summaryrefslogtreecommitdiffstats
path: root/drivers/staging/greybus/operation.c
blob: c64b2bf47a43d12a6982404e90cde17de3d08692 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
/*
 * Greybus operations
 *
 * Copyright 2014-2015 Google Inc.
 * Copyright 2014-2015 Linaro Ltd.
 *
 * Released under the GPLv2 only.
 */

#include <linux/kernel.h>
#include <linux/slab.h>
#include <linux/module.h>
#include <linux/workqueue.h>

#include "greybus.h"

/* The default amount of time a request is given to complete */
#define OPERATION_TIMEOUT_DEFAULT	1000	/* milliseconds */

/*
 * XXX This needs to be coordinated with host driver parameters
 * XXX May need to reduce to allow for message header within a page
 */
#define GB_OPERATION_MESSAGE_SIZE_MAX	4096

static struct kmem_cache *gb_operation_cache;
static struct kmem_cache *gb_simple_message_cache;

/* Workqueue to handle Greybus operation completions. */
static struct workqueue_struct *gb_operation_workqueue;

/* Protects the cookie representing whether a message is in flight */
static DEFINE_MUTEX(gb_message_mutex);

/*
 * All operation messages (both requests and responses) begin with
 * a header that encodes the size of the message (header included).
 * This header also contains a unique identifier, that associates a
 * response message with its operation.  The header contains an
 * operation type field, whose interpretation is dependent on what
 * type of protocol is used over the connection.  The high bit
 * (0x80) of the operation type field is used to indicate whether
 * the message is a request (clear) or a response (set).
 *
 * Response messages include an additional result byte, which
 * communicates the result of the corresponding request.  A zero
 * result value means the operation completed successfully.  Any
 * other value indicates an error; in this case, the payload of the
 * response message (if any) is ignored.  The result byte must be
 * zero in the header for a request message.
 *
 * The wire format for all numeric fields in the header is little
 * endian.  Any operation-specific data begins immediately after the
 * header, and is 64-bit aligned.
 */
struct gb_operation_msg_hdr {
	__le16	size;		/* Size in bytes of header + payload */
	__le16	operation_id;	/* Operation unique id */
	__u8	type;		/* E.g GB_I2C_TYPE_* or GB_GPIO_TYPE_* */
	__u8	result;		/* Result of request (in responses only) */
	/* 2 bytes pad, must be zero (ignore when read) */
} __aligned(sizeof(u64));

/*
 * Protects access to connection operations lists, as well as
 * updates to operation->errno.
 */
static DEFINE_SPINLOCK(gb_operations_lock);

/*
 * Set an operation's result.
 *
 * Initially an outgoing operation's errno value is -EBADR.
 * If no error occurs before sending the request message the only
 * valid value operation->errno can be set to is -EINPROGRESS,
 * indicating the request has been (or rather is about to be) sent.
 * At that point nobody should be looking at the result until the
 * response arrives.
 *
 * The first time the result gets set after the request has been
 * sent, that result "sticks."  That is, if two concurrent threads
 * race to set the result, the first one wins.  The return value
 * tells the caller whether its result was recorded; if not the
 * caller has nothing more to do.
 *
 * The result value -EILSEQ is reserved to signal an implementation
 * error; if it's ever observed, the code performing the request has
 * done something fundamentally wrong.  It is an error to try to set
 * the result to -EBADR, and attempts to do so result in a warning,
 * and -EILSEQ is used instead.  Similarly, the only valid result
 * value to set for an operation in initial state is -EINPROGRESS.
 * Attempts to do otherwise will also record a (successful) -EILSEQ
 * operation result.
 */
static bool gb_operation_result_set(struct gb_operation *operation, int result)
{
	unsigned long flags;
	int prev;

