display problem with matroxfb

display problem with matroxfb

[f5ibh]

Hi !

I've a matrox mystique with 8Mb RAM.
I've a problem when I use matroxfb instead vesafb.
If I enable CONFIG_FB_VESA, I get the nice logo and all is right for me.
If I enable CONFIG_FB_MATROX, the beginning of each line is in the middle
of the screen and the cursor position does not match the prompt position.
Nevetheless, the screen is 'readable' (no garbage). I've read the files
in ../Documentation/fb and the Framebuffer-HOWTO.

This appends with both 2.2.xx and 2.4.x kernels.

Do I missed something ?

----
Regards
		Jean-Luc
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: display problem with matroxfb

[Petr Vandrovec]

On Sun, Jan 21, 2001 at 03:48:44PM +0100, f5ibh wrote:
> Hi !
> 
> I've a matrox mystique with 8Mb RAM.
> I've a problem when I use matroxfb instead vesafb.
> If I enable CONFIG_FB_VESA, I get the nice logo and all is right for me.
> If I enable CONFIG_FB_MATROX, the beginning of each line is in the middle
> of the screen and the cursor position does not match the prompt position.
> Nevetheless, the screen is 'readable' (no garbage). I've read the files
> in ../Documentation/fb and the Framebuffer-HOWTO.

Are you sure that you did not enabled both vesafb and matroxfb? They cannot
work together. Also, does this happen only in 8bpp mode, or does this
happen in other color depths too?
						Petr Vandrovec
						vandrove@vc.cvut.cz


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: display problem with matroxfb

[f5ibh]

Hi Petr,

Thank you for the answer.

>Are you sure that you did not enabled both vesafb and matroxfb? They cannot
>work together. Also, does this happen only in 8bpp mode, or does this
>happen in other color depths too?

Yes, sure. I've read the docs and tested with vesafb enabled OR matroxfb
enabled, never both.
This happens with both 8bpp or 16bpp. 

Is there a specific configuration file for the matrox ? 

---
Regards
		Jean-Luc
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: display problem with matroxfb

[Petr Vandrovec]

On 22 Jan 01 at 22:11, f5ibh wrote:
> >Are you sure that you did not enabled both vesafb and matroxfb? They cannot
> >work together. Also, does this happen only in 8bpp mode, or does this
> >happen in other color depths too?
> 
> Yes, sure. I've read the docs and tested with vesafb enabled OR matroxfb
> enabled, never both.
> This happens with both 8bpp or 16bpp. 
> 
> Is there a specific configuration file for the matrox ? 

No, only kernel options described in linux/Documentation/fb/matroxfb.txt.
Can you try 'video=matrox:vesa:0x105'? And does this happen if you'll
do 'fbset -accel false', or if you boot with 'video=matrox:noaccel' ?
                                        Thanks,
                                                Petr Vandrovec
                                                vandrove@vc.cvut.cz
                                                
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: display problem with matroxfb

[f5ibh]

Hi Petr,

I've tested the parameters you gave to me, with no success : all is right with
vesafb and wrong with matroxfb. vesa:0x105 does not work, the monitor is out of
sync (I've tried 0x305 too). 

Here are parts of the syslog at system boot for matroxfb and vesafb :

matroxfb: Matrox Mystique (PCI) detected
matroxfb: MTRR's turned on
matroxfb: 640x480x8bpp (virtual: 640x13075)
matroxfb: framebuffer at 0xE7000000, mapped to 0xc8805000, size 8388608
Console: switching to colour frame buffer device 80x30
fb0: MATROX VGA frame buffer device


vesafb: framebuffer at 0xe7000000, mapped to 0xc8000000, size 8192k
vesafb: mode is 640x480x8, linelength=640, pages=11
vesafb: protected mode interface info at c000:7820
vesafb: scrolling: redraw
Console: switching to colour frame buffer device 80x30
fb0: VESA VGA frame buffer device

After booting and having the display shifted to the middle of the screen, I've
played a bit with fbset and the -left, -right, -move, -match options. After a
while I got an 'acceptable' picture (some pixels missing on the left). At this
point, fbset -s  give me the following :

mode "640x480-60"
    # D: 25.176 MHz, H: 31.628 kHz, V: 60.243 Hz
    geometry 640 480 640 480 8
    timings 39721 50 10 32 11 96 2
    accel true
    rgba 8/0,8/0,8/0,0/0
endmode
		    
I've replaced these values in /etc/fb.modes instead of the original values and
after login in the system, entering 'fbset -match -a' set the virtual consoles
correctly.

