1 files changed, 313 insertions, 116 deletions

diff --git a/drivers/crypto/caam/caamhash.c b/drivers/crypto/caam/caamhash.c
index bb1a2cdf1951..d7483e4d0ce2 100644
--- a/drivers/crypto/caam/caamhash.c
+++ b/drivers/crypto/caam/caamhash.c
@@ -3,7 +3,7 @@
  * caam - Freescale FSL CAAM support for ahash functions of crypto API
  *
  * Copyright 2011 Freescale Semiconductor, Inc.
- * Copyright 2018 NXP
+ * Copyright 2018-2019 NXP
  *
  * Based on caamalg.c crypto API driver.
  *
@@ -98,13 +98,14 @@ struct caam_hash_ctx {
 	u32 sh_desc_update_first[DESC_HASH_MAX_USED_LEN] ____cacheline_aligned;
 	u32 sh_desc_fin[DESC_HASH_MAX_USED_LEN] ____cacheline_aligned;
 	u32 sh_desc_digest[DESC_HASH_MAX_USED_LEN] ____cacheline_aligned;
+	u8 key[CAAM_MAX_HASH_KEY_SIZE] ____cacheline_aligned;
 	dma_addr_t sh_desc_update_dma ____cacheline_aligned;
 	dma_addr_t sh_desc_update_first_dma;
 	dma_addr_t sh_desc_fin_dma;
 	dma_addr_t sh_desc_digest_dma;
+	dma_addr_t key_dma;
 	enum dma_data_direction dir;
 	struct device *jrdev;
-	u8 key[CAAM_MAX_HASH_KEY_SIZE];
 	int ctx_len;
 	struct alginfo adata;
 };
@@ -113,6 +114,7 @@ struct caam_hash_ctx {
 struct caam_hash_state {
 	dma_addr_t buf_dma;
 	dma_addr_t ctx_dma;
+	int ctx_dma_len;
 	u8 buf_0[CAAM_MAX_HASH_BLOCK_SIZE] ____cacheline_aligned;
 	int buflen_0;
 	u8 buf_1[CAAM_MAX_HASH_BLOCK_SIZE] ____cacheline_aligned;
@@ -158,6 +160,11 @@ static inline int *alt_buflen(struct caam_hash_state *state)
 	return state->current_buf ? &state->buflen_0 : &state->buflen_1;
 }
 
+static inline bool is_cmac_aes(u32 algtype)
+{
+	return (algtype & (OP_ALG_ALGSEL_MASK | OP_ALG_AAI_MASK)) ==
+	       (OP_ALG_ALGSEL_AES | OP_ALG_AAI_CMAC);
+}
 /* Common job descriptor seq in/out ptr routines */
 
 /* Map state->caam_ctx, and append seq_out_ptr command that points to it */
@@ -165,6 +172,7 @@ static inline int map_seq_out_ptr_ctx(u32 *desc, struct device *jrdev,
 				      struct caam_hash_state *state,
 				      int ctx_len)
 {
+	state->ctx_dma_len = ctx_len;
 	state->ctx_dma = dma_map_single(jrdev, state->caam_ctx,
 					ctx_len, DMA_FROM_DEVICE);
 	if (dma_mapping_error(jrdev, state->ctx_dma)) {
@@ -178,18 +186,6 @@ static inline int map_seq_out_ptr_ctx(u32 *desc, struct device *jrdev,
 	return 0;
 }
 
-/* Map req->result, and append seq_out_ptr command that points to it */
-static inline dma_addr_t map_seq_out_ptr_result(u32 *desc, struct device *jrdev,
-						u8 *result, int digestsize)
-{
-	dma_addr_t dst_dma;
-
-	dst_dma = dma_map_single(jrdev, result, digestsize, DMA_FROM_DEVICE);
-	append_seq_out_ptr(desc, dst_dma, digestsize, 0);
-
-	return dst_dma;
-}
-
 /* Map current buffer in state (if length > 0) and put it in link table */
 static inline int buf_map_to_sec4_sg(struct device *jrdev,
 				     struct sec4_sg_entry *sec4_sg,
@@ -218,6 +214,7 @@ static inline int ctx_map_to_sec4_sg(struct device *jrdev,
 				     struct caam_hash_state *state, int ctx_len,
 				     struct sec4_sg_entry *sec4_sg, u32 flag)
 {
+	state->ctx_dma_len = ctx_len;
 	state->ctx_dma = dma_map_single(jrdev, state->caam_ctx, ctx_len, flag);
 	if (dma_mapping_error(jrdev, state->ctx_dma)) {
 		dev_err(jrdev, "unable to map ctx\n");
@@ -292,14 +289,119 @@ static int ahash_set_sh_desc(struct crypto_ahash *ahash)
 	return 0;
 }
 
