[LinuxBIOS] vga bios troubles

[Jon Dufresne]

I followed the guide at this site http://linuxbios.org/VGA_support to
get my onboard video to work. I have yet to be successful.

I am a little confused about the Config.lb section. This is what the guid says:

device pci 9.0 on  # PCI
        chip drivers/pci/onboard
                device pci 9.0 on end
                register "rom_address" = "0xfff80000" #512k image
        end
end

I don't get why you list the "device pci 9.0" twice. I tried that and
when I did my vga device never got the driver from onboard.c, instead
it would always get the generic pci driver. So I changed it to only
include the device below the chip line. This is what my Config.lb
looks like

chip northbridge/intel/i855gm
    chip cpu/intel/socket_mPGA479M
    	device apic 0 on end
    end
    device pci_domain 0 on
    	device pci 0.0 on end # host bridge
    	device pci 0.1 on end # memory configuration
    	device pci 0.3 on end # process configuration
    	device pci 1.0 on end
    	
    	chip drivers/pci/onboard
    		device pci 2.0 on end # vga controller
    		register "rom_address" = "0xfff80000" #512k image
    	end

chip southbridge/intel/i82801dbm
    		chip superio/ite/it8712f end
    	end
    end
end

One thing that is also strange is that when I run "lspci" in a working
environment I see two vga devices one at 0:2.0 and one at 0:2.1, I
tried listing both but LB complained that I had a leftover static
device at 0:2.1, it appears that only one of these devices is visible
at boot.

So I continued with the code above, and in the debug output I can see
where linuxbios attempts to run the proprietary vga bios. I see:

PCI: 00:02.0 init
rom address for PCI: 00:02.0 = fff80000
PCI Expansion ROM, signature 0xaa55, INIT size 0xc800, data ptr 0x0040
PCI ROM Image, Vendor 8086, Device 3582,
PCI ROM Image,  Class Code 030000, Code Type 00
copying VGA ROM Image from 0xfff80000 to 0xc0000, 0xc800 bytes
entering emulator

Now, 0xC0000 is controlled by my PAM registers, so I took my best
guess and set it read/write to SDRAM. When I do this LB stays inside
the emulator for a while and eventually kicks out with this message:

halt_sys: file /root/LinuxBIOSv2/src/devices/emulator/x86emu/ops.c, line 4387

I never see anything on my monitor. I looked at the ops.c file but
couldn't figure out what was going on. I have been fooling around with
this for quite some time, changing every little thing I can think of,
but no luck.

My northbridge is an Intel 855 gme, I used the perl script to extract
the vga bios which is 50KiB, I subtracted that size from the ROM_SIZE,
and I concatenated it with the linuxBIOS.

Anyone have an idea how to fix this?

Thanks,
Jon

[Luis Correia]

Hi!

On 12/7/06, Jon Dufresne  wrote:
> I followed the guide at this site http://linuxbios.org/VGA_support to
> get my onboard video to work. I have yet to be successful.
>
> I am a little confused about the Config.lb section. This is what the guid says:
>
> device pci 9.0 on  # PCI
>         chip drivers/pci/onboard
>                 device pci 9.0 on end
>                 register "rom_address" = "0xfff80000" #512k image
>         end
> end
>
> I don't get why you list the "device pci 9.0" twice. I tried that and
> when I did my vga device never got the driver from onboard.c, instead
> it would always get the generic pci driver. So I changed it to only
> include the device below the chip line. This is what my Config.lb
> looks like
>
> chip northbridge/intel/i855gm
>         chip cpu/intel/socket_mPGA479M
>                 device apic 0 on end
>         end
>         device pci_domain 0 on
>                 device pci 0.0 on end # host bridge
>                 device pci 0.1 on end # memory configuration
>                 device pci 0.3 on end # process configuration
>                 device pci 1.0 on end
>             chip drivers/pci/onboard
>                     device pci 2.0 on end # vga controller
>                     register "rom_address" = "0xfff80000" #512k image
>             end
>
>             chip southbridge/intel/i82801dbm
>                     chip superio/ite/it8712f end
>             end
>     end
>
> end
>
> One thing that is also strange is that when I run "lspci" in a working
> environment I see two vga devices one at 0:2.0 and one at 0:2.1, I
> tried listing both but LB complained that I had a leftover static
> device at 0:2.1, it appears that only one of these devices is visible
> at boot.

