/*
|
/*
|
* linux/net/sunrpc/xprt.c
|
* linux/net/sunrpc/xprt.c
|
*
|
*
|
* This is a generic RPC call interface supporting congestion avoidance,
|
* This is a generic RPC call interface supporting congestion avoidance,
|
* and asynchronous calls.
|
* and asynchronous calls.
|
*
|
*
|
* The interface works like this:
|
* The interface works like this:
|
*
|
*
|
* - When a process places a call, it allocates a request slot if
|
* - When a process places a call, it allocates a request slot if
|
* one is available. Otherwise, it sleeps on the backlog queue
|
* one is available. Otherwise, it sleeps on the backlog queue
|
* (xprt_reserve).
|
* (xprt_reserve).
|
* - Next, the caller puts together the RPC message, stuffs it into
|
* - Next, the caller puts together the RPC message, stuffs it into
|
* the request struct, and calls xprt_call().
|
* the request struct, and calls xprt_call().
|
* - xprt_call transmits the message and installs the caller on the
|
* - xprt_call transmits the message and installs the caller on the
|
* socket's wait list. At the same time, it installs a timer that
|
* socket's wait list. At the same time, it installs a timer that
|
* is run after the packet's timeout has expired.
|
* is run after the packet's timeout has expired.
|
* - When a packet arrives, the data_ready handler walks the list of
|
* - When a packet arrives, the data_ready handler walks the list of
|
* pending requests for that socket. If a matching XID is found, the
|
* pending requests for that socket. If a matching XID is found, the
|
* caller is woken up, and the timer removed.
|
* caller is woken up, and the timer removed.
|
* - When no reply arrives within the timeout interval, the timer is
|
* - When no reply arrives within the timeout interval, the timer is
|
* fired by the kernel and runs xprt_timer(). It either adjusts the
|
* fired by the kernel and runs xprt_timer(). It either adjusts the
|
* timeout values (minor timeout) or wakes up the caller with a status
|
* timeout values (minor timeout) or wakes up the caller with a status
|
* of -ETIMEDOUT.
|
* of -ETIMEDOUT.
|
* - When the caller receives a notification from RPC that a reply arrived,
|
* - When the caller receives a notification from RPC that a reply arrived,
|
* it should release the RPC slot, and process the reply.
|
* it should release the RPC slot, and process the reply.
|
* If the call timed out, it may choose to retry the operation by
|
* If the call timed out, it may choose to retry the operation by
|
* adjusting the initial timeout value, and simply calling rpc_call
|
* adjusting the initial timeout value, and simply calling rpc_call
|
* again.
|
* again.
|
*
|
*
|
* Support for async RPC is done through a set of RPC-specific scheduling
|
* Support for async RPC is done through a set of RPC-specific scheduling
|
* primitives that `transparently' work for processes as well as async
|
* primitives that `transparently' work for processes as well as async
|
* tasks that rely on callbacks.
|
* tasks that rely on callbacks.
|
*
|
*
|
* Copyright (C) 1995-1997, Olaf Kirch <okir@monad.swb.de>
|
* Copyright (C) 1995-1997, Olaf Kirch <okir@monad.swb.de>
|
*
|
*
|
* TCP callback races fixes (C) 1998 Red Hat Software <alan@redhat.com>
|
* TCP callback races fixes (C) 1998 Red Hat Software <alan@redhat.com>
|
* TCP send fixes (C) 1998 Red Hat Software <alan@redhat.com>
|
* TCP send fixes (C) 1998 Red Hat Software <alan@redhat.com>
|
* TCP NFS related read + write fixes
|
* TCP NFS related read + write fixes
|
* (C) 1999 Dave Airlie, University of Limerick, Ireland <airlied@linux.ie>
|
* (C) 1999 Dave Airlie, University of Limerick, Ireland <airlied@linux.ie>
|
*
|
*
|
* Rewrite of larges part of the code in order to stabilize TCP stuff.
|
* Rewrite of larges part of the code in order to stabilize TCP stuff.
|
* Fix behaviour when socket buffer is full.
|
* Fix behaviour when socket buffer is full.
|
* (C) 1999 Trond Myklebust <trond.myklebust@fys.uio.no>
|
* (C) 1999 Trond Myklebust <trond.myklebust@fys.uio.no>
|
*/
|
*/
|
|
|
#define __KERNEL_SYSCALLS__
|
#define __KERNEL_SYSCALLS__
|
|
|
#include <linux/version.h>
|
#include <linux/version.h>
|
#include <linux/types.h>
|
#include <linux/types.h>
|
#include <linux/slab.h>
|
#include <linux/slab.h>
|
#include <linux/capability.h>
|
#include <linux/capability.h>
|
#include <linux/sched.h>
|
#include <linux/sched.h>
|
#include <linux/errno.h>
|
#include <linux/errno.h>
|
#include <linux/socket.h>
|
#include <linux/socket.h>
|
#include <linux/in.h>
|
#include <linux/in.h>
|
#include <linux/net.h>
|
#include <linux/net.h>
|
#include <linux/mm.h>
|
#include <linux/mm.h>
|
#include <linux/udp.h>
|
#include <linux/udp.h>
|
#include <linux/unistd.h>
|
#include <linux/unistd.h>
|
#include <linux/sunrpc/clnt.h>
|
#include <linux/sunrpc/clnt.h>
|
#include <linux/file.h>
|
#include <linux/file.h>
|
|
|
#include <net/sock.h>
|
#include <net/sock.h>
|
#include <net/checksum.h>
|
#include <net/checksum.h>
|
#include <net/udp.h>
|
#include <net/udp.h>
|
#include <net/tcp.h>
|
#include <net/tcp.h>
|
|
|
#include <asm/uaccess.h>
|
#include <asm/uaccess.h>
|
|
|
/*
|
/*
|
* Local variables
|
* Local variables
|
*/
|
*/
|
|
|
#ifdef RPC_DEBUG
|
#ifdef RPC_DEBUG
|
# undef RPC_DEBUG_DATA
|
# undef RPC_DEBUG_DATA
|
# define RPCDBG_FACILITY RPCDBG_XPRT
|
# define RPCDBG_FACILITY RPCDBG_XPRT
|
#endif
|
#endif
|
|
|
#define XPRT_MAX_BACKOFF (8)
|
#define XPRT_MAX_BACKOFF (8)
|
|
|
/*
|
/*
|
* Local functions
|
* Local functions
|
*/
|
*/
|
static void xprt_request_init(struct rpc_task *, struct rpc_xprt *);
|
static void xprt_request_init(struct rpc_task *, struct rpc_xprt *);
|
static void do_xprt_transmit(struct rpc_task *);
|
static void do_xprt_transmit(struct rpc_task *);
|
static inline void do_xprt_reserve(struct rpc_task *);
|
static inline void do_xprt_reserve(struct rpc_task *);
|
static void xprt_disconnect(struct rpc_xprt *);
|
static void xprt_disconnect(struct rpc_xprt *);
|
static void xprt_connect_status(struct rpc_task *task);
|
static void xprt_connect_status(struct rpc_task *task);
|
static struct socket *xprt_create_socket(int, struct rpc_timeout *, int);
|
static struct socket *xprt_create_socket(int, struct rpc_timeout *, int);
|
static int xprt_bind_socket(struct rpc_xprt *, struct socket *);
|
static int xprt_bind_socket(struct rpc_xprt *, struct socket *);
|
static int __xprt_get_cong(struct rpc_xprt *, struct rpc_task *);
|
static int __xprt_get_cong(struct rpc_xprt *, struct rpc_task *);
|
|
|
#ifdef RPC_DEBUG_DATA
|
#ifdef RPC_DEBUG_DATA
|
/*
|
/*
|
* Print the buffer contents (first 128 bytes only--just enough for
|
* Print the buffer contents (first 128 bytes only--just enough for
|
* diropres return).
|
* diropres return).
|
*/
|
*/
|
static void
|
static void
|
xprt_pktdump(char *msg, u32 *packet, unsigned int count)
|
xprt_pktdump(char *msg, u32 *packet, unsigned int count)
|
{
|
{
|
u8 *buf = (u8 *) packet;
|
u8 *buf = (u8 *) packet;
|
int j;
|
int j;
|
|
|
dprintk("RPC: %s\n", msg);
|
dprintk("RPC: %s\n", msg);
|
for (j = 0; j < count && j < 128; j += 4) {
|
for (j = 0; j < count && j < 128; j += 4) {
|
if (!(j & 31)) {
|
if (!(j & 31)) {
|
if (j)
|
if (j)
|
dprintk("\n");
|
dprintk("\n");
|
dprintk("0x%04x ", j);
|
dprintk("0x%04x ", j);
|
}
|
}
|
dprintk("%02x%02x%02x%02x ",
|
dprintk("%02x%02x%02x%02x ",
|
buf[j], buf[j+1], buf[j+2], buf[j+3]);
|
buf[j], buf[j+1], buf[j+2], buf[j+3]);
|
}
|
}
|
dprintk("\n");
|
dprintk("\n");
|
}
|
}
|
#else
|
#else
|
static inline void
|
static inline void
|
xprt_pktdump(char *msg, u32 *packet, unsigned int count)
|
xprt_pktdump(char *msg, u32 *packet, unsigned int count)
|
{
|
{
|
/* NOP */
|
/* NOP */
|
}
|
}
|
#endif
|
#endif
|
|
|
/*
|
/*
|
* Look up RPC transport given an INET socket
|
* Look up RPC transport given an INET socket
|
*/
|
*/
|
static inline struct rpc_xprt *
|
static inline struct rpc_xprt *
|
xprt_from_sock(struct sock *sk)
|
xprt_from_sock(struct sock *sk)
|
{
|
{
|
return (struct rpc_xprt *) sk->user_data;
|
return (struct rpc_xprt *) sk->user_data;
|
}
|
}
|
|
|
/*
|
/*
|
* Serialize write access to sockets, in order to prevent different
|
* Serialize write access to sockets, in order to prevent different
|
* requests from interfering with each other.
|
* requests from interfering with each other.
|
* Also prevents TCP socket connections from colliding with writes.
|
* Also prevents TCP socket connections from colliding with writes.
|
*/
|
*/
|
static int
|
static int
|
__xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
|
__xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
|
{
|
{
|
struct rpc_rqst *req = task->tk_rqstp;
|
struct rpc_rqst *req = task->tk_rqstp;
|
if (!xprt->snd_task) {
|
if (!xprt->snd_task) {
|
if (xprt->nocong || __xprt_get_cong(xprt, task)) {
|
if (xprt->nocong || __xprt_get_cong(xprt, task)) {
|
xprt->snd_task = task;
|
xprt->snd_task = task;
|
if (req) {
|
if (req) {
|
req->rq_bytes_sent = 0;
|
req->rq_bytes_sent = 0;
|
req->rq_ntrans++;
|
req->rq_ntrans++;
|
}
|
}
|
}
|
}
|
}
|
}
|
if (xprt->snd_task != task) {
|
if (xprt->snd_task != task) {
|
dprintk("RPC: %4d TCP write queue full\n", task->tk_pid);
|
dprintk("RPC: %4d TCP write queue full\n", task->tk_pid);
|
task->tk_timeout = 0;
|
task->tk_timeout = 0;
|
task->tk_status = -EAGAIN;
|
task->tk_status = -EAGAIN;
|
if (req && req->rq_ntrans)
|
if (req && req->rq_ntrans)
|
rpc_sleep_on(&xprt->resend, task, NULL, NULL);
|
rpc_sleep_on(&xprt->resend, task, NULL, NULL);
|
else
|
else
|
rpc_sleep_on(&xprt->sending, task, NULL, NULL);
|
rpc_sleep_on(&xprt->sending, task, NULL, NULL);
|
}
|
}
|
return xprt->snd_task == task;
|
return xprt->snd_task == task;
|
}
|
}
|
|
|
static inline int
|
static inline int
|
xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
|
xprt_lock_write(struct rpc_xprt *xprt, struct rpc_task *task)
|
{
|
{
|
int retval;
|
int retval;
|
spin_lock_bh(&xprt->sock_lock);
|
spin_lock_bh(&xprt->sock_lock);
|
retval = __xprt_lock_write(xprt, task);
|
retval = __xprt_lock_write(xprt, task);
|
spin_unlock_bh(&xprt->sock_lock);
|
spin_unlock_bh(&xprt->sock_lock);
|
return retval;
|
return retval;
|
}
|
}
|
|
|
static void
|
static void
|
__xprt_lock_write_next(struct rpc_xprt *xprt)
|
__xprt_lock_write_next(struct rpc_xprt *xprt)
|
{
|
{
|
struct rpc_task *task;
|
struct rpc_task *task;
|
|
|
if (xprt->snd_task)
|
if (xprt->snd_task)
|
return;
|
return;
|
task = rpc_wake_up_next(&xprt->resend);
|
task = rpc_wake_up_next(&xprt->resend);
|
if (!task) {
|
if (!task) {
|
if (!xprt->nocong && RPCXPRT_CONGESTED(xprt))
|
if (!xprt->nocong && RPCXPRT_CONGESTED(xprt))
|
return;
|
return;
|
task = rpc_wake_up_next(&xprt->sending);
|
task = rpc_wake_up_next(&xprt->sending);
|
if (!task)
|
if (!task)
|
return;
|
return;
|
}
|
}
|
if (xprt->nocong || __xprt_get_cong(xprt, task)) {
|
if (xprt->nocong || __xprt_get_cong(xprt, task)) {
|
struct rpc_rqst *req = task->tk_rqstp;
|
struct rpc_rqst *req = task->tk_rqstp;
|
xprt->snd_task = task;
|
xprt->snd_task = task;
|
if (req) {
|
if (req) {
|
req->rq_bytes_sent = 0;
|
req->rq_bytes_sent = 0;
|
req->rq_ntrans++;
|
req->rq_ntrans++;
|
}
|
}
|
}
|
}
|
}
|
}
|
|
|
/*
|
/*
|
* Releases the socket for use by other requests.
