All,
I finally got some cycles to look at why I was getting a "trap divied error" when trying to run coreboot v2 + LAB on the SimNOW Cheetah BSD. It looks like linuxrc in busybox is the culprit. It gets to the point in kernel init where it tries to run the linuxrc "program" and that is where it crashes, everything is Ok up to that point.
Has anyone else seen a similar issue?
Marc
-----Original Message----- From: coreboot-bounces+mylesgw=gmail.com@coreboot.org [mailto:coreboot- bounces+mylesgw=gmail.com@coreboot.org] On Behalf Of Marc Karasek Sent: Tuesday, May 06, 2008 2:26 PM To: coreboot@coreboot.org Subject: [coreboot] SimNOW V2 LAB problem
All,
I finally got some cycles to look at why I was getting a "trap divied error" when trying to run coreboot v2 + LAB on the SimNOW Cheetah BSD. It looks like linuxrc in busybox is the culprit. It gets to the point in kernel init where it tries to run the linuxrc "program" and that is where it crashes, everything is Ok up to that point.
Has anyone else seen a similar issue?
I have the same issue when I build on my FC 6 and later systems, but the same code works fine on FC 5.
Thanks, Myles
Crap, another compiler/tools problem. I hate these.. what version of gcc/binutils did you use under FC5?
********************* Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415 *********************
Myles Watson wrote:
-----Original Message----- From: coreboot-bounces+mylesgw=gmail.com@coreboot.org [mailto:coreboot- bounces+mylesgw=gmail.com@coreboot.org] On Behalf Of Marc Karasek Sent: Tuesday, May 06, 2008 2:26 PM To: coreboot@coreboot.org Subject: [coreboot] SimNOW V2 LAB problem
All,
I finally got some cycles to look at why I was getting a "trap divied error" when trying to run coreboot v2 + LAB on the SimNOW Cheetah BSD. It looks like linuxrc in busybox is the culprit. It gets to the point in kernel init where it tries to run the linuxrc "program" and that is where it crashes, everything is Ok up to that point.
Has anyone else seen a similar issue?
I have the same issue when I build on my FC 6 and later systems, but the same code works fine on FC 5.
Thanks, Myles
-----Original Message----- From: Marc.Karasek@Sun.COM [mailto:Marc.Karasek@Sun.COM] Sent: Tuesday, May 06, 2008 3:22 PM To: Myles Watson Cc: coreboot@coreboot.org Subject: Re: [coreboot] SimNOW V2 LAB problem
Crap, another compiler/tools problem. I hate these.. what version of gcc/binutils did you use under FC5?
gcc 4.1.1 for x86_64 release 51.fc5 binutils 2.16.91.0.6 for x86_64 release 5
I wouldn't have upgraded, but I couldn't build a working version of qemu with those tools. Now I have two broken systems :(
Thanks, Myles
I am going to try to compile with a newer version of busybox. Since the latest is 1.10.1 & buildrom uses 1.1.3. That is quite a lot of revisions...
********************* Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415 *********************
Myles Watson wrote:
-----Original Message----- From: Marc.Karasek@Sun.COM [mailto:Marc.Karasek@Sun.COM] Sent: Tuesday, May 06, 2008 3:22 PM To: Myles Watson Cc: coreboot@coreboot.org Subject: Re: [coreboot] SimNOW V2 LAB problem
Crap, another compiler/tools problem. I hate these.. what version of gcc/binutils did you use under FC5?
gcc 4.1.1 for x86_64 release 51.fc5 binutils 2.16.91.0.6 for x86_64 release 5
I wouldn't have upgraded, but I couldn't build a working version of qemu with those tools. Now I have two broken systems :(
Thanks, Myles
Myles,
When you build the broken one: 1) Is it a real exec file? Or is the file like 496 bytes? What does file busybox report? 2) What version of uclibc was the FC5 system using? 3) can you do a make in the busybox directory outside of buildrom in work/busybox/busybox-a.b.c/ and look at the output from this, i.e. is the busybox file have a correct size and is dynamic linked, etc..
