2 * Copyright (c) 2011 Alexander Motin <mav@FreeBSD.org>
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 #include <sys/cdefs.h>
28 __FBSDID("$FreeBSD$");
30 #include <sys/param.h>
32 #include <sys/endian.h>
33 #include <sys/kernel.h>
35 #include <sys/limits.h>
37 #include <sys/malloc.h>
38 #include <sys/mutex.h>
39 #include <sys/systm.h>
40 #include <sys/taskqueue.h>
41 #include <geom/geom.h>
42 #include "geom/raid/g_raid.h"
43 #include "g_raid_md_if.h"
45 static MALLOC_DEFINE(M_MD_NVIDIA, "md_nvidia_data", "GEOM_RAID NVIDIA metadata");
47 struct nvidia_raid_conf {
49 #define NVIDIA_MAGIC "NVIDIA "
56 uint32_t total_sectors;
63 #define NVIDIA_MAGIC0 0x00640044
65 uint64_t volume_id[2];
67 #define NVIDIA_S_IDLE 0
68 #define NVIDIA_S_INIT 2
69 #define NVIDIA_S_REBUILD 3
70 #define NVIDIA_S_UPGRADE 4
71 #define NVIDIA_S_SYNC 5
74 uint8_t orig_array_width;
76 #define NVIDIA_T_RAID0 0x0080
77 #define NVIDIA_T_RAID1 0x0081
78 #define NVIDIA_T_RAID3 0x0083
79 #define NVIDIA_T_RAID5 0x0085 /* RLQ = 00/02? */
80 #define NVIDIA_T_RAID5_SYM 0x0095 /* RLQ = 03 */
81 #define NVIDIA_T_RAID10 0x008a
82 #define NVIDIA_T_RAID01 0x8180
83 #define NVIDIA_T_CONCAT 0x00ff
86 uint32_t strip_sectors;
90 uint32_t stripe_sectors;
91 uint32_t stripe_bytes;
94 uint32_t orig_total_sectors;
96 #define NVIDIA_S_BOOTABLE 0x00000001
97 #define NVIDIA_S_DEGRADED 0x00000002
102 struct g_raid_md_nvidia_perdisk {
103 struct nvidia_raid_conf *pd_meta;
108 struct g_raid_md_nvidia_object {
109 struct g_raid_md_object mdio_base;
110 uint64_t mdio_volume_id[2];
111 struct nvidia_raid_conf *mdio_meta;
112 struct callout mdio_start_co; /* STARTING state timer. */
113 int mdio_total_disks;
114 int mdio_disks_present;
117 struct root_hold_token *mdio_rootmount; /* Root mount delay token. */
120 static g_raid_md_create_t g_raid_md_create_nvidia;
121 static g_raid_md_taste_t g_raid_md_taste_nvidia;
122 static g_raid_md_event_t g_raid_md_event_nvidia;
123 static g_raid_md_ctl_t g_raid_md_ctl_nvidia;
124 static g_raid_md_write_t g_raid_md_write_nvidia;
125 static g_raid_md_fail_disk_t g_raid_md_fail_disk_nvidia;
126 static g_raid_md_free_disk_t g_raid_md_free_disk_nvidia;
127 static g_raid_md_free_t g_raid_md_free_nvidia;
129 static kobj_method_t g_raid_md_nvidia_methods[] = {
130 KOBJMETHOD(g_raid_md_create, g_raid_md_create_nvidia),
131 KOBJMETHOD(g_raid_md_taste, g_raid_md_taste_nvidia),
132 KOBJMETHOD(g_raid_md_event, g_raid_md_event_nvidia),
133 KOBJMETHOD(g_raid_md_ctl, g_raid_md_ctl_nvidia),
134 KOBJMETHOD(g_raid_md_write, g_raid_md_write_nvidia),
135 KOBJMETHOD(g_raid_md_fail_disk, g_raid_md_fail_disk_nvidia),
136 KOBJMETHOD(g_raid_md_free_disk, g_raid_md_free_disk_nvidia),
137 KOBJMETHOD(g_raid_md_free, g_raid_md_free_nvidia),
141 static struct g_raid_md_class g_raid_md_nvidia_class = {
143 g_raid_md_nvidia_methods,
144 sizeof(struct g_raid_md_nvidia_object),
148 static int NVIDIANodeID = 1;
151 g_raid_md_nvidia_print(struct nvidia_raid_conf *meta)
154 if (g_raid_debug < 1)
157 printf("********* ATA NVIDIA RAID Metadata *********\n");
158 printf("nvidia_id <%.8s>\n", meta->nvidia_id);
159 printf("config_size %u\n", meta->config_size);
160 printf("checksum 0x%08x\n", meta->checksum);
161 printf("version 0x%04x\n", meta->version);
162 printf("disk_number %d\n", meta->disk_number);
163 printf("dummy_0 0x%02x\n", meta->dummy_0);
164 printf("total_sectors %u\n", meta->total_sectors);
165 printf("sector_size %u\n", meta->sector_size);
166 printf("name <%.16s>\n", meta->name);
167 printf("revision 0x%02x%02x%02x%02x\n",
168 meta->revision[0], meta->revision[1],
169 meta->revision[2], meta->revision[3]);
170 printf("disk_status 0x%08x\n", meta->disk_status);
171 printf("magic_0 0x%08x\n", meta->magic_0);
172 printf("volume_id 0x%016jx%016jx\n",
173 meta->volume_id[1], meta->volume_id[0]);
174 printf("state 0x%02x\n", meta->state);
175 printf("array_width %u\n", meta->array_width);
176 printf("total_disks %u\n", meta->total_disks);
177 printf("orig_array_width %u\n", meta->orig_array_width);
178 printf("type 0x%04x\n", meta->type);
179 printf("dummy_3 0x%04x\n", meta->dummy_3);
180 printf("strip_sectors %u\n", meta->strip_sectors);
181 printf("strip_bytes %u\n", meta->strip_bytes);
182 printf("strip_shift %u\n", meta->strip_shift);
183 printf("strip_mask 0x%08x\n", meta->strip_mask);
184 printf("stripe_sectors %u\n", meta->stripe_sectors);
185 printf("stripe_bytes %u\n", meta->stripe_bytes);
186 printf("rebuild_lba %u\n", meta->rebuild_lba);
187 printf("orig_type 0x%04x\n", meta->orig_type);
188 printf("orig_total_sectors %u\n", meta->orig_total_sectors);
189 printf("status 0x%08x\n", meta->status);
190 printf("=================================================\n");
193 static struct nvidia_raid_conf *
194 nvidia_meta_copy(struct nvidia_raid_conf *meta)
196 struct nvidia_raid_conf *nmeta;
198 nmeta = malloc(sizeof(*meta), M_MD_NVIDIA, M_WAITOK);
199 memcpy(nmeta, meta, sizeof(*meta));
204 nvidia_meta_translate_disk(struct nvidia_raid_conf *meta, int md_disk_pos)
