linux-next: s390 build failure

linux-next: s390 build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 1338 bytes --]

Hi Martin,

Starting with next-20080613, the linux-next build (s390 defconfig s390x
compiler) has failed with these errors:

In file included from /scratch/michael/kisskb-build/src/include/linux/spinlock.h:87,
                 from /scratch/michael/kisskb-build/src/include/linux/smp.h:11,
                 from /scratch/michael/kisskb-build/src/include/linux/kernel_stat.h:4,
                 from /scratch/michael/kisskb-build/src/mm/memory.c:41:
include2/asm/spinlock.h: In function '__raw_spin_lock':
include2/asm/spinlock.h:69: error: implicit declaration of function 'smp_processor_id'

There are others as well, but I was concentrating on this.  Bisecting
leads to commit 2448ad3d8e64a9b58acca34a3d010cfa0be88cc1 which is the
merge of the block tree.  The block tree contains a commit
(37e81b98dba08cce1d8eedb2bdaa58022bf31ee0 "Add generic helpers for arch
IPI function calls") that adds an include of linux/spinlock.h to
linux/smp.h. (asm-s390/spinlock.h includes linux/smp.h ...)

But the block tree itself is OK.  So something in the s390 tree is
interacting with the above commit to cause the failure.  It is not
immediately obvious to me what that is. (Or I may have screwed up the
bisect :-))

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: linux-next: s390 build failure

[Jens Axboe]

On Fri, Jun 20 2008, Stephen Rothwell wrote:
> Hi Martin,
> 
> Starting with next-20080613, the linux-next build (s390 defconfig s390x
> compiler) has failed with these errors:
> 
> In file included from /scratch/michael/kisskb-build/src/include/linux/spinlock.h:87,
>                  from /scratch/michael/kisskb-build/src/include/linux/smp.h:11,
>                  from /scratch/michael/kisskb-build/src/include/linux/kernel_stat.h:4,
>                  from /scratch/michael/kisskb-build/src/mm/memory.c:41:
> include2/asm/spinlock.h: In function '__raw_spin_lock':
> include2/asm/spinlock.h:69: error: implicit declaration of function 'smp_processor_id'
> 
> There are others as well, but I was concentrating on this.  Bisecting
> leads to commit 2448ad3d8e64a9b58acca34a3d010cfa0be88cc1 which is the
> merge of the block tree.  The block tree contains a commit
> (37e81b98dba08cce1d8eedb2bdaa58022bf31ee0 "Add generic helpers for arch
> IPI function calls") that adds an include of linux/spinlock.h to
> linux/smp.h. (asm-s390/spinlock.h includes linux/smp.h ...)
> 
> But the block tree itself is OK.  So something in the s390 tree is
> interacting with the above commit to cause the failure.  It is not
> immediately obvious to me what that is. (Or I may have screwed up the
> bisect :-))

Heiko was looking into this, I had forgotten about it. Heiko, did you
find an adequate solution to this?

-- 
Jens Axboe

Re: linux-next: s390 build failure

[Martin Schwidefsky]

On Fri, 2008-06-20 at 12:42 +1000, Stephen Rothwell wrote:
> In file included from /scratch/michael/kisskb-build/src/include/linux/spinlock.h:87,
>                  from /scratch/michael/kisskb-build/src/include/linux/smp.h:11,
>                  from /scratch/michael/kisskb-build/src/include/linux/kernel_stat.h:4,
>                  from /scratch/michael/kisskb-build/src/mm/memory.c:41:
> include2/asm/spinlock.h: In function '__raw_spin_lock':
> include2/asm/spinlock.h:69: error: implicit declaration of function 'smp_processor_id'
> 
> There are others as well, but I was concentrating on this.  Bisecting
> leads to commit 2448ad3d8e64a9b58acca34a3d010cfa0be88cc1 which is the
> merge of the block tree.  The block tree contains a commit
> (37e81b98dba08cce1d8eedb2bdaa58022bf31ee0 "Add generic helpers for arch
> IPI function calls") that adds an include of linux/spinlock.h to
> linux/smp.h. (asm-s390/spinlock.h includes linux/smp.h ...)
> 
> But the block tree itself is OK.  So something in the s390 tree is
> interacting with the above commit to cause the failure.  It is not
> immediately obvious to me what that is. (Or I may have screwed up the
> bisect :-))

Hmm, the minimal test case for this is:

