aboutsummaryrefslogtreecommitdiff
path: root/sys/cam/nvme/nvme_da.c
blob: 892dee5a9959d289e0c83a0b77130bdf5852a262 (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
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
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
/*-
 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
 *
 * Copyright (c) 2015 Netflix, Inc.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer,
 *    without modification, immediately at the beginning of the file.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 * Derived from ata_da.c:
 * Copyright (c) 2009 Alexander Motin <mav@FreeBSD.org>
 */

#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");

#include <sys/param.h>

#ifdef _KERNEL
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/bio.h>
#include <sys/sysctl.h>
#include <sys/taskqueue.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/conf.h>
#include <sys/devicestat.h>
#include <sys/eventhandler.h>
#include <sys/malloc.h>
#include <sys/cons.h>
#include <sys/proc.h>
#include <sys/reboot.h>
#include <sys/sbuf.h>
#include <geom/geom.h>
#include <geom/geom_disk.h>
#endif /* _KERNEL */

#ifndef _KERNEL
#include <stdio.h>
#include <string.h>
#endif /* _KERNEL */

#include <cam/cam.h>
#include <cam/cam_ccb.h>
#include <cam/cam_periph.h>
#include <cam/cam_xpt_periph.h>
#include <cam/cam_sim.h>
#include <cam/cam_iosched.h>

#include <cam/nvme/nvme_all.h>

typedef enum {
	NDA_STATE_NORMAL
} nda_state;

typedef enum {
	NDA_FLAG_OPEN		= 0x0001,
	NDA_FLAG_DIRTY		= 0x0002,
	NDA_FLAG_SCTX_INIT	= 0x0004,
} nda_flags;
#define NDA_FLAG_STRING		\
	"\020"			\
	"\001OPEN"		\
	"\002DIRTY"		\
	"\003SCTX_INIT"

typedef enum {
	NDA_Q_4K   = 0x01,
	NDA_Q_NONE = 0x00,
} nda_quirks;
	
#define NDA_Q_BIT_STRING	\
	"\020"			\
	"\001Bit 0"

typedef enum {
	NDA_CCB_BUFFER_IO	= 0x01,
	NDA_CCB_DUMP            = 0x02,
	NDA_CCB_TRIM            = 0x03,
	NDA_CCB_PASS            = 0x04,
	NDA_CCB_TYPE_MASK	= 0x0F,
} nda_ccb_state;

/* Offsets into our private area for storing information */
#define ccb_state	ccb_h.ppriv_field0
#define ccb_bp		ccb_h.ppriv_ptr1	/* For NDA_CCB_BUFFER_IO */
#define ccb_trim	ccb_h.ppriv_ptr1	/* For NDA_CCB_TRIM */

struct nda_softc {
	struct   cam_iosched_softc *cam_iosched;
	int			outstanding_cmds;	/* Number of active commands */
	int			refcount;		/* Active xpt_action() calls */
	nda_state		state;
	nda_flags		flags;
	nda_quirks		quirks;
	int			unmappedio;
	quad_t			deletes;
	uint32_t		nsid;			/* Namespace ID for this nda device */
	struct disk		*disk;
	struct task		sysctl_task;
	struct sysctl_ctx_list	sysctl_ctx;
	struct sysctl_oid	*sysctl_tree;
	uint64_t		trim_count;
	uint64_t		trim_ranges;
	uint64_t		trim_lbas;
#ifdef CAM_TEST_FAILURE
	int			force_read_error;
	int			force_write_error;
	int			periodic_read_error;
	int			periodic_read_count;
#endif
#ifdef CAM_IO_STATS
	struct sysctl_ctx_list	sysctl_stats_ctx;
	struct sysctl_oid	*sysctl_stats_tree;
	u_int			timeouts;
	u_int			errors;
	u_int			invalidations;
#endif
};

struct nda_trim_request {
	struct nvme_dsm_range	dsm[NVME_MAX_DSM_TRIM / sizeof(struct nvme_dsm_range)];
	TAILQ_HEAD(, bio) bps;
};
_Static_assert(NVME_MAX_DSM_TRIM % sizeof(struct nvme_dsm_range) == 0,
    "NVME_MAX_DSM_TRIM must be an integral number of ranges");

/* Need quirk table */

static	disk_ioctl_t	ndaioctl;
static	disk_strategy_t	ndastrategy;
static	dumper_t	ndadump;
static	periph_init_t	ndainit;
static	void		ndaasync(void *callback_arg, u_int32_t code,
				struct cam_path *path, void *arg);
static	void		ndasysctlinit(void *context, int pending);
static	int		ndaflagssysctl(SYSCTL_HANDLER_ARGS);
static	periph_ctor_t	ndaregister;
static	periph_dtor_t	ndacleanup;
static	periph_start_t	ndastart;
static	periph_oninv_t	ndaoninvalidate;
static	void		ndadone(struct cam_periph *periph,
			       union ccb *done_ccb);
static  int		ndaerror(union ccb *ccb, u_int32_t cam_flags,
				u_int32_t sense_flags);
static void		ndashutdown(void *arg, int howto);
static void		ndasuspend(void *arg);

#ifndef	NDA_DEFAULT_SEND_ORDERED
#define	NDA_DEFAULT_SEND_ORDERED	1
#endif
#ifndef NDA_DEFAULT_TIMEOUT
#define NDA_DEFAULT_TIMEOUT 30	/* Timeout in seconds */
#endif
#ifndef	NDA_DEFAULT_RETRY
#define	NDA_DEFAULT_RETRY	4
#endif
#ifndef NDA_MAX_TRIM_ENTRIES
#define NDA_MAX_TRIM_ENTRIES  (NVME_MAX_DSM_TRIM / sizeof(struct nvme_dsm_range))/* Number of DSM trims to use, max 256 */
#endif

static SYSCTL_NODE(_kern_cam, OID_AUTO, nda, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
    "CAM Direct Access Disk driver");

