aboutsummaryrefslogtreecommitdiffstats
path: root/net
diff options
context:
space:
mode:
authorChuck Lever <cel@citi.umich.edu>2005-08-11 16:25:23 -0400
committerTrond Myklebust <Trond.Myklebust@netapp.com>2005-09-23 12:38:12 -0400
commita246b0105bbd9a70a698f69baae2042996f2a0e9 (patch)
tree6c8831d8579a7fdc5201d3e9c20270cb1420eeda /net
parent094bb20b9fcab3a1652a77741caba6b78097d622 (diff)
[PATCH] RPC: introduce client-side transport switch
Move the bulk of client-side socket-specific code into a separate source file, net/sunrpc/xprtsock.c. Test-plan: Millions of fsx operations. Performance characterization such as "sio" or "iozone". Destructive testing (unplugging the network temporarily, server reboots). Connectathon with v2, v3, and v4. Version: Thu, 11 Aug 2005 16:03:38 -0400 Signed-off-by: Chuck Lever <cel@netapp.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Diffstat (limited to 'net')
-rw-r--r--net/sunrpc/Makefile2
-rw-r--r--net/sunrpc/clnt.c3
-rw-r--r--net/sunrpc/sysctl.c3
-rw-r--r--net/sunrpc/xdr.c102
-rw-r--r--net/sunrpc/xprt.c916
-rw-r--r--net/sunrpc/xprtsock.c1021
6 files changed, 1070 insertions, 977 deletions
diff --git a/net/sunrpc/Makefile b/net/sunrpc/Makefile
index f0a955627177..cdcab9ca4c60 100644
--- a/net/sunrpc/Makefile
+++ b/net/sunrpc/Makefile
@@ -6,7 +6,7 @@
6obj-$(CONFIG_SUNRPC) += sunrpc.o 6obj-$(CONFIG_SUNRPC) += sunrpc.o
7obj-$(CONFIG_SUNRPC_GSS) += auth_gss/ 7obj-$(CONFIG_SUNRPC_GSS) += auth_gss/
8 8
9sunrpc-y := clnt.o xprt.o socklib.o sched.o \ 9sunrpc-y := clnt.o xprt.o socklib.o xprtsock.o sched.o \
10 auth.o auth_null.o auth_unix.o \ 10 auth.o auth_null.o auth_unix.o \
11 svc.o svcsock.o svcauth.o svcauth_unix.o \ 11 svc.o svcsock.o svcauth.o svcauth_unix.o \
12 pmap_clnt.o timer.o xdr.o \ 12 pmap_clnt.o timer.o xdr.o \
diff --git a/net/sunrpc/clnt.c b/net/sunrpc/clnt.c
index 2d3cf0a52d82..ab50c3c9e6a8 100644
--- a/net/sunrpc/clnt.c
+++ b/net/sunrpc/clnt.c
@@ -525,8 +525,7 @@ rpc_setbufsize(struct rpc_clnt *clnt, unsigned int sndsize, unsigned int rcvsize
525 xprt->rcvsize = 0; 525 xprt->rcvsize = 0;
526 if (rcvsize) 526 if (rcvsize)
527 xprt->rcvsize = rcvsize + RPC_SLACK_SPACE; 527 xprt->rcvsize = rcvsize + RPC_SLACK_SPACE;
528 if (xprt_connected(xprt)) 528 xprt->ops->set_buffer_size(xprt);
529 xprt_sock_setbufsize(xprt);
530} 529}
531 530
532/* 531/*
diff --git a/net/sunrpc/sysctl.c b/net/sunrpc/sysctl.c
index 1b9616a12e24..ef483262f17f 100644
--- a/net/sunrpc/sysctl.c
+++ b/net/sunrpc/sysctl.c
@@ -119,6 +119,9 @@ done:
119 return 0; 119 return 0;
120} 120}
121 121
122unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE;
123unsigned int xprt_tcp_slot_table_entries = RPC_DEF_SLOT_TABLE;
124
122static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE; 125static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE;
123static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE; 126static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE;
124 127
diff --git a/net/sunrpc/xdr.c b/net/sunrpc/xdr.c
index 9cc12aeed22c..32df43372ee9 100644
--- a/net/sunrpc/xdr.c
+++ b/net/sunrpc/xdr.c
@@ -6,15 +6,12 @@
6 * Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de> 6 * Copyright (C) 1995, 1996 Olaf Kirch <okir@monad.swb.de>
7 */ 7 */
8 8
9#include <linux/module.h>
9#include <linux/types.h> 10#include <linux/types.h>
10#include <linux/socket.h>
11#include <linux/string.h> 11#include <linux/string.h>
12#include <linux/kernel.h> 12#include <linux/kernel.h>
13#include <linux/pagemap.h> 13#include <linux/pagemap.h>
14#include <linux/errno.h> 14#include <linux/errno.h>
15#include <linux/in.h>
16#include <linux/net.h>
17#include <net/sock.h>
18#include <linux/sunrpc/xdr.h> 15#include <linux/sunrpc/xdr.h>
19#include <linux/sunrpc/msg_prot.h> 16#include <linux/sunrpc/msg_prot.h>
20 17
@@ -177,103 +174,6 @@ xdr_inline_pages(struct xdr_buf *xdr, unsigned int offset,
177} 174}
178 175
179 176
180int
181xdr_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen,
182 struct xdr_buf *xdr, unsigned int base, int msgflags)
183{
184 struct page **ppage = xdr->pages;
185 unsigned int len, pglen = xdr->page_len;
186 int err, ret = 0;
187 ssize_t (*sendpage)(struct socket *, struct page *, int, size_t, int);
188
189 len = xdr->head[0].iov_len;
190 if (base < len || (addr != NULL && base == 0)) {
191 struct kvec iov = {
192 .iov_base = xdr->head[0].iov_base + base,
193 .iov_len = len - base,
194 };
195 struct msghdr msg = {
196 .msg_name = addr,
197 .msg_namelen = addrlen,
198 .msg_flags = msgflags,
199 };
200 if (xdr->len > len)
201 msg.msg_flags |= MSG_MORE;
202
203 if (iov.iov_len != 0)
204 err = kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len);
205 else
206 err = kernel_sendmsg(sock, &msg, NULL, 0, 0);
207 if (ret == 0)
208 ret = err;
209 else if (err > 0)
210 ret += err;
211 if (err != iov.iov_len)
212 goto out;
213 base = 0;
214 } else
215 base -= len;
216
217 if (pglen == 0)
218 goto copy_tail;
219 if (base >= pglen) {
220 base -= pglen;
221 goto copy_tail;
222 }
223 if (base || xdr->page_base) {
224 pglen -= base;
225 base += xdr->page_base;
226 ppage += base >> PAGE_CACHE_SHIFT;
227 base &= ~PAGE_CACHE_MASK;
228 }
229
230 sendpage = sock->ops->sendpage ? : sock_no_sendpage;
231 do {
232 int flags = msgflags;
233
234 len = PAGE_CACHE_SIZE;
235 if (base)
236 len -= base;
237 if (pglen < len)
238 len = pglen;
239
240 if (pglen != len || xdr->tail[0].iov_len != 0)
241 flags |= MSG_MORE;
242
243 /* Hmm... We might be dealing with highmem pages */
244 if (PageHighMem(*ppage))
245 sendpage = sock_no_sendpage;
246 err = sendpage(sock, *ppage, base, len, flags);
247 if (ret == 0)
248 ret = err;
249 else if (err > 0)
250 ret += err;
251 if (err != len)
252 goto out;
253 base = 0;
254 ppage++;
255 } while ((pglen -= len) != 0);
256copy_tail:
257 len = xdr->tail[0].iov_len;
258 if (base < len) {
259 struct kvec iov = {
260 .iov_base = xdr->tail[0].iov_base + base,
261 .iov_len = len - base,
262 };
263 struct msghdr msg = {
264 .msg_flags = msgflags,
265 };
266 err = kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len);
267 if (ret == 0)
268 ret = err;
269 else if (err > 0)
270 ret += err;
271 }
272out:
273 return ret;
274}
275
276
277/* 177/*
278 * Helper routines for doing 'memmove' like operations on a struct xdr_buf 178 * Helper routines for doing 'memmove' like operations on a struct xdr_buf
279 * 179 *
diff --git a/net/sunrpc/xprt.c b/net/sunrpc/xprt.c
index 67444f494fea..4342acf4d1cd 100644
--- a/net/sunrpc/xprt.c
+++ b/net/sunrpc/xprt.c
@@ -32,37 +32,16 @@
32 * tasks that rely on callbacks. 32 * tasks that rely on callbacks.
33 * 33 *
34 * Copyright (C) 1995-1997, Olaf Kirch <okir@monad.swb.de> 34 * Copyright (C) 1995-1997, Olaf Kirch <okir@monad.swb.de>
35 *
36 * TCP callback races fixes (C) 1998 Red Hat Software <alan@redhat.com>
37 * TCP send fixes (C) 1998 Red Hat Software <alan@redhat.com>
38 * TCP NFS related read + write fixes
39 * (C) 1999 Dave Airlie, University of Limerick, Ireland <airlied@linux.ie>
40 *
41 * Rewrite of larges part of the code in order to stabilize TCP stuff.
42 * Fix behaviour when socket buffer is full.
43 * (C) 1999 Trond Myklebust <trond.myklebust@fys.uio.no>
44 */ 35 */
45 36
37#include <linux/module.h>
38
46#include <linux/types.h> 39#include <linux/types.h>
47#include <linux/slab.h> 40#include <linux/interrupt.h>
48#include <linux/capability.h>
49#include <linux/sched.h>
50#include <linux/errno.h>
51#include <linux/socket.h>
52#include <linux/in.h>
53#include <linux/net.h>
54#include <linux/mm.h>
55#include <linux/udp.h>
56#include <linux/tcp.h>
57#include <linux/sunrpc/clnt.h>
58#include <linux/file.h>
59#include <linux/workqueue.h> 41#include <linux/workqueue.h>
60#include <linux/random.h> 42#include <linux/random.h>
61 43
62#include <net/sock.h> 44#include <linux/sunrpc/clnt.h>
63#include <net/checksum.h>
64#include <net/udp.h>
65#include <net/tcp.h>
66 45
67/* 46/*
68 * Local variables 47 * Local variables
@@ -74,64 +53,17 @@
74#endif 53#endif
75 54
76#define XPRT_MAX_BACKOFF (8) 55#define XPRT_MAX_BACKOFF (8)
77#define XPRT_IDLE_TIMEOUT (5*60*HZ)
78#define XPRT_MAX_RESVPORT (800)
79 56
80/* 57/*
81 * Local functions 58 * Local functions
82 */ 59 */
83static void xprt_request_init(struct rpc_task *, struct rpc_xprt *); 60static void xprt_request_init(struct rpc_task *, struct rpc_xprt *);
84static inline void do_xprt_reserve(struct rpc_task *); 61static inline void do_xprt_reserve(struct rpc_task *);
85static void xprt_disconnect(struct rpc_xprt *);
86static void xprt_connect_status(struct rpc_task *task); 62static void xprt_connect_status(struct rpc_task *task);
87static struct rpc_xprt * xprt_setup(int proto, struct sockaddr_in *ap,
88 struct rpc_timeout *to);
89static struct socket *xprt_create_socket(struct rpc_xprt *, int, int);
90static void xprt_bind_socket(struct rpc_xprt *, struct socket *);
91static int __xprt_get_cong(struct rpc_xprt *, struct rpc_task *); 63static int __xprt_get_cong(struct rpc_xprt *, struct rpc_task *);
92 64
93static int xprt_clear_backlog(struct rpc_xprt *xprt); 65static int xprt_clear_backlog(struct rpc_xprt *xprt);
94 66
95#ifdef RPC_DEBUG_DATA
96/*
97 * Print the buffer contents (first 128 bytes only--just enough for
98 * diropres return).
