blob: bf005d3c65efbffb46f5f05967de3806d9ac9b7d [file] [log] [blame]
Chuck Levera246b012005-08-11 16:25:23 -04001/*
2 * linux/net/sunrpc/xprtsock.c
3 *
4 * Client-side transport implementation for sockets.
5 *
Alan Cox113aa832008-10-13 19:01:08 -07006 * TCP callback races fixes (C) 1998 Red Hat
7 * TCP send fixes (C) 1998 Red Hat
Chuck Levera246b012005-08-11 16:25:23 -04008 * 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>
Chuck Lever55aa4f52005-08-11 16:25:47 -040014 *
15 * IP socket transport implementation, (C) 2005 Chuck Lever <cel@netapp.com>
Chuck Lever8f9d5b12007-08-06 11:57:53 -040016 *
17 * IPv6 support contributed by Gilles Quillard, Bull Open Source, 2005.
18 * <gilles.quillard@bull.net>
Chuck Levera246b012005-08-11 16:25:23 -040019 */
20
21#include <linux/types.h>
22#include <linux/slab.h>
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -040023#include <linux/module.h>
Chuck Levera246b012005-08-11 16:25:23 -040024#include <linux/capability.h>
Chuck Levera246b012005-08-11 16:25:23 -040025#include <linux/pagemap.h>
26#include <linux/errno.h>
27#include <linux/socket.h>
28#include <linux/in.h>
29#include <linux/net.h>
30#include <linux/mm.h>
31#include <linux/udp.h>
32#include <linux/tcp.h>
33#include <linux/sunrpc/clnt.h>
Chuck Lever02107142006-01-03 09:55:49 +010034#include <linux/sunrpc/sched.h>
Rahul Iyer4cfc7e62009-09-10 17:32:28 +030035#include <linux/sunrpc/svcsock.h>
\"Talpey, Thomas\49c36fc2007-09-10 13:47:31 -040036#include <linux/sunrpc/xprtsock.h>
Chuck Levera246b012005-08-11 16:25:23 -040037#include <linux/file.h>
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -040038#ifdef CONFIG_NFS_V4_1
39#include <linux/sunrpc/bc_xprt.h>
40#endif
Chuck Levera246b012005-08-11 16:25:23 -040041
42#include <net/sock.h>
43#include <net/checksum.h>
44#include <net/udp.h>
45#include <net/tcp.h>
46
Rahul Iyer4cfc7e62009-09-10 17:32:28 +030047#include "sunrpc.h"
Chuck Lever9903cd12005-08-11 16:25:26 -040048/*
Chuck Leverc556b752005-11-01 12:24:48 -050049 * xprtsock tunables
50 */
51unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE;
52unsigned int xprt_tcp_slot_table_entries = RPC_DEF_SLOT_TABLE;
53
54unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT;
55unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT;
56
Trond Myklebust7d1e8252009-03-11 14:38:03 -040057#define XS_TCP_LINGER_TO (15U * HZ)
Trond Myklebust25fe6142009-03-11 14:38:03 -040058static unsigned int xs_tcp_fin_timeout __read_mostly = XS_TCP_LINGER_TO;
Trond Myklebust7d1e8252009-03-11 14:38:03 -040059
Chuck Leverc556b752005-11-01 12:24:48 -050060/*
Chuck Leverfbf76682006-12-05 16:35:54 -050061 * We can register our own files under /proc/sys/sunrpc by
62 * calling register_sysctl_table() again. The files in that
63 * directory become the union of all files registered there.
64 *
65 * We simply need to make sure that we don't collide with
66 * someone else's file names!
67 */
68
69#ifdef RPC_DEBUG
70
71static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE;
72static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE;
73static unsigned int xprt_min_resvport_limit = RPC_MIN_RESVPORT;
74static unsigned int xprt_max_resvport_limit = RPC_MAX_RESVPORT;
75
76static struct ctl_table_header *sunrpc_table_header;
77
78/*
79 * FIXME: changing the UDP slot table size should also resize the UDP
80 * socket buffers for existing UDP transports
81 */
82static ctl_table xs_tunables_table[] = {
83 {
Chuck Leverfbf76682006-12-05 16:35:54 -050084 .procname = "udp_slot_table_entries",
85 .data = &xprt_udp_slot_table_entries,
86 .maxlen = sizeof(unsigned int),
87 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -080088 .proc_handler = proc_dointvec_minmax,
Chuck Leverfbf76682006-12-05 16:35:54 -050089 .extra1 = &min_slot_table_size,
90 .extra2 = &max_slot_table_size
91 },
92 {
Chuck Leverfbf76682006-12-05 16:35:54 -050093 .procname = "tcp_slot_table_entries",
94 .data = &xprt_tcp_slot_table_entries,
95 .maxlen = sizeof(unsigned int),
96 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -080097 .proc_handler = proc_dointvec_minmax,
Chuck Leverfbf76682006-12-05 16:35:54 -050098 .extra1 = &min_slot_table_size,
99 .extra2 = &max_slot_table_size
100 },
101 {
Chuck Leverfbf76682006-12-05 16:35:54 -0500102 .procname = "min_resvport",
103 .data = &xprt_min_resvport,
104 .maxlen = sizeof(unsigned int),
105 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -0800106 .proc_handler = proc_dointvec_minmax,
Chuck Leverfbf76682006-12-05 16:35:54 -0500107 .extra1 = &xprt_min_resvport_limit,
108 .extra2 = &xprt_max_resvport_limit
109 },
110 {
Chuck Leverfbf76682006-12-05 16:35:54 -0500111 .procname = "max_resvport",
112 .data = &xprt_max_resvport,
113 .maxlen = sizeof(unsigned int),
114 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -0800115 .proc_handler = proc_dointvec_minmax,
Chuck Leverfbf76682006-12-05 16:35:54 -0500116 .extra1 = &xprt_min_resvport_limit,
117 .extra2 = &xprt_max_resvport_limit
118 },
119 {
Trond Myklebust25fe6142009-03-11 14:38:03 -0400120 .procname = "tcp_fin_timeout",
121 .data = &xs_tcp_fin_timeout,
122 .maxlen = sizeof(xs_tcp_fin_timeout),
123 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -0800124 .proc_handler = proc_dointvec_jiffies,
Trond Myklebust25fe6142009-03-11 14:38:03 -0400125 },
Eric W. Biedermanf8572d82009-11-05 13:32:03 -0800126 { },
Chuck Leverfbf76682006-12-05 16:35:54 -0500127};
128
129static ctl_table sunrpc_table[] = {
130 {
Chuck Leverfbf76682006-12-05 16:35:54 -0500131 .procname = "sunrpc",
132 .mode = 0555,
133 .child = xs_tunables_table
134 },
Eric W. Biedermanf8572d82009-11-05 13:32:03 -0800135 { },
Chuck Leverfbf76682006-12-05 16:35:54 -0500136};
137
138#endif
139
140/*
Chuck Lever03bf4b72005-08-25 16:25:55 -0700141 * Wait duration for a reply from the RPC portmapper.
142 */
143#define XS_BIND_TO (60U * HZ)
144
145/*
146 * Delay if a UDP socket connect error occurs. This is most likely some
147 * kind of resource problem on the local host.
148 */
149#define XS_UDP_REEST_TO (2U * HZ)
150
151/*
152 * The reestablish timeout allows clients to delay for a bit before attempting
153 * to reconnect to a server that just dropped our connection.
154 *
155 * We implement an exponential backoff when trying to reestablish a TCP
156 * transport connection with the server. Some servers like to drop a TCP
157 * connection when they are overworked, so we start with a short timeout and
158 * increase over time if the server is down or not responding.
159 */
160#define XS_TCP_INIT_REEST_TO (3U * HZ)
161#define XS_TCP_MAX_REEST_TO (5U * 60 * HZ)
162
163/*
164 * TCP idle timeout; client drops the transport socket if it is idle
165 * for this long. Note that we also timeout UDP sockets to prevent
166 * holding port numbers when there is no RPC traffic.
167 */
168#define XS_IDLE_DISC_TO (5U * 60 * HZ)
169
Chuck Levera246b012005-08-11 16:25:23 -0400170#ifdef RPC_DEBUG
171# undef RPC_DEBUG_DATA
Chuck Lever9903cd12005-08-11 16:25:26 -0400172# define RPCDBG_FACILITY RPCDBG_TRANS
Chuck Levera246b012005-08-11 16:25:23 -0400173#endif
174
Chuck Levera246b012005-08-11 16:25:23 -0400175#ifdef RPC_DEBUG_DATA
Chuck Lever9903cd12005-08-11 16:25:26 -0400176static void xs_pktdump(char *msg, u32 *packet, unsigned int count)
Chuck Levera246b012005-08-11 16:25:23 -0400177{
Chuck Lever9903cd12005-08-11 16:25:26 -0400178 u8 *buf = (u8 *) packet;
179 int j;
Chuck Levera246b012005-08-11 16:25:23 -0400180
Chuck Lever46121cf2007-01-31 12:14:08 -0500181 dprintk("RPC: %s\n", msg);
Chuck Levera246b012005-08-11 16:25:23 -0400182 for (j = 0; j < count && j < 128; j += 4) {
183 if (!(j & 31)) {
184 if (j)
185 dprintk("\n");
186 dprintk("0x%04x ", j);
187 }
188 dprintk("%02x%02x%02x%02x ",
189 buf[j], buf[j+1], buf[j+2], buf[j+3]);
190 }
191 dprintk("\n");
192}
193#else
Chuck Lever9903cd12005-08-11 16:25:26 -0400194static inline void xs_pktdump(char *msg, u32 *packet, unsigned int count)
Chuck Levera246b012005-08-11 16:25:23 -0400195{
196 /* NOP */
197}
198#endif
199
Chuck Leverffc2e512006-12-05 16:35:11 -0500200struct sock_xprt {
201 struct rpc_xprt xprt;
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500202
203 /*
204 * Network layer
205 */
206 struct socket * sock;
207 struct sock * inet;
Chuck Lever51971132006-12-05 16:35:19 -0500208
209 /*
210 * State of TCP reply receive
211 */
212 __be32 tcp_fraghdr,
Trond Myklebustb76ce562010-06-16 13:57:32 -0400213 tcp_xid,
214 tcp_calldir;
Chuck Lever51971132006-12-05 16:35:19 -0500215
216 u32 tcp_offset,
217 tcp_reclen;
218
219 unsigned long tcp_copied,
220 tcp_flags;
Chuck Leverc8475462006-12-05 16:35:26 -0500221
222 /*
223 * Connection of transports
224 */
Trond Myklebust34161db2006-12-07 15:48:15 -0500225 struct delayed_work connect_worker;
Chuck Leverfbfffbd2009-08-09 15:09:46 -0400226 struct sockaddr_storage srcaddr;
227 unsigned short srcport;
Chuck Lever7c6e0662006-12-05 16:35:30 -0500228
229 /*
230 * UDP socket buffer size parameters
231 */
232 size_t rcvsize,
233 sndsize;
Chuck Lever314dfd72006-12-05 16:35:34 -0500234
235 /*
236 * Saved socket callback addresses
237 */
238 void (*old_data_ready)(struct sock *, int);
239 void (*old_state_change)(struct sock *);
240 void (*old_write_space)(struct sock *);
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -0400241 void (*old_error_report)(struct sock *);
Chuck Leverffc2e512006-12-05 16:35:11 -0500242};
243
Chuck Levere136d092006-12-05 16:35:23 -0500244/*
245 * TCP receive state flags
246 */
247#define TCP_RCV_LAST_FRAG (1UL << 0)
248#define TCP_RCV_COPY_FRAGHDR (1UL << 1)
249#define TCP_RCV_COPY_XID (1UL << 2)
250#define TCP_RCV_COPY_DATA (1UL << 3)
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -0400251#define TCP_RCV_READ_CALLDIR (1UL << 4)
252#define TCP_RCV_COPY_CALLDIR (1UL << 5)
Ricardo Labiaga18dca022009-04-01 09:22:53 -0400253
254/*
255 * TCP RPC flags
256 */
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -0400257#define TCP_RPC_REPLY (1UL << 6)
Chuck Levere136d092006-12-05 16:35:23 -0500258
Chuck Lever95392c52007-08-06 11:57:58 -0400259static inline struct sockaddr *xs_addr(struct rpc_xprt *xprt)
Chuck Leveredb267a2006-08-22 20:06:18 -0400260{
Chuck Lever95392c52007-08-06 11:57:58 -0400261 return (struct sockaddr *) &xprt->addr;
262}
263
264static inline struct sockaddr_in *xs_addr_in(struct rpc_xprt *xprt)
265{
266 return (struct sockaddr_in *) &xprt->addr;
267}
268
269static inline struct sockaddr_in6 *xs_addr_in6(struct rpc_xprt *xprt)
270{
271 return (struct sockaddr_in6 *) &xprt->addr;
272}
273
Chuck Leverc877b842009-08-09 15:09:36 -0400274static void xs_format_common_peer_addresses(struct rpc_xprt *xprt)
Chuck Leveredb267a2006-08-22 20:06:18 -0400275{
Chuck Leverc877b842009-08-09 15:09:36 -0400276 struct sockaddr *sap = xs_addr(xprt);
Chuck Lever9dc3b092009-08-09 15:09:46 -0400277 struct sockaddr_in6 *sin6;
278 struct sockaddr_in *sin;
Chuck Leverc877b842009-08-09 15:09:36 -0400279 char buf[128];
Chuck Leveredb267a2006-08-22 20:06:18 -0400280
Chuck Leverc877b842009-08-09 15:09:36 -0400281 (void)rpc_ntop(sap, buf, sizeof(buf));
282 xprt->address_strings[RPC_DISPLAY_ADDR] = kstrdup(buf, GFP_KERNEL);
283
Chuck Lever9dc3b092009-08-09 15:09:46 -0400284 switch (sap->sa_family) {
285 case AF_INET:
286 sin = xs_addr_in(xprt);
Joe Perchesfc0b57912010-03-08 12:15:28 -0800287 snprintf(buf, sizeof(buf), "%08x", ntohl(sin->sin_addr.s_addr));
Chuck Lever9dc3b092009-08-09 15:09:46 -0400288 break;
289 case AF_INET6:
290 sin6 = xs_addr_in6(xprt);
Joe Perchesfc0b57912010-03-08 12:15:28 -0800291 snprintf(buf, sizeof(buf), "%pi6", &sin6->sin6_addr);
Chuck Lever9dc3b092009-08-09 15:09:46 -0400292 break;
293 default:
294 BUG();
Chuck Leveredb267a2006-08-22 20:06:18 -0400295 }
Chuck Lever9dc3b092009-08-09 15:09:46 -0400296 xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = kstrdup(buf, GFP_KERNEL);
Chuck Leveredb267a2006-08-22 20:06:18 -0400297}
298
Chuck Lever9dc3b092009-08-09 15:09:46 -0400299static void xs_format_common_peer_ports(struct rpc_xprt *xprt)
Chuck Lever4b6473f2007-08-06 11:57:12 -0400300{
Chuck Lever9dc3b092009-08-09 15:09:46 -0400301 struct sockaddr *sap = xs_addr(xprt);
302 char buf[128];
Chuck Lever4b6473f2007-08-06 11:57:12 -0400303
Joe Perches81160e662010-03-08 12:15:59 -0800304 snprintf(buf, sizeof(buf), "%u", rpc_get_port(sap));
Chuck Leverc877b842009-08-09 15:09:36 -0400305 xprt->address_strings[RPC_DISPLAY_PORT] = kstrdup(buf, GFP_KERNEL);
Chuck Lever4b6473f2007-08-06 11:57:12 -0400306
Joe Perches81160e662010-03-08 12:15:59 -0800307 snprintf(buf, sizeof(buf), "%4hx", rpc_get_port(sap));
Chuck Leverc877b842009-08-09 15:09:36 -0400308 xprt->address_strings[RPC_DISPLAY_HEX_PORT] = kstrdup(buf, GFP_KERNEL);
309}
Chuck Lever4b6473f2007-08-06 11:57:12 -0400310
Chuck Lever9dc3b092009-08-09 15:09:46 -0400311static void xs_format_peer_addresses(struct rpc_xprt *xprt,
312 const char *protocol,
313 const char *netid)
Chuck Leveredb267a2006-08-22 20:06:18 -0400314{
Chuck Leverb454ae92008-01-07 18:34:48 -0500315 xprt->address_strings[RPC_DISPLAY_PROTO] = protocol;
Chuck Leverb454ae92008-01-07 18:34:48 -0500316 xprt->address_strings[RPC_DISPLAY_NETID] = netid;
Chuck Leverc877b842009-08-09 15:09:36 -0400317 xs_format_common_peer_addresses(xprt);
Chuck Lever9dc3b092009-08-09 15:09:46 -0400318 xs_format_common_peer_ports(xprt);
Chuck Lever4b6473f2007-08-06 11:57:12 -0400319}
320
Chuck Lever9dc3b092009-08-09 15:09:46 -0400321static void xs_update_peer_port(struct rpc_xprt *xprt)
Chuck Lever4b6473f2007-08-06 11:57:12 -0400322{
Chuck Lever9dc3b092009-08-09 15:09:46 -0400323 kfree(xprt->address_strings[RPC_DISPLAY_HEX_PORT]);
324 kfree(xprt->address_strings[RPC_DISPLAY_PORT]);
Chuck Lever4b6473f2007-08-06 11:57:12 -0400325
Chuck Lever9dc3b092009-08-09 15:09:46 -0400326 xs_format_common_peer_ports(xprt);
Chuck Leveredb267a2006-08-22 20:06:18 -0400327}
328
329static void xs_free_peer_addresses(struct rpc_xprt *xprt)
330{
Chuck Lever33e01dc2008-01-14 12:32:20 -0500331 unsigned int i;
332
333 for (i = 0; i < RPC_DISPLAY_MAX; i++)
334 switch (i) {
335 case RPC_DISPLAY_PROTO:
336 case RPC_DISPLAY_NETID:
337 continue;
338 default:
339 kfree(xprt->address_strings[i]);
340 }
Chuck Leveredb267a2006-08-22 20:06:18 -0400341}
342
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400343#define XS_SENDMSG_FLAGS (MSG_DONTWAIT | MSG_NOSIGNAL)
344
Trond Myklebust24c56842006-10-17 15:06:22 -0400345static int xs_send_kvec(struct socket *sock, struct sockaddr *addr, int addrlen, struct kvec *vec, unsigned int base, int more)
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400346{
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400347 struct msghdr msg = {
348 .msg_name = addr,
349 .msg_namelen = addrlen,
Trond Myklebust24c56842006-10-17 15:06:22 -0400350 .msg_flags = XS_SENDMSG_FLAGS | (more ? MSG_MORE : 0),
351 };
352 struct kvec iov = {
353 .iov_base = vec->iov_base + base,
354 .iov_len = vec->iov_len - base,
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400355 };
356
Trond Myklebust24c56842006-10-17 15:06:22 -0400357 if (iov.iov_len != 0)
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400358 return kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len);