//static int nda_retry_count = NDA_DEFAULT_RETRY;
static int nda_send_ordered = NDA_DEFAULT_SEND_ORDERED;
static int nda_default_timeout = NDA_DEFAULT_TIMEOUT;
static int nda_max_trim_entries = NDA_MAX_TRIM_ENTRIES;
static int nda_enable_biospeedup = 1;
static int nda_nvd_compat = 1;
SYSCTL_INT(_kern_cam_nda, OID_AUTO, max_trim, CTLFLAG_RDTUN,
    &nda_max_trim_entries, NDA_MAX_TRIM_ENTRIES,
    "Maximum number of BIO_DELETE to send down as a DSM TRIM.");
SYSCTL_INT(_kern_cam_nda, OID_AUTO, enable_biospeedup, CTLFLAG_RDTUN,
    &nda_enable_biospeedup, 0, "Enable BIO_SPEEDUP processing.");
SYSCTL_INT(_kern_cam_nda, OID_AUTO, nvd_compat, CTLFLAG_RDTUN,
    &nda_nvd_compat, 1, "Enable creation of nvd aliases.");

/*
 * All NVMe media is non-rotational, so all nvme device instances
 * share this to implement the sysctl.
 */
static int nda_rotating_media = 0;

static struct periph_driver ndadriver =
{
	ndainit, "nda",
	TAILQ_HEAD_INITIALIZER(ndadriver.units), /* generation */ 0
};

PERIPHDRIVER_DECLARE(nda, ndadriver);

static MALLOC_DEFINE(M_NVMEDA, "nvme_da", "nvme_da buffers");

/*
 * nice wrappers. Maybe these belong in nvme_all.c instead of
 * here, but this is the only place that uses these. Should
 * we ever grow another NVME periph, we should move them
 * all there wholesale.
 */

static void
nda_nvme_flush(struct nda_softc *softc, struct ccb_nvmeio *nvmeio)
{
	cam_fill_nvmeio(nvmeio,
	    0,			/* retries */
	    ndadone,		/* cbfcnp */
	    CAM_DIR_NONE,	/* flags */
	    NULL,		/* data_ptr */
	    0,			/* dxfer_len */
	    nda_default_timeout * 1000); /* timeout 30s */
	nvme_ns_flush_cmd(&nvmeio->cmd, softc->nsid);
}

static void
nda_nvme_trim(struct nda_softc *softc, struct ccb_nvmeio *nvmeio,
    void *payload, uint32_t num_ranges)
{
	cam_fill_nvmeio(nvmeio,
	    0,			/* retries */
	    ndadone,		/* cbfcnp */
	    CAM_DIR_OUT,	/* flags */
	    payload,		/* data_ptr */
	    num_ranges * sizeof(struct nvme_dsm_range), /* dxfer_len */
	    nda_default_timeout * 1000); /* timeout 30s */
	nvme_ns_trim_cmd(&nvmeio->cmd, softc->nsid, num_ranges);
}

static void
nda_nvme_write(struct nda_softc *softc, struct ccb_nvmeio *nvmeio,
    void *payload, uint64_t lba, uint32_t len, uint32_t count)
{
	cam_fill_nvmeio(nvmeio,
	    0,			/* retries */
	    ndadone,		/* cbfcnp */
	    CAM_DIR_OUT,	/* flags */
	    payload,		/* data_ptr */
	    len,		/* dxfer_len */
	    nda_default_timeout * 1000); /* timeout 30s */
	nvme_ns_write_cmd(&nvmeio->cmd, softc->nsid, lba, count);
}

static void
nda_nvme_rw_bio(struct nda_softc *softc, struct ccb_nvmeio *nvmeio,
    struct bio *bp, uint32_t rwcmd)
{
	int flags = rwcmd == NVME_OPC_READ ? CAM_DIR_IN : CAM_DIR_OUT;
	void *payload;
	uint64_t lba;
	uint32_t count;

	if (bp->bio_flags & BIO_UNMAPPED) {
		flags |= CAM_DATA_BIO;
		payload = bp;
	} else {
		payload = bp->bio_data;
	}

	lba = bp->bio_pblkno;
	count = bp->bio_bcount / softc->disk->d_sectorsize;

	cam_fill_nvmeio(nvmeio,
	    0,			/* retries */
	    ndadone,		/* cbfcnp */
	    flags,		/* flags */
	    payload,		/* data_ptr */
	    bp->bio_bcount,	/* dxfer_len */
	    nda_default_timeout * 1000); /* timeout 30s */
	nvme_ns_rw_cmd(&nvmeio->cmd, rwcmd, softc->nsid, lba, count);
}

static int
ndaopen(struct disk *dp)
{
	struct cam_periph *periph;
	struct nda_softc *softc;
	int error;

	periph = (struct cam_periph *)dp->d_drv1;
	if (cam_periph_acquire(periph) != 0) {
		return(ENXIO);
	}

	cam_periph_lock(periph);
	if ((error = cam_periph_hold(periph, PRIBIO|PCATCH)) != 0) {
		cam_periph_unlock(periph);
		cam_periph_release(periph);
		return (error);
	}

	CAM_DEBUG(periph->path, CAM_DEBUG_TRACE | CAM_DEBUG_PERIPH,
	    ("ndaopen\n"));

	softc = (struct nda_softc *)periph->softc;
	softc->flags |= NDA_FLAG_OPEN;

	cam_periph_unhold(periph);
	cam_periph_unlock(periph);
	return (0);
}

static int
ndaclose(struct disk *dp)
{
	struct	cam_periph *periph;
	struct	nda_softc *softc;
	union ccb *ccb;
	int error;

	periph = (struct cam_periph *)dp->d_drv1;
	softc = (struct nda_softc *)periph->softc;
	cam_periph_lock(periph);