#include <linux/smp.h>

* linux/smp.h defines __LINUX_SMP_H then includes linux/spinlock.h
* linux/spinlock.h includes some more headers, then asm/spinlock.h
* asm/spinlock.h includes linux/smp.h with no effect because linux/smp.h
has already defined __LINUX_SMP_H, then uses smp_processor_id() which is
not defined yet
* on return from the linux/spinlock.h include linux/smp.h defines
smp_processor_id()

Relavant commits that cause this problem:

1)
commit 3888f22c9cc8b4eca034f11b13d7f7957627fcfb
Author: Jens Axboe <jens.axboe@oracle.com>
Date:   Fri Jun 6 10:58:32 2008 +0200
Add generic helpers for arch IPI function calls

This commit introduces the #include <linux/spinlock.h> to linux/smp.h

-- 
blue skies,
  Martin.

"Reality continues to ruin my life." - Calvin.

Re: linux-next: s390 build failure

[Heiko Carstens]

On Fri, Jun 20, 2008 at 01:42:48PM +0200, Jens Axboe wrote:
> On Fri, Jun 20 2008, Stephen Rothwell wrote:
> > Hi Martin,
> > 
> > Starting with next-20080613, the linux-next build (s390 defconfig s390x
> > compiler) has failed with these errors:
> > 
> > In file included from /scratch/michael/kisskb-build/src/include/linux/spinlock.h:87,
> >                  from /scratch/michael/kisskb-build/src/include/linux/smp.h:11,
> >                  from /scratch/michael/kisskb-build/src/include/linux/kernel_stat.h:4,
> >                  from /scratch/michael/kisskb-build/src/mm/memory.c:41:
> > include2/asm/spinlock.h: In function '__raw_spin_lock':
> > include2/asm/spinlock.h:69: error: implicit declaration of function 'smp_processor_id'
> > 
> > There are others as well, but I was concentrating on this.  Bisecting
> > leads to commit 2448ad3d8e64a9b58acca34a3d010cfa0be88cc1 which is the
> > merge of the block tree.  The block tree contains a commit
> > (37e81b98dba08cce1d8eedb2bdaa58022bf31ee0 "Add generic helpers for arch
> > IPI function calls") that adds an include of linux/spinlock.h to
> > linux/smp.h. (asm-s390/spinlock.h includes linux/smp.h ...)
> > 
> > But the block tree itself is OK.  So something in the s390 tree is
> > interacting with the above commit to cause the failure.  It is not
> > immediately obvious to me what that is. (Or I may have screwed up the
> > bisect :-))
> 
> Heiko was looking into this, I had forgotten about it. Heiko, did you
> find an adequate solution to this?

My only conclusion back then was that you shouldn't add the new
#include <linux/spinlock.h> to include <linux/smp.h> (see also Martin's mail).

Re: linux-next: s390 build failure

[Jens Axboe]

On Sat, Jun 21 2008, Heiko Carstens wrote:
> On Fri, Jun 20, 2008 at 01:42:48PM +0200, Jens Axboe wrote:
> > On Fri, Jun 20 2008, Stephen Rothwell wrote:
> > > Hi Martin,
> > > 
> > > Starting with next-20080613, the linux-next build (s390 defconfig s390x
> > > compiler) has failed with these errors:
> > > 
> > > In file included from /scratch/michael/kisskb-build/src/include/linux/spinlock.h:87,
> > >                  from /scratch/michael/kisskb-build/src/include/linux/smp.h:11,
> > >                  from /scratch/michael/kisskb-build/src/include/linux/kernel_stat.h:4,
> > >                  from /scratch/michael/kisskb-build/src/mm/memory.c:41:
> > > include2/asm/spinlock.h: In function '__raw_spin_lock':
> > > include2/asm/spinlock.h:69: error: implicit declaration of function 'smp_processor_id'
> > > 
> > > There are others as well, but I was concentrating on this.  Bisecting
> > > leads to commit 2448ad3d8e64a9b58acca34a3d010cfa0be88cc1 which is the
> > > merge of the block tree.  The block tree contains a commit
> > > (37e81b98dba08cce1d8eedb2bdaa58022bf31ee0 "Add generic helpers for arch
> > > IPI function calls") that adds an include of linux/spinlock.h to
> > > linux/smp.h. (asm-s390/spinlock.h includes linux/smp.h ...)
> > > 
> > > But the block tree itself is OK.  So something in the s390 tree is
> > > interacting with the above commit to cause the failure.  It is not
> > > immediately obvious to me what that is. (Or I may have screwed up the
> > > bisect :-))
> > 
> > Heiko was looking into this, I had forgotten about it. Heiko, did you
> > find an adequate solution to this?
> 
> My only conclusion back then was that you shouldn't add the new
> #include <linux/spinlock.h> to include <linux/smp.h> (see also Martin's mail).

