xprtsock.c

linux 内核源代码

/*
 * linux/net/sunrpc/xprtsock.c
 *
 * Client-side transport implementation for sockets.
 *
 * TCP callback races fixes (C) 1998 Red Hat Software <[e-mail]>
 * TCP send fixes (C) 1998 Red Hat Software <[e-mail]>
 * TCP NFS related read + write fixes
 *  (C) 1999 Dave Airlie, University of Limerick, Ireland <[e-mail]>
 *
 * Rewrite of larges part of the code in order to stabilize TCP stuff.
 * Fix behaviour when socket buffer is full.
 *  (C) 1999 Trond Myklebust <[e-mail]>
 *
 * IP socket transport implementation, (C) 2005 Chuck Lever <[e-mail]>
 *
 * IPv6 support contributed by Gilles Quillard, Bull Open Source, 2005.
 *   <[e-mail]>
 */

#include <linux/types.h>
#include <linux/slab.h>
#include <linux/module.h>
#include <linux/capability.h>
#include <linux/pagemap.h>
#include <linux/errno.h>
#include <linux/socket.h>
#include <linux/in.h>
#include <linux/net.h>
#include <linux/mm.h>
#include <linux/udp.h>
#include <linux/tcp.h>
#include <linux/sunrpc/clnt.h>
#include <linux/sunrpc/sched.h>
#include <linux/sunrpc/xprtsock.h>
#include <linux/file.h>

#include <net/sock.h>
#include <net/checksum.h>
#include <net/udp.h>
#include <net/tcp.h>

/*
 * xprtsock tunables
 */
unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE;
unsigned int xprt_tcp_slot_table_entries = RPC_DEF_SLOT_TABLE;

unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT;
unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT;

/*
 * We can register our own files under /proc/sys/sunrpc by
 * calling register_sysctl_table() again.  The files in that
 * directory become the union of all files registered there.
 *
 * We simply need to make sure that we don't collide with
 * someone else's file names!
 */

#ifdef RPC_DEBUG

static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE;
static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE;
static unsigned int xprt_min_resvport_limit = RPC_MIN_RESVPORT;
static unsigned int xprt_max_resvport_limit = RPC_MAX_RESVPORT;

static struct ctl_table_header *sunrpc_table_header;

/*
 * FIXME: changing the UDP slot table size should also resize the UDP
 *        socket buffers for existing UDP transports
 */
static ctl_table xs_tunables_table[] = {
	{
		.ctl_name	= CTL_SLOTTABLE_UDP,
		.procname	= "udp_slot_table_entries",
		.data		= &xprt_udp_slot_table_entries,
		.maxlen		= sizeof(unsigned int),
		.mode		= 0644,
		.proc_handler	= &proc_dointvec_minmax,
		.strategy	= &sysctl_intvec,
		.extra1		= &min_slot_table_size,
		.extra2		= &max_slot_table_size
	},
	{
		.ctl_name	= CTL_SLOTTABLE_TCP,
		.procname	= "tcp_slot_table_entries",
		.data		= &xprt_tcp_slot_table_entries,
		.maxlen		= sizeof(unsigned int),
		.mode		= 0644,
		.proc_handler	= &proc_dointvec_minmax,
		.strategy	= &sysctl_intvec,
		.extra1		= &min_slot_table_size,
		.extra2		= &max_slot_table_size
	},
	{
		.ctl_name	= CTL_MIN_RESVPORT,
		.procname	= "min_resvport",
		.data		= &xprt_min_resvport,
		.maxlen		= sizeof(unsigned int),
		.mode		= 0644,
		.proc_handler	= &proc_dointvec_minmax,
		.strategy	= &sysctl_intvec,
		.extra1		= &xprt_min_resvport_limit,
		.extra2		= &xprt_max_resvport_limit
	},
	{
		.ctl_name	= CTL_MAX_RESVPORT,
		.procname	= "max_resvport",
		.data		= &xprt_max_resvport,
		.maxlen		= sizeof(unsigned int),
		.mode		= 0644,
		.proc_handler	= &proc_dointvec_minmax,
		.strategy	= &sysctl_intvec,
		.extra1		= &xprt_min_resvport_limit,
		.extra2		= &xprt_max_resvport_limit
	},
	{
		.ctl_name = 0,
	},
};

static ctl_table sunrpc_table[] = {
	{
		.ctl_name	= CTL_SUNRPC,
		.procname	= "sunrpc",
		.mode		= 0555,
		.child		= xs_tunables_table
	},
	{
		.ctl_name = 0,
	},
};