How can I pass the parameters at boot time ? 
I've tried :
video=matrox:vesa:0x301,pixclock:39721,left:50,right:10,upper:32,lower:11,hslen:96,vslen:2
... without any success...

--
Regards
		Jean-Luc
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: display problem with matroxfb

[Petr Vandrovec]

On 23 Jan 01 at 21:34, f5ibh wrote:
> 
> After booting and having the display shifted to the middle of the screen, I've

Stop. Do you see right portion of screen in left, left portion in right,
with black column in between, or is picture only shifted, without
wraparound? I thought that you see wrapped display...

> played a bit with fbset and the -left, -right, -move, -match options. After a
> while I got an 'acceptable' picture (some pixels missing on the left). At this
> point, fbset -s  give me the following :
> 
> mode "640x480-60"
>     # D: 25.176 MHz, H: 31.628 kHz, V: 60.243 Hz
>     geometry 640 480 640 480 8
>     timings 39721 50 10 32 11 96 2
>     accel true
>     rgba 8/0,8/0,8/0,0/0
> endmode

Are you sure that it is source of problem? Default matroxfb settings
for 640x480 is 'timmings 39721 48 16 33 10 96 2' - As horizontal position
is in multiple of 8, I cannot understand, how moving picture one column
(8 pixels) right(!) (and even worse - there is no rounding in code,
so you only changed right screen margin from 16 to 8 - this should
move picture even more to right side of screen) and one scanline up could 
make so drastical change...

> How can I pass the parameters at boot time ? 
> I've tried :
> video=matrox:vesa:0x301,pixclock:39721,left:50,right:10,upper:32,lower:11,hslen:96,vslen:2
> ... without any success...

After boot look at /proc/cmdline. If it is cutted at column 64 (or 79),
you have to upgrade your LILO. RedHat6.2 uses LILO with this dumb limitation.
Also, you do not have to specify vesa,pixclock,hslen and vslen, as you leave
them on defaults. So 'video=matrox:left:50,right:10,upper:32,lower:11'
should work... But I think that only 'right:' really matters.
                                                Best regards,
                                                    Petr Vandrovec
                                                    vandrove@vc.cvut.cz

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: display problem with matroxfb

[f5ibh]

Hi Petr,

> Stop. Do you see right portion of screen in left, left portion in right,
> with black column in between, or is picture only shifted, without
> wraparound? I thought that you see wrapped display...

The display is shifted, WITH wraparound : the end of the line is at the
beginning of the following one ...

> Are you sure that it is source of problem? Default matroxfb settings
> for 640x480 is 'timmings 39721 48 16 33 10 96 2' - As horizontal position

No, I'm not sure ... so I've tested the parameters you mention : 
timmings 39721 48 16 33 10 96 2

I copied the following block in /etc/fb.modes

mode "640x480-60"
    # D: 25.200 MHz, H: 31.500 kHz, V: 59.999 Hz
    geometry 640 480 640 480 8
    timings 39683 48 16 33 10 96 2
    accel true
    rgba 8/0,8/0,8/0,0/0
endmode

With these parameters, the command 'fbset -match -a' is enough to reset the
display to normal operation. So, you was right.

> After boot look at /proc/cmdline. If it is cutted at column 64 (or 79),
> you have to upgrade your LILO. RedHat6.2 uses LILO with this dumb limitation.

I use Debian 2.2 and ... grub. But I've tested with lilo too. With the sae
results. The cmdline is not truncated.

> Also, you do not have to specify vesa,pixclock,hslen and vslen, as you leave
> them on defaults. So 'video=matrox:left:50,right:10,upper:32,lower:11'
> should work... But I think that only 'right:' really matters.

This does not work ... I've to use the fbset command to reset the display to
normal operation after login in (or in an automatic command at boot time in
/et/init.d). This works, but I've lost the logo  ;-)

--
Regards
		Jean-Luc
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: display problem with matroxfb

[Petr Vandrovec]

Grr. Did not pass through due to DUL blacklist...

