2 * Copyright (c) 2011 Alexander Motin <mav@FreeBSD.org>
3 * Copyright (c) 2000 - 2008 Søren Schmidt <sos@FreeBSD.org>
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28 #include <sys/cdefs.h>
29 __FBSDID("$FreeBSD$");
31 #include <sys/param.h>
33 #include <sys/endian.h>
34 #include <sys/kernel.h>
36 #include <sys/limits.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/raid/g_raid.h"
44 #include "g_raid_md_if.h"
46 static MALLOC_DEFINE(M_MD_SII, "md_sii_data", "GEOM_RAID SiI metadata");
48 struct sii_raid_conf {
49 uint16_t ata_params_00_53[54];
50 uint64_t total_sectors; /* 54 - 57 */
51 uint16_t ata_params_58_81[72];
52 uint16_t product_id; /* 130 */
53 uint16_t vendor_id; /* 131 */
54 uint16_t version_minor; /* 132 */
55 uint16_t version_major; /* 133 */
56 uint8_t timestamp[6]; /* 134 - 136 */
57 uint16_t strip_sectors; /* 137 */
59 uint8_t disk_number; /* 139 */
61 #define SII_T_RAID0 0x00
62 #define SII_T_RAID1 0x01
63 #define SII_T_RAID01 0x02
64 #define SII_T_SPARE 0x03
65 #define SII_T_CONCAT 0x04
66 #define SII_T_RAID5 0x10
67 #define SII_T_RESERVED 0xfd
68 #define SII_T_JBOD 0xff
70 uint8_t raid0_disks; /* 140 */
72 uint8_t raid1_disks; /* 141 */
74 uint64_t rebuild_lba; /* 142 - 145 */
75 uint32_t generation; /* 146 - 147 */
76 uint8_t disk_status; /* 148 */
77 #define SII_S_CURRENT 0x01
78 #define SII_S_REBUILD 0x02
79 #define SII_S_DROPPED 0x03
80 #define SII_S_REMOVED 0x04
83 #define SII_S_ONLINE 0x01
84 #define SII_S_AVAILABLE 0x02
86 uint8_t raid_location; /* 149 */
87 uint8_t disk_location;
88 uint8_t auto_rebuild; /* 150 */
89 #define SII_R_REBUILD 0x00
90 #define SII_R_NOREBUILD 0xff
93 uint8_t name[16]; /* 151 - 158 */
94 uint16_t checksum; /* 159 */
95 uint16_t ata_params_160_255[96];
98 struct g_raid_md_sii_perdisk {
99 struct sii_raid_conf *pd_meta;
104 struct g_raid_md_sii_object {
105 struct g_raid_md_object mdio_base;
106 uint8_t mdio_timestamp[6];
107 uint8_t mdio_location;
108 uint32_t mdio_generation;
109 struct sii_raid_conf *mdio_meta;
110 struct callout mdio_start_co; /* STARTING state timer. */
111 int mdio_total_disks;
112 int mdio_disks_present;
115 struct root_hold_token *mdio_rootmount; /* Root mount delay token. */
118 static g_raid_md_create_t g_raid_md_create_sii;
119 static g_raid_md_taste_t g_raid_md_taste_sii;
120 static g_raid_md_event_t g_raid_md_event_sii;
121 static g_raid_md_ctl_t g_raid_md_ctl_sii;
122 static g_raid_md_write_t g_raid_md_write_sii;
123 static g_raid_md_fail_disk_t g_raid_md_fail_disk_sii;
124 static g_raid_md_free_disk_t g_raid_md_free_disk_sii;
125 static g_raid_md_free_t g_raid_md_free_sii;
127 static kobj_method_t g_raid_md_sii_methods[] = {
128 KOBJMETHOD(g_raid_md_create, g_raid_md_create_sii),
129 KOBJMETHOD(g_raid_md_taste, g_raid_md_taste_sii),
130 KOBJMETHOD(g_raid_md_event, g_raid_md_event_sii),
131 KOBJMETHOD(g_raid_md_ctl, g_raid_md_ctl_sii),
132 KOBJMETHOD(g_raid_md_write, g_raid_md_write_sii),
133 KOBJMETHOD(g_raid_md_fail_disk, g_raid_md_fail_disk_sii),
134 KOBJMETHOD(g_raid_md_free_disk, g_raid_md_free_disk_sii),
135 KOBJMETHOD(g_raid_md_free, g_raid_md_free_sii),
139 static struct g_raid_md_class g_raid_md_sii_class = {
141 g_raid_md_sii_methods,
142 sizeof(struct g_raid_md_sii_object),
148 g_raid_md_sii_print(struct sii_raid_conf *meta)
151 if (g_raid_debug < 1)
154 printf("********* ATA SiI RAID Metadata *********\n");
155 printf("total_sectors %llu\n",
156 (long long unsigned)meta->total_sectors);
157 printf("product_id 0x%04x\n", meta->product_id);
158 printf("vendor_id 0x%04x\n", meta->vendor_id);
159 printf("version_minor 0x%04x\n", meta->version_minor);
160 printf("version_major 0x%04x\n", meta->version_major);
161 printf("timestamp 0x%02x%02x%02x%02x%02x%02x\n",
162 meta->timestamp[5], meta->timestamp[4], meta->timestamp[3],
163 meta->timestamp[2], meta->timestamp[1], meta->timestamp[0]);
164 printf("strip_sectors %d\n", meta->strip_sectors);
165 printf("disk_number %d\n", meta->disk_number);
166 printf("type 0x%02x\n", meta->type);
167 printf("raid0_disks %d\n", meta->raid0_disks);
168 printf("raid0_ident %d\n", meta->raid0_ident);
169 printf("raid1_disks %d\n", meta->raid1_disks);
170 printf("raid1_ident %d\n", meta->raid1_ident);
171 printf("rebuild_lba %llu\n",
172 (long long unsigned)meta->rebuild_lba);
173 printf("generation %d\n", meta->generation);
174 printf("disk_status %d\n", meta->disk_status);
175 printf("raid_status %d\n", meta->raid_status);
176 printf("raid_location %d\n", meta->raid_location);
177 printf("disk_location %d\n", meta->disk_location);
178 printf("auto_rebuild %d\n", meta->auto_rebuild);
179 printf("name <%.16s>\n", meta->name);
180 printf("checksum 0x%04x\n", meta->checksum);
181 printf("=================================================\n");
184 static struct sii_raid_conf *
185 sii_meta_copy(struct sii_raid_conf *meta)
187 struct sii_raid_conf *nmeta;
189 nmeta = malloc(sizeof(*meta), M_MD_SII, M_WAITOK);
190 memcpy(nmeta, meta, sizeof(*meta));
195 sii_meta_total_disks(struct sii_raid_conf *meta)
198 switch (meta->type) {
202 return (meta->raid0_disks);
204 return (meta->raid1_disks);
206 return (meta->raid0_disks * meta->raid1_disks);
215 sii_meta_disk_pos(struct sii_raid_conf *meta, struct sii_raid_conf *pdmeta)
218 if (pdmeta->type == SII_T_SPARE)
221 if (memcmp(&meta->timestamp, &pdmeta->timestamp, 6) != 0)