Maybe it is the second head that is visible from PCI but
un-implemented by hardware.
> So I continued with the code above, and in the debug output I can see
> where linuxbios attempts to run the proprietary vga bios. I see:
>
> PCI: 00:02.0 init
> rom address for PCI: 00:02.0 = fff80000
> PCI Expansion ROM, signature 0xaa55, INIT size 0xc800, data ptr 0x0040
> PCI ROM Image, Vendor 8086, Device 3582,
> PCI ROM Image,  Class Code 030000, Code Type 00
> copying VGA ROM Image from 0xfff80000 to 0xc0000, 0xc800 bytes
> entering emulator
>
> Now, 0xC0000 is controlled by my PAM registers, so I took my best
> guess and set it read/write to SDRAM. When I do this LB stays inside
> the emulator for a while and eventually kicks out with this message:
>
> halt_sys: file /root/LinuxBIOSv2/src/devices/emulator/x86emu/ops.c, line 4387
>
> I never see anything on my monitor. I looked at the ops.c file but
> couldn't figure out what was going on. I have been fooling around with
> this for quite some time, changing every little thing I can think of,
> but no luck.
>
> My northbridge is an Intel 855 gme, I used the perl script to extract
> the vga bios which is 50KiB, I subtracted that size from the ROM_SIZE,
> and I concatenated it with the linuxBIOS.
>
> Anyone have an idea how to fix this?
>
> Thanks,
> Jon

Luis

[Jon Dufresne]

On 12/7/06, Luis Correia  wrote:
> Hi!
>
> On 12/7/06, Jon Dufresne  wrote:
> > I followed the guide at this site http://linuxbios.org/VGA_support to
> > get my onboard video to work. I have yet to be successful.
> >
> > I am a little confused about the Config.lb section. This is what the guid says:
> >
> > device pci 9.0 on  # PCI
> >         chip drivers/pci/onboard
> >                 device pci 9.0 on end
> >                 register "rom_address" = "0xfff80000" #512k image
> >         end
> > end
> >
> > I don't get why you list the "device pci 9.0" twice. I tried that and
> > when I did my vga device never got the driver from onboard.c, instead
> > it would always get the generic pci driver. So I changed it to only
> > include the device below the chip line. This is what my Config.lb
> > looks like
> >
> > chip northbridge/intel/i855gm
> >         chip cpu/intel/socket_mPGA479M
> >                 device apic 0 on end
> >         end
> >         device pci_domain 0 on
> >                 device pci 0.0 on end # host bridge
> >                 device pci 0.1 on end # memory configuration
> >                 device pci 0.3 on end # process configuration
> >                 device pci 1.0 on end
> >             chip drivers/pci/onboard
> >                     device pci 2.0 on end # vga controller
> >                     register "rom_address" = "0xfff80000" #512k image
> >             end
> >
> >             chip southbridge/intel/i82801dbm
> >                     chip superio/ite/it8712f end
> >             end
> >     end
> >
> > end
> >
> > One thing that is also strange is that when I run "lspci" in a working
> > environment I see two vga devices one at 0:2.0 and one at 0:2.1, I
> > tried listing both but LB complained that I had a leftover static
> > device at 0:2.1, it appears that only one of these devices is visible
> > at boot.
>
> Maybe it is the second head that is visible from PCI but
> un-implemented by hardware.
> > So I continued with the code above, and in the debug output I can see
> > where linuxbios attempts to run the proprietary vga bios. I see:
> >
> > PCI: 00:02.0 init
> > rom address for PCI: 00:02.0 = fff80000
> > PCI Expansion ROM, signature 0xaa55, INIT size 0xc800, data ptr 0x0040
> > PCI ROM Image, Vendor 8086, Device 3582,
> > PCI ROM Image,  Class Code 030000, Code Type 00
> > copying VGA ROM Image from 0xfff80000 to 0xc0000, 0xc800 bytes
> > entering emulator
> >
> > Now, 0xC0000 is controlled by my PAM registers, so I took my best
> > guess and set it read/write to SDRAM. When I do this LB stays inside
> > the emulator for a while and eventually kicks out with this message:
> >
> > halt_sys: file /root/LinuxBIOSv2/src/devices/emulator/x86emu/ops.c, line 4387
> >
> > I never see anything on my monitor. I looked at the ops.c file but
> > couldn't figure out what was going on. I have been fooling around with
> > this for quite some time, changing every little thing I can think of,
> > but no luck.
> >
> > My northbridge is an Intel 855 gme, I used the perl script to extract
> > the vga bios which is 50KiB, I subtracted that size from the ROM_SIZE,
> > and I concatenated it with the linuxBIOS.
> >
> > Anyone have an idea how to fix this?
> >
> > Thanks,
> > Jon
>
> Luis

