]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blob - fs/afs/rxrpc.c
ASoC: sti: fix missing clk_disable_unprepare() on error in uni_player_start()
[mirror_ubuntu-zesty-kernel.git] / fs / afs / rxrpc.c
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 static int afs_wait_atomic_t(atomic_t *p)
69 {
70 schedule();
71 return 0;
72 }
73
74 /*
75 * open an RxRPC socket and bind it to be a server for callback notifications
76 * - the socket is left in blocking mode and non-blocking ops use MSG_DONTWAIT
77 */
78 int afs_open_socket(void)
79 {
80 struct sockaddr_rxrpc srx;
81 struct socket *socket;
82 int ret;
83
84 _enter("");
85
86 skb_queue_head_init(&afs_incoming_calls);
87
88 ret = -ENOMEM;
89 afs_async_calls = create_singlethread_workqueue("kafsd");
90 if (!afs_async_calls)
91 goto error_0;
92
93 ret = sock_create_kern(&init_net, AF_RXRPC, SOCK_DGRAM, PF_INET, &socket);
94 if (ret < 0)
95 goto error_1;
96
97 socket->sk->sk_allocation = GFP_NOFS;
98
99 /* bind the callback manager's address to make this a server socket */
100 srx.srx_family = AF_RXRPC;
101 srx.srx_service = CM_SERVICE;
102 srx.transport_type = SOCK_DGRAM;
103 srx.transport_len = sizeof(srx.transport.sin);
104 srx.transport.sin.sin_family = AF_INET;
105 srx.transport.sin.sin_port = htons(AFS_CM_PORT);
106 memset(&srx.transport.sin.sin_addr, 0,
107 sizeof(srx.transport.sin.sin_addr));
108
109 ret = kernel_bind(socket, (struct sockaddr *) &srx, sizeof(srx));
110 if (ret < 0)
111 goto error_2;
112
113 ret = kernel_listen(socket, INT_MAX);
114 if (ret < 0)
115 goto error_2;
116
117 rxrpc_kernel_intercept_rx_messages(socket, afs_rx_interceptor);
118
119 afs_socket = socket;
120 _leave(" = 0");
121 return 0;
122
123 error_2:
124 sock_release(socket);
125 error_1:
126 destroy_workqueue(afs_async_calls);
127 error_0:
128 _leave(" = %d", ret);
129 return ret;
130 }
131
132 /*
133 * close the RxRPC socket AFS was using
134 */
135 void afs_close_socket(void)
136 {
137 _enter("");
138
139 wait_on_atomic_t(&afs_outstanding_calls, afs_wait_atomic_t,
140 TASK_UNINTERRUPTIBLE);
141 _debug("no outstanding calls");
142
143 sock_release(afs_socket);
144
145 _debug("dework");
146 destroy_workqueue(afs_async_calls);
147
148 ASSERTCMP(atomic_read(&afs_outstanding_skbs), ==, 0);
149 _leave("");
150 }
151
152 /*
153 * note that the data in a socket buffer is now delivered and that the buffer
154 * should be freed
155 */
156 static void afs_data_delivered(struct sk_buff *skb)
157 {
158 if (!skb) {
159 _debug("DLVR NULL [%d]", atomic_read(&afs_outstanding_skbs));
160 dump_stack();
161 } else {
162 _debug("DLVR %p{%u} [%d]",
163 skb, skb->mark, atomic_read(&afs_outstanding_skbs));
164 if (atomic_dec_return(&afs_outstanding_skbs) == -1)
165 BUG();
166 rxrpc_kernel_data_delivered(skb);
167 }
168 }
169
170 /*
171 * free a socket buffer
172 */
173 static void afs_free_skb(struct sk_buff *skb)
174 {
175 if (!skb) {
176 _debug("FREE NULL [%d]", atomic_read(&afs_outstanding_skbs));
177 dump_stack();
178 } else {
179 _debug("FREE %p{%u} [%d]",
180 skb, skb->mark, atomic_read(&afs_outstanding_skbs));
181 if (atomic_dec_return(&afs_outstanding_skbs) == -1)
182 BUG();
183 rxrpc_kernel_free_skb(skb);
184 }
185 }
186
187 /*
188 * free a call
189 */
190 static void afs_free_call(struct afs_call *call)