#endif

/*
 * How many times to try sending a request on a socket before waiting
 * for the socket buffer to clear.
 */
#define XS_SENDMSG_RETRY	(10U)

/*
 * Time out for an RPC UDP socket connect.  UDP socket connects are
 * synchronous, but we set a timeout anyway in case of resource
 * exhaustion on the local host.
 */
#define XS_UDP_CONN_TO		(5U * HZ)

/*
 * Wait duration for an RPC TCP connection to be established.  Solaris
 * NFS over TCP uses 60 seconds, for example, which is in line with how
 * long a server takes to reboot.
 */
#define XS_TCP_CONN_TO		(60U * HZ)

/*
 * Wait duration for a reply from the RPC portmapper.
 */
#define XS_BIND_TO		(60U * HZ)

/*
 * Delay if a UDP socket connect error occurs.  This is most likely some
 * kind of resource problem on the local host.
 */
#define XS_UDP_REEST_TO		(2U * HZ)

/*
 * The reestablish timeout allows clients to delay for a bit before attempting
 * to reconnect to a server that just dropped our connection.
 *
 * We implement an exponential backoff when trying to reestablish a TCP
 * transport connection with the server.  Some servers like to drop a TCP
 * connection when they are overworked, so we start with a short timeout and
 * increase over time if the server is down or not responding.
 */
#define XS_TCP_INIT_REEST_TO	(3U * HZ)
#define XS_TCP_MAX_REEST_TO	(5U * 60 * HZ)

/*
 * TCP idle timeout; client drops the transport socket if it is idle
 * for this long.  Note that we also timeout UDP sockets to prevent
 * holding port numbers when there is no RPC traffic.
 */
#define XS_IDLE_DISC_TO		(5U * 60 * HZ)

#ifdef RPC_DEBUG
# undef  RPC_DEBUG_DATA
# define RPCDBG_FACILITY	RPCDBG_TRANS
#endif

#ifdef RPC_DEBUG_DATA
static void xs_pktdump(char *msg, u32 *packet, unsigned int count)
{
	u8 *buf = (u8 *) packet;
	int j;

	dprintk("RPC:       %s\n", msg);
	for (j = 0; j < count && j < 128; j += 4) {
		if (!(j & 31)) {
			if (j)
				dprintk("\n");
			dprintk("0x%04x ", j);
		}
		dprintk("%02x%02x%02x%02x ",
			buf[j], buf[j+1], buf[j+2], buf[j+3]);
	}
	dprintk("\n");
}
#else
static inline void xs_pktdump(char *msg, u32 *packet, unsigned int count)
{
	/* NOP */
}
#endif

struct sock_xprt {
	struct rpc_xprt		xprt;

	/*
	 * Network layer
	 */
	struct socket *		sock;
	struct sock *		inet;

	/*
	 * State of TCP reply receive
	 */
	__be32			tcp_fraghdr,
				tcp_xid;

	u32			tcp_offset,
				tcp_reclen;

	unsigned long		tcp_copied,
				tcp_flags;

	/*
	 * Connection of transports
	 */
	struct delayed_work	connect_worker;
	struct sockaddr_storage	addr;
	unsigned short		port;

	/*
	 * UDP socket buffer size parameters
	 */
	size_t			rcvsize,
				sndsize;

	/*
	 * Saved socket callback addresses
	 */
	void			(*old_data_ready)(struct sock *, int);
	void			(*old_state_change)(struct sock *);
	void			(*old_write_space)(struct sock *);
};