That makes sense, because many of the registers are documented the
same including the exact same VID/DID. However, I don't understand why
this shows up with an lspci but not with linuxbios. This could just be
a peculiarity of pci that I don't understand.

Jon

[Luis Correia]

Hi!

On 12/7/06, Jon Dufresne  wrote:
> On 12/7/06, Luis Correia  wrote:
>
> That makes sense, because many of the registers are documented the
> same including the exact same VID/DID. However, I don't understand why
> this shows up with an lspci but not with linuxbios. This could just be
> a peculiarity of pci that I don't understand.

Define the PIC device twice with the proper ID's but describe only
once the VGA BIOS.

Maybe then it will show up with LB.

BTW, does the board have two vga output devices? (even if one of them
is a LVDS connector)
> Jon

Luis

[Jon Dufresne]

The board is slightly custom, so when I looked at it I orignally
thought it had one, but I looked at the manual for the board it is
based on and I see this

* Integrated Intel 855GME graphic engine with two LVDS channels

So it does indeed look like there are two outputs. I just don't have
physical access to the second one.

Jon

On 12/7/06, Luis Correia  wrote:
> Hi!
>
> On 12/7/06, Jon Dufresne  wrote:
> > On 12/7/06, Luis Correia  wrote:
> >
> > That makes sense, because many of the registers are documented the
> > same including the exact same VID/DID. However, I don't understand why
> > this shows up with an lspci but not with linuxbios. This could just be
> > a peculiarity of pci that I don't understand.
>
> Define the PIC device twice with the proper ID's but describe only
> once the VGA BIOS.
>
> Maybe then it will show up with LB.
>
> BTW, does the board have two vga output devices? (even if one of them
> is a LVDS connector)
> > Jon
>
> Luis

[Jon Dufresne]

> > Define the PIC device twice with the proper ID's but describe only
> > once the VGA BIOS.
> >
> > Maybe then it will show up with LB.

Sorry, I'm not sure exactly what you mean by this. Is ths in Config.lb
or in a source file. I took a guess and I made this in the northbridge
folder called i855gm_vga.c

#include <console/console.h>
#include <device/device.h>
#include <device/pci.h>
#include <device/pci_ids.h>
#include "i855gm.h"

static struct device_operations vga_ops  = {
    .read_resources   = pci_dev_read_resources,
    .set_resources    = pci_dev_set_resources,
    .enable_resources = pci_dev_enable_resources,
    .enable		  = i855gm_enable,
    .init             = 0,
    .scan_bus         = 0,
};

static struct pci_driver vga_driver0 __pci_driver = {
    .ops    = &vga_ops,
    .vendor = PCI_VENDOR_ID_INTEL,
    /* add into pci_ids.h PCI_DEVICE_ID_INTEL_855GME_VGA */
    .device = 0x3582,
};

static struct pci_driver vga_driver1 __pci_driver = {
    .ops    = &vga_ops,
    .vendor = PCI_VENDOR_ID_INTEL,
    /* add into pci_ids.h PCI_DEVICE_ID_INTEL_855GME_VGA */
    .device = 0x3582,
};

[Adam Talbot]

