]> git.ipfire.org Git - thirdparty/kernel/stable.git/commit
Merge branch 'net-sleepable-ndo_set_rx_mode'
authorPaolo Abeni <pabeni@redhat.com>
Tue, 21 Apr 2026 10:50:26 +0000 (12:50 +0200)
committerPaolo Abeni <pabeni@redhat.com>
Tue, 21 Apr 2026 10:50:26 +0000 (12:50 +0200)
commitedaa48dc2c071cf2ab0611ee504bbd4c544fc178
tree20044ff1883003104534576496d228414979c4a0
parent4c1367a2d7aad643a6f87c6931b13cc1a25e8ca7
parentc4dde411bc366f568dbe33366253bbfea049e8ea
Merge branch 'net-sleepable-ndo_set_rx_mode'

Stanislav Fomichev says:

====================
net: sleepable ndo_set_rx_mode

This series adds a new ndo_set_rx_mode_async callback that enables
drivers to handle address list updates in a sleepable context. The
current ndo_set_rx_mode is called under the netif_addr_lock spinlock
with BHs disabled, which prevents drivers from sleeping. This is
problematic for ops-locked drivers that need to sleep.

The approach:
1. Add snapshot/reconcile infrastructure for address lists
2. Introduce dev_rx_mode_work that takes snapshots under the lock,
   drops the lock, calls the driver, then reconciles changes back
3. Move promiscuity handling into the scheduled work as well
4. Convert existing ops-locked drivers to ndo_set_rx_mode_async
5. Add a warning for ops-locked drivers still using ndo_set_rx_mode
6. Add a selftest exercising the team+bridge+macvlan topology that
   triggers the addr_lock -> ops_lock ordering issue
====================

Link: https://patch.msgid.link/20260416185712.2155425-1-sdf@fomichev.me
Signed-off-by: Paolo Abeni <pabeni@redhat.com>