blob: 890b03f8d8771ea325cf41be4f9bddd50a07fc12 [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>
Chuck Lever176e21e2011-05-09 15:22:44 -040022#include <linux/string.h>
Chuck Levera246b012005-08-11 16:25:23 -040023#include <linux/slab.h>
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -040024#include <linux/module.h>
Chuck Levera246b012005-08-11 16:25:23 -040025#include <linux/capability.h>
Chuck Levera246b012005-08-11 16:25:23 -040026#include <linux/pagemap.h>
27#include <linux/errno.h>
28#include <linux/socket.h>
29#include <linux/in.h>
30#include <linux/net.h>
31#include <linux/mm.h>
Chuck Lever176e21e2011-05-09 15:22:44 -040032#include <linux/un.h>
Chuck Levera246b012005-08-11 16:25:23 -040033#include <linux/udp.h>
34#include <linux/tcp.h>
35#include <linux/sunrpc/clnt.h>
Chuck Lever02107142006-01-03 09:55:49 +010036#include <linux/sunrpc/sched.h>
Rahul Iyer4cfc7e62009-09-10 17:32:28 +030037#include <linux/sunrpc/svcsock.h>
\"Talpey, Thomas\49c36fc2007-09-10 13:47:31 -040038#include <linux/sunrpc/xprtsock.h>
Chuck Levera246b012005-08-11 16:25:23 -040039#include <linux/file.h>
Trond Myklebust9e00abc2011-07-13 19:20:49 -040040#ifdef CONFIG_SUNRPC_BACKCHANNEL
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -040041#include <linux/sunrpc/bc_xprt.h>
42#endif
Chuck Levera246b012005-08-11 16:25:23 -040043
44#include <net/sock.h>
45#include <net/checksum.h>
46#include <net/udp.h>
47#include <net/tcp.h>
48
Rahul Iyer4cfc7e62009-09-10 17:32:28 +030049#include "sunrpc.h"
Chuck Lever176e21e2011-05-09 15:22:44 -040050
51static void xs_close(struct rpc_xprt *xprt);
52
Chuck Lever9903cd12005-08-11 16:25:26 -040053/*
Chuck Leverc556b752005-11-01 12:24:48 -050054 * xprtsock tunables
55 */
Trond Myklebust09acfea2012-03-11 15:22:54 -040056static unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE;
57static unsigned int xprt_tcp_slot_table_entries = RPC_MIN_SLOT_TABLE;
58static unsigned int xprt_max_tcp_slot_table_entries = RPC_MAX_SLOT_TABLE;
Chuck Leverc556b752005-11-01 12:24:48 -050059
Trond Myklebust09acfea2012-03-11 15:22:54 -040060static unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT;
61static unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT;
Chuck Leverc556b752005-11-01 12:24:48 -050062
Trond Myklebust7d1e8252009-03-11 14:38:03 -040063#define XS_TCP_LINGER_TO (15U * HZ)
Trond Myklebust25fe6142009-03-11 14:38:03 -040064static unsigned int xs_tcp_fin_timeout __read_mostly = XS_TCP_LINGER_TO;
Trond Myklebust7d1e8252009-03-11 14:38:03 -040065
Chuck Leverc556b752005-11-01 12:24:48 -050066/*
Chuck Leverfbf76682006-12-05 16:35:54 -050067 * We can register our own files under /proc/sys/sunrpc by
68 * calling register_sysctl_table() again. The files in that
69 * directory become the union of all files registered there.
70 *
71 * We simply need to make sure that we don't collide with
72 * someone else's file names!
73 */
74
75#ifdef RPC_DEBUG
76
77static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE;
78static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE;
Trond Myklebustd9ba1312011-07-17 18:11:30 -040079static unsigned int max_tcp_slot_table_limit = RPC_MAX_SLOT_TABLE_LIMIT;
Chuck Leverfbf76682006-12-05 16:35:54 -050080static unsigned int xprt_min_resvport_limit = RPC_MIN_RESVPORT;
81static unsigned int xprt_max_resvport_limit = RPC_MAX_RESVPORT;
82
83static struct ctl_table_header *sunrpc_table_header;
84
85/*
86 * FIXME: changing the UDP slot table size should also resize the UDP
87 * socket buffers for existing UDP transports
88 */
89static ctl_table xs_tunables_table[] = {
90 {
Chuck Leverfbf76682006-12-05 16:35:54 -050091 .procname = "udp_slot_table_entries",
92 .data = &xprt_udp_slot_table_entries,
93 .maxlen = sizeof(unsigned int),
94 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -080095 .proc_handler = proc_dointvec_minmax,
Chuck Leverfbf76682006-12-05 16:35:54 -050096 .extra1 = &min_slot_table_size,
97 .extra2 = &max_slot_table_size
98 },
99 {
Chuck Leverfbf76682006-12-05 16:35:54 -0500100 .procname = "tcp_slot_table_entries",
101 .data = &xprt_tcp_slot_table_entries,
102 .maxlen = sizeof(unsigned int),
103 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -0800104 .proc_handler = proc_dointvec_minmax,
Chuck Leverfbf76682006-12-05 16:35:54 -0500105 .extra1 = &min_slot_table_size,
106 .extra2 = &max_slot_table_size
107 },
108 {
Trond Myklebustd9ba1312011-07-17 18:11:30 -0400109 .procname = "tcp_max_slot_table_entries",
110 .data = &xprt_max_tcp_slot_table_entries,
111 .maxlen = sizeof(unsigned int),
112 .mode = 0644,
113 .proc_handler = proc_dointvec_minmax,
114 .extra1 = &min_slot_table_size,
115 .extra2 = &max_tcp_slot_table_limit
116 },
117 {
Chuck Leverfbf76682006-12-05 16:35:54 -0500118 .procname = "min_resvport",
119 .data = &xprt_min_resvport,
120 .maxlen = sizeof(unsigned int),
121 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -0800122 .proc_handler = proc_dointvec_minmax,
Chuck Leverfbf76682006-12-05 16:35:54 -0500123 .extra1 = &xprt_min_resvport_limit,
124 .extra2 = &xprt_max_resvport_limit
125 },
126 {
Chuck Leverfbf76682006-12-05 16:35:54 -0500127 .procname = "max_resvport",
128 .data = &xprt_max_resvport,
129 .maxlen = sizeof(unsigned int),
130 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -0800131 .proc_handler = proc_dointvec_minmax,
Chuck Leverfbf76682006-12-05 16:35:54 -0500132 .extra1 = &xprt_min_resvport_limit,
133 .extra2 = &xprt_max_resvport_limit
134 },
135 {
Trond Myklebust25fe6142009-03-11 14:38:03 -0400136 .procname = "tcp_fin_timeout",
137 .data = &xs_tcp_fin_timeout,
138 .maxlen = sizeof(xs_tcp_fin_timeout),
139 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -0800140 .proc_handler = proc_dointvec_jiffies,
Trond Myklebust25fe6142009-03-11 14:38:03 -0400141 },
Eric W. Biedermanf8572d82009-11-05 13:32:03 -0800142 { },
Chuck Leverfbf76682006-12-05 16:35:54 -0500143};
144
145static ctl_table sunrpc_table[] = {
146 {
Chuck Leverfbf76682006-12-05 16:35:54 -0500147 .procname = "sunrpc",
148 .mode = 0555,
149 .child = xs_tunables_table
150 },
Eric W. Biedermanf8572d82009-11-05 13:32:03 -0800151 { },
Chuck Leverfbf76682006-12-05 16:35:54 -0500152};
153
154#endif
155
156/*
Chuck Lever03bf4b72005-08-25 16:25:55 -0700157 * Wait duration for a reply from the RPC portmapper.
158 */
159#define XS_BIND_TO (60U * HZ)
160
161/*
162 * Delay if a UDP socket connect error occurs. This is most likely some
163 * kind of resource problem on the local host.
164 */
165#define XS_UDP_REEST_TO (2U * HZ)
166
167/*
168 * The reestablish timeout allows clients to delay for a bit before attempting
169 * to reconnect to a server that just dropped our connection.
170 *
171 * We implement an exponential backoff when trying to reestablish a TCP
172 * transport connection with the server. Some servers like to drop a TCP
173 * connection when they are overworked, so we start with a short timeout and
174 * increase over time if the server is down or not responding.
175 */
176#define XS_TCP_INIT_REEST_TO (3U * HZ)
177#define XS_TCP_MAX_REEST_TO (5U * 60 * HZ)
178
179/*
180 * TCP idle timeout; client drops the transport socket if it is idle
181 * for this long. Note that we also timeout UDP sockets to prevent
182 * holding port numbers when there is no RPC traffic.
183 */
184#define XS_IDLE_DISC_TO (5U * 60 * HZ)
185
Chuck Levera246b012005-08-11 16:25:23 -0400186#ifdef RPC_DEBUG
187# undef RPC_DEBUG_DATA
Chuck Lever9903cd12005-08-11 16:25:26 -0400188# define RPCDBG_FACILITY RPCDBG_TRANS
Chuck Levera246b012005-08-11 16:25:23 -0400189#endif
190
Chuck Levera246b012005-08-11 16:25:23 -0400191#ifdef RPC_DEBUG_DATA
Chuck Lever9903cd12005-08-11 16:25:26 -0400192static void xs_pktdump(char *msg, u32 *packet, unsigned int count)
Chuck Levera246b012005-08-11 16:25:23 -0400193{
Chuck Lever9903cd12005-08-11 16:25:26 -0400194 u8 *buf = (u8 *) packet;
195 int j;
Chuck Levera246b012005-08-11 16:25:23 -0400196
Chuck Lever46121cf2007-01-31 12:14:08 -0500197 dprintk("RPC: %s\n", msg);
Chuck Levera246b012005-08-11 16:25:23 -0400198 for (j = 0; j < count && j < 128; j += 4) {
199 if (!(j & 31)) {
200 if (j)
201 dprintk("\n");
202 dprintk("0x%04x ", j);
203 }
204 dprintk("%02x%02x%02x%02x ",
205 buf[j], buf[j+1], buf[j+2], buf[j+3]);
206 }
207 dprintk("\n");
208}
209#else
Chuck Lever9903cd12005-08-11 16:25:26 -0400210static inline void xs_pktdump(char *msg, u32 *packet, unsigned int count)
Chuck Levera246b012005-08-11 16:25:23 -0400211{
212 /* NOP */
213}
214#endif
215
Chuck Leverffc2e512006-12-05 16:35:11 -0500216struct sock_xprt {
217 struct rpc_xprt xprt;
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500218
219 /*
220 * Network layer
221 */
222 struct socket * sock;
223 struct sock * inet;
Chuck Lever51971132006-12-05 16:35:19 -0500224
225 /*
226 * State of TCP reply receive
227 */
228 __be32 tcp_fraghdr,
Trond Myklebustb76ce562010-06-16 13:57:32 -0400229 tcp_xid,
230 tcp_calldir;
Chuck Lever51971132006-12-05 16:35:19 -0500231
232 u32 tcp_offset,
233 tcp_reclen;
234
235 unsigned long tcp_copied,
236 tcp_flags;
Chuck Leverc8475462006-12-05 16:35:26 -0500237
238 /*
239 * Connection of transports
240 */
Trond Myklebust34161db2006-12-07 15:48:15 -0500241 struct delayed_work connect_worker;
Chuck Leverfbfffbd2009-08-09 15:09:46 -0400242 struct sockaddr_storage srcaddr;
243 unsigned short srcport;
Chuck Lever7c6e0662006-12-05 16:35:30 -0500244
245 /*
246 * UDP socket buffer size parameters
247 */
248 size_t rcvsize,
249 sndsize;
Chuck Lever314dfd72006-12-05 16:35:34 -0500250
251 /*
252 * Saved socket callback addresses
253 */
254 void (*old_data_ready)(struct sock *, int);
255 void (*old_state_change)(struct sock *);
256 void (*old_write_space)(struct sock *);
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -0400257 void (*old_error_report)(struct sock *);
Chuck Leverffc2e512006-12-05 16:35:11 -0500258};
259
Chuck Levere136d092006-12-05 16:35:23 -0500260/*
261 * TCP receive state flags
262 */
263#define TCP_RCV_LAST_FRAG (1UL << 0)
264#define TCP_RCV_COPY_FRAGHDR (1UL << 1)
265#define TCP_RCV_COPY_XID (1UL << 2)
266#define TCP_RCV_COPY_DATA (1UL << 3)
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -0400267#define TCP_RCV_READ_CALLDIR (1UL << 4)
268#define TCP_RCV_COPY_CALLDIR (1UL << 5)
Ricardo Labiaga18dca022009-04-01 09:22:53 -0400269
270/*
271 * TCP RPC flags
272 */
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -0400273#define TCP_RPC_REPLY (1UL << 6)
Chuck Levere136d092006-12-05 16:35:23 -0500274
Chuck Lever95392c52007-08-06 11:57:58 -0400275static inline struct sockaddr *xs_addr(struct rpc_xprt *xprt)
Chuck Leveredb267a2006-08-22 20:06:18 -0400276{
Chuck Lever95392c52007-08-06 11:57:58 -0400277 return (struct sockaddr *) &xprt->addr;
278}
279
Chuck Lever176e21e2011-05-09 15:22:44 -0400280static inline struct sockaddr_un *xs_addr_un(struct rpc_xprt *xprt)
281{
282 return (struct sockaddr_un *) &xprt->addr;
283}
284
Chuck Lever95392c52007-08-06 11:57:58 -0400285static inline struct sockaddr_in *xs_addr_in(struct rpc_xprt *xprt)
286{
287 return (struct sockaddr_in *) &xprt->addr;
288}
289
290static inline struct sockaddr_in6 *xs_addr_in6(struct rpc_xprt *xprt)
291{
292 return (struct sockaddr_in6 *) &xprt->addr;
293}
294
Chuck Leverc877b842009-08-09 15:09:36 -0400295static void xs_format_common_peer_addresses(struct rpc_xprt *xprt)
Chuck Leveredb267a2006-08-22 20:06:18 -0400296{
Chuck Leverc877b842009-08-09 15:09:36 -0400297 struct sockaddr *sap = xs_addr(xprt);
Chuck Lever9dc3b092009-08-09 15:09:46 -0400298 struct sockaddr_in6 *sin6;
299 struct sockaddr_in *sin;
Chuck Lever176e21e2011-05-09 15:22:44 -0400300 struct sockaddr_un *sun;
Chuck Leverc877b842009-08-09 15:09:36 -0400301 char buf[128];
Chuck Leveredb267a2006-08-22 20:06:18 -0400302
Chuck Lever9dc3b092009-08-09 15:09:46 -0400303 switch (sap->sa_family) {
Chuck Lever176e21e2011-05-09 15:22:44 -0400304 case AF_LOCAL:
305 sun = xs_addr_un(xprt);
306 strlcpy(buf, sun->sun_path, sizeof(buf));
307 xprt->address_strings[RPC_DISPLAY_ADDR] =
308 kstrdup(buf, GFP_KERNEL);
309 break;
Chuck Lever9dc3b092009-08-09 15:09:46 -0400310 case AF_INET:
Chuck Lever176e21e2011-05-09 15:22:44 -0400311 (void)rpc_ntop(sap, buf, sizeof(buf));
312 xprt->address_strings[RPC_DISPLAY_ADDR] =
313 kstrdup(buf, GFP_KERNEL);
Chuck Lever9dc3b092009-08-09 15:09:46 -0400314 sin = xs_addr_in(xprt);
Joe Perchesfc0b5792010-03-08 12:15:28 -0800315 snprintf(buf, sizeof(buf), "%08x", ntohl(sin->sin_addr.s_addr));
Chuck Lever9dc3b092009-08-09 15:09:46 -0400316 break;
317 case AF_INET6:
Chuck Lever176e21e2011-05-09 15:22:44 -0400318 (void)rpc_ntop(sap, buf, sizeof(buf));
319 xprt->address_strings[RPC_DISPLAY_ADDR] =
320 kstrdup(buf, GFP_KERNEL);
Chuck Lever9dc3b092009-08-09 15:09:46 -0400321 sin6 = xs_addr_in6(xprt);
Joe Perchesfc0b5792010-03-08 12:15:28 -0800322 snprintf(buf, sizeof(buf), "%pi6", &sin6->sin6_addr);
Chuck Lever9dc3b092009-08-09 15:09:46 -0400323 break;
324 default:
325 BUG();
Chuck Leveredb267a2006-08-22 20:06:18 -0400326 }
Chuck Lever176e21e2011-05-09 15:22:44 -0400327
Chuck Lever9dc3b092009-08-09 15:09:46 -0400328 xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = kstrdup(buf, GFP_KERNEL);
Chuck Leveredb267a2006-08-22 20:06:18 -0400329}
330
Chuck Lever9dc3b092009-08-09 15:09:46 -0400331static void xs_format_common_peer_ports(struct rpc_xprt *xprt)
Chuck Lever4b6473f2007-08-06 11:57:12 -0400332{
Chuck Lever9dc3b092009-08-09 15:09:46 -0400333 struct sockaddr *sap = xs_addr(xprt);
334 char buf[128];
Chuck Lever4b6473f2007-08-06 11:57:12 -0400335
Joe Perches81160e662010-03-08 12:15:59 -0800336 snprintf(buf, sizeof(buf), "%u", rpc_get_port(sap));
Chuck Leverc877b842009-08-09 15:09:36 -0400337 xprt->address_strings[RPC_DISPLAY_PORT] = kstrdup(buf, GFP_KERNEL);
Chuck Lever4b6473f2007-08-06 11:57:12 -0400338
Joe Perches81160e662010-03-08 12:15:59 -0800339 snprintf(buf, sizeof(buf), "%4hx", rpc_get_port(sap));
Chuck Leverc877b842009-08-09 15:09:36 -0400340 xprt->address_strings[RPC_DISPLAY_HEX_PORT] = kstrdup(buf, GFP_KERNEL);
341}
Chuck Lever4b6473f2007-08-06 11:57:12 -0400342
Chuck Lever9dc3b092009-08-09 15:09:46 -0400343static void xs_format_peer_addresses(struct rpc_xprt *xprt,
344 const char *protocol,
345 const char *netid)
Chuck Leveredb267a2006-08-22 20:06:18 -0400346{
Chuck Leverb454ae92008-01-07 18:34:48 -0500347 xprt->address_strings[RPC_DISPLAY_PROTO] = protocol;
Chuck Leverb454ae92008-01-07 18:34:48 -0500348 xprt->address_strings[RPC_DISPLAY_NETID] = netid;
Chuck Leverc877b842009-08-09 15:09:36 -0400349 xs_format_common_peer_addresses(xprt);
Chuck Lever9dc3b092009-08-09 15:09:46 -0400350 xs_format_common_peer_ports(xprt);
Chuck Lever4b6473f2007-08-06 11:57:12 -0400351}
352
Chuck Lever9dc3b092009-08-09 15:09:46 -0400353static void xs_update_peer_port(struct rpc_xprt *xprt)
Chuck Lever4b6473f2007-08-06 11:57:12 -0400354{
Chuck Lever9dc3b092009-08-09 15:09:46 -0400355 kfree(xprt->address_strings[RPC_DISPLAY_HEX_PORT]);
356 kfree(xprt->address_strings[RPC_DISPLAY_PORT]);
Chuck Lever4b6473f2007-08-06 11:57:12 -0400357
Chuck Lever9dc3b092009-08-09 15:09:46 -0400358 xs_format_common_peer_ports(xprt);
Chuck Leveredb267a2006-08-22 20:06:18 -0400359}
360
361static void xs_free_peer_addresses(struct rpc_xprt *xprt)
362{
Chuck Lever33e01dc2008-01-14 12:32:20 -0500363 unsigned int i;
364
365 for (i = 0; i < RPC_DISPLAY_MAX; i++)
366 switch (i) {
367 case RPC_DISPLAY_PROTO:
368 case RPC_DISPLAY_NETID:
369 continue;
370 default:
371 kfree(xprt->address_strings[i]);
372 }
Chuck Leveredb267a2006-08-22 20:06:18 -0400373}
374
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400375#define XS_SENDMSG_FLAGS (MSG_DONTWAIT | MSG_NOSIGNAL)
376
Trond Myklebust24c56842006-10-17 15:06:22 -0400377static 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 -0400378{
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400379 struct msghdr msg = {
380 .msg_name = addr,
381 .msg_namelen = addrlen,
Trond Myklebust24c56842006-10-17 15:06:22 -0400382 .msg_flags = XS_SENDMSG_FLAGS | (more ? MSG_MORE : 0),
383 };
384 struct kvec iov = {
385 .iov_base = vec->iov_base + base,
386 .iov_len = vec->iov_len - base,
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400387 };
388
Trond Myklebust24c56842006-10-17 15:06:22 -0400389 if (iov.iov_len != 0)
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400390 return kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len);
391 return kernel_sendmsg(sock, &msg, NULL, 0, 0);
392}
393
Trond Myklebust24c56842006-10-17 15:06:22 -0400394static int xs_send_pagedata(struct socket *sock, struct xdr_buf *xdr, unsigned int base, int more)
