xref: /freebsd/sys/geom/raid/md_jmicron.c (revision e32fecd0c2c3ee37c47ee100f169e7eb0282a873)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 2010 Alexander Motin <mav@FreeBSD.org>
5  * Copyright (c) 2000 - 2008 Søren Schmidt <sos@FreeBSD.org>
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  */
29 
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
32 
33 #include <sys/param.h>
34 #include <sys/bio.h>
35 #include <sys/endian.h>
36 #include <sys/kernel.h>
37 #include <sys/kobj.h>
38 #include <sys/limits.h>
39 #include <sys/lock.h>
40 #include <sys/malloc.h>
41 #include <sys/mutex.h>
42 #include <sys/systm.h>
43 #include <sys/taskqueue.h>
44 #include <geom/geom.h>
45 #include <geom/geom_dbg.h>
46 #include "geom/raid/g_raid.h"
47 #include "g_raid_md_if.h"
48 
49 static MALLOC_DEFINE(M_MD_JMICRON, "md_jmicron_data", "GEOM_RAID JMicron metadata");
50 
51 #define	JMICRON_MAX_DISKS	8
52 #define	JMICRON_MAX_SPARE	2
53 
54 struct jmicron_raid_conf {
55     uint8_t		signature[2];
56 #define	JMICRON_MAGIC		"JM"
57 
58     uint16_t		version;
59 #define	JMICRON_VERSION		0x0001
60 
61     uint16_t		checksum;
62     uint8_t		filler_1[10];
63     uint32_t		disk_id;
64     uint32_t		offset;
65     uint32_t		disk_sectors_high;
66     uint16_t		disk_sectors_low;
67     uint8_t		filler_2[2];
68     uint8_t		name[16];
69     uint8_t		type;
70 #define	JMICRON_T_RAID0		0
71 #define	JMICRON_T_RAID1		1
72 #define	JMICRON_T_RAID01	2
73 #define	JMICRON_T_CONCAT	3
74 #define	JMICRON_T_RAID5		5
75 
76     uint8_t		stripe_shift;
77     uint16_t		flags;
78 #define	JMICRON_F_READY		0x0001
79 #define	JMICRON_F_BOOTABLE	0x0002
80 #define	JMICRON_F_BADSEC	0x0004
81 #define	JMICRON_F_ACTIVE	0x0010
82 #define	JMICRON_F_UNSYNC	0x0020
83 #define	JMICRON_F_NEWEST	0x0040
84 
85     uint8_t		filler_3[4];
86     uint32_t		spare[JMICRON_MAX_SPARE];
87     uint32_t		disks[JMICRON_MAX_DISKS];
88 #define	JMICRON_DISK_MASK	0xFFFFFFF0
89 #define	JMICRON_SEG_MASK	0x0000000F
90     uint8_t		filler_4[32];
91     uint8_t		filler_5[384];
92 };
93 
94 struct g_raid_md_jmicron_perdisk {
95 	struct jmicron_raid_conf	*pd_meta;
96 	int				 pd_disk_pos;
97 	int				 pd_disk_id;
98 	off_t				 pd_disk_size;
99 };
100 
101 struct g_raid_md_jmicron_object {
102 	struct g_raid_md_object	 mdio_base;
103 	uint32_t		 mdio_config_id;
104 	struct jmicron_raid_conf	*mdio_meta;
105 	struct callout		 mdio_start_co;	/* STARTING state timer. */
106 	int			 mdio_total_disks;
107 	int			 mdio_disks_present;
108 	int			 mdio_started;
109 	int			 mdio_incomplete;
110 	struct root_hold_token	*mdio_rootmount; /* Root mount delay token. */
111 };
112 
113 static g_raid_md_create_t g_raid_md_create_jmicron;
114 static g_raid_md_taste_t g_raid_md_taste_jmicron;
115 static g_raid_md_event_t g_raid_md_event_jmicron;
116 static g_raid_md_ctl_t g_raid_md_ctl_jmicron;
117 static g_raid_md_write_t g_raid_md_write_jmicron;
118 static g_raid_md_fail_disk_t g_raid_md_fail_disk_jmicron;
119 static g_raid_md_free_disk_t g_raid_md_free_disk_jmicron;
120 static g_raid_md_free_t g_raid_md_free_jmicron;
121 
122 static kobj_method_t g_raid_md_jmicron_methods[] = {
123 	KOBJMETHOD(g_raid_md_create,	g_raid_md_create_jmicron),
124 	KOBJMETHOD(g_raid_md_taste,	g_raid_md_taste_jmicron),
125 	KOBJMETHOD(g_raid_md_event,	g_raid_md_event_jmicron),
126 	KOBJMETHOD(g_raid_md_ctl,	g_raid_md_ctl_jmicron),
127 	KOBJMETHOD(g_raid_md_write,	g_raid_md_write_jmicron),
128 	KOBJMETHOD(g_raid_md_fail_disk,	g_raid_md_fail_disk_jmicron),
129 	KOBJMETHOD(g_raid_md_free_disk,	g_raid_md_free_disk_jmicron),
130 	KOBJMETHOD(g_raid_md_free,	g_raid_md_free_jmicron),
131 	{ 0, 0 }
132 };
133 
134 static struct g_raid_md_class g_raid_md_jmicron_class = {
135 	"JMicron",
136 	g_raid_md_jmicron_methods,
137 	sizeof(struct g_raid_md_jmicron_object),
138 	.mdc_enable = 1,
139 	.mdc_priority = 100
140 };
141 
142 static void
143 g_raid_md_jmicron_print(struct jmicron_raid_conf *meta)
144 {
145 	int k;
146 
147 	if (g_raid_debug < 1)
148 		return;
149 
150 	printf("********* ATA JMicron RAID Metadata *********\n");
151 	printf("signature           <%c%c>\n", meta->signature[0], meta->signature[1]);
152 	printf("version             %04x\n", meta->version);
153 	printf("checksum            0x%04x\n", meta->checksum);
154 	printf("disk_id             0x%08x\n", meta->disk_id);
155 	printf("offset              0x%08x\n", meta->offset);
156 	printf("disk_sectors_high   0x%08x\n", meta->disk_sectors_high);
157 	printf("disk_sectors_low    0x%04x\n", meta->disk_sectors_low);
158 	printf("name                <%.16s>\n", meta->name);
159 	printf("type                %d\n", meta->type);
160 	printf("stripe_shift        %d\n", meta->stripe_shift);
161 	printf("flags               %04x\n", meta->flags);
162 	printf("spare              ");
163 	for (k = 0; k < JMICRON_MAX_SPARE; k++)
164 		printf(" 0x%08x", meta->spare[k]);
165 	printf("\n");
166 	printf("disks              ");
167 	for (k = 0; k < JMICRON_MAX_DISKS; k++)
168 		printf(" 0x%08x", meta->disks[k]);
169 	printf("\n");
170 	printf("=================================================\n");
171 }
172 
173 static struct jmicron_raid_conf *
174 jmicron_meta_copy(struct jmicron_raid_conf *meta)
175 {
176 	struct jmicron_raid_conf *nmeta;
177 
178 	nmeta = malloc(sizeof(*meta), M_MD_JMICRON, M_WAITOK);
179 	memcpy(nmeta, meta, sizeof(*meta));
180 	return (nmeta);
181 }
182 
183 static int
184 jmicron_meta_total_disks(struct jmicron_raid_conf *meta)
185 {
186 	int pos;
187 
188 	for (pos = 0; pos < JMICRON_MAX_DISKS; pos++) {
189 		if (meta->disks[pos] == 0)
190 			break;
191 	}
192 	return (pos);
193 }
194 
195 static int
196 jmicron_meta_total_spare(struct jmicron_raid_conf *meta)
197 {
198 	int pos, n;
199 
200 	n = 0;
201 	for (pos = 0; pos < JMICRON_MAX_SPARE; pos++) {
202 		if (meta->spare[pos] != 0)
203 			n++;
204 	}
205 	return (n);
206 }
207 
208 /*
209  * Generate fake Configuration ID based on disk IDs.
210  * Note: it will change after each disk set change.
211  */
212 static uint32_t
213 jmicron_meta_config_id(struct jmicron_raid_conf *meta)
214 {
215 	int pos;
216 	uint32_t config_id;
217 
218 	config_id = 0;
219 	for (pos = 0; pos < JMICRON_MAX_DISKS; pos++)
220 		config_id += meta->disks[pos] << pos;
221 	return (config_id);
222 }
223 
224 static void
225 jmicron_meta_get_name(struct jmicron_raid_conf *meta, char *buf)
226 {
227 	int i;
228 
229 	strncpy(buf, meta->name, 16);
230 	buf[16] = 0;
231 	for (i = 15; i >= 0; i--) {
232 		if (buf[i] > 0x20)
233 			break;
234 		buf[i] = 0;
235 	}
236 }
237 
238 static void
239 jmicron_meta_put_name(struct jmicron_raid_conf *meta, char *buf)
240 {
241 
242 	memset(meta->name, 0x20, 16);
243 	memcpy(meta->name, buf, MIN(strlen(buf), 16));
244 }
245 
246 static int
247 jmicron_meta_find_disk(struct jmicron_raid_conf *meta, uint32_t id)
248 {
249 	int pos;
250 
251 	id &= JMICRON_DISK_MASK;
252 	for (pos = 0; pos < JMICRON_MAX_DISKS; pos++) {
253 		if ((meta->disks[pos] & JMICRON_DISK_MASK) == id)
254 			return (pos);
255 	}
256 	for (pos = 0; pos < JMICRON_MAX_SPARE; pos++) {
257 		if ((meta->spare[pos] & JMICRON_DISK_MASK) == id)
258 			return (-3);
259 	}
260 	return (-1);
261 }
262 
263 static struct jmicron_raid_conf *
264 jmicron_meta_read(struct g_consumer *cp)
265 {
266 	struct g_provider *pp;
267 	struct jmicron_raid_conf *meta;
268 	char *buf;
269 	int error, i;
270 	uint16_t checksum, *ptr;
271 
272 	pp = cp->provider;
273 	if (pp->sectorsize < sizeof(*meta))
274 		return (NULL);
275 	/* Read the anchor sector. */
276 	buf = g_read_data(cp,
277 	    pp->mediasize - pp->sectorsize, pp->sectorsize, &error);
278 	if (buf == NULL) {
279 		G_RAID_DEBUG(1, "Cannot read metadata from %s (error=%d).",
280 		    pp->name, error);
281 		return (NULL);
282 	}
283 	meta = (struct jmicron_raid_conf *)buf;
284 
285 	/* Check if this is an JMicron RAID struct */
286 	if (strncmp(meta->signature, JMICRON_MAGIC, strlen(JMICRON_MAGIC))) {
287 		G_RAID_DEBUG(1, "JMicron signature check failed on %s", pp->name);
288 		g_free(buf);
289 		return (NULL);
290 	}
291 	meta = malloc(sizeof(*meta), M_MD_JMICRON, M_WAITOK);
292 	memcpy(meta, buf, min(sizeof(*meta), pp->sectorsize));
293 	g_free(buf);
294 
295 	/* Check metadata checksum. */
296 	for (checksum = 0, ptr = (uint16_t *)meta, i = 0; i < 64; i++)
297 		checksum += *ptr++;
298 	if (checksum != 0) {
299 		G_RAID_DEBUG(1, "JMicron checksum check failed on %s", pp->name);
300 		free(meta, M_MD_JMICRON);
301 		return (NULL);
302 	}
303 
304 	return (meta);
305 }
306 
307 static int
308 jmicron_meta_write(struct g_consumer *cp, struct jmicron_raid_conf *meta)
309 {
310 	struct g_provider *pp;
311 	char *buf;
312 	int error, i;
313 	uint16_t checksum, *ptr;
314 
315 	pp = cp->provider;
316 
317 	/* Recalculate checksum for case if metadata were changed. */
318 	meta->checksum = 0;
319 	for (checksum = 0, ptr = (uint16_t *)meta, i = 0; i < 64; i++)
320 		checksum += *ptr++;
321 	meta->checksum -= checksum;
322 
323 	/* Create and fill buffer. */
324 	buf = malloc(pp->sectorsize, M_MD_JMICRON, M_WAITOK | M_ZERO);
325 	memcpy(buf, meta, sizeof(*meta));
326 
327 	error = g_write_data(cp,
328 	    pp->mediasize - pp->sectorsize, buf, pp->sectorsize);
329 	if (error != 0) {
330 		G_RAID_DEBUG(1, "Cannot write metadata to %s (error=%d).",
331 		    pp->name, error);
332 	}
333 
334 	free(buf, M_MD_JMICRON);
335 	return (error);
336 }
337 
338 static int
339 jmicron_meta_erase(struct g_consumer *cp)
340 {
341 	struct g_provider *pp;
342 	char *buf;
343 	int error;
344 
345 	pp = cp->provider;
346 	buf = malloc(pp->sectorsize, M_MD_JMICRON, M_WAITOK | M_ZERO);
347 	error = g_write_data(cp,
348 	    pp->mediasize - pp->sectorsize, buf, pp->sectorsize);
349 	if (error != 0) {
350 		G_RAID_DEBUG(1, "Cannot erase metadata on %s (error=%d).",
351 		    pp->name, error);
352 	}
353 	free(buf, M_MD_JMICRON);
354 	return (error);
355 }
356 
357 static struct g_raid_disk *
358 g_raid_md_jmicron_get_disk(struct g_raid_softc *sc, int id)
359 {
360 	struct g_raid_disk	*disk;
361 	struct g_raid_md_jmicron_perdisk *pd;
362 
363 	TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
364 		pd = (struct g_raid_md_jmicron_perdisk *)disk->d_md_data;
365 		if (pd->pd_disk_pos == id)
366 			break;
367 	}
368 	return (disk);
369 }
370 
371 static int
372 g_raid_md_jmicron_supported(int level, int qual, int disks, int force)
373 {
374 
375 	if (disks > 8)
376 		return (0);
377 	switch (level) {
378 	case G_RAID_VOLUME_RL_RAID0:
379 		if (disks < 1)
380 			return (0);
381 		if (!force && (disks < 2 || disks > 6))
382 			return (0);
383 		break;
384 	case G_RAID_VOLUME_RL_RAID1:
385 		if (disks < 1)
386 			return (0);
387 		if (!force && (disks != 2))
388 			return (0);
389 		break;
390 	case G_RAID_VOLUME_RL_RAID1E:
391 		if (disks < 2)
392 			return (0);
393 		if (!force && (disks != 4))
394 			return (0);
395 		break;
396 	case G_RAID_VOLUME_RL_SINGLE:
397 		if (disks != 1)
398 			return (0);
399 		if (!force)
400 			return (0);
401 		break;
402 	case G_RAID_VOLUME_RL_CONCAT:
403 		if (disks < 2)
404 			return (0);
405 		break;
406 	case G_RAID_VOLUME_RL_RAID5:
407 		if (disks < 3)
408 			return (0);
409 		if (qual != G_RAID_VOLUME_RLQ_R5LA)
410 			return (0);
411 		if (!force)
412 			return (0);
413 		break;
414 	default:
415 		return (0);
416 	}
417 	if (level != G_RAID_VOLUME_RL_RAID5 && qual != G_RAID_VOLUME_RLQ_NONE)
418 		return (0);
419 	return (1);
420 }
421 
422 static int
423 g_raid_md_jmicron_start_disk(struct g_raid_disk *disk)
424 {
425 	struct g_raid_softc *sc;
426 	struct g_raid_subdisk *sd, *tmpsd;
427 	struct g_raid_disk *olddisk, *tmpdisk;
428 	struct g_raid_md_object *md;
429 	struct g_raid_md_jmicron_object *mdi;
430 	struct g_raid_md_jmicron_perdisk *pd, *oldpd;
431 	struct jmicron_raid_conf *meta;
432 	int disk_pos, resurrection = 0;
433 
434 	sc = disk->d_softc;
435 	md = sc->sc_md;
436 	mdi = (struct g_raid_md_jmicron_object *)md;
437 	meta = mdi->mdio_meta;
438 	pd = (struct g_raid_md_jmicron_perdisk *)disk->d_md_data;
439 	olddisk = NULL;
440 
441 	/* Find disk position in metadata by its serial. */
442 	if (pd->pd_meta != NULL)
443 		disk_pos = jmicron_meta_find_disk(meta, pd->pd_disk_id);
444 	else
445 		disk_pos = -1;
446 	if (disk_pos < 0) {
447 		G_RAID_DEBUG1(1, sc, "Unknown, probably new or stale disk");
448 		/* If we are in the start process, that's all for now. */
449 		if (!mdi->mdio_started)
450 			goto nofit;
451 		/*
452 		 * If we have already started - try to get use of the disk.
453 		 * Try to replace OFFLINE disks first, then FAILED.
454 		 */
455 		TAILQ_FOREACH(tmpdisk, &sc->sc_disks, d_next) {
456 			if (tmpdisk->d_state != G_RAID_DISK_S_OFFLINE &&
457 			    tmpdisk->d_state != G_RAID_DISK_S_FAILED)
458 				continue;
459 			/* Make sure this disk is big enough. */
460 			TAILQ_FOREACH(sd, &tmpdisk->d_subdisks, sd_next) {
461 				if (sd->sd_offset + sd->sd_size + 512 >
462 				    pd->pd_disk_size) {
463 					G_RAID_DEBUG1(1, sc,
464 					    "Disk too small (%ju < %ju)",
465 					    pd->pd_disk_size,
466 					    sd->sd_offset + sd->sd_size + 512);
467 					break;
468 				}
469 			}
470 			if (sd != NULL)
471 				continue;
472 			if (tmpdisk->d_state == G_RAID_DISK_S_OFFLINE) {
473 				olddisk = tmpdisk;
474 				break;
475 			} else if (olddisk == NULL)
476 				olddisk = tmpdisk;
477 		}
478 		if (olddisk == NULL) {
479 nofit:
480 			if (disk_pos == -3 || pd->pd_disk_pos == -3) {
481 				g_raid_change_disk_state(disk,
482 				    G_RAID_DISK_S_SPARE);
483 				return (1);
484 			} else {
485 				g_raid_change_disk_state(disk,
486 				    G_RAID_DISK_S_STALE);
487 				return (0);
488 			}
489 		}
490 		oldpd = (struct g_raid_md_jmicron_perdisk *)olddisk->d_md_data;
491 		disk_pos = oldpd->pd_disk_pos;
492 		resurrection = 1;
493 	}
494 
495 	if (olddisk == NULL) {
496 		/* Find placeholder by position. */
497 		olddisk = g_raid_md_jmicron_get_disk(sc, disk_pos);
498 		if (olddisk == NULL)
499 			panic("No disk at position %d!", disk_pos);
500 		if (olddisk->d_state != G_RAID_DISK_S_OFFLINE) {
501 			G_RAID_DEBUG1(1, sc, "More than one disk for pos %d",
502 			    disk_pos);
503 			g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE);
504 			return (0);
505 		}
506 		oldpd = (struct g_raid_md_jmicron_perdisk *)olddisk->d_md_data;
507 	}
508 
509 	/* Replace failed disk or placeholder with new disk. */
510 	TAILQ_FOREACH_SAFE(sd, &olddisk->d_subdisks, sd_next, tmpsd) {
511 		TAILQ_REMOVE(&olddisk->d_subdisks, sd, sd_next);
512 		TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
513 		sd->sd_disk = disk;
514 	}
515 	oldpd->pd_disk_pos = -2;
516 	pd->pd_disk_pos = disk_pos;
517 	/* Update global metadata just in case. */
518 	meta->disks[disk_pos] = pd->pd_disk_id;
519 
520 	/* If it was placeholder -- destroy it. */
521 	if (olddisk->d_state == G_RAID_DISK_S_OFFLINE) {
522 		g_raid_destroy_disk(olddisk);
523 	} else {
524 		/* Otherwise, make it STALE_FAILED. */
525 		g_raid_change_disk_state(olddisk, G_RAID_DISK_S_STALE_FAILED);
526 	}
527 
528 	/* Welcome the new disk. */
529 	g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
530 	TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
531 		/*
532 		 * Different disks may have different sizes/offsets,
533 		 * especially in concat mode. Update.
534 		 */
535 		if (!resurrection) {
536 			sd->sd_offset =
537 			    (off_t)pd->pd_meta->offset * 16 * 512; //ZZZ
538 			sd->sd_size =
539 			    (((off_t)pd->pd_meta->disk_sectors_high << 16) +
540 			      pd->pd_meta->disk_sectors_low) * 512;
541 		}
542 
543 		if (resurrection) {
544 			/* Stale disk, almost same as new. */
545 			g_raid_change_subdisk_state(sd,
546 			    G_RAID_SUBDISK_S_NEW);
547 		} else if ((meta->flags & JMICRON_F_BADSEC) != 0 &&
548 		    (pd->pd_meta->flags & JMICRON_F_BADSEC) == 0) {
549 			/* Cold-inserted or rebuilding disk. */
550 			g_raid_change_subdisk_state(sd,
551 			    G_RAID_SUBDISK_S_NEW);
552 		} else if (pd->pd_meta->flags & JMICRON_F_UNSYNC) {
553 			/* Dirty or resyncing disk.. */
554 			g_raid_change_subdisk_state(sd,
555 			    G_RAID_SUBDISK_S_STALE);
556 		} else {
557 			/* Up to date disk. */
558 			g_raid_change_subdisk_state(sd,
559 			    G_RAID_SUBDISK_S_ACTIVE);
560 		}
561 		g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
562 		    G_RAID_EVENT_SUBDISK);
563 	}
564 
565 	/* Update status of our need for spare. */
566 	if (mdi->mdio_started) {
567 		mdi->mdio_incomplete =
568 		    (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) <
569 		     mdi->mdio_total_disks);
570 	}
571 
572 	return (resurrection);
573 }
574 
575 static void
576 g_disk_md_jmicron_retaste(void *arg, int pending)
577 {
578 
579 	G_RAID_DEBUG(1, "Array is not complete, trying to retaste.");
580 	g_retaste(&g_raid_class);
581 	free(arg, M_MD_JMICRON);
582 }
583 
584 static void
585 g_raid_md_jmicron_refill(struct g_raid_softc *sc)
586 {
587 	struct g_raid_md_object *md;
588 	struct g_raid_md_jmicron_object *mdi;
589 	struct g_raid_disk *disk;
590 	struct task *task;
591 	int update, na;
592 
593 	md = sc->sc_md;
594 	mdi = (struct g_raid_md_jmicron_object *)md;
595 	update = 0;
596 	do {
597 		/* Make sure we miss anything. */
598 		na = g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE);
599 		if (na == mdi->mdio_total_disks)
600 			break;
601 
602 		G_RAID_DEBUG1(1, md->mdo_softc,
603 		    "Array is not complete (%d of %d), "
604 		    "trying to refill.", na, mdi->mdio_total_disks);
605 
606 		/* Try to get use some of STALE disks. */
607 		TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
608 			if (disk->d_state == G_RAID_DISK_S_STALE) {
609 				update += g_raid_md_jmicron_start_disk(disk);
610 				if (disk->d_state == G_RAID_DISK_S_ACTIVE)