208 if (md_disk_pos >= 0 && meta->type == NVIDIA_T_RAID01) {
209 disk_pos = (md_disk_pos / meta->array_width) +
210 (md_disk_pos % meta->array_width) * meta->array_width;
212 disk_pos = md_disk_pos;
217 nvidia_meta_get_name(struct nvidia_raid_conf *meta, char *buf)
221 strncpy(buf, meta->name, 16);
223 for (i = 15; i >= 0; i--) {
231 nvidia_meta_put_name(struct nvidia_raid_conf *meta, char *buf)
234 memset(meta->name, 0x20, 16);
235 memcpy(meta->name, buf, MIN(strlen(buf), 16));
238 static struct nvidia_raid_conf *
239 nvidia_meta_read(struct g_consumer *cp)
241 struct g_provider *pp;
242 struct nvidia_raid_conf *meta;
245 uint32_t checksum, *ptr;
249 /* Read the anchor sector. */
250 buf = g_read_data(cp,
251 pp->mediasize - 2 * pp->sectorsize, pp->sectorsize, &error);
253 G_RAID_DEBUG(1, "Cannot read metadata from %s (error=%d).",
257 meta = malloc(sizeof(*meta), M_MD_NVIDIA, M_WAITOK);
258 memcpy(meta, buf, min(sizeof(*meta), pp->sectorsize));
261 /* Check if this is an NVIDIA RAID struct */
262 if (strncmp(meta->nvidia_id, NVIDIA_MAGIC, strlen(NVIDIA_MAGIC))) {
263 G_RAID_DEBUG(1, "NVIDIA signature check failed on %s", pp->name);
264 free(meta, M_MD_NVIDIA);
267 if (meta->config_size > 128 ||
268 meta->config_size < 30) {
269 G_RAID_DEBUG(1, "NVIDIA metadata size looks wrong: %d",
271 free(meta, M_MD_NVIDIA);
275 /* Check metadata checksum. */
276 for (checksum = 0, ptr = (uint32_t *)meta,
277 i = 0; i < meta->config_size; i++)
280 G_RAID_DEBUG(1, "NVIDIA checksum check failed on %s", pp->name);
281 free(meta, M_MD_NVIDIA);
285 /* Check volume state. */
286 if (meta->state != NVIDIA_S_IDLE && meta->state != NVIDIA_S_INIT &&
287 meta->state != NVIDIA_S_REBUILD && meta->state != NVIDIA_S_SYNC) {
288 G_RAID_DEBUG(1, "NVIDIA unknown state on %s (0x%02x)",
289 pp->name, meta->state);
290 free(meta, M_MD_NVIDIA);
294 /* Check raid type. */
295 if (meta->type != NVIDIA_T_RAID0 && meta->type != NVIDIA_T_RAID1 &&
296 meta->type != NVIDIA_T_RAID3 && meta->type != NVIDIA_T_RAID5 &&
297 meta->type != NVIDIA_T_RAID5_SYM &&
298 meta->type != NVIDIA_T_RAID01 && meta->type != NVIDIA_T_CONCAT) {
299 G_RAID_DEBUG(1, "NVIDIA unknown RAID level on %s (0x%02x)",
300 pp->name, meta->type);
301 free(meta, M_MD_NVIDIA);
309 nvidia_meta_write(struct g_consumer *cp, struct nvidia_raid_conf *meta)
311 struct g_provider *pp;
314 uint32_t checksum, *ptr;
318 /* Recalculate checksum for case if metadata were changed. */
320 for (checksum = 0, ptr = (uint32_t *)meta,
321 i = 0; i < meta->config_size; i++)
323 meta->checksum -= checksum;
325 /* Create and fill buffer. */
326 buf = malloc(pp->sectorsize, M_MD_NVIDIA, M_WAITOK | M_ZERO);
327 memcpy(buf, meta, sizeof(*meta));
329 /* Write metadata. */
330 error = g_write_data(cp,
331 pp->mediasize - 2 * pp->sectorsize, buf, pp->sectorsize);
333 G_RAID_DEBUG(1, "Cannot write metadata to %s (error=%d).",
337 free(buf, M_MD_NVIDIA);
342 nvidia_meta_erase(struct g_consumer *cp)
344 struct g_provider *pp;
349 buf = malloc(pp->sectorsize, M_MD_NVIDIA, M_WAITOK | M_ZERO);
350 error = g_write_data(cp,
351 pp->mediasize - 2 * pp->sectorsize, buf, pp->sectorsize);
353 G_RAID_DEBUG(1, "Cannot erase metadata on %s (error=%d).",
356 free(buf, M_MD_NVIDIA);
360 static struct g_raid_disk *
361 g_raid_md_nvidia_get_disk(struct g_raid_softc *sc, int id)
363 struct g_raid_disk *disk;
364 struct g_raid_md_nvidia_perdisk *pd;
366 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
367 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
368 if (pd->pd_disk_pos == id)
375 g_raid_md_nvidia_supported(int level, int qual, int disks, int force)
379 case G_RAID_VOLUME_RL_RAID0:
382 if (!force && (disks < 2 || disks > 6))
385 case G_RAID_VOLUME_RL_RAID1:
388 if (!force && (disks != 2))
391 case G_RAID_VOLUME_RL_RAID1E:
396 if (!force && (disks < 4))
399 case G_RAID_VOLUME_RL_SINGLE:
403 case G_RAID_VOLUME_RL_CONCAT:
407 case G_RAID_VOLUME_RL_RAID5:
414 if (qual != G_RAID_VOLUME_RLQ_NONE)
420 g_raid_md_nvidia_start_disk(struct g_raid_disk *disk)
422 struct g_raid_softc *sc;
423 struct g_raid_subdisk *sd, *tmpsd;
424 struct g_raid_disk *olddisk, *tmpdisk;
425 struct g_raid_md_object *md;
426 struct g_raid_md_nvidia_object *mdi;
427 struct g_raid_md_nvidia_perdisk *pd, *oldpd;
428 struct nvidia_raid_conf *meta;
429 int disk_pos, resurrection = 0;
433 mdi = (struct g_raid_md_nvidia_object *)md;
434 meta = mdi->mdio_meta;
435 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
438 /* Find disk position in metadata by it's serial. */
439 if (pd->pd_meta != NULL) {
440 disk_pos = pd->pd_meta->disk_number;
441 if (disk_pos >= meta->total_disks || mdi->mdio_started)
445 /* For RAID0+1 we need to translate order. */
446 disk_pos = nvidia_meta_translate_disk(meta, disk_pos);
448 G_RAID_DEBUG1(1, sc, "Unknown, probably new or stale disk");
449 /* If we are in the start process, that's all for now. */
450 if (!mdi->mdio_started)
453 * If we have already started - try to get use of the disk.
