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_NVIDIA, "md_nvidia_data", "GEOM_RAID NVIDIA metadata");
48 struct nvidia_raid_conf {
50 #define NVIDIA_MAGIC "NVIDIA "
57 uint32_t total_sectors;
64 #define NVIDIA_MAGIC0 0x00640044
66 uint64_t volume_id[2];
68 #define NVIDIA_S_IDLE 0
69 #define NVIDIA_S_INIT 2
70 #define NVIDIA_S_REBUILD 3
71 #define NVIDIA_S_UPGRADE 4
72 #define NVIDIA_S_SYNC 5
75 uint8_t orig_array_width;
77 #define NVIDIA_T_RAID0 0x0080
78 #define NVIDIA_T_RAID1 0x0081
79 #define NVIDIA_T_RAID3 0x0083
80 #define NVIDIA_T_RAID5 0x0085 /* RLQ = 00/02? */
81 #define NVIDIA_T_RAID5_SYM 0x0095 /* RLQ = 03 */
82 #define NVIDIA_T_RAID10 0x008a
83 #define NVIDIA_T_RAID01 0x8180
84 #define NVIDIA_T_CONCAT 0x00ff
87 uint32_t strip_sectors;
91 uint32_t stripe_sectors;
92 uint32_t stripe_bytes;
95 uint32_t orig_total_sectors;
97 #define NVIDIA_S_BOOTABLE 0x00000001
98 #define NVIDIA_S_DEGRADED 0x00000002
103 struct g_raid_md_nvidia_perdisk {
104 struct nvidia_raid_conf *pd_meta;
109 struct g_raid_md_nvidia_object {
110 struct g_raid_md_object mdio_base;
111 uint64_t mdio_volume_id[2];
112 struct nvidia_raid_conf *mdio_meta;
113 struct callout mdio_start_co; /* STARTING state timer. */
114 int mdio_total_disks;
115 int mdio_disks_present;
118 struct root_hold_token *mdio_rootmount; /* Root mount delay token. */
121 static g_raid_md_create_t g_raid_md_create_nvidia;
122 static g_raid_md_taste_t g_raid_md_taste_nvidia;
123 static g_raid_md_event_t g_raid_md_event_nvidia;
124 static g_raid_md_ctl_t g_raid_md_ctl_nvidia;
125 static g_raid_md_write_t g_raid_md_write_nvidia;
126 static g_raid_md_fail_disk_t g_raid_md_fail_disk_nvidia;
127 static g_raid_md_free_disk_t g_raid_md_free_disk_nvidia;
128 static g_raid_md_free_t g_raid_md_free_nvidia;
130 static kobj_method_t g_raid_md_nvidia_methods[] = {
131 KOBJMETHOD(g_raid_md_create, g_raid_md_create_nvidia),
132 KOBJMETHOD(g_raid_md_taste, g_raid_md_taste_nvidia),
133 KOBJMETHOD(g_raid_md_event, g_raid_md_event_nvidia),
134 KOBJMETHOD(g_raid_md_ctl, g_raid_md_ctl_nvidia),
135 KOBJMETHOD(g_raid_md_write, g_raid_md_write_nvidia),
136 KOBJMETHOD(g_raid_md_fail_disk, g_raid_md_fail_disk_nvidia),
137 KOBJMETHOD(g_raid_md_free_disk, g_raid_md_free_disk_nvidia),
138 KOBJMETHOD(g_raid_md_free, g_raid_md_free_nvidia),
142 static struct g_raid_md_class g_raid_md_nvidia_class = {
144 g_raid_md_nvidia_methods,
145 sizeof(struct g_raid_md_nvidia_object),
150 static int NVIDIANodeID = 1;
153 g_raid_md_nvidia_print(struct nvidia_raid_conf *meta)
156 if (g_raid_debug < 1)
159 printf("********* ATA NVIDIA RAID Metadata *********\n");
160 printf("nvidia_id <%.8s>\n", meta->nvidia_id);
161 printf("config_size %u\n", meta->config_size);
162 printf("checksum 0x%08x\n", meta->checksum);
163 printf("version 0x%04x\n", meta->version);
164 printf("disk_number %d\n", meta->disk_number);
165 printf("dummy_0 0x%02x\n", meta->dummy_0);
166 printf("total_sectors %u\n", meta->total_sectors);
167 printf("sector_size %u\n", meta->sector_size);
168 printf("name <%.16s>\n", meta->name);
169 printf("revision 0x%02x%02x%02x%02x\n",
170 meta->revision[0], meta->revision[1],
171 meta->revision[2], meta->revision[3]);
172 printf("disk_status 0x%08x\n", meta->disk_status);
173 printf("magic_0 0x%08x\n", meta->magic_0);
174 printf("volume_id 0x%016jx%016jx\n",
175 meta->volume_id[1], meta->volume_id[0]);
176 printf("state 0x%02x\n", meta->state);
177 printf("array_width %u\n", meta->array_width);
178 printf("total_disks %u\n", meta->total_disks);
179 printf("orig_array_width %u\n", meta->orig_array_width);
180 printf("type 0x%04x\n", meta->type);
181 printf("dummy_3 0x%04x\n", meta->dummy_3);
182 printf("strip_sectors %u\n", meta->strip_sectors);
183 printf("strip_bytes %u\n", meta->strip_bytes);
184 printf("strip_shift %u\n", meta->strip_shift);
185 printf("strip_mask 0x%08x\n", meta->strip_mask);
186 printf("stripe_sectors %u\n", meta->stripe_sectors);
187 printf("stripe_bytes %u\n", meta->stripe_bytes);
188 printf("rebuild_lba %u\n", meta->rebuild_lba);
189 printf("orig_type 0x%04x\n", meta->orig_type);
190 printf("orig_total_sectors %u\n", meta->orig_total_sectors);
191 printf("status 0x%08x\n", meta->status);
192 printf("=================================================\n");
195 static struct nvidia_raid_conf *
196 nvidia_meta_copy(struct nvidia_raid_conf *meta)
198 struct nvidia_raid_conf *nmeta;
200 nmeta = malloc(sizeof(*meta), M_MD_NVIDIA, M_WAITOK);
201 memcpy(nmeta, meta, sizeof(*meta));
206 nvidia_meta_translate_disk(struct nvidia_raid_conf *meta, int md_disk_pos)
210 if (md_disk_pos >= 0 && meta->type == NVIDIA_T_RAID01) {
211 disk_pos = (md_disk_pos / meta->array_width) +
212 (md_disk_pos % meta->array_width) * meta->array_width;
214 disk_pos = md_disk_pos;
219 nvidia_meta_get_name(struct nvidia_raid_conf *meta, char *buf)
