* [PATCH] net/mlx5: fix miniCQEs number calculation
@ 2024-10-28 16:42 Alexander Kozyrev
2024-10-28 17:08 ` Slava Ovsiienko
2024-10-28 17:14 ` [PATCH v2] " Alexander Kozyrev
0 siblings, 2 replies; 3+ messages in thread
From: Alexander Kozyrev @ 2024-10-28 16:42 UTC (permalink / raw)
To: dev; +Cc: stable, rasland, viacheslavo, matan, dsosnowski, bingz, suanmingm
Use the information from the CQE, not from the title packet,
for getting the number of miniCQEs in the compressed CQEs array.
This way we can avoid segfaults in the rxq_cq_decompress_v()
in case of mbuf corruption (due to double mbuf free, for example).
Fixes: 6cb559d67b ("net/mlx5: add vectorized Rx/Tx burst for x86")
Signed-off-by: Alexander Kozyrev <akozyrev@nvidia.com>
---
drivers/net/mlx5/mlx5_rxtx_vec_altivec.h | 3 +--
drivers/net/mlx5/mlx5_rxtx_vec_neon.h | 3 +--
drivers/net/mlx5/mlx5_rxtx_vec_sse.h | 3 +--
3 files changed, 3 insertions(+), 6 deletions(-)
diff --git a/drivers/net/mlx5/mlx5_rxtx_vec_altivec.h b/drivers/net/mlx5/mlx5_rxtx_vec_altivec.h
index b2bbc4ba17..3c6a14670d 100644
--- a/drivers/net/mlx5/mlx5_rxtx_vec_altivec.h
+++ b/drivers/net/mlx5/mlx5_rxtx_vec_altivec.h
@@ -98,8 +98,7 @@ rxq_cq_decompress_v(struct mlx5_rxq_data *rxq, volatile struct mlx5_cqe *cq,
11, 10, 9, 8}; /* bswap32, rss */
/* Restore the compressed count. Must be 16 bits. */
uint16_t mcqe_n = (rxq->cqe_comp_layout) ?
- (MLX5_CQE_NUM_MINIS(cq->op_own) + 1) :
- t_pkt->data_len + (rxq->crc_present * RTE_ETHER_CRC_LEN);
+ (MLX5_CQE_NUM_MINIS(cq->op_own) + 1) : rte_be_to_cpu_32(cqe->byte_cnt)
uint16_t pkts_n = mcqe_n;
const __vector unsigned char rearm =
(__vector unsigned char)vec_vsx_ld(0,
diff --git a/drivers/net/mlx5/mlx5_rxtx_vec_neon.h b/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
index 510f60b25d..ba79b460a4 100644
--- a/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
+++ b/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
@@ -97,8 +97,7 @@ rxq_cq_decompress_v(struct mlx5_rxq_data *rxq, volatile struct mlx5_cqe *cq,
};
/* Restore the compressed count. Must be 16 bits. */
uint16_t mcqe_n = (rxq->cqe_comp_layout) ?
- (MLX5_CQE_NUM_MINIS(cq->op_own) + 1) :
- t_pkt->data_len + (rxq->crc_present * RTE_ETHER_CRC_LEN);
+ (MLX5_CQE_NUM_MINIS(cq->op_own) + 1) : rte_be_to_cpu_32(cq->byte_cnt);
uint16_t pkts_n = mcqe_n;
const uint64x2_t rearm =
vld1q_u64((void *)&t_pkt->rearm_data);
diff --git a/drivers/net/mlx5/mlx5_rxtx_vec_sse.h b/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
index e71d6c303f..bd4010f2fe 100644
--- a/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
+++ b/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
@@ -96,8 +96,7 @@ rxq_cq_decompress_v(struct mlx5_rxq_data *rxq, volatile struct mlx5_cqe *cq,
-1, -1, -1, -1 /* skip packet_type */);
/* Restore the compressed count. Must be 16 bits. */
uint16_t mcqe_n = (rxq->cqe_comp_layout) ?
