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),
147 g_raid_md_sii_print(struct sii_raid_conf *meta)
150 if (g_raid_debug < 1)
153 printf("********* ATA SiI RAID Metadata *********\n");
154 printf("total_sectors %llu\n",
155 (long long unsigned)meta->total_sectors);
156 printf("product_id 0x%04x\n", meta->product_id);
157 printf("vendor_id 0x%04x\n", meta->vendor_id);
158 printf("version_minor 0x%04x\n", meta->version_minor);
159 printf("version_major 0x%04x\n", meta->version_major);
160 printf("timestamp 0x%02x%02x%02x%02x%02x%02x\n",
161 meta->timestamp[5], meta->timestamp[4], meta->timestamp[3],
162 meta->timestamp[2], meta->timestamp[1], meta->timestamp[0]);
163 printf("strip_sectors %d\n", meta->strip_sectors);
164 printf("disk_number %d\n", meta->disk_number);
165 printf("type 0x%02x\n", meta->type);
166 printf("raid0_disks %d\n", meta->raid0_disks);
167 printf("raid0_ident %d\n", meta->raid0_ident);
168 printf("raid1_disks %d\n", meta->raid1_disks);
169 printf("raid1_ident %d\n", meta->raid1_ident);
170 printf("rebuild_lba %llu\n",
171 (long long unsigned)meta->rebuild_lba);
172 printf("generation %d\n", meta->generation);
173 printf("disk_status %d\n", meta->disk_status);
174 printf("raid_status %d\n", meta->raid_status);
175 printf("raid_location %d\n", meta->raid_location);
176 printf("disk_location %d\n", meta->disk_location);
177 printf("auto_rebuild %d\n", meta->auto_rebuild);
178 printf("name <%.16s>\n", meta->name);
179 printf("checksum 0x%04x\n", meta->checksum);
180 printf("=================================================\n");
183 static struct sii_raid_conf *
184 sii_meta_copy(struct sii_raid_conf *meta)
186 struct sii_raid_conf *nmeta;
188 nmeta = malloc(sizeof(*meta), M_MD_SII, M_WAITOK);
189 memcpy(nmeta, meta, sizeof(*meta));
194 sii_meta_total_disks(struct sii_raid_conf *meta)
197 switch (meta->type) {
201 return (meta->raid0_disks);
203 return (meta->raid1_disks);
205 return (meta->raid0_disks * meta->raid1_disks);
214 sii_meta_disk_pos(struct sii_raid_conf *meta, struct sii_raid_conf *pdmeta)
217 if (pdmeta->type == SII_T_SPARE)
220 if (memcmp(&meta->timestamp, &pdmeta->timestamp, 6) != 0)
223 switch (pdmeta->type) {
228 return (pdmeta->disk_number);
230 return (pdmeta->raid1_ident * pdmeta->raid1_disks +
231 pdmeta->raid0_ident);
239 sii_meta_get_name(struct sii_raid_conf *meta, char *buf)
243 strncpy(buf, meta->name, 16);
245 for (i = 15; i >= 0; i--) {
253 sii_meta_put_name(struct sii_raid_conf *meta, char *buf)
256 memset(meta->name, 0x20, 16);
257 memcpy(meta->name, buf, MIN(strlen(buf), 16));
260 static struct sii_raid_conf *
261 sii_meta_read(struct g_consumer *cp)
263 struct g_provider *pp;
264 struct sii_raid_conf *meta;
267 uint16_t checksum, *ptr;
271 /* Read the anchor sector. */
272 buf = g_read_data(cp,
273 pp->mediasize - pp->sectorsize, pp->sectorsize, &error);
275 G_RAID_DEBUG(1, "Cannot read metadata from %s (error=%d).",
279 meta = malloc(sizeof(*meta), M_MD_SII, M_WAITOK);
280 memcpy(meta, buf, min(sizeof(*meta), pp->sectorsize));
283 /* Check vendor ID. */
284 if (meta->vendor_id != 0x1095) {
285 G_RAID_DEBUG(1, "SiI vendor ID check failed on %s (0x%04x)",
286 pp->name, meta->vendor_id);
287 free(meta, M_MD_SII);
291 /* Check metadata major version. */
292 if (meta->version_major != 2) {
293 G_RAID_DEBUG(1, "SiI version check failed on %s (%d.%d)",
294 pp->name, meta->version_major, meta->version_minor);
295 free(meta, M_MD_SII);
299 /* Check metadata checksum. */
300 for (checksum = 0, ptr = (uint16_t *)meta, i = 0; i <= 159; i++)
303 G_RAID_DEBUG(1, "SiI checksum check failed on %s", pp->name);
304 free(meta, M_MD_SII);
308 /* Check raid type. */
309 if (meta->type != SII_T_RAID0 && meta->type != SII_T_RAID1 &&
310 meta->type != SII_T_RAID01 && meta->type != SII_T_SPARE &&
311 meta->type != SII_T_RAID5 && meta->type != SII_T_CONCAT &&
312 meta->type != SII_T_JBOD) {
313 G_RAID_DEBUG(1, "SiI unknown RAID level on %s (0x%02x)",
314 pp->name, meta->type);
315 free(meta, M_MD_SII);
323 sii_meta_write(struct g_consumer *cp, struct sii_raid_conf *meta)
325 struct g_provider *pp;
328 uint16_t checksum, *ptr;
332 /* Recalculate checksum for case if metadata were changed. */
334 for (checksum = 0, ptr = (uint16_t *)meta, i = 0; i < 159; i++)
336 meta->checksum -= checksum;
338 /* Create and fill buffer. */
339 buf = malloc(pp->sectorsize, M_MD_SII, M_WAITOK | M_ZERO);
340 memcpy(buf, meta, sizeof(*meta));
342 /* Write 4 copies of metadata. */
343 for (i = 0; i < 4; i++) {
344 error = g_write_data(cp,
345 pp->mediasize - (pp->sectorsize * (1 + 0x200 * i)),
346 buf, pp->sectorsize);
348 G_RAID_DEBUG(1, "Cannot write metadata to %s (error=%d).",
359 sii_meta_erase(struct g_consumer *cp)
361 struct g_provider *pp;
366 buf = malloc(pp->sectorsize, M_MD_SII, M_WAITOK | M_ZERO);
367 /* Write 4 copies of metadata. */
