OK, I've committed code that turns on shadow dram in F segment, so PIRQ will work; also committed what should be working SMBUS code in vt8231_early_smbus.c; now the fun begins. I could use advice on: - what SMBUS bytes we need to care about - what we should program the north with using those bytes. - how to do it.
Which is to say, I think I know, but let's get this done in a speedy fashion. Momentum is on our side here.
The EPIA will be a good demo platform.
Why this sudden burst of EPIA work from me? Two reasons: - I want to encourage people to look at V2, so we need a demo system less complex than the K8 systems. - We are teaching a tutorial on linuxbios, bproc, etc. and will be doing it hands-on with a roomful of EPIAs, 2 or 3 per tutorial attendee. See: http://lacsi.lanl.gov/symposium/agenda.shtml Come if you want. You can see how we build clusters that boot in seconds. The software we're showing is the same stuff that is running on our 1024-node Pink cluster, and will run on our 1408-node K8 cluster.
ron
At the end of the compilation, I get this.
gcc -nostdlib -nostartfiles -static -o linuxbios -T ldscript.ld crt0.o nm -n linuxbios | sort > linuxbios.map objcopy -O binary linuxbios linuxbios.strip gcc -o buildrom /home/randall/src/linuxbios/freebios2/util/buildrom/buildrom.c ./buildrom linuxbios.strip linuxbios.rom ../../../../tg3--ide_disk.zelf 0x10000 0x20000 ../../../../tg3--ide_disk.zelf: No such file or directory make[1]: *** [linuxbios.rom] Error 2 make[1]: Leaving directory `/home/randall/src/linuxbios/freebios2/targets/via/epia/epia/fallback' make: *** [fallback-rom] Error 1
tg3--ide-disk.zelf was never built. Is this missing:
from the buildrom output, tg3--ide-disk.zelf is supposed to be the payload.
Any ideas?
On Tue September 30 2003 8:52 pm, Randall H. Craig wrote:
tg3--ide-disk.zelf was never built. Is this missing:
from the buildrom output, tg3--ide-disk.zelf is supposed to be the payload.
Any ideas?
I am guessing that this is an etherboot image. What is the pci.ids for the ethernet device?
Cheers,
Randall
On Tue, 30 Sep 2003, Randall H. Craig wrote:
At the end of the compilation, I get this.
gcc -nostdlib -nostartfiles -static -o linuxbios -T ldscript.ld crt0.o nm -n linuxbios | sort > linuxbios.map objcopy -O binary linuxbios linuxbios.strip gcc -o buildrom /home/randall/src/linuxbios/freebios2/util/buildrom/buildrom.c ./buildrom linuxbios.strip linuxbios.rom ../../../../tg3--ide_disk.zelf 0x10000 0x20000 ../../../../tg3--ide_disk.zelf: No such file or directory make[1]: *** [linuxbios.rom] Error 2 make[1]: Leaving directory `/home/randall/src/linuxbios/freebios2/targets/via/epia/epia/fallback' make: *** [fallback-rom] Error 1
tg3--ide-disk.zelf was never built. Is this missing:
from the buildrom output, tg3--ide-disk.zelf is supposed to be the payload.
I'll send you that payload under a seperate note.
but it's only useful for loading an elfimage from an ide-flash. I have no via-rhine etherboot, which is what you need.
ron
Hi Ron,
I'm able to build and run LinuxBIOS on EPIA with some treaking on Config.lb. However, it fails to jump to payload(etherboot). I suspect LinuxBIOS is jumping to a wrong location. Is it possible to set ZKERNEL_START in v2? BTW, ROM_SIZE for default setup should be 262144 instead of 512k, since most EPIA doesn't come with 512k flash.
-Andrew
On Tue, Sep 30, 2003 at 08:08:09PM -0600, ron minnich wrote:
OK, I've committed code that turns on shadow dram in F segment, so PIRQ will work; also committed what should be working SMBUS code in vt8231_early_smbus.c; now the fun begins. I could use advice on:
- what SMBUS bytes we need to care about
- what we should program the north with using those bytes.
- how to do it.
Which is to say, I think I know, but let's get this done in a speedy fashion. Momentum is on our side here.
The EPIA will be a good demo platform.
Why this sudden burst of EPIA work from me? Two reasons:
- I want to encourage people to look at V2, so we need a demo system less complex than the K8 systems.
- We are teaching a tutorial on linuxbios, bproc, etc. and will be doing it hands-on with a roomful of EPIAs, 2 or 3 per tutorial attendee. See: http://lacsi.lanl.gov/symposium/agenda.shtml Come if you want. You can see how we build clusters that boot in seconds. The software we're showing is the same stuff that is running on our 1024-node Pink cluster, and will run on our 1408-node K8 cluster.
ron
Linuxbios mailing list Linuxbios@clustermatic.org http://www.clustermatic.org/mailman/listinfo/linuxbios
On Wed, 1 Oct 2003, Andrew Ip wrote:
I'm able to build and run LinuxBIOS on EPIA with some treaking on Config.lb.
send me the tweaks, although they should not have been needed.
