Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > linux.debian.kernel > #89024 > unrolled thread
| Started by | Henrik Ahlgren <pablo@seestieto.com> |
|---|---|
| First post | 2025-08-28 17:10 +0200 |
| Last post | 2025-08-28 17:10 +0200 |
| Articles | 1 — 1 participant |
Back to article view | Back to linux.debian.kernel
Bug#1112333: linux-image-6.12.41+deb13-amd64: amdgpu crash with darktable/opencl/rocm Henrik Ahlgren <pablo@seestieto.com> - 2025-08-28 17:10 +0200
| From | Henrik Ahlgren <pablo@seestieto.com> |
|---|---|
| Date | 2025-08-28 17:10 +0200 |
| Subject | Bug#1112333: linux-image-6.12.41+deb13-amd64: amdgpu crash with darktable/opencl/rocm |
| Message-ID | <LoMqd-b2H5-1@gated-at.bofh.it> |
[Multipart message — attachments visible in raw view] — view raw
Package: src:linux Version: 6.12.41-1 Severity: important Dear Maintainer, When utilizing darktable photo editing software (5.0.1-2) with GPU acceleration on AMD Ryzen 5 3400G integrated graphics, the amdgpu kernel driver crashes after a period of time. This results in the failure of not only darktable itself but the entire GNOME/Wayland session. A core dump from Xwayland is generated in $HOME. I've installed at least the following trixie packages to enable opencl: librocm-smi-dev librocm-smi64-1 rocm-cmake rocm-device-libs libclang-common-17-dev I can consistently reproduce this issue by utilizing darktable and performing certain operations that induce GPU load, such as generating thumbnails while scrolling through a series of newly imported images. Same thing happens with other ROCm usage scenarios as well, like attempting to run local LLM models with ollama (integrated GPUs are not really supported, but it could work without this crash). -- Package-specific info:
Back to top | Article view | linux.debian.kernel
csiph-web