On Tue, Nov 07, 2023 at 09:03:59AM +0000, Laurence Tratt wrote:
> Across a few Intel machines I've had, I've often seen warnings about
> "potential atomic update failures". On my current desktop with an
> i5-13600K I've now seen actual failures:
>
> drm:pid51847:intel_pipe_update_start *ERROR* [drm] *ERROR* Potential atomic
> update failure on pipe A
> drm:pid51847:intel_pipe_update_end *ERROR* [drm] *ERROR* Atomic update
> failure on pipe A (start=84142 end=84143) time 3 us, min 2544, max 2559,
> scanline start 2555, end 2560
>
> I think these correlate with occasional roughly-second-long locks in X.
In the last couple of days (with a Nov 17th snapshot) I've twice
experienced an interesting variation on this: the machine became almost
unusable for about 10s, with the mouse lurching slowly around the screen
like someone at the end of a long Friday night out. After today's version I
ended up with this in my dmesg:
Asynchronous wait on fence :Xorg[79196]:498526 timed out
(hint:0xffffffff81b8c2c0s)
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
i915_vma_coredump_create: stub
pool_fini: stub
drm:pid13130:__intel_engine_reset_bh *NOTICE* [drm] Resetting rcs0 for
stopped heartbeat on rcs0
drm:pid13130:gen8_engine_reset_prepare *ERROR* [drm] *ERROR* rcs0 reset
request timed out: {request: 00000001, RESET_CTL: 00000001}
drm:pid13130:intel_gt_reset *NOTICE* [drm] Resetting chip for stopped
heartbeat on rcs0
drm:pid13130:gen8_engine_reset_prepare *ERROR* [drm] *ERROR* rcs0 reset
request timed out: {request: 00000001, RESET_CTL: 00000001}
drm:pid13130:gen8_engine_reset_prepare *ERROR* [drm] *ERROR* rcs0 reset
request timed out: {request: 00000001, RESET_CTL: 00000001}
drm:pid13130:mark_guilty *NOTICE* [drm] firefox[92248] context reset due to
GPU hang
As that last line suggests, I was using Firefox at the time.
Laurie