+static int axcbc_set_sh_desc(struct crypto_ahash *ahash)
+{
+	struct caam_hash_ctx *ctx = crypto_ahash_ctx(ahash);
+	int digestsize = crypto_ahash_digestsize(ahash);
+	struct device *jrdev = ctx->jrdev;
+	u32 *desc;
+
+	/* key is loaded from memory for UPDATE and FINALIZE states */
+	ctx->adata.key_dma = ctx->key_dma;
+
+	/* shared descriptor for ahash_update */
+	desc = ctx->sh_desc_update;
+	cnstr_shdsc_sk_hash(desc, &ctx->adata, OP_ALG_AS_UPDATE,
+			    ctx->ctx_len, ctx->ctx_len, 0);
+	dma_sync_single_for_device(jrdev, ctx->sh_desc_update_dma,
+				   desc_bytes(desc), ctx->dir);
+	print_hex_dump_debug("axcbc update shdesc@" __stringify(__LINE__)" : ",
+			     DUMP_PREFIX_ADDRESS, 16, 4, desc, desc_bytes(desc),
+			     1);
+
+	/* shared descriptor for ahash_{final,finup} */
+	desc = ctx->sh_desc_fin;
+	cnstr_shdsc_sk_hash(desc, &ctx->adata, OP_ALG_AS_FINALIZE,
+			    digestsize, ctx->ctx_len, 0);
+	dma_sync_single_for_device(jrdev, ctx->sh_desc_fin_dma,
+				   desc_bytes(desc), ctx->dir);
+	print_hex_dump_debug("axcbc finup shdesc@" __stringify(__LINE__)" : ",
+			     DUMP_PREFIX_ADDRESS, 16, 4, desc, desc_bytes(desc),
+			     1);
+
+	/* key is immediate data for INIT and INITFINAL states */
+	ctx->adata.key_virt = ctx->key;
+
+	/* shared descriptor for first invocation of ahash_update */
+	desc = ctx->sh_desc_update_first;
+	cnstr_shdsc_sk_hash(desc, &ctx->adata, OP_ALG_AS_INIT, ctx->ctx_len,
+			    ctx->ctx_len, ctx->key_dma);
+	dma_sync_single_for_device(jrdev, ctx->sh_desc_update_first_dma,
+				   desc_bytes(desc), ctx->dir);
+	print_hex_dump_debug("axcbc update first shdesc@" __stringify(__LINE__)" : ",
+			     DUMP_PREFIX_ADDRESS, 16, 4, desc, desc_bytes(desc),
+			     1);
+
+	/* shared descriptor for ahash_digest */
+	desc = ctx->sh_desc_digest;
+	cnstr_shdsc_sk_hash(desc, &ctx->adata, OP_ALG_AS_INITFINAL,
+			    digestsize, ctx->ctx_len, 0);
+	dma_sync_single_for_device(jrdev, ctx->sh_desc_digest_dma,
+				   desc_bytes(desc), ctx->dir);
+	print_hex_dump_debug("axcbc digest shdesc@" __stringify(__LINE__)" : ",
+			     DUMP_PREFIX_ADDRESS, 16, 4, desc, desc_bytes(desc),
+			     1);
+	return 0;
+}
+
+static int acmac_set_sh_desc(struct crypto_ahash *ahash)
+{
+	struct caam_hash_ctx *ctx = crypto_ahash_ctx(ahash);
+	int digestsize = crypto_ahash_digestsize(ahash);
+	struct device *jrdev = ctx->jrdev;
+	u32 *desc;
+
+	/* shared descriptor for ahash_update */
+	desc = ctx->sh_desc_update;
+	cnstr_shdsc_sk_hash(desc, &ctx->adata, OP_ALG_AS_UPDATE,
+			    ctx->ctx_len, ctx->ctx_len, 0);
+	dma_sync_single_for_device(jrdev, ctx->sh_desc_update_dma,
+				   desc_bytes(desc), ctx->dir);
+	print_hex_dump_debug("acmac update shdesc@" __stringify(__LINE__)" : ",
+			     DUMP_PREFIX_ADDRESS, 16, 4, desc,
+			     desc_bytes(desc), 1);
+
+	/* shared descriptor for ahash_{final,finup} */
+	desc = ctx->sh_desc_fin;
+	cnstr_shdsc_sk_hash(desc, &ctx->adata, OP_ALG_AS_FINALIZE,
+			    digestsize, ctx->ctx_len, 0);
+	dma_sync_single_for_device(jrdev, ctx->sh_desc_fin_dma,
+				   desc_bytes(desc), ctx->dir);
+	print_hex_dump_debug("acmac finup shdesc@" __stringify(__LINE__)" : ",
+			     DUMP_PREFIX_ADDRESS, 16, 4, desc,
+			     desc_bytes(desc), 1);
+
+	/* shared descriptor for first invocation of ahash_update */
+	desc = ctx->sh_desc_update_first;
+	cnstr_shdsc_sk_hash(desc, &ctx->adata, OP_ALG_AS_INIT, ctx->ctx_len,
+			    ctx->ctx_len, 0);
+	dma_sync_single_for_device(jrdev, ctx->sh_desc_update_first_dma,
+				   desc_bytes(desc), ctx->dir);
+	print_hex_dump_debug("acmac update first shdesc@" __stringify(__LINE__)" : ",
+			     DUMP_PREFIX_ADDRESS, 16, 4, desc,
+			     desc_bytes(desc), 1);
+
+	/* shared descriptor for ahash_digest */
+	desc = ctx->sh_desc_digest;
+	cnstr_shdsc_sk_hash(desc, &ctx->adata, OP_ALG_AS_INITFINAL,
+			    digestsize, ctx->ctx_len, 0);
+	dma_sync_single_for_device(jrdev, ctx->sh_desc_digest_dma,
+				   desc_bytes(desc), ctx->dir);
+	print_hex_dump_debug("acmac digest shdesc@" __stringify(__LINE__)" : ",
+			     DUMP_PREFIX_ADDRESS, 16, 4, desc,
+			     desc_bytes(desc), 1);
+
+	return 0;
+}
+
 /* Digest hash size if it is too large */
-static int hash_digest_key(struct caam_hash_ctx *ctx, const u8 *key_in,
-			   u32 *keylen, u8 *key_out, u32 digestsize)
+static int hash_digest_key(struct caam_hash_ctx *ctx, u32 *keylen, u8 *key,
+			   u32 digestsize)
 {
 	struct device *jrdev = ctx->jrdev;
 	u32 *desc;
 	struct split_key_result result;
-	dma_addr_t src_dma, dst_dma;
+	dma_addr_t key_dma;
 	int ret;
 
