ENGR00215945-1: Rework scatterlist handling for bi-endian platforms

Former versions of this (ARM) branch of this driver reworked the hardware-
readable scatter/gather list to operate as a set of 32-bit integers,
rather than a packed structure of smaller sizes, which cannot burst-read
correctly on a little-endian platform.

Integration of caamhash.c revealed subtle ways in which the ordering of
items written to a hardware s/g list could create bugs, such as the
"final" bit being written to an entry that would later be updated with
a size, inadvertently erasing the bit (e.g. such as sg_to_sec4_sg_last()
before sg_to_sec4_sg()).

Since fields must be ORed in to operate correctly using any order of
operations, changed allocations of the combination of extended descriptor
structs + hardware scatterlists to use kzalloc() instead of kmalloc(), so
as to ensure that residue data would not be ORed in with the correct data.

Signed-off-by: Steve Cornelius <steve.cornelius@freescale.com>
Signed-off-by: Terry Lv <r65388@freescale.com>

3 files changed, 32 insertions, 28 deletions

diff --git a/drivers/crypto/caam/caamalg.c b/drivers/crypto/caam/caamalg.c
index 1d3d89aba09e..87b365f8089f 100644
--- a/drivers/crypto/caam/caamalg.c
+++ b/drivers/crypto/caam/caamalg.c
@@ -1391,7 +1391,7 @@ static struct aead_edesc *aead_giv_edesc_alloc(struct aead_givcrypt_request
 	dma_sync_single_for_device(jrdev, iv_dma, ivsize, DMA_TO_DEVICE);
 
 	/* allocate space for base edesc and hw desc commands, link tables */
-	edesc = kmalloc(sizeof(struct aead_edesc) + desc_bytes +
+	edesc = kzalloc(sizeof(struct aead_edesc) + desc_bytes +
 			sec4_sg_bytes, GFP_DMA | flags);
 	if (!edesc) {
 		dev_err(jrdev, "could not allocate extended descriptor\n");
@@ -1538,7 +1538,7 @@ static struct ablkcipher_edesc *ablkcipher_edesc_alloc(struct ablkcipher_request
 			sizeof(struct sec4_sg_entry);
 
 	/* allocate space for base edesc and hw desc commands, link tables */
-	edesc = kmalloc(sizeof(struct ablkcipher_edesc) + desc_bytes +
+	edesc = kzalloc(sizeof(struct ablkcipher_edesc) + desc_bytes +
 			sec4_sg_bytes, GFP_DMA | flags);
 	if (!edesc) {
 		dev_err(jrdev, "could not allocate extended descriptor\n");
diff --git a/drivers/crypto/caam/caamhash.c b/drivers/crypto/caam/caamhash.c
index 7d1fd28046cb..fe6af79419db 100644
--- a/drivers/crypto/caam/caamhash.c
+++ b/drivers/crypto/caam/caamhash.c
@@ -824,7 +824,7 @@ static int ahash_update_ctx(struct ahash_request *req)
 		 * allocate space for base edesc and hw desc commands,
 		 * link tables
 		 */
-		edesc = kmalloc(sizeof(struct ahash_edesc) + DESC_JOB_IO_LEN +
+		edesc = kzalloc(sizeof(struct ahash_edesc) + DESC_JOB_IO_LEN +
 				sec4_sg_bytes, GFP_DMA | flags);
 		if (!edesc) {
 			dev_err(jrdev,
@@ -841,9 +841,6 @@ static int ahash_update_ctx(struct ahash_request *req)
 						     sec4_sg_bytes,
 						     DMA_TO_DEVICE);
 
-		dma_sync_single_for_device(jrdev, edesc->sec4_sg_dma,
-					   sec4_sg_bytes, DMA_TO_DEVICE);
-
 		ctx_map_to_sec4_sg(desc, jrdev, state, ctx->ctx_len,
 				   edesc->sec4_sg, DMA_BIDIRECTIONAL);
 
@@ -876,6 +873,9 @@ static int ahash_update_ctx(struct ahash_request *req)
 
 		append_seq_out_ptr(desc, state->ctx_dma, ctx->ctx_len, 0);
 
+		dma_sync_single_for_device(jrdev, edesc->sec4_sg_dma,
+					   sec4_sg_bytes, DMA_TO_DEVICE);
+
 #ifdef DEBUG
 		print_hex_dump(KERN_ERR, "jobdesc@"xstr(__LINE__)": ",
 			       DUMP_PREFIX_ADDRESS, 16, 4, desc,
@@ -929,7 +929,7 @@ static int ahash_final_ctx(struct ahash_request *req)
 	sec4_sg_bytes = (1 + (buflen ? 1 : 0)) * sizeof(struct sec4_sg_entry);
 
