blob: ddf0602603bdef4ac0a6ad29a9815bae56f04626 [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>
Jeff Layton59766872013-02-04 12:50:00 -050036#include <linux/sunrpc/addr.h>
Chuck Lever02107142006-01-03 09:55:49 +010037#include <linux/sunrpc/sched.h>
Rahul Iyer4cfc7e62009-09-10 17:32:28 +030038#include <linux/sunrpc/svcsock.h>
\"Talpey, Thomas\49c36fc2007-09-10 13:47:31 -040039#include <linux/sunrpc/xprtsock.h>
Chuck Levera246b012005-08-11 16:25:23 -040040#include <linux/file.h>
Trond Myklebust9e00abc2011-07-13 19:20:49 -040041#ifdef CONFIG_SUNRPC_BACKCHANNEL
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -040042#include <linux/sunrpc/bc_xprt.h>
43#endif
Chuck Levera246b012005-08-11 16:25:23 -040044
45#include <net/sock.h>
46#include <net/checksum.h>
47#include <net/udp.h>
48#include <net/tcp.h>
49
Rahul Iyer4cfc7e62009-09-10 17:32:28 +030050#include "sunrpc.h"
Chuck Lever176e21e2011-05-09 15:22:44 -040051
52static void xs_close(struct rpc_xprt *xprt);
53
Chuck Lever9903cd12005-08-11 16:25:26 -040054/*
Chuck Leverc556b752005-11-01 12:24:48 -050055 * xprtsock tunables
56 */
Trond Myklebust09acfea2012-03-11 15:22:54 -040057static unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE;
58static unsigned int xprt_tcp_slot_table_entries = RPC_MIN_SLOT_TABLE;
59static unsigned int xprt_max_tcp_slot_table_entries = RPC_MAX_SLOT_TABLE;
Chuck Leverc556b752005-11-01 12:24:48 -050060
Trond Myklebust09acfea2012-03-11 15:22:54 -040061static unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT;
62static unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT;
Chuck Leverc556b752005-11-01 12:24:48 -050063
Trond Myklebust7d1e8252009-03-11 14:38:03 -040064#define XS_TCP_LINGER_TO (15U * HZ)
Trond Myklebust25fe6142009-03-11 14:38:03 -040065static unsigned int xs_tcp_fin_timeout __read_mostly = XS_TCP_LINGER_TO;
Trond Myklebust7d1e8252009-03-11 14:38:03 -040066
Chuck Leverc556b752005-11-01 12:24:48 -050067/*
Chuck Leverfbf76682006-12-05 16:35:54 -050068 * We can register our own files under /proc/sys/sunrpc by
69 * calling register_sysctl_table() again. The files in that
70 * directory become the union of all files registered there.
71 *
72 * We simply need to make sure that we don't collide with
73 * someone else's file names!
74 */
75
76#ifdef RPC_DEBUG
77
78static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE;
79static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE;
Trond Myklebustd9ba1312011-07-17 18:11:30 -040080static unsigned int max_tcp_slot_table_limit = RPC_MAX_SLOT_TABLE_LIMIT;
Chuck Leverfbf76682006-12-05 16:35:54 -050081static unsigned int xprt_min_resvport_limit = RPC_MIN_RESVPORT;
82static unsigned int xprt_max_resvport_limit = RPC_MAX_RESVPORT;
83
84static struct ctl_table_header *sunrpc_table_header;
85
86/*
87 * FIXME: changing the UDP slot table size should also resize the UDP
88 * socket buffers for existing UDP transports
89 */
Joe Perchesfe2c6332013-06-11 23:04:25 -070090static struct ctl_table xs_tunables_table[] = {
Chuck Leverfbf76682006-12-05 16:35:54 -050091 {
Chuck Leverfbf76682006-12-05 16:35:54 -050092 .procname = "udp_slot_table_entries",
93 .data = &xprt_udp_slot_table_entries,
94 .maxlen = sizeof(unsigned int),
95 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -080096 .proc_handler = proc_dointvec_minmax,
Chuck Leverfbf76682006-12-05 16:35:54 -050097 .extra1 = &min_slot_table_size,
98 .extra2 = &max_slot_table_size
99 },
100 {
Chuck Leverfbf76682006-12-05 16:35:54 -0500101 .procname = "tcp_slot_table_entries",
102 .data = &xprt_tcp_slot_table_entries,
103 .maxlen = sizeof(unsigned int),
104 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -0800105 .proc_handler = proc_dointvec_minmax,
Chuck Leverfbf76682006-12-05 16:35:54 -0500106 .extra1 = &min_slot_table_size,
107 .extra2 = &max_slot_table_size
108 },
109 {
Trond Myklebustd9ba1312011-07-17 18:11:30 -0400110 .procname = "tcp_max_slot_table_entries",
111 .data = &xprt_max_tcp_slot_table_entries,
112 .maxlen = sizeof(unsigned int),
113 .mode = 0644,
114 .proc_handler = proc_dointvec_minmax,
115 .extra1 = &min_slot_table_size,
116 .extra2 = &max_tcp_slot_table_limit
117 },
118 {
Chuck Leverfbf76682006-12-05 16:35:54 -0500119 .procname = "min_resvport",
120 .data = &xprt_min_resvport,
121 .maxlen = sizeof(unsigned int),
122 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -0800123 .proc_handler = proc_dointvec_minmax,
Chuck Leverfbf76682006-12-05 16:35:54 -0500124 .extra1 = &xprt_min_resvport_limit,
125 .extra2 = &xprt_max_resvport_limit
126 },
127 {
Chuck Leverfbf76682006-12-05 16:35:54 -0500128 .procname = "max_resvport",
129 .data = &xprt_max_resvport,
130 .maxlen = sizeof(unsigned int),
131 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -0800132 .proc_handler = proc_dointvec_minmax,
Chuck Leverfbf76682006-12-05 16:35:54 -0500133 .extra1 = &xprt_min_resvport_limit,
134 .extra2 = &xprt_max_resvport_limit
135 },
136 {
Trond Myklebust25fe6142009-03-11 14:38:03 -0400137 .procname = "tcp_fin_timeout",
138 .data = &xs_tcp_fin_timeout,
139 .maxlen = sizeof(xs_tcp_fin_timeout),
140 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -0800141 .proc_handler = proc_dointvec_jiffies,
Trond Myklebust25fe6142009-03-11 14:38:03 -0400142 },
Eric W. Biedermanf8572d82009-11-05 13:32:03 -0800143 { },
Chuck Leverfbf76682006-12-05 16:35:54 -0500144};
145
Joe Perchesfe2c6332013-06-11 23:04:25 -0700146static struct ctl_table sunrpc_table[] = {
Chuck Leverfbf76682006-12-05 16:35:54 -0500147 {
Chuck Leverfbf76682006-12-05 16:35:54 -0500148 .procname = "sunrpc",
149 .mode = 0555,
150 .child = xs_tunables_table
151 },
Eric W. Biedermanf8572d82009-11-05 13:32:03 -0800152 { },
Chuck Leverfbf76682006-12-05 16:35:54 -0500153};
154
155#endif
156
157/*
Chuck Lever03bf4b72005-08-25 16:25:55 -0700158 * Wait duration for a reply from the RPC portmapper.
159 */
160#define XS_BIND_TO (60U * HZ)
161
162/*
163 * Delay if a UDP socket connect error occurs. This is most likely some
164 * kind of resource problem on the local host.
165 */
166#define XS_UDP_REEST_TO (2U * HZ)
167
168/*
169 * The reestablish timeout allows clients to delay for a bit before attempting
170 * to reconnect to a server that just dropped our connection.
171 *
172 * We implement an exponential backoff when trying to reestablish a TCP
173 * transport connection with the server. Some servers like to drop a TCP
174 * connection when they are overworked, so we start with a short timeout and
175 * increase over time if the server is down or not responding.
176 */
177#define XS_TCP_INIT_REEST_TO (3U * HZ)
178#define XS_TCP_MAX_REEST_TO (5U * 60 * HZ)
179
180/*
181 * TCP idle timeout; client drops the transport socket if it is idle
182 * for this long. Note that we also timeout UDP sockets to prevent
183 * holding port numbers when there is no RPC traffic.
184 */
185#define XS_IDLE_DISC_TO (5U * 60 * HZ)
186
Chuck Levera246b012005-08-11 16:25:23 -0400187#ifdef RPC_DEBUG
188# undef RPC_DEBUG_DATA
Chuck Lever9903cd12005-08-11 16:25:26 -0400189# define RPCDBG_FACILITY RPCDBG_TRANS
Chuck Levera246b012005-08-11 16:25:23 -0400190#endif
191
Chuck Levera246b012005-08-11 16:25:23 -0400192#ifdef RPC_DEBUG_DATA
Chuck Lever9903cd12005-08-11 16:25:26 -0400193static void xs_pktdump(char *msg, u32 *packet, unsigned int count)
Chuck Levera246b012005-08-11 16:25:23 -0400194{
Chuck Lever9903cd12005-08-11 16:25:26 -0400195 u8 *buf = (u8 *) packet;
196 int j;
Chuck Levera246b012005-08-11 16:25:23 -0400197
Chuck Lever46121cf2007-01-31 12:14:08 -0500198 dprintk("RPC: %s\n", msg);
Chuck Levera246b012005-08-11 16:25:23 -0400199 for (j = 0; j < count && j < 128; j += 4) {
200 if (!(j & 31)) {
201 if (j)
202 dprintk("\n");
203 dprintk("0x%04x ", j);
204 }
205 dprintk("%02x%02x%02x%02x ",
206 buf[j], buf[j+1], buf[j+2], buf[j+3]);
207 }
208 dprintk("\n");
209}
210#else
Chuck Lever9903cd12005-08-11 16:25:26 -0400211static inline void xs_pktdump(char *msg, u32 *packet, unsigned int count)
Chuck Levera246b012005-08-11 16:25:23 -0400212{
213 /* NOP */
214}
215#endif
216
Chuck Leverffc2e512006-12-05 16:35:11 -0500217struct sock_xprt {
218 struct rpc_xprt xprt;
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500219
220 /*
221 * Network layer
222 */
223 struct socket * sock;
224 struct sock * inet;
Chuck Lever51971132006-12-05 16:35:19 -0500225
226 /*
227 * State of TCP reply receive
228 */
229 __be32 tcp_fraghdr,
Trond Myklebustb76ce562010-06-16 13:57:32 -0400230 tcp_xid,
231 tcp_calldir;
Chuck Lever51971132006-12-05 16:35:19 -0500232
233 u32 tcp_offset,
234 tcp_reclen;
235
236 unsigned long tcp_copied,
237 tcp_flags;
Chuck Leverc8475462006-12-05 16:35:26 -0500238
239 /*
240 * Connection of transports
241 */
Trond Myklebust34161db2006-12-07 15:48:15 -0500242 struct delayed_work connect_worker;
Chuck Leverfbfffbd2009-08-09 15:09:46 -0400243 struct sockaddr_storage srcaddr;
244 unsigned short srcport;
Chuck Lever7c6e0662006-12-05 16:35:30 -0500245
246 /*
247 * UDP socket buffer size parameters
248 */
249 size_t rcvsize,
250 sndsize;
Chuck Lever314dfd72006-12-05 16:35:34 -0500251
252 /*
253 * Saved socket callback addresses
254 */
255 void (*old_data_ready)(struct sock *, int);
256 void (*old_state_change)(struct sock *);
257 void (*old_write_space)(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 Perchesfc0b57912010-03-08 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 Perchesfc0b57912010-03-08 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 Myklebuste06799f2007-11-05 15:44:12 -0500740 xs_tcp_shutdown(xprt);
Chuck Lever262965f2005-08-11 16:25:56 -0400741 case -ECONNREFUSED:
742 case -ENOTCONN:
Trond Myklebustb9d2bb22012-10-23 11:40:02 -0400743 case -EPIPE:
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 Myklebust43cedbf0e2011-07-17 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;
Trond Myklebusta4f08352013-01-08 09:10:21 -0500774 set_bit(XPRT_CLOSE_WAIT, &xprt->state);
Trond Myklebuste0ab53d2006-07-27 17:22:50 -0400775out_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;
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -0400784}
785
786static void xs_restore_old_callbacks(struct sock_xprt *transport, struct sock *sk)
787{
788 sk->sk_data_ready = transport->old_data_ready;
789 sk->sk_state_change = transport->old_state_change;
790 sk->sk_write_space = transport->old_write_space;
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -0400791}
792
Chuck Leverfe315e72009-03-11 14:10:21 -0400793static void xs_reset_transport(struct sock_xprt *transport)
Chuck Levera246b012005-08-11 16:25:23 -0400794{
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500795 struct socket *sock = transport->sock;
796 struct sock *sk = transport->inet;
Chuck Levera246b012005-08-11 16:25:23 -0400797
Chuck Leverfe315e72009-03-11 14:10:21 -0400798 if (sk == NULL)
799 return;
Chuck Lever9903cd12005-08-11 16:25:26 -0400800
Trond Myklebust246408d2011-03-22 18:40:10 -0400801 transport->srcport = 0;
802
Chuck Levera246b012005-08-11 16:25:23 -0400803 write_lock_bh(&sk->sk_callback_lock);
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500804 transport->inet = NULL;
805 transport->sock = NULL;
Chuck Levera246b012005-08-11 16:25:23 -0400806
Chuck Lever9903cd12005-08-11 16:25:26 -0400807 sk->sk_user_data = NULL;
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -0400808
809 xs_restore_old_callbacks(transport, sk);
Chuck Levera246b012005-08-11 16:25:23 -0400810 write_unlock_bh(&sk->sk_callback_lock);
811
Chuck Lever9903cd12005-08-11 16:25:26 -0400812 sk->sk_no_check = 0;
Chuck Levera246b012005-08-11 16:25:23 -0400813
814 sock_release(sock);
Chuck Leverfe315e72009-03-11 14:10:21 -0400815}
816
817/**
818 * xs_close - close a socket
819 * @xprt: transport
820 *
821 * This is used when all requests are complete; ie, no DRC state remains
822 * on the server we want to save.
Trond Myklebustf75e6742009-04-21 17:18:20 -0400823 *
824 * The caller _must_ be holding XPRT_LOCKED in order to avoid issues with
825 * xs_reset_transport() zeroing the socket from underneath a writer.
Chuck Leverfe315e72009-03-11 14:10:21 -0400826 */
827static void xs_close(struct rpc_xprt *xprt)
828{
829 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
830
831 dprintk("RPC: xs_close xprt %p\n", xprt);
832
833 xs_reset_transport(transport);
Neil Brown61d0a8e2009-09-23 14:36:37 -0400834 xprt->reestablish_timeout = 0;
Chuck Leverfe315e72009-03-11 14:10:21 -0400835
Trond Myklebust632e3bd2006-01-03 09:55:55 +0100836 smp_mb__before_clear_bit();
Trond Myklebust7d1e8252009-03-11 14:38:03 -0400837 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
Trond Myklebust632e3bd2006-01-03 09:55:55 +0100838 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
Trond Myklebust3b948ae2007-11-05 17:42:39 -0500839 clear_bit(XPRT_CLOSING, &xprt->state);
Trond Myklebust632e3bd2006-01-03 09:55:55 +0100840 smp_mb__after_clear_bit();
Trond Myklebust62da3b22007-11-06 18:44:20 -0500841 xprt_disconnect_done(xprt);
Chuck Levera246b012005-08-11 16:25:23 -0400842}
843
Trond Myklebustf75e6742009-04-21 17:18:20 -0400844static void xs_tcp_close(struct rpc_xprt *xprt)
845{
846 if (test_and_clear_bit(XPRT_CONNECTION_CLOSE, &xprt->state))
847 xs_close(xprt);
848 else
849 xs_tcp_shutdown(xprt);
850}
851
J. Bruce Fields190b1ecf2013-03-08 13:13:08 -0500852static void xs_local_destroy(struct rpc_xprt *xprt)
853{
854 xs_close(xprt);
855 xs_free_peer_addresses(xprt);
856 xprt_free(xprt);
857 module_put(THIS_MODULE);
858}
859
Chuck Lever9903cd12005-08-11 16:25:26 -0400860/**
861 * xs_destroy - prepare to shutdown a transport
862 * @xprt: doomed transport
863 *
864 */
865static void xs_destroy(struct rpc_xprt *xprt)
Chuck Levera246b012005-08-11 16:25:23 -0400866{
Chuck Leverc8475462006-12-05 16:35:26 -0500867 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
868
Chuck Lever46121cf2007-01-31 12:14:08 -0500869 dprintk("RPC: xs_destroy xprt %p\n", xprt);
Chuck Lever9903cd12005-08-11 16:25:26 -0400870
Tejun Heoafe2c512010-12-14 16:21:17 +0100871 cancel_delayed_work_sync(&transport->connect_worker);
Chuck Levera246b012005-08-11 16:25:23 -0400872
J. Bruce Fields190b1ecf2013-03-08 13:13:08 -0500873 xs_local_destroy(xprt);
Chuck Levera246b012005-08-11 16:25:23 -0400874}
875
Chuck Lever9903cd12005-08-11 16:25:26 -0400876static inline struct rpc_xprt *xprt_from_sock(struct sock *sk)
Chuck Levera246b012005-08-11 16:25:23 -0400877{
Chuck Lever9903cd12005-08-11 16:25:26 -0400878 return (struct rpc_xprt *) sk->sk_user_data;
879}
880
Chuck Lever176e21e2011-05-09 15:22:44 -0400881static int xs_local_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb)
882{
883 struct xdr_skb_reader desc = {
884 .skb = skb,
885 .offset = sizeof(rpc_fraghdr),
886 .count = skb->len - sizeof(rpc_fraghdr),
887 };
888
889 if (xdr_partial_copy_from_skb(xdr, 0, &desc, xdr_skb_read_bits) < 0)
890 return -1;
891 if (desc.count)
892 return -1;
893 return 0;
894}
895
896/**
897 * xs_local_data_ready - "data ready" callback for AF_LOCAL sockets
898 * @sk: socket with data to read
899 * @len: how much data to read
900 *
901 * Currently this assumes we can read the whole reply in a single gulp.