191 {
192 _debug("DONE %p{%s} [%d]",
193 call, call->type->name, atomic_read(&afs_outstanding_calls));
194
195 ASSERTCMP(call->rxcall, ==, NULL);
196 ASSERT(!work_pending(&call->async_work));
197 ASSERT(skb_queue_empty(&call->rx_queue));
198 ASSERT(call->type->name != NULL);
199
200 kfree(call->request);
201 kfree(call);
202
203 if (atomic_dec_and_test(&afs_outstanding_calls))
204 wake_up_atomic_t(&afs_outstanding_calls);
205 }
206
207 /*
208 * End a call but do not free it
209 */
210 static void afs_end_call_nofree(struct afs_call *call)
211 {
212 if (call->rxcall) {
213 rxrpc_kernel_end_call(call->rxcall);
214 call->rxcall = NULL;
215 }
216 if (call->type->destructor)
217 call->type->destructor(call);
218 }
219
220 /*
221 * End a call and free it
222 */
223 static void afs_end_call(struct afs_call *call)
224 {
225 afs_end_call_nofree(call);
226 afs_free_call(call);
227 }
228
229 /*
230 * allocate a call with flat request and reply buffers
231 */
232 struct afs_call *afs_alloc_flat_call(const struct afs_call_type *type,
233 size_t request_size, size_t reply_size)
234 {
235 struct afs_call *call;
236
237 call = kzalloc(sizeof(*call), GFP_NOFS);
238 if (!call)
239 goto nomem_call;
240
241 _debug("CALL %p{%s} [%d]",
242 call, type->name, atomic_read(&afs_outstanding_calls));
243 atomic_inc(&afs_outstanding_calls);
244
245 call->type = type;
246 call->request_size = request_size;
247 call->reply_max = reply_size;
248
249 if (request_size) {
250 call->request = kmalloc(request_size, GFP_NOFS);
251 if (!call->request)
252 goto nomem_free;
253 }
254
255 if (reply_size) {
256 call->buffer = kmalloc(reply_size, GFP_NOFS);
257 if (!call->buffer)
258 goto nomem_free;
259 }
260
261 init_waitqueue_head(&call->waitq);
262 skb_queue_head_init(&call->rx_queue);
263 return call;
264
265 nomem_free:
266 afs_free_call(call);
267 nomem_call:
268 return NULL;
269 }
270
271 /*
272 * clean up a call with flat buffer
273 */
274 void afs_flat_call_destructor(struct afs_call *call)
275 {
276 _enter("");
277
278 kfree(call->request);
279 call->request = NULL;
280 kfree(call->buffer);
281 call->buffer = NULL;
282 }
283
284 /*
285 * attach the data from a bunch of pages on an inode to a call
286 */
287 static int afs_send_pages(struct afs_call *call, struct msghdr *msg,
288 struct kvec *iov)
289 {
290 struct page *pages[8];
291 unsigned count, n, loop, offset, to;
292 pgoff_t first = call->first, last = call->last;
293 int ret;
294
295 _enter("");
296
297 offset = call->first_offset;
298 call->first_offset = 0;
299
300 do {
301 _debug("attach %lx-%lx", first, last);
302
303 count = last - first + 1;
304 if (count > ARRAY_SIZE(pages))
305 count = ARRAY_SIZE(pages);
306 n = find_get_pages_contig(call->mapping, first, count, pages);
307 ASSERTCMP(n, ==, count);
308
309 loop = 0;
310 do {
311 msg->msg_flags = 0;
312 to = PAGE_SIZE;
313 if (first + loop >= last)
314 to = call->last_to;
315 else
316 msg->msg_flags = MSG_MORE;
317 iov->iov_base = kmap(pages[loop]) + offset;
318 iov->iov_len = to - offset;
319 offset = 0;
320
321 _debug("- range %u-%u%s",
322 offset, to, msg->msg_flags ? " [more]" : "");
323 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC,
324 iov, 1, to - offset);
325
326 /* have to change the state *before* sending the last
