stacktrace.c:(.text+0x1b0): undefined reference to `.smp_send_safe_nmi_ipi'

stacktrace.c:(.text+0x1b0): undefined reference to `.smp_send_safe_nmi_ipi'

[Christophe LEROY]

See http://kisskb.ellerman.id.au/kisskb/buildresult/13395345/ for details

Seems like that function only exists when CONFIG_NMI_IPI is defined.

Problem introduced by commit 5cc05910f26e6fd6da15f052f86f6150e4b91664 
https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/arch/powerpc/kernel/stacktrace.c?h=next-20180614&id=5cc05910f26e6fd6da15f052f86f6150e4b91664


What do you recommend to fix that ?


Christophe

Re: stacktrace.c:(.text+0x1b0): undefined reference to `.smp_send_safe_nmi_ipi'

[Nicholas Piggin]

On Thu, 14 Jun 2018 12:06:51 +0200
Christophe LEROY <christophe.leroy@c-s.fr> wrote:

> See http://kisskb.ellerman.id.au/kisskb/buildresult/13395345/ for details
> 
> Seems like that function only exists when CONFIG_NMI_IPI is defined.
> 
> Problem introduced by commit 5cc05910f26e6fd6da15f052f86f6150e4b91664 
> https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/arch/powerpc/kernel/stacktrace.c?h=next-20180614&id=5cc05910f26e6fd6da15f052f86f6150e4b91664
> 
> 
> What do you recommend to fix that ?

We could just do smp_call_function in the case we have no NMI_IPI?

Thanks,
Nick

Re: stacktrace.c:(.text+0x1b0): undefined reference to `.smp_send_safe_nmi_ipi'

[Christophe LEROY]

Le 14/06/2018 à 12:25, Nicholas Piggin a écrit :
> On Thu, 14 Jun 2018 12:06:51 +0200
> Christophe LEROY <christophe.leroy@c-s.fr> wrote:
> 
>> See http://kisskb.ellerman.id.au/kisskb/buildresult/13395345/ for details
>>
>> Seems like that function only exists when CONFIG_NMI_IPI is defined.
>>
>> Problem introduced by commit 5cc05910f26e6fd6da15f052f86f6150e4b91664
>> https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/arch/powerpc/kernel/stacktrace.c?h=next-20180614&id=5cc05910f26e6fd6da15f052f86f6150e4b91664
>>
>>
>> What do you recommend to fix that ?
> 
> We could just do smp_call_function in the case we have no NMI_IPI?

???

Can you detail what you have in mind ? Do smp_call_function() to call 
what ? Instead of the call to smp_send_safe_nmi_ipi() ?

Thanks
Christophe

> 
> Thanks,
> Nick
> 

Re: stacktrace.c:(.text+0x1b0): undefined reference to `.smp_send_safe_nmi_ipi'

[Nicholas Piggin]

On Thu, 14 Jun 2018 12:51:28 +0200
Christophe LEROY <christophe.leroy@c-s.fr> wrote:

> Le 14/06/2018 =C3=A0 12:25, Nicholas Piggin a =C3=A9crit=C2=A0:
> > On Thu, 14 Jun 2018 12:06:51 +0200
> > Christophe LEROY <christophe.leroy@c-s.fr> wrote:
> >  =20
> >> See http://kisskb.ellerman.id.au/kisskb/buildresult/13395345/ for deta=
ils
> >>
> >> Seems like that function only exists when CONFIG_NMI_IPI is defined.
> >>
> >> Problem introduced by commit 5cc05910f26e6fd6da15f052f86f6150e4b91664
> >> https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/co=
mmit/arch/powerpc/kernel/stacktrace.c?h=3Dnext-20180614&id=3D5cc05910f26e6f=
d6da15f052f86f6150e4b91664
> >>
> >>
> >> What do you recommend to fix that ? =20
> >=20
> > We could just do smp_call_function in the case we have no NMI_IPI? =20
>=20
> ???
>=20
> Can you detail what you have in mind ? Do smp_call_function() to call=20
> what ? Instead of the call to smp_send_safe_nmi_ipi() ?

Yes I was thinking that but now I take a better look there's places that
call it with irqs off so that might be a bit risky. I guess just making
it depend on NMI_IPI for now is the minimum fix.

Thanks,
Nick

Re: stacktrace.c:(.text+0x1b0): undefined reference to `.smp_send_safe_nmi_ipi'

[Michael Ellerman]

Christophe LEROY <christophe.leroy@c-s.fr> writes:

> See http://kisskb.ellerman.id.au/kisskb/buildresult/13395345/ for details
>
> Seems like that function only exists when CONFIG_NMI_IPI is defined.
>
> Problem introduced by commit 5cc05910f26e6fd6da15f052f86f6150e4b91664 
> https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/arch/powerpc/kernel/stacktrace.c?h=next-20180614&id=5cc05910f26e6fd6da15f052f86f6150e4b91664
>
>
> What do you recommend to fix that ?

It should just use CONFIG_NMI_IPI instead of CONFIG_PPC_BOOK3S_64 I
think.

I'll do a patch.

cheers