rxrpc: Rewrite the data and ack handling code

Rewrite the data and ack handling code such that:

 (1) Parsing of received ACK and ABORT packets and the distribution and the
     filing of DATA packets happens entirely within the data_ready context
     called from the UDP socket.  This allows us to process and discard ACK
     and ABORT packets much more quickly (they're no longer stashed on a
     queue for a background thread to process).

 (2) We avoid calling skb_clone(), pskb_pull() and pskb_trim().  We instead
     keep track of the offset and length of the content of each packet in
     the sk_buff metadata.  This means we don't do any allocation in the
     receive path.

 (3) Jumbo DATA packet parsing is now done in data_ready context.  Rather
     than cloning the packet once for each subpacket and pulling/trimming
     it, we file the packet multiple times with an annotation for each
     indicating which subpacket is there.  From that we can directly
     calculate the offset and length.

 (4) A call's receive queue can be accessed without taking locks (memory
     barriers do have to be used, though).

 (5) Incoming calls are set up from preallocated resources and immediately
     made live.  They can than have packets queued upon them and ACKs
     generated.  If insufficient resources exist, DATA packet #1 is given a
     BUSY reply and other DATA packets are discarded).

 (6) sk_buffs no longer take a ref on their parent call.

To make this work, the following changes are made:

 (1) Each call's receive buffer is now a circular buffer of sk_buff
     pointers (rxtx_buffer) rather than a number of sk_buff_heads spread
     between the call and the socket.  This permits each sk_buff to be in
     the buffer multiple times.  The receive buffer is reused for the
     transmit buffer.

 (2) A circular buffer of annotations (rxtx_annotations) is kept parallel
     to the data buffer.  Transmission phase annotations indicate whether a
     buffered packet has been ACK'd or not and whether it needs
     retransmission.

     Receive phase annotations indicate whether a slot holds a whole packet
     or a jumbo subpacket and, if the latter, which subpacket.  They also
     note whether the packet has been decrypted in place.

 (3) DATA packet window tracking is much simplified.  Each phase has just
     two numbers representing the window (rx_hard_ack/rx_top and
     tx_hard_ack/tx_top).

     The hard_ack number is the sequence number before base of the window,
     representing the last packet the other side says it has consumed.
     hard_ack starts from 0 and the first packet is sequence number 1.

     The top number is the sequence number of the highest-numbered packet
     residing in the buffer.  Packets between hard_ack+1 and top are
     soft-ACK'd to indicate they've been received, but not yet consumed.

     Four macros, before(), before_eq(), after() and after_eq() are added
     to compare sequence numbers within the window.  This allows for the
     top of the window to wrap when the hard-ack sequence number gets close
     to the limit.

     Two flags, RXRPC_CALL_RX_LAST and RXRPC_CALL_TX_LAST, are added also
     to indicate when rx_top and tx_top point at the packets with the
     LAST_PACKET bit set, indicating the end of the phase.

 (4) Calls are queued on the socket 'receive queue' rather than packets.
     This means that we don't need have to invent dummy packets to queue to
     indicate abnormal/terminal states and we don't have to keep metadata
     packets (such as ABORTs) around

 (5) The offset and length of a (sub)packet's content are now passed to
     the verify_packet security op.  This is currently expected to decrypt
     the packet in place and validate it.

     However, there's now nowhere to store the revised offset and length of
     the actual data within the decrypted blob (there may be a header and
     padding to skip) because an sk_buff may represent multiple packets, so
     a locate_data security op is added to retrieve these details from the
     sk_buff content when needed.

 (6) recvmsg() now has to handle jumbo subpackets, where each subpacket is
     individually secured and needs to be individually decrypted.  The code
     to do this is broken out into rxrpc_recvmsg_data() and shared with the
     kernel API.  It now iterates over the call's receive buffer rather
     than walking the socket receive queue.

Additional changes:

 (1) The timers are condensed to a single timer that is set for the soonest
     of three timeouts (delayed ACK generation, DATA retransmission and
     call lifespan).

 (2) Transmission of ACK and ABORT packets is effected immediately from
     process-context socket ops/kernel API calls that cause them instead of
     them being punted off to a background work item.  The data_ready
     handler still has to defer to the background, though.

 (3) A shutdown op is added to the AF_RXRPC socket so that the AFS
     filesystem can shut down the socket and flush its own work items
     before closing the socket to deal with any in-progress service calls.

Future additional changes that will need to be considered:

 (1) Make sure that a call doesn't hog the front of the queue by receiving
     data from the network as fast as userspace is consuming it to the
     exclusion of other calls.

 (2) Transmit delayed ACKs from within recvmsg() when we've consumed
     sufficiently more packets to avoid the background work item needing to
     run.

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

1 files changed, 424 insertions, 340 deletions

diff --git a/net/rxrpc/recvmsg.c b/net/rxrpc/recvmsg.c
index 6876ffb3b410..20d0b5c6f81b 100644
--- a/net/rxrpc/recvmsg.c
+++ b/net/rxrpc/recvmsg.c
@@ -19,319 +19,479 @@
 #include "ar-internal.h"
 
 /*
- * receive a message from an RxRPC socket
- * - we need to be careful about two or more threads calling recvmsg
- *   simultaneously
+ * Post a call for attention by the socket or kernel service.  Further
+ * notifications are suppressed by putting recvmsg_link on a dummy queue.
  */
-int rxrpc_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
-		  int flags)
+void rxrpc_notify_socket(struct rxrpc_call *call)
 {
-	struct rxrpc_skb_priv *sp;
-	struct rxrpc_call *call = NULL, *continue_call = NULL;
-	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
-	struct sk_buff *skb;
-	long timeo;
-	int copy, ret, ullen, offset, copied = 0;
-	u32 abort_code;
+	struct rxrpc_sock *rx;
+	struct sock *sk;
 
-	DEFINE_WAIT(wait);
+	_enter("%d", call->debug_id);
 
-	_enter(",,,%zu,%d", len, flags);
+	if (!list_empty(&call->recvmsg_link))
+		return;
+
+	rcu_read_lock();
+
+	rx = rcu_dereference(call->socket);
+	sk = &rx->sk;
+	if (rx && sk->sk_state < RXRPC_CLOSE) {
+		if (call->notify_rx) {
+			call->notify_rx(sk, call, call->user_call_ID);
+		} else {
+			write_lock_bh(&rx->recvmsg_lock);
+			if (list_empty(&call->recvmsg_link)) {
+				rxrpc_get_call(call, rxrpc_call_got);
+				list_add_tail(&call->recvmsg_link, &rx->recvmsg_q);
+			}
+			write_unlock_bh(&rx->recvmsg_lock);
 
-	if (flags & (MSG_OOB | MSG_TRUNC))
-		return -EOPNOTSUPP;
+			if (!sock_flag(sk, SOCK_DEAD)) {
+				_debug("call %ps", sk->sk_data_ready);
+				sk->sk_data_ready(sk);
+			}
+		}
+	}
 
-	ullen = msg->msg_flags & MSG_CMSG_COMPAT ? 4 : sizeof(unsigned long);
+	rcu_read_unlock();
+	_leave("");
+}
 
-	timeo = sock_rcvtimeo(&rx->sk, flags & MSG_DONTWAIT);
-	msg->msg_flags |= MSG_MORE;
+/*
+ * Pass a call terminating message to userspace.
+ */
+static int rxrpc_recvmsg_term(struct rxrpc_call *call, struct msghdr *msg)
+{
+	u32 tmp = 0;
+	int ret;
 
-	lock_sock(&rx->sk);
+	switch (call->completion) {
+	case RXRPC_CALL_SUCCEEDED:
+		ret = 0;
+		if (rxrpc_is_service_call(call))
+			ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ACK, 0, &tmp);
+		break;
+	case RXRPC_CALL_REMOTELY_ABORTED:
+		tmp = call->abort_code;
+		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
+		break;
+	case RXRPC_CALL_LOCALLY_ABORTED:
+		tmp = call->abort_code;
+		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &tmp);
+		break;
+	case RXRPC_CALL_NETWORK_ERROR:
+		tmp = call->error;
+		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NET_ERROR, 4, &tmp);
+		break;
+	case RXRPC_CALL_LOCAL_ERROR:
+		tmp = call->error;
+		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_LOCAL_ERROR, 4, &tmp);
+		break;
+	default:
+		pr_err("Invalid terminal call state %u\n", call->state);
+		BUG();
+		break;
+	}
 
-	for (;;) {
-		/* return immediately if a client socket has no outstanding
-		 * calls */
-		if (RB_EMPTY_ROOT(&rx->calls)) {
-			if (copied)
-				goto out;
-			if (rx->sk.sk_state != RXRPC_SERVER_LISTENING) {
-				release_sock(&rx->sk);
-				if (continue_call)
-					rxrpc_put_call(continue_call,
-						       rxrpc_call_put);
-				return -ENODATA;
-			}
-		}
+	return ret;
+}
 
-		/* get the next message on the Rx queue */
-		skb = skb_peek(&rx->sk.sk_receive_queue);
-		if (!skb) {
-			/* nothing remains on the queue */
-			if (copied &&
-			    (flags & MSG_PEEK || timeo == 0))
-				goto out;
+/*
+ * Pass back notification of a new call.  The call is added to the
+ * to-be-accepted list.  This means that the next call to be accepted might not
+ * be the last call seen awaiting acceptance, but unless we leave this on the
+ * front of the queue and block all other messages until someone gives us a
+ * user_ID for it, there's not a lot we can do.
+ */
+static int rxrpc_recvmsg_new_call(struct rxrpc_sock *rx,
+				  struct rxrpc_call *call,
+				  struct msghdr *msg, int flags)
+{
+	int tmp = 0, ret;
 
-			/* wait for a message to turn up */
-			release_sock(&rx->sk);
-			prepare_to_wait_exclusive(sk_sleep(&rx->sk), &wait,
-						  TASK_INTERRUPTIBLE);
-			ret = sock_error(&rx->sk);
-			if (ret)
-				goto wait_error;
-
-			if (skb_queue_empty(&rx->sk.sk_receive_queue)) {
-				if (signal_pending(current))
-					goto wait_interrupted;
-				timeo = schedule_timeout(timeo);
-			}
-			finish_wait(sk_sleep(&rx->sk), &wait);
-			lock_sock(&rx->sk);
-			continue;
-		}
+	ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NEW_CALL, 0, &tmp);
 
-	peek_next_packet:
-		rxrpc_see_skb(skb);
-		sp = rxrpc_skb(skb);
-		call = sp->call;
-		ASSERT(call != NULL);
-		rxrpc_see_call(call);
-
-		_debug("next pkt %s", rxrpc_pkts[sp->hdr.type]);
-
-		/* make sure we wait for the state to be updated in this call */
-		spin_lock_bh(&call->lock);
-		spin_unlock_bh(&call->lock);
-
-		if (test_bit(RXRPC_CALL_RELEASED, &call->flags)) {
-			_debug("packet from released call");
-			if (skb_dequeue(&rx->sk.sk_receive_queue) != skb)
-				BUG();
-			rxrpc_free_skb(skb);
-			continue;
-		}
+	if (ret == 0 && !(flags & MSG_PEEK)) {
+		_debug("to be accepted");
+		write_lock_bh(&rx->recvmsg_lock);
+		list_del_init(&call->recvmsg_link);
+		write_unlock_bh(&rx->recvmsg_lock);
 
-		/* determine whether to continue last data receive */
-		if (continue_call) {
-			_debug("maybe cont");
-			if (call != continue_call ||
-			    skb->mark != RXRPC_SKB_MARK_DATA) {
-				release_sock(&rx->sk);
-				rxrpc_put_call(continue_call, rxrpc_call_put);
-				_leave(" = %d [noncont]", copied);
-				return copied;
-			}
-		}
+		write_lock(&rx->call_lock);
+		list_add_tail(&call->accept_link, &rx->to_be_accepted);
+		write_unlock(&rx->call_lock);
+	}
 
-		rxrpc_get_call(call, rxrpc_call_got);
+	return ret;
+}
 
-		/* copy the peer address and timestamp */
-		if (!continue_call) {
-			if (msg->msg_name) {
-				size_t len =
-					sizeof(call->conn->params.peer->srx);
-				memcpy(msg->msg_name,
-				       &call->conn->params.peer->srx, len);
-				msg->msg_namelen = len;
-			}
-			sock_recv_timestamp(msg, &rx->sk, skb);
-		}
+/*
+ * End the packet reception phase.
+ */
+static void rxrpc_end_rx_phase(struct rxrpc_call *call)
+{
+	_enter("%d,%s", call->debug_id, rxrpc_call_states[call->state]);
 
-		/* receive the message */
-		if (skb->mark != RXRPC_SKB_MARK_DATA)
-			goto receive_non_data_message;
+	if (call->state == RXRPC_CALL_CLIENT_RECV_REPLY) {
+		rxrpc_propose_ACK(call, RXRPC_ACK_IDLE, 0, 0, true, false);
+		rxrpc_send_call_packet(call, RXRPC_PACKET_TYPE_ACK);
+	} else {
+		rxrpc_propose_ACK(call, RXRPC_ACK_IDLE, 0, 0, false, false);
+	}
 
-		_debug("recvmsg DATA #%u { %d, %d }",
-		       sp->hdr.seq, skb->len, sp->offset);
+	write_lock_bh(&call->state_lock);
 
-		if (!continue_call) {
-			/* only set the control data once per recvmsg() */
-			ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
-				       ullen, &call->user_call_ID);
-			if (ret < 0)
-				goto copy_error;
-			ASSERT(test_bit(RXRPC_CALL_HAS_USERID, &call->flags));
-		}
+	switch (call->state) {
+	case RXRPC_CALL_CLIENT_RECV_REPLY:
+		__rxrpc_call_completed(call);
+		break;
 
-		ASSERTCMP(sp->hdr.seq, >=, call->rx_data_recv);
-		ASSERTCMP(sp->hdr.seq, <=, call->rx_data_recv + 1);
-		call->rx_data_recv = sp->hdr.seq;
+	case RXRPC_CALL_SERVER_RECV_REQUEST:
+		call->state = RXRPC_CALL_SERVER_ACK_REQUEST;
+		break;
+	default:
+		break;
+	}
 
-		ASSERTCMP(sp->hdr.seq, >, call->rx_data_eaten);
+	write_unlock_bh(&call->state_lock);
+}
 
-		offset = sp->offset;
-		copy = skb->len - offset;
-		if (copy > len - copied)
-			copy = len - copied;
+/*
+ * Discard a packet we've used up and advance the Rx window by one.
+ */
+static void rxrpc_rotate_rx_window(struct rxrpc_call *call)
+{
+	struct sk_buff *skb;
+	rxrpc_seq_t hard_ack, top;
+	int ix;
+
+	_enter("%d", call->debug_id);
+
+	hard_ack = call->rx_hard_ack;
+	top = smp_load_acquire(&call->rx_top);
+	ASSERT(before(hard_ack, top));
+
+	hard_ack++;
+	ix = hard_ack & RXRPC_RXTX_BUFF_MASK;
+	skb = call->rxtx_buffer[ix];
+	rxrpc_see_skb(skb);
+	call->rxtx_buffer[ix] = NULL;
+	call->rxtx_annotations[ix] = 0;
+	/* Barrier against rxrpc_input_data(). */
+	smp_store_release(&call->rx_hard_ack, hard_ack);
 
-		ret = skb_copy_datagram_msg(skb, offset, msg, copy);
+	rxrpc_free_skb(skb);
 
+	_debug("%u,%u,%lx", hard_ack, top, call->flags);
+	if (hard_ack == top && test_bit(RXRPC_CALL_RX_LAST, &call->flags))
+		rxrpc_end_rx_phase(call);
+}
+
+/*
+ * Decrypt and verify a (sub)packet.  The packet's length may be changed due to
+ * padding, but if this is the case, the packet length will be resident in the
+ * socket buffer.  Note that we can't modify the master skb info as the skb may
+ * be the home to multiple subpackets.
+ */
+static int rxrpc_verify_packet(struct rxrpc_call *call, struct sk_buff *skb,
+			       u8 annotation,
+			       unsigned int offset, unsigned int len)
+{
+	struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
+	rxrpc_seq_t seq = sp->hdr.seq;
+	u16 cksum = sp->hdr.cksum;
+
+	_enter("");
+
+	/* For all but the head jumbo subpacket, the security checksum is in a
+	 * jumbo header immediately prior to the data.
+	 */
+	if ((annotation & RXRPC_RX_ANNO_JUMBO) > 1) {
+		__be16 tmp;
+		if (skb_copy_bits(skb, offset - 2, &tmp, 2) < 0)
+			BUG();
+		cksum = ntohs(tmp);
+		seq += (annotation & RXRPC_RX_ANNO_JUMBO) - 1;
+	}
+
+	return call->conn->security->verify_packet(call, skb, offset, len,
+						   seq, cksum);
+}
+
+/*
+ * Locate the data within a packet.  This is complicated by:
+ *
+ * (1) An skb may contain a jumbo packet - so we have to find the appropriate
+ *     subpacket.
+ *
+ * (2) The (sub)packets may be encrypted and, if so, the encrypted portion
+ *     contains an extra header which includes the true length of the data,
+ *     excluding any encrypted padding.
+ */
+static int rxrpc_locate_data(struct rxrpc_call *call, struct sk_buff *skb,
+			     u8 *_annotation,
+			     unsigned int *_offset, unsigned int *_len)
+{
+	struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
+	unsigned int offset = *_offset;
+	unsigned int len = *_len;
+	int ret;
+	u8 annotation = *_annotation;
+
+	if (offset > 0)
+		return 0;
+
+	/* Locate the subpacket */
+	offset = sp->offset;
+	len = skb->len - sp->offset;
+	if ((annotation & RXRPC_RX_ANNO_JUMBO) > 0) {
+		offset += (((annotation & RXRPC_RX_ANNO_JUMBO) - 1) *
+			   RXRPC_JUMBO_SUBPKTLEN);
+		len = (annotation & RXRPC_RX_ANNO_JLAST) ?
+			skb->len - offset : RXRPC_JUMBO_SUBPKTLEN;
+	}
+
+	if (!(annotation & RXRPC_RX_ANNO_VERIFIED)) {
+		ret = rxrpc_verify_packet(call, skb, annotation, offset, len);
 		if (ret < 0)
-			goto copy_error;
+			return ret;
+		*_annotation |= RXRPC_RX_ANNO_VERIFIED;
+	}
 
-		/* handle piecemeal consumption of data packets */
-		_debug("copied %d+%d", copy, copied);
+	*_offset = offset;
+	*_len = len;
+	call->conn->security->locate_data(call, skb, _offset, _len);
+	return 0;
+}
 
-		offset += copy;
-		copied += copy;
+/*
+ * Deliver messages to a call.  This keeps processing packets until the buffer
+ * is filled and we find either more DATA (returns 0) or the end of the DATA
+ * (returns 1).  If more packets are required, it returns -EAGAIN.
+ */
+static int rxrpc_recvmsg_data(struct socket *sock, struct rxrpc_call *call,
+			      struct msghdr *msg, struct iov_iter *iter,
+			      size_t len, int flags, size_t *_offset)
+{
+	struct rxrpc_skb_priv *sp;
+	struct sk_buff *skb;
+	rxrpc_seq_t hard_ack, top, seq;
+	size_t remain;
+	bool last;
+	unsigned int rx_pkt_offset, rx_pkt_len;
+	int ix, copy, ret = 0;
+
+	_enter("");
+
+	rx_pkt_offset = call->rx_pkt_offset;
+	rx_pkt_len = call->rx_pkt_len;
+
+	/* Barriers against rxrpc_input_data(). */
+	hard_ack = call->rx_hard_ack;
+	top = smp_load_acquire(&call->rx_top);
+	for (seq = hard_ack + 1; before_eq(seq, top); seq++) {
+		ix = seq & RXRPC_RXTX_BUFF_MASK;
+		skb = call->rxtx_buffer[ix];
+		if (!skb)
+			break;
+		smp_rmb();
+		rxrpc_see_skb(skb);
+		sp = rxrpc_skb(skb);
 
-		if (!(flags & MSG_PEEK))
-			sp->offset = offset;
+		if (msg)
+			sock_recv_timestamp(msg, sock->sk, skb);
+
+		ret = rxrpc_locate_data(call, skb, &call->rxtx_annotations[ix],
+					&rx_pkt_offset, &rx_pkt_len);
+		_debug("recvmsg %x DATA #%u { %d, %d }",
+		       sp->hdr.callNumber, seq, rx_pkt_offset, rx_pkt_len);
+
+		/* We have to handle short, empty and used-up DATA packets. */
+		remain = len - *_offset;
+		copy = rx_pkt_len;
+		if (copy > remain)
+			copy = remain;
+		if (copy > 0) {
+			ret = skb_copy_datagram_iter(skb, rx_pkt_offset, iter,
+						     copy);
+			if (ret < 0)
+				goto out;
+
+			/* handle piecemeal consumption of data packets */
+			_debug("copied %d @%zu", copy, *_offset);
+
+			rx_pkt_offset += copy;
+			rx_pkt_len -= copy;
+			*_offset += copy;
+		}
 
-		if (sp->offset < skb->len) {
+		if (rx_pkt_len > 0) {
 			_debug("buffer full");
-			ASSERTCMP(copied, ==, len);
+			ASSERTCMP(*_offset, ==, len);
 			break;
 		}
 
-		/* we transferred the whole data packet */
+		/* The whole packet has been transferred. */
+		last = sp->hdr.flags & RXRPC_LAST_PACKET;
 		if (!(flags & MSG_PEEK))
-			rxrpc_kernel_data_consumed(call, skb);
-
-		if (sp->hdr.flags & RXRPC_LAST_PACKET) {
-			_debug("last");
-			if (rxrpc_conn_is_client(call->conn)) {
-				 /* last byte of reply received */
-				ret = copied;
-				goto terminal_message;
-			}
+			rxrpc_rotate_rx_window(call);
+		rx_pkt_offset = 0;
+		rx_pkt_len = 0;
 
-			/* last bit of request received */
-			if (!(flags & MSG_PEEK)) {
-				_debug("eat packet");
-				if (skb_dequeue(&rx->sk.sk_receive_queue) !=
-				    skb)
-					BUG();
-				rxrpc_free_skb(skb);
-			}
-			msg->msg_flags &= ~MSG_MORE;
-			break;
-		}
+		ASSERTIFCMP(last, seq, ==, top);
+	}
 
-		/* move on to the next data message */
-		_debug("next");
-		if (!continue_call)
-			continue_call = sp->call;
-		else
-			rxrpc_put_call(call, rxrpc_call_put);
-		call = NULL;
-
-		if (flags & MSG_PEEK) {
-			_debug("peek next");
-			skb = skb->next;
-			if (skb == (struct sk_buff *) &rx->sk.sk_receive_queue)
-				break;
-			goto peek_next_packet;
-		}
+	if (after(seq, top)) {
+		ret = -EAGAIN;
+		if (test_bit(RXRPC_CALL_RX_LAST, &call->flags))
+			ret = 1;
+	}
+out:
+	if (!(flags & MSG_PEEK)) {
+		call->rx_pkt_offset = rx_pkt_offset;
+		call->rx_pkt_len = rx_pkt_len;
+	}
+	_leave(" = %d [%u/%u]", ret, seq, top);
+	return ret;
+}
 
-		_debug("eat packet");
-		if (skb_dequeue(&rx->sk.sk_receive_queue) != skb)
-			BUG();
-		rxrpc_free_skb(skb);
+/*
+ * Receive a message from an RxRPC socket
+ * - we need to be careful about two or more threads calling recvmsg
+ *   simultaneously
+ */
+int rxrpc_recvmsg(struct socket *sock, struct msghdr *msg, size_t len,
+		  int flags)
+{
+	struct rxrpc_call *call;
+	struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
+	struct list_head *l;
+	size_t copied = 0;
+	long timeo;
+	int ret;
+
+	DEFINE_WAIT(wait);
+
+	_enter(",,,%zu,%d", len, flags);
+
+	if (flags & (MSG_OOB | MSG_TRUNC))
+		return -EOPNOTSUPP;
+
+	timeo = sock_rcvtimeo(&rx->sk, flags & MSG_DONTWAIT);
+
+try_again:
+	lock_sock(&rx->sk);
+
+	/* Return immediately if a client socket has no outstanding calls */
+	if (RB_EMPTY_ROOT(&rx->calls) &&
+	    list_empty(&rx->recvmsg_q) &&
+	    rx->sk.sk_state != RXRPC_SERVER_LISTENING) {
+		release_sock(&rx->sk);
+		return -ENODATA;
 	}
 
-	/* end of non-terminal data packet reception for the moment */
-	_debug("end rcv data");
-out:
-	release_sock(&rx->sk);
-	if (call)
-		rxrpc_put_call(call, rxrpc_call_put);
-	if (continue_call)
-		rxrpc_put_call(continue_call, rxrpc_call_put);
-	_leave(" = %d [data]", copied);
-	return copied;
-
-	/* handle non-DATA messages such as aborts, incoming connections and
-	 * final ACKs */
-receive_non_data_message:
-	_debug("non-data");
-
-	if (skb->mark == RXRPC_SKB_MARK_NEW_CALL) {
-		_debug("RECV NEW CALL");
-		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NEW_CALL, 0, &abort_code);
-		if (ret < 0)
-			goto copy_error;
-		if (!(flags & MSG_PEEK)) {
-			if (skb_dequeue(&rx->sk.sk_receive_queue) != skb)
-				BUG();
-			rxrpc_free_skb(skb);
+	if (list_empty(&rx->recvmsg_q)) {
+		ret = -EWOULDBLOCK;
+		if (timeo == 0)
+			goto error_no_call;
+
+		release_sock(&rx->sk);
+
+		/* Wait for something to happen */
+		prepare_to_wait_exclusive(sk_sleep(&rx->sk), &wait,
+					  TASK_INTERRUPTIBLE);
+		ret = sock_error(&rx->sk);
+		if (ret)
+			goto wait_error;
+
+		if (list_empty(&rx->recvmsg_q)) {
+			if (signal_pending(current))
+				goto wait_interrupted;
+			timeo = schedule_timeout(timeo);
 		}
-		goto out;
+		finish_wait(sk_sleep(&rx->sk), &wait);
+		goto try_again;
 	}
 
-	ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
-		       ullen, &call->user_call_ID);
-	if (ret < 0)
-		goto copy_error;
-	ASSERT(test_bit(RXRPC_CALL_HAS_USERID, &call->flags));
+	/* Find the next call and dequeue it if we're not just peeking.  If we
+	 * do dequeue it, that comes with a ref that we will need to release.
+	 */
+	write_lock_bh(&rx->recvmsg_lock);
+	l = rx->recvmsg_q.next;
+	call = list_entry(l, struct rxrpc_call, recvmsg_link);
+	if (!(flags & MSG_PEEK))
+		list_del_init(&call->recvmsg_link);
+	else
+		rxrpc_get_call(call, rxrpc_call_got);
+	write_unlock_bh(&rx->recvmsg_lock);
 
-	switch (skb->mark) {
-	case RXRPC_SKB_MARK_DATA:
+	_debug("recvmsg call %p", call);
+
+	if (test_bit(RXRPC_CALL_RELEASED, &call->flags))
 		BUG();
-	case RXRPC_SKB_MARK_FINAL_ACK:
-		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ACK, 0, &abort_code);
-		break;
-	case RXRPC_SKB_MARK_BUSY:
-		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_BUSY, 0, &abort_code);
-		break;
-	case RXRPC_SKB_MARK_REMOTE_ABORT:
-		abort_code = call->abort_code;
-		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &abort_code);
-		break;
-	case RXRPC_SKB_MARK_LOCAL_ABORT:
-		abort_code = call->abort_code;
-		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &abort_code);
-		if (call->error) {
-			abort_code = call->error;
-			ret = put_cmsg(msg, SOL_RXRPC, RXRPC_LOCAL_ERROR, 4,
-				       &abort_code);
+
+	if (test_bit(RXRPC_CALL_HAS_USERID, &call->flags)) {
+		if (flags & MSG_CMSG_COMPAT) {
+			unsigned int id32 = call->user_call_ID;
+
+			ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
+				       sizeof(unsigned int), &id32);
+		} else {
+			ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
+				       sizeof(unsigned long),
+				       &call->user_call_ID);
 		}
+		if (ret < 0)
+			goto error;
+	}
+
+	if (msg->msg_name) {
+		size_t len = sizeof(call->conn->params.peer->srx);
+		memcpy(msg->msg_name, &call->conn->params.peer->srx, len);
+		msg->msg_namelen = len;
+	}
+
+	switch (call->state) {
+	case RXRPC_CALL_SERVER_ACCEPTING:
+		ret = rxrpc_recvmsg_new_call(rx, call, msg, flags);
 		break;
-	case RXRPC_SKB_MARK_NET_ERROR:
-		_debug("RECV NET ERROR %d", sp->error);
-		abort_code = sp->error;
-		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NET_ERROR, 4, &abort_code);
-		break;
-	case RXRPC_SKB_MARK_LOCAL_ERROR:
-		_debug("RECV LOCAL ERROR %d", sp->error);
-		abort_code = sp->error;
-		ret = put_cmsg(msg, SOL_RXRPC, RXRPC_LOCAL_ERROR, 4,
-			       &abort_code);
+	case RXRPC_CALL_CLIENT_RECV_REPLY:
+	case RXRPC_CALL_SERVER_RECV_REQUEST:
+	case RXRPC_CALL_SERVER_ACK_REQUEST:
+		ret = rxrpc_recvmsg_data(sock, call, msg, &msg->msg_iter, len,
+					 flags, &copied);
+		if (ret == -EAGAIN)
+			ret = 0;
 		break;
 	default:
-		pr_err("Unknown packet mark %u\n", skb->mark);
-		BUG();
+		ret = 0;
 		break;
 	}
 
 	if (ret < 0)
-		goto copy_error;
-
-terminal_message:
-	_debug("terminal");
-	msg->msg_flags &= ~MSG_MORE;
-	msg->msg_flags |= MSG_EOR;
+		goto error;
 
-	if (!(flags & MSG_PEEK)) {
-		_net("free terminal skb %p", skb);
-		if (skb_dequeue(&rx->sk.sk_receive_queue) != skb)
-			BUG();
-		rxrpc_free_skb(skb);
-		rxrpc_release_call(rx, call);
+	if (call->state == RXRPC_CALL_COMPLETE) {
+		ret = rxrpc_recvmsg_term(call, msg);
+		if (ret < 0)
+			goto error;
+		if (!(flags & MSG_PEEK))
+			rxrpc_release_call(rx, call);
+		msg->msg_flags |= MSG_EOR;
+		ret = 1;
 	}
 
-	release_sock(&rx->sk);
-	rxrpc_put_call(call, rxrpc_call_put);
-	if (continue_call)
-		rxrpc_put_call(continue_call, rxrpc_call_put);
-	_leave(" = %d", ret);
-	return ret;
+	if (ret == 0)
+		msg->msg_flags |= MSG_MORE;
+	else
+		msg->msg_flags &= ~MSG_MORE;
+	ret = copied;
 
-copy_error:
-	_debug("copy error");
-	release_sock(&rx->sk);
+error:
 	rxrpc_put_call(call, rxrpc_call_put);
-	if (continue_call)
-		rxrpc_put_call(continue_call, rxrpc_call_put);
+error_no_call:
+	release_sock(&rx->sk);
 	_leave(" = %d", ret);
 	return ret;
 
@@ -339,85 +499,8 @@ wait_interrupted:
 	ret = sock_intr_errno(timeo);
 wait_error:
 	finish_wait(sk_sleep(&rx->sk), &wait);
-	if (continue_call)
-		rxrpc_put_call(continue_call, rxrpc_call_put);
-	if (copied)
-		copied = ret;
-	_leave(" = %d [waitfail %d]", copied, ret);
-	return copied;
-
-}
-
-/*
- * Deliver messages to a call.  This keeps processing packets until the buffer
- * is filled and we find either more DATA (returns 0) or the end of the DATA
- * (returns 1).  If more packets are required, it returns -EAGAIN.
- *
- * TODO: Note that this is hacked in at the moment and will be replaced.
- */
-static int temp_deliver_data(struct socket *sock, struct rxrpc_call *call,
-			     struct iov_iter *iter, size_t size,
-			     size_t *_offset)
-{
-	struct rxrpc_skb_priv *sp;
-	struct sk_buff *skb;
-	size_t remain;
-	int ret, copy;
-
-	_enter("%d", call->debug_id);
-
-next:
-	local_bh_disable();
-	skb = skb_dequeue(&call->knlrecv_queue);
-	local_bh_enable();
-	if (!skb) {
-		if (test_bit(RXRPC_CALL_RX_NO_MORE, &call->flags))
-			return 1;
-		_leave(" = -EAGAIN [empty]");
-		return -EAGAIN;
-	}
-
-	sp = rxrpc_skb(skb);
-	_debug("dequeued %p %u/%zu", skb, sp->offset, size);
-
-	switch (skb->mark) {
-	case RXRPC_SKB_MARK_DATA:
-		remain = size - *_offset;
-		if (remain > 0) {
-			copy = skb->len - sp->offset;
-			if (copy > remain)
-				copy = remain;
-			ret = skb_copy_datagram_iter(skb, sp->offset, iter,
-						     copy);
-			if (ret < 0)
-				goto requeue_and_leave;
-
-			/* handle piecemeal consumption of data packets */
-			sp->offset += copy;
-			*_offset += copy;
-		}
-
-		if (sp->offset < skb->len)
-			goto partially_used_skb;
-
-		/* We consumed the whole packet */
-		ASSERTCMP(sp->offset, ==, skb->len);
-		if (sp->hdr.flags & RXRPC_LAST_PACKET)
-			set_bit(RXRPC_CALL_RX_NO_MORE, &call->flags);
-		rxrpc_kernel_data_consumed(call, skb);
-		rxrpc_free_skb(skb);
-		goto next;
-
-	default:
-		rxrpc_free_skb(skb);
-		goto next;
-	}
-
-partially_used_skb:
-	ASSERTCMP(*_offset, ==, size);
-	ret = 0;
-requeue_and_leave:
-	skb_queue_head(&call->knlrecv_queue, skb);
+	release_sock(&rx->sk);
+	_leave(" = %d [wait]", ret);
 	return ret;
 }
 
@@ -453,8 +536,9 @@ int rxrpc_kernel_recv_data(struct socket *sock, struct rxrpc_call *call,
 	struct kvec iov;
 	int ret;
 
-	_enter("{%d,%s},%zu,%d",
-	       call->debug_id, rxrpc_call_states[call->state], size, want_more);
+	_enter("{%d,%s},%zu/%zu,%d",
+	       call->debug_id, rxrpc_call_states[call->state],
+	       *_offset, size, want_more);
 
 	ASSERTCMP(*_offset, <=, size);
 	ASSERTCMP(call->state, !=, RXRPC_CALL_SERVER_ACCEPTING);
@@ -469,7 +553,8 @@ int rxrpc_kernel_recv_data(struct socket *sock, struct rxrpc_call *call,
 	case RXRPC_CALL_CLIENT_RECV_REPLY:
 	case RXRPC_CALL_SERVER_RECV_REQUEST:
 	case RXRPC_CALL_SERVER_ACK_REQUEST:
-		ret = temp_deliver_data(sock, call, &iter, size, _offset);
+		ret = rxrpc_recvmsg_data(sock, call, NULL, &iter, size, 0,
+					 _offset);
 		if (ret < 0)
 			goto out;
 
@@ -494,7 +579,6 @@ int rxrpc_kernel_recv_data(struct socket *sock, struct rxrpc_call *call,
 		goto call_complete;
 
 	default:
-		*_offset = 0;
 		ret = -EINPROGRESS;
 		goto out;
 	}