]>
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> | |
176e21ee | 22 | #include <linux/string.h> |
a246b010 | 23 | #include <linux/slab.h> |
bc25571e | 24 | #include <linux/module.h> |
a246b010 | 25 | #include <linux/capability.h> |
a246b010 CL |
26 | #include <linux/pagemap.h> |
27 | #include <linux/errno.h> | |
28 | #include <linux/socket.h> | |
29 | #include <linux/in.h> | |
30 | #include <linux/net.h> | |
31 | #include <linux/mm.h> | |
176e21ee | 32 | #include <linux/un.h> |
a246b010 CL |
33 | #include <linux/udp.h> |
34 | #include <linux/tcp.h> | |
35 | #include <linux/sunrpc/clnt.h> | |
5976687a | 36 | #include <linux/sunrpc/addr.h> |
02107148 | 37 | #include <linux/sunrpc/sched.h> |
4cfc7e60 | 38 | #include <linux/sunrpc/svcsock.h> |
49c36fcc | 39 | #include <linux/sunrpc/xprtsock.h> |
a246b010 | 40 | #include <linux/file.h> |
9e00abc3 | 41 | #ifdef CONFIG_SUNRPC_BACKCHANNEL |
44b98efd RL |
42 | #include <linux/sunrpc/bc_xprt.h> |
43 | #endif | |
a246b010 CL |
44 | |
45 | #include <net/sock.h> | |
46 | #include <net/checksum.h> | |
47 | #include <net/udp.h> | |
48 | #include <net/tcp.h> | |
49 | ||
40b5ea0c TM |
50 | #include <trace/events/sunrpc.h> |
51 | ||
4cfc7e60 | 52 | #include "sunrpc.h" |
176e21ee CL |
53 | |
54 | static void xs_close(struct rpc_xprt *xprt); | |
55 | ||
c556b754 CL |
56 | /* |
57 | * xprtsock tunables | |
58 | */ | |
09acfea5 TM |
59 | static unsigned int xprt_udp_slot_table_entries = RPC_DEF_SLOT_TABLE; |
60 | static unsigned int xprt_tcp_slot_table_entries = RPC_MIN_SLOT_TABLE; | |
61 | static unsigned int xprt_max_tcp_slot_table_entries = RPC_MAX_SLOT_TABLE; | |
c556b754 | 62 | |
09acfea5 TM |
63 | static unsigned int xprt_min_resvport = RPC_DEF_MIN_RESVPORT; |
64 | static unsigned int xprt_max_resvport = RPC_DEF_MAX_RESVPORT; | |
c556b754 | 65 | |
7d1e8255 | 66 | #define XS_TCP_LINGER_TO (15U * HZ) |
25fe6142 | 67 | static unsigned int xs_tcp_fin_timeout __read_mostly = XS_TCP_LINGER_TO; |
7d1e8255 | 68 | |
fbf76683 CL |
69 | /* |
70 | * We can register our own files under /proc/sys/sunrpc by | |
71 | * calling register_sysctl_table() again. The files in that | |
72 | * directory become the union of all files registered there. | |
73 | * | |
74 | * We simply need to make sure that we don't collide with | |
75 | * someone else's file names! | |
76 | */ | |
77 | ||
78 | #ifdef RPC_DEBUG | |
79 | ||
80 | static unsigned int min_slot_table_size = RPC_MIN_SLOT_TABLE; | |
81 | static unsigned int max_slot_table_size = RPC_MAX_SLOT_TABLE; | |
d9ba131d | 82 | static unsigned int max_tcp_slot_table_limit = RPC_MAX_SLOT_TABLE_LIMIT; |
fbf76683 CL |
83 | static unsigned int xprt_min_resvport_limit = RPC_MIN_RESVPORT; |
84 | static unsigned int xprt_max_resvport_limit = RPC_MAX_RESVPORT; | |
85 | ||
86 | static struct ctl_table_header *sunrpc_table_header; | |
87 | ||
88 | /* | |
89 | * FIXME: changing the UDP slot table size should also resize the UDP | |
90 | * socket buffers for existing UDP transports | |
91 | */ | |
fe2c6338 | 92 | static struct ctl_table xs_tunables_table[] = { |
fbf76683 | 93 | { |
fbf76683 CL |
94 | .procname = "udp_slot_table_entries", |
95 | .data = &xprt_udp_slot_table_entries, | |
96 | .maxlen = sizeof(unsigned int), | |
97 | .mode = 0644, | |
6d456111 | 98 | .proc_handler = proc_dointvec_minmax, |
fbf76683 CL |
99 | .extra1 = &min_slot_table_size, |
100 | .extra2 = &max_slot_table_size | |
101 | }, | |
102 | { | |
fbf76683 CL |
103 | .procname = "tcp_slot_table_entries", |
104 | .data = &xprt_tcp_slot_table_entries, | |
105 | .maxlen = sizeof(unsigned int), | |
106 | .mode = 0644, | |
6d456111 | 107 | .proc_handler = proc_dointvec_minmax, |
fbf76683 CL |
108 | .extra1 = &min_slot_table_size, |
109 | .extra2 = &max_slot_table_size | |
110 | }, | |
d9ba131d TM |
111 | { |
112 | .procname = "tcp_max_slot_table_entries", | |
113 | .data = &xprt_max_tcp_slot_table_entries, | |
114 | .maxlen = sizeof(unsigned int), | |
115 | .mode = 0644, | |
116 | .proc_handler = proc_dointvec_minmax, | |
117 | .extra1 = &min_slot_table_size, | |
118 | .extra2 = &max_tcp_slot_table_limit | |
119 | }, | |
fbf76683 | 120 | { |
fbf76683 CL |
121 | .procname = "min_resvport", |
122 | .data = &xprt_min_resvport, | |
123 | .maxlen = sizeof(unsigned int), | |
124 | .mode = 0644, | |
6d456111 | 125 | .proc_handler = proc_dointvec_minmax, |
fbf76683 CL |
126 | .extra1 = &xprt_min_resvport_limit, |
127 | .extra2 = &xprt_max_resvport_limit | |
128 | }, | |
129 | { | |
fbf76683 CL |
130 | .procname = "max_resvport", |
131 | .data = &xprt_max_resvport, | |
132 | .maxlen = sizeof(unsigned int), | |
133 | .mode = 0644, | |
6d456111 | 134 | .proc_handler = proc_dointvec_minmax, |
fbf76683 CL |
135 | .extra1 = &xprt_min_resvport_limit, |
136 | .extra2 = &xprt_max_resvport_limit | |
137 | }, | |
25fe6142 TM |
138 | { |
139 | .procname = "tcp_fin_timeout", | |
140 | .data = &xs_tcp_fin_timeout, | |
141 | .maxlen = sizeof(xs_tcp_fin_timeout), | |
142 | .mode = 0644, | |
6d456111 | 143 | .proc_handler = proc_dointvec_jiffies, |
fbf76683 | 144 | }, |
f8572d8f | 145 | { }, |
fbf76683 CL |
146 | }; |
147 | ||
fe2c6338 | 148 | static struct ctl_table sunrpc_table[] = { |
fbf76683 | 149 | { |
fbf76683 CL |
150 | .procname = "sunrpc", |
151 | .mode = 0555, | |
152 | .child = xs_tunables_table | |
153 | }, | |
f8572d8f | 154 | { }, |
fbf76683 CL |
155 | }; |
156 | ||
157 | #endif | |
158 | ||
03bf4b70 CL |
159 | /* |
160 | * Wait duration for a reply from the RPC portmapper. | |
161 | */ | |
162 | #define XS_BIND_TO (60U * HZ) | |
163 | ||
164 | /* | |
165 | * Delay if a UDP socket connect error occurs. This is most likely some | |
166 | * kind of resource problem on the local host. | |
167 | */ | |
168 | #define XS_UDP_REEST_TO (2U * HZ) | |
169 | ||
170 | /* | |
171 | * The reestablish timeout allows clients to delay for a bit before attempting | |
172 | * to reconnect to a server that just dropped our connection. | |
173 | * | |
174 | * We implement an exponential backoff when trying to reestablish a TCP | |
175 | * transport connection with the server. Some servers like to drop a TCP | |
176 | * connection when they are overworked, so we start with a short timeout and | |
177 | * increase over time if the server is down or not responding. | |
178 | */ | |
179 | #define XS_TCP_INIT_REEST_TO (3U * HZ) | |
180 | #define XS_TCP_MAX_REEST_TO (5U * 60 * HZ) | |
181 | ||
182 | /* | |
183 | * TCP idle timeout; client drops the transport socket if it is idle | |
184 | * for this long. Note that we also timeout UDP sockets to prevent | |
185 | * holding port numbers when there is no RPC traffic. | |
186 | */ | |
187 | #define XS_IDLE_DISC_TO (5U * 60 * HZ) | |
188 | ||
a246b010 CL |
189 | #ifdef RPC_DEBUG |
190 | # undef RPC_DEBUG_DATA | |
9903cd1c | 191 | # define RPCDBG_FACILITY RPCDBG_TRANS |
a246b010 CL |
192 | #endif |
193 | ||
a246b010 | 194 | #ifdef RPC_DEBUG_DATA |
9903cd1c | 195 | static void xs_pktdump(char *msg, u32 *packet, unsigned int count) |
a246b010 | 196 | { |
9903cd1c CL |
197 | u8 *buf = (u8 *) packet; |
198 | int j; | |
a246b010 | 199 | |
46121cf7 | 200 | dprintk("RPC: %s\n", msg); |
a246b010 CL |
201 | for (j = 0; j < count && j < 128; j += 4) { |
202 | if (!(j & 31)) { | |
203 | if (j) | |
204 | dprintk("\n"); | |
205 | dprintk("0x%04x ", j); | |
206 | } | |
207 | dprintk("%02x%02x%02x%02x ", | |
208 | buf[j], buf[j+1], buf[j+2], buf[j+3]); | |
209 | } | |
210 | dprintk("\n"); | |
211 | } | |
212 | #else | |
9903cd1c | 213 | static inline void xs_pktdump(char *msg, u32 *packet, unsigned int count) |
a246b010 CL |
214 | { |
215 | /* NOP */ | |
216 | } | |
217 | #endif | |
218 | ||
ffc2e518 CL |
219 | struct sock_xprt { |
220 | struct rpc_xprt xprt; | |
ee0ac0c2 CL |
221 | |
222 | /* | |
223 | * Network layer | |
224 | */ | |
225 | struct socket * sock; | |
226 | struct sock * inet; | |
51971139 CL |
227 | |
228 | /* | |
229 | * State of TCP reply receive | |
230 | */ | |
231 | __be32 tcp_fraghdr, | |
b76ce561 TM |
232 | tcp_xid, |
233 | tcp_calldir; | |
51971139 CL |
234 | |
235 | u32 tcp_offset, | |
236 | tcp_reclen; | |
237 | ||
238 | unsigned long tcp_copied, | |
239 | tcp_flags; | |
c8475461 CL |
240 | |
241 | /* | |
242 | * Connection of transports | |
243 | */ | |
34161db6 | 244 | struct delayed_work connect_worker; |
fbfffbd5 CL |
245 | struct sockaddr_storage srcaddr; |
246 | unsigned short srcport; | |
7c6e066e CL |
247 | |
248 | /* | |
249 | * UDP socket buffer size parameters | |
250 | */ | |
251 | size_t rcvsize, | |
252 | sndsize; | |
314dfd79 CL |
253 | |
254 | /* | |
255 | * Saved socket callback addresses | |
256 | */ | |
676d2369 | 257 | void (*old_data_ready)(struct sock *); |
314dfd79 CL |
258 | void (*old_state_change)(struct sock *); |
259 | void (*old_write_space)(struct sock *); | |
2118071d | 260 | void (*old_error_report)(struct sock *); |
ffc2e518 CL |
261 | }; |
262 | ||
e136d092 CL |
263 | /* |
264 | * TCP receive state flags | |
265 | */ | |
266 | #define TCP_RCV_LAST_FRAG (1UL << 0) | |
267 | #define TCP_RCV_COPY_FRAGHDR (1UL << 1) | |
268 | #define TCP_RCV_COPY_XID (1UL << 2) | |
269 | #define TCP_RCV_COPY_DATA (1UL << 3) | |
f4a2e418 RL |
270 | #define TCP_RCV_READ_CALLDIR (1UL << 4) |
271 | #define TCP_RCV_COPY_CALLDIR (1UL << 5) | |
18dca02a RL |
272 | |
273 | /* | |
274 | * TCP RPC flags | |
275 | */ | |
f4a2e418 | 276 | #define TCP_RPC_REPLY (1UL << 6) |
e136d092 | 277 | |
2118071d TM |
278 | static inline struct rpc_xprt *xprt_from_sock(struct sock *sk) |
279 | { | |
280 | return (struct rpc_xprt *) sk->sk_user_data; | |
281 | } | |
282 | ||
95392c59 CL |
283 | static inline struct sockaddr *xs_addr(struct rpc_xprt *xprt) |
284 | { | |
285 | return (struct sockaddr *) &xprt->addr; | |
286 | } | |
287 | ||
176e21ee CL |
288 | static inline struct sockaddr_un *xs_addr_un(struct rpc_xprt *xprt) |
289 | { | |
290 | return (struct sockaddr_un *) &xprt->addr; | |
291 | } | |
292 | ||
95392c59 | 293 | static inline struct sockaddr_in *xs_addr_in(struct rpc_xprt *xprt) |
edb267a6 | 294 | { |
95392c59 CL |
295 | return (struct sockaddr_in *) &xprt->addr; |
296 | } | |
297 | ||
298 | static inline struct sockaddr_in6 *xs_addr_in6(struct rpc_xprt *xprt) | |
299 | { | |
300 | return (struct sockaddr_in6 *) &xprt->addr; | |
301 | } | |
302 | ||
c877b849 | 303 | static void xs_format_common_peer_addresses(struct rpc_xprt *xprt) |
edb267a6 | 304 | { |
c877b849 | 305 | struct sockaddr *sap = xs_addr(xprt); |
9dc3b095 CL |
306 | struct sockaddr_in6 *sin6; |
307 | struct sockaddr_in *sin; | |
176e21ee | 308 | struct sockaddr_un *sun; |
c877b849 | 309 | char buf[128]; |
edb267a6 | 310 | |
9dc3b095 | 311 | switch (sap->sa_family) { |
176e21ee CL |
312 | case AF_LOCAL: |
313 | sun = xs_addr_un(xprt); | |
314 | strlcpy(buf, sun->sun_path, sizeof(buf)); | |
315 | xprt->address_strings[RPC_DISPLAY_ADDR] = | |
316 | kstrdup(buf, GFP_KERNEL); | |
317 | break; | |
9dc3b095 | 318 | case AF_INET: |
176e21ee CL |
319 | (void)rpc_ntop(sap, buf, sizeof(buf)); |
320 | xprt->address_strings[RPC_DISPLAY_ADDR] = | |
321 | kstrdup(buf, GFP_KERNEL); | |
9dc3b095 | 322 | sin = xs_addr_in(xprt); |
fc0b5791 | 323 | snprintf(buf, sizeof(buf), "%08x", ntohl(sin->sin_addr.s_addr)); |
9dc3b095 CL |
324 | break; |
325 | case AF_INET6: | |
176e21ee CL |
326 | (void)rpc_ntop(sap, buf, sizeof(buf)); |
327 | xprt->address_strings[RPC_DISPLAY_ADDR] = | |
328 | kstrdup(buf, GFP_KERNEL); | |
9dc3b095 | 329 | sin6 = xs_addr_in6(xprt); |
fc0b5791 | 330 | snprintf(buf, sizeof(buf), "%pi6", &sin6->sin6_addr); |
9dc3b095 CL |
331 | break; |
332 | default: | |
333 | BUG(); | |
756805e7 | 334 | } |
176e21ee | 335 | |
9dc3b095 | 336 | xprt->address_strings[RPC_DISPLAY_HEX_ADDR] = kstrdup(buf, GFP_KERNEL); |
edb267a6 CL |
337 | } |
338 | ||
9dc3b095 | 339 | static void xs_format_common_peer_ports(struct rpc_xprt *xprt) |
4b6473fb | 340 | { |
9dc3b095 CL |
341 | struct sockaddr *sap = xs_addr(xprt); |
342 | char buf[128]; | |
4b6473fb | 343 | |
81160e66 | 344 | snprintf(buf, sizeof(buf), "%u", rpc_get_port(sap)); |
c877b849 | 345 | xprt->address_strings[RPC_DISPLAY_PORT] = kstrdup(buf, GFP_KERNEL); |
4b6473fb | 346 | |
81160e66 | 347 | snprintf(buf, sizeof(buf), "%4hx", rpc_get_port(sap)); |
c877b849 CL |
348 | xprt->address_strings[RPC_DISPLAY_HEX_PORT] = kstrdup(buf, GFP_KERNEL); |
349 | } | |
4b6473fb | 350 | |
9dc3b095 CL |
351 | static void xs_format_peer_addresses(struct rpc_xprt *xprt, |
352 | const char *protocol, | |
353 | const char *netid) | |
c877b849 | 354 | { |
b454ae90 | 355 | xprt->address_strings[RPC_DISPLAY_PROTO] = protocol; |
b454ae90 | 356 | xprt->address_strings[RPC_DISPLAY_NETID] = netid; |
c877b849 | 357 | xs_format_common_peer_addresses(xprt); |
9dc3b095 | 358 | xs_format_common_peer_ports(xprt); |
edb267a6 | 359 | } |
4b6473fb | 360 | |
9dc3b095 | 361 | static void xs_update_peer_port(struct rpc_xprt *xprt) |
4b6473fb | 362 | { |
9dc3b095 CL |
363 | kfree(xprt->address_strings[RPC_DISPLAY_HEX_PORT]); |
364 | kfree(xprt->address_strings[RPC_DISPLAY_PORT]); | |
4417c8c4 | 365 | |
9dc3b095 | 366 | xs_format_common_peer_ports(xprt); |
edb267a6 CL |
367 | } |
368 | ||
369 | static void xs_free_peer_addresses(struct rpc_xprt *xprt) | |
370 | { | |
33e01dc7 CL |
371 | unsigned int i; |
372 | ||
373 | for (i = 0; i < RPC_DISPLAY_MAX; i++) | |
374 | switch (i) { | |
375 | case RPC_DISPLAY_PROTO: | |
376 | case RPC_DISPLAY_NETID: | |
377 | continue; | |
378 | default: | |
379 | kfree(xprt->address_strings[i]); | |
380 | } | |
edb267a6 CL |
381 | } |
382 | ||
b4b5cc85 CL |
383 | #define XS_SENDMSG_FLAGS (MSG_DONTWAIT | MSG_NOSIGNAL) |
384 | ||
24c5684b | 385 | static int xs_send_kvec(struct socket *sock, struct sockaddr *addr, int addrlen, struct kvec *vec, unsigned int base, int more) |
b4b5cc85 | 386 | { |
b4b5cc85 CL |
387 | struct msghdr msg = { |
388 | .msg_name = addr, | |
389 | .msg_namelen = addrlen, | |
24c5684b TM |
390 | .msg_flags = XS_SENDMSG_FLAGS | (more ? MSG_MORE : 0), |
391 | }; | |
392 | struct kvec iov = { | |
393 | .iov_base = vec->iov_base + base, | |
394 | .iov_len = vec->iov_len - base, | |
b4b5cc85 CL |
395 | }; |
396 | ||
24c5684b | 397 | if (iov.iov_len != 0) |
b4b5cc85 CL |
398 | return kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len); |
399 | return kernel_sendmsg(sock, &msg, NULL, 0, 0); | |
400 | } | |
401 | ||
a6b31d18 | 402 | static int xs_send_pagedata(struct socket *sock, struct xdr_buf *xdr, unsigned int base, int more, bool zerocopy) |
b4b5cc85 | 403 | { |
a6b31d18 TM |
404 | ssize_t (*do_sendpage)(struct socket *sock, struct page *page, |
405 | int offset, size_t size, int flags); | |
24c5684b TM |
406 | struct page **ppage; |
407 | unsigned int remainder; | |
408 | int err, sent = 0; | |
409 | ||
410 | remainder = xdr->page_len - base; | |
411 | base += xdr->page_base; | |
412 | ppage = xdr->pages + (base >> PAGE_SHIFT); | |
413 | base &= ~PAGE_MASK; | |
a6b31d18 TM |
414 | do_sendpage = sock->ops->sendpage; |
415 | if (!zerocopy) | |
416 | do_sendpage = sock_no_sendpage; | |
24c5684b TM |
417 | for(;;) { |
418 | unsigned int len = min_t(unsigned int, PAGE_SIZE - base, remainder); | |
419 | int flags = XS_SENDMSG_FLAGS; | |
b4b5cc85 | 420 | |
24c5684b TM |
421 | remainder -= len; |
422 | if (remainder != 0 || more) | |
423 | flags |= MSG_MORE; | |
