Bug 17823

Complementary information, when I downgrade the driver without a power cycle (just restarting X), I can switch to VGA but pixel clocks are totally messed. I have to power cycle in order to restore correct settings

>> When using Intel driver 2.4.2, I was able to use xrandr to switch from LVDS to
VGA-1.
What's the previous working driver version (before 2.4.2)? You seem to make a typo.

(In reply to comment #2)
> >> When using Intel driver 2.4.2, I was able to use xrandr to switch from LVDS to
> VGA-1.
> What's the previous working driver version (before 2.4.2)? You seem to make a
> typo.
> 

Oups, sorry, previous version was 2.2.1

Sounds like you have an SDVO attached VGA port on this system and the SDVO code is confused about its type.  There have been some fixes to the master branch though lately, does this still happen with the 2.5.1 driver or git master?

I built 2.5.1 driver and unfortunately it is exactly the same. 

Created attachment 20588 [details] [review]
prefer VGA over TV output

It looks like your SDVO capabilities indicate both VGA and TV outputs.  Does your machine have both?  This patch should make the driver preferentially initialize the VGA output instead, hopefully making things work again for you.  Can you give it a try?

Could you attach X log with ModeDebug option on? And  your machine seems only have VGA and no TV right?

Created attachment 20590 [details]
2.5.1 Debug logs

No, this is a very small netbook (Fujitsu U810/U1010) with no TV output, just Panel and VGA. 

I attached the debug output with 2.5.1 build

I think your SDVO chip is buggy as it announced non-existing outputs. I think
Jesse's patch will fix for you.

Patch applied, it's working very well, thanks !!!

Is this something that could be applied for anyone and can I notify Mandriva about it ?

Unfortunately we can't push this patch now. I've just checked on a 915G LF board which has SDVO-TV output, the capability flag looks like yours, means SDVO chip announces support for VGA and TV, but this board has only TV output. With Jesse's patch, we will have fault type on this board.

It looks we might have to find out what's the actually current SDVO output type.
Still open this one for reference.

Any way I can help ?

We need to check SDVO spec again to see if any command can be used for this, or we might provide config option. You may use Jesse's patch for now, and help to test with any future patches. Thanks!

Created attachment 20826 [details] [review]
SDVO use attached display flag for output setup

Ulhume, could you help to try with this patch in place of Jesse's origin patch? This is applied to current git master. Please help to report back the result with X log with ModeDebug on.

Thanks.

Created attachment 21135 [details]
Xorg log file with patch applied on last code base (2.6.99)

So I tried the patch, it's working well. I had at one point a strange behavior (attached log file). When using xrandr, the LVDS panel was shown as "UNKNOWN-1". I had to restart the server in order to get LVDS back. 

But for my "core" problem this patch seems to be fine. I have VGA-1 and I can switch to it. 

There is a last problem with pm-suspend. It seems that when going back to life, the backlight stays off. I have to switch to text mode and go back to graphic mode in order to switch on the backlight. I'm not sure this is intel driver related so you tell me. 

Comment on attachment 20826 [details] [review]
SDVO use attached display flag for output setup

This does wrong to get attached display in initialization, as that refused an existing output without monitor to be setup in X start. I'm reworking the SDVO multi-function encoder support patch.

Created attachment 21225 [details] [review]
Handle multifunc SDVO encoder

Here's my current patch for handle multiple function SDVO encoder. Please help to verify this, thanks!

It works nicely here. Do you need me to do some specific test/command output ? 

Created attachment 21722 [details]
SDVO: handle multifunction encoder (updated for HEAD version)

This is juste an updated patch working with last GIT version. Any chance to get this mainstream ?

Thanks, my updated patch has been on intel-gfx@lists.freedesktop.org for people to review and comment. I'll do more testing and push it if without problem.

On this for sure, I test your patch every day and it's working perfectly here. Many thanks for this !!!

Created attachment 21820 [details] [review]
Add new option for SDVO output setting

I've reworked the fix in another way, to add an option for SDVO output setting if our driver's default priority list is not right according with current connect in multifunction SDVO encoder case.

