[PATCH 1/3 V2] crypto: Fix the pointer voodoo in unaligned ahash

On Thursday, March 13, 2014 at 02:56:25 AM, Herbert Xu wrote:

On Thu, Mar 13, 2014 at 02:20:29AM +0100, Marek Vasut wrote:

quoted

OK, understood. But shall we not preserve the request intact in case a
crypto- api function called crypto_ahash_final() with request which has
.priv already set? Then we would have really funny corruption of the
request going on and I'm not sure that'd be nice.

The priv field is only ever used by ahash.c so how can this happen?
The crypto API refers to code in the API itself, excluding drivers
and users.

OK, I agree with you that people plumbing in the API itself will know what 
they're doing.

btw. can you please check the V3 of 3/3 for the fixup of the base.completion() 
call ? I will then do tests and roll V4 of the series.

Best regards,
Marek Vasut