359 return kernel_sendmsg(sock, &msg, NULL, 0, 0);
360}
361
Trond Myklebust24c56842006-10-17 15:06:22 -0400362static int xs_send_pagedata(struct socket *sock, struct xdr_buf *xdr, unsigned int base, int more)
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400363{
Trond Myklebust24c56842006-10-17 15:06:22 -0400364 struct page **ppage;
365 unsigned int remainder;
366 int err, sent = 0;
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400367
Trond Myklebust24c56842006-10-17 15:06:22 -0400368 remainder = xdr->page_len - base;
369 base += xdr->page_base;
370 ppage = xdr->pages + (base >> PAGE_SHIFT);
371 base &= ~PAGE_MASK;
372 for(;;) {
373 unsigned int len = min_t(unsigned int, PAGE_SIZE - base, remainder);
374 int flags = XS_SENDMSG_FLAGS;
375
376 remainder -= len;
377 if (remainder != 0 || more)
378 flags |= MSG_MORE;
379 err = sock->ops->sendpage(sock, *ppage, base, len, flags);
380 if (remainder == 0 || err != len)
381 break;
382 sent += err;
383 ppage++;
384 base = 0;
385 }
386 if (sent == 0)
387 return err;
388 if (err > 0)
389 sent += err;
390 return sent;
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400391}
392
Chuck Lever9903cd12005-08-11 16:25:26 -0400393/**
394 * xs_sendpages - write pages directly to a socket
395 * @sock: socket to send on
396 * @addr: UDP only -- address of destination
397 * @addrlen: UDP only -- length of destination address
398 * @xdr: buffer containing this request
399 * @base: starting position in the buffer
400 *
Chuck Levera246b012005-08-11 16:25:23 -0400401 */
Trond Myklebust24c56842006-10-17 15:06:22 -0400402static int xs_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen, struct xdr_buf *xdr, unsigned int base)
Chuck Levera246b012005-08-11 16:25:23 -0400403{
Trond Myklebust24c56842006-10-17 15:06:22 -0400404 unsigned int remainder = xdr->len - base;
405 int err, sent = 0;
Chuck Levera246b012005-08-11 16:25:23 -0400406
Chuck Lever262965f2005-08-11 16:25:56 -0400407 if (unlikely(!sock))
Trond Myklebustfba91af2009-03-11 14:06:41 -0400408 return -ENOTSOCK;
Chuck Lever262965f2005-08-11 16:25:56 -0400409
410 clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags);
Trond Myklebust24c56842006-10-17 15:06:22 -0400411 if (base != 0) {
412 addr = NULL;
413 addrlen = 0;
414 }
Chuck Lever262965f2005-08-11 16:25:56 -0400415
Trond Myklebust24c56842006-10-17 15:06:22 -0400416 if (base < xdr->head[0].iov_len || addr != NULL) {
417 unsigned int len = xdr->head[0].iov_len - base;
418 remainder -= len;
419 err = xs_send_kvec(sock, addr, addrlen, &xdr->head[0], base, remainder != 0);
420 if (remainder == 0 || err != len)
Chuck Levera246b012005-08-11 16:25:23 -0400421 goto out;
Trond Myklebust24c56842006-10-17 15:06:22 -0400422 sent += err;
Chuck Levera246b012005-08-11 16:25:23 -0400423 base = 0;
424 } else
Trond Myklebust24c56842006-10-17 15:06:22 -0400425 base -= xdr->head[0].iov_len;
Chuck Levera246b012005-08-11 16:25:23 -0400426
Trond Myklebust24c56842006-10-17 15:06:22 -0400427 if (base < xdr->page_len) {
428 unsigned int len = xdr->page_len - base;
429 remainder -= len;
430 err = xs_send_pagedata(sock, xdr, base, remainder != 0);
431 if (remainder == 0 || err != len)
Chuck Levera246b012005-08-11 16:25:23 -0400432 goto out;
Trond Myklebust24c56842006-10-17 15:06:22 -0400433 sent += err;
Chuck Levera246b012005-08-11 16:25:23 -0400434 base = 0;
Trond Myklebust24c56842006-10-17 15:06:22 -0400435 } else
436 base -= xdr->page_len;
437
438 if (base >= xdr->tail[0].iov_len)
439 return sent;
440 err = xs_send_kvec(sock, NULL, 0, &xdr->tail[0], base, 0);
Chuck Levera246b012005-08-11 16:25:23 -0400441out:
Trond Myklebust24c56842006-10-17 15:06:22 -0400442 if (sent == 0)
443 return err;
444 if (err > 0)
445 sent += err;
446 return sent;
Chuck Levera246b012005-08-11 16:25:23 -0400447}
448
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400449static void xs_nospace_callback(struct rpc_task *task)
450{
451 struct sock_xprt *transport = container_of(task->tk_rqstp->rq_xprt, struct sock_xprt, xprt);
452
453 transport->inet->sk_write_pending--;
454 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
455}
456
Chuck Lever9903cd12005-08-11 16:25:26 -0400457/**
Chuck Lever262965f2005-08-11 16:25:56 -0400458 * xs_nospace - place task on wait queue if transmit was incomplete
459 * @task: task to put to sleep
Chuck Lever9903cd12005-08-11 16:25:26 -0400460 *
Chuck Levera246b012005-08-11 16:25:23 -0400461 */
Trond Myklebust5e3771c2009-03-11 14:38:01 -0400462static int xs_nospace(struct rpc_task *task)
Chuck Levera246b012005-08-11 16:25:23 -0400463{
Chuck Lever262965f2005-08-11 16:25:56 -0400464 struct rpc_rqst *req = task->tk_rqstp;
465 struct rpc_xprt *xprt = req->rq_xprt;
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500466 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Trond Myklebust5e3771c2009-03-11 14:38:01 -0400467 int ret = 0;
Chuck Levera246b012005-08-11 16:25:23 -0400468
Chuck Lever46121cf2007-01-31 12:14:08 -0500469 dprintk("RPC: %5u xmit incomplete (%u left of %u)\n",
Chuck Lever262965f2005-08-11 16:25:56 -0400470 task->tk_pid, req->rq_slen - req->rq_bytes_sent,
471 req->rq_slen);
472
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400473 /* Protect against races with write_space */
474 spin_lock_bh(&xprt->transport_lock);
Chuck Lever262965f2005-08-11 16:25:56 -0400475
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400476 /* Don't race with disconnect */
477 if (xprt_connected(xprt)) {
478 if (test_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags)) {
Trond Myklebust5e3771c2009-03-11 14:38:01 -0400479 ret = -EAGAIN;
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400480 /*
481 * Notify TCP that we're limited by the application
482 * window size
483 */
484 set_bit(SOCK_NOSPACE, &transport->sock->flags);
485 transport->inet->sk_write_pending++;
486 /* ...and wait for more buffer space */
487 xprt_wait_for_buffer_space(task, xs_nospace_callback);
488 }
489 } else {
490 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
Trond Myklebust5e3771c2009-03-11 14:38:01 -0400491 ret = -ENOTCONN;
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400492 }
Chuck Lever262965f2005-08-11 16:25:56 -0400493
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400494 spin_unlock_bh(&xprt->transport_lock);
Trond Myklebust5e3771c2009-03-11 14:38:01 -0400495 return ret;
Chuck Lever262965f2005-08-11 16:25:56 -0400496}
497
498/**
499 * xs_udp_send_request - write an RPC request to a UDP socket
500 * @task: address of RPC task that manages the state of an RPC request
501 *
502 * Return values:
503 * 0: The request has been sent
504 * EAGAIN: The socket was blocked, please call again later to
505 * complete the request
506 * ENOTCONN: Caller needs to invoke connect logic then call again
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300507 * other: Some other error occurred, the request was not sent
Chuck Lever262965f2005-08-11 16:25:56 -0400508 */
509static int xs_udp_send_request(struct rpc_task *task)
510{
511 struct rpc_rqst *req = task->tk_rqstp;
512 struct rpc_xprt *xprt = req->rq_xprt;
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500513 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Lever262965f2005-08-11 16:25:56 -0400514 struct xdr_buf *xdr = &req->rq_snd_buf;
515 int status;
Chuck Levera246b012005-08-11 16:25:23 -0400516
Chuck Lever9903cd12005-08-11 16:25:26 -0400517 xs_pktdump("packet data:",
Chuck Levera246b012005-08-11 16:25:23 -0400518 req->rq_svec->iov_base,
519 req->rq_svec->iov_len);
520
Trond Myklebust01d37c42009-03-11 14:09:39 -0400521 if (!xprt_bound(xprt))
522 return -ENOTCONN;
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500523 status = xs_sendpages(transport->sock,
Chuck Lever95392c52007-08-06 11:57:58 -0400524 xs_addr(xprt),
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500525 xprt->addrlen, xdr,
526 req->rq_bytes_sent);
Chuck Levera246b012005-08-11 16:25:23 -0400527
Chuck Lever46121cf2007-01-31 12:14:08 -0500528 dprintk("RPC: xs_udp_send_request(%u) = %d\n",
Chuck Lever262965f2005-08-11 16:25:56 -0400529 xdr->len - req->rq_bytes_sent, status);
Chuck Levera246b012005-08-11 16:25:23 -0400530
Trond Myklebust21997002007-10-01 11:43:37 -0400531 if (status >= 0) {
Trond Myklebustd60dbb22010-05-13 12:51:49 -0400532 req->rq_xmit_bytes_sent += status;
Trond Myklebust21997002007-10-01 11:43:37 -0400533 if (status >= req->rq_slen)
534 return 0;
535 /* Still some bytes left; set up for a retry later. */
Chuck Lever262965f2005-08-11 16:25:56 -0400536 status = -EAGAIN;
Trond Myklebust21997002007-10-01 11:43:37 -0400537 }
Chuck Levera246b012005-08-11 16:25:23 -0400538
Chuck Lever262965f2005-08-11 16:25:56 -0400539 switch (status) {
Trond Myklebustfba91af2009-03-11 14:06:41 -0400540 case -ENOTSOCK:
541 status = -ENOTCONN;
542 /* Should we call xs_close() here? */
543 break;
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400544 case -EAGAIN:
Trond Myklebust5e3771c2009-03-11 14:38:01 -0400545 status = xs_nospace(task);
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400546 break;
Trond Myklebustc8485e42009-03-11 14:37:59 -0400547 default:
548 dprintk("RPC: sendmsg returned unrecognized error %d\n",
549 -status);
Chuck Lever262965f2005-08-11 16:25:56 -0400550 case -ENETUNREACH:
551 case -EPIPE:
Chuck Levera246b012005-08-11 16:25:23 -0400552 case -ECONNREFUSED:
553 /* When the server has died, an ICMP port unreachable message
Chuck Lever9903cd12005-08-11 16:25:26 -0400554 * prompts ECONNREFUSED. */
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400555 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
Chuck Levera246b012005-08-11 16:25:23 -0400556 }
Bian Naimeng5fe46e92010-03-08 14:49:01 +0800557
Chuck Lever262965f2005-08-11 16:25:56 -0400558 return status;
Chuck Levera246b012005-08-11 16:25:23 -0400559}
560
Trond Myklebuste06799f2007-11-05 15:44:12 -0500561/**
562 * xs_tcp_shutdown - gracefully shut down a TCP socket
563 * @xprt: transport
564 *
565 * Initiates a graceful shutdown of the TCP socket by calling the
566 * equivalent of shutdown(SHUT_WR);
567 */
568static void xs_tcp_shutdown(struct rpc_xprt *xprt)
569{
570 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
571 struct socket *sock = transport->sock;
572
573 if (sock != NULL)
574 kernel_sock_shutdown(sock, SHUT_WR);
575}
576
Chuck Lever808012f2005-08-25 16:25:49 -0700577static inline void xs_encode_tcp_record_marker(struct xdr_buf *buf)
578{
579 u32 reclen = buf->len - sizeof(rpc_fraghdr);
580 rpc_fraghdr *base = buf->head[0].iov_base;
581 *base = htonl(RPC_LAST_STREAM_FRAGMENT | reclen);
582}
583
Chuck Lever9903cd12005-08-11 16:25:26 -0400584/**
Chuck Lever262965f2005-08-11 16:25:56 -0400585 * xs_tcp_send_request - write an RPC request to a TCP socket
Chuck Lever9903cd12005-08-11 16:25:26 -0400586 * @task: address of RPC task that manages the state of an RPC request
587 *
588 * Return values:
Chuck Lever262965f2005-08-11 16:25:56 -0400589 * 0: The request has been sent
590 * EAGAIN: The socket was blocked, please call again later to
591 * complete the request
592 * ENOTCONN: Caller needs to invoke connect logic then call again
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300593 * other: Some other error occurred, the request was not sent
Chuck Lever9903cd12005-08-11 16:25:26 -0400594 *
595 * XXX: In the case of soft timeouts, should we eventually give up
Chuck Lever262965f2005-08-11 16:25:56 -0400596 * if sendmsg is not able to make progress?
Chuck Lever9903cd12005-08-11 16:25:26 -0400597 */
Chuck Lever262965f2005-08-11 16:25:56 -0400598static int xs_tcp_send_request(struct rpc_task *task)
Chuck Levera246b012005-08-11 16:25:23 -0400599{
600 struct rpc_rqst *req = task->tk_rqstp;
601 struct rpc_xprt *xprt = req->rq_xprt;
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500602 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Lever262965f2005-08-11 16:25:56 -0400603 struct xdr_buf *xdr = &req->rq_snd_buf;
Chuck Leverb595bb12007-08-06 11:56:42 -0400604 int status;
Chuck Levera246b012005-08-11 16:25:23 -0400605
Chuck Lever808012f2005-08-25 16:25:49 -0700606 xs_encode_tcp_record_marker(&req->rq_snd_buf);
Chuck Levera246b012005-08-11 16:25:23 -0400607
Chuck Lever262965f2005-08-11 16:25:56 -0400608 xs_pktdump("packet data:",
609 req->rq_svec->iov_base,
610 req->rq_svec->iov_len);
Chuck Levera246b012005-08-11 16:25:23 -0400611
612 /* Continue transmitting the packet/record. We must be careful
613 * to cope with writespace callbacks arriving _after_ we have
Chuck Lever262965f2005-08-11 16:25:56 -0400614 * called sendmsg(). */
Chuck Levera246b012005-08-11 16:25:23 -0400615 while (1) {
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500616 status = xs_sendpages(transport->sock,
617 NULL, 0, xdr, req->rq_bytes_sent);
Chuck Levera246b012005-08-11 16:25:23 -0400618
Chuck Lever46121cf2007-01-31 12:14:08 -0500619 dprintk("RPC: xs_tcp_send_request(%u) = %d\n",
Chuck Lever262965f2005-08-11 16:25:56 -0400620 xdr->len - req->rq_bytes_sent, status);
621
622 if (unlikely(status < 0))
Chuck Levera246b012005-08-11 16:25:23 -0400623 break;
624
Chuck Lever262965f2005-08-11 16:25:56 -0400625 /* If we've sent the entire packet, immediately
626 * reset the count of bytes sent. */
627 req->rq_bytes_sent += status;
Trond Myklebustd60dbb22010-05-13 12:51:49 -0400628 req->rq_xmit_bytes_sent += status;
Chuck Lever262965f2005-08-11 16:25:56 -0400629 if (likely(req->rq_bytes_sent >= req->rq_slen)) {
630 req->rq_bytes_sent = 0;
631 return 0;
Chuck Levera246b012005-08-11 16:25:23 -0400632 }
633
Trond Myklebust06b4b682008-04-16 16:51:38 -0400634 if (status != 0)
635 continue;
Chuck Levera246b012005-08-11 16:25:23 -0400636 status = -EAGAIN;
Trond Myklebust06b4b682008-04-16 16:51:38 -0400637 break;
Chuck Levera246b012005-08-11 16:25:23 -0400638 }
639
Chuck Lever262965f2005-08-11 16:25:56 -0400640 switch (status) {
Trond Myklebustfba91af2009-03-11 14:06:41 -0400641 case -ENOTSOCK:
642 status = -ENOTCONN;
643 /* Should we call xs_close() here? */
644 break;
Chuck Lever262965f2005-08-11 16:25:56 -0400645 case -EAGAIN:
Trond Myklebust5e3771c2009-03-11 14:38:01 -0400646 status = xs_nospace(task);
Chuck Lever262965f2005-08-11 16:25:56 -0400647 break;
648 default:
Chuck Lever46121cf2007-01-31 12:14:08 -0500649 dprintk("RPC: sendmsg returned unrecognized error %d\n",
Chuck Lever262965f2005-08-11 16:25:56 -0400650 -status);
Chuck Lever262965f2005-08-11 16:25:56 -0400651 case -ECONNRESET:
Trond Myklebust55420c22009-03-11 15:29:24 -0400652 case -EPIPE:
Trond Myklebuste06799f2007-11-05 15:44:12 -0500653 xs_tcp_shutdown(xprt);
Chuck Lever262965f2005-08-11 16:25:56 -0400654 case -ECONNREFUSED:
655 case -ENOTCONN:
Chuck Lever262965f2005-08-11 16:25:56 -0400656 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
Chuck Levera246b012005-08-11 16:25:23 -0400657 }
Bian Naimeng5fe46e92010-03-08 14:49:01 +0800658
Chuck Levera246b012005-08-11 16:25:23 -0400659 return status;
660}
661
Chuck Lever9903cd12005-08-11 16:25:26 -0400662/**
Trond Myklebuste0ab53d2006-07-27 17:22:50 -0400663 * xs_tcp_release_xprt - clean up after a tcp transmission
664 * @xprt: transport
665 * @task: rpc task
666 *
667 * This cleans up if an error causes us to abort the transmission of a request.