Origin patch doesn't work right in case there's no monitor connected when X start, and later when connected, current randr output can't be changed easily after output setup. And for user multi-head config it also can't handle correctly. It looks we should detect or initialize the connect type earlier in SDVO setup. I'm asking hw guys for possible method to detect real output type. If that can't be done, we can use this patch instead. So you provide option
like Option "SDVOBOutput" "VGA-1" should work, and normal multi-head config can work too.

Rather than doing this, why not create outputs covers reported capabilities?

1)RGB->VGA
2)Since there will be only one of CVBS/SVID/YPRPB/SCART, these capabilities can be grouped to create one output called TV to cover them all. Any one capability report connected, the TV output is connected.

as long as the output is reported as disconnected, it won't be trouble, right?

Hi zhenyu,

Does the latest intel driver includes your patches that ccan solve bug 18687 as well? 

Should I try to use your patch in comment #24?

Even git master does not include zhenyu's patch, although there is no better solution at the moment.

Is there any timeline or schedule to release a new driver with sdvo fix?

Intel,

we are looking forward to you reply for question posted in comment#26/28 so that Novell can get it  to include the fix in for us.

Lew/Guek. Even if intel would make a release with this patch, Novell won't update to this release, since it might include regressions and we no longer have time for extensive testing. Question is, if Intel considers to put this patch in a new release, or if it needs to be handled differently anyway and the patch in comment #24 is just a hack. I don't know. Comment #25 sounds like this.

yeah, new option might not be ideal solution. Or we just go with my previous approach, I'll ask keith/eric for comments.

Created attachment 23100 [details] [review]
Updated patch for multifunction SDVO device

Here's refreshed patch against git master. Please help to test and verify.

I just did a git clone from git://anongit.freedesktop.org/git/xorg/driver/xf86-video-intel. May I know what version is it? I cannot build it due to libdrm from SLE 11 is still 2.4.1.

Driver's version is marked in configure.ac.

You can grab libdrm 2.4.4 tarball from http://dri.freedesktop.org/libdrm/, and build yourself.

(In reply to comment #33)
> I just did a git clone from
> git://anongit.freedesktop.org/git/xorg/driver/xf86-video-intel. May I know what
> version is it? I cannot build it due to libdrm from SLE 11 is still 2.4.1.
> 

Lew Shih Ming, are you able to continue this with zhenyu's suggestion?

(In reply to comment #35)
> (In reply to comment #33)
> > I just did a git clone from
> > git://anongit.freedesktop.org/git/xorg/driver/xf86-video-intel. May I know what
> > version is it? I cannot build it due to libdrm from SLE 11 is still 2.4.1.
> > 
> Lew Shih Ming, are you able to continue this with zhenyu's suggestion?

Hi Zhenyu and Michael,

I am still working on it. I just got the machine back for testing and will verify it by today. Will update you guys once I applied the patch to the git-master.

(In reply to comment #34)
> Driver's version is marked in configure.ac.
> You can grab libdrm 2.4.4 tarball from http://dri.freedesktop.org/libdrm/, and
> build yourself.

Zhenyu,

I trying to compile the libdrm-2.4.5. but encounter some issue. From the readme of the tarball, I followed the instruction doing ./configure --prefix=usr/ --exec-prefix=/ and make install.

Then I compile xf86-video-intel by doing ./autogen.sh. It stop at a point when it says:-
checking for DRM... configure: error: Package requirements (libdrm >=2.4.5) were not met:
Requested 'libdrm >= 2.4.5' but version of libdrm is 2.4.1

Then I cannot carry on and do make. Could you advice?

> I trying to compile the libdrm-2.4.5. but encounter some issue. From the readme
> of the tarball, I followed the instruction doing ./configure --prefix=usr/
> --exec-prefix=/ and make install.
> 
> Then I compile xf86-video-intel by doing ./autogen.sh. It stop at a point when
> it says:-
> checking for DRM... configure: error: Package requirements (libdrm >=2.4.5)
> were not met:
> Requested 'libdrm >= 2.4.5' but version of libdrm is 2.4.1
> 
> Then I cannot carry on and do make. Could you advice?
> 

following the below instruction to build libdrm and xf86-video-intel
1) libdrm
./autogen.sh --prefix=/usr --libdir=/usr/lib (note: if you use x64, please use /usr/lib64)
and then
make install

2) xf86-video-intel
Should be the same as libdrm. Just note x64 setting.