Marc
********************* Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415 *********************
Myles Watson wrote:
-----Original Message----- From: Marc.Karasek@Sun.COM [mailto:Marc.Karasek@Sun.COM] Sent: Tuesday, May 06, 2008 3:22 PM To: Myles Watson Cc: coreboot@coreboot.org Subject: Re: [coreboot] SimNOW V2 LAB problem
Crap, another compiler/tools problem. I hate these.. what version of gcc/binutils did you use under FC5?
gcc 4.1.1 for x86_64 release 51.fc5 binutils 2.16.91.0.6 for x86_64 release 5
I wouldn't have upgraded, but I couldn't build a working version of qemu with those tools. Now I have two broken systems :(
Thanks, Myles
On Thu, May 8, 2008 at 12:37 PM, Marc Karasek Marc.Karasek@sun.com wrote:
Myles,
When you build the broken one:
- Is it a real exec file? Or is the file like 496 bytes? What does file
busybox report?
initrd-rootfs/bin/busybox is about 200K
Did you mean a different file? I think the error would come sooner if busybox were linked incorrectly.
- What version of uclibc was the FC5 system using?
The same as the FC6. 9.29. I bumped it up for SimNow because 9.28 didn't have a file necessary for x86_64.
- can you do a make in the busybox directory outside of buildrom in
work/busybox/busybox-a.b.c/ and look at the output from this, i.e. is the busybox file have a correct size and is dynamic linked, etc..
I don't think I have that problem. Maybe you're talking about a different file?
Myles
Marc
Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415
Myles Watson wrote:
-----Original Message----- From: Marc.Karasek@Sun.COM [mailto:Marc.Karasek@Sun.COM] Sent: Tuesday, May 06, 2008 3:22 PM To: Myles Watson Cc: coreboot@coreboot.org Subject: Re: [coreboot] SimNOW V2 LAB problem
Crap, another compiler/tools problem. I hate these.. what version of gcc/binutils did you use under FC5?
gcc 4.1.1 for x86_64 release 51.fc5 binutils 2.16.91.0.6 for x86_64 release 5
I wouldn't have upgraded, but I couldn't build a working version of qemu with those tools. Now I have two broken systems :(
Thanks, Myles
Myles,
Disregard my feeble wanderings.. I was thinking about my attempt to move busybox to 1.10.1 and the problems I was seeing there.
I think though the two are related. Seems that uClibc 0.9.29 no longer supports the toolchain wrappers for -nostdlib and forcing a linking with native tools (gcc/binutils) and a compiled uClibc library. This was pulled in 0.9.22 according to the uclibc FAQ. Maybe it was just a one off that FC5 still worked. What I see when I try 1.10.1 is it does not link properly and I get a 496 byte busybox file that says it is statically linked but is actually not even a valid exec file. I am working on trying to get buildroot to build me a valid environment that I can then use to build busybox correctly under buildrom. Either that or I pull the uclibc stuff from buildrom and see what glibc does in terms of size to the binary.
Marc
********************* Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415 *********************
Myles Watson wrote:
On Thu, May 8, 2008 at 12:37 PM, Marc Karasek Marc.Karasek@sun.com wrote:
Myles,
When you build the broken one:
- Is it a real exec file? Or is the file like 496 bytes? What does file
busybox report?
initrd-rootfs/bin/busybox is about 200K
Did you mean a different file? I think the error would come sooner if busybox were linked incorrectly.
- What version of uclibc was the FC5 system using?
The same as the FC6. 9.29. I bumped it up for SimNow because 9.28 didn't have a file necessary for x86_64.
- can you do a make in the busybox directory outside of buildrom in
work/busybox/busybox-a.b.c/ and look at the output from this, i.e. is the busybox file have a correct size and is dynamic linked, etc..
I don't think I have that problem. Maybe you're talking about a different file?
Myles
Marc
Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415
Myles Watson wrote:
-----Original Message----- From: Marc.Karasek@Sun.COM [mailto:Marc.Karasek@Sun.COM] Sent: Tuesday, May 06, 2008 3:22 PM To: Myles Watson Cc: coreboot@coreboot.org Subject: Re: [coreboot] SimNOW V2 LAB problem
Crap, another compiler/tools problem. I hate these.. what version of gcc/binutils did you use under FC5?
gcc 4.1.1 for x86_64 release 51.fc5 binutils 2.16.91.0.6 for x86_64 release 5
I wouldn't have upgraded, but I couldn't build a working version of qemu with those tools. Now I have two broken systems :(
Thanks, Myles
Disregard my feeble wanderings.. I was thinking about my attempt to move busybox to 1.10.1 and the problems I was seeing there.
I think though the two are related. Seems that uClibc 0.9.29 no longer supports the toolchain wrappers for -nostdlib and forcing a linking with native tools (gcc/binutils) and a compiled uClibc library. This was pulled in 0.9.22 according to the uclibc FAQ. Maybe it was just a one off that FC5 still worked.
Strange. It worked for 0.9.28 as well. Those are the two versions I've used in buildrom.