/*
 * TCP receive state flags
 */
#define TCP_RCV_LAST_FRAG	(1UL << 0)
#define TCP_RCV_COPY_FRAGHDR	(1UL << 1)
#define TCP_RCV_COPY_XID	(1UL << 2)
#define TCP_RCV_COPY_DATA	(1UL << 3)

static inline struct sockaddr *xs_addr(struct rpc_xprt *xprt)
{
	return (struct sockaddr *) &xprt->addr;
}

static inline struct sockaddr_in *xs_addr_in(struct rpc_xprt *xprt)
{
	return (struct sockaddr_in *) &xprt->addr;
}

static inline struct sockaddr_in6 *xs_addr_in6(struct rpc_xprt *xprt)
{
	return (struct sockaddr_in6 *) &xprt->addr;
}

static void xs_format_ipv4_peer_addresses(struct rpc_xprt *xprt)
{
	struct sockaddr_in *addr = xs_addr_in(xprt);
	char *buf;

	buf = kzalloc(20, GFP_KERNEL);
	if (buf) {
		snprintf(buf, 20, NIPQUAD_FMT,
				NIPQUAD(addr->sin_addr.s_addr));
	}
	xprt->address_strings[RPC_DISPLAY_ADDR] = buf;

	buf = kzalloc(8, GFP_KERNEL);
	if (buf) {
		snprintf(buf, 8, "%u",
				ntohs(addr->sin_port));
	}
	xprt->address_strings[RPC_DISPLAY_PORT] = buf;

	buf = kzalloc(8, GFP_KERNEL);
	if (buf) {
		if (xprt->prot == IPPROTO_UDP)
			snprintf(buf, 8, "udp");
		else
			snprintf(buf, 8, "tcp");
	}
	xprt->address_strings[RPC_DISPLAY_PROTO] = buf;

	buf = kzalloc(48, GFP_KERNEL);
	if (buf) {
		snprintf(buf, 48, "addr="NIPQUAD_FMT" port=%u proto=%s",
			NIPQUAD(addr->sin_addr.s_addr),
			ntohs(addr->sin_port),
			xprt->prot == IPPROTO_UDP ? "udp" : "tcp");
	}
	xprt->address_strings[RPC_DISPLAY_ALL] = buf;

	buf = kzalloc(10, GFP_KERNEL);
	if (buf) {
		snprintf(buf, 10, "%02x%02x%02x%02x",
				NIPQUAD(addr->sin_addr.s_addr));
	}
	xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = buf;

	buf = kzalloc(8, GFP_KERNEL);
	if (buf) {
		snprintf(buf, 8, "%4hx",
				ntohs(addr->sin_port));
	}
	xprt->address_strings[RPC_DISPLAY_HEX_PORT] = buf;

	buf = kzalloc(30, GFP_KERNEL);
	if (buf) {
		snprintf(buf, 30, NIPQUAD_FMT".%u.%u",
				NIPQUAD(addr->sin_addr.s_addr),
				ntohs(addr->sin_port) >> 8,
				ntohs(addr->sin_port) & 0xff);
	}
	xprt->address_strings[RPC_DISPLAY_UNIVERSAL_ADDR] = buf;

	xprt->address_strings[RPC_DISPLAY_NETID] =
		kstrdup(xprt->prot == IPPROTO_UDP ?
			RPCBIND_NETID_UDP : RPCBIND_NETID_TCP, GFP_KERNEL);
}

static void xs_format_ipv6_peer_addresses(struct rpc_xprt *xprt)
{
	struct sockaddr_in6 *addr = xs_addr_in6(xprt);
	char *buf;

	buf = kzalloc(40, GFP_KERNEL);
	if (buf) {
		snprintf(buf, 40, NIP6_FMT,
				NIP6(addr->sin6_addr));
	}
	xprt->address_strings[RPC_DISPLAY_ADDR] = buf;

	buf = kzalloc(8, GFP_KERNEL);
	if (buf) {
		snprintf(buf, 8, "%u",
				ntohs(addr->sin6_port));
	}
	xprt->address_strings[RPC_DISPLAY_PORT] = buf;

	buf = kzalloc(8, GFP_KERNEL);
	if (buf) {
		if (xprt->prot == IPPROTO_UDP)
			snprintf(buf, 8, "udp");
		else
			snprintf(buf, 8, "tcp");
	}
	xprt->address_strings[RPC_DISPLAY_PROTO] = buf;

