Skip to contentSkip to conversation

linux-next: ia64 allmodconfig

DISCUSSION Andrew Morton 5 participants 15 messages Aug 21, 2008 – Aug 22, 2008

lore
Share & tools
lore
#0 lore raw

arch/ia64/kernel/built-in.o(.text+0x4862): In function `efi_initialize_iomem_resources':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x12a92): In function pfm_context_create':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x1bc02): In function sys_perfmonctl':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x37c82): In function dup_state_stack':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x382b2): In function desc_prologue':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x3ad12): more undefined references to __ia64_save_stack_nonlocal' follow

ia64 seems to break a lot.

#1 lore raw
Quoted 12 lines · Andrew Morton

arch/ia64/kernel/built-in.o(.text+0x4862): In function `efi_initialize_iomem_resources':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x12a92): In function pfm_context_create':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x1bc02): In function sys_perfmonctl':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x37c82): In function dup_state_stack':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x382b2): In function desc_prologue':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x3ad12): more undefined references to __ia64_save_stack_nonlocal' follow

Quoted 1 line

ia64 seems to break a lot.

I just did a git-checkout of
git://git.kernel.org/pub/scm/linux/kernel/git/sfr/linux-next.git
linux-next/master and got commit:
e85df409b2ad8528fa772379393c81406c417bbf

When I
make allmodconfig; make -j 24 compressed, I get no such warnings. I do still see a failure from the SFC driver :(

Are you sure something isn't a little dirty in your work area? Did you maybe do a make allyesconfig, allnoconfig, allmodconfig sequence?

Thanks,
Robin

#2 lore raw
In reply to Andrew Morton
Quoted 20 lines · Robin Holt
Quoted 12 lines · Andrew Morton

arch/ia64/kernel/built-in.o(.text+0x4862): In function `efi_initialize_iomem_resources':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x12a92): In function pfm_context_create':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x1bc02): In function sys_perfmonctl':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x37c82): In function dup_state_stack':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x382b2): In function desc_prologue':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x3ad12): more undefined references to __ia64_save_stack_nonlocal' follow

Quoted 1 line

ia64 seems to break a lot.

I just did a git-checkout of
git://git.kernel.org/pub/scm/linux/kernel/git/sfr/linux-next.git
linux-next/master and got commit:
e85df409b2ad8528fa772379393c81406c417bbf

Quoted 6 lines

When I
make allmodconfig; make -j 24 compressed, I get no such warnings. I do still see a failure from the SFC driver :(

Are you sure something isn't a little dirty in your work area? Did you maybe do a make allyesconfig, allnoconfig, allmodconfig sequence?

Oh, just realized this could be a gcc version issue as well. Could you provide that as well? I had used the SLES installed version (4.1.2).

Thanks,
Robin

#3 lore raw
In reply to Andrew Morton
Quoted 21 lines · On Thu, 21 Aug 2008 08:04:23 -0500 Robin Holt <[email protected]> wrote:
Quoted 20 lines · Robin Holt
Quoted 12 lines · Andrew Morton

arch/ia64/kernel/built-in.o(.text+0x4862): In function `efi_initialize_iomem_resources':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x12a92): In function pfm_context_create':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x1bc02): In function sys_perfmonctl':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x37c82): In function dup_state_stack':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x382b2): In function desc_prologue':
: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x3ad12): more undefined references to __ia64_save_stack_nonlocal' follow

Quoted 1 line

ia64 seems to break a lot.

I just did a git-checkout of
git://git.kernel.org/pub/scm/linux/kernel/git/sfr/linux-next.git
linux-next/master and got commit:
e85df409b2ad8528fa772379393c81406c417bbf

Quoted 6 lines
Quoted 6 lines

When I
make allmodconfig; make -j 24 compressed, I get no such warnings. I do still see a failure from the SFC driver :(

Are you sure something isn't a little dirty in your work area? Did you maybe do a make allyesconfig, allnoconfig, allmodconfig sequence?

make mrproper
make allmodconfig
make -j12

Quoted 2 lines

Oh, just realized this could be a gcc version issue as well. Could you provide that as well? I had used the SLES installed version (4.1.2).