a6b31d18 | 424 | err = do_sendpage(sock, *ppage, base, len, flags); |
24c5684b TM |
425 | if (remainder == 0 || err != len) |
426 | break; | |
427 | sent += err; | |
428 | ppage++; | |
429 | base = 0; | |
430 | } | |
431 | if (sent == 0) | |
432 | return err; | |
433 | if (err > 0) | |
434 | sent += err; | |
435 | return sent; | |
b4b5cc85 CL |
436 | } |
437 | ||
9903cd1c CL |
438 | /** |
439 | * xs_sendpages - write pages directly to a socket | |
440 | * @sock: socket to send on | |
441 | * @addr: UDP only -- address of destination | |
442 | * @addrlen: UDP only -- length of destination address | |
443 | * @xdr: buffer containing this request | |
444 | * @base: starting position in the buffer | |
a6b31d18 | 445 | * @zerocopy: true if it is safe to use sendpage() |
9903cd1c | 446 | * |
a246b010 | 447 | */ |
a6b31d18 | 448 | static int xs_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen, struct xdr_buf *xdr, unsigned int base, bool zerocopy) |
a246b010 | 449 | { |
24c5684b TM |
450 | unsigned int remainder = xdr->len - base; |
451 | int err, sent = 0; | |
a246b010 | 452 | |
262965f5 | 453 | if (unlikely(!sock)) |
fba91afb | 454 | return -ENOTSOCK; |
262965f5 CL |
455 | |
456 | clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags); | |
24c5684b TM |
457 | if (base != 0) { |
458 | addr = NULL; | |
459 | addrlen = 0; | |
460 | } | |
262965f5 | 461 | |
24c5684b TM |
462 | if (base < xdr->head[0].iov_len || addr != NULL) { |
463 | unsigned int len = xdr->head[0].iov_len - base; | |
464 | remainder -= len; | |
465 | err = xs_send_kvec(sock, addr, addrlen, &xdr->head[0], base, remainder != 0); | |
466 | if (remainder == 0 || err != len) | |
a246b010 | 467 | goto out; |
24c5684b | 468 | sent += err; |
a246b010 CL |
469 | base = 0; |
470 | } else | |
24c5684b | 471 | base -= xdr->head[0].iov_len; |
a246b010 | 472 | |
24c5684b TM |
473 | if (base < xdr->page_len) { |
474 | unsigned int len = xdr->page_len - base; | |
475 | remainder -= len; | |
a6b31d18 | 476 | err = xs_send_pagedata(sock, xdr, base, remainder != 0, zerocopy); |
24c5684b | 477 | if (remainder == 0 || err != len) |
a246b010 | 478 | goto out; |
24c5684b | 479 | sent += err; |
a246b010 | 480 | base = 0; |
24c5684b TM |
481 | } else |
482 | base -= xdr->page_len; | |
483 | ||
484 | if (base >= xdr->tail[0].iov_len) | |
485 | return sent; | |
486 | err = xs_send_kvec(sock, NULL, 0, &xdr->tail[0], base, 0); | |
a246b010 | 487 | out: |
24c5684b TM |
488 | if (sent == 0) |
489 | return err; | |
490 | if (err > 0) | |
491 | sent += err; | |
492 | return sent; | |
a246b010 CL |
493 | } |
494 | ||
b6ddf64f TM |
495 | static void xs_nospace_callback(struct rpc_task *task) |
496 | { | |
497 | struct sock_xprt *transport = container_of(task->tk_rqstp->rq_xprt, struct sock_xprt, xprt); | |
498 | ||
499 | transport->inet->sk_write_pending--; | |
500 | clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); | |
501 | } | |
502 | ||
9903cd1c | 503 | /** |
262965f5 CL |
504 | * xs_nospace - place task on wait queue if transmit was incomplete |
505 | * @task: task to put to sleep | |
9903cd1c | 506 | * |
a246b010 | 507 | */ |
5e3771ce | 508 | static int xs_nospace(struct rpc_task *task) |
a246b010 | 509 | { |
262965f5 CL |
510 | struct rpc_rqst *req = task->tk_rqstp; |
511 | struct rpc_xprt *xprt = req->rq_xprt; | |
ee0ac0c2 | 512 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
06ea0bfe | 513 | struct sock *sk = transport->inet; |
24ca9a84 | 514 | int ret = -EAGAIN; |
a246b010 | 515 | |
46121cf7 | 516 | dprintk("RPC: %5u xmit incomplete (%u left of %u)\n", |
262965f5 CL |
517 | task->tk_pid, req->rq_slen - req->rq_bytes_sent, |
518 | req->rq_slen); | |
519 | ||
b6ddf64f TM |
520 | /* Protect against races with write_space */ |
521 | spin_lock_bh(&xprt->transport_lock); | |
522 | ||
523 | /* Don't race with disconnect */ | |
524 | if (xprt_connected(xprt)) { | |
525 | if (test_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags)) { | |
526 | /* | |
527 | * Notify TCP that we're limited by the application | |
528 | * window size | |
529 | */ | |
530 | set_bit(SOCK_NOSPACE, &transport->sock->flags); | |
06ea0bfe | 531 | sk->sk_write_pending++; |
b6ddf64f TM |
532 | /* ...and wait for more buffer space */ |
533 | xprt_wait_for_buffer_space(task, xs_nospace_callback); | |
534 | } | |
535 | } else { | |
536 | clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); | |
5e3771ce | 537 | ret = -ENOTCONN; |
b6ddf64f | 538 | } |
262965f5 | 539 | |
b6ddf64f | 540 | spin_unlock_bh(&xprt->transport_lock); |
06ea0bfe TM |
541 | |
542 | /* Race breaker in case memory is freed before above code is called */ | |
543 | sk->sk_write_space(sk); | |
5e3771ce | 544 | return ret; |
262965f5 CL |
545 | } |
546 | ||
61677eee CL |
547 | /* |
548 | * Construct a stream transport record marker in @buf. | |
549 | */ | |
550 | static inline void xs_encode_stream_record_marker(struct xdr_buf *buf) | |
551 | { | |
552 | u32 reclen = buf->len - sizeof(rpc_fraghdr); | |
553 | rpc_fraghdr *base = buf->head[0].iov_base; | |
554 | *base = cpu_to_be32(RPC_LAST_STREAM_FRAGMENT | reclen); | |
555 | } | |
556 | ||
176e21ee CL |
557 | /** |
558 | * xs_local_send_request - write an RPC request to an AF_LOCAL socket | |
559 | * @task: RPC task that manages the state of an RPC request | |
560 | * | |
561 | * Return values: | |
562 | * 0: The request has been sent | |
563 | * EAGAIN: The socket was blocked, please call again later to | |
564 | * complete the request | |
565 | * ENOTCONN: Caller needs to invoke connect logic then call again | |
566 | * other: Some other error occured, the request was not sent | |
567 | */ | |
568 | static int xs_local_send_request(struct rpc_task *task) | |
569 | { | |
570 | struct rpc_rqst *req = task->tk_rqstp; | |
571 | struct rpc_xprt *xprt = req->rq_xprt; | |
572 | struct sock_xprt *transport = | |
573 | container_of(xprt, struct sock_xprt, xprt); | |
574 | struct xdr_buf *xdr = &req->rq_snd_buf; | |
575 | int status; | |
576 | ||
577 | xs_encode_stream_record_marker(&req->rq_snd_buf); | |
578 | ||
579 | xs_pktdump("packet data:", | |
580 | req->rq_svec->iov_base, req->rq_svec->iov_len); | |
581 | ||
582 | status = xs_sendpages(transport->sock, NULL, 0, | |
a6b31d18 | 583 | xdr, req->rq_bytes_sent, true); |
176e21ee CL |
584 | dprintk("RPC: %s(%u) = %d\n", |
585 | __func__, xdr->len - req->rq_bytes_sent, status); | |
586 | if (likely(status >= 0)) { | |
587 | req->rq_bytes_sent += status; | |
588 | req->rq_xmit_bytes_sent += status; | |
589 | if (likely(req->rq_bytes_sent >= req->rq_slen)) { | |
590 | req->rq_bytes_sent = 0; | |
591 | return 0; | |
592 | } | |
593 | status = -EAGAIN; | |
594 | } | |
595 | ||
596 | switch (status) { | |
3601c4a9 | 597 | case -ENOBUFS: |
176e21ee CL |
598 | case -EAGAIN: |
599 | status = xs_nospace(task); | |
600 | break; | |
601 | default: | |
602 | dprintk("RPC: sendmsg returned unrecognized error %d\n", | |
603 | -status); | |
604 | case -EPIPE: | |
605 | xs_close(xprt); | |
606 | status = -ENOTCONN; | |
607 | } | |
608 | ||
609 | return status; | |
610 | } | |
611 | ||
262965f5 CL |
612 | /** |
613 | * xs_udp_send_request - write an RPC request to a UDP socket | |
614 | * @task: address of RPC task that manages the state of an RPC request | |
615 | * | |
616 | * Return values: | |
617 | * 0: The request has been sent | |
618 | * EAGAIN: The socket was blocked, please call again later to | |
619 | * complete the request | |
620 | * ENOTCONN: Caller needs to invoke connect logic then call again | |
25985edc | 621 | * other: Some other error occurred, the request was not sent |
262965f5 CL |
622 | */ |
623 | static int xs_udp_send_request(struct rpc_task *task) | |
624 | { | |
625 | struct rpc_rqst *req = task->tk_rqstp; | |
626 | struct rpc_xprt *xprt = req->rq_xprt; | |
ee0ac0c2 | 627 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
262965f5 CL |
628 | struct xdr_buf *xdr = &req->rq_snd_buf; |
629 | int status; | |
a246b010 | 630 | |
9903cd1c | 631 | xs_pktdump("packet data:", |
a246b010 CL |
632 | req->rq_svec->iov_base, |
633 | req->rq_svec->iov_len); | |
634 | ||
01d37c42 TM |
635 | if (!xprt_bound(xprt)) |
636 | return -ENOTCONN; | |
ee0ac0c2 | 637 | status = xs_sendpages(transport->sock, |
95392c59 | 638 | xs_addr(xprt), |
ee0ac0c2 | 639 | xprt->addrlen, xdr, |
a6b31d18 | 640 | req->rq_bytes_sent, true); |
a246b010 | 641 | |
46121cf7 | 642 | dprintk("RPC: xs_udp_send_request(%u) = %d\n", |
262965f5 | 643 | xdr->len - req->rq_bytes_sent, status); |
a246b010 | 644 | |
2199700f | 645 | if (status >= 0) { |
d60dbb20 | 646 | req->rq_xmit_bytes_sent += status; |
2199700f TM |
647 | if (status >= req->rq_slen) |
648 | return 0; | |
649 | /* Still some bytes left; set up for a retry later. */ | |
262965f5 | 650 | status = -EAGAIN; |
2199700f | 651 | } |
a246b010 | 652 | |
262965f5 | 653 | switch (status) { |
fba91afb TM |
654 | case -ENOTSOCK: |
655 | status = -ENOTCONN; | |
656 | /* Should we call xs_close() here? */ | |
657 | break; | |
b6ddf64f | 658 | case -EAGAIN: |
5e3771ce | 659 | status = xs_nospace(task); |
b6ddf64f | 660 | break; |
c8485e4d TM |
661 | default: |
662 | dprintk("RPC: sendmsg returned unrecognized error %d\n", | |
663 | -status); | |
262965f5 | 664 | case -ENETUNREACH: |
3601c4a9 | 665 | case -ENOBUFS: |
262965f5 | 666 | case -EPIPE: |
a246b010 CL |
667 | case -ECONNREFUSED: |
668 | /* When the server has died, an ICMP port unreachable message | |
9903cd1c | 669 | * prompts ECONNREFUSED. */ |
b6ddf64f | 670 | clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); |
a246b010 | 671 | } |
5fe46e9d | 672 | |
262965f5 | 673 | return status; |
a246b010 CL |
674 | } |
675 | ||
e06799f9 TM |
676 | /** |
677 | * xs_tcp_shutdown - gracefully shut down a TCP socket | |
678 | * @xprt: transport | |
679 | * | |
680 | * Initiates a graceful shutdown of the TCP socket by calling the | |
681 | * equivalent of shutdown(SHUT_WR); | |
682 | */ | |
683 | static void xs_tcp_shutdown(struct rpc_xprt *xprt) | |
684 | { | |
685 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); | |
686 | struct socket *sock = transport->sock; | |
687 | ||
40b5ea0c | 688 | if (sock != NULL) { |
e06799f9 | 689 | kernel_sock_shutdown(sock, SHUT_WR); |
40b5ea0c TM |
690 | trace_rpc_socket_shutdown(xprt, sock); |
691 | } | |
e06799f9 TM |
692 | } |
693 | ||
9903cd1c | 694 | /** |
262965f5 | 695 | * xs_tcp_send_request - write an RPC request to a TCP socket |
9903cd1c CL |
696 | * @task: address of RPC task that manages the state of an RPC request |
697 | * | |
698 | * Return values: | |
262965f5 CL |
699 | * 0: The request has been sent |
700 | * EAGAIN: The socket was blocked, please call again later to | |
701 | * complete the request | |
702 | * ENOTCONN: Caller needs to invoke connect logic then call again | |
25985edc | 703 | * other: Some other error occurred, the request was not sent |
9903cd1c CL |
704 | * |
705 | * XXX: In the case of soft timeouts, should we eventually give up | |
262965f5 | 706 | * if sendmsg is not able to make progress? |
9903cd1c | 707 | */ |
262965f5 | 708 | static int xs_tcp_send_request(struct rpc_task *task) |
a246b010 CL |
709 | { |
710 | struct rpc_rqst *req = task->tk_rqstp; | |
711 | struct rpc_xprt *xprt = req->rq_xprt; | |
ee0ac0c2 | 712 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
262965f5 | 713 | struct xdr_buf *xdr = &req->rq_snd_buf; |
a6b31d18 | 714 | bool zerocopy = true; |
b595bb15 | 715 | int status; |
a246b010 | 716 | |
61677eee | 717 | xs_encode_stream_record_marker(&req->rq_snd_buf); |
a246b010 | 718 | |
262965f5 CL |
719 | xs_pktdump("packet data:", |
720 | req->rq_svec->iov_base, | |
721 | req->rq_svec->iov_len); | |
a6b31d18 TM |
722 | /* Don't use zero copy if this is a resend. If the RPC call |
723 | * completes while the socket holds a reference to the pages, | |
724 | * then we may end up resending corrupted data. | |
725 | */ | |
726 | if (task->tk_flags & RPC_TASK_SENT) | |
727 | zerocopy = false; | |
a246b010 CL |
728 | |
729 | /* Continue transmitting the packet/record. We must be careful | |
730 | * to cope with writespace callbacks arriving _after_ we have | |
262965f5 | 731 | * called sendmsg(). */ |
a246b010 | 732 | while (1) { |
ee0ac0c2 | 733 | status = xs_sendpages(transport->sock, |
a6b31d18 TM |
734 | NULL, 0, xdr, req->rq_bytes_sent, |
735 | zerocopy); | |
a246b010 | 736 | |
46121cf7 | 737 | dprintk("RPC: xs_tcp_send_request(%u) = %d\n", |
262965f5 | 738 | xdr->len - req->rq_bytes_sent, status); |
a246b010 | 739 | |
262965f5 | 740 | if (unlikely(status < 0)) |
a246b010 | 741 | break; |
a246b010 | 742 | |
262965f5 CL |
743 | /* If we've sent the entire packet, immediately |
744 | * reset the count of bytes sent. */ | |
745 | req->rq_bytes_sent += status; | |
d60dbb20 | 746 | req->rq_xmit_bytes_sent += status; |
262965f5 CL |
747 | if (likely(req->rq_bytes_sent >= req->rq_slen)) { |
748 | req->rq_bytes_sent = 0; | |
749 | return 0; | |
750 | } | |
a246b010 | 751 | |
06b4b681 TM |
752 | if (status != 0) |
753 | continue; | |
a246b010 | 754 | status = -EAGAIN; |
06b4b681 | 755 | break; |
a246b010 CL |
756 | } |
757 | ||
262965f5 | 758 | switch (status) { |
fba91afb TM |
759 | case -ENOTSOCK: |
760 | status = -ENOTCONN; | |
761 | /* Should we call xs_close() here? */ | |
762 | break; | |
3601c4a9 | 763 | case -ENOBUFS: |
262965f5 | 764 | case -EAGAIN: |
5e3771ce | 765 | status = xs_nospace(task); |
262965f5 | 766 | break; |
c8485e4d TM |
767 | default: |
768 | dprintk("RPC: sendmsg returned unrecognized error %d\n", | |
769 | -status); | |
262965f5 | 770 | case -ECONNRESET: |
2a9e1cfa TM |
771 | xs_tcp_shutdown(xprt); |
772 | case -ECONNREFUSED: | |
262965f5 | 773 | case -ENOTCONN: |
b9d2bb2e | 774 | case -EPIPE: |
b6ddf64f | 775 | clear_bit(SOCK_ASYNC_NOSPACE, &transport->sock->flags); |
a246b010 | 776 | } |
5fe46e9d | 777 | |
a246b010 CL |
778 | return status; |
779 | } | |
780 | ||
e0ab53de TM |
781 | /** |
782 | * xs_tcp_release_xprt - clean up after a tcp transmission | |
783 | * @xprt: transport | |
784 | * @task: rpc task | |
785 | * | |
786 | * This cleans up if an error causes us to abort the transmission of a request. | |
787 | * In this case, the socket may need to be reset in order to avoid confusing | |
788 | * the server. | |
789 | */ | |
790 | static void xs_tcp_release_xprt(struct rpc_xprt *xprt, struct rpc_task *task) | |
791 | { | |
792 | struct rpc_rqst *req; | |
793 | ||
794 | if (task != xprt->snd_task) | |
795 | return; | |
796 | if (task == NULL) | |
797 | goto out_release; | |
798 | req = task->tk_rqstp; | |
43cedbf0 TM |
799 | if (req == NULL) |
800 | goto out_release; | |
e0ab53de TM |
801 | if (req->rq_bytes_sent == 0) |
802 | goto out_release; | |
803 | if (req->rq_bytes_sent == req->rq_snd_buf.len) | |
804 | goto out_release; | |
a4f0835c | 805 | set_bit(XPRT_CLOSE_WAIT, &xprt->state); |
e0ab53de TM |
806 | out_release: |
807 | xprt_release_xprt(xprt, task); | |
808 | } | |
809 | ||
2a9e1cfa TM |
810 | static void xs_save_old_callbacks(struct sock_xprt *transport, struct sock *sk) |
811 | { | |
812 | transport->old_data_ready = sk->sk_data_ready; | |
813 | transport->old_state_change = sk->sk_state_change; | |
814 | transport->old_write_space = sk->sk_write_space; | |
2118071d | 815 | transport->old_error_report = sk->sk_error_report; |
2a9e1cfa TM |
816 | } |