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400395{
Trond Myklebust24c56842006-10-17 15:06:22 -0400396 struct page **ppage;
397 unsigned int remainder;
398 int err, sent = 0;
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400399
Trond Myklebust24c56842006-10-17 15:06:22 -0400400 remainder = xdr->page_len - base;
401 base += xdr->page_base;
402 ppage = xdr->pages + (base >> PAGE_SHIFT);
403 base &= ~PAGE_MASK;
404 for(;;) {
405 unsigned int len = min_t(unsigned int, PAGE_SIZE - base, remainder);
406 int flags = XS_SENDMSG_FLAGS;
407
408 remainder -= len;
409 if (remainder != 0 || more)
410 flags |= MSG_MORE;
411 err = sock->ops->sendpage(sock, *ppage, base, len, flags);
412 if (remainder == 0 || err != len)
413 break;
414 sent += err;
415 ppage++;
416 base = 0;
417 }
418 if (sent == 0)
419 return err;
420 if (err > 0)
421 sent += err;
422 return sent;
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400423}
424
Chuck Lever9903cd12005-08-11 16:25:26 -0400425/**
426 * xs_sendpages - write pages directly to a socket
427 * @sock: socket to send on
428 * @addr: UDP only -- address of destination
429 * @addrlen: UDP only -- length of destination address
430 * @xdr: buffer containing this request
431 * @base: starting position in the buffer
432 *
Chuck Levera246b012005-08-11 16:25:23 -0400433 */
Trond Myklebust24c56842006-10-17 15:06:22 -0400434static 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 -0400435{
Trond Myklebust24c56842006-10-17 15:06:22 -0400436 unsigned int remainder = xdr->len - base;
437 int err, sent = 0;
Chuck Levera246b012005-08-11 16:25:23 -0400438
Chuck Lever262965f2005-08-11 16:25:56 -0400439 if (unlikely(!sock))
Trond Myklebustfba91af2009-03-11 14:06:41 -0400440 return -ENOTSOCK;
Chuck Lever262965f2005-08-11 16:25:56 -0400441
442 clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags);
Trond Myklebust24c56842006-10-17 15:06:22 -0400443 if (base != 0) {
444 addr = NULL;
445 addrlen = 0;
446 }
Chuck Lever262965f2005-08-11 16:25:56 -0400447
Trond Myklebust24c56842006-10-17 15:06:22 -0400448 if (base < xdr->head[0].iov_len || addr != NULL) {
449 unsigned int len = xdr->head[0].iov_len - base;
450 remainder -= len;
451 err = xs_send_kvec(sock, addr, addrlen, &xdr->head[0], base, remainder != 0);
452 if (remainder == 0 || err != len)
Chuck Levera246b012005-08-11 16:25:23 -0400453 goto out;
Trond Myklebust24c56842006-10-17 15:06:22 -0400454 sent += err;
Chuck Levera246b012005-08-11 16:25:23 -0400455 base = 0;
456 } else
Trond Myklebust24c56842006-10-17 15:06:22 -0400457 base -= xdr->head[0].iov_len;
Chuck Levera246b012005-08-11 16:25:23 -0400458
Trond Myklebust24c56842006-10-17 15:06:22 -0400459 if (base < xdr->page_len) {
460 unsigned int len = xdr->page_len - base;
461 remainder -= len;
462 err = xs_send_pagedata(sock, xdr, base, remainder != 0);
463 if (remainder == 0 || err != len)
Chuck Levera246b012005-08-11 16:25:23 -0400464 goto out;
Trond Myklebust24c56842006-10-17 15:06:22 -0400465 sent += err;
Chuck Levera246b012005-08-11 16:25:23 -0400466 base = 0;
Trond Myklebust24c56842006-10-17 15:06:22 -0400467 } else
468 base -= xdr->page_len;
469
470 if (base >= xdr->tail[0].iov_len)
471 return sent;
472 err = xs_send_kvec(sock, NULL, 0, &xdr->tail[0], base, 0);
Chuck Levera246b012005-08-11 16:25:23 -0400473out:
Trond Myklebust24c56842006-10-17 15:06:22 -0400474 if (sent == 0)
475 return err;
476 if (err > 0)
477 sent += err;
478 return sent;
Chuck Levera246b012005-08-11 16:25:23 -0400479}
480
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400481static void xs_nospace_callback(struct rpc_task *task)
482{
483 struct sock_xprt *transport = container_of(task->tk_rqstp->rq_xprt, struct sock_xprt, xprt);
484
485 transport->inet->sk_write_pending--;
486 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
487}
488
Chuck Lever9903cd12005-08-11 16:25:26 -0400489/**
Chuck Lever262965f2005-08-11 16:25:56 -0400490 * xs_nospace - place task on wait queue if transmit was incomplete
491 * @task: task to put to sleep
Chuck Lever9903cd12005-08-11 16:25:26 -0400492 *
Chuck Levera246b012005-08-11 16:25:23 -0400493 */
Trond Myklebust5e3771c2009-03-11 14:38:01 -0400494static int xs_nospace(struct rpc_task *task)
Chuck Levera246b012005-08-11 16:25:23 -0400495{
Chuck Lever262965f2005-08-11 16:25:56 -0400496 struct rpc_rqst *req = task->tk_rqstp;
497 struct rpc_xprt *xprt = req->rq_xprt;
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500498 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Trond Myklebust24ca9a82011-11-22 14:44:28 +0200499 int ret = -EAGAIN;
Chuck Levera246b012005-08-11 16:25:23 -0400500
Chuck Lever46121cf2007-01-31 12:14:08 -0500501 dprintk("RPC: %5u xmit incomplete (%u left of %u)\n",
Chuck Lever262965f2005-08-11 16:25:56 -0400502 task->tk_pid, req->rq_slen - req->rq_bytes_sent,
503 req->rq_slen);
504
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400505 /* Protect against races with write_space */
506 spin_lock_bh(&xprt->transport_lock);
Chuck Lever262965f2005-08-11 16:25:56 -0400507
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400508 /* Don't race with disconnect */
509 if (xprt_connected(xprt)) {
510 if (test_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags)) {
511 /*
512 * Notify TCP that we're limited by the application
513 * window size
514 */
515 set_bit(SOCK_NOSPACE, &transport->sock->flags);
516 transport->inet->sk_write_pending++;
517 /* ...and wait for more buffer space */
518 xprt_wait_for_buffer_space(task, xs_nospace_callback);
519 }
520 } else {
521 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
Trond Myklebust5e3771c2009-03-11 14:38:01 -0400522 ret = -ENOTCONN;
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400523 }
Chuck Lever262965f2005-08-11 16:25:56 -0400524
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400525 spin_unlock_bh(&xprt->transport_lock);
Trond Myklebust5e3771c2009-03-11 14:38:01 -0400526 return ret;
Chuck Lever262965f2005-08-11 16:25:56 -0400527}
528
Chuck Lever61677ee2011-05-09 15:22:34 -0400529/*
530 * Construct a stream transport record marker in @buf.
531 */
532static inline void xs_encode_stream_record_marker(struct xdr_buf *buf)
533{
534 u32 reclen = buf->len - sizeof(rpc_fraghdr);
535 rpc_fraghdr *base = buf->head[0].iov_base;
536 *base = cpu_to_be32(RPC_LAST_STREAM_FRAGMENT | reclen);
537}
538
Chuck Lever262965f2005-08-11 16:25:56 -0400539/**
Chuck Lever176e21e2011-05-09 15:22:44 -0400540 * xs_local_send_request - write an RPC request to an AF_LOCAL socket
541 * @task: RPC task that manages the state of an RPC request
542 *
543 * Return values:
544 * 0: The request has been sent
545 * EAGAIN: The socket was blocked, please call again later to
546 * complete the request
547 * ENOTCONN: Caller needs to invoke connect logic then call again
548 * other: Some other error occured, the request was not sent
549 */
550static int xs_local_send_request(struct rpc_task *task)
551{
552 struct rpc_rqst *req = task->tk_rqstp;
553 struct rpc_xprt *xprt = req->rq_xprt;
554 struct sock_xprt *transport =
555 container_of(xprt, struct sock_xprt, xprt);
556 struct xdr_buf *xdr = &req->rq_snd_buf;
557 int status;
558
559 xs_encode_stream_record_marker(&req->rq_snd_buf);
560
561 xs_pktdump("packet data:",
562 req->rq_svec->iov_base, req->rq_svec->iov_len);
563
564 status = xs_sendpages(transport->sock, NULL, 0,
565 xdr, req->rq_bytes_sent);
566 dprintk("RPC: %s(%u) = %d\n",
567 __func__, xdr->len - req->rq_bytes_sent, status);
568 if (likely(status >= 0)) {
569 req->rq_bytes_sent += status;
570 req->rq_xmit_bytes_sent += status;
571 if (likely(req->rq_bytes_sent >= req->rq_slen)) {
572 req->rq_bytes_sent = 0;
573 return 0;
574 }
575 status = -EAGAIN;
576 }
577
578 switch (status) {
579 case -EAGAIN:
580 status = xs_nospace(task);
581 break;
582 default:
583 dprintk("RPC: sendmsg returned unrecognized error %d\n",
584 -status);
585 case -EPIPE:
586 xs_close(xprt);
587 status = -ENOTCONN;
588 }
589
590 return status;
591}
592
593/**
Chuck Lever262965f2005-08-11 16:25:56 -0400594 * xs_udp_send_request - write an RPC request to a UDP socket
595 * @task: address of RPC task that manages the state of an RPC request
596 *
597 * Return values:
598 * 0: The request has been sent
599 * EAGAIN: The socket was blocked, please call again later to
600 * complete the request
601 * ENOTCONN: Caller needs to invoke connect logic then call again
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300602 * other: Some other error occurred, the request was not sent
Chuck Lever262965f2005-08-11 16:25:56 -0400603 */
604static int xs_udp_send_request(struct rpc_task *task)
605{
606 struct rpc_rqst *req = task->tk_rqstp;
607 struct rpc_xprt *xprt = req->rq_xprt;
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500608 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Lever262965f2005-08-11 16:25:56 -0400609 struct xdr_buf *xdr = &req->rq_snd_buf;
610 int status;
Chuck Levera246b012005-08-11 16:25:23 -0400611
Chuck Lever9903cd12005-08-11 16:25:26 -0400612 xs_pktdump("packet data:",
Chuck Levera246b012005-08-11 16:25:23 -0400613 req->rq_svec->iov_base,
614 req->rq_svec->iov_len);
615
Trond Myklebust01d37c42009-03-11 14:09:39 -0400616 if (!xprt_bound(xprt))
617 return -ENOTCONN;
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500618 status = xs_sendpages(transport->sock,
Chuck Lever95392c52007-08-06 11:57:58 -0400619 xs_addr(xprt),
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500620 xprt->addrlen, xdr,
621 req->rq_bytes_sent);
Chuck Levera246b012005-08-11 16:25:23 -0400622
Chuck Lever46121cf2007-01-31 12:14:08 -0500623 dprintk("RPC: xs_udp_send_request(%u) = %d\n",
Chuck Lever262965f2005-08-11 16:25:56 -0400624 xdr->len - req->rq_bytes_sent, status);
Chuck Levera246b012005-08-11 16:25:23 -0400625
Trond Myklebust21997002007-10-01 11:43:37 -0400626 if (status >= 0) {
Trond Myklebustd60dbb22010-05-13 12:51:49 -0400627 req->rq_xmit_bytes_sent += status;
Trond Myklebust21997002007-10-01 11:43:37 -0400628 if (status >= req->rq_slen)
629 return 0;
630 /* Still some bytes left; set up for a retry later. */
Chuck Lever262965f2005-08-11 16:25:56 -0400631 status = -EAGAIN;
Trond Myklebust21997002007-10-01 11:43:37 -0400632 }
Chuck Levera246b012005-08-11 16:25:23 -0400633
Chuck Lever262965f2005-08-11 16:25:56 -0400634 switch (status) {
Trond Myklebustfba91af2009-03-11 14:06:41 -0400635 case -ENOTSOCK:
636 status = -ENOTCONN;
637 /* Should we call xs_close() here? */
638 break;
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400639 case -EAGAIN:
Trond Myklebust5e3771c2009-03-11 14:38:01 -0400640 status = xs_nospace(task);
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400641 break;
Trond Myklebustc8485e42009-03-11 14:37:59 -0400642 default:
643 dprintk("RPC: sendmsg returned unrecognized error %d\n",
644 -status);
Chuck Lever262965f2005-08-11 16:25:56 -0400645 case -ENETUNREACH:
646 case -EPIPE:
Chuck Levera246b012005-08-11 16:25:23 -0400647 case -ECONNREFUSED:
648 /* When the server has died, an ICMP port unreachable message
Chuck Lever9903cd12005-08-11 16:25:26 -0400649 * prompts ECONNREFUSED. */
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400650 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
Chuck Levera246b012005-08-11 16:25:23 -0400651 }
Bian Naimeng5fe46e92010-03-08 14:49:01 +0800652
Chuck Lever262965f2005-08-11 16:25:56 -0400653 return status;
Chuck Levera246b012005-08-11 16:25:23 -0400654}
655
Trond Myklebuste06799f2007-11-05 15:44:12 -0500656/**
657 * xs_tcp_shutdown - gracefully shut down a TCP socket
658 * @xprt: transport
659 *
660 * Initiates a graceful shutdown of the TCP socket by calling the
661 * equivalent of shutdown(SHUT_WR);
662 */
663static void xs_tcp_shutdown(struct rpc_xprt *xprt)
664{
665 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
666 struct socket *sock = transport->sock;
667
668 if (sock != NULL)
669 kernel_sock_shutdown(sock, SHUT_WR);
670}
671
Chuck Lever9903cd12005-08-11 16:25:26 -0400672/**
Chuck Lever262965f2005-08-11 16:25:56 -0400673 * xs_tcp_send_request - write an RPC request to a TCP socket
Chuck Lever9903cd12005-08-11 16:25:26 -0400674 * @task: address of RPC task that manages the state of an RPC request
675 *
676 * Return values:
Chuck Lever262965f2005-08-11 16:25:56 -0400677 * 0: The request has been sent
678 * EAGAIN: The socket was blocked, please call again later to
679 * complete the request
680 * ENOTCONN: Caller needs to invoke connect logic then call again
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300681 * other: Some other error occurred, the request was not sent
Chuck Lever9903cd12005-08-11 16:25:26 -0400682 *
683 * XXX: In the case of soft timeouts, should we eventually give up
Chuck Lever262965f2005-08-11 16:25:56 -0400684 * if sendmsg is not able to make progress?
Chuck Lever9903cd12005-08-11 16:25:26 -0400685 */
Chuck Lever262965f2005-08-11 16:25:56 -0400686static int xs_tcp_send_request(struct rpc_task *task)
Chuck Levera246b012005-08-11 16:25:23 -0400687{
688 struct rpc_rqst *req = task->tk_rqstp;
689 struct rpc_xprt *xprt = req->rq_xprt;
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500690 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Lever262965f2005-08-11 16:25:56 -0400691 struct xdr_buf *xdr = &req->rq_snd_buf;
Chuck Leverb595bb12007-08-06 11:56:42 -0400692 int status;
Chuck Levera246b012005-08-11 16:25:23 -0400693
Chuck Lever61677ee2011-05-09 15:22:34 -0400694 xs_encode_stream_record_marker(&req->rq_snd_buf);
Chuck Levera246b012005-08-11 16:25:23 -0400695
Chuck Lever262965f2005-08-11 16:25:56 -0400696 xs_pktdump("packet data:",
697 req->rq_svec->iov_base,
698 req->rq_svec->iov_len);
Chuck Levera246b012005-08-11 16:25:23 -0400699
700 /* Continue transmitting the packet/record. We must be careful
701 * to cope with writespace callbacks arriving _after_ we have
Chuck Lever262965f2005-08-11 16:25:56 -0400702 * called sendmsg(). */
Chuck Levera246b012005-08-11 16:25:23 -0400703 while (1) {
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500704 status = xs_sendpages(transport->sock,
705 NULL, 0, xdr, req->rq_bytes_sent);
Chuck Levera246b012005-08-11 16:25:23 -0400706
Chuck Lever46121cf2007-01-31 12:14:08 -0500707 dprintk("RPC: xs_tcp_send_request(%u) = %d\n",
Chuck Lever262965f2005-08-11 16:25:56 -0400708 xdr->len - req->rq_bytes_sent, status);
709
710 if (unlikely(status < 0))
Chuck Levera246b012005-08-11 16:25:23 -0400711 break;
712
Chuck Lever262965f2005-08-11 16:25:56 -0400713 /* If we've sent the entire packet, immediately
714 * reset the count of bytes sent. */
715 req->rq_bytes_sent += status;
Trond Myklebustd60dbb22010-05-13 12:51:49 -0400716 req->rq_xmit_bytes_sent += status;
Chuck Lever262965f2005-08-11 16:25:56 -0400717 if (likely(req->rq_bytes_sent >= req->rq_slen)) {
718 req->rq_bytes_sent = 0;
719 return 0;
Chuck Levera246b012005-08-11 16:25:23 -0400720 }
721
Trond Myklebust06b4b682008-04-16 16:51:38 -0400722 if (status != 0)
723 continue;
Chuck Levera246b012005-08-11 16:25:23 -0400724 status = -EAGAIN;
Trond Myklebust06b4b682008-04-16 16:51:38 -0400725 break;
Chuck Levera246b012005-08-11 16:25:23 -0400726 }
727
Chuck Lever262965f2005-08-11 16:25:56 -0400728 switch (status) {
Trond Myklebustfba91af2009-03-11 14:06:41 -0400729 case -ENOTSOCK:
730 status = -ENOTCONN;
731 /* Should we call xs_close() here? */
732 break;
Chuck Lever262965f2005-08-11 16:25:56 -0400733 case -EAGAIN:
Trond Myklebust5e3771c2009-03-11 14:38:01 -0400734 status = xs_nospace(task);
Chuck Lever262965f2005-08-11 16:25:56 -0400735 break;
736 default:
Chuck Lever46121cf2007-01-31 12:14:08 -0500737 dprintk("RPC: sendmsg returned unrecognized error %d\n",
Chuck Lever262965f2005-08-11 16:25:56 -0400738 -status);
Chuck Lever262965f2005-08-11 16:25:56 -0400739 case -ECONNRESET:
Trond Myklebust55420c22009-03-11 15:29:24 -0400740 case -EPIPE:
Trond Myklebuste06799f2007-11-05 15:44:12 -0500741 xs_tcp_shutdown(xprt);
Chuck Lever262965f2005-08-11 16:25:56 -0400742 case -ECONNREFUSED:
743 case -ENOTCONN:
Chuck Lever262965f2005-08-11 16:25:56 -0400744 clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags);
Chuck Levera246b012005-08-11 16:25:23 -0400745 }
Bian Naimeng5fe46e92010-03-08 14:49:01 +0800746
Chuck Levera246b012005-08-11 16:25:23 -0400747 return status;
748}
749
Chuck Lever9903cd12005-08-11 16:25:26 -0400750/**
Trond Myklebuste0ab53d2006-07-27 17:22:50 -0400751 * xs_tcp_release_xprt - clean up after a tcp transmission
752 * @xprt: transport
753 * @task: rpc task
754 *
755 * This cleans up if an error causes us to abort the transmission of a request.
