On 24/04/17 14:01, Nikolay Aleksandrov wrote: > On 24/04/17 10:25, Xin Long wrote: >> During removing a bridge device, if the bridge is still up, a new mdb entry >> still can be added in br_multicast_add_group() after all mdb entries are >> removed in br_multicast_dev_del(). Like the path: >> >> mld_ifc_timer_expire -> >> mld_sendpack -> ... >> br_multicast_rcv -> >> br_multicast_add_group >> >> The new mp's timer will be set up. If the timer expires after the bridge >> is freed, it may cause use-after-free panic in br_multicast_group_expired. >> This can happen when ip link remove a bridge or destroy a netns with a >> bridge device inside. >> >> As we can see in br_del_bridge, brctl is also supposed to remove a bridge >> device after it's shutdown. >> >> This patch is to call dev_close at the beginning of br_dev_delete so that >> netif_running check in br_multicast_add_group can avoid this issue. But >> to keep consistent with before, it will not remove the IFF_UP check in >> br_del_bridge for brctl. >> >> Reported-by: Jianwen Ji <jiji@xxxxxxxxxx> >> Signed-off-by: Xin Long <lucien.xin@xxxxxxxxx> >> --- >> net/bridge/br_if.c | 2 ++ >> 1 file changed, 2 insertions(+) >> > > +CC bridge maintainers > > I can see how this could happen, could you also provide the traceback ? > > The patch looks good to me, actually I think it fixes another issue with > mcast stats where the percpu pointer can be accessed after it's freed if > an mcast packet can get sent via br->dev after the br_multicast_dev_del() call. Never mind the another issue part, Ido's recent ndo_uninit() patch fixed it. > This is definitely stable material, if I'm not mistaken the issue is there since > the introduction of br_dev_delete: > commit e10177abf842 > Author: Satish Ashok <sashok@xxxxxxxxxxxxxxxxxxx> > Date: Wed Jul 15 07:16:51 2015 -0700 > > bridge: multicast: fix handling of temp and perm entries > > > > Acked-by: Nikolay Aleksandrov <nikolay@xxxxxxxxxxxxxxxxxxx> > > > > > > >