|
* Releases the socket for use by other requests.
|
*/
|
*/
|
static void
|
static void
|
__xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
|
__xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
|
{
|
{
|
if (xprt->snd_task == task)
|
if (xprt->snd_task == task)
|
xprt->snd_task = NULL;
|
xprt->snd_task = NULL;
|
__xprt_lock_write_next(xprt);
|
__xprt_lock_write_next(xprt);
|
}
|
}
|
|
|
static inline void
|
static inline void
|
xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
|
xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
|
{
|
{
|
spin_lock_bh(&xprt->sock_lock);
|
spin_lock_bh(&xprt->sock_lock);
|
__xprt_release_write(xprt, task);
|
__xprt_release_write(xprt, task);
|
spin_unlock_bh(&xprt->sock_lock);
|
spin_unlock_bh(&xprt->sock_lock);
|
}
|
}
|
|
|
/*
|
/*
|
* Write data to socket.
|
* Write data to socket.
|
*/
|
*/
|
static inline int
|
static inline int
|
xprt_sendmsg(struct rpc_xprt *xprt, struct rpc_rqst *req)
|
xprt_sendmsg(struct rpc_xprt *xprt, struct rpc_rqst *req)
|
{
|
{
|
struct socket *sock = xprt->sock;
|
struct socket *sock = xprt->sock;
|
struct msghdr msg;
|
struct msghdr msg;
|
struct xdr_buf *xdr = &req->rq_snd_buf;
|
struct xdr_buf *xdr = &req->rq_snd_buf;
|
struct iovec niv[MAX_IOVEC];
|
struct iovec niv[MAX_IOVEC];
|
unsigned int niov, slen, skip;
|
unsigned int niov, slen, skip;
|
mm_segment_t oldfs;
|
mm_segment_t oldfs;
|
int result;
|
int result;
|
|
|
if (!sock)
|
if (!sock)
|
return -ENOTCONN;
|
return -ENOTCONN;
|
|
|
xprt_pktdump("packet data:",
|
xprt_pktdump("packet data:",
|
req->rq_svec->iov_base,
|
req->rq_svec->iov_base,
|
req->rq_svec->iov_len);
|
req->rq_svec->iov_len);
|
|
|
/* Dont repeat bytes */
|
/* Dont repeat bytes */
|
skip = req->rq_bytes_sent;
|
skip = req->rq_bytes_sent;
|
slen = xdr->len - skip;
|
slen = xdr->len - skip;
|
oldfs = get_fs(); set_fs(get_ds());
|
oldfs = get_fs(); set_fs(get_ds());
|
do {
|
do {
|
unsigned int slen_part, n;
|
unsigned int slen_part, n;
|
|
|
niov = xdr_kmap(niv, xdr, skip);
|
niov = xdr_kmap(niv, xdr, skip);
|
if (!niov) {
|
if (!niov) {
|
result = -EAGAIN;
|
result = -EAGAIN;
|
break;
|
break;
|
}
|
}
|
|
|
msg.msg_flags = MSG_DONTWAIT|MSG_NOSIGNAL;
|
msg.msg_flags = MSG_DONTWAIT|MSG_NOSIGNAL;
|
msg.msg_iov = niv;
|
msg.msg_iov = niv;
|
msg.msg_iovlen = niov;
|
msg.msg_iovlen = niov;
|
msg.msg_name = (struct sockaddr *) &xprt->addr;
|
msg.msg_name = (struct sockaddr *) &xprt->addr;
|
msg.msg_namelen = sizeof(xprt->addr);
|
msg.msg_namelen = sizeof(xprt->addr);
|
msg.msg_control = NULL;
|
msg.msg_control = NULL;
|
msg.msg_controllen = 0;
|
msg.msg_controllen = 0;
|
|
|
slen_part = 0;
|
slen_part = 0;
|
for (n = 0; n < niov; n++)
|
for (n = 0; n < niov; n++)
|
slen_part += niv[n].iov_len;
|
slen_part += niv[n].iov_len;
|
|
|
clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags);
|
clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags);
|
result = sock_sendmsg(sock, &msg, slen_part);
|
result = sock_sendmsg(sock, &msg, slen_part);
|
|
|
xdr_kunmap(xdr, skip, niov);
|
xdr_kunmap(xdr, skip, niov);
|
|
|
skip += slen_part;
|
skip += slen_part;
|
slen -= slen_part;
|
slen -= slen_part;
|
} while (result >= 0 && slen);
|
} while (result >= 0 && slen);
|
set_fs(oldfs);
|
set_fs(oldfs);
|
|
|
dprintk("RPC: xprt_sendmsg(%d) = %d\n", slen, result);
|
dprintk("RPC: xprt_sendmsg(%d) = %d\n", slen, result);
|
|
|
if (result >= 0)
|
if (result >= 0)
|
return result;
|
return result;
|
|
|
switch (result) {
|
switch (result) {
|
case -ECONNREFUSED:
|
case -ECONNREFUSED:
|
/* When the server has died, an ICMP port unreachable message
|
/* When the server has died, an ICMP port unreachable message
|
* prompts ECONNREFUSED.
|
* prompts ECONNREFUSED.
|
*/
|
*/
|
case -EAGAIN:
|
case -EAGAIN:
|
break;
|
break;
|
case -ECONNRESET:
|
case -ECONNRESET:
|
case -ENOTCONN:
|
case -ENOTCONN:
|
case -EPIPE:
|
case -EPIPE:
|
/* connection broken */
|
/* connection broken */
|
if (xprt->stream)
|
if (xprt->stream)
|
result = -ENOTCONN;
|
result = -ENOTCONN;
|
break;
|
break;
|
default:
|
default:
|
printk(KERN_NOTICE "RPC: sendmsg returned error %d\n", -result);
|
printk(KERN_NOTICE "RPC: sendmsg returned error %d\n", -result);
|
}
|
}
|
return result;
|
return result;
|
}
|
}
|
|
|
/*
|
/*
|
* Van Jacobson congestion avoidance. Check if the congestion window
|
* Van Jacobson congestion avoidance. Check if the congestion window
|
* overflowed. Put the task to sleep if this is the case.
|
* overflowed. Put the task to sleep if this is the case.
|
*/
|
*/
|
static int
|
static int
|
__xprt_get_cong(struct rpc_xprt *xprt, struct rpc_task *task)
|
__xprt_get_cong(struct rpc_xprt *xprt, struct rpc_task *task)
|
{
|
{
|
struct rpc_rqst *req = task->tk_rqstp;
|
struct rpc_rqst *req = task->tk_rqstp;
|
|
|
if (req->rq_cong)
|
if (req->rq_cong)
|
return 1;
|
return 1;
|
dprintk("RPC: %4d xprt_cwnd_limited cong = %ld cwnd = %ld\n",
|
dprintk("RPC: %4d xprt_cwnd_limited cong = %ld cwnd = %ld\n",
|
task->tk_pid, xprt->cong, xprt->cwnd);
|
task->tk_pid, xprt->cong, xprt->cwnd);
|
if (RPCXPRT_CONGESTED(xprt))
|
if (RPCXPRT_CONGESTED(xprt))
|
return 0;
|
return 0;
|
req->rq_cong = 1;
|
req->rq_cong = 1;
|
xprt->cong += RPC_CWNDSCALE;
|
xprt->cong += RPC_CWNDSCALE;
|
return 1;
|
return 1;
|
}
|
}
|
|
|
/*
|
/*
|
* Adjust the congestion window, and wake up the next task
|
* Adjust the congestion window, and wake up the next task
|
* that has been sleeping due to congestion
|
* that has been sleeping due to congestion
|
*/
|
*/
|
static void
|
static void
|
__xprt_put_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
|
__xprt_put_cong(struct rpc_xprt *xprt, struct rpc_rqst *req)
|
{
|
{
|
if (!req->rq_cong)
|
if (!req->rq_cong)
|
return;
|
return;
|
req->rq_cong = 0;
|
req->rq_cong = 0;
|
xprt->cong -= RPC_CWNDSCALE;
|
xprt->cong -= RPC_CWNDSCALE;
|
__xprt_lock_write_next(xprt);
|
__xprt_lock_write_next(xprt);
|
}
|
}
|
|
|
/*
|
/*
|
* Adjust RPC congestion window
|
* Adjust RPC congestion window
|
* We use a time-smoothed congestion estimator to avoid heavy oscillation.
|
* We use a time-smoothed congestion estimator to avoid heavy oscillation.
|
*/
|
*/
|
static void
|
static void
|
xprt_adjust_cwnd(struct rpc_xprt *xprt, int result)
|
xprt_adjust_cwnd(struct rpc_xprt *xprt, int result)
|
{
|
{
|
unsigned long cwnd;
|
unsigned long cwnd;
|
|
|
cwnd = xprt->cwnd;
|
cwnd = xprt->cwnd;
|
if (result >= 0 && cwnd <= xprt->cong) {
|
if (result >= 0 && cwnd <= xprt->cong) {
|
/* The (cwnd >> 1) term makes sure
|
/* The (cwnd >> 1) term makes sure
|
* the result gets rounded properly. */
|
* the result gets rounded properly. */
|
cwnd += (RPC_CWNDSCALE * RPC_CWNDSCALE + (cwnd >> 1)) / cwnd;
|
cwnd += (RPC_CWNDSCALE * RPC_CWNDSCALE + (cwnd >> 1)) / cwnd;
|
if (cwnd > RPC_MAXCWND)
|
if (cwnd > RPC_MAXCWND)
|
cwnd = RPC_MAXCWND;
|
cwnd = RPC_MAXCWND;
|
__xprt_lock_write_next(xprt);
|
__xprt_lock_write_next(xprt);
|
} else if (result == -ETIMEDOUT) {
|
} else if (result == -ETIMEDOUT) {
|
cwnd >>= 1;
|
cwnd >>= 1;
|
if (cwnd < RPC_CWNDSCALE)
|
if (cwnd < RPC_CWNDSCALE)
|
cwnd = RPC_CWNDSCALE;
|
cwnd = RPC_CWNDSCALE;
|
}
|
}
|
dprintk("RPC: cong %ld, cwnd was %ld, now %ld\n",
|
dprintk("RPC: cong %ld, cwnd was %ld, now %ld\n",
|
xprt->cong, xprt->cwnd, cwnd);
|
xprt->cong, xprt->cwnd, cwnd);
|
xprt->cwnd = cwnd;
|
xprt->cwnd = cwnd;
|
}
|
}
|
|
|
/*
|
/*
|
* Adjust timeout values etc for next retransmit
|
* Adjust timeout values etc for next retransmit
|
*/
|
*/
|
int
|
int
|
xprt_adjust_timeout(struct rpc_timeout *to)
|
xprt_adjust_timeout(struct rpc_timeout *to)
|
{
|
{
|
if (to->to_retries > 0) {
|
if (to->to_retries > 0) {
|
if (to->to_exponential)
|
if (to->to_exponential)
|
to->to_current <<= 1;
|
to->to_current <<= 1;
|
else
|
else
|
to->to_current += to->to_increment;
|
to->to_current += to->to_increment;
|
if (to->to_maxval && to->to_current >= to->to_maxval)
|
if (to->to_maxval && to->to_current >= to->to_maxval)
|
to->to_current = to->to_maxval;
|
to->to_current = to->to_maxval;
|
} else {
|
} else {
|
if (to->to_exponential)
|
if (to->to_exponential)
|
to->to_initval <<= 1;
|
to->to_initval <<= 1;
|
else
|
else
|
to->to_initval += to->to_increment;
|
to->to_initval += to->to_increment;
|
if (to->to_maxval && to->to_initval >= to->to_maxval)
|
if (to->to_maxval && to->to_initval >= to->to_maxval)
|
to->to_initval = to->to_maxval;
|
to->to_initval = to->to_maxval;
|
to->to_current = to->to_initval;
|
to->to_current = to->to_initval;
|
}
|
}
|
|
|
if (!to->to_current) {
|
if (!to->to_current) {
|
printk(KERN_WARNING "xprt_adjust_timeout: to_current = 0!\n");
|
printk(KERN_WARNING "xprt_adjust_timeout: to_current = 0!\n");
|
to->to_current = 5 * HZ;
|
to->to_current = 5 * HZ;
|
}
|
}
|
pprintk("RPC: %lu %s\n", jiffies,
|
pprintk("RPC: %lu %s\n", jiffies,
|
to->to_retries? "retrans" : "timeout");
|
to->to_retries? "retrans" : "timeout");
|
return to->to_retries-- > 0;
|
return to->to_retries-- > 0;
|
}
|
}
|
|
|
/*
|
/*
|
* Close down a transport socket
|
* Close down a transport socket
|
*/
|
*/
|
static void
|
static void
|
xprt_close(struct rpc_xprt *xprt)
|
xprt_close(struct rpc_xprt *xprt)
|
{
|
{
|
struct socket *sock = xprt->sock;
|
struct socket *sock = xprt->sock;
|
struct sock *sk = xprt->inet;
|
struct sock *sk = xprt->inet;
|
|
|
if (!sk)
|
if (!sk)
|
return;
|
return;
|
|
|
write_lock_bh(&sk->callback_lock);
|
write_lock_bh(&sk->callback_lock);
|
xprt->inet = NULL;
|
xprt->inet = NULL;
|
xprt->sock = NULL;
|
xprt->sock = NULL;
|
|
|
sk->user_data = NULL;
|
sk->user_data = NULL;
|
sk->data_ready = xprt->old_data_ready;
|
sk->data_ready = xprt->old_data_ready;
|
sk->state_change = xprt->old_state_change;
|
sk->state_change = xprt->old_state_change;
|
sk->write_space = xprt->old_write_space;
|
sk->write_space = xprt->old_write_space;
|
write_unlock_bh(&sk->callback_lock);
|
write_unlock_bh(&sk->callback_lock);
|
|
|
xprt_disconnect(xprt);
|
xprt_disconnect(xprt);
|
sk->no_check = 0;
|
sk->no_check = 0;
|
|
|
sock_release(sock);
|
sock_release(sock);
|
}
|
}
|
|
|
/*
|
/*
|
* Mark a transport as disconnected
|
* Mark a transport as disconnected
|
*/
|
*/
|
static void
|
static void
|
xprt_disconnect(struct rpc_xprt *xprt)
|
xprt_disconnect(struct rpc_xprt *xprt)
|
{
|
{
|
dprintk("RPC: disconnected transport %p\n", xprt);
|
dprintk("RPC: disconnected transport %p\n", xprt);
|
spin_lock_bh(&xprt->sock_lock);
|
spin_lock_bh(&xprt->sock_lock);
|
xprt_clear_connected(xprt);
|
xprt_clear_connected(xprt);
|
rpc_wake_up_status(&xprt->pending, -ENOTCONN);
|
rpc_wake_up_status(&xprt->pending, -ENOTCONN);
|
spin_unlock_bh(&xprt->sock_lock);
|
spin_unlock_bh(&xprt->sock_lock);
|
}
|
}
|
|
|
/*
|
/*
|
* Reconnect a broken TCP connection.