----- Transcript of session follows -----
... while talking to vger.kernel.org.:
>>> RCPT To:<linux-kernel@vger.kernel.org>
<<< 550 5.7.1 Policy analysis reported: See <URL:http://mail-abuse.org/dul/> rcpt=<linux-kernel@vger.kernel.org>
550 5.1.1 linux-kernel@vger.kernel.org... User unknown

----- Original message -----

Date: Wed, 24 Jan 2001 01:53:09 +0100
From: Petr Vandrovec <vandrove@vc.cvut.cz>
To: f5ibh <f5ibh@db0bm.ampr.org>
Cc: linux-kernel@vger.kernel.org
Subject: Re: display problem with matroxfb
User-Agent: Mutt/1.3.12i
In-Reply-To: <200101232319.AAA08536@db0bm.ampr.org>; from f5ibh@db0bm.ampr.org on Wed, Jan 24, 2001 at 12:19:14AM +0100

On Wed, Jan 24, 2001 at 12:19:14AM +0100, f5ibh wrote:
> I copied the following block in /etc/fb.modes
> 
> mode "640x480-60"
>     # D: 25.200 MHz, H: 31.500 kHz, V: 59.999 Hz
>     geometry 640 480 640 480 8
>     timings 39683 48 16 33 10 96 2
>     accel true
>     rgba 8/0,8/0,8/0,0/0
> endmode
> 
> With these parameters, the command 'fbset -match -a' is enough to reset the
> display to normal operation. So, you was right.

It is stranger and stranger with each email from you :-(

> > Also, you do not have to specify vesa,pixclock,hslen and vslen, as you leave
> > them on defaults. So 'video=matrox:left:50,right:10,upper:32,lower:11'
> > should work... But I think that only 'right:' really matters.
> 
> This does not work ... I've to use the fbset command to reset the display to
> normal operation after login in (or in an automatic command at boot time in
> /et/init.d). This works, but I've lost the logo  ;-)

Can you try 'video=matrox:init' ? And 'video=matrox:nopan'? First (init) should
reinitialize hardware from scratch. Maybe it is programmed to some unusual
delay between memory and CRT - but in such case shift cannot be more than 64 
pixels (8 chars).

And nopan disables vertical panning (by default matroxfb sets vyres to largest 
possible value, equivalent to 'geometry 640 480 640 65535 8'). But if there is 
problem with vertical panning, screen should jump to left and right when scrolling 
up, not stay just moved right.

If everything above fails, there is function matrox_set_var in 
linux/drivers/video/matrox/matroxfb_base.c. This function contains also follwing
lines:

        mga_setr(M_CRTC_INDEX, 0x0D, p0);
        mga_setr(M_CRTC_INDEX, 0x0C, p1);
#ifdef CONFIG_FB_MATROX_32MB
        if (ACCESS_FBINFO(devflags.support32MB))
                mga_setr(M_EXTVGA_INDEX, 0x08, p3);
#endif
        mga_setr(M_EXTVGA_INDEX, 0x00, p2);

Can you just duplicate this block once more? Documentation says that value written
to M_CRTC_INDEX/C,D is applied after write to M_EXTVGA_INDEX/0, but maybe that your
piece of hardware thinks that it is just other way around (I'm sure that Millennium
and Gxx0 hardware needs CRTC0D/0C first, and CRTCEXT0 last, but who knows).
					Best regards,
						Petr Vandrovec
						vandrove@vc.cvut.cz



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: display problem with matroxfb

[f5ibh]

Hi Petr,

> Grr. Did not pass through due to DUL blacklist...
What is DUL blacklist ?
If the problem is with my e-mail address (it is an hamradio address), you can
use <jean-luc.coulon@wanadoo.fr> instead.

> Can you try 'video=matrox:init' ? And 'video=matrox:nopan'?

Bingo ! 'init' does not work but 'nopam' give me a normal display without the
fbset gimnick.

Thank you for the help.

Next step will be to test dual head. I've already installed an AGP ATI (S3
chipset) together with the PCI matrox Mystique and it seems to work. If I
succeed, I will attach a normal VGA display on the ATI board and a 19" IBM (not
multisync) on the Matrox. The ATI card is a cheap one with only 4Mb RAM. I've
upgraded my Matrox Mystique to 8Mb, so I hope it will work.
Now, I will have a look on how to use XF68_FBDev (if needed).

