Commit | Line | Data |
---|---|---|
a246b010 CL |
1 | /* |
2 | * linux/net/sunrpc/xprtsock.c | |
3 | * | |
4 | * Client-side transport implementation for sockets. | |
5 | * | |
113aa838 AC |
6 | * TCP callback races fixes (C) 1998 Red Hat |
7 | * TCP send fixes (C) 1998 Red Hat | |
a246b010 CL |
8 | * TCP NFS related read + write fixes |
9 | * (C) 1999 Dave Airlie, University of Limerick, Ireland <airlied@linux.ie> | |
10 | * | |
11 | * Rewrite of larges part of the code in order to stabilize TCP stuff. | |
12 | * Fix behaviour when socket buffer is full. | |
13 | * (C) 1999 Trond Myklebust <trond.myklebust@fys.uio.no> | |
55aa4f58 CL |
14 | * |
15 | * IP socket transport implementation, (C) 2005 Chuck Lever <cel@netapp.com> | |
8f9d5b1a CL |
16 | * |
17 | * IPv6 support contributed by Gilles Quillard, Bull Open Source, 2005. | |
18 | * <gilles.quillard@bull.net> | |
a246b010 CL |
19 | */ |
20 | ||
21 | #include <linux/types.h> | |
22 | #include <linux/slab.h> | |
bc25571e | 23 | #include <linux/module.h> |
a246b010 | 24 | #include <linux/capability.h> |
a246b010 CL |
25 | #include <linux/pagemap.h> |
26 | #include <linux/errno.h> | |
27 | #include <linux/socket.h> | |
28 | #include <linux/in.h> | |
29 | #include <linux/net.h> | |
30 | #include <linux/mm.h> | |
31 | #include <linux/udp.h> | |
32 | #include <linux/tcp.h> | |
33 | #include <linux/sunrpc/clnt.h> | |
02107148 | 34 | #include <linux/sunrpc/sched.h> |
49c36fcc | 35 | #include <linux/sunrpc/xprtsock.h> |
a246b010 CL |
36 | #include <linux/file.h> |
37 | ||
38 | #include <net/sock.h> | |
39 | #include <net/checksum.h> | |
40 | #include <net/udp.h> | |
41 | #include <net/tcp.h> | |
42 | ||
c556b754 CL |
43 | /* |
44 | * xprtsock tunables | |
45 | */ | |
46 | unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE; | |
47 | unsigned int xprt_tcp_slot_table_entries = RPC_DEF_SLOT_TABLE; | |
48 | ||
49 | unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT; | |
50 | unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT; | |
51 | ||
fbf76683 CL |
52 | /* |
53 | * We can register our own files under /proc/sys/sunrpc by | |
54 | * calling register_sysctl_table() again. The files in that | |
55 | * directory become the union of all files registered there. | |
56 | * | |
57 | * We simply need to make sure that we don't collide with | |
58 | * someone else's file names! | |
59 | */ | |
60 | ||
61 | #ifdef RPC_DEBUG | |
62 | ||
63 | static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE; | |
64 | static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE; | |
65 | static unsigned int xprt_min_resvport_limit = RPC_MIN_RESVPORT; | |
66 | static unsigned int xprt_max_resvport_limit = RPC_MAX_RESVPORT; | |
67 | ||
68 | static struct ctl_table_header *sunrpc_table_header; | |
69 | ||
70 | /* | |
71 | * FIXME: changing the UDP slot table size should also resize the UDP | |
72 | * socket buffers for existing UDP transports | |
73 | */ | |
74 | static ctl_table xs_tunables_table[] = { | |
75 | { | |
76 | .ctl_name = CTL_SLOTTABLE_UDP, | |
77 | .procname = "udp_slot_table_entries", | |
78 | .data = &xprt_udp_slot_table_entries, | |
79 | .maxlen = sizeof(unsigned int), | |
80 | .mode = 0644, | |
81 | .proc_handler = &proc_dointvec_minmax, | |
82 | .strategy = &sysctl_intvec, | |
83 | .extra1 = &min_slot_table_size, | |
84 | .extra2 = &max_slot_table_size | |
85 | }, | |
86 | { | |
87 | .ctl_name = CTL_SLOTTABLE_TCP, | |
88 | .procname = "tcp_slot_table_entries", | |
89 | .data = &xprt_tcp_slot_table_entries, | |
90 | .maxlen = sizeof(unsigned int), | |
91 | .mode = 0644, | |
92 | .proc_handler = &proc_dointvec_minmax, | |
93 | .strategy = &sysctl_intvec, | |
94 | .extra1 = &min_slot_table_size, | |
95 | .extra2 = &max_slot_table_size | |
96 | }, | |
97 | { | |
98 | .ctl_name = CTL_MIN_RESVPORT, | |
99 | .procname = "min_resvport", | |
100 | .data = &xprt_min_resvport, | |
101 | .maxlen = sizeof(unsigned int), | |
102 | .mode = 0644, | |
103 | .proc_handler = &proc_dointvec_minmax, | |
104 | .strategy = &sysctl_intvec, | |
105 | .extra1 = &xprt_min_resvport_limit, | |
106 | .extra2 = &xprt_max_resvport_limit | |
107 | }, | |
108 | { | |
109 | .ctl_name = CTL_MAX_RESVPORT, | |
110 | .procname = "max_resvport", | |
111 | .data = &xprt_max_resvport, | |
112 | .maxlen = sizeof(unsigned int), | |
113 | .mode = 0644, | |
114 | .proc_handler = &proc_dointvec_minmax, | |
115 | .strategy = &sysctl_intvec, | |
116 | .extra1 = &xprt_min_resvport_limit, | |
117 | .extra2 = &xprt_max_resvport_limit | |
118 | }, | |
119 | { | |
120 | .ctl_name = 0, | |
121 | }, | |
122 | }; | |
123 | ||
124 | static ctl_table sunrpc_table[] = { | |
125 | { | |
126 | .ctl_name = CTL_SUNRPC, | |
127 | .procname = "sunrpc", | |
128 | .mode = 0555, | |
129 | .child = xs_tunables_table | |
130 | }, | |
131 | { | |
132 | .ctl_name = 0, | |
133 | }, | |
134 | }; | |
135 | ||
136 | #endif | |
137 | ||
03bf4b70 CL |
138 | /* |
139 | * Time out for an RPC UDP socket connect. UDP socket connects are | |
140 | * synchronous, but we set a timeout anyway in case of resource | |
141 | * exhaustion on the local host. | |
142 | */ | |
143 | #define XS_UDP_CONN_TO (5U * HZ) | |
144 | ||
145 | /* | |
146 | * Wait duration for an RPC TCP connection to be established. Solaris | |
147 | * NFS over TCP uses 60 seconds, for example, which is in line with how | |
148 | * long a server takes to reboot. | |
149 | */ | |
150 | #define XS_TCP_CONN_TO (60U * HZ) | |
151 | ||
152 | /* | |
153 | * Wait duration for a reply from the RPC portmapper. | |
154 | */ | |
155 | #define XS_BIND_TO (60U * HZ) | |
156 | ||
157 | /* | |
158 | * Delay if a UDP socket connect error occurs. This is most likely some | |
159 | * kind of resource problem on the local host. | |
160 | */ | |
161 | #define XS_UDP_REEST_TO (2U * HZ) | |
162 | ||
163 | /* | |
164 | * The reestablish timeout allows clients to delay for a bit before attempting | |
165 | * to reconnect to a server that just dropped our connection. | |
166 | * | |
167 | * We implement an exponential backoff when trying to reestablish a TCP | |
168 | * transport connection with the server. Some servers like to drop a TCP | |
169 | * connection when they are overworked, so we start with a short timeout and | |
170 | * increase over time if the server is down or not responding. | |
171 | */ | |
172 | #define XS_TCP_INIT_REEST_TO (3U * HZ) | |
173 | #define XS_TCP_MAX_REEST_TO (5U * 60 * HZ) | |
174 | ||
175 | /* | |
176 | * TCP idle timeout; client drops the transport socket if it is idle | |
177 | * for this long. Note that we also timeout UDP sockets to prevent | |
178 | * holding port numbers when there is no RPC traffic. | |
179 | */ | |
180 | #define XS_IDLE_DISC_TO (5U * 60 * HZ) | |
181 | ||
a246b010 CL |
182 | #ifdef RPC_DEBUG |
183 | # undef RPC_DEBUG_DATA | |
9903cd1c | 184 | # define RPCDBG_FACILITY RPCDBG_TRANS |
a246b010 CL |
185 | #endif |
186 | ||
a246b010 | 187 | #ifdef RPC_DEBUG_DATA |
9903cd1c | 188 | static void xs_pktdump(char *msg, u32 *packet, unsigned int count) |
a246b010 | 189 | { |
9903cd1c CL |
190 | u8 *buf = (u8 *) packet; |
191 | int j; | |
a246b010 | 192 | |
46121cf7 | 193 | dprintk("RPC: %s\n", msg); |
a246b010 CL |
194 | for (j = 0; j < count && j < 128; j += 4) { |
195 | if (!(j & 31)) { | |
196 | if (j) | |
197 | dprintk("\n"); | |
198 | dprintk("0x%04x ", j); | |
199 | } | |
200 | dprintk("%02x%02x%02x%02x ", | |
201 | buf[j], buf[j+1], buf[j+2], buf[j+3]); | |
202 | } | |
203 | dprintk("\n"); | |
204 | } | |
205 | #else | |
9903cd1c | 206 | static inline void xs_pktdump(char *msg, u32 *packet, unsigned int count) |
a246b010 CL |
207 | { |
208 | /* NOP */ | |
209 | } | |
210 | #endif | |
211 | ||
ffc2e518 CL |
212 | struct sock_xprt { |
213 | struct rpc_xprt xprt; | |
ee0ac0c2 CL |
214 | |
215 | /* | |
216 | * Network layer | |
217 | */ | |
218 | struct socket * sock; | |
219 | struct sock * inet; | |
51971139 CL |
220 | |
221 | /* | |
222 | * State of TCP reply receive | |
223 | */ | |
224 | __be32 tcp_fraghdr, | |
225 | tcp_xid; | |
226 | ||
227 | u32 tcp_offset, | |
228 | tcp_reclen; | |
229 | ||
230 | unsigned long tcp_copied, | |
231 | tcp_flags; | |
c8475461 CL |
232 | |
233 | /* | |
234 | * Connection of transports | |
235 | */ | |
34161db6 | 236 | struct delayed_work connect_worker; |
d3bc9a1d | 237 | struct sockaddr_storage addr; |
c8475461 | 238 | unsigned short port; |
7c6e066e CL |
239 | |
240 | /* | |
241 | * UDP socket buffer size parameters | |
242 | */ | |
243 | size_t rcvsize, | |
244 | sndsize; | |
314dfd79 CL |
245 | |
246 | /* | |
247 | * Saved socket callback addresses | |
248 | */ | |
249 | void (*old_data_ready)(struct sock *, int); | |
250 | void (*old_state_change)(struct sock *); | |
251 | void (*old_write_space)(struct sock *); | |
2a9e1cfa | 252 | void (*old_error_report)(struct sock *); |
ffc2e518 CL |
253 | }; |
254 | ||
e136d092 CL |
255 | /* |
256 | * TCP receive state flags | |
257 | */ | |
258 | #define TCP_RCV_LAST_FRAG (1UL << 0) | |
259 | #define TCP_RCV_COPY_FRAGHDR (1UL << 1) | |
260 | #define TCP_RCV_COPY_XID (1UL << 2) | |
261 | #define TCP_RCV_COPY_DATA (1UL << 3) | |
262 | ||
95392c59 CL |
263 | static inline struct sockaddr *xs_addr(struct rpc_xprt *xprt) |
264 | { | |
265 | return (struct sockaddr *) &xprt->addr; | |
266 | } | |
267 | ||
268 | static inline struct sockaddr_in *xs_addr_in(struct rpc_xprt *xprt) | |
edb267a6 | 269 | { |
95392c59 CL |
270 | return (struct sockaddr_in *) &xprt->addr; |
271 | } | |
272 | ||
273 | static inline struct sockaddr_in6 *xs_addr_in6(struct rpc_xprt *xprt) | |
274 | { | |
275 | return (struct sockaddr_in6 *) &xprt->addr; | |
276 | } | |
277 | ||
b454ae90 CL |
278 | static void xs_format_ipv4_peer_addresses(struct rpc_xprt *xprt, |
279 | const char *protocol, | |
280 | const char *netid) | |
edb267a6 | 281 | { |
95392c59 | 282 | struct sockaddr_in *addr = xs_addr_in(xprt); |
edb267a6 CL |
283 | char *buf; |
284 | ||
285 | buf = kzalloc(20, GFP_KERNEL); | |
286 | if (buf) { | |
bda243df | 287 | snprintf(buf, 20, NIPQUAD_FMT, |
edb267a6 CL |
288 | NIPQUAD(addr->sin_addr.s_addr)); |
289 | } | |
290 | xprt->address_strings[RPC_DISPLAY_ADDR] = buf; | |
291 | ||
292 | buf = kzalloc(8, GFP_KERNEL); | |
293 | if (buf) { | |
294 | snprintf(buf, 8, "%u", | |
295 | ntohs(addr->sin_port)); | |
296 | } | |
297 | xprt->address_strings[RPC_DISPLAY_PORT] = buf; | |
298 | ||
