[PATCH] RPC: introduce client-side transport switch

 Move the bulk of client-side socket-specific code into a separate source
 file, net/sunrpc/xprtsock.c.

 Test-plan:
 Millions of fsx operations.  Performance characterization such as "sio" or
 "iozone".  Destructive testing (unplugging the network temporarily, server
 reboots).  Connectathon with v2, v3, and v4.

 Version: Thu, 11 Aug 2005 16:03:38 -0400
Signed-off-by: Chuck Lever <cel@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>

include/linux/sunrpc/xdr.h

@@ -165,11 +165,6 @@ extern int csum_partial_copy_to_xdr(struct xdr_buf *, struct sk_buff *);
 extern ssize_t xdr_partial_copy_from_skb(struct xdr_buf *, unsigned int,
 		skb_reader_t *, skb_read_actor_t);
 
-struct socket;
-struct sockaddr;
-extern int xdr_sendpages(struct socket *, struct sockaddr *, int,
-		struct xdr_buf *, unsigned int, int);
-
 extern int xdr_encode_word(struct xdr_buf *, int, u32);
 extern int xdr_decode_word(struct xdr_buf *, int, u32 *);
 

include/linux/sunrpc/xprt.h

@@ -59,7 +59,13 @@ extern unsigned int xprt_tcp_slot_table_entries;
  */
 #define RPC_REESTABLISH_TIMEOUT	(15*HZ)
 
-/* RPC call and reply header size as number of 32bit words (verifier
+/*
+ * RPC transport idle timeout.
+ */
+#define RPC_IDLE_DISCONNECT_TIMEOUT	(5*60*HZ)
+
+/*
+ * RPC call and reply header size as number of 32bit words (verifier
  * size computed separately)
  */
 #define RPC_CALLHDRSIZE		6
@@ -121,12 +127,19 @@ struct rpc_rqst {
 #define rq_svec			rq_snd_buf.head
 #define rq_slen			rq_snd_buf.len
 
-#define XPRT_LAST_FRAG		(1 << 0)
-#define XPRT_COPY_RECM		(1 << 1)
-#define XPRT_COPY_XID		(1 << 2)
-#define XPRT_COPY_DATA		(1 << 3)
+struct rpc_task;
+struct rpc_xprt;
+
+struct rpc_xprt_ops {
+	void		(*set_buffer_size)(struct rpc_xprt *xprt);
+	void		(*connect)(struct rpc_task *task);
+	int		(*send_request)(struct rpc_task *task);
+	void		(*close)(struct rpc_xprt *xprt);
+	void		(*destroy)(struct rpc_xprt *xprt);
+};
 
 struct rpc_xprt {
+	struct rpc_xprt_ops *	ops;		/* transport methods */
 	struct socket *		sock;		/* BSD socket layer */
 	struct sock *		inet;		/* INET layer */
 
@@ -199,14 +212,22 @@ struct rpc_xprt {
 	wait_queue_head_t	cong_wait;
 };
 
+#define XPRT_LAST_FRAG		(1 << 0)
+#define XPRT_COPY_RECM		(1 << 1)
+#define XPRT_COPY_XID		(1 << 2)
+#define XPRT_COPY_DATA		(1 << 3)
+
 #ifdef __KERNEL__
 
 struct rpc_xprt *	xprt_create_proto(int proto, struct sockaddr_in *addr,
 					struct rpc_timeout *toparms);
+void			xprt_disconnect(struct rpc_xprt *);
 int			xprt_destroy(struct rpc_xprt *);
 void			xprt_set_timeout(struct rpc_timeout *, unsigned int,
 					unsigned long);
-
+struct rpc_rqst *	xprt_lookup_rqst(struct rpc_xprt *, u32);
+void			xprt_complete_rqst(struct rpc_xprt *,
+					struct rpc_rqst *, int);
 void			xprt_reserve(struct rpc_task *);
 int			xprt_prepare_transmit(struct rpc_task *);
 void			xprt_transmit(struct rpc_task *);
@@ -214,7 +235,10 @@ void			xprt_receive(struct rpc_task *);
 int			xprt_adjust_timeout(struct rpc_rqst *req);
 void			xprt_release(struct rpc_task *);
 void			xprt_connect(struct rpc_task *);
-void			xprt_sock_setbufsize(struct rpc_xprt *);
+int			xs_setup_udp(struct rpc_xprt *,
+					struct rpc_timeout *);
+int			xs_setup_tcp(struct rpc_xprt *,
+					struct rpc_timeout *);
 
 #define XPRT_LOCKED	0
 #define XPRT_CONNECT	1

net/sunrpc/Makefile

@@ -6,7 +6,7 @@
 obj-$(CONFIG_SUNRPC) += sunrpc.o
 obj-$(CONFIG_SUNRPC_GSS) += auth_gss/
 