|
* Reconnect a broken TCP connection.
|
*
|
*
|
*/
|
*/
|
void
|
void
|
xprt_connect(struct rpc_task *task)
|
xprt_connect(struct rpc_task *task)
|
{
|
{
|
struct rpc_xprt *xprt = task->tk_xprt;
|
struct rpc_xprt *xprt = task->tk_xprt;
|
struct socket *sock = xprt->sock;
|
struct socket *sock = xprt->sock;
|
struct sock *inet;
|
struct sock *inet;
|
int status;
|
int status;
|
|
|
dprintk("RPC: %4d xprt_connect %p connected %d\n",
|
dprintk("RPC: %4d xprt_connect %p connected %d\n",
|
task->tk_pid, xprt, xprt_connected(xprt));
|
task->tk_pid, xprt, xprt_connected(xprt));
|
if (xprt->shutdown)
|
if (xprt->shutdown)
|
return;
|
return;
|
|
|
if (!xprt->addr.sin_port) {
|
if (!xprt->addr.sin_port) {
|
task->tk_status = -EIO;
|
task->tk_status = -EIO;
|
return;
|
return;
|
}
|
}
|
|
|
if (!xprt_lock_write(xprt, task))
|
if (!xprt_lock_write(xprt, task))
|
return;
|
return;
|
if (xprt_connected(xprt))
|
if (xprt_connected(xprt))
|
goto out_write;
|
goto out_write;
|
|
|
if (task->tk_rqstp)
|
if (task->tk_rqstp)
|
task->tk_rqstp->rq_bytes_sent = 0;
|
task->tk_rqstp->rq_bytes_sent = 0;
|
|
|
xprt_close(xprt);
|
xprt_close(xprt);
|
/* Create an unconnected socket */
|
/* Create an unconnected socket */
|
sock = xprt_create_socket(xprt->prot, &xprt->timeout, xprt->resvport);
|
sock = xprt_create_socket(xprt->prot, &xprt->timeout, xprt->resvport);
|
if (!sock) {
|
if (!sock) {
|
/* couldn't create socket or bind to reserved port;
|
/* couldn't create socket or bind to reserved port;
|
* this is likely a permanent error, so cause an abort */
|
* this is likely a permanent error, so cause an abort */
|
task->tk_status = -EIO;
|
task->tk_status = -EIO;
|
goto out_write;
|
goto out_write;
|
}
|
}
|
xprt_bind_socket(xprt, sock);
|
xprt_bind_socket(xprt, sock);
|
|
|
if (!xprt->stream)
|
if (!xprt->stream)
|
goto out_write;
|
goto out_write;
|
|
|
inet = sock->sk;
|
inet = sock->sk;
|
|
|
/* Now connect it asynchronously. */
|
/* Now connect it asynchronously. */
|
dprintk("RPC: %4d connecting new socket\n", task->tk_pid);
|
dprintk("RPC: %4d connecting new socket\n", task->tk_pid);
|
status = sock->ops->connect(sock, (struct sockaddr *) &xprt->addr,
|
status = sock->ops->connect(sock, (struct sockaddr *) &xprt->addr,
|
sizeof(xprt->addr), O_NONBLOCK);
|
sizeof(xprt->addr), O_NONBLOCK);
|
dprintk("RPC: %4d connect status %d connected %d\n",
|
dprintk("RPC: %4d connect status %d connected %d\n",
|
task->tk_pid, status, xprt_connected(xprt));
|
task->tk_pid, status, xprt_connected(xprt));
|
|
|
if (status >= 0)
|
if (status >= 0)
|
return;
|
return;
|
|
|
switch (status) {
|
switch (status) {
|
case -EALREADY:
|
case -EALREADY:
|
case -EINPROGRESS:
|
case -EINPROGRESS:
|
/* Protect against TCP socket state changes */
|
/* Protect against TCP socket state changes */
|
lock_sock(inet);
|
lock_sock(inet);
|
if (inet->state != TCP_ESTABLISHED) {
|
if (inet->state != TCP_ESTABLISHED) {
|
dprintk("RPC: %4d waiting for connection\n",
|
dprintk("RPC: %4d waiting for connection\n",
|
task->tk_pid);
|
task->tk_pid);
|
task->tk_timeout = RPC_CONNECT_TIMEOUT;
|
task->tk_timeout = RPC_CONNECT_TIMEOUT;
|
/* if the socket is already closing, delay briefly */
|
/* if the socket is already closing, delay briefly */
|
if ((1<<inet->state) & ~(TCPF_SYN_SENT|TCPF_SYN_RECV))
|
if ((1<<inet->state) & ~(TCPF_SYN_SENT|TCPF_SYN_RECV))
|
task->tk_timeout = RPC_REESTABLISH_TIMEOUT;
|
task->tk_timeout = RPC_REESTABLISH_TIMEOUT;
|
rpc_sleep_on(&xprt->pending, task, xprt_connect_status,
|
rpc_sleep_on(&xprt->pending, task, xprt_connect_status,
|
NULL);
|
NULL);
|
}
|
}
|
release_sock(inet);
|
release_sock(inet);
|
break;
|
break;
|
case -ECONNREFUSED:
|
case -ECONNREFUSED:
|
case -ECONNRESET:
|
case -ECONNRESET:
|
case -ENOTCONN:
|
case -ENOTCONN:
|
if (!task->tk_client->cl_softrtry) {
|
if (!task->tk_client->cl_softrtry) {
|
rpc_delay(task, RPC_REESTABLISH_TIMEOUT);
|
rpc_delay(task, RPC_REESTABLISH_TIMEOUT);
|
task->tk_status = -ENOTCONN;
|
task->tk_status = -ENOTCONN;
|
break;
|
break;
|
}
|
}
|
default:
|
default:
|
/* Report myriad other possible returns. If this file
|
/* Report myriad other possible returns. If this file
|
* system is soft mounted, just error out, like Solaris. */
|
* system is soft mounted, just error out, like Solaris. */
|
if (task->tk_client->cl_softrtry) {
|
if (task->tk_client->cl_softrtry) {
|
printk(KERN_WARNING
|
printk(KERN_WARNING
|
"RPC: error %d connecting to server %s, exiting\n",
|
"RPC: error %d connecting to server %s, exiting\n",
|
-status, task->tk_client->cl_server);
|
-status, task->tk_client->cl_server);
|
task->tk_status = -EIO;
|
task->tk_status = -EIO;
|
goto out_write;
|
goto out_write;
|
}
|
}
|
printk(KERN_WARNING "RPC: error %d connecting to server %s\n",
|
printk(KERN_WARNING "RPC: error %d connecting to server %s\n",
|
-status, task->tk_client->cl_server);
|
-status, task->tk_client->cl_server);
|
/* This will prevent anybody else from connecting */
|
/* This will prevent anybody else from connecting */
|
rpc_delay(task, RPC_REESTABLISH_TIMEOUT);
|
rpc_delay(task, RPC_REESTABLISH_TIMEOUT);
|
task->tk_status = status;
|
task->tk_status = status;
|
break;
|
break;
|
}
|
}
|
return;
|
return;
|
out_write:
|
out_write:
|
xprt_release_write(xprt, task);
|
xprt_release_write(xprt, task);
|
}
|
}
|
|
|
/*
|
/*
|
* We arrive here when awoken from waiting on connection establishment.
|
* We arrive here when awoken from waiting on connection establishment.
|
*/
|
*/
|
static void
|
static void
|
xprt_connect_status(struct rpc_task *task)
|
xprt_connect_status(struct rpc_task *task)
|
{
|
{
|
struct rpc_xprt *xprt = task->tk_xprt;
|
struct rpc_xprt *xprt = task->tk_xprt;
|
|
|
if (task->tk_status >= 0) {
|
if (task->tk_status >= 0) {
|
dprintk("RPC: %4d xprt_connect_status: connection established\n",
|
dprintk("RPC: %4d xprt_connect_status: connection established\n",
|
task->tk_pid);
|
task->tk_pid);
|
return;
|
return;
|
}
|
}
|
|
|
/* if soft mounted, cause this RPC to fail */
|
/* if soft mounted, cause this RPC to fail */
|
if (task->tk_client->cl_softrtry)
|
if (task->tk_client->cl_softrtry)
|
task->tk_status = -EIO;
|
task->tk_status = -EIO;
|
|
|
switch (task->tk_status) {
|
switch (task->tk_status) {
|
case -ENOTCONN:
|
case -ENOTCONN:
|
rpc_delay(task, RPC_REESTABLISH_TIMEOUT);
|
rpc_delay(task, RPC_REESTABLISH_TIMEOUT);
|
return;
|
return;
|
case -ETIMEDOUT:
|
case -ETIMEDOUT:
|
dprintk("RPC: %4d xprt_connect_status: timed out\n",
|
dprintk("RPC: %4d xprt_connect_status: timed out\n",
|
task->tk_pid);
|
task->tk_pid);
|
break;
|
break;
|
default:
|
default:
|
printk(KERN_ERR "RPC: error %d connecting to server %s\n",
|
printk(KERN_ERR "RPC: error %d connecting to server %s\n",
|
-task->tk_status, task->tk_client->cl_server);
|
-task->tk_status, task->tk_client->cl_server);
|
}
|
}
|
xprt_release_write(xprt, task);
|
xprt_release_write(xprt, task);
|
}
|
}
|
|
|
/*
|
/*
|
* Look up the RPC request corresponding to a reply, and then lock it.
|
* Look up the RPC request corresponding to a reply, and then lock it.
|
*/
|
*/
|
static inline struct rpc_rqst *
|
static inline struct rpc_rqst *
|
xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid)
|
xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid)
|
{
|
{
|
struct list_head *pos;
|
struct list_head *pos;
|
struct rpc_rqst *req = NULL;
|
struct rpc_rqst *req = NULL;
|
|
|
list_for_each(pos, &xprt->recv) {
|
list_for_each(pos, &xprt->recv) {
|
struct rpc_rqst *entry = list_entry(pos, struct rpc_rqst, rq_list);
|
struct rpc_rqst *entry = list_entry(pos, struct rpc_rqst, rq_list);
|
if (entry->rq_xid == xid) {
|
if (entry->rq_xid == xid) {
|
req = entry;
|
req = entry;
|
break;
|
break;
|
}
|
}
|
}
|
}
|
return req;
|
return req;
|
}
|
}
|
|
|
/*
|
/*
|
* Complete reply received.
|
* Complete reply received.
|
* The TCP code relies on us to remove the request from xprt->pending.
|
* The TCP code relies on us to remove the request from xprt->pending.
