[PATCH net-next v3] netdevsim: implement peer queue flow control

[PATCH net-next v3] netdevsim: implement peer queue flow control

[Breno Leitao]

Add flow control mechanism between paired netdevsim devices to stop the
TX queue during high traffic scenarios. When a receive queue becomes
congested (approaching NSIM_RING_SIZE limit), the corresponding transmit
queue on the peer device is stopped using netif_subqueue_try_stop().

Once the receive queue has sufficient capacity again, the peer's
transmit queue is resumed with netif_tx_wake_queue().

Key changes:
  * Add nsim_stop_peer_tx_queue() to pause peer TX when RX queue is full
  * Add nsim_start_peer_tx_queue() to resume peer TX when RX queue drains
  * Implement queue mapping validation to ensure TX/RX queue count match
  * Wake all queues during device unlinking to prevent stuck queues
  * Use RCU protection when accessing peer device references
  * wake the queues when changing the queue numbers
  * Remove IFF_NO_QUEUE given it will enqueue packets now

The flow control only activates when devices have matching TX/RX queue
counts to ensure proper queue mapping.

Suggested-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Breno Leitao <leitao@debian.org>
---
Changes in v3:
- Wake the queues if we need to in nsim_set_channels() (Jakub)
- Remove IFF_NO_QUEUE (Jakub)
- Improve function parameters (Jakub) 
- Link to v2: https://lore.kernel.org/r/20250703-netdev_flow_control-v2-1-ab00341c9cc1@debian.org

Changes in v2:
- Move the RCU locks inside the function (David)
- Use better helpers for waking up all the queues (Jakub)
- Link to v1: https://lore.kernel.org/r/20250701-netdev_flow_control-v1-1-240329fc91b1@debian.org
---
 drivers/net/netdevsim/bus.c     |  3 ++
 drivers/net/netdevsim/ethtool.c | 21 +++++++++++++
 drivers/net/netdevsim/netdev.c  | 68 ++++++++++++++++++++++++++++++++++++-----
 3 files changed, 85 insertions(+), 7 deletions(-)

diff --git a/drivers/net/netdevsim/bus.c b/drivers/net/netdevsim/bus.c
index 64c0cdd31bf85..1ba52471f3fbc 100644
--- a/drivers/net/netdevsim/bus.c
+++ b/drivers/net/netdevsim/bus.c
@@ -366,6 +366,9 @@ static ssize_t unlink_device_store(const struct bus_type *bus, const char *buf,
 	err = 0;
 	RCU_INIT_POINTER(nsim->peer, NULL);
 	RCU_INIT_POINTER(peer->peer, NULL);
+	synchronize_net();
+	netif_tx_wake_all_queues(dev);
+	netif_tx_wake_all_queues(peer->netdev);
 
 out_put_netns:
 	put_net(ns);
diff --git a/drivers/net/netdevsim/ethtool.c b/drivers/net/netdevsim/ethtool.c
index 4d191a3293c74..f631d90c428ac 100644
--- a/drivers/net/netdevsim/ethtool.c
+++ b/drivers/net/netdevsim/ethtool.c
@@ -101,6 +101,22 @@ nsim_get_channels(struct net_device *dev, struct ethtool_channels *ch)
 	ch->combined_count = ns->ethtool.channels;
 }
 
+static void
+nsim_wake_queues(struct net_device *dev)
+{
+	struct netdevsim *ns = netdev_priv(dev);
+	struct netdevsim *peer;
+
+	synchronize_net();
+	netif_tx_wake_all_queues(dev);
+
+	rcu_read_lock();
+	peer = rcu_dereference(ns->peer);
+	if (peer)
+		netif_tx_wake_all_queues(peer->netdev);
+	rcu_read_unlock();
+}
+
 static int
 nsim_set_channels(struct net_device *dev, struct ethtool_channels *ch)
 {
@@ -113,6 +129,11 @@ nsim_set_channels(struct net_device *dev, struct ethtool_channels *ch)
 		return err;
 
 	ns->ethtool.channels = ch->combined_count;
+
+	/* Only wake up queues if devices are linked */
+	if (rcu_access_pointer(ns->peer))
+		nsim_wake_queues(dev);
+
 	return 0;
 }
 
diff --git a/drivers/net/netdevsim/netdev.c b/drivers/net/netdevsim/netdev.c
index e36d3e846c2dc..48da5d4b3a89a 100644
--- a/drivers/net/netdevsim/netdev.c
+++ b/drivers/net/netdevsim/netdev.c
@@ -37,7 +37,53 @@ MODULE_IMPORT_NS("NETDEV_INTERNAL");
 
