I've got the newest working version of the emulator for anyone who wants to try it.
This won't work on K8 systems minus some hacks, due to K8 differences, contact me if interested.
ron
ron minnich rminnich@lanl.gov writes:
I've got the newest working version of the emulator for anyone who wants to try it.
This won't work on K8 systems minus some hacks, due to K8 differences, contact me if interested.
This will or won't work on K8 systems?
Eric
On 4 Feb 2004, Eric W. Biederman wrote:
This won't work on K8 systems minus some hacks, due to K8 differences, contact me if interested.
This will or won't work on K8 systems?
What we've found so far: - K8 mtrr settings are incorrect for the fixed MTRRs for VGA. There are extension bits that are more or less random that we need to fix in linuxbios. - Previously mentioned routing stuff needs to be set in the K8 northbridge - GART table settings are not correct, so GART is not yet there.
But we're still pretty close. We have shell scripts and a quick hack to linuxbios and the cards are *almost* there. Once we get all the bits, we can start to get linuxbios set up correctly.
ron
On Wed, 2004-02-04 at 20:34, ron minnich wrote:
On 4 Feb 2004, Eric W. Biederman wrote:
This won't work on K8 systems minus some hacks, due to K8 differences, contact me if interested.
This will or won't work on K8 systems?
What we've found so far:
- K8 mtrr settings are incorrect for the fixed MTRRs for VGA. There are extension bits that are more or less random that we need to fix in linuxbios.
- Previously mentioned routing stuff needs to be set in the K8 northbridge
- GART table settings are not correct, so GART is not yet there.
But we're still pretty close. We have shell scripts and a quick hack to linuxbios and the cards are *almost* there. Once we get all the bits, we can start to get linuxbios set up correctly.
It is done. We have AGP 8x and HW 3D acceleration now. We can even run tuxracer ;-).
Ollie
On Thu, Feb 05, 2004 at 10:20:05AM -0700, Li-Ta Lo wrote:
It is done. We have AGP 8x and HW 3D acceleration now. We can even run tuxracer ;-).
Great work!
Does it automagically work with many different cards? :p
//Peter
On Thu, 2004-02-05 at 10:45, Peter Stuge wrote:
On Thu, Feb 05, 2004 at 10:20:05AM -0700, Li-Ta Lo wrote:
It is done. We have AGP 8x and HW 3D acceleration now. We can even run tuxracer ;-).
Great work!
Does it automagically work with many different cards? :p
Not yet. The bios on Nvidia card has a specific feature to not calling itself by intX. Other cards like ATI want to install the intX handlers and then call them. Currently the emulator can not handle this but it is very simple to add this.
Of course, we need to try on more vendor/models.
Ollie
* Li-Ta Lo ollie@lanl.gov [040205 18:58]:
Not yet. The bios on Nvidia card has a specific feature to not calling itself by intX. Other cards like ATI want to install the intX handlers and then call them. Currently the emulator can not handle this but it is very simple to add this.
Cool. If this works, SCSI option roms could also be used to boot from after they installed an int13 handler...
Stefan
On Thu, 2004-02-05 at 11:28, Stefan Reinauer wrote:
- Li-Ta Lo ollie@lanl.gov [040205 18:58]:
Not yet. The bios on Nvidia card has a specific feature to not calling itself by intX. Other cards like ATI want to install the intX handlers and then call them. Currently the emulator can not handle this but it is very simple to add this.
Cool. If this works, SCSI option roms could also be used to boot from after they installed an int13 handler...
Actually, I suspect that this has already been implemented in the original XFree86 version. Ron has removed/disgarded some good things when porting it to LB.
Ollie
On Thu, 5 Feb 2004, Li-Ta Lo wrote:
Actually, I suspect that this has already been implemented in the original XFree86 version. Ron has removed/disgarded some good things when porting it to LB.
yes, I sure did, there was stuff in there I did not like but it turns out we need it :=(
ron
On Thu, 5 Feb 2004, Stefan Reinauer wrote:
Cool. If this works, SCSI option roms could also be used to boot from after they installed an int13 handler...
that makes my brain hurt :-)
ron
On Thu, 5 Feb 2004, Peter Stuge wrote:
Great work!