611 					break;
612 			}
613 		}
614 		if (disk != NULL)
615 			continue;
616 
617 		/* Try to get use some of SPARE disks. */
618 		TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
619 			if (disk->d_state == G_RAID_DISK_S_SPARE) {
620 				update += g_raid_md_jmicron_start_disk(disk);
621 				if (disk->d_state == G_RAID_DISK_S_ACTIVE)
622 					break;
623 			}
624 		}
625 	} while (disk != NULL);
626 
627 	/* Write new metadata if we changed something. */
628 	if (update)
629 		g_raid_md_write_jmicron(md, NULL, NULL, NULL);
630 
631 	/* Update status of our need for spare. */
632 	mdi->mdio_incomplete = (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) <
633 	    mdi->mdio_total_disks);
634 
635 	/* Request retaste hoping to find spare. */
636 	if (mdi->mdio_incomplete) {
637 		task = malloc(sizeof(struct task),
638 		    M_MD_JMICRON, M_WAITOK | M_ZERO);
639 		TASK_INIT(task, 0, g_disk_md_jmicron_retaste, task);
640 		taskqueue_enqueue(taskqueue_swi, task);
641 	}
642 }
643 
644 static void
645 g_raid_md_jmicron_start(struct g_raid_softc *sc)
646 {
647 	struct g_raid_md_object *md;
648 	struct g_raid_md_jmicron_object *mdi;
649 	struct g_raid_md_jmicron_perdisk *pd;
650 	struct jmicron_raid_conf *meta;
651 	struct g_raid_volume *vol;
652 	struct g_raid_subdisk *sd;
653 	struct g_raid_disk *disk;
654 	off_t size;
655 	int j, disk_pos;
656 	char buf[17];
657 
658 	md = sc->sc_md;
659 	mdi = (struct g_raid_md_jmicron_object *)md;
660 	meta = mdi->mdio_meta;
661 
662 	/* Create volumes and subdisks. */
663 	jmicron_meta_get_name(meta, buf);
664 	vol = g_raid_create_volume(sc, buf, -1);
665 	size = ((off_t)meta->disk_sectors_high << 16) + meta->disk_sectors_low;
666 	size *= 512; //ZZZ
667 	vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE;
668 	if (meta->type == JMICRON_T_RAID0) {
669 		vol->v_raid_level = G_RAID_VOLUME_RL_RAID0;
670 		vol->v_mediasize = size * mdi->mdio_total_disks;
671 	} else if (meta->type == JMICRON_T_RAID1) {
672 		vol->v_raid_level = G_RAID_VOLUME_RL_RAID1;
673 		vol->v_mediasize = size;
674 	} else if (meta->type == JMICRON_T_RAID01) {
675 		vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E;
676 		vol->v_mediasize = size * mdi->mdio_total_disks / 2;
677 	} else if (meta->type == JMICRON_T_CONCAT) {
678 		if (mdi->mdio_total_disks == 1)
679 			vol->v_raid_level = G_RAID_VOLUME_RL_SINGLE;
680 		else
681 			vol->v_raid_level = G_RAID_VOLUME_RL_CONCAT;
682 		vol->v_mediasize = 0;
683 	} else if (meta->type == JMICRON_T_RAID5) {
684 		vol->v_raid_level = G_RAID_VOLUME_RL_RAID5;
685 		vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_R5LA;
686 		vol->v_mediasize = size * (mdi->mdio_total_disks - 1);
687 	} else {
688 		vol->v_raid_level = G_RAID_VOLUME_RL_UNKNOWN;
689 		vol->v_mediasize = 0;
690 	}
691 	vol->v_strip_size = 1024 << meta->stripe_shift; //ZZZ
692 	vol->v_disks_count = mdi->mdio_total_disks;
693 	vol->v_sectorsize = 512; //ZZZ
694 	for (j = 0; j < vol->v_disks_count; j++) {
695 		sd = &vol->v_subdisks[j];
696 		sd->sd_offset = (off_t)meta->offset * 16 * 512; //ZZZ
697 		sd->sd_size = size;
698 	}
699 	g_raid_start_volume(vol);
700 
701 	/* Create disk placeholders to store data for later writing. */
702 	for (disk_pos = 0; disk_pos < mdi->mdio_total_disks; disk_pos++) {
703 		pd = malloc(sizeof(*pd), M_MD_JMICRON, M_WAITOK | M_ZERO);
704 		pd->pd_disk_pos = disk_pos;
705 		pd->pd_disk_id = meta->disks[disk_pos];
706 		disk = g_raid_create_disk(sc);
707 		disk->d_md_data = (void *)pd;
708 		disk->d_state = G_RAID_DISK_S_OFFLINE;
709 		sd = &vol->v_subdisks[disk_pos];
710 		sd->sd_disk = disk;
711 		TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
712 	}
713 
714 	/* Make all disks found till the moment take their places. */
715 	do {
716 		TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
717 			if (disk->d_state == G_RAID_DISK_S_NONE) {
718 				g_raid_md_jmicron_start_disk(disk);
719 				break;
720 			}
721 		}
722 	} while (disk != NULL);
723 
724 	mdi->mdio_started = 1;
725 	G_RAID_DEBUG1(0, sc, "Array started.");
726 	g_raid_md_write_jmicron(md, NULL, NULL, NULL);
727 
728 	/* Pickup any STALE/SPARE disks to refill array if needed. */
729 	g_raid_md_jmicron_refill(sc);
730 
731 	g_raid_event_send(vol, G_RAID_VOLUME_E_START, G_RAID_EVENT_VOLUME);
732 
733 	callout_stop(&mdi->mdio_start_co);
734 	G_RAID_DEBUG1(1, sc, "root_mount_rel %p", mdi->mdio_rootmount);
735 	root_mount_rel(mdi->mdio_rootmount);
736 	mdi->mdio_rootmount = NULL;
737 }
738 
739 static void
740 g_raid_md_jmicron_new_disk(struct g_raid_disk *disk)
741 {
742 	struct g_raid_softc *sc;
743 	struct g_raid_md_object *md;
744 	struct g_raid_md_jmicron_object *mdi;
745 	struct jmicron_raid_conf *pdmeta;
746 	struct g_raid_md_jmicron_perdisk *pd;
747 
748 	sc = disk->d_softc;
749 	md = sc->sc_md;
750 	mdi = (struct g_raid_md_jmicron_object *)md;
751 	pd = (struct g_raid_md_jmicron_perdisk *)disk->d_md_data;
752 	pdmeta = pd->pd_meta;
753 
754 	if (mdi->mdio_started) {
755 		if (g_raid_md_jmicron_start_disk(disk))
756 			g_raid_md_write_jmicron(md, NULL, NULL, NULL);
757 	} else {
758 		/*
759 		 * If we haven't started yet - update common metadata
760 		 * to get subdisks details, avoiding data from spare disks.
761 		 */
762 		if (mdi->mdio_meta == NULL ||
763 		    jmicron_meta_find_disk(mdi->mdio_meta,
764 		     mdi->mdio_meta->disk_id) == -3) {
765 			if (mdi->mdio_meta != NULL)
766 				free(mdi->mdio_meta, M_MD_JMICRON);
767 			mdi->mdio_meta = jmicron_meta_copy(pdmeta);
768 			mdi->mdio_total_disks = jmicron_meta_total_disks(pdmeta);
769 		}
770 		mdi->mdio_meta->flags |= pdmeta->flags & JMICRON_F_BADSEC;
771 
772 		mdi->mdio_disks_present++;
773 		G_RAID_DEBUG1(1, sc, "Matching disk (%d of %d+%d up)",
774 		    mdi->mdio_disks_present,
775 		    mdi->mdio_total_disks,
776 		    jmicron_meta_total_spare(mdi->mdio_meta));
777 
778 		/* If we collected all needed disks - start array. */
779 		if (mdi->mdio_disks_present == mdi->mdio_total_disks +
780 		    jmicron_meta_total_spare(mdi->mdio_meta))
781 			g_raid_md_jmicron_start(sc);
782 	}
783 }
784 
785 static void
786 g_raid_jmicron_go(void *arg)
787 {
788 	struct g_raid_softc *sc;
789 	struct g_raid_md_object *md;
790 	struct g_raid_md_jmicron_object *mdi;
791 
792 	sc = arg;
793 	md = sc->sc_md;
794 	mdi = (struct g_raid_md_jmicron_object *)md;
795 	if (!mdi->mdio_started) {
796 		G_RAID_DEBUG1(0, sc, "Force array start due to timeout.");
797 		g_raid_event_send(sc, G_RAID_NODE_E_START, 0);
798 	}
799 }
800 
801 static int
802 g_raid_md_create_jmicron(struct g_raid_md_object *md, struct g_class *mp,
803     struct g_geom **gp)
804 {
805 	struct g_raid_softc *sc;
806 	struct g_raid_md_jmicron_object *mdi;
807 	char name[16];
808 
809 	mdi = (struct g_raid_md_jmicron_object *)md;
810 	mdi->mdio_config_id = arc4random();
811 	snprintf(name, sizeof(name), "JMicron-%08x", mdi->mdio_config_id);
812 	sc = g_raid_create_node(mp, name, md);
813 	if (sc == NULL)
814 		return (G_RAID_MD_TASTE_FAIL);
815 	md->mdo_softc = sc;
816 	*gp = sc->sc_geom;
817 	return (G_RAID_MD_TASTE_NEW);
818 }
819 
820 static int
821 g_raid_md_taste_jmicron(struct g_raid_md_object *md, struct g_class *mp,
822                               struct g_consumer *cp, struct g_geom **gp)
823 {
824 	struct g_consumer *rcp;
825 	struct g_provider *pp;
826 	struct g_raid_md_jmicron_object *mdi, *mdi1;
827 	struct g_raid_softc *sc;
828 	struct g_raid_disk *disk;
829 	struct jmicron_raid_conf *meta;
830 	struct g_raid_md_jmicron_perdisk *pd;
831 	struct g_geom *geom;
832 	int disk_pos, result, spare, len;
833 	char name[16];
834 	uint16_t vendor;
835 
836 	G_RAID_DEBUG(1, "Tasting JMicron on %s", cp->provider->name);
837 	mdi = (struct g_raid_md_jmicron_object *)md;
838 	pp = cp->provider;
839 
840 	/* Read metadata from device. */
841 	meta = NULL;
842 	g_topology_unlock();
843 	vendor = 0xffff;
844 	len = sizeof(vendor);
845 	if (pp->geom->rank == 1)
846 		g_io_getattr("GEOM::hba_vendor", cp, &len, &vendor);
847 	meta = jmicron_meta_read(cp);
848 	g_topology_lock();
849 	if (meta == NULL) {
850 		if (g_raid_aggressive_spare) {
851 			if (vendor == 0x197b) {
852 				G_RAID_DEBUG(1,
853 				    "No JMicron metadata, forcing spare.");
854 				spare = 2;
855 				goto search;
856 			} else {
857 				G_RAID_DEBUG(1,
858 				    "JMicron vendor mismatch 0x%04x != 0x197b",
859 				    vendor);
860 			}
861 		}
862 		return (G_RAID_MD_TASTE_FAIL);
863 	}
864 
865 	/* Check this disk position in obtained metadata. */
866 	disk_pos = jmicron_meta_find_disk(meta, meta->disk_id);
867 	if (disk_pos == -1) {
868 		G_RAID_DEBUG(1, "JMicron disk_id %08x not found",
869 		    meta->disk_id);
870 		goto fail1;
871 	}
872 
873 	/* Metadata valid. Print it. */
874 	g_raid_md_jmicron_print(meta);
875 	G_RAID_DEBUG(1, "JMicron disk position %d", disk_pos);
876 	spare = (disk_pos == -2) ? 1 : 0;
877 
878 search:
879 	/* Search for matching node. */
880 	sc = NULL;
881 	mdi1 = NULL;
882 	LIST_FOREACH(geom, &mp->geom, geom) {
883 		sc = geom->softc;
884 		if (sc == NULL)
885 			continue;
886 		if (sc->sc_stopping != 0)
887 			continue;
888 		if (sc->sc_md->mdo_class != md->mdo_class)
889 			continue;
890 		mdi1 = (struct g_raid_md_jmicron_object *)sc->sc_md;
891 		if (spare == 2) {
892 			if (mdi1->mdio_incomplete)
893 				break;
894 		} else {
895 			if (mdi1->mdio_config_id ==
896 			    jmicron_meta_config_id(meta))
897 				break;
898 		}
899 	}
900 
901 	/* Found matching node. */
902 	if (geom != NULL) {
903 		G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name);
904 		result = G_RAID_MD_TASTE_EXISTING;
905 
906 	} else if (spare) { /* Not found needy node -- left for later. */
907 		G_RAID_DEBUG(1, "Spare is not needed at this time");
908 		goto fail1;
909 
910 	} else { /* Not found matching node -- create one. */
911 		result = G_RAID_MD_TASTE_NEW;
912 		mdi->mdio_config_id = jmicron_meta_config_id(meta);
913 		snprintf(name, sizeof(name), "JMicron-%08x",
914 		    mdi->mdio_config_id);
915 		sc = g_raid_create_node(mp, name, md);
916 		md->mdo_softc = sc;
917 		geom = sc->sc_geom;
918 		callout_init(&mdi->mdio_start_co, 1);
919 		callout_reset(&mdi->mdio_start_co, g_raid_start_timeout * hz,
920 		    g_raid_jmicron_go, sc);
921 		mdi->mdio_rootmount = root_mount_hold("GRAID-JMicron");
922 		G_RAID_DEBUG1(1, sc, "root_mount_hold %p", mdi->mdio_rootmount);
923 	}
924 
925 	/* There is no return after this point, so we close passed consumer. */
926 	g_access(cp, -1, 0, 0);
927 
928 	rcp = g_new_consumer(geom);
929 	rcp->flags |= G_CF_DIRECT_RECEIVE;
930 	g_attach(rcp, pp);
931 	if (g_access(rcp, 1, 1, 1) != 0)
932 		; //goto fail1;
933 
934 	g_topology_unlock();
935 	sx_xlock(&sc->sc_lock);
936 
937 	pd = malloc(sizeof(*pd), M_MD_JMICRON, M_WAITOK | M_ZERO);
938 	pd->pd_meta = meta;
939 	if (spare == 2) {
940 		pd->pd_disk_pos = -3;
941 		pd->pd_disk_id = arc4random() & JMICRON_DISK_MASK;
942 	} else {
943 		pd->pd_disk_pos = -1;
944 		pd->pd_disk_id = meta->disk_id;
945 	}
946 	pd->pd_disk_size = pp->mediasize;
947 	disk = g_raid_create_disk(sc);
948 	disk->d_md_data = (void *)pd;
949 	disk->d_consumer = rcp;
950 	rcp->private = disk;
951 
952 	g_raid_get_disk_info(disk);
953 
954 	g_raid_md_jmicron_new_disk(disk);
955 
956 	sx_xunlock(&sc->sc_lock);
957 	g_topology_lock();
958 	*gp = geom;
959 	return (result);
960 fail1:
961 	free(meta, M_MD_JMICRON);
962 	return (G_RAID_MD_TASTE_FAIL);
963 }
964 
965 static int
966 g_raid_md_event_jmicron(struct g_raid_md_object *md,
967     struct g_raid_disk *disk, u_int event)
968 {
969 	struct g_raid_softc *sc;
970 	struct g_raid_subdisk *sd;
971 	struct g_raid_md_jmicron_object *mdi;
972 	struct g_raid_md_jmicron_perdisk *pd;
973 
974 	sc = md->mdo_softc;
975 	mdi = (struct g_raid_md_jmicron_object *)md;
976 	if (disk == NULL) {
977 		switch (event) {
978 		case G_RAID_NODE_E_START:
979 			if (!mdi->mdio_started)
980 				g_raid_md_jmicron_start(sc);
981 			return (0);
982 		}
983 		return (-1);
984 	}
985 	pd = (struct g_raid_md_jmicron_perdisk *)disk->d_md_data;
986 	switch (event) {
987 	case G_RAID_DISK_E_DISCONNECTED:
988 		/* If disk was assigned, just update statuses. */
989 		if (pd->pd_disk_pos >= 0) {
990 			g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
991 			if (disk->d_consumer) {
992 				g_raid_kill_consumer(sc, disk->d_consumer);
993 				disk->d_consumer = NULL;
994 			}
995 			TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
996 				g_raid_change_subdisk_state(sd,
997 				    G_RAID_SUBDISK_S_NONE);
998 				g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
999 				    G_RAID_EVENT_SUBDISK);
1000 			}
1001 		} else {
1002 			/* Otherwise -- delete. */
1003 			g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
1004 			g_raid_destroy_disk(disk);
1005 		}
1006 
1007 		/* Write updated metadata to all disks. */
1008 		g_raid_md_write_jmicron(md, NULL, NULL, NULL);
1009 
1010 		/* Check if anything left except placeholders. */
1011 		if (g_raid_ndisks(sc, -1) ==
1012 		    g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1013 			g_raid_destroy_node(sc, 0);
1014 		else
1015 			g_raid_md_jmicron_refill(sc);
1016 		return (0);
1017 	}
1018 	return (-2);
1019 }
1020 
1021 static int
1022 g_raid_md_ctl_jmicron(struct g_raid_md_object *md,
1023     struct gctl_req *req)
1024 {
1025 	struct g_raid_softc *sc;
1026 	struct g_raid_volume *vol;
1027 	struct g_raid_subdisk *sd;
1028 	struct g_raid_disk *disk;
1029 	struct g_raid_md_jmicron_object *mdi;
1030 	struct g_raid_md_jmicron_perdisk *pd;
1031 	struct g_consumer *cp;
1032 	struct g_provider *pp;
1033 	char arg[16];
1034 	const char *verb, *volname, *levelname, *diskname;
1035 	int *nargs, *force;
1036 	off_t size, sectorsize, strip;
1037 	intmax_t *sizearg, *striparg;
1038 	int numdisks, i, len, level, qual, update;
1039 	int error;
1040 
1041 	sc = md->mdo_softc;
1042 	mdi = (struct g_raid_md_jmicron_object *)md;
1043 	verb = gctl_get_param(req, "verb", NULL);
1044 	nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
1045 	error = 0;
1046 	if (strcmp(verb, "label") == 0) {
1047 		if (*nargs < 4) {
1048 			gctl_error(req, "Invalid number of arguments.");
1049 			return (-1);
1050 		}
1051 		volname = gctl_get_asciiparam(req, "arg1");
1052 		if (volname == NULL) {
1053 			gctl_error(req, "No volume name.");
1054 			return (-2);
1055 		}
1056 		levelname = gctl_get_asciiparam(req, "arg2");
1057 		if (levelname == NULL) {
1058 			gctl_error(req, "No RAID level.");
1059 			return (-3);
1060 		}
1061 		if (strcasecmp(levelname, "RAID5") == 0)
1062 			levelname = "RAID5-LA";
1063 		if (g_raid_volume_str2level(levelname, &level, &qual)) {
1064 			gctl_error(req, "Unknown RAID level '%s'.", levelname);
1065 			return (-4);
1066 		}
1067 		numdisks = *nargs - 3;
1068 		force = gctl_get_paraml(req, "force", sizeof(*force));
1069 		if (!g_raid_md_jmicron_supported(level, qual, numdisks,
1070 		    force ? *force : 0)) {
1071 			gctl_error(req, "Unsupported RAID level "
1072 			    "(0x%02x/0x%02x), or number of disks (%d).",
1073 			    level, qual, numdisks);
1074 			return (-5);
1075 		}
1076 
1077 		/* Search for disks, connect them and probe. */
1078 		size = 0x7fffffffffffffffllu;
1079 		sectorsize = 0;
1080 		for (i = 0; i < numdisks; i++) {
1081 			snprintf(arg, sizeof(arg), "arg%d", i + 3);
1082 			diskname = gctl_get_asciiparam(req, arg);
1083 			if (diskname == NULL) {
1084 				gctl_error(req, "No disk name (%s).", arg);
1085 				error = -6;
1086 				break;
1087 			}
1088 			if (strcmp(diskname, "NONE") == 0) {
1089 				cp = NULL;
1090 				pp = NULL;
1091 			} else {
1092 				g_topology_lock();
1093 				cp = g_raid_open_consumer(sc, diskname);
1094 				if (cp == NULL) {
1095 					gctl_error(req, "Can't open '%s'.",
1096 					    diskname);
1097 					g_topology_unlock();
1098 					error = -7;
1099 					break;
1100 				}
1101 				pp = cp->provider;
1102 			}
1103 			pd = malloc(sizeof(*pd), M_MD_JMICRON, M_WAITOK | M_ZERO);
1104 			pd->pd_disk_pos = i;
1105 			pd->pd_disk_id = arc4random() & JMICRON_DISK_MASK;
1106 			disk = g_raid_create_disk(sc);
1107 			disk->d_md_data = (void *)pd;
1108 			disk->d_consumer = cp;
1109 			if (cp == NULL)
1110 				continue;
1111 			cp->private = disk;
1112 			g_topology_unlock();
1113 
1114 			g_raid_get_disk_info(disk);
1115 
1116 			pd->pd_disk_size = pp->mediasize;
1117 			if (size > pp->mediasize)
1118 				size = pp->mediasize;
1119 			if (sectorsize < pp->sectorsize)
1120 				sectorsize = pp->sectorsize;
1121 		}
1122 		if (error != 0)
1123 			return (error);
1124 
1125 		if (sectorsize <= 0) {
1126 			gctl_error(req, "Can't get sector size.");
1127 			return (-8);
1128 		}
1129 
1130 		/* Reserve space for metadata. */
1131 		size -= sectorsize;
1132 
1133 		/* Handle size argument. */
1134 		len = sizeof(*sizearg);
1135 		sizearg = gctl_get_param(req, "size", &len);
1136 		if (sizearg != NULL && len == sizeof(*sizearg) &&
1137 		    *sizearg > 0) {
1138 			if (*sizearg > size) {
1139 				gctl_error(req, "Size too big %lld > %lld.",
1140 				    (long long)*sizearg, (long long)size);
1141 				return (-9);
1142 			}
1143 			size = *sizearg;
1144 		}
1145 
1146 		/* Handle strip argument. */
1147 		strip = 131072;
1148 		len = sizeof(*striparg);
1149 		striparg = gctl_get_param(req, "strip", &len);
1150 		if (striparg != NULL && len == sizeof(*striparg) &&
1151 		    *striparg > 0) {
1152 			if (*striparg < sectorsize) {
1153 				gctl_error(req, "Strip size too small.");
1154 				return (-10);
1155 			}
1156 			if (*striparg % sectorsize != 0) {
1157 				gctl_error(req, "Incorrect strip size.");
1158 				return (-11);
1159 			}
1160 			if (strip > 65535 * sectorsize) {
1161 				gctl_error(req, "Strip size too big.");
1162 				return (-12);
1163 			}
1164 			strip = *striparg;
1165 		}
1166 
1167 		/* Round size down to strip or sector. */
1168 		if (level == G_RAID_VOLUME_RL_RAID1)
1169 			size -= (size % sectorsize);
1170 		else if (level == G_RAID_VOLUME_RL_RAID1E &&
1171 		    (numdisks & 1) != 0)
1172 			size -= (size % (2 * strip));
1173 		else
1174 			size -= (size % strip);
1175 		if (size <= 0) {
1176 			gctl_error(req, "Size too small.");
1177 			return (-13);
1178 		}
1179 		if (size > 0xffffffffffffllu * sectorsize) {
1180 			gctl_error(req, "Size too big.");
1181 			return (-14);
1182 		}
1183 
1184 		/* We have all we need, create things: volume, ... */
1185 		mdi->mdio_total_disks = numdisks;
1186 		mdi->mdio_started = 1;
1187 		vol = g_raid_create_volume(sc, volname, -1);
1188 		vol->v_md_data = (void *)(intptr_t)0;
1189 		vol->v_raid_level = level;
1190 		vol->v_raid_level_qualifier = qual;
1191 		vol->v_strip_size = strip;
1192 		vol->v_disks_count = numdisks;
1193 		if (level == G_RAID_VOLUME_RL_RAID0 ||
1194 		    level == G_RAID_VOLUME_RL_CONCAT ||
1195 		    level == G_RAID_VOLUME_RL_SINGLE)
1196 			vol->v_mediasize = size * numdisks;
1197 		else if (level == G_RAID_VOLUME_RL_RAID1)
1198 			vol->v_mediasize = size;
1199 		else if (level == G_RAID_VOLUME_RL_RAID5)
1200 			vol->v_mediasize = size * (numdisks - 1);
1201 		else { /* RAID1E */
1202 			vol->v_mediasize = ((size * numdisks) / strip / 2) *
1203 			    strip;
1204 		}
1205 		vol->v_sectorsize = sectorsize;
1206 		g_raid_start_volume(vol);
1207 
1208 		/* , and subdisks. */
1209 		TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1210 			pd = (struct g_raid_md_jmicron_perdisk *)disk->d_md_data;
1211 			sd = &vol->v_subdisks[pd->pd_disk_pos];
1212 			sd->sd_disk = disk;
1213 			sd->sd_offset = 0;
1214 			sd->sd_size = size;
1215 			TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1216 			if (sd->sd_disk->d_consumer != NULL) {
1217 				g_raid_change_disk_state(disk,
1218 				    G_RAID_DISK_S_ACTIVE);
1219 				g_raid_change_subdisk_state(sd,
1220 				    G_RAID_SUBDISK_S_ACTIVE);
1221 				g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
1222 				    G_RAID_EVENT_SUBDISK);
1223 			} else {
1224 				g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1225 			}
1226 		}
1227 
1228 		/* Write metadata based on created entities. */
1229 		G_RAID_DEBUG1(0, sc, "Array started.");
1230 		g_raid_md_write_jmicron(md, NULL, NULL, NULL);
1231 
1232 		/* Pickup any STALE/SPARE disks to refill array if needed. */
1233 		g_raid_md_jmicron_refill(sc);
1234 
1235 		g_raid_event_send(vol, G_RAID_VOLUME_E_START,
1236 		    G_RAID_EVENT_VOLUME);
1237 		return (0);
1238 	}
1239 	if (strcmp(verb, "delete") == 0) {
1240 		/* Check if some volume is still open. */
1241 		force = gctl_get_paraml(req, "force", sizeof(*force));
1242 		if (force != NULL && *force == 0 &&
1243 		    g_raid_nopens(sc) != 0) {
1244 			gctl_error(req, "Some volume is still open.");
1245 			return (-4);
1246 		}
1247 
1248 		TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1249 			if (disk->d_consumer)
1250 				jmicron_meta_erase(disk->d_consumer);
1251 		}
1252 		g_raid_destroy_node(sc, 0);
1253 		return (0);
1254 	}
1255 	if (strcmp(verb, "remove") == 0 ||
1256 	    strcmp(verb, "fail") == 0) {
1257 		if (*nargs < 2) {
1258 			gctl_error(req, "Invalid number of arguments.");
1259 			return (-1);
1260 		}
1261 		for (i = 1; i < *nargs; i++) {
1262 			snprintf(arg, sizeof(arg), "arg%d", i);
1263 			diskname = gctl_get_asciiparam(req, arg);
1264 			if (diskname == NULL) {
1265 				gctl_error(req, "No disk name (%s).", arg);
1266 				error = -2;
1267 				break;
1268 			}
1269 			if (strncmp(diskname, _PATH_DEV, 5) == 0)
1270 				diskname += 5;
1271 
1272 			TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1273 				if (disk->d_consumer != NULL &&
1274 				    disk->d_consumer->provider != NULL &&
1275 				    strcmp(disk->d_consumer->provider->name,
1276 				     diskname) == 0)
1277 					break;
1278 			}
1279 			if (disk == NULL) {
1280 				gctl_error(req, "Disk '%s' not found.",
1281 				    diskname);
1282 				error = -3;
1283 				break;
1284 			}
1285 
1286 			if (strcmp(verb, "fail") == 0) {
1287 				g_raid_md_fail_disk_jmicron(md, NULL, disk);
1288 				continue;
1289 			}
1290 
1291 			pd = (struct g_raid_md_jmicron_perdisk *)disk->d_md_data;
1292 
1293 			/* Erase metadata on deleting disk. */
1294 			jmicron_meta_erase(disk->d_consumer);
1295 
1296 			/* If disk was assigned, just update statuses. */
1297 			if (pd->pd_disk_pos >= 0) {
1298 				g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1299 				g_raid_kill_consumer(sc, disk->d_consumer);
1300 				disk->d_consumer = NULL;
1301 				TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
1302 					g_raid_change_subdisk_state(sd,
1303 					    G_RAID_SUBDISK_S_NONE);
1304 					g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
1305 					    G_RAID_EVENT_SUBDISK);
1306 				}
1307 			} else {
1308 				/* Otherwise -- delete. */
1309 				g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
1310 				g_raid_destroy_disk(disk);
1311 			}
1312 		}
1313 
1314 		/* Write updated metadata to remaining disks. */
1315 		g_raid_md_write_jmicron(md, NULL, NULL, NULL);
1316 
1317 		/* Check if anything left except placeholders. */
1318 		if (g_raid_ndisks(sc, -1) ==
1319 		    g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1320 			g_raid_destroy_node(sc, 0);
1321 		else
1322 			g_raid_md_jmicron_refill(sc);
1323 		return (error);
1324 	}
1325 	if (strcmp(verb, "insert") == 0) {
1326 		if (*nargs < 2) {
1327 			gctl_error(req, "Invalid number of arguments.");
1328 			return (-1);
1329 		}
1330 		update = 0;
1331 		for (i = 1; i < *nargs; i++) {
1332 			/* Get disk name. */
1333 			snprintf(arg, sizeof(arg), "arg%d", i);
1334 			diskname = gctl_get_asciiparam(req, arg);
1335 			if (diskname == NULL) {
1336 				gctl_error(req, "No disk name (%s).", arg);
1337 				error = -3;
1338 				break;
1339 			}
1340 
1341 			/* Try to find provider with specified name. */
1342 			g_topology_lock();
1343 			cp = g_raid_open_consumer(sc, diskname);
1344 			if (cp == NULL) {
1345 				gctl_error(req, "Can't open disk '%s'.",
1346 				    diskname);
1347 				g_topology_unlock();
1348 				error = -4;
1349 				break;
1350 			}
1351 			pp = cp->provider;
1352 
1353 			pd = malloc(sizeof(*pd), M_MD_JMICRON, M_WAITOK | M_ZERO);
1354 			pd->pd_disk_pos = -3;
1355 			pd->pd_disk_id = arc4random() & JMICRON_DISK_MASK;
1356 			pd->pd_disk_size = pp->mediasize;
1357 
1358 			disk = g_raid_create_disk(sc);
1359 			disk->d_consumer = cp;
1360 			disk->d_md_data = (void *)pd;
1361 			cp->private = disk;
1362 			g_topology_unlock();
1363 
1364 			g_raid_get_disk_info(disk);
1365 
1366 			/* Welcome the "new" disk. */
1367 			update += g_raid_md_jmicron_start_disk(disk);
1368 			if (disk->d_state != G_RAID_DISK_S_ACTIVE &&
1369 			    disk->d_state != G_RAID_DISK_S_SPARE) {
1370 				gctl_error(req, "Disk '%s' doesn't fit.",
1371 				    diskname);