368 for (i = 0; i < 4; i++) {
369 error = g_write_data(cp,
370 pp->mediasize - (pp->sectorsize * (1 + 0x200 * i)),
371 buf, pp->sectorsize);
373 G_RAID_DEBUG(1, "Cannot erase metadata on %s (error=%d).",
382 sii_meta_write_spare(struct g_consumer *cp)
384 struct sii_raid_conf *meta;
387 meta = malloc(sizeof(*meta), M_MD_SII, M_WAITOK | M_ZERO);
388 meta->total_sectors = cp->provider->mediasize /
389 cp->provider->sectorsize - 0x800;
390 meta->vendor_id = 0x1095;
391 meta->version_minor = 0;
392 meta->version_major = 2;
393 meta->timestamp[0] = arc4random();
394 meta->timestamp[1] = arc4random();
395 meta->timestamp[2] = arc4random();
396 meta->timestamp[3] = arc4random();
397 meta->timestamp[4] = arc4random();
398 meta->timestamp[5] = arc4random();
399 meta->type = SII_T_SPARE;
400 meta->generation = 1;
401 meta->raid1_ident = 0xff;
402 meta->raid_location = arc4random();
403 error = sii_meta_write(cp, meta);
404 free(meta, M_MD_SII);
408 static struct g_raid_disk *
409 g_raid_md_sii_get_disk(struct g_raid_softc *sc, int id)
411 struct g_raid_disk *disk;
412 struct g_raid_md_sii_perdisk *pd;
414 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
415 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data;
416 if (pd->pd_disk_pos == id)
423 g_raid_md_sii_supported(int level, int qual, int disks, int force)
429 case G_RAID_VOLUME_RL_RAID0:
432 if (!force && (disks < 2 || disks > 6))
435 case G_RAID_VOLUME_RL_RAID1:
438 if (!force && (disks != 2))
441 case G_RAID_VOLUME_RL_RAID1E:
446 if (!force && (disks < 4))
449 case G_RAID_VOLUME_RL_SINGLE:
453 case G_RAID_VOLUME_RL_CONCAT:
457 case G_RAID_VOLUME_RL_RAID5:
460 if (qual != G_RAID_VOLUME_RLQ_R5LS)
466 if (level != G_RAID_VOLUME_RL_RAID5 && qual != G_RAID_VOLUME_RLQ_NONE)
472 g_raid_md_sii_start_disk(struct g_raid_disk *disk)
474 struct g_raid_softc *sc;
475 struct g_raid_subdisk *sd, *tmpsd;
476 struct g_raid_disk *olddisk, *tmpdisk;
477 struct g_raid_md_object *md;
478 struct g_raid_md_sii_object *mdi;
479 struct g_raid_md_sii_perdisk *pd, *oldpd;
480 struct sii_raid_conf *meta;
481 int disk_pos, resurrection = 0;
485 mdi = (struct g_raid_md_sii_object *)md;
486 meta = mdi->mdio_meta;
487 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data;
490 /* Find disk position in metadata by it's serial. */
491 if (pd->pd_meta != NULL)
492 disk_pos = sii_meta_disk_pos(meta, pd->pd_meta);
496 G_RAID_DEBUG1(1, sc, "Unknown, probably new or stale disk");
497 /* If we are in the start process, that's all for now. */
498 if (!mdi->mdio_started)
501 * If we have already started - try to get use of the disk.
502 * Try to replace OFFLINE disks first, then FAILED.
504 TAILQ_FOREACH(tmpdisk, &sc->sc_disks, d_next) {
505 if (tmpdisk->d_state != G_RAID_DISK_S_OFFLINE &&
506 tmpdisk->d_state != G_RAID_DISK_S_FAILED)
508 /* Make sure this disk is big enough. */
509 TAILQ_FOREACH(sd, &tmpdisk->d_subdisks, sd_next) {
510 if (sd->sd_offset + sd->sd_size + 512 >
513 "Disk too small (%ju < %ju)",
515 sd->sd_offset + sd->sd_size + 512);
521 if (tmpdisk->d_state == G_RAID_DISK_S_OFFLINE) {
524 } else if (olddisk == NULL)
527 if (olddisk == NULL) {
529 if (disk_pos == -3 || pd->pd_disk_pos == -3) {
530 g_raid_change_disk_state(disk,
531 G_RAID_DISK_S_SPARE);
534 g_raid_change_disk_state(disk,
535 G_RAID_DISK_S_STALE);
539 oldpd = (struct g_raid_md_sii_perdisk *)olddisk->d_md_data;
540 disk_pos = oldpd->pd_disk_pos;
544 if (olddisk == NULL) {
545 /* Find placeholder by position. */
546 olddisk = g_raid_md_sii_get_disk(sc, disk_pos);
548 panic("No disk at position %d!", disk_pos);
549 if (olddisk->d_state != G_RAID_DISK_S_OFFLINE) {
550 G_RAID_DEBUG1(1, sc, "More then one disk for pos %d",
552 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE);
555 oldpd = (struct g_raid_md_sii_perdisk *)olddisk->d_md_data;
558 /* Replace failed disk or placeholder with new disk. */
559 TAILQ_FOREACH_SAFE(sd, &olddisk->d_subdisks, sd_next, tmpsd) {
560 TAILQ_REMOVE(&olddisk->d_subdisks, sd, sd_next);
561 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
564 oldpd->pd_disk_pos = -2;
565 pd->pd_disk_pos = disk_pos;
567 /* If it was placeholder -- destroy it. */
568 if (olddisk->d_state == G_RAID_DISK_S_OFFLINE) {
569 g_raid_destroy_disk(olddisk);
571 /* Otherwise, make it STALE_FAILED. */
572 g_raid_change_disk_state(olddisk, G_RAID_DISK_S_STALE_FAILED);
575 /* Welcome the new disk. */
577 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
578 else if (pd->pd_meta->disk_status == SII_S_CURRENT ||
579 pd->pd_meta->disk_status == SII_S_REBUILD)
580 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
582 g_raid_change_disk_state(disk, G_RAID_DISK_S_FAILED);
583 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
586 * Different disks may have different sizes,
587 * in concat mode. Update from real disk size.