902 */
903static void xs_local_data_ready(struct sock *sk, int len)
904{
905 struct rpc_task *task;
906 struct rpc_xprt *xprt;
907 struct rpc_rqst *rovr;
908 struct sk_buff *skb;
909 int err, repsize, copied;
910 u32 _xid;
911 __be32 *xp;
912
913 read_lock_bh(&sk->sk_callback_lock);
914 dprintk("RPC: %s...\n", __func__);
915 xprt = xprt_from_sock(sk);
916 if (xprt == NULL)
917 goto out;
918
919 skb = skb_recv_datagram(sk, 0, 1, &err);
920 if (skb == NULL)
921 goto out;
922
Chuck Lever176e21e2011-05-09 15:22:44 -0400923 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
Chuck Levera246b012005-08-11 16:25:23 -0400984 repsize = skb->len - sizeof(struct udphdr);
985 if (repsize < 4) {
Chuck Lever46121cf2007-01-31 12:14:08 -0500986 dprintk("RPC: impossible RPC reply size %d!\n", repsize);
Chuck Levera246b012005-08-11 16:25:23 -0400987 goto dropit;
988 }
989
990 /* Copy the XID from the skb... */
991 xp = skb_header_pointer(skb, sizeof(struct udphdr),
992 sizeof(_xid), &_xid);
993 if (xp == NULL)
994 goto dropit;
995
996 /* Look up and lock the request corresponding to the given XID */
Chuck Lever4a0f8c02005-08-11 16:25:32 -0400997 spin_lock(&xprt->transport_lock);
Chuck Levera246b012005-08-11 16:25:23 -0400998 rovr = xprt_lookup_rqst(xprt, *xp);
999 if (!rovr)
1000 goto out_unlock;
1001 task = rovr->rq_task;
1002
Chuck Levera246b012005-08-11 16:25:23 -04001003 if ((copied = rovr->rq_private_buf.buflen) > repsize)
1004 copied = repsize;
1005
1006 /* Suck it into the iovec, verify checksum if not done by hw. */
Herbert Xu1781f7f2007-12-11 11:30:32 -08001007 if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb)) {
1008 UDPX_INC_STATS_BH(sk, UDP_MIB_INERRORS);
Chuck Levera246b012005-08-11 16:25:23 -04001009 goto out_unlock;
Herbert Xu1781f7f2007-12-11 11:30:32 -08001010 }
1011
1012 UDPX_INC_STATS_BH(sk, UDP_MIB_INDATAGRAMS);
Chuck Levera246b012005-08-11 16:25:23 -04001013
Trond Myklebust6a24dfb2013-01-08 09:48:15 -05001014 xprt_adjust_cwnd(xprt, task, copied);
Chuck Lever1570c1e2005-08-25 16:25:52 -07001015 xprt_complete_rqst(task, copied);
Chuck Levera246b012005-08-11 16:25:23 -04001016
1017 out_unlock:
Chuck Lever4a0f8c02005-08-11 16:25:32 -04001018 spin_unlock(&xprt->transport_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001019 dropit:
1020 skb_free_datagram(sk, skb);
1021 out:
Eric Dumazetf064af12010-09-22 12:43:39 +00001022 read_unlock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001023}
1024
Trond Myklebusta519fc72012-09-12 16:49:15 -04001025/*
1026 * Helper function to force a TCP close if the server is sending
1027 * junk and/or it has put us in CLOSE_WAIT
1028 */
1029static void xs_tcp_force_close(struct rpc_xprt *xprt)
1030{
1031 set_bit(XPRT_CONNECTION_CLOSE, &xprt->state);
1032 xprt_force_disconnect(xprt);
1033}
1034
Chuck Leverdd456472006-12-05 16:35:44 -05001035static inline void xs_tcp_read_fraghdr(struct rpc_xprt *xprt, struct xdr_skb_reader *desc)
Chuck Levera246b012005-08-11 16:25:23 -04001036{
Chuck Lever51971132006-12-05 16:35:19 -05001037 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -04001038 size_t len, used;
1039 char *p;
1040
Chuck Lever51971132006-12-05 16:35:19 -05001041 p = ((char *) &transport->tcp_fraghdr) + transport->tcp_offset;
1042 len = sizeof(transport->tcp_fraghdr) - transport->tcp_offset;
Chuck Lever9d292312006-12-05 16:35:41 -05001043 used = xdr_skb_read_bits(desc, p, len);
Chuck Lever51971132006-12-05 16:35:19 -05001044 transport->tcp_offset += used;
Chuck Levera246b012005-08-11 16:25:23 -04001045 if (used != len)
1046 return;
Chuck Lever808012f2005-08-25 16:25:49 -07001047
Chuck Lever51971132006-12-05 16:35:19 -05001048 transport->tcp_reclen = ntohl(transport->tcp_fraghdr);
1049 if (transport->tcp_reclen & RPC_LAST_STREAM_FRAGMENT)
Chuck Levere136d092006-12-05 16:35:23 -05001050 transport->tcp_flags |= TCP_RCV_LAST_FRAG;
Chuck Levera246b012005-08-11 16:25:23 -04001051 else
Chuck Levere136d092006-12-05 16:35:23 -05001052 transport->tcp_flags &= ~TCP_RCV_LAST_FRAG;
Chuck Lever51971132006-12-05 16:35:19 -05001053 transport->tcp_reclen &= RPC_FRAGMENT_SIZE_MASK;
Chuck Lever808012f2005-08-25 16:25:49 -07001054
Chuck Levere136d092006-12-05 16:35:23 -05001055 transport->tcp_flags &= ~TCP_RCV_COPY_FRAGHDR;
Chuck Lever51971132006-12-05 16:35:19 -05001056 transport->tcp_offset = 0;
Chuck Lever808012f2005-08-25 16:25:49 -07001057
Chuck Levera246b012005-08-11 16:25:23 -04001058 /* Sanity check of the record length */
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001059 if (unlikely(transport->tcp_reclen < 8)) {
Chuck Lever46121cf2007-01-31 12:14:08 -05001060 dprintk("RPC: invalid TCP record fragment length\n");
Trond Myklebusta519fc72012-09-12 16:49:15 -04001061 xs_tcp_force_close(xprt);
Chuck Lever9903cd12005-08-11 16:25:26 -04001062 return;
Chuck Levera246b012005-08-11 16:25:23 -04001063 }
Chuck Lever46121cf2007-01-31 12:14:08 -05001064 dprintk("RPC: reading TCP record fragment of length %d\n",
Chuck Lever51971132006-12-05 16:35:19 -05001065 transport->tcp_reclen);
Chuck Levera246b012005-08-11 16:25:23 -04001066}
1067
Chuck Lever51971132006-12-05 16:35:19 -05001068static void xs_tcp_check_fraghdr(struct sock_xprt *transport)
Chuck Levera246b012005-08-11 16:25:23 -04001069{
Chuck Lever51971132006-12-05 16:35:19 -05001070 if (transport->tcp_offset == transport->tcp_reclen) {
Chuck Levere136d092006-12-05 16:35:23 -05001071 transport->tcp_flags |= TCP_RCV_COPY_FRAGHDR;
Chuck Lever51971132006-12-05 16:35:19 -05001072 transport->tcp_offset = 0;
Chuck Levere136d092006-12-05 16:35:23 -05001073 if (transport->tcp_flags & TCP_RCV_LAST_FRAG) {
1074 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
1075 transport->tcp_flags |= TCP_RCV_COPY_XID;
Chuck Lever51971132006-12-05 16:35:19 -05001076 transport->tcp_copied = 0;
Chuck Levera246b012005-08-11 16:25:23 -04001077 }
1078 }
1079}
1080
Chuck Leverdd456472006-12-05 16:35:44 -05001081static inline void xs_tcp_read_xid(struct sock_xprt *transport, struct xdr_skb_reader *desc)
Chuck Levera246b012005-08-11 16:25:23 -04001082{
1083 size_t len, used;
1084 char *p;
1085
Chuck Lever51971132006-12-05 16:35:19 -05001086 len = sizeof(transport->tcp_xid) - transport->tcp_offset;
Chuck Lever46121cf2007-01-31 12:14:08 -05001087 dprintk("RPC: reading XID (%Zu bytes)\n", len);
Chuck Lever51971132006-12-05 16:35:19 -05001088 p = ((char *) &transport->tcp_xid) + transport->tcp_offset;
Chuck Lever9d292312006-12-05 16:35:41 -05001089 used = xdr_skb_read_bits(desc, p, len);
Chuck Lever51971132006-12-05 16:35:19 -05001090 transport->tcp_offset += used;
Chuck Levera246b012005-08-11 16:25:23 -04001091 if (used != len)
1092 return;
Chuck Levere136d092006-12-05 16:35:23 -05001093 transport->tcp_flags &= ~TCP_RCV_COPY_XID;
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001094 transport->tcp_flags |= TCP_RCV_READ_CALLDIR;
Chuck Lever51971132006-12-05 16:35:19 -05001095 transport->tcp_copied = 4;
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001096 dprintk("RPC: reading %s XID %08x\n",
1097 (transport->tcp_flags & TCP_RPC_REPLY) ? "reply for"
1098 : "request with",
Chuck Lever51971132006-12-05 16:35:19 -05001099 ntohl(transport->tcp_xid));
1100 xs_tcp_check_fraghdr(transport);
Chuck Levera246b012005-08-11 16:25:23 -04001101}
1102
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001103static inline void xs_tcp_read_calldir(struct sock_xprt *transport,
1104 struct xdr_skb_reader *desc)
Chuck Levera246b012005-08-11 16:25:23 -04001105{
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001106 size_t len, used;
1107 u32 offset;
Trond Myklebustb76ce562010-06-16 13:57:32 -04001108 char *p;
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001109
1110 /*
1111 * We want transport->tcp_offset to be 8 at the end of this routine
1112 * (4 bytes for the xid and 4 bytes for the call/reply flag).
1113 * When this function is called for the first time,
1114 * transport->tcp_offset is 4 (after having already read the xid).
1115 */
1116 offset = transport->tcp_offset - sizeof(transport->tcp_xid);
Trond Myklebustb76ce562010-06-16 13:57:32 -04001117 len = sizeof(transport->tcp_calldir) - offset;
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001118 dprintk("RPC: reading CALL/REPLY flag (%Zu bytes)\n", len);
Trond Myklebustb76ce562010-06-16 13:57:32 -04001119 p = ((char *) &transport->tcp_calldir) + offset;
1120 used = xdr_skb_read_bits(desc, p, len);
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001121 transport->tcp_offset += used;
1122 if (used != len)
1123 return;
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001124 transport->tcp_flags &= ~TCP_RCV_READ_CALLDIR;
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001125 /*
1126 * We don't yet have the XDR buffer, so we will write the calldir
1127 * out after we get the buffer from the 'struct rpc_rqst'
1128 */
Trond Myklebustb76ce562010-06-16 13:57:32 -04001129 switch (ntohl(transport->tcp_calldir)) {
1130 case RPC_REPLY:
1131 transport->tcp_flags |= TCP_RCV_COPY_CALLDIR;
1132 transport->tcp_flags |= TCP_RCV_COPY_DATA;
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001133 transport->tcp_flags |= TCP_RPC_REPLY;
Trond Myklebustb76ce562010-06-16 13:57:32 -04001134 break;
1135 case RPC_CALL:
1136 transport->tcp_flags |= TCP_RCV_COPY_CALLDIR;
1137 transport->tcp_flags |= TCP_RCV_COPY_DATA;
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001138 transport->tcp_flags &= ~TCP_RPC_REPLY;
Trond Myklebustb76ce562010-06-16 13:57:32 -04001139 break;
1140 default:
1141 dprintk("RPC: invalid request message type\n");
Trond Myklebusta519fc72012-09-12 16:49:15 -04001142 xs_tcp_force_close(&transport->xprt);
Trond Myklebustb76ce562010-06-16 13:57:32 -04001143 }
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001144 xs_tcp_check_fraghdr(transport);
1145}
1146
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001147static inline void xs_tcp_read_common(struct rpc_xprt *xprt,
1148 struct xdr_skb_reader *desc,
1149 struct rpc_rqst *req)
Chuck Levera246b012005-08-11 16:25:23 -04001150{
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001151 struct sock_xprt *transport =
1152 container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -04001153 struct xdr_buf *rcvbuf;
1154 size_t len;
1155 ssize_t r;
1156
Chuck Levera246b012005-08-11 16:25:23 -04001157 rcvbuf = &req->rq_private_buf;
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001158
1159 if (transport->tcp_flags & TCP_RCV_COPY_CALLDIR) {
1160 /*
1161 * Save the RPC direction in the XDR buffer
1162 */
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001163 memcpy(rcvbuf->head[0].iov_base + transport->tcp_copied,
Trond Myklebustb76ce562010-06-16 13:57:32 -04001164 &transport->tcp_calldir,
1165 sizeof(transport->tcp_calldir));
1166 transport->tcp_copied += sizeof(transport->tcp_calldir);
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001167 transport->tcp_flags &= ~TCP_RCV_COPY_CALLDIR;
Chuck Levera246b012005-08-11 16:25:23 -04001168 }
1169
Chuck Levera246b012005-08-11 16:25:23 -04001170 len = desc->count;
Chuck Lever51971132006-12-05 16:35:19 -05001171 if (len > transport->tcp_reclen - transport->tcp_offset) {
Chuck Leverdd456472006-12-05 16:35:44 -05001172 struct xdr_skb_reader my_desc;
Chuck Levera246b012005-08-11 16:25:23 -04001173
Chuck Lever51971132006-12-05 16:35:19 -05001174 len = transport->tcp_reclen - transport->tcp_offset;
Chuck Levera246b012005-08-11 16:25:23 -04001175 memcpy(&my_desc, desc, sizeof(my_desc));
1176 my_desc.count = len;
Chuck Lever51971132006-12-05 16:35:19 -05001177 r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied,
Chuck Lever9d292312006-12-05 16:35:41 -05001178 &my_desc, xdr_skb_read_bits);
Chuck Levera246b012005-08-11 16:25:23 -04001179 desc->count -= r;
1180 desc->offset += r;
1181 } else
Chuck Lever51971132006-12-05 16:35:19 -05001182 r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied,
Chuck Lever9d292312006-12-05 16:35:41 -05001183 desc, xdr_skb_read_bits);
Chuck Levera246b012005-08-11 16:25:23 -04001184
1185 if (r > 0) {
Chuck Lever51971132006-12-05 16:35:19 -05001186 transport->tcp_copied += r;
1187 transport->tcp_offset += r;
Chuck Levera246b012005-08-11 16:25:23 -04001188 }
1189 if (r != len) {
1190 /* Error when copying to the receive buffer,
1191 * usually because we weren't able to allocate
1192 * additional buffer pages. All we can do now
Chuck Levere136d092006-12-05 16:35:23 -05001193 * is turn off TCP_RCV_COPY_DATA, so the request
Chuck Levera246b012005-08-11 16:25:23 -04001194 * will not receive any additional updates,
1195 * and time out.
1196 * Any remaining data from this record will
1197 * be discarded.