 #define NSIM_RING_SIZE		256
 
-static int nsim_napi_rx(struct nsim_rq *rq, struct sk_buff *skb)
+static void nsim_start_peer_tx_queue(struct net_device *dev, struct nsim_rq *rq)
+{
+	struct netdevsim *ns = netdev_priv(dev);
+	struct net_device *peer_dev;
+	struct netdevsim *peer_ns;
+	struct netdev_queue *txq;
+	u16 idx;
+
+	idx = rq->napi.index;
+	rcu_read_lock();
+	peer_ns = rcu_dereference(ns->peer);
+	if (!peer_ns)
+		goto out;
+
+	/* TX device */
+	peer_dev = peer_ns->netdev;
+	if (dev->real_num_tx_queues != peer_dev->num_rx_queues)
+		goto out;
+
+	txq = netdev_get_tx_queue(peer_dev, idx);
+	if (!(netif_tx_queue_stopped(txq)))
+		goto out;
+
+	netif_tx_wake_queue(txq);
+out:
+	rcu_read_unlock();
+}
+
+static void nsim_stop_tx_queue(struct net_device *tx_dev,
+			       struct net_device *rx_dev,
+			       struct nsim_rq *rq,
+			       u16 idx)
+{
+	/* If different queues size, do not stop, since it is not
+	 * easy to find which TX queue is mapped here
+	 */
+	if (rx_dev->real_num_tx_queues != tx_dev->num_rx_queues)
+		return;
+
+	/* rq is the queue on the receive side */
+	netif_subqueue_try_stop(tx_dev, idx,
+				NSIM_RING_SIZE - skb_queue_len(&rq->skb_queue),
+				NSIM_RING_SIZE / 2);
+}
+
+static int nsim_napi_rx(struct net_device *tx_dev, struct net_device *rx_dev,
+			struct nsim_rq *rq, struct sk_buff *skb)
 {
 	if (skb_queue_len(&rq->skb_queue) > NSIM_RING_SIZE) {
 		dev_kfree_skb_any(skb);
@@ -45,13 +91,22 @@ static int nsim_napi_rx(struct nsim_rq *rq, struct sk_buff *skb)
 	}
 
 	skb_queue_tail(&rq->skb_queue, skb);
+
+	/* Stop the peer TX queue avoiding dropping packets later */
+	if (skb_queue_len(&rq->skb_queue) >= NSIM_RING_SIZE)
+		nsim_stop_tx_queue(tx_dev, rx_dev, rq,
+				   skb_get_queue_mapping(skb));
+
 	return NET_RX_SUCCESS;
 }
 
-static int nsim_forward_skb(struct net_device *dev, struct sk_buff *skb,
+static int nsim_forward_skb(struct net_device *tx_dev,
+			    struct net_device *rx_dev,
+			    struct sk_buff *skb,
 			    struct nsim_rq *rq)
 {
-	return __dev_forward_skb(dev, skb) ?: nsim_napi_rx(rq, skb);
+	return __dev_forward_skb(rx_dev, skb) ?:
+		nsim_napi_rx(tx_dev, rx_dev, rq, skb);
 }
 
 static netdev_tx_t nsim_start_xmit(struct sk_buff *skb, struct net_device *dev)
@@ -86,7 +141,7 @@ static netdev_tx_t nsim_start_xmit(struct sk_buff *skb, struct net_device *dev)
 		skb_linearize(skb);
 
 	skb_tx_timestamp(skb);
-	if (unlikely(nsim_forward_skb(peer_dev, skb, rq) == NET_RX_DROP))
+	if (unlikely(nsim_forward_skb(dev, peer_dev, skb, rq) == NET_RX_DROP))
 		goto out_drop_cnt;
 
 	if (!hrtimer_active(&rq->napi_timer))
@@ -351,6 +406,7 @@ static int nsim_rcv(struct nsim_rq *rq, int budget)
 			dev_dstats_rx_dropped(dev);
 	}
 
+	nsim_start_peer_tx_queue(dev, rq);
 	return i;
 }
 
@@ -864,10 +920,8 @@ static void nsim_setup(struct net_device *dev)
 	ether_setup(dev);
 	eth_hw_addr_random(dev);
 
-	dev->tx_queue_len = 0;
 	dev->flags &= ~IFF_MULTICAST;
-	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE |
-			   IFF_NO_QUEUE;
+	dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
 	dev->features |= NETIF_F_HIGHDMA |
 			 NETIF_F_SG |
 			 NETIF_F_FRAGLIST |

---
base-commit: 5fa1fa9184ef1af929ae49c715d4a25667e84df5
change-id: 20250630-netdev_flow_control-2b2d37965377

Best regards,
--  
Breno Leitao <leitao@debian.org>

Re: [PATCH net-next v3] netdevsim: implement peer queue flow control

[Jakub Kicinski]

On Fri, 11 Jul 2025 10:06:59 -0700 Breno Leitao wrote:
> +	if (!(netif_tx_queue_stopped(txq)))

Unnecessary parenthesis here, will drop when applying.

Re: [PATCH net-next v3] netdevsim: implement peer queue flow control

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to netdev/net-next.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Fri, 11 Jul 2025 10:06:59 -0700 you wrote:
> Add flow control mechanism between paired netdevsim devices to stop the
> TX queue during high traffic scenarios. When a receive queue becomes
> congested (approaching NSIM_RING_SIZE limit), the corresponding transmit
> queue on the peer device is stopped using netif_subqueue_try_stop().
> 
> Once the receive queue has sufficient capacity again, the peer's
> transmit queue is resumed with netif_tx_wake_queue().
> 
> [...]

Here is the summary with links:
  - [net-next,v3] netdevsim: implement peer queue flow control
    https://git.kernel.org/netdev/net-next/c/ff2ac4df58ad

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html