[PATCH 4/5] powerpc: Define config option for processors without broadcast TLBIE

[Nicholas Piggin]

On Wed Jan 29, 2025 at 5:10 PM AEST, Paul Mackerras wrote:
> On Wed, Jan 29, 2025 at 04:14:25PM +1000, Nicholas Piggin wrote:
>> On Wed Jan 29, 2025 at 8:53 AM AEST, Paul Mackerras wrote:
>> > Power ISA v3.1 implementations in the Linux Compliancy Subset and
>> > lower are not required to implement broadcast TLBIE, and in fact
>> > Microwatt doesn't.  To avoid the need to specify "disable_tlbie" on
>> > the kernel command line on SMP Microwatt systems, this defines a
>> > config option that asserts that broadcast TLBIE should never be used
>> > (the kernel will instead use IPIs to trigger local TLBIEs on other
>> > CPUs when required).
>> >
>> > Signed-off-by: Paul Mackerras <paulus at ozlabs.org>
>> > ---
>> >  arch/powerpc/mm/book3s64/pgtable.c       | 10 ++++++++--
>> >  arch/powerpc/platforms/Kconfig.cputype   | 12 ++++++++++++
>> >  arch/powerpc/platforms/microwatt/Kconfig |  1 +
>> >  3 files changed, 21 insertions(+), 2 deletions(-)
>> >
>> > diff --git a/arch/powerpc/mm/book3s64/pgtable.c b/arch/powerpc/mm/book3s64/pgtable.c
>> > index 374542528080..14ee96e2a581 100644
>> > --- a/arch/powerpc/mm/book3s64/pgtable.c
>> > +++ b/arch/powerpc/mm/book3s64/pgtable.c
>> > @@ -588,10 +588,16 @@ int pmd_move_must_withdraw(struct spinlock *new_pmd_ptl,
>> >  }
>> >  #endif
>> >  
>> > +#ifndef CONFIG_PPC_RADIX_NO_BROADCAST_TLBIE
>> 
>> Hate to bikeshed, but would it be annoying to make this an affirmative
>> option?
>
> I guess we'd have to make all the platforms that do have broadcast
> tlbie (and a book3s-64 MMU with radix) select that option.  Which
> would be powernv and pseries, I would think.  If that's correct then
> it's probably not too annoying.  Should I do that in v2?

I think you're right, powernv and pseries. If you wouldn't mind doing
it please.

Thanks,
Nick