<html>
<head>
<base href="https://bugs.freedesktop.org/">
</head>
<body>
<p>
<div>
<b><a class="bz_bug_link
bz_status_NEEDINFO "
title="NEEDINFO - i915 slab shrink cause a panic"
href="https://bugs.freedesktop.org/show_bug.cgi?id=109005#c7">Comment # 7</a>
on <a class="bz_bug_link
bz_status_NEEDINFO "
title="NEEDINFO - i915 slab shrink cause a panic"
href="https://bugs.freedesktop.org/show_bug.cgi?id=109005">bug 109005</a>
from <span class="vcard"><a class="email" href="mailto:bin.yang@intel.com" title="Yang Bin <bin.yang@intel.com>"> <span class="fn">Yang Bin</span></a>
</span></b>
<pre>I had reproduced this issue successfully and found the root cause as below.
i915_gem_wait_for_idle() waits for all requests being completed and
calls i915_retire_requests() to retire them. It assumes the
active_requests should be zero finally.
In i915_retire_requests(), it will retire all requests on the active
rings. Unfortunately, active_requests is increased in
i915_request_alloc() and reduced in i915_request_retire(), but the
request is added into active rings in i915_request_add().
If i915_gem_wait_for_idle() is called between i915_request_alloc()
and i915_request_add(), this request will not be retired. Then, the
active_requests will not be zero in the end.
Normally, i915_request_alloc() and i915_request_add() will be called
in sequence with drm.struct_mutex locked. But in
intel_vgpu_create_workload(), it will pre-allocate the request and
call i915_request_add() in the workload thread for performance
optimization. The above issue will be triggered.
I had submitted a RFC patch and continue to discuss it in maillist. Please
refer to <a href="https://lkml.org/lkml/2018/12/20/78">https://lkml.org/lkml/2018/12/20/78</a></pre>
</div>
</p>
<hr>
<span>You are receiving this mail because:</span>
<ul>
<li>You are on the CC list for the bug.</li>
<li>You are the QA Contact for the bug.</li>
<li>You are the assignee for the bug.</li>
</ul>
</body>
</html>