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),
149 static int NVIDIANodeID = 1;
152 g_raid_md_nvidia_print(struct nvidia_raid_conf *meta)
155 if (g_raid_debug < 1)
158 printf("********* ATA NVIDIA RAID Metadata *********\n");
159 printf("nvidia_id <%.8s>\n", meta->nvidia_id);
160 printf("config_size %u\n", meta->config_size);
161 printf("checksum 0x%08x\n", meta->checksum);
162 printf("version 0x%04x\n", meta->version);
163 printf("disk_number %d\n", meta->disk_number);
164 printf("dummy_0 0x%02x\n", meta->dummy_0);
165 printf("total_sectors %u\n", meta->total_sectors);
166 printf("sector_size %u\n", meta->sector_size);
167 printf("name <%.16s>\n", meta->name);
168 printf("revision 0x%02x%02x%02x%02x\n",
169 meta->revision[0], meta->revision[1],
170 meta->revision[2], meta->revision[3]);
171 printf("disk_status 0x%08x\n", meta->disk_status);
172 printf("magic_0 0x%08x\n", meta->magic_0);
173 printf("volume_id 0x%016jx%016jx\n",
174 meta->volume_id[1], meta->volume_id[0]);
175 printf("state 0x%02x\n", meta->state);
176 printf("array_width %u\n", meta->array_width);
177 printf("total_disks %u\n", meta->total_disks);
178 printf("orig_array_width %u\n", meta->orig_array_width);
179 printf("type 0x%04x\n", meta->type);
180 printf("dummy_3 0x%04x\n", meta->dummy_3);
181 printf("strip_sectors %u\n", meta->strip_sectors);
182 printf("strip_bytes %u\n", meta->strip_bytes);
183 printf("strip_shift %u\n", meta->strip_shift);
184 printf("strip_mask 0x%08x\n", meta->strip_mask);
185 printf("stripe_sectors %u\n", meta->stripe_sectors);
186 printf("stripe_bytes %u\n", meta->stripe_bytes);
187 printf("rebuild_lba %u\n", meta->rebuild_lba);
188 printf("orig_type 0x%04x\n", meta->orig_type);
189 printf("orig_total_sectors %u\n", meta->orig_total_sectors);
190 printf("status 0x%08x\n", meta->status);
191 printf("=================================================\n");
194 static struct nvidia_raid_conf *
195 nvidia_meta_copy(struct nvidia_raid_conf *meta)
197 struct nvidia_raid_conf *nmeta;
199 nmeta = malloc(sizeof(*meta), M_MD_NVIDIA, M_WAITOK);
200 memcpy(nmeta, meta, sizeof(*meta));
205 nvidia_meta_translate_disk(struct nvidia_raid_conf *meta, int md_disk_pos)
209 if (md_disk_pos >= 0 && meta->type == NVIDIA_T_RAID01) {
210 disk_pos = (md_disk_pos / meta->array_width) +
211 (md_disk_pos % meta->array_width) * meta->array_width;
213 disk_pos = md_disk_pos;
218 nvidia_meta_get_name(struct nvidia_raid_conf *meta, char *buf)
222 strncpy(buf, meta->name, 16);
224 for (i = 15; i >= 0; i--) {
232 nvidia_meta_put_name(struct nvidia_raid_conf *meta, char *buf)
235 memset(meta->name, 0x20, 16);
236 memcpy(meta->name, buf, MIN(strlen(buf), 16));
239 static struct nvidia_raid_conf *
240 nvidia_meta_read(struct g_consumer *cp)
242 struct g_provider *pp;
243 struct nvidia_raid_conf *meta;
246 uint32_t checksum, *ptr;
250 /* Read the anchor sector. */
251 buf = g_read_data(cp,
252 pp->mediasize - 2 * pp->sectorsize, pp->sectorsize, &error);
254 G_RAID_DEBUG(1, "Cannot read metadata from %s (error=%d).",
258 meta = malloc(sizeof(*meta), M_MD_NVIDIA, M_WAITOK);
259 memcpy(meta, buf, min(sizeof(*meta), pp->sectorsize));
262 /* Check if this is an NVIDIA RAID struct */
263 if (strncmp(meta->nvidia_id, NVIDIA_MAGIC, strlen(NVIDIA_MAGIC))) {
264 G_RAID_DEBUG(1, "NVIDIA signature check failed on %s", pp->name);
265 free(meta, M_MD_NVIDIA);
268 if (meta->config_size > 128 ||
269 meta->config_size < 30) {
270 G_RAID_DEBUG(1, "NVIDIA metadata size looks wrong: %d",
272 free(meta, M_MD_NVIDIA);
276 /* Check metadata checksum. */
277 for (checksum = 0, ptr = (uint32_t *)meta,
278 i = 0; i < meta->config_size; i++)
281 G_RAID_DEBUG(1, "NVIDIA checksum check failed on %s", pp->name);
282 free(meta, M_MD_NVIDIA);
286 /* Check volume state. */
287 if (meta->state != NVIDIA_S_IDLE && meta->state != NVIDIA_S_INIT &&
288 meta->state != NVIDIA_S_REBUILD && meta->state != NVIDIA_S_SYNC) {
289 G_RAID_DEBUG(1, "NVIDIA unknown state on %s (0x%02x)",
290 pp->name, meta->state);
291 free(meta, M_MD_NVIDIA);
295 /* Check raid type. */
296 if (meta->type != NVIDIA_T_RAID0 && meta->type != NVIDIA_T_RAID1 &&
297 meta->type != NVIDIA_T_RAID3 && meta->type != NVIDIA_T_RAID5 &&
298 meta->type != NVIDIA_T_RAID5_SYM &&
299 meta->type != NVIDIA_T_RAID01 && meta->type != NVIDIA_T_CONCAT) {
300 G_RAID_DEBUG(1, "NVIDIA unknown RAID level on %s (0x%02x)",
301 pp->name, meta->type);
302 free(meta, M_MD_NVIDIA);
310 nvidia_meta_write(struct g_consumer *cp, struct nvidia_raid_conf *meta)
312 struct g_provider *pp;
315 uint32_t checksum, *ptr;
319 /* Recalculate checksum for case if metadata were changed. */