99 */
100static void
101xprt_pktdump(char *msg, u32 *packet, unsigned int count)
102{
103 u8 *buf = (u8 *) packet;
104 int j;
105
106 dprintk("RPC: %s\n", msg);
107 for (j = 0; j < count && j < 128; j += 4) {
108 if (!(j & 31)) {
109 if (j)
110 dprintk("\n");
111 dprintk("0x%04x ", j);
112 }
113 dprintk("%02x%02x%02x%02x ",
114 buf[j], buf[j+1], buf[j+2], buf[j+3]);
115 }
116 dprintk("\n");
117}
118#else
119static inline void
120xprt_pktdump(char *msg, u32 *packet, unsigned int count)
121{
122 /* NOP */
123}
124#endif
125
126/*
127 * Look up RPC transport given an INET socket
128 */
129static inline struct rpc_xprt *
130xprt_from_sock(struct sock *sk)
131{
132 return (struct rpc_xprt *) sk->sk_user_data;
133}
134
135/* 67/*
136 * Serialize write access to sockets, in order to prevent different 68 * Serialize write access to sockets, in order to prevent different
137 * requests from interfering with each other. 69 * requests from interfering with each other.
@@ -235,62 +167,6 @@ xprt_release_write(struct rpc_xprt *xprt, struct rpc_task *task)
235} 167}
236 168
237/* 169/*
238 * Write data to socket.
239 */
240static inline int
241xprt_sendmsg(struct rpc_xprt *xprt, struct rpc_rqst *req)
242{
243 struct socket *sock = xprt->sock;
244 struct xdr_buf *xdr = &req->rq_snd_buf;
245 struct sockaddr *addr = NULL;
246 int addrlen = 0;
247 unsigned int skip;
248 int result;
249
250 if (!sock)
251 return -ENOTCONN;
252
253 xprt_pktdump("packet data:",
254 req->rq_svec->iov_base,
255 req->rq_svec->iov_len);
256
257 /* For UDP, we need to provide an address */
258 if (!xprt->stream) {
259 addr = (struct sockaddr *) &xprt->addr;
260 addrlen = sizeof(xprt->addr);
261 }
262 /* Dont repeat bytes */
263 skip = req->rq_bytes_sent;
264
265 clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags);
266 result = xdr_sendpages(sock, addr, addrlen, xdr, skip, MSG_DONTWAIT);
267
268 dprintk("RPC: xprt_sendmsg(%d) = %d\n", xdr->len - skip, result);
269
270 if (result >= 0)
271 return result;
272
273 switch (result) {
274 case -ECONNREFUSED:
275 /* When the server has died, an ICMP port unreachable message
276 * prompts ECONNREFUSED.
277 */
278 case -EAGAIN:
279 break;
280 case -ECONNRESET:
281 case -ENOTCONN:
282 case -EPIPE:
283 /* connection broken */
284 if (xprt->stream)
285 result = -ENOTCONN;
286 break;
287 default:
288 printk(KERN_NOTICE "RPC: sendmsg returned error %d\n", -result);
289 }
290 return result;
291}
292
293/*
294 * Van Jacobson congestion avoidance. Check if the congestion window 170 * Van Jacobson congestion avoidance. Check if the congestion window
295 * overflowed. Put the task to sleep if this is the case. 171 * overflowed. Put the task to sleep if this is the case.
296 */ 172 */
@@ -405,48 +281,20 @@ int xprt_adjust_timeout(struct rpc_rqst *req)
405 return status; 281 return status;
406} 282}
407 283
408/*
409 * Close down a transport socket
410 */
411static void
412xprt_close(struct rpc_xprt *xprt)
413{
414 struct socket *sock = xprt->sock;
415 struct sock *sk = xprt->inet;
416
417 if (!sk)
418 return;
419
420 write_lock_bh(&sk->sk_callback_lock);
421 xprt->inet = NULL;
422 xprt->sock = NULL;
423
424 sk->sk_user_data = NULL;
425 sk->sk_data_ready = xprt->old_data_ready;
426 sk->sk_state_change = xprt->old_state_change;
427 sk->sk_write_space = xprt->old_write_space;
428 write_unlock_bh(&sk->sk_callback_lock);
429
430 sk->sk_no_check = 0;
431
432 sock_release(sock);
433}
434
435static void 284static void
436xprt_socket_autoclose(void *args) 285xprt_socket_autoclose(void *args)
437{ 286{
438 struct rpc_xprt *xprt = (struct rpc_xprt *)args; 287 struct rpc_xprt *xprt = (struct rpc_xprt *)args;
439 288
440 xprt_disconnect(xprt); 289 xprt_disconnect(xprt);
441 xprt_close(xprt); 290 xprt->ops->close(xprt);
442 xprt_release_write(xprt, NULL); 291 xprt_release_write(xprt, NULL);
443} 292}
444 293
445/* 294/*
446 * Mark a transport as disconnected 295 * Mark a transport as disconnected
447 */ 296 */
448static void 297void xprt_disconnect(struct rpc_xprt *xprt)
449xprt_disconnect(struct rpc_xprt *xprt)
450{ 298{
451 dprintk("RPC: disconnected transport %p\n", xprt); 299 dprintk("RPC: disconnected transport %p\n", xprt);
452 spin_lock_bh(&xprt->sock_lock); 300 spin_lock_bh(&xprt->sock_lock);
@@ -479,57 +327,6 @@ out_abort:
479 spin_unlock(&xprt->sock_lock); 327 spin_unlock(&xprt->sock_lock);
480} 328}
481 329
482static void xprt_socket_connect(void *args)
483{
484 struct rpc_xprt *xprt = (struct rpc_xprt *)args;
485 struct socket *sock = xprt->sock;
486 int status = -EIO;
487
488 if (xprt->shutdown || xprt->addr.sin_port == 0)
489 goto out;
490
491 /*
492 * Start by resetting any existing state
493 */
494 xprt_close(xprt);
495 sock = xprt_create_socket(xprt, xprt->prot, xprt->resvport);
496 if (sock == NULL) {
497 /* couldn't create socket or bind to reserved port;
498 * this is likely a permanent error, so cause an abort */
499 goto out;
500 }
501 xprt_bind_socket(xprt, sock);
502 xprt_sock_setbufsize(xprt);
503
504 status = 0;
505 if (!xprt->stream)
506 goto out;
507
508 /*
509 * Tell the socket layer to start connecting...
510 */
511 status = sock->ops->connect(sock, (struct sockaddr *) &xprt->addr,
512 sizeof(xprt->addr), O_NONBLOCK);
513 dprintk("RPC: %p connect status %d connected %d sock state %d\n",
514 xprt, -status, xprt_connected(xprt), sock->sk->sk_state);
515 if (status < 0) {
516 switch (status) {
517 case -EINPROGRESS:
518 case -EALREADY:
519 goto out_clear;
520 }
521 }
522out:
523 if (status < 0)
524 rpc_wake_up_status(&xprt->pending, status);
525 else
526 rpc_wake_up(&xprt->pending);
527out_clear:
528 smp_mb__before_clear_bit();
529 clear_bit(XPRT_CONNECTING, &xprt->sockstate);
530 smp_mb__after_clear_bit();
531}
532
533/* 330/*
534 * Attempt to connect a TCP socket. 331 * Attempt to connect a TCP socket.
535 * 332 *
@@ -552,30 +349,16 @@ void xprt_connect(struct rpc_task *task)
552 if (!xprt_lock_write(xprt, task)) 349 if (!xprt_lock_write(xprt, task))
553 return; 350 return;
554 if (xprt_connected(xprt)) 351 if (xprt_connected(xprt))
555 goto out_write; 352 xprt_release_write(xprt, task);
353 else {
354 if (task->tk_rqstp)
355 task->tk_rqstp->rq_bytes_sent = 0;
556 356
557 if (task->tk_rqstp) 357 task->tk_timeout = RPC_CONNECT_TIMEOUT;
558 task->tk_rqstp->rq_bytes_sent = 0; 358 rpc_sleep_on(&xprt->pending, task, xprt_connect_status, NULL);
559 359 xprt->ops->connect(task);
560 task->tk_timeout = RPC_CONNECT_TIMEOUT;
561 rpc_sleep_on(&xprt->pending, task, xprt_connect_status, NULL);
562 if (!test_and_set_bit(XPRT_CONNECTING, &xprt->sockstate)) {
563 /* Note: if we are here due to a dropped connection
564 * we delay reconnecting by RPC_REESTABLISH_TIMEOUT/HZ
565 * seconds
566 */
567 if (xprt->sock != NULL)
568 schedule_delayed_work(&xprt->sock_connect,
569 RPC_REESTABLISH_TIMEOUT);
570 else {
571 schedule_work(&xprt->sock_connect);
572 if (!RPC_IS_ASYNC(task))
573 flush_scheduled_work();
574 }
575 } 360 }
576 return; 361 return;
577 out_write:
578 xprt_release_write(xprt, task);
579} 362}
580 363
581/* 364/*
@@ -624,8 +407,7 @@ xprt_connect_status(struct rpc_task *task)
624/* 407/*
625 * Look up the RPC request corresponding to a reply, and then lock it. 408 * Look up the RPC request corresponding to a reply, and then lock it.
626 */ 409 */
627static inline struct rpc_rqst * 410struct rpc_rqst *xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid)
628xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid)
629{ 411{
630 struct list_head *pos; 412 struct list_head *pos;
631 struct rpc_rqst *req = NULL; 413 struct rpc_rqst *req = NULL;
@@ -644,8 +426,7 @@ xprt_lookup_rqst(struct rpc_xprt *xprt, u32 xid)
644 * Complete reply received. 426 * Complete reply received.
645 * The TCP code relies on us to remove the request from xprt->pending. 427 * The TCP code relies on us to remove the request from xprt->pending.
646 */ 428 */
647static void 429void xprt_complete_rqst(struct rpc_xprt *xprt, struct rpc_rqst *req, int copied)
648xprt_complete_rqst(struct rpc_xprt *xprt, struct rpc_rqst *req, int copied)
649{ 430{
650 struct rpc_task *task = req->rq_task; 431 struct rpc_task *task = req->rq_task;
651 struct rpc_clnt *clnt = task->tk_client; 432 struct rpc_clnt *clnt = task->tk_client;
@@ -692,409 +473,6 @@ xprt_complete_rqst(struct rpc_xprt *xprt, struct rpc_rqst *req, int copied)
692} 473}
693 474
694/* 475/*
695 * Input handler for RPC replies. Called from a bottom half and hence
696 * atomic.
