blob: 65e759569e4873619735b966cd055b9d634d2d3a [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
Trond Myklebust40b5ea02013-09-04 12:16:23 -040050#include <trace/events/sunrpc.h>
51
Rahul Iyer4cfc7e62009-09-10 17:32:28 +030052#include "sunrpc.h"
Chuck Lever176e21e2011-05-09 15:22:44 -040053
54static void xs_close(struct rpc_xprt *xprt);
55
Chuck Lever9903cd12005-08-11 16:25:26 -040056/*
Chuck Leverc556b752005-11-01 12:24:48 -050057 * xprtsock tunables
58 */
Trond Myklebust09acfea2012-03-11 15:22:54 -040059static unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE;
60static unsigned int xprt_tcp_slot_table_entries = RPC_MIN_SLOT_TABLE;
61static unsigned int xprt_max_tcp_slot_table_entries = RPC_MAX_SLOT_TABLE;
Chuck Leverc556b752005-11-01 12:24:48 -050062
Trond Myklebust09acfea2012-03-11 15:22:54 -040063static unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT;
64static unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT;
Chuck Leverc556b752005-11-01 12:24:48 -050065
Trond Myklebust54c09872015-02-09 11:01:02 -050066#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
67
Trond Myklebust7d1e8252009-03-11 14:38:03 -040068#define XS_TCP_LINGER_TO (15U * HZ)
Trond Myklebust25fe6142009-03-11 14:38:03 -040069static unsigned int xs_tcp_fin_timeout __read_mostly = XS_TCP_LINGER_TO;
Trond Myklebust7d1e8252009-03-11 14:38:03 -040070
Chuck Leverc556b752005-11-01 12:24:48 -050071/*
Chuck Leverfbf76682006-12-05 16:35:54 -050072 * We can register our own files under /proc/sys/sunrpc by
73 * calling register_sysctl_table() again. The files in that
74 * directory become the union of all files registered there.
75 *
76 * We simply need to make sure that we don't collide with
77 * someone else's file names!
78 */
79
Chuck Leverfbf76682006-12-05 16:35:54 -050080static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE;
81static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE;
Trond Myklebustd9ba1312011-07-17 18:11:30 -040082static unsigned int max_tcp_slot_table_limit = RPC_MAX_SLOT_TABLE_LIMIT;
Chuck Leverfbf76682006-12-05 16:35:54 -050083static unsigned int xprt_min_resvport_limit = RPC_MIN_RESVPORT;
84static unsigned int xprt_max_resvport_limit = RPC_MAX_RESVPORT;
85
86static struct ctl_table_header *sunrpc_table_header;
87
88/*
89 * FIXME: changing the UDP slot table size should also resize the UDP
90 * socket buffers for existing UDP transports
91 */
Joe Perchesfe2c6332013-06-11 23:04:25 -070092static struct ctl_table xs_tunables_table[] = {
Chuck Leverfbf76682006-12-05 16:35:54 -050093 {
Chuck Leverfbf76682006-12-05 16:35:54 -050094 .procname = "udp_slot_table_entries",
95 .data = &xprt_udp_slot_table_entries,
96 .maxlen = sizeof(unsigned int),
97 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -080098 .proc_handler = proc_dointvec_minmax,
Chuck Leverfbf76682006-12-05 16:35:54 -050099 .extra1 = &min_slot_table_size,
100 .extra2 = &max_slot_table_size
101 },
102 {
Chuck Leverfbf76682006-12-05 16:35:54 -0500103 .procname = "tcp_slot_table_entries",
104 .data = &xprt_tcp_slot_table_entries,
105 .maxlen = sizeof(unsigned int),
106 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -0800107 .proc_handler = proc_dointvec_minmax,
Chuck Leverfbf76682006-12-05 16:35:54 -0500108 .extra1 = &min_slot_table_size,
109 .extra2 = &max_slot_table_size
110 },
111 {
Trond Myklebustd9ba1312011-07-17 18:11:30 -0400112 .procname = "tcp_max_slot_table_entries",
113 .data = &xprt_max_tcp_slot_table_entries,
114 .maxlen = sizeof(unsigned int),
115 .mode = 0644,
116 .proc_handler = proc_dointvec_minmax,
117 .extra1 = &min_slot_table_size,
118 .extra2 = &max_tcp_slot_table_limit
119 },
120 {
Chuck Leverfbf76682006-12-05 16:35:54 -0500121 .procname = "min_resvport",
122 .data = &xprt_min_resvport,
123 .maxlen = sizeof(unsigned int),
124 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -0800125 .proc_handler = proc_dointvec_minmax,
Chuck Leverfbf76682006-12-05 16:35:54 -0500126 .extra1 = &xprt_min_resvport_limit,
127 .extra2 = &xprt_max_resvport_limit
128 },
129 {
Chuck Leverfbf76682006-12-05 16:35:54 -0500130 .procname = "max_resvport",
131 .data = &xprt_max_resvport,
132 .maxlen = sizeof(unsigned int),
133 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -0800134 .proc_handler = proc_dointvec_minmax,
Chuck Leverfbf76682006-12-05 16:35:54 -0500135 .extra1 = &xprt_min_resvport_limit,
136 .extra2 = &xprt_max_resvport_limit
137 },
138 {
Trond Myklebust25fe6142009-03-11 14:38:03 -0400139 .procname = "tcp_fin_timeout",
140 .data = &xs_tcp_fin_timeout,
141 .maxlen = sizeof(xs_tcp_fin_timeout),
142 .mode = 0644,
Eric W. Biederman6d456112009-11-16 03:11:48 -0800143 .proc_handler = proc_dointvec_jiffies,
Trond Myklebust25fe6142009-03-11 14:38:03 -0400144 },
Eric W. Biedermanf8572d82009-11-05 13:32:03 -0800145 { },
Chuck Leverfbf76682006-12-05 16:35:54 -0500146};
147
Joe Perchesfe2c6332013-06-11 23:04:25 -0700148static struct ctl_table sunrpc_table[] = {
Chuck Leverfbf76682006-12-05 16:35:54 -0500149 {
Chuck Leverfbf76682006-12-05 16:35:54 -0500150 .procname = "sunrpc",
151 .mode = 0555,
152 .child = xs_tunables_table
153 },
Eric W. Biedermanf8572d82009-11-05 13:32:03 -0800154 { },
Chuck Leverfbf76682006-12-05 16:35:54 -0500155};
156
157#endif
158
159/*
Chuck Lever03bf4b72005-08-25 16:25:55 -0700160 * Wait duration for a reply from the RPC portmapper.
161 */
162#define XS_BIND_TO (60U * HZ)
163
164/*
165 * Delay if a UDP socket connect error occurs. This is most likely some
166 * kind of resource problem on the local host.
167 */
168#define XS_UDP_REEST_TO (2U * HZ)
169
170/*
171 * The reestablish timeout allows clients to delay for a bit before attempting
172 * to reconnect to a server that just dropped our connection.
173 *
174 * We implement an exponential backoff when trying to reestablish a TCP
175 * transport connection with the server. Some servers like to drop a TCP
176 * connection when they are overworked, so we start with a short timeout and
177 * increase over time if the server is down or not responding.
178 */
179#define XS_TCP_INIT_REEST_TO (3U * HZ)
180#define XS_TCP_MAX_REEST_TO (5U * 60 * HZ)
181
182/*
183 * TCP idle timeout; client drops the transport socket if it is idle
184 * for this long. Note that we also timeout UDP sockets to prevent
185 * holding port numbers when there is no RPC traffic.
186 */
187#define XS_IDLE_DISC_TO (5U * 60 * HZ)
188
Jeff Laytonf895b252014-11-17 16:58:04 -0500189#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
Chuck Levera246b012005-08-11 16:25:23 -0400190# undef RPC_DEBUG_DATA
Chuck Lever9903cd12005-08-11 16:25:26 -0400191# define RPCDBG_FACILITY RPCDBG_TRANS
Chuck Levera246b012005-08-11 16:25:23 -0400192#endif
193
Chuck Levera246b012005-08-11 16:25:23 -0400194#ifdef RPC_DEBUG_DATA
Chuck Lever9903cd12005-08-11 16:25:26 -0400195static void xs_pktdump(char *msg, u32 *packet, unsigned int count)
Chuck Levera246b012005-08-11 16:25:23 -0400196{
Chuck Lever9903cd12005-08-11 16:25:26 -0400197 u8 *buf = (u8 *) packet;
198 int j;
Chuck Levera246b012005-08-11 16:25:23 -0400199
Chuck Lever46121cf2007-01-31 12:14:08 -0500200 dprintk("RPC: %s\n", msg);
Chuck Levera246b012005-08-11 16:25:23 -0400201 for (j = 0; j < count && j < 128; j += 4) {
202 if (!(j & 31)) {
203 if (j)
204 dprintk("\n");
205 dprintk("0x%04x ", j);
206 }
207 dprintk("%02x%02x%02x%02x ",
208 buf[j], buf[j+1], buf[j+2], buf[j+3]);
209 }
210 dprintk("\n");
211}
212#else
Chuck Lever9903cd12005-08-11 16:25:26 -0400213static inline void xs_pktdump(char *msg, u32 *packet, unsigned int count)
Chuck Levera246b012005-08-11 16:25:23 -0400214{
215 /* NOP */
216}
217#endif
218
Trond Myklebust21180712013-12-31 13:22:59 -0500219static inline struct rpc_xprt *xprt_from_sock(struct sock *sk)
220{
221 return (struct rpc_xprt *) sk->sk_user_data;
222}
223
Chuck Lever95392c52007-08-06 11:57:58 -0400224static inline struct sockaddr *xs_addr(struct rpc_xprt *xprt)
Chuck Leveredb267a2006-08-22 20:06:18 -0400225{
Chuck Lever95392c52007-08-06 11:57:58 -0400226 return (struct sockaddr *) &xprt->addr;
227}
228
Chuck Lever176e21e2011-05-09 15:22:44 -0400229static inline struct sockaddr_un *xs_addr_un(struct rpc_xprt *xprt)
230{
231 return (struct sockaddr_un *) &xprt->addr;
232}
233
Chuck Lever95392c52007-08-06 11:57:58 -0400234static inline struct sockaddr_in *xs_addr_in(struct rpc_xprt *xprt)
235{
236 return (struct sockaddr_in *) &xprt->addr;
237}
238
239static inline struct sockaddr_in6 *xs_addr_in6(struct rpc_xprt *xprt)
240{
241 return (struct sockaddr_in6 *) &xprt->addr;
242}
243
Chuck Leverc877b842009-08-09 15:09:36 -0400244static void xs_format_common_peer_addresses(struct rpc_xprt *xprt)
Chuck Leveredb267a2006-08-22 20:06:18 -0400245{
Chuck Leverc877b842009-08-09 15:09:36 -0400246 struct sockaddr *sap = xs_addr(xprt);
Chuck Lever9dc3b092009-08-09 15:09:46 -0400247 struct sockaddr_in6 *sin6;
248 struct sockaddr_in *sin;
Chuck Lever176e21e2011-05-09 15:22:44 -0400249 struct sockaddr_un *sun;
Chuck Leverc877b842009-08-09 15:09:36 -0400250 char buf[128];
Chuck Leveredb267a2006-08-22 20:06:18 -0400251
Chuck Lever9dc3b092009-08-09 15:09:46 -0400252 switch (sap->sa_family) {
Chuck Lever176e21e2011-05-09 15:22:44 -0400253 case AF_LOCAL:
254 sun = xs_addr_un(xprt);
255 strlcpy(buf, sun->sun_path, sizeof(buf));
256 xprt->address_strings[RPC_DISPLAY_ADDR] =
257 kstrdup(buf, GFP_KERNEL);
258 break;
Chuck Lever9dc3b092009-08-09 15:09:46 -0400259 case AF_INET:
Chuck Lever176e21e2011-05-09 15:22:44 -0400260 (void)rpc_ntop(sap, buf, sizeof(buf));
261 xprt->address_strings[RPC_DISPLAY_ADDR] =
262 kstrdup(buf, GFP_KERNEL);
Chuck Lever9dc3b092009-08-09 15:09:46 -0400263 sin = xs_addr_in(xprt);
Joe Perchesfc0b57912010-03-08 12:15:28 -0800264 snprintf(buf, sizeof(buf), "%08x", ntohl(sin->sin_addr.s_addr));
Chuck Lever9dc3b092009-08-09 15:09:46 -0400265 break;
266 case AF_INET6:
Chuck Lever176e21e2011-05-09 15:22:44 -0400267 (void)rpc_ntop(sap, buf, sizeof(buf));
268 xprt->address_strings[RPC_DISPLAY_ADDR] =
269 kstrdup(buf, GFP_KERNEL);
Chuck Lever9dc3b092009-08-09 15:09:46 -0400270 sin6 = xs_addr_in6(xprt);
Joe Perchesfc0b57912010-03-08 12:15:28 -0800271 snprintf(buf, sizeof(buf), "%pi6", &sin6->sin6_addr);
Chuck Lever9dc3b092009-08-09 15:09:46 -0400272 break;
273 default:
274 BUG();
Chuck Leveredb267a2006-08-22 20:06:18 -0400275 }
Chuck Lever176e21e2011-05-09 15:22:44 -0400276
Chuck Lever9dc3b092009-08-09 15:09:46 -0400277 xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = kstrdup(buf, GFP_KERNEL);
Chuck Leveredb267a2006-08-22 20:06:18 -0400278}
279
Chuck Lever9dc3b092009-08-09 15:09:46 -0400280static void xs_format_common_peer_ports(struct rpc_xprt *xprt)
Chuck Lever4b6473f2007-08-06 11:57:12 -0400281{
Chuck Lever9dc3b092009-08-09 15:09:46 -0400282 struct sockaddr *sap = xs_addr(xprt);
283 char buf[128];
Chuck Lever4b6473f2007-08-06 11:57:12 -0400284
Joe Perches81160e662010-03-08 12:15:59 -0800285 snprintf(buf, sizeof(buf), "%u", rpc_get_port(sap));
Chuck Leverc877b842009-08-09 15:09:36 -0400286 xprt->address_strings[RPC_DISPLAY_PORT] = kstrdup(buf, GFP_KERNEL);
Chuck Lever4b6473f2007-08-06 11:57:12 -0400287
Joe Perches81160e662010-03-08 12:15:59 -0800288 snprintf(buf, sizeof(buf), "%4hx", rpc_get_port(sap));
Chuck Leverc877b842009-08-09 15:09:36 -0400289 xprt->address_strings[RPC_DISPLAY_HEX_PORT] = kstrdup(buf, GFP_KERNEL);
290}
Chuck Lever4b6473f2007-08-06 11:57:12 -0400291
Chuck Lever9dc3b092009-08-09 15:09:46 -0400292static void xs_format_peer_addresses(struct rpc_xprt *xprt,
293 const char *protocol,
294 const char *netid)
Chuck Leveredb267a2006-08-22 20:06:18 -0400295{
Chuck Leverb454ae92008-01-07 18:34:48 -0500296 xprt->address_strings[RPC_DISPLAY_PROTO] = protocol;
Chuck Leverb454ae92008-01-07 18:34:48 -0500297 xprt->address_strings[RPC_DISPLAY_NETID] = netid;
Chuck Leverc877b842009-08-09 15:09:36 -0400298 xs_format_common_peer_addresses(xprt);
Chuck Lever9dc3b092009-08-09 15:09:46 -0400299 xs_format_common_peer_ports(xprt);
Chuck Lever4b6473f2007-08-06 11:57:12 -0400300}
301
Chuck Lever9dc3b092009-08-09 15:09:46 -0400302static void xs_update_peer_port(struct rpc_xprt *xprt)
Chuck Lever4b6473f2007-08-06 11:57:12 -0400303{
Chuck Lever9dc3b092009-08-09 15:09:46 -0400304 kfree(xprt->address_strings[RPC_DISPLAY_HEX_PORT]);
305 kfree(xprt->address_strings[RPC_DISPLAY_PORT]);
Chuck Lever4b6473f2007-08-06 11:57:12 -0400306
Chuck Lever9dc3b092009-08-09 15:09:46 -0400307 xs_format_common_peer_ports(xprt);
Chuck Leveredb267a2006-08-22 20:06:18 -0400308}
309
310static void xs_free_peer_addresses(struct rpc_xprt *xprt)
311{
Chuck Lever33e01dc2008-01-14 12:32:20 -0500312 unsigned int i;
313
314 for (i = 0; i < RPC_DISPLAY_MAX; i++)
315 switch (i) {
316 case RPC_DISPLAY_PROTO:
317 case RPC_DISPLAY_NETID:
318 continue;
319 default:
320 kfree(xprt->address_strings[i]);
321 }
Chuck Leveredb267a2006-08-22 20:06:18 -0400322}
323
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400324#define XS_SENDMSG_FLAGS (MSG_DONTWAIT | MSG_NOSIGNAL)
325
Trond Myklebust24c56842006-10-17 15:06:22 -0400326static 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 -0400327{
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400328 struct msghdr msg = {
329 .msg_name = addr,
330 .msg_namelen = addrlen,
Trond Myklebust24c56842006-10-17 15:06:22 -0400331 .msg_flags = XS_SENDMSG_FLAGS | (more ? MSG_MORE : 0),
332 };
333 struct kvec iov = {
334 .iov_base = vec->iov_base + base,
335 .iov_len = vec->iov_len - base,
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400336 };
337
Trond Myklebust24c56842006-10-17 15:06:22 -0400338 if (iov.iov_len != 0)
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400339 return kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len);
340 return kernel_sendmsg(sock, &msg, NULL, 0, 0);
341}
342
Jason Baronf279cd02014-09-24 18:08:00 +0000343static int xs_send_pagedata(struct socket *sock, struct xdr_buf *xdr, unsigned int base, int more, bool zerocopy, int *sent_p)
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400344{
Trond Myklebusta6b31d12013-11-08 16:03:50 -0500345 ssize_t (*do_sendpage)(struct socket *sock, struct page *page,
346 int offset, size_t size, int flags);
Trond Myklebust24c56842006-10-17 15:06:22 -0400347 struct page **ppage;
348 unsigned int remainder;
Jason Baronf279cd02014-09-24 18:08:00 +0000349 int err;
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400350
Trond Myklebust24c56842006-10-17 15:06:22 -0400351 remainder = xdr->page_len - base;
352 base += xdr->page_base;
353 ppage = xdr->pages + (base >> PAGE_SHIFT);
354 base &= ~PAGE_MASK;
Trond Myklebusta6b31d12013-11-08 16:03:50 -0500355 do_sendpage = sock->ops->sendpage;
356 if (!zerocopy)
357 do_sendpage = sock_no_sendpage;
Trond Myklebust24c56842006-10-17 15:06:22 -0400358 for(;;) {
359 unsigned int len = min_t(unsigned int, PAGE_SIZE - base, remainder);
360 int flags = XS_SENDMSG_FLAGS;
361
362 remainder -= len;
Trond Myklebust31303d62015-10-06 15:59:20 -0400363 if (more)
Trond Myklebust24c56842006-10-17 15:06:22 -0400364 flags |= MSG_MORE;
Trond Myklebust31303d62015-10-06 15:59:20 -0400365 if (remainder != 0)
366 flags |= MSG_SENDPAGE_NOTLAST | MSG_MORE;
Trond Myklebusta6b31d12013-11-08 16:03:50 -0500367 err = do_sendpage(sock, *ppage, base, len, flags);
Trond Myklebust24c56842006-10-17 15:06:22 -0400368 if (remainder == 0 || err != len)
369 break;
Jason Baronf279cd02014-09-24 18:08:00 +0000370 *sent_p += err;
Trond Myklebust24c56842006-10-17 15:06:22 -0400371 ppage++;
372 base = 0;
373 }
Jason Baronf279cd02014-09-24 18:08:00 +0000374 if (err > 0) {
375 *sent_p += err;
376 err = 0;
377 }
378 return err;
Chuck Leverb4b5cc82005-08-11 16:25:29 -0400379}
380
Chuck Lever9903cd12005-08-11 16:25:26 -0400381/**
382 * xs_sendpages - write pages directly to a socket
383 * @sock: socket to send on
384 * @addr: UDP only -- address of destination
385 * @addrlen: UDP only -- length of destination address
386 * @xdr: buffer containing this request
387 * @base: starting position in the buffer
Trond Myklebusta6b31d12013-11-08 16:03:50 -0500388 * @zerocopy: true if it is safe to use sendpage()
Jason Baronf279cd02014-09-24 18:08:00 +0000389 * @sent_p: return the total number of bytes successfully queued for sending
Chuck Lever9903cd12005-08-11 16:25:26 -0400390 *
Chuck Levera246b012005-08-11 16:25:23 -0400391 */
Jason Baronf279cd02014-09-24 18:08:00 +0000392static int xs_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen, struct xdr_buf *xdr, unsigned int base, bool zerocopy, int *sent_p)
Chuck Levera246b012005-08-11 16:25:23 -0400393{
Trond Myklebust24c56842006-10-17 15:06:22 -0400394 unsigned int remainder = xdr->len - base;
Jason Baronf279cd02014-09-24 18:08:00 +0000395 int err = 0;
396 int sent = 0;
Chuck Levera246b012005-08-11 16:25:23 -0400397
Chuck Lever262965f2005-08-11 16:25:56 -0400398 if (unlikely(!sock))
Trond Myklebustfba91af2009-03-11 14:06:41 -0400399 return -ENOTSOCK;
Chuck Lever262965f2005-08-11 16:25:56 -0400400
Trond Myklebust24c56842006-10-17 15:06:22 -0400401 if (base != 0) {
402 addr = NULL;
403 addrlen = 0;
404 }
Chuck Lever262965f2005-08-11 16:25:56 -0400405
Trond Myklebust24c56842006-10-17 15:06:22 -0400406 if (base < xdr->head[0].iov_len || addr != NULL) {
407 unsigned int len = xdr->head[0].iov_len - base;
408 remainder -= len;
409 err = xs_send_kvec(sock, addr, addrlen, &xdr->head[0], base, remainder != 0);
410 if (remainder == 0 || err != len)
Chuck Levera246b012005-08-11 16:25:23 -0400411 goto out;
Jason Baronf279cd02014-09-24 18:08:00 +0000412 *sent_p += err;
Chuck Levera246b012005-08-11 16:25:23 -0400413 base = 0;
414 } else
Trond Myklebust24c56842006-10-17 15:06:22 -0400415 base -= xdr->head[0].iov_len;
Chuck Levera246b012005-08-11 16:25:23 -0400416
Trond Myklebust24c56842006-10-17 15:06:22 -0400417 if (base < xdr->page_len) {
418 unsigned int len = xdr->page_len - base;
419 remainder -= len;
Jason Baronf279cd02014-09-24 18:08:00 +0000420 err = xs_send_pagedata(sock, xdr, base, remainder != 0, zerocopy, &sent);
421 *sent_p += sent;
422 if (remainder == 0 || sent != len)
Chuck Levera246b012005-08-11 16:25:23 -0400423 goto out;
424 base = 0;
Trond Myklebust24c56842006-10-17 15:06:22 -0400425 } else
426 base -= xdr->page_len;
427
428 if (base >= xdr->tail[0].iov_len)
Jason Baronf279cd02014-09-24 18:08:00 +0000429 return 0;
Trond Myklebust24c56842006-10-17 15:06:22 -0400430 err = xs_send_kvec(sock, NULL, 0, &xdr->tail[0], base, 0);
Chuck Levera246b012005-08-11 16:25:23 -0400431out:
Jason Baronf279cd02014-09-24 18:08:00 +0000432 if (err > 0) {
433 *sent_p += err;
434 err = 0;
435 }
436 return err;
Chuck Levera246b012005-08-11 16:25:23 -0400437}
438
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400439static void xs_nospace_callback(struct rpc_task *task)
440{
441 struct sock_xprt *transport = container_of(task->tk_rqstp->rq_xprt, struct sock_xprt, xprt);
442
443 transport->inet->sk_write_pending--;
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400444}
445
Chuck Lever9903cd12005-08-11 16:25:26 -0400446/**
Chuck Lever262965f2005-08-11 16:25:56 -0400447 * xs_nospace - place task on wait queue if transmit was incomplete
448 * @task: task to put to sleep
Chuck Lever9903cd12005-08-11 16:25:26 -0400449 *
Chuck Levera246b012005-08-11 16:25:23 -0400450 */
Trond Myklebust5e3771c2009-03-11 14:38:01 -0400451static int xs_nospace(struct rpc_task *task)
Chuck Levera246b012005-08-11 16:25:23 -0400452{
Chuck Lever262965f2005-08-11 16:25:56 -0400453 struct rpc_rqst *req = task->tk_rqstp;
454 struct rpc_xprt *xprt = req->rq_xprt;
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500455 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Trond Myklebust06ea0bf2014-02-11 09:15:54 -0500456 struct sock *sk = transport->inet;
Trond Myklebust24ca9a82011-11-22 14:44:28 +0200457 int ret = -EAGAIN;
Chuck Levera246b012005-08-11 16:25:23 -0400458
Chuck Lever46121cf2007-01-31 12:14:08 -0500459 dprintk("RPC: %5u xmit incomplete (%u left of %u)\n",
Chuck Lever262965f2005-08-11 16:25:56 -0400460 task->tk_pid, req->rq_slen - req->rq_bytes_sent,
461 req->rq_slen);
462
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400463 /* Protect against races with write_space */
464 spin_lock_bh(&xprt->transport_lock);
Chuck Lever262965f2005-08-11 16:25:56 -0400465
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400466 /* Don't race with disconnect */
467 if (xprt_connected(xprt)) {
Trond Myklebust13331a52016-01-06 08:57:06 -0500468 /* wait for more buffer space */
469 sk->sk_write_pending++;
470 xprt_wait_for_buffer_space(task, xs_nospace_callback);
471 } else
Trond Myklebust5e3771c2009-03-11 14:38:01 -0400472 ret = -ENOTCONN;
Chuck Lever262965f2005-08-11 16:25:56 -0400473
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400474 spin_unlock_bh(&xprt->transport_lock);
Trond Myklebust06ea0bf2014-02-11 09:15:54 -0500475
476 /* Race breaker in case memory is freed before above code is called */
477 sk->sk_write_space(sk);
Trond Myklebust5e3771c2009-03-11 14:38:01 -0400478 return ret;
Chuck Lever262965f2005-08-11 16:25:56 -0400479}
480
Chuck Lever61677ee2011-05-09 15:22:34 -0400481/*
482 * Construct a stream transport record marker in @buf.