|
*/
|
*/
|
static void
|
static void
|
xprt_complete_rqst(struct rpc_xprt *xprt, struct rpc_rqst *req, int copied)
|
xprt_complete_rqst(struct rpc_xprt *xprt, struct rpc_rqst *req, int copied)
|
{
|
{
|
struct rpc_task *task = req->rq_task;
|
struct rpc_task *task = req->rq_task;
|
struct rpc_clnt *clnt = task->tk_client;
|
struct rpc_clnt *clnt = task->tk_client;
|
|
|
/* Adjust congestion window */
|
/* Adjust congestion window */
|
if (!xprt->nocong) {
|
if (!xprt->nocong) {
|
int timer = rpcproc_timer(clnt, task->tk_msg.rpc_proc);
|
int timer = rpcproc_timer(clnt, task->tk_msg.rpc_proc);
|
xprt_adjust_cwnd(xprt, copied);
|
xprt_adjust_cwnd(xprt, copied);
|
__xprt_put_cong(xprt, req);
|
__xprt_put_cong(xprt, req);
|
if (req->rq_ntrans == 1) {
|
if (req->rq_ntrans == 1) {
|
if (timer)
|
if (timer)
|
rpc_update_rtt(&clnt->cl_rtt, timer, (long)jiffies - req->rq_xtime);
|
rpc_update_rtt(&clnt->cl_rtt, timer, (long)jiffies - req->rq_xtime);
|
}
|
}
|
rpc_set_timeo(&clnt->cl_rtt, timer, req->rq_ntrans - 1);
|
rpc_set_timeo(&clnt->cl_rtt, timer, req->rq_ntrans - 1);
|
}
|
}
|
|
|
#ifdef RPC_PROFILE
|
#ifdef RPC_PROFILE
|
/* Profile only reads for now */
|
/* Profile only reads for now */
|
if (copied > 1024) {
|
if (copied > 1024) {
|
static unsigned long nextstat = 0;
|
static unsigned long nextstat = 0;
|
static unsigned long pkt_rtt = 0, pkt_len = 0, pkt_cnt = 0;
|
static unsigned long pkt_rtt = 0, pkt_len = 0, pkt_cnt = 0;
|
|
|
pkt_cnt++;
|
pkt_cnt++;
|
pkt_len += req->rq_slen + copied;
|
pkt_len += req->rq_slen + copied;
|
pkt_rtt += jiffies - req->rq_xtime;
|
pkt_rtt += jiffies - req->rq_xtime;
|
if (time_before(nextstat, jiffies)) {
|
if (time_before(nextstat, jiffies)) {
|
printk("RPC: %lu %ld cwnd\n", jiffies, xprt->cwnd);
|
printk("RPC: %lu %ld cwnd\n", jiffies, xprt->cwnd);
|
printk("RPC: %ld %ld %ld %ld stat\n",
|
printk("RPC: %ld %ld %ld %ld stat\n",
|
jiffies, pkt_cnt, pkt_len, pkt_rtt);
|
jiffies, pkt_cnt, pkt_len, pkt_rtt);
|
pkt_rtt = pkt_len = pkt_cnt = 0;
|
pkt_rtt = pkt_len = pkt_cnt = 0;
|
nextstat = jiffies + 5 * HZ;
|
nextstat = jiffies + 5 * HZ;
|
}
|
}
|
}
|
}
|
#endif
|
#endif
|
|
|
dprintk("RPC: %4d has input (%d bytes)\n", task->tk_pid, copied);
|
dprintk("RPC: %4d has input (%d bytes)\n", task->tk_pid, copied);
|
req->rq_received = copied;
|
req->rq_received = copied;
|
list_del_init(&req->rq_list);
|
list_del_init(&req->rq_list);
|
|
|
/* ... and wake up the process. */
|
/* ... and wake up the process. */
|
rpc_wake_up_task(task);
|
rpc_wake_up_task(task);
|
return;
|
return;
|
}
|
}
|
|
|
static size_t
|
static size_t
|
skb_read_bits(skb_reader_t *desc, void *to, size_t len)
|
skb_read_bits(skb_reader_t *desc, void *to, size_t len)
|
{
|
{
|
if (len > desc->count)
|
if (len > desc->count)
|
len = desc->count;
|
len = desc->count;
|
skb_copy_bits(desc->skb, desc->offset, to, len);
|
skb_copy_bits(desc->skb, desc->offset, to, len);
|
desc->count -= len;
|
desc->count -= len;
|
desc->offset += len;
|
desc->offset += len;
|
return len;
|
return len;
|
}
|
}
|
|
|
static size_t
|
static size_t
|
skb_read_and_csum_bits(skb_reader_t *desc, void *to, size_t len)
|
skb_read_and_csum_bits(skb_reader_t *desc, void *to, size_t len)
|
{
|
{
|
unsigned int csum2, pos;
|
unsigned int csum2, pos;
|
|
|
if (len > desc->count)
|
if (len > desc->count)
|
len = desc->count;
|
len = desc->count;
|
pos = desc->offset;
|
pos = desc->offset;
|
csum2 = skb_copy_and_csum_bits(desc->skb, pos, to, len, 0);
|
csum2 = skb_copy_and_csum_bits(desc->skb, pos, to, len, 0);
|
desc->csum = csum_block_add(desc->csum, csum2, pos);
|
desc->csum = csum_block_add(desc->csum, csum2, pos);
|
desc->count -= len;
|
desc->count -= len;
|
desc->offset += len;
|
desc->offset += len;
|
return len;
|
return len;
|
}
|
}
|
|
|
/*
|
/*
|
* We have set things up such that we perform the checksum of the UDP
|
* We have set things up such that we perform the checksum of the UDP
|
* packet in parallel with the copies into the RPC client iovec. -DaveM
|
* packet in parallel with the copies into the RPC client iovec. -DaveM
|
*/
|
*/
|
static int
|
static int
|
csum_partial_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb)
|
csum_partial_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb)
|
{
|
{
|
skb_reader_t desc;
|
skb_reader_t desc;
|
|
|
desc.skb = skb;
|
desc.skb = skb;
|
desc.offset = sizeof(struct udphdr);
|
desc.offset = sizeof(struct udphdr);
|
desc.count = skb->len - desc.offset;
|
desc.count = skb->len - desc.offset;
|
|
|
if (skb->ip_summed == CHECKSUM_UNNECESSARY)
|
if (skb->ip_summed == CHECKSUM_UNNECESSARY)
|
goto no_checksum;
|
goto no_checksum;
|
|
|
desc.csum = csum_partial(skb->data, desc.offset, skb->csum);
|
desc.csum = csum_partial(skb->data, desc.offset, skb->csum);
|
xdr_partial_copy_from_skb(xdr, 0, &desc, skb_read_and_csum_bits);
|
xdr_partial_copy_from_skb(xdr, 0, &desc, skb_read_and_csum_bits);
|
if (desc.offset != skb->len) {
|
if (desc.offset != skb->len) {
|
unsigned int csum2;
|
unsigned int csum2;
|
csum2 = skb_checksum(skb, desc.offset, skb->len - desc.offset, 0);
|
csum2 = skb_checksum(skb, desc.offset, skb->len - desc.offset, 0);
|
desc.csum = csum_block_add(desc.csum, csum2, desc.offset);
|
desc.csum = csum_block_add(desc.csum, csum2, desc.offset);
|
}
|
}
|
if ((unsigned short)csum_fold(desc.csum))
|
if ((unsigned short)csum_fold(desc.csum))
|
return -1;
|
return -1;
|
return 0;
|
return 0;
|
no_checksum:
|
no_checksum:
|
xdr_partial_copy_from_skb(xdr, 0, &desc, skb_read_bits);
|
xdr_partial_copy_from_skb(xdr, 0, &desc, skb_read_bits);
|
return 0;
|
return 0;
|
}
|
}
|
|
|
/*
|
/*
|
* Input handler for RPC replies. Called from a bottom half and hence
|
* Input handler for RPC replies. Called from a bottom half and hence
|
* atomic.
|
* atomic.
|
*/
|
*/
|
static void
|
static void
|
udp_data_ready(struct sock *sk, int len)
|
udp_data_ready(struct sock *sk, int len)
|
{
|
{
|
struct rpc_task *task;
|
struct rpc_task *task;
|
struct rpc_xprt *xprt;
|
struct rpc_xprt *xprt;
|
struct rpc_rqst *rovr;
|
struct rpc_rqst *rovr;
|
struct sk_buff *skb;
|
struct sk_buff *skb;
|
int err, repsize, copied;
|
int err, repsize, copied;
|
|
|
read_lock(&sk->callback_lock);
|
read_lock(&sk->callback_lock);
|
dprintk("RPC: udp_data_ready...\n");
|
dprintk("RPC: udp_data_ready...\n");
|
if (sk->dead || !(xprt = xprt_from_sock(sk))) {
|
if (sk->dead || !(xprt = xprt_from_sock(sk))) {
|
printk("RPC: udp_data_ready request not found!\n");
|
printk("RPC: udp_data_ready request not found!\n");
|
goto out;
|
goto out;
|
}
|
}
|
|
|
dprintk("RPC: udp_data_ready client %p\n", xprt);
|
dprintk("RPC: udp_data_ready client %p\n", xprt);
|
|
|
if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL)
|
if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL)
|
goto out;
|
goto out;
|
|
|
if (xprt->shutdown)
|
if (xprt->shutdown)
|
goto dropit;
|
goto dropit;
|
|
|
repsize = skb->len - sizeof(struct udphdr);
|
repsize = skb->len - sizeof(struct udphdr);
|
if (repsize < 4) {
|
if (repsize < 4) {
|
printk("RPC: impossible RPC reply size %d!\n", repsize);
|
printk("RPC: impossible RPC reply size %d!\n", repsize);
|
goto dropit;
|
goto dropit;
|
}
|
}
|
|
|
/* Look up and lock the request corresponding to the given XID */
|
/* Look up and lock the request corresponding to the given XID */
|
spin_lock(&xprt->sock_lock);
|
spin_lock(&xprt->sock_lock);
|
rovr = xprt_lookup_rqst(xprt, *(u32 *) (skb->h.raw + sizeof(struct udphdr)));
|
rovr = xprt_lookup_rqst(xprt, *(u32 *) (skb->h.raw + sizeof(struct udphdr)));
|
if (!rovr)
|
if (!rovr)
|
goto out_unlock;
|
goto out_unlock;
|
task = rovr->rq_task;
|
task = rovr->rq_task;
|
|
|
dprintk("RPC: %4d received reply\n", task->tk_pid);
|
dprintk("RPC: %4d received reply\n", task->tk_pid);
|
xprt_pktdump("packet data:",
|
xprt_pktdump("packet data:",
|
(u32 *) (skb->h.raw+sizeof(struct udphdr)), repsize);
|
(u32 *) (skb->h.raw+sizeof(struct udphdr)), repsize);
|
|
|
if ((copied = rovr->rq_private_buf.len) > repsize)
|
if ((copied = rovr->rq_private_buf.len) > repsize)
|
copied = repsize;
|
copied = repsize;
|
|
|
/* Suck it into the iovec, verify checksum if not done by hw. */
|
/* Suck it into the iovec, verify checksum if not done by hw. */
|
if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb))
|
if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb))
|
goto out_unlock;
|
goto out_unlock;
|
|
|
/* Something worked... */
|
/* Something worked... */
|
dst_confirm(skb->dst);
|
dst_confirm(skb->dst);
|
|
|
xprt_complete_rqst(xprt, rovr, copied);
|
xprt_complete_rqst(xprt, rovr, copied);
|
|
|
out_unlock:
|
out_unlock:
|
spin_unlock(&xprt->sock_lock);
|
spin_unlock(&xprt->sock_lock);
|
dropit:
|
dropit:
|
skb_free_datagram(sk, skb);
|
skb_free_datagram(sk, skb);
|
out:
|
out:
|
if (sk->sleep && waitqueue_active(sk->sleep))
|
if (sk->sleep && waitqueue_active(sk->sleep))
|
wake_up_interruptible(sk->sleep);
|
wake_up_interruptible(sk->sleep);
|
read_unlock(&sk->callback_lock);
|
read_unlock(&sk->callback_lock);
|
}
|
}
|
|
|
/*
|
/*
|
* Copy from an skb into memory and shrink the skb.
|
* Copy from an skb into memory and shrink the skb.
