<html>
<head>
<base href="https://bugs.freedesktop.org/">
</head>
<body>
<p>
<div>
<b><a class="bz_bug_link
bz_status_NEW "
title="NEW - amdgpu 0000:07:00.0: swiotlb buffer is full (sz: 2097152 bytes)"
href="https://bugs.freedesktop.org/show_bug.cgi?id=104082#c25">Comment # 25</a>
on <a class="bz_bug_link
bz_status_NEW "
title="NEW - amdgpu 0000:07:00.0: swiotlb buffer is full (sz: 2097152 bytes)"
href="https://bugs.freedesktop.org/show_bug.cgi?id=104082">bug 104082</a>
from <span class="vcard"><a class="email" href="mailto:ckoenig.leichtzumerken@gmail.com" title="Christian König <ckoenig.leichtzumerken@gmail.com>"> <span class="fn">Christian König</span></a>
</span></b>
<pre>(In reply to Andreas Kilgus from <a href="show_bug.cgi?id=104082#c24">comment #24</a>)
<span class="quote">> (In reply to Christian König from <a href="show_bug.cgi?id=104082#c23">comment #23</a>)
> > Take a look at /proc/pagetypeinfo. There must at least be pages at order 9
> > or higher for an 2MB swiotlb allocation to succeed.
>
> OK, I had a look and if I got it right: memory fragmentation lets the malloc
> fail. <nitpick>I still would not call that to "run out of memory" - there is
> enough memory available, its current layout is just not in the desired
> shape.</nitpick> ;)</span >
Yeah, correct.
One interesting point is that this only seems to happen when DC is enabled, so
that points to a possible memory leak there.
Would be interesting to monitor amdgpu_gem_info if there are an increasing
number of buffers around when DC is enabled.</pre>
</div>
</p>
<hr>
<span>You are receiving this mail because:</span>
<ul>
<li>You are the assignee for the bug.</li>
</ul>
</body>
</html>