rxrpc: Fix deadlock between call creation and sendmsg/recvmsg

All the routines by which rxrpc is accessed from the outside are serialised
by means of the socket lock (sendmsg, recvmsg, bind,
rxrpc_kernel_begin_call(), ...) and this presents a problem:

 (1) If a number of calls on the same socket are in the process of
     connection to the same peer, a maximum of four concurrent live calls
     are permitted before further calls need to wait for a slot.

 (2) If a call is waiting for a slot, it is deep inside sendmsg() or
     rxrpc_kernel_begin_call() and the entry function is holding the socket
     lock.

 (3) sendmsg() and recvmsg() or the in-kernel equivalents are prevented
     from servicing the other calls as they need to take the socket lock to
     do so.

 (4) The socket is stuck until a call is aborted and makes its slot
     available to the waiter.

Fix this by:

 (1) Provide each call with a mutex ('user_mutex') that arbitrates access
     by the users of rxrpc separately for each specific call.

 (2) Make rxrpc_sendmsg() and rxrpc_recvmsg() unlock the socket as soon as
     they've got a call and taken its mutex.

     Note that I'm returning EWOULDBLOCK from recvmsg() if MSG_DONTWAIT is
     set but someone else has the lock.  Should I instead only return
     EWOULDBLOCK if there's nothing currently to be done on a socket, and
     sleep in this particular instance because there is something to be
     done, but we appear to be blocked by the interrupt handler doing its
     ping?

 (3) Make rxrpc_new_client_call() unlock the socket after allocating a new
     call, locking its user mutex and adding it to the socket's call tree.
     The call is returned locked so that sendmsg() can add data to it
     immediately.

     From the moment the call is in the socket tree, it is subject to
     access by sendmsg() and recvmsg() - even if it isn't connected yet.

 (4) Lock new service calls in the UDP data_ready handler (in
     rxrpc_new_incoming_call()) because they may already be in the socket's
     tree and the data_ready handler makes them live immediately if a user
     ID has already been preassigned.

     Note that the new call is locked before any notifications are sent
     that it is live, so doing mutex_trylock() *ought* to always succeed.
     Userspace is prevented from doing sendmsg() on calls that are in a
     too-early state in rxrpc_do_sendmsg().

 (5) Make rxrpc_new_incoming_call() return the call with the user mutex
     held so that a ping can be scheduled immediately under it.

     Note that it might be worth moving the ping call into
     rxrpc_new_incoming_call() and then we can drop the mutex there.

 (6) Make rxrpc_accept_call() take the lock on the call it is accepting and
     release the socket after adding the call to the socket's tree.  This
     is slightly tricky as we've dequeued the call by that point and have
     to requeue it.

     Note that requeuing emits a trace event.

 (7) Make rxrpc_kernel_send_data() and rxrpc_kernel_recv_data() take the
     new mutex immediately and don't bother with the socket mutex at all.

This patch has the nice bonus that calls on the same socket are now to some
extent parallelisable.

Note that we might want to move rxrpc_service_prealloc() calls out from the
socket lock and give it its own lock, so that we don't hang progress in
other calls because we're waiting for the allocator.

We probably also want to avoid calling rxrpc_notify_socket() from within
the socket lock (rxrpc_accept_call()).

Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Marc Dionne <marc.c.dionne@auristor.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 48 insertions, 0 deletions

diff --git a/net/rxrpc/call_accept.c b/net/rxrpc/call_accept.c
index 7c4c64ab8da2..0ed181f53f32 100644
--- a/net/rxrpc/call_accept.c
+++ b/net/rxrpc/call_accept.c
@@ -323,6 +323,8 @@ static struct rxrpc_call *rxrpc_alloc_incoming_call(struct rxrpc_sock *rx,
  *
  * If we want to report an error, we mark the skb with the packet type and
  * abort code and return NULL.
+ *
+ * The call is returned with the user access mutex held.
  */
 struct rxrpc_call *rxrpc_new_incoming_call(struct rxrpc_local *local,
 					   struct rxrpc_connection *conn,
@@ -371,6 +373,18 @@ found_service:
 	trace_rxrpc_receive(call, rxrpc_receive_incoming,
 			    sp->hdr.serial, sp->hdr.seq);
 