1198 */
Chuck Levere136d092006-12-05 16:35:23 -05001199 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
Chuck Lever46121cf2007-01-31 12:14:08 -05001200 dprintk("RPC: XID %08x truncated request\n",
Chuck Lever51971132006-12-05 16:35:19 -05001201 ntohl(transport->tcp_xid));
Chuck Lever46121cf2007-01-31 12:14:08 -05001202 dprintk("RPC: xprt = %p, tcp_copied = %lu, "
1203 "tcp_offset = %u, tcp_reclen = %u\n",
1204 xprt, transport->tcp_copied,
1205 transport->tcp_offset, transport->tcp_reclen);
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001206 return;
Chuck Levera246b012005-08-11 16:25:23 -04001207 }
1208
Chuck Lever46121cf2007-01-31 12:14:08 -05001209 dprintk("RPC: XID %08x read %Zd bytes\n",
Chuck Lever51971132006-12-05 16:35:19 -05001210 ntohl(transport->tcp_xid), r);
Chuck Lever46121cf2007-01-31 12:14:08 -05001211 dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, "
1212 "tcp_reclen = %u\n", xprt, transport->tcp_copied,
1213 transport->tcp_offset, transport->tcp_reclen);
Chuck Levera246b012005-08-11 16:25:23 -04001214
Chuck Lever51971132006-12-05 16:35:19 -05001215 if (transport->tcp_copied == req->rq_private_buf.buflen)
Chuck Levere136d092006-12-05 16:35:23 -05001216 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
Chuck Lever51971132006-12-05 16:35:19 -05001217 else if (transport->tcp_offset == transport->tcp_reclen) {
Chuck Levere136d092006-12-05 16:35:23 -05001218 if (transport->tcp_flags & TCP_RCV_LAST_FRAG)
1219 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
Chuck Levera246b012005-08-11 16:25:23 -04001220 }
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001221}
1222
1223/*
1224 * Finds the request corresponding to the RPC xid and invokes the common
1225 * tcp read code to read the data.
1226 */
1227static inline int xs_tcp_read_reply(struct rpc_xprt *xprt,
1228 struct xdr_skb_reader *desc)
1229{
1230 struct sock_xprt *transport =
1231 container_of(xprt, struct sock_xprt, xprt);
1232 struct rpc_rqst *req;
1233
1234 dprintk("RPC: read reply XID %08x\n", ntohl(transport->tcp_xid));
1235
1236 /* Find and lock the request corresponding to this xid */
1237 spin_lock(&xprt->transport_lock);
1238 req = xprt_lookup_rqst(xprt, transport->tcp_xid);
1239 if (!req) {
1240 dprintk("RPC: XID %08x request not found!\n",
1241 ntohl(transport->tcp_xid));
1242 spin_unlock(&xprt->transport_lock);
1243 return -1;
1244 }
1245
1246 xs_tcp_read_common(xprt, desc, req);
1247
Chuck Levere136d092006-12-05 16:35:23 -05001248 if (!(transport->tcp_flags & TCP_RCV_COPY_DATA))
Chuck Lever51971132006-12-05 16:35:19 -05001249 xprt_complete_rqst(req->rq_task, transport->tcp_copied);
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001250
Chuck Lever4a0f8c02005-08-11 16:25:32 -04001251 spin_unlock(&xprt->transport_lock);
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001252 return 0;
1253}
1254
Trond Myklebust9e00abc2011-07-13 19:20:49 -04001255#if defined(CONFIG_SUNRPC_BACKCHANNEL)
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001256/*
1257 * Obtains an rpc_rqst previously allocated and invokes the common
1258 * tcp read code to read the data. The result is placed in the callback
1259 * queue.
1260 * If we're unable to obtain the rpc_rqst we schedule the closing of the
1261 * connection and return -1.
1262 */
1263static inline int xs_tcp_read_callback(struct rpc_xprt *xprt,
1264 struct xdr_skb_reader *desc)
1265{
1266 struct sock_xprt *transport =
1267 container_of(xprt, struct sock_xprt, xprt);
1268 struct rpc_rqst *req;
1269
1270 req = xprt_alloc_bc_request(xprt);
1271 if (req == NULL) {
1272 printk(KERN_WARNING "Callback slot table overflowed\n");
1273 xprt_force_disconnect(xprt);
1274 return -1;
1275 }
1276
1277 req->rq_xid = transport->tcp_xid;
1278 dprintk("RPC: read callback XID %08x\n", ntohl(req->rq_xid));
1279 xs_tcp_read_common(xprt, desc, req);
1280
1281 if (!(transport->tcp_flags & TCP_RCV_COPY_DATA)) {
1282 struct svc_serv *bc_serv = xprt->bc_serv;
1283
1284 /*
1285 * Add callback request to callback list. The callback
1286 * service sleeps on the sv_cb_waitq waiting for new
1287 * requests. Wake it up after adding enqueing the
1288 * request.
1289 */
1290 dprintk("RPC: add callback request to list\n");
1291 spin_lock(&bc_serv->sv_cb_lock);
1292 list_add(&req->rq_bc_list, &bc_serv->sv_cb_list);
1293 spin_unlock(&bc_serv->sv_cb_lock);
1294 wake_up(&bc_serv->sv_cb_waitq);
1295 }
1296
1297 req->rq_private_buf.len = transport->tcp_copied;
1298
1299 return 0;
1300}
1301
1302static inline int _xs_tcp_read_data(struct rpc_xprt *xprt,
1303 struct xdr_skb_reader *desc)
1304{
1305 struct sock_xprt *transport =
1306 container_of(xprt, struct sock_xprt, xprt);
1307
1308 return (transport->tcp_flags & TCP_RPC_REPLY) ?
1309 xs_tcp_read_reply(xprt, desc) :
1310 xs_tcp_read_callback(xprt, desc);
1311}
1312#else
1313static inline int _xs_tcp_read_data(struct rpc_xprt *xprt,
1314 struct xdr_skb_reader *desc)
1315{
1316 return xs_tcp_read_reply(xprt, desc);
1317}
Trond Myklebust9e00abc2011-07-13 19:20:49 -04001318#endif /* CONFIG_SUNRPC_BACKCHANNEL */
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001319
1320/*
1321 * Read data off the transport. This can be either an RPC_CALL or an
1322 * RPC_REPLY. Relay the processing to helper functions.
1323 */
1324static void xs_tcp_read_data(struct rpc_xprt *xprt,
1325 struct xdr_skb_reader *desc)
1326{
1327 struct sock_xprt *transport =
1328 container_of(xprt, struct sock_xprt, xprt);
1329
1330 if (_xs_tcp_read_data(xprt, desc) == 0)
1331 xs_tcp_check_fraghdr(transport);
1332 else {
1333 /*
1334 * The transport_lock protects the request handling.
1335 * There's no need to hold it to update the tcp_flags.
1336 */
1337 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
1338 }
Chuck Levera246b012005-08-11 16:25:23 -04001339}
1340
Chuck Leverdd456472006-12-05 16:35:44 -05001341static inline void xs_tcp_read_discard(struct sock_xprt *transport, struct xdr_skb_reader *desc)
Chuck Levera246b012005-08-11 16:25:23 -04001342{
1343 size_t len;
1344
Chuck Lever51971132006-12-05 16:35:19 -05001345 len = transport->tcp_reclen - transport->tcp_offset;
Chuck Levera246b012005-08-11 16:25:23 -04001346 if (len > desc->count)
1347 len = desc->count;
1348 desc->count -= len;
1349 desc->offset += len;
Chuck Lever51971132006-12-05 16:35:19 -05001350 transport->tcp_offset += len;
Chuck Lever46121cf2007-01-31 12:14:08 -05001351 dprintk("RPC: discarded %Zu bytes\n", len);
Chuck Lever51971132006-12-05 16:35:19 -05001352 xs_tcp_check_fraghdr(transport);
Chuck Levera246b012005-08-11 16:25:23 -04001353}
1354
Chuck Lever9903cd12005-08-11 16:25:26 -04001355static 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 -04001356{
1357 struct rpc_xprt *xprt = rd_desc->arg.data;
Chuck Lever51971132006-12-05 16:35:19 -05001358 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Leverdd456472006-12-05 16:35:44 -05001359 struct xdr_skb_reader desc = {
Chuck Levera246b012005-08-11 16:25:23 -04001360 .skb = skb,
1361 .offset = offset,
1362 .count = len,
Chuck Lever9903cd12005-08-11 16:25:26 -04001363 };
Chuck Levera246b012005-08-11 16:25:23 -04001364
Chuck Lever46121cf2007-01-31 12:14:08 -05001365 dprintk("RPC: xs_tcp_data_recv started\n");
Chuck Levera246b012005-08-11 16:25:23 -04001366 do {
1367 /* Read in a new fragment marker if necessary */
1368 /* Can we ever really expect to get completely empty fragments? */
Chuck Levere136d092006-12-05 16:35:23 -05001369 if (transport->tcp_flags & TCP_RCV_COPY_FRAGHDR) {
Chuck Lever9903cd12005-08-11 16:25:26 -04001370 xs_tcp_read_fraghdr(xprt, &desc);
Chuck Levera246b012005-08-11 16:25:23 -04001371 continue;
1372 }
1373 /* Read in the xid if necessary */
Chuck Levere136d092006-12-05 16:35:23 -05001374 if (transport->tcp_flags & TCP_RCV_COPY_XID) {
Chuck Lever51971132006-12-05 16:35:19 -05001375 xs_tcp_read_xid(transport, &desc);
Chuck Levera246b012005-08-11 16:25:23 -04001376 continue;
1377 }
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001378 /* Read in the call/reply flag */
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001379 if (transport->tcp_flags & TCP_RCV_READ_CALLDIR) {
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001380 xs_tcp_read_calldir(transport, &desc);
1381 continue;
1382 }
Chuck Levera246b012005-08-11 16:25:23 -04001383 /* Read in the request data */
Chuck Levere136d092006-12-05 16:35:23 -05001384 if (transport->tcp_flags & TCP_RCV_COPY_DATA) {
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001385 xs_tcp_read_data(xprt, &desc);
Chuck Levera246b012005-08-11 16:25:23 -04001386 continue;
1387 }
1388 /* Skip over any trailing bytes on short reads */
Chuck Lever51971132006-12-05 16:35:19 -05001389 xs_tcp_read_discard(transport, &desc);
Chuck Levera246b012005-08-11 16:25:23 -04001390 } while (desc.count);
Chuck Lever46121cf2007-01-31 12:14:08 -05001391 dprintk("RPC: xs_tcp_data_recv done\n");
Chuck Levera246b012005-08-11 16:25:23 -04001392 return len - desc.count;
1393}
1394
Chuck Lever9903cd12005-08-11 16:25:26 -04001395/**
1396 * xs_tcp_data_ready - "data ready" callback for TCP sockets
1397 * @sk: socket with data to read
1398 * @bytes: how much data to read
1399 *
1400 */
1401static void xs_tcp_data_ready(struct sock *sk, int bytes)
Chuck Levera246b012005-08-11 16:25:23 -04001402{
1403 struct rpc_xprt *xprt;
1404 read_descriptor_t rd_desc;
Trond Myklebustff2d7db2008-02-25 21:40:51 -08001405 int read;
Chuck Levera246b012005-08-11 16:25:23 -04001406
Chuck Lever46121cf2007-01-31 12:14:08 -05001407 dprintk("RPC: xs_tcp_data_ready...\n");
1408
Eric Dumazetf064af12010-09-22 12:43:39 +00001409 read_lock_bh(&sk->sk_callback_lock);
Chuck Lever9903cd12005-08-11 16:25:26 -04001410 if (!(xprt = xprt_from_sock(sk)))
Chuck Levera246b012005-08-11 16:25:23 -04001411 goto out;
Neil Brown61d0a8e2009-09-23 14:36:37 -04001412 /* Any data means we had a useful conversation, so
1413 * the we don't need to delay the next reconnect
1414 */
1415 if (xprt->reestablish_timeout)
1416 xprt->reestablish_timeout = 0;
1417
Chuck Lever9903cd12005-08-11 16:25:26 -04001418 /* We use rd_desc to pass struct xprt to xs_tcp_data_recv */
Chuck Levera246b012005-08-11 16:25:23 -04001419 rd_desc.arg.data = xprt;
Trond Myklebustff2d7db2008-02-25 21:40:51 -08001420 do {
1421 rd_desc.count = 65536;
1422 read = tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv);
1423 } while (read > 0);
Chuck Levera246b012005-08-11 16:25:23 -04001424out:
Eric Dumazetf064af12010-09-22 12:43:39 +00001425 read_unlock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001426}
1427
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001428/*
1429 * Do the equivalent of linger/linger2 handling for dealing with
1430 * broken servers that don't close the socket in a timely
1431 * fashion
1432 */
1433static void xs_tcp_schedule_linger_timeout(struct rpc_xprt *xprt,
1434 unsigned long timeout)
1435{
1436 struct sock_xprt *transport;
1437
1438 if (xprt_test_and_set_connecting(xprt))
1439 return;
1440 set_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1441 transport = container_of(xprt, struct sock_xprt, xprt);
1442 queue_delayed_work(rpciod_workqueue, &transport->connect_worker,
1443 timeout);
1444}
1445
1446static void xs_tcp_cancel_linger_timeout(struct rpc_xprt *xprt)
1447{
1448 struct sock_xprt *transport;
1449
1450 transport = container_of(xprt, struct sock_xprt, xprt);
1451
1452 if (!test_bit(XPRT_CONNECTION_ABORT, &xprt->state) ||
1453 !cancel_delayed_work(&transport->connect_worker))
1454 return;
1455 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1456 xprt_clear_connecting(xprt);
1457}
1458
Trond Myklebust4bc1e682012-10-23 17:50:07 -04001459static void xs_sock_reset_connection_flags(struct rpc_xprt *xprt)
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001460{
1461 smp_mb__before_clear_bit();
Trond Myklebusta519fc72012-09-12 16:49:15 -04001462 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1463 clear_bit(XPRT_CONNECTION_CLOSE, &xprt->state);
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001464 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
1465 clear_bit(XPRT_CLOSING, &xprt->state);
1466 smp_mb__after_clear_bit();
Trond Myklebust4bc1e682012-10-23 17:50:07 -04001467}
1468
1469static void xs_sock_mark_closed(struct rpc_xprt *xprt)
1470{
1471 xs_sock_reset_connection_flags(xprt);
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001472 /* Mark transport as closed and wake up all pending tasks */
1473 xprt_disconnect_done(xprt);
1474}
1475
Chuck Lever9903cd12005-08-11 16:25:26 -04001476/**
1477 * xs_tcp_state_change - callback to handle TCP socket state changes
1478 * @sk: socket whose state has changed
1479 *
1480 */
1481static void xs_tcp_state_change(struct sock *sk)
Chuck Levera246b012005-08-11 16:25:23 -04001482{
Chuck Lever9903cd12005-08-11 16:25:26 -04001483 struct rpc_xprt *xprt;
Chuck Levera246b012005-08-11 16:25:23 -04001484
Eric Dumazetf064af12010-09-22 12:43:39 +00001485 read_lock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001486 if (!(xprt = xprt_from_sock(sk)))
1487 goto out;
Chuck Lever46121cf2007-01-31 12:14:08 -05001488 dprintk("RPC: xs_tcp_state_change client %p...\n", xprt);
Andy Chittenden669502f2010-08-10 10:19:53 -04001489 dprintk("RPC: state %x conn %d dead %d zapped %d sk_shutdown %d\n",
Chuck Lever46121cf2007-01-31 12:14:08 -05001490 sk->sk_state, xprt_connected(xprt),
1491 sock_flag(sk, SOCK_DEAD),
Andy Chittenden669502f2010-08-10 10:19:53 -04001492 sock_flag(sk, SOCK_ZAPPED),
1493 sk->sk_shutdown);
Chuck Levera246b012005-08-11 16:25:23 -04001494
1495 switch (sk->sk_state) {
1496 case TCP_ESTABLISHED:
Eric Dumazetf064af12010-09-22 12:43:39 +00001497 spin_lock(&xprt->transport_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001498 if (!xprt_test_and_set_connected(xprt)) {
Chuck Lever51971132006-12-05 16:35:19 -05001499 struct sock_xprt *transport = container_of(xprt,
1500 struct sock_xprt, xprt);
1501
Chuck Levera246b012005-08-11 16:25:23 -04001502 /* Reset TCP record info */
Chuck Lever51971132006-12-05 16:35:19 -05001503 transport->tcp_offset = 0;
1504 transport->tcp_reclen = 0;
1505 transport->tcp_copied = 0;
Chuck Levere136d092006-12-05 16:35:23 -05001506 transport->tcp_flags =
1507 TCP_RCV_COPY_FRAGHDR | TCP_RCV_COPY_XID;
Chuck Lever51971132006-12-05 16:35:19 -05001508
Trond Myklebust2a491992009-03-11 14:38:00 -04001509 xprt_wake_pending_tasks(xprt, -EAGAIN);
Chuck Levera246b012005-08-11 16:25:23 -04001510 }
Eric Dumazetf064af12010-09-22 12:43:39 +00001511 spin_unlock(&xprt->transport_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001512 break;
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001513 case TCP_FIN_WAIT1:
1514 /* The client initiated a shutdown of the socket */
Trond Myklebust7c1d71c2008-04-17 16:52:57 -04001515 xprt->connect_cookie++;
Trond Myklebust663b8852008-01-01 18:42:12 -05001516 xprt->reestablish_timeout = 0;
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001517 set_bit(XPRT_CLOSING, &xprt->state);
1518 smp_mb__before_clear_bit();
1519 clear_bit(XPRT_CONNECTED, &xprt->state);
Trond Myklebustef803672007-12-31 16:19:17 -05001520 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001521 smp_mb__after_clear_bit();
Trond Myklebust25fe6142009-03-11 14:38:03 -04001522 xs_tcp_schedule_linger_timeout(xprt, xs_tcp_fin_timeout);
Chuck Levera246b012005-08-11 16:25:23 -04001523 break;
Trond Myklebust632e3bd2006-01-03 09:55:55 +01001524 case TCP_CLOSE_WAIT:
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001525 /* The server initiated a shutdown of the socket */
Trond Myklebust7c1d71c2008-04-17 16:52:57 -04001526 xprt->connect_cookie++;
Trond Myklebustd0bea452012-10-23 11:35:47 -04001527 clear_bit(XPRT_CONNECTED, &xprt->state);
Trond Myklebusta519fc72012-09-12 16:49:15 -04001528 xs_tcp_force_close(xprt);
Trond Myklebust663b8852008-01-01 18:42:12 -05001529 case TCP_CLOSING:
1530 /*
1531 * If the server closed down the connection, make sure that
1532 * we back off before reconnecting
1533 */
1534 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
1535 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001536 break;
1537 case TCP_LAST_ACK:
Trond Myklebust670f9452009-03-11 14:37:58 -04001538 set_bit(XPRT_CLOSING, &xprt->state);
Trond Myklebust25fe6142009-03-11 14:38:03 -04001539 xs_tcp_schedule_linger_timeout(xprt, xs_tcp_fin_timeout);
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001540 smp_mb__before_clear_bit();
1541 clear_bit(XPRT_CONNECTED, &xprt->state);
1542 smp_mb__after_clear_bit();
1543 break;
1544 case TCP_CLOSE:
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001545 xs_tcp_cancel_linger_timeout(xprt);
1546 xs_sock_mark_closed(xprt);
Chuck Levera246b012005-08-11 16:25:23 -04001547 }
1548 out:
Eric Dumazetf064af12010-09-22 12:43:39 +00001549 read_unlock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001550}
1551
Ilpo Järvinen1f0fa152009-02-06 23:48:33 -08001552static void xs_write_space(struct sock *sk)
1553{
1554 struct socket *sock;
1555 struct rpc_xprt *xprt;
1556
1557 if (unlikely(!(sock = sk->sk_socket)))
1558 return;
1559 clear_bit(SOCK_NOSPACE, &sock->flags);
1560
1561 if (unlikely(!(xprt = xprt_from_sock(sk))))
1562 return;
1563 if (test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags) == 0)
1564 return;
1565
1566 xprt_write_space(xprt);
1567}
1568
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04001569/**
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001570 * xs_udp_write_space - callback invoked when socket buffer space
1571 * becomes available
Chuck Lever9903cd12005-08-11 16:25:26 -04001572 * @sk: socket whose state has changed
1573 *
Chuck Levera246b012005-08-11 16:25:23 -04001574 * Called when more output buffer space is available for this socket.