|
*/
|
*/
|
static inline size_t
|
static inline size_t
|
tcp_copy_data(skb_reader_t *desc, void *p, size_t len)
|
tcp_copy_data(skb_reader_t *desc, void *p, size_t len)
|
{
|
{
|
if (len > desc->count)
|
if (len > desc->count)
|
len = desc->count;
|
len = desc->count;
|
skb_copy_bits(desc->skb, desc->offset, p, len);
|
skb_copy_bits(desc->skb, desc->offset, p, len);
|
desc->offset += len;
|
desc->offset += len;
|
desc->count -= len;
|
desc->count -= len;
|
return len;
|
return len;
|
}
|
}
|
|
|
/*
|
/*
|
* TCP read fragment marker
|
* TCP read fragment marker
|
*/
|
*/
|
static inline void
|
static inline void
|
tcp_read_fraghdr(struct rpc_xprt *xprt, skb_reader_t *desc)
|
tcp_read_fraghdr(struct rpc_xprt *xprt, skb_reader_t *desc)
|
{
|
{
|
size_t len, used;
|
size_t len, used;
|
char *p;
|
char *p;
|
|
|
p = ((char *) &xprt->tcp_recm) + xprt->tcp_offset;
|
p = ((char *) &xprt->tcp_recm) + xprt->tcp_offset;
|
len = sizeof(xprt->tcp_recm) - xprt->tcp_offset;
|
len = sizeof(xprt->tcp_recm) - xprt->tcp_offset;
|
used = tcp_copy_data(desc, p, len);
|
used = tcp_copy_data(desc, p, len);
|
xprt->tcp_offset += used;
|
xprt->tcp_offset += used;
|
if (used != len)
|
if (used != len)
|
return;
|
return;
|
xprt->tcp_reclen = ntohl(xprt->tcp_recm);
|
xprt->tcp_reclen = ntohl(xprt->tcp_recm);
|
if (xprt->tcp_reclen & 0x80000000)
|
if (xprt->tcp_reclen & 0x80000000)
|
xprt->tcp_flags |= XPRT_LAST_FRAG;
|
xprt->tcp_flags |= XPRT_LAST_FRAG;
|
else
|
else
|
xprt->tcp_flags &= ~XPRT_LAST_FRAG;
|
xprt->tcp_flags &= ~XPRT_LAST_FRAG;
|
xprt->tcp_reclen &= 0x7fffffff;
|
xprt->tcp_reclen &= 0x7fffffff;
|
xprt->tcp_flags &= ~XPRT_COPY_RECM;
|
xprt->tcp_flags &= ~XPRT_COPY_RECM;
|
xprt->tcp_offset = 0;
|
xprt->tcp_offset = 0;
|
/* Sanity check of the record length */
|
/* Sanity check of the record length */
|
if (xprt->tcp_reclen < 4) {
|
if (xprt->tcp_reclen < 4) {
|
printk(KERN_ERR "RPC: Invalid TCP record fragment length\n");
|
printk(KERN_ERR "RPC: Invalid TCP record fragment length\n");
|
xprt_disconnect(xprt);
|
xprt_disconnect(xprt);
|
}
|
}
|
dprintk("RPC: reading TCP record fragment of length %d\n",
|
dprintk("RPC: reading TCP record fragment of length %d\n",
|
xprt->tcp_reclen);
|
xprt->tcp_reclen);
|
}
|
}
|
|
|
static void
|
static void
|
tcp_check_recm(struct rpc_xprt *xprt)
|
tcp_check_recm(struct rpc_xprt *xprt)
|
{
|
{
|
if (xprt->tcp_offset == xprt->tcp_reclen) {
|
if (xprt->tcp_offset == xprt->tcp_reclen) {
|
xprt->tcp_flags |= XPRT_COPY_RECM;
|
xprt->tcp_flags |= XPRT_COPY_RECM;
|
xprt->tcp_offset = 0;
|
xprt->tcp_offset = 0;
|
if (xprt->tcp_flags & XPRT_LAST_FRAG) {
|
if (xprt->tcp_flags & XPRT_LAST_FRAG) {
|
xprt->tcp_flags &= ~XPRT_COPY_DATA;
|
xprt->tcp_flags &= ~XPRT_COPY_DATA;
|
xprt->tcp_flags |= XPRT_COPY_XID;
|
xprt->tcp_flags |= XPRT_COPY_XID;
|
xprt->tcp_copied = 0;
|
xprt->tcp_copied = 0;
|
}
|
}
|
}
|
}
|
}
|
}
|
|
|
/*
|
/*
|
* TCP read xid
|
* TCP read xid
|
*/
|
*/
|
static inline void
|
static inline void
|
tcp_read_xid(struct rpc_xprt *xprt, skb_reader_t *desc)
|
tcp_read_xid(struct rpc_xprt *xprt, skb_reader_t *desc)
|
{
|
{
|
size_t len, used;
|
size_t len, used;
|
char *p;
|
char *p;
|
|
|
len = sizeof(xprt->tcp_xid) - xprt->tcp_offset;
|
len = sizeof(xprt->tcp_xid) - xprt->tcp_offset;
|
dprintk("RPC: reading XID (%Zu bytes)\n", len);
|
dprintk("RPC: reading XID (%Zu bytes)\n", len);
|
p = ((char *) &xprt->tcp_xid) + xprt->tcp_offset;
|
p = ((char *) &xprt->tcp_xid) + xprt->tcp_offset;
|
used = tcp_copy_data(desc, p, len);
|
used = tcp_copy_data(desc, p, len);
|
xprt->tcp_offset += used;
|
xprt->tcp_offset += used;
|
if (used != len)
|
if (used != len)
|
return;
|
return;
|
xprt->tcp_flags &= ~XPRT_COPY_XID;
|
xprt->tcp_flags &= ~XPRT_COPY_XID;
|
xprt->tcp_flags |= XPRT_COPY_DATA;
|
xprt->tcp_flags |= XPRT_COPY_DATA;
|
xprt->tcp_copied = 4;
|
xprt->tcp_copied = 4;
|
dprintk("RPC: reading reply for XID %08x\n", xprt->tcp_xid);
|
dprintk("RPC: reading reply for XID %08x\n", xprt->tcp_xid);
|
tcp_check_recm(xprt);
|
tcp_check_recm(xprt);
|
}
|
}
|
|
|
/*
|
/*
|
* TCP read and complete request
|
* TCP read and complete request
|
*/
|
*/
|
static inline void
|
static inline void
|
tcp_read_request(struct rpc_xprt *xprt, skb_reader_t *desc)
|
tcp_read_request(struct rpc_xprt *xprt, skb_reader_t *desc)
|
{
|
{
|
struct rpc_rqst *req;
|
struct rpc_rqst *req;
|
struct xdr_buf *rcvbuf;
|
struct xdr_buf *rcvbuf;
|
size_t len;
|
size_t len;
|
|
|
/* Find and lock the request corresponding to this xid */
|
/* Find and lock the request corresponding to this xid */
|
spin_lock(&xprt->sock_lock);
|
spin_lock(&xprt->sock_lock);
|
req = xprt_lookup_rqst(xprt, xprt->tcp_xid);
|
req = xprt_lookup_rqst(xprt, xprt->tcp_xid);
|
if (!req) {
|
if (!req) {
|
xprt->tcp_flags &= ~XPRT_COPY_DATA;
|
xprt->tcp_flags &= ~XPRT_COPY_DATA;
|
dprintk("RPC: XID %08x request not found!\n",
|
dprintk("RPC: XID %08x request not found!\n",
|
xprt->tcp_xid);
|
xprt->tcp_xid);
|
spin_unlock(&xprt->sock_lock);
|
spin_unlock(&xprt->sock_lock);
|
return;
|
return;
|
}
|
}
|
|
|
rcvbuf = &req->rq_private_buf;
|
rcvbuf = &req->rq_private_buf;
|
len = desc->count;
|
len = desc->count;
|
if (len > xprt->tcp_reclen - xprt->tcp_offset) {
|
if (len > xprt->tcp_reclen - xprt->tcp_offset) {
|
skb_reader_t my_desc;
|
skb_reader_t my_desc;
|
|
|
len = xprt->tcp_reclen - xprt->tcp_offset;
|
len = xprt->tcp_reclen - xprt->tcp_offset;
|
memcpy(&my_desc, desc, sizeof(my_desc));
|
memcpy(&my_desc, desc, sizeof(my_desc));
|
my_desc.count = len;
|
my_desc.count = len;
|
xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied,
|
xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied,
|
&my_desc, tcp_copy_data);
|
&my_desc, tcp_copy_data);
|
desc->count -= len;
|
desc->count -= len;
|
desc->offset += len;
|
desc->offset += len;
|
} else
|
} else
|
xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied,
|
xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied,
|
desc, tcp_copy_data);
|
desc, tcp_copy_data);
|
xprt->tcp_copied += len;
|
xprt->tcp_copied += len;
|
xprt->tcp_offset += len;
|
xprt->tcp_offset += len;
|
|
|
if (xprt->tcp_copied == req->rq_private_buf.len)
|
if (xprt->tcp_copied == req->rq_private_buf.len)
|
xprt->tcp_flags &= ~XPRT_COPY_DATA;
|
xprt->tcp_flags &= ~XPRT_COPY_DATA;
|
else if (xprt->tcp_offset == xprt->tcp_reclen) {
|
else if (xprt->tcp_offset == xprt->tcp_reclen) {
|
if (xprt->tcp_flags & XPRT_LAST_FRAG)
|
if (xprt->tcp_flags & XPRT_LAST_FRAG)
|
xprt->tcp_flags &= ~XPRT_COPY_DATA;
|
xprt->tcp_flags &= ~XPRT_COPY_DATA;
|
}
|
}
|
|
|
if (!(xprt->tcp_flags & XPRT_COPY_DATA)) {
|
if (!(xprt->tcp_flags & XPRT_COPY_DATA)) {
|
dprintk("RPC: %4d received reply complete\n",
|
dprintk("RPC: %4d received reply complete\n",
|
req->rq_task->tk_pid);
|
req->rq_task->tk_pid);
|
xprt_complete_rqst(xprt, req, xprt->tcp_copied);
|
xprt_complete_rqst(xprt, req, xprt->tcp_copied);
|
}
|
}
|
spin_unlock(&xprt->sock_lock);
|
spin_unlock(&xprt->sock_lock);
|
tcp_check_recm(xprt);
|
tcp_check_recm(xprt);
|
}
|
}
|
|
|
/*
|
/*
|
* TCP discard extra bytes from a short read
|
* TCP discard extra bytes from a short read
|
*/
|
*/
|
static inline void
|
static inline void
|
tcp_read_discard(struct rpc_xprt *xprt, skb_reader_t *desc)
|
tcp_read_discard(struct rpc_xprt *xprt, skb_reader_t *desc)
|
{
|
{
|
size_t len;
|
size_t len;
|
|
|
len = xprt->tcp_reclen - xprt->tcp_offset;
|
len = xprt->tcp_reclen - xprt->tcp_offset;
|
if (len > desc->count)
|
if (len > desc->count)
|
len = desc->count;
|
len = desc->count;
|
desc->count -= len;
|
desc->count -= len;
|
desc->offset += len;
|
desc->offset += len;
|
xprt->tcp_offset += len;
|
xprt->tcp_offset += len;
|
tcp_check_recm(xprt);
|
tcp_check_recm(xprt);
|
}
|
}
|
|
|
/*
|
/*
|
* TCP record receive routine
|
* TCP record receive routine
|
* We first have to grab the record marker, then the XID, then the data.
|
* We first have to grab the record marker, then the XID, then the data.
|
*/
|
*/
|
static int
|
static int
|
tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb,
|
tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb,
|
unsigned int offset, size_t len)
|
unsigned int offset, size_t len)
|
{
|
{
|
struct rpc_xprt *xprt = (struct rpc_xprt *)rd_desc->buf;
|
struct rpc_xprt *xprt = (struct rpc_xprt *)rd_desc->buf;
|
skb_reader_t desc = { skb, offset, len };
|
skb_reader_t desc = { skb, offset, len };
|
|
|
dprintk("RPC: tcp_data_recv\n");
|
dprintk("RPC: tcp_data_recv\n");
|
do {
|
do {
|
/* Read in a new fragment marker if necessary */
|
/* Read in a new fragment marker if necessary */
|
/* Can we ever really expect to get completely empty fragments? */
|
/* Can we ever really expect to get completely empty fragments? */
|
if (xprt->tcp_flags & XPRT_COPY_RECM) {
|
if (xprt->tcp_flags & XPRT_COPY_RECM) {
|
tcp_read_fraghdr(xprt, &desc);
|
tcp_read_fraghdr(xprt, &desc);
|
continue;
|
continue;
|
}
|
}
|
/* Read in the xid if necessary */
|
/* Read in the xid if necessary */
|
if (xprt->tcp_flags & XPRT_COPY_XID) {
|
if (xprt->tcp_flags & XPRT_COPY_XID) {
|
tcp_read_xid(xprt, &desc);
|
tcp_read_xid(xprt, &desc);
|
continue;
|
continue;
|
}
|
}
|
/* Read in the request data */
|
/* Read in the request data */
|
if (xprt->tcp_flags & XPRT_COPY_DATA) {
|
if (xprt->tcp_flags & XPRT_COPY_DATA) {
|
tcp_read_request(xprt, &desc);
|
tcp_read_request(xprt, &desc);
|
continue;
|
continue;
|
}
|
}
|
/* Skip over any trailing bytes on short reads */
|
/* Skip over any trailing bytes on short reads */
|
tcp_read_discard(xprt, &desc);
|
tcp_read_discard(xprt, &desc);
|
} while (desc.count);
|
} while (desc.count);
|
dprintk("RPC: tcp_data_recv done\n");
|
dprintk("RPC: tcp_data_recv done\n");
|
return len - desc.count;
|
return len - desc.count;
|
}
|
}
|
|
|
static void tcp_data_ready(struct sock *sk, int bytes)
|
static void tcp_data_ready(struct sock *sk, int bytes)
|
{
|
{
|
struct rpc_xprt *xprt;
|
struct rpc_xprt *xprt;
|
read_descriptor_t rd_desc;
|
read_descriptor_t rd_desc;
|
|
|
read_lock(&sk->callback_lock);
|
read_lock(&sk->callback_lock);
|
dprintk("RPC: tcp_data_ready...\n");
|
dprintk("RPC: tcp_data_ready...\n");
|
if (!(xprt = xprt_from_sock(sk))) {
|
if (!(xprt = xprt_from_sock(sk))) {
|
printk("RPC: tcp_data_ready socket info not found!\n");
|
printk("RPC: tcp_data_ready socket info not found!\n");
|
goto out;
|
goto out;
|
}
|
}
|
if (xprt->shutdown)
|
if (xprt->shutdown)
|
goto out;
|
goto out;
|
|
|
/* We use rd_desc to pass struct xprt to tcp_data_recv */
|
/* We use rd_desc to pass struct xprt to tcp_data_recv */
|
rd_desc.buf = (char *)xprt;
|
rd_desc.buf = (char *)xprt;
|
rd_desc.count = 65536;
|
rd_desc.count = 65536;
|
tcp_read_sock(sk, &rd_desc, tcp_data_recv);
|
tcp_read_sock(sk, &rd_desc, tcp_data_recv);
|
out:
|
out:
|
read_unlock(&sk->callback_lock);
|
read_unlock(&sk->callback_lock);
|
}
|
}
|
|
|
static void
|
static void
|
tcp_state_change(struct sock *sk)
|
tcp_state_change(struct sock *sk)
|
{
|
{
|
struct rpc_xprt *xprt;
|
struct rpc_xprt *xprt;
|
|
|
read_lock(&sk->callback_lock);
|
read_lock(&sk->callback_lock);
|
if (!(xprt = xprt_from_sock(sk)))
|
if (!(xprt = xprt_from_sock(sk)))
|
goto out;
|
goto out;
|
dprintk("RPC: tcp_state_change client %p...\n", xprt);
|
dprintk("RPC: tcp_state_change client %p...\n", xprt);
|
dprintk("RPC: state %x conn %d dead %d zapped %d\n",
|
dprintk("RPC: state %x conn %d dead %d zapped %d\n",
|
sk->state, xprt_connected(xprt),
|
sk->state, xprt_connected(xprt),
|
sk->dead, sk->zapped);
|
sk->dead, sk->zapped);
|
|
|
switch (sk->state) {
|
switch (sk->state) {
|
case TCP_ESTABLISHED:
|
case TCP_ESTABLISHED:
|
if (xprt_test_and_set_connected(xprt))
|
if (xprt_test_and_set_connected(xprt))
|
break;
|
break;
|
|
|
/* Reset TCP record info */
|
/* Reset TCP record info */
|
xprt->tcp_offset = 0;
|
xprt->tcp_offset = 0;
|
xprt->tcp_reclen = 0;
|
xprt->tcp_reclen = 0;
|
xprt->tcp_copied = 0;
|
xprt->tcp_copied = 0;
|
xprt->tcp_flags = XPRT_COPY_RECM | XPRT_COPY_XID;
|
xprt->tcp_flags = XPRT_COPY_RECM | XPRT_COPY_XID;
|
|
|
spin_lock_bh(&xprt->sock_lock);
|
spin_lock_bh(&xprt->sock_lock);
|
if (xprt->snd_task && xprt->snd_task->tk_rpcwait == &xprt->pending)
|
if (xprt->snd_task && xprt->snd_task->tk_rpcwait == &xprt->pending)
|
rpc_wake_up_task(xprt->snd_task);
|
rpc_wake_up_task(xprt->snd_task);
|
spin_unlock_bh(&xprt->sock_lock);
|
spin_unlock_bh(&xprt->sock_lock);
|
break;
|
break;
|
case TCP_SYN_SENT:
|
case TCP_SYN_SENT:
|
case TCP_SYN_RECV:
|
case TCP_SYN_RECV:
|
break;
|
break;
|
default:
|
default:
|
xprt_disconnect(xprt);
|
xprt_disconnect(xprt);
|
break;
|
break;
|
}
|
}
|
out:
|
out:
|
if (sk->sleep && waitqueue_active(sk->sleep))
|
if (sk->sleep && waitqueue_active(sk->sleep))
|
wake_up_interruptible_all(sk->sleep);
|
wake_up_interruptible_all(sk->sleep);
|
read_unlock(&sk->callback_lock);
|
read_unlock(&sk->callback_lock);
|
}
|
}
|
|
|
/*
|
/*
|
* Called when more output buffer space is available for this socket.
