| Message ID | 1469647047-7544-4-git-send-email-jarod@redhat.com (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
On Wed, Jul 27, 2016 at 03:17:24PM -0400, Jarod Wilson wrote: > At present, the size of eseg->inline_hdr_start is 16 bits, while > MLX5_ETH_L2_INLINE_HEADER_SIZE is 18, so there are attempts made to copy > 18 bits into 16 bits of storage. The mlx5_dbg() statement in > copy_eth_inline_header() suggests that perhaps > MLX5_ETH_L2_INLINE_HEADER_SIZE should be only 16, not 18. So either that > needs to be changed, or the inline_hdr_start array needs to be bumped up > to 3 bytes instead of 2. Ugh. Now I see what's going on. The copy is actually designed to copy 18 _bytes_, not bits, into inline_hdr_start[2] and inline_hdr[16]. Is there a particular reason those two aren't just a single array?
On Wed, Jul 27, 2016 at 05:26:10PM -0400, Jarod Wilson wrote: > On Wed, Jul 27, 2016 at 03:17:24PM -0400, Jarod Wilson wrote: > > At present, the size of eseg->inline_hdr_start is 16 bits, while > > MLX5_ETH_L2_INLINE_HEADER_SIZE is 18, so there are attempts made to copy > > 18 bits into 16 bits of storage. The mlx5_dbg() statement in > > copy_eth_inline_header() suggests that perhaps > > MLX5_ETH_L2_INLINE_HEADER_SIZE should be only 16, not 18. So either that > > needs to be changed, or the inline_hdr_start array needs to be bumped up > > to 3 bytes instead of 2. > > Ugh. Now I see what's going on. The copy is actually designed to copy 18 > _bytes_, not bits, into inline_hdr_start[2] and inline_hdr[16]. Is there a > particular reason those two aren't just a single array? Drop this one. I've got a new patch together that just merges the two. I've looked over the code, and can't see anything that actually uses inline_hdr separate from inline_hdr_start.
diff --git a/src/wqe.h b/src/wqe.h index c2622d5..c0e176d 100644 --- a/src/wqe.h +++ b/src/wqe.h @@ -77,7 +77,7 @@ struct mlx5_eqe_qp_srq { }; enum { - MLX5_ETH_L2_INLINE_HEADER_SIZE = 18, + MLX5_ETH_L2_INLINE_HEADER_SIZE = 16, }; enum {
At present, the size of eseg->inline_hdr_start is 16 bits, while MLX5_ETH_L2_INLINE_HEADER_SIZE is 18, so there are attempts made to copy 18 bits into 16 bits of storage. The mlx5_dbg() statement in copy_eth_inline_header() suggests that perhaps MLX5_ETH_L2_INLINE_HEADER_SIZE should be only 16, not 18. So either that needs to be changed, or the inline_hdr_start array needs to be bumped up to 3 bytes instead of 2. CC: Yishai Hadas <yishaih@mellanox.com> Signed-off-by: Jarod Wilson <jarod@redhat.com> --- src/wqe.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)