On Tue, 2 Jul 2024 at 19:16, Eric Biggers <ebiggers@xxxxxxxxxx> wrote:
>
> Hi Ard,
>
> On Tue, Jul 02, 2024 at 09:41:19AM +0200, Ard Biesheuvel wrote:
> > > int verity_hash(struct dm_verity *v, struct dm_verity_io *io,
> > > const u8 *data, size_t len, u8 *digest, bool may_sleep)
> > > {
> > > - struct ahash_request *req = verity_io_hash_req(v, io);
> > > int r;
> > > - struct crypto_wait wait;
> > > -
> > > - r = verity_hash_init(v, req, &wait, may_sleep);
> > > - if (unlikely(r < 0))
> > > - goto out;
> > >
> > > - r = verity_hash_update(v, req, data, len, &wait);
> > > - if (unlikely(r < 0))
> > > - goto out;
> > > + if (static_branch_unlikely(&ahash_enabled) && !v->shash_tfm) {
> >
> > Is the static key really worth the hassle? Couldn't this just be
> >
> > if (unlikely(!v->shash_tfm)) {
> >
> > so that the ahash logic is moved to the cold path? We need to
> > dereference v->shash_tfm right away in any case, and if it is never
> > NULL, the branch predictor should be able to remember that.
>
> The value of the static key is indeed marginal. I included it because of the
> precedent of dm-verity's existing use_bh_wq_enabled static key, which exists for
> a similar purpose. As long as we're going through the trouble of doing that, I
> think it makes sense to use the same pattern for ahash too. It's another rarely
> needed option that can be patched in in the very rare case that it's needed.
>
If it's an existing pattern, fair enough.