Ward: You've used lab haven't you? What toolchain do you use?
Thanks, Myles
On Thu, May 08, 2008 at 03:52:21PM -0600, Myles Watson wrote:
Disregard my feeble wanderings.. I was thinking about my attempt to move busybox to 1.10.1 and the problems I was seeing there.
I think though the two are related. Seems that uClibc 0.9.29 no longer supports the toolchain wrappers for -nostdlib and forcing a linking with native tools (gcc/binutils) and a compiled uClibc library. This was pulled in 0.9.22 according to the uclibc FAQ. Maybe it was just a one off that FC5 still worked.
Strange. It worked for 0.9.28 as well. Those are the two versions I've used in buildrom.
Ward: You've used lab haven't you? What toolchain do you use?
Yep, I've got LAB systems running. I build on gNewsense deltad typically, which is basically Ubuntu Dapper, 32 bit.
Thanks, Ward.
That is one thing.. I am building on a 64 bit platform FC8_64. Maybe the 64 bit stuff is the problem. I can try to build it on a 32 bit platform and see what happens.
I should be able to run the rom file on SimNOW in 32bit mode, right???
/********************* Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415 *********************/
Ward Vandewege wrote:
On Thu, May 08, 2008 at 03:52:21PM -0600, Myles Watson wrote:
Disregard my feeble wanderings.. I was thinking about my attempt to move busybox to 1.10.1 and the problems I was seeing there.
I think though the two are related. Seems that uClibc 0.9.29 no longer supports the toolchain wrappers for -nostdlib and forcing a linking with native tools (gcc/binutils) and a compiled uClibc library. This was pulled in 0.9.22 according to the uclibc FAQ. Maybe it was just a one off that FC5 still worked.
Strange. It worked for 0.9.28 as well. Those are the two versions I've used in buildrom.
Ward: You've used lab haven't you? What toolchain do you use?
Yep, I've got LAB systems running. I build on gNewsense deltad typically, which is basically Ubuntu Dapper, 32 bit.
Thanks, Ward.
-----Original Message----- From: Marc.Karasek@Sun.COM [mailto:Marc.Karasek@Sun.COM] Sent: Thursday, May 08, 2008 10:50 PM To: Myles Watson; Marc.Karasek@Sun.COM; coreboot@coreboot.org Subject: Re: [coreboot] SimNOW V2 LAB problem
That is one thing.. I am building on a 64 bit platform FC8_64. Maybe the 64 bit stuff is the problem. I can try to build it on a 32 bit platform and see what happens.
I should be able to run the rom file on SimNOW in 32bit mode, right???
Yes. Either way it starts out in 32-bit mode.
Thanks, Myles
/********************* Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415 *********************/
Ward Vandewege wrote:
On Thu, May 08, 2008 at 03:52:21PM -0600, Myles Watson wrote:
Disregard my feeble wanderings.. I was thinking about my attempt to move busybox to 1.10.1 and the problems I was seeing there.
I think though the two are related. Seems that uClibc 0.9.29 no
longer
supports the toolchain wrappers for -nostdlib and forcing a linking
with
native tools (gcc/binutils) and a compiled uClibc library. This was pulled in 0.9.22 according to the uclibc FAQ. Maybe it was just a one off that FC5 still worked.
Strange. It worked for 0.9.28 as well. Those are the two versions
I've
used in buildrom.
Ward: You've used lab haven't you? What toolchain do you use?
Yep, I've got LAB systems running. I build on gNewsense deltad
typically,
which is basically Ubuntu Dapper, 32 bit.
Thanks, Ward.
On 08/05/08 20:32 -0400, Ward Vandewege wrote:
On Thu, May 08, 2008 at 03:52:21PM -0600, Myles Watson wrote:
Disregard my feeble wanderings.. I was thinking about my attempt to move busybox to 1.10.1 and the problems I was seeing there.
I think though the two are related. Seems that uClibc 0.9.29 no longer supports the toolchain wrappers for -nostdlib and forcing a linking with native tools (gcc/binutils) and a compiled uClibc library. This was pulled in 0.9.22 according to the uclibc FAQ. Maybe it was just a one off that FC5 still worked.
Strange. It worked for 0.9.28 as well. Those are the two versions I've used in buildrom.
Ward: You've used lab haven't you? What toolchain do you use?
Yep, I've got LAB systems running. I build on gNewsense deltad typically, which is basically Ubuntu Dapper, 32 bit.
It might be time for us to ping the uClibc / busybox teams for some guidance.