454 * Try to replace OFFLINE disks first, then FAILED.
456 TAILQ_FOREACH(tmpdisk, &sc->sc_disks, d_next) {
457 if (tmpdisk->d_state != G_RAID_DISK_S_OFFLINE &&
458 tmpdisk->d_state != G_RAID_DISK_S_FAILED)
460 /* Make sure this disk is big enough. */
461 TAILQ_FOREACH(sd, &tmpdisk->d_subdisks, sd_next) {
462 if (sd->sd_offset + sd->sd_size + 2 * 512 >
465 "Disk too small (%ju < %ju)",
467 sd->sd_offset + sd->sd_size + 512);
473 if (tmpdisk->d_state == G_RAID_DISK_S_OFFLINE) {
476 } else if (olddisk == NULL)
479 if (olddisk == NULL) {
481 g_raid_change_disk_state(disk, G_RAID_DISK_S_SPARE);
484 oldpd = (struct g_raid_md_nvidia_perdisk *)olddisk->d_md_data;
485 disk_pos = oldpd->pd_disk_pos;
489 if (olddisk == NULL) {
490 /* Find placeholder by position. */
491 olddisk = g_raid_md_nvidia_get_disk(sc, disk_pos);
493 panic("No disk at position %d!", disk_pos);
494 if (olddisk->d_state != G_RAID_DISK_S_OFFLINE) {
495 G_RAID_DEBUG1(1, sc, "More then one disk for pos %d",
497 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE);
500 oldpd = (struct g_raid_md_nvidia_perdisk *)olddisk->d_md_data;
503 /* Replace failed disk or placeholder with new disk. */
504 TAILQ_FOREACH_SAFE(sd, &olddisk->d_subdisks, sd_next, tmpsd) {
505 TAILQ_REMOVE(&olddisk->d_subdisks, sd, sd_next);
506 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
509 oldpd->pd_disk_pos = -2;
510 pd->pd_disk_pos = disk_pos;
512 /* If it was placeholder -- destroy it. */
513 if (olddisk->d_state == G_RAID_DISK_S_OFFLINE) {
514 g_raid_destroy_disk(olddisk);
516 /* Otherwise, make it STALE_FAILED. */
517 g_raid_change_disk_state(olddisk, G_RAID_DISK_S_STALE_FAILED);
520 /* Welcome the new disk. */
522 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
523 else// if (pd->pd_meta->disk_status == NVIDIA_S_CURRENT ||
524 //pd->pd_meta->disk_status == NVIDIA_S_REBUILD)
525 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
527 // g_raid_change_disk_state(disk, G_RAID_DISK_S_FAILED);
528 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
531 * Different disks may have different sizes,
532 * in concat mode. Update from real disk size.
534 if (meta->type == NVIDIA_T_CONCAT)
535 sd->sd_size = pd->pd_disk_size - 0x800 * 512;
538 /* New or ex-spare disk. */
539 g_raid_change_subdisk_state(sd,
540 G_RAID_SUBDISK_S_NEW);
541 } else if (meta->state == NVIDIA_S_REBUILD &&
542 (pd->pd_meta->disk_status & 0x100)) {
543 /* Rebuilding disk. */
544 g_raid_change_subdisk_state(sd,
545 G_RAID_SUBDISK_S_REBUILD);
546 sd->sd_rebuild_pos = (off_t)pd->pd_meta->rebuild_lba /
547 meta->array_width * pd->pd_meta->sector_size;
548 } else if (meta->state == NVIDIA_S_SYNC) {
549 /* Resyncing/dirty disk. */
550 g_raid_change_subdisk_state(sd,
551 G_RAID_SUBDISK_S_RESYNC);
552 sd->sd_rebuild_pos = (off_t)pd->pd_meta->rebuild_lba /
553 meta->array_width * pd->pd_meta->sector_size;
555 /* Up to date disk. */
556 g_raid_change_subdisk_state(sd,
557 G_RAID_SUBDISK_S_ACTIVE);
559 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
560 G_RAID_EVENT_SUBDISK);
563 /* Update status of our need for spare. */
564 if (mdi->mdio_started) {
565 mdi->mdio_incomplete =
566 (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) <
567 mdi->mdio_total_disks);
570 return (resurrection);
574 g_disk_md_nvidia_retaste(void *arg, int pending)
577 G_RAID_DEBUG(1, "Array is not complete, trying to retaste.");
578 g_retaste(&g_raid_class);
579 free(arg, M_MD_NVIDIA);
583 g_raid_md_nvidia_refill(struct g_raid_softc *sc)
585 struct g_raid_md_object *md;
586 struct g_raid_md_nvidia_object *mdi;
587 struct g_raid_disk *disk;
592 mdi = (struct g_raid_md_nvidia_object *)md;
595 /* Make sure we miss anything. */
596 na = g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE);
597 if (na == mdi->mdio_total_disks)
600 G_RAID_DEBUG1(1, md->mdo_softc,
601 "Array is not complete (%d of %d), "
602 "trying to refill.", na, mdi->mdio_total_disks);
604 /* Try to get use some of STALE disks. */
605 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
606 if (disk->d_state == G_RAID_DISK_S_STALE) {
607 update += g_raid_md_nvidia_start_disk(disk);
608 if (disk->d_state == G_RAID_DISK_S_ACTIVE)
615 /* Try to get use some of SPARE disks. */
616 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
617 if (disk->d_state == G_RAID_DISK_S_SPARE) {
618 update += g_raid_md_nvidia_start_disk(disk);
619 if (disk->d_state == G_RAID_DISK_S_ACTIVE)
623 } while (disk != NULL);
625 /* Write new metadata if we changed something. */
627 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
629 /* Update status of our need for spare. */
630 mdi->mdio_incomplete = (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) <
631 mdi->mdio_total_disks);
633 /* Request retaste hoping to find spare. */
634 if (mdi->mdio_incomplete) {
635 task = malloc(sizeof(struct task),
636 M_MD_NVIDIA, M_WAITOK | M_ZERO);
637 TASK_INIT(task, 0, g_disk_md_nvidia_retaste, task);
638 taskqueue_enqueue(taskqueue_swi, task);
643 g_raid_md_nvidia_start(struct g_raid_softc *sc)