	if (result == -EINPROGRESS) {
		/*
		 * -EINPROGRESS is used to indicate the request is
		 * in flight.  It should be the first result value
		 * set after the initial -EBADR.  Issue a warning
		 * and record an implementation error if it's
		 * set at any other time.
		 */
		spin_lock_irqsave(&gb_operations_lock, flags);
		prev = operation->errno;
		if (prev == -EBADR)
			operation->errno = result;
		else
			operation->errno = -EILSEQ;
		spin_unlock_irqrestore(&gb_operations_lock, flags);
		WARN_ON(prev != -EBADR);

		return true;
	}

	/*
	 * The first result value set after a request has been sent
	 * will be the final result of the operation.  Subsequent
	 * attempts to set the result are ignored.
	 *
	 * Note that -EBADR is a reserved "initial state" result
	 * value.  Attempts to set this value result in a warning,
	 * and the result code is set to -EILSEQ instead.
	 */
	if (WARN_ON(result == -EBADR))
		result = -EILSEQ; /* Nobody should be setting -EBADR */

	spin_lock_irqsave(&gb_operations_lock, flags);
	prev = operation->errno;
	if (prev == -EINPROGRESS)
		operation->errno = result;	/* First and final result */
	spin_unlock_irqrestore(&gb_operations_lock, flags);

	return prev == -EINPROGRESS;
}

int gb_operation_result(struct gb_operation *operation)
{
	int result = operation->errno;

	WARN_ON(result == -EBADR);
	WARN_ON(result == -EINPROGRESS);

	return result;
}
EXPORT_SYMBOL_GPL(gb_operation_result);

static struct gb_operation *
gb_operation_find(struct gb_connection *connection, u16 operation_id)
{
	struct gb_operation *operation;
	unsigned long flags;
	bool found = false;

	spin_lock_irqsave(&gb_operations_lock, flags);
	list_for_each_entry(operation, &connection->operations, links)
		if (operation->id == operation_id) {
			found = true;
			break;
		}
	spin_unlock_irqrestore(&gb_operations_lock, flags);

	return found ? operation : NULL;
}

static int gb_message_send(struct gb_message *message)
{
	size_t message_size = sizeof(*message->header) + message->payload_size;
	struct gb_connection *connection = message->operation->connection;
	int ret = 0;
	void *cookie;

	mutex_lock(&gb_message_mutex);
	cookie = connection->hd->driver->buffer_send(connection->hd,
					connection->hd_cport_id,
					message->header,
					message_size,
					GFP_KERNEL);
	if (IS_ERR(cookie))
		ret = PTR_ERR(cookie);
	else
		message->cookie = cookie;
	mutex_unlock(&gb_message_mutex);

	return ret;
}

/*
 * Cancel a message whose buffer we have passed to the host device
 * layer to be sent.
 */
static void gb_message_cancel(struct gb_message *message)
{
	mutex_lock(&gb_message_mutex);
	if (message->cookie) {
		struct greybus_host_device *hd;

		hd = message->operation->connection->hd;
		hd->driver->buffer_cancel(message->cookie);
	}
	mutex_unlock(&gb_message_mutex);
}

static void gb_operation_request_handle(struct gb_operation *operation)
{
	struct gb_protocol *protocol = operation->connection->protocol;

	if (!protocol)
		return;

	/*
	 * If the protocol has no incoming request handler, report
	 * an error and mark the request bad.
	 */
	if (protocol->request_recv) {
		protocol->request_recv(operation->type, operation);
		return;
	}

	dev_err(&operation->connection->dev,
		"unexpected incoming request type 0x%02hhx\n", operation->type);
	if (gb_operation_result_set(operation, -EPROTONOSUPPORT))
		queue_work(gb_operation_workqueue, &operation->work);
	else
		WARN(true, "failed to mark request bad\n");
}