668 * In this case, the socket may need to be reset in order to avoid confusing
669 * the server.
670 */
671static void xs_tcp_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
672{
673 struct rpc_rqst *req;
674
675 if (task != xprt->snd_task)
676 return;
677 if (task == NULL)
678 goto out_release;
679 req = task->tk_rqstp;
680 if (req->rq_bytes_sent == 0)
681 goto out_release;
682 if (req->rq_bytes_sent == req->rq_snd_buf.len)
683 goto out_release;
684 set_bit(XPRT_CLOSE_WAIT, &task->tk_xprt->state);
685out_release:
686 xprt_release_xprt(xprt, task);
687}
688
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -0400689static void xs_save_old_callbacks(struct sock_xprt *transport, struct sock *sk)
690{
691 transport->old_data_ready = sk->sk_data_ready;
692 transport->old_state_change = sk->sk_state_change;
693 transport->old_write_space = sk->sk_write_space;
694 transport->old_error_report = sk->sk_error_report;
695}
696
697static void xs_restore_old_callbacks(struct sock_xprt *transport, struct sock *sk)
698{
699 sk->sk_data_ready = transport->old_data_ready;
700 sk->sk_state_change = transport->old_state_change;
701 sk->sk_write_space = transport->old_write_space;
702 sk->sk_error_report = transport->old_error_report;
703}
704
Chuck Leverfe315e72009-03-11 14:10:21 -0400705static void xs_reset_transport(struct sock_xprt *transport)
Chuck Levera246b012005-08-11 16:25:23 -0400706{
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500707 struct socket *sock = transport->sock;
708 struct sock *sk = transport->inet;
Chuck Levera246b012005-08-11 16:25:23 -0400709
Chuck Leverfe315e72009-03-11 14:10:21 -0400710 if (sk == NULL)
711 return;
Chuck Lever9903cd12005-08-11 16:25:26 -0400712
Trond Myklebust246408d2011-03-22 18:40:10 -0400713 transport->srcport = 0;
714
Chuck Levera246b012005-08-11 16:25:23 -0400715 write_lock_bh(&sk->sk_callback_lock);
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500716 transport->inet = NULL;
717 transport->sock = NULL;
Chuck Levera246b012005-08-11 16:25:23 -0400718
Chuck Lever9903cd12005-08-11 16:25:26 -0400719 sk->sk_user_data = NULL;
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -0400720
721 xs_restore_old_callbacks(transport, sk);
Chuck Levera246b012005-08-11 16:25:23 -0400722 write_unlock_bh(&sk->sk_callback_lock);
723
Chuck Lever9903cd12005-08-11 16:25:26 -0400724 sk->sk_no_check = 0;
Chuck Levera246b012005-08-11 16:25:23 -0400725
726 sock_release(sock);
Chuck Leverfe315e72009-03-11 14:10:21 -0400727}
728
729/**
730 * xs_close - close a socket
731 * @xprt: transport
732 *
733 * This is used when all requests are complete; ie, no DRC state remains
734 * on the server we want to save.
Trond Myklebustf75e6742009-04-21 17:18:20 -0400735 *
736 * The caller _must_ be holding XPRT_LOCKED in order to avoid issues with
737 * xs_reset_transport() zeroing the socket from underneath a writer.
Chuck Leverfe315e72009-03-11 14:10:21 -0400738 */
739static void xs_close(struct rpc_xprt *xprt)
740{
741 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
742
743 dprintk("RPC: xs_close xprt %p\n", xprt);
744
745 xs_reset_transport(transport);
Neil Brown61d0a8e2009-09-23 14:36:37 -0400746 xprt->reestablish_timeout = 0;
Chuck Leverfe315e72009-03-11 14:10:21 -0400747
Trond Myklebust632e3bd2006-01-03 09:55:55 +0100748 smp_mb__before_clear_bit();
Trond Myklebust7d1e8252009-03-11 14:38:03 -0400749 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
Trond Myklebust632e3bd2006-01-03 09:55:55 +0100750 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
Trond Myklebust3b948ae2007-11-05 17:42:39 -0500751 clear_bit(XPRT_CLOSING, &xprt->state);
Trond Myklebust632e3bd2006-01-03 09:55:55 +0100752 smp_mb__after_clear_bit();
Trond Myklebust62da3b22007-11-06 18:44:20 -0500753 xprt_disconnect_done(xprt);
Chuck Levera246b012005-08-11 16:25:23 -0400754}
755
Trond Myklebustf75e6742009-04-21 17:18:20 -0400756static void xs_tcp_close(struct rpc_xprt *xprt)
757{
758 if (test_and_clear_bit(XPRT_CONNECTION_CLOSE, &xprt->state))
759 xs_close(xprt);
760 else
761 xs_tcp_shutdown(xprt);
762}
763
Chuck Lever9903cd12005-08-11 16:25:26 -0400764/**
765 * xs_destroy - prepare to shutdown a transport
766 * @xprt: doomed transport
767 *
768 */
769static void xs_destroy(struct rpc_xprt *xprt)
Chuck Levera246b012005-08-11 16:25:23 -0400770{
Chuck Leverc8475462006-12-05 16:35:26 -0500771 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
772
Chuck Lever46121cf2007-01-31 12:14:08 -0500773 dprintk("RPC: xs_destroy xprt %p\n", xprt);
Chuck Lever9903cd12005-08-11 16:25:26 -0400774
Tejun Heoafe2c512010-12-14 16:21:17 +0100775 cancel_delayed_work_sync(&transport->connect_worker);
Chuck Levera246b012005-08-11 16:25:23 -0400776
Chuck Lever9903cd12005-08-11 16:25:26 -0400777 xs_close(xprt);
Chuck Leveredb267a2006-08-22 20:06:18 -0400778 xs_free_peer_addresses(xprt);
Pavel Emelyanove204e622010-09-29 16:03:13 +0400779 xprt_free(xprt);
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -0400780 module_put(THIS_MODULE);
Chuck Levera246b012005-08-11 16:25:23 -0400781}
782
Chuck Lever9903cd12005-08-11 16:25:26 -0400783static inline struct rpc_xprt *xprt_from_sock(struct sock *sk)
Chuck Levera246b012005-08-11 16:25:23 -0400784{
Chuck Lever9903cd12005-08-11 16:25:26 -0400785 return (struct rpc_xprt *) sk->sk_user_data;
786}
787
788/**
789 * xs_udp_data_ready - "data ready" callback for UDP sockets
790 * @sk: socket with data to read
791 * @len: how much data to read
792 *
793 */
794static void xs_udp_data_ready(struct sock *sk, int len)
795{
796 struct rpc_task *task;
797 struct rpc_xprt *xprt;
Chuck Levera246b012005-08-11 16:25:23 -0400798 struct rpc_rqst *rovr;
Chuck Lever9903cd12005-08-11 16:25:26 -0400799 struct sk_buff *skb;
Chuck Levera246b012005-08-11 16:25:23 -0400800 int err, repsize, copied;
Alexey Dobriyand8ed0292006-09-26 22:29:38 -0700801 u32 _xid;
802 __be32 *xp;
Chuck Levera246b012005-08-11 16:25:23 -0400803
Eric Dumazetf064af12010-09-22 12:43:39 +0000804 read_lock_bh(&sk->sk_callback_lock);
Chuck Lever46121cf2007-01-31 12:14:08 -0500805 dprintk("RPC: xs_udp_data_ready...\n");
Chuck Lever9903cd12005-08-11 16:25:26 -0400806 if (!(xprt = xprt_from_sock(sk)))
Chuck Levera246b012005-08-11 16:25:23 -0400807 goto out;
Chuck Levera246b012005-08-11 16:25:23 -0400808
809 if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL)
810 goto out;
811
812 if (xprt->shutdown)
813 goto dropit;
814
815 repsize = skb->len - sizeof(struct udphdr);
816 if (repsize < 4) {
Chuck Lever46121cf2007-01-31 12:14:08 -0500817 dprintk("RPC: impossible RPC reply size %d!\n", repsize);
Chuck Levera246b012005-08-11 16:25:23 -0400818 goto dropit;
819 }
820
821 /* Copy the XID from the skb... */
822 xp = skb_header_pointer(skb, sizeof(struct udphdr),
823 sizeof(_xid), &_xid);
824 if (xp == NULL)
825 goto dropit;
826
827 /* Look up and lock the request corresponding to the given XID */
Chuck Lever4a0f8c02005-08-11 16:25:32 -0400828 spin_lock(&xprt->transport_lock);
Chuck Levera246b012005-08-11 16:25:23 -0400829 rovr = xprt_lookup_rqst(xprt, *xp);
830 if (!rovr)
831 goto out_unlock;
832 task = rovr->rq_task;
833
Chuck Levera246b012005-08-11 16:25:23 -0400834 if ((copied = rovr->rq_private_buf.buflen) > repsize)
835 copied = repsize;
836
837 /* Suck it into the iovec, verify checksum if not done by hw. */
Herbert Xu1781f7f2007-12-11 11:30:32 -0800838 if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb)) {
839 UDPX_INC_STATS_BH(sk, UDP_MIB_INERRORS);
Chuck Levera246b012005-08-11 16:25:23 -0400840 goto out_unlock;
Herbert Xu1781f7f2007-12-11 11:30:32 -0800841 }
842
843 UDPX_INC_STATS_BH(sk, UDP_MIB_INDATAGRAMS);
Chuck Levera246b012005-08-11 16:25:23 -0400844
845 /* Something worked... */
Eric Dumazetadf30902009-06-02 05:19:30 +0000846 dst_confirm(skb_dst(skb));
Chuck Levera246b012005-08-11 16:25:23 -0400847
Chuck Lever1570c1e2005-08-25 16:25:52 -0700848 xprt_adjust_cwnd(task, copied);
Chuck Lever1570c1e2005-08-25 16:25:52 -0700849 xprt_complete_rqst(task, copied);
Chuck Levera246b012005-08-11 16:25:23 -0400850
851 out_unlock:
Chuck Lever4a0f8c02005-08-11 16:25:32 -0400852 spin_unlock(&xprt->transport_lock);
Chuck Levera246b012005-08-11 16:25:23 -0400853 dropit:
854 skb_free_datagram(sk, skb);
855 out:
Eric Dumazetf064af12010-09-22 12:43:39 +0000856 read_unlock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -0400857}
858
Chuck Leverdd456472006-12-05 16:35:44 -0500859static inline void xs_tcp_read_fraghdr(struct rpc_xprt *xprt, struct xdr_skb_reader *desc)
Chuck Levera246b012005-08-11 16:25:23 -0400860{
Chuck Lever51971132006-12-05 16:35:19 -0500861 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -0400862 size_t len, used;
863 char *p;
864
Chuck Lever51971132006-12-05 16:35:19 -0500865 p = ((char *) &transport->tcp_fraghdr) + transport->tcp_offset;
866 len = sizeof(transport->tcp_fraghdr) - transport->tcp_offset;
Chuck Lever9d292312006-12-05 16:35:41 -0500867 used = xdr_skb_read_bits(desc, p, len);
Chuck Lever51971132006-12-05 16:35:19 -0500868 transport->tcp_offset += used;
Chuck Levera246b012005-08-11 16:25:23 -0400869 if (used != len)
870 return;
Chuck Lever808012f2005-08-25 16:25:49 -0700871
Chuck Lever51971132006-12-05 16:35:19 -0500872 transport->tcp_reclen = ntohl(transport->tcp_fraghdr);
873 if (transport->tcp_reclen & RPC_LAST_STREAM_FRAGMENT)
Chuck Levere136d092006-12-05 16:35:23 -0500874 transport->tcp_flags |= TCP_RCV_LAST_FRAG;
Chuck Levera246b012005-08-11 16:25:23 -0400875 else
Chuck Levere136d092006-12-05 16:35:23 -0500876 transport->tcp_flags &= ~TCP_RCV_LAST_FRAG;
Chuck Lever51971132006-12-05 16:35:19 -0500877 transport->tcp_reclen &= RPC_FRAGMENT_SIZE_MASK;
Chuck Lever808012f2005-08-25 16:25:49 -0700878
Chuck Levere136d092006-12-05 16:35:23 -0500879 transport->tcp_flags &= ~TCP_RCV_COPY_FRAGHDR;
Chuck Lever51971132006-12-05 16:35:19 -0500880 transport->tcp_offset = 0;
Chuck Lever808012f2005-08-25 16:25:49 -0700881
Chuck Levera246b012005-08-11 16:25:23 -0400882 /* Sanity check of the record length */
Ricardo Labiaga18dca022009-04-01 09:22:53 -0400883 if (unlikely(transport->tcp_reclen < 8)) {
Chuck Lever46121cf2007-01-31 12:14:08 -0500884 dprintk("RPC: invalid TCP record fragment length\n");
Trond Myklebust3ebb0672007-11-06 18:40:12 -0500885 xprt_force_disconnect(xprt);
Chuck Lever9903cd12005-08-11 16:25:26 -0400886 return;
Chuck Levera246b012005-08-11 16:25:23 -0400887 }
Chuck Lever46121cf2007-01-31 12:14:08 -0500888 dprintk("RPC: reading TCP record fragment of length %d\n",
Chuck Lever51971132006-12-05 16:35:19 -0500889 transport->tcp_reclen);
Chuck Levera246b012005-08-11 16:25:23 -0400890}
891
Chuck Lever51971132006-12-05 16:35:19 -0500892static void xs_tcp_check_fraghdr(struct sock_xprt *transport)
Chuck Levera246b012005-08-11 16:25:23 -0400893{
Chuck Lever51971132006-12-05 16:35:19 -0500894 if (transport->tcp_offset == transport->tcp_reclen) {
Chuck Levere136d092006-12-05 16:35:23 -0500895 transport->tcp_flags |= TCP_RCV_COPY_FRAGHDR;
Chuck Lever51971132006-12-05 16:35:19 -0500896 transport->tcp_offset = 0;
Chuck Levere136d092006-12-05 16:35:23 -0500897 if (transport->tcp_flags & TCP_RCV_LAST_FRAG) {
898 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
899 transport->tcp_flags |= TCP_RCV_COPY_XID;
Chuck Lever51971132006-12-05 16:35:19 -0500900 transport->tcp_copied = 0;
Chuck Levera246b012005-08-11 16:25:23 -0400901 }
902 }
903}
904
Chuck Leverdd456472006-12-05 16:35:44 -0500905static inline void xs_tcp_read_xid(struct sock_xprt *transport, struct xdr_skb_reader *desc)
Chuck Levera246b012005-08-11 16:25:23 -0400906{
907 size_t len, used;
908 char *p;
909
Chuck Lever51971132006-12-05 16:35:19 -0500910 len = sizeof(transport->tcp_xid) - transport->tcp_offset;
Chuck Lever46121cf2007-01-31 12:14:08 -0500911 dprintk("RPC: reading XID (%Zu bytes)\n", len);
Chuck Lever51971132006-12-05 16:35:19 -0500912 p = ((char *) &transport->tcp_xid) + transport->tcp_offset;
Chuck Lever9d292312006-12-05 16:35:41 -0500913 used = xdr_skb_read_bits(desc, p, len);
Chuck Lever51971132006-12-05 16:35:19 -0500914 transport->tcp_offset += used;
Chuck Levera246b012005-08-11 16:25:23 -0400915 if (used != len)
916 return;
Chuck Levere136d092006-12-05 16:35:23 -0500917 transport->tcp_flags &= ~TCP_RCV_COPY_XID;
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -0400918 transport->tcp_flags |= TCP_RCV_READ_CALLDIR;
Chuck Lever51971132006-12-05 16:35:19 -0500919 transport->tcp_copied = 4;
Ricardo Labiaga18dca022009-04-01 09:22:53 -0400920 dprintk("RPC: reading %s XID %08x\n",
921 (transport->tcp_flags & TCP_RPC_REPLY) ? "reply for"
922 : "request with",
Chuck Lever51971132006-12-05 16:35:19 -0500923 ntohl(transport->tcp_xid));
924 xs_tcp_check_fraghdr(transport);
Chuck Levera246b012005-08-11 16:25:23 -0400925}
926
Ricardo Labiaga18dca022009-04-01 09:22:53 -0400927static inline void xs_tcp_read_calldir(struct sock_xprt *transport,
928 struct xdr_skb_reader *desc)
Chuck Levera246b012005-08-11 16:25:23 -0400929{
Ricardo Labiaga18dca022009-04-01 09:22:53 -0400930 size_t len, used;
931 u32 offset;
Trond Myklebustb76ce562010-06-16 13:57:32 -0400932 char *p;
Ricardo Labiaga18dca022009-04-01 09:22:53 -0400933
934 /*
935 * We want transport->tcp_offset to be 8 at the end of this routine
936 * (4 bytes for the xid and 4 bytes for the call/reply flag).
937 * When this function is called for the first time,
938 * transport->tcp_offset is 4 (after having already read the xid).
