https://lists.freedesktop.org/archives/amd-gfx/2025-June/126067.htmlSo this is why AMDGPU can't reset properly?
-
https://lists.freedesktop.org/archives/amd-gfx/2025-June/126067.html
So this is why AMDGPU can't reset properly? Because Digital Restrictions Management hardware built into it intentionally can't be initialized more than once after a power cycle or hard reset, for "security reasons"???? Fuck that shit. #amdgpu #endDRM
-
https://lists.freedesktop.org/archives/amd-gfx/2025-June/126067.html
So this is why AMDGPU can't reset properly? Because Digital Restrictions Management hardware built into it intentionally can't be initialized more than once after a power cycle or hard reset, for "security reasons"???? Fuck that shit. #amdgpu #endDRM
FYI, I specifically meant the reset cases required for kexec, or for mapping a GPU into a VM and back again. This does not apply to user space GPU hang recovery. Sorry if I implied that.
-
FYI, I specifically meant the reset cases required for kexec, or for mapping a GPU into a VM and back again. This does not apply to user space GPU hang recovery. Sorry if I implied that.
@chris we are finding our selves having to drain hypervisors with AMD MI200s running VMs and K8s nodes every now and then, which is a pain. Could this be related? Can't believe it... 🫣
-
undefined fosstodon.org ha condiviso questa discussione