net: sock_def_readable() and friends RCU conversion

sk_callback_lock rwlock actually protects sk->sk_sleep pointer, so we
need two atomic operations (and associated dirtying) per incoming
packet.

RCU conversion is pretty much needed :

1) Add a new structure, called "struct socket_wq" to hold all fields
that will need rcu_read_lock() protection (currently: a
wait_queue_head_t and a struct fasync_struct pointer).

[Future patch will add a list anchor for wakeup coalescing]

2) Attach one of such structure to each "struct socket" created in
sock_alloc_inode().

3) Respect RCU grace period when freeing a "struct socket_wq"

4) Change sk_sleep pointer in "struct sock" by sk_wq, pointer to "struct
socket_wq"

5) Change sk_sleep() function to use new sk->sk_wq instead of
sk->sk_sleep

6) Change sk_has_sleeper() to wq_has_sleeper() that must be used inside
a rcu_read_lock() section.

7) Change all sk_has_sleeper() callers to :
  - Use rcu_read_lock() instead of read_lock(&sk->sk_callback_lock)
  - Use wq_has_sleeper() to eventually wakeup tasks.
  - Use rcu_read_unlock() instead of read_unlock(&sk->sk_callback_lock)

8) sock_wake_async() is modified to use rcu protection as well.

9) Exceptions :
  macvtap, drivers/net/tun.c, af_unix use integrated "struct socket_wq"
instead of dynamically allocated ones. They dont need rcu freeing.

Some cleanups or followups are probably needed, (possible
sk_callback_lock conversion to a spinlock for example...).

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

3 files changed, 39 insertions, 33 deletions

diff --git a/include/linux/net.h b/include/linux/net.h
index 4157b5d..2b4deee 100644
--- a/include/linux/net.h
+++ b/include/linux/net.h
@@ -59,6 +59,7 @@ typedef enum {
 #include <linux/wait.h>
 #include <linux/fcntl.h>	/* For O_CLOEXEC and O_NONBLOCK */
 #include <linux/kmemcheck.h>
+#include <linux/rcupdate.h>
 
 struct poll_table_struct;
 struct pipe_inode_info;
@@ -116,6 +117,12 @@ enum sock_shutdown_cmd {
 	SHUT_RDWR	= 2,
 };
 
+struct socket_wq {
+	wait_queue_head_t	wait;
+	struct fasync_struct	*fasync_list;
+	struct rcu_head		rcu;
+} ____cacheline_aligned_in_smp;
+
 /**
  *  struct socket - general BSD socket
  *  @state: socket state (%SS_CONNECTED, etc)
@@ -135,11 +142,8 @@ struct socket {
 	kmemcheck_bitfield_end(type);
 
 	unsigned long		flags;
-	/*
-	 * Please keep fasync_list & wait fields in the same cache line
-	 */
-	struct fasync_struct	*fasync_list;
-	wait_queue_head_t	wait;
+
+	struct socket_wq	*wq;
 
 	struct file		*file;
 	struct sock		*sk;
diff --git a/include/net/af_unix.h b/include/net/af_unix.h
index 1614d78..20725e2 100644
--- a/include/net/af_unix.h
+++ b/include/net/af_unix.h
@@ -30,7 +30,7 @@ struct unix_skb_parms {
 #endif
 };
 
-#define UNIXCB(skb) 	(*(struct unix_skb_parms*)&((skb)->cb))
+#define UNIXCB(skb) 	(*(struct unix_skb_parms *)&((skb)->cb))
 #define UNIXCREDS(skb)	(&UNIXCB((skb)).creds)
 #define UNIXSID(skb)	(&UNIXCB((skb)).secid)
 
@@ -45,21 +45,23 @@ struct unix_skb_parms {
 struct unix_sock {
 	/* WARNING: sk has to be the first member */
 	struct sock		sk;
-        struct unix_address     *addr;
-        struct dentry		*dentry;
-        struct vfsmount		*mnt;
+	struct unix_address     *addr;
+	struct dentry		*dentry;
+	struct vfsmount		*mnt;
 	struct mutex		readlock;
-        struct sock		*peer;
-        struct sock		*other;
+	struct sock		*peer;
+	struct sock		*other;
 	struct list_head	link;
-        atomic_long_t           inflight;
-        spinlock_t		lock;
+	atomic_long_t		inflight;
+	spinlock_t		lock;
 	unsigned int		gc_candidate : 1;
 	unsigned int		gc_maybe_cycle : 1;
-        wait_queue_head_t       peer_wait;
+	struct socket_wq	peer_wq;
 };
 #define unix_sk(__sk) ((struct unix_sock *)__sk)
 