Jordan
-----Original Message----- From: Jordan Crouse [mailto:jordan.crouse@amd.com] Sent: Friday, May 09, 2008 9:16 AM To: Myles Watson; Marc.Karasek@Sun.COM; coreboot@coreboot.org Subject: Re: SimNOW V2 LAB problem
On 08/05/08 20:32 -0400, Ward Vandewege wrote:
On Thu, May 08, 2008 at 03:52:21PM -0600, Myles Watson wrote:
Disregard my feeble wanderings.. I was thinking about my attempt to move busybox to 1.10.1 and the problems I was seeing there.
I think though the two are related. Seems that uClibc 0.9.29 no
longer
supports the toolchain wrappers for -nostdlib and forcing a linking
with
native tools (gcc/binutils) and a compiled uClibc library. This was pulled in 0.9.22 according to the uclibc FAQ. Maybe it was just a
one
off that FC5 still worked.
Strange. It worked for 0.9.28 as well. Those are the two versions
I've
used in buildrom.
Ward: You've used lab haven't you? What toolchain do you use?
Yep, I've got LAB systems running. I build on gNewsense deltad
typically,
which is basically Ubuntu Dapper, 32 bit.
It might be time for us to ping the uClibc / busybox teams for some guidance.
So you think it's a broken uClibc / busybox problem, not a compiler/binutils problem?
I'm inclined to think that it's a toolchain issue since I can't build v2 on my 32-bit system. I don't know how to tell the difference, though.
Having the serial port continue working through the kernel on SimNow would help, though.
Thanks, Myles
On 09/05/08 09:40 -0600, Myles Watson wrote:
-----Original Message----- From: Jordan Crouse [mailto:jordan.crouse@amd.com] Sent: Friday, May 09, 2008 9:16 AM To: Myles Watson; Marc.Karasek@Sun.COM; coreboot@coreboot.org Subject: Re: SimNOW V2 LAB problem
On 08/05/08 20:32 -0400, Ward Vandewege wrote:
On Thu, May 08, 2008 at 03:52:21PM -0600, Myles Watson wrote:
Disregard my feeble wanderings.. I was thinking about my attempt to move busybox to 1.10.1 and the problems I was seeing there.
I think though the two are related. Seems that uClibc 0.9.29 no
longer
supports the toolchain wrappers for -nostdlib and forcing a linking
with
native tools (gcc/binutils) and a compiled uClibc library. This was pulled in 0.9.22 according to the uclibc FAQ. Maybe it was just a
one
off that FC5 still worked.
Strange. It worked for 0.9.28 as well. Those are the two versions
I've
used in buildrom.
Ward: You've used lab haven't you? What toolchain do you use?
Yep, I've got LAB systems running. I build on gNewsense deltad
typically,
which is basically Ubuntu Dapper, 32 bit.
It might be time for us to ping the uClibc / busybox teams for some guidance.
So you think it's a broken uClibc / busybox problem, not a compiler/binutils problem?
I'm inclined to think that it's a toolchain issue since I can't build v2 on my 32-bit system. I don't know how to tell the difference, though.
Having the serial port continue working through the kernel on SimNow would help, though.
D'oh - its not working through the kernel? Even with console=ttyS0,115200 ?
That sounds ominous.
Jordan
Nope, from what I recall the serial console stops when he kernel loads. I will have to verify this, but...
********************* Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415 *********************
Jordan Crouse wrote:
On 09/05/08 09:40 -0600, Myles Watson wrote:
-----Original Message----- From: Jordan Crouse [mailto:jordan.crouse@amd.com] Sent: Friday, May 09, 2008 9:16 AM To: Myles Watson; Marc.Karasek@Sun.COM; coreboot@coreboot.org Subject: Re: SimNOW V2 LAB problem
On 08/05/08 20:32 -0400, Ward Vandewege wrote:
On Thu, May 08, 2008 at 03:52:21PM -0600, Myles Watson wrote:
Disregard my feeble wanderings.. I was thinking about my attempt to move busybox to 1.10.1 and the problems I was seeing there.
I think though the two are related. Seems that uClibc 0.9.29 no
longer
supports the toolchain wrappers for -nostdlib and forcing a linking
with
native tools (gcc/binutils) and a compiled uClibc library. This was pulled in 0.9.22 according to the uclibc FAQ. Maybe it was just a
one
off that FC5 still worked.
Strange. It worked for 0.9.28 as well. Those are the two versions
I've
used in buildrom.
Ward: You've used lab haven't you? What toolchain do you use?
Yep, I've got LAB systems running. I build on gNewsense deltad
typically,
which is basically Ubuntu Dapper, 32 bit.