645 struct g_raid_md_object *md;
646 struct g_raid_md_nvidia_object *mdi;
647 struct g_raid_md_nvidia_perdisk *pd;
648 struct nvidia_raid_conf *meta;
649 struct g_raid_volume *vol;
650 struct g_raid_subdisk *sd;
651 struct g_raid_disk *disk;
657 mdi = (struct g_raid_md_nvidia_object *)md;
658 meta = mdi->mdio_meta;
660 /* Create volumes and subdisks. */
661 nvidia_meta_get_name(meta, buf);
662 vol = g_raid_create_volume(sc, buf, -1);
663 vol->v_mediasize = (off_t)meta->total_sectors * 512;
664 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE;
665 if (meta->type == NVIDIA_T_RAID0) {
666 vol->v_raid_level = G_RAID_VOLUME_RL_RAID0;
667 size = vol->v_mediasize / mdi->mdio_total_disks;
668 } else if (meta->type == NVIDIA_T_RAID1) {
669 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1;
670 size = vol->v_mediasize;
671 } else if (meta->type == NVIDIA_T_RAID01) {
672 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E;
673 size = vol->v_mediasize / (mdi->mdio_total_disks / 2);
674 } else if (meta->type == NVIDIA_T_CONCAT) {
675 if (mdi->mdio_total_disks == 1)
676 vol->v_raid_level = G_RAID_VOLUME_RL_SINGLE;
678 vol->v_raid_level = G_RAID_VOLUME_RL_CONCAT;
680 } else if (meta->type == NVIDIA_T_RAID5) {
681 vol->v_raid_level = G_RAID_VOLUME_RL_RAID5;
682 size = vol->v_mediasize / (mdi->mdio_total_disks - 1);
683 } else if (meta->type == NVIDIA_T_RAID5_SYM) {
684 vol->v_raid_level = G_RAID_VOLUME_RL_RAID5;
685 // vol->v_raid_level_qualifier = 0x03;
686 size = vol->v_mediasize / (mdi->mdio_total_disks - 1);
688 vol->v_raid_level = G_RAID_VOLUME_RL_UNKNOWN;
691 vol->v_strip_size = meta->strip_sectors * 512; //ZZZ
692 vol->v_disks_count = mdi->mdio_total_disks;
693 vol->v_sectorsize = 512; //ZZZ
694 for (j = 0; j < vol->v_disks_count; j++) {
695 sd = &vol->v_subdisks[j];
699 g_raid_start_volume(vol);
701 /* Create disk placeholders to store data for later writing. */
702 for (disk_pos = 0; disk_pos < mdi->mdio_total_disks; disk_pos++) {
703 pd = malloc(sizeof(*pd), M_MD_NVIDIA, M_WAITOK | M_ZERO);
704 pd->pd_disk_pos = disk_pos;
705 disk = g_raid_create_disk(sc);
706 disk->d_md_data = (void *)pd;
707 disk->d_state = G_RAID_DISK_S_OFFLINE;
708 sd = &vol->v_subdisks[disk_pos];
710 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
713 /* Make all disks found till the moment take their places. */
715 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
716 if (disk->d_state == G_RAID_DISK_S_NONE) {
717 g_raid_md_nvidia_start_disk(disk);
721 } while (disk != NULL);
723 mdi->mdio_started = 1;
724 G_RAID_DEBUG1(0, sc, "Array started.");
725 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
727 /* Pickup any STALE/SPARE disks to refill array if needed. */
728 g_raid_md_nvidia_refill(sc);
730 g_raid_event_send(vol, G_RAID_VOLUME_E_START, G_RAID_EVENT_VOLUME);
732 callout_stop(&mdi->mdio_start_co);
733 G_RAID_DEBUG1(1, sc, "root_mount_rel %p", mdi->mdio_rootmount);
734 root_mount_rel(mdi->mdio_rootmount);
735 mdi->mdio_rootmount = NULL;
739 g_raid_md_nvidia_new_disk(struct g_raid_disk *disk)
741 struct g_raid_softc *sc;
742 struct g_raid_md_object *md;
743 struct g_raid_md_nvidia_object *mdi;
744 struct nvidia_raid_conf *pdmeta;
745 struct g_raid_md_nvidia_perdisk *pd;
749 mdi = (struct g_raid_md_nvidia_object *)md;
750 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
751 pdmeta = pd->pd_meta;
753 if (mdi->mdio_started) {
754 if (g_raid_md_nvidia_start_disk(disk))
755 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
757 if (mdi->mdio_meta == NULL ||
758 mdi->mdio_meta->disk_number >= mdi->mdio_meta->total_disks) {
759 G_RAID_DEBUG1(1, sc, "Newer disk");
760 if (mdi->mdio_meta != NULL)
761 free(mdi->mdio_meta, M_MD_NVIDIA);
762 mdi->mdio_meta = nvidia_meta_copy(pdmeta);
763 mdi->mdio_total_disks = pdmeta->total_disks;
764 mdi->mdio_disks_present = 1;
765 } else if (pdmeta->disk_number < mdi->mdio_meta->total_disks) {
766 mdi->mdio_disks_present++;
767 G_RAID_DEBUG1(1, sc, "Matching disk (%d of %d up)",
768 mdi->mdio_disks_present,
769 mdi->mdio_total_disks);
771 G_RAID_DEBUG1(1, sc, "Spare disk");
773 /* If we collected all needed disks - start array. */
774 if (mdi->mdio_disks_present == mdi->mdio_total_disks)
775 g_raid_md_nvidia_start(sc);
780 g_raid_nvidia_go(void *arg)
782 struct g_raid_softc *sc;
783 struct g_raid_md_object *md;
784 struct g_raid_md_nvidia_object *mdi;
788 mdi = (struct g_raid_md_nvidia_object *)md;
789 if (!mdi->mdio_started) {
790 G_RAID_DEBUG1(0, sc, "Force array start due to timeout.");
791 g_raid_event_send(sc, G_RAID_NODE_E_START, 0);
796 g_raid_md_create_nvidia(struct g_raid_md_object *md, struct g_class *mp,
799 struct g_raid_softc *sc;
800 struct g_raid_md_nvidia_object *mdi;
803 mdi = (struct g_raid_md_nvidia_object *)md;
804 arc4rand(&mdi->mdio_volume_id, 16, 0);
805 snprintf(name, sizeof(name), "NVIDIA-%d",
806 atomic_fetchadd_int(&NVIDIANodeID, 1));
807 sc = g_raid_create_node(mp, name, md);
809 return (G_RAID_MD_TASTE_FAIL);
812 return (G_RAID_MD_TASTE_NEW);