939 */
940 offset = transport->tcp_offset - sizeof(transport->tcp_xid);
Trond Myklebustb76ce562010-06-16 13:57:32 -0400941 len = sizeof(transport->tcp_calldir) - offset;
Ricardo Labiaga18dca022009-04-01 09:22:53 -0400942 dprintk("RPC: reading CALL/REPLY flag (%Zu bytes)\n", len);
Trond Myklebustb76ce562010-06-16 13:57:32 -0400943 p = ((char *) &transport->tcp_calldir) + offset;
944 used = xdr_skb_read_bits(desc, p, len);
Ricardo Labiaga18dca022009-04-01 09:22:53 -0400945 transport->tcp_offset += used;
946 if (used != len)
947 return;
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -0400948 transport->tcp_flags &= ~TCP_RCV_READ_CALLDIR;
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -0400949 /*
950 * We don't yet have the XDR buffer, so we will write the calldir
951 * out after we get the buffer from the 'struct rpc_rqst'
952 */
Trond Myklebustb76ce562010-06-16 13:57:32 -0400953 switch (ntohl(transport->tcp_calldir)) {
954 case RPC_REPLY:
955 transport->tcp_flags |= TCP_RCV_COPY_CALLDIR;
956 transport->tcp_flags |= TCP_RCV_COPY_DATA;
Ricardo Labiaga18dca022009-04-01 09:22:53 -0400957 transport->tcp_flags |= TCP_RPC_REPLY;
Trond Myklebustb76ce562010-06-16 13:57:32 -0400958 break;
959 case RPC_CALL:
960 transport->tcp_flags |= TCP_RCV_COPY_CALLDIR;
961 transport->tcp_flags |= TCP_RCV_COPY_DATA;
Ricardo Labiaga18dca022009-04-01 09:22:53 -0400962 transport->tcp_flags &= ~TCP_RPC_REPLY;
Trond Myklebustb76ce562010-06-16 13:57:32 -0400963 break;
964 default:
965 dprintk("RPC: invalid request message type\n");
966 xprt_force_disconnect(&transport->xprt);
967 }
Ricardo Labiaga18dca022009-04-01 09:22:53 -0400968 xs_tcp_check_fraghdr(transport);
969}
970
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -0400971static inline void xs_tcp_read_common(struct rpc_xprt *xprt,
972 struct xdr_skb_reader *desc,
973 struct rpc_rqst *req)
Chuck Levera246b012005-08-11 16:25:23 -0400974{
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -0400975 struct sock_xprt *transport =
976 container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -0400977 struct xdr_buf *rcvbuf;
978 size_t len;
979 ssize_t r;
980
Chuck Levera246b012005-08-11 16:25:23 -0400981 rcvbuf = &req->rq_private_buf;
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -0400982
983 if (transport->tcp_flags & TCP_RCV_COPY_CALLDIR) {
984 /*
985 * Save the RPC direction in the XDR buffer
986 */
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -0400987 memcpy(rcvbuf->head[0].iov_base + transport->tcp_copied,
Trond Myklebustb76ce562010-06-16 13:57:32 -0400988 &transport->tcp_calldir,
989 sizeof(transport->tcp_calldir));
990 transport->tcp_copied += sizeof(transport->tcp_calldir);
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -0400991 transport->tcp_flags &= ~TCP_RCV_COPY_CALLDIR;
Chuck Levera246b012005-08-11 16:25:23 -0400992 }
993
Chuck Levera246b012005-08-11 16:25:23 -0400994 len = desc->count;
Chuck Lever51971132006-12-05 16:35:19 -0500995 if (len > transport->tcp_reclen - transport->tcp_offset) {
Chuck Leverdd456472006-12-05 16:35:44 -0500996 struct xdr_skb_reader my_desc;
Chuck Levera246b012005-08-11 16:25:23 -0400997
Chuck Lever51971132006-12-05 16:35:19 -0500998 len = transport->tcp_reclen - transport->tcp_offset;
Chuck Levera246b012005-08-11 16:25:23 -0400999 memcpy(&my_desc, desc, sizeof(my_desc));
1000 my_desc.count = len;
Chuck Lever51971132006-12-05 16:35:19 -05001001 r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied,
Chuck Lever9d292312006-12-05 16:35:41 -05001002 &my_desc, xdr_skb_read_bits);
Chuck Levera246b012005-08-11 16:25:23 -04001003 desc->count -= r;
1004 desc->offset += r;
1005 } else
Chuck Lever51971132006-12-05 16:35:19 -05001006 r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied,
Chuck Lever9d292312006-12-05 16:35:41 -05001007 desc, xdr_skb_read_bits);
Chuck Levera246b012005-08-11 16:25:23 -04001008
1009 if (r > 0) {
Chuck Lever51971132006-12-05 16:35:19 -05001010 transport->tcp_copied += r;
1011 transport->tcp_offset += r;
Chuck Levera246b012005-08-11 16:25:23 -04001012 }
1013 if (r != len) {
1014 /* Error when copying to the receive buffer,
1015 * usually because we weren't able to allocate
1016 * additional buffer pages. All we can do now
Chuck Levere136d092006-12-05 16:35:23 -05001017 * is turn off TCP_RCV_COPY_DATA, so the request
Chuck Levera246b012005-08-11 16:25:23 -04001018 * will not receive any additional updates,
1019 * and time out.
1020 * Any remaining data from this record will
1021 * be discarded.
1022 */
Chuck Levere136d092006-12-05 16:35:23 -05001023 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
Chuck Lever46121cf2007-01-31 12:14:08 -05001024 dprintk("RPC: XID %08x truncated request\n",
Chuck Lever51971132006-12-05 16:35:19 -05001025 ntohl(transport->tcp_xid));
Chuck Lever46121cf2007-01-31 12:14:08 -05001026 dprintk("RPC: xprt = %p, tcp_copied = %lu, "
1027 "tcp_offset = %u, tcp_reclen = %u\n",
1028 xprt, transport->tcp_copied,
1029 transport->tcp_offset, transport->tcp_reclen);
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001030 return;
Chuck Levera246b012005-08-11 16:25:23 -04001031 }
1032
Chuck Lever46121cf2007-01-31 12:14:08 -05001033 dprintk("RPC: XID %08x read %Zd bytes\n",
Chuck Lever51971132006-12-05 16:35:19 -05001034 ntohl(transport->tcp_xid), r);
Chuck Lever46121cf2007-01-31 12:14:08 -05001035 dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, "
1036 "tcp_reclen = %u\n", xprt, transport->tcp_copied,
1037 transport->tcp_offset, transport->tcp_reclen);
Chuck Levera246b012005-08-11 16:25:23 -04001038
Chuck Lever51971132006-12-05 16:35:19 -05001039 if (transport->tcp_copied == req->rq_private_buf.buflen)
Chuck Levere136d092006-12-05 16:35:23 -05001040 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
Chuck Lever51971132006-12-05 16:35:19 -05001041 else if (transport->tcp_offset == transport->tcp_reclen) {
Chuck Levere136d092006-12-05 16:35:23 -05001042 if (transport->tcp_flags & TCP_RCV_LAST_FRAG)
1043 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
Chuck Levera246b012005-08-11 16:25:23 -04001044 }
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001045}
1046
1047/*
1048 * Finds the request corresponding to the RPC xid and invokes the common
1049 * tcp read code to read the data.
1050 */
1051static inline int xs_tcp_read_reply(struct rpc_xprt *xprt,
1052 struct xdr_skb_reader *desc)
1053{
1054 struct sock_xprt *transport =
1055 container_of(xprt, struct sock_xprt, xprt);
1056 struct rpc_rqst *req;
1057
1058 dprintk("RPC: read reply XID %08x\n", ntohl(transport->tcp_xid));
1059
1060 /* Find and lock the request corresponding to this xid */
1061 spin_lock(&xprt->transport_lock);
1062 req = xprt_lookup_rqst(xprt, transport->tcp_xid);
1063 if (!req) {
1064 dprintk("RPC: XID %08x request not found!\n",
1065 ntohl(transport->tcp_xid));
1066 spin_unlock(&xprt->transport_lock);
1067 return -1;
1068 }
1069
1070 xs_tcp_read_common(xprt, desc, req);
1071
Chuck Levere136d092006-12-05 16:35:23 -05001072 if (!(transport->tcp_flags & TCP_RCV_COPY_DATA))
Chuck Lever51971132006-12-05 16:35:19 -05001073 xprt_complete_rqst(req->rq_task, transport->tcp_copied);
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001074
Chuck Lever4a0f8c02005-08-11 16:25:32 -04001075 spin_unlock(&xprt->transport_lock);
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001076 return 0;
1077}
1078
1079#if defined(CONFIG_NFS_V4_1)
1080/*
1081 * Obtains an rpc_rqst previously allocated and invokes the common
1082 * tcp read code to read the data. The result is placed in the callback
1083 * queue.
1084 * If we're unable to obtain the rpc_rqst we schedule the closing of the
1085 * connection and return -1.
1086 */
1087static inline int xs_tcp_read_callback(struct rpc_xprt *xprt,
1088 struct xdr_skb_reader *desc)
1089{
1090 struct sock_xprt *transport =
1091 container_of(xprt, struct sock_xprt, xprt);
1092 struct rpc_rqst *req;
1093
1094 req = xprt_alloc_bc_request(xprt);
1095 if (req == NULL) {
1096 printk(KERN_WARNING "Callback slot table overflowed\n");
1097 xprt_force_disconnect(xprt);
1098 return -1;
1099 }
1100
1101 req->rq_xid = transport->tcp_xid;
1102 dprintk("RPC: read callback XID %08x\n", ntohl(req->rq_xid));
1103 xs_tcp_read_common(xprt, desc, req);
1104
1105 if (!(transport->tcp_flags & TCP_RCV_COPY_DATA)) {
1106 struct svc_serv *bc_serv = xprt->bc_serv;
1107
1108 /*
1109 * Add callback request to callback list. The callback
1110 * service sleeps on the sv_cb_waitq waiting for new
1111 * requests. Wake it up after adding enqueing the
1112 * request.
1113 */
1114 dprintk("RPC: add callback request to list\n");
1115 spin_lock(&bc_serv->sv_cb_lock);
1116 list_add(&req->rq_bc_list, &bc_serv->sv_cb_list);
1117 spin_unlock(&bc_serv->sv_cb_lock);
1118 wake_up(&bc_serv->sv_cb_waitq);
1119 }
1120
1121 req->rq_private_buf.len = transport->tcp_copied;
1122
1123 return 0;
1124}
1125
1126static inline int _xs_tcp_read_data(struct rpc_xprt *xprt,
1127 struct xdr_skb_reader *desc)
1128{
1129 struct sock_xprt *transport =
1130 container_of(xprt, struct sock_xprt, xprt);
1131
1132 return (transport->tcp_flags & TCP_RPC_REPLY) ?
1133 xs_tcp_read_reply(xprt, desc) :
1134 xs_tcp_read_callback(xprt, desc);
1135}
1136#else
1137static inline int _xs_tcp_read_data(struct rpc_xprt *xprt,
1138 struct xdr_skb_reader *desc)
1139{
1140 return xs_tcp_read_reply(xprt, desc);
1141}
1142#endif /* CONFIG_NFS_V4_1 */
1143
1144/*
1145 * Read data off the transport. This can be either an RPC_CALL or an
1146 * RPC_REPLY. Relay the processing to helper functions.
1147 */
1148static void xs_tcp_read_data(struct rpc_xprt *xprt,
1149 struct xdr_skb_reader *desc)
1150{
1151 struct sock_xprt *transport =
1152 container_of(xprt, struct sock_xprt, xprt);
1153
1154 if (_xs_tcp_read_data(xprt, desc) == 0)
1155 xs_tcp_check_fraghdr(transport);
1156 else {
1157 /*
1158 * The transport_lock protects the request handling.
1159 * There's no need to hold it to update the tcp_flags.
1160 */
1161 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
1162 }
Chuck Levera246b012005-08-11 16:25:23 -04001163}
1164
Chuck Leverdd456472006-12-05 16:35:44 -05001165static inline void xs_tcp_read_discard(struct sock_xprt *transport, struct xdr_skb_reader *desc)
Chuck Levera246b012005-08-11 16:25:23 -04001166{
1167 size_t len;
1168
Chuck Lever51971132006-12-05 16:35:19 -05001169 len = transport->tcp_reclen - transport->tcp_offset;
Chuck Levera246b012005-08-11 16:25:23 -04001170 if (len > desc->count)
1171 len = desc->count;
1172 desc->count -= len;
1173 desc->offset += len;
Chuck Lever51971132006-12-05 16:35:19 -05001174 transport->tcp_offset += len;
Chuck Lever46121cf2007-01-31 12:14:08 -05001175 dprintk("RPC: discarded %Zu bytes\n", len);
Chuck Lever51971132006-12-05 16:35:19 -05001176 xs_tcp_check_fraghdr(transport);
Chuck Levera246b012005-08-11 16:25:23 -04001177}
1178
Chuck Lever9903cd12005-08-11 16:25:26 -04001179static int xs_tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb, unsigned int offset, size_t len)
Chuck Levera246b012005-08-11 16:25:23 -04001180{
1181 struct rpc_xprt *xprt = rd_desc->arg.data;
Chuck Lever51971132006-12-05 16:35:19 -05001182 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Leverdd456472006-12-05 16:35:44 -05001183 struct xdr_skb_reader desc = {
Chuck Levera246b012005-08-11 16:25:23 -04001184 .skb = skb,
1185 .offset = offset,
1186 .count = len,
Chuck Lever9903cd12005-08-11 16:25:26 -04001187 };
Chuck Levera246b012005-08-11 16:25:23 -04001188
Chuck Lever46121cf2007-01-31 12:14:08 -05001189 dprintk("RPC: xs_tcp_data_recv started\n");
Chuck Levera246b012005-08-11 16:25:23 -04001190 do {
1191 /* Read in a new fragment marker if necessary */
1192 /* Can we ever really expect to get completely empty fragments? */
Chuck Levere136d092006-12-05 16:35:23 -05001193 if (transport->tcp_flags & TCP_RCV_COPY_FRAGHDR) {
Chuck Lever9903cd12005-08-11 16:25:26 -04001194 xs_tcp_read_fraghdr(xprt, &desc);
Chuck Levera246b012005-08-11 16:25:23 -04001195 continue;
1196 }
1197 /* Read in the xid if necessary */
Chuck Levere136d092006-12-05 16:35:23 -05001198 if (transport->tcp_flags & TCP_RCV_COPY_XID) {
Chuck Lever51971132006-12-05 16:35:19 -05001199 xs_tcp_read_xid(transport, &desc);
Chuck Levera246b012005-08-11 16:25:23 -04001200 continue;
1201 }
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001202 /* Read in the call/reply flag */
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001203 if (transport->tcp_flags & TCP_RCV_READ_CALLDIR) {
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001204 xs_tcp_read_calldir(transport, &desc);
1205 continue;
1206 }
Chuck Levera246b012005-08-11 16:25:23 -04001207 /* Read in the request data */
Chuck Levere136d092006-12-05 16:35:23 -05001208 if (transport->tcp_flags & TCP_RCV_COPY_DATA) {
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001209 xs_tcp_read_data(xprt, &desc);
Chuck Levera246b012005-08-11 16:25:23 -04001210 continue;
1211 }
1212 /* Skip over any trailing bytes on short reads */
Chuck Lever51971132006-12-05 16:35:19 -05001213 xs_tcp_read_discard(transport, &desc);
Chuck Levera246b012005-08-11 16:25:23 -04001214 } while (desc.count);
Chuck Lever46121cf2007-01-31 12:14:08 -05001215 dprintk("RPC: xs_tcp_data_recv done\n");
Chuck Levera246b012005-08-11 16:25:23 -04001216 return len - desc.count;
1217}
1218
Chuck Lever9903cd12005-08-11 16:25:26 -04001219/**
1220 * xs_tcp_data_ready - "data ready" callback for TCP sockets
1221 * @sk: socket with data to read
1222 * @bytes: how much data to read
1223 *
1224 */
1225static void xs_tcp_data_ready(struct sock *sk, int bytes)
Chuck Levera246b012005-08-11 16:25:23 -04001226{
1227 struct rpc_xprt *xprt;
1228 read_descriptor_t rd_desc;
