[Bug 91704] Tonga startx fails since mesa/drm 56d8dd6 amdgpu: make vamgr per device v2

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=91704

--- Comment #12 from Mathias Tillman <master.homer at gmail.com> ---
Created attachment 117870
  --> https://bugs.freedesktop.org/attachment.cgi?id=117870&action=edit
Debug output

I've done some testing by adding debug output in amdgpu_device.c and while
testing your va_test does indeed work, I think I have found the reason as to
why startx does not work. Basically, the hash generation is broken, if you
check util_hash_table_get in the logs, you can see that it returns null (in the
startx output), even when the fd exists in the hash table. This seems to be
because drmGetPrimaryDeviceNameFromFd in fd_hash returns the directory name of
the dri device, as opposed to the dri device itself (/dev/dri/ vs
/dev/dri/card0) - you can see that in the log at lines 14 and 25. Strangely it
works just fine the first time amdgpu_device_initialize is called (line 4), and
in your va_test.

To explain the log a bit more, the amdgpu_callback output is a foreach of the
fd_tab hash table right before the util_hash_table_get call. The rest should be
self-explanatory I think.

I will see if I can find out why drmGetPrimaryDeviceNameFromFd returns the
directory name, so if you have any ideas, please let me know.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/dri-devel/attachments/20150823/02058882/attachment.html>