xref: /freebsd/sys/geom/raid/md_jmicron.c (revision edf8578117e8844e02c0121147f45e4609b30680)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
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 #include <sys/param.h>
32 #include <sys/bio.h>
33 #include <sys/endian.h>
34 #include <sys/kernel.h>
35 #include <sys/kobj.h>
36 #include <sys/limits.h>
37 #include <sys/lock.h>
38 #include <sys/malloc.h>
39 #include <sys/mutex.h>
40 #include <sys/systm.h>
41 #include <sys/taskqueue.h>
42 #include <geom/geom.h>
43 #include <geom/geom_dbg.h>
44 #include "geom/raid/g_raid.h"
45 #include "g_raid_md_if.h"
46 
47 static MALLOC_DEFINE(M_MD_JMICRON, "md_jmicron_data", "GEOM_RAID JMicron metadata");
48 
49 #define	JMICRON_MAX_DISKS	8
50 #define	JMICRON_MAX_SPARE	2
51 
52 struct jmicron_raid_conf {
53     uint8_t		signature[2];
54 #define	JMICRON_MAGIC		"JM"
55 
56     uint16_t		version;
57 #define	JMICRON_VERSION		0x0001
58 
59     uint16_t		checksum;
60     uint8_t		filler_1[10];
61     uint32_t		disk_id;
62     uint32_t		offset;
63     uint32_t		disk_sectors_high;
64     uint16_t		disk_sectors_low;
65     uint8_t		filler_2[2];
66     uint8_t		name[16];
67     uint8_t		type;
68 #define	JMICRON_T_RAID0		0
69 #define	JMICRON_T_RAID1		1
70 #define	JMICRON_T_RAID01	2
71 #define	JMICRON_T_CONCAT	3
72 #define	JMICRON_T_RAID5		5
73 
74     uint8_t		stripe_shift;
75     uint16_t		flags;
76 #define	JMICRON_F_READY		0x0001
77 #define	JMICRON_F_BOOTABLE	0x0002
78 #define	JMICRON_F_BADSEC	0x0004
79 #define	JMICRON_F_ACTIVE	0x0010
80 #define	JMICRON_F_UNSYNC	0x0020
81 #define	JMICRON_F_NEWEST	0x0040
82 
83     uint8_t		filler_3[4];
84     uint32_t		spare[JMICRON_MAX_SPARE];
85     uint32_t		disks[JMICRON_MAX_DISKS];
86 #define	JMICRON_DISK_MASK	0xFFFFFFF0
87 #define	JMICRON_SEG_MASK	0x0000000F
88     uint8_t		filler_4[32];
89     uint8_t		filler_5[384];
90 };
91 
92 struct g_raid_md_jmicron_perdisk {
93 	struct jmicron_raid_conf	*pd_meta;
94 	int				 pd_disk_pos;
95 	int				 pd_disk_id;
96 	off_t				 pd_disk_size;
97 };
98 
99 struct g_raid_md_jmicron_object {
100 	struct g_raid_md_object	 mdio_base;
101 	uint32_t		 mdio_config_id;
102 	struct jmicron_raid_conf	*mdio_meta;
103 	struct callout		 mdio_start_co;	/* STARTING state timer. */
104 	int			 mdio_total_disks;
105 	int			 mdio_disks_present;
106 	int			 mdio_started;
107 	int			 mdio_incomplete;
108 	struct root_hold_token	*mdio_rootmount; /* Root mount delay token. */
109 };
110 
111 static g_raid_md_create_t g_raid_md_create_jmicron;
112 static g_raid_md_taste_t g_raid_md_taste_jmicron;
113 static g_raid_md_event_t g_raid_md_event_jmicron;
114 static g_raid_md_ctl_t g_raid_md_ctl_jmicron;
115 static g_raid_md_write_t g_raid_md_write_jmicron;
116 static g_raid_md_fail_disk_t g_raid_md_fail_disk_jmicron;
117 static g_raid_md_free_disk_t g_raid_md_free_disk_jmicron;
118 static g_raid_md_free_t g_raid_md_free_jmicron;
119 
120 static kobj_method_t g_raid_md_jmicron_methods[] = {
121 	KOBJMETHOD(g_raid_md_create,	g_raid_md_create_jmicron),
122 	KOBJMETHOD(g_raid_md_taste,	g_raid_md_taste_jmicron),
123 	KOBJMETHOD(g_raid_md_event,	g_raid_md_event_jmicron),
124 	KOBJMETHOD(g_raid_md_ctl,	g_raid_md_ctl_jmicron),
125 	KOBJMETHOD(g_raid_md_write,	g_raid_md_write_jmicron),
126 	KOBJMETHOD(g_raid_md_fail_disk,	g_raid_md_fail_disk_jmicron),
127 	KOBJMETHOD(g_raid_md_free_disk,	g_raid_md_free_disk_jmicron),
128 	KOBJMETHOD(g_raid_md_free,	g_raid_md_free_jmicron),
129 	{ 0, 0 }
130 };
131 
132 static struct g_raid_md_class g_raid_md_jmicron_class = {
133 	"JMicron",
134 	g_raid_md_jmicron_methods,
135 	sizeof(struct g_raid_md_jmicron_object),
136 	.mdc_enable = 1,
137 	.mdc_priority = 100
138 };
139 
140 static void
141 g_raid_md_jmicron_print(struct jmicron_raid_conf *meta)
142 {
143 	int k;
144 
145 	if (g_raid_debug < 1)
146 		return;
147 
148 	printf("********* ATA JMicron RAID Metadata *********\n");
149 	printf("signature           <%c%c>\n", meta->signature[0], meta->signature[1]);
150 	printf("version             %04x\n", meta->version);
151 	printf("checksum            0x%04x\n", meta->checksum);
152 	printf("disk_id             0x%08x\n", meta->disk_id);
153 	printf("offset              0x%08x\n", meta->offset);
154 	printf("disk_sectors_high   0x%08x\n", meta->disk_sectors_high);
155 	printf("disk_sectors_low    0x%04x\n", meta->disk_sectors_low);
156 	printf("name                <%.16s>\n", meta->name);
157 	printf("type                %d\n", meta->type);
158 	printf("stripe_shift        %d\n", meta->stripe_shift);
159 	printf("flags               %04x\n", meta->flags);
160 	printf("spare              ");
161 	for (k = 0; k < JMICRON_MAX_SPARE; k++)
162 		printf(" 0x%08x", meta->spare[k]);
163 	printf("\n");
164 	printf("disks              ");
165 	for (k = 0; k < JMICRON_MAX_DISKS; k++)
166 		printf(" 0x%08x", meta->disks[k]);
167 	printf("\n");
168 	printf("=================================================\n");
169 }
170 
171 static struct jmicron_raid_conf *
172 jmicron_meta_copy(struct jmicron_raid_conf *meta)
173 {
174 	struct jmicron_raid_conf *nmeta;
175 
176 	nmeta = malloc(sizeof(*meta), M_MD_JMICRON, M_WAITOK);
177 	memcpy(nmeta, meta, sizeof(*meta));
178 	return (nmeta);
179 }
180 
181 static int
182 jmicron_meta_total_disks(struct jmicron_raid_conf *meta)
183 {
184 	int pos;
185 
186 	for (pos = 0; pos < JMICRON_MAX_DISKS; pos++) {
187 		if (meta->disks[pos] == 0)
188 			break;
189 	}
190 	return (pos);
191 }
192 
193 static int
194 jmicron_meta_total_spare(struct jmicron_raid_conf *meta)
195 {
196 	int pos, n;
197 
198 	n = 0;
199 	for (pos = 0; pos < JMICRON_MAX_SPARE; pos++) {
200 		if (meta->spare[pos] != 0)
201 			n++;
202 	}
203 	return (n);
204 }
205 
206 /*
207  * Generate fake Configuration ID based on disk IDs.
208  * Note: it will change after each disk set change.
