On 03/20, Loktionov, Aleksandr wrote:

quoted

-----Original Message-----
From: Intel-wired-lan <redacted> On Behalf
Of Stanislav Fomichev
Sent: Friday, March 20, 2026 2:25 AM
To: netdev@vger.kernel.org
Cc: davem@davemloft.net; edumazet@google.com; kuba@kernel.org;
pabeni@redhat.com; horms@kernel.org; corbet@lwn.net;
skhan@linuxfoundation.org; andrew+netdev@lunn.ch;
michael.chan@broadcom.com; pavan.chebbi@broadcom.com; Nguyen, Anthony
L [off-list ref]; Kitszel, Przemyslaw
[off-list ref]; saeedm@nvidia.com; tariqt@nvidia.com;
mbloch@nvidia.com; alexanderduyck@fb.com; kernel-team@meta.com;
johannes@sipsolutions.net; sd@queasysnail.net; jianbol@nvidia.com;
dtatulea@nvidia.com; sdf@fomichev.me; mohsin.bashr@gmail.com; Keller,
Jacob E [off-list ref]; willemb@google.com;
skhawaja@google.com; bestswngs@gmail.com; Loktionov, Aleksandr
[off-list ref]; kees@kernel.org; linux-
doc@vger.kernel.org; linux-kernel@vger.kernel.org; intel-wired-
lan@lists.osuosl.org; linux-rdma@vger.kernel.org; linux-
wireless@vger.kernel.org; linux-kselftest@vger.kernel.org;
leon@kernel.org
Subject: [Intel-wired-lan] [PATCH net-next v3 03/13] net: introduce
ndo_set_rx_mode_async and dev_rx_mode_work

Add ndo_set_rx_mode_async callback that drivers can implement instead
of the legacy ndo_set_rx_mode. The legacy callback runs under the
netif_addr_lock spinlock with BHs disabled, preventing drivers from
sleeping. The async variant runs from a work queue with rtnl_lock and
netdev_lock_ops held, in fully sleepable context.

When __dev_set_rx_mode() sees ndo_set_rx_mode_async, it schedules
dev_rx_mode_work instead of calling the driver inline. The work
function takes two snapshots of each address list (uc/mc) under the
addr_lock, then drops the lock and calls the driver with the work
copies. After the driver returns, it reconciles the snapshots back to
the real lists under the lock.

Reviewed-by: Aleksandr Loktionov <redacted>
Signed-off-by: Stanislav Fomichev <sdf@fomichev.me>
---
 Documentation/networking/netdevices.rst |  8 +++
 include/linux/netdevice.h               | 20 ++++++
 net/core/dev.c                          | 95 +++++++++++++++++++++++-
-
 3 files changed, 116 insertions(+), 7 deletions(-)

diff --git a/Documentation/networking/netdevices.rst

b/Documentation/networking/netdevices.rst
index 35704d115312..dc83d78d3b27 100644

--- a/Documentation/networking/netdevices.rst
+++ b/Documentation/networking/netdevices.rst

@@ -289,6 +289,14 @@ struct net_device synchronization rules
 ndo_set_rx_mode:
 	Synchronization: netif_addr_lock spinlock.
 	Context: BHs disabled

...

quoted

to
+device
+ * and configure RX filtering.
+ * @dev: device
+ *
+ * When the device doesn't support unicast filtering it is put in
+promiscuous
+ * mode while unicast addresses are present.
  */
 void __dev_set_rx_mode(struct net_device *dev)  {
 	const struct net_device_ops *ops = dev->netdev_ops;

 	/* dev_open will call this function so the list will stay sane.
*/
-	if (!(dev->flags&IFF_UP))
+	if (!netif_up_and_present(dev))
 		return;

-	if (!netif_device_present(dev))
+	if (ops->ndo_set_rx_mode_async) {
+		queue_work(rx_mode_wq, &dev->rx_mode_work);
 		return;

This early return skips the legacy core fallback below.
Before this patch, __dev_set_rx_mode() continued into the
existing unicast-filter handling when the device did not
advertise IFF_UNICAST_FLT.

After this patch, any driver that implements
ndo_set_rx_mode_async but does not set IFF_UNICAST_FLT
will never hit that fallback path.

I believe this is addressed later in "net: move promiscuity handling into
dev_rx_mode_work"? That should take care of doing __dev_set_promiscuity
for !IFF_UNICAST_FLT+ndo_set_rx_mode_async. Not sure if there is a
better way to rearrange the chunks in the patches.

        if (ops->ndo_set_rx_mode_async) {
                ...

+		promisc_inc = dev_uc_promisc_update(dev);
+
+		netif_addr_unlock_bh(dev);
+	} else {
+		netif_addr_lock_bh(dev);
+		promisc_inc = dev_uc_promisc_update(dev);
+		netif_addr_unlock_bh(dev);
+	}
+
+	if (promisc_inc)
+		__dev_set_promiscuity(dev, promisc_inc, false);
+