321 for (checksum = 0, ptr = (uint32_t *)meta,
322 i = 0; i < meta->config_size; i++)
324 meta->checksum -= checksum;
326 /* Create and fill buffer. */
327 buf = malloc(pp->sectorsize, M_MD_NVIDIA, M_WAITOK | M_ZERO);
328 memcpy(buf, meta, sizeof(*meta));
330 /* Write metadata. */
331 error = g_write_data(cp,
332 pp->mediasize - 2 * pp->sectorsize, buf, pp->sectorsize);
334 G_RAID_DEBUG(1, "Cannot write metadata to %s (error=%d).",
338 free(buf, M_MD_NVIDIA);
343 nvidia_meta_erase(struct g_consumer *cp)
345 struct g_provider *pp;
350 buf = malloc(pp->sectorsize, M_MD_NVIDIA, M_WAITOK | M_ZERO);
351 error = g_write_data(cp,
352 pp->mediasize - 2 * pp->sectorsize, buf, pp->sectorsize);
354 G_RAID_DEBUG(1, "Cannot erase metadata on %s (error=%d).",
357 free(buf, M_MD_NVIDIA);
361 static struct g_raid_disk *
362 g_raid_md_nvidia_get_disk(struct g_raid_softc *sc, int id)
364 struct g_raid_disk *disk;
365 struct g_raid_md_nvidia_perdisk *pd;
367 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
368 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
369 if (pd->pd_disk_pos == id)
376 g_raid_md_nvidia_supported(int level, int qual, int disks, int force)
380 case G_RAID_VOLUME_RL_RAID0:
383 if (!force && (disks < 2 || disks > 6))
386 case G_RAID_VOLUME_RL_RAID1:
389 if (!force && (disks != 2))
392 case G_RAID_VOLUME_RL_RAID1E:
397 if (!force && (disks < 4))
400 case G_RAID_VOLUME_RL_SINGLE:
404 case G_RAID_VOLUME_RL_CONCAT:
408 case G_RAID_VOLUME_RL_RAID5:
415 if (qual != G_RAID_VOLUME_RLQ_NONE)
421 g_raid_md_nvidia_start_disk(struct g_raid_disk *disk)
423 struct g_raid_softc *sc;
424 struct g_raid_subdisk *sd, *tmpsd;
425 struct g_raid_disk *olddisk, *tmpdisk;
426 struct g_raid_md_object *md;
427 struct g_raid_md_nvidia_object *mdi;
428 struct g_raid_md_nvidia_perdisk *pd, *oldpd;
429 struct nvidia_raid_conf *meta;
430 int disk_pos, resurrection = 0;
434 mdi = (struct g_raid_md_nvidia_object *)md;
435 meta = mdi->mdio_meta;
436 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
439 /* Find disk position in metadata by it's serial. */
440 if (pd->pd_meta != NULL) {
441 disk_pos = pd->pd_meta->disk_number;
442 if (disk_pos >= meta->total_disks || mdi->mdio_started)
446 /* For RAID0+1 we need to translate order. */
447 disk_pos = nvidia_meta_translate_disk(meta, disk_pos);
449 G_RAID_DEBUG1(1, sc, "Unknown, probably new or stale disk");
450 /* If we are in the start process, that's all for now. */
451 if (!mdi->mdio_started)
454 * If we have already started - try to get use of the disk.
455 * Try to replace OFFLINE disks first, then FAILED.
457 TAILQ_FOREACH(tmpdisk, &sc->sc_disks, d_next) {
458 if (tmpdisk->d_state != G_RAID_DISK_S_OFFLINE &&
459 tmpdisk->d_state != G_RAID_DISK_S_FAILED)
461 /* Make sure this disk is big enough. */
462 TAILQ_FOREACH(sd, &tmpdisk->d_subdisks, sd_next) {
463 if (sd->sd_offset + sd->sd_size + 2 * 512 >
466 "Disk too small (%ju < %ju)",
468 sd->sd_offset + sd->sd_size + 512);
474 if (tmpdisk->d_state == G_RAID_DISK_S_OFFLINE) {
477 } else if (olddisk == NULL)
480 if (olddisk == NULL) {
482 g_raid_change_disk_state(disk, G_RAID_DISK_S_SPARE);
485 oldpd = (struct g_raid_md_nvidia_perdisk *)olddisk->d_md_data;
486 disk_pos = oldpd->pd_disk_pos;
490 if (olddisk == NULL) {
491 /* Find placeholder by position. */
492 olddisk = g_raid_md_nvidia_get_disk(sc, disk_pos);
494 panic("No disk at position %d!", disk_pos);
495 if (olddisk->d_state != G_RAID_DISK_S_OFFLINE) {
496 G_RAID_DEBUG1(1, sc, "More then one disk for pos %d",
498 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE);
501 oldpd = (struct g_raid_md_nvidia_perdisk *)olddisk->d_md_data;
504 /* Replace failed disk or placeholder with new disk. */
505 TAILQ_FOREACH_SAFE(sd, &olddisk->d_subdisks, sd_next, tmpsd) {
506 TAILQ_REMOVE(&olddisk->d_subdisks, sd, sd_next);
507 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
510 oldpd->pd_disk_pos = -2;
511 pd->pd_disk_pos = disk_pos;
513 /* If it was placeholder -- destroy it. */
514 if (olddisk->d_state == G_RAID_DISK_S_OFFLINE) {
515 g_raid_destroy_disk(olddisk);
517 /* Otherwise, make it STALE_FAILED. */
518 g_raid_change_disk_state(olddisk, G_RAID_DISK_S_STALE_FAILED);
521 /* Welcome the new disk. */
523 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
524 else// if (pd->pd_meta->disk_status == NVIDIA_S_CURRENT ||
525 //pd->pd_meta->disk_status == NVIDIA_S_REBUILD)
526 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
528 // g_raid_change_disk_state(disk, G_RAID_DISK_S_FAILED);
529 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
532 * Different disks may have different sizes,
533 * in concat mode. Update from real disk size.