 	/* allocate space for base edesc and hw desc commands, link tables */
-	edesc = kmalloc(sizeof(struct ahash_edesc) + DESC_JOB_IO_LEN +
+	edesc = kzalloc(sizeof(struct ahash_edesc) + DESC_JOB_IO_LEN +
 			sec4_sg_bytes, GFP_DMA | flags);
 	if (!edesc) {
 		dev_err(jrdev, "could not allocate extended descriptor\n");
@@ -945,8 +945,6 @@ static int ahash_final_ctx(struct ahash_request *req)
 			 DESC_JOB_IO_LEN;
 	edesc->sec4_sg_dma = dma_map_single(jrdev, edesc->sec4_sg,
 					    sec4_sg_bytes, DMA_TO_DEVICE);
-	dma_sync_single_for_device(jrdev, edesc->sec4_sg_dma, sec4_sg_bytes,
-				   DMA_TO_DEVICE);
 	edesc->src_nents = 0;
 
 	ctx_map_to_sec4_sg(desc, jrdev, state, ctx->ctx_len, edesc->sec4_sg,
@@ -963,6 +961,9 @@ static int ahash_final_ctx(struct ahash_request *req)
 	edesc->dst_dma = map_seq_out_ptr_result(desc, jrdev, req->result,
 						digestsize);
 
+	dma_sync_single_for_device(jrdev, edesc->sec4_sg_dma, sec4_sg_bytes,
+				   DMA_TO_DEVICE);
+
 #ifdef DEBUG
 	print_hex_dump(KERN_ERR, "jobdesc@"xstr(__LINE__)": ",
 		       DUMP_PREFIX_ADDRESS, 16, 4, desc, desc_bytes(desc), 1);
@@ -1007,7 +1008,7 @@ static int ahash_finup_ctx(struct ahash_request *req)
 			 sizeof(struct sec4_sg_entry);
 
 	/* allocate space for base edesc and hw desc commands, link tables */
-	edesc = kmalloc(sizeof(struct ahash_edesc) + DESC_JOB_IO_LEN +
+	edesc = kzalloc(sizeof(struct ahash_edesc) + DESC_JOB_IO_LEN +
 			sec4_sg_bytes, GFP_DMA | flags);
 	if (!edesc) {
 		dev_err(jrdev, "could not allocate extended descriptor\n");
@@ -1025,8 +1026,6 @@ static int ahash_finup_ctx(struct ahash_request *req)
 			 DESC_JOB_IO_LEN;
 	edesc->sec4_sg_dma = dma_map_single(jrdev, edesc->sec4_sg,
 					    sec4_sg_bytes, DMA_TO_DEVICE);
-	dma_sync_single_for_device(jrdev, edesc->sec4_sg_dma, sec4_sg_bytes,
-				   DMA_TO_DEVICE);
 
 	ctx_map_to_sec4_sg(desc, jrdev, state, ctx->ctx_len, edesc->sec4_sg,
 			   DMA_TO_DEVICE);
@@ -1044,6 +1043,9 @@ static int ahash_finup_ctx(struct ahash_request *req)
 	edesc->dst_dma = map_seq_out_ptr_result(desc, jrdev, req->result,
 						digestsize);
 
+	dma_sync_single_for_device(jrdev, edesc->sec4_sg_dma, sec4_sg_bytes,
+				   DMA_TO_DEVICE);
+
 #ifdef DEBUG
 	print_hex_dump(KERN_ERR, "jobdesc@"xstr(__LINE__)": ",
 		       DUMP_PREFIX_ADDRESS, 16, 4, desc, desc_bytes(desc), 1);
@@ -1084,7 +1086,7 @@ static int ahash_digest(struct ahash_request *req)
 	sec4_sg_bytes = src_nents * sizeof(struct sec4_sg_entry);
 
