord from `kenokeefe`: The reporter reran with `RX_VK_CHECKOINTS=1` and provided checkpoint output. Note the variable as typed is `CHECKOINTS` (not `CHECKPOINTS`); no confirmation was provided whether that is a typo or the setting was recognized. Output: ```text [hos Presentation Thread] vk: batched submit 113543 failed (-4) [hos Presentation Thread] vk: device lost: reading the buffer markers [hos Presentation Thread] vk: device lost: 116756 markers set, 64511 reached; the last whose earlier work all finished: 115904 [hos Presentation Thread] vk: finished before: marker 115898 (upload of texture 00000010ce9cea00, frame 26081) [hos Presentation Thread] vk: finished before: marker 115899 (upload of texture 00000010ce9cea00, frame 26081) [hos Presentation Thread] vk: finished before: marker 115900 (upload of texture 00000010ce9cea00, frame 26081) [hos Presentation Thread] vk: finished before: marker 115901 (dispatch of program d5ae882ba38e7737, frame 26081) [hos Presentation Thread] vk: finished before: marker 115902 (draw of program d5ae882ba38e7737, frame 26081) [hos Presentation Thread] vk: finished before: marker 115903 (upload of texture 00000010ce4e5a00, frame 26081) [hos Presentation Thread] vk: RUNNING after: marker 115904 (upload of texture 00000010ce4e5a00, frame 26081) [hos Presentation Thread] vk: running: marker 115905 (dispatch of program d5ae882ba38e7737, frame 26081) [hos Presentation Thread] vk: running: marker 115906 (draw of program d5ae882ba38e7737, frame 26081) [hos Presentation Thread] vk: running: marker 115907 (dispatch of program 2bcc4174a654992f, frame 26081) [hos Presentation Thread] vk: running: marker 115908 (draw of program 2bcc4174a654992f, frame 26081) [hos Presentation Thread] vk: running: marker 115909 (dispatch of program 2bcc4174a654992f, frame 26081) [hos Presentation Thread] vk: running: marker 115910 (draw of program 2bcc4174a654992f, frame 26081) [hos Presentation Thread] vk: running: marker 115911 (dispatch of program 2bcc4174a654992f, frame 26081) [hos Presentation Thread] vk: running: marker 115912 (draw of program 2bcc4174a654992f, frame 26081) ```
Author
Member

From a DM to the bot on Discord from kenokeefe:

Reporter supplied the exact longshot_env.txt contents for the latest run: RX_VK_MESH_SHADERS=compute, RX_VK_MESH_NOCULL=2c56dc1006942f74, and RX_VK_CHECKPOINTS=1. This confirms the checkpoint setting is correctly spelled in that run; the prior run had the typo RX_VK_CHECKOINTS.

From a DM to the bot on Discord from `kenokeefe`: Reporter supplied the exact `longshot_env.txt` contents for the latest run: `RX_VK_MESH_SHADERS=compute`, `RX_VK_MESH_NOCULL=2c56dc1006942f74`, and `RX_VK_CHECKPOINTS=1`. This confirms the checkpoint setting is correctly spelled in that run; the prior run had the typo `RX_VK_CHECKOINTS`.
Sign in to join this conversation.
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
MagicBeanDev/Longshot#1737
No description provided.