<html>
<head>
<meta content="text/html; charset=utf-8" http-equiv="Content-Type">
</head>
<body text="#000000" bgcolor="#FFFFFF">
<div class="moz-cite-prefix">Hi Nocolai,<br>
<br>
If we don't already have an option for this try to double the size
of the VM area allocate for each BO in userspace.<br>
<br>
That should give you a nice hole between each BO and so should
help to catch cases when somebody writes over the end of a BO.<br>
<br>
Regards,<br>
Christian.<br>
<br>
Am 22.06.2016 um 09:50 schrieb Nicolai Hähnle:<br>
</div>
<blockquote cite="mid:576A4332.1090409@gmail.com" type="cite">Hi
Mads,
<br>
<br>
setting R600_DEBUG=nodma in the X server should work around your
problem for now.
<br>
<br>
Marek, perhaps an out-of-bounds check for tiled texture memory
access similar to the linear access check is necessary? I wonder
if you've seen something about that in the docs.
<br>
<br>
I've annotated the sDMA IB dump. It's a linear-to-display-tiled
copy on Carrizo. I tried to reproduce with the attached patch, but
failed to do so even with amdgpu.vm_debug=1. With the patch, I get
DMA copies that are identical to the one that causes the VM fault
except for a different bank_height and macro_tile_aspect, so the
issue is likely related to those.
<br>
<br>
Nicolai
<br>
<br>
On 21.06.2016 19:32, Nicolai Hähnle wrote:
<br>
<blockquote type="cite">On 21.06.2016 19:16, Mads wrote:
<br>
<blockquote type="cite">I sent this for 1.5 hours ago, but since
it hasn't arrived to the
<br>
mailing list yet, I try again...
<br>
</blockquote>
<br>
It arrived, no worries :)
<br>
<br>
I'll take a look later.
<br>
<br>
Nicolai
<br>
<br>
<blockquote type="cite">
<br>
On 2016-06-21 17:48, Mads wrote:
<br>
<br>
<blockquote type="cite">On 2016-06-21 10:12, Mads wrote:
<br>
<br>
On 2016-06-21 09:39, Nicolai Hähnle wrote:
<br>
<br>
Thanks. However, I still don't think this is going to help.
Your
<br>
earlier trace experiments showed that the problematic SDMA
commands
<br>
came from the X server, _not_ from plasmashell.
<br>
<br>
So what we see here is likely just the first set of GPU
commands sent
<br>
by plasmashell after the VM fault occurred. Since the
plasmashell
<br>
process is unable to tell who caused the VM fault, it takes
the blame
<br>
incorrectly. Are you sure the X server is using your
self-compiled
<br>
radeonsi_dri.so and has the environment variable set? If it
creates a
<br>
ddebug_dump, it might be somewhere else (it's based off the
HOME
<br>
environment variable, which may be different).
<br>
I'll take a second look to see if there's an X dump there
too, but
<br>
unfortunately it'll be in about ~8 hours before I have the
machine at
<br>
hand again..
<br>
<br>
And yes, I'm sure, everything is built through portage, so
there is no
<br>
"self-compiled" on the system per se. There's always just
one lib
<br>
available at any time :)
<br>
</blockquote>
<br>
You were right! X didn't have R600_DEBUG=check_vm in
environment (no
<br>
login shell/sourcing of /etc/profile).
<br>
<br>
Here's what i ran:
<br>
<br>
<blockquote type="cite">$ XAUTHORITY=.Xauthority DISPLAY=:0
LIBGL_DEBUG=verbose dolphin
<br>
libGL: pci id for fd 9: 1002:9874, driver radeonsi
<br>
libGL: OpenDriver: trying /usr/lib64/dri/tls/radeonsi_dri.so
<br>
libGL: OpenDriver: trying /usr/lib64/dri/radeonsi_dri.so
<br>
si_vm_fault_occured: failed to parse line '
Either
<br>
enable ECC checking or force module loading by setting
<br>
'ecc_enable_override'.
<br>
'
<br>
libGL: Using DRI3 for screen 0
<br>
Trying to convert empty KLocalizedString to QString.
<br>
Cannot creat accessible child interface for object:
<br>
PlacesView(0x118d670) index: 5
<br>
QPixmap::scaled: Pixmap is a null pixmap
<br>
QPixmap::scaled: Pixmap is a null pixmap
<br>
(... etc ...)
<br>
The X11 connection broke (error 1). Did the X11 server die?
<br>
</blockquote>
<br>
Attaching dmesg and ddebug_dump.
<br>
<br>
- Mads
<br>
</blockquote>
</blockquote>
<br>
<br>
<fieldset class="mimeAttachmentHeader"></fieldset>
<br>
<pre wrap="">_______________________________________________
amd-gfx mailing list
<a class="moz-txt-link-abbreviated" href="mailto:amd-gfx@lists.freedesktop.org">amd-gfx@lists.freedesktop.org</a>
<a class="moz-txt-link-freetext" href="https://lists.freedesktop.org/mailman/listinfo/amd-gfx">https://lists.freedesktop.org/mailman/listinfo/amd-gfx</a>
</pre>
</blockquote>
<br>
</body>
</html>