xref: /openbmc/linux/fs/afs/rxrpc.c (revision 92a76f6d)
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 afs_call *);
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 static void afs_async_workfn(struct work_struct *work)
62 {
63 	struct afs_call *call = container_of(work, struct afs_call, async_work);
64 
65 	call->async_workfn(call);
66 }
67 
68 /*
69  * open an RxRPC socket and bind it to be a server for callback notifications
70  * - the socket is left in blocking mode and non-blocking ops use MSG_DONTWAIT
71  */
72 int afs_open_socket(void)
73 {
74 	struct sockaddr_rxrpc srx;
75 	struct socket *socket;
76 	int ret;
77 
78 	_enter("");
79 
80 	skb_queue_head_init(&afs_incoming_calls);
81 
82 	afs_async_calls = create_singlethread_workqueue("kafsd");
83 	if (!afs_async_calls) {
84 		_leave(" = -ENOMEM [wq]");
85 		return -ENOMEM;
86 	}
87 
88 	ret = sock_create_kern(&init_net, AF_RXRPC, SOCK_DGRAM, PF_INET, &socket);
89 	if (ret < 0) {
90 		destroy_workqueue(afs_async_calls);
91 		_leave(" = %d [socket]", ret);
92 		return ret;
93 	}
94 
95 	socket->sk->sk_allocation = GFP_NOFS;
96 
97 	/* bind the callback manager's address to make this a server socket */
98 	srx.srx_family			= AF_RXRPC;
99 	srx.srx_service			= CM_SERVICE;
100 	srx.transport_type		= SOCK_DGRAM;
101 	srx.transport_len		= sizeof(srx.transport.sin);
102 	srx.transport.sin.sin_family	= AF_INET;
103 	srx.transport.sin.sin_port	= htons(AFS_CM_PORT);
104 	memset(&srx.transport.sin.sin_addr, 0,
105 	       sizeof(srx.transport.sin.sin_addr));
106 
107 	ret = kernel_bind(socket, (struct sockaddr *) &srx, sizeof(srx));
108 	if (ret < 0) {
109 		sock_release(socket);
110 		destroy_workqueue(afs_async_calls);
111 		_leave(" = %d [bind]", ret);
112 		return ret;
113 	}
114 
115 	rxrpc_kernel_intercept_rx_messages(socket, afs_rx_interceptor);
116 
117 	afs_socket = socket;
118 	_leave(" = 0");
119 	return 0;
120 }
121 
122 /*
123  * close the RxRPC socket AFS was using
124  */
125 void afs_close_socket(void)
126 {
127 	_enter("");
128 
129 	sock_release(afs_socket);
130 
131 	_debug("dework");
132 	destroy_workqueue(afs_async_calls);
133 
134 	ASSERTCMP(atomic_read(&afs_outstanding_skbs), ==, 0);
135 	ASSERTCMP(atomic_read(&afs_outstanding_calls), ==, 0);
136 	_leave("");
137 }
138 
139 /*
140  * note that the data in a socket buffer is now delivered and that the buffer
141  * should be freed
142  */
143 static void afs_data_delivered(struct sk_buff *skb)
144 {
145 	if (!skb) {
146 		_debug("DLVR NULL [%d]", atomic_read(&afs_outstanding_skbs));
147 		dump_stack();
148 	} else {
149 		_debug("DLVR %p{%u} [%d]",
150 		       skb, skb->mark, atomic_read(&afs_outstanding_skbs));
151 		if (atomic_dec_return(&afs_outstanding_skbs) == -1)
152 			BUG();
153 		rxrpc_kernel_data_delivered(skb);
154 	}
155 }
156 
157 /*
158  * free a socket buffer
159  */
160 static void afs_free_skb(struct sk_buff *skb)
161 {
162 	if (!skb) {
163 		_debug("FREE NULL [%d]", atomic_read(&afs_outstanding_skbs));
164 		dump_stack();
165 	} else {
166 		_debug("FREE %p{%u} [%d]",
167 		       skb, skb->mark, atomic_read(&afs_outstanding_skbs));
168 		if (atomic_dec_return(&afs_outstanding_skbs) == -1)
169 			BUG();
170 		rxrpc_kernel_free_skb(skb);
171 	}
172 }
173 
174 /*
175  * free a call
176  */
177 static void afs_free_call(struct afs_call *call)
178 {
179 	_debug("DONE %p{%s} [%d]",
180 	       call, call->type->name, atomic_read(&afs_outstanding_calls));
181 	if (atomic_dec_return(&afs_outstanding_calls) == -1)
182 		BUG();
183 
184 	ASSERTCMP(call->rxcall, ==, NULL);
185 	ASSERT(!work_pending(&call->async_work));
186 	ASSERT(skb_queue_empty(&call->rx_queue));
187 	ASSERT(call->type->name != NULL);
188 
189 	kfree(call->request);
190 	kfree(call);
191 }
192 
193 /*
194  * End a call but do not free it
195  */
196 static void afs_end_call_nofree(struct afs_call *call)
197 {
198 	if (call->rxcall) {
199 		rxrpc_kernel_end_call(call->rxcall);
200 		call->rxcall = NULL;
201 	}
202 	if (call->type->destructor)
203 		call->type->destructor(call);
204 }
205 
206 /*
207  * End a call and free it
208  */
209 static void afs_end_call(struct afs_call *call)
210 {
211 	afs_end_call_nofree(call);
212 	afs_free_call(call);
213 }
214 
215 /*
216  * allocate a call with flat request and reply buffers
217  */
218 struct afs_call *afs_alloc_flat_call(const struct afs_call_type *type,
219 				     size_t request_size, size_t reply_size)
220 {
221 	struct afs_call *call;
222 
223 	call = kzalloc(sizeof(*call), GFP_NOFS);
224 	if (!call)
225 		goto nomem_call;
226 
227 	_debug("CALL %p{%s} [%d]",
228 	       call, type->name, atomic_read(&afs_outstanding_calls));
229 	atomic_inc(&afs_outstanding_calls);
230 
231 	call->type = type;
232 	call->request_size = request_size;
233 	call->reply_max = reply_size;
234 
235 	if (request_size) {
236 		call->request = kmalloc(request_size, GFP_NOFS);
237 		if (!call->request)
238 			goto nomem_free;
239 	}
240 
241 	if (reply_size) {
242 		call->buffer = kmalloc(reply_size, GFP_NOFS);
243 		if (!call->buffer)
244 			goto nomem_free;
245 	}
246 
247 	init_waitqueue_head(&call->waitq);
248 	skb_queue_head_init(&call->rx_queue);
249 	return call;
250 
251 nomem_free:
252 	afs_free_call(call);
253 nomem_call:
254 	return NULL;
255 }
256 
257 /*
258  * clean up a call with flat buffer
259  */
260 void afs_flat_call_destructor(struct afs_call *call)
261 {
262 	_enter("");
263 
264 	kfree(call->request);
265 	call->request = NULL;
266 	kfree(call->buffer);
267 	call->buffer = NULL;
268 }
269 
270 /*
271  * attach the data from a bunch of pages on an inode to a call
272  */
273 static int afs_send_pages(struct afs_call *call, struct msghdr *msg,
274 			  struct kvec *iov)
275 {
276 	struct page *pages[8];
277 	unsigned count, n, loop, offset, to;
278 	pgoff_t first = call->first, last = call->last;
279 	int ret;
280 
281 	_enter("");
282 
283 	offset = call->first_offset;
284 	call->first_offset = 0;
285 
286 	do {
287 		_debug("attach %lx-%lx", first, last);
288 
289 		count = last - first + 1;
290 		if (count > ARRAY_SIZE(pages))
291 			count = ARRAY_SIZE(pages);
292 		n = find_get_pages_contig(call->mapping, first, count, pages);
293 		ASSERTCMP(n, ==, count);
294 
295 		loop = 0;
296 		do {
297 			msg->msg_flags = 0;
298 			to = PAGE_SIZE;
299 			if (first + loop >= last)
300 				to = call->last_to;
301 			else
302 				msg->msg_flags = MSG_MORE;
303 			iov->iov_base = kmap(pages[loop]) + offset;
304 			iov->iov_len = to - offset;
305 			offset = 0;
306 
307 			_debug("- range %u-%u%s",
308 			       offset, to, msg->msg_flags ? " [more]" : "");
309 			iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC,
310 				      iov, 1, to - offset);
311 
312 			/* have to change the state *before* sending the last
313 			 * packet as RxRPC might give us the reply before it
314 			 * returns from sending the request */
315 			if (first + loop >= last)
316 				call->state = AFS_CALL_AWAIT_REPLY;
317 			ret = rxrpc_kernel_send_data(call->rxcall, msg,
318 						     to - offset);
319 			kunmap(pages[loop]);
320 			if (ret < 0)
321 				break;
322 		} while (++loop < count);
323 		first += count;
324 
325 		for (loop = 0; loop < count; loop++)
326 			put_page(pages[loop]);
327 		if (ret < 0)
328 			break;
329 	} while (first <= last);
330 
331 	_leave(" = %d", ret);
332 	return ret;
333 }
334 
335 /*
336  * initiate a call
337  */
338 int afs_make_call(struct in_addr *addr, struct afs_call *call, gfp_t gfp,
339 		  const struct afs_wait_mode *wait_mode)
340 {
341 	struct sockaddr_rxrpc srx;
342 	struct rxrpc_call *rxcall;
343 	struct msghdr msg;
344 	struct kvec iov[1];
345 	int ret;
346 	struct sk_buff *skb;
347 
348 	_enter("%x,{%d},", addr->s_addr, ntohs(call->port));
349 
350 	ASSERT(call->type != NULL);
351 	ASSERT(call->type->name != NULL);
352 
353 	_debug("____MAKE %p{%s,%x} [%d]____",
354 	       call, call->type->name, key_serial(call->key),
355 	       atomic_read(&afs_outstanding_calls));
356 
357 	call->wait_mode = wait_mode;
358 	call->async_workfn = afs_process_async_call;
359 	INIT_WORK(&call->async_work, afs_async_workfn);
360 
361 	memset(&srx, 0, sizeof(srx));
362 	srx.srx_family = AF_RXRPC;
363 	srx.srx_service = call->service_id;
364 	srx.transport_type = SOCK_DGRAM;
365 	srx.transport_len = sizeof(srx.transport.sin);
366 	srx.transport.sin.sin_family = AF_INET;
367 	srx.transport.sin.sin_port = call->port;
368 	memcpy(&srx.transport.sin.sin_addr, addr, 4);
369 
370 	/* create a call */
371 	rxcall = rxrpc_kernel_begin_call(afs_socket, &srx, call->key,
372 					 (unsigned long) call, gfp);
373 	call->key = NULL;
374 	if (IS_ERR(rxcall)) {
375 		ret = PTR_ERR(rxcall);
376 		goto error_kill_call;
377 	}
378 
379 	call->rxcall = rxcall;
380 
381 	/* send the request */
382 	iov[0].iov_base	= call->request;
383 	iov[0].iov_len	= call->request_size;
384 
385 	msg.msg_name		= NULL;
386 	msg.msg_namelen		= 0;
387 	iov_iter_kvec(&msg.msg_iter, WRITE | ITER_KVEC, iov, 1,
388 		      call->request_size);
389 	msg.msg_control		= NULL;
390 	msg.msg_controllen	= 0;
391 	msg.msg_flags		= (call->send_pages ? MSG_MORE : 0);
392 
393 	/* have to change the state *before* sending the last packet as RxRPC
394 	 * might give us the reply before it returns from sending the
395 	 * request */
396 	if (!call->send_pages)
397 		call->state = AFS_CALL_AWAIT_REPLY;
