nfsd41: sunrpc: Added rpc server-side backchannel handling

When the call direction is a reply, copy the xid and call direction into the
req->rq_private_buf.head[0].iov_base otherwise rpc_verify_header returns
rpc_garbage.

Signed-off-by: Rahul Iyer <[e-mail]>
Signed-off-by: Mike Sager <[e-mail]>
Signed-off-by: Marc Eshel <[e-mail]>
Signed-off-by: Benny Halevy <[e-mail]>
Signed-off-by: Ricardo Labiaga <[e-mail]>
Signed-off-by: Andy Adamson <[e-mail]>
Signed-off-by: Benny Halevy <[e-mail]>
[get rid of CONFIG_NFSD_V4_1]
[sunrpc: refactoring of svc_tcp_recvfrom]
[nfsd41: sunrpc: create common send routine for the fore and the back channels]
[nfsd41: sunrpc: Use free_page() to free server backchannel pages]
[nfsd41: sunrpc: Document server backchannel locking]
[nfsd41: sunrpc: remove bc_connect_worker()]
[nfsd41: sunrpc: Define xprt_server_backchannel()[
[nfsd41: sunrpc: remove bc_close and bc_init_auto_disconnect dummy functions]
[nfsd41: sunrpc: eliminate unneeded switch statement in xs_setup_tcp()]
[nfsd41: sunrpc: Don't auto close the server backchannel connection]
[nfsd41: sunrpc: Remove unused functions]
Signed-off-by: Alexandros Batsakis <[e-mail]>
Signed-off-by: Ricardo Labiaga <[e-mail]>
Signed-off-by: Benny Halevy <[e-mail]>
[nfsd41: change bc_sock to bc_xprt]
[nfsd41: sunrpc: move struct rpc_buffer def into a common header file]
[nfsd41: sunrpc: use rpc_sleep in bc_send_request so not to block on mutex]
[removed cosmetic changes]
Signed-off-by: Benny Halevy <[e-mail]>
[sunrpc: add new xprt class for nfsv4.1 backchannel]
[sunrpc: v2.1 change handling of auto_close and init_auto_disconnect operations for the nfsv4.1 backchannel]
Signed-off-by: Alexandros Batsakis <[e-mail]>
[reverted more cosmetic leftovers]
[got rid of xprt_server_backchannel]
[separated "nfsd41: sunrpc: add new xprt class for nfsv4.1 backchannel"]
Signed-off-by: Benny Halevy <[e-mail]>
Cc: Trond Myklebust <[e-mail]>
[sunrpc: change idle timeout value for the backchannel]
Signed-off-by: Alexandros Batsakis <[e-mail]>
Signed-off-by: Benny Halevy <[e-mail]>
Acked-by: Trond Myklebust <[e-mail]>
Signed-off-by: J. Bruce Fields <[e-mail]>

include/linux/sunrpc/svc_xprt.h

···
 	size_t			xpt_locallen;	/* length of address */
 	struct sockaddr_storage	xpt_remote;	/* remote peer's address */
 	size_t			xpt_remotelen;	/* length of address */
+	struct rpc_wait_queue	xpt_bc_pending;	/* backchannel wait queue */
 };
 
 int	svc_reg_xprt_class(struct svc_xprt_class *);

include/linux/sunrpc/svcsock.h

···
 	/* private TCP part */
 	u32			sk_reclen;	/* length of record */
 	u32			sk_tcplen;	/* current read length */
+	struct rpc_xprt		*sk_bc_xprt;	/* NFSv4.1 backchannel xprt */
 };
 
 /*

include/linux/sunrpc/xprt.h

···
 	spinlock_t		reserve_lock;	/* lock slot table */
 	u32			xid;		/* Next XID value to use */
 	struct rpc_task *	snd_task;	/* Task blocked in send */
+	struct svc_xprt		*bc_xprt;	/* NFSv4.1 backchannel */
 #if defined(CONFIG_NFS_V4_1)
 	struct svc_serv		*bc_serv;       /* The RPC service which will */
 						/* process the callback */

net/sunrpc/sunrpc.h

···
 		(task->tk_msg.rpc_proc->p_decode != NULL);
 }
 
+int svc_send_common(struct socket *sock, struct xdr_buf *xdr,
+		    struct page *headpage, unsigned long headoffset,
+		    struct page *tailpage, unsigned long tailoffset);
+
 #endif /* _NET_SUNRPC_SUNRPC_H */
 

net/sunrpc/svc_xprt.c

···
 	mutex_init(&xprt->xpt_mutex);
 	spin_lock_init(&xprt->xpt_lock);
 	set_bit(XPT_BUSY, &xprt->xpt_flags);
+	rpc_init_wait_queue(&xprt->xpt_bc_pending, "xpt_bc_pending");
 }
 EXPORT_SYMBOL_GPL(svc_xprt_init);
 