817 | ||
818 | static void xs_restore_old_callbacks(struct sock_xprt *transport, struct sock *sk) | |
819 | { | |
820 | sk->sk_data_ready = transport->old_data_ready; | |
821 | sk->sk_state_change = transport->old_state_change; | |
822 | sk->sk_write_space = transport->old_write_space; | |
2118071d TM |
823 | sk->sk_error_report = transport->old_error_report; |
824 | } | |
825 | ||
826 | /** | |
827 | * xs_error_report - callback to handle TCP socket state errors | |
828 | * @sk: socket | |
829 | * | |
830 | * Note: we don't call sock_error() since there may be a rpc_task | |
831 | * using the socket, and so we don't want to clear sk->sk_err. | |
832 | */ | |
833 | static void xs_error_report(struct sock *sk) | |
834 | { | |
835 | struct rpc_xprt *xprt; | |
836 | int err; | |
837 | ||
838 | read_lock_bh(&sk->sk_callback_lock); | |
839 | if (!(xprt = xprt_from_sock(sk))) | |
840 | goto out; | |
841 | ||
842 | err = -sk->sk_err; | |
843 | if (err == 0) | |
844 | goto out; | |
845 | dprintk("RPC: xs_error_report client %p, error=%d...\n", | |
846 | xprt, -err); | |
e8353c76 | 847 | trace_rpc_socket_error(xprt, sk->sk_socket, err); |
2118071d TM |
848 | xprt_wake_pending_tasks(xprt, err); |
849 | out: | |
850 | read_unlock_bh(&sk->sk_callback_lock); | |
2a9e1cfa TM |
851 | } |
852 | ||
fe315e76 | 853 | static void xs_reset_transport(struct sock_xprt *transport) |
a246b010 | 854 | { |
ee0ac0c2 CL |
855 | struct socket *sock = transport->sock; |
856 | struct sock *sk = transport->inet; | |
a246b010 | 857 | |
fe315e76 CL |
858 | if (sk == NULL) |
859 | return; | |
9903cd1c | 860 | |
246408dc TM |
861 | transport->srcport = 0; |
862 | ||
a246b010 | 863 | write_lock_bh(&sk->sk_callback_lock); |
ee0ac0c2 CL |
864 | transport->inet = NULL; |
865 | transport->sock = NULL; | |
a246b010 | 866 | |
9903cd1c | 867 | sk->sk_user_data = NULL; |
2a9e1cfa TM |
868 | |
869 | xs_restore_old_callbacks(transport, sk); | |
a246b010 CL |
870 | write_unlock_bh(&sk->sk_callback_lock); |
871 | ||
40b5ea0c | 872 | trace_rpc_socket_close(&transport->xprt, sock); |
a246b010 | 873 | sock_release(sock); |
fe315e76 CL |
874 | } |
875 | ||
876 | /** | |
877 | * xs_close - close a socket | |
878 | * @xprt: transport | |
879 | * | |
880 | * This is used when all requests are complete; ie, no DRC state remains | |
881 | * on the server we want to save. | |
f75e6745 TM |
882 | * |
883 | * The caller _must_ be holding XPRT_LOCKED in order to avoid issues with | |
884 | * xs_reset_transport() zeroing the socket from underneath a writer. | |
fe315e76 CL |
885 | */ |
886 | static void xs_close(struct rpc_xprt *xprt) | |
887 | { | |
888 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); | |
889 | ||
890 | dprintk("RPC: xs_close xprt %p\n", xprt); | |
891 | ||
93dc41bd N |
892 | cancel_delayed_work_sync(&transport->connect_worker); |
893 | ||
fe315e76 | 894 | xs_reset_transport(transport); |
61d0a8e6 | 895 | xprt->reestablish_timeout = 0; |
fe315e76 | 896 | |
4e857c58 | 897 | smp_mb__before_atomic(); |
7d1e8255 | 898 | clear_bit(XPRT_CONNECTION_ABORT, &xprt->state); |
632e3bdc | 899 | clear_bit(XPRT_CLOSE_WAIT, &xprt->state); |
3b948ae5 | 900 | clear_bit(XPRT_CLOSING, &xprt->state); |
4e857c58 | 901 | smp_mb__after_atomic(); |
62da3b24 | 902 | xprt_disconnect_done(xprt); |
a246b010 CL |
903 | } |
904 | ||
f75e6745 TM |
905 | static void xs_tcp_close(struct rpc_xprt *xprt) |
906 | { | |
907 | if (test_and_clear_bit(XPRT_CONNECTION_CLOSE, &xprt->state)) | |
908 | xs_close(xprt); | |
909 | else | |
910 | xs_tcp_shutdown(xprt); | |
911 | } | |
912 | ||
315f3812 KM |
913 | static void xs_xprt_free(struct rpc_xprt *xprt) |
914 | { | |
915 | xs_free_peer_addresses(xprt); | |
916 | xprt_free(xprt); | |
917 | } | |
918 | ||
9903cd1c CL |
919 | /** |
920 | * xs_destroy - prepare to shutdown a transport | |
921 | * @xprt: doomed transport | |
922 | * | |
923 | */ | |
924 | static void xs_destroy(struct rpc_xprt *xprt) | |
a246b010 | 925 | { |
46121cf7 | 926 | dprintk("RPC: xs_destroy xprt %p\n", xprt); |
9903cd1c | 927 | |
a1311d87 | 928 | xs_close(xprt); |
315f3812 | 929 | xs_xprt_free(xprt); |
a1311d87 | 930 | module_put(THIS_MODULE); |
a246b010 CL |
931 | } |
932 | ||
176e21ee CL |
933 | static int xs_local_copy_to_xdr(struct xdr_buf *xdr, struct sk_buff *skb) |
934 | { | |
935 | struct xdr_skb_reader desc = { | |
936 | .skb = skb, | |
937 | .offset = sizeof(rpc_fraghdr), | |
938 | .count = skb->len - sizeof(rpc_fraghdr), | |
939 | }; | |
940 | ||
941 | if (xdr_partial_copy_from_skb(xdr, 0, &desc, xdr_skb_read_bits) < 0) | |
942 | return -1; | |
943 | if (desc.count) | |
944 | return -1; | |
945 | return 0; | |
946 | } | |
947 | ||
948 | /** | |
949 | * xs_local_data_ready - "data ready" callback for AF_LOCAL sockets | |
950 | * @sk: socket with data to read | |
951 | * @len: how much data to read | |
952 | * | |
953 | * Currently this assumes we can read the whole reply in a single gulp. | |
954 | */ | |
676d2369 | 955 | static void xs_local_data_ready(struct sock *sk) |
176e21ee CL |
956 | { |
957 | struct rpc_task *task; | |
958 | struct rpc_xprt *xprt; | |
959 | struct rpc_rqst *rovr; | |
960 | struct sk_buff *skb; | |
961 | int err, repsize, copied; | |
962 | u32 _xid; | |
963 | __be32 *xp; | |
964 | ||
965 | read_lock_bh(&sk->sk_callback_lock); | |
966 | dprintk("RPC: %s...\n", __func__); | |
967 | xprt = xprt_from_sock(sk); | |
968 | if (xprt == NULL) | |
969 | goto out; | |
970 | ||
971 | skb = skb_recv_datagram(sk, 0, 1, &err); | |
972 | if (skb == NULL) | |
973 | goto out; | |
974 | ||
176e21ee CL |
975 | repsize = skb->len - sizeof(rpc_fraghdr); |
976 | if (repsize < 4) { | |
977 | dprintk("RPC: impossible RPC reply size %d\n", repsize); | |
978 | goto dropit; | |
979 | } | |
980 | ||
981 | /* Copy the XID from the skb... */ | |
982 | xp = skb_header_pointer(skb, sizeof(rpc_fraghdr), sizeof(_xid), &_xid); | |
983 | if (xp == NULL) | |
984 | goto dropit; | |
985 | ||
986 | /* Look up and lock the request corresponding to the given XID */ | |
987 | spin_lock(&xprt->transport_lock); | |
988 | rovr = xprt_lookup_rqst(xprt, *xp); | |
989 | if (!rovr) | |
990 | goto out_unlock; | |
991 | task = rovr->rq_task; | |
992 | ||
993 | copied = rovr->rq_private_buf.buflen; | |
994 | if (copied > repsize) | |
995 | copied = repsize; | |
996 | ||
997 | if (xs_local_copy_to_xdr(&rovr->rq_private_buf, skb)) { | |
998 | dprintk("RPC: sk_buff copy failed\n"); | |
999 | goto out_unlock; | |
1000 | } | |
1001 | ||
1002 | xprt_complete_rqst(task, copied); | |
1003 | ||
1004 | out_unlock: | |
1005 | spin_unlock(&xprt->transport_lock); | |
1006 | dropit: | |
1007 | skb_free_datagram(sk, skb); | |
1008 | out: | |
1009 | read_unlock_bh(&sk->sk_callback_lock); | |
1010 | } | |
1011 | ||
9903cd1c CL |
1012 | /** |
1013 | * xs_udp_data_ready - "data ready" callback for UDP sockets | |
1014 | * @sk: socket with data to read | |
1015 | * @len: how much data to read | |
1016 | * | |
a246b010 | 1017 | */ |
676d2369 | 1018 | static void xs_udp_data_ready(struct sock *sk) |
a246b010 | 1019 | { |
9903cd1c CL |
1020 | struct rpc_task *task; |
1021 | struct rpc_xprt *xprt; | |
a246b010 | 1022 | struct rpc_rqst *rovr; |
9903cd1c | 1023 | struct sk_buff *skb; |
a246b010 | 1024 | int err, repsize, copied; |
d8ed029d AD |
1025 | u32 _xid; |
1026 | __be32 *xp; | |
a246b010 | 1027 | |
f064af1e | 1028 | read_lock_bh(&sk->sk_callback_lock); |
46121cf7 | 1029 | dprintk("RPC: xs_udp_data_ready...\n"); |
9903cd1c | 1030 | if (!(xprt = xprt_from_sock(sk))) |
a246b010 | 1031 | goto out; |
a246b010 CL |
1032 | |
1033 | if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL) | |
1034 | goto out; | |
1035 | ||
a246b010 CL |
1036 | repsize = skb->len - sizeof(struct udphdr); |
1037 | if (repsize < 4) { | |
46121cf7 | 1038 | dprintk("RPC: impossible RPC reply size %d!\n", repsize); |
a246b010 CL |
1039 | goto dropit; |
1040 | } | |
1041 | ||
1042 | /* Copy the XID from the skb... */ | |
1043 | xp = skb_header_pointer(skb, sizeof(struct udphdr), | |
1044 | sizeof(_xid), &_xid); | |
1045 | if (xp == NULL) | |
1046 | goto dropit; | |
1047 | ||
1048 | /* Look up and lock the request corresponding to the given XID */ | |
4a0f8c04 | 1049 | spin_lock(&xprt->transport_lock); |
a246b010 CL |
1050 | rovr = xprt_lookup_rqst(xprt, *xp); |
1051 | if (!rovr) | |
1052 | goto out_unlock; | |
1053 | task = rovr->rq_task; | |
1054 | ||
a246b010 CL |
1055 | if ((copied = rovr->rq_private_buf.buflen) > repsize) |
1056 | copied = repsize; | |
1057 | ||
1058 | /* Suck it into the iovec, verify checksum if not done by hw. */ | |
1781f7f5 HX |
1059 | if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb)) { |
1060 | UDPX_INC_STATS_BH(sk, UDP_MIB_INERRORS); | |
a246b010 | 1061 | goto out_unlock; |
1781f7f5 HX |
1062 | } |
1063 | ||
1064 | UDPX_INC_STATS_BH(sk, UDP_MIB_INDATAGRAMS); | |
a246b010 | 1065 | |
6a24dfb6 | 1066 | xprt_adjust_cwnd(xprt, task, copied); |
1570c1e4 | 1067 | xprt_complete_rqst(task, copied); |
a246b010 CL |
1068 | |
1069 | out_unlock: | |
4a0f8c04 | 1070 | spin_unlock(&xprt->transport_lock); |
a246b010 CL |
1071 | dropit: |
1072 | skb_free_datagram(sk, skb); | |
1073 | out: | |
f064af1e | 1074 | read_unlock_bh(&sk->sk_callback_lock); |
a246b010 CL |
1075 | } |
1076 | ||
a519fc7a TM |
1077 | /* |
1078 | * Helper function to force a TCP close if the server is sending | |
1079 | * junk and/or it has put us in CLOSE_WAIT | |
1080 | */ | |
1081 | static void xs_tcp_force_close(struct rpc_xprt *xprt) | |
1082 | { | |
1083 | set_bit(XPRT_CONNECTION_CLOSE, &xprt->state); | |
1084 | xprt_force_disconnect(xprt); | |
1085 | } | |
1086 | ||
dd456471 | 1087 | static inline void xs_tcp_read_fraghdr(struct rpc_xprt *xprt, struct xdr_skb_reader *desc) |
a246b010 | 1088 | { |
51971139 | 1089 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
a246b010 CL |
1090 | size_t len, used; |
1091 | char *p; | |
1092 | ||
51971139 CL |
1093 | p = ((char *) &transport->tcp_fraghdr) + transport->tcp_offset; |
1094 | len = sizeof(transport->tcp_fraghdr) - transport->tcp_offset; | |
9d292316 | 1095 | used = xdr_skb_read_bits(desc, p, len); |
51971139 | 1096 | transport->tcp_offset += used; |
a246b010 CL |
1097 | if (used != len) |
1098 | return; | |
808012fb | 1099 | |
51971139 CL |
1100 | transport->tcp_reclen = ntohl(transport->tcp_fraghdr); |
1101 | if (transport->tcp_reclen & RPC_LAST_STREAM_FRAGMENT) | |
e136d092 | 1102 | transport->tcp_flags |= TCP_RCV_LAST_FRAG; |
a246b010 | 1103 | else |
e136d092 | 1104 | transport->tcp_flags &= ~TCP_RCV_LAST_FRAG; |
51971139 | 1105 | transport->tcp_reclen &= RPC_FRAGMENT_SIZE_MASK; |
808012fb | 1106 | |
e136d092 | 1107 | transport->tcp_flags &= ~TCP_RCV_COPY_FRAGHDR; |
51971139 | 1108 | transport->tcp_offset = 0; |
808012fb | 1109 | |
a246b010 | 1110 | /* Sanity check of the record length */ |
18dca02a | 1111 | if (unlikely(transport->tcp_reclen < 8)) { |
46121cf7 | 1112 | dprintk("RPC: invalid TCP record fragment length\n"); |
a519fc7a | 1113 | xs_tcp_force_close(xprt); |
9903cd1c | 1114 | return; |
a246b010 | 1115 | } |
46121cf7 | 1116 | dprintk("RPC: reading TCP record fragment of length %d\n", |
51971139 | 1117 | transport->tcp_reclen); |
a246b010 CL |
1118 | } |
1119 | ||
51971139 | 1120 | static void xs_tcp_check_fraghdr(struct sock_xprt *transport) |
a246b010 | 1121 | { |
51971139 | 1122 | if (transport->tcp_offset == transport->tcp_reclen) { |
e136d092 | 1123 | transport->tcp_flags |= TCP_RCV_COPY_FRAGHDR; |
51971139 | 1124 | transport->tcp_offset = 0; |
e136d092 CL |
1125 | if (transport->tcp_flags & TCP_RCV_LAST_FRAG) { |
1126 | transport->tcp_flags &= ~TCP_RCV_COPY_DATA; | |
1127 | transport->tcp_flags |= TCP_RCV_COPY_XID; | |
51971139 | 1128 | transport->tcp_copied = 0; |
a246b010 CL |
1129 | } |
1130 | } | |
1131 | } | |
1132 | ||
dd456471 | 1133 | static inline void xs_tcp_read_xid(struct sock_xprt *transport, struct xdr_skb_reader *desc) |
a246b010 CL |
1134 | { |
1135 | size_t len, used; | |
1136 | char *p; | |
1137 | ||
51971139 | 1138 | len = sizeof(transport->tcp_xid) - transport->tcp_offset; |
46121cf7 | 1139 | dprintk("RPC: reading XID (%Zu bytes)\n", len); |
51971139 | 1140 | p = ((char *) &transport->tcp_xid) + transport->tcp_offset; |
9d292316 | 1141 | used = xdr_skb_read_bits(desc, p, len); |
51971139 | 1142 | transport->tcp_offset += used; |
a246b010 CL |
1143 | if (used != len) |
1144 | return; | |
e136d092 | 1145 | transport->tcp_flags &= ~TCP_RCV_COPY_XID; |
f4a2e418 | 1146 | transport->tcp_flags |= TCP_RCV_READ_CALLDIR; |
51971139 | 1147 | transport->tcp_copied = 4; |
18dca02a RL |
1148 | dprintk("RPC: reading %s XID %08x\n", |
1149 | (transport->tcp_flags & TCP_RPC_REPLY) ? "reply for" | |
1150 | : "request with", | |
51971139 CL |
1151 | ntohl(transport->tcp_xid)); |
1152 | xs_tcp_check_fraghdr(transport); | |
a246b010 CL |
1153 | } |
1154 | ||
18dca02a RL |
1155 | static inline void xs_tcp_read_calldir(struct sock_xprt *transport, |
1156 | struct xdr_skb_reader *desc) | |
a246b010 | 1157 | { |
18dca02a RL |
1158 | size_t len, used; |
1159 | u32 offset; | |
b76ce561 | 1160 | char *p; |
18dca02a RL |
1161 | |
1162 | /* | |
1163 | * We want transport->tcp_offset to be 8 at the end of this routine | |
1164 | * (4 bytes for the xid and 4 bytes for the call/reply flag). | |
1165 | * When this function is called for the first time, | |
1166 | * transport->tcp_offset is 4 (after having already read the xid). | |
1167 | */ | |
1168 | offset = transport->tcp_offset - sizeof(transport->tcp_xid); | |
b76ce561 | 1169 | len = sizeof(transport->tcp_calldir) - offset; |
18dca02a | 1170 | dprintk("RPC: reading CALL/REPLY flag (%Zu bytes)\n", len); |
b76ce561 TM |
1171 | p = ((char *) &transport->tcp_calldir) + offset; |
1172 | used = xdr_skb_read_bits(desc, p, len); | |
18dca02a RL |
1173 | transport->tcp_offset += used; |
1174 | if (used != len) | |
1175 | return; | |
f4a2e418 | 1176 | transport->tcp_flags &= ~TCP_RCV_READ_CALLDIR; |
f4a2e418 RL |
1177 | /* |
1178 | * We don't yet have the XDR buffer, so we will write the calldir | |
1179 | * out after we get the buffer from the 'struct rpc_rqst' | |
1180 | */ | |
b76ce561 TM |
1181 | switch (ntohl(transport->tcp_calldir)) { |
1182 | case RPC_REPLY: | |
1183 | transport->tcp_flags |= TCP_RCV_COPY_CALLDIR; | |
1184 | transport->tcp_flags |= TCP_RCV_COPY_DATA; | |
18dca02a | 1185 | transport->tcp_flags |= TCP_RPC_REPLY; |
b76ce561 TM |
1186 | break; |
1187 | case RPC_CALL: | |
1188 | transport->tcp_flags |= TCP_RCV_COPY_CALLDIR; | |
1189 | transport->tcp_flags |= TCP_RCV_COPY_DATA; | |
18dca02a | 1190 | transport->tcp_flags &= ~TCP_RPC_REPLY; |
b76ce561 TM |
1191 | break; |
1192 | default: | |
1193 | dprintk("RPC: invalid request message type\n"); | |
a519fc7a | 1194 | xs_tcp_force_close(&transport->xprt); |
b76ce561 | 1195 | } |
18dca02a RL |
1196 | xs_tcp_check_fraghdr(transport); |
1197 | } | |
1198 | ||
44b98efd RL |
1199 | static inline void xs_tcp_read_common(struct rpc_xprt *xprt, |
1200 | struct xdr_skb_reader *desc, | |
1201 | struct rpc_rqst *req) | |
a246b010 | 1202 | { |
44b98efd RL |
1203 | struct sock_xprt *transport = |
1204 | container_of(xprt, struct sock_xprt, xprt); | |
a246b010 CL |
1205 | struct xdr_buf *rcvbuf; |
1206 | size_t len; | |
1207 | ssize_t r; | |
1208 | ||
a246b010 | 1209 | rcvbuf = &req->rq_private_buf; |
f4a2e418 RL |
