net: team: do not use dynamic lockdep key

team interface has used a dynamic lockdep key to avoid false-positive
lockdep deadlock detection. Virtual interfaces such as team usually
have their own lock for protecting private data.
These interfaces can be nested.
team0
  |
team1

Each interface's lock is actually different(team0->lock and team1->lock).
So,
mutex_lock(&team0->lock);
mutex_lock(&team1->lock);
mutex_unlock(&team1->lock);
mutex_unlock(&team0->lock);
The above case is absolutely safe. But lockdep warns about deadlock.
Because the lockdep understands these two locks are same. This is a
false-positive lockdep warning.

So, in order to avoid this problem, the team interfaces started to use
dynamic lockdep key. The false-positive problem was fixed, but it
introduced a new problem.

When the new team virtual interface is created, it registers a dynamic
lockdep key(creates dynamic lockdep key) and uses it. But there is the
limitation of the number of lockdep keys.
So, If so many team interfaces are created, it consumes all lockdep keys.
Then, the lockdep stops to work and warns about it.

In order to fix this problem, team interfaces use the subclass instead
of the dynamic key. So, when a new team interface is created, it doesn't
register(create) a new lockdep, but uses existed subclass key instead.
It is already used by the bonding interface for a similar case.

As the bonding interface does, the subclass variable is the same as
the 'dev->nested_level'. This variable indicates the depth in the stacked
interface graph.

The 'dev->nested_level' is protected by RTNL and RCU.
So, 'mutex_lock_nested()' for 'team->lock' requires RTNL or RCU.
In the current code, 'team->lock' is usually acquired under RTNL, there is
no problem with using 'dev->nested_level'.

The 'team_nl_team_get()' and The 'lb_stats_refresh()' functions acquire
'team->lock' without RTNL.
But these don't iterate their own ports nested so they don't need nested
lock.

Reproducer:
   for i in {0..1000}
   do
           ip link add team$i type team
           ip link add dummy$i master team$i type dummy
           ip link set dummy$i up
           ip link set team$i up
   done

Splat looks like:
   BUG: MAX_LOCKDEP_ENTRIES too low!
   turning off the locking correctness validator.
   Please attach the output of /proc/lock_stat to the bug report
   CPU: 0 PID: 4104 Comm: ip Not tainted 6.5.0-rc7+ #45
   Call Trace:
    <TASK>
   dump_stack_lvl+0x64/0xb0
   add_lock_to_list+0x30d/0x5e0
   check_prev_add+0x73a/0x23a0
   ...
   sock_def_readable+0xfe/0x4f0
   netlink_broadcast+0x76b/0xac0
   nlmsg_notify+0x69/0x1d0
   dev_open+0xed/0x130
   ...

Reported-by: syzbot+9bbbacfbf1e04d5221f7@syzkaller.appspotmail.com
Fixes: 369f61bee0f5 ("team: fix nested locking lockdep warning")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 29 insertions, 1 deletions

diff --git a/include/linux/if_team.h b/include/linux/if_team.h
index 1b9b15a492fa..12d4447fc8ab 100644
--- a/include/linux/if_team.h
+++ b/include/linux/if_team.h
@@ -221,10 +221,38 @@ struct team {
 		atomic_t count_pending;
 		struct delayed_work dw;
 	} mcast_rejoin;
-	struct lock_class_key team_lock_key;
 	long mode_priv[TEAM_MODE_PRIV_LONGS];
 };
 
+static inline void __team_lock(struct team *team)
+{
+	mutex_lock(&team->lock);
+}
+
+static inline int team_trylock(struct team *team)
+{
+	return mutex_trylock(&team->lock);
+}
+
+#ifdef CONFIG_LOCKDEP
+static inline void team_lock(struct team *team)
+{
+	ASSERT_RTNL();
+	mutex_lock_nested(&team->lock, team->dev->nested_level);
+}
+
+#else
+static inline void team_lock(struct team *team)
+{
+	__team_lock(team);
+}
+#endif
+
+static inline void team_unlock(struct team *team)
+{
+	mutex_unlock(&team->lock);
+}
+
 static inline int team_dev_queue_xmit(struct team *team, struct team_port *port,
 				      struct sk_buff *skb)
 {