On Thu, Mar 06, 2025 at 11:40:31AM +0800, Herbert Xu wrote: > On Wed, Mar 05, 2025 at 07:36:58PM -0800, Eric Biggers wrote: > > > > That's exactly what happens to struct skcipher_walk. This patch adds two > > redundant pointers to it. Yes it's allocated on the stack, but it's still not a > > great result. > > I forgot to mention that :) > > I marked a few places in the patch with XXXs to indicate where > the API is being abused. skcipher_walk happens to be one of them > where it's mixing the new done calls with the old map call. So > I will come back to this and fix it to use the new next call instead. > > At that point I intend to have exactly one virtual pointer each > for src/dst in skcipher_walk, probably the new one that I've just > added to scatterwalk. > > Thanks, I don't think it will be quite that simple, since the skcipher_walk code relies on the different parts being split up so that it can do things like calculate the length before it starts mapping anything. If you can make it work, we can do that. But until that additional patch is ready I don't think it makes sense to merge this one, as it leaves things half-baked with the redundant pointers. - Eric