It might be time for us to ping the uClibc / busybox teams for some guidance.
So you think it's a broken uClibc / busybox problem, not a compiler/binutils problem?
I'm inclined to think that it's a toolchain issue since I can't build v2 on my 32-bit system. I don't know how to tell the difference, though.
Having the serial port continue working through the kernel on SimNow would help, though.
D'oh - its not working through the kernel? Even with console=ttyS0,115200 ?
That sounds ominous.
Jordan
-----Original Message----- From: Jordan Crouse [mailto:jordan.crouse@amd.com] Sent: Friday, May 09, 2008 9:46 AM To: Myles Watson Cc: Marc.Karasek@Sun.COM; coreboot@coreboot.org Subject: Re: SimNOW V2 LAB problem
On 09/05/08 09:40 -0600, Myles Watson wrote:
-----Original Message----- From: Jordan Crouse [mailto:jordan.crouse@amd.com] Sent: Friday, May 09, 2008 9:16 AM To: Myles Watson; Marc.Karasek@Sun.COM; coreboot@coreboot.org Subject: Re: SimNOW V2 LAB problem
On 08/05/08 20:32 -0400, Ward Vandewege wrote:
On Thu, May 08, 2008 at 03:52:21PM -0600, Myles Watson wrote:
Disregard my feeble wanderings.. I was thinking about my
attempt to
move busybox to 1.10.1 and the problems I was seeing there.
I think though the two are related. Seems that uClibc 0.9.29 no
longer
supports the toolchain wrappers for -nostdlib and forcing a
linking
with
native tools (gcc/binutils) and a compiled uClibc library. This
was
pulled in 0.9.22 according to the uclibc FAQ. Maybe it was just
a
one
off that FC5 still worked.
Strange. It worked for 0.9.28 as well. Those are the two
versions
I've
used in buildrom.
Ward: You've used lab haven't you? What toolchain do you use?
Yep, I've got LAB systems running. I build on gNewsense deltad
typically,
which is basically Ubuntu Dapper, 32 bit.
It might be time for us to ping the uClibc / busybox teams for some guidance.
So you think it's a broken uClibc / busybox problem, not a
compiler/binutils
problem?
I'm inclined to think that it's a toolchain issue since I can't build v2
on
my 32-bit system. I don't know how to tell the difference, though.
Having the serial port continue working through the kernel on SimNow
would
help, though.
D'oh - its not working through the kernel? Even with console=ttyS0,115200 ?
That's right.
I tried to get help from SimNow support, but they didn't seem to be using the same version of SimNow as I was. They couldn't boot the ROMs I sent them, even though they booted fine for me. I eventually gave up.
Myles
On 09/05/08 09:55 -0600, Myles Watson wrote:
-----Original Message----- From: Jordan Crouse [mailto:jordan.crouse@amd.com] Sent: Friday, May 09, 2008 9:46 AM To: Myles Watson Cc: Marc.Karasek@Sun.COM; coreboot@coreboot.org Subject: Re: SimNOW V2 LAB problem
On 09/05/08 09:40 -0600, Myles Watson wrote:
-----Original Message----- From: Jordan Crouse [mailto:jordan.crouse@amd.com] Sent: Friday, May 09, 2008 9:16 AM To: Myles Watson; Marc.Karasek@Sun.COM; coreboot@coreboot.org Subject: Re: SimNOW V2 LAB problem
On 08/05/08 20:32 -0400, Ward Vandewege wrote:
On Thu, May 08, 2008 at 03:52:21PM -0600, Myles Watson wrote:
> Disregard my feeble wanderings.. I was thinking about my
attempt to
> move busybox to 1.10.1 and the problems I was seeing there. > > I think though the two are related. Seems that uClibc 0.9.29 no
longer
> supports the toolchain wrappers for -nostdlib and forcing a
linking
with
> native tools (gcc/binutils) and a compiled uClibc library. This
was
> pulled in 0.9.22 according to the uclibc FAQ. Maybe it was just
a
one
> off that FC5 still worked.
Strange. It worked for 0.9.28 as well. Those are the two
versions
I've
used in buildrom.
Ward: You've used lab haven't you? What toolchain do you use?
Yep, I've got LAB systems running. I build on gNewsense deltad
typically,
which is basically Ubuntu Dapper, 32 bit.
It might be time for us to ping the uClibc / busybox teams for some guidance.
So you think it's a broken uClibc / busybox problem, not a
compiler/binutils
problem?
