Re: [PATCH] crypto: marvell/cesa - Avoid empty transfer descriptor

[Herbert Xu <herbert@gondor.apana.org.au>]

On Sun, May 11, 2025 at 06:39:43PM +0200, Klaus Kudielka wrote:
>
> Here the log after modprobe, with the new printk patch:

Thanks.  I'm starting to get the feeling that the partial hash
is corrupted.

Please apply this patch on top of the printk patch to confirm this.

Cheers,
-- 
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
--
diff --git a/drivers/crypto/marvell/cesa/hash.c b/drivers/crypto/marvell/cesa/hash.c
index 6815eddc9068..230501fe843b 100644
--- a/drivers/crypto/marvell/cesa/hash.c
+++ b/drivers/crypto/marvell/cesa/hash.c
@@ -374,6 +374,12 @@ static void mv_cesa_ahash_complete(struct crypto_async_request *req)
 
 		memcpy(ahashreq->result, data, digsize);
 	} else {
+		struct {
+			u32 digest[8];
+			u64 len;
+		} state;
+
+		memcpy(state.digest, creq->state, digsize);
 		for (i = 0; i < digsize / 4; i++)
 			creq->state[i] = readl_relaxed(engine->regs +
 						       CESA_IVDIG(i));
@@ -393,6 +399,21 @@ static void mv_cesa_ahash_complete(struct crypto_async_request *req)
 				for (i = 0; i < digsize / 4; i++)
 					result[i] = cpu_to_be32(creq->state[i]);
 			}
+		} else {
+			HASH_FBREQ_ON_STACK(fbreq, ahashreq);
+
+			crypto_ahash_import_core(fbreq, &state);
+			crypto_ahash_update(fbreq);
+			crypto_ahash_export_core(fbreq, &state);
+			if (memcmp(state.digest, creq->state, digsize)) {
+				pr_err("mv_cesa_ahash_complete partial hash mismatch\n");
+				print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET,
+						16, 1,
+						state.digest, digsize, false);
+				print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET,
+						16, 1,
+						creq->state, digsize, false);
+			}
 		}
 	}