1210 | |
1211 | if (transport->tcp_flags & TCP_RCV_COPY_CALLDIR) { | |
1212 | /* | |
1213 | * Save the RPC direction in the XDR buffer | |
1214 | */ | |
f4a2e418 | 1215 | memcpy(rcvbuf->head[0].iov_base + transport->tcp_copied, |
b76ce561 TM |
1216 | &transport->tcp_calldir, |
1217 | sizeof(transport->tcp_calldir)); | |
1218 | transport->tcp_copied += sizeof(transport->tcp_calldir); | |
f4a2e418 | 1219 | transport->tcp_flags &= ~TCP_RCV_COPY_CALLDIR; |
a246b010 CL |
1220 | } |
1221 | ||
a246b010 | 1222 | len = desc->count; |
51971139 | 1223 | if (len > transport->tcp_reclen - transport->tcp_offset) { |
dd456471 | 1224 | struct xdr_skb_reader my_desc; |
a246b010 | 1225 | |
51971139 | 1226 | len = transport->tcp_reclen - transport->tcp_offset; |
a246b010 CL |
1227 | memcpy(&my_desc, desc, sizeof(my_desc)); |
1228 | my_desc.count = len; | |
51971139 | 1229 | r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied, |
9d292316 | 1230 | &my_desc, xdr_skb_read_bits); |
a246b010 CL |
1231 | desc->count -= r; |
1232 | desc->offset += r; | |
1233 | } else | |
51971139 | 1234 | r = xdr_partial_copy_from_skb(rcvbuf, transport->tcp_copied, |
9d292316 | 1235 | desc, xdr_skb_read_bits); |
a246b010 CL |
1236 | |
1237 | if (r > 0) { | |
51971139 CL |
1238 | transport->tcp_copied += r; |
1239 | transport->tcp_offset += r; | |
a246b010 CL |
1240 | } |
1241 | if (r != len) { | |
1242 | /* Error when copying to the receive buffer, | |
1243 | * usually because we weren't able to allocate | |
1244 | * additional buffer pages. All we can do now | |
e136d092 | 1245 | * is turn off TCP_RCV_COPY_DATA, so the request |
a246b010 CL |
1246 | * will not receive any additional updates, |
1247 | * and time out. | |
1248 | * Any remaining data from this record will | |
1249 | * be discarded. | |
1250 | */ | |
e136d092 | 1251 | transport->tcp_flags &= ~TCP_RCV_COPY_DATA; |
46121cf7 | 1252 | dprintk("RPC: XID %08x truncated request\n", |
51971139 | 1253 | ntohl(transport->tcp_xid)); |
46121cf7 CL |
1254 | dprintk("RPC: xprt = %p, tcp_copied = %lu, " |
1255 | "tcp_offset = %u, tcp_reclen = %u\n", | |
1256 | xprt, transport->tcp_copied, | |
1257 | transport->tcp_offset, transport->tcp_reclen); | |
44b98efd | 1258 | return; |
a246b010 CL |
1259 | } |
1260 | ||
46121cf7 | 1261 | dprintk("RPC: XID %08x read %Zd bytes\n", |
51971139 | 1262 | ntohl(transport->tcp_xid), r); |
46121cf7 CL |
1263 | dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, " |
1264 | "tcp_reclen = %u\n", xprt, transport->tcp_copied, | |
1265 | transport->tcp_offset, transport->tcp_reclen); | |
51971139 CL |
1266 | |
1267 | if (transport->tcp_copied == req->rq_private_buf.buflen) | |
e136d092 | 1268 | transport->tcp_flags &= ~TCP_RCV_COPY_DATA; |
51971139 | 1269 | else if (transport->tcp_offset == transport->tcp_reclen) { |
e136d092 CL |
1270 | if (transport->tcp_flags & TCP_RCV_LAST_FRAG) |
1271 | transport->tcp_flags &= ~TCP_RCV_COPY_DATA; | |
a246b010 | 1272 | } |
44b98efd RL |
1273 | } |
1274 | ||
1275 | /* | |
1276 | * Finds the request corresponding to the RPC xid and invokes the common | |
1277 | * tcp read code to read the data. | |
1278 | */ | |
1279 | static inline int xs_tcp_read_reply(struct rpc_xprt *xprt, | |
1280 | struct xdr_skb_reader *desc) | |
1281 | { | |
1282 | struct sock_xprt *transport = | |
1283 | container_of(xprt, struct sock_xprt, xprt); | |
1284 | struct rpc_rqst *req; | |
1285 | ||
1286 | dprintk("RPC: read reply XID %08x\n", ntohl(transport->tcp_xid)); | |
1287 | ||
1288 | /* Find and lock the request corresponding to this xid */ | |
1289 | spin_lock(&xprt->transport_lock); | |
1290 | req = xprt_lookup_rqst(xprt, transport->tcp_xid); | |
1291 | if (!req) { | |
1292 | dprintk("RPC: XID %08x request not found!\n", | |
1293 | ntohl(transport->tcp_xid)); | |
1294 | spin_unlock(&xprt->transport_lock); | |
1295 | return -1; | |
1296 | } | |
1297 | ||
1298 | xs_tcp_read_common(xprt, desc, req); | |
1299 | ||
e136d092 | 1300 | if (!(transport->tcp_flags & TCP_RCV_COPY_DATA)) |
51971139 | 1301 | xprt_complete_rqst(req->rq_task, transport->tcp_copied); |
44b98efd | 1302 | |
4a0f8c04 | 1303 | spin_unlock(&xprt->transport_lock); |
44b98efd RL |
1304 | return 0; |
1305 | } | |
1306 | ||
9e00abc3 | 1307 | #if defined(CONFIG_SUNRPC_BACKCHANNEL) |
44b98efd RL |
1308 | /* |
1309 | * Obtains an rpc_rqst previously allocated and invokes the common | |
1310 | * tcp read code to read the data. The result is placed in the callback | |
1311 | * queue. | |
1312 | * If we're unable to obtain the rpc_rqst we schedule the closing of the | |
1313 | * connection and return -1. | |
1314 | */ | |
2ea24497 | 1315 | static int xs_tcp_read_callback(struct rpc_xprt *xprt, |
44b98efd RL |
1316 | struct xdr_skb_reader *desc) |
1317 | { | |
1318 | struct sock_xprt *transport = | |
1319 | container_of(xprt, struct sock_xprt, xprt); | |
1320 | struct rpc_rqst *req; | |
1321 | ||
2ea24497 TM |
1322 | /* Look up and lock the request corresponding to the given XID */ |
1323 | spin_lock(&xprt->transport_lock); | |
1324 | req = xprt_lookup_bc_request(xprt, transport->tcp_xid); | |
44b98efd | 1325 | if (req == NULL) { |
2ea24497 | 1326 | spin_unlock(&xprt->transport_lock); |
44b98efd RL |
1327 | printk(KERN_WARNING "Callback slot table overflowed\n"); |
1328 | xprt_force_disconnect(xprt); | |
1329 | return -1; | |
1330 | } | |
1331 | ||
44b98efd RL |
1332 | dprintk("RPC: read callback XID %08x\n", ntohl(req->rq_xid)); |
1333 | xs_tcp_read_common(xprt, desc, req); | |
1334 | ||
2ea24497 TM |
1335 | if (!(transport->tcp_flags & TCP_RCV_COPY_DATA)) |
1336 | xprt_complete_bc_request(req, transport->tcp_copied); | |
1337 | spin_unlock(&xprt->transport_lock); | |
44b98efd RL |
1338 | |
1339 | return 0; | |
1340 | } | |
1341 | ||
1342 | static inline int _xs_tcp_read_data(struct rpc_xprt *xprt, | |
1343 | struct xdr_skb_reader *desc) | |
1344 | { | |
1345 | struct sock_xprt *transport = | |
1346 | container_of(xprt, struct sock_xprt, xprt); | |
1347 | ||
1348 | return (transport->tcp_flags & TCP_RPC_REPLY) ? | |
1349 | xs_tcp_read_reply(xprt, desc) : | |
1350 | xs_tcp_read_callback(xprt, desc); | |
1351 | } | |
1352 | #else | |
1353 | static inline int _xs_tcp_read_data(struct rpc_xprt *xprt, | |
1354 | struct xdr_skb_reader *desc) | |
1355 | { | |
1356 | return xs_tcp_read_reply(xprt, desc); | |
1357 | } | |
9e00abc3 | 1358 | #endif /* CONFIG_SUNRPC_BACKCHANNEL */ |
44b98efd RL |
1359 | |
1360 | /* | |
1361 | * Read data off the transport. This can be either an RPC_CALL or an | |
1362 | * RPC_REPLY. Relay the processing to helper functions. | |
1363 | */ | |
1364 | static void xs_tcp_read_data(struct rpc_xprt *xprt, | |
1365 | struct xdr_skb_reader *desc) | |
1366 | { | |
1367 | struct sock_xprt *transport = | |
1368 | container_of(xprt, struct sock_xprt, xprt); | |
1369 | ||
1370 | if (_xs_tcp_read_data(xprt, desc) == 0) | |
1371 | xs_tcp_check_fraghdr(transport); | |
1372 | else { | |
1373 | /* | |
1374 | * The transport_lock protects the request handling. | |
1375 | * There's no need to hold it to update the tcp_flags. | |
1376 | */ | |
1377 | transport->tcp_flags &= ~TCP_RCV_COPY_DATA; | |
1378 | } | |
a246b010 CL |
1379 | } |
1380 | ||
dd456471 | 1381 | static inline void xs_tcp_read_discard(struct sock_xprt *transport, struct xdr_skb_reader *desc) |
a246b010 CL |
1382 | { |
1383 | size_t len; | |
1384 | ||
51971139 | 1385 | len = transport->tcp_reclen - transport->tcp_offset; |
a246b010 CL |
1386 | if (len > desc->count) |
1387 | len = desc->count; | |
1388 | desc->count -= len; | |
1389 | desc->offset += len; | |
51971139 | 1390 | transport->tcp_offset += len; |
46121cf7 | 1391 | dprintk("RPC: discarded %Zu bytes\n", len); |
51971139 | 1392 | xs_tcp_check_fraghdr(transport); |
a246b010 CL |
1393 | } |
1394 | ||
9903cd1c | 1395 | static int xs_tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb, unsigned int offset, size_t len) |
a246b010 CL |
1396 | { |
1397 | struct rpc_xprt *xprt = rd_desc->arg.data; | |
51971139 | 1398 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
dd456471 | 1399 | struct xdr_skb_reader desc = { |
a246b010 CL |
1400 | .skb = skb, |
1401 | .offset = offset, | |
1402 | .count = len, | |
9903cd1c | 1403 | }; |
a246b010 | 1404 | |
46121cf7 | 1405 | dprintk("RPC: xs_tcp_data_recv started\n"); |
a246b010 CL |
1406 | do { |
1407 | /* Read in a new fragment marker if necessary */ | |
1408 | /* Can we ever really expect to get completely empty fragments? */ | |
e136d092 | 1409 | if (transport->tcp_flags & TCP_RCV_COPY_FRAGHDR) { |
9903cd1c | 1410 | xs_tcp_read_fraghdr(xprt, &desc); |
a246b010 CL |
1411 | continue; |
1412 | } | |
1413 | /* Read in the xid if necessary */ | |
e136d092 | 1414 | if (transport->tcp_flags & TCP_RCV_COPY_XID) { |
51971139 | 1415 | xs_tcp_read_xid(transport, &desc); |
a246b010 CL |
1416 | continue; |
1417 | } | |
18dca02a | 1418 | /* Read in the call/reply flag */ |
f4a2e418 | 1419 | if (transport->tcp_flags & TCP_RCV_READ_CALLDIR) { |
18dca02a RL |
1420 | xs_tcp_read_calldir(transport, &desc); |
1421 | continue; | |
1422 | } | |
a246b010 | 1423 | /* Read in the request data */ |
e136d092 | 1424 | if (transport->tcp_flags & TCP_RCV_COPY_DATA) { |
44b98efd | 1425 | xs_tcp_read_data(xprt, &desc); |
a246b010 CL |
1426 | continue; |
1427 | } | |
1428 | /* Skip over any trailing bytes on short reads */ | |
51971139 | 1429 | xs_tcp_read_discard(transport, &desc); |
a246b010 | 1430 | } while (desc.count); |
46121cf7 | 1431 | dprintk("RPC: xs_tcp_data_recv done\n"); |
a246b010 CL |
1432 | return len - desc.count; |
1433 | } | |
1434 | ||
9903cd1c CL |
1435 | /** |
1436 | * xs_tcp_data_ready - "data ready" callback for TCP sockets | |
1437 | * @sk: socket with data to read | |
1438 | * @bytes: how much data to read | |
1439 | * | |
1440 | */ | |
676d2369 | 1441 | static void xs_tcp_data_ready(struct sock *sk) |
a246b010 CL |
1442 | { |
1443 | struct rpc_xprt *xprt; | |
1444 | read_descriptor_t rd_desc; | |
ff2d7db8 | 1445 | int read; |
a246b010 | 1446 | |
46121cf7 CL |
1447 | dprintk("RPC: xs_tcp_data_ready...\n"); |
1448 | ||
f064af1e | 1449 | read_lock_bh(&sk->sk_callback_lock); |
9903cd1c | 1450 | if (!(xprt = xprt_from_sock(sk))) |
a246b010 | 1451 | goto out; |
61d0a8e6 NB |
1452 | /* Any data means we had a useful conversation, so |
1453 | * the we don't need to delay the next reconnect | |
1454 | */ | |
1455 | if (xprt->reestablish_timeout) | |
1456 | xprt->reestablish_timeout = 0; | |
1457 | ||
9903cd1c | 1458 | /* We use rd_desc to pass struct xprt to xs_tcp_data_recv */ |
a246b010 | 1459 | rd_desc.arg.data = xprt; |
ff2d7db8 TM |
1460 | do { |
1461 | rd_desc.count = 65536; | |
1462 | read = tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv); | |
1463 | } while (read > 0); | |
a246b010 | 1464 | out: |
f064af1e | 1465 | read_unlock_bh(&sk->sk_callback_lock); |
a246b010 CL |
1466 | } |
1467 | ||
7d1e8255 TM |
1468 | /* |
1469 | * Do the equivalent of linger/linger2 handling for dealing with | |
1470 | * broken servers that don't close the socket in a timely | |
1471 | * fashion | |
1472 | */ | |
1473 | static void xs_tcp_schedule_linger_timeout(struct rpc_xprt *xprt, | |
1474 | unsigned long timeout) | |
1475 | { | |
1476 | struct sock_xprt *transport; | |
1477 | ||
1478 | if (xprt_test_and_set_connecting(xprt)) | |
1479 | return; | |
1480 | set_bit(XPRT_CONNECTION_ABORT, &xprt->state); | |
1481 | transport = container_of(xprt, struct sock_xprt, xprt); | |
1482 | queue_delayed_work(rpciod_workqueue, &transport->connect_worker, | |
1483 | timeout); | |
1484 | } | |
1485 | ||
1486 | static void xs_tcp_cancel_linger_timeout(struct rpc_xprt *xprt) | |
1487 | { | |
1488 | struct sock_xprt *transport; | |
1489 | ||
1490 | transport = container_of(xprt, struct sock_xprt, xprt); | |
1491 | ||
1492 | if (!test_bit(XPRT_CONNECTION_ABORT, &xprt->state) || | |
1493 | !cancel_delayed_work(&transport->connect_worker)) | |
1494 | return; | |
1495 | clear_bit(XPRT_CONNECTION_ABORT, &xprt->state); | |
1496 | xprt_clear_connecting(xprt); | |
1497 | } | |
1498 | ||
4bc1e68e | 1499 | static void xs_sock_reset_connection_flags(struct rpc_xprt *xprt) |
7d1e8255 | 1500 | { |
4e857c58 | 1501 | smp_mb__before_atomic(); |
a519fc7a TM |
1502 | clear_bit(XPRT_CONNECTION_ABORT, &xprt->state); |
1503 | clear_bit(XPRT_CONNECTION_CLOSE, &xprt->state); | |
7d1e8255 TM |
1504 | clear_bit(XPRT_CLOSE_WAIT, &xprt->state); |
1505 | clear_bit(XPRT_CLOSING, &xprt->state); | |
4e857c58 | 1506 | smp_mb__after_atomic(); |
4bc1e68e TM |
1507 | } |
1508 | ||
1509 | static void xs_sock_mark_closed(struct rpc_xprt *xprt) | |
1510 | { | |
1511 | xs_sock_reset_connection_flags(xprt); | |
7d1e8255 TM |
1512 | /* Mark transport as closed and wake up all pending tasks */ |
1513 | xprt_disconnect_done(xprt); | |
1514 | } | |
1515 | ||
9903cd1c CL |
1516 | /** |
1517 | * xs_tcp_state_change - callback to handle TCP socket state changes | |
1518 | * @sk: socket whose state has changed | |
1519 | * | |
1520 | */ | |
1521 | static void xs_tcp_state_change(struct sock *sk) | |
a246b010 | 1522 | { |
9903cd1c | 1523 | struct rpc_xprt *xprt; |
a246b010 | 1524 | |
f064af1e | 1525 | read_lock_bh(&sk->sk_callback_lock); |
a246b010 CL |
1526 | if (!(xprt = xprt_from_sock(sk))) |
1527 | goto out; | |
46121cf7 | 1528 | dprintk("RPC: xs_tcp_state_change client %p...\n", xprt); |
669502ff | 1529 | dprintk("RPC: state %x conn %d dead %d zapped %d sk_shutdown %d\n", |
46121cf7 CL |
1530 | sk->sk_state, xprt_connected(xprt), |
1531 | sock_flag(sk, SOCK_DEAD), | |
669502ff AC |
1532 | sock_flag(sk, SOCK_ZAPPED), |
1533 | sk->sk_shutdown); | |
a246b010 | 1534 | |
40b5ea0c | 1535 | trace_rpc_socket_state_change(xprt, sk->sk_socket); |
a246b010 CL |
1536 | switch (sk->sk_state) { |
1537 | case TCP_ESTABLISHED: | |
f064af1e | 1538 | spin_lock(&xprt->transport_lock); |
a246b010 | 1539 | if (!xprt_test_and_set_connected(xprt)) { |
51971139 CL |
1540 | struct sock_xprt *transport = container_of(xprt, |
1541 | struct sock_xprt, xprt); | |
1542 | ||
a246b010 | 1543 | /* Reset TCP record info */ |
51971139 CL |
1544 | transport->tcp_offset = 0; |
1545 | transport->tcp_reclen = 0; | |
1546 | transport->tcp_copied = 0; | |
e136d092 CL |
1547 | transport->tcp_flags = |
1548 | TCP_RCV_COPY_FRAGHDR | TCP_RCV_COPY_XID; | |
8b71798c | 1549 | xprt->connect_cookie++; |
51971139 | 1550 | |
2a491991 | 1551 | xprt_wake_pending_tasks(xprt, -EAGAIN); |
a246b010 | 1552 | } |
f064af1e | 1553 | spin_unlock(&xprt->transport_lock); |
a246b010 | 1554 | break; |
3b948ae5 TM |
1555 | case TCP_FIN_WAIT1: |
1556 | /* The client initiated a shutdown of the socket */ | |
7c1d71cf | 1557 | xprt->connect_cookie++; |
663b8858 | 1558 | xprt->reestablish_timeout = 0; |
3b948ae5 | 1559 | set_bit(XPRT_CLOSING, &xprt->state); |
4e857c58 | 1560 | smp_mb__before_atomic(); |
3b948ae5 | 1561 | clear_bit(XPRT_CONNECTED, &xprt->state); |
ef803670 | 1562 | clear_bit(XPRT_CLOSE_WAIT, &xprt->state); |
4e857c58 | 1563 | smp_mb__after_atomic(); |
25fe6142 | 1564 | xs_tcp_schedule_linger_timeout(xprt, xs_tcp_fin_timeout); |
a246b010 | 1565 | break; |
632e3bdc | 1566 | case TCP_CLOSE_WAIT: |
3b948ae5 | 1567 | /* The server initiated a shutdown of the socket */ |
7c1d71cf | 1568 | xprt->connect_cookie++; |
d0bea455 | 1569 | clear_bit(XPRT_CONNECTED, &xprt->state); |
a519fc7a | 1570 | xs_tcp_force_close(xprt); |
663b8858 TM |
1571 | case TCP_CLOSING: |
1572 | /* | |
1573 | * If the server closed down the connection, make sure that | |
1574 | * we back off before reconnecting | |
1575 | */ | |
1576 | if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO) | |
1577 | xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; | |
3b948ae5 TM |
1578 | break; |
1579 | case TCP_LAST_ACK: | |
670f9457 | 1580 | set_bit(XPRT_CLOSING, &xprt->state); |
25fe6142 | 1581 | xs_tcp_schedule_linger_timeout(xprt, xs_tcp_fin_timeout); |
4e857c58 | 1582 | smp_mb__before_atomic(); |
3b948ae5 | 1583 | clear_bit(XPRT_CONNECTED, &xprt->state); |
4e857c58 | 1584 | smp_mb__after_atomic(); |
3b948ae5 TM |
1585 | break; |
1586 | case TCP_CLOSE: | |
7d1e8255 TM |
1587 | xs_tcp_cancel_linger_timeout(xprt); |
1588 | xs_sock_mark_closed(xprt); | |
a246b010 CL |
1589 | } |
1590 | out: | |
f064af1e | 1591 | read_unlock_bh(&sk->sk_callback_lock); |
a246b010 CL |
1592 | } |
1593 | ||
1f0fa154 IJ |
1594 | static void xs_write_space(struct sock *sk) |
1595 | { | |
1596 | struct socket *sock; | |
1597 | struct rpc_xprt *xprt; | |
1598 | ||
1599 | if (unlikely(!(sock = sk->sk_socket))) | |
1600 | return; | |
1601 | clear_bit(SOCK_NOSPACE, &sock->flags); | |
1602 | ||
1603 | if (unlikely(!(xprt = xprt_from_sock(sk)))) | |
1604 | return; | |
1605 | if (test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags) == 0) | |
1606 | return; | |
1607 | ||
1608 | xprt_write_space(xprt); | |
1609 | } | |
1610 | ||
9903cd1c | 1611 | /** |
c7b2cae8 CL |
1612 | * xs_udp_write_space - callback invoked when socket buffer space |
1613 | * becomes available | |
9903cd1c CL |
1614 | * @sk: socket whose state has changed |
1615 | * | |
a246b010 CL |
1616 | * Called when more output buffer space is available for this socket. |
1617 | * We try not to wake our writers until they can make "significant" | |
c7b2cae8 | 1618 | * progress, otherwise we'll waste resources thrashing kernel_sendmsg |
a246b010 CL |
1619 | * with a bunch of small requests. |
1620 | */ | |
c7b2cae8 | 1621 | static void xs_udp_write_space(struct sock *sk) |
a246b010 | 1622 | { |
f064af1e | 1623 | read_lock_bh(&sk->sk_callback_lock); |
a246b010 | 1624 | |
c7b2cae8 | 1625 | /* from net/core/sock.c:sock_def_write_space */ |
1f0fa154 IJ |
1626 | if (sock_writeable(sk)) |
1627 | xs_write_space(sk); | |
a246b010 | 1628 | |
f064af1e | 1629 | read_unlock_bh(&sk->sk_callback_lock); |
c7b2cae8 | 1630 | } |
a246b010 | 1631 | |
c7b2cae8 CL |
1632 | /** |
1633 | * xs_tcp_write_space - callback invoked when socket buffer space | |
1634 | * becomes available | |
1635 | * @sk: socket whose state has changed | |
1636 | * | |
1637 | * Called when more output buffer space is available for this socket. | |
1638 | * We try not to wake our writers until they can make "significant" | |
1639 | * progress, otherwise we'll waste resources thrashing kernel_sendmsg | |
1640 | * with a bunch of small requests. | |
1641 | */ | |
1642 | static void xs_tcp_write_space(struct sock *sk) | |
1643 | { | |
f064af1e | 1644 | read_lock_bh(&sk->sk_callback_lock); |
c7b2cae8 CL |
1645 | |
1646 | /* from net/core/stream.c:sk_stream_write_space */ | |
64dc6130 | 1647 | if (sk_stream_is_writeable(sk)) |
1f0fa154 | 1648 | xs_write_space(sk); |
b6ddf64f | 1649 | |
f064af1e | 1650 | read_unlock_bh(&sk->sk_callback_lock); |
a246b010 CL |
1651 | } |
1652 | ||
470056c2 | 1653 | static void xs_udp_do_set_buffer_size(struct rpc_xprt *xprt) |
a246b010 | 1654 | { |
ee0ac0c2 CL |
1655 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
1656 | struct sock *sk = transport->inet; | |
a246b010 | 1657 | |
7c6e066e | 1658 | if (transport->rcvsize) { |
a246b010 | 1659 | sk->sk_userlocks |= SOCK_RCVBUF_LOCK; |
7c6e066e | 1660 | sk->sk_rcvbuf = transport->rcvsize * xprt->max_reqs * 2; |
a246b010 | 1661 | } |
7c6e066e | 1662 | if (transport->sndsize) { |
a246b010 | 1663 | sk->sk_userlocks |= SOCK_SNDBUF_LOCK; |
7c6e066e | 1664 | sk->sk_sndbuf = transport->sndsize * xprt->max_reqs * 2; |
a246b010 CL |
1665 | sk->sk_write_space(sk); |
1666 | } | |
1667 | } | |
1668 | ||
43118c29 | 1669 | /** |
470056c2 | 1670 | * xs_udp_set_buffer_size - set send and receive limits |
43118c29 | 1671 | * @xprt: generic transport |
470056c2 CL |
1672 | * @sndsize: requested size of send buffer, in bytes |
1673 | * @rcvsize: requested size of receive buffer, in bytes | |
43118c29 | 1674 | * |
470056c2 | 1675 | * Set socket send and receive buffer size limits. |
43118c29 | 1676 | */ |
470056c2 | 1677 | static void xs_udp_set_buffer_size(struct rpc_xprt *xprt, size_t sndsize, size_t rcvsize) |
43118c29 | 1678 | { |
7c6e066e CL |
1679 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
1680 | ||
1681 | transport->sndsize = 0; | |
470056c2 | 1682 | if (sndsize) |
7c6e066e CL |
1683 | transport->sndsize = sndsize + 1024; |
1684 | transport->rcvsize = 0; | |
470056c2 | 1685 | if (rcvsize) |
7c6e066e | 1686 | transport->rcvsize = rcvsize + 1024; |
470056c2 CL |
1687 | |
1688 | xs_udp_do_set_buffer_size(xprt); | |
43118c29 CL |
1689 | } |
1690 | ||
46c0ee8b CL |
1691 | /** |
1692 | * xs_udp_timer - called when a retransmit timeout occurs on a UDP transport | |
1693 | * @task: task that timed out | |
1694 | * | |
1695 | * Adjust the congestion window after a retransmit timeout has occurred. | |
1696 | */ | |
6a24dfb6 | 1697 | static void xs_udp_timer(struct rpc_xprt *xprt, struct rpc_task *task) |
46c0ee8b | 1698 | { |
6a24dfb6 | 1699 | xprt_adjust_cwnd(xprt, task, -ETIMEDOUT); |
46c0ee8b CL |
1700 | } |
1701 | ||
b85d8806 CL |
1702 | static unsigned short xs_get_random_port(void) |
1703 | { | |
1704 | unsigned short range = xprt_max_resvport - xprt_min_resvport; | |
63862b5b | 1705 | unsigned short rand = (unsigned short) prandom_u32() % range; |
b85d8806 CL |
1706 | return rand + xprt_min_resvport; |
1707 | } | |
1708 | ||
92200412 CL |
1709 | /** |
1710 | * xs_set_port - reset the port number in the remote endpoint address | |
1711 | * @xprt: generic transport | |
1712 | * @port: new port number | |
1713 | * | |
1714 | */ | |
1715 | static void xs_set_port(struct rpc_xprt *xprt, unsigned short port) | |
1716 | { | |
46121cf7 | 1717 | dprintk("RPC: setting port for xprt %p to %u\n", xprt, port); |
c4efcb1d | 1718 | |
9dc3b095 CL |
1719 | rpc_set_port(xs_addr(xprt), port); |
1720 | xs_update_peer_port(xprt); | |
92200412 CL |
1721 | } |
1722 | ||
5d4ec932 | 1723 | static unsigned short xs_get_srcport(struct sock_xprt *transport) |
67a391d7 | 1724 | { |
fbfffbd5 | 1725 | unsigned short port = transport->srcport; |
67a391d7 TM |
1726 | |
1727 | if (port == 0 && transport->xprt.resvport) | |
1728 | port = xs_get_random_port(); | |
1729 | return port; | |
1730 | } | |
1731 | ||
baaf4e48 | 1732 | static unsigned short xs_next_srcport(struct sock_xprt *transport, unsigned short port) |
67a391d7 | 1733 | { |
fbfffbd5 CL |
1734 | if (transport->srcport != 0) |
1735 | transport->srcport = 0; | |
67a391d7 TM |
1736 | if (!transport->xprt.resvport) |
1737 | return 0; | |
1738 | if (port <= xprt_min_resvport || port > xprt_max_resvport) | |
1739 | return xprt_max_resvport; | |
1740 | return --port; | |
1741 | } | |
beb59b68 | 1742 | static int xs_bind(struct sock_xprt *transport, struct socket *sock) |
a246b010 | 1743 | { |
beb59b68 | 1744 | struct sockaddr_storage myaddr; |
67a391d7 | 1745 | int err, nloop = 0; |
5d4ec932 | 1746 | unsigned short port = xs_get_srcport(transport); |
67a391d7 | 1747 | unsigned short last; |
a246b010 | 1748 | |
beb59b68 | 1749 | memcpy(&myaddr, &transport->srcaddr, transport->xprt.addrlen); |
a246b010 | 1750 | do { |
beb59b68 PE |
1751 | rpc_set_port((struct sockaddr *)&myaddr, port); |
1752 | err = kernel_bind(sock, (struct sockaddr *)&myaddr, | |
1753 | transport->xprt.addrlen); | |
67a391d7 | 1754 | if (port == 0) |
d3bc9a1d | 1755 | break; |
a246b010 | 1756 | if (err == 0) { |
fbfffbd5 | 1757 | transport->srcport = port; |
d3bc9a1d | 1758 | break; |
a246b010 | 1759 | } |
67a391d7 | 1760 | last = port; |
baaf4e48 | 1761 | port = xs_next_srcport(transport, port); |
67a391d7 TM |
1762 | if (port > last) |
1763 | nloop++; | |
1764 | } while (err == -EADDRINUSE && nloop != 2); | |
90058d37 | 1765 | |
4232e863 | 1766 | if (myaddr.ss_family == AF_INET) |
beb59b68 PE |
1767 | dprintk("RPC: %s %pI4:%u: %s (%d)\n", __func__, |
1768 | &((struct sockaddr_in *)&myaddr)->sin_addr, | |
1769 | port, err ? "failed" : "ok", err); | |
1770 | else | |
1771 | dprintk("RPC: %s %pI6:%u: %s (%d)\n", __func__, | |
1772 | &((struct sockaddr_in6 *)&myaddr)->sin6_addr, | |
1773 | port, err ? "failed" : "ok", err); | |
a246b010 CL |
1774 | return err; |
1775 | } | |
1776 | ||
176e21ee CL |
1777 | /* |
1778 | * We don't support autobind on AF_LOCAL sockets | |
1779 | */ | |
1780 | static void xs_local_rpcbind(struct rpc_task *task) | |
1781 | { | |
3dc0da27 TM |
1782 | rcu_read_lock(); |
1783 | xprt_set_bound(rcu_dereference(task->tk_client->cl_xprt)); | |
1784 | rcu_read_unlock(); | |
176e21ee CL |
1785 | } |
1786 | ||
1787 | static void xs_local_set_port(struct rpc_xprt *xprt, unsigned short port) | |
1788 | { | |
1789 | } | |
beb59b68 | 1790 | |
ed07536e PZ |
1791 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
1792 | static struct lock_class_key xs_key[2]; | |
1793 | static struct lock_class_key xs_slock_key[2]; | |
1794 | ||
176e21ee CL |
1795 | static inline void xs_reclassify_socketu(struct socket *sock) |
1796 | { | |
1797 | struct sock *sk = sock->sk; | |
1798 | ||
176e21ee CL |
1799 | sock_lock_init_class_and_name(sk, "slock-AF_LOCAL-RPC", |
1800 | &xs_slock_key[1], "sk_lock-AF_LOCAL-RPC", &xs_key[1]); | |
1801 | } | |
1802 | ||
8945ee5e | 1803 | static inline void xs_reclassify_socket4(struct socket *sock) |
ed07536e PZ |
1804 | { |
1805 | struct sock *sk = sock->sk; | |
8945ee5e | 1806 | |
8945ee5e CL |
1807 | sock_lock_init_class_and_name(sk, "slock-AF_INET-RPC", |
1808 | &xs_slock_key[0], "sk_lock-AF_INET-RPC", &xs_key[0]); | |
1809 | } | |
ed07536e | 1810 | |
8945ee5e CL |
1811 | static inline void xs_reclassify_socket6(struct socket *sock) |
1812 | { | |
1813 | struct sock *sk = sock->sk; | |
ed07536e | 1814 | |
8945ee5e CL |
1815 | sock_lock_init_class_and_name(sk, "slock-AF_INET6-RPC", |
1816 | &xs_slock_key[1], "sk_lock-AF_INET6-RPC", &xs_key[1]); | |
ed07536e | 1817 | } |
6bc9638a PE |
1818 | |
1819 | static inline void xs_reclassify_socket(int family, struct socket *sock) | |
1820 | { | |
1b7a1819 WAA |
1821 | WARN_ON_ONCE(sock_owned_by_user(sock->sk)); |
1822 | if (sock_owned_by_user(sock->sk)) | |
1823 | return; | |
1824 | ||
4232e863 | 1825 | switch (family) { |
176e21ee CL |
1826 | case AF_LOCAL: |
1827 | xs_reclassify_socketu(sock); | |
1828 | break; | |
4232e863 | 1829 | case AF_INET: |
6bc9638a | 1830 | xs_reclassify_socket4(sock); |
4232e863 CL |
1831 | break; |
1832 | case AF_INET6: | |
6bc9638a | 1833 | xs_reclassify_socket6(sock); |
4232e863 CL |
1834 | break; |
1835 | } | |
6bc9638a | 1836 | } |
ed07536e | 1837 | #else |
176e21ee CL |
1838 | static inline void xs_reclassify_socketu(struct socket *sock) |
1839 | { | |
1840 | } | |
1841 | ||
8945ee5e CL |
1842 | static inline void xs_reclassify_socket4(struct socket *sock) |
1843 | { | |
1844 | } | |
1845 | ||
1846 | static inline void xs_reclassify_socket6(struct socket *sock) | |
ed07536e PZ |
1847 | { |
1848 | } | |
6bc9638a PE |
1849 | |
1850 | static inline void xs_reclassify_socket(int family, struct socket *sock) | |
1851 | { | |
1852 | } | |
ed07536e PZ |
1853 | #endif |
1854 | ||
93dc41bd N |
1855 | static void xs_dummy_setup_socket(struct work_struct *work) |
1856 | { | |
1857 | } | |
1858 | ||
6bc9638a PE |
1859 | static struct socket *xs_create_sock(struct rpc_xprt *xprt, |
1860 | struct sock_xprt *transport, int family, int type, int protocol) | |
22f79326 PE |
1861 | { |
1862 | struct socket *sock; | |
1863 | int err; | |
1864 | ||
6bc9638a | 1865 | err = __sock_create(xprt->xprt_net, family, type, protocol, &sock, 1); |
22f79326 PE |
1866 | if (err < 0) { |
1867 | dprintk("RPC: can't create %d transport socket (%d).\n", | |
1868 | protocol, -err); | |
1869 | goto out; | |
1870 | } | |
6bc9638a | 1871 | xs_reclassify_socket(family, sock); |
22f79326 | 1872 | |
4cea288a BH |
1873 | err = xs_bind(transport, sock); |
1874 | if (err) { | |
22f79326 PE |
1875 | sock_release(sock); |
1876 | goto out; | |
1877 | } | |
1878 | ||
1879 | return sock; | |
1880 | out: | |
1881 | return ERR_PTR(err); | |
1882 | } | |
1883 | ||
176e21ee CL |
1884 | static int xs_local_finish_connecting(struct rpc_xprt *xprt, |
1885 | struct socket *sock) | |
1886 | { | |
1887 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, | |
1888 | xprt); | |
1889 | ||
1890 | if (!transport->inet) { | |
1891 | struct sock *sk = sock->sk; | |
1892 | ||
1893 | write_lock_bh(&sk->sk_callback_lock); | |
1894 | ||
1895 | xs_save_old_callbacks(transport, sk); | |
1896 | ||
1897 | sk->sk_user_data = xprt; | |
1898 | sk->sk_data_ready = xs_local_data_ready; | |
1899 | sk->sk_write_space = xs_udp_write_space; | |
2118071d | 1900 | sk->sk_error_report = xs_error_report; |
176e21ee CL |
1901 | sk->sk_allocation = GFP_ATOMIC; |
1902 | ||
1903 | xprt_clear_connected(xprt); | |
1904 | ||
1905 | /* Reset to new socket */ | |
1906 | transport->sock = sock; | |
1907 | transport->inet = sk; | |
1908 | ||
1909 | write_unlock_bh(&sk->sk_callback_lock); | |
1910 | } | |
1911 | ||
1912 | /* Tell the socket layer to start connecting... */ | |
1913 | xprt->stat.connect_count++; | |
1914 | xprt->stat.connect_start = jiffies; | |
1915 | return kernel_connect(sock, xs_addr(xprt), xprt->addrlen, 0); | |
1916 | } | |
1917 | ||
1918 | /** | |
1919 | * xs_local_setup_socket - create AF_LOCAL socket, connect to a local endpoint | |
1920 | * @xprt: RPC transport to connect | |
1921 | * @transport: socket transport to connect | |
1922 | * @create_sock: function to create a socket of the correct type | |
176e21ee | 1923 | */ |
dc107402 | 1924 | static int xs_local_setup_socket(struct sock_xprt *transport) |
176e21ee | 1925 | { |
176e21ee CL |
1926 | struct rpc_xprt *xprt = &transport->xprt; |
1927 | struct socket *sock; | |
1928 | int status = -EIO; | |
1929 | ||
5cf02d09 JL |
1930 | current->flags |= PF_FSTRANS; |
1931 | ||
176e21ee CL |
1932 | clear_bit(XPRT_CONNECTION_ABORT, &xprt->state); |
1933 | status = __sock_create(xprt->xprt_net, AF_LOCAL, | |
1934 | SOCK_STREAM, 0, &sock, 1); | |
1935 | if (status < 0) { | |
1936 | dprintk("RPC: can't create AF_LOCAL " | |
1937 | "transport socket (%d).\n", -status); | |
1938 | goto out; | |
1939 | } | |
1940 | xs_reclassify_socketu(sock); | |
1941 | ||
1942 | dprintk("RPC: worker connecting xprt %p via AF_LOCAL to %s\n", | |
1943 | xprt, xprt->address_strings[RPC_DISPLAY_ADDR]); | |
1944 | ||
1945 | status = xs_local_finish_connecting(xprt, sock); | |
40b5ea0c | 1946 | trace_rpc_socket_connect(xprt, sock, status); |
176e21ee CL |
1947 | switch (status) { |
1948 | case 0: | |
1949 | dprintk("RPC: xprt %p connected to %s\n", | |
1950 | xprt, xprt->address_strings[RPC_DISPLAY_ADDR]); | |
1951 | xprt_set_connected(xprt); | |
3601c4a9 | 1952 | case -ENOBUFS: |
176e21ee CL |
1953 | break; |
1954 | case -ENOENT: | |
1955 | dprintk("RPC: xprt %p: socket %s does not exist\n", | |
1956 | xprt, xprt->address_strings[RPC_DISPLAY_ADDR]); | |
1957 | break; | |
4a20a988 TM |
1958 | case -ECONNREFUSED: |
1959 | dprintk("RPC: xprt %p: connection refused for %s\n", | |
1960 | xprt, xprt->address_strings[RPC_DISPLAY_ADDR]); | |
1961 | break; | |
176e21ee CL |
1962 | default: |
1963 | printk(KERN_ERR "%s: unhandled error (%d) connecting to %s\n", | |
1964 | __func__, -status, | |
1965 | xprt->address_strings[RPC_DISPLAY_ADDR]); | |
1966 | } | |
1967 | ||
1968 | out: | |
1969 | xprt_clear_connecting(xprt); | |
1970 | xprt_wake_pending_tasks(xprt, status); | |
5cf02d09 | 1971 | current->flags &= ~PF_FSTRANS; |
dc107402 BF |
1972 | return status; |
1973 | } | |
1974 | ||
b6669737 | 1975 | static void xs_local_connect(struct rpc_xprt *xprt, struct rpc_task *task) |
dc107402 | 1976 | { |
dc107402 BF |
1977 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
1978 | int ret; | |
1979 | ||
1980 | if (RPC_IS_ASYNC(task)) { | |
1981 | /* | |
1982 | * We want the AF_LOCAL connect to be resolved in the | |
1983 | * filesystem namespace of the process making the rpc | |
1984 | * call. Thus we connect synchronously. | |
1985 | * | |
1986 | * If we want to support asynchronous AF_LOCAL calls, | |
1987 | * we'll need to figure out how to pass a namespace to | |
1988 | * connect. | |
1989 | */ | |
1990 | rpc_exit(task, -ENOTCONN); | |
1991 | return; | |
1992 | } | |
1993 | ret = xs_local_setup_socket(transport); | |
1994 | if (ret && !RPC_IS_SOFTCONN(task)) | |
1995 | msleep_interruptible(15000); | |
176e21ee CL |
1996 | } |
1997 | ||
a564b8f0 MG |
1998 | #ifdef CONFIG_SUNRPC_SWAP |
1999 | static void xs_set_memalloc(struct rpc_xprt *xprt) | |
2000 | { | |
2001 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, | |
2002 | xprt); | |
2003 | ||
2004 | if (xprt->swapper) | |
2005 | sk_set_memalloc(transport->inet); | |
2006 | } | |
2007 | ||
2008 | /** | |
2009 | * xs_swapper - Tag this transport as being used for swap. | |
2010 | * @xprt: transport to tag | |
2011 | * @enable: enable/disable | |
2012 | * | |
2013 | */ | |
2014 | int xs_swapper(struct rpc_xprt *xprt, int enable) | |
2015 | { | |
2016 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, | |
2017 | xprt); | |
2018 | int err = 0; | |
2019 | ||
2020 | if (enable) { | |
2021 | xprt->swapper++; | |
2022 | xs_set_memalloc(xprt); | |
2023 | } else if (xprt->swapper) { | |
2024 | xprt->swapper--; | |
2025 | sk_clear_memalloc(transport->inet); | |
2026 | } | |
2027 | ||
2028 | return err; | |
2029 | } | |
2030 | EXPORT_SYMBOL_GPL(xs_swapper); | |
2031 | #else | |
2032 | static void xs_set_memalloc(struct rpc_xprt *xprt) | |
2033 | { | |
2034 | } | |
2035 | #endif | |
2036 | ||
16be2d20 CL |
2037 | static void xs_udp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock) |
2038 | { | |
2039 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); | |
2040 | ||
2041 | if (!transport->inet) { | |
2042 | struct sock *sk = sock->sk; | |
2043 | ||
2044 | write_lock_bh(&sk->sk_callback_lock); | |
2045 | ||
2a9e1cfa TM |
2046 | xs_save_old_callbacks(transport, sk); |
2047 | ||
16be2d20 | 2048 | sk->sk_user_data = xprt; |
16be2d20 CL |
2049 | sk->sk_data_ready = xs_udp_data_ready; |
2050 | sk->sk_write_space = xs_udp_write_space; | |
16be2d20 CL |
2051 | sk->sk_allocation = GFP_ATOMIC; |
2052 | ||
2053 | xprt_set_connected(xprt); | |
2054 | ||
2055 | /* Reset to new socket */ | |
2056 | transport->sock = sock; | |
2057 | transport->inet = sk; | |
2058 | ||
a564b8f0 MG |
2059 | xs_set_memalloc(xprt); |
2060 | ||
16be2d20 CL |
2061 | write_unlock_bh(&sk->sk_callback_lock); |
2062 | } | |
2063 | xs_udp_do_set_buffer_size(xprt); | |
2064 | } | |
2065 | ||
8c14ff2a | 2066 | static void xs_udp_setup_socket(struct work_struct *work) |
a246b010 | 2067 | { |
34161db6 TM |
2068 | struct sock_xprt *transport = |
2069 | container_of(work, struct sock_xprt, connect_worker.work); | |
c8475461 | 2070 | struct rpc_xprt *xprt = &transport->xprt; |
ee0ac0c2 | 2071 | struct socket *sock = transport->sock; |
b65c0310 | 2072 | int status = -EIO; |
9903cd1c | 2073 | |
5cf02d09 JL |
2074 | current->flags |= PF_FSTRANS; |
2075 | ||
b0d93ad5 | 2076 | /* Start by resetting any existing state */ |
fe315e76 | 2077 | xs_reset_transport(transport); |
8c14ff2a PE |
2078 | sock = xs_create_sock(xprt, transport, |
2079 | xs_addr(xprt)->sa_family, SOCK_DGRAM, IPPROTO_UDP); | |
b65c0310 | 2080 | if (IS_ERR(sock)) |
b0d93ad5 | 2081 | goto out; |
68e220bd | 2082 | |
c740eff8 CL |
2083 | dprintk("RPC: worker connecting xprt %p via %s to " |
2084 | "%s (port %s)\n", xprt, | |
2085 | xprt->address_strings[RPC_DISPLAY_PROTO], | |
2086 | xprt->address_strings[RPC_DISPLAY_ADDR], | |
2087 | xprt->address_strings[RPC_DISPLAY_PORT]); | |
68e220bd CL |
2088 | |
2089 | xs_udp_finish_connecting(xprt, sock); | |
40b5ea0c | 2090 | trace_rpc_socket_connect(xprt, sock, 0); |
b0d93ad5 CL |
2091 | status = 0; |
2092 | out: | |
b0d93ad5 | 2093 | xprt_clear_connecting(xprt); |
7d1e8255 | 2094 | xprt_wake_pending_tasks(xprt, status); |
5cf02d09 | 2095 | current->flags &= ~PF_FSTRANS; |
a246b010 CL |
2096 | } |
2097 | ||
3167e12c CL |
2098 | /* |
2099 | * We need to preserve the port number so the reply cache on the server can | |
2100 | * find our cached RPC replies when we get around to reconnecting. | |
2101 | */ | |
58dddac9 | 2102 | static void xs_abort_connection(struct sock_xprt *transport) |
3167e12c CL |
2103 | { |
2104 | int result; | |
3167e12c CL |
2105 | struct sockaddr any; |
2106 | ||
58dddac9 | 2107 | dprintk("RPC: disconnecting xprt %p to reuse port\n", transport); |
3167e12c CL |
2108 | |
2109 | /* | |
2110 | * Disconnect the transport socket by doing a connect operation | |
2111 | * with AF_UNSPEC. This should return immediately... | |
2112 | */ | |
2113 | memset(&any, 0, sizeof(any)); | |
2114 | any.sa_family = AF_UNSPEC; | |
ee0ac0c2 | 2115 | result = kernel_connect(transport->sock, &any, sizeof(any), 0); |
40b5ea0c TM |
2116 | trace_rpc_socket_reset_connection(&transport->xprt, |
2117 | transport->sock, result); | |
7d1e8255 | 2118 | if (!result) |
4bc1e68e TM |
2119 | xs_sock_reset_connection_flags(&transport->xprt); |
2120 | dprintk("RPC: AF_UNSPEC connect return code %d\n", result); | |
3167e12c CL |
2121 | } |
2122 | ||
58dddac9 | 2123 | static void xs_tcp_reuse_connection(struct sock_xprt *transport) |
40d2549d TM |
2124 | { |
2125 | unsigned int state = transport->inet->sk_state; | |
2126 | ||
669502ff AC |
2127 | if (state == TCP_CLOSE && transport->sock->state == SS_UNCONNECTED) { |
2128 | /* we don't need to abort the connection if the socket | |
2129 | * hasn't undergone a shutdown | |
2130 | */ | |
2131 | if (transport->inet->sk_shutdown == 0) | |
2132 | return; | |
2133 | dprintk("RPC: %s: TCP_CLOSEd and sk_shutdown set to %d\n", | |
2134 | __func__, transport->inet->sk_shutdown); | |
2135 | } | |
2136 | if ((1 << state) & (TCPF_ESTABLISHED|TCPF_SYN_SENT)) { | |
2137 | /* we don't need to abort the connection if the socket | |
2138 | * hasn't undergone a shutdown | |
2139 | */ | |
2140 | if (transport->inet->sk_shutdown == 0) | |
2141 | return; | |
2142 | dprintk("RPC: %s: ESTABLISHED/SYN_SENT " | |
2143 | "sk_shutdown set to %d\n", | |
2144 | __func__, transport->inet->sk_shutdown); | |
2145 | } | |
58dddac9 | 2146 | xs_abort_connection(transport); |
40d2549d TM |
2147 | } |
2148 | ||
16be2d20 | 2149 | static int xs_tcp_finish_connecting(struct rpc_xprt *xprt, struct socket *sock) |
a246b010 | 2150 | { |
16be2d20 | 2151 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
fe19a96b | 2152 | int ret = -ENOTCONN; |
edb267a6 | 2153 | |
ee0ac0c2 | 2154 | if (!transport->inet) { |
b0d93ad5 | 2155 | struct sock *sk = sock->sk; |
7f260e85 TM |
2156 | unsigned int keepidle = xprt->timeout->to_initval / HZ; |
2157 | unsigned int keepcnt = xprt->timeout->to_retries + 1; | |
2158 | unsigned int opt_on = 1; | |
2159 | ||
2160 | /* TCP Keepalive options */ | |
2161 | kernel_setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE, | |
2162 | (char *)&opt_on, sizeof(opt_on)); | |
2163 | kernel_setsockopt(sock, SOL_TCP, TCP_KEEPIDLE, | |
2164 | (char *)&keepidle, sizeof(keepidle)); | |
2165 | kernel_setsockopt(sock, SOL_TCP, TCP_KEEPINTVL, | |
2166 | (char *)&keepidle, sizeof(keepidle)); | |
2167 | kernel_setsockopt(sock, SOL_TCP, TCP_KEEPCNT, | |
2168 | (char *)&keepcnt, sizeof(keepcnt)); | |
b0d93ad5 CL |
2169 | |
2170 | write_lock_bh(&sk->sk_callback_lock); | |
2171 | ||
2a9e1cfa TM |
2172 | xs_save_old_callbacks(transport, sk); |
2173 | ||
b0d93ad5 | 2174 | sk->sk_user_data = xprt; |
b0d93ad5 CL |
2175 | sk->sk_data_ready = xs_tcp_data_ready; |
2176 | sk->sk_state_change = xs_tcp_state_change; | |
2177 | sk->sk_write_space = xs_tcp_write_space; | |
2118071d | 2178 | sk->sk_error_report = xs_error_report; |
b079fa7b | 2179 | sk->sk_allocation = GFP_ATOMIC; |
3167e12c CL |
2180 | |
2181 | /* socket options */ | |
2182 | sk->sk_userlocks |= SOCK_BINDPORT_LOCK; | |
2183 | sock_reset_flag(sk, SOCK_LINGER); | |
2184 | tcp_sk(sk)->linger2 = 0; | |
2185 | tcp_sk(sk)->nonagle |= TCP_NAGLE_OFF; | |
b0d93ad5 CL |
2186 | |
2187 | xprt_clear_connected(xprt); | |
2188 | ||
2189 | /* Reset to new socket */ | |
ee0ac0c2 CL |
2190 | transport->sock = sock; |
2191 | transport->inet = sk; | |
b0d93ad5 CL |
2192 | |
2193 | write_unlock_bh(&sk->sk_callback_lock); | |
2194 | } | |
2195 | ||
01d37c42 | 2196 | if (!xprt_bound(xprt)) |
fe19a96b | 2197 | goto out; |
01d37c42 | 2198 | |
a564b8f0 MG |
2199 | xs_set_memalloc(xprt); |
2200 | ||
b0d93ad5 | 2201 | /* Tell the socket layer to start connecting... */ |
262ca07d CL |
2202 | xprt->stat.connect_count++; |
2203 | xprt->stat.connect_start = jiffies; | |
fe19a96b TM |
2204 | ret = kernel_connect(sock, xs_addr(xprt), xprt->addrlen, O_NONBLOCK); |
2205 | switch (ret) { | |
2206 | case 0: | |
2207 | case -EINPROGRESS: | |
2208 | /* SYN_SENT! */ | |
fe19a96b TM |
2209 | if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO) |
2210 | xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; | |
2211 | } | |
2212 | out: | |
2213 | return ret; | |
16be2d20 CL |
2214 | } |
2215 | ||
9903cd1c | 2216 | /** |
b61d59ff TM |
2217 | * xs_tcp_setup_socket - create a TCP socket and connect to a remote endpoint |
2218 | * @xprt: RPC transport to connect | |
2219 | * @transport: socket transport to connect | |
2220 | * @create_sock: function to create a socket of the correct type | |
9903cd1c CL |
2221 | * |
2222 | * Invoked by a work queue tasklet. | |
a246b010 | 2223 | */ |
cdd518d5 | 2224 | static void xs_tcp_setup_socket(struct work_struct *work) |
a246b010 | 2225 | { |
cdd518d5 PE |
2226 | struct sock_xprt *transport = |
2227 | container_of(work, struct sock_xprt, connect_worker.work); | |
ee0ac0c2 | 2228 | struct socket *sock = transport->sock; |
a9f5f0f7 | 2229 | struct rpc_xprt *xprt = &transport->xprt; |
b61d59ff | 2230 | int status = -EIO; |
a246b010 | 2231 | |
5cf02d09 JL |
2232 | current->flags |= PF_FSTRANS; |
2233 | ||
ee0ac0c2 | 2234 | if (!sock) { |
7d1e8255 | 2235 | clear_bit(XPRT_CONNECTION_ABORT, &xprt->state); |
cdd518d5 PE |
2236 | sock = xs_create_sock(xprt, transport, |
2237 | xs_addr(xprt)->sa_family, SOCK_STREAM, IPPROTO_TCP); | |
b61d59ff TM |
2238 | if (IS_ERR(sock)) { |
2239 | status = PTR_ERR(sock); | |
3167e12c CL |
2240 | goto out; |
2241 | } | |
7d1e8255 TM |
2242 | } else { |
2243 | int abort_and_exit; | |
a246b010 | 2244 | |
7d1e8255 TM |
2245 | abort_and_exit = test_and_clear_bit(XPRT_CONNECTION_ABORT, |
2246 | &xprt->state); | |
3167e12c | 2247 | /* "close" the socket, preserving the local port */ |
58dddac9 | 2248 | xs_tcp_reuse_connection(transport); |
a246b010 | 2249 | |
7d1e8255 TM |
2250 | if (abort_and_exit) |
2251 | goto out_eagain; | |
2252 | } | |
a246b010 | 2253 | |
c740eff8 CL |
2254 | dprintk("RPC: worker connecting xprt %p via %s to " |
2255 | "%s (port %s)\n", xprt, | |
2256 | xprt->address_strings[RPC_DISPLAY_PROTO], | |
2257 | xprt->address_strings[RPC_DISPLAY_ADDR], | |
2258 | xprt->address_strings[RPC_DISPLAY_PORT]); | |
edb267a6 | 2259 | |
16be2d20 | 2260 | status = xs_tcp_finish_connecting(xprt, sock); |
40b5ea0c | 2261 | trace_rpc_socket_connect(xprt, sock, status); |
46121cf7 CL |
2262 | dprintk("RPC: %p connect status %d connected %d sock state %d\n", |
2263 | xprt, -status, xprt_connected(xprt), | |
2264 | sock->sk->sk_state); | |
2a491991 | 2265 | switch (status) { |
f75e6745 TM |
2266 | default: |
2267 | printk("%s: connect returned unhandled error %d\n", | |
2268 | __func__, status); | |
2269 | case -EADDRNOTAVAIL: | |
2270 | /* We're probably in TIME_WAIT. Get rid of existing socket, | |
2271 | * and retry | |
2272 | */ | |
a519fc7a | 2273 | xs_tcp_force_close(xprt); |
88b5ed73 | 2274 | break; |
2a491991 TM |
2275 | case 0: |
2276 | case -EINPROGRESS: | |
2277 | case -EALREADY: | |
7d1e8255 | 2278 | xprt_clear_connecting(xprt); |
5cf02d09 | 2279 | current->flags &= ~PF_FSTRANS; |
7d1e8255 | 2280 | return; |
9fcfe0c8 TM |
2281 | case -EINVAL: |
2282 | /* Happens, for instance, if the user specified a link | |
2283 | * local IPv6 address without a scope-id. | |
2284 | */ | |
3ed5e2a2 TM |
2285 | case -ECONNREFUSED: |
2286 | case -ECONNRESET: | |
2287 | case -ENETUNREACH: | |
3601c4a9 | 2288 | case -ENOBUFS: |
3ed5e2a2 | 2289 | /* retry with existing socket, after a delay */ |
9fcfe0c8 | 2290 | goto out; |
a246b010 | 2291 | } |
7d1e8255 | 2292 | out_eagain: |
2a491991 | 2293 | status = -EAGAIN; |
a246b010 | 2294 | out: |
68e220bd | 2295 | xprt_clear_connecting(xprt); |
7d1e8255 | 2296 | xprt_wake_pending_tasks(xprt, status); |
5cf02d09 | 2297 | current->flags &= ~PF_FSTRANS; |
68e220bd | 2298 | } |
b0d93ad5 | 2299 | |
9903cd1c CL |
2300 | /** |
2301 | * xs_connect - connect a socket to a remote endpoint | |
1b092092 | 2302 | * @xprt: pointer to transport structure |
9903cd1c CL |
2303 | * @task: address of RPC task that manages state of connect request |
2304 | * | |
2305 | * TCP: If the remote end dropped the connection, delay reconnecting. | |
03bf4b70 CL |
2306 | * |
2307 | * UDP socket connects are synchronous, but we use a work queue anyway | |
2308 | * to guarantee that even unprivileged user processes can set up a | |
2309 | * socket on a privileged port. | |
2310 | * | |
2311 | * If a UDP socket connect fails, the delay behavior here prevents | |
2312 | * retry floods (hard mounts). | |
9903cd1c | 2313 | */ |
1b092092 | 2314 | static void xs_connect(struct rpc_xprt *xprt, struct rpc_task *task) |
a246b010 | 2315 | { |
ee0ac0c2 | 2316 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
a246b010 | 2317 | |
09a21c41 | 2318 | if (transport->sock != NULL && !RPC_IS_SOFTCONN(task)) { |
46121cf7 CL |
2319 | dprintk("RPC: xs_connect delayed xprt %p for %lu " |
2320 | "seconds\n", | |
03bf4b70 | 2321 | xprt, xprt->reestablish_timeout / HZ); |
c1384c9c TM |
2322 | queue_delayed_work(rpciod_workqueue, |
2323 | &transport->connect_worker, | |
2324 | xprt->reestablish_timeout); | |
03bf4b70 | 2325 | xprt->reestablish_timeout <<= 1; |
61d0a8e6 NB |
2326 | if (xprt->reestablish_timeout < XS_TCP_INIT_REEST_TO) |
2327 | xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; | |
03bf4b70 CL |
2328 | if (xprt->reestablish_timeout > XS_TCP_MAX_REEST_TO) |
2329 | xprt->reestablish_timeout = XS_TCP_MAX_REEST_TO; | |
b0d93ad5 | 2330 | } else { |
46121cf7 | 2331 | dprintk("RPC: xs_connect scheduled xprt %p\n", xprt); |
c1384c9c TM |
2332 | queue_delayed_work(rpciod_workqueue, |
2333 | &transport->connect_worker, 0); | |
a246b010 CL |
2334 | } |
2335 | } | |
2336 | ||
176e21ee CL |
2337 | /** |
2338 | * xs_local_print_stats - display AF_LOCAL socket-specifc stats | |
2339 | * @xprt: rpc_xprt struct containing statistics | |
2340 | * @seq: output file | |
2341 | * | |
2342 | */ | |
2343 | static void xs_local_print_stats(struct rpc_xprt *xprt, struct seq_file *seq) | |
2344 | { | |
2345 | long idle_time = 0; | |
2346 | ||
2347 | if (xprt_connected(xprt)) | |
2348 | idle_time = (long)(jiffies - xprt->last_used) / HZ; | |
2349 | ||
2350 | seq_printf(seq, "\txprt:\tlocal %lu %lu %lu %ld %lu %lu %lu " | |
15a45206 | 2351 | "%llu %llu %lu %llu %llu\n", |
176e21ee CL |
2352 | xprt->stat.bind_count, |
2353 | xprt->stat.connect_count, | |
2354 | xprt->stat.connect_time, | |
2355 | idle_time, | |
2356 | xprt->stat.sends, | |
2357 | xprt->stat.recvs, | |
2358 | xprt->stat.bad_xids, | |
2359 | xprt->stat.req_u, | |
15a45206 AA |
2360 | xprt->stat.bklog_u, |
2361 | xprt->stat.max_slots, | |
2362 | xprt->stat.sending_u, | |
2363 | xprt->stat.pending_u); | |
176e21ee CL |
2364 | } |
2365 | ||
262ca07d CL |
2366 | /** |
2367 | * xs_udp_print_stats - display UDP socket-specifc stats | |
2368 | * @xprt: rpc_xprt struct containing statistics | |
2369 | * @seq: output file | |
2370 | * | |
2371 | */ | |
2372 | static void xs_udp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq) | |
2373 | { | |
c8475461 CL |
2374 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
2375 | ||
15a45206 AA |
2376 | seq_printf(seq, "\txprt:\tudp %u %lu %lu %lu %lu %llu %llu " |
2377 | "%lu %llu %llu\n", | |
fbfffbd5 | 2378 | transport->srcport, |
262ca07d CL |
2379 | xprt->stat.bind_count, |
2380 | xprt->stat.sends, | |
2381 | xprt->stat.recvs, | |
2382 | xprt->stat.bad_xids, | |
2383 | xprt->stat.req_u, | |
15a45206 AA |
2384 | xprt->stat.bklog_u, |
2385 | xprt->stat.max_slots, | |
2386 | xprt->stat.sending_u, | |
2387 | xprt->stat.pending_u); | |
262ca07d CL |
2388 | } |
2389 | ||
2390 | /** | |
2391 | * xs_tcp_print_stats - display TCP socket-specifc stats | |
2392 | * @xprt: rpc_xprt struct containing statistics | |
2393 | * @seq: output file | |
2394 | * | |
2395 | */ | |
2396 | static void xs_tcp_print_stats(struct rpc_xprt *xprt, struct seq_file *seq) | |
2397 | { | |
c8475461 | 2398 | struct sock_xprt *transport = container_of(xprt, struct sock_xprt, xprt); |
262ca07d CL |
2399 | long idle_time = 0; |
2400 | ||
2401 | if (xprt_connected(xprt)) | |
2402 | idle_time = (long)(jiffies - xprt->last_used) / HZ; | |
2403 | ||
15a45206 AA |
2404 | seq_printf(seq, "\txprt:\ttcp %u %lu %lu %lu %ld %lu %lu %lu " |
2405 | "%llu %llu %lu %llu %llu\n", | |
fbfffbd5 | 2406 | transport->srcport, |
262ca07d CL |
2407 | xprt->stat.bind_count, |
2408 | xprt->stat.connect_count, | |
2409 | xprt->stat.connect_time, | |
2410 | idle_time, | |
2411 | xprt->stat.sends, | |
2412 | xprt->stat.recvs, | |
2413 | xprt->stat.bad_xids, | |
2414 | xprt->stat.req_u, | |
15a45206 AA |
2415 | xprt->stat.bklog_u, |
2416 | xprt->stat.max_slots, | |
2417 | xprt->stat.sending_u, | |
2418 | xprt->stat.pending_u); | |
262ca07d CL |
2419 | } |
2420 | ||
4cfc7e60 RI |
2421 | /* |
2422 | * Allocate a bunch of pages for a scratch buffer for the rpc code. The reason | |
2423 | * we allocate pages instead doing a kmalloc like rpc_malloc is because we want | |
2424 | * to use the server side send routines. | |
2425 | */ | |
5a51f13a | 2426 | static void *bc_malloc(struct rpc_task *task, size_t size) |
4cfc7e60 RI |
2427 | { |
2428 | struct page *page; | |
2429 | struct rpc_buffer *buf; | |
2430 | ||
b8a13d03 WAA |
2431 | WARN_ON_ONCE(size > PAGE_SIZE - sizeof(struct rpc_buffer)); |
2432 | if (size > PAGE_SIZE - sizeof(struct rpc_buffer)) | |
2433 | return NULL; | |
4cfc7e60 | 2434 | |
b8a13d03 | 2435 | page = alloc_page(GFP_KERNEL); |
4cfc7e60 RI |
2436 | if (!page) |
2437 | return NULL; | |
2438 | ||
2439 | buf = page_address(page); | |
2440 | buf->len = PAGE_SIZE; | |
2441 | ||
2442 | return buf->data; | |
2443 | } | |
2444 | ||
2445 | /* | |
2446 | * Free the space allocated in the bc_alloc routine | |
2447 | */ | |
5a51f13a | 2448 | static void bc_free(void *buffer) |
4cfc7e60 RI |
2449 | { |
2450 | struct rpc_buffer *buf; | |
2451 | ||
2452 | if (!buffer) | |
2453 | return; | |
2454 | ||
2455 | buf = container_of(buffer, struct rpc_buffer, data); | |
2456 | free_page((unsigned long)buf); | |
2457 | } | |
2458 | ||
2459 | /* | |
2460 | * Use the svc_sock to send the callback. Must be called with svsk->sk_mutex | |
2461 | * held. Borrows heavily from svc_tcp_sendto and xs_tcp_send_request. | |
2462 | */ | |
2463 | static int bc_sendto(struct rpc_rqst *req) | |
2464 | { | |
2465 | int len; | |
2466 | struct xdr_buf *xbufp = &req->rq_snd_buf; | |
2467 | struct rpc_xprt *xprt = req->rq_xprt; | |
2468 | struct sock_xprt *transport = | |
2469 | container_of(xprt, struct sock_xprt, xprt); | |
2470 | struct socket *sock = transport->sock; | |
2471 | unsigned long headoff; | |
2472 | unsigned long tailoff; | |
2473 | ||
61677eee | 2474 | xs_encode_stream_record_marker(xbufp); |
4cfc7e60 RI |
2475 | |
2476 | tailoff = (unsigned long)xbufp->tail[0].iov_base & ~PAGE_MASK; | |
2477 | headoff = (unsigned long)xbufp->head[0].iov_base & ~PAGE_MASK; | |
2478 | len = svc_send_common(sock, xbufp, | |
2479 | virt_to_page(xbufp->head[0].iov_base), headoff, | |
2480 | xbufp->tail[0].iov_base, tailoff); | |
2481 | ||
2482 | if (len != xbufp->len) { | |
2483 | printk(KERN_NOTICE "Error sending entire callback!\n"); | |
2484 | len = -EAGAIN; | |
2485 | } | |
2486 | ||
2487 | return len; | |
2488 | } | |
2489 | ||
2490 | /* | |
2491 | * The send routine. Borrows from svc_send | |
2492 | */ | |
2493 | static int bc_send_request(struct rpc_task *task) | |
2494 | { | |
2495 | struct rpc_rqst *req = task->tk_rqstp; | |
2496 | struct svc_xprt *xprt; | |
4cfc7e60 RI |
2497 | u32 len; |
2498 | ||
2499 | dprintk("sending request with xid: %08x\n", ntohl(req->rq_xid)); | |
2500 | /* | |
2501 | * Get the server socket associated with this callback xprt | |
2502 | */ | |
2503 | xprt = req->rq_xprt->bc_xprt; | |
4cfc7e60 RI |
2504 | |
2505 | /* | |
2506 | * Grab the mutex to serialize data as the connection is shared | |
2507 | * with the fore channel | |
2508 | */ | |
2509 | if (!mutex_trylock(&xprt->xpt_mutex)) { | |
2510 | rpc_sleep_on(&xprt->xpt_bc_pending, task, NULL); | |
2511 | if (!mutex_trylock(&xprt->xpt_mutex)) | |
2512 | return -EAGAIN; | |
2513 | rpc_wake_up_queued_task(&xprt->xpt_bc_pending, task); | |
2514 | } | |
2515 | if (test_bit(XPT_DEAD, &xprt->xpt_flags)) | |
2516 | len = -ENOTCONN; | |
2517 | else | |
2518 | len = bc_sendto(req); | |
2519 | mutex_unlock(&xprt->xpt_mutex); | |
2520 | ||
2521 | if (len > 0) | |
2522 | len = 0; | |
2523 | ||
2524 | return len; | |
2525 | } | |
2526 | ||
2527 | /* | |
2528 | * The close routine. Since this is client initiated, we do nothing | |
2529 | */ | |
2530 | ||
2531 | static void bc_close(struct rpc_xprt *xprt) | |
2532 | { | |
4cfc7e60 RI |
2533 | } |
2534 | ||
2535 | /* | |
2536 | * The xprt destroy routine. Again, because this connection is client | |
2537 | * initiated, we do nothing | |
2538 | */ | |
2539 | ||
2540 | static void bc_destroy(struct rpc_xprt *xprt) | |
2541 | { | |
47f72efa KM |
2542 | dprintk("RPC: bc_destroy xprt %p\n", xprt); |
2543 | ||
2544 | xs_xprt_free(xprt); | |
2545 | module_put(THIS_MODULE); | |
4cfc7e60 RI |
2546 | } |
2547 | ||
176e21ee CL |
2548 | static struct rpc_xprt_ops xs_local_ops = { |
2549 | .reserve_xprt = xprt_reserve_xprt, | |
2550 | .release_xprt = xs_tcp_release_xprt, | |
f39c1bfb | 2551 | .alloc_slot = xprt_alloc_slot, |
176e21ee CL |
2552 | .rpcbind = xs_local_rpcbind, |
2553 | .set_port = xs_local_set_port, | |
dc107402 | 2554 | .connect = xs_local_connect, |
176e21ee CL |
2555 | .buf_alloc = rpc_malloc, |
2556 | .buf_free = rpc_free, | |
2557 | .send_request = xs_local_send_request, | |
2558 | .set_retrans_timeout = xprt_set_retrans_timeout_def, | |
2559 | .close = xs_close, | |
a1311d87 | 2560 | .destroy = xs_destroy, |
176e21ee CL |
2561 | .print_stats = xs_local_print_stats, |
2562 | }; | |
2563 | ||
262965f5 | 2564 | static struct rpc_xprt_ops xs_udp_ops = { |
43118c29 | 2565 | .set_buffer_size = xs_udp_set_buffer_size, |
12a80469 | 2566 | .reserve_xprt = xprt_reserve_xprt_cong, |
49e9a890 | 2567 | .release_xprt = xprt_release_xprt_cong, |
f39c1bfb | 2568 | .alloc_slot = xprt_alloc_slot, |
45160d62 | 2569 | .rpcbind = rpcb_getport_async, |
92200412 | 2570 | .set_port = xs_set_port, |
262965f5 | 2571 | .connect = xs_connect, |
02107148 CL |
2572 | .buf_alloc = rpc_malloc, |
2573 | .buf_free = rpc_free, | |
262965f5 | 2574 | .send_request = xs_udp_send_request, |
fe3aca29 | 2575 | .set_retrans_timeout = xprt_set_retrans_timeout_rtt, |
46c0ee8b | 2576 | .timer = xs_udp_timer, |
a58dd398 | 2577 | .release_request = xprt_release_rqst_cong, |
262965f5 CL |
2578 | .close = xs_close, |
2579 | .destroy = xs_destroy, | |
262ca07d | 2580 | .print_stats = xs_udp_print_stats, |
262965f5 CL |
2581 | }; |
2582 | ||
2583 | static struct rpc_xprt_ops xs_tcp_ops = { | |
12a80469 | 2584 | .reserve_xprt = xprt_reserve_xprt, |
e0ab53de | 2585 | .release_xprt = xs_tcp_release_xprt, |
f39c1bfb | 2586 | .alloc_slot = xprt_lock_and_alloc_slot, |
45160d62 | 2587 | .rpcbind = rpcb_getport_async, |
92200412 | 2588 | .set_port = xs_set_port, |
0b9e7943 | 2589 | .connect = xs_connect, |
02107148 CL |
2590 | .buf_alloc = rpc_malloc, |
2591 | .buf_free = rpc_free, | |
262965f5 | 2592 | .send_request = xs_tcp_send_request, |
fe3aca29 | 2593 | .set_retrans_timeout = xprt_set_retrans_timeout_def, |
f75e6745 | 2594 | .close = xs_tcp_close, |
9903cd1c | 2595 | .destroy = xs_destroy, |
262ca07d | 2596 | .print_stats = xs_tcp_print_stats, |
a246b010 CL |
2597 | }; |
2598 | ||
4cfc7e60 RI |
2599 | /* |
2600 | * The rpc_xprt_ops for the server backchannel | |
2601 | */ | |
2602 | ||
2603 | static struct rpc_xprt_ops bc_tcp_ops = { | |
2604 | .reserve_xprt = xprt_reserve_xprt, | |
2605 | .release_xprt = xprt_release_xprt, | |
84e28a30 | 2606 | .alloc_slot = xprt_alloc_slot, |
4cfc7e60 RI |
2607 | .buf_alloc = bc_malloc, |
2608 | .buf_free = bc_free, | |
2609 | .send_request = bc_send_request, | |
2610 | .set_retrans_timeout = xprt_set_retrans_timeout_def, | |
2611 | .close = bc_close, | |
2612 | .destroy = bc_destroy, | |
2613 | .print_stats = xs_tcp_print_stats, | |
2614 | }; | |
2615 | ||
92476850 CL |
2616 | static int xs_init_anyaddr(const int family, struct sockaddr *sap) |
2617 | { | |
2618 | static const struct sockaddr_in sin = { | |
2619 | .sin_family = AF_INET, | |
2620 | .sin_addr.s_addr = htonl(INADDR_ANY), | |
2621 | }; | |
2622 | static const struct sockaddr_in6 sin6 = { | |
2623 | .sin6_family = AF_INET6, | |
2624 | .sin6_addr = IN6ADDR_ANY_INIT, | |
2625 | }; | |
2626 | ||
2627 | switch (family) { | |
176e21ee CL |
2628 | case AF_LOCAL: |
2629 | break; | |
92476850 CL |
2630 | case AF_INET: |
2631 | memcpy(sap, &sin, sizeof(sin)); | |
2632 | break; | |
2633 | case AF_INET6: | |
2634 | memcpy(sap, &sin6, sizeof(sin6)); | |
2635 | break; | |
2636 | default: | |
2637 | dprintk("RPC: %s: Bad address family\n", __func__); | |
2638 | return -EAFNOSUPPORT; | |
2639 | } | |
2640 | return 0; | |
2641 | } | |
2642 | ||
3c341b0b | 2643 | static struct rpc_xprt *xs_setup_xprt(struct xprt_create *args, |
d9ba131d TM |
2644 | unsigned int slot_table_size, |
2645 | unsigned int max_slot_table_size) | |
c8541ecd CL |
2646 | { |
2647 | struct rpc_xprt *xprt; | |
ffc2e518 | 2648 | struct sock_xprt *new; |
c8541ecd | 2649 | |
96802a09 | 2650 | if (args->addrlen > sizeof(xprt->addr)) { |
46121cf7 | 2651 | dprintk("RPC: xs_setup_xprt: address too large\n"); |
c8541ecd CL |
2652 | return ERR_PTR(-EBADF); |
2653 | } | |
2654 | ||
d9ba131d TM |
2655 | xprt = xprt_alloc(args->net, sizeof(*new), slot_table_size, |
2656 | max_slot_table_size); | |
bd1722d4 | 2657 | if (xprt == NULL) { |
46121cf7 CL |
2658 | dprintk("RPC: xs_setup_xprt: couldn't allocate " |
2659 | "rpc_xprt\n"); | |
c8541ecd CL |
2660 | return ERR_PTR(-ENOMEM); |
2661 | } | |
c8541ecd | 2662 | |
bd1722d4 | 2663 | new = container_of(xprt, struct sock_xprt, xprt); |
96802a09 FM |
2664 | memcpy(&xprt->addr, args->dstaddr, args->addrlen); |
2665 | xprt->addrlen = args->addrlen; | |
d3bc9a1d | 2666 | if (args->srcaddr) |
fbfffbd5 | 2667 | memcpy(&new->srcaddr, args->srcaddr, args->addrlen); |
92476850 CL |
2668 | else { |
2669 | int err; | |
2670 | err = xs_init_anyaddr(args->dstaddr->sa_family, | |
2671 | (struct sockaddr *)&new->srcaddr); | |
2aa13531 SK |
2672 | if (err != 0) { |
2673 | xprt_free(xprt); | |
92476850 | 2674 | return ERR_PTR(err); |
2aa13531 | 2675 | } |
92476850 | 2676 | } |
c8541ecd CL |
2677 | |
2678 | return xprt; | |
2679 | } | |
2680 | ||
176e21ee CL |
2681 | static const struct rpc_timeout xs_local_default_timeout = { |
2682 | .to_initval = 10 * HZ, | |
2683 | .to_maxval = 10 * HZ, | |
2684 | .to_retries = 2, | |
2685 | }; | |
2686 | ||
2687 | /** | |
2688 | * xs_setup_local - Set up transport to use an AF_LOCAL socket | |
2689 | * @args: rpc transport creation arguments | |
2690 | * | |
2691 | * AF_LOCAL is a "tpi_cots_ord" transport, just like TCP | |
2692 | */ | |
2693 | static struct rpc_xprt *xs_setup_local(struct xprt_create *args) | |
2694 | { | |
2695 | struct sockaddr_un *sun = (struct sockaddr_un *)args->dstaddr; | |
2696 | struct sock_xprt *transport; | |
2697 | struct rpc_xprt *xprt; | |
2698 | struct rpc_xprt *ret; | |
2699 | ||
d9ba131d TM |
2700 | xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries, |
2701 | xprt_max_tcp_slot_table_entries); | |
176e21ee CL |
2702 | if (IS_ERR(xprt)) |
2703 | return xprt; | |
2704 | transport = container_of(xprt, struct sock_xprt, xprt); | |
2705 | ||
2706 | xprt->prot = 0; | |
2707 | xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32); | |
2708 | xprt->max_payload = RPC_MAX_FRAGMENT_SIZE; | |
2709 | ||
2710 | xprt->bind_timeout = XS_BIND_TO; | |
2711 | xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; | |
2712 | xprt->idle_timeout = XS_IDLE_DISC_TO; | |
2713 | ||
2714 | xprt->ops = &xs_local_ops; | |
2715 | xprt->timeout = &xs_local_default_timeout; | |
2716 | ||
93dc41bd N |
2717 | INIT_DELAYED_WORK(&transport->connect_worker, |
2718 | xs_dummy_setup_socket); | |
2719 | ||
176e21ee CL |
2720 | switch (sun->sun_family) { |
2721 | case AF_LOCAL: | |
2722 | if (sun->sun_path[0] != '/') { | |
2723 | dprintk("RPC: bad AF_LOCAL address: %s\n", | |
2724 | sun->sun_path); | |
2725 | ret = ERR_PTR(-EINVAL); | |
2726 | goto out_err; | |
2727 | } | |
2728 | xprt_set_bound(xprt); | |
176e21ee | 2729 | xs_format_peer_addresses(xprt, "local", RPCBIND_NETID_LOCAL); |
7073ea87 BF |
2730 | ret = ERR_PTR(xs_local_setup_socket(transport)); |
2731 | if (ret) | |
2732 | goto out_err; | |
176e21ee CL |
2733 | break; |
2734 | default: | |
2735 | ret = ERR_PTR(-EAFNOSUPPORT); | |
2736 | goto out_err; | |
2737 | } | |
2738 | ||
2739 | dprintk("RPC: set up xprt to %s via AF_LOCAL\n", | |
2740 | xprt->address_strings[RPC_DISPLAY_ADDR]); | |
2741 | ||
2742 | if (try_module_get(THIS_MODULE)) | |
2743 | return xprt; | |
2744 | ret = ERR_PTR(-EINVAL); | |