756 * In this case, the socket may need to be reset in order to avoid confusing
757 * the server.
758 */
759static void xs_tcp_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
760{
761 struct rpc_rqst *req;
762
763 if (task != xprt->snd_task)
764 return;
765 if (task == NULL)
766 goto out_release;
767 req = task->tk_rqstp;
Trond Myklebust43cedbf2011-07-17 16:01:03 -0400768 if (req == NULL)
769 goto out_release;
Trond Myklebuste0ab53d2006-07-27 17:22:50 -0400770 if (req->rq_bytes_sent == 0)
771 goto out_release;
772 if (req->rq_bytes_sent == req->rq_snd_buf.len)
773 goto out_release;
774 set_bit(XPRT_CLOSE_WAIT, &task->tk_xprt->state);
775out_release:
776 xprt_release_xprt(xprt, task);
777}
778
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -0400779static void xs_save_old_callbacks(struct sock_xprt *transport, struct sock *sk)
780{
781 transport->old_data_ready = sk->sk_data_ready;
782 transport->old_state_change = sk->sk_state_change;
783 transport->old_write_space = sk->sk_write_space;
784 transport->old_error_report = sk->sk_error_report;
785}
786
787static void xs_restore_old_callbacks(struct sock_xprt *transport, struct sock *sk)
788{
789 sk->sk_data_ready = transport->old_data_ready;
790 sk->sk_state_change = transport->old_state_change;
791 sk->sk_write_space = transport->old_write_space;
792 sk->sk_error_report = transport->old_error_report;
793}
794
Chuck Leverfe315e72009-03-11 14:10:21 -0400795static void xs_reset_transport(struct sock_xprt *transport)
Chuck Levera246b012005-08-11 16:25:23 -0400796{
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500797 struct socket *sock = transport->sock;
798 struct sock *sk = transport->inet;
Chuck Levera246b012005-08-11 16:25:23 -0400799
Chuck Leverfe315e72009-03-11 14:10:21 -0400800 if (sk == NULL)
801 return;
Chuck Lever9903cd12005-08-11 16:25:26 -0400802
Trond Myklebust246408d2011-03-22 18:40:10 -0400803 transport->srcport = 0;
804
Chuck Levera246b012005-08-11 16:25:23 -0400805 write_lock_bh(&sk->sk_callback_lock);
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500806 transport->inet = NULL;
807 transport->sock = NULL;
Chuck Levera246b012005-08-11 16:25:23 -0400808
Chuck Lever9903cd12005-08-11 16:25:26 -0400809 sk->sk_user_data = NULL;
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -0400810
811 xs_restore_old_callbacks(transport, sk);
Chuck Levera246b012005-08-11 16:25:23 -0400812 write_unlock_bh(&sk->sk_callback_lock);
813
Chuck Lever9903cd12005-08-11 16:25:26 -0400814 sk->sk_no_check = 0;
Chuck Levera246b012005-08-11 16:25:23 -0400815
816 sock_release(sock);
Chuck Leverfe315e72009-03-11 14:10:21 -0400817}
818
819/**
820 * xs_close - close a socket
821 * @xprt: transport
822 *
823 * This is used when all requests are complete; ie, no DRC state remains
824 * on the server we want to save.
Trond Myklebustf75e6742009-04-21 17:18:20 -0400825 *
826 * The caller _must_ be holding XPRT_LOCKED in order to avoid issues with
827 * xs_reset_transport() zeroing the socket from underneath a writer.
Chuck Leverfe315e72009-03-11 14:10:21 -0400828 */
829static void xs_close(struct rpc_xprt *xprt)
830{
831 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
832
833 dprintk("RPC: xs_close xprt %p\n", xprt);
834
835 xs_reset_transport(transport);
Neil Brown61d0a8e2009-09-23 14:36:37 -0400836 xprt->reestablish_timeout = 0;
Chuck Leverfe315e72009-03-11 14:10:21 -0400837
Trond Myklebust632e3bd2006-01-03 09:55:55 +0100838 smp_mb__before_clear_bit();
Trond Myklebust7d1e8252009-03-11 14:38:03 -0400839 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
Trond Myklebust632e3bd2006-01-03 09:55:55 +0100840 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
Trond Myklebust3b948ae2007-11-05 17:42:39 -0500841 clear_bit(XPRT_CLOSING, &xprt->state);
Trond Myklebust632e3bd2006-01-03 09:55:55 +0100842 smp_mb__after_clear_bit();
Trond Myklebust62da3b22007-11-06 18:44:20 -0500843 xprt_disconnect_done(xprt);
Chuck Levera246b012005-08-11 16:25:23 -0400844}
845
Trond Myklebustf75e6742009-04-21 17:18:20 -0400846static void xs_tcp_close(struct rpc_xprt *xprt)
847{
848 if (test_and_clear_bit(XPRT_CONNECTION_CLOSE, &xprt->state))
849 xs_close(xprt);
850 else
851 xs_tcp_shutdown(xprt);
852}
853
Chuck Lever9903cd12005-08-11 16:25:26 -0400854/**
855 * xs_destroy - prepare to shutdown a transport
856 * @xprt: doomed transport
857 *
858 */
859static void xs_destroy(struct rpc_xprt *xprt)
Chuck Levera246b012005-08-11 16:25:23 -0400860{
Chuck Leverc8475462006-12-05 16:35:26 -0500861 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
862
Chuck Lever46121cf2007-01-31 12:14:08 -0500863 dprintk("RPC: xs_destroy xprt %p\n", xprt);
Chuck Lever9903cd12005-08-11 16:25:26 -0400864
Tejun Heoafe2c512010-12-14 16:21:17 +0100865 cancel_delayed_work_sync(&transport->connect_worker);
Chuck Levera246b012005-08-11 16:25:23 -0400866
Chuck Lever9903cd12005-08-11 16:25:26 -0400867 xs_close(xprt);
Chuck Leveredb267a2006-08-22 20:06:18 -0400868 xs_free_peer_addresses(xprt);
Pavel Emelyanove204e622010-09-29 16:03:13 +0400869 xprt_free(xprt);
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -0400870 module_put(THIS_MODULE);
Chuck Levera246b012005-08-11 16:25:23 -0400871}
872
Chuck Lever9903cd12005-08-11 16:25:26 -0400873static inline struct rpc_xprt *xprt_from_sock(struct sock *sk)
Chuck Levera246b012005-08-11 16:25:23 -0400874{
Chuck Lever9903cd12005-08-11 16:25:26 -0400875 return (struct rpc_xprt *) sk->sk_user_data;
876}
877
Chuck Lever176e21e2011-05-09 15:22:44 -0400878static int xs_local_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb)
879{
880 struct xdr_skb_reader desc = {
881 .skb = skb,
882 .offset = sizeof(rpc_fraghdr),
883 .count = skb->len - sizeof(rpc_fraghdr),
884 };
885
886 if (xdr_partial_copy_from_skb(xdr, 0, &desc, xdr_skb_read_bits) < 0)
887 return -1;
888 if (desc.count)
889 return -1;
890 return 0;
891}
892
893/**
894 * xs_local_data_ready - "data ready" callback for AF_LOCAL sockets
895 * @sk: socket with data to read
896 * @len: how much data to read
897 *
898 * Currently this assumes we can read the whole reply in a single gulp.
899 */
900static void xs_local_data_ready(struct sock *sk, int len)
901{
902 struct rpc_task *task;
903 struct rpc_xprt *xprt;
904 struct rpc_rqst *rovr;
905 struct sk_buff *skb;
906 int err, repsize, copied;
907 u32 _xid;
908 __be32 *xp;
909
910 read_lock_bh(&sk->sk_callback_lock);
911 dprintk("RPC: %s...\n", __func__);
912 xprt = xprt_from_sock(sk);
913 if (xprt == NULL)
914 goto out;
915
916 skb = skb_recv_datagram(sk, 0, 1, &err);
917 if (skb == NULL)
918 goto out;
919
920 if (xprt->shutdown)
921 goto dropit;
922
923 repsize = skb->len - sizeof(rpc_fraghdr);
924 if (repsize < 4) {
925 dprintk("RPC: impossible RPC reply size %d\n", repsize);
926 goto dropit;
927 }
928
929 /* Copy the XID from the skb... */
930 xp = skb_header_pointer(skb, sizeof(rpc_fraghdr), sizeof(_xid), &_xid);
931 if (xp == NULL)
932 goto dropit;
933
934 /* Look up and lock the request corresponding to the given XID */
935 spin_lock(&xprt->transport_lock);
936 rovr = xprt_lookup_rqst(xprt, *xp);
937 if (!rovr)
938 goto out_unlock;
939 task = rovr->rq_task;
940
941 copied = rovr->rq_private_buf.buflen;
942 if (copied > repsize)
943 copied = repsize;
944
945 if (xs_local_copy_to_xdr(&rovr->rq_private_buf, skb)) {
946 dprintk("RPC: sk_buff copy failed\n");
947 goto out_unlock;
948 }
949
950 xprt_complete_rqst(task, copied);
951
952 out_unlock:
953 spin_unlock(&xprt->transport_lock);
954 dropit:
955 skb_free_datagram(sk, skb);
956 out:
957 read_unlock_bh(&sk->sk_callback_lock);
958}
959
Chuck Lever9903cd12005-08-11 16:25:26 -0400960/**
961 * xs_udp_data_ready - "data ready" callback for UDP sockets
962 * @sk: socket with data to read
963 * @len: how much data to read
964 *
965 */
966static void xs_udp_data_ready(struct sock *sk, int len)
967{
968 struct rpc_task *task;
969 struct rpc_xprt *xprt;
Chuck Levera246b012005-08-11 16:25:23 -0400970 struct rpc_rqst *rovr;
Chuck Lever9903cd12005-08-11 16:25:26 -0400971 struct sk_buff *skb;
Chuck Levera246b012005-08-11 16:25:23 -0400972 int err, repsize, copied;
Alexey Dobriyand8ed0292006-09-26 22:29:38 -0700973 u32 _xid;
974 __be32 *xp;
Chuck Levera246b012005-08-11 16:25:23 -0400975
Eric Dumazetf064af12010-09-22 12:43:39 +0000976 read_lock_bh(&sk->sk_callback_lock);
Chuck Lever46121cf2007-01-31 12:14:08 -0500977 dprintk("RPC: xs_udp_data_ready...\n");
Chuck Lever9903cd12005-08-11 16:25:26 -0400978 if (!(xprt = xprt_from_sock(sk)))
Chuck Levera246b012005-08-11 16:25:23 -0400979 goto out;
Chuck Levera246b012005-08-11 16:25:23 -0400980
981 if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL)
982 goto out;
983
984 if (xprt->shutdown)
985 goto dropit;
986
987 repsize = skb->len - sizeof(struct udphdr);
988 if (repsize < 4) {
Chuck Lever46121cf2007-01-31 12:14:08 -0500989 dprintk("RPC: impossible RPC reply size %d!\n", repsize);
Chuck Levera246b012005-08-11 16:25:23 -0400990 goto dropit;
991 }
992
993 /* Copy the XID from the skb... */
994 xp = skb_header_pointer(skb, sizeof(struct udphdr),
995 sizeof(_xid), &_xid);
996 if (xp == NULL)
997 goto dropit;
998
999 /* Look up and lock the request corresponding to the given XID */
Chuck Lever4a0f8c02005-08-11 16:25:32 -04001000 spin_lock(&xprt->transport_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001001 rovr = xprt_lookup_rqst(xprt, *xp);
1002 if (!rovr)
1003 goto out_unlock;
1004 task = rovr->rq_task;
1005
Chuck Levera246b012005-08-11 16:25:23 -04001006 if ((copied = rovr->rq_private_buf.buflen) > repsize)
1007 copied = repsize;
1008
1009 /* Suck it into the iovec, verify checksum if not done by hw. */
Herbert Xu1781f7f2007-12-11 11:30:32 -08001010 if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb)) {
1011 UDPX_INC_STATS_BH(sk, UDP_MIB_INERRORS);
Chuck Levera246b012005-08-11 16:25:23 -04001012 goto out_unlock;
Herbert Xu1781f7f2007-12-11 11:30:32 -08001013 }
1014
1015 UDPX_INC_STATS_BH(sk, UDP_MIB_INDATAGRAMS);
Chuck Levera246b012005-08-11 16:25:23 -04001016
1017 /* Something worked... */
Eric Dumazetadf30902009-06-02 05:19:30 +00001018 dst_confirm(skb_dst(skb));
Chuck Levera246b012005-08-11 16:25:23 -04001019
Chuck Lever1570c1e2005-08-25 16:25:52 -07001020 xprt_adjust_cwnd(task, copied);
Chuck Lever1570c1e2005-08-25 16:25:52 -07001021 xprt_complete_rqst(task, copied);
Chuck Levera246b012005-08-11 16:25:23 -04001022
1023 out_unlock:
Chuck Lever4a0f8c02005-08-11 16:25:32 -04001024 spin_unlock(&xprt->transport_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001025 dropit:
1026 skb_free_datagram(sk, skb);
1027 out:
Eric Dumazetf064af12010-09-22 12:43:39 +00001028 read_unlock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001029}
1030
Chuck Leverdd456472006-12-05 16:35:44 -05001031static inline void xs_tcp_read_fraghdr(struct rpc_xprt *xprt, struct xdr_skb_reader *desc)
Chuck Levera246b012005-08-11 16:25:23 -04001032{
Chuck Lever51971132006-12-05 16:35:19 -05001033 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -04001034 size_t len, used;
1035 char *p;
1036
Chuck Lever51971132006-12-05 16:35:19 -05001037 p = ((char *) &transport->tcp_fraghdr) + transport->tcp_offset;
1038 len = sizeof(transport->tcp_fraghdr) - transport->tcp_offset;
Chuck Lever9d292312006-12-05 16:35:41 -05001039 used = xdr_skb_read_bits(desc, p, len);
Chuck Lever51971132006-12-05 16:35:19 -05001040 transport->tcp_offset += used;
Chuck Levera246b012005-08-11 16:25:23 -04001041 if (used != len)
1042 return;
Chuck Lever808012f2005-08-25 16:25:49 -07001043
Chuck Lever51971132006-12-05 16:35:19 -05001044 transport->tcp_reclen = ntohl(transport->tcp_fraghdr);
1045 if (transport->tcp_reclen & RPC_LAST_STREAM_FRAGMENT)
Chuck Levere136d092006-12-05 16:35:23 -05001046 transport->tcp_flags |= TCP_RCV_LAST_FRAG;
Chuck Levera246b012005-08-11 16:25:23 -04001047 else
Chuck Levere136d092006-12-05 16:35:23 -05001048 transport->tcp_flags &= ~TCP_RCV_LAST_FRAG;
Chuck Lever51971132006-12-05 16:35:19 -05001049 transport->tcp_reclen &= RPC_FRAGMENT_SIZE_MASK;
Chuck Lever808012f2005-08-25 16:25:49 -07001050
Chuck Levere136d092006-12-05 16:35:23 -05001051 transport->tcp_flags &= ~TCP_RCV_COPY_FRAGHDR;
Chuck Lever51971132006-12-05 16:35:19 -05001052 transport->tcp_offset = 0;
Chuck Lever808012f2005-08-25 16:25:49 -07001053
Chuck Levera246b012005-08-11 16:25:23 -04001054 /* Sanity check of the record length */
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001055 if (unlikely(transport->tcp_reclen < 8)) {
Chuck Lever46121cf2007-01-31 12:14:08 -05001056 dprintk("RPC: invalid TCP record fragment length\n");
Trond Myklebust3ebb0672007-11-06 18:40:12 -05001057 xprt_force_disconnect(xprt);
Chuck Lever9903cd12005-08-11 16:25:26 -04001058 return;
Chuck Levera246b012005-08-11 16:25:23 -04001059 }
Chuck Lever46121cf2007-01-31 12:14:08 -05001060 dprintk("RPC: reading TCP record fragment of length %d\n",
Chuck Lever51971132006-12-05 16:35:19 -05001061 transport->tcp_reclen);
Chuck Levera246b012005-08-11 16:25:23 -04001062}
1063
Chuck Lever51971132006-12-05 16:35:19 -05001064static void xs_tcp_check_fraghdr(struct sock_xprt *transport)
Chuck Levera246b012005-08-11 16:25:23 -04001065{
Chuck Lever51971132006-12-05 16:35:19 -05001066 if (transport->tcp_offset == transport->tcp_reclen) {
Chuck Levere136d092006-12-05 16:35:23 -05001067 transport->tcp_flags |= TCP_RCV_COPY_FRAGHDR;
Chuck Lever51971132006-12-05 16:35:19 -05001068 transport->tcp_offset = 0;
Chuck Levere136d092006-12-05 16:35:23 -05001069 if (transport->tcp_flags & TCP_RCV_LAST_FRAG) {
1070 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
1071 transport->tcp_flags |= TCP_RCV_COPY_XID;
Chuck Lever51971132006-12-05 16:35:19 -05001072 transport->tcp_copied = 0;
Chuck Levera246b012005-08-11 16:25:23 -04001073 }
1074 }
1075}
1076
Chuck Leverdd456472006-12-05 16:35:44 -05001077static inline void xs_tcp_read_xid(struct sock_xprt *transport, struct xdr_skb_reader *desc)
Chuck Levera246b012005-08-11 16:25:23 -04001078{
1079 size_t len, used;
1080 char *p;
1081
Chuck Lever51971132006-12-05 16:35:19 -05001082 len = sizeof(transport->tcp_xid) - transport->tcp_offset;
Chuck Lever46121cf2007-01-31 12:14:08 -05001083 dprintk("RPC: reading XID (%Zu bytes)\n", len);
Chuck Lever51971132006-12-05 16:35:19 -05001084 p = ((char *) &transport->tcp_xid) + transport->tcp_offset;
Chuck Lever9d292312006-12-05 16:35:41 -05001085 used = xdr_skb_read_bits(desc, p, len);
Chuck Lever51971132006-12-05 16:35:19 -05001086 transport->tcp_offset += used;
Chuck Levera246b012005-08-11 16:25:23 -04001087 if (used != len)
1088 return;
Chuck Levere136d092006-12-05 16:35:23 -05001089 transport->tcp_flags &= ~TCP_RCV_COPY_XID;
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001090 transport->tcp_flags |= TCP_RCV_READ_CALLDIR;
Chuck Lever51971132006-12-05 16:35:19 -05001091 transport->tcp_copied = 4;
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001092 dprintk("RPC: reading %s XID %08x\n",
1093 (transport->tcp_flags & TCP_RPC_REPLY) ? "reply for"
1094 : "request with",
Chuck Lever51971132006-12-05 16:35:19 -05001095 ntohl(transport->tcp_xid));
1096 xs_tcp_check_fraghdr(transport);
Chuck Levera246b012005-08-11 16:25:23 -04001097}
1098
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001099static inline void xs_tcp_read_calldir(struct sock_xprt *transport,
1100 struct xdr_skb_reader *desc)
Chuck Levera246b012005-08-11 16:25:23 -04001101{
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001102 size_t len, used;
1103 u32 offset;
Trond Myklebustb76ce562010-06-16 13:57:32 -04001104 char *p;
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001105
1106 /*
1107 * We want transport->tcp_offset to be 8 at the end of this routine
1108 * (4 bytes for the xid and 4 bytes for the call/reply flag).
1109 * When this function is called for the first time,
1110 * transport->tcp_offset is 4 (after having already read the xid).