Trond Myklebustff2d7db2008-02-25 21:40:51 -08001229 int read;
Chuck Levera246b012005-08-11 16:25:23 -04001230
Chuck Lever46121cf2007-01-31 12:14:08 -05001231 dprintk("RPC: xs_tcp_data_ready...\n");
1232
Eric Dumazetf064af12010-09-22 12:43:39 +00001233 read_lock_bh(&sk->sk_callback_lock);
Chuck Lever9903cd12005-08-11 16:25:26 -04001234 if (!(xprt = xprt_from_sock(sk)))
Chuck Levera246b012005-08-11 16:25:23 -04001235 goto out;
Chuck Levera246b012005-08-11 16:25:23 -04001236 if (xprt->shutdown)
1237 goto out;
1238
Neil Brown61d0a8e2009-09-23 14:36:37 -04001239 /* Any data means we had a useful conversation, so
1240 * the we don't need to delay the next reconnect
1241 */
1242 if (xprt->reestablish_timeout)
1243 xprt->reestablish_timeout = 0;
1244
Chuck Lever9903cd12005-08-11 16:25:26 -04001245 /* We use rd_desc to pass struct xprt to xs_tcp_data_recv */
Chuck Levera246b012005-08-11 16:25:23 -04001246 rd_desc.arg.data = xprt;
Trond Myklebustff2d7db2008-02-25 21:40:51 -08001247 do {
1248 rd_desc.count = 65536;
1249 read = tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv);
1250 } while (read > 0);
Chuck Levera246b012005-08-11 16:25:23 -04001251out:
Eric Dumazetf064af12010-09-22 12:43:39 +00001252 read_unlock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001253}
1254
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001255/*
1256 * Do the equivalent of linger/linger2 handling for dealing with
1257 * broken servers that don't close the socket in a timely
1258 * fashion
1259 */
1260static void xs_tcp_schedule_linger_timeout(struct rpc_xprt *xprt,
1261 unsigned long timeout)
1262{
1263 struct sock_xprt *transport;
1264
1265 if (xprt_test_and_set_connecting(xprt))
1266 return;
1267 set_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1268 transport = container_of(xprt, struct sock_xprt, xprt);
1269 queue_delayed_work(rpciod_workqueue, &transport->connect_worker,
1270 timeout);
1271}
1272
1273static void xs_tcp_cancel_linger_timeout(struct rpc_xprt *xprt)
1274{
1275 struct sock_xprt *transport;
1276
1277 transport = container_of(xprt, struct sock_xprt, xprt);
1278
1279 if (!test_bit(XPRT_CONNECTION_ABORT, &xprt->state) ||
1280 !cancel_delayed_work(&transport->connect_worker))
1281 return;
1282 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1283 xprt_clear_connecting(xprt);
1284}
1285
1286static void xs_sock_mark_closed(struct rpc_xprt *xprt)
1287{
1288 smp_mb__before_clear_bit();
1289 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
1290 clear_bit(XPRT_CLOSING, &xprt->state);
1291 smp_mb__after_clear_bit();
1292 /* Mark transport as closed and wake up all pending tasks */
1293 xprt_disconnect_done(xprt);
1294}
1295
Chuck Lever9903cd12005-08-11 16:25:26 -04001296/**
1297 * xs_tcp_state_change - callback to handle TCP socket state changes
1298 * @sk: socket whose state has changed
1299 *
1300 */
1301static void xs_tcp_state_change(struct sock *sk)
Chuck Levera246b012005-08-11 16:25:23 -04001302{
Chuck Lever9903cd12005-08-11 16:25:26 -04001303 struct rpc_xprt *xprt;
Chuck Levera246b012005-08-11 16:25:23 -04001304
Eric Dumazetf064af12010-09-22 12:43:39 +00001305 read_lock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001306 if (!(xprt = xprt_from_sock(sk)))
1307 goto out;
Chuck Lever46121cf2007-01-31 12:14:08 -05001308 dprintk("RPC: xs_tcp_state_change client %p...\n", xprt);
Andy Chittenden669502f2010-08-10 10:19:53 -04001309 dprintk("RPC: state %x conn %d dead %d zapped %d sk_shutdown %d\n",
Chuck Lever46121cf2007-01-31 12:14:08 -05001310 sk->sk_state, xprt_connected(xprt),
1311 sock_flag(sk, SOCK_DEAD),
Andy Chittenden669502f2010-08-10 10:19:53 -04001312 sock_flag(sk, SOCK_ZAPPED),
1313 sk->sk_shutdown);
Chuck Levera246b012005-08-11 16:25:23 -04001314
1315 switch (sk->sk_state) {
1316 case TCP_ESTABLISHED:
Eric Dumazetf064af12010-09-22 12:43:39 +00001317 spin_lock(&xprt->transport_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001318 if (!xprt_test_and_set_connected(xprt)) {
Chuck Lever51971132006-12-05 16:35:19 -05001319 struct sock_xprt *transport = container_of(xprt,
1320 struct sock_xprt, xprt);
1321
Chuck Levera246b012005-08-11 16:25:23 -04001322 /* Reset TCP record info */
Chuck Lever51971132006-12-05 16:35:19 -05001323 transport->tcp_offset = 0;
1324 transport->tcp_reclen = 0;
1325 transport->tcp_copied = 0;
Chuck Levere136d092006-12-05 16:35:23 -05001326 transport->tcp_flags =
1327 TCP_RCV_COPY_FRAGHDR | TCP_RCV_COPY_XID;
Chuck Lever51971132006-12-05 16:35:19 -05001328
Trond Myklebust2a491992009-03-11 14:38:00 -04001329 xprt_wake_pending_tasks(xprt, -EAGAIN);
Chuck Levera246b012005-08-11 16:25:23 -04001330 }
Eric Dumazetf064af12010-09-22 12:43:39 +00001331 spin_unlock(&xprt->transport_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001332 break;
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001333 case TCP_FIN_WAIT1:
1334 /* The client initiated a shutdown of the socket */
Trond Myklebust7c1d71c2008-04-17 16:52:57 -04001335 xprt->connect_cookie++;
Trond Myklebust663b8852008-01-01 18:42:12 -05001336 xprt->reestablish_timeout = 0;
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001337 set_bit(XPRT_CLOSING, &xprt->state);
1338 smp_mb__before_clear_bit();
1339 clear_bit(XPRT_CONNECTED, &xprt->state);
Trond Myklebustef803672007-12-31 16:19:17 -05001340 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001341 smp_mb__after_clear_bit();
Trond Myklebust25fe6142009-03-11 14:38:03 -04001342 xs_tcp_schedule_linger_timeout(xprt, xs_tcp_fin_timeout);
Chuck Levera246b012005-08-11 16:25:23 -04001343 break;
Trond Myklebust632e3bd2006-01-03 09:55:55 +01001344 case TCP_CLOSE_WAIT:
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001345 /* The server initiated a shutdown of the socket */
Trond Myklebust66af1e52007-11-06 10:18:36 -05001346 xprt_force_disconnect(xprt);
Trond Myklebust663b8852008-01-01 18:42:12 -05001347 case TCP_SYN_SENT:
Trond Myklebust7c1d71c2008-04-17 16:52:57 -04001348 xprt->connect_cookie++;
Trond Myklebust663b8852008-01-01 18:42:12 -05001349 case TCP_CLOSING:
1350 /*
1351 * If the server closed down the connection, make sure that
1352 * we back off before reconnecting
1353 */
1354 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
1355 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001356 break;
1357 case TCP_LAST_ACK:
Trond Myklebust670f9452009-03-11 14:37:58 -04001358 set_bit(XPRT_CLOSING, &xprt->state);
Trond Myklebust25fe6142009-03-11 14:38:03 -04001359 xs_tcp_schedule_linger_timeout(xprt, xs_tcp_fin_timeout);
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001360 smp_mb__before_clear_bit();
1361 clear_bit(XPRT_CONNECTED, &xprt->state);
1362 smp_mb__after_clear_bit();
1363 break;
1364 case TCP_CLOSE:
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001365 xs_tcp_cancel_linger_timeout(xprt);
1366 xs_sock_mark_closed(xprt);
Chuck Levera246b012005-08-11 16:25:23 -04001367 }
1368 out:
Eric Dumazetf064af12010-09-22 12:43:39 +00001369 read_unlock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001370}
1371
Chuck Lever9903cd12005-08-11 16:25:26 -04001372/**
Trond Myklebust482f32e2009-03-11 14:38:00 -04001373 * xs_error_report - callback mainly for catching socket errors
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04001374 * @sk: socket
1375 */
Trond Myklebust482f32e2009-03-11 14:38:00 -04001376static void xs_error_report(struct sock *sk)
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04001377{
1378 struct rpc_xprt *xprt;
1379
Eric Dumazetf064af12010-09-22 12:43:39 +00001380 read_lock_bh(&sk->sk_callback_lock);
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04001381 if (!(xprt = xprt_from_sock(sk)))
1382 goto out;
1383 dprintk("RPC: %s client %p...\n"
1384 "RPC: error %d\n",
1385 __func__, xprt, sk->sk_err);
Trond Myklebust482f32e2009-03-11 14:38:00 -04001386 xprt_wake_pending_tasks(xprt, -EAGAIN);
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04001387out:
Eric Dumazetf064af12010-09-22 12:43:39 +00001388 read_unlock_bh(&sk->sk_callback_lock);
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04001389}
1390
Ilpo Järvinen1f0fa152009-02-06 23:48:33 -08001391static void xs_write_space(struct sock *sk)
1392{
1393 struct socket *sock;
1394 struct rpc_xprt *xprt;
1395
1396 if (unlikely(!(sock = sk->sk_socket)))
1397 return;
1398 clear_bit(SOCK_NOSPACE, &sock->flags);
1399
1400 if (unlikely(!(xprt = xprt_from_sock(sk))))
1401 return;
1402 if (test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags) == 0)
1403 return;
1404
1405 xprt_write_space(xprt);
1406}
1407
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04001408/**
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001409 * xs_udp_write_space - callback invoked when socket buffer space
1410 * becomes available
Chuck Lever9903cd12005-08-11 16:25:26 -04001411 * @sk: socket whose state has changed
1412 *
Chuck Levera246b012005-08-11 16:25:23 -04001413 * Called when more output buffer space is available for this socket.
1414 * We try not to wake our writers until they can make "significant"
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001415 * progress, otherwise we'll waste resources thrashing kernel_sendmsg
Chuck Levera246b012005-08-11 16:25:23 -04001416 * with a bunch of small requests.
1417 */
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001418static void xs_udp_write_space(struct sock *sk)
Chuck Levera246b012005-08-11 16:25:23 -04001419{
Eric Dumazetf064af12010-09-22 12:43:39 +00001420 read_lock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001421
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001422 /* from net/core/sock.c:sock_def_write_space */
Ilpo Järvinen1f0fa152009-02-06 23:48:33 -08001423 if (sock_writeable(sk))
1424 xs_write_space(sk);
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001425
Eric Dumazetf064af12010-09-22 12:43:39 +00001426 read_unlock_bh(&sk->sk_callback_lock);
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001427}
Chuck Levera246b012005-08-11 16:25:23 -04001428
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001429/**
1430 * xs_tcp_write_space - callback invoked when socket buffer space
1431 * becomes available
1432 * @sk: socket whose state has changed
1433 *
1434 * Called when more output buffer space is available for this socket.
1435 * We try not to wake our writers until they can make "significant"
1436 * progress, otherwise we'll waste resources thrashing kernel_sendmsg
1437 * with a bunch of small requests.
1438 */
1439static void xs_tcp_write_space(struct sock *sk)
1440{
Eric Dumazetf064af12010-09-22 12:43:39 +00001441 read_lock_bh(&sk->sk_callback_lock);
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001442
1443 /* from net/core/stream.c:sk_stream_write_space */
Ilpo Järvinen1f0fa152009-02-06 23:48:33 -08001444 if (sk_stream_wspace(sk) >= sk_stream_min_wspace(sk))
1445 xs_write_space(sk);
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001446
Eric Dumazetf064af12010-09-22 12:43:39 +00001447 read_unlock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001448}
1449
Chuck Lever470056c2005-08-25 16:25:56 -07001450static void xs_udp_do_set_buffer_size(struct rpc_xprt *xprt)
Chuck Levera246b012005-08-11 16:25:23 -04001451{
Chuck Leveree0ac0c2006-12-05 16:35:15 -05001452 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1453 struct sock *sk = transport->inet;
Chuck Levera246b012005-08-11 16:25:23 -04001454
Chuck Lever7c6e0662006-12-05 16:35:30 -05001455 if (transport->rcvsize) {
Chuck Levera246b012005-08-11 16:25:23 -04001456 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
Chuck Lever7c6e0662006-12-05 16:35:30 -05001457 sk->sk_rcvbuf = transport->rcvsize * xprt->max_reqs * 2;
Chuck Levera246b012005-08-11 16:25:23 -04001458 }
Chuck Lever7c6e0662006-12-05 16:35:30 -05001459 if (transport->sndsize) {
Chuck Levera246b012005-08-11 16:25:23 -04001460 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
Chuck Lever7c6e0662006-12-05 16:35:30 -05001461 sk->sk_sndbuf = transport->sndsize * xprt->max_reqs * 2;
Chuck Levera246b012005-08-11 16:25:23 -04001462 sk->sk_write_space(sk);
1463 }
1464}
1465
Chuck Lever43118c22005-08-25 16:25:49 -07001466/**
Chuck Lever470056c2005-08-25 16:25:56 -07001467 * xs_udp_set_buffer_size - set send and receive limits
Chuck Lever43118c22005-08-25 16:25:49 -07001468 * @xprt: generic transport
Chuck Lever470056c2005-08-25 16:25:56 -07001469 * @sndsize: requested size of send buffer, in bytes
1470 * @rcvsize: requested size of receive buffer, in bytes
Chuck Lever43118c22005-08-25 16:25:49 -07001471 *
Chuck Lever470056c2005-08-25 16:25:56 -07001472 * Set socket send and receive buffer size limits.
Chuck Lever43118c22005-08-25 16:25:49 -07001473 */
Chuck Lever470056c2005-08-25 16:25:56 -07001474static void xs_udp_set_buffer_size(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize)
Chuck Lever43118c22005-08-25 16:25:49 -07001475{
Chuck Lever7c6e0662006-12-05 16:35:30 -05001476 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1477
1478 transport->sndsize = 0;
Chuck Lever470056c2005-08-25 16:25:56 -07001479 if (sndsize)
Chuck Lever7c6e0662006-12-05 16:35:30 -05001480 transport->sndsize = sndsize + 1024;
1481 transport->rcvsize = 0;
Chuck Lever470056c2005-08-25 16:25:56 -07001482 if (rcvsize)
Chuck Lever7c6e0662006-12-05 16:35:30 -05001483 transport->rcvsize = rcvsize + 1024;
Chuck Lever470056c2005-08-25 16:25:56 -07001484
1485 xs_udp_do_set_buffer_size(xprt);
Chuck Lever43118c22005-08-25 16:25:49 -07001486}
1487
Chuck Lever46c0ee82005-08-25 16:25:52 -07001488/**
1489 * xs_udp_timer - called when a retransmit timeout occurs on a UDP transport
1490 * @task: task that timed out
1491 *
1492 * Adjust the congestion window after a retransmit timeout has occurred.
