Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
a246b010 CL |
2 | /* |
3 | * linux/net/sunrpc/xprtsock.c | |
4 | * | |
5 | * Client-side transport implementation for sockets. | |
6 | * | |
113aa838 AC |
7 | * TCP callback races fixes (C) 1998 Red Hat |
8 | * TCP send fixes (C) 1998 Red Hat | |
a246b010 CL |
9 | * TCP NFS related read + write fixes |
10 | * (C) 1999 Dave Airlie, University of Limerick, Ireland <airlied@linux.ie> | |
11 | * | |
12 | * Rewrite of larges part of the code in order to stabilize TCP stuff. | |
13 | * Fix behaviour when socket buffer is full. | |
14 | * (C) 1999 Trond Myklebust <trond.myklebust@fys.uio.no> | |
55aa4f58 CL |
15 | * |
16 | * IP socket transport implementation, (C) 2005 Chuck Lever <cel@netapp.com> | |
8f9d5b1a CL |
17 | * |
18 | * IPv6 support contributed by Gilles Quillard, Bull Open Source, 2005. | |
19 | * <gilles.quillard@bull.net> | |
a246b010 CL |
20 | */ |
21 | ||
22 | #include <linux/types.h> | |
176e21ee | 23 | #include <linux/string.h> |
a246b010 | 24 | #include <linux/slab.h> |
bc25571e | 25 | #include <linux/module.h> |
a246b010 | 26 | #include <linux/capability.h> |
a246b010 CL |
27 | #include <linux/pagemap.h> |
28 | #include <linux/errno.h> | |
29 | #include <linux/socket.h> | |
30 | #include <linux/in.h> | |
31 | #include <linux/net.h> | |
32 | #include <linux/mm.h> | |
176e21ee | 33 | #include <linux/un.h> |
a246b010 CL |
34 | #include <linux/udp.h> |
35 | #include <linux/tcp.h> | |
36 | #include <linux/sunrpc/clnt.h> | |
5976687a | 37 | #include <linux/sunrpc/addr.h> |
02107148 | 38 | #include <linux/sunrpc/sched.h> |
4cfc7e60 | 39 | #include <linux/sunrpc/svcsock.h> |
49c36fcc | 40 | #include <linux/sunrpc/xprtsock.h> |
a246b010 | 41 | #include <linux/file.h> |
9e00abc3 | 42 | #ifdef CONFIG_SUNRPC_BACKCHANNEL |
44b98efd RL |
43 | #include <linux/sunrpc/bc_xprt.h> |
44 | #endif | |
a246b010 CL |
45 | |
46 | #include <net/sock.h> | |
47 | #include <net/checksum.h> | |
48 | #include <net/udp.h> | |
49 | #include <net/tcp.h> | |
277e4ab7 | 50 | #include <linux/bvec.h> |
6a829eb8 | 51 | #include <linux/highmem.h> |
277e4ab7 | 52 | #include <linux/uio.h> |
a1231fda | 53 | #include <linux/sched/mm.h> |
a246b010 | 54 | |
40b5ea0c TM |
55 | #include <trace/events/sunrpc.h> |
56 | ||
9e55eef4 | 57 | #include "socklib.h" |
4cfc7e60 | 58 | #include "sunrpc.h" |
176e21ee CL |
59 | |
60 | static void xs_close(struct rpc_xprt *xprt); | |
7196dbb0 TM |
61 | static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt, |
62 | struct socket *sock); | |
176e21ee | 63 | |
c556b754 CL |
64 | /* |
65 | * xprtsock tunables | |
66 | */ | |
09acfea5 TM |
67 | static unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE; |
68 | static unsigned int xprt_tcp_slot_table_entries = RPC_MIN_SLOT_TABLE; | |
69 | static unsigned int xprt_max_tcp_slot_table_entries = RPC_MAX_SLOT_TABLE; | |
c556b754 | 70 | |
09acfea5 TM |
71 | static unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT; |
72 | static unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT; | |
c556b754 | 73 | |
7d1e8255 | 74 | #define XS_TCP_LINGER_TO (15U * HZ) |
25fe6142 | 75 | static unsigned int xs_tcp_fin_timeout __read_mostly = XS_TCP_LINGER_TO; |
7d1e8255 | 76 | |
fbf76683 CL |
77 | /* |
78 | * We can register our own files under /proc/sys/sunrpc by | |
79 | * calling register_sysctl_table() again. The files in that | |
80 | * directory become the union of all files registered there. | |
81 | * | |
82 | * We simply need to make sure that we don't collide with | |
83 | * someone else's file names! | |
84 | */ | |
85 | ||
fbf76683 CL |
86 | static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE; |
87 | static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE; | |
d9ba131d | 88 | static unsigned int max_tcp_slot_table_limit = RPC_MAX_SLOT_TABLE_LIMIT; |
fbf76683 CL |
89 | static unsigned int xprt_min_resvport_limit = RPC_MIN_RESVPORT; |
90 | static unsigned int xprt_max_resvport_limit = RPC_MAX_RESVPORT; | |
91 | ||
92 | static struct ctl_table_header *sunrpc_table_header; | |
93 | ||
94 | /* | |
95 | * FIXME: changing the UDP slot table size should also resize the UDP | |
96 | * socket buffers for existing UDP transports | |
97 | */ | |
fe2c6338 | 98 | static struct ctl_table xs_tunables_table[] = { |
fbf76683 | 99 | { |
fbf76683 CL |
100 | .procname = "udp_slot_table_entries", |
101 | .data = &xprt_udp_slot_table_entries, | |
102 | .maxlen = sizeof(unsigned int), | |
103 | .mode = 0644, | |
6d456111 | 104 | .proc_handler = proc_dointvec_minmax, |
fbf76683 CL |
105 | .extra1 = &min_slot_table_size, |
106 | .extra2 = &max_slot_table_size | |
107 | }, | |
108 | { | |
fbf76683 CL |
109 | .procname = "tcp_slot_table_entries", |
110 | .data = &xprt_tcp_slot_table_entries, | |
111 | .maxlen = sizeof(unsigned int), | |
112 | .mode = 0644, | |
6d456111 | 113 | .proc_handler = proc_dointvec_minmax, |
fbf76683 CL |
114 | .extra1 = &min_slot_table_size, |
115 | .extra2 = &max_slot_table_size | |
116 | }, | |
d9ba131d TM |
117 | { |
118 | .procname = "tcp_max_slot_table_entries", | |
119 | .data = &xprt_max_tcp_slot_table_entries, | |
120 | .maxlen = sizeof(unsigned int), | |
121 | .mode = 0644, | |
122 | .proc_handler = proc_dointvec_minmax, | |
123 | .extra1 = &min_slot_table_size, | |
124 | .extra2 = &max_tcp_slot_table_limit | |
125 | }, | |
fbf76683 | 126 | { |
fbf76683 CL |
127 | .procname = "min_resvport", |
128 | .data = &xprt_min_resvport, | |
129 | .maxlen = sizeof(unsigned int), | |
130 | .mode = 0644, | |
6d456111 | 131 | .proc_handler = proc_dointvec_minmax, |
fbf76683 | 132 | .extra1 = &xprt_min_resvport_limit, |
826799e6 | 133 | .extra2 = &xprt_max_resvport_limit |
fbf76683 CL |
134 | }, |
135 | { | |
fbf76683 CL |
136 | .procname = "max_resvport", |
137 | .data = &xprt_max_resvport, | |
138 | .maxlen = sizeof(unsigned int), | |
139 | .mode = 0644, | |
6d456111 | 140 | .proc_handler = proc_dointvec_minmax, |
826799e6 | 141 | .extra1 = &xprt_min_resvport_limit, |
fbf76683 CL |
142 | .extra2 = &xprt_max_resvport_limit |
143 | }, | |
25fe6142 TM |
144 | { |
145 | .procname = "tcp_fin_timeout", | |
146 | .data = &xs_tcp_fin_timeout, | |
147 | .maxlen = sizeof(xs_tcp_fin_timeout), | |
148 | .mode = 0644, | |
6d456111 | 149 | .proc_handler = proc_dointvec_jiffies, |
fbf76683 | 150 | }, |
f8572d8f | 151 | { }, |
fbf76683 CL |
152 | }; |
153 | ||
fe2c6338 | 154 | static struct ctl_table sunrpc_table[] = { |
fbf76683 | 155 | { |
fbf76683 CL |
156 | .procname = "sunrpc", |
157 | .mode = 0555, | |
158 | .child = xs_tunables_table | |
159 | }, | |
f8572d8f | 160 | { }, |
fbf76683 CL |
161 | }; |
162 | ||
03bf4b70 CL |
163 | /* |
164 | * Wait duration for a reply from the RPC portmapper. | |
165 | */ | |
166 | #define XS_BIND_TO (60U * HZ) | |
167 | ||
168 | /* | |
169 | * Delay if a UDP socket connect error occurs. This is most likely some | |
170 | * kind of resource problem on the local host. | |
171 | */ | |
172 | #define XS_UDP_REEST_TO (2U * HZ) | |
173 | ||
174 | /* | |
175 | * The reestablish timeout allows clients to delay for a bit before attempting | |
176 | * to reconnect to a server that just dropped our connection. | |
177 | * | |
178 | * We implement an exponential backoff when trying to reestablish a TCP | |
179 | * transport connection with the server. Some servers like to drop a TCP | |
180 | * connection when they are overworked, so we start with a short timeout and | |
181 | * increase over time if the server is down or not responding. | |
182 | */ | |
183 | #define XS_TCP_INIT_REEST_TO (3U * HZ) | |
03bf4b70 CL |
184 | |
185 | /* | |
186 | * TCP idle timeout; client drops the transport socket if it is idle | |
187 | * for this long. Note that we also timeout UDP sockets to prevent | |
188 | * holding port numbers when there is no RPC traffic. | |
189 | */ | |
190 | #define XS_IDLE_DISC_TO (5U * 60 * HZ) | |
191 | ||
f895b252 | 192 | #if IS_ENABLED(CONFIG_SUNRPC_DEBUG) |
a246b010 | 193 | # undef RPC_DEBUG_DATA |
9903cd1c | 194 | # define RPCDBG_FACILITY RPCDBG_TRANS |
a246b010 CL |
195 | #endif |
196 | ||
a246b010 | 197 | #ifdef RPC_DEBUG_DATA |
9903cd1c | 198 | static void xs_pktdump(char *msg, u32 *packet, unsigned int count) |
a246b010 | 199 | { |
9903cd1c CL |
200 | u8 *buf = (u8 *) packet; |
201 | int j; | |
a246b010 | 202 | |
46121cf7 | 203 | dprintk("RPC: %s\n", msg); |
a246b010 CL |
204 | for (j = 0; j < count && j < 128; j += 4) { |
205 | if (!(j & 31)) { | |
206 | if (j) | |
207 | dprintk("\n"); | |
208 | dprintk("0x%04x ", j); | |
209 | } | |
210 | dprintk("%02x%02x%02x%02x ", | |
211 | buf[j], buf[j+1], buf[j+2], buf[j+3]); | |
212 | } | |
213 | dprintk("\n"); | |
214 | } | |
215 | #else | |
9903cd1c | 216 | static inline void xs_pktdump(char *msg, u32 *packet, unsigned int count) |
a246b010 CL |
217 | { |
218 | /* NOP */ | |
219 | } | |
220 | #endif | |
221 | ||
2118071d TM |
222 | static inline struct rpc_xprt *xprt_from_sock(struct sock *sk) |
223 | { | |
224 | return (struct rpc_xprt *) sk->sk_user_data; | |
225 | } | |
226 | ||
95392c59 CL |
227 | static inline struct sockaddr *xs_addr(struct rpc_xprt *xprt) |
228 | { | |
229 | return (struct sockaddr *) &xprt->addr; | |
230 | } | |
231 | ||
176e21ee CL |
232 | static inline struct sockaddr_un *xs_addr_un(struct rpc_xprt *xprt) |
233 | { | |
234 | return (struct sockaddr_un *) &xprt->addr; | |
235 | } | |
236 | ||
95392c59 | 237 | static inline struct sockaddr_in *xs_addr_in(struct rpc_xprt *xprt) |
edb267a6 | 238 | { |
95392c59 CL |
239 | return (struct sockaddr_in *) &xprt->addr; |
240 | } | |
241 | ||
242 | static inline struct sockaddr_in6 *xs_addr_in6(struct rpc_xprt *xprt) | |
243 | { | |
244 | return (struct sockaddr_in6 *) &xprt->addr; | |
245 | } | |
246 | ||
c877b849 | 247 | static void xs_format_common_peer_addresses(struct rpc_xprt *xprt) |
edb267a6 | 248 | { |
c877b849 | 249 | struct sockaddr *sap = xs_addr(xprt); |
9dc3b095 CL |
250 | struct sockaddr_in6 *sin6; |
251 | struct sockaddr_in *sin; | |
176e21ee | 252 | struct sockaddr_un *sun; |
c877b849 | 253 | char buf[128]; |
edb267a6 | 254 | |
9dc3b095 | 255 | switch (sap->sa_family) { |
176e21ee CL |
256 | case AF_LOCAL: |
257 | sun = xs_addr_un(xprt); | |
258 | strlcpy(buf, sun->sun_path, sizeof(buf)); | |
259 | xprt->address_strings[RPC_DISPLAY_ADDR] = | |
260 | kstrdup(buf, GFP_KERNEL); | |
261 | break; | |
9dc3b095 | 262 | case AF_INET: |
176e21ee CL |
263 | (void)rpc_ntop(sap, buf, sizeof(buf)); |
264 | xprt->address_strings[RPC_DISPLAY_ADDR] = | |
265 | kstrdup(buf, GFP_KERNEL); | |
9dc3b095 | 266 | sin = xs_addr_in(xprt); |
fc0b5791 | 267 | snprintf(buf, sizeof(buf), "%08x", ntohl(sin->sin_addr.s_addr)); |
9dc3b095 CL |
268 | break; |
269 | case AF_INET6: | |
176e21ee CL |
270 | (void)rpc_ntop(sap, buf, sizeof(buf)); |
271 | xprt->address_strings[RPC_DISPLAY_ADDR] = | |
272 | kstrdup(buf, GFP_KERNEL); | |
9dc3b095 | 273 | sin6 = xs_addr_in6(xprt); |
fc0b5791 | 274 | snprintf(buf, sizeof(buf), "%pi6", &sin6->sin6_addr); |
9dc3b095 CL |
275 | break; |
276 | default: | |
277 | BUG(); | |
756805e7 | 278 | } |
176e21ee | 279 | |
9dc3b095 | 280 | xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = kstrdup(buf, GFP_KERNEL); |
edb267a6 CL |
281 | } |
282 | ||
9dc3b095 | 283 | static void xs_format_common_peer_ports(struct rpc_xprt *xprt) |
4b6473fb | 284 | { |
9dc3b095 CL |
285 | struct sockaddr *sap = xs_addr(xprt); |
286 | char buf[128]; | |
4b6473fb | 287 | |
81160e66 | 288 | snprintf(buf, sizeof(buf), "%u", rpc_get_port(sap)); |
c877b849 | 289 | xprt->address_strings[RPC_DISPLAY_PORT] = kstrdup(buf, GFP_KERNEL); |
4b6473fb | 290 | |
81160e66 | 291 | snprintf(buf, sizeof(buf), "%4hx", rpc_get_port(sap)); |
c877b849 CL |
292 | xprt->address_strings[RPC_DISPLAY_HEX_PORT] = kstrdup(buf, GFP_KERNEL); |
293 | } | |
4b6473fb | 294 | |
9dc3b095 CL |
295 | static void xs_format_peer_addresses(struct rpc_xprt *xprt, |
296 | const char *protocol, | |
297 | const char *netid) | |
c877b849 | 298 | { |
b454ae90 | 299 | xprt->address_strings[RPC_DISPLAY_PROTO] = protocol; |
b454ae90 | 300 | xprt->address_strings[RPC_DISPLAY_NETID] = netid; |
c877b849 | 301 | xs_format_common_peer_addresses(xprt); |
9dc3b095 | 302 | xs_format_common_peer_ports(xprt); |
edb267a6 | 303 | } |
4b6473fb | 304 | |
9dc3b095 | 305 | static void xs_update_peer_port(struct rpc_xprt *xprt) |
4b6473fb | 306 | { |
9dc3b095 CL |
307 | kfree(xprt->address_strings[RPC_DISPLAY_HEX_PORT]); |
308 | kfree(xprt->address_strings[RPC_DISPLAY_PORT]); | |
4417c8c4 | 309 | |
9dc3b095 | 310 | xs_format_common_peer_ports(xprt); |
edb267a6 CL |
311 | } |
312 | ||
313 | static void xs_free_peer_addresses(struct rpc_xprt *xprt) | |
314 | { | |
33e01dc7 CL |
315 | unsigned int i; |
316 | ||
317 | for (i = 0; i < RPC_DISPLAY_MAX; i++) | |
318 | switch (i) { | |
319 | case RPC_DISPLAY_PROTO: | |
320 | case RPC_DISPLAY_NETID: | |
321 | continue; | |
322 | default: | |
323 | kfree(xprt->address_strings[i]); | |
324 | } | |
edb267a6 CL |
325 | } |
326 | ||
277e4ab7 TM |
327 | static size_t |
328 | xs_alloc_sparse_pages(struct xdr_buf *buf, size_t want, gfp_t gfp) | |
329 | { | |
330 | size_t i,n; | |
331 | ||
16e5e90f | 332 | if (!want || !(buf->flags & XDRBUF_SPARSE_PAGES)) |
277e4ab7 | 333 | return want; |
277e4ab7 TM |
334 | n = (buf->page_base + want + PAGE_SIZE - 1) >> PAGE_SHIFT; |
335 | for (i = 0; i < n; i++) { | |
336 | if (buf->pages[i]) | |
337 | continue; | |
338 | buf->bvec[i].bv_page = buf->pages[i] = alloc_page(gfp); | |
339 | if (!buf->pages[i]) { | |
16e5e90f TM |
340 | i *= PAGE_SIZE; |
341 | return i > buf->page_base ? i - buf->page_base : 0; | |
277e4ab7 TM |
342 | } |
343 | } | |
344 | return want; | |
345 | } | |
346 | ||
347 | static ssize_t | |
348 | xs_sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags, size_t seek) | |
349 | { | |
350 | ssize_t ret; | |
351 | if (seek != 0) | |
352 | iov_iter_advance(&msg->msg_iter, seek); | |
353 | ret = sock_recvmsg(sock, msg, flags); | |
354 | return ret > 0 ? ret + seek : ret; | |
355 | } | |
356 | ||
357 | static ssize_t | |
358 | xs_read_kvec(struct socket *sock, struct msghdr *msg, int flags, | |
359 | struct kvec *kvec, size_t count, size_t seek) | |
360 | { | |
0e9b4a82 | 361 | iov_iter_kvec(&msg->msg_iter, READ, kvec, 1, count); |
277e4ab7 TM |
362 | return xs_sock_recvmsg(sock, msg, flags, seek); |
363 | } | |
364 | ||
365 | static ssize_t | |
366 | xs_read_bvec(struct socket *sock, struct msghdr *msg, int flags, | |
367 | struct bio_vec *bvec, unsigned long nr, size_t count, | |
368 | size_t seek) | |
369 | { | |
0e9b4a82 | 370 | iov_iter_bvec(&msg->msg_iter, READ, bvec, nr, count); |
277e4ab7 TM |
371 | return xs_sock_recvmsg(sock, msg, flags, seek); |
372 | } | |
373 | ||
374 | static ssize_t | |
375 | xs_read_discard(struct socket *sock, struct msghdr *msg, int flags, | |
376 | size_t count) | |
377 | { | |
b76a5afd TM |
378 | iov_iter_discard(&msg->msg_iter, READ, count); |
379 | return sock_recvmsg(sock, msg, flags); | |
277e4ab7 TM |
380 | } |
381 | ||
6a829eb8 TM |
382 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
383 | static void | |
384 | xs_flush_bvec(const struct bio_vec *bvec, size_t count, size_t seek) | |
385 | { | |
386 | struct bvec_iter bi = { | |
387 | .bi_size = count, | |
388 | }; | |
389 | struct bio_vec bv; | |
390 | ||
391 | bvec_iter_advance(bvec, &bi, seek & PAGE_MASK); | |
392 | for_each_bvec(bv, bvec, bi, bi) | |
393 | flush_dcache_page(bv.bv_page); | |
394 | } | |
395 | #else | |
396 | static inline void | |
397 | xs_flush_bvec(const struct bio_vec *bvec, size_t count, size_t seek) | |
398 | { | |
399 | } | |
400 | #endif | |
401 | ||
277e4ab7 TM |
402 | static ssize_t |
403 | xs_read_xdr_buf(struct socket *sock, struct msghdr *msg, int flags, | |
404 | struct xdr_buf *buf, size_t count, size_t seek, size_t *read) | |
405 | { | |