	CAM_DEBUG(periph->path, CAM_DEBUG_TRACE | CAM_DEBUG_PERIPH,
	    ("ndaclose\n"));

	if ((softc->flags & NDA_FLAG_DIRTY) != 0 &&
	    (periph->flags & CAM_PERIPH_INVALID) == 0 &&
	    cam_periph_hold(periph, PRIBIO) == 0) {

		ccb = cam_periph_getccb(periph, CAM_PRIORITY_NORMAL);
		nda_nvme_flush(softc, &ccb->nvmeio);
		error = cam_periph_runccb(ccb, ndaerror, /*cam_flags*/0,
		    /*sense_flags*/0, softc->disk->d_devstat);

		if (error != 0)
			xpt_print(periph->path, "Synchronize cache failed\n");
		else
			softc->flags &= ~NDA_FLAG_DIRTY;
		xpt_release_ccb(ccb);
		cam_periph_unhold(periph);
	}

	softc->flags &= ~NDA_FLAG_OPEN;

	while (softc->refcount != 0)
		cam_periph_sleep(periph, &softc->refcount, PRIBIO, "ndaclose", 1);
	KASSERT(softc->outstanding_cmds == 0,
	    ("nda %d outstanding commands", softc->outstanding_cmds));
	cam_periph_unlock(periph);
	cam_periph_release(periph);
	return (0);	
}

static void
ndaschedule(struct cam_periph *periph)
{
	struct nda_softc *softc = (struct nda_softc *)periph->softc;

	if (softc->state != NDA_STATE_NORMAL)
		return;

	cam_iosched_schedule(softc->cam_iosched, periph);
}

static int
ndaioctl(struct disk *dp, u_long cmd, void *data, int fflag,
    struct thread *td)
{
	struct cam_periph *periph;
	struct nda_softc *softc;

	periph = (struct cam_periph *)dp->d_drv1;
	softc = (struct nda_softc *)periph->softc;

	switch (cmd) {
	case NVME_IO_TEST:
	case NVME_BIO_TEST:
		/*
		 * These don't map well to the underlying CCBs, so
		 * they are usupported via CAM.
		 */
		return (ENOTTY);
	case NVME_GET_NSID:
	{
		struct nvme_get_nsid *gnsid = (struct nvme_get_nsid *)data;
		struct ccb_pathinq cpi;

		xpt_path_inq(&cpi, periph->path);
		strncpy(gnsid->cdev, cpi.xport_specific.nvme.dev_name,
		    sizeof(gnsid->cdev));
		gnsid->nsid = cpi.xport_specific.nvme.nsid;
		return (0);
	}
	case NVME_PASSTHROUGH_CMD:
	{
		struct nvme_pt_command *pt;
		union ccb *ccb;
		struct cam_periph_map_info mapinfo;
		u_int maxmap = dp->d_maxsize;
		int error;

		/*
		 * Create a NVME_IO CCB to do the passthrough command.
		 */
		pt = (struct nvme_pt_command *)data;
		ccb = xpt_alloc_ccb();
		xpt_setup_ccb(&ccb->ccb_h, periph->path, CAM_PRIORITY_NORMAL);
		ccb->ccb_state = NDA_CCB_PASS;
		cam_fill_nvmeio(&ccb->nvmeio,
		    0,			/* Retries */
		    ndadone,
		    (pt->is_read ? CAM_DIR_IN : CAM_DIR_OUT) | CAM_DATA_VADDR,
		    pt->buf,
		    pt->len,
		    nda_default_timeout * 1000);
		memcpy(&ccb->nvmeio.cmd, &pt->cmd, sizeof(pt->cmd));

		/*
		 * Wire the user memory in this request for the I/O
		 */
		memset(&mapinfo, 0, sizeof(mapinfo));
		error = cam_periph_mapmem(ccb, &mapinfo, maxmap);
		if (error)
			goto out;

		/*
		 * Lock the periph and run the command.
		 */
		cam_periph_lock(periph);
		cam_periph_runccb(ccb, NULL, CAM_RETRY_SELTO,
		    SF_RETRY_UA | SF_NO_PRINT, NULL);

		/*
		 * Tear down mapping and return status.
		 */
		cam_periph_unlock(periph);
		cam_periph_unmapmem(ccb, &mapinfo);
		error = (ccb->ccb_h.status == CAM_REQ_CMP) ? 0 : EIO;
out:
		cam_periph_lock(periph);
		xpt_release_ccb(ccb);
		cam_periph_unlock(periph);
		return (error);
	}
	default:
		break;
	}
	return (ENOTTY);
}

/*
 * Actually translate the requested transfer into one the physical driver
 * can understand.  The transfer is described by a buf and will include
 * only one physical transfer.
 */
static void
ndastrategy(struct bio *bp)
{
	struct cam_periph *periph;
	struct nda_softc *softc;
	
	periph = (struct cam_periph *)bp->bio_disk->d_drv1;
	softc = (struct nda_softc *)periph->softc;

	cam_periph_lock(periph);

	CAM_DEBUG(periph->path, CAM_DEBUG_TRACE, ("ndastrategy(%p)\n", bp));

	/*
	 * If the device has been made invalid, error out
	 */
	if ((periph->flags & CAM_PERIPH_INVALID) != 0) {
		cam_periph_unlock(periph);
		biofinish(bp, NULL, ENXIO);
		return;
	}
	
	if (bp->bio_cmd == BIO_DELETE)
		softc->deletes++;

	/*
	 * Place it in the queue of disk activities for this disk
	 */
	cam_iosched_queue_work(softc->cam_iosched, bp);

	/*
	 * Schedule ourselves for performing the work.
	 */
	ndaschedule(periph);
	cam_periph_unlock(periph);

	return;
}

static int
ndadump(void *arg, void *virtual, vm_offset_t physical, off_t offset, size_t length)
{
	struct	    cam_periph *periph;
	struct	    nda_softc *softc;
	u_int	    secsize;
	struct ccb_nvmeio nvmeio;
	struct	    disk *dp;
	uint64_t    lba;
	uint32_t    count;
	int	    error = 0;

	dp = arg;
	periph = dp->d_drv1;
	softc = (struct nda_softc *)periph->softc;
	secsize = softc->disk->d_sectorsize;
	lba = offset / secsize;
	count = length / secsize;
	
	if ((periph->flags & CAM_PERIPH_INVALID) != 0)
		return (ENXIO);

