Commit | Line | Data |
---|---|---|
a246b010 CL |
1 | /* |
2 | * linux/net/sunrpc/xprtsock.c | |
3 | * | |
4 | * Client-side transport implementation for sockets. | |
5 | * | |
6 | * TCP callback races fixes (C) 1998 Red Hat Software <alan@redhat.com> | |
7 | * TCP send fixes (C) 1998 Red Hat Software <alan@redhat.com> | |
8 | * TCP NFS related read + write fixes | |
9 | * (C) 1999 Dave Airlie, University of Limerick, Ireland <airlied@linux.ie> | |
10 | * | |
11 | * Rewrite of larges part of the code in order to stabilize TCP stuff. | |
12 | * Fix behaviour when socket buffer is full. | |
13 | * (C) 1999 Trond Myklebust <trond.myklebust@fys.uio.no> | |
55aa4f58 CL |
14 | * |
15 | * IP socket transport implementation, (C) 2005 Chuck Lever <cel@netapp.com> | |
a246b010 CL |
16 | */ |
17 | ||
18 | #include <linux/types.h> | |
19 | #include <linux/slab.h> | |
20 | #include <linux/capability.h> | |
21 | #include <linux/sched.h> | |
22 | #include <linux/pagemap.h> | |
23 | #include <linux/errno.h> | |
24 | #include <linux/socket.h> | |
25 | #include <linux/in.h> | |
26 | #include <linux/net.h> | |
27 | #include <linux/mm.h> | |
28 | #include <linux/udp.h> | |
29 | #include <linux/tcp.h> | |
30 | #include <linux/sunrpc/clnt.h> | |
31 | #include <linux/file.h> | |
32 | ||
33 | #include <net/sock.h> | |
34 | #include <net/checksum.h> | |
35 | #include <net/udp.h> | |
36 | #include <net/tcp.h> | |
37 | ||
9903cd1c CL |
38 | /* |
39 | * Maximum port number to use when requesting a reserved port. | |
40 | */ | |
41 | #define XS_MAX_RESVPORT (800U) | |
42 | ||
a246b010 CL |
43 | #ifdef RPC_DEBUG |
44 | # undef RPC_DEBUG_DATA | |
9903cd1c | 45 | # define RPCDBG_FACILITY RPCDBG_TRANS |
a246b010 CL |
46 | #endif |
47 | ||
a246b010 | 48 | #ifdef RPC_DEBUG_DATA |
9903cd1c | 49 | static void xs_pktdump(char *msg, u32 *packet, unsigned int count) |
a246b010 | 50 | { |
9903cd1c CL |
51 | u8 *buf = (u8 *) packet; |
52 | int j; | |
a246b010 CL |
53 | |
54 | dprintk("RPC: %s\n", msg); | |
55 | for (j = 0; j < count && j < 128; j += 4) { | |
56 | if (!(j & 31)) { | |
57 | if (j) | |
58 | dprintk("\n"); | |
59 | dprintk("0x%04x ", j); | |
60 | } | |
61 | dprintk("%02x%02x%02x%02x ", | |
62 | buf[j], buf[j+1], buf[j+2], buf[j+3]); | |
63 | } | |
64 | dprintk("\n"); | |
65 | } | |
66 | #else | |
9903cd1c | 67 | static inline void xs_pktdump(char *msg, u32 *packet, unsigned int count) |
a246b010 CL |
68 | { |
69 | /* NOP */ | |
70 | } | |
71 | #endif | |
72 | ||
b4b5cc85 CL |
73 | #define XS_SENDMSG_FLAGS (MSG_DONTWAIT | MSG_NOSIGNAL) |
74 | ||
75 | static inline int xs_send_head(struct socket *sock, struct sockaddr *addr, int addrlen, struct xdr_buf *xdr, unsigned int base, unsigned int len) | |
76 | { | |
77 | struct kvec iov = { | |
78 | .iov_base = xdr->head[0].iov_base + base, | |
79 | .iov_len = len - base, | |
80 | }; | |
81 | struct msghdr msg = { | |
82 | .msg_name = addr, | |
83 | .msg_namelen = addrlen, | |
84 | .msg_flags = XS_SENDMSG_FLAGS, | |
85 | }; | |
86 | ||
87 | if (xdr->len > len) | |
88 | msg.msg_flags |= MSG_MORE; | |
89 | ||
90 | if (likely(iov.iov_len)) | |
91 | return kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len); | |
92 | return kernel_sendmsg(sock, &msg, NULL, 0, 0); | |
93 | } | |
94 | ||
95 | static int xs_send_tail(struct socket *sock, struct xdr_buf *xdr, unsigned int base, unsigned int len) | |
96 | { | |
97 | struct kvec iov = { | |
98 | .iov_base = xdr->tail[0].iov_base + base, | |
99 | .iov_len = len - base, | |
100 | }; | |
101 | struct msghdr msg = { | |
102 | .msg_flags = XS_SENDMSG_FLAGS, | |
103 | }; | |
104 | ||
105 | return kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len); | |
106 | } | |
107 | ||
9903cd1c CL |
108 | /** |
109 | * xs_sendpages - write pages directly to a socket | |
110 | * @sock: socket to send on | |
111 | * @addr: UDP only -- address of destination | |
112 | * @addrlen: UDP only -- length of destination address | |
113 | * @xdr: buffer containing this request | |
114 | * @base: starting position in the buffer | |
115 | * | |
a246b010 | 116 | */ |
b4b5cc85 | 117 | static int xs_sendpages(struct socket *sock, struct sockaddr *addr, int addrlen, struct xdr_buf *xdr, unsigned int base) |
a246b010 CL |
118 | { |
119 | struct page **ppage = xdr->pages; | |
120 | unsigned int len, pglen = xdr->page_len; | |
121 | int err, ret = 0; | |
122 | ssize_t (*sendpage)(struct socket *, struct page *, int, size_t, int); | |
123 | ||
124 | len = xdr->head[0].iov_len; | |
125 | if (base < len || (addr != NULL && base == 0)) { | |
b4b5cc85 | 126 | err = xs_send_head(sock, addr, addrlen, xdr, base, len); |
a246b010 CL |
127 | if (ret == 0) |
128 | ret = err; | |
129 | else if (err > 0) | |
130 | ret += err; | |
b4b5cc85 | 131 | if (err != (len - base)) |
a246b010 CL |
132 | goto out; |
133 | base = 0; | |
134 | } else | |
135 | base -= len; | |
136 | ||
b4b5cc85 | 137 | if (unlikely(pglen == 0)) |
