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