1575 * We try not to wake our writers until they can make "significant"
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001576 * progress, otherwise we'll waste resources thrashing kernel_sendmsg
Chuck Levera246b012005-08-11 16:25:23 -04001577 * with a bunch of small requests.
1578 */
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001579static void xs_udp_write_space(struct sock *sk)
Chuck Levera246b012005-08-11 16:25:23 -04001580{
Eric Dumazetf064af12010-09-22 12:43:39 +00001581 read_lock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001582
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001583 /* from net/core/sock.c:sock_def_write_space */
Ilpo Järvinen1f0fa152009-02-06 23:48:33 -08001584 if (sock_writeable(sk))
1585 xs_write_space(sk);
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001586
Eric Dumazetf064af12010-09-22 12:43:39 +00001587 read_unlock_bh(&sk->sk_callback_lock);
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001588}
Chuck Levera246b012005-08-11 16:25:23 -04001589
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001590/**
1591 * xs_tcp_write_space - callback invoked when socket buffer space
1592 * becomes available
1593 * @sk: socket whose state has changed
1594 *
1595 * Called when more output buffer space is available for this socket.
1596 * We try not to wake our writers until they can make "significant"
1597 * progress, otherwise we'll waste resources thrashing kernel_sendmsg
1598 * with a bunch of small requests.
1599 */
1600static void xs_tcp_write_space(struct sock *sk)
1601{
Eric Dumazetf064af12010-09-22 12:43:39 +00001602 read_lock_bh(&sk->sk_callback_lock);
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001603
1604 /* from net/core/stream.c:sk_stream_write_space */
Ilpo Järvinen1f0fa152009-02-06 23:48:33 -08001605 if (sk_stream_wspace(sk) >= sk_stream_min_wspace(sk))
1606 xs_write_space(sk);
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001607
Eric Dumazetf064af12010-09-22 12:43:39 +00001608 read_unlock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001609}
1610
Chuck Lever470056c2005-08-25 16:25:56 -07001611static void xs_udp_do_set_buffer_size(struct rpc_xprt *xprt)
Chuck Levera246b012005-08-11 16:25:23 -04001612{
Chuck Leveree0ac0c2006-12-05 16:35:15 -05001613 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1614 struct sock *sk = transport->inet;
Chuck Levera246b012005-08-11 16:25:23 -04001615
Chuck Lever7c6e0662006-12-05 16:35:30 -05001616 if (transport->rcvsize) {
Chuck Levera246b012005-08-11 16:25:23 -04001617 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
Chuck Lever7c6e0662006-12-05 16:35:30 -05001618 sk->sk_rcvbuf = transport->rcvsize * xprt->max_reqs * 2;
Chuck Levera246b012005-08-11 16:25:23 -04001619 }
Chuck Lever7c6e0662006-12-05 16:35:30 -05001620 if (transport->sndsize) {
Chuck Levera246b012005-08-11 16:25:23 -04001621 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
Chuck Lever7c6e0662006-12-05 16:35:30 -05001622 sk->sk_sndbuf = transport->sndsize * xprt->max_reqs * 2;
Chuck Levera246b012005-08-11 16:25:23 -04001623 sk->sk_write_space(sk);
1624 }
1625}
1626
Chuck Lever43118c22005-08-25 16:25:49 -07001627/**
Chuck Lever470056c2005-08-25 16:25:56 -07001628 * xs_udp_set_buffer_size - set send and receive limits
Chuck Lever43118c22005-08-25 16:25:49 -07001629 * @xprt: generic transport
Chuck Lever470056c2005-08-25 16:25:56 -07001630 * @sndsize: requested size of send buffer, in bytes
1631 * @rcvsize: requested size of receive buffer, in bytes
Chuck Lever43118c22005-08-25 16:25:49 -07001632 *
Chuck Lever470056c2005-08-25 16:25:56 -07001633 * Set socket send and receive buffer size limits.
Chuck Lever43118c22005-08-25 16:25:49 -07001634 */
Chuck Lever470056c2005-08-25 16:25:56 -07001635static void xs_udp_set_buffer_size(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize)
Chuck Lever43118c22005-08-25 16:25:49 -07001636{
Chuck Lever7c6e0662006-12-05 16:35:30 -05001637 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1638
1639 transport->sndsize = 0;
Chuck Lever470056c2005-08-25 16:25:56 -07001640 if (sndsize)
Chuck Lever7c6e0662006-12-05 16:35:30 -05001641 transport->sndsize = sndsize + 1024;
1642 transport->rcvsize = 0;
Chuck Lever470056c2005-08-25 16:25:56 -07001643 if (rcvsize)
Chuck Lever7c6e0662006-12-05 16:35:30 -05001644 transport->rcvsize = rcvsize + 1024;
Chuck Lever470056c2005-08-25 16:25:56 -07001645
1646 xs_udp_do_set_buffer_size(xprt);
Chuck Lever43118c22005-08-25 16:25:49 -07001647}
1648
Chuck Lever46c0ee82005-08-25 16:25:52 -07001649/**
1650 * xs_udp_timer - called when a retransmit timeout occurs on a UDP transport
1651 * @task: task that timed out
1652 *
1653 * Adjust the congestion window after a retransmit timeout has occurred.
1654 */
Trond Myklebust6a24dfb2013-01-08 09:48:15 -05001655static void xs_udp_timer(struct rpc_xprt *xprt, struct rpc_task *task)
Chuck Lever46c0ee82005-08-25 16:25:52 -07001656{
Trond Myklebust6a24dfb2013-01-08 09:48:15 -05001657 xprt_adjust_cwnd(xprt, task, -ETIMEDOUT);
Chuck Lever46c0ee82005-08-25 16:25:52 -07001658}
1659
Chuck Leverb85d8802006-05-25 01:40:49 -04001660static unsigned short xs_get_random_port(void)
1661{
1662 unsigned short range = xprt_max_resvport - xprt_min_resvport;
1663 unsigned short rand = (unsigned short) net_random() % range;
1664 return rand + xprt_min_resvport;
1665}
1666
Chuck Lever92200412006-01-03 09:55:51 +01001667/**
1668 * xs_set_port - reset the port number in the remote endpoint address
1669 * @xprt: generic transport
1670 * @port: new port number
1671 *
1672 */
1673static void xs_set_port(struct rpc_xprt *xprt, unsigned short port)
1674{
Chuck Lever46121cf2007-01-31 12:14:08 -05001675 dprintk("RPC: setting port for xprt %p to %u\n", xprt, port);
Chuck Leverc4efcb12006-08-22 20:06:19 -04001676
Chuck Lever9dc3b092009-08-09 15:09:46 -04001677 rpc_set_port(xs_addr(xprt), port);
1678 xs_update_peer_port(xprt);
Chuck Lever92200412006-01-03 09:55:51 +01001679}
1680
Pavel Emelyanov5d4ec932010-10-04 16:51:23 +04001681static unsigned short xs_get_srcport(struct sock_xprt *transport)
Trond Myklebust67a391d2007-11-05 17:40:58 -05001682{
Chuck Leverfbfffbd2009-08-09 15:09:46 -04001683 unsigned short port = transport->srcport;
Trond Myklebust67a391d2007-11-05 17:40:58 -05001684
1685 if (port == 0 && transport->xprt.resvport)
1686 port = xs_get_random_port();
1687 return port;
1688}
1689
Pavel Emelyanovbaaf4e42010-10-04 16:51:56 +04001690static unsigned short xs_next_srcport(struct sock_xprt *transport, unsigned short port)
Trond Myklebust67a391d2007-11-05 17:40:58 -05001691{
Chuck Leverfbfffbd2009-08-09 15:09:46 -04001692 if (transport->srcport != 0)
1693 transport->srcport = 0;
Trond Myklebust67a391d2007-11-05 17:40:58 -05001694 if (!transport->xprt.resvport)
1695 return 0;
1696 if (port <= xprt_min_resvport || port > xprt_max_resvport)
1697 return xprt_max_resvport;
1698 return --port;
1699}
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001700static int xs_bind(struct sock_xprt *transport, struct socket *sock)
Chuck Levera246b012005-08-11 16:25:23 -04001701{
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001702 struct sockaddr_storage myaddr;
Trond Myklebust67a391d2007-11-05 17:40:58 -05001703 int err, nloop = 0;
Pavel Emelyanov5d4ec932010-10-04 16:51:23 +04001704 unsigned short port = xs_get_srcport(transport);
Trond Myklebust67a391d2007-11-05 17:40:58 -05001705 unsigned short last;
Chuck Levera246b012005-08-11 16:25:23 -04001706
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001707 memcpy(&myaddr, &transport->srcaddr, transport->xprt.addrlen);
Chuck Levera246b012005-08-11 16:25:23 -04001708 do {
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001709 rpc_set_port((struct sockaddr *)&myaddr, port);
1710 err = kernel_bind(sock, (struct sockaddr *)&myaddr,
1711 transport->xprt.addrlen);
Trond Myklebust67a391d2007-11-05 17:40:58 -05001712 if (port == 0)
Frank van Maarseveend3bc9a12007-07-09 22:23:35 +02001713 break;
Chuck Levera246b012005-08-11 16:25:23 -04001714 if (err == 0) {
Chuck Leverfbfffbd2009-08-09 15:09:46 -04001715 transport->srcport = port;
Frank van Maarseveend3bc9a12007-07-09 22:23:35 +02001716 break;
Chuck Levera246b012005-08-11 16:25:23 -04001717 }
Trond Myklebust67a391d2007-11-05 17:40:58 -05001718 last = port;
Pavel Emelyanovbaaf4e42010-10-04 16:51:56 +04001719 port = xs_next_srcport(transport, port);
Trond Myklebust67a391d2007-11-05 17:40:58 -05001720 if (port > last)
1721 nloop++;
1722 } while (err == -EADDRINUSE && nloop != 2);
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001723
Chuck Lever4232e862010-10-20 11:52:51 -04001724 if (myaddr.ss_family == AF_INET)
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001725 dprintk("RPC: %s %pI4:%u: %s (%d)\n", __func__,
1726 &((struct sockaddr_in *)&myaddr)->sin_addr,
1727 port, err ? "failed" : "ok", err);
1728 else
1729 dprintk("RPC: %s %pI6:%u: %s (%d)\n", __func__,
1730 &((struct sockaddr_in6 *)&myaddr)->sin6_addr,
1731 port, err ? "failed" : "ok", err);
Chuck Levera246b012005-08-11 16:25:23 -04001732 return err;
1733}
1734
Chuck Lever176e21e2011-05-09 15:22:44 -04001735/*
1736 * We don't support autobind on AF_LOCAL sockets
1737 */
1738static void xs_local_rpcbind(struct rpc_task *task)
1739{
Trond Myklebust3dc0da22013-01-08 09:31:13 -05001740 rcu_read_lock();
1741 xprt_set_bound(rcu_dereference(task->tk_client->cl_xprt));
1742 rcu_read_unlock();
Chuck Lever176e21e2011-05-09 15:22:44 -04001743}
1744
1745static void xs_local_set_port(struct rpc_xprt *xprt, unsigned short port)
1746{
1747}
Chuck Levera246b012005-08-11 16:25:23 -04001748
Peter Zijlstraed075362006-12-06 20:35:24 -08001749#ifdef CONFIG_DEBUG_LOCK_ALLOC
1750static struct lock_class_key xs_key[2];
1751static struct lock_class_key xs_slock_key[2];
1752
Chuck Lever176e21e2011-05-09 15:22:44 -04001753static inline void xs_reclassify_socketu(struct socket *sock)
1754{
1755 struct sock *sk = sock->sk;
1756
Chuck Lever176e21e2011-05-09 15:22:44 -04001757 sock_lock_init_class_and_name(sk, "slock-AF_LOCAL-RPC",
1758 &xs_slock_key[1], "sk_lock-AF_LOCAL-RPC", &xs_key[1]);
1759}
1760
Chuck Lever8945ee52007-08-06 11:58:04 -04001761static inline void xs_reclassify_socket4(struct socket *sock)
Peter Zijlstraed075362006-12-06 20:35:24 -08001762{
1763 struct sock *sk = sock->sk;
Chuck Lever8945ee52007-08-06 11:58:04 -04001764
Chuck Lever8945ee52007-08-06 11:58:04 -04001765 sock_lock_init_class_and_name(sk, "slock-AF_INET-RPC",
1766 &xs_slock_key[0], "sk_lock-AF_INET-RPC", &xs_key[0]);
1767}
Peter Zijlstraed075362006-12-06 20:35:24 -08001768
Chuck Lever8945ee52007-08-06 11:58:04 -04001769static inline void xs_reclassify_socket6(struct socket *sock)
1770{
1771 struct sock *sk = sock->sk;
Peter Zijlstraed075362006-12-06 20:35:24 -08001772
Chuck Lever8945ee52007-08-06 11:58:04 -04001773 sock_lock_init_class_and_name(sk, "slock-AF_INET6-RPC",
1774 &xs_slock_key[1], "sk_lock-AF_INET6-RPC", &xs_key[1]);
Peter Zijlstraed075362006-12-06 20:35:24 -08001775}
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001776
1777static inline void xs_reclassify_socket(int family, struct socket *sock)
1778{
Weston Andros Adamson1b7a1812012-10-23 10:43:39 -04001779 WARN_ON_ONCE(sock_owned_by_user(sock->sk));
1780 if (sock_owned_by_user(sock->sk))
1781 return;
1782
Chuck Lever4232e862010-10-20 11:52:51 -04001783 switch (family) {
Chuck Lever176e21e2011-05-09 15:22:44 -04001784 case AF_LOCAL:
1785 xs_reclassify_socketu(sock);
1786 break;
Chuck Lever4232e862010-10-20 11:52:51 -04001787 case AF_INET:
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001788 xs_reclassify_socket4(sock);
Chuck Lever4232e862010-10-20 11:52:51 -04001789 break;
1790 case AF_INET6:
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001791 xs_reclassify_socket6(sock);
Chuck Lever4232e862010-10-20 11:52:51 -04001792 break;
1793 }
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001794}
Peter Zijlstraed075362006-12-06 20:35:24 -08001795#else
Chuck Lever176e21e2011-05-09 15:22:44 -04001796static inline void xs_reclassify_socketu(struct socket *sock)
1797{
1798}
1799
Chuck Lever8945ee52007-08-06 11:58:04 -04001800static inline void xs_reclassify_socket4(struct socket *sock)
1801{
1802}
1803
1804static inline void xs_reclassify_socket6(struct socket *sock)
Peter Zijlstraed075362006-12-06 20:35:24 -08001805{
1806}
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001807
1808static inline void xs_reclassify_socket(int family, struct socket *sock)
1809{
1810}
Peter Zijlstraed075362006-12-06 20:35:24 -08001811#endif
1812
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001813static struct socket *xs_create_sock(struct rpc_xprt *xprt,
1814 struct sock_xprt *transport, int family, int type, int protocol)
Pavel Emelyanov22f79322010-10-04 16:54:26 +04001815{
1816 struct socket *sock;
1817 int err;
1818
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001819 err = __sock_create(xprt->xprt_net, family, type, protocol, &sock, 1);
Pavel Emelyanov22f79322010-10-04 16:54:26 +04001820 if (err < 0) {
1821 dprintk("RPC: can't create %d transport socket (%d).\n",
1822 protocol, -err);
1823 goto out;
1824 }
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001825 xs_reclassify_socket(family, sock);
Pavel Emelyanov22f79322010-10-04 16:54:26 +04001826
Ben Hutchings4cea2882011-02-22 21:54:34 +00001827 err = xs_bind(transport, sock);
1828 if (err) {
Pavel Emelyanov22f79322010-10-04 16:54:26 +04001829 sock_release(sock);
1830 goto out;
1831 }
1832
1833 return sock;
1834out:
1835 return ERR_PTR(err);
1836}
1837