 	desc = kmalloc(CAAM_CMD_SZ * 8 + CAAM_PTR_SZ * 2, GFP_KERNEL | GFP_DMA);
@@ -310,18 +412,9 @@ static int hash_digest_key(struct caam_hash_ctx *ctx, const u8 *key_in,
 
 	init_job_desc(desc, 0);
 
-	src_dma = dma_map_single(jrdev, (void *)key_in, *keylen,
-				 DMA_TO_DEVICE);
-	if (dma_mapping_error(jrdev, src_dma)) {
-		dev_err(jrdev, "unable to map key input memory\n");
-		kfree(desc);
-		return -ENOMEM;
-	}
-	dst_dma = dma_map_single(jrdev, (void *)key_out, digestsize,
-				 DMA_FROM_DEVICE);
-	if (dma_mapping_error(jrdev, dst_dma)) {
-		dev_err(jrdev, "unable to map key output memory\n");
-		dma_unmap_single(jrdev, src_dma, *keylen, DMA_TO_DEVICE);
+	key_dma = dma_map_single(jrdev, key, *keylen, DMA_BIDIRECTIONAL);
+	if (dma_mapping_error(jrdev, key_dma)) {
+		dev_err(jrdev, "unable to map key memory\n");
 		kfree(desc);
 		return -ENOMEM;
 	}
@@ -329,16 +422,16 @@ static int hash_digest_key(struct caam_hash_ctx *ctx, const u8 *key_in,
 	/* Job descriptor to perform unkeyed hash on key_in */
 	append_operation(desc, ctx->adata.algtype | OP_ALG_ENCRYPT |
 			 OP_ALG_AS_INITFINAL);
-	append_seq_in_ptr(desc, src_dma, *keylen, 0);
+	append_seq_in_ptr(desc, key_dma, *keylen, 0);
 	append_seq_fifo_load(desc, *keylen, FIFOLD_CLASS_CLASS2 |
 			     FIFOLD_TYPE_LAST2 | FIFOLD_TYPE_MSG);
-	append_seq_out_ptr(desc, dst_dma, digestsize, 0);
+	append_seq_out_ptr(desc, key_dma, digestsize, 0);
 	append_seq_store(desc, digestsize, LDST_CLASS_2_CCB |
 			 LDST_SRCDST_BYTE_CONTEXT);
 
 #ifdef DEBUG
 	print_hex_dump(KERN_ERR, "key_in@"__stringify(__LINE__)": ",
-		       DUMP_PREFIX_ADDRESS, 16, 4, key_in, *keylen, 1);
+		       DUMP_PREFIX_ADDRESS, 16, 4, key, *keylen, 1);
 	print_hex_dump(KERN_ERR, "jobdesc@"__stringify(__LINE__)": ",
 		       DUMP_PREFIX_ADDRESS, 16, 4, desc, desc_bytes(desc), 1);
 #endif
@@ -354,12 +447,10 @@ static int hash_digest_key(struct caam_hash_ctx *ctx, const u8 *key_in,
 #ifdef DEBUG
 		print_hex_dump(KERN_ERR,
 			       "digested key@"__stringify(__LINE__)": ",
-			       DUMP_PREFIX_ADDRESS, 16, 4, key_in,
-			       digestsize, 1);
+			       DUMP_PREFIX_ADDRESS, 16, 4, key, digestsize, 1);
 #endif
 	}
-	dma_unmap_single(jrdev, src_dma, *keylen, DMA_TO_DEVICE);
-	dma_unmap_single(jrdev, dst_dma, digestsize, DMA_FROM_DEVICE);
+	dma_unmap_single(jrdev, key_dma, *keylen, DMA_BIDIRECTIONAL);
 
 	*keylen = digestsize;
 
@@ -383,13 +474,10 @@ static int ahash_setkey(struct crypto_ahash *ahash,
 #endif
 
 	if (keylen > blocksize) {
-		hashed_key = kmalloc_array(digestsize,
-					   sizeof(*hashed_key),
-					   GFP_KERNEL | GFP_DMA);
+		hashed_key = kmemdup(key, keylen, GFP_KERNEL | GFP_DMA);
 		if (!hashed_key)
 			return -ENOMEM;
-		ret = hash_digest_key(ctx, key, &keylen, hashed_key,
-				      digestsize);
+		ret = hash_digest_key(ctx, &keylen, hashed_key, digestsize);
 		if (ret)
 			goto bad_free_key;
 		key = hashed_key;
@@ -424,9 +512,39 @@ static int ahash_setkey(struct crypto_ahash *ahash,
 	return -EINVAL;
 }
 