It's the one in http://userweb.kernel.org/~akpm/cross-compilers/

gcc-3.4.5-glibc-2.3.6
GNU ld version 2.15

#4 lore raw
In reply to Andrew Morton
Quoted 4 lines

I just did a git-checkout of
git://git.kernel.org/pub/scm/linux/kernel/git/sfr/linux-next.git
linux-next/master and got commit:
e85df409b2ad8528fa772379393c81406c417bbf

Hmmm that looks like the commit for next-20080821, but I see build failures for 0820 and 0821.

Using "git diff next-20080819 next-20080820" shows the only change in arch/ia64 is an innocuous looking one to arch/ia64/kvm/irq.h The only plausible looking stuff is the slab.h and slub_def.h changes because the first error is reported in efi_initialize_iomem_resources() which does make a kzalloc() call ... and commenting out the kzalloc() moves the first reported error out of this function to some other one.

I see the same error with the generic_defconfig ... so it isn't special to allmodconfig.

I'm using gcc version:
gcc version 3.4.6 20060404 (Red Hat 3.4.6-8)

So this might be a 3.x vs. 4.x issue.

Heck if I can see where these "__ia64_save_stack_nonlocal" symbols are coming from though. I thought that there is some cpp trickery concatenating tokens ... but a "cc -E" of efi.c didn't give me any clues :-(

-Tony

#5 lore raw
In reply to Andrew Morton
Quoted 21 lines · Tony
Quoted 4 lines

I just did a git-checkout of
git://git.kernel.org/pub/scm/linux/kernel/git/sfr/linux-next.git
linux-next/master and got commit:
e85df409b2ad8528fa772379393c81406c417bbf

Hmmm that looks like the commit for next-20080821, but I see build failures for 0820 and 0821.

Using "git diff next-20080819 next-20080820" shows the only change in arch/ia64 is an innocuous looking one to arch/ia64/kvm/irq.h The only plausible looking stuff is the slab.h and slub_def.h changes because the first error is reported in efi_initialize_iomem_resources() which does make a kzalloc() call ... and commenting out the kzalloc() moves the first reported error out of this function to some other one.

I see the same error with the generic_defconfig ... so it isn't special to allmodconfig.

I'm using gcc version:
gcc version 3.4.6 20060404 (Red Hat 3.4.6-8)

Quoted 6 lines

So this might be a 3.x vs. 4.x issue.

Heck if I can see where these "__ia64_save_stack_nonlocal" symbols are coming from though. I thought that there is some cpp trickery concatenating tokens ... but a "cc -E" of efi.c didn't give me any clues :-(

Not sure if this is helpful, but google pointed me at:
http://www.mail-archive.com/[email protected]/msg116574.html

Robin

#6 lore raw
In reply to Andrew Morton
Quoted 2 lines

Not sure if this is helpful, but google pointed me at:
http://www.mail-archive.com/[email protected]/msg116574.html

That said "Fixed in 4.0.0" ... so I grabbed and built a newer gcc (4.2.4) and lo and behold the next-20080821 tag builds and boots just fine[*] with it.

Still need to figure out exactly what broke the build with a 3.x gcc to see if there is an easy workaround. Though 4.0.0 was released over three years ago ... is it time to give up on the ancient 3.x technology?

-Tony

[*] If you ignore that the 4.2.4 compiler is about 15% slower than my trusty old 3.4.6 :-(

#7 lore raw
In reply to Andrew Morton
Quoted 12 lines · Tony
Quoted 2 lines

Not sure if this is helpful, but google pointed me at:
http://www.mail-archive.com/[email protected]/msg116574.html

That said "Fixed in 4.0.0" ... so I grabbed and built a newer gcc (4.2.4) and lo and behold the next-20080821 tag builds and boots just fine[*] with it.

Still need to figure out exactly what broke the build with a 3.x gcc to see if there is an easy workaround. Though 4.0.0 was released over three years ago ... is it time to give up on the ancient 3.x technology?

Quoted 4 lines

-Tony

[*] If you ignore that the 4.2.4 compiler is about 15% slower than my trusty old 3.4.6 :-(

I have been doing a bisect using akpm's compiler to see if I can find the individual commit. Unfortunately, my laptop isn't the fastest build machine in the world.