However, it fails to jump to payload(etherboot). I suspect LinuxBIOS is jumping to a wrong location. Is it possible to set ZKERNEL_START in v2?
ZKERNEL_START is not really needed any more. I don't think. Am I wrong?
BTW, ROM_SIZE for default setup should be 262144 instead of 512k, since most EPIA doesn't come with 512k flash.
OK, I will try to commit that tomorrow.
please send me output.
ron
Hi Ron,
Are you able to use etherboot as payload?
ZKERNEL_START is not really needed any more. I don't think. Am I wrong?
Is it changed to CONFIG_ROM_STREAM_START?
send me the tweaks, although they should not have been needed.
It is just the size. I can commit it now. We can say LinuxBIOS is a 24x7 project. :)
-Andrew
Here is the part of the log. ============================================================================ Copying IRQ routing tables to 0xf0000...done. Verifing priq routing tables copy at 0xf0000...failed Wrote linuxbios table at: 00000500 - 00000aec checksum 7d48
Welcome to elfboot, the open sourced starter. January 2002, Eric Biederman. Version 1.3
23:stream_init() - rom_stream: 0xfffe0000 - 0xfffeffff Found ELF candiate at offset 0 header_offset is 0 Try to load at offset 0x0 malloc Enter, size 32, free_mem_ptr 0001432c malloc 0x0001432c New segment addr 0x400000 size 0xaec0 offset 0x60 filesize 0x6fa1 (cleaned up) New segment addr 0x400000 size 0xaec0 offset 0x60 filesize 0x6fa1 lb: [0x0000000000004000, 0x0000000000017b4c) Loading Segment: addr: 0x0000000000400000 memsz: 0x000000000000aec0 filesz: 0x01[ 0x0000000000400000, 0000000000406fa1, 0x000000000040aec0) <- 0000000000000060 Clearing Segment: addr: 0x0000000000406fa1 memsz: 0x0000000000003f1f Loaded segments verified segments closed down stream Jumping to boot code at 0x400000 entry = 0x00400000 lb_start = 0x00004000 lb_size = 0x00013b4c adjust = 0x03fe84b4 buffer = 0x03fd8968 elf_boot_notes = 0x0000d400 adjusted_boot_notes = 0x03ff58b4 ROM segment 0x0000 length 0x0000 reloc 0x0000 Etherboot 5.0.6 (GPL) ELF for [VIA 86C100] Boot from (N)etwork or from (L)ocal? clocks_per_tick = 504240 N I am now initializing the ide system
============================================================================
On Tue, Sep 30, 2003 at 11:36:25PM -0600, ron minnich wrote:
On Wed, 1 Oct 2003, Andrew Ip wrote:
I'm able to build and run LinuxBIOS on EPIA with some treaking on Config.lb.
However, it fails to jump to payload(etherboot). I suspect LinuxBIOS is jumping to a wrong location. Is it possible to set ZKERNEL_START in v2?
ZKERNEL_START is not really needed any more. I don't think. Am I wrong?
BTW, ROM_SIZE for default setup should be 262144 instead of 512k, since most EPIA doesn't come with 512k flash.
OK, I will try to commit that tomorrow.
please send me output.
ron
Changes:
new epia target for 512k: targets/via/epia/Config.512kflash.lb
epia defaults to 256k flash
buildtarget now takes either a directory, and uses directory/Config.lb, or takes a file
e.g.
buildtarget via/epia will use via/epia/Config.lb, and buildtarget via/epia/Config.512kflash.lb will use that file.
ron
don't commit! I fixed that for you already.
ron
OK, for VGA, here we go again.
I just found out that a well-known linuxbios company ships their boards in some cases with the old bios in it because .... the field engineers want vga.
Here is my plan.
I will combine vgabios + idt into one file, put it in pc80, make it a standard device, have it turn itself on with the standard static configuration techniques, and we'll have vga.
The understanding here is that this is code that we hope to deprecate in favor of the emulation that Stefan is working on. But the pressure to do something in the short term is getting irresistable, this code works, so I think we take a short term focus to help the user community.
I'll try to get this in this week.
ron
On Wed, 1 Oct 2003, Andrew Ip wrote:
ZKERNEL_START is not really needed any more. I don't think. Am I wrong?
Is it changed to CONFIG_ROM_STREAM_START?
Think so. Those names are terrible (we need to fix them) but I think that's it.
yes. You may need to get the latest cvs ... I have a few important fixes in there. I just put them in tonight, not tested (I can't) but should work.
ron
ron minnich rminnich@lanl.gov writes:
OK, for VGA, here we go again.