406 | size_t want, seek_init = seek, offset = 0; | |
407 | ssize_t ret; | |
408 | ||
e92053a5 TM |
409 | want = min_t(size_t, count, buf->head[0].iov_len); |
410 | if (seek < want) { | |
277e4ab7 TM |
411 | ret = xs_read_kvec(sock, msg, flags, &buf->head[0], want, seek); |
412 | if (ret <= 0) | |
413 | goto sock_err; | |
414 | offset += ret; | |
415 | if (offset == count || msg->msg_flags & (MSG_EOR|MSG_TRUNC)) | |
416 | goto out; | |
417 | if (ret != want) | |
c4433055 | 418 | goto out; |
277e4ab7 TM |
419 | seek = 0; |
420 | } else { | |
e92053a5 TM |
421 | seek -= want; |
422 | offset += want; | |
277e4ab7 | 423 | } |
16e5e90f TM |
424 | |
425 | want = xs_alloc_sparse_pages(buf, | |
426 | min_t(size_t, count - offset, buf->page_len), | |
12a3ad61 | 427 | GFP_KERNEL); |
16e5e90f | 428 | if (seek < want) { |
277e4ab7 TM |
429 | ret = xs_read_bvec(sock, msg, flags, buf->bvec, |
430 | xdr_buf_pagecount(buf), | |
431 | want + buf->page_base, | |
432 | seek + buf->page_base); | |
433 | if (ret <= 0) | |
434 | goto sock_err; | |
6a829eb8 | 435 | xs_flush_bvec(buf->bvec, ret, seek + buf->page_base); |
ac9645c8 DA |
436 | ret -= buf->page_base; |
437 | offset += ret; | |
277e4ab7 TM |
438 | if (offset == count || msg->msg_flags & (MSG_EOR|MSG_TRUNC)) |
439 | goto out; | |
440 | if (ret != want) | |
c4433055 | 441 | goto out; |
277e4ab7 TM |
442 | seek = 0; |
443 | } else { | |
16e5e90f TM |
444 | seek -= want; |
445 | offset += want; | |
277e4ab7 | 446 | } |
16e5e90f | 447 | |
e92053a5 TM |
448 | want = min_t(size_t, count - offset, buf->tail[0].iov_len); |
449 | if (seek < want) { | |
277e4ab7 TM |
450 | ret = xs_read_kvec(sock, msg, flags, &buf->tail[0], want, seek); |
451 | if (ret <= 0) | |
452 | goto sock_err; | |
453 | offset += ret; | |
454 | if (offset == count || msg->msg_flags & (MSG_EOR|MSG_TRUNC)) | |
455 | goto out; | |
456 | if (ret != want) | |
c4433055 | 457 | goto out; |
9734ad57 | 458 | } else if (offset < seek_init) |
e92053a5 | 459 | offset = seek_init; |
277e4ab7 | 460 | ret = -EMSGSIZE; |
277e4ab7 TM |
461 | out: |
462 | *read = offset - seek_init; | |
463 | return ret; | |
277e4ab7 TM |
464 | sock_err: |
465 | offset += seek; | |
466 | goto out; | |
467 | } | |
468 | ||
469 | static void | |
470 | xs_read_header(struct sock_xprt *transport, struct xdr_buf *buf) | |
471 | { | |
472 | if (!transport->recv.copied) { | |
473 | if (buf->head[0].iov_len >= transport->recv.offset) | |
474 | memcpy(buf->head[0].iov_base, | |
475 | &transport->recv.xid, | |
476 | transport->recv.offset); | |
477 | transport->recv.copied = transport->recv.offset; | |
478 | } | |
479 | } | |
480 | ||
481 | static bool | |
482 | xs_read_stream_request_done(struct sock_xprt *transport) | |
483 | { | |
484 | return transport->recv.fraghdr & cpu_to_be32(RPC_LAST_STREAM_FRAGMENT); | |
485 | } | |
486 | ||
e92053a5 TM |
487 | static void |
488 | xs_read_stream_check_eor(struct sock_xprt *transport, | |
489 | struct msghdr *msg) | |
490 | { | |
491 | if (xs_read_stream_request_done(transport)) | |
492 | msg->msg_flags |= MSG_EOR; | |
493 | } | |
494 | ||
277e4ab7 TM |
495 | static ssize_t |
496 | xs_read_stream_request(struct sock_xprt *transport, struct msghdr *msg, | |
497 | int flags, struct rpc_rqst *req) | |
498 | { | |
499 | struct xdr_buf *buf = &req->rq_private_buf; | |
3f649ab7 KC |
500 | size_t want, read; |
501 | ssize_t ret; | |
277e4ab7 TM |
502 | |
503 | xs_read_header(transport, buf); | |
504 | ||
505 | want = transport->recv.len - transport->recv.offset; | |
e92053a5 TM |
506 | if (want != 0) { |
507 | ret = xs_read_xdr_buf(transport->sock, msg, flags, buf, | |
508 | transport->recv.copied + want, | |
509 | transport->recv.copied, | |
510 | &read); | |
511 | transport->recv.offset += read; | |
512 | transport->recv.copied += read; | |
727fcc64 | 513 | } |
e92053a5 | 514 | |
727fcc64 | 515 | if (transport->recv.offset == transport->recv.len) |
e92053a5 | 516 | xs_read_stream_check_eor(transport, msg); |
277e4ab7 | 517 | |
e92053a5 TM |
518 | if (want == 0) |
519 | return 0; | |
520 | ||
277e4ab7 | 521 | switch (ret) { |
c4433055 TM |
522 | default: |
523 | break; | |
26781eab | 524 | case -EFAULT: |
277e4ab7 | 525 | case -EMSGSIZE: |
26781eab | 526 | msg->msg_flags |= MSG_TRUNC; |
c4433055 | 527 | return read; |
277e4ab7 TM |
528 | case 0: |
529 | return -ESHUTDOWN; | |
277e4ab7 | 530 | } |
c4433055 | 531 | return ret < 0 ? ret : read; |
277e4ab7 TM |
532 | } |
533 | ||
534 | static size_t | |
535 | xs_read_stream_headersize(bool isfrag) | |
536 | { | |
537 | if (isfrag) | |
538 | return sizeof(__be32); | |
539 | return 3 * sizeof(__be32); | |
540 | } | |
541 | ||
542 | static ssize_t | |
543 | xs_read_stream_header(struct sock_xprt *transport, struct msghdr *msg, | |
544 | int flags, size_t want, size_t seek) | |
545 | { | |
546 | struct kvec kvec = { | |
547 | .iov_base = &transport->recv.fraghdr, | |
548 | .iov_len = want, | |
549 | }; | |
550 | return xs_read_kvec(transport->sock, msg, flags, &kvec, want, seek); | |
551 | } | |
552 | ||
553 | #if defined(CONFIG_SUNRPC_BACKCHANNEL) | |
554 | static ssize_t | |
555 | xs_read_stream_call(struct sock_xprt *transport, struct msghdr *msg, int flags) | |
556 | { | |
557 | struct rpc_xprt *xprt = &transport->xprt; | |
558 | struct rpc_rqst *req; | |
559 | ssize_t ret; | |
560 | ||
561 | /* Look up and lock the request corresponding to the given XID */ | |
562 | req = xprt_lookup_bc_request(xprt, transport->recv.xid); | |
563 | if (!req) { | |
564 | printk(KERN_WARNING "Callback slot table overflowed\n"); | |
565 | return -ESHUTDOWN; | |
566 | } | |
45835a63 TM |
567 | if (transport->recv.copied && !req->rq_private_buf.len) |
568 | return -ESHUTDOWN; | |
277e4ab7 TM |
569 | |
570 | ret = xs_read_stream_request(transport, msg, flags, req); | |
571 | if (msg->msg_flags & (MSG_EOR|MSG_TRUNC)) | |
c4433055 | 572 | xprt_complete_bc_request(req, transport->recv.copied); |
45835a63 TM |
573 | else |
574 | req->rq_private_buf.len = transport->recv.copied; | |
277e4ab7 TM |
575 | |
576 | return ret; | |
577 | } | |
578 | #else /* CONFIG_SUNRPC_BACKCHANNEL */ | |
579 | static ssize_t | |
580 | xs_read_stream_call(struct sock_xprt *transport, struct msghdr *msg, int flags) | |
581 | { | |
582 | return -ESHUTDOWN; | |
583 | } | |
584 | #endif /* CONFIG_SUNRPC_BACKCHANNEL */ | |
585 | ||
586 | static ssize_t | |
587 | xs_read_stream_reply(struct sock_xprt *transport, struct msghdr *msg, int flags) | |
588 | { | |
589 | struct rpc_xprt *xprt = &transport->xprt; | |
590 | struct rpc_rqst *req; | |
591 | ssize_t ret = 0; | |
592 | ||
593 | /* Look up and lock the request corresponding to the given XID */ | |
594 | spin_lock(&xprt->queue_lock); | |
595 | req = xprt_lookup_rqst(xprt, transport->recv.xid); | |
45835a63 | 596 | if (!req || (transport->recv.copied && !req->rq_private_buf.len)) { |
277e4ab7 TM |
597 | msg->msg_flags |= MSG_TRUNC; |
598 | goto out; | |
599 | } | |
600 | xprt_pin_rqst(req); | |
601 | spin_unlock(&xprt->queue_lock); | |
602 | ||
603 | ret = xs_read_stream_request(transport, msg, flags, req); | |
604 | ||
605 | spin_lock(&xprt->queue_lock); | |
606 | if (msg->msg_flags & (MSG_EOR|MSG_TRUNC)) | |
c4433055 | 607 | xprt_complete_rqst(req->rq_task, transport->recv.copied); |
45835a63 TM |
608 | else |
609 | req->rq_private_buf.len = transport->recv.copied; | |
277e4ab7 TM |
610 | xprt_unpin_rqst(req); |
611 | out: | |
612 | spin_unlock(&xprt->queue_lock); | |
613 | return ret; | |
614 | } | |
615 | ||
616 | static ssize_t | |
617 | xs_read_stream(struct sock_xprt *transport, int flags) | |
618 | { | |
619 | struct msghdr msg = { 0 }; | |
620 | size_t want, read = 0; | |
621 | ssize_t ret = 0; | |
622 | ||
623 | if (transport->recv.len == 0) { | |
624 | want = xs_read_stream_headersize(transport->recv.copied != 0); | |
625 | ret = xs_read_stream_header(transport, &msg, flags, want, | |
626 | transport->recv.offset); | |
627 | if (ret <= 0) | |
628 | goto out_err; | |
629 | transport->recv.offset = ret; | |
c4433055 TM |
630 | if (transport->recv.offset != want) |
631 | return transport->recv.offset; | |
277e4ab7 TM |
632 | transport->recv.len = be32_to_cpu(transport->recv.fraghdr) & |
633 | RPC_FRAGMENT_SIZE_MASK; | |
634 | transport->recv.offset -= sizeof(transport->recv.fraghdr); | |
635 | read = ret; | |
636 | } | |
637 | ||
638 | switch (be32_to_cpu(transport->recv.calldir)) { | |
c4433055 TM |
639 | default: |
640 | msg.msg_flags |= MSG_TRUNC; | |
641 | break; | |
277e4ab7 TM |
642 | case RPC_CALL: |
643 | ret = xs_read_stream_call(transport, &msg, flags); | |
644 | break; | |
645 | case RPC_REPLY: | |
646 | ret = xs_read_stream_reply(transport, &msg, flags); | |
647 | } | |
648 | if (msg.msg_flags & MSG_TRUNC) { | |
649 | transport->recv.calldir = cpu_to_be32(-1); | |
650 | transport->recv.copied = -1; | |
651 | } | |
652 | if (ret < 0) | |
653 | goto out_err; | |
654 | read += ret; | |
655 | if (transport->recv.offset < transport->recv.len) { | |
c4433055 TM |
656 | if (!(msg.msg_flags & MSG_TRUNC)) |
657 | return read; | |
b76a5afd | 658 | msg.msg_flags = 0; |
277e4ab7 TM |
659 | ret = xs_read_discard(transport->sock, &msg, flags, |
660 | transport->recv.len - transport->recv.offset); | |
661 | if (ret <= 0) | |
662 | goto out_err; | |
663 | transport->recv.offset += ret; | |
664 | read += ret; | |
665 | if (transport->recv.offset != transport->recv.len) | |
c4433055 | 666 | return read; |
277e4ab7 TM |
667 | } |
668 | if (xs_read_stream_request_done(transport)) { | |
c50b8ee0 | 669 | trace_xs_stream_read_request(transport); |
277e4ab7 TM |
670 | transport->recv.copied = 0; |
671 | } | |
672 | transport->recv.offset = 0; | |
673 | transport->recv.len = 0; | |
674 | return read; | |
675 | out_err: | |
79462857 | 676 | return ret != 0 ? ret : -ESHUTDOWN; |
277e4ab7 TM |
677 | } |
678 | ||
0ffe86f4 TM |
679 | static __poll_t xs_poll_socket(struct sock_xprt *transport) |
680 | { | |
a73881c9 TM |
681 | return transport->sock->ops->poll(transport->file, transport->sock, |
682 | NULL); | |
0ffe86f4 TM |
683 | } |
684 | ||
685 | static bool xs_poll_socket_readable(struct sock_xprt *transport) | |
686 | { | |
687 | __poll_t events = xs_poll_socket(transport); | |
688 | ||
689 | return (events & (EPOLLIN | EPOLLRDNORM)) && !(events & EPOLLRDHUP); | |
690 | } | |
691 | ||
692 | static void xs_poll_check_readable(struct sock_xprt *transport) | |
693 | { | |
694 | ||
695 | clear_bit(XPRT_SOCK_DATA_READY, &transport->sock_state); | |
696 | if (!xs_poll_socket_readable(transport)) | |
697 | return; | |
698 | if (!test_and_set_bit(XPRT_SOCK_DATA_READY, &transport->sock_state)) | |
699 | queue_work(xprtiod_workqueue, &transport->recv_worker); | |
700 | } | |
701 | ||
c50b8ee0 TM |
702 | static void xs_stream_data_receive(struct sock_xprt *transport) |
703 | { | |
704 | size_t read = 0; | |
705 | ssize_t ret = 0; | |
706 | ||
707 | mutex_lock(&transport->recv_mutex); | |
708 | if (transport->sock == NULL) | |
709 | goto out; | |
c50b8ee0 TM |
710 | for (;;) { |
711 | ret = xs_read_stream(transport, MSG_DONTWAIT); | |
c4433055 | 712 | if (ret < 0) |
c50b8ee0 TM |
713 | break; |
714 | read += ret; | |
715 | cond_resched(); | |
716 | } | |
5f52a9d4 TM |
717 | if (ret == -ESHUTDOWN) |
718 | kernel_sock_shutdown(transport->sock, SHUT_RDWR); | |
719 | else | |
720 | xs_poll_check_readable(transport); | |
c50b8ee0 TM |
721 | out: |
722 | mutex_unlock(&transport->recv_mutex); | |
723 | trace_xs_stream_read_data(&transport->xprt, ret, read); | |
724 | } | |
725 | ||
726 | static void xs_stream_data_receive_workfn(struct work_struct *work) | |
727 | { | |
728 | struct sock_xprt *transport = | |
729 | container_of(work, struct sock_xprt, recv_worker); | |
a1231fda TM |
730 | unsigned int pflags = memalloc_nofs_save(); |
731 | ||
c50b8ee0 | 732 | xs_stream_data_receive(transport); |
a1231fda | 733 | memalloc_nofs_restore(pflags); |
c50b8ee0 TM |
734 | } |
735 | ||
550aebfe TM |
736 | static void |
737 | xs_stream_reset_connect(struct sock_xprt *transport) | |
738 | { | |
739 | transport->recv.offset = 0; | |
740 | transport->recv.len = 0; | |
741 | transport->recv.copied = 0; | |
742 | transport->xmit.offset = 0; | |
ae053551 TM |
743 | } |
744 | ||
745 | static void | |
746 | xs_stream_start_connect(struct sock_xprt *transport) | |
747 | { | |
550aebfe TM |
748 | transport->xprt.stat.connect_count++; |
749 | transport->xprt.stat.connect_start = jiffies; | |
750 | } | |
751 | ||
b4b5cc85 CL |
752 | #define XS_SENDMSG_FLAGS (MSG_DONTWAIT | MSG_NOSIGNAL) |
753 | ||
9903cd1c | 754 | /** |
c544577d | 755 | * xs_nospace - handle transmit was incomplete |
50f484e2 | 756 | * @req: pointer to RPC request |
9903cd1c | 757 | * |
a246b010 | 758 | */ |
c544577d | 759 | static int xs_nospace(struct rpc_rqst *req) |
a246b010 | 760 | { |
262965f5 | 761 | struct rpc_xprt *xprt = req->rq_xprt; |
ee0ac0c2 | 762 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
06ea0bfe | 763 | struct sock *sk = transport->inet; |
24ca9a84 | 764 | int ret = -EAGAIN; |
a246b010 | 765 | |
015747d2 | 766 | trace_rpc_socket_nospace(req, transport); |
262965f5 | 767 | |
b6ddf64f | 768 | /* Protect against races with write_space */ |
b5e92419 | 769 | spin_lock(&xprt->transport_lock); |
b6ddf64f TM |
770 | |
771 | /* Don't race with disconnect */ | |
772 | if (xprt_connected(xprt)) { | |
13331a55 TM |
773 | /* wait for more buffer space */ |
774 | sk->sk_write_pending++; | |
c544577d | 775 | xprt_wait_for_buffer_space(xprt); |
13331a55 | 776 | } else |
5e3771ce | 777 | ret = -ENOTCONN; |
262965f5 | 778 | |
b5e92419 | 779 | spin_unlock(&xprt->transport_lock); |
06ea0bfe TM |
780 | |
781 | /* Race breaker in case memory is freed before above code is called */ | |
d48f9ce7 DV |
782 | if (ret == -EAGAIN) { |
783 | struct socket_wq *wq; | |
784 | ||
785 | rcu_read_lock(); | |
786 | wq = rcu_dereference(sk->sk_wq); | |
787 | set_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags); | |
788 | rcu_read_unlock(); | |
789 | ||
790 | sk->sk_write_space(sk); | |
791 | } | |
5e3771ce | 792 | return ret; |
262965f5 CL |
793 | } |
794 | ||
277e4ab7 TM |
795 | static void |
796 | xs_stream_prepare_request(struct rpc_rqst *req) | |
797 | { | |
75369089 | 798 | xdr_free_bvec(&req->rq_rcv_buf); |
12a3ad61 | 799 | req->rq_task->tk_status = xdr_alloc_bvec(&req->rq_rcv_buf, GFP_KERNEL); |
277e4ab7 TM |
800 | } |
801 | ||
4cd34e7c TM |
802 | /* |
803 | * Determine if the previous message in the stream was aborted before it | |
804 | * could complete transmission. | |
805 | */ | |
806 | static bool | |
807 | xs_send_request_was_aborted(struct sock_xprt *transport, struct rpc_rqst *req) | |
808 | { | |
809 | return transport->xmit.offset != 0 && req->rq_bytes_sent == 0; | |
810 | } | |
811 | ||
61677eee | 812 | /* |
06b5fc3a | 813 | * Return the stream record marker field for a record of length < 2^31-1 |
61677eee | 814 | */ |
06b5fc3a TM |
815 | static rpc_fraghdr |
816 | xs_stream_record_marker(struct xdr_buf *xdr) | |
61677eee | 817 | { |
06b5fc3a TM |
818 | if (!xdr->len) |
819 | return 0; | |
820 | return cpu_to_be32(RPC_LAST_STREAM_FRAGMENT | (u32)xdr->len); | |
61677eee CL |
821 | } |
822 | ||
176e21ee CL |
823 | /** |
824 | * xs_local_send_request - write an RPC request to an AF_LOCAL socket | |
50f484e2 | 825 | * @req: pointer to RPC request |
176e21ee CL |
826 | * |
827 | * Return values: | |
828 | * 0: The request has been sent | |
829 | * EAGAIN: The socket was blocked, please call again later to | |
830 | * complete the request | |
831 | * ENOTCONN: Caller needs to invoke connect logic then call again | |
12b20ce3 | 832 | * other: Some other error occurred, the request was not sent |
176e21ee | 833 | */ |
adfa7144 | 834 | static int xs_local_send_request(struct rpc_rqst *req) |
176e21ee | 835 | { |
176e21ee CL |
836 | struct rpc_xprt *xprt = req->rq_xprt; |
837 | struct sock_xprt *transport = | |