+static int axcbc_setkey(struct crypto_ahash *ahash, const u8 *key,
+			unsigned int keylen)
+{
+	struct caam_hash_ctx *ctx = crypto_ahash_ctx(ahash);
+	struct device *jrdev = ctx->jrdev;
+
+	memcpy(ctx->key, key, keylen);
+	dma_sync_single_for_device(jrdev, ctx->key_dma, keylen, DMA_TO_DEVICE);
+	ctx->adata.keylen = keylen;
+
+	print_hex_dump_debug("axcbc ctx.key@" __stringify(__LINE__)" : ",
+			     DUMP_PREFIX_ADDRESS, 16, 4, ctx->key, keylen, 1);
+
+	return axcbc_set_sh_desc(ahash);
+}
+
+static int acmac_setkey(struct crypto_ahash *ahash, const u8 *key,
+			unsigned int keylen)
+{
+	struct caam_hash_ctx *ctx = crypto_ahash_ctx(ahash);
+
+	/* key is immediate data for all cmac shared descriptors */
+	ctx->adata.key_virt = key;
+	ctx->adata.keylen = keylen;
+
+	print_hex_dump_debug("acmac ctx.key@" __stringify(__LINE__)" : ",
+			     DUMP_PREFIX_ADDRESS, 16, 4, key, keylen, 1);
+
+	return acmac_set_sh_desc(ahash);
+}
+
 /*
  * ahash_edesc - s/w-extended ahash descriptor
- * @dst_dma: physical mapped address of req->result
  * @sec4_sg_dma: physical mapped address of h/w link table
  * @src_nents: number of segments in input scatterlist
  * @sec4_sg_bytes: length of dma mapped sec4_sg space
@@ -434,7 +552,6 @@ static int ahash_setkey(struct crypto_ahash *ahash,
  * @sec4_sg: h/w link table
  */
 struct ahash_edesc {
-	dma_addr_t dst_dma;
 	dma_addr_t sec4_sg_dma;
 	int src_nents;
 	int sec4_sg_bytes;
@@ -450,8 +567,6 @@ static inline void ahash_unmap(struct device *dev,
 
 	if (edesc->src_nents)
 		dma_unmap_sg(dev, req->src, edesc->src_nents, DMA_TO_DEVICE);
-	if (edesc->dst_dma)
-		dma_unmap_single(dev, edesc->dst_dma, dst_len, DMA_FROM_DEVICE);
 
 	if (edesc->sec4_sg_bytes)
 		dma_unmap_single(dev, edesc->sec4_sg_dma,
@@ -468,12 +583,10 @@ static inline void ahash_unmap_ctx(struct device *dev,
 			struct ahash_edesc *edesc,
 			struct ahash_request *req, int dst_len, u32 flag)
 {
-	struct crypto_ahash *ahash = crypto_ahash_reqtfm(req);
-	struct caam_hash_ctx *ctx = crypto_ahash_ctx(ahash);
 	struct caam_hash_state *state = ahash_request_ctx(req);
 
 	if (state->ctx_dma) {
-		dma_unmap_single(dev, state->ctx_dma, ctx->ctx_len, flag);
+		dma_unmap_single(dev, state->ctx_dma, state->ctx_dma_len, flag);
 		state->ctx_dma = 0;
 	}
 	ahash_unmap(dev, edesc, req, dst_len);
@@ -486,9 +599,9 @@ static void ahash_done(struct device *jrdev, u32 *desc, u32 err,
 	struct ahash_edesc *edesc;
 	struct crypto_ahash *ahash = crypto_ahash_reqtfm(req);
 	int digestsize = crypto_ahash_digestsize(ahash);
+	struct caam_hash_state *state = ahash_request_ctx(req);
 #ifdef DEBUG
 	struct caam_hash_ctx *ctx = crypto_ahash_ctx(ahash);
-	struct caam_hash_state *state = ahash_request_ctx(req);
 
 	dev_err(jrdev, "%s %d: err 0x%x\n", __func__, __LINE__, err);
 #endif
@@ -497,17 +610,14 @@ static void ahash_done(struct device *jrdev, u32 *desc, u32 err,
 	if (err)
 		caam_jr_strstatus(jrdev, err);
 
-	ahash_unmap(jrdev, edesc, req, digestsize);
+	ahash_unmap_ctx(jrdev, edesc, req, digestsize, DMA_FROM_DEVICE);
+	memcpy(req->result, state->caam_ctx, digestsize);
 	kfree(edesc);
 