327 * packet as RxRPC might give us the reply before it
328 * returns from sending the request */
329 if (first + loop >= last)
330 call->state = AFS_CALL_AWAIT_REPLY;
331 ret = rxrpc_kernel_send_data(call->rxcall, msg,
332 to - offset);
333 kunmap(pages[loop]);
334 if (ret < 0)
335 break;
336 } while (++loop < count);
337 first += count;
338
339 for (loop = 0; loop < count; loop++)
340 put_page(pages[loop]);
341 if (ret < 0)
342 break;
343 } while (first <= last);
344
345 _leave(" = %d", ret);
346 return ret;
347 }
348
349 /*
350 * initiate a call
351 */
352 int afs_make_call(struct in_addr *addr, struct afs_call *call, gfp_t gfp,
353 const struct afs_wait_mode *wait_mode)
354 {
355 struct sockaddr_rxrpc srx;
356 struct rxrpc_call *rxcall;
357 struct msghdr msg;
358 struct kvec iov[1];
359 int ret;
360 struct sk_buff *skb;
361
362 _enter("%x,{%d},", addr->s_addr, ntohs(call->port));
363
364 ASSERT(call->type != NULL);
365 ASSERT(call->type->name != NULL);
366
367 _debug("____MAKE %p{%s,%x} [%d]____",
368 call, call->type->name, key_serial(call->key),
369 atomic_read(&afs_outstanding_calls));
370
371 call->wait_mode = wait_mode;
372 call->async_workfn = afs_process_async_call;
373 INIT_WORK(&call->async_work, afs_async_workfn);
374
375 memset(&srx, 0, sizeof(srx));
376 srx.srx_family = AF_RXRPC;
377 srx.srx_service = call->service_id;
378 srx.transport_type = SOCK_DGRAM;
379 srx.transport_len = sizeof(srx.transport.sin);
380 srx.transport.sin.sin_family = AF_INET;
381 srx.transport.sin.sin_port = call->port;
382 memcpy(&srx.transport.sin.sin_addr, addr, 4);
383
384 /* create a call */
385 rxcall = rxrpc_kernel_begin_call(afs_socket, &srx, call->key,
386 (unsigned long) call, gfp);
387 call->key = NULL;
388 if (IS_ERR(rxcall)) {
389 ret = PTR_ERR(rxcall);
390 goto error_kill_call;
391 }
392
393 call->rxcall = rxcall;
394
395 /* send the request */
396 iov[0].iov_base = call->request;
397 iov[0].iov_len = call->request_size;
398
399 msg.msg_name = NULL;
400 msg.msg_namelen = 0;
401 iov_iter_kvec(&msg.msg_iter, WRITE | ITER_KVEC, iov, 1,
402 call->request_size);
403 msg.msg_control = NULL;
404 msg.msg_controllen = 0;
405 msg.msg_flags = (call->send_pages ? MSG_MORE : 0);
406
407 /* have to change the state *before* sending the last packet as RxRPC
408 * might give us the reply before it returns from sending the
409 * request */
410 if (!call->send_pages)
411 call->state = AFS_CALL_AWAIT_REPLY;
412 ret = rxrpc_kernel_send_data(rxcall, &msg, call->request_size);
413 if (ret < 0)
414 goto error_do_abort;
415
416 if (call->send_pages) {
417 ret = afs_send_pages(call, &msg, iov);
418 if (ret < 0)
419 goto error_do_abort;
420 }
421
422 /* at this point, an async call may no longer exist as it may have
423 * already completed */
424 return wait_mode->wait(call);
425
426 error_do_abort:
427 rxrpc_kernel_abort_call(rxcall, RX_USER_ABORT);
428 while ((skb = skb_dequeue(&call->rx_queue)))
429 afs_free_skb(skb);
430 error_kill_call:
431 afs_end_call(call);
432 _leave(" = %d", ret);
433 return ret;
434 }
435
436 /*
437 * Handles intercepted messages that were arriving in the socket's Rx queue.
438 *
439 * Called from the AF_RXRPC call processor in waitqueue process context. For
440 * each call, it is guaranteed this will be called in order of packet to be
441 * delivered.