816 g_raid_md_taste_nvidia(struct g_raid_md_object *md, struct g_class *mp,
817 struct g_consumer *cp, struct g_geom **gp)
819 struct g_consumer *rcp;
820 struct g_provider *pp;
821 struct g_raid_md_nvidia_object *mdi, *mdi1;
822 struct g_raid_softc *sc;
823 struct g_raid_disk *disk;
824 struct nvidia_raid_conf *meta;
825 struct g_raid_md_nvidia_perdisk *pd;
827 int error, result, spare, len;
831 G_RAID_DEBUG(1, "Tasting NVIDIA on %s", cp->provider->name);
832 mdi = (struct g_raid_md_nvidia_object *)md;
835 /* Read metadata from device. */
838 if (g_access(cp, 1, 0, 0) != 0)
839 return (G_RAID_MD_TASTE_FAIL);
842 if (pp->geom->rank == 1)
843 g_io_getattr("GEOM::hba_vendor", cp, &len, &vendor);
844 meta = nvidia_meta_read(cp);
846 g_access(cp, -1, 0, 0);
848 if (g_raid_aggressive_spare) {
849 if (vendor == 0x10de) {
851 "No NVIDIA metadata, forcing spare.");
856 "NVIDIA vendor mismatch 0x%04x != 0x10de",
860 return (G_RAID_MD_TASTE_FAIL);
863 /* Metadata valid. Print it. */
864 g_raid_md_nvidia_print(meta);
865 G_RAID_DEBUG(1, "NVIDIA disk position %d", meta->disk_number);
866 spare = 0;//(meta->type == NVIDIA_T_SPARE) ? 1 : 0;
869 /* Search for matching node. */
872 LIST_FOREACH(geom, &mp->geom, geom) {
876 if (sc->sc_stopping != 0)
878 if (sc->sc_md->mdo_class != md->mdo_class)
880 mdi1 = (struct g_raid_md_nvidia_object *)sc->sc_md;
882 if (mdi1->mdio_incomplete)
885 if (memcmp(&mdi1->mdio_volume_id,
886 &meta->volume_id, 16) == 0)
891 /* Found matching node. */
893 G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name);
894 result = G_RAID_MD_TASTE_EXISTING;
896 } else if (spare) { /* Not found needy node -- left for later. */
897 G_RAID_DEBUG(1, "Spare is not needed at this time");
900 } else { /* Not found matching node -- create one. */
901 result = G_RAID_MD_TASTE_NEW;
902 memcpy(&mdi->mdio_volume_id, &meta->volume_id, 16);
903 snprintf(name, sizeof(name), "NVIDIA-%d",
904 atomic_fetchadd_int(&NVIDIANodeID, 1));
905 sc = g_raid_create_node(mp, name, md);
908 callout_init(&mdi->mdio_start_co, 1);
909 callout_reset(&mdi->mdio_start_co, g_raid_start_timeout * hz,
910 g_raid_nvidia_go, sc);
911 mdi->mdio_rootmount = root_mount_hold("GRAID-NVIDIA");
912 G_RAID_DEBUG1(1, sc, "root_mount_hold %p", mdi->mdio_rootmount);
915 rcp = g_new_consumer(geom);
917 if (g_access(rcp, 1, 1, 1) != 0)
921 sx_xlock(&sc->sc_lock);
923 pd = malloc(sizeof(*pd), M_MD_NVIDIA, M_WAITOK | M_ZERO);
926 pd->pd_disk_pos = -3;
928 pd->pd_disk_pos = -1;
930 pd->pd_disk_size = pp->mediasize;
931 disk = g_raid_create_disk(sc);
932 disk->d_md_data = (void *)pd;
933 disk->d_consumer = rcp;
936 /* Read kernel dumping information. */
937 disk->d_kd.offset = 0;
938 disk->d_kd.length = OFF_MAX;
939 len = sizeof(disk->d_kd);
940 error = g_io_getattr("GEOM::kerneldump", rcp, &len, &disk->d_kd);
941 if (disk->d_kd.di.dumper == NULL)
942 G_RAID_DEBUG1(2, sc, "Dumping not supported by %s: %d.",
943 rcp->provider->name, error);
945 g_raid_md_nvidia_new_disk(disk);
947 sx_xunlock(&sc->sc_lock);
952 free(meta, M_MD_NVIDIA);
953 return (G_RAID_MD_TASTE_FAIL);
957 g_raid_md_event_nvidia(struct g_raid_md_object *md,
958 struct g_raid_disk *disk, u_int event)
960 struct g_raid_softc *sc;
961 struct g_raid_subdisk *sd;
962 struct g_raid_md_nvidia_object *mdi;
963 struct g_raid_md_nvidia_perdisk *pd;
966 mdi = (struct g_raid_md_nvidia_object *)md;
969 case G_RAID_NODE_E_START:
970 if (!mdi->mdio_started) {
971 /* Bump volume ID to drop missing disks. */
972 arc4rand(&mdi->mdio_volume_id, 16, 0);
973 g_raid_md_nvidia_start(sc);
979 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
981 case G_RAID_DISK_E_DISCONNECTED:
982 /* If disk was assigned, just update statuses. */
983 if (pd->pd_disk_pos >= 0) {
984 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
985 if (disk->d_consumer) {
986 g_raid_kill_consumer(sc, disk->d_consumer);
987 disk->d_consumer = NULL;
989 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
990 g_raid_change_subdisk_state(sd,
991 G_RAID_SUBDISK_S_NONE);
992 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
993 G_RAID_EVENT_SUBDISK);
996 /* Otherwise -- delete. */
997 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
998 g_raid_destroy_disk(disk);
1001 if (mdi->mdio_started) {
1002 /* Bump volume ID to prevent disk resurrection. */
1003 if (pd->pd_disk_pos >= 0)
1004 arc4rand(&mdi->mdio_volume_id, 16, 0);
1006 /* Write updated metadata to all disks. */
1007 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
1010 /* Check if anything left except placeholders. */
1011 if (g_raid_ndisks(sc, -1) ==
1012 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1013 g_raid_destroy_node(sc, 0);
1015 g_raid_md_nvidia_refill(sc);
1022 g_raid_md_ctl_nvidia(struct g_raid_md_object *md,
1023 struct gctl_req *req)
1025 struct g_raid_softc *sc;
1026 struct g_raid_volume *vol;
1027 struct g_raid_subdisk *sd;
1028 struct g_raid_disk *disk;
1029 struct g_raid_md_nvidia_object *mdi;
1030 struct g_raid_md_nvidia_perdisk *pd;
1031 struct g_consumer *cp;