223 strncpy(buf, meta->name, 16);
225 for (i = 15; i >= 0; i--) {
233 nvidia_meta_put_name(struct nvidia_raid_conf *meta, char *buf)
236 memset(meta->name, 0x20, 16);
237 memcpy(meta->name, buf, MIN(strlen(buf), 16));
240 static struct nvidia_raid_conf *
241 nvidia_meta_read(struct g_consumer *cp)
243 struct g_provider *pp;
244 struct nvidia_raid_conf *meta;
247 uint32_t checksum, *ptr;
251 /* Read the anchor sector. */
252 buf = g_read_data(cp,
253 pp->mediasize - 2 * pp->sectorsize, pp->sectorsize, &error);
255 G_RAID_DEBUG(1, "Cannot read metadata from %s (error=%d).",
259 meta = malloc(sizeof(*meta), M_MD_NVIDIA, M_WAITOK);
260 memcpy(meta, buf, min(sizeof(*meta), pp->sectorsize));
263 /* Check if this is an NVIDIA RAID struct */
264 if (strncmp(meta->nvidia_id, NVIDIA_MAGIC, strlen(NVIDIA_MAGIC))) {
265 G_RAID_DEBUG(1, "NVIDIA signature check failed on %s", pp->name);
266 free(meta, M_MD_NVIDIA);
269 if (meta->config_size > 128 ||
270 meta->config_size < 30) {
271 G_RAID_DEBUG(1, "NVIDIA metadata size looks wrong: %d",
273 free(meta, M_MD_NVIDIA);
277 /* Check metadata checksum. */
278 for (checksum = 0, ptr = (uint32_t *)meta,
279 i = 0; i < meta->config_size; i++)
282 G_RAID_DEBUG(1, "NVIDIA checksum check failed on %s", pp->name);
283 free(meta, M_MD_NVIDIA);
287 /* Check volume state. */
288 if (meta->state != NVIDIA_S_IDLE && meta->state != NVIDIA_S_INIT &&
289 meta->state != NVIDIA_S_REBUILD && meta->state != NVIDIA_S_SYNC) {
290 G_RAID_DEBUG(1, "NVIDIA unknown state on %s (0x%02x)",
291 pp->name, meta->state);
292 free(meta, M_MD_NVIDIA);
296 /* Check raid type. */
297 if (meta->type != NVIDIA_T_RAID0 && meta->type != NVIDIA_T_RAID1 &&
298 meta->type != NVIDIA_T_RAID3 && meta->type != NVIDIA_T_RAID5 &&
299 meta->type != NVIDIA_T_RAID5_SYM &&
300 meta->type != NVIDIA_T_RAID01 && meta->type != NVIDIA_T_CONCAT) {
301 G_RAID_DEBUG(1, "NVIDIA unknown RAID level on %s (0x%02x)",
302 pp->name, meta->type);
303 free(meta, M_MD_NVIDIA);
311 nvidia_meta_write(struct g_consumer *cp, struct nvidia_raid_conf *meta)
313 struct g_provider *pp;
316 uint32_t checksum, *ptr;
320 /* Recalculate checksum for case if metadata were changed. */
322 for (checksum = 0, ptr = (uint32_t *)meta,
323 i = 0; i < meta->config_size; i++)
325 meta->checksum -= checksum;
327 /* Create and fill buffer. */
328 buf = malloc(pp->sectorsize, M_MD_NVIDIA, M_WAITOK | M_ZERO);
329 memcpy(buf, meta, sizeof(*meta));
331 /* Write metadata. */
332 error = g_write_data(cp,
333 pp->mediasize - 2 * pp->sectorsize, buf, pp->sectorsize);
335 G_RAID_DEBUG(1, "Cannot write metadata to %s (error=%d).",
339 free(buf, M_MD_NVIDIA);
344 nvidia_meta_erase(struct g_consumer *cp)
346 struct g_provider *pp;
351 buf = malloc(pp->sectorsize, M_MD_NVIDIA, M_WAITOK | M_ZERO);
352 error = g_write_data(cp,
353 pp->mediasize - 2 * pp->sectorsize, buf, pp->sectorsize);
355 G_RAID_DEBUG(1, "Cannot erase metadata on %s (error=%d).",
358 free(buf, M_MD_NVIDIA);
362 static struct g_raid_disk *
363 g_raid_md_nvidia_get_disk(struct g_raid_softc *sc, int id)
365 struct g_raid_disk *disk;
366 struct g_raid_md_nvidia_perdisk *pd;
368 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
369 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
370 if (pd->pd_disk_pos == id)
377 g_raid_md_nvidia_supported(int level, int qual, int disks, int force)
381 case G_RAID_VOLUME_RL_RAID0:
384 if (!force && (disks < 2 || disks > 6))
387 case G_RAID_VOLUME_RL_RAID1:
390 if (!force && (disks != 2))
393 case G_RAID_VOLUME_RL_RAID1E:
398 if (!force && (disks < 4))
401 case G_RAID_VOLUME_RL_SINGLE:
405 case G_RAID_VOLUME_RL_CONCAT:
409 case G_RAID_VOLUME_RL_RAID5:
412 if (qual != G_RAID_VOLUME_RLQ_R5LA &&
413 qual != G_RAID_VOLUME_RLQ_R5LS)
419 if (level != G_RAID_VOLUME_RL_RAID5 && qual != G_RAID_VOLUME_RLQ_NONE)
425 g_raid_md_nvidia_start_disk(struct g_raid_disk *disk)
427 struct g_raid_softc *sc;
428 struct g_raid_subdisk *sd, *tmpsd;
429 struct g_raid_disk *olddisk, *tmpdisk;
430 struct g_raid_md_object *md;
431 struct g_raid_md_nvidia_object *mdi;
432 struct g_raid_md_nvidia_perdisk *pd, *oldpd;
433 struct nvidia_raid_conf *meta;
434 int disk_pos, resurrection = 0;
438 mdi = (struct g_raid_md_nvidia_object *)md;
439 meta = mdi->mdio_meta;
440 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
443 /* Find disk position in metadata by it's serial. */
444 if (pd->pd_meta != NULL) {
445 disk_pos = pd->pd_meta->disk_number;
446 if (disk_pos >= meta->total_disks || mdi->mdio_started)
450 /* For RAID0+1 we need to translate order. */
451 disk_pos = nvidia_meta_translate_disk(meta, disk_pos);
453 G_RAID_DEBUG1(1, sc, "Unknown, probably new or stale disk");
454 /* If we are in the start process, that's all for now. */
455 if (!mdi->mdio_started)
458 * If we have already started - try to get use of the disk.