|
* Called when more output buffer space is available for this socket.
|
* We try not to wake our writers until they can make "significant"
|
* We try not to wake our writers until they can make "significant"
|
* progress, otherwise we'll waste resources thrashing sock_sendmsg
|
* progress, otherwise we'll waste resources thrashing sock_sendmsg
|
* with a bunch of small requests.
|
* with a bunch of small requests.
|
*/
|
*/
|
static void
|
static void
|
xprt_write_space(struct sock *sk)
|
xprt_write_space(struct sock *sk)
|
{
|
{
|
struct rpc_xprt *xprt;
|
struct rpc_xprt *xprt;
|
struct socket *sock;
|
struct socket *sock;
|
|
|
read_lock(&sk->callback_lock);
|
read_lock(&sk->callback_lock);
|
if (!(xprt = xprt_from_sock(sk)) || !(sock = sk->socket))
|
if (!(xprt = xprt_from_sock(sk)) || !(sock = sk->socket))
|
goto out;
|
goto out;
|
if (xprt->shutdown)
|
if (xprt->shutdown)
|
goto out;
|
goto out;
|
|
|
/* Wait until we have enough socket memory */
|
/* Wait until we have enough socket memory */
|
if (xprt->stream) {
|
if (xprt->stream) {
|
/* from net/ipv4/tcp.c:tcp_write_space */
|
/* from net/ipv4/tcp.c:tcp_write_space */
|
if (tcp_wspace(sk) < tcp_min_write_space(sk))
|
if (tcp_wspace(sk) < tcp_min_write_space(sk))
|
goto out;
|
goto out;
|
} else {
|
} else {
|
/* from net/core/sock.c:sock_def_write_space */
|
/* from net/core/sock.c:sock_def_write_space */
|
if (!sock_writeable(sk))
|
if (!sock_writeable(sk))
|
goto out;
|
goto out;
|
}
|
}
|
|
|
if (!test_and_clear_bit(SOCK_NOSPACE, &sock->flags))
|
if (!test_and_clear_bit(SOCK_NOSPACE, &sock->flags))
|
goto out;
|
goto out;
|
|
|
spin_lock_bh(&xprt->sock_lock);
|
spin_lock_bh(&xprt->sock_lock);
|
if (xprt->snd_task && xprt->snd_task->tk_rpcwait == &xprt->pending)
|
if (xprt->snd_task && xprt->snd_task->tk_rpcwait == &xprt->pending)
|
rpc_wake_up_task(xprt->snd_task);
|
rpc_wake_up_task(xprt->snd_task);
|
spin_unlock_bh(&xprt->sock_lock);
|
spin_unlock_bh(&xprt->sock_lock);
|
if (sk->sleep && waitqueue_active(sk->sleep))
|
if (sk->sleep && waitqueue_active(sk->sleep))
|
wake_up_interruptible(sk->sleep);
|
wake_up_interruptible(sk->sleep);
|
out:
|
out:
|
read_unlock(&sk->callback_lock);
|
read_unlock(&sk->callback_lock);
|
}
|
}
|
|
|
/*
|
/*
|
* RPC receive timeout handler.
|
* RPC receive timeout handler.
|
*/
|
*/
|
static void
|
static void
|
xprt_timer(struct rpc_task *task)
|
xprt_timer(struct rpc_task *task)
|
{
|
{
|
struct rpc_rqst *req = task->tk_rqstp;
|
struct rpc_rqst *req = task->tk_rqstp;
|
struct rpc_xprt *xprt = req->rq_xprt;
|
struct rpc_xprt *xprt = req->rq_xprt;
|
|
|
spin_lock(&xprt->sock_lock);
|
spin_lock(&xprt->sock_lock);
|
if (req->rq_received)
|
if (req->rq_received)
|
goto out;
|
goto out;
|
|
|
xprt_adjust_cwnd(req->rq_xprt, -ETIMEDOUT);
|
xprt_adjust_cwnd(req->rq_xprt, -ETIMEDOUT);
|
__xprt_put_cong(xprt, req);
|
__xprt_put_cong(xprt, req);
|
|
|
dprintk("RPC: %4d xprt_timer (%s request)\n",
|
dprintk("RPC: %4d xprt_timer (%s request)\n",
|
task->tk_pid, req ? "pending" : "backlogged");
|
task->tk_pid, req ? "pending" : "backlogged");
|
|
|
task->tk_status = -ETIMEDOUT;
|
task->tk_status = -ETIMEDOUT;
|
out:
|
out:
|
task->tk_timeout = 0;
|
task->tk_timeout = 0;
|
rpc_wake_up_task(task);
|
rpc_wake_up_task(task);
|
spin_unlock(&xprt->sock_lock);
|
spin_unlock(&xprt->sock_lock);
|
}
|
}
|
|
|
/*
|
/*
|
* Place the actual RPC call.
|
* Place the actual RPC call.
|
* We have to copy the iovec because sendmsg fiddles with its contents.
|
* We have to copy the iovec because sendmsg fiddles with its contents.
|
*/
|
*/
|
void
|
void
|
xprt_transmit(struct rpc_task *task)
|
xprt_transmit(struct rpc_task *task)
|
{
|
{
|
struct rpc_rqst *req = task->tk_rqstp;
|
struct rpc_rqst *req = task->tk_rqstp;
|
struct rpc_xprt *xprt = req->rq_xprt;
|
struct rpc_xprt *xprt = req->rq_xprt;
|
|
|
dprintk("RPC: %4d xprt_transmit(%x)\n", task->tk_pid,
|
dprintk("RPC: %4d xprt_transmit(%x)\n", task->tk_pid,
|
*(u32 *)(req->rq_svec[0].iov_base));
|
*(u32 *)(req->rq_svec[0].iov_base));
|
|
|
if (xprt->shutdown)
|
if (xprt->shutdown)
|
task->tk_status = -EIO;
|
task->tk_status = -EIO;
|
|
|
if (task->tk_status < 0)
|
if (task->tk_status < 0)
|
return;
|
return;
|
|
|
if (task->tk_rpcwait)
|
if (task->tk_rpcwait)
|
rpc_remove_wait_queue(task);
|
rpc_remove_wait_queue(task);
|
|
|
/* set up everything as needed. */
|
/* set up everything as needed. */
|
/* Write the record marker */
|
/* Write the record marker */
|
if (xprt->stream) {
|
if (xprt->stream) {
|
u32 *marker = req->rq_svec[0].iov_base;
|
u32 *marker = req->rq_svec[0].iov_base;
|
|
|
*marker = htonl(0x80000000|(req->rq_slen-sizeof(*marker)));
|
*marker = htonl(0x80000000|(req->rq_slen-sizeof(*marker)));
|
}
|
}
|
|
|
spin_lock_bh(&xprt->sock_lock);
|
spin_lock_bh(&xprt->sock_lock);
|
if (req->rq_received != 0 && !req->rq_bytes_sent)
|
if (req->rq_received != 0 && !req->rq_bytes_sent)
|
goto out_notrans;
|
goto out_notrans;
|
|
|
if (!__xprt_lock_write(xprt, task))
|
if (!__xprt_lock_write(xprt, task))
|
goto out_notrans;
|
goto out_notrans;
|
|
|
if (!xprt_connected(xprt)) {
|
if (!xprt_connected(xprt)) {
|
task->tk_status = -ENOTCONN;
|
task->tk_status = -ENOTCONN;
|
goto out_notrans;
|
goto out_notrans;
|
}
|
}
|
|
|
if (list_empty(&req->rq_list)) {
|
if (list_empty(&req->rq_list)) {
|
/* Update the softirq receive buffer */
|
/* Update the softirq receive buffer */
|
memcpy(&req->rq_private_buf, &req->rq_rcv_buf,
|
memcpy(&req->rq_private_buf, &req->rq_rcv_buf,
|
sizeof(req->rq_private_buf));
|
sizeof(req->rq_private_buf));
|
list_add_tail(&req->rq_list, &xprt->recv);
|
list_add_tail(&req->rq_list, &xprt->recv);
|
}
|
}
|
spin_unlock_bh(&xprt->sock_lock);
|
spin_unlock_bh(&xprt->sock_lock);
|
|
|
do_xprt_transmit(task);
|
do_xprt_transmit(task);
|
return;
|
return;
|
out_notrans:
|
out_notrans:
|
spin_unlock_bh(&xprt->sock_lock);
|
spin_unlock_bh(&xprt->sock_lock);
|
}
|
}
|
|
|
static void
|
static void
|
do_xprt_transmit(struct rpc_task *task)
|
do_xprt_transmit(struct rpc_task *task)
|
{
|
{
|
struct rpc_clnt *clnt = task->tk_client;
|
struct rpc_clnt *clnt = task->tk_client;
|
struct rpc_rqst *req = task->tk_rqstp;
|
struct rpc_rqst *req = task->tk_rqstp;
|
struct rpc_xprt *xprt = req->rq_xprt;
|
struct rpc_xprt *xprt = req->rq_xprt;
|
int status, retry = 0;
|
int status, retry = 0;
|
|
|
|
|
/* Continue transmitting the packet/record. We must be careful
|
/* Continue transmitting the packet/record. We must be careful
|
* to cope with writespace callbacks arriving _after_ we have
|
* to cope with writespace callbacks arriving _after_ we have
|
* called xprt_sendmsg().
|
* called xprt_sendmsg().
|
*/
|
*/
|
while (1) {
|
while (1) {
|
req->rq_xtime = jiffies;
|
req->rq_xtime = jiffies;
|
status = xprt_sendmsg(xprt, req);
|
status = xprt_sendmsg(xprt, req);
|
|
|
if (status < 0)
|
if (status < 0)
|
break;
|
break;
|
|
|
if (xprt->stream) {
|
if (xprt->stream) {
|
req->rq_bytes_sent += status;
|
req->rq_bytes_sent += status;
|
|
|
/* If we've sent the entire packet, immediately
|
/* If we've sent the entire packet, immediately
|
* reset the count of bytes sent. */
|
* reset the count of bytes sent. */
|
if (req->rq_bytes_sent >= req->rq_slen) {
|
if (req->rq_bytes_sent >= req->rq_slen) {
|
req->rq_bytes_sent = 0;
|
req->rq_bytes_sent = 0;
|
goto out_receive;
|
goto out_receive;
|
}
|
}
|
} else {
|
} else {
|
if (status >= req->rq_slen)
|
if (status >= req->rq_slen)
|
goto out_receive;
|
goto out_receive;
|
status = -EAGAIN;
|
status = -EAGAIN;
|
break;
|
break;
|
}
|
}
|
|
|
dprintk("RPC: %4d xmit incomplete (%d left of %d)\n",
|
dprintk("RPC: %4d xmit incomplete (%d left of %d)\n",
|
task->tk_pid, req->rq_slen - req->rq_bytes_sent,
|
task->tk_pid, req->rq_slen - req->rq_bytes_sent,
|
req->rq_slen);
|
req->rq_slen);
|
|
|
status = -EAGAIN;
|
status = -EAGAIN;
|
if (retry++ > 50)
|
if (retry++ > 50)
|
break;
|
break;
|
}
|
}
|
|
|
/* If we're doing a resend and have received a reply already,
|
/* If we're doing a resend and have received a reply already,
|
* then exit early.
|
* then exit early.
|
* Note, though, that we can't do this if we've already started
|
* Note, though, that we can't do this if we've already started
|
* resending down a TCP stream.
|
* resending down a TCP stream.