Does it automagically work with many different cards? :p
it's still held together with rubber bands, there is work to do now to get linuxbios to do the right things.
What you have to do right now by hand (wrmsr, setpci, etc.) - get a video bios image - Do some msr writes to correctly set fixed msr with extensions - get the base address for gart (read vga card for this) - set IORR1, and some other K8 northbridge settings based on gart base - run testbios (newest one)
I think that's about it.
ron
On Thu, 2004-02-05 at 11:09, ron minnich wrote:
On Thu, 5 Feb 2004, Peter Stuge wrote:
Great work!
Does it automagically work with many different cards? :p
it's still held together with rubber bands, there is work to do now to get linuxbios to do the right things.
What you have to do right now by hand (wrmsr, setpci, etc.)
- get a video bios image
- Do some msr writes to correctly set fixed msr with extensions
- get the base address for gart (read vga card for this)
Should it be from AGP bridge ??
- set IORR1, and some other K8 northbridge settings based on gart base
- run testbios (newest one)
Ollie
I'd like a copy of the latest version. I'll run it against our ATI M1 bios
I sent it to Richard.
I am putting the scripts we use here for info.
ron jotun root # cat FIXGART.sh #!/bin/sh insmod /usr/src/linux/arch/i386/kernel/msr.o # fix the fixed MTRR /root/wrmsr 0xC0010010 0x1e0601 0 /root/wrmsr 0x259 0 0
#fix the IORR1 crap on BOTH northbridges
/root/rdmsr 0xC0010018 /root/wrmsr 0xC0010018 0xd0000000 0xff /root/rdmsr 0xC0010019 /root/wrmsr 0xC0010019 0xd8000000 0xff
# set up 0xd in the whatever setpci -s 0:18.3 94.l setpci -s 0:18.3 94.l=68 setpci -s 0:18.3 94.l setpci -s 0:18.3 90.b setpci -s 0:18.3 90.b=3 setpci -s 0:18.3 90.b setpci -s 0:19.3 94.l setpci -s 0:19.3 94.l=68 setpci -s 0:19.3 94.l setpci -s 0:19.3 90.b setpci -s 0:19.3 90.b=3 setpci -s 0:19.3 90.b # 8151 setpci -s 1:1.0 10.l # setpci -s 1:1.0 10.l=f0000008 # setpci -s 1:1.0 10.l setpci -s 1:1.0 b4.l # setpci -s 1:1.0 b4.l=00010f20 # setpci -s 1:1.0 b4.l jotun root # cat DO_GEFORCEFX.sh #! /bin/sh #insmod /usr/src/linux/arch/i386/kernel/msr.o #/root/wrmsr 0xC0010010 0x1e0601 0 #/root/wrmsr 0x259 0 0 /root/vgabios//testbios --abseg /dev/mem -s 65536 /root/nvbios.bin 2>&1
/tmp/out
modprobe nvidia NVreg_EnableAGPSBA=1 NVreg_EnableAGPFW=1 #modprobe agpgart agp_try_unsupported=1 modprobe agpgart jotun root #
It should have modprobe agpgart agp_try_unsupported=1 uncommented. I had difficulty loading the module with agp_try_unsupported in kernel 2.6.2, looking into it further right now...
On Fri, 6 Feb 2004, ron minnich wrote:
I sent it to Richard.
