aboutsummaryrefslogtreecommitdiffstats
path: root/module/zfs/zfs_fm.c
blob: fedceee19af9ef4182e0220625f85acf55031f2f (plain)
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
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */
/*
 * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
 * Use is subject to license terms.
 */

/*
 * Copyright (c) 2012 by Delphix. All rights reserved.
 */

#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/vdev.h>
#include <sys/vdev_impl.h>
#include <sys/zio.h>
#include <sys/zio_checksum.h>

#include <sys/fm/fs/zfs.h>
#include <sys/fm/protocol.h>
#include <sys/fm/util.h>
#include <sys/sysevent.h>

/*
 * This general routine is responsible for generating all the different ZFS
 * ereports.  The payload is dependent on the class, and which arguments are
 * supplied to the function:
 *
 * 	EREPORT			POOL	VDEV	IO
 * 	block			X	X	X
 * 	data			X		X
 * 	device			X	X
 * 	pool			X
 *
 * If we are in a loading state, all errors are chained together by the same
 * SPA-wide ENA (Error Numeric Association).
 *
 * For isolated I/O requests, we get the ENA from the zio_t. The propagation
 * gets very complicated due to RAID-Z, gang blocks, and vdev caching.  We want
 * to chain together all ereports associated with a logical piece of data.  For
 * read I/Os, there  are basically three 'types' of I/O, which form a roughly
 * layered diagram:
 *
 *      +---------------+
 * 	| Aggregate I/O |	No associated logical data or device
 * 	+---------------+
 *              |
 *              V
 * 	+---------------+	Reads associated with a piece of logical data.
 * 	|   Read I/O    |	This includes reads on behalf of RAID-Z,
 * 	+---------------+       mirrors, gang blocks, retries, etc.
 *              |
 *              V
 * 	+---------------+	Reads associated with a particular device, but
 * 	| Physical I/O  |	no logical data.  Issued as part of vdev caching
 * 	+---------------+	and I/O aggregation.
 *
 * Note that 'physical I/O' here is not the same terminology as used in the rest
 * of ZIO.  Typically, 'physical I/O' simply means that there is no attached
 * blockpointer.  But I/O with no associated block pointer can still be related
 * to a logical piece of data (i.e. RAID-Z requests).
 *
 * Purely physical I/O always have unique ENAs.  They are not related to a
 * particular piece of logical data, and therefore cannot be chained together.
 * We still generate an ereport, but the DE doesn't correlate it with any
 * logical piece of data.  When such an I/O fails, the delegated I/O requests
 * will issue a retry, which will trigger the 'real' ereport with the correct
 * ENA.
 *
 * We keep track of the ENA for a ZIO chain through the 'io_logical' member.
 * When a new logical I/O is issued, we set this to point to itself.  Child I/Os
 * then inherit this pointer, so that when it is first set subsequent failures
 * will use the same ENA.  For vdev cache fill and queue aggregation I/O,
 * this pointer is set to NULL, and no ereport will be generated (since it
 * doesn't actually correspond to any particular device or piece of data,
 * and the caller will always retry without caching or queueing anyway).
 *
 * For checksum errors, we want to include more information about the actual
 * error which occurs.  Accordingly, we build an ereport when the error is
 * noticed, but instead of sending it in immediately, we hang it off of the
 * io_cksum_report field of the logical IO.  When the logical IO completes
 * (successfully or not), zfs_ereport_finish_checksum() is called with the
 * good and bad versions of the buffer (if available), and we annotate the
 * ereport with information about the differences.
 */
#ifdef _KERNEL
static void
zfs_zevent_post_cb(nvlist_t *nvl, nvlist_t *detector)
{
	if (nvl)
		fm_nvlist_destroy(nvl, FM_NVA_FREE);

	if (detector)
		fm_nvlist_destroy(detector, FM_NVA_FREE);
}

/*
 * We want to rate limit ZIO delay and checksum events so as to not
 * flood ZED when a disk is acting up.
 *
 * Returns 1 if we're ratelimiting, 0 if not.
 */
static int
zfs_is_ratelimiting_event(const char *subclass, vdev_t *vd)
{
	int rc = 0;
	/*
	 * __ratelimit() returns 1 if we're *not* ratelimiting and 0 if we
	 * are.  Invert it to get our return value.
	 */
	if (strcmp(subclass, FM_EREPORT_ZFS_DELAY) == 0) {
		rc = !zfs_ratelimit(&vd->vdev_delay_rl);
	} else if (strcmp(subclass, FM_EREPORT_ZFS_CHECKSUM) == 0) {
		rc = !zfs_ratelimit(&vd->vdev_checksum_rl);
	}

	if (rc)	{
		/* We're rate limiting */
		fm_erpt_dropped_increment();
	}

	return (rc);
}

static void
zfs_ereport_start(nvlist_t **ereport_out, nvlist_t **detector_out,
    const char *subclass, spa_t *spa, vdev_t *vd, zio_t *zio,
    uint64_t stateoroffset, uint64_t size)
{
	nvlist_t *ereport, *detector;

	uint64_t ena;
	char class[64];