b454ae90 | 299 | xprt->address_strings[RPC_DISPLAY_PROTO] = protocol; |
edb267a6 CL |
300 | |
301 | buf = kzalloc(48, GFP_KERNEL); | |
302 | if (buf) { | |
bda243df | 303 | snprintf(buf, 48, "addr="NIPQUAD_FMT" port=%u proto=%s", |
edb267a6 CL |
304 | NIPQUAD(addr->sin_addr.s_addr), |
305 | ntohs(addr->sin_port), | |
b454ae90 | 306 | protocol); |
edb267a6 CL |
307 | } |
308 | xprt->address_strings[RPC_DISPLAY_ALL] = buf; | |
fbfe3cc6 CL |
309 | |
310 | buf = kzalloc(10, GFP_KERNEL); | |
311 | if (buf) { | |
312 | snprintf(buf, 10, "%02x%02x%02x%02x", | |
313 | NIPQUAD(addr->sin_addr.s_addr)); | |
314 | } | |
315 | xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = buf; | |
316 | ||
317 | buf = kzalloc(8, GFP_KERNEL); | |
318 | if (buf) { | |
319 | snprintf(buf, 8, "%4hx", | |
320 | ntohs(addr->sin_port)); | |
321 | } | |
322 | xprt->address_strings[RPC_DISPLAY_HEX_PORT] = buf; | |
756805e7 CL |
323 | |
324 | buf = kzalloc(30, GFP_KERNEL); | |
325 | if (buf) { | |
326 | snprintf(buf, 30, NIPQUAD_FMT".%u.%u", | |
327 | NIPQUAD(addr->sin_addr.s_addr), | |
328 | ntohs(addr->sin_port) >> 8, | |
329 | ntohs(addr->sin_port) & 0xff); | |
330 | } | |
331 | xprt->address_strings[RPC_DISPLAY_UNIVERSAL_ADDR] = buf; | |
4417c8c4 | 332 | |
b454ae90 | 333 | xprt->address_strings[RPC_DISPLAY_NETID] = netid; |
edb267a6 CL |
334 | } |
335 | ||
b454ae90 CL |
336 | static void xs_format_ipv6_peer_addresses(struct rpc_xprt *xprt, |
337 | const char *protocol, | |
338 | const char *netid) | |
4b6473fb | 339 | { |
95392c59 | 340 | struct sockaddr_in6 *addr = xs_addr_in6(xprt); |
4b6473fb CL |
341 | char *buf; |
342 | ||
343 | buf = kzalloc(40, GFP_KERNEL); | |
344 | if (buf) { | |
345 | snprintf(buf, 40, NIP6_FMT, | |
346 | NIP6(addr->sin6_addr)); | |
347 | } | |
348 | xprt->address_strings[RPC_DISPLAY_ADDR] = buf; | |
349 | ||
350 | buf = kzalloc(8, GFP_KERNEL); | |
351 | if (buf) { | |
352 | snprintf(buf, 8, "%u", | |
353 | ntohs(addr->sin6_port)); | |
354 | } | |
355 | xprt->address_strings[RPC_DISPLAY_PORT] = buf; | |
356 | ||
b454ae90 | 357 | xprt->address_strings[RPC_DISPLAY_PROTO] = protocol; |
4b6473fb CL |
358 | |
359 | buf = kzalloc(64, GFP_KERNEL); | |
360 | if (buf) { | |
361 | snprintf(buf, 64, "addr="NIP6_FMT" port=%u proto=%s", | |
362 | NIP6(addr->sin6_addr), | |
363 | ntohs(addr->sin6_port), | |
b454ae90 | 364 | protocol); |
4b6473fb CL |
365 | } |
366 | xprt->address_strings[RPC_DISPLAY_ALL] = buf; | |
367 | ||
368 | buf = kzalloc(36, GFP_KERNEL); | |
369 | if (buf) { | |
370 | snprintf(buf, 36, NIP6_SEQFMT, | |
371 | NIP6(addr->sin6_addr)); | |
372 | } | |
373 | xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = buf; | |
374 | ||
375 | buf = kzalloc(8, GFP_KERNEL); | |
376 | if (buf) { | |
377 | snprintf(buf, 8, "%4hx", | |
378 | ntohs(addr->sin6_port)); | |
379 | } | |
380 | xprt->address_strings[RPC_DISPLAY_HEX_PORT] = buf; | |
756805e7 CL |
381 | |
382 | buf = kzalloc(50, GFP_KERNEL); | |
383 | if (buf) { | |
384 | snprintf(buf, 50, NIP6_FMT".%u.%u", | |
385 | NIP6(addr->sin6_addr), | |
386 | ntohs(addr->sin6_port) >> 8, | |
387 | ntohs(addr->sin6_port) & 0xff); | |
388 | } | |
389 | xprt->address_strings[RPC_DISPLAY_UNIVERSAL_ADDR] = buf; | |
4417c8c4 | 390 | |
b454ae90 | 391 | xprt->address_strings[RPC_DISPLAY_NETID] = netid; |
edb267a6 CL |
392 | } |
393 | ||
394 | static void xs_free_peer_addresses(struct rpc_xprt *xprt) | |
395 | { | |
33e01dc7 CL |
396 | unsigned int i; |
397 | ||
398 | for (i = 0; i < RPC_DISPLAY_MAX; i++) | |
399 | switch (i) { | |
400 | case RPC_DISPLAY_PROTO: | |
401 | case RPC_DISPLAY_NETID: | |
402 | continue; | |
403 | default: | |
404 | kfree(xprt->address_strings[i]); | |
405 | } | |
edb267a6 CL |
406 | } |
407 | ||
b4b5cc85 CL |
408 | #define XS_SENDMSG_FLAGS (MSG_DONTWAIT | MSG_NOSIGNAL) |
409 | ||
24c5684b | 410 | static int xs_send_kvec(struct socket *sock, struct sockaddr *addr, int addrlen, struct kvec *vec, unsigned int base, int more) |
b4b5cc85 | 411 | { |
b4b5cc85 CL |
412 | struct msghdr msg = { |
413 | .msg_name = addr, | |
414 | .msg_namelen = addrlen, | |
24c5684b TM |
415 | .msg_flags = XS_SENDMSG_FLAGS | (more ? MSG_MORE : 0), |
416 | }; | |
417 | struct kvec iov = { | |
418 | .iov_base = vec->iov_base + base, | |
419 | .iov_len = vec->iov_len - base, | |
b4b5cc85 CL |
420 | }; |
421 | ||
24c5684b | 422 | if (iov.iov_len != 0) |
b4b5cc85 CL |
423 | return kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len); |
424 | return kernel_sendmsg(sock, &msg, NULL, 0, 0); | |
425 | } | |
426 | ||
24c5684b | 427 | static int xs_send_pagedata(struct socket *sock, struct xdr_buf *xdr, unsigned int base, int more) |
b4b5cc85 | 428 | { |
24c5684b TM |
429 | struct page **ppage; |
430 | unsigned int remainder; | |
431 | int err, sent = 0; | |
432 | ||
433 | remainder = xdr->page_len - base; | |
434 | base += xdr->page_base; | |
435 | ppage = xdr->pages + (base >> PAGE_SHIFT); | |
436 | base &= ~PAGE_MASK; | |
437 | for(;;) { | |
438 | unsigned int len = min_t(unsigned int, PAGE_SIZE - base, remainder); | |
439 | int flags = XS_SENDMSG_FLAGS; | |
b4b5cc85 | 440 | |
24c5684b TM |
441 | remainder -= len; |
442 | if (remainder != 0 || more) | |
443 | flags |= MSG_MORE; | |
444 | err = sock->ops->sendpage(sock, *ppage, base, len, flags); | |
445 | if (remainder == 0 || err != len) | |
446 | break; | |
447 | sent += err; | |
448 | ppage++; | |
449 | base = 0; | |
450 | } | |
451 | if (sent == 0) | |
452 | return err; | |
453 | if (err > 0) | |
454 | sent += err; | |
455 | return sent; | |
b4b5cc85 CL |
456 | } |
457 | ||
9903cd1c CL |
458 | /** |
459 | * xs_sendpages - write pages directly to a socket | |
460 | * @sock: socket to send on | |
461 | * @addr: UDP only -- address of destination | |
462 | * @addrlen: UDP only -- length of destination address | |
463 | * @xdr: buffer containing this request | |
464 | * @base: starting position in the buffer | |
465 | * | |
a246b010 | 466 | */ |
24c5684b | 467 | static int xs_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen, struct xdr_buf *xdr, unsigned int base) |
a246b010 | 468 | { |
24c5684b TM |
469 | unsigned int remainder = xdr->len - base; |
470 | int err, sent = 0; | |
a246b010 | 471 | |
262965f5 CL |
472 | if (unlikely(!sock)) |
473 | return -ENOTCONN; | |
474 | ||
475 | clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags); | |
24c5684b TM |
476 | if (base != 0) { |
477 | addr = NULL; | |
478 | addrlen = 0; | |
479 | } | |
262965f5 | 480 | |
24c5684b TM |
481 | if (base < xdr->head[0].iov_len || addr != NULL) { |
482 | unsigned int len = xdr->head[0].iov_len - base; | |
483 | remainder -= len; | |
484 | err = xs_send_kvec(sock, addr, addrlen, &xdr->head[0], base, remainder != 0); | |
485 | if (remainder == 0 || err != len) | |
a246b010 | 486 | goto out; |
24c5684b | 487 | sent += err; |
a246b010 CL |
488 | base = 0; |
489 | } else | |
24c5684b | 490 | base -= xdr->head[0].iov_len; |
a246b010 | 491 | |
24c5684b TM |
492 | if (base < xdr->page_len) { |
493 | unsigned int len = xdr->page_len - base; | |
494 | remainder -= len; | |
495 | err = xs_send_pagedata(sock, xdr, base, remainder != 0); | |
496 | if (remainder == 0 || err != len) | |
a246b010 | 497 | goto out; |
24c5684b | 498 | sent += err; |
a246b010 | 499 | base = 0; |
24c5684b TM |
500 | } else |
501 | base -= xdr->page_len; | |
502 | ||
503 | if (base >= xdr->tail[0].iov_len) | |
504 | return sent; | |
505 | err = xs_send_kvec(sock, NULL, 0, &xdr->tail[0], base, 0); | |
a246b010 | 506 | out: |
24c5684b TM |
507 | if (sent == 0) |
508 | return err; | |
509 | if (err > 0) | |
510 | sent += err; | |
511 | return sent; | |
a246b010 CL |
512 | } |
513 | ||
b6ddf64f TM |
514 | static void xs_nospace_callback(struct rpc_task *task) |
515 | { | |
516 | struct sock_xprt *transport = container_of(task->tk_rqstp->rq_xprt, struct sock_xprt, xprt); | |
517 | ||
518 | transport->inet->sk_write_pending--; | |
519 | clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); | |
520 | } | |
521 | ||
9903cd1c | 522 | /** |
262965f5 CL |
523 | * xs_nospace - place task on wait queue if transmit was incomplete |
524 | * @task: task to put to sleep | |
9903cd1c | 525 | * |
a246b010 | 526 | */ |
262965f5 | 527 | static void xs_nospace(struct rpc_task *task) |
a246b010 | 528 | { |
262965f5 CL |
529 | struct rpc_rqst *req = task->tk_rqstp; |
530 | struct rpc_xprt *xprt = req->rq_xprt; | |
ee0ac0c2 | 531 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
a246b010 | 532 | |
46121cf7 | 533 | dprintk("RPC: %5u xmit incomplete (%u left of %u)\n", |
262965f5 CL |
534 | task->tk_pid, req->rq_slen - req->rq_bytes_sent, |
535 | req->rq_slen); | |
536 | ||
b6ddf64f TM |
537 | /* Protect against races with write_space */ |
538 | spin_lock_bh(&xprt->transport_lock); | |
539 | ||
540 | /* Don't race with disconnect */ | |
541 | if (xprt_connected(xprt)) { | |
542 | if (test_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags)) { | |
543 | /* | |
544 | * Notify TCP that we're limited by the application | |
545 | * window size | |
546 | */ | |
547 | set_bit(SOCK_NOSPACE, &transport->sock->flags); | |
548 | transport->inet->sk_write_pending++; | |
549 | /* ...and wait for more buffer space */ | |
550 | xprt_wait_for_buffer_space(task, xs_nospace_callback); | |
551 | } | |
552 | } else { | |
553 | clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); | |
554 | task->tk_status = -ENOTCONN; | |
555 | } | |
262965f5 | 556 | |
b6ddf64f | 557 | spin_unlock_bh(&xprt->transport_lock); |
262965f5 CL |
558 | } |
559 | ||
560 | /** | |
561 | * xs_udp_send_request - write an RPC request to a UDP socket | |
562 | * @task: address of RPC task that manages the state of an RPC request | |
563 | * | |
564 | * Return values: | |
565 | * 0: The request has been sent | |
566 | * EAGAIN: The socket was blocked, please call again later to | |
567 | * complete the request | |
568 | * ENOTCONN: Caller needs to invoke connect logic then call again | |
569 | * other: Some other error occured, the request was not sent | |
570 | */ | |
571 | static int xs_udp_send_request(struct rpc_task *task) | |
572 | { | |
573 | struct rpc_rqst *req = task->tk_rqstp; | |
574 | struct rpc_xprt *xprt = req->rq_xprt; | |
ee0ac0c2 | 575 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
262965f5 CL |
576 | struct xdr_buf *xdr = &req->rq_snd_buf; |
577 | int status; | |
a246b010 | 578 | |
