On Monday, December 30, 2013 05:01:13 PM Herbert Xu wrote: > On Fri, Dec 27, 2013 at 01:21:36AM +0100, Marek Vasut wrote: > > > > > > - complete(data, err); > > > > + areq->base.complete = complete; > > > > + areq->base.data = data; > > > > + > > > > + complete(&areq->base, err); > > > > > > This looks completely bogus. While restoring areq isn't wrong per > > > se, calling complete with &areq->base makes no sense. The original > > > completion data is in the variable "data". > > > > Is there some documentation for this so I can understand why this is wrong, > > please? I really don't quite get it, sorry. Actually, is there some > > documentation for writing crypto API drivers at all please ? > > Well it's wrong because the completion function (req->base.complete) > is meant to take data (req->base.data) as its first argument. So > giving it a pointer to req->base makes no sense. > The crypto_completion_t typdef is defined as: typedef void (*crypto_completion_t)(struct crypto_async_request *req, int err); so I believe &areq->base is the proper first argument (which is actually just the req parameter on the ahash_op_unaligned_done function). If you are going to restore areq, you really should restore all fields that were changed - result, base.complete, base.data - and set priv to NULL. Since the ahash_request_priv structure is freed in ahash_op_unaligned_finish you'll need to save the value of priv->result in order to restore areq->result (u8 *result = priv->result; or similar). Additionally, you should probably also fix up ahash_def_finup_done2 and ahash_def_finup_done1. Thanks, Tom > Cheers, > -- To unsubscribe from this list: send the line "unsubscribe linux-crypto" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html