697 */
698static void
699udp_data_ready(struct sock *sk, int len)
700{
701 struct rpc_task *task;
702 struct rpc_xprt *xprt;
703 struct rpc_rqst *rovr;
704 struct sk_buff *skb;
705 int err, repsize, copied;
706 u32 _xid, *xp;
707
708 read_lock(&sk->sk_callback_lock);
709 dprintk("RPC: udp_data_ready...\n");
710 if (!(xprt = xprt_from_sock(sk))) {
711 printk("RPC: udp_data_ready request not found!\n");
712 goto out;
713 }
714
715 dprintk("RPC: udp_data_ready client %p\n", xprt);
716
717 if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL)
718 goto out;
719
720 if (xprt->shutdown)
721 goto dropit;
722
723 repsize = skb->len - sizeof(struct udphdr);
724 if (repsize < 4) {
725 printk("RPC: impossible RPC reply size %d!\n", repsize);
726 goto dropit;
727 }
728
729 /* Copy the XID from the skb... */
730 xp = skb_header_pointer(skb, sizeof(struct udphdr),
731 sizeof(_xid), &_xid);
732 if (xp == NULL)
733 goto dropit;
734
735 /* Look up and lock the request corresponding to the given XID */
736 spin_lock(&xprt->sock_lock);
737 rovr = xprt_lookup_rqst(xprt, *xp);
738 if (!rovr)
739 goto out_unlock;
740 task = rovr->rq_task;
741
742 dprintk("RPC: %4d received reply\n", task->tk_pid);
743
744 if ((copied = rovr->rq_private_buf.buflen) > repsize)
745 copied = repsize;
746
747 /* Suck it into the iovec, verify checksum if not done by hw. */
748 if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb))
749 goto out_unlock;
750
751 /* Something worked... */
752 dst_confirm(skb->dst);
753
754 xprt_complete_rqst(xprt, rovr, copied);
755
756 out_unlock:
757 spin_unlock(&xprt->sock_lock);
758 dropit:
759 skb_free_datagram(sk, skb);
760 out:
761 read_unlock(&sk->sk_callback_lock);
762}
763
764/*
765 * Copy from an skb into memory and shrink the skb.
766 */
767static inline size_t
768tcp_copy_data(skb_reader_t *desc, void *p, size_t len)
769{
770 if (len > desc->count)
771 len = desc->count;
772 if (skb_copy_bits(desc->skb, desc->offset, p, len)) {
773 dprintk("RPC: failed to copy %zu bytes from skb. %zu bytes remain\n",
774 len, desc->count);
775 return 0;
776 }
777 desc->offset += len;
778 desc->count -= len;
779 dprintk("RPC: copied %zu bytes from skb. %zu bytes remain\n",
780 len, desc->count);
781 return len;
782}
783
784/*
785 * TCP read fragment marker
786 */
787static inline void
788tcp_read_fraghdr(struct rpc_xprt *xprt, skb_reader_t *desc)
789{
790 size_t len, used;
791 char *p;
792
793 p = ((char *) &xprt->tcp_recm) + xprt->tcp_offset;
794 len = sizeof(xprt->tcp_recm) - xprt->tcp_offset;
795 used = tcp_copy_data(desc, p, len);
796 xprt->tcp_offset += used;
797 if (used != len)
798 return;
799 xprt->tcp_reclen = ntohl(xprt->tcp_recm);
800 if (xprt->tcp_reclen & 0x80000000)
801 xprt->tcp_flags |= XPRT_LAST_FRAG;
802 else
803 xprt->tcp_flags &= ~XPRT_LAST_FRAG;
804 xprt->tcp_reclen &= 0x7fffffff;
805 xprt->tcp_flags &= ~XPRT_COPY_RECM;
806 xprt->tcp_offset = 0;
807 /* Sanity check of the record length */
808 if (xprt->tcp_reclen < 4) {
809 printk(KERN_ERR "RPC: Invalid TCP record fragment length\n");
810 xprt_disconnect(xprt);
811 }
812 dprintk("RPC: reading TCP record fragment of length %d\n",
813 xprt->tcp_reclen);
814}
815
816static void
817tcp_check_recm(struct rpc_xprt *xprt)
818{
819 dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u, tcp_flags = %lx\n",
820 xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen, xprt->tcp_flags);
821 if (xprt->tcp_offset == xprt->tcp_reclen) {
822 xprt->tcp_flags |= XPRT_COPY_RECM;
823 xprt->tcp_offset = 0;
824 if (xprt->tcp_flags & XPRT_LAST_FRAG) {
825 xprt->tcp_flags &= ~XPRT_COPY_DATA;
826 xprt->tcp_flags |= XPRT_COPY_XID;
827 xprt->tcp_copied = 0;
828 }
829 }
830}
831
832/*
833 * TCP read xid
834 */
835static inline void
836tcp_read_xid(struct rpc_xprt *xprt, skb_reader_t *desc)
837{
838 size_t len, used;
839 char *p;
840
841 len = sizeof(xprt->tcp_xid) - xprt->tcp_offset;
842 dprintk("RPC: reading XID (%Zu bytes)\n", len);
843 p = ((char *) &xprt->tcp_xid) + xprt->tcp_offset;
844 used = tcp_copy_data(desc, p, len);
845 xprt->tcp_offset += used;
846 if (used != len)
847 return;
848 xprt->tcp_flags &= ~XPRT_COPY_XID;
849 xprt->tcp_flags |= XPRT_COPY_DATA;
850 xprt->tcp_copied = 4;
851 dprintk("RPC: reading reply for XID %08x\n",
852 ntohl(xprt->tcp_xid));
853 tcp_check_recm(xprt);
854}
855
856/*
857 * TCP read and complete request
858 */
859static inline void
860tcp_read_request(struct rpc_xprt *xprt, skb_reader_t *desc)
861{
862 struct rpc_rqst *req;
863 struct xdr_buf *rcvbuf;
864 size_t len;
865 ssize_t r;
866
867 /* Find and lock the request corresponding to this xid */
868 spin_lock(&xprt->sock_lock);
869 req = xprt_lookup_rqst(xprt, xprt->tcp_xid);
870 if (!req) {
871 xprt->tcp_flags &= ~XPRT_COPY_DATA;
872 dprintk("RPC: XID %08x request not found!\n",
873 ntohl(xprt->tcp_xid));
874 spin_unlock(&xprt->sock_lock);
875 return;
876 }
877
878 rcvbuf = &req->rq_private_buf;
879 len = desc->count;
880 if (len > xprt->tcp_reclen - xprt->tcp_offset) {
881 skb_reader_t my_desc;
882
883 len = xprt->tcp_reclen - xprt->tcp_offset;
884 memcpy(&my_desc, desc, sizeof(my_desc));
885 my_desc.count = len;
886 r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied,
887 &my_desc, tcp_copy_data);
888 desc->count -= r;
889 desc->offset += r;
890 } else
891 r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied,
892 desc, tcp_copy_data);
893
894 if (r > 0) {
895 xprt->tcp_copied += r;
896 xprt->tcp_offset += r;
897 }
898 if (r != len) {
899 /* Error when copying to the receive buffer,
900 * usually because we weren't able to allocate
901 * additional buffer pages. All we can do now
902 * is turn off XPRT_COPY_DATA, so the request
903 * will not receive any additional updates,
904 * and time out.
905 * Any remaining data from this record will
906 * be discarded.
907 */
908 xprt->tcp_flags &= ~XPRT_COPY_DATA;
909 dprintk("RPC: XID %08x truncated request\n",
910 ntohl(xprt->tcp_xid));
911 dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n",
912 xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen);
913 goto out;
914 }
915
916 dprintk("RPC: XID %08x read %Zd bytes\n",
917 ntohl(xprt->tcp_xid), r);
918 dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n",
919 xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen);
920
921 if (xprt->tcp_copied == req->rq_private_buf.buflen)
922 xprt->tcp_flags &= ~XPRT_COPY_DATA;
923 else if (xprt->tcp_offset == xprt->tcp_reclen) {
924 if (xprt->tcp_flags & XPRT_LAST_FRAG)
925 xprt->tcp_flags &= ~XPRT_COPY_DATA;
926 }
927
928out:
929 if (!(xprt->tcp_flags & XPRT_COPY_DATA)) {
930 dprintk("RPC: %4d received reply complete\n",
931 req->rq_task->tk_pid);
932 xprt_complete_rqst(xprt, req, xprt->tcp_copied);
933 }
934 spin_unlock(&xprt->sock_lock);
935 tcp_check_recm(xprt);
936}
937
938/*
939 * TCP discard extra bytes from a short read
940 */
941static inline void
942tcp_read_discard(struct rpc_xprt *xprt, skb_reader_t *desc)
943{
944 size_t len;
945
946 len = xprt->tcp_reclen - xprt->tcp_offset;
947 if (len > desc->count)
948 len = desc->count;
949 desc->count -= len;
950 desc->offset += len;
951 xprt->tcp_offset += len;
952 dprintk("RPC: discarded %Zu bytes\n", len);
953 tcp_check_recm(xprt);
954}
955
956/*
957 * TCP record receive routine
958 * We first have to grab the record marker, then the XID, then the data.
959 */
960static int
961tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb,
962 unsigned int offset, size_t len)
963{
964 struct rpc_xprt *xprt = rd_desc->arg.data;
965 skb_reader_t desc = {
966 .skb = skb,
967 .offset = offset,
968 .count = len,
969 .csum = 0
970 };
971
972 dprintk("RPC: tcp_data_recv\n");
973 do {
974 /* Read in a new fragment marker if necessary */
975 /* Can we ever really expect to get completely empty fragments? */
976 if (xprt->tcp_flags & XPRT_COPY_RECM) {
977 tcp_read_fraghdr(xprt, &desc);
978 continue;
979 }
980 /* Read in the xid if necessary */
981 if (xprt->tcp_flags & XPRT_COPY_XID) {
982 tcp_read_xid(xprt, &desc);
983 continue;
984 }
985 /* Read in the request data */
986 if (xprt->tcp_flags & XPRT_COPY_DATA) {
987 tcp_read_request(xprt, &desc);
988 continue;
989 }
990 /* Skip over any trailing bytes on short reads */
991 tcp_read_discard(xprt, &desc);
992 } while (desc.count);
993 dprintk("RPC: tcp_data_recv done\n");
994 return len - desc.count;
995}
996
997static void tcp_data_ready(struct sock *sk, int bytes)
998{
999 struct rpc_xprt *xprt;
1000 read_descriptor_t rd_desc;
1001
1002 read_lock(&sk->sk_callback_lock);
1003 dprintk("RPC: tcp_data_ready...\n");
1004 if (!(xprt = xprt_from_sock(sk))) {
1005 printk("RPC: tcp_data_ready socket info not found!\n");
1006 goto out;
1007 }
1008 if (xprt->shutdown)
1009 goto out;
1010
1011 /* We use rd_desc to pass struct xprt to tcp_data_recv */
1012 rd_desc.arg.data = xprt;
1013 rd_desc.count = 65536;
1014 tcp_read_sock(sk, &rd_desc, tcp_data_recv);
1015out:
1016 read_unlock(&sk->sk_callback_lock);
1017}
1018
1019static void
1020tcp_state_change(struct sock *sk)
1021{
1022 struct rpc_xprt *xprt;
1023
1024 read_lock(&sk->sk_callback_lock);
1025 if (!(xprt = xprt_from_sock(sk)))
1026 goto out;
1027 dprintk("RPC: tcp_state_change client %p...\n", xprt);
1028 dprintk("RPC: state %x conn %d dead %d zapped %d\n",
1029 sk->sk_state, xprt_connected(xprt),
1030 sock_flag(sk, SOCK_DEAD),
1031 sock_flag(sk, SOCK_ZAPPED));
1032
1033 switch (sk->sk_state) {
1034 case TCP_ESTABLISHED:
1035 spin_lock_bh(&xprt->sock_lock);
1036 if (!xprt_test_and_set_connected(xprt)) {
1037 /* Reset TCP record info */
1038 xprt->tcp_offset = 0;
1039 xprt->tcp_reclen = 0;
1040 xprt->tcp_copied = 0;
1041 xprt->tcp_flags = XPRT_COPY_RECM | XPRT_COPY_XID;
1042 rpc_wake_up(&xprt->pending);
1043 }
1044 spin_unlock_bh(&xprt->sock_lock);
1045 break;
1046 case TCP_SYN_SENT:
1047 case TCP_SYN_RECV:
1048 break;
1049 default:
1050 xprt_disconnect(xprt);
1051 break;
1052 }
1053 out:
1054 read_unlock(&sk->sk_callback_lock);
1055}
1056
1057/*
1058 * Called when more output buffer space is available for this socket.