2745 | out_err: | |
315f3812 | 2746 | xs_xprt_free(xprt); |
176e21ee CL |
2747 | return ret; |
2748 | } | |
2749 | ||
2881ae74 TM |
2750 | static const struct rpc_timeout xs_udp_default_timeout = { |
2751 | .to_initval = 5 * HZ, | |
2752 | .to_maxval = 30 * HZ, | |
2753 | .to_increment = 5 * HZ, | |
2754 | .to_retries = 5, | |
2755 | }; | |
2756 | ||
9903cd1c CL |
2757 | /** |
2758 | * xs_setup_udp - Set up transport to use a UDP socket | |
96802a09 | 2759 | * @args: rpc transport creation arguments |
9903cd1c CL |
2760 | * |
2761 | */ | |
483066d6 | 2762 | static struct rpc_xprt *xs_setup_udp(struct xprt_create *args) |
a246b010 | 2763 | { |
8f9d5b1a | 2764 | struct sockaddr *addr = args->dstaddr; |
c8541ecd | 2765 | struct rpc_xprt *xprt; |
c8475461 | 2766 | struct sock_xprt *transport; |
0a68b0be | 2767 | struct rpc_xprt *ret; |
a246b010 | 2768 | |
d9ba131d TM |
2769 | xprt = xs_setup_xprt(args, xprt_udp_slot_table_entries, |
2770 | xprt_udp_slot_table_entries); | |
c8541ecd CL |
2771 | if (IS_ERR(xprt)) |
2772 | return xprt; | |
c8475461 | 2773 | transport = container_of(xprt, struct sock_xprt, xprt); |
a246b010 | 2774 | |
ec739ef0 | 2775 | xprt->prot = IPPROTO_UDP; |
808012fb | 2776 | xprt->tsh_size = 0; |
a246b010 CL |
2777 | /* XXX: header size can vary due to auth type, IPv6, etc. */ |
2778 | xprt->max_payload = (1U << 16) - (MAX_HEADER << 3); | |
2779 | ||
03bf4b70 | 2780 | xprt->bind_timeout = XS_BIND_TO; |
03bf4b70 CL |
2781 | xprt->reestablish_timeout = XS_UDP_REEST_TO; |
2782 | xprt->idle_timeout = XS_IDLE_DISC_TO; | |
a246b010 | 2783 | |
262965f5 | 2784 | xprt->ops = &xs_udp_ops; |
a246b010 | 2785 | |
ba7392bb | 2786 | xprt->timeout = &xs_udp_default_timeout; |
a246b010 | 2787 | |
8f9d5b1a CL |
2788 | switch (addr->sa_family) { |
2789 | case AF_INET: | |
2790 | if (((struct sockaddr_in *)addr)->sin_port != htons(0)) | |
2791 | xprt_set_bound(xprt); | |
2792 | ||
2793 | INIT_DELAYED_WORK(&transport->connect_worker, | |
8c14ff2a | 2794 | xs_udp_setup_socket); |
9dc3b095 | 2795 | xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP); |
8f9d5b1a CL |
2796 | break; |
2797 | case AF_INET6: | |
2798 | if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0)) | |
2799 | xprt_set_bound(xprt); | |
2800 | ||
2801 | INIT_DELAYED_WORK(&transport->connect_worker, | |
8c14ff2a | 2802 | xs_udp_setup_socket); |
9dc3b095 | 2803 | xs_format_peer_addresses(xprt, "udp", RPCBIND_NETID_UDP6); |
8f9d5b1a CL |
2804 | break; |
2805 | default: | |
0a68b0be BF |
2806 | ret = ERR_PTR(-EAFNOSUPPORT); |
2807 | goto out_err; | |
8f9d5b1a CL |
2808 | } |
2809 | ||
c740eff8 CL |
2810 | if (xprt_bound(xprt)) |
2811 | dprintk("RPC: set up xprt to %s (port %s) via %s\n", | |
2812 | xprt->address_strings[RPC_DISPLAY_ADDR], | |
2813 | xprt->address_strings[RPC_DISPLAY_PORT], | |
2814 | xprt->address_strings[RPC_DISPLAY_PROTO]); | |
2815 | else | |
2816 | dprintk("RPC: set up xprt to %s (autobind) via %s\n", | |
2817 | xprt->address_strings[RPC_DISPLAY_ADDR], | |
2818 | xprt->address_strings[RPC_DISPLAY_PROTO]); | |
edb267a6 | 2819 | |
bc25571e TT |
2820 | if (try_module_get(THIS_MODULE)) |
2821 | return xprt; | |
0a68b0be BF |
2822 | ret = ERR_PTR(-EINVAL); |
2823 | out_err: | |
315f3812 | 2824 | xs_xprt_free(xprt); |
0a68b0be | 2825 | return ret; |
a246b010 CL |
2826 | } |
2827 | ||
2881ae74 TM |
2828 | static const struct rpc_timeout xs_tcp_default_timeout = { |
2829 | .to_initval = 60 * HZ, | |
2830 | .to_maxval = 60 * HZ, | |
2831 | .to_retries = 2, | |
2832 | }; | |
2833 | ||
9903cd1c CL |
2834 | /** |
2835 | * xs_setup_tcp - Set up transport to use a TCP socket | |
96802a09 | 2836 | * @args: rpc transport creation arguments |
9903cd1c CL |
2837 | * |
2838 | */ | |
483066d6 | 2839 | static struct rpc_xprt *xs_setup_tcp(struct xprt_create *args) |
a246b010 | 2840 | { |
8f9d5b1a | 2841 | struct sockaddr *addr = args->dstaddr; |
c8541ecd | 2842 | struct rpc_xprt *xprt; |
c8475461 | 2843 | struct sock_xprt *transport; |
0a68b0be | 2844 | struct rpc_xprt *ret; |
b7993ceb TM |
2845 | unsigned int max_slot_table_size = xprt_max_tcp_slot_table_entries; |
2846 | ||
2847 | if (args->flags & XPRT_CREATE_INFINITE_SLOTS) | |
2848 | max_slot_table_size = RPC_MAX_SLOT_TABLE_LIMIT; | |
a246b010 | 2849 | |
d9ba131d | 2850 | xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries, |
b7993ceb | 2851 | max_slot_table_size); |
c8541ecd CL |
2852 | if (IS_ERR(xprt)) |
2853 | return xprt; | |
c8475461 | 2854 | transport = container_of(xprt, struct sock_xprt, xprt); |
a246b010 | 2855 | |
ec739ef0 | 2856 | xprt->prot = IPPROTO_TCP; |
808012fb | 2857 | xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32); |
808012fb | 2858 | xprt->max_payload = RPC_MAX_FRAGMENT_SIZE; |
a246b010 | 2859 | |
03bf4b70 | 2860 | xprt->bind_timeout = XS_BIND_TO; |
03bf4b70 CL |
2861 | xprt->reestablish_timeout = XS_TCP_INIT_REEST_TO; |
2862 | xprt->idle_timeout = XS_IDLE_DISC_TO; | |
a246b010 | 2863 | |
262965f5 | 2864 | xprt->ops = &xs_tcp_ops; |
ba7392bb | 2865 | xprt->timeout = &xs_tcp_default_timeout; |
a246b010 | 2866 | |
8f9d5b1a CL |
2867 | switch (addr->sa_family) { |
2868 | case AF_INET: | |
2869 | if (((struct sockaddr_in *)addr)->sin_port != htons(0)) | |
2870 | xprt_set_bound(xprt); | |
2871 | ||
9dc3b095 | 2872 | INIT_DELAYED_WORK(&transport->connect_worker, |
cdd518d5 | 2873 | xs_tcp_setup_socket); |
9dc3b095 | 2874 | xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP); |
8f9d5b1a CL |
2875 | break; |
2876 | case AF_INET6: | |
2877 | if (((struct sockaddr_in6 *)addr)->sin6_port != htons(0)) | |
2878 | xprt_set_bound(xprt); | |
2879 | ||
9dc3b095 | 2880 | INIT_DELAYED_WORK(&transport->connect_worker, |
cdd518d5 | 2881 | xs_tcp_setup_socket); |
9dc3b095 | 2882 | xs_format_peer_addresses(xprt, "tcp", RPCBIND_NETID_TCP6); |
8f9d5b1a CL |
2883 | break; |
2884 | default: | |
0a68b0be BF |
2885 | ret = ERR_PTR(-EAFNOSUPPORT); |
2886 | goto out_err; | |
8f9d5b1a CL |
2887 | } |
2888 | ||
c740eff8 CL |
2889 | if (xprt_bound(xprt)) |
2890 | dprintk("RPC: set up xprt to %s (port %s) via %s\n", | |
2891 | xprt->address_strings[RPC_DISPLAY_ADDR], | |
2892 | xprt->address_strings[RPC_DISPLAY_PORT], | |
2893 | xprt->address_strings[RPC_DISPLAY_PROTO]); | |
2894 | else | |
2895 | dprintk("RPC: set up xprt to %s (autobind) via %s\n", | |
2896 | xprt->address_strings[RPC_DISPLAY_ADDR], | |
2897 | xprt->address_strings[RPC_DISPLAY_PROTO]); | |
2898 | ||
bc25571e TT |
2899 | if (try_module_get(THIS_MODULE)) |
2900 | return xprt; | |
0a68b0be BF |
2901 | ret = ERR_PTR(-EINVAL); |
2902 | out_err: | |
315f3812 | 2903 | xs_xprt_free(xprt); |
0a68b0be | 2904 | return ret; |
a246b010 | 2905 | } |
282b32e1 | 2906 | |
f300baba AB |
2907 | /** |
2908 | * xs_setup_bc_tcp - Set up transport to use a TCP backchannel socket | |
2909 | * @args: rpc transport creation arguments | |
2910 | * | |
2911 | */ | |
2912 | static struct rpc_xprt *xs_setup_bc_tcp(struct xprt_create *args) | |
2913 | { | |
2914 | struct sockaddr *addr = args->dstaddr; | |
2915 | struct rpc_xprt *xprt; | |
2916 | struct sock_xprt *transport; | |
2917 | struct svc_sock *bc_sock; | |
0a68b0be | 2918 | struct rpc_xprt *ret; |
f300baba | 2919 | |
d9ba131d TM |
2920 | xprt = xs_setup_xprt(args, xprt_tcp_slot_table_entries, |
2921 | xprt_tcp_slot_table_entries); | |
f300baba AB |
2922 | if (IS_ERR(xprt)) |
2923 | return xprt; | |
2924 | transport = container_of(xprt, struct sock_xprt, xprt); | |
2925 | ||
2926 | xprt->prot = IPPROTO_TCP; | |
2927 | xprt->tsh_size = sizeof(rpc_fraghdr) / sizeof(u32); | |
2928 | xprt->max_payload = RPC_MAX_FRAGMENT_SIZE; | |
2929 | xprt->timeout = &xs_tcp_default_timeout; | |
2930 | ||
2931 | /* backchannel */ | |
2932 | xprt_set_bound(xprt); | |
2933 | xprt->bind_timeout = 0; | |
f300baba AB |
2934 | xprt->reestablish_timeout = 0; |
2935 | xprt->idle_timeout = 0; | |
2936 | ||
f300baba AB |
2937 | xprt->ops = &bc_tcp_ops; |
2938 | ||
2939 | switch (addr->sa_family) { | |
2940 | case AF_INET: | |
2941 | xs_format_peer_addresses(xprt, "tcp", | |
2942 | RPCBIND_NETID_TCP); | |
2943 | break; | |
2944 | case AF_INET6: | |
2945 | xs_format_peer_addresses(xprt, "tcp", | |
2946 | RPCBIND_NETID_TCP6); | |
2947 | break; | |
2948 | default: | |
0a68b0be BF |
2949 | ret = ERR_PTR(-EAFNOSUPPORT); |
2950 | goto out_err; | |
f300baba AB |
2951 | } |
2952 | ||
50fa0d40 PE |
2953 | dprintk("RPC: set up xprt to %s (port %s) via %s\n", |
2954 | xprt->address_strings[RPC_DISPLAY_ADDR], | |
2955 | xprt->address_strings[RPC_DISPLAY_PORT], | |
2956 | xprt->address_strings[RPC_DISPLAY_PROTO]); | |
f300baba | 2957 | |
99de8ea9 BF |
2958 | /* |
2959 | * Once we've associated a backchannel xprt with a connection, | |
28303ca3 WM |
2960 | * we want to keep it around as long as the connection lasts, |
2961 | * in case we need to start using it for a backchannel again; | |
2962 | * this reference won't be dropped until bc_xprt is destroyed. | |
99de8ea9 BF |
2963 | */ |
2964 | xprt_get(xprt); | |
2965 | args->bc_xprt->xpt_bc_xprt = xprt; | |
2966 | xprt->bc_xprt = args->bc_xprt; | |
2967 | bc_sock = container_of(args->bc_xprt, struct svc_sock, sk_xprt); | |
2968 | transport->sock = bc_sock->sk_sock; | |
2969 | transport->inet = bc_sock->sk_sk; | |
2970 | ||
f300baba AB |
2971 | /* |
2972 | * Since we don't want connections for the backchannel, we set | |
2973 | * the xprt status to connected | |
2974 | */ | |
2975 | xprt_set_connected(xprt); | |
2976 | ||
f300baba AB |
2977 | if (try_module_get(THIS_MODULE)) |
2978 | return xprt; | |
642aab58 KM |
2979 | |
2980 | args->bc_xprt->xpt_bc_xprt = NULL; | |
99de8ea9 | 2981 | xprt_put(xprt); |
0a68b0be BF |
2982 | ret = ERR_PTR(-EINVAL); |
2983 | out_err: | |
315f3812 | 2984 | xs_xprt_free(xprt); |
0a68b0be | 2985 | return ret; |
f300baba AB |
2986 | } |
2987 | ||
176e21ee CL |
2988 | static struct xprt_class xs_local_transport = { |
2989 | .list = LIST_HEAD_INIT(xs_local_transport.list), | |
2990 | .name = "named UNIX socket", | |
2991 | .owner = THIS_MODULE, | |
2992 | .ident = XPRT_TRANSPORT_LOCAL, | |
2993 | .setup = xs_setup_local, | |
2994 | }; | |
2995 | ||
bc25571e TT |
2996 | static struct xprt_class xs_udp_transport = { |
2997 | .list = LIST_HEAD_INIT(xs_udp_transport.list), | |
2998 | .name = "udp", | |
2999 | .owner = THIS_MODULE, | |
f300baba | 3000 | .ident = XPRT_TRANSPORT_UDP, |
bc25571e TT |
3001 | .setup = xs_setup_udp, |
3002 | }; | |
3003 | ||
3004 | static struct xprt_class xs_tcp_transport = { | |
3005 | .list = LIST_HEAD_INIT(xs_tcp_transport.list), | |
3006 | .name = "tcp", | |
3007 | .owner = THIS_MODULE, | |
f300baba | 3008 | .ident = XPRT_TRANSPORT_TCP, |
bc25571e TT |
3009 | .setup = xs_setup_tcp, |
3010 | }; | |
3011 | ||
f300baba AB |
3012 | static struct xprt_class xs_bc_tcp_transport = { |
3013 | .list = LIST_HEAD_INIT(xs_bc_tcp_transport.list), | |
3014 | .name = "tcp NFSv4.1 backchannel", | |
3015 | .owner = THIS_MODULE, | |
3016 | .ident = XPRT_TRANSPORT_BC_TCP, | |
3017 | .setup = xs_setup_bc_tcp, | |
3018 | }; | |
3019 | ||
282b32e1 | 3020 | /** |
bc25571e | 3021 | * init_socket_xprt - set up xprtsock's sysctls, register with RPC client |
282b32e1 CL |
3022 | * |
3023 | */ | |
3024 | int init_socket_xprt(void) | |
3025 | { | |
fbf76683 | 3026 | #ifdef RPC_DEBUG |
2b1bec5f | 3027 | if (!sunrpc_table_header) |
0b4d4147 | 3028 | sunrpc_table_header = register_sysctl_table(sunrpc_table); |
fbf76683 CL |
3029 | #endif |
3030 | ||
176e21ee | 3031 | xprt_register_transport(&xs_local_transport); |
bc25571e TT |
3032 | xprt_register_transport(&xs_udp_transport); |
3033 | xprt_register_transport(&xs_tcp_transport); | |
f300baba | 3034 | xprt_register_transport(&xs_bc_tcp_transport); |
bc25571e | 3035 | |
282b32e1 CL |
3036 | return 0; |
3037 | } | |
3038 | ||
3039 | /** | |
bc25571e | 3040 | * cleanup_socket_xprt - remove xprtsock's sysctls, unregister |
282b32e1 CL |
3041 | * |
3042 | */ | |
3043 | void cleanup_socket_xprt(void) | |
3044 | { | |
fbf76683 CL |
3045 | #ifdef RPC_DEBUG |
3046 | if (sunrpc_table_header) { | |
3047 | unregister_sysctl_table(sunrpc_table_header); | |
3048 | sunrpc_table_header = NULL; | |
3049 | } | |
3050 | #endif | |
bc25571e | 3051 | |
176e21ee | 3052 | xprt_unregister_transport(&xs_local_transport); |
bc25571e TT |
3053 | xprt_unregister_transport(&xs_udp_transport); |
3054 | xprt_unregister_transport(&xs_tcp_transport); | |
f300baba | 3055 | xprt_unregister_transport(&xs_bc_tcp_transport); |
282b32e1 | 3056 | } |
cbf11071 | 3057 | |
9bbb9e5a RR |
3058 | static int param_set_uint_minmax(const char *val, |
3059 | const struct kernel_param *kp, | |
cbf11071 TM |
3060 | unsigned int min, unsigned int max) |
3061 | { | |
00cfaa94 | 3062 | unsigned int num; |
cbf11071 TM |
3063 | int ret; |
3064 | ||
3065 | if (!val) | |
3066 | return -EINVAL; | |
00cfaa94 | 3067 | ret = kstrtouint(val, 0, &num); |
cbf11071 TM |
3068 | if (ret == -EINVAL || num < min || num > max) |
3069 | return -EINVAL; | |
3070 | *((unsigned int *)kp->arg) = num; | |
3071 | return 0; | |
3072 | } | |
3073 | ||
9bbb9e5a | 3074 | static int param_set_portnr(const char *val, const struct kernel_param *kp) |
cbf11071 TM |
3075 | { |
3076 | return param_set_uint_minmax(val, kp, | |
3077 | RPC_MIN_RESVPORT, | |
3078 | RPC_MAX_RESVPORT); | |
3079 | } | |
3080 | ||
9bbb9e5a RR |
3081 | static struct kernel_param_ops param_ops_portnr = { |
3082 | .set = param_set_portnr, | |
3083 | .get = param_get_uint, | |
3084 | }; | |
3085 | ||
cbf11071 TM |
3086 | #define param_check_portnr(name, p) \ |
3087 | __param_check(name, p, unsigned int); | |
3088 | ||
3089 | module_param_named(min_resvport, xprt_min_resvport, portnr, 0644); | |
3090 | module_param_named(max_resvport, xprt_max_resvport, portnr, 0644); | |
3091 | ||
9bbb9e5a RR |
3092 | static int param_set_slot_table_size(const char *val, |
3093 | const struct kernel_param *kp) | |
cbf11071 TM |
3094 | { |
3095 | return param_set_uint_minmax(val, kp, | |
3096 | RPC_MIN_SLOT_TABLE, | |
3097 | RPC_MAX_SLOT_TABLE); | |
3098 | } | |
3099 | ||
9bbb9e5a RR |
3100 | static struct kernel_param_ops param_ops_slot_table_size = { |
3101 | .set = param_set_slot_table_size, | |
3102 | .get = param_get_uint, | |
3103 | }; | |
3104 | ||
cbf11071 TM |
3105 | #define param_check_slot_table_size(name, p) \ |
3106 | __param_check(name, p, unsigned int); | |
3107 | ||
d9ba131d TM |
3108 | static int param_set_max_slot_table_size(const char *val, |
3109 | const struct kernel_param *kp) | |
3110 | { | |
3111 | return param_set_uint_minmax(val, kp, | |
3112 | RPC_MIN_SLOT_TABLE, | |
3113 | RPC_MAX_SLOT_TABLE_LIMIT); | |
3114 | } | |
3115 | ||
3116 | static struct kernel_param_ops param_ops_max_slot_table_size = { | |
3117 | .set = param_set_max_slot_table_size, | |
3118 | .get = param_get_uint, | |
3119 | }; | |
3120 | ||
3121 | #define param_check_max_slot_table_size(name, p) \ | |
3122 | __param_check(name, p, unsigned int); | |
3123 | ||
cbf11071 TM |
3124 | module_param_named(tcp_slot_table_entries, xprt_tcp_slot_table_entries, |
3125 | slot_table_size, 0644); | |
d9ba131d TM |
3126 | module_param_named(tcp_max_slot_table_entries, xprt_max_tcp_slot_table_entries, |
3127 | max_slot_table_size, 0644); | |
cbf11071 TM |
3128 | module_param_named(udp_slot_table_entries, xprt_udp_slot_table_entries, |
3129 | slot_table_size, 0644); | |
3130 |