···
 	else
 		len = xprt->xpt_ops->xpo_sendto(rqstp);
 	mutex_unlock(&xprt->xpt_mutex);
+	rpc_wake_up(&xprt->xpt_bc_pending);
 	svc_xprt_release(rqstp);
 
 	if (len == -ECONNREFUSED || len == -ENOTCONN || len == -EAGAIN)

net/sunrpc/svcsock.c

···
 #include <linux/sunrpc/msg_prot.h>
 #include <linux/sunrpc/svcsock.h>
 #include <linux/sunrpc/stats.h>
+#include <linux/sunrpc/xprt.h>
 
 #define RPCDBG_FACILITY	RPCDBG_SVCXPRT
 
···
 }
 
 /*
- * Generic sendto routine
+ * send routine intended to be shared by the fore- and back-channel
  */
-static int svc_sendto(struct svc_rqst *rqstp, struct xdr_buf *xdr)
+int svc_send_common(struct socket *sock, struct xdr_buf *xdr,
+		    struct page *headpage, unsigned long headoffset,
+		    struct page *tailpage, unsigned long tailoffset)
 {
-	struct svc_sock	*svsk =
-		container_of(rqstp->rq_xprt, struct svc_sock, sk_xprt);
-	struct socket	*sock = svsk->sk_sock;
-	int		slen;
-	union {
-		struct cmsghdr	hdr;
-		long		all[SVC_PKTINFO_SPACE / sizeof(long)];
-	} buffer;
-	struct cmsghdr *cmh = &buffer.hdr;
-	int		len = 0;
 	int		result;
 	int		size;
 	struct page	**ppage = xdr->pages;
 	size_t		base = xdr->page_base;
 	unsigned int	pglen = xdr->page_len;
 	unsigned int	flags = MSG_MORE;
-	RPC_IFDEBUG(char buf[RPC_MAX_ADDRBUFLEN]);
+	int		slen;
+	int		len = 0;
 
 	slen = xdr->len;
-
-	if (rqstp->rq_prot == IPPROTO_UDP) {
-		struct msghdr msg = {
-			.msg_name	= &rqstp->rq_addr,
-			.msg_namelen	= rqstp->rq_addrlen,
-			.msg_control	= cmh,
-			.msg_controllen	= sizeof(buffer),
-			.msg_flags	= MSG_MORE,
-		};
-
-		svc_set_cmsg_data(rqstp, cmh);
-
-		if (sock_sendmsg(sock, &msg, 0) < 0)
-			goto out;
-	}
 