1059 * We try not to wake our writers until they can make "significant"
1060 * progress, otherwise we'll waste resources thrashing sock_sendmsg
1061 * with a bunch of small requests.
1062 */
1063static void
1064xprt_write_space(struct sock *sk)
1065{
1066 struct rpc_xprt *xprt;
1067 struct socket *sock;
1068
1069 read_lock(&sk->sk_callback_lock);
1070 if (!(xprt = xprt_from_sock(sk)) || !(sock = sk->sk_socket))
1071 goto out;
1072 if (xprt->shutdown)
1073 goto out;
1074
1075 /* Wait until we have enough socket memory */
1076 if (xprt->stream) {
1077 /* from net/core/stream.c:sk_stream_write_space */
1078 if (sk_stream_wspace(sk) < sk_stream_min_wspace(sk))
1079 goto out;
1080 } else {
1081 /* from net/core/sock.c:sock_def_write_space */
1082 if (!sock_writeable(sk))
1083 goto out;
1084 }
1085
1086 if (!test_and_clear_bit(SOCK_NOSPACE, &sock->flags))
1087 goto out;
1088
1089 spin_lock_bh(&xprt->sock_lock);
1090 if (xprt->snd_task)
1091 rpc_wake_up_task(xprt->snd_task);
1092 spin_unlock_bh(&xprt->sock_lock);
1093out:
1094 read_unlock(&sk->sk_callback_lock);
1095}
1096
1097/*
1098 * RPC receive timeout handler. 476 * RPC receive timeout handler.
1099 */ 477 */
1100static void 478static void
@@ -1161,19 +539,10 @@ xprt_transmit(struct rpc_task *task)
1161 struct rpc_clnt *clnt = task->tk_client; 539 struct rpc_clnt *clnt = task->tk_client;
1162 struct rpc_rqst *req = task->tk_rqstp; 540 struct rpc_rqst *req = task->tk_rqstp;
1163 struct rpc_xprt *xprt = req->rq_xprt; 541 struct rpc_xprt *xprt = req->rq_xprt;
1164 int status, retry = 0; 542 int status;
1165
1166 543
1167 dprintk("RPC: %4d xprt_transmit(%u)\n", task->tk_pid, req->rq_slen); 544 dprintk("RPC: %4d xprt_transmit(%u)\n", task->tk_pid, req->rq_slen);
1168 545
1169 /* set up everything as needed. */
1170 /* Write the record marker */
1171 if (xprt->stream) {
1172 u32 *marker = req->rq_svec[0].iov_base;
1173
1174 *marker = htonl(0x80000000|(req->rq_slen-sizeof(*marker)));
1175 }
1176
1177 smp_rmb(); 546 smp_rmb();
1178 if (!req->rq_received) { 547 if (!req->rq_received) {
1179 if (list_empty(&req->rq_list)) { 548 if (list_empty(&req->rq_list)) {
@@ -1191,41 +560,9 @@ xprt_transmit(struct rpc_task *task)
1191 } else if (!req->rq_bytes_sent) 560 } else if (!req->rq_bytes_sent)
1192 return; 561 return;
1193 562
1194 /* Continue transmitting the packet/record. We must be careful 563 status = xprt->ops->send_request(task);
1195 * to cope with writespace callbacks arriving _after_ we have 564 if (!status)
1196 * called xprt_sendmsg(). 565 goto out_receive;
1197 */
1198 while (1) {
1199 req->rq_xtime = jiffies;
1200 status = xprt_sendmsg(xprt, req);
1201
1202 if (status < 0)
1203 break;
1204
1205 if (xprt->stream) {
1206 req->rq_bytes_sent += status;
1207
1208 /* If we've sent the entire packet, immediately
1209 * reset the count of bytes sent. */
1210 if (req->rq_bytes_sent >= req->rq_slen) {
1211 req->rq_bytes_sent = 0;
1212 goto out_receive;
1213 }
1214 } else {
1215 if (status >= req->rq_slen)
1216 goto out_receive;
1217 status = -EAGAIN;
1218 break;
1219 }
1220
1221 dprintk("RPC: %4d xmit incomplete (%d left of %d)\n",
1222 task->tk_pid, req->rq_slen - req->rq_bytes_sent,
1223 req->rq_slen);
1224
1225 status = -EAGAIN;
1226 if (retry++ > 50)
1227 break;
1228 }
1229 566
1230 /* Note: at this point, task->tk_sleeping has not yet been set, 567 /* Note: at this point, task->tk_sleeping has not yet been set,
1231 * hence there is no danger of the waking up task being put on 568 * hence there is no danger of the waking up task being put on
@@ -1234,26 +571,10 @@ xprt_transmit(struct rpc_task *task)
1234 task->tk_status = status; 571 task->tk_status = status;
1235 572
1236 switch (status) { 573 switch (status) {
1237 case -EAGAIN:
1238 if (test_bit(SOCK_ASYNC_NOSPACE, &xprt->sock->flags)) {
1239 /* Protect against races with xprt_write_space */
1240 spin_lock_bh(&xprt->sock_lock);
1241 /* Don't race with disconnect */
1242 if (!xprt_connected(xprt))
1243 task->tk_status = -ENOTCONN;
1244 else if (test_bit(SOCK_NOSPACE, &xprt->sock->flags)) {
1245 task->tk_timeout = req->rq_timeout;
1246 rpc_sleep_on(&xprt->pending, task, NULL, NULL);
1247 }
1248 spin_unlock_bh(&xprt->sock_lock);
1249 return;
1250 }
1251 /* Keep holding the socket if it is blocked */
1252 rpc_delay(task, HZ>>4);
1253 return;
1254 case -ECONNREFUSED: 574 case -ECONNREFUSED:
1255 task->tk_timeout = RPC_REESTABLISH_TIMEOUT; 575 task->tk_timeout = RPC_REESTABLISH_TIMEOUT;
1256 rpc_sleep_on(&xprt->sending, task, NULL, NULL); 576 rpc_sleep_on(&xprt->sending, task, NULL, NULL);
577 case -EAGAIN:
1257 case -ENOTCONN: 578 case -ENOTCONN:
1258 return; 579 return;
1259 default: 580 default:
@@ -1367,7 +688,8 @@ xprt_release(struct rpc_task *task)
1367 list_del(&req->rq_list); 688 list_del(&req->rq_list);
1368 xprt->last_used = jiffies; 689 xprt->last_used = jiffies;
1369 if (list_empty(&xprt->recv) && !xprt->shutdown) 690 if (list_empty(&xprt->recv) && !xprt->shutdown)
1370 mod_timer(&xprt->timer, xprt->last_used + XPRT_IDLE_TIMEOUT); 691 mod_timer(&xprt->timer,
692 xprt->last_used + RPC_IDLE_DISCONNECT_TIMEOUT);
1371 spin_unlock_bh(&xprt->sock_lock); 693 spin_unlock_bh(&xprt->sock_lock);
1372 task->tk_rqstp = NULL; 694 task->tk_rqstp = NULL;
1373 memset(req, 0, sizeof(*req)); /* mark unused */ 695 memset(req, 0, sizeof(*req)); /* mark unused */
@@ -1381,18 +703,6 @@ xprt_release(struct rpc_task *task)
1381} 703}
1382 704
1383/* 705/*
1384 * Set default timeout parameters
1385 */
1386static void
1387xprt_default_timeout(struct rpc_timeout *to, int proto)
1388{
1389 if (proto == IPPROTO_UDP)
1390 xprt_set_timeout(to, 5, 5 * HZ);
1391 else
1392 xprt_set_timeout(to, 2, 60 * HZ);
1393}
1394
1395/*
1396 * Set constant timeout 706 * Set constant timeout
1397 */ 707 */
1398void 708void
@@ -1405,68 +715,51 @@ xprt_set_timeout(struct rpc_timeout *to, unsigned int retr, unsigned long incr)
1405 to->to_exponential = 0; 715 to->to_exponential = 0;
1406} 716}
1407 717
1408unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE;
1409unsigned int xprt_tcp_slot_table_entries = RPC_DEF_SLOT_TABLE;
1410
1411/* 718/*
1412 * Initialize an RPC client 719 * Initialize an RPC client
1413 */ 720 */
1414static struct rpc_xprt * 721static struct rpc_xprt *
1415xprt_setup(int proto, struct sockaddr_in *ap, struct rpc_timeout *to) 722xprt_setup(int proto, struct sockaddr_in *ap, struct rpc_timeout *to)
1416{ 723{
724 int result;
1417 struct rpc_xprt *xprt; 725 struct rpc_xprt *xprt;
1418 unsigned int entries;
1419 size_t slot_table_size;
1420 struct rpc_rqst *req; 726 struct rpc_rqst *req;
1421 727
1422 dprintk("RPC: setting up %s transport...\n",
1423 proto == IPPROTO_UDP? "UDP" : "TCP");
1424
1425 entries = (proto == IPPROTO_TCP)?