	/*
	 * If we are doing a spa_tryimport() or in recovery mode,
	 * ignore errors.
	 */
	if (spa_load_state(spa) == SPA_LOAD_TRYIMPORT ||
	    spa_load_state(spa) == SPA_LOAD_RECOVER)
		return;

	/*
	 * If we are in the middle of opening a pool, and the previous attempt
	 * failed, don't bother logging any new ereports - we're just going to
	 * get the same diagnosis anyway.
	 */
	if (spa_load_state(spa) != SPA_LOAD_NONE &&
	    spa->spa_last_open_failed)
		return;

	if (zio != NULL) {
		/*
		 * If this is not a read or write zio, ignore the error.  This
		 * can occur if the DKIOCFLUSHWRITECACHE ioctl fails.
		 */
		if (zio->io_type != ZIO_TYPE_READ &&
		    zio->io_type != ZIO_TYPE_WRITE)
			return;

		if (vd != NULL) {
			/*
			 * If the vdev has already been marked as failing due
			 * to a failed probe, then ignore any subsequent I/O
			 * errors, as the DE will automatically fault the vdev
			 * on the first such failure.  This also catches cases
			 * where vdev_remove_wanted is set and the device has
			 * not yet been asynchronously placed into the REMOVED
			 * state.
			 */
			if (zio->io_vd == vd && !vdev_accessible(vd, zio))
				return;

			/*
			 * Ignore checksum errors for reads from DTL regions of
			 * leaf vdevs.
			 */
			if (zio->io_type == ZIO_TYPE_READ &&
			    zio->io_error == ECKSUM &&
			    vd->vdev_ops->vdev_op_leaf &&
			    vdev_dtl_contains(vd, DTL_MISSING, zio->io_txg, 1))
				return;
		}
	}

	/*
	 * For probe failure, we want to avoid posting ereports if we've
	 * already removed the device in the meantime.
	 */
	if (vd != NULL &&
	    strcmp(subclass, FM_EREPORT_ZFS_PROBE_FAILURE) == 0 &&
	    (vd->vdev_remove_wanted || vd->vdev_state == VDEV_STATE_REMOVED))
		return;

	if ((strcmp(subclass, FM_EREPORT_ZFS_DELAY) == 0) &&
	    (zio != NULL) && (!zio->io_timestamp)) {
		/* Ignore bogus delay events */
		return;
	}

	if ((ereport = fm_nvlist_create(NULL)) == NULL)
		return;

	if ((detector = fm_nvlist_create(NULL)) == NULL) {
		fm_nvlist_destroy(ereport, FM_NVA_FREE);
		return;
	}

	/*
	 * Serialize ereport generation
	 */
	mutex_enter(&spa->spa_errlist_lock);

	/*
	 * Determine the ENA to use for this event.  If we are in a loading
	 * state, use a SPA-wide ENA.  Otherwise, if we are in an I/O state, use
	 * a root zio-wide ENA.  Otherwise, simply use a unique ENA.
	 */
	if (spa_load_state(spa) != SPA_LOAD_NONE) {
		if (spa->spa_ena == 0)
			spa->spa_ena = fm_ena_generate(0, FM_ENA_FMT1);
		ena = spa->spa_ena;
	} else if (zio != NULL && zio->io_logical != NULL) {
		if (zio->io_logical->io_ena == 0)
			zio->io_logical->io_ena =
			    fm_ena_generate(0, FM_ENA_FMT1);
		ena = zio->io_logical->io_ena;
	} else {
		ena = fm_ena_generate(0, FM_ENA_FMT1);
	}

	/*
	 * Construct the full class, detector, and other standard FMA fields.
	 */
	(void) snprintf(class, sizeof (class), "%s.%s",
	    ZFS_ERROR_CLASS, subclass);

	fm_fmri_zfs_set(detector, FM_ZFS_SCHEME_VERSION, spa_guid(spa),
	    vd != NULL ? vd->vdev_guid : 0);

	fm_ereport_set(ereport, FM_EREPORT_VERSION, class, ena, detector, NULL);

	/*
	 * Construct the per-ereport payload, depending on which parameters are
	 * passed in.
	 */

	/*
	 * Generic payload members common to all ereports.
	 */
	fm_payload_set(ereport,
	    FM_EREPORT_PAYLOAD_ZFS_POOL, DATA_TYPE_STRING, spa_name(spa),
	    FM_EREPORT_PAYLOAD_ZFS_POOL_GUID, DATA_TYPE_UINT64, spa_guid(spa),
	    FM_EREPORT_PAYLOAD_ZFS_POOL_STATE, DATA_TYPE_UINT64,
	    (uint64_t)spa_state(spa),
	    FM_EREPORT_PAYLOAD_ZFS_POOL_CONTEXT, DATA_TYPE_INT32,
	    (int32_t)spa_load_state(spa), NULL);

	fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_POOL_FAILMODE,
	    DATA_TYPE_STRING,
	    spa_get_failmode(spa) == ZIO_FAILURE_MODE_WAIT ?
	    FM_EREPORT_FAILMODE_WAIT :
	    spa_get_failmode(spa) == ZIO_FAILURE_MODE_CONTINUE ?
	    FM_EREPORT_FAILMODE_CONTINUE : FM_EREPORT_FAILMODE_PANIC,
	    NULL);