/*
 * Complete an operation in non-atomic context.  For incoming
 * requests, the callback function is the request handler, and
 * the operation result should be -EINPROGRESS at this point.
 *
 * For outgoing requests, the operation result value should have
 * been set before queueing this.  The operation callback function
 * allows the original requester to know the request has completed
 * and its result is available.
 */
static void gb_operation_work(struct work_struct *work)
{
	struct gb_operation *operation;

	operation = container_of(work, struct gb_operation, work);

	operation->callback(operation);

	gb_operation_put(operation);
}


/*
 * Given a pointer to the header in a message sent on a given host
 * device, return the associated message structure.  (This "header"
 * is just the buffer pointer we supply to the host device for
 * sending.)
 */
static struct gb_message *
gb_hd_message_find(struct greybus_host_device *hd, void *header)
{
	struct gb_message *message;
	u8 *result;

	result = (u8 *)header - hd->buffer_headroom - sizeof(*message);
	message = (struct gb_message *)result;

	return message;
}

static void gb_operation_message_init(struct greybus_host_device *hd,
				struct gb_message *message, u16 operation_id,
				size_t payload_size, u8 type)
{
	struct gb_operation_msg_hdr *header;
	u8 *buffer;

	buffer = &message->buffer[0];
	header = (struct gb_operation_msg_hdr *)(buffer + hd->buffer_headroom);

	message->header = header;
	message->payload = payload_size ? header + 1 : NULL;
	message->payload_size = payload_size;

	/*
	 * The type supplied for incoming message buffers will be
	 * 0x00.  Such buffers will be overwritten by arriving data
	 * so there's no need to initialize the message header.
	 */
	if (type != GB_OPERATION_TYPE_INVALID) {
		u16 message_size = (u16)(sizeof(*header) + payload_size);

		/*
		 * For a request, the operation id gets filled in
		 * when the message is sent.  For a response, it
		 * will be copied from the request by the caller.
		 *
		 * The result field in a request message must be
		 * zero.  It will be set just prior to sending for
		 * a response.
		 */
		header->size = cpu_to_le16(message_size);
		header->operation_id = 0;
		header->type = type;
		header->result = 0;
	}
}

/*
 * Allocate a message to be used for an operation request or response.
 * Both types of message contain a common header.  The request message
 * for an outgoing operation is outbound, as is the response message
 * for an incoming operation.  The message header for an outbound
 * message is partially initialized here.
 *
 * The headers for inbound messages don't need to be initialized;
 * they'll be filled in by arriving data.
 *
 * Our message structure consists of:
 *	message structure
 *	headroom
 *	message header  \_ these combined are
 *	message payload /  the message size
 */
static struct gb_message *
gb_operation_message_alloc(struct greybus_host_device *hd, u8 type,
				size_t payload_size, gfp_t gfp_flags)
{
	struct gb_message *message;
	struct gb_operation_msg_hdr *header;
	size_t message_size = payload_size + sizeof(*header);
	size_t size;

	if (hd->buffer_size_max > GB_OPERATION_MESSAGE_SIZE_MAX) {
		pr_warn("limiting buffer size to %u\n",
			GB_OPERATION_MESSAGE_SIZE_MAX);
		hd->buffer_size_max = GB_OPERATION_MESSAGE_SIZE_MAX;
	}

	/* Allocate the message.  Use the slab cache for simple messages */
	if (payload_size) {
		if (message_size > hd->buffer_size_max) {
			pr_warn("requested message size too big (%zu > %zu)\n",
				message_size, hd->buffer_size_max);
			return NULL;
		}

		size = sizeof(*message) + hd->buffer_headroom + message_size;
		message = kzalloc(size, gfp_flags);
	} else {
		message = kmem_cache_zalloc(gb_simple_message_cache, gfp_flags);
	}
	if (!message)
		return NULL;

	/* Initialize the message.  Operation id is filled in later. */
	gb_operation_message_init(hd, message, 0, payload_size, type);

	return message;
}

static void gb_operation_message_free(struct gb_message *message)
{
	if (message->payload_size)
		kfree(message);
	else
		kmem_cache_free(gb_simple_message_cache, message);
}