Chuck Lever176e21e2011-05-09 15:22:44 -04001838static int xs_local_finish_connecting(struct rpc_xprt *xprt,
1839 struct socket *sock)
1840{
1841 struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
1842 xprt);
1843
1844 if (!transport->inet) {
1845 struct sock *sk = sock->sk;
1846
1847 write_lock_bh(&sk->sk_callback_lock);
1848
1849 xs_save_old_callbacks(transport, sk);
1850
1851 sk->sk_user_data = xprt;
1852 sk->sk_data_ready = xs_local_data_ready;
1853 sk->sk_write_space = xs_udp_write_space;
Chuck Lever176e21e2011-05-09 15:22:44 -04001854 sk->sk_allocation = GFP_ATOMIC;
1855
1856 xprt_clear_connected(xprt);
1857
1858 /* Reset to new socket */
1859 transport->sock = sock;
1860 transport->inet = sk;
1861
1862 write_unlock_bh(&sk->sk_callback_lock);
1863 }
1864
1865 /* Tell the socket layer to start connecting... */
1866 xprt->stat.connect_count++;
1867 xprt->stat.connect_start = jiffies;
1868 return kernel_connect(sock, xs_addr(xprt), xprt->addrlen, 0);
1869}
1870
1871/**
1872 * xs_local_setup_socket - create AF_LOCAL socket, connect to a local endpoint
1873 * @xprt: RPC transport to connect
1874 * @transport: socket transport to connect
1875 * @create_sock: function to create a socket of the correct type
Chuck Lever176e21e2011-05-09 15:22:44 -04001876 */
J. Bruce Fieldsdc107402013-02-20 17:52:19 -05001877static int xs_local_setup_socket(struct sock_xprt *transport)
Chuck Lever176e21e2011-05-09 15:22:44 -04001878{
Chuck Lever176e21e2011-05-09 15:22:44 -04001879 struct rpc_xprt *xprt = &transport->xprt;
1880 struct socket *sock;
1881 int status = -EIO;
1882
Jeff Layton5cf02d02012-07-23 13:58:51 -04001883 current->flags |= PF_FSTRANS;
1884
Chuck Lever176e21e2011-05-09 15:22:44 -04001885 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
1886 status = __sock_create(xprt->xprt_net, AF_LOCAL,
1887 SOCK_STREAM, 0, &sock, 1);
1888 if (status < 0) {
1889 dprintk("RPC: can't create AF_LOCAL "
1890 "transport socket (%d).\n", -status);
1891 goto out;
1892 }
1893 xs_reclassify_socketu(sock);
1894
1895 dprintk("RPC: worker connecting xprt %p via AF_LOCAL to %s\n",
1896 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1897
1898 status = xs_local_finish_connecting(xprt, sock);
1899 switch (status) {
1900 case 0:
1901 dprintk("RPC: xprt %p connected to %s\n",
1902 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1903 xprt_set_connected(xprt);
1904 break;
1905 case -ENOENT:
1906 dprintk("RPC: xprt %p: socket %s does not exist\n",
1907 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1908 break;
Trond Myklebust4a20a982012-12-15 17:02:29 -05001909 case -ECONNREFUSED:
1910 dprintk("RPC: xprt %p: connection refused for %s\n",
1911 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1912 break;
Chuck Lever176e21e2011-05-09 15:22:44 -04001913 default:
1914 printk(KERN_ERR "%s: unhandled error (%d) connecting to %s\n",
1915 __func__, -status,
1916 xprt->address_strings[RPC_DISPLAY_ADDR]);
1917 }
1918
1919out:
1920 xprt_clear_connecting(xprt);
1921 xprt_wake_pending_tasks(xprt, status);
Jeff Layton5cf02d02012-07-23 13:58:51 -04001922 current->flags &= ~PF_FSTRANS;
J. Bruce Fieldsdc107402013-02-20 17:52:19 -05001923 return status;
1924}
1925
Linus Torvaldsb6669732013-02-28 18:02:55 -08001926static void xs_local_connect(struct rpc_xprt *xprt, struct rpc_task *task)
J. Bruce Fieldsdc107402013-02-20 17:52:19 -05001927{
J. Bruce Fieldsdc107402013-02-20 17:52:19 -05001928 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1929 int ret;
1930
1931 if (RPC_IS_ASYNC(task)) {
1932 /*
1933 * We want the AF_LOCAL connect to be resolved in the
1934 * filesystem namespace of the process making the rpc
1935 * call. Thus we connect synchronously.
1936 *
1937 * If we want to support asynchronous AF_LOCAL calls,
1938 * we'll need to figure out how to pass a namespace to
1939 * connect.
1940 */
1941 rpc_exit(task, -ENOTCONN);
1942 return;
1943 }
1944 ret = xs_local_setup_socket(transport);
1945 if (ret && !RPC_IS_SOFTCONN(task))
1946 msleep_interruptible(15000);
Chuck Lever176e21e2011-05-09 15:22:44 -04001947}
1948
Mel Gormana564b8f2012-07-31 16:45:12 -07001949#ifdef CONFIG_SUNRPC_SWAP
1950static void xs_set_memalloc(struct rpc_xprt *xprt)
1951{
1952 struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
1953 xprt);
1954
1955 if (xprt->swapper)
1956 sk_set_memalloc(transport->inet);
1957}
1958
1959/**
1960 * xs_swapper - Tag this transport as being used for swap.
1961 * @xprt: transport to tag
1962 * @enable: enable/disable
1963 *
1964 */
1965int xs_swapper(struct rpc_xprt *xprt, int enable)
1966{
1967 struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
1968 xprt);
1969 int err = 0;
1970
1971 if (enable) {
1972 xprt->swapper++;
1973 xs_set_memalloc(xprt);
1974 } else if (xprt->swapper) {
1975 xprt->swapper--;
1976 sk_clear_memalloc(transport->inet);
1977 }
1978
1979 return err;
1980}
1981EXPORT_SYMBOL_GPL(xs_swapper);
1982#else
1983static void xs_set_memalloc(struct rpc_xprt *xprt)
1984{
1985}
1986#endif
1987
Chuck Lever16be2d22007-08-06 11:57:38 -04001988static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
Chuck Levera246b012005-08-11 16:25:23 -04001989{
Chuck Lever16be2d22007-08-06 11:57:38 -04001990 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Leveredb267a2006-08-22 20:06:18 -04001991
Chuck Leveree0ac0c2006-12-05 16:35:15 -05001992 if (!transport->inet) {
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001993 struct sock *sk = sock->sk;
1994
1995 write_lock_bh(&sk->sk_callback_lock);
1996
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04001997 xs_save_old_callbacks(transport, sk);
1998
Chuck Leverb0d93ad2005-08-11 16:25:53 -04001999 sk->sk_user_data = xprt;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002000 sk->sk_data_ready = xs_udp_data_ready;
2001 sk->sk_write_space = xs_udp_write_space;
2002 sk->sk_no_check = UDP_CSUM_NORCV;
Trond Myklebustb079fa7b2005-12-13 16:13:52 -05002003 sk->sk_allocation = GFP_ATOMIC;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002004
2005 xprt_set_connected(xprt);
2006
2007 /* Reset to new socket */
Chuck Leveree0ac0c2006-12-05 16:35:15 -05002008 transport->sock = sock;
2009 transport->inet = sk;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002010
Mel Gormana564b8f2012-07-31 16:45:12 -07002011 xs_set_memalloc(xprt);
2012
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002013 write_unlock_bh(&sk->sk_callback_lock);
2014 }
Chuck Lever470056c2005-08-25 16:25:56 -07002015 xs_udp_do_set_buffer_size(xprt);
Chuck Lever16be2d22007-08-06 11:57:38 -04002016}
2017
Pavel Emelyanov8c14ff22010-10-04 16:58:02 +04002018static void xs_udp_setup_socket(struct work_struct *work)
Chuck Levera246b012005-08-11 16:25:23 -04002019{
2020 struct sock_xprt *transport =
2021 container_of(work, struct sock_xprt, connect_worker.work);
2022 struct rpc_xprt *xprt = &transport->xprt;
2023 struct socket *sock = transport->sock;
Pavel Emelyanovb65c0312010-10-04 16:53:46 +04002024 int status = -EIO;
Chuck Levera246b012005-08-11 16:25:23 -04002025
Jeff Layton5cf02d02012-07-23 13:58:51 -04002026 current->flags |= PF_FSTRANS;
2027
Chuck Levera246b012005-08-11 16:25:23 -04002028 /* Start by resetting any existing state */
Chuck Leverfe315e72009-03-11 14:10:21 -04002029 xs_reset_transport(transport);
Pavel Emelyanov8c14ff22010-10-04 16:58:02 +04002030 sock = xs_create_sock(xprt, transport,
2031 xs_addr(xprt)->sa_family, SOCK_DGRAM, IPPROTO_UDP);
Pavel Emelyanovb65c0312010-10-04 16:53:46 +04002032 if (IS_ERR(sock))
Chuck Levera246b012005-08-11 16:25:23 -04002033 goto out;
Chuck Lever68e220b2007-08-06 11:57:48 -04002034
Chuck Leverc740eff2009-08-09 15:09:46 -04002035 dprintk("RPC: worker connecting xprt %p via %s to "
2036 "%s (port %s)\n", xprt,
2037 xprt->address_strings[RPC_DISPLAY_PROTO],
2038 xprt->address_strings[RPC_DISPLAY_ADDR],
2039 xprt->address_strings[RPC_DISPLAY_PORT]);
Chuck Lever68e220b2007-08-06 11:57:48 -04002040
2041 xs_udp_finish_connecting(xprt, sock);
Chuck Levera246b012005-08-11 16:25:23 -04002042 status = 0;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002043out:
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002044 xprt_clear_connecting(xprt);
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002045 xprt_wake_pending_tasks(xprt, status);
Jeff Layton5cf02d02012-07-23 13:58:51 -04002046 current->flags &= ~PF_FSTRANS;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002047}
2048
Chuck Lever3167e122005-08-25 16:25:55 -07002049/*
2050 * We need to preserve the port number so the reply cache on the server can
2051 * find our cached RPC replies when we get around to reconnecting.
2052 */
Pavel Emelyanov58dddac2010-10-04 16:52:55 +04002053static void xs_abort_connection(struct sock_xprt *transport)
Chuck Lever3167e122005-08-25 16:25:55 -07002054{
2055 int result;
Chuck Lever3167e122005-08-25 16:25:55 -07002056 struct sockaddr any;
2057
Pavel Emelyanov58dddac2010-10-04 16:52:55 +04002058 dprintk("RPC: disconnecting xprt %p to reuse port\n", transport);
Chuck Lever3167e122005-08-25 16:25:55 -07002059
2060 /*
2061 * Disconnect the transport socket by doing a connect operation
2062 * with AF_UNSPEC. This should return immediately...
2063 */
2064 memset(&any, 0, sizeof(any));
2065 any.sa_family = AF_UNSPEC;
Chuck Leveree0ac0c2006-12-05 16:35:15 -05002066 result = kernel_connect(transport->sock, &any, sizeof(any), 0);
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002067 if (!result)
Trond Myklebust4bc1e682012-10-23 17:50:07 -04002068 xs_sock_reset_connection_flags(&transport->xprt);
2069 dprintk("RPC: AF_UNSPEC connect return code %d\n", result);
Chuck Lever3167e122005-08-25 16:25:55 -07002070}
2071
Pavel Emelyanov58dddac2010-10-04 16:52:55 +04002072static void xs_tcp_reuse_connection(struct sock_xprt *transport)
Trond Myklebust40d25492009-03-11 14:37:58 -04002073{
2074 unsigned int state = transport->inet->sk_state;
2075
Andy Chittenden669502f2010-08-10 10:19:53 -04002076 if (state == TCP_CLOSE && transport->sock->state == SS_UNCONNECTED) {
2077 /* we don't need to abort the connection if the socket
2078 * hasn't undergone a shutdown
2079 */
2080 if (transport->inet->sk_shutdown == 0)
2081 return;
2082 dprintk("RPC: %s: TCP_CLOSEd and sk_shutdown set to %d\n",
2083 __func__, transport->inet->sk_shutdown);
2084 }
2085 if ((1 << state) & (TCPF_ESTABLISHED|TCPF_SYN_SENT)) {
2086 /* we don't need to abort the connection if the socket
2087 * hasn't undergone a shutdown
2088 */
2089 if (transport->inet->sk_shutdown == 0)
2090 return;
2091 dprintk("RPC: %s: ESTABLISHED/SYN_SENT "
2092 "sk_shutdown set to %d\n",
2093 __func__, transport->inet->sk_shutdown);
2094 }
Pavel Emelyanov58dddac2010-10-04 16:52:55 +04002095 xs_abort_connection(transport);
Trond Myklebust40d25492009-03-11 14:37:58 -04002096}
2097
Chuck Lever16be2d22007-08-06 11:57:38 -04002098static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002099{
Chuck Lever16be2d22007-08-06 11:57:38 -04002100 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Trond Myklebustfe19a962011-03-18 20:21:23 -04002101 int ret = -ENOTCONN;
Chuck Leveredb267a2006-08-22 20:06:18 -04002102
Chuck Leveree0ac0c2006-12-05 16:35:15 -05002103 if (!transport->inet) {
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002104 struct sock *sk = sock->sk;
2105
2106 write_lock_bh(&sk->sk_callback_lock);
2107
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04002108 xs_save_old_callbacks(transport, sk);
2109
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002110 sk->sk_user_data = xprt;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002111 sk->sk_data_ready = xs_tcp_data_ready;
2112 sk->sk_state_change = xs_tcp_state_change;
2113 sk->sk_write_space = xs_tcp_write_space;
Trond Myklebustb079fa7b2005-12-13 16:13:52 -05002114 sk->sk_allocation = GFP_ATOMIC;
Chuck Lever3167e122005-08-25 16:25:55 -07002115
2116 /* socket options */
2117 sk->sk_userlocks |= SOCK_BINDPORT_LOCK;
2118 sock_reset_flag(sk, SOCK_LINGER);
2119 tcp_sk(sk)->linger2 = 0;
2120 tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002121
2122 xprt_clear_connected(xprt);
2123
2124 /* Reset to new socket */
Chuck Leveree0ac0c2006-12-05 16:35:15 -05002125 transport->sock = sock;
2126 transport->inet = sk;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002127
2128 write_unlock_bh(&sk->sk_callback_lock);
2129 }
2130
Trond Myklebust01d37c42009-03-11 14:09:39 -04002131 if (!xprt_bound(xprt))
Trond Myklebustfe19a962011-03-18 20:21:23 -04002132 goto out;
Trond Myklebust01d37c42009-03-11 14:09:39 -04002133
Mel Gormana564b8f2012-07-31 16:45:12 -07002134 xs_set_memalloc(xprt);
2135
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002136 /* Tell the socket layer to start connecting... */
Chuck Lever262ca072006-03-20 13:44:16 -05002137 xprt->stat.connect_count++;
2138 xprt->stat.connect_start = jiffies;
Trond Myklebustfe19a962011-03-18 20:21:23 -04002139 ret = kernel_connect(sock, xs_addr(xprt), xprt->addrlen, O_NONBLOCK);
2140 switch (ret) {
2141 case 0:
2142 case -EINPROGRESS:
2143 /* SYN_SENT! */
2144 xprt->connect_cookie++;
2145 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
2146 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2147 }
2148out:
2149 return ret;
Chuck Lever16be2d22007-08-06 11:57:38 -04002150}
2151
2152/**
Trond Myklebustb61d59f2009-03-11 14:38:04 -04002153 * xs_tcp_setup_socket - create a TCP socket and connect to a remote endpoint
2154 * @xprt: RPC transport to connect
2155 * @transport: socket transport to connect
2156 * @create_sock: function to create a socket of the correct type
Chuck Lever16be2d22007-08-06 11:57:38 -04002157 *
2158 * Invoked by a work queue tasklet.
