[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

--- Comment #13 from Eero Tamminen <eero.t.tamminen at intel.com> ---
(In reply to Chris Wilson from comment #1)
> The oom may be symptomatic of the same failure, if it is struggling to
> allocate memory, then it is entirely possible for the machine to be starved
> an no further progress made under the mutex for a long time.

(In reply to Mark Janes from comment #5)
> /proc/meminfo from hung system:
> 
> MemTotal:        7848548 kB
> MemFree:          328460 kB
> MemAvailable:    6171272 kB
...
> SwapTotal:       8068092 kB
> SwapFree:        7879420 kB

Memory usage at this point doesn't seem to be problem, but I think that's
because the OOM kill causing program may have already gone.  You would need to
track memory usage before the OOM.

Btw. for i915, it's relevant to track also driver memory as that's not really
visible in meminfo until other things start getting pushed to swap:
  sudo watch cat /sys/kernel/debug/dri/0/i915_gem_objects


> VmallocTotal:   34359738367 kB

34 TB allocated on machine with 8 GB of RAM.  That's a lot of (currently)
unused memory...

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20181112/2cd81830/attachment.html>