Thanks for pointing that out. I wasn't thinking from that point of view.

    According to Documentation, rtl8150_set_multicast (the
ndo_set_rx_mode callback) should
    rely on the netif_addr_lock spinlock, not the netif_tx_lock
manipulated by netif
    stop/start/wake queue functions.

    However, There is no need to use the netif_addr_lock in the driver
directly because
    the core function (dev_set_rx_mode) invoking this function locks
and unlocks the lock
    correctly.

    Synchronization is therefore handled by the core, making it safe
to remove that lock.

    From what I have seen, every network driver assumes this for the
ndo_set_rx_mode callback.

    I am not sure what the historical context was for using the
tx_lock as the synchronization
    mechanism here but it's definitely not valid in the modern networking stack.

Thanks. Please include an explanation in V2. Also, please read:

https://www.kernel.org/doc/html/latest/process/maintainer-netdev.html

	Andrew