rxrpc: Preallocate peers, conns and calls for incoming service requests

Make it possible for the data_ready handler called from the UDP transport
socket to completely instantiate an rxrpc_call structure and make it
immediately live by preallocating all the memory it might need.  The idea
is to cut out the background thread usage as much as possible.

[Note that the preallocated structs are not actually used in this patch -
 that will be done in a future patch.]

If insufficient resources are available in the preallocation buffers, it
will be possible to discard the DATA packet in the data_ready handler or
schedule a BUSY packet without the need to schedule an attempt at
allocation in a background thread.

To this end:

 (1) Preallocate rxrpc_peer, rxrpc_connection and rxrpc_call structs to a
     maximum number each of the listen backlog size.  The backlog size is
     limited to a maxmimum of 32.  Only this many of each can be in the
     preallocation buffer.

 (2) For userspace sockets, the preallocation is charged initially by
     listen() and will be recharged by accepting or rejecting pending
     new incoming calls.

 (3) For kernel services {,re,dis}charging of the preallocation buffers is
     handled manually.  Two notifier callbacks have to be provided before
     kernel_listen() is invoked:

     (a) An indication that a new call has been instantiated.  This can be
     	 used to trigger background recharging.

     (b) An indication that a call is being discarded.  This is used when
     	 the socket is being released.

     A function, rxrpc_kernel_charge_accept() is called by the kernel
     service to preallocate a single call.  It should be passed the user ID
     to be used for that call and a callback to associate the rxrpc call
     with the kernel service's side of the ID.

 (4) Discard the preallocation when the socket is closed.

 (5) Temporarily bump the refcount on the call allocated in
     rxrpc_incoming_call() so that rxrpc_release_call() can ditch the
     preallocation ref on service calls unconditionally.  This will no
     longer be necessary once the preallocation is used.

Note that this does not yet control the number of active service calls on a
client - that will come in a later patch.

A future development would be to provide a setsockopt() call that allows a
userspace server to manually charge the preallocation buffer.  This would
allow user call IDs to be provided in advance and the awkward manual accept
stage to be bypassed.

Signed-off-by: David Howells <dhowells@redhat.com>

Author: dhowells

fs/afs/rxrpc.c

@@ -18,6 +18,7 @@
 
 struct socket *afs_socket; /* my RxRPC socket */
 static struct workqueue_struct *afs_async_calls;
+static struct afs_call *afs_spare_incoming_call;
 static atomic_t afs_outstanding_calls;
 
 static void afs_free_call(struct afs_call *);
@@ -26,7 +27,8 @@ static int afs_wait_for_call_to_complete(struct afs_call *);
 static void afs_wake_up_async_call(struct sock *, struct rxrpc_call *, unsigned long);
 static int afs_dont_wait_for_call_to_complete(struct afs_call *);
 static void afs_process_async_call(struct work_struct *);
-static void afs_rx_new_call(struct sock *);
+static void afs_rx_new_call(struct sock *, struct rxrpc_call *, unsigned long);
+static void afs_rx_discard_new_call(struct rxrpc_call *, unsigned long);
 static int afs_deliver_cm_op_id(struct afs_call *);
 
 /* synchronous call management */
@@ -54,8 +56,10 @@ static const struct afs_call_type afs_RXCMxxxx = {
 };
 
 static void afs_collect_incoming_call(struct work_struct *);
+static void afs_charge_preallocation(struct work_struct *);
 
 static DECLARE_WORK(afs_collect_incoming_call_work, afs_collect_incoming_call);
+static DECLARE_WORK(afs_charge_preallocation_work, afs_charge_preallocation);
 
 static int afs_wait_atomic_t(atomic_t *p)
 {
@@ -100,13 +104,15 @@ int afs_open_socket(void)
 	if (ret < 0)
 		goto error_2;
 
-	rxrpc_kernel_new_call_notification(socket, afs_rx_new_call);
+	rxrpc_kernel_new_call_notification(socket, afs_rx_new_call,
+					   afs_rx_discard_new_call);
 
 	ret = kernel_listen(socket, INT_MAX);
 	if (ret < 0)
 		goto error_2;
 
 	afs_socket = socket;
+	afs_charge_preallocation(NULL);
 	_leave(" = 0");
 	return 0;
 