	/* xpt_get_ccb returns a zero'd allocation for the ccb, mimic that here */
	memset(&nvmeio, 0, sizeof(nvmeio));
	if (length > 0) {
		xpt_setup_ccb(&nvmeio.ccb_h, periph->path, CAM_PRIORITY_NORMAL);
		nvmeio.ccb_state = NDA_CCB_DUMP;
		nda_nvme_write(softc, &nvmeio, virtual, lba, length, count);
		error = cam_periph_runccb((union ccb *)&nvmeio, cam_periph_error,
		    0, SF_NO_RECOVERY | SF_NO_RETRY, NULL);
		if (error != 0)
			printf("Aborting dump due to I/O error %d.\n", error);

		return (error);
	}
	
	/* Flush */
	xpt_setup_ccb(&nvmeio.ccb_h, periph->path, CAM_PRIORITY_NORMAL);

	nvmeio.ccb_state = NDA_CCB_DUMP;
	nda_nvme_flush(softc, &nvmeio);
	error = cam_periph_runccb((union ccb *)&nvmeio, cam_periph_error,
	    0, SF_NO_RECOVERY | SF_NO_RETRY, NULL);
	if (error != 0)
		xpt_print(periph->path, "flush cmd failed\n");
	return (error);
}

static void
ndainit(void)
{
	cam_status status;

	/*
	 * Install a global async callback.  This callback will
	 * receive async callbacks like "new device found".
	 */
	status = xpt_register_async(AC_FOUND_DEVICE, ndaasync, NULL, NULL);

	if (status != CAM_REQ_CMP) {
		printf("nda: Failed to attach master async callback "
		       "due to status 0x%x!\n", status);
	} else if (nda_send_ordered) {

		/* Register our event handlers */
		if ((EVENTHANDLER_REGISTER(power_suspend, ndasuspend,
					   NULL, EVENTHANDLER_PRI_LAST)) == NULL)
		    printf("ndainit: power event registration failed!\n");
		if ((EVENTHANDLER_REGISTER(shutdown_post_sync, ndashutdown,
					   NULL, SHUTDOWN_PRI_DEFAULT)) == NULL)
		    printf("ndainit: shutdown event registration failed!\n");
	}
}

/*
 * Callback from GEOM, called when it has finished cleaning up its
 * resources.
 */
static void
ndadiskgonecb(struct disk *dp)
{
	struct cam_periph *periph;

	periph = (struct cam_periph *)dp->d_drv1;

	cam_periph_release(periph);
}

static void
ndaoninvalidate(struct cam_periph *periph)
{
	struct nda_softc *softc;

	softc = (struct nda_softc *)periph->softc;

	/*
	 * De-register any async callbacks.
	 */
	xpt_register_async(0, ndaasync, periph, periph->path);
#ifdef CAM_IO_STATS
	softc->invalidations++;
#endif

	/*
	 * Return all queued I/O with ENXIO.
	 * XXX Handle any transactions queued to the card
	 *     with XPT_ABORT_CCB.
	 */
	cam_iosched_flush(softc->cam_iosched, NULL, ENXIO);

	disk_gone(softc->disk);
}

static void
ndacleanup(struct cam_periph *periph)
{
	struct nda_softc *softc;

	softc = (struct nda_softc *)periph->softc;

	cam_periph_unlock(periph);

	cam_iosched_fini(softc->cam_iosched);

	/*
	 * If we can't free the sysctl tree, oh well...
	 */
	if ((softc->flags & NDA_FLAG_SCTX_INIT) != 0) {
#ifdef CAM_IO_STATS
		if (sysctl_ctx_free(&softc->sysctl_stats_ctx) != 0)
			xpt_print(periph->path,
			    "can't remove sysctl stats context\n");
#endif
		if (sysctl_ctx_free(&softc->sysctl_ctx) != 0)
			xpt_print(periph->path,
			    "can't remove sysctl context\n");
	}

	disk_destroy(softc->disk);
	free(softc, M_DEVBUF);
	cam_periph_lock(periph);
}

static void
ndaasync(void *callback_arg, u_int32_t code,
	struct cam_path *path, void *arg)
{
	struct cam_periph *periph;

	periph = (struct cam_periph *)callback_arg;
	switch (code) {
	case AC_FOUND_DEVICE:
	{
		struct ccb_getdev *cgd;
		cam_status status;
 
		cgd = (struct ccb_getdev *)arg;
		if (cgd == NULL)
			break;

		if (cgd->protocol != PROTO_NVME)
			break;

		/*
		 * Allocate a peripheral instance for
		 * this device and start the probe
		 * process.
		 */
		status = cam_periph_alloc(ndaregister, ndaoninvalidate,
					  ndacleanup, ndastart,
					  "nda", CAM_PERIPH_BIO,
					  path, ndaasync,
					  AC_FOUND_DEVICE, cgd);

		if (status != CAM_REQ_CMP
		 && status != CAM_REQ_INPROG)
			printf("ndaasync: Unable to attach to new device "
				"due to status 0x%x\n", status);
		break;
	}
	case AC_ADVINFO_CHANGED:
	{
		uintptr_t buftype;

		buftype = (uintptr_t)arg;
		if (buftype == CDAI_TYPE_PHYS_PATH) {
			struct nda_softc *softc;

			softc = periph->softc;
			disk_attr_changed(softc->disk, "GEOM::physpath",
					  M_NOWAIT);
		}
		break;
	}
	case AC_LOST_DEVICE:
	default:
		cam_periph_async(periph, code, path, arg);
		break;
	}
}

static void
ndasysctlinit(void *context, int pending)
{
	struct cam_periph *periph;
	struct nda_softc *softc;
	char tmpstr[32], tmpstr2[16];

	periph = (struct cam_periph *)context;

	/* periph was held for us when this task was enqueued */
	if ((periph->flags & CAM_PERIPH_INVALID) != 0) {
		cam_periph_release(periph);
		return;
	}

	softc = (struct nda_softc *)periph->softc;
	snprintf(tmpstr, sizeof(tmpstr), "CAM NDA unit %d", periph->unit_number);
	snprintf(tmpstr2, sizeof(tmpstr2), "%d", periph->unit_number);

	sysctl_ctx_init(&softc->sysctl_ctx);
	softc->flags |= NDA_FLAG_SCTX_INIT;
	softc->sysctl_tree = SYSCTL_ADD_NODE_WITH_LABEL(&softc->sysctl_ctx,
		SYSCTL_STATIC_CHILDREN(_kern_cam_nda), OID_AUTO, tmpstr2,
		CTLFLAG_RD | CTLFLAG_MPSAFE, 0, tmpstr, "device_index");
	if (softc->sysctl_tree == NULL) {
		printf("ndasysctlinit: unable to allocate sysctl tree\n");
		cam_periph_release(periph);
		return;
	}

	SYSCTL_ADD_INT(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree),
	    OID_AUTO, "unmapped_io", CTLFLAG_RD,
	    &softc->unmappedio, 0, "Unmapped I/O leaf");

	SYSCTL_ADD_QUAD(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree),
	    OID_AUTO, "deletes", CTLFLAG_RD,
	    &softc->deletes, "Number of BIO_DELETE requests");

	SYSCTL_ADD_UQUAD(&softc->sysctl_ctx,
		SYSCTL_CHILDREN(softc->sysctl_tree), OID_AUTO,
		"trim_count", CTLFLAG_RD, &softc->trim_count,
		"Total number of unmap/dsm commands sent");
	SYSCTL_ADD_UQUAD(&softc->sysctl_ctx,
		SYSCTL_CHILDREN(softc->sysctl_tree), OID_AUTO,
		"trim_ranges", CTLFLAG_RD, &softc->trim_ranges,
		"Total number of ranges in unmap/dsm commands");
	SYSCTL_ADD_UQUAD(&softc->sysctl_ctx,
		SYSCTL_CHILDREN(softc->sysctl_tree), OID_AUTO,
		"trim_lbas", CTLFLAG_RD, &softc->trim_lbas,
		"Total lbas in the unmap/dsm commands sent");

	SYSCTL_ADD_INT(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree),
	    OID_AUTO, "rotating", CTLFLAG_RD, &nda_rotating_media, 1,
	    "Rotating media");

	SYSCTL_ADD_PROC(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree),
	    OID_AUTO, "flags", CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE,
	    softc, 0, ndaflagssysctl, "A",
	    "Flags for drive");

#ifdef CAM_IO_STATS
	softc->sysctl_stats_tree = SYSCTL_ADD_NODE(&softc->sysctl_stats_ctx,
		SYSCTL_CHILDREN(softc->sysctl_tree), OID_AUTO, "stats",
		CTLFLAG_RD | CTLFLAG_MPSAFE, 0, "Statistics");
	if (softc->sysctl_stats_tree == NULL) {
		printf("ndasysctlinit: unable to allocate sysctl tree for stats\n");
		cam_periph_release(periph);
		return;
	}
	SYSCTL_ADD_INT(&softc->sysctl_stats_ctx,
		SYSCTL_CHILDREN(softc->sysctl_stats_tree),
		OID_AUTO, "timeouts", CTLFLAG_RD,
		&softc->timeouts, 0,
		"Device timeouts reported by the SIM");
	SYSCTL_ADD_INT(&softc->sysctl_stats_ctx,
		SYSCTL_CHILDREN(softc->sysctl_stats_tree),
		OID_AUTO, "errors", CTLFLAG_RD,
		&softc->errors, 0,
		"Transport errors reported by the SIM.");
	SYSCTL_ADD_INT(&softc->sysctl_stats_ctx,
		SYSCTL_CHILDREN(softc->sysctl_stats_tree),
		OID_AUTO, "pack_invalidations", CTLFLAG_RD,
		&softc->invalidations, 0,
		"Device pack invalidations.");
#endif

#ifdef CAM_TEST_FAILURE
	SYSCTL_ADD_PROC(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree),
		OID_AUTO, "invalidate", CTLTYPE_U64 | CTLFLAG_RW | CTLFLAG_MPSAFE,
		periph, 0, cam_periph_invalidate_sysctl, "I",
		"Write 1 to invalidate the drive immediately");
#endif

	cam_iosched_sysctl_init(softc->cam_iosched, &softc->sysctl_ctx,
	    softc->sysctl_tree);

	cam_periph_release(periph);
}

static int
ndaflagssysctl(SYSCTL_HANDLER_ARGS)
{
	struct sbuf sbuf;
	struct nda_softc *softc = arg1;
	int error;

	sbuf_new_for_sysctl(&sbuf, NULL, 0, req);
	if (softc->flags != 0)
		sbuf_printf(&sbuf, "0x%b", (unsigned)softc->flags, NDA_FLAG_STRING);
	else
		sbuf_printf(&sbuf, "0");
	error = sbuf_finish(&sbuf);
	sbuf_delete(&sbuf);

	return (error);
}

static int
ndagetattr(struct bio *bp)
{
	int ret;
	struct cam_periph *periph;

	if (g_handleattr_int(bp, "GEOM::canspeedup", nda_enable_biospeedup))
		return (EJUSTRETURN);

	periph = (struct cam_periph *)bp->bio_disk->d_drv1;
	cam_periph_lock(periph);
	ret = xpt_getattr(bp->bio_data, bp->bio_length, bp->bio_attribute,
	    periph->path);
	cam_periph_unlock(periph);
	if (ret == 0)
		bp->bio_completed = bp->bio_length;
	return ret;
}

static cam_status
ndaregister(struct cam_periph *periph, void *arg)
{
	struct nda_softc *softc;
	struct disk *disk;
	struct ccb_pathinq cpi;
	const struct nvme_namespace_data *nsd;
	const struct nvme_controller_data *cd;
	char   announce_buf[80];
	uint8_t flbas_fmt, lbads, vwc_present;
	u_int maxio;
	int quirks;

	nsd = nvme_get_identify_ns(periph);
	cd = nvme_get_identify_cntrl(periph);

	softc = (struct nda_softc *)malloc(sizeof(*softc), M_DEVBUF,
	    M_NOWAIT | M_ZERO);

	if (softc == NULL) {
		printf("ndaregister: Unable to probe new device. "
		    "Unable to allocate softc\n");
		return(CAM_REQ_CMP_ERR);
	}

	if (cam_iosched_init(&softc->cam_iosched, periph) != 0) {
		printf("ndaregister: Unable to probe new device. "
		       "Unable to allocate iosched memory\n");
		free(softc, M_DEVBUF);
		return(CAM_REQ_CMP_ERR);
	}

	/* ident_data parsing */

	periph->softc = softc;
	softc->quirks = NDA_Q_NONE;
	xpt_path_inq(&cpi, periph->path);
	TASK_INIT(&softc->sysctl_task, 0, ndasysctlinit, periph);

	/*
	 * The name space ID is the lun, save it for later I/O
	 */
	softc->nsid = (uint32_t)xpt_path_lun_id(periph->path);

	/*
	 * Register this media as a disk
	 */
	(void)cam_periph_hold(periph, PRIBIO);
	cam_periph_unlock(periph);
	snprintf(announce_buf, sizeof(announce_buf),
	    "kern.cam.nda.%d.quirks", periph->unit_number);
	quirks = softc->quirks;
	TUNABLE_INT_FETCH(announce_buf, &quirks);
	softc->quirks = quirks;
	cam_iosched_set_sort_queue(softc->cam_iosched, 0);
	softc->disk = disk = disk_alloc();
	disk->d_rotation_rate = DISK_RR_NON_ROTATING;
	disk->d_open = ndaopen;
	disk->d_close = ndaclose;
	disk->d_strategy = ndastrategy;
	disk->d_ioctl = ndaioctl;
	disk->d_getattr = ndagetattr;
	disk->d_dump = ndadump;
	disk->d_gone = ndadiskgonecb;
	disk->d_name = "nda";
	disk->d_drv1 = periph;
	disk->d_unit = periph->unit_number;
	maxio = cpi.maxio;		/* Honor max I/O size of SIM */
	if (maxio == 0)
		maxio = DFLTPHYS;	/* traditional default */
	else if (maxio > MAXPHYS)
		maxio = MAXPHYS;	/* for safety */
	disk->d_maxsize = maxio;
	flbas_fmt = (nsd->flbas >> NVME_NS_DATA_FLBAS_FORMAT_SHIFT) &
		NVME_NS_DATA_FLBAS_FORMAT_MASK;
	lbads = (nsd->lbaf[flbas_fmt] >> NVME_NS_DATA_LBAF_LBADS_SHIFT) &
		NVME_NS_DATA_LBAF_LBADS_MASK;
	disk->d_sectorsize = 1 << lbads;
	disk->d_mediasize = (off_t)(disk->d_sectorsize * nsd->nsze);
	disk->d_delmaxsize = disk->d_mediasize;
	disk->d_flags = DISKFLAG_DIRECT_COMPLETION;
	if (nvme_ctrlr_has_dataset_mgmt(cd))
		disk->d_flags |= DISKFLAG_CANDELETE;
	vwc_present = (cd->vwc >> NVME_CTRLR_DATA_VWC_PRESENT_SHIFT) &
		NVME_CTRLR_DATA_VWC_PRESENT_MASK;
	if (vwc_present)
		disk->d_flags |= DISKFLAG_CANFLUSHCACHE;
	if ((cpi.hba_misc & PIM_UNMAPPED) != 0) {
		disk->d_flags |= DISKFLAG_UNMAPPED_BIO;
		softc->unmappedio = 1;
	}
	/*
	 * d_ident and d_descr are both far bigger than the length of either
	 *  the serial or model number strings.
	 */
	cam_strvis(disk->d_descr, cd->mn,
	    NVME_MODEL_NUMBER_LENGTH, sizeof(disk->d_descr));
	cam_strvis(disk->d_ident, cd->sn,
	    NVME_SERIAL_NUMBER_LENGTH, sizeof(disk->d_ident));
	disk->d_hba_vendor = cpi.hba_vendor;
	disk->d_hba_device = cpi.hba_device;
	disk->d_hba_subvendor = cpi.hba_subvendor;
	disk->d_hba_subdevice = cpi.hba_subdevice;
	snprintf(disk->d_attachment, sizeof(disk->d_attachment),
	    "%s%d", cpi.dev_name, cpi.unit_number);
	if (((nsd->nsfeat >> NVME_NS_DATA_NSFEAT_NPVALID_SHIFT) &
	    NVME_NS_DATA_NSFEAT_NPVALID_MASK) != 0 && nsd->npwg != 0)
		disk->d_stripesize = ((nsd->npwg + 1) * disk->d_sectorsize);
	else
		disk->d_stripesize = nsd->noiob * disk->d_sectorsize;
	disk->d_stripeoffset = 0;
	disk->d_devstat = devstat_new_entry(periph->periph_name,
	    periph->unit_number, disk->d_sectorsize,
	    DEVSTAT_ALL_SUPPORTED,
	    DEVSTAT_TYPE_DIRECT | XPORT_DEVSTAT_TYPE(cpi.transport),
	    DEVSTAT_PRIORITY_DISK);
	/*
	 * Add alias for older nvd drives to ease transition.
	 */
	if (nda_nvd_compat)
		disk_add_alias(disk, "nvd");

	/*
	 * Acquire a reference to the periph before we register with GEOM.
	 * We'll release this reference once GEOM calls us back (via
	 * ndadiskgonecb()) telling us that our provider has been freed.
	 */
	if (cam_periph_acquire(periph) != 0) {
		xpt_print(periph->path, "%s: lost periph during "
			  "registration!\n", __func__);
		cam_periph_lock(periph);
		return (CAM_REQ_CMP_ERR);
	}
	disk_create(softc->disk, DISK_VERSION);
	cam_periph_lock(periph);
	cam_periph_unhold(periph);

	snprintf(announce_buf, sizeof(announce_buf),
		"%juMB (%ju %u byte sectors)",
	    (uintmax_t)((uintmax_t)disk->d_mediasize / (1024*1024)),
		(uintmax_t)disk->d_mediasize / disk->d_sectorsize,
		disk->d_sectorsize);
	xpt_announce_periph(periph, announce_buf);
	xpt_announce_quirks(periph, softc->quirks, NDA_Q_BIT_STRING);

	/*
	 * Create our sysctl variables, now that we know
	 * we have successfully attached.
	 */
	if (cam_periph_acquire(periph) == 0)
		taskqueue_enqueue(taskqueue_thread, &softc->sysctl_task);