589 if (meta->type == SII_T_CONCAT || meta->type == SII_T_JBOD)
590 sd->sd_size = pd->pd_disk_size - 0x800 * 512;
593 /* New or ex-spare disk. */
594 g_raid_change_subdisk_state(sd,
595 G_RAID_SUBDISK_S_NEW);
596 } else if (pd->pd_meta->disk_status == SII_S_REBUILD) {
597 /* Rebuilding disk. */
598 g_raid_change_subdisk_state(sd,
599 G_RAID_SUBDISK_S_REBUILD);
600 if (pd->pd_meta->generation == meta->generation)
601 sd->sd_rebuild_pos = pd->pd_meta->rebuild_lba * 512;
603 sd->sd_rebuild_pos = 0;
604 } else if (pd->pd_meta->disk_status == SII_S_CURRENT) {
605 if (pd->pd_meta->raid_status == SII_S_ONLINE ||
606 pd->pd_meta->generation != meta->generation) {
607 /* Dirty or resyncing disk. */
608 g_raid_change_subdisk_state(sd,
609 G_RAID_SUBDISK_S_STALE);
611 /* Up to date disk. */
612 g_raid_change_subdisk_state(sd,
613 G_RAID_SUBDISK_S_ACTIVE);
616 g_raid_change_subdisk_state(sd,
617 G_RAID_SUBDISK_S_FAILED);
619 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
620 G_RAID_EVENT_SUBDISK);
623 /* Update status of our need for spare. */
624 if (mdi->mdio_started) {
625 mdi->mdio_incomplete =
626 (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) <
627 mdi->mdio_total_disks);
630 return (resurrection);
634 g_disk_md_sii_retaste(void *arg, int pending)
637 G_RAID_DEBUG(1, "Array is not complete, trying to retaste.");
638 g_retaste(&g_raid_class);
643 g_raid_md_sii_refill(struct g_raid_softc *sc)
645 struct g_raid_md_object *md;
646 struct g_raid_md_sii_object *mdi;
647 struct g_raid_disk *disk;
652 mdi = (struct g_raid_md_sii_object *)md;
655 /* Make sure we miss anything. */
656 na = g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE);
657 if (na == mdi->mdio_total_disks)
660 G_RAID_DEBUG1(1, md->mdo_softc,
661 "Array is not complete (%d of %d), "
662 "trying to refill.", na, mdi->mdio_total_disks);
664 /* Try to get use some of STALE disks. */
665 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
666 if (disk->d_state == G_RAID_DISK_S_STALE) {
667 update += g_raid_md_sii_start_disk(disk);
668 if (disk->d_state == G_RAID_DISK_S_ACTIVE)
675 /* Try to get use some of SPARE disks. */
676 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
677 if (disk->d_state == G_RAID_DISK_S_SPARE) {
678 update += g_raid_md_sii_start_disk(disk);
679 if (disk->d_state == G_RAID_DISK_S_ACTIVE)
683 } while (disk != NULL);
685 /* Write new metadata if we changed something. */
687 g_raid_md_write_sii(md, NULL, NULL, NULL);
689 /* Update status of our need for spare. */
690 mdi->mdio_incomplete = (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) <
691 mdi->mdio_total_disks);
693 /* Request retaste hoping to find spare. */
694 if (mdi->mdio_incomplete) {
695 task = malloc(sizeof(struct task),
696 M_MD_SII, M_WAITOK | M_ZERO);
697 TASK_INIT(task, 0, g_disk_md_sii_retaste, task);
698 taskqueue_enqueue(taskqueue_swi, task);
703 g_raid_md_sii_start(struct g_raid_softc *sc)
705 struct g_raid_md_object *md;
706 struct g_raid_md_sii_object *mdi;
707 struct g_raid_md_sii_perdisk *pd;
708 struct sii_raid_conf *meta;
709 struct g_raid_volume *vol;
710 struct g_raid_subdisk *sd;
711 struct g_raid_disk *disk, *best;
714 uint32_t gendiff, bestgendiff;
718 mdi = (struct g_raid_md_sii_object *)md;
719 meta = mdi->mdio_meta;
721 /* Create volumes and subdisks. */
722 sii_meta_get_name(meta, buf);
723 vol = g_raid_create_volume(sc, buf, -1);
724 vol->v_mediasize = (off_t)meta->total_sectors * 512;
725 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE;
726 if (meta->type == SII_T_RAID0) {
727 vol->v_raid_level = G_RAID_VOLUME_RL_RAID0;
728 size = vol->v_mediasize / mdi->mdio_total_disks;
729 } else if (meta->type == SII_T_RAID1) {
730 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1;
731 size = vol->v_mediasize;
732 } else if (meta->type == SII_T_RAID01) {
733 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E;
734 size = vol->v_mediasize / (mdi->mdio_total_disks / 2);
735 } else if (meta->type == SII_T_CONCAT) {
736 if (mdi->mdio_total_disks == 1)
737 vol->v_raid_level = G_RAID_VOLUME_RL_SINGLE;
739 vol->v_raid_level = G_RAID_VOLUME_RL_CONCAT;
741 } else if (meta->type == SII_T_RAID5) {
742 vol->v_raid_level = G_RAID_VOLUME_RL_RAID5;
743 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_R5LS;
744 size = vol->v_mediasize / (mdi->mdio_total_disks - 1);
745 } else if (meta->type == SII_T_JBOD) {
746 vol->v_raid_level = G_RAID_VOLUME_RL_SINGLE;
749 vol->v_raid_level = G_RAID_VOLUME_RL_UNKNOWN;
752 vol->v_strip_size = meta->strip_sectors * 512; //ZZZ
753 vol->v_disks_count = mdi->mdio_total_disks;
754 vol->v_sectorsize = 512; //ZZZ
755 for (j = 0; j < vol->v_disks_count; j++) {
756 sd = &vol->v_subdisks[j];
760 g_raid_start_volume(vol);
762 /* Create disk placeholders to store data for later writing. */
763 for (disk_pos = 0; disk_pos < mdi->mdio_total_disks; disk_pos++) {
764 pd = malloc(sizeof(*pd), M_MD_SII, M_WAITOK | M_ZERO);
765 pd->pd_disk_pos = disk_pos;
766 disk = g_raid_create_disk(sc);
767 disk->d_md_data = (void *)pd;
768 disk->d_state = G_RAID_DISK_S_OFFLINE;
769 sd = &vol->v_subdisks[disk_pos];
771 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
775 * Make all disks found till the moment take their places
776 * in order of their generation numbers.