- (MLX5_CQE_NUM_MINIS(cq->op_own) + 1) :
- t_pkt->data_len + (rxq->crc_present * RTE_ETHER_CRC_LEN);
+ (MLX5_CQE_NUM_MINIS(cq->op_own) + 1) : rte_be_to_cpu_32(cq->byte_cnt);
uint16_t pkts_n = mcqe_n;
const __m128i rearm =
_mm_loadu_si128((__m128i *)&t_pkt->rearm_data);
--
2.43.5
^ permalink raw reply [flat|nested] 3+ messages in thread* RE: [PATCH] net/mlx5: fix miniCQEs number calculation
2024-10-28 16:42 [PATCH] net/mlx5: fix miniCQEs number calculation Alexander Kozyrev
@ 2024-10-28 17:08 ` Slava Ovsiienko
2024-10-28 17:14 ` [PATCH v2] " Alexander Kozyrev
1 sibling, 0 replies; 3+ messages in thread
From: Slava Ovsiienko @ 2024-10-28 17:08 UTC (permalink / raw)
To: Alexander Kozyrev, dev
Cc: stable, Raslan Darawsheh, Matan Azrad, Dariusz Sosnowski,
Bing Zhao, Suanming Mou
Acked-by: Viacheslav Ovsiienko <viacheslavo@nvidia.com>
> -----Original Message-----
> From: Alexander Kozyrev <akozyrev@nvidia.com>
> Sent: Monday, October 28, 2024 6:43 PM
> To: dev@dpdk.org
> Cc: stable@dpdk.org; Raslan Darawsheh <rasland@nvidia.com>; Slava
> Ovsiienko <viacheslavo@nvidia.com>; Matan Azrad <matan@nvidia.com>;
> Dariusz Sosnowski <dsosnowski@nvidia.com>; Bing Zhao
> <bingz@nvidia.com>; Suanming Mou <suanmingm@nvidia.com>
> Subject: [PATCH] net/mlx5: fix miniCQEs number calculation
>
> Use the information from the CQE, not from the title packet, for getting the
> number of miniCQEs in the compressed CQEs array.
> This way we can avoid segfaults in the rxq_cq_decompress_v() in case of
> mbuf corruption (due to double mbuf free, for example).
>
> Fixes: 6cb559d67b ("net/mlx5: add vectorized Rx/Tx burst for x86")
>
> Signed-off-by: Alexander Kozyrev <akozyrev@nvidia.com>
> ---
> drivers/net/mlx5/mlx5_rxtx_vec_altivec.h | 3 +--
> drivers/net/mlx5/mlx5_rxtx_vec_neon.h | 3 +--
> drivers/net/mlx5/mlx5_rxtx_vec_sse.h | 3 +--
> 3 files changed, 3 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/net/mlx5/mlx5_rxtx_vec_altivec.h
> b/drivers/net/mlx5/mlx5_rxtx_vec_altivec.h
> index b2bbc4ba17..3c6a14670d 100644
> --- a/drivers/net/mlx5/mlx5_rxtx_vec_altivec.h
> +++ b/drivers/net/mlx5/mlx5_rxtx_vec_altivec.h
> @@ -98,8 +98,7 @@ rxq_cq_decompress_v(struct mlx5_rxq_data *rxq,
> volatile struct mlx5_cqe *cq,
> 11, 10, 9, 8}; /* bswap32, rss */
> /* Restore the compressed count. Must be 16 bits. */
> uint16_t mcqe_n = (rxq->cqe_comp_layout) ?
> - (MLX5_CQE_NUM_MINIS(cq->op_own) + 1) :
> - t_pkt->data_len + (rxq->crc_present *
> RTE_ETHER_CRC_LEN);
> + (MLX5_CQE_NUM_MINIS(cq->op_own) + 1) :
> +rte_be_to_cpu_32(cqe->byte_cnt)
> uint16_t pkts_n = mcqe_n;
> const __vector unsigned char rearm =
> (__vector unsigned char)vec_vsx_ld(0, diff --git
> a/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
> b/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
> index 510f60b25d..ba79b460a4 100644
> --- a/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
> +++ b/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
> @@ -97,8 +97,7 @@ rxq_cq_decompress_v(struct mlx5_rxq_data *rxq,
> volatile struct mlx5_cqe *cq,
> };
> /* Restore the compressed count. Must be 16 bits. */
> uint16_t mcqe_n = (rxq->cqe_comp_layout) ?
> - (MLX5_CQE_NUM_MINIS(cq->op_own) + 1) :
> - t_pkt->data_len + (rxq->crc_present *
> RTE_ETHER_CRC_LEN);
> + (MLX5_CQE_NUM_MINIS(cq->op_own) + 1) :
> +rte_be_to_cpu_32(cq->byte_cnt);
> uint16_t pkts_n = mcqe_n;
> const uint64x2_t rearm =
> vld1q_u64((void *)&t_pkt->rearm_data); diff --git
> a/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
> b/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
> index e71d6c303f..bd4010f2fe 100644
> --- a/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
> +++ b/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
> @@ -96,8 +96,7 @@ rxq_cq_decompress_v(struct mlx5_rxq_data *rxq,
> volatile struct mlx5_cqe *cq,
> -1, -1, -1, -1 /* skip packet_type */);
> /* Restore the compressed count. Must be 16 bits. */
> uint16_t mcqe_n = (rxq->cqe_comp_layout) ?
