xref: /linux/fs/afs/rxrpc.c (revision eb2bce7f5e7ac1ca6da434461217fadf3c688d2c)
1 /* Maintain an RxRPC server socket to do AFS communications through
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 <net/sock.h>
13 #include <net/af_rxrpc.h>
14 #include <rxrpc/packet.h>
15 #include "internal.h"
16 #include "afs_cm.h"
17 
18 static struct socket *afs_socket; /* my RxRPC socket */
19 static struct workqueue_struct *afs_async_calls;
20 static atomic_t afs_outstanding_calls;
21 static atomic_t afs_outstanding_skbs;
22 
23 static void afs_wake_up_call_waiter(struct afs_call *);
24 static int afs_wait_for_call_to_complete(struct afs_call *);
25 static void afs_wake_up_async_call(struct afs_call *);
26 static int afs_dont_wait_for_call_to_complete(struct afs_call *);
27 static void afs_process_async_call(struct work_struct *);
28 static void afs_rx_interceptor(struct sock *, unsigned long, struct sk_buff *);
29 static int afs_deliver_cm_op_id(struct afs_call *, struct sk_buff *, bool);
30 
31 /* synchronous call management */
32 const struct afs_wait_mode afs_sync_call = {
33 	.rx_wakeup	= afs_wake_up_call_waiter,
34 	.wait		= afs_wait_for_call_to_complete,
35 };
36 
37 /* asynchronous call management */
38 const struct afs_wait_mode afs_async_call = {
39 	.rx_wakeup	= afs_wake_up_async_call,
40 	.wait		= afs_dont_wait_for_call_to_complete,
41 };
42 
43 /* asynchronous incoming call management */
44 static const struct afs_wait_mode afs_async_incoming_call = {
45 	.rx_wakeup	= afs_wake_up_async_call,
46 };
47 
48 /* asynchronous incoming call initial processing */
49 static const struct afs_call_type afs_RXCMxxxx = {
50 	.name		= "CB.xxxx",
51 	.deliver	= afs_deliver_cm_op_id,
52 	.abort_to_error	= afs_abort_to_error,
53 };
54 
55 static void afs_collect_incoming_call(struct work_struct *);
56 
57 static struct sk_buff_head afs_incoming_calls;
58 static DECLARE_WORK(afs_collect_incoming_call_work, afs_collect_incoming_call);
59 
60 /*
61  * open an RxRPC socket and bind it to be a server for callback notifications
62  * - the socket is left in blocking mode and non-blocking ops use MSG_DONTWAIT
63  */
64 int afs_open_socket(void)
65 {
66 	struct sockaddr_rxrpc srx;
67 	struct socket *socket;
68 	int ret;
69 
70 	_enter("");
71 
72 	skb_queue_head_init(&afs_incoming_calls);
73 
74 	afs_async_calls = create_singlethread_workqueue("kafsd");
75 	if (!afs_async_calls) {
76 		_leave(" = -ENOMEM [wq]");
77 		return -ENOMEM;
78 	}
79 
80 	ret = sock_create_kern(AF_RXRPC, SOCK_DGRAM, PF_INET, &socket);
81 	if (ret < 0) {
82 		destroy_workqueue(afs_async_calls);
83 		_leave(" = %d [socket]", ret);
84 		return ret;
85 	}
86 
87 	socket->sk->sk_allocation = GFP_NOFS;
88 
89 	/* bind the callback manager's address to make this a server socket */
90 	srx.srx_family			= AF_RXRPC;
91 	srx.srx_service			= CM_SERVICE;
92 	srx.transport_type		= SOCK_DGRAM;
93 	srx.transport_len		= sizeof(srx.transport.sin);
94 	srx.transport.sin.sin_family	= AF_INET;
95 	srx.transport.sin.sin_port	= htons(AFS_CM_PORT);
96 	memset(&srx.transport.sin.sin_addr, 0,
97 	       sizeof(srx.transport.sin.sin_addr));
98 
99 	ret = kernel_bind(socket, (struct sockaddr *) &srx, sizeof(srx));
100 	if (ret < 0) {
101 		sock_release(socket);
102 		_leave(" = %d [bind]", ret);
103 		return ret;
104 	}
105 
106 	rxrpc_kernel_intercept_rx_messages(socket, afs_rx_interceptor);
107 
108 	afs_socket = socket;
109 	_leave(" = 0");
110 	return 0;
111 }
112 
113 /*
114  * close the RxRPC socket AFS was using
115  */
116 void afs_close_socket(void)
117 {
118 	_enter("");
119 
120 	sock_release(afs_socket);
121 
122 	_debug("dework");
123 	destroy_workqueue(afs_async_calls);
124 
125 	ASSERTCMP(atomic_read(&afs_outstanding_skbs), ==, 0);
126 	ASSERTCMP(atomic_read(&afs_outstanding_calls), ==, 0);
127 	_leave("");
128 }
129 
130 /*
131  * note that the data in a socket buffer is now delivered and that the buffer
132  * should be freed
133  */
134 static void afs_data_delivered(struct sk_buff *skb)
135 {
136 	if (!skb) {
137 		_debug("DLVR NULL [%d]", atomic_read(&afs_outstanding_skbs));
138 		dump_stack();
139 	} else {
140 		_debug("DLVR %p{%u} [%d]",
141 		       skb, skb->mark, atomic_read(&afs_outstanding_skbs));
142 		if (atomic_dec_return(&afs_outstanding_skbs) == -1)
143 			BUG();
144 		rxrpc_kernel_data_delivered(skb);
145 	}
146 }
147 
148 /*
149  * free a socket buffer
150  */
151 static void afs_free_skb(struct sk_buff *skb)
152 {
153 	if (!skb) {
154 		_debug("FREE NULL [%d]", atomic_read(&afs_outstanding_skbs));
155 		dump_stack();
156 	} else {
157 		_debug("FREE %p{%u} [%d]",
158 		       skb, skb->mark, atomic_read(&afs_outstanding_skbs));
159 		if (atomic_dec_return(&afs_outstanding_skbs) == -1)
160 			BUG();
161 		rxrpc_kernel_free_skb(skb);
162 	}
163 }
164 
165 /*
166  * free a call
167  */
168 static void afs_free_call(struct afs_call *call)
169 {
170 	_debug("DONE %p{%s} [%d]",
171 	       call, call->type->name, atomic_read(&afs_outstanding_calls));
172 	if (atomic_dec_return(&afs_outstanding_calls) == -1)
173 		BUG();
174 
175 	ASSERTCMP(call->rxcall, ==, NULL);
176 	ASSERT(!work_pending(&call->async_work));
177 	ASSERT(skb_queue_empty(&call->rx_queue));
178 	ASSERT(call->type->name != NULL);
179 
180 	kfree(call->request);
181 	kfree(call);
182 }
183 
184 /*
185  * allocate a call with flat request and reply buffers
186  */
187 struct afs_call *afs_alloc_flat_call(const struct afs_call_type *type,
188 				     size_t request_size, size_t reply_size)
189 {
190 	struct afs_call *call;
191 
192 	call = kzalloc(sizeof(*call), GFP_NOFS);
193 	if (!call)
194 		goto nomem_call;
195 
196 	_debug("CALL %p{%s} [%d]",
197 	       call, type->name, atomic_read(&afs_outstanding_calls));
198 	atomic_inc(&afs_outstanding_calls);
199 
200 	call->type = type;
201 	call->request_size = request_size;
202 	call->reply_max = reply_size;
203 
204 	if (request_size) {
205 		call->request = kmalloc(request_size, GFP_NOFS);
206 		if (!call->request)
207 			goto nomem_free;
208 	}
209 
210 	if (reply_size) {
211 		call->buffer = kmalloc(reply_size, GFP_NOFS);
212 		if (!call->buffer)
213 			goto nomem_free;
214 	}
215 
216 	init_waitqueue_head(&call->waitq);
217 	skb_queue_head_init(&call->rx_queue);
218 	return call;
219 
220 nomem_free:
221 	afs_free_call(call);
222 nomem_call:
223 	return NULL;
224 }
225 
226 /*
227  * clean up a call with flat buffer
228  */
229 void afs_flat_call_destructor(struct afs_call *call)
230 {
231 	_enter("");
232 
233 	kfree(call->request);
234 	call->request = NULL;
235 	kfree(call->buffer);
236 	call->buffer = NULL;
237 }
238 
239 /*
240  * initiate a call
241  */
242 int afs_make_call(struct in_addr *addr, struct afs_call *call, gfp_t gfp,
243 		  const struct afs_wait_mode *wait_mode)
244 {
245 	struct sockaddr_rxrpc srx;
246 	struct rxrpc_call *rxcall;
247 	struct msghdr msg;
248 	struct kvec iov[1];
249 	int ret;
250 
251 	_enter("%x,{%d},", addr->s_addr, ntohs(call->port));
252 
253 	ASSERT(call->type != NULL);
254 	ASSERT(call->type->name != NULL);
255 
256 	_debug("MAKE %p{%s} [%d]",
257 	       call, call->type->name, atomic_read(&afs_outstanding_calls));
258 
259 	call->wait_mode = wait_mode;
260 	INIT_WORK(&call->async_work, afs_process_async_call);
261 
262 	memset(&srx, 0, sizeof(srx));
263 	srx.srx_family = AF_RXRPC;
264 	srx.srx_service = call->service_id;
265 	srx.transport_type = SOCK_DGRAM;
266 	srx.transport_len = sizeof(srx.transport.sin);
267 	srx.transport.sin.sin_family = AF_INET;
268 	srx.transport.sin.sin_port = call->port;
269 	memcpy(&srx.transport.sin.sin_addr, addr, 4);
270 
271 	/* create a call */
272 	rxcall = rxrpc_kernel_begin_call(afs_socket, &srx, call->key,
273 					 (unsigned long) call, gfp);
274 	call->key = NULL;
275 	if (IS_ERR(rxcall)) {
276 		ret = PTR_ERR(rxcall);
277 		goto error_kill_call;
278 	}
279 
280 	call->rxcall = rxcall;
281 
282 	/* send the request */
283 	iov[0].iov_base	= call->request;
284 	iov[0].iov_len	= call->request_size;
285 
286 	msg.msg_name		= NULL;
287 	msg.msg_namelen		= 0;
288 	msg.msg_iov		= (struct iovec *) iov;
289 	msg.msg_iovlen		= 1;
290 	msg.msg_control		= NULL;
291 	msg.msg_controllen	= 0;
292 	msg.msg_flags		= 0;
293 
294 	/* have to change the state *before* sending the last packet as RxRPC
295 	 * might give us the reply before it returns from sending the
296 	 * request */
297 	call->state = AFS_CALL_AWAIT_REPLY;
298 	ret = rxrpc_kernel_send_data(rxcall, &msg, call->request_size);
299 	if (ret < 0)
300 		goto error_do_abort;
301 
302 	/* at this point, an async call may no longer exist as it may have
303 	 * already completed */
304 	return wait_mode->wait(call);
305 
306 error_do_abort:
307 	rxrpc_kernel_abort_call(rxcall, RX_USER_ABORT);
308 	rxrpc_kernel_end_call(rxcall);
309 	call->rxcall = NULL;
310 error_kill_call:
311 	call->type->destructor(call);
312 	afs_free_call(call);
313 	_leave(" = %d", ret);
314 	return ret;
315 }
316 
317 /*
318  * handles intercepted messages that were arriving in the socket's Rx queue
319  * - called with the socket receive queue lock held to ensure message ordering
320  * - called with softirqs disabled
321  */
322 static void afs_rx_interceptor(struct sock *sk, unsigned long user_call_ID,
323 			       struct sk_buff *skb)
324 {
325 	struct afs_call *call = (struct afs_call *) user_call_ID;
326 
327 	_enter("%p,,%u", call, skb->mark);
328 
329 	_debug("ICPT %p{%u} [%d]",
330 	       skb, skb->mark, atomic_read(&afs_outstanding_skbs));
331 
332 	ASSERTCMP(sk, ==, afs_socket->sk);
333 	atomic_inc(&afs_outstanding_skbs);
334 
335 	if (!call) {
336 		/* its an incoming call for our callback service */
337 		skb_queue_tail(&afs_incoming_calls, skb);
338 		schedule_work(&afs_collect_incoming_call_work);
339 	} else {
340 		/* route the messages directly to the appropriate call */
341 		skb_queue_tail(&call->rx_queue, skb);
342 		call->wait_mode->rx_wakeup(call);
343 	}
344 
345 	_leave("");
346 }
347 
348 /*
349  * deliver messages to a call
350  */
351 static void afs_deliver_to_call(struct afs_call *call)
352 {
353 	struct sk_buff *skb;
354 	bool last;
355 	u32 abort_code;
356 	int ret;
357 
358 	_enter("");
359 
360 	while ((call->state == AFS_CALL_AWAIT_REPLY ||
361 		call->state == AFS_CALL_AWAIT_OP_ID ||
362 		call->state == AFS_CALL_AWAIT_REQUEST ||
363 		call->state == AFS_CALL_AWAIT_ACK) &&
364 	       (skb = skb_dequeue(&call->rx_queue))) {
365 		switch (skb->mark) {
366 		case RXRPC_SKB_MARK_DATA:
367 			_debug("Rcv DATA");
368 			last = rxrpc_kernel_is_data_last(skb);
369 			ret = call->type->deliver(call, skb, last);
370 			switch (ret) {
371 			case 0:
372 				if (last &&
373 				    call->state == AFS_CALL_AWAIT_REPLY)
374 					call->state = AFS_CALL_COMPLETE;
375 				break;
376 			case -ENOTCONN:
377 				abort_code = RX_CALL_DEAD;
378 				goto do_abort;
379 			case -ENOTSUPP:
380 				abort_code = RX_INVALID_OPERATION;
381 				goto do_abort;
382 			default:
383 				abort_code = RXGEN_CC_UNMARSHAL;
384 				if (call->state != AFS_CALL_AWAIT_REPLY)
385 					abort_code = RXGEN_SS_UNMARSHAL;
386 			do_abort:
387 				rxrpc_kernel_abort_call(call->rxcall,
388 							abort_code);
389 				call->error = ret;
390 				call->state = AFS_CALL_ERROR;
391 				break;
392 			}
393 			afs_data_delivered(skb);
394 			skb = NULL;
395 			continue;
396 		case RXRPC_SKB_MARK_FINAL_ACK:
397 			_debug("Rcv ACK");
398 			call->state = AFS_CALL_COMPLETE;
399 			break;
400 		case RXRPC_SKB_MARK_BUSY:
401 			_debug("Rcv BUSY");
402 			call->error = -EBUSY;
403 			call->state = AFS_CALL_BUSY;
404 			break;
405 		case RXRPC_SKB_MARK_REMOTE_ABORT:
406 			abort_code = rxrpc_kernel_get_abort_code(skb);
407 			call->error = call->type->abort_to_error(abort_code);
408 			call->state = AFS_CALL_ABORTED;
409 			_debug("Rcv ABORT %u -> %d", abort_code, call->error);
410 			break;
411 		case RXRPC_SKB_MARK_NET_ERROR:
412 			call->error = -rxrpc_kernel_get_error_number(skb);
413 			call->state = AFS_CALL_ERROR;
414 			_debug("Rcv NET ERROR %d", call->error);
415 			break;
416 		case RXRPC_SKB_MARK_LOCAL_ERROR:
417 			call->error = -rxrpc_kernel_get_error_number(skb);
418 			call->state = AFS_CALL_ERROR;
419 			_debug("Rcv LOCAL ERROR %d", call->error);
420 			break;
421 		default:
422 			BUG();
423 			break;
424 		}
425 
426 		afs_free_skb(skb);
427 	}
428 
429 	/* make sure the queue is empty if the call is done with (we might have
430 	 * aborted the call early because of an unmarshalling error) */
431 	if (call->state >= AFS_CALL_COMPLETE) {
432 		while ((skb = skb_dequeue(&call->rx_queue)))
433 			afs_free_skb(skb);
434 		if (call->incoming) {
435 			rxrpc_kernel_end_call(call->rxcall);
436 			call->rxcall = NULL;
437 			call->type->destructor(call);
438 			afs_free_call(call);
439 		}
440 	}
441 
442 	_leave("");
443 }
444 
445 /*
446  * wait synchronously for a call to complete
447  */
448 static int afs_wait_for_call_to_complete(struct afs_call *call)
449 {
450 	struct sk_buff *skb;
451 	int ret;
452 
453 	DECLARE_WAITQUEUE(myself, current);
454 
455 	_enter("");
456 
457 	add_wait_queue(&call->waitq, &myself);
458 	for (;;) {
459 		set_current_state(TASK_INTERRUPTIBLE);
460 
461 		/* deliver any messages that are in the queue */
462 		if (!skb_queue_empty(&call->rx_queue)) {
463 			__set_current_state(TASK_RUNNING);
464 			afs_deliver_to_call(call);
465 			continue;
466 		}
467 
468 		ret = call->error;
469 		if (call->state >= AFS_CALL_COMPLETE)
470 			break;
471 		ret = -EINTR;
472 		if (signal_pending(current))
473 			break;
474 		schedule();
475 	}
476 
477 	remove_wait_queue(&call->waitq, &myself);
478 	__set_current_state(TASK_RUNNING);
479 
480 	/* kill the call */
481 	if (call->state < AFS_CALL_COMPLETE) {
482 		_debug("call incomplete");
483 		rxrpc_kernel_abort_call(call->rxcall, RX_CALL_DEAD);
484 		while ((skb = skb_dequeue(&call->rx_queue)))
485 			afs_free_skb(skb);
486 	}
487 
488 	_debug("call complete");
489 	rxrpc_kernel_end_call(call->rxcall);
490 	call->rxcall = NULL;
491 	call->type->destructor(call);
492 	afs_free_call(call);
493 	_leave(" = %d", ret);
494 	return ret;
495 }
496 
497 /*
498  * wake up a waiting call
499  */
500 static void afs_wake_up_call_waiter(struct afs_call *call)
501 {
502 	wake_up(&call->waitq);
503 }
504 
505 /*
506  * wake up an asynchronous call
507  */
508 static void afs_wake_up_async_call(struct afs_call *call)
509 {
510 	_enter("");
511 	queue_work(afs_async_calls, &call->async_work);
512 }
513 
514 /*
515  * put a call into asynchronous mode
516  * - mustn't touch the call descriptor as the call my have completed by the
517  *   time we get here
518  */
519 static int afs_dont_wait_for_call_to_complete(struct afs_call *call)
520 {
521 	_enter("");
522 	return -EINPROGRESS;
523 }
524 
525 /*
526  * delete an asynchronous call
527  */
528 static void afs_delete_async_call(struct work_struct *work)
529 {
530 	struct afs_call *call =
531 		container_of(work, struct afs_call, async_work);
532 
533 	_enter("");
534 
535 	afs_free_call(call);
536 
537 	_leave("");
538 }
539 
540 /*
541  * perform processing on an asynchronous call
542  * - on a multiple-thread workqueue this work item may try to run on several
543  *   CPUs at the same time
544  */
545 static void afs_process_async_call(struct work_struct *work)
546 {
547 	struct afs_call *call =