1032 struct g_provider *pp;
1034 const char *verb, *volname, *levelname, *diskname;
1036 off_t size, sectorsize, strip;
1037 intmax_t *sizearg, *striparg;
1038 int numdisks, i, len, level, qual, update;
1042 mdi = (struct g_raid_md_nvidia_object *)md;
1043 verb = gctl_get_param(req, "verb", NULL);
1044 nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
1046 if (strcmp(verb, "label") == 0) {
1049 gctl_error(req, "Invalid number of arguments.");
1052 volname = gctl_get_asciiparam(req, "arg1");
1053 if (volname == NULL) {
1054 gctl_error(req, "No volume name.");
1057 levelname = gctl_get_asciiparam(req, "arg2");
1058 if (levelname == NULL) {
1059 gctl_error(req, "No RAID level.");
1062 if (g_raid_volume_str2level(levelname, &level, &qual)) {
1063 gctl_error(req, "Unknown RAID level '%s'.", levelname);
1066 numdisks = *nargs - 3;
1067 force = gctl_get_paraml(req, "force", sizeof(*force));
1068 if (!g_raid_md_nvidia_supported(level, qual, numdisks,
1069 force ? *force : 0)) {
1070 gctl_error(req, "Unsupported RAID level "
1071 "(0x%02x/0x%02x), or number of disks (%d).",
1072 level, qual, numdisks);
1076 /* Search for disks, connect them and probe. */
1077 size = 0x7fffffffffffffffllu;
1079 for (i = 0; i < numdisks; i++) {
1080 snprintf(arg, sizeof(arg), "arg%d", i + 3);
1081 diskname = gctl_get_asciiparam(req, arg);
1082 if (diskname == NULL) {
1083 gctl_error(req, "No disk name (%s).", arg);
1087 if (strcmp(diskname, "NONE") == 0) {
1092 cp = g_raid_open_consumer(sc, diskname);
1094 gctl_error(req, "Can't open '%s'.",
1096 g_topology_unlock();
1102 pd = malloc(sizeof(*pd), M_MD_NVIDIA, M_WAITOK | M_ZERO);
1103 pd->pd_disk_pos = i;
1104 disk = g_raid_create_disk(sc);
1105 disk->d_md_data = (void *)pd;
1106 disk->d_consumer = cp;
1110 g_topology_unlock();
1112 /* Read kernel dumping information. */
1113 disk->d_kd.offset = 0;
1114 disk->d_kd.length = OFF_MAX;
1115 len = sizeof(disk->d_kd);
1116 g_io_getattr("GEOM::kerneldump", cp, &len, &disk->d_kd);
1117 if (disk->d_kd.di.dumper == NULL)
1118 G_RAID_DEBUG1(2, sc,
1119 "Dumping not supported by %s.",
1120 cp->provider->name);
1122 pd->pd_disk_size = pp->mediasize;
1123 if (size > pp->mediasize)
1124 size = pp->mediasize;
1125 if (sectorsize < pp->sectorsize)
1126 sectorsize = pp->sectorsize;
1131 if (sectorsize <= 0) {
1132 gctl_error(req, "Can't get sector size.");
1136 /* Reserve space for metadata. */
1137 size -= 2 * sectorsize;
1139 /* Handle size argument. */
1140 len = sizeof(*sizearg);
1141 sizearg = gctl_get_param(req, "size", &len);
1142 if (sizearg != NULL && len == sizeof(*sizearg) &&
1144 if (*sizearg > size) {
1145 gctl_error(req, "Size too big %lld > %lld.",
1146 (long long)*sizearg, (long long)size);
1152 /* Handle strip argument. */
1154 len = sizeof(*striparg);
1155 striparg = gctl_get_param(req, "strip", &len);
1156 if (striparg != NULL && len == sizeof(*striparg) &&
1158 if (*striparg < sectorsize) {
1159 gctl_error(req, "Strip size too small.");
1162 if (*striparg % sectorsize != 0) {
1163 gctl_error(req, "Incorrect strip size.");
1166 if (strip > 65535 * sectorsize) {
1167 gctl_error(req, "Strip size too big.");
1173 /* Round size down to strip or sector. */
1174 if (level == G_RAID_VOLUME_RL_RAID1)
1175 size -= (size % sectorsize);
1176 else if (level == G_RAID_VOLUME_RL_RAID1E &&
1177 (numdisks & 1) != 0)
1178 size -= (size % (2 * strip));
1180 size -= (size % strip);
1182 gctl_error(req, "Size too small.");
1185 if (size > 0xffffffffffffllu * sectorsize) {
1186 gctl_error(req, "Size too big.");
1190 /* We have all we need, create things: volume, ... */
1191 mdi->mdio_total_disks = numdisks;
1192 mdi->mdio_started = 1;
1193 vol = g_raid_create_volume(sc, volname, -1);
1194 vol->v_md_data = (void *)(intptr_t)0;
1195 vol->v_raid_level = level;
1196 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE;
1197 vol->v_strip_size = strip;
1198 vol->v_disks_count = numdisks;
1199 if (level == G_RAID_VOLUME_RL_RAID0 ||
1200 level == G_RAID_VOLUME_RL_CONCAT ||
1201 level == G_RAID_VOLUME_RL_SINGLE)
1202 vol->v_mediasize = size * numdisks;
1203 else if (level == G_RAID_VOLUME_RL_RAID1)
1204 vol->v_mediasize = size;
1205 else if (level == G_RAID_VOLUME_RL_RAID5)
1206 vol->v_mediasize = size * (numdisks - 1);
1208 vol->v_mediasize = ((size * numdisks) / strip / 2) *
1211 vol->v_sectorsize = sectorsize;
1212 g_raid_start_volume(vol);
1214 /* , and subdisks. */
1215 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1216 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
1217 sd = &vol->v_subdisks[pd->pd_disk_pos];
1221 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1222 if (sd->sd_disk->d_consumer != NULL) {
1223 g_raid_change_disk_state(disk,
1224 G_RAID_DISK_S_ACTIVE);
1225 g_raid_change_subdisk_state(sd,
1226 G_RAID_SUBDISK_S_ACTIVE);
1227 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
1228 G_RAID_EVENT_SUBDISK);
1230 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1234 /* Write metadata based on created entities. */
