[LinuxBIOS] Build broken?

[beneo]

Hi, 
I download a lateset snapshot from linuxBIOS.org yesterday. I don't know if it is bad luck or something, I can not successfully build any BIOS from the tree I downloaded. I tried following three platform.

tyan/s2891
tyan/s2895
amd/serenade

Does anybody has similar issue or know what happened?

Below are the details on the things I got for these three different target platform

1. Both tyan/s2891 and tyan/s2895 have similar behavior

I changed a few lines in Options.lb:

default USE_DCACHE_RAM=1
default CONFIG_USE_INIT=1
default DEFAULT_CONSOLE_LOGLEVEL=7
default  MAXIMUM_CONSOLE_LOGLEVEL=7

I get a error output like below: (Note: Changing options in the Options.lb above were not the cuase of the problem. Without changing USE_DCACHE_RAM and CONFIG_USE_INIT, I still get same results)

------------
gcc -m32 -nostdlib -nostartfiles -static -o linuxbios_ram -T /export/1/new/src/config/linuxbios_ram.ld linuxbios_ram.o
linuxbios_ram.o(.text+0x13db9): In function `smp_write_config_table':
: undefined reference to `get_apicid_base'
collect2: ld returned 1 exit status
make[1]: *** [linuxbios_ram] Error 1
make[1]: Leaving directory `/export/1/new/targets/tyan/s2895/s2895/normal'
make: *** [normal/linuxbios.rom] Error 1

2. amd/serenade gets the build error like below. It probably is the side product of the tree enabled dual core and cache as RAM?
-----
gcc -m32 -nostdlib -nostartfiles -static -o linuxbios_ram -T /export/1/new/src/config/linuxbios_ram.ld linuxbios_ram.o
linuxbios_ram.o(.text+0x14d2): In function `model_fxx_init':
: undefined reference to `get_node_core_id'
linuxbios_ram.o(.text+0x14e6): In function `model_fxx_init':
: undefined reference to `amd_sibling_init'
collect2: ld returned 1 exit status
make[1]: *** [linuxbios_ram] Error 1
make[1]: Leaving directory `/export/1/new/targets/amd/serenade/serenade/normal'
make: *** [normal/linuxbios.rom] Error 1

Any help would be appriciated.

Thanks

Beneo

[David Hendricks]

Probably bad luck. This has been known to happen, and IIRC one of the
points at the recent LinuxBIOS summit was to post 'stable' tarballs
where everything should build without hassle, even if it means rolling
back some features. Stay tuned...

On Tue, 8 Nov 2005 17:01:37 -0800
"beneo"  wrote:
> Hi, 
> I download a lateset snapshot from linuxBIOS.org yesterday. I don't
> know if it is bad luck or something, I can not successfully build any
> BIOS from the tree I downloaded. I tried following three platform.
>
> tyan/s2891
> tyan/s2895
> amd/serenade
>
> Does anybody has similar issue or know what happened?
>
> Below are the details on the things I got for these three different
> target platform
> Both tyan/s2891 and tyan/s2895 have similar behavior
>
> I changed a few lines in Options.lb:
>
> default USE_DCACHE_RAM=1
> default CONFIG_USE_INIT=1
> default DEFAULT_CONSOLE_LOGLEVEL=7
> default  MAXIMUM_CONSOLE_LOGLEVEL=7
>
> I get a error output like below: (Note: Changing options in the
> Options.lb above were not the cuase of the problem. Without changing
> USE_DCACHE_RAM and CONFIG_USE_INIT, I still get same results)
>
> gcc -m32 -nostdlib -nostartfiles -static -o linuxbios_ram
> -T /export/1/new/src/config/linuxbios_ram.ld linuxbios_ram.o
> linuxbios_ram.o(.text+0x13db9): In function
> `smp_write_config_table': : undefined reference to `get_apicid_base'
> collect2: ld returned 1 exit status make[1]: *** [linuxbios_ram]
> Error 1 make[1]: Leaving directory
> `/export/1/new/targets/tyan/s2895/s2895/normal' make: ***
> [normal/linuxbios.rom] Error 1
> amd/serenade gets the build error like below. It probably is the
>
> side product of the tree enabled dual core and cache as RAM? -----
> gcc -m32 -nostdlib -nostartfiles -static -o linuxbios_ram
> -T /export/1/new/src/config/linuxbios_ram.ld linuxbios_ram.o
> linuxbios_ram.o(.text+0x14d2): In function `model_fxx_init': :
> undefined reference to `get_node_core_id'
> linuxbios_ram.o(.text+0x14e6): In function `model_fxx_init': :
> undefined reference to `amd_sibling_init' collect2: ld returned 1
> exit status make[1]: *** [linuxbios_ram] Error 1
> make[1]: Leaving directory
> `/export/1/new/targets/amd/serenade/serenade/normal' make: ***
> [normal/linuxbios.rom] Error 1
>
> Any help would be appriciated.
>
> Thanks
>
> Beneo