483 */
484static inline void xs_encode_stream_record_marker(struct xdr_buf *buf)
485{
486 u32 reclen = buf->len - sizeof(rpc_fraghdr);
487 rpc_fraghdr *base = buf->head[0].iov_base;
488 *base = cpu_to_be32(RPC_LAST_STREAM_FRAGMENT | reclen);
489}
490
Chuck Lever262965f2005-08-11 16:25:56 -0400491/**
Chuck Lever176e21e2011-05-09 15:22:44 -0400492 * xs_local_send_request - write an RPC request to an AF_LOCAL socket
493 * @task: RPC task that manages the state of an RPC request
494 *
495 * Return values:
496 * 0: The request has been sent
497 * EAGAIN: The socket was blocked, please call again later to
498 * complete the request
499 * ENOTCONN: Caller needs to invoke connect logic then call again
500 * other: Some other error occured, the request was not sent
501 */
502static int xs_local_send_request(struct rpc_task *task)
503{
504 struct rpc_rqst *req = task->tk_rqstp;
505 struct rpc_xprt *xprt = req->rq_xprt;
506 struct sock_xprt *transport =
507 container_of(xprt, struct sock_xprt, xprt);
508 struct xdr_buf *xdr = &req->rq_snd_buf;
509 int status;
Jason Baronf279cd02014-09-24 18:08:00 +0000510 int sent = 0;
Chuck Lever176e21e2011-05-09 15:22:44 -0400511
512 xs_encode_stream_record_marker(&req->rq_snd_buf);
513
514 xs_pktdump("packet data:",
515 req->rq_svec->iov_base, req->rq_svec->iov_len);
516
Jason Baronf279cd02014-09-24 18:08:00 +0000517 status = xs_sendpages(transport->sock, NULL, 0, xdr, req->rq_bytes_sent,
518 true, &sent);
Chuck Lever176e21e2011-05-09 15:22:44 -0400519 dprintk("RPC: %s(%u) = %d\n",
520 __func__, xdr->len - req->rq_bytes_sent, status);
NeilBrown743c69e2015-07-27 10:55:35 +1000521
522 if (status == -EAGAIN && sock_writeable(transport->inet))
523 status = -ENOBUFS;
524
Jason Baronf279cd02014-09-24 18:08:00 +0000525 if (likely(sent > 0) || status == 0) {
526 req->rq_bytes_sent += sent;
527 req->rq_xmit_bytes_sent += sent;
Chuck Lever176e21e2011-05-09 15:22:44 -0400528 if (likely(req->rq_bytes_sent >= req->rq_slen)) {
529 req->rq_bytes_sent = 0;
530 return 0;
531 }
532 status = -EAGAIN;
533 }
534
535 switch (status) {
Trond Myklebust3601c4a2014-06-30 13:42:19 -0400536 case -ENOBUFS:
Trond Myklebustb5872f02015-07-03 09:32:23 -0400537 break;
Chuck Lever176e21e2011-05-09 15:22:44 -0400538 case -EAGAIN:
539 status = xs_nospace(task);
540 break;
541 default:
542 dprintk("RPC: sendmsg returned unrecognized error %d\n",
543 -status);
544 case -EPIPE:
545 xs_close(xprt);
546 status = -ENOTCONN;
547 }
548
549 return status;
550}
551
552/**
Chuck Lever262965f2005-08-11 16:25:56 -0400553 * xs_udp_send_request - write an RPC request to a UDP socket
554 * @task: address of RPC task that manages the state of an RPC request
555 *
556 * Return values:
557 * 0: The request has been sent
558 * EAGAIN: The socket was blocked, please call again later to
559 * complete the request
560 * ENOTCONN: Caller needs to invoke connect logic then call again
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300561 * other: Some other error occurred, the request was not sent
Chuck Lever262965f2005-08-11 16:25:56 -0400562 */
563static int xs_udp_send_request(struct rpc_task *task)
564{
565 struct rpc_rqst *req = task->tk_rqstp;
566 struct rpc_xprt *xprt = req->rq_xprt;
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500567 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Lever262965f2005-08-11 16:25:56 -0400568 struct xdr_buf *xdr = &req->rq_snd_buf;
Jason Baronf279cd02014-09-24 18:08:00 +0000569 int sent = 0;
Chuck Lever262965f2005-08-11 16:25:56 -0400570 int status;
Chuck Levera246b012005-08-11 16:25:23 -0400571
Chuck Lever9903cd12005-08-11 16:25:26 -0400572 xs_pktdump("packet data:",
Chuck Levera246b012005-08-11 16:25:23 -0400573 req->rq_svec->iov_base,
574 req->rq_svec->iov_len);
575
Trond Myklebust01d37c42009-03-11 14:09:39 -0400576 if (!xprt_bound(xprt))
577 return -ENOTCONN;
Jason Baronf279cd02014-09-24 18:08:00 +0000578 status = xs_sendpages(transport->sock, xs_addr(xprt), xprt->addrlen,
579 xdr, req->rq_bytes_sent, true, &sent);
Chuck Levera246b012005-08-11 16:25:23 -0400580
Chuck Lever46121cf2007-01-31 12:14:08 -0500581 dprintk("RPC: xs_udp_send_request(%u) = %d\n",
Chuck Lever262965f2005-08-11 16:25:56 -0400582 xdr->len - req->rq_bytes_sent, status);
Chuck Levera246b012005-08-11 16:25:23 -0400583
Jason Baron3dedbb52014-09-24 18:08:04 +0000584 /* firewall is blocking us, don't return -EAGAIN or we end up looping */
585 if (status == -EPERM)
586 goto process_status;
587
NeilBrown743c69e2015-07-27 10:55:35 +1000588 if (status == -EAGAIN && sock_writeable(transport->inet))
589 status = -ENOBUFS;
590
Jason Baronf279cd02014-09-24 18:08:00 +0000591 if (sent > 0 || status == 0) {
592 req->rq_xmit_bytes_sent += sent;
593 if (sent >= req->rq_slen)
Trond Myklebust21997002007-10-01 11:43:37 -0400594 return 0;
595 /* Still some bytes left; set up for a retry later. */
Chuck Lever262965f2005-08-11 16:25:56 -0400596 status = -EAGAIN;
Trond Myklebust21997002007-10-01 11:43:37 -0400597 }
Chuck Levera246b012005-08-11 16:25:23 -0400598
Jason Baron3dedbb52014-09-24 18:08:04 +0000599process_status:
Chuck Lever262965f2005-08-11 16:25:56 -0400600 switch (status) {
Trond Myklebustfba91af2009-03-11 14:06:41 -0400601 case -ENOTSOCK:
602 status = -ENOTCONN;
603 /* Should we call xs_close() here? */
604 break;
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400605 case -EAGAIN:
Trond Myklebust5e3771c2009-03-11 14:38:01 -0400606 status = xs_nospace(task);
Trond Myklebustb6ddf642008-04-17 18:52:19 -0400607 break;
Chuck Lever262965f2005-08-11 16:25:56 -0400608 case -ENETUNREACH:
Trond Myklebust3601c4a2014-06-30 13:42:19 -0400609 case -ENOBUFS:
Chuck Lever262965f2005-08-11 16:25:56 -0400610 case -EPIPE:
Chuck Levera246b012005-08-11 16:25:23 -0400611 case -ECONNREFUSED:
Jason Baron3dedbb52014-09-24 18:08:04 +0000612 case -EPERM:
Chuck Levera246b012005-08-11 16:25:23 -0400613 /* When the server has died, an ICMP port unreachable message
Chuck Lever9903cd12005-08-11 16:25:26 -0400614 * prompts ECONNREFUSED. */
Trond Myklebust13331a52016-01-06 08:57:06 -0500615 break;
616 default:
617 dprintk("RPC: sendmsg returned unrecognized error %d\n",
618 -status);
Chuck Levera246b012005-08-11 16:25:23 -0400619 }
Bian Naimeng5fe46e92010-03-08 14:49:01 +0800620
Chuck Lever262965f2005-08-11 16:25:56 -0400621 return status;
Chuck Levera246b012005-08-11 16:25:23 -0400622}
623
Trond Myklebuste06799f2007-11-05 15:44:12 -0500624/**
Chuck Lever262965f2005-08-11 16:25:56 -0400625 * xs_tcp_send_request - write an RPC request to a TCP socket
Chuck Lever9903cd12005-08-11 16:25:26 -0400626 * @task: address of RPC task that manages the state of an RPC request
627 *
628 * Return values:
Chuck Lever262965f2005-08-11 16:25:56 -0400629 * 0: The request has been sent
630 * EAGAIN: The socket was blocked, please call again later to
631 * complete the request
632 * ENOTCONN: Caller needs to invoke connect logic then call again
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300633 * other: Some other error occurred, the request was not sent
Chuck Lever9903cd12005-08-11 16:25:26 -0400634 *
635 * XXX: In the case of soft timeouts, should we eventually give up
Chuck Lever262965f2005-08-11 16:25:56 -0400636 * if sendmsg is not able to make progress?
Chuck Lever9903cd12005-08-11 16:25:26 -0400637 */
Chuck Lever262965f2005-08-11 16:25:56 -0400638static int xs_tcp_send_request(struct rpc_task *task)
Chuck Levera246b012005-08-11 16:25:23 -0400639{
640 struct rpc_rqst *req = task->tk_rqstp;
641 struct rpc_xprt *xprt = req->rq_xprt;
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500642 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Lever262965f2005-08-11 16:25:56 -0400643 struct xdr_buf *xdr = &req->rq_snd_buf;
Trond Myklebusta6b31d12013-11-08 16:03:50 -0500644 bool zerocopy = true;
Chuck Leverb595bb12007-08-06 11:56:42 -0400645 int status;
Jason Baronf279cd02014-09-24 18:08:00 +0000646 int sent;
Chuck Levera246b012005-08-11 16:25:23 -0400647
Chuck Lever61677ee2011-05-09 15:22:34 -0400648 xs_encode_stream_record_marker(&req->rq_snd_buf);
Chuck Levera246b012005-08-11 16:25:23 -0400649
Chuck Lever262965f2005-08-11 16:25:56 -0400650 xs_pktdump("packet data:",
651 req->rq_svec->iov_base,
652 req->rq_svec->iov_len);
Trond Myklebusta6b31d12013-11-08 16:03:50 -0500653 /* Don't use zero copy if this is a resend. If the RPC call
654 * completes while the socket holds a reference to the pages,
655 * then we may end up resending corrupted data.
656 */
657 if (task->tk_flags & RPC_TASK_SENT)
658 zerocopy = false;
Chuck Levera246b012005-08-11 16:25:23 -0400659
660 /* Continue transmitting the packet/record. We must be careful
661 * to cope with writespace callbacks arriving _after_ we have
Chuck Lever262965f2005-08-11 16:25:56 -0400662 * called sendmsg(). */
Chuck Levera246b012005-08-11 16:25:23 -0400663 while (1) {
Jason Baronf279cd02014-09-24 18:08:00 +0000664 sent = 0;
665 status = xs_sendpages(transport->sock, NULL, 0, xdr,
666 req->rq_bytes_sent, zerocopy, &sent);
Chuck Levera246b012005-08-11 16:25:23 -0400667
Chuck Lever46121cf2007-01-31 12:14:08 -0500668 dprintk("RPC: xs_tcp_send_request(%u) = %d\n",
Chuck Lever262965f2005-08-11 16:25:56 -0400669 xdr->len - req->rq_bytes_sent, status);
670
Chuck Lever262965f2005-08-11 16:25:56 -0400671 /* If we've sent the entire packet, immediately
672 * reset the count of bytes sent. */
Jason Baronf279cd02014-09-24 18:08:00 +0000673 req->rq_bytes_sent += sent;
674 req->rq_xmit_bytes_sent += sent;
Chuck Lever262965f2005-08-11 16:25:56 -0400675 if (likely(req->rq_bytes_sent >= req->rq_slen)) {
676 req->rq_bytes_sent = 0;
677 return 0;
Chuck Levera246b012005-08-11 16:25:23 -0400678 }
679
Trond Myklebustf580dd02015-07-11 17:48:52 +0200680 if (status < 0)
681 break;
682 if (sent == 0) {
683 status = -EAGAIN;
684 break;
685 }
Chuck Levera246b012005-08-11 16:25:23 -0400686 }
NeilBrown743c69e2015-07-27 10:55:35 +1000687 if (status == -EAGAIN && sk_stream_is_writeable(transport->inet))
688 status = -ENOBUFS;
Chuck Levera246b012005-08-11 16:25:23 -0400689
Chuck Lever262965f2005-08-11 16:25:56 -0400690 switch (status) {
Trond Myklebustfba91af2009-03-11 14:06:41 -0400691 case -ENOTSOCK:
692 status = -ENOTCONN;
693 /* Should we call xs_close() here? */
694 break;
Chuck Lever262965f2005-08-11 16:25:56 -0400695 case -EAGAIN:
Trond Myklebust5e3771c2009-03-11 14:38:01 -0400696 status = xs_nospace(task);
Chuck Lever262965f2005-08-11 16:25:56 -0400697 break;
Chuck Lever262965f2005-08-11 16:25:56 -0400698 case -ECONNRESET:
Chuck Lever262965f2005-08-11 16:25:56 -0400699 case -ECONNREFUSED:
700 case -ENOTCONN:
Trond Myklebust3913c782015-02-08 21:44:04 -0500701 case -EADDRINUSE:
Trond Myklebustb5872f02015-07-03 09:32:23 -0400702 case -ENOBUFS:
Trond Myklebustb9d2bb22012-10-23 11:40:02 -0400703 case -EPIPE:
Trond Myklebust13331a52016-01-06 08:57:06 -0500704 break;
705 default:
706 dprintk("RPC: sendmsg returned unrecognized error %d\n",
707 -status);
Chuck Levera246b012005-08-11 16:25:23 -0400708 }
Bian Naimeng5fe46e92010-03-08 14:49:01 +0800709
Chuck Levera246b012005-08-11 16:25:23 -0400710 return status;
711}
712
Chuck Lever9903cd12005-08-11 16:25:26 -0400713/**
Trond Myklebuste0ab53d2006-07-27 17:22:50 -0400714 * xs_tcp_release_xprt - clean up after a tcp transmission
715 * @xprt: transport
716 * @task: rpc task
717 *
718 * This cleans up if an error causes us to abort the transmission of a request.
719 * In this case, the socket may need to be reset in order to avoid confusing
720 * the server.
721 */
722static void xs_tcp_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task)
723{
724 struct rpc_rqst *req;
725
726 if (task != xprt->snd_task)
727 return;
728 if (task == NULL)
729 goto out_release;
730 req = task->tk_rqstp;
Trond Myklebust43cedbf0e2011-07-17 16:01:03 -0400731 if (req == NULL)
732 goto out_release;
Trond Myklebuste0ab53d2006-07-27 17:22:50 -0400733 if (req->rq_bytes_sent == 0)
734 goto out_release;
735 if (req->rq_bytes_sent == req->rq_snd_buf.len)
736 goto out_release;
Trond Myklebusta4f08352013-01-08 09:10:21 -0500737 set_bit(XPRT_CLOSE_WAIT, &xprt->state);
Trond Myklebuste0ab53d2006-07-27 17:22:50 -0400738out_release:
739 xprt_release_xprt(xprt, task);
740}
741
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -0400742static void xs_save_old_callbacks(struct sock_xprt *transport, struct sock *sk)
743{
744 transport->old_data_ready = sk->sk_data_ready;
745 transport->old_state_change = sk->sk_state_change;
746 transport->old_write_space = sk->sk_write_space;
Trond Myklebust21180712013-12-31 13:22:59 -0500747 transport->old_error_report = sk->sk_error_report;
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -0400748}
749
750static void xs_restore_old_callbacks(struct sock_xprt *transport, struct sock *sk)
751{
752 sk->sk_data_ready = transport->old_data_ready;
753 sk->sk_state_change = transport->old_state_change;
754 sk->sk_write_space = transport->old_write_space;
Trond Myklebust21180712013-12-31 13:22:59 -0500755 sk->sk_error_report = transport->old_error_report;
756}
757
Trond Myklebustb70ae912015-02-09 09:41:32 -0500758static void xs_sock_reset_connection_flags(struct rpc_xprt *xprt)
759{
760 smp_mb__before_atomic();
761 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
762 clear_bit(XPRT_CLOSING, &xprt->state);
763 smp_mb__after_atomic();
764}
765
766static void xs_sock_mark_closed(struct rpc_xprt *xprt)
767{
768 xs_sock_reset_connection_flags(xprt);
769 /* Mark transport as closed and wake up all pending tasks */
770 xprt_disconnect_done(xprt);
771}
772
Trond Myklebust21180712013-12-31 13:22:59 -0500773/**
774 * xs_error_report - callback to handle TCP socket state errors
775 * @sk: socket
776 *
777 * Note: we don't call sock_error() since there may be a rpc_task
778 * using the socket, and so we don't want to clear sk->sk_err.
779 */
780static void xs_error_report(struct sock *sk)
781{
782 struct rpc_xprt *xprt;
783 int err;
784
785 read_lock_bh(&sk->sk_callback_lock);
786 if (!(xprt = xprt_from_sock(sk)))
787 goto out;
788
789 err = -sk->sk_err;
790 if (err == 0)
791 goto out;
Trond Myklebustb70ae912015-02-09 09:41:32 -0500792 /* Is this a reset event? */
793 if (sk->sk_state == TCP_CLOSE)
794 xs_sock_mark_closed(xprt);
Trond Myklebust21180712013-12-31 13:22:59 -0500795 dprintk("RPC: xs_error_report client %p, error=%d...\n",
796 xprt, -err);
Trond Myklebuste8353c762013-12-31 13:39:22 -0500797 trace_rpc_socket_error(xprt, sk->sk_socket, err);
Trond Myklebust21180712013-12-31 13:22:59 -0500798 xprt_wake_pending_tasks(xprt, err);
799 out:
800 read_unlock_bh(&sk->sk_callback_lock);
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -0400801}
802
Chuck Leverfe315e72009-03-11 14:10:21 -0400803static void xs_reset_transport(struct sock_xprt *transport)
Chuck Levera246b012005-08-11 16:25:23 -0400804{
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500805 struct socket *sock = transport->sock;
806 struct sock *sk = transport->inet;
Trond Myklebust6cc7e902015-02-08 18:35:25 -0500807 struct rpc_xprt *xprt = &transport->xprt;
Chuck Levera246b012005-08-11 16:25:23 -0400808
Chuck Leverfe315e72009-03-11 14:10:21 -0400809 if (sk == NULL)
810 return;
Chuck Lever9903cd12005-08-11 16:25:26 -0400811
Jeff Layton264d1df2015-06-03 16:14:27 -0400812 if (atomic_read(&transport->xprt.swapper))
813 sk_clear_memalloc(sk);
814
Trond Myklebust09939202015-08-29 19:11:21 -0700815 kernel_sock_shutdown(sock, SHUT_RDWR);
816
Trond Myklebustedc1b012015-10-05 10:53:49 -0400817 mutex_lock(&transport->recv_mutex);
Chuck Levera246b012005-08-11 16:25:23 -0400818 write_lock_bh(&sk->sk_callback_lock);
Chuck Leveree0ac0c2006-12-05 16:35:15 -0500819 transport->inet = NULL;
820 transport->sock = NULL;
Chuck Levera246b012005-08-11 16:25:23 -0400821
Chuck Lever9903cd12005-08-11 16:25:26 -0400822 sk->sk_user_data = NULL;
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -0400823
824 xs_restore_old_callbacks(transport, sk);
Trond Myklebust0c787892015-08-29 13:36:30 -0700825 xprt_clear_connected(xprt);
Chuck Levera246b012005-08-11 16:25:23 -0400826 write_unlock_bh(&sk->sk_callback_lock);
Trond Myklebust6cc7e902015-02-08 18:35:25 -0500827 xs_sock_reset_connection_flags(xprt);
Trond Myklebustedc1b012015-10-05 10:53:49 -0400828 mutex_unlock(&transport->recv_mutex);
Chuck Levera246b012005-08-11 16:25:23 -0400829
Trond Myklebust6cc7e902015-02-08 18:35:25 -0500830 trace_rpc_socket_close(xprt, sock);
Chuck Levera246b012005-08-11 16:25:23 -0400831 sock_release(sock);
Chuck Leverfe315e72009-03-11 14:10:21 -0400832}
833
834/**
835 * xs_close - close a socket
836 * @xprt: transport
837 *
838 * This is used when all requests are complete; ie, no DRC state remains
839 * on the server we want to save.
