[Intel-gfx] bo problem, file-max limit reached within two days

[Knut Petersen]

Am 12.04.2012 10:03, schrieb Daniel Vetter:
> Can you try the drm-intel-next-queued branch from my git repo at http://cgit.freedesktop.org/~danvet/drm-intel/ That contains a patch from Chris Wilson to mitigate mmio offset exhaustion, one possible reason for a -ENOSPC failure. -Daniel 

It´s a short time for a test, but drm-intel-next-queued does not look better than other kernels:

date;  cat /proc/sys/fs/file-nr;  cat /sys/kernel/debug/dri/0/i915_gem_objects

Do 12. Apr 13:21:50 CEST 2012
9323    0       204854
1275 objects, 81797120 bytes
1251 [1251] objects, 84410368 [84410368] bytes in gtt
   19 [19] active objects, 11051008 [11051008] bytes
   2 [2] pinned objects, 5373952 [5373952] bytes
   1230 [1230] inactive objects, 67985408 [67985408] bytes
   0 [0] freed objects, 0 [0] bytes
3 pinned mappable objects, 13762560 bytes
786 fault mappable objects, 16318464 bytes
268435456 [268435456] gtt total

Do 12. Apr 14:09:41 CEST 2012
12387   0       204854
4317 objects, 82264064 bytes
4282 [4282] objects, 84713472 [84713472] bytes in gtt
   15 [15] active objects, 11304960 [11304960] bytes
   2 [2] pinned objects, 5373952 [5373952] bytes
   4265 [4265] inactive objects, 68034560 [68034560] bytes
   0 [0] freed objects, 0 [0] bytes
3 pinned mappable objects, 13762560 bytes
3809 fault mappable objects, 27975680 bytes
268435456 [268435456] gtt total

Do 12. Apr 14:47:18 CEST 2012
15616   0       204854
7015 objects, 148389888 bytes
905 [905] objects, 73834496 [73834496] bytes in gtt
   15 [15] active objects, 11304960 [11304960] bytes
   2 [2] pinned objects, 5373952 [5373952] bytes
   888 [888] inactive objects, 57155584 [57155584] bytes
   0 [0] freed objects, 0 [0] bytes
3 pinned mappable objects, 13762560 bytes
699 fault mappable objects, 13668352 bytes
268435456 [268435456] gtt total


Files, total number of objects, inactive objects and fault mappable objects raise at a rate comparable
to kernels 3.3.1, 3.2.* and 3.1*.

I searched about 15 months of log files - both error messages appeared for the first time about 14 days ago.
Kernel 3.2 is much older, and the system survived _much_ longer uptimes previously. Either it´s not a kernel problem,
or that kernel problem is exposed by some recent (max 6 weeks) change in Xorg, KDE, ... etc. Unfortunately there were
a lot of changes in that time.

cu,
  Knut