+	/* Lock the call to prevent rxrpc_kernel_send/recv_data() and
+	 * sendmsg()/recvmsg() inconveniently stealing the mutex once the
+	 * notification is generated.
+	 *
+	 * The BUG should never happen because the kernel should be well
+	 * behaved enough not to access the call before the first notification
+	 * event and userspace is prevented from doing so until the state is
+	 * appropriate.
+	 */
+	if (!mutex_trylock(&call->user_mutex))
+		BUG();
+
 	/* Make the call live. */
 	rxrpc_incoming_call(rx, call, skb);
 	conn = call->conn;
@@ -429,10 +443,12 @@ out:
 /*
  * handle acceptance of a call by userspace
  * - assign the user call ID to the call at the front of the queue
+ * - called with the socket locked.
  */
 struct rxrpc_call *rxrpc_accept_call(struct rxrpc_sock *rx,
 				     unsigned long user_call_ID,
 				     rxrpc_notify_rx_t notify_rx)
+	__releases(&rx->sk.sk_lock.slock)
 {
 	struct rxrpc_call *call;
 	struct rb_node *parent, **pp;
@@ -446,6 +462,7 @@ struct rxrpc_call *rxrpc_accept_call(struct rxrpc_sock *rx,
 
 	if (list_empty(&rx->to_be_accepted)) {
 		write_unlock(&rx->call_lock);
+		release_sock(&rx->sk);
 		kleave(" = -ENODATA [empty]");
 		return ERR_PTR(-ENODATA);
 	}
@@ -470,10 +487,39 @@ struct rxrpc_call *rxrpc_accept_call(struct rxrpc_sock *rx,
 	 */
 	call = list_entry(rx->to_be_accepted.next,
 			  struct rxrpc_call, accept_link);
+	write_unlock(&rx->call_lock);
+
+	/* We need to gain the mutex from the interrupt handler without
+	 * upsetting lockdep, so we have to release it there and take it here.
+	 * We are, however, still holding the socket lock, so other accepts
+	 * must wait for us and no one can add the user ID behind our backs.
+	 */
+	if (mutex_lock_interruptible(&call->user_mutex) < 0) {
+		release_sock(&rx->sk);
+		kleave(" = -ERESTARTSYS");
+		return ERR_PTR(-ERESTARTSYS);
+	}
+
+	write_lock(&rx->call_lock);
 	list_del_init(&call->accept_link);
 	sk_acceptq_removed(&rx->sk);
 	rxrpc_see_call(call);
 
+	/* Find the user ID insertion point. */
+	pp = &rx->calls.rb_node;
+	parent = NULL;
+	while (*pp) {
+		parent = *pp;
+		call = rb_entry(parent, struct rxrpc_call, sock_node);
+
+		if (user_call_ID < call->user_call_ID)
+			pp = &(*pp)->rb_left;
+		else if (user_call_ID > call->user_call_ID)
+			pp = &(*pp)->rb_right;
+		else
+			BUG();
+	}
+
 	write_lock_bh(&call->state_lock);
 	switch (call->state) {
 	case RXRPC_CALL_SERVER_ACCEPTING:
@@ -499,6 +545,7 @@ struct rxrpc_call *rxrpc_accept_call(struct rxrpc_sock *rx,
 	write_unlock(&rx->call_lock);
 	rxrpc_notify_socket(call);
 	rxrpc_service_prealloc(rx, GFP_KERNEL);
+	release_sock(&rx->sk);
 	_leave(" = %p{%d}", call, call->debug_id);
 	return call;
 
@@ -515,6 +562,7 @@ id_in_use:
 	write_unlock(&rx->call_lock);
 out:
 	rxrpc_service_prealloc(rx, GFP_KERNEL);
+	release_sock(&rx->sk);
 	_leave(" = %d", ret);
 	return ERR_PTR(ret);
 }