Re: Matrox G550, dual head, 1680x1050

Re: Matrox G550, dual head, 1680x1050

[Hans-Jürgen Koch]

Am Freitag, 2. März 2007 16:24 schrieb Tilman Sauerbeck:
> Hans-Jürgen Koch [2007-03-02 16:10]:
> > Am Freitag 02 März 2007 15:39 schrieb Tilman Sauerbeck:
> > > Hans-Jürgen Koch [2007-03-02 10:53]:
> > > > I'm the proud owner of two NEC MultiSync 2070WNX monitors (1680x1050)
> > > > and a Matrox G550 dual DVI card. I'm now trying to set up a dual head
> > > > configuration with this equipment.
> > > >
> > > > [snip]
> > > >
> > > > Is this a known limitation of matroxfb? Or a X server bug?
> > >
> > > AFAIK matroxfb doesn't support the G550 dual DVI card. The mga X driver
> > > certainly doesn't support it yet.
> >
> > Well, at the moment I'm using the VGA outputs. Current status is:
>
> I believe/suspect the hardware doesn't care whether you use the 2VGA
> cable or the 2DVI cable. Unfortunately, I only have the dual-dvi cable
> so I cannot be sure. Last time I had a close look at dual-dvi was around
> christmas and I don't remember all the details ;)
>
> > * Without X (boot, VT) I get clean 1280x1024 on both monitors. The
> > monitors interpolate that to their native resolution, both show the same
> > image.

Meanwhile, I managed to get 1680x1050 to work with matroxfb. Here's proof that 
both monitors really work:

http://www.hjk-az.de/temp/g550-dualhead-fbcon-small.jpg

The left monitor is the primary head (/dev/fb0) and shows text on a VT. There 
I issued a command "fbi -d /dev/fb1 image.jpg" that shows a JPEG image on the 
second monitor. Note that there is no X running (The black-on-white text is 
due to a small patch I usually apply to vt.c).


>
> > * With a xinerama configuration, X starts without any (EE) errors.
> > Monitor1 shows a clean 1680x1050 image, Monitor2 a distorted image. I can
> > move the mouse or windows from one monitor to the other. The reason for
> > the distortion on monitor2 is most probably that the card is still
> > configured for 1280x1024 while the X server thinks it's 1680x1050.

Meanwhile I doubt this explanation. Here's what fbset has to say:

$ fbset -fb /dev/fb1 --show

mode "1680x1050-60"
    # D: 135.501 MHz, H: 65.396 kHz, V: 59.997 Hz
    geometry 1680 1050 1680 1050 32
    timings 7380 32 224 32 4 136 4
    rgba 8/16,8/8,8/0,8/24
endmode

So, even fbset thinks it's 1680x1050. But where does the distortion come 
from???

>
> I'm curious, can you describe the distortion in more detail/take a
> photo?

Yes, now I've got one:

http://www.hjk-az.de/temp/g550-xinerama-error-small.jpg

This shows a KDE login screen. Note that KDE scaled the background image with 
the Debian logo to cover both screens. The right screen is distorted, 
but "working": If I move a window over to the second screen, I see changes 
there.

Conclusion:

@matroxfb developers: Thanks a lot, guys! Well done! Most people I asked about 
having a dual-head framebuffer console at 1680x1050 thought it would be 
impossible.

@mga xorg developers: Well... Can you tell me why the kernel guys can 
initialize and use the second head and you can't? If you can give me hints 
where in the source I can start to look for the bug, I'm willing to spend 
some time there. Is there something I could try in my configuration?

Any hints?

Thanks,
Hans

Re: Matrox G550, dual head, 1680x1050

[Petr Vandrovec]

Hans-Jürgen Koch wrote:
> Am Freitag, 2. März 2007 16:24 schrieb Tilman Sauerbeck:
> Meanwhile I doubt this explanation. Here's what fbset has to say:
> 
> $ fbset -fb /dev/fb1 --show
> 
> mode "1680x1050-60"
>     # D: 135.501 MHz, H: 65.396 kHz, V: 59.997 Hz
>     geometry 1680 1050 1680 1050 32
>     timings 7380 32 224 32 4 136 4
>     rgba 8/16,8/8,8/0,8/24
> endmode
> 
> So, even fbset thinks it's 1680x1050. But where does the distortion come 
> from???