1426 xprt_tcp_slot_table_entries : xprt_udp_slot_table_entries;
1427
1428 if ((xprt = kmalloc(sizeof(struct rpc_xprt), GFP_KERNEL)) == NULL) 728 if ((xprt = kmalloc(sizeof(struct rpc_xprt), GFP_KERNEL)) == NULL)
1429 return ERR_PTR(-ENOMEM); 729 return ERR_PTR(-ENOMEM);
1430 memset(xprt, 0, sizeof(*xprt)); /* Nnnngh! */ 730 memset(xprt, 0, sizeof(*xprt)); /* Nnnngh! */
1431 xprt->max_reqs = entries;
1432 slot_table_size = entries * sizeof(xprt->slot[0]);
1433 xprt->slot = kmalloc(slot_table_size, GFP_KERNEL);
1434 if (xprt->slot == NULL) {
1435 kfree(xprt);
1436 return ERR_PTR(-ENOMEM);
1437 }
1438 memset(xprt->slot, 0, slot_table_size);
1439 731
1440 xprt->addr = *ap; 732 xprt->addr = *ap;
1441 xprt->prot = proto; 733
1442 xprt->stream = (proto == IPPROTO_TCP)? 1 : 0; 734 switch (proto) {
1443 if (xprt->stream) { 735 case IPPROTO_UDP:
1444 xprt->cwnd = RPC_MAXCWND(xprt); 736 result = xs_setup_udp(xprt, to);
1445 xprt->nocong = 1; 737 break;
1446 xprt->max_payload = (1U << 31) - 1; 738 case IPPROTO_TCP:
1447 } else { 739 result = xs_setup_tcp(xprt, to);
1448 xprt->cwnd = RPC_INITCWND; 740 break;
1449 xprt->max_payload = (1U << 16) - (MAX_HEADER << 3); 741 default:
742 printk(KERN_ERR "RPC: unrecognized transport protocol: %d\n",
743 proto);
744 result = -EIO;
745 break;
746 }
747 if (result) {
748 kfree(xprt);
749 return ERR_PTR(result);
1450 } 750 }
751
1451 spin_lock_init(&xprt->sock_lock); 752 spin_lock_init(&xprt->sock_lock);
1452 spin_lock_init(&xprt->xprt_lock); 753 spin_lock_init(&xprt->xprt_lock);
1453 init_waitqueue_head(&xprt->cong_wait); 754 init_waitqueue_head(&xprt->cong_wait);
1454 755
1455 INIT_LIST_HEAD(&xprt->free); 756 INIT_LIST_HEAD(&xprt->free);
1456 INIT_LIST_HEAD(&xprt->recv); 757 INIT_LIST_HEAD(&xprt->recv);
1457 INIT_WORK(&xprt->sock_connect, xprt_socket_connect, xprt);
1458 INIT_WORK(&xprt->task_cleanup, xprt_socket_autoclose, xprt); 758 INIT_WORK(&xprt->task_cleanup, xprt_socket_autoclose, xprt);
1459 init_timer(&xprt->timer); 759 init_timer(&xprt->timer);
1460 xprt->timer.function = xprt_init_autodisconnect; 760 xprt->timer.function = xprt_init_autodisconnect;
1461 xprt->timer.data = (unsigned long) xprt; 761 xprt->timer.data = (unsigned long) xprt;
1462 xprt->last_used = jiffies; 762 xprt->last_used = jiffies;
1463 xprt->port = XPRT_MAX_RESVPORT;
1464
1465 /* Set timeout parameters */
1466 if (to) {
1467 xprt->timeout = *to;
1468 } else
1469 xprt_default_timeout(&xprt->timeout, xprt->prot);
1470 763
1471 rpc_init_wait_queue(&xprt->pending, "xprt_pending"); 764 rpc_init_wait_queue(&xprt->pending, "xprt_pending");
1472 rpc_init_wait_queue(&xprt->sending, "xprt_sending"); 765 rpc_init_wait_queue(&xprt->sending, "xprt_sending");
@@ -1474,14 +767,11 @@ xprt_setup(int proto, struct sockaddr_in *ap, struct rpc_timeout *to)
1474 rpc_init_priority_wait_queue(&xprt->backlog, "xprt_backlog"); 767 rpc_init_priority_wait_queue(&xprt->backlog, "xprt_backlog");
1475 768
1476 /* initialize free list */ 769 /* initialize free list */
1477 for (req = &xprt->slot[entries-1]; req >= &xprt->slot[0]; req--) 770 for (req = &xprt->slot[xprt->max_reqs-1]; req >= &xprt->slot[0]; req--)
1478 list_add(&req->rq_list, &xprt->free); 771 list_add(&req->rq_list, &xprt->free);
1479 772
1480 xprt_init_xid(xprt); 773 xprt_init_xid(xprt);
1481 774
1482 /* Check whether we want to use a reserved port */
1483 xprt->resvport = capable(CAP_NET_BIND_SERVICE) ? 1 : 0;
1484
1485 dprintk("RPC: created transport %p with %u slots\n", xprt, 775 dprintk("RPC: created transport %p with %u slots\n", xprt,
1486 xprt->max_reqs); 776 xprt->max_reqs);
1487 777
@@ -1489,120 +779,6 @@ xprt_setup(int proto, struct sockaddr_in *ap, struct rpc_timeout *to)
1489} 779}
1490 780
1491/* 781/*
1492 * Bind to a reserved port
1493 */
1494static inline int xprt_bindresvport(struct rpc_xprt *xprt, struct socket *sock)
1495{
1496 struct sockaddr_in myaddr = {
1497 .sin_family = AF_INET,
1498 };
1499 int err, port;
1500
1501 /* Were we already bound to a given port? Try to reuse it */
1502 port = xprt->port;
1503 do {
1504 myaddr.sin_port = htons(port);
1505 err = sock->ops->bind(sock, (struct sockaddr *) &myaddr,
1506 sizeof(myaddr));
1507 if (err == 0) {
1508 xprt->port = port;
1509 return 0;
1510 }
1511 if (--port == 0)
1512 port = XPRT_MAX_RESVPORT;
1513 } while (err == -EADDRINUSE && port != xprt->port);
1514
1515 printk("RPC: Can't bind to reserved port (%d).\n", -err);
1516 return err;
1517}
1518
1519static void
1520xprt_bind_socket(struct rpc_xprt *xprt, struct socket *sock)
1521{
1522 struct sock *sk = sock->sk;
1523
1524 if (xprt->inet)
1525 return;
1526
1527 write_lock_bh(&sk->sk_callback_lock);
1528 sk->sk_user_data = xprt;
1529 xprt->old_data_ready = sk->sk_data_ready;
1530 xprt->old_state_change = sk->sk_state_change;
1531 xprt->old_write_space = sk->sk_write_space;
1532 if (xprt->prot == IPPROTO_UDP) {
1533 sk->sk_data_ready = udp_data_ready;
1534 sk->sk_no_check = UDP_CSUM_NORCV;
1535 xprt_set_connected(xprt);
1536 } else {
1537 tcp_sk(sk)->nonagle = 1; /* disable Nagle's algorithm */
1538 sk->sk_data_ready = tcp_data_ready;
1539 sk->sk_state_change = tcp_state_change;
1540 xprt_clear_connected(xprt);
1541 }
1542 sk->sk_write_space = xprt_write_space;
1543
1544 /* Reset to new socket */
1545 xprt->sock = sock;
1546 xprt->inet = sk;
1547 write_unlock_bh(&sk->sk_callback_lock);
1548
1549 return;
1550}
1551
1552/*
1553 * Set socket buffer length
1554 */
1555void
1556xprt_sock_setbufsize(struct rpc_xprt *xprt)
1557{
1558 struct sock *sk = xprt->inet;
1559
1560 if (xprt->stream)
1561 return;
1562 if (xprt->rcvsize) {
1563 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
1564 sk->sk_rcvbuf = xprt->rcvsize * xprt->max_reqs * 2;
1565 }
1566 if (xprt->sndsize) {
1567 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
1568 sk->sk_sndbuf = xprt->sndsize * xprt->max_reqs * 2;
1569 sk->sk_write_space(sk);
1570 }
1571}
1572
1573/*
1574 * Datastream sockets are created here, but xprt_connect will create
1575 * and connect stream sockets.
1576 */
1577static struct socket * xprt_create_socket(struct rpc_xprt *xprt, int proto, int resvport)
1578{
1579 struct socket *sock;
1580 int type, err;
1581
1582 dprintk("RPC: xprt_create_socket(%s %d)\n",
1583 (proto == IPPROTO_UDP)? "udp" : "tcp", proto);
1584
1585 type = (proto == IPPROTO_UDP)? SOCK_DGRAM : SOCK_STREAM;
1586
1587 if ((err = sock_create_kern(PF_INET, type, proto, &sock)) < 0) {
1588 printk("RPC: can't create socket (%d).\n", -err);
1589 return NULL;
1590 }
1591
1592 /* If the caller has the capability, bind to a reserved port */
1593 if (resvport && xprt_bindresvport(xprt, sock) < 0) {
1594 printk("RPC: can't bind to reserved port.\n");
1595 goto failed;
1596 }
1597
1598 return sock;
1599
1600failed:
1601 sock_release(sock);
1602 return NULL;
1603}
1604
1605/*
1606 * Create an RPC client transport given the protocol and peer address. 782 * Create an RPC client transport given the protocol and peer address.
1607 */ 783 */
1608struct rpc_xprt * 784struct rpc_xprt *
@@ -1631,10 +807,6 @@ xprt_shutdown(struct rpc_xprt *xprt)
1631 rpc_wake_up(&xprt->backlog); 807 rpc_wake_up(&xprt->backlog);
1632 wake_up(&xprt->cong_wait); 808 wake_up(&xprt->cong_wait);
1633 del_timer_sync(&xprt->timer); 809 del_timer_sync(&xprt->timer);
1634
1635 /* synchronously wait for connect worker to finish */
1636 cancel_delayed_work(&xprt->sock_connect);
1637 flush_scheduled_work();
1638} 810}
1639 811
1640/* 812/*
@@ -1655,9 +827,7 @@ xprt_destroy(struct rpc_xprt *xprt)
1655{ 827{
1656 dprintk("RPC: destroying transport %p\n", xprt); 828 dprintk("RPC: destroying transport %p\n", xprt);
1657 xprt_shutdown(xprt); 829 xprt_shutdown(xprt);
1658 xprt_disconnect(xprt); 830 xprt->ops->destroy(xprt);
1659 xprt_close(xprt);
1660 kfree(xprt->slot);
1661 kfree(xprt); 831 kfree(xprt);
1662 832
1663 return 0; 833 return 0;
diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
new file mode 100644
index 000000000000..fa1180ac4823
--- /dev/null
+++ b/net/sunrpc/xprtsock.c
@@ -0,0 +1,1021 @@
1/*
2 * linux/net/sunrpc/xprtsock.c
3 *
4 * Client-side transport implementation for sockets.
5 *
6 * TCP callback races fixes (C) 1998 Red Hat Software <alan@redhat.com>
7 * TCP send fixes (C) 1998 Red Hat Software <alan@redhat.com>
8 * TCP NFS related read + write fixes
9 * (C) 1999 Dave Airlie, University of Limerick, Ireland <airlied@linux.ie>
10 *
11 * Rewrite of larges part of the code in order to stabilize TCP stuff.
12 * Fix behaviour when socket buffer is full.
