[PATCH] target/iscsi: fix digest computation for chained SGs

[PATCH] target/iscsi: fix digest computation for chained SGs

[Spencer Baugh]

From: Alexei Potashnik <alexei@purestorage.com>

Current implementation assumes that all the buffers of an IO are linked
with a single SG list. Which makes it fail if SG chaining is used.

Signed-off-by: Alexei Potashnik <alexei@purestorage.com>
---
 drivers/target/iscsi/iscsi_target.c | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/drivers/target/iscsi/iscsi_target.c b/drivers/target/iscsi/iscsi_target.c
index 4e68b62..a4cf58c 100644
--- a/drivers/target/iscsi/iscsi_target.c
+++ b/drivers/target/iscsi/iscsi_target.c
@@ -1209,7 +1209,6 @@ static u32 iscsit_do_crypto_hash_sg(
 	u8 *pad_bytes)
 {
 	u32 data_crc;
-	u32 i;
 	struct scatterlist *sg;
 	unsigned int page_off;
 
@@ -1218,15 +1217,15 @@ static u32 iscsit_do_crypto_hash_sg(
 	sg = cmd->first_data_sg;
 	page_off = cmd->first_data_sg_off;
 
-	i = 0;
 	while (data_length) {
-		u32 cur_len = min_t(u32, data_length, (sg[i].length - page_off));
+		u32 cur_len = min_t(u32, data_length, (sg->length - page_off));
 
-		crypto_hash_update(hash, &sg[i], cur_len);
+		crypto_hash_update(hash, sg, cur_len);
 
 		data_length -= cur_len;
 		page_off = 0;
-		i++;
+		/* iscsit_map_iovec has already checked for invalid sg pointers */
+		sg = sg_next(sg);
 	}
 
 	if (padding) {
-- 
2.4.3

[PATCH] target/iscsi: fix digest computation for chained SGs

[Spencer Baugh]

From: Alexei Potashnik <alexei@purestorage.com>

Current implementation assumes that all the buffers of an IO are linked
with a single SG list. Which makes it fail if SG chaining is used.

Signed-off-by: Alexei Potashnik <alexei@purestorage.com>
---
 drivers/target/iscsi/iscsi_target.c | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/drivers/target/iscsi/iscsi_target.c b/drivers/target/iscsi/iscsi_target.c
index 4e68b62..a4cf58c 100644
--- a/drivers/target/iscsi/iscsi_target.c
+++ b/drivers/target/iscsi/iscsi_target.c
@@ -1209,7 +1209,6 @@ static u32 iscsit_do_crypto_hash_sg(
 	u8 *pad_bytes)
 {
 	u32 data_crc;
-	u32 i;
 	struct scatterlist *sg;
 	unsigned int page_off;
 
@@ -1218,15 +1217,15 @@ static u32 iscsit_do_crypto_hash_sg(
 	sg = cmd->first_data_sg;
 	page_off = cmd->first_data_sg_off;
 
-	i = 0;
 	while (data_length) {
-		u32 cur_len = min_t(u32, data_length, (sg[i].length - page_off));
+		u32 cur_len = min_t(u32, data_length, (sg->length - page_off));
 
-		crypto_hash_update(hash, &sg[i], cur_len);
+		crypto_hash_update(hash, sg, cur_len);
 
 		data_length -= cur_len;
 		page_off = 0;
-		i++;
+		/* iscsit_map_iovec has already checked for invalid sg pointers */
+		sg = sg_next(sg);
 	}
 
 	if (padding) {
-- 
2.4.3

Re: [PATCH] target/iscsi: fix digest computation for chained SGs

[Sagi Grimberg]

On 7/21/2015 3:12 AM, Spencer Baugh wrote:
> From: Alexei Potashnik <alexei@purestorage.com>
>
> Current implementation assumes that all the buffers of an IO are linked
> with a single SG list. Which makes it fail if SG chaining is used.
>
> Signed-off-by: Alexei Potashnik <alexei@purestorage.com>
> ---
>   drivers/target/iscsi/iscsi_target.c | 9 ++++-----
>   1 file changed, 4 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/target/iscsi/iscsi_target.c b/drivers/target/iscsi/iscsi_target.c
> index 4e68b62..a4cf58c 100644
> --- a/drivers/target/iscsi/iscsi_target.c
> +++ b/drivers/target/iscsi/iscsi_target.c
> @@ -1209,7 +1209,6 @@ static u32 iscsit_do_crypto_hash_sg(
>   	u8 *pad_bytes)
>   {
>   	u32 data_crc;
> -	u32 i;
>   	struct scatterlist *sg;
>   	unsigned int page_off;
>
> @@ -1218,15 +1217,15 @@ static u32 iscsit_do_crypto_hash_sg(
>   	sg = cmd->first_data_sg;
>   	page_off = cmd->first_data_sg_off;
>
> -	i = 0;
>   	while (data_length) {
> -		u32 cur_len = min_t(u32, data_length, (sg[i].length - page_off));
> +		u32 cur_len = min_t(u32, data_length, (sg->length - page_off));
>
> -		crypto_hash_update(hash, &sg[i], cur_len);
> +		crypto_hash_update(hash, sg, cur_len);
>
>   		data_length -= cur_len;
>   		page_off = 0;
> -		i++;
> +		/* iscsit_map_iovec has already checked for invalid sg pointers */
> +		sg = sg_next(sg);
>   	}
>
>   	if (padding) {
>

How were you able to get a chained SG list in the target code?

In any event, looks good,

Reviewed-by: Sagi Grimberg <sagig@mellanox.com>

Re: [PATCH] target/iscsi: fix digest computation for chained SGs

[Roland Dreier]

On Tue, Jul 21, 2015 at 1:57 AM, Sagi Grimberg <sagig@dev.mellanox.co.il> wrote:
> How were you able to get a chained SG list in the target code?

Local hack.  So this bug can't be hit in current mainline code, but
patch improves the code and removes a hidden booby-trap, so I think it
makes sense to apply.

Re: [PATCH] target/iscsi: fix digest computation for chained SGs

[Nicholas A. Bellinger]

On Tue, 2015-07-21 at 18:38 -0700, Roland Dreier wrote:
> On Tue, Jul 21, 2015 at 1:57 AM, Sagi Grimberg <sagig@dev.mellanox.co.il> wrote:
> > How were you able to get a chained SG list in the target code?
> 
> Local hack.  So this bug can't be hit in current mainline code, but
> patch improves the code and removes a hidden booby-trap, so I think it
> makes sense to apply.
> --

Applied to target-pending/for-next.

Thanks Roland & Co.