I did objdump the built-in.o and there is no such symbol in that file. 'strings' agrees.

Thanks,
Robin

#8 lore raw
In reply to Andrew Morton
Quoted 3 lines · Tony

Still need to figure out exactly what broke the build with a 3.x gcc to see if there is an easy workaround. Though 4.0.0 was released over three years ago ... is it time to give up on the ancient 3.x technology?

Yes. I've got a hard time seeing why anyone would use GCC 3.x to build current Linux kernels for Itanium (except when using an acient distro as the base for kernel development).

Quoted 2 lines

[*] If you ignore that the 4.2.4 compiler is about 15% slower than my trusty old 3.4.6 :-(

The resulting binary should be faster, though. ;-) How about giving GCC 4.3 a try, by the way? IIRC, GCC 4.2 has some aliasing issues that have been addressed.

Gerald

#9 lore raw
In reply to Andrew Morton
Quoted 1 line

The resulting binary should be faster, though. ;-)

Ah, but I spend much more time compiling kernels than I do running them ... so this isn't a good tradeoff for me ;-)

Quoted 3 lines

How about giving
GCC 4.3 a try, by the way? IIRC, GCC 4.2 has some aliasing issues that have been addressed.

I tried 4.3.1 and got:

configure: error: Building GCC requires GMP 4.1+ and MPFR 2.3.0+.

So I'll have to do some other upgrading and installing on my ancient build box to make this work :-(

-Tony

#10 lore raw
In reply to Andrew Morton

[holt@robinlt allmodconfig]$ git-bisect good e63352fa06fb8c476e56e51323d80f2e7baee67f is first bad commit commit e63352fa06fb8c476e56e51323d80f2e7baee67f

Author: Eduard - Gabriel Munteanu <[email protected]>
Date:   Tue Aug 19 20:43:26 2008 +0300

kmemtrace: SLUB hooks.

This adds hooks for the SLUB allocator, to allow tracing with kmemtrace.

:040000 040000 913348ac82982bd048439179f5d9e2231cc42ee6 984c6232d4eb257d8b31fe104e8731d8d5d6af18 M include
:040000 040000 eca09754017963bda58bc53e2863f59431df0ea7 92bc3efb5a7b9bc254f17def2c31b3c0a030122e M mm

To sum up, This commit when compiled for IA64 using Andrew's gcc 3.4.5 cross compiler and allmodconfig results in a build failure with:

arch/ia64/kernel/built-in.o(.text+0x4852): In function `efi_initialize_iomem_resources':
/work/git-linus/allmodconfig/arch/ia64/kernel/efi.c:1200: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x12992): In function pfm_context_create':
/work/git-linus/allmodconfig/arch/ia64/kernel/perfmon.c:2683: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x1ba22): In function sys_perfmonctl':
/work/git-linus/allmodconfig/arch/ia64/kernel/perfmon.c:4808: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x379a2): In function dup_state_stack':
/work/git-linus/allmodconfig/arch/ia64/kernel/unwind.c:645: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x37ec2): In function desc_prologue':
/work/git-linus/allmodconfig/arch/ia64/kernel/unwind.c:809: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x3a822):/work/git-linus/allmodconfig/arch/ia64/kernel/unwind_decoder.c:368: more undefined references to __ia64_save_stack_nonlocal' follow

With gcc 4.1.2 it compiles fine.

Thanks,
Robin

Quoted 21 lines · Tony
Quoted 1 line

The resulting binary should be faster, though. ;-)

Ah, but I spend much more time compiling kernels than I do running them ... so this isn't a good tradeoff for me ;-)

Quoted 3 lines

How about giving
GCC 4.3 a try, by the way? IIRC, GCC 4.2 has some aliasing issues that have been addressed.

I tried 4.3.1 and got:

configure: error: Building GCC requires GMP 4.1+ and MPFR 2.3.0+.

So I'll have to do some other upgrading and installing on my ancient build box to make this work :-(

-Tony
--
To unsubscribe from this list: send the line "unsubscribe linux-ia64" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#11 lore raw
In reply to Andrew Morton

Argh, I forgot to add the relavent people.