1111 */
1112 offset = transport->tcp_offset - sizeof(transport->tcp_xid);
Trond Myklebustb76ce562010-06-16 13:57:32 -04001113 len = sizeof(transport->tcp_calldir) - offset;
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001114 dprintk("RPC: reading CALL/REPLY flag (%Zu bytes)\n", len);
Trond Myklebustb76ce562010-06-16 13:57:32 -04001115 p = ((char *) &transport->tcp_calldir) + offset;
1116 used = xdr_skb_read_bits(desc, p, len);
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001117 transport->tcp_offset += used;
1118 if (used != len)
1119 return;
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001120 transport->tcp_flags &= ~TCP_RCV_READ_CALLDIR;
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001121 /*
1122 * We don't yet have the XDR buffer, so we will write the calldir
1123 * out after we get the buffer from the 'struct rpc_rqst'
1124 */
Trond Myklebustb76ce562010-06-16 13:57:32 -04001125 switch (ntohl(transport->tcp_calldir)) {
1126 case RPC_REPLY:
1127 transport->tcp_flags |= TCP_RCV_COPY_CALLDIR;
1128 transport->tcp_flags |= TCP_RCV_COPY_DATA;
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001129 transport->tcp_flags |= TCP_RPC_REPLY;
Trond Myklebustb76ce562010-06-16 13:57:32 -04001130 break;
1131 case RPC_CALL:
1132 transport->tcp_flags |= TCP_RCV_COPY_CALLDIR;
1133 transport->tcp_flags |= TCP_RCV_COPY_DATA;
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001134 transport->tcp_flags &= ~TCP_RPC_REPLY;
Trond Myklebustb76ce562010-06-16 13:57:32 -04001135 break;
1136 default:
1137 dprintk("RPC: invalid request message type\n");
1138 xprt_force_disconnect(&transport->xprt);
1139 }
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001140 xs_tcp_check_fraghdr(transport);
1141}
1142
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001143static inline void xs_tcp_read_common(struct rpc_xprt *xprt,
1144 struct xdr_skb_reader *desc,
1145 struct rpc_rqst *req)
Chuck Levera246b012005-08-11 16:25:23 -04001146{
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001147 struct sock_xprt *transport =
1148 container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -04001149 struct xdr_buf *rcvbuf;
1150 size_t len;
1151 ssize_t r;
1152
Chuck Levera246b012005-08-11 16:25:23 -04001153 rcvbuf = &req->rq_private_buf;
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001154
1155 if (transport->tcp_flags & TCP_RCV_COPY_CALLDIR) {
1156 /*
1157 * Save the RPC direction in the XDR buffer
1158 */
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001159 memcpy(rcvbuf->head[0].iov_base + transport->tcp_copied,
Trond Myklebustb76ce562010-06-16 13:57:32 -04001160 &transport->tcp_calldir,
1161 sizeof(transport->tcp_calldir));
1162 transport->tcp_copied += sizeof(transport->tcp_calldir);
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001163 transport->tcp_flags &= ~TCP_RCV_COPY_CALLDIR;
Chuck Levera246b012005-08-11 16:25:23 -04001164 }
1165
Chuck Levera246b012005-08-11 16:25:23 -04001166 len = desc->count;
Chuck Lever51971132006-12-05 16:35:19 -05001167 if (len > transport->tcp_reclen - transport->tcp_offset) {
Chuck Leverdd456472006-12-05 16:35:44 -05001168 struct xdr_skb_reader my_desc;
Chuck Levera246b012005-08-11 16:25:23 -04001169
Chuck Lever51971132006-12-05 16:35:19 -05001170 len = transport->tcp_reclen - transport->tcp_offset;
Chuck Levera246b012005-08-11 16:25:23 -04001171 memcpy(&my_desc, desc, sizeof(my_desc));
1172 my_desc.count = len;
Chuck Lever51971132006-12-05 16:35:19 -05001173 r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied,
Chuck Lever9d292312006-12-05 16:35:41 -05001174 &my_desc, xdr_skb_read_bits);
Chuck Levera246b012005-08-11 16:25:23 -04001175 desc->count -= r;
1176 desc->offset += r;
1177 } else
Chuck Lever51971132006-12-05 16:35:19 -05001178 r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied,
Chuck Lever9d292312006-12-05 16:35:41 -05001179 desc, xdr_skb_read_bits);
Chuck Levera246b012005-08-11 16:25:23 -04001180
1181 if (r > 0) {
Chuck Lever51971132006-12-05 16:35:19 -05001182 transport->tcp_copied += r;
1183 transport->tcp_offset += r;
Chuck Levera246b012005-08-11 16:25:23 -04001184 }
1185 if (r != len) {
1186 /* Error when copying to the receive buffer,
1187 * usually because we weren't able to allocate
1188 * additional buffer pages. All we can do now
Chuck Levere136d092006-12-05 16:35:23 -05001189 * is turn off TCP_RCV_COPY_DATA, so the request
Chuck Levera246b012005-08-11 16:25:23 -04001190 * will not receive any additional updates,
1191 * and time out.
1192 * Any remaining data from this record will
1193 * be discarded.
1194 */
Chuck Levere136d092006-12-05 16:35:23 -05001195 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
Chuck Lever46121cf2007-01-31 12:14:08 -05001196 dprintk("RPC: XID %08x truncated request\n",
Chuck Lever51971132006-12-05 16:35:19 -05001197 ntohl(transport->tcp_xid));
Chuck Lever46121cf2007-01-31 12:14:08 -05001198 dprintk("RPC: xprt = %p, tcp_copied = %lu, "
1199 "tcp_offset = %u, tcp_reclen = %u\n",
1200 xprt, transport->tcp_copied,
1201 transport->tcp_offset, transport->tcp_reclen);
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001202 return;
Chuck Levera246b012005-08-11 16:25:23 -04001203 }
1204
Chuck Lever46121cf2007-01-31 12:14:08 -05001205 dprintk("RPC: XID %08x read %Zd bytes\n",
Chuck Lever51971132006-12-05 16:35:19 -05001206 ntohl(transport->tcp_xid), r);
Chuck Lever46121cf2007-01-31 12:14:08 -05001207 dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, "
1208 "tcp_reclen = %u\n", xprt, transport->tcp_copied,
1209 transport->tcp_offset, transport->tcp_reclen);
Chuck Levera246b012005-08-11 16:25:23 -04001210
Chuck Lever51971132006-12-05 16:35:19 -05001211 if (transport->tcp_copied == req->rq_private_buf.buflen)
Chuck Levere136d092006-12-05 16:35:23 -05001212 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
Chuck Lever51971132006-12-05 16:35:19 -05001213 else if (transport->tcp_offset == transport->tcp_reclen) {
Chuck Levere136d092006-12-05 16:35:23 -05001214 if (transport->tcp_flags & TCP_RCV_LAST_FRAG)
1215 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
Chuck Levera246b012005-08-11 16:25:23 -04001216 }
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001217}
1218
1219/*
1220 * Finds the request corresponding to the RPC xid and invokes the common
1221 * tcp read code to read the data.
1222 */
1223static inline int xs_tcp_read_reply(struct rpc_xprt *xprt,
1224 struct xdr_skb_reader *desc)
1225{
1226 struct sock_xprt *transport =
1227 container_of(xprt, struct sock_xprt, xprt);
1228 struct rpc_rqst *req;
1229
1230 dprintk("RPC: read reply XID %08x\n", ntohl(transport->tcp_xid));
1231
1232 /* Find and lock the request corresponding to this xid */
1233 spin_lock(&xprt->transport_lock);
1234 req = xprt_lookup_rqst(xprt, transport->tcp_xid);
1235 if (!req) {
1236 dprintk("RPC: XID %08x request not found!\n",
1237 ntohl(transport->tcp_xid));
1238 spin_unlock(&xprt->transport_lock);
1239 return -1;
1240 }
1241
1242 xs_tcp_read_common(xprt, desc, req);
1243
Chuck Levere136d092006-12-05 16:35:23 -05001244 if (!(transport->tcp_flags & TCP_RCV_COPY_DATA))
Chuck Lever51971132006-12-05 16:35:19 -05001245 xprt_complete_rqst(req->rq_task, transport->tcp_copied);
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001246
Chuck Lever4a0f8c02005-08-11 16:25:32 -04001247 spin_unlock(&xprt->transport_lock);
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001248 return 0;
1249}
1250
Trond Myklebust9e00abc2011-07-13 19:20:49 -04001251#if defined(CONFIG_SUNRPC_BACKCHANNEL)
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001252/*
1253 * Obtains an rpc_rqst previously allocated and invokes the common
1254 * tcp read code to read the data. The result is placed in the callback
1255 * queue.
1256 * If we're unable to obtain the rpc_rqst we schedule the closing of the
1257 * connection and return -1.
1258 */
1259static inline int xs_tcp_read_callback(struct rpc_xprt *xprt,
1260 struct xdr_skb_reader *desc)
1261{
1262 struct sock_xprt *transport =
1263 container_of(xprt, struct sock_xprt, xprt);
1264 struct rpc_rqst *req;
1265
1266 req = xprt_alloc_bc_request(xprt);
1267 if (req == NULL) {
1268 printk(KERN_WARNING "Callback slot table overflowed\n");
1269 xprt_force_disconnect(xprt);
1270 return -1;
1271 }
1272
1273 req->rq_xid = transport->tcp_xid;
1274 dprintk("RPC: read callback XID %08x\n", ntohl(req->rq_xid));
1275 xs_tcp_read_common(xprt, desc, req);
1276
1277 if (!(transport->tcp_flags & TCP_RCV_COPY_DATA)) {
1278 struct svc_serv *bc_serv = xprt->bc_serv;
1279
1280 /*
1281 * Add callback request to callback list. The callback
1282 * service sleeps on the sv_cb_waitq waiting for new
1283 * requests. Wake it up after adding enqueing the
1284 * request.
1285 */
1286 dprintk("RPC: add callback request to list\n");
1287 spin_lock(&bc_serv->sv_cb_lock);
1288 list_add(&req->rq_bc_list, &bc_serv->sv_cb_list);
1289 spin_unlock(&bc_serv->sv_cb_lock);
1290 wake_up(&bc_serv->sv_cb_waitq);
1291 }
1292
1293 req->rq_private_buf.len = transport->tcp_copied;
1294
1295 return 0;
1296}
1297
1298static inline int _xs_tcp_read_data(struct rpc_xprt *xprt,
1299 struct xdr_skb_reader *desc)
1300{
1301 struct sock_xprt *transport =
1302 container_of(xprt, struct sock_xprt, xprt);
1303
1304 return (transport->tcp_flags & TCP_RPC_REPLY) ?
1305 xs_tcp_read_reply(xprt, desc) :
1306 xs_tcp_read_callback(xprt, desc);
1307}
1308#else
1309static inline int _xs_tcp_read_data(struct rpc_xprt *xprt,
1310 struct xdr_skb_reader *desc)
1311{
1312 return xs_tcp_read_reply(xprt, desc);
1313}
Trond Myklebust9e00abc2011-07-13 19:20:49 -04001314#endif /* CONFIG_SUNRPC_BACKCHANNEL */
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001315
1316/*
1317 * Read data off the transport. This can be either an RPC_CALL or an
1318 * RPC_REPLY. Relay the processing to helper functions.
1319 */
1320static void xs_tcp_read_data(struct rpc_xprt *xprt,
1321 struct xdr_skb_reader *desc)
1322{
1323 struct sock_xprt *transport =
1324 container_of(xprt, struct sock_xprt, xprt);
1325
1326 if (_xs_tcp_read_data(xprt, desc) == 0)
1327 xs_tcp_check_fraghdr(transport);
1328 else {
1329 /*
1330 * The transport_lock protects the request handling.
1331 * There's no need to hold it to update the tcp_flags.
1332 */
1333 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
1334 }
Chuck Levera246b012005-08-11 16:25:23 -04001335}
1336
Chuck Leverdd456472006-12-05 16:35:44 -05001337static inline void xs_tcp_read_discard(struct sock_xprt *transport, struct xdr_skb_reader *desc)
Chuck Levera246b012005-08-11 16:25:23 -04001338{
1339 size_t len;
1340
Chuck Lever51971132006-12-05 16:35:19 -05001341 len = transport->tcp_reclen - transport->tcp_offset;
Chuck Levera246b012005-08-11 16:25:23 -04001342 if (len > desc->count)
1343 len = desc->count;
1344 desc->count -= len;
1345 desc->offset += len;
Chuck Lever51971132006-12-05 16:35:19 -05001346 transport->tcp_offset += len;
Chuck Lever46121cf2007-01-31 12:14:08 -05001347 dprintk("RPC: discarded %Zu bytes\n", len);
Chuck Lever51971132006-12-05 16:35:19 -05001348 xs_tcp_check_fraghdr(transport);
Chuck Levera246b012005-08-11 16:25:23 -04001349}
1350
Chuck Lever9903cd12005-08-11 16:25:26 -04001351static 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 -04001352{
1353 struct rpc_xprt *xprt = rd_desc->arg.data;
Chuck Lever51971132006-12-05 16:35:19 -05001354 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Leverdd456472006-12-05 16:35:44 -05001355 struct xdr_skb_reader desc = {
Chuck Levera246b012005-08-11 16:25:23 -04001356 .skb = skb,
1357 .offset = offset,
1358 .count = len,
Chuck Lever9903cd12005-08-11 16:25:26 -04001359 };
Chuck Levera246b012005-08-11 16:25:23 -04001360
Chuck Lever46121cf2007-01-31 12:14:08 -05001361 dprintk("RPC: xs_tcp_data_recv started\n");
Chuck Levera246b012005-08-11 16:25:23 -04001362 do {
1363 /* Read in a new fragment marker if necessary */
1364 /* Can we ever really expect to get completely empty fragments? */
Chuck Levere136d092006-12-05 16:35:23 -05001365 if (transport->tcp_flags & TCP_RCV_COPY_FRAGHDR) {
Chuck Lever9903cd12005-08-11 16:25:26 -04001366 xs_tcp_read_fraghdr(xprt, &desc);
Chuck Levera246b012005-08-11 16:25:23 -04001367 continue;
1368 }
1369 /* Read in the xid if necessary */
Chuck Levere136d092006-12-05 16:35:23 -05001370 if (transport->tcp_flags & TCP_RCV_COPY_XID) {
Chuck Lever51971132006-12-05 16:35:19 -05001371 xs_tcp_read_xid(transport, &desc);
Chuck Levera246b012005-08-11 16:25:23 -04001372 continue;
1373 }
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001374 /* Read in the call/reply flag */
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001375 if (transport->tcp_flags & TCP_RCV_READ_CALLDIR) {
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001376 xs_tcp_read_calldir(transport, &desc);
1377 continue;
1378 }
Chuck Levera246b012005-08-11 16:25:23 -04001379 /* Read in the request data */
Chuck Levere136d092006-12-05 16:35:23 -05001380 if (transport->tcp_flags & TCP_RCV_COPY_DATA) {
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001381 xs_tcp_read_data(xprt, &desc);
Chuck Levera246b012005-08-11 16:25:23 -04001382 continue;
1383 }
1384 /* Skip over any trailing bytes on short reads */
Chuck Lever51971132006-12-05 16:35:19 -05001385 xs_tcp_read_discard(transport, &desc);
Chuck Levera246b012005-08-11 16:25:23 -04001386 } while (desc.count);
Chuck Lever46121cf2007-01-31 12:14:08 -05001387 dprintk("RPC: xs_tcp_data_recv done\n");
Chuck Levera246b012005-08-11 16:25:23 -04001388 return len - desc.count;
1389}
1390
Chuck Lever9903cd12005-08-11 16:25:26 -04001391/**
1392 * xs_tcp_data_ready - "data ready" callback for TCP sockets
1393 * @sk: socket with data to read
1394 * @bytes: how much data to read
1395 *
1396 */
1397static void xs_tcp_data_ready(struct sock *sk, int bytes)
Chuck Levera246b012005-08-11 16:25:23 -04001398{
1399 struct rpc_xprt *xprt;
1400 read_descriptor_t rd_desc;
Trond Myklebustff2d7db2008-02-25 21:40:51 -08001401 int read;
Chuck Levera246b012005-08-11 16:25:23 -04001402
Chuck Lever46121cf2007-01-31 12:14:08 -05001403 dprintk("RPC: xs_tcp_data_ready...\n");
1404
Eric Dumazetf064af12010-09-22 12:43:39 +00001405 read_lock_bh(&sk->sk_callback_lock);
Chuck Lever9903cd12005-08-11 16:25:26 -04001406 if (!(xprt = xprt_from_sock(sk)))
Chuck Levera246b012005-08-11 16:25:23 -04001407 goto out;
Chuck Levera246b012005-08-11 16:25:23 -04001408 if (xprt->shutdown)
1409 goto out;
1410
Neil Brown61d0a8e2009-09-23 14:36:37 -04001411 /* Any data means we had a useful conversation, so
1412 * the we don't need to delay the next reconnect
1413 */
1414 if (xprt->reestablish_timeout)
1415 xprt->reestablish_timeout = 0;
1416
Chuck Lever9903cd12005-08-11 16:25:26 -04001417 /* We use rd_desc to pass struct xprt to xs_tcp_data_recv */
Chuck Levera246b012005-08-11 16:25:23 -04001418 rd_desc.arg.data = xprt;
Trond Myklebustff2d7db2008-02-25 21:40:51 -08001419 do {
1420 rd_desc.count = 65536;
1421 read = tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv);
1422 } while (read > 0);
Chuck Levera246b012005-08-11 16:25:23 -04001423out:
Eric Dumazetf064af12010-09-22 12:43:39 +00001424 read_unlock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001425}
1426
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001427/*
1428 * Do the equivalent of linger/linger2 handling for dealing with
1429 * broken servers that don't close the socket in a timely
1430 * fashion
1431 */
1432static void xs_tcp_schedule_linger_timeout(struct rpc_xprt *xprt,
1433 unsigned long timeout)
1434{
1435 struct sock_xprt *transport;
1436
1437 if (xprt_test_and_set_connecting(xprt))
1438 return;
1439 set_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1440 transport = container_of(xprt, struct sock_xprt, xprt);
1441 queue_delayed_work(rpciod_workqueue, &transport->connect_worker,
1442 timeout);
1443}
1444
1445static void xs_tcp_cancel_linger_timeout(struct rpc_xprt *xprt)
1446{
1447 struct sock_xprt *transport;
1448
1449 transport = container_of(xprt, struct sock_xprt, xprt);
1450
1451 if (!test_bit(XPRT_CONNECTION_ABORT, &xprt->state) ||
1452 !cancel_delayed_work(&transport->connect_worker))
1453 return;
1454 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1455 xprt_clear_connecting(xprt);
1456}
1457
1458static void xs_sock_mark_closed(struct rpc_xprt *xprt)
1459{
1460 smp_mb__before_clear_bit();
1461 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
1462 clear_bit(XPRT_CLOSING, &xprt->state);
1463 smp_mb__after_clear_bit();
1464 /* Mark transport as closed and wake up all pending tasks */
1465 xprt_disconnect_done(xprt);
1466}
1467
Chuck Lever9903cd12005-08-11 16:25:26 -04001468/**
1469 * xs_tcp_state_change - callback to handle TCP socket state changes
1470 * @sk: socket whose state has changed
1471 *
1472 */
1473static void xs_tcp_state_change(struct sock *sk)
Chuck Levera246b012005-08-11 16:25:23 -04001474{
Chuck Lever9903cd12005-08-11 16:25:26 -04001475 struct rpc_xprt *xprt;
Chuck Levera246b012005-08-11 16:25:23 -04001476
Eric Dumazetf064af12010-09-22 12:43:39 +00001477 read_lock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001478 if (!(xprt = xprt_from_sock(sk)))
1479 goto out;
Chuck Lever46121cf2007-01-31 12:14:08 -05001480 dprintk("RPC: xs_tcp_state_change client %p...\n", xprt);
Andy Chittenden669502f2010-08-10 10:19:53 -04001481 dprintk("RPC: state %x conn %d dead %d zapped %d sk_shutdown %d\n",
Chuck Lever46121cf2007-01-31 12:14:08 -05001482 sk->sk_state, xprt_connected(xprt),
1483 sock_flag(sk, SOCK_DEAD),
Andy Chittenden669502f2010-08-10 10:19:53 -04001484 sock_flag(sk, SOCK_ZAPPED),
1485 sk->sk_shutdown);
Chuck Levera246b012005-08-11 16:25:23 -04001486
1487 switch (sk->sk_state) {
1488 case TCP_ESTABLISHED:
Eric Dumazetf064af12010-09-22 12:43:39 +00001489 spin_lock(&xprt->transport_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001490 if (!xprt_test_and_set_connected(xprt)) {
Chuck Lever51971132006-12-05 16:35:19 -05001491 struct sock_xprt *transport = container_of(xprt,
1492 struct sock_xprt, xprt);
1493
Chuck Levera246b012005-08-11 16:25:23 -04001494 /* Reset TCP record info */
Chuck Lever51971132006-12-05 16:35:19 -05001495 transport->tcp_offset = 0;
1496 transport->tcp_reclen = 0;
1497 transport->tcp_copied = 0;
Chuck Levere136d092006-12-05 16:35:23 -05001498 transport->tcp_flags =
1499 TCP_RCV_COPY_FRAGHDR | TCP_RCV_COPY_XID;
Chuck Lever51971132006-12-05 16:35:19 -05001500
Trond Myklebust2a491992009-03-11 14:38:00 -04001501 xprt_wake_pending_tasks(xprt, -EAGAIN);
Chuck Levera246b012005-08-11 16:25:23 -04001502 }
Eric Dumazetf064af12010-09-22 12:43:39 +00001503 spin_unlock(&xprt->transport_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001504 break;
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001505 case TCP_FIN_WAIT1:
1506 /* The client initiated a shutdown of the socket */
Trond Myklebust7c1d71c2008-04-17 16:52:57 -04001507 xprt->connect_cookie++;
Trond Myklebust663b8852008-01-01 18:42:12 -05001508 xprt->reestablish_timeout = 0;
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001509 set_bit(XPRT_CLOSING, &xprt->state);
1510 smp_mb__before_clear_bit();
1511 clear_bit(XPRT_CONNECTED, &xprt->state);
Trond Myklebustef803672007-12-31 16:19:17 -05001512 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001513 smp_mb__after_clear_bit();
Trond Myklebust25fe6142009-03-11 14:38:03 -04001514 xs_tcp_schedule_linger_timeout(xprt, xs_tcp_fin_timeout);
Chuck Levera246b012005-08-11 16:25:23 -04001515 break;
Trond Myklebust632e3bd2006-01-03 09:55:55 +01001516 case TCP_CLOSE_WAIT:
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001517 /* The server initiated a shutdown of the socket */
Trond Myklebust66af1e52007-11-06 10:18:36 -05001518 xprt_force_disconnect(xprt);
Trond Myklebust7c1d71c2008-04-17 16:52:57 -04001519 xprt->connect_cookie++;
Trond Myklebust663b8852008-01-01 18:42:12 -05001520 case TCP_CLOSING:
1521 /*
1522 * If the server closed down the connection, make sure that
1523 * we back off before reconnecting
1524 */
1525 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
1526 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001527 break;
1528 case TCP_LAST_ACK:
Trond Myklebust670f9452009-03-11 14:37:58 -04001529 set_bit(XPRT_CLOSING, &xprt->state);
Trond Myklebust25fe6142009-03-11 14:38:03 -04001530 xs_tcp_schedule_linger_timeout(xprt, xs_tcp_fin_timeout);
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001531 smp_mb__before_clear_bit();
1532 clear_bit(XPRT_CONNECTED, &xprt->state);
1533 smp_mb__after_clear_bit();
1534 break;
1535 case TCP_CLOSE:
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001536 xs_tcp_cancel_linger_timeout(xprt);
1537 xs_sock_mark_closed(xprt);
Chuck Levera246b012005-08-11 16:25:23 -04001538 }
1539 out:
Eric Dumazetf064af12010-09-22 12:43:39 +00001540 read_unlock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001541}
1542
Chuck Lever9903cd12005-08-11 16:25:26 -04001543/**
Trond Myklebust482f32e2009-03-11 14:38:00 -04001544 * xs_error_report - callback mainly for catching socket errors
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04001545 * @sk: socket
1546 */
Trond Myklebust482f32e2009-03-11 14:38:00 -04001547static void xs_error_report(struct sock *sk)
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04001548{
1549 struct rpc_xprt *xprt;
1550
Eric Dumazetf064af12010-09-22 12:43:39 +00001551 read_lock_bh(&sk->sk_callback_lock);
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04001552 if (!(xprt = xprt_from_sock(sk)))
1553 goto out;
1554 dprintk("RPC: %s client %p...\n"
1555 "RPC: error %d\n",
1556 __func__, xprt, sk->sk_err);
Trond Myklebust482f32e2009-03-11 14:38:00 -04001557 xprt_wake_pending_tasks(xprt, -EAGAIN);
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04001558out:
Eric Dumazetf064af12010-09-22 12:43:39 +00001559 read_unlock_bh(&sk->sk_callback_lock);
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04001560}
1561
Ilpo Järvinen1f0fa152009-02-06 23:48:33 -08001562static void xs_write_space(struct sock *sk)
1563{
1564 struct socket *sock;
1565 struct rpc_xprt *xprt;
1566
1567 if (unlikely(!(sock = sk->sk_socket)))
1568 return;
1569 clear_bit(SOCK_NOSPACE, &sock->flags);
1570
1571 if (unlikely(!(xprt = xprt_from_sock(sk))))
1572 return;
1573 if (test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags) == 0)
1574 return;
1575
1576 xprt_write_space(xprt);
1577}
1578
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04001579/**
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001580 * xs_udp_write_space - callback invoked when socket buffer space
1581 * becomes available
Chuck Lever9903cd12005-08-11 16:25:26 -04001582 * @sk: socket whose state has changed
1583 *
Chuck Levera246b012005-08-11 16:25:23 -04001584 * Called when more output buffer space is available for this socket.