Can you try setting vxres for second head from 1680 to 1696, or 1728? 
Maybe X look at vxres and not at framebuffer line length (what fbset -i 
reports for line length?)

If you are using mga driver for second head in X, then probably X driver 
does not know correctly what are hardware restrictions for second CRTC 
(I think they've changed between G400 and G450), asks for 1680 but gets 
1664, and this causes those diagonal lines you see.
								Petr


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: Matrox G550, dual head, 1680x1050

[Geert Uytterhoeven]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: TEXT/PLAIN; charset=UTF-8, Size: 1200 bytes --]

On Sun, 4 Mar 2007, Petr Vandrovec wrote:
> Hans-Jürgen Koch wrote:
> > Am Freitag, 2. März 2007 16:24 schrieb Tilman Sauerbeck:
> > Meanwhile I doubt this explanation. Here's what fbset has to say:
> > 
> > $ fbset -fb /dev/fb1 --show
> > 
> > mode "1680x1050-60"
> >     # D: 135.501 MHz, H: 65.396 kHz, V: 59.997 Hz
> >     geometry 1680 1050 1680 1050 32
> >     timings 7380 32 224 32 4 136 4
> >     rgba 8/16,8/8,8/0,8/24
> > endmode
> > 
> > So, even fbset thinks it's 1680x1050. But where does the distortion come 
> > from???
> 
> Can you try setting vxres for second head from 1680 to 1696, or 1728? 
> Maybe X look at vxres and not at framebuffer line length (what fbset -i 
> reports for line length?)

These days at least the X server in Debian looks correctly at the frame buffer
line length, else it wouldn't work on PS3 with non-fullscreen modes.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds

[-- Attachment #2: Type: text/plain, Size: 345 bytes --]

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

[-- Attachment #3: Type: text/plain, Size: 182 bytes --]

_______________________________________________
Linux-fbdev-devel mailing list
Linux-fbdev-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-fbdev-devel

Re: [Linux-fbdev-devel] Matrox G550, dual head, 1680x1050

[Michel Dänzer]

On Mon, 2007-03-05 at 07:33 +0100, Geert Uytterhoeven wrote:
> On Sun, 4 Mar 2007, Petr Vandrovec wrote:
> > Hans-Jrgen Koch wrote:
> > > Am Freitag, 2. Mrz 2007 16:24 schrieb Tilman Sauerbeck:
> > > Meanwhile I doubt this explanation. Here's what fbset has to say:
> > > 
> > > $ fbset -fb /dev/fb1 --show
> > > 
> > > mode "1680x1050-60"
> > >     # D: 135.501 MHz, H: 65.396 kHz, V: 59.997 Hz
> > >     geometry 1680 1050 1680 1050 32
> > >     timings 7380 32 224 32 4 136 4
> > >     rgba 8/16,8/8,8/0,8/24
> > > endmode
> > > 
> > > So, even fbset thinks it's 1680x1050. But where does the distortion come 
> > > from???
> > 
> > Can you try setting vxres for second head from 1680 to 1696, or 1728? 
> > Maybe X look at vxres and not at framebuffer line length (what fbset -i 
> > reports for line length?)
> 
> These days at least the X server in Debian looks correctly at the frame buffer
> line length, else it wouldn't work on PS3 with non-fullscreen modes.

It depends on the driver. The fbdev and radeon drivers should handle it
correctly, not sure about others.


-- 
Earthling Michel Dänzer           |          http://tungstengraphics.com
Libre software enthusiast         |          Debian, X and DRI developer

Re: [Linux-fbdev-devel] Matrox G550, dual head, 1680x1050

[Hans-Jürgen Koch]

Am Montag, 5. März 2007 05:59 schrieb Petr Vandrovec:
> Hans-Jürgen Koch wrote:
> > Am Freitag, 2. März 2007 16:24 schrieb Tilman Sauerbeck:
> > Meanwhile I doubt this explanation. Here's what fbset has to say:
> >
> > $ fbset -fb /dev/fb1 --show
> >
> > mode "1680x1050-60"
> >     # D: 135.501 MHz, H: 65.396 kHz, V: 59.997 Hz
> >     geometry 1680 1050 1680 1050 32
> >     timings 7380 32 224 32 4 136 4
> >     rgba 8/16,8/8,8/0,8/24
> > endmode
> >
> > So, even fbset thinks it's 1680x1050. But where does the distortion come
> > from???
>
> Can you try setting vxres for second head from 1680 to 1696, or 1728?
> Maybe X look at vxres and not at framebuffer line length (what fbset -i
> reports for line length?)