13 * (C) 1999 Trond Myklebust <trond.myklebust@fys.uio.no>
14 */
15
16#include <linux/types.h>
17#include <linux/slab.h>
18#include <linux/capability.h>
19#include <linux/sched.h>
20#include <linux/pagemap.h>
21#include <linux/errno.h>
22#include <linux/socket.h>
23#include <linux/in.h>
24#include <linux/net.h>
25#include <linux/mm.h>
26#include <linux/udp.h>
27#include <linux/tcp.h>
28#include <linux/sunrpc/clnt.h>
29#include <linux/file.h>
30
31#include <net/sock.h>
32#include <net/checksum.h>
33#include <net/udp.h>
34#include <net/tcp.h>
35
36#ifdef RPC_DEBUG
37# undef RPC_DEBUG_DATA
38# define RPCDBG_FACILITY RPCDBG_XPRT
39#endif
40
41#define XPRT_MAX_RESVPORT (800)
42
43#ifdef RPC_DEBUG_DATA
44/*
45 * Print the buffer contents (first 128 bytes only--just enough for
46 * diropres return).
47 */
48static void
49xprt_pktdump(char *msg, u32 *packet, unsigned int count)
50{
51 u8 *buf = (u8 *) packet;
52 int j;
53
54 dprintk("RPC: %s\n", msg);
55 for (j = 0; j < count && j < 128; j += 4) {
56 if (!(j & 31)) {
57 if (j)
58 dprintk("\n");
59 dprintk("0x%04x ", j);
60 }
61 dprintk("%02x%02x%02x%02x ",
62 buf[j], buf[j+1], buf[j+2], buf[j+3]);
63 }
64 dprintk("\n");
65}
66#else
67static inline void
68xprt_pktdump(char *msg, u32 *packet, unsigned int count)
69{
70 /* NOP */
71}
72#endif
73
74/*
75 * Look up RPC transport given an INET socket
76 */
77static inline struct rpc_xprt *
78xprt_from_sock(struct sock *sk)
79{
80 return (struct rpc_xprt *) sk->sk_user_data;
81}
82
83static int
84xdr_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen,
85 struct xdr_buf *xdr, unsigned int base, int msgflags)
86{
87 struct page **ppage = xdr->pages;
88 unsigned int len, pglen = xdr->page_len;
89 int err, ret = 0;
90 ssize_t (*sendpage)(struct socket *, struct page *, int, size_t, int);
91
92 len = xdr->head[0].iov_len;
93 if (base < len || (addr != NULL && base == 0)) {
94 struct kvec iov = {
95 .iov_base = xdr->head[0].iov_base + base,
96 .iov_len = len - base,
97 };
98 struct msghdr msg = {
99 .msg_name = addr,
100 .msg_namelen = addrlen,
101 .msg_flags = msgflags,
102 };
103 if (xdr->len > len)
104 msg.msg_flags |= MSG_MORE;
105
106 if (iov.iov_len != 0)
107 err = kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len);
108 else
109 err = kernel_sendmsg(sock, &msg, NULL, 0, 0);
110 if (ret == 0)
111 ret = err;
112 else if (err > 0)
113 ret += err;
114 if (err != iov.iov_len)
115 goto out;
116 base = 0;
117 } else
118 base -= len;
119
120 if (pglen == 0)
121 goto copy_tail;
122 if (base >= pglen) {
123 base -= pglen;
124 goto copy_tail;
125 }
126 if (base || xdr->page_base) {
127 pglen -= base;
128 base += xdr->page_base;
129 ppage += base >> PAGE_CACHE_SHIFT;
130 base &= ~PAGE_CACHE_MASK;
131 }
132
133 sendpage = sock->ops->sendpage ? : sock_no_sendpage;
134 do {
135 int flags = msgflags;
136
137 len = PAGE_CACHE_SIZE;
138 if (base)
139 len -= base;
140 if (pglen < len)
141 len = pglen;
142
143 if (pglen != len || xdr->tail[0].iov_len != 0)
144 flags |= MSG_MORE;
145
146 /* Hmm... We might be dealing with highmem pages */
147 if (PageHighMem(*ppage))
148 sendpage = sock_no_sendpage;
149 err = sendpage(sock, *ppage, base, len, flags);
150 if (ret == 0)
151 ret = err;
152 else if (err > 0)
153 ret += err;
154 if (err != len)
155 goto out;
156 base = 0;
157 ppage++;
158 } while ((pglen -= len) != 0);
159copy_tail:
160 len = xdr->tail[0].iov_len;
161 if (base < len) {
162 struct kvec iov = {
163 .iov_base = xdr->tail[0].iov_base + base,
164 .iov_len = len - base,
165 };
166 struct msghdr msg = {
167 .msg_flags = msgflags,
168 };
169 err = kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len);
170 if (ret == 0)
171 ret = err;
172 else if (err > 0)
173 ret += err;
174 }
175out:
176 return ret;
177}
178
179/*
180 * Write data to socket.
181 */
182static inline int
183xprt_sendmsg(struct rpc_xprt *xprt, struct rpc_rqst *req)
184{
185 struct socket *sock = xprt->sock;
186 struct xdr_buf *xdr = &req->rq_snd_buf;
187 struct sockaddr *addr = NULL;
188 int addrlen = 0;
189 unsigned int skip;
190 int result;
191
192 if (!sock)
193 return -ENOTCONN;
194
195 xprt_pktdump("packet data:",
196 req->rq_svec->iov_base,
197 req->rq_svec->iov_len);
198
199 /* For UDP, we need to provide an address */
200 if (!xprt->stream) {
201 addr = (struct sockaddr *) &xprt->addr;
202 addrlen = sizeof(xprt->addr);
203 }
204 /* Dont repeat bytes */
205 skip = req->rq_bytes_sent;
206
207 clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags);
208 result = xdr_sendpages(sock, addr, addrlen, xdr, skip, MSG_DONTWAIT);
209
210 dprintk("RPC: xprt_sendmsg(%d) = %d\n", xdr->len - skip, result);
211
212 if (result >= 0)
213 return result;
214
215 switch (result) {
216 case -ECONNREFUSED:
217 /* When the server has died, an ICMP port unreachable message
218 * prompts ECONNREFUSED.
219 */
220 case -EAGAIN:
221 break;
222 case -ECONNRESET:
223 case -ENOTCONN:
224 case -EPIPE:
225 /* connection broken */
226 if (xprt->stream)
227 result = -ENOTCONN;
228 break;
229 default:
230 printk(KERN_NOTICE "RPC: sendmsg returned error %d\n", -result);
231 }
232 return result;
233}
234
235static int
236xprt_send_request(struct rpc_task *task)
237{
238 struct rpc_rqst *req = task->tk_rqstp;
239 struct rpc_xprt *xprt = req->rq_xprt;
240 int status, retry = 0;
241
242 /* set up everything as needed. */
243 /* Write the record marker */
244 if (xprt->stream) {
245 u32 *marker = req->rq_svec[0].iov_base;
246
247 *marker = htonl(0x80000000|(req->rq_slen-sizeof(*marker)));
248 }
249
250 /* Continue transmitting the packet/record. We must be careful
251 * to cope with writespace callbacks arriving _after_ we have
252 * called xprt_sendmsg().
253 */
254 while (1) {
255 req->rq_xtime = jiffies;
256 status = xprt_sendmsg(xprt, req);
257
258 if (status < 0)
259 break;
260
261 if (xprt->stream) {
262 req->rq_bytes_sent += status;
263
264 /* If we've sent the entire packet, immediately
265 * reset the count of bytes sent. */
266 if (req->rq_bytes_sent >= req->rq_slen) {
267 req->rq_bytes_sent = 0;
268 return 0;
269 }
270 } else {
271 if (status >= req->rq_slen)
272 return 0;
273 status = -EAGAIN;
274 break;
275 }
276
277 dprintk("RPC: %4d xmit incomplete (%d left of %d)\n",
278 task->tk_pid, req->rq_slen - req->rq_bytes_sent,
279 req->rq_slen);
280
281 status = -EAGAIN;
282 if (retry++ > 50)
283 break;
284 }
285
286 if (status == -EAGAIN) {
287 if (test_bit(SOCK_ASYNC_NOSPACE, &xprt->sock->flags)) {
288 /* Protect against races with xprt_write_space */
289 spin_lock_bh(&xprt->sock_lock);
290 /* Don't race with disconnect */
291 if (!xprt_connected(xprt))
292 task->tk_status = -ENOTCONN;
293 else if (test_bit(SOCK_NOSPACE, &xprt->sock->flags)) {
294 task->tk_timeout = req->rq_timeout;
295 rpc_sleep_on(&xprt->pending, task, NULL, NULL);
296 }
297 spin_unlock_bh(&xprt->sock_lock);
298 return status;
299 }
300 /* Keep holding the socket if it is blocked */
301 rpc_delay(task, HZ>>4);
302 }
303 return status;
304}
305
306/*
307 * Close down a transport socket
308 */
309static void
310xprt_close(struct rpc_xprt *xprt)
311{
312 struct socket *sock = xprt->sock;
313 struct sock *sk = xprt->inet;
314
315 if (!sk)
316 return;
317
318 write_lock_bh(&sk->sk_callback_lock);
319 xprt->inet = NULL;
320 xprt->sock = NULL;
321
322 sk->sk_user_data = NULL;
323 sk->sk_data_ready = xprt->old_data_ready;
324 sk->sk_state_change = xprt->old_state_change;
325 sk->sk_write_space = xprt->old_write_space;
326 write_unlock_bh(&sk->sk_callback_lock);
327
328 sk->sk_no_check = 0;
329
330 sock_release(sock);
331}
332
333static void xprt_socket_destroy(struct rpc_xprt *xprt)
334{
335 cancel_delayed_work(&xprt->sock_connect);
336 flush_scheduled_work();
337
338 xprt_disconnect(xprt);
339 xprt_close(xprt);
340 kfree(xprt->slot);
341}
342
343/*
344 * Input handler for RPC replies. Called from a bottom half and hence
345 * atomic.
346 */
347static void
348udp_data_ready(struct sock *sk, int len)
349{
350 struct rpc_task *task;
351 struct rpc_xprt *xprt;
352 struct rpc_rqst *rovr;
353 struct sk_buff *skb;
354 int err, repsize, copied;
355 u32 _xid, *xp;
356
357 read_lock(&sk->sk_callback_lock);
358 dprintk("RPC: udp_data_ready...\n");
359 if (!(xprt = xprt_from_sock(sk))) {
360 printk("RPC: udp_data_ready request not found!\n");
361 goto out;
362 }
363
364 dprintk("RPC: udp_data_ready client %p\n", xprt);
365
366 if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL)
367 goto out;
368
369 if (xprt->shutdown)
370 goto dropit;
371
372 repsize = skb->len - sizeof(struct udphdr);
373 if (repsize < 4) {
374 printk("RPC: impossible RPC reply size %d!\n", repsize);
375 goto dropit;
376 }
377
378 /* Copy the XID from the skb... */
379 xp = skb_header_pointer(skb, sizeof(struct udphdr),
380 sizeof(_xid), &_xid);
381 if (xp == NULL)
382 goto dropit;
383
384 /* Look up and lock the request corresponding to the given XID */
385 spin_lock(&xprt->sock_lock);
386 rovr = xprt_lookup_rqst(xprt, *xp);
387 if (!rovr)
388 goto out_unlock;
389 task = rovr->rq_task;
390
391 dprintk("RPC: %4d received reply\n", task->tk_pid);
392
393 if ((copied = rovr->rq_private_buf.buflen) > repsize)
394 copied = repsize;
395
396 /* Suck it into the iovec, verify checksum if not done by hw. */
397 if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb))
398 goto out_unlock;
399
400 /* Something worked... */
401 dst_confirm(skb->dst);
402
403 xprt_complete_rqst(xprt, rovr, copied);
404
405 out_unlock:
406 spin_unlock(&xprt->sock_lock);
407 dropit:
408 skb_free_datagram(sk, skb);
409 out:
410 read_unlock(&sk->sk_callback_lock);
411}
412
413/*
414 * Copy from an skb into memory and shrink the skb.