> - (MLX5_CQE_NUM_MINIS(cq->op_own) + 1) :
> - t_pkt->data_len + (rxq->crc_present *
> RTE_ETHER_CRC_LEN);
> + (MLX5_CQE_NUM_MINIS(cq->op_own) + 1) :
> +rte_be_to_cpu_32(cq->byte_cnt);
> uint16_t pkts_n = mcqe_n;
> const __m128i rearm =
> _mm_loadu_si128((__m128i *)&t_pkt->rearm_data);
> --
> 2.43.5
^ permalink raw reply [flat|nested] 3+ messages in thread* [PATCH v2] net/mlx5: fix miniCQEs number calculation
2024-10-28 16:42 [PATCH] net/mlx5: fix miniCQEs number calculation Alexander Kozyrev
2024-10-28 17:08 ` Slava Ovsiienko
@ 2024-10-28 17:14 ` Alexander Kozyrev
1 sibling, 0 replies; 3+ messages in thread
From: Alexander Kozyrev @ 2024-10-28 17:14 UTC (permalink / raw)
To: dev; +Cc: stable, rasland, viacheslavo, matan, dsosnowski, bingz, suanmingm
Use the information from the CQE, not from the title packet,
for getting the number of miniCQEs in the compressed CQEs array.
This way we can avoid segfaults in the rxq_cq_decompress_v()
in case of mbuf corruption (due to double mbuf free, for example).
Fixes: 6cb559d67b ("net/mlx5: add vectorized Rx/Tx burst for x86")
Signed-off-by: Alexander Kozyrev <akozyrev@nvidia.com>
---
drivers/net/mlx5/mlx5_rxtx_vec_altivec.h | 3 +--
drivers/net/mlx5/mlx5_rxtx_vec_neon.h | 3 +--
drivers/net/mlx5/mlx5_rxtx_vec_sse.h | 3 +--
3 files changed, 3 insertions(+), 6 deletions(-)
diff --git a/drivers/net/mlx5/mlx5_rxtx_vec_altivec.h b/drivers/net/mlx5/mlx5_rxtx_vec_altivec.h
index b2bbc4ba17..265c81d8dc 100644
--- a/drivers/net/mlx5/mlx5_rxtx_vec_altivec.h
+++ b/drivers/net/mlx5/mlx5_rxtx_vec_altivec.h
@@ -98,8 +98,7 @@ rxq_cq_decompress_v(struct mlx5_rxq_data *rxq, volatile struct mlx5_cqe *cq,
11, 10, 9, 8}; /* bswap32, rss */
/* Restore the compressed count. Must be 16 bits. */
uint16_t mcqe_n = (rxq->cqe_comp_layout) ?
- (MLX5_CQE_NUM_MINIS(cq->op_own) + 1) :
- t_pkt->data_len + (rxq->crc_present * RTE_ETHER_CRC_LEN);
+ (MLX5_CQE_NUM_MINIS(cq->op_own) + 1U) : rte_be_to_cpu_32(cqe->byte_cnt)
uint16_t pkts_n = mcqe_n;
const __vector unsigned char rearm =
(__vector unsigned char)vec_vsx_ld(0,
diff --git a/drivers/net/mlx5/mlx5_rxtx_vec_neon.h b/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
index 510f60b25d..a4d4f3b582 100644
--- a/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
+++ b/drivers/net/mlx5/mlx5_rxtx_vec_neon.h
@@ -97,8 +97,7 @@ rxq_cq_decompress_v(struct mlx5_rxq_data *rxq, volatile struct mlx5_cqe *cq,
};
/* Restore the compressed count. Must be 16 bits. */
uint16_t mcqe_n = (rxq->cqe_comp_layout) ?
- (MLX5_CQE_NUM_MINIS(cq->op_own) + 1) :
- t_pkt->data_len + (rxq->crc_present * RTE_ETHER_CRC_LEN);
+ (MLX5_CQE_NUM_MINIS(cq->op_own) + 1U) : rte_be_to_cpu_32(cq->byte_cnt);
uint16_t pkts_n = mcqe_n;
const uint64x2_t rearm =
vld1q_u64((void *)&t_pkt->rearm_data);
diff --git a/drivers/net/mlx5/mlx5_rxtx_vec_sse.h b/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
index e71d6c303f..0273fdb54a 100644
--- a/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
+++ b/drivers/net/mlx5/mlx5_rxtx_vec_sse.h
@@ -96,8 +96,7 @@ rxq_cq_decompress_v(struct mlx5_rxq_data *rxq, volatile struct mlx5_cqe *cq,
-1, -1, -1, -1 /* skip packet_type */);
/* Restore the compressed count. Must be 16 bits. */
uint16_t mcqe_n = (rxq->cqe_comp_layout) ?
- (MLX5_CQE_NUM_MINIS(cq->op_own) + 1) :
- t_pkt->data_len + (rxq->crc_present * RTE_ETHER_CRC_LEN);
+ (MLX5_CQE_NUM_MINIS(cq->op_own) + 1U) : rte_be_to_cpu_32(cq->byte_cnt);
uint16_t pkts_n = mcqe_n;
const __m128i rearm =
_mm_loadu_si128((__m128i *)&t_pkt->rearm_data);
--
2.43.5
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2024-10-28 17:16 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-10-28 16:42 [PATCH] net/mlx5: fix miniCQEs number calculation Alexander Kozyrev
2024-10-28 17:08 ` Slava Ovsiienko
2024-10-28 17:14 ` [PATCH v2] " Alexander Kozyrev
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).