net/mlx5/hws: fix return value of send queue action

PATCH In review Alex Vesker 2 participants 6 messages Mar 23, 2023 Network drivers

Share & tools

Alex Vesker

author

2023-03-23 12:34 UTC #0 lore raw

PATCH 1/1 net/mlx5/hws: fix return value of send queue action

The rte_errno should be set to a positive error value while the ret of the function should return a negative value this aligns code to other mlx5dr API functions.

Fixes 3eb748869d2d ("net/mlx5/hws: add send layer")

Cc [email protected]

Signed-off-by Alex Vesker <[email protected]>

Reviewed-by Erez Shitrit <[email protected]>

Acked-by Matan Azrad [email protected]

---
 drivers/net/mlx5/hws/mlx5dr_send.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

drivers/net/mlx5/hws/mlx5dr_send.c +2 -2

index 51aaf5c8e2..d650c55124 100644
@@ -1007,8 +1007,8 @@ int mlx5dr_send_queue_action(struct mlx5dr_context *ctx,
 
 		break;
 	default:
-		rte_errno = -EINVAL;
-		return rte_errno;
+		rte_errno = EINVAL;
+		return -rte_errno;
 	}
 
 	return 0;

Signature · 1 line

2.18.1

Alex Vesker

author

2023-03-23 12:34 UTC #1 lore raw

PATCH 1/1 net/mlx5/hws: fix send sync drain empty queue check

The function that checks if the queue is empty used on queue action for SYNC and ASYNC drain didn't function correctly since cur_post is a free running value and not cyclic. The fix is bitwise AND cur_post to get the real value.

Fixes 90488887ee33 ("net/mlx5/hws: support synchronous drain")

Signed-off-by Alex Vesker <[email protected]>

Reviewed-by Erez Shitrit <[email protected]>

Acked-by Matan Azrad [email protected]

---
 drivers/net/mlx5/hws/mlx5dr_send.h | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

drivers/net/mlx5/hws/mlx5dr_send.h +4 -1

index d0977ec851..c1e8616f7e 100644
@@ -255,7 +255,10 @@ void mlx5dr_send_engine_flush_queue(struct mlx5dr_send_engine *queue);
 
 static inline bool mlx5dr_send_engine_empty(struct mlx5dr_send_engine *queue)
 {
-	return (queue->send_ring->send_sq.cur_post == queue->send_ring->send_cq.poll_wqe);
+	struct mlx5dr_send_ring_sq *send_sq = &queue->send_ring->send_sq;
+	struct mlx5dr_send_ring_cq *send_cq = &queue->send_ring->send_cq;
+
+	return ((send_sq->cur_post & send_sq->buf_mask) == send_cq->poll_wqe);
 }
 
 static inline bool mlx5dr_send_engine_full(struct mlx5dr_send_engine *queue)

Signature · 1 line

2.18.1

Raslan Darawsheh

2023-03-23 19:43 UTC #2 lore raw

RE: [PATCH] net/mlx5/hws: fix send sync drain empty queue check

Hi,

Quoted 15 lines

-----Original Message-----
From: Alex Vesker <[email protected]>
Sent: Thursday, March 23, 2023 2:34 PM
To: Alex Vesker <[email protected]>; Slava Ovsiienko <[email protected]>; NBU-Contact-Thomas Monjalon (EXTERNAL) <[email protected]>; Suanming Mou <[email protected]>; Matan Azrad <[email protected]>
Cc: [email protected]; Ori Kam <[email protected]>
Subject: [PATCH] net/mlx5/hws: fix send sync drain empty queue check
The function that checks if the queue is empty used on queue action for SYNC and ASYNC drain didn't function correctly since cur_post is a free running value and not cyclic.
The fix is bitwise AND cur_post to get the real value.

Quoted 4 lines

Fixes 90488887ee33 ("net/mlx5/hws: support synchronous drain")
Signed-off-by Alex Vesker <[email protected]>
Reviewed-by Erez Shitrit <[email protected]>
Acked-by Matan Azrad [email protected]

Fixed Acked-by tag,
Patch applied to next-net-mlx,

Kindest regards
Raslan Darawsheh

Alex Vesker

author

2023-03-23 12:35 UTC #3 lore raw

PATCH 1/1 net/mlx5/hws: fix IPv4 frag matching

In reply to Alex Vesker

From: Hamdan Igbaria <[email protected]>

Fix IPv4 frag matching in case fragment_offset field is set to non zero value in the mask.
fragment_offset value is converted using the following logic:
-In case fragment_offset value was set to 0x3fff, then we will
match only on ip_fragmented bit.
-Otherwise we will match fragment_offset based on spec and last
same as any other field.