|
*/
|
*/
|
task->tk_status = status;
|
task->tk_status = status;
|
|
|
switch (status) {
|
switch (status) {
|
case -EAGAIN:
|
case -EAGAIN:
|
if (test_bit(SOCK_ASYNC_NOSPACE, &xprt->sock->flags)) {
|
if (test_bit(SOCK_ASYNC_NOSPACE, &xprt->sock->flags)) {
|
/* Protect against races with xprt_write_space */
|
/* Protect against races with xprt_write_space */
|
spin_lock_bh(&xprt->sock_lock);
|
spin_lock_bh(&xprt->sock_lock);
|
/* Don't race with disconnect */
|
/* Don't race with disconnect */
|
if (!xprt_connected(xprt))
|
if (!xprt_connected(xprt))
|
task->tk_status = -ENOTCONN;
|
task->tk_status = -ENOTCONN;
|
else if (test_bit(SOCK_NOSPACE, &xprt->sock->flags)) {
|
else if (test_bit(SOCK_NOSPACE, &xprt->sock->flags)) {
|
task->tk_timeout = req->rq_timeout.to_current;
|
task->tk_timeout = req->rq_timeout.to_current;
|
rpc_sleep_on(&xprt->pending, task, NULL, NULL);
|
rpc_sleep_on(&xprt->pending, task, NULL, NULL);
|
}
|
}
|
spin_unlock_bh(&xprt->sock_lock);
|
spin_unlock_bh(&xprt->sock_lock);
|
return;
|
return;
|
}
|
}
|
/* Keep holding the socket if it is blocked */
|
/* Keep holding the socket if it is blocked */
|
rpc_delay(task, HZ>>4);
|
rpc_delay(task, HZ>>4);
|
return;
|
return;
|
case -ECONNREFUSED:
|
case -ECONNREFUSED:
|
task->tk_timeout = RPC_REESTABLISH_TIMEOUT;
|
task->tk_timeout = RPC_REESTABLISH_TIMEOUT;
|
rpc_sleep_on(&xprt->sending, task, NULL, NULL);
|
rpc_sleep_on(&xprt->sending, task, NULL, NULL);
|
case -ENOTCONN:
|
case -ENOTCONN:
|
return;
|
return;
|
default:
|
default:
|
if (xprt->stream)
|
if (xprt->stream)
|
xprt_disconnect(xprt);
|
xprt_disconnect(xprt);
|
}
|
}
|
xprt_release_write(xprt, task);
|
xprt_release_write(xprt, task);
|
return;
|
return;
|
out_receive:
|
out_receive:
|
dprintk("RPC: %4d xmit complete\n", task->tk_pid);
|
dprintk("RPC: %4d xmit complete\n", task->tk_pid);
|
spin_lock_bh(&xprt->sock_lock);
|
spin_lock_bh(&xprt->sock_lock);
|
/* Set the task's receive timeout value */
|
/* Set the task's receive timeout value */
|
if (!xprt->nocong) {
|
if (!xprt->nocong) {
|
int timer = rpcproc_timer(clnt, task->tk_msg.rpc_proc);
|
int timer = rpcproc_timer(clnt, task->tk_msg.rpc_proc);
|
task->tk_timeout = rpc_calc_rto(&clnt->cl_rtt, timer);
|
task->tk_timeout = rpc_calc_rto(&clnt->cl_rtt, timer);
|
task->tk_timeout <<= rpc_ntimeo(&clnt->cl_rtt, timer);
|
task->tk_timeout <<= rpc_ntimeo(&clnt->cl_rtt, timer);
|
task->tk_timeout <<= clnt->cl_timeout.to_retries
|
task->tk_timeout <<= clnt->cl_timeout.to_retries
|
- req->rq_timeout.to_retries;
|
- req->rq_timeout.to_retries;
|
if (task->tk_timeout > req->rq_timeout.to_maxval)
|
if (task->tk_timeout > req->rq_timeout.to_maxval)
|
task->tk_timeout = req->rq_timeout.to_maxval;
|
task->tk_timeout = req->rq_timeout.to_maxval;
|
} else
|
} else
|
task->tk_timeout = req->rq_timeout.to_current;
|
task->tk_timeout = req->rq_timeout.to_current;
|
/* Don't race with disconnect */
|
/* Don't race with disconnect */
|
if (!xprt_connected(xprt))
|
if (!xprt_connected(xprt))
|
task->tk_status = -ENOTCONN;
|
task->tk_status = -ENOTCONN;
|
else if (!req->rq_received)
|
else if (!req->rq_received)
|
rpc_sleep_on(&xprt->pending, task, NULL, xprt_timer);
|
rpc_sleep_on(&xprt->pending, task, NULL, xprt_timer);
|
__xprt_release_write(xprt, task);
|
__xprt_release_write(xprt, task);
|
spin_unlock_bh(&xprt->sock_lock);
|
spin_unlock_bh(&xprt->sock_lock);
|
}
|
}
|
|
|
/*
|
/*
|
* Reserve an RPC call slot.
|
* Reserve an RPC call slot.
|
*/
|
*/
|
void
|
void
|
xprt_reserve(struct rpc_task *task)
|
xprt_reserve(struct rpc_task *task)
|
{
|
{
|
struct rpc_xprt *xprt = task->tk_xprt;
|
struct rpc_xprt *xprt = task->tk_xprt;
|
|
|
task->tk_status = -EIO;
|
task->tk_status = -EIO;
|
if (!xprt->shutdown) {
|
if (!xprt->shutdown) {
|
spin_lock(&xprt->xprt_lock);
|
spin_lock(&xprt->xprt_lock);
|
do_xprt_reserve(task);
|
do_xprt_reserve(task);
|
spin_unlock(&xprt->xprt_lock);
|
spin_unlock(&xprt->xprt_lock);
|
}
|
}
|
}
|
}
|
|
|
static inline void
|
static inline void
|
do_xprt_reserve(struct rpc_task *task)
|
do_xprt_reserve(struct rpc_task *task)
|
{
|
{
|
struct rpc_xprt *xprt = task->tk_xprt;
|
struct rpc_xprt *xprt = task->tk_xprt;
|
|
|
task->tk_status = 0;
|
task->tk_status = 0;
|
if (task->tk_rqstp)
|
if (task->tk_rqstp)
|
return;
|
return;
|
if (xprt->free) {
|
if (xprt->free) {
|
struct rpc_rqst *req = xprt->free;
|
struct rpc_rqst *req = xprt->free;
|
xprt->free = req->rq_next;
|
xprt->free = req->rq_next;
|
req->rq_next = NULL;
|
req->rq_next = NULL;
|
task->tk_rqstp = req;
|
task->tk_rqstp = req;
|
xprt_request_init(task, xprt);
|
xprt_request_init(task, xprt);
|
return;
|
return;
|
}
|
}
|
dprintk("RPC: waiting for request slot\n");
|
dprintk("RPC: waiting for request slot\n");
|
task->tk_status = -EAGAIN;
|
task->tk_status = -EAGAIN;
|
task->tk_timeout = 0;
|
task->tk_timeout = 0;
|
rpc_sleep_on(&xprt->backlog, task, NULL, NULL);
|
rpc_sleep_on(&xprt->backlog, task, NULL, NULL);
|
}
|
}
|
|
|
/*
|
/*
|
* Allocate a 'unique' XID
|
* Allocate a 'unique' XID
|
*/
|
*/
|
static u32
|
static u32
|
xprt_alloc_xid(void)
|
xprt_alloc_xid(void)
|
{
|
{
|
static spinlock_t xid_lock = SPIN_LOCK_UNLOCKED;
|
static spinlock_t xid_lock = SPIN_LOCK_UNLOCKED;
|
static int need_init = 1;
|
static int need_init = 1;
|
static u32 xid;
|
static u32 xid;
|
u32 ret;
|
u32 ret;
|
|
|
spin_lock(&xid_lock);
|
spin_lock(&xid_lock);
|
if (unlikely(need_init)) {
|
if (unlikely(need_init)) {
|
xid = CURRENT_TIME << 12;
|
xid = CURRENT_TIME << 12;
|
need_init = 0;
|
need_init = 0;
|
}
|
}
|
ret = xid++;
|
ret = xid++;
|
spin_unlock(&xid_lock);
|
spin_unlock(&xid_lock);
|
return ret;
|
return ret;
|
}
|
}
|
|
|
/*
|
/*
|
* Initialize RPC request
|
* Initialize RPC request
|
*/
|
*/
|
static void
|
static void
|
xprt_request_init(struct rpc_task *task, struct rpc_xprt *xprt)
|
xprt_request_init(struct rpc_task *task, struct rpc_xprt *xprt)
|
{
|
{
|
struct rpc_rqst *req = task->tk_rqstp;
|
struct rpc_rqst *req = task->tk_rqstp;
|
|
|
req->rq_timeout = xprt->timeout;
|
req->rq_timeout = xprt->timeout;
|
req->rq_task = task;
|
req->rq_task = task;
|
req->rq_xprt = xprt;
|
req->rq_xprt = xprt;
|
req->rq_xid = xprt_alloc_xid();
|
req->rq_xid = xprt_alloc_xid();
|
INIT_LIST_HEAD(&req->rq_list);
|
INIT_LIST_HEAD(&req->rq_list);
|
dprintk("RPC: %4d reserved req %p xid %08x\n", task->tk_pid,
|
dprintk("RPC: %4d reserved req %p xid %08x\n", task->tk_pid,
|
req, req->rq_xid);
|
req, req->rq_xid);
|
}
|
}
|
|
|
/*
|
/*
|
* Release an RPC call slot
|
* Release an RPC call slot
|
*/
|
*/
|
void
|
void
|
xprt_release(struct rpc_task *task)
|
xprt_release(struct rpc_task *task)
|
{
|
{
|
struct rpc_xprt *xprt = task->tk_xprt;
|
struct rpc_xprt *xprt = task->tk_xprt;
|
struct rpc_rqst *req;
|
struct rpc_rqst *req;
|
|
|
if (!(req = task->tk_rqstp))
|
if (!(req = task->tk_rqstp))
|
return;
|
return;
|
spin_lock_bh(&xprt->sock_lock);
|
spin_lock_bh(&xprt->sock_lock);
|
__xprt_release_write(xprt, task);
|
__xprt_release_write(xprt, task);
|
__xprt_put_cong(xprt, req);
|
__xprt_put_cong(xprt, req);
|
if (!list_empty(&req->rq_list))
|
if (!list_empty(&req->rq_list))
|
list_del(&req->rq_list);
|
list_del(&req->rq_list);
|
spin_unlock_bh(&xprt->sock_lock);
|
spin_unlock_bh(&xprt->sock_lock);
|
task->tk_rqstp = NULL;
|
task->tk_rqstp = NULL;
|
memset(req, 0, sizeof(*req)); /* mark unused */
|
memset(req, 0, sizeof(*req)); /* mark unused */
|
|
|
dprintk("RPC: %4d release request %p\n", task->tk_pid, req);
|
dprintk("RPC: %4d release request %p\n", task->tk_pid, req);
|
|
|
spin_lock(&xprt->xprt_lock);
|
spin_lock(&xprt->xprt_lock);
|
req->rq_next = xprt->free;
|
req->rq_next = xprt->free;
|
xprt->free = req;
|
xprt->free = req;
|
|
|
xprt_clear_backlog(xprt);
|
xprt_clear_backlog(xprt);
|
spin_unlock(&xprt->xprt_lock);
|
spin_unlock(&xprt->xprt_lock);
|
}
|
}
|
|
|
/*
|
/*
|
* Set default timeout parameters
|
* Set default timeout parameters
|
*/
|
*/
|
void
|
void
|
xprt_default_timeout(struct rpc_timeout *to, int proto)
|
xprt_default_timeout(struct rpc_timeout *to, int proto)
|
{
|
{
|
if (proto == IPPROTO_UDP)
|
if (proto == IPPROTO_UDP)
|
xprt_set_timeout(to, 5, 5 * HZ);
|
xprt_set_timeout(to, 5, 5 * HZ);
|
else
|
else
|
xprt_set_timeout(to, 5, 60 * HZ);
|
xprt_set_timeout(to, 5, 60 * HZ);
|
}
|
}
|
|
|
/*
|
/*
|
* Set constant timeout
|
* Set constant timeout
|
*/
|
*/
|
void
|
void
|
xprt_set_timeout(struct rpc_timeout *to, unsigned int retr, unsigned long incr)
|
xprt_set_timeout(struct rpc_timeout *to, unsigned int retr, unsigned long incr)
|
{
|
{
|
to->to_current =
|
to->to_current =
|
to->to_initval =
|
to->to_initval =
|
to->to_increment = incr;
|
to->to_increment = incr;
|
to->to_maxval = incr * retr;
|
to->to_maxval = incr * retr;
|
to->to_retries = retr;
|
to->to_retries = retr;
|
to->to_exponential = 0;
|
to->to_exponential = 0;
|
}
|
}
|
|
|
/*
|
/*
|
* Initialize an RPC client
|
* Initialize an RPC client
|
*/
|
*/
|
static struct rpc_xprt *
|
static struct rpc_xprt *
|
xprt_setup(int proto, struct sockaddr_in *ap, struct rpc_timeout *to)
|
xprt_setup(int proto, struct sockaddr_in *ap, struct rpc_timeout *to)
|
{
|
{
|
struct rpc_xprt *xprt;
|
struct rpc_xprt *xprt;
|
struct rpc_rqst *req;
|
struct rpc_rqst *req;
|
int i;
|
int i;
|
|
|
dprintk("RPC: setting up %s transport...\n",
|
dprintk("RPC: setting up %s transport...\n",
|
proto == IPPROTO_UDP? "UDP" : "TCP");
|
proto == IPPROTO_UDP? "UDP" : "TCP");
|
|
|
if ((xprt = kmalloc(sizeof(struct rpc_xprt), GFP_KERNEL)) == NULL)
|
if ((xprt = kmalloc(sizeof(struct rpc_xprt), GFP_KERNEL)) == NULL)
|
return NULL;
|
return NULL;
|
memset(xprt, 0, sizeof(*xprt)); /* Nnnngh! */
|
memset(xprt, 0, sizeof(*xprt)); /* Nnnngh! */
|
|
|
xprt->addr = *ap;
|
xprt->addr = *ap;
|
xprt->prot = proto;
|
xprt->prot = proto;
|
xprt->stream = (proto == IPPROTO_TCP)? 1 : 0;
|
xprt->stream = (proto == IPPROTO_TCP)? 1 : 0;
|
if (xprt->stream) {
|
if (xprt->stream) {
|
xprt->cwnd = RPC_MAXCWND;
|
xprt->cwnd = RPC_MAXCWND;
|
xprt->nocong = 1;
|
xprt->nocong = 1;
|
} else
|
} else
|
xprt->cwnd = RPC_INITCWND;
|
xprt->cwnd = RPC_INITCWND;
|
spin_lock_init(&xprt->sock_lock);
|
spin_lock_init(&xprt->sock_lock);
|
spin_lock_init(&xprt->xprt_lock);
|
spin_lock_init(&xprt->xprt_lock);
|
init_waitqueue_head(&xprt->cong_wait);
|
init_waitqueue_head(&xprt->cong_wait);
|
|
|
INIT_LIST_HEAD(&xprt->recv);
|
INIT_LIST_HEAD(&xprt->recv);
|
|
|
/* Set timeout parameters */
|
/* Set timeout parameters */
|
if (to) {
|
if (to) {
|
xprt->timeout = *to;
|
xprt->timeout = *to;
|
xprt->timeout.to_current = to->to_initval;
|
xprt->timeout.to_current = to->to_initval;
|
} else
|
} else
|
xprt_default_timeout(&xprt->timeout, xprt->prot);
|
xprt_default_timeout(&xprt->timeout, xprt->prot);
|
|
|
INIT_RPC_WAITQ(&xprt->pending, "xprt_pending");
|
INIT_RPC_WAITQ(&xprt->pending, "xprt_pending");
|
INIT_RPC_WAITQ(&xprt->sending, "xprt_sending");
|
INIT_RPC_WAITQ(&xprt->sending, "xprt_sending");
|
INIT_RPC_WAITQ(&xprt->resend, "xprt_resend");
|
INIT_RPC_WAITQ(&xprt->resend, "xprt_resend");
|
INIT_RPC_WAITQ(&xprt->backlog, "xprt_backlog");
|
INIT_RPC_WAITQ(&xprt->backlog, "xprt_backlog");
|
|
|
/* initialize free list */
|
/* initialize free list */
|
for (i = 0, req = xprt->slot; i < RPC_MAXREQS-1; i++, req++)
|
for (i = 0, req = xprt->slot; i < RPC_MAXREQS-1; i++, req++)
|
req->rq_next = req + 1;
|
req->rq_next = req + 1;
|
req->rq_next = NULL;
|
req->rq_next = NULL;
|
xprt->free = xprt->slot;
|
xprt->free = xprt->slot;
|
|
|
/* Check whether we want to use a reserved port */
|
/* Check whether we want to use a reserved port */
|
xprt->resvport = capable(CAP_NET_BIND_SERVICE) ? 1 : 0;
|
xprt->resvport = capable(CAP_NET_BIND_SERVICE) ? 1 : 0;
|
|
|
dprintk("RPC: created transport %p\n", xprt);
|
dprintk("RPC: created transport %p\n", xprt);
|
|
|
return xprt;
|
return xprt;
|
}
|
}
|
|
|
/*
|
/*
|
* Bind to a reserved port
|
* Bind to a reserved port
|
*/
|
*/
|
static inline int
|
static inline int
|
xprt_bindresvport(struct socket *sock)
|
xprt_bindresvport(struct socket *sock)
|
{
|
{
|
struct sockaddr_in myaddr;
|
struct sockaddr_in myaddr;
|
int err, port;
|
int err, port;
|
kernel_cap_t saved_cap = current->cap_effective;
|
kernel_cap_t saved_cap = current->cap_effective;
|
|
|
/* Override capabilities.
