On 11/28/2023 9:51 PM, Mario Limonciello wrote:
Hi,
In amd-staging-drm-next 46fe6312082c ("drm/amdgpu: update retry times
for psp BL wait") and upstream a11156ff6f41 ("drm/amdgpu: update retry
times for psp BL wait") the number of loops for
psp_v13_0_wait_for_bootloader() to try again increased significantly.
It went from 10 loops to 20k loops. Essentially this means that the
function can "effectively" no longer fail.
PSP_VMBX_POLLING_LIMIT to 20k is introduced by this - f2328c2ba0e84
("drm/amdgpu: update retry times for psp vmbx wait")
20k is too much even for PSP 13.0.6. Will reduce it to 3000 (~5mins) for
13.0.6 and for others keep the default 10.
Thanks,
Lijo
I've got an issue I'm looking at where runtime resume for a dGPU fails,
and because of this change the system gets stuck in a never ending busy
loop instead of cleanly returning an error code to the caller. The
outcome is the system appears hung while the 20k loops run instead of
just the dGPU failing to resume.
Is this 20k value really required? Or can we reduce it back to
something more manageable?
Thanks,