459 * Try to replace OFFLINE disks first, then FAILED.
461 TAILQ_FOREACH(tmpdisk, &sc->sc_disks, d_next) {
462 if (tmpdisk->d_state != G_RAID_DISK_S_OFFLINE &&
463 tmpdisk->d_state != G_RAID_DISK_S_FAILED)
465 /* Make sure this disk is big enough. */
466 TAILQ_FOREACH(sd, &tmpdisk->d_subdisks, sd_next) {
467 if (sd->sd_offset + sd->sd_size + 2 * 512 >
470 "Disk too small (%ju < %ju)",
472 sd->sd_offset + sd->sd_size + 512);
478 if (tmpdisk->d_state == G_RAID_DISK_S_OFFLINE) {
481 } else if (olddisk == NULL)
484 if (olddisk == NULL) {
486 g_raid_change_disk_state(disk, G_RAID_DISK_S_SPARE);
489 oldpd = (struct g_raid_md_nvidia_perdisk *)olddisk->d_md_data;
490 disk_pos = oldpd->pd_disk_pos;
494 if (olddisk == NULL) {
495 /* Find placeholder by position. */
496 olddisk = g_raid_md_nvidia_get_disk(sc, disk_pos);
498 panic("No disk at position %d!", disk_pos);
499 if (olddisk->d_state != G_RAID_DISK_S_OFFLINE) {
500 G_RAID_DEBUG1(1, sc, "More then one disk for pos %d",
502 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE);
505 oldpd = (struct g_raid_md_nvidia_perdisk *)olddisk->d_md_data;
508 /* Replace failed disk or placeholder with new disk. */
509 TAILQ_FOREACH_SAFE(sd, &olddisk->d_subdisks, sd_next, tmpsd) {
510 TAILQ_REMOVE(&olddisk->d_subdisks, sd, sd_next);
511 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
514 oldpd->pd_disk_pos = -2;
515 pd->pd_disk_pos = disk_pos;
517 /* If it was placeholder -- destroy it. */
518 if (olddisk->d_state == G_RAID_DISK_S_OFFLINE) {
519 g_raid_destroy_disk(olddisk);
521 /* Otherwise, make it STALE_FAILED. */
522 g_raid_change_disk_state(olddisk, G_RAID_DISK_S_STALE_FAILED);
525 /* Welcome the new disk. */
527 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
528 else// if (pd->pd_meta->disk_status == NVIDIA_S_CURRENT ||
529 //pd->pd_meta->disk_status == NVIDIA_S_REBUILD)
530 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
532 // g_raid_change_disk_state(disk, G_RAID_DISK_S_FAILED);
533 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
536 * Different disks may have different sizes,
537 * in concat mode. Update from real disk size.
539 if (meta->type == NVIDIA_T_CONCAT)
540 sd->sd_size = pd->pd_disk_size - 0x800 * 512;
543 /* New or ex-spare disk. */
544 g_raid_change_subdisk_state(sd,
545 G_RAID_SUBDISK_S_NEW);
546 } else if (meta->state == NVIDIA_S_REBUILD &&
547 (pd->pd_meta->disk_status & 0x100)) {
548 /* Rebuilding disk. */
549 g_raid_change_subdisk_state(sd,
550 G_RAID_SUBDISK_S_REBUILD);
551 sd->sd_rebuild_pos = (off_t)pd->pd_meta->rebuild_lba /
552 meta->array_width * pd->pd_meta->sector_size;
553 } else if (meta->state == NVIDIA_S_SYNC) {
554 /* Resyncing/dirty disk. */
555 g_raid_change_subdisk_state(sd,
556 G_RAID_SUBDISK_S_RESYNC);
557 sd->sd_rebuild_pos = (off_t)pd->pd_meta->rebuild_lba /
558 meta->array_width * pd->pd_meta->sector_size;
560 /* Up to date disk. */
561 g_raid_change_subdisk_state(sd,
562 G_RAID_SUBDISK_S_ACTIVE);
564 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
565 G_RAID_EVENT_SUBDISK);
568 /* Update status of our need for spare. */
569 if (mdi->mdio_started) {
570 mdi->mdio_incomplete =
571 (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) <
572 mdi->mdio_total_disks);
575 return (resurrection);
579 g_disk_md_nvidia_retaste(void *arg, int pending)
582 G_RAID_DEBUG(1, "Array is not complete, trying to retaste.");
583 g_retaste(&g_raid_class);
584 free(arg, M_MD_NVIDIA);
588 g_raid_md_nvidia_refill(struct g_raid_softc *sc)
590 struct g_raid_md_object *md;
591 struct g_raid_md_nvidia_object *mdi;
592 struct g_raid_disk *disk;
597 mdi = (struct g_raid_md_nvidia_object *)md;
600 /* Make sure we miss anything. */
601 na = g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE);
602 if (na == mdi->mdio_total_disks)
605 G_RAID_DEBUG1(1, md->mdo_softc,
606 "Array is not complete (%d of %d), "
607 "trying to refill.", na, mdi->mdio_total_disks);
609 /* Try to get use some of STALE disks. */
610 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
611 if (disk->d_state == G_RAID_DISK_S_STALE) {
612 update += g_raid_md_nvidia_start_disk(disk);
613 if (disk->d_state == G_RAID_DISK_S_ACTIVE)
620 /* Try to get use some of SPARE disks. */
621 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
622 if (disk->d_state == G_RAID_DISK_S_SPARE) {
623 update += g_raid_md_nvidia_start_disk(disk);
624 if (disk->d_state == G_RAID_DISK_S_ACTIVE)
628 } while (disk != NULL);
630 /* Write new metadata if we changed something. */
632 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
634 /* Update status of our need for spare. */
635 mdi->mdio_incomplete = (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) <
636 mdi->mdio_total_disks);
638 /* Request retaste hoping to find spare. */
639 if (mdi->mdio_incomplete) {
640 task = malloc(sizeof(struct task),
641 M_MD_NVIDIA, M_WAITOK | M_ZERO);