224 switch (pdmeta->type) {
229 return (pdmeta->disk_number);
231 return (pdmeta->raid1_ident * pdmeta->raid1_disks +
232 pdmeta->raid0_ident);
240 sii_meta_get_name(struct sii_raid_conf *meta, char *buf)
244 strncpy(buf, meta->name, 16);
246 for (i = 15; i >= 0; i--) {
254 sii_meta_put_name(struct sii_raid_conf *meta, char *buf)
257 memset(meta->name, 0x20, 16);
258 memcpy(meta->name, buf, MIN(strlen(buf), 16));
261 static struct sii_raid_conf *
262 sii_meta_read(struct g_consumer *cp)
264 struct g_provider *pp;
265 struct sii_raid_conf *meta;
268 uint16_t checksum, *ptr;
272 /* Read the anchor sector. */
273 buf = g_read_data(cp,
274 pp->mediasize - pp->sectorsize, pp->sectorsize, &error);
276 G_RAID_DEBUG(1, "Cannot read metadata from %s (error=%d).",
280 meta = malloc(sizeof(*meta), M_MD_SII, M_WAITOK);
281 memcpy(meta, buf, min(sizeof(*meta), pp->sectorsize));
284 /* Check vendor ID. */
285 if (meta->vendor_id != 0x1095) {
286 G_RAID_DEBUG(1, "SiI vendor ID check failed on %s (0x%04x)",
287 pp->name, meta->vendor_id);
288 free(meta, M_MD_SII);
292 /* Check metadata major version. */
293 if (meta->version_major != 2) {
294 G_RAID_DEBUG(1, "SiI version check failed on %s (%d.%d)",
295 pp->name, meta->version_major, meta->version_minor);
296 free(meta, M_MD_SII);
300 /* Check metadata checksum. */
301 for (checksum = 0, ptr = (uint16_t *)meta, i = 0; i <= 159; i++)
304 G_RAID_DEBUG(1, "SiI checksum check failed on %s", pp->name);
305 free(meta, M_MD_SII);
309 /* Check raid type. */
310 if (meta->type != SII_T_RAID0 && meta->type != SII_T_RAID1 &&
311 meta->type != SII_T_RAID01 && meta->type != SII_T_SPARE &&
312 meta->type != SII_T_RAID5 && meta->type != SII_T_CONCAT &&
313 meta->type != SII_T_JBOD) {
314 G_RAID_DEBUG(1, "SiI unknown RAID level on %s (0x%02x)",
315 pp->name, meta->type);
316 free(meta, M_MD_SII);
324 sii_meta_write(struct g_consumer *cp, struct sii_raid_conf *meta)
326 struct g_provider *pp;
329 uint16_t checksum, *ptr;
333 /* Recalculate checksum for case if metadata were changed. */
335 for (checksum = 0, ptr = (uint16_t *)meta, i = 0; i < 159; i++)
337 meta->checksum -= checksum;
339 /* Create and fill buffer. */
340 buf = malloc(pp->sectorsize, M_MD_SII, M_WAITOK | M_ZERO);
341 memcpy(buf, meta, sizeof(*meta));
343 /* Write 4 copies of metadata. */
344 for (i = 0; i < 4; i++) {
345 error = g_write_data(cp,
346 pp->mediasize - (pp->sectorsize * (1 + 0x200 * i)),
347 buf, pp->sectorsize);
349 G_RAID_DEBUG(1, "Cannot write metadata to %s (error=%d).",
360 sii_meta_erase(struct g_consumer *cp)
362 struct g_provider *pp;
367 buf = malloc(pp->sectorsize, M_MD_SII, M_WAITOK | M_ZERO);
368 /* Write 4 copies of metadata. */
369 for (i = 0; i < 4; i++) {
370 error = g_write_data(cp,
371 pp->mediasize - (pp->sectorsize * (1 + 0x200 * i)),
372 buf, pp->sectorsize);
374 G_RAID_DEBUG(1, "Cannot erase metadata on %s (error=%d).",
383 sii_meta_write_spare(struct g_consumer *cp)
385 struct sii_raid_conf *meta;
388 meta = malloc(sizeof(*meta), M_MD_SII, M_WAITOK | M_ZERO);
389 meta->total_sectors = cp->provider->mediasize /
390 cp->provider->sectorsize - 0x800;
391 meta->vendor_id = 0x1095;
392 meta->version_minor = 0;
393 meta->version_major = 2;
394 meta->timestamp[0] = arc4random();
395 meta->timestamp[1] = arc4random();
396 meta->timestamp[2] = arc4random();
397 meta->timestamp[3] = arc4random();
398 meta->timestamp[4] = arc4random();
399 meta->timestamp[5] = arc4random();
400 meta->type = SII_T_SPARE;
401 meta->generation = 1;
402 meta->raid1_ident = 0xff;
403 meta->raid_location = arc4random();
404 error = sii_meta_write(cp, meta);
405 free(meta, M_MD_SII);
409 static struct g_raid_disk *
410 g_raid_md_sii_get_disk(struct g_raid_softc *sc, int id)
412 struct g_raid_disk *disk;
413 struct g_raid_md_sii_perdisk *pd;
415 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
416 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data;
417 if (pd->pd_disk_pos == id)
424 g_raid_md_sii_supported(int level, int qual, int disks, int force)
430 case G_RAID_VOLUME_RL_RAID0:
433 if (!force && (disks < 2 || disks > 6))
436 case G_RAID_VOLUME_RL_RAID1:
439 if (!force && (disks != 2))
442 case G_RAID_VOLUME_RL_RAID1E:
447 if (!force && (disks < 4))
450 case G_RAID_VOLUME_RL_SINGLE:
454 case G_RAID_VOLUME_RL_CONCAT:
458 case G_RAID_VOLUME_RL_RAID5:
461 if (qual != G_RAID_VOLUME_RLQ_R5LS)
467 if (level != G_RAID_VOLUME_RL_RAID5 && qual != G_RAID_VOLUME_RLQ_NONE)
473 g_raid_md_sii_start_disk(struct g_raid_disk *disk)
475 struct g_raid_softc *sc;
476 struct g_raid_subdisk *sd, *tmpsd;
477 struct g_raid_disk *olddisk, *tmpdisk;
478 struct g_raid_md_object *md;
479 struct g_raid_md_sii_object *mdi;
480 struct g_raid_md_sii_perdisk *pd, *oldpd;
481 struct sii_raid_conf *meta;
482 int disk_pos, resurrection = 0;
486 mdi = (struct g_raid_md_sii_object *)md;
487 meta = mdi->mdio_meta;
488 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data;
491 /* Find disk position in metadata by it's serial. */
492 if (pd->pd_meta != NULL)
493 disk_pos = sii_meta_disk_pos(meta, pd->pd_meta);
497 G_RAID_DEBUG1(1, sc, "Unknown, probably new or stale disk");
498 /* If we are in the start process, that's all for now. */
499 if (!mdi->mdio_started)
502 * If we have already started - try to get use of the disk.
