On Fri, Aug 29, 2014 at 03:58:55PM +0200, Steffen Trumtrar wrote:

On Fri, Aug 29, 2014 at 09:44:42PM +0800, Herbert Xu wrote:

quoted

On Mon, Aug 18, 2014 at 03:13:39PM +0200, Steffen Trumtrar wrote:

quoted

+struct sahara_sha_reqctx {
+	unsigned int mode;
+	unsigned int digest_size;
+	unsigned int context_size;
+	u8 buf[SAHARA_MAX_SHA_BLOCK_SIZE];
+	unsigned int buf_cnt;
+	unsigned int hash_rem;
+	struct sg_table data_sg;
+	struct scatterlist buf_sg;
+	unsigned int sg_in_idx;
+	unsigned int result_idx;
+};

So where is the hash state stored?

The hash state is stored in the context_base variable in sahara_dev.
Freescale calls it context in their driver (the only source of information
I have to develop a GPL driver).
To simplify the driver I allocated the context also as dma_alloc_coherent,
same as the key_base for AES.

Hmm that's a bit of a problem because our API requires the hash
state to be stored in the request as otherwise a new request
belonging to a different hash computation will cause corruption.

Cheers,
-- 
Email: Herbert Xu [off-list ref]
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt