xref: /openbmc/linux/fs/afs/rxrpc.c (revision d5532ee7)
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 <linux/slab.h>
13 #include <net/sock.h>
14 #include <net/af_rxrpc.h>
15 #include <rxrpc/packet.h>
16 #include "internal.h"
17 #include "afs_cm.h"
18 
19 static struct socket *afs_socket; /* my RxRPC socket */
20 static struct workqueue_struct *afs_async_calls;
21 static atomic_t afs_outstanding_calls;
22 static atomic_t afs_outstanding_skbs;
23 
24 static void afs_wake_up_call_waiter(struct afs_call *);
25 static int afs_wait_for_call_to_complete(struct afs_call *);
26 static void afs_wake_up_async_call(struct afs_call *);
27 static int afs_dont_wait_for_call_to_complete(struct afs_call *);
28 static void afs_process_async_call(struct work_struct *);
29 static void afs_rx_interceptor(struct sock *, unsigned long, struct sk_buff *);
30 static int afs_deliver_cm_op_id(struct afs_call *, struct sk_buff *, bool);
31 
32 /* synchronous call management */
33 const struct afs_wait_mode afs_sync_call = {
34 	.rx_wakeup	= afs_wake_up_call_waiter,
35 	.wait		= afs_wait_for_call_to_complete,
36 };
37 
38 /* asynchronous call management */
39 const struct afs_wait_mode afs_async_call = {
40 	.rx_wakeup	= afs_wake_up_async_call,
41 	.wait		= afs_dont_wait_for_call_to_complete,
42 };
43 
44 /* asynchronous incoming call management */
45 static const struct afs_wait_mode afs_async_incoming_call = {
46 	.rx_wakeup	= afs_wake_up_async_call,
47 };
48 
49 /* asynchronous incoming call initial processing */
50 static const struct afs_call_type afs_RXCMxxxx = {
51 	.name		= "CB.xxxx",
52 	.deliver	= afs_deliver_cm_op_id,
53 	.abort_to_error	= afs_abort_to_error,
54 };
55 
56 static void afs_collect_incoming_call(struct work_struct *);
57 
58 static struct sk_buff_head afs_incoming_calls;
59 static DECLARE_WORK(afs_collect_incoming_call_work, afs_collect_incoming_call);
60 
61 /*
62  * open an RxRPC socket and bind it to be a server for callback notifications
63  * - the socket is left in blocking mode and non-blocking ops use MSG_DONTWAIT
64  */
65 int afs_open_socket(void)
66 {
67 	struct sockaddr_rxrpc srx;
68 	struct socket *socket;
69 	int ret;
70 
71 	_enter("");
72 
73 	skb_queue_head_init(&afs_incoming_calls);
74 
75 	afs_async_calls = create_singlethread_workqueue("kafsd");
76 	if (!afs_async_calls) {
77 		_leave(" = -ENOMEM [wq]");
78 		return -ENOMEM;
79 	}
80 
81 	ret = sock_create_kern(AF_RXRPC, SOCK_DGRAM, PF_INET, &socket);
82 	if (ret < 0) {
83 		destroy_workqueue(afs_async_calls);
84 		_leave(" = %d [socket]", ret);
85 		return ret;
86 	}
87 
88 	socket->sk->sk_allocation = GFP_NOFS;
89 
90 	/* bind the callback manager's address to make this a server socket */
91 	srx.srx_family			= AF_RXRPC;
92 	srx.srx_service			= CM_SERVICE;
93 	srx.transport_type		= SOCK_DGRAM;
94 	srx.transport_len		= sizeof(srx.transport.sin);
95 	srx.transport.sin.sin_family	= AF_INET;
96 	srx.transport.sin.sin_port	= htons(AFS_CM_PORT);
97 	memset(&srx.transport.sin.sin_addr, 0,
98 	       sizeof(srx.transport.sin.sin_addr));
99 
100 	ret = kernel_bind(socket, (struct sockaddr *) &srx, sizeof(srx));
101 	if (ret < 0) {
102 		sock_release(socket);
103 		destroy_workqueue(afs_async_calls);
104 		_leave(" = %d [bind]", ret);
105 		return ret;
106 	}
107 
108 	rxrpc_kernel_intercept_rx_messages(socket, afs_rx_interceptor);
109 
110 	afs_socket = socket;
111 	_leave(" = 0");
112 	return 0;
113 }
114 
115 /*
116  * close the RxRPC socket AFS was using
117  */
118 void afs_close_socket(void)
119 {
120 	_enter("");
121 
122 	sock_release(afs_socket);
123 
124 	_debug("dework");
125 	destroy_workqueue(afs_async_calls);
126 
127 	ASSERTCMP(atomic_read(&afs_outstanding_skbs), ==, 0);
128 	ASSERTCMP(atomic_read(&afs_outstanding_calls), ==, 0);
129 	_leave("");
130 }
131 
132 /*
133  * note that the data in a socket buffer is now delivered and that the buffer
134  * should be freed
135  */
136 static void afs_data_delivered(struct sk_buff *skb)
137 {
138 	if (!skb) {
139 		_debug("DLVR NULL [%d]", atomic_read(&afs_outstanding_skbs));
140 		dump_stack();
141 	} else {
142 		_debug("DLVR %p{%u} [%d]",
143 		       skb, skb->mark, atomic_read(&afs_outstanding_skbs));
144 		if (atomic_dec_return(&afs_outstanding_skbs) == -1)
145 			BUG();
146 		rxrpc_kernel_data_delivered(skb);
147 	}
148 }
149 
150 /*
151  * free a socket buffer
152  */
153 static void afs_free_skb(struct sk_buff *skb)
154 {
155 	if (!skb) {
156 		_debug("FREE NULL [%d]", atomic_read(&afs_outstanding_skbs));
157 		dump_stack();
158 	} else {
159 		_debug("FREE %p{%u} [%d]",
160 		       skb, skb->mark, atomic_read(&afs_outstanding_skbs));
161 		if (atomic_dec_return(&afs_outstanding_skbs) == -1)
162 			BUG();
163 		rxrpc_kernel_free_skb(skb);
164 	}
165 }
166 
167 /*
168  * free a call
169  */
170 static void afs_free_call(struct afs_call *call)
171 {
172 	_debug("DONE %p{%s} [%d]",
173 	       call, call->type->name, atomic_read(&afs_outstanding_calls));
174 	if (atomic_dec_return(&afs_outstanding_calls) == -1)
175 		BUG();
176 
177 	ASSERTCMP(call->rxcall, ==, NULL);
178 	ASSERT(!work_pending(&call->async_work));
179 	ASSERT(skb_queue_empty(&call->rx_queue));
180 	ASSERT(call->type->name != NULL);
181 
182 	kfree(call->request);
183 	kfree(call);
184 }
185 
186 /*
187  * allocate a call with flat request and reply buffers
188  */
189 struct afs_call *afs_alloc_flat_call(const struct afs_call_type *type,
190 				     size_t request_size, size_t reply_size)
191 {
192 	struct afs_call *call;
193 
194 	call = kzalloc(sizeof(*call), GFP_NOFS);
195 	if (!call)
196 		goto nomem_call;
197 
198 	_debug("CALL %p{%s} [%d]",
199 	       call, type->name, atomic_read(&afs_outstanding_calls));
200 	atomic_inc(&afs_outstanding_calls);
201 
202 	call->type = type;
203 	call->request_size = request_size;
204 	call->reply_max = reply_size;
205 
206 	if (request_size) {
207 		call->request = kmalloc(request_size, GFP_NOFS);
208 		if (!call->request)
209 			goto nomem_free;
210 	}
211 
212 	if (reply_size) {
213 		call->buffer = kmalloc(reply_size, GFP_NOFS);
214 		if (!call->buffer)
215 			goto nomem_free;
216 	}
217 
218 	init_waitqueue_head(&call->waitq);
219 	skb_queue_head_init(&call->rx_queue);
220 	return call;
221 
222 nomem_free:
223 	afs_free_call(call);
224 nomem_call:
225 	return NULL;
226 }
227 
228 /*
229  * clean up a call with flat buffer
230  */
231 void afs_flat_call_destructor(struct afs_call *call)
232 {
233 	_enter("");
234 
235 	kfree(call->request);
236 	call->request = NULL;
237 	kfree(call->buffer);
238 	call->buffer = NULL;
239 }
240 
241 /*
242  * attach the data from a bunch of pages on an inode to a call
243  */
244 static int afs_send_pages(struct afs_call *call, struct msghdr *msg,
