Bug 10645 - XV plays only on external VGA
Summary: XV plays only on external VGA
Status: RESOLVED FIXED
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: git
Hardware: x86 (IA32) Linux (All)
: medium normal
Assignee: JM Ibanez
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords:
: 9807 10158 10732 (view as bug list)
Depends on:
Blocks: 10158
  Show dependency treegraph
 
Reported: 2007-04-14 07:47 UTC by JM Ibanez
Modified: 2007-05-29 15:16 UTC (History)
6 users (show)

See Also:
i915 platform:
i915 features:


Attachments
xvinfo(1) output (1.84 KB, text/plain)
2007-04-14 07:49 UTC, JM Ibanez
no flags Details
lspci(8) output (10.67 KB, text/plain)
2007-04-14 07:50 UTC, JM Ibanez
no flags Details
Xorg logfile (36.00 KB, text/plain)
2007-04-14 07:50 UTC, JM Ibanez
no flags Details
PATCH: Perform pipe selection on XV based on clip rect (2.70 KB, patch)
2007-04-15 03:17 UTC, JM Ibanez
no flags Details | Splinter Review
PATCH: Translate dstBox coordinates to output-local coordinates (3.19 KB, patch)
2007-04-16 11:52 UTC, JM Ibanez
no flags Details | Splinter Review
xorg.conf (3.97 KB, application/octet-stream)
2007-04-17 17:15 UTC, Conn
no flags Details
Xorg.0.log (unpatched RC4 driver) (33.10 KB, text/plain)
2007-04-17 17:16 UTC, Conn
no flags Details
Xorg.0.log (patched RC4 driver) (33.02 KB, text/plain)
2007-04-17 17:17 UTC, Conn
no flags Details

Description JM Ibanez 2007-04-14 07:47:44 UTC
When playing a movie (either using MPlayer or Totem, both through XV), the Intel driver seems to want to output only on one pipe. On my laptop, LVDS is attached to pipe B, while VGA is attached to pipe A.

When the XV overlay window is on the laptop screen, only the overlay color (blue) is shown, without output; when it is on an attached VGA output, XV output is shown there.

Similarly, when the display is cloned, only the external VGA output shows, while the laptop display shows blue.

When there are no external VGA outputs connected (i.e. only the laptop LCD panel), playing movies via XV shows just the overlay color.

See as follows:

- on clone:
 +------------------+  +------------------+
 |                  |  |                  |
 |  external/VGA    |  |   laptop/LVDS    |
 |                  |  |                  |
 +------------------+  +------------------+
   displays here          blue screen here

- on mergefb:
 +------------------+  +------------------+
 |                  |  |                  |
 |  external/VGA    |  |   laptop/LVDS    |
 |                  |  |                  |
 +------------------+  +------------------+
 overlay window here,
 displays here


 +------------------+  +------------------+
 |                  |  |                  |
 |  external/VGA    |  |   laptop/LVDS    |
 |                  |  |                  |
 +------------------+  +------------------+
                       overlay window here,
                       blue screen
Comment 1 JM Ibanez 2007-04-14 07:49:23 UTC
Created attachment 9601 [details]
xvinfo(1) output

Output from xvinfo(1) on my system.
Comment 2 JM Ibanez 2007-04-14 07:50:14 UTC
Created attachment 9602 [details]
lspci(8) output

Output from lspci -vv
Comment 3 JM Ibanez 2007-04-14 07:50:55 UTC
Created attachment 9603 [details]
Xorg logfile
Comment 4 JM Ibanez 2007-04-14 10:08:40 UTC
In src/i830_video.c, the pipe is selected via pPriv->pipe; however, this never is set to a value other than 0, which means only pipe A is ever selected. Previously, pPriv->pipe was set to pI830->pipe (which no longer exists as a member of _I830Rec).
Comment 5 JM Ibanez 2007-04-15 03:17:48 UTC
Created attachment 9610 [details] [review]
PATCH: Perform pipe selection on XV based on clip rect

As I understand it, bug #10645 is caused by the XV code using just pipe 0 (A) instead of selecting the pipe based on where the clip rectangle is. The following patch introduces logic that checks where the clip rectangle really is and selects that pipe for display via XVideo.