1585 * We try not to wake our writers until they can make "significant"
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001586 * progress, otherwise we'll waste resources thrashing kernel_sendmsg
Chuck Levera246b012005-08-11 16:25:23 -04001587 * with a bunch of small requests.
1588 */
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001589static void xs_udp_write_space(struct sock *sk)
Chuck Levera246b012005-08-11 16:25:23 -04001590{
Eric Dumazetf064af12010-09-22 12:43:39 +00001591 read_lock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001592
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001593 /* from net/core/sock.c:sock_def_write_space */
Ilpo Järvinen1f0fa152009-02-06 23:48:33 -08001594 if (sock_writeable(sk))
1595 xs_write_space(sk);
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001596
Eric Dumazetf064af12010-09-22 12:43:39 +00001597 read_unlock_bh(&sk->sk_callback_lock);
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001598}
Chuck Levera246b012005-08-11 16:25:23 -04001599
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001600/**
1601 * xs_tcp_write_space - callback invoked when socket buffer space
1602 * becomes available
1603 * @sk: socket whose state has changed
1604 *
1605 * Called when more output buffer space is available for this socket.
1606 * We try not to wake our writers until they can make "significant"
1607 * progress, otherwise we'll waste resources thrashing kernel_sendmsg
1608 * with a bunch of small requests.
1609 */
1610static void xs_tcp_write_space(struct sock *sk)
1611{
Eric Dumazetf064af12010-09-22 12:43:39 +00001612 read_lock_bh(&sk->sk_callback_lock);
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001613
1614 /* from net/core/stream.c:sk_stream_write_space */
Ilpo Järvinen1f0fa152009-02-06 23:48:33 -08001615 if (sk_stream_wspace(sk) >= sk_stream_min_wspace(sk))
1616 xs_write_space(sk);
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001617
Eric Dumazetf064af12010-09-22 12:43:39 +00001618 read_unlock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001619}
1620
Chuck Lever470056c2005-08-25 16:25:56 -07001621static void xs_udp_do_set_buffer_size(struct rpc_xprt *xprt)
Chuck Levera246b012005-08-11 16:25:23 -04001622{
Chuck Leveree0ac0c2006-12-05 16:35:15 -05001623 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1624 struct sock *sk = transport->inet;
Chuck Levera246b012005-08-11 16:25:23 -04001625
Chuck Lever7c6e0662006-12-05 16:35:30 -05001626 if (transport->rcvsize) {
Chuck Levera246b012005-08-11 16:25:23 -04001627 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
Chuck Lever7c6e0662006-12-05 16:35:30 -05001628 sk->sk_rcvbuf = transport->rcvsize * xprt->max_reqs * 2;
Chuck Levera246b012005-08-11 16:25:23 -04001629 }
Chuck Lever7c6e0662006-12-05 16:35:30 -05001630 if (transport->sndsize) {
Chuck Levera246b012005-08-11 16:25:23 -04001631 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
Chuck Lever7c6e0662006-12-05 16:35:30 -05001632 sk->sk_sndbuf = transport->sndsize * xprt->max_reqs * 2;
Chuck Levera246b012005-08-11 16:25:23 -04001633 sk->sk_write_space(sk);
1634 }
1635}
1636
Chuck Lever43118c22005-08-25 16:25:49 -07001637/**
Chuck Lever470056c2005-08-25 16:25:56 -07001638 * xs_udp_set_buffer_size - set send and receive limits
Chuck Lever43118c22005-08-25 16:25:49 -07001639 * @xprt: generic transport
Chuck Lever470056c2005-08-25 16:25:56 -07001640 * @sndsize: requested size of send buffer, in bytes
1641 * @rcvsize: requested size of receive buffer, in bytes
Chuck Lever43118c22005-08-25 16:25:49 -07001642 *
Chuck Lever470056c2005-08-25 16:25:56 -07001643 * Set socket send and receive buffer size limits.
Chuck Lever43118c22005-08-25 16:25:49 -07001644 */
Chuck Lever470056c2005-08-25 16:25:56 -07001645static void xs_udp_set_buffer_size(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize)
Chuck Lever43118c22005-08-25 16:25:49 -07001646{
Chuck Lever7c6e0662006-12-05 16:35:30 -05001647 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1648
1649 transport->sndsize = 0;
Chuck Lever470056c2005-08-25 16:25:56 -07001650 if (sndsize)
Chuck Lever7c6e0662006-12-05 16:35:30 -05001651 transport->sndsize = sndsize + 1024;
1652 transport->rcvsize = 0;
Chuck Lever470056c2005-08-25 16:25:56 -07001653 if (rcvsize)
Chuck Lever7c6e0662006-12-05 16:35:30 -05001654 transport->rcvsize = rcvsize + 1024;
Chuck Lever470056c2005-08-25 16:25:56 -07001655
1656 xs_udp_do_set_buffer_size(xprt);
Chuck Lever43118c22005-08-25 16:25:49 -07001657}
1658
Chuck Lever46c0ee82005-08-25 16:25:52 -07001659/**
1660 * xs_udp_timer - called when a retransmit timeout occurs on a UDP transport
1661 * @task: task that timed out
1662 *
1663 * Adjust the congestion window after a retransmit timeout has occurred.
1664 */
1665static void xs_udp_timer(struct rpc_task *task)
1666{
1667 xprt_adjust_cwnd(task, -ETIMEDOUT);
1668}
1669
Chuck Leverb85d8802006-05-25 01:40:49 -04001670static unsigned short xs_get_random_port(void)
1671{
1672 unsigned short range = xprt_max_resvport - xprt_min_resvport;
1673 unsigned short rand = (unsigned short) net_random() % range;
1674 return rand + xprt_min_resvport;
1675}
1676
Chuck Lever92200412006-01-03 09:55:51 +01001677/**
1678 * xs_set_port - reset the port number in the remote endpoint address
1679 * @xprt: generic transport
1680 * @port: new port number
1681 *
1682 */
1683static void xs_set_port(struct rpc_xprt *xprt, unsigned short port)
1684{
Chuck Lever46121cf2007-01-31 12:14:08 -05001685 dprintk("RPC: setting port for xprt %p to %u\n", xprt, port);
Chuck Leverc4efcb12006-08-22 20:06:19 -04001686
Chuck Lever9dc3b092009-08-09 15:09:46 -04001687 rpc_set_port(xs_addr(xprt), port);
1688 xs_update_peer_port(xprt);
Chuck Lever92200412006-01-03 09:55:51 +01001689}
1690
Pavel Emelyanov5d4ec932010-10-04 16:51:23 +04001691static unsigned short xs_get_srcport(struct sock_xprt *transport)
Trond Myklebust67a391d2007-11-05 17:40:58 -05001692{
Chuck Leverfbfffbd2009-08-09 15:09:46 -04001693 unsigned short port = transport->srcport;
Trond Myklebust67a391d2007-11-05 17:40:58 -05001694
1695 if (port == 0 && transport->xprt.resvport)
1696 port = xs_get_random_port();
1697 return port;
1698}
1699
Pavel Emelyanovbaaf4e42010-10-04 16:51:56 +04001700static unsigned short xs_next_srcport(struct sock_xprt *transport, unsigned short port)
Trond Myklebust67a391d2007-11-05 17:40:58 -05001701{
Chuck Leverfbfffbd2009-08-09 15:09:46 -04001702 if (transport->srcport != 0)
1703 transport->srcport = 0;
Trond Myklebust67a391d2007-11-05 17:40:58 -05001704 if (!transport->xprt.resvport)
1705 return 0;
1706 if (port <= xprt_min_resvport || port > xprt_max_resvport)
1707 return xprt_max_resvport;
1708 return --port;
1709}
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001710static int xs_bind(struct sock_xprt *transport, struct socket *sock)
Chuck Levera246b012005-08-11 16:25:23 -04001711{
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001712 struct sockaddr_storage myaddr;
Trond Myklebust67a391d2007-11-05 17:40:58 -05001713 int err, nloop = 0;
Pavel Emelyanov5d4ec932010-10-04 16:51:23 +04001714 unsigned short port = xs_get_srcport(transport);
Trond Myklebust67a391d2007-11-05 17:40:58 -05001715 unsigned short last;
Chuck Levera246b012005-08-11 16:25:23 -04001716
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001717 memcpy(&myaddr, &transport->srcaddr, transport->xprt.addrlen);
Chuck Levera246b012005-08-11 16:25:23 -04001718 do {
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001719 rpc_set_port((struct sockaddr *)&myaddr, port);
1720 err = kernel_bind(sock, (struct sockaddr *)&myaddr,
1721 transport->xprt.addrlen);
Trond Myklebust67a391d2007-11-05 17:40:58 -05001722 if (port == 0)
Frank van Maarseveend3bc9a12007-07-09 22:23:35 +02001723 break;
Chuck Levera246b012005-08-11 16:25:23 -04001724 if (err == 0) {
Chuck Leverfbfffbd2009-08-09 15:09:46 -04001725 transport->srcport = port;
Frank van Maarseveend3bc9a12007-07-09 22:23:35 +02001726 break;
Chuck Levera246b012005-08-11 16:25:23 -04001727 }
Trond Myklebust67a391d2007-11-05 17:40:58 -05001728 last = port;
Pavel Emelyanovbaaf4e42010-10-04 16:51:56 +04001729 port = xs_next_srcport(transport, port);
Trond Myklebust67a391d2007-11-05 17:40:58 -05001730 if (port > last)
1731 nloop++;
1732 } while (err == -EADDRINUSE && nloop != 2);
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001733
Chuck Lever4232e862010-10-20 11:52:51 -04001734 if (myaddr.ss_family == AF_INET)
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001735 dprintk("RPC: %s %pI4:%u: %s (%d)\n", __func__,
1736 &((struct sockaddr_in *)&myaddr)->sin_addr,
1737 port, err ? "failed" : "ok", err);
1738 else
1739 dprintk("RPC: %s %pI6:%u: %s (%d)\n", __func__,
1740 &((struct sockaddr_in6 *)&myaddr)->sin6_addr,
1741 port, err ? "failed" : "ok", err);
Chuck Levera246b012005-08-11 16:25:23 -04001742 return err;
1743}
1744
Chuck Lever176e21e2011-05-09 15:22:44 -04001745/*
1746 * We don't support autobind on AF_LOCAL sockets
1747 */
1748static void xs_local_rpcbind(struct rpc_task *task)
1749{
1750 xprt_set_bound(task->tk_xprt);
1751}
1752
1753static void xs_local_set_port(struct rpc_xprt *xprt, unsigned short port)
1754{
1755}
Chuck Levera246b012005-08-11 16:25:23 -04001756
Peter Zijlstraed075362006-12-06 20:35:24 -08001757#ifdef CONFIG_DEBUG_LOCK_ALLOC
1758static struct lock_class_key xs_key[2];
1759static struct lock_class_key xs_slock_key[2];
1760
Chuck Lever176e21e2011-05-09 15:22:44 -04001761static inline void xs_reclassify_socketu(struct socket *sock)
1762{
1763 struct sock *sk = sock->sk;
1764
1765 BUG_ON(sock_owned_by_user(sk));
1766 sock_lock_init_class_and_name(sk, "slock-AF_LOCAL-RPC",
1767 &xs_slock_key[1], "sk_lock-AF_LOCAL-RPC", &xs_key[1]);
1768}
1769
Chuck Lever8945ee52007-08-06 11:58:04 -04001770static inline void xs_reclassify_socket4(struct socket *sock)
Peter Zijlstraed075362006-12-06 20:35:24 -08001771{
1772 struct sock *sk = sock->sk;
Chuck Lever8945ee52007-08-06 11:58:04 -04001773
John Heffner02b3d342007-09-12 10:42:12 +02001774 BUG_ON(sock_owned_by_user(sk));
Chuck Lever8945ee52007-08-06 11:58:04 -04001775 sock_lock_init_class_and_name(sk, "slock-AF_INET-RPC",
1776 &xs_slock_key[0], "sk_lock-AF_INET-RPC", &xs_key[0]);
1777}
Peter Zijlstraed075362006-12-06 20:35:24 -08001778
Chuck Lever8945ee52007-08-06 11:58:04 -04001779static inline void xs_reclassify_socket6(struct socket *sock)
1780{
1781 struct sock *sk = sock->sk;
Peter Zijlstraed075362006-12-06 20:35:24 -08001782
Linus Torvaldsf4921af2007-10-15 10:46:05 -07001783 BUG_ON(sock_owned_by_user(sk));
Chuck Lever8945ee52007-08-06 11:58:04 -04001784 sock_lock_init_class_and_name(sk, "slock-AF_INET6-RPC",
1785 &xs_slock_key[1], "sk_lock-AF_INET6-RPC", &xs_key[1]);
Peter Zijlstraed075362006-12-06 20:35:24 -08001786}
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001787
1788static inline void xs_reclassify_socket(int family, struct socket *sock)
1789{
Chuck Lever4232e862010-10-20 11:52:51 -04001790 switch (family) {
Chuck Lever176e21e2011-05-09 15:22:44 -04001791 case AF_LOCAL:
1792 xs_reclassify_socketu(sock);
1793 break;
Chuck Lever4232e862010-10-20 11:52:51 -04001794 case AF_INET:
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001795 xs_reclassify_socket4(sock);
Chuck Lever4232e862010-10-20 11:52:51 -04001796 break;
1797 case AF_INET6:
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001798 xs_reclassify_socket6(sock);
Chuck Lever4232e862010-10-20 11:52:51 -04001799 break;
1800 }
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001801}
Peter Zijlstraed075362006-12-06 20:35:24 -08001802#else
Chuck Lever176e21e2011-05-09 15:22:44 -04001803static inline void xs_reclassify_socketu(struct socket *sock)
1804{
1805}
1806
Chuck Lever8945ee52007-08-06 11:58:04 -04001807static inline void xs_reclassify_socket4(struct socket *sock)
1808{
1809}
1810
1811static inline void xs_reclassify_socket6(struct socket *sock)
Peter Zijlstraed075362006-12-06 20:35:24 -08001812{
1813}
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001814
1815static inline void xs_reclassify_socket(int family, struct socket *sock)
1816{
1817}
Peter Zijlstraed075362006-12-06 20:35:24 -08001818#endif
1819
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001820static struct socket *xs_create_sock(struct rpc_xprt *xprt,
1821 struct sock_xprt *transport, int family, int type, int protocol)
Pavel Emelyanov22f79322010-10-04 16:54:26 +04001822{
1823 struct socket *sock;
1824 int err;
1825
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001826 err = __sock_create(xprt->xprt_net, family, type, protocol, &sock, 1);
Pavel Emelyanov22f79322010-10-04 16:54:26 +04001827 if (err < 0) {
1828 dprintk("RPC: can't create %d transport socket (%d).\n",
1829 protocol, -err);
1830 goto out;
1831 }
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001832 xs_reclassify_socket(family, sock);
Pavel Emelyanov22f79322010-10-04 16:54:26 +04001833
Ben Hutchings4cea2882011-02-22 21:54:34 +00001834 err = xs_bind(transport, sock);
1835 if (err) {
Pavel Emelyanov22f79322010-10-04 16:54:26 +04001836 sock_release(sock);
1837 goto out;
1838 }
1839
1840 return sock;
1841out:
1842 return ERR_PTR(err);
1843}
1844
Chuck Lever176e21e2011-05-09 15:22:44 -04001845static int xs_local_finish_connecting(struct rpc_xprt *xprt,
1846 struct socket *sock)
1847{
1848 struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
1849 xprt);
1850
1851 if (!transport->inet) {
1852 struct sock *sk = sock->sk;
1853
1854 write_lock_bh(&sk->sk_callback_lock);
1855
1856 xs_save_old_callbacks(transport, sk);
1857
1858 sk->sk_user_data = xprt;
1859 sk->sk_data_ready = xs_local_data_ready;
1860 sk->sk_write_space = xs_udp_write_space;
1861 sk->sk_error_report = xs_error_report;
1862 sk->sk_allocation = GFP_ATOMIC;
1863
1864 xprt_clear_connected(xprt);
1865
1866 /* Reset to new socket */
1867 transport->sock = sock;
1868 transport->inet = sk;
1869
1870 write_unlock_bh(&sk->sk_callback_lock);
1871 }
1872
1873 /* Tell the socket layer to start connecting... */
1874 xprt->stat.connect_count++;
1875 xprt->stat.connect_start = jiffies;
1876 return kernel_connect(sock, xs_addr(xprt), xprt->addrlen, 0);
1877}
1878
1879/**
1880 * xs_local_setup_socket - create AF_LOCAL socket, connect to a local endpoint
1881 * @xprt: RPC transport to connect
1882 * @transport: socket transport to connect
1883 * @create_sock: function to create a socket of the correct type
1884 *
1885 * Invoked by a work queue tasklet.