/*
 * Map an enum gb_operation_status value (which is represented in a
 * message as a single byte) to an appropriate Linux negative errno.
 */
static int gb_operation_status_map(u8 status)
{
	switch (status) {
	case GB_OP_SUCCESS:
		return 0;
	case GB_OP_INTERRUPTED:
		return -EINTR;
	case GB_OP_TIMEOUT:
		return -ETIMEDOUT;
	case GB_OP_NO_MEMORY:
		return -ENOMEM;
	case GB_OP_PROTOCOL_BAD:
		return -EPROTONOSUPPORT;
	case GB_OP_OVERFLOW:
		return -EMSGSIZE;
	case GB_OP_INVALID:
		return -EINVAL;
	case GB_OP_RETRY:
		return -EAGAIN;
	case GB_OP_NONEXISTENT:
		return -ENODEV;
	case GB_OP_MALFUNCTION:
		return -EILSEQ;
	case GB_OP_UNKNOWN_ERROR:
	default:
		return -EIO;
	}
}

/*
 * Map a Linux errno value (from operation->errno) into the value
 * that should represent it in a response message status sent
 * over the wire.  Returns an enum gb_operation_status value (which
 * is represented in a message as a single byte).
 */
static u8 gb_operation_errno_map(int errno)
{
	switch (errno) {
	case 0:
		return GB_OP_SUCCESS;
	case -EINTR:
		return GB_OP_INTERRUPTED;
	case -ETIMEDOUT:
		return GB_OP_TIMEOUT;
	case -ENOMEM:
		return GB_OP_NO_MEMORY;
	case -EPROTONOSUPPORT:
		return GB_OP_PROTOCOL_BAD;
	case -EMSGSIZE:
		return GB_OP_OVERFLOW;	/* Could be underflow too */
	case -EINVAL:
		return GB_OP_INVALID;
	case -EAGAIN:
		return GB_OP_RETRY;
	case -EILSEQ:
		return GB_OP_MALFUNCTION;
	case -ENODEV:
		return GB_OP_NONEXISTENT;
	case -EIO:
	default:
		return GB_OP_UNKNOWN_ERROR;
	}
}

bool gb_operation_response_alloc(struct gb_operation *operation,
					size_t response_size)
{
	struct greybus_host_device *hd = operation->connection->hd;
	struct gb_operation_msg_hdr *request_header;
	struct gb_message *response;
	u8 type;

	type = operation->type | GB_OPERATION_TYPE_RESPONSE;
	response = gb_operation_message_alloc(hd, type, response_size,
						GFP_KERNEL);
	if (!response)
		return false;
	response->operation = operation;

	/*
	 * Size and type get initialized when the message is
	 * allocated.  The errno will be set before sending.  All
	 * that's left is the operation id, which we copy from the
	 * request message header (as-is, in little-endian order).
	 */
	request_header = operation->request->header;
	response->header->operation_id = request_header->operation_id;
	operation->response = response;

	return true;
}
EXPORT_SYMBOL_GPL(gb_operation_response_alloc);

/*
 * Create a Greybus operation to be sent over the given connection.
 * The request buffer will be big enough for a payload of the given
 * size.
 *
 * For outgoing requests, the request message's header will be
 * initialized with the type of the request and the message size.
 * Outgoing operations must also specify the response buffer size,
 * which must be sufficient to hold all expected response data.  The
 * response message header will eventually be overwritten, so there's
 * no need to initialize it here.
 *
 * Request messages for incoming operations can arrive in interrupt
 * context, so they must be allocated with GFP_ATOMIC.  In this case
 * the request buffer will be immediately overwritten, so there is
 * no need to initialize the message header.  Responsibility for
 * allocating a response buffer lies with the incoming request
 * handler for a protocol.  So we don't allocate that here.
 *
 * Returns a pointer to the new operation or a null pointer if an
 * error occurs.
 */
static struct gb_operation *
gb_operation_create_common(struct gb_connection *connection, u8 type,
				size_t request_size, size_t response_size)
{
	struct greybus_host_device *hd = connection->hd;
	struct gb_operation *operation;
	unsigned long flags;
	gfp_t gfp_flags;

