[Bug 109385] [CI][DRMTIP] igt at perf_pmu@* - dmesg-warn / incomplete - WARNING: possible circular locking dependency detected

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=109385

--- Comment #2 from Chris Wilson <chris at chris-wilson.co.uk> ---
This is close to being a WONTFIX as for us it is nigh on impossible to fix.
perf allocates from inside per-cpu smpboot callback, tainting the cpuhp mutex
used for stop-machine. We need to use stop-machine from any context to prevent
the HW issue... Or we need to do something like revoke all mmaps before any
GGTT update (which sounds even worse than stop-machine). Even breaking away
from struct_mutex won't help here as the same chain will be created for
ggtt->vm.mutex (both required from inside i915_gem_fault and then taking
stop-machine for the ggtt insertion).

So the most realistic coarse of action would be to avoid the same perf lock
being required inside cpuhp and around the perf_event.

-- 
You are receiving this mail because:
You are the assignee for the bug.
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/20190118/74751f58/attachment.html>