	if (vd != NULL) {
		vdev_t *pvd = vd->vdev_parent;
		vdev_queue_t *vq = &vd->vdev_queue;
		vdev_stat_t *vs = &vd->vdev_stat;
		vdev_t *spare_vd;
		uint64_t *spare_guids;
		char **spare_paths;
		int i, spare_count;

		fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID,
		    DATA_TYPE_UINT64, vd->vdev_guid,
		    FM_EREPORT_PAYLOAD_ZFS_VDEV_TYPE,
		    DATA_TYPE_STRING, vd->vdev_ops->vdev_op_type, NULL);
		if (vd->vdev_path != NULL)
			fm_payload_set(ereport,
			    FM_EREPORT_PAYLOAD_ZFS_VDEV_PATH,
			    DATA_TYPE_STRING, vd->vdev_path, NULL);
		if (vd->vdev_devid != NULL)
			fm_payload_set(ereport,
			    FM_EREPORT_PAYLOAD_ZFS_VDEV_DEVID,
			    DATA_TYPE_STRING, vd->vdev_devid, NULL);
		if (vd->vdev_fru != NULL)
			fm_payload_set(ereport,
			    FM_EREPORT_PAYLOAD_ZFS_VDEV_FRU,
			    DATA_TYPE_STRING, vd->vdev_fru, NULL);
		if (vd->vdev_enc_sysfs_path != NULL)
			fm_payload_set(ereport,
			    FM_EREPORT_PAYLOAD_ZFS_VDEV_ENC_SYSFS_PATH,
			    DATA_TYPE_STRING, vd->vdev_enc_sysfs_path, NULL);
		if (vd->vdev_ashift)
			fm_payload_set(ereport,
			    FM_EREPORT_PAYLOAD_ZFS_VDEV_ASHIFT,
			    DATA_TYPE_UINT64, vd->vdev_ashift, NULL);

		if (vq != NULL) {
			fm_payload_set(ereport,
			    FM_EREPORT_PAYLOAD_ZFS_VDEV_COMP_TS,
			    DATA_TYPE_UINT64, vq->vq_io_complete_ts, NULL);
			fm_payload_set(ereport,
			    FM_EREPORT_PAYLOAD_ZFS_VDEV_DELTA_TS,
			    DATA_TYPE_UINT64, vq->vq_io_delta_ts, NULL);
		}

		if (vs != NULL) {
			fm_payload_set(ereport,
			    FM_EREPORT_PAYLOAD_ZFS_VDEV_READ_ERRORS,
			    DATA_TYPE_UINT64, vs->vs_read_errors,
			    FM_EREPORT_PAYLOAD_ZFS_VDEV_WRITE_ERRORS,
			    DATA_TYPE_UINT64, vs->vs_write_errors,
			    FM_EREPORT_PAYLOAD_ZFS_VDEV_CKSUM_ERRORS,
			    DATA_TYPE_UINT64, vs->vs_checksum_errors, NULL);
		}

		if (pvd != NULL) {
			fm_payload_set(ereport,
			    FM_EREPORT_PAYLOAD_ZFS_PARENT_GUID,
			    DATA_TYPE_UINT64, pvd->vdev_guid,
			    FM_EREPORT_PAYLOAD_ZFS_PARENT_TYPE,
			    DATA_TYPE_STRING, pvd->vdev_ops->vdev_op_type,
			    NULL);
			if (pvd->vdev_path)
				fm_payload_set(ereport,
				    FM_EREPORT_PAYLOAD_ZFS_PARENT_PATH,
				    DATA_TYPE_STRING, pvd->vdev_path, NULL);
			if (pvd->vdev_devid)
				fm_payload_set(ereport,
				    FM_EREPORT_PAYLOAD_ZFS_PARENT_DEVID,
				    DATA_TYPE_STRING, pvd->vdev_devid, NULL);
		}

		spare_count = spa->spa_spares.sav_count;
		spare_paths = kmem_zalloc(sizeof (char *) * spare_count,
		    KM_SLEEP);
		spare_guids = kmem_zalloc(sizeof (uint64_t) * spare_count,
		    KM_SLEEP);

		for (i = 0; i < spare_count; i++) {
			spare_vd = spa->spa_spares.sav_vdevs[i];
			if (spare_vd) {
				spare_paths[i] = spare_vd->vdev_path;
				spare_guids[i] = spare_vd->vdev_guid;
			}
		}

		fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_VDEV_SPARE_PATHS,
		    DATA_TYPE_STRING_ARRAY, spare_count, spare_paths,
		    FM_EREPORT_PAYLOAD_ZFS_VDEV_SPARE_GUIDS,
		    DATA_TYPE_UINT64_ARRAY, spare_count, spare_guids, NULL);

		kmem_free(spare_guids, sizeof (uint64_t) * spare_count);
		kmem_free(spare_paths, sizeof (char *) * spare_count);
	}

