Created attachment 114501 [details]
spinlock bad magic trace

Hello,
few days ago i tried to switch to 3.10.18 and have caught "spinlock bad magic".

[    6.120692] Freeing unused kernel memory: 1144k freed
[   19.062504] BUG: spinlock bad magic on CPU#6, swapper/6/0
[   19.068011]  lock: 0xffff8810529ec000, .magic: 3a310000, .owner: <none>/-1, .owner_cpu: 825110576
[   19.076984] CPU: 6 PID: 0 Comm: swapper/6 Not tainted 3.10.17 #1
[   19.083065] Hardware name: Supermicro X9DRG-HF/X9DRG-HF, BIOS 1.0c 08/22/2012
[   19.090278]  0000000000000000 ffff88107fc03b70 ffffffff817823c4 ffff88107fc03b90
[   19.098057]  ffffffff81782450 ffff8810529ec000 ffffffff81a21568 ffff88107fc03bb0
[   19.105841]  ffffffff81782476 ffff8810529ec000 ffff8810529ec000 ffff88107fc03bd0
[   19.113624] Call Trace:
[   19.116149]  <IRQ>  [<ffffffff817823c4>] dump_stack+0x19/0x1b
[   19.122114]  [<ffffffff81782450>] spin_dump+0x8a/0x8f
[   19.127242]  [<ffffffff81782476>] spin_bug+0x21/0x26
[   19.132291]  [<ffffffff81357f22>] do_raw_spin_lock+0xb2/0xc0
[   19.138029]  [<ffffffff8178db8e>] _raw_spin_lock+0xe/0x10
[   19.143506]  [<ffffffff81760d03>] br_stp_rcv+0x73/0x370
....................................

It looks like uninitialized br->lock in in br_stp_rcv
./net/bridge/br_stp_bpdu.c +158

        p = br_port_get_rcu(dev);

        br = p->br;
        spin_lock(&br->lock); <- here
-----------------------------------------------

Bisect leads me to 716ec052d2280d511e10e90ad54a86f5b5d4dcc2 (mainline): 960b8e5018a552f62cfbc0dfe94be7b6ba178f13(stable)
Before that br_port_get_rcu has internal check for br_port_exists, without it scenario like this is possible:

- stp broadcasts 01:80:c2:00:00:00 is coming on some interface
- bridge module loaded but there is no bridge on that interface (br->lock is not initialized).
- interface is in promiscuous mod
 (llc_rcv droping PACKET_OTHERHOST to protect us in promiscuous mode but seems like not as for broadcasts like this)
- rx_handler_data is initialised (it was a macvlan in my case but could be team of somewhat else)

So It seems like STP needs its own IFF_BRIDGE_PORT check.