838 | container_of(xprt, struct sock_xprt, xprt); | |
839 | struct xdr_buf *xdr = &req->rq_snd_buf; | |
7e3d3620 TM |
840 | rpc_fraghdr rm = xs_stream_record_marker(xdr); |
841 | unsigned int msglen = rm ? req->rq_slen + sizeof(rm) : req->rq_slen; | |
9e55eef4 CL |
842 | struct msghdr msg = { |
843 | .msg_flags = XS_SENDMSG_FLAGS, | |
844 | }; | |
3f649ab7 | 845 | unsigned int sent; |
176e21ee CL |
846 | int status; |
847 | ||
4cd34e7c TM |
848 | /* Close the stream if the previous transmission was incomplete */ |
849 | if (xs_send_request_was_aborted(transport, req)) { | |
850 | xs_close(xprt); | |
851 | return -ENOTCONN; | |
852 | } | |
853 | ||
176e21ee CL |
854 | xs_pktdump("packet data:", |
855 | req->rq_svec->iov_base, req->rq_svec->iov_len); | |
856 | ||
78215759 | 857 | req->rq_xtime = ktime_get(); |
9e55eef4 CL |
858 | status = xprt_sock_sendmsg(transport->sock, &msg, xdr, |
859 | transport->xmit.offset, rm, &sent); | |
176e21ee | 860 | dprintk("RPC: %s(%u) = %d\n", |
6c7a64e5 | 861 | __func__, xdr->len - transport->xmit.offset, status); |
743c69e7 N |
862 | |
863 | if (status == -EAGAIN && sock_writeable(transport->inet)) | |
864 | status = -ENOBUFS; | |
865 | ||
f279cd00 | 866 | if (likely(sent > 0) || status == 0) { |
6c7a64e5 TM |
867 | transport->xmit.offset += sent; |
868 | req->rq_bytes_sent = transport->xmit.offset; | |
7e3d3620 | 869 | if (likely(req->rq_bytes_sent >= msglen)) { |
6c7a64e5 | 870 | req->rq_xmit_bytes_sent += transport->xmit.offset; |
6c7a64e5 | 871 | transport->xmit.offset = 0; |
176e21ee CL |
872 | return 0; |
873 | } | |
874 | status = -EAGAIN; | |
875 | } | |
876 | ||
877 | switch (status) { | |
3601c4a9 | 878 | case -ENOBUFS: |
b5872f0c | 879 | break; |
176e21ee | 880 | case -EAGAIN: |
c544577d | 881 | status = xs_nospace(req); |
176e21ee CL |
882 | break; |
883 | default: | |
884 | dprintk("RPC: sendmsg returned unrecognized error %d\n", | |
885 | -status); | |
df561f66 | 886 | fallthrough; |
176e21ee CL |
887 | case -EPIPE: |
888 | xs_close(xprt); | |
889 | status = -ENOTCONN; | |
890 | } | |
891 | ||
892 | return status; | |
893 | } | |
894 | ||
262965f5 CL |
895 | /** |
896 | * xs_udp_send_request - write an RPC request to a UDP socket | |
50f484e2 | 897 | * @req: pointer to RPC request |
262965f5 CL |
898 | * |
899 | * Return values: | |
900 | * 0: The request has been sent | |
901 | * EAGAIN: The socket was blocked, please call again later to | |
902 | * complete the request | |
903 | * ENOTCONN: Caller needs to invoke connect logic then call again | |
25985edc | 904 | * other: Some other error occurred, the request was not sent |
262965f5 | 905 | */ |
adfa7144 | 906 | static int xs_udp_send_request(struct rpc_rqst *req) |
262965f5 | 907 | { |
262965f5 | 908 | struct rpc_xprt *xprt = req->rq_xprt; |
ee0ac0c2 | 909 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
262965f5 | 910 | struct xdr_buf *xdr = &req->rq_snd_buf; |
9e55eef4 CL |
911 | struct msghdr msg = { |
912 | .msg_name = xs_addr(xprt), | |
913 | .msg_namelen = xprt->addrlen, | |
914 | .msg_flags = XS_SENDMSG_FLAGS, | |
915 | }; | |
3f649ab7 | 916 | unsigned int sent; |
262965f5 | 917 | int status; |
a246b010 | 918 | |
9903cd1c | 919 | xs_pktdump("packet data:", |
a246b010 CL |
920 | req->rq_svec->iov_base, |
921 | req->rq_svec->iov_len); | |
922 | ||
01d37c42 TM |
923 | if (!xprt_bound(xprt)) |
924 | return -ENOTCONN; | |
75891f50 TM |
925 | |
926 | if (!xprt_request_get_cong(xprt, req)) | |
927 | return -EBADSLT; | |
928 | ||
78215759 | 929 | req->rq_xtime = ktime_get(); |
9e55eef4 | 930 | status = xprt_sock_sendmsg(transport->sock, &msg, xdr, 0, 0, &sent); |
a246b010 | 931 | |
46121cf7 | 932 | dprintk("RPC: xs_udp_send_request(%u) = %d\n", |
6c7a64e5 | 933 | xdr->len, status); |
a246b010 | 934 | |
3dedbb5c JB |
935 | /* firewall is blocking us, don't return -EAGAIN or we end up looping */ |
936 | if (status == -EPERM) | |
937 | goto process_status; | |
938 | ||
743c69e7 N |
939 | if (status == -EAGAIN && sock_writeable(transport->inet)) |
940 | status = -ENOBUFS; | |
941 | ||
f279cd00 JB |
942 | if (sent > 0 || status == 0) { |
943 | req->rq_xmit_bytes_sent += sent; | |
944 | if (sent >= req->rq_slen) | |
2199700f TM |
945 | return 0; |
946 | /* Still some bytes left; set up for a retry later. */ | |
262965f5 | 947 | status = -EAGAIN; |
2199700f | 948 | } |
a246b010 | 949 | |
3dedbb5c | 950 | process_status: |
262965f5 | 951 | switch (status) { |
fba91afb TM |
952 | case -ENOTSOCK: |
953 | status = -ENOTCONN; | |
954 | /* Should we call xs_close() here? */ | |
955 | break; | |
b6ddf64f | 956 | case -EAGAIN: |
c544577d | 957 | status = xs_nospace(req); |
b6ddf64f | 958 | break; |
262965f5 | 959 | case -ENETUNREACH: |
3601c4a9 | 960 | case -ENOBUFS: |
262965f5 | 961 | case -EPIPE: |
a246b010 | 962 | case -ECONNREFUSED: |
3dedbb5c | 963 | case -EPERM: |
a246b010 | 964 | /* When the server has died, an ICMP port unreachable message |
9903cd1c | 965 | * prompts ECONNREFUSED. */ |
13331a55 TM |
966 | break; |
967 | default: | |
968 | dprintk("RPC: sendmsg returned unrecognized error %d\n", | |
969 | -status); | |
a246b010 | 970 | } |
5fe46e9d | 971 | |
262965f5 | 972 | return status; |
a246b010 CL |
973 | } |
974 | ||
9903cd1c | 975 | /** |
262965f5 | 976 | * xs_tcp_send_request - write an RPC request to a TCP socket |
50f484e2 | 977 | * @req: pointer to RPC request |
9903cd1c CL |
978 | * |
979 | * Return values: | |
262965f5 CL |
980 | * 0: The request has been sent |
981 | * EAGAIN: The socket was blocked, please call again later to | |
982 | * complete the request | |
983 | * ENOTCONN: Caller needs to invoke connect logic then call again | |
25985edc | 984 | * other: Some other error occurred, the request was not sent |
9903cd1c CL |
985 | * |
986 | * XXX: In the case of soft timeouts, should we eventually give up | |
262965f5 | 987 | * if sendmsg is not able to make progress? |
9903cd1c | 988 | */ |
adfa7144 | 989 | static int xs_tcp_send_request(struct rpc_rqst *req) |
a246b010 | 990 | { |
a246b010 | 991 | struct rpc_xprt *xprt = req->rq_xprt; |
ee0ac0c2 | 992 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
262965f5 | 993 | struct xdr_buf *xdr = &req->rq_snd_buf; |
7e3d3620 TM |
994 | rpc_fraghdr rm = xs_stream_record_marker(xdr); |
995 | unsigned int msglen = rm ? req->rq_slen + sizeof(rm) : req->rq_slen; | |
9e55eef4 CL |
996 | struct msghdr msg = { |
997 | .msg_flags = XS_SENDMSG_FLAGS, | |
998 | }; | |
9ffadfbc | 999 | bool vm_wait = false; |
3f649ab7 | 1000 | unsigned int sent; |
b595bb15 | 1001 | int status; |
a246b010 | 1002 | |
4cd34e7c TM |
1003 | /* Close the stream if the previous transmission was incomplete */ |
1004 | if (xs_send_request_was_aborted(transport, req)) { | |
1005 | if (transport->sock != NULL) | |
1006 | kernel_sock_shutdown(transport->sock, SHUT_RDWR); | |
1007 | return -ENOTCONN; | |
1008 | } | |
1009 | ||
262965f5 CL |
1010 | xs_pktdump("packet data:", |
1011 | req->rq_svec->iov_base, | |
1012 | req->rq_svec->iov_len); | |
a246b010 | 1013 | |
7196dbb0 TM |
1014 | if (test_bit(XPRT_SOCK_UPD_TIMEOUT, &transport->sock_state)) |
1015 | xs_tcp_set_socket_timeouts(xprt, transport->sock); | |
1016 | ||
a246b010 CL |
1017 | /* Continue transmitting the packet/record. We must be careful |
1018 | * to cope with writespace callbacks arriving _after_ we have | |
262965f5 | 1019 | * called sendmsg(). */ |
78215759 | 1020 | req->rq_xtime = ktime_get(); |
a246b010 | 1021 | while (1) { |
9e55eef4 CL |
1022 | status = xprt_sock_sendmsg(transport->sock, &msg, xdr, |
1023 | transport->xmit.offset, rm, &sent); | |
a246b010 | 1024 | |
46121cf7 | 1025 | dprintk("RPC: xs_tcp_send_request(%u) = %d\n", |
6c7a64e5 | 1026 | xdr->len - transport->xmit.offset, status); |
a246b010 | 1027 | |
262965f5 CL |
1028 | /* If we've sent the entire packet, immediately |
1029 | * reset the count of bytes sent. */ | |
6c7a64e5 TM |
1030 | transport->xmit.offset += sent; |
1031 | req->rq_bytes_sent = transport->xmit.offset; | |
7e3d3620 | 1032 | if (likely(req->rq_bytes_sent >= msglen)) { |
6c7a64e5 | 1033 | req->rq_xmit_bytes_sent += transport->xmit.offset; |
6c7a64e5 | 1034 | transport->xmit.offset = 0; |
262965f5 CL |
1035 | return 0; |
1036 | } | |
a246b010 | 1037 | |
9ffadfbc TM |
1038 | WARN_ON_ONCE(sent == 0 && status == 0); |
1039 | ||
1040 | if (status == -EAGAIN ) { | |
1041 | /* | |
1042 | * Return EAGAIN if we're sure we're hitting the | |
1043 | * socket send buffer limits. | |
1044 | */ | |
1045 | if (test_bit(SOCK_NOSPACE, &transport->sock->flags)) | |
1046 | break; | |
1047 | /* | |
1048 | * Did we hit a memory allocation failure? | |
1049 | */ | |
1050 | if (sent == 0) { | |
1051 | status = -ENOBUFS; | |
1052 | if (vm_wait) | |
1053 | break; | |
1054 | /* Retry, knowing now that we're below the | |
1055 | * socket send buffer limit | |
1056 | */ | |
1057 | vm_wait = true; | |
1058 | } | |
1059 | continue; | |
1060 | } | |
f580dd04 TM |
1061 | if (status < 0) |
1062 | break; | |
9ffadfbc | 1063 | vm_wait = false; |
a246b010 CL |
1064 | } |
1065 | ||
262965f5 | 1066 | switch (status) { |
fba91afb TM |
1067 | case -ENOTSOCK: |
1068 | status = -ENOTCONN; | |
1069 | /* Should we call xs_close() here? */ | |
1070 | break; | |
262965f5 | 1071 | case -EAGAIN: |
c544577d | 1072 | status = xs_nospace(req); |
262965f5 | 1073 | break; |
262965f5 | 1074 | case -ECONNRESET: |
2a9e1cfa | 1075 | case -ECONNREFUSED: |
262965f5 | 1076 | case -ENOTCONN: |
3913c78c | 1077 | case -EADDRINUSE: |
b5872f0c | 1078 | case -ENOBUFS: |
b9d2bb2e | 1079 | case -EPIPE: |
13331a55 TM |
1080 | break; |
1081 | default: | |
1082 | dprintk("RPC: sendmsg returned unrecognized error %d\n", | |
1083 | -status); | |
a246b010 | 1084 | } |
5fe46e9d | 1085 | |
a246b010 CL |
1086 | return status; |
1087 | } | |
1088 | ||
2a9e1cfa TM |
1089 | static void xs_save_old_callbacks(struct sock_xprt *transport, struct sock *sk) |
1090 | { | |
1091 | transport->old_data_ready = sk->sk_data_ready; | |
1092 | transport->old_state_change = sk->sk_state_change; | |
1093 | transport->old_write_space = sk->sk_write_space; | |
2118071d | 1094 | transport->old_error_report = sk->sk_error_report; |
2a9e1cfa TM |
1095 | } |
1096 | ||
1097 | static void xs_restore_old_callbacks(struct sock_xprt *transport, struct sock *sk) | |
1098 | { | |
1099 | sk->sk_data_ready = transport->old_data_ready; | |
1100 | sk->sk_state_change = transport->old_state_change; | |
1101 | sk->sk_write_space = transport->old_write_space; | |
2118071d TM |
1102 | sk->sk_error_report = transport->old_error_report; |
1103 | } | |
1104 | ||
42d42a5b TM |
1105 | static void xs_sock_reset_state_flags(struct rpc_xprt *xprt) |
1106 | { | |
1107 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); | |
1108 | ||
1109 | clear_bit(XPRT_SOCK_DATA_READY, &transport->sock_state); | |
4f8943f8 TM |
1110 | clear_bit(XPRT_SOCK_WAKE_ERROR, &transport->sock_state); |
1111 | clear_bit(XPRT_SOCK_WAKE_WRITE, &transport->sock_state); | |
1112 | clear_bit(XPRT_SOCK_WAKE_DISCONNECT, &transport->sock_state); | |
1113 | } | |
1114 | ||
1115 | static void xs_run_error_worker(struct sock_xprt *transport, unsigned int nr) | |
1116 | { | |
1117 | set_bit(nr, &transport->sock_state); | |
1118 | queue_work(xprtiod_workqueue, &transport->error_worker); | |
42d42a5b TM |
1119 | } |
1120 | ||
b70ae915 TM |
1121 | static void xs_sock_reset_connection_flags(struct rpc_xprt *xprt) |
1122 | { | |
1123 | smp_mb__before_atomic(); | |
1124 | clear_bit(XPRT_CLOSE_WAIT, &xprt->state); | |
1125 | clear_bit(XPRT_CLOSING, &xprt->state); | |
42d42a5b | 1126 | xs_sock_reset_state_flags(xprt); |
b70ae915 TM |
1127 | smp_mb__after_atomic(); |
1128 | } | |
1129 | ||
2118071d TM |
1130 | /** |
1131 | * xs_error_report - callback to handle TCP socket state errors | |
1132 | * @sk: socket | |
1133 | * | |
1134 | * Note: we don't call sock_error() since there may be a rpc_task | |
1135 | * using the socket, and so we don't want to clear sk->sk_err. | |
1136 | */ | |
1137 | static void xs_error_report(struct sock *sk) | |
1138 | { | |
4f8943f8 | 1139 | struct sock_xprt *transport; |
2118071d | 1140 | struct rpc_xprt *xprt; |
2118071d TM |
1141 | |
1142 | read_lock_bh(&sk->sk_callback_lock); | |
1143 | if (!(xprt = xprt_from_sock(sk))) | |
1144 | goto out; | |
1145 | ||
4f8943f8 | 1146 | transport = container_of(xprt, struct sock_xprt, xprt); |
af84537d BC |
1147 | transport->xprt_err = -sk->sk_err; |
1148 | if (transport->xprt_err == 0) | |
2118071d TM |
1149 | goto out; |
1150 | dprintk("RPC: xs_error_report client %p, error=%d...\n", | |
af84537d BC |
1151 | xprt, -transport->xprt_err); |
1152 | trace_rpc_socket_error(xprt, sk->sk_socket, transport->xprt_err); | |
1153 | ||
1154 | /* barrier ensures xprt_err is set before XPRT_SOCK_WAKE_ERROR */ | |
1155 | smp_mb__before_atomic(); | |
4f8943f8 | 1156 | xs_run_error_worker(transport, XPRT_SOCK_WAKE_ERROR); |
2118071d TM |
1157 | out: |
1158 | read_unlock_bh(&sk->sk_callback_lock); | |
2a9e1cfa TM |
1159 | } |
1160 | ||
fe315e76 | 1161 | static void xs_reset_transport(struct sock_xprt *transport) |
a246b010 | 1162 | { |
ee0ac0c2 CL |
1163 | struct socket *sock = transport->sock; |
1164 | struct sock *sk = transport->inet; | |
6cc7e908 | 1165 | struct rpc_xprt *xprt = &transport->xprt; |
a73881c9 | 1166 | struct file *filp = transport->file; |
a246b010 | 1167 | |
fe315e76 CL |
1168 | if (sk == NULL) |
1169 | return; | |
9903cd1c | 1170 | |
264d1df3 JL |
1171 | if (atomic_read(&transport->xprt.swapper)) |
1172 | sk_clear_memalloc(sk); | |
1173 | ||
09939204 TM |
1174 | kernel_sock_shutdown(sock, SHUT_RDWR); |
1175 | ||
edc1b01c | 1176 | mutex_lock(&transport->recv_mutex); |
a246b010 | 1177 | write_lock_bh(&sk->sk_callback_lock); |
ee0ac0c2 CL |
1178 | transport->inet = NULL; |
1179 | transport->sock = NULL; | |
a73881c9 | 1180 | transport->file = NULL; |
a246b010 | 1181 | |
9903cd1c | 1182 | sk->sk_user_data = NULL; |
2a9e1cfa TM |
1183 | |
1184 | xs_restore_old_callbacks(transport, sk); | |
0c78789e | 1185 | xprt_clear_connected(xprt); |
a246b010 | 1186 | write_unlock_bh(&sk->sk_callback_lock); |
6cc7e908 | 1187 | xs_sock_reset_connection_flags(xprt); |
ae053551 TM |
1188 | /* Reset stream record info */ |
1189 | xs_stream_reset_connect(transport); | |
edc1b01c | 1190 | mutex_unlock(&transport->recv_mutex); |
a246b010 | 1191 | |
6cc7e908 | 1192 | trace_rpc_socket_close(xprt, sock); |
a73881c9 | 1193 | fput(filp); |
0445f92c TM |
1194 | |
1195 | xprt_disconnect_done(xprt); | |
fe315e76 CL |
1196 | } |
1197 | ||
1198 | /** | |
1199 | * xs_close - close a socket | |
1200 | * @xprt: transport | |
1201 | * | |
1202 | * This is used when all requests are complete; ie, no DRC state remains | |
1203 | * on the server we want to save. | |
f75e6745 TM |
1204 | * |
1205 | * The caller _must_ be holding XPRT_LOCKED in order to avoid issues with | |
1206 | * xs_reset_transport() zeroing the socket from underneath a writer. | |
fe315e76 CL |
1207 | */ |
1208 | static void xs_close(struct rpc_xprt *xprt) | |
1209 | { | |
1210 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); | |
1211 | ||
1212 | dprintk("RPC: xs_close xprt %p\n", xprt); | |
1213 | ||
1214 | xs_reset_transport(transport); | |
61d0a8e6 | 1215 | xprt->reestablish_timeout = 0; |
a246b010 CL |
1216 | } |
1217 | ||
4a068258 CL |
1218 | static void xs_inject_disconnect(struct rpc_xprt *xprt) |
1219 | { | |
1220 | dprintk("RPC: injecting transport disconnect on xprt=%p\n", | |
1221 | xprt); | |
1222 | xprt_disconnect_done(xprt); | |
1223 | } | |
1224 | ||
315f3812 KM |
1225 | static void xs_xprt_free(struct rpc_xprt *xprt) |
1226 | { | |
1227 | xs_free_peer_addresses(xprt); | |
1228 | xprt_free(xprt); | |
1229 | } | |
1230 | ||
9903cd1c CL |
1231 | /** |
1232 | * xs_destroy - prepare to shutdown a transport | |
1233 | * @xprt: doomed transport | |
1234 | * | |
1235 | */ | |
1236 | static void xs_destroy(struct rpc_xprt *xprt) | |
a246b010 | 1237 | { |
03c78827 TM |
1238 | struct sock_xprt *transport = container_of(xprt, |
1239 | struct sock_xprt, xprt); | |
46121cf7 | 1240 | dprintk("RPC: xs_destroy xprt %p\n", xprt); |
9903cd1c | 1241 | |
03c78827 | 1242 | cancel_delayed_work_sync(&transport->connect_worker); |
a1311d87 | 1243 | xs_close(xprt); |
edc1b01c | 1244 | cancel_work_sync(&transport->recv_worker); |
b5e92419 | 1245 | cancel_work_sync(&transport->error_worker); |
315f3812 | 1246 | xs_xprt_free(xprt); |
a1311d87 | 1247 | module_put(THIS_MODULE); |
a246b010 CL |
1248 | } |
1249 | ||
9903cd1c | 1250 | /** |
f9b2ee71 TM |
1251 | * xs_udp_data_read_skb - receive callback for UDP sockets |
1252 | * @xprt: transport | |
1253 | * @sk: socket | |
1254 | * @skb: skbuff | |
9903cd1c | 1255 | * |
a246b010 | 1256 | */ |
f9b2ee71 TM |
1257 | static void xs_udp_data_read_skb(struct rpc_xprt *xprt, |
1258 | struct sock *sk, | |
1259 | struct sk_buff *skb) | |
a246b010 | 1260 | { |
9903cd1c | 1261 | struct rpc_task *task; |
a246b010 | 1262 | struct rpc_rqst *rovr; |
f9b2ee71 | 1263 | int repsize, copied; |
d8ed029d AD |
1264 | u32 _xid; |
1265 | __be32 *xp; | |
a246b010 | 1266 | |
1da8c681 | 1267 | repsize = skb->len; |
a246b010 | 1268 | if (repsize < 4) { |
46121cf7 | 1269 | dprintk("RPC: impossible RPC reply size %d!\n", repsize); |
f9b2ee71 | 1270 | return; |
a246b010 CL |
1271 | } |
1272 | ||
1273 | /* Copy the XID from the skb... */ | |
1da8c681 | 1274 | xp = skb_header_pointer(skb, 0, sizeof(_xid), &_xid); |
a246b010 | 1275 | if (xp == NULL) |
f9b2ee71 | 1276 | return; |
a246b010 CL |
1277 | |
1278 | /* Look up and lock the request corresponding to the given XID */ | |
75c84151 | 1279 | spin_lock(&xprt->queue_lock); |
a246b010 CL |
1280 | rovr = xprt_lookup_rqst(xprt, *xp); |
1281 | if (!rovr) | |
1282 | goto out_unlock; | |
729749bb | 1283 | xprt_pin_rqst(rovr); |
ecd465ee | 1284 | xprt_update_rtt(rovr->rq_task); |
75c84151 | 1285 | spin_unlock(&xprt->queue_lock); |
a246b010 CL |
1286 | task = rovr->rq_task; |
1287 | ||
a246b010 CL |
1288 | if ((copied = rovr->rq_private_buf.buflen) > repsize) |
1289 | copied = repsize; | |
1290 | ||
1291 | /* Suck it into the iovec, verify checksum if not done by hw. */ | |
1781f7f5 | 1292 | if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb)) { |
75c84151 | 1293 | spin_lock(&xprt->queue_lock); |
0afa6b44 | 1294 | __UDPX_INC_STATS(sk, UDP_MIB_INERRORS); |
729749bb | 1295 | goto out_unpin; |
1781f7f5 HX |
1296 | } |
1297 | ||
a246b010 | 1298 | |
b5e92419 | 1299 | spin_lock(&xprt->transport_lock); |
6a24dfb6 | 1300 | xprt_adjust_cwnd(xprt, task, copied); |
b5e92419 | 1301 | spin_unlock(&xprt->transport_lock); |
75c84151 | 1302 | spin_lock(&xprt->queue_lock); |
1570c1e4 | 1303 | xprt_complete_rqst(task, copied); |
0afa6b44 | 1304 | __UDPX_INC_STATS(sk, UDP_MIB_INDATAGRAMS); |
729749bb TM |
1305 | out_unpin: |
1306 | xprt_unpin_rqst(rovr); | |
a246b010 | 1307 | out_unlock: |
75c84151 | 1308 | spin_unlock(&xprt->queue_lock); |
f9b2ee71 TM |
1309 | } |
1310 | ||
1311 | static void xs_udp_data_receive(struct sock_xprt *transport) | |
1312 | { | |
1313 | struct sk_buff *skb; | |
1314 | struct sock *sk; | |
1315 | int err; | |
1316 | ||
1317 | mutex_lock(&transport->recv_mutex); | |
1318 | sk = transport->inet; | |
1319 | if (sk == NULL) | |
1320 | goto out; | |
1321 | for (;;) { | |
7c13f97f | 1322 | skb = skb_recv_udp(sk, 0, 1, &err); |
4f546149 | 1323 | if (skb == NULL) |
f9b2ee71 | 1324 | break; |
4f546149 TM |
1325 | xs_udp_data_read_skb(&transport->xprt, sk, skb); |
1326 | consume_skb(skb); | |
1327 | cond_resched(); | |
f9b2ee71 | 1328 | } |
0ffe86f4 | 1329 | xs_poll_check_readable(transport); |
f9b2ee71 TM |
1330 | out: |
1331 | mutex_unlock(&transport->recv_mutex); | |
1332 | } | |
1333 | ||
1334 | static void xs_udp_data_receive_workfn(struct work_struct *work) | |
1335 | { | |
1336 | struct sock_xprt *transport = | |
1337 | container_of(work, struct sock_xprt, recv_worker); | |
a1231fda TM |
1338 | unsigned int pflags = memalloc_nofs_save(); |
1339 | ||
f9b2ee71 | 1340 | xs_udp_data_receive(transport); |
a1231fda | 1341 | memalloc_nofs_restore(pflags); |
f9b2ee71 TM |
1342 | } |
1343 | ||
1344 | /** | |
1345 | * xs_data_ready - "data ready" callback for UDP sockets | |
1346 | * @sk: socket with data to read | |
1347 | * | |
1348 | */ | |
1349 | static void xs_data_ready(struct sock *sk) | |
1350 | { | |
1351 | struct rpc_xprt *xprt; | |
1352 | ||
1353 | read_lock_bh(&sk->sk_callback_lock); | |
1354 | dprintk("RPC: xs_data_ready...\n"); | |
1355 | xprt = xprt_from_sock(sk); | |
1356 | if (xprt != NULL) { | |
1357 | struct sock_xprt *transport = container_of(xprt, | |
1358 | struct sock_xprt, xprt); | |
5157b956 TM |
1359 | transport->old_data_ready(sk); |
1360 | /* Any data means we had a useful conversation, so | |
1361 | * then we don't need to delay the next reconnect | |
1362 | */ | |
1363 | if (xprt->reestablish_timeout) | |
1364 | xprt->reestablish_timeout = 0; | |
42d42a5b | 1365 | if (!test_and_set_bit(XPRT_SOCK_DATA_READY, &transport->sock_state)) |
40a5f1b1 | 1366 | queue_work(xprtiod_workqueue, &transport->recv_worker); |
f9b2ee71 | 1367 | } |
f064af1e | 1368 | read_unlock_bh(&sk->sk_callback_lock); |
a246b010 CL |
1369 | } |
1370 | ||
a519fc7a TM |
1371 | /* |
1372 | * Helper function to force a TCP close if the server is sending | |
1373 | * junk and/or it has put us in CLOSE_WAIT | |
1374 | */ | |
1375 | static void xs_tcp_force_close(struct rpc_xprt *xprt) | |
1376 | { | |
a519fc7a TM |
1377 | xprt_force_disconnect(xprt); |
1378 | } | |
1379 | ||
9e00abc3 | 1380 | #if defined(CONFIG_SUNRPC_BACKCHANNEL) |
6b26cc8c CL |
1381 | static size_t xs_tcp_bc_maxpayload(struct rpc_xprt *xprt) |
1382 | { | |
1383 | return PAGE_SIZE; | |
1384 | } | |
9e00abc3 | 1385 | #endif /* CONFIG_SUNRPC_BACKCHANNEL */ |
44b98efd | 1386 | |
9903cd1c CL |
1387 | /** |
1388 | * xs_tcp_state_change - callback to handle TCP socket state changes | |
1389 | * @sk: socket whose state has changed | |
1390 | * | |
1391 | */ | |
1392 | static void xs_tcp_state_change(struct sock *sk) | |
a246b010 | 1393 | { |
9903cd1c | 1394 | struct rpc_xprt *xprt; |
0fdea1e8 | 1395 | struct sock_xprt *transport; |
a246b010 | 1396 | |
f064af1e | 1397 | read_lock_bh(&sk->sk_callback_lock); |
a246b010 CL |
1398 | if (!(xprt = xprt_from_sock(sk))) |
1399 | goto out; | |
46121cf7 | 1400 | dprintk("RPC: xs_tcp_state_change client %p...\n", xprt); |
669502ff | 1401 | dprintk("RPC: state %x conn %d dead %d zapped %d sk_shutdown %d\n", |
46121cf7 CL |
1402 | sk->sk_state, xprt_connected(xprt), |
1403 | sock_flag(sk, SOCK_DEAD), | |
669502ff AC |
1404 | sock_flag(sk, SOCK_ZAPPED), |
1405 | sk->sk_shutdown); | |
a246b010 | 1406 | |
0fdea1e8 | 1407 | transport = container_of(xprt, struct sock_xprt, xprt); |
40b5ea0c | 1408 | trace_rpc_socket_state_change(xprt, sk->sk_socket); |
a246b010 CL |
1409 | switch (sk->sk_state) { |
1410 | case TCP_ESTABLISHED: | |
a246b010 | 1411 | if (!xprt_test_and_set_connected(xprt)) { |
8b71798c | 1412 | xprt->connect_cookie++; |
0fdea1e8 TM |
1413 | clear_bit(XPRT_SOCK_CONNECTING, &transport->sock_state); |
1414 | xprt_clear_connecting(xprt); | |
51971139 | 1415 | |
3968a8a5 CL |
1416 | xprt->stat.connect_count++; |
1417 | xprt->stat.connect_time += (long)jiffies - | |
1418 | xprt->stat.connect_start; | |
4f8943f8 | 1419 | xs_run_error_worker(transport, XPRT_SOCK_WAKE_PENDING); |
a246b010 | 1420 | } |
a246b010 | 1421 | break; |
3b948ae5 TM |
1422 | case TCP_FIN_WAIT1: |
1423 | /* The client initiated a shutdown of the socket */ | |
7c1d71cf | 1424 | xprt->connect_cookie++; |
663b8858 | 1425 | xprt->reestablish_timeout = 0; |
3b948ae5 | 1426 | set_bit(XPRT_CLOSING, &xprt->state); |
4e857c58 | 1427 | smp_mb__before_atomic(); |
3b948ae5 | 1428 | clear_bit(XPRT_CONNECTED, &xprt->state); |
ef803670 | 1429 | clear_bit(XPRT_CLOSE_WAIT, &xprt->state); |
4e857c58 | 1430 | smp_mb__after_atomic(); |
a246b010 | 1431 | break; |
632e3bdc | 1432 | case TCP_CLOSE_WAIT: |
3b948ae5 | 1433 | /* The server initiated a shutdown of the socket */ |
7c1d71cf | 1434 | xprt->connect_cookie++; |
d0bea455 | 1435 | clear_bit(XPRT_CONNECTED, &xprt->state); |
4f8943f8 | 1436 | xs_run_error_worker(transport, XPRT_SOCK_WAKE_DISCONNECT); |
df561f66 | 1437 | fallthrough; |
663b8858 TM |
1438 | case TCP_CLOSING: |
1439 | /* | |
1440 | * If the server closed down the connection, make sure that | |
1441 | * we back off before reconnecting | |
1442 | */ | |
1443 | if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO) | |
1444 | xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; | |
3b948ae5 TM |
1445 | break; |
1446 | case TCP_LAST_ACK: | |
670f9457 | 1447 | set_bit(XPRT_CLOSING, &xprt->state); |
4e857c58 | 1448 | smp_mb__before_atomic(); |
3b948ae5 | 1449 | clear_bit(XPRT_CONNECTED, &xprt->state); |
4e857c58 | 1450 | smp_mb__after_atomic(); |
3b948ae5 TM |
1451 | break; |
1452 | case TCP_CLOSE: | |
0fdea1e8 TM |
1453 | if (test_and_clear_bit(XPRT_SOCK_CONNECTING, |
1454 | &transport->sock_state)) | |
1455 | xprt_clear_connecting(xprt); | |
9b30889c | 1456 | clear_bit(XPRT_CLOSING, &xprt->state); |
9b30889c | 1457 | /* Trigger the socket release */ |
4f8943f8 | 1458 | xs_run_error_worker(transport, XPRT_SOCK_WAKE_DISCONNECT); |
a246b010 CL |
1459 | } |
1460 | out: | |
f064af1e | 1461 | read_unlock_bh(&sk->sk_callback_lock); |
a246b010 CL |
1462 | } |
1463 | ||
1f0fa154 IJ |
1464 | static void xs_write_space(struct sock *sk) |
1465 | { | |
13331a55 | 1466 | struct socket_wq *wq; |
4f8943f8 | 1467 | struct sock_xprt *transport; |
1f0fa154 IJ |
1468 | struct rpc_xprt *xprt; |
1469 | ||
13331a55 | 1470 | if (!sk->sk_socket) |
1f0fa154 | 1471 | return; |
13331a55 | 1472 | clear_bit(SOCK_NOSPACE, &sk->sk_socket->flags); |
1f0fa154 IJ |
1473 | |
1474 | if (unlikely(!(xprt = xprt_from_sock(sk)))) | |
1475 | return; | |
4f8943f8 | 1476 | transport = container_of(xprt, struct sock_xprt, xprt); |
13331a55 TM |
1477 | rcu_read_lock(); |
1478 | wq = rcu_dereference(sk->sk_wq); | |
1479 | if (!wq || test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags) == 0) | |
1480 | goto out; | |
1f0fa154 | 1481 | |
4f8943f8 TM |
1482 | xs_run_error_worker(transport, XPRT_SOCK_WAKE_WRITE); |
1483 | sk->sk_write_pending--; | |
13331a55 TM |
1484 | out: |
1485 | rcu_read_unlock(); | |
1f0fa154 IJ |
1486 | } |
1487 | ||
9903cd1c | 1488 | /** |
c7b2cae8 CL |
1489 | * xs_udp_write_space - callback invoked when socket buffer space |
1490 | * becomes available | |
9903cd1c CL |
1491 | * @sk: socket whose state has changed |
1492 | * | |
a246b010 CL |
1493 | * Called when more output buffer space is available for this socket. |
1494 | * We try not to wake our writers until they can make "significant" | |
c7b2cae8 | 1495 | * progress, otherwise we'll waste resources thrashing kernel_sendmsg |
a246b010 CL |
1496 | * with a bunch of small requests. |
1497 | */ | |
c7b2cae8 | 1498 | static void xs_udp_write_space(struct sock *sk) |
a246b010 | 1499 | { |
f064af1e | 1500 | read_lock_bh(&sk->sk_callback_lock); |
a246b010 | 1501 | |
c7b2cae8 | 1502 | /* from net/core/sock.c:sock_def_write_space */ |
1f0fa154 IJ |
1503 | if (sock_writeable(sk)) |
1504 | xs_write_space(sk); | |
a246b010 | 1505 | |
f064af1e | 1506 | read_unlock_bh(&sk->sk_callback_lock); |
c7b2cae8 | 1507 | } |
a246b010 | 1508 | |
c7b2cae8 CL |
1509 | /** |
1510 | * xs_tcp_write_space - callback invoked when socket buffer space | |
1511 | * becomes available | |
1512 | * @sk: socket whose state has changed | |
1513 | * | |
1514 | * Called when more output buffer space is available for this socket. | |
1515 | * We try not to wake our writers until they can make "significant" | |
1516 | * progress, otherwise we'll waste resources thrashing kernel_sendmsg | |
1517 | * with a bunch of small requests. | |
1518 | */ | |
1519 | static void xs_tcp_write_space(struct sock *sk) | |
1520 | { | |
f064af1e | 1521 | read_lock_bh(&sk->sk_callback_lock); |
c7b2cae8 CL |
1522 | |
1523 | /* from net/core/stream.c:sk_stream_write_space */ | |
64dc6130 | 1524 | if (sk_stream_is_writeable(sk)) |
1f0fa154 | 1525 | xs_write_space(sk); |
b6ddf64f | 1526 | |
f064af1e | 1527 | read_unlock_bh(&sk->sk_callback_lock); |
a246b010 CL |
1528 | } |
1529 | ||
470056c2 | 1530 | static void xs_udp_do_set_buffer_size(struct rpc_xprt *xprt) |
a246b010 | 1531 | { |
ee0ac0c2 CL |
1532 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
1533 | struct sock *sk = transport->inet; | |
a246b010 | 1534 | |
7c6e066e | 1535 | if (transport->rcvsize) { |
a246b010 | 1536 | sk->sk_userlocks |= SOCK_RCVBUF_LOCK; |
7c6e066e | 1537 | sk->sk_rcvbuf = transport->rcvsize * xprt->max_reqs * 2; |
a246b010 | 1538 | } |
7c6e066e | 1539 | if (transport->sndsize) { |
a246b010 | 1540 | sk->sk_userlocks |= SOCK_SNDBUF_LOCK; |
7c6e066e | 1541 | sk->sk_sndbuf = transport->sndsize * xprt->max_reqs * 2; |
a246b010 CL |
1542 | sk->sk_write_space(sk); |
1543 | } | |
1544 | } | |
1545 | ||
43118c29 | 1546 | /** |
470056c2 | 1547 | * xs_udp_set_buffer_size - set send and receive limits |
43118c29 | 1548 | * @xprt: generic transport |
470056c2 CL |
1549 | * @sndsize: requested size of send buffer, in bytes |
1550 | * @rcvsize: requested size of receive buffer, in bytes | |
43118c29 | 1551 | * |
470056c2 | 1552 | * Set socket send and receive buffer size limits. |
43118c29 | 1553 | */ |
470056c2 | 1554 | static void xs_udp_set_buffer_size(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize) |
43118c29 | 1555 | { |
7c6e066e CL |
1556 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
1557 | ||
1558 | transport->sndsize = 0; | |
470056c2 | 1559 | if (sndsize) |
7c6e066e CL |
1560 | transport->sndsize = sndsize + 1024; |
1561 | transport->rcvsize = 0; | |
470056c2 | 1562 | if (rcvsize) |
7c6e066e | 1563 | transport->rcvsize = rcvsize + 1024; |
470056c2 CL |
1564 | |
1565 | xs_udp_do_set_buffer_size(xprt); | |
43118c29 CL |
1566 | } |
1567 | ||
46c0ee8b CL |
1568 | /** |
1569 | * xs_udp_timer - called when a retransmit timeout occurs on a UDP transport | |
acf0a39f | 1570 | * @xprt: controlling transport |
46c0ee8b CL |
1571 | * @task: task that timed out |
1572 | * | |
1573 | * Adjust the congestion window after a retransmit timeout has occurred. | |
1574 | */ | |
6a24dfb6 | 1575 | static void xs_udp_timer(struct rpc_xprt *xprt, struct rpc_task *task) |
46c0ee8b | 1576 | { |
b5e92419 | 1577 | spin_lock(&xprt->transport_lock); |
6a24dfb6 | 1578 | xprt_adjust_cwnd(xprt, task, -ETIMEDOUT); |
b5e92419 | 1579 | spin_unlock(&xprt->transport_lock); |
46c0ee8b CL |
1580 | } |
1581 | ||
826799e6 | 1582 | static int xs_get_random_port(void) |
b85d8806 | 1583 | { |
826799e6 BF |
1584 | unsigned short min = xprt_min_resvport, max = xprt_max_resvport; |
1585 | unsigned short range; | |
1586 | unsigned short rand; | |
1587 | ||
1588 | if (max < min) | |
1589 | return -EADDRINUSE; | |
1590 | range = max - min + 1; | |
1591 | rand = (unsigned short) prandom_u32() % range; | |
1592 | return rand + min; | |
b85d8806 CL |
1593 | } |
1594 | ||
4dda9c8a TM |
1595 | static unsigned short xs_sock_getport(struct socket *sock) |
1596 | { | |
1597 | struct sockaddr_storage buf; | |
4dda9c8a TM |
1598 | unsigned short port = 0; |
1599 | ||
9b2c45d4 | 1600 | if (kernel_getsockname(sock, (struct sockaddr *)&buf) < 0) |
4dda9c8a TM |
1601 | goto out; |
1602 | switch (buf.ss_family) { | |
1603 | case AF_INET6: | |
1604 | port = ntohs(((struct sockaddr_in6 *)&buf)->sin6_port); | |
1605 | break; | |
1606 | case AF_INET: | |
1607 | port = ntohs(((struct sockaddr_in *)&buf)->sin_port); | |
1608 | } | |
1609 | out: | |
1610 | return port; | |
1611 | } | |
1612 | ||
92200412 CL |
1613 | /** |
1614 | * xs_set_port - reset the port number in the remote endpoint address | |
1615 | * @xprt: generic transport | |