503 * Try to replace OFFLINE disks first, then FAILED.
505 TAILQ_FOREACH(tmpdisk, &sc->sc_disks, d_next) {
506 if (tmpdisk->d_state != G_RAID_DISK_S_OFFLINE &&
507 tmpdisk->d_state != G_RAID_DISK_S_FAILED)
509 /* Make sure this disk is big enough. */
510 TAILQ_FOREACH(sd, &tmpdisk->d_subdisks, sd_next) {
511 if (sd->sd_offset + sd->sd_size + 512 >
514 "Disk too small (%ju < %ju)",
516 sd->sd_offset + sd->sd_size + 512);
522 if (tmpdisk->d_state == G_RAID_DISK_S_OFFLINE) {
525 } else if (olddisk == NULL)
528 if (olddisk == NULL) {
530 if (disk_pos == -3 || pd->pd_disk_pos == -3) {
531 g_raid_change_disk_state(disk,
532 G_RAID_DISK_S_SPARE);
535 g_raid_change_disk_state(disk,
536 G_RAID_DISK_S_STALE);
540 oldpd = (struct g_raid_md_sii_perdisk *)olddisk->d_md_data;
541 disk_pos = oldpd->pd_disk_pos;
545 if (olddisk == NULL) {
546 /* Find placeholder by position. */
547 olddisk = g_raid_md_sii_get_disk(sc, disk_pos);
549 panic("No disk at position %d!", disk_pos);
550 if (olddisk->d_state != G_RAID_DISK_S_OFFLINE) {
551 G_RAID_DEBUG1(1, sc, "More then one disk for pos %d",
553 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE);
556 oldpd = (struct g_raid_md_sii_perdisk *)olddisk->d_md_data;
559 /* Replace failed disk or placeholder with new disk. */
560 TAILQ_FOREACH_SAFE(sd, &olddisk->d_subdisks, sd_next, tmpsd) {
561 TAILQ_REMOVE(&olddisk->d_subdisks, sd, sd_next);
562 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
565 oldpd->pd_disk_pos = -2;
566 pd->pd_disk_pos = disk_pos;
568 /* If it was placeholder -- destroy it. */
569 if (olddisk->d_state == G_RAID_DISK_S_OFFLINE) {
570 g_raid_destroy_disk(olddisk);
572 /* Otherwise, make it STALE_FAILED. */
573 g_raid_change_disk_state(olddisk, G_RAID_DISK_S_STALE_FAILED);
576 /* Welcome the new disk. */
578 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
579 else if (pd->pd_meta->disk_status == SII_S_CURRENT ||
580 pd->pd_meta->disk_status == SII_S_REBUILD)
581 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
583 g_raid_change_disk_state(disk, G_RAID_DISK_S_FAILED);
584 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
587 * Different disks may have different sizes,
588 * in concat mode. Update from real disk size.
590 if (meta->type == SII_T_CONCAT || meta->type == SII_T_JBOD)
591 sd->sd_size = pd->pd_disk_size - 0x800 * 512;
594 /* New or ex-spare disk. */
595 g_raid_change_subdisk_state(sd,
596 G_RAID_SUBDISK_S_NEW);
597 } else if (pd->pd_meta->disk_status == SII_S_REBUILD) {
598 /* Rebuilding disk. */
599 g_raid_change_subdisk_state(sd,
600 G_RAID_SUBDISK_S_REBUILD);
601 if (pd->pd_meta->generation == meta->generation)
602 sd->sd_rebuild_pos = pd->pd_meta->rebuild_lba * 512;
604 sd->sd_rebuild_pos = 0;
605 } else if (pd->pd_meta->disk_status == SII_S_CURRENT) {
606 if (pd->pd_meta->raid_status == SII_S_ONLINE ||
607 pd->pd_meta->generation != meta->generation) {
608 /* Dirty or resyncing disk. */
609 g_raid_change_subdisk_state(sd,
610 G_RAID_SUBDISK_S_STALE);
612 /* Up to date disk. */
613 g_raid_change_subdisk_state(sd,
614 G_RAID_SUBDISK_S_ACTIVE);
617 g_raid_change_subdisk_state(sd,
618 G_RAID_SUBDISK_S_FAILED);
620 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
621 G_RAID_EVENT_SUBDISK);
624 /* Update status of our need for spare. */
625 if (mdi->mdio_started) {
626 mdi->mdio_incomplete =
627 (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) <
628 mdi->mdio_total_disks);
631 return (resurrection);
635 g_disk_md_sii_retaste(void *arg, int pending)
638 G_RAID_DEBUG(1, "Array is not complete, trying to retaste.");
639 g_retaste(&g_raid_class);
644 g_raid_md_sii_refill(struct g_raid_softc *sc)
646 struct g_raid_md_object *md;
647 struct g_raid_md_sii_object *mdi;
648 struct g_raid_disk *disk;
653 mdi = (struct g_raid_md_sii_object *)md;
656 /* Make sure we miss anything. */
657 na = g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE);
658 if (na == mdi->mdio_total_disks)
661 G_RAID_DEBUG1(1, md->mdo_softc,
662 "Array is not complete (%d of %d), "
663 "trying to refill.", na, mdi->mdio_total_disks);
665 /* Try to get use some of STALE disks. */
666 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
667 if (disk->d_state == G_RAID_DISK_S_STALE) {
668 update += g_raid_md_sii_start_disk(disk);
669 if (disk->d_state == G_RAID_DISK_S_ACTIVE)
676 /* Try to get use some of SPARE disks. */
677 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
678 if (disk->d_state == G_RAID_DISK_S_SPARE) {
679 update += g_raid_md_sii_start_disk(disk);
680 if (disk->d_state == G_RAID_DISK_S_ACTIVE)
684 } while (disk != NULL);
686 /* Write new metadata if we changed something. */
688 g_raid_md_write_sii(md, NULL, NULL, NULL);
690 /* Update status of our need for spare. */
691 mdi->mdio_incomplete = (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) <
692 mdi->mdio_total_disks);
694 /* Request retaste hoping to find spare. */
695 if (mdi->mdio_incomplete) {
696 task = malloc(sizeof(struct task),
697 M_MD_SII, M_WAITOK | M_ZERO);
698 TASK_INIT(task, 0, g_disk_md_sii_retaste, task);
699 taskqueue_enqueue(taskqueue_swi, task);
704 g_raid_md_sii_start(struct g_raid_softc *sc)
706 struct g_raid_md_object *md;