442 */
443 static void afs_rx_interceptor(struct sock *sk, unsigned long user_call_ID,
444 struct sk_buff *skb)
445 {
446 struct afs_call *call = (struct afs_call *) user_call_ID;
447
448 _enter("%p,,%u", call, skb->mark);
449
450 _debug("ICPT %p{%u} [%d]",
451 skb, skb->mark, atomic_read(&afs_outstanding_skbs));
452
453 ASSERTCMP(sk, ==, afs_socket->sk);
454 atomic_inc(&afs_outstanding_skbs);
455
456 if (!call) {
457 /* its an incoming call for our callback service */
458 skb_queue_tail(&afs_incoming_calls, skb);
459 queue_work(afs_wq, &afs_collect_incoming_call_work);
460 } else {
461 /* route the messages directly to the appropriate call */
462 skb_queue_tail(&call->rx_queue, skb);
463 call->wait_mode->rx_wakeup(call);
464 }
465
466 _leave("");
467 }
468
469 /*
470 * deliver messages to a call
471 */
472 static void afs_deliver_to_call(struct afs_call *call)
473 {
474 struct sk_buff *skb;
475 bool last;
476 u32 abort_code;
477 int ret;
478
479 _enter("");
480
481 while ((call->state == AFS_CALL_AWAIT_REPLY ||
482 call->state == AFS_CALL_AWAIT_OP_ID ||
483 call->state == AFS_CALL_AWAIT_REQUEST ||
484 call->state == AFS_CALL_AWAIT_ACK) &&
485 (skb = skb_dequeue(&call->rx_queue))) {
486 switch (skb->mark) {
487 case RXRPC_SKB_MARK_DATA:
488 _debug("Rcv DATA");
489 last = rxrpc_kernel_is_data_last(skb);
490 ret = call->type->deliver(call, skb, last);
491 switch (ret) {
492 case 0:
493 if (last &&
494 call->state == AFS_CALL_AWAIT_REPLY)
495 call->state = AFS_CALL_COMPLETE;
496 break;
497 case -ENOTCONN:
498 abort_code = RX_CALL_DEAD;
499 goto do_abort;
500 case -ENOTSUPP:
501 abort_code = RX_INVALID_OPERATION;
502 goto do_abort;
503 default:
504 abort_code = RXGEN_CC_UNMARSHAL;
505 if (call->state != AFS_CALL_AWAIT_REPLY)
506 abort_code = RXGEN_SS_UNMARSHAL;
507 do_abort:
508 rxrpc_kernel_abort_call(call->rxcall,
509 abort_code);
510 call->error = ret;
511 call->state = AFS_CALL_ERROR;
512 break;
513 }
514 afs_data_delivered(skb);
515 skb = NULL;
516 continue;
517 case RXRPC_SKB_MARK_FINAL_ACK:
518 _debug("Rcv ACK");
519 call->state = AFS_CALL_COMPLETE;
520 break;
521 case RXRPC_SKB_MARK_BUSY:
522 _debug("Rcv BUSY");
523 call->error = -EBUSY;
524 call->state = AFS_CALL_BUSY;
525 break;
526 case RXRPC_SKB_MARK_REMOTE_ABORT:
527 abort_code = rxrpc_kernel_get_abort_code(skb);
528 call->error = call->type->abort_to_error(abort_code);
529 call->state = AFS_CALL_ABORTED;
530 _debug("Rcv ABORT %u -> %d", abort_code, call->error);
531 break;
532 case RXRPC_SKB_MARK_LOCAL_ABORT:
533 abort_code = rxrpc_kernel_get_abort_code(skb);
534 call->error = call->type->abort_to_error(abort_code);
535 call->state = AFS_CALL_ABORTED;
536 _debug("Loc ABORT %u -> %d", abort_code, call->error);
537 break;
538 case RXRPC_SKB_MARK_NET_ERROR:
539 call->error = -rxrpc_kernel_get_error_number(skb);
540 call->state = AFS_CALL_ERROR;
541 _debug("Rcv NET ERROR %d", call->error);
542 break;
543 case RXRPC_SKB_MARK_LOCAL_ERROR:
544 call->error = -rxrpc_kernel_get_error_number(skb);
545 call->state = AFS_CALL_ERROR;
546 _debug("Rcv LOCAL ERROR %d", call->error);
547 break;
548 default:
549 BUG();
550 break;
551 }
552
553 afs_free_skb(skb);
554 }
555
556 /* make sure the queue is empty if the call is done with (we might have
557 * aborted the call early because of an unmarshalling error) */
558 if (call->state >= AFS_CALL_COMPLETE) {
559 while ((skb = skb_dequeue(&call->rx_queue)))
560 afs_free_skb(skb);
561 if (call->incoming)
562 afs_end_call(call);
563 }
564
565 _leave("");
566 }
567
568 /*
569 * wait synchronously for a call to complete
570 */
571 static int afs_wait_for_call_to_complete(struct afs_call *call)
572 {
573 struct sk_buff *skb;
574 int ret;
575
576 DECLARE_WAITQUEUE(myself, current);