Thanks a lot, Petr, you're on the right track! With

fbset -fb /dev/fb1 -g 1680 1050 1696 1050 32 -t 7380 224 32 32 4 136 4

the distortion disappears. At the moment I have to do this manually after X 
has started. I'll have to experiment a bit with my configurations, but at 
least now I have working xinerama.

Thanks again for your help.

Hans

Re: Matrox G550, dual head, 1680x1050

[Hans-Jürgen Koch]

Am Montag, 5. März 2007 05:59 schrieb Petr Vandrovec:

>
> Can you try setting vxres for second head from 1680 to 1696, or 1728?
> Maybe X look at vxres and not at framebuffer line length (what fbset -i
> reports for line length?)
>

Here are more details:

1.) I boot my kernel (vanilla 2.6.20.1) with the following parameters:

video=matroxfb:outputs:120,pixclock:7380,xres:1680,yres:1050,upper:32,
               lower:4,left:32,right:224,hslen:136,vslen:4,depth:32

This initializes the primary head correctly. The second head is probably 
wrong, because the monitor's on screen message complains about a non-ideal 
resolution.

2.) X starts with Xinerama=On, using this modeline for each monitor:

Modeline "1680x1050@60" 154.20 1680 1712 2296 2328 1050 1071 1081 1103

3.) After that, fbset -i reports this:

$ fbset -i -fb /dev/fb0

mode "1680x1050-60"
    # D: 135.501 MHz, H: 65.396 kHz, V: 59.997 Hz
    geometry 1680 1050 1696 2473 32
    timings 7380 32 224 32 4 136 4
    accel true
    rgba 8/16,8/8,8/0,8/24
endmode

Frame buffer device information:
    Name        : MATROX
    Address     : 0xd4000000
    Size        : 16777216
    Type        : PACKED PIXELS
    Visual      : TRUECOLOR
    XPanStep    : 8
    YPanStep    : 1
    YWrapStep   : 0
    LineLength  : 6784
    MMIO Address: 0xd6000000
    MMIO Size   : 16384
    Accelerator : Matrox G400


$ fbset -i -fb /dev/fb1

mode "640x480-60"
    # D: 25.176 MHz, H: 31.469 kHz, V: 59.942 Hz
    geometry 640 480 640 480 32
    timings 39721 48 16 33 10 96 2
    rgba 8/16,8/8,8/0,8/24
endmode

Frame buffer device information:
    Name        : MATROX DH
    Address     : 0xd5800000
    Size        : 8388608
    Type        : PACKED PIXELS
    Visual      : TRUECOLOR
    XPanStep    : 8
    YPanStep    : 1
    YWrapStep   : 0
    LineLength  : 2560
    MMIO Address: 0xd6000000
    MMIO Size   : 16384
    Accelerator : No

As you can see, the second monitor fell back to VGA resolution, but the logs 
tell that X thinks it successfully switched to 1680x1050. The KDE control 
center reports the same. The image on the second monitor is distorted.

4.) I then issue the following command:

$ fbset -fb /dev/fb1 -g 1680 1050 1696 1050 32 -t 7380 224 32 32 4 136 4

With that, the image on the second monitor immediately becomes clear, and I 
have a fully working xinerama configuration then.

As far as I can see, fbset is the only way to initialize the second head 
properly. Neither matroxfb nor mga do it correctly with 1680x1050.