707 struct g_raid_md_sii_object *mdi;
708 struct g_raid_md_sii_perdisk *pd;
709 struct sii_raid_conf *meta;
710 struct g_raid_volume *vol;
711 struct g_raid_subdisk *sd;
712 struct g_raid_disk *disk, *best;
715 uint32_t gendiff, bestgendiff;
719 mdi = (struct g_raid_md_sii_object *)md;
720 meta = mdi->mdio_meta;
722 /* Create volumes and subdisks. */
723 sii_meta_get_name(meta, buf);
724 vol = g_raid_create_volume(sc, buf, -1);
725 vol->v_mediasize = (off_t)meta->total_sectors * 512;
726 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE;
727 if (meta->type == SII_T_RAID0) {
728 vol->v_raid_level = G_RAID_VOLUME_RL_RAID0;
729 size = vol->v_mediasize / mdi->mdio_total_disks;
730 } else if (meta->type == SII_T_RAID1) {
731 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1;
732 size = vol->v_mediasize;
733 } else if (meta->type == SII_T_RAID01) {
734 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E;
735 size = vol->v_mediasize / (mdi->mdio_total_disks / 2);
736 } else if (meta->type == SII_T_CONCAT) {
737 if (mdi->mdio_total_disks == 1)
738 vol->v_raid_level = G_RAID_VOLUME_RL_SINGLE;
740 vol->v_raid_level = G_RAID_VOLUME_RL_CONCAT;
742 } else if (meta->type == SII_T_RAID5) {
743 vol->v_raid_level = G_RAID_VOLUME_RL_RAID5;
744 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_R5LS;
745 size = vol->v_mediasize / (mdi->mdio_total_disks - 1);
746 } else if (meta->type == SII_T_JBOD) {
747 vol->v_raid_level = G_RAID_VOLUME_RL_SINGLE;
750 vol->v_raid_level = G_RAID_VOLUME_RL_UNKNOWN;
753 vol->v_strip_size = meta->strip_sectors * 512; //ZZZ
754 vol->v_disks_count = mdi->mdio_total_disks;
755 vol->v_sectorsize = 512; //ZZZ
756 for (j = 0; j < vol->v_disks_count; j++) {
757 sd = &vol->v_subdisks[j];
761 g_raid_start_volume(vol);
763 /* Create disk placeholders to store data for later writing. */
764 for (disk_pos = 0; disk_pos < mdi->mdio_total_disks; disk_pos++) {
765 pd = malloc(sizeof(*pd), M_MD_SII, M_WAITOK | M_ZERO);
766 pd->pd_disk_pos = disk_pos;
767 disk = g_raid_create_disk(sc);
768 disk->d_md_data = (void *)pd;
769 disk->d_state = G_RAID_DISK_S_OFFLINE;
770 sd = &vol->v_subdisks[disk_pos];
772 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
776 * Make all disks found till the moment take their places
777 * in order of their generation numbers.
781 bestgendiff = 0xffffffff;
782 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
783 if (disk->d_state != G_RAID_DISK_S_NONE)
785 pd = disk->d_md_data;
786 if (pd->pd_meta == NULL)
787 gendiff = 0xfffffffe;
789 gendiff = meta->generation -
790 pd->pd_meta->generation;
791 if (gendiff < bestgendiff) {
793 bestgendiff = gendiff;
797 g_raid_md_sii_start_disk(best);
798 } while (best != NULL);
800 mdi->mdio_started = 1;
801 G_RAID_DEBUG1(0, sc, "Array started.");
802 g_raid_md_write_sii(md, NULL, NULL, NULL);
804 /* Pickup any STALE/SPARE disks to refill array if needed. */
805 g_raid_md_sii_refill(sc);
807 g_raid_event_send(vol, G_RAID_VOLUME_E_START, G_RAID_EVENT_VOLUME);
809 callout_stop(&mdi->mdio_start_co);
810 G_RAID_DEBUG1(1, sc, "root_mount_rel %p", mdi->mdio_rootmount);
811 root_mount_rel(mdi->mdio_rootmount);
812 mdi->mdio_rootmount = NULL;
816 g_raid_md_sii_new_disk(struct g_raid_disk *disk)
818 struct g_raid_softc *sc;
819 struct g_raid_md_object *md;
820 struct g_raid_md_sii_object *mdi;
821 struct sii_raid_conf *pdmeta;
822 struct g_raid_md_sii_perdisk *pd;
826 mdi = (struct g_raid_md_sii_object *)md;
827 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data;
828 pdmeta = pd->pd_meta;
830 if (mdi->mdio_started) {
831 if (g_raid_md_sii_start_disk(disk))
832 g_raid_md_write_sii(md, NULL, NULL, NULL);
834 if (mdi->mdio_meta == NULL ||
835 ((int32_t)(pdmeta->generation - mdi->mdio_generation)) > 0) {
836 G_RAID_DEBUG1(1, sc, "Newer disk");
837 if (mdi->mdio_meta != NULL)
838 free(mdi->mdio_meta, M_MD_SII);
839 mdi->mdio_meta = sii_meta_copy(pdmeta);
840 mdi->mdio_generation = mdi->mdio_meta->generation;
841 mdi->mdio_total_disks = sii_meta_total_disks(pdmeta);
842 mdi->mdio_disks_present = 1;
843 } else if (pdmeta->generation == mdi->mdio_generation) {
844 mdi->mdio_disks_present++;
845 G_RAID_DEBUG1(1, sc, "Matching disk (%d of %d up)",
846 mdi->mdio_disks_present,
847 mdi->mdio_total_disks);
849 G_RAID_DEBUG1(1, sc, "Older disk");
852 /* If we collected all needed disks - start array. */
853 if (mdi->mdio_disks_present == mdi->mdio_total_disks)
854 g_raid_md_sii_start(sc);
859 g_raid_sii_go(void *arg)
861 struct g_raid_softc *sc;
862 struct g_raid_md_object *md;
863 struct g_raid_md_sii_object *mdi;
867 mdi = (struct g_raid_md_sii_object *)md;
868 if (!mdi->mdio_started) {
869 G_RAID_DEBUG1(0, sc, "Force array start due to timeout.");
870 g_raid_event_send(sc, G_RAID_NODE_E_START, 0);
875 g_raid_md_create_sii(struct g_raid_md_object *md, struct g_class *mp,
878 struct g_raid_softc *sc;
879 struct g_raid_md_sii_object *mdi;
882 mdi = (struct g_raid_md_sii_object *)md;
883 mdi->mdio_timestamp[5] = arc4random();
884 mdi->mdio_timestamp[4] = arc4random();
885 mdi->mdio_timestamp[3] = arc4random();
886 mdi->mdio_timestamp[2] = arc4random();
887 mdi->mdio_timestamp[1] = arc4random();