I've run this patch and it does display on the correct pipe for local LCD, but no output is given on external. Still debugging that.
Comment 6 JM Ibanez 2007-04-15 03:21:12 UTC
My Git tree now is configured to turn on overlay debugging (i830_video.c:150 -- switched from '#if 0' to '#if 1'). The debugging output tells me that the code is clipping it when the XV overlay is on the external VGA (where external VGA is --right-of the LVDS)

I've also noticed that when the LVDS is --right-of VGA (i.e. the VGA is at 0,0 of the merged framebuffer), the XV output does show on the external VGA but is clipped (I've been testing it with a 1440x900 external LCD display).
Comment 7 JM Ibanez 2007-04-16 11:52:39 UTC
Created attachment 9619 [details] [review]
PATCH: Translate dstBox coordinates to output-local coordinates
Comment 8 JM Ibanez 2007-04-16 12:24:50 UTC
*** Bug 10158 has been marked as a duplicate of this bug. ***
Comment 9 Daniel Stone 2007-04-17 02:22:36 UTC
*** Bug 9807 has been marked as a duplicate of this bug. ***
Comment 10 JM Ibanez 2007-04-17 02:40:46 UTC
Daniel: regarding comment #9 in bug #9807, is this a side-effect of the patches I introduced? Can you attach the Xorg log (/var/log/Xorg.0.log.old maybe, since your system powers down)?
Comment 11 Daniel Stone 2007-04-17 02:54:38 UTC
no, it's nothing to do with your patch.  it's been going on for a while, and is some kind of electrical, rather than software, issue: if it's plugged in while the laptop is off, it won't turn on at all.
Comment 12 Conn 2007-04-17 17:06:26 UTC
System: Dell Inspiron 510m w/Intel 82855GM graphics, no external monitors attached. Ubuntu Feisty w/latest updates, xserver-xorg-video-intel 1.9.94-1ubuntu3

Using the RC4 driver I have several issues (which I will post as separate bugs when I find out which aren't related to this).

1. Switching to a VT sometimes causes an irrecoverable system crash, and changing modes (e.g. running ppracer) always causes a crash (no, DRI is functioning correctly for windowed apps).
2. Video playback never works; sometimes the movie plays with a blank window, but mostly the server crashes, restarts and the system freezes with the cursor onscreen before GDM can kick in.

I applied JM Ibanez's two patches and recompiled, and now video playback is working with absolutely no crashes reproducible. Aside from issues with bad default contrast and saturation (which is probably unrelated to the patches), I just wanted to let you know these patches helped. I'll attach two Xorg.0.log files, using unpatched and patched driver. If any more logs will be handy, let me know.
Comment 13 Conn 2007-04-17 17:15:58 UTC
Created attachment 9645 [details]
xorg.conf
Comment 14 Conn 2007-04-17 17:16:32 UTC
Created attachment 9646 [details]
Xorg.0.log (unpatched RC4 driver)
Comment 15 Conn 2007-04-17 17:17:01 UTC
Created attachment 9647 [details]
Xorg.0.log (patched RC4 driver)
Comment 16 Daniel Stone 2007-04-24 02:42:17 UTC
*** Bug 10732 has been marked as a duplicate of this bug. ***
Comment 17 Raúl 2007-05-01 04:04:31 UTC
Hello.
I'm interested in this bug. This is closed as of April, 17th but I wonder if patches will be included or if they are accepted.

Thanks
Comment 18 Brice Goglin 2007-05-10 10:31:24 UTC
I don't see why this bug is marked as fixed. Did anything get committed in git yet?

Also, it is said in #10466 that when the patches are applied, the XV overlay is now always on LVDS. It does not look like a good solution to me.
Comment 19 Samuel Thibault 2007-05-24 15:50:05 UTC
I tried today's git (with the recent overlay fixes), and it works great on my 855GM, I can even move the overlay window between screens, the overlay nicely switches appropriately!

That said, sometimes I'd like to show the overlay on both screen. For this, I tried xrandr --output VGA --crtc 1 --auto --output LVDS --crtc 1 --auto (since xrandr --verbose tells me CRTCs: 0 1 for VGA and CRTCs: 1 for LVDS), but this xrandr call fails with error BadMatch on request 156(RANDR) 21. Shouldn't that be possible?
Comment 20 Eric Anholt 2007-05-29 12:27:33 UTC
This bug is now fixed, with the overlay choosing the CRTC with the most coverage of the window, and XV_PIPE being available to override it.  To Samuel, no, you are often not able to clone different types of outputs on one CRTC due to clocking differences.
Comment 21 Samuel Thibault 2007-05-29 15:16:06 UTC
> To Samuel, no, you are often not able to clone different types of outputs on
> one CRTC due to clocking differences.

Well, yes, I knew that, I should have been a bit more precise: when running
xrandr --verbose, part of the output is:

VGA connected
...
1024x768 (0x4e) 65.0MHz
	h: width  1024 start 1048 end 1184 total 1344 skew 0 clock 48.4KHz
	v: height  768 start  771 end  777 total  806        clock 60.0Hz
...
LVDS connected
...
1024x768 (0x4e) 65.0MHz
	h: width  1024 start 1048 end 1184 total 1344 skew 0 clock 48.4KHz
	v: height  768 start  771 end  777 total  806        clock 60.0Hz

So I tried

xrandr --output VGA --mode 0x4e --crtc 1 --output LVDS --mode 0x4e --crtc 1

but this fails. Since the clock values are exactly the same, shouldn't this
work?


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.