<html><head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
</head>
<body>
<p>JingWen - could you maybe give those patches a try on SRIOV XGMI
system ? If you see issues maybe you could let me connect and
debug. My SRIOV XGMI system which Shayun kindly arranged for me is
not loading the driver with my drm-misc-next branch even without
my patches.<br>
</p>
<p>Andrey<br>
</p>
<div class="moz-cite-prefix">On 2022-01-17 14:21, Andrey Grodzovsky
wrote:<br>
</div>
<blockquote type="cite" cite="mid:769b97dd-c6f9-88fe-a26b-34bfd617e257@amd.com">
<p><br>
</p>
<div class="moz-cite-prefix">On 2022-01-17 2:17 p.m., Christian
König wrote:<br>
</div>
<blockquote type="cite" cite="mid:096c0884-7e32-40ed-7570-b65f19104f5f@gmail.com">
Am 17.01.22 um 20:14 schrieb Andrey Grodzovsky:<br>
<blockquote type="cite" cite="mid:62f9f1c2-312b-760e-75f7-e86421333be3@amd.com">
<p>Ping on the question</p>
</blockquote>
<br>
Oh, my! That was already more than a week ago and is completely
swapped out of my head again.<br>
<br>
<blockquote type="cite" cite="mid:62f9f1c2-312b-760e-75f7-e86421333be3@amd.com">
<p>Andrey<br>
</p>
<div class="moz-cite-prefix">On 2022-01-05 1:11 p.m., Andrey
Grodzovsky wrote:<br>
</div>
<blockquote type="cite" cite="mid:c64c933f-498d-a2d9-fe63-058c6f1bed9c@amd.com">
<blockquote type="cite" style="color: #007cff;">
<blockquote type="cite" style="color: #007cff;">Also, what
about having the reset_active or in_reset flag in the
reset_domain itself? <br>
</blockquote>
<br>
Of hand that sounds like a good idea. <br>
</blockquote>
<br>
<br>
What then about the adev->reset_sem semaphore ? Should we
also move this to reset_domain ? Both of the moves have
functional <br>
implications only for XGMI case because there will be
contention over accessing those single instance variables
from multiple devices <br>
while now each device has it's own copy. <br>
</blockquote>
</blockquote>
<br>
Since this is a rw semaphore that should be unproblematic I
think. It could just be that the cache line of the lock then
plays ping/pong between the CPU cores.<br>
<br>
<blockquote type="cite" cite="mid:62f9f1c2-312b-760e-75f7-e86421333be3@amd.com">
<blockquote type="cite" cite="mid:c64c933f-498d-a2d9-fe63-058c6f1bed9c@amd.com"> <br>
What benefit the centralization into reset_domain gives - is
it for example to prevent one device in a hive trying to
access through MMIO another one's <br>
VRAM (shared FB memory) while the other one goes through
reset ? <br>
</blockquote>
</blockquote>
<br>
I think that this is the killer argument for a centralized lock,
yes.<br>
</blockquote>
<p><br>
</p>
<p>np, i will add a patch with centralizing both flag into reset
domain and resend.</p>
<p>Andrey</p>
<p><br>
</p>
<blockquote type="cite" cite="mid:096c0884-7e32-40ed-7570-b65f19104f5f@gmail.com"> <br>
Christian.<br>
<br>
<blockquote type="cite" cite="mid:62f9f1c2-312b-760e-75f7-e86421333be3@amd.com">
<blockquote type="cite" cite="mid:c64c933f-498d-a2d9-fe63-058c6f1bed9c@amd.com"> <br>
Andrey </blockquote>
</blockquote>
<br>
</blockquote>
</blockquote>
</body>
</html>