888 mdi->mdio_timestamp[0] = arc4random();
889 mdi->mdio_location = arc4random();
890 mdi->mdio_generation = 0;
891 snprintf(name, sizeof(name), "SiI-%02x%02x%02x%02x%02x%02x",
892 mdi->mdio_timestamp[5], mdi->mdio_timestamp[4],
893 mdi->mdio_timestamp[3], mdi->mdio_timestamp[2],
894 mdi->mdio_timestamp[1], mdi->mdio_timestamp[0]);
895 sc = g_raid_create_node(mp, name, md);
897 return (G_RAID_MD_TASTE_FAIL);
900 return (G_RAID_MD_TASTE_NEW);
904 g_raid_md_taste_sii(struct g_raid_md_object *md, struct g_class *mp,
905 struct g_consumer *cp, struct g_geom **gp)
907 struct g_consumer *rcp;
908 struct g_provider *pp;
909 struct g_raid_md_sii_object *mdi, *mdi1;
910 struct g_raid_softc *sc;
911 struct g_raid_disk *disk;
912 struct sii_raid_conf *meta;
913 struct g_raid_md_sii_perdisk *pd;
915 int error, disk_pos, result, spare, len;
919 G_RAID_DEBUG(1, "Tasting SiI on %s", cp->provider->name);
920 mdi = (struct g_raid_md_sii_object *)md;
923 /* Read metadata from device. */
926 if (g_access(cp, 1, 0, 0) != 0)
927 return (G_RAID_MD_TASTE_FAIL);
930 if (pp->geom->rank == 1)
931 g_io_getattr("GEOM::hba_vendor", cp, &len, &vendor);
932 meta = sii_meta_read(cp);
934 g_access(cp, -1, 0, 0);
936 if (g_raid_aggressive_spare) {
937 if (vendor == 0x1095) {
939 "No SiI metadata, forcing spare.");
944 "SiI vendor mismatch 0x%04x != 0x1095",
948 return (G_RAID_MD_TASTE_FAIL);
951 /* Check this disk position in obtained metadata. */
952 disk_pos = sii_meta_disk_pos(meta, meta);
953 if (disk_pos == -1) {
954 G_RAID_DEBUG(1, "SiI disk position not found");
958 /* Metadata valid. Print it. */
959 g_raid_md_sii_print(meta);
960 G_RAID_DEBUG(1, "SiI disk position %d", disk_pos);
961 spare = (meta->type == SII_T_SPARE) ? 1 : 0;
964 /* Search for matching node. */
967 LIST_FOREACH(geom, &mp->geom, geom) {
971 if (sc->sc_stopping != 0)
973 if (sc->sc_md->mdo_class != md->mdo_class)
975 mdi1 = (struct g_raid_md_sii_object *)sc->sc_md;
977 if (mdi1->mdio_incomplete)
980 if (mdi1->mdio_location == meta->raid_location &&
981 memcmp(&mdi1->mdio_timestamp,
982 &meta->timestamp, 6) == 0)
987 /* Found matching node. */
989 G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name);
990 result = G_RAID_MD_TASTE_EXISTING;
992 } else if (spare) { /* Not found needy node -- left for later. */
993 G_RAID_DEBUG(1, "Spare is not needed at this time");
996 } else { /* Not found matching node -- create one. */
997 result = G_RAID_MD_TASTE_NEW;
998 memcpy(&mdi->mdio_timestamp, &meta->timestamp, 6);
999 mdi->mdio_location = meta->raid_location;
1000 snprintf(name, sizeof(name), "SiI-%02x%02x%02x%02x%02x%02x",
1001 mdi->mdio_timestamp[5], mdi->mdio_timestamp[4],
1002 mdi->mdio_timestamp[3], mdi->mdio_timestamp[2],
1003 mdi->mdio_timestamp[1], mdi->mdio_timestamp[0]);
1004 sc = g_raid_create_node(mp, name, md);
1007 callout_init(&mdi->mdio_start_co, 1);
1008 callout_reset(&mdi->mdio_start_co, g_raid_start_timeout * hz,
1010 mdi->mdio_rootmount = root_mount_hold("GRAID-SiI");
1011 G_RAID_DEBUG1(1, sc, "root_mount_hold %p", mdi->mdio_rootmount);
1014 rcp = g_new_consumer(geom);
1016 if (g_access(rcp, 1, 1, 1) != 0)
1019 g_topology_unlock();
1020 sx_xlock(&sc->sc_lock);
1022 pd = malloc(sizeof(*pd), M_MD_SII, M_WAITOK | M_ZERO);
1025 pd->pd_disk_pos = -3;
1027 pd->pd_disk_pos = -1;
1029 pd->pd_disk_size = pp->mediasize;
1030 disk = g_raid_create_disk(sc);
1031 disk->d_md_data = (void *)pd;
1032 disk->d_consumer = rcp;
1033 rcp->private = disk;
1035 /* Read kernel dumping information. */
1036 disk->d_kd.offset = 0;
1037 disk->d_kd.length = OFF_MAX;
1038 len = sizeof(disk->d_kd);
1039 error = g_io_getattr("GEOM::kerneldump", rcp, &len, &disk->d_kd);
1040 if (disk->d_kd.di.dumper == NULL)
1041 G_RAID_DEBUG1(2, sc, "Dumping not supported by %s: %d.",
1042 rcp->provider->name, error);
1044 g_raid_md_sii_new_disk(disk);
1046 sx_xunlock(&sc->sc_lock);
1051 free(meta, M_MD_SII);
1052 return (G_RAID_MD_TASTE_FAIL);
1056 g_raid_md_event_sii(struct g_raid_md_object *md,
1057 struct g_raid_disk *disk, u_int event)
1059 struct g_raid_softc *sc;
1060 struct g_raid_subdisk *sd;
1061 struct g_raid_md_sii_object *mdi;
1062 struct g_raid_md_sii_perdisk *pd;
1065 mdi = (struct g_raid_md_sii_object *)md;
1068 case G_RAID_NODE_E_START:
1069 if (!mdi->mdio_started)
1070 g_raid_md_sii_start(sc);
1075 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data;
1077 case G_RAID_DISK_E_DISCONNECTED:
1078 /* If disk was assigned, just update statuses. */
1079 if (pd->pd_disk_pos >= 0) {
1080 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1081 if (disk->d_consumer) {
1082 g_raid_kill_consumer(sc, disk->d_consumer);
1083 disk->d_consumer = NULL;
1085 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
1086 g_raid_change_subdisk_state(sd,
1087 G_RAID_SUBDISK_S_NONE);
1088 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
1089 G_RAID_EVENT_SUBDISK);
1092 /* Otherwise -- delete. */
1093 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
1094 g_raid_destroy_disk(disk);
1097 /* Write updated metadata to all disks. */
1098 g_raid_md_write_sii(md, NULL, NULL, NULL);
1100 /* Check if anything left except placeholders. */