[Beneo]

David,

Do you know a tar ball with recent date that is relatively good that I can
download?

Thanks

Beneo

----- Original Message ----- 
From: "David Hendricks" 
To: 
Sent: Tuesday, November 08, 2005 9:26 PM
Subject: Re: [LinuxBIOS] Build broken?
> Probably bad luck. This has been known to happen, and IIRC one of the
> points at the recent LinuxBIOS summit was to post 'stable' tarballs
> where everything should build without hassle, even if it means rolling
> back some features. Stay tuned...
>
> On Tue, 8 Nov 2005 17:01:37 -0800
> "beneo"  wrote:
> > Hi,
> > I download a lateset snapshot from linuxBIOS.org yesterday. I don't
> > know if it is bad luck or something, I can not successfully build any
> > BIOS from the tree I downloaded. I tried following three platform.
> >
> > tyan/s2891
> > tyan/s2895
> > amd/serenade
> >
> > Does anybody has similar issue or know what happened?
> >
> > Below are the details on the things I got for these three different
> > target platform
> > Both tyan/s2891 and tyan/s2895 have similar behavior
> >
> > I changed a few lines in Options.lb:
> >
> > default USE_DCACHE_RAM=1
> > default CONFIG_USE_INIT=1
> > default DEFAULT_CONSOLE_LOGLEVEL=7
> > default  MAXIMUM_CONSOLE_LOGLEVEL=7
> >
> > I get a error output like below: (Note: Changing options in the
> > Options.lb above were not the cuase of the problem. Without changing
> > USE_DCACHE_RAM and CONFIG_USE_INIT, I still get same results)
> >
> > gcc -m32 -nostdlib -nostartfiles -static -o linuxbios_ram
> > -T /export/1/new/src/config/linuxbios_ram.ld linuxbios_ram.o
> > linuxbios_ram.o(.text+0x13db9): In function
> > `smp_write_config_table': : undefined reference to `get_apicid_base'
> > collect2: ld returned 1 exit status make[1]: *** [linuxbios_ram]
> > Error 1 make[1]: Leaving directory
> > `/export/1/new/targets/tyan/s2895/s2895/normal' make: ***
> > [normal/linuxbios.rom] Error 1
> > amd/serenade gets the build error like below. It probably is the
> >
> > side product of the tree enabled dual core and cache as RAM? -----
> > gcc -m32 -nostdlib -nostartfiles -static -o linuxbios_ram
> > -T /export/1/new/src/config/linuxbios_ram.ld linuxbios_ram.o
> > linuxbios_ram.o(.text+0x14d2): In function `model_fxx_init': :
> > undefined reference to `get_node_core_id'
> > linuxbios_ram.o(.text+0x14e6): In function `model_fxx_init': :
> > undefined reference to `amd_sibling_init' collect2: ld returned 1
> > exit status make[1]: *** [linuxbios_ram] Error 1
> > make[1]: Leaving directory
> > `/export/1/new/targets/amd/serenade/serenade/normal' make: ***
> > [normal/linuxbios.rom] Error 1
> >
> > Any help would be appriciated.
> >
> > Thanks
> >
> > Beneo

[Stefan Reinauer]

* Beneo  [051109 19:16]:
> David,
>
> Do you know a tar ball with recent date that is relatively good that I can
> download?

http://snapshots.linuxbios.org/

rev 2064 should be relatively good.