780 bestgendiff = 0xffffffff;
781 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
782 if (disk->d_state != G_RAID_DISK_S_NONE)
784 pd = disk->d_md_data;
785 if (pd->pd_meta == NULL)
786 gendiff = 0xfffffffe;
788 gendiff = meta->generation -
789 pd->pd_meta->generation;
790 if (gendiff < bestgendiff) {
792 bestgendiff = gendiff;
796 g_raid_md_sii_start_disk(best);
797 } while (best != NULL);
799 mdi->mdio_started = 1;
800 G_RAID_DEBUG1(0, sc, "Array started.");
801 g_raid_md_write_sii(md, NULL, NULL, NULL);
803 /* Pickup any STALE/SPARE disks to refill array if needed. */
804 g_raid_md_sii_refill(sc);
806 g_raid_event_send(vol, G_RAID_VOLUME_E_START, G_RAID_EVENT_VOLUME);
808 callout_stop(&mdi->mdio_start_co);
809 G_RAID_DEBUG1(1, sc, "root_mount_rel %p", mdi->mdio_rootmount);
810 root_mount_rel(mdi->mdio_rootmount);
811 mdi->mdio_rootmount = NULL;
815 g_raid_md_sii_new_disk(struct g_raid_disk *disk)
817 struct g_raid_softc *sc;
818 struct g_raid_md_object *md;
819 struct g_raid_md_sii_object *mdi;
820 struct sii_raid_conf *pdmeta;
821 struct g_raid_md_sii_perdisk *pd;
825 mdi = (struct g_raid_md_sii_object *)md;
826 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data;
827 pdmeta = pd->pd_meta;
829 if (mdi->mdio_started) {
830 if (g_raid_md_sii_start_disk(disk))
831 g_raid_md_write_sii(md, NULL, NULL, NULL);
833 if (mdi->mdio_meta == NULL ||
834 ((int32_t)(pdmeta->generation - mdi->mdio_generation)) > 0) {
835 G_RAID_DEBUG1(1, sc, "Newer disk");
836 if (mdi->mdio_meta != NULL)
837 free(mdi->mdio_meta, M_MD_SII);
838 mdi->mdio_meta = sii_meta_copy(pdmeta);
839 mdi->mdio_generation = mdi->mdio_meta->generation;
840 mdi->mdio_total_disks = sii_meta_total_disks(pdmeta);
841 mdi->mdio_disks_present = 1;
842 } else if (pdmeta->generation == mdi->mdio_generation) {
843 mdi->mdio_disks_present++;
844 G_RAID_DEBUG1(1, sc, "Matching disk (%d of %d up)",
845 mdi->mdio_disks_present,
846 mdi->mdio_total_disks);
848 G_RAID_DEBUG1(1, sc, "Older disk");
851 /* If we collected all needed disks - start array. */
852 if (mdi->mdio_disks_present == mdi->mdio_total_disks)
853 g_raid_md_sii_start(sc);
858 g_raid_sii_go(void *arg)
860 struct g_raid_softc *sc;
861 struct g_raid_md_object *md;
862 struct g_raid_md_sii_object *mdi;
866 mdi = (struct g_raid_md_sii_object *)md;
867 if (!mdi->mdio_started) {
868 G_RAID_DEBUG1(0, sc, "Force array start due to timeout.");
869 g_raid_event_send(sc, G_RAID_NODE_E_START, 0);
874 g_raid_md_create_sii(struct g_raid_md_object *md, struct g_class *mp,
877 struct g_raid_softc *sc;
878 struct g_raid_md_sii_object *mdi;
881 mdi = (struct g_raid_md_sii_object *)md;
882 mdi->mdio_timestamp[5] = arc4random();
883 mdi->mdio_timestamp[4] = arc4random();
884 mdi->mdio_timestamp[3] = arc4random();
885 mdi->mdio_timestamp[2] = arc4random();
886 mdi->mdio_timestamp[1] = arc4random();
887 mdi->mdio_timestamp[0] = arc4random();
888 mdi->mdio_location = arc4random();
889 mdi->mdio_generation = 0;
890 snprintf(name, sizeof(name), "SiI-%02x%02x%02x%02x%02x%02x",
891 mdi->mdio_timestamp[5], mdi->mdio_timestamp[4],
892 mdi->mdio_timestamp[3], mdi->mdio_timestamp[2],
893 mdi->mdio_timestamp[1], mdi->mdio_timestamp[0]);
894 sc = g_raid_create_node(mp, name, md);
896 return (G_RAID_MD_TASTE_FAIL);
899 return (G_RAID_MD_TASTE_NEW);
903 g_raid_md_taste_sii(struct g_raid_md_object *md, struct g_class *mp,
904 struct g_consumer *cp, struct g_geom **gp)
906 struct g_consumer *rcp;
907 struct g_provider *pp;
908 struct g_raid_md_sii_object *mdi, *mdi1;
909 struct g_raid_softc *sc;
910 struct g_raid_disk *disk;
911 struct sii_raid_conf *meta;
912 struct g_raid_md_sii_perdisk *pd;
914 int error, disk_pos, result, spare, len;
918 G_RAID_DEBUG(1, "Tasting SiI on %s", cp->provider->name);
919 mdi = (struct g_raid_md_sii_object *)md;
922 /* Read metadata from device. */
925 if (g_access(cp, 1, 0, 0) != 0)
926 return (G_RAID_MD_TASTE_FAIL);
929 if (pp->geom->rank == 1)
930 g_io_getattr("GEOM::hba_vendor", cp, &len, &vendor);
931 meta = sii_meta_read(cp);
933 g_access(cp, -1, 0, 0);
935 if (g_raid_aggressive_spare) {
936 if (vendor == 0x1095) {
938 "No SiI metadata, forcing spare.");
943 "SiI vendor mismatch 0x%04x != 0x1095",
947 return (G_RAID_MD_TASTE_FAIL);
950 /* Check this disk position in obtained metadata. */
951 disk_pos = sii_meta_disk_pos(meta, meta);
952 if (disk_pos == -1) {
953 G_RAID_DEBUG(1, "SiI disk position not found");
957 /* Metadata valid. Print it. */
958 g_raid_md_sii_print(meta);
959 G_RAID_DEBUG(1, "SiI disk position %d", disk_pos);
960 spare = (meta->type == SII_T_SPARE) ? 1 : 0;
963 /* Search for matching node. */
966 LIST_FOREACH(geom, &mp->geom, geom) {
970 if (sc->sc_stopping != 0)