1101 if (g_raid_ndisks(sc, -1) ==
1102 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1103 g_raid_destroy_node(sc, 0);
1105 g_raid_md_sii_refill(sc);
1112 g_raid_md_ctl_sii(struct g_raid_md_object *md,
1113 struct gctl_req *req)
1115 struct g_raid_softc *sc;
1116 struct g_raid_volume *vol;
1117 struct g_raid_subdisk *sd;
1118 struct g_raid_disk *disk;
1119 struct g_raid_md_sii_object *mdi;
1120 struct g_raid_md_sii_perdisk *pd;
1121 struct g_consumer *cp;
1122 struct g_provider *pp;
1124 const char *verb, *volname, *levelname, *diskname;
1126 off_t size, sectorsize, strip;
1127 intmax_t *sizearg, *striparg;
1128 int numdisks, i, len, level, qual, update;
1132 mdi = (struct g_raid_md_sii_object *)md;
1133 verb = gctl_get_param(req, "verb", NULL);
1134 nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
1136 if (strcmp(verb, "label") == 0) {
1139 gctl_error(req, "Invalid number of arguments.");
1142 volname = gctl_get_asciiparam(req, "arg1");
1143 if (volname == NULL) {
1144 gctl_error(req, "No volume name.");
1147 levelname = gctl_get_asciiparam(req, "arg2");
1148 if (levelname == NULL) {
1149 gctl_error(req, "No RAID level.");
1152 if (strcasecmp(levelname, "RAID5") == 0)
1153 levelname = "RAID5-LS";
1154 if (g_raid_volume_str2level(levelname, &level, &qual)) {
1155 gctl_error(req, "Unknown RAID level '%s'.", levelname);
1158 numdisks = *nargs - 3;
1159 force = gctl_get_paraml(req, "force", sizeof(*force));
1160 if (!g_raid_md_sii_supported(level, qual, numdisks,
1161 force ? *force : 0)) {
1162 gctl_error(req, "Unsupported RAID level "
1163 "(0x%02x/0x%02x), or number of disks (%d).",
1164 level, qual, numdisks);
1168 /* Search for disks, connect them and probe. */
1169 size = 0x7fffffffffffffffllu;
1171 for (i = 0; i < numdisks; i++) {
1172 snprintf(arg, sizeof(arg), "arg%d", i + 3);
1173 diskname = gctl_get_asciiparam(req, arg);
1174 if (diskname == NULL) {
1175 gctl_error(req, "No disk name (%s).", arg);
1179 if (strcmp(diskname, "NONE") == 0) {
1184 cp = g_raid_open_consumer(sc, diskname);
1186 gctl_error(req, "Can't open '%s'.",
1188 g_topology_unlock();
1194 pd = malloc(sizeof(*pd), M_MD_SII, M_WAITOK | M_ZERO);
1195 pd->pd_disk_pos = i;
1196 disk = g_raid_create_disk(sc);
1197 disk->d_md_data = (void *)pd;
1198 disk->d_consumer = cp;
1202 g_topology_unlock();
1204 /* Read kernel dumping information. */
1205 disk->d_kd.offset = 0;
1206 disk->d_kd.length = OFF_MAX;
1207 len = sizeof(disk->d_kd);
1208 g_io_getattr("GEOM::kerneldump", cp, &len, &disk->d_kd);
1209 if (disk->d_kd.di.dumper == NULL)
1210 G_RAID_DEBUG1(2, sc,
1211 "Dumping not supported by %s.",
1212 cp->provider->name);
1214 pd->pd_disk_size = pp->mediasize;
1215 if (size > pp->mediasize)
1216 size = pp->mediasize;
1217 if (sectorsize < pp->sectorsize)
1218 sectorsize = pp->sectorsize;
1223 if (sectorsize <= 0) {
1224 gctl_error(req, "Can't get sector size.");
1228 /* Reserve space for metadata. */
1229 size -= 0x800 * sectorsize;
1231 /* Handle size argument. */
1232 len = sizeof(*sizearg);
1233 sizearg = gctl_get_param(req, "size", &len);
1234 if (sizearg != NULL && len == sizeof(*sizearg) &&
1236 if (*sizearg > size) {
1237 gctl_error(req, "Size too big %lld > %lld.",
1238 (long long)*sizearg, (long long)size);
1244 /* Handle strip argument. */
1246 len = sizeof(*striparg);
1247 striparg = gctl_get_param(req, "strip", &len);
1248 if (striparg != NULL && len == sizeof(*striparg) &&
1250 if (*striparg < sectorsize) {
1251 gctl_error(req, "Strip size too small.");
1254 if (*striparg % sectorsize != 0) {
1255 gctl_error(req, "Incorrect strip size.");
1258 if (strip > 65535 * sectorsize) {
1259 gctl_error(req, "Strip size too big.");
1265 /* Round size down to strip or sector. */
1266 if (level == G_RAID_VOLUME_RL_RAID1)
1267 size -= (size % sectorsize);
1268 else if (level == G_RAID_VOLUME_RL_RAID1E &&
1269 (numdisks & 1) != 0)
1270 size -= (size % (2 * strip));
1272 size -= (size % strip);
1274 gctl_error(req, "Size too small.");
1277 if (size > 0xffffffffffffllu * sectorsize) {
1278 gctl_error(req, "Size too big.");
1282 /* We have all we need, create things: volume, ... */
1283 mdi->mdio_total_disks = numdisks;
1284 mdi->mdio_started = 1;
1285 vol = g_raid_create_volume(sc, volname, -1);
1286 vol->v_md_data = (void *)(intptr_t)0;
1287 vol->v_raid_level = level;
1288 vol->v_raid_level_qualifier = qual;
1289 vol->v_strip_size = strip;
1290 vol->v_disks_count = numdisks;
1291 if (level == G_RAID_VOLUME_RL_RAID0 ||
1292 level == G_RAID_VOLUME_RL_CONCAT ||
1293 level == G_RAID_VOLUME_RL_SINGLE)
1294 vol->v_mediasize = size * numdisks;
1295 else if (level == G_RAID_VOLUME_RL_RAID1)
1296 vol->v_mediasize = size;
1297 else if (level == G_RAID_VOLUME_RL_RAID5)
1298 vol->v_mediasize = size * (numdisks - 1);
1300 vol->v_mediasize = ((size * numdisks) / strip / 2) *
1303 vol->v_sectorsize = sectorsize;
1304 g_raid_start_volume(vol);
1306 /* , and subdisks. */
1307 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1308 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data;
1309 sd = &vol->v_subdisks[pd->pd_disk_pos];
1313 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1314 if (sd->sd_disk->d_consumer != NULL) {
