[LinuxBIOS] compile errors on tyan S2881

[Ward Vandewege]

Hi there,

We have a couple of Tyan S2881 boxes that we want to run linuxbios on.

I checked out the latest SVN repository (revision 2117), but get compile
errors for the S2881 tree.

This is the case on Debian testing/unstable with GCC 3.4.5 20051015
(prerelease) (Debian 3.4.4-9) and GCC 4.0.2 (Debian 4.0.2-2), as well as on
Ubuntu Breezy with GCC 4.0.2 20050808 (prerelease) (Ubuntu 4.0.1-4ubuntu9).

There are a lot more warnings when compiling with GCC 4, but it always dies
on this:

LinuxBIOSv2/src/cpu/amd/car/disable_cache_as_ram.c: At top level:
  LinuxBIOSv2/src/cpu/amd/car/disable_cache_as_ram.c:3: error:
    syntax error before 'volatile'

Am I missing a dependency, or is this a bug?

Any help would be greatly appreciated.

Thanks,
Ward.

[yhlu]

please go
https://openbios.org/roundup/linuxbios/issue41

to get the patch for that..

Also I have add inine attribute((__always_inline__))  to some must
inline function to make it works gcc already....

We will apply the patch into the public tree soon....

YH

On 12/2/05, Ward Vandewege  wrote:
> Hi there,
>
> We have a couple of Tyan S2881 boxes that we want to run linuxbios on.
>
> I checked out the latest SVN repository (revision 2117), but get compile
> errors for the S2881 tree.
>
> This is the case on Debian testing/unstable with GCC 3.4.5 20051015
> (prerelease) (Debian 3.4.4-9) and GCC 4.0.2 (Debian 4.0.2-2), as well as on
> Ubuntu Breezy with GCC 4.0.2 20050808 (prerelease) (Ubuntu 4.0.1-4ubuntu9).
>
> There are a lot more warnings when compiling with GCC 4, but it always dies
> on this:
>
> LinuxBIOSv2/src/cpu/amd/car/disable_cache_as_ram.c: At top level:
>  LinuxBIOSv2/src/cpu/amd/car/disable_cache_as_ram.c:3: error:
>    syntax error before 'volatile'
>
> Am I missing a dependency, or is this a bug?
>
> Any help would be greatly appreciated.
>
> Thanks,
> Ward.

[Ward Vandewege]

On Fri, Dec 02, 2005 at 12:36:23PM -0800, yhlu wrote:
> please go
> https://openbios.org/roundup/linuxbios/issue41
>
> to get the patch for that..
>
> Also I have add inine attribute((__always_inline__))  to some must
> inline function to make it works gcc already....
>
> We will apply the patch into the public tree soon....

Thanks! That helped it get past that particular error. Now it dies during
linking, though...

gcc -m32 -nostdlib -nostartfiles -static -o linuxbios -T ldscript.ld crt0.o init.o
  /usr/bin/ld: section .rom [00000000fffb0000 -> 00000000fffb011f] overlaps section .ram [00000000fffaa000 -> 00000000fffb2fcc]
  /usr/bin/ld: section .init [00000000fffb0120 -> 00000000fffb5faf] overlaps section .ram [00000000fffaa000 -> 00000000fffb2fcc]
  /usr/bin/ld: linuxbios: section .rom lma 0xfffb0000 overlaps previous sections
  / usr/bin/ld: linuxbios: section .init lma 0xfffb0120 overlaps previous sections
  /usr/bin/ld: linuxbios: section .data lma 0xfffb2fd0 overlaps previous sections
  /usr/bin/ld: linuxbios: section .init.data lma 0xfffb2fd0 overlaps previous sections
  collect2: ld returned 1 exit status
  make[1]: *** [linuxbios] Error 1

Thanks,
Ward.
> On 12/2/05, Ward Vandewege  wrote:
> > Hi there,
> >
> > We have a couple of Tyan S2881 boxes that we want to run linuxbios on.
> >
> > I checked out the latest SVN repository (revision 2117), but get compile
> > errors for the S2881 tree.
> >
> > This is the case on Debian testing/unstable with GCC 3.4.5 20051015
> > (prerelease) (Debian 3.4.4-9) and GCC 4.0.2 (Debian 4.0.2-2), as well as on
> > Ubuntu Breezy with GCC 4.0.2 20050808 (prerelease) (Ubuntu 4.0.1-4ubuntu9).
> >
> > There are a lot more warnings when compiling with GCC 4, but it always dies
> > on this:
> >
> > LinuxBIOSv2/src/cpu/amd/car/disable_cache_as_ram.c: At top level:
> >  LinuxBIOSv2/src/cpu/amd/car/disable_cache_as_ram.c:3: error:
> >    syntax error before 'volatile'
> >
> > Am I missing a dependency, or is this a bug?
> >
> > Any help would be greatly appreciated.
> >
> > Thanks,
> > Ward.
>
> !DSPAM:4390b0e942921951315283!