972 if (sc->sc_md->mdo_class != md->mdo_class)
974 mdi1 = (struct g_raid_md_sii_object *)sc->sc_md;
976 if (mdi1->mdio_incomplete)
979 if (mdi1->mdio_location == meta->raid_location &&
980 memcmp(&mdi1->mdio_timestamp,
981 &meta->timestamp, 6) == 0)
986 /* Found matching node. */
988 G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name);
989 result = G_RAID_MD_TASTE_EXISTING;
991 } else if (spare) { /* Not found needy node -- left for later. */
992 G_RAID_DEBUG(1, "Spare is not needed at this time");
995 } else { /* Not found matching node -- create one. */
996 result = G_RAID_MD_TASTE_NEW;
997 memcpy(&mdi->mdio_timestamp, &meta->timestamp, 6);
998 mdi->mdio_location = meta->raid_location;
999 snprintf(name, sizeof(name), "SiI-%02x%02x%02x%02x%02x%02x",
1000 mdi->mdio_timestamp[5], mdi->mdio_timestamp[4],
1001 mdi->mdio_timestamp[3], mdi->mdio_timestamp[2],
1002 mdi->mdio_timestamp[1], mdi->mdio_timestamp[0]);
1003 sc = g_raid_create_node(mp, name, md);
1006 callout_init(&mdi->mdio_start_co, 1);
1007 callout_reset(&mdi->mdio_start_co, g_raid_start_timeout * hz,
1009 mdi->mdio_rootmount = root_mount_hold("GRAID-SiI");
1010 G_RAID_DEBUG1(1, sc, "root_mount_hold %p", mdi->mdio_rootmount);
1013 rcp = g_new_consumer(geom);
1015 if (g_access(rcp, 1, 1, 1) != 0)
1018 g_topology_unlock();
1019 sx_xlock(&sc->sc_lock);
1021 pd = malloc(sizeof(*pd), M_MD_SII, M_WAITOK | M_ZERO);
1024 pd->pd_disk_pos = -3;
1026 pd->pd_disk_pos = -1;
1028 pd->pd_disk_size = pp->mediasize;
1029 disk = g_raid_create_disk(sc);
1030 disk->d_md_data = (void *)pd;
1031 disk->d_consumer = rcp;
1032 rcp->private = disk;
1034 /* Read kernel dumping information. */
1035 disk->d_kd.offset = 0;
1036 disk->d_kd.length = OFF_MAX;
1037 len = sizeof(disk->d_kd);
1038 error = g_io_getattr("GEOM::kerneldump", rcp, &len, &disk->d_kd);
1039 if (disk->d_kd.di.dumper == NULL)
1040 G_RAID_DEBUG1(2, sc, "Dumping not supported by %s: %d.",
1041 rcp->provider->name, error);
1043 g_raid_md_sii_new_disk(disk);
1045 sx_xunlock(&sc->sc_lock);
1050 free(meta, M_MD_SII);
1051 return (G_RAID_MD_TASTE_FAIL);
1055 g_raid_md_event_sii(struct g_raid_md_object *md,
1056 struct g_raid_disk *disk, u_int event)
1058 struct g_raid_softc *sc;
1059 struct g_raid_subdisk *sd;
1060 struct g_raid_md_sii_object *mdi;
1061 struct g_raid_md_sii_perdisk *pd;
1064 mdi = (struct g_raid_md_sii_object *)md;
1067 case G_RAID_NODE_E_START:
1068 if (!mdi->mdio_started)
1069 g_raid_md_sii_start(sc);
1074 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data;
1076 case G_RAID_DISK_E_DISCONNECTED:
1077 /* If disk was assigned, just update statuses. */
1078 if (pd->pd_disk_pos >= 0) {
1079 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1080 if (disk->d_consumer) {
1081 g_raid_kill_consumer(sc, disk->d_consumer);
1082 disk->d_consumer = NULL;
1084 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
1085 g_raid_change_subdisk_state(sd,
1086 G_RAID_SUBDISK_S_NONE);
1087 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
1088 G_RAID_EVENT_SUBDISK);
1091 /* Otherwise -- delete. */
1092 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
1093 g_raid_destroy_disk(disk);
1096 /* Write updated metadata to all disks. */
1097 g_raid_md_write_sii(md, NULL, NULL, NULL);
1099 /* Check if anything left except placeholders. */
1100 if (g_raid_ndisks(sc, -1) ==
1101 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1102 g_raid_destroy_node(sc, 0);
1104 g_raid_md_sii_refill(sc);
1111 g_raid_md_ctl_sii(struct g_raid_md_object *md,
1112 struct gctl_req *req)
1114 struct g_raid_softc *sc;
1115 struct g_raid_volume *vol;
1116 struct g_raid_subdisk *sd;
1117 struct g_raid_disk *disk;
1118 struct g_raid_md_sii_object *mdi;
1119 struct g_raid_md_sii_perdisk *pd;
1120 struct g_consumer *cp;
1121 struct g_provider *pp;
1123 const char *verb, *volname, *levelname, *diskname;
1125 off_t size, sectorsize, strip;
1126 intmax_t *sizearg, *striparg;
1127 int numdisks, i, len, level, qual, update;
1131 mdi = (struct g_raid_md_sii_object *)md;
1132 verb = gctl_get_param(req, "verb", NULL);
1133 nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
1135 if (strcmp(verb, "label") == 0) {
1138 gctl_error(req, "Invalid number of arguments.");
1141 volname = gctl_get_asciiparam(req, "arg1");
1142 if (volname == NULL) {
1143 gctl_error(req, "No volume name.");
1146 levelname = gctl_get_asciiparam(req, "arg2");
1147 if (levelname == NULL) {
1148 gctl_error(req, "No RAID level.");
1151 if (strcasecmp(levelname, "RAID5") == 0)
1152 levelname = "RAID5-LS";
1153 if (g_raid_volume_str2level(levelname, &level, &qual)) {
1154 gctl_error(req, "Unknown RAID level '%s'.", levelname);
1157 numdisks = *nargs - 3;
1158 force = gctl_get_paraml(req, "force", sizeof(*force));
1159 if (!g_raid_md_sii_supported(level, qual, numdisks,
1160 force ? *force : 0)) {
1161 gctl_error(req, "Unsupported RAID level "
1162 "(0x%02x/0x%02x), or number of disks (%d).",
1163 level, qual, numdisks);
1167 /* Search for disks, connect them and probe. */
1168 size = 0x7fffffffffffffffllu;
1170 for (i = 0; i < numdisks; i++) {
1171 snprintf(arg, sizeof(arg), "arg%d", i + 3);
1172 diskname = gctl_get_asciiparam(req, arg);
1173 if (diskname == NULL) {
1174 gctl_error(req, "No disk name (%s).", arg);
1178 if (strcmp(diskname, "NONE") == 0) {
1183 cp = g_raid_open_consumer(sc, diskname);
1185 gctl_error(req, "Can't open '%s'.",
1187 g_topology_unlock();
1193 pd = malloc(sizeof(*pd), M_MD_SII, M_WAITOK | M_ZERO);
1194 pd->pd_disk_pos = i;
1195 disk = g_raid_create_disk(sc);
1196 disk->d_md_data = (void *)pd;
1197 disk->d_consumer = cp;
1201 g_topology_unlock();
1203 /* Read kernel dumping information. */
1204 disk->d_kd.offset = 0;
1205 disk->d_kd.length = OFF_MAX;
1206 len = sizeof(disk->d_kd);
1207 g_io_getattr("GEOM::kerneldump", cp, &len, &disk->d_kd);
1208 if (disk->d_kd.di.dumper == NULL)
1209 G_RAID_DEBUG1(2, sc,
1210 "Dumping not supported by %s.",
1211 cp->provider->name);
1213 pd->pd_disk_size = pp->mediasize;
1214 if (size > pp->mediasize)
1215 size = pp->mediasize;
1216 if (sectorsize < pp->sectorsize)
1217 sectorsize = pp->sectorsize;
1222 if (sectorsize <= 0) {
1223 gctl_error(req, "Can't get sector size.");
1227 /* Reserve space for metadata. */
1228 size -= 0x800 * sectorsize;
1230 /* Handle size argument. */
1231 len = sizeof(*sizearg);
1232 sizearg = gctl_get_param(req, "size", &len);
1233 if (sizearg != NULL && len == sizeof(*sizearg) &&
1235 if (*sizearg > size) {
1236 gctl_error(req, "Size too big %lld > %lld.",
1237 (long long)*sizearg, (long long)size);
1243 /* Handle strip argument. */
1245 len = sizeof(*striparg);
1246 striparg = gctl_get_param(req, "strip", &len);
1247 if (striparg != NULL && len == sizeof(*striparg) &&
1249 if (*striparg < sectorsize) {
1250 gctl_error(req, "Strip size too small.");
1253 if (*striparg % sectorsize != 0) {
1254 gctl_error(req, "Incorrect strip size.");
1257 if (strip > 65535 * sectorsize) {
1258 gctl_error(req, "Strip size too big.");
1264 /* Round size down to strip or sector. */
1265 if (level == G_RAID_VOLUME_RL_RAID1)
1266 size -= (size % sectorsize);
1267 else if (level == G_RAID_VOLUME_RL_RAID1E &&
1268 (numdisks & 1) != 0)
1269 size -= (size % (2 * strip));
1271 size -= (size % strip);
1273 gctl_error(req, "Size too small.");
1276 if (size > 0xffffffffffffllu * sectorsize) {
1277 gctl_error(req, "Size too big.");
1281 /* We have all we need, create things: volume, ... */
1282 mdi->mdio_total_disks = numdisks;
1283 mdi->mdio_started = 1;
1284 vol = g_raid_create_volume(sc, volname, -1);
1285 vol->v_md_data = (void *)(intptr_t)0;
1286 vol->v_raid_level = level;
1287 vol->v_raid_level_qualifier = qual;
1288 vol->v_strip_size = strip;
1289 vol->v_disks_count = numdisks;
1290 if (level == G_RAID_VOLUME_RL_RAID0 ||
1291 level == G_RAID_VOLUME_RL_CONCAT ||
1292 level == G_RAID_VOLUME_RL_SINGLE)
1293 vol->v_mediasize = size * numdisks;
1294 else if (level == G_RAID_VOLUME_RL_RAID1)
1295 vol->v_mediasize = size;
1296 else if (level == G_RAID_VOLUME_RL_RAID5)
1297 vol->v_mediasize = size * (numdisks - 1);
1299 vol->v_mediasize = ((size * numdisks) / strip / 2) *
1302 vol->v_sectorsize = sectorsize;
1303 g_raid_start_volume(vol);
1305 /* , and subdisks. */
1306 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1307 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data;
1308 sd = &vol->v_subdisks[pd->pd_disk_pos];
1312 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1313 if (sd->sd_disk->d_consumer != NULL) {
1314 g_raid_change_disk_state(disk,
1315 G_RAID_DISK_S_ACTIVE);
1316 g_raid_change_subdisk_state(sd,
1317 G_RAID_SUBDISK_S_ACTIVE);
1318 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
1319 G_RAID_EVENT_SUBDISK);
1321 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1325 /* Write metadata based on created entities. */
1326 G_RAID_DEBUG1(0, sc, "Array started.");
1327 g_raid_md_write_sii(md, NULL, NULL, NULL);
1329 /* Pickup any STALE/SPARE disks to refill array if needed. */
1330 g_raid_md_sii_refill(sc);
1332 g_raid_event_send(vol, G_RAID_VOLUME_E_START,
1333 G_RAID_EVENT_VOLUME);
1336 if (strcmp(verb, "delete") == 0) {
1338 /* Check if some volume is still open. */
1339 force = gctl_get_paraml(req, "force", sizeof(*force));
1340 if (force != NULL && *force == 0 &&
1341 g_raid_nopens(sc) != 0) {
1342 gctl_error(req, "Some volume is still open.");
1346 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1347 if (disk->d_consumer)
