[v2] event/cnxk: fix incorrect mbuf offset calculation
Checks
Commit Message
From: Pavan Nikhilesh <pbhagavatula@marvell.com>
Fix incorrect mbuf offset calculation when HEADROOM exceeds 128B
while processing event vectors.
Fixes: 7fbbc981d54f("event/cnxk: support vectorized Rx event fast path")
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
---
v2 Changes:
- Remove internal Change-Id.
drivers/net/cnxk/cn10k_rx.h | 8 +++++---
1 file changed, 5 insertions(+), 3 deletions(-)
--
2.25.1
Comments
On Tue, Oct 25, 2022 at 9:41 PM <pbhagavatula@marvell.com> wrote:
>
> From: Pavan Nikhilesh <pbhagavatula@marvell.com>
>
> Fix incorrect mbuf offset calculation when HEADROOM exceeds 128B
> while processing event vectors.
>
> Fixes: 7fbbc981d54f("event/cnxk: support vectorized Rx event fast path")
Cc: stable@dpdk.org
Applied to dpdk-next-net-eventdev/for-main. Thanks
>
> Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
> ---
> v2 Changes:
> - Remove internal Change-Id.
>
> drivers/net/cnxk/cn10k_rx.h | 8 +++++---
> 1 file changed, 5 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/net/cnxk/cn10k_rx.h b/drivers/net/cnxk/cn10k_rx.h
> index 46488d442e..f562a75245 100644
> --- a/drivers/net/cnxk/cn10k_rx.h
> +++ b/drivers/net/cnxk/cn10k_rx.h
> @@ -1201,9 +1201,11 @@ cn10k_nix_recv_pkts_vector(void *args, struct rte_mbuf **mbufs, uint16_t pkts,
> mbuf23 = vqsubq_u64(mbuf23, data_off);
> } else {
> mbuf01 =
> - vsubq_u64(vld1q_u64((uint64_t *)cq0), data_off);
> - mbuf23 = vsubq_u64(vld1q_u64((uint64_t *)(cq0 + 16)),
> - data_off);
> + vsubq_u64(vld1q_u64((uint64_t *)cq0),
> + vdupq_n_u64(sizeof(struct rte_mbuf)));
> + mbuf23 =
> + vsubq_u64(vld1q_u64((uint64_t *)(cq0 + 16)),
> + vdupq_n_u64(sizeof(struct rte_mbuf)));
> }
>
> /* Move mbufs to scalar registers for future use */
> --
> 2.25.1
>
@@ -1201,9 +1201,11 @@ cn10k_nix_recv_pkts_vector(void *args, struct rte_mbuf **mbufs, uint16_t pkts,
mbuf23 = vqsubq_u64(mbuf23, data_off);
} else {
mbuf01 =
- vsubq_u64(vld1q_u64((uint64_t *)cq0), data_off);
- mbuf23 = vsubq_u64(vld1q_u64((uint64_t *)(cq0 + 16)),
- data_off);
+ vsubq_u64(vld1q_u64((uint64_t *)cq0),
+ vdupq_n_u64(sizeof(struct rte_mbuf)));
+ mbuf23 =
+ vsubq_u64(vld1q_u64((uint64_t *)(cq0 + 16)),
+ vdupq_n_u64(sizeof(struct rte_mbuf)));
}
/* Move mbufs to scalar registers for future use */