[dpdk-dev] net/mlx5: fix multi segment packet send

Message ID 1482766116-6202-1-git-send-email-shahafs@mellanox.com (mailing list archive)
State Accepted, archived
Delegated to: Ferruh Yigit
Headers

Checks

Context Check Description
ci/checkpatch success coding style OK
ci/Intel compilation success Compilation OK

Commit Message

Shahaf Shuler Dec. 26, 2016, 3:28 p.m. UTC
  Dseg pointer is not initialised when the first segment is inlined
causing a segmentation fault in such situation.

Fixes: 2a66cf378954 ("net/mlx5: support inline send")

CC: stable@dpdk.org
Signed-off-by: Shahaf Shuler <shahafs@mellanox.com>
---
 drivers/net/mlx5/mlx5_rxtx.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
  

Comments

Adrien Mazarguil Dec. 27, 2016, 10:21 a.m. UTC | #1
Hi Shahaf,

On Mon, Dec 26, 2016 at 05:28:36PM +0200, Shahaf Shuler wrote:
> Dseg pointer is not initialised when the first segment is inlined
> causing a segmentation fault in such situation.
> 
> Fixes: 2a66cf378954 ("net/mlx5: support inline send")
> 
> CC: stable@dpdk.org
> Signed-off-by: Shahaf Shuler <shahafs@mellanox.com>

Thanks for fixing this bug, a few comments below.

> ---
>  drivers/net/mlx5/mlx5_rxtx.c | 6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/mlx5/mlx5_rxtx.c b/drivers/net/mlx5/mlx5_rxtx.c
> index 97810e8..d6688c6 100644
> --- a/drivers/net/mlx5/mlx5_rxtx.c
> +++ b/drivers/net/mlx5/mlx5_rxtx.c
> @@ -483,7 +483,7 @@
>  				assert(addr <= addr_end);
>  			}
>  			/*
> -			 * 2 DWORDs consumed by the WQE header + 1 DSEG +
> +			 * 2 DWORDs consumed by the WQE header + ETH segment +
>  			 * the size of the inline part of the packet.
>  			 */
>  			ds = 2 + MLX5_WQE_DS(pkt_inline_sz - 2);
> @@ -498,6 +498,10 @@
>  			} else if (!segs_n) {
>  				goto next_pkt;
>  			} else {
> +				/* dseg will be advance as part of next_seg*/

Nit-picking here, there is a missing space in the above comment.

> +				dseg = (volatile rte_v128u32_t *)

rte_v128u32_t does not exist (yet) in the tree, this patch therefore depends
on "eal: define generic vector types" [1]. Such dependencies should be
mentioned as a notes section of a patch (after a three-dash line).

Regarding stable@dpdk.org, un case the vector types patch is not applied on
the stable branch, you'll also have to provide your own definition.

> +					((uintptr_t)wqe +
> +					 ((ds - 1) * MLX5_WQE_DWORD_SIZE));
>  				goto next_seg;
>  			}
>  		} else {
> -- 
> 1.8.3.1
> 

Otherwise,

Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>

[1] http://dpdk.org/ml/archives/dev/2016-November/050261.html
  
Ferruh Yigit Jan. 5, 2017, 6:55 p.m. UTC | #2
On 12/27/2016 10:21 AM, Adrien Mazarguil wrote:
> Hi Shahaf,
> 
> On Mon, Dec 26, 2016 at 05:28:36PM +0200, Shahaf Shuler wrote:
>> Dseg pointer is not initialised when the first segment is inlined
>> causing a segmentation fault in such situation.
>>
>> Fixes: 2a66cf378954 ("net/mlx5: support inline send")
>>
>> CC: stable@dpdk.org
>> Signed-off-by: Shahaf Shuler <shahafs@mellanox.com>
...
> Acked-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>

Applied to dpdk-next-net/master, thanks.

(whitespace int the comment fixed)
  

Patch

diff --git a/drivers/net/mlx5/mlx5_rxtx.c b/drivers/net/mlx5/mlx5_rxtx.c
index 97810e8..d6688c6 100644
--- a/drivers/net/mlx5/mlx5_rxtx.c
+++ b/drivers/net/mlx5/mlx5_rxtx.c
@@ -483,7 +483,7 @@ 
 				assert(addr <= addr_end);
 			}
 			/*
-			 * 2 DWORDs consumed by the WQE header + 1 DSEG +
+			 * 2 DWORDs consumed by the WQE header + ETH segment +
 			 * the size of the inline part of the packet.
 			 */
 			ds = 2 + MLX5_WQE_DS(pkt_inline_sz - 2);
@@ -498,6 +498,10 @@ 
 			} else if (!segs_n) {
 				goto next_pkt;
 			} else {
+				/* dseg will be advance as part of next_seg*/
+				dseg = (volatile rte_v128u32_t *)
+					((uintptr_t)wqe +
+					 ((ds - 1) * MLX5_WQE_DWORD_SIZE));
 				goto next_seg;
 			}
 		} else {