[Richard Smith]

> /usr/bin/ld: linuxbios: section .rom lma 0xfffb0000 overlaps previous sections
>   / usr/bin/ld: linuxbios: section .init lma 0xfffb0120 overlaps previous sections
>   /usr/bin/ld: linuxbios: section .data lma 0xfffb2fd0 overlaps previous sections
>   /usr/bin/ld: linuxbios: section .init.data lma 0xfffb2fd0 overlaps previous sections
>   collect2: ld returned 1 exit status
>   make[1]: *** [linuxbios] Error 1

I don't know a whole lot about that board but normally that means your
ROM_IMAGE_SIZE is too small.

[yhlu]

what is your payload?

I mean the payload definition in target/tyan/s2881/Config.lb

Please try to disable VGA in MB Config.lb and remve the
line in target Config: Option ROM_SIZE=475136

YH

On 12/2/05, Ward Vandewege  wrote:
> On Fri, Dec 02, 2005 at 12:36:23PM -0800, yhlu wrote:
> > please go
> > https://openbios.org/roundup/linuxbios/issue41
> >
> > to get the patch for that..
> >
> > Also I have add inine attribute((__always_inline__))  to some must
> > inline function to make it works gcc already....
> >
> > We will apply the patch into the public tree soon....
>
> Thanks! That helped it get past that particular error. Now it dies during
> linking, though...
>
> gcc -m32 -nostdlib -nostartfiles -static -o linuxbios -T ldscript.ld crt0.o init.o
>  /usr/bin/ld: section .rom [00000000fffb0000 -> 00000000fffb011f] overlaps section .ram [00000000fffaa000 -> 00000000fffb2fcc]
>  /usr/bin/ld: section .init [00000000fffb0120 -> 00000000fffb5faf] overlaps section .ram [00000000fffaa000 -> 00000000fffb2fcc]
>  /usr/bin/ld: linuxbios: section .rom lma 0xfffb0000 overlaps previous sections
>  / usr/bin/ld: linuxbios: section .init lma 0xfffb0120 overlaps previous sections
>  /usr/bin/ld: linuxbios: section .data lma 0xfffb2fd0 overlaps previous sections
>  /usr/bin/ld: linuxbios: section .init.data lma 0xfffb2fd0 overlaps previous sections
>  collect2: ld returned 1 exit status
>  make[1]: *** [linuxbios] Error 1
>
> Thanks,
> Ward.
> > On 12/2/05, Ward Vandewege  wrote:
> > > Hi there,
> > >
> > > We have a couple of Tyan S2881 boxes that we want to run linuxbios on.
> > >
> > > I checked out the latest SVN repository (revision 2117), but get compile
> > > errors for the S2881 tree.
> > >
> > > This is the case on Debian testing/unstable with GCC 3.4.5 20051015
> > > (prerelease) (Debian 3.4.4-9) and GCC 4.0.2 (Debian 4.0.2-2), as well as on
> > > Ubuntu Breezy with GCC 4.0.2 20050808 (prerelease) (Ubuntu 4.0.1-4ubuntu9).
> > >
> > > There are a lot more warnings when compiling with GCC 4, but it always dies
> > > on this:
> > >
> > > LinuxBIOSv2/src/cpu/amd/car/disable_cache_as_ram.c: At top level:
> > >  LinuxBIOSv2/src/cpu/amd/car/disable_cache_as_ram.c:3: error:
> > >    syntax error before 'volatile'
> > >
> > > Am I missing a dependency, or is this a bug?
> > >
> > > Any help would be greatly appreciated.
> > >
> > > Thanks,
> > > Ward.
> >
> > !DSPAM:4390b0e942921951315283!

[yhlu]

the ROM_IMAGE_SIZE should be 0x20000 in target MB Config.