1886 */
1887static void xs_local_setup_socket(struct work_struct *work)
1888{
1889 struct sock_xprt *transport =
1890 container_of(work, struct sock_xprt, connect_worker.work);
1891 struct rpc_xprt *xprt = &transport->xprt;
1892 struct socket *sock;
1893 int status = -EIO;
1894
1895 if (xprt->shutdown)
1896 goto out;
1897
1898 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1899 status = __sock_create(xprt->xprt_net, AF_LOCAL,
1900 SOCK_STREAM, 0, &sock, 1);
1901 if (status < 0) {
1902 dprintk("RPC: can't create AF_LOCAL "
1903 "transport socket (%d).\n", -status);
1904 goto out;
1905 }
1906 xs_reclassify_socketu(sock);
1907
1908 dprintk("RPC: worker connecting xprt %p via AF_LOCAL to %s\n",
1909 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1910
1911 status = xs_local_finish_connecting(xprt, sock);
1912 switch (status) {
1913 case 0:
1914 dprintk("RPC: xprt %p connected to %s\n",
1915 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1916 xprt_set_connected(xprt);
1917 break;
1918 case -ENOENT:
1919 dprintk("RPC: xprt %p: socket %s does not exist\n",
1920 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1921 break;
1922 default:
1923 printk(KERN_ERR "%s: unhandled error (%d) connecting to %s\n",
1924 __func__, -status,
1925 xprt->address_strings[RPC_DISPLAY_ADDR]);
1926 }
1927
1928out:
1929 xprt_clear_connecting(xprt);
1930 xprt_wake_pending_tasks(xprt, status);
1931}
1932
Chuck Lever16be2d22007-08-06 11:57:38 -04001933static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
Chuck Levera246b012005-08-11 16:25:23 -04001934{
Chuck Lever16be2d22007-08-06 11:57:38 -04001935 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Leveredb267a2006-08-22 20:06:18 -04001936
Chuck Leveree0ac0c2006-12-05 16:35:15 -05001937 if (!transport->inet) {
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001938 struct sock *sk = sock->sk;
1939
1940 write_lock_bh(&sk->sk_callback_lock);
1941
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04001942 xs_save_old_callbacks(transport, sk);
1943
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001944 sk->sk_user_data = xprt;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001945 sk->sk_data_ready = xs_udp_data_ready;
1946 sk->sk_write_space = xs_udp_write_space;
Trond Myklebust482f32e2009-03-11 14:38:00 -04001947 sk->sk_error_report = xs_error_report;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001948 sk->sk_no_check = UDP_CSUM_NORCV;
Trond Myklebustb079fa72005-12-13 16:13:52 -05001949 sk->sk_allocation = GFP_ATOMIC;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001950
1951 xprt_set_connected(xprt);
1952
1953 /* Reset to new socket */
Chuck Leveree0ac0c2006-12-05 16:35:15 -05001954 transport->sock = sock;
1955 transport->inet = sk;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001956
1957 write_unlock_bh(&sk->sk_callback_lock);
1958 }
Chuck Lever470056c2005-08-25 16:25:56 -07001959 xs_udp_do_set_buffer_size(xprt);
Chuck Lever16be2d22007-08-06 11:57:38 -04001960}
1961
Pavel Emelyanov8c14ff22010-10-04 16:58:02 +04001962static void xs_udp_setup_socket(struct work_struct *work)
Chuck Levera246b012005-08-11 16:25:23 -04001963{
1964 struct sock_xprt *transport =
1965 container_of(work, struct sock_xprt, connect_worker.work);
1966 struct rpc_xprt *xprt = &transport->xprt;
1967 struct socket *sock = transport->sock;
Pavel Emelyanovb65c0312010-10-04 16:53:46 +04001968 int status = -EIO;
Chuck Levera246b012005-08-11 16:25:23 -04001969
Trond Myklebust01d37c42009-03-11 14:09:39 -04001970 if (xprt->shutdown)
Chuck Lever9903cd12005-08-11 16:25:26 -04001971 goto out;
1972
Chuck Levera246b012005-08-11 16:25:23 -04001973 /* Start by resetting any existing state */
Chuck Leverfe315e72009-03-11 14:10:21 -04001974 xs_reset_transport(transport);
Pavel Emelyanov8c14ff22010-10-04 16:58:02 +04001975 sock = xs_create_sock(xprt, transport,
1976 xs_addr(xprt)->sa_family, SOCK_DGRAM, IPPROTO_UDP);
Pavel Emelyanovb65c0312010-10-04 16:53:46 +04001977 if (IS_ERR(sock))
Chuck Levera246b012005-08-11 16:25:23 -04001978 goto out;
Chuck Lever68e220b2007-08-06 11:57:48 -04001979
Chuck Leverc740eff2009-08-09 15:09:46 -04001980 dprintk("RPC: worker connecting xprt %p via %s to "
1981 "%s (port %s)\n", xprt,
1982 xprt->address_strings[RPC_DISPLAY_PROTO],
1983 xprt->address_strings[RPC_DISPLAY_ADDR],
1984 xprt->address_strings[RPC_DISPLAY_PORT]);
Chuck Lever68e220b2007-08-06 11:57:48 -04001985
1986 xs_udp_finish_connecting(xprt, sock);
Chuck Levera246b012005-08-11 16:25:23 -04001987 status = 0;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001988out:
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001989 xprt_clear_connecting(xprt);
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001990 xprt_wake_pending_tasks(xprt, status);
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001991}
1992
Chuck Lever3167e122005-08-25 16:25:55 -07001993/*
1994 * We need to preserve the port number so the reply cache on the server can
1995 * find our cached RPC replies when we get around to reconnecting.
1996 */
Pavel Emelyanov58dddac2010-10-04 16:52:55 +04001997static void xs_abort_connection(struct sock_xprt *transport)
Chuck Lever3167e122005-08-25 16:25:55 -07001998{
1999 int result;
Chuck Lever3167e122005-08-25 16:25:55 -07002000 struct sockaddr any;
2001
Pavel Emelyanov58dddac2010-10-04 16:52:55 +04002002 dprintk("RPC: disconnecting xprt %p to reuse port\n", transport);
Chuck Lever3167e122005-08-25 16:25:55 -07002003
2004 /*
2005 * Disconnect the transport socket by doing a connect operation
2006 * with AF_UNSPEC. This should return immediately...
2007 */
2008 memset(&any, 0, sizeof(any));
2009 any.sa_family = AF_UNSPEC;
Chuck Leveree0ac0c2006-12-05 16:35:15 -05002010 result = kernel_connect(transport->sock, &any, sizeof(any), 0);
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002011 if (!result)
Pavel Emelyanov58dddac2010-10-04 16:52:55 +04002012 xs_sock_mark_closed(&transport->xprt);
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002013 else
Chuck Lever46121cf2007-01-31 12:14:08 -05002014 dprintk("RPC: AF_UNSPEC connect return code %d\n",
Chuck Lever3167e122005-08-25 16:25:55 -07002015 result);
2016}
2017
Pavel Emelyanov58dddac2010-10-04 16:52:55 +04002018static void xs_tcp_reuse_connection(struct sock_xprt *transport)
Trond Myklebust40d25492009-03-11 14:37:58 -04002019{
2020 unsigned int state = transport->inet->sk_state;
2021
Andy Chittenden669502f2010-08-10 10:19:53 -04002022 if (state == TCP_CLOSE && transport->sock->state == SS_UNCONNECTED) {
2023 /* we don't need to abort the connection if the socket
2024 * hasn't undergone a shutdown
2025 */
2026 if (transport->inet->sk_shutdown == 0)
2027 return;
2028 dprintk("RPC: %s: TCP_CLOSEd and sk_shutdown set to %d\n",
2029 __func__, transport->inet->sk_shutdown);
2030 }
2031 if ((1 << state) & (TCPF_ESTABLISHED|TCPF_SYN_SENT)) {
2032 /* we don't need to abort the connection if the socket
2033 * hasn't undergone a shutdown
2034 */
2035 if (transport->inet->sk_shutdown == 0)
2036 return;
2037 dprintk("RPC: %s: ESTABLISHED/SYN_SENT "
2038 "sk_shutdown set to %d\n",
2039 __func__, transport->inet->sk_shutdown);
2040 }
Pavel Emelyanov58dddac2010-10-04 16:52:55 +04002041 xs_abort_connection(transport);
Trond Myklebust40d25492009-03-11 14:37:58 -04002042}
2043
Chuck Lever16be2d22007-08-06 11:57:38 -04002044static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002045{
Chuck Lever16be2d22007-08-06 11:57:38 -04002046 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Trond Myklebustfe19a962011-03-18 20:21:23 -04002047 int ret = -ENOTCONN;
Chuck Leveredb267a2006-08-22 20:06:18 -04002048
Chuck Leveree0ac0c2006-12-05 16:35:15 -05002049 if (!transport->inet) {
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002050 struct sock *sk = sock->sk;
2051
2052 write_lock_bh(&sk->sk_callback_lock);
2053
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04002054 xs_save_old_callbacks(transport, sk);
2055
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002056 sk->sk_user_data = xprt;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002057 sk->sk_data_ready = xs_tcp_data_ready;
2058 sk->sk_state_change = xs_tcp_state_change;
2059 sk->sk_write_space = xs_tcp_write_space;
Trond Myklebust482f32e2009-03-11 14:38:00 -04002060 sk->sk_error_report = xs_error_report;
Trond Myklebustb079fa72005-12-13 16:13:52 -05002061 sk->sk_allocation = GFP_ATOMIC;
Chuck Lever3167e122005-08-25 16:25:55 -07002062
2063 /* socket options */
2064 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
2065 sock_reset_flag(sk, SOCK_LINGER);
2066 tcp_sk(sk)->linger2 = 0;
2067 tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002068
2069 xprt_clear_connected(xprt);
2070
2071 /* Reset to new socket */
Chuck Leveree0ac0c2006-12-05 16:35:15 -05002072 transport->sock = sock;
2073 transport->inet = sk;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002074
2075 write_unlock_bh(&sk->sk_callback_lock);
2076 }
2077
Trond Myklebust01d37c42009-03-11 14:09:39 -04002078 if (!xprt_bound(xprt))
Trond Myklebustfe19a962011-03-18 20:21:23 -04002079 goto out;
Trond Myklebust01d37c42009-03-11 14:09:39 -04002080
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002081 /* Tell the socket layer to start connecting... */
Chuck Lever262ca072006-03-20 13:44:16 -05002082 xprt->stat.connect_count++;
2083 xprt->stat.connect_start = jiffies;
Trond Myklebustfe19a962011-03-18 20:21:23 -04002084 ret = kernel_connect(sock, xs_addr(xprt), xprt->addrlen, O_NONBLOCK);
2085 switch (ret) {
2086 case 0:
2087 case -EINPROGRESS:
2088 /* SYN_SENT! */
2089 xprt->connect_cookie++;
2090 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
2091 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2092 }
2093out:
2094 return ret;
Chuck Lever16be2d22007-08-06 11:57:38 -04002095}
2096
2097/**
Trond Myklebustb61d59f2009-03-11 14:38:04 -04002098 * xs_tcp_setup_socket - create a TCP socket and connect to a remote endpoint
2099 * @xprt: RPC transport to connect
2100 * @transport: socket transport to connect
2101 * @create_sock: function to create a socket of the correct type
Chuck Lever16be2d22007-08-06 11:57:38 -04002102 *
2103 * Invoked by a work queue tasklet.
2104 */
Pavel Emelyanovcdd518d2010-10-04 16:57:40 +04002105static void xs_tcp_setup_socket(struct work_struct *work)
Chuck Lever16be2d22007-08-06 11:57:38 -04002106{
Pavel Emelyanovcdd518d2010-10-04 16:57:40 +04002107 struct sock_xprt *transport =
2108 container_of(work, struct sock_xprt, connect_worker.work);
Chuck Lever16be2d22007-08-06 11:57:38 -04002109 struct socket *sock = transport->sock;
Pavel Emelyanova9f5f0f2010-10-04 16:52:25 +04002110 struct rpc_xprt *xprt = &transport->xprt;
Trond Myklebustb61d59f2009-03-11 14:38:04 -04002111 int status = -EIO;
Chuck Lever16be2d22007-08-06 11:57:38 -04002112
Trond Myklebust01d37c42009-03-11 14:09:39 -04002113 if (xprt->shutdown)
Chuck Lever16be2d22007-08-06 11:57:38 -04002114 goto out;
2115
2116 if (!sock) {
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002117 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
Pavel Emelyanovcdd518d2010-10-04 16:57:40 +04002118 sock = xs_create_sock(xprt, transport,
2119 xs_addr(xprt)->sa_family, SOCK_STREAM, IPPROTO_TCP);
Trond Myklebustb61d59f2009-03-11 14:38:04 -04002120 if (IS_ERR(sock)) {
2121 status = PTR_ERR(sock);
Chuck Lever16be2d22007-08-06 11:57:38 -04002122 goto out;
2123 }
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002124 } else {
2125 int abort_and_exit;
2126
2127 abort_and_exit = test_and_clear_bit(XPRT_CONNECTION_ABORT,
2128 &xprt->state);
Chuck Lever16be2d22007-08-06 11:57:38 -04002129 /* "close" the socket, preserving the local port */
Pavel Emelyanov58dddac2010-10-04 16:52:55 +04002130 xs_tcp_reuse_connection(transport);
Chuck Lever16be2d22007-08-06 11:57:38 -04002131
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002132 if (abort_and_exit)
2133 goto out_eagain;
2134 }
2135
Chuck Leverc740eff2009-08-09 15:09:46 -04002136 dprintk("RPC: worker connecting xprt %p via %s to "
2137 "%s (port %s)\n", xprt,
2138 xprt->address_strings[RPC_DISPLAY_PROTO],
2139 xprt->address_strings[RPC_DISPLAY_ADDR],
2140 xprt->address_strings[RPC_DISPLAY_PORT]);
Chuck Lever16be2d22007-08-06 11:57:38 -04002141
2142 status = xs_tcp_finish_connecting(xprt, sock);
Chuck Lever46121cf2007-01-31 12:14:08 -05002143 dprintk("RPC: %p connect status %d connected %d sock state %d\n",
2144 xprt, -status, xprt_connected(xprt),
2145 sock->sk->sk_state);
Trond Myklebust2a491992009-03-11 14:38:00 -04002146 switch (status) {
Trond Myklebustf75e6742009-04-21 17:18:20 -04002147 default:
2148 printk("%s: connect returned unhandled error %d\n",
2149 __func__, status);
2150 case -EADDRNOTAVAIL:
2151 /* We're probably in TIME_WAIT. Get rid of existing socket,
2152 * and retry
2153 */
2154 set_bit(XPRT_CONNECTION_CLOSE, &xprt->state);
2155 xprt_force_disconnect(xprt);
Trond Myklebust88b5ed72009-06-17 13:22:57 -07002156 break;
Trond Myklebust8a2cec22009-03-11 14:38:01 -04002157 case -ECONNREFUSED:
2158 case -ECONNRESET:
2159 case -ENETUNREACH:
2160 /* retry with existing socket, after a delay */
Trond Myklebust2a491992009-03-11 14:38:00 -04002161 case 0:
2162 case -EINPROGRESS:
2163 case -EALREADY:
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002164 xprt_clear_connecting(xprt);
2165 return;
Trond Myklebust9fcfe0c2010-03-02 13:06:21 -05002166 case -EINVAL:
2167 /* Happens, for instance, if the user specified a link
2168 * local IPv6 address without a scope-id.
2169 */
2170 goto out;
Chuck Levera246b012005-08-11 16:25:23 -04002171 }
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002172out_eagain:
Trond Myklebust2a491992009-03-11 14:38:00 -04002173 status = -EAGAIN;
Chuck Levera246b012005-08-11 16:25:23 -04002174out:
Chuck Lever2226feb2005-08-11 16:25:38 -04002175 xprt_clear_connecting(xprt);
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002176 xprt_wake_pending_tasks(xprt, status);
Chuck Levera246b012005-08-11 16:25:23 -04002177}
2178
Chuck Lever68e220b2007-08-06 11:57:48 -04002179/**
Chuck Lever9903cd12005-08-11 16:25:26 -04002180 * xs_connect - connect a socket to a remote endpoint
2181 * @task: address of RPC task that manages state of connect request
2182 *
2183 * TCP: If the remote end dropped the connection, delay reconnecting.
Chuck Lever03bf4b72005-08-25 16:25:55 -07002184 *
2185 * UDP socket connects are synchronous, but we use a work queue anyway
2186 * to guarantee that even unprivileged user processes can set up a
2187 * socket on a privileged port.
2188 *
2189 * If a UDP socket connect fails, the delay behavior here prevents
2190 * retry floods (hard mounts).