+#define peer_wait peer_wq.wait
+
 #ifdef CONFIG_SYSCTL
 extern int unix_sysctl_register(struct net *net);
 extern void unix_sysctl_unregister(struct net *net);
diff --git a/include/net/sock.h b/include/net/sock.h
index e1777db..cc7f91e 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -159,7 +159,7 @@ struct sock_common {
   *	@sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
   *	@sk_lock:	synchronizer
   *	@sk_rcvbuf: size of receive buffer in bytes
-  *	@sk_sleep: sock wait queue
+  *	@sk_wq: sock wait queue and async head
   *	@sk_dst_cache: destination cache
   *	@sk_dst_lock: destination cache lock
   *	@sk_policy: flow policy
@@ -257,7 +257,7 @@ struct sock {
 		struct sk_buff *tail;
 		int len;
 	} sk_backlog;
-	wait_queue_head_t	*sk_sleep;
+	struct socket_wq	*sk_wq;
 	struct dst_entry	*sk_dst_cache;
 #ifdef CONFIG_XFRM
 	struct xfrm_policy	*sk_policy[2];
@@ -1219,7 +1219,7 @@ static inline void sk_set_socket(struct sock *sk, struct socket *sock)
 
 static inline wait_queue_head_t *sk_sleep(struct sock *sk)
 {
-	return sk->sk_sleep;
+	return &sk->sk_wq->wait;
 }
 /* Detach socket from process context.
  * Announce socket dead, detach it from wait queue and inode.
@@ -1233,14 +1233,14 @@ static inline void sock_orphan(struct sock *sk)
 	write_lock_bh(&sk->sk_callback_lock);
 	sock_set_flag(sk, SOCK_DEAD);
 	sk_set_socket(sk, NULL);
-	sk->sk_sleep  = NULL;
+	sk->sk_wq  = NULL;
 	write_unlock_bh(&sk->sk_callback_lock);
 }
 
 static inline void sock_graft(struct sock *sk, struct socket *parent)
 {
 	write_lock_bh(&sk->sk_callback_lock);
-	sk->sk_sleep = &parent->wait;
+	rcu_assign_pointer(sk->sk_wq, parent->wq);
 	parent->sk = sk;
 	sk_set_socket(sk, parent);
 	security_sock_graft(sk, parent);
@@ -1392,12 +1392,12 @@ static inline int sk_has_allocations(const struct sock *sk)
 }
 
 /**
- * sk_has_sleeper - check if there are any waiting processes
- * @sk: socket
+ * wq_has_sleeper - check if there are any waiting processes
+ * @sk: struct socket_wq
  *
- * Returns true if socket has waiting processes
+ * Returns true if socket_wq has waiting processes
  *
- * The purpose of the sk_has_sleeper and sock_poll_wait is to wrap the memory
+ * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
  * barrier call. They were added due to the race found within the tcp code.
  *
  * Consider following tcp code paths:
@@ -1410,9 +1410,10 @@ static inline int sk_has_allocations(const struct sock *sk)
  *   ...                 ...
  *   tp->rcv_nxt check   sock_def_readable
  *   ...                 {
- *   schedule               ...
- *                          if (sk_sleep(sk) && waitqueue_active(sk_sleep(sk)))
- *                              wake_up_interruptible(sk_sleep(sk))
+ *   schedule               rcu_read_lock();
+ *                          wq = rcu_dereference(sk->sk_wq);
+ *                          if (wq && waitqueue_active(&wq->wait))
+ *                              wake_up_interruptible(&wq->wait)
  *                          ...
  *                       }
  *
@@ -1421,19 +1422,18 @@ static inline int sk_has_allocations(const struct sock *sk)
  * could then endup calling schedule and sleep forever if there are no more
  * data on the socket.
  *
- * The sk_has_sleeper is always called right after a call to read_lock, so we
- * can use smp_mb__after_lock barrier.
  */
-static inline int sk_has_sleeper(struct sock *sk)
+static inline bool wq_has_sleeper(struct socket_wq *wq)
 {
+
 	/*
 	 * We need to be sure we are in sync with the
 	 * add_wait_queue modifications to the wait queue.
 	 *
 	 * This memory barrier is paired in the sock_poll_wait.
 	 */
-	smp_mb__after_lock();
-	return sk_sleep(sk) && waitqueue_active(sk_sleep(sk));
+	smp_mb();
+	return wq && waitqueue_active(&wq->wait);
 }
 
 /**
@@ -1442,7 +1442,7 @@ static inline int sk_has_sleeper(struct sock *sk)
  * @wait_address:   socket wait queue
  * @p:              poll_table
  *
- * See the comments in the sk_has_sleeper function.
+ * See the comments in the wq_has_sleeper function.
  */
 static inline void sock_poll_wait(struct file *filp,
 		wait_queue_head_t *wait_address, poll_table *p)
@@ -1453,7 +1453,7 @@ static inline void sock_poll_wait(struct file *filp,
 		 * We need to be sure we are in sync with the
 		 * socket flags modification.
 		 *
-		 * This memory barrier is paired in the sk_has_sleeper.
+		 * This memory barrier is paired in the wq_has_sleeper.
 		*/
 		smp_mb();
 	}