9903cd1c | 579 | xs_pktdump("packet data:", |
a246b010 CL |
580 | req->rq_svec->iov_base, |
581 | req->rq_svec->iov_len); | |
582 | ||
ee0ac0c2 | 583 | status = xs_sendpages(transport->sock, |
95392c59 | 584 | xs_addr(xprt), |
ee0ac0c2 CL |
585 | xprt->addrlen, xdr, |
586 | req->rq_bytes_sent); | |
a246b010 | 587 | |
46121cf7 | 588 | dprintk("RPC: xs_udp_send_request(%u) = %d\n", |
262965f5 | 589 | xdr->len - req->rq_bytes_sent, status); |
a246b010 | 590 | |
2199700f TM |
591 | if (status >= 0) { |
592 | task->tk_bytes_sent += status; | |
593 | if (status >= req->rq_slen) | |
594 | return 0; | |
595 | /* Still some bytes left; set up for a retry later. */ | |
262965f5 | 596 | status = -EAGAIN; |
2199700f | 597 | } |
a246b010 | 598 | |
262965f5 | 599 | switch (status) { |
b6ddf64f TM |
600 | case -EAGAIN: |
601 | xs_nospace(task); | |
602 | break; | |
262965f5 CL |
603 | case -ENETUNREACH: |
604 | case -EPIPE: | |
a246b010 CL |
605 | case -ECONNREFUSED: |
606 | /* When the server has died, an ICMP port unreachable message | |
9903cd1c | 607 | * prompts ECONNREFUSED. */ |
b6ddf64f | 608 | clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); |
a246b010 CL |
609 | break; |
610 | default: | |
b6ddf64f | 611 | clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); |
46121cf7 | 612 | dprintk("RPC: sendmsg returned unrecognized error %d\n", |
262965f5 | 613 | -status); |
a246b010 | 614 | } |
262965f5 CL |
615 | |
616 | return status; | |
a246b010 CL |
617 | } |
618 | ||
e06799f9 TM |
619 | /** |
620 | * xs_tcp_shutdown - gracefully shut down a TCP socket | |
621 | * @xprt: transport | |
622 | * | |
623 | * Initiates a graceful shutdown of the TCP socket by calling the | |
624 | * equivalent of shutdown(SHUT_WR); | |
625 | */ | |
626 | static void xs_tcp_shutdown(struct rpc_xprt *xprt) | |
627 | { | |
628 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); | |
629 | struct socket *sock = transport->sock; | |
630 | ||
631 | if (sock != NULL) | |
632 | kernel_sock_shutdown(sock, SHUT_WR); | |
633 | } | |
634 | ||
808012fb CL |
635 | static inline void xs_encode_tcp_record_marker(struct xdr_buf *buf) |
636 | { | |
637 | u32 reclen = buf->len - sizeof(rpc_fraghdr); | |
638 | rpc_fraghdr *base = buf->head[0].iov_base; | |
639 | *base = htonl(RPC_LAST_STREAM_FRAGMENT | reclen); | |
640 | } | |
641 | ||
9903cd1c | 642 | /** |
262965f5 | 643 | * xs_tcp_send_request - write an RPC request to a TCP socket |
9903cd1c CL |
644 | * @task: address of RPC task that manages the state of an RPC request |
645 | * | |
646 | * Return values: | |
262965f5 CL |
647 | * 0: The request has been sent |
648 | * EAGAIN: The socket was blocked, please call again later to | |
649 | * complete the request | |
650 | * ENOTCONN: Caller needs to invoke connect logic then call again | |
651 | * other: Some other error occured, the request was not sent | |
9903cd1c CL |
652 | * |
653 | * XXX: In the case of soft timeouts, should we eventually give up | |
262965f5 | 654 | * if sendmsg is not able to make progress? |
9903cd1c | 655 | */ |
262965f5 | 656 | static int xs_tcp_send_request(struct rpc_task *task) |
a246b010 CL |
657 | { |
658 | struct rpc_rqst *req = task->tk_rqstp; | |
659 | struct rpc_xprt *xprt = req->rq_xprt; | |
ee0ac0c2 | 660 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
262965f5 | 661 | struct xdr_buf *xdr = &req->rq_snd_buf; |
b595bb15 | 662 | int status; |
a246b010 | 663 | |
808012fb | 664 | xs_encode_tcp_record_marker(&req->rq_snd_buf); |
a246b010 | 665 | |
262965f5 CL |
666 | xs_pktdump("packet data:", |
667 | req->rq_svec->iov_base, | |
668 | req->rq_svec->iov_len); | |
a246b010 CL |
669 | |
670 | /* Continue transmitting the packet/record. We must be careful | |
671 | * to cope with writespace callbacks arriving _after_ we have | |
262965f5 | 672 | * called sendmsg(). */ |
a246b010 | 673 | while (1) { |
ee0ac0c2 CL |
674 | status = xs_sendpages(transport->sock, |
675 | NULL, 0, xdr, req->rq_bytes_sent); | |
a246b010 | 676 | |
46121cf7 | 677 | dprintk("RPC: xs_tcp_send_request(%u) = %d\n", |
262965f5 | 678 | xdr->len - req->rq_bytes_sent, status); |
a246b010 | 679 | |
262965f5 | 680 | if (unlikely(status < 0)) |
a246b010 | 681 | break; |
a246b010 | 682 | |
262965f5 CL |
683 | /* If we've sent the entire packet, immediately |
684 | * reset the count of bytes sent. */ | |
685 | req->rq_bytes_sent += status; | |
ef759a2e | 686 | task->tk_bytes_sent += status; |
262965f5 CL |
687 | if (likely(req->rq_bytes_sent >= req->rq_slen)) { |
688 | req->rq_bytes_sent = 0; | |
689 | return 0; | |
690 | } | |
a246b010 | 691 | |
06b4b681 TM |
692 | if (status != 0) |
693 | continue; | |
a246b010 | 694 | status = -EAGAIN; |
06b4b681 | 695 | break; |
a246b010 CL |
696 | } |
697 | ||
262965f5 CL |
698 | switch (status) { |
699 | case -EAGAIN: | |
700 | xs_nospace(task); | |
701 | break; | |
262965f5 | 702 | case -ECONNRESET: |
2a9e1cfa TM |
703 | xs_tcp_shutdown(xprt); |
704 | case -ECONNREFUSED: | |
262965f5 CL |
705 | case -ENOTCONN: |
706 | case -EPIPE: | |
707 | status = -ENOTCONN; | |
b6ddf64f | 708 | clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); |
262965f5 CL |
709 | break; |
710 | default: | |
46121cf7 | 711 | dprintk("RPC: sendmsg returned unrecognized error %d\n", |
262965f5 | 712 | -status); |
b6ddf64f | 713 | clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); |
e06799f9 | 714 | xs_tcp_shutdown(xprt); |
a246b010 | 715 | } |
262965f5 | 716 | |
a246b010 CL |
717 | return status; |
718 | } | |
719 | ||
e0ab53de TM |
720 | /** |
721 | * xs_tcp_release_xprt - clean up after a tcp transmission | |
722 | * @xprt: transport | |
723 | * @task: rpc task | |
724 | * | |
725 | * This cleans up if an error causes us to abort the transmission of a request. | |
726 | * In this case, the socket may need to be reset in order to avoid confusing | |
727 | * the server. | |
728 | */ | |
729 | static void xs_tcp_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task) | |
730 | { | |
731 | struct rpc_rqst *req; | |
732 | ||
733 | if (task != xprt->snd_task) | |
734 | return; | |
735 | if (task == NULL) | |
736 | goto out_release; | |
737 | req = task->tk_rqstp; | |
738 | if (req->rq_bytes_sent == 0) | |
739 | goto out_release; | |
740 | if (req->rq_bytes_sent == req->rq_snd_buf.len) | |
741 | goto out_release; | |
742 | set_bit(XPRT_CLOSE_WAIT, &task->tk_xprt->state); | |
743 | out_release: | |
744 | xprt_release_xprt(xprt, task); | |
745 | } | |
746 | ||
2a9e1cfa TM |
747 | static void xs_save_old_callbacks(struct sock_xprt *transport, struct sock *sk) |
748 | { | |
749 | transport->old_data_ready = sk->sk_data_ready; | |
750 | transport->old_state_change = sk->sk_state_change; | |
751 | transport->old_write_space = sk->sk_write_space; | |
752 | transport->old_error_report = sk->sk_error_report; | |
753 | } | |
754 | ||
755 | static void xs_restore_old_callbacks(struct sock_xprt *transport, struct sock *sk) | |
756 | { | |
757 | sk->sk_data_ready = transport->old_data_ready; | |
758 | sk->sk_state_change = transport->old_state_change; | |
759 | sk->sk_write_space = transport->old_write_space; | |
760 | sk->sk_error_report = transport->old_error_report; | |
761 | } | |
762 | ||
9903cd1c CL |
763 | /** |
764 | * xs_close - close a socket | |
765 | * @xprt: transport | |
766 | * | |
3167e12c CL |
767 | * This is used when all requests are complete; ie, no DRC state remains |
768 | * on the server we want to save. | |
a246b010 | 769 | */ |
9903cd1c | 770 | static void xs_close(struct rpc_xprt *xprt) |
a246b010 | 771 | { |
ee0ac0c2 CL |
772 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
773 | struct socket *sock = transport->sock; | |
774 | struct sock *sk = transport->inet; | |
a246b010 CL |
775 | |
776 | if (!sk) | |
632e3bdc | 777 | goto clear_close_wait; |
a246b010 | 778 | |
46121cf7 | 779 | dprintk("RPC: xs_close xprt %p\n", xprt); |
9903cd1c | 780 | |
a246b010 | 781 | write_lock_bh(&sk->sk_callback_lock); |
ee0ac0c2 CL |
782 | transport->inet = NULL; |
783 | transport->sock = NULL; | |
a246b010 | 784 | |
9903cd1c | 785 | sk->sk_user_data = NULL; |
2a9e1cfa TM |
786 | |
787 | xs_restore_old_callbacks(transport, sk); | |
a246b010 CL |
788 | write_unlock_bh(&sk->sk_callback_lock); |
789 | ||
9903cd1c | 790 | sk->sk_no_check = 0; |
a246b010 CL |
791 | |
792 | sock_release(sock); | |
632e3bdc TM |
793 | clear_close_wait: |
794 | smp_mb__before_clear_bit(); | |
795 | clear_bit(XPRT_CLOSE_WAIT, &xprt->state); | |
3b948ae5 | 796 | clear_bit(XPRT_CLOSING, &xprt->state); |
632e3bdc | 797 | smp_mb__after_clear_bit(); |
62da3b24 | 798 | xprt_disconnect_done(xprt); |
a246b010 CL |
799 | } |
800 | ||
9903cd1c CL |
801 | /** |
802 | * xs_destroy - prepare to shutdown a transport | |
803 | * @xprt: doomed transport | |
804 | * | |
805 | */ | |
806 | static void xs_destroy(struct rpc_xprt *xprt) | |
a246b010 | 807 | { |
c8475461 CL |
808 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
809 | ||
46121cf7 | 810 | dprintk("RPC: xs_destroy xprt %p\n", xprt); |
9903cd1c | 811 | |
c1384c9c | 812 | cancel_rearming_delayed_work(&transport->connect_worker); |
a246b010 | 813 | |
9903cd1c | 814 | xs_close(xprt); |
edb267a6 | 815 | xs_free_peer_addresses(xprt); |
a246b010 | 816 | kfree(xprt->slot); |
c8541ecd | 817 | kfree(xprt); |
bc25571e | 818 | module_put(THIS_MODULE); |
a246b010 CL |
819 | } |
820 | ||
9903cd1c CL |
821 | static inline struct rpc_xprt *xprt_from_sock(struct sock *sk) |
822 | { | |
823 | return (struct rpc_xprt *) sk->sk_user_data; | |
824 | } | |
825 | ||
826 | /** | |
827 | * xs_udp_data_ready - "data ready" callback for UDP sockets | |
828 | * @sk: socket with data to read | |
829 | * @len: how much data to read | |
830 | * | |
a246b010 | 831 | */ |
9903cd1c | 832 | static void xs_udp_data_ready(struct sock *sk, int len) |
a246b010 | 833 | { |
9903cd1c CL |
834 | struct rpc_task *task; |
835 | struct rpc_xprt *xprt; | |
a246b010 | 836 | struct rpc_rqst *rovr; |
9903cd1c | 837 | struct sk_buff *skb; |
a246b010 | 838 | int err, repsize, copied; |
d8ed029d AD |
839 | u32 _xid; |
840 | __be32 *xp; | |
a246b010 CL |
841 | |
842 | read_lock(&sk->sk_callback_lock); | |
46121cf7 | 843 | dprintk("RPC: xs_udp_data_ready...\n"); |
9903cd1c | 844 | if (!(xprt = xprt_from_sock(sk))) |
a246b010 | 845 | goto out; |
a246b010 CL |
846 | |
847 | if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL) | |
848 | goto out; | |
849 | ||