 	/* send head */
 	if (slen == xdr->head[0].iov_len)
 		flags = 0;
-	len = kernel_sendpage(sock, rqstp->rq_respages[0], 0,
+	len = kernel_sendpage(sock, headpage, headoffset,
 				  xdr->head[0].iov_len, flags);
 	if (len != xdr->head[0].iov_len)
 		goto out;
···
 		base = 0;
 		ppage++;
 	}
+
 	/* send tail */
 	if (xdr->tail[0].iov_len) {
-		result = kernel_sendpage(sock, rqstp->rq_respages[0],
-					     ((unsigned long)xdr->tail[0].iov_base)
-						& (PAGE_SIZE-1),
-					     xdr->tail[0].iov_len, 0);
-
+		result = kernel_sendpage(sock, tailpage, tailoffset,
+				   xdr->tail[0].iov_len, 0);
 		if (result > 0)
 			len += result;
 	}
+
+out:
+	return len;
+}
+
+
+/*
+ * Generic sendto routine
+ */
+static int svc_sendto(struct svc_rqst *rqstp, struct xdr_buf *xdr)
+{
+	struct svc_sock	*svsk =
+		container_of(rqstp->rq_xprt, struct svc_sock, sk_xprt);
+	struct socket	*sock = svsk->sk_sock;
+	union {
+		struct cmsghdr	hdr;
+		long		all[SVC_PKTINFO_SPACE / sizeof(long)];
+	} buffer;
+	struct cmsghdr *cmh = &buffer.hdr;
+	int		len = 0;
+	unsigned long tailoff;
+	unsigned long headoff;
+	RPC_IFDEBUG(char buf[RPC_MAX_ADDRBUFLEN]);
+
+	if (rqstp->rq_prot == IPPROTO_UDP) {
+		struct msghdr msg = {
+			.msg_name	= &rqstp->rq_addr,
+			.msg_namelen	= rqstp->rq_addrlen,
+			.msg_control	= cmh,
+			.msg_controllen	= sizeof(buffer),
+			.msg_flags	= MSG_MORE,
+		};
+
+		svc_set_cmsg_data(rqstp, cmh);
+
+		if (sock_sendmsg(sock, &msg, 0) < 0)
+			goto out;
+	}
+
+	tailoff = ((unsigned long)xdr->tail[0].iov_base) & (PAGE_SIZE-1);
+	headoff = 0;
+	len = svc_send_common(sock, xdr, rqstp->rq_respages[0], headoff,
+			       rqstp->rq_respages[0], tailoff);
+
 out:
 	dprintk("svc: socket %p sendto([%p %Zu... ], %d) = %d (addr %s)\n",
 		svsk, xdr->head[0].iov_base, xdr->head[0].iov_len,
···
 	return -EAGAIN;
 }
 
+static int svc_process_calldir(struct svc_sock *svsk, struct svc_rqst *rqstp,
+			       struct rpc_rqst **reqpp, struct kvec *vec)
+{
+	struct rpc_rqst *req = NULL;
+	u32 *p;
+	u32 xid;
+	u32 calldir;
+	int len;
+
+	len = svc_recvfrom(rqstp, vec, 1, 8);
+	if (len < 0)
+		goto error;
+
+	p = (u32 *)rqstp->rq_arg.head[0].iov_base;
+	xid = *p++;
+	calldir = *p;
+
+	if (calldir == 0) {
+		/* REQUEST is the most common case */
+		vec[0] = rqstp->rq_arg.head[0];
+	} else {
+		/* REPLY */
+		if (svsk->sk_bc_xprt)
+			req = xprt_lookup_rqst(svsk->sk_bc_xprt, xid);
+
+		if (!req) {
+			printk(KERN_NOTICE
+				"%s: Got unrecognized reply: "
+				"calldir 0x%x sk_bc_xprt %p xid %08x\n",
+				__func__, ntohl(calldir),
+				svsk->sk_bc_xprt, xid);
+			vec[0] = rqstp->rq_arg.head[0];
+			goto out;
+		}
+
+		memcpy(&req->rq_private_buf, &req->rq_rcv_buf,
+		       sizeof(struct xdr_buf));
+		/* copy the xid and call direction */
+		memcpy(req->rq_private_buf.head[0].iov_base,
+		       rqstp->rq_arg.head[0].iov_base, 8);
+		vec[0] = req->rq_private_buf.head[0];
+	}
+ out:
+	vec[0].iov_base += 8;
+	vec[0].iov_len -= 8;
+	len = svsk->sk_reclen - 8;
+ error:
+	*reqpp = req;
+	return len;
+}
+
 /*
  * Receive data from a TCP socket.
  */
···
 	int		len;
 	struct kvec *vec;
 	int pnum, vlen;
+	struct rpc_rqst *req = NULL;
 
 	dprintk("svc: tcp_recv %p data %d conn %d close %d\n",
 		svsk, test_bit(XPT_DATA, &svsk->sk_xprt.xpt_flags),
···
 	vec = rqstp->rq_vec;
 	vec[0] = rqstp->rq_arg.head[0];
 	vlen = PAGE_SIZE;
+
+	/*
+	 * We have enough data for the whole tcp record. Let's try and read the
+	 * first 8 bytes to get the xid and the call direction. We can use this
+	 * to figure out if this is a call or a reply to a callback. If
+	 * sk_reclen is < 8 (xid and calldir), then this is a malformed packet.
+	 * In that case, don't bother with the calldir and just read the data.
+	 * It will be rejected in svc_process.
+	 */
+	if (len >= 8) {
+		len = svc_process_calldir(svsk, rqstp, &req, vec);
+		if (len < 0)
+			goto err_again;
+		vlen -= 8;
+	}
+
 	pnum = 1;
 	while (vlen < len) {
-		vec[pnum].iov_base = page_address(rqstp->rq_pages[pnum]);
+		vec[pnum].iov_base = (req) ?
+			page_address(req->rq_private_buf.pages[pnum - 1]) :
+			page_address(rqstp->rq_pages[pnum]);
 		vec[pnum].iov_len = PAGE_SIZE;
 		pnum++;
 		vlen += PAGE_SIZE;
···
 	if (len < 0)
 		goto err_again;
 