398 	ret = rxrpc_kernel_send_data(rxcall, &msg, call->request_size);
399 	if (ret < 0)
400 		goto error_do_abort;
401 
402 	if (call->send_pages) {
403 		ret = afs_send_pages(call, &msg, iov);
404 		if (ret < 0)
405 			goto error_do_abort;
406 	}
407 
408 	/* at this point, an async call may no longer exist as it may have
409 	 * already completed */
410 	return wait_mode->wait(call);
411 
412 error_do_abort:
413 	rxrpc_kernel_abort_call(rxcall, RX_USER_ABORT);
414 	while ((skb = skb_dequeue(&call->rx_queue)))
415 		afs_free_skb(skb);
416 error_kill_call:
417 	afs_end_call(call);
418 	_leave(" = %d", ret);
419 	return ret;
420 }
421 
422 /*
423  * handles intercepted messages that were arriving in the socket's Rx queue
424  * - called with the socket receive queue lock held to ensure message ordering
425  * - called with softirqs disabled
426  */
427 static void afs_rx_interceptor(struct sock *sk, unsigned long user_call_ID,
428 			       struct sk_buff *skb)
429 {
430 	struct afs_call *call = (struct afs_call *) user_call_ID;
431 
432 	_enter("%p,,%u", call, skb->mark);
433 
434 	_debug("ICPT %p{%u} [%d]",
435 	       skb, skb->mark, atomic_read(&afs_outstanding_skbs));
436 
437 	ASSERTCMP(sk, ==, afs_socket->sk);
438 	atomic_inc(&afs_outstanding_skbs);
439 
440 	if (!call) {
441 		/* its an incoming call for our callback service */
442 		skb_queue_tail(&afs_incoming_calls, skb);
443 		queue_work(afs_wq, &afs_collect_incoming_call_work);
444 	} else {
445 		/* route the messages directly to the appropriate call */
446 		skb_queue_tail(&call->rx_queue, skb);
447 		call->wait_mode->rx_wakeup(call);
448 	}
449 
450 	_leave("");
451 }
452 
453 /*
454  * deliver messages to a call
455  */
456 static void afs_deliver_to_call(struct afs_call *call)
457 {
458 	struct sk_buff *skb;
459 	bool last;
460 	u32 abort_code;
461 	int ret;
462 
463 	_enter("");
464 
465 	while ((call->state == AFS_CALL_AWAIT_REPLY ||
466 		call->state == AFS_CALL_AWAIT_OP_ID ||
467 		call->state == AFS_CALL_AWAIT_REQUEST ||
468 		call->state == AFS_CALL_AWAIT_ACK) &&
469 	       (skb = skb_dequeue(&call->rx_queue))) {
470 		switch (skb->mark) {
471 		case RXRPC_SKB_MARK_DATA:
472 			_debug("Rcv DATA");
473 			last = rxrpc_kernel_is_data_last(skb);
474 			ret = call->type->deliver(call, skb, last);
475 			switch (ret) {
476 			case 0:
477 				if (last &&
478 				    call->state == AFS_CALL_AWAIT_REPLY)
479 					call->state = AFS_CALL_COMPLETE;
480 				break;
481 			case -ENOTCONN:
482 				abort_code = RX_CALL_DEAD;
483 				goto do_abort;
484 			case -ENOTSUPP:
485 				abort_code = RX_INVALID_OPERATION;
486 				goto do_abort;
487 			default:
488 				abort_code = RXGEN_CC_UNMARSHAL;
489 				if (call->state != AFS_CALL_AWAIT_REPLY)
490 					abort_code = RXGEN_SS_UNMARSHAL;
491 			do_abort:
492 				rxrpc_kernel_abort_call(call->rxcall,
493 							abort_code);
494 				call->error = ret;
495 				call->state = AFS_CALL_ERROR;
496 				break;
497 			}
498 			afs_data_delivered(skb);
499 			skb = NULL;
500 			continue;
501 		case RXRPC_SKB_MARK_FINAL_ACK:
502 			_debug("Rcv ACK");
503 			call->state = AFS_CALL_COMPLETE;