a246b010 | 138 | goto copy_tail; |
b4b5cc85 | 139 | if (unlikely(base >= pglen)) { |
a246b010 CL |
140 | base -= pglen; |
141 | goto copy_tail; | |
142 | } | |
143 | if (base || xdr->page_base) { | |
144 | pglen -= base; | |
9903cd1c | 145 | base += xdr->page_base; |
a246b010 CL |
146 | ppage += base >> PAGE_CACHE_SHIFT; |
147 | base &= ~PAGE_CACHE_MASK; | |
148 | } | |
149 | ||
150 | sendpage = sock->ops->sendpage ? : sock_no_sendpage; | |
151 | do { | |
b4b5cc85 | 152 | int flags = XS_SENDMSG_FLAGS; |
a246b010 CL |
153 | |
154 | len = PAGE_CACHE_SIZE; | |
155 | if (base) | |
156 | len -= base; | |
157 | if (pglen < len) | |
158 | len = pglen; | |
159 | ||
160 | if (pglen != len || xdr->tail[0].iov_len != 0) | |
161 | flags |= MSG_MORE; | |
162 | ||
163 | /* Hmm... We might be dealing with highmem pages */ | |
164 | if (PageHighMem(*ppage)) | |
165 | sendpage = sock_no_sendpage; | |
166 | err = sendpage(sock, *ppage, base, len, flags); | |
167 | if (ret == 0) | |
168 | ret = err; | |
169 | else if (err > 0) | |
170 | ret += err; | |
171 | if (err != len) | |
172 | goto out; | |
173 | base = 0; | |
174 | ppage++; | |
175 | } while ((pglen -= len) != 0); | |
176 | copy_tail: | |
177 | len = xdr->tail[0].iov_len; | |
178 | if (base < len) { | |
b4b5cc85 | 179 | err = xs_send_tail(sock, xdr, base, len); |
a246b010 CL |
180 | if (ret == 0) |
181 | ret = err; | |
182 | else if (err > 0) | |
183 | ret += err; | |
184 | } | |
185 | out: | |
186 | return ret; | |
187 | } | |
188 | ||
9903cd1c CL |
189 | /** |
190 | * xs_sendmsg - write an RPC request to a socket | |
191 | * @xprt: generic transport | |
192 | * @req: the RPC request to write | |
193 | * | |
a246b010 | 194 | */ |
9903cd1c | 195 | static int xs_sendmsg(struct rpc_xprt *xprt, struct rpc_rqst *req) |
a246b010 | 196 | { |
9903cd1c CL |
197 | struct socket *sock = xprt->sock; |
198 | struct xdr_buf *xdr = &req->rq_snd_buf; | |
a246b010 CL |
199 | struct sockaddr *addr = NULL; |
200 | int addrlen = 0; | |
9903cd1c CL |
201 | unsigned int skip; |
202 | int result; | |
a246b010 CL |
203 | |
204 | if (!sock) | |
205 | return -ENOTCONN; | |
206 | ||
9903cd1c | 207 | xs_pktdump("packet data:", |
a246b010 CL |
208 | req->rq_svec->iov_base, |
209 | req->rq_svec->iov_len); | |
210 | ||
211 | /* For UDP, we need to provide an address */ | |
212 | if (!xprt->stream) { | |
213 | addr = (struct sockaddr *) &xprt->addr; | |
214 | addrlen = sizeof(xprt->addr); | |
215 | } | |
9903cd1c | 216 | /* Don't repeat bytes */ |
a246b010 CL |
217 | skip = req->rq_bytes_sent; |
218 | ||
219 | clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags); | |
b4b5cc85 | 220 | result = xs_sendpages(sock, addr, addrlen, xdr, skip); |
a246b010 | 221 | |
9903cd1c | 222 | dprintk("RPC: xs_sendmsg(%d) = %d\n", xdr->len - skip, result); |
a246b010 CL |
223 | |
224 | if (result >= 0) | |
225 | return result; | |
226 | ||
227 | switch (result) { | |
228 | case -ECONNREFUSED: | |
229 | /* When the server has died, an ICMP port unreachable message | |
9903cd1c | 230 | * prompts ECONNREFUSED. */ |
a246b010 CL |
231 | case -EAGAIN: |
232 | break; | |
233 | case -ECONNRESET: | |
234 | case -ENOTCONN: | |
235 | case -EPIPE: | |
236 | /* connection broken */ | |
237 | if (xprt->stream) | |
238 | result = -ENOTCONN; | |
239 | break; | |
240 | default: | |
9903cd1c | 241 | break; |
a246b010 CL |
242 | } |
243 | return result; | |
244 | } | |
245 | ||
9903cd1c CL |
246 | /** |
247 | * xs_send_request - write an RPC request to a socket | |
248 | * @task: address of RPC task that manages the state of an RPC request | |
249 | * | |
250 | * Return values: | |
251 | * 0: The request has been sent | |
252 | * EAGAIN: The socket was blocked, please call again later to | |
253 | * complete the request | |
254 | * other: Some other error occured, the request was not sent | |
255 | * | |
256 | * XXX: In the case of soft timeouts, should we eventually give up | |
257 | * if the socket is not able to make progress? | |
258 | */ | |
259 | static int xs_send_request(struct rpc_task *task) | |
a246b010 CL |
260 | { |
261 | struct rpc_rqst *req = task->tk_rqstp; | |
262 | struct rpc_xprt *xprt = req->rq_xprt; | |
263 | int status, retry = 0; | |
264 | ||
265 | /* set up everything as needed. */ | |
266 | /* Write the record marker */ | |
267 | if (xprt->stream) { | |
9903cd1c | 268 | u32 *marker = req->rq_svec[0].iov_base; |
a246b010 CL |
269 | |
270 | *marker = htonl(0x80000000|(req->rq_slen-sizeof(*marker))); | |
271 | } | |
272 | ||
273 | /* Continue transmitting the packet/record. We must be careful | |
274 | * to cope with writespace callbacks arriving _after_ we have | |
9903cd1c | 275 | * called sendmsg(). |
a246b010 CL |
276 | */ |
277 | while (1) { | |
278 | req->rq_xtime = jiffies; | |
9903cd1c | 279 | status = xs_sendmsg(xprt, req); |
a246b010 CL |
280 | |
281 | if (status < 0) | |
282 | break; | |
283 | ||
284 | if (xprt->stream) { | |
285 | req->rq_bytes_sent += status; | |
286 | ||
287 | /* If we've sent the entire packet, immediately | |
288 | * reset the count of bytes sent. */ | |
289 | if (req->rq_bytes_sent >= req->rq_slen) { | |
290 | req->rq_bytes_sent = 0; | |
291 | return 0; | |
292 | } | |
293 | } else { | |
294 | if (status >= req->rq_slen) | |
295 | return 0; | |
296 | status = -EAGAIN; | |
297 | break; | |
298 | } | |
299 | ||
300 | dprintk("RPC: %4d xmit incomplete (%d left of %d)\n", | |
301 | task->tk_pid, req->rq_slen - req->rq_bytes_sent, | |
302 | req->rq_slen); | |
303 | ||
304 | status = -EAGAIN; | |
305 | if (retry++ > 50) | |
306 | break; | |
307 | } | |
308 | ||
309 | if (status == -EAGAIN) { | |
310 | if (test_bit(SOCK_ASYNC_NOSPACE, &xprt->sock->flags)) { | |
c7b2cae8 | 311 | /* Protect against races with write_space */ |
4a0f8c04 | 312 | spin_lock_bh(&xprt->transport_lock); |
a246b010 CL |
313 | /* Don't race with disconnect */ |
314 | if (!xprt_connected(xprt)) | |
315 | task->tk_status = -ENOTCONN; | |
c7b2cae8 CL |
316 | else if (test_bit(SOCK_NOSPACE, &xprt->sock->flags)) |
317 | xprt_wait_for_buffer_space(task); | |
4a0f8c04 | 318 | spin_unlock_bh(&xprt->transport_lock); |
a246b010 CL |
319 | return status; |
320 | } | |
321 | /* Keep holding the socket if it is blocked */ | |
322 | rpc_delay(task, HZ>>4); | |
323 | } | |
324 | return status; | |
325 | } | |
326 | ||
9903cd1c CL |
327 | /** |
328 | * xs_close - close a socket | |
329 | * @xprt: transport | |
330 | * | |
a246b010 | 331 | */ |
9903cd1c | 332 | static void xs_close(struct rpc_xprt *xprt) |
a246b010 | 333 | { |
9903cd1c CL |
334 | struct socket *sock = xprt->sock; |
335 | struct sock *sk = xprt->inet; | |
a246b010 CL |
336 | |
337 | if (!sk) | |
338 | return; | |
339 | ||
9903cd1c CL |
340 | dprintk("RPC: xs_close xprt %p\n", xprt); |
341 | ||
a246b010 CL |
342 | write_lock_bh(&sk->sk_callback_lock); |
343 | xprt->inet = NULL; | |
344 | xprt->sock = NULL; | |
345 | ||
9903cd1c CL |
346 | sk->sk_user_data = NULL; |
347 | sk->sk_data_ready = xprt->old_data_ready; | |
a246b010 | 348 | sk->sk_state_change = xprt->old_state_change; |
9903cd1c | 349 | sk->sk_write_space = xprt->old_write_space; |
a246b010 CL |
350 | write_unlock_bh(&sk->sk_callback_lock); |
351 | ||
9903cd1c | 352 | sk->sk_no_check = 0; |
a246b010 CL |
353 | |
354 | sock_release(sock); | |
355 | } | |
356 | ||
9903cd1c CL |
357 | /** |
358 | * xs_destroy - prepare to shutdown a transport | |
359 | * @xprt: doomed transport | |
360 | * | |
361 | */ | |
362 | static void xs_destroy(struct rpc_xprt *xprt) | |
a246b010 | 363 | { |
9903cd1c CL |
364 | dprintk("RPC: xs_destroy xprt %p\n", xprt); |
365 | ||
55aa4f58 | 366 | cancel_delayed_work(&xprt->connect_worker); |
a246b010 CL |
367 | flush_scheduled_work(); |
368 | ||
369 | xprt_disconnect(xprt); | |
9903cd1c | 370 | xs_close(xprt); |
a246b010 CL |
371 | kfree(xprt->slot); |
372 | } | |
373 | ||
9903cd1c CL |
374 | static inline struct rpc_xprt *xprt_from_sock(struct sock *sk) |
375 | { | |
376 | return (struct rpc_xprt *) sk->sk_user_data; | |
377 | } | |
378 | ||
379 | /** | |
380 | * xs_udp_data_ready - "data ready" callback for UDP sockets | |
381 | * @sk: socket with data to read | |
382 | * @len: how much data to read | |
383 | * | |
a246b010 | 384 | */ |
9903cd1c | 385 | static void xs_udp_data_ready(struct sock *sk, int len) |
a246b010 | 386 | { |
9903cd1c CL |
387 | struct rpc_task *task; |
388 | struct rpc_xprt *xprt; | |
a246b010 | 389 | struct rpc_rqst *rovr; |
9903cd1c | 390 | struct sk_buff *skb; |
a246b010 CL |
391 | int err, repsize, copied; |
392 | u32 _xid, *xp; | |
393 | ||
394 | read_lock(&sk->sk_callback_lock); | |
9903cd1c CL |
395 | dprintk("RPC: xs_udp_data_ready...\n"); |
396 | if (!(xprt = xprt_from_sock(sk))) | |
a246b010 | 397 | goto out; |
a246b010 CL |
398 | |
399 | if ((skb = skb_recv_datagram(sk, 0, 1, &err)) == NULL) | |
400 | goto out; | |
401 | ||
402 | if (xprt->shutdown) | |
403 | goto dropit; | |
404 | ||
405 | repsize = skb->len - sizeof(struct udphdr); | |
406 | if (repsize < 4) { | |
9903cd1c | 407 | dprintk("RPC: impossible RPC reply size %d!\n", repsize); |
a246b010 CL |
408 | goto dropit; |
409 | } | |
410 | ||
411 | /* Copy the XID from the skb... */ | |
412 | xp = skb_header_pointer(skb, sizeof(struct udphdr), | |
413 | sizeof(_xid), &_xid); | |
414 | if (xp == NULL) | |
415 | goto dropit; | |
416 | ||
417 | /* Look up and lock the request corresponding to the given XID */ | |
4a0f8c04 | 418 | spin_lock(&xprt->transport_lock); |
a246b010 CL |
419 | rovr = xprt_lookup_rqst(xprt, *xp); |
420 | if (!rovr) | |
421 | goto out_unlock; | |
422 | task = rovr->rq_task; | |
423 | ||
424 | dprintk("RPC: %4d received reply\n", task->tk_pid); | |
425 | ||
426 | if ((copied = rovr->rq_private_buf.buflen) > repsize) | |
427 | copied = repsize; | |
428 | ||
429 | /* Suck it into the iovec, verify checksum if not done by hw. */ | |
430 | if (csum_partial_copy_to_xdr(&rovr->rq_private_buf, skb)) | |
431 | goto out_unlock; | |
432 | ||
433 | /* Something worked... */ | |
434 | dst_confirm(skb->dst); | |
435 | ||
436 | xprt_complete_rqst(xprt, rovr, copied); | |
437 | ||
438 | out_unlock: | |
4a0f8c04 | 439 | spin_unlock(&xprt->transport_lock); |
a246b010 CL |
440 | dropit: |
441 | skb_free_datagram(sk, skb); | |
442 | out: | |
443 | read_unlock(&sk->sk_callback_lock); | |
444 | } | |
445 | ||
9903cd1c | 446 | static inline size_t xs_tcp_copy_data(skb_reader_t *desc, void *p, size_t len) |
a246b010 CL |
447 | { |
448 | if (len > desc->count) | |
449 | len = desc->count; | |
450 | if (skb_copy_bits(desc->skb, desc->offset, p, len)) { | |
451 | dprintk("RPC: failed to copy %zu bytes from skb. %zu bytes remain\n", | |
452 | len, desc->count); | |
453 | return 0; | |
454 | } | |
455 | desc->offset += len; | |
456 | desc->count -= len; | |
457 | dprintk("RPC: copied %zu bytes from skb. %zu bytes remain\n", | |
458 | len, desc->count); | |
459 | return len; | |
460 | } | |
461 | ||
9903cd1c | 462 | static inline void xs_tcp_read_fraghdr(struct rpc_xprt *xprt, skb_reader_t *desc) |
a246b010 CL |
463 | { |
464 | size_t len, used; | |
465 | char *p; | |
466 | ||
467 | p = ((char *) &xprt->tcp_recm) + xprt->tcp_offset; | |
468 | len = sizeof(xprt->tcp_recm) - xprt->tcp_offset; | |
9903cd1c | 469 | used = xs_tcp_copy_data(desc, p, len); |
a246b010 CL |
470 | xprt->tcp_offset += used; |
471 | if (used != len) | |
472 | return; | |
473 | xprt->tcp_reclen = ntohl(xprt->tcp_recm); | |
474 | if (xprt->tcp_reclen & 0x80000000) | |
475 | xprt->tcp_flags |= XPRT_LAST_FRAG; | |
476 | else | |
477 | xprt->tcp_flags &= ~XPRT_LAST_FRAG; | |
478 | xprt->tcp_reclen &= 0x7fffffff; | |
479 | xprt->tcp_flags &= ~XPRT_COPY_RECM; | |
480 | xprt->tcp_offset = 0; | |
481 | /* Sanity check of the record length */ | |
482 | if (xprt->tcp_reclen < 4) { | |
9903cd1c | 483 | dprintk("RPC: invalid TCP record fragment length\n"); |
a246b010 | 484 | xprt_disconnect(xprt); |
9903cd1c | 485 | return; |
a246b010 CL |
486 | } |
487 | dprintk("RPC: reading TCP record fragment of length %d\n", | |
488 | xprt->tcp_reclen); | |
489 | } | |
490 | ||
9903cd1c | 491 | static void xs_tcp_check_recm(struct rpc_xprt *xprt) |
a246b010 CL |
492 | { |
493 | dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u, tcp_flags = %lx\n", | |
494 | xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen, xprt->tcp_flags); | |
495 | if (xprt->tcp_offset == xprt->tcp_reclen) { | |
496 | xprt->tcp_flags |= XPRT_COPY_RECM; | |
497 | xprt->tcp_offset = 0; | |
498 | if (xprt->tcp_flags & XPRT_LAST_FRAG) { | |
499 | xprt->tcp_flags &= ~XPRT_COPY_DATA; | |
500 | xprt->tcp_flags |= XPRT_COPY_XID; | |
501 | xprt->tcp_copied = 0; | |
502 | } | |
503 | } | |
504 | } | |
505 | ||
9903cd1c | 506 | static inline void xs_tcp_read_xid(struct rpc_xprt *xprt, skb_reader_t *desc) |
a246b010 CL |
507 | { |
508 | size_t len, used; | |
509 | char *p; | |
510 | ||
511 | len = sizeof(xprt->tcp_xid) - xprt->tcp_offset; | |
512 | dprintk("RPC: reading XID (%Zu bytes)\n", len); | |
513 | p = ((char *) &xprt->tcp_xid) + xprt->tcp_offset; | |
9903cd1c | 514 | used = xs_tcp_copy_data(desc, p, len); |
a246b010 CL |
515 | xprt->tcp_offset += used; |
516 | if (used != len) | |
517 | return; | |
518 | xprt->tcp_flags &= ~XPRT_COPY_XID; | |
519 | xprt->tcp_flags |= XPRT_COPY_DATA; | |
520 | xprt->tcp_copied = 4; | |
521 | dprintk("RPC: reading reply for XID %08x\n", | |
522 | ntohl(xprt->tcp_xid)); | |
9903cd1c | 523 | xs_tcp_check_recm(xprt); |
a246b010 CL |
524 | } |
525 | ||
9903cd1c | 526 | static inline void xs_tcp_read_request(struct rpc_xprt *xprt, skb_reader_t *desc) |
a246b010 CL |
527 | { |
528 | struct rpc_rqst *req; | |
529 | struct xdr_buf *rcvbuf; | |
530 | size_t len; | |
531 | ssize_t r; | |
532 | ||
533 | /* Find and lock the request corresponding to this xid */ | |
4a0f8c04 | 534 | spin_lock(&xprt->transport_lock); |
a246b010 CL |
535 | req = xprt_lookup_rqst(xprt, xprt->tcp_xid); |
536 | if (!req) { | |
537 | xprt->tcp_flags &= ~XPRT_COPY_DATA; | |
538 | dprintk("RPC: XID %08x request not found!\n", | |
539 | ntohl(xprt->tcp_xid)); | |
4a0f8c04 | 540 | spin_unlock(&xprt->transport_lock); |
a246b010 CL |
541 | return; |
542 | } | |
543 | ||
544 | rcvbuf = &req->rq_private_buf; | |
545 | len = desc->count; | |
546 | if (len > xprt->tcp_reclen - xprt->tcp_offset) { | |
547 | skb_reader_t my_desc; | |
548 | ||
549 | len = xprt->tcp_reclen - xprt->tcp_offset; | |
550 | memcpy(&my_desc, desc, sizeof(my_desc)); | |
551 | my_desc.count = len; | |
552 | r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied, | |
9903cd1c | 553 | &my_desc, xs_tcp_copy_data); |
a246b010 CL |
554 | desc->count -= r; |
555 | desc->offset += r; | |
556 | } else | |
557 | r = xdr_partial_copy_from_skb(rcvbuf, xprt->tcp_copied, | |
9903cd1c | 558 | desc, xs_tcp_copy_data); |
a246b010 CL |
559 | |
560 | if (r > 0) { | |
561 | xprt->tcp_copied += r; | |
562 | xprt->tcp_offset += r; | |
563 | } | |
564 | if (r != len) { | |
565 | /* Error when copying to the receive buffer, | |
566 | * usually because we weren't able to allocate | |
567 | * additional buffer pages. All we can do now | |
568 | * is turn off XPRT_COPY_DATA, so the request | |
569 | * will not receive any additional updates, | |
570 | * and time out. | |
571 | * Any remaining data from this record will | |
572 | * be discarded. | |
573 | */ | |
574 | xprt->tcp_flags &= ~XPRT_COPY_DATA; | |
575 | dprintk("RPC: XID %08x truncated request\n", | |
576 | ntohl(xprt->tcp_xid)); | |
577 | dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n", | |
578 | xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen); | |
579 | goto out; | |
580 | } | |
581 | ||
582 | dprintk("RPC: XID %08x read %Zd bytes\n", | |
583 | ntohl(xprt->tcp_xid), r); | |
584 | dprintk("RPC: xprt = %p, tcp_copied = %lu, tcp_offset = %u, tcp_reclen = %u\n", | |
585 | xprt, xprt->tcp_copied, xprt->tcp_offset, xprt->tcp_reclen); | |
586 | ||
587 | if (xprt->tcp_copied == req->rq_private_buf.buflen) | |
588 | xprt->tcp_flags &= ~XPRT_COPY_DATA; | |
589 | else if (xprt->tcp_offset == xprt->tcp_reclen) { | |
590 | if (xprt->tcp_flags & XPRT_LAST_FRAG) | |
591 | xprt->tcp_flags &= ~XPRT_COPY_DATA; | |
592 | } | |
593 | ||
594 | out: | |
595 | if (!(xprt->tcp_flags & XPRT_COPY_DATA)) { | |
596 | dprintk("RPC: %4d received reply complete\n", | |
597 | req->rq_task->tk_pid); | |
598 | xprt_complete_rqst(xprt, req, xprt->tcp_copied); | |
599 | } | |
4a0f8c04 | 600 | spin_unlock(&xprt->transport_lock); |
9903cd1c | 601 | xs_tcp_check_recm(xprt); |
a246b010 CL |
602 | } |
603 | ||
9903cd1c | 604 | static inline void xs_tcp_read_discard(struct rpc_xprt *xprt, skb_reader_t *desc) |
a246b010 CL |
605 | { |
606 | size_t len; | |
607 | ||
608 | len = xprt->tcp_reclen - xprt->tcp_offset; | |
609 | if (len > desc->count) | |
610 | len = desc->count; | |
611 | desc->count -= len; | |
612 | desc->offset += len; | |
613 | xprt->tcp_offset += len; | |
614 | dprintk("RPC: discarded %Zu bytes\n", len); | |
9903cd1c | 615 | xs_tcp_check_recm(xprt); |
a246b010 CL |
616 | } |
617 | ||
9903cd1c | 618 | static int xs_tcp_data_recv(read_descriptor_t *rd_desc, struct sk_buff *skb, unsigned int offset, size_t len) |
a246b010 CL |
619 | { |
620 | struct rpc_xprt *xprt = rd_desc->arg.data; | |
621 | skb_reader_t desc = { | |
622 | .skb = skb, | |
623 | .offset = offset, | |
624 | .count = len, | |
625 | .csum = 0 | |
9903cd1c | 626 | }; |
a246b010 | 627 | |
9903cd1c | 628 | dprintk("RPC: xs_tcp_data_recv started\n"); |
a246b010 CL |
629 | do { |
630 | /* Read in a new fragment marker if necessary */ | |
631 | /* Can we ever really expect to get completely empty fragments? */ | |
632 | if (xprt->tcp_flags & XPRT_COPY_RECM) { | |
9903cd1c | 633 | xs_tcp_read_fraghdr(xprt, &desc); |
a246b010 CL |
634 | continue; |
635 | } | |
636 | /* Read in the xid if necessary */ | |
637 | if (xprt->tcp_flags & XPRT_COPY_XID) { | |
9903cd1c | 638 | xs_tcp_read_xid(xprt, &desc); |
a246b010 CL |
639 | continue; |
640 | } | |
641 | /* Read in the request data */ | |
642 | if (xprt->tcp_flags & XPRT_COPY_DATA) { | |
9903cd1c | 643 | xs_tcp_read_request(xprt, &desc); |
a246b010 CL |
644 | continue; |
645 | } | |
646 | /* Skip over any trailing bytes on short reads */ | |
9903cd1c | 647 | xs_tcp_read_discard(xprt, &desc); |
a246b010 | 648 | } while (desc.count); |
9903cd1c | 649 | dprintk("RPC: xs_tcp_data_recv done\n"); |
a246b010 CL |
650 | return len - desc.count; |
651 | } | |
652 | ||
9903cd1c CL |
653 | /** |
654 | * xs_tcp_data_ready - "data ready" callback for TCP sockets | |
655 | * @sk: socket with data to read | |
656 | * @bytes: how much data to read | |
657 | * | |
658 | */ | |
659 | static void xs_tcp_data_ready(struct sock *sk, int bytes) | |
a246b010 CL |
660 | { |
661 | struct rpc_xprt *xprt; | |
662 | read_descriptor_t rd_desc; | |
663 | ||
664 | read_lock(&sk->sk_callback_lock); | |
9903cd1c CL |
665 | dprintk("RPC: xs_tcp_data_ready...\n"); |
666 | if (!(xprt = xprt_from_sock(sk))) | |
a246b010 | 667 | goto out; |
a246b010 CL |
668 | if (xprt->shutdown) |
669 | goto out; | |
670 | ||
9903cd1c | 671 | /* We use rd_desc to pass struct xprt to xs_tcp_data_recv */ |
a246b010 CL |
672 | rd_desc.arg.data = xprt; |
673 | rd_desc.count = 65536; | |
9903cd1c | 674 | tcp_read_sock(sk, &rd_desc, xs_tcp_data_recv); |
a246b010 CL |
675 | out: |
676 | read_unlock(&sk->sk_callback_lock); | |
677 | } | |
678 | ||
9903cd1c CL |
679 | /** |
680 | * xs_tcp_state_change - callback to handle TCP socket state changes | |
681 | * @sk: socket whose state has changed | |
682 | * | |
683 | */ | |
684 | static void xs_tcp_state_change(struct sock *sk) | |
a246b010 | 685 | { |
9903cd1c | 686 | struct rpc_xprt *xprt; |
a246b010 CL |
687 | |
688 | read_lock(&sk->sk_callback_lock); | |
689 | if (!(xprt = xprt_from_sock(sk))) | |
690 | goto out; | |
9903cd1c | 691 | dprintk("RPC: xs_tcp_state_change client %p...\n", xprt); |
a246b010 CL |
692 | dprintk("RPC: state %x conn %d dead %d zapped %d\n", |
693 | sk->sk_state, xprt_connected(xprt), | |
694 | sock_flag(sk, SOCK_DEAD), | |
695 | sock_flag(sk, SOCK_ZAPPED)); | |
696 | ||
697 | switch (sk->sk_state) { | |
698 | case TCP_ESTABLISHED: | |
4a0f8c04 | 699 | spin_lock_bh(&xprt->transport_lock); |
a246b010 CL |
700 | if (!xprt_test_and_set_connected(xprt)) { |
701 | /* Reset TCP record info */ | |
702 | xprt->tcp_offset = 0; | |
703 | xprt->tcp_reclen = 0; | |
704 | xprt->tcp_copied = 0; | |
705 | xprt->tcp_flags = XPRT_COPY_RECM | XPRT_COPY_XID; | |
44fbac22 | 706 | xprt_wake_pending_tasks(xprt, 0); |
a246b010 | 707 | } |
4a0f8c04 | 708 | spin_unlock_bh(&xprt->transport_lock); |
a246b010 CL |
709 | break; |
710 | case TCP_SYN_SENT: | |
711 | case TCP_SYN_RECV: | |
712 | break; | |
713 | default: | |
714 | xprt_disconnect(xprt); | |
715 | break; | |
716 | } | |
717 | out: | |
718 | read_unlock(&sk->sk_callback_lock); | |
719 | } | |
720 | ||
9903cd1c | 721 | /** |
c7b2cae8 CL |
722 | * xs_udp_write_space - callback invoked when socket buffer space |
723 | * becomes available | |
9903cd1c CL |
724 | * @sk: socket whose state has changed |
725 | * | |
a246b010 CL |
726 | * Called when more output buffer space is available for this socket. |
727 | * We try not to wake our writers until they can make "significant" | |
c7b2cae8 | 728 | * progress, otherwise we'll waste resources thrashing kernel_sendmsg |
a246b010 CL |
729 | * with a bunch of small requests. |
730 | */ | |
c7b2cae8 | 731 | static void xs_udp_write_space(struct sock *sk) |
a246b010 | 732 | { |
a246b010 | 733 | read_lock(&sk->sk_callback_lock); |
a246b010 | 734 | |
c7b2cae8 CL |
735 | /* from net/core/sock.c:sock_def_write_space */ |
736 | if (sock_writeable(sk)) { | |
737 | struct socket *sock; | |
738 | struct rpc_xprt *xprt; | |
739 | ||
740 | if (unlikely(!(sock = sk->sk_socket))) | |
a246b010 | 741 | goto out; |
c7b2cae8 CL |
742 | if (unlikely(!(xprt = xprt_from_sock(sk)))) |
743 | goto out; | |
744 | if (unlikely(!test_and_clear_bit(SOCK_NOSPACE, &sock->flags))) | |
a246b010 | 745 | goto out; |
c7b2cae8 CL |
746 | |
747 | xprt_write_space(xprt); | |
a246b010 CL |
748 | } |
749 | ||
c7b2cae8 CL |
750 | out: |
751 | read_unlock(&sk->sk_callback_lock); | |
752 | } | |
a246b010 | 753 | |
c7b2cae8 CL |
754 | /** |
755 | * xs_tcp_write_space - callback invoked when socket buffer space | |
756 | * becomes available | |
757 | * @sk: socket whose state has changed | |
758 | * | |
759 | * Called when more output buffer space is available for this socket. | |
760 | * We try not to wake our writers until they can make "significant" | |
761 | * progress, otherwise we'll waste resources thrashing kernel_sendmsg | |
762 | * with a bunch of small requests. | |
763 | */ | |
764 | static void xs_tcp_write_space(struct sock *sk) | |
765 | { | |
766 | read_lock(&sk->sk_callback_lock); | |
767 | ||
768 | /* from net/core/stream.c:sk_stream_write_space */ | |
769 | if (sk_stream_wspace(sk) >= sk_stream_min_wspace(sk)) { | |
770 | struct socket *sock; | |
771 | struct rpc_xprt *xprt; | |
772 | ||
773 | if (unlikely(!(sock = sk->sk_socket))) | |
774 | goto out; | |
775 | if (unlikely(!(xprt = xprt_from_sock(sk)))) | |
776 | goto out; | |
777 | if (unlikely(!test_and_clear_bit(SOCK_NOSPACE, &sock->flags))) | |
778 | goto out; | |
779 | ||
780 | xprt_write_space(xprt); | |
781 | } | |
782 | ||
783 | out: | |
a246b010 CL |
784 | read_unlock(&sk->sk_callback_lock); |
785 | } | |
786 | ||
9903cd1c CL |
787 | /** |
788 | * xs_set_buffer_size - set send and receive limits | |
789 | * @xprt: generic transport | |
790 | * | |
791 | * Set socket send and receive limits based on the | |
792 | * sndsize and rcvsize fields in the generic transport | |
793 | * structure. This applies only to UDP sockets. | |
a246b010 | 794 | */ |
9903cd1c | 795 | static void xs_set_buffer_size(struct rpc_xprt *xprt) |
a246b010 CL |
796 | { |
797 | struct sock *sk = xprt->inet; | |
798 | ||
799 | if (xprt->stream) | |
800 | return; | |
801 | if (xprt->rcvsize) { | |
802 | sk->sk_userlocks |= SOCK_RCVBUF_LOCK; | |
803 | sk->sk_rcvbuf = xprt->rcvsize * xprt->max_reqs * 2; | |
804 | } | |
805 | if (xprt->sndsize) { | |
806 | sk->sk_userlocks |= SOCK_SNDBUF_LOCK; | |
807 | sk->sk_sndbuf = xprt->sndsize * xprt->max_reqs * 2; | |
808 | sk->sk_write_space(sk); | |
809 | } | |
810 | } | |
811 | ||
9903cd1c | 812 | static int xs_bindresvport(struct rpc_xprt *xprt, struct socket *sock) |
a246b010 CL |
813 | { |
814 | struct sockaddr_in myaddr = { | |
815 | .sin_family = AF_INET, | |
816 | }; | |
9903cd1c | 817 | int err, port; |
a246b010 CL |
818 | |
819 | /* Were we already bound to a given port? Try to reuse it */ | |
820 | port = xprt->port; | |
821 | do { | |
822 | myaddr.sin_port = htons(port); | |
823 | err = sock->ops->bind(sock, (struct sockaddr *) &myaddr, | |
824 | sizeof(myaddr)); | |
825 | if (err == 0) { | |
826 | xprt->port = port; | |
9903cd1c CL |
827 | dprintk("RPC: xs_bindresvport bound to port %u\n", |
828 | port); | |
a246b010 CL |
829 | return 0; |
830 | } | |
831 | if (--port == 0) | |
9903cd1c | 832 | port = XS_MAX_RESVPORT; |
a246b010 CL |
833 | } while (err == -EADDRINUSE && port != xprt->port); |
834 | ||
9903cd1c | 835 | dprintk("RPC: can't bind to reserved port (%d).\n", -err); |
a246b010 CL |
836 | return err; |
837 | } | |
838 | ||
b0d93ad5 CL |
839 | /** |
840 | * xs_udp_connect_worker - set up a UDP socket | |
841 | * @args: RPC transport to connect | |
842 | * | |
843 | * Invoked by a work queue tasklet. | |
844 | */ | |
845 | static void xs_udp_connect_worker(void *args) | |
a246b010 | 846 | { |
b0d93ad5 CL |
847 | struct rpc_xprt *xprt = (struct rpc_xprt *) args; |
848 | struct socket *sock = xprt->sock; | |
849 | int err, status = -EIO; | |
9903cd1c | 850 | |
b0d93ad5 CL |
851 | if (xprt->shutdown || xprt->addr.sin_port == 0) |
852 | goto out; | |
9903cd1c | 853 | |
b0d93ad5 | 854 | dprintk("RPC: xs_udp_connect_worker for xprt %p\n", xprt); |
9903cd1c | 855 | |
b0d93ad5 CL |
856 | /* Start by resetting any existing state */ |
857 | xs_close(xprt); | |
9903cd1c | 858 | |
b0d93ad5 CL |
859 | if ((err = sock_create_kern(PF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock)) < 0) { |
860 | dprintk("RPC: can't create UDP transport socket (%d).\n", -err); | |
861 | goto out; | |
862 | } | |
9903cd1c | 863 | |
b0d93ad5 CL |
864 | if (xprt->resvport && xs_bindresvport(xprt, sock) < 0) { |
865 | sock_release(sock); | |
866 | goto out; | |
867 | } | |
9903cd1c | 868 | |
b0d93ad5 CL |
869 | if (!xprt->inet) { |
870 | struct sock *sk = sock->sk; | |
a246b010 | 871 | |
b0d93ad5 | 872 | write_lock_bh(&sk->sk_callback_lock); |
a246b010 | 873 | |
b0d93ad5 CL |
874 | sk->sk_user_data = xprt; |
875 | xprt->old_data_ready = sk->sk_data_ready; | |
876 | xprt->old_state_change = sk->sk_state_change; | |
877 | xprt->old_write_space = sk->sk_write_space; | |
9903cd1c | 878 | sk->sk_data_ready = xs_udp_data_ready; |
c7b2cae8 | 879 | sk->sk_write_space = xs_udp_write_space; |
a246b010 | 880 | sk->sk_no_check = UDP_CSUM_NORCV; |
b0d93ad5 | 881 | |
a246b010 | 882 | xprt_set_connected(xprt); |
a246b010 | 883 | |
b0d93ad5 CL |
884 | /* Reset to new socket */ |
885 | xprt->sock = sock; | |
886 | xprt->inet = sk; | |
a246b010 | 887 | |
b0d93ad5 CL |
888 | write_unlock_bh(&sk->sk_callback_lock); |
889 | } | |
890 | xs_set_buffer_size(xprt); | |
891 | status = 0; | |
892 | out: | |
893 | xprt_wake_pending_tasks(xprt, status); | |
894 | xprt_clear_connecting(xprt); | |
a246b010 CL |
895 | } |
896 | ||
9903cd1c | 897 | /** |
b0d93ad5 | 898 | * xs_tcp_connect_worker - connect a TCP socket to a remote endpoint |
9903cd1c CL |
899 | * @args: RPC transport to connect |
900 | * | |
901 | * Invoked by a work queue tasklet. | |
a246b010 | 902 | */ |
b0d93ad5 | 903 | static void xs_tcp_connect_worker(void *args) |
a246b010 CL |
904 | { |
905 | struct rpc_xprt *xprt = (struct rpc_xprt *)args; | |
906 | struct socket *sock = xprt->sock; | |
b0d93ad5 | 907 | int err, status = -EIO; |
a246b010 CL |
908 | |
909 | if (xprt->shutdown || xprt->addr.sin_port == 0) | |
910 | goto out; | |
911 | ||
b0d93ad5 | 912 | dprintk("RPC: xs_tcp_connect_worker for xprt %p\n", xprt); |
9903cd1c | 913 | |
b0d93ad5 | 914 | /* Start by resetting any existing socket state */ |
9903cd1c | 915 | xs_close(xprt); |
b0d93ad5 CL |
916 | |
917 | if ((err = sock_create_kern(PF_INET, SOCK_STREAM, IPPROTO_TCP, &sock)) < 0) { | |
918 | dprintk("RPC: can't create TCP transport socket (%d).\n", -err); | |
a246b010 CL |
919 | goto out; |
920 | } | |
a246b010 | 921 | |
b0d93ad5 CL |
922 | if (xprt->resvport && xs_bindresvport(xprt, sock) < 0) { |
923 | sock_release(sock); | |
a246b010 | 924 | goto out; |
b0d93ad5 | 925 | } |
a246b010 | 926 | |
b0d93ad5 CL |
927 | if (!xprt->inet) { |
928 | struct sock *sk = sock->sk; | |
929 | ||
930 | write_lock_bh(&sk->sk_callback_lock); | |
931 | ||
932 | sk->sk_user_data = xprt; | |
933 | xprt->old_data_ready = sk->sk_data_ready; | |
934 | xprt->old_state_change = sk->sk_state_change; | |
935 | xprt->old_write_space = sk->sk_write_space; | |
936 | sk->sk_data_ready = xs_tcp_data_ready; | |
937 | sk->sk_state_change = xs_tcp_state_change; | |
938 | sk->sk_write_space = xs_tcp_write_space; | |
939 | tcp_sk(sk)->nonagle = 1; | |
940 | ||
941 | xprt_clear_connected(xprt); | |
942 | ||
943 | /* Reset to new socket */ | |
944 | xprt->sock = sock; | |
945 | xprt->inet = sk; | |
946 | ||
947 | write_unlock_bh(&sk->sk_callback_lock); | |
948 | } | |
949 | ||
950 | /* Tell the socket layer to start connecting... */ | |
a246b010 CL |