850 | if (xprt->shutdown) | |
851 | goto dropit; | |
852 | ||
853 | repsize = skb->len - sizeof(struct udphdr); | |
854 | if (repsize < 4) { | |
46121cf7 | 855 | dprintk("RPC: impossible RPC reply size %d!\n", repsize); |
a246b010 CL |
856 | goto dropit; |
857 | } | |
858 | ||
859 | /* Copy the XID from the skb... */ | |
860 | xp = skb_header_pointer(skb, sizeof(struct udphdr), | |
861 | sizeof(_xid), &_xid); | |
862 | if (xp == NULL) | |
863 | goto dropit; | |
864 | ||
865 | /* Look up and lock the request corresponding to the given XID */ | |
4a0f8c04 | 866 | spin_lock(&xprt->transport_lock); |
a246b010 CL |
867 | rovr = xprt_lookup_rqst(xprt, *xp); |
868 | if (!rovr) | |
869 | goto out_unlock; | |
870 | task = rovr->rq_task; | |
871 | ||
a246b010 CL |
872 | if ((copied = rovr->rq_private_buf.buflen) > repsize) |
873 | copied = repsize; | |
874 | ||
875 | /* Suck it into the iovec, verify checksum if not done by hw. */ | |
1781f7f5 HX |
876 | if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb)) { |
877 | UDPX_INC_STATS_BH(sk, UDP_MIB_INERRORS); | |
a246b010 | 878 | goto out_unlock; |
1781f7f5 HX |
879 | } |
880 | ||
881 | UDPX_INC_STATS_BH(sk, UDP_MIB_INDATAGRAMS); | |
a246b010 CL |
882 | |
883 | /* Something worked... */ | |
884 | dst_confirm(skb->dst); | |
885 | ||
1570c1e4 CL |
886 | xprt_adjust_cwnd(task, copied); |
887 | xprt_update_rtt(task); | |
888 | xprt_complete_rqst(task, copied); | |
a246b010 CL |
889 | |
890 | out_unlock: | |
4a0f8c04 | 891 | spin_unlock(&xprt->transport_lock); |
a246b010 CL |
892 | dropit: |
893 | skb_free_datagram(sk, skb); | |
894 | out: | |
895 | read_unlock(&sk->sk_callback_lock); | |
896 | } | |
897 | ||
dd456471 | 898 | static inline void xs_tcp_read_fraghdr(struct rpc_xprt *xprt, struct xdr_skb_reader *desc) |
a246b010 | 899 | { |
51971139 | 900 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
a246b010 CL |
901 | size_t len, used; |
902 | char *p; | |
903 | ||
51971139 CL |
904 | p = ((char *) &transport->tcp_fraghdr) + transport->tcp_offset; |
905 | len = sizeof(transport->tcp_fraghdr) - transport->tcp_offset; | |
9d292316 | 906 | used = xdr_skb_read_bits(desc, p, len); |
51971139 | 907 | transport->tcp_offset += used; |
a246b010 CL |
908 | if (used != len) |
909 | return; | |
808012fb | 910 | |
51971139 CL |
911 | transport->tcp_reclen = ntohl(transport->tcp_fraghdr); |
912 | if (transport->tcp_reclen & RPC_LAST_STREAM_FRAGMENT) | |
e136d092 | 913 | transport->tcp_flags |= TCP_RCV_LAST_FRAG; |
a246b010 | 914 | else |
e136d092 | 915 | transport->tcp_flags &= ~TCP_RCV_LAST_FRAG; |
51971139 | 916 | transport->tcp_reclen &= RPC_FRAGMENT_SIZE_MASK; |
808012fb | 917 | |
e136d092 | 918 | transport->tcp_flags &= ~TCP_RCV_COPY_FRAGHDR; |
51971139 | 919 | transport->tcp_offset = 0; |
808012fb | 920 | |
a246b010 | 921 | /* Sanity check of the record length */ |
51971139 | 922 | if (unlikely(transport->tcp_reclen < 4)) { |
46121cf7 | 923 | dprintk("RPC: invalid TCP record fragment length\n"); |
3ebb067d | 924 | xprt_force_disconnect(xprt); |
9903cd1c | 925 | return; |
a246b010 | 926 | } |
46121cf7 | 927 | dprintk("RPC: reading TCP record fragment of length %d\n", |
51971139 | 928 | transport->tcp_reclen); |
a246b010 CL |
929 | } |
930 | ||
51971139 | 931 | static void xs_tcp_check_fraghdr(struct sock_xprt *transport) |
a246b010 | 932 | { |
51971139 | 933 | if (transport->tcp_offset == transport->tcp_reclen) { |
e136d092 | 934 | transport->tcp_flags |= TCP_RCV_COPY_FRAGHDR; |
51971139 | 935 | transport->tcp_offset = 0; |
e136d092 CL |
936 | if (transport->tcp_flags & TCP_RCV_LAST_FRAG) { |
937 | transport->tcp_flags &= ~TCP_RCV_COPY_DATA; | |
938 | transport->tcp_flags |= TCP_RCV_COPY_XID; | |
51971139 | 939 | transport->tcp_copied = 0; |
a246b010 CL |
940 | } |
941 | } | |
942 | } | |
943 | ||
dd456471 | 944 | static inline void xs_tcp_read_xid(struct sock_xprt *transport, struct xdr_skb_reader *desc) |
a246b010 CL |
945 | { |
946 | size_t len, used; | |
947 | char *p; | |
948 | ||
51971139 | 949 | len = sizeof(transport->tcp_xid) - transport->tcp_offset; |
46121cf7 | 950 | dprintk("RPC: reading XID (%Zu bytes)\n", len); |
51971139 | 951 | p = ((char *) &transport->tcp_xid) + transport->tcp_offset; |
9d292316 | 952 | used = xdr_skb_read_bits(desc, p, len); |
51971139 | 953 | transport->tcp_offset += used; |
a246b010 CL |
954 | if (used != len) |
955 | return; | |
e136d092 CL |
956 | transport->tcp_flags &= ~TCP_RCV_COPY_XID; |
957 | transport->tcp_flags |= TCP_RCV_COPY_DATA; | |
51971139 | 958 | transport->tcp_copied = 4; |
46121cf7 | 959 | dprintk("RPC: reading reply for XID %08x\n", |
51971139 CL |
960 | ntohl(transport->tcp_xid)); |
961 | xs_tcp_check_fraghdr(transport); | |
a246b010 CL |
962 | } |
963 | ||
dd456471 | 964 | static inline void xs_tcp_read_request(struct rpc_xprt *xprt, struct xdr_skb_reader *desc) |
a246b010 | 965 | { |
51971139 | 966 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
a246b010 CL |
967 | struct rpc_rqst *req; |
968 | struct xdr_buf *rcvbuf; | |
969 | size_t len; | |
970 | ssize_t r; | |
971 | ||
972 | /* Find and lock the request corresponding to this xid */ | |
4a0f8c04 | 973 | spin_lock(&xprt->transport_lock); |
51971139 | 974 | req = xprt_lookup_rqst(xprt, transport->tcp_xid); |
a246b010 | 975 | if (!req) { |
e136d092 | 976 | transport->tcp_flags &= ~TCP_RCV_COPY_DATA; |
46121cf7 | 977 | dprintk("RPC: XID %08x request not found!\n", |
51971139 | 978 | ntohl(transport->tcp_xid)); |
4a0f8c04 | 979 | spin_unlock(&xprt->transport_lock); |
a246b010 CL |
980 | return; |
981 | } | |
982 | ||
983 | rcvbuf = &req->rq_private_buf; | |
984 | len = desc->count; | |
51971139 | 985 | if (len > transport->tcp_reclen - transport->tcp_offset) { |
dd456471 | 986 | struct xdr_skb_reader my_desc; |
a246b010 | 987 | |
51971139 | 988 | len = transport->tcp_reclen - transport->tcp_offset; |
a246b010 CL |
989 | memcpy(&my_desc, desc, sizeof(my_desc)); |
990 | my_desc.count = len; | |
51971139 | 991 | r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied, |
9d292316 | 992 | &my_desc, xdr_skb_read_bits); |
a246b010 CL |
993 | desc->count -= r; |
994 | desc->offset += r; | |
995 | } else | |
51971139 | 996 | r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied, |
9d292316 | 997 | desc, xdr_skb_read_bits); |
a246b010 CL |
998 | |
999 | if (r > 0) { | |
51971139 CL |
1000 | transport->tcp_copied += r; |
1001 | transport->tcp_offset += r; | |
a246b010 CL |
1002 | } |
1003 | if (r != len) { | |
1004 | /* Error when copying to the receive buffer, | |
1005 | * usually because we weren't able to allocate | |
1006 | * additional buffer pages. All we can do now | |
e136d092 | 1007 | * is turn off TCP_RCV_COPY_DATA, so the request |
a246b010 CL |
1008 | * will not receive any additional updates, |
1009 | * and time out. | |
1010 | * Any remaining data from this record will | |
1011 | * be discarded. | |
1012 | */ | |
e136d092 | 1013 | transport->tcp_flags &= ~TCP_RCV_COPY_DATA; |
46121cf7 | 1014 | dprintk("RPC: XID %08x truncated request\n", |
51971139 | 1015 | ntohl(transport->tcp_xid)); |
46121cf7 CL |
1016 | dprintk("RPC: xprt = %p, tcp_copied = %lu, " |
1017 | "tcp_offset = %u, tcp_reclen = %u\n", | |
1018 | xprt, transport->tcp_copied, | |
1019 | transport->tcp_offset, transport->tcp_reclen); | |
a246b010 CL |
1020 | goto out; |
1021 | } | |
1022 | ||
46121cf7 | 1023 | dprintk("RPC: XID %08x read %Zd bytes\n", |
51971139 | 1024 | ntohl(transport->tcp_xid), r); |
46121cf7 CL |
1025 | dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, " |
1026 | "tcp_reclen = %u\n", xprt, transport->tcp_copied, | |
1027 | transport->tcp_offset, transport->tcp_reclen); | |
51971139 CL |
1028 | |
1029 | if (transport->tcp_copied == req->rq_private_buf.buflen) | |
e136d092 | 1030 | transport->tcp_flags &= ~TCP_RCV_COPY_DATA; |
51971139 | 1031 | else if (transport->tcp_offset == transport->tcp_reclen) { |
e136d092 CL |
1032 | if (transport->tcp_flags & TCP_RCV_LAST_FRAG) |
1033 | transport->tcp_flags &= ~TCP_RCV_COPY_DATA; | |
a246b010 CL |
1034 | } |
1035 | ||
1036 | out: | |
e136d092 | 1037 | if (!(transport->tcp_flags & TCP_RCV_COPY_DATA)) |
51971139 | 1038 | xprt_complete_rqst(req->rq_task, transport->tcp_copied); |
4a0f8c04 | 1039 | spin_unlock(&xprt->transport_lock); |
51971139 | 1040 | xs_tcp_check_fraghdr(transport); |
a246b010 CL |
1041 | } |
1042 | ||
dd456471 | 1043 | static inline void xs_tcp_read_discard(struct sock_xprt *transport, struct xdr_skb_reader *desc) |
a246b010 CL |
1044 | { |
1045 | size_t len; | |
1046 | ||
51971139 | 1047 | len = transport->tcp_reclen - transport->tcp_offset; |
a246b010 CL |
1048 | if (len > desc->count) |
1049 | len = desc->count; | |
1050 | desc->count -= len; | |
1051 | desc->offset += len; | |
51971139 | 1052 | transport->tcp_offset += len; |
46121cf7 | 1053 | dprintk("RPC: discarded %Zu bytes\n", len); |
51971139 | 1054 | xs_tcp_check_fraghdr(transport); |
a246b010 CL |
1055 | } |
1056 | ||
9903cd1c | 1057 | static int xs_tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb, unsigned int offset, size_t len) |
a246b010 CL |
1058 | { |
1059 | struct rpc_xprt *xprt = rd_desc->arg.data; | |
51971139 | 1060 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
dd456471 | 1061 | struct xdr_skb_reader desc = { |
a246b010 CL |
1062 | .skb = skb, |
1063 | .offset = offset, | |
1064 | .count = len, | |
9903cd1c | 1065 | }; |
a246b010 | 1066 | |
46121cf7 | 1067 | dprintk("RPC: xs_tcp_data_recv started\n"); |
a246b010 CL |
1068 | do { |
1069 | /* Read in a new fragment marker if necessary */ | |
1070 | /* Can we ever really expect to get completely empty fragments? */ | |
e136d092 | 1071 | if (transport->tcp_flags & TCP_RCV_COPY_FRAGHDR) { |
9903cd1c | 1072 | xs_tcp_read_fraghdr(xprt, &desc); |
a246b010 CL |
1073 | continue; |
1074 | } | |
1075 | /* Read in the xid if necessary */ | |
e136d092 | 1076 | if (transport->tcp_flags & TCP_RCV_COPY_XID) { |
51971139 | 1077 | xs_tcp_read_xid(transport, &desc); |
a246b010 CL |
1078 | continue; |
1079 | } | |
1080 | /* Read in the request data */ | |
e136d092 | 1081 | if (transport->tcp_flags & TCP_RCV_COPY_DATA) { |
9903cd1c | 1082 | xs_tcp_read_request(xprt, &desc); |
a246b010 CL |
1083 | continue; |
1084 | } | |
1085 | /* Skip over any trailing bytes on short reads */ | |
51971139 | 1086 | xs_tcp_read_discard(transport, &desc); |
a246b010 | 1087 | } while (desc.count); |