	/*
	 * An incoming request will pass an invalid operation type,
	 * because the header will get overwritten anyway.  These
	 * occur in interrupt context, so we must use GFP_ATOMIC.
	 */
	if (type == GB_OPERATION_TYPE_INVALID)
		gfp_flags = GFP_ATOMIC;
	else
		gfp_flags = GFP_KERNEL;
	operation = kmem_cache_zalloc(gb_operation_cache, gfp_flags);
	if (!operation)
		return NULL;
	operation->connection = connection;

	operation->request = gb_operation_message_alloc(hd, type, request_size,
							gfp_flags);
	if (!operation->request)
		goto err_cache;
	operation->request->operation = operation;

	/* Allocate the response buffer for outgoing operations */
	if (type != GB_OPERATION_TYPE_INVALID) {
		if (!gb_operation_response_alloc(operation, response_size))
			goto err_request;
		operation->type = type;
	}
	operation->errno = -EBADR;  /* Initial value--means "never set" */

	INIT_WORK(&operation->work, gb_operation_work);
	init_completion(&operation->completion);
	kref_init(&operation->kref);

	spin_lock_irqsave(&gb_operations_lock, flags);
	list_add_tail(&operation->links, &connection->operations);
	spin_unlock_irqrestore(&gb_operations_lock, flags);

	return operation;

err_request:
	gb_operation_message_free(operation->request);
err_cache:
	kmem_cache_free(gb_operation_cache, operation);

	return NULL;
}

/*
 * Create a new operation associated with the given connection.  The
 * request and response sizes provided are the number of bytes
 * required to hold the request/response payload only.  Both of
 * these are allowed to be 0.  Note that 0x00 is reserved as an
 * invalid operation type for all protocols, and this is enforced
 * here.
 */
struct gb_operation *gb_operation_create(struct gb_connection *connection,
					u8 type, size_t request_size,
					size_t response_size)
{
	if (WARN_ON_ONCE(type == GB_OPERATION_TYPE_INVALID))
		return NULL;
	if (WARN_ON_ONCE(type & GB_OPERATION_TYPE_RESPONSE))
		type &= ~GB_OPERATION_TYPE_RESPONSE;

	return gb_operation_create_common(connection, type,
					request_size, response_size);
}
EXPORT_SYMBOL_GPL(gb_operation_create);

static struct gb_operation *
gb_operation_create_incoming(struct gb_connection *connection, u16 id,
				u8 type, void *data, size_t request_size)
{
	struct gb_operation *operation;

	operation = gb_operation_create_common(connection,
					GB_OPERATION_TYPE_INVALID,
					request_size, 0);
	if (operation) {
		operation->id = id;
		operation->type = type;
		memcpy(operation->request->header, data, request_size);
	}

	return operation;
}

/*
 * Get an additional reference on an operation.
 */
void gb_operation_get(struct gb_operation *operation)
{
	kref_get(&operation->kref);
}
EXPORT_SYMBOL_GPL(gb_operation_get);

/*
 * Destroy a previously created operation.
 */
static void _gb_operation_destroy(struct kref *kref)
{
	struct gb_operation *operation;
	unsigned long flags;

	operation = container_of(kref, struct gb_operation, kref);

	/* XXX Make sure it's not in flight */
	spin_lock_irqsave(&gb_operations_lock, flags);
	list_del(&operation->links);
	spin_unlock_irqrestore(&gb_operations_lock, flags);

	gb_operation_message_free(operation->response);
	gb_operation_message_free(operation->request);

	kmem_cache_free(gb_operation_cache, operation);
}

/*
 * Drop a reference on an operation, and destroy it when the last
 * one is gone.
 */
void gb_operation_put(struct gb_operation *operation)
{
	if (!WARN_ON(!operation))
		kref_put(&operation->kref, _gb_operation_destroy);
}
EXPORT_SYMBOL_GPL(gb_operation_put);

/* Tell the requester we're done */
static void gb_operation_sync_callback(struct gb_operation *operation)
{
	complete(&operation->completion);
}