504 			break;
505 		case RXRPC_SKB_MARK_BUSY:
506 			_debug("Rcv BUSY");
507 			call->error = -EBUSY;
508 			call->state = AFS_CALL_BUSY;
509 			break;
510 		case RXRPC_SKB_MARK_REMOTE_ABORT:
511 			abort_code = rxrpc_kernel_get_abort_code(skb);
512 			call->error = call->type->abort_to_error(abort_code);
513 			call->state = AFS_CALL_ABORTED;
514 			_debug("Rcv ABORT %u -> %d", abort_code, call->error);
515 			break;
516 		case RXRPC_SKB_MARK_NET_ERROR:
517 			call->error = -rxrpc_kernel_get_error_number(skb);
518 			call->state = AFS_CALL_ERROR;
519 			_debug("Rcv NET ERROR %d", call->error);
520 			break;
521 		case RXRPC_SKB_MARK_LOCAL_ERROR:
522 			call->error = -rxrpc_kernel_get_error_number(skb);
523 			call->state = AFS_CALL_ERROR;
524 			_debug("Rcv LOCAL ERROR %d", call->error);
525 			break;
526 		default:
527 			BUG();
528 			break;
529 		}
530 
531 		afs_free_skb(skb);
532 	}
533 
534 	/* make sure the queue is empty if the call is done with (we might have
535 	 * aborted the call early because of an unmarshalling error) */
536 	if (call->state >= AFS_CALL_COMPLETE) {
537 		while ((skb = skb_dequeue(&call->rx_queue)))
538 			afs_free_skb(skb);
539 		if (call->incoming)
540 			afs_end_call(call);
541 	}
542 
543 	_leave("");
544 }
545 
546 /*
547  * wait synchronously for a call to complete
548  */
549 static int afs_wait_for_call_to_complete(struct afs_call *call)
550 {
551 	struct sk_buff *skb;
552 	int ret;
553 
554 	DECLARE_WAITQUEUE(myself, current);
555 
556 	_enter("");
557 
558 	add_wait_queue(&call->waitq, &myself);
559 	for (;;) {
560 		set_current_state(TASK_INTERRUPTIBLE);
561 
562 		/* deliver any messages that are in the queue */
563 		if (!skb_queue_empty(&call->rx_queue)) {
564 			__set_current_state(TASK_RUNNING);
565 			afs_deliver_to_call(call);
566 			continue;
567 		}
568 
569 		ret = call->error;
570 		if (call->state >= AFS_CALL_COMPLETE)
571 			break;
572 		ret = -EINTR;
573 		if (signal_pending(current))
574 			break;
575 		schedule();
576 	}
577 
578 	remove_wait_queue(&call->waitq, &myself);
579 	__set_current_state(TASK_RUNNING);
580 
581 	/* kill the call */
582 	if (call->state < AFS_CALL_COMPLETE) {
583 		_debug("call incomplete");
584 		rxrpc_kernel_abort_call(call->rxcall, RX_CALL_DEAD);
585 		while ((skb = skb_dequeue(&call->rx_queue)))
586 			afs_free_skb(skb);
587 	}
588 
589 	_debug("call complete");
590 	afs_end_call(call);
591 	_leave(" = %d", ret);
592 	return ret;
593 }
594 
595 /*
596  * wake up a waiting call
597  */
598 static void afs_wake_up_call_waiter(struct afs_call *call)
599 {
600 	wake_up(&call->waitq);
601 }
602 
603 /*
604  * wake up an asynchronous call
605  */
606 static void afs_wake_up_async_call(struct afs_call *call)
607 {
608 	_enter("");
609 	queue_work(afs_async_calls, &call->async_work);
610 }
611 
612 /*
613  * put a call into asynchronous mode
614  * - mustn't touch the call descriptor as the call my have completed by the
615  *   time we get here
616  */
617 static int afs_dont_wait_for_call_to_complete(struct afs_call *call)
618 {
619 	_enter("");
620 	return -EINPROGRESS;
621 }
622 
623 /*
624  * delete an asynchronous call
625  */
626 static void afs_delete_async_call(struct afs_call *call)
627 {
628 	_enter("");
629 
630 	afs_free_call(call);
631 
632 	_leave("");
633 }
634 
635 /*
636  * perform processing on an asynchronous call
637  * - on a multiple-thread workqueue this work item may try to run on several
638  *   CPUs at the same time
639  */
640 static void afs_process_async_call(struct afs_call *call)
641 {
642 	_enter("");
643 
644 	if (!skb_queue_empty(&call->rx_queue))
645 		afs_deliver_to_call(call);
646 
647 	if (call->state >= AFS_CALL_COMPLETE && call->wait_mode) {
648 		if (call->wait_mode->async_complete)
649 			call->wait_mode->async_complete(call->reply,
650 							call->error);
651 		call->reply = NULL;
652 
653 		/* kill the call */
654 		afs_end_call_nofree(call);
655 
656 		/* we can't just delete the call because the work item may be
657 		 * queued */
658 		call->async_workfn = afs_delete_async_call;
659 		queue_work(afs_async_calls, &call->async_work);
660 	}
661 
662 	_leave("");
663 }
664 
665 /*
666  * empty a socket buffer into a flat reply buffer
667  */
668 void afs_transfer_reply(struct afs_call *call, struct sk_buff *skb)
669 {
670 	size_t len = skb->len;
671 
672 	if (skb_copy_bits(skb, 0, call->buffer + call->reply_size, len) < 0)
673 		BUG();
674 	call->reply_size += len;
675 }
676 
677 /*
678  * accept the backlog of incoming calls
679  */
680 static void afs_collect_incoming_call(struct work_struct *work)
681 {
682 	struct rxrpc_call *rxcall;
683 	struct afs_call *call = NULL;
684 	struct sk_buff *skb;
685 
686 	while ((skb = skb_dequeue(&afs_incoming_calls))) {
687 		_debug("new call");
688 
689 		/* don't need the notification */
690 		afs_free_skb(skb);
691 
692 		if (!call) {
693 			call = kzalloc(sizeof(struct afs_call), GFP_KERNEL);
694 			if (!call) {
695 				rxrpc_kernel_reject_call(afs_socket);
696 				return;
697 			}
698 
699 			call->async_workfn = afs_process_async_call;
700 			INIT_WORK(&call->async_work, afs_async_workfn);
701 			call->wait_mode = &afs_async_incoming_call;
702 			call->type = &afs_RXCMxxxx;
703 			init_waitqueue_head(&call->waitq);
704 			skb_queue_head_init(&call->rx_queue);
705 			call->state = AFS_CALL_AWAIT_OP_ID;
706 
707 			_debug("CALL %p{%s} [%d]",
708 			       call, call->type->name,
709 			       atomic_read(&afs_outstanding_calls));
710 			atomic_inc(&afs_outstanding_calls);
711 		}
712 
713 		rxcall = rxrpc_kernel_accept_call(afs_socket,
714 						  (unsigned long) call);
715 		if (!IS_ERR(rxcall)) {
716 			call->rxcall = rxcall;
717 			call = NULL;
718 		}
719 	}
720 
721 	if (call)
722 		afs_free_call(call);
723 }
724 
725 /*
726  * grab the operation ID from an incoming cache manager call
727  */
728 static int afs_deliver_cm_op_id(struct afs_call *call, struct sk_buff *skb,
729 				bool last)
730 {
731 	size_t len = skb->len;
732 	void *oibuf = (void *) &call->operation_ID;
733 
734 	_enter("{%u},{%zu},%d", call->offset, len, last);
735 
736 	ASSERTCMP(call->offset, <, 4);
737 
738 	/* the operation ID forms the first four bytes of the request data */