46121cf7 | 1088 | dprintk("RPC: xs_tcp_data_recv done\n"); |
a246b010 CL |
1089 | return len - desc.count; |
1090 | } | |
1091 | ||
9903cd1c CL |
1092 | /** |
1093 | * xs_tcp_data_ready - "data ready" callback for TCP sockets | |
1094 | * @sk: socket with data to read | |
1095 | * @bytes: how much data to read | |
1096 | * | |
1097 | */ | |
1098 | static void xs_tcp_data_ready(struct sock *sk, int bytes) | |
a246b010 CL |
1099 | { |
1100 | struct rpc_xprt *xprt; | |
1101 | read_descriptor_t rd_desc; | |
ff2d7db8 | 1102 | int read; |
a246b010 | 1103 | |
46121cf7 CL |
1104 | dprintk("RPC: xs_tcp_data_ready...\n"); |
1105 | ||
a246b010 | 1106 | read_lock(&sk->sk_callback_lock); |
9903cd1c | 1107 | if (!(xprt = xprt_from_sock(sk))) |
a246b010 | 1108 | goto out; |
a246b010 CL |
1109 | if (xprt->shutdown) |
1110 | goto out; | |
1111 | ||
9903cd1c | 1112 | /* We use rd_desc to pass struct xprt to xs_tcp_data_recv */ |
a246b010 | 1113 | rd_desc.arg.data = xprt; |
ff2d7db8 TM |
1114 | do { |
1115 | rd_desc.count = 65536; | |
1116 | read = tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv); | |
1117 | } while (read > 0); | |
a246b010 CL |
1118 | out: |
1119 | read_unlock(&sk->sk_callback_lock); | |
1120 | } | |
1121 | ||
9903cd1c CL |
1122 | /** |
1123 | * xs_tcp_state_change - callback to handle TCP socket state changes | |
1124 | * @sk: socket whose state has changed | |
1125 | * | |
1126 | */ | |
1127 | static void xs_tcp_state_change(struct sock *sk) | |
a246b010 | 1128 | { |
9903cd1c | 1129 | struct rpc_xprt *xprt; |
a246b010 CL |
1130 | |
1131 | read_lock(&sk->sk_callback_lock); | |
1132 | if (!(xprt = xprt_from_sock(sk))) | |
1133 | goto out; | |
46121cf7 CL |
1134 | dprintk("RPC: xs_tcp_state_change client %p...\n", xprt); |
1135 | dprintk("RPC: state %x conn %d dead %d zapped %d\n", | |
1136 | sk->sk_state, xprt_connected(xprt), | |
1137 | sock_flag(sk, SOCK_DEAD), | |
1138 | sock_flag(sk, SOCK_ZAPPED)); | |
a246b010 CL |
1139 | |
1140 | switch (sk->sk_state) { | |
1141 | case TCP_ESTABLISHED: | |
4a0f8c04 | 1142 | spin_lock_bh(&xprt->transport_lock); |
a246b010 | 1143 | if (!xprt_test_and_set_connected(xprt)) { |
51971139 CL |
1144 | struct sock_xprt *transport = container_of(xprt, |
1145 | struct sock_xprt, xprt); | |
1146 | ||
a246b010 | 1147 | /* Reset TCP record info */ |
51971139 CL |
1148 | transport->tcp_offset = 0; |
1149 | transport->tcp_reclen = 0; | |
1150 | transport->tcp_copied = 0; | |
e136d092 CL |
1151 | transport->tcp_flags = |
1152 | TCP_RCV_COPY_FRAGHDR | TCP_RCV_COPY_XID; | |
51971139 | 1153 | |
44fbac22 | 1154 | xprt_wake_pending_tasks(xprt, 0); |
a246b010 | 1155 | } |
4a0f8c04 | 1156 | spin_unlock_bh(&xprt->transport_lock); |
a246b010 | 1157 | break; |
3b948ae5 TM |
1158 | case TCP_FIN_WAIT1: |
1159 | /* The client initiated a shutdown of the socket */ | |
7c1d71cf | 1160 | xprt->connect_cookie++; |
663b8858 | 1161 | xprt->reestablish_timeout = 0; |
3b948ae5 TM |
1162 | set_bit(XPRT_CLOSING, &xprt->state); |
1163 | smp_mb__before_clear_bit(); | |
1164 | clear_bit(XPRT_CONNECTED, &xprt->state); | |
ef803670 | 1165 | clear_bit(XPRT_CLOSE_WAIT, &xprt->state); |
3b948ae5 | 1166 | smp_mb__after_clear_bit(); |
a246b010 | 1167 | break; |
632e3bdc | 1168 | case TCP_CLOSE_WAIT: |
3b948ae5 TM |
1169 | /* The server initiated a shutdown of the socket */ |
1170 | set_bit(XPRT_CLOSING, &xprt->state); | |
66af1e55 | 1171 | xprt_force_disconnect(xprt); |
663b8858 | 1172 | case TCP_SYN_SENT: |
7c1d71cf | 1173 | xprt->connect_cookie++; |
663b8858 TM |
1174 | case TCP_CLOSING: |
1175 | /* | |
1176 | * If the server closed down the connection, make sure that | |
1177 | * we back off before reconnecting | |
1178 | */ | |
1179 | if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO) | |
1180 | xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; | |
3b948ae5 TM |
1181 | break; |
1182 | case TCP_LAST_ACK: | |
1183 | smp_mb__before_clear_bit(); | |
1184 | clear_bit(XPRT_CONNECTED, &xprt->state); | |
1185 | smp_mb__after_clear_bit(); | |
1186 | break; | |
1187 | case TCP_CLOSE: | |
1188 | smp_mb__before_clear_bit(); | |
ef803670 | 1189 | clear_bit(XPRT_CLOSE_WAIT, &xprt->state); |
3b948ae5 TM |
1190 | clear_bit(XPRT_CLOSING, &xprt->state); |
1191 | smp_mb__after_clear_bit(); | |
1192 | /* Mark transport as closed and wake up all pending tasks */ | |
62da3b24 | 1193 | xprt_disconnect_done(xprt); |
a246b010 CL |
1194 | } |
1195 | out: | |
1196 | read_unlock(&sk->sk_callback_lock); | |
1197 | } | |
1198 | ||
2a9e1cfa TM |
1199 | /** |
1200 | * xs_tcp_error_report - callback mainly for catching RST events | |
1201 | * @sk: socket | |
1202 | */ | |
1203 | static void xs_tcp_error_report(struct sock *sk) | |
1204 | { | |
1205 | struct rpc_xprt *xprt; | |
1206 | ||
1207 | read_lock(&sk->sk_callback_lock); | |
1208 | if (sk->sk_err != ECONNRESET || sk->sk_state != TCP_ESTABLISHED) | |
1209 | goto out; | |
1210 | if (!(xprt = xprt_from_sock(sk))) | |
1211 | goto out; | |
1212 | dprintk("RPC: %s client %p...\n" | |
1213 | "RPC: error %d\n", | |
1214 | __func__, xprt, sk->sk_err); | |
1215 | ||
1216 | xprt_force_disconnect(xprt); | |
1217 | out: | |
1218 | read_unlock(&sk->sk_callback_lock); | |
1219 | } | |
1220 | ||
9903cd1c | 1221 | /** |
c7b2cae8 CL |
1222 | * xs_udp_write_space - callback invoked when socket buffer space |
1223 | * becomes available | |
9903cd1c CL |
1224 | * @sk: socket whose state has changed |
1225 | * | |
a246b010 CL |
1226 | * Called when more output buffer space is available for this socket. |
1227 | * We try not to wake our writers until they can make "significant" | |
c7b2cae8 | 1228 | * progress, otherwise we'll waste resources thrashing kernel_sendmsg |
a246b010 CL |
1229 | * with a bunch of small requests. |
1230 | */ | |
c7b2cae8 | 1231 | static void xs_udp_write_space(struct sock *sk) |
a246b010 | 1232 | { |
a246b010 | 1233 | read_lock(&sk->sk_callback_lock); |
a246b010 | 1234 | |
c7b2cae8 CL |
1235 | /* from net/core/sock.c:sock_def_write_space */ |
1236 | if (sock_writeable(sk)) { | |
1237 | struct socket *sock; | |
1238 | struct rpc_xprt *xprt; | |
1239 | ||
1240 | if (unlikely(!(sock = sk->sk_socket))) | |
a246b010 | 1241 | goto out; |
b6ddf64f TM |
1242 | clear_bit(SOCK_NOSPACE, &sock->flags); |
1243 | ||
c7b2cae8 CL |
1244 | if (unlikely(!(xprt = xprt_from_sock(sk)))) |
1245 | goto out; | |
b6ddf64f | 1246 | if (test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags) == 0) |
a246b010 | 1247 | goto out; |
c7b2cae8 CL |
1248 | |
1249 | xprt_write_space(xprt); | |
a246b010 CL |
1250 | } |
1251 | ||
c7b2cae8 CL |
1252 | out: |
1253 | read_unlock(&sk->sk_callback_lock); | |
1254 | } | |
a246b010 | 1255 | |
c7b2cae8 CL |
1256 | /** |
1257 | * xs_tcp_write_space - callback invoked when socket buffer space | |
1258 | * becomes available | |
1259 | * @sk: socket whose state has changed | |
1260 | * | |
1261 | * Called when more output buffer space is available for this socket. | |
1262 | * We try not to wake our writers until they can make "significant" | |
1263 | * progress, otherwise we'll waste resources thrashing kernel_sendmsg | |
1264 | * with a bunch of small requests. | |
1265 | */ | |
1266 | static void xs_tcp_write_space(struct sock *sk) | |
1267 | { | |
1268 | read_lock(&sk->sk_callback_lock); | |
1269 | ||
1270 | /* from net/core/stream.c:sk_stream_write_space */ | |
1271 | if (sk_stream_wspace(sk) >= sk_stream_min_wspace(sk)) { | |
1272 | struct socket *sock; | |
1273 | struct rpc_xprt *xprt; | |
1274 | ||
1275 | if (unlikely(!(sock = sk->sk_socket))) | |
1276 | goto out; | |
b6ddf64f TM |
1277 | clear_bit(SOCK_NOSPACE, &sock->flags); |
1278 | ||
c7b2cae8 CL |
1279 | if (unlikely(!(xprt = xprt_from_sock(sk)))) |
1280 | goto out; | |
b6ddf64f | 1281 | if (test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags) == 0) |
c7b2cae8 CL |
1282 | goto out; |
1283 | ||
1284 | xprt_write_space(xprt); | |
1285 | } | |
1286 | ||
1287 | out: | |
a246b010 CL |
1288 | read_unlock(&sk->sk_callback_lock); |
1289 | } | |
1290 | ||
470056c2 | 1291 | static void xs_udp_do_set_buffer_size(struct rpc_xprt *xprt) |
a246b010 | 1292 | { |
ee0ac0c2 CL |
1293 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
1294 | struct sock *sk = transport->inet; | |
a246b010 | 1295 | |
7c6e066e | 1296 | if (transport->rcvsize) { |
a246b010 | 1297 | sk->sk_userlocks |= SOCK_RCVBUF_LOCK; |
7c6e066e | 1298 | sk->sk_rcvbuf = transport->rcvsize * xprt->max_reqs * 2; |
a246b010 | 1299 | } |
7c6e066e | 1300 | if (transport->sndsize) { |
a246b010 | 1301 | sk->sk_userlocks |= SOCK_SNDBUF_LOCK; |
7c6e066e | 1302 | sk->sk_sndbuf = transport->sndsize * xprt->max_reqs * 2; |
a246b010 CL |
1303 | sk->sk_write_space(sk); |
1304 | } | |
1305 | } | |
1306 | ||
43118c29 | 1307 | /** |
470056c2 | 1308 | * xs_udp_set_buffer_size - set send and receive limits |
43118c29 | 1309 | * @xprt: generic transport |
470056c2 CL |
1310 | * @sndsize: requested size of send buffer, in bytes |
1311 | * @rcvsize: requested size of receive buffer, in bytes | |
43118c29 | 1312 | * |
470056c2 | 1313 | * Set socket send and receive buffer size limits. |
43118c29 | 1314 | */ |
470056c2 | 1315 | static void xs_udp_set_buffer_size(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize) |
43118c29 | 1316 | { |
7c6e066e CL |
1317 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
1318 | ||
1319 | transport->sndsize = 0; | |
470056c2 | 1320 | if (sndsize) |
7c6e066e CL |
1321 | transport->sndsize = sndsize + 1024; |
1322 | transport->rcvsize = 0; | |
470056c2 | 1323 | if (rcvsize) |
7c6e066e | 1324 | transport->rcvsize = rcvsize + 1024; |
470056c2 CL |
1325 | |
1326 | xs_udp_do_set_buffer_size(xprt); | |
43118c29 CL |
1327 | } |
1328 | ||
46c0ee8b CL |
1329 | /** |
1330 | * xs_udp_timer - called when a retransmit timeout occurs on a UDP transport | |
1331 | * @task: task that timed out | |
1332 | * | |
1333 | * Adjust the congestion window after a retransmit timeout has occurred. | |
1334 | */ | |
1335 | static void xs_udp_timer(struct rpc_task *task) | |
1336 | { | |
1337 | xprt_adjust_cwnd(task, -ETIMEDOUT); | |
1338 | } | |
1339 | ||
b85d8806 CL |
1340 | static unsigned short xs_get_random_port(void) |
1341 | { | |
1342 | unsigned short range = xprt_max_resvport - xprt_min_resvport; | |
1343 | unsigned short rand = (unsigned short) net_random() % range; | |
1344 | return rand + xprt_min_resvport; | |
1345 | } | |
1346 | ||