/*
 * Send an operation request message. The caller has filled in any payload so
 * the request message is ready to go. The callback function supplied will be
 * called when the response message has arrived indicating the operation is
 * complete. In that case, the callback function is responsible for fetching
 * the result of the operation using gb_operation_result() if desired, and
 * dropping the initial reference to the operation.
 */
int gb_operation_request_send(struct gb_operation *operation,
				gb_operation_callback callback)
{
	struct gb_connection *connection = operation->connection;
	struct gb_operation_msg_hdr *header;
	unsigned int cycle;

	if (!callback)
		return -EINVAL;

	if (connection->state != GB_CONNECTION_STATE_ENABLED)
		return -ENOTCONN;

	/*
	 * First, get an extra reference on the operation.
	 * It'll be dropped when the operation completes.
	 */
	gb_operation_get(operation);

	/*
	 * Record the callback function, which is executed in
	 * non-atomic (workqueue) context when the final result
	 * of an operation has been set.
	 */
	operation->callback = callback;

	/*
	 * Assign the operation's id, and store it in the request header.
	 * Zero is a reserved operation id.
	 */
	cycle = (unsigned int)atomic_inc_return(&connection->op_cycle);
	operation->id = (u16)(cycle % U16_MAX + 1);
	header = operation->request->header;
	header->operation_id = cpu_to_le16(operation->id);

	/* All set, send the request */
	gb_operation_result_set(operation, -EINPROGRESS);

	return gb_message_send(operation->request);
}
EXPORT_SYMBOL_GPL(gb_operation_request_send);

/*
 * Send a synchronous operation.  This function is expected to
 * block, returning only when the response has arrived, (or when an
 * error is detected.  The return value is the result of the
 * operation.
 */
int gb_operation_request_send_sync(struct gb_operation *operation)
{
	int ret;
	unsigned long timeout;

	ret = gb_operation_request_send(operation, gb_operation_sync_callback);
	if (ret)
		return ret;

	timeout = msecs_to_jiffies(OPERATION_TIMEOUT_DEFAULT);
	ret = wait_for_completion_interruptible_timeout(&operation->completion, timeout);
	if (ret < 0) {
		/* Cancel the operation if interrupted */
		gb_operation_cancel(operation, -ECANCELED);
	} else if (ret == 0) {
		/* Cancel the operation if op timed out */
		gb_operation_cancel(operation, -ETIMEDOUT);
	}

	return gb_operation_result(operation);
}
EXPORT_SYMBOL_GPL(gb_operation_request_send_sync);

/*
 * Send a response for an incoming operation request.  A non-zero
 * errno indicates a failed operation.
 *
 * If there is any response payload, the incoming request handler is
 * responsible for allocating the response message.  Otherwise the
 * it can simply supply the result errno; this function will
 * allocate the response message if necessary.
 */
int gb_operation_response_send(struct gb_operation *operation, int errno)
{
	/* Record the result */
	if (!gb_operation_result_set(operation, errno)) {
		pr_err("request result already set\n");
		return -EIO;	/* Shouldn't happen */
	}

	if (!operation->response) {
		if (!gb_operation_response_alloc(operation, 0)) {
			pr_err("error allocating response\n");
			/* XXX Respond with pre-allocated -ENOMEM? */
			return -ENOMEM;
		}
	}

	/* Fill in the response header and send it */
	operation->response->header->result = gb_operation_errno_map(errno);

	return gb_message_send(operation->response);
}
EXPORT_SYMBOL_GPL(gb_operation_response_send);

/*
 * This function is called when a buffer send request has completed.
 * The "header" is the message header--the beginning of what we
 * asked to have sent.
 */
void
greybus_data_sent(struct greybus_host_device *hd, void *header, int status)
{
	struct gb_message *message;
	struct gb_operation *operation;

	/* Get the message and record that it is no longer in flight */
	message = gb_hd_message_find(hd, header);
	message->cookie = NULL;