642 TASK_INIT(task, 0, g_disk_md_nvidia_retaste, task);
643 taskqueue_enqueue(taskqueue_swi, task);
648 g_raid_md_nvidia_start(struct g_raid_softc *sc)
650 struct g_raid_md_object *md;
651 struct g_raid_md_nvidia_object *mdi;
652 struct g_raid_md_nvidia_perdisk *pd;
653 struct nvidia_raid_conf *meta;
654 struct g_raid_volume *vol;
655 struct g_raid_subdisk *sd;
656 struct g_raid_disk *disk;
662 mdi = (struct g_raid_md_nvidia_object *)md;
663 meta = mdi->mdio_meta;
665 /* Create volumes and subdisks. */
666 nvidia_meta_get_name(meta, buf);
667 vol = g_raid_create_volume(sc, buf, -1);
668 vol->v_mediasize = (off_t)meta->total_sectors * 512;
669 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE;
670 if (meta->type == NVIDIA_T_RAID0) {
671 vol->v_raid_level = G_RAID_VOLUME_RL_RAID0;
672 size = vol->v_mediasize / mdi->mdio_total_disks;
673 } else if (meta->type == NVIDIA_T_RAID1) {
674 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1;
675 size = vol->v_mediasize;
676 } else if (meta->type == NVIDIA_T_RAID01) {
677 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E;
678 size = vol->v_mediasize / (mdi->mdio_total_disks / 2);
679 } else if (meta->type == NVIDIA_T_CONCAT) {
680 if (mdi->mdio_total_disks == 1)
681 vol->v_raid_level = G_RAID_VOLUME_RL_SINGLE;
683 vol->v_raid_level = G_RAID_VOLUME_RL_CONCAT;
685 } else if (meta->type == NVIDIA_T_RAID5) {
686 vol->v_raid_level = G_RAID_VOLUME_RL_RAID5;
687 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_R5LA;
688 size = vol->v_mediasize / (mdi->mdio_total_disks - 1);
689 } else if (meta->type == NVIDIA_T_RAID5_SYM) {
690 vol->v_raid_level = G_RAID_VOLUME_RL_RAID5;
691 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_R5LS;
692 size = vol->v_mediasize / (mdi->mdio_total_disks - 1);
694 vol->v_raid_level = G_RAID_VOLUME_RL_UNKNOWN;
697 vol->v_strip_size = meta->strip_sectors * 512; //ZZZ
698 vol->v_disks_count = mdi->mdio_total_disks;
699 vol->v_sectorsize = 512; //ZZZ
700 for (j = 0; j < vol->v_disks_count; j++) {
701 sd = &vol->v_subdisks[j];
705 g_raid_start_volume(vol);
707 /* Create disk placeholders to store data for later writing. */
708 for (disk_pos = 0; disk_pos < mdi->mdio_total_disks; disk_pos++) {
709 pd = malloc(sizeof(*pd), M_MD_NVIDIA, M_WAITOK | M_ZERO);
710 pd->pd_disk_pos = disk_pos;
711 disk = g_raid_create_disk(sc);
712 disk->d_md_data = (void *)pd;
713 disk->d_state = G_RAID_DISK_S_OFFLINE;
714 sd = &vol->v_subdisks[disk_pos];
716 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
719 /* Make all disks found till the moment take their places. */
721 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
722 if (disk->d_state == G_RAID_DISK_S_NONE) {
723 g_raid_md_nvidia_start_disk(disk);
727 } while (disk != NULL);
729 mdi->mdio_started = 1;
730 G_RAID_DEBUG1(0, sc, "Array started.");
731 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
733 /* Pickup any STALE/SPARE disks to refill array if needed. */
734 g_raid_md_nvidia_refill(sc);
736 g_raid_event_send(vol, G_RAID_VOLUME_E_START, G_RAID_EVENT_VOLUME);
738 callout_stop(&mdi->mdio_start_co);
739 G_RAID_DEBUG1(1, sc, "root_mount_rel %p", mdi->mdio_rootmount);
740 root_mount_rel(mdi->mdio_rootmount);
741 mdi->mdio_rootmount = NULL;
745 g_raid_md_nvidia_new_disk(struct g_raid_disk *disk)
747 struct g_raid_softc *sc;
748 struct g_raid_md_object *md;
749 struct g_raid_md_nvidia_object *mdi;
750 struct nvidia_raid_conf *pdmeta;
751 struct g_raid_md_nvidia_perdisk *pd;
755 mdi = (struct g_raid_md_nvidia_object *)md;
756 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
757 pdmeta = pd->pd_meta;
759 if (mdi->mdio_started) {
760 if (g_raid_md_nvidia_start_disk(disk))
761 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
763 if (mdi->mdio_meta == NULL ||
764 mdi->mdio_meta->disk_number >= mdi->mdio_meta->total_disks) {
765 G_RAID_DEBUG1(1, sc, "Newer disk");
766 if (mdi->mdio_meta != NULL)
767 free(mdi->mdio_meta, M_MD_NVIDIA);
768 mdi->mdio_meta = nvidia_meta_copy(pdmeta);
769 mdi->mdio_total_disks = pdmeta->total_disks;
770 mdi->mdio_disks_present = 1;
771 } else if (pdmeta->disk_number < mdi->mdio_meta->total_disks) {
772 mdi->mdio_disks_present++;
773 G_RAID_DEBUG1(1, sc, "Matching disk (%d of %d up)",
774 mdi->mdio_disks_present,
775 mdi->mdio_total_disks);
777 G_RAID_DEBUG1(1, sc, "Spare disk");
779 /* If we collected all needed disks - start array. */
780 if (mdi->mdio_disks_present == mdi->mdio_total_disks)
781 g_raid_md_nvidia_start(sc);
786 g_raid_nvidia_go(void *arg)
788 struct g_raid_softc *sc;
789 struct g_raid_md_object *md;
790 struct g_raid_md_nvidia_object *mdi;
794 mdi = (struct g_raid_md_nvidia_object *)md;
795 if (!mdi->mdio_started) {
796 G_RAID_DEBUG1(0, sc, "Force array start due to timeout.");
797 g_raid_event_send(sc, G_RAID_NODE_E_START, 0);
802 g_raid_md_create_nvidia(struct g_raid_md_object *md, struct g_class *mp,
805 struct g_raid_softc *sc;
