Bug 89603 - [SKL]gem_userptr_blits/coherency-sync takes more than 10 minutes
Summary: [SKL]gem_userptr_blits/coherency-sync takes more than 10 minutes
Status: CLOSED WORKSFORME
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Intel (show other bugs)
Version: unspecified
Hardware: All Linux (All)
: medium normal
Assignee: Rami
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2015-03-17 07:35 UTC by lu hua
Modified: 2016-02-25 17:01 UTC (History)
2 users (show)

See Also:
i915 platform: SKL
i915 features: GEM/Other


Attachments
dmesg (119.23 KB, text/plain)
2015-03-17 07:35 UTC, lu hua
no flags Details

Description lu hua 2015-03-17 07:35:30 UTC
Created attachment 114374 [details]
dmesg

==System Environment==
--------------------------
Regression:  not sure, always fails on SKL

no-working platforms: SKL

==kernel==
--------------------------
drm-intel-nightly/4f7ab66fd78ed940a5fcfcaad7e9f1bf8b1abe76
commit 4f7ab66fd78ed940a5fcfcaad7e9f1bf8b1abe76
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Mon Mar 16 18:51:08 2015 +0100

    drm-intel-nightly: 2015y-03m-16d-17h-50m-27s UTC integration manifest


==Bug detailed description==
-----------------------------
It always has this issue on SKL with drm-intel-nightly kernel.
gem_userptr_blits/coherency-unsync also has this issue.

output:
IGT-Version: 1.10-g505645c (x86_64) (Linux: 4.0.0-rc4_drm-intel-nightly_4f7ab6_20150317+ x86_64)
Aperture size is 4096 MiB
Total RAM is 3613 MiB
Not enough RAM to run test, reducing buffer count.
Testing unsynchronized mappings...
Received signal SIGTERM.
Testing synchronized mappings...
Using 2x2709 1MiB buffers
Verifying initialisation...
Cyclic blits cpu->gpu, forward...
Cyclic blits gpu->cpu, backward...
Random blits...
Subtest coherency-sync: SUCCESS (844.768s)
Received signal SIGTERM.

real    14m5.006s
user    0m10.787s
sys     0m58.486s

==Reproduce steps==
---------------------------- 
1.  time ./gem_userptr_blits --run-subtest coherency-sync
Comment 1 Jesse Barnes 2015-03-18 21:22:34 UTC
This is probably a general problem with the blit tests since they try to allocate more memory than the aperture size.  If we had a way to pin a large buffer in the GTT and then use the usable space to determine the allocation size instead, the test would probably run a lot faster...  As it is, I think it's swapping the machine to death, at least here.
Comment 2 Chris Wilson 2015-03-18 21:40:28 UTC
Here, the approach should be similar to that taken with the recent fixes to gem_tiled_blits: do a basic test for simple coherency tests, then a heavier test for thrashing with an intel_require_memory() to skip on machines with insufficient memory to run the test sanely.
Comment 3 cprigent 2015-08-03 22:33:16 UTC
Bug scrub: assigned to Rami
Rami, please check if it is still reproduced.
Comment 4 Rami 2015-08-24 15:46:17 UTC
Tested on skl:

System Environment:
===================
Linux : Ubuntu 14.04 LTS 64 bits
Kernel : 
BIOS : SKLSE2R1.R00.X093.B02.1507222151
ME FW : 11.0.0.1163
Ksc (EC FW): 1.16

Software: 
=========
Software 
Linux : Ubuntu 14.04 LTS 64 bits
Kernel : 
BIOS : SKLSE2R1.R00.X093.B02.1507222151
ME FW : 11.0.0.1163
Ksc (EC FW): 1.16

Kernel: drm-intel-nightly 30263ef782e0548c114fb6f2771b58b27c56ba0d 4.2.0-rc7 from git://anongit.freedesktop.org/drm-intel
Mesa: mesa-10.6.3 ddc976368fef367e464472ebcc2ac4fd89eb9fd8 from http://cgit.freedesktop.org/mesa/mesa/ 
Xf86_video_intel: 2.99.917 baec802b21387d04aebb10ac29e719a1800c5aa0 from http://cgit.freedesktop.org/xorg/driver/xf86-video-intel/ 
Libdrm: libdrm-2.4.62 ba4b5ac010ab85406ec52e3906e13d58cd9aa782 from http://cgit.freedesktop.org/mesa/drm/ 
Cairo: 1.14.2 93422b3cb5e0ef8104b8194c8873124ce2f5ea2d from http://cgit.freedesktop.org/cairo 
libva: libva-1.6.0 a8008998bc0d4a76ae6927607c048e52ba50fd0e from http://cgit.freedesktop.org/libva/ 
intel-driver: 1.6.0 32268c46d538667d437dc9266aa4c183e51c1286 from http://cgit.freedesktop.org/vaapi/intel-driver 
Xserver: xorg-server-1.17.2 2123f7682d522619f101b05fb75efa75dabbe371 from http://cgit.freedesktop.org/xorg/xserver

Actual Result:
==============
IGT-Version: 1.11-g0e25e5d (x86_64) (Linux: 4.2.0-rc8-drm-intel-nightly-ww35-patch-bug91210+ x86_64)
Aperture size is 4096 MiB
Total RAM is 3836 MiB
Testing unsynchronized mappings...
Testing synchronized mappings...
Using 2x2730 1MiB buffers
Verifying initialisation...
Cyclic blits cpu->gpu, forward...
Cyclic blits gpu->cpu, backward...
Random blits...
Subtest coherency-sync: SUCCESS (394.632s)

real	6m34.832s
user	0m11.205s
sys	1m30.234s
Comment 5 Chris Wilson 2015-08-24 15:49:31 UTC
Hmm.. What happened?
Comment 6 Rami 2015-08-24 16:02:29 UTC
What did you want to know exactly?
Comment 7 cprigent 2016-02-25 17:01:25 UTC
gem_userptr_blits/coherency-sync is Skip in 0.031s.

Platform: SKY LAKE Y A0 QUAL
CPU : Intel(R) Core(TM) m5-6Y57 CPU @ 1.10GHz (family: 6, model: 78 stepping: 3)
MCP : SKL-Y D1 2+2 (ou ULX-D1)
QDF : QJK9
CPU : SKL D0
Chipset PCH: Sunrise Point LP C0
CRB : SKY LAKE Y LPDDR3 RVP3 CRB FAB2
Software
Linux OS: Ubuntu 15.04 64 bits
BIOS : BRAS.X64.B088.R00.1510270350
TXE FW : 2.0.0.2093
Ksc : 1.08
  kernel 4.5.0-rc5 08fc1b1 from from http://cgit.freedesktop.org/drm-intel
  commit 08fc1b101049694778bff7559e1d05250d2e7072
  Author: Ville Syrjälä <ville.syrjala@linux.intel.com>
  Date: Mon Feb 22 19:31:15 2016 +0200
  drm-intel-nightly: 2016y-02m-22d-17h-30m-27s UTC integration manifest
drm: tag libdrm-2.4.66 9b77443 from http://cgit.freedesktop.org/mesa/drm/
IGT 1.13 babcf40 from https://cgit.freedesktop.org/xorg/app/intel-gpu-tools/
One eDP connected
Comment 8 cprigent 2016-02-25 17:01:42 UTC
So closed as not reproducible


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.