Bug 67889 - transformed fb may leave the scanout blank
Summary: transformed fb may leave the scanout blank
Status: VERIFIED FIXED
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: unspecified
Hardware: Other Linux (All)
: high critical
Assignee: Chris Wilson
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2013-08-08 07:46 UTC by kobeqin
Modified: 2014-01-02 05:39 UTC (History)
5 users (show)

See Also:
i915 platform:
i915 features:


Attachments
dmesg for crandr_scale(drm.debug=6) (68.11 KB, text/plain)
2013-08-08 07:46 UTC, kobeqin
no flags Details
Xorg.0.log (18.21 KB, text/plain)
2013-08-08 07:49 UTC, kobeqin
no flags Details

Description kobeqin 2013-08-08 07:46:55 UTC
Created attachment 83813 [details]
dmesg for crandr_scale(drm.debug=6)

Description,
Blackscreen will occur after panel scaling via xrandr at the second time. Try to rotate, blackscreen also will occur at the second time.

Reproduced steps,
1. Boot up system with VGA connected
2. $xinit &
3. $xrandr --output VGA1 --scale 5x5
4. $xrandr --output VGA1 --scale 10x10

Or
3. $xrandr --output VGA1 --rotate left
4. $xrandr --output VGA1 --rotate inverted

Test Environment, 

Platform: Baytrail-M 
00:00.0 Host bridge [0600]: Intel Corporation ValleyView SSA-CUnit [8086:0f00] (rev 06)
00:02.0 VGA compatible controller [0300]: Intel Corporation ValleyView Gen7 [8086:0f31] (rev 06)

Kernel:         3.11.0rc3_nightlytop_89ff7e_20130807
Libdrm:         (master)libdrm-2.4.46-26-g3c967e715528ee52195c178c4d09d03b643f0c06
Mesa:           (master)12522041d6e17ac176a8e86c590789bfbbe9def5
Xserver:                (master)xorg-server-1.14.99.1-176-gf67d022523c59a27f3bf8791aa9ca6624318b1fd
Xf86_video_intel:               (master)2.21.14-5-g6004870d53221caf05a330fcfc8a72e68b2b8898
Cairo:          (master)46d9db96d460fea72f0420102e8a90c6a7231f79
Libva:          (staging)bc49c42bcbfb24286c06b3e7cdb171b2a01c6f80
Libva_intel_driver:             (staging)b3afeef8092dc4eb7cb73fce672ddf7a55205f34
Comment 1 kobeqin 2013-08-08 07:49:01 UTC
Created attachment 83814 [details]
Xorg.0.log
Comment 2 Gordon Jin 2013-09-23 08:19:36 UTC
any developer can reproduce this?
Comment 3 shui yangwei 2013-09-23 09:49:35 UTC
On -latest nightly: 

Screen will blackscreen after panel scaling via xrandr at the first time.
xrandr --output VGA1 --scale 10x10

I tried rotate, it works good.
Comment 4 Chris Wilson 2013-09-26 10:33:22 UTC
That would be a 19200x10800 framebuffer. :)
Comment 5 Chris Wilson 2013-09-26 13:44:02 UTC
How much memory do you have on your byt-m? I have 2GiB, and using a 19200x10800 I hit an allocation failure when downsampling to the scanout, hence fill the output with black.
Comment 6 shui yangwei 2013-09-26 14:11:21 UTC
(In reply to comment #5)
> How much memory do you have on your byt-m? I have 2GiB, and using a
> 19200x10800 I hit an allocation failure when downsampling to the scanout,
> hence fill the output with black.

Our machine use mode 16800x10500:
----------------------
VGA1 connected 16800x10500+0+0 (normal left inverted right x axis y axis) 474mm x 296mm
   1680x1050      60.0*+
   1280x1024      75.0     60.0
   1152x864       75.0
   1024x768       75.1     60.0


machine boot up memory info:
-----------------------
[root@x-byt04 ~]# cat /proc/meminfo
MemTotal:        1910716 kB
MemFree:         1686332 kB
Buffers:           25112 kB
Cached:           116548 kB



The memory of our byt-m is also 2GiB.
After running one time "xrandr --output VGA1 --scale 10x10":
-----------------------
[root@x-byt04 ~]# cat /proc/meminfo
MemTotal:        1910716 kB
MemFree:          974536 kB
Buffers:           25180 kB
Cached:           820776 kB
.....
.....

running "xrandr --output VGA1 --scale 10x10" for the second time, meminfo have no changes.
Comment 7 Chris Wilson 2013-09-27 10:14:57 UTC
I've pushed a series of patches to harden our allocation and handling of very large framebuffers and I am able to start firefox at 19200x10800 despite the system continually reporting ENOMEM/ENOSPC - though eventually we do hit an untrapped signal and die.
Comment 8 Gordon Jin 2014-01-02 02:11:41 UTC
Yangwei, please verify on behalf of Kobe.
Comment 9 shui yangwei 2014-01-02 05:39:30 UTC
Retesting with latest -nightly kernel, this bug is gone. I tried many times scale setting and rotation, the Baytrail machine can work properly.

Verified here.


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.