415 */
416static inline size_t
417tcp_copy_data(skb_reader_t *desc, void *p, size_t len)
418{
419 if (len > desc->count)
420 len = desc->count;
421 if (skb_copy_bits(desc->skb, desc->offset, p, len)) {
422 dprintk("RPC: failed to copy %zu bytes from skb. %zu bytes remain\n",
423 len, desc->count);
424 return 0;
425 }
426 desc->offset += len;
427 desc->count -= len;
428 dprintk("RPC: copied %zu bytes from skb. %zu bytes remain\n",
429 len, desc->count);
430 return len;
431}
432
433/*
434 * TCP read fragment marker
435 */
436static inline void
437tcp_read_fraghdr(struct rpc_xprt *xprt, skb_reader_t *desc)
438{
439 size_t len, used;
440 char *p;
441
442 p = ((char *) &xprt->tcp_recm) + xprt->tcp_offset;
443 len = sizeof(xprt->tcp_recm) - xprt->tcp_offset;
444 used = tcp_copy_data(desc, p, len);
445 xprt->tcp_offset += used;
446 if (used != len)
447 return;
448 xprt->tcp_reclen = ntohl(xprt->tcp_recm);
449 if (xprt->tcp_reclen & 0x80000000)
450 xprt->tcp_flags |= XPRT_LAST_FRAG;
451 else
452 xprt->tcp_flags &= ~XPRT_LAST_FRAG;
453 xprt->tcp_reclen &= 0x7fffffff;
454 xprt->tcp_flags &= ~XPRT_COPY_RECM;
455 xprt->tcp_offset = 0;
456 /* Sanity check of the record length */
457 if (xprt->tcp_reclen < 4) {
458 printk(KERN_ERR "RPC: Invalid TCP record fragment length\n");
459 xprt_disconnect(xprt);
460 }
461 dprintk("RPC: reading TCP record fragment of length %d\n",
462 xprt->tcp_reclen);
463}
464
465static void
466tcp_check_recm(struct rpc_xprt *xprt)
467{
468 dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u, tcp_flags = %lx\n",
469 xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen, xprt->tcp_flags);
470 if (xprt->tcp_offset == xprt->tcp_reclen) {
471 xprt->tcp_flags |= XPRT_COPY_RECM;
472 xprt->tcp_offset = 0;
473 if (xprt->tcp_flags & XPRT_LAST_FRAG) {
474 xprt->tcp_flags &= ~XPRT_COPY_DATA;
475 xprt->tcp_flags |= XPRT_COPY_XID;
476 xprt->tcp_copied = 0;
477 }
478 }
479}
480
481/*
482 * TCP read xid
483 */
484static inline void
485tcp_read_xid(struct rpc_xprt *xprt, skb_reader_t *desc)
486{
487 size_t len, used;
488 char *p;
489
490 len = sizeof(xprt->tcp_xid) - xprt->tcp_offset;
491 dprintk("RPC: reading XID (%Zu bytes)\n", len);
492 p = ((char *) &xprt->tcp_xid) + xprt->tcp_offset;
493 used = tcp_copy_data(desc, p, len);
494 xprt->tcp_offset += used;
495 if (used != len)
496 return;
497 xprt->tcp_flags &= ~XPRT_COPY_XID;
498 xprt->tcp_flags |= XPRT_COPY_DATA;
499 xprt->tcp_copied = 4;
500 dprintk("RPC: reading reply for XID %08x\n",
501 ntohl(xprt->tcp_xid));
502 tcp_check_recm(xprt);
503}
504
505/*
506 * TCP read and complete request
507 */
508static inline void
509tcp_read_request(struct rpc_xprt *xprt, skb_reader_t *desc)
510{
511 struct rpc_rqst *req;
512 struct xdr_buf *rcvbuf;
513 size_t len;
514 ssize_t r;
515
516 /* Find and lock the request corresponding to this xid */
517 spin_lock(&xprt->sock_lock);
518 req = xprt_lookup_rqst(xprt, xprt->tcp_xid);
519 if (!req) {
520 xprt->tcp_flags &= ~XPRT_COPY_DATA;
521 dprintk("RPC: XID %08x request not found!\n",
522 ntohl(xprt->tcp_xid));
523 spin_unlock(&xprt->sock_lock);
524 return;
525 }
526
527 rcvbuf = &req->rq_private_buf;
528 len = desc->count;
529 if (len > xprt->tcp_reclen - xprt->tcp_offset) {
530 skb_reader_t my_desc;
531
532 len = xprt->tcp_reclen - xprt->tcp_offset;
533 memcpy(&my_desc, desc, sizeof(my_desc));
534 my_desc.count = len;
535 r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied,
536 &my_desc, tcp_copy_data);
537 desc->count -= r;
538 desc->offset += r;
539 } else
540 r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied,
541 desc, tcp_copy_data);
542
543 if (r > 0) {
544 xprt->tcp_copied += r;
545 xprt->tcp_offset += r;
546 }
547 if (r != len) {
548 /* Error when copying to the receive buffer,
549 * usually because we weren't able to allocate
550 * additional buffer pages. All we can do now
551 * is turn off XPRT_COPY_DATA, so the request
552 * will not receive any additional updates,
553 * and time out.
554 * Any remaining data from this record will
555 * be discarded.
556 */
557 xprt->tcp_flags &= ~XPRT_COPY_DATA;
558 dprintk("RPC: XID %08x truncated request\n",
559 ntohl(xprt->tcp_xid));
560 dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n",
561 xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen);
562 goto out;
563 }
564
565 dprintk("RPC: XID %08x read %Zd bytes\n",
566 ntohl(xprt->tcp_xid), r);
567 dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n",
568 xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen);
569
570 if (xprt->tcp_copied == req->rq_private_buf.buflen)
571 xprt->tcp_flags &= ~XPRT_COPY_DATA;
572 else if (xprt->tcp_offset == xprt->tcp_reclen) {
573 if (xprt->tcp_flags & XPRT_LAST_FRAG)
574 xprt->tcp_flags &= ~XPRT_COPY_DATA;
575 }
576
577out:
578 if (!(xprt->tcp_flags & XPRT_COPY_DATA)) {
579 dprintk("RPC: %4d received reply complete\n",
580 req->rq_task->tk_pid);
581 xprt_complete_rqst(xprt, req, xprt->tcp_copied);
582 }
583 spin_unlock(&xprt->sock_lock);
584 tcp_check_recm(xprt);
585}
586
587/*
588 * TCP discard extra bytes from a short read
589 */
590static inline void
591tcp_read_discard(struct rpc_xprt *xprt, skb_reader_t *desc)
592{
593 size_t len;
594
595 len = xprt->tcp_reclen - xprt->tcp_offset;
596 if (len > desc->count)
597 len = desc->count;
598 desc->count -= len;
599 desc->offset += len;
600 xprt->tcp_offset += len;
601 dprintk("RPC: discarded %Zu bytes\n", len);
602 tcp_check_recm(xprt);
603}
604
605/*
606 * TCP record receive routine
607 * We first have to grab the record marker, then the XID, then the data.
608 */
609static int
610tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb,
611 unsigned int offset, size_t len)
612{
613 struct rpc_xprt *xprt = rd_desc->arg.data;
614 skb_reader_t desc = {
615 .skb = skb,
616 .offset = offset,
617 .count = len,
618 .csum = 0
619 };
620
621 dprintk("RPC: tcp_data_recv\n");
622 do {
623 /* Read in a new fragment marker if necessary */
624 /* Can we ever really expect to get completely empty fragments? */
625 if (xprt->tcp_flags & XPRT_COPY_RECM) {
626 tcp_read_fraghdr(xprt, &desc);
627 continue;
628 }
629 /* Read in the xid if necessary */
630 if (xprt->tcp_flags & XPRT_COPY_XID) {
631 tcp_read_xid(xprt, &desc);
632 continue;
633 }
634 /* Read in the request data */
635 if (xprt->tcp_flags & XPRT_COPY_DATA) {
636 tcp_read_request(xprt, &desc);
637 continue;
638 }
639 /* Skip over any trailing bytes on short reads */
640 tcp_read_discard(xprt, &desc);
641 } while (desc.count);
642 dprintk("RPC: tcp_data_recv done\n");
643 return len - desc.count;
644}
645
646static void tcp_data_ready(struct sock *sk, int bytes)
647{
648 struct rpc_xprt *xprt;
649 read_descriptor_t rd_desc;
650
651 read_lock(&sk->sk_callback_lock);
652 dprintk("RPC: tcp_data_ready...\n");
653 if (!(xprt = xprt_from_sock(sk))) {
654 printk("RPC: tcp_data_ready socket info not found!\n");
655 goto out;
656 }
657 if (xprt->shutdown)
658 goto out;
659
660 /* We use rd_desc to pass struct xprt to tcp_data_recv */
661 rd_desc.arg.data = xprt;
662 rd_desc.count = 65536;
663 tcp_read_sock(sk, &rd_desc, tcp_data_recv);
664out:
665 read_unlock(&sk->sk_callback_lock);
666}
667
668static void
669tcp_state_change(struct sock *sk)
670{
671 struct rpc_xprt *xprt;
672
673 read_lock(&sk->sk_callback_lock);
674 if (!(xprt = xprt_from_sock(sk)))
675 goto out;
676 dprintk("RPC: tcp_state_change client %p...\n", xprt);
677 dprintk("RPC: state %x conn %d dead %d zapped %d\n",
678 sk->sk_state, xprt_connected(xprt),
679 sock_flag(sk, SOCK_DEAD),
680 sock_flag(sk, SOCK_ZAPPED));
681
682 switch (sk->sk_state) {
683 case TCP_ESTABLISHED:
684 spin_lock_bh(&xprt->sock_lock);
685 if (!xprt_test_and_set_connected(xprt)) {
686 /* Reset TCP record info */
687 xprt->tcp_offset = 0;
688 xprt->tcp_reclen = 0;
689 xprt->tcp_copied = 0;
690 xprt->tcp_flags = XPRT_COPY_RECM | XPRT_COPY_XID;
691 rpc_wake_up(&xprt->pending);
692 }
693 spin_unlock_bh(&xprt->sock_lock);
694 break;
695 case TCP_SYN_SENT:
696 case TCP_SYN_RECV:
697 break;
698 default:
699 xprt_disconnect(xprt);
700 break;
701 }
702 out:
703 read_unlock(&sk->sk_callback_lock);
704}
705
706/*
707 * Called when more output buffer space is available for this socket.