1493 */
1494static void xs_udp_timer(struct rpc_task *task)
1495{
1496 xprt_adjust_cwnd(task, -ETIMEDOUT);
1497}
1498
Chuck Leverb85d8802006-05-25 01:40:49 -04001499static unsigned short xs_get_random_port(void)
1500{
1501 unsigned short range = xprt_max_resvport - xprt_min_resvport;
1502 unsigned short rand = (unsigned short) net_random() % range;
1503 return rand + xprt_min_resvport;
1504}
1505
Chuck Lever92200412006-01-03 09:55:51 +01001506/**
1507 * xs_set_port - reset the port number in the remote endpoint address
1508 * @xprt: generic transport
1509 * @port: new port number
1510 *
1511 */
1512static void xs_set_port(struct rpc_xprt *xprt, unsigned short port)
1513{
Chuck Lever46121cf2007-01-31 12:14:08 -05001514 dprintk("RPC: setting port for xprt %p to %u\n", xprt, port);
Chuck Leverc4efcb12006-08-22 20:06:19 -04001515
Chuck Lever9dc3b092009-08-09 15:09:46 -04001516 rpc_set_port(xs_addr(xprt), port);
1517 xs_update_peer_port(xprt);
Chuck Lever92200412006-01-03 09:55:51 +01001518}
1519
Pavel Emelyanov5d4ec932010-10-04 16:51:23 +04001520static unsigned short xs_get_srcport(struct sock_xprt *transport)
Trond Myklebust67a391d2007-11-05 17:40:58 -05001521{
Chuck Leverfbfffbd2009-08-09 15:09:46 -04001522 unsigned short port = transport->srcport;
Trond Myklebust67a391d2007-11-05 17:40:58 -05001523
1524 if (port == 0 && transport->xprt.resvport)
1525 port = xs_get_random_port();
1526 return port;
1527}
1528
Pavel Emelyanovbaaf4e42010-10-04 16:51:56 +04001529static unsigned short xs_next_srcport(struct sock_xprt *transport, unsigned short port)
Trond Myklebust67a391d2007-11-05 17:40:58 -05001530{
Chuck Leverfbfffbd2009-08-09 15:09:46 -04001531 if (transport->srcport != 0)
1532 transport->srcport = 0;
Trond Myklebust67a391d2007-11-05 17:40:58 -05001533 if (!transport->xprt.resvport)
1534 return 0;
1535 if (port <= xprt_min_resvport || port > xprt_max_resvport)
1536 return xprt_max_resvport;
1537 return --port;
1538}
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001539static int xs_bind(struct sock_xprt *transport, struct socket *sock)
Chuck Levera246b012005-08-11 16:25:23 -04001540{
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001541 struct sockaddr_storage myaddr;
Trond Myklebust67a391d2007-11-05 17:40:58 -05001542 int err, nloop = 0;
Pavel Emelyanov5d4ec932010-10-04 16:51:23 +04001543 unsigned short port = xs_get_srcport(transport);
Trond Myklebust67a391d2007-11-05 17:40:58 -05001544 unsigned short last;
Chuck Levera246b012005-08-11 16:25:23 -04001545
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001546 memcpy(&myaddr, &transport->srcaddr, transport->xprt.addrlen);
Chuck Levera246b012005-08-11 16:25:23 -04001547 do {
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001548 rpc_set_port((struct sockaddr *)&myaddr, port);
1549 err = kernel_bind(sock, (struct sockaddr *)&myaddr,
1550 transport->xprt.addrlen);
Trond Myklebust67a391d2007-11-05 17:40:58 -05001551 if (port == 0)
Frank van Maarseveend3bc9a12007-07-09 22:23:35 +02001552 break;
Chuck Levera246b012005-08-11 16:25:23 -04001553 if (err == 0) {
Chuck Leverfbfffbd2009-08-09 15:09:46 -04001554 transport->srcport = port;
Frank van Maarseveend3bc9a12007-07-09 22:23:35 +02001555 break;
Chuck Levera246b012005-08-11 16:25:23 -04001556 }
Trond Myklebust67a391d2007-11-05 17:40:58 -05001557 last = port;
Pavel Emelyanovbaaf4e42010-10-04 16:51:56 +04001558 port = xs_next_srcport(transport, port);
Trond Myklebust67a391d2007-11-05 17:40:58 -05001559 if (port > last)
1560 nloop++;
1561 } while (err == -EADDRINUSE && nloop != 2);
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001562
Chuck Lever4232e862010-10-20 11:52:51 -04001563 if (myaddr.ss_family == AF_INET)
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001564 dprintk("RPC: %s %pI4:%u: %s (%d)\n", __func__,
1565 &((struct sockaddr_in *)&myaddr)->sin_addr,
1566 port, err ? "failed" : "ok", err);
1567 else
1568 dprintk("RPC: %s %pI6:%u: %s (%d)\n", __func__,
1569 &((struct sockaddr_in6 *)&myaddr)->sin6_addr,
1570 port, err ? "failed" : "ok", err);
Chuck Levera246b012005-08-11 16:25:23 -04001571 return err;
1572}
1573
Chuck Levera246b012005-08-11 16:25:23 -04001574
Peter Zijlstraed075362006-12-06 20:35:24 -08001575#ifdef CONFIG_DEBUG_LOCK_ALLOC
1576static struct lock_class_key xs_key[2];
1577static struct lock_class_key xs_slock_key[2];
1578
Chuck Lever8945ee52007-08-06 11:58:04 -04001579static inline void xs_reclassify_socket4(struct socket *sock)
Peter Zijlstraed075362006-12-06 20:35:24 -08001580{
1581 struct sock *sk = sock->sk;
Chuck Lever8945ee52007-08-06 11:58:04 -04001582
John Heffner02b3d342007-09-12 10:42:12 +02001583 BUG_ON(sock_owned_by_user(sk));
Chuck Lever8945ee52007-08-06 11:58:04 -04001584 sock_lock_init_class_and_name(sk, "slock-AF_INET-RPC",
1585 &xs_slock_key[0], "sk_lock-AF_INET-RPC", &xs_key[0]);
1586}
Peter Zijlstraed075362006-12-06 20:35:24 -08001587
Chuck Lever8945ee52007-08-06 11:58:04 -04001588static inline void xs_reclassify_socket6(struct socket *sock)
1589{
1590 struct sock *sk = sock->sk;
Peter Zijlstraed075362006-12-06 20:35:24 -08001591
Linus Torvaldsf4921af2007-10-15 10:46:05 -07001592 BUG_ON(sock_owned_by_user(sk));
Chuck Lever8945ee52007-08-06 11:58:04 -04001593 sock_lock_init_class_and_name(sk, "slock-AF_INET6-RPC",
1594 &xs_slock_key[1], "sk_lock-AF_INET6-RPC", &xs_key[1]);
Peter Zijlstraed075362006-12-06 20:35:24 -08001595}
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001596
1597static inline void xs_reclassify_socket(int family, struct socket *sock)
1598{
Chuck Lever4232e862010-10-20 11:52:51 -04001599 switch (family) {
1600 case AF_INET:
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001601 xs_reclassify_socket4(sock);
Chuck Lever4232e862010-10-20 11:52:51 -04001602 break;
1603 case AF_INET6:
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001604 xs_reclassify_socket6(sock);
Chuck Lever4232e862010-10-20 11:52:51 -04001605 break;
1606 }
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001607}
Peter Zijlstraed075362006-12-06 20:35:24 -08001608#else
Chuck Lever8945ee52007-08-06 11:58:04 -04001609static inline void xs_reclassify_socket4(struct socket *sock)
1610{
1611}
1612
1613static inline void xs_reclassify_socket6(struct socket *sock)
Peter Zijlstraed075362006-12-06 20:35:24 -08001614{
1615}
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001616
1617static inline void xs_reclassify_socket(int family, struct socket *sock)
1618{
1619}
Peter Zijlstraed075362006-12-06 20:35:24 -08001620#endif
1621
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001622static struct socket *xs_create_sock(struct rpc_xprt *xprt,
1623 struct sock_xprt *transport, int family, int type, int protocol)
Pavel Emelyanov22f79322010-10-04 16:54:26 +04001624{
1625 struct socket *sock;
1626 int err;
1627
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001628 err = __sock_create(xprt->xprt_net, family, type, protocol, &sock, 1);
Pavel Emelyanov22f79322010-10-04 16:54:26 +04001629 if (err < 0) {
1630 dprintk("RPC: can't create %d transport socket (%d).\n",
1631 protocol, -err);
1632 goto out;
1633 }
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001634 xs_reclassify_socket(family, sock);
Pavel Emelyanov22f79322010-10-04 16:54:26 +04001635
Ben Hutchings4cea2882011-02-22 21:54:34 +00001636 err = xs_bind(transport, sock);
1637 if (err) {
Pavel Emelyanov22f79322010-10-04 16:54:26 +04001638 sock_release(sock);
1639 goto out;
1640 }
1641
1642 return sock;
1643out:
1644 return ERR_PTR(err);
1645}
1646
Chuck Lever16be2d22007-08-06 11:57:38 -04001647static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
Chuck Levera246b012005-08-11 16:25:23 -04001648{
Chuck Lever16be2d22007-08-06 11:57:38 -04001649 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Leveredb267a2006-08-22 20:06:18 -04001650
Chuck Leveree0ac0c2006-12-05 16:35:15 -05001651 if (!transport->inet) {
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001652 struct sock *sk = sock->sk;
1653
1654 write_lock_bh(&sk->sk_callback_lock);
1655
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04001656 xs_save_old_callbacks(transport, sk);
1657
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001658 sk->sk_user_data = xprt;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001659 sk->sk_data_ready = xs_udp_data_ready;
1660 sk->sk_write_space = xs_udp_write_space;
Trond Myklebust482f32e2009-03-11 14:38:00 -04001661 sk->sk_error_report = xs_error_report;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001662 sk->sk_no_check = UDP_CSUM_NORCV;
Trond Myklebustb079fa7b2005-12-13 16:13:52 -05001663 sk->sk_allocation = GFP_ATOMIC;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001664
1665 xprt_set_connected(xprt);
1666
1667 /* Reset to new socket */
Chuck Leveree0ac0c2006-12-05 16:35:15 -05001668 transport->sock = sock;
1669 transport->inet = sk;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001670
1671 write_unlock_bh(&sk->sk_callback_lock);
1672 }
Chuck Lever470056c2005-08-25 16:25:56 -07001673 xs_udp_do_set_buffer_size(xprt);
Chuck Lever16be2d22007-08-06 11:57:38 -04001674}
1675
Pavel Emelyanov8c14ff22010-10-04 16:58:02 +04001676static void xs_udp_setup_socket(struct work_struct *work)
Chuck Levera246b012005-08-11 16:25:23 -04001677{
1678 struct sock_xprt *transport =
1679 container_of(work, struct sock_xprt, connect_worker.work);
1680 struct rpc_xprt *xprt = &transport->xprt;
1681 struct socket *sock = transport->sock;
Pavel Emelyanovb65c0312010-10-04 16:53:46 +04001682 int status = -EIO;
Chuck Levera246b012005-08-11 16:25:23 -04001683
Trond Myklebust01d37c42009-03-11 14:09:39 -04001684 if (xprt->shutdown)
Chuck Lever9903cd12005-08-11 16:25:26 -04001685 goto out;
1686
Chuck Levera246b012005-08-11 16:25:23 -04001687 /* Start by resetting any existing state */
Chuck Leverfe315e72009-03-11 14:10:21 -04001688 xs_reset_transport(transport);
Pavel Emelyanov8c14ff22010-10-04 16:58:02 +04001689 sock = xs_create_sock(xprt, transport,
1690 xs_addr(xprt)->sa_family, SOCK_DGRAM, IPPROTO_UDP);
Pavel Emelyanovb65c0312010-10-04 16:53:46 +04001691 if (IS_ERR(sock))
Chuck Levera246b012005-08-11 16:25:23 -04001692 goto out;
Chuck Lever68e220b2007-08-06 11:57:48 -04001693
Chuck Leverc740eff2009-08-09 15:09:46 -04001694 dprintk("RPC: worker connecting xprt %p via %s to "
1695 "%s (port %s)\n", xprt,
1696 xprt->address_strings[RPC_DISPLAY_PROTO],
1697 xprt->address_strings[RPC_DISPLAY_ADDR],
1698 xprt->address_strings[RPC_DISPLAY_PORT]);
Chuck Lever68e220b2007-08-06 11:57:48 -04001699
1700 xs_udp_finish_connecting(xprt, sock);
Chuck Levera246b012005-08-11 16:25:23 -04001701 status = 0;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001702out:
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001703 xprt_clear_connecting(xprt);
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001704 xprt_wake_pending_tasks(xprt, status);
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001705}
1706
Chuck Lever3167e122005-08-25 16:25:55 -07001707/*
1708 * We need to preserve the port number so the reply cache on the server can
1709 * find our cached RPC replies when we get around to reconnecting.
1710 */
Pavel Emelyanov58dddac2010-10-04 16:52:55 +04001711static void xs_abort_connection(struct sock_xprt *transport)
Chuck Lever3167e122005-08-25 16:25:55 -07001712{
1713 int result;
Chuck Lever3167e122005-08-25 16:25:55 -07001714 struct sockaddr any;
1715
Pavel Emelyanov58dddac2010-10-04 16:52:55 +04001716 dprintk("RPC: disconnecting xprt %p to reuse port\n", transport);
Chuck Lever3167e122005-08-25 16:25:55 -07001717
1718 /*
1719 * Disconnect the transport socket by doing a connect operation
1720 * with AF_UNSPEC. This should return immediately...
1721 */
1722 memset(&any, 0, sizeof(any));
1723 any.sa_family = AF_UNSPEC;
Chuck Leveree0ac0c2006-12-05 16:35:15 -05001724 result = kernel_connect(transport->sock, &any, sizeof(any), 0);
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001725 if (!result)
Pavel Emelyanov58dddac2010-10-04 16:52:55 +04001726 xs_sock_mark_closed(&transport->xprt);
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001727 else
Chuck Lever46121cf2007-01-31 12:14:08 -05001728 dprintk("RPC: AF_UNSPEC connect return code %d\n",
Chuck Lever3167e122005-08-25 16:25:55 -07001729 result);
1730}
1731
Pavel Emelyanov58dddac2010-10-04 16:52:55 +04001732static void xs_tcp_reuse_connection(struct sock_xprt *transport)
Trond Myklebust40d25492009-03-11 14:37:58 -04001733{
1734 unsigned int state = transport->inet->sk_state;
1735
Andy Chittenden669502f2010-08-10 10:19:53 -04001736 if (state == TCP_CLOSE && transport->sock->state == SS_UNCONNECTED) {
1737 /* we don't need to abort the connection if the socket
1738 * hasn't undergone a shutdown
1739 */
1740 if (transport->inet->sk_shutdown == 0)
1741 return;
1742 dprintk("RPC: %s: TCP_CLOSEd and sk_shutdown set to %d\n",
1743 __func__, transport->inet->sk_shutdown);
1744 }
1745 if ((1 << state) & (TCPF_ESTABLISHED|TCPF_SYN_SENT)) {
1746 /* we don't need to abort the connection if the socket
1747 * hasn't undergone a shutdown
1748 */
1749 if (transport->inet->sk_shutdown == 0)
1750 return;
1751 dprintk("RPC: %s: ESTABLISHED/SYN_SENT "
1752 "sk_shutdown set to %d\n",
1753 __func__, transport->inet->sk_shutdown);
1754 }
Pavel Emelyanov58dddac2010-10-04 16:52:55 +04001755 xs_abort_connection(transport);
Trond Myklebust40d25492009-03-11 14:37:58 -04001756}
1757
Chuck Lever16be2d22007-08-06 11:57:38 -04001758static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001759{
Chuck Lever16be2d22007-08-06 11:57:38 -04001760 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Leveredb267a2006-08-22 20:06:18 -04001761
Chuck Leveree0ac0c2006-12-05 16:35:15 -05001762 if (!transport->inet) {
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001763 struct sock *sk = sock->sk;
1764
1765 write_lock_bh(&sk->sk_callback_lock);
1766
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04001767 xs_save_old_callbacks(transport, sk);
1768
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001769 sk->sk_user_data = xprt;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001770 sk->sk_data_ready = xs_tcp_data_ready;
1771 sk->sk_state_change = xs_tcp_state_change;
1772 sk->sk_write_space = xs_tcp_write_space;
Trond Myklebust482f32e2009-03-11 14:38:00 -04001773 sk->sk_error_report = xs_error_report;
Trond Myklebustb079fa7b2005-12-13 16:13:52 -05001774 sk->sk_allocation = GFP_ATOMIC;
Chuck Lever3167e122005-08-25 16:25:55 -07001775
1776 /* socket options */
1777 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
1778 sock_reset_flag(sk, SOCK_LINGER);
1779 tcp_sk(sk)->linger2 = 0;
1780 tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001781
1782 xprt_clear_connected(xprt);
1783
1784 /* Reset to new socket */
Chuck Leveree0ac0c2006-12-05 16:35:15 -05001785 transport->sock = sock;
1786 transport->inet = sk;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001787
1788 write_unlock_bh(&sk->sk_callback_lock);
1789 }
1790
Trond Myklebust01d37c42009-03-11 14:09:39 -04001791 if (!xprt_bound(xprt))
1792 return -ENOTCONN;
1793
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001794 /* Tell the socket layer to start connecting... */
Chuck Lever262ca072006-03-20 13:44:16 -05001795 xprt->stat.connect_count++;
1796 xprt->stat.connect_start = jiffies;
Chuck Lever95392c52007-08-06 11:57:58 -04001797 return kernel_connect(sock, xs_addr(xprt), xprt->addrlen, O_NONBLOCK);
Chuck Lever16be2d22007-08-06 11:57:38 -04001798}
1799
1800/**
Trond Myklebustb61d59f2009-03-11 14:38:04 -04001801 * xs_tcp_setup_socket - create a TCP socket and connect to a remote endpoint
1802 * @xprt: RPC transport to connect
1803 * @transport: socket transport to connect
1804 * @create_sock: function to create a socket of the correct type
Chuck Lever16be2d22007-08-06 11:57:38 -04001805 *
1806 * Invoked by a work queue tasklet.
1807 */
Pavel Emelyanovcdd518d2010-10-04 16:57:40 +04001808static void xs_tcp_setup_socket(struct work_struct *work)
Chuck Lever16be2d22007-08-06 11:57:38 -04001809{
Pavel Emelyanovcdd518d2010-10-04 16:57:40 +04001810 struct sock_xprt *transport =
1811 container_of(work, struct sock_xprt, connect_worker.work);
Chuck Lever16be2d22007-08-06 11:57:38 -04001812 struct socket *sock = transport->sock;
Pavel Emelyanova9f5f0f2010-10-04 16:52:25 +04001813 struct rpc_xprt *xprt = &transport->xprt;
Trond Myklebustb61d59f2009-03-11 14:38:04 -04001814 int status = -EIO;
Chuck Lever16be2d22007-08-06 11:57:38 -04001815
Trond Myklebust01d37c42009-03-11 14:09:39 -04001816 if (xprt->shutdown)
Chuck Lever16be2d22007-08-06 11:57:38 -04001817 goto out;
1818
1819 if (!sock) {
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001820 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
Pavel Emelyanovcdd518d2010-10-04 16:57:40 +04001821 sock = xs_create_sock(xprt, transport,
1822 xs_addr(xprt)->sa_family, SOCK_STREAM, IPPROTO_TCP);
Trond Myklebustb61d59f2009-03-11 14:38:04 -04001823 if (IS_ERR(sock)) {
1824 status = PTR_ERR(sock);
Chuck Lever16be2d22007-08-06 11:57:38 -04001825 goto out;
1826 }
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001827 } else {
1828 int abort_and_exit;
1829
1830 abort_and_exit = test_and_clear_bit(XPRT_CONNECTION_ABORT,
1831 &xprt->state);
Chuck Lever16be2d22007-08-06 11:57:38 -04001832 /* "close" the socket, preserving the local port */
Pavel Emelyanov58dddac2010-10-04 16:52:55 +04001833 xs_tcp_reuse_connection(transport);
Chuck Lever16be2d22007-08-06 11:57:38 -04001834
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001835 if (abort_and_exit)
1836 goto out_eagain;
1837 }
1838
Chuck Leverc740eff2009-08-09 15:09:46 -04001839 dprintk("RPC: worker connecting xprt %p via %s to "
1840 "%s (port %s)\n", xprt,
1841 xprt->address_strings[RPC_DISPLAY_PROTO],
1842 xprt->address_strings[RPC_DISPLAY_ADDR],
1843 xprt->address_strings[RPC_DISPLAY_PORT]);
Chuck Lever16be2d22007-08-06 11:57:38 -04001844
1845 status = xs_tcp_finish_connecting(xprt, sock);
Chuck Lever46121cf2007-01-31 12:14:08 -05001846 dprintk("RPC: %p connect status %d connected %d sock state %d\n",
1847 xprt, -status, xprt_connected(xprt),
1848 sock->sk->sk_state);
Trond Myklebust2a491992009-03-11 14:38:00 -04001849 switch (status) {
Trond Myklebustf75e6742009-04-21 17:18:20 -04001850 default:
1851 printk("%s: connect returned unhandled error %d\n",
1852 __func__, status);
1853 case -EADDRNOTAVAIL:
1854 /* We're probably in TIME_WAIT. Get rid of existing socket,
1855 * and retry
1856 */
1857 set_bit(XPRT_CONNECTION_CLOSE, &xprt->state);
1858 xprt_force_disconnect(xprt);
Trond Myklebust88b5ed72009-06-17 13:22:57 -07001859 break;
Trond Myklebust8a2cec22009-03-11 14:38:01 -04001860 case -ECONNREFUSED:
1861 case -ECONNRESET:
1862 case -ENETUNREACH:
1863 /* retry with existing socket, after a delay */
Trond Myklebust2a491992009-03-11 14:38:00 -04001864 case 0:
1865 case -EINPROGRESS:
1866 case -EALREADY:
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001867 xprt_clear_connecting(xprt);
1868 return;
Trond Myklebust9fcfe0c2010-03-02 13:06:21 -05001869 case -EINVAL:
1870 /* Happens, for instance, if the user specified a link
1871 * local IPv6 address without a scope-id.