I am putting the scripts we use here for info.
ron jotun root # cat FIXGART.sh #!/bin/sh insmod /usr/src/linux/arch/i386/kernel/msr.o # fix the fixed MTRR /root/wrmsr 0xC0010010 0x1e0601 0 /root/wrmsr 0x259 0 0
#fix the IORR1 crap on BOTH northbridges
/root/rdmsr 0xC0010018 /root/wrmsr 0xC0010018 0xd0000000 0xff /root/rdmsr 0xC0010019 /root/wrmsr 0xC0010019 0xd8000000 0xff
# set up 0xd in the whatever setpci -s 0:18.3 94.l setpci -s 0:18.3 94.l=68 setpci -s 0:18.3 94.l setpci -s 0:18.3 90.b setpci -s 0:18.3 90.b=3 setpci -s 0:18.3 90.b setpci -s 0:19.3 94.l setpci -s 0:19.3 94.l=68 setpci -s 0:19.3 94.l setpci -s 0:19.3 90.b setpci -s 0:19.3 90.b=3 setpci -s 0:19.3 90.b # 8151 setpci -s 1:1.0 10.l # setpci -s 1:1.0 10.l=f0000008 # setpci -s 1:1.0 10.l setpci -s 1:1.0 b4.l # setpci -s 1:1.0 b4.l=00010f20 # setpci -s 1:1.0 b4.l jotun root # cat DO_GEFORCEFX.sh #! /bin/sh #insmod /usr/src/linux/arch/i386/kernel/msr.o #/root/wrmsr 0xC0010010 0x1e0601 0 #/root/wrmsr 0x259 0 0 /root/vgabios//testbios --abseg /dev/mem -s 65536 /root/nvbios.bin 2>&1
/tmp/out
modprobe nvidia NVreg_EnableAGPSBA=1 NVreg_EnableAGPFW=1 #modprobe agpgart agp_try_unsupported=1 modprobe agpgart jotun root #
Linuxbios mailing list Linuxbios@clustermatic.org http://www.clustermatic.org/mailman/listinfo/linuxbios
For anyone using the S2885 or another mainboard with the 8151, there seems to be a problem with the agpgart driver in 2.6 kernels. Further description is here: http://www.x86-64.org/lists/discuss/msg04393.html . I could only get Linux agpgart to detect the AMD 8151 when compiled statically into the 2.6.1 kernel with the patch from x86-64.org. This problem also seems to exist in 2.6.2. Using our current approach, agpgart is loaded as a module after Linux is booted and we've run testbios. Looks like 2.6 users might have to wait for Ron to get the aforementioned fixes into LinuxBIOS so we can get run the VGA BIOS before the kernel tries to boot.
On Fri, 6 Feb 2004, Hendricks David W. wrote:
It should have modprobe agpgart agp_try_unsupported=1 uncommented. I had difficulty loading the module with agp_try_unsupported in kernel 2.6.2, looking into it further right now...
On Fri, 6 Feb 2004, ron minnich wrote:
I sent it to Richard.
I am putting the scripts we use here for info.
ron jotun root # cat FIXGART.sh #!/bin/sh insmod /usr/src/linux/arch/i386/kernel/msr.o # fix the fixed MTRR /root/wrmsr 0xC0010010 0x1e0601 0 /root/wrmsr 0x259 0 0
#fix the IORR1 crap on BOTH northbridges
/root/rdmsr 0xC0010018 /root/wrmsr 0xC0010018 0xd0000000 0xff /root/rdmsr 0xC0010019 /root/wrmsr 0xC0010019 0xd8000000 0xff
# set up 0xd in the whatever setpci -s 0:18.3 94.l setpci -s 0:18.3 94.l=68 setpci -s 0:18.3 94.l setpci -s 0:18.3 90.b setpci -s 0:18.3 90.b=3 setpci -s 0:18.3 90.b setpci -s 0:19.3 94.l setpci -s 0:19.3 94.l=68 setpci -s 0:19.3 94.l setpci -s 0:19.3 90.b setpci -s 0:19.3 90.b=3 setpci -s 0:19.3 90.b # 8151 setpci -s 1:1.0 10.l # setpci -s 1:1.0 10.l=f0000008 # setpci -s 1:1.0 10.l setpci -s 1:1.0 b4.l # setpci -s 1:1.0 b4.l=00010f20 # setpci -s 1:1.0 b4.l jotun root # cat DO_GEFORCEFX.sh #! /bin/sh #insmod /usr/src/linux/arch/i386/kernel/msr.o #/root/wrmsr 0xC0010010 0x1e0601 0 #/root/wrmsr 0x259 0 0 /root/vgabios//testbios --abseg /dev/mem -s 65536 /root/nvbios.bin 2>&1
/tmp/out
modprobe nvidia NVreg_EnableAGPSBA=1 NVreg_EnableAGPFW=1 #modprobe agpgart agp_try_unsupported=1 modprobe agpgart jotun root #
Linuxbios mailing list Linuxbios@clustermatic.org http://www.clustermatic.org/mailman/listinfo/linuxbios
Linuxbios mailing list Linuxbios@clustermatic.org http://www.clustermatic.org/mailman/listinfo/linuxbios
ron minnich wrote:
Ok.. No change for me.