548 		container_of(work, struct afs_call, async_work);
549 
550 	_enter("");
551 
552 	if (!skb_queue_empty(&call->rx_queue))
553 		afs_deliver_to_call(call);
554 
555 	if (call->state >= AFS_CALL_COMPLETE && call->wait_mode) {
556 		if (call->wait_mode->async_complete)
557 			call->wait_mode->async_complete(call->reply,
558 							call->error);
559 		call->reply = NULL;
560 
561 		/* kill the call */
562 		rxrpc_kernel_end_call(call->rxcall);
563 		call->rxcall = NULL;
564 		if (call->type->destructor)
565 			call->type->destructor(call);
566 
567 		/* we can't just delete the call because the work item may be
568 		 * queued */
569 		PREPARE_WORK(&call->async_work, afs_delete_async_call);
570 		queue_work(afs_async_calls, &call->async_work);
571 	}
572 
573 	_leave("");
574 }
575 
576 /*
577  * empty a socket buffer into a flat reply buffer
578  */
579 void afs_transfer_reply(struct afs_call *call, struct sk_buff *skb)
580 {
581 	size_t len = skb->len;
582 
583 	if (skb_copy_bits(skb, 0, call->buffer + call->reply_size, len) < 0)
584 		BUG();
585 	call->reply_size += len;
586 }
587 
588 /*
589  * accept the backlog of incoming calls
590  */
591 static void afs_collect_incoming_call(struct work_struct *work)
592 {
593 	struct rxrpc_call *rxcall;
594 	struct afs_call *call = NULL;
595 	struct sk_buff *skb;
596 
597 	while ((skb = skb_dequeue(&afs_incoming_calls))) {
598 		_debug("new call");
599 
600 		/* don't need the notification */
601 		afs_free_skb(skb);
602 
603 		if (!call) {
604 			call = kzalloc(sizeof(struct afs_call), GFP_KERNEL);
605 			if (!call) {
606 				rxrpc_kernel_reject_call(afs_socket);
607 				return;
608 			}
609 
610 			INIT_WORK(&call->async_work, afs_process_async_call);
611 			call->wait_mode = &afs_async_incoming_call;
612 			call->type = &afs_RXCMxxxx;
613 			init_waitqueue_head(&call->waitq);
614 			skb_queue_head_init(&call->rx_queue);
615 			call->state = AFS_CALL_AWAIT_OP_ID;
616 
617 			_debug("CALL %p{%s} [%d]",
618 			       call, call->type->name,
619 			       atomic_read(&afs_outstanding_calls));
620 			atomic_inc(&afs_outstanding_calls);
621 		}
622 
623 		rxcall = rxrpc_kernel_accept_call(afs_socket,
624 						  (unsigned long) call);
625 		if (!IS_ERR(rxcall)) {
626 			call->rxcall = rxcall;
627 			call = NULL;
628 		}
629 	}
630 
631 	if (call)
632 		afs_free_call(call);
633 }
634 
635 /*
636  * grab the operation ID from an incoming cache manager call
637  */
638 static int afs_deliver_cm_op_id(struct afs_call *call, struct sk_buff *skb,
639 				bool last)
640 {
641 	size_t len = skb->len;
642 	void *oibuf = (void *) &call->operation_ID;
643 
644 	_enter("{%u},{%zu},%d", call->offset, len, last);
645 
646 	ASSERTCMP(call->offset, <, 4);
647 
648 	/* the operation ID forms the first four bytes of the request data */
649 	len = min_t(size_t, len, 4 - call->offset);
650 	if (skb_copy_bits(skb, 0, oibuf + call->offset, len) < 0)
651 		BUG();
652 	if (!pskb_pull(skb, len))
653 		BUG();
654 	call->offset += len;
655 
656 	if (call->offset < 4) {
657 		if (last) {
658 			_leave(" = -EBADMSG [op ID short]");