1872 */
1873 goto out;
Chuck Levera246b012005-08-11 16:25:23 -04001874 }
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001875out_eagain:
Trond Myklebust2a491992009-03-11 14:38:00 -04001876 status = -EAGAIN;
Chuck Levera246b012005-08-11 16:25:23 -04001877out:
Chuck Lever2226feb2005-08-11 16:25:38 -04001878 xprt_clear_connecting(xprt);
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001879 xprt_wake_pending_tasks(xprt, status);
Chuck Levera246b012005-08-11 16:25:23 -04001880}
1881
Chuck Lever68e220b2007-08-06 11:57:48 -04001882/**
Chuck Lever9903cd12005-08-11 16:25:26 -04001883 * xs_connect - connect a socket to a remote endpoint
1884 * @task: address of RPC task that manages state of connect request
1885 *
1886 * TCP: If the remote end dropped the connection, delay reconnecting.
Chuck Lever03bf4b72005-08-25 16:25:55 -07001887 *
1888 * UDP socket connects are synchronous, but we use a work queue anyway
1889 * to guarantee that even unprivileged user processes can set up a
1890 * socket on a privileged port.
1891 *
1892 * If a UDP socket connect fails, the delay behavior here prevents
1893 * retry floods (hard mounts).
Chuck Lever9903cd12005-08-11 16:25:26 -04001894 */
1895static void xs_connect(struct rpc_task *task)
Chuck Levera246b012005-08-11 16:25:23 -04001896{
1897 struct rpc_xprt *xprt = task->tk_xprt;
Chuck Leveree0ac0c2006-12-05 16:35:15 -05001898 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -04001899
Chuck Lever09a21c42009-12-03 15:58:56 -05001900 if (transport->sock != NULL && !RPC_IS_SOFTCONN(task)) {
Chuck Lever46121cf2007-01-31 12:14:08 -05001901 dprintk("RPC: xs_connect delayed xprt %p for %lu "
1902 "seconds\n",
Chuck Lever03bf4b72005-08-25 16:25:55 -07001903 xprt, xprt->reestablish_timeout / HZ);
Trond Myklebustc1384c92007-06-14 18:00:42 -04001904 queue_delayed_work(rpciod_workqueue,
1905 &transport->connect_worker,
1906 xprt->reestablish_timeout);
Chuck Lever03bf4b72005-08-25 16:25:55 -07001907 xprt->reestablish_timeout <<= 1;
Neil Brown61d0a8e2009-09-23 14:36:37 -04001908 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
1909 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
Chuck Lever03bf4b72005-08-25 16:25:55 -07001910 if (xprt->reestablish_timeout > XS_TCP_MAX_REEST_TO)
1911 xprt->reestablish_timeout = XS_TCP_MAX_REEST_TO;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001912 } else {
Chuck Lever46121cf2007-01-31 12:14:08 -05001913 dprintk("RPC: xs_connect scheduled xprt %p\n", xprt);
Trond Myklebustc1384c92007-06-14 18:00:42 -04001914 queue_delayed_work(rpciod_workqueue,
1915 &transport->connect_worker, 0);
Chuck Levera246b012005-08-11 16:25:23 -04001916 }
1917}
1918
Chuck Lever262ca072006-03-20 13:44:16 -05001919/**
1920 * xs_udp_print_stats - display UDP socket-specifc stats
1921 * @xprt: rpc_xprt struct containing statistics
1922 * @seq: output file
1923 *
1924 */
1925static void xs_udp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
1926{
Chuck Leverc8475462006-12-05 16:35:26 -05001927 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1928
Chuck Lever262ca072006-03-20 13:44:16 -05001929 seq_printf(seq, "\txprt:\tudp %u %lu %lu %lu %lu %Lu %Lu\n",
Chuck Leverfbfffbd2009-08-09 15:09:46 -04001930 transport->srcport,
Chuck Lever262ca072006-03-20 13:44:16 -05001931 xprt->stat.bind_count,
1932 xprt->stat.sends,
1933 xprt->stat.recvs,
1934 xprt->stat.bad_xids,
1935 xprt->stat.req_u,
1936 xprt->stat.bklog_u);
1937}
1938
1939/**
1940 * xs_tcp_print_stats - display TCP socket-specifc stats
1941 * @xprt: rpc_xprt struct containing statistics
1942 * @seq: output file
1943 *
1944 */
1945static void xs_tcp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
1946{
Chuck Leverc8475462006-12-05 16:35:26 -05001947 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Lever262ca072006-03-20 13:44:16 -05001948 long idle_time = 0;
1949
1950 if (xprt_connected(xprt))
1951 idle_time = (long)(jiffies - xprt->last_used) / HZ;
1952
1953 seq_printf(seq, "\txprt:\ttcp %u %lu %lu %lu %ld %lu %lu %lu %Lu %Lu\n",
Chuck Leverfbfffbd2009-08-09 15:09:46 -04001954 transport->srcport,
Chuck Lever262ca072006-03-20 13:44:16 -05001955 xprt->stat.bind_count,
1956 xprt->stat.connect_count,
1957 xprt->stat.connect_time,
1958 idle_time,
1959 xprt->stat.sends,
1960 xprt->stat.recvs,
1961 xprt->stat.bad_xids,
1962 xprt->stat.req_u,
1963 xprt->stat.bklog_u);
1964}
1965
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03001966/*
1967 * Allocate a bunch of pages for a scratch buffer for the rpc code. The reason
1968 * we allocate pages instead doing a kmalloc like rpc_malloc is because we want
1969 * to use the server side send routines.
1970 */
H Hartley Sweeten5a51f132010-01-14 15:38:31 -07001971static void *bc_malloc(struct rpc_task *task, size_t size)
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03001972{
1973 struct page *page;
1974 struct rpc_buffer *buf;
1975
1976 BUG_ON(size > PAGE_SIZE - sizeof(struct rpc_buffer));
1977 page = alloc_page(GFP_KERNEL);
1978
1979 if (!page)
1980 return NULL;
1981
1982 buf = page_address(page);
1983 buf->len = PAGE_SIZE;
1984
1985 return buf->data;
1986}
1987
1988/*
1989 * Free the space allocated in the bc_alloc routine
1990 */
H Hartley Sweeten5a51f132010-01-14 15:38:31 -07001991static void bc_free(void *buffer)
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03001992{
1993 struct rpc_buffer *buf;
1994
1995 if (!buffer)
1996 return;
1997
1998 buf = container_of(buffer, struct rpc_buffer, data);
1999 free_page((unsigned long)buf);
2000}
2001
2002/*
2003 * Use the svc_sock to send the callback. Must be called with svsk->sk_mutex
2004 * held. Borrows heavily from svc_tcp_sendto and xs_tcp_send_request.
2005 */
2006static int bc_sendto(struct rpc_rqst *req)
2007{
2008 int len;
2009 struct xdr_buf *xbufp = &req->rq_snd_buf;
2010 struct rpc_xprt *xprt = req->rq_xprt;
2011 struct sock_xprt *transport =
2012 container_of(xprt, struct sock_xprt, xprt);
2013 struct socket *sock = transport->sock;
2014 unsigned long headoff;
2015 unsigned long tailoff;
2016
2017 /*
2018 * Set up the rpc header and record marker stuff
2019 */
2020 xs_encode_tcp_record_marker(xbufp);
2021
2022 tailoff = (unsigned long)xbufp->tail[0].iov_base & ~PAGE_MASK;
2023 headoff = (unsigned long)xbufp->head[0].iov_base & ~PAGE_MASK;
2024 len = svc_send_common(sock, xbufp,
2025 virt_to_page(xbufp->head[0].iov_base), headoff,
2026 xbufp->tail[0].iov_base, tailoff);
2027
2028 if (len != xbufp->len) {
2029 printk(KERN_NOTICE "Error sending entire callback!\n");
2030 len = -EAGAIN;
2031 }
2032
2033 return len;
2034}
2035
2036/*
2037 * The send routine. Borrows from svc_send
2038 */
2039static int bc_send_request(struct rpc_task *task)
2040{
2041 struct rpc_rqst *req = task->tk_rqstp;
2042 struct svc_xprt *xprt;
2043 struct svc_sock *svsk;
2044 u32 len;
2045
2046 dprintk("sending request with xid: %08x\n", ntohl(req->rq_xid));
2047 /*
2048 * Get the server socket associated with this callback xprt
2049 */
2050 xprt = req->rq_xprt->bc_xprt;
2051 svsk = container_of(xprt, struct svc_sock, sk_xprt);
2052
2053 /*
2054 * Grab the mutex to serialize data as the connection is shared
2055 * with the fore channel
2056 */
2057 if (!mutex_trylock(&xprt->xpt_mutex)) {
2058 rpc_sleep_on(&xprt->xpt_bc_pending, task, NULL);
2059 if (!mutex_trylock(&xprt->xpt_mutex))
2060 return -EAGAIN;
2061 rpc_wake_up_queued_task(&xprt->xpt_bc_pending, task);
2062 }
2063 if (test_bit(XPT_DEAD, &xprt->xpt_flags))
2064 len = -ENOTCONN;
2065 else
2066 len = bc_sendto(req);
2067 mutex_unlock(&xprt->xpt_mutex);
2068
2069 if (len > 0)
2070 len = 0;
2071
2072 return len;
2073}
2074
2075/*
2076 * The close routine. Since this is client initiated, we do nothing
2077 */
2078
2079static void bc_close(struct rpc_xprt *xprt)
2080{
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002081}
2082
2083/*
2084 * The xprt destroy routine. Again, because this connection is client
2085 * initiated, we do nothing
2086 */
2087
2088static void bc_destroy(struct rpc_xprt *xprt)
2089{
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002090}
2091
Chuck Lever262965f2005-08-11 16:25:56 -04002092static struct rpc_xprt_ops xs_udp_ops = {
Chuck Lever43118c22005-08-25 16:25:49 -07002093 .set_buffer_size = xs_udp_set_buffer_size,
Chuck Lever12a80462005-08-25 16:25:51 -07002094 .reserve_xprt = xprt_reserve_xprt_cong,
Chuck Lever49e9a892005-08-25 16:25:51 -07002095 .release_xprt = xprt_release_xprt_cong,
Chuck Lever45160d62007-07-01 12:13:17 -04002096 .rpcbind = rpcb_getport_async,
Chuck Lever92200412006-01-03 09:55:51 +01002097 .set_port = xs_set_port,
Chuck Lever9903cd12005-08-11 16:25:26 -04002098 .connect = xs_connect,
Chuck Lever02107142006-01-03 09:55:49 +01002099 .buf_alloc = rpc_malloc,
2100 .buf_free = rpc_free,
Chuck Lever262965f2005-08-11 16:25:56 -04002101 .send_request = xs_udp_send_request,
Chuck Leverfe3aca22005-08-25 16:25:50 -07002102 .set_retrans_timeout = xprt_set_retrans_timeout_rtt,
Chuck Lever46c0ee82005-08-25 16:25:52 -07002103 .timer = xs_udp_timer,
Chuck Levera58dd392005-08-25 16:25:53 -07002104 .release_request = xprt_release_rqst_cong,
Chuck Lever262965f2005-08-11 16:25:56 -04002105 .close = xs_close,
2106 .destroy = xs_destroy,
Chuck Lever262ca072006-03-20 13:44:16 -05002107 .print_stats = xs_udp_print_stats,
Chuck Lever262965f2005-08-11 16:25:56 -04002108};
2109
2110static struct rpc_xprt_ops xs_tcp_ops = {
Chuck Lever12a80462005-08-25 16:25:51 -07002111 .reserve_xprt = xprt_reserve_xprt,
Trond Myklebuste0ab53d2006-07-27 17:22:50 -04002112 .release_xprt = xs_tcp_release_xprt,
Chuck Lever45160d62007-07-01 12:13:17 -04002113 .rpcbind = rpcb_getport_async,
Chuck Lever92200412006-01-03 09:55:51 +01002114 .set_port = xs_set_port,
Trond Myklebust0b9e7942010-04-16 16:41:57 -04002115 .connect = xs_connect,
Chuck Lever02107142006-01-03 09:55:49 +01002116 .buf_alloc = rpc_malloc,
2117 .buf_free = rpc_free,
Chuck Lever262965f2005-08-11 16:25:56 -04002118 .send_request = xs_tcp_send_request,
Chuck Leverfe3aca22005-08-25 16:25:50 -07002119 .set_retrans_timeout = xprt_set_retrans_timeout_def,
Trond Myklebustf75e6742009-04-21 17:18:20 -04002120 .close = xs_tcp_close,
Chuck Lever9903cd12005-08-11 16:25:26 -04002121 .destroy = xs_destroy,
Chuck Lever262ca072006-03-20 13:44:16 -05002122 .print_stats = xs_tcp_print_stats,
Chuck Levera246b012005-08-11 16:25:23 -04002123};
2124
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002125/*
2126 * The rpc_xprt_ops for the server backchannel
2127 */
2128
2129static struct rpc_xprt_ops bc_tcp_ops = {
2130 .reserve_xprt = xprt_reserve_xprt,
2131 .release_xprt = xprt_release_xprt,
2132 .buf_alloc = bc_malloc,
2133 .buf_free = bc_free,
2134 .send_request = bc_send_request,
2135 .set_retrans_timeout = xprt_set_retrans_timeout_def,
2136 .close = bc_close,
2137 .destroy = bc_destroy,
2138 .print_stats = xs_tcp_print_stats,
2139};
2140
Chuck Lever92476852010-10-20 11:53:01 -04002141static int xs_init_anyaddr(const int family, struct sockaddr *sap)
2142{
2143 static const struct sockaddr_in sin = {
2144 .sin_family = AF_INET,
2145 .sin_addr.s_addr = htonl(INADDR_ANY),
2146 };
2147 static const struct sockaddr_in6 sin6 = {
2148 .sin6_family = AF_INET6,
2149 .sin6_addr = IN6ADDR_ANY_INIT,
2150 };
2151
2152 switch (family) {
2153 case AF_INET:
2154 memcpy(sap, &sin, sizeof(sin));
2155 break;
2156 case AF_INET6:
2157 memcpy(sap, &sin6, sizeof(sin6));
2158 break;
2159 default:
2160 dprintk("RPC: %s: Bad address family\n", __func__);
2161 return -EAFNOSUPPORT;
2162 }
2163 return 0;
2164}
2165
\"Talpey, Thomas\3c341b0b2007-09-10 13:47:07 -04002166static struct rpc_xprt *xs_setup_xprt(struct xprt_create *args,
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002167 unsigned int slot_table_size)
Chuck Leverc8541ec2006-10-17 14:44:27 -04002168{
2169 struct rpc_xprt *xprt;
Chuck Leverffc2e512006-12-05 16:35:11 -05002170 struct sock_xprt *new;
Chuck Leverc8541ec2006-10-17 14:44:27 -04002171
Frank van Maarseveen96802a02007-07-08 13:08:54 +02002172 if (args->addrlen > sizeof(xprt->addr)) {
Chuck Lever46121cf2007-01-31 12:14:08 -05002173 dprintk("RPC: xs_setup_xprt: address too large\n");
Chuck Leverc8541ec2006-10-17 14:44:27 -04002174 return ERR_PTR(-EBADF);
2175 }
2176
Pavel Emelyanov37aa2132010-09-29 16:05:43 +04002177 xprt = xprt_alloc(args->net, sizeof(*new), slot_table_size);
Pavel Emelyanovbd1722d2010-09-29 16:02:43 +04002178 if (xprt == NULL) {
Chuck Lever46121cf2007-01-31 12:14:08 -05002179 dprintk("RPC: xs_setup_xprt: couldn't allocate "
2180 "rpc_xprt\n");
Chuck Leverc8541ec2006-10-17 14:44:27 -04002181 return ERR_PTR(-ENOMEM);
2182 }
2183
Pavel Emelyanovbd1722d2010-09-29 16:02:43 +04002184 new = container_of(xprt, struct sock_xprt, xprt);
Frank van Maarseveen96802a02007-07-08 13:08:54 +02002185 memcpy(&xprt->addr, args->dstaddr, args->addrlen);
2186 xprt->addrlen = args->addrlen;
Frank van Maarseveend3bc9a12007-07-09 22:23:35 +02002187 if (args->srcaddr)
Chuck Leverfbfffbd2009-08-09 15:09:46 -04002188 memcpy(&new->srcaddr, args->srcaddr, args->addrlen);
Chuck Lever92476852010-10-20 11:53:01 -04002189 else {
2190 int err;
2191 err = xs_init_anyaddr(args->dstaddr->sa_family,
2192 (struct sockaddr *)&new->srcaddr);
2193 if (err != 0)
2194 return ERR_PTR(err);
2195 }
Chuck Leverc8541ec2006-10-17 14:44:27 -04002196
2197 return xprt;
2198}
2199
Trond Myklebust2881ae72007-12-20 16:03:54 -05002200static const struct rpc_timeout xs_udp_default_timeout = {
2201 .to_initval = 5 * HZ,
2202 .to_maxval = 30 * HZ,
2203 .to_increment = 5 * HZ,
2204 .to_retries = 5,
2205};
2206
Chuck Lever9903cd12005-08-11 16:25:26 -04002207/**
2208 * xs_setup_udp - Set up transport to use a UDP socket
Frank van Maarseveen96802a02007-07-08 13:08:54 +02002209 * @args: rpc transport creation arguments
Chuck Lever9903cd12005-08-11 16:25:26 -04002210 *
2211 */
Adrian Bunk483066d2007-10-24 18:24:02 +02002212static struct rpc_xprt *xs_setup_udp(struct xprt_create *args)
Chuck Levera246b012005-08-11 16:25:23 -04002213{
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002214 struct sockaddr *addr = args->dstaddr;
Chuck Leverc8541ec2006-10-17 14:44:27 -04002215 struct rpc_xprt *xprt;
Chuck Leverc8475462006-12-05 16:35:26 -05002216 struct sock_xprt *transport;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002217 struct rpc_xprt *ret;
Chuck Levera246b012005-08-11 16:25:23 -04002218
Frank van Maarseveen96802a02007-07-08 13:08:54 +02002219 xprt = xs_setup_xprt(args, xprt_udp_slot_table_entries);
Chuck Leverc8541ec2006-10-17 14:44:27 -04002220 if (IS_ERR(xprt))
2221 return xprt;
Chuck Leverc8475462006-12-05 16:35:26 -05002222 transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -04002223
Chuck Leverec739ef2006-08-22 20:06:15 -04002224 xprt->prot = IPPROTO_UDP;
Chuck Lever808012f2005-08-25 16:25:49 -07002225 xprt->tsh_size = 0;
Chuck Levera246b012005-08-11 16:25:23 -04002226 /* XXX: header size can vary due to auth type, IPv6, etc. */
2227 xprt->max_payload = (1U << 16) - (MAX_HEADER << 3);
2228
Chuck Lever03bf4b72005-08-25 16:25:55 -07002229 xprt->bind_timeout = XS_BIND_TO;
Chuck Lever03bf4b72005-08-25 16:25:55 -07002230 xprt->reestablish_timeout = XS_UDP_REEST_TO;