+	/*
+	 * Account for the 8 bytes we read earlier
+	 */
+	len += 8;
+
+	if (req) {
+		xprt_complete_rqst(req->rq_task, len);
+		len = 0;
+		goto out;
+	}
 	dprintk("svc: TCP complete record (%d bytes)\n", len);
 	rqstp->rq_arg.len = len;
 	rqstp->rq_arg.page_base = 0;
···
 	rqstp->rq_xprt_ctxt   = NULL;
 	rqstp->rq_prot	      = IPPROTO_TCP;
 
+out:
 	/* Reset TCP read info */
 	svsk->sk_reclen = 0;
 	svsk->sk_tcplen = 0;

net/sunrpc/xprt.c

···
 	spin_unlock_bh(&xprt->transport_lock);
 }
 
+static inline int xprt_has_timer(struct rpc_xprt *xprt)
+{
+	return xprt->idle_timeout != 0;
+}
+
 /**
  * xprt_prepare_transmit - reserve the transport before sending a request
  * @task: RPC task about to send a request
···
 	if (!list_empty(&req->rq_list))
 		list_del(&req->rq_list);
 	xprt->last_used = jiffies;
-	if (list_empty(&xprt->recv))
+	if (list_empty(&xprt->recv) && xprt_has_timer(xprt))
 		mod_timer(&xprt->timer,
 				xprt->last_used + xprt->idle_timeout);
 	spin_unlock_bh(&xprt->transport_lock);
···
 #endif /* CONFIG_NFS_V4_1 */
 
 	INIT_WORK(&xprt->task_cleanup, xprt_autoclose);
-	setup_timer(&xprt->timer, xprt_init_autodisconnect,
-			(unsigned long)xprt);
+	if (xprt_has_timer(xprt))
+		setup_timer(&xprt->timer, xprt_init_autodisconnect,
+			    (unsigned long)xprt);
+	else
+		init_timer(&xprt->timer);
 	xprt->last_used = jiffies;
 	xprt->cwnd = RPC_INITCWND;
 	xprt->bind_index = 0;
···
 
 	dprintk("RPC:       created transport %p with %u slots\n", xprt,
 			xprt->max_reqs);
-
 	return xprt;
 }
 

net/sunrpc/xprtsock.c

···
 #include <linux/tcp.h>
 #include <linux/sunrpc/clnt.h>
 #include <linux/sunrpc/sched.h>
+#include <linux/sunrpc/svcsock.h>
 #include <linux/sunrpc/xprtsock.h>
 #include <linux/file.h>
 #ifdef CONFIG_NFS_V4_1
···
 #include <net/udp.h>
 #include <net/tcp.h>
 
+#include "sunrpc.h"
 /*
  * xprtsock tunables
  */
···
 			xprt->stat.bklog_u);
 }
 
