On Thu, 27 Feb 2025 at 12:57, Arnd Bergmann <arnd@xxxxxxxx> wrote: > .. > > An unrelated issue I noticed is that CRYPTO_LIB_CHACHA20POLY1305 > depends on CRYPTO in order to pull in CRYPTO_ALGAPI, this > looks like a mistake and could be resolved by moving > crypto/scatterwalk.c into lib/crypto/ with its own symbol. > That should be a separate patch of course. > The only use of the scatterwalk API in lib/crypto is from chacha20poly1305_crypt_sg_inplace(), and it occurs on a cold path where the digest buffer is covered by a separate scatterlist entry. I'll look into whether we could just use the sg_miter API there instead, which is what the rest of the code already uses.