659 			return -EBADMSG;
660 		}
661 		_leave(" = 0 [incomplete]");
662 		return 0;
663 	}
664 
665 	call->state = AFS_CALL_AWAIT_REQUEST;
666 
667 	/* ask the cache manager to route the call (it'll change the call type
668 	 * if successful) */
669 	if (!afs_cm_incoming_call(call))
670 		return -ENOTSUPP;
671 
672 	/* pass responsibility for the remainer of this message off to the
673 	 * cache manager op */
674 	return call->type->deliver(call, skb, last);
675 }
676 
677 /*
678  * send an empty reply
679  */
680 void afs_send_empty_reply(struct afs_call *call)
681 {
682 	struct msghdr msg;
683 	struct iovec iov[1];
684 
685 	_enter("");
686 
687 	iov[0].iov_base		= NULL;
688 	iov[0].iov_len		= 0;
689 	msg.msg_name		= NULL;
690 	msg.msg_namelen		= 0;
691 	msg.msg_iov		= iov;
692 	msg.msg_iovlen		= 0;
693 	msg.msg_control		= NULL;
694 	msg.msg_controllen	= 0;
695 	msg.msg_flags		= 0;
696 
697 	call->state = AFS_CALL_AWAIT_ACK;
698 	switch (rxrpc_kernel_send_data(call->rxcall, &msg, 0)) {
699 	case 0:
700 		_leave(" [replied]");
701 		return;
702 
703 	case -ENOMEM:
704 		_debug("oom");
705 		rxrpc_kernel_abort_call(call->rxcall, RX_USER_ABORT);
706 	default:
707 		rxrpc_kernel_end_call(call->rxcall);
708 		call->rxcall = NULL;
709 		call->type->destructor(call);
710 		afs_free_call(call);
711 		_leave(" [error]");
712 		return;
713 	}
714 }
715 
716 /*
717  * send a simple reply
718  */
719 void afs_send_simple_reply(struct afs_call *call, const void *buf, size_t len)
720 {
721 	struct msghdr msg;
722 	struct iovec iov[1];
723 
724 	_enter("");
725 
726 	iov[0].iov_base		= (void *) buf;
727 	iov[0].iov_len		= len;
728 	msg.msg_name		= NULL;
729 	msg.msg_namelen		= 0;
730 	msg.msg_iov		= iov;
731 	msg.msg_iovlen		= 1;
732 	msg.msg_control		= NULL;
733 	msg.msg_controllen	= 0;
734 	msg.msg_flags		= 0;
735 
736 	call->state = AFS_CALL_AWAIT_ACK;
737 	switch (rxrpc_kernel_send_data(call->rxcall, &msg, len)) {
738 	case 0:
739 		_leave(" [replied]");
740 		return;
741 
742 	case -ENOMEM:
743 		_debug("oom");
744 		rxrpc_kernel_abort_call(call->rxcall, RX_USER_ABORT);
745 	default:
746 		rxrpc_kernel_end_call(call->rxcall);
747 		call->rxcall = NULL;
748 		call->type->destructor(call);
749 		afs_free_call(call);
750 		_leave(" [error]");
751 		return;
752 	}
753 }
754 
755 /*
756  * extract a piece of data from the received data socket buffers
757  */
758 int afs_extract_data(struct afs_call *call, struct sk_buff *skb,
759 		     bool last, void *buf, size_t count)
760 {
761 	size_t len = skb->len;
762 
763 	_enter("{%u},{%zu},%d,,%zu", call->offset, len, last, count);
764 
765 	ASSERTCMP(call->offset, <, count);
766 
767 	len = min_t(size_t, len, count - call->offset);
768 	if (skb_copy_bits(skb, 0, buf + call->offset, len) < 0 ||
769 	    !pskb_pull(skb, len))
770 		BUG();
771 	call->offset += len;
772 
773 	if (call->offset < count) {
774 		if (last) {
775 			_leave(" = -EBADMSG [%d < %zu]", call->offset, count);
776 			return -EBADMSG;
777 		}
778 		_leave(" = -EAGAIN");
779 		return -EAGAIN;
780 	}
781 	return 0;
782 }
783