Re: [PATCH 3/6] 8xx: invalidate non present TLBs

[Benjamin Herrenschmidt <benh@kernel.crashing.org>]

Hoy Dan !

While you are around ... I have a question :-)

Do you happen to remember what the story is with the invalidation of
"unpopulated" (aka invalid) entries ?

IE. We create those in the 8xx TLB miss when the PTE is !present (or the
PMD is absent). Those then cause a TLB error on the next access which
allows us to process the page fault. But when/how are those invalid
entries supposed to be invalidated ?

The doco seems to hint that at least in the case of an entry with the
wrong C bit (store to an entry with C=0), the HW automatically
invalidates it before taking the TLB Error but that's all I found.

Is there a general HW policy on 8xx to invalidate TLB entries that cause
TLB errors ? Or is the kernel expected to do it most of the time ?

Cheers,
Ben.