Fixes c55c2bf35333 ("net/mlx5/hws: add definer layer")

Cc [email protected]

Signed-off-by Hamdan Igbaria <[email protected]>

Reviewed-by Alex Vesker <[email protected]>

Acked-by Matan Azrad [email protected]

---
 drivers/net/mlx5/hws/mlx5dr_definer.c | 11 +++++++++--
 drivers/net/mlx5/hws/mlx5dr_definer.h |  9 +++++++--
 2 files changed, 16 insertions(+), 4 deletions(-)

drivers/net/mlx5/hws/mlx5dr_definer.c +9 -2

index 379f4065e7..f92d3e8e1f 100644
@@ -127,6 +127,7 @@ struct mlx5dr_definer_conv_data {
 	X(SET,		ipv4_version,		STE_IPV4,		rte_ipv4_hdr) \
 	X(SET_BE16,	ipv4_frag,		v->fragment_offset,	rte_ipv4_hdr) \
 	X(SET_BE16,	ipv4_len,		v->total_length,	rte_ipv4_hdr) \
+	X(SET,          ip_fragmented,          !!v->fragment_offset,   rte_ipv4_hdr) \
 	X(SET_BE16,	ipv6_payload_len,	v->hdr.payload_len,	rte_flow_item_ipv6) \
 	X(SET,		ipv6_proto,		v->hdr.proto,		rte_flow_item_ipv6) \
 	X(SET,		ipv6_routing_hdr,	IPPROTO_ROUTING,	rte_flow_item_ipv6) \

@@ -735,8 +736,14 @@ mlx5dr_definer_conv_item_ipv4(struct mlx5dr_definer_conv_data *cd,
 	if (m->fragment_offset) {
 		fc = &cd->fc[DR_CALC_FNAME(IP_FRAG, inner)];
 		fc->item_idx = item_idx;
-		fc->tag_set = &mlx5dr_definer_ipv4_frag_set;
-		DR_CALC_SET(fc, eth_l3, fragment_offset, inner);
+		if (rte_be_to_cpu_16(m->fragment_offset) == 0x3fff) {
+			fc->tag_set = &mlx5dr_definer_ip_fragmented_set;
+			DR_CALC_SET(fc, eth_l2, ip_fragmented, inner);
+		} else {
+			fc->is_range = l && l->fragment_offset;
+			fc->tag_set = &mlx5dr_definer_ipv4_frag_set;
+			DR_CALC_SET(fc, eth_l3, ipv4_frag, inner);
+		}
 	}
 
 	if (m->next_proto_id) {

drivers/net/mlx5/hws/mlx5dr_definer.h +7 -2

index 0cd83db756..90ec4ce845 100644
@@ -226,8 +226,13 @@ struct mlx5_ifc_definer_hl_eth_l3_bits {
 	u8 time_to_live_hop_limit[0x8];
 	u8 protocol_next_header[0x8];
 	u8 identification[0x10];
-	u8 flags[0x3];
-	u8 fragment_offset[0xd];
+	union {
+		u8 ipv4_frag[0x10];
+		struct {
+			u8 flags[0x3];
+			u8 fragment_offset[0xd];
+		};
+	};
 	u8 ipv4_total_length[0x10];
 	u8 checksum[0x10];
 	u8 reserved_at_60[0xc];

Signature · 1 line

2.18.1

Raslan Darawsheh

2023-03-23 19:43 UTC #4 lore raw

RE: [PATCH] net/mlx5/hws: fix IPv4 frag matching

Hi,

Quoted 21 lines

-----Original Message-----
From: Alex Vesker <[email protected]>
Sent: Thursday, March 23, 2023 2:34 PM
To: Alex Vesker <[email protected]>; Slava Ovsiienko <[email protected]>; NBU-Contact-Thomas Monjalon (EXTERNAL) <[email protected]>; Suanming Mou <[email protected]>; Matan Azrad <[email protected]>
Cc: [email protected]; Ori Kam <[email protected]>; Hamdan Igbaria <[email protected]>; [email protected]
Subject: [PATCH] net/mlx5/hws: fix IPv4 frag matching
From: Hamdan Igbaria <[email protected]>
Fix IPv4 frag matching in case fragment_offset field is set to non zero value in the mask.
fragment_offset value is converted using the following logic:
-In case fragment_offset value was set to 0x3fff, then we will match only on ip_fragmented bit.
-Otherwise we will match fragment_offset based on spec and last same as any other field.

Quoted 6 lines

Fixes c55c2bf35333 ("net/mlx5/hws: add definer layer")
Cc [email protected]
Signed-off-by Hamdan Igbaria <[email protected]>
Reviewed-by Alex Vesker <[email protected]>
Acked-by Matan Azrad [email protected]

Fixed Acked-by tag,
Patch applied to next-net-mlx,

Kindest regards,
Raslan Darawsheh

Raslan Darawsheh

2023-03-23 19:42 UTC #5 lore raw

In reply to Alex Vesker

Hi,

Quoted 14 lines

-----Original Message-----
From: Alex Vesker <[email protected]>
Sent: Thursday, March 23, 2023 2:34 PM
To: Alex Vesker <[email protected]>; Slava Ovsiienko <[email protected]>; NBU-Contact-Thomas Monjalon (EXTERNAL) <[email protected]>; Suanming Mou <[email protected]>; Matan Azrad <[email protected]>
Cc: [email protected]; Ori Kam <[email protected]>; [email protected]
Subject: [PATCH] net/mlx5/hws: fix return value of send queue action
The rte_errno should be set to a positive error value while the ret of the function should return a negative value this aligns code to other mlx5dr API functions.

Quoted 5 lines

Fixes 3eb748869d2d ("net/mlx5/hws: add send layer")
Cc [email protected]
Signed-off-by Alex Vesker <[email protected]>
Reviewed-by Erez Shitrit <[email protected]>
Acked-by Matan Azrad [email protected]

Fixed Acked-by tag,
Patch applied to next-net-mlx,

Kindest regards,
Raslan Darawsheh