OK, I changed it to not include spinlock.h and kill the
call_function_lock decleration. If generic smp helpers are now used,
smp.c will include functions for acquiring and releasing the ipi call
lock instead.

-- 
Jens Axboe

Re: linux-next: s390 build failure

[Heiko Carstens]

On Fri, 20 Mar 2009 11:00:22 +0100
Ingo Molnar <mingo@elte.hu> wrote:
> Btw., my cross-build scripts fail on s390 on latest -git, with:
> 
>  Assembler messages:
>  FATAL: can't create scripts/mod/.tmp_empty.o: Invalid bfd target
>  make[3]: *** [scripts/mod/empty.o] Error 2
>  make[3]: *** Waiting for unfinished jobs....

Haven't seen this yet.

> have i made some mistake in setting them up? I have this cross build 
> environment:
> 
> aldebaran:~/linux/linux> /opt/crossgcc/cross/bin/s390-linux-gcc -v
> Using built-in specs.
> Target: s390-linux
> Configured with: /home/mingo/work/projects/crossgcc/gcc/configure 
> --target=s390-linux --enable-targets=all 
> --prefix=/opt/crossgcc/cross --enable-languages=c --without-headers 
> --enable-sjlj-exceptions --with-system-libunwind --disable-nls 
> --disable-threads --disable-shared --disable-libmudflap 
> --disable-libssp --disable-libgomp --disable-decimal-float 
> --enable-checking=release
> Thread model: single
> gcc version 4.3.3 20081123 (prerelease) (GCC) 
> aldebaran:~/linux/linux> /opt/crossgcc/cross/bin/s390-linux-ld -v
> GNU ld (GNU Binutils) 2.19

ah, it's probably that you specified --target=s390-linux , that will
build a 32 bit compiler, not a 64 bit one. You probably want
s390x-ibm-linux. That gcc version can build both 32 and 64 bit binaries.

Latest cross compiler I have on my notebook that works:

heiko@osiris:~> s390x-ibm-linux-gcc-4.3.1 -v
Using built-in specs.
Target: s390x-ibm-linux
Configured with: ../gcc-4.3.1/configure --prefix=/home/heiko/cross/
--disable-nls --disable-checking --enable-languages=c
--target=s390x-ibm-linux --disable-threads --disable-shared
--without-headers --disable-libmudflap --disable-libssp
--disable-multilib --disable-libunwind-exceptions --disable-libgomp
--disable-werror --program-prefix=s390x-4.3.1-
Thread model: single gcc
version 4.3.1 (GCC) 


heiko@osiris:~> s390x-4.3.1-ld -v    
GNU ld (Linux/GNU Binutils) 2.19.51.0.1.20090106

> and i use:
> 
> CROSS="CROSS_COMPILE=$ARCH-linux-"
> 
> make O=$BUILDDIR $CROSS ARCH=$ARCH mrproper defconfig
> 
> where ARCH=s390 in this case.

should work.

Re: linux-next: s390 build failure

[Ingo Molnar]

* Heiko Carstens <heiko.carstens@de.ibm.com> wrote:

> On Fri, 20 Mar 2009 10:42:08 +1100
> Stephen Rothwell <sfr@canb.auug.org.au> wrote:
> > The linux-next build for s390 allnoconfig failed like this:
> > 
> > In file included from arch/s390/include/asm/tlb.h:29,
> >                  from arch/s390/mm/pgtable.c:25:
> > arch/s390/include/asm/smp.h:95: error: redefinition of 'smp_send_stop'
> > include/linux/smp.h:125: error: previous definition of 'smp_send_stop' was here
> > 
> > Caused by commit d1dedb52acd98bd5e13e1ff4c4d045d58bbd16fe ("panic, smp:
> > provide smp_send_stop() wrapper on UP too") from the tip-core tree.
> 
> Ingo, could you pick up the patch below?

Applied, thanks guys!

