On Oct 03, 2023 / 11:43, Herbert Xu wrote: > The lskcipher conversion broke essiv. The superficial reason is > that it was missed during the lskcipher addition as it does something > similar to cryptd by being capable of both creating skciphers as well > as aead algorithms. > > However, this also revealed that users of skcipher spawns were > dereferencing fields that may be specific to skcipher_alg. This > is invalid as the underlying algorithm may be an lskcipher_alg. > Only fields in skcipher_alg_common may be used. > > This series adds the necessary helper for this and converts all > users of skcipher spawns to use the new helper. Thanks Herbert. I applied this series on top of the kernel next-20230929 and confirmed the failure I reported [1] goes away. Also I ran my test sets on some dm-crypt devices and saw no regression. Looks good from testing point of view. [1] https://lore.kernel.org/dm-devel/4u6ffcrzr5xg6tzoczkfnuqy7v2e3w6243oxdsu3g4uughh6go@6owks5linnxi/