Robin

Quoted 60 lines · Robin Holt

[holt@robinlt allmodconfig]$ git-bisect good e63352fa06fb8c476e56e51323d80f2e7baee67f is first bad commit commit e63352fa06fb8c476e56e51323d80f2e7baee67f

Author: Eduard - Gabriel Munteanu <[email protected]>
Date:   Tue Aug 19 20:43:26 2008 +0300

kmemtrace: SLUB hooks.

This adds hooks for the SLUB allocator, to allow tracing with kmemtrace.

:040000 040000 913348ac82982bd048439179f5d9e2231cc42ee6 984c6232d4eb257d8b31fe104e8731d8d5d6af18 M include
:040000 040000 eca09754017963bda58bc53e2863f59431df0ea7 92bc3efb5a7b9bc254f17def2c31b3c0a030122e M mm

To sum up, This commit when compiled for IA64 using Andrew's gcc 3.4.5 cross compiler and allmodconfig results in a build failure with:

arch/ia64/kernel/built-in.o(.text+0x4852): In function `efi_initialize_iomem_resources':
/work/git-linus/allmodconfig/arch/ia64/kernel/efi.c:1200: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x12992): In function pfm_context_create':
/work/git-linus/allmodconfig/arch/ia64/kernel/perfmon.c:2683: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x1ba22): In function sys_perfmonctl':
/work/git-linus/allmodconfig/arch/ia64/kernel/perfmon.c:4808: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x379a2): In function dup_state_stack':
/work/git-linus/allmodconfig/arch/ia64/kernel/unwind.c:645: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x37ec2): In function desc_prologue':
/work/git-linus/allmodconfig/arch/ia64/kernel/unwind.c:809: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x3a822):/work/git-linus/allmodconfig/arch/ia64/kernel/unwind_decoder.c:368: more undefined references to __ia64_save_stack_nonlocal' follow

With gcc 4.1.2 it compiles fine.

Thanks,
Robin

Quoted 21 lines · Tony
Quoted 1 line

The resulting binary should be faster, though. ;-)

Ah, but I spend much more time compiling kernels than I do running them ... so this isn't a good tradeoff for me ;-)

Quoted 3 lines

How about giving
GCC 4.3 a try, by the way? IIRC, GCC 4.2 has some aliasing issues that have been addressed.

I tried 4.3.1 and got:

configure: error: Building GCC requires GMP 4.1+ and MPFR 2.3.0+.

So I'll have to do some other upgrading and installing on my ancient build box to make this work :-(

-Tony
--
To unsubscribe from this list: send the line "unsubscribe linux-ia64" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#12 lore raw
In reply to Andrew Morton
Quoted 38 lines · Robin Holt

Argh, I forgot to add the relavent people.

Robin

Quoted 33 lines · Robin Holt

[holt@robinlt allmodconfig]$ git-bisect good e63352fa06fb8c476e56e51323d80f2e7baee67f is first bad commit commit e63352fa06fb8c476e56e51323d80f2e7baee67f

Author: Eduard - Gabriel Munteanu <[email protected]>
Date:   Tue Aug 19 20:43:26 2008 +0300

kmemtrace: SLUB hooks.

This adds hooks for the SLUB allocator, to allow tracing with kmemtrace.

:040000 040000 913348ac82982bd048439179f5d9e2231cc42ee6 984c6232d4eb257d8b31fe104e8731d8d5d6af18 M include
:040000 040000 eca09754017963bda58bc53e2863f59431df0ea7 92bc3efb5a7b9bc254f17def2c31b3c0a030122e M mm

To sum up, This commit when compiled for IA64 using Andrew's gcc 3.4.5 cross compiler and allmodconfig results in a build failure with:

arch/ia64/kernel/built-in.o(.text+0x4852): In function `efi_initialize_iomem_resources':
/work/git-linus/allmodconfig/arch/ia64/kernel/efi.c:1200: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x12992): In function pfm_context_create':
/work/git-linus/allmodconfig/arch/ia64/kernel/perfmon.c:2683: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x1ba22): In function sys_perfmonctl':
/work/git-linus/allmodconfig/arch/ia64/kernel/perfmon.c:4808: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x379a2): In function dup_state_stack':
/work/git-linus/allmodconfig/arch/ia64/kernel/unwind.c:645: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x37ec2): In function desc_prologue':
/work/git-linus/allmodconfig/arch/ia64/kernel/unwind.c:809: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x3a822):/work/git-linus/allmodconfig/arch/ia64/kernel/unwind_decoder.c:368: more undefined references to __ia64_save_stack_nonlocal' follow

With gcc 4.1.2 it compiles fine.

So efi_initialize_iomem_resources calls kzalloc() which calls an inline kmalloc() function. Is this with our without CONFIG_KMEMTRACE enabled? Any idea what this '__ia64_save_stack_nonlocal' thing is?

Eduard, did you have plans to move the hooks out-of-line as suggested by Christoph? That patch might "cure" this thing here...

Pekka

#13 lore raw
In reply to Andrew Morton
Quoted 38 lines · Pekka Enberg
Quoted 37 lines · Robin Holt

Argh, I forgot to add the relavent people.

Robin

Quoted 32 lines · Robin Holt

[holt@robinlt allmodconfig]$ git-bisect good e63352fa06fb8c476e56e51323d80f2e7baee67f is first bad commit commit e63352fa06fb8c476e56e51323d80f2e7baee67f

Author: Eduard - Gabriel Munteanu <[email protected]>
Date:   Tue Aug 19 20:43:26 2008 +0300

kmemtrace: SLUB hooks.

This adds hooks for the SLUB allocator, to allow tracing with kmemtrace.

:040000 040000 913348ac82982bd048439179f5d9e2231cc42ee6 984c6232d4eb257d8b31fe104e8731d8d5d6af18 M include
:040000 040000 eca09754017963bda58bc53e2863f59431df0ea7 92bc3efb5a7b9bc254f17def2c31b3c0a030122e M mm

To sum up, This commit when compiled for IA64 using Andrew's gcc 3.4.5 cross compiler and allmodconfig results in a build failure with:

arch/ia64/kernel/built-in.o(.text+0x4852): In function `efi_initialize_iomem_resources':
/work/git-linus/allmodconfig/arch/ia64/kernel/efi.c:1200: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x12992): In function pfm_context_create':
/work/git-linus/allmodconfig/arch/ia64/kernel/perfmon.c:2683: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x1ba22): In function sys_perfmonctl':
/work/git-linus/allmodconfig/arch/ia64/kernel/perfmon.c:4808: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x379a2): In function dup_state_stack':
/work/git-linus/allmodconfig/arch/ia64/kernel/unwind.c:645: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x37ec2): In function desc_prologue':
/work/git-linus/allmodconfig/arch/ia64/kernel/unwind.c:809: undefined reference to __ia64_save_stack_nonlocal' arch/ia64/kernel/built-in.o(.text+0x3a822):/work/git-linus/allmodconfig/arch/ia64/kernel/unwind_decoder.c:368: more undefined references to __ia64_save_stack_nonlocal' follow

Quoted 5 lines
Quoted 1 line
Quoted 1 line

With gcc 4.1.2 it compiles fine.

So efi_initialize_iomem_resources calls kzalloc() which calls an inline kmalloc() function. Is this with our without CONFIG_KMEMTRACE enabled? Any idea what this '__ia64_save_stack_nonlocal' thing is?

This failed with either CONFIG_KMEMTRACE =y or =n.

Quoted 2 lines

Eduard, did you have plans to move the hooks out-of-line as suggested by Christoph? That patch might "cure" this thing here...

I probably could not test this until the weekend at the earliest. If you would like to test it yourself, just git-checkout the linux-next/master tree and use Andrew's cross compiler to make allmodconfig and make compressed.

Thanks,
Robin

#14 lore raw
In reply to Andrew Morton
Quoted 6 lines

I tried 4.3.1 and got:

configure: error: Building GCC requires GMP 4.1+ and MPFR 2.3.0+.

So I'll have to do some other upgrading and installing on my ancient build box to make this work :-(

I moved my build tools over to a newer, faster box. gcc4.3.1 built with no problems ... and built some kernels which seem to be working just fine.

-Tony