245 			  struct kvec *iov)
246 {
247 	struct page *pages[8];
248 	unsigned count, n, loop, offset, to;
249 	pgoff_t first = call->first, last = call->last;
250 	int ret;
251 
252 	_enter("");
253 
254 	offset = call->first_offset;
255 	call->first_offset = 0;
256 
257 	do {
258 		_debug("attach %lx-%lx", first, last);
259 
260 		count = last - first + 1;
261 		if (count > ARRAY_SIZE(pages))
262 			count = ARRAY_SIZE(pages);
263 		n = find_get_pages_contig(call->mapping, first, count, pages);
264 		ASSERTCMP(n, ==, count);
265 
266 		loop = 0;
267 		do {
268 			msg->msg_flags = 0;
269 			to = PAGE_SIZE;
270 			if (first + loop >= last)
271 				to = call->last_to;
272 			else
273 				msg->msg_flags = MSG_MORE;
274 			iov->iov_base = kmap(pages[loop]) + offset;
275 			iov->iov_len = to - offset;
276 			offset = 0;
277 
278 			_debug("- range %u-%u%s",
279 			       offset, to, msg->msg_flags ? " [more]" : "");
280 			msg->msg_iov = (struct iovec *) iov;
281 			msg->msg_iovlen = 1;
282 
283 			/* have to change the state *before* sending the last
284 			 * packet as RxRPC might give us the reply before it
285 			 * returns from sending the request */
286 			if (first + loop >= last)
287 				call->state = AFS_CALL_AWAIT_REPLY;
288 			ret = rxrpc_kernel_send_data(call->rxcall, msg,
289 						     to - offset);
290 			kunmap(pages[loop]);
291 			if (ret < 0)
292 				break;
293 		} while (++loop < count);
294 		first += count;
295 
296 		for (loop = 0; loop < count; loop++)
297 			put_page(pages[loop]);
298 		if (ret < 0)
299 			break;
300 	} while (first <= last);
301 
302 	_leave(" = %d", ret);
303 	return ret;
304 }
305 
306 /*
307  * initiate a call
308  */
309 int afs_make_call(struct in_addr *addr, struct afs_call *call, gfp_t gfp,
310 		  const struct afs_wait_mode *wait_mode)
311 {
312 	struct sockaddr_rxrpc srx;
313 	struct rxrpc_call *rxcall;
314 	struct msghdr msg;
315 	struct kvec iov[1];
316 	int ret;
317 
318 	_enter("%x,{%d},", addr->s_addr, ntohs(call->port));
319 
320 	ASSERT(call->type != NULL);
321 	ASSERT(call->type->name != NULL);
322 
323 	_debug("____MAKE %p{%s,%x} [%d]____",
324 	       call, call->type->name, key_serial(call->key),
325 	       atomic_read(&afs_outstanding_calls));
326 
327 	call->wait_mode = wait_mode;
328 	INIT_WORK(&call->async_work, afs_process_async_call);
329 
330 	memset(&srx, 0, sizeof(srx));
331 	srx.srx_family = AF_RXRPC;
332 	srx.srx_service = call->service_id;
333 	srx.transport_type = SOCK_DGRAM;
334 	srx.transport_len = sizeof(srx.transport.sin);
335 	srx.transport.sin.sin_family = AF_INET;
336 	srx.transport.sin.sin_port = call->port;
337 	memcpy(&srx.transport.sin.sin_addr, addr, 4);
338 
339 	/* create a call */
340 	rxcall = rxrpc_kernel_begin_call(afs_socket, &srx, call->key,
341 					 (unsigned long) call, gfp);
342 	call->key = NULL;
343 	if (IS_ERR(rxcall)) {
344 		ret = PTR_ERR(rxcall);
345 		goto error_kill_call;
346 	}
347 
348 	call->rxcall = rxcall;
349 
350 	/* send the request */
351 	iov[0].iov_base	= call->request;
352 	iov[0].iov_len	= call->request_size;
353 
354 	msg.msg_name		= NULL;
355 	msg.msg_namelen		= 0;
356 	msg.msg_iov		= (struct iovec *) iov;
357 	msg.msg_iovlen		= 1;
358 	msg.msg_control		= NULL;
359 	msg.msg_controllen	= 0;
360 	msg.msg_flags		= (call->send_pages ? MSG_MORE : 0);
361 
362 	/* have to change the state *before* sending the last packet as RxRPC
363 	 * might give us the reply before it returns from sending the
364 	 * request */
365 	if (!call->send_pages)
366 		call->state = AFS_CALL_AWAIT_REPLY;
367 	ret = rxrpc_kernel_send_data(rxcall, &msg, call->request_size);
368 	if (ret < 0)
369 		goto error_do_abort;
370 
371 	if (call->send_pages) {
372 		ret = afs_send_pages(call, &msg, iov);
373 		if (ret < 0)
374 			goto error_do_abort;
375 	}
376 
377 	/* at this point, an async call may no longer exist as it may have
378 	 * already completed */
379 	return wait_mode->wait(call);
380 
381 error_do_abort:
382 	rxrpc_kernel_abort_call(rxcall, RX_USER_ABORT);
383 	rxrpc_kernel_end_call(rxcall);
384 	call->rxcall = NULL;
385 error_kill_call:
386 	call->type->destructor(call);
387 	afs_free_call(call);
388 	_leave(" = %d", ret);
389 	return ret;
390 }
391 
392 /*
393  * handles intercepted messages that were arriving in the socket's Rx queue
394  * - called with the socket receive queue lock held to ensure message ordering
395  * - called with softirqs disabled
396  */
397 static void afs_rx_interceptor(struct sock *sk, unsigned long user_call_ID,
398 			       struct sk_buff *skb)
399 {
400 	struct afs_call *call = (struct afs_call *) user_call_ID;
401 
402 	_enter("%p,,%u", call, skb->mark);
403 
404 	_debug("ICPT %p{%u} [%d]",
405 	       skb, skb->mark, atomic_read(&afs_outstanding_skbs));
406 
407 	ASSERTCMP(sk, ==, afs_socket->sk);
408 	atomic_inc(&afs_outstanding_skbs);
409 
410 	if (!call) {
411 		/* its an incoming call for our callback service */
412 		skb_queue_tail(&afs_incoming_calls, skb);
413 		schedule_work(&afs_collect_incoming_call_work);
414 	} else {
415 		/* route the messages directly to the appropriate call */
416 		skb_queue_tail(&call->rx_queue, skb);
417 		call->wait_mode->rx_wakeup(call);
418 	}
419 
420 	_leave("");
421 }
422 
423 /*
424  * deliver messages to a call
425  */
426 static void afs_deliver_to_call(struct afs_call *call)
427 {
428 	struct sk_buff *skb;
429 	bool last;
430 	u32 abort_code;
431 	int ret;
432 
433 	_enter("");
434 
435 	while ((call->state == AFS_CALL_AWAIT_REPLY ||
436 		call->state == AFS_CALL_AWAIT_OP_ID ||
437 		call->state == AFS_CALL_AWAIT_REQUEST ||
438 		call->state == AFS_CALL_AWAIT_ACK) &&
439 	       (skb = skb_dequeue(&call->rx_queue))) {
440 		switch (skb->mark) {
441 		case RXRPC_SKB_MARK_DATA:
442 			_debug("Rcv DATA");
443 			last = rxrpc_kernel_is_data_last(skb);
444 			ret = call->type->deliver(call, skb, last);
445 			switch (ret) {
446 			case 0:
447 				if (last &&
448 				    call->state == AFS_CALL_AWAIT_REPLY)
449 					call->state = AFS_CALL_COMPLETE;
450 				break;
451 			case -ENOTCONN:
452 				abort_code = RX_CALL_DEAD;
453 				goto do_abort;
454 			case -ENOTSUPP:
455 				abort_code = RX_INVALID_OPERATION;
456 				goto do_abort;
457 			default:
458 				abort_code = RXGEN_CC_UNMARSHAL;
459 				if (call->state != AFS_CALL_AWAIT_REPLY)
460 					abort_code = RXGEN_SS_UNMARSHAL;
461 			do_abort:
462 				rxrpc_kernel_abort_call(call->rxcall,
463 							abort_code);
464 				call->error = ret;
465 				call->state = AFS_CALL_ERROR;
466 				break;
467 			}
468 			afs_data_delivered(skb);
469 			skb = NULL;
470 			continue;
471 		case RXRPC_SKB_MARK_FINAL_ACK:
472 			_debug("Rcv ACK");
473 			call->state = AFS_CALL_COMPLETE;
474 			break;
475 		case RXRPC_SKB_MARK_BUSY:
476 			_debug("Rcv BUSY");
477 			call->error = -EBUSY;
478 			call->state = AFS_CALL_BUSY;
479 			break;
480 		case RXRPC_SKB_MARK_REMOTE_ABORT:
481 			abort_code = rxrpc_kernel_get_abort_code(skb);
482 			call->error = call->type->abort_to_error(abort_code);
483 			call->state = AFS_CALL_ABORTED;
484 			_debug("Rcv ABORT %u -> %d", abort_code, call->error);
485 			break;
486 		case RXRPC_SKB_MARK_NET_ERROR:
487 			call->error = -rxrpc_kernel_get_error_number(skb);
488 			call->state = AFS_CALL_ERROR;
489 			_debug("Rcv NET ERROR %d", call->error);
490 			break;
491 		case RXRPC_SKB_MARK_LOCAL_ERROR:
492 			call->error = -rxrpc_kernel_get_error_number(skb);
493 			call->state = AFS_CALL_ERROR;
494 			_debug("Rcv LOCAL ERROR %d", call->error);
495 			break;
496 		default:
497 			BUG();
498 			break;
499 		}
500 
501 		afs_free_skb(skb);
502 	}
503 
504 	/* make sure the queue is empty if the call is done with (we might have
505 	 * aborted the call early because of an unmarshalling error) */
506 	if (call->state >= AFS_CALL_COMPLETE) {
507 		while ((skb = skb_dequeue(&call->rx_queue)))
508 			afs_free_skb(skb);
509 		if (call->incoming) {
510 			rxrpc_kernel_end_call(call->rxcall);
511 			call->rxcall = NULL;
512 			call->type->destructor(call);
513 			afs_free_call(call);
514 		}
515 	}
516 
517 	_leave("");
518 }
519 
520 /*
521  * wait synchronously for a call to complete
522  */
523 static int afs_wait_for_call_to_complete(struct afs_call *call)
524 {
525 	struct sk_buff *skb;
526 	int ret;
527 
528 	DECLARE_WAITQUEUE(myself, current);
529 
530 	_enter("");
531 
532 	add_wait_queue(&call->waitq, &myself);
533 	for (;;) {
534 		set_current_state(TASK_INTERRUPTIBLE);
535 
536 		/* deliver any messages that are in the queue */
537 		if (!skb_queue_empty(&call->rx_queue)) {
538 			__set_current_state(TASK_RUNNING);
539 			afs_deliver_to_call(call);
540 			continue;
541 		}
542 
543 		ret = call->error;
544 		if (call->state >= AFS_CALL_COMPLETE)
545 			break;
546 		ret = -EINTR;
547 		if (signal_pending(current))
548 			break;
549 		schedule();
550 	}
551 
552 	remove_wait_queue(&call->waitq, &myself);
553 	__set_current_state(TASK_RUNNING);
554 
555 	/* kill the call */
556 	if (call->state < AFS_CALL_COMPLETE) {
557 		_debug("call incomplete");
558 		rxrpc_kernel_abort_call(call->rxcall, RX_CALL_DEAD);
559 		while ((skb = skb_dequeue(&call->rx_queue)))
560 			afs_free_skb(skb);
561 	}
562 
563 	_debug("call complete");
564 	rxrpc_kernel_end_call(call->rxcall);
565 	call->rxcall = NULL;
566 	call->type->destructor(call);
567 	afs_free_call(call);
568 	_leave(" = %d", ret);
569 	return ret;
570 }
571 
572 /*
573  * wake up a waiting call
574  */
575 static void afs_wake_up_call_waiter(struct afs_call *call)
576 {
577 	wake_up(&call->waitq);
578 }
579 
580 /*
581  * wake up an asynchronous call
582  */
583 static void afs_wake_up_async_call(struct afs_call *call)
584 {
585 	_enter("");
586 	queue_work(afs_async_calls, &call->async_work);
587 }
588 
589 /*
590  * put a call into asynchronous mode
591  * - mustn't touch the call descriptor as the call my have completed by the
592  *   time we get here
593  */
594 static int afs_dont_wait_for_call_to_complete(struct afs_call *call)
595 {
596 	_enter("");
597 	return -EINPROGRESS;
598 }
599 
600 /*
601  * delete an asynchronous call
602  */
603 static void afs_delete_async_call(struct work_struct *work)
604 {
605 	struct afs_call *call =
606 		container_of(work, struct afs_call, async_work);
607 
608 	_enter("");
609 
610 	afs_free_call(call);
611 
612 	_leave("");
613 }
614 
615 /*
616  * perform processing on an asynchronous call
617  * - on a multiple-thread workqueue this work item may try to run on several
618  *   CPUs at the same time
619  */
620 static void afs_process_async_call(struct work_struct *work)
621 {
622 	struct afs_call *call =
623 		container_of(work, struct afs_call, async_work);
624 
625 	_enter("");
626 
627 	if (!skb_queue_empty(&call->rx_queue))
628 		afs_deliver_to_call(call);
629 
630 	if (call->state >= AFS_CALL_COMPLETE && call->wait_mode) {
631 		if (call->wait_mode->async_complete)
632 			call->wait_mode->async_complete(call->reply,
633 							call->error);
634 		call->reply = NULL;
635 
636 		/* kill the call */
637 		rxrpc_kernel_end_call(call->rxcall);
638 		call->rxcall = NULL;
639 		if (call->type->destructor)
640 			call->type->destructor(call);
641 
642 		/* we can't just delete the call because the work item may be
643 		 * queued */
644 		PREPARE_WORK(&call->async_work, afs_delete_async_call);
645 		queue_work(afs_async_calls, &call->async_work);
646 	}
647 
648 	_leave("");
649 }
650 
651 /*
652  * empty a socket buffer into a flat reply buffer
653  */
654 void afs_transfer_reply(struct afs_call *call, struct sk_buff *skb)
655 {
656 	size_t len = skb->len;
657 
658 	if (skb_copy_bits(skb, 0, call->buffer + call->reply_size, len) < 0)
659 		BUG();
660 	call->reply_size += len;
661 }
662 
663 /*
664  * accept the backlog of incoming calls
665  */
666 static void afs_collect_incoming_call(struct work_struct *work)
667 {
668 	struct rxrpc_call *rxcall;
669 	struct afs_call *call = NULL;
670 	struct sk_buff *skb;
671 
672 	while ((skb = skb_dequeue(&afs_incoming_calls))) {
673 		_debug("new call");
674 
675 		/* don't need the notification */
676 		afs_free_skb(skb);
677 
678 		if (!call) {
679 			call = kzalloc(sizeof(struct afs_call), GFP_KERNEL);
680 			if (!call) {
681 				rxrpc_kernel_reject_call(afs_socket);
682 				return;
683 			}
684 
685 			INIT_WORK(&call->async_work, afs_process_async_call);
686 			call->wait_mode = &afs_async_incoming_call;
687 			call->type = &afs_RXCMxxxx;
688 			init_waitqueue_head(&call->waitq);
689 			skb_queue_head_init(&call->rx_queue);
690 			call->state = AFS_CALL_AWAIT_OP_ID;
691 
692 			_debug("CALL %p{%s} [%d]",
693 			       call, call->type->name,
694 			       atomic_read(&afs_outstanding_calls));
695 			atomic_inc(&afs_outstanding_calls);
696 		}
697 
698 		rxcall = rxrpc_kernel_accept_call(afs_socket,
699 						  (unsigned long) call);
700 		if (!IS_ERR(rxcall)) {
701 			call->rxcall = rxcall;
702 			call = NULL;
703 		}
704 	}
705 
706 	if (call)
707 		afs_free_call(call);
708 }
709 
710 /*
711  * grab the operation ID from an incoming cache manager call
712  */
713 static int afs_deliver_cm_op_id(struct afs_call *call, struct sk_buff *skb,
714 				bool last)
715 {
716 	size_t len = skb->len;
717 	void *oibuf = (void *) &call->operation_ID;
718 
719 	_enter("{%u},{%zu},%d", call->offset, len, last);
720 
721 	ASSERTCMP(call->offset, <, 4);
722 
723 	/* the operation ID forms the first four bytes of the request data */
