In rxrpc_open_socket(), it sets up the socket and then sets up the I/O
thread that will handle it.  This is a problem, however, as there's a gap
between the two phases in which a packet may come into rxrpc_encap_rcv()
from the UDP packet but we oops when trying to wake the not-yet created I/O
thread.

As a quick fix, just make rxrpc_encap_rcv() discard the packet if there's
no I/O thread yet.

A better, but more intrusive fix would perhaps be to rearrange things such
that the socket creation is done by the I/O thread.

Fixes: a275da62e8c1 ("rxrpc: Create a per-local endpoint receive queue and I/O thread")
Signed-off-by: David Howells <dhowells@redhat.com>
cc: yuxuanzhe@outlook.com
cc: Marc Dionne <marc.dionne@auristor.com>
cc: "David S. Miller" <davem@davemloft.net>
cc: Eric Dumazet <edumazet@google.com>
cc: Jakub Kicinski <kuba@kernel.org>
cc: Paolo Abeni <pabeni@redhat.com>
cc: Simon Horman <horms@kernel.org>
cc: linux-afs@lists.infradead.org
cc: netdev@vger.kernel.org
---
 net/rxrpc/ar-internal.h  |  2 +-
 net/rxrpc/io_thread.c    | 10 ++++++++--
 net/rxrpc/local_object.c |  2 +-
 3 files changed, 10 insertions(+), 4 deletions(-)

diff --git a/net/rxrpc/ar-internal.h b/net/rxrpc/ar-internal.h
index 80d682f89b23..d0fd37bdcfe9 100644
--- a/net/rxrpc/ar-internal.h
+++ b/net/rxrpc/ar-internal.h
@@ -1056,7 +1056,7 @@ bool rxrpc_direct_abort(struct sk_buff *skb, enum rxrpc_abort_reason why,
 int rxrpc_io_thread(void *data);
 static inline void rxrpc_wake_up_io_thread(struct rxrpc_local *local)
 {
-	wake_up_process(local->io_thread);
+	wake_up_process(READ_ONCE(local->io_thread));
 }
 
 static inline bool rxrpc_protocol_error(struct sk_buff *skb, enum rxrpc_abort_reason why)
diff --git a/net/rxrpc/io_thread.c b/net/rxrpc/io_thread.c
index 0300baa9afcd..07c74c77d802 100644
--- a/net/rxrpc/io_thread.c
+++ b/net/rxrpc/io_thread.c
@@ -27,11 +27,17 @@ int rxrpc_encap_rcv(struct sock *udp_sk, struct sk_buff *skb)
 {
 	struct sk_buff_head *rx_queue;
 	struct rxrpc_local *local = rcu_dereference_sk_user_data(udp_sk);
+	struct task_struct *io_thread;
 
 	if (unlikely(!local)) {
 		kfree_skb(skb);
 		return 0;
 	}
+	io_thread = READ_ONCE(local->io_thread);
+	if (!io_thread) {
+		kfree_skb(skb);
+		return 0;
+	}
 	if (skb->tstamp == 0)
 		skb->tstamp = ktime_get_real();
 
@@ -47,7 +53,7 @@ int rxrpc_encap_rcv(struct sock *udp_sk, struct sk_buff *skb)
 #endif
 
 	skb_queue_tail(rx_queue, skb);
-	rxrpc_wake_up_io_thread(local);
+	wake_up_process(io_thread);
 	return 0;
 }
 
@@ -565,7 +571,7 @@ int rxrpc_io_thread(void *data)
 	__set_current_state(TASK_RUNNING);
 	rxrpc_see_local(local, rxrpc_local_stop);
 	rxrpc_destroy_local(local);
-	local->io_thread = NULL;
+	WRITE_ONCE(local->io_thread, NULL);
 	rxrpc_see_local(local, rxrpc_local_stopped);
 	return 0;
 }
diff --git a/net/rxrpc/local_object.c b/net/rxrpc/local_object.c
index 504453c688d7..f9623ace2201 100644
--- a/net/rxrpc/local_object.c
+++ b/net/rxrpc/local_object.c
@@ -232,7 +232,7 @@ static int rxrpc_open_socket(struct rxrpc_local *local, struct net *net)
 	}
 
 	wait_for_completion(&local->io_thread_ready);
-	local->io_thread = io_thread;
+	WRITE_ONCE(local->io_thread, io_thread);
 	_leave(" = 0");
 	return 0;

On Tue, Oct 1, 2024 at 3:27 PM David Howells <dhowells@redhat.com> wrote:
>
> In rxrpc_open_socket(), it sets up the socket and then sets up the I/O
> thread that will handle it.  This is a problem, however, as there's a gap
> between the two phases in which a packet may come into rxrpc_encap_rcv()
> from the UDP packet but we oops when trying to wake the not-yet created I/O
> thread.
>
> As a quick fix, just make rxrpc_encap_rcv() discard the packet if there's
> no I/O thread yet.
>
> A better, but more intrusive fix would perhaps be to rearrange things such
> that the socket creation is done by the I/O thread.
>
> Fixes: a275da62e8c1 ("rxrpc: Create a per-local endpoint receive queue and I/O thread")
> Signed-off-by: David Howells <dhowells@redhat.com>

Reviewed-by: Eric Dumazet <edumazet@google.com>