| Summary: | xorg resource leak and crash in Intel Ivybridge | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Product: | DRI | Reporter: | Nikolay <mar.kolya> | ||||||
| Component: | DRM/Intel | Assignee: | Intel GFX Bugs mailing list <intel-gfx-bugs> | ||||||
| Status: | CLOSED FIXED | QA Contact: | Intel GFX Bugs mailing list <intel-gfx-bugs> | ||||||
| Severity: | normal | ||||||||
| Priority: | medium | CC: | David.Biesack, intel-gfx-bugs, nemesis | ||||||
| Version: | unspecified | ||||||||
| Hardware: | x86-64 (AMD64) | ||||||||
| OS: | Linux (All) | ||||||||
| Whiteboard: | |||||||||
| i915 platform: | i915 features: | ||||||||
| Attachments: |
|
||||||||
Created attachment 119582 [details]
crash stack trace
commit 7490b9ec263b87b3669096579ec0f0066ec328cb Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Thu Nov 12 09:17:17 2015 +0000 sna: Wait upon the same ring when out-of-memory The current out-of-memory allocation code was waiting upon the wrong ring id instead of using the index - causing an issue if forced to wait upon the BLT ring. If we still cannot allocate, make sure that all caches are dropped. References: https://bugs.freedesktop.org/show_bug.cgi?id=92911 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Can you keep an eye on /sys/kernel/debug/dri/0/i915_gem_objects? Thanks for very quick response! Number of '/drm mm' lines in maps file went down to ~250 and it doesn't seem to grow even with hangout running. $ sudo cat /sys/kernel/debug/dri/0/i915_gem_objects 2616 objects, 1292308480 bytes 438 [33] objects, 865861632 [240902144] bytes in gtt 54 [13] active objects, 407191552 [74022912] bytes 384 [20] inactive objects, 458670080 [166879232] bytes 1192 unbound objects, 319500288 bytes 172 purgeable objects, 12177408 bytes 3 pinned mappable objects, 121995264 bytes 14 fault mappable objects, 10797056 bytes 2147483648 [268435456] gtt total [k]batch pool: 6 objects, 36864 bytes (24576 active, 12288 inactive, 36864 global, 0 shared, 0 unbound) Xorg: 2143 objects, 686604288 bytes (80031744 active, 310317056 inactive, 390348800 global, 534630400 shared, 234741760 unbound) compiz: 382 objects, 535212032 bytes (348803072 active, 103641088 inactive, 452444160 global, 211742720 shared, 21983232 unbound) Compositor: 141 objects, 327192576 bytes (57970688 active, 170000384 inactive, 227971072 global, 45486080 shared, 98963456 unbound) Also, subjectively system is somewhat more responsive. Thanks! commit 2d26643cab33a32847afaf13b50d326d09d58bf7 Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Fri Nov 13 19:03:36 2015 +0000 sna/dri2: Drop the reference on the fence when complete Fixes regression from commit 8d9e496670f48b4eec64dfe1bcedb49793cf3073 Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Wed Jul 22 11:14:01 2015 +0100 sna/dri2: Take over the placeholder vblank After noting the fence was complete, we would clear it. But I forgot that we actually held a reference on to it, and so we would leak the 64k batch, and starve the system of available memory in about 18 minutes of SwapBuffers. Reported-by: Arkadiusz Miskiewicz <arekm@maven.pl> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=92911 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> *** Bug 94092 has been marked as a duplicate of this bug. *** in what release is this fix available? I downloaded the latest release of https://download.01.org/gfx/ubuntu/15.10/main/pool/main/i/intel-linux-graphics-installer/intel-linux-graphics-installer_1.2.1-0intel2_amd64.deb That file has date 2015-11-16 (in the parent directory of the download) Bug 94092 was marked as a dup of this, but I experienced Bug 94092 using that deb file. See https://bugs.freedesktop.org/attachment.cgi?id=121709 -- does that show which version of the driver is running and does it contain this fix? thanks David, this fix was not backported to Wily yet. The relevant bug thread is here: https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/1522727 Please engage the Ubuntu maintainer to get a backport. |
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.
Created attachment 119581 [details] xorg log X session crashes after some use, several times a day under normal desktop usage. I'll attach xorg log I've noticed one more strange thing that is possibly related. Output of this command: cat /proc/{xorg pid}/maps | grep mm | wc -l (this looks for lines like this: 7fd313cc8000-7fd313d09000 rw-s 00000000 00:05 41728 /drm mm object (deleted)) goes to 10's of thousands. It seems to be growing quicker if some videa application is being used (google hangouts). Stopping browser (and hangouts) doesn't seem to release those mmaps. The frequency of the crashing seems to be roughly correlated to time of hangout usage and therefore it looks like crashes and number of mmaps growing without bounds seems related. This all happens with current iobaf build on Ubuntu Trusty (Mint). Rolling back to vivid lts packages seems to be keeping number of mmaped regions to less than 150 and no crashes so far.