 	/* allocate space for base edesc and hw desc commands, link tables */
-	edesc = kmalloc(sizeof(struct ahash_edesc) + sec4_sg_bytes +
+	edesc = kzalloc(sizeof(struct ahash_edesc) + sec4_sg_bytes +
 			DESC_JOB_IO_LEN, GFP_DMA | flags);
 	if (!edesc) {
 		dev_err(jrdev, "could not allocate extended descriptor\n");
@@ -1095,8 +1097,6 @@ static int ahash_digest(struct ahash_request *req)
 	edesc->sec4_sg_dma = dma_map_single(jrdev, edesc->sec4_sg,
 					    sec4_sg_bytes, DMA_TO_DEVICE);
 	edesc->sec4_sg_bytes = sec4_sg_bytes;
-	dma_sync_single_for_device(jrdev, edesc->sec4_sg_dma, sec4_sg_bytes,
-				   DMA_TO_DEVICE);
 	edesc->src_nents = src_nents;
 	edesc->chained = chained;
 
@@ -1114,6 +1114,9 @@ static int ahash_digest(struct ahash_request *req)
 	}
 	append_seq_in_ptr(desc, src_dma, req->nbytes, options);
 
+	dma_sync_single_for_device(jrdev, edesc->sec4_sg_dma,
+				   edesc->sec4_sg_bytes, DMA_TO_DEVICE);
+
 	edesc->dst_dma = map_seq_out_ptr_result(desc, jrdev, req->result,
 						digestsize);
 
@@ -1152,7 +1155,7 @@ static int ahash_final_no_ctx(struct ahash_request *req)
 	int sh_len;
 
 	/* allocate space for base edesc and hw desc commands, link tables */
-	edesc = kmalloc(sizeof(struct ahash_edesc) + DESC_JOB_IO_LEN,
+	edesc = kzalloc(sizeof(struct ahash_edesc) + DESC_JOB_IO_LEN,
 			GFP_DMA | flags);
 	if (!edesc) {
 		dev_err(jrdev, "could not allocate extended descriptor\n");
@@ -1164,8 +1167,6 @@ static int ahash_final_no_ctx(struct ahash_request *req)
 	init_job_desc_shared(desc, ptr, sh_len, HDR_SHARE_DEFER | HDR_REVERSE);
 
 	state->buf_dma = dma_map_single(jrdev, buf, buflen, DMA_TO_DEVICE);
-	dma_sync_single_for_device(jrdev, state->buf_dma, buflen,
-				   DMA_TO_DEVICE);
 
 	append_seq_in_ptr(desc, state->buf_dma, buflen, 0);
 
@@ -1173,6 +1174,8 @@ static int ahash_final_no_ctx(struct ahash_request *req)
 						digestsize);
 	edesc->src_nents = 0;
 
+	dma_sync_single_for_device(jrdev, state->buf_dma, buflen,
+				   DMA_TO_DEVICE);
 #ifdef DEBUG
 	print_hex_dump(KERN_ERR, "jobdesc@"xstr(__LINE__)": ",
 		       DUMP_PREFIX_ADDRESS, 16, 4, desc, desc_bytes(desc), 1);
@@ -1225,7 +1228,7 @@ static int ahash_update_no_ctx(struct ahash_request *req)
 		 * allocate space for base edesc and hw desc commands,
 		 * link tables
 		 */
-		edesc = kmalloc(sizeof(struct ahash_edesc) + DESC_JOB_IO_LEN +
+		edesc = kzalloc(sizeof(struct ahash_edesc) + DESC_JOB_IO_LEN +
 				sec4_sg_bytes, GFP_DMA | flags);
 		if (!edesc) {
 			dev_err(jrdev,
@@ -1241,8 +1244,6 @@ static int ahash_update_no_ctx(struct ahash_request *req)
 		edesc->sec4_sg_dma = dma_map_single(jrdev, edesc->sec4_sg,
 						    sec4_sg_bytes,
 						    DMA_TO_DEVICE);
-		dma_sync_single_for_device(jrdev, edesc->sec4_sg_dma,
-					   sec4_sg_bytes, DMA_TO_DEVICE);
 
 		state->buf_dma = buf_map_to_sec4_sg(jrdev, edesc->sec4_sg,
 						    buf, *buflen);
@@ -1263,6 +1264,8 @@ static int ahash_update_no_ctx(struct ahash_request *req)
 
 		map_seq_out_ptr_ctx(desc, jrdev, state, ctx->ctx_len);
 
+		dma_sync_single_for_device(jrdev, edesc->sec4_sg_dma,
+					   sec4_sg_bytes, DMA_TO_DEVICE);
 #ifdef DEBUG
 		print_hex_dump(KERN_ERR, "jobdesc@"xstr(__LINE__)": ",
 			       DUMP_PREFIX_ADDRESS, 16, 4, desc,
@@ -1324,7 +1327,7 @@ static int ahash_finup_no_ctx(struct ahash_request *req)
 			 sizeof(struct sec4_sg_entry);
 