@@ -126,6 +132,12 @@ void afs_close_socket(void)
 {
 	_enter("");
 
+	if (afs_spare_incoming_call) {
+		atomic_inc(&afs_outstanding_calls);
+		afs_free_call(afs_spare_incoming_call);
+		afs_spare_incoming_call = NULL;
+	}
+
 	_debug("outstanding %u", atomic_read(&afs_outstanding_calls));
 	wait_on_atomic_t(&afs_outstanding_calls, afs_wait_atomic_t,
 			 TASK_UNINTERRUPTIBLE);
@@ -635,12 +647,65 @@ static void afs_collect_incoming_call(struct work_struct *work)
 		afs_free_call(call);
 }
 
+static void afs_rx_attach(struct rxrpc_call *rxcall, unsigned long user_call_ID)
+{
+	struct afs_call *call = (struct afs_call *)user_call_ID;
+
+	call->rxcall = rxcall;
+}
+
+/*
+ * Charge the incoming call preallocation.
+ */
+static void afs_charge_preallocation(struct work_struct *work)
+{
+	struct afs_call *call = afs_spare_incoming_call;
+
+	for (;;) {
+		if (!call) {
+			call = kzalloc(sizeof(struct afs_call), GFP_KERNEL);
+			if (!call)
+				break;
+
+			INIT_WORK(&call->async_work, afs_process_async_call);
+			call->wait_mode = &afs_async_incoming_call;
+			call->type = &afs_RXCMxxxx;
+			init_waitqueue_head(&call->waitq);
+			call->state = AFS_CALL_AWAIT_OP_ID;
+		}
+
+		if (rxrpc_kernel_charge_accept(afs_socket,
+					       afs_wake_up_async_call,
+					       afs_rx_attach,
+					       (unsigned long)call,
+					       GFP_KERNEL) < 0)
+			break;
+		call = NULL;
+	}
+	afs_spare_incoming_call = call;
+}
+
+/*
+ * Discard a preallocated call when a socket is shut down.
+ */
+static void afs_rx_discard_new_call(struct rxrpc_call *rxcall,
+				    unsigned long user_call_ID)
+{
+	struct afs_call *call = (struct afs_call *)user_call_ID;
+
+	atomic_inc(&afs_outstanding_calls);
+	call->rxcall = NULL;
+	afs_free_call(call);
+}
+
 /*
  * Notification of an incoming call.
  */
-static void afs_rx_new_call(struct sock *sk)
+static void afs_rx_new_call(struct sock *sk, struct rxrpc_call *rxcall,
+			    unsigned long user_call_ID)
 {
 	queue_work(afs_wq, &afs_collect_incoming_call_work);
+	queue_work(afs_wq, &afs_charge_preallocation_work);
 }
 
 /*

include/net/af_rxrpc.h

@@ -21,10 +21,14 @@ struct rxrpc_call;
 
 typedef void (*rxrpc_notify_rx_t)(struct sock *, struct rxrpc_call *,
 				  unsigned long);
-typedef void (*rxrpc_notify_new_call_t)(struct sock *);
+typedef void (*rxrpc_notify_new_call_t)(struct sock *, struct rxrpc_call *,
+					unsigned long);
+typedef void (*rxrpc_discard_new_call_t)(struct rxrpc_call *, unsigned long);
+typedef void (*rxrpc_user_attach_call_t)(struct rxrpc_call *, unsigned long);
 
 void rxrpc_kernel_new_call_notification(struct socket *,
-					rxrpc_notify_new_call_t);
+					rxrpc_notify_new_call_t,
+					rxrpc_discard_new_call_t);
 struct rxrpc_call *rxrpc_kernel_begin_call(struct socket *,
 					   struct sockaddr_rxrpc *,
 					   struct key *,
@@ -43,5 +47,7 @@ struct rxrpc_call *rxrpc_kernel_accept_call(struct socket *, unsigned long,
 int rxrpc_kernel_reject_call(struct socket *);
 void rxrpc_kernel_get_peer(struct socket *, struct rxrpc_call *,
 			   struct sockaddr_rxrpc *);
+int rxrpc_kernel_charge_accept(struct socket *, rxrpc_notify_rx_t,
+			       rxrpc_user_attach_call_t, unsigned long, gfp_t);
 