577
578 _enter("");
579
580 add_wait_queue(&call->waitq, &myself);
581 for (;;) {
582 set_current_state(TASK_INTERRUPTIBLE);
583
584 /* deliver any messages that are in the queue */
585 if (!skb_queue_empty(&call->rx_queue)) {
586 __set_current_state(TASK_RUNNING);
587 afs_deliver_to_call(call);
588 continue;
589 }
590
591 ret = call->error;
592 if (call->state >= AFS_CALL_COMPLETE)
593 break;
594 ret = -EINTR;
595 if (signal_pending(current))
596 break;
597 schedule();
598 }
599
600 remove_wait_queue(&call->waitq, &myself);
601 __set_current_state(TASK_RUNNING);
602
603 /* kill the call */
604 if (call->state < AFS_CALL_COMPLETE) {
605 _debug("call incomplete");
606 rxrpc_kernel_abort_call(call->rxcall, RX_CALL_DEAD);
607 while ((skb = skb_dequeue(&call->rx_queue)))
608 afs_free_skb(skb);
609 }
610
611 _debug("call complete");
612 afs_end_call(call);
613 _leave(" = %d", ret);
614 return ret;
615 }
616
617 /*
618 * wake up a waiting call
619 */
620 static void afs_wake_up_call_waiter(struct afs_call *call)
621 {
622 wake_up(&call->waitq);
623 }
624
625 /*
626 * wake up an asynchronous call
627 */
628 static void afs_wake_up_async_call(struct afs_call *call)
629 {
630 _enter("");
631 queue_work(afs_async_calls, &call->async_work);
632 }
633
634 /*
635 * put a call into asynchronous mode
636 * - mustn't touch the call descriptor as the call my have completed by the
637 * time we get here
638 */
639 static int afs_dont_wait_for_call_to_complete(struct afs_call *call)
640 {
641 _enter("");
642 return -EINPROGRESS;
643 }
644
645 /*
646 * delete an asynchronous call
647 */
648 static void afs_delete_async_call(struct afs_call *call)
649 {
650 _enter("");
651
652 afs_free_call(call);
653
654 _leave("");
655 }
656
657 /*
658 * perform processing on an asynchronous call
659 * - on a multiple-thread workqueue this work item may try to run on several
660 * CPUs at the same time
661 */
662 static void afs_process_async_call(struct afs_call *call)
663 {
664 _enter("");
665
666 if (!skb_queue_empty(&call->rx_queue))
667 afs_deliver_to_call(call);
668
669 if (call->state >= AFS_CALL_COMPLETE && call->wait_mode) {
670 if (call->wait_mode->async_complete)
671 call->wait_mode->async_complete(call->reply,
672 call->error);
673 call->reply = NULL;
674
675 /* kill the call */
676 afs_end_call_nofree(call);
677
678 /* we can't just delete the call because the work item may be
679 * queued */
680 call->async_workfn = afs_delete_async_call;
681 queue_work(afs_async_calls, &call->async_work);
682 }
683
684 _leave("");
685 }
686
687 /*
688 * empty a socket buffer into a flat reply buffer
689 */
690 void afs_transfer_reply(struct afs_call *call, struct sk_buff *skb)
691 {
692 size_t len = skb->len;
693
694 if (skb_copy_bits(skb, 0, call->buffer + call->reply_size, len) < 0)
695 BUG();
696 call->reply_size += len;
697 }
698
699 /*
700 * accept the backlog of incoming calls
701 */
702 static void afs_collect_incoming_call(struct work_struct *work)
703 {
704 struct rxrpc_call *rxcall;
705 struct afs_call *call = NULL;
706 struct sk_buff *skb;
707
708 while ((skb = skb_dequeue(&afs_incoming_calls))) {
709 _debug("new call");
710
711 /* don't need the notification */
712 afs_free_skb(skb);
713
714 if (!call) {
715 call = kzalloc(sizeof(struct afs_call), GFP_KERNEL);
716 if (!call) {
717 rxrpc_kernel_reject_call(afs_socket);
718 return;
719 }
720
721 call->async_workfn = afs_process_async_call;
722 INIT_WORK(&call->async_work, afs_async_workfn);
723 call->wait_mode = &afs_async_incoming_call;
724 call->type = &afs_RXCMxxxx;
725 init_waitqueue_head(&call->waitq);
726 skb_queue_head_init(&call->rx_queue);
727 call->state = AFS_CALL_AWAIT_OP_ID;
728
729 _debug("CALL %p{%s} [%d]",
