[PATCH 0/2] Faster MMU lookups for Book3s v3
benh at kernel.crashing.org
Fri Jul 2 12:50:29 EST 2010
On Thu, 2010-07-01 at 14:52 +0200, Alexander Graf wrote:
> Page ageing is difficult. The HTAB has a hardware set referenced bit,
> but we don't have a guarantee that the entry is still there when we look
> for it. Something else could have overwritten it by then, but the entry
> could still be lingering around in the TLB.
> So I think the only reasonable way to implement page ageing is to unmap
> pages. And that's slow, because it means we have to map them again on
> access. Bleks. Or we could look for the HTAB entry and only unmap them
> if the entry is moot.
Well, not quite.
We -could- use the HW reference bit. However, that means that whenever
we flush the hash PTE we get a snapshot of the HW bit and copy it over
to the PTE.
That's not -that- bad for normal invalidations. However, it's a problem
potentially for eviction. IE. When a hash bucket is full, we
pseudo-randomly evict a slot. If we were to use the HW ref bit, we would
need a way to go back to the PTE from the hash bucket to perform that
update (or something really tricky like sticking it in a list somewhere,
and have the young test walk that list when non-empty, etc...)
More information about the Linuxppc-dev