2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2011 Alexander Motin <mav@FreeBSD.org>
5 * Copyright (c) 2000 - 2008 Søren Schmidt <sos@FreeBSD.org>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
33 #include <sys/param.h>
35 #include <sys/endian.h>
36 #include <sys/kernel.h>
38 #include <sys/limits.h>
40 #include <sys/malloc.h>
41 #include <sys/mutex.h>
42 #include <sys/systm.h>
43 #include <sys/taskqueue.h>
44 #include <geom/geom.h>
45 #include <geom/geom_dbg.h>
46 #include "geom/raid/g_raid.h"
47 #include "g_raid_md_if.h"
49 static MALLOC_DEFINE(M_MD_NVIDIA, "md_nvidia_data", "GEOM_RAID NVIDIA metadata");
51 struct nvidia_raid_conf {
53 #define NVIDIA_MAGIC "NVIDIA "
60 uint32_t total_sectors;
67 #define NVIDIA_MAGIC0 0x00640044
69 uint64_t volume_id[2];
71 #define NVIDIA_S_IDLE 0
72 #define NVIDIA_S_INIT 2
73 #define NVIDIA_S_REBUILD 3
74 #define NVIDIA_S_UPGRADE 4
75 #define NVIDIA_S_SYNC 5
78 uint8_t orig_array_width;
80 #define NVIDIA_T_RAID0 0x0080
81 #define NVIDIA_T_RAID1 0x0081
82 #define NVIDIA_T_RAID3 0x0083
83 #define NVIDIA_T_RAID5 0x0085 /* RLQ = 00/02? */
84 #define NVIDIA_T_RAID5_SYM 0x0095 /* RLQ = 03 */
85 #define NVIDIA_T_RAID10 0x008a
86 #define NVIDIA_T_RAID01 0x8180
87 #define NVIDIA_T_CONCAT 0x00ff
90 uint32_t strip_sectors;
94 uint32_t stripe_sectors;
95 uint32_t stripe_bytes;
98 uint32_t orig_total_sectors;
100 #define NVIDIA_S_BOOTABLE 0x00000001
101 #define NVIDIA_S_DEGRADED 0x00000002
106 struct g_raid_md_nvidia_perdisk {
107 struct nvidia_raid_conf *pd_meta;
112 struct g_raid_md_nvidia_object {
113 struct g_raid_md_object mdio_base;
114 uint64_t mdio_volume_id[2];
115 struct nvidia_raid_conf *mdio_meta;
116 struct callout mdio_start_co; /* STARTING state timer. */
117 int mdio_total_disks;
118 int mdio_disks_present;
121 struct root_hold_token *mdio_rootmount; /* Root mount delay token. */
124 static g_raid_md_create_t g_raid_md_create_nvidia;
125 static g_raid_md_taste_t g_raid_md_taste_nvidia;
126 static g_raid_md_event_t g_raid_md_event_nvidia;
127 static g_raid_md_ctl_t g_raid_md_ctl_nvidia;
128 static g_raid_md_write_t g_raid_md_write_nvidia;
129 static g_raid_md_fail_disk_t g_raid_md_fail_disk_nvidia;
130 static g_raid_md_free_disk_t g_raid_md_free_disk_nvidia;
131 static g_raid_md_free_t g_raid_md_free_nvidia;
133 static kobj_method_t g_raid_md_nvidia_methods[] = {
134 KOBJMETHOD(g_raid_md_create, g_raid_md_create_nvidia),
135 KOBJMETHOD(g_raid_md_taste, g_raid_md_taste_nvidia),
136 KOBJMETHOD(g_raid_md_event, g_raid_md_event_nvidia),
137 KOBJMETHOD(g_raid_md_ctl, g_raid_md_ctl_nvidia),
138 KOBJMETHOD(g_raid_md_write, g_raid_md_write_nvidia),
139 KOBJMETHOD(g_raid_md_fail_disk, g_raid_md_fail_disk_nvidia),
140 KOBJMETHOD(g_raid_md_free_disk, g_raid_md_free_disk_nvidia),
141 KOBJMETHOD(g_raid_md_free, g_raid_md_free_nvidia),
145 static struct g_raid_md_class g_raid_md_nvidia_class = {
147 g_raid_md_nvidia_methods,
148 sizeof(struct g_raid_md_nvidia_object),
153 static int NVIDIANodeID = 1;
156 g_raid_md_nvidia_print(struct nvidia_raid_conf *meta)
159 if (g_raid_debug < 1)
162 printf("********* ATA NVIDIA RAID Metadata *********\n");
163 printf("nvidia_id <%.8s>\n", meta->nvidia_id);
164 printf("config_size %u\n", meta->config_size);
165 printf("checksum 0x%08x\n", meta->checksum);
166 printf("version 0x%04x\n", meta->version);
167 printf("disk_number %d\n", meta->disk_number);
168 printf("dummy_0 0x%02x\n", meta->dummy_0);
169 printf("total_sectors %u\n", meta->total_sectors);
170 printf("sector_size %u\n", meta->sector_size);
171 printf("name <%.16s>\n", meta->name);
172 printf("revision 0x%02x%02x%02x%02x\n",
173 meta->revision[0], meta->revision[1],
174 meta->revision[2], meta->revision[3]);
175 printf("disk_status 0x%08x\n", meta->disk_status);
176 printf("magic_0 0x%08x\n", meta->magic_0);
177 printf("volume_id 0x%016jx%016jx\n",
178 meta->volume_id[1], meta->volume_id[0]);
179 printf("state 0x%02x\n", meta->state);
180 printf("array_width %u\n", meta->array_width);
181 printf("total_disks %u\n", meta->total_disks);
182 printf("orig_array_width %u\n", meta->orig_array_width);