Btw., my cross-build scripts fail on s390 on latest -git, with:

 Assembler messages:
 FATAL: can't create scripts/mod/.tmp_empty.o: Invalid bfd target
 make[3]: *** [scripts/mod/empty.o] Error 2
 make[3]: *** Waiting for unfinished jobs....

have i made some mistake in setting them up? I have this cross build 
environment:

aldebaran:~/linux/linux> /opt/crossgcc/cross/bin/s390-linux-gcc -v
Using built-in specs.
Target: s390-linux
Configured with: /home/mingo/work/projects/crossgcc/gcc/configure 
--target=s390-linux --enable-targets=all 
--prefix=/opt/crossgcc/cross --enable-languages=c --without-headers 
--enable-sjlj-exceptions --with-system-libunwind --disable-nls 
--disable-threads --disable-shared --disable-libmudflap 
--disable-libssp --disable-libgomp --disable-decimal-float 
--enable-checking=release
Thread model: single
gcc version 4.3.3 20081123 (prerelease) (GCC) 
aldebaran:~/linux/linux> /opt/crossgcc/cross/bin/s390-linux-ld -v
GNU ld (GNU Binutils) 2.19

and i use:

CROSS="CROSS_COMPILE=$ARCH-linux-"

make O=$BUILDDIR $CROSS ARCH=$ARCH mrproper defconfig

where ARCH=s390 in this case.

	Ingo

Re: linux-next: s390 build failure

[Heiko Carstens]

On Fri, 20 Mar 2009 10:42:08 +1100
Stephen Rothwell <sfr@canb.auug.org.au> wrote:
> The linux-next build for s390 allnoconfig failed like this:
> 
> In file included from arch/s390/include/asm/tlb.h:29,
>                  from arch/s390/mm/pgtable.c:25:
> arch/s390/include/asm/smp.h:95: error: redefinition of 'smp_send_stop'
> include/linux/smp.h:125: error: previous definition of 'smp_send_stop' was here
> 
> Caused by commit d1dedb52acd98bd5e13e1ff4c4d045d58bbd16fe ("panic, smp:
> provide smp_send_stop() wrapper on UP too") from the tip-core tree.

Ingo, could you pick up the patch below?


Subject: [PATCH] s390: remove arch specific smp_send_stop

From: Heiko Carstens <heiko.carstens@de.ibm.com>

Remove arch specific smp_send_stop for !CONFIG_SMP since it conflicts
with a new generic version.

Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
---
 arch/s390/include/asm/smp.h |    6 ------
 1 file changed, 6 deletions(-)

Index: linux-next/arch/s390/include/asm/smp.h
===================================================================
--- linux-next.orig/arch/s390/include/asm/smp.h
+++ linux-next/arch/s390/include/asm/smp.h
@@ -92,12 +92,6 @@ extern void arch_send_call_function_ipi(
 #endif
 
 #ifndef CONFIG_SMP
-static inline void smp_send_stop(void)
-{
-	/* Disable all interrupts/machine checks */
-	__load_psw_mask(psw_kernel_bits & ~PSW_MASK_MCHECK);
-}
-
 #define hard_smp_processor_id()		0
 #define smp_cpu_not_running(cpu)	1
 #endif

linux-next: s390 build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 575 bytes --]

Hi all,

Today's (and yesterday's) linux-next build of s390 defconfig failed like
this:

In file included from arch/s390/include/asm/tlb.h:30,
                 from arch/s390/mm/init.c:37:
arch/s390/include/asm/tlbflush.h: In function '__tlb_flush_full':
arch/s390/include/asm/tlbflush.h:54: error: incompatible type for argument 2 of 'cpumask_equal'

Caused by a bad merge on my part.  It will be fixed today as the
conflicting commit has been removed.