92200412 CL |
1347 | /** |
1348 | * xs_set_port - reset the port number in the remote endpoint address | |
1349 | * @xprt: generic transport | |
1350 | * @port: new port number | |
1351 | * | |
1352 | */ | |
1353 | static void xs_set_port(struct rpc_xprt *xprt, unsigned short port) | |
1354 | { | |
95392c59 | 1355 | struct sockaddr *addr = xs_addr(xprt); |
c4efcb1d | 1356 | |
46121cf7 | 1357 | dprintk("RPC: setting port for xprt %p to %u\n", xprt, port); |
c4efcb1d | 1358 | |
20612005 CL |
1359 | switch (addr->sa_family) { |
1360 | case AF_INET: | |
1361 | ((struct sockaddr_in *)addr)->sin_port = htons(port); | |
1362 | break; | |
1363 | case AF_INET6: | |
1364 | ((struct sockaddr_in6 *)addr)->sin6_port = htons(port); | |
1365 | break; | |
1366 | default: | |
1367 | BUG(); | |
1368 | } | |
92200412 CL |
1369 | } |
1370 | ||
67a391d7 TM |
1371 | static unsigned short xs_get_srcport(struct sock_xprt *transport, struct socket *sock) |
1372 | { | |
1373 | unsigned short port = transport->port; | |
1374 | ||
1375 | if (port == 0 && transport->xprt.resvport) | |
1376 | port = xs_get_random_port(); | |
1377 | return port; | |
1378 | } | |
1379 | ||
1380 | static unsigned short xs_next_srcport(struct sock_xprt *transport, struct socket *sock, unsigned short port) | |
1381 | { | |
1382 | if (transport->port != 0) | |
1383 | transport->port = 0; | |
1384 | if (!transport->xprt.resvport) | |
1385 | return 0; | |
1386 | if (port <= xprt_min_resvport || port > xprt_max_resvport) | |
1387 | return xprt_max_resvport; | |
1388 | return --port; | |
1389 | } | |
1390 | ||
7dc753f0 | 1391 | static int xs_bind4(struct sock_xprt *transport, struct socket *sock) |
a246b010 CL |
1392 | { |
1393 | struct sockaddr_in myaddr = { | |
1394 | .sin_family = AF_INET, | |
1395 | }; | |
d3bc9a1d | 1396 | struct sockaddr_in *sa; |
67a391d7 TM |
1397 | int err, nloop = 0; |
1398 | unsigned short port = xs_get_srcport(transport, sock); | |
1399 | unsigned short last; | |
a246b010 | 1400 | |
d3bc9a1d FM |
1401 | sa = (struct sockaddr_in *)&transport->addr; |
1402 | myaddr.sin_addr = sa->sin_addr; | |
a246b010 CL |
1403 | do { |
1404 | myaddr.sin_port = htons(port); | |
e6242e92 | 1405 | err = kernel_bind(sock, (struct sockaddr *) &myaddr, |
a246b010 | 1406 | sizeof(myaddr)); |
67a391d7 | 1407 | if (port == 0) |
d3bc9a1d | 1408 | break; |
a246b010 | 1409 | if (err == 0) { |
c8475461 | 1410 | transport->port = port; |
d3bc9a1d | 1411 | break; |
a246b010 | 1412 | } |
67a391d7 TM |
1413 | last = port; |
1414 | port = xs_next_srcport(transport, sock, port); | |
1415 | if (port > last) | |
1416 | nloop++; | |
1417 | } while (err == -EADDRINUSE && nloop != 2); | |
7dc753f0 | 1418 | dprintk("RPC: %s "NIPQUAD_FMT":%u: %s (%d)\n", |
0dc47877 | 1419 | __func__, NIPQUAD(myaddr.sin_addr), |
7dc753f0 | 1420 | port, err ? "failed" : "ok", err); |
a246b010 CL |
1421 | return err; |
1422 | } | |
1423 | ||
90058d37 CL |
1424 | static int xs_bind6(struct sock_xprt *transport, struct socket *sock) |
1425 | { | |
1426 | struct sockaddr_in6 myaddr = { | |
1427 | .sin6_family = AF_INET6, | |
1428 | }; | |
1429 | struct sockaddr_in6 *sa; | |
67a391d7 TM |
1430 | int err, nloop = 0; |
1431 | unsigned short port = xs_get_srcport(transport, sock); | |
1432 | unsigned short last; | |
90058d37 | 1433 | |
90058d37 CL |
1434 | sa = (struct sockaddr_in6 *)&transport->addr; |
1435 | myaddr.sin6_addr = sa->sin6_addr; | |
1436 | do { | |
1437 | myaddr.sin6_port = htons(port); | |
1438 | err = kernel_bind(sock, (struct sockaddr *) &myaddr, | |
1439 | sizeof(myaddr)); | |
67a391d7 | 1440 | if (port == 0) |
90058d37 CL |
1441 | break; |
1442 | if (err == 0) { | |
1443 | transport->port = port; | |
1444 | break; | |
1445 | } | |
67a391d7 TM |
1446 | last = port; |
1447 | port = xs_next_srcport(transport, sock, port); | |
1448 | if (port > last) | |
1449 | nloop++; | |
1450 | } while (err == -EADDRINUSE && nloop != 2); | |
90058d37 CL |
1451 | dprintk("RPC: xs_bind6 "NIP6_FMT":%u: %s (%d)\n", |
1452 | NIP6(myaddr.sin6_addr), port, err ? "failed" : "ok", err); | |
a246b010 CL |
1453 | return err; |
1454 | } | |
1455 | ||
ed07536e PZ |
1456 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
1457 | static struct lock_class_key xs_key[2]; | |
1458 | static struct lock_class_key xs_slock_key[2]; | |
1459 | ||
8945ee5e | 1460 | static inline void xs_reclassify_socket4(struct socket *sock) |
ed07536e PZ |
1461 | { |
1462 | struct sock *sk = sock->sk; | |
8945ee5e | 1463 | |
02b3d346 | 1464 | BUG_ON(sock_owned_by_user(sk)); |
8945ee5e CL |
1465 | sock_lock_init_class_and_name(sk, "slock-AF_INET-RPC", |
1466 | &xs_slock_key[0], "sk_lock-AF_INET-RPC", &xs_key[0]); | |
1467 | } | |
ed07536e | 1468 | |
8945ee5e CL |
1469 | static inline void xs_reclassify_socket6(struct socket *sock) |
1470 | { | |
1471 | struct sock *sk = sock->sk; | |
ed07536e | 1472 | |
f4921aff | 1473 | BUG_ON(sock_owned_by_user(sk)); |
8945ee5e CL |
1474 | sock_lock_init_class_and_name(sk, "slock-AF_INET6-RPC", |
1475 | &xs_slock_key[1], "sk_lock-AF_INET6-RPC", &xs_key[1]); | |
ed07536e PZ |
1476 | } |
1477 | #else | |
8945ee5e CL |
1478 | static inline void xs_reclassify_socket4(struct socket *sock) |
1479 | { | |
1480 | } | |
1481 | ||
1482 | static inline void xs_reclassify_socket6(struct socket *sock) | |
ed07536e PZ |
1483 | { |
1484 | } | |
1485 | #endif | |
1486 | ||
16be2d20 CL |
1487 | static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock) |
1488 | { | |
1489 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); | |
1490 | ||
1491 | if (!transport->inet) { | |
1492 | struct sock *sk = sock->sk; | |
1493 | ||
1494 | write_lock_bh(&sk->sk_callback_lock); | |
1495 | ||
2a9e1cfa TM |
1496 | xs_save_old_callbacks(transport, sk); |
1497 | ||
16be2d20 | 1498 | sk->sk_user_data = xprt; |
16be2d20 CL |
1499 | sk->sk_data_ready = xs_udp_data_ready; |
1500 | sk->sk_write_space = xs_udp_write_space; | |
1501 | sk->sk_no_check = UDP_CSUM_NORCV; | |
1502 | sk->sk_allocation = GFP_ATOMIC; | |
1503 | ||
1504 | xprt_set_connected(xprt); | |
1505 | ||
1506 | /* Reset to new socket */ | |
1507 | transport->sock = sock; | |
1508 | transport->inet = sk; | |
1509 | ||
1510 | write_unlock_bh(&sk->sk_callback_lock); | |
1511 | } | |
1512 | xs_udp_do_set_buffer_size(xprt); | |
1513 | } | |
1514 | ||
b0d93ad5 | 1515 | /** |
9c3d72de | 1516 | * xs_udp_connect_worker4 - set up a UDP socket |
65f27f38 | 1517 | * @work: RPC transport to connect |
b0d93ad5 CL |
1518 | * |
1519 | * Invoked by a work queue tasklet. | |
1520 | */ | |
9c3d72de | 1521 | static void xs_udp_connect_worker4(struct work_struct *work) |
a246b010 | 1522 | { |
34161db6 TM |
1523 | struct sock_xprt *transport = |
1524 | container_of(work, struct sock_xprt, connect_worker.work); | |
c8475461 | 1525 | struct rpc_xprt *xprt = &transport->xprt; |
ee0ac0c2 | 1526 | struct socket *sock = transport->sock; |
b0d93ad5 | 1527 | int err, status = -EIO; |
9903cd1c | 1528 | |
ec739ef0 | 1529 | if (xprt->shutdown || !xprt_bound(xprt)) |
b0d93ad5 | 1530 | goto out; |
9903cd1c | 1531 | |
b0d93ad5 CL |
1532 | /* Start by resetting any existing state */ |
1533 | xs_close(xprt); | |
9903cd1c | 1534 | |
b0d93ad5 | 1535 | if ((err = sock_create_kern(PF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock)) < 0) { |
46121cf7 | 1536 | dprintk("RPC: can't create UDP transport socket (%d).\n", -err); |
b0d93ad5 CL |
1537 | goto out; |
1538 | } | |
8945ee5e | 1539 | xs_reclassify_socket4(sock); |
9903cd1c | 1540 | |
7dc753f0 | 1541 | if (xs_bind4(transport, sock)) { |
b0d93ad5 CL |
1542 | sock_release(sock); |
1543 | goto out; | |
1544 | } | |
9903cd1c | 1545 | |
46121cf7 | 1546 | dprintk("RPC: worker connecting xprt %p to address: %s\n", |
7559c7a2 | 1547 | xprt, xprt->address_strings[RPC_DISPLAY_ALL]); |
edb267a6 | 1548 | |
16be2d20 | 1549 | xs_udp_finish_connecting(xprt, sock); |
b0d93ad5 CL |
1550 | status = 0; |
1551 | out: | |
1552 | xprt_wake_pending_tasks(xprt, status); | |
1553 | xprt_clear_connecting(xprt); | |
a246b010 | 1554 | } |
a246b010 | 1555 | |
68e220bd CL |
1556 | /** |
1557 | * xs_udp_connect_worker6 - set up a UDP socket | |
1558 | * @work: RPC transport to connect | |
1559 | * | |
1560 | * Invoked by a work queue tasklet. | |
1561 | */ | |
1562 | static void xs_udp_connect_worker6(struct work_struct *work) | |
1563 | { | |
1564 | struct sock_xprt *transport = | |
1565 | container_of(work, struct sock_xprt, connect_worker.work); | |
1566 | struct rpc_xprt *xprt = &transport->xprt; | |
1567 | struct socket *sock = transport->sock; | |
1568 | int err, status = -EIO; | |
a246b010 | 1569 | |
68e220bd CL |
1570 | if (xprt->shutdown || !xprt_bound(xprt)) |
1571 | goto out; | |
b0d93ad5 | 1572 | |
68e220bd CL |
1573 | /* Start by resetting any existing state */ |
1574 | xs_close(xprt); | |
a246b010 | 1575 | |
68e220bd CL |
1576 | if ((err = sock_create_kern(PF_INET6, SOCK_DGRAM, IPPROTO_UDP, &sock)) < 0) { |
1577 | dprintk("RPC: can't create UDP transport socket (%d).\n", -err); | |
1578 | goto out; | |
1579 | } | |
8945ee5e | 1580 | xs_reclassify_socket6(sock); |
a246b010 | 1581 | |
68e220bd CL |
1582 | if (xs_bind6(transport, sock) < 0) { |
1583 | sock_release(sock); | |
1584 | goto out; | |
b0d93ad5 | 1585 | } |
68e220bd CL |
1586 | |
1587 | dprintk("RPC: worker connecting xprt %p to address: %s\n", | |
1588 | xprt, xprt->address_strings[RPC_DISPLAY_ALL]); | |
1589 | ||
1590 | xs_udp_finish_connecting(xprt, sock); | |
b0d93ad5 CL |
1591 | status = 0; |
1592 | out: | |
1593 | xprt_wake_pending_tasks(xprt, status); | |
1594 | xprt_clear_connecting(xprt); | |
a246b010 CL |
1595 | } |
1596 | ||
3167e12c CL |
1597 | /* |
1598 | * We need to preserve the port number so the reply cache on the server can | |
1599 | * find our cached RPC replies when we get around to reconnecting. | |
1600 | */ | |
1601 | static void xs_tcp_reuse_connection(struct rpc_xprt *xprt) | |
1602 | { | |
1603 | int result; | |
ee0ac0c2 | 1604 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
3167e12c CL |
1605 | struct sockaddr any; |
1606 | ||
46121cf7 | 1607 | dprintk("RPC: disconnecting xprt %p to reuse port\n", xprt); |
3167e12c CL |
1608 | |
1609 | /* | |
1610 | * Disconnect the transport socket by doing a connect operation | |
1611 | * with AF_UNSPEC. This should return immediately... | |
1612 | */ | |
1613 | memset(&any, 0, sizeof(any)); | |
1614 | any.sa_family = AF_UNSPEC; | |