 #ifdef DEBUG
 	print_hex_dump(KERN_ERR, "ctx@"__stringify(__LINE__)": ",
 		       DUMP_PREFIX_ADDRESS, 16, 4, state->caam_ctx,
 		       ctx->ctx_len, 1);
-	if (req->result)
-		print_hex_dump(KERN_ERR, "result@"__stringify(__LINE__)": ",
-			       DUMP_PREFIX_ADDRESS, 16, 4, req->result,
-			       digestsize, 1);
 #endif
 
 	req->base.complete(&req->base, err);
@@ -555,9 +665,9 @@ static void ahash_done_ctx_src(struct device *jrdev, u32 *desc, u32 err,
 	struct ahash_edesc *edesc;
 	struct crypto_ahash *ahash = crypto_ahash_reqtfm(req);
 	int digestsize = crypto_ahash_digestsize(ahash);
+	struct caam_hash_state *state = ahash_request_ctx(req);
 #ifdef DEBUG
 	struct caam_hash_ctx *ctx = crypto_ahash_ctx(ahash);
-	struct caam_hash_state *state = ahash_request_ctx(req);
 
 	dev_err(jrdev, "%s %d: err 0x%x\n", __func__, __LINE__, err);
 #endif
@@ -566,17 +676,14 @@ static void ahash_done_ctx_src(struct device *jrdev, u32 *desc, u32 err,
 	if (err)
 		caam_jr_strstatus(jrdev, err);
 
-	ahash_unmap_ctx(jrdev, edesc, req, digestsize, DMA_TO_DEVICE);
+	ahash_unmap_ctx(jrdev, edesc, req, digestsize, DMA_BIDIRECTIONAL);
+	memcpy(req->result, state->caam_ctx, digestsize);
 	kfree(edesc);
 
 #ifdef DEBUG
 	print_hex_dump(KERN_ERR, "ctx@"__stringify(__LINE__)": ",
 		       DUMP_PREFIX_ADDRESS, 16, 4, state->caam_ctx,
 		       ctx->ctx_len, 1);
-	if (req->result)
-		print_hex_dump(KERN_ERR, "result@"__stringify(__LINE__)": ",
-			       DUMP_PREFIX_ADDRESS, 16, 4, req->result,
-			       digestsize, 1);
 #endif
 
 	req->base.complete(&req->base, err);
@@ -688,6 +795,7 @@ static int ahash_update_ctx(struct ahash_request *req)
 	u8 *buf = current_buf(state);
 	int *buflen = current_buflen(state);
 	u8 *next_buf = alt_buf(state);
+	int blocksize = crypto_ahash_blocksize(ahash);
 	int *next_buflen = alt_buflen(state), last_buflen;
 	int in_len = *buflen + req->nbytes, to_hash;
 	u32 *desc;
@@ -696,9 +804,20 @@ static int ahash_update_ctx(struct ahash_request *req)
 	int ret = 0;
 
 	last_buflen = *next_buflen;
-	*next_buflen = in_len & (crypto_tfm_alg_blocksize(&ahash->base) - 1);
+	*next_buflen = in_len & (blocksize - 1);
 	to_hash = in_len - *next_buflen;
 
+	/*
+	 * For XCBC and CMAC, if to_hash is multiple of block size,
+	 * keep last block in internal buffer
+	 */
+	if ((is_xcbc_aes(ctx->adata.algtype) ||
+	     is_cmac_aes(ctx->adata.algtype)) && to_hash >= blocksize &&
+	     (*next_buflen == 0)) {
+		*next_buflen = blocksize;
+		to_hash -= blocksize;
+	}
+
 	if (to_hash) {
 		src_nents = sg_nents_for_len(req->src,
 					     req->nbytes - (*next_buflen));
@@ -801,7 +920,7 @@ static int ahash_update_ctx(struct ahash_request *req)
 #endif
 
 	return ret;
- unmap_ctx:
+unmap_ctx:
 	ahash_unmap_ctx(jrdev, edesc, req, ctx->ctx_len, DMA_BIDIRECTIONAL);
 	kfree(edesc);
 	return ret;
@@ -837,7 +956,7 @@ static int ahash_final_ctx(struct ahash_request *req)
 	edesc->sec4_sg_bytes = sec4_sg_bytes;
 
 	ret = ctx_map_to_sec4_sg(jrdev, state, ctx->ctx_len,
-				 edesc->sec4_sg, DMA_TO_DEVICE);
+				 edesc->sec4_sg, DMA_BIDIRECTIONAL);
 	if (ret)
 		goto unmap_ctx;
 
@@ -857,14 +976,7 @@ static int ahash_final_ctx(struct ahash_request *req)
 
 	append_seq_in_ptr(desc, edesc->sec4_sg_dma, ctx->ctx_len + buflen,
 			  LDST_SGF);
-
-	edesc->dst_dma = map_seq_out_ptr_result(desc, jrdev, req->result,
-						digestsize);
-	if (dma_mapping_error(jrdev, edesc->dst_dma)) {
-		dev_err(jrdev, "unable to map dst\n");
-		ret = -ENOMEM;
-		goto unmap_ctx;
-	}
+	append_seq_out_ptr(desc, state->ctx_dma, digestsize, 0);
 