951 | status = sock->ops->connect(sock, (struct sockaddr *) &xprt->addr, |
952 | sizeof(xprt->addr), O_NONBLOCK); | |
953 | dprintk("RPC: %p connect status %d connected %d sock state %d\n", | |
954 | xprt, -status, xprt_connected(xprt), sock->sk->sk_state); | |
955 | if (status < 0) { | |
956 | switch (status) { | |
957 | case -EINPROGRESS: | |
958 | case -EALREADY: | |
959 | goto out_clear; | |
960 | } | |
961 | } | |
962 | out: | |
44fbac22 | 963 | xprt_wake_pending_tasks(xprt, status); |
a246b010 | 964 | out_clear: |
2226feb6 | 965 | xprt_clear_connecting(xprt); |
a246b010 CL |
966 | } |
967 | ||
9903cd1c CL |
968 | /** |
969 | * xs_connect - connect a socket to a remote endpoint | |
970 | * @task: address of RPC task that manages state of connect request | |
971 | * | |
972 | * TCP: If the remote end dropped the connection, delay reconnecting. | |
973 | */ | |
974 | static void xs_connect(struct rpc_task *task) | |
a246b010 CL |
975 | { |
976 | struct rpc_xprt *xprt = task->tk_xprt; | |
977 | ||
b0d93ad5 CL |
978 | if (xprt_test_and_set_connecting(xprt)) |
979 | return; | |
980 | ||
981 | if (xprt->sock != NULL) { | |
982 | dprintk("RPC: xs_connect delayed xprt %p\n", xprt); | |
983 | schedule_delayed_work(&xprt->connect_worker, | |
a246b010 | 984 | RPC_REESTABLISH_TIMEOUT); |
b0d93ad5 CL |
985 | } else { |
986 | dprintk("RPC: xs_connect scheduled xprt %p\n", xprt); | |
987 | schedule_work(&xprt->connect_worker); | |
988 | ||
989 | /* flush_scheduled_work can sleep... */ | |
990 | if (!RPC_IS_ASYNC(task)) | |
991 | flush_scheduled_work(); | |
a246b010 CL |
992 | } |
993 | } | |
994 | ||
9903cd1c CL |
995 | static struct rpc_xprt_ops xs_ops = { |
996 | .set_buffer_size = xs_set_buffer_size, | |
997 | .connect = xs_connect, | |
998 | .send_request = xs_send_request, | |
999 | .close = xs_close, | |
1000 | .destroy = xs_destroy, | |
a246b010 CL |
1001 | }; |
1002 | ||
1003 | extern unsigned int xprt_udp_slot_table_entries; | |
1004 | extern unsigned int xprt_tcp_slot_table_entries; | |
1005 | ||
9903cd1c CL |
1006 | /** |
1007 | * xs_setup_udp - Set up transport to use a UDP socket | |
1008 | * @xprt: transport to set up | |
1009 | * @to: timeout parameters | |
1010 | * | |
1011 | */ | |
a246b010 CL |
1012 | int xs_setup_udp(struct rpc_xprt *xprt, struct rpc_timeout *to) |
1013 | { | |
1014 | size_t slot_table_size; | |
1015 | ||
1016 | dprintk("RPC: setting up udp-ipv4 transport...\n"); | |
1017 | ||
1018 | xprt->max_reqs = xprt_udp_slot_table_entries; | |
1019 | slot_table_size = xprt->max_reqs * sizeof(xprt->slot[0]); | |
1020 | xprt->slot = kmalloc(slot_table_size, GFP_KERNEL); | |
1021 | if (xprt->slot == NULL) | |
1022 | return -ENOMEM; | |
1023 | memset(xprt->slot, 0, slot_table_size); | |
1024 | ||
1025 | xprt->prot = IPPROTO_UDP; | |
9903cd1c | 1026 | xprt->port = XS_MAX_RESVPORT; |
a246b010 CL |
1027 | xprt->stream = 0; |
1028 | xprt->nocong = 0; | |
1029 | xprt->cwnd = RPC_INITCWND; | |
1030 | xprt->resvport = capable(CAP_NET_BIND_SERVICE) ? 1 : 0; | |
1031 | /* XXX: header size can vary due to auth type, IPv6, etc. */ | |
1032 | xprt->max_payload = (1U << 16) - (MAX_HEADER << 3); | |
1033 | ||
b0d93ad5 | 1034 | INIT_WORK(&xprt->connect_worker, xs_udp_connect_worker, xprt); |
a246b010 | 1035 | |
9903cd1c | 1036 | xprt->ops = &xs_ops; |
a246b010 CL |
1037 | |
1038 | if (to) | |
1039 | xprt->timeout = *to; | |
1040 | else | |
9903cd1c | 1041 | xprt_set_timeout(&xprt->timeout, 5, 5 * HZ); |
a246b010 CL |
1042 | |
1043 | return 0; | |
1044 | } | |
1045 | ||
9903cd1c CL |
1046 | /** |
1047 | * xs_setup_tcp - Set up transport to use a TCP socket | |
1048 | * @xprt: transport to set up | |
1049 | * @to: timeout parameters | |
1050 | * | |
1051 | */ | |
a246b010 CL |
1052 | int xs_setup_tcp(struct rpc_xprt *xprt, struct rpc_timeout *to) |
1053 | { | |
1054 | size_t slot_table_size; | |
1055 | ||
1056 | dprintk("RPC: setting up tcp-ipv4 transport...\n"); | |
1057 | ||
1058 | xprt->max_reqs = xprt_tcp_slot_table_entries; | |
1059 | slot_table_size = xprt->max_reqs * sizeof(xprt->slot[0]); | |
1060 | xprt->slot = kmalloc(slot_table_size, GFP_KERNEL); | |
1061 | if (xprt->slot == NULL) | |
1062 | return -ENOMEM; | |
1063 | memset(xprt->slot, 0, slot_table_size); | |
1064 | ||
1065 | xprt->prot = IPPROTO_TCP; | |
9903cd1c | 1066 | xprt->port = XS_MAX_RESVPORT; |
a246b010 CL |
1067 | xprt->stream = 1; |
1068 | xprt->nocong = 1; | |
1069 | xprt->cwnd = RPC_MAXCWND(xprt); | |
1070 | xprt->resvport = capable(CAP_NET_BIND_SERVICE) ? 1 : 0; | |
1071 | xprt->max_payload = (1U << 31) - 1; | |
1072 | ||
b0d93ad5 | 1073 | INIT_WORK(&xprt->connect_worker, xs_tcp_connect_worker, xprt); |
a246b010 | 1074 | |
9903cd1c | 1075 | xprt->ops = &xs_ops; |
a246b010 CL |
1076 | |
1077 | if (to) | |
1078 | xprt->timeout = *to; | |
1079 | else | |
9903cd1c | 1080 | xprt_set_timeout(&xprt->timeout, 2, 60 * HZ); |
a246b010 CL |
1081 | |
1082 | return 0; | |
1083 | } |