Re: [PATCH crypto-stable] crypto: arch/lib - limit simd usage to PAGE_SIZE chunks

From: Jason A. Donenfeld
Date: Wed Apr 22 2020 - 03:38:56 EST


On Wed, Apr 22, 2020 at 1:23 AM Ard Biesheuvel <ardb@xxxxxxxxxx> wrote:
> So I think 4 KB is indeed a reasonable quantum of work here. Only
> PAGE_SIZE is not necessarily equal to 4 KB on arm64, so we should use
> SZ_4K instead.

Will wait to hear the rt guys' opinion, but sure, we can do SZ_4K
explicitly. If we go with that in the end, v2 will adjust the other
places that are already using PAGE_SIZE.

> So do we have any callers that are likely to pass more than 4 KB of
> input at a time?

Network packets can be big -- ethernet jumbo packets are 9k, for
example -- so that means this could potentially ipsec, wireguard, and
maybe wifi too. (Crypto api users might go through another layer of
indirection that splits things up smaller, maybe.)