1235 G_RAID_DEBUG1(0, sc, "Array started.");
1236 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
1238 /* Pickup any STALE/SPARE disks to refill array if needed. */
1239 g_raid_md_nvidia_refill(sc);
1241 g_raid_event_send(vol, G_RAID_VOLUME_E_START,
1242 G_RAID_EVENT_VOLUME);
1245 if (strcmp(verb, "delete") == 0) {
1247 /* Check if some volume is still open. */
1248 force = gctl_get_paraml(req, "force", sizeof(*force));
1249 if (force != NULL && *force == 0 &&
1250 g_raid_nopens(sc) != 0) {
1251 gctl_error(req, "Some volume is still open.");
1255 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1256 if (disk->d_consumer)
1257 nvidia_meta_erase(disk->d_consumer);
1259 g_raid_destroy_node(sc, 0);
1262 if (strcmp(verb, "remove") == 0 ||
1263 strcmp(verb, "fail") == 0) {
1265 gctl_error(req, "Invalid number of arguments.");
1268 for (i = 1; i < *nargs; i++) {
1269 snprintf(arg, sizeof(arg), "arg%d", i);
1270 diskname = gctl_get_asciiparam(req, arg);
1271 if (diskname == NULL) {
1272 gctl_error(req, "No disk name (%s).", arg);
1276 if (strncmp(diskname, "/dev/", 5) == 0)
1279 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1280 if (disk->d_consumer != NULL &&
1281 disk->d_consumer->provider != NULL &&
1282 strcmp(disk->d_consumer->provider->name,
1287 gctl_error(req, "Disk '%s' not found.",
1293 if (strcmp(verb, "fail") == 0) {
1294 g_raid_md_fail_disk_nvidia(md, NULL, disk);
1298 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
1300 /* Erase metadata on deleting disk. */
1301 nvidia_meta_erase(disk->d_consumer);
1303 /* If disk was assigned, just update statuses. */
1304 if (pd->pd_disk_pos >= 0) {
1305 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1306 g_raid_kill_consumer(sc, disk->d_consumer);
1307 disk->d_consumer = NULL;
1308 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
1309 g_raid_change_subdisk_state(sd,
1310 G_RAID_SUBDISK_S_NONE);
1311 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
1312 G_RAID_EVENT_SUBDISK);
1315 /* Otherwise -- delete. */
1316 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
1317 g_raid_destroy_disk(disk);
1321 /* Write updated metadata to remaining disks. */
1322 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
1324 /* Check if anything left except placeholders. */
1325 if (g_raid_ndisks(sc, -1) ==
1326 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1327 g_raid_destroy_node(sc, 0);
1329 g_raid_md_nvidia_refill(sc);
1332 if (strcmp(verb, "insert") == 0) {
1334 gctl_error(req, "Invalid number of arguments.");
1338 for (i = 1; i < *nargs; i++) {
1339 /* Get disk name. */
1340 snprintf(arg, sizeof(arg), "arg%d", i);
1341 diskname = gctl_get_asciiparam(req, arg);
1342 if (diskname == NULL) {
1343 gctl_error(req, "No disk name (%s).", arg);
1348 /* Try to find provider with specified name. */
1350 cp = g_raid_open_consumer(sc, diskname);
1352 gctl_error(req, "Can't open disk '%s'.",
1354 g_topology_unlock();
1360 pd = malloc(sizeof(*pd), M_MD_NVIDIA, M_WAITOK | M_ZERO);
1361 pd->pd_disk_pos = -3;
1362 pd->pd_disk_size = pp->mediasize;
1364 disk = g_raid_create_disk(sc);
1365 disk->d_consumer = cp;
1366 disk->d_md_data = (void *)pd;
1368 g_topology_unlock();
1370 /* Read kernel dumping information. */
1371 disk->d_kd.offset = 0;
1372 disk->d_kd.length = OFF_MAX;
1373 len = sizeof(disk->d_kd);
1374 g_io_getattr("GEOM::kerneldump", cp, &len, &disk->d_kd);
1375 if (disk->d_kd.di.dumper == NULL)
1376 G_RAID_DEBUG1(2, sc,
1377 "Dumping not supported by %s.",
1378 cp->provider->name);
1380 /* Welcome the "new" disk. */
1381 update += g_raid_md_nvidia_start_disk(disk);
1382 if (disk->d_state != G_RAID_DISK_S_SPARE &&
1383 disk->d_state != G_RAID_DISK_S_ACTIVE) {
1384 gctl_error(req, "Disk '%s' doesn't fit.",
1386 g_raid_destroy_disk(disk);
1392 /* Write new metadata if we changed something. */
1394 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
1397 gctl_error(req, "Command '%s' is not supported.", verb);
1402 g_raid_md_write_nvidia(struct g_raid_md_object *md, struct g_raid_volume *tvol,
1403 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
1405 struct g_raid_softc *sc;
1406 struct g_raid_volume *vol;
1407 struct g_raid_subdisk *sd;
1408 struct g_raid_disk *disk;
1409 struct g_raid_md_nvidia_object *mdi;
1410 struct g_raid_md_nvidia_perdisk *pd;
1411 struct nvidia_raid_conf *meta;
1415 mdi = (struct g_raid_md_nvidia_object *)md;
1417 if (sc->sc_stopping == G_RAID_DESTROY_HARD)
1420 /* There is only one volume. */
1421 vol = TAILQ_FIRST(&sc->sc_volumes);
1423 /* Fill global fields. */
1424 meta = malloc(sizeof(*meta), M_MD_NVIDIA, M_WAITOK | M_ZERO);
1426 memcpy(meta, mdi->mdio_meta, sizeof(*meta));
1427 memcpy(meta->nvidia_id, NVIDIA_MAGIC, sizeof(NVIDIA_MAGIC) - 1);
1428 meta->config_size = 30;
1429 meta->version = 0x0064;
1430 meta->total_sectors = vol->v_mediasize / vol->v_sectorsize;
1431 meta->sector_size = vol->v_sectorsize;
1432 nvidia_meta_put_name(meta, vol->v_name);
1433 meta->magic_0 = NVIDIA_MAGIC0;
1434 memcpy(&meta->volume_id, &mdi->mdio_volume_id, 16);