2159 */
Pavel Emelyanovcdd518d2010-10-04 16:57:40 +04002160static void xs_tcp_setup_socket(struct work_struct *work)
Chuck Lever16be2d22007-08-06 11:57:38 -04002161{
Pavel Emelyanovcdd518d2010-10-04 16:57:40 +04002162 struct sock_xprt *transport =
2163 container_of(work, struct sock_xprt, connect_worker.work);
Chuck Lever16be2d22007-08-06 11:57:38 -04002164 struct socket *sock = transport->sock;
Pavel Emelyanova9f5f0f2010-10-04 16:52:25 +04002165 struct rpc_xprt *xprt = &transport->xprt;
Trond Myklebustb61d59f2009-03-11 14:38:04 -04002166 int status = -EIO;
Chuck Lever16be2d22007-08-06 11:57:38 -04002167
Jeff Layton5cf02d02012-07-23 13:58:51 -04002168 current->flags |= PF_FSTRANS;
2169
Chuck Lever16be2d22007-08-06 11:57:38 -04002170 if (!sock) {
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002171 clear_bit(XPRT_CONNECTION_ABORT, &xprt->state);
Pavel Emelyanovcdd518d2010-10-04 16:57:40 +04002172 sock = xs_create_sock(xprt, transport,
2173 xs_addr(xprt)->sa_family, SOCK_STREAM, IPPROTO_TCP);
Trond Myklebustb61d59f2009-03-11 14:38:04 -04002174 if (IS_ERR(sock)) {
2175 status = PTR_ERR(sock);
Chuck Lever16be2d22007-08-06 11:57:38 -04002176 goto out;
2177 }
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002178 } else {
2179 int abort_and_exit;
2180
2181 abort_and_exit = test_and_clear_bit(XPRT_CONNECTION_ABORT,
2182 &xprt->state);
Chuck Lever16be2d22007-08-06 11:57:38 -04002183 /* "close" the socket, preserving the local port */
Pavel Emelyanov58dddac2010-10-04 16:52:55 +04002184 xs_tcp_reuse_connection(transport);
Chuck Lever16be2d22007-08-06 11:57:38 -04002185
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002186 if (abort_and_exit)
2187 goto out_eagain;
2188 }
2189
Chuck Leverc740eff2009-08-09 15:09:46 -04002190 dprintk("RPC: worker connecting xprt %p via %s to "
2191 "%s (port %s)\n", xprt,
2192 xprt->address_strings[RPC_DISPLAY_PROTO],
2193 xprt->address_strings[RPC_DISPLAY_ADDR],
2194 xprt->address_strings[RPC_DISPLAY_PORT]);
Chuck Lever16be2d22007-08-06 11:57:38 -04002195
2196 status = xs_tcp_finish_connecting(xprt, sock);
Chuck Lever46121cf2007-01-31 12:14:08 -05002197 dprintk("RPC: %p connect status %d connected %d sock state %d\n",
2198 xprt, -status, xprt_connected(xprt),
2199 sock->sk->sk_state);
Trond Myklebust2a491992009-03-11 14:38:00 -04002200 switch (status) {
Trond Myklebustf75e6742009-04-21 17:18:20 -04002201 default:
2202 printk("%s: connect returned unhandled error %d\n",
2203 __func__, status);
2204 case -EADDRNOTAVAIL:
2205 /* We're probably in TIME_WAIT. Get rid of existing socket,
2206 * and retry
2207 */
Trond Myklebusta519fc72012-09-12 16:49:15 -04002208 xs_tcp_force_close(xprt);
Trond Myklebust88b5ed72009-06-17 13:22:57 -07002209 break;
Trond Myklebust2a491992009-03-11 14:38:00 -04002210 case 0:
2211 case -EINPROGRESS:
2212 case -EALREADY:
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002213 xprt_clear_connecting(xprt);
Jeff Layton5cf02d02012-07-23 13:58:51 -04002214 current->flags &= ~PF_FSTRANS;
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002215 return;
Trond Myklebust9fcfe0c2010-03-02 13:06:21 -05002216 case -EINVAL:
2217 /* Happens, for instance, if the user specified a link
2218 * local IPv6 address without a scope-id.
2219 */
Trond Myklebust3ed5e2a2013-03-04 17:29:33 -05002220 case -ECONNREFUSED:
2221 case -ECONNRESET:
2222 case -ENETUNREACH:
2223 /* retry with existing socket, after a delay */
Trond Myklebust9fcfe0c2010-03-02 13:06:21 -05002224 goto out;
Chuck Levera246b012005-08-11 16:25:23 -04002225 }
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002226out_eagain:
Trond Myklebust2a491992009-03-11 14:38:00 -04002227 status = -EAGAIN;
Chuck Levera246b012005-08-11 16:25:23 -04002228out:
Chuck Lever2226feb2005-08-11 16:25:38 -04002229 xprt_clear_connecting(xprt);
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002230 xprt_wake_pending_tasks(xprt, status);
Jeff Layton5cf02d02012-07-23 13:58:51 -04002231 current->flags &= ~PF_FSTRANS;
Chuck Levera246b012005-08-11 16:25:23 -04002232}
2233
Chuck Lever68e220b2007-08-06 11:57:48 -04002234/**
Chuck Lever9903cd12005-08-11 16:25:26 -04002235 * xs_connect - connect a socket to a remote endpoint
Trond Myklebust1b092092013-01-08 09:26:49 -05002236 * @xprt: pointer to transport structure
Chuck Lever9903cd12005-08-11 16:25:26 -04002237 * @task: address of RPC task that manages state of connect request
2238 *
2239 * TCP: If the remote end dropped the connection, delay reconnecting.
Chuck Lever03bf4b72005-08-25 16:25:55 -07002240 *
2241 * UDP socket connects are synchronous, but we use a work queue anyway
2242 * to guarantee that even unprivileged user processes can set up a
2243 * socket on a privileged port.
2244 *
2245 * If a UDP socket connect fails, the delay behavior here prevents
2246 * retry floods (hard mounts).
Chuck Lever9903cd12005-08-11 16:25:26 -04002247 */
Trond Myklebust1b092092013-01-08 09:26:49 -05002248static void xs_connect(struct rpc_xprt *xprt, struct rpc_task *task)
Chuck Levera246b012005-08-11 16:25:23 -04002249{
Chuck Leveree0ac0c2006-12-05 16:35:15 -05002250 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -04002251
Chuck Lever09a21c42009-12-03 15:58:56 -05002252 if (transport->sock != NULL && !RPC_IS_SOFTCONN(task)) {
Chuck Lever46121cf2007-01-31 12:14:08 -05002253 dprintk("RPC: xs_connect delayed xprt %p for %lu "
2254 "seconds\n",
Chuck Lever03bf4b72005-08-25 16:25:55 -07002255 xprt, xprt->reestablish_timeout / HZ);
Trond Myklebustc1384c92007-06-14 18:00:42 -04002256 queue_delayed_work(rpciod_workqueue,
2257 &transport->connect_worker,
2258 xprt->reestablish_timeout);
Chuck Lever03bf4b72005-08-25 16:25:55 -07002259 xprt->reestablish_timeout <<= 1;
Neil Brown61d0a8e2009-09-23 14:36:37 -04002260 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
2261 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
Chuck Lever03bf4b72005-08-25 16:25:55 -07002262 if (xprt->reestablish_timeout > XS_TCP_MAX_REEST_TO)
2263 xprt->reestablish_timeout = XS_TCP_MAX_REEST_TO;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002264 } else {
Chuck Lever46121cf2007-01-31 12:14:08 -05002265 dprintk("RPC: xs_connect scheduled xprt %p\n", xprt);
Trond Myklebustc1384c92007-06-14 18:00:42 -04002266 queue_delayed_work(rpciod_workqueue,
2267 &transport->connect_worker, 0);
Chuck Levera246b012005-08-11 16:25:23 -04002268 }
2269}
2270
Chuck Lever262ca072006-03-20 13:44:16 -05002271/**
Chuck Lever176e21e2011-05-09 15:22:44 -04002272 * xs_local_print_stats - display AF_LOCAL socket-specifc stats
2273 * @xprt: rpc_xprt struct containing statistics
2274 * @seq: output file
2275 *
2276 */
2277static void xs_local_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2278{
2279 long idle_time = 0;
2280
2281 if (xprt_connected(xprt))
2282 idle_time = (long)(jiffies - xprt->last_used) / HZ;
2283
2284 seq_printf(seq, "\txprt:\tlocal %lu %lu %lu %ld %lu %lu %lu "
Andy Adamson15a45202012-02-14 16:19:18 -05002285 "%llu %llu %lu %llu %llu\n",
Chuck Lever176e21e2011-05-09 15:22:44 -04002286 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 Lever176e21e2011-05-09 15:22:44 -04002298}
2299
2300/**
Chuck Lever262ca072006-03-20 13:44:16 -05002301 * xs_udp_print_stats - display UDP socket-specifc stats
2302 * @xprt: rpc_xprt struct containing statistics
2303 * @seq: output file
2304 *
2305 */
2306static void xs_udp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2307{
Chuck Leverc8475462006-12-05 16:35:26 -05002308 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2309
Andy Adamson15a45202012-02-14 16:19:18 -05002310 seq_printf(seq, "\txprt:\tudp %u %lu %lu %lu %lu %llu %llu "
2311 "%lu %llu %llu\n",
Chuck Leverfbfffbd2009-08-09 15:09:46 -04002312 transport->srcport,
Chuck Lever262ca072006-03-20 13:44:16 -05002313 xprt->stat.bind_count,
2314 xprt->stat.sends,
2315 xprt->stat.recvs,
2316 xprt->stat.bad_xids,
2317 xprt->stat.req_u,
Andy Adamson15a45202012-02-14 16:19:18 -05002318 xprt->stat.bklog_u,
2319 xprt->stat.max_slots,
2320 xprt->stat.sending_u,
2321 xprt->stat.pending_u);
Chuck Lever262ca072006-03-20 13:44:16 -05002322}
2323
2324/**
2325 * xs_tcp_print_stats - display TCP socket-specifc stats
2326 * @xprt: rpc_xprt struct containing statistics
2327 * @seq: output file
2328 *
2329 */
2330static void xs_tcp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2331{
Chuck Leverc8475462006-12-05 16:35:26 -05002332 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Lever262ca072006-03-20 13:44:16 -05002333 long idle_time = 0;
2334
2335 if (xprt_connected(xprt))
2336 idle_time = (long)(jiffies - xprt->last_used) / HZ;
2337
Andy Adamson15a45202012-02-14 16:19:18 -05002338 seq_printf(seq, "\txprt:\ttcp %u %lu %lu %lu %ld %lu %lu %lu "
2339 "%llu %llu %lu %llu %llu\n",
Chuck Leverfbfffbd2009-08-09 15:09:46 -04002340 transport->srcport,
Chuck Lever262ca072006-03-20 13:44:16 -05002341 xprt->stat.bind_count,
2342 xprt->stat.connect_count,
2343 xprt->stat.connect_time,
2344 idle_time,
2345 xprt->stat.sends,
2346 xprt->stat.recvs,
2347 xprt->stat.bad_xids,
2348 xprt->stat.req_u,
Andy Adamson15a45202012-02-14 16:19:18 -05002349 xprt->stat.bklog_u,
2350 xprt->stat.max_slots,
2351 xprt->stat.sending_u,
2352 xprt->stat.pending_u);
Chuck Lever262ca072006-03-20 13:44:16 -05002353}
2354
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002355/*
2356 * Allocate a bunch of pages for a scratch buffer for the rpc code. The reason
2357 * we allocate pages instead doing a kmalloc like rpc_malloc is because we want
2358 * to use the server side send routines.
2359 */
H Hartley Sweeten5a51f132010-01-14 15:38:31 -07002360static void *bc_malloc(struct rpc_task *task, size_t size)
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002361{
2362 struct page *page;
2363 struct rpc_buffer *buf;
2364
Weston Andros Adamsonb8a13d02012-10-23 10:43:43 -04002365 WARN_ON_ONCE(size > PAGE_SIZE - sizeof(struct rpc_buffer));
2366 if (size > PAGE_SIZE - sizeof(struct rpc_buffer))
2367 return NULL;
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002368
Weston Andros Adamsonb8a13d02012-10-23 10:43:43 -04002369 page = alloc_page(GFP_KERNEL);
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002370 if (!page)
2371 return NULL;
2372
2373 buf = page_address(page);
2374 buf->len = PAGE_SIZE;
2375
2376 return buf->data;
2377}
2378
2379/*
2380 * Free the space allocated in the bc_alloc routine
2381 */
H Hartley Sweeten5a51f132010-01-14 15:38:31 -07002382static void bc_free(void *buffer)
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002383{
2384 struct rpc_buffer *buf;
2385
2386 if (!buffer)
2387 return;
2388
2389 buf = container_of(buffer, struct rpc_buffer, data);
2390 free_page((unsigned long)buf);
2391}
2392
2393/*
2394 * Use the svc_sock to send the callback. Must be called with svsk->sk_mutex
2395 * held. Borrows heavily from svc_tcp_sendto and xs_tcp_send_request.
