On 7/2/2020 7:36 AM, Herbert Xu wrote: > The arc4 algorithm requires storing state in the request context > in order to allow more than one encrypt/decrypt operation. As this > driver does not seem to do that, it means that using it for more > than one operation is broken. > The fact that smth. is broken doesn't necessarily means it has to be removed. Looking at the HW capabilities, I am sure the implementation could be modified to save/restore the internal state to/from the request context. Anyhow I would like to know if only the correctness is being debated, or this patch should be dealt with in the larger context of removing crypto API based ecb(arc4) altogether: [RFC PATCH 0/7] crypto: get rid of ecb(arc4) https://lore.kernel.org/linux-crypto/20200702101947.682-1-ardb@xxxxxxxxxx/ Thanks, Horia
