2 files changed, 110 insertions, 8 deletions

diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index 7cfbdb215ba2..44f6a181a754 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -71,6 +71,13 @@ static kmem_cache_t *skbuff_head_cache __read_mostly;
 static kmem_cache_t *skbuff_fclone_cache __read_mostly;
 
 /*
+ * lockdep: lock class key used by skb_queue_head_init():
+ */
+struct lock_class_key skb_queue_lock_key;
+
+EXPORT_SYMBOL(skb_queue_lock_key);
+
+/*
  *      Keep out-of-line to prevent kernel bloat.
  *      __builtin_return_address is not used because it is not always
  *      reliable.
diff --git a/net/core/sock.c b/net/core/sock.c
index 533b9317144b..51fcfbc041a7 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -129,6 +129,53 @@
 #include <net/tcp.h>
 #endif
 
+/*
+ * Each address family might have different locking rules, so we have
+ * one slock key per address family:
+ */
+static struct lock_class_key af_family_keys[AF_MAX];
+static struct lock_class_key af_family_slock_keys[AF_MAX];
+
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+/*
+ * Make lock validator output more readable. (we pre-construct these
+ * strings build-time, so that runtime initialization of socket
+ * locks is fast):
+ */
+static const char *af_family_key_strings[AF_MAX+1] = {
+  "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX"     , "sk_lock-AF_INET"     ,
+  "sk_lock-AF_AX25"  , "sk_lock-AF_IPX"      , "sk_lock-AF_APPLETALK",
+  "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE"   , "sk_lock-AF_ATMPVC"   ,
+  "sk_lock-AF_X25"   , "sk_lock-AF_INET6"    , "sk_lock-AF_ROSE"     ,
+  "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI"  , "sk_lock-AF_SECURITY" ,
+  "sk_lock-AF_KEY"   , "sk_lock-AF_NETLINK"  , "sk_lock-AF_PACKET"   ,
+  "sk_lock-AF_ASH"   , "sk_lock-AF_ECONET"   , "sk_lock-AF_ATMSVC"   ,
+  "sk_lock-21"       , "sk_lock-AF_SNA"      , "sk_lock-AF_IRDA"     ,
+  "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE"  , "sk_lock-AF_LLC"      ,
+  "sk_lock-27"       , "sk_lock-28"          , "sk_lock-29"          ,
+  "sk_lock-AF_TIPC"  , "sk_lock-AF_BLUETOOTH", "sk_lock-AF_MAX"
+};
+static const char *af_family_slock_key_strings[AF_MAX+1] = {
+  "slock-AF_UNSPEC", "slock-AF_UNIX"     , "slock-AF_INET"     ,
+  "slock-AF_AX25"  , "slock-AF_IPX"      , "slock-AF_APPLETALK",
+  "slock-AF_NETROM", "slock-AF_BRIDGE"   , "slock-AF_ATMPVC"   ,
+  "slock-AF_X25"   , "slock-AF_INET6"    , "slock-AF_ROSE"     ,
+  "slock-AF_DECnet", "slock-AF_NETBEUI"  , "slock-AF_SECURITY" ,
+  "slock-AF_KEY"   , "slock-AF_NETLINK"  , "slock-AF_PACKET"   ,
+  "slock-AF_ASH"   , "slock-AF_ECONET"   , "slock-AF_ATMSVC"   ,
+  "slock-21"       , "slock-AF_SNA"      , "slock-AF_IRDA"     ,
+  "slock-AF_PPPOX" , "slock-AF_WANPIPE"  , "slock-AF_LLC"      ,
+  "slock-27"       , "slock-28"          , "slock-29"          ,
+  "slock-AF_TIPC"  , "slock-AF_BLUETOOTH", "slock-AF_MAX"
+};
+#endif
+
+/*
+ * sk_callback_lock locking rules are per-address-family,
+ * so split the lock classes by using a per-AF key:
+ */
+static struct lock_class_key af_callback_keys[AF_MAX];
+
 /* Take into consideration the size of the struct sk_buff overhead in the
  * determination of these values, since that is non-constant across
  * platforms.  This makes socket queueing behavior and performance
@@ -237,9 +284,16 @@ int sk_receive_skb(struct sock *sk, struct sk_buff *skb)
         skb->dev = NULL;
 
         bh_lock_sock(sk);
-        if (!sock_owned_by_user(sk))
+        if (!sock_owned_by_user(sk)) {
+                /*
+                 * trylock + unlock semantics:
+                 */
+                mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
+
                 rc = sk->sk_backlog_rcv(sk, skb);
-        else
+
+                mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
+        } else
                 sk_add_backlog(sk, skb);
         bh_unlock_sock(sk);
 out:
@@ -749,6 +803,33 @@ lenout:
         return 0;
 }
 