 #ifdef DEBUG
 	print_hex_dump(KERN_ERR, "jobdesc@"__stringify(__LINE__)": ",
@@ -877,7 +989,7 @@ static int ahash_final_ctx(struct ahash_request *req)
 
 	return -EINPROGRESS;
  unmap_ctx:
-	ahash_unmap_ctx(jrdev, edesc, req, digestsize, DMA_FROM_DEVICE);
+	ahash_unmap_ctx(jrdev, edesc, req, digestsize, DMA_BIDIRECTIONAL);
 	kfree(edesc);
 	return ret;
 }
@@ -931,7 +1043,7 @@ static int ahash_finup_ctx(struct ahash_request *req)
 	edesc->src_nents = src_nents;
 
 	ret = ctx_map_to_sec4_sg(jrdev, state, ctx->ctx_len,
-				 edesc->sec4_sg, DMA_TO_DEVICE);
+				 edesc->sec4_sg, DMA_BIDIRECTIONAL);
 	if (ret)
 		goto unmap_ctx;
 
@@ -945,13 +1057,7 @@ static int ahash_finup_ctx(struct ahash_request *req)
 	if (ret)
 		goto unmap_ctx;
 
-	edesc->dst_dma = map_seq_out_ptr_result(desc, jrdev, req->result,
-						digestsize);
-	if (dma_mapping_error(jrdev, edesc->dst_dma)) {
-		dev_err(jrdev, "unable to map dst\n");
-		ret = -ENOMEM;
-		goto unmap_ctx;
-	}
+	append_seq_out_ptr(desc, state->ctx_dma, digestsize, 0);
 
 #ifdef DEBUG
 	print_hex_dump(KERN_ERR, "jobdesc@"__stringify(__LINE__)": ",
@@ -964,7 +1070,7 @@ static int ahash_finup_ctx(struct ahash_request *req)
 
 	return -EINPROGRESS;
  unmap_ctx:
-	ahash_unmap_ctx(jrdev, edesc, req, digestsize, DMA_FROM_DEVICE);
+	ahash_unmap_ctx(jrdev, edesc, req, digestsize, DMA_BIDIRECTIONAL);
 	kfree(edesc);
 	return ret;
 }
@@ -1023,10 +1129,8 @@ static int ahash_digest(struct ahash_request *req)
 
 	desc = edesc->hw_desc;
 
-	edesc->dst_dma = map_seq_out_ptr_result(desc, jrdev, req->result,
-						digestsize);
-	if (dma_mapping_error(jrdev, edesc->dst_dma)) {
-		dev_err(jrdev, "unable to map dst\n");
+	ret = map_seq_out_ptr_ctx(desc, jrdev, state, digestsize);
+	if (ret) {
 		ahash_unmap(jrdev, edesc, req, digestsize);
 		kfree(edesc);
 		return -ENOMEM;
@@ -1041,7 +1145,7 @@ static int ahash_digest(struct ahash_request *req)
 	if (!ret) {
 		ret = -EINPROGRESS;
 	} else {
-		ahash_unmap(jrdev, edesc, req, digestsize);
+		ahash_unmap_ctx(jrdev, edesc, req, digestsize, DMA_FROM_DEVICE);
 		kfree(edesc);
 	}
 
@@ -1083,12 +1187,9 @@ static int ahash_final_no_ctx(struct ahash_request *req)
 		append_seq_in_ptr(desc, state->buf_dma, buflen, 0);
 	}
 
-	edesc->dst_dma = map_seq_out_ptr_result(desc, jrdev, req->result,
-						digestsize);
-	if (dma_mapping_error(jrdev, edesc->dst_dma)) {
-		dev_err(jrdev, "unable to map dst\n");
+	ret = map_seq_out_ptr_ctx(desc, jrdev, state, digestsize);
+	if (ret)
 		goto unmap;
-	}
 
 #ifdef DEBUG
 	print_hex_dump(KERN_ERR, "jobdesc@"__stringify(__LINE__)": ",
@@ -1099,7 +1200,7 @@ static int ahash_final_no_ctx(struct ahash_request *req)
 	if (!ret) {
 		ret = -EINPROGRESS;
 	} else {
-		ahash_unmap(jrdev, edesc, req, digestsize);
+		ahash_unmap_ctx(jrdev, edesc, req, digestsize, DMA_FROM_DEVICE);
 		kfree(edesc);
 	}
 
@@ -1122,6 +1223,7 @@ static int ahash_update_no_ctx(struct ahash_request *req)
 		       GFP_KERNEL : GFP_ATOMIC;
 	u8 *buf = current_buf(state);
 	int *buflen = current_buflen(state);
+	int blocksize = crypto_ahash_blocksize(ahash);
 	u8 *next_buf = alt_buf(state);
 	int *next_buflen = alt_buflen(state);
 	int in_len = *buflen + req->nbytes, to_hash;
@@ -1130,9 +1232,20 @@ static int ahash_update_no_ctx(struct ahash_request *req)
 	u32 *desc;
 	int ret = 0;
 
-	*next_buflen = in_len & (crypto_tfm_alg_blocksize(&ahash->base) - 1);
+	*next_buflen = in_len & (blocksize - 1);
 	to_hash = in_len - *next_buflen;
 
+	/*
+	 * For XCBC and CMAC, if to_hash is multiple of block size,
+	 * keep last block in internal buffer
+	 */
+	if ((is_xcbc_aes(ctx->adata.algtype) ||
+	     is_cmac_aes(ctx->adata.algtype)) && to_hash >= blocksize &&
+	     (*next_buflen == 0)) {
+		*next_buflen = blocksize;
+		to_hash -= blocksize;
+	}
+
 	if (to_hash) {
 		src_nents = sg_nents_for_len(req->src,
 					     req->nbytes - *next_buflen);
@@ -1298,12 +1411,9 @@ static int ahash_finup_no_ctx(struct ahash_request *req)
 		goto unmap;
 	}
 