	if (zio != NULL) {
		/*
		 * Payload common to all I/Os.
		 */
		fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_ERR,
		    DATA_TYPE_INT32, zio->io_error, NULL);
		fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_FLAGS,
		    DATA_TYPE_INT32, zio->io_flags, NULL);
		fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_STAGE,
		    DATA_TYPE_UINT32, zio->io_stage, NULL);
		fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_PIPELINE,
		    DATA_TYPE_UINT32, zio->io_pipeline, NULL);
		fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_DELAY,
		    DATA_TYPE_UINT64, zio->io_delay, NULL);
		fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_TIMESTAMP,
		    DATA_TYPE_UINT64, zio->io_timestamp, NULL);
		fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_DELTA,
		    DATA_TYPE_UINT64, zio->io_delta, NULL);

		/*
		 * If the 'size' parameter is non-zero, it indicates this is a
		 * RAID-Z or other I/O where the physical offset and length are
		 * provided for us, instead of within the zio_t.
		 */
		if (vd != NULL) {
			if (size)
				fm_payload_set(ereport,
				    FM_EREPORT_PAYLOAD_ZFS_ZIO_OFFSET,
				    DATA_TYPE_UINT64, stateoroffset,
				    FM_EREPORT_PAYLOAD_ZFS_ZIO_SIZE,
				    DATA_TYPE_UINT64, size, NULL);
			else
				fm_payload_set(ereport,
				    FM_EREPORT_PAYLOAD_ZFS_ZIO_OFFSET,
				    DATA_TYPE_UINT64, zio->io_offset,
				    FM_EREPORT_PAYLOAD_ZFS_ZIO_SIZE,
				    DATA_TYPE_UINT64, zio->io_size, NULL);
		}

		/*
		 * Payload for I/Os with corresponding logical information.
		 */
		if (zio->io_logical != NULL)
			fm_payload_set(ereport,
			    FM_EREPORT_PAYLOAD_ZFS_ZIO_OBJSET,
			    DATA_TYPE_UINT64,
			    zio->io_logical->io_bookmark.zb_objset,
			    FM_EREPORT_PAYLOAD_ZFS_ZIO_OBJECT,
			    DATA_TYPE_UINT64,
			    zio->io_logical->io_bookmark.zb_object,
			    FM_EREPORT_PAYLOAD_ZFS_ZIO_LEVEL,
			    DATA_TYPE_INT64,
			    zio->io_logical->io_bookmark.zb_level,
			    FM_EREPORT_PAYLOAD_ZFS_ZIO_BLKID,
			    DATA_TYPE_UINT64,
			    zio->io_logical->io_bookmark.zb_blkid, NULL);
	} else if (vd != NULL) {
		/*
		 * If we have a vdev but no zio, this is a device fault, and the
		 * 'stateoroffset' parameter indicates the previous state of the
		 * vdev.
		 */
		fm_payload_set(ereport,
		    FM_EREPORT_PAYLOAD_ZFS_PREV_STATE,
		    DATA_TYPE_UINT64, stateoroffset, NULL);
	}

	mutex_exit(&spa->spa_errlist_lock);

	*ereport_out = ereport;
	*detector_out = detector;
}

/* if it's <= 128 bytes, save the corruption directly */
#define	ZFM_MAX_INLINE		(128 / sizeof (uint64_t))

#define	MAX_RANGES		16

typedef struct zfs_ecksum_info {
	/* histograms of set and cleared bits by bit number in a 64-bit word */
	uint16_t zei_histogram_set[sizeof (uint64_t) * NBBY];
	uint16_t zei_histogram_cleared[sizeof (uint64_t) * NBBY];

	/* inline arrays of bits set and cleared. */
	uint64_t zei_bits_set[ZFM_MAX_INLINE];
	uint64_t zei_bits_cleared[ZFM_MAX_INLINE];

	/*
	 * for each range, the number of bits set and cleared.  The Hamming
	 * distance between the good and bad buffers is the sum of them all.
	 */
	uint32_t zei_range_sets[MAX_RANGES];
	uint32_t zei_range_clears[MAX_RANGES];

	struct zei_ranges {
		uint32_t	zr_start;
		uint32_t	zr_end;
	} zei_ranges[MAX_RANGES];

	size_t	zei_range_count;
	uint32_t zei_mingap;
	uint32_t zei_allowed_mingap;

} zfs_ecksum_info_t;

static void
update_histogram(uint64_t value_arg, uint16_t *hist, uint32_t *count)
{
	size_t i;
	size_t bits = 0;
	uint64_t value = BE_64(value_arg);

	/* We store the bits in big-endian (largest-first) order */
	for (i = 0; i < 64; i++) {
		if (value & (1ull << i)) {
			if (hist[63 - i] < UINT16_MAX)
				hist[63 - i]++;
			++bits;
		}
	}
	/* update the count of bits changed */
	*count += bits;
}