Just asking the dumb question.  Is the 855 chip set supported?
-Adam

Jon Dufresne wrote:
> > > Define the PIC device twice with the proper ID's but describe only
> > > once the VGA BIOS.
> > >
> > > Maybe then it will show up with LB.
>
> Sorry, I'm not sure exactly what you mean by this. Is ths in Config.lb
> or in a source file. I took a guess and I made this in the northbridge
> folder called i855gm_vga.c
>
> #include <console/console.h>
> #include <device/device.h>
> #include <device/pci.h>
> #include <device/pci_ids.h>
> #include "i855gm.h"
>
> static struct device_operations vga_ops  = {
>    .read_resources   = pci_dev_read_resources,
>    .set_resources    = pci_dev_set_resources,
>    .enable_resources = pci_dev_enable_resources,
>    .enable		  = i855gm_enable,
>    .init             = 0,
>    .scan_bus         = 0,
> };
>
> static struct pci_driver vga_driver0 __pci_driver = {
>    .ops    = &vga_ops,
>    .vendor = PCI_VENDOR_ID_INTEL,
>    /* add into pci_ids.h PCI_DEVICE_ID_INTEL_855GME_VGA */
>    .device = 0x3582,
> };
>
> static struct pci_driver vga_driver1 __pci_driver = {
>    .ops    = &vga_ops,
>    .vendor = PCI_VENDOR_ID_INTEL,
>    /* add into pci_ids.h PCI_DEVICE_ID_INTEL_855GME_VGA */
>    .device = 0x3582,
> };

[Jon Dufresne]

There is an 855pm that looks incomplete, I took it and have been
modifying it to work with my board which is 855gme (close but
different). Are you saying there is code I need to write in the
northbridge to get VGA to work beyond the guide? It is fine if I do
and I will do it, I was just didn't realize there was more than what
the guide mentions.

Thanks,
Jon

On 12/7/06, Adam Talbot  wrote:
> Just asking the dumb question.  Is the 855 chip set supported?
> -Adam
>
> Jon Dufresne wrote:
> > > > Define the PIC device twice with the proper ID's but describe only
> > > > once the VGA BIOS.
> > > >
> > > > Maybe then it will show up with LB.
> >
> > Sorry, I'm not sure exactly what you mean by this. Is ths in Config.lb
> > or in a source file. I took a guess and I made this in the northbridge
> > folder called i855gm_vga.c
> >
> > #include <console/console.h>
> > #include <device/device.h>
> > #include <device/pci.h>
> > #include <device/pci_ids.h>
> > #include "i855gm.h"
> >
> > static struct device_operations vga_ops  = {
> >       .read_resources   = pci_dev_read_resources,
> >       .set_resources    = pci_dev_set_resources,
> >       .enable_resources = pci_dev_enable_resources,
> >       .enable           = i855gm_enable,
> >       .init             = 0,
> >       .scan_bus         = 0,
> > };
> >
> > static struct pci_driver vga_driver0 __pci_driver = {
> >       .ops    = &vga_ops,
> >       .vendor = PCI_VENDOR_ID_INTEL,
> >       /* add into pci_ids.h PCI_DEVICE_ID_INTEL_855GME_VGA */
> >       .device = 0x3582,
> > };
> >
> > static struct pci_driver vga_driver1 __pci_driver = {
> >       .ops    = &vga_ops,
> >       .vendor = PCI_VENDOR_ID_INTEL,
> >       /* add into pci_ids.h PCI_DEVICE_ID_INTEL_855GME_VGA */
> >       .device = 0x3582,
> > };

[Adam Talbot]

I could be wrong, but I think I was the last person to work on the 855
over a year ago.  I was never able to get it running.  I would start
with getting the system to boot using a console connection.  Once that
is working, move on to VGA bios support.
-Adam

