Re: [PATCH v3 2/4] crypto: talitos - fix hash on SEC1.

Le 14/06/2019 à 13:32, Horia Geanta a écrit :

On 6/13/2019 3:48 PM, Christophe Leroy wrote:

quoted

@@ -336,15 +336,18 @@ static void flush_channel(struct device *dev, int ch, int error, int reset_ch)
  	tail = priv->chan[ch].tail;
  	while (priv->chan[ch].fifo[tail].desc) {
  		__be32 hdr;
+		struct talitos_edesc *edesc;
  
  		request = &priv->chan[ch].fifo[tail];
+		edesc = container_of(request->desc, struct talitos_edesc, desc);

Not needed for all cases, should be moved to the block that uses it.

Ok.

quoted

  
  		/* descriptors with their done bits set don't get the error */
  		rmb();
  		if (!is_sec1)
  			hdr = request->desc->hdr;
  		else if (request->desc->next_desc)
-			hdr = (request->desc + 1)->hdr1;
+			hdr = ((struct talitos_desc *)
+			       (edesc->buf + edesc->dma_len))->hdr1;
  		else
  			hdr = request->desc->hdr1;
  

[snip]

quoted

@@ -2058,7 +2065,18 @@ static int ahash_process_req(struct ahash_request *areq, unsigned int nbytes)
  		sg_copy_to_buffer(areq->src, nents,
  				  ctx_buf + req_ctx->nbuf, offset);
  		req_ctx->nbuf += offset;
-		req_ctx->psrc = areq->src;
+		for (sg = areq->src; sg && offset >= sg->length;
+		     offset -= sg->length, sg = sg_next(sg))
+			;
+		if (offset) {
+			sg_init_table(req_ctx->bufsl, 2);
+			sg_set_buf(req_ctx->bufsl, sg_virt(sg) + offset,
+				   sg->length - offset);
+			sg_chain(req_ctx->bufsl, 2, sg_next(sg));
+			req_ctx->psrc = req_ctx->bufsl;

Isn't this what scatterwalk_ffwd() does?

Thanks for pointing this, I wasn't aware of that function. Looking at it 
it seems to do the same. Unfortunately, some tests fails with 'wrong 
result' when using it instead.

Comparing the results of scatterwalk_ffwd() with what I get with my open 
codying, I see the following difference:

scatterwalk_ffwd() uses sg_page(sg) + sg->offset + len

while my open codying results in virt_to_page(sg_virt(sg) + len)

When sg->offset + len is greater than PAGE_SIZE, the resulting SG entry 
is different allthough valid in both cases. I think this difference 
results in sg_copy_to_buffer() failing. I'm still investigating. Any idea ?

Christophe

Horia