/*
 * We've now filled up the range array, and need to increase "mingap" and
 * shrink the range list accordingly.  zei_mingap is always the smallest
 * distance between array entries, so we set the new_allowed_gap to be
 * one greater than that.  We then go through the list, joining together
 * any ranges which are closer than the new_allowed_gap.
 *
 * By construction, there will be at least one.  We also update zei_mingap
 * to the new smallest gap, to prepare for our next invocation.
 */
static void
zei_shrink_ranges(zfs_ecksum_info_t *eip)
{
	uint32_t mingap = UINT32_MAX;
	uint32_t new_allowed_gap = eip->zei_mingap + 1;

	size_t idx, output;
	size_t max = eip->zei_range_count;

	struct zei_ranges *r = eip->zei_ranges;

	ASSERT3U(eip->zei_range_count, >, 0);
	ASSERT3U(eip->zei_range_count, <=, MAX_RANGES);

	output = idx = 0;
	while (idx < max - 1) {
		uint32_t start = r[idx].zr_start;
		uint32_t end = r[idx].zr_end;

		while (idx < max - 1) {
			uint32_t nstart, nend, gap;

			idx++;
			nstart = r[idx].zr_start;
			nend = r[idx].zr_end;

			gap = nstart - end;
			if (gap < new_allowed_gap) {
				end = nend;
				continue;
			}
			if (gap < mingap)
				mingap = gap;
			break;
		}
		r[output].zr_start = start;
		r[output].zr_end = end;
		output++;
	}
	ASSERT3U(output, <, eip->zei_range_count);
	eip->zei_range_count = output;
	eip->zei_mingap = mingap;
	eip->zei_allowed_mingap = new_allowed_gap;
}

static void
zei_add_range(zfs_ecksum_info_t *eip, int start, int end)
{
	struct zei_ranges *r = eip->zei_ranges;
	size_t count = eip->zei_range_count;

	if (count >= MAX_RANGES) {
		zei_shrink_ranges(eip);
		count = eip->zei_range_count;
	}
	if (count == 0) {
		eip->zei_mingap = UINT32_MAX;
		eip->zei_allowed_mingap = 1;
	} else {
		int gap = start - r[count - 1].zr_end;

		if (gap < eip->zei_allowed_mingap) {
			r[count - 1].zr_end = end;
			return;
		}
		if (gap < eip->zei_mingap)
			eip->zei_mingap = gap;
	}
	r[count].zr_start = start;
	r[count].zr_end = end;
	eip->zei_range_count++;
}

static size_t
zei_range_total_size(zfs_ecksum_info_t *eip)
{
	struct zei_ranges *r = eip->zei_ranges;
	size_t count = eip->zei_range_count;
	size_t result = 0;
	size_t idx;

	for (idx = 0; idx < count; idx++)
		result += (r[idx].zr_end - r[idx].zr_start);

	return (result);
}

static zfs_ecksum_info_t *
annotate_ecksum(nvlist_t *ereport, zio_bad_cksum_t *info,
    const abd_t *goodabd, const abd_t *badabd, size_t size,
    boolean_t drop_if_identical)
{
	const uint64_t *good;
	const uint64_t *bad;

	uint64_t allset = 0;
	uint64_t allcleared = 0;

	size_t nui64s = size / sizeof (uint64_t);

	size_t inline_size;
	int no_inline = 0;
	size_t idx;
	size_t range;

	size_t offset = 0;
	ssize_t start = -1;

	zfs_ecksum_info_t *eip = kmem_zalloc(sizeof (*eip), KM_SLEEP);

	/* don't do any annotation for injected checksum errors */
	if (info != NULL && info->zbc_injected)
		return (eip);

	if (info != NULL && info->zbc_has_cksum) {
		fm_payload_set(ereport,
		    FM_EREPORT_PAYLOAD_ZFS_CKSUM_EXPECTED,
		    DATA_TYPE_UINT64_ARRAY,
		    sizeof (info->zbc_expected) / sizeof (uint64_t),
		    (uint64_t *)&info->zbc_expected,
		    FM_EREPORT_PAYLOAD_ZFS_CKSUM_ACTUAL,
		    DATA_TYPE_UINT64_ARRAY,
		    sizeof (info->zbc_actual) / sizeof (uint64_t),
		    (uint64_t *)&info->zbc_actual,
		    FM_EREPORT_PAYLOAD_ZFS_CKSUM_ALGO,
		    DATA_TYPE_STRING,
		    info->zbc_checksum_name,
		    NULL);

		if (info->zbc_byteswapped) {
			fm_payload_set(ereport,
			    FM_EREPORT_PAYLOAD_ZFS_CKSUM_BYTESWAP,
			    DATA_TYPE_BOOLEAN, 1,
			    NULL);
		}
	}

	if (badabd == NULL || goodabd == NULL)
		return (eip);

	ASSERT3U(size, ==, nui64s * sizeof (uint64_t));
	ASSERT3U(size, <=, SPA_MAXBLOCKSIZE);
	ASSERT3U(size, <=, UINT32_MAX);

	good = (const uint64_t *) abd_borrow_buf_copy((abd_t *)goodabd, size);
	bad = (const uint64_t *) abd_borrow_buf_copy((abd_t *)badabd, size);