1435 meta->state = NVIDIA_S_IDLE;
1436 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
1437 meta->array_width = 1;
1438 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E)
1439 meta->array_width = vol->v_disks_count / 2;
1440 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID5)
1441 meta->array_width = vol->v_disks_count - 1;
1443 meta->array_width = vol->v_disks_count;
1444 meta->total_disks = vol->v_disks_count;
1445 meta->orig_array_width = meta->array_width;
1446 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0)
1447 meta->type = NVIDIA_T_RAID0;
1448 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
1449 meta->type = NVIDIA_T_RAID1;
1450 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E)
1451 meta->type = NVIDIA_T_RAID01;
1452 else if (vol->v_raid_level == G_RAID_VOLUME_RL_CONCAT ||
1453 vol->v_raid_level == G_RAID_VOLUME_RL_SINGLE)
1454 meta->type = NVIDIA_T_CONCAT;
1455 // else if (vol->v_raid_level_qualifier == 0)
1456 // meta->type = NVIDIA_T_RAID5;
1458 meta->type = NVIDIA_T_RAID5_SYM;
1459 meta->strip_sectors = vol->v_strip_size / vol->v_sectorsize;
1460 meta->strip_bytes = vol->v_strip_size;
1461 meta->strip_shift = ffs(meta->strip_sectors) - 1;
1462 meta->strip_mask = meta->strip_sectors - 1;
1463 meta->stripe_sectors = meta->strip_sectors * meta->orig_array_width;
1464 meta->stripe_bytes = meta->stripe_sectors * vol->v_sectorsize;
1465 meta->rebuild_lba = 0;
1466 meta->orig_type = meta->type;
1467 meta->orig_total_sectors = meta->total_sectors;
1470 for (i = 0; i < vol->v_disks_count; i++) {
1471 sd = &vol->v_subdisks[i];
1472 if ((sd->sd_state == G_RAID_SUBDISK_S_STALE ||
1473 sd->sd_state == G_RAID_SUBDISK_S_RESYNC ||
1475 meta->state != NVIDIA_S_REBUILD)
1476 meta->state = NVIDIA_S_SYNC;
1477 else if (sd->sd_state == G_RAID_SUBDISK_S_NEW ||
1478 sd->sd_state == G_RAID_SUBDISK_S_REBUILD)
1479 meta->state = NVIDIA_S_REBUILD;
1482 /* We are done. Print meta data and store them to disks. */
1483 if (mdi->mdio_meta != NULL)
1484 free(mdi->mdio_meta, M_MD_NVIDIA);
1485 mdi->mdio_meta = meta;
1487 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1488 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
1489 if (disk->d_state != G_RAID_DISK_S_ACTIVE &&
1490 disk->d_state != G_RAID_DISK_S_SPARE)
1492 if (pd->pd_meta != NULL) {
1493 free(pd->pd_meta, M_MD_NVIDIA);
1496 pd->pd_meta = nvidia_meta_copy(meta);
1497 if ((sd = TAILQ_FIRST(&disk->d_subdisks)) != NULL) {
1498 /* For RAID0+1 we need to translate order. */
1499 pd->pd_meta->disk_number =
1500 nvidia_meta_translate_disk(meta, sd->sd_pos);
1501 if (sd->sd_state != G_RAID_SUBDISK_S_ACTIVE) {
1502 pd->pd_meta->disk_status = 0x100;
1503 pd->pd_meta->rebuild_lba =
1504 sd->sd_rebuild_pos / vol->v_sectorsize *
1508 pd->pd_meta->disk_number = meta->total_disks + spares++;
1509 G_RAID_DEBUG(1, "Writing NVIDIA metadata to %s",
1510 g_raid_get_diskname(disk));
1511 g_raid_md_nvidia_print(pd->pd_meta);
1512 nvidia_meta_write(disk->d_consumer, pd->pd_meta);
1518 g_raid_md_fail_disk_nvidia(struct g_raid_md_object *md,
1519 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
1521 struct g_raid_softc *sc;
1522 struct g_raid_md_nvidia_perdisk *pd;
1523 struct g_raid_subdisk *sd;
1526 pd = (struct g_raid_md_nvidia_perdisk *)tdisk->d_md_data;
1528 /* We can't fail disk that is not a part of array now. */
1529 if (pd->pd_disk_pos < 0)
1532 /* Erase metadata to prevent disks's later resurrection. */
1533 if (tdisk->d_consumer)
1534 nvidia_meta_erase(tdisk->d_consumer);
1536 /* Change states. */
1537 g_raid_change_disk_state(tdisk, G_RAID_DISK_S_FAILED);
1538 TAILQ_FOREACH(sd, &tdisk->d_subdisks, sd_next) {
1539 g_raid_change_subdisk_state(sd,
1540 G_RAID_SUBDISK_S_FAILED);
1541 g_raid_event_send(sd, G_RAID_SUBDISK_E_FAILED,
1542 G_RAID_EVENT_SUBDISK);
1545 /* Write updated metadata to remaining disks. */
1546 g_raid_md_write_nvidia(md, NULL, NULL, tdisk);
1548 /* Check if anything left except placeholders. */
1549 if (g_raid_ndisks(sc, -1) ==
1550 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1551 g_raid_destroy_node(sc, 0);
1553 g_raid_md_nvidia_refill(sc);
1558 g_raid_md_free_disk_nvidia(struct g_raid_md_object *md,
1559 struct g_raid_disk *disk)
1561 struct g_raid_md_nvidia_perdisk *pd;
1563 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
1564 if (pd->pd_meta != NULL) {
1565 free(pd->pd_meta, M_MD_NVIDIA);
1568 free(pd, M_MD_NVIDIA);
1569 disk->d_md_data = NULL;
1574 g_raid_md_free_nvidia(struct g_raid_md_object *md)
1576 struct g_raid_md_nvidia_object *mdi;
1578 mdi = (struct g_raid_md_nvidia_object *)md;
1579 if (!mdi->mdio_started) {
1580 mdi->mdio_started = 0;
1581 callout_stop(&mdi->mdio_start_co);
1582 G_RAID_DEBUG1(1, md->mdo_softc,
1583 "root_mount_rel %p", mdi->mdio_rootmount);
1584 root_mount_rel(mdi->mdio_rootmount);
1585 mdi->mdio_rootmount = NULL;
1587 if (mdi->mdio_meta != NULL) {
1588 free(mdi->mdio_meta, M_MD_NVIDIA);
1589 mdi->mdio_meta = NULL;
1594 G_RAID_MD_DECLARE(g_raid_md_nvidia);