matroxfb not working after trying to upgrade to 2.6.3

matroxfb not working after trying to upgrade to 2.6.3

[Juha]

Hello list,

I've got a problem with matroxfb not working in either 2.4.23 or 2.6.3 after
trying to upgrade to 2.6 series. My videocard is a G400 and it's built as a
module. Previously, when I modprobed the module, it would register itself in
/etc/log/syslog, but now nothing is printed in the syslog when inserting the
modules. And I don't understand, what has changed. These are the parameters I
give to lilo on boot:

Feb 23 14:06:27 linux kernel: Kernel command line: auto BOOT_IMAGE=Linux ro
root=303 video=matrox:vesa:0x115

And this is my .config:

CONFIG_FB_MATROX=m
CONFIG_FB_MATROX_G450=y
CONFIG_FB_MATROX_G100=y
CONFIG_FB_MATROX_I2C=m
CONFIG_FB_MATROX_MAVEN=m

and lsmod:

matroxfb_crtc2          8308   0  (unused)
matroxfb_maven         13308   0  (unused)
i2c-matroxfb            2772   0  (unused)
matroxfb_base          20580  63  [matroxfb_crtc2 i2c-matroxfb]
matroxfb_g450           4708   0  [matroxfb_base]
matroxfb_DAC1064        9500   0  [matroxfb_crtc2 matroxfb_base]
g450_pll                3536   0  [matroxfb_g450 matroxfb_DAC1064]
matroxfb_accel          9448   0  [matroxfb_base matroxfb_DAC1064]
matroxfb_misc          15364   0  [matroxfb_crtc2 matroxfb_maven i2c-matroxfb
matroxfb_base matroxfb_g450 matroxfb_DAC1064 g450_pll matroxfb_accel]

but when I try to use matroxset, with both 2.4 and 2.6, I get:

bash:# matroxset -f /dev/fb0 -m 2
ioctl failed: Device or resource busy

I'd be very grateful for any help,  juhis

Re: matroxfb not working after trying to upgrade to 2.6.3

[Juha Pahkala]

Juha <juhis <at> trinity.is-a-geek.com> writes:

> 
> 



by the way, I forgot this one the first time:

I had to apply this patch, found from the lkml archives, in order for
matroxfb_maven not to segfault when inserting the module.

--- gold-2.6/drivers/video/matrox/matroxfb_maven.c      2003-10-25
14:42:44.000000000 -0400
+++ linux-2.6.0/drivers/video/matrox/matroxfb_maven.c   2003-12-22
21:55:04.082725504 -0500
@@ -1249,6 +1249,7 @@
                err = -ENOMEM;
                goto ERROR0;
        }
+       memset(new_client, 0, sizeof(*new_client) + sizeof(*data));
        data = (struct maven_data*)(new_client + 1);
        i2c_set_clientdata(new_client, data);
        new_client->addr = address;



juhis

Re: matroxfb not working after trying to upgrade to 2.6.3

[Petr Vandrovec]

On 23 Feb 04 at 12:35, Juha wrote:
> I've got a problem with matroxfb not working in either 2.4.23 or 2.6.3 after
> trying to upgrade to 2.6 series. My videocard is a G400 and it's built as a
> module. Previously, when I modprobed the module, it would register itself in
> /etc/log/syslog, but now nothing is printed in the syslog when inserting the
> modules. And I don't understand, what has changed. These are the parameters I
> give to lilo on boot:
> 
> Feb 23 14:06:27 linux kernel: Kernel command line: auto BOOT_IMAGE=Linux ro
> root=303 video=matrox:vesa:0x115
> 
> And this is my .config:
> 
> CONFIG_FB_MATROX=m

Do not build it as a module. It is not going to work in usual configurations.
Also, did you built CONFIG_FRAMEBUFFER_CONSOLE as a module and insmod-ded it
AFTER matroxfb?

> but when I try to use matroxset, with both 2.4 and 2.6, I get:
> 
> bash:# matroxset -f /dev/fb0 -m 2
> ioctl failed: Device or resource busy

Probably output #1 is already used by someone else. Build everything
to the kernel, you'll get much saner defaults then, as it forces 
correct order for crtc2 & maven initializations.

What reports 'matroxset -f /dev/fb0 -m' and 'matroxset -f /dev/fb1 -m' ?
Does not second one report that /dev/fb1 is currently displayed on that
output?
                                                                Petr
                                                                

Re: matroxfb not working after trying to upgrade to 2.6.3

[Juha Pahkala]

Petr Vandrovec <VANDROVE <at> vc.cvut.cz> writes:

> > CONFIG_FB_MATROX=m
> 
> Do not build it as a module. It is not going to work in usual configurations.

Building FB_MATROX into the kernel fixed the problem, so thanks alot Petr for
the quick help!!

> What reports 'matroxset -f /dev/fb0 -m' and 'matroxset -f /dev/fb1 -m' ?
> Does not second one report that /dev/fb1 is currently displayed on that
> output?

The fb0 was mapped to crtc1 and fb1 to crtc2, and even matroxset -m 5 option
wouldn't unmap them. But everything works now :)

juhis