535 if (meta->type == NVIDIA_T_CONCAT)
536 sd->sd_size = pd->pd_disk_size - 0x800 * 512;
539 /* New or ex-spare disk. */
540 g_raid_change_subdisk_state(sd,
541 G_RAID_SUBDISK_S_NEW);
542 } else if (meta->state == NVIDIA_S_REBUILD &&
543 (pd->pd_meta->disk_status & 0x100)) {
544 /* Rebuilding disk. */
545 g_raid_change_subdisk_state(sd,
546 G_RAID_SUBDISK_S_REBUILD);
547 sd->sd_rebuild_pos = (off_t)pd->pd_meta->rebuild_lba /
548 meta->array_width * pd->pd_meta->sector_size;
549 } else if (meta->state == NVIDIA_S_SYNC) {
550 /* Resyncing/dirty disk. */
551 g_raid_change_subdisk_state(sd,
552 G_RAID_SUBDISK_S_RESYNC);
553 sd->sd_rebuild_pos = (off_t)pd->pd_meta->rebuild_lba /
554 meta->array_width * pd->pd_meta->sector_size;
556 /* Up to date disk. */
557 g_raid_change_subdisk_state(sd,
558 G_RAID_SUBDISK_S_ACTIVE);
560 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
561 G_RAID_EVENT_SUBDISK);
564 /* Update status of our need for spare. */
565 if (mdi->mdio_started) {
566 mdi->mdio_incomplete =
567 (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) <
568 mdi->mdio_total_disks);
571 return (resurrection);
575 g_disk_md_nvidia_retaste(void *arg, int pending)
578 G_RAID_DEBUG(1, "Array is not complete, trying to retaste.");
579 g_retaste(&g_raid_class);
580 free(arg, M_MD_NVIDIA);
584 g_raid_md_nvidia_refill(struct g_raid_softc *sc)
586 struct g_raid_md_object *md;
587 struct g_raid_md_nvidia_object *mdi;
588 struct g_raid_disk *disk;
593 mdi = (struct g_raid_md_nvidia_object *)md;
596 /* Make sure we miss anything. */
597 na = g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE);
598 if (na == mdi->mdio_total_disks)
601 G_RAID_DEBUG1(1, md->mdo_softc,
602 "Array is not complete (%d of %d), "
603 "trying to refill.", na, mdi->mdio_total_disks);
605 /* Try to get use some of STALE disks. */
606 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
607 if (disk->d_state == G_RAID_DISK_S_STALE) {
608 update += g_raid_md_nvidia_start_disk(disk);
609 if (disk->d_state == G_RAID_DISK_S_ACTIVE)
616 /* Try to get use some of SPARE disks. */
617 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
618 if (disk->d_state == G_RAID_DISK_S_SPARE) {
619 update += g_raid_md_nvidia_start_disk(disk);
620 if (disk->d_state == G_RAID_DISK_S_ACTIVE)
624 } while (disk != NULL);
626 /* Write new metadata if we changed something. */
628 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
630 /* Update status of our need for spare. */
631 mdi->mdio_incomplete = (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) <
632 mdi->mdio_total_disks);
634 /* Request retaste hoping to find spare. */
635 if (mdi->mdio_incomplete) {
636 task = malloc(sizeof(struct task),
637 M_MD_NVIDIA, M_WAITOK | M_ZERO);
638 TASK_INIT(task, 0, g_disk_md_nvidia_retaste, task);
639 taskqueue_enqueue(taskqueue_swi, task);
644 g_raid_md_nvidia_start(struct g_raid_softc *sc)
646 struct g_raid_md_object *md;
647 struct g_raid_md_nvidia_object *mdi;
648 struct g_raid_md_nvidia_perdisk *pd;
649 struct nvidia_raid_conf *meta;
650 struct g_raid_volume *vol;
651 struct g_raid_subdisk *sd;
652 struct g_raid_disk *disk;
658 mdi = (struct g_raid_md_nvidia_object *)md;
659 meta = mdi->mdio_meta;
661 /* Create volumes and subdisks. */
662 nvidia_meta_get_name(meta, buf);
663 vol = g_raid_create_volume(sc, buf, -1);
664 vol->v_mediasize = (off_t)meta->total_sectors * 512;
665 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE;
666 if (meta->type == NVIDIA_T_RAID0) {
667 vol->v_raid_level = G_RAID_VOLUME_RL_RAID0;
668 size = vol->v_mediasize / mdi->mdio_total_disks;
669 } else if (meta->type == NVIDIA_T_RAID1) {
670 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1;
671 size = vol->v_mediasize;
672 } else if (meta->type == NVIDIA_T_RAID01) {
673 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E;
674 size = vol->v_mediasize / (mdi->mdio_total_disks / 2);
675 } else if (meta->type == NVIDIA_T_CONCAT) {
676 if (mdi->mdio_total_disks == 1)
677 vol->v_raid_level = G_RAID_VOLUME_RL_SINGLE;
679 vol->v_raid_level = G_RAID_VOLUME_RL_CONCAT;
681 } else if (meta->type == NVIDIA_T_RAID5) {
682 vol->v_raid_level = G_RAID_VOLUME_RL_RAID5;
683 size = vol->v_mediasize / (mdi->mdio_total_disks - 1);
684 } else if (meta->type == NVIDIA_T_RAID5_SYM) {
685 vol->v_raid_level = G_RAID_VOLUME_RL_RAID5;
686 // vol->v_raid_level_qualifier = 0x03;
687 size = vol->v_mediasize / (mdi->mdio_total_disks - 1);
689 vol->v_raid_level = G_RAID_VOLUME_RL_UNKNOWN;