209  */
210 static uint32_t
211 jmicron_meta_config_id(struct jmicron_raid_conf *meta)
212 {
213 	int pos;
214 	uint32_t config_id;
215 
216 	config_id = 0;
217 	for (pos = 0; pos < JMICRON_MAX_DISKS; pos++)
218 		config_id += meta->disks[pos] << pos;
219 	return (config_id);
220 }
221 
222 static void
223 jmicron_meta_get_name(struct jmicron_raid_conf *meta, char *buf)
224 {
225 	int i;
226 
227 	strncpy(buf, meta->name, 16);
228 	buf[16] = 0;
229 	for (i = 15; i >= 0; i--) {
230 		if (buf[i] > 0x20)
231 			break;
232 		buf[i] = 0;
233 	}
234 }
235 
236 static void
237 jmicron_meta_put_name(struct jmicron_raid_conf *meta, char *buf)
238 {
239 
240 	memset(meta->name, 0x20, 16);
241 	memcpy(meta->name, buf, MIN(strlen(buf), 16));
242 }
243 
244 static int
245 jmicron_meta_find_disk(struct jmicron_raid_conf *meta, uint32_t id)
246 {
247 	int pos;
248 
249 	id &= JMICRON_DISK_MASK;
250 	for (pos = 0; pos < JMICRON_MAX_DISKS; pos++) {
251 		if ((meta->disks[pos] & JMICRON_DISK_MASK) == id)
252 			return (pos);
253 	}
254 	for (pos = 0; pos < JMICRON_MAX_SPARE; pos++) {
255 		if ((meta->spare[pos] & JMICRON_DISK_MASK) == id)
256 			return (-3);
257 	}
258 	return (-1);
259 }
260 
261 static struct jmicron_raid_conf *
262 jmicron_meta_read(struct g_consumer *cp)
263 {
264 	struct g_provider *pp;
265 	struct jmicron_raid_conf *meta;
266 	char *buf;
267 	int error, i;
268 	uint16_t checksum, *ptr;
269 
270 	pp = cp->provider;
271 	if (pp->sectorsize < sizeof(*meta))
272 		return (NULL);
273 	/* Read the anchor sector. */
274 	buf = g_read_data(cp,
275 	    pp->mediasize - pp->sectorsize, pp->sectorsize, &error);
276 	if (buf == NULL) {
277 		G_RAID_DEBUG(1, "Cannot read metadata from %s (error=%d).",
278 		    pp->name, error);
279 		return (NULL);
280 	}
281 	meta = (struct jmicron_raid_conf *)buf;
282 
283 	/* Check if this is an JMicron RAID struct */
284 	if (strncmp(meta->signature, JMICRON_MAGIC, strlen(JMICRON_MAGIC))) {
285 		G_RAID_DEBUG(1, "JMicron signature check failed on %s", pp->name);
286 		g_free(buf);
287 		return (NULL);
288 	}
289 	meta = malloc(sizeof(*meta), M_MD_JMICRON, M_WAITOK);
290 	memcpy(meta, buf, min(sizeof(*meta), pp->sectorsize));
291 	g_free(buf);
292 
293 	/* Check metadata checksum. */
294 	for (checksum = 0, ptr = (uint16_t *)meta, i = 0; i < 64; i++)
295 		checksum += *ptr++;
296 	if (checksum != 0) {
297 		G_RAID_DEBUG(1, "JMicron checksum check failed on %s", pp->name);
298 		free(meta, M_MD_JMICRON);
299 		return (NULL);
300 	}
301 
302 	return (meta);
303 }
304 
305 static int
306 jmicron_meta_write(struct g_consumer *cp, struct jmicron_raid_conf *meta)
307 {
308 	struct g_provider *pp;
309 	char *buf;
310 	int error, i;
311 	uint16_t checksum, *ptr;
312 
313 	pp = cp->provider;
314 
315 	/* Recalculate checksum for case if metadata were changed. */
316 	meta->checksum = 0;
317 	for (checksum = 0, ptr = (uint16_t *)meta, i = 0; i < 64; i++)
318 		checksum += *ptr++;
319 	meta->checksum -= checksum;
320 
321 	/* Create and fill buffer. */
322 	buf = malloc(pp->sectorsize, M_MD_JMICRON, M_WAITOK | M_ZERO);
323 	memcpy(buf, meta, sizeof(*meta));
324 
325 	error = g_write_data(cp,
326 	    pp->mediasize - pp->sectorsize, buf, pp->sectorsize);
327 	if (error != 0) {
328 		G_RAID_DEBUG(1, "Cannot write metadata to %s (error=%d).",
329 		    pp->name, error);
330 	}
331 
332 	free(buf, M_MD_JMICRON);
333 	return (error);
334 }
335 
336 static int
337 jmicron_meta_erase(struct g_consumer *cp)
338 {
339 	struct g_provider *pp;
340 	char *buf;
341 	int error;
342 
343 	pp = cp->provider;
344 	buf = malloc(pp->sectorsize, M_MD_JMICRON, M_WAITOK | M_ZERO);
345 	error = g_write_data(cp,
346 	    pp->mediasize - pp->sectorsize, buf, pp->sectorsize);
347 	if (error != 0) {
348 		G_RAID_DEBUG(1, "Cannot erase metadata on %s (error=%d).",
349 		    pp->name, error);
350 	}
351 	free(buf, M_MD_JMICRON);
352 	return (error);
353 }
354 
355 static struct g_raid_disk *
356 g_raid_md_jmicron_get_disk(struct g_raid_softc *sc, int id)
357 {
358 	struct g_raid_disk	*disk;
359 	struct g_raid_md_jmicron_perdisk *pd;
360 
361 	TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
362 		pd = (struct g_raid_md_jmicron_perdisk *)disk->d_md_data;
363 		if (pd->pd_disk_pos == id)
364 			break;
365 	}
366 	return (disk);
367 }
368 
369 static int
370 g_raid_md_jmicron_supported(int level, int qual, int disks, int force)
371 {
372 
373 	if (disks > 8)
374 		return (0);
375 	switch (level) {
376 	case G_RAID_VOLUME_RL_RAID0:
377 		if (disks < 1)
378 			return (0);
379 		if (!force && (disks < 2 || disks > 6))
380 			return (0);
381 		break;
382 	case G_RAID_VOLUME_RL_RAID1:
383 		if (disks < 1)
384 			return (0);
385 		if (!force && (disks != 2))
386 			return (0);
387 		break;
388 	case G_RAID_VOLUME_RL_RAID1E:
389 		if (disks < 2)
390 			return (0);
391 		if (!force && (disks != 4))
392 			return (0);
393 		break;
394 	case G_RAID_VOLUME_RL_SINGLE:
395 		if (disks != 1)
396 			return (0);
397 		if (!force)
398 			return (0);
399 		break;
400 	case G_RAID_VOLUME_RL_CONCAT:
401 		if (disks < 2)
402 			return (0);
403 		break;
404 	case G_RAID_VOLUME_RL_RAID5:
405 		if (disks < 3)
406 			return (0);
407 		if (qual != G_RAID_VOLUME_RLQ_R5LA)
408 			return (0);
409 		if (!force)
410 			return (0);
411 		break;
412 	default:
413 		return (0);
414 	}
415 	if (level != G_RAID_VOLUME_RL_RAID5 && qual != G_RAID_VOLUME_RLQ_NONE)
416 		return (0);
417 	return (1);
418 }
419 
420 static int
421 g_raid_md_jmicron_start_disk(struct g_raid_disk *disk)
422 {
423 	struct g_raid_softc *sc;
424 	struct g_raid_subdisk *sd, *tmpsd;
425 	struct g_raid_disk *olddisk, *tmpdisk;
426 	struct g_raid_md_object *md;
427 	struct g_raid_md_jmicron_object *mdi;
428 	struct g_raid_md_jmicron_perdisk *pd, *oldpd;
429 	struct jmicron_raid_conf *meta;
430 	int disk_pos, resurrection = 0;
431 
432 	sc = disk->d_softc;
433 	md = sc->sc_md;
434 	mdi = (struct g_raid_md_jmicron_object *)md;
435 	meta = mdi->mdio_meta;
436 	pd = (struct g_raid_md_jmicron_perdisk *)disk->d_md_data;
437 	olddisk = NULL;
438 
439 	/* Find disk position in metadata by its serial. */
440 	if (pd->pd_meta != NULL)
441 		disk_pos = jmicron_meta_find_disk(meta, pd->pd_disk_id);
442 	else
443 		disk_pos = -1;
444 	if (disk_pos < 0) {
445 		G_RAID_DEBUG1(1, sc, "Unknown, probably new or stale disk");
446 		/* If we are in the start process, that's all for now. */
447 		if (!mdi->mdio_started)
448 			goto nofit;
449 		/*
450 		 * If we have already started - try to get use of the disk.