2396 */
2397static int bc_sendto(struct rpc_rqst *req)
2398{
2399 int len;
2400 struct xdr_buf *xbufp = &req->rq_snd_buf;
2401 struct rpc_xprt *xprt = req->rq_xprt;
2402 struct sock_xprt *transport =
2403 container_of(xprt, struct sock_xprt, xprt);
2404 struct socket *sock = transport->sock;
2405 unsigned long headoff;
2406 unsigned long tailoff;
2407
Chuck Lever61677ee2011-05-09 15:22:34 -04002408 xs_encode_stream_record_marker(xbufp);
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002409
2410 tailoff = (unsigned long)xbufp->tail[0].iov_base & ~PAGE_MASK;
2411 headoff = (unsigned long)xbufp->head[0].iov_base & ~PAGE_MASK;
2412 len = svc_send_common(sock, xbufp,
2413 virt_to_page(xbufp->head[0].iov_base), headoff,
2414 xbufp->tail[0].iov_base, tailoff);
2415
2416 if (len != xbufp->len) {
2417 printk(KERN_NOTICE "Error sending entire callback!\n");
2418 len = -EAGAIN;
2419 }
2420
2421 return len;
2422}
2423
2424/*
2425 * The send routine. Borrows from svc_send
2426 */
2427static int bc_send_request(struct rpc_task *task)
2428{
2429 struct rpc_rqst *req = task->tk_rqstp;
2430 struct svc_xprt *xprt;
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002431 u32 len;
2432
2433 dprintk("sending request with xid: %08x\n", ntohl(req->rq_xid));
2434 /*
2435 * Get the server socket associated with this callback xprt
2436 */
2437 xprt = req->rq_xprt->bc_xprt;
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002438
2439 /*
2440 * Grab the mutex to serialize data as the connection is shared
2441 * with the fore channel
2442 */
2443 if (!mutex_trylock(&xprt->xpt_mutex)) {
2444 rpc_sleep_on(&xprt->xpt_bc_pending, task, NULL);
2445 if (!mutex_trylock(&xprt->xpt_mutex))
2446 return -EAGAIN;
2447 rpc_wake_up_queued_task(&xprt->xpt_bc_pending, task);
2448 }
2449 if (test_bit(XPT_DEAD, &xprt->xpt_flags))
2450 len = -ENOTCONN;
2451 else
2452 len = bc_sendto(req);
2453 mutex_unlock(&xprt->xpt_mutex);
2454
2455 if (len > 0)
2456 len = 0;
2457
2458 return len;
2459}
2460
2461/*
2462 * The close routine. Since this is client initiated, we do nothing
2463 */
2464
2465static void bc_close(struct rpc_xprt *xprt)
2466{
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002467}
2468
2469/*
2470 * The xprt destroy routine. Again, because this connection is client
2471 * initiated, we do nothing
2472 */
2473
2474static void bc_destroy(struct rpc_xprt *xprt)
2475{
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002476}
2477
Chuck Lever176e21e2011-05-09 15:22:44 -04002478static struct rpc_xprt_ops xs_local_ops = {
2479 .reserve_xprt = xprt_reserve_xprt,
2480 .release_xprt = xs_tcp_release_xprt,
Trond Myklebustf39c1bf2012-09-07 11:08:50 -04002481 .alloc_slot = xprt_alloc_slot,
Chuck Lever176e21e2011-05-09 15:22:44 -04002482 .rpcbind = xs_local_rpcbind,
2483 .set_port = xs_local_set_port,
J. Bruce Fieldsdc107402013-02-20 17:52:19 -05002484 .connect = xs_local_connect,
Chuck Lever176e21e2011-05-09 15:22:44 -04002485 .buf_alloc = rpc_malloc,
2486 .buf_free = rpc_free,
2487 .send_request = xs_local_send_request,
2488 .set_retrans_timeout = xprt_set_retrans_timeout_def,
2489 .close = xs_close,
J. Bruce Fields190b1ecf2013-03-08 13:13:08 -05002490 .destroy = xs_local_destroy,
Chuck Lever176e21e2011-05-09 15:22:44 -04002491 .print_stats = xs_local_print_stats,
2492};
2493
Chuck Lever262965f2005-08-11 16:25:56 -04002494static struct rpc_xprt_ops xs_udp_ops = {
Chuck Lever43118c22005-08-25 16:25:49 -07002495 .set_buffer_size = xs_udp_set_buffer_size,
Chuck Lever12a80462005-08-25 16:25:51 -07002496 .reserve_xprt = xprt_reserve_xprt_cong,
Chuck Lever49e9a892005-08-25 16:25:51 -07002497 .release_xprt = xprt_release_xprt_cong,
Trond Myklebustf39c1bf2012-09-07 11:08:50 -04002498 .alloc_slot = xprt_alloc_slot,
Chuck Lever45160d62007-07-01 12:13:17 -04002499 .rpcbind = rpcb_getport_async,
Chuck Lever92200412006-01-03 09:55:51 +01002500 .set_port = xs_set_port,
Chuck Lever9903cd12005-08-11 16:25:26 -04002501 .connect = xs_connect,
Chuck Lever02107142006-01-03 09:55:49 +01002502 .buf_alloc = rpc_malloc,
2503 .buf_free = rpc_free,
Chuck Lever262965f2005-08-11 16:25:56 -04002504 .send_request = xs_udp_send_request,
Chuck Leverfe3aca22005-08-25 16:25:50 -07002505 .set_retrans_timeout = xprt_set_retrans_timeout_rtt,
Chuck Lever46c0ee82005-08-25 16:25:52 -07002506 .timer = xs_udp_timer,
Chuck Levera58dd392005-08-25 16:25:53 -07002507 .release_request = xprt_release_rqst_cong,
Chuck Lever262965f2005-08-11 16:25:56 -04002508 .close = xs_close,
2509 .destroy = xs_destroy,
Chuck Lever262ca072006-03-20 13:44:16 -05002510 .print_stats = xs_udp_print_stats,
Chuck Lever262965f2005-08-11 16:25:56 -04002511};
2512
2513static struct rpc_xprt_ops xs_tcp_ops = {
Chuck Lever12a80462005-08-25 16:25:51 -07002514 .reserve_xprt = xprt_reserve_xprt,
Trond Myklebuste0ab53d2006-07-27 17:22:50 -04002515 .release_xprt = xs_tcp_release_xprt,
Trond Myklebustf39c1bf2012-09-07 11:08:50 -04002516 .alloc_slot = xprt_lock_and_alloc_slot,
Chuck Lever45160d62007-07-01 12:13:17 -04002517 .rpcbind = rpcb_getport_async,
Chuck Lever92200412006-01-03 09:55:51 +01002518 .set_port = xs_set_port,
Trond Myklebust0b9e7942010-04-16 16:41:57 -04002519 .connect = xs_connect,
Chuck Lever02107142006-01-03 09:55:49 +01002520 .buf_alloc = rpc_malloc,
2521 .buf_free = rpc_free,
Chuck Lever262965f2005-08-11 16:25:56 -04002522 .send_request = xs_tcp_send_request,
Chuck Leverfe3aca22005-08-25 16:25:50 -07002523 .set_retrans_timeout = xprt_set_retrans_timeout_def,
Trond Myklebustf75e6742009-04-21 17:18:20 -04002524 .close = xs_tcp_close,
Chuck Lever9903cd12005-08-11 16:25:26 -04002525 .destroy = xs_destroy,
Chuck Lever262ca072006-03-20 13:44:16 -05002526 .print_stats = xs_tcp_print_stats,
Chuck Levera246b012005-08-11 16:25:23 -04002527};
2528
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002529/*
2530 * The rpc_xprt_ops for the server backchannel
2531 */
2532
2533static struct rpc_xprt_ops bc_tcp_ops = {
2534 .reserve_xprt = xprt_reserve_xprt,
2535 .release_xprt = xprt_release_xprt,
Bryan Schumaker84e28a32012-09-24 13:39:01 -04002536 .alloc_slot = xprt_alloc_slot,
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002537 .buf_alloc = bc_malloc,
2538 .buf_free = bc_free,
2539 .send_request = bc_send_request,
2540 .set_retrans_timeout = xprt_set_retrans_timeout_def,
2541 .close = bc_close,
2542 .destroy = bc_destroy,
2543 .print_stats = xs_tcp_print_stats,
2544};
2545
Chuck Lever92476852010-10-20 11:53:01 -04002546static int xs_init_anyaddr(const int family, struct sockaddr *sap)
2547{
2548 static const struct sockaddr_in sin = {
2549 .sin_family = AF_INET,
2550 .sin_addr.s_addr = htonl(INADDR_ANY),
2551 };
2552 static const struct sockaddr_in6 sin6 = {
2553 .sin6_family = AF_INET6,
2554 .sin6_addr = IN6ADDR_ANY_INIT,
2555 };
2556
2557 switch (family) {
Chuck Lever176e21e2011-05-09 15:22:44 -04002558 case AF_LOCAL:
2559 break;
Chuck Lever92476852010-10-20 11:53:01 -04002560 case AF_INET:
2561 memcpy(sap, &sin, sizeof(sin));
2562 break;
2563 case AF_INET6:
2564 memcpy(sap, &sin6, sizeof(sin6));
2565 break;
2566 default:
2567 dprintk("RPC: %s: Bad address family\n", __func__);
2568 return -EAFNOSUPPORT;
2569 }
2570 return 0;
2571}
2572
\"Talpey, Thomas\3c341b0b2007-09-10 13:47:07 -04002573static struct rpc_xprt *xs_setup_xprt(struct xprt_create *args,
Trond Myklebustd9ba1312011-07-17 18:11:30 -04002574 unsigned int slot_table_size,
2575 unsigned int max_slot_table_size)
Chuck Leverc8541ec2006-10-17 14:44:27 -04002576{
2577 struct rpc_xprt *xprt;
Chuck Leverffc2e512006-12-05 16:35:11 -05002578 struct sock_xprt *new;
Chuck Leverc8541ec2006-10-17 14:44:27 -04002579
Frank van Maarseveen96802a02007-07-08 13:08:54 +02002580 if (args->addrlen > sizeof(xprt->addr)) {
Chuck Lever46121cf2007-01-31 12:14:08 -05002581 dprintk("RPC: xs_setup_xprt: address too large\n");
Chuck Leverc8541ec2006-10-17 14:44:27 -04002582 return ERR_PTR(-EBADF);
2583 }
2584
Trond Myklebustd9ba1312011-07-17 18:11:30 -04002585 xprt = xprt_alloc(args->net, sizeof(*new), slot_table_size,
2586 max_slot_table_size);
Pavel Emelyanovbd1722d2010-09-29 16:02:43 +04002587 if (xprt == NULL) {
Chuck Lever46121cf2007-01-31 12:14:08 -05002588 dprintk("RPC: xs_setup_xprt: couldn't allocate "
2589 "rpc_xprt\n");
Chuck Leverc8541ec2006-10-17 14:44:27 -04002590 return ERR_PTR(-ENOMEM);
2591 }
2592
Pavel Emelyanovbd1722d2010-09-29 16:02:43 +04002593 new = container_of(xprt, struct sock_xprt, xprt);
Frank van Maarseveen96802a02007-07-08 13:08:54 +02002594 memcpy(&xprt->addr, args->dstaddr, args->addrlen);
2595 xprt->addrlen = args->addrlen;
Frank van Maarseveend3bc9a12007-07-09 22:23:35 +02002596 if (args->srcaddr)
Chuck Leverfbfffbd2009-08-09 15:09:46 -04002597 memcpy(&new->srcaddr, args->srcaddr, args->addrlen);
Chuck Lever92476852010-10-20 11:53:01 -04002598 else {
2599 int err;
2600 err = xs_init_anyaddr(args->dstaddr->sa_family,
2601 (struct sockaddr *)&new->srcaddr);
Stanislav Kinsbursky2aa13532011-11-10 14:33:23 +03002602 if (err != 0) {
2603 xprt_free(xprt);
Chuck Lever92476852010-10-20 11:53:01 -04002604 return ERR_PTR(err);
Stanislav Kinsbursky2aa13532011-11-10 14:33:23 +03002605 }
Chuck Lever92476852010-10-20 11:53:01 -04002606 }
Chuck Leverc8541ec2006-10-17 14:44:27 -04002607
2608 return xprt;
2609}
2610
Chuck Lever176e21e2011-05-09 15:22:44 -04002611static const struct rpc_timeout xs_local_default_timeout = {
2612 .to_initval = 10 * HZ,
2613 .to_maxval = 10 * HZ,
2614 .to_retries = 2,
2615};
2616
2617/**
2618 * xs_setup_local - Set up transport to use an AF_LOCAL socket
2619 * @args: rpc transport creation arguments
2620 *
2621 * AF_LOCAL is a "tpi_cots_ord" transport, just like TCP
2622 */
2623static struct rpc_xprt *xs_setup_local(struct xprt_create *args)
2624{
2625 struct sockaddr_un *sun = (struct sockaddr_un *)args->dstaddr;
2626 struct sock_xprt *transport;
2627 struct rpc_xprt *xprt;
2628 struct rpc_xprt *ret;
2629
Trond Myklebustd9ba1312011-07-17 18:11:30 -04002630 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
2631 xprt_max_tcp_slot_table_entries);
Chuck Lever176e21e2011-05-09 15:22:44 -04002632 if (IS_ERR(xprt))
2633 return xprt;
2634 transport = container_of(xprt, struct sock_xprt, xprt);
2635
2636 xprt->prot = 0;
2637 xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
2638 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
2639
2640 xprt->bind_timeout = XS_BIND_TO;
2641 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2642 xprt->idle_timeout = XS_IDLE_DISC_TO;
2643
2644 xprt->ops = &xs_local_ops;
2645 xprt->timeout = &xs_local_default_timeout;
2646
2647 switch (sun->sun_family) {
2648 case AF_LOCAL:
2649 if (sun->sun_path[0] != '/') {
2650 dprintk("RPC: bad AF_LOCAL address: %s\n",
2651 sun->sun_path);
2652 ret = ERR_PTR(-EINVAL);
2653 goto out_err;
2654 }
2655 xprt_set_bound(xprt);
Chuck Lever176e21e2011-05-09 15:22:44 -04002656 xs_format_peer_addresses(xprt, "local", RPCBIND_NETID_LOCAL);
J. Bruce Fields7073ea82013-02-21 10:14:22 -05002657 ret = ERR_PTR(xs_local_setup_socket(transport));
2658 if (ret)
2659 goto out_err;
Chuck Lever176e21e2011-05-09 15:22:44 -04002660 break;
2661 default:
2662 ret = ERR_PTR(-EAFNOSUPPORT);
2663 goto out_err;
2664 }
2665
2666 dprintk("RPC: set up xprt to %s via AF_LOCAL\n",
2667 xprt->address_strings[RPC_DISPLAY_ADDR]);
2668
2669 if (try_module_get(THIS_MODULE))
2670 return xprt;
2671 ret = ERR_PTR(-EINVAL);
2672out_err:
2673 xprt_free(xprt);
2674 return ret;
2675}
2676
Trond Myklebust2881ae72007-12-20 16:03:54 -05002677static const struct rpc_timeout xs_udp_default_timeout = {
2678 .to_initval = 5 * HZ,
2679 .to_maxval = 30 * HZ,
2680 .to_increment = 5 * HZ,
2681 .to_retries = 5,
2682};
2683
Chuck Lever9903cd12005-08-11 16:25:26 -04002684/**
2685 * xs_setup_udp - Set up transport to use a UDP socket
Frank van Maarseveen96802a02007-07-08 13:08:54 +02002686 * @args: rpc transport creation arguments
Chuck Lever9903cd12005-08-11 16:25:26 -04002687 *
2688 */
Adrian Bunk483066d2007-10-24 18:24:02 +02002689static struct rpc_xprt *xs_setup_udp(struct xprt_create *args)
Chuck Levera246b012005-08-11 16:25:23 -04002690{
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002691 struct sockaddr *addr = args->dstaddr;
Chuck Leverc8541ec2006-10-17 14:44:27 -04002692 struct rpc_xprt *xprt;
Chuck Leverc8475462006-12-05 16:35:26 -05002693 struct sock_xprt *transport;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002694 struct rpc_xprt *ret;
Chuck Levera246b012005-08-11 16:25:23 -04002695
Trond Myklebustd9ba1312011-07-17 18:11:30 -04002696 xprt = xs_setup_xprt(args, xprt_udp_slot_table_entries,
2697 xprt_udp_slot_table_entries);
Chuck Leverc8541ec2006-10-17 14:44:27 -04002698 if (IS_ERR(xprt))
2699 return xprt;
Chuck Leverc8475462006-12-05 16:35:26 -05002700 transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -04002701
Chuck Leverec739ef2006-08-22 20:06:15 -04002702 xprt->prot = IPPROTO_UDP;
Chuck Lever808012f2005-08-25 16:25:49 -07002703 xprt->tsh_size = 0;
Chuck Levera246b012005-08-11 16:25:23 -04002704 /* XXX: header size can vary due to auth type, IPv6, etc. */
2705 xprt->max_payload = (1U << 16) - (MAX_HEADER << 3);
2706
Chuck Lever03bf4b72005-08-25 16:25:55 -07002707 xprt->bind_timeout = XS_BIND_TO;
Chuck Lever03bf4b72005-08-25 16:25:55 -07002708 xprt->reestablish_timeout = XS_UDP_REEST_TO;
2709 xprt->idle_timeout = XS_IDLE_DISC_TO;
Chuck Levera246b012005-08-11 16:25:23 -04002710
Chuck Lever262965f2005-08-11 16:25:56 -04002711 xprt->ops = &xs_udp_ops;
Chuck Levera246b012005-08-11 16:25:23 -04002712
Trond Myklebustba7392b2007-12-20 16:03:55 -05002713 xprt->timeout = &xs_udp_default_timeout;
Chuck Levera246b012005-08-11 16:25:23 -04002714
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002715 switch (addr->sa_family) {
2716 case AF_INET:
2717 if (((struct sockaddr_in *)addr)->sin_port != htons(0))
2718 xprt_set_bound(xprt);
2719
2720 INIT_DELAYED_WORK(&transport->connect_worker,
Pavel Emelyanov8c14ff22010-10-04 16:58:02 +04002721 xs_udp_setup_socket);
Chuck Lever9dc3b092009-08-09 15:09:46 -04002722 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP);
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002723 break;
2724 case AF_INET6:
2725 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
2726 xprt_set_bound(xprt);
2727
2728 INIT_DELAYED_WORK(&transport->connect_worker,
Pavel Emelyanov8c14ff22010-10-04 16:58:02 +04002729 xs_udp_setup_socket);
Chuck Lever9dc3b092009-08-09 15:09:46 -04002730 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP6);
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002731 break;
2732 default:
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002733 ret = ERR_PTR(-EAFNOSUPPORT);
2734 goto out_err;
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002735 }
2736
Chuck Leverc740eff2009-08-09 15:09:46 -04002737 if (xprt_bound(xprt))
2738 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
2739 xprt->address_strings[RPC_DISPLAY_ADDR],
2740 xprt->address_strings[RPC_DISPLAY_PORT],
2741 xprt->address_strings[RPC_DISPLAY_PROTO]);
2742 else