+/*
+ * Initialize an sk_lock.
+ *
+ * (We also register the sk_lock with the lock validator.)
+ */
+static void inline sock_lock_init(struct sock *sk)
+{
+        spin_lock_init(&sk->sk_lock.slock);
+        sk->sk_lock.owner = NULL;
+        init_waitqueue_head(&sk->sk_lock.wq);
+        /*
+         * Make sure we are not reinitializing a held lock:
+         */
+        debug_check_no_locks_freed((void *)&sk->sk_lock, sizeof(sk->sk_lock));
+
+        /*
+         * Mark both the sk_lock and the sk_lock.slock as a
+         * per-address-family lock class:
+         */
+        lockdep_set_class_and_name(&sk->sk_lock.slock,
+                                   af_family_slock_keys + sk->sk_family,
+                                   af_family_slock_key_strings[sk->sk_family]);
+        lockdep_init_map(&sk->sk_lock.dep_map,
+                         af_family_key_strings[sk->sk_family],
+                         af_family_keys + sk->sk_family);
+}
+
 /**
  *      sk_alloc - All socket objects are allocated here
  *      @family: protocol family
@@ -848,6 +929,8 @@ struct sock *sk_clone(const struct sock *sk, const gfp_t priority)
 
                 rwlock_init(&newsk->sk_dst_lock);
                 rwlock_init(&newsk->sk_callback_lock);
+                lockdep_set_class(&newsk->sk_callback_lock,
+                                   af_callback_keys + newsk->sk_family);
 
                 newsk->sk_dst_cache     = NULL;
                 newsk->sk_wmem_queued   = 0;
@@ -1422,6 +1505,8 @@ void sock_init_data(struct socket *sock, struct sock *sk)
 
         rwlock_init(&sk->sk_dst_lock);
         rwlock_init(&sk->sk_callback_lock);
+        lockdep_set_class(&sk->sk_callback_lock,
+                           af_callback_keys + sk->sk_family);
 
         sk->sk_state_change     =       sock_def_wakeup;
         sk->sk_data_ready       =       sock_def_readable;
@@ -1449,24 +1534,34 @@ void sock_init_data(struct socket *sock, struct sock *sk)
 void fastcall lock_sock(struct sock *sk)
 {
         might_sleep();
-        spin_lock_bh(&(sk->sk_lock.slock));
+        spin_lock_bh(&sk->sk_lock.slock);
         if (sk->sk_lock.owner)
                 __lock_sock(sk);
         sk->sk_lock.owner = (void *)1;
-        spin_unlock_bh(&(sk->sk_lock.slock));
+        spin_unlock(&sk->sk_lock.slock);
+        /*
+         * The sk_lock has mutex_lock() semantics here:
+         */
+        mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
+        local_bh_enable();
 }
 
 EXPORT_SYMBOL(lock_sock);
 
 void fastcall release_sock(struct sock *sk)
 {
-        spin_lock_bh(&(sk->sk_lock.slock));
+        /*
+         * The sk_lock has mutex_unlock() semantics:
+         */
+        mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
+
+        spin_lock_bh(&sk->sk_lock.slock);
         if (sk->sk_backlog.tail)
                 __release_sock(sk);
         sk->sk_lock.owner = NULL;
-        if (waitqueue_active(&(sk->sk_lock.wq)))
-                wake_up(&(sk->sk_lock.wq));
-        spin_unlock_bh(&(sk->sk_lock.slock));
+        if (waitqueue_active(&sk->sk_lock.wq))
+                wake_up(&sk->sk_lock.wq);
+        spin_unlock_bh(&sk->sk_lock.slock);
 }
 EXPORT_SYMBOL(release_sock);