Trond Myklebustf75e6742009-04-21 17:18:20 -0400840 *
841 * The caller _must_ be holding XPRT_LOCKED in order to avoid issues with
842 * xs_reset_transport() zeroing the socket from underneath a writer.
Chuck Leverfe315e72009-03-11 14:10:21 -0400843 */
844static void xs_close(struct rpc_xprt *xprt)
845{
846 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
847
848 dprintk("RPC: xs_close xprt %p\n", xprt);
849
850 xs_reset_transport(transport);
Neil Brown61d0a8e2009-09-23 14:36:37 -0400851 xprt->reestablish_timeout = 0;
Chuck Leverfe315e72009-03-11 14:10:21 -0400852
Trond Myklebust62da3b22007-11-06 18:44:20 -0500853 xprt_disconnect_done(xprt);
Chuck Levera246b012005-08-11 16:25:23 -0400854}
855
Chuck Lever4a068252015-05-11 14:02:25 -0400856static void xs_inject_disconnect(struct rpc_xprt *xprt)
857{
858 dprintk("RPC: injecting transport disconnect on xprt=%p\n",
859 xprt);
860 xprt_disconnect_done(xprt);
861}
862
Kinglong Mee315f3812014-03-24 11:07:22 +0800863static void xs_xprt_free(struct rpc_xprt *xprt)
864{
865 xs_free_peer_addresses(xprt);
866 xprt_free(xprt);
867}
868
Chuck Lever9903cd12005-08-11 16:25:26 -0400869/**
870 * xs_destroy - prepare to shutdown a transport
871 * @xprt: doomed transport
872 *
873 */
874static void xs_destroy(struct rpc_xprt *xprt)
Chuck Levera246b012005-08-11 16:25:23 -0400875{
Trond Myklebust03c788272015-09-17 10:42:27 -0400876 struct sock_xprt *transport = container_of(xprt,
877 struct sock_xprt, xprt);
Chuck Lever46121cf2007-01-31 12:14:08 -0500878 dprintk("RPC: xs_destroy xprt %p\n", xprt);
Chuck Lever9903cd12005-08-11 16:25:26 -0400879
Trond Myklebust03c788272015-09-17 10:42:27 -0400880 cancel_delayed_work_sync(&transport->connect_worker);
Trond Myklebusta1311d82013-10-31 09:18:49 -0400881 xs_close(xprt);
Trond Myklebustedc1b012015-10-05 10:53:49 -0400882 cancel_work_sync(&transport->recv_worker);
Kinglong Mee315f3812014-03-24 11:07:22 +0800883 xs_xprt_free(xprt);
Trond Myklebusta1311d82013-10-31 09:18:49 -0400884 module_put(THIS_MODULE);
Chuck Levera246b012005-08-11 16:25:23 -0400885}
886
Chuck Lever176e21e2011-05-09 15:22:44 -0400887static int xs_local_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb)
888{
889 struct xdr_skb_reader desc = {
890 .skb = skb,
891 .offset = sizeof(rpc_fraghdr),
892 .count = skb->len - sizeof(rpc_fraghdr),
893 };
894
895 if (xdr_partial_copy_from_skb(xdr, 0, &desc, xdr_skb_read_bits) < 0)
896 return -1;
897 if (desc.count)
898 return -1;
899 return 0;
900}
901
902/**
Trond Myklebusta2648092015-10-06 17:03:00 -0400903 * xs_local_data_read_skb
904 * @xprt: transport
905 * @sk: socket
906 * @skb: skbuff
Chuck Lever176e21e2011-05-09 15:22:44 -0400907 *
908 * Currently this assumes we can read the whole reply in a single gulp.
909 */
Trond Myklebusta2648092015-10-06 17:03:00 -0400910static void xs_local_data_read_skb(struct rpc_xprt *xprt,
911 struct sock *sk,
912 struct sk_buff *skb)
Chuck Lever176e21e2011-05-09 15:22:44 -0400913{
914 struct rpc_task *task;
Chuck Lever176e21e2011-05-09 15:22:44 -0400915 struct rpc_rqst *rovr;
Trond Myklebusta2648092015-10-06 17:03:00 -0400916 int repsize, copied;
Chuck Lever176e21e2011-05-09 15:22:44 -0400917 u32 _xid;
918 __be32 *xp;
919
Chuck Lever176e21e2011-05-09 15:22:44 -0400920 repsize = skb->len - sizeof(rpc_fraghdr);
921 if (repsize < 4) {
922 dprintk("RPC: impossible RPC reply size %d\n", repsize);
Trond Myklebusta2648092015-10-06 17:03:00 -0400923 return;
Chuck Lever176e21e2011-05-09 15:22:44 -0400924 }
925
926 /* Copy the XID from the skb... */
927 xp = skb_header_pointer(skb, sizeof(rpc_fraghdr), sizeof(_xid), &_xid);
928 if (xp == NULL)
Trond Myklebusta2648092015-10-06 17:03:00 -0400929 return;
Chuck Lever176e21e2011-05-09 15:22:44 -0400930
931 /* Look up and lock the request corresponding to the given XID */
Trond Myklebusta2648092015-10-06 17:03:00 -0400932 spin_lock_bh(&xprt->transport_lock);
Chuck Lever176e21e2011-05-09 15:22:44 -0400933 rovr = xprt_lookup_rqst(xprt, *xp);
934 if (!rovr)
935 goto out_unlock;
936 task = rovr->rq_task;
937
938 copied = rovr->rq_private_buf.buflen;
939 if (copied > repsize)
940 copied = repsize;
941
942 if (xs_local_copy_to_xdr(&rovr->rq_private_buf, skb)) {
943 dprintk("RPC: sk_buff copy failed\n");
944 goto out_unlock;
945 }
946
947 xprt_complete_rqst(task, copied);
948
949 out_unlock:
Trond Myklebusta2648092015-10-06 17:03:00 -0400950 spin_unlock_bh(&xprt->transport_lock);
951}
952
953static void xs_local_data_receive(struct sock_xprt *transport)
954{
955 struct sk_buff *skb;
956 struct sock *sk;
957 int err;
958
959 mutex_lock(&transport->recv_mutex);
960 sk = transport->inet;
961 if (sk == NULL)
962 goto out;
963 for (;;) {
964 skb = skb_recv_datagram(sk, 0, 1, &err);
965 if (skb == NULL)
966 break;
967 xs_local_data_read_skb(&transport->xprt, sk, skb);
968 skb_free_datagram(sk, skb);
969 }
970out:
971 mutex_unlock(&transport->recv_mutex);
972}
973
974static void xs_local_data_receive_workfn(struct work_struct *work)
975{
976 struct sock_xprt *transport =
977 container_of(work, struct sock_xprt, recv_worker);
978 xs_local_data_receive(transport);
Chuck Lever176e21e2011-05-09 15:22:44 -0400979}
980
Chuck Lever9903cd12005-08-11 16:25:26 -0400981/**
Trond Myklebustf9b2ee72015-10-06 16:26:05 -0400982 * xs_udp_data_read_skb - receive callback for UDP sockets
983 * @xprt: transport
984 * @sk: socket
985 * @skb: skbuff
Chuck Lever9903cd12005-08-11 16:25:26 -0400986 *
987 */
Trond Myklebustf9b2ee72015-10-06 16:26:05 -0400988static void xs_udp_data_read_skb(struct rpc_xprt *xprt,
989 struct sock *sk,
990 struct sk_buff *skb)
Chuck Lever9903cd12005-08-11 16:25:26 -0400991{
992 struct rpc_task *task;
Chuck Levera246b012005-08-11 16:25:23 -0400993 struct rpc_rqst *rovr;
Trond Myklebustf9b2ee72015-10-06 16:26:05 -0400994 int repsize, copied;
Alexey Dobriyand8ed0292006-09-26 22:29:38 -0700995 u32 _xid;
996 __be32 *xp;
Chuck Levera246b012005-08-11 16:25:23 -0400997
Chuck Levera246b012005-08-11 16:25:23 -0400998 repsize = skb->len - sizeof(struct udphdr);
999 if (repsize < 4) {
Chuck Lever46121cf2007-01-31 12:14:08 -05001000 dprintk("RPC: impossible RPC reply size %d!\n", repsize);
Trond Myklebustf9b2ee72015-10-06 16:26:05 -04001001 return;
Chuck Levera246b012005-08-11 16:25:23 -04001002 }
1003
1004 /* Copy the XID from the skb... */
1005 xp = skb_header_pointer(skb, sizeof(struct udphdr),
1006 sizeof(_xid), &_xid);
1007 if (xp == NULL)
Trond Myklebustf9b2ee72015-10-06 16:26:05 -04001008 return;
Chuck Levera246b012005-08-11 16:25:23 -04001009
1010 /* Look up and lock the request corresponding to the given XID */
Trond Myklebustf9b2ee72015-10-06 16:26:05 -04001011 spin_lock_bh(&xprt->transport_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001012 rovr = xprt_lookup_rqst(xprt, *xp);
1013 if (!rovr)
1014 goto out_unlock;
1015 task = rovr->rq_task;
1016
Chuck Levera246b012005-08-11 16:25:23 -04001017 if ((copied = rovr->rq_private_buf.buflen) > repsize)
1018 copied = repsize;
1019
1020 /* Suck it into the iovec, verify checksum if not done by hw. */
Herbert Xu1781f7f2007-12-11 11:30:32 -08001021 if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb)) {
1022 UDPX_INC_STATS_BH(sk, UDP_MIB_INERRORS);
Chuck Levera246b012005-08-11 16:25:23 -04001023 goto out_unlock;
Herbert Xu1781f7f2007-12-11 11:30:32 -08001024 }
1025
1026 UDPX_INC_STATS_BH(sk, UDP_MIB_INDATAGRAMS);
Chuck Levera246b012005-08-11 16:25:23 -04001027
Trond Myklebust6a24dfb2013-01-08 09:48:15 -05001028 xprt_adjust_cwnd(xprt, task, copied);
Chuck Lever1570c1e2005-08-25 16:25:52 -07001029 xprt_complete_rqst(task, copied);
Chuck Levera246b012005-08-11 16:25:23 -04001030
1031 out_unlock:
Trond Myklebustf9b2ee72015-10-06 16:26:05 -04001032 spin_unlock_bh(&xprt->transport_lock);
1033}
1034
1035static void xs_udp_data_receive(struct sock_xprt *transport)
1036{
1037 struct sk_buff *skb;
1038 struct sock *sk;
1039 int err;
1040
1041 mutex_lock(&transport->recv_mutex);
1042 sk = transport->inet;
1043 if (sk == NULL)
1044 goto out;
1045 for (;;) {
1046 skb = skb_recv_datagram(sk, 0, 1, &err);
1047 if (skb == NULL)
1048 break;
1049 xs_udp_data_read_skb(&transport->xprt, sk, skb);
1050 skb_free_datagram(sk, skb);
1051 }
1052out:
1053 mutex_unlock(&transport->recv_mutex);
1054}
1055
1056static void xs_udp_data_receive_workfn(struct work_struct *work)
1057{
1058 struct sock_xprt *transport =
1059 container_of(work, struct sock_xprt, recv_worker);
1060 xs_udp_data_receive(transport);
1061}
1062
1063/**
1064 * xs_data_ready - "data ready" callback for UDP sockets
1065 * @sk: socket with data to read
1066 *
1067 */
1068static void xs_data_ready(struct sock *sk)
1069{
1070 struct rpc_xprt *xprt;
1071
1072 read_lock_bh(&sk->sk_callback_lock);
1073 dprintk("RPC: xs_data_ready...\n");
1074 xprt = xprt_from_sock(sk);
1075 if (xprt != NULL) {
1076 struct sock_xprt *transport = container_of(xprt,
1077 struct sock_xprt, xprt);
1078 queue_work(rpciod_workqueue, &transport->recv_worker);
1079 }
Eric Dumazetf064af12010-09-22 12:43:39 +00001080 read_unlock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001081}
1082
Trond Myklebusta519fc72012-09-12 16:49:15 -04001083/*
1084 * Helper function to force a TCP close if the server is sending
1085 * junk and/or it has put us in CLOSE_WAIT
1086 */
1087static void xs_tcp_force_close(struct rpc_xprt *xprt)
1088{
Trond Myklebusta519fc72012-09-12 16:49:15 -04001089 xprt_force_disconnect(xprt);
1090}
1091
Chuck Leverdd456472006-12-05 16:35:44 -05001092static inline void xs_tcp_read_fraghdr(struct rpc_xprt *xprt, struct xdr_skb_reader *desc)
Chuck Levera246b012005-08-11 16:25:23 -04001093{
Chuck Lever51971132006-12-05 16:35:19 -05001094 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -04001095 size_t len, used;
1096 char *p;
1097
Chuck Lever51971132006-12-05 16:35:19 -05001098 p = ((char *) &transport->tcp_fraghdr) + transport->tcp_offset;
1099 len = sizeof(transport->tcp_fraghdr) - transport->tcp_offset;
Chuck Lever9d292312006-12-05 16:35:41 -05001100 used = xdr_skb_read_bits(desc, p, len);
Chuck Lever51971132006-12-05 16:35:19 -05001101 transport->tcp_offset += used;
Chuck Levera246b012005-08-11 16:25:23 -04001102 if (used != len)
1103 return;
Chuck Lever808012f2005-08-25 16:25:49 -07001104
Chuck Lever51971132006-12-05 16:35:19 -05001105 transport->tcp_reclen = ntohl(transport->tcp_fraghdr);
1106 if (transport->tcp_reclen & RPC_LAST_STREAM_FRAGMENT)
Chuck Levere136d092006-12-05 16:35:23 -05001107 transport->tcp_flags |= TCP_RCV_LAST_FRAG;
Chuck Levera246b012005-08-11 16:25:23 -04001108 else
Chuck Levere136d092006-12-05 16:35:23 -05001109 transport->tcp_flags &= ~TCP_RCV_LAST_FRAG;
Chuck Lever51971132006-12-05 16:35:19 -05001110 transport->tcp_reclen &= RPC_FRAGMENT_SIZE_MASK;
Chuck Lever808012f2005-08-25 16:25:49 -07001111
Chuck Levere136d092006-12-05 16:35:23 -05001112 transport->tcp_flags &= ~TCP_RCV_COPY_FRAGHDR;
Chuck Lever51971132006-12-05 16:35:19 -05001113 transport->tcp_offset = 0;
Chuck Lever808012f2005-08-25 16:25:49 -07001114
Chuck Levera246b012005-08-11 16:25:23 -04001115 /* Sanity check of the record length */
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001116 if (unlikely(transport->tcp_reclen < 8)) {
Chuck Lever46121cf2007-01-31 12:14:08 -05001117 dprintk("RPC: invalid TCP record fragment length\n");
Trond Myklebusta519fc72012-09-12 16:49:15 -04001118 xs_tcp_force_close(xprt);
Chuck Lever9903cd12005-08-11 16:25:26 -04001119 return;
Chuck Levera246b012005-08-11 16:25:23 -04001120 }
Chuck Lever46121cf2007-01-31 12:14:08 -05001121 dprintk("RPC: reading TCP record fragment of length %d\n",
Chuck Lever51971132006-12-05 16:35:19 -05001122 transport->tcp_reclen);
Chuck Levera246b012005-08-11 16:25:23 -04001123}
1124
Chuck Lever51971132006-12-05 16:35:19 -05001125static void xs_tcp_check_fraghdr(struct sock_xprt *transport)
Chuck Levera246b012005-08-11 16:25:23 -04001126{
Chuck Lever51971132006-12-05 16:35:19 -05001127 if (transport->tcp_offset == transport->tcp_reclen) {
Chuck Levere136d092006-12-05 16:35:23 -05001128 transport->tcp_flags |= TCP_RCV_COPY_FRAGHDR;
Chuck Lever51971132006-12-05 16:35:19 -05001129 transport->tcp_offset = 0;
Chuck Levere136d092006-12-05 16:35:23 -05001130 if (transport->tcp_flags & TCP_RCV_LAST_FRAG) {
1131 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
1132 transport->tcp_flags |= TCP_RCV_COPY_XID;
Chuck Lever51971132006-12-05 16:35:19 -05001133 transport->tcp_copied = 0;
Chuck Levera246b012005-08-11 16:25:23 -04001134 }
1135 }
1136}
1137
Chuck Leverdd456472006-12-05 16:35:44 -05001138static inline void xs_tcp_read_xid(struct sock_xprt *transport, struct xdr_skb_reader *desc)
Chuck Levera246b012005-08-11 16:25:23 -04001139{
1140 size_t len, used;
1141 char *p;
1142
Chuck Lever51971132006-12-05 16:35:19 -05001143 len = sizeof(transport->tcp_xid) - transport->tcp_offset;
Chuck Lever46121cf2007-01-31 12:14:08 -05001144 dprintk("RPC: reading XID (%Zu bytes)\n", len);
Chuck Lever51971132006-12-05 16:35:19 -05001145 p = ((char *) &transport->tcp_xid) + transport->tcp_offset;
Chuck Lever9d292312006-12-05 16:35:41 -05001146 used = xdr_skb_read_bits(desc, p, len);
Chuck Lever51971132006-12-05 16:35:19 -05001147 transport->tcp_offset += used;
Chuck Levera246b012005-08-11 16:25:23 -04001148 if (used != len)
1149 return;
Chuck Levere136d092006-12-05 16:35:23 -05001150 transport->tcp_flags &= ~TCP_RCV_COPY_XID;
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001151 transport->tcp_flags |= TCP_RCV_READ_CALLDIR;
Chuck Lever51971132006-12-05 16:35:19 -05001152 transport->tcp_copied = 4;
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001153 dprintk("RPC: reading %s XID %08x\n",
1154 (transport->tcp_flags & TCP_RPC_REPLY) ? "reply for"
1155 : "request with",
Chuck Lever51971132006-12-05 16:35:19 -05001156 ntohl(transport->tcp_xid));
1157 xs_tcp_check_fraghdr(transport);
Chuck Levera246b012005-08-11 16:25:23 -04001158}
1159
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001160static inline void xs_tcp_read_calldir(struct sock_xprt *transport,
1161 struct xdr_skb_reader *desc)
Chuck Levera246b012005-08-11 16:25:23 -04001162{
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001163 size_t len, used;
1164 u32 offset;
Trond Myklebustb76ce562010-06-16 13:57:32 -04001165 char *p;
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001166
1167 /*
1168 * We want transport->tcp_offset to be 8 at the end of this routine
1169 * (4 bytes for the xid and 4 bytes for the call/reply flag).
1170 * When this function is called for the first time,
1171 * transport->tcp_offset is 4 (after having already read the xid).
1172 */
1173 offset = transport->tcp_offset - sizeof(transport->tcp_xid);
Trond Myklebustb76ce562010-06-16 13:57:32 -04001174 len = sizeof(transport->tcp_calldir) - offset;
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001175 dprintk("RPC: reading CALL/REPLY flag (%Zu bytes)\n", len);
Trond Myklebustb76ce562010-06-16 13:57:32 -04001176 p = ((char *) &transport->tcp_calldir) + offset;
1177 used = xdr_skb_read_bits(desc, p, len);
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001178 transport->tcp_offset += used;
1179 if (used != len)
1180 return;
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001181 transport->tcp_flags &= ~TCP_RCV_READ_CALLDIR;
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001182 /*
1183 * We don't yet have the XDR buffer, so we will write the calldir
1184 * out after we get the buffer from the 'struct rpc_rqst'
1185 */
Trond Myklebustb76ce562010-06-16 13:57:32 -04001186 switch (ntohl(transport->tcp_calldir)) {
1187 case RPC_REPLY:
1188 transport->tcp_flags |= TCP_RCV_COPY_CALLDIR;
1189 transport->tcp_flags |= TCP_RCV_COPY_DATA;
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001190 transport->tcp_flags |= TCP_RPC_REPLY;
Trond Myklebustb76ce562010-06-16 13:57:32 -04001191 break;
1192 case RPC_CALL:
1193 transport->tcp_flags |= TCP_RCV_COPY_CALLDIR;
1194 transport->tcp_flags |= TCP_RCV_COPY_DATA;
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001195 transport->tcp_flags &= ~TCP_RPC_REPLY;
Trond Myklebustb76ce562010-06-16 13:57:32 -04001196 break;
1197 default:
1198 dprintk("RPC: invalid request message type\n");
Trond Myklebusta519fc72012-09-12 16:49:15 -04001199 xs_tcp_force_close(&transport->xprt);
Trond Myklebustb76ce562010-06-16 13:57:32 -04001200 }
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001201 xs_tcp_check_fraghdr(transport);
1202}
1203
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001204static inline void xs_tcp_read_common(struct rpc_xprt *xprt,
1205 struct xdr_skb_reader *desc,
1206 struct rpc_rqst *req)
Chuck Levera246b012005-08-11 16:25:23 -04001207{
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001208 struct sock_xprt *transport =
1209 container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -04001210 struct xdr_buf *rcvbuf;
1211 size_t len;
1212 ssize_t r;
1213
Chuck Levera246b012005-08-11 16:25:23 -04001214 rcvbuf = &req->rq_private_buf;
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001215
1216 if (transport->tcp_flags & TCP_RCV_COPY_CALLDIR) {
1217 /*
1218 * Save the RPC direction in the XDR buffer
1219 */
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001220 memcpy(rcvbuf->head[0].iov_base + transport->tcp_copied,
Trond Myklebustb76ce562010-06-16 13:57:32 -04001221 &transport->tcp_calldir,
1222 sizeof(transport->tcp_calldir));
1223 transport->tcp_copied += sizeof(transport->tcp_calldir);
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001224 transport->tcp_flags &= ~TCP_RCV_COPY_CALLDIR;
Chuck Levera246b012005-08-11 16:25:23 -04001225 }
1226
Chuck Levera246b012005-08-11 16:25:23 -04001227 len = desc->count;
Chuck Lever51971132006-12-05 16:35:19 -05001228 if (len > transport->tcp_reclen - transport->tcp_offset) {
Chuck Leverdd456472006-12-05 16:35:44 -05001229 struct xdr_skb_reader my_desc;
Chuck Levera246b012005-08-11 16:25:23 -04001230
Chuck Lever51971132006-12-05 16:35:19 -05001231 len = transport->tcp_reclen - transport->tcp_offset;
Chuck Levera246b012005-08-11 16:25:23 -04001232 memcpy(&my_desc, desc, sizeof(my_desc));
1233 my_desc.count = len;
Chuck Lever51971132006-12-05 16:35:19 -05001234 r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied,
Chuck Lever9d292312006-12-05 16:35:41 -05001235 &my_desc, xdr_skb_read_bits);
Chuck Levera246b012005-08-11 16:25:23 -04001236 desc->count -= r;
1237 desc->offset += r;
1238 } else
Chuck Lever51971132006-12-05 16:35:19 -05001239 r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied,
Chuck Lever9d292312006-12-05 16:35:41 -05001240 desc, xdr_skb_read_bits);
Chuck Levera246b012005-08-11 16:25:23 -04001241
1242 if (r > 0) {
Chuck Lever51971132006-12-05 16:35:19 -05001243 transport->tcp_copied += r;
1244 transport->tcp_offset += r;
Chuck Levera246b012005-08-11 16:25:23 -04001245 }
1246 if (r != len) {
1247 /* Error when copying to the receive buffer,
1248 * usually because we weren't able to allocate
1249 * additional buffer pages. All we can do now
Chuck Levere136d092006-12-05 16:35:23 -05001250 * is turn off TCP_RCV_COPY_DATA, so the request
Chuck Levera246b012005-08-11 16:25:23 -04001251 * will not receive any additional updates,
1252 * and time out.
