[PATCH 4.1 42/78] crypto: nx - respect sg limit bounds when building sg lists for SHA

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

4.1-stable review patch.  If anyone has any objections, please let me know.

------------------

From: Jan Stancek <jstancek@redhat.com>

commit d3392f41f6d3cd0a034bd0aca47fabea2b47218e upstream.

Commit 000851119e80 changed sha256/512 update functions to
pass more data to nx_build_sg_list(), which ends with
sg list overflows and usually with update functions failing
for data larger than max_sg_len * NX_PAGE_SIZE.

This happens because:
- both "total" and "to_process" are updated, which leads to
  "to_process" getting overflowed for some data lengths
  For example:
    In first iteration "total" is 50, and let's assume "to_process"
    is 30 due to sg limits. At the end of first iteration "total" is
    set to 20. At start of 2nd iteration "to_process" overflows on:
      to_process = total - to_process;
- "in_sg" is not reset to nx_ctx->in_sg after each iteration
- nx_build_sg_list() is hitting overflow because the amount of data
  passed to it would require more than sgmax elements
- as consequence of previous item, data stored in overflowed sg list
  may no longer be aligned to SHA*_BLOCK_SIZE

This patch changes sha256/512 update functions so that "to_process"
respects sg limits and never tries to pass more data to
nx_build_sg_list() to avoid overflows. "to_process" is calculated
as minimum of "total" and sg limits at start of every iteration.

Fixes: 000851119e80 ("crypto: nx - Fix SHA concurrence issue and sg
		      limit bounds")
Signed-off-by: Jan Stancek <jstancek@redhat.com>
Cc: Leonidas Da Silva Barbosa <leosilva@linux.vnet.ibm.com>
Cc: Marcelo Henrique Cerri <mhcerri@linux.vnet.ibm.com>
Cc: Fionnuala Gunter <fin@linux.vnet.ibm.com>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

---
 drivers/crypto/nx/nx-sha256.c |   27 ++++++++++++++++-----------
 drivers/crypto/nx/nx-sha512.c |   28 ++++++++++++++++------------
 2 files changed, 32 insertions(+), 23 deletions(-)

--- a/drivers/crypto/nx/nx-sha256.c
+++ b/drivers/crypto/nx/nx-sha256.c
@@ -71,7 +71,6 @@ static int nx_sha256_update(struct shash
 	struct sha256_state *sctx = shash_desc_ctx(desc);
 	struct nx_crypto_ctx *nx_ctx = crypto_tfm_ctx(&desc->tfm->base);
 	struct nx_csbcpb *csbcpb = (struct nx_csbcpb *)nx_ctx->csbcpb;
-	struct nx_sg *in_sg;
 	struct nx_sg *out_sg;
 	u64 to_process = 0, leftover, total;
 	unsigned long irq_flags;
@@ -97,7 +96,6 @@ static int nx_sha256_update(struct shash
 	NX_CPB_FDM(csbcpb) |= NX_FDM_INTERMEDIATE;
 	NX_CPB_FDM(csbcpb) |= NX_FDM_CONTINUATION;
 
-	in_sg = nx_ctx->in_sg;
 	max_sg_len = min_t(u64, nx_ctx->ap->sglen,
 			nx_driver.of.max_sg_len/sizeof(struct nx_sg));
 	max_sg_len = min_t(u64, max_sg_len,
@@ -114,17 +112,12 @@ static int nx_sha256_update(struct shash
 	}
 
 	do {
-		/*
-		 * to_process: the SHA256_BLOCK_SIZE data chunk to process in
-		 * this update. This value is also restricted by the sg list
-		 * limits.
-		 */
-		to_process = total - to_process;
-		to_process = to_process & ~(SHA256_BLOCK_SIZE - 1);
+		int used_sgs = 0;
+		struct nx_sg *in_sg = nx_ctx->in_sg;
 
 		if (buf_len) {
 			data_len = buf_len;
-			in_sg = nx_build_sg_list(nx_ctx->in_sg,
+			in_sg = nx_build_sg_list(in_sg,
 						 (u8 *) sctx->buf,
 						 &data_len,
 						 max_sg_len);
@@ -133,15 +126,27 @@ static int nx_sha256_update(struct shash
 				rc = -EINVAL;
 				goto out;
 			}
+			used_sgs = in_sg - nx_ctx->in_sg;
 		}
 