I'm inclined to think that it's a toolchain issue since I can't build v2
on
my 32-bit system. I don't know how to tell the difference, though.
Having the serial port continue working through the kernel on SimNow
would
help, though.
D'oh - its not working through the kernel? Even with console=ttyS0,115200 ?
That's right.
I tried to get help from SimNow support, but they didn't seem to be using the same version of SimNow as I was. They couldn't boot the ROMs I sent them, even though they booted fine for me. I eventually gave up.
Okay, I'll give it a look.
Jordan
Jordan,
I d/l buildrom to my notebook to build a 32bit version to try it with SimNOW. It compiles but gets a linker error : <coreboot: section `.ram' can't be allocated in segment 0> Any thoughts on what could be causing this???
Marc
********************* Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415 *********************
Jordan Crouse wrote:
On 08/05/08 20:32 -0400, Ward Vandewege wrote:
On Thu, May 08, 2008 at 03:52:21PM -0600, Myles Watson wrote:
Disregard my feeble wanderings.. I was thinking about my attempt to move busybox to 1.10.1 and the problems I was seeing there.
I think though the two are related. Seems that uClibc 0.9.29 no longer supports the toolchain wrappers for -nostdlib and forcing a linking with native tools (gcc/binutils) and a compiled uClibc library. This was pulled in 0.9.22 according to the uclibc FAQ. Maybe it was just a one off that FC5 still worked.
Strange. It worked for 0.9.28 as well. Those are the two versions I've used in buildrom.
Ward: You've used lab haven't you? What toolchain do you use?
Yep, I've got LAB systems running. I build on gNewsense deltad typically, which is basically Ubuntu Dapper, 32 bit.
It might be time for us to ping the uClibc / busybox teams for some guidance.
Jordan
On 09/05/08 11:41 -0400, Marc Karasek wrote:
Jordan,
I d/l buildrom to my notebook to build a 32bit version to try it with SimNOW. It compiles but gets a linker error : <coreboot: section `.ram' can't be allocated in segment 0> Any thoughts on what could be causing this???
This is from coreboot itself? That sounds really evil. We clearly have some major toolchain churn here.
Marc
Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415
Jordan Crouse wrote:
On 08/05/08 20:32 -0400, Ward Vandewege wrote:
On Thu, May 08, 2008 at 03:52:21PM -0600, Myles Watson wrote:
Disregard my feeble wanderings.. I was thinking about my attempt to move busybox to 1.10.1 and the problems I was seeing there.
I think though the two are related. Seems that uClibc 0.9.29 no longer supports the toolchain wrappers for -nostdlib and forcing a linking with native tools (gcc/binutils) and a compiled uClibc library. This was pulled in 0.9.22 according to the uclibc FAQ. Maybe it was just a one off that FC5 still worked.
Strange. It worked for 0.9.28 as well. Those are the two versions I've used in buildrom.
Ward: You've used lab haven't you? What toolchain do you use?
Yep, I've got LAB systems running. I build on gNewsense deltad typically, which is basically Ubuntu Dapper, 32 bit.
It might be time for us to ping the uClibc / busybox teams for some guidance.
Jordan
Good News I just tried coreboot-v3 with the same system and it does compile...
********************* Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415 *********************
Jordan Crouse wrote:
On 09/05/08 11:41 -0400, Marc Karasek wrote:
Jordan,
I d/l buildrom to my notebook to build a 32bit version to try it with SimNOW. It compiles but gets a linker error : <coreboot: section `.ram' can't be allocated in segment 0> Any thoughts on what could be causing this???
This is from coreboot itself? That sounds really evil. We clearly have some major toolchain churn here.
Marc
Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415
Jordan Crouse wrote:
On 08/05/08 20:32 -0400, Ward Vandewege wrote:
On Thu, May 08, 2008 at 03:52:21PM -0600, Myles Watson wrote:
Disregard my feeble wanderings.. I was thinking about my attempt to move busybox to 1.10.1 and the problems I was seeing there.
I think though the two are related. Seems that uClibc 0.9.29 no longer supports the toolchain wrappers for -nostdlib and forcing a linking with native tools (gcc/binutils) and a compiled uClibc library. This was pulled in 0.9.22 according to the uclibc FAQ. Maybe it was just a one off that FC5 still worked.
Strange. It worked for 0.9.28 as well. Those are the two versions I've used in buildrom.
Ward: You've used lab haven't you? What toolchain do you use?
Yep, I've got LAB systems running. I build on gNewsense deltad typically, which is basically Ubuntu Dapper, 32 bit.
It might be time for us to ping the uClibc / busybox teams for some guidance.