1348 sii_meta_erase(disk->d_consumer);
1350 g_raid_destroy_node(sc, 0);
1353 if (strcmp(verb, "remove") == 0 ||
1354 strcmp(verb, "fail") == 0) {
1356 gctl_error(req, "Invalid number of arguments.");
1359 for (i = 1; i < *nargs; i++) {
1360 snprintf(arg, sizeof(arg), "arg%d", i);
1361 diskname = gctl_get_asciiparam(req, arg);
1362 if (diskname == NULL) {
1363 gctl_error(req, "No disk name (%s).", arg);
1367 if (strncmp(diskname, "/dev/", 5) == 0)
1370 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1371 if (disk->d_consumer != NULL &&
1372 disk->d_consumer->provider != NULL &&
1373 strcmp(disk->d_consumer->provider->name,
1378 gctl_error(req, "Disk '%s' not found.",
1384 if (strcmp(verb, "fail") == 0) {
1385 g_raid_md_fail_disk_sii(md, NULL, disk);
1389 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data;
1391 /* Erase metadata on deleting disk. */
1392 sii_meta_erase(disk->d_consumer);
1394 /* If disk was assigned, just update statuses. */
1395 if (pd->pd_disk_pos >= 0) {
1396 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1397 g_raid_kill_consumer(sc, disk->d_consumer);
1398 disk->d_consumer = NULL;
1399 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
1400 g_raid_change_subdisk_state(sd,
1401 G_RAID_SUBDISK_S_NONE);
1402 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
1403 G_RAID_EVENT_SUBDISK);
1406 /* Otherwise -- delete. */
1407 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
1408 g_raid_destroy_disk(disk);
1412 /* Write updated metadata to remaining disks. */
1413 g_raid_md_write_sii(md, NULL, NULL, NULL);
1415 /* Check if anything left except placeholders. */
1416 if (g_raid_ndisks(sc, -1) ==
1417 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1418 g_raid_destroy_node(sc, 0);
1420 g_raid_md_sii_refill(sc);
1423 if (strcmp(verb, "insert") == 0) {
1425 gctl_error(req, "Invalid number of arguments.");
1429 for (i = 1; i < *nargs; i++) {
1430 /* Get disk name. */
1431 snprintf(arg, sizeof(arg), "arg%d", i);
1432 diskname = gctl_get_asciiparam(req, arg);
1433 if (diskname == NULL) {
1434 gctl_error(req, "No disk name (%s).", arg);
1439 /* Try to find provider with specified name. */
1441 cp = g_raid_open_consumer(sc, diskname);
1443 gctl_error(req, "Can't open disk '%s'.",
1445 g_topology_unlock();
1451 pd = malloc(sizeof(*pd), M_MD_SII, M_WAITOK | M_ZERO);
1452 pd->pd_disk_pos = -3;
1453 pd->pd_disk_size = pp->mediasize;
1455 disk = g_raid_create_disk(sc);
1456 disk->d_consumer = cp;
1457 disk->d_md_data = (void *)pd;
1459 g_topology_unlock();
1461 /* Read kernel dumping information. */
1462 disk->d_kd.offset = 0;
1463 disk->d_kd.length = OFF_MAX;
1464 len = sizeof(disk->d_kd);
1465 g_io_getattr("GEOM::kerneldump", cp, &len, &disk->d_kd);
1466 if (disk->d_kd.di.dumper == NULL)
1467 G_RAID_DEBUG1(2, sc,
1468 "Dumping not supported by %s.",
1469 cp->provider->name);
1471 /* Welcome the "new" disk. */
1472 update += g_raid_md_sii_start_disk(disk);
1473 if (disk->d_state == G_RAID_DISK_S_SPARE) {
1474 sii_meta_write_spare(cp);
1475 g_raid_destroy_disk(disk);
1476 } else if (disk->d_state != G_RAID_DISK_S_ACTIVE) {
1477 gctl_error(req, "Disk '%s' doesn't fit.",
1479 g_raid_destroy_disk(disk);
1485 /* Write new metadata if we changed something. */
1487 g_raid_md_write_sii(md, NULL, NULL, NULL);
1490 gctl_error(req, "Command '%s' is not supported.", verb);
1495 g_raid_md_write_sii(struct g_raid_md_object *md, struct g_raid_volume *tvol,
1496 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
1498 struct g_raid_softc *sc;
1499 struct g_raid_volume *vol;
1500 struct g_raid_subdisk *sd;
1501 struct g_raid_disk *disk;
1502 struct g_raid_md_sii_object *mdi;
1503 struct g_raid_md_sii_perdisk *pd;
1504 struct sii_raid_conf *meta;
1508 mdi = (struct g_raid_md_sii_object *)md;
1510 if (sc->sc_stopping == G_RAID_DESTROY_HARD)
1513 /* Bump generation. Newly written metadata may differ from previous. */
1514 mdi->mdio_generation++;
1516 /* There is only one volume. */
1517 vol = TAILQ_FIRST(&sc->sc_volumes);
1519 /* Fill global fields. */
1520 meta = malloc(sizeof(*meta), M_MD_SII, M_WAITOK | M_ZERO);
1522 memcpy(meta, mdi->mdio_meta, sizeof(*meta));
1523 meta->total_sectors = vol->v_mediasize / vol->v_sectorsize;
1524 meta->vendor_id = 0x1095;
1525 meta->version_minor = 0;
1526 meta->version_major = 2;
1527 memcpy(&meta->timestamp, &mdi->mdio_timestamp, 6);
1528 meta->strip_sectors = vol->v_strip_size / vol->v_sectorsize;
1529 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0) {
1530 meta->type = SII_T_RAID0;
1531 meta->raid0_disks = vol->v_disks_count;
1532 meta->raid1_disks = 0xff;
1533 } else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1) {