What you come across for libdrm >=2.4.5?
1) you got wrong code. Seems it should master code. 
2) you install libdrm.pc to wrong directory. That is why you should set --libdir=/usr/lib(or /usr/lib64)

Wish it can help you.

(In reply to comment #38)
> > I trying to compile the libdrm-2.4.5. but encounter some issue. From the readme
> > of the tarball, I followed the instruction doing ./configure --prefix=usr/
> > --exec-prefix=/ and make install.
> > 
> > Then I compile xf86-video-intel by doing ./autogen.sh. It stop at a point when
> > it says:-
> > checking for DRM... configure: error: Package requirements (libdrm >=2.4.5)
> > were not met:
> > Requested 'libdrm >= 2.4.5' but version of libdrm is 2.4.1
> > 
> > Then I cannot carry on and do make. Could you advice?
> > 
> following the below instruction to build libdrm and xf86-video-intel
> 1) libdrm
> ./autogen.sh --prefix=/usr --libdir=/usr/lib (note: if you use x64, please use
> /usr/lib64)
> and then
> make install
> 2) xf86-video-intel
> Should be the same as libdrm. Just note x64 setting.
> What you come across for libdrm >=2.4.5?
> 1) you got wrong code. Seems it should master code. 
> 2) you install libdrm.pc to wrong directory. That is why you should set
> --libdir=/usr/lib(or /usr/lib64)
> Wish it can help you.

Thanks for your help. I had successfully built the intel driver. FYI, there is no autogen.sh script in libdrm directory. So I use configure script to do the trick. I will continue with the dual display testing.

(In reply to comment #34)
> Driver's version is marked in configure.ac.
> You can grab libdrm 2.4.4 tarball from http://dri.freedesktop.org/libdrm/, and
> build yourself.

Hi Zhenyu,

I had built both libdrm and xfree86-video-intel on SLED11 RC4. The kernel is 2.6.27.15-2-pae. After rebooting my system, I could not login from gnome-login. I have not patch with the updated patch from comment #32. Could it have break some of the xorg code?

I don't understand if you haven't used that patch, how could it break anything? 

Please try to apply that patch to your xf86-video-intel, then build and install again.

(In reply to comment #41)
> I don't understand if you haven't used that patch, how could it break anything? 
> Please try to apply that patch to your xf86-video-intel, then build and install
> again.

I had tested with the patch. It's still the same problem. Do you need me to collect any logs?

Created attachment 23323 [details]
video clip of gnome-login problem

This is a video clip that I had taken on the gnome-login problem. The resolution of the clip is very low but hopefully you guys can get the idea of what I meant. 

After I enter with my user account and password, it will start to login without any authcentication error. Then the screen will start to refresh and it comes back to the same login screen.

Well, this can be everything. Xserver crashing or Gnome desktop terminating the Xserver to kill the Xsessions immediately - for whatever reasons.

Could you paste the X log in failure case?

Created attachment 23365 [details]
Xorg.0.log collected at AltF1  before Xlogin in AltF7

Xorg.0.log collected at AltF1  before Xlogin in AltF7

Not able to proceed to ALTF1 once the terminal goes through the Xlogin in ALTF7 and relaunch login screen.

this doesn't look like caused by zhenyu's patch. The mis-detected-integrated-TV should be a problem of our bios parsing code... 

Could you ignore integrated TV in xorg.conf? 

Section "Monitor"
    Identifier "TV"
    Option "Ignore" "true"
EndSection

Section "Device"
    ...
    Driver "intel"
    ...
    Option "monitor-TV" "TV"
EndSection

Created attachment 23429 [details]
xorg.conf file with disable monitor-tv

Hi Zhenyu,

I had included the new section to disable monitor-TV. It still does not work. Is there any workaround for the bios parsing code?

