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
19static struct socket *afs_socket; /* my RxRPC socket */
20static struct workqueue_struct *afs_async_calls;
21static atomic_t afs_outstanding_calls;
22static atomic_t afs_outstanding_skbs;
23
24static void afs_wake_up_call_waiter(struct afs_call *);
25static int afs_wait_for_call_to_complete(struct afs_call *);
26static void afs_wake_up_async_call(struct afs_call *);
27static int afs_dont_wait_for_call_to_complete(struct afs_call *);
28static void afs_process_async_call(struct afs_call *);
29static void afs_rx_interceptor(struct sock *, unsigned long, struct sk_buff *);
30static int afs_deliver_cm_op_id(struct afs_call *, struct sk_buff *, bool);
31
32/* synchronous call management */
33const 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 */
39const 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 */
45static 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 */
50static 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
56static void afs_collect_incoming_call(struct work_struct *);
57
58static struct sk_buff_head afs_incoming_calls;
59static DECLARE_WORK(afs_collect_incoming_call_work, afs_collect_incoming_call);
60
61static 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 */
72int 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(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 */
125void 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 */
143static 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 */
160static 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 */
177static 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 */
196static 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 */
209static 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 */
218struct 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
251nomem_free:
252	afs_free_call(call);
253nomem_call:
254	return NULL;
255}
256
257/*
258 * clean up a call with flat buffer
259 */
260void 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 */
273static 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			msg->msg_iov = (struct iovec *) iov;
310			msg->msg_iovlen = 1;
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 */
338int 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	msg.msg_iov		= (struct iovec *) iov;
388	msg.msg_iovlen		= 1;
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
412error_do_abort:
413	rxrpc_kernel_abort_call(rxcall, RX_USER_ABORT);
414	while ((skb = skb_dequeue(&call->rx_queue)))
415		afs_free_skb(skb);
416error_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 */
427static 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 */
456static 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 */
549static 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 */
598static 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 */
606static 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 */
617static 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 */
626static 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 */
640static 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 */
668void 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 */
680static 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 */
728static 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 */
770void afs_send_empty_reply(struct afs_call *call)
771{
772	struct msghdr msg;
773	struct iovec iov[1];
774
775	_enter("");
776
777	iov[0].iov_base		= NULL;
778	iov[0].iov_len		= 0;
779	msg.msg_name		= NULL;
780	msg.msg_namelen		= 0;
781	msg.msg_iov		= iov;
782	msg.msg_iovlen		= 0;
783	msg.msg_control		= NULL;
784	msg.msg_controllen	= 0;
785	msg.msg_flags		= 0;
786
787	call->state = AFS_CALL_AWAIT_ACK;
788	switch (rxrpc_kernel_send_data(call->rxcall, &msg, 0)) {
789	case 0:
790		_leave(" [replied]");
791		return;
792
793	case -ENOMEM:
794		_debug("oom");
795		rxrpc_kernel_abort_call(call->rxcall, RX_USER_ABORT);
796	default:
797		afs_end_call(call);
798		_leave(" [error]");
799		return;
800	}
801}
802
803/*
804 * send a simple reply
805 */
806void afs_send_simple_reply(struct afs_call *call, const void *buf, size_t len)
807{
808	struct msghdr msg;
809	struct iovec iov[1];
810	int n;
811
812	_enter("");
813
814	iov[0].iov_base		= (void *) buf;
815	iov[0].iov_len		= len;
816	msg.msg_name		= NULL;
817	msg.msg_namelen		= 0;
818	msg.msg_iov		= iov;
819	msg.msg_iovlen		= 1;
820	msg.msg_control		= NULL;
821	msg.msg_controllen	= 0;
822	msg.msg_flags		= 0;
823
824	call->state = AFS_CALL_AWAIT_ACK;
825	n = rxrpc_kernel_send_data(call->rxcall, &msg, len);
826	if (n >= 0) {
827		/* Success */
828		_leave(" [replied]");
829		return;
830	}
831
832	if (n == -ENOMEM) {
833		_debug("oom");
834		rxrpc_kernel_abort_call(call->rxcall, RX_USER_ABORT);
835	}
836	afs_end_call(call);
837	_leave(" [error]");
838}
839
840/*
841 * extract a piece of data from the received data socket buffers
842 */
843int afs_extract_data(struct afs_call *call, struct sk_buff *skb,
844		     bool last, void *buf, size_t count)
845{
846	size_t len = skb->len;
847
848	_enter("{%u},{%zu},%d,,%zu", call->offset, len, last, count);
849
850	ASSERTCMP(call->offset, <, count);
851
852	len = min_t(size_t, len, count - call->offset);
853	if (skb_copy_bits(skb, 0, buf + call->offset, len) < 0 ||
854	    !pskb_pull(skb, len))
855		BUG();
856	call->offset += len;
857
858	if (call->offset < count) {
859		if (last) {
860			_leave(" = -EBADMSG [%d < %zu]", call->offset, count);
861			return -EBADMSG;
862		}
863		_leave(" = -EAGAIN");
864		return -EAGAIN;
865	}
866	return 0;
867}
868