	/* build up the range list by comparing the two buffers. */
	for (idx = 0; idx < nui64s; idx++) {
		if (good[idx] == bad[idx]) {
			if (start == -1)
				continue;

			zei_add_range(eip, start, idx);
			start = -1;
		} else {
			if (start != -1)
				continue;

			start = idx;
		}
	}
	if (start != -1)
		zei_add_range(eip, start, idx);

	/* See if it will fit in our inline buffers */
	inline_size = zei_range_total_size(eip);
	if (inline_size > ZFM_MAX_INLINE)
		no_inline = 1;

	/*
	 * If there is no change and we want to drop if the buffers are
	 * identical, do so.
	 */
	if (inline_size == 0 && drop_if_identical) {
		kmem_free(eip, sizeof (*eip));
		abd_return_buf((abd_t *)goodabd, (void *)good, size);
		abd_return_buf((abd_t *)badabd, (void *)bad, size);
		return (NULL);
	}

	/*
	 * Now walk through the ranges, filling in the details of the
	 * differences.  Also convert our uint64_t-array offsets to byte
	 * offsets.
	 */
	for (range = 0; range < eip->zei_range_count; range++) {
		size_t start = eip->zei_ranges[range].zr_start;
		size_t end = eip->zei_ranges[range].zr_end;

		for (idx = start; idx < end; idx++) {
			uint64_t set, cleared;

			// bits set in bad, but not in good
			set = ((~good[idx]) & bad[idx]);
			// bits set in good, but not in bad
			cleared = (good[idx] & (~bad[idx]));

			allset |= set;
			allcleared |= cleared;

			if (!no_inline) {
				ASSERT3U(offset, <, inline_size);
				eip->zei_bits_set[offset] = set;
				eip->zei_bits_cleared[offset] = cleared;
				offset++;
			}

			update_histogram(set, eip->zei_histogram_set,
			    &eip->zei_range_sets[range]);
			update_histogram(cleared, eip->zei_histogram_cleared,
			    &eip->zei_range_clears[range]);
		}

		/* convert to byte offsets */
		eip->zei_ranges[range].zr_start	*= sizeof (uint64_t);
		eip->zei_ranges[range].zr_end	*= sizeof (uint64_t);
	}

	abd_return_buf((abd_t *)goodabd, (void *)good, size);
	abd_return_buf((abd_t *)badabd, (void *)bad, size);

	eip->zei_allowed_mingap	*= sizeof (uint64_t);
	inline_size		*= sizeof (uint64_t);

	/* fill in ereport */
	fm_payload_set(ereport,
	    FM_EREPORT_PAYLOAD_ZFS_BAD_OFFSET_RANGES,
	    DATA_TYPE_UINT32_ARRAY, 2 * eip->zei_range_count,
	    (uint32_t *)eip->zei_ranges,
	    FM_EREPORT_PAYLOAD_ZFS_BAD_RANGE_MIN_GAP,
	    DATA_TYPE_UINT32, eip->zei_allowed_mingap,
	    FM_EREPORT_PAYLOAD_ZFS_BAD_RANGE_SETS,
	    DATA_TYPE_UINT32_ARRAY, eip->zei_range_count, eip->zei_range_sets,
	    FM_EREPORT_PAYLOAD_ZFS_BAD_RANGE_CLEARS,
	    DATA_TYPE_UINT32_ARRAY, eip->zei_range_count, eip->zei_range_clears,
	    NULL);

	if (!no_inline) {
		fm_payload_set(ereport,
		    FM_EREPORT_PAYLOAD_ZFS_BAD_SET_BITS,
		    DATA_TYPE_UINT8_ARRAY,
		    inline_size, (uint8_t *)eip->zei_bits_set,
		    FM_EREPORT_PAYLOAD_ZFS_BAD_CLEARED_BITS,
		    DATA_TYPE_UINT8_ARRAY,
		    inline_size, (uint8_t *)eip->zei_bits_cleared,
		    NULL);
	} else {
		fm_payload_set(ereport,
		    FM_EREPORT_PAYLOAD_ZFS_BAD_SET_HISTOGRAM,
		    DATA_TYPE_UINT16_ARRAY,
		    NBBY * sizeof (uint64_t), eip->zei_histogram_set,
		    FM_EREPORT_PAYLOAD_ZFS_BAD_CLEARED_HISTOGRAM,
		    DATA_TYPE_UINT16_ARRAY,
		    NBBY * sizeof (uint64_t), eip->zei_histogram_cleared,
		    NULL);
	}
	return (eip);
}
#endif

void
zfs_ereport_post(const char *subclass, spa_t *spa, vdev_t *vd, zio_t *zio,
    uint64_t stateoroffset, uint64_t size)
{
#ifdef _KERNEL
	nvlist_t *ereport = NULL;
	nvlist_t *detector = NULL;

	if (zfs_is_ratelimiting_event(subclass, vd))
		return;

	zfs_ereport_start(&ereport, &detector,
	    subclass, spa, vd, zio, stateoroffset, size);

	if (ereport == NULL)
		return;

