[Bug 108456] userptr deadlock in Vulkan CTS: mmu_notifier vs workers vs struct_mutex

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=108456

Mark Janes <mark.a.janes at intel.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |francesco.balestrieri at intel
                   |                            |.com

--- Comment #20 from Mark Janes <mark.a.janes at intel.com> ---
(In reply to Francesco Balestrieri from comment #18)
> https://patchwork.freedesktop.org/patch/258901/ 
> 
> Did it help?

I ran vulkancts on this system, and found that found that it encountered GPU
hang on at least 25% of vulkancts runs.  Usually, this test crashed together
with the GPU Hang.

dEQP-VK.memory.pipeline_barrier.host_write_uniform_buffer.1048576


The GPU hang was unrecoverable.  The system locked up and required a manual
power off.  I was able to catch the dmesg log:

[ 1166.700509] [drm] GPU HANG: ecode 9:0:0x84d5fffb, in deqp-vk [18358],
reason: no progress on rcs0, action: reset
[ 1166.700510] [drm] GPU hangs can indicate a bug anywhere in the entire gfx
stack, including userspace.
[ 1166.700510] [drm] Please file a _new_ bug report on bugs.freedesktop.org
against DRI -> DRM/Intel
[ 1166.700511] [drm] drm/i915 developers can then reassign to the right
component if it's not a kernel issue.
[ 1166.700511] [drm] The gpu crash dump is required to analyze gpu hangs, so
please always attach it.
[ 1166.700512] [drm] GPU crash dump saved to /sys/class/drm/card0/error
[ 1166.700591] i915 0000:00:02.0: Resetting rcs0 for no progress on rcs0


Because the system was hard-hung, I could not retrieve the error state.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20181203/0ea6e6b2/attachment-0001.html>