Cheers,
Hans

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: Matrox G550, dual head, 1680x1050

[Petr Vandrovec]

Hans-Jürgen Koch wrote:
> Am Montag, 5. März 2007 05:59 schrieb Petr Vandrovec:
> 
>> Can you try setting vxres for second head from 1680 to 1696, or 1728?
>> Maybe X look at vxres and not at framebuffer line length (what fbset -i
>> reports for line length?)
>>
> 
> Here are more details:
> 
> 1.) I boot my kernel (vanilla 2.6.20.1) with the following parameters:
> 
> video=matroxfb:outputs:120,pixclock:7380,xres:1680,yres:1050,upper:32,
>                lower:4,left:32,right:224,hslen:136,vslen:4,depth:32
> 
> This initializes the primary head correctly. The second head is probably 
> wrong, because the monitor's on screen message complains about a non-ideal 
> resolution.

Second head always starts in 640x480/32bpp.  You have to use fbset to 
switch it to different mode, so matroxfb works fine as far as I can tell.

> As you can see, the second monitor fell back to VGA resolution, but the logs 
> tell that X thinks it successfully switched to 1680x1050. The KDE control 
> center reports the same. The image on the second monitor is distorted.

Because X (mga driver) was programming hardware...

> 4.) I then issue the following command:
> 
> $ fbset -fb /dev/fb1 -g 1680 1050 1696 1050 32 -t 7380 224 32 32 4 136 4
> 
> With that, the image on the second monitor immediately becomes clear, and I 
> have a fully working xinerama configuration then.

Yes, and this corrects X programming mistake.

> As far as I can see, fbset is the only way to initialize the second head 
> properly. Neither matroxfb nor mga do it correctly with 1680x1050.

Well, fbset == matroxfb, so I do not quite understand your complaint.
							Petr


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: [Linux-fbdev-devel] Matrox G550, dual head, 1680x1050

[Hans-Jürgen Koch]

Am Dienstag, 6. März 2007 10:25 schrieb Petr Vandrovec:
> Hans-Jürgen Koch wrote:
> > Am Montag, 5. März 2007 05:59 schrieb Petr Vandrovec:
> >> Can you try setting vxres for second head from 1680 to 1696, or 1728?
> >> Maybe X look at vxres and not at framebuffer line length (what fbset -i
> >> reports for line length?)
> >
> > Here are more details:
> >
> > 1.) I boot my kernel (vanilla 2.6.20.1) with the following parameters:
> >
> > video=matroxfb:outputs:120,pixclock:7380,xres:1680,yres:1050,upper:32,
> >                lower:4,left:32,right:224,hslen:136,vslen:4,depth:32
> >
> > This initializes the primary head correctly. The second head is probably
> > wrong, because the monitor's on screen message complains about a
> > non-ideal resolution.
>
> Second head always starts in 640x480/32bpp.  You have to use fbset to
> switch it to different mode, so matroxfb works fine as far as I can tell.
>
> > As you can see, the second monitor fell back to VGA resolution, but the
> > logs tell that X thinks it successfully switched to 1680x1050. The KDE
> > control center reports the same. The image on the second monitor is
> > distorted.
>
> Because X (mga driver) was programming hardware...
>
> > 4.) I then issue the following command:
> >
> > $ fbset -fb /dev/fb1 -g 1680 1050 1696 1050 32 -t 7380 224 32 32 4 136 4
> >
> > With that, the image on the second monitor immediately becomes clear, and
> > I have a fully working xinerama configuration then.
>
> Yes, and this corrects X programming mistake.
>
> > As far as I can see, fbset is the only way to initialize the second head
> > properly. Neither matroxfb nor mga do it correctly with 1680x1050.
>
> Well, fbset == matroxfb, so I do not quite understand your complaint.
> 							

If matroxfb deliberately initializes the second head to 640x480/32bpp, then it 
is OK. I thought the kernel boot parameters would apply to both heads. But on 
second thought, I agree, it has to be that way. Otherwise you'd run into 
trouble if you had two different monitors connected as there are no 
individual settings for both heads.

Thanks for your explanations!

Hans

Re: Matrox G550, dual head, 1680x1050

[Geert Uytterhoeven]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: TEXT/PLAIN; charset=UTF-8, Size: 1448 bytes --]

On Tue, 6 Mar 2007, Petr Vandrovec wrote:
> Hans-Jürgen Koch wrote:
> > This initializes the primary head correctly. The second head is probably 
> > wrong, because the monitor's on screen message complains about a non-ideal 
> > resolution.
> 
> Second head always starts in 640x480/32bpp.  You have to use fbset to 
> switch it to different mode, so matroxfb works fine as far as I can tell.