Jon Dufresne wrote:
> There is an 855pm that looks incomplete, I took it and have been
> modifying it to work with my board which is 855gme (close but
> different). Are you saying there is code I need to write in the
> northbridge to get VGA to work beyond the guide? It is fine if I do
> and I will do it, I was just didn't realize there was more than what
> the guide mentions.
>
> Thanks,
> Jon
>
> On 12/7/06, Adam Talbot  wrote:
> > Just asking the dumb question.  Is the 855 chip set supported?
> > -Adam
> >
> > Jon Dufresne wrote:
> > > > > Define the PIC device twice with the proper ID's but describe only
> > > > > once the VGA BIOS.
> > > > >
> > > > > Maybe then it will show up with LB.
> > >
> > > Sorry, I'm not sure exactly what you mean by this. Is ths in Config.lb
> > > or in a source file. I took a guess and I made this in the northbridge
> > > folder called i855gm_vga.c
> > >
> > > #include <console/console.h>
> > > #include <device/device.h>
> > > #include <device/pci.h>
> > > #include <device/pci_ids.h>
> > > #include "i855gm.h"
> > >
> > > static struct device_operations vga_ops  = {
> > >       .read_resources   = pci_dev_read_resources,
> > >       .set_resources    = pci_dev_set_resources,
> > >       .enable_resources = pci_dev_enable_resources,
> > >       .enable           = i855gm_enable,
> > >       .init             = 0,
> > >       .scan_bus         = 0,
> > > };
> > >
> > > static struct pci_driver vga_driver0 __pci_driver = {
> > >       .ops    = &vga_ops,
> > >       .vendor = PCI_VENDOR_ID_INTEL,
> > >       /* add into pci_ids.h PCI_DEVICE_ID_INTEL_855GME_VGA */
> > >       .device = 0x3582,
> > > };
> > >
> > > static struct pci_driver vga_driver1 __pci_driver = {
> > >       .ops    = &vga_ops,
> > >       .vendor = PCI_VENDOR_ID_INTEL,
> > >       /* add into pci_ids.h PCI_DEVICE_ID_INTEL_855GME_VGA */
> > >       .device = 0x3582,
> > > };

[Jon Dufresne]

ok, Thanks for the tip.

Right now I can get into filo, and from there direct it to a livecd,
however I am guessing the livecd doesn't deliver serial output and I
think it dies somewhere booting the kernel.

What should I do to watch the boot process over a serial connection?

Thanks,
Jon

On 12/7/06, Adam Talbot  wrote:
> I could be wrong, but I think I was the last person to work on the 855
> over a year ago.  I was never able to get it running.  I would start
> with getting the system to boot using a console connection.  Once that
> is working, move on to VGA bios support.
> -Adam
>
> Jon Dufresne wrote:
> > There is an 855pm that looks incomplete, I took it and have been
> > modifying it to work with my board which is 855gme (close but
> > different). Are you saying there is code I need to write in the
> > northbridge to get VGA to work beyond the guide? It is fine if I do
> > and I will do it, I was just didn't realize there was more than what
> > the guide mentions.
> >
> > Thanks,
> > Jon
> >
> > On 12/7/06, Adam Talbot  wrote:
> > > Just asking the dumb question.  Is the 855 chip set supported?
> > > -Adam
> > >
> > > Jon Dufresne wrote:
> > > > > > Define the PIC device twice with the proper ID's but describe only
> > > > > > once the VGA BIOS.
> > > > > >
> > > > > > Maybe then it will show up with LB.
> > > >
> > > > Sorry, I'm not sure exactly what you mean by this. Is ths in Config.lb
> > > > or in a source file. I took a guess and I made this in the northbridge
> > > > folder called i855gm_vga.c
> > > >
> > > > #include <console/console.h>
> > > > #include <device/device.h>
> > > > #include <device/pci.h>
> > > > #include <device/pci_ids.h>
> > > > #include "i855gm.h"
> > > >
> > > > static struct device_operations vga_ops  = {
> > > >       .read_resources   = pci_dev_read_resources,
> > > >       .set_resources    = pci_dev_set_resources,
> > > >       .enable_resources = pci_dev_enable_resources,
> > > >       .enable           = i855gm_enable,
> > > >       .init             = 0,
> > > >       .scan_bus         = 0,
> > > > };
> > > >
> > > > static struct pci_driver vga_driver0 __pci_driver = {
> > > >       .ops    = &vga_ops,
> > > >       .vendor = PCI_VENDOR_ID_INTEL,
> > > >       /* add into pci_ids.h PCI_DEVICE_ID_INTEL_855GME_VGA */
> > > >       .device = 0x3582,
> > > > };
> > > >
> > > > static struct pci_driver vga_driver1 __pci_driver = {
> > > >       .ops    = &vga_ops,
> > > >       .vendor = PCI_VENDOR_ID_INTEL,
> > > >       /* add into pci_ids.h PCI_DEVICE_ID_INTEL_855GME_VGA */
> > > >       .device = 0x3582,
> > > > };