--
Regards
		Jean-Luc
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: display problem with matroxfb

[Petr Vandrovec]

On 24 Jan 01 at 10:32, f5ibh wrote:
> > Grr. Did not pass through due to DUL blacklist...
> What is DUL blacklist ?
> If the problem is with my e-mail address (it is an hamradio address), you can
> use <jean-luc.coulon@wanadoo.fr> instead.

No. Problem is with linux-kernel@... It started blacklisting direct
mails from dialups, so I have to implement some workaround on my side
(as I do not want to use provider's SMTP gateway - I do not trust them).

> > Can you try 'video=matrox:init' ? And 'video=matrox:nopan'?
> 
> Bingo ! 'init' does not work but 'nopam' give me a normal display without the
> fbset gimnick.

It looks like that there is some problem with screen offset computation then.
Can you try 'video=matrox:cross4MB' instead of '...:nopan'? I did not
have 8MB Mystique for testing, and it even looks reasonable - Mystique
does not use WRAM memory, so there is no need for WRAM workarounds.
If it will work, I'll cook twoliner for Linus.
 
> Next step will be to test dual head. I've already installed an AGP ATI (S3
> chipset) together with the PCI matrox Mystique and it seems to work. If I
> succeed, I will attach a normal VGA display on the ATI board and a 19" IBM (not
> multisync) on the Matrox. The ATI card is a cheap one with only 4Mb RAM. I've
> upgraded my Matrox Mystique to 8Mb, so I hope it will work.
> Now, I will have a look on how to use XF68_FBDev (if needed).

XF4.0.x should work reasonably well. Or you can run accelerated XF on mga:
matroxfb is compatible with accelerated XF 3.3.x, and with accelerated
XF 4.0.x WITHOUT enabled DRI (as DRI code reprograms hardware even if
X are on background) (and 'Option "UseFBDev"' is required if you are
using both heads of G400/G450 with XF4).
                                                Best regards,
                                                    Petr Vandrovec
                                                    vandrove@vc.cvut.cz
                                                    
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: display problem with matroxfb

[f5ibh]

Hi Petr,

> It looks like that there is some problem with screen offset computation then.
> Can you try 'video=matrox:cross4MB' instead of '...:nopan'? I did not

video=matrox:cross4MB  works too !

> XF4.0.x should work reasonably well. Or you can run accelerated XF on mga:
> matroxfb is compatible with accelerated XF 3.3.x, and with accelerated
> XF 4.0.x WITHOUT enabled DRI (as DRI code reprograms hardware even if
> X are on background) (and 'Option "UseFBDev"' is required if you are
> using both heads of G400/G450 with XF4).

I am lost !!
I've XFree86 Version 3.3.6
I've and ATI AGP video card with an S3 chipset and 4Mb plus the Matrox Mystique
(PCI) with 8Mb.

I've told the BIOS to boot the AGP first. So I boot normaly with the display on
the ATI card. Then the system switches to the matrox (and Framebuffer). At this
time, I move manually the display to the Matrox board and all si right (I'm
doing this test before doing some cabling / modification on the 19" IBM
screen). If I start Xwindows with the SVGA accelerated server then it starts on
the ATI card. If I start it with the FBdev server, I get it on the Matrox card.
I've read some stuff in Framebuffer-HOWTO but I think I've still some
experiment to do.
IFAIK, there is no DRI option for the Mystique ... ??

--------

Best Regards

		Jean-Luc
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: display problem with matroxfb

[Rafael E. Herrera]

f5ibh wrote:
> > XF4.0.x should work reasonably well. Or you can run accelerated XF on mga:
> > matroxfb is compatible with accelerated XF 3.3.x, and with accelerated
> > XF 4.0.x WITHOUT enabled DRI (as DRI code reprograms hardware even if
> > X are on background) (and 'Option "UseFBDev"' is required if you are
> > using both heads of G400/G450 with XF4).
> 
> I am lost !!
> I've XFree86 Version 3.3.6
> I've and ATI AGP video card with an S3 chipset and 4Mb plus the Matrox Mystique
> (PCI) with 8Mb.

I think XFree86 3.3.6 will support one card at the time, not both, so
you won't get 
a dual-head server with that version. Upgrade to XFree86 4.0.x.

-- 
     Rafael
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: display problem with matroxfb

[Trevor Hemsley]

On Tue, 23 Jan 2001 21:57:56, "Petr Vandrovec" <VANDROVE@vc.cvut.cz> 
wrote:

> you do not have to specify vesa,pixclock,hslen and vslen, as you leave
> them on defaults. 