692 vol->v_strip_size = meta->strip_sectors * 512; //ZZZ
693 vol->v_disks_count = mdi->mdio_total_disks;
694 vol->v_sectorsize = 512; //ZZZ
695 for (j = 0; j < vol->v_disks_count; j++) {
696 sd = &vol->v_subdisks[j];
700 g_raid_start_volume(vol);
702 /* Create disk placeholders to store data for later writing. */
703 for (disk_pos = 0; disk_pos < mdi->mdio_total_disks; disk_pos++) {
704 pd = malloc(sizeof(*pd), M_MD_NVIDIA, M_WAITOK | M_ZERO);
705 pd->pd_disk_pos = disk_pos;
706 disk = g_raid_create_disk(sc);
707 disk->d_md_data = (void *)pd;
708 disk->d_state = G_RAID_DISK_S_OFFLINE;
709 sd = &vol->v_subdisks[disk_pos];
711 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
714 /* Make all disks found till the moment take their places. */
716 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
717 if (disk->d_state == G_RAID_DISK_S_NONE) {
718 g_raid_md_nvidia_start_disk(disk);
722 } while (disk != NULL);
724 mdi->mdio_started = 1;
725 G_RAID_DEBUG1(0, sc, "Array started.");
726 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
728 /* Pickup any STALE/SPARE disks to refill array if needed. */
729 g_raid_md_nvidia_refill(sc);
731 g_raid_event_send(vol, G_RAID_VOLUME_E_START, G_RAID_EVENT_VOLUME);
733 callout_stop(&mdi->mdio_start_co);
734 G_RAID_DEBUG1(1, sc, "root_mount_rel %p", mdi->mdio_rootmount);
735 root_mount_rel(mdi->mdio_rootmount);
736 mdi->mdio_rootmount = NULL;
740 g_raid_md_nvidia_new_disk(struct g_raid_disk *disk)
742 struct g_raid_softc *sc;
743 struct g_raid_md_object *md;
744 struct g_raid_md_nvidia_object *mdi;
745 struct nvidia_raid_conf *pdmeta;
746 struct g_raid_md_nvidia_perdisk *pd;
750 mdi = (struct g_raid_md_nvidia_object *)md;
751 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
752 pdmeta = pd->pd_meta;
754 if (mdi->mdio_started) {
755 if (g_raid_md_nvidia_start_disk(disk))
756 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
758 if (mdi->mdio_meta == NULL ||
759 mdi->mdio_meta->disk_number >= mdi->mdio_meta->total_disks) {
760 G_RAID_DEBUG1(1, sc, "Newer disk");
761 if (mdi->mdio_meta != NULL)
762 free(mdi->mdio_meta, M_MD_NVIDIA);
763 mdi->mdio_meta = nvidia_meta_copy(pdmeta);
764 mdi->mdio_total_disks = pdmeta->total_disks;
765 mdi->mdio_disks_present = 1;
766 } else if (pdmeta->disk_number < mdi->mdio_meta->total_disks) {
767 mdi->mdio_disks_present++;
768 G_RAID_DEBUG1(1, sc, "Matching disk (%d of %d up)",
769 mdi->mdio_disks_present,
770 mdi->mdio_total_disks);
772 G_RAID_DEBUG1(1, sc, "Spare disk");
774 /* If we collected all needed disks - start array. */
775 if (mdi->mdio_disks_present == mdi->mdio_total_disks)
776 g_raid_md_nvidia_start(sc);
781 g_raid_nvidia_go(void *arg)
783 struct g_raid_softc *sc;
784 struct g_raid_md_object *md;
785 struct g_raid_md_nvidia_object *mdi;
789 mdi = (struct g_raid_md_nvidia_object *)md;
790 if (!mdi->mdio_started) {
791 G_RAID_DEBUG1(0, sc, "Force array start due to timeout.");
792 g_raid_event_send(sc, G_RAID_NODE_E_START, 0);
797 g_raid_md_create_nvidia(struct g_raid_md_object *md, struct g_class *mp,
800 struct g_raid_softc *sc;
801 struct g_raid_md_nvidia_object *mdi;
804 mdi = (struct g_raid_md_nvidia_object *)md;
805 arc4rand(&mdi->mdio_volume_id, 16, 0);
806 snprintf(name, sizeof(name), "NVIDIA-%d",
807 atomic_fetchadd_int(&NVIDIANodeID, 1));
808 sc = g_raid_create_node(mp, name, md);
810 return (G_RAID_MD_TASTE_FAIL);
813 return (G_RAID_MD_TASTE_NEW);
817 g_raid_md_taste_nvidia(struct g_raid_md_object *md, struct g_class *mp,
818 struct g_consumer *cp, struct g_geom **gp)
820 struct g_consumer *rcp;
821 struct g_provider *pp;
822 struct g_raid_md_nvidia_object *mdi, *mdi1;
823 struct g_raid_softc *sc;
824 struct g_raid_disk *disk;
825 struct nvidia_raid_conf *meta;
826 struct g_raid_md_nvidia_perdisk *pd;
828 int error, result, spare, len;
832 G_RAID_DEBUG(1, "Tasting NVIDIA on %s", cp->provider->name);
833 mdi = (struct g_raid_md_nvidia_object *)md;
836 /* Read metadata from device. */
839 if (g_access(cp, 1, 0, 0) != 0)
840 return (G_RAID_MD_TASTE_FAIL);
843 if (pp->geom->rank == 1)
844 g_io_getattr("GEOM::hba_vendor", cp, &len, &vendor);
845 meta = nvidia_meta_read(cp);
847 g_access(cp, -1, 0, 0);
849 if (g_raid_aggressive_spare) {
850 if (vendor == 0x10de) {
852 "No NVIDIA metadata, forcing spare.");
857 "NVIDIA vendor mismatch 0x%04x != 0x10de",
861 return (G_RAID_MD_TASTE_FAIL);
864 /* Metadata valid. Print it. */
865 g_raid_md_nvidia_print(meta);
866 G_RAID_DEBUG(1, "NVIDIA disk position %d", meta->disk_number);
867 spare = 0;//(meta->type == NVIDIA_T_SPARE) ? 1 : 0;
870 /* Search for matching node. */
873 LIST_FOREACH(geom, &mp->geom, geom) {
877 if (sc->sc_stopping != 0)
879 if (sc->sc_md->mdo_class != md->mdo_class)
881 mdi1 = (struct g_raid_md_nvidia_object *)sc->sc_md;
883 if (mdi1->mdio_incomplete)
886 if (memcmp(&mdi1->mdio_volume_id,
887 &meta->volume_id, 16) == 0)
892 /* Found matching node. */
894 G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name);
895 result = G_RAID_MD_TASTE_EXISTING;
897 } else if (spare) { /* Not found needy node -- left for later. */
898 G_RAID_DEBUG(1, "Spare is not needed at this time");
901 } else { /* Not found matching node -- create one. */
902 result = G_RAID_MD_TASTE_NEW;
903 memcpy(&mdi->mdio_volume_id, &meta->volume_id, 16);
904 snprintf(name, sizeof(name), "NVIDIA-%d",
905 atomic_fetchadd_int(&NVIDIANodeID, 1));
906 sc = g_raid_create_node(mp, name, md);
909 callout_init(&mdi->mdio_start_co, 1);
910 callout_reset(&mdi->mdio_start_co, g_raid_start_timeout * hz,
911 g_raid_nvidia_go, sc);
912 mdi->mdio_rootmount = root_mount_hold("GRAID-NVIDIA");
913 G_RAID_DEBUG1(1, sc, "root_mount_hold %p", mdi->mdio_rootmount);
916 rcp = g_new_consumer(geom);
918 if (g_access(rcp, 1, 1, 1) != 0)
922 sx_xlock(&sc->sc_lock);
924 pd = malloc(sizeof(*pd), M_MD_NVIDIA, M_WAITOK | M_ZERO);
927 pd->pd_disk_pos = -3;
929 pd->pd_disk_pos = -1;
931 pd->pd_disk_size = pp->mediasize;
932 disk = g_raid_create_disk(sc);
933 disk->d_md_data = (void *)pd;
934 disk->d_consumer = rcp;
937 /* Read kernel dumping information. */
938 disk->d_kd.offset = 0;
939 disk->d_kd.length = OFF_MAX;
940 len = sizeof(disk->d_kd);
941 error = g_io_getattr("GEOM::kerneldump", rcp, &len, &disk->d_kd);
942 if (disk->d_kd.di.dumper == NULL)
943 G_RAID_DEBUG1(2, sc, "Dumping not supported by %s: %d.",
944 rcp->provider->name, error);
946 g_raid_md_nvidia_new_disk(disk);
948 sx_xunlock(&sc->sc_lock);
953 free(meta, M_MD_NVIDIA);
954 return (G_RAID_MD_TASTE_FAIL);
958 g_raid_md_event_nvidia(struct g_raid_md_object *md,
959 struct g_raid_disk *disk, u_int event)
961 struct g_raid_softc *sc;
962 struct g_raid_subdisk *sd;
963 struct g_raid_md_nvidia_object *mdi;
964 struct g_raid_md_nvidia_perdisk *pd;
967 mdi = (struct g_raid_md_nvidia_object *)md;
970 case G_RAID_NODE_E_START:
971 if (!mdi->mdio_started) {
972 /* Bump volume ID to drop missing disks. */
973 arc4rand(&mdi->mdio_volume_id, 16, 0);
974 g_raid_md_nvidia_start(sc);
980 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
982 case G_RAID_DISK_E_DISCONNECTED:
983 /* If disk was assigned, just update statuses. */
984 if (pd->pd_disk_pos >= 0) {
985 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
986 if (disk->d_consumer) {
987 g_raid_kill_consumer(sc, disk->d_consumer);
988 disk->d_consumer = NULL;
990 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
991 g_raid_change_subdisk_state(sd,
992 G_RAID_SUBDISK_S_NONE);
993 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
994 G_RAID_EVENT_SUBDISK);
997 /* Otherwise -- delete. */
998 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
999 g_raid_destroy_disk(disk);
1002 if (mdi->mdio_started) {
1003 /* Bump volume ID to prevent disk resurrection. */
1004 if (pd->pd_disk_pos >= 0)
1005 arc4rand(&mdi->mdio_volume_id, 16, 0);
1007 /* Write updated metadata to all disks. */
1008 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
1011 /* Check if anything left except placeholders. */
1012 if (g_raid_ndisks(sc, -1) ==
1013 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1014 g_raid_destroy_node(sc, 0);
1016 g_raid_md_nvidia_refill(sc);
1023 g_raid_md_ctl_nvidia(struct g_raid_md_object *md,
1024 struct gctl_req *req)
1026 struct g_raid_softc *sc;
1027 struct g_raid_volume *vol;
1028 struct g_raid_subdisk *sd;
1029 struct g_raid_disk *disk;
1030 struct g_raid_md_nvidia_object *mdi;
1031 struct g_raid_md_nvidia_perdisk *pd;
1032 struct g_consumer *cp;
1033 struct g_provider *pp;
1035 const char *verb, *volname, *levelname, *diskname;
1037 off_t size, sectorsize, strip, volsize;
1038 intmax_t *sizearg, *striparg;
1039 int numdisks, i, len, level, qual, update;
1043 mdi = (struct g_raid_md_nvidia_object *)md;
1044 verb = gctl_get_param(req, "verb", NULL);
1045 nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
1047 if (strcmp(verb, "label") == 0) {
1050 gctl_error(req, "Invalid number of arguments.");
1053 volname = gctl_get_asciiparam(req, "arg1");
1054 if (volname == NULL) {
1055 gctl_error(req, "No volume name.");
1058 levelname = gctl_get_asciiparam(req, "arg2");
1059 if (levelname == NULL) {
1060 gctl_error(req, "No RAID level.");
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 /* Read kernel dumping information. */
1114 disk->d_kd.offset = 0;
1115 disk->d_kd.length = OFF_MAX;
1116 len = sizeof(disk->d_kd);
1117 g_io_getattr("GEOM::kerneldump", cp, &len, &disk->d_kd);
1118 if (disk->d_kd.di.dumper == NULL)
1119 G_RAID_DEBUG1(2, sc,
1120 "Dumping not supported by %s.",
1121 cp->provider->name);
1123 pd->pd_disk_size = pp->mediasize;
1124 if (size > pp->mediasize)
1125 size = pp->mediasize;
1126 if (sectorsize < pp->sectorsize)
1127 sectorsize = pp->sectorsize;
1132 if (sectorsize <= 0) {
1133 gctl_error(req, "Can't get sector size.");
1137 /* Reserve space for metadata. */
1138 size -= 2 * sectorsize;
1140 /* Handle size argument. */
1141 len = sizeof(*sizearg);
1142 sizearg = gctl_get_param(req, "size", &len);
1143 if (sizearg != NULL && len == sizeof(*sizearg) &&
1145 if (*sizearg > size) {
1146 gctl_error(req, "Size too big %lld > %lld.",
1147 (long long)*sizearg, (long long)size);
1153 /* Handle strip argument. */
1155 len = sizeof(*striparg);
1156 striparg = gctl_get_param(req, "strip", &len);
1157 if (striparg != NULL && len == sizeof(*striparg) &&
1159 if (*striparg < sectorsize) {
1160 gctl_error(req, "Strip size too small.");
1163 if (*striparg % sectorsize != 0) {
1164 gctl_error(req, "Incorrect strip size.");
1167 if (strip > 65535 * sectorsize) {
1168 gctl_error(req, "Strip size too big.");
1174 /* Round size down to strip or sector. */
1175 if (level == G_RAID_VOLUME_RL_RAID1)
1176 size -= (size % sectorsize);
1177 else if (level == G_RAID_VOLUME_RL_RAID1E &&
1178 (numdisks & 1) != 0)
1179 size -= (size % (2 * strip));
1181 size -= (size % strip);
1183 gctl_error(req, "Size too small.");
1187 if (level == G_RAID_VOLUME_RL_RAID0 ||
1188 level == G_RAID_VOLUME_RL_CONCAT ||
1189 level == G_RAID_VOLUME_RL_SINGLE)
1190 volsize = size * numdisks;
1191 else if (level == G_RAID_VOLUME_RL_RAID1)
1193 else if (level == G_RAID_VOLUME_RL_RAID5)
1194 volsize = size * (numdisks - 1);
1196 volsize = ((size * numdisks) / strip / 2) *
1199 if (volsize > 0xffffffffllu * sectorsize) {
1200 gctl_error(req, "Size too big.");
1204 /* We have all we need, create things: volume, ... */
1205 mdi->mdio_total_disks = numdisks;
1206 mdi->mdio_started = 1;
1207 vol = g_raid_create_volume(sc, volname, -1);
1208 vol->v_md_data = (void *)(intptr_t)0;
1209 vol->v_raid_level = level;
1210 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE;
1211 vol->v_strip_size = strip;
1212 vol->v_disks_count = numdisks;
1213 vol->v_mediasize = volsize;
1214 vol->v_sectorsize = sectorsize;
1215 g_raid_start_volume(vol);
1217 /* , and subdisks. */
1218 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1219 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
1220 sd = &vol->v_subdisks[pd->pd_disk_pos];
1224 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1225 if (sd->sd_disk->d_consumer != NULL) {
1226 g_raid_change_disk_state(disk,
1227 G_RAID_DISK_S_ACTIVE);
1228 g_raid_change_subdisk_state(sd,
1229 G_RAID_SUBDISK_S_ACTIVE);
1230 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
1231 G_RAID_EVENT_SUBDISK);
1233 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1237 /* Write metadata based on created entities. */
1238 G_RAID_DEBUG1(0, sc, "Array started.");
1239 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
1241 /* Pickup any STALE/SPARE disks to refill array if needed. */
1242 g_raid_md_nvidia_refill(sc);
1244 g_raid_event_send(vol, G_RAID_VOLUME_E_START,
1245 G_RAID_EVENT_VOLUME);
1248 if (strcmp(verb, "delete") == 0) {
1250 /* Check if some volume is still open. */
1251 force = gctl_get_paraml(req, "force", sizeof(*force));
1252 if (force != NULL && *force == 0 &&
1253 g_raid_nopens(sc) != 0) {
1254 gctl_error(req, "Some volume is still open.");
1258 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1259 if (disk->d_consumer)
1260 nvidia_meta_erase(disk->d_consumer);
1262 g_raid_destroy_node(sc, 0);
1265 if (strcmp(verb, "remove") == 0 ||
1266 strcmp(verb, "fail") == 0) {
1268 gctl_error(req, "Invalid number of arguments.");
1271 for (i = 1; i < *nargs; i++) {
1272 snprintf(arg, sizeof(arg), "arg%d", i);
1273 diskname = gctl_get_asciiparam(req, arg);
1274 if (diskname == NULL) {
1275 gctl_error(req, "No disk name (%s).", arg);
1279 if (strncmp(diskname, "/dev/", 5) == 0)
