Re: [PATCH 3/4] mm/tlb, x86/mm: Support invalidating TLB caches for RCU_TABLE_FREE

From: Nicholas Piggin
Date: Mon Aug 27 2018 - 04:54:18 EST

On Mon, 27 Aug 2018 10:20:45 +0200
Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:

> On Mon, Aug 27, 2018 at 06:09:50PM +1000, Benjamin Herrenschmidt wrote:
> > Sadly our architecture requires a precise match between the page size
> > specified in the tlbie instruction and the entry in the TLB or it won't
> > be flushed.
> Argh.. OK I see. That is rather unfortunate and does seem to require
> something along the lines of tlb_remove_check_page_size_change().

Or we can do better with some more of our own data in mmu_gather,
but things that probably few or no other architectures want. I've
held off trying to put any crap in generic code because there's
other lower hanging fruit still, but I'd really rather just give
archs the ability to put their own data in there. I don't really
see a downside to it (divergence of course, but the existing
proliferation of code is much harder to follow than some data that
would be maintained and used purely by the arch, and beats having
to implement entirely your own mmu_gather).