-	edesc->dst_dma = map_seq_out_ptr_result(desc, jrdev, req->result,
-						digestsize);
-	if (dma_mapping_error(jrdev, edesc->dst_dma)) {
-		dev_err(jrdev, "unable to map dst\n");
+	ret = map_seq_out_ptr_ctx(desc, jrdev, state, digestsize);
+	if (ret)
 		goto unmap;
-	}
 
 #ifdef DEBUG
 	print_hex_dump(KERN_ERR, "jobdesc@"__stringify(__LINE__)": ",
@@ -1314,7 +1424,7 @@ static int ahash_finup_no_ctx(struct ahash_request *req)
 	if (!ret) {
 		ret = -EINPROGRESS;
 	} else {
-		ahash_unmap(jrdev, edesc, req, digestsize);
+		ahash_unmap_ctx(jrdev, edesc, req, digestsize, DMA_FROM_DEVICE);
 		kfree(edesc);
 	}
 
@@ -1338,15 +1448,26 @@ static int ahash_update_first(struct ahash_request *req)
 	u8 *next_buf = alt_buf(state);
 	int *next_buflen = alt_buflen(state);
 	int to_hash;
+	int blocksize = crypto_ahash_blocksize(ahash);
 	u32 *desc;
 	int src_nents, mapped_nents;
 	struct ahash_edesc *edesc;
 	int ret = 0;
 
-	*next_buflen = req->nbytes & (crypto_tfm_alg_blocksize(&ahash->base) -
-				      1);
+	*next_buflen = req->nbytes & (blocksize - 1);
 	to_hash = req->nbytes - *next_buflen;
 
+	/*
+	 * For XCBC and CMAC, if to_hash is multiple of block size,
+	 * keep last block in internal buffer
+	 */
+	if ((is_xcbc_aes(ctx->adata.algtype) ||
+	     is_cmac_aes(ctx->adata.algtype)) && to_hash >= blocksize &&
+	     (*next_buflen == 0)) {
+		*next_buflen = blocksize;
+		to_hash -= blocksize;
+	}
+
 	if (to_hash) {
 		src_nents = sg_nents_for_len(req->src,
 					     req->nbytes - *next_buflen);
@@ -1446,6 +1567,7 @@ static int ahash_init(struct ahash_request *req)
 	state->final = ahash_final_no_ctx;
 
 	state->ctx_dma = 0;
+	state->ctx_dma_len = 0;
 	state->current_buf = 0;
 	state->buf_dma = 0;
 	state->buflen_0 = 0;
@@ -1654,6 +1776,44 @@ static struct caam_hash_template driver_hash[] = {
 			},
 		},
 		.alg_type = OP_ALG_ALGSEL_MD5,
+	}, {
+		.hmac_name = "xcbc(aes)",
+		.hmac_driver_name = "xcbc-aes-caam",
+		.blocksize = AES_BLOCK_SIZE,
+		.template_ahash = {
+			.init = ahash_init,
+			.update = ahash_update,
+			.final = ahash_final,
+			.finup = ahash_finup,
+			.digest = ahash_digest,
+			.export = ahash_export,
+			.import = ahash_import,
+			.setkey = axcbc_setkey,
+			.halg = {
+				.digestsize = AES_BLOCK_SIZE,
+				.statesize = sizeof(struct caam_export_state),
+			},
+		 },
+		.alg_type = OP_ALG_ALGSEL_AES | OP_ALG_AAI_XCBC_MAC,
+	}, {
+		.hmac_name = "cmac(aes)",
+		.hmac_driver_name = "cmac-aes-caam",
+		.blocksize = AES_BLOCK_SIZE,
+		.template_ahash = {
+			.init = ahash_init,
+			.update = ahash_update,
+			.final = ahash_final,
+			.finup = ahash_finup,
+			.digest = ahash_digest,
+			.export = ahash_export,
+			.import = ahash_import,
+			.setkey = acmac_setkey,
+			.halg = {
+				.digestsize = AES_BLOCK_SIZE,
+				.statesize = sizeof(struct caam_export_state),
+			},
+		 },
+		.alg_type = OP_ALG_ALGSEL_AES | OP_ALG_AAI_CMAC,
 	},
 };
 