806 struct g_raid_md_nvidia_object *mdi;
809 mdi = (struct g_raid_md_nvidia_object *)md;
810 arc4rand(&mdi->mdio_volume_id, 16, 0);
811 snprintf(name, sizeof(name), "NVIDIA-%d",
812 atomic_fetchadd_int(&NVIDIANodeID, 1));
813 sc = g_raid_create_node(mp, name, md);
815 return (G_RAID_MD_TASTE_FAIL);
818 return (G_RAID_MD_TASTE_NEW);
822 g_raid_md_taste_nvidia(struct g_raid_md_object *md, struct g_class *mp,
823 struct g_consumer *cp, struct g_geom **gp)
825 struct g_consumer *rcp;
826 struct g_provider *pp;
827 struct g_raid_md_nvidia_object *mdi, *mdi1;
828 struct g_raid_softc *sc;
829 struct g_raid_disk *disk;
830 struct nvidia_raid_conf *meta;
831 struct g_raid_md_nvidia_perdisk *pd;
833 int result, spare, len;
837 G_RAID_DEBUG(1, "Tasting NVIDIA on %s", cp->provider->name);
838 mdi = (struct g_raid_md_nvidia_object *)md;
841 /* Read metadata from device. */
846 if (pp->geom->rank == 1)
847 g_io_getattr("GEOM::hba_vendor", cp, &len, &vendor);
848 meta = nvidia_meta_read(cp);
851 if (g_raid_aggressive_spare) {
852 if (vendor == 0x10de) {
854 "No NVIDIA metadata, forcing spare.");
859 "NVIDIA vendor mismatch 0x%04x != 0x10de",
863 return (G_RAID_MD_TASTE_FAIL);
866 /* Metadata valid. Print it. */
867 g_raid_md_nvidia_print(meta);
868 G_RAID_DEBUG(1, "NVIDIA disk position %d", meta->disk_number);
869 spare = 0;//(meta->type == NVIDIA_T_SPARE) ? 1 : 0;
872 /* Search for matching node. */
875 LIST_FOREACH(geom, &mp->geom, geom) {
879 if (sc->sc_stopping != 0)
881 if (sc->sc_md->mdo_class != md->mdo_class)
883 mdi1 = (struct g_raid_md_nvidia_object *)sc->sc_md;
885 if (mdi1->mdio_incomplete)
888 if (memcmp(&mdi1->mdio_volume_id,
889 &meta->volume_id, 16) == 0)
894 /* Found matching node. */
896 G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name);
897 result = G_RAID_MD_TASTE_EXISTING;
899 } else if (spare) { /* Not found needy node -- left for later. */
900 G_RAID_DEBUG(1, "Spare is not needed at this time");
903 } else { /* Not found matching node -- create one. */
904 result = G_RAID_MD_TASTE_NEW;
905 memcpy(&mdi->mdio_volume_id, &meta->volume_id, 16);
906 snprintf(name, sizeof(name), "NVIDIA-%d",
907 atomic_fetchadd_int(&NVIDIANodeID, 1));
908 sc = g_raid_create_node(mp, name, md);
911 callout_init(&mdi->mdio_start_co, 1);
912 callout_reset(&mdi->mdio_start_co, g_raid_start_timeout * hz,
913 g_raid_nvidia_go, sc);
914 mdi->mdio_rootmount = root_mount_hold("GRAID-NVIDIA");
915 G_RAID_DEBUG1(1, sc, "root_mount_hold %p", mdi->mdio_rootmount);
918 /* There is no return after this point, so we close passed consumer. */
919 g_access(cp, -1, 0, 0);
921 rcp = g_new_consumer(geom);
923 if (g_access(rcp, 1, 1, 1) != 0)
927 sx_xlock(&sc->sc_lock);
929 pd = malloc(sizeof(*pd), M_MD_NVIDIA, M_WAITOK | M_ZERO);
932 pd->pd_disk_pos = -3;
934 pd->pd_disk_pos = -1;
936 pd->pd_disk_size = pp->mediasize;
937 disk = g_raid_create_disk(sc);
938 disk->d_md_data = (void *)pd;
939 disk->d_consumer = rcp;
942 g_raid_get_disk_info(disk);
944 g_raid_md_nvidia_new_disk(disk);
946 sx_xunlock(&sc->sc_lock);
951 free(meta, M_MD_NVIDIA);
952 return (G_RAID_MD_TASTE_FAIL);
956 g_raid_md_event_nvidia(struct g_raid_md_object *md,
957 struct g_raid_disk *disk, u_int event)
959 struct g_raid_softc *sc;
960 struct g_raid_subdisk *sd;
961 struct g_raid_md_nvidia_object *mdi;
962 struct g_raid_md_nvidia_perdisk *pd;
965 mdi = (struct g_raid_md_nvidia_object *)md;
968 case G_RAID_NODE_E_START:
969 if (!mdi->mdio_started) {
970 /* Bump volume ID to drop missing disks. */
971 arc4rand(&mdi->mdio_volume_id, 16, 0);
972 g_raid_md_nvidia_start(sc);
978 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
980 case G_RAID_DISK_E_DISCONNECTED:
981 /* If disk was assigned, just update statuses. */
982 if (pd->pd_disk_pos >= 0) {
983 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
984 if (disk->d_consumer) {
985 g_raid_kill_consumer(sc, disk->d_consumer);
986 disk->d_consumer = NULL;
988 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
989 g_raid_change_subdisk_state(sd,
990 G_RAID_SUBDISK_S_NONE);
991 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
992 G_RAID_EVENT_SUBDISK);
995 /* Otherwise -- delete. */
996 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
997 g_raid_destroy_disk(disk);
1000 if (mdi->mdio_started) {
1001 /* Bump volume ID to prevent disk resurrection. */
1002 if (pd->pd_disk_pos >= 0)
1003 arc4rand(&mdi->mdio_volume_id, 16, 0);
1005 /* Write updated metadata to all disks. */
1006 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
1009 /* Check if anything left except placeholders. */
1010 if (g_raid_ndisks(sc, -1) ==
1011 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1012 g_raid_destroy_node(sc, 0);
1014 g_raid_md_nvidia_refill(sc);
1021 g_raid_md_ctl_nvidia(struct g_raid_md_object *md,
1022 struct gctl_req *req)
1024 struct g_raid_softc *sc;
1025 struct g_raid_volume *vol;
1026 struct g_raid_subdisk *sd;
1027 struct g_raid_disk *disk;
1028 struct g_raid_md_nvidia_object *mdi;
1029 struct g_raid_md_nvidia_perdisk *pd;
1030 struct g_consumer *cp;
1031 struct g_provider *pp;
1033 const char *verb, *volname, *levelname, *diskname;
1035 off_t size, sectorsize, strip, volsize;
1036 intmax_t *sizearg, *striparg;
1037 int numdisks, i, len, level, qual, update;
1041 mdi = (struct g_raid_md_nvidia_object *)md;
1042 verb = gctl_get_param(req, "verb", NULL);
1043 nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
1045 if (strcmp(verb, "label") == 0) {
1048 gctl_error(req, "Invalid number of arguments.");
1051 volname = gctl_get_asciiparam(req, "arg1");
1052 if (volname == NULL) {
1053 gctl_error(req, "No volume name.");
1056 levelname = gctl_get_asciiparam(req, "arg2");
1057 if (levelname == NULL) {
1058 gctl_error(req, "No RAID level.");
1061 if (strcasecmp(levelname, "RAID5") == 0)
1062 levelname = "RAID5-LS";
1063 if (g_raid_volume_str2level(levelname, &level, &qual)) {
1064 gctl_error(req, "Unknown RAID level '%s'.", levelname);
1067 numdisks = *nargs - 3;
1068 force = gctl_get_paraml(req, "force", sizeof(*force));
1069 if (!g_raid_md_nvidia_supported(level, qual, numdisks,
1070 force ? *force : 0)) {
1071 gctl_error(req, "Unsupported RAID level "
1072 "(0x%02x/0x%02x), or number of disks (%d).",
1073 level, qual, numdisks);
1077 /* Search for disks, connect them and probe. */
1078 size = 0x7fffffffffffffffllu;
1080 for (i = 0; i < numdisks; i++) {
1081 snprintf(arg, sizeof(arg), "arg%d", i + 3);
1082 diskname = gctl_get_asciiparam(req, arg);
1083 if (diskname == NULL) {
1084 gctl_error(req, "No disk name (%s).", arg);
1088 if (strcmp(diskname, "NONE") == 0) {
1093 cp = g_raid_open_consumer(sc, diskname);
1095 gctl_error(req, "Can't open '%s'.",
1097 g_topology_unlock();
1103 pd = malloc(sizeof(*pd), M_MD_NVIDIA, M_WAITOK | M_ZERO);
1104 pd->pd_disk_pos = i;
1105 disk = g_raid_create_disk(sc);
1106 disk->d_md_data = (void *)pd;
1107 disk->d_consumer = cp;
1111 g_topology_unlock();
1113 g_raid_get_disk_info(disk);
1115 pd->pd_disk_size = pp->mediasize;
1116 if (size > pp->mediasize)
1117 size = pp->mediasize;
1118 if (sectorsize < pp->sectorsize)
1119 sectorsize = pp->sectorsize;
1124 if (sectorsize <= 0) {
1125 gctl_error(req, "Can't get sector size.");
1129 /* Reserve space for metadata. */
1130 size -= 2 * sectorsize;
1132 /* Handle size argument. */
1133 len = sizeof(*sizearg);
1134 sizearg = gctl_get_param(req, "size", &len);
1135 if (sizearg != NULL && len == sizeof(*sizearg) &&
1137 if (*sizearg > size) {
1138 gctl_error(req, "Size too big %lld > %lld.",
1139 (long long)*sizearg, (long long)size);
1145 /* Handle strip argument. */
1147 len = sizeof(*striparg);
1148 striparg = gctl_get_param(req, "strip", &len);
1149 if (striparg != NULL && len == sizeof(*striparg) &&
1151 if (*striparg < sectorsize) {
1152 gctl_error(req, "Strip size too small.");
1155 if (*striparg % sectorsize != 0) {
1156 gctl_error(req, "Incorrect strip size.");
1159 if (strip > 65535 * sectorsize) {
1160 gctl_error(req, "Strip size too big.");
1166 /* Round size down to strip or sector. */
1167 if (level == G_RAID_VOLUME_RL_RAID1)
1168 size -= (size % sectorsize);
1169 else if (level == G_RAID_VOLUME_RL_RAID1E &&
1170 (numdisks & 1) != 0)
1171 size -= (size % (2 * strip));
1173 size -= (size % strip);
1175 gctl_error(req, "Size too small.");
1179 if (level == G_RAID_VOLUME_RL_RAID0 ||
1180 level == G_RAID_VOLUME_RL_CONCAT ||
1181 level == G_RAID_VOLUME_RL_SINGLE)
1182 volsize = size * numdisks;
1183 else if (level == G_RAID_VOLUME_RL_RAID1)
1185 else if (level == G_RAID_VOLUME_RL_RAID5)
1186 volsize = size * (numdisks - 1);
1188 volsize = ((size * numdisks) / strip / 2) *
1191 if (volsize > 0xffffffffllu * sectorsize) {
1192 gctl_error(req, "Size too big.");
1196 /* We have all we need, create things: volume, ... */
1197 mdi->mdio_total_disks = numdisks;
1198 mdi->mdio_started = 1;
1199 vol = g_raid_create_volume(sc, volname, -1);
1200 vol->v_md_data = (void *)(intptr_t)0;
1201 vol->v_raid_level = level;
1202 vol->v_raid_level_qualifier = qual;
1203 vol->v_strip_size = strip;
1204 vol->v_disks_count = numdisks;
1205 vol->v_mediasize = volsize;
1206 vol->v_sectorsize = sectorsize;
1207 g_raid_start_volume(vol);
1209 /* , and subdisks. */
1210 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1211 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
1212 sd = &vol->v_subdisks[pd->pd_disk_pos];
1216 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1217 if (sd->sd_disk->d_consumer != NULL) {
1218 g_raid_change_disk_state(disk,
1219 G_RAID_DISK_S_ACTIVE);
1220 g_raid_change_subdisk_state(sd,
1221 G_RAID_SUBDISK_S_ACTIVE);
1222 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
1223 G_RAID_EVENT_SUBDISK);
1225 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1229 /* Write metadata based on created entities. */