 	/* allocate space for base edesc and hw desc commands, link tables */
-	edesc = kmalloc(sizeof(struct ahash_edesc) + DESC_JOB_IO_LEN +
+	edesc = kzalloc(sizeof(struct ahash_edesc) + DESC_JOB_IO_LEN +
 			sec4_sg_bytes, GFP_DMA | flags);
 	if (!edesc) {
 		dev_err(jrdev, "could not allocate extended descriptor\n");
@@ -1342,8 +1345,6 @@ static int ahash_finup_no_ctx(struct ahash_request *req)
 			 DESC_JOB_IO_LEN;
 	edesc->sec4_sg_dma = dma_map_single(jrdev, edesc->sec4_sg,
 					    sec4_sg_bytes, DMA_TO_DEVICE);
-	dma_sync_single_for_device(jrdev, edesc->sec4_sg_dma, sec4_sg_bytes,
-				   DMA_TO_DEVICE);
 
 	state->buf_dma = try_buf_map_to_sec4_sg(jrdev, edesc->sec4_sg, buf,
 						state->buf_dma, buflen,
@@ -1358,6 +1359,9 @@ static int ahash_finup_no_ctx(struct ahash_request *req)
 	edesc->dst_dma = map_seq_out_ptr_result(desc, jrdev, req->result,
 						digestsize);
 
+	dma_sync_single_for_device(jrdev, edesc->sec4_sg_dma, sec4_sg_bytes,
+				   DMA_TO_DEVICE);
+
 #ifdef DEBUG
 	print_hex_dump(KERN_ERR, "jobdesc@"xstr(__LINE__)": ",
 		       DUMP_PREFIX_ADDRESS, 16, 4, desc, desc_bytes(desc), 1);
@@ -1412,7 +1416,7 @@ static int ahash_update_first(struct ahash_request *req)
 		 * allocate space for base edesc and hw desc commands,
 		 * link tables
 		 */
-		edesc = kmalloc(sizeof(struct ahash_edesc) + DESC_JOB_IO_LEN +
+		edesc = kzalloc(sizeof(struct ahash_edesc) + DESC_JOB_IO_LEN +
 				sec4_sg_bytes, GFP_DMA | flags);
 		if (!edesc) {
 			dev_err(jrdev,
@@ -1428,8 +1432,6 @@ static int ahash_update_first(struct ahash_request *req)
 		edesc->sec4_sg_dma = dma_map_single(jrdev, edesc->sec4_sg,
 						    sec4_sg_bytes,
 						    DMA_TO_DEVICE);
-		dma_sync_single_for_device(jrdev, edesc->sec4_sg_dma,
-					   sec4_sg_bytes, DMA_TO_DEVICE);
 
 		if (src_nents) {
 			sg_to_sec4_sg_last(req->src, src_nents,
@@ -1453,6 +1455,8 @@ static int ahash_update_first(struct ahash_request *req)
 
 		map_seq_out_ptr_ctx(desc, jrdev, state, ctx->ctx_len);
 
+		dma_sync_single_for_device(jrdev, edesc->sec4_sg_dma,
+					   sec4_sg_bytes, DMA_TO_DEVICE);
 #ifdef DEBUG
 		print_hex_dump(KERN_ERR, "jobdesc@"xstr(__LINE__)": ",
 			       DUMP_PREFIX_ADDRESS, 16, 4, desc,
diff --git a/drivers/crypto/caam/sg_sw_sec4.h b/drivers/crypto/caam/sg_sw_sec4.h
index b2286ecce87b..e05fc58c9637 100644
--- a/drivers/crypto/caam/sg_sw_sec4.h
+++ b/drivers/crypto/caam/sg_sw_sec4.h
@@ -17,7 +17,7 @@ static inline void dma_to_sec4_sg_one(struct sec4_sg_entry *sec4_sg_ptr,
 	sec4_sg_ptr->reserved = 0;	/* ensure MSB half is zeroed */
 #endif
 	sec4_sg_ptr->ptr = dma;
-	sec4_sg_ptr->len = (len & SEC4_SG_LEN_MASK);
+	sec4_sg_ptr->len |= (len & SEC4_SG_LEN_MASK);
 	/* Does not add in buffer pool ID's at this time */
 	sec4_sg_ptr->bpid_offset = (offset & SEC4_SG_OFFS_MASK);
 #ifdef DEBUG