1253 * Any remaining data from this record will
1254 * be discarded.
1255 */
Chuck Levere136d092006-12-05 16:35:23 -05001256 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
Chuck Lever46121cf2007-01-31 12:14:08 -05001257 dprintk("RPC: XID %08x truncated request\n",
Chuck Lever51971132006-12-05 16:35:19 -05001258 ntohl(transport->tcp_xid));
Chuck Lever46121cf2007-01-31 12:14:08 -05001259 dprintk("RPC: xprt = %p, tcp_copied = %lu, "
1260 "tcp_offset = %u, tcp_reclen = %u\n",
1261 xprt, transport->tcp_copied,
1262 transport->tcp_offset, transport->tcp_reclen);
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001263 return;
Chuck Levera246b012005-08-11 16:25:23 -04001264 }
1265
Chuck Lever46121cf2007-01-31 12:14:08 -05001266 dprintk("RPC: XID %08x read %Zd bytes\n",
Chuck Lever51971132006-12-05 16:35:19 -05001267 ntohl(transport->tcp_xid), r);
Chuck Lever46121cf2007-01-31 12:14:08 -05001268 dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, "
1269 "tcp_reclen = %u\n", xprt, transport->tcp_copied,
1270 transport->tcp_offset, transport->tcp_reclen);
Chuck Levera246b012005-08-11 16:25:23 -04001271
Chuck Lever51971132006-12-05 16:35:19 -05001272 if (transport->tcp_copied == req->rq_private_buf.buflen)
Chuck Levere136d092006-12-05 16:35:23 -05001273 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
Chuck Lever51971132006-12-05 16:35:19 -05001274 else if (transport->tcp_offset == transport->tcp_reclen) {
Chuck Levere136d092006-12-05 16:35:23 -05001275 if (transport->tcp_flags & TCP_RCV_LAST_FRAG)
1276 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
Chuck Levera246b012005-08-11 16:25:23 -04001277 }
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001278}
1279
1280/*
1281 * Finds the request corresponding to the RPC xid and invokes the common
1282 * tcp read code to read the data.
1283 */
1284static inline int xs_tcp_read_reply(struct rpc_xprt *xprt,
1285 struct xdr_skb_reader *desc)
1286{
1287 struct sock_xprt *transport =
1288 container_of(xprt, struct sock_xprt, xprt);
1289 struct rpc_rqst *req;
1290
1291 dprintk("RPC: read reply XID %08x\n", ntohl(transport->tcp_xid));
1292
1293 /* Find and lock the request corresponding to this xid */
Trond Myklebustedc1b012015-10-05 10:53:49 -04001294 spin_lock_bh(&xprt->transport_lock);
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001295 req = xprt_lookup_rqst(xprt, transport->tcp_xid);
1296 if (!req) {
1297 dprintk("RPC: XID %08x request not found!\n",
1298 ntohl(transport->tcp_xid));
Trond Myklebustedc1b012015-10-05 10:53:49 -04001299 spin_unlock_bh(&xprt->transport_lock);
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001300 return -1;
1301 }
1302
1303 xs_tcp_read_common(xprt, desc, req);
1304
Chuck Levere136d092006-12-05 16:35:23 -05001305 if (!(transport->tcp_flags & TCP_RCV_COPY_DATA))
Chuck Lever51971132006-12-05 16:35:19 -05001306 xprt_complete_rqst(req->rq_task, transport->tcp_copied);
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001307
Trond Myklebustedc1b012015-10-05 10:53:49 -04001308 spin_unlock_bh(&xprt->transport_lock);
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001309 return 0;
1310}
1311
Trond Myklebust9e00abc2011-07-13 19:20:49 -04001312#if defined(CONFIG_SUNRPC_BACKCHANNEL)
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001313/*
1314 * Obtains an rpc_rqst previously allocated and invokes the common
1315 * tcp read code to read the data. The result is placed in the callback
1316 * queue.
1317 * If we're unable to obtain the rpc_rqst we schedule the closing of the
1318 * connection and return -1.
1319 */
Trond Myklebust2ea24492014-02-10 11:18:39 -05001320static int xs_tcp_read_callback(struct rpc_xprt *xprt,
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001321 struct xdr_skb_reader *desc)
1322{
1323 struct sock_xprt *transport =
1324 container_of(xprt, struct sock_xprt, xprt);
1325 struct rpc_rqst *req;
1326
Trond Myklebust2ea24492014-02-10 11:18:39 -05001327 /* Look up and lock the request corresponding to the given XID */
Trond Myklebustedc1b012015-10-05 10:53:49 -04001328 spin_lock_bh(&xprt->transport_lock);
Trond Myklebust2ea24492014-02-10 11:18:39 -05001329 req = xprt_lookup_bc_request(xprt, transport->tcp_xid);
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001330 if (req == NULL) {
Trond Myklebustedc1b012015-10-05 10:53:49 -04001331 spin_unlock_bh(&xprt->transport_lock);
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001332 printk(KERN_WARNING "Callback slot table overflowed\n");
1333 xprt_force_disconnect(xprt);
1334 return -1;
1335 }
1336
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001337 dprintk("RPC: read callback XID %08x\n", ntohl(req->rq_xid));
1338 xs_tcp_read_common(xprt, desc, req);
1339
Trond Myklebust2ea24492014-02-10 11:18:39 -05001340 if (!(transport->tcp_flags & TCP_RCV_COPY_DATA))
1341 xprt_complete_bc_request(req, transport->tcp_copied);
Trond Myklebustedc1b012015-10-05 10:53:49 -04001342 spin_unlock_bh(&xprt->transport_lock);
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001343
1344 return 0;
1345}
1346
1347static inline int _xs_tcp_read_data(struct rpc_xprt *xprt,
1348 struct xdr_skb_reader *desc)
1349{
1350 struct sock_xprt *transport =
1351 container_of(xprt, struct sock_xprt, xprt);
1352
1353 return (transport->tcp_flags & TCP_RPC_REPLY) ?
1354 xs_tcp_read_reply(xprt, desc) :
1355 xs_tcp_read_callback(xprt, desc);
1356}
Chuck Lever76566772015-10-24 17:28:32 -04001357
1358static int xs_tcp_bc_up(struct svc_serv *serv, struct net *net)
1359{
1360 int ret;
1361
1362 ret = svc_create_xprt(serv, "tcp-bc", net, PF_INET, 0,
1363 SVC_SOCK_ANONYMOUS);
1364 if (ret < 0)
1365 return ret;
1366 return 0;
1367}
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001368#else
1369static inline int _xs_tcp_read_data(struct rpc_xprt *xprt,
1370 struct xdr_skb_reader *desc)
1371{
1372 return xs_tcp_read_reply(xprt, desc);
1373}
Trond Myklebust9e00abc2011-07-13 19:20:49 -04001374#endif /* CONFIG_SUNRPC_BACKCHANNEL */
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001375
1376/*
1377 * Read data off the transport. This can be either an RPC_CALL or an
1378 * RPC_REPLY. Relay the processing to helper functions.
1379 */
1380static void xs_tcp_read_data(struct rpc_xprt *xprt,
1381 struct xdr_skb_reader *desc)
1382{
1383 struct sock_xprt *transport =
1384 container_of(xprt, struct sock_xprt, xprt);
1385
1386 if (_xs_tcp_read_data(xprt, desc) == 0)
1387 xs_tcp_check_fraghdr(transport);
1388 else {
1389 /*
1390 * The transport_lock protects the request handling.
1391 * There's no need to hold it to update the tcp_flags.
1392 */
1393 transport->tcp_flags &= ~TCP_RCV_COPY_DATA;
1394 }
Chuck Levera246b012005-08-11 16:25:23 -04001395}
1396
Chuck Leverdd456472006-12-05 16:35:44 -05001397static inline void xs_tcp_read_discard(struct sock_xprt *transport, struct xdr_skb_reader *desc)
Chuck Levera246b012005-08-11 16:25:23 -04001398{
1399 size_t len;
1400
Chuck Lever51971132006-12-05 16:35:19 -05001401 len = transport->tcp_reclen - transport->tcp_offset;
Chuck Levera246b012005-08-11 16:25:23 -04001402 if (len > desc->count)
1403 len = desc->count;
1404 desc->count -= len;
1405 desc->offset += len;
Chuck Lever51971132006-12-05 16:35:19 -05001406 transport->tcp_offset += len;
Chuck Lever46121cf2007-01-31 12:14:08 -05001407 dprintk("RPC: discarded %Zu bytes\n", len);
Chuck Lever51971132006-12-05 16:35:19 -05001408 xs_tcp_check_fraghdr(transport);
Chuck Levera246b012005-08-11 16:25:23 -04001409}
1410
Chuck Lever9903cd12005-08-11 16:25:26 -04001411static 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 -04001412{
1413 struct rpc_xprt *xprt = rd_desc->arg.data;
Chuck Lever51971132006-12-05 16:35:19 -05001414 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Leverdd456472006-12-05 16:35:44 -05001415 struct xdr_skb_reader desc = {
Chuck Levera246b012005-08-11 16:25:23 -04001416 .skb = skb,
1417 .offset = offset,
1418 .count = len,
Chuck Lever9903cd12005-08-11 16:25:26 -04001419 };
Chuck Levera246b012005-08-11 16:25:23 -04001420
Chuck Lever46121cf2007-01-31 12:14:08 -05001421 dprintk("RPC: xs_tcp_data_recv started\n");
Chuck Levera246b012005-08-11 16:25:23 -04001422 do {
Jeff Layton1a867a02014-10-28 14:24:14 -04001423 trace_xs_tcp_data_recv(transport);
Chuck Levera246b012005-08-11 16:25:23 -04001424 /* Read in a new fragment marker if necessary */
1425 /* Can we ever really expect to get completely empty fragments? */
Chuck Levere136d092006-12-05 16:35:23 -05001426 if (transport->tcp_flags & TCP_RCV_COPY_FRAGHDR) {
Chuck Lever9903cd12005-08-11 16:25:26 -04001427 xs_tcp_read_fraghdr(xprt, &desc);
Chuck Levera246b012005-08-11 16:25:23 -04001428 continue;
1429 }
1430 /* Read in the xid if necessary */
Chuck Levere136d092006-12-05 16:35:23 -05001431 if (transport->tcp_flags & TCP_RCV_COPY_XID) {
Chuck Lever51971132006-12-05 16:35:19 -05001432 xs_tcp_read_xid(transport, &desc);
Chuck Levera246b012005-08-11 16:25:23 -04001433 continue;
1434 }
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001435 /* Read in the call/reply flag */
Ricardo Labiagaf4a2e412009-04-01 09:22:54 -04001436 if (transport->tcp_flags & TCP_RCV_READ_CALLDIR) {
Ricardo Labiaga18dca022009-04-01 09:22:53 -04001437 xs_tcp_read_calldir(transport, &desc);
1438 continue;
1439 }
Chuck Levera246b012005-08-11 16:25:23 -04001440 /* Read in the request data */
Chuck Levere136d092006-12-05 16:35:23 -05001441 if (transport->tcp_flags & TCP_RCV_COPY_DATA) {
Ricardo Labiaga44b98ef2009-04-01 09:23:02 -04001442 xs_tcp_read_data(xprt, &desc);
Chuck Levera246b012005-08-11 16:25:23 -04001443 continue;
1444 }
1445 /* Skip over any trailing bytes on short reads */
Chuck Lever51971132006-12-05 16:35:19 -05001446 xs_tcp_read_discard(transport, &desc);
Chuck Levera246b012005-08-11 16:25:23 -04001447 } while (desc.count);
Jeff Layton1a867a02014-10-28 14:24:14 -04001448 trace_xs_tcp_data_recv(transport);
Chuck Lever46121cf2007-01-31 12:14:08 -05001449 dprintk("RPC: xs_tcp_data_recv done\n");
Chuck Levera246b012005-08-11 16:25:23 -04001450 return len - desc.count;
1451}
1452
Trond Myklebust66d7a562015-10-05 11:19:20 -04001453static void xs_tcp_data_receive(struct sock_xprt *transport)
1454{
1455 struct rpc_xprt *xprt = &transport->xprt;
1456 struct sock *sk;
1457 read_descriptor_t rd_desc = {
1458 .count = 2*1024*1024,
1459 .arg.data = xprt,
1460 };
1461 unsigned long total = 0;
1462 int read = 0;
1463
Trond Myklebustedc1b012015-10-05 10:53:49 -04001464 mutex_lock(&transport->recv_mutex);
Trond Myklebust66d7a562015-10-05 11:19:20 -04001465 sk = transport->inet;
Trond Myklebustedc1b012015-10-05 10:53:49 -04001466 if (sk == NULL)
1467 goto out;
Trond Myklebust66d7a562015-10-05 11:19:20 -04001468
1469 /* We use rd_desc to pass struct xprt to xs_tcp_data_recv */
1470 for (;;) {
Trond Myklebustedc1b012015-10-05 10:53:49 -04001471 lock_sock(sk);
Trond Myklebust66d7a562015-10-05 11:19:20 -04001472 read = tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv);
Trond Myklebustedc1b012015-10-05 10:53:49 -04001473 release_sock(sk);
Trond Myklebust66d7a562015-10-05 11:19:20 -04001474 if (read <= 0)
1475 break;
1476 total += read;
1477 rd_desc.count = 65536;
1478 }
Trond Myklebustedc1b012015-10-05 10:53:49 -04001479out:
1480 mutex_unlock(&transport->recv_mutex);
Trond Myklebust66d7a562015-10-05 11:19:20 -04001481 trace_xs_tcp_data_ready(xprt, read, total);
1482}
1483
Trond Myklebustedc1b012015-10-05 10:53:49 -04001484static void xs_tcp_data_receive_workfn(struct work_struct *work)
1485{
1486 struct sock_xprt *transport =
1487 container_of(work, struct sock_xprt, recv_worker);
1488 xs_tcp_data_receive(transport);
1489}
1490
Chuck Lever9903cd12005-08-11 16:25:26 -04001491/**
1492 * xs_tcp_data_ready - "data ready" callback for TCP sockets
1493 * @sk: socket with data to read
Chuck Lever9903cd12005-08-11 16:25:26 -04001494 *
1495 */
David S. Miller676d2362014-04-11 16:15:36 -04001496static void xs_tcp_data_ready(struct sock *sk)
Chuck Levera246b012005-08-11 16:25:23 -04001497{
Trond Myklebust66d7a562015-10-05 11:19:20 -04001498 struct sock_xprt *transport;
Chuck Levera246b012005-08-11 16:25:23 -04001499 struct rpc_xprt *xprt;
Chuck Levera246b012005-08-11 16:25:23 -04001500
Chuck Lever46121cf2007-01-31 12:14:08 -05001501 dprintk("RPC: xs_tcp_data_ready...\n");
1502
Eric Dumazetf064af12010-09-22 12:43:39 +00001503 read_lock_bh(&sk->sk_callback_lock);
Trond Myklebust66d7a562015-10-05 11:19:20 -04001504 if (!(xprt = xprt_from_sock(sk)))
Chuck Levera246b012005-08-11 16:25:23 -04001505 goto out;
Trond Myklebust66d7a562015-10-05 11:19:20 -04001506 transport = container_of(xprt, struct sock_xprt, xprt);
1507
Neil Brown61d0a8e2009-09-23 14:36:37 -04001508 /* Any data means we had a useful conversation, so
1509 * the we don't need to delay the next reconnect
1510 */
1511 if (xprt->reestablish_timeout)
1512 xprt->reestablish_timeout = 0;
Trond Myklebustedc1b012015-10-05 10:53:49 -04001513 queue_work(rpciod_workqueue, &transport->recv_worker);
Neil Brown61d0a8e2009-09-23 14:36:37 -04001514
Chuck Levera246b012005-08-11 16:25:23 -04001515out:
Eric Dumazetf064af12010-09-22 12:43:39 +00001516 read_unlock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001517}
1518
Chuck Lever9903cd12005-08-11 16:25:26 -04001519/**
1520 * xs_tcp_state_change - callback to handle TCP socket state changes
1521 * @sk: socket whose state has changed
1522 *
1523 */
1524static void xs_tcp_state_change(struct sock *sk)
Chuck Levera246b012005-08-11 16:25:23 -04001525{
Chuck Lever9903cd12005-08-11 16:25:26 -04001526 struct rpc_xprt *xprt;
Trond Myklebust0fdea1e2015-09-16 23:43:17 -04001527 struct sock_xprt *transport;
Chuck Levera246b012005-08-11 16:25:23 -04001528
Eric Dumazetf064af12010-09-22 12:43:39 +00001529 read_lock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001530 if (!(xprt = xprt_from_sock(sk)))
1531 goto out;
Chuck Lever46121cf2007-01-31 12:14:08 -05001532 dprintk("RPC: xs_tcp_state_change client %p...\n", xprt);
Andy Chittenden669502f2010-08-10 10:19:53 -04001533 dprintk("RPC: state %x conn %d dead %d zapped %d sk_shutdown %d\n",
Chuck Lever46121cf2007-01-31 12:14:08 -05001534 sk->sk_state, xprt_connected(xprt),
1535 sock_flag(sk, SOCK_DEAD),
Andy Chittenden669502f2010-08-10 10:19:53 -04001536 sock_flag(sk, SOCK_ZAPPED),
1537 sk->sk_shutdown);
Chuck Levera246b012005-08-11 16:25:23 -04001538
Trond Myklebust0fdea1e2015-09-16 23:43:17 -04001539 transport = container_of(xprt, struct sock_xprt, xprt);
Trond Myklebust40b5ea02013-09-04 12:16:23 -04001540 trace_rpc_socket_state_change(xprt, sk->sk_socket);
Chuck Levera246b012005-08-11 16:25:23 -04001541 switch (sk->sk_state) {
1542 case TCP_ESTABLISHED:
Eric Dumazetf064af12010-09-22 12:43:39 +00001543 spin_lock(&xprt->transport_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001544 if (!xprt_test_and_set_connected(xprt)) {
Chuck Lever51971132006-12-05 16:35:19 -05001545
Chuck Levera246b012005-08-11 16:25:23 -04001546 /* Reset TCP record info */
Chuck Lever51971132006-12-05 16:35:19 -05001547 transport->tcp_offset = 0;
1548 transport->tcp_reclen = 0;
1549 transport->tcp_copied = 0;
Chuck Levere136d092006-12-05 16:35:23 -05001550 transport->tcp_flags =
1551 TCP_RCV_COPY_FRAGHDR | TCP_RCV_COPY_XID;
Trond Myklebust8b717982013-09-26 10:18:04 -04001552 xprt->connect_cookie++;
Trond Myklebust0fdea1e2015-09-16 23:43:17 -04001553 clear_bit(XPRT_SOCK_CONNECTING, &transport->sock_state);
1554 xprt_clear_connecting(xprt);
Chuck Lever51971132006-12-05 16:35:19 -05001555
Trond Myklebust2a491992009-03-11 14:38:00 -04001556 xprt_wake_pending_tasks(xprt, -EAGAIN);
Chuck Levera246b012005-08-11 16:25:23 -04001557 }
Eric Dumazetf064af12010-09-22 12:43:39 +00001558 spin_unlock(&xprt->transport_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001559 break;
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001560 case TCP_FIN_WAIT1:
1561 /* The client initiated a shutdown of the socket */
Trond Myklebust7c1d71c2008-04-17 16:52:57 -04001562 xprt->connect_cookie++;
Trond Myklebust663b8852008-01-01 18:42:12 -05001563 xprt->reestablish_timeout = 0;
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001564 set_bit(XPRT_CLOSING, &xprt->state);
Peter Zijlstra4e857c52014-03-17 18:06:10 +01001565 smp_mb__before_atomic();
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001566 clear_bit(XPRT_CONNECTED, &xprt->state);
Trond Myklebustef803672007-12-31 16:19:17 -05001567 clear_bit(XPRT_CLOSE_WAIT, &xprt->state);
Peter Zijlstra4e857c52014-03-17 18:06:10 +01001568 smp_mb__after_atomic();
Chuck Levera246b012005-08-11 16:25:23 -04001569 break;
Trond Myklebust632e3bd2006-01-03 09:55:55 +01001570 case TCP_CLOSE_WAIT:
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001571 /* The server initiated a shutdown of the socket */
Trond Myklebust7c1d71c2008-04-17 16:52:57 -04001572 xprt->connect_cookie++;
Trond Myklebustd0bea452012-10-23 11:35:47 -04001573 clear_bit(XPRT_CONNECTED, &xprt->state);
Trond Myklebusta519fc72012-09-12 16:49:15 -04001574 xs_tcp_force_close(xprt);
Trond Myklebust663b8852008-01-01 18:42:12 -05001575 case TCP_CLOSING:
1576 /*
1577 * If the server closed down the connection, make sure that
1578 * we back off before reconnecting
1579 */
1580 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
1581 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001582 break;
1583 case TCP_LAST_ACK:
Trond Myklebust670f9452009-03-11 14:37:58 -04001584 set_bit(XPRT_CLOSING, &xprt->state);
Peter Zijlstra4e857c52014-03-17 18:06:10 +01001585 smp_mb__before_atomic();
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001586 clear_bit(XPRT_CONNECTED, &xprt->state);
Peter Zijlstra4e857c52014-03-17 18:06:10 +01001587 smp_mb__after_atomic();
Trond Myklebust3b948ae2007-11-05 17:42:39 -05001588 break;
1589 case TCP_CLOSE:
Trond Myklebust0fdea1e2015-09-16 23:43:17 -04001590 if (test_and_clear_bit(XPRT_SOCK_CONNECTING,
1591 &transport->sock_state))
1592 xprt_clear_connecting(xprt);
Trond Myklebust7d1e8252009-03-11 14:38:03 -04001593 xs_sock_mark_closed(xprt);
Chuck Levera246b012005-08-11 16:25:23 -04001594 }
1595 out:
Eric Dumazetf064af12010-09-22 12:43:39 +00001596 read_unlock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001597}
1598
Ilpo Järvinen1f0fa152009-02-06 23:48:33 -08001599static void xs_write_space(struct sock *sk)
1600{
Trond Myklebust13331a52016-01-06 08:57:06 -05001601 struct socket_wq *wq;
Ilpo Järvinen1f0fa152009-02-06 23:48:33 -08001602 struct rpc_xprt *xprt;
1603
Trond Myklebust13331a52016-01-06 08:57:06 -05001604 if (!sk->sk_socket)
Ilpo Järvinen1f0fa152009-02-06 23:48:33 -08001605 return;
Trond Myklebust13331a52016-01-06 08:57:06 -05001606 clear_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
Ilpo Järvinen1f0fa152009-02-06 23:48:33 -08001607
1608 if (unlikely(!(xprt = xprt_from_sock(sk))))
1609 return;
Trond Myklebust13331a52016-01-06 08:57:06 -05001610 rcu_read_lock();
1611 wq = rcu_dereference(sk->sk_wq);
1612 if (!wq || test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags) == 0)
1613 goto out;
Ilpo Järvinen1f0fa152009-02-06 23:48:33 -08001614
1615 xprt_write_space(xprt);
Trond Myklebust13331a52016-01-06 08:57:06 -05001616out:
1617 rcu_read_unlock();
Ilpo Järvinen1f0fa152009-02-06 23:48:33 -08001618}
1619
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04001620/**
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001621 * xs_udp_write_space - callback invoked when socket buffer space
1622 * becomes available
Chuck Lever9903cd12005-08-11 16:25:26 -04001623 * @sk: socket whose state has changed
1624 *
Chuck Levera246b012005-08-11 16:25:23 -04001625 * Called when more output buffer space is available for this socket.
1626 * We try not to wake our writers until they can make "significant"
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001627 * progress, otherwise we'll waste resources thrashing kernel_sendmsg
Chuck Levera246b012005-08-11 16:25:23 -04001628 * with a bunch of small requests.
1629 */
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001630static void xs_udp_write_space(struct sock *sk)
Chuck Levera246b012005-08-11 16:25:23 -04001631{
Eric Dumazetf064af12010-09-22 12:43:39 +00001632 read_lock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001633
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001634 /* from net/core/sock.c:sock_def_write_space */
Ilpo Järvinen1f0fa152009-02-06 23:48:33 -08001635 if (sock_writeable(sk))
1636 xs_write_space(sk);
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001637
Eric Dumazetf064af12010-09-22 12:43:39 +00001638 read_unlock_bh(&sk->sk_callback_lock);
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001639}
Chuck Levera246b012005-08-11 16:25:23 -04001640
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001641/**
1642 * xs_tcp_write_space - callback invoked when socket buffer space
1643 * becomes available
1644 * @sk: socket whose state has changed
1645 *
1646 * Called when more output buffer space is available for this socket.
1647 * We try not to wake our writers until they can make "significant"
1648 * progress, otherwise we'll waste resources thrashing kernel_sendmsg
1649 * with a bunch of small requests.
1650 */
1651static void xs_tcp_write_space(struct sock *sk)
1652{
Eric Dumazetf064af12010-09-22 12:43:39 +00001653 read_lock_bh(&sk->sk_callback_lock);
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001654
1655 /* from net/core/stream.c:sk_stream_write_space */
Eric Dumazet64dc6132013-07-22 20:26:31 -07001656 if (sk_stream_is_writeable(sk))
Ilpo Järvinen1f0fa152009-02-06 23:48:33 -08001657 xs_write_space(sk);
Chuck Leverc7b2cae2005-08-11 16:25:50 -04001658
Eric Dumazetf064af12010-09-22 12:43:39 +00001659 read_unlock_bh(&sk->sk_callback_lock);
Chuck Levera246b012005-08-11 16:25:23 -04001660}
1661
Chuck Lever470056c2005-08-25 16:25:56 -07001662static void xs_udp_do_set_buffer_size(struct rpc_xprt *xprt)
Chuck Levera246b012005-08-11 16:25:23 -04001663{
Chuck Leveree0ac0c2006-12-05 16:35:15 -05001664 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1665 struct sock *sk = transport->inet;
Chuck Levera246b012005-08-11 16:25:23 -04001666
Chuck Lever7c6e0662006-12-05 16:35:30 -05001667 if (transport->rcvsize) {
Chuck Levera246b012005-08-11 16:25:23 -04001668 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
Chuck Lever7c6e0662006-12-05 16:35:30 -05001669 sk->sk_rcvbuf = transport->rcvsize * xprt->max_reqs * 2;
Chuck Levera246b012005-08-11 16:25:23 -04001670 }
Chuck Lever7c6e0662006-12-05 16:35:30 -05001671 if (transport->sndsize) {
Chuck Levera246b012005-08-11 16:25:23 -04001672 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
Chuck Lever7c6e0662006-12-05 16:35:30 -05001673 sk->sk_sndbuf = transport->sndsize * xprt->max_reqs * 2;
Chuck Levera246b012005-08-11 16:25:23 -04001674 sk->sk_write_space(sk);
1675 }
1676}
1677
Chuck Lever43118c22005-08-25 16:25:49 -07001678/**
Chuck Lever470056c2005-08-25 16:25:56 -07001679 * xs_udp_set_buffer_size - set send and receive limits
Chuck Lever43118c22005-08-25 16:25:49 -07001680 * @xprt: generic transport
Chuck Lever470056c2005-08-25 16:25:56 -07001681 * @sndsize: requested size of send buffer, in bytes
1682 * @rcvsize: requested size of receive buffer, in bytes
Chuck Lever43118c22005-08-25 16:25:49 -07001683 *
Chuck Lever470056c2005-08-25 16:25:56 -07001684 * Set socket send and receive buffer size limits.
Chuck Lever43118c22005-08-25 16:25:49 -07001685 */
Chuck Lever470056c2005-08-25 16:25:56 -07001686static void xs_udp_set_buffer_size(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize)
Chuck Lever43118c22005-08-25 16:25:49 -07001687{
Chuck Lever7c6e0662006-12-05 16:35:30 -05001688 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
1689
1690 transport->sndsize = 0;
Chuck Lever470056c2005-08-25 16:25:56 -07001691 if (sndsize)
Chuck Lever7c6e0662006-12-05 16:35:30 -05001692 transport->sndsize = sndsize + 1024;
1693 transport->rcvsize = 0;
Chuck Lever470056c2005-08-25 16:25:56 -07001694 if (rcvsize)
Chuck Lever7c6e0662006-12-05 16:35:30 -05001695 transport->rcvsize = rcvsize + 1024;
Chuck Lever470056c2005-08-25 16:25:56 -07001696
1697 xs_udp_do_set_buffer_size(xprt);
Chuck Lever43118c22005-08-25 16:25:49 -07001698}
1699
Chuck Lever46c0ee82005-08-25 16:25:52 -07001700/**
1701 * xs_udp_timer - called when a retransmit timeout occurs on a UDP transport
1702 * @task: task that timed out
1703 *
1704 * Adjust the congestion window after a retransmit timeout has occurred.
1705 */
Trond Myklebust6a24dfb2013-01-08 09:48:15 -05001706static void xs_udp_timer(struct rpc_xprt *xprt, struct rpc_task *task)
Chuck Lever46c0ee82005-08-25 16:25:52 -07001707{
Trond Myklebust6a24dfb2013-01-08 09:48:15 -05001708 xprt_adjust_cwnd(xprt, task, -ETIMEDOUT);
Chuck Lever46c0ee82005-08-25 16:25:52 -07001709}
1710
Chuck Leverb85d8802006-05-25 01:40:49 -04001711static unsigned short xs_get_random_port(void)
1712{
1713 unsigned short range = xprt_max_resvport - xprt_min_resvport;
Aruna-Hewapathirane63862b52014-01-11 07:15:59 -05001714 unsigned short rand = (unsigned short) prandom_u32() % range;
Chuck Leverb85d8802006-05-25 01:40:49 -04001715 return rand + xprt_min_resvport;
1716}
1717
Chuck Lever92200412006-01-03 09:55:51 +01001718/**
Trond Myklebust4dda9c82015-02-08 15:00:06 -05001719 * xs_set_reuseaddr_port - set the socket's port and address reuse options
1720 * @sock: socket
1721 *
1722 * Note that this function has to be called on all sockets that share the
1723 * same port, and it must be called before binding.
1724 */
1725static void xs_sock_set_reuseport(struct socket *sock)
1726{
Trond Myklebust402e23b2015-02-09 17:20:14 -05001727 int opt = 1;
Trond Myklebust4dda9c82015-02-08 15:00:06 -05001728
Trond Myklebust402e23b2015-02-09 17:20:14 -05001729 kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEPORT,
1730 (char *)&opt, sizeof(opt));
Trond Myklebust4dda9c82015-02-08 15:00:06 -05001731}
1732
1733static unsigned short xs_sock_getport(struct socket *sock)
1734{
1735 struct sockaddr_storage buf;
1736 int buflen;
1737 unsigned short port = 0;
1738
1739 if (kernel_getsockname(sock, (struct sockaddr *)&buf, &buflen) < 0)
1740 goto out;
1741 switch (buf.ss_family) {
1742 case AF_INET6:
1743 port = ntohs(((struct sockaddr_in6 *)&buf)->sin6_port);
1744 break;
1745 case AF_INET:
1746 port = ntohs(((struct sockaddr_in *)&buf)->sin_port);
1747 }
1748out:
1749 return port;
1750}
1751
1752/**
Chuck Lever92200412006-01-03 09:55:51 +01001753 * xs_set_port - reset the port number in the remote endpoint address
1754 * @xprt: generic transport
1755 * @port: new port number
1756 *
1757 */
1758static void xs_set_port(struct rpc_xprt *xprt, unsigned short port)
1759{
Chuck Lever46121cf2007-01-31 12:14:08 -05001760 dprintk("RPC: setting port for xprt %p to %u\n", xprt, port);
Chuck Leverc4efcb12006-08-22 20:06:19 -04001761
Chuck Lever9dc3b092009-08-09 15:09:46 -04001762 rpc_set_port(xs_addr(xprt), port);
1763 xs_update_peer_port(xprt);
Chuck Lever92200412006-01-03 09:55:51 +01001764}
1765
Trond Myklebust4dda9c82015-02-08 15:00:06 -05001766static void xs_set_srcport(struct sock_xprt *transport, struct socket *sock)
1767{
1768 if (transport->srcport == 0)
1769 transport->srcport = xs_sock_getport(sock);
1770}
1771
Pavel Emelyanov5d4ec932010-10-04 16:51:23 +04001772static unsigned short xs_get_srcport(struct sock_xprt *transport)
Trond Myklebust67a391d2007-11-05 17:40:58 -05001773{
Chuck Leverfbfffbd2009-08-09 15:09:46 -04001774 unsigned short port = transport->srcport;
Trond Myklebust67a391d2007-11-05 17:40:58 -05001775
1776 if (port == 0 && transport->xprt.resvport)
1777 port = xs_get_random_port();
1778 return port;
1779}
1780
Pavel Emelyanovbaaf4e42010-10-04 16:51:56 +04001781static unsigned short xs_next_srcport(struct sock_xprt *transport, unsigned short port)
Trond Myklebust67a391d2007-11-05 17:40:58 -05001782{
Chuck Leverfbfffbd2009-08-09 15:09:46 -04001783 if (transport->srcport != 0)
1784 transport->srcport = 0;
Trond Myklebust67a391d2007-11-05 17:40:58 -05001785 if (!transport->xprt.resvport)
1786 return 0;
1787 if (port <= xprt_min_resvport || port > xprt_max_resvport)
1788 return xprt_max_resvport;
1789 return --port;
1790}
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001791static int xs_bind(struct sock_xprt *transport, struct socket *sock)
Chuck Levera246b012005-08-11 16:25:23 -04001792{
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001793 struct sockaddr_storage myaddr;
Trond Myklebust67a391d2007-11-05 17:40:58 -05001794 int err, nloop = 0;
Pavel Emelyanov5d4ec932010-10-04 16:51:23 +04001795 unsigned short port = xs_get_srcport(transport);
Trond Myklebust67a391d2007-11-05 17:40:58 -05001796 unsigned short last;
Chuck Levera246b012005-08-11 16:25:23 -04001797
Chris Perl0f7a6222014-09-05 15:40:21 -04001798 /*
1799 * If we are asking for any ephemeral port (i.e. port == 0 &&
1800 * transport->xprt.resvport == 0), don't bind. Let the local
1801 * port selection happen implicitly when the socket is used
1802 * (for example at connect time).
1803 *
1804 * This ensures that we can continue to establish TCP
1805 * connections even when all local ephemeral ports are already
1806 * a part of some TCP connection. This makes no difference
1807 * for UDP sockets, but also doens't harm them.
1808 *
1809 * If we're asking for any reserved port (i.e. port == 0 &&
1810 * transport->xprt.resvport == 1) xs_get_srcport above will
1811 * ensure that port is non-zero and we will bind as needed.
1812 */
1813 if (port == 0)
1814 return 0;
1815
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001816 memcpy(&myaddr, &transport->srcaddr, transport->xprt.addrlen);
Chuck Levera246b012005-08-11 16:25:23 -04001817 do {
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001818 rpc_set_port((struct sockaddr *)&myaddr, port);
1819 err = kernel_bind(sock, (struct sockaddr *)&myaddr,
1820 transport->xprt.addrlen);
Chuck Levera246b012005-08-11 16:25:23 -04001821 if (err == 0) {
Chuck Leverfbfffbd2009-08-09 15:09:46 -04001822 transport->srcport = port;
Frank van Maarseveend3bc9a12007-07-09 22:23:35 +02001823 break;
Chuck Levera246b012005-08-11 16:25:23 -04001824 }
Trond Myklebust67a391d2007-11-05 17:40:58 -05001825 last = port;
Pavel Emelyanovbaaf4e42010-10-04 16:51:56 +04001826 port = xs_next_srcport(transport, port);
Trond Myklebust67a391d2007-11-05 17:40:58 -05001827 if (port > last)
1828 nloop++;
1829 } while (err == -EADDRINUSE && nloop != 2);
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001830
Chuck Lever4232e862010-10-20 11:52:51 -04001831 if (myaddr.ss_family == AF_INET)
Pavel Emelyanovbeb59b62010-10-05 15:53:08 +04001832 dprintk("RPC: %s %pI4:%u: %s (%d)\n", __func__,
1833 &((struct sockaddr_in *)&myaddr)->sin_addr,
1834 port, err ? "failed" : "ok", err);
1835 else
1836 dprintk("RPC: %s %pI6:%u: %s (%d)\n", __func__,
1837 &((struct sockaddr_in6 *)&myaddr)->sin6_addr,
1838 port, err ? "failed" : "ok", err);
Chuck Levera246b012005-08-11 16:25:23 -04001839 return err;
1840}
1841
Chuck Lever176e21e2011-05-09 15:22:44 -04001842/*
1843 * We don't support autobind on AF_LOCAL sockets
1844 */
1845static void xs_local_rpcbind(struct rpc_task *task)
1846{
Trond Myklebustfb43d172016-01-30 16:39:26 -05001847 xprt_set_bound(task->tk_xprt);
Chuck Lever176e21e2011-05-09 15:22:44 -04001848}
1849
1850static void xs_local_set_port(struct rpc_xprt *xprt, unsigned short port)
1851{
1852}
Chuck Levera246b012005-08-11 16:25:23 -04001853
Peter Zijlstraed075362006-12-06 20:35:24 -08001854#ifdef CONFIG_DEBUG_LOCK_ALLOC
1855static struct lock_class_key xs_key[2];
1856static struct lock_class_key xs_slock_key[2];
1857
Chuck Lever176e21e2011-05-09 15:22:44 -04001858static inline void xs_reclassify_socketu(struct socket *sock)
1859{
1860 struct sock *sk = sock->sk;
1861
Chuck Lever176e21e2011-05-09 15:22:44 -04001862 sock_lock_init_class_and_name(sk, "slock-AF_LOCAL-RPC",
1863 &xs_slock_key[1], "sk_lock-AF_LOCAL-RPC", &xs_key[1]);
1864}
1865
Chuck Lever8945ee52007-08-06 11:58:04 -04001866static inline void xs_reclassify_socket4(struct socket *sock)
Peter Zijlstraed075362006-12-06 20:35:24 -08001867{
1868 struct sock *sk = sock->sk;
Chuck Lever8945ee52007-08-06 11:58:04 -04001869
Chuck Lever8945ee52007-08-06 11:58:04 -04001870 sock_lock_init_class_and_name(sk, "slock-AF_INET-RPC",
1871 &xs_slock_key[0], "sk_lock-AF_INET-RPC", &xs_key[0]);
1872}
Peter Zijlstraed075362006-12-06 20:35:24 -08001873
Chuck Lever8945ee52007-08-06 11:58:04 -04001874static inline void xs_reclassify_socket6(struct socket *sock)
1875{
1876 struct sock *sk = sock->sk;
Peter Zijlstraed075362006-12-06 20:35:24 -08001877
Chuck Lever8945ee52007-08-06 11:58:04 -04001878 sock_lock_init_class_and_name(sk, "slock-AF_INET6-RPC",
1879 &xs_slock_key[1], "sk_lock-AF_INET6-RPC", &xs_key[1]);
Peter Zijlstraed075362006-12-06 20:35:24 -08001880}
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001881
1882static inline void xs_reclassify_socket(int family, struct socket *sock)
1883{
Weston Andros Adamson1b7a1812012-10-23 10:43:39 -04001884 WARN_ON_ONCE(sock_owned_by_user(sock->sk));
1885 if (sock_owned_by_user(sock->sk))
1886 return;
1887
Chuck Lever4232e862010-10-20 11:52:51 -04001888 switch (family) {
Chuck Lever176e21e2011-05-09 15:22:44 -04001889 case AF_LOCAL:
1890 xs_reclassify_socketu(sock);
1891 break;
Chuck Lever4232e862010-10-20 11:52:51 -04001892 case AF_INET:
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001893 xs_reclassify_socket4(sock);
Chuck Lever4232e862010-10-20 11:52:51 -04001894 break;
1895 case AF_INET6:
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001896 xs_reclassify_socket6(sock);
Chuck Lever4232e862010-10-20 11:52:51 -04001897 break;
1898 }
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001899}
Peter Zijlstraed075362006-12-06 20:35:24 -08001900#else
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001901static inline void xs_reclassify_socket(int family, struct socket *sock)
1902{
1903}
Peter Zijlstraed075362006-12-06 20:35:24 -08001904#endif
1905
NeilBrown93dc41b2013-10-31 16:14:36 +11001906static void xs_dummy_setup_socket(struct work_struct *work)
1907{
1908}
1909
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001910static struct socket *xs_create_sock(struct rpc_xprt *xprt,
Trond Myklebust4dda9c82015-02-08 15:00:06 -05001911 struct sock_xprt *transport, int family, int type,
1912 int protocol, bool reuseport)
Pavel Emelyanov22f79322010-10-04 16:54:26 +04001913{
1914 struct socket *sock;
1915 int err;
1916
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001917 err = __sock_create(xprt->xprt_net, family, type, protocol, &sock, 1);
Pavel Emelyanov22f79322010-10-04 16:54:26 +04001918 if (err < 0) {
1919 dprintk("RPC: can't create %d transport socket (%d).\n",
1920 protocol, -err);
1921 goto out;
1922 }
Pavel Emelyanov6bc96382010-10-04 16:56:38 +04001923 xs_reclassify_socket(family, sock);
Pavel Emelyanov22f79322010-10-04 16:54:26 +04001924
Trond Myklebust4dda9c82015-02-08 15:00:06 -05001925 if (reuseport)
1926 xs_sock_set_reuseport(sock);
1927
Ben Hutchings4cea2882011-02-22 21:54:34 +00001928 err = xs_bind(transport, sock);
1929 if (err) {
Pavel Emelyanov22f79322010-10-04 16:54:26 +04001930 sock_release(sock);
1931 goto out;
1932 }
1933
1934 return sock;
1935out:
1936 return ERR_PTR(err);
1937}
1938
Chuck Lever176e21e2011-05-09 15:22:44 -04001939static int xs_local_finish_connecting(struct rpc_xprt *xprt,
1940 struct socket *sock)
1941{
1942 struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
1943 xprt);
1944
1945 if (!transport->inet) {
1946 struct sock *sk = sock->sk;
1947
1948 write_lock_bh(&sk->sk_callback_lock);
1949
1950 xs_save_old_callbacks(transport, sk);
1951
1952 sk->sk_user_data = xprt;
Trond Myklebusta2648092015-10-06 17:03:00 -04001953 sk->sk_data_ready = xs_data_ready;
Chuck Lever176e21e2011-05-09 15:22:44 -04001954 sk->sk_write_space = xs_udp_write_space;
Trond Myklebust21180712013-12-31 13:22:59 -05001955 sk->sk_error_report = xs_error_report;
Trond Myklebustc2126152015-08-19 21:46:15 -05001956 sk->sk_allocation = GFP_NOIO;
Chuck Lever176e21e2011-05-09 15:22:44 -04001957
1958 xprt_clear_connected(xprt);
1959
1960 /* Reset to new socket */
1961 transport->sock = sock;
1962 transport->inet = sk;
1963
1964 write_unlock_bh(&sk->sk_callback_lock);
1965 }
1966
1967 /* Tell the socket layer to start connecting... */
1968 xprt->stat.connect_count++;
1969 xprt->stat.connect_start = jiffies;
1970 return kernel_connect(sock, xs_addr(xprt), xprt->addrlen, 0);
1971}
1972
1973/**
1974 * xs_local_setup_socket - create AF_LOCAL socket, connect to a local endpoint
Chuck Lever176e21e2011-05-09 15:22:44 -04001975 * @transport: socket transport to connect
Chuck Lever176e21e2011-05-09 15:22:44 -04001976 */
J. Bruce Fieldsdc107402013-02-20 17:52:19 -05001977static int xs_local_setup_socket(struct sock_xprt *transport)
Chuck Lever176e21e2011-05-09 15:22:44 -04001978{
Chuck Lever176e21e2011-05-09 15:22:44 -04001979 struct rpc_xprt *xprt = &transport->xprt;
1980 struct socket *sock;
1981 int status = -EIO;
1982
Chuck Lever176e21e2011-05-09 15:22:44 -04001983 status = __sock_create(xprt->xprt_net, AF_LOCAL,
1984 SOCK_STREAM, 0, &sock, 1);
1985 if (status < 0) {
1986 dprintk("RPC: can't create AF_LOCAL "
1987 "transport socket (%d).\n", -status);
1988 goto out;
1989 }
Stefan Hajnoczid1358912015-12-02 14:17:52 +08001990 xs_reclassify_socket(AF_LOCAL, sock);
Chuck Lever176e21e2011-05-09 15:22:44 -04001991
1992 dprintk("RPC: worker connecting xprt %p via AF_LOCAL to %s\n",
1993 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
1994
1995 status = xs_local_finish_connecting(xprt, sock);
Trond Myklebust40b5ea02013-09-04 12:16:23 -04001996 trace_rpc_socket_connect(xprt, sock, status);
Chuck Lever176e21e2011-05-09 15:22:44 -04001997 switch (status) {
1998 case 0:
1999 dprintk("RPC: xprt %p connected to %s\n",
2000 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
2001 xprt_set_connected(xprt);
Trond Myklebust3601c4a2014-06-30 13:42:19 -04002002 case -ENOBUFS:
Chuck Lever176e21e2011-05-09 15:22:44 -04002003 break;
2004 case -ENOENT:
2005 dprintk("RPC: xprt %p: socket %s does not exist\n",
2006 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
2007 break;
Trond Myklebust4a20a982012-12-15 17:02:29 -05002008 case -ECONNREFUSED:
2009 dprintk("RPC: xprt %p: connection refused for %s\n",
2010 xprt, xprt->address_strings[RPC_DISPLAY_ADDR]);
2011 break;
Chuck Lever176e21e2011-05-09 15:22:44 -04002012 default:
2013 printk(KERN_ERR "%s: unhandled error (%d) connecting to %s\n",
2014 __func__, -status,
2015 xprt->address_strings[RPC_DISPLAY_ADDR]);
2016 }
2017
2018out:
2019 xprt_clear_connecting(xprt);
2020 xprt_wake_pending_tasks(xprt, status);
J. Bruce Fieldsdc107402013-02-20 17:52:19 -05002021 return status;
2022}
2023
Linus Torvaldsb6669732013-02-28 18:02:55 -08002024static void xs_local_connect(struct rpc_xprt *xprt, struct rpc_task *task)
J. Bruce Fieldsdc107402013-02-20 17:52:19 -05002025{
J. Bruce Fieldsdc107402013-02-20 17:52:19 -05002026 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2027 int ret;
2028
2029 if (RPC_IS_ASYNC(task)) {
2030 /*
2031 * We want the AF_LOCAL connect to be resolved in the
2032 * filesystem namespace of the process making the rpc
2033 * call. Thus we connect synchronously.
2034 *
2035 * If we want to support asynchronous AF_LOCAL calls,
2036 * we'll need to figure out how to pass a namespace to
2037 * connect.
2038 */
2039 rpc_exit(task, -ENOTCONN);
2040 return;
2041 }
2042 ret = xs_local_setup_socket(transport);
2043 if (ret && !RPC_IS_SOFTCONN(task))
2044 msleep_interruptible(15000);
Chuck Lever176e21e2011-05-09 15:22:44 -04002045}
2046
Jeff Layton3c87ef62015-06-03 16:14:25 -04002047#if IS_ENABLED(CONFIG_SUNRPC_SWAP)
Jeff Laytond6e971d2015-06-03 16:14:28 -04002048/*
2049 * Note that this should be called with XPRT_LOCKED held (or when we otherwise
2050 * know that we have exclusive access to the socket), to guard against
2051 * races with xs_reset_transport.
2052 */
Mel Gormana564b8f2012-07-31 16:45:12 -07002053static void xs_set_memalloc(struct rpc_xprt *xprt)
2054{
2055 struct sock_xprt *transport = container_of(xprt, struct sock_xprt,
2056 xprt);
2057
Jeff Laytond6e971d2015-06-03 16:14:28 -04002058 /*
2059 * If there's no sock, then we have nothing to set. The
2060 * reconnecting process will get it for us.
2061 */
2062 if (!transport->inet)
2063 return;
Jeff Layton8e228132015-06-03 16:14:26 -04002064 if (atomic_read(&xprt->swapper))
Mel Gormana564b8f2012-07-31 16:45:12 -07002065 sk_set_memalloc(transport->inet);
2066}
2067
2068/**
Jeff Laytond67fa4d2015-06-03 16:14:29 -04002069 * xs_enable_swap - Tag this transport as being used for swap.
Mel Gormana564b8f2012-07-31 16:45:12 -07002070 * @xprt: transport to tag
Mel Gormana564b8f2012-07-31 16:45:12 -07002071 *
Jeff Layton8e228132015-06-03 16:14:26 -04002072 * Take a reference to this transport on behalf of the rpc_clnt, and
2073 * optionally mark it for swapping if it wasn't already.
Mel Gormana564b8f2012-07-31 16:45:12 -07002074 */
Jeff Laytond67fa4d2015-06-03 16:14:29 -04002075static int
2076xs_enable_swap(struct rpc_xprt *xprt)
Mel Gormana564b8f2012-07-31 16:45:12 -07002077{
Jeff Laytond6e971d2015-06-03 16:14:28 -04002078 struct sock_xprt *xs = container_of(xprt, struct sock_xprt, xprt);
Mel Gormana564b8f2012-07-31 16:45:12 -07002079
Jeff Laytond6e971d2015-06-03 16:14:28 -04002080 if (atomic_inc_return(&xprt->swapper) != 1)
2081 return 0;
2082 if (wait_on_bit_lock(&xprt->state, XPRT_LOCKED, TASK_KILLABLE))
2083 return -ERESTARTSYS;
2084 if (xs->inet)
2085 sk_set_memalloc(xs->inet);
2086 xprt_release_xprt(xprt, NULL);
Jeff Layton8e228132015-06-03 16:14:26 -04002087 return 0;
Mel Gormana564b8f2012-07-31 16:45:12 -07002088}
Jeff Layton8e228132015-06-03 16:14:26 -04002089
2090/**
Jeff Laytond67fa4d2015-06-03 16:14:29 -04002091 * xs_disable_swap - Untag this transport as being used for swap.
Jeff Layton8e228132015-06-03 16:14:26 -04002092 * @xprt: transport to tag
2093 *
2094 * Drop a "swapper" reference to this xprt on behalf of the rpc_clnt. If the
2095 * swapper refcount goes to 0, untag the socket as a memalloc socket.
2096 */
Jeff Laytond67fa4d2015-06-03 16:14:29 -04002097static void
2098xs_disable_swap(struct rpc_xprt *xprt)
Jeff Layton8e228132015-06-03 16:14:26 -04002099{
Jeff Laytond6e971d2015-06-03 16:14:28 -04002100 struct sock_xprt *xs = container_of(xprt, struct sock_xprt, xprt);
Jeff Layton8e228132015-06-03 16:14:26 -04002101
Jeff Laytond6e971d2015-06-03 16:14:28 -04002102 if (!atomic_dec_and_test(&xprt->swapper))
2103 return;
2104 if (wait_on_bit_lock(&xprt->state, XPRT_LOCKED, TASK_KILLABLE))
2105 return;
2106 if (xs->inet)
2107 sk_clear_memalloc(xs->inet);
2108 xprt_release_xprt(xprt, NULL);
Mel Gormana564b8f2012-07-31 16:45:12 -07002109}
Mel Gormana564b8f2012-07-31 16:45:12 -07002110#else
2111static void xs_set_memalloc(struct rpc_xprt *xprt)
2112{
2113}
Jeff Laytond67fa4d2015-06-03 16:14:29 -04002114
2115static int
2116xs_enable_swap(struct rpc_xprt *xprt)
2117{
2118 return -EINVAL;
2119}
2120
2121static void
2122xs_disable_swap(struct rpc_xprt *xprt)
2123{
2124}
Mel Gormana564b8f2012-07-31 16:45:12 -07002125#endif
2126
Chuck Lever16be2d22007-08-06 11:57:38 -04002127static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
Chuck Levera246b012005-08-11 16:25:23 -04002128{
Chuck Lever16be2d22007-08-06 11:57:38 -04002129 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Leveredb267a2006-08-22 20:06:18 -04002130
Chuck Leveree0ac0c2006-12-05 16:35:15 -05002131 if (!transport->inet) {
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002132 struct sock *sk = sock->sk;
2133
2134 write_lock_bh(&sk->sk_callback_lock);
2135
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04002136 xs_save_old_callbacks(transport, sk);
2137
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002138 sk->sk_user_data = xprt;
Trond Myklebustf9b2ee72015-10-06 16:26:05 -04002139 sk->sk_data_ready = xs_data_ready;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002140 sk->sk_write_space = xs_udp_write_space;
Trond Myklebustc2126152015-08-19 21:46:15 -05002141 sk->sk_allocation = GFP_NOIO;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002142
2143 xprt_set_connected(xprt);
2144
2145 /* Reset to new socket */
Chuck Leveree0ac0c2006-12-05 16:35:15 -05002146 transport->sock = sock;
2147 transport->inet = sk;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002148
Mel Gormana564b8f2012-07-31 16:45:12 -07002149 xs_set_memalloc(xprt);
2150
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002151 write_unlock_bh(&sk->sk_callback_lock);
2152 }
Chuck Lever470056c2005-08-25 16:25:56 -07002153 xs_udp_do_set_buffer_size(xprt);
Chuck Lever16be2d22007-08-06 11:57:38 -04002154}
2155
Pavel Emelyanov8c14ff22010-10-04 16:58:02 +04002156static void xs_udp_setup_socket(struct work_struct *work)
Chuck Levera246b012005-08-11 16:25:23 -04002157{
2158 struct sock_xprt *transport =
2159 container_of(work, struct sock_xprt, connect_worker.work);
2160 struct rpc_xprt *xprt = &transport->xprt;
2161 struct socket *sock = transport->sock;
Pavel Emelyanovb65c0312010-10-04 16:53:46 +04002162 int status = -EIO;
Chuck Levera246b012005-08-11 16:25:23 -04002163
Pavel Emelyanov8c14ff22010-10-04 16:58:02 +04002164 sock = xs_create_sock(xprt, transport,
Trond Myklebust4dda9c82015-02-08 15:00:06 -05002165 xs_addr(xprt)->sa_family, SOCK_DGRAM,
2166 IPPROTO_UDP, false);
Pavel Emelyanovb65c0312010-10-04 16:53:46 +04002167 if (IS_ERR(sock))
Chuck Levera246b012005-08-11 16:25:23 -04002168 goto out;
Chuck Lever68e220b2007-08-06 11:57:48 -04002169
Chuck Leverc740eff2009-08-09 15:09:46 -04002170 dprintk("RPC: worker connecting xprt %p via %s to "
2171 "%s (port %s)\n", xprt,
2172 xprt->address_strings[RPC_DISPLAY_PROTO],
2173 xprt->address_strings[RPC_DISPLAY_ADDR],
2174 xprt->address_strings[RPC_DISPLAY_PORT]);
Chuck Lever68e220b2007-08-06 11:57:48 -04002175
2176 xs_udp_finish_connecting(xprt, sock);
Trond Myklebust40b5ea02013-09-04 12:16:23 -04002177 trace_rpc_socket_connect(xprt, sock, 0);
Chuck Levera246b012005-08-11 16:25:23 -04002178 status = 0;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002179out:
Trond Myklebust718ba5b2015-02-08 18:19:25 -05002180 xprt_unlock_connect(xprt, transport);
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002181 xprt_clear_connecting(xprt);
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002182 xprt_wake_pending_tasks(xprt, status);
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002183}
2184
Trond Myklebust4876cc72015-06-19 16:17:57 -04002185/**
2186 * xs_tcp_shutdown - gracefully shut down a TCP socket
2187 * @xprt: transport
2188 *
2189 * Initiates a graceful shutdown of the TCP socket by calling the
2190 * equivalent of shutdown(SHUT_RDWR);
2191 */
2192static void xs_tcp_shutdown(struct rpc_xprt *xprt)
2193{
2194 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2195 struct socket *sock = transport->sock;
2196
2197 if (sock == NULL)
2198 return;
2199 if (xprt_connected(xprt)) {
2200 kernel_sock_shutdown(sock, SHUT_RDWR);
2201 trace_rpc_socket_shutdown(xprt, sock);
2202 } else
2203 xs_reset_transport(transport);
2204}
2205
Chuck Lever16be2d22007-08-06 11:57:38 -04002206static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock)
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002207{
Chuck Lever16be2d22007-08-06 11:57:38 -04002208 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Trond Myklebustfe19a962011-03-18 20:21:23 -04002209 int ret = -ENOTCONN;
Chuck Leveredb267a2006-08-22 20:06:18 -04002210
Chuck Leveree0ac0c2006-12-05 16:35:15 -05002211 if (!transport->inet) {
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002212 struct sock *sk = sock->sk;
Trond Myklebust7f260e82013-09-24 11:25:22 -04002213 unsigned int keepidle = xprt->timeout->to_initval / HZ;
2214 unsigned int keepcnt = xprt->timeout->to_retries + 1;
2215 unsigned int opt_on = 1;
Trond Myklebust775f06a2015-06-20 15:31:54 -04002216 unsigned int timeo;
Trond Myklebust7f260e82013-09-24 11:25:22 -04002217
2218 /* TCP Keepalive options */
2219 kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE,
2220 (char *)&opt_on, sizeof(opt_on));
2221 kernel_setsockopt(sock, SOL_TCP, TCP_KEEPIDLE,
2222 (char *)&keepidle, sizeof(keepidle));
2223 kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL,
2224 (char *)&keepidle, sizeof(keepidle));
2225 kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT,
2226 (char *)&keepcnt, sizeof(keepcnt));
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002227
Trond Myklebust775f06a2015-06-20 15:31:54 -04002228 /* TCP user timeout (see RFC5482) */
2229 timeo = jiffies_to_msecs(xprt->timeout->to_initval) *
2230 (xprt->timeout->to_retries + 1);
2231 kernel_setsockopt(sock, SOL_TCP, TCP_USER_TIMEOUT,
2232 (char *)&timeo, sizeof(timeo));
2233
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002234 write_lock_bh(&sk->sk_callback_lock);
2235
Trond Myklebust2a9e1cf2008-10-28 15:21:39 -04002236 xs_save_old_callbacks(transport, sk);
2237
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002238 sk->sk_user_data = xprt;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002239 sk->sk_data_ready = xs_tcp_data_ready;
2240 sk->sk_state_change = xs_tcp_state_change;
2241 sk->sk_write_space = xs_tcp_write_space;
Trond Myklebust21180712013-12-31 13:22:59 -05002242 sk->sk_error_report = xs_error_report;
Trond Myklebustc2126152015-08-19 21:46:15 -05002243 sk->sk_allocation = GFP_NOIO;
Chuck Lever3167e122005-08-25 16:25:55 -07002244
2245 /* socket options */
Chuck Lever3167e122005-08-25 16:25:55 -07002246 sock_reset_flag(sk, SOCK_LINGER);
Chuck Lever3167e122005-08-25 16:25:55 -07002247 tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002248
2249 xprt_clear_connected(xprt);
2250
2251 /* Reset to new socket */
Chuck Leveree0ac0c2006-12-05 16:35:15 -05002252 transport->sock = sock;
2253 transport->inet = sk;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002254
2255 write_unlock_bh(&sk->sk_callback_lock);
2256 }
2257
Trond Myklebust01d37c42009-03-11 14:09:39 -04002258 if (!xprt_bound(xprt))
Trond Myklebustfe19a962011-03-18 20:21:23 -04002259 goto out;
Trond Myklebust01d37c42009-03-11 14:09:39 -04002260
Mel Gormana564b8f2012-07-31 16:45:12 -07002261 xs_set_memalloc(xprt);
2262
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002263 /* Tell the socket layer to start connecting... */
Chuck Lever262ca072006-03-20 13:44:16 -05002264 xprt->stat.connect_count++;
2265 xprt->stat.connect_start = jiffies;
Trond Myklebust0fdea1e2015-09-16 23:43:17 -04002266 set_bit(XPRT_SOCK_CONNECTING, &transport->sock_state);
Trond Myklebustfe19a962011-03-18 20:21:23 -04002267 ret = kernel_connect(sock, xs_addr(xprt), xprt->addrlen, O_NONBLOCK);
2268 switch (ret) {
2269 case 0:
Trond Myklebust4dda9c82015-02-08 15:00:06 -05002270 xs_set_srcport(transport, sock);
Trond Myklebustfe19a962011-03-18 20:21:23 -04002271 case -EINPROGRESS:
2272 /* SYN_SENT! */
Trond Myklebustfe19a962011-03-18 20:21:23 -04002273 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
2274 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2275 }
2276out:
2277 return ret;
Chuck Lever16be2d22007-08-06 11:57:38 -04002278}
2279
2280/**
Trond Myklebustb61d59f2009-03-11 14:38:04 -04002281 * xs_tcp_setup_socket - create a TCP socket and connect to a remote endpoint
Chuck Lever16be2d22007-08-06 11:57:38 -04002282 *
2283 * Invoked by a work queue tasklet.
2284 */
Pavel Emelyanovcdd518d2010-10-04 16:57:40 +04002285static void xs_tcp_setup_socket(struct work_struct *work)
Chuck Lever16be2d22007-08-06 11:57:38 -04002286{
Pavel Emelyanovcdd518d2010-10-04 16:57:40 +04002287 struct sock_xprt *transport =
2288 container_of(work, struct sock_xprt, connect_worker.work);
Chuck Lever16be2d22007-08-06 11:57:38 -04002289 struct socket *sock = transport->sock;
Pavel Emelyanova9f5f0f2010-10-04 16:52:25 +04002290 struct rpc_xprt *xprt = &transport->xprt;
Trond Myklebustb61d59f2009-03-11 14:38:04 -04002291 int status = -EIO;
Chuck Lever16be2d22007-08-06 11:57:38 -04002292
Chuck Lever16be2d22007-08-06 11:57:38 -04002293 if (!sock) {
Pavel Emelyanovcdd518d2010-10-04 16:57:40 +04002294 sock = xs_create_sock(xprt, transport,
Trond Myklebust4dda9c82015-02-08 15:00:06 -05002295 xs_addr(xprt)->sa_family, SOCK_STREAM,
2296 IPPROTO_TCP, true);
Trond Myklebustb61d59f2009-03-11 14:38:04 -04002297 if (IS_ERR(sock)) {
2298 status = PTR_ERR(sock);
Chuck Lever16be2d22007-08-06 11:57:38 -04002299 goto out;
2300 }
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002301 }
2302
Chuck Leverc740eff2009-08-09 15:09:46 -04002303 dprintk("RPC: worker connecting xprt %p via %s to "
2304 "%s (port %s)\n", xprt,
2305 xprt->address_strings[RPC_DISPLAY_PROTO],
2306 xprt->address_strings[RPC_DISPLAY_ADDR],
2307 xprt->address_strings[RPC_DISPLAY_PORT]);
Chuck Lever16be2d22007-08-06 11:57:38 -04002308
2309 status = xs_tcp_finish_connecting(xprt, sock);
Trond Myklebust40b5ea02013-09-04 12:16:23 -04002310 trace_rpc_socket_connect(xprt, sock, status);
Chuck Lever46121cf2007-01-31 12:14:08 -05002311 dprintk("RPC: %p connect status %d connected %d sock state %d\n",
2312 xprt, -status, xprt_connected(xprt),
2313 sock->sk->sk_state);
Trond Myklebust2a491992009-03-11 14:38:00 -04002314 switch (status) {
Trond Myklebustf75e6742009-04-21 17:18:20 -04002315 default:
2316 printk("%s: connect returned unhandled error %d\n",
2317 __func__, status);
2318 case -EADDRNOTAVAIL:
2319 /* We're probably in TIME_WAIT. Get rid of existing socket,
2320 * and retry
2321 */
Trond Myklebusta519fc72012-09-12 16:49:15 -04002322 xs_tcp_force_close(xprt);
Trond Myklebust88b5ed72009-06-17 13:22:57 -07002323 break;
Trond Myklebust2a491992009-03-11 14:38:00 -04002324 case 0:
2325 case -EINPROGRESS:
2326 case -EALREADY:
Trond Myklebust718ba5b2015-02-08 18:19:25 -05002327 xprt_unlock_connect(xprt, transport);
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002328 return;
Trond Myklebust9fcfe0c2010-03-02 13:06:21 -05002329 case -EINVAL:
2330 /* Happens, for instance, if the user specified a link
2331 * local IPv6 address without a scope-id.
2332 */
Trond Myklebust3ed5e2a2013-03-04 17:29:33 -05002333 case -ECONNREFUSED:
2334 case -ECONNRESET:
2335 case -ENETUNREACH:
Trond Myklebust3913c782015-02-08 21:44:04 -05002336 case -EADDRINUSE:
Trond Myklebust3601c4a2014-06-30 13:42:19 -04002337 case -ENOBUFS:
Trond Myklebust3ed5e2a2013-03-04 17:29:33 -05002338 /* retry with existing socket, after a delay */
Trond Myklebust4efdd922015-02-08 15:34:28 -05002339 xs_tcp_force_close(xprt);
Trond Myklebust9fcfe0c2010-03-02 13:06:21 -05002340 goto out;
Chuck Levera246b012005-08-11 16:25:23 -04002341 }
Trond Myklebust2a491992009-03-11 14:38:00 -04002342 status = -EAGAIN;
Chuck Levera246b012005-08-11 16:25:23 -04002343out:
Trond Myklebust718ba5b2015-02-08 18:19:25 -05002344 xprt_unlock_connect(xprt, transport);
Chuck Lever2226feb2005-08-11 16:25:38 -04002345 xprt_clear_connecting(xprt);
Trond Myklebust7d1e8252009-03-11 14:38:03 -04002346 xprt_wake_pending_tasks(xprt, status);
Chuck Levera246b012005-08-11 16:25:23 -04002347}
2348
Chuck Lever68e220b2007-08-06 11:57:48 -04002349/**
Chuck Lever9903cd12005-08-11 16:25:26 -04002350 * xs_connect - connect a socket to a remote endpoint
Trond Myklebust1b092092013-01-08 09:26:49 -05002351 * @xprt: pointer to transport structure
Chuck Lever9903cd12005-08-11 16:25:26 -04002352 * @task: address of RPC task that manages state of connect request
2353 *
2354 * TCP: If the remote end dropped the connection, delay reconnecting.
Chuck Lever03bf4b72005-08-25 16:25:55 -07002355 *
2356 * UDP socket connects are synchronous, but we use a work queue anyway
2357 * to guarantee that even unprivileged user processes can set up a
2358 * socket on a privileged port.
2359 *
2360 * If a UDP socket connect fails, the delay behavior here prevents
2361 * retry floods (hard mounts).
Chuck Lever9903cd12005-08-11 16:25:26 -04002362 */
Trond Myklebust1b092092013-01-08 09:26:49 -05002363static void xs_connect(struct rpc_xprt *xprt, struct rpc_task *task)
Chuck Levera246b012005-08-11 16:25:23 -04002364{
Chuck Leveree0ac0c2006-12-05 16:35:15 -05002365 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -04002366
Trond Myklebust718ba5b2015-02-08 18:19:25 -05002367 WARN_ON_ONCE(!xprt_lock_connect(xprt, task, transport));
2368
Trond Myklebust99b1a4c2015-08-13 15:33:51 -04002369 if (transport->sock != NULL) {
Chuck Lever46121cf2007-01-31 12:14:08 -05002370 dprintk("RPC: xs_connect delayed xprt %p for %lu "
2371 "seconds\n",
Chuck Lever03bf4b72005-08-25 16:25:55 -07002372 xprt, xprt->reestablish_timeout / HZ);
Trond Myklebust99b1a4c2015-08-13 15:33:51 -04002373
2374 /* Start by resetting any existing state */
2375 xs_reset_transport(transport);
2376
Trond Myklebustc1384c92007-06-14 18:00:42 -04002377 queue_delayed_work(rpciod_workqueue,
2378 &transport->connect_worker,
2379 xprt->reestablish_timeout);
Chuck Lever03bf4b72005-08-25 16:25:55 -07002380 xprt->reestablish_timeout <<= 1;
Neil Brown61d0a8e2009-09-23 14:36:37 -04002381 if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO)
2382 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
Chuck Lever03bf4b72005-08-25 16:25:55 -07002383 if (xprt->reestablish_timeout > XS_TCP_MAX_REEST_TO)
2384 xprt->reestablish_timeout = XS_TCP_MAX_REEST_TO;
Chuck Leverb0d93ad2005-08-11 16:25:53 -04002385 } else {
Chuck Lever46121cf2007-01-31 12:14:08 -05002386 dprintk("RPC: xs_connect scheduled xprt %p\n", xprt);
Trond Myklebustc1384c92007-06-14 18:00:42 -04002387 queue_delayed_work(rpciod_workqueue,
2388 &transport->connect_worker, 0);
Chuck Levera246b012005-08-11 16:25:23 -04002389 }
2390}
2391
Chuck Lever262ca072006-03-20 13:44:16 -05002392/**
Chuck Lever176e21e2011-05-09 15:22:44 -04002393 * xs_local_print_stats - display AF_LOCAL socket-specifc stats
2394 * @xprt: rpc_xprt struct containing statistics
2395 * @seq: output file
2396 *
2397 */
2398static void xs_local_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2399{
2400 long idle_time = 0;
2401
2402 if (xprt_connected(xprt))
2403 idle_time = (long)(jiffies - xprt->last_used) / HZ;
2404
2405 seq_printf(seq, "\txprt:\tlocal %lu %lu %lu %ld %lu %lu %lu "
Andy Adamson15a45202012-02-14 16:19:18 -05002406 "%llu %llu %lu %llu %llu\n",
Chuck Lever176e21e2011-05-09 15:22:44 -04002407 xprt->stat.bind_count,
2408 xprt->stat.connect_count,
2409 xprt->stat.connect_time,
2410 idle_time,
2411 xprt->stat.sends,
2412 xprt->stat.recvs,
2413 xprt->stat.bad_xids,
2414 xprt->stat.req_u,
Andy Adamson15a45202012-02-14 16:19:18 -05002415 xprt->stat.bklog_u,
2416 xprt->stat.max_slots,
2417 xprt->stat.sending_u,
2418 xprt->stat.pending_u);
Chuck Lever176e21e2011-05-09 15:22:44 -04002419}
2420
2421/**
Chuck Lever262ca072006-03-20 13:44:16 -05002422 * xs_udp_print_stats - display UDP socket-specifc stats
2423 * @xprt: rpc_xprt struct containing statistics
2424 * @seq: output file
2425 *
2426 */
2427static void xs_udp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2428{
Chuck Leverc8475462006-12-05 16:35:26 -05002429 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
2430
Andy Adamson15a45202012-02-14 16:19:18 -05002431 seq_printf(seq, "\txprt:\tudp %u %lu %lu %lu %lu %llu %llu "
2432 "%lu %llu %llu\n",
Chuck Leverfbfffbd2009-08-09 15:09:46 -04002433 transport->srcport,
Chuck Lever262ca072006-03-20 13:44:16 -05002434 xprt->stat.bind_count,
2435 xprt->stat.sends,
2436 xprt->stat.recvs,
2437 xprt->stat.bad_xids,
2438 xprt->stat.req_u,
Andy Adamson15a45202012-02-14 16:19:18 -05002439 xprt->stat.bklog_u,
2440 xprt->stat.max_slots,
2441 xprt->stat.sending_u,
2442 xprt->stat.pending_u);
Chuck Lever262ca072006-03-20 13:44:16 -05002443}
2444
2445/**
2446 * xs_tcp_print_stats - display TCP socket-specifc stats
2447 * @xprt: rpc_xprt struct containing statistics
2448 * @seq: output file
2449 *
2450 */
2451static void xs_tcp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq)
2452{
Chuck Leverc8475462006-12-05 16:35:26 -05002453 struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Lever262ca072006-03-20 13:44:16 -05002454 long idle_time = 0;
2455
2456 if (xprt_connected(xprt))
2457 idle_time = (long)(jiffies - xprt->last_used) / HZ;
2458
Andy Adamson15a45202012-02-14 16:19:18 -05002459 seq_printf(seq, "\txprt:\ttcp %u %lu %lu %lu %ld %lu %lu %lu "
2460 "%llu %llu %lu %llu %llu\n",
Chuck Leverfbfffbd2009-08-09 15:09:46 -04002461 transport->srcport,
Chuck Lever262ca072006-03-20 13:44:16 -05002462 xprt->stat.bind_count,
2463 xprt->stat.connect_count,
2464 xprt->stat.connect_time,
2465 idle_time,
2466 xprt->stat.sends,
2467 xprt->stat.recvs,
2468 xprt->stat.bad_xids,
2469 xprt->stat.req_u,
Andy Adamson15a45202012-02-14 16:19:18 -05002470 xprt->stat.bklog_u,
2471 xprt->stat.max_slots,
2472 xprt->stat.sending_u,
2473 xprt->stat.pending_u);
Chuck Lever262ca072006-03-20 13:44:16 -05002474}
2475
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002476/*
2477 * Allocate a bunch of pages for a scratch buffer for the rpc code. The reason
2478 * we allocate pages instead doing a kmalloc like rpc_malloc is because we want
2479 * to use the server side send routines.
2480 */
H Hartley Sweeten5a51f132010-01-14 15:38:31 -07002481static void *bc_malloc(struct rpc_task *task, size_t size)
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002482{
2483 struct page *page;
2484 struct rpc_buffer *buf;
2485
Weston Andros Adamsonb8a13d02012-10-23 10:43:43 -04002486 WARN_ON_ONCE(size > PAGE_SIZE - sizeof(struct rpc_buffer));
2487 if (size > PAGE_SIZE - sizeof(struct rpc_buffer))
2488 return NULL;
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002489
Weston Andros Adamsonb8a13d02012-10-23 10:43:43 -04002490 page = alloc_page(GFP_KERNEL);
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002491 if (!page)
2492 return NULL;
2493
2494 buf = page_address(page);
2495 buf->len = PAGE_SIZE;
2496
2497 return buf->data;
2498}
2499
2500/*
2501 * Free the space allocated in the bc_alloc routine
2502 */
H Hartley Sweeten5a51f132010-01-14 15:38:31 -07002503static void bc_free(void *buffer)
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002504{
2505 struct rpc_buffer *buf;
2506
2507 if (!buffer)
2508 return;
2509
2510 buf = container_of(buffer, struct rpc_buffer, data);
2511 free_page((unsigned long)buf);
2512}
2513
2514/*
2515 * Use the svc_sock to send the callback. Must be called with svsk->sk_mutex
2516 * held. Borrows heavily from svc_tcp_sendto and xs_tcp_send_request.
2517 */
2518static int bc_sendto(struct rpc_rqst *req)
2519{
2520 int len;
2521 struct xdr_buf *xbufp = &req->rq_snd_buf;
2522 struct rpc_xprt *xprt = req->rq_xprt;
2523 struct sock_xprt *transport =
2524 container_of(xprt, struct sock_xprt, xprt);
2525 struct socket *sock = transport->sock;
2526 unsigned long headoff;
2527 unsigned long tailoff;
2528
Chuck Lever61677ee2011-05-09 15:22:34 -04002529 xs_encode_stream_record_marker(xbufp);
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002530
2531 tailoff = (unsigned long)xbufp->tail[0].iov_base & ~PAGE_MASK;
2532 headoff = (unsigned long)xbufp->head[0].iov_base & ~PAGE_MASK;
2533 len = svc_send_common(sock, xbufp,
2534 virt_to_page(xbufp->head[0].iov_base), headoff,
2535 xbufp->tail[0].iov_base, tailoff);
2536
2537 if (len != xbufp->len) {
2538 printk(KERN_NOTICE "Error sending entire callback!\n");
2539 len = -EAGAIN;
2540 }
2541
2542 return len;
2543}
2544
2545/*
2546 * The send routine. Borrows from svc_send
2547 */
2548static int bc_send_request(struct rpc_task *task)
2549{
2550 struct rpc_rqst *req = task->tk_rqstp;
2551 struct svc_xprt *xprt;
Andrzej Hajda7fc56132015-09-24 16:00:09 +02002552 int len;
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002553
2554 dprintk("sending request with xid: %08x\n", ntohl(req->rq_xid));
2555 /*
2556 * Get the server socket associated with this callback xprt
2557 */
2558 xprt = req->rq_xprt->bc_xprt;
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002559
2560 /*
2561 * Grab the mutex to serialize data as the connection is shared
2562 * with the fore channel
2563 */
2564 if (!mutex_trylock(&xprt->xpt_mutex)) {
2565 rpc_sleep_on(&xprt->xpt_bc_pending, task, NULL);
2566 if (!mutex_trylock(&xprt->xpt_mutex))
2567 return -EAGAIN;
2568 rpc_wake_up_queued_task(&xprt->xpt_bc_pending, task);
2569 }
2570 if (test_bit(XPT_DEAD, &xprt->xpt_flags))
2571 len = -ENOTCONN;
2572 else
2573 len = bc_sendto(req);
2574 mutex_unlock(&xprt->xpt_mutex);
2575
2576 if (len > 0)
2577 len = 0;
2578
2579 return len;
2580}
2581
2582/*
2583 * The close routine. Since this is client initiated, we do nothing
2584 */
2585
2586static void bc_close(struct rpc_xprt *xprt)
2587{
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002588}
2589
2590/*
2591 * The xprt destroy routine. Again, because this connection is client
2592 * initiated, we do nothing
2593 */
2594
2595static void bc_destroy(struct rpc_xprt *xprt)
2596{
Kinglong Mee47f72ef2014-03-24 11:58:16 +08002597 dprintk("RPC: bc_destroy xprt %p\n", xprt);
2598
2599 xs_xprt_free(xprt);
2600 module_put(THIS_MODULE);
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002601}
2602
Chuck Lever176e21e2011-05-09 15:22:44 -04002603static struct rpc_xprt_ops xs_local_ops = {
2604 .reserve_xprt = xprt_reserve_xprt,
2605 .release_xprt = xs_tcp_release_xprt,
Trond Myklebustf39c1bf2012-09-07 11:08:50 -04002606 .alloc_slot = xprt_alloc_slot,
Chuck Lever176e21e2011-05-09 15:22:44 -04002607 .rpcbind = xs_local_rpcbind,
2608 .set_port = xs_local_set_port,
J. Bruce Fieldsdc107402013-02-20 17:52:19 -05002609 .connect = xs_local_connect,
Chuck Lever176e21e2011-05-09 15:22:44 -04002610 .buf_alloc = rpc_malloc,
2611 .buf_free = rpc_free,
2612 .send_request = xs_local_send_request,
2613 .set_retrans_timeout = xprt_set_retrans_timeout_def,
2614 .close = xs_close,
Trond Myklebusta1311d82013-10-31 09:18:49 -04002615 .destroy = xs_destroy,
Chuck Lever176e21e2011-05-09 15:22:44 -04002616 .print_stats = xs_local_print_stats,
Jeff Laytond67fa4d2015-06-03 16:14:29 -04002617 .enable_swap = xs_enable_swap,
2618 .disable_swap = xs_disable_swap,
Chuck Lever176e21e2011-05-09 15:22:44 -04002619};
2620
Chuck Lever262965f2005-08-11 16:25:56 -04002621static struct rpc_xprt_ops xs_udp_ops = {
Chuck Lever43118c22005-08-25 16:25:49 -07002622 .set_buffer_size = xs_udp_set_buffer_size,
Chuck Lever12a80462005-08-25 16:25:51 -07002623 .reserve_xprt = xprt_reserve_xprt_cong,
Chuck Lever49e9a892005-08-25 16:25:51 -07002624 .release_xprt = xprt_release_xprt_cong,
Trond Myklebustf39c1bf2012-09-07 11:08:50 -04002625 .alloc_slot = xprt_alloc_slot,
Chuck Lever45160d62007-07-01 12:13:17 -04002626 .rpcbind = rpcb_getport_async,
Chuck Lever92200412006-01-03 09:55:51 +01002627 .set_port = xs_set_port,
Chuck Lever9903cd12005-08-11 16:25:26 -04002628 .connect = xs_connect,
Chuck Lever02107142006-01-03 09:55:49 +01002629 .buf_alloc = rpc_malloc,
2630 .buf_free = rpc_free,
Chuck Lever262965f2005-08-11 16:25:56 -04002631 .send_request = xs_udp_send_request,
Chuck Leverfe3aca22005-08-25 16:25:50 -07002632 .set_retrans_timeout = xprt_set_retrans_timeout_rtt,
Chuck Lever46c0ee82005-08-25 16:25:52 -07002633 .timer = xs_udp_timer,
Chuck Levera58dd392005-08-25 16:25:53 -07002634 .release_request = xprt_release_rqst_cong,
Chuck Lever262965f2005-08-11 16:25:56 -04002635 .close = xs_close,
2636 .destroy = xs_destroy,
Chuck Lever262ca072006-03-20 13:44:16 -05002637 .print_stats = xs_udp_print_stats,
Jeff Laytond67fa4d2015-06-03 16:14:29 -04002638 .enable_swap = xs_enable_swap,
2639 .disable_swap = xs_disable_swap,
Chuck Lever4a068252015-05-11 14:02:25 -04002640 .inject_disconnect = xs_inject_disconnect,
Chuck Lever262965f2005-08-11 16:25:56 -04002641};
2642
2643static struct rpc_xprt_ops xs_tcp_ops = {
Chuck Lever12a80462005-08-25 16:25:51 -07002644 .reserve_xprt = xprt_reserve_xprt,
Trond Myklebuste0ab53d2006-07-27 17:22:50 -04002645 .release_xprt = xs_tcp_release_xprt,
Trond Myklebustf39c1bf2012-09-07 11:08:50 -04002646 .alloc_slot = xprt_lock_and_alloc_slot,
Chuck Lever45160d62007-07-01 12:13:17 -04002647 .rpcbind = rpcb_getport_async,
Chuck Lever92200412006-01-03 09:55:51 +01002648 .set_port = xs_set_port,
Trond Myklebust0b9e7942010-04-16 16:41:57 -04002649 .connect = xs_connect,
Chuck Lever02107142006-01-03 09:55:49 +01002650 .buf_alloc = rpc_malloc,
2651 .buf_free = rpc_free,
Chuck Lever262965f2005-08-11 16:25:56 -04002652 .send_request = xs_tcp_send_request,
Chuck Leverfe3aca22005-08-25 16:25:50 -07002653 .set_retrans_timeout = xprt_set_retrans_timeout_def,
Trond Myklebustc627d312015-02-10 11:06:04 -05002654 .close = xs_tcp_shutdown,
Chuck Lever9903cd12005-08-11 16:25:26 -04002655 .destroy = xs_destroy,
Chuck Lever262ca072006-03-20 13:44:16 -05002656 .print_stats = xs_tcp_print_stats,
Jeff Laytond67fa4d2015-06-03 16:14:29 -04002657 .enable_swap = xs_enable_swap,
2658 .disable_swap = xs_disable_swap,
Chuck Lever4a068252015-05-11 14:02:25 -04002659 .inject_disconnect = xs_inject_disconnect,
Chuck Lever42e5c3e2015-10-24 17:27:35 -04002660#ifdef CONFIG_SUNRPC_BACKCHANNEL
2661 .bc_setup = xprt_setup_bc,
Chuck Lever76566772015-10-24 17:28:32 -04002662 .bc_up = xs_tcp_bc_up,
Chuck Lever42e5c3e2015-10-24 17:27:35 -04002663 .bc_free_rqst = xprt_free_bc_rqst,
2664 .bc_destroy = xprt_destroy_bc,
2665#endif
Chuck Levera246b012005-08-11 16:25:23 -04002666};
2667
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002668/*
2669 * The rpc_xprt_ops for the server backchannel
2670 */
2671
2672static struct rpc_xprt_ops bc_tcp_ops = {
2673 .reserve_xprt = xprt_reserve_xprt,
2674 .release_xprt = xprt_release_xprt,
Bryan Schumaker84e28a32012-09-24 13:39:01 -04002675 .alloc_slot = xprt_alloc_slot,
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002676 .buf_alloc = bc_malloc,
2677 .buf_free = bc_free,
2678 .send_request = bc_send_request,
2679 .set_retrans_timeout = xprt_set_retrans_timeout_def,
2680 .close = bc_close,
2681 .destroy = bc_destroy,
2682 .print_stats = xs_tcp_print_stats,
Jeff Laytond67fa4d2015-06-03 16:14:29 -04002683 .enable_swap = xs_enable_swap,
2684 .disable_swap = xs_disable_swap,
Chuck Lever4a068252015-05-11 14:02:25 -04002685 .inject_disconnect = xs_inject_disconnect,
Rahul Iyer4cfc7e62009-09-10 17:32:28 +03002686};
2687
Chuck Lever92476852010-10-20 11:53:01 -04002688static int xs_init_anyaddr(const int family, struct sockaddr *sap)
2689{
2690 static const struct sockaddr_in sin = {
2691 .sin_family = AF_INET,
2692 .sin_addr.s_addr = htonl(INADDR_ANY),
2693 };
2694 static const struct sockaddr_in6 sin6 = {
2695 .sin6_family = AF_INET6,
2696 .sin6_addr = IN6ADDR_ANY_INIT,
2697 };
2698
2699 switch (family) {
Chuck Lever176e21e2011-05-09 15:22:44 -04002700 case AF_LOCAL:
2701 break;
Chuck Lever92476852010-10-20 11:53:01 -04002702 case AF_INET:
2703 memcpy(sap, &sin, sizeof(sin));
2704 break;
2705 case AF_INET6:
2706 memcpy(sap, &sin6, sizeof(sin6));
2707 break;
2708 default:
2709 dprintk("RPC: %s: Bad address family\n", __func__);
2710 return -EAFNOSUPPORT;
2711 }
2712 return 0;
2713}
2714
\"Talpey, Thomas\3c341b0b2007-09-10 13:47:07 -04002715static struct rpc_xprt *xs_setup_xprt(struct xprt_create *args,
Trond Myklebustd9ba1312011-07-17 18:11:30 -04002716 unsigned int slot_table_size,
2717 unsigned int max_slot_table_size)
Chuck Leverc8541ec2006-10-17 14:44:27 -04002718{
2719 struct rpc_xprt *xprt;
Chuck Leverffc2e512006-12-05 16:35:11 -05002720 struct sock_xprt *new;
Chuck Leverc8541ec2006-10-17 14:44:27 -04002721
Frank van Maarseveen96802a02007-07-08 13:08:54 +02002722 if (args->addrlen > sizeof(xprt->addr)) {
Chuck Lever46121cf2007-01-31 12:14:08 -05002723 dprintk("RPC: xs_setup_xprt: address too large\n");
Chuck Leverc8541ec2006-10-17 14:44:27 -04002724 return ERR_PTR(-EBADF);
2725 }
2726
Trond Myklebustd9ba1312011-07-17 18:11:30 -04002727 xprt = xprt_alloc(args->net, sizeof(*new), slot_table_size,
2728 max_slot_table_size);
Pavel Emelyanovbd1722d2010-09-29 16:02:43 +04002729 if (xprt == NULL) {
Chuck Lever46121cf2007-01-31 12:14:08 -05002730 dprintk("RPC: xs_setup_xprt: couldn't allocate "
2731 "rpc_xprt\n");
Chuck Leverc8541ec2006-10-17 14:44:27 -04002732 return ERR_PTR(-ENOMEM);
2733 }
2734
Pavel Emelyanovbd1722d2010-09-29 16:02:43 +04002735 new = container_of(xprt, struct sock_xprt, xprt);
Trond Myklebustedc1b012015-10-05 10:53:49 -04002736 mutex_init(&new->recv_mutex);
Frank van Maarseveen96802a02007-07-08 13:08:54 +02002737 memcpy(&xprt->addr, args->dstaddr, args->addrlen);
2738 xprt->addrlen = args->addrlen;
Frank van Maarseveend3bc9a12007-07-09 22:23:35 +02002739 if (args->srcaddr)
Chuck Leverfbfffbd2009-08-09 15:09:46 -04002740 memcpy(&new->srcaddr, args->srcaddr, args->addrlen);
Chuck Lever92476852010-10-20 11:53:01 -04002741 else {
2742 int err;
2743 err = xs_init_anyaddr(args->dstaddr->sa_family,
2744 (struct sockaddr *)&new->srcaddr);
Stanislav Kinsbursky2aa13532011-11-10 14:33:23 +03002745 if (err != 0) {
2746 xprt_free(xprt);
Chuck Lever92476852010-10-20 11:53:01 -04002747 return ERR_PTR(err);
Stanislav Kinsbursky2aa13532011-11-10 14:33:23 +03002748 }
Chuck Lever92476852010-10-20 11:53:01 -04002749 }
Chuck Leverc8541ec2006-10-17 14:44:27 -04002750
2751 return xprt;
2752}
2753
Chuck Lever176e21e2011-05-09 15:22:44 -04002754static const struct rpc_timeout xs_local_default_timeout = {
2755 .to_initval = 10 * HZ,
2756 .to_maxval = 10 * HZ,
2757 .to_retries = 2,
2758};
2759
2760/**
2761 * xs_setup_local - Set up transport to use an AF_LOCAL socket
2762 * @args: rpc transport creation arguments
2763 *
2764 * AF_LOCAL is a "tpi_cots_ord" transport, just like TCP
2765 */
2766static struct rpc_xprt *xs_setup_local(struct xprt_create *args)
2767{
2768 struct sockaddr_un *sun = (struct sockaddr_un *)args->dstaddr;
2769 struct sock_xprt *transport;
2770 struct rpc_xprt *xprt;
2771 struct rpc_xprt *ret;
2772
Trond Myklebustd9ba1312011-07-17 18:11:30 -04002773 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
2774 xprt_max_tcp_slot_table_entries);
Chuck Lever176e21e2011-05-09 15:22:44 -04002775 if (IS_ERR(xprt))
2776 return xprt;
2777 transport = container_of(xprt, struct sock_xprt, xprt);
2778
2779 xprt->prot = 0;
2780 xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
2781 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
2782
2783 xprt->bind_timeout = XS_BIND_TO;
2784 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2785 xprt->idle_timeout = XS_IDLE_DISC_TO;
2786
2787 xprt->ops = &xs_local_ops;
2788 xprt->timeout = &xs_local_default_timeout;
2789
Trond Myklebusta2648092015-10-06 17:03:00 -04002790 INIT_WORK(&transport->recv_worker, xs_local_data_receive_workfn);
NeilBrown93dc41b2013-10-31 16:14:36 +11002791 INIT_DELAYED_WORK(&transport->connect_worker,
2792 xs_dummy_setup_socket);
2793
Chuck Lever176e21e2011-05-09 15:22:44 -04002794 switch (sun->sun_family) {
2795 case AF_LOCAL:
2796 if (sun->sun_path[0] != '/') {
2797 dprintk("RPC: bad AF_LOCAL address: %s\n",
2798 sun->sun_path);
2799 ret = ERR_PTR(-EINVAL);
2800 goto out_err;
2801 }
2802 xprt_set_bound(xprt);
Chuck Lever176e21e2011-05-09 15:22:44 -04002803 xs_format_peer_addresses(xprt, "local", RPCBIND_NETID_LOCAL);
J. Bruce Fields7073ea82013-02-21 10:14:22 -05002804 ret = ERR_PTR(xs_local_setup_socket(transport));
2805 if (ret)