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

linux-next: s390 build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 583 bytes --]

Hi all,

The linux-next build for s390 allnoconfig failed like this:

In file included from arch/s390/include/asm/tlb.h:29,
                 from arch/s390/mm/pgtable.c:25:
arch/s390/include/asm/smp.h:95: error: redefinition of 'smp_send_stop'
include/linux/smp.h:125: error: previous definition of 'smp_send_stop' was here

Caused by commit d1dedb52acd98bd5e13e1ff4c4d045d58bbd16fe ("panic, smp:
provide smp_send_stop() wrapper on UP too") from the tip-core tree.
-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: linux-next: s390 build failure

[Greg KH]

On Mon, Jun 23, 2008 at 05:22:26PM +0200, Cornelia Huck wrote:
> On Fri, 20 Jun 2008 09:58:43 +0200,
> Cornelia Huck <cornelia.huck@de.ibm.com> wrote:
> 
> > I could also try to split out
> > dev_set_name() conversion from the init_name stuff, but I fear that
> > doesn't make much sense...
> 
> Well, so I did just that... I'll reply to this mail with two patches:
> 
> - "s390: bus_id -> dev_set_name() changes" should replace the patch of
> the same title. It removes the dependency on init_name.
> 
> - "s390: bus_id -> dev_set_name() for css and ccw busses" should go
> after "driver core: add init_name to struct device". It depends on
> init_name and consequently should be commented out until that patch is
> re-activated.
> 
> Greg: Could you please add these to your tree?

I've now done so, thanks for sending them to me.

greg k-h

Re: linux-next: s390 build failure

[Cornelia Huck]

On Fri, 20 Jun 2008 09:58:43 +0200,
Cornelia Huck <cornelia.huck@de.ibm.com> wrote:

> I could also try to split out
> dev_set_name() conversion from the init_name stuff, but I fear that
> doesn't make much sense...

Well, so I did just that... I'll reply to this mail with two patches:

- "s390: bus_id -> dev_set_name() changes" should replace the patch of
the same title. It removes the dependency on init_name.

- "s390: bus_id -> dev_set_name() for css and ccw busses" should go
after "driver core: add init_name to struct device". It depends on
init_name and consequently should be commented out until that patch is
re-activated.

Greg: Could you please add these to your tree?

Re: linux-next: s390 build failure

[Cornelia Huck]

On Fri, 20 Jun 2008 13:52:57 +1000,
Stephen Rothwell <sfr@canb.auug.org.au> wrote:

> Bisecting lead to the commit "s390: bus_id -> dev_set_name() changes" in
> the driver-core tree.  I suspect this was caused by the removal of
> "driver core: add init_name to struct device" from the driver-core tree
> on June 11 (even though the former patch was earlier in the series then
> the latter).

Yes, this patch should go after the introduction of init_name (and be
commented out in Greg's series file). I could also try to split out
dev_set_name() conversion from the init_name stuff, but I fear that
doesn't make much sense...

linux-next: s390 build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 658 bytes --]

Hi Greg, Cornelia,

Yesterday's (and back to next-20080612) linux-next build (s390
allnoconfig (and others) with an s390s compiler) fails with this error:

drivers/s390/cio/cio.c:573: error: 'struct device' has no member named 'init_name'

Bisecting lead to the commit "s390: bus_id -> dev_set_name() changes" in
the driver-core tree.  I suspect this was caused by the removal of
"driver core: add init_name to struct device" from the driver-core tree
on June 11 (even though the former patch was earlier in the series then
the latter).

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: linux-next: s390 build failure

[Greg KH]

On Mon, May 12, 2008 at 06:03:35PM +1000, Stephen Rothwell wrote:
> Hi Greg,
> 
> Today's linux-next build (s390 defconfig) fails like this:
> 
> drivers/s390/cio/device_fsm.c:53: error: syntax error before ';' token
> drivers/s390/cio/device_fsm.c:54: error: syntax error before ';' token
> 
> Caused by commit 67da378c80848bb778ee8caecadcdf17e1e310e0 ("driver-core:
> prepare for removal of 20 char limit from struct device") from the
> driver-core tree. Missing close parentheses on added dev_name() calls.
> 
> A touch more care, please.

I have updated version of this in my to-apply queue, I've dropped this
now so linux-next tomorrow should be just fine.  I'll add it back in
when I have the cross-build system set up here so that I can test things
to prevent this kind of mess from happening in the future.

Sorry about that.

greg k-h

linux-next: s390 build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 573 bytes --]

Hi Greg,

Today's linux-next build (s390 defconfig) fails like this:

drivers/s390/cio/device_fsm.c:53: error: syntax error before ';' token
drivers/s390/cio/device_fsm.c:54: error: syntax error before ';' token

Caused by commit 67da378c80848bb778ee8caecadcdf17e1e310e0 ("driver-core:
prepare for removal of 20 char limit from struct device") from the
driver-core tree. Missing close parentheses on added dev_name() calls.

A touch more care, please.

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]