	buf = kzalloc(64, GFP_KERNEL);
	if (buf) {
		snprintf(buf, 64, "addr="NIP6_FMT" port=%u proto=%s",
				NIP6(addr->sin6_addr),
				ntohs(addr->sin6_port),
				xprt->prot == IPPROTO_UDP ? "udp" : "tcp");
	}
	xprt->address_strings[RPC_DISPLAY_ALL] = buf;

	buf = kzalloc(36, GFP_KERNEL);
	if (buf) {
		snprintf(buf, 36, NIP6_SEQFMT,
				NIP6(addr->sin6_addr));
	}
	xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = buf;

	buf = kzalloc(8, GFP_KERNEL);
	if (buf) {
		snprintf(buf, 8, "%4hx",
				ntohs(addr->sin6_port));
	}
	xprt->address_strings[RPC_DISPLAY_HEX_PORT] = buf;

	buf = kzalloc(50, GFP_KERNEL);
	if (buf) {
		snprintf(buf, 50, NIP6_FMT".%u.%u",
				NIP6(addr->sin6_addr),
				ntohs(addr->sin6_port) >> 8,
				ntohs(addr->sin6_port) & 0xff);
	}
	xprt->address_strings[RPC_DISPLAY_UNIVERSAL_ADDR] = buf;

	xprt->address_strings[RPC_DISPLAY_NETID] =
		kstrdup(xprt->prot == IPPROTO_UDP ?
			RPCBIND_NETID_UDP6 : RPCBIND_NETID_TCP6, GFP_KERNEL);
}

static void xs_free_peer_addresses(struct rpc_xprt *xprt)
{
	int i;

	for (i = 0; i < RPC_DISPLAY_MAX; i++)
		kfree(xprt->address_strings[i]);
}

#define XS_SENDMSG_FLAGS	(MSG_DONTWAIT | MSG_NOSIGNAL)

static int xs_send_kvec(struct socket *sock, struct sockaddr *addr, int addrlen, struct kvec *vec, unsigned int base, int more)
{
	struct msghdr msg = {
		.msg_name	= addr,
		.msg_namelen	= addrlen,
		.msg_flags	= XS_SENDMSG_FLAGS | (more ? MSG_MORE : 0),
	};
	struct kvec iov = {
		.iov_base	= vec->iov_base + base,
		.iov_len	= vec->iov_len - base,
	};

	if (iov.iov_len != 0)
		return kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len);
	return kernel_sendmsg(sock, &msg, NULL, 0, 0);
}

static int xs_send_pagedata(struct socket *sock, struct xdr_buf *xdr, unsigned int base, int more)
{
	struct page **ppage;
	unsigned int remainder;
	int err, sent = 0;

	remainder = xdr->page_len - base;
	base += xdr->page_base;
	ppage = xdr->pages + (base >> PAGE_SHIFT);
	base &= ~PAGE_MASK;
	for(;;) {
		unsigned int len = min_t(unsigned int, PAGE_SIZE - base, remainder);
		int flags = XS_SENDMSG_FLAGS;

		remainder -= len;
		if (remainder != 0 || more)
			flags |= MSG_MORE;
		err = sock->ops->sendpage(sock, *ppage, base, len, flags);
		if (remainder == 0 || err != len)
			break;
		sent += err;
		ppage++;
		base = 0;
	}
	if (sent == 0)
		return err;
	if (err > 0)
		sent += err;
	return sent;
}

/**
 * xs_sendpages - write pages directly to a socket
 * @sock: socket to send on
 * @addr: UDP only -- address of destination
 * @addrlen: UDP only -- length of destination address
 * @xdr: buffer containing this request
 * @base: starting position in the buffer
 *
 */
static int xs_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen, struct xdr_buf *xdr, unsigned int base)
{
	unsigned int remainder = xdr->len - base;
	int err, sent = 0;

	if (unlikely(!sock))
		return -ENOTCONN;

	clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags);
	if (base != 0) {
		addr = NULL;
		addrlen = 0;
	}

	if (base < xdr->head[0].iov_len || addr != NULL) {
		unsigned int len = xdr->head[0].iov_len - base;
		remainder -= len;
		err = xs_send_kvec(sock, addr, addrlen, &xdr->head[0], base, remainder != 0);
		if (remainder == 0 || err != len)
			goto out;
		sent += err;
		base = 0;