	/*
	 * Register for device going away and info about the drive
	 * changing (though with NVMe, it can't)
	 */
	xpt_register_async(AC_LOST_DEVICE | AC_ADVINFO_CHANGED,
	    ndaasync, periph, periph->path);

	softc->state = NDA_STATE_NORMAL;
	return(CAM_REQ_CMP);
}

static void
ndastart(struct cam_periph *periph, union ccb *start_ccb)
{
	struct nda_softc *softc = (struct nda_softc *)periph->softc;
	struct ccb_nvmeio *nvmeio = &start_ccb->nvmeio;

	CAM_DEBUG(periph->path, CAM_DEBUG_TRACE, ("ndastart\n"));

	switch (softc->state) {
	case NDA_STATE_NORMAL:
	{
		struct bio *bp;

		bp = cam_iosched_next_bio(softc->cam_iosched);
		CAM_DEBUG(periph->path, CAM_DEBUG_TRACE, ("ndastart: bio %p\n", bp));
		if (bp == NULL) {
			xpt_release_ccb(start_ccb);
			break;
		}

		switch (bp->bio_cmd) {
		case BIO_WRITE:
			softc->flags |= NDA_FLAG_DIRTY;
			/* FALLTHROUGH */
		case BIO_READ:
		{
#ifdef CAM_TEST_FAILURE
			int fail = 0;

			/*
			 * Support the failure ioctls.  If the command is a
			 * read, and there are pending forced read errors, or
			 * if a write and pending write errors, then fail this
			 * operation with EIO.  This is useful for testing
			 * purposes.  Also, support having every Nth read fail.
			 *
			 * This is a rather blunt tool.
			 */
			if (bp->bio_cmd == BIO_READ) {
				if (softc->force_read_error) {
					softc->force_read_error--;
					fail = 1;
				}
				if (softc->periodic_read_error > 0) {
					if (++softc->periodic_read_count >=
					    softc->periodic_read_error) {
						softc->periodic_read_count = 0;
						fail = 1;
					}
				}
			} else {
				if (softc->force_write_error) {
					softc->force_write_error--;
					fail = 1;
				}
			}
			if (fail) {
				biofinish(bp, NULL, EIO);
				xpt_release_ccb(start_ccb);
				ndaschedule(periph);
				return;
			}
#endif
			KASSERT((bp->bio_flags & BIO_UNMAPPED) == 0 ||
			    round_page(bp->bio_bcount + bp->bio_ma_offset) /
			    PAGE_SIZE == bp->bio_ma_n,
			    ("Short bio %p", bp));
			nda_nvme_rw_bio(softc, &start_ccb->nvmeio, bp, bp->bio_cmd == BIO_READ ?
			    NVME_OPC_READ : NVME_OPC_WRITE);
			break;
		}
		case BIO_DELETE:
		{
			struct nvme_dsm_range *dsm_range, *dsm_end;
			struct nda_trim_request *trim;
			struct bio *bp1;
			int ents;
			uint32_t totalcount = 0, ranges = 0;

			trim = malloc(sizeof(*trim), M_NVMEDA, M_ZERO | M_NOWAIT);
			if (trim == NULL) {
				biofinish(bp, NULL, ENOMEM);
				xpt_release_ccb(start_ccb);
				ndaschedule(periph);
				return;
			}
			TAILQ_INIT(&trim->bps);
			bp1 = bp;
			ents = min(nitems(trim->dsm), nda_max_trim_entries);
			ents = max(ents, 1);
			dsm_range = trim->dsm;
			dsm_end = dsm_range + ents;
			do {
				TAILQ_INSERT_TAIL(&trim->bps, bp1, bio_queue);
				dsm_range->length =
				    htole32(bp1->bio_bcount / softc->disk->d_sectorsize);
				dsm_range->starting_lba =
				    htole64(bp1->bio_offset / softc->disk->d_sectorsize);
				ranges++;
				totalcount += dsm_range->length;
				dsm_range++;
				if (dsm_range >= dsm_end)
					break;
				bp1 = cam_iosched_next_trim(softc->cam_iosched);
				/* XXX -- Could collapse adjacent ranges, but we don't for now */
				/* XXX -- Could limit based on total payload size */
			} while (bp1 != NULL);
			start_ccb->ccb_trim = trim;
			nda_nvme_trim(softc, &start_ccb->nvmeio, trim->dsm,
			    dsm_range - trim->dsm);
			start_ccb->ccb_state = NDA_CCB_TRIM;
			softc->trim_count++;
			softc->trim_ranges += ranges;
			softc->trim_lbas += totalcount;
			/*
			 * Note: We can have multiple TRIMs in flight, so we don't call
			 * cam_iosched_submit_trim(softc->cam_iosched);
			 * since that forces the I/O scheduler to only schedule one at a time.
			 * On NVMe drives, this is a performance disaster.
			 */
			goto out;
		}
		case BIO_FLUSH:
			nda_nvme_flush(softc, nvmeio);
			break;
		default:
			biofinish(bp, NULL, EOPNOTSUPP);
			xpt_release_ccb(start_ccb);
			ndaschedule(periph);
			return;
		}
		start_ccb->ccb_state = NDA_CCB_BUFFER_IO;
		start_ccb->ccb_bp = bp;
out:
		start_ccb->ccb_h.flags |= CAM_UNLOCKED;
		softc->outstanding_cmds++;
		softc->refcount++;			/* For submission only */
		cam_periph_unlock(periph);
		xpt_action(start_ccb);
		cam_periph_lock(periph);
		softc->refcount--;			/* Submission done */

		/* May have more work to do, so ensure we stay scheduled */
		ndaschedule(periph);
		break;
		}
	}
}

static void
ndadone(struct cam_periph *periph, union ccb *done_ccb)
{
	struct nda_softc *softc;
	struct ccb_nvmeio *nvmeio = &done_ccb->nvmeio;
	struct cam_path *path;
	int state;

	softc = (struct nda_softc *)periph->softc;
	path = done_ccb->ccb_h.path;

