patch 'net/af_packet: fix socket close on device stop' has been queued to stable release 23.11.4

[Xueming Li <xuemingl@nvidia.com>]

Hi,

FYI, your patch has been queued to stable release 23.11.4

Note it hasn't been pushed to http://dpdk.org/browse/dpdk-stable yet.
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=6594bba195d9047e43e30d22170baab5f103d08a

Thanks.

Xueming Li <xuemingl@nvidia.com>

---
From 6594bba195d9047e43e30d22170baab5f103d08a Mon Sep 17 00:00:00 2001
From: Tudor Cornea <tudor.cornea@gmail.com>
Date: Tue, 4 Feb 2025 18:45:08 +0200
Subject: [PATCH] net/af_packet: fix socket close on device stop
Cc: Xueming Li <xuemingl@nvidia.com>

[ upstream commit 872e846f6bb31afbdd508903380cf05a3d313a3a ]

Currently, if we call rte_eth_dev_stop(), the sockets are closed.
If we attempt to start the port again, socket related operations
will not work correctly.

This can be alleviated by closing the socket at the same place in
which we currently free the memory, in eth_dev_close().

If an application calls rte_eth_dev_stop() on a port managed
by the af_packet PMD, the port becomes unusable. This is in contrast
with ports managed by other drivers (e.g virtio).

I also managed to reproduce the issue using testpmd.

sudo ip link add test-veth0 type veth peer name test-veth1

sudo ip link set test-veth0 up
sudo ip link set test-veth1 up

AF_PACKET_ARGS=\
"blocksz=4096,framesz=2048,framecnt=512,qpairs=1,qdisc_bypass=0"

sudo ./dpdk-testpmd \
        -l 0-3 \
        -m 1024 \
        --no-huge \
        --no-shconf \
        --no-pci \
        --vdev=net_af_packet0,iface=test-veth0,${AF_PACKET_ARGS} \
        --vdev=net_af_packet1,iface=test-veth1,${AF_PACKET_ARGS} \
        -- \
        -i

testpmd> start tx_first

Forwarding will start, and we will see traffic on the interfaces.

testpmd> stop
testpmd> port stop 0
Stopping ports...
Checking link statuses...
Done
testpmd> port stop 1
Stopping ports...
Checking link statuses...
Done

testpmd> port start 0
AFPACKET: eth_dev_macaddr_set(): receive socket not found
Port 0: CA:65:81:63:81:B2
Checking link statuses...
Done
testpmd> port start 1
AFPACKET: eth_dev_macaddr_set(): receive socket not found
Port 1: CA:12:D0:BE:93:3F
Checking link statuses...
Done

testpmd> start tx_first

When we start forwarding again, we can see that there is no traffic
on the interfaces. This does not happen when testing with other PMD
drivers (e.g virtio).

With the patch, the port should re-initialize correctly.

testpmd> port start 0
Port 0: CA:65:81:63:81:B2
Checking link statuses...
Done

Fixes: 364e08f2bbc0 ("af_packet: add PMD for AF_PACKET-based virtual devices")

Signed-off-by: Tudor Cornea <tudor.cornea@gmail.com>
---
 drivers/net/af_packet/rte_eth_af_packet.c | 30 +++++++++++------------
 1 file changed, 15 insertions(+), 15 deletions(-)

diff --git a/drivers/net/af_packet/rte_eth_af_packet.c b/drivers/net/af_packet/rte_eth_af_packet.c
index 6b7b16f348..0b059bfd0b 100644
--- a/drivers/net/af_packet/rte_eth_af_packet.c
+++ b/drivers/net/af_packet/rte_eth_af_packet.c
@@ -332,27 +332,12 @@ static int
 eth_dev_stop(struct rte_eth_dev *dev)
 {
 	unsigned i;
-	int sockfd;
 	struct pmd_internals *internals = dev->data->dev_private;

 	for (i = 0; i < internals->nb_queues; i++) {
-		sockfd = internals->rx_queue[i].sockfd;
-		if (sockfd != -1)
-			close(sockfd);
-
-		/* Prevent use after free in case tx fd == rx fd */
-		if (sockfd != internals->tx_queue[i].sockfd) {
-			sockfd = internals->tx_queue[i].sockfd;
-			if (sockfd != -1)
-				close(sockfd);
-		}
-
-		internals->rx_queue[i].sockfd = -1;
-		internals->tx_queue[i].sockfd = -1;
 		dev->data->rx_queue_state[i] = RTE_ETH_QUEUE_STATE_STOPPED;
 		dev->data->tx_queue_state[i] = RTE_ETH_QUEUE_STATE_STOPPED;
 	}
-
 	dev->data->dev_link.link_status = RTE_ETH_LINK_DOWN;
 	return 0;
 }
@@ -447,6 +432,7 @@ eth_dev_close(struct rte_eth_dev *dev)
 	struct pmd_internals *internals;
 	struct tpacket_req *req;
 	unsigned int q;
+	int sockfd;

 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
 		return 0;
@@ -457,6 +443,20 @@ eth_dev_close(struct rte_eth_dev *dev)
 	internals = dev->data->dev_private;
 	req = &internals->req;
 	for (q = 0; q < internals->nb_queues; q++) {
+		sockfd = internals->rx_queue[q].sockfd;
+		if (sockfd != -1)
+			close(sockfd);
+
+		/* Prevent use after free in case tx fd == rx fd */
+		if (sockfd != internals->tx_queue[q].sockfd) {
+			sockfd = internals->tx_queue[q].sockfd;
+			if (sockfd != -1)
+				close(sockfd);
+		}
+
+		internals->rx_queue[q].sockfd = -1;
+		internals->tx_queue[q].sockfd = -1;
+
 		munmap(internals->rx_queue[q].map,
 			2 * req->tp_block_size * req->tp_block_nr);
 		rte_free(internals->rx_queue[q].rd);
--
2.34.1

---
  Diff of the applied patch vs upstream commit (please double-check if non-empty:
---
--- -	2025-02-18 19:39:01.988448413 +0800
+++ 0035-net-af_packet-fix-socket-close-on-device-stop.patch	2025-02-18 19:39:00.528244065 +0800
@@ -1 +1 @@
-From 872e846f6bb31afbdd508903380cf05a3d313a3a Mon Sep 17 00:00:00 2001
+From 6594bba195d9047e43e30d22170baab5f103d08a Mon Sep 17 00:00:00 2001
@@ -4,0 +5,3 @@
+Cc: Xueming Li <xuemingl@nvidia.com>
+
+[ upstream commit 872e846f6bb31afbdd508903380cf05a3d313a3a ]
@@ -77 +79,0 @@
-Cc: stable@dpdk.org
@@ -85 +87 @@
-index 7fcd27ee0c..ab34285ed9 100644
+index 6b7b16f348..0b059bfd0b 100644
@@ -88 +90 @@
-@@ -363,27 +363,12 @@ static int
+@@ -332,27 +332,12 @@ static int
@@ -116 +118 @@
-@@ -510,6 +495,7 @@ eth_dev_close(struct rte_eth_dev *dev)
+@@ -447,6 +432,7 @@ eth_dev_close(struct rte_eth_dev *dev)
@@ -124 +126 @@
-@@ -520,6 +506,20 @@ eth_dev_close(struct rte_eth_dev *dev)
+@@ -457,6 +443,20 @@ eth_dev_close(struct rte_eth_dev *dev)