16:09 karolherbst: if the kernel rejects a command submission, but nothing gets printed in dmesg, what's the best way to debug this and figure out what went wrong?
16:22 soreau: karolherbst: probably dumb questions for you since I'm just throwing spaghetti, but is the problem reproducible? did you crank up the kernel/drm debug log level? maybe there's some AMD_DEBUG option that can help?
16:22 karolherbst: it is reproducible, but haven't tried cranking up the drm debug log level
16:23 karolherbst: what's the level I should set so it's not flodding my logs tho
16:23 soreau: TBD
17:05 karolherbst: anything in here looking like an error? https://gist.githubusercontent.com/karolherbst/0a0b484de21d46d076c4550e8749706f/raw/4c7a092f64303d6e148bf5c11b2f2fbb873461e2/gistfile1.txt
17:05 karolherbst: all I see is "radv/amdgpu: The CS has been rejected, see dmesg for more information (-2)."
17:05 karolherbst: (in userspace)
17:05 karolherbst: ohh "amdgpu 0000:03:00.0: [drm:drm_ioctl] comm="test_vulkan", pid=824978, ret=-2"
17:05 karolherbst: doesn't tell me much tho
17:18 soreau: does -2 mean errno is 'no such file or directory'?
17:20 karolherbst: ENOENT
17:20 karolherbst: I think it's also thrown if a listed bo isn't valid anymore?
17:20 karolherbst: dunno tbh
17:20 karolherbst: amdgpu has quite a lot of places where it's returned
17:22 karolherbst: "RADV_DEBUG=allbos" "fixes" it...
17:22 soreau: karolherbst: any relevancy? https://gitlab.freedesktop.org/mesa/mesa/-/issues/13273
17:22 karolherbst: no
17:24 soreau: hm, maybe file an issue with what you've found so far
17:26 karolherbst: I doubt it's a radv bug tbh
17:26 karolherbst: there are also tons of VVL errors in the test
17:26 karolherbst: probably should fix them first
17:28 soreau: heh, prolly
17:34 _ds__: It's -ENOENT, yes.
17:34 _ds__: (that -2)