|
/* Override capabilities.
|
* They were checked in xprt_create_proto i.e. at mount time
|
* They were checked in xprt_create_proto i.e. at mount time
|
*/
|
*/
|
cap_raise (current->cap_effective, CAP_NET_BIND_SERVICE);
|
cap_raise (current->cap_effective, CAP_NET_BIND_SERVICE);
|
|
|
memset(&myaddr, 0, sizeof(myaddr));
|
memset(&myaddr, 0, sizeof(myaddr));
|
myaddr.sin_family = AF_INET;
|
myaddr.sin_family = AF_INET;
|
port = 800;
|
port = 800;
|
do {
|
do {
|
myaddr.sin_port = htons(port);
|
myaddr.sin_port = htons(port);
|
err = sock->ops->bind(sock, (struct sockaddr *) &myaddr,
|
err = sock->ops->bind(sock, (struct sockaddr *) &myaddr,
|
sizeof(myaddr));
|
sizeof(myaddr));
|
} while (err == -EADDRINUSE && --port > 0);
|
} while (err == -EADDRINUSE && --port > 0);
|
current->cap_effective = saved_cap;
|
current->cap_effective = saved_cap;
|
|
|
if (err < 0)
|
if (err < 0)
|
printk("RPC: Can't bind to reserved port (%d).\n", -err);
|
printk("RPC: Can't bind to reserved port (%d).\n", -err);
|
|
|
return err;
|
return err;
|
}
|
}
|
|
|
static int
|
static int
|
xprt_bind_socket(struct rpc_xprt *xprt, struct socket *sock)
|
xprt_bind_socket(struct rpc_xprt *xprt, struct socket *sock)
|
{
|
{
|
struct sock *sk = sock->sk;
|
struct sock *sk = sock->sk;
|
|
|
if (xprt->inet)
|
if (xprt->inet)
|
return -EBUSY;
|
return -EBUSY;
|
|
|
write_lock_bh(&sk->callback_lock);
|
write_lock_bh(&sk->callback_lock);
|
sk->user_data = xprt;
|
sk->user_data = xprt;
|
xprt->old_data_ready = sk->data_ready;
|
xprt->old_data_ready = sk->data_ready;
|
xprt->old_state_change = sk->state_change;
|
xprt->old_state_change = sk->state_change;
|
xprt->old_write_space = sk->write_space;
|
xprt->old_write_space = sk->write_space;
|
if (xprt->prot == IPPROTO_UDP) {
|
if (xprt->prot == IPPROTO_UDP) {
|
sk->data_ready = udp_data_ready;
|
sk->data_ready = udp_data_ready;
|
sk->no_check = UDP_CSUM_NORCV;
|
sk->no_check = UDP_CSUM_NORCV;
|
xprt_set_connected(xprt);
|
xprt_set_connected(xprt);
|
} else {
|
} else {
|
struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);
|
struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);
|
tp->nonagle = 1; /* disable Nagle's algorithm */
|
tp->nonagle = 1; /* disable Nagle's algorithm */
|
sk->data_ready = tcp_data_ready;
|
sk->data_ready = tcp_data_ready;
|
sk->state_change = tcp_state_change;
|
sk->state_change = tcp_state_change;
|
xprt_clear_connected(xprt);
|
xprt_clear_connected(xprt);
|
}
|
}
|
sk->write_space = xprt_write_space;
|
sk->write_space = xprt_write_space;
|
|
|
/* Reset to new socket */
|
/* Reset to new socket */
|
xprt->sock = sock;
|
xprt->sock = sock;
|
xprt->inet = sk;
|
xprt->inet = sk;
|
write_unlock_bh(&sk->callback_lock);
|
write_unlock_bh(&sk->callback_lock);
|
|
|
return 0;
|
return 0;
|
}
|
}
|
|
|
/*
|
/*
|
* Set socket buffer length
|
* Set socket buffer length
|
*/
|
*/
|
void
|
void
|
xprt_sock_setbufsize(struct rpc_xprt *xprt)
|
xprt_sock_setbufsize(struct rpc_xprt *xprt)
|
{
|
{
|
struct sock *sk = xprt->inet;
|
struct sock *sk = xprt->inet;
|
|
|
if (xprt->stream)
|
if (xprt->stream)
|
return;
|
return;
|
if (xprt->rcvsize) {
|
if (xprt->rcvsize) {
|
sk->userlocks |= SOCK_RCVBUF_LOCK;
|
sk->userlocks |= SOCK_RCVBUF_LOCK;
|
sk->rcvbuf = xprt->rcvsize * RPC_MAXCONG * 2;
|
sk->rcvbuf = xprt->rcvsize * RPC_MAXCONG * 2;
|
}
|
}
|
if (xprt->sndsize) {
|
if (xprt->sndsize) {
|
sk->userlocks |= SOCK_SNDBUF_LOCK;
|
sk->userlocks |= SOCK_SNDBUF_LOCK;
|
sk->sndbuf = xprt->sndsize * RPC_MAXCONG * 2;
|
sk->sndbuf = xprt->sndsize * RPC_MAXCONG * 2;
|
sk->write_space(sk);
|
sk->write_space(sk);
|
}
|
}
|
}
|
}
|
|
|
/*
|
/*
|
* Create a client socket given the protocol and peer address.
|
* Create a client socket given the protocol and peer address.
|
*/
|
*/
|
static struct socket *
|
static struct socket *
|
xprt_create_socket(int proto, struct rpc_timeout *to, int resvport)
|
xprt_create_socket(int proto, struct rpc_timeout *to, int resvport)
|
{
|
{
|
struct socket *sock;
|
struct socket *sock;
|
int type, err;
|
int type, err;
|
|
|
dprintk("RPC: xprt_create_socket(%s %d)\n",
|
dprintk("RPC: xprt_create_socket(%s %d)\n",
|
(proto == IPPROTO_UDP)? "udp" : "tcp", proto);
|
(proto == IPPROTO_UDP)? "udp" : "tcp", proto);
|
|
|
type = (proto == IPPROTO_UDP)? SOCK_DGRAM : SOCK_STREAM;
|
type = (proto == IPPROTO_UDP)? SOCK_DGRAM : SOCK_STREAM;
|
|
|
if ((err = sock_create(PF_INET, type, proto, &sock)) < 0) {
|
if ((err = sock_create(PF_INET, type, proto, &sock)) < 0) {
|
printk("RPC: can't create socket (%d).\n", -err);
|
printk("RPC: can't create socket (%d).\n", -err);
|
goto failed;
|
goto failed;
|
}
|
}
|
|
|
/* bind to a reserved port */
|
/* bind to a reserved port */
|
if (resvport && xprt_bindresvport(sock) < 0)
|
if (resvport && xprt_bindresvport(sock) < 0)
|
goto failed;
|
goto failed;
|
|
|
return sock;
|
return sock;
|
|
|
failed:
|
failed:
|
sock_release(sock);
|
sock_release(sock);
|
return NULL;
|
return NULL;
|
}
|
}
|
|
|
/*
|
/*
|
* Create an RPC client transport given the protocol and peer address.
|
* Create an RPC client transport given the protocol and peer address.
|
*/
|
*/
|
struct rpc_xprt *
|
struct rpc_xprt *
|
xprt_create_proto(int proto, struct sockaddr_in *sap, struct rpc_timeout *to)
|
xprt_create_proto(int proto, struct sockaddr_in *sap, struct rpc_timeout *to)
|
{
|
{
|
struct rpc_xprt *xprt;
|
struct rpc_xprt *xprt;
|
|
|
xprt = xprt_setup(proto, sap, to);
|
xprt = xprt_setup(proto, sap, to);
|
if (!xprt)
|
if (!xprt)
|
goto out_bad;
|
goto out_bad;
|
|
|
dprintk("RPC: xprt_create_proto created xprt %p\n", xprt);
|
dprintk("RPC: xprt_create_proto created xprt %p\n", xprt);
|
return xprt;
|
return xprt;
|
out_bad:
|
out_bad:
|
dprintk("RPC: xprt_create_proto failed\n");
|
dprintk("RPC: xprt_create_proto failed\n");
|
if (xprt)
|
if (xprt)
|
kfree(xprt);
|
kfree(xprt);
|
return NULL;
|
return NULL;
|
}
|
}
|
|
|
/*
|
/*
|
* Prepare for transport shutdown.
|
* Prepare for transport shutdown.
|
*/
|
*/
|
void
|
void
|
xprt_shutdown(struct rpc_xprt *xprt)
|
xprt_shutdown(struct rpc_xprt *xprt)
|
{
|
{
|
xprt->shutdown = 1;
|
xprt->shutdown = 1;
|
rpc_wake_up(&xprt->sending);
|
rpc_wake_up(&xprt->sending);
|
rpc_wake_up(&xprt->resend);
|
rpc_wake_up(&xprt->resend);
|
rpc_wake_up(&xprt->pending);
|
rpc_wake_up(&xprt->pending);
|
rpc_wake_up(&xprt->backlog);
|
rpc_wake_up(&xprt->backlog);
|
if (waitqueue_active(&xprt->cong_wait))
|
if (waitqueue_active(&xprt->cong_wait))
|
wake_up(&xprt->cong_wait);
|
wake_up(&xprt->cong_wait);
|
}
|
}
|
|
|
/*
|
/*
|
* Clear the xprt backlog queue
|
* Clear the xprt backlog queue
|
*/
|
*/
|
int
|
int
|
xprt_clear_backlog(struct rpc_xprt *xprt) {
|
xprt_clear_backlog(struct rpc_xprt *xprt) {
|
rpc_wake_up_next(&xprt->backlog);
|
rpc_wake_up_next(&xprt->backlog);
|
if (waitqueue_active(&xprt->cong_wait))
|
if (waitqueue_active(&xprt->cong_wait))
|
wake_up(&xprt->cong_wait);
|
wake_up(&xprt->cong_wait);
|
return 1;
|
return 1;
|
}
|
}
|
|
|
/*
|
/*
|
* Destroy an RPC transport, killing off all requests.
|
* Destroy an RPC transport, killing off all requests.
|
*/
|
*/
|
int
|
int
|
xprt_destroy(struct rpc_xprt *xprt)
|
xprt_destroy(struct rpc_xprt *xprt)
|
{
|
{
|
dprintk("RPC: destroying transport %p\n", xprt);
|
dprintk("RPC: destroying transport %p\n", xprt);
|
xprt_shutdown(xprt);
|
xprt_shutdown(xprt);
|
xprt_close(xprt);
|
xprt_close(xprt);
|
kfree(xprt);
|
kfree(xprt);
|
|
|
return 0;
|
return 0;
|
}
|
}
|
|
|