	/* Cleanup is handled by the callback function */
	zfs_zevent_post(ereport, detector, zfs_zevent_post_cb);
#endif
}

void
zfs_ereport_start_checksum(spa_t *spa, vdev_t *vd,
    struct zio *zio, uint64_t offset, uint64_t length, void *arg,
    zio_bad_cksum_t *info)
{
	zio_cksum_report_t *report;


#ifdef _KERNEL
	if (zfs_is_ratelimiting_event(FM_EREPORT_ZFS_CHECKSUM, vd))
		return;
#endif

	report = kmem_zalloc(sizeof (*report), KM_SLEEP);

	if (zio->io_vsd != NULL)
		zio->io_vsd_ops->vsd_cksum_report(zio, report, arg);
	else
		zio_vsd_default_cksum_report(zio, report, arg);

	/* copy the checksum failure information if it was provided */
	if (info != NULL) {
		report->zcr_ckinfo = kmem_zalloc(sizeof (*info), KM_SLEEP);
		bcopy(info, report->zcr_ckinfo, sizeof (*info));
	}

	report->zcr_align = 1ULL << vd->vdev_top->vdev_ashift;
	report->zcr_length = length;

#ifdef _KERNEL
	zfs_ereport_start(&report->zcr_ereport, &report->zcr_detector,
	    FM_EREPORT_ZFS_CHECKSUM, spa, vd, zio, offset, length);

	if (report->zcr_ereport == NULL) {
		zfs_ereport_free_checksum(report);
		return;
	}
#endif

	mutex_enter(&spa->spa_errlist_lock);
	report->zcr_next = zio->io_logical->io_cksum_report;
	zio->io_logical->io_cksum_report = report;
	mutex_exit(&spa->spa_errlist_lock);
}

void
zfs_ereport_finish_checksum(zio_cksum_report_t *report, const abd_t *good_data,
    const abd_t *bad_data, boolean_t drop_if_identical)
{
#ifdef _KERNEL
	zfs_ecksum_info_t *info;

	info = annotate_ecksum(report->zcr_ereport, report->zcr_ckinfo,
	    good_data, bad_data, report->zcr_length, drop_if_identical);
	if (info != NULL)
		zfs_zevent_post(report->zcr_ereport,
		    report->zcr_detector, zfs_zevent_post_cb);
	else
		zfs_zevent_post_cb(report->zcr_ereport, report->zcr_detector);

	report->zcr_ereport = report->zcr_detector = NULL;
	if (info != NULL)
		kmem_free(info, sizeof (*info));
#endif
}

void
zfs_ereport_free_checksum(zio_cksum_report_t *rpt)
{
#ifdef _KERNEL
	if (rpt->zcr_ereport != NULL) {
		fm_nvlist_destroy(rpt->zcr_ereport,
		    FM_NVA_FREE);
		fm_nvlist_destroy(rpt->zcr_detector,
		    FM_NVA_FREE);
	}
#endif
	rpt->zcr_free(rpt->zcr_cbdata, rpt->zcr_cbinfo);

	if (rpt->zcr_ckinfo != NULL)
		kmem_free(rpt->zcr_ckinfo, sizeof (*rpt->zcr_ckinfo));

	kmem_free(rpt, sizeof (*rpt));
}


void
zfs_ereport_post_checksum(spa_t *spa, vdev_t *vd,
    struct zio *zio, uint64_t offset, uint64_t length,
    const abd_t *good_data, const abd_t *bad_data, zio_bad_cksum_t *zbc)
{
#ifdef _KERNEL
	nvlist_t *ereport = NULL;
	nvlist_t *detector = NULL;
	zfs_ecksum_info_t *info;

	zfs_ereport_start(&ereport, &detector,
	    FM_EREPORT_ZFS_CHECKSUM, spa, vd, zio, offset, length);

	if (ereport == NULL)
		return;

	info = annotate_ecksum(ereport, zbc, good_data, bad_data, length,
	    B_FALSE);

	if (info != NULL) {
		zfs_zevent_post(ereport, detector, zfs_zevent_post_cb);
		kmem_free(info, sizeof (*info));
	}
#endif
}

static void
zfs_post_common(spa_t *spa, vdev_t *vd, const char *type, const char *name,
    nvlist_t *aux)
{
#ifdef _KERNEL
	nvlist_t *resource;
	char class[64];

	if (spa_load_state(spa) == SPA_LOAD_TRYIMPORT)
		return;

	if ((resource = fm_nvlist_create(NULL)) == NULL)
		return;