1282 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1283 if (disk->d_consumer != NULL &&
1284 disk->d_consumer->provider != NULL &&
1285 strcmp(disk->d_consumer->provider->name,
1290 gctl_error(req, "Disk '%s' not found.",
1296 if (strcmp(verb, "fail") == 0) {
1297 g_raid_md_fail_disk_nvidia(md, NULL, disk);
1301 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
1303 /* Erase metadata on deleting disk. */
1304 nvidia_meta_erase(disk->d_consumer);
1306 /* If disk was assigned, just update statuses. */
1307 if (pd->pd_disk_pos >= 0) {
1308 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1309 g_raid_kill_consumer(sc, disk->d_consumer);
1310 disk->d_consumer = NULL;
1311 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
1312 g_raid_change_subdisk_state(sd,
1313 G_RAID_SUBDISK_S_NONE);
1314 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
1315 G_RAID_EVENT_SUBDISK);
1318 /* Otherwise -- delete. */
1319 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
1320 g_raid_destroy_disk(disk);
1324 /* Write updated metadata to remaining disks. */
1325 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
1327 /* Check if anything left except placeholders. */
1328 if (g_raid_ndisks(sc, -1) ==
1329 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1330 g_raid_destroy_node(sc, 0);
1332 g_raid_md_nvidia_refill(sc);
1335 if (strcmp(verb, "insert") == 0) {
1337 gctl_error(req, "Invalid number of arguments.");
1341 for (i = 1; i < *nargs; i++) {
1342 /* Get disk name. */
1343 snprintf(arg, sizeof(arg), "arg%d", i);
1344 diskname = gctl_get_asciiparam(req, arg);
1345 if (diskname == NULL) {
1346 gctl_error(req, "No disk name (%s).", arg);
1351 /* Try to find provider with specified name. */
1353 cp = g_raid_open_consumer(sc, diskname);
1355 gctl_error(req, "Can't open disk '%s'.",
1357 g_topology_unlock();
1363 pd = malloc(sizeof(*pd), M_MD_NVIDIA, M_WAITOK | M_ZERO);
1364 pd->pd_disk_pos = -3;
1365 pd->pd_disk_size = pp->mediasize;
1367 disk = g_raid_create_disk(sc);
1368 disk->d_consumer = cp;
1369 disk->d_md_data = (void *)pd;
1371 g_topology_unlock();
1373 /* Read kernel dumping information. */
1374 disk->d_kd.offset = 0;
1375 disk->d_kd.length = OFF_MAX;
1376 len = sizeof(disk->d_kd);
1377 g_io_getattr("GEOM::kerneldump", cp, &len, &disk->d_kd);
1378 if (disk->d_kd.di.dumper == NULL)
1379 G_RAID_DEBUG1(2, sc,
1380 "Dumping not supported by %s.",
1381 cp->provider->name);
1383 /* Welcome the "new" disk. */
1384 update += g_raid_md_nvidia_start_disk(disk);
1385 if (disk->d_state != G_RAID_DISK_S_SPARE &&
1386 disk->d_state != G_RAID_DISK_S_ACTIVE) {
1387 gctl_error(req, "Disk '%s' doesn't fit.",
1389 g_raid_destroy_disk(disk);
1395 /* Write new metadata if we changed something. */
1397 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
1400 gctl_error(req, "Command '%s' is not supported.", verb);
1405 g_raid_md_write_nvidia(struct g_raid_md_object *md, struct g_raid_volume *tvol,
1406 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
1408 struct g_raid_softc *sc;
1409 struct g_raid_volume *vol;
1410 struct g_raid_subdisk *sd;
1411 struct g_raid_disk *disk;
1412 struct g_raid_md_nvidia_object *mdi;
1413 struct g_raid_md_nvidia_perdisk *pd;
1414 struct nvidia_raid_conf *meta;
1418 mdi = (struct g_raid_md_nvidia_object *)md;
1420 if (sc->sc_stopping == G_RAID_DESTROY_HARD)
1423 /* There is only one volume. */
1424 vol = TAILQ_FIRST(&sc->sc_volumes);
1426 /* Fill global fields. */
1427 meta = malloc(sizeof(*meta), M_MD_NVIDIA, M_WAITOK | M_ZERO);
1429 memcpy(meta, mdi->mdio_meta, sizeof(*meta));
1430 memcpy(meta->nvidia_id, NVIDIA_MAGIC, sizeof(NVIDIA_MAGIC) - 1);
1431 meta->config_size = 30;
1432 meta->version = 0x0064;
1433 meta->total_sectors = vol->v_mediasize / vol->v_sectorsize;
1434 meta->sector_size = vol->v_sectorsize;
1435 nvidia_meta_put_name(meta, vol->v_name);
1436 meta->magic_0 = NVIDIA_MAGIC0;
1437 memcpy(&meta->volume_id, &mdi->mdio_volume_id, 16);
1438 meta->state = NVIDIA_S_IDLE;
1439 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
1440 meta->array_width = 1;
1441 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E)
1442 meta->array_width = vol->v_disks_count / 2;
1443 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID5)
1444 meta->array_width = vol->v_disks_count - 1;
1446 meta->array_width = vol->v_disks_count;
1447 meta->total_disks = vol->v_disks_count;
1448 meta->orig_array_width = meta->array_width;