[Adam Talbot]

If you are able to get to filo some one has done some good work on the
855...  ok, live CD.  You need to pass console options to the kernel on
the liveCD. Something like `kernel-version console=tty0
console=ttyS0,115200`  But this can be a pain as you may not have a
console at this point :-).    Having had this exact problem, I found an
old hard drive and installed a small version of Linux, just for
testing.  Setup my test Linux for console redirect and never used VGA
bios until I got every thing else working.

What board are you using? (web link)
-Adam

Jon Dufresne wrote:
> ok, Thanks for the tip.
>
> Right now I can get into filo, and from there direct it to a livecd,
> however I am guessing the livecd doesn't deliver serial output and I
> think it dies somewhere booting the kernel.
>
> What should I do to watch the boot process over a serial connection?
>
> Thanks,
> Jon
>
> On 12/7/06, Adam Talbot  wrote:
> > I could be wrong, but I think I was the last person to work on the 855
> > over a year ago.  I was never able to get it running.  I would start
> > with getting the system to boot using a console connection.  Once that
> > is working, move on to VGA bios support.
> > -Adam
> >
> > Jon Dufresne wrote:
> > > There is an 855pm that looks incomplete, I took it and have been
> > > modifying it to work with my board which is 855gme (close but
> > > different). Are you saying there is code I need to write in the
> > > northbridge to get VGA to work beyond the guide? It is fine if I do
> > > and I will do it, I was just didn't realize there was more than what
> > > the guide mentions.
> > >
> > > Thanks,
> > > Jon
> > >
> > > On 12/7/06, Adam Talbot  wrote:
> > > > Just asking the dumb question.  Is the 855 chip set supported?
> > > > -Adam
> > > >
> > > > Jon Dufresne wrote:
> > > > > > > Define the PIC device twice with the proper ID's but describe
> >
> > only
> > > > > > > once the VGA BIOS.
> > > > > > >
> > > > > > > Maybe then it will show up with LB.
> > > > > > >
> > > > >
> > > > > Sorry, I'm not sure exactly what you mean by this. Is ths in
> >
> > Config.lb
> > > > > or in a source file. I took a guess and I made this in the
> >
> > northbridge
> > > > > folder called i855gm_vga.c
> > > > >
> > > > > #include <console/console.h>
> > > > > #include <device/device.h>
> > > > > #include <device/pci.h>
> > > > > #include <device/pci_ids.h>
> > > > > #include "i855gm.h"
> > > > >
> > > > > static struct device_operations vga_ops  = {
> > > > >       .read_resources   = pci_dev_read_resources,
> > > > >       .set_resources    = pci_dev_set_resources,
> > > > >       .enable_resources = pci_dev_enable_resources,
> > > > >       .enable           = i855gm_enable,
> > > > >       .init             = 0,
> > > > >       .scan_bus         = 0,
> > > > > };
> > > > >
> > > > > static struct pci_driver vga_driver0 __pci_driver = {
> > > > >       .ops    = &vga_ops,
> > > > >       .vendor = PCI_VENDOR_ID_INTEL,
> > > > >       /* add into pci_ids.h PCI_DEVICE_ID_INTEL_855GME_VGA */
> > > > >       .device = 0x3582,
> > > > > };
> > > > >
> > > > > static struct pci_driver vga_driver1 __pci_driver = {
> > > > >       .ops    = &vga_ops,
> > > > >       .vendor = PCI_VENDOR_ID_INTEL,
> > > > >       /* add into pci_ids.h PCI_DEVICE_ID_INTEL_855GME_VGA */
> > > > >       .device = 0x3582,
> > > > > };