Please checkt that

You seem set that to ROM_IMAGE_SIZE to 0x16000, So change that to
0x20000 will solve the problem...

YH

On 12/2/05, Richard Smith  wrote:
> > /usr/bin/ld: linuxbios: section .rom lma 0xfffb0000 overlaps previous sections
> >   / usr/bin/ld: linuxbios: section .init lma 0xfffb0120 overlaps previous sections
> >   /usr/bin/ld: linuxbios: section .data lma 0xfffb2fd0 overlaps previous sections
> >   /usr/bin/ld: linuxbios: section .init.data lma 0xfffb2fd0 overlaps previous sections
> >   collect2: ld returned 1 exit status
> >   make[1]: *** [linuxbios] Error 1
>
> I don't know a whole lot about that board but normally that means your
> ROM_IMAGE_SIZE is too small.

[Stefan Reinauer]

* yhlu  [051202 22:32]:
> what is your payload?
>
> I mean the payload definition in target/tyan/s2881/Config.lb
>
> Please try to disable VGA in MB Config.lb and remve the
> line in target Config: Option ROM_SIZE=475136

Is that no longer possible?

Stefan

[Stefan Reinauer]

* yhlu  [051202 22:37]:
> the ROM_IMAGE_SIZE should be 0x20000 in target MB Config.

Can you please explain this again? Does this mean that linuxbios+payload
is at least 128KB?

Or is this LinuxBIOS without payload?

No LinuxBIOS Normal+Fallback on machines with 128kb only anymore?

Stefan

[yhlu]

In MB Option.lb, We define FALLBACK_SIZE=0x40000 ===> it means 256K
for Fallback image.

In the Fallback last 64K will be linuxbios_rom(crt0.S and init.o),
In Traget you can set ROM_IMAGE_SIZE to LinuxBIOS_RAM + LinuxBIOS_rom
So it could be 0x20000, or more 0x30000, then linuxbios_ram will use
64K or 128K.

then space for payloads will be 128k or 64k....

For Normal image, if the total image is 512K, the Normal image is
about 256K too.
but some time we need to allocate 48K from it to ATIX.rom...
then the paload space will be small..

anyway, you can set ROM_IMAGE_SIZE to any size
1. >64K
2. ROM_IMAGE_SIZE-64K > linuxbios_ram size
3. leave enough size for payload...

YH

On 12/2/05, Stefan Reinauer  wrote:
> yhlu  [051202 22:37]:
> > the ROM_IMAGE_SIZE should be 0x20000 in target MB Config.
>
> Can you please explain this again? Does this mean that linuxbios+payload
> is at least 128KB?
>
> Or is this LinuxBIOS without payload?
>
> No LinuxBIOS Normal+Fallback on machines with 128kb only anymore?
>   Stefan

[Stefan Reinauer]

* yhlu  [051202 23:07]:
> In MB Option.lb, We define FALLBACK_SIZE=0x40000 ===> it means 256K
> for Fallback image.
>
> In the Fallback last 64K will be linuxbios_rom(crt0.S and init.o),

only crt0.S and init.o? Are they really that big? What about packing a
really silent LinuxBIOS image without payload in 32k again? I remember
some mainboards allowed that.

Say I want a system with a flash utility in fallback (so FALLBACK_SIZE
would be 64k) and the real payload in normal, so I'd want 448k for that.
Possible?
> For Normal image, if the total image is 512K, the Normal image is
> about 256K too.
> but some time we need to allocate 48K from it to ATIX.rom...
> then the paload space will be small..

I'd rather want a smaller Fallback than a smaller normal. Fallback is
thought of as a recovery thing only.
> anyway, you can set ROM_IMAGE_SIZE to any size
> > 64K
> ROM_IMAGE_SIZE-64K > linuxbios_ram size
> leave enough size for payload...

thanks for the explanation.

[yhlu]

please check the arch/i386/init/ldscript.lb

the line
.=_ROMBASE + ROM_IMAGE_SIZE - 0x10000;

will force linuxbios_rom at last 64K

the better solution will be
for fallback image

if .eram < (_ROMBASE + ROM_IMAGE_SIZE - 0x10000)
. (_ROMBASE + ROM_IMAGE_SIZE - 0x10000)
endif

then you can try to reduce ROM_IMAGE_SIZE to squash linuxbios_rom to the rom.

or the best solution to get the linuxbios_rom size and reverse the start point..

YH