	/*
	 * If the message was a response, we just need to drop our
	 * reference to the operation.  If an error occurred, report
	 * it.
	 *
	 * For requests, if there's no error, there's nothing more
	 * to do until the response arrives.  If an error occurred
	 * attempting to send it, record that as the result of
	 * the operation and schedule its completion.
	 */
	operation = message->operation;
	if (message == operation->response) {
		if (status)
			pr_err("error %d sending response\n", status);
		gb_operation_put(operation);
	} else if (status) {
		if (gb_operation_result_set(operation, status))
			queue_work(gb_operation_workqueue, &operation->work);
	}
}
EXPORT_SYMBOL_GPL(greybus_data_sent);

/*
 * We've received data on a connection, and it doesn't look like a
 * response, so we assume it's a request.
 *
 * This is called in interrupt context, so just copy the incoming
 * data into the request buffer and handle the rest via workqueue.
 */
static void gb_connection_recv_request(struct gb_connection *connection,
				       u16 operation_id, u8 type,
				       void *data, size_t size)
{
	struct gb_operation *operation;

	operation = gb_operation_create_incoming(connection, operation_id,
						type, data, size);
	if (!operation) {
		dev_err(&connection->dev, "can't create operation\n");
		return;		/* XXX Respond with pre-allocated ENOMEM */
	}

	/*
	 * Incoming requests are handled by arranging for the
	 * request handler to be the operation's callback function.
	 *
	 * The last thing the handler does is send a response
	 * message. The original reference to the operation will be
	 * dropped when the response has been sent.
	 */
	operation->callback = gb_operation_request_handle;
	if (gb_operation_result_set(operation, -EINPROGRESS))
		queue_work(gb_operation_workqueue, &operation->work);
}

/*
 * We've received data that appears to be an operation response
 * message.  Look up the operation, and record that we've received
 * its response.
 *
 * This is called in interrupt context, so just copy the incoming
 * data into the response buffer and handle the rest via workqueue.
 */
static void gb_connection_recv_response(struct gb_connection *connection,
			u16 operation_id, u8 result, void *data, size_t size)
{
	struct gb_operation *operation;
	struct gb_message *message;
	int errno = gb_operation_status_map(result);
	size_t message_size;

	operation = gb_operation_find(connection, operation_id);
	if (!operation) {
		dev_err(&connection->dev, "operation not found\n");
		return;
	}

	message = operation->response;
	message_size = sizeof(*message->header) + message->payload_size;
	if (!errno && size != message_size) {
		dev_err(&connection->dev, "bad message size (%zu != %zu)\n",
			size, message_size);
		errno = -EMSGSIZE;
	}

	/* We must ignore the payload if a bad status is returned */
	if (errno)
		size = sizeof(*message->header);
	memcpy(message->header, data, size);

	/* The rest will be handled in work queue context */
	if (gb_operation_result_set(operation, errno))
		queue_work(gb_operation_workqueue, &operation->work);
}

/*
 * Handle data arriving on a connection.  As soon as we return the
 * supplied data buffer will be reused (so unless we do something
 * with, it's effectively dropped).
 */
void gb_connection_recv(struct gb_connection *connection,
				void *data, size_t size)
{
	struct gb_operation_msg_hdr *header;
	size_t msg_size;
	u16 operation_id;

	if (connection->state != GB_CONNECTION_STATE_ENABLED) {
		dev_err(&connection->dev, "dropping %zu received bytes\n",
			size);
		return;
	}

	if (size < sizeof(*header)) {
		dev_err(&connection->dev, "message too small\n");
		return;
	}

	header = data;
	msg_size = le16_to_cpu(header->size);
	if (msg_size > size) {
		dev_err(&connection->dev, "incomplete message\n");
		return;		/* XXX Should still complete operation */
	}

	operation_id = le16_to_cpu(header->operation_id);
	if (header->type & GB_OPERATION_TYPE_RESPONSE)
		gb_connection_recv_response(connection, operation_id,
						header->result, data, msg_size);
	else
		gb_connection_recv_request(connection, operation_id,
						header->type, data, msg_size);
}