708 * We try not to wake our writers until they can make "significant"
709 * progress, otherwise we'll waste resources thrashing sock_sendmsg
710 * with a bunch of small requests.
711 */
712static void
713xprt_write_space(struct sock *sk)
714{
715 struct rpc_xprt *xprt;
716 struct socket *sock;
717
718 read_lock(&sk->sk_callback_lock);
719 if (!(xprt = xprt_from_sock(sk)) || !(sock = sk->sk_socket))
720 goto out;
721 if (xprt->shutdown)
722 goto out;
723
724 /* Wait until we have enough socket memory */
725 if (xprt->stream) {
726 /* from net/core/stream.c:sk_stream_write_space */
727 if (sk_stream_wspace(sk) < sk_stream_min_wspace(sk))
728 goto out;
729 } else {
730 /* from net/core/sock.c:sock_def_write_space */
731 if (!sock_writeable(sk))
732 goto out;
733 }
734
735 if (!test_and_clear_bit(SOCK_NOSPACE, &sock->flags))
736 goto out;
737
738 spin_lock_bh(&xprt->sock_lock);
739 if (xprt->snd_task)
740 rpc_wake_up_task(xprt->snd_task);
741 spin_unlock_bh(&xprt->sock_lock);
742out:
743 read_unlock(&sk->sk_callback_lock);
744}
745
746/*
747 * Set socket buffer length
748 */
749static void
750xprt_sock_setbufsize(struct rpc_xprt *xprt)
751{
752 struct sock *sk = xprt->inet;
753
754 if (xprt->stream)
755 return;
756 if (xprt->rcvsize) {
757 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
758 sk->sk_rcvbuf = xprt->rcvsize * xprt->max_reqs * 2;
759 }
760 if (xprt->sndsize) {
761 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
762 sk->sk_sndbuf = xprt->sndsize * xprt->max_reqs * 2;
763 sk->sk_write_space(sk);
764 }
765}
766
767/*
768 * Bind to a reserved port
769 */
770static inline int xprt_bindresvport(struct rpc_xprt *xprt, struct socket *sock)
771{
772 struct sockaddr_in myaddr = {
773 .sin_family = AF_INET,
774 };
775 int err, port;
776
777 /* Were we already bound to a given port? Try to reuse it */
778 port = xprt->port;
779 do {
780 myaddr.sin_port = htons(port);
781 err = sock->ops->bind(sock, (struct sockaddr *) &myaddr,
782 sizeof(myaddr));
783 if (err == 0) {
784 xprt->port = port;
785 return 0;
786 }
787 if (--port == 0)
788 port = XPRT_MAX_RESVPORT;
789 } while (err == -EADDRINUSE && port != xprt->port);
790
791 printk("RPC: Can't bind to reserved port (%d).\n", -err);
792 return err;
793}
794
795static void
796xprt_bind_socket(struct rpc_xprt *xprt, struct socket *sock)
797{
798 struct sock *sk = sock->sk;
799
800 if (xprt->inet)
801 return;
802
803 write_lock_bh(&sk->sk_callback_lock);
804 sk->sk_user_data = xprt;
805 xprt->old_data_ready = sk->sk_data_ready;
806 xprt->old_state_change = sk->sk_state_change;
807 xprt->old_write_space = sk->sk_write_space;
808 if (xprt->prot == IPPROTO_UDP) {
809 sk->sk_data_ready = udp_data_ready;
810 sk->sk_no_check = UDP_CSUM_NORCV;
811 xprt_set_connected(xprt);
812 } else {
813 tcp_sk(sk)->nonagle = 1; /* disable Nagle's algorithm */
814 sk->sk_data_ready = tcp_data_ready;
815 sk->sk_state_change = tcp_state_change;
816 xprt_clear_connected(xprt);
817 }
818 sk->sk_write_space = xprt_write_space;
819
820 /* Reset to new socket */
821 xprt->sock = sock;
822 xprt->inet = sk;
823 write_unlock_bh(&sk->sk_callback_lock);
824
825 return;
826}
827
828/*
829 * Datastream sockets are created here, but xprt_connect will create
830 * and connect stream sockets.
831 */
832static struct socket * xprt_create_socket(struct rpc_xprt *xprt, int proto, int resvport)
833{
834 struct socket *sock;
835 int type, err;
836
837 dprintk("RPC: xprt_create_socket(%s %d)\n",
838 (proto == IPPROTO_UDP)? "udp" : "tcp", proto);
839
840 type = (proto == IPPROTO_UDP)? SOCK_DGRAM : SOCK_STREAM;
841
842 if ((err = sock_create_kern(PF_INET, type, proto, &sock)) < 0) {
843 printk("RPC: can't create socket (%d).\n", -err);
844 return NULL;
845 }
846
847 /* If the caller has the capability, bind to a reserved port */
848 if (resvport && xprt_bindresvport(xprt, sock) < 0) {
849 printk("RPC: can't bind to reserved port.\n");
850 goto failed;
851 }
852
853 return sock;
854
855failed:
856 sock_release(sock);
857 return NULL;
858}
859
860static void xprt_socket_connect(void *args)
861{
862 struct rpc_xprt *xprt = (struct rpc_xprt *)args;
863 struct socket *sock = xprt->sock;
864 int status = -EIO;
865
866 if (xprt->shutdown || xprt->addr.sin_port == 0)
867 goto out;
868
869 /*
870 * Start by resetting any existing state
871 */
872 xprt_close(xprt);
873 sock = xprt_create_socket(xprt, xprt->prot, xprt->resvport);
874 if (sock == NULL) {
875 /* couldn't create socket or bind to reserved port;
876 * this is likely a permanent error, so cause an abort */
877 goto out;
878 }
879 xprt_bind_socket(xprt, sock);
880 xprt_sock_setbufsize(xprt);
881
882 status = 0;
883 if (!xprt->stream)
884 goto out;
885
886 /*
887 * Tell the socket layer to start connecting...
888 */
889 status = sock->ops->connect(sock, (struct sockaddr *) &xprt->addr,
890 sizeof(xprt->addr), O_NONBLOCK);
891 dprintk("RPC: %p connect status %d connected %d sock state %d\n",
892 xprt, -status, xprt_connected(xprt), sock->sk->sk_state);
893 if (status < 0) {
894 switch (status) {
895 case -EINPROGRESS:
896 case -EALREADY:
897 goto out_clear;
898 }
899 }
900out:
901 if (status < 0)
902 rpc_wake_up_status(&xprt->pending, status);
903 else
904 rpc_wake_up(&xprt->pending);
905out_clear:
906 smp_mb__before_clear_bit();
907 clear_bit(XPRT_CONNECTING, &xprt->sockstate);
908 smp_mb__after_clear_bit();
909}
910
911static void
912xprt_connect_sock(struct rpc_task *task)
913{
914 struct rpc_xprt *xprt = task->tk_xprt;
915
916 if (!test_and_set_bit(XPRT_CONNECTING, &xprt->sockstate)) {
917 /* Note: if we are here due to a dropped connection
918 * we delay reconnecting by RPC_REESTABLISH_TIMEOUT/HZ
919 * seconds
920 */
921 if (xprt->sock != NULL)
922 schedule_delayed_work(&xprt->sock_connect,
923 RPC_REESTABLISH_TIMEOUT);
924 else {
925 schedule_work(&xprt->sock_connect);
926 /* flush_scheduled_work can sleep... */
927 if (!RPC_IS_ASYNC(task))
928 flush_scheduled_work();
929 }
930 }
931}
932
933/*
934 * Set default timeout parameters
935 */
936static void
937xprt_default_timeout(struct rpc_timeout *to, int proto)
938{
939 if (proto == IPPROTO_UDP)
940 xprt_set_timeout(to, 5, 5 * HZ);
941 else
942 xprt_set_timeout(to, 2, 60 * HZ);
943}
944
945static struct rpc_xprt_ops xprt_socket_ops = {
946 .set_buffer_size = xprt_sock_setbufsize,
947 .connect = xprt_connect_sock,
948 .send_request = xprt_send_request,
949 .close = xprt_close,
950 .destroy = xprt_socket_destroy,
951};
952
953extern unsigned int xprt_udp_slot_table_entries;
954extern unsigned int xprt_tcp_slot_table_entries;
955
956int xs_setup_udp(struct rpc_xprt *xprt, struct rpc_timeout *to)
957{
958 size_t slot_table_size;
959
960 dprintk("RPC: setting up udp-ipv4 transport...\n");
961
962 xprt->max_reqs = xprt_udp_slot_table_entries;
963 slot_table_size = xprt->max_reqs * sizeof(xprt->slot[0]);
964 xprt->slot = kmalloc(slot_table_size, GFP_KERNEL);
965 if (xprt->slot == NULL)
966 return -ENOMEM;
967 memset(xprt->slot, 0, slot_table_size);
968
969 xprt->prot = IPPROTO_UDP;
970 xprt->port = XPRT_MAX_RESVPORT;
971 xprt->stream = 0;
972 xprt->nocong = 0;
973 xprt->cwnd = RPC_INITCWND;
974 xprt->resvport = capable(CAP_NET_BIND_SERVICE) ? 1 : 0;
975 /* XXX: header size can vary due to auth type, IPv6, etc. */
976 xprt->max_payload = (1U << 16) - (MAX_HEADER << 3);
977
978 INIT_WORK(&xprt->sock_connect, xprt_socket_connect, xprt);
979
980 xprt->ops = &xprt_socket_ops;
981
982 if (to)
983 xprt->timeout = *to;
984 else
985 xprt_default_timeout(to, xprt->prot);
986
987 return 0;
988}
989
990int xs_setup_tcp(struct rpc_xprt *xprt, struct rpc_timeout *to)
991{
992 size_t slot_table_size;
993
994 dprintk("RPC: setting up tcp-ipv4 transport...\n");
995
996 xprt->max_reqs = xprt_tcp_slot_table_entries;
997 slot_table_size = xprt->max_reqs * sizeof(xprt->slot[0]);
998 xprt->slot = kmalloc(slot_table_size, GFP_KERNEL);
999 if (xprt->slot == NULL)
1000 return -ENOMEM;
1001 memset(xprt->slot, 0, slot_table_size);
1002
1003 xprt->prot = IPPROTO_TCP;
1004 xprt->port = XPRT_MAX_RESVPORT;
1005 xprt->stream = 1;
1006 xprt->nocong = 1;
1007 xprt->cwnd = RPC_MAXCWND(xprt);
1008 xprt->resvport = capable(CAP_NET_BIND_SERVICE) ? 1 : 0;
1009 xprt->max_payload = (1U << 31) - 1;
1010
1011 INIT_WORK(&xprt->sock_connect, xprt_socket_connect, xprt);
1012
1013 xprt->ops = &xprt_socket_ops;
1014
1015 if (to)
1016 xprt->timeout = *to;
1017 else
1018 xprt_default_timeout(to, xprt->prot);
1019
1020 return 0;
1021}
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-25 12:09:33 -0500