linux-next: scsi-post-merge tree build failure

linux-next: scsi-post-merge tree build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 578 bytes --]

Hi Paul, Ben,

Today's linux-next build (powerpc ppc44x_config) failed like this:

arch/powerpc/kvm/powerpc.c: In function 'kvm_arch_vcpu_put':
arch/powerpc/kvm/powerpc.c:333: error: implicit declaration of function '_tlbil_all'

Probably caused by commit 2a4aca1144394653269720ffbb5a325a77abd5fa
("powerpc/mm: Split low level tlb invalidate for nohash processors").
Reverting just that commit was too difficult, so I have reverted the
powerpc tree for today.
-- 
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: scsi-post-merge tree build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 150 bytes --]

I meant the powerpc tree, of course ...

-- 
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: scsi-post-merge tree build failure

[Benjamin Herrenschmidt]

On Tue, 2008-12-23 at 01:25 +1100, Stephen Rothwell wrote:
> Hi Paul, Ben,
> 
> Today's linux-next build (powerpc ppc44x_config) failed like this:
> 
> arch/powerpc/kvm/powerpc.c: In function 'kvm_arch_vcpu_put':
> arch/powerpc/kvm/powerpc.c:333: error: implicit declaration of function '_tlbil_all'
> 
> Probably caused by commit 2a4aca1144394653269720ffbb5a325a77abd5fa
> ("powerpc/mm: Split low level tlb invalidate for nohash processors").
> Reverting just that commit was too difficult, so I have reverted the
> powerpc tree for today.

You can't revert that without reverting my whole series.

Adding a #include ../mm/mmu_decl.h to the kvm code should fix it for
now.

I need to talk to hollis about the right way to do that stuff in
the long run but it might be it... those _tlbil things are low level
stuff that aren't supposed to be used by the outside world which is
why I moved the declarations there, but KVM is also low level :-)

Cheers,
Ben.

testing (Was: Re: linux-next: scsi-post-merge tree build failure)

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 1640 bytes --]

Hi Ben,

On Tue, 23 Dec 2008 07:56:32 +1100 Benjamin Herrenschmidt <benh@kernel.crashing.org> wrote:
>
> On Tue, 2008-12-23 at 01:25 +1100, Stephen Rothwell wrote:
> > Hi Paul, Ben,
> > 
> > Today's linux-next build (powerpc ppc44x_config) failed like this:
> > 
> > arch/powerpc/kvm/powerpc.c: In function 'kvm_arch_vcpu_put':
> > arch/powerpc/kvm/powerpc.c:333: error: implicit declaration of function '_tlbil_all'
> > 
> > Probably caused by commit 2a4aca1144394653269720ffbb5a325a77abd5fa
> > ("powerpc/mm: Split low level tlb invalidate for nohash processors").
> > Reverting just that commit was too difficult, so I have reverted the
> > powerpc tree for today.
> 
> You can't revert that without reverting my whole series.

I know that ... I tried.

> Adding a #include ../mm/mmu_decl.h to the kvm code should fix it for
> now.
> 
> I need to talk to hollis about the right way to do that stuff in
> the long run but it might be it... those _tlbil things are low level
> stuff that aren't supposed to be used by the outside world which is
> why I moved the declarations there, but KVM is also low level :-)

So, why is this series in powerpc/next at all if some of the modified
files haven't even been compiled?  A simple ppc44x_defconfig build of the
powerpc/next tree fails - which I would have thought would be part of the
testing since KVM is only implemented for 44x on powerpc.

(Sorry, I am annoyed - it was late when I finally found this build
failure on top of everything else.)

-- 
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: testing (Was: Re: linux-next: scsi-post-merge tree build failure)

[Benjamin Herrenschmidt]

On Tue, 2008-12-23 at 10:04 +1100, Stephen Rothwell wrote:
> So, why is this series in powerpc/next at all if some of the modified
> files haven't even been compiled?  

Because nobody spotted the problem while the series was out for review ?

> A simple ppc44x_defconfig build of the
> powerpc/next tree fails - which I would have thought would be part of the
> testing since KVM is only implemented for 44x on powerpc.
> 
> (Sorry, I am annoyed - it was late when I finally found this build
> failure on top of everything else.)

I have about 3 44x configs in my build tests. KVM happens not to be
enabled in any of them :-( I'll add this ppc4xx_defconfig to the lot, I
didn't know it existed in fact.

Cheers,
Ben.

Re: testing (Was: Re: linux-next: scsi-post-merge tree build failure)

[Josh Boyer]

On Tue, Dec 23, 2008 at 10:04:56AM +1100, Stephen Rothwell wrote:
>Hi Ben,
>
>On Tue, 23 Dec 2008 07:56:32 +1100 Benjamin Herrenschmidt <benh@kernel.crashing.org> wrote:
>>
>> On Tue, 2008-12-23 at 01:25 +1100, Stephen Rothwell wrote:
>> > Hi Paul, Ben,
>> > 
>> > Today's linux-next build (powerpc ppc44x_config) failed like this:
>> > 
>> > arch/powerpc/kvm/powerpc.c: In function 'kvm_arch_vcpu_put':
>> > arch/powerpc/kvm/powerpc.c:333: error: implicit declaration of function '_tlbil_all'
>> > 
>> > Probably caused by commit 2a4aca1144394653269720ffbb5a325a77abd5fa
>> > ("powerpc/mm: Split low level tlb invalidate for nohash processors").
>> > Reverting just that commit was too difficult, so I have reverted the
>> > powerpc tree for today.
>> 
>> You can't revert that without reverting my whole series.
>
>I know that ... I tried.
>
>> Adding a #include ../mm/mmu_decl.h to the kvm code should fix it for
>> now.
>> 
>> I need to talk to hollis about the right way to do that stuff in
>> the long run but it might be it... those _tlbil things are low level
>> stuff that aren't supposed to be used by the outside world which is
>> why I moved the declarations there, but KVM is also low level :-)
>
>So, why is this series in powerpc/next at all if some of the modified
>files haven't even been compiled?  A simple ppc44x_defconfig build of the
>powerpc/next tree fails - which I would have thought would be part of the
>testing since KVM is only implemented for 44x on powerpc.

Well, they were, and then they weren't.  I tested this series starting
with the ppc44x_defconfig.  However, I was incrementally compiling them
and I also had the 16/64K pages patch applied and 64K enabled.  That
resulted in a kernel that was too big for the wrapper, so I went in
and turned some stuff off to make the kernel smaller.  That included
the KVM stuff.  By the time I was done building and testing the whole
series, it had remained disabled.

That might sound like a cop-out, but it's what happend.  I should have
done a simple build after I was done.  Sorry Stephen.

josh