	CAM_DEBUG(path, CAM_DEBUG_TRACE, ("ndadone\n"));

	state = nvmeio->ccb_state & NDA_CCB_TYPE_MASK;
	switch (state) {
	case NDA_CCB_BUFFER_IO:
	case NDA_CCB_TRIM:
	{
		int error;

		cam_periph_lock(periph);
		if ((done_ccb->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) {
			error = ndaerror(done_ccb, 0, 0);
			if (error == ERESTART) {
				/* A retry was scheduled, so just return. */
				cam_periph_unlock(periph);
				return;
			}
			if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0)
				cam_release_devq(path,
						 /*relsim_flags*/0,
						 /*reduction*/0,
						 /*timeout*/0,
						 /*getcount_only*/0);
		} else {
			if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0)
				panic("REQ_CMP with QFRZN");
			error = 0;
		}
		if (state == NDA_CCB_BUFFER_IO) {
			struct bio *bp;

			bp = (struct bio *)done_ccb->ccb_bp;
			bp->bio_error = error;
			if (error != 0) {
				bp->bio_resid = bp->bio_bcount;
				bp->bio_flags |= BIO_ERROR;
			} else {
				bp->bio_resid = 0;
			}
			softc->outstanding_cmds--;

			/*
			 * We need to call cam_iosched before we call biodone so that we
			 * don't measure any activity that happens in the completion
			 * routine, which in the case of sendfile can be quite
			 * extensive.
			 */
			cam_iosched_bio_complete(softc->cam_iosched, bp, done_ccb);
			xpt_release_ccb(done_ccb);
			ndaschedule(periph);
			cam_periph_unlock(periph);
			biodone(bp);
		} else { /* state == NDA_CCB_TRIM */
			struct nda_trim_request *trim;
			struct bio *bp1, *bp2;
			TAILQ_HEAD(, bio) queue;

			trim = nvmeio->ccb_trim;
			TAILQ_INIT(&queue);
			TAILQ_CONCAT(&queue, &trim->bps, bio_queue);
			free(trim, M_NVMEDA);

			/*
			 * Since we can have multiple trims in flight, we don't
			 * need to call this here.
			 * cam_iosched_trim_done(softc->cam_iosched);
			 */
			/*
			 * The the I/O scheduler that we're finishing the I/O
			 * so we can keep book. The first one we pass in the CCB
			 * which has the timing information. The rest we pass in NULL
			 * so we can keep proper counts.
			 */
			bp1 = TAILQ_FIRST(&queue);
			cam_iosched_bio_complete(softc->cam_iosched, bp1, done_ccb);
			xpt_release_ccb(done_ccb);
			softc->outstanding_cmds--;
			ndaschedule(periph);
			cam_periph_unlock(periph);
			while ((bp2 = TAILQ_FIRST(&queue)) != NULL) {
				TAILQ_REMOVE(&queue, bp2, bio_queue);
				bp2->bio_error = error;
				if (error != 0) {
					bp2->bio_flags |= BIO_ERROR;
					bp2->bio_resid = bp1->bio_bcount;
				} else
					bp2->bio_resid = 0;
				if (bp1 != bp2)
					cam_iosched_bio_complete(softc->cam_iosched, bp2, NULL);
				biodone(bp2);
			}
		}
		return;
	}
	case NDA_CCB_DUMP:
		/* No-op.  We're polling */
		return;
	case NDA_CCB_PASS:
		return;
	default:
		break;
	}
	xpt_release_ccb(done_ccb);
}

static int
ndaerror(union ccb *ccb, u_int32_t cam_flags, u_int32_t sense_flags)
{
	struct nda_softc *softc;
	struct cam_periph *periph;

	periph = xpt_path_periph(ccb->ccb_h.path);
	softc = (struct nda_softc *)periph->softc;

	switch (ccb->ccb_h.status & CAM_STATUS_MASK) {
	case CAM_CMD_TIMEOUT:
#ifdef CAM_IO_STATS
		softc->timeouts++;
#endif
		break;
	case CAM_REQ_ABORTED:
	case CAM_REQ_CMP_ERR:
	case CAM_REQ_TERMIO:
	case CAM_UNREC_HBA_ERROR:
	case CAM_DATA_RUN_ERR:
	case CAM_ATA_STATUS_ERROR:
#ifdef CAM_IO_STATS
		softc->errors++;
#endif
		break;
	default:
		break;
	}

	return(cam_periph_error(ccb, cam_flags, sense_flags));
}

/*
 * Step through all NDA peripheral drivers, and if the device is still open,
 * sync the disk cache to physical media.
 */
static void
ndaflush(void)
{
	struct cam_periph *periph;
	struct nda_softc *softc;
	union ccb *ccb;
	int error;

	CAM_PERIPH_FOREACH(periph, &ndadriver) {
		softc = (struct nda_softc *)periph->softc;

		if (SCHEDULER_STOPPED()) {
			/*
			 * If we paniced with the lock held or the periph is not
			 * open, do not recurse.  Otherwise, call ndadump since
			 * that avoids the sleeping cam_periph_getccb does if no
			 * CCBs are available.
			 */
			if (!cam_periph_owned(periph) &&
			    (softc->flags & NDA_FLAG_OPEN)) {
				ndadump(softc->disk, NULL, 0, 0, 0);
			}
			continue;
		}

		/*
		 * We only sync the cache if the drive is still open
		 */
		cam_periph_lock(periph);
		if ((softc->flags & NDA_FLAG_OPEN) == 0) {
			cam_periph_unlock(periph);
			continue;
		}

		ccb = cam_periph_getccb(periph, CAM_PRIORITY_NORMAL);
		nda_nvme_flush(softc, &ccb->nvmeio);
		error = cam_periph_runccb(ccb, ndaerror, /*cam_flags*/0,
		    /*sense_flags*/ SF_NO_RECOVERY | SF_NO_RETRY,
		    softc->disk->d_devstat);
		if (error != 0)
			xpt_print(periph->path, "Synchronize cache failed\n");
		xpt_release_ccb(ccb);
		cam_periph_unlock(periph);
	}
}

static void
ndashutdown(void *arg, int howto)
{

	ndaflush();
}

static void
ndasuspend(void *arg)
{

	ndaflush();
}