1616 | * @port: new port number | |
1617 | * | |
1618 | */ | |
1619 | static void xs_set_port(struct rpc_xprt *xprt, unsigned short port) | |
1620 | { | |
46121cf7 | 1621 | dprintk("RPC: setting port for xprt %p to %u\n", xprt, port); |
c4efcb1d | 1622 | |
9dc3b095 CL |
1623 | rpc_set_port(xs_addr(xprt), port); |
1624 | xs_update_peer_port(xprt); | |
92200412 CL |
1625 | } |
1626 | ||
4dda9c8a TM |
1627 | static void xs_set_srcport(struct sock_xprt *transport, struct socket *sock) |
1628 | { | |
e6237b6f | 1629 | if (transport->srcport == 0 && transport->xprt.reuseport) |
4dda9c8a TM |
1630 | transport->srcport = xs_sock_getport(sock); |
1631 | } | |
1632 | ||
826799e6 | 1633 | static int xs_get_srcport(struct sock_xprt *transport) |
67a391d7 | 1634 | { |
826799e6 | 1635 | int port = transport->srcport; |
67a391d7 TM |
1636 | |
1637 | if (port == 0 && transport->xprt.resvport) | |
1638 | port = xs_get_random_port(); | |
1639 | return port; | |
1640 | } | |
1641 | ||
baaf4e48 | 1642 | static unsigned short xs_next_srcport(struct sock_xprt *transport, unsigned short port) |
67a391d7 | 1643 | { |
fbfffbd5 CL |
1644 | if (transport->srcport != 0) |
1645 | transport->srcport = 0; | |
67a391d7 TM |
1646 | if (!transport->xprt.resvport) |
1647 | return 0; | |
1648 | if (port <= xprt_min_resvport || port > xprt_max_resvport) | |
1649 | return xprt_max_resvport; | |
1650 | return --port; | |
1651 | } | |
beb59b68 | 1652 | static int xs_bind(struct sock_xprt *transport, struct socket *sock) |
a246b010 | 1653 | { |
beb59b68 | 1654 | struct sockaddr_storage myaddr; |
67a391d7 | 1655 | int err, nloop = 0; |
826799e6 | 1656 | int port = xs_get_srcport(transport); |
67a391d7 | 1657 | unsigned short last; |
a246b010 | 1658 | |
0f7a622c CP |
1659 | /* |
1660 | * If we are asking for any ephemeral port (i.e. port == 0 && | |
1661 | * transport->xprt.resvport == 0), don't bind. Let the local | |
1662 | * port selection happen implicitly when the socket is used | |
1663 | * (for example at connect time). | |
1664 | * | |
1665 | * This ensures that we can continue to establish TCP | |
1666 | * connections even when all local ephemeral ports are already | |
1667 | * a part of some TCP connection. This makes no difference | |
12b20ce3 | 1668 | * for UDP sockets, but also doesn't harm them. |
0f7a622c CP |
1669 | * |
1670 | * If we're asking for any reserved port (i.e. port == 0 && | |
1671 | * transport->xprt.resvport == 1) xs_get_srcport above will | |
1672 | * ensure that port is non-zero and we will bind as needed. | |
1673 | */ | |
826799e6 BF |
1674 | if (port <= 0) |
1675 | return port; | |
0f7a622c | 1676 | |
beb59b68 | 1677 | memcpy(&myaddr, &transport->srcaddr, transport->xprt.addrlen); |
a246b010 | 1678 | do { |
beb59b68 PE |
1679 | rpc_set_port((struct sockaddr *)&myaddr, port); |
1680 | err = kernel_bind(sock, (struct sockaddr *)&myaddr, | |
1681 | transport->xprt.addrlen); | |
a246b010 | 1682 | if (err == 0) { |
fbfffbd5 | 1683 | transport->srcport = port; |
d3bc9a1d | 1684 | break; |
a246b010 | 1685 | } |
67a391d7 | 1686 | last = port; |
baaf4e48 | 1687 | port = xs_next_srcport(transport, port); |
67a391d7 TM |
1688 | if (port > last) |
1689 | nloop++; | |
1690 | } while (err == -EADDRINUSE && nloop != 2); | |
90058d37 | 1691 | |
4232e863 | 1692 | if (myaddr.ss_family == AF_INET) |
beb59b68 PE |
1693 | dprintk("RPC: %s %pI4:%u: %s (%d)\n", __func__, |
1694 | &((struct sockaddr_in *)&myaddr)->sin_addr, | |
1695 | port, err ? "failed" : "ok", err); | |
1696 | else | |
1697 | dprintk("RPC: %s %pI6:%u: %s (%d)\n", __func__, | |
1698 | &((struct sockaddr_in6 *)&myaddr)->sin6_addr, | |
1699 | port, err ? "failed" : "ok", err); | |
a246b010 CL |
1700 | return err; |
1701 | } | |
1702 | ||
176e21ee CL |
1703 | /* |
1704 | * We don't support autobind on AF_LOCAL sockets | |
1705 | */ | |
1706 | static void xs_local_rpcbind(struct rpc_task *task) | |
1707 | { | |
fb43d172 | 1708 | xprt_set_bound(task->tk_xprt); |
176e21ee CL |
1709 | } |
1710 | ||
1711 | static void xs_local_set_port(struct rpc_xprt *xprt, unsigned short port) | |
1712 | { | |
1713 | } | |
beb59b68 | 1714 | |
ed07536e PZ |
1715 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
1716 | static struct lock_class_key xs_key[2]; | |
1717 | static struct lock_class_key xs_slock_key[2]; | |
1718 | ||
176e21ee CL |
1719 | static inline void xs_reclassify_socketu(struct socket *sock) |
1720 | { | |
1721 | struct sock *sk = sock->sk; | |
1722 | ||
176e21ee CL |
1723 | sock_lock_init_class_and_name(sk, "slock-AF_LOCAL-RPC", |
1724 | &xs_slock_key[1], "sk_lock-AF_LOCAL-RPC", &xs_key[1]); | |
1725 | } | |
1726 | ||
8945ee5e | 1727 | static inline void xs_reclassify_socket4(struct socket *sock) |
ed07536e PZ |
1728 | { |
1729 | struct sock *sk = sock->sk; | |
8945ee5e | 1730 | |
8945ee5e CL |
1731 | sock_lock_init_class_and_name(sk, "slock-AF_INET-RPC", |
1732 | &xs_slock_key[0], "sk_lock-AF_INET-RPC", &xs_key[0]); | |
1733 | } | |
ed07536e | 1734 | |
8945ee5e CL |
1735 | static inline void xs_reclassify_socket6(struct socket *sock) |
1736 | { | |
1737 | struct sock *sk = sock->sk; | |
ed07536e | 1738 | |
8945ee5e CL |
1739 | sock_lock_init_class_and_name(sk, "slock-AF_INET6-RPC", |
1740 | &xs_slock_key[1], "sk_lock-AF_INET6-RPC", &xs_key[1]); | |
ed07536e | 1741 | } |
6bc9638a PE |
1742 | |
1743 | static inline void xs_reclassify_socket(int family, struct socket *sock) | |
1744 | { | |
fafc4e1e | 1745 | if (WARN_ON_ONCE(!sock_allow_reclassification(sock->sk))) |
1b7a1819 WAA |
1746 | return; |
1747 | ||
4232e863 | 1748 | switch (family) { |
176e21ee CL |
1749 | case AF_LOCAL: |
1750 | xs_reclassify_socketu(sock); | |
1751 | break; | |
4232e863 | 1752 | case AF_INET: |
6bc9638a | 1753 | xs_reclassify_socket4(sock); |
4232e863 CL |
1754 | break; |
1755 | case AF_INET6: | |
6bc9638a | 1756 | xs_reclassify_socket6(sock); |
4232e863 CL |
1757 | break; |
1758 | } | |
6bc9638a | 1759 | } |
ed07536e | 1760 | #else |
6bc9638a PE |
1761 | static inline void xs_reclassify_socket(int family, struct socket *sock) |
1762 | { | |
1763 | } | |
ed07536e PZ |
1764 | #endif |
1765 | ||
93dc41bd N |
1766 | static void xs_dummy_setup_socket(struct work_struct *work) |
1767 | { | |
1768 | } | |
1769 | ||
6bc9638a | 1770 | static struct socket *xs_create_sock(struct rpc_xprt *xprt, |
4dda9c8a TM |
1771 | struct sock_xprt *transport, int family, int type, |
1772 | int protocol, bool reuseport) | |
22f79326 | 1773 | { |
a73881c9 | 1774 | struct file *filp; |
22f79326 PE |
1775 | struct socket *sock; |
1776 | int err; | |
1777 | ||
6bc9638a | 1778 | err = __sock_create(xprt->xprt_net, family, type, protocol, &sock, 1); |
22f79326 PE |
1779 | if (err < 0) { |
1780 | dprintk("RPC: can't create %d transport socket (%d).\n", | |
1781 | protocol, -err); | |
1782 | goto out; | |
1783 | } | |
6bc9638a | 1784 | xs_reclassify_socket(family, sock); |
22f79326 | 1785 | |
4dda9c8a | 1786 | if (reuseport) |
fe31a326 | 1787 | sock_set_reuseport(sock->sk); |
4dda9c8a | 1788 | |
4cea288a BH |
1789 | err = xs_bind(transport, sock); |
1790 | if (err) { | |
22f79326 PE |
1791 | sock_release(sock); |
1792 | goto out; | |
1793 | } | |
1794 | ||
a73881c9 TM |
1795 | filp = sock_alloc_file(sock, O_NONBLOCK, NULL); |
1796 | if (IS_ERR(filp)) | |
1797 | return ERR_CAST(filp); | |
1798 | transport->file = filp; | |
1799 | ||
22f79326 PE |
1800 | return sock; |
1801 | out: | |
1802 | return ERR_PTR(err); | |
1803 | } | |
1804 | ||
176e21ee CL |
1805 | static int xs_local_finish_connecting(struct rpc_xprt *xprt, |
1806 | struct socket *sock) | |
1807 | { | |
1808 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, | |
1809 | xprt); | |
1810 | ||
1811 | if (!transport->inet) { | |
1812 | struct sock *sk = sock->sk; | |
1813 | ||
1814 | write_lock_bh(&sk->sk_callback_lock); | |
1815 | ||
1816 | xs_save_old_callbacks(transport, sk); | |
1817 | ||
1818 | sk->sk_user_data = xprt; | |
a2648094 | 1819 | sk->sk_data_ready = xs_data_ready; |
176e21ee | 1820 | sk->sk_write_space = xs_udp_write_space; |
b4411457 | 1821 | sock_set_flag(sk, SOCK_FASYNC); |
2118071d | 1822 | sk->sk_error_report = xs_error_report; |
176e21ee CL |
1823 | |
1824 | xprt_clear_connected(xprt); | |
1825 | ||
1826 | /* Reset to new socket */ | |
1827 | transport->sock = sock; | |
1828 | transport->inet = sk; | |
1829 | ||
1830 | write_unlock_bh(&sk->sk_callback_lock); | |
1831 | } | |
1832 | ||
ae053551 | 1833 | xs_stream_start_connect(transport); |
6c7a64e5 | 1834 | |
176e21ee CL |
1835 | return kernel_connect(sock, xs_addr(xprt), xprt->addrlen, 0); |
1836 | } | |
1837 | ||
1838 | /** | |
1839 | * xs_local_setup_socket - create AF_LOCAL socket, connect to a local endpoint | |
176e21ee | 1840 | * @transport: socket transport to connect |
176e21ee | 1841 | */ |
dc107402 | 1842 | static int xs_local_setup_socket(struct sock_xprt *transport) |
176e21ee | 1843 | { |
176e21ee | 1844 | struct rpc_xprt *xprt = &transport->xprt; |
a73881c9 | 1845 | struct file *filp; |
176e21ee | 1846 | struct socket *sock; |
68e9a246 | 1847 | int status; |
176e21ee | 1848 | |
176e21ee CL |
1849 | status = __sock_create(xprt->xprt_net, AF_LOCAL, |
1850 | SOCK_STREAM, 0, &sock, 1); | |
1851 | if (status < 0) { | |
1852 | dprintk("RPC: can't create AF_LOCAL " | |
1853 | "transport socket (%d).\n", -status); | |
1854 | goto out; | |
1855 | } | |
d1358917 | 1856 | xs_reclassify_socket(AF_LOCAL, sock); |
176e21ee | 1857 | |
a73881c9 TM |
1858 | filp = sock_alloc_file(sock, O_NONBLOCK, NULL); |
1859 | if (IS_ERR(filp)) { | |
1860 | status = PTR_ERR(filp); | |
1861 | goto out; | |
1862 | } | |
1863 | transport->file = filp; | |
1864 | ||
176e21ee CL |
1865 | dprintk("RPC: worker connecting xprt %p via AF_LOCAL to %s\n", |
1866 | xprt, xprt->address_strings[RPC_DISPLAY_ADDR]); | |
1867 | ||
1868 | status = xs_local_finish_connecting(xprt, sock); | |
40b5ea0c | 1869 | trace_rpc_socket_connect(xprt, sock, status); |
176e21ee CL |
1870 | switch (status) { |
1871 | case 0: | |
1872 | dprintk("RPC: xprt %p connected to %s\n", | |
1873 | xprt, xprt->address_strings[RPC_DISPLAY_ADDR]); | |
3968a8a5 CL |
1874 | xprt->stat.connect_count++; |
1875 | xprt->stat.connect_time += (long)jiffies - | |
1876 | xprt->stat.connect_start; | |
176e21ee | 1877 | xprt_set_connected(xprt); |
3601c4a9 | 1878 | case -ENOBUFS: |
176e21ee CL |
1879 | break; |
1880 | case -ENOENT: | |
1881 | dprintk("RPC: xprt %p: socket %s does not exist\n", | |
1882 | xprt, xprt->address_strings[RPC_DISPLAY_ADDR]); | |
1883 | break; | |
4a20a988 TM |
1884 | case -ECONNREFUSED: |
1885 | dprintk("RPC: xprt %p: connection refused for %s\n", | |
1886 | xprt, xprt->address_strings[RPC_DISPLAY_ADDR]); | |
1887 | break; | |
176e21ee CL |
1888 | default: |
1889 | printk(KERN_ERR "%s: unhandled error (%d) connecting to %s\n", | |
1890 | __func__, -status, | |
1891 | xprt->address_strings[RPC_DISPLAY_ADDR]); | |
1892 | } | |
1893 | ||
1894 | out: | |
1895 | xprt_clear_connecting(xprt); | |
1896 | xprt_wake_pending_tasks(xprt, status); | |
dc107402 BF |
1897 | return status; |
1898 | } | |
1899 | ||
b6669737 | 1900 | static void xs_local_connect(struct rpc_xprt *xprt, struct rpc_task *task) |
dc107402 | 1901 | { |
dc107402 BF |
1902 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
1903 | int ret; | |
1904 | ||
1905 | if (RPC_IS_ASYNC(task)) { | |
1906 | /* | |
1907 | * We want the AF_LOCAL connect to be resolved in the | |
1908 | * filesystem namespace of the process making the rpc | |
1909 | * call. Thus we connect synchronously. | |
1910 | * | |
1911 | * If we want to support asynchronous AF_LOCAL calls, | |
1912 | * we'll need to figure out how to pass a namespace to | |
1913 | * connect. | |
1914 | */ | |
5ad64b36 | 1915 | task->tk_rpc_status = -ENOTCONN; |
dc107402 BF |
1916 | rpc_exit(task, -ENOTCONN); |
1917 | return; | |
1918 | } | |
1919 | ret = xs_local_setup_socket(transport); | |
1920 | if (ret && !RPC_IS_SOFTCONN(task)) | |
1921 | msleep_interruptible(15000); | |
176e21ee CL |
1922 | } |
1923 | ||
3c87ef6e | 1924 | #if IS_ENABLED(CONFIG_SUNRPC_SWAP) |
d6e971d8 JL |
1925 | /* |
1926 | * Note that this should be called with XPRT_LOCKED held (or when we otherwise | |
1927 | * know that we have exclusive access to the socket), to guard against | |
1928 | * races with xs_reset_transport. | |
1929 | */ | |
a564b8f0 MG |
1930 | static void xs_set_memalloc(struct rpc_xprt *xprt) |
1931 | { | |
1932 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, | |
1933 | xprt); | |
1934 | ||
d6e971d8 JL |
1935 | /* |
1936 | * If there's no sock, then we have nothing to set. The | |
1937 | * reconnecting process will get it for us. | |
1938 | */ | |
1939 | if (!transport->inet) | |
1940 | return; | |
8e228133 | 1941 | if (atomic_read(&xprt->swapper)) |
a564b8f0 MG |
1942 | sk_set_memalloc(transport->inet); |
1943 | } | |
1944 | ||
1945 | /** | |
d67fa4d8 | 1946 | * xs_enable_swap - Tag this transport as being used for swap. |
a564b8f0 | 1947 | * @xprt: transport to tag |
a564b8f0 | 1948 | * |
8e228133 JL |
1949 | * Take a reference to this transport on behalf of the rpc_clnt, and |
1950 | * optionally mark it for swapping if it wasn't already. | |
a564b8f0 | 1951 | */ |
d67fa4d8 JL |
1952 | static int |
1953 | xs_enable_swap(struct rpc_xprt *xprt) | |
a564b8f0 | 1954 | { |
d6e971d8 | 1955 | struct sock_xprt *xs = container_of(xprt, struct sock_xprt, xprt); |
a564b8f0 | 1956 | |
d6e971d8 JL |
1957 | if (atomic_inc_return(&xprt->swapper) != 1) |
1958 | return 0; | |
1959 | if (wait_on_bit_lock(&xprt->state, XPRT_LOCKED, TASK_KILLABLE)) | |
1960 | return -ERESTARTSYS; | |
1961 | if (xs->inet) | |
1962 | sk_set_memalloc(xs->inet); | |
1963 | xprt_release_xprt(xprt, NULL); | |
8e228133 JL |
1964 | return 0; |
1965 | } | |
a564b8f0 | 1966 | |
8e228133 | 1967 | /** |
d67fa4d8 | 1968 | * xs_disable_swap - Untag this transport as being used for swap. |
8e228133 JL |
1969 | * @xprt: transport to tag |
1970 | * | |
1971 | * Drop a "swapper" reference to this xprt on behalf of the rpc_clnt. If the | |
1972 | * swapper refcount goes to 0, untag the socket as a memalloc socket. | |
1973 | */ | |
d67fa4d8 JL |
1974 | static void |
1975 | xs_disable_swap(struct rpc_xprt *xprt) | |
8e228133 | 1976 | { |
d6e971d8 | 1977 | struct sock_xprt *xs = container_of(xprt, struct sock_xprt, xprt); |
8e228133 | 1978 | |
d6e971d8 JL |
1979 | if (!atomic_dec_and_test(&xprt->swapper)) |
1980 | return; | |
1981 | if (wait_on_bit_lock(&xprt->state, XPRT_LOCKED, TASK_KILLABLE)) | |
1982 | return; | |
1983 | if (xs->inet) | |
1984 | sk_clear_memalloc(xs->inet); | |
1985 | xprt_release_xprt(xprt, NULL); | |
a564b8f0 | 1986 | } |
a564b8f0 MG |
1987 | #else |
1988 | static void xs_set_memalloc(struct rpc_xprt *xprt) | |
1989 | { | |
1990 | } | |
d67fa4d8 JL |
1991 | |
1992 | static int | |
1993 | xs_enable_swap(struct rpc_xprt *xprt) | |
1994 | { | |
1995 | return -EINVAL; | |
1996 | } | |
1997 | ||
1998 | static void | |
1999 | xs_disable_swap(struct rpc_xprt *xprt) | |
2000 | { | |
2001 | } | |
a564b8f0 MG |
2002 | #endif |
2003 | ||
16be2d20 CL |
2004 | static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock) |
2005 | { | |
2006 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); | |
2007 | ||
2008 | if (!transport->inet) { | |
2009 | struct sock *sk = sock->sk; | |
2010 | ||
2011 | write_lock_bh(&sk->sk_callback_lock); | |
2012 | ||
2a9e1cfa TM |
2013 | xs_save_old_callbacks(transport, sk); |
2014 | ||
16be2d20 | 2015 | sk->sk_user_data = xprt; |
f9b2ee71 | 2016 | sk->sk_data_ready = xs_data_ready; |
16be2d20 | 2017 | sk->sk_write_space = xs_udp_write_space; |
b4411457 | 2018 | sock_set_flag(sk, SOCK_FASYNC); |
16be2d20 CL |
2019 | |
2020 | xprt_set_connected(xprt); | |
2021 | ||
2022 | /* Reset to new socket */ | |
2023 | transport->sock = sock; | |
2024 | transport->inet = sk; | |
2025 | ||
a564b8f0 MG |
2026 | xs_set_memalloc(xprt); |
2027 | ||
16be2d20 CL |
2028 | write_unlock_bh(&sk->sk_callback_lock); |
2029 | } | |
2030 | xs_udp_do_set_buffer_size(xprt); | |
02910177 TM |
2031 | |
2032 | xprt->stat.connect_start = jiffies; | |
16be2d20 CL |
2033 | } |
2034 | ||
8c14ff2a | 2035 | static void xs_udp_setup_socket(struct work_struct *work) |
a246b010 | 2036 | { |
34161db6 TM |
2037 | struct sock_xprt *transport = |
2038 | container_of(work, struct sock_xprt, connect_worker.work); | |
c8475461 | 2039 | struct rpc_xprt *xprt = &transport->xprt; |
d099b8af | 2040 | struct socket *sock; |
b65c0310 | 2041 | int status = -EIO; |
9903cd1c | 2042 | |
8c14ff2a | 2043 | sock = xs_create_sock(xprt, transport, |
4dda9c8a TM |
2044 | xs_addr(xprt)->sa_family, SOCK_DGRAM, |
2045 | IPPROTO_UDP, false); | |
b65c0310 | 2046 | if (IS_ERR(sock)) |
b0d93ad5 | 2047 | goto out; |
68e220bd | 2048 | |
c740eff8 CL |
2049 | dprintk("RPC: worker connecting xprt %p via %s to " |
2050 | "%s (port %s)\n", xprt, | |
2051 | xprt->address_strings[RPC_DISPLAY_PROTO], | |
2052 | xprt->address_strings[RPC_DISPLAY_ADDR], | |
2053 | xprt->address_strings[RPC_DISPLAY_PORT]); | |
68e220bd CL |
2054 | |
2055 | xs_udp_finish_connecting(xprt, sock); | |
40b5ea0c | 2056 | trace_rpc_socket_connect(xprt, sock, 0); |
b0d93ad5 CL |
2057 | status = 0; |
2058 | out: | |
b0d93ad5 | 2059 | xprt_clear_connecting(xprt); |
cf76785d | 2060 | xprt_unlock_connect(xprt, transport); |
7d1e8255 | 2061 | xprt_wake_pending_tasks(xprt, status); |
a246b010 CL |
2062 | } |
2063 | ||
4876cc77 TM |
2064 | /** |
2065 | * xs_tcp_shutdown - gracefully shut down a TCP socket | |
2066 | * @xprt: transport | |
2067 | * | |
2068 | * Initiates a graceful shutdown of the TCP socket by calling the | |
2069 | * equivalent of shutdown(SHUT_RDWR); | |
2070 | */ | |
2071 | static void xs_tcp_shutdown(struct rpc_xprt *xprt) | |
2072 | { | |
2073 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); | |
2074 | struct socket *sock = transport->sock; | |
9b30889c | 2075 | int skst = transport->inet ? transport->inet->sk_state : TCP_CLOSE; |
4876cc77 TM |
2076 | |
2077 | if (sock == NULL) | |
2078 | return; | |
9b30889c TM |
2079 | switch (skst) { |
2080 | default: | |
4876cc77 TM |
2081 | kernel_sock_shutdown(sock, SHUT_RDWR); |
2082 | trace_rpc_socket_shutdown(xprt, sock); | |
9b30889c TM |
2083 | break; |
2084 | case TCP_CLOSE: | |
2085 | case TCP_TIME_WAIT: | |
4876cc77 | 2086 | xs_reset_transport(transport); |
9b30889c | 2087 | } |
4876cc77 TM |
2088 | } |
2089 | ||
8d1b8c62 TM |
2090 | static void xs_tcp_set_socket_timeouts(struct rpc_xprt *xprt, |
2091 | struct socket *sock) | |
2092 | { | |
7196dbb0 TM |
2093 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
2094 | unsigned int keepidle; | |
2095 | unsigned int keepcnt; | |
8d1b8c62 TM |
2096 | unsigned int timeo; |
2097 | ||
b5e92419 | 2098 | spin_lock(&xprt->transport_lock); |
7196dbb0 TM |
2099 | keepidle = DIV_ROUND_UP(xprt->timeout->to_initval, HZ); |
2100 | keepcnt = xprt->timeout->to_retries + 1; | |
2101 | timeo = jiffies_to_msecs(xprt->timeout->to_initval) * | |
2102 | (xprt->timeout->to_retries + 1); | |
2103 | clear_bit(XPRT_SOCK_UPD_TIMEOUT, &transport->sock_state); | |
b5e92419 | 2104 | spin_unlock(&xprt->transport_lock); |
7196dbb0 | 2105 | |
8d1b8c62 | 2106 | /* TCP Keepalive options */ |
ce3d9544 | 2107 | sock_set_keepalive(sock->sk); |
71c48eb8 | 2108 | tcp_sock_set_keepidle(sock->sk, keepidle); |
d41ecaac | 2109 | tcp_sock_set_keepintvl(sock->sk, keepidle); |
480aeb96 | 2110 | tcp_sock_set_keepcnt(sock->sk, keepcnt); |
8d1b8c62 TM |
2111 | |
2112 | /* TCP user timeout (see RFC5482) */ | |
c488aead | 2113 | tcp_sock_set_user_timeout(sock->sk, timeo); |
8d1b8c62 TM |
2114 | } |
2115 | ||
7196dbb0 TM |
2116 | static void xs_tcp_set_connect_timeout(struct rpc_xprt *xprt, |
2117 | unsigned long connect_timeout, | |
2118 | unsigned long reconnect_timeout) | |
2119 | { | |
2120 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); | |
2121 | struct rpc_timeout to; | |
2122 | unsigned long initval; | |
2123 | ||
b5e92419 | 2124 | spin_lock(&xprt->transport_lock); |
7196dbb0 TM |
2125 | if (reconnect_timeout < xprt->max_reconnect_timeout) |
2126 | xprt->max_reconnect_timeout = reconnect_timeout; | |
2127 | if (connect_timeout < xprt->connect_timeout) { | |
2128 | memcpy(&to, xprt->timeout, sizeof(to)); | |
2129 | initval = DIV_ROUND_UP(connect_timeout, to.to_retries + 1); | |
2130 | /* Arbitrary lower limit */ | |
2131 | if (initval < XS_TCP_INIT_REEST_TO << 1) | |
2132 | initval = XS_TCP_INIT_REEST_TO << 1; | |
2133 | to.to_initval = initval; | |
2134 | to.to_maxval = initval; | |
2135 | memcpy(&transport->tcp_timeout, &to, | |
2136 | sizeof(transport->tcp_timeout)); | |
2137 | xprt->timeout = &transport->tcp_timeout; | |
2138 | xprt->connect_timeout = connect_timeout; | |
2139 | } | |
2140 | set_bit(XPRT_SOCK_UPD_TIMEOUT, &transport->sock_state); | |
b5e92419 | 2141 | spin_unlock(&xprt->transport_lock); |
7196dbb0 TM |
2142 | } |
2143 | ||
16be2d20 | 2144 | static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock) |
a246b010 | 2145 | { |
16be2d20 | 2146 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
fe19a96b | 2147 | int ret = -ENOTCONN; |
edb267a6 | 2148 | |
ee0ac0c2 | 2149 | if (!transport->inet) { |
b0d93ad5 | 2150 | struct sock *sk = sock->sk; |
7f260e85 | 2151 | |
d88e4d82 N |
2152 | /* Avoid temporary address, they are bad for long-lived |
2153 | * connections such as NFS mounts. | |
2154 | * RFC4941, section 3.6 suggests that: | |
2155 | * Individual applications, which have specific | |
2156 | * knowledge about the normal duration of connections, | |
2157 | * MAY override this as appropriate. | |
2158 | */ | |
18d5ad62 CH |
2159 | if (xs_addr(xprt)->sa_family == PF_INET6) { |
2160 | ip6_sock_set_addr_preferences(sk, | |
2161 | IPV6_PREFER_SRC_PUBLIC); | |
2162 | } | |
d88e4d82 | 2163 | |
8d1b8c62 | 2164 | xs_tcp_set_socket_timeouts(xprt, sock); |
775f06ab | 2165 | |
b0d93ad5 CL |
2166 | write_lock_bh(&sk->sk_callback_lock); |
2167 | ||
2a9e1cfa TM |
2168 | xs_save_old_callbacks(transport, sk); |
2169 | ||
b0d93ad5 | 2170 | sk->sk_user_data = xprt; |
5157b956 | 2171 | sk->sk_data_ready = xs_data_ready; |
b0d93ad5 CL |
2172 | sk->sk_state_change = xs_tcp_state_change; |
2173 | sk->sk_write_space = xs_tcp_write_space; | |
b4411457 | 2174 | sock_set_flag(sk, SOCK_FASYNC); |
2118071d | 2175 | sk->sk_error_report = xs_error_report; |
3167e12c CL |
2176 | |
2177 | /* socket options */ | |
3167e12c | 2178 | sock_reset_flag(sk, SOCK_LINGER); |
3167e12c | 2179 | tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF; |
b0d93ad5 CL |
2180 | |
2181 | xprt_clear_connected(xprt); | |
2182 | ||
2183 | /* Reset to new socket */ | |
ee0ac0c2 CL |
2184 | transport->sock = sock; |
2185 | transport->inet = sk; | |
b0d93ad5 CL |
2186 | |
2187 | write_unlock_bh(&sk->sk_callback_lock); | |
2188 | } | |
2189 | ||
01d37c42 | 2190 | if (!xprt_bound(xprt)) |
fe19a96b | 2191 | goto out; |
01d37c42 | 2192 | |
a564b8f0 MG |
2193 | xs_set_memalloc(xprt); |
2194 | ||
ae053551 | 2195 | xs_stream_start_connect(transport); |
e1806c7b | 2196 | |
b0d93ad5 | 2197 | /* Tell the socket layer to start connecting... */ |
0fdea1e8 | 2198 | set_bit(XPRT_SOCK_CONNECTING, &transport->sock_state); |
fe19a96b TM |
2199 | ret = kernel_connect(sock, xs_addr(xprt), xprt->addrlen, O_NONBLOCK); |
2200 | switch (ret) { | |
2201 | case 0: | |
4dda9c8a | 2202 | xs_set_srcport(transport, sock); |
df561f66 | 2203 | fallthrough; |
fe19a96b TM |
2204 | case -EINPROGRESS: |
2205 | /* SYN_SENT! */ | |
fe19a96b TM |
2206 | if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO) |
2207 | xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; | |
1f4c17a0 TM |
2208 | break; |
2209 | case -EADDRNOTAVAIL: | |
2210 | /* Source port number is unavailable. Try a new one! */ | |
2211 | transport->srcport = 0; | |
fe19a96b TM |
2212 | } |
2213 | out: | |
2214 | return ret; | |
16be2d20 CL |
2215 | } |
2216 | ||
9903cd1c | 2217 | /** |
b61d59ff | 2218 | * xs_tcp_setup_socket - create a TCP socket and connect to a remote endpoint |
acf0a39f | 2219 | * @work: queued work item |
9903cd1c CL |
2220 | * |
2221 | * Invoked by a work queue tasklet. | |
a246b010 | 2222 | */ |
cdd518d5 | 2223 | static void xs_tcp_setup_socket(struct work_struct *work) |
a246b010 | 2224 | { |
cdd518d5 PE |
2225 | struct sock_xprt *transport = |
2226 | container_of(work, struct sock_xprt, connect_worker.work); | |
ee0ac0c2 | 2227 | struct socket *sock = transport->sock; |
a9f5f0f7 | 2228 | struct rpc_xprt *xprt = &transport->xprt; |
b61d59ff | 2229 | int status = -EIO; |
a246b010 | 2230 | |
ee0ac0c2 | 2231 | if (!sock) { |
cdd518d5 | 2232 | sock = xs_create_sock(xprt, transport, |
4dda9c8a TM |
2233 | xs_addr(xprt)->sa_family, SOCK_STREAM, |
2234 | IPPROTO_TCP, true); | |
b61d59ff TM |
2235 | if (IS_ERR(sock)) { |
2236 | status = PTR_ERR(sock); | |
3167e12c CL |
2237 | goto out; |
2238 | } | |
7d1e8255 | 2239 | } |
a246b010 | 2240 | |
c740eff8 CL |
2241 | dprintk("RPC: worker connecting xprt %p via %s to " |
2242 | "%s (port %s)\n", xprt, | |
2243 | xprt->address_strings[RPC_DISPLAY_PROTO], | |
2244 | xprt->address_strings[RPC_DISPLAY_ADDR], | |
2245 | xprt->address_strings[RPC_DISPLAY_PORT]); | |
edb267a6 | 2246 | |
16be2d20 | 2247 | status = xs_tcp_finish_connecting(xprt, sock); |
40b5ea0c | 2248 | trace_rpc_socket_connect(xprt, sock, status); |
46121cf7 CL |
2249 | dprintk("RPC: %p connect status %d connected %d sock state %d\n", |
2250 | xprt, -status, xprt_connected(xprt), | |
2251 | sock->sk->sk_state); | |
2a491991 | 2252 | switch (status) { |
f75e6745 TM |
2253 | default: |
2254 | printk("%s: connect returned unhandled error %d\n", | |
2255 | __func__, status); | |
df561f66 | 2256 | fallthrough; |
f75e6745 TM |
2257 | case -EADDRNOTAVAIL: |
2258 | /* We're probably in TIME_WAIT. Get rid of existing socket, | |
2259 | * and retry | |
2260 | */ | |
a519fc7a | 2261 | xs_tcp_force_close(xprt); |
88b5ed73 | 2262 | break; |
2a491991 TM |
2263 | case 0: |
2264 | case -EINPROGRESS: | |
2265 | case -EALREADY: | |
718ba5b8 | 2266 | xprt_unlock_connect(xprt, transport); |
7d1e8255 | 2267 | return; |
9fcfe0c8 TM |
2268 | case -EINVAL: |
2269 | /* Happens, for instance, if the user specified a link | |
2270 | * local IPv6 address without a scope-id. | |
2271 | */ | |
3ed5e2a2 TM |
2272 | case -ECONNREFUSED: |
2273 | case -ECONNRESET: | |
eb5b46fa | 2274 | case -ENETDOWN: |
3ed5e2a2 | 2275 | case -ENETUNREACH: |
4ba161a7 | 2276 | case -EHOSTUNREACH: |
3913c78c | 2277 | case -EADDRINUSE: |
3601c4a9 | 2278 | case -ENOBUFS: |
8c71139d CM |
2279 | /* xs_tcp_force_close() wakes tasks with a fixed error code. |
2280 | * We need to wake them first to ensure the correct error code. | |
6ea44adc N |
2281 | */ |
2282 | xprt_wake_pending_tasks(xprt, status); | |
4efdd92c | 2283 | xs_tcp_force_close(xprt); |
9fcfe0c8 | 2284 | goto out; |
a246b010 | 2285 | } |
2a491991 | 2286 | status = -EAGAIN; |
a246b010 | 2287 | out: |
68e220bd | 2288 | xprt_clear_connecting(xprt); |
cf76785d | 2289 | xprt_unlock_connect(xprt, transport); |
7d1e8255 | 2290 | xprt_wake_pending_tasks(xprt, status); |
68e220bd | 2291 | } |
b0d93ad5 | 2292 | |
9903cd1c CL |
2293 | /** |
2294 | * xs_connect - connect a socket to a remote endpoint | |
1b092092 | 2295 | * @xprt: pointer to transport structure |
9903cd1c CL |
2296 | * @task: address of RPC task that manages state of connect request |
2297 | * | |
2298 | * TCP: If the remote end dropped the connection, delay reconnecting. | |
03bf4b70 CL |
2299 | * |
2300 | * UDP socket connects are synchronous, but we use a work queue anyway | |
2301 | * to guarantee that even unprivileged user processes can set up a | |
2302 | * socket on a privileged port. | |
2303 | * | |
2304 | * If a UDP socket connect fails, the delay behavior here prevents | |
2305 | * retry floods (hard mounts). | |
9903cd1c | 2306 | */ |
1b092092 | 2307 | static void xs_connect(struct rpc_xprt *xprt, struct rpc_task *task) |
a246b010 | 2308 | { |
ee0ac0c2 | 2309 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
02910177 | 2310 | unsigned long delay = 0; |
a246b010 | 2311 | |
718ba5b8 TM |
2312 | WARN_ON_ONCE(!xprt_lock_connect(xprt, task, transport)); |
2313 | ||
99b1a4c3 | 2314 | if (transport->sock != NULL) { |
46121cf7 CL |
2315 | dprintk("RPC: xs_connect delayed xprt %p for %lu " |
2316 | "seconds\n", | |
03bf4b70 | 2317 | xprt, xprt->reestablish_timeout / HZ); |
99b1a4c3 TM |
2318 | |
2319 | /* Start by resetting any existing state */ | |
2320 | xs_reset_transport(transport); | |
2321 | ||
675dd90a CL |
2322 | delay = xprt_reconnect_delay(xprt); |
2323 | xprt_reconnect_backoff(xprt, XS_TCP_INIT_REEST_TO); | |
02910177 | 2324 | |
02910177 | 2325 | } else |
46121cf7 | 2326 | dprintk("RPC: xs_connect scheduled xprt %p\n", xprt); |
02910177 TM |
2327 | |
2328 | queue_delayed_work(xprtiod_workqueue, | |
2329 | &transport->connect_worker, | |
2330 | delay); | |
a246b010 CL |
2331 | } |
2332 | ||
4f8943f8 TM |
2333 | static void xs_wake_disconnect(struct sock_xprt *transport) |
2334 | { | |
2335 | if (test_and_clear_bit(XPRT_SOCK_WAKE_DISCONNECT, &transport->sock_state)) | |
2336 | xs_tcp_force_close(&transport->xprt); | |
2337 | } | |
2338 | ||
2339 | static void xs_wake_write(struct sock_xprt *transport) | |
2340 | { | |
2341 | if (test_and_clear_bit(XPRT_SOCK_WAKE_WRITE, &transport->sock_state)) | |
2342 | xprt_write_space(&transport->xprt); | |
2343 | } | |
2344 | ||
2345 | static void xs_wake_error(struct sock_xprt *transport) | |
2346 | { | |
2347 | int sockerr; | |
4f8943f8 TM |
2348 | |
2349 | if (!test_bit(XPRT_SOCK_WAKE_ERROR, &transport->sock_state)) | |
2350 | return; | |
2351 | mutex_lock(&transport->recv_mutex); | |
2352 | if (transport->sock == NULL) | |
2353 | goto out; | |
2354 | if (!test_and_clear_bit(XPRT_SOCK_WAKE_ERROR, &transport->sock_state)) | |
2355 | goto out; | |
af84537d | 2356 | sockerr = xchg(&transport->xprt_err, 0); |
4f8943f8 TM |
2357 | if (sockerr < 0) |
2358 | xprt_wake_pending_tasks(&transport->xprt, sockerr); | |
2359 | out: | |
2360 | mutex_unlock(&transport->recv_mutex); | |
2361 | } | |
2362 | ||
2363 | static void xs_wake_pending(struct sock_xprt *transport) | |
2364 | { | |
2365 | if (test_and_clear_bit(XPRT_SOCK_WAKE_PENDING, &transport->sock_state)) | |
2366 | xprt_wake_pending_tasks(&transport->xprt, -EAGAIN); | |
2367 | } | |
2368 | ||
2369 | static void xs_error_handle(struct work_struct *work) | |
2370 | { | |
2371 | struct sock_xprt *transport = container_of(work, | |
2372 | struct sock_xprt, error_worker); | |
2373 | ||
2374 | xs_wake_disconnect(transport); | |
2375 | xs_wake_write(transport); | |
2376 | xs_wake_error(transport); | |
2377 | xs_wake_pending(transport); | |
2378 | } | |
2379 | ||
176e21ee | 2380 | /** |
12b20ce3 | 2381 | * xs_local_print_stats - display AF_LOCAL socket-specific stats |
176e21ee CL |
2382 | * @xprt: rpc_xprt struct containing statistics |
2383 | * @seq: output file | |
2384 | * | |
2385 | */ | |
2386 | static void xs_local_print_stats(struct rpc_xprt *xprt, struct seq_file *seq) | |
2387 | { | |
2388 | long idle_time = 0; | |
2389 | ||
2390 | if (xprt_connected(xprt)) | |
2391 | idle_time = (long)(jiffies - xprt->last_used) / HZ; | |
2392 | ||
2393 | seq_printf(seq, "\txprt:\tlocal %lu %lu %lu %ld %lu %lu %lu " | |
15a45206 | 2394 | "%llu %llu %lu %llu %llu\n", |
176e21ee CL |
2395 | xprt->stat.bind_count, |
2396 | xprt->stat.connect_count, | |
8440a886 | 2397 | xprt->stat.connect_time / HZ, |
176e21ee CL |
2398 | idle_time, |
2399 | xprt->stat.sends, | |
2400 | xprt->stat.recvs, | |
2401 | xprt->stat.bad_xids, | |
2402 | xprt->stat.req_u, | |
15a45206 AA |
2403 | xprt->stat.bklog_u, |
2404 | xprt->stat.max_slots, | |
2405 | xprt->stat.sending_u, | |
2406 | xprt->stat.pending_u); | |
176e21ee CL |
2407 | } |
2408 | ||
262ca07d | 2409 | /** |
12b20ce3 | 2410 | * xs_udp_print_stats - display UDP socket-specific stats |
262ca07d CL |
2411 | * @xprt: rpc_xprt struct containing statistics |
2412 | * @seq: output file | |
2413 | * | |
2414 | */ | |
2415 | static void xs_udp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq) | |
2416 | { | |
c8475461 CL |
2417 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
2418 | ||
15a45206 AA |
2419 | seq_printf(seq, "\txprt:\tudp %u %lu %lu %lu %lu %llu %llu " |
2420 | "%lu %llu %llu\n", | |
fbfffbd5 | 2421 | transport->srcport, |
262ca07d CL |
2422 | xprt->stat.bind_count, |
2423 | xprt->stat.sends, | |
2424 | xprt->stat.recvs, | |
2425 | xprt->stat.bad_xids, | |
2426 | xprt->stat.req_u, | |
15a45206 AA |
2427 | xprt->stat.bklog_u, |
2428 | xprt->stat.max_slots, | |
2429 | xprt->stat.sending_u, | |
2430 | xprt->stat.pending_u); | |
262ca07d CL |
2431 | } |
2432 | ||
2433 | /** | |
12b20ce3 | 2434 | * xs_tcp_print_stats - display TCP socket-specific stats |
262ca07d CL |
2435 | * @xprt: rpc_xprt struct containing statistics |
2436 | * @seq: output file | |
2437 | * | |
2438 | */ | |
2439 | static void xs_tcp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq) | |
2440 | { | |
c8475461 | 2441 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
262ca07d CL |
2442 | long idle_time = 0; |
2443 | ||
2444 | if (xprt_connected(xprt)) | |
2445 | idle_time = (long)(jiffies - xprt->last_used) / HZ; | |
2446 | ||
15a45206 AA |
2447 | seq_printf(seq, "\txprt:\ttcp %u %lu %lu %lu %ld %lu %lu %lu " |
2448 | "%llu %llu %lu %llu %llu\n", | |
fbfffbd5 | 2449 | transport->srcport, |
262ca07d CL |
2450 | xprt->stat.bind_count, |
2451 | xprt->stat.connect_count, | |
8440a886 | 2452 | xprt->stat.connect_time / HZ, |
262ca07d CL |
2453 | idle_time, |
2454 | xprt->stat.sends, | |
2455 | xprt->stat.recvs, | |
2456 | xprt->stat.bad_xids, | |
2457 | xprt->stat.req_u, | |
15a45206 AA |
2458 | xprt->stat.bklog_u, |
2459 | xprt->stat.max_slots, | |
2460 | xprt->stat.sending_u, | |
2461 | xprt->stat.pending_u); | |
262ca07d CL |
2462 | } |
2463 | ||
4cfc7e60 RI |
2464 | /* |
2465 | * Allocate a bunch of pages for a scratch buffer for the rpc code. The reason | |
2466 | * we allocate pages instead doing a kmalloc like rpc_malloc is because we want | |
2467 | * to use the server side send routines. | |
2468 | */ | |
5fe6eaa1 | 2469 | static int bc_malloc(struct rpc_task *task) |
4cfc7e60 | 2470 | { |
5fe6eaa1 CL |
2471 | struct rpc_rqst *rqst = task->tk_rqstp; |
2472 | size_t size = rqst->rq_callsize; | |
4cfc7e60 RI |
2473 | struct page *page; |
2474 | struct rpc_buffer *buf; | |
2475 | ||
5fe6eaa1 CL |
2476 | if (size > PAGE_SIZE - sizeof(struct rpc_buffer)) { |
2477 | WARN_ONCE(1, "xprtsock: large bc buffer request (size %zu)\n", | |
2478 | size); | |
2479 | return -EINVAL; | |
2480 | } | |
4cfc7e60 | 2481 | |
b8a13d03 | 2482 | page = alloc_page(GFP_KERNEL); |
4cfc7e60 | 2483 | if (!page) |
5fe6eaa1 | 2484 | return -ENOMEM; |
4cfc7e60 RI |
2485 | |
2486 | buf = page_address(page); | |
2487 | buf->len = PAGE_SIZE; | |
2488 | ||
5fe6eaa1 | 2489 | rqst->rq_buffer = buf->data; |
18e601d6 | 2490 | rqst->rq_rbuffer = (char *)rqst->rq_buffer + rqst->rq_callsize; |
5fe6eaa1 | 2491 | return 0; |
4cfc7e60 RI |
2492 | } |
2493 | ||
2494 | /* | |
2495 | * Free the space allocated in the bc_alloc routine | |
2496 | */ | |
3435c74a | 2497 | static void bc_free(struct rpc_task *task) |
4cfc7e60 | 2498 | { |
3435c74a | 2499 | void *buffer = task->tk_rqstp->rq_buffer; |
4cfc7e60 RI |
2500 | struct rpc_buffer *buf; |
2501 | ||
4cfc7e60 RI |
2502 | buf = container_of(buffer, struct rpc_buffer, data); |
2503 | free_page((unsigned long)buf); | |
2504 | } | |
2505 | ||
4cfc7e60 RI |
2506 | static int bc_sendto(struct rpc_rqst *req) |
2507 | { | |
da1661b9 | 2508 | struct xdr_buf *xdr = &req->rq_snd_buf; |
4cfc7e60 | 2509 | struct sock_xprt *transport = |
067fb11b | 2510 | container_of(req->rq_xprt, struct sock_xprt, xprt); |
067fb11b | 2511 | struct msghdr msg = { |
da1661b9 | 2512 | .msg_flags = 0, |
067fb11b CL |
2513 | }; |
2514 | rpc_fraghdr marker = cpu_to_be32(RPC_LAST_STREAM_FRAGMENT | | |
da1661b9 CL |
2515 | (u32)xdr->len); |
2516 | unsigned int sent = 0; | |
2517 | int err; | |
4cfc7e60 | 2518 | |
8729aaba | 2519 | req->rq_xtime = ktime_get(); |
da1661b9 CL |
2520 | err = xprt_sock_sendmsg(transport->sock, &msg, xdr, 0, marker, &sent); |
2521 | xdr_free_bvec(xdr); | |
2522 | if (err < 0 || sent != (xdr->len + sizeof(marker))) | |
067fb11b | 2523 | return -EAGAIN; |
da1661b9 | 2524 | return sent; |
4cfc7e60 RI |
2525 | } |
2526 | ||
ca4faf54 CL |
2527 | /** |
2528 | * bc_send_request - Send a backchannel Call on a TCP socket | |
2529 | * @req: rpc_rqst containing Call message to be sent | |
2530 | * | |
2531 | * xpt_mutex ensures @rqstp's whole message is written to the socket | |
2532 | * without interruption. | |
2533 | * | |
2534 | * Return values: | |
2535 | * %0 if the message was sent successfully | |
2536 | * %ENOTCONN if the message was not sent | |
4cfc7e60 | 2537 | */ |
adfa7144 | 2538 | static int bc_send_request(struct rpc_rqst *req) |
4cfc7e60 | 2539 | { |
4cfc7e60 | 2540 | struct svc_xprt *xprt; |
7fc56136 | 2541 | int len; |
4cfc7e60 | 2542 | |
4cfc7e60 RI |
2543 | /* |
2544 | * Get the server socket associated with this callback xprt | |
2545 | */ | |
2546 | xprt = req->rq_xprt->bc_xprt; | |
4cfc7e60 RI |
2547 | |
2548 | /* | |
2549 | * Grab the mutex to serialize data as the connection is shared | |
2550 | * with the fore channel | |
2551 | */ | |
c544577d | 2552 | mutex_lock(&xprt->xpt_mutex); |
4cfc7e60 RI |
2553 | if (test_bit(XPT_DEAD, &xprt->xpt_flags)) |
2554 | len = -ENOTCONN; | |
2555 | else | |
2556 | len = bc_sendto(req); | |
2557 | mutex_unlock(&xprt->xpt_mutex); | |
2558 | ||
2559 | if (len > 0) | |
2560 | len = 0; | |
2561 | ||
2562 | return len; | |
2563 | } | |
2564 | ||
2565 | /* | |
2566 | * The close routine. Since this is client initiated, we do nothing | |
2567 | */ | |
2568 | ||
2569 | static void bc_close(struct rpc_xprt *xprt) | |
2570 | { | |
6221f1d9 | 2571 | xprt_disconnect_done(xprt); |
4cfc7e60 RI |
2572 | } |
2573 | ||
2574 | /* | |
2575 | * The xprt destroy routine. Again, because this connection is client | |
2576 | * initiated, we do nothing | |
2577 | */ | |
2578 | ||
2579 | static void bc_destroy(struct rpc_xprt *xprt) | |
2580 | { | |
47f72efa KM |
2581 | dprintk("RPC: bc_destroy xprt %p\n", xprt); |
2582 | ||
2583 | xs_xprt_free(xprt); | |
2584 | module_put(THIS_MODULE); | |
4cfc7e60 RI |
2585 | } |
2586 | ||
d31ae254 | 2587 | static const struct rpc_xprt_ops xs_local_ops = { |
176e21ee | 2588 | .reserve_xprt = xprt_reserve_xprt, |
4cd34e7c | 2589 | .release_xprt = xprt_release_xprt, |
f39c1bfb | 2590 | .alloc_slot = xprt_alloc_slot, |
a9cde23a | 2591 | .free_slot = xprt_free_slot, |
176e21ee CL |
2592 | .rpcbind = xs_local_rpcbind, |
2593 | .set_port = xs_local_set_port, | |
dc107402 | 2594 | .connect = xs_local_connect, |
176e21ee CL |
2595 | .buf_alloc = rpc_malloc, |
2596 | .buf_free = rpc_free, | |
550aebfe | 2597 | .prepare_request = xs_stream_prepare_request, |
176e21ee | 2598 | .send_request = xs_local_send_request, |
8ba6a92d | 2599 | .wait_for_reply_request = xprt_wait_for_reply_request_def, |
176e21ee | 2600 | .close = xs_close, |
a1311d87 | 2601 | .destroy = xs_destroy, |
176e21ee | 2602 | .print_stats = xs_local_print_stats, |
d67fa4d8 JL |
2603 | .enable_swap = xs_enable_swap, |
2604 | .disable_swap = xs_disable_swap, | |
176e21ee CL |
2605 | }; |
2606 | ||
d31ae254 | 2607 | static const struct rpc_xprt_ops xs_udp_ops = { |
43118c29 | 2608 | .set_buffer_size = xs_udp_set_buffer_size, |
12a80469 | 2609 | .reserve_xprt = xprt_reserve_xprt_cong, |
49e9a890 | 2610 | .release_xprt = xprt_release_xprt_cong, |
f39c1bfb | 2611 | .alloc_slot = xprt_alloc_slot, |
a9cde23a | 2612 | .free_slot = xprt_free_slot, |
45160d62 | 2613 | .rpcbind = rpcb_getport_async, |
92200412 | 2614 | .set_port = xs_set_port, |
262965f5 | 2615 | .connect = xs_connect, |
02107148 CL |
2616 | .buf_alloc = rpc_malloc, |
2617 | .buf_free = rpc_free, | |
262965f5 | 2618 | .send_request = xs_udp_send_request, |
8ba6a92d | 2619 | .wait_for_reply_request = xprt_wait_for_reply_request_rtt, |
46c0ee8b | 2620 | .timer = xs_udp_timer, |
a58dd398 | 2621 | .release_request = xprt_release_rqst_cong, |
262965f5 CL |
2622 | .close = xs_close, |
2623 | .destroy = xs_destroy, | |
262ca07d | 2624 | .print_stats = xs_udp_print_stats, |
d67fa4d8 JL |
2625 | .enable_swap = xs_enable_swap, |
2626 | .disable_swap = xs_disable_swap, | |
4a068258 | 2627 | .inject_disconnect = xs_inject_disconnect, |
262965f5 CL |
2628 | }; |
2629 | ||
d31ae254 | 2630 | static const struct rpc_xprt_ops xs_tcp_ops = { |
12a80469 | 2631 | .reserve_xprt = xprt_reserve_xprt, |
4cd34e7c | 2632 | .release_xprt = xprt_release_xprt, |
36bd7de9 | 2633 | .alloc_slot = xprt_alloc_slot, |
a9cde23a | 2634 | .free_slot = xprt_free_slot, |
45160d62 | 2635 | .rpcbind = rpcb_getport_async, |
92200412 | 2636 | .set_port = xs_set_port, |
0b9e7943 | 2637 | .connect = xs_connect, |
02107148 CL |
2638 | .buf_alloc = rpc_malloc, |
2639 | .buf_free = rpc_free, | |
277e4ab7 | 2640 | .prepare_request = xs_stream_prepare_request, |
262965f5 | 2641 | .send_request = xs_tcp_send_request, |
8ba6a92d | 2642 | .wait_for_reply_request = xprt_wait_for_reply_request_def, |
c627d31b | 2643 | .close = xs_tcp_shutdown, |
9903cd1c | 2644 | .destroy = xs_destroy, |
7196dbb0 | 2645 | .set_connect_timeout = xs_tcp_set_connect_timeout, |
262ca07d | 2646 | .print_stats = xs_tcp_print_stats, |
d67fa4d8 JL |
2647 | .enable_swap = xs_enable_swap, |
2648 | .disable_swap = xs_disable_swap, | |
4a068258 | 2649 | .inject_disconnect = xs_inject_disconnect, |
42e5c3e2 CL |
2650 | #ifdef CONFIG_SUNRPC_BACKCHANNEL |
2651 | .bc_setup = xprt_setup_bc, | |
6b26cc8c | 2652 | .bc_maxpayload = xs_tcp_bc_maxpayload, |
7402a4fe | 2653 | .bc_num_slots = xprt_bc_max_slots, |
42e5c3e2 CL |
2654 | .bc_free_rqst = xprt_free_bc_rqst, |
2655 | .bc_destroy = xprt_destroy_bc, | |
2656 | #endif | |
a246b010 CL |
2657 | }; |
2658 | ||
4cfc7e60 RI |
2659 | /* |
2660 | * The rpc_xprt_ops for the server backchannel | |
2661 | */ | |
2662 | ||
d31ae254 | 2663 | static const struct rpc_xprt_ops bc_tcp_ops = { |
4cfc7e60 RI |
2664 | .reserve_xprt = xprt_reserve_xprt, |
2665 | .release_xprt = xprt_release_xprt, | |
84e28a30 | 2666 | .alloc_slot = xprt_alloc_slot, |
a9cde23a | 2667 | .free_slot = xprt_free_slot, |
4cfc7e60 RI |
2668 | .buf_alloc = bc_malloc, |
2669 | .buf_free = bc_free, | |
2670 | .send_request = bc_send_request, | |
8ba6a92d | 2671 | .wait_for_reply_request = xprt_wait_for_reply_request_def, |
4cfc7e60 RI |
2672 | .close = bc_close, |
2673 | .destroy = bc_destroy, | |
2674 | .print_stats = xs_tcp_print_stats, | |
d67fa4d8 JL |
2675 | .enable_swap = xs_enable_swap, |
2676 | .disable_swap = xs_disable_swap, | |
4a068258 | 2677 | .inject_disconnect = xs_inject_disconnect, |
4cfc7e60 RI |
2678 | }; |
2679 | ||
92476850 CL |
2680 | static int xs_init_anyaddr(const int family, struct sockaddr *sap) |
2681 | { | |
2682 | static const struct sockaddr_in sin = { | |
2683 | .sin_family = AF_INET, | |
2684 | .sin_addr.s_addr = htonl(INADDR_ANY), | |
2685 | }; | |
2686 | static const struct sockaddr_in6 sin6 = { | |
2687 | .sin6_family = AF_INET6, | |
2688 | .sin6_addr = IN6ADDR_ANY_INIT, | |
2689 | }; | |
2690 | ||
2691 | switch (family) { | |
176e21ee CL |
2692 | case AF_LOCAL: |
2693 | break; | |
92476850 CL |
2694 | case AF_INET: |
2695 | memcpy(sap, &sin, sizeof(sin)); | |
2696 | break; | |
2697 | case AF_INET6: | |
2698 | memcpy(sap, &sin6, sizeof(sin6)); | |
2699 | break; | |
2700 | default: | |
2701 | dprintk("RPC: %s: Bad address family\n", __func__); | |
2702 | return -EAFNOSUPPORT; | |
2703 | } | |
2704 | return 0; | |
2705 | } | |
2706 | ||
3c341b0b | 2707 | static struct rpc_xprt *xs_setup_xprt(struct xprt_create *args, |
d9ba131d TM |
2708 | unsigned int slot_table_size, |
2709 | unsigned int max_slot_table_size) | |
c8541ecd CL |
2710 | { |
2711 | struct rpc_xprt *xprt; | |
ffc2e518 | 2712 | struct sock_xprt *new; |
c8541ecd | 2713 | |
96802a09 | 2714 | if (args->addrlen > sizeof(xprt->addr)) { |
46121cf7 | 2715 | dprintk("RPC: xs_setup_xprt: address too large\n"); |
c8541ecd CL |
2716 | return ERR_PTR(-EBADF); |
2717 | } | |
2718 | ||
d9ba131d TM |
2719 | xprt = xprt_alloc(args->net, sizeof(*new), slot_table_size, |
2720 | max_slot_table_size); | |
bd1722d4 | 2721 | if (xprt == NULL) { |
46121cf7 CL |
2722 | dprintk("RPC: xs_setup_xprt: couldn't allocate " |
2723 | "rpc_xprt\n"); | |
c8541ecd CL |
2724 | return ERR_PTR(-ENOMEM); |
2725 | } | |
c8541ecd | 2726 | |
bd1722d4 | 2727 | new = container_of(xprt, struct sock_xprt, xprt); |
edc1b01c | 2728 | mutex_init(&new->recv_mutex); |
96802a09 FM |
2729 | memcpy(&xprt->addr, args->dstaddr, args->addrlen); |
2730 | xprt->addrlen = args->addrlen; | |
d3bc9a1d | 2731 | if (args->srcaddr) |
fbfffbd5 | 2732 | memcpy(&new->srcaddr, args->srcaddr, args->addrlen); |
92476850 CL |
2733 | else { |
2734 | int err; | |
2735 | err = xs_init_anyaddr(args->dstaddr->sa_family, | |
2736 | (struct sockaddr *)&new->srcaddr); | |
2aa13531 SK |
2737 | if (err != 0) { |
2738 | xprt_free(xprt); | |
92476850 | 2739 | return ERR_PTR(err); |
2aa13531 | 2740 | } |
92476850 | 2741 | } |
c8541ecd CL |
2742 | |
2743 | return xprt; | |
2744 | } | |
2745 | ||
176e21ee CL |
2746 | static const struct rpc_timeout xs_local_default_timeout = { |
2747 | .to_initval = 10 * HZ, | |
2748 | .to_maxval = 10 * HZ, | |
2749 | .to_retries = 2, | |
2750 | }; | |
2751 | ||
2752 | /** | |
2753 | * xs_setup_local - Set up transport to use an AF_LOCAL socket | |
2754 | * @args: rpc transport creation arguments | |
2755 | * | |
2756 | * AF_LOCAL is a "tpi_cots_ord" transport, just like TCP | |
2757 | */ | |
2758 | static struct rpc_xprt *xs_setup_local(struct xprt_create *args) | |
2759 | { | |
2760 | struct sockaddr_un *sun = (struct sockaddr_un *)args->dstaddr; | |
2761 | struct sock_xprt *transport; | |
2762 | struct rpc_xprt *xprt; | |
2763 | struct rpc_xprt *ret; | |
2764 | ||
d9ba131d TM |
2765 | xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries, |
2766 | xprt_max_tcp_slot_table_entries); | |
176e21ee CL |
2767 | if (IS_ERR(xprt)) |
2768 | return xprt; | |
2769 | transport = container_of(xprt, struct sock_xprt, xprt); | |
2770 | ||
2771 | xprt->prot = 0; | |
176e21ee CL |
2772 | xprt->max_payload = RPC_MAX_FRAGMENT_SIZE; |
2773 | ||
2774 | xprt->bind_timeout = XS_BIND_TO; | |
2775 | xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; | |
2776 | xprt->idle_timeout = XS_IDLE_DISC_TO; | |
2777 | ||
2778 | xprt->ops = &xs_local_ops; | |
2779 | xprt->timeout = &xs_local_default_timeout; | |
2780 | ||
550aebfe | 2781 | INIT_WORK(&transport->recv_worker, xs_stream_data_receive_workfn); |
4f8943f8 | 2782 | INIT_WORK(&transport->error_worker, xs_error_handle); |
550aebfe | 2783 | INIT_DELAYED_WORK(&transport->connect_worker, xs_dummy_setup_socket); |
93dc41bd | 2784 | |
176e21ee CL |
2785 | switch (sun->sun_family) { |
2786 | case AF_LOCAL: | |
2787 | if (sun->sun_path[0] != '/') { | |
2788 | dprintk("RPC: bad AF_LOCAL address: %s\n", | |
2789 | sun->sun_path); | |
2790 | ret = ERR_PTR(-EINVAL); | |
2791 | goto out_err; | |
2792 | } | |
2793 | xprt_set_bound(xprt); | |
176e21ee | 2794 | xs_format_peer_addresses(xprt, "local", RPCBIND_NETID_LOCAL); |
7073ea87 BF |
2795 | ret = ERR_PTR(xs_local_setup_socket(transport)); |
2796 | if (ret) | |
2797 | goto out_err; | |
176e21ee CL |
2798 | break; |
2799 | default: | |
2800 | ret = ERR_PTR(-EAFNOSUPPORT); | |
2801 | goto out_err; | |
2802 | } | |
2803 | ||
2804 | dprintk("RPC: set up xprt to %s via AF_LOCAL\n", | |
2805 | xprt->address_strings[RPC_DISPLAY_ADDR]); | |
2806 | ||
