On Tue, Mar 11, 2025 at 10:44:31AM -0700, Eric Biggers wrote: > > I guess you think this is fixing a bug in the case where sg->offset > PAGE_SIZE? > Is that case even supported? It is supposed to be the offset into a page. Supported? Obviously not since this bug has been there since the very start :) But is it conceivable to create such a scatterlist, it certainly seems to be. If we were to create a scatterlist from a single order-n folio, then it's quite reasonable for the offset to be greater than PAGE_SIZE. > Even if so, a simpler fix (1 line) would be to use: > 'sg->length >= nbytes && sg->offset + nbytes <= PAGE_SIZE' That would just mean more use of the fallback code path. Not a big deal but I was feeling generous. Thanks, -- Email: Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx> Home Page: http://gondor.apana.org.au/~herbert/ PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt