Bug 31547 - [bisected piketon]some cairo test regressed
Summary: [bisected piketon]some cairo test regressed
Status: VERIFIED FIXED
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: git
Hardware: All Linux (All)
: medium normal
Assignee: Chris Wilson
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2010-11-11 00:54 UTC by fangxun
Modified: 2010-12-05 22:38 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:


Attachments

Description fangxun 2010-11-11 00:54:35 UTC
System Environment:
--------------------------
Arch    x86_64
Platform    piketon
Libdrm:  (master)2.4.22-14-g877b2ce15b80975b4dac42657bdfb0a3da833e1c
Mesa:   (master)5b8ed2f6d2e14303c55aeca65cdbc20791e820f6
Xserver:  (server-1.9-branch)xorg-server-1.9.2-7-g46314e1e7ad05d6ff6a2f722b09a76f2931db7f5
Xf86_video_intel: (master)2.13.901-2-g3c5b1399e29ef577b8b91655b5e1c215d1b6dfbb
Kernel: (drm-intel-fixes)69669455b049c0f1f04bb306625c5d4db6838b11


Bug detailed description:
-------------------------
cairo test operator-source, clear-source, device-offset-fractional, filter-nearest-offset, pixman-rotate and xcomposite-projection regresses on piketon.
Bisect shows 3c5b1399e29ef577b8b91655b5e1c215d1b6dfbb is first bad commit.
commit 3c5b1399e29ef577b8b91655b5e1c215d1b6dfbb
Author:     Chris Wilson <chris@chris-wilson.co.uk>
AuthorDate: Tue Nov 9 20:20:06 2010 +0000
Commit:     Chris Wilson <chris@chris-wilson.co.uk>
CommitDate: Tue Nov 9 20:20:06 2010 +0000

    i915: Disable maximum state addresses

    As the kernel controls the relocation of state buffers, we should not
    hard code the maximum permissible value for them.

    Fixes an eventual hang with full-gtt.

    Reported-by: Peter Clifton <pcjc2@cam.ac.uk>
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Comment 1 Chris Wilson 2010-11-11 01:11:25 UTC
Interesting. Maximum GTT address on that branch is 0x10000000 so there should have been no functional change. :(

Are the failures 100% reproducible? i.e. something like

 xinit &
 CAIRO_TEST_TARGET=xlib DISPLAyY=:0 ./cairo-test-suite xcomposite-projection

is sufficient to reproduce?
Comment 2 fangxun 2010-11-11 01:41:34 UTC
Yes, it is sufficient to reproduce this.

Output on screen:
[root@x-pk1 test]#CAIRO_TEST_TARGET=xlib ./cairo-test-suite xcomposite-projection

TESTING cairo-test-suite
Compiled against cairo 1.11.1, running on 1.9.7.
Compiled against pixman 0.19.5, running on 0.21.1.

TESTING xcomposite-projection
xcomposite-projection.xlib.argb32 [0]:  FAIL
xcomposite-projection.xlib.rgb24 [0]:   FAIL
xcomposite-projection.xlib-fallback.rgb24 [0]:  PASS
xcomposite-projection: FAIL (xlib)
0 Passed, 1 Failed [0 crashed, 0 expected], 0 Skipped
xlib (argb32): 1 failed - xcomposite-projection
xlib (rgb24): 1 failed - xcomposite-projection
Comment 3 Chris Wilson 2010-12-03 07:12:16 UTC
Works for me :(

I wonder if it was related to:

commit de18a29e0fa3904894b4e02fae0e712cd43f740c
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Sat Nov 27 22:30:41 2010 +0100

    drm/i915: fix regression due to ba3d8d749b01548b9
    
    We don't track gpu flush request in any special way. So even with
    obj->write_domain == 0, a gpu flush might be outstanding but no
    yet executed. Even worse, the latest request might use the object
    only for reading. So and unconditional call to object_wait_rendering
    is needed for !pipelined.
    
    Hence revert that patch fully and untangle the flushing from the
    synchronization again.
    
    Reported-by: Keith Packard <keithp@keithp.com>
    Tested-by: Keith Packard <keithp@keithp.com>
    Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Comment 4 Chris Wilson 2010-12-05 03:47:59 UTC
I've been unable to reproduce, so I am assuming that the flush fix is good.
Comment 5 Chris Wilson 2010-12-05 03:48:35 UTC
(Not that I understand how it relates to the bisected culprit.)
Comment 6 fangxun 2010-12-05 22:38:01 UTC
It works well with current code. Verified.


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.