@@ -1695,14 +1855,45 @@ static int caam_hash_cra_init(struct crypto_tfm *tfm)
 	}
 
 	priv = dev_get_drvdata(ctx->jrdev->parent);
-	ctx->dir = priv->era >= 6 ? DMA_BIDIRECTIONAL : DMA_TO_DEVICE;
+
+	if (is_xcbc_aes(caam_hash->alg_type)) {
+		ctx->dir = DMA_TO_DEVICE;
+		ctx->adata.algtype = OP_TYPE_CLASS1_ALG | caam_hash->alg_type;
+		ctx->ctx_len = 48;
+
+		ctx->key_dma = dma_map_single_attrs(ctx->jrdev, ctx->key,
+						    ARRAY_SIZE(ctx->key),
+						    DMA_BIDIRECTIONAL,
+						    DMA_ATTR_SKIP_CPU_SYNC);
+		if (dma_mapping_error(ctx->jrdev, ctx->key_dma)) {
+			dev_err(ctx->jrdev, "unable to map key\n");
+			caam_jr_free(ctx->jrdev);
+			return -ENOMEM;
+		}
+	} else if (is_cmac_aes(caam_hash->alg_type)) {
+		ctx->dir = DMA_TO_DEVICE;
+		ctx->adata.algtype = OP_TYPE_CLASS1_ALG | caam_hash->alg_type;
+		ctx->ctx_len = 32;
+	} else {
+		ctx->dir = priv->era >= 6 ? DMA_BIDIRECTIONAL : DMA_TO_DEVICE;
+		ctx->adata.algtype = OP_TYPE_CLASS2_ALG | caam_hash->alg_type;
+		ctx->ctx_len = runninglen[(ctx->adata.algtype &
+					   OP_ALG_ALGSEL_SUBMASK) >>
+					  OP_ALG_ALGSEL_SHIFT];
+	}
 
 	dma_addr = dma_map_single_attrs(ctx->jrdev, ctx->sh_desc_update,
-					offsetof(struct caam_hash_ctx,
-						 sh_desc_update_dma),
+					offsetof(struct caam_hash_ctx, key),
 					ctx->dir, DMA_ATTR_SKIP_CPU_SYNC);
 	if (dma_mapping_error(ctx->jrdev, dma_addr)) {
 		dev_err(ctx->jrdev, "unable to map shared descriptors\n");
+
+		if (is_xcbc_aes(caam_hash->alg_type))
+			dma_unmap_single_attrs(ctx->jrdev, ctx->key_dma,
+					       ARRAY_SIZE(ctx->key),
+					       DMA_BIDIRECTIONAL,
+					       DMA_ATTR_SKIP_CPU_SYNC);
+
 		caam_jr_free(ctx->jrdev);
 		return -ENOMEM;
 	}
@@ -1716,16 +1907,14 @@ static int caam_hash_cra_init(struct crypto_tfm *tfm)
 	ctx->sh_desc_digest_dma = dma_addr + offsetof(struct caam_hash_ctx,
 						      sh_desc_digest);
 
-	/* copy descriptor header template value */
-	ctx->adata.algtype = OP_TYPE_CLASS2_ALG | caam_hash->alg_type;
-
-	ctx->ctx_len = runninglen[(ctx->adata.algtype &
-				   OP_ALG_ALGSEL_SUBMASK) >>
-				  OP_ALG_ALGSEL_SHIFT];
-
 	crypto_ahash_set_reqsize(__crypto_ahash_cast(tfm),
 				 sizeof(struct caam_hash_state));
-	return ahash_set_sh_desc(ahash);
+
+	/*
+	 * For keyed hash algorithms shared descriptors
+	 * will be created later in setkey() callback
+	 */
+	return alg->setkey ? 0 : ahash_set_sh_desc(ahash);
 }
 
 static void caam_hash_cra_exit(struct crypto_tfm *tfm)
@@ -1733,9 +1922,12 @@ static void caam_hash_cra_exit(struct crypto_tfm *tfm)
 	struct caam_hash_ctx *ctx = crypto_tfm_ctx(tfm);
 
 	dma_unmap_single_attrs(ctx->jrdev, ctx->sh_desc_update_dma,
-			       offsetof(struct caam_hash_ctx,
-					sh_desc_update_dma),
+			       offsetof(struct caam_hash_ctx, key),
 			       ctx->dir, DMA_ATTR_SKIP_CPU_SYNC);
+	if (is_xcbc_aes(ctx->adata.algtype))
+		dma_unmap_single_attrs(ctx->jrdev, ctx->key_dma,
+				       ARRAY_SIZE(ctx->key), DMA_BIDIRECTIONAL,
+				       DMA_ATTR_SKIP_CPU_SYNC);
 	caam_jr_free(ctx->jrdev);
 }
 
@@ -1866,14 +2058,16 @@ static int __init caam_algapi_hash_init(void)
 		struct caam_hash_template *alg = driver_hash + i;
 
 		/* If MD size is not supported by device, skip registration */
-		if (alg->template_ahash.halg.digestsize > md_limit)
+		if (is_mdha(alg->alg_type) &&
+		    alg->template_ahash.halg.digestsize > md_limit)
 			continue;
 
 		/* register hmac version */
 		t_alg = caam_hash_alloc(alg, true);
 		if (IS_ERR(t_alg)) {
 			err = PTR_ERR(t_alg);
-			pr_warn("%s alg allocation failed\n", alg->driver_name);
+			pr_warn("%s alg allocation failed\n",
+				alg->hmac_driver_name);
 			continue;
 		}
 
@@ -1886,6 +2080,9 @@ static int __init caam_algapi_hash_init(void)
 		} else
 			list_add_tail(&t_alg->entry, &hash_list);
 
+		if ((alg->alg_type & OP_ALG_ALGSEL_MASK) == OP_ALG_ALGSEL_AES)
+			continue;
+
 		/* register unkeyed version */
 		t_alg = caam_hash_alloc(alg, false);
 		if (IS_ERR(t_alg)) {