2807 | if (try_module_get(THIS_MODULE)) | |
2808 | return xprt; | |
2809 | ret = ERR_PTR(-EINVAL); | |
2810 | out_err: | |
315f3812 | 2811 | xs_xprt_free(xprt); |
176e21ee CL |
2812 | return ret; |
2813 | } | |
2814 | ||
2881ae74 TM |
2815 | static const struct rpc_timeout xs_udp_default_timeout = { |
2816 | .to_initval = 5 * HZ, | |
2817 | .to_maxval = 30 * HZ, | |
2818 | .to_increment = 5 * HZ, | |
2819 | .to_retries = 5, | |
2820 | }; | |
2821 | ||
9903cd1c CL |
2822 | /** |
2823 | * xs_setup_udp - Set up transport to use a UDP socket | |
96802a09 | 2824 | * @args: rpc transport creation arguments |
9903cd1c CL |
2825 | * |
2826 | */ | |
483066d6 | 2827 | static struct rpc_xprt *xs_setup_udp(struct xprt_create *args) |
a246b010 | 2828 | { |
8f9d5b1a | 2829 | struct sockaddr *addr = args->dstaddr; |
c8541ecd | 2830 | struct rpc_xprt *xprt; |
c8475461 | 2831 | struct sock_xprt *transport; |
0a68b0be | 2832 | struct rpc_xprt *ret; |
a246b010 | 2833 | |
d9ba131d TM |
2834 | xprt = xs_setup_xprt(args, xprt_udp_slot_table_entries, |
2835 | xprt_udp_slot_table_entries); | |
c8541ecd CL |
2836 | if (IS_ERR(xprt)) |
2837 | return xprt; | |
c8475461 | 2838 | transport = container_of(xprt, struct sock_xprt, xprt); |
a246b010 | 2839 | |
ec739ef0 | 2840 | xprt->prot = IPPROTO_UDP; |
a246b010 CL |
2841 | /* XXX: header size can vary due to auth type, IPv6, etc. */ |
2842 | xprt->max_payload = (1U << 16) - (MAX_HEADER << 3); | |
2843 | ||
03bf4b70 | 2844 | xprt->bind_timeout = XS_BIND_TO; |
03bf4b70 CL |
2845 | xprt->reestablish_timeout = XS_UDP_REEST_TO; |
2846 | xprt->idle_timeout = XS_IDLE_DISC_TO; | |
a246b010 | 2847 | |
262965f5 | 2848 | xprt->ops = &xs_udp_ops; |
a246b010 | 2849 | |
ba7392bb | 2850 | xprt->timeout = &xs_udp_default_timeout; |
a246b010 | 2851 | |
f9b2ee71 | 2852 | INIT_WORK(&transport->recv_worker, xs_udp_data_receive_workfn); |
4f8943f8 | 2853 | INIT_WORK(&transport->error_worker, xs_error_handle); |
edc1b01c TM |
2854 | INIT_DELAYED_WORK(&transport->connect_worker, xs_udp_setup_socket); |
2855 | ||
8f9d5b1a CL |
2856 | switch (addr->sa_family) { |
2857 | case AF_INET: | |
2858 | if (((struct sockaddr_in *)addr)->sin_port != htons(0)) | |
2859 | xprt_set_bound(xprt); | |
2860 | ||
9dc3b095 | 2861 | xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP); |
8f9d5b1a CL |
2862 | break; |
2863 | case AF_INET6: | |
2864 | if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0)) | |
2865 | xprt_set_bound(xprt); | |
2866 | ||
9dc3b095 | 2867 | xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP6); |
8f9d5b1a CL |
2868 | break; |
2869 | default: | |
0a68b0be BF |
2870 | ret = ERR_PTR(-EAFNOSUPPORT); |
2871 | goto out_err; | |
8f9d5b1a CL |
2872 | } |
2873 | ||
c740eff8 CL |
2874 | if (xprt_bound(xprt)) |
2875 | dprintk("RPC: set up xprt to %s (port %s) via %s\n", | |
2876 | xprt->address_strings[RPC_DISPLAY_ADDR], | |
2877 | xprt->address_strings[RPC_DISPLAY_PORT], | |
2878 | xprt->address_strings[RPC_DISPLAY_PROTO]); | |
2879 | else | |
2880 | dprintk("RPC: set up xprt to %s (autobind) via %s\n", | |
2881 | xprt->address_strings[RPC_DISPLAY_ADDR], | |
2882 | xprt->address_strings[RPC_DISPLAY_PROTO]); | |
edb267a6 | 2883 | |
bc25571e TT |
2884 | if (try_module_get(THIS_MODULE)) |
2885 | return xprt; | |
0a68b0be BF |
2886 | ret = ERR_PTR(-EINVAL); |
2887 | out_err: | |
315f3812 | 2888 | xs_xprt_free(xprt); |
0a68b0be | 2889 | return ret; |
a246b010 CL |
2890 | } |
2891 | ||
2881ae74 TM |
2892 | static const struct rpc_timeout xs_tcp_default_timeout = { |
2893 | .to_initval = 60 * HZ, | |
2894 | .to_maxval = 60 * HZ, | |
2895 | .to_retries = 2, | |
2896 | }; | |
2897 | ||
9903cd1c CL |
2898 | /** |
2899 | * xs_setup_tcp - Set up transport to use a TCP socket | |
96802a09 | 2900 | * @args: rpc transport creation arguments |
9903cd1c CL |
2901 | * |
2902 | */ | |
483066d6 | 2903 | static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args) |
a246b010 | 2904 | { |
8f9d5b1a | 2905 | struct sockaddr *addr = args->dstaddr; |
c8541ecd | 2906 | struct rpc_xprt *xprt; |
c8475461 | 2907 | struct sock_xprt *transport; |
0a68b0be | 2908 | struct rpc_xprt *ret; |
b7993ceb TM |
2909 | unsigned int max_slot_table_size = xprt_max_tcp_slot_table_entries; |
2910 | ||
2911 | if (args->flags & XPRT_CREATE_INFINITE_SLOTS) | |
2912 | max_slot_table_size = RPC_MAX_SLOT_TABLE_LIMIT; | |
a246b010 | 2913 | |
d9ba131d | 2914 | xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries, |
b7993ceb | 2915 | max_slot_table_size); |
c8541ecd CL |
2916 | if (IS_ERR(xprt)) |
2917 | return xprt; | |
c8475461 | 2918 | transport = container_of(xprt, struct sock_xprt, xprt); |
a246b010 | 2919 | |
ec739ef0 | 2920 | xprt->prot = IPPROTO_TCP; |
808012fb | 2921 | xprt->max_payload = RPC_MAX_FRAGMENT_SIZE; |
a246b010 | 2922 | |
03bf4b70 | 2923 | xprt->bind_timeout = XS_BIND_TO; |
03bf4b70 CL |
2924 | xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; |
2925 | xprt->idle_timeout = XS_IDLE_DISC_TO; | |
a246b010 | 2926 | |
262965f5 | 2927 | xprt->ops = &xs_tcp_ops; |
ba7392bb | 2928 | xprt->timeout = &xs_tcp_default_timeout; |
a246b010 | 2929 | |
3851f1cd | 2930 | xprt->max_reconnect_timeout = xprt->timeout->to_maxval; |
7196dbb0 TM |
2931 | xprt->connect_timeout = xprt->timeout->to_initval * |
2932 | (xprt->timeout->to_retries + 1); | |
3851f1cd | 2933 | |
c50b8ee0 | 2934 | INIT_WORK(&transport->recv_worker, xs_stream_data_receive_workfn); |
4f8943f8 | 2935 | INIT_WORK(&transport->error_worker, xs_error_handle); |
edc1b01c TM |
2936 | INIT_DELAYED_WORK(&transport->connect_worker, xs_tcp_setup_socket); |
2937 | ||
8f9d5b1a CL |
2938 | switch (addr->sa_family) { |
2939 | case AF_INET: | |
2940 | if (((struct sockaddr_in *)addr)->sin_port != htons(0)) | |
2941 | xprt_set_bound(xprt); | |
2942 | ||
9dc3b095 | 2943 | xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP); |
8f9d5b1a CL |
2944 | break; |
2945 | case AF_INET6: | |
2946 | if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0)) | |
2947 | xprt_set_bound(xprt); | |
2948 | ||
9dc3b095 | 2949 | xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP6); |
8f9d5b1a CL |
2950 | break; |
2951 | default: | |
0a68b0be BF |
2952 | ret = ERR_PTR(-EAFNOSUPPORT); |
2953 | goto out_err; | |
8f9d5b1a CL |
2954 | } |
2955 | ||
c740eff8 CL |
2956 | if (xprt_bound(xprt)) |
2957 | dprintk("RPC: set up xprt to %s (port %s) via %s\n", | |
2958 | xprt->address_strings[RPC_DISPLAY_ADDR], | |
2959 | xprt->address_strings[RPC_DISPLAY_PORT], | |
2960 | xprt->address_strings[RPC_DISPLAY_PROTO]); | |
2961 | else | |
2962 | dprintk("RPC: set up xprt to %s (autobind) via %s\n", | |
2963 | xprt->address_strings[RPC_DISPLAY_ADDR], | |
2964 | xprt->address_strings[RPC_DISPLAY_PROTO]); | |
2965 | ||
bc25571e TT |
2966 | if (try_module_get(THIS_MODULE)) |
2967 | return xprt; | |
0a68b0be BF |
2968 | ret = ERR_PTR(-EINVAL); |
2969 | out_err: | |
315f3812 | 2970 | xs_xprt_free(xprt); |
0a68b0be | 2971 | return ret; |
a246b010 | 2972 | } |
282b32e1 | 2973 | |
f300baba AB |
2974 | /** |
2975 | * xs_setup_bc_tcp - Set up transport to use a TCP backchannel socket | |
2976 | * @args: rpc transport creation arguments | |
2977 | * | |
2978 | */ | |
2979 | static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args) | |
2980 | { | |
2981 | struct sockaddr *addr = args->dstaddr; | |
2982 | struct rpc_xprt *xprt; | |
2983 | struct sock_xprt *transport; | |
2984 | struct svc_sock *bc_sock; | |
0a68b0be | 2985 | struct rpc_xprt *ret; |
f300baba | 2986 | |
d9ba131d TM |
2987 | xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries, |
2988 | xprt_tcp_slot_table_entries); | |
f300baba AB |
2989 | if (IS_ERR(xprt)) |
2990 | return xprt; | |
2991 | transport = container_of(xprt, struct sock_xprt, xprt); | |
2992 | ||
2993 | xprt->prot = IPPROTO_TCP; | |
f300baba AB |
2994 | xprt->max_payload = RPC_MAX_FRAGMENT_SIZE; |
2995 | xprt->timeout = &xs_tcp_default_timeout; | |
2996 | ||
2997 | /* backchannel */ | |
2998 | xprt_set_bound(xprt); | |
2999 | xprt->bind_timeout = 0; | |
f300baba AB |
3000 | xprt->reestablish_timeout = 0; |
3001 | xprt->idle_timeout = 0; | |
3002 | ||
f300baba AB |
3003 | xprt->ops = &bc_tcp_ops; |
3004 | ||
3005 | switch (addr->sa_family) { | |
3006 | case AF_INET: | |
3007 | xs_format_peer_addresses(xprt, "tcp", | |
3008 | RPCBIND_NETID_TCP); | |
3009 | break; | |
3010 | case AF_INET6: | |
3011 | xs_format_peer_addresses(xprt, "tcp", | |
3012 | RPCBIND_NETID_TCP6); | |
3013 | break; | |
3014 | default: | |
0a68b0be BF |
3015 | ret = ERR_PTR(-EAFNOSUPPORT); |
3016 | goto out_err; | |
f300baba AB |
3017 | } |
3018 | ||
50fa0d40 PE |
3019 | dprintk("RPC: set up xprt to %s (port %s) via %s\n", |
3020 | xprt->address_strings[RPC_DISPLAY_ADDR], | |
3021 | xprt->address_strings[RPC_DISPLAY_PORT], | |
3022 | xprt->address_strings[RPC_DISPLAY_PROTO]); | |
f300baba | 3023 | |
99de8ea9 BF |
3024 | /* |
3025 | * Once we've associated a backchannel xprt with a connection, | |
28303ca3 WM |
3026 | * we want to keep it around as long as the connection lasts, |
3027 | * in case we need to start using it for a backchannel again; | |
3028 | * this reference won't be dropped until bc_xprt is destroyed. | |
99de8ea9 BF |
3029 | */ |
3030 | xprt_get(xprt); | |
3031 | args->bc_xprt->xpt_bc_xprt = xprt; | |
3032 | xprt->bc_xprt = args->bc_xprt; | |
3033 | bc_sock = container_of(args->bc_xprt, struct svc_sock, sk_xprt); | |
3034 | transport->sock = bc_sock->sk_sock; | |
3035 | transport->inet = bc_sock->sk_sk; | |
3036 | ||
f300baba AB |
3037 | /* |
3038 | * Since we don't want connections for the backchannel, we set | |
3039 | * the xprt status to connected | |
3040 | */ | |
3041 | xprt_set_connected(xprt); | |
3042 | ||
f300baba AB |
3043 | if (try_module_get(THIS_MODULE)) |
3044 | return xprt; | |
642aab58 KM |
3045 | |
3046 | args->bc_xprt->xpt_bc_xprt = NULL; | |
39a9beab | 3047 | args->bc_xprt->xpt_bc_xps = NULL; |
99de8ea9 | 3048 | xprt_put(xprt); |
0a68b0be BF |
3049 | ret = ERR_PTR(-EINVAL); |
3050 | out_err: | |
315f3812 | 3051 | xs_xprt_free(xprt); |
0a68b0be | 3052 | return ret; |
f300baba AB |
3053 | } |
3054 | ||
176e21ee CL |
3055 | static struct xprt_class xs_local_transport = { |
3056 | .list = LIST_HEAD_INIT(xs_local_transport.list), | |
3057 | .name = "named UNIX socket", | |
3058 | .owner = THIS_MODULE, | |
3059 | .ident = XPRT_TRANSPORT_LOCAL, | |
3060 | .setup = xs_setup_local, | |
d5aa6b22 | 3061 | .netid = { "" }, |
176e21ee CL |
3062 | }; |
3063 | ||
bc25571e TT |
3064 | static struct xprt_class xs_udp_transport = { |
3065 | .list = LIST_HEAD_INIT(xs_udp_transport.list), | |
3066 | .name = "udp", | |
3067 | .owner = THIS_MODULE, | |
f300baba | 3068 | .ident = XPRT_TRANSPORT_UDP, |
bc25571e | 3069 | .setup = xs_setup_udp, |
d5aa6b22 | 3070 | .netid = { "udp", "udp6", "" }, |
bc25571e TT |
3071 | }; |
3072 | ||
3073 | static struct xprt_class xs_tcp_transport = { | |
3074 | .list = LIST_HEAD_INIT(xs_tcp_transport.list), | |
3075 | .name = "tcp", | |
3076 | .owner = THIS_MODULE, | |
f300baba | 3077 | .ident = XPRT_TRANSPORT_TCP, |
bc25571e | 3078 | .setup = xs_setup_tcp, |
d5aa6b22 | 3079 | .netid = { "tcp", "tcp6", "" }, |
bc25571e TT |
3080 | }; |
3081 | ||
f300baba AB |
3082 | static struct xprt_class xs_bc_tcp_transport = { |
3083 | .list = LIST_HEAD_INIT(xs_bc_tcp_transport.list), | |
3084 | .name = "tcp NFSv4.1 backchannel", | |
3085 | .owner = THIS_MODULE, | |
3086 | .ident = XPRT_TRANSPORT_BC_TCP, | |
3087 | .setup = xs_setup_bc_tcp, | |
d5aa6b22 | 3088 | .netid = { "" }, |
f300baba AB |
3089 | }; |
3090 | ||
282b32e1 | 3091 | /** |
bc25571e | 3092 | * init_socket_xprt - set up xprtsock's sysctls, register with RPC client |
282b32e1 CL |
3093 | * |
3094 | */ | |
3095 | int init_socket_xprt(void) | |
3096 | { | |
2b1bec5f | 3097 | if (!sunrpc_table_header) |
0b4d4147 | 3098 | sunrpc_table_header = register_sysctl_table(sunrpc_table); |
fbf76683 | 3099 | |
176e21ee | 3100 | xprt_register_transport(&xs_local_transport); |
bc25571e TT |
3101 | xprt_register_transport(&xs_udp_transport); |
3102 | xprt_register_transport(&xs_tcp_transport); | |
f300baba | 3103 | xprt_register_transport(&xs_bc_tcp_transport); |
bc25571e | 3104 | |
282b32e1 CL |
3105 | return 0; |
3106 | } | |
3107 | ||
3108 | /** | |
bc25571e | 3109 | * cleanup_socket_xprt - remove xprtsock's sysctls, unregister |
282b32e1 CL |
3110 | * |
3111 | */ | |
3112 | void cleanup_socket_xprt(void) | |
3113 | { | |
fbf76683 CL |
3114 | if (sunrpc_table_header) { |
3115 | unregister_sysctl_table(sunrpc_table_header); | |
3116 | sunrpc_table_header = NULL; | |
3117 | } | |
bc25571e | 3118 | |
176e21ee | 3119 | xprt_unregister_transport(&xs_local_transport); |
bc25571e TT |
3120 | xprt_unregister_transport(&xs_udp_transport); |
3121 | xprt_unregister_transport(&xs_tcp_transport); | |
f300baba | 3122 | xprt_unregister_transport(&xs_bc_tcp_transport); |
282b32e1 | 3123 | } |
cbf11071 | 3124 | |
9bbb9e5a RR |
3125 | static int param_set_uint_minmax(const char *val, |
3126 | const struct kernel_param *kp, | |
cbf11071 TM |
3127 | unsigned int min, unsigned int max) |
3128 | { | |
00cfaa94 | 3129 | unsigned int num; |
cbf11071 TM |
3130 | int ret; |
3131 | ||
3132 | if (!val) | |
3133 | return -EINVAL; | |
00cfaa94 | 3134 | ret = kstrtouint(val, 0, &num); |
9761a246 DC |
3135 | if (ret) |
3136 | return ret; | |
3137 | if (num < min || num > max) | |
cbf11071 TM |
3138 | return -EINVAL; |
3139 | *((unsigned int *)kp->arg) = num; | |
3140 | return 0; | |
3141 | } | |
3142 | ||
9bbb9e5a | 3143 | static int param_set_portnr(const char *val, const struct kernel_param *kp) |
cbf11071 | 3144 | { |
ffb6ca33 | 3145 | return param_set_uint_minmax(val, kp, |
826799e6 | 3146 | RPC_MIN_RESVPORT, |
cbf11071 TM |
3147 | RPC_MAX_RESVPORT); |
3148 | } | |
3149 | ||
9c27847d | 3150 | static const struct kernel_param_ops param_ops_portnr = { |
9bbb9e5a RR |
3151 | .set = param_set_portnr, |
3152 | .get = param_get_uint, | |
3153 | }; | |
3154 | ||
cbf11071 TM |
3155 | #define param_check_portnr(name, p) \ |
3156 | __param_check(name, p, unsigned int); | |
3157 | ||
3158 | module_param_named(min_resvport, xprt_min_resvport, portnr, 0644); | |
3159 | module_param_named(max_resvport, xprt_max_resvport, portnr, 0644); | |
3160 | ||
9bbb9e5a RR |
3161 | static int param_set_slot_table_size(const char *val, |
3162 | const struct kernel_param *kp) | |
cbf11071 TM |
3163 | { |
3164 | return param_set_uint_minmax(val, kp, | |
3165 | RPC_MIN_SLOT_TABLE, | |
3166 | RPC_MAX_SLOT_TABLE); | |
3167 | } | |
3168 | ||
9c27847d | 3169 | static const struct kernel_param_ops param_ops_slot_table_size = { |
9bbb9e5a RR |
3170 | .set = param_set_slot_table_size, |
3171 | .get = param_get_uint, | |
3172 | }; | |
3173 | ||
cbf11071 TM |
3174 | #define param_check_slot_table_size(name, p) \ |
3175 | __param_check(name, p, unsigned int); | |
3176 | ||
d9ba131d TM |
3177 | static int param_set_max_slot_table_size(const char *val, |
3178 | const struct kernel_param *kp) | |
3179 | { | |
3180 | return param_set_uint_minmax(val, kp, | |
3181 | RPC_MIN_SLOT_TABLE, | |
3182 | RPC_MAX_SLOT_TABLE_LIMIT); | |
3183 | } | |
3184 | ||
9c27847d | 3185 | static const struct kernel_param_ops param_ops_max_slot_table_size = { |
d9ba131d TM |
3186 | .set = param_set_max_slot_table_size, |
3187 | .get = param_get_uint, | |
3188 | }; | |
3189 | ||
3190 | #define param_check_max_slot_table_size(name, p) \ | |
3191 | __param_check(name, p, unsigned int); | |
3192 | ||
cbf11071 TM |
3193 | module_param_named(tcp_slot_table_entries, xprt_tcp_slot_table_entries, |
3194 | slot_table_size, 0644); | |
d9ba131d TM |
3195 | module_param_named(tcp_max_slot_table_entries, xprt_max_tcp_slot_table_entries, |
3196 | max_slot_table_size, 0644); | |
cbf11071 TM |
3197 | module_param_named(udp_slot_table_entries, xprt_udp_slot_table_entries, |
3198 | slot_table_size, 0644); |