Here is my plan.
I will combine vgabios + idt into one file, put it in pc80, make it a standard device, have it turn itself on with the standard static configuration techniques, and we'll have vga.
The understanding here is that this is code that we hope to deprecate in favor of the emulation that Stefan is working on. But the pressure to do something in the short term is getting irresistable, this code works, so I think we take a short term focus to help the user community.
I'll try to get this in this week.
Ok. That sounds like a plan. As long as it does not require special cases in the configuration it should impose no extra support burden.
Adam Agnew has pointed me at a frame buffer driver for the ATI Rage XL that should be capable of initializing the card without bios support. If that works we have an even better solution.
So that would make an even more elegant driver for the cards that have on board VGA support. And given that the ATI Rage XL is the most common chip I have seen on motherboards with LinuxBIOS it should be a big help.
Eric
I have via-rhine etherboot and if anyone need it, I can mail it to him.
----- Original Message ----- From: "ron minnich" rminnich@lanl.gov To: "Randall H. Craig" randall@tdl.com Cc: linuxbios@clustermatic.org Sent: Wednesday, October 01, 2003 1:03 PM Subject: Re: V2 VIA EPIA Build Errors
On Tue, 30 Sep 2003, Randall H. Craig wrote:
At the end of the compilation, I get this.
gcc -nostdlib -nostartfiles -static -o linuxbios -T ldscript.ld crt0.o nm -n linuxbios | sort > linuxbios.map objcopy -O binary linuxbios linuxbios.strip gcc -o buildrom /home/randall/src/linuxbios/freebios2/util/buildrom/buildrom.c ./buildrom linuxbios.strip linuxbios.rom ../../../../tg3--ide_disk.zelf 0x10000 0x20000 ../../../../tg3--ide_disk.zelf: No such file or directory make[1]: *** [linuxbios.rom] Error 2 make[1]: Leaving directory `/home/randall/src/linuxbios/freebios2/targets/via/epia/epia/fallback' make: *** [fallback-rom] Error 1
tg3--ide-disk.zelf was never built. Is this missing:
from the buildrom output, tg3--ide-disk.zelf is supposed to be the payload.
I'll send you that payload under a seperate note.
but it's only useful for loading an elfimage from an ide-flash. I have no via-rhine etherboot, which is what you need.
ron
Linuxbios mailing list Linuxbios@clustermatic.org http://www.clustermatic.org/mailman/listinfo/linuxbios
On 1 Oct 2003, Eric W. Biederman wrote:
Adam Agnew has pointed me at a frame buffer driver for the ATI Rage XL that should be capable of initializing the card without bios support. If that works we have an even better solution.
That's been my hope all along -- that we could do this. But nvdia has made it clear, again and again, that they won't play, and nvidia is too important to ignore.
thanks
ron
Oh, I can see what is going to happen to buildrom now that Ron has discovered it... :-)
Greg
At 11:44 PM -0600 30/9/03, ron minnich wrote:
Changes:
new epia target for 512k: targets/via/epia/Config.512kflash.lb
epia defaults to 256k flash
buildtarget now takes either a directory, and uses directory/Config.lb, or takes a file
e.g.
buildtarget via/epia will use via/epia/Config.lb, and buildtarget via/epia/Config.512kflash.lb will use that file.
ron
Linuxbios mailing list Linuxbios@clustermatic.org http://www.clustermatic.org/mailman/listinfo/linuxbios
On 1 Oct 2003, Eric W. Biederman wrote:
Right. However I think we can start setting the precedent that onboard video is handled via open source code. Plug in video cards are an entirely different problem.
We can set this precedent, and we can try to cleave to it as best we can. But the market in this case will control our actions more than we might like.
I don't like it; you know my feelings on this going back four years ago: I wanted to figure out those VGA registers and do all open source. But that didn't quite work out.
ron
Andrew, you should be able to build a working linuxbios image for EPIA by:
cd targets ./buildtarget via/epia etc.
If you can try to figure out why the IRQ table is not working, that would help. I'm stuck. It's the same table as on V1, it checks out when copied to ram, but Linux is just unhappy with it. Not sure why. If anyone wants to take a look that would help.
ron
ron minnich rminnich@lanl.gov writes:
On 1 Oct 2003, Eric W. Biederman wrote:
Adam Agnew has pointed me at a frame buffer driver for the ATI Rage XL that should be capable of initializing the card without bios support. If that works we have an even better solution.
That's been my hope all along -- that we could do this. But nvdia has made it clear, again and again, that they won't play, and nvidia is too important to ignore.
Right. However I think we can start setting the precedent that onboard video is handled via open source code. Plug in video cards are an entirely different problem.
Eric
-
Andrew Ip -
ebiederman＠lnxi.com -
Greg Watson -
Randall H. Craig -
rimy2000 -
ron minnich