/*
 * Cancel an operation, and record the given error to indicate why.
 */
void gb_operation_cancel(struct gb_operation *operation, int errno)
{
	if (gb_operation_result_set(operation, errno)) {
		gb_message_cancel(operation->request);
		gb_message_cancel(operation->response);
	}
	gb_operation_put(operation);
}
EXPORT_SYMBOL_GPL(gb_operation_cancel);

/**
 * gb_operation_sync: implement a "simple" synchronous gb operation.
 * @connection: the Greybus connection to send this to
 * @type: the type of operation to send
 * @request: pointer to a memory buffer to copy the request from
 * @request_size: size of @request
 * @response: pointer to a memory buffer to copy the response to
 * @response_size: the size of @response.
 *
 * This function implements a simple synchronous Greybus operation.  It sends
 * the provided operation request and waits (sleeps) until the corresponding
 * operation response message has been successfully received, or an error
 * occurs.  @request and @response are buffers to hold the request and response
 * data respectively, and if they are not NULL, their size must be specified in
 * @request_size and @response_size.
 *
 * If a response payload is to come back, and @response is not NULL,
 * @response_size number of bytes will be copied into @response if the operation
 * is successful.
 *
 * If there is an error, the response buffer is left alone.
 */
int gb_operation_sync(struct gb_connection *connection, int type,
		      void *request, int request_size,
		      void *response, int response_size)
{
	struct gb_operation *operation;
	int ret;

	if ((response_size && !response) ||
	    (request_size && !request))
		return -EINVAL;

	operation = gb_operation_create(connection, type,
					request_size, response_size);
	if (!operation)
		return -ENOMEM;

	if (request_size)
		memcpy(operation->request->payload, request, request_size);

	ret = gb_operation_request_send_sync(operation);
	if (ret) {
		dev_err(&connection->dev, "synchronous operation failed: %d\n",
			ret);
	} else {
		if (response_size) {
			memcpy(response, operation->response->payload,
			       response_size);
		}
	}
	gb_operation_destroy(operation);

	return ret;
}
EXPORT_SYMBOL_GPL(gb_operation_sync);

int gb_operation_init(void)
{
	size_t size;

	BUILD_BUG_ON(GB_OPERATION_MESSAGE_SIZE_MAX >
			U16_MAX - sizeof(struct gb_operation_msg_hdr));

	/*
	 * A message structure consists of:
	 *  - the message structure itself
	 *  - the headroom set aside for the host device
	 *  - the message header
	 *  - space for the message payload
	 * Messages with no payload are a fairly common case and
	 * have a known fixed maximum size, so we use a slab cache
	 * for them.
	 */
	size = sizeof(struct gb_message) + GB_BUFFER_HEADROOM_MAX +
				sizeof(struct gb_operation_msg_hdr);
	gb_simple_message_cache = kmem_cache_create("gb_simple_message_cache",
							size, 0, 0, NULL);
	if (!gb_simple_message_cache)
		return -ENOMEM;

	gb_operation_cache = kmem_cache_create("gb_operation_cache",
				sizeof(struct gb_operation), 0, 0, NULL);
	if (!gb_operation_cache)
		goto err_simple;

	gb_operation_workqueue = alloc_workqueue("greybus_operation", 0, 1);
	if (!gb_operation_workqueue)
		goto err_operation;

	return 0;
err_operation:
	kmem_cache_destroy(gb_operation_cache);
	gb_operation_cache = NULL;
err_simple:
	kmem_cache_destroy(gb_simple_message_cache);
	gb_simple_message_cache = NULL;

	return -ENOMEM;
}

void gb_operation_exit(void)
{
	destroy_workqueue(gb_operation_workqueue);
	gb_operation_workqueue = NULL;
	kmem_cache_destroy(gb_operation_cache);
	gb_operation_cache = NULL;
	kmem_cache_destroy(gb_simple_message_cache);
	gb_simple_message_cache = NULL;
}