From 60034d3d146b11922ab1db613bce062dddc0327a Mon Sep 17 00:00:00 2001 From: David Howells Date: Tue, 30 Jul 2019 14:42:50 +0100 Subject: rxrpc: Fix potential deadlock There is a potential deadlock in rxrpc_peer_keepalive_dispatch() whereby rxrpc_put_peer() is called with the peer_hash_lock held, but if it reduces the peer's refcount to 0, rxrpc_put_peer() calls __rxrpc_put_peer() - which the tries to take the already held lock. Fix this by providing a version of rxrpc_put_peer() that can be called in situations where the lock is already held. The bug may produce the following lockdep report: ============================================ WARNING: possible recursive locking detected 5.2.0-next-20190718 #41 Not tainted -------------------------------------------- kworker/0:3/21678 is trying to acquire lock: 00000000aa5eecdf (&(&rxnet->peer_hash_lock)->rlock){+.-.}, at: spin_lock_bh /./include/linux/spinlock.h:343 [inline] 00000000aa5eecdf (&(&rxnet->peer_hash_lock)->rlock){+.-.}, at: __rxrpc_put_peer /net/rxrpc/peer_object.c:415 [inline] 00000000aa5eecdf (&(&rxnet->peer_hash_lock)->rlock){+.-.}, at: rxrpc_put_peer+0x2d3/0x6a0 /net/rxrpc/peer_object.c:435 but task is already holding lock: 00000000aa5eecdf (&(&rxnet->peer_hash_lock)->rlock){+.-.}, at: spin_lock_bh /./include/linux/spinlock.h:343 [inline] 00000000aa5eecdf (&(&rxnet->peer_hash_lock)->rlock){+.-.}, at: rxrpc_peer_keepalive_dispatch /net/rxrpc/peer_event.c:378 [inline] 00000000aa5eecdf (&(&rxnet->peer_hash_lock)->rlock){+.-.}, at: rxrpc_peer_keepalive_worker+0x6b3/0xd02 /net/rxrpc/peer_event.c:430 Fixes: 330bdcfadcee ("rxrpc: Fix the keepalive generator [ver #2]") Reported-by: syzbot+72af434e4b3417318f84@syzkaller.appspotmail.com Signed-off-by: David Howells Reviewed-by: Marc Dionne Reviewed-by: Jeffrey Altman --- net/rxrpc/ar-internal.h | 1 + net/rxrpc/peer_event.c | 2 +- net/rxrpc/peer_object.c | 18 ++++++++++++++++++ 3 files changed, 20 insertions(+), 1 deletion(-) (limited to 'net') diff --git a/net/rxrpc/ar-internal.h b/net/rxrpc/ar-internal.h index 80335b4ee4fd..822f45386e31 100644 --- a/net/rxrpc/ar-internal.h +++ b/net/rxrpc/ar-internal.h @@ -1061,6 +1061,7 @@ void rxrpc_destroy_all_peers(struct rxrpc_net *); struct rxrpc_peer *rxrpc_get_peer(struct rxrpc_peer *); struct rxrpc_peer *rxrpc_get_peer_maybe(struct rxrpc_peer *); void rxrpc_put_peer(struct rxrpc_peer *); +void rxrpc_put_peer_locked(struct rxrpc_peer *); /* * proc.c diff --git a/net/rxrpc/peer_event.c b/net/rxrpc/peer_event.c index 9f2f45c09e58..7666ec72d37e 100644 --- a/net/rxrpc/peer_event.c +++ b/net/rxrpc/peer_event.c @@ -378,7 +378,7 @@ static void rxrpc_peer_keepalive_dispatch(struct rxrpc_net *rxnet, spin_lock_bh(&rxnet->peer_hash_lock); list_add_tail(&peer->keepalive_link, &rxnet->peer_keepalive[slot & mask]); - rxrpc_put_peer(peer); + rxrpc_put_peer_locked(peer); } spin_unlock_bh(&rxnet->peer_hash_lock); diff --git a/net/rxrpc/peer_object.c b/net/rxrpc/peer_object.c index 9d3ce81cf8ae..9c3ac96f71cb 100644 --- a/net/rxrpc/peer_object.c +++ b/net/rxrpc/peer_object.c @@ -436,6 +436,24 @@ void rxrpc_put_peer(struct rxrpc_peer *peer) } } +/* + * Drop a ref on a peer record where the caller already holds the + * peer_hash_lock. + */ +void rxrpc_put_peer_locked(struct rxrpc_peer *peer) +{ + const void *here = __builtin_return_address(0); + int n; + + n = atomic_dec_return(&peer->usage); + trace_rxrpc_peer(peer, rxrpc_peer_put, n, here); + if (n == 0) { + hash_del_rcu(&peer->hash_link); + list_del_init(&peer->keepalive_link); + kfree_rcu(peer, rcu); + } +} + /* * Make sure all peer records have been discarded. */ -- cgit v1.2.3 From c69565ee6681e151e2bb80502930a16e04b553d1 Mon Sep 17 00:00:00 2001 From: David Howells Date: Tue, 30 Jul 2019 14:42:50 +0100 Subject: rxrpc: Fix the lack of notification when sendmsg() fails on a DATA packet Fix the fact that a notification isn't sent to the recvmsg side to indicate a call failed when sendmsg() fails to transmit a DATA packet with the error ENETUNREACH, EHOSTUNREACH or ECONNREFUSED. Without this notification, the afs client just sits there waiting for the call to complete in some manner (which it's not now going to do), which also pins the rxrpc call in place. This can be seen if the client has a scope-level IPv6 address, but not a global-level IPv6 address, and we try and transmit an operation to a server's IPv6 address. Looking in /proc/net/rxrpc/calls shows completed calls just sat there with an abort code of RX_USER_ABORT and an error code of -ENETUNREACH. Fixes: c54e43d752c7 ("rxrpc: Fix missing start of call timeout") Signed-off-by: David Howells Reviewed-by: Marc Dionne Reviewed-by: Jeffrey Altman --- net/rxrpc/sendmsg.c | 1 + 1 file changed, 1 insertion(+) (limited to 'net') diff --git a/net/rxrpc/sendmsg.c b/net/rxrpc/sendmsg.c index 5d3f33ce6d41..bae14438f869 100644 --- a/net/rxrpc/sendmsg.c +++ b/net/rxrpc/sendmsg.c @@ -226,6 +226,7 @@ static int rxrpc_queue_packet(struct rxrpc_sock *rx, struct rxrpc_call *call, rxrpc_set_call_completion(call, RXRPC_CALL_LOCAL_ERROR, 0, ret); + rxrpc_notify_socket(call); goto out; } _debug("need instant resend %d", ret); -- cgit v1.2.3