724 	len = min_t(size_t, len, 4 - call->offset);
725 	if (skb_copy_bits(skb, 0, oibuf + call->offset, len) < 0)
726 		BUG();
727 	if (!pskb_pull(skb, len))
728 		BUG();
729 	call->offset += len;
730 
731 	if (call->offset < 4) {
732 		if (last) {
733 			_leave(" = -EBADMSG [op ID short]");
734 			return -EBADMSG;
735 		}
736 		_leave(" = 0 [incomplete]");
737 		return 0;
738 	}
739 
740 	call->state = AFS_CALL_AWAIT_REQUEST;
741 
742 	/* ask the cache manager to route the call (it'll change the call type
743 	 * if successful) */
744 	if (!afs_cm_incoming_call(call))
745 		return -ENOTSUPP;
746 
747 	/* pass responsibility for the remainer of this message off to the
748 	 * cache manager op */
749 	return call->type->deliver(call, skb, last);
750 }
751 
752 /*
753  * send an empty reply
754  */
755 void afs_send_empty_reply(struct afs_call *call)
756 {
757 	struct msghdr msg;
758 	struct iovec iov[1];
759 
760 	_enter("");
761 
762 	iov[0].iov_base		= NULL;
763 	iov[0].iov_len		= 0;
764 	msg.msg_name		= NULL;
765 	msg.msg_namelen		= 0;
766 	msg.msg_iov		= iov;
767 	msg.msg_iovlen		= 0;
768 	msg.msg_control		= NULL;
769 	msg.msg_controllen	= 0;
770 	msg.msg_flags		= 0;
771 
772 	call->state = AFS_CALL_AWAIT_ACK;
773 	switch (rxrpc_kernel_send_data(call->rxcall, &msg, 0)) {
774 	case 0:
775 		_leave(" [replied]");
776 		return;
777 
778 	case -ENOMEM:
779 		_debug("oom");
780 		rxrpc_kernel_abort_call(call->rxcall, RX_USER_ABORT);
781 	default:
782 		rxrpc_kernel_end_call(call->rxcall);
783 		call->rxcall = NULL;
784 		call->type->destructor(call);
785 		afs_free_call(call);
786 		_leave(" [error]");
787 		return;
788 	}
789 }
790 
791 /*
792  * send a simple reply
793  */
794 void afs_send_simple_reply(struct afs_call *call, const void *buf, size_t len)
795 {
796 	struct msghdr msg;
797 	struct iovec iov[1];
798 	int n;
799 
800 	_enter("");
801 
802 	iov[0].iov_base		= (void *) buf;
803 	iov[0].iov_len		= len;
804 	msg.msg_name		= NULL;
805 	msg.msg_namelen		= 0;
806 	msg.msg_iov		= iov;
807 	msg.msg_iovlen		= 1;
808 	msg.msg_control		= NULL;
809 	msg.msg_controllen	= 0;
810 	msg.msg_flags		= 0;
811 
812 	call->state = AFS_CALL_AWAIT_ACK;
813 	n = rxrpc_kernel_send_data(call->rxcall, &msg, len);
814 	if (n >= 0) {
815 		_leave(" [replied]");
816 		return;
817 	}
818 	if (n == -ENOMEM) {
819 		_debug("oom");
820 		rxrpc_kernel_abort_call(call->rxcall, RX_USER_ABORT);
821 	}
822 	rxrpc_kernel_end_call(call->rxcall);
823 	call->rxcall = NULL;
824 	call->type->destructor(call);
825 	afs_free_call(call);
826 	_leave(" [error]");
827 }
828 
829 /*
830  * extract a piece of data from the received data socket buffers
831  */
832 int afs_extract_data(struct afs_call *call, struct sk_buff *skb,
833 		     bool last, void *buf, size_t count)
834 {
835 	size_t len = skb->len;
836 
837 	_enter("{%u},{%zu},%d,,%zu", call->offset, len, last, count);
838 
839 	ASSERTCMP(call->offset, <, count);
840 
841 	len = min_t(size_t, len, count - call->offset);
842 	if (skb_copy_bits(skb, 0, buf + call->offset, len) < 0 ||
843 	    !pskb_pull(skb, len))
844 		BUG();
845 	call->offset += len;
846 
847 	if (call->offset < count) {
848 		if (last) {
849 			_leave(" = -EBADMSG [%d < %zu]", call->offset, count);
850 			return -EBADMSG;
851 		}
852 		_leave(" = -EAGAIN");
853 		return -EAGAIN;
854 	}
855 	return 0;
856 }
857