1315 g_raid_change_disk_state(disk,
1316 G_RAID_DISK_S_ACTIVE);
1317 g_raid_change_subdisk_state(sd,
1318 G_RAID_SUBDISK_S_ACTIVE);
1319 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
1320 G_RAID_EVENT_SUBDISK);
1322 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1326 /* Write metadata based on created entities. */
1327 G_RAID_DEBUG1(0, sc, "Array started.");
1328 g_raid_md_write_sii(md, NULL, NULL, NULL);
1330 /* Pickup any STALE/SPARE disks to refill array if needed. */
1331 g_raid_md_sii_refill(sc);
1333 g_raid_event_send(vol, G_RAID_VOLUME_E_START,
1334 G_RAID_EVENT_VOLUME);
1337 if (strcmp(verb, "delete") == 0) {
1339 /* Check if some volume is still open. */
1340 force = gctl_get_paraml(req, "force", sizeof(*force));
1341 if (force != NULL && *force == 0 &&
1342 g_raid_nopens(sc) != 0) {
1343 gctl_error(req, "Some volume is still open.");
1347 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1348 if (disk->d_consumer)
1349 sii_meta_erase(disk->d_consumer);
1351 g_raid_destroy_node(sc, 0);
1354 if (strcmp(verb, "remove") == 0 ||
1355 strcmp(verb, "fail") == 0) {
1357 gctl_error(req, "Invalid number of arguments.");
1360 for (i = 1; i < *nargs; i++) {
1361 snprintf(arg, sizeof(arg), "arg%d", i);
1362 diskname = gctl_get_asciiparam(req, arg);
1363 if (diskname == NULL) {
1364 gctl_error(req, "No disk name (%s).", arg);
1368 if (strncmp(diskname, "/dev/", 5) == 0)
1371 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1372 if (disk->d_consumer != NULL &&
1373 disk->d_consumer->provider != NULL &&
1374 strcmp(disk->d_consumer->provider->name,
1379 gctl_error(req, "Disk '%s' not found.",
1385 if (strcmp(verb, "fail") == 0) {
1386 g_raid_md_fail_disk_sii(md, NULL, disk);
1390 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data;
1392 /* Erase metadata on deleting disk. */
1393 sii_meta_erase(disk->d_consumer);
1395 /* If disk was assigned, just update statuses. */
1396 if (pd->pd_disk_pos >= 0) {
1397 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1398 g_raid_kill_consumer(sc, disk->d_consumer);
1399 disk->d_consumer = NULL;
1400 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
1401 g_raid_change_subdisk_state(sd,
1402 G_RAID_SUBDISK_S_NONE);
1403 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
1404 G_RAID_EVENT_SUBDISK);
1407 /* Otherwise -- delete. */
1408 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
1409 g_raid_destroy_disk(disk);
1413 /* Write updated metadata to remaining disks. */
1414 g_raid_md_write_sii(md, NULL, NULL, NULL);
1416 /* Check if anything left except placeholders. */
1417 if (g_raid_ndisks(sc, -1) ==
1418 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1419 g_raid_destroy_node(sc, 0);
1421 g_raid_md_sii_refill(sc);
1424 if (strcmp(verb, "insert") == 0) {
1426 gctl_error(req, "Invalid number of arguments.");
1430 for (i = 1; i < *nargs; i++) {
1431 /* Get disk name. */
1432 snprintf(arg, sizeof(arg), "arg%d", i);
1433 diskname = gctl_get_asciiparam(req, arg);
1434 if (diskname == NULL) {
1435 gctl_error(req, "No disk name (%s).", arg);
1440 /* Try to find provider with specified name. */
1442 cp = g_raid_open_consumer(sc, diskname);
1444 gctl_error(req, "Can't open disk '%s'.",
1446 g_topology_unlock();
1452 pd = malloc(sizeof(*pd), M_MD_SII, M_WAITOK | M_ZERO);
1453 pd->pd_disk_pos = -3;
1454 pd->pd_disk_size = pp->mediasize;
1456 disk = g_raid_create_disk(sc);
1457 disk->d_consumer = cp;
1458 disk->d_md_data = (void *)pd;
1460 g_topology_unlock();
1462 /* Read kernel dumping information. */
1463 disk->d_kd.offset = 0;
1464 disk->d_kd.length = OFF_MAX;
1465 len = sizeof(disk->d_kd);
1466 g_io_getattr("GEOM::kerneldump", cp, &len, &disk->d_kd);
1467 if (disk->d_kd.di.dumper == NULL)
1468 G_RAID_DEBUG1(2, sc,
1469 "Dumping not supported by %s.",
1470 cp->provider->name);
1472 /* Welcome the "new" disk. */
1473 update += g_raid_md_sii_start_disk(disk);
1474 if (disk->d_state == G_RAID_DISK_S_SPARE) {
1475 sii_meta_write_spare(cp);
1476 g_raid_destroy_disk(disk);
1477 } else if (disk->d_state != G_RAID_DISK_S_ACTIVE) {
1478 gctl_error(req, "Disk '%s' doesn't fit.",
1480 g_raid_destroy_disk(disk);
1486 /* Write new metadata if we changed something. */
1488 g_raid_md_write_sii(md, NULL, NULL, NULL);
1491 gctl_error(req, "Command '%s' is not supported.", verb);
1496 g_raid_md_write_sii(struct g_raid_md_object *md, struct g_raid_volume *tvol,
1497 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
1499 struct g_raid_softc *sc;
1500 struct g_raid_volume *vol;
1501 struct g_raid_subdisk *sd;
1502 struct g_raid_disk *disk;
1503 struct g_raid_md_sii_object *mdi;
1504 struct g_raid_md_sii_perdisk *pd;
1505 struct sii_raid_conf *meta;
1509 mdi = (struct g_raid_md_sii_object *)md;
1511 if (sc->sc_stopping == G_RAID_DESTROY_HARD)
1514 /* Bump generation. Newly written metadata may differ from previous. */
1515 mdi->mdio_generation++;
1517 /* There is only one volume. */
1518 vol = TAILQ_FIRST(&sc->sc_volumes);
1520 /* Fill global fields. */
1521 meta = malloc(sizeof(*meta), M_MD_SII, M_WAITOK | M_ZERO);
1523 memcpy(meta, mdi->mdio_meta, sizeof(*meta));
1524 meta->total_sectors = vol->v_mediasize / vol->v_sectorsize;
