patch 'net/ark: fix index arithmetic' has been queued to stable release 23.11.2

[Xueming Li]

Hi,

FYI, your patch has been queued to stable release 23.11.2

Note it hasn't been pushed to http://dpdk.org/browse/dpdk-stable yet.
It will be pushed if I get no objections before 08/14/24. So please
shout if anyone has objections.

Also note that after the patch there's a diff of the upstream commit vs the
patch applied to the branch. This will indicate if there was any rebasing
needed to apply to the stable branch. If there were code changes for rebasing
(ie: not only metadata diffs), please double check that the rebase was
correctly done.

Queued patches are on a temporary branch at:
https://git.dpdk.org/dpdk-stable/log/?h=23.11-staging

This queued commit can be viewed at:
https://git.dpdk.org/dpdk-stable/commit/?h=23.11-staging&id=e2b3b0a5d4cee408f6d1638010669df467e7398b

Thanks.

Xueming Li <xuemingl at nvidia.com>

---
>From e2b3b0a5d4cee408f6d1638010669df467e7398b Mon Sep 17 00:00:00 2001
From: Ed Czeck <ed.czeck at atomicrules.com>
Date: Wed, 17 Jul 2024 16:38:30 -0400
Subject: [PATCH] net/ark: fix index arithmetic
Cc: Xueming Li <xuemingl at nvidia.com>

[ upstream commit 73c0e26c8ed9f59f867ff251fbb2a50c8dfa06c5 ]

Behavior for signed integer overflow is not defined which can causes
undesired behavior at values near max and min bounds.
The used of unsigned is defined as to use modulo arithmetic which is the
desired behavior.
This patch replaces int32_t with uint32_t except for necessary casts.

Fixes: 9ee9e0d3b85e ("net/ark: update to reflect FPGA updates")

Signed-off-by: Ed Czeck <ed.czeck at atomicrules.com>
Acked-by: Ferruh Yigit <ferruh.yigit at amd.com>
---
 drivers/net/ark/ark_ethdev_tx.c | 24 +++++++++++++++---------
 1 file changed, 15 insertions(+), 9 deletions(-)

diff --git a/drivers/net/ark/ark_ethdev_tx.c b/drivers/net/ark/ark_ethdev_tx.c
index 4792754f19..8f1f90b1a4 100644
--- a/drivers/net/ark/ark_ethdev_tx.c
+++ b/drivers/net/ark/ark_ethdev_tx.c
@@ -39,8 +39,8 @@ struct ark_tx_queue {
 	uint32_t queue_mask;
 
 	/* 3 indexes to the paired data rings. */
-	int32_t prod_index;		/* where to put the next one */
-	int32_t free_index;		/* mbuf has been freed */
+	uint32_t prod_index;		/* where to put the next one */
+	uint32_t free_index;		/* mbuf has been freed */
 
 	/* The queue Id is used to identify the HW Q */
 	uint16_t phys_qid;
@@ -49,7 +49,7 @@ struct ark_tx_queue {
 
 	/* next cache line - fields written by device */
 	RTE_MARKER cacheline1 __rte_cache_min_aligned;
-	volatile int32_t cons_index;		/* hw is done, can be freed */
+	volatile uint32_t cons_index;		/* hw is done, can be freed */
 } __rte_cache_aligned;
 
 /* Forward declarations */
@@ -108,7 +108,7 @@ eth_ark_xmit_pkts(void *vtxq, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 	uint32_t user_meta[5];
 
 	int stat;
-	int32_t prod_index_limit;
+	uint32_t prod_index_limit;
 	uint16_t nb;
 	uint8_t user_len = 0;
 	const uint32_t min_pkt_len = ARK_MIN_TX_PKTLEN;
@@ -123,8 +123,13 @@ eth_ark_xmit_pkts(void *vtxq, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 	/* leave 4 elements mpu data */
 	prod_index_limit = queue->queue_size + queue->free_index - 4;
 
+	/* Populate the buffer bringing prod_index up to or slightly beyond
+	 * prod_index_limit. Prod_index will increment by 2 or more each
+	 * iteration.  Note: indexes are uint32_t, cast to (signed) int32_t
+	 * to catch the slight overage case;  e.g. (200 - 201)
+	 */
 	for (nb = 0;
-	     (nb < nb_pkts) && (prod_index_limit - queue->prod_index) > 0;
+	     (nb < nb_pkts) && (int32_t)(prod_index_limit - queue->prod_index) > 0;
 	     ++nb) {
 		mbuf = tx_pkts[nb];
 
@@ -194,13 +199,13 @@ eth_ark_tx_jumbo(struct ark_tx_queue *queue, struct rte_mbuf *mbuf,
 		 uint32_t *user_meta, uint8_t meta_cnt)
 {
 	struct rte_mbuf *next;
-	int32_t free_queue_space;
+	uint32_t free_queue_space;
 	uint8_t flags = ARK_DDM_SOP;
 
 	free_queue_space = queue->queue_mask -
 		(queue->prod_index - queue->free_index);
 	/* We need up to 4 mbufs for first header and 2 for subsequent ones */
-	if (unlikely(free_queue_space < (2 + (2 * mbuf->nb_segs))))
+	if (unlikely(free_queue_space < (2U + (2U * mbuf->nb_segs))))
 		return -1;
 
 	while (mbuf != NULL) {
@@ -392,10 +397,11 @@ free_completed_tx(struct ark_tx_queue *queue)
 {
 	struct rte_mbuf *mbuf;
 	union ark_tx_meta *meta;
-	int32_t top_index;
+	uint32_t top_index;
 
 	top_index = queue->cons_index;	/* read once */
-	while ((top_index - queue->free_index) > 0) {
+
+	while ((int32_t)(top_index - queue->free_index) > 0) {
 		meta = &queue->meta_q[queue->free_index & queue->queue_mask];
 		if (likely((meta->flags & ARK_DDM_SOP) != 0)) {
 			mbuf = queue->bufs[queue->free_index &
-- 
2.34.1

---
  Diff of the applied patch vs upstream commit (please double-check if non-empty:
---
--- -	2024-08-12 20:44:06.642847471 +0800
+++ 0130-net-ark-fix-index-arithmetic.patch	2024-08-12 20:44:02.525069379 +0800
@@ -1 +1 @@
-From 73c0e26c8ed9f59f867ff251fbb2a50c8dfa06c5 Mon Sep 17 00:00:00 2001
+From e2b3b0a5d4cee408f6d1638010669df467e7398b Mon Sep 17 00:00:00 2001
@@ -4,0 +5,3 @@
+Cc: Xueming Li <xuemingl at nvidia.com>
+
+[ upstream commit 73c0e26c8ed9f59f867ff251fbb2a50c8dfa06c5 ]
@@ -13 +15,0 @@
-Cc: stable at dpdk.org
@@ -22 +24 @@
-index 9c89c85f50..ca6cd297a1 100644
+index 4792754f19..8f1f90b1a4 100644
@@ -25 +27 @@
-@@ -39,8 +39,8 @@ struct __rte_cache_aligned ark_tx_queue {
+@@ -39,8 +39,8 @@ struct ark_tx_queue {
@@ -36 +38 @@
-@@ -49,7 +49,7 @@ struct __rte_cache_aligned ark_tx_queue {
+@@ -49,7 +49,7 @@ struct ark_tx_queue {
@@ -39 +41 @@
- 	alignas(RTE_CACHE_LINE_MIN_SIZE) RTE_MARKER cacheline1;
+ 	RTE_MARKER cacheline1 __rte_cache_min_aligned;
@@ -42 +44 @@
- };
+ } __rte_cache_aligned;