Chuck Lever9903cd12005-08-11 16:25:26 -04002191 */
2192static void xs_connect(struct rpc_task *task)
Chuck Levera246b012005-08-11 16:25:23 -04002193{
2194 struct rpc_xprt *xprt = task->tk_xprt;
Chuck Leveree0ac0c2006-12-05 16:35:15 -05002195 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -04002196
Chuck Lever09a21c42009-12-03 15:58:56 -05002197 if (transport->sock != NULL && !RPC_IS_SOFTCONN(task)) {
Chuck Lever46121cf2007-01-31 12:14:08 -05002198 dprintk("RPC: xs_connect delayed xprt %p for %lu "
2199 "seconds\n",
Chuck Lever03bf4b72005-08-25 16:25:55 -07002200 xprt, xprt->reestablish_timeout / HZ);
Trond Myklebustc1384c92007-06-14 18:00:42 -04002201 queue_delayed_work(rpciod_workqueue,
2202 &transport->connect_worker,
2203 xprt->reestablish_timeout);
Chuck Lever03bf4b72005-08-25 16:25:55 -07002204 xprt->reestablish_timeout <<= 1;
Neil Brown61d0a8e2009-09-23 14:36:37 -04002205 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
2206 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
Chuck Lever03bf4b72005-08-25 16:25:55 -07002207 if (xprt->reestablish_timeout > XS_TCP_MAX_REEST_TO)
2208 xprt->reestablish_timeout = XS_TCP_MAX_REEST_TO;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002209 } else {
Chuck Lever46121cf2007-01-31 12:14:08 -05002210 dprintk("RPC: xs_connect scheduled xprt %p\n", xprt);
Trond Myklebustc1384c92007-06-14 18:00:42 -04002211 queue_delayed_work(rpciod_workqueue,
2212 &transport->connect_worker, 0);
Chuck Levera246b012005-08-11 16:25:23 -04002213 }
2214}
2215
Chuck Lever262ca072006-03-20 13:44:16 -05002216/**
Chuck Lever176e21e2011-05-09 15:22:44 -04002217 * xs_local_print_stats - display AF_LOCAL socket-specifc stats
2218 * @xprt: rpc_xprt struct containing statistics
2219 * @seq: output file
2220 *
2221 */
2222static void xs_local_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2223{
2224 long idle_time = 0;
2225
2226 if (xprt_connected(xprt))
2227 idle_time = (long)(jiffies - xprt->last_used) / HZ;
2228
2229 seq_printf(seq, "\txprt:\tlocal %lu %lu %lu %ld %lu %lu %lu "
Andy Adamson15a45202012-02-14 16:19:18 -05002230 "%llu %llu %lu %llu %llu\n",
Chuck Lever176e21e2011-05-09 15:22:44 -04002231 xprt->stat.bind_count,
2232 xprt->stat.connect_count,
2233 xprt->stat.connect_time,
2234 idle_time,
2235 xprt->stat.sends,
2236 xprt->stat.recvs,
2237 xprt->stat.bad_xids,
2238 xprt->stat.req_u,
Andy Adamson15a45202012-02-14 16:19:18 -05002239 xprt->stat.bklog_u,
2240 xprt->stat.max_slots,
2241 xprt->stat.sending_u,
2242 xprt->stat.pending_u);
Chuck Lever176e21e2011-05-09 15:22:44 -04002243}
2244
2245/**
Chuck Lever262ca072006-03-20 13:44:16 -05002246 * xs_udp_print_stats - display UDP socket-specifc stats
2247 * @xprt: rpc_xprt struct containing statistics
2248 * @seq: output file
2249 *
2250 */
2251static void xs_udp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2252{
Chuck Leverc8475462006-12-05 16:35:26 -05002253 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2254
Andy Adamson15a45202012-02-14 16:19:18 -05002255 seq_printf(seq, "\txprt:\tudp %u %lu %lu %lu %lu %llu %llu "
2256 "%lu %llu %llu\n",
Chuck Leverfbfffbd2009-08-09 15:09:46 -04002257 transport->srcport,
Chuck Lever262ca072006-03-20 13:44:16 -05002258 xprt->stat.bind_count,
2259 xprt->stat.sends,
2260 xprt->stat.recvs,
2261 xprt->stat.bad_xids,
2262 xprt->stat.req_u,
Andy Adamson15a45202012-02-14 16:19:18 -05002263 xprt->stat.bklog_u,
2264 xprt->stat.max_slots,
2265 xprt->stat.sending_u,
2266 xprt->stat.pending_u);
Chuck Lever262ca072006-03-20 13:44:16 -05002267}
2268
2269/**
2270 * xs_tcp_print_stats - display TCP socket-specifc stats
2271 * @xprt: rpc_xprt struct containing statistics
2272 * @seq: output file
2273 *
2274 */
2275static void xs_tcp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2276{
Chuck Leverc8475462006-12-05 16:35:26 -05002277 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Lever262ca072006-03-20 13:44:16 -05002278 long idle_time = 0;
2279
2280 if (xprt_connected(xprt))
2281 idle_time = (long)(jiffies - xprt->last_used) / HZ;
2282
Andy Adamson15a45202012-02-14 16:19:18 -05002283 seq_printf(seq, "\txprt:\ttcp %u %lu %lu %lu %ld %lu %lu %lu "
2284 "%llu %llu %lu %llu %llu\n",
Chuck Leverfbfffbd2009-08-09 15:09:46 -04002285 transport->srcport,
Chuck Lever262ca072006-03-20 13:44:16 -05002286 xprt->stat.bind_count,
2287 xprt->stat.connect_count,
2288 xprt->stat.connect_time,
2289 idle_time,
2290 xprt->stat.sends,
2291 xprt->stat.recvs,
2292 xprt->stat.bad_xids,
2293 xprt->stat.req_u,
Andy Adamson15a45202012-02-14 16:19:18 -05002294 xprt->stat.bklog_u,
2295 xprt->stat.max_slots,
2296 xprt->stat.sending_u,
2297 xprt->stat.pending_u);
Chuck Lever262ca072006-03-20 13:44:16 -05002298}
2299
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002300/*
2301 * Allocate a bunch of pages for a scratch buffer for the rpc code. The reason
2302 * we allocate pages instead doing a kmalloc like rpc_malloc is because we want
2303 * to use the server side send routines.
2304 */
H Hartley Sweeten5a51f132010-01-14 15:38:31 -07002305static void *bc_malloc(struct rpc_task *task, size_t size)
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002306{
2307 struct page *page;
2308 struct rpc_buffer *buf;
2309
2310 BUG_ON(size > PAGE_SIZE - sizeof(struct rpc_buffer));
2311 page = alloc_page(GFP_KERNEL);
2312
2313 if (!page)
2314 return NULL;
2315
2316 buf = page_address(page);
2317 buf->len = PAGE_SIZE;
2318
2319 return buf->data;
2320}
2321
2322/*
2323 * Free the space allocated in the bc_alloc routine
2324 */
H Hartley Sweeten5a51f132010-01-14 15:38:31 -07002325static void bc_free(void *buffer)
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002326{
2327 struct rpc_buffer *buf;
2328
2329 if (!buffer)
2330 return;
2331
2332 buf = container_of(buffer, struct rpc_buffer, data);
2333 free_page((unsigned long)buf);
2334}
2335
2336/*
2337 * Use the svc_sock to send the callback. Must be called with svsk->sk_mutex
2338 * held. Borrows heavily from svc_tcp_sendto and xs_tcp_send_request.
2339 */
2340static int bc_sendto(struct rpc_rqst *req)
2341{
2342 int len;
2343 struct xdr_buf *xbufp = &req->rq_snd_buf;
2344 struct rpc_xprt *xprt = req->rq_xprt;
2345 struct sock_xprt *transport =
2346 container_of(xprt, struct sock_xprt, xprt);
2347 struct socket *sock = transport->sock;
2348 unsigned long headoff;
2349 unsigned long tailoff;
2350
Chuck Lever61677ee2011-05-09 15:22:34 -04002351 xs_encode_stream_record_marker(xbufp);
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002352
2353 tailoff = (unsigned long)xbufp->tail[0].iov_base & ~PAGE_MASK;
2354 headoff = (unsigned long)xbufp->head[0].iov_base & ~PAGE_MASK;
2355 len = svc_send_common(sock, xbufp,
2356 virt_to_page(xbufp->head[0].iov_base), headoff,
2357 xbufp->tail[0].iov_base, tailoff);
2358
2359 if (len != xbufp->len) {
2360 printk(KERN_NOTICE "Error sending entire callback!\n");
2361 len = -EAGAIN;
2362 }
2363
2364 return len;
2365}
2366
2367/*
2368 * The send routine. Borrows from svc_send
2369 */
2370static int bc_send_request(struct rpc_task *task)
2371{
2372 struct rpc_rqst *req = task->tk_rqstp;
2373 struct svc_xprt *xprt;
2374 struct svc_sock *svsk;
2375 u32 len;
2376
2377 dprintk("sending request with xid: %08x\n", ntohl(req->rq_xid));
2378 /*
2379 * Get the server socket associated with this callback xprt
2380 */
2381 xprt = req->rq_xprt->bc_xprt;
2382 svsk = container_of(xprt, struct svc_sock, sk_xprt);
2383
2384 /*
2385 * Grab the mutex to serialize data as the connection is shared
2386 * with the fore channel
2387 */
2388 if (!mutex_trylock(&xprt->xpt_mutex)) {
2389 rpc_sleep_on(&xprt->xpt_bc_pending, task, NULL);
2390 if (!mutex_trylock(&xprt->xpt_mutex))
2391 return -EAGAIN;
2392 rpc_wake_up_queued_task(&xprt->xpt_bc_pending, task);
2393 }
2394 if (test_bit(XPT_DEAD, &xprt->xpt_flags))
2395 len = -ENOTCONN;
2396 else
2397 len = bc_sendto(req);
2398 mutex_unlock(&xprt->xpt_mutex);
2399
2400 if (len > 0)
2401 len = 0;
2402
2403 return len;
2404}
2405
2406/*
2407 * The close routine. Since this is client initiated, we do nothing
2408 */
2409
2410static void bc_close(struct rpc_xprt *xprt)
2411{
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002412}
2413
2414/*
2415 * The xprt destroy routine. Again, because this connection is client
2416 * initiated, we do nothing
2417 */
2418
2419static void bc_destroy(struct rpc_xprt *xprt)
2420{
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002421}
2422
Chuck Lever176e21e2011-05-09 15:22:44 -04002423static struct rpc_xprt_ops xs_local_ops = {
2424 .reserve_xprt = xprt_reserve_xprt,
2425 .release_xprt = xs_tcp_release_xprt,
2426 .rpcbind = xs_local_rpcbind,
2427 .set_port = xs_local_set_port,
2428 .connect = xs_connect,
2429 .buf_alloc = rpc_malloc,
2430 .buf_free = rpc_free,
2431 .send_request = xs_local_send_request,
2432 .set_retrans_timeout = xprt_set_retrans_timeout_def,
2433 .close = xs_close,
2434 .destroy = xs_destroy,
2435 .print_stats = xs_local_print_stats,
2436};
2437
Chuck Lever262965f2005-08-11 16:25:56 -04002438static struct rpc_xprt_ops xs_udp_ops = {
Chuck Lever43118c22005-08-25 16:25:49 -07002439 .set_buffer_size = xs_udp_set_buffer_size,
Chuck Lever12a80462005-08-25 16:25:51 -07002440 .reserve_xprt = xprt_reserve_xprt_cong,
Chuck Lever49e9a892005-08-25 16:25:51 -07002441 .release_xprt = xprt_release_xprt_cong,
Chuck Lever45160d62007-07-01 12:13:17 -04002442 .rpcbind = rpcb_getport_async,
Chuck Lever92200412006-01-03 09:55:51 +01002443 .set_port = xs_set_port,
Chuck Lever9903cd12005-08-11 16:25:26 -04002444 .connect = xs_connect,
Chuck Lever02107142006-01-03 09:55:49 +01002445 .buf_alloc = rpc_malloc,
2446 .buf_free = rpc_free,
Chuck Lever262965f2005-08-11 16:25:56 -04002447 .send_request = xs_udp_send_request,
Chuck Leverfe3aca22005-08-25 16:25:50 -07002448 .set_retrans_timeout = xprt_set_retrans_timeout_rtt,
Chuck Lever46c0ee82005-08-25 16:25:52 -07002449 .timer = xs_udp_timer,
Chuck Levera58dd392005-08-25 16:25:53 -07002450 .release_request = xprt_release_rqst_cong,
Chuck Lever262965f2005-08-11 16:25:56 -04002451 .close = xs_close,
2452 .destroy = xs_destroy,
Chuck Lever262ca072006-03-20 13:44:16 -05002453 .print_stats = xs_udp_print_stats,
Chuck Lever262965f2005-08-11 16:25:56 -04002454};
2455
2456static struct rpc_xprt_ops xs_tcp_ops = {
Chuck Lever12a80462005-08-25 16:25:51 -07002457 .reserve_xprt = xprt_reserve_xprt,
Trond Myklebuste0ab53d2006-07-27 17:22:50 -04002458 .release_xprt = xs_tcp_release_xprt,
Chuck Lever45160d62007-07-01 12:13:17 -04002459 .rpcbind = rpcb_getport_async,
Chuck Lever92200412006-01-03 09:55:51 +01002460 .set_port = xs_set_port,
Trond Myklebust0b9e7942010-04-16 16:41:57 -04002461 .connect = xs_connect,
Chuck Lever02107142006-01-03 09:55:49 +01002462 .buf_alloc = rpc_malloc,
2463 .buf_free = rpc_free,
Chuck Lever262965f2005-08-11 16:25:56 -04002464 .send_request = xs_tcp_send_request,
Chuck Leverfe3aca22005-08-25 16:25:50 -07002465 .set_retrans_timeout = xprt_set_retrans_timeout_def,
Trond Myklebustf75e6742009-04-21 17:18:20 -04002466 .close = xs_tcp_close,
Chuck Lever9903cd12005-08-11 16:25:26 -04002467 .destroy = xs_destroy,
Chuck Lever262ca072006-03-20 13:44:16 -05002468 .print_stats = xs_tcp_print_stats,
Chuck Levera246b012005-08-11 16:25:23 -04002469};
2470
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002471/*
2472 * The rpc_xprt_ops for the server backchannel
2473 */
2474
2475static struct rpc_xprt_ops bc_tcp_ops = {
2476 .reserve_xprt = xprt_reserve_xprt,
2477 .release_xprt = xprt_release_xprt,
J. Bruce Fields92769102012-03-23 15:25:03 -04002478 .rpcbind = xs_local_rpcbind,
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002479 .buf_alloc = bc_malloc,
2480 .buf_free = bc_free,
2481 .send_request = bc_send_request,
2482 .set_retrans_timeout = xprt_set_retrans_timeout_def,
2483 .close = bc_close,
2484 .destroy = bc_destroy,
2485 .print_stats = xs_tcp_print_stats,
2486};
2487
Chuck Lever92476852010-10-20 11:53:01 -04002488static int xs_init_anyaddr(const int family, struct sockaddr *sap)
2489{
2490 static const struct sockaddr_in sin = {
2491 .sin_family = AF_INET,
2492 .sin_addr.s_addr = htonl(INADDR_ANY),
2493 };
2494 static const struct sockaddr_in6 sin6 = {
2495 .sin6_family = AF_INET6,
2496 .sin6_addr = IN6ADDR_ANY_INIT,
2497 };
2498
2499 switch (family) {
Chuck Lever176e21e2011-05-09 15:22:44 -04002500 case AF_LOCAL:
2501 break;
Chuck Lever92476852010-10-20 11:53:01 -04002502 case AF_INET:
2503 memcpy(sap, &sin, sizeof(sin));
2504 break;
2505 case AF_INET6:
2506 memcpy(sap, &sin6, sizeof(sin6));
2507 break;
2508 default:
2509 dprintk("RPC: %s: Bad address family\n", __func__);
2510 return -EAFNOSUPPORT;
2511 }
2512 return 0;
2513}
2514
\"Talpey, Thomas\3c341b02007-09-10 13:47:07 -04002515static struct rpc_xprt *xs_setup_xprt(struct xprt_create *args,
Trond Myklebustd9ba1312011-07-17 18:11:30 -04002516 unsigned int slot_table_size,
2517 unsigned int max_slot_table_size)
Chuck Leverc8541ec2006-10-17 14:44:27 -04002518{
2519 struct rpc_xprt *xprt;
Chuck Leverffc2e512006-12-05 16:35:11 -05002520 struct sock_xprt *new;
Chuck Leverc8541ec2006-10-17 14:44:27 -04002521
Frank van Maarseveen96802a02007-07-08 13:08:54 +02002522 if (args->addrlen > sizeof(xprt->addr)) {
Chuck Lever46121cf2007-01-31 12:14:08 -05002523 dprintk("RPC: xs_setup_xprt: address too large\n");
Chuck Leverc8541ec2006-10-17 14:44:27 -04002524 return ERR_PTR(-EBADF);
2525 }
2526
Trond Myklebustd9ba1312011-07-17 18:11:30 -04002527 xprt = xprt_alloc(args->net, sizeof(*new), slot_table_size,
2528 max_slot_table_size);
Pavel Emelyanovbd1722d2010-09-29 16:02:43 +04002529 if (xprt == NULL) {
Chuck Lever46121cf2007-01-31 12:14:08 -05002530 dprintk("RPC: xs_setup_xprt: couldn't allocate "
2531 "rpc_xprt\n");
Chuck Leverc8541ec2006-10-17 14:44:27 -04002532 return ERR_PTR(-ENOMEM);
2533 }
2534
Pavel Emelyanovbd1722d2010-09-29 16:02:43 +04002535 new = container_of(xprt, struct sock_xprt, xprt);
Frank van Maarseveen96802a02007-07-08 13:08:54 +02002536 memcpy(&xprt->addr, args->dstaddr, args->addrlen);
2537 xprt->addrlen = args->addrlen;
Frank van Maarseveend3bc9a12007-07-09 22:23:35 +02002538 if (args->srcaddr)
Chuck Leverfbfffbd2009-08-09 15:09:46 -04002539 memcpy(&new->srcaddr, args->srcaddr, args->addrlen);
Chuck Lever92476852010-10-20 11:53:01 -04002540 else {
2541 int err;
2542 err = xs_init_anyaddr(args->dstaddr->sa_family,
2543 (struct sockaddr *)&new->srcaddr);
Stanislav Kinsbursky2aa13532011-11-10 14:33:23 +03002544 if (err != 0) {
2545 xprt_free(xprt);
Chuck Lever92476852010-10-20 11:53:01 -04002546 return ERR_PTR(err);
Stanislav Kinsbursky2aa13532011-11-10 14:33:23 +03002547 }
Chuck Lever92476852010-10-20 11:53:01 -04002548 }
Chuck Leverc8541ec2006-10-17 14:44:27 -04002549
2550 return xprt;
2551}
2552
Chuck Lever176e21e2011-05-09 15:22:44 -04002553static const struct rpc_timeout xs_local_default_timeout = {
2554 .to_initval = 10 * HZ,
2555 .to_maxval = 10 * HZ,
2556 .to_retries = 2,
2557};
2558
2559/**
2560 * xs_setup_local - Set up transport to use an AF_LOCAL socket
2561 * @args: rpc transport creation arguments
2562 *
2563 * AF_LOCAL is a "tpi_cots_ord" transport, just like TCP
2564 */
2565static struct rpc_xprt *xs_setup_local(struct xprt_create *args)
2566{
2567 struct sockaddr_un *sun = (struct sockaddr_un *)args->dstaddr;
2568 struct sock_xprt *transport;
2569 struct rpc_xprt *xprt;
2570 struct rpc_xprt *ret;
2571
Trond Myklebustd9ba1312011-07-17 18:11:30 -04002572 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
2573 xprt_max_tcp_slot_table_entries);
Chuck Lever176e21e2011-05-09 15:22:44 -04002574 if (IS_ERR(xprt))
2575 return xprt;
2576 transport = container_of(xprt, struct sock_xprt, xprt);
2577
2578 xprt->prot = 0;
2579 xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
2580 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
2581
2582 xprt->bind_timeout = XS_BIND_TO;
2583 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2584 xprt->idle_timeout = XS_IDLE_DISC_TO;
2585
2586 xprt->ops = &xs_local_ops;
2587 xprt->timeout = &xs_local_default_timeout;
2588
2589 switch (sun->sun_family) {
2590 case AF_LOCAL:
2591 if (sun->sun_path[0] != '/') {
2592 dprintk("RPC: bad AF_LOCAL address: %s\n",
2593 sun->sun_path);
2594 ret = ERR_PTR(-EINVAL);
2595 goto out_err;
2596 }
2597 xprt_set_bound(xprt);
2598 INIT_DELAYED_WORK(&transport->connect_worker,
2599 xs_local_setup_socket);
2600 xs_format_peer_addresses(xprt, "local", RPCBIND_NETID_LOCAL);
2601 break;
2602 default:
2603 ret = ERR_PTR(-EAFNOSUPPORT);
2604 goto out_err;
2605 }
2606
2607 dprintk("RPC: set up xprt to %s via AF_LOCAL\n",
2608 xprt->address_strings[RPC_DISPLAY_ADDR]);
2609
2610 if (try_module_get(THIS_MODULE))
2611 return xprt;
2612 ret = ERR_PTR(-EINVAL);
2613out_err:
2614 xprt_free(xprt);
2615 return ret;
2616}
2617
Trond Myklebust2881ae72007-12-20 16:03:54 -05002618static const struct rpc_timeout xs_udp_default_timeout = {
2619 .to_initval = 5 * HZ,
2620 .to_maxval = 30 * HZ,
2621 .to_increment = 5 * HZ,
2622 .to_retries = 5,
2623};
2624
Chuck Lever9903cd12005-08-11 16:25:26 -04002625/**
2626 * xs_setup_udp - Set up transport to use a UDP socket
Frank van Maarseveen96802a02007-07-08 13:08:54 +02002627 * @args: rpc transport creation arguments
Chuck Lever9903cd12005-08-11 16:25:26 -04002628 *
2629 */
Adrian Bunk483066d2007-10-24 18:24:02 +02002630static struct rpc_xprt *xs_setup_udp(struct xprt_create *args)
Chuck Levera246b012005-08-11 16:25:23 -04002631{
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002632 struct sockaddr *addr = args->dstaddr;