ee0ac0c2 | 1615 | result = kernel_connect(transport->sock, &any, sizeof(any), 0); |
3167e12c | 1616 | if (result) |
46121cf7 | 1617 | dprintk("RPC: AF_UNSPEC connect return code %d\n", |
3167e12c CL |
1618 | result); |
1619 | } | |
1620 | ||
16be2d20 | 1621 | static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock) |
a246b010 | 1622 | { |
16be2d20 | 1623 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
edb267a6 | 1624 | |
ee0ac0c2 | 1625 | if (!transport->inet) { |
b0d93ad5 CL |
1626 | struct sock *sk = sock->sk; |
1627 | ||
1628 | write_lock_bh(&sk->sk_callback_lock); | |
1629 | ||
2a9e1cfa TM |
1630 | xs_save_old_callbacks(transport, sk); |
1631 | ||
b0d93ad5 | 1632 | sk->sk_user_data = xprt; |
b0d93ad5 CL |
1633 | sk->sk_data_ready = xs_tcp_data_ready; |
1634 | sk->sk_state_change = xs_tcp_state_change; | |
1635 | sk->sk_write_space = xs_tcp_write_space; | |
2a9e1cfa | 1636 | sk->sk_error_report = xs_tcp_error_report; |
b079fa7b | 1637 | sk->sk_allocation = GFP_ATOMIC; |
3167e12c CL |
1638 | |
1639 | /* socket options */ | |
1640 | sk->sk_userlocks |= SOCK_BINDPORT_LOCK; | |
1641 | sock_reset_flag(sk, SOCK_LINGER); | |
1642 | tcp_sk(sk)->linger2 = 0; | |
1643 | tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF; | |
b0d93ad5 CL |
1644 | |
1645 | xprt_clear_connected(xprt); | |
1646 | ||
1647 | /* Reset to new socket */ | |
ee0ac0c2 CL |
1648 | transport->sock = sock; |
1649 | transport->inet = sk; | |
b0d93ad5 CL |
1650 | |
1651 | write_unlock_bh(&sk->sk_callback_lock); | |
1652 | } | |
1653 | ||
1654 | /* Tell the socket layer to start connecting... */ | |
262ca07d CL |
1655 | xprt->stat.connect_count++; |
1656 | xprt->stat.connect_start = jiffies; | |
95392c59 | 1657 | return kernel_connect(sock, xs_addr(xprt), xprt->addrlen, O_NONBLOCK); |
16be2d20 CL |
1658 | } |
1659 | ||
9903cd1c | 1660 | /** |
9c3d72de | 1661 | * xs_tcp_connect_worker4 - connect a TCP socket to a remote endpoint |
65f27f38 | 1662 | * @work: RPC transport to connect |
9903cd1c CL |
1663 | * |
1664 | * Invoked by a work queue tasklet. | |
a246b010 | 1665 | */ |
9c3d72de | 1666 | static void xs_tcp_connect_worker4(struct work_struct *work) |
a246b010 | 1667 | { |
34161db6 TM |
1668 | struct sock_xprt *transport = |
1669 | container_of(work, struct sock_xprt, connect_worker.work); | |
c8475461 | 1670 | struct rpc_xprt *xprt = &transport->xprt; |
ee0ac0c2 | 1671 | struct socket *sock = transport->sock; |
b0d93ad5 | 1672 | int err, status = -EIO; |
a246b010 | 1673 | |
ec739ef0 | 1674 | if (xprt->shutdown || !xprt_bound(xprt)) |
a246b010 CL |
1675 | goto out; |
1676 | ||
ee0ac0c2 | 1677 | if (!sock) { |
3167e12c CL |
1678 | /* start from scratch */ |
1679 | if ((err = sock_create_kern(PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock)) < 0) { | |
16be2d20 | 1680 | dprintk("RPC: can't create TCP transport socket (%d).\n", -err); |
3167e12c CL |
1681 | goto out; |
1682 | } | |
8945ee5e | 1683 | xs_reclassify_socket4(sock); |
a246b010 | 1684 | |
16be2d20 | 1685 | if (xs_bind4(transport, sock) < 0) { |
3167e12c CL |
1686 | sock_release(sock); |
1687 | goto out; | |
1688 | } | |
1689 | } else | |
1690 | /* "close" the socket, preserving the local port */ | |
1691 | xs_tcp_reuse_connection(xprt); | |
a246b010 | 1692 | |
46121cf7 | 1693 | dprintk("RPC: worker connecting xprt %p to address: %s\n", |
7559c7a2 | 1694 | xprt, xprt->address_strings[RPC_DISPLAY_ALL]); |
edb267a6 | 1695 | |
16be2d20 | 1696 | status = xs_tcp_finish_connecting(xprt, sock); |
46121cf7 CL |
1697 | dprintk("RPC: %p connect status %d connected %d sock state %d\n", |
1698 | xprt, -status, xprt_connected(xprt), | |
1699 | sock->sk->sk_state); | |
a246b010 CL |
1700 | if (status < 0) { |
1701 | switch (status) { | |
1702 | case -EINPROGRESS: | |
1703 | case -EALREADY: | |
1704 | goto out_clear; | |
3167e12c CL |
1705 | case -ECONNREFUSED: |
1706 | case -ECONNRESET: | |
1707 | /* retry with existing socket, after a delay */ | |
1708 | break; | |
1709 | default: | |
1710 | /* get rid of existing socket, and retry */ | |
e06799f9 | 1711 | xs_tcp_shutdown(xprt); |
a246b010 CL |
1712 | } |
1713 | } | |
1714 | out: | |
44fbac22 | 1715 | xprt_wake_pending_tasks(xprt, status); |
68e220bd CL |
1716 | out_clear: |
1717 | xprt_clear_connecting(xprt); | |
1718 | } | |
b0d93ad5 | 1719 | |
68e220bd CL |
1720 | /** |
1721 | * xs_tcp_connect_worker6 - connect a TCP socket to a remote endpoint | |
1722 | * @work: RPC transport to connect | |
1723 | * | |
1724 | * Invoked by a work queue tasklet. | |
1725 | */ | |
1726 | static void xs_tcp_connect_worker6(struct work_struct *work) | |
1727 | { | |
1728 | struct sock_xprt *transport = | |
1729 | container_of(work, struct sock_xprt, connect_worker.work); | |
1730 | struct rpc_xprt *xprt = &transport->xprt; | |
1731 | struct socket *sock = transport->sock; | |
1732 | int err, status = -EIO; | |
3167e12c | 1733 | |
68e220bd CL |
1734 | if (xprt->shutdown || !xprt_bound(xprt)) |
1735 | goto out; | |
b0d93ad5 | 1736 | |
68e220bd CL |
1737 | if (!sock) { |
1738 | /* start from scratch */ | |
1739 | if ((err = sock_create_kern(PF_INET6, SOCK_STREAM, IPPROTO_TCP, &sock)) < 0) { | |
1740 | dprintk("RPC: can't create TCP transport socket (%d).\n", -err); | |
1741 | goto out; | |
1742 | } | |
8945ee5e | 1743 | xs_reclassify_socket6(sock); |
b0d93ad5 | 1744 | |
68e220bd CL |
1745 | if (xs_bind6(transport, sock) < 0) { |
1746 | sock_release(sock); | |
1747 | goto out; | |
1748 | } | |
1749 | } else | |
1750 | /* "close" the socket, preserving the local port */ | |
1751 | xs_tcp_reuse_connection(xprt); | |
b0d93ad5 | 1752 | |
68e220bd CL |
1753 | dprintk("RPC: worker connecting xprt %p to address: %s\n", |
1754 | xprt, xprt->address_strings[RPC_DISPLAY_ALL]); | |
b0d93ad5 | 1755 | |
68e220bd | 1756 | status = xs_tcp_finish_connecting(xprt, sock); |
46121cf7 | 1757 | dprintk("RPC: %p connect status %d connected %d sock state %d\n", |
68e220bd | 1758 | xprt, -status, xprt_connected(xprt), sock->sk->sk_state); |
a246b010 CL |
1759 | if (status < 0) { |
1760 | switch (status) { | |
1761 | case -EINPROGRESS: | |
1762 | case -EALREADY: | |
1763 | goto out_clear; | |
3167e12c CL |
1764 | case -ECONNREFUSED: |
1765 | case -ECONNRESET: | |
1766 | /* retry with existing socket, after a delay */ | |
1767 | break; | |
1768 | default: | |
1769 | /* get rid of existing socket, and retry */ | |
e06799f9 | 1770 | xs_tcp_shutdown(xprt); |
a246b010 CL |
1771 | } |
1772 | } | |
1773 | out: | |
44fbac22 | 1774 | xprt_wake_pending_tasks(xprt, status); |
a246b010 | 1775 | out_clear: |
2226feb6 | 1776 | xprt_clear_connecting(xprt); |
a246b010 CL |
1777 | } |
1778 | ||
9903cd1c CL |
1779 | /** |
1780 | * xs_connect - connect a socket to a remote endpoint | |
1781 | * @task: address of RPC task that manages state of connect request | |
1782 | * | |
1783 | * TCP: If the remote end dropped the connection, delay reconnecting. | |
03bf4b70 CL |
1784 | * |
1785 | * UDP socket connects are synchronous, but we use a work queue anyway | |
1786 | * to guarantee that even unprivileged user processes can set up a | |
1787 | * socket on a privileged port. | |
1788 | * | |
1789 | * If a UDP socket connect fails, the delay behavior here prevents | |
1790 | * retry floods (hard mounts). | |
9903cd1c CL |
1791 | */ |
1792 | static void xs_connect(struct rpc_task *task) | |
a246b010 CL |
1793 | { |
1794 | struct rpc_xprt *xprt = task->tk_xprt; | |
ee0ac0c2 | 1795 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
a246b010 | 1796 | |
b0d93ad5 CL |
1797 | if (xprt_test_and_set_connecting(xprt)) |
1798 | return; | |
1799 | ||
ee0ac0c2 | 1800 | if (transport->sock != NULL) { |
46121cf7 CL |
1801 | dprintk("RPC: xs_connect delayed xprt %p for %lu " |
1802 | "seconds\n", | |
03bf4b70 | 1803 | xprt, xprt->reestablish_timeout / HZ); |
c1384c9c TM |
1804 | queue_delayed_work(rpciod_workqueue, |
1805 | &transport->connect_worker, | |
1806 | xprt->reestablish_timeout); | |
03bf4b70 CL |
1807 | xprt->reestablish_timeout <<= 1; |
1808 | if (xprt->reestablish_timeout > XS_TCP_MAX_REEST_TO) | |
1809 | xprt->reestablish_timeout = XS_TCP_MAX_REEST_TO; | |
b0d93ad5 | 1810 | } else { |
46121cf7 | 1811 | dprintk("RPC: xs_connect scheduled xprt %p\n", xprt); |
c1384c9c TM |
1812 | queue_delayed_work(rpciod_workqueue, |
1813 | &transport->connect_worker, 0); | |
a246b010 CL |
1814 | } |
1815 | } | |
1816 | ||
e06799f9 TM |
1817 | static void xs_tcp_connect(struct rpc_task *task) |
1818 | { | |
1819 | struct rpc_xprt *xprt = task->tk_xprt; | |
1820 | ||
1821 | /* Initiate graceful shutdown of the socket if not already done */ | |
1822 | if (test_bit(XPRT_CONNECTED, &xprt->state)) | |
1823 | xs_tcp_shutdown(xprt); | |
1824 | /* Exit if we need to wait for socket shutdown to complete */ | |
1825 | if (test_bit(XPRT_CLOSING, &xprt->state)) | |
1826 | return; | |
1827 | xs_connect(task); | |
1828 | } | |
1829 | ||
262ca07d CL |
1830 | /** |
1831 | * xs_udp_print_stats - display UDP socket-specifc stats | |
1832 | * @xprt: rpc_xprt struct containing statistics | |
1833 | * @seq: output file | |
1834 | * | |
1835 | */ | |
1836 | static void xs_udp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq) | |
1837 | { | |
c8475461 CL |
1838 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
1839 | ||
262ca07d | 1840 | seq_printf(seq, "\txprt:\tudp %u %lu %lu %lu %lu %Lu %Lu\n", |
c8475461 | 1841 | transport->port, |
262ca07d CL |
1842 | xprt->stat.bind_count, |
1843 | xprt->stat.sends, | |
1844 | xprt->stat.recvs, | |
1845 | xprt->stat.bad_xids, | |
1846 | xprt->stat.req_u, | |
1847 | xprt->stat.bklog_u); | |
1848 | } | |
1849 | ||
1850 | /** | |
1851 | * xs_tcp_print_stats - display TCP socket-specifc stats | |
1852 | * @xprt: rpc_xprt struct containing statistics | |
1853 | * @seq: output file | |
1854 | * | |
1855 | */ | |
1856 | static void xs_tcp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq) | |
1857 | { | |
c8475461 | 1858 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
262ca07d CL |
1859 | long idle_time = 0; |
1860 | ||
1861 | if (xprt_connected(xprt)) | |
1862 | idle_time = (long)(jiffies - xprt->last_used) / HZ; | |
1863 | ||
1864 | seq_printf(seq, "\txprt:\ttcp %u %lu %lu %lu %ld %lu %lu %lu %Lu %Lu\n", | |
c8475461 | 1865 | transport->port, |
262ca07d CL |
1866 | xprt->stat.bind_count, |
1867 | xprt->stat.connect_count, | |
1868 | xprt->stat.connect_time, | |
1869 | idle_time, | |