2743 dprintk("RPC: set up xprt to %s (autobind) via %s\n",
2744 xprt->address_strings[RPC_DISPLAY_ADDR],
2745 xprt->address_strings[RPC_DISPLAY_PROTO]);
Chuck Leveredb267a2006-08-22 20:06:18 -04002746
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002747 if (try_module_get(THIS_MODULE))
2748 return xprt;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002749 ret = ERR_PTR(-EINVAL);
2750out_err:
Pavel Emelyanove204e622010-09-29 16:03:13 +04002751 xprt_free(xprt);
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002752 return ret;
Chuck Levera246b012005-08-11 16:25:23 -04002753}
2754
Trond Myklebust2881ae72007-12-20 16:03:54 -05002755static const struct rpc_timeout xs_tcp_default_timeout = {
2756 .to_initval = 60 * HZ,
2757 .to_maxval = 60 * HZ,
2758 .to_retries = 2,
2759};
2760
Chuck Lever9903cd12005-08-11 16:25:26 -04002761/**
2762 * xs_setup_tcp - Set up transport to use a TCP socket
Frank van Maarseveen96802a02007-07-08 13:08:54 +02002763 * @args: rpc transport creation arguments
Chuck Lever9903cd12005-08-11 16:25:26 -04002764 *
2765 */
Adrian Bunk483066d2007-10-24 18:24:02 +02002766static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args)
Chuck Levera246b012005-08-11 16:25:23 -04002767{
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002768 struct sockaddr *addr = args->dstaddr;
Chuck Leverc8541ec2006-10-17 14:44:27 -04002769 struct rpc_xprt *xprt;
Chuck Leverc8475462006-12-05 16:35:26 -05002770 struct sock_xprt *transport;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002771 struct rpc_xprt *ret;
Trond Myklebustb7993ce2013-04-14 11:42:00 -04002772 unsigned int max_slot_table_size = xprt_max_tcp_slot_table_entries;
2773
2774 if (args->flags & XPRT_CREATE_INFINITE_SLOTS)
2775 max_slot_table_size = RPC_MAX_SLOT_TABLE_LIMIT;
Chuck Levera246b012005-08-11 16:25:23 -04002776
Trond Myklebustd9ba1312011-07-17 18:11:30 -04002777 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
Trond Myklebustb7993ce2013-04-14 11:42:00 -04002778 max_slot_table_size);
Chuck Leverc8541ec2006-10-17 14:44:27 -04002779 if (IS_ERR(xprt))
2780 return xprt;
Chuck Leverc8475462006-12-05 16:35:26 -05002781 transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -04002782
Chuck Leverec739ef2006-08-22 20:06:15 -04002783 xprt->prot = IPPROTO_TCP;
Chuck Lever808012f2005-08-25 16:25:49 -07002784 xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
Chuck Lever808012f2005-08-25 16:25:49 -07002785 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
Chuck Levera246b012005-08-11 16:25:23 -04002786
Chuck Lever03bf4b72005-08-25 16:25:55 -07002787 xprt->bind_timeout = XS_BIND_TO;
Chuck Lever03bf4b72005-08-25 16:25:55 -07002788 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2789 xprt->idle_timeout = XS_IDLE_DISC_TO;
Chuck Levera246b012005-08-11 16:25:23 -04002790
Chuck Lever262965f2005-08-11 16:25:56 -04002791 xprt->ops = &xs_tcp_ops;
Trond Myklebustba7392b2007-12-20 16:03:55 -05002792 xprt->timeout = &xs_tcp_default_timeout;
Chuck Levera246b012005-08-11 16:25:23 -04002793
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002794 switch (addr->sa_family) {
2795 case AF_INET:
2796 if (((struct sockaddr_in *)addr)->sin_port != htons(0))
2797 xprt_set_bound(xprt);
2798
Chuck Lever9dc3b092009-08-09 15:09:46 -04002799 INIT_DELAYED_WORK(&transport->connect_worker,
Pavel Emelyanovcdd518d2010-10-04 16:57:40 +04002800 xs_tcp_setup_socket);
Chuck Lever9dc3b092009-08-09 15:09:46 -04002801 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP);
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002802 break;
2803 case AF_INET6:
2804 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
2805 xprt_set_bound(xprt);
2806
Chuck Lever9dc3b092009-08-09 15:09:46 -04002807 INIT_DELAYED_WORK(&transport->connect_worker,
Pavel Emelyanovcdd518d2010-10-04 16:57:40 +04002808 xs_tcp_setup_socket);
Chuck Lever9dc3b092009-08-09 15:09:46 -04002809 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP6);
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002810 break;
2811 default:
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002812 ret = ERR_PTR(-EAFNOSUPPORT);
2813 goto out_err;
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002814 }
2815
Chuck Leverc740eff2009-08-09 15:09:46 -04002816 if (xprt_bound(xprt))
2817 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
2818 xprt->address_strings[RPC_DISPLAY_ADDR],
2819 xprt->address_strings[RPC_DISPLAY_PORT],
2820 xprt->address_strings[RPC_DISPLAY_PROTO]);
2821 else
2822 dprintk("RPC: set up xprt to %s (autobind) via %s\n",
2823 xprt->address_strings[RPC_DISPLAY_ADDR],
2824 xprt->address_strings[RPC_DISPLAY_PROTO]);
2825
Chuck Leveredb267a2006-08-22 20:06:18 -04002826
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002827 if (try_module_get(THIS_MODULE))
2828 return xprt;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002829 ret = ERR_PTR(-EINVAL);
2830out_err:
Pavel Emelyanove204e622010-09-29 16:03:13 +04002831 xprt_free(xprt);
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002832 return ret;
Chuck Levera246b012005-08-11 16:25:23 -04002833}
Chuck Lever282b32e2006-12-05 16:35:51 -05002834
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002835/**
2836 * xs_setup_bc_tcp - Set up transport to use a TCP backchannel socket
2837 * @args: rpc transport creation arguments
2838 *
2839 */
2840static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args)
2841{
2842 struct sockaddr *addr = args->dstaddr;
2843 struct rpc_xprt *xprt;
2844 struct sock_xprt *transport;
2845 struct svc_sock *bc_sock;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002846 struct rpc_xprt *ret;
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002847
J. Bruce Fieldsf0418aa2010-12-08 13:48:19 -05002848 if (args->bc_xprt->xpt_bc_xprt) {
2849 /*
2850 * This server connection already has a backchannel
2851 * export; we can't create a new one, as we wouldn't be
2852 * able to match replies based on xid any more. So,
2853 * reuse the already-existing one:
2854 */
2855 return args->bc_xprt->xpt_bc_xprt;
2856 }
Trond Myklebustd9ba1312011-07-17 18:11:30 -04002857 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
2858 xprt_tcp_slot_table_entries);
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002859 if (IS_ERR(xprt))
2860 return xprt;
2861 transport = container_of(xprt, struct sock_xprt, xprt);
2862
2863 xprt->prot = IPPROTO_TCP;
2864 xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
2865 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
2866 xprt->timeout = &xs_tcp_default_timeout;
2867
2868 /* backchannel */
2869 xprt_set_bound(xprt);
2870 xprt->bind_timeout = 0;
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002871 xprt->reestablish_timeout = 0;
2872 xprt->idle_timeout = 0;
2873
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002874 xprt->ops = &bc_tcp_ops;
2875
2876 switch (addr->sa_family) {
2877 case AF_INET:
2878 xs_format_peer_addresses(xprt, "tcp",
2879 RPCBIND_NETID_TCP);
2880 break;
2881 case AF_INET6:
2882 xs_format_peer_addresses(xprt, "tcp",
2883 RPCBIND_NETID_TCP6);
2884 break;
2885 default:
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002886 ret = ERR_PTR(-EAFNOSUPPORT);
2887 goto out_err;
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002888 }
2889
Pavel Emelyanov50fa0d42010-10-05 20:49:35 +04002890 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
2891 xprt->address_strings[RPC_DISPLAY_ADDR],
2892 xprt->address_strings[RPC_DISPLAY_PORT],
2893 xprt->address_strings[RPC_DISPLAY_PROTO]);
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002894
2895 /*
J. Bruce Fields99de8ea2010-12-08 12:45:44 -05002896 * Once we've associated a backchannel xprt with a connection,
2897 * we want to keep it around as long as long as the connection
2898 * lasts, in case we need to start using it for a backchannel
2899 * again; this reference won't be dropped until bc_xprt is
2900 * destroyed.
2901 */
2902 xprt_get(xprt);
2903 args->bc_xprt->xpt_bc_xprt = xprt;
2904 xprt->bc_xprt = args->bc_xprt;
2905 bc_sock = container_of(args->bc_xprt, struct svc_sock, sk_xprt);
2906 transport->sock = bc_sock->sk_sock;
2907 transport->inet = bc_sock->sk_sk;
2908
2909 /*
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002910 * Since we don't want connections for the backchannel, we set
2911 * the xprt status to connected
2912 */
2913 xprt_set_connected(xprt);
2914
2915
2916 if (try_module_get(THIS_MODULE))
2917 return xprt;
J. Bruce Fields99de8ea2010-12-08 12:45:44 -05002918 xprt_put(xprt);
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002919 ret = ERR_PTR(-EINVAL);
2920out_err:
Pavel Emelyanove204e622010-09-29 16:03:13 +04002921 xprt_free(xprt);
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002922 return ret;
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002923}
2924
Chuck Lever176e21e2011-05-09 15:22:44 -04002925static struct xprt_class xs_local_transport = {
2926 .list = LIST_HEAD_INIT(xs_local_transport.list),
2927 .name = "named UNIX socket",
2928 .owner = THIS_MODULE,
2929 .ident = XPRT_TRANSPORT_LOCAL,
2930 .setup = xs_setup_local,
2931};
2932
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002933static struct xprt_class xs_udp_transport = {
2934 .list = LIST_HEAD_INIT(xs_udp_transport.list),
2935 .name = "udp",
2936 .owner = THIS_MODULE,
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002937 .ident = XPRT_TRANSPORT_UDP,
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002938 .setup = xs_setup_udp,
2939};
2940
2941static struct xprt_class xs_tcp_transport = {
2942 .list = LIST_HEAD_INIT(xs_tcp_transport.list),
2943 .name = "tcp",
2944 .owner = THIS_MODULE,
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002945 .ident = XPRT_TRANSPORT_TCP,
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002946 .setup = xs_setup_tcp,
2947};
2948
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002949static struct xprt_class xs_bc_tcp_transport = {
2950 .list = LIST_HEAD_INIT(xs_bc_tcp_transport.list),
2951 .name = "tcp NFSv4.1 backchannel",
2952 .owner = THIS_MODULE,
2953 .ident = XPRT_TRANSPORT_BC_TCP,
2954 .setup = xs_setup_bc_tcp,
2955};
2956
Chuck Lever282b32e2006-12-05 16:35:51 -05002957/**
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002958 * init_socket_xprt - set up xprtsock's sysctls, register with RPC client
Chuck Lever282b32e2006-12-05 16:35:51 -05002959 *
2960 */
2961int init_socket_xprt(void)
2962{
Chuck Leverfbf76682006-12-05 16:35:54 -05002963#ifdef RPC_DEBUG
Eric W. Biederman2b1bec52007-02-14 00:33:24 -08002964 if (!sunrpc_table_header)
Eric W. Biederman0b4d4142007-02-14 00:34:09 -08002965 sunrpc_table_header = register_sysctl_table(sunrpc_table);
Chuck Leverfbf76682006-12-05 16:35:54 -05002966#endif
2967
Chuck Lever176e21e2011-05-09 15:22:44 -04002968 xprt_register_transport(&xs_local_transport);
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002969 xprt_register_transport(&xs_udp_transport);
2970 xprt_register_transport(&xs_tcp_transport);
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002971 xprt_register_transport(&xs_bc_tcp_transport);
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002972
Chuck Lever282b32e2006-12-05 16:35:51 -05002973 return 0;
2974}
2975
2976/**
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002977 * cleanup_socket_xprt - remove xprtsock's sysctls, unregister
Chuck Lever282b32e2006-12-05 16:35:51 -05002978 *
2979 */
2980void cleanup_socket_xprt(void)
2981{
Chuck Leverfbf76682006-12-05 16:35:54 -05002982#ifdef RPC_DEBUG
2983 if (sunrpc_table_header) {
2984 unregister_sysctl_table(sunrpc_table_header);
2985 sunrpc_table_header = NULL;
2986 }
2987#endif
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002988
Chuck Lever176e21e2011-05-09 15:22:44 -04002989 xprt_unregister_transport(&xs_local_transport);
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002990 xprt_unregister_transport(&xs_udp_transport);
2991 xprt_unregister_transport(&xs_tcp_transport);
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002992 xprt_unregister_transport(&xs_bc_tcp_transport);
Chuck Lever282b32e2006-12-05 16:35:51 -05002993}
Trond Myklebustcbf11072009-08-09 15:06:19 -04002994
Rusty Russell9bbb9e52010-08-11 23:04:12 -06002995static int param_set_uint_minmax(const char *val,
2996 const struct kernel_param *kp,
Trond Myklebustcbf11072009-08-09 15:06:19 -04002997 unsigned int min, unsigned int max)
2998{
2999 unsigned long num;
3000 int ret;
3001
3002 if (!val)
3003 return -EINVAL;
3004 ret = strict_strtoul(val, 0, &num);
3005 if (ret == -EINVAL || num < min || num > max)
3006 return -EINVAL;
3007 *((unsigned int *)kp->arg) = num;
3008 return 0;
3009}
3010
Rusty Russell9bbb9e52010-08-11 23:04:12 -06003011static int param_set_portnr(const char *val, const struct kernel_param *kp)
Trond Myklebustcbf11072009-08-09 15:06:19 -04003012{
3013 return param_set_uint_minmax(val, kp,
3014 RPC_MIN_RESVPORT,
3015 RPC_MAX_RESVPORT);
3016}
3017
Rusty Russell9bbb9e52010-08-11 23:04:12 -06003018static struct kernel_param_ops param_ops_portnr = {
3019 .set = param_set_portnr,
3020 .get = param_get_uint,
3021};
3022
Trond Myklebustcbf11072009-08-09 15:06:19 -04003023#define param_check_portnr(name, p) \
3024 __param_check(name, p, unsigned int);
3025
3026module_param_named(min_resvport, xprt_min_resvport, portnr, 0644);
3027module_param_named(max_resvport, xprt_max_resvport, portnr, 0644);
3028
Rusty Russell9bbb9e52010-08-11 23:04:12 -06003029static int param_set_slot_table_size(const char *val,
3030 const struct kernel_param *kp)
Trond Myklebustcbf11072009-08-09 15:06:19 -04003031{
3032 return param_set_uint_minmax(val, kp,
3033 RPC_MIN_SLOT_TABLE,
3034 RPC_MAX_SLOT_TABLE);
3035}
3036
Rusty Russell9bbb9e52010-08-11 23:04:12 -06003037static struct kernel_param_ops param_ops_slot_table_size = {
3038 .set = param_set_slot_table_size,
3039 .get = param_get_uint,
3040};
3041
Trond Myklebustcbf11072009-08-09 15:06:19 -04003042#define param_check_slot_table_size(name, p) \
3043 __param_check(name, p, unsigned int);
3044
Trond Myklebustd9ba1312011-07-17 18:11:30 -04003045static int param_set_max_slot_table_size(const char *val,
3046 const struct kernel_param *kp)
3047{
3048 return param_set_uint_minmax(val, kp,
3049 RPC_MIN_SLOT_TABLE,
3050 RPC_MAX_SLOT_TABLE_LIMIT);
3051}
3052
3053static struct kernel_param_ops param_ops_max_slot_table_size = {
3054 .set = param_set_max_slot_table_size,
3055 .get = param_get_uint,
3056};
3057
3058#define param_check_max_slot_table_size(name, p) \
3059 __param_check(name, p, unsigned int);
3060
Trond Myklebustcbf11072009-08-09 15:06:19 -04003061module_param_named(tcp_slot_table_entries, xprt_tcp_slot_table_entries,
3062 slot_table_size, 0644);
Trond Myklebustd9ba1312011-07-17 18:11:30 -04003063module_param_named(tcp_max_slot_table_entries, xprt_max_tcp_slot_table_entries,
3064 max_slot_table_size, 0644);
Trond Myklebustcbf11072009-08-09 15:06:19 -04003065module_param_named(udp_slot_table_entries, xprt_udp_slot_table_entries,
3066 slot_table_size, 0644);
3067