 #endif /* _NET_RXRPC_H */

net/rxrpc/af_rxrpc.c

@@ -193,7 +193,7 @@ static int rxrpc_listen(struct socket *sock, int backlog)
 {
 	struct sock *sk = sock->sk;
 	struct rxrpc_sock *rx = rxrpc_sk(sk);
-	unsigned int max;
+	unsigned int max, old;
 	int ret;
 
 	_enter("%p,%d", rx, backlog);
@@ -212,9 +212,13 @@ static int rxrpc_listen(struct socket *sock, int backlog)
 			backlog = max;
 		else if (backlog < 0 || backlog > max)
 			break;
+		old = sk->sk_max_ack_backlog;
 		sk->sk_max_ack_backlog = backlog;
-		rx->sk.sk_state = RXRPC_SERVER_LISTENING;
-		ret = 0;
+		ret = rxrpc_service_prealloc(rx, GFP_KERNEL);
+		if (ret == 0)
+			rx->sk.sk_state = RXRPC_SERVER_LISTENING;
+		else
+			sk->sk_max_ack_backlog = old;
 		break;
 	default:
 		ret = -EBUSY;
@@ -303,16 +307,19 @@ EXPORT_SYMBOL(rxrpc_kernel_end_call);
  * rxrpc_kernel_new_call_notification - Get notifications of new calls
  * @sock: The socket to intercept received messages on
  * @notify_new_call: Function to be called when new calls appear
+ * @discard_new_call: Function to discard preallocated calls
  *
  * Allow a kernel service to be given notifications about new calls.
  */
 void rxrpc_kernel_new_call_notification(
 	struct socket *sock,
-	rxrpc_notify_new_call_t notify_new_call)
+	rxrpc_notify_new_call_t notify_new_call,
+	rxrpc_discard_new_call_t discard_new_call)
 {
 	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
 
 	rx->notify_new_call = notify_new_call;
+	rx->discard_new_call = discard_new_call;
 }
 EXPORT_SYMBOL(rxrpc_kernel_new_call_notification);
 
@@ -622,6 +629,7 @@ static int rxrpc_release_sock(struct sock *sk)
 	}
 
 	/* try to flush out this socket */
+	rxrpc_discard_prealloc(rx);
 	rxrpc_release_calls_on_socket(rx);
 	flush_workqueue(rxrpc_workqueue);
 	rxrpc_purge_queue(&sk->sk_receive_queue);

net/rxrpc/ar-internal.h

@@ -63,20 +63,43 @@ enum {
 	RXRPC_CLOSE,			/* socket is being closed */
 };
 