2231 xprt->idle_timeout = XS_IDLE_DISC_TO;
Chuck Levera246b012005-08-11 16:25:23 -04002232
Chuck Lever262965f2005-08-11 16:25:56 -04002233 xprt->ops = &xs_udp_ops;
Chuck Levera246b012005-08-11 16:25:23 -04002234
Trond Myklebustba7392b2007-12-20 16:03:55 -05002235 xprt->timeout = &xs_udp_default_timeout;
Chuck Levera246b012005-08-11 16:25:23 -04002236
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002237 switch (addr->sa_family) {
2238 case AF_INET:
2239 if (((struct sockaddr_in *)addr)->sin_port != htons(0))
2240 xprt_set_bound(xprt);
2241
2242 INIT_DELAYED_WORK(&transport->connect_worker,
Pavel Emelyanov8c14ff22010-10-04 16:58:02 +04002243 xs_udp_setup_socket);
Chuck Lever9dc3b092009-08-09 15:09:46 -04002244 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP);
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002245 break;
2246 case AF_INET6:
2247 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
2248 xprt_set_bound(xprt);
2249
2250 INIT_DELAYED_WORK(&transport->connect_worker,
Pavel Emelyanov8c14ff22010-10-04 16:58:02 +04002251 xs_udp_setup_socket);
Chuck Lever9dc3b092009-08-09 15:09:46 -04002252 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP6);
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002253 break;
2254 default:
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002255 ret = ERR_PTR(-EAFNOSUPPORT);
2256 goto out_err;
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002257 }
2258
Chuck Leverc740eff2009-08-09 15:09:46 -04002259 if (xprt_bound(xprt))
2260 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
2261 xprt->address_strings[RPC_DISPLAY_ADDR],
2262 xprt->address_strings[RPC_DISPLAY_PORT],
2263 xprt->address_strings[RPC_DISPLAY_PROTO]);
2264 else
2265 dprintk("RPC: set up xprt to %s (autobind) via %s\n",
2266 xprt->address_strings[RPC_DISPLAY_ADDR],
2267 xprt->address_strings[RPC_DISPLAY_PROTO]);
Chuck Leveredb267a2006-08-22 20:06:18 -04002268
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002269 if (try_module_get(THIS_MODULE))
2270 return xprt;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002271 ret = ERR_PTR(-EINVAL);
2272out_err:
Pavel Emelyanove204e622010-09-29 16:03:13 +04002273 xprt_free(xprt);
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002274 return ret;
Chuck Levera246b012005-08-11 16:25:23 -04002275}
2276
Trond Myklebust2881ae72007-12-20 16:03:54 -05002277static const struct rpc_timeout xs_tcp_default_timeout = {
2278 .to_initval = 60 * HZ,
2279 .to_maxval = 60 * HZ,
2280 .to_retries = 2,
2281};
2282
Chuck Lever9903cd12005-08-11 16:25:26 -04002283/**
2284 * xs_setup_tcp - Set up transport to use a TCP socket
Frank van Maarseveen96802a02007-07-08 13:08:54 +02002285 * @args: rpc transport creation arguments
Chuck Lever9903cd12005-08-11 16:25:26 -04002286 *
2287 */
Adrian Bunk483066d2007-10-24 18:24:02 +02002288static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args)
Chuck Levera246b012005-08-11 16:25:23 -04002289{
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002290 struct sockaddr *addr = args->dstaddr;
Chuck Leverc8541ec2006-10-17 14:44:27 -04002291 struct rpc_xprt *xprt;
Chuck Leverc8475462006-12-05 16:35:26 -05002292 struct sock_xprt *transport;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002293 struct rpc_xprt *ret;
Chuck Levera246b012005-08-11 16:25:23 -04002294
Frank van Maarseveen96802a02007-07-08 13:08:54 +02002295 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries);
Chuck Leverc8541ec2006-10-17 14:44:27 -04002296 if (IS_ERR(xprt))
2297 return xprt;
Chuck Leverc8475462006-12-05 16:35:26 -05002298 transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -04002299
Chuck Leverec739ef2006-08-22 20:06:15 -04002300 xprt->prot = IPPROTO_TCP;
Chuck Lever808012f2005-08-25 16:25:49 -07002301 xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
Chuck Lever808012f2005-08-25 16:25:49 -07002302 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
Chuck Levera246b012005-08-11 16:25:23 -04002303
Chuck Lever03bf4b72005-08-25 16:25:55 -07002304 xprt->bind_timeout = XS_BIND_TO;
Chuck Lever03bf4b72005-08-25 16:25:55 -07002305 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2306 xprt->idle_timeout = XS_IDLE_DISC_TO;
Chuck Levera246b012005-08-11 16:25:23 -04002307
Chuck Lever262965f2005-08-11 16:25:56 -04002308 xprt->ops = &xs_tcp_ops;
Trond Myklebustba7392b2007-12-20 16:03:55 -05002309 xprt->timeout = &xs_tcp_default_timeout;
Chuck Levera246b012005-08-11 16:25:23 -04002310
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002311 switch (addr->sa_family) {
2312 case AF_INET:
2313 if (((struct sockaddr_in *)addr)->sin_port != htons(0))
2314 xprt_set_bound(xprt);
2315
Chuck Lever9dc3b092009-08-09 15:09:46 -04002316 INIT_DELAYED_WORK(&transport->connect_worker,
Pavel Emelyanovcdd518d2010-10-04 16:57:40 +04002317 xs_tcp_setup_socket);
Chuck Lever9dc3b092009-08-09 15:09:46 -04002318 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP);
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002319 break;
2320 case AF_INET6:
2321 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
2322 xprt_set_bound(xprt);
2323
Chuck Lever9dc3b092009-08-09 15:09:46 -04002324 INIT_DELAYED_WORK(&transport->connect_worker,
Pavel Emelyanovcdd518d2010-10-04 16:57:40 +04002325 xs_tcp_setup_socket);
Chuck Lever9dc3b092009-08-09 15:09:46 -04002326 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP6);
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002327 break;
2328 default:
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002329 ret = ERR_PTR(-EAFNOSUPPORT);
2330 goto out_err;
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002331 }
2332
Chuck Leverc740eff2009-08-09 15:09:46 -04002333 if (xprt_bound(xprt))
2334 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
2335 xprt->address_strings[RPC_DISPLAY_ADDR],
2336 xprt->address_strings[RPC_DISPLAY_PORT],
2337 xprt->address_strings[RPC_DISPLAY_PROTO]);
2338 else
2339 dprintk("RPC: set up xprt to %s (autobind) via %s\n",
2340 xprt->address_strings[RPC_DISPLAY_ADDR],
2341 xprt->address_strings[RPC_DISPLAY_PROTO]);
2342
Chuck Leveredb267a2006-08-22 20:06:18 -04002343
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002344 if (try_module_get(THIS_MODULE))
2345 return xprt;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002346 ret = ERR_PTR(-EINVAL);
2347out_err:
Pavel Emelyanove204e622010-09-29 16:03:13 +04002348 xprt_free(xprt);
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002349 return ret;
Chuck Levera246b012005-08-11 16:25:23 -04002350}
Chuck Lever282b32e2006-12-05 16:35:51 -05002351
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002352/**
2353 * xs_setup_bc_tcp - Set up transport to use a TCP backchannel socket
2354 * @args: rpc transport creation arguments
2355 *
2356 */
2357static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args)
2358{
2359 struct sockaddr *addr = args->dstaddr;
2360 struct rpc_xprt *xprt;
2361 struct sock_xprt *transport;
2362 struct svc_sock *bc_sock;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002363 struct rpc_xprt *ret;
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002364
J. Bruce Fieldsf0418aa2010-12-08 13:48:19 -05002365 if (args->bc_xprt->xpt_bc_xprt) {
2366 /*
2367 * This server connection already has a backchannel
2368 * export; we can't create a new one, as we wouldn't be
2369 * able to match replies based on xid any more. So,
2370 * reuse the already-existing one:
2371 */
2372 return args->bc_xprt->xpt_bc_xprt;
2373 }
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002374 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries);
2375 if (IS_ERR(xprt))
2376 return xprt;
2377 transport = container_of(xprt, struct sock_xprt, xprt);
2378
2379 xprt->prot = IPPROTO_TCP;
2380 xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
2381 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
2382 xprt->timeout = &xs_tcp_default_timeout;
2383
2384 /* backchannel */
2385 xprt_set_bound(xprt);
2386 xprt->bind_timeout = 0;
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002387 xprt->reestablish_timeout = 0;
2388 xprt->idle_timeout = 0;
2389
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002390 xprt->ops = &bc_tcp_ops;
2391
2392 switch (addr->sa_family) {
2393 case AF_INET:
2394 xs_format_peer_addresses(xprt, "tcp",
2395 RPCBIND_NETID_TCP);
2396 break;
2397 case AF_INET6:
2398 xs_format_peer_addresses(xprt, "tcp",
2399 RPCBIND_NETID_TCP6);
2400 break;
2401 default:
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002402 ret = ERR_PTR(-EAFNOSUPPORT);
2403 goto out_err;
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002404 }
2405
Pavel Emelyanov50fa0d42010-10-05 20:49:35 +04002406 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
2407 xprt->address_strings[RPC_DISPLAY_ADDR],
2408 xprt->address_strings[RPC_DISPLAY_PORT],
2409 xprt->address_strings[RPC_DISPLAY_PROTO]);
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002410
2411 /*
J. Bruce Fields99de8ea2010-12-08 12:45:44 -05002412 * Once we've associated a backchannel xprt with a connection,
2413 * we want to keep it around as long as long as the connection
2414 * lasts, in case we need to start using it for a backchannel
2415 * again; this reference won't be dropped until bc_xprt is
2416 * destroyed.
2417 */
2418 xprt_get(xprt);
2419 args->bc_xprt->xpt_bc_xprt = xprt;
2420 xprt->bc_xprt = args->bc_xprt;
2421 bc_sock = container_of(args->bc_xprt, struct svc_sock, sk_xprt);
2422 transport->sock = bc_sock->sk_sock;
2423 transport->inet = bc_sock->sk_sk;
2424
2425 /*
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002426 * Since we don't want connections for the backchannel, we set
2427 * the xprt status to connected
2428 */
2429 xprt_set_connected(xprt);
2430
2431
2432 if (try_module_get(THIS_MODULE))
2433 return xprt;
J. Bruce Fields99de8ea2010-12-08 12:45:44 -05002434 xprt_put(xprt);
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002435 ret = ERR_PTR(-EINVAL);
2436out_err:
Pavel Emelyanove204e622010-09-29 16:03:13 +04002437 xprt_free(xprt);
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002438 return ret;
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002439}
2440
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002441static struct xprt_class xs_udp_transport = {
2442 .list = LIST_HEAD_INIT(xs_udp_transport.list),
2443 .name = "udp",
2444 .owner = THIS_MODULE,
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002445 .ident = XPRT_TRANSPORT_UDP,
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002446 .setup = xs_setup_udp,
2447};
2448
2449static struct xprt_class xs_tcp_transport = {
2450 .list = LIST_HEAD_INIT(xs_tcp_transport.list),
2451 .name = "tcp",
2452 .owner = THIS_MODULE,
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002453 .ident = XPRT_TRANSPORT_TCP,
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002454 .setup = xs_setup_tcp,
2455};
2456
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002457static struct xprt_class xs_bc_tcp_transport = {
2458 .list = LIST_HEAD_INIT(xs_bc_tcp_transport.list),
2459 .name = "tcp NFSv4.1 backchannel",
2460 .owner = THIS_MODULE,
2461 .ident = XPRT_TRANSPORT_BC_TCP,
2462 .setup = xs_setup_bc_tcp,
2463};
2464
Chuck Lever282b32e2006-12-05 16:35:51 -05002465/**
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002466 * init_socket_xprt - set up xprtsock's sysctls, register with RPC client
Chuck Lever282b32e2006-12-05 16:35:51 -05002467 *
2468 */
2469int init_socket_xprt(void)
2470{
Chuck Leverfbf76682006-12-05 16:35:54 -05002471#ifdef RPC_DEBUG
Eric W. Biederman2b1bec52007-02-14 00:33:24 -08002472 if (!sunrpc_table_header)
Eric W. Biederman0b4d4142007-02-14 00:34:09 -08002473 sunrpc_table_header = register_sysctl_table(sunrpc_table);
Chuck Leverfbf76682006-12-05 16:35:54 -05002474#endif
2475
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002476 xprt_register_transport(&xs_udp_transport);
2477 xprt_register_transport(&xs_tcp_transport);
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002478 xprt_register_transport(&xs_bc_tcp_transport);
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002479
Chuck Lever282b32e2006-12-05 16:35:51 -05002480 return 0;
2481}
2482
2483/**
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002484 * cleanup_socket_xprt - remove xprtsock's sysctls, unregister
Chuck Lever282b32e2006-12-05 16:35:51 -05002485 *
2486 */
2487void cleanup_socket_xprt(void)
2488{
Chuck Leverfbf76682006-12-05 16:35:54 -05002489#ifdef RPC_DEBUG
2490 if (sunrpc_table_header) {
2491 unregister_sysctl_table(sunrpc_table_header);
2492 sunrpc_table_header = NULL;
2493 }
2494#endif
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002495
2496 xprt_unregister_transport(&xs_udp_transport);
2497 xprt_unregister_transport(&xs_tcp_transport);
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002498 xprt_unregister_transport(&xs_bc_tcp_transport);
Chuck Lever282b32e2006-12-05 16:35:51 -05002499}
Trond Myklebustcbf11072009-08-09 15:06:19 -04002500
Rusty Russell9bbb9e52010-08-11 23:04:12 -06002501static int param_set_uint_minmax(const char *val,
2502 const struct kernel_param *kp,
Trond Myklebustcbf11072009-08-09 15:06:19 -04002503 unsigned int min, unsigned int max)
2504{
2505 unsigned long num;
2506 int ret;
2507
2508 if (!val)
2509 return -EINVAL;
2510 ret = strict_strtoul(val, 0, &num);
2511 if (ret == -EINVAL || num < min || num > max)
2512 return -EINVAL;
2513 *((unsigned int *)kp->arg) = num;
2514 return 0;
2515}
2516
Rusty Russell9bbb9e52010-08-11 23:04:12 -06002517static int param_set_portnr(const char *val, const struct kernel_param *kp)
Trond Myklebustcbf11072009-08-09 15:06:19 -04002518{
2519 return param_set_uint_minmax(val, kp,
2520 RPC_MIN_RESVPORT,
2521 RPC_MAX_RESVPORT);
2522}
2523
Rusty Russell9bbb9e52010-08-11 23:04:12 -06002524static struct kernel_param_ops param_ops_portnr = {
2525 .set = param_set_portnr,
2526 .get = param_get_uint,
2527};
2528
Trond Myklebustcbf11072009-08-09 15:06:19 -04002529#define param_check_portnr(name, p) \
2530 __param_check(name, p, unsigned int);
2531
2532module_param_named(min_resvport, xprt_min_resvport, portnr, 0644);
2533module_param_named(max_resvport, xprt_max_resvport, portnr, 0644);
2534
Rusty Russell9bbb9e52010-08-11 23:04:12 -06002535static int param_set_slot_table_size(const char *val,
2536 const struct kernel_param *kp)
Trond Myklebustcbf11072009-08-09 15:06:19 -04002537{
2538 return param_set_uint_minmax(val, kp,
2539 RPC_MIN_SLOT_TABLE,
2540 RPC_MAX_SLOT_TABLE);
2541}
2542
Rusty Russell9bbb9e52010-08-11 23:04:12 -06002543static struct kernel_param_ops param_ops_slot_table_size = {
2544 .set = param_set_slot_table_size,
2545 .get = param_get_uint,
2546};
2547
Trond Myklebustcbf11072009-08-09 15:06:19 -04002548#define param_check_slot_table_size(name, p) \
2549 __param_check(name, p, unsigned int);
2550
2551module_param_named(tcp_slot_table_entries, xprt_tcp_slot_table_entries,
2552 slot_table_size, 0644);
2553module_param_named(udp_slot_table_entries, xprt_udp_slot_table_entries,
2554 slot_table_size, 0644);
2555