Jordan
-----Original Message----- From: Marc.Karasek@Sun.COM [mailto:Marc.Karasek@Sun.COM] Sent: Friday, May 09, 2008 10:34 AM To: Jordan Crouse Cc: Myles Watson; coreboot@coreboot.org Subject: Re: SimNOW V2 LAB problem
Good News I just tried coreboot-v3 with the same system and it does compile...
Yes. It's weird. I can even compile v2 with an identical toolchain (release versions) on my x86_64 box.
Here's the original thread.
http://coreboot.org/pipermail/coreboot/2008-February/029993.html
Thanks, Myles
Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415
Jordan Crouse wrote:
On 09/05/08 11:41 -0400, Marc Karasek wrote:
Jordan,
I d/l buildrom to my notebook to build a 32bit version to try it with SimNOW. It compiles but gets a linker error : <coreboot: section `.ram' can't be allocated in segment 0> Any thoughts on what could be causing this???
This is from coreboot itself? That sounds really evil. We clearly have some major toolchain churn here.
Marc
Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415
Jordan Crouse wrote:
On 08/05/08 20:32 -0400, Ward Vandewege wrote:
On Thu, May 08, 2008 at 03:52:21PM -0600, Myles Watson wrote:
> Disregard my feeble wanderings.. I was thinking about my attempt
to
> move busybox to 1.10.1 and the problems I was seeing there. > > I think though the two are related. Seems that uClibc 0.9.29 no
longer
> supports the toolchain wrappers for -nostdlib and forcing a linking > with > native tools (gcc/binutils) and a compiled uClibc library. This
was
> pulled in 0.9.22 according to the uclibc FAQ. Maybe it was just a
one
> off that FC5 still worked. > Strange. It worked for 0.9.28 as well. Those are the two versions
I've
used in buildrom.
Ward: You've used lab haven't you? What toolchain do you use?
Yep, I've got LAB systems running. I build on gNewsense deltad
typically,
which is basically Ubuntu Dapper, 32 bit.
It might be time for us to ping the uClibc / busybox teams for some guidance.
Jordan
I do know about the .id failure.
This has to do with where the .id section is being placed in the rom image. For some reason, I never found out why, you need to put it at 0xFFFFE000 or lower. Any address from there to the top fails with the .id message.
I really do not like toolchain problems..
We should standardize on a set toolchain. Maybe using the buildroot tool to build a toolchain we can then use to build coreboot. Having something that is a "standard" would help with these issues...
Marc
********************* Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415 *********************
Myles Watson wrote:
-----Original Message----- From: Marc.Karasek@Sun.COM [mailto:Marc.Karasek@Sun.COM] Sent: Friday, May 09, 2008 10:34 AM To: Jordan Crouse Cc: Myles Watson; coreboot@coreboot.org Subject: Re: SimNOW V2 LAB problem
Good News I just tried coreboot-v3 with the same system and it does compile...
Yes. It's weird. I can even compile v2 with an identical toolchain (release versions) on my x86_64 box.
Here's the original thread.
http://coreboot.org/pipermail/coreboot/2008-February/029993.html
Thanks, Myles
Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415
Jordan Crouse wrote:
On 09/05/08 11:41 -0400, Marc Karasek wrote:
Jordan,
I d/l buildrom to my notebook to build a 32bit version to try it with SimNOW. It compiles but gets a linker error : <coreboot: section `.ram' can't be allocated in segment 0> Any thoughts on what could be causing this???
This is from coreboot itself? That sounds really evil. We clearly have some major toolchain churn here.
Marc
Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415
Jordan Crouse wrote:
On 08/05/08 20:32 -0400, Ward Vandewege wrote:
On Thu, May 08, 2008 at 03:52:21PM -0600, Myles Watson wrote:
>> Disregard my feeble wanderings.. I was thinking about my attempt >>
to
>> move busybox to 1.10.1 and the problems I was seeing there. >> >> I think though the two are related. Seems that uClibc 0.9.29 no >>
longer
>> supports the toolchain wrappers for -nostdlib and forcing a linking >> with >> native tools (gcc/binutils) and a compiled uClibc library. This >>
was
>> pulled in 0.9.22 according to the uclibc FAQ. Maybe it was just a >>
one
>> off that FC5 still worked. >> >> > Strange. It worked for 0.9.28 as well. Those are the two versions >
I've
> used in buildrom. > > Ward: You've used lab haven't you? What toolchain do you use? > > Yep, I've got LAB systems running. I build on gNewsense deltad
typically,
which is basically Ubuntu Dapper, 32 bit.
