]> git.karo-electronics.de Git - karo-tx-linux.git/commitdiff
bridge: fix lockdep addr_list_lock false positive splat
authorNikolay Aleksandrov <nikolay@cumulusnetworks.com>
Fri, 15 Jan 2016 18:03:54 +0000 (19:03 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 31 Jan 2016 19:29:01 +0000 (11:29 -0800)
[ Upstream commit c6894dec8ea9ae05747124dce98b3b5c2e69b168 ]

After promisc mode management was introduced a bridge device could do
dev_set_promiscuity from its ndo_change_rx_flags() callback which in
turn can be called after the bridge's addr_list_lock has been taken
(e.g. by dev_uc_add). This causes a false positive lockdep splat because
the port interfaces' addr_list_lock is taken when br_manage_promisc()
runs after the bridge's addr list lock was already taken.
To remove the false positive introduce a custom bridge addr_list_lock
class and set it on bridge init.
A simple way to reproduce this is with the following:
$ brctl addbr br0
$ ip l add l br0 br0.100 type vlan id 100
$ ip l set br0 up
$ ip l set br0.100 up
$ echo 1 > /sys/class/net/br0/bridge/vlan_filtering
$ brctl addif br0 eth0
Splat:
[   43.684325] =============================================
[   43.684485] [ INFO: possible recursive locking detected ]
[   43.684636] 4.4.0-rc8+ #54 Not tainted
[   43.684755] ---------------------------------------------
[   43.684906] brctl/1187 is trying to acquire lock:
[   43.685047]  (_xmit_ETHER){+.....}, at: [<ffffffff8150169e>] dev_set_rx_mode+0x1e/0x40
[   43.685460]  but task is already holding lock:
[   43.685618]  (_xmit_ETHER){+.....}, at: [<ffffffff815072a7>] dev_uc_add+0x27/0x80
[   43.686015]  other info that might help us debug this:
[   43.686316]  Possible unsafe locking scenario:

[   43.686743]        CPU0
[   43.686967]        ----
[   43.687197]   lock(_xmit_ETHER);
[   43.687544]   lock(_xmit_ETHER);
[   43.687886] *** DEADLOCK ***