1534 meta->type = SII_T_RAID1;
1535 meta->raid0_disks = 0xff;
1536 meta->raid1_disks = vol->v_disks_count;
1537 } else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E) {
1538 meta->type = SII_T_RAID01;
1539 meta->raid0_disks = vol->v_disks_count / 2;
1540 meta->raid1_disks = 2;
1541 } else if (vol->v_raid_level == G_RAID_VOLUME_RL_CONCAT ||
1542 vol->v_raid_level == G_RAID_VOLUME_RL_SINGLE) {
1543 meta->type = SII_T_JBOD;
1544 meta->raid0_disks = vol->v_disks_count;
1545 meta->raid1_disks = 0xff;
1547 meta->type = SII_T_RAID5;
1548 meta->raid0_disks = vol->v_disks_count;
1549 meta->raid1_disks = 0xff;
1551 meta->generation = mdi->mdio_generation;
1552 meta->raid_status = vol->v_dirty ? SII_S_ONLINE : SII_S_AVAILABLE;
1553 for (i = 0; i < vol->v_disks_count; i++) {
1554 sd = &vol->v_subdisks[i];
1555 if (sd->sd_state == G_RAID_SUBDISK_S_STALE ||
1556 sd->sd_state == G_RAID_SUBDISK_S_RESYNC)
1557 meta->raid_status = SII_S_ONLINE;
1559 meta->raid_location = mdi->mdio_location;
1560 sii_meta_put_name(meta, vol->v_name);
1562 /* We are done. Print meta data and store them to disks. */
1563 if (mdi->mdio_meta != NULL)
1564 free(mdi->mdio_meta, M_MD_SII);
1565 mdi->mdio_meta = meta;
1566 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1567 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data;
1568 if (disk->d_state != G_RAID_DISK_S_ACTIVE)
1570 if (pd->pd_meta != NULL) {
1571 free(pd->pd_meta, M_MD_SII);
1574 pd->pd_meta = sii_meta_copy(meta);
1575 if ((sd = TAILQ_FIRST(&disk->d_subdisks)) != NULL) {
1576 if (sd->sd_state < G_RAID_SUBDISK_S_NEW)
1577 pd->pd_meta->disk_status = SII_S_DROPPED;
1578 else if (sd->sd_state < G_RAID_SUBDISK_S_STALE) {
1579 pd->pd_meta->disk_status = SII_S_REBUILD;
1580 pd->pd_meta->rebuild_lba =
1581 sd->sd_rebuild_pos / vol->v_sectorsize;
1583 pd->pd_meta->disk_status = SII_S_CURRENT;
1584 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1) {
1585 pd->pd_meta->disk_number = sd->sd_pos;
1586 pd->pd_meta->raid0_ident = 0xff;
1587 pd->pd_meta->raid1_ident = 0;
1588 } else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E) {
1589 pd->pd_meta->disk_number = sd->sd_pos / meta->raid1_disks;
1590 pd->pd_meta->raid0_ident = sd->sd_pos % meta->raid1_disks;
1591 pd->pd_meta->raid1_ident = sd->sd_pos / meta->raid1_disks;
1593 pd->pd_meta->disk_number = sd->sd_pos;
1594 pd->pd_meta->raid0_ident = 0;
1595 pd->pd_meta->raid1_ident = 0xff;
1598 G_RAID_DEBUG(1, "Writing SiI metadata to %s",
1599 g_raid_get_diskname(disk));
1600 g_raid_md_sii_print(pd->pd_meta);
1601 sii_meta_write(disk->d_consumer, pd->pd_meta);
1607 g_raid_md_fail_disk_sii(struct g_raid_md_object *md,
1608 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
1610 struct g_raid_softc *sc;
1611 struct g_raid_md_sii_perdisk *pd;
1612 struct g_raid_subdisk *sd;
1615 pd = (struct g_raid_md_sii_perdisk *)tdisk->d_md_data;
1617 /* We can't fail disk that is not a part of array now. */
1618 if (pd->pd_disk_pos < 0)
1622 * Mark disk as failed in metadata and try to write that metadata
1623 * to the disk itself to prevent it's later resurrection as STALE.
1625 if (tdisk->d_consumer) {
1627 pd->pd_meta->disk_status = SII_S_REMOVED;
1628 sii_meta_write(tdisk->d_consumer, pd->pd_meta);
1630 sii_meta_erase(tdisk->d_consumer);
1633 /* Change states. */
1634 g_raid_change_disk_state(tdisk, G_RAID_DISK_S_FAILED);
1635 TAILQ_FOREACH(sd, &tdisk->d_subdisks, sd_next) {
1636 g_raid_change_subdisk_state(sd,
1637 G_RAID_SUBDISK_S_FAILED);
1638 g_raid_event_send(sd, G_RAID_SUBDISK_E_FAILED,
1639 G_RAID_EVENT_SUBDISK);
1642 /* Write updated metadata to remaining disks. */
1643 g_raid_md_write_sii(md, NULL, NULL, tdisk);
1645 /* Check if anything left except placeholders. */
1646 if (g_raid_ndisks(sc, -1) ==
1647 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1648 g_raid_destroy_node(sc, 0);
1650 g_raid_md_sii_refill(sc);
1655 g_raid_md_free_disk_sii(struct g_raid_md_object *md,
1656 struct g_raid_disk *disk)
1658 struct g_raid_md_sii_perdisk *pd;
1660 pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data;
1661 if (pd->pd_meta != NULL) {
1662 free(pd->pd_meta, M_MD_SII);
1666 disk->d_md_data = NULL;
1671 g_raid_md_free_sii(struct g_raid_md_object *md)
1673 struct g_raid_md_sii_object *mdi;
1675 mdi = (struct g_raid_md_sii_object *)md;
1676 if (!mdi->mdio_started) {
1677 mdi->mdio_started = 0;
1678 callout_stop(&mdi->mdio_start_co);
1679 G_RAID_DEBUG1(1, md->mdo_softc,
1680 "root_mount_rel %p", mdi->mdio_rootmount);
1681 root_mount_rel(mdi->mdio_rootmount);
1682 mdi->mdio_rootmount = NULL;
1684 if (mdi->mdio_meta != NULL) {
1685 free(mdi->mdio_meta, M_MD_SII);
1686 mdi->mdio_meta = NULL;
1691 G_RAID_MD_DECLARE(g_raid_md_sii);