A slight update is that I now know what all those int15 calls are and what they do. Except I can't tell cause its NDA stuff.
I can tell you that I added that support into int15.c and it didn't make any differnce.
Somthing is still not quite right with the emulation since both under ADLO and IDT mode I can get a VSYNC but not with the emulator.
I'm going to try and add my int15 stuff to ADLO and IDT and see what happens.
Oh and Ron would it be too much trouble to enable/disable all your extra printfs of ins and outs with the -DDEBUG or perhaps a command line switch. Took me awhile to find that they weren't part of the DEBUG setup.
On Sat, 7 Feb 2004, Richard Smith wrote:
Oh and Ron would it be too much trouble to enable/disable all your extra printfs of ins and outs with the -DDEBUG or perhaps a command line switch. Took me awhile to find that they weren't part of the DEBUG setup.
yeah, sorry, a very quick hack I forgot to take out.
I'll take it out.
ron
Wonderfull news... I got this back today from ATI. Looks like someone with some clout there finally gets it.
So I'm about to start porting some biosless init code I received over to Linux. Then work it into linux bios.
I take back everything I said bad about ATI. *grin*
Tell all your OEM buddies that ATI Mobility M1 video chips are about to become LinuxBIOS friendly in the near future.
------- Sorry for the lengthy delay. There is a lot of detail here. Basically, you can develop code with our documentation, and then release it under GPL - no issues there.
If you want to produce a biosless chip init procedure, you can use the documentation I pointed you to. If you have any questions, feel free to email me about those documents. ---------
Woohoo! Does this also apply to their latest R3xx core Radeons such as the 8500 and up? Or only their old crummy ones?
On Mon, 9 Feb 2004, Richard Smith wrote:
Wonderfull news... I got this back today from ATI. Looks like someone with some clout there finally gets it.
So I'm about to start porting some biosless init code I received over to Linux. Then work it into linux bios.
I take back everything I said bad about ATI. *grin*
Tell all your OEM buddies that ATI Mobility M1 video chips are about to become LinuxBIOS friendly in the near future.
Sorry for the lengthy delay. There is a lot of detail here. Basically, you can develop code with our documentation, and then release it under GPL - no issues there.
If you want to produce a biosless chip init procedure, you can use the documentation I pointed you to. If you have any questions, feel free to email me about those documents.
Hendricks David W. wrote:
Woohoo! Does this also apply to their latest R3xx core Radeons such as the 8500 and up? Or only their old crummy ones?
Bitworks's agreement only covers the Rage Mobility M1 which is several years old and Mach64 based.
I don't know about the new cores. Probally not as they are not trying to sell those into the embedded market.
Hello (again) from Gregg C Levine Richard, I've got a Mach64 chip here, or at least one part of the set, here. Do you know where I can track down the literature for it? I've been resisting designing it into a design because of that problem. And yes, I've got a number of ATI based video cards here, probably both either PCI, or AGP based.
And yes, this is indeed wonderful news. ------------------- Gregg C Levine hansolofalcon@worldnet.att.net ------------------------------------------------------------ "The Force will be with you...Always." Obi-Wan Kenobi "Use the Force, Luke." Obi-Wan Kenobi (This company dedicates this E-Mail to General Obi-Wan Kenobi ) (This company dedicates this E-Mail to Master Yoda )
-----Original Message----- From: linuxbios-admin@clustermatic.org [mailto:linuxbios- admin@clustermatic.org] On Behalf Of Richard Smith Sent: Monday, February 09, 2004 6:16 PM To: dwh@lanl.gov Cc: linuxbios@clustermatic.org Subject: Re: ATI gets a clue.