1449 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0)
1450 meta->type = NVIDIA_T_RAID0;
1451 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
1452 meta->type = NVIDIA_T_RAID1;
1453 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E)
1454 meta->type = NVIDIA_T_RAID01;
1455 else if (vol->v_raid_level == G_RAID_VOLUME_RL_CONCAT ||
1456 vol->v_raid_level == G_RAID_VOLUME_RL_SINGLE)
1457 meta->type = NVIDIA_T_CONCAT;
1458 // else if (vol->v_raid_level_qualifier == 0)
1459 // meta->type = NVIDIA_T_RAID5;
1461 meta->type = NVIDIA_T_RAID5_SYM;
1462 meta->strip_sectors = vol->v_strip_size / vol->v_sectorsize;
1463 meta->strip_bytes = vol->v_strip_size;
1464 meta->strip_shift = ffs(meta->strip_sectors) - 1;
1465 meta->strip_mask = meta->strip_sectors - 1;
1466 meta->stripe_sectors = meta->strip_sectors * meta->orig_array_width;
1467 meta->stripe_bytes = meta->stripe_sectors * vol->v_sectorsize;
1468 meta->rebuild_lba = 0;
1469 meta->orig_type = meta->type;
1470 meta->orig_total_sectors = meta->total_sectors;
1473 for (i = 0; i < vol->v_disks_count; i++) {
1474 sd = &vol->v_subdisks[i];
1475 if ((sd->sd_state == G_RAID_SUBDISK_S_STALE ||
1476 sd->sd_state == G_RAID_SUBDISK_S_RESYNC ||
1478 meta->state != NVIDIA_S_REBUILD)
1479 meta->state = NVIDIA_S_SYNC;
1480 else if (sd->sd_state == G_RAID_SUBDISK_S_NEW ||
1481 sd->sd_state == G_RAID_SUBDISK_S_REBUILD)
1482 meta->state = NVIDIA_S_REBUILD;
1485 /* We are done. Print meta data and store them to disks. */
1486 if (mdi->mdio_meta != NULL)
1487 free(mdi->mdio_meta, M_MD_NVIDIA);
1488 mdi->mdio_meta = meta;
1490 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1491 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
1492 if (disk->d_state != G_RAID_DISK_S_ACTIVE &&
1493 disk->d_state != G_RAID_DISK_S_SPARE)
1495 if (pd->pd_meta != NULL) {
1496 free(pd->pd_meta, M_MD_NVIDIA);
1499 pd->pd_meta = nvidia_meta_copy(meta);
1500 if ((sd = TAILQ_FIRST(&disk->d_subdisks)) != NULL) {
1501 /* For RAID0+1 we need to translate order. */
1502 pd->pd_meta->disk_number =
1503 nvidia_meta_translate_disk(meta, sd->sd_pos);
1504 if (sd->sd_state != G_RAID_SUBDISK_S_ACTIVE) {
1505 pd->pd_meta->disk_status = 0x100;
1506 pd->pd_meta->rebuild_lba =
1507 sd->sd_rebuild_pos / vol->v_sectorsize *
1511 pd->pd_meta->disk_number = meta->total_disks + spares++;
1512 G_RAID_DEBUG(1, "Writing NVIDIA metadata to %s",
1513 g_raid_get_diskname(disk));
1514 g_raid_md_nvidia_print(pd->pd_meta);
1515 nvidia_meta_write(disk->d_consumer, pd->pd_meta);
1521 g_raid_md_fail_disk_nvidia(struct g_raid_md_object *md,
1522 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
1524 struct g_raid_softc *sc;
1525 struct g_raid_md_nvidia_perdisk *pd;
1526 struct g_raid_subdisk *sd;
1529 pd = (struct g_raid_md_nvidia_perdisk *)tdisk->d_md_data;
1531 /* We can't fail disk that is not a part of array now. */
1532 if (pd->pd_disk_pos < 0)
1535 /* Erase metadata to prevent disks's later resurrection. */
1536 if (tdisk->d_consumer)
1537 nvidia_meta_erase(tdisk->d_consumer);
1539 /* Change states. */
1540 g_raid_change_disk_state(tdisk, G_RAID_DISK_S_FAILED);
1541 TAILQ_FOREACH(sd, &tdisk->d_subdisks, sd_next) {
1542 g_raid_change_subdisk_state(sd,
1543 G_RAID_SUBDISK_S_FAILED);
1544 g_raid_event_send(sd, G_RAID_SUBDISK_E_FAILED,
1545 G_RAID_EVENT_SUBDISK);
1548 /* Write updated metadata to remaining disks. */
1549 g_raid_md_write_nvidia(md, NULL, NULL, tdisk);
1551 /* Check if anything left except placeholders. */
1552 if (g_raid_ndisks(sc, -1) ==
1553 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1554 g_raid_destroy_node(sc, 0);
1556 g_raid_md_nvidia_refill(sc);
1561 g_raid_md_free_disk_nvidia(struct g_raid_md_object *md,
1562 struct g_raid_disk *disk)
1564 struct g_raid_md_nvidia_perdisk *pd;
1566 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
1567 if (pd->pd_meta != NULL) {
1568 free(pd->pd_meta, M_MD_NVIDIA);
1571 free(pd, M_MD_NVIDIA);
1572 disk->d_md_data = NULL;
1577 g_raid_md_free_nvidia(struct g_raid_md_object *md)
1579 struct g_raid_md_nvidia_object *mdi;
1581 mdi = (struct g_raid_md_nvidia_object *)md;
1582 if (!mdi->mdio_started) {
1583 mdi->mdio_started = 0;
1584 callout_stop(&mdi->mdio_start_co);
1585 G_RAID_DEBUG1(1, md->mdo_softc,
1586 "root_mount_rel %p", mdi->mdio_rootmount);
1587 root_mount_rel(mdi->mdio_rootmount);
1588 mdi->mdio_rootmount = NULL;
1590 if (mdi->mdio_meta != NULL) {
1591 free(mdi->mdio_meta, M_MD_NVIDIA);
1592 mdi->mdio_meta = NULL;
1597 G_RAID_MD_DECLARE(g_raid_md_nvidia);