Re: [RESEND RFC PATCH v1 4/5] arm64/mm: Delay tlbi in contpte_convert() under BBML2

[Will Deacon <will@kernel.org>]

On Wed, Dec 11, 2024 at 04:01:40PM +0000, Mikołaj Lenczewski wrote:
> When converting a region via contpte_convert() to use mTHP, we have two
> different goals. We have to mark each entry as contiguous, and we would
> like to smear the dirty and young (access) bits across all entries in
> the contiguous block. Currently, we do this by first accumulating the
> dirty and young bits in the block, using an atomic
> __ptep_get_and_clear() and the relevant pte_{dirty,young}() calls,
> performing a tlbi, and finally smearing the correct bits across the
> block using __set_ptes().
> 
> This approach works fine for BBM level 0, but with support for BBM level
> 2 we are allowed to reorder the tlbi to after setting the pagetable
> entries. This reordering means that other threads will not see an
> invalid pagetable entry, instead operating on stale data, until we have
> performed our smearing and issued the invalidation. Avoiding this
> invalid entry reduces faults in other threads, and thus improves
> performance marginally (more so when there are more threads).

Please provide the performance data.

Will