739 	len = min_t(size_t, len, 4 - call->offset);
740 	if (skb_copy_bits(skb, 0, oibuf + call->offset, len) < 0)
741 		BUG();
742 	if (!pskb_pull(skb, len))
743 		BUG();
744 	call->offset += len;
745 
746 	if (call->offset < 4) {
747 		if (last) {
748 			_leave(" = -EBADMSG [op ID short]");
749 			return -EBADMSG;
750 		}
751 		_leave(" = 0 [incomplete]");
752 		return 0;
753 	}
754 
755 	call->state = AFS_CALL_AWAIT_REQUEST;
756 
757 	/* ask the cache manager to route the call (it'll change the call type
758 	 * if successful) */
759 	if (!afs_cm_incoming_call(call))
760 		return -ENOTSUPP;
761 
762 	/* pass responsibility for the remainer of this message off to the
763 	 * cache manager op */
764 	return call->type->deliver(call, skb, last);
765 }
766 
767 /*
768  * send an empty reply
769  */
770 void afs_send_empty_reply(struct afs_call *call)
771 {
772 	struct msghdr msg;
773 
774 	_enter("");
775 
776 	msg.msg_name		= NULL;
777 	msg.msg_namelen		= 0;
778 	iov_iter_kvec(&msg.msg_iter, WRITE | ITER_KVEC, NULL, 0, 0);
779 	msg.msg_control		= NULL;
780 	msg.msg_controllen	= 0;
781 	msg.msg_flags		= 0;
782 
783 	call->state = AFS_CALL_AWAIT_ACK;
784 	switch (rxrpc_kernel_send_data(call->rxcall, &msg, 0)) {
785 	case 0:
786 		_leave(" [replied]");
787 		return;
788 
789 	case -ENOMEM:
790 		_debug("oom");
791 		rxrpc_kernel_abort_call(call->rxcall, RX_USER_ABORT);
792 	default:
793 		afs_end_call(call);
794 		_leave(" [error]");
795 		return;
796 	}
797 }
798 
799 /*
800  * send a simple reply
801  */
802 void afs_send_simple_reply(struct afs_call *call, const void *buf, size_t len)
803 {
804 	struct msghdr msg;
805 	struct kvec iov[1];
806 	int n;
807 
808 	_enter("");
809 
810 	iov[0].iov_base		= (void *) buf;
811 	iov[0].iov_len		= len;
812 	msg.msg_name		= NULL;
813 	msg.msg_namelen		= 0;
814 	iov_iter_kvec(&msg.msg_iter, WRITE | ITER_KVEC, iov, 1, len);
815 	msg.msg_control		= NULL;
816 	msg.msg_controllen	= 0;
817 	msg.msg_flags		= 0;
818 
819 	call->state = AFS_CALL_AWAIT_ACK;
820 	n = rxrpc_kernel_send_data(call->rxcall, &msg, len);
821 	if (n >= 0) {
822 		/* Success */
823 		_leave(" [replied]");
824 		return;
825 	}
826 
827 	if (n == -ENOMEM) {
828 		_debug("oom");
829 		rxrpc_kernel_abort_call(call->rxcall, RX_USER_ABORT);
830 	}
831 	afs_end_call(call);
832 	_leave(" [error]");
833 }
834 
835 /*
836  * extract a piece of data from the received data socket buffers
837  */
838 int afs_extract_data(struct afs_call *call, struct sk_buff *skb,
839 		     bool last, void *buf, size_t count)
840 {
841 	size_t len = skb->len;
842 
843 	_enter("{%u},{%zu},%d,,%zu", call->offset, len, last, count);
844 
845 	ASSERTCMP(call->offset, <, count);
846 
847 	len = min_t(size_t, len, count - call->offset);
848 	if (skb_copy_bits(skb, 0, buf + call->offset, len) < 0 ||
849 	    !pskb_pull(skb, len))
850 		BUG();
851 	call->offset += len;
852 
853 	if (call->offset < count) {
854 		if (last) {
855 			_leave(" = -EBADMSG [%d < %zu]", call->offset, count);
856 			return -EBADMSG;
857 		}
858 		_leave(" = -EAGAIN");
859 		return -EAGAIN;
860 	}
861 	return 0;
862 }
863