On Fri, Sep 18, 2026 at 03:45:40PM -0400, Sasha Levin wrote:
> > /tmp/kci/linux/drivers/net/ethernet/ti/cpsw.c:310:34: note: in expansion of macro 'container_of'
> >   310 |         struct cpsw_priv *priv = container_of(work, struct cpsw_priv, rx_mode_work);
> > /tmp/kci/linux/include/linux/stddef.h:16:33: error: 'struct cpsw_priv' has no member named 'rx_mode_work'
> 
> The queue took 0b8c878d1173 ("net: cpsw: Execute ndo_set_rx_mode callback in a
> work queue") without its sibling c0b5dc73a38f ("net: cpsw_new: Execute
> ndo_set_rx_mode callback in a work queue"), which is the commit that adds
> rx_mode_work to struct cpsw_priv. It is dropped from 6.12, 6.6 and 5.15 now.
> Note that the cpsw.c half applies textually clean on all three trees on its own
> and only fails at compile.
> 
> Kevin, could you send a backport of the pair for 6.12, 6.6 and 5.15?

Sure.

Thanks,
Kevin

> 6.6 and
> 5.15 have no disable_work_sync(), so those two need adapting. 6.1 and 5.10 do
> not need it, the Fixes: commit 1767bb2d47b7 ("ipv6: mcast: Don't hold RTNL for
> IPV6_ADD_MEMBERSHIP and MCAST_JOIN_GROUP.") is not there.
> 
> -- 
> Thanks,
> Sasha