fix section mismatch warnings
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / rxrpc / ar-recvmsg.c
1 /* RxRPC recvmsg() implementation
2 *
3 * Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 */
11
12 #include <linux/net.h>
13 #include <linux/skbuff.h>
14 #include <linux/export.h>
15 #include <net/sock.h>
16 #include <net/af_rxrpc.h>
17 #include "ar-internal.h"
18
19 /*
20 * removal a call's user ID from the socket tree to make the user ID available
21 * again and so that it won't be seen again in association with that call
22 */
23 void rxrpc_remove_user_ID(struct rxrpc_sock *rx, struct rxrpc_call *call)
24 {
25 _debug("RELEASE CALL %d", call->debug_id);
26
27 if (test_bit(RXRPC_CALL_HAS_USERID, &call->flags)) {
28 write_lock_bh(&rx->call_lock);
29 rb_erase(&call->sock_node, &call->socket->calls);
30 clear_bit(RXRPC_CALL_HAS_USERID, &call->flags);
31 write_unlock_bh(&rx->call_lock);
32 }
33
34 read_lock_bh(&call->state_lock);
35 if (!test_bit(RXRPC_CALL_RELEASED, &call->flags) &&
36 !test_and_set_bit(RXRPC_CALL_RELEASE, &call->events))
37 rxrpc_queue_call(call);
38 read_unlock_bh(&call->state_lock);
39 }
40
41 /*
42 * receive a message from an RxRPC socket
43 * - we need to be careful about two or more threads calling recvmsg
44 * simultaneously
45 */
46 int rxrpc_recvmsg(struct kiocb *iocb, struct socket *sock,
47 struct msghdr *msg, size_t len, int flags)
48 {
49 struct rxrpc_skb_priv *sp;
50 struct rxrpc_call *call = NULL, *continue_call = NULL;
51 struct rxrpc_sock *rx = rxrpc_sk(sock->sk);
52 struct sk_buff *skb;
53 long timeo;
54 int copy, ret, ullen, offset, copied = 0;
55 u32 abort_code;
56
57 DEFINE_WAIT(wait);
58
59 _enter(",,,%zu,%d", len, flags);
60
61 if (flags & (MSG_OOB | MSG_TRUNC))
62 return -EOPNOTSUPP;
63
64 ullen = msg->msg_flags & MSG_CMSG_COMPAT ? 4 : sizeof(unsigned long);
65
66 timeo = sock_rcvtimeo(&rx->sk, flags & MSG_DONTWAIT);
67 msg->msg_flags |= MSG_MORE;
68
69 lock_sock(&rx->sk);
70
71 for (;;) {
72 /* return immediately if a client socket has no outstanding
73 * calls */
74 if (RB_EMPTY_ROOT(&rx->calls)) {
75 if (copied)
76 goto out;
77 if (rx->sk.sk_state != RXRPC_SERVER_LISTENING) {
78 release_sock(&rx->sk);
79 if (continue_call)
80 rxrpc_put_call(continue_call);
81 return -ENODATA;
82 }
83 }
84
85 /* get the next message on the Rx queue */
86 skb = skb_peek(&rx->sk.sk_receive_queue);
87 if (!skb) {
88 /* nothing remains on the queue */
89 if (copied &&
90 (flags & MSG_PEEK || timeo == 0))
91 goto out;
92
93 /* wait for a message to turn up */
94 release_sock(&rx->sk);
95 prepare_to_wait_exclusive(sk_sleep(&rx->sk), &wait,
96 TASK_INTERRUPTIBLE);
97 ret = sock_error(&rx->sk);
98 if (ret)
99 goto wait_error;
100
101 if (skb_queue_empty(&rx->sk.sk_receive_queue)) {
102 if (signal_pending(current))
103 goto wait_interrupted;
104 timeo = schedule_timeout(timeo);
105 }
106 finish_wait(sk_sleep(&rx->sk), &wait);
107 lock_sock(&rx->sk);
108 continue;
109 }
110
111 peek_next_packet:
112 sp = rxrpc_skb(skb);
113 call = sp->call;
114 ASSERT(call != NULL);
115
116 _debug("next pkt %s", rxrpc_pkts[sp->hdr.type]);
117
118 /* make sure we wait for the state to be updated in this call */
119 spin_lock_bh(&call->lock);
120 spin_unlock_bh(&call->lock);
121
122 if (test_bit(RXRPC_CALL_RELEASED, &call->flags)) {
123 _debug("packet from released call");
124 if (skb_dequeue(&rx->sk.sk_receive_queue) != skb)
125 BUG();
126 rxrpc_free_skb(skb);
127 continue;
128 }
129
130 /* determine whether to continue last data receive */
131 if (continue_call) {
132 _debug("maybe cont");
133 if (call != continue_call ||
134 skb->mark != RXRPC_SKB_MARK_DATA) {
135 release_sock(&rx->sk);
136 rxrpc_put_call(continue_call);
137 _leave(" = %d [noncont]", copied);
138 return copied;
139 }
140 }
141
142 rxrpc_get_call(call);
143
144 /* copy the peer address and timestamp */
145 if (!continue_call) {
146 if (msg->msg_name) {
147 size_t len =
148 sizeof(call->conn->trans->peer->srx);
149 memcpy(msg->msg_name,
150 &call->conn->trans->peer->srx, len);
151 msg->msg_namelen = len;
152 }
153 sock_recv_ts_and_drops(msg, &rx->sk, skb);
154 }
155
156 /* receive the message */
157 if (skb->mark != RXRPC_SKB_MARK_DATA)
158 goto receive_non_data_message;
159
160 _debug("recvmsg DATA #%u { %d, %d }",
161 ntohl(sp->hdr.seq), skb->len, sp->offset);
162
163 if (!continue_call) {
164 /* only set the control data once per recvmsg() */
165 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
166 ullen, &call->user_call_ID);
167 if (ret < 0)
168 goto copy_error;
169 ASSERT(test_bit(RXRPC_CALL_HAS_USERID, &call->flags));
170 }
171
172 ASSERTCMP(ntohl(sp->hdr.seq), >=, call->rx_data_recv);
173 ASSERTCMP(ntohl(sp->hdr.seq), <=, call->rx_data_recv + 1);
174 call->rx_data_recv = ntohl(sp->hdr.seq);
175
176 ASSERTCMP(ntohl(sp->hdr.seq), >, call->rx_data_eaten);
177
178 offset = sp->offset;
179 copy = skb->len - offset;
180 if (copy > len - copied)
181 copy = len - copied;
182
183 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
184 ret = skb_copy_datagram_iovec(skb, offset,
185 msg->msg_iov, copy);
186 } else {
187 ret = skb_copy_and_csum_datagram_iovec(skb, offset,
188 msg->msg_iov);
189 if (ret == -EINVAL)
190 goto csum_copy_error;
191 }
192
193 if (ret < 0)
194 goto copy_error;
195
196 /* handle piecemeal consumption of data packets */
197 _debug("copied %d+%d", copy, copied);
198
199 offset += copy;
200 copied += copy;
201
202 if (!(flags & MSG_PEEK))
203 sp->offset = offset;
204
205 if (sp->offset < skb->len) {
206 _debug("buffer full");
207 ASSERTCMP(copied, ==, len);
208 break;
209 }
210
211 /* we transferred the whole data packet */
212 if (sp->hdr.flags & RXRPC_LAST_PACKET) {
213 _debug("last");
214 if (call->conn->out_clientflag) {
215 /* last byte of reply received */
216 ret = copied;
217 goto terminal_message;
218 }
219
220 /* last bit of request received */
221 if (!(flags & MSG_PEEK)) {
222 _debug("eat packet");
223 if (skb_dequeue(&rx->sk.sk_receive_queue) !=
224 skb)
225 BUG();
226 rxrpc_free_skb(skb);
227 }
228 msg->msg_flags &= ~MSG_MORE;
229 break;
230 }
231
232 /* move on to the next data message */
233 _debug("next");
234 if (!continue_call)
235 continue_call = sp->call;
236 else
237 rxrpc_put_call(call);
238 call = NULL;
239
240 if (flags & MSG_PEEK) {
241 _debug("peek next");
242 skb = skb->next;
243 if (skb == (struct sk_buff *) &rx->sk.sk_receive_queue)
244 break;
245 goto peek_next_packet;
246 }
247
248 _debug("eat packet");
249 if (skb_dequeue(&rx->sk.sk_receive_queue) != skb)
250 BUG();
251 rxrpc_free_skb(skb);
252 }
253
254 /* end of non-terminal data packet reception for the moment */
255 _debug("end rcv data");
256 out:
257 release_sock(&rx->sk);
258 if (call)
259 rxrpc_put_call(call);
260 if (continue_call)
261 rxrpc_put_call(continue_call);
262 _leave(" = %d [data]", copied);
263 return copied;
264
265 /* handle non-DATA messages such as aborts, incoming connections and
266 * final ACKs */
267 receive_non_data_message:
268 _debug("non-data");
269
270 if (skb->mark == RXRPC_SKB_MARK_NEW_CALL) {
271 _debug("RECV NEW CALL");
272 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NEW_CALL, 0, &abort_code);
273 if (ret < 0)
274 goto copy_error;
275 if (!(flags & MSG_PEEK)) {
276 if (skb_dequeue(&rx->sk.sk_receive_queue) != skb)
277 BUG();
278 rxrpc_free_skb(skb);
279 }
280 goto out;
281 }
282
283 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_USER_CALL_ID,
284 ullen, &call->user_call_ID);
285 if (ret < 0)
286 goto copy_error;
287 ASSERT(test_bit(RXRPC_CALL_HAS_USERID, &call->flags));
288
289 switch (skb->mark) {
290 case RXRPC_SKB_MARK_DATA:
291 BUG();
292 case RXRPC_SKB_MARK_FINAL_ACK:
293 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ACK, 0, &abort_code);
294 break;
295 case RXRPC_SKB_MARK_BUSY:
296 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_BUSY, 0, &abort_code);
297 break;
298 case RXRPC_SKB_MARK_REMOTE_ABORT:
299 abort_code = call->abort_code;
300 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_ABORT, 4, &abort_code);
301 break;
302 case RXRPC_SKB_MARK_NET_ERROR:
303 _debug("RECV NET ERROR %d", sp->error);
304 abort_code = sp->error;
305 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_NET_ERROR, 4, &abort_code);
306 break;
307 case RXRPC_SKB_MARK_LOCAL_ERROR:
308 _debug("RECV LOCAL ERROR %d", sp->error);
309 abort_code = sp->error;
310 ret = put_cmsg(msg, SOL_RXRPC, RXRPC_LOCAL_ERROR, 4,
311 &abort_code);
312 break;
313 default:
314 BUG();
315 break;
316 }
317
318 if (ret < 0)
319 goto copy_error;
320
321 terminal_message:
322 _debug("terminal");
323 msg->msg_flags &= ~MSG_MORE;
324 msg->msg_flags |= MSG_EOR;
325
326 if (!(flags & MSG_PEEK)) {
327 _net("free terminal skb %p", skb);
328 if (skb_dequeue(&rx->sk.sk_receive_queue) != skb)
329 BUG();
330 rxrpc_free_skb(skb);
331 rxrpc_remove_user_ID(rx, call);
332 }
333
334 release_sock(&rx->sk);
335 rxrpc_put_call(call);
336 if (continue_call)
337 rxrpc_put_call(continue_call);
338 _leave(" = %d", ret);
339 return ret;
340
341 copy_error:
342 _debug("copy error");
343 release_sock(&rx->sk);
344 rxrpc_put_call(call);
345 if (continue_call)
346 rxrpc_put_call(continue_call);
347 _leave(" = %d", ret);
348 return ret;
349
350 csum_copy_error:
351 _debug("csum error");
352 release_sock(&rx->sk);
353 if (continue_call)
354 rxrpc_put_call(continue_call);
355 rxrpc_kill_skb(skb);
356 skb_kill_datagram(&rx->sk, skb, flags);
357 rxrpc_put_call(call);
358 return -EAGAIN;
359
360 wait_interrupted:
361 ret = sock_intr_errno(timeo);
362 wait_error:
363 finish_wait(sk_sleep(&rx->sk), &wait);
364 if (continue_call)
365 rxrpc_put_call(continue_call);
366 if (copied)
367 copied = ret;
368 _leave(" = %d [waitfail %d]", copied, ret);
369 return copied;
370
371 }
372
373 /**
374 * rxrpc_kernel_data_delivered - Record delivery of data message
375 * @skb: Message holding data
376 *
377 * Record the delivery of a data message. This permits RxRPC to keep its
378 * tracking correct. The socket buffer will be deleted.
379 */
380 void rxrpc_kernel_data_delivered(struct sk_buff *skb)
381 {
382 struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
383 struct rxrpc_call *call = sp->call;
384
385 ASSERTCMP(ntohl(sp->hdr.seq), >=, call->rx_data_recv);
386 ASSERTCMP(ntohl(sp->hdr.seq), <=, call->rx_data_recv + 1);
387 call->rx_data_recv = ntohl(sp->hdr.seq);
388
389 ASSERTCMP(ntohl(sp->hdr.seq), >, call->rx_data_eaten);
390 rxrpc_free_skb(skb);
391 }
392
393 EXPORT_SYMBOL(rxrpc_kernel_data_delivered);
394
395 /**
396 * rxrpc_kernel_is_data_last - Determine if data message is last one
397 * @skb: Message holding data
398 *
399 * Determine if data message is last one for the parent call.
400 */
401 bool rxrpc_kernel_is_data_last(struct sk_buff *skb)
402 {
403 struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
404
405 ASSERTCMP(skb->mark, ==, RXRPC_SKB_MARK_DATA);
406
407 return sp->hdr.flags & RXRPC_LAST_PACKET;
408 }
409
410 EXPORT_SYMBOL(rxrpc_kernel_is_data_last);
411
412 /**
413 * rxrpc_kernel_get_abort_code - Get the abort code from an RxRPC abort message
414 * @skb: Message indicating an abort
415 *
416 * Get the abort code from an RxRPC abort message.
417 */
418 u32 rxrpc_kernel_get_abort_code(struct sk_buff *skb)
419 {
420 struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
421
422 ASSERTCMP(skb->mark, ==, RXRPC_SKB_MARK_REMOTE_ABORT);
423
424 return sp->call->abort_code;
425 }
426
427 EXPORT_SYMBOL(rxrpc_kernel_get_abort_code);
428
429 /**
430 * rxrpc_kernel_get_error - Get the error number from an RxRPC error message
431 * @skb: Message indicating an error
432 *
433 * Get the error number from an RxRPC error message.
434 */
435 int rxrpc_kernel_get_error_number(struct sk_buff *skb)
436 {
437 struct rxrpc_skb_priv *sp = rxrpc_skb(skb);
438
439 return sp->error;
440 }
441
442 EXPORT_SYMBOL(rxrpc_kernel_get_error_number);