Chuck Leverc8541ec2006-10-17 14:44:27 -04002633 struct rpc_xprt *xprt;
Chuck Leverc8475462006-12-05 16:35:26 -05002634 struct sock_xprt *transport;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002635 struct rpc_xprt *ret;
Chuck Levera246b012005-08-11 16:25:23 -04002636
Trond Myklebustd9ba1312011-07-17 18:11:30 -04002637 xprt = xs_setup_xprt(args, xprt_udp_slot_table_entries,
2638 xprt_udp_slot_table_entries);
Chuck Leverc8541ec2006-10-17 14:44:27 -04002639 if (IS_ERR(xprt))
2640 return xprt;
Chuck Leverc8475462006-12-05 16:35:26 -05002641 transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -04002642
Chuck Leverec739ef2006-08-22 20:06:15 -04002643 xprt->prot = IPPROTO_UDP;
Chuck Lever808012f2005-08-25 16:25:49 -07002644 xprt->tsh_size = 0;
Chuck Levera246b012005-08-11 16:25:23 -04002645 /* XXX: header size can vary due to auth type, IPv6, etc. */
2646 xprt->max_payload = (1U << 16) - (MAX_HEADER << 3);
2647
Chuck Lever03bf4b72005-08-25 16:25:55 -07002648 xprt->bind_timeout = XS_BIND_TO;
Chuck Lever03bf4b72005-08-25 16:25:55 -07002649 xprt->reestablish_timeout = XS_UDP_REEST_TO;
2650 xprt->idle_timeout = XS_IDLE_DISC_TO;
Chuck Levera246b012005-08-11 16:25:23 -04002651
Chuck Lever262965f2005-08-11 16:25:56 -04002652 xprt->ops = &xs_udp_ops;
Chuck Levera246b012005-08-11 16:25:23 -04002653
Trond Myklebustba7392b2007-12-20 16:03:55 -05002654 xprt->timeout = &xs_udp_default_timeout;
Chuck Levera246b012005-08-11 16:25:23 -04002655
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002656 switch (addr->sa_family) {
2657 case AF_INET:
2658 if (((struct sockaddr_in *)addr)->sin_port != htons(0))
2659 xprt_set_bound(xprt);
2660
2661 INIT_DELAYED_WORK(&transport->connect_worker,
Pavel Emelyanov8c14ff22010-10-04 16:58:02 +04002662 xs_udp_setup_socket);
Chuck Lever9dc3b092009-08-09 15:09:46 -04002663 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP);
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002664 break;
2665 case AF_INET6:
2666 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
2667 xprt_set_bound(xprt);
2668
2669 INIT_DELAYED_WORK(&transport->connect_worker,
Pavel Emelyanov8c14ff22010-10-04 16:58:02 +04002670 xs_udp_setup_socket);
Chuck Lever9dc3b092009-08-09 15:09:46 -04002671 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP6);
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002672 break;
2673 default:
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002674 ret = ERR_PTR(-EAFNOSUPPORT);
2675 goto out_err;
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002676 }
2677
Chuck Leverc740eff2009-08-09 15:09:46 -04002678 if (xprt_bound(xprt))
2679 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
2680 xprt->address_strings[RPC_DISPLAY_ADDR],
2681 xprt->address_strings[RPC_DISPLAY_PORT],
2682 xprt->address_strings[RPC_DISPLAY_PROTO]);
2683 else
2684 dprintk("RPC: set up xprt to %s (autobind) via %s\n",
2685 xprt->address_strings[RPC_DISPLAY_ADDR],
2686 xprt->address_strings[RPC_DISPLAY_PROTO]);
Chuck Leveredb267a2006-08-22 20:06:18 -04002687
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002688 if (try_module_get(THIS_MODULE))
2689 return xprt;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002690 ret = ERR_PTR(-EINVAL);
2691out_err:
Pavel Emelyanove204e622010-09-29 16:03:13 +04002692 xprt_free(xprt);
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002693 return ret;
Chuck Levera246b012005-08-11 16:25:23 -04002694}
2695
Trond Myklebust2881ae72007-12-20 16:03:54 -05002696static const struct rpc_timeout xs_tcp_default_timeout = {
2697 .to_initval = 60 * HZ,
2698 .to_maxval = 60 * HZ,
2699 .to_retries = 2,
2700};
2701
Chuck Lever9903cd12005-08-11 16:25:26 -04002702/**
2703 * xs_setup_tcp - Set up transport to use a TCP socket
Frank van Maarseveen96802a02007-07-08 13:08:54 +02002704 * @args: rpc transport creation arguments
Chuck Lever9903cd12005-08-11 16:25:26 -04002705 *
2706 */
Adrian Bunk483066d2007-10-24 18:24:02 +02002707static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args)
Chuck Levera246b012005-08-11 16:25:23 -04002708{
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002709 struct sockaddr *addr = args->dstaddr;
Chuck Leverc8541ec2006-10-17 14:44:27 -04002710 struct rpc_xprt *xprt;
Chuck Leverc8475462006-12-05 16:35:26 -05002711 struct sock_xprt *transport;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002712 struct rpc_xprt *ret;
Chuck Levera246b012005-08-11 16:25:23 -04002713
Trond Myklebustd9ba1312011-07-17 18:11:30 -04002714 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
2715 xprt_max_tcp_slot_table_entries);
Chuck Leverc8541ec2006-10-17 14:44:27 -04002716 if (IS_ERR(xprt))
2717 return xprt;
Chuck Leverc8475462006-12-05 16:35:26 -05002718 transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -04002719
Chuck Leverec739ef2006-08-22 20:06:15 -04002720 xprt->prot = IPPROTO_TCP;
Chuck Lever808012f2005-08-25 16:25:49 -07002721 xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
Chuck Lever808012f2005-08-25 16:25:49 -07002722 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
Chuck Levera246b012005-08-11 16:25:23 -04002723
Chuck Lever03bf4b72005-08-25 16:25:55 -07002724 xprt->bind_timeout = XS_BIND_TO;
Chuck Lever03bf4b72005-08-25 16:25:55 -07002725 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2726 xprt->idle_timeout = XS_IDLE_DISC_TO;
Chuck Levera246b012005-08-11 16:25:23 -04002727
Chuck Lever262965f2005-08-11 16:25:56 -04002728 xprt->ops = &xs_tcp_ops;
Trond Myklebustba7392b2007-12-20 16:03:55 -05002729 xprt->timeout = &xs_tcp_default_timeout;
Chuck Levera246b012005-08-11 16:25:23 -04002730
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002731 switch (addr->sa_family) {
2732 case AF_INET:
2733 if (((struct sockaddr_in *)addr)->sin_port != htons(0))
2734 xprt_set_bound(xprt);
2735
Chuck Lever9dc3b092009-08-09 15:09:46 -04002736 INIT_DELAYED_WORK(&transport->connect_worker,
Pavel Emelyanovcdd518d2010-10-04 16:57:40 +04002737 xs_tcp_setup_socket);
Chuck Lever9dc3b092009-08-09 15:09:46 -04002738 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP);
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002739 break;
2740 case AF_INET6:
2741 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
2742 xprt_set_bound(xprt);
2743
Chuck Lever9dc3b092009-08-09 15:09:46 -04002744 INIT_DELAYED_WORK(&transport->connect_worker,
Pavel Emelyanovcdd518d2010-10-04 16:57:40 +04002745 xs_tcp_setup_socket);
Chuck Lever9dc3b092009-08-09 15:09:46 -04002746 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP6);
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002747 break;
2748 default:
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002749 ret = ERR_PTR(-EAFNOSUPPORT);
2750 goto out_err;
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002751 }
2752
Chuck Leverc740eff2009-08-09 15:09:46 -04002753 if (xprt_bound(xprt))
2754 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
2755 xprt->address_strings[RPC_DISPLAY_ADDR],
2756 xprt->address_strings[RPC_DISPLAY_PORT],
2757 xprt->address_strings[RPC_DISPLAY_PROTO]);
2758 else
2759 dprintk("RPC: set up xprt to %s (autobind) via %s\n",
2760 xprt->address_strings[RPC_DISPLAY_ADDR],
2761 xprt->address_strings[RPC_DISPLAY_PROTO]);
2762
Chuck Leveredb267a2006-08-22 20:06:18 -04002763
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002764 if (try_module_get(THIS_MODULE))
2765 return xprt;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002766 ret = ERR_PTR(-EINVAL);
2767out_err:
Pavel Emelyanove204e622010-09-29 16:03:13 +04002768 xprt_free(xprt);
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002769 return ret;
Chuck Levera246b012005-08-11 16:25:23 -04002770}
Chuck Lever282b32e2006-12-05 16:35:51 -05002771
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002772/**
2773 * xs_setup_bc_tcp - Set up transport to use a TCP backchannel socket
2774 * @args: rpc transport creation arguments
2775 *
2776 */
2777static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args)
2778{
2779 struct sockaddr *addr = args->dstaddr;
2780 struct rpc_xprt *xprt;
2781 struct sock_xprt *transport;
2782 struct svc_sock *bc_sock;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002783 struct rpc_xprt *ret;
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002784
J. Bruce Fieldsf0418aa2010-12-08 13:48:19 -05002785 if (args->bc_xprt->xpt_bc_xprt) {
2786 /*
2787 * This server connection already has a backchannel
2788 * export; we can't create a new one, as we wouldn't be
2789 * able to match replies based on xid any more. So,
2790 * reuse the already-existing one:
2791 */
2792 return args->bc_xprt->xpt_bc_xprt;
2793 }
Trond Myklebustd9ba1312011-07-17 18:11:30 -04002794 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
2795 xprt_tcp_slot_table_entries);
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002796 if (IS_ERR(xprt))
2797 return xprt;
2798 transport = container_of(xprt, struct sock_xprt, xprt);
2799
2800 xprt->prot = IPPROTO_TCP;
2801 xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
2802 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
2803 xprt->timeout = &xs_tcp_default_timeout;
2804
2805 /* backchannel */
2806 xprt_set_bound(xprt);
2807 xprt->bind_timeout = 0;
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002808 xprt->reestablish_timeout = 0;
2809 xprt->idle_timeout = 0;
2810
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002811 xprt->ops = &bc_tcp_ops;
2812
2813 switch (addr->sa_family) {
2814 case AF_INET:
2815 xs_format_peer_addresses(xprt, "tcp",
2816 RPCBIND_NETID_TCP);
2817 break;
2818 case AF_INET6:
2819 xs_format_peer_addresses(xprt, "tcp",
2820 RPCBIND_NETID_TCP6);
2821 break;
2822 default:
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002823 ret = ERR_PTR(-EAFNOSUPPORT);
2824 goto out_err;
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002825 }
2826
Pavel Emelyanov50fa0d42010-10-05 20:49:35 +04002827 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
2828 xprt->address_strings[RPC_DISPLAY_ADDR],
2829 xprt->address_strings[RPC_DISPLAY_PORT],
2830 xprt->address_strings[RPC_DISPLAY_PROTO]);
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002831
2832 /*
J. Bruce Fields99de8ea2010-12-08 12:45:44 -05002833 * Once we've associated a backchannel xprt with a connection,
2834 * we want to keep it around as long as long as the connection
2835 * lasts, in case we need to start using it for a backchannel
2836 * again; this reference won't be dropped until bc_xprt is
2837 * destroyed.
2838 */
2839 xprt_get(xprt);
2840 args->bc_xprt->xpt_bc_xprt = xprt;
2841 xprt->bc_xprt = args->bc_xprt;
2842 bc_sock = container_of(args->bc_xprt, struct svc_sock, sk_xprt);
2843 transport->sock = bc_sock->sk_sock;
2844 transport->inet = bc_sock->sk_sk;
2845
2846 /*
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002847 * Since we don't want connections for the backchannel, we set
2848 * the xprt status to connected
2849 */
2850 xprt_set_connected(xprt);
2851
2852
2853 if (try_module_get(THIS_MODULE))
2854 return xprt;
J. Bruce Fields99de8ea2010-12-08 12:45:44 -05002855 xprt_put(xprt);
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002856 ret = ERR_PTR(-EINVAL);
2857out_err:
Pavel Emelyanove204e622010-09-29 16:03:13 +04002858 xprt_free(xprt);
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002859 return ret;
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002860}
2861
Chuck Lever176e21e2011-05-09 15:22:44 -04002862static struct xprt_class xs_local_transport = {
2863 .list = LIST_HEAD_INIT(xs_local_transport.list),
2864 .name = "named UNIX socket",
2865 .owner = THIS_MODULE,
2866 .ident = XPRT_TRANSPORT_LOCAL,
2867 .setup = xs_setup_local,
2868};
2869
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002870static struct xprt_class xs_udp_transport = {
2871 .list = LIST_HEAD_INIT(xs_udp_transport.list),
2872 .name = "udp",
2873 .owner = THIS_MODULE,
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002874 .ident = XPRT_TRANSPORT_UDP,
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002875 .setup = xs_setup_udp,
2876};
2877
2878static struct xprt_class xs_tcp_transport = {
2879 .list = LIST_HEAD_INIT(xs_tcp_transport.list),
2880 .name = "tcp",
2881 .owner = THIS_MODULE,
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002882 .ident = XPRT_TRANSPORT_TCP,
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002883 .setup = xs_setup_tcp,
2884};
2885
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002886static struct xprt_class xs_bc_tcp_transport = {
2887 .list = LIST_HEAD_INIT(xs_bc_tcp_transport.list),
2888 .name = "tcp NFSv4.1 backchannel",
2889 .owner = THIS_MODULE,
2890 .ident = XPRT_TRANSPORT_BC_TCP,
2891 .setup = xs_setup_bc_tcp,
2892};
2893
Chuck Lever282b32e2006-12-05 16:35:51 -05002894/**
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002895 * init_socket_xprt - set up xprtsock's sysctls, register with RPC client
Chuck Lever282b32e2006-12-05 16:35:51 -05002896 *
2897 */
2898int init_socket_xprt(void)
2899{
Chuck Leverfbf76682006-12-05 16:35:54 -05002900#ifdef RPC_DEBUG
Eric W. Biederman2b1bec52007-02-14 00:33:24 -08002901 if (!sunrpc_table_header)
Eric W. Biederman0b4d4142007-02-14 00:34:09 -08002902 sunrpc_table_header = register_sysctl_table(sunrpc_table);
Chuck Leverfbf76682006-12-05 16:35:54 -05002903#endif
2904
Chuck Lever176e21e2011-05-09 15:22:44 -04002905 xprt_register_transport(&xs_local_transport);
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002906 xprt_register_transport(&xs_udp_transport);
2907 xprt_register_transport(&xs_tcp_transport);
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002908 xprt_register_transport(&xs_bc_tcp_transport);
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002909
Chuck Lever282b32e2006-12-05 16:35:51 -05002910 return 0;
2911}
2912
2913/**
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002914 * cleanup_socket_xprt - remove xprtsock's sysctls, unregister
Chuck Lever282b32e2006-12-05 16:35:51 -05002915 *
2916 */
2917void cleanup_socket_xprt(void)
2918{
Chuck Leverfbf76682006-12-05 16:35:54 -05002919#ifdef RPC_DEBUG
2920 if (sunrpc_table_header) {
2921 unregister_sysctl_table(sunrpc_table_header);
2922 sunrpc_table_header = NULL;
2923 }
2924#endif
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002925
Chuck Lever176e21e2011-05-09 15:22:44 -04002926 xprt_unregister_transport(&xs_local_transport);
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002927 xprt_unregister_transport(&xs_udp_transport);
2928 xprt_unregister_transport(&xs_tcp_transport);
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002929 xprt_unregister_transport(&xs_bc_tcp_transport);
Chuck Lever282b32e2006-12-05 16:35:51 -05002930}
Trond Myklebustcbf11072009-08-09 15:06:19 -04002931
Rusty Russell9bbb9e52010-08-11 23:04:12 -06002932static int param_set_uint_minmax(const char *val,
2933 const struct kernel_param *kp,
Trond Myklebustcbf11072009-08-09 15:06:19 -04002934 unsigned int min, unsigned int max)
2935{
2936 unsigned long num;
2937 int ret;
2938
2939 if (!val)
2940 return -EINVAL;
2941 ret = strict_strtoul(val, 0, &num);
2942 if (ret == -EINVAL || num < min || num > max)
2943 return -EINVAL;
2944 *((unsigned int *)kp->arg) = num;
2945 return 0;
2946}
2947
Rusty Russell9bbb9e52010-08-11 23:04:12 -06002948static int param_set_portnr(const char *val, const struct kernel_param *kp)
Trond Myklebustcbf11072009-08-09 15:06:19 -04002949{
2950 return param_set_uint_minmax(val, kp,
2951 RPC_MIN_RESVPORT,
2952 RPC_MAX_RESVPORT);
2953}
2954
Rusty Russell9bbb9e52010-08-11 23:04:12 -06002955static struct kernel_param_ops param_ops_portnr = {
2956 .set = param_set_portnr,
2957 .get = param_get_uint,
2958};
2959
Trond Myklebustcbf11072009-08-09 15:06:19 -04002960#define param_check_portnr(name, p) \
2961 __param_check(name, p, unsigned int);
2962
2963module_param_named(min_resvport, xprt_min_resvport, portnr, 0644);
2964module_param_named(max_resvport, xprt_max_resvport, portnr, 0644);
2965
Rusty Russell9bbb9e52010-08-11 23:04:12 -06002966static int param_set_slot_table_size(const char *val,
2967 const struct kernel_param *kp)
Trond Myklebustcbf11072009-08-09 15:06:19 -04002968{
2969 return param_set_uint_minmax(val, kp,
2970 RPC_MIN_SLOT_TABLE,
2971 RPC_MAX_SLOT_TABLE);
2972}
2973
Rusty Russell9bbb9e52010-08-11 23:04:12 -06002974static struct kernel_param_ops param_ops_slot_table_size = {
2975 .set = param_set_slot_table_size,
2976 .get = param_get_uint,
2977};
2978
Trond Myklebustcbf11072009-08-09 15:06:19 -04002979#define param_check_slot_table_size(name, p) \
2980 __param_check(name, p, unsigned int);
2981
Trond Myklebustd9ba1312011-07-17 18:11:30 -04002982static int param_set_max_slot_table_size(const char *val,
2983 const struct kernel_param *kp)
2984{
2985 return param_set_uint_minmax(val, kp,
2986 RPC_MIN_SLOT_TABLE,
2987 RPC_MAX_SLOT_TABLE_LIMIT);
2988}
2989
2990static struct kernel_param_ops param_ops_max_slot_table_size = {
2991 .set = param_set_max_slot_table_size,
2992 .get = param_get_uint,
2993};
2994
2995#define param_check_max_slot_table_size(name, p) \
2996 __param_check(name, p, unsigned int);
2997
Trond Myklebustcbf11072009-08-09 15:06:19 -04002998module_param_named(tcp_slot_table_entries, xprt_tcp_slot_table_entries,
2999 slot_table_size, 0644);
Trond Myklebustd9ba1312011-07-17 18:11:30 -04003000module_param_named(tcp_max_slot_table_entries, xprt_max_tcp_slot_table_entries,
3001 max_slot_table_size, 0644);
Trond Myklebustcbf11072009-08-09 15:06:19 -04003002module_param_named(udp_slot_table_entries, xprt_udp_slot_table_entries,
3003 slot_table_size, 0644);
3004