+/*
+ * Allocate a bunch of pages for a scratch buffer for the rpc code. The reason
+ * we allocate pages instead doing a kmalloc like rpc_malloc is because we want
+ * to use the server side send routines.
+ */
+void *bc_malloc(struct rpc_task *task, size_t size)
+{
+	struct page *page;
+	struct rpc_buffer *buf;
+
+	BUG_ON(size > PAGE_SIZE - sizeof(struct rpc_buffer));
+	page = alloc_page(GFP_KERNEL);
+
+	if (!page)
+		return NULL;
+
+	buf = page_address(page);
+	buf->len = PAGE_SIZE;
+
+	return buf->data;
+}
+
+/*
+ * Free the space allocated in the bc_alloc routine
+ */
+void bc_free(void *buffer)
+{
+	struct rpc_buffer *buf;
+
+	if (!buffer)
+		return;
+
+	buf = container_of(buffer, struct rpc_buffer, data);
+	free_page((unsigned long)buf);
+}
+
+/*
+ * Use the svc_sock to send the callback. Must be called with svsk->sk_mutex
+ * held. Borrows heavily from svc_tcp_sendto and xs_tcp_send_request.
+ */
+static int bc_sendto(struct rpc_rqst *req)
+{
+	int len;
+	struct xdr_buf *xbufp = &req->rq_snd_buf;
+	struct rpc_xprt *xprt = req->rq_xprt;
+	struct sock_xprt *transport =
+				container_of(xprt, struct sock_xprt, xprt);
+	struct socket *sock = transport->sock;
+	unsigned long headoff;
+	unsigned long tailoff;
+
+	/*
+	 * Set up the rpc header and record marker stuff
+	 */
+	xs_encode_tcp_record_marker(xbufp);
+
+	tailoff = (unsigned long)xbufp->tail[0].iov_base & ~PAGE_MASK;
+	headoff = (unsigned long)xbufp->head[0].iov_base & ~PAGE_MASK;
+	len = svc_send_common(sock, xbufp,
+			      virt_to_page(xbufp->head[0].iov_base), headoff,
+			      xbufp->tail[0].iov_base, tailoff);
+
+	if (len != xbufp->len) {
+		printk(KERN_NOTICE "Error sending entire callback!\n");
+		len = -EAGAIN;
+	}
+
+	return len;
+}
+
+/*
+ * The send routine. Borrows from svc_send
+ */
+static int bc_send_request(struct rpc_task *task)
+{
+	struct rpc_rqst *req = task->tk_rqstp;
+	struct svc_xprt	*xprt;
+	struct svc_sock         *svsk;
+	u32                     len;
+
+	dprintk("sending request with xid: %08x\n", ntohl(req->rq_xid));
+	/*
+	 * Get the server socket associated with this callback xprt
+	 */
+	xprt = req->rq_xprt->bc_xprt;
+	svsk = container_of(xprt, struct svc_sock, sk_xprt);
+
+	/*
+	 * Grab the mutex to serialize data as the connection is shared
+	 * with the fore channel
+	 */
+	if (!mutex_trylock(&xprt->xpt_mutex)) {
+		rpc_sleep_on(&xprt->xpt_bc_pending, task, NULL);
+		if (!mutex_trylock(&xprt->xpt_mutex))
+			return -EAGAIN;
+		rpc_wake_up_queued_task(&xprt->xpt_bc_pending, task);
+	}
+	if (test_bit(XPT_DEAD, &xprt->xpt_flags))
+		len = -ENOTCONN;
+	else
+		len = bc_sendto(req);
+	mutex_unlock(&xprt->xpt_mutex);
+
+	if (len > 0)
+		len = 0;
+
+	return len;
+}
+
+/*
+ * The close routine. Since this is client initiated, we do nothing
+ */
+
+static void bc_close(struct rpc_xprt *xprt)
+{
+	return;
+}
+
+/*
+ * The xprt destroy routine. Again, because this connection is client
+ * initiated, we do nothing
+ */
+
+static void bc_destroy(struct rpc_xprt *xprt)
+{
+	return;
+}
+
 static struct rpc_xprt_ops xs_udp_ops = {
 	.set_buffer_size	= xs_udp_set_buffer_size,
 	.reserve_xprt		= xprt_reserve_xprt_cong,
···
 #endif /* CONFIG_NFS_V4_1 */
 	.close			= xs_tcp_close,
 	.destroy		= xs_destroy,
+	.print_stats		= xs_tcp_print_stats,
+};
+
+/*
+ * The rpc_xprt_ops for the server backchannel
+ */
+
+static struct rpc_xprt_ops bc_tcp_ops = {
+	.reserve_xprt		= xprt_reserve_xprt,
+	.release_xprt		= xprt_release_xprt,
+	.buf_alloc		= bc_malloc,
+	.buf_free		= bc_free,
+	.send_request		= bc_send_request,
+	.set_retrans_timeout	= xprt_set_retrans_timeout_def,
+	.close			= bc_close,
+	.destroy		= bc_destroy,
 	.print_stats		= xs_tcp_print_stats,
 };