+		/* to_process: SHA256_BLOCK_SIZE aligned chunk to be
+		 * processed in this iteration. This value is restricted
+		 * by sg list limits and number of sgs we already used
+		 * for leftover data. (see above)
+		 * In ideal case, we could allow NX_PAGE_SIZE * max_sg_len,
+		 * but because data may not be aligned, we need to account
+		 * for that too. */
+		to_process = min_t(u64, total,
+			(max_sg_len - 1 - used_sgs) * NX_PAGE_SIZE);
+		to_process = to_process & ~(SHA256_BLOCK_SIZE - 1);
+
 		data_len = to_process - buf_len;
 		in_sg = nx_build_sg_list(in_sg, (u8 *) data,
 					 &data_len, max_sg_len);
 
 		nx_ctx->op.inlen = (nx_ctx->in_sg - in_sg) * sizeof(struct nx_sg);
 
-		to_process = (data_len + buf_len);
+		to_process = data_len + buf_len;
 		leftover = total - to_process;
 
 		/*
--- a/drivers/crypto/nx/nx-sha512.c
+++ b/drivers/crypto/nx/nx-sha512.c
@@ -71,7 +71,6 @@ static int nx_sha512_update(struct shash
 	struct sha512_state *sctx = shash_desc_ctx(desc);
 	struct nx_crypto_ctx *nx_ctx = crypto_tfm_ctx(&desc->tfm->base);
 	struct nx_csbcpb *csbcpb = (struct nx_csbcpb *)nx_ctx->csbcpb;
-	struct nx_sg *in_sg;
 	struct nx_sg *out_sg;
 	u64 to_process, leftover = 0, total;
 	unsigned long irq_flags;
@@ -97,7 +96,6 @@ static int nx_sha512_update(struct shash
 	NX_CPB_FDM(csbcpb) |= NX_FDM_INTERMEDIATE;
 	NX_CPB_FDM(csbcpb) |= NX_FDM_CONTINUATION;
 
-	in_sg = nx_ctx->in_sg;
 	max_sg_len = min_t(u64, nx_ctx->ap->sglen,
 			nx_driver.of.max_sg_len/sizeof(struct nx_sg));
 	max_sg_len = min_t(u64, max_sg_len,
@@ -114,18 +112,12 @@ static int nx_sha512_update(struct shash
 	}
 
 	do {
-		/*
-		 * to_process: the SHA512_BLOCK_SIZE data chunk to process in
-		 * this update. This value is also restricted by the sg list
-		 * limits.
-		 */
-		to_process = total - leftover;
-		to_process = to_process & ~(SHA512_BLOCK_SIZE - 1);
-		leftover = total - to_process;
+		int used_sgs = 0;
+		struct nx_sg *in_sg = nx_ctx->in_sg;
 
 		if (buf_len) {
 			data_len = buf_len;
-			in_sg = nx_build_sg_list(nx_ctx->in_sg,
+			in_sg = nx_build_sg_list(in_sg,
 						 (u8 *) sctx->buf,
 						 &data_len, max_sg_len);
 
@@ -133,8 +125,20 @@ static int nx_sha512_update(struct shash
 				rc = -EINVAL;
 				goto out;
 			}
+			used_sgs = in_sg - nx_ctx->in_sg;
 		}
 
+		/* to_process: SHA512_BLOCK_SIZE aligned chunk to be
+		 * processed in this iteration. This value is restricted
+		 * by sg list limits and number of sgs we already used
+		 * for leftover data. (see above)
+		 * In ideal case, we could allow NX_PAGE_SIZE * max_sg_len,
+		 * but because data may not be aligned, we need to account
+		 * for that too. */
+		to_process = min_t(u64, total,
+			(max_sg_len - 1 - used_sgs) * NX_PAGE_SIZE);
+		to_process = to_process & ~(SHA512_BLOCK_SIZE - 1);
+
 		data_len = to_process - buf_len;
 		in_sg = nx_build_sg_list(in_sg, (u8 *) data,
 					 &data_len, max_sg_len);
@@ -146,7 +150,7 @@ static int nx_sha512_update(struct shash
 			goto out;
 		}
 
-		to_process = (data_len + buf_len);
+		to_process = data_len + buf_len;
 		leftover = total - to_process;
 
 		/*