1230 G_RAID_DEBUG1(0, sc, "Array started.");
1231 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
1233 /* Pickup any STALE/SPARE disks to refill array if needed. */
1234 g_raid_md_nvidia_refill(sc);
1236 g_raid_event_send(vol, G_RAID_VOLUME_E_START,
1237 G_RAID_EVENT_VOLUME);
1240 if (strcmp(verb, "delete") == 0) {
1242 /* Check if some volume is still open. */
1243 force = gctl_get_paraml(req, "force", sizeof(*force));
1244 if (force != NULL && *force == 0 &&
1245 g_raid_nopens(sc) != 0) {
1246 gctl_error(req, "Some volume is still open.");
1250 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1251 if (disk->d_consumer)
1252 nvidia_meta_erase(disk->d_consumer);
1254 g_raid_destroy_node(sc, 0);
1257 if (strcmp(verb, "remove") == 0 ||
1258 strcmp(verb, "fail") == 0) {
1260 gctl_error(req, "Invalid number of arguments.");
1263 for (i = 1; i < *nargs; i++) {
1264 snprintf(arg, sizeof(arg), "arg%d", i);
1265 diskname = gctl_get_asciiparam(req, arg);
1266 if (diskname == NULL) {
1267 gctl_error(req, "No disk name (%s).", arg);
1271 if (strncmp(diskname, "/dev/", 5) == 0)
1274 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1275 if (disk->d_consumer != NULL &&
1276 disk->d_consumer->provider != NULL &&
1277 strcmp(disk->d_consumer->provider->name,
1282 gctl_error(req, "Disk '%s' not found.",
1288 if (strcmp(verb, "fail") == 0) {
1289 g_raid_md_fail_disk_nvidia(md, NULL, disk);
1293 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
1295 /* Erase metadata on deleting disk. */
1296 nvidia_meta_erase(disk->d_consumer);
1298 /* If disk was assigned, just update statuses. */
1299 if (pd->pd_disk_pos >= 0) {
1300 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1301 g_raid_kill_consumer(sc, disk->d_consumer);
1302 disk->d_consumer = NULL;
1303 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
1304 g_raid_change_subdisk_state(sd,
1305 G_RAID_SUBDISK_S_NONE);
1306 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
1307 G_RAID_EVENT_SUBDISK);
1310 /* Otherwise -- delete. */
1311 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
1312 g_raid_destroy_disk(disk);
1316 /* Write updated metadata to remaining disks. */
1317 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
1319 /* Check if anything left except placeholders. */
1320 if (g_raid_ndisks(sc, -1) ==
1321 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1322 g_raid_destroy_node(sc, 0);
1324 g_raid_md_nvidia_refill(sc);
1327 if (strcmp(verb, "insert") == 0) {
1329 gctl_error(req, "Invalid number of arguments.");
1333 for (i = 1; i < *nargs; i++) {
1334 /* Get disk name. */
1335 snprintf(arg, sizeof(arg), "arg%d", i);
1336 diskname = gctl_get_asciiparam(req, arg);
1337 if (diskname == NULL) {
1338 gctl_error(req, "No disk name (%s).", arg);
1343 /* Try to find provider with specified name. */
1345 cp = g_raid_open_consumer(sc, diskname);
1347 gctl_error(req, "Can't open disk '%s'.",
1349 g_topology_unlock();
1355 pd = malloc(sizeof(*pd), M_MD_NVIDIA, M_WAITOK | M_ZERO);
1356 pd->pd_disk_pos = -3;
1357 pd->pd_disk_size = pp->mediasize;
1359 disk = g_raid_create_disk(sc);
1360 disk->d_consumer = cp;
1361 disk->d_md_data = (void *)pd;
1363 g_topology_unlock();
1365 g_raid_get_disk_info(disk);
1367 /* Welcome the "new" disk. */
1368 update += g_raid_md_nvidia_start_disk(disk);
1369 if (disk->d_state != G_RAID_DISK_S_SPARE &&
1370 disk->d_state != G_RAID_DISK_S_ACTIVE) {
1371 gctl_error(req, "Disk '%s' doesn't fit.",
1373 g_raid_destroy_disk(disk);
1379 /* Write new metadata if we changed something. */
1381 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
1384 gctl_error(req, "Command '%s' is not supported.", verb);
1389 g_raid_md_write_nvidia(struct g_raid_md_object *md, struct g_raid_volume *tvol,
1390 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
1392 struct g_raid_softc *sc;
1393 struct g_raid_volume *vol;
1394 struct g_raid_subdisk *sd;
1395 struct g_raid_disk *disk;
1396 struct g_raid_md_nvidia_object *mdi;
1397 struct g_raid_md_nvidia_perdisk *pd;
1398 struct nvidia_raid_conf *meta;
1402 mdi = (struct g_raid_md_nvidia_object *)md;
1404 if (sc->sc_stopping == G_RAID_DESTROY_HARD)
1407 /* There is only one volume. */
1408 vol = TAILQ_FIRST(&sc->sc_volumes);
1410 /* Fill global fields. */
1411 meta = malloc(sizeof(*meta), M_MD_NVIDIA, M_WAITOK | M_ZERO);
1413 memcpy(meta, mdi->mdio_meta, sizeof(*meta));
1414 memcpy(meta->nvidia_id, NVIDIA_MAGIC, sizeof(NVIDIA_MAGIC) - 1);
1415 meta->config_size = 30;
1416 meta->version = 0x0064;
1417 meta->total_sectors = vol->v_mediasize / vol->v_sectorsize;
1418 meta->sector_size = vol->v_sectorsize;
1419 nvidia_meta_put_name(meta, vol->v_name);
1420 meta->magic_0 = NVIDIA_MAGIC0;
1421 memcpy(&meta->volume_id, &mdi->mdio_volume_id, 16);
1422 meta->state = NVIDIA_S_IDLE;
1423 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
1424 meta->array_width = 1;