1525 meta->vendor_id = 0x1095;
1526 meta->version_minor = 0;
1527 meta->version_major = 2;
1528 memcpy(&meta->timestamp, &mdi->mdio_timestamp, 6);
1529 meta->strip_sectors = vol->v_strip_size / vol->v_sectorsize;
1530 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0) {
1531 meta->type = SII_T_RAID0;
1532 meta->raid0_disks = vol->v_disks_count;
1533 meta->raid1_disks = 0xff;
1534 } else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1) {
1535 meta->type = SII_T_RAID1;
1536 meta->raid0_disks = 0xff;
1537 meta->raid1_disks = vol->v_disks_count;
1538 } else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E) {
1539 meta->type = SII_T_RAID01;
1540 meta->raid0_disks = vol->v_disks_count / 2;
1541 meta->raid1_disks = 2;
1542 } else if (vol->v_raid_level == G_RAID_VOLUME_RL_CONCAT ||
1543 vol->v_raid_level == G_RAID_VOLUME_RL_SINGLE) {
1544 meta->type = SII_T_JBOD;
1545 meta->raid0_disks = vol->v_disks_count;
1546 meta->raid1_disks = 0xff;
1548 meta->type = SII_T_RAID5;
1549 meta->raid0_disks = vol->v_disks_count;
1550 meta->raid1_disks = 0xff;
1552 meta->generation = mdi->mdio_generation;
1553 meta->raid_status = vol->v_dirty ? SII_S_ONLINE : SII_S_AVAILABLE;
1554 for (i = 0; i < vol->v_disks_count; i++) {
1555 sd = &vol->v_subdisks[i];
1556 if (sd->sd_state == G_RAID_SUBDISK_S_STALE ||
1557 sd->sd_state == G_RAID_SUBDISK_S_RESYNC)
1558 meta->raid_status = SII_S_ONLINE;
1560 meta->raid_location = mdi->mdio_location;
1561 sii_meta_put_name(meta, vol->v_name);
1563 /* We are done. Print meta data and store them to disks. */
1564 if (mdi->mdio_meta != NULL)
1565 free(mdi->mdio_meta, M_MD_SII);
1566 mdi->mdio_meta = meta;
1567 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1568 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data;
1569 if (disk->d_state != G_RAID_DISK_S_ACTIVE)
1571 if (pd->pd_meta != NULL) {
1572 free(pd->pd_meta, M_MD_SII);
1575 pd->pd_meta = sii_meta_copy(meta);
1576 if ((sd = TAILQ_FIRST(&disk->d_subdisks)) != NULL) {
1577 if (sd->sd_state < G_RAID_SUBDISK_S_NEW)
1578 pd->pd_meta->disk_status = SII_S_DROPPED;
1579 else if (sd->sd_state < G_RAID_SUBDISK_S_STALE) {
1580 pd->pd_meta->disk_status = SII_S_REBUILD;
1581 pd->pd_meta->rebuild_lba =
1582 sd->sd_rebuild_pos / vol->v_sectorsize;
1584 pd->pd_meta->disk_status = SII_S_CURRENT;
1585 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1) {
1586 pd->pd_meta->disk_number = sd->sd_pos;
1587 pd->pd_meta->raid0_ident = 0xff;
1588 pd->pd_meta->raid1_ident = 0;
1589 } else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E) {
1590 pd->pd_meta->disk_number = sd->sd_pos / meta->raid1_disks;
1591 pd->pd_meta->raid0_ident = sd->sd_pos % meta->raid1_disks;
1592 pd->pd_meta->raid1_ident = sd->sd_pos / meta->raid1_disks;
1594 pd->pd_meta->disk_number = sd->sd_pos;
1595 pd->pd_meta->raid0_ident = 0;
1596 pd->pd_meta->raid1_ident = 0xff;
1599 G_RAID_DEBUG(1, "Writing SiI metadata to %s",
1600 g_raid_get_diskname(disk));
1601 g_raid_md_sii_print(pd->pd_meta);
1602 sii_meta_write(disk->d_consumer, pd->pd_meta);
1608 g_raid_md_fail_disk_sii(struct g_raid_md_object *md,
1609 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
1611 struct g_raid_softc *sc;
1612 struct g_raid_md_sii_perdisk *pd;
1613 struct g_raid_subdisk *sd;
1616 pd = (struct g_raid_md_sii_perdisk *)tdisk->d_md_data;
1618 /* We can't fail disk that is not a part of array now. */
1619 if (pd->pd_disk_pos < 0)
1623 * Mark disk as failed in metadata and try to write that metadata
1624 * to the disk itself to prevent it's later resurrection as STALE.
1626 if (tdisk->d_consumer) {
1628 pd->pd_meta->disk_status = SII_S_REMOVED;
1629 sii_meta_write(tdisk->d_consumer, pd->pd_meta);
1631 sii_meta_erase(tdisk->d_consumer);
1634 /* Change states. */
1635 g_raid_change_disk_state(tdisk, G_RAID_DISK_S_FAILED);
1636 TAILQ_FOREACH(sd, &tdisk->d_subdisks, sd_next) {
1637 g_raid_change_subdisk_state(sd,
1638 G_RAID_SUBDISK_S_FAILED);
1639 g_raid_event_send(sd, G_RAID_SUBDISK_E_FAILED,
1640 G_RAID_EVENT_SUBDISK);
1643 /* Write updated metadata to remaining disks. */
1644 g_raid_md_write_sii(md, NULL, NULL, tdisk);
1646 /* Check if anything left except placeholders. */
1647 if (g_raid_ndisks(sc, -1) ==
1648 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1649 g_raid_destroy_node(sc, 0);
1651 g_raid_md_sii_refill(sc);
1656 g_raid_md_free_disk_sii(struct g_raid_md_object *md,
1657 struct g_raid_disk *disk)
1659 struct g_raid_md_sii_perdisk *pd;
1661 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data;
1662 if (pd->pd_meta != NULL) {
1663 free(pd->pd_meta, M_MD_SII);
1667 disk->d_md_data = NULL;
1672 g_raid_md_free_sii(struct g_raid_md_object *md)
1674 struct g_raid_md_sii_object *mdi;
1676 mdi = (struct g_raid_md_sii_object *)md;
1677 if (!mdi->mdio_started) {
1678 mdi->mdio_started = 0;
1679 callout_stop(&mdi->mdio_start_co);
1680 G_RAID_DEBUG1(1, md->mdo_softc,
1681 "root_mount_rel %p", mdi->mdio_rootmount);
1682 root_mount_rel(mdi->mdio_rootmount);
1683 mdi->mdio_rootmount = NULL;
1685 if (mdi->mdio_meta != NULL) {
1686 free(mdi->mdio_meta, M_MD_SII);
1687 mdi->mdio_meta = NULL;
1692 G_RAID_MD_DECLARE(sii, "SiI");