It might be time for us to ping the uClibc / busybox teams for some guidance.
Jordan
On Fri, May 9, 2008 at 9:58 AM, Marc Karasek Marc.Karasek@sun.com wrote:
We should standardize on a set toolchain. Maybe using the buildroot tool to build a toolchain we can then use to build coreboot. Having something that is a "standard" would help with these issues...
hmm. This has been mentioned before :-)
I think I agree with you. gcc and friends are just not stable enough to be relied on in my experience.
ron
Even better news version V2 - 3092 does not have the .ram in section 0 problem. It has the .id problem I faced before, but i know how to fix that by moving the .id section to a lower fixed value in the rom.
V2 now compiles, the image is too big. 491143 bytes to large. I will increase the size of the rom to 2MB and that should do it. ?? Why is this image so much bigger than the one I built before?
********************* Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415 *********************
Marc Karasek wrote:
Good News I just tried coreboot-v3 with the same system and it does compile...
Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415
Jordan Crouse wrote:
On 09/05/08 11:41 -0400, Marc Karasek wrote:
Jordan,
I d/l buildrom to my notebook to build a 32bit version to try it with SimNOW. It compiles but gets a linker error : <coreboot: section `.ram' can't be allocated in segment 0> Any thoughts on what could be causing this???
This is from coreboot itself? That sounds really evil. We clearly have some major toolchain churn here.
Marc
Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415
Jordan Crouse wrote:
On 08/05/08 20:32 -0400, Ward Vandewege wrote:
On Thu, May 08, 2008 at 03:52:21PM -0600, Myles Watson wrote:
> Disregard my feeble wanderings.. I was thinking about my attempt to > move busybox to 1.10.1 and the problems I was seeing there. > > I think though the two are related. Seems that uClibc 0.9.29 no longer > supports the toolchain wrappers for -nostdlib and forcing a linking > with > native tools (gcc/binutils) and a compiled uClibc library. This was > pulled in 0.9.22 according to the uclibc FAQ. Maybe it was just a one > off that FC5 still worked. > > Strange. It worked for 0.9.28 as well. Those are the two versions I've used in buildrom.
Ward: You've used lab haven't you? What toolchain do you use?
Yep, I've got LAB systems running. I build on gNewsense deltad typically, which is basically Ubuntu Dapper, 32 bit.
It might be time for us to ping the uClibc / busybox teams for some guidance.
Jordan
To try to get a better handle on this:
I am going to put together a matrix of what compiles what does not on the two machines I have here. (32 & 64 bit). Along with any errors. As soon as I have this i am going to send it to the list.
Marc
********************* Marc Karasek MTS Sun Microsystems mailto:marc.karasek@sun.com ph:770.360.6415 *********************
Ward Vandewege wrote:
On Thu, May 08, 2008 at 03:52:21PM -0600, Myles Watson wrote:
Disregard my feeble wanderings.. I was thinking about my attempt to move busybox to 1.10.1 and the problems I was seeing there.
I think though the two are related. Seems that uClibc 0.9.29 no longer supports the toolchain wrappers for -nostdlib and forcing a linking with native tools (gcc/binutils) and a compiled uClibc library. This was pulled in 0.9.22 according to the uclibc FAQ. Maybe it was just a one off that FC5 still worked.
Strange. It worked for 0.9.28 as well. Those are the two versions I've used in buildrom.
Ward: You've used lab haven't you? What toolchain do you use?
Yep, I've got LAB systems running. I build on gNewsense deltad typically, which is basically Ubuntu Dapper, 32 bit.
Thanks, Ward.
On 06/05/08 15:19 -0600, Myles Watson wrote:
-----Original Message----- From: coreboot-bounces+mylesgw=gmail.com@coreboot.org [mailto:coreboot- bounces+mylesgw=gmail.com@coreboot.org] On Behalf Of Marc Karasek Sent: Tuesday, May 06, 2008 2:26 PM To: coreboot@coreboot.org Subject: [coreboot] SimNOW V2 LAB problem
All,
I finally got some cycles to look at why I was getting a "trap divied error" when trying to run coreboot v2 + LAB on the SimNOW Cheetah BSD. It looks like linuxrc in busybox is the culprit. It gets to the point in kernel init where it tries to run the linuxrc "program" and that is where it crashes, everything is Ok up to that point.
Has anyone else seen a similar issue?
I have the same issue when I build on my FC 6 and later systems, but the same code works fine on FC 5.
Sounds like possible uclibc breakage? </shot in the dark>
Jordan