2806 goto out_err;
Chuck Lever176e21e2011-05-09 15:22:44 -04002807 break;
2808 default:
2809 ret = ERR_PTR(-EAFNOSUPPORT);
2810 goto out_err;
2811 }
2812
2813 dprintk("RPC: set up xprt to %s via AF_LOCAL\n",
2814 xprt->address_strings[RPC_DISPLAY_ADDR]);
2815
2816 if (try_module_get(THIS_MODULE))
2817 return xprt;
2818 ret = ERR_PTR(-EINVAL);
2819out_err:
Kinglong Mee315f3812014-03-24 11:07:22 +08002820 xs_xprt_free(xprt);
Chuck Lever176e21e2011-05-09 15:22:44 -04002821 return ret;
2822}
2823
Trond Myklebust2881ae72007-12-20 16:03:54 -05002824static const struct rpc_timeout xs_udp_default_timeout = {
2825 .to_initval = 5 * HZ,
2826 .to_maxval = 30 * HZ,
2827 .to_increment = 5 * HZ,
2828 .to_retries = 5,
2829};
2830
Chuck Lever9903cd12005-08-11 16:25:26 -04002831/**
2832 * xs_setup_udp - Set up transport to use a UDP socket
Frank van Maarseveen96802a02007-07-08 13:08:54 +02002833 * @args: rpc transport creation arguments
Chuck Lever9903cd12005-08-11 16:25:26 -04002834 *
2835 */
Adrian Bunk483066d2007-10-24 18:24:02 +02002836static struct rpc_xprt *xs_setup_udp(struct xprt_create *args)
Chuck Levera246b012005-08-11 16:25:23 -04002837{
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002838 struct sockaddr *addr = args->dstaddr;
Chuck Leverc8541ec2006-10-17 14:44:27 -04002839 struct rpc_xprt *xprt;
Chuck Leverc8475462006-12-05 16:35:26 -05002840 struct sock_xprt *transport;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002841 struct rpc_xprt *ret;
Chuck Levera246b012005-08-11 16:25:23 -04002842
Trond Myklebustd9ba1312011-07-17 18:11:30 -04002843 xprt = xs_setup_xprt(args, xprt_udp_slot_table_entries,
2844 xprt_udp_slot_table_entries);
Chuck Leverc8541ec2006-10-17 14:44:27 -04002845 if (IS_ERR(xprt))
2846 return xprt;
Chuck Leverc8475462006-12-05 16:35:26 -05002847 transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -04002848
Chuck Leverec739ef2006-08-22 20:06:15 -04002849 xprt->prot = IPPROTO_UDP;
Chuck Lever808012f2005-08-25 16:25:49 -07002850 xprt->tsh_size = 0;
Chuck Levera246b012005-08-11 16:25:23 -04002851 /* XXX: header size can vary due to auth type, IPv6, etc. */
2852 xprt->max_payload = (1U << 16) - (MAX_HEADER << 3);
2853
Chuck Lever03bf4b72005-08-25 16:25:55 -07002854 xprt->bind_timeout = XS_BIND_TO;
Chuck Lever03bf4b72005-08-25 16:25:55 -07002855 xprt->reestablish_timeout = XS_UDP_REEST_TO;
2856 xprt->idle_timeout = XS_IDLE_DISC_TO;
Chuck Levera246b012005-08-11 16:25:23 -04002857
Chuck Lever262965f2005-08-11 16:25:56 -04002858 xprt->ops = &xs_udp_ops;
Chuck Levera246b012005-08-11 16:25:23 -04002859
Trond Myklebustba7392b2007-12-20 16:03:55 -05002860 xprt->timeout = &xs_udp_default_timeout;
Chuck Levera246b012005-08-11 16:25:23 -04002861
Trond Myklebustf9b2ee72015-10-06 16:26:05 -04002862 INIT_WORK(&transport->recv_worker, xs_udp_data_receive_workfn);
Trond Myklebustedc1b012015-10-05 10:53:49 -04002863 INIT_DELAYED_WORK(&transport->connect_worker, xs_udp_setup_socket);
2864
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002865 switch (addr->sa_family) {
2866 case AF_INET:
2867 if (((struct sockaddr_in *)addr)->sin_port != htons(0))
2868 xprt_set_bound(xprt);
2869
Chuck Lever9dc3b092009-08-09 15:09:46 -04002870 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP);
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002871 break;
2872 case AF_INET6:
2873 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
2874 xprt_set_bound(xprt);
2875
Chuck Lever9dc3b092009-08-09 15:09:46 -04002876 xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP6);
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002877 break;
2878 default:
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002879 ret = ERR_PTR(-EAFNOSUPPORT);
2880 goto out_err;
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002881 }
2882
Chuck Leverc740eff2009-08-09 15:09:46 -04002883 if (xprt_bound(xprt))
2884 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
2885 xprt->address_strings[RPC_DISPLAY_ADDR],
2886 xprt->address_strings[RPC_DISPLAY_PORT],
2887 xprt->address_strings[RPC_DISPLAY_PROTO]);
2888 else
2889 dprintk("RPC: set up xprt to %s (autobind) via %s\n",
2890 xprt->address_strings[RPC_DISPLAY_ADDR],
2891 xprt->address_strings[RPC_DISPLAY_PROTO]);
Chuck Leveredb267a2006-08-22 20:06:18 -04002892
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002893 if (try_module_get(THIS_MODULE))
2894 return xprt;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002895 ret = ERR_PTR(-EINVAL);
2896out_err:
Kinglong Mee315f3812014-03-24 11:07:22 +08002897 xs_xprt_free(xprt);
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002898 return ret;
Chuck Levera246b012005-08-11 16:25:23 -04002899}
2900
Trond Myklebust2881ae72007-12-20 16:03:54 -05002901static const struct rpc_timeout xs_tcp_default_timeout = {
2902 .to_initval = 60 * HZ,
2903 .to_maxval = 60 * HZ,
2904 .to_retries = 2,
2905};
2906
Chuck Lever9903cd12005-08-11 16:25:26 -04002907/**
2908 * xs_setup_tcp - Set up transport to use a TCP socket
Frank van Maarseveen96802a02007-07-08 13:08:54 +02002909 * @args: rpc transport creation arguments
Chuck Lever9903cd12005-08-11 16:25:26 -04002910 *
2911 */
Adrian Bunk483066d2007-10-24 18:24:02 +02002912static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args)
Chuck Levera246b012005-08-11 16:25:23 -04002913{
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002914 struct sockaddr *addr = args->dstaddr;
Chuck Leverc8541ec2006-10-17 14:44:27 -04002915 struct rpc_xprt *xprt;
Chuck Leverc8475462006-12-05 16:35:26 -05002916 struct sock_xprt *transport;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002917 struct rpc_xprt *ret;
Trond Myklebustb7993ce2013-04-14 11:42:00 -04002918 unsigned int max_slot_table_size = xprt_max_tcp_slot_table_entries;
2919
2920 if (args->flags & XPRT_CREATE_INFINITE_SLOTS)
2921 max_slot_table_size = RPC_MAX_SLOT_TABLE_LIMIT;
Chuck Levera246b012005-08-11 16:25:23 -04002922
Trond Myklebustd9ba1312011-07-17 18:11:30 -04002923 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
Trond Myklebustb7993ce2013-04-14 11:42:00 -04002924 max_slot_table_size);
Chuck Leverc8541ec2006-10-17 14:44:27 -04002925 if (IS_ERR(xprt))
2926 return xprt;
Chuck Leverc8475462006-12-05 16:35:26 -05002927 transport = container_of(xprt, struct sock_xprt, xprt);
Chuck Levera246b012005-08-11 16:25:23 -04002928
Chuck Leverec739ef2006-08-22 20:06:15 -04002929 xprt->prot = IPPROTO_TCP;
Chuck Lever808012f2005-08-25 16:25:49 -07002930 xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
Chuck Lever808012f2005-08-25 16:25:49 -07002931 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
Chuck Levera246b012005-08-11 16:25:23 -04002932
Chuck Lever03bf4b72005-08-25 16:25:55 -07002933 xprt->bind_timeout = XS_BIND_TO;
Chuck Lever03bf4b72005-08-25 16:25:55 -07002934 xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO;
2935 xprt->idle_timeout = XS_IDLE_DISC_TO;
Chuck Levera246b012005-08-11 16:25:23 -04002936
Chuck Lever262965f2005-08-11 16:25:56 -04002937 xprt->ops = &xs_tcp_ops;
Trond Myklebustba7392b2007-12-20 16:03:55 -05002938 xprt->timeout = &xs_tcp_default_timeout;
Chuck Levera246b012005-08-11 16:25:23 -04002939
Trond Myklebustedc1b012015-10-05 10:53:49 -04002940 INIT_WORK(&transport->recv_worker, xs_tcp_data_receive_workfn);
2941 INIT_DELAYED_WORK(&transport->connect_worker, xs_tcp_setup_socket);
2942
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002943 switch (addr->sa_family) {
2944 case AF_INET:
2945 if (((struct sockaddr_in *)addr)->sin_port != htons(0))
2946 xprt_set_bound(xprt);
2947
Chuck Lever9dc3b092009-08-09 15:09:46 -04002948 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP);
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002949 break;
2950 case AF_INET6:
2951 if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0))
2952 xprt_set_bound(xprt);
2953
Chuck Lever9dc3b092009-08-09 15:09:46 -04002954 xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP6);
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002955 break;
2956 default:
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002957 ret = ERR_PTR(-EAFNOSUPPORT);
2958 goto out_err;
Chuck Lever8f9d5b12007-08-06 11:57:53 -04002959 }
2960
Chuck Leverc740eff2009-08-09 15:09:46 -04002961 if (xprt_bound(xprt))
2962 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
2963 xprt->address_strings[RPC_DISPLAY_ADDR],
2964 xprt->address_strings[RPC_DISPLAY_PORT],
2965 xprt->address_strings[RPC_DISPLAY_PROTO]);
2966 else
2967 dprintk("RPC: set up xprt to %s (autobind) via %s\n",
2968 xprt->address_strings[RPC_DISPLAY_ADDR],
2969 xprt->address_strings[RPC_DISPLAY_PROTO]);
2970
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04002971 if (try_module_get(THIS_MODULE))
2972 return xprt;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002973 ret = ERR_PTR(-EINVAL);
2974out_err:
Kinglong Mee315f3812014-03-24 11:07:22 +08002975 xs_xprt_free(xprt);
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002976 return ret;
Chuck Levera246b012005-08-11 16:25:23 -04002977}
Chuck Lever282b32e2006-12-05 16:35:51 -05002978
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002979/**
2980 * xs_setup_bc_tcp - Set up transport to use a TCP backchannel socket
2981 * @args: rpc transport creation arguments
2982 *
2983 */
2984static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args)
2985{
2986 struct sockaddr *addr = args->dstaddr;
2987 struct rpc_xprt *xprt;
2988 struct sock_xprt *transport;
2989 struct svc_sock *bc_sock;
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04002990 struct rpc_xprt *ret;
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002991
Trond Myklebustd9ba1312011-07-17 18:11:30 -04002992 xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries,
2993 xprt_tcp_slot_table_entries);
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03002994 if (IS_ERR(xprt))
2995 return xprt;
2996 transport = container_of(xprt, struct sock_xprt, xprt);
2997
2998 xprt->prot = IPPROTO_TCP;
2999 xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32);
3000 xprt->max_payload = RPC_MAX_FRAGMENT_SIZE;
3001 xprt->timeout = &xs_tcp_default_timeout;
3002
3003 /* backchannel */
3004 xprt_set_bound(xprt);
3005 xprt->bind_timeout = 0;
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03003006 xprt->reestablish_timeout = 0;
3007 xprt->idle_timeout = 0;
3008
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03003009 xprt->ops = &bc_tcp_ops;
3010
3011 switch (addr->sa_family) {
3012 case AF_INET:
3013 xs_format_peer_addresses(xprt, "tcp",
3014 RPCBIND_NETID_TCP);
3015 break;
3016 case AF_INET6:
3017 xs_format_peer_addresses(xprt, "tcp",
3018 RPCBIND_NETID_TCP6);
3019 break;
3020 default:
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04003021 ret = ERR_PTR(-EAFNOSUPPORT);
3022 goto out_err;
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03003023 }
3024
Pavel Emelyanov50fa0d42010-10-05 20:49:35 +04003025 dprintk("RPC: set up xprt to %s (port %s) via %s\n",
3026 xprt->address_strings[RPC_DISPLAY_ADDR],
3027 xprt->address_strings[RPC_DISPLAY_PORT],
3028 xprt->address_strings[RPC_DISPLAY_PROTO]);
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03003029
3030 /*
J. Bruce Fields99de8ea2010-12-08 12:45:44 -05003031 * Once we've associated a backchannel xprt with a connection,
Weng Meiling28303ca2013-11-30 17:56:44 +08003032 * we want to keep it around as long as the connection lasts,
3033 * in case we need to start using it for a backchannel again;
3034 * this reference won't be dropped until bc_xprt is destroyed.
J. Bruce Fields99de8ea2010-12-08 12:45:44 -05003035 */
3036 xprt_get(xprt);
3037 args->bc_xprt->xpt_bc_xprt = xprt;
3038 xprt->bc_xprt = args->bc_xprt;
3039 bc_sock = container_of(args->bc_xprt, struct svc_sock, sk_xprt);
3040 transport->sock = bc_sock->sk_sock;
3041 transport->inet = bc_sock->sk_sk;
3042
3043 /*
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03003044 * Since we don't want connections for the backchannel, we set
3045 * the xprt status to connected
3046 */
3047 xprt_set_connected(xprt);
3048
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03003049 if (try_module_get(THIS_MODULE))
3050 return xprt;
Kinglong Mee642aab52014-03-24 12:00:28 +08003051
3052 args->bc_xprt->xpt_bc_xprt = NULL;
J. Bruce Fields99de8ea2010-12-08 12:45:44 -05003053 xprt_put(xprt);
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04003054 ret = ERR_PTR(-EINVAL);
3055out_err:
Kinglong Mee315f3812014-03-24 11:07:22 +08003056 xs_xprt_free(xprt);
J. Bruce Fields0a68b0b2010-05-26 08:42:24 -04003057 return ret;
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03003058}
3059
Chuck Lever176e21e2011-05-09 15:22:44 -04003060static struct xprt_class xs_local_transport = {
3061 .list = LIST_HEAD_INIT(xs_local_transport.list),
3062 .name = "named UNIX socket",
3063 .owner = THIS_MODULE,
3064 .ident = XPRT_TRANSPORT_LOCAL,
3065 .setup = xs_setup_local,
3066};
3067
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04003068static struct xprt_class xs_udp_transport = {
3069 .list = LIST_HEAD_INIT(xs_udp_transport.list),
3070 .name = "udp",
3071 .owner = THIS_MODULE,
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03003072 .ident = XPRT_TRANSPORT_UDP,
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04003073 .setup = xs_setup_udp,
3074};
3075
3076static struct xprt_class xs_tcp_transport = {
3077 .list = LIST_HEAD_INIT(xs_tcp_transport.list),
3078 .name = "tcp",
3079 .owner = THIS_MODULE,
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03003080 .ident = XPRT_TRANSPORT_TCP,
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04003081 .setup = xs_setup_tcp,
3082};
3083
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03003084static struct xprt_class xs_bc_tcp_transport = {
3085 .list = LIST_HEAD_INIT(xs_bc_tcp_transport.list),
3086 .name = "tcp NFSv4.1 backchannel",
3087 .owner = THIS_MODULE,
3088 .ident = XPRT_TRANSPORT_BC_TCP,
3089 .setup = xs_setup_bc_tcp,
3090};
3091
Chuck Lever282b32e2006-12-05 16:35:51 -05003092/**
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04003093 * init_socket_xprt - set up xprtsock's sysctls, register with RPC client
Chuck Lever282b32e2006-12-05 16:35:51 -05003094 *
3095 */
3096int init_socket_xprt(void)
3097{
Jeff Laytonf895b252014-11-17 16:58:04 -05003098#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
Eric W. Biederman2b1bec52007-02-14 00:33:24 -08003099 if (!sunrpc_table_header)
Eric W. Biederman0b4d4142007-02-14 00:34:09 -08003100 sunrpc_table_header = register_sysctl_table(sunrpc_table);
Chuck Leverfbf76682006-12-05 16:35:54 -05003101#endif
3102
Chuck Lever176e21e2011-05-09 15:22:44 -04003103 xprt_register_transport(&xs_local_transport);
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04003104 xprt_register_transport(&xs_udp_transport);
3105 xprt_register_transport(&xs_tcp_transport);
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03003106 xprt_register_transport(&xs_bc_tcp_transport);
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04003107
Chuck Lever282b32e2006-12-05 16:35:51 -05003108 return 0;
3109}
3110
3111/**
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04003112 * cleanup_socket_xprt - remove xprtsock's sysctls, unregister
Chuck Lever282b32e2006-12-05 16:35:51 -05003113 *
3114 */
3115void cleanup_socket_xprt(void)
3116{
Jeff Laytonf895b252014-11-17 16:58:04 -05003117#if IS_ENABLED(CONFIG_SUNRPC_DEBUG)
Chuck Leverfbf76682006-12-05 16:35:54 -05003118 if (sunrpc_table_header) {
3119 unregister_sysctl_table(sunrpc_table_header);
3120 sunrpc_table_header = NULL;
3121 }
3122#endif
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04003123
Chuck Lever176e21e2011-05-09 15:22:44 -04003124 xprt_unregister_transport(&xs_local_transport);
\"Talpey, Thomas\bc255712007-09-10 13:46:39 -04003125 xprt_unregister_transport(&xs_udp_transport);
3126 xprt_unregister_transport(&xs_tcp_transport);
Alexandros Batsakisf300bab2009-09-10 17:33:30 +03003127 xprt_unregister_transport(&xs_bc_tcp_transport);
Chuck Lever282b32e2006-12-05 16:35:51 -05003128}
Trond Myklebustcbf11072009-08-09 15:06:19 -04003129
Rusty Russell9bbb9e52010-08-11 23:04:12 -06003130static int param_set_uint_minmax(const char *val,
3131 const struct kernel_param *kp,
Trond Myklebustcbf11072009-08-09 15:06:19 -04003132 unsigned int min, unsigned int max)
3133{
Daniel Walter00cfaa92014-06-21 13:06:38 +01003134 unsigned int num;
Trond Myklebustcbf11072009-08-09 15:06:19 -04003135 int ret;
3136
3137 if (!val)
3138 return -EINVAL;
Daniel Walter00cfaa92014-06-21 13:06:38 +01003139 ret = kstrtouint(val, 0, &num);
Trond Myklebustcbf11072009-08-09 15:06:19 -04003140 if (ret == -EINVAL || num < min || num > max)
3141 return -EINVAL;
3142 *((unsigned int *)kp->arg) = num;
3143 return 0;
3144}
3145
Rusty Russell9bbb9e52010-08-11 23:04:12 -06003146static int param_set_portnr(const char *val, const struct kernel_param *kp)
Trond Myklebustcbf11072009-08-09 15:06:19 -04003147{
3148 return param_set_uint_minmax(val, kp,
3149 RPC_MIN_RESVPORT,
3150 RPC_MAX_RESVPORT);
3151}
3152
Luis R. Rodriguez9c278472015-05-27 11:09:38 +09303153static const struct kernel_param_ops param_ops_portnr = {
Rusty Russell9bbb9e52010-08-11 23:04:12 -06003154 .set = param_set_portnr,
3155 .get = param_get_uint,
3156};
3157
Trond Myklebustcbf11072009-08-09 15:06:19 -04003158#define param_check_portnr(name, p) \
3159 __param_check(name, p, unsigned int);
3160
3161module_param_named(min_resvport, xprt_min_resvport, portnr, 0644);
3162module_param_named(max_resvport, xprt_max_resvport, portnr, 0644);
3163
Rusty Russell9bbb9e52010-08-11 23:04:12 -06003164static int param_set_slot_table_size(const char *val,
3165 const struct kernel_param *kp)
Trond Myklebustcbf11072009-08-09 15:06:19 -04003166{
3167 return param_set_uint_minmax(val, kp,
3168 RPC_MIN_SLOT_TABLE,
3169 RPC_MAX_SLOT_TABLE);
3170}
3171
Luis R. Rodriguez9c278472015-05-27 11:09:38 +09303172static const struct kernel_param_ops param_ops_slot_table_size = {
Rusty Russell9bbb9e52010-08-11 23:04:12 -06003173 .set = param_set_slot_table_size,
3174 .get = param_get_uint,
3175};
3176
Trond Myklebustcbf11072009-08-09 15:06:19 -04003177#define param_check_slot_table_size(name, p) \
3178 __param_check(name, p, unsigned int);
3179
Trond Myklebustd9ba1312011-07-17 18:11:30 -04003180static int param_set_max_slot_table_size(const char *val,
3181 const struct kernel_param *kp)
3182{
3183 return param_set_uint_minmax(val, kp,
3184 RPC_MIN_SLOT_TABLE,
3185 RPC_MAX_SLOT_TABLE_LIMIT);
3186}
3187
Luis R. Rodriguez9c278472015-05-27 11:09:38 +09303188static const struct kernel_param_ops param_ops_max_slot_table_size = {
Trond Myklebustd9ba1312011-07-17 18:11:30 -04003189 .set = param_set_max_slot_table_size,
3190 .get = param_get_uint,
3191};
3192
3193#define param_check_max_slot_table_size(name, p) \
3194 __param_check(name, p, unsigned int);
3195
Trond Myklebustcbf11072009-08-09 15:06:19 -04003196module_param_named(tcp_slot_table_entries, xprt_tcp_slot_table_entries,
3197 slot_table_size, 0644);
Trond Myklebustd9ba1312011-07-17 18:11:30 -04003198module_param_named(tcp_max_slot_table_entries, xprt_max_tcp_slot_table_entries,
3199 max_slot_table_size, 0644);
Trond Myklebustcbf11072009-08-09 15:06:19 -04003200module_param_named(udp_slot_table_entries, xprt_udp_slot_table_entries,
3201 slot_table_size, 0644);
3202