(In reply to comment #49)
> Created an attachment (id=23429) [details]
> xorg.conf file with disable monitor-tv
> 
> Hi Zhenyu,
> 
> I had included the new section to disable monitor-TV. It still does not work.
> Is there any workaround for the bios parsing code?
> 

please ignore my comment in comment#47, it shouldn't be the problem of your symptom..zhenyu's solution in comment# 48 should be enough, if it's caused by that... 

have you attached monitor to both VGA ports on your platform? To test zhenyu's patch, we need it. thanks.

Created attachment 23433 [details]
xorg log after disabling monitor-tv

This is the xorg log I had collected after configuring xorg.conf with the disabling of monitor-tv. The login problem still exists with/without 2 monitors connected.

from the log, the SDVO output has been successfully detected as VGA-1 now, rather than TV-1 as before. This is what zhenyu's patch is about..

For the issue you saw in comment# 43, it should be caused by something else. Have you tried to turns option like NoAccel, EXANoCompsite on? Compiz or other fancy desktop feature could be the culprit, too.

Created attachment 23467 [details]
xorg.conf for 2 monitors

Thanks Michael for your tips. I am able to login after uninstalling compiz completely. 

Zhengyu, this is the xorg.conf file I used to configure dual monitors manually after unsuccessfull attempts of using the previous xorg.conf file that I had used to disable LVDS. I am attaching the log file in the next comment.

Created attachment 23468 [details]
xorg log files after enabling 2 monitors

Pushed the patch. Close.
commit 0d20bbbc2005a51f427a9ae6b6a66dbbb101dbab
Author: Zhenyu Wang <zhenyu.z.wang@intel.com>
Date:   Thu Feb 19 14:24:24 2009 +0800

    SDVO: handle multifunction encoder (try 2)
    
    For SDVO encoder that advertise multiple functions,
    we have to get attached display to determine current
    output, and update output's name according with
    current type.

(In reply to comment #55)
> Pushed the patch. Close.
> commit 0d20bbbc2005a51f427a9ae6b6a66dbbb101dbab
> Author: Zhenyu Wang <zhenyu.z.wang@intel.com>
> Date:   Thu Feb 19 14:24:24 2009 +0800
>     SDVO: handle multifunction encoder (try 2)
>     For SDVO encoder that advertise multiple functions,
>     we have to get attached display to determine current
>     output, and update output's name according with
>     current type.

Hi Zhenyu,

Do you mean that you had committed the patch to git? 

The fix have successfully replaced TV-1 with VGA-1. Does this meant that dual displays (2 external VGA) should be fixed too? I still have problem displaying on my 2nd VGA interface.

(In reply to comment #57)
> The fix have successfully replaced TV-1 with VGA-1. Does this meant that dual
> displays (2 external VGA) should be fixed too? I still have problem displaying
> on my 2nd VGA interface.
> 

in comment# 53, you've confirmed getting rid of compiz will make your dual head work, havn't you?

what issue do you run into now?

In comment #53, I said that I have unsuccessfull attempts of using the previous xorg.conf file which was working previously. Which means I still have problem making my 2nd monitor to display even thought xrandr is reflecting VGA-1 instead of TV-1. And I had attached my xorg.conf and log file for someone to review it. I am waiting for further comments. Sorry for the misinterpretation.

(In reply to comment #59)
> In comment #53, I said that I have unsuccessfull attempts of using the previous
> xorg.conf file which was working previously. Which means I still have problem
> making my 2nd monitor to display even thought xrandr is reflecting VGA-1
> instead of TV-1. And I had attached my xorg.conf and log file for someone to
> review it. I am waiting for further comments. Sorry for the misinterpretation.
> 

what did you see on the 2nd VGA monitor ( I assume it's the one connected to SDVO VGA )? Blank screen?

Yes. You are correct. It's using SDVO. Only primary monitor is able to display while the secondary is blank.

If so, there should be another bug for SDVO VGA issue, don't pollute this closed bug track.