Talking of defaults for matroxfb, would you consider limiting the fv: 
value default to something reasonable that'll work on all monitors? It
took me several recompiles/reboots to get a setting that would not put
my monitor into auto-powerdown. If you defaulted to fv:60 then it 
would work on 99.9% of monitors and then people could override that 
upwards. I have a Philips 201B 21" monitor and was using 

append="video=matrox:vesa:400"

and this was setting too high a vertical refresh rate for the monitors
capabilities. Adding fv:85 lets it work. The card is a Matrox 
Millennium G200 8MB SDRAM.

trevor@trevor4:/usr/src/linux > grep ^C .config|grep FB
CONFIG_FB=y            
CONFIG_FB_MATROX=y     
CONFIG_FB_MATROX_G100=y
CONFIG_FBCON_ADVANCED=y
CONFIG_FBCON_CFB8=y    
CONFIG_FBCON_CFB16=y   
CONFIG_FBCON_CFB24=y   
CONFIG_FBCON_CFB32=y   
CONFIG_FBCON_FONTS=y   

-- 
Trevor Hemsley, Brighton, UK.
Trevor-Hemsley@dial.pipex.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: display problem with matroxfb

[Petr Vandrovec]

> > you do not have to specify vesa,pixclock,hslen and vslen, as you leave
> > them on defaults. 
> 
> Talking of defaults for matroxfb, would you consider limiting the fv: 
> value default to something reasonable that'll work on all monitors? It
> took me several recompiles/reboots to get a setting that would not put
> my monitor into auto-powerdown. If you defaulted to fv:60 then it 
> would work on 99.9% of monitors and then people could override that 
> upwards. I have a Philips 201B 21" monitor and was using 
> 
> append="video=matrox:vesa:400"
> 
> and this was setting too high a vertical refresh rate for the monitors
> capabilities. Adding fv:85 lets it work. The card is a Matrox 
> Millennium G200 8MB SDRAM.

Are you sure that it did not run out of horizontal sync, or something
like that? vesa:400 == vesa:0x190 == 1152x864/60Hz... And it powers
up in 60Hz, at least here ;-)

See timmings array in drivers/video/matrox/matroxfb_base.c - all videomodes
except XXXx400 powerups in fv=60Hz unless you specified fh/fv/pixclock.
XXXx400 powerups with fv=70Hz, like standard VGA does.
						Petr Vandrovec
						vandrove@vc.cvut.cz

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: display problem with matroxfb

[Trevor Hemsley]

On Mon, 29 Jan 2001 00:30:34, Petr Vandrovec <vandrove@vc.cvut.cz> 
wrote:

> > > you do not have to specify vesa,pixclock,hslen and vslen, as you leave
> > > them on defaults. 
> > 
> > Talking of defaults for matroxfb, would you consider limiting the fv: 
> > value default to something reasonable that'll work on all monitors? It
> > took me several recompiles/reboots to get a setting that would not put
> > my monitor into auto-powerdown. If you defaulted to fv:60 then it 
> > would work on 99.9% of monitors and then people could override that 
> > upwards. I have a Philips 201B 21" monitor and was using 
> > 
> > append="video=matrox:vesa:400"
> > 
> > and this was setting too high a vertical refresh rate for the monitors
> > capabilities. Adding fv:85 lets it work. The card is a Matrox 
> > Millennium G200 8MB SDRAM.
> 
> Are you sure that it did not run out of horizontal sync, or something
> like that? vesa:400 == vesa:0x190 == 1152x864/60Hz... And it powers
> up in 60Hz, at least here ;-)
> 
> See timmings array in drivers/video/matrox/matroxfb_base.c - all videomodes
> except XXXx400 powerups in fv=60Hz unless you specified fh/fv/pixclock.
> XXXx400 powerups with fv=70Hz, like standard VGA does.

Looks like one you've already fixed. I've retested and recreated it 
but it only happens with 2.2.13 not with the other two kernel sources 
I have (2.2.18 and 2.4.0). I also misinformed you about the way to 
recreate it, I had specified only append="matrox" in my lilo.conf. It 
was 2.2.13 that I did my experimentation on to get it to work in the 
first place and never bothered to retest afterwards. Sorry for the 
false alarm...

-- 
Trevor Hemsley, Brighton, UK.
Trevor-Hemsley@dial.pipex.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/