On Sun, Mar 13, 2022 at 04:10:30PM +0200, Vladimir Oltean wrote:
> Hi Daniel,
>
> On Sun, Mar 13, 2022 at 03:03:07PM +0100, Daniel Suchy wrote:
> > Hello,
> >
> > I noticed boot problems on my Turris Omnia (with Marvell 88E6176 switch
> > chip) after "net: dsa: mv88e6xxx: flush switchdev FDB workqueue before
> > removing VLAN" commit https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=2566a89b9e163b2fcd104d6005e0149f197b8a48
> >
> > Within logs I catched hung kernel tasks (see below), at least first is
> > related to DSA subsystem.
> >
> > When I revert this patch, everything works as expected and without any
> > issues.
> >
> > In my setup, I have few vlans on affected switch (i'm using ifupdown2 v3.0
> > with iproute2 5.16 for configuration).
> >
> > It seems your this patch introduces some new problem (at least for 5.15
> > kernels). I suggest revert this patch.
> >
> > - Daniel
>
> Oh wow, I'm terribly sorry. Yes, this patch shouldn't have been
> backported to kernel 5.15 and below, but I guess I missed the
> backport notification email and forgot to tell Greg about this.
> Patch "net: dsa: mv88e6xxx: flush switchdev FDB workqueue before
> removing VLAN" needs to be immediately reverted from these trees.
>
> Greg, to avoid this from happening in the future, would something like
> this work? Is this parsed in some way?
>
> Depends-on: 0faf890fc519 ("net: dsa: drop rtnl_lock from dsa_slave_switchdev_event_work") # which first appeared in v5.16
Looks like only 5.15 is affected. I re-checked the backport notification
emails, and the patch failed to apply on 5.10, 5.4, 4.19 and 4.14, as
intended.