Stefan

[beneo]

I tried rev 2064, on two targets: tyan/s2891 and amd/serenade, both gave me
an error like this:

---
gcc -m32 -nostdlib -nostartfiles -static -o linuxbios -T ldscript.ld crt0.o
/usr/bin/ld: section .reset [fffdfff0 -> fffdffff] overlaps section .rom
[fffd7da8 -> fffe08df]
/usr/bin/ld: section .id [fffdffd7 -> fffdffef] overlaps section .rom
[fffd7da8 -> fffe08df]
collect2: ld returned 1 exit status
make[1]: *** [linuxbios] Error 1
---

Do I need to do some ajustment? or do I need to try other revision?

Thanks

Beneo

----- Original Message ----- 
From: "Stefan Reinauer" 
To: "Beneo" 
Cc: "David Hendricks" ; 
Sent: Wednesday, November 09, 2005 10:24 AM
Subject: Re: [LinuxBIOS] Build broken?
> Beneo  [051109 19:16]:
> > David,
> >
> > Do you know a tar ball with recent date that is relatively good that I

can
> > download?
>
> http://snapshots.linuxbios.org/
>
> rev 2064 should be relatively good.
>
> Stefan

[Ronald G Minnich]

beneo wrote:
> I tried rev 2064, on two targets: tyan/s2891 and amd/serenade, both gave me
> an error like this:
>
> gcc -m32 -nostdlib -nostartfiles -static -o linuxbios -T ldscript.ld crt0.o
> /usr/bin/ld: section .reset [fffdfff0 -> fffdffff] overlaps section .rom
> [fffd7da8 -> fffe08df]
> /usr/bin/ld: section .id [fffdffd7 -> fffdffef] overlaps section .rom
> [fffd7da8 -> fffe08df]
> collect2: ld returned 1 exit status
> make[1]: *** [linuxbios] Error 1

This is the 'your flash part is too small or your payload is too big' 
error. I apologize for the lousy error diagnostic, this is on our 
Post-sc2005 list of things to fix.

ron

[beneo]

Thanks for the pointer. I tried it, but still has some issues, hope a guru
can point me a little more.

I changed the ROM_IMAGE_SIZE to 128k from 64K for both normal and fall back,
the PAYLOAD_SIZE changed to 128K too.

But I still get a error message like this:
-----
objcopy --gap-fill 0xff -O binary linuxbios linuxbios.strip
gcc -o buildrom
/root/dl/lb/v2-2064/LinuxBIOSv2-2064/util/buildrom/buildrom.c
./buildrom linuxbios.strip linuxbios.rom /export/1/payloads/filo.zelf
0x20000 0x40000
linuxbios image is 131080 bytes; only 131072 allowed
Linuxbios input file larger than allowed size!
-----

This only happens to my fallback image. I traced the code, indeed, My
linuxbios.strip file has a size = 131080, which is too large. But the size
of my linuxbios file was only 124570. the objcopy seems created a too large
linuxbios.strip for me.

More interesting thing is that I have a older tree, (8 month older tree) In
that tree, my linuxbios file size was 122982, the objcopy successfully
created a linuxbios.strip file with a size 65536. It must used some
compression. Since I don't know much about objcopy, I can not tell what it
used. The only difference I can tell is that objcopy was look like this:
objcopy -O binary linuxbios linuxbios.strip. (Note: without "--gap-fill 0xff
" option, but don't know if it would make any difference)

Does anybody know what I did wrong or missed?

By the way, my linuxbios_ram.rom size is 31614, crt0.o size is 87388.
Payload size is 42784.

Thanks

beneo

----- Original Message ----- 
From: "Ronald G Minnich" 
To: "beneo" 
Cc: 
Sent: Wednesday, November 09, 2005 12:00 PM
Subject: Re: [LinuxBIOS] Build broken?
> beneo wrote:
> > I tried rev 2064, on two targets: tyan/s2891 and amd/serenade, both gave

me
> > an error like this:
> >
> > gcc -m32 -nostdlib -nostartfiles -static -o linuxbios -T ldscript.ld

crt0.o
> > /usr/bin/ld: section .reset [fffdfff0 -> fffdffff] overlaps section .rom
> > [fffd7da8 -> fffe08df]
> > /usr/bin/ld: section .id [fffdffd7 -> fffdffef] overlaps section .rom
> > [fffd7da8 -> fffe08df]
> > collect2: ld returned 1 exit status
> > make[1]: *** [linuxbios] Error 1
>
> This is the 'your flash part is too small or your payload is too big'
> error. I apologize for the lousy error diagnostic, this is on our
> Post-sc2005 list of things to fix.
>
> ron