1870 | xprt->stat.sends, | |
1871 | xprt->stat.recvs, | |
1872 | xprt->stat.bad_xids, | |
1873 | xprt->stat.req_u, | |
1874 | xprt->stat.bklog_u); | |
1875 | } | |
1876 | ||
262965f5 | 1877 | static struct rpc_xprt_ops xs_udp_ops = { |
43118c29 | 1878 | .set_buffer_size = xs_udp_set_buffer_size, |
12a80469 | 1879 | .reserve_xprt = xprt_reserve_xprt_cong, |
49e9a890 | 1880 | .release_xprt = xprt_release_xprt_cong, |
45160d62 | 1881 | .rpcbind = rpcb_getport_async, |
92200412 | 1882 | .set_port = xs_set_port, |
262965f5 | 1883 | .connect = xs_connect, |
02107148 CL |
1884 | .buf_alloc = rpc_malloc, |
1885 | .buf_free = rpc_free, | |
262965f5 | 1886 | .send_request = xs_udp_send_request, |
fe3aca29 | 1887 | .set_retrans_timeout = xprt_set_retrans_timeout_rtt, |
46c0ee8b | 1888 | .timer = xs_udp_timer, |
a58dd398 | 1889 | .release_request = xprt_release_rqst_cong, |
262965f5 CL |
1890 | .close = xs_close, |
1891 | .destroy = xs_destroy, | |
262ca07d | 1892 | .print_stats = xs_udp_print_stats, |
262965f5 CL |
1893 | }; |
1894 | ||
1895 | static struct rpc_xprt_ops xs_tcp_ops = { | |
12a80469 | 1896 | .reserve_xprt = xprt_reserve_xprt, |
e0ab53de | 1897 | .release_xprt = xs_tcp_release_xprt, |
45160d62 | 1898 | .rpcbind = rpcb_getport_async, |
92200412 | 1899 | .set_port = xs_set_port, |
e06799f9 | 1900 | .connect = xs_tcp_connect, |
02107148 CL |
1901 | .buf_alloc = rpc_malloc, |
1902 | .buf_free = rpc_free, | |
262965f5 | 1903 | .send_request = xs_tcp_send_request, |
fe3aca29 | 1904 | .set_retrans_timeout = xprt_set_retrans_timeout_def, |
e06799f9 | 1905 | .close = xs_tcp_shutdown, |
9903cd1c | 1906 | .destroy = xs_destroy, |
262ca07d | 1907 | .print_stats = xs_tcp_print_stats, |
a246b010 CL |
1908 | }; |
1909 | ||
3c341b0b | 1910 | static struct rpc_xprt *xs_setup_xprt(struct xprt_create *args, |
bc25571e | 1911 | unsigned int slot_table_size) |
c8541ecd CL |
1912 | { |
1913 | struct rpc_xprt *xprt; | |
ffc2e518 | 1914 | struct sock_xprt *new; |
c8541ecd | 1915 | |
96802a09 | 1916 | if (args->addrlen > sizeof(xprt->addr)) { |
46121cf7 | 1917 | dprintk("RPC: xs_setup_xprt: address too large\n"); |
c8541ecd CL |
1918 | return ERR_PTR(-EBADF); |
1919 | } | |
1920 | ||
ffc2e518 CL |
1921 | new = kzalloc(sizeof(*new), GFP_KERNEL); |
1922 | if (new == NULL) { | |
46121cf7 CL |
1923 | dprintk("RPC: xs_setup_xprt: couldn't allocate " |
1924 | "rpc_xprt\n"); | |
c8541ecd CL |
1925 | return ERR_PTR(-ENOMEM); |
1926 | } | |
ffc2e518 | 1927 | xprt = &new->xprt; |
c8541ecd CL |
1928 | |
1929 | xprt->max_reqs = slot_table_size; | |
1930 | xprt->slot = kcalloc(xprt->max_reqs, sizeof(struct rpc_rqst), GFP_KERNEL); | |
1931 | if (xprt->slot == NULL) { | |
1932 | kfree(xprt); | |
46121cf7 CL |
1933 | dprintk("RPC: xs_setup_xprt: couldn't allocate slot " |
1934 | "table\n"); | |
c8541ecd CL |
1935 | return ERR_PTR(-ENOMEM); |
1936 | } | |
1937 | ||
96802a09 FM |
1938 | memcpy(&xprt->addr, args->dstaddr, args->addrlen); |
1939 | xprt->addrlen = args->addrlen; | |
d3bc9a1d FM |
1940 | if (args->srcaddr) |
1941 | memcpy(&new->addr, args->srcaddr, args->addrlen); | |
c8541ecd CL |
1942 | |
1943 | return xprt; | |
1944 | } | |
1945 | ||
2881ae74 TM |
1946 | static const struct rpc_timeout xs_udp_default_timeout = { |
1947 | .to_initval = 5 * HZ, | |
1948 | .to_maxval = 30 * HZ, | |
1949 | .to_increment = 5 * HZ, | |
1950 | .to_retries = 5, | |
1951 | }; | |
1952 | ||
9903cd1c CL |
1953 | /** |
1954 | * xs_setup_udp - Set up transport to use a UDP socket | |
96802a09 | 1955 | * @args: rpc transport creation arguments |
9903cd1c CL |
1956 | * |
1957 | */ | |
483066d6 | 1958 | static struct rpc_xprt *xs_setup_udp(struct xprt_create *args) |
a246b010 | 1959 | { |
8f9d5b1a | 1960 | struct sockaddr *addr = args->dstaddr; |
c8541ecd | 1961 | struct rpc_xprt *xprt; |
c8475461 | 1962 | struct sock_xprt *transport; |
a246b010 | 1963 | |
96802a09 | 1964 | xprt = xs_setup_xprt(args, xprt_udp_slot_table_entries); |
c8541ecd CL |
1965 | if (IS_ERR(xprt)) |
1966 | return xprt; | |
c8475461 | 1967 | transport = container_of(xprt, struct sock_xprt, xprt); |
a246b010 | 1968 | |
ec739ef0 | 1969 | xprt->prot = IPPROTO_UDP; |
808012fb | 1970 | xprt->tsh_size = 0; |
a246b010 CL |
1971 | /* XXX: header size can vary due to auth type, IPv6, etc. */ |
1972 | xprt->max_payload = (1U << 16) - (MAX_HEADER << 3); | |
1973 | ||
03bf4b70 CL |
1974 | xprt->bind_timeout = XS_BIND_TO; |
1975 | xprt->connect_timeout = XS_UDP_CONN_TO; | |
1976 | xprt->reestablish_timeout = XS_UDP_REEST_TO; | |
1977 | xprt->idle_timeout = XS_IDLE_DISC_TO; | |
a246b010 | 1978 | |
262965f5 | 1979 | xprt->ops = &xs_udp_ops; |
a246b010 | 1980 | |
ba7392bb | 1981 | xprt->timeout = &xs_udp_default_timeout; |
a246b010 | 1982 | |
8f9d5b1a CL |
1983 | switch (addr->sa_family) { |
1984 | case AF_INET: | |
1985 | if (((struct sockaddr_in *)addr)->sin_port != htons(0)) | |
1986 | xprt_set_bound(xprt); | |
1987 | ||
1988 | INIT_DELAYED_WORK(&transport->connect_worker, | |
1989 | xs_udp_connect_worker4); | |
b454ae90 | 1990 | xs_format_ipv4_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP); |
8f9d5b1a CL |
1991 | break; |
1992 | case AF_INET6: | |
1993 | if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0)) | |
1994 | xprt_set_bound(xprt); | |
1995 | ||
1996 | INIT_DELAYED_WORK(&transport->connect_worker, | |
1997 | xs_udp_connect_worker6); | |
b454ae90 | 1998 | xs_format_ipv6_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP6); |
8f9d5b1a CL |
1999 | break; |
2000 | default: | |
2001 | kfree(xprt); | |
2002 | return ERR_PTR(-EAFNOSUPPORT); | |
2003 | } | |
2004 | ||
46121cf7 | 2005 | dprintk("RPC: set up transport to address %s\n", |
7559c7a2 | 2006 | xprt->address_strings[RPC_DISPLAY_ALL]); |
edb267a6 | 2007 | |
bc25571e TT |
2008 | if (try_module_get(THIS_MODULE)) |
2009 | return xprt; | |
2010 | ||
2011 | kfree(xprt->slot); | |
2012 | kfree(xprt); | |
2013 | return ERR_PTR(-EINVAL); | |
a246b010 CL |
2014 | } |
2015 | ||
2881ae74 TM |
2016 | static const struct rpc_timeout xs_tcp_default_timeout = { |
2017 | .to_initval = 60 * HZ, | |
2018 | .to_maxval = 60 * HZ, | |
2019 | .to_retries = 2, | |
2020 | }; | |
2021 | ||
9903cd1c CL |
2022 | /** |
2023 | * xs_setup_tcp - Set up transport to use a TCP socket | |
96802a09 | 2024 | * @args: rpc transport creation arguments |
9903cd1c CL |
2025 | * |
2026 | */ | |
483066d6 | 2027 | static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args) |
a246b010 | 2028 | { |
8f9d5b1a | 2029 | struct sockaddr *addr = args->dstaddr; |
c8541ecd | 2030 | struct rpc_xprt *xprt; |
c8475461 | 2031 | struct sock_xprt *transport; |
a246b010 | 2032 | |
96802a09 | 2033 | xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries); |
c8541ecd CL |
2034 | if (IS_ERR(xprt)) |
2035 | return xprt; | |
c8475461 | 2036 | transport = container_of(xprt, struct sock_xprt, xprt); |
a246b010 | 2037 | |
ec739ef0 | 2038 | xprt->prot = IPPROTO_TCP; |
808012fb | 2039 | xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32); |
808012fb | 2040 | xprt->max_payload = RPC_MAX_FRAGMENT_SIZE; |
a246b010 | 2041 | |
03bf4b70 CL |
2042 | xprt->bind_timeout = XS_BIND_TO; |
2043 | xprt->connect_timeout = XS_TCP_CONN_TO; | |
2044 | xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; | |
2045 | xprt->idle_timeout = XS_IDLE_DISC_TO; | |
a246b010 | 2046 | |
262965f5 | 2047 | xprt->ops = &xs_tcp_ops; |
ba7392bb | 2048 | xprt->timeout = &xs_tcp_default_timeout; |
a246b010 | 2049 | |
8f9d5b1a CL |
2050 | switch (addr->sa_family) { |
2051 | case AF_INET: | |
2052 | if (((struct sockaddr_in *)addr)->sin_port != htons(0)) | |
2053 | xprt_set_bound(xprt); | |
2054 | ||
2055 | INIT_DELAYED_WORK(&transport->connect_worker, xs_tcp_connect_worker4); | |
b454ae90 | 2056 | xs_format_ipv4_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP); |
8f9d5b1a CL |
2057 | break; |
2058 | case AF_INET6: | |
2059 | if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0)) | |
2060 | xprt_set_bound(xprt); | |
2061 | ||
2062 | INIT_DELAYED_WORK(&transport->connect_worker, xs_tcp_connect_worker6); | |
b454ae90 | 2063 | xs_format_ipv6_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP6); |
8f9d5b1a CL |
2064 | break; |
2065 | default: | |
2066 | kfree(xprt); | |
2067 | return ERR_PTR(-EAFNOSUPPORT); | |
2068 | } | |
2069 | ||
46121cf7 | 2070 | dprintk("RPC: set up transport to address %s\n", |
7559c7a2 | 2071 | xprt->address_strings[RPC_DISPLAY_ALL]); |
edb267a6 | 2072 | |
bc25571e TT |
2073 | if (try_module_get(THIS_MODULE)) |
2074 | return xprt; | |
2075 | ||
2076 | kfree(xprt->slot); | |
2077 | kfree(xprt); | |
2078 | return ERR_PTR(-EINVAL); | |
a246b010 | 2079 | } |
282b32e1 | 2080 | |
bc25571e TT |
2081 | static struct xprt_class xs_udp_transport = { |
2082 | .list = LIST_HEAD_INIT(xs_udp_transport.list), | |
2083 | .name = "udp", | |
2084 | .owner = THIS_MODULE, | |
4fa016eb | 2085 | .ident = IPPROTO_UDP, |
bc25571e TT |
2086 | .setup = xs_setup_udp, |
2087 | }; | |
2088 | ||
2089 | static struct xprt_class xs_tcp_transport = { | |
2090 | .list = LIST_HEAD_INIT(xs_tcp_transport.list), | |
2091 | .name = "tcp", | |
2092 | .owner = THIS_MODULE, | |
4fa016eb | 2093 | .ident = IPPROTO_TCP, |
bc25571e TT |
2094 | .setup = xs_setup_tcp, |
2095 | }; | |
2096 | ||
282b32e1 | 2097 | /** |
bc25571e | 2098 | * init_socket_xprt - set up xprtsock's sysctls, register with RPC client |
282b32e1 CL |
2099 | * |
2100 | */ | |
2101 | int init_socket_xprt(void) | |
2102 | { | |
fbf76683 | 2103 | #ifdef RPC_DEBUG |
2b1bec5f | 2104 | if (!sunrpc_table_header) |
0b4d4147 | 2105 | sunrpc_table_header = register_sysctl_table(sunrpc_table); |
fbf76683 CL |
2106 | #endif |
2107 | ||
bc25571e TT |
2108 | xprt_register_transport(&xs_udp_transport); |
2109 | xprt_register_transport(&xs_tcp_transport); | |
2110 | ||
282b32e1 CL |
2111 | return 0; |
2112 | } | |
2113 | ||
2114 | /** | |
bc25571e | 2115 | * cleanup_socket_xprt - remove xprtsock's sysctls, unregister |
282b32e1 CL |
2116 | * |
2117 | */ | |
2118 | void cleanup_socket_xprt(void) | |
2119 | { | |
fbf76683 CL |
2120 | #ifdef RPC_DEBUG |
2121 | if (sunrpc_table_header) { | |
2122 | unregister_sysctl_table(sunrpc_table_header); | |
2123 | sunrpc_table_header = NULL; | |
2124 | } | |
2125 | #endif | |
bc25571e TT |
2126 | |
2127 | xprt_unregister_transport(&xs_udp_transport); | |
2128 | xprt_unregister_transport(&xs_tcp_transport); | |
282b32e1 | 2129 | } |