	(void) snprintf(class, sizeof (class), "%s.%s.%s", type,
	    ZFS_ERROR_CLASS, name);
	VERIFY0(nvlist_add_uint8(resource, FM_VERSION, FM_RSRC_VERSION));
	VERIFY0(nvlist_add_string(resource, FM_CLASS, class));
	VERIFY0(nvlist_add_string(resource,
	    FM_EREPORT_PAYLOAD_ZFS_POOL, spa_name(spa)));
	VERIFY0(nvlist_add_uint64(resource,
	    FM_EREPORT_PAYLOAD_ZFS_POOL_GUID, spa_guid(spa)));
	VERIFY0(nvlist_add_uint64(resource,
	    FM_EREPORT_PAYLOAD_ZFS_POOL_STATE, spa_state(spa)));
	VERIFY0(nvlist_add_int32(resource,
	    FM_EREPORT_PAYLOAD_ZFS_POOL_CONTEXT, spa_load_state(spa)));

	if (vd) {
		VERIFY0(nvlist_add_uint64(resource,
		    FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID, vd->vdev_guid));
		VERIFY0(nvlist_add_uint64(resource,
		    FM_EREPORT_PAYLOAD_ZFS_VDEV_STATE, vd->vdev_state));
		if (vd->vdev_path != NULL)
			VERIFY0(nvlist_add_string(resource,
			    FM_EREPORT_PAYLOAD_ZFS_VDEV_PATH, vd->vdev_path));
		if (vd->vdev_devid != NULL)
			VERIFY0(nvlist_add_string(resource,
			    FM_EREPORT_PAYLOAD_ZFS_VDEV_DEVID, vd->vdev_devid));
		if (vd->vdev_fru != NULL)
			VERIFY0(nvlist_add_string(resource,
			    FM_EREPORT_PAYLOAD_ZFS_VDEV_FRU, vd->vdev_fru));
		if (vd->vdev_enc_sysfs_path != NULL)
			VERIFY0(nvlist_add_string(resource,
			    FM_EREPORT_PAYLOAD_ZFS_VDEV_ENC_SYSFS_PATH,
			    vd->vdev_enc_sysfs_path));
		/* also copy any optional payload data */
		if (aux) {
			nvpair_t *elem = NULL;

			while ((elem = nvlist_next_nvpair(aux, elem)) != NULL)
				(void) nvlist_add_nvpair(resource, elem);
		}
	}

	zfs_zevent_post(resource, NULL, zfs_zevent_post_cb);
#endif
}

/*
 * The 'resource.fs.zfs.removed' event is an internal signal that the given vdev
 * has been removed from the system.  This will cause the DE to ignore any
 * recent I/O errors, inferring that they are due to the asynchronous device
 * removal.
 */
void
zfs_post_remove(spa_t *spa, vdev_t *vd)
{
	zfs_post_common(spa, vd, FM_RSRC_CLASS, FM_RESOURCE_REMOVED, NULL);
}

/*
 * The 'resource.fs.zfs.autoreplace' event is an internal signal that the pool
 * has the 'autoreplace' property set, and therefore any broken vdevs will be
 * handled by higher level logic, and no vdev fault should be generated.
 */
void
zfs_post_autoreplace(spa_t *spa, vdev_t *vd)
{
	zfs_post_common(spa, vd, FM_RSRC_CLASS, FM_RESOURCE_AUTOREPLACE, NULL);
}

/*
 * The 'resource.fs.zfs.statechange' event is an internal signal that the
 * given vdev has transitioned its state to DEGRADED or HEALTHY.  This will
 * cause the retire agent to repair any outstanding fault management cases
 * open because the device was not found (fault.fs.zfs.device).
 */
void
zfs_post_state_change(spa_t *spa, vdev_t *vd, uint64_t laststate)
{
#ifdef _KERNEL
	nvlist_t *aux;

	/*
	 * Add optional supplemental keys to payload
	 */
	aux = fm_nvlist_create(NULL);
	if (vd && aux) {
		if (vd->vdev_physpath) {
			(void) nvlist_add_string(aux,
			    FM_EREPORT_PAYLOAD_ZFS_VDEV_PHYSPATH,
			    vd->vdev_physpath);
		}
		if (vd->vdev_enc_sysfs_path) {
			(void) nvlist_add_string(aux,
			    FM_EREPORT_PAYLOAD_ZFS_VDEV_ENC_SYSFS_PATH,
			    vd->vdev_enc_sysfs_path);
		}

		(void) nvlist_add_uint64(aux,
		    FM_EREPORT_PAYLOAD_ZFS_VDEV_LASTSTATE, laststate);
	}

	zfs_post_common(spa, vd, FM_RSRC_CLASS, FM_RESOURCE_STATECHANGE,
	    aux);

	if (aux)
		fm_nvlist_destroy(aux, FM_NVA_FREE);
#endif
}

/*
 * The 'sysevent.fs.zfs.*' events are signals posted to notify user space of
 * change in the pool.  All sysevents are listed in sys/sysevent/eventdefs.h
 * and are designed to be consumed by the ZFS Event Daemon (ZED).  For
 * additional details refer to the zed(8) man page.
 */
void
zfs_post_sysevent(spa_t *spa, vdev_t *vd, const char *name)
{
	zfs_post_common(spa, vd, FM_SYSEVENT_CLASS, name, NULL);
}

#if defined(_KERNEL) && defined(HAVE_SPL)
EXPORT_SYMBOL(zfs_ereport_post);
EXPORT_SYMBOL(zfs_ereport_post_checksum);
EXPORT_SYMBOL(zfs_post_remove);
EXPORT_SYMBOL(zfs_post_autoreplace);
EXPORT_SYMBOL(zfs_post_state_change);
EXPORT_SYMBOL(zfs_post_sysevent);
#endif /* _KERNEL */