451 		 * Try to replace OFFLINE disks first, then FAILED.
452 		 */
453 		TAILQ_FOREACH(tmpdisk, &sc->sc_disks, d_next) {
454 			if (tmpdisk->d_state != G_RAID_DISK_S_OFFLINE &&
455 			    tmpdisk->d_state != G_RAID_DISK_S_FAILED)
456 				continue;
457 			/* Make sure this disk is big enough. */
458 			TAILQ_FOREACH(sd, &tmpdisk->d_subdisks, sd_next) {
459 				if (sd->sd_offset + sd->sd_size + 512 >
460 				    pd->pd_disk_size) {
461 					G_RAID_DEBUG1(1, sc,
462 					    "Disk too small (%ju < %ju)",
463 					    pd->pd_disk_size,
464 					    sd->sd_offset + sd->sd_size + 512);
465 					break;
466 				}
467 			}
468 			if (sd != NULL)
469 				continue;
470 			if (tmpdisk->d_state == G_RAID_DISK_S_OFFLINE) {
471 				olddisk = tmpdisk;
472 				break;
473 			} else if (olddisk == NULL)
474 				olddisk = tmpdisk;
475 		}
476 		if (olddisk == NULL) {
477 nofit:
478 			if (disk_pos == -3 || pd->pd_disk_pos == -3) {
479 				g_raid_change_disk_state(disk,
480 				    G_RAID_DISK_S_SPARE);
481 				return (1);
482 			} else {
483 				g_raid_change_disk_state(disk,
484 				    G_RAID_DISK_S_STALE);
485 				return (0);
486 			}
487 		}
488 		oldpd = (struct g_raid_md_jmicron_perdisk *)olddisk->d_md_data;
489 		disk_pos = oldpd->pd_disk_pos;
490 		resurrection = 1;
491 	}
492 
493 	if (olddisk == NULL) {
494 		/* Find placeholder by position. */
495 		olddisk = g_raid_md_jmicron_get_disk(sc, disk_pos);
496 		if (olddisk == NULL)
497 			panic("No disk at position %d!", disk_pos);
498 		if (olddisk->d_state != G_RAID_DISK_S_OFFLINE) {
499 			G_RAID_DEBUG1(1, sc, "More than one disk for pos %d",
500 			    disk_pos);
501 			g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE);
502 			return (0);
503 		}
504 		oldpd = (struct g_raid_md_jmicron_perdisk *)olddisk->d_md_data;
505 	}
506 
507 	/* Replace failed disk or placeholder with new disk. */
508 	TAILQ_FOREACH_SAFE(sd, &olddisk->d_subdisks, sd_next, tmpsd) {
509 		TAILQ_REMOVE(&olddisk->d_subdisks, sd, sd_next);
510 		TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
511 		sd->sd_disk = disk;
512 	}
513 	oldpd->pd_disk_pos = -2;
514 	pd->pd_disk_pos = disk_pos;
515 	/* Update global metadata just in case. */
516 	meta->disks[disk_pos] = pd->pd_disk_id;
517 
518 	/* If it was placeholder -- destroy it. */
519 	if (olddisk->d_state == G_RAID_DISK_S_OFFLINE) {
520 		g_raid_destroy_disk(olddisk);
521 	} else {
522 		/* Otherwise, make it STALE_FAILED. */
523 		g_raid_change_disk_state(olddisk, G_RAID_DISK_S_STALE_FAILED);
524 	}
525 
526 	/* Welcome the new disk. */
527 	g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
528 	TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
529 		/*
530 		 * Different disks may have different sizes/offsets,
531 		 * especially in concat mode. Update.
532 		 */
533 		if (!resurrection) {
534 			sd->sd_offset =
535 			    (off_t)pd->pd_meta->offset * 16 * 512; //ZZZ
536 			sd->sd_size =
537 			    (((off_t)pd->pd_meta->disk_sectors_high << 16) +
538 			      pd->pd_meta->disk_sectors_low) * 512;
539 		}
540 
541 		if (resurrection) {
542 			/* Stale disk, almost same as new. */
543 			g_raid_change_subdisk_state(sd,
544 			    G_RAID_SUBDISK_S_NEW);
545 		} else if ((meta->flags & JMICRON_F_BADSEC) != 0 &&
546 		    (pd->pd_meta->flags & JMICRON_F_BADSEC) == 0) {
547 			/* Cold-inserted or rebuilding disk. */
548 			g_raid_change_subdisk_state(sd,
549 			    G_RAID_SUBDISK_S_NEW);
550 		} else if (pd->pd_meta->flags & JMICRON_F_UNSYNC) {
551 			/* Dirty or resyncing disk.. */
552 			g_raid_change_subdisk_state(sd,
553 			    G_RAID_SUBDISK_S_STALE);
554 		} else {
555 			/* Up to date disk. */
556 			g_raid_change_subdisk_state(sd,
557 			    G_RAID_SUBDISK_S_ACTIVE);
558 		}
559 		g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
560 		    G_RAID_EVENT_SUBDISK);
561 	}
562 
563 	/* Update status of our need for spare. */
564 	if (mdi->mdio_started) {
565 		mdi->mdio_incomplete =
566 		    (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) <
567 		     mdi->mdio_total_disks);
568 	}
569 
570 	return (resurrection);
571 }
572 
573 static void
574 g_disk_md_jmicron_retaste(void *arg, int pending)
575 {
576 
577 	G_RAID_DEBUG(1, "Array is not complete, trying to retaste.");
578 	g_retaste(&g_raid_class);
579 	free(arg, M_MD_JMICRON);
580 }
581 
582 static void
583 g_raid_md_jmicron_refill(struct g_raid_softc *sc)
584 {
585 	struct g_raid_md_object *md;
586 	struct g_raid_md_jmicron_object *mdi;
587 	struct g_raid_disk *disk;
588 	struct task *task;
589 	int update, na;
590 
591 	md = sc->sc_md;
592 	mdi = (struct g_raid_md_jmicron_object *)md;
593 	update = 0;
594 	do {
595 		/* Make sure we miss anything. */
596 		na = g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE);
597 		if (na == mdi->mdio_total_disks)
598 			break;
599 
600 		G_RAID_DEBUG1(1, md->mdo_softc,
601 		    "Array is not complete (%d of %d), "
602 		    "trying to refill.", na, mdi->mdio_total_disks);
603 
604 		/* Try to get use some of STALE disks. */
605 		TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
606 			if (disk->d_state == G_RAID_DISK_S_STALE) {
607 				update += g_raid_md_jmicron_start_disk(disk);
608 				if (disk->d_state == G_RAID_DISK_S_ACTIVE)
609 					break;
610 			}
611 		}
612 		if (disk != NULL)
613 			continue;
614 
615 		/* Try to get use some of SPARE disks. */
616 		TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
617 			if (disk->d_state == G_RAID_DISK_S_SPARE) {
618 				update += g_raid_md_jmicron_start_disk(disk);
619 				if (disk->d_state == G_RAID_DISK_S_ACTIVE)
620 					break;
621 			}
622 		}
623 	} while (disk != NULL);
624 
625 	/* Write new metadata if we changed something. */
626 	if (update)
627 		g_raid_md_write_jmicron(md, NULL, NULL, NULL);
628 
629 	/* Update status of our need for spare. */
630 	mdi->mdio_incomplete = (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) <
631 	    mdi->mdio_total_disks);
632 
633 	/* Request retaste hoping to find spare. */
634 	if (mdi->mdio_incomplete) {
635 		task = malloc(sizeof(struct task),
636 		    M_MD_JMICRON, M_WAITOK | M_ZERO);
637 		TASK_INIT(task, 0, g_disk_md_jmicron_retaste, task);
638 		taskqueue_enqueue(taskqueue_swi, task);
639 	}
640 }
641 
642 static void
643 g_raid_md_jmicron_start(struct g_raid_softc *sc)
644 {
645 	struct g_raid_md_object *md;
646 	struct g_raid_md_jmicron_object *mdi;
647 	struct g_raid_md_jmicron_perdisk *pd;
648 	struct jmicron_raid_conf *meta;
649 	struct g_raid_volume *vol;
650 	struct g_raid_subdisk *sd;
651 	struct g_raid_disk *disk;
652 	off_t size;
653 	int j, disk_pos;
654 	char buf[17];
655 
656 	md = sc->sc_md;
657 	mdi = (struct g_raid_md_jmicron_object *)md;
658 	meta = mdi->mdio_meta;
659 
660 	/* Create volumes and subdisks. */
661 	jmicron_meta_get_name(meta, buf);
662 	vol = g_raid_create_volume(sc, buf, -1);
663 	size = ((off_t)meta->disk_sectors_high << 16) + meta->disk_sectors_low;
664 	size *= 512; //ZZZ
665 	vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE;
666 	if (meta->type == JMICRON_T_RAID0) {
667 		vol->v_raid_level = G_RAID_VOLUME_RL_RAID0;
668 		vol->v_mediasize = size * mdi->mdio_total_disks;
669 	} else if (meta->type == JMICRON_T_RAID1) {
670 		vol->v_raid_level = G_RAID_VOLUME_RL_RAID1;
671 		vol->v_mediasize = size;
672 	} else if (meta->type == JMICRON_T_RAID01) {
673 		vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E;
674 		vol->v_mediasize = size * mdi->mdio_total_disks / 2;
675 	} else if (meta->type == JMICRON_T_CONCAT) {
676 		if (mdi->mdio_total_disks == 1)
677 			vol->v_raid_level = G_RAID_VOLUME_RL_SINGLE;
678 		else
679 			vol->v_raid_level = G_RAID_VOLUME_RL_CONCAT;
680 		vol->v_mediasize = 0;
681 	} else if (meta->type == JMICRON_T_RAID5) {
682 		vol->v_raid_level = G_RAID_VOLUME_RL_RAID5;
683 		vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_R5LA;
684 		vol->v_mediasize = size * (mdi->mdio_total_disks - 1);
685 	} else {
686 		vol->v_raid_level = G_RAID_VOLUME_RL_UNKNOWN;
687 		vol->v_mediasize = 0;
688 	}
689 	vol->v_strip_size = 1024 << meta->stripe_shift; //ZZZ
690 	vol->v_disks_count = mdi->mdio_total_disks;
691 	vol->v_sectorsize = 512; //ZZZ
692 	for (j = 0; j < vol->v_disks_count; j++) {
693 		sd = &vol->v_subdisks[j];
694 		sd->sd_offset = (off_t)meta->offset * 16 * 512; //ZZZ
695 		sd->sd_size = size;
696 	}
697 	g_raid_start_volume(vol);
698 
699 	/* Create disk placeholders to store data for later writing. */
700 	for (disk_pos = 0; disk_pos < mdi->mdio_total_disks; disk_pos++) {
701 		pd = malloc(sizeof(*pd), M_MD_JMICRON, M_WAITOK | M_ZERO);
702 		pd->pd_disk_pos = disk_pos;
703 		pd->pd_disk_id = meta->disks[disk_pos];
704 		disk = g_raid_create_disk(sc);
705 		disk->d_md_data = (void *)pd;
706 		disk->d_state = G_RAID_DISK_S_OFFLINE;
707 		sd = &vol->v_subdisks[disk_pos];
708 		sd->sd_disk = disk;
709 		TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
710 	}
711 
712 	/* Make all disks found till the moment take their places. */
713 	do {
714 		TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
715 			if (disk->d_state == G_RAID_DISK_S_NONE) {
716 				g_raid_md_jmicron_start_disk(disk);
717 				break;
718 			}
719 		}
720 	} while (disk != NULL);
721 
722 	mdi->mdio_started = 1;
723 	G_RAID_DEBUG1(0, sc, "Array started.");
724 	g_raid_md_write_jmicron(md, NULL, NULL, NULL);
725 
726 	/* Pickup any STALE/SPARE disks to refill array if needed. */
727 	g_raid_md_jmicron_refill(sc);
728 
729 	g_raid_event_send(vol, G_RAID_VOLUME_E_START, G_RAID_EVENT_VOLUME);
730 
731 	callout_stop(&mdi->mdio_start_co);
732 	G_RAID_DEBUG1(1, sc, "root_mount_rel %p", mdi->mdio_rootmount);
733 	root_mount_rel(mdi->mdio_rootmount);
734 	mdi->mdio_rootmount = NULL;
735 }
736 
737 static void
738 g_raid_md_jmicron_new_disk(struct g_raid_disk *disk)
739 {
740 	struct g_raid_softc *sc;
741 	struct g_raid_md_object *md;
742 	struct g_raid_md_jmicron_object *mdi;
743 	struct jmicron_raid_conf *pdmeta;
744 	struct g_raid_md_jmicron_perdisk *pd;
745 
746 	sc = disk->d_softc;
747 	md = sc->sc_md;
748 	mdi = (struct g_raid_md_jmicron_object *)md;
749 	pd = (struct g_raid_md_jmicron_perdisk *)disk->d_md_data;
750 	pdmeta = pd->pd_meta;
751 
752 	if (mdi->mdio_started) {
753 		if (g_raid_md_jmicron_start_disk(disk))
754 			g_raid_md_write_jmicron(md, NULL, NULL, NULL);
755 	} else {
756 		/*
757 		 * If we haven't started yet - update common metadata
758 		 * to get subdisks details, avoiding data from spare disks.
759 		 */
760 		if (mdi->mdio_meta == NULL ||
761 		    jmicron_meta_find_disk(mdi->mdio_meta,
762 		     mdi->mdio_meta->disk_id) == -3) {
763 			if (mdi->mdio_meta != NULL)
764 				free(mdi->mdio_meta, M_MD_JMICRON);
765 			mdi->mdio_meta = jmicron_meta_copy(pdmeta);
766 			mdi->mdio_total_disks = jmicron_meta_total_disks(pdmeta);
767 		}
768 		mdi->mdio_meta->flags |= pdmeta->flags & JMICRON_F_BADSEC;
769 
770 		mdi->mdio_disks_present++;
771 		G_RAID_DEBUG1(1, sc, "Matching disk (%d of %d+%d up)",
772 		    mdi->mdio_disks_present,
773 		    mdi->mdio_total_disks,
774 		    jmicron_meta_total_spare(mdi->mdio_meta));
775 
776 		/* If we collected all needed disks - start array. */
777 		if (mdi->mdio_disks_present == mdi->mdio_total_disks +
778 		    jmicron_meta_total_spare(mdi->mdio_meta))
779 			g_raid_md_jmicron_start(sc);
780 	}
781 }
782 
783 static void
784 g_raid_jmicron_go(void *arg)
785 {
786 	struct g_raid_softc *sc;
787 	struct g_raid_md_object *md;
788 	struct g_raid_md_jmicron_object *mdi;
789 
790 	sc = arg;
791 	md = sc->sc_md;
792 	mdi = (struct g_raid_md_jmicron_object *)md;
793 	if (!mdi->mdio_started) {
794 		G_RAID_DEBUG1(0, sc, "Force array start due to timeout.");
795 		g_raid_event_send(sc, G_RAID_NODE_E_START, 0);
796 	}
797 }
798 
799 static int
800 g_raid_md_create_jmicron(struct g_raid_md_object *md, struct g_class *mp,
801     struct g_geom **gp)
802 {
803 	struct g_raid_softc *sc;
804 	struct g_raid_md_jmicron_object *mdi;
805 	char name[16];
806 
807 	mdi = (struct g_raid_md_jmicron_object *)md;
808 	mdi->mdio_config_id = arc4random();
809 	snprintf(name, sizeof(name), "JMicron-%08x", mdi->mdio_config_id);
810 	sc = g_raid_create_node(mp, name, md);
811 	if (sc == NULL)
812 		return (G_RAID_MD_TASTE_FAIL);
813 	md->mdo_softc = sc;
814 	*gp = sc->sc_geom;
815 	return (G_RAID_MD_TASTE_NEW);
816 }
817 
818 static int
819 g_raid_md_taste_jmicron(struct g_raid_md_object *md, struct g_class *mp,
820                               struct g_consumer *cp, struct g_geom **gp)
821 {
822 	struct g_consumer *rcp;
823 	struct g_provider *pp;
824 	struct g_raid_md_jmicron_object *mdi, *mdi1;
825 	struct g_raid_softc *sc;
826 	struct g_raid_disk *disk;
827 	struct jmicron_raid_conf *meta;
828 	struct g_raid_md_jmicron_perdisk *pd;
829 	struct g_geom *geom;
830 	int disk_pos, result, spare, len;
831 	char name[16];
832 	uint16_t vendor;
833 
834 	G_RAID_DEBUG(1, "Tasting JMicron on %s", cp->provider->name);
835 	mdi = (struct g_raid_md_jmicron_object *)md;
836 	pp = cp->provider;
837 
838 	/* Read metadata from device. */
839 	meta = NULL;
840 	g_topology_unlock();
841 	vendor = 0xffff;
842 	len = sizeof(vendor);
843 	if (pp->geom->rank == 1)
844 		g_io_getattr("GEOM::hba_vendor", cp, &len, &vendor);
845 	meta = jmicron_meta_read(cp);
846 	g_topology_lock();
847 	if (meta == NULL) {
848 		if (g_raid_aggressive_spare) {
849 			if (vendor == 0x197b) {
850 				G_RAID_DEBUG(1,
851 				    "No JMicron metadata, forcing spare.");
852 				spare = 2;
853 				goto search;
854 			} else {
855 				G_RAID_DEBUG(1,
856 				    "JMicron vendor mismatch 0x%04x != 0x197b",
857 				    vendor);
858 			}
859 		}
860 		return (G_RAID_MD_TASTE_FAIL);
861 	}
862 
863 	/* Check this disk position in obtained metadata. */
864 	disk_pos = jmicron_meta_find_disk(meta, meta->disk_id);
865 	if (disk_pos == -1) {
866 		G_RAID_DEBUG(1, "JMicron disk_id %08x not found",
867 		    meta->disk_id);
868 		goto fail1;
869 	}
870 
871 	/* Metadata valid. Print it. */
872 	g_raid_md_jmicron_print(meta);
873 	G_RAID_DEBUG(1, "JMicron disk position %d", disk_pos);
874 	spare = (disk_pos == -2) ? 1 : 0;
875 
876 search:
877 	/* Search for matching node. */
878 	sc = NULL;
879 	mdi1 = NULL;
880 	LIST_FOREACH(geom, &mp->geom, geom) {
881 		sc = geom->softc;
882 		if (sc == NULL)
883 			continue;
884 		if (sc->sc_stopping != 0)
885 			continue;
886 		if (sc->sc_md->mdo_class != md->mdo_class)
887 			continue;
888 		mdi1 = (struct g_raid_md_jmicron_object *)sc->sc_md;
889 		if (spare == 2) {
890 			if (mdi1->mdio_incomplete)
891 				break;
892 		} else {
893 			if (mdi1->mdio_config_id ==
894 			    jmicron_meta_config_id(meta))
895 				break;
896 		}
897 	}
898 
899 	/* Found matching node. */
900 	if (geom != NULL) {
901 		G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name);
902 		result = G_RAID_MD_TASTE_EXISTING;
903 
904 	} else if (spare) { /* Not found needy node -- left for later. */
905 		G_RAID_DEBUG(1, "Spare is not needed at this time");
906 		goto fail1;
907 
908 	} else { /* Not found matching node -- create one. */
909 		result = G_RAID_MD_TASTE_NEW;
910 		mdi->mdio_config_id = jmicron_meta_config_id(meta);
911 		snprintf(name, sizeof(name), "JMicron-%08x",
912 		    mdi->mdio_config_id);
913 		sc = g_raid_create_node(mp, name, md);
914 		md->mdo_softc = sc;
915 		geom = sc->sc_geom;
916 		callout_init(&mdi->mdio_start_co, 1);
917 		callout_reset(&mdi->mdio_start_co, g_raid_start_timeout * hz,
918 		    g_raid_jmicron_go, sc);
919 		mdi->mdio_rootmount = root_mount_hold("GRAID-JMicron");
920 		G_RAID_DEBUG1(1, sc, "root_mount_hold %p", mdi->mdio_rootmount);
921 	}
922 
923 	/* There is no return after this point, so we close passed consumer. */
924 	g_access(cp, -1, 0, 0);
925 
926 	rcp = g_new_consumer(geom);
927 	rcp->flags |= G_CF_DIRECT_RECEIVE;
928 	g_attach(rcp, pp);
929 	if (g_access(rcp, 1, 1, 1) != 0)
930 		; //goto fail1;
931 
932 	g_topology_unlock();
933 	sx_xlock(&sc->sc_lock);
934 
935 	pd = malloc(sizeof(*pd), M_MD_JMICRON, M_WAITOK | M_ZERO);
936 	pd->pd_meta = meta;
937 	if (spare == 2) {
938 		pd->pd_disk_pos = -3;
939 		pd->pd_disk_id = arc4random() & JMICRON_DISK_MASK;
940 	} else {
941 		pd->pd_disk_pos = -1;
942 		pd->pd_disk_id = meta->disk_id;
943 	}
944 	pd->pd_disk_size = pp->mediasize;
945 	disk = g_raid_create_disk(sc);
946 	disk->d_md_data = (void *)pd;
947 	disk->d_consumer = rcp;
948 	rcp->private = disk;
949 
950 	g_raid_get_disk_info(disk);
951 
952 	g_raid_md_jmicron_new_disk(disk);
953 
954 	sx_xunlock(&sc->sc_lock);
955 	g_topology_lock();
956 	*gp = geom;
957 	return (result);
958 fail1:
959 	free(meta, M_MD_JMICRON);
960 	return (G_RAID_MD_TASTE_FAIL);
961 }
962 
963 static int
964 g_raid_md_event_jmicron(struct g_raid_md_object *md,
965     struct g_raid_disk *disk, u_int event)
966 {
967 	struct g_raid_softc *sc;
968 	struct g_raid_subdisk *sd;
969 	struct g_raid_md_jmicron_object *mdi;
970 	struct g_raid_md_jmicron_perdisk *pd;
971 
972 	sc = md->mdo_softc;
973 	mdi = (struct g_raid_md_jmicron_object *)md;
974 	if (disk == NULL) {
975 		switch (event) {
976 		case G_RAID_NODE_E_START:
977 			if (!mdi->mdio_started)
978 				g_raid_md_jmicron_start(sc);
979 			return (0);
980 		}
981 		return (-1);
982 	}
983 	pd = (struct g_raid_md_jmicron_perdisk *)disk->d_md_data;
984 	switch (event) {
985 	case G_RAID_DISK_E_DISCONNECTED:
986 		/* If disk was assigned, just update statuses. */
987 		if (pd->pd_disk_pos >= 0) {
988 			g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
989 			if (disk->d_consumer) {
990 				g_raid_kill_consumer(sc, disk->d_consumer);
991 				disk->d_consumer = NULL;
992 			}
993 			TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
994 				g_raid_change_subdisk_state(sd,
995 				    G_RAID_SUBDISK_S_NONE);
996 				g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
997 				    G_RAID_EVENT_SUBDISK);
998 			}
999 		} else {
1000 			/* Otherwise -- delete. */
1001 			g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
1002 			g_raid_destroy_disk(disk);
1003 		}
1004 
1005 		/* Write updated metadata to all disks. */
1006 		g_raid_md_write_jmicron(md, NULL, NULL, NULL);
1007 
1008 		/* Check if anything left except placeholders. */
1009 		if (g_raid_ndisks(sc, -1) ==
1010 		    g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1011 			g_raid_destroy_node(sc, 0);
1012 		else
1013 			g_raid_md_jmicron_refill(sc);
1014 		return (0);
1015 	}
1016 	return (-2);
1017 }
1018 
1019 static int
1020 g_raid_md_ctl_jmicron(struct g_raid_md_object *md,
1021     struct gctl_req *req)
1022 {
1023 	struct g_raid_softc *sc;
1024 	struct g_raid_volume *vol;
1025 	struct g_raid_subdisk *sd;
1026 	struct g_raid_disk *disk;
1027 	struct g_raid_md_jmicron_object *mdi;
1028 	struct g_raid_md_jmicron_perdisk *pd;
1029 	struct g_consumer *cp;
1030 	struct g_provider *pp;
1031 	char arg[16];
1032 	const char *verb, *volname, *levelname, *diskname;
1033 	int *nargs, *force;
1034 	off_t size, sectorsize, strip;
1035 	intmax_t *sizearg, *striparg;
1036 	int numdisks, i, len, level, qual, update;
1037 	int error;
1038 
1039 	sc = md->mdo_softc;
1040 	mdi = (struct g_raid_md_jmicron_object *)md;
1041 	verb = gctl_get_param(req, "verb", NULL);
1042 	nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
1043 	error = 0;
1044 	if (strcmp(verb, "label") == 0) {
1045 		if (*nargs < 4) {
1046 			gctl_error(req, "Invalid number of arguments.");
1047 			return (-1);
1048 		}
1049 		volname = gctl_get_asciiparam(req, "arg1");
1050 		if (volname == NULL) {
1051 			gctl_error(req, "No volume name.");
1052 			return (-2);
1053 		}
1054 		levelname = gctl_get_asciiparam(req, "arg2");
1055 		if (levelname == NULL) {
1056 			gctl_error(req, "No RAID level.");
1057 			return (-3);
1058 		}
1059 		if (strcasecmp(levelname, "RAID5") == 0)
1060 			levelname = "RAID5-LA";
1061 		if (g_raid_volume_str2level(levelname, &level, &qual)) {
1062 			gctl_error(req, "Unknown RAID level '%s'.", levelname);
1063 			return (-4);
1064 		}
1065 		numdisks = *nargs - 3;
1066 		force = gctl_get_paraml(req, "force", sizeof(*force));
1067 		if (!g_raid_md_jmicron_supported(level, qual, numdisks,
1068 		    force ? *force : 0)) {
1069 			gctl_error(req, "Unsupported RAID level "
1070 			    "(0x%02x/0x%02x), or number of disks (%d).",
1071 			    level, qual, numdisks);
1072 			return (-5);
1073 		}
1074 
1075 		/* Search for disks, connect them and probe. */
1076 		size = 0x7fffffffffffffffllu;
1077 		sectorsize = 0;
1078 		for (i = 0; i < numdisks; i++) {
1079 			snprintf(arg, sizeof(arg), "arg%d", i + 3);
1080 			diskname = gctl_get_asciiparam(req, arg);
1081 			if (diskname == NULL) {
1082 				gctl_error(req, "No disk name (%s).", arg);
1083 				error = -6;
1084 				break;
1085 			}
1086 			if (strcmp(diskname, "NONE") == 0) {
1087 				cp = NULL;
1088 				pp = NULL;
1089 			} else {
1090 				g_topology_lock();
1091 				cp = g_raid_open_consumer(sc, diskname);
1092 				if (cp == NULL) {
1093 					gctl_error(req, "Can't open '%s'.",
1094 					    diskname);
1095 					g_topology_unlock();
1096 					error = -7;
1097 					break;
1098 				}
1099 				pp = cp->provider;
1100 			}
1101 			pd = malloc(sizeof(*pd), M_MD_JMICRON, M_WAITOK | M_ZERO);
1102 			pd->pd_disk_pos = i;
1103 			pd->pd_disk_id = arc4random() & JMICRON_DISK_MASK;
1104 			disk = g_raid_create_disk(sc);
1105 			disk->d_md_data = (void *)pd;
1106 			disk->d_consumer = cp;
1107 			if (cp == NULL)
1108 				continue;
1109 			cp->private = disk;
1110 			g_topology_unlock();
1111 
1112 			g_raid_get_disk_info(disk);
1113 
1114 			pd->pd_disk_size = pp->mediasize;
1115 			if (size > pp->mediasize)
1116 				size = pp->mediasize;
1117 			if (sectorsize < pp->sectorsize)
1118 				sectorsize = pp->sectorsize;
1119 		}
1120 		if (error != 0)
1121 			return (error);
1122 
1123 		if (sectorsize <= 0) {
1124 			gctl_error(req, "Can't get sector size.");
1125 			return (-8);
1126 		}
1127 
1128 		/* Reserve space for metadata. */
1129 		size -= sectorsize;
1130 
1131 		/* Handle size argument. */
1132 		len = sizeof(*sizearg);
1133 		sizearg = gctl_get_param(req, "size", &len);
1134 		if (sizearg != NULL && len == sizeof(*sizearg) &&
1135 		    *sizearg > 0) {
1136 			if (*sizearg > size) {
1137 				gctl_error(req, "Size too big %lld > %lld.",
1138 				    (long long)*sizearg, (long long)size);
1139 				return (-9);
1140 			}
1141 			size = *sizearg;
1142 		}
1143 
1144 		/* Handle strip argument. */
1145 		strip = 131072;
1146 		len = sizeof(*striparg);
1147 		striparg = gctl_get_param(req, "strip", &len);
1148 		if (striparg != NULL && len == sizeof(*striparg) &&
1149 		    *striparg > 0) {
1150 			if (*striparg < sectorsize) {
1151 				gctl_error(req, "Strip size too small.");
1152 				return (-10);
1153 			}
1154 			if (*striparg % sectorsize != 0) {
1155 				gctl_error(req, "Incorrect strip size.");
1156 				return (-11);
1157 			}
1158 			if (strip > 65535 * sectorsize) {
1159 				gctl_error(req, "Strip size too big.");
1160 				return (-12);
1161 			}
1162 			strip = *striparg;
1163 		}
1164 
1165 		/* Round size down to strip or sector. */
1166 		if (level == G_RAID_VOLUME_RL_RAID1)
1167 			size -= (size % sectorsize);
1168 		else if (level == G_RAID_VOLUME_RL_RAID1E &&
1169 		    (numdisks & 1) != 0)
1170 			size -= (size % (2 * strip));
1171 		else
1172 			size -= (size % strip);
1173 		if (size <= 0) {
1174 			gctl_error(req, "Size too small.");
1175 			return (-13);
1176 		}
1177 		if (size > 0xffffffffffffllu * sectorsize) {
1178 			gctl_error(req, "Size too big.");
1179 			return (-14);
1180 		}
1181 
1182 		/* We have all we need, create things: volume, ... */
1183 		mdi->mdio_total_disks = numdisks;
1184 		mdi->mdio_started = 1;
1185 		vol = g_raid_create_volume(sc, volname, -1);
1186 		vol->v_md_data = (void *)(intptr_t)0;
1187 		vol->v_raid_level = level;
1188 		vol->v_raid_level_qualifier = qual;
1189 		vol->v_strip_size = strip;
1190 		vol->v_disks_count = numdisks;
1191 		if (level == G_RAID_VOLUME_RL_RAID0 ||
1192 		    level == G_RAID_VOLUME_RL_CONCAT ||
1193 		    level == G_RAID_VOLUME_RL_SINGLE)
1194 			vol->v_mediasize = size * numdisks;
1195 		else if (level == G_RAID_VOLUME_RL_RAID1)
1196 			vol->v_mediasize = size;
1197 		else if (level == G_RAID_VOLUME_RL_RAID5)
1198 			vol->v_mediasize = size * (numdisks - 1);
1199 		else { /* RAID1E */
1200 			vol->v_mediasize = ((size * numdisks) / strip / 2) *
1201 			    strip;
1202 		}
1203 		vol->v_sectorsize = sectorsize;
1204 		g_raid_start_volume(vol);
1205 
1206 		/* , and subdisks. */
1207 		TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1208 			pd = (struct g_raid_md_jmicron_perdisk *)disk->d_md_data;
1209 			sd = &vol->v_subdisks[pd->pd_disk_pos];
1210 			sd->sd_disk = disk;
1211 			sd->sd_offset = 0;
1212 			sd->sd_size = size;
1213 			TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1214 			if (sd->sd_disk->d_consumer != NULL) {
1215 				g_raid_change_disk_state(disk,
1216 				    G_RAID_DISK_S_ACTIVE);
1217 				g_raid_change_subdisk_state(sd,
1218 				    G_RAID_SUBDISK_S_ACTIVE);
1219 				g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
1220 				    G_RAID_EVENT_SUBDISK);
1221 			} else {
1222 				g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1223 			}
1224 		}
1225 
1226 		/* Write metadata based on created entities. */
1227 		G_RAID_DEBUG1(0, sc, "Array started.");
1228 		g_raid_md_write_jmicron(md, NULL, NULL, NULL);
1229 
1230 		/* Pickup any STALE/SPARE disks to refill array if needed. */
1231 		g_raid_md_jmicron_refill(sc);
1232 
1233 		g_raid_event_send(vol, G_RAID_VOLUME_E_START,
1234 		    G_RAID_EVENT_VOLUME);
1235 		return (0);
1236 	}
1237 	if (strcmp(verb, "delete") == 0) {
1238 		/* Check if some volume is still open. */
1239 		force = gctl_get_paraml(req, "force", sizeof(*force));
1240 		if (force != NULL && *force == 0 &&
1241 		    g_raid_nopens(sc) != 0) {
1242 			gctl_error(req, "Some volume is still open.");
1243 			return (-4);
1244 		}
1245 
1246 		TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1247 			if (disk->d_consumer)
1248 				jmicron_meta_erase(disk->d_consumer);
1249 		}
1250 		g_raid_destroy_node(sc, 0);
1251 		return (0);
1252 	}
1253 	if (strcmp(verb, "remove") == 0 ||
1254 	    strcmp(verb, "fail") == 0) {
1255 		if (*nargs < 2) {
1256 			gctl_error(req, "Invalid number of arguments.");
1257 			return (-1);
1258 		}
1259 		for (i = 1; i < *nargs; i++) {
1260 			snprintf(arg, sizeof(arg), "arg%d", i);
1261 			diskname = gctl_get_asciiparam(req, arg);
1262 			if (diskname == NULL) {
1263 				gctl_error(req, "No disk name (%s).", arg);
1264 				error = -2;
1265 				break;
1266 			}
1267 			if (strncmp(diskname, _PATH_DEV, 5) == 0)
1268 				diskname += 5;
1269 
1270 			TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1271 				if (disk->d_consumer != NULL &&
1272 				    disk->d_consumer->provider != NULL &&
1273 				    strcmp(disk->d_consumer->provider->name,
1274 				     diskname) == 0)
1275 					break;
1276 			}
1277 			if (disk == NULL) {
1278 				gctl_error(req, "Disk '%s' not found.",
1279 				    diskname);
1280 				error = -3;
1281 				break;
1282 			}
1283 
1284 			if (strcmp(verb, "fail") == 0) {
1285 				g_raid_md_fail_disk_jmicron(md, NULL, disk);
1286 				continue;
1287 			}
1288 
1289 			pd = (struct g_raid_md_jmicron_perdisk *)disk->d_md_data;
1290 
1291 			/* Erase metadata on deleting disk. */
1292 			jmicron_meta_erase(disk->d_consumer);
1293 
1294 			/* If disk was assigned, just update statuses. */
1295 			if (pd->pd_disk_pos >= 0) {
1296 				g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1297 				g_raid_kill_consumer(sc, disk->d_consumer);
1298 				disk->d_consumer = NULL;
1299 				TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
1300 					g_raid_change_subdisk_state(sd,
1301 					    G_RAID_SUBDISK_S_NONE);
1302 					g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
1303 					    G_RAID_EVENT_SUBDISK);
1304 				}
1305 			} else {
1306 				/* Otherwise -- delete. */
1307 				g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
1308 				g_raid_destroy_disk(disk);
1309 			}
1310 		}
1311 
1312 		/* Write updated metadata to remaining disks. */
1313 		g_raid_md_write_jmicron(md, NULL, NULL, NULL);
1314 
1315 		/* Check if anything left except placeholders. */
1316 		if (g_raid_ndisks(sc, -1) ==
1317 		    g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1318 			g_raid_destroy_node(sc, 0);
1319 		else
1320 			g_raid_md_jmicron_refill(sc);
1321 		return (error);
1322 	}
1323 	if (strcmp(verb, "insert") == 0) {
1324 		if (*nargs < 2) {
1325 			gctl_error(req, "Invalid number of arguments.");
1326 			return (-1);
1327 		}
1328 		update = 0;
1329 		for (i = 1; i < *nargs; i++) {
1330 			/* Get disk name. */
1331 			snprintf(arg, sizeof(arg), "arg%d", i);
1332 			diskname = gctl_get_asciiparam(req, arg);
1333 			if (diskname == NULL) {
1334 				gctl_error(req, "No disk name (%s).", arg);
1335 				error = -3;
1336 				break;
1337 			}
1338 
1339 			/* Try to find provider with specified name. */
1340 			g_topology_lock();
1341 			cp = g_raid_open_consumer(sc, diskname);
1342 			if (cp == NULL) {
1343 				gctl_error(req, "Can't open disk '%s'.",
1344 				    diskname);
1345 				g_topology_unlock();
1346 				error = -4;
1347 				break;
1348 			}
1349 			pp = cp->provider;
1350 
1351 			pd = malloc(sizeof(*pd), M_MD_JMICRON, M_WAITOK | M_ZERO);
1352 			pd->pd_disk_pos = -3;
1353 			pd->pd_disk_id = arc4random() & JMICRON_DISK_MASK;
1354 			pd->pd_disk_size = pp->mediasize;
1355 
1356 			disk = g_raid_create_disk(sc);
1357 			disk->d_consumer = cp;
1358 			disk->d_md_data = (void *)pd;
1359 			cp->private = disk;
1360 			g_topology_unlock();
1361 
1362 			g_raid_get_disk_info(disk);
1363 
1364 			/* Welcome the "new" disk. */
1365 			update += g_raid_md_jmicron_start_disk(disk);
1366 			if (disk->d_state != G_RAID_DISK_S_ACTIVE &&
1367 			    disk->d_state != G_RAID_DISK_S_SPARE) {
1368 				gctl_error(req, "Disk '%s' doesn't fit.",
1369 				    diskname);
1370 				g_raid_destroy_disk(disk);
1371 				error = -8;
1372 				break;
1373 			}
1374 		}
1375 
1376 		/* Write new metadata if we changed something. */
1377 		if (update)
1378 			g_raid_md_write_jmicron(md, NULL, NULL, NULL);
1379 		return (error);
1380 	}
1381 	gctl_error(req, "Command '%s' is not supported.", verb);
1382 	return (-100);
1383 }
1384 
1385 static int
1386 g_raid_md_write_jmicron(struct g_raid_md_object *md, struct g_raid_volume *tvol,
1387     struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
1388 {
1389 	struct g_raid_softc *sc;
1390 	struct g_raid_volume *vol;
1391 	struct g_raid_subdisk *sd;
1392 	struct g_raid_disk *disk;
1393 	struct g_raid_md_jmicron_object *mdi;
1394 	struct g_raid_md_jmicron_perdisk *pd;
1395 	struct jmicron_raid_conf *meta;
1396 	int i, spares;
1397 
1398 	sc = md->mdo_softc;
1399 	mdi = (struct g_raid_md_jmicron_object *)md;
1400 
1401 	if (sc->sc_stopping == G_RAID_DESTROY_HARD)
1402 		return (0);
1403 
1404 	/* There is only one volume. */
1405 	vol = TAILQ_FIRST(&sc->sc_volumes);
1406 
1407 	/* Fill global fields. */
1408 	meta = malloc(sizeof(*meta), M_MD_JMICRON, M_WAITOK | M_ZERO);
1409 	strncpy(meta->signature, JMICRON_MAGIC, 2);
1410 	meta->version = JMICRON_VERSION;
1411 	jmicron_meta_put_name(meta, vol->v_name);
1412 	if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0)
1413 		meta->type = JMICRON_T_RAID0;
1414 	else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
1415 		meta->type = JMICRON_T_RAID1;
1416 	else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E)
1417 		meta->type = JMICRON_T_RAID01;
1418 	else if (vol->v_raid_level == G_RAID_VOLUME_RL_CONCAT ||
1419 	    vol->v_raid_level == G_RAID_VOLUME_RL_SINGLE)
1420 		meta->type = JMICRON_T_CONCAT;
1421 	else
1422 		meta->type = JMICRON_T_RAID5;
1423 	meta->stripe_shift = fls(vol->v_strip_size / 2048);
1424 	meta->flags = JMICRON_F_READY | JMICRON_F_BOOTABLE;
1425 	for (i = 0; i < vol->v_disks_count; i++) {
1426 		sd = &vol->v_subdisks[i];
1427 		if (sd->sd_disk == NULL || sd->sd_disk->d_md_data == NULL)
1428 			meta->disks[i] = 0xffffffff;
1429 		else {
1430 			pd = (struct g_raid_md_jmicron_perdisk *)
1431 			    sd->sd_disk->d_md_data;
1432 			meta->disks[i] = pd->pd_disk_id;
1433 		}
1434 		if (sd->sd_state < G_RAID_SUBDISK_S_STALE)
1435 			meta->flags |= JMICRON_F_BADSEC;
1436 		if (vol->v_dirty)
1437 			meta->flags |= JMICRON_F_UNSYNC;
1438 	}
1439 
1440 	/* Put spares to their slots. */
1441 	spares = 0;
1442 	TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1443 		pd = (struct g_raid_md_jmicron_perdisk *)disk->d_md_data;
1444 		if (disk->d_state != G_RAID_DISK_S_SPARE)
1445 			continue;
1446 		meta->spare[spares] = pd->pd_disk_id;
1447 		if (++spares >= 2)
1448 			break;
1449 	}
1450 
1451 	/* We are done. Print meta data and store them to disks. */
1452 	if (mdi->mdio_meta != NULL)
1453 		free(mdi->mdio_meta, M_MD_JMICRON);
1454 	mdi->mdio_meta = meta;
1455 	TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1456 		pd = (struct g_raid_md_jmicron_perdisk *)disk->d_md_data;
1457 		if (disk->d_state != G_RAID_DISK_S_ACTIVE &&
1458 		    disk->d_state != G_RAID_DISK_S_SPARE)
1459 			continue;
1460 		if (pd->pd_meta != NULL) {
1461 			free(pd->pd_meta, M_MD_JMICRON);
1462 			pd->pd_meta = NULL;
1463 		}
1464 		pd->pd_meta = jmicron_meta_copy(meta);
1465 		pd->pd_meta->disk_id = pd->pd_disk_id;
1466 		if ((sd = TAILQ_FIRST(&disk->d_subdisks)) != NULL) {
1467 			pd->pd_meta->offset =
1468 			    (sd->sd_offset / 512) / 16;
1469 			pd->pd_meta->disk_sectors_high =
1470 			    (sd->sd_size / 512) >> 16;
1471 			pd->pd_meta->disk_sectors_low =
1472 			    (sd->sd_size / 512) & 0xffff;
1473 			if (sd->sd_state < G_RAID_SUBDISK_S_STALE)
1474 				pd->pd_meta->flags &= ~JMICRON_F_BADSEC;
1475 			else if (sd->sd_state < G_RAID_SUBDISK_S_ACTIVE)
1476 				pd->pd_meta->flags |= JMICRON_F_UNSYNC;
1477 		}
1478 		G_RAID_DEBUG(1, "Writing JMicron metadata to %s",
1479 		    g_raid_get_diskname(disk));
1480 		g_raid_md_jmicron_print(pd->pd_meta);
1481 		jmicron_meta_write(disk->d_consumer, pd->pd_meta);
1482 	}
1483 	return (0);
1484 }
1485 
1486 static int
1487 g_raid_md_fail_disk_jmicron(struct g_raid_md_object *md,
1488     struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
1489 {
1490 	struct g_raid_softc *sc;
1491 	struct g_raid_md_jmicron_perdisk *pd;
1492 	struct g_raid_subdisk *sd;
1493 
1494 	sc = md->mdo_softc;
1495 	pd = (struct g_raid_md_jmicron_perdisk *)tdisk->d_md_data;
1496 
1497 	/* We can't fail disk that is not a part of array now. */
1498 	if (pd->pd_disk_pos < 0)
1499 		return (-1);
1500 
1501 	if (tdisk->d_consumer)
1502 		jmicron_meta_erase(tdisk->d_consumer);
1503 
1504 	/* Change states. */
1505 	g_raid_change_disk_state(tdisk, G_RAID_DISK_S_FAILED);
1506 	TAILQ_FOREACH(sd, &tdisk->d_subdisks, sd_next) {
1507 		g_raid_change_subdisk_state(sd,
1508 		    G_RAID_SUBDISK_S_FAILED);
1509 		g_raid_event_send(sd, G_RAID_SUBDISK_E_FAILED,
1510 		    G_RAID_EVENT_SUBDISK);
1511 	}
1512 
1513 	/* Write updated metadata to remaining disks. */
1514 	g_raid_md_write_jmicron(md, NULL, NULL, tdisk);
1515 
1516 	/* Check if anything left except placeholders. */
1517 	if (g_raid_ndisks(sc, -1) ==
1518 	    g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1519 		g_raid_destroy_node(sc, 0);
1520 	else
1521 		g_raid_md_jmicron_refill(sc);
1522 	return (0);
1523 }
1524 
1525 static int
1526 g_raid_md_free_disk_jmicron(struct g_raid_md_object *md,
1527     struct g_raid_disk *disk)
1528 {
1529 	struct g_raid_md_jmicron_perdisk *pd;
1530 
1531 	pd = (struct g_raid_md_jmicron_perdisk *)disk->d_md_data;
1532 	if (pd->pd_meta != NULL) {
1533 		free(pd->pd_meta, M_MD_JMICRON);
1534 		pd->pd_meta = NULL;
1535 	}
1536 	free(pd, M_MD_JMICRON);
1537 	disk->d_md_data = NULL;
1538 	return (0);
1539 }
1540 
1541 static int
1542 g_raid_md_free_jmicron(struct g_raid_md_object *md)
1543 {
1544 	struct g_raid_md_jmicron_object *mdi;
1545 
1546 	mdi = (struct g_raid_md_jmicron_object *)md;
1547 	if (!mdi->mdio_started) {
1548 		mdi->mdio_started = 0;
1549 		callout_stop(&mdi->mdio_start_co);
1550 		G_RAID_DEBUG1(1, md->mdo_softc,
1551 		    "root_mount_rel %p", mdi->mdio_rootmount);
1552 		root_mount_rel(mdi->mdio_rootmount);
1553 		mdi->mdio_rootmount = NULL;
1554 	}
1555 	if (mdi->mdio_meta != NULL) {
1556 		free(mdi->mdio_meta, M_MD_JMICRON);
1557 		mdi->mdio_meta = NULL;
1558 	}
1559 	return (0);
1560 }
1561 
1562 G_RAID_MD_DECLARE(jmicron, "JMicron");
1563