1425 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E)
1426 meta->array_width = vol->v_disks_count / 2;
1427 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID5)
1428 meta->array_width = vol->v_disks_count - 1;
1430 meta->array_width = vol->v_disks_count;
1431 meta->total_disks = vol->v_disks_count;
1432 meta->orig_array_width = meta->array_width;
1433 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0)
1434 meta->type = NVIDIA_T_RAID0;
1435 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
1436 meta->type = NVIDIA_T_RAID1;
1437 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E)
1438 meta->type = NVIDIA_T_RAID01;
1439 else if (vol->v_raid_level == G_RAID_VOLUME_RL_CONCAT ||
1440 vol->v_raid_level == G_RAID_VOLUME_RL_SINGLE)
1441 meta->type = NVIDIA_T_CONCAT;
1442 else if (vol->v_raid_level_qualifier == G_RAID_VOLUME_RLQ_R5LA)
1443 meta->type = NVIDIA_T_RAID5;
1445 meta->type = NVIDIA_T_RAID5_SYM;
1446 meta->strip_sectors = vol->v_strip_size / vol->v_sectorsize;
1447 meta->strip_bytes = vol->v_strip_size;
1448 meta->strip_shift = ffs(meta->strip_sectors) - 1;
1449 meta->strip_mask = meta->strip_sectors - 1;
1450 meta->stripe_sectors = meta->strip_sectors * meta->orig_array_width;
1451 meta->stripe_bytes = meta->stripe_sectors * vol->v_sectorsize;
1452 meta->rebuild_lba = 0;
1453 meta->orig_type = meta->type;
1454 meta->orig_total_sectors = meta->total_sectors;
1457 for (i = 0; i < vol->v_disks_count; i++) {
1458 sd = &vol->v_subdisks[i];
1459 if ((sd->sd_state == G_RAID_SUBDISK_S_STALE ||
1460 sd->sd_state == G_RAID_SUBDISK_S_RESYNC ||
1462 meta->state != NVIDIA_S_REBUILD)
1463 meta->state = NVIDIA_S_SYNC;
1464 else if (sd->sd_state == G_RAID_SUBDISK_S_NEW ||
1465 sd->sd_state == G_RAID_SUBDISK_S_REBUILD)
1466 meta->state = NVIDIA_S_REBUILD;
1469 /* We are done. Print meta data and store them to disks. */
1470 if (mdi->mdio_meta != NULL)
1471 free(mdi->mdio_meta, M_MD_NVIDIA);
1472 mdi->mdio_meta = meta;
1474 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1475 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
1476 if (disk->d_state != G_RAID_DISK_S_ACTIVE &&
1477 disk->d_state != G_RAID_DISK_S_SPARE)
1479 if (pd->pd_meta != NULL) {
1480 free(pd->pd_meta, M_MD_NVIDIA);
1483 pd->pd_meta = nvidia_meta_copy(meta);
1484 if ((sd = TAILQ_FIRST(&disk->d_subdisks)) != NULL) {
1485 /* For RAID0+1 we need to translate order. */
1486 pd->pd_meta->disk_number =
1487 nvidia_meta_translate_disk(meta, sd->sd_pos);
1488 if (sd->sd_state != G_RAID_SUBDISK_S_ACTIVE) {
1489 pd->pd_meta->disk_status = 0x100;
1490 pd->pd_meta->rebuild_lba =
1491 sd->sd_rebuild_pos / vol->v_sectorsize *
1495 pd->pd_meta->disk_number = meta->total_disks + spares++;
1496 G_RAID_DEBUG(1, "Writing NVIDIA metadata to %s",
1497 g_raid_get_diskname(disk));
1498 g_raid_md_nvidia_print(pd->pd_meta);
1499 nvidia_meta_write(disk->d_consumer, pd->pd_meta);
1505 g_raid_md_fail_disk_nvidia(struct g_raid_md_object *md,
1506 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
1508 struct g_raid_softc *sc;
1509 struct g_raid_md_nvidia_perdisk *pd;
1510 struct g_raid_subdisk *sd;
1513 pd = (struct g_raid_md_nvidia_perdisk *)tdisk->d_md_data;
1515 /* We can't fail disk that is not a part of array now. */
1516 if (pd->pd_disk_pos < 0)
1519 /* Erase metadata to prevent disks's later resurrection. */
1520 if (tdisk->d_consumer)
1521 nvidia_meta_erase(tdisk->d_consumer);
1523 /* Change states. */
1524 g_raid_change_disk_state(tdisk, G_RAID_DISK_S_FAILED);
1525 TAILQ_FOREACH(sd, &tdisk->d_subdisks, sd_next) {
1526 g_raid_change_subdisk_state(sd,
1527 G_RAID_SUBDISK_S_FAILED);
1528 g_raid_event_send(sd, G_RAID_SUBDISK_E_FAILED,
1529 G_RAID_EVENT_SUBDISK);
1532 /* Write updated metadata to remaining disks. */
1533 g_raid_md_write_nvidia(md, NULL, NULL, tdisk);
1535 /* Check if anything left except placeholders. */
1536 if (g_raid_ndisks(sc, -1) ==
1537 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1538 g_raid_destroy_node(sc, 0);
1540 g_raid_md_nvidia_refill(sc);
1545 g_raid_md_free_disk_nvidia(struct g_raid_md_object *md,
1546 struct g_raid_disk *disk)
1548 struct g_raid_md_nvidia_perdisk *pd;
1550 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
1551 if (pd->pd_meta != NULL) {
1552 free(pd->pd_meta, M_MD_NVIDIA);
1555 free(pd, M_MD_NVIDIA);
1556 disk->d_md_data = NULL;
1561 g_raid_md_free_nvidia(struct g_raid_md_object *md)
1563 struct g_raid_md_nvidia_object *mdi;
1565 mdi = (struct g_raid_md_nvidia_object *)md;
1566 if (!mdi->mdio_started) {
1567 mdi->mdio_started = 0;
1568 callout_stop(&mdi->mdio_start_co);
1569 G_RAID_DEBUG1(1, md->mdo_softc,
1570 "root_mount_rel %p", mdi->mdio_rootmount);
1571 root_mount_rel(mdi->mdio_rootmount);
1572 mdi->mdio_rootmount = NULL;
1574 if (mdi->mdio_meta != NULL) {
1575 free(mdi->mdio_meta, M_MD_NVIDIA);
1576 mdi->mdio_meta = NULL;
1581 G_RAID_MD_DECLARE(nvidia, "NVIDIA");