1372 				g_raid_destroy_disk(disk);
1373 				error = -8;
1374 				break;
1375 			}
1376 		}
1377 
1378 		/* Write new metadata if we changed something. */
1379 		if (update)
1380 			g_raid_md_write_jmicron(md, NULL, NULL, NULL);
1381 		return (error);
1382 	}
1383 	gctl_error(req, "Command '%s' is not supported.", verb);
1384 	return (-100);
1385 }
1386 
1387 static int
1388 g_raid_md_write_jmicron(struct g_raid_md_object *md, struct g_raid_volume *tvol,
1389     struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
1390 {
1391 	struct g_raid_softc *sc;
1392 	struct g_raid_volume *vol;
1393 	struct g_raid_subdisk *sd;
1394 	struct g_raid_disk *disk;
1395 	struct g_raid_md_jmicron_object *mdi;
1396 	struct g_raid_md_jmicron_perdisk *pd;
1397 	struct jmicron_raid_conf *meta;
1398 	int i, spares;
1399 
1400 	sc = md->mdo_softc;
1401 	mdi = (struct g_raid_md_jmicron_object *)md;
1402 
1403 	if (sc->sc_stopping == G_RAID_DESTROY_HARD)
1404 		return (0);
1405 
1406 	/* There is only one volume. */
1407 	vol = TAILQ_FIRST(&sc->sc_volumes);
1408 
1409 	/* Fill global fields. */
1410 	meta = malloc(sizeof(*meta), M_MD_JMICRON, M_WAITOK | M_ZERO);
1411 	strncpy(meta->signature, JMICRON_MAGIC, 2);
1412 	meta->version = JMICRON_VERSION;
1413 	jmicron_meta_put_name(meta, vol->v_name);
1414 	if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0)
1415 		meta->type = JMICRON_T_RAID0;
1416 	else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
1417 		meta->type = JMICRON_T_RAID1;
1418 	else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E)
1419 		meta->type = JMICRON_T_RAID01;
1420 	else if (vol->v_raid_level == G_RAID_VOLUME_RL_CONCAT ||
1421 	    vol->v_raid_level == G_RAID_VOLUME_RL_SINGLE)
1422 		meta->type = JMICRON_T_CONCAT;
1423 	else
1424 		meta->type = JMICRON_T_RAID5;
1425 	meta->stripe_shift = fls(vol->v_strip_size / 2048);
1426 	meta->flags = JMICRON_F_READY | JMICRON_F_BOOTABLE;
1427 	for (i = 0; i < vol->v_disks_count; i++) {
1428 		sd = &vol->v_subdisks[i];
1429 		if (sd->sd_disk == NULL || sd->sd_disk->d_md_data == NULL)
1430 			meta->disks[i] = 0xffffffff;
1431 		else {
1432 			pd = (struct g_raid_md_jmicron_perdisk *)
1433 			    sd->sd_disk->d_md_data;
1434 			meta->disks[i] = pd->pd_disk_id;
1435 		}
1436 		if (sd->sd_state < G_RAID_SUBDISK_S_STALE)
1437 			meta->flags |= JMICRON_F_BADSEC;
1438 		if (vol->v_dirty)
1439 			meta->flags |= JMICRON_F_UNSYNC;
1440 	}
1441 
1442 	/* Put spares to their slots. */
1443 	spares = 0;
1444 	TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1445 		pd = (struct g_raid_md_jmicron_perdisk *)disk->d_md_data;
1446 		if (disk->d_state != G_RAID_DISK_S_SPARE)
1447 			continue;
1448 		meta->spare[spares] = pd->pd_disk_id;
1449 		if (++spares >= 2)
1450 			break;
1451 	}
1452 
1453 	/* We are done. Print meta data and store them to disks. */
1454 	if (mdi->mdio_meta != NULL)
1455 		free(mdi->mdio_meta, M_MD_JMICRON);
1456 	mdi->mdio_meta = meta;
1457 	TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1458 		pd = (struct g_raid_md_jmicron_perdisk *)disk->d_md_data;
1459 		if (disk->d_state != G_RAID_DISK_S_ACTIVE &&
1460 		    disk->d_state != G_RAID_DISK_S_SPARE)
1461 			continue;
1462 		if (pd->pd_meta != NULL) {
1463 			free(pd->pd_meta, M_MD_JMICRON);
1464 			pd->pd_meta = NULL;
1465 		}
1466 		pd->pd_meta = jmicron_meta_copy(meta);
1467 		pd->pd_meta->disk_id = pd->pd_disk_id;
1468 		if ((sd = TAILQ_FIRST(&disk->d_subdisks)) != NULL) {
1469 			pd->pd_meta->offset =
1470 			    (sd->sd_offset / 512) / 16;
1471 			pd->pd_meta->disk_sectors_high =
1472 			    (sd->sd_size / 512) >> 16;
1473 			pd->pd_meta->disk_sectors_low =
1474 			    (sd->sd_size / 512) & 0xffff;
1475 			if (sd->sd_state < G_RAID_SUBDISK_S_STALE)
1476 				pd->pd_meta->flags &= ~JMICRON_F_BADSEC;
1477 			else if (sd->sd_state < G_RAID_SUBDISK_S_ACTIVE)
1478 				pd->pd_meta->flags |= JMICRON_F_UNSYNC;
1479 		}
1480 		G_RAID_DEBUG(1, "Writing JMicron metadata to %s",
1481 		    g_raid_get_diskname(disk));
1482 		g_raid_md_jmicron_print(pd->pd_meta);
1483 		jmicron_meta_write(disk->d_consumer, pd->pd_meta);
1484 	}
1485 	return (0);
1486 }
1487 
1488 static int
1489 g_raid_md_fail_disk_jmicron(struct g_raid_md_object *md,
1490     struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
1491 {
1492 	struct g_raid_softc *sc;
1493 	struct g_raid_md_jmicron_perdisk *pd;
1494 	struct g_raid_subdisk *sd;
1495 
1496 	sc = md->mdo_softc;
1497 	pd = (struct g_raid_md_jmicron_perdisk *)tdisk->d_md_data;
1498 
1499 	/* We can't fail disk that is not a part of array now. */
1500 	if (pd->pd_disk_pos < 0)
1501 		return (-1);
1502 
1503 	if (tdisk->d_consumer)
1504 		jmicron_meta_erase(tdisk->d_consumer);
1505 
1506 	/* Change states. */
1507 	g_raid_change_disk_state(tdisk, G_RAID_DISK_S_FAILED);
1508 	TAILQ_FOREACH(sd, &tdisk->d_subdisks, sd_next) {
1509 		g_raid_change_subdisk_state(sd,
1510 		    G_RAID_SUBDISK_S_FAILED);
1511 		g_raid_event_send(sd, G_RAID_SUBDISK_E_FAILED,
1512 		    G_RAID_EVENT_SUBDISK);
1513 	}
1514 
1515 	/* Write updated metadata to remaining disks. */
1516 	g_raid_md_write_jmicron(md, NULL, NULL, tdisk);
1517 
1518 	/* Check if anything left except placeholders. */
1519 	if (g_raid_ndisks(sc, -1) ==
1520 	    g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1521 		g_raid_destroy_node(sc, 0);
1522 	else
1523 		g_raid_md_jmicron_refill(sc);
1524 	return (0);
1525 }
1526 
1527 static int
1528 g_raid_md_free_disk_jmicron(struct g_raid_md_object *md,
1529     struct g_raid_disk *disk)
1530 {
1531 	struct g_raid_md_jmicron_perdisk *pd;
1532 
1533 	pd = (struct g_raid_md_jmicron_perdisk *)disk->d_md_data;
1534 	if (pd->pd_meta != NULL) {
1535 		free(pd->pd_meta, M_MD_JMICRON);
1536 		pd->pd_meta = NULL;
1537 	}
1538 	free(pd, M_MD_JMICRON);
1539 	disk->d_md_data = NULL;
1540 	return (0);
1541 }
1542 
1543 static int
1544 g_raid_md_free_jmicron(struct g_raid_md_object *md)
1545 {
1546 	struct g_raid_md_jmicron_object *mdi;
1547 
1548 	mdi = (struct g_raid_md_jmicron_object *)md;
1549 	if (!mdi->mdio_started) {
1550 		mdi->mdio_started = 0;
1551 		callout_stop(&mdi->mdio_start_co);
1552 		G_RAID_DEBUG1(1, md->mdo_softc,
1553 		    "root_mount_rel %p", mdi->mdio_rootmount);
1554 		root_mount_rel(mdi->mdio_rootmount);
1555 		mdi->mdio_rootmount = NULL;
1556 	}
1557 	if (mdi->mdio_meta != NULL) {
1558 		free(mdi->mdio_meta, M_MD_JMICRON);
1559 		mdi->mdio_meta = NULL;
1560 	}
1561 	return (0);
1562 }
1563 
1564 G_RAID_MD_DECLARE(jmicron, "JMicron");
1565