+/*
+ * Service backlog preallocation.
+ *
+ * This contains circular buffers of preallocated peers, connections and calls
+ * for incoming service calls and their head and tail pointers.  This allows
+ * calls to be set up in the data_ready handler, thereby avoiding the need to
+ * shuffle packets around so much.
+ */
+struct rxrpc_backlog {
+	unsigned short		peer_backlog_head;
+	unsigned short		peer_backlog_tail;
+	unsigned short		conn_backlog_head;
+	unsigned short		conn_backlog_tail;
+	unsigned short		call_backlog_head;
+	unsigned short		call_backlog_tail;
+#define RXRPC_BACKLOG_MAX	32
+	struct rxrpc_peer	*peer_backlog[RXRPC_BACKLOG_MAX];
+	struct rxrpc_connection	*conn_backlog[RXRPC_BACKLOG_MAX];
+	struct rxrpc_call	*call_backlog[RXRPC_BACKLOG_MAX];
+};
+
 /*
  * RxRPC socket definition
  */
 struct rxrpc_sock {
 	/* WARNING: sk has to be the first member */
 	struct sock		sk;
 	rxrpc_notify_new_call_t	notify_new_call; /* Func to notify of new call */
+	rxrpc_discard_new_call_t discard_new_call; /* Func to discard a new call */
 	struct rxrpc_local	*local;		/* local endpoint */
 	struct hlist_node	listen_link;	/* link in the local endpoint's listen list */
 	struct list_head	secureq;	/* calls awaiting connection security clearance */
 	struct list_head	acceptq;	/* calls awaiting acceptance */
+	struct rxrpc_backlog	*backlog;	/* Preallocation for services */
 	struct key		*key;		/* security for this socket */
 	struct key		*securities;	/* list of server security descriptors */
-	struct rb_root		calls;		/* outstanding calls on this socket */
+	struct rb_root		calls;		/* User ID -> call mapping */
 	unsigned long		flags;
 #define RXRPC_SOCK_CONNECTED		0	/* connect_srx is set */
 	rwlock_t		call_lock;	/* lock for calls */
@@ -290,6 +313,7 @@ enum rxrpc_conn_cache_state {
 enum rxrpc_conn_proto_state {
 	RXRPC_CONN_UNUSED,		/* Connection not yet attempted */
 	RXRPC_CONN_CLIENT,		/* Client connection */
+	RXRPC_CONN_SERVICE_PREALLOC,	/* Service connection preallocation */
 	RXRPC_CONN_SERVICE_UNSECURED,	/* Service unsecured connection */
 	RXRPC_CONN_SERVICE_CHALLENGING,	/* Service challenging for security */
 	RXRPC_CONN_SERVICE,		/* Service secured connection */
@@ -408,6 +432,7 @@ enum rxrpc_call_state {
 	RXRPC_CALL_CLIENT_AWAIT_REPLY,	/* - client awaiting reply */
 	RXRPC_CALL_CLIENT_RECV_REPLY,	/* - client receiving reply phase */
 	RXRPC_CALL_CLIENT_FINAL_ACK,	/* - client sending final ACK phase */
+	RXRPC_CALL_SERVER_PREALLOC,	/* - service preallocation */
 	RXRPC_CALL_SERVER_SECURING,	/* - server securing request connection */
 	RXRPC_CALL_SERVER_ACCEPTING,	/* - server accepting request */
 	RXRPC_CALL_SERVER_RECV_REQUEST,	/* - server receiving request */
@@ -534,6 +559,8 @@ extern struct workqueue_struct *rxrpc_workqueue;
 /*
  * call_accept.c
  */
+int rxrpc_service_prealloc(struct rxrpc_sock *, gfp_t);
+void rxrpc_discard_prealloc(struct rxrpc_sock *);
 void rxrpc_accept_incoming_calls(struct rxrpc_local *);
 struct rxrpc_call *rxrpc_accept_call(struct rxrpc_sock *, unsigned long,
 				     rxrpc_notify_rx_t);
@@ -557,6 +584,7 @@ extern struct list_head rxrpc_calls;
 extern rwlock_t rxrpc_call_lock;
 
 struct rxrpc_call *rxrpc_find_call_by_user_ID(struct rxrpc_sock *, unsigned long);
+struct rxrpc_call *rxrpc_alloc_call(gfp_t);
 struct rxrpc_call *rxrpc_new_client_call(struct rxrpc_sock *,
 					 struct rxrpc_conn_parameters *,
 					 struct sockaddr_rxrpc *,
@@ -573,6 +601,7 @@ void rxrpc_get_call(struct rxrpc_call *, enum rxrpc_call_trace);
 void rxrpc_put_call(struct rxrpc_call *, enum rxrpc_call_trace);
 void rxrpc_get_call_for_skb(struct rxrpc_call *, struct sk_buff *);
 void rxrpc_put_call_for_skb(struct rxrpc_call *, struct sk_buff *);
+void rxrpc_cleanup_call(struct rxrpc_call *);
 void __exit rxrpc_destroy_all_calls(void);
 
 static inline bool rxrpc_is_service_call(const struct rxrpc_call *call)
@@ -757,6 +786,7 @@ struct rxrpc_connection *rxrpc_find_service_conn_rcu(struct rxrpc_peer *,
 struct rxrpc_connection *rxrpc_incoming_connection(struct rxrpc_local *,
 						   struct sockaddr_rxrpc *,
 						   struct sk_buff *);
+struct rxrpc_connection *rxrpc_prealloc_service_connection(gfp_t);
 void rxrpc_unpublish_service_conn(struct rxrpc_connection *);
 
 /*