Hendricks David W. wrote:
Woohoo! Does this also apply to their latest R3xx core Radeons
such as the
8500 and up? Or only their old crummy ones?
Bitworks's agreement only covers the Rage Mobility M1 which is
several
years old and Mach64 based.
I don't know about the new cores. Probally not as they are not
trying
to sell those into the embedded market.
-- Richard A. Smith rsmith@bitworks.com
Linuxbios mailing list Linuxbios@clustermatic.org http://www.clustermatic.org/mailman/listinfo/linuxbios
Gregg C Levine wrote:
Hello (again) from Gregg C Levine
Richard, I've got a Mach64 chip here, or at least one part of the set, here. Do you know where I can track down the literature for it? I've been resisting designing it into a design because of that problem. And yes, I've got a number of ATI based video cards here, probably both either PCI, or AGP based.
You have to set up an NDA with ATI to get the info. I'm allowed to release my _code_ but the docs describing all the registers are protected by the NDA and are not publicly available.
Richard Smith wrote:
You have to set up an NDA with ATI to get the info. I'm allowed to release my _code_ but the docs describing all the registers are protected by the NDA and are not publicly available.
Are they allowing you to release source or just a binary?
If it's source then it gets tricky as to how you comment your code since you're under NDA. There have been several issues that have come up on this list and all I could do was offer a binary solution.
-Bari
Bari Ari wrote:
You have to set up an NDA with ATI to get the info. I'm allowed to release my _code_ but the docs describing all the registers are protected by the NDA and are not publicly available.
Are they allowing you to release source or just a binary?
If it's source then it gets tricky as to how you comment your code since you're under NDA. There have been several issues that have come up on this list and all I could do was offer a binary solution.
I specifically asked about source and I was very detailed in my request so I am interpreting thier response to source really means source. I plan on integrating it into the LinuxBios code for our board so binary only is not an option.
Yes commenting could be abused if I started listing all the registers, thier descriptions and meanings in my comments. I don't plan on commenting any different than I normally would though. Which pretty much only involves heavly commenting things that don't work like the docs say they should.
Most of the info is already present in the atyfb kernel framebuffer driver anyway.
I've requested an "offical" permission statement so I'll just have to wait and see what in that to really know.
Really nice news... i'll keep dreaming that nvidia will do the same.
Wonderfull news... I got this back today from ATI. Looks like someone with some clout there finally gets it.
So I'm about to start porting some biosless init code I received over to Linux. Then work it into linux bios.
I take back everything I said bad about ATI. *grin*
Tell all your OEM buddies that ATI Mobility M1 video chips are about to become LinuxBIOS friendly in the near future.
Sorry for the lengthy delay. There is a lot of detail here. Basically, you can develop code with our documentation, and then release it under GPL - no issues there.
If you want to produce a biosless chip init procedure, you can use the documentation I pointed you to. If you have any questions, feel free to email me about those documents.
-- Richard A. Smith rsmith@bitworks.com
Linuxbios mailing list Linuxbios@clustermatic.org http://www.clustermatic.org/mailman/listinfo/linuxbios
On Mon, Feb 09, 2004 at 04:42:37PM -0600, Richard Smith wrote:
Wonderfull news... I got this back today from ATI. Looks like someone with some clout there finally gets it.
This is great news :)
I remain in awe at what has been achieved in the time I have been lurking on this list.
congrats to everyone working on linuxbios,
v
-
Bari Ari -
Carlos Silva -
ebiederman＠lnxi.com -
Gregg C Levine -
Hendricks David W. -
Li-Ta Lo -
Peter Stuge -
Richard Smith -
ron minnich -
Stefan Reinauer -
Vincent Touquet