730 call, call->type->name,
731 atomic_read(&afs_outstanding_calls));
732 atomic_inc(&afs_outstanding_calls);
733 }
734
735 rxcall = rxrpc_kernel_accept_call(afs_socket,
736 (unsigned long) call);
737 if (!IS_ERR(rxcall)) {
738 call->rxcall = rxcall;
739 call = NULL;
740 }
741 }
742
743 if (call)
744 afs_free_call(call);
745 }
746
747 /*
748 * grab the operation ID from an incoming cache manager call
749 */
750 static int afs_deliver_cm_op_id(struct afs_call *call, struct sk_buff *skb,
751 bool last)
752 {
753 size_t len = skb->len;
754 void *oibuf = (void *) &call->operation_ID;
755
756 _enter("{%u},{%zu},%d", call->offset, len, last);
757
758 ASSERTCMP(call->offset, <, 4);
759
760 /* the operation ID forms the first four bytes of the request data */
761 len = min_t(size_t, len, 4 - call->offset);
762 if (skb_copy_bits(skb, 0, oibuf + call->offset, len) < 0)
763 BUG();
764 if (!pskb_pull(skb, len))
765 BUG();
766 call->offset += len;
767
768 if (call->offset < 4) {
769 if (last) {
770 _leave(" = -EBADMSG [op ID short]");
771 return -EBADMSG;
772 }
773 _leave(" = 0 [incomplete]");
774 return 0;
775 }
776
777 call->state = AFS_CALL_AWAIT_REQUEST;
778
779 /* ask the cache manager to route the call (it'll change the call type
780 * if successful) */
781 if (!afs_cm_incoming_call(call))
782 return -ENOTSUPP;
783
784 /* pass responsibility for the remainer of this message off to the
785 * cache manager op */
786 return call->type->deliver(call, skb, last);
787 }
788
789 /*
790 * send an empty reply
791 */
792 void afs_send_empty_reply(struct afs_call *call)
793 {
794 struct msghdr msg;
795
796 _enter("");
797
798 msg.msg_name = NULL;
799 msg.msg_namelen = 0;
800 iov_iter_kvec(&msg.msg_iter, WRITE | ITER_KVEC, NULL, 0, 0);
801 msg.msg_control = NULL;
802 msg.msg_controllen = 0;
803 msg.msg_flags = 0;
804
805 call->state = AFS_CALL_AWAIT_ACK;
806 switch (rxrpc_kernel_send_data(call->rxcall, &msg, 0)) {
807 case 0:
808 _leave(" [replied]");
809 return;
810
811 case -ENOMEM:
812 _debug("oom");
813 rxrpc_kernel_abort_call(call->rxcall, RX_USER_ABORT);
814 default:
815 afs_end_call(call);
816 _leave(" [error]");
817 return;
818 }
819 }
820
821 /*
822 * send a simple reply
823 */
824 void afs_send_simple_reply(struct afs_call *call, const void *buf, size_t len)
825 {
826 struct msghdr msg;
827 struct kvec iov[1];
828 int n;
829
830 _enter("");
831
832 iov[0].iov_base = (void *) buf;
833 iov[0].iov_len = len;
834 msg.msg_name = NULL;
835 msg.msg_namelen = 0;
836 iov_iter_kvec(&msg.msg_iter, WRITE | ITER_KVEC, iov, 1, len);
837 msg.msg_control = NULL;
838 msg.msg_controllen = 0;
839 msg.msg_flags = 0;
840
841 call->state = AFS_CALL_AWAIT_ACK;
842 n = rxrpc_kernel_send_data(call->rxcall, &msg, len);
843 if (n >= 0) {
844 /* Success */
845 _leave(" [replied]");
846 return;
847 }
848
849 if (n == -ENOMEM) {
850 _debug("oom");
851 rxrpc_kernel_abort_call(call->rxcall, RX_USER_ABORT);
852 }
853 afs_end_call(call);
854 _leave(" [error]");
855 }
856
857 /*
858 * extract a piece of data from the received data socket buffers
859 */
860 int afs_extract_data(struct afs_call *call, struct sk_buff *skb,
861 bool last, void *buf, size_t count)
862 {
863 size_t len = skb->len;
864
865 _enter("{%u},{%zu},%d,,%zu", call->offset, len, last, count);
866
867 ASSERTCMP(call->offset, <, count);
868
869 len = min_t(size_t, len, count - call->offset);
870 if (skb_copy_bits(skb, 0, buf + call->offset, len) < 0 ||
871 !pskb_pull(skb, len))
872 BUG();
873 call->offset += len;
874
875 if (call->offset < count) {
876 if (last) {
877 _leave(" = -EBADMSG [%d < %zu]", call->offset, count);
878 return -EBADMSG;
879 }
880 _leave(" = -EAGAIN");
881 return -EAGAIN;
882 }
883 return 0;
884 }