-sunrpc-y := clnt.o xprt.o socklib.o sched.o \
+sunrpc-y := clnt.o xprt.o socklib.o xprtsock.o sched.o \
 	    auth.o auth_null.o auth_unix.o \
 	    svc.o svcsock.o svcauth.o svcauth_unix.o \
 	    pmap_clnt.o timer.o xdr.o \

net/sunrpc/clnt.c

@@ -525,8 +525,7 @@ rpc_setbufsize(struct rpc_clnt *clnt, unsigned int sndsize, unsigned int rcvsize
 	xprt->rcvsize = 0;
 	if (rcvsize)
 		xprt->rcvsize = rcvsize + RPC_SLACK_SPACE;
-	if (xprt_connected(xprt))
-		xprt_sock_setbufsize(xprt);
+	xprt->ops->set_buffer_size(xprt);
 }
 
 /*

net/sunrpc/sysctl.c

@@ -119,6 +119,9 @@ proc_dodebug(ctl_table *table, int write, struct file *file,
 	return 0;
 }
 
+unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE;
+unsigned int xprt_tcp_slot_table_entries = RPC_DEF_SLOT_TABLE;
+
 static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE;
 static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE;
 

net/sunrpc/xdr.c

@@ -6,15 +6,12 @@
  * Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de>
  */
 
+#include <linux/module.h>
 #include <linux/types.h>
-#include <linux/socket.h>
 #include <linux/string.h>
 #include <linux/kernel.h>
 #include <linux/pagemap.h>
 #include <linux/errno.h>
-#include <linux/in.h>
-#include <linux/net.h>
-#include <net/sock.h>
 #include <linux/sunrpc/xdr.h>
 #include <linux/sunrpc/msg_prot.h>
 
@@ -177,103 +174,6 @@ xdr_inline_pages(struct xdr_buf *xdr, unsigned int offset,
 }
 
 
-int
-xdr_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen,
-		struct xdr_buf *xdr, unsigned int base, int msgflags)
-{
-	struct page **ppage = xdr->pages;
-	unsigned int len, pglen = xdr->page_len;
-	int err, ret = 0;
-	ssize_t (*sendpage)(struct socket *, struct page *, int, size_t, int);
-
-	len = xdr->head[0].iov_len;
-	if (base < len || (addr != NULL && base == 0)) {
-		struct kvec iov = {
-			.iov_base = xdr->head[0].iov_base + base,
-			.iov_len  = len - base,
-		};
-		struct msghdr msg = {
-			.msg_name    = addr,
-			.msg_namelen = addrlen,
-			.msg_flags   = msgflags,
-		};
-		if (xdr->len > len)
-			msg.msg_flags |= MSG_MORE;
-
-		if (iov.iov_len != 0)
-			err = kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len);
-		else
-			err = kernel_sendmsg(sock, &msg, NULL, 0, 0);
-		if (ret == 0)
-			ret = err;
-		else if (err > 0)
-			ret += err;
-		if (err != iov.iov_len)
-			goto out;
-		base = 0;
-	} else
-		base -= len;
-
-	if (pglen == 0)
-		goto copy_tail;
-	if (base >= pglen) {
-		base -= pglen;
-		goto copy_tail;
-	}
-	if (base || xdr->page_base) {
-		pglen -= base;
-		base  += xdr->page_base;
-		ppage += base >> PAGE_CACHE_SHIFT;
-		base &= ~PAGE_CACHE_MASK;
-	}
-
-	sendpage = sock->ops->sendpage ? : sock_no_sendpage;
-	do {
-		int flags = msgflags;
-
-		len = PAGE_CACHE_SIZE;
-		if (base)
-			len -= base;
-		if (pglen < len)
-			len = pglen;
-
-		if (pglen != len || xdr->tail[0].iov_len != 0)
-			flags |= MSG_MORE;
-
-		/* Hmm... We might be dealing with highmem pages */
-		if (PageHighMem(*ppage))
-			sendpage = sock_no_sendpage;
-		err = sendpage(sock, *ppage, base, len, flags);
-		if (ret == 0)
-			ret = err;
-		else if (err > 0)
-			ret += err;
-		if (err != len)
-			goto out;
-		base = 0;
-		ppage++;
-	} while ((pglen -= len) != 0);
-copy_tail:
-	len = xdr->tail[0].iov_len;
-	if (base < len) {
-		struct kvec iov = {
-			.iov_base = xdr->tail[0].iov_base + base,
-			.iov_len  = len - base,
-		};
-		struct msghdr msg = {
-			.msg_flags   = msgflags,
-		};
-		err = kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len);
-		if (ret == 0)
-			ret = err;
-		else if (err > 0)
-			ret += err;
-	}
-out:
-	return ret;
-}
-
-
 /*
  * Helper routines for doing 'memmove' like operations on a struct xdr_buf
  *