[   43.688438]  May be due to missing lock nesting notation

[   43.688882] 2 locks held by brctl/1187:
[   43.689134]  #0:  (rtnl_mutex){+.+.+.}, at: [<ffffffff81510317>] rtnl_lock+0x17/0x20
[   43.689852]  #1:  (_xmit_ETHER){+.....}, at: [<ffffffff815072a7>] dev_uc_add+0x27/0x80
[   43.690575] stack backtrace:
[   43.690970] CPU: 0 PID: 1187 Comm: brctl Not tainted 4.4.0-rc8+ #54
[   43.691270] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.8.1-20150318_183358- 04/01/2014
[   43.691770]  ffffffff826a25c0 ffff8800369fb8e0 ffffffff81360ceb ffffffff826a25c0
[   43.692425]  ffff8800369fb9b8 ffffffff810d0466 ffff8800369fb968 ffffffff81537139
[   43.693071]  ffff88003a08c880 0000000000000000 00000000ffffffff 0000000002080020
[   43.693709] Call Trace:
[   43.693931]  [<ffffffff81360ceb>] dump_stack+0x4b/0x70
[   43.694199]  [<ffffffff810d0466>] __lock_acquire+0x1e46/0x1e90
[   43.694483]  [<ffffffff81537139>] ? netlink_broadcast_filtered+0x139/0x3e0
[   43.694789]  [<ffffffff8153b5da>] ? nlmsg_notify+0x5a/0xc0
[   43.695064]  [<ffffffff810d10f5>] lock_acquire+0xe5/0x1f0
[   43.695340]  [<ffffffff8150169e>] ? dev_set_rx_mode+0x1e/0x40
[   43.695623]  [<ffffffff815edea5>] _raw_spin_lock_bh+0x45/0x80
[   43.695901]  [<ffffffff8150169e>] ? dev_set_rx_mode+0x1e/0x40
[   43.696180]  [<ffffffff8150169e>] dev_set_rx_mode+0x1e/0x40
[   43.696460]  [<ffffffff8150189c>] dev_set_promiscuity+0x3c/0x50
[   43.696750]  [<ffffffffa0586845>] br_port_set_promisc+0x25/0x50 [bridge]
[   43.697052]  [<ffffffffa05869aa>] br_manage_promisc+0x8a/0xe0 [bridge]
[   43.697348]  [<ffffffffa05826ee>] br_dev_change_rx_flags+0x1e/0x20 [bridge]
[   43.697655]  [<ffffffff81501532>] __dev_set_promiscuity+0x132/0x1f0
[   43.697943]  [<ffffffff81501672>] __dev_set_rx_mode+0x82/0x90
[   43.698223]  [<ffffffff815072de>] dev_uc_add+0x5e/0x80
[   43.698498]  [<ffffffffa05b3c62>] vlan_device_event+0x542/0x650 [8021q]
[   43.698798]  [<ffffffff8109886d>] notifier_call_chain+0x5d/0x80
[   43.699083]  [<ffffffff810988b6>] raw_notifier_call_chain+0x16/0x20
[   43.699374]  [<ffffffff814f456e>] call_netdevice_notifiers_info+0x6e/0x80
[   43.699678]  [<ffffffff814f4596>] call_netdevice_notifiers+0x16/0x20
[   43.699973]  [<ffffffffa05872be>] br_add_if+0x47e/0x4c0 [bridge]
[   43.700259]  [<ffffffffa058801e>] add_del_if+0x6e/0x80 [bridge]
[   43.700548]  [<ffffffffa0588b5f>] br_dev_ioctl+0xaf/0xc0 [bridge]
[   43.700836]  [<ffffffff8151a7ac>] dev_ifsioc+0x30c/0x3c0
[   43.701106]  [<ffffffff8151aac9>] dev_ioctl+0xf9/0x6f0
[   43.701379]  [<ffffffff81254345>] ? mntput_no_expire+0x5/0x450
[   43.701665]  [<ffffffff812543ee>] ? mntput_no_expire+0xae/0x450
[   43.701947]  [<ffffffff814d7b02>] sock_do_ioctl+0x42/0x50
[   43.702219]  [<ffffffff814d8175>] sock_ioctl+0x1e5/0x290
[   43.702500]  [<ffffffff81242d0b>] do_vfs_ioctl+0x2cb/0x5c0
[   43.702771]  [<ffffffff81243079>] SyS_ioctl+0x79/0x90
[   43.703033]  [<ffffffff815eebb6>] entry_SYSCALL_64_fastpath+0x16/0x7a

CC: Vlad Yasevich <vyasevic@redhat.com>
CC: Stephen Hemminger <stephen@networkplumber.org>
CC: Bridge list <bridge@lists.linux-foundation.org>
CC: Andy Gospodarek <gospo@cumulusnetworks.com>
CC: Roopa Prabhu <roopa@cumulusnetworks.com>
Fixes: 2796d0c648c9 ("bridge: Automatically manage port promiscuous mode.")
Reported-by: Andy Gospodarek <gospo@cumulusnetworks.com>
Signed-off-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
net/bridge/br_device.c

index 5e88d3e17546618a728bb429fc61e20d5f8d79dc..2c8095a5d824feeba6afa186074d590fd9777b79 100644 (file)
@@ -28,6 +28,8 @@
 const struct nf_br_ops __rcu *nf_br_ops __read_mostly;
 EXPORT_SYMBOL_GPL(nf_br_ops);
 
+static struct lock_class_key bridge_netdev_addr_lock_key;
+
 /* net device transmit always called with BH disabled */
 netdev_tx_t br_dev_xmit(struct sk_buff *skb, struct net_device *dev)
 {
@@ -87,6 +89,11 @@ out:
        return NETDEV_TX_OK;
 }
 
+static void br_set_lockdep_class(struct net_device *dev)
+{
+       lockdep_set_class(&dev->addr_list_lock, &bridge_netdev_addr_lock_key);
+}
+
 static int br_dev_init(struct net_device *dev)
 {
        struct net_bridge *br = netdev_priv(dev);
@@ -99,6 +106,7 @@ static int br_dev_init(struct net_device *dev)
        err = br_vlan_init(br);
        if (err)
                free_percpu(br->stats);
+       br_set_lockdep_class(dev);
 
        return err;
 }