> > $ fbset -fb /dev/fb1 -g 1680 1050 1696 1050 32 -t 7380 224 32 32 4 136 4
                                      ^^^^
> > 
> > With that, the image on the second monitor immediately becomes clear, and I 
> > have a fully working xinerama configuration then.
> 
> Yes, and this corrects X programming mistake.
> 
> > As far as I can see, fbset is the only way to initialize the second head 
> > properly. Neither matroxfb nor mga do it correctly with 1680x1050.
> 
> Well, fbset == matroxfb, so I do not quite understand your complaint.

If I followed this discussion correctly, there's still an issue with matroxfb,
as it doesn't seem to round up xres_virtual to 1696 automatically.
You have to do that yourself.

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds

[-- Attachment #2: Type: text/plain, Size: 345 bytes --]

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

[-- Attachment #3: Type: text/plain, Size: 182 bytes --]

_______________________________________________
Linux-fbdev-devel mailing list
Linux-fbdev-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-fbdev-devel

Re: Matrox G550, dual head, 1680x1050

[Petr Vandrovec]

Geert Uytterhoeven wrote:
> On Tue, 6 Mar 2007, Petr Vandrovec wrote:
>> Hans-J�rgen Koch wrote:
>>> This initializes the primary head correctly. The second head is probably 
>>> wrong, because the monitor's on screen message complains about a non-ideal 
>>> resolution.
>> Second head always starts in 640x480/32bpp.  You have to use fbset to 
>> switch it to different mode, so matroxfb works fine as far as I can tell.
> 
>>> $ fbset -fb /dev/fb1 -g 1680 1050 1696 1050 32 -t 7380 224 32 32 4 136 4
>                                       ^^^^
>>> With that, the image on the second monitor immediately becomes clear, and I 
>>> have a fully working xinerama configuration then.
>> Yes, and this corrects X programming mistake.
>>
>>> As far as I can see, fbset is the only way to initialize the second head 
>>> properly. Neither matroxfb nor mga do it correctly with 1680x1050.
>> Well, fbset == matroxfb, so I do not quite understand your complaint.
> 
> If I followed this discussion correctly, there's still an issue with matroxfb,
> as it doesn't seem to round up xres_virtual to 1696 automatically.
> You have to do that yourself.

Well, as far as I can tell 1680 is legal line length for 32bit color 
mode.  In Hans-Jurgen scenario X program display themselves (otherwise 
fbset would not report 640x480 on /dev/fb1 once X start), and apparently 
screw up rounding.  Having to use 1696 should be necessary only for 
16bpp, and for such case matroxfb already contains code to round up 
value to multiple of 32...

So unless somebody can reproduce problem with using fbdev driver on both 
primary and secondary head I do not believe there is any problem to fix 
on matroxfb side.
								Petr



-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
Linux-fbdev-devel mailing list
Linux-fbdev-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-fbdev-devel

Re: Matrox G550, dual head, 1680x1050

[Tilman Sauerbeck]

[-- Attachment #1.1: Type: text/plain, Size: 1053 bytes --]

Hans-Jürgen Koch [2007-03-04 22:47]:
> @matroxfb developers: Thanks a lot, guys! Well done! Most people I asked about 
> having a dual-head framebuffer console at 1680x1050 thought it would be 
> impossible.
> 
> @mga xorg developers: Well... Can you tell me why the kernel guys can 
> initialize and use the second head and you can't? If you can give me hints 
> where in the source I can start to look for the bug, I'm willing to spend 
> some time there. Is there something I could try in my configuration?

matroxfb is much better organized wrt dual-head configurations. It also
has support for some hardware that xf86-video-mga doesn't support yet.
I'm working on that.

I think you might be hitting this bug:
https://bugs.freedesktop.org/show_bug.cgi?id=9854

I just attached a temporary fix that you could try.

Regards,
Tilman

-- 
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?

[-- Attachment #1.2: Type: application/pgp-signature, Size: 189 bytes --]

[-- Attachment #2: Type: text/plain, Size: 143 bytes --]

_______________________________________________
xorg mailing list
xorg@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/xorg