183 printf("type 0x%04x\n", meta->type);
184 printf("dummy_3 0x%04x\n", meta->dummy_3);
185 printf("strip_sectors %u\n", meta->strip_sectors);
186 printf("strip_bytes %u\n", meta->strip_bytes);
187 printf("strip_shift %u\n", meta->strip_shift);
188 printf("strip_mask 0x%08x\n", meta->strip_mask);
189 printf("stripe_sectors %u\n", meta->stripe_sectors);
190 printf("stripe_bytes %u\n", meta->stripe_bytes);
191 printf("rebuild_lba %u\n", meta->rebuild_lba);
192 printf("orig_type 0x%04x\n", meta->orig_type);
193 printf("orig_total_sectors %u\n", meta->orig_total_sectors);
194 printf("status 0x%08x\n", meta->status);
195 printf("=================================================\n");
198 static struct nvidia_raid_conf *
199 nvidia_meta_copy(struct nvidia_raid_conf *meta)
201 struct nvidia_raid_conf *nmeta;
203 nmeta = malloc(sizeof(*meta), M_MD_NVIDIA, M_WAITOK);
204 memcpy(nmeta, meta, sizeof(*meta));
209 nvidia_meta_translate_disk(struct nvidia_raid_conf *meta, int md_disk_pos)
213 if (md_disk_pos >= 0 && meta->type == NVIDIA_T_RAID01) {
214 disk_pos = (md_disk_pos / meta->array_width) +
215 (md_disk_pos % meta->array_width) * meta->array_width;
217 disk_pos = md_disk_pos;
222 nvidia_meta_get_name(struct nvidia_raid_conf *meta, char *buf)
226 strncpy(buf, meta->name, 16);
228 for (i = 15; i >= 0; i--) {
236 nvidia_meta_put_name(struct nvidia_raid_conf *meta, char *buf)
239 memset(meta->name, 0x20, 16);
240 memcpy(meta->name, buf, MIN(strlen(buf), 16));
243 static struct nvidia_raid_conf *
244 nvidia_meta_read(struct g_consumer *cp)
246 struct g_provider *pp;
247 struct nvidia_raid_conf *meta;
250 uint32_t checksum, *ptr;
254 /* Read the anchor sector. */
255 buf = g_read_data(cp,
256 pp->mediasize - 2 * pp->sectorsize, pp->sectorsize, &error);
258 G_RAID_DEBUG(1, "Cannot read metadata from %s (error=%d).",
262 meta = (struct nvidia_raid_conf *)buf;
264 /* Check if this is an NVIDIA RAID struct */
265 if (strncmp(meta->nvidia_id, NVIDIA_MAGIC, strlen(NVIDIA_MAGIC))) {
266 G_RAID_DEBUG(1, "NVIDIA signature check failed on %s", pp->name);
270 if (meta->config_size > 128 ||
271 meta->config_size < 30) {
272 G_RAID_DEBUG(1, "NVIDIA metadata size looks wrong: %d",
277 meta = malloc(sizeof(*meta), M_MD_NVIDIA, M_WAITOK);
278 memcpy(meta, buf, min(sizeof(*meta), pp->sectorsize));
281 /* Check metadata checksum. */
282 for (checksum = 0, ptr = (uint32_t *)meta,
283 i = 0; i < meta->config_size; i++)
286 G_RAID_DEBUG(1, "NVIDIA checksum check failed on %s", pp->name);
287 free(meta, M_MD_NVIDIA);
291 /* Check volume state. */
292 if (meta->state != NVIDIA_S_IDLE && meta->state != NVIDIA_S_INIT &&
293 meta->state != NVIDIA_S_REBUILD && meta->state != NVIDIA_S_SYNC) {
294 G_RAID_DEBUG(1, "NVIDIA unknown state on %s (0x%02x)",
295 pp->name, meta->state);
296 free(meta, M_MD_NVIDIA);
300 /* Check raid type. */
301 if (meta->type != NVIDIA_T_RAID0 && meta->type != NVIDIA_T_RAID1 &&
302 meta->type != NVIDIA_T_RAID3 && meta->type != NVIDIA_T_RAID5 &&
303 meta->type != NVIDIA_T_RAID5_SYM &&
304 meta->type != NVIDIA_T_RAID01 && meta->type != NVIDIA_T_CONCAT) {
305 G_RAID_DEBUG(1, "NVIDIA unknown RAID level on %s (0x%02x)",
306 pp->name, meta->type);
307 free(meta, M_MD_NVIDIA);
315 nvidia_meta_write(struct g_consumer *cp, struct nvidia_raid_conf *meta)
317 struct g_provider *pp;
320 uint32_t checksum, *ptr;
324 /* Recalculate checksum for case if metadata were changed. */
326 for (checksum = 0, ptr = (uint32_t *)meta,
327 i = 0; i < meta->config_size; i++)
329 meta->checksum -= checksum;
331 /* Create and fill buffer. */
332 buf = malloc(pp->sectorsize, M_MD_NVIDIA, M_WAITOK | M_ZERO);
333 memcpy(buf, meta, sizeof(*meta));
335 /* Write metadata. */
336 error = g_write_data(cp,
337 pp->mediasize - 2 * pp->sectorsize, buf, pp->sectorsize);
339 G_RAID_DEBUG(1, "Cannot write metadata to %s (error=%d).",
343 free(buf, M_MD_NVIDIA);
348 nvidia_meta_erase(struct g_consumer *cp)
350 struct g_provider *pp;
355 buf = malloc(pp->sectorsize, M_MD_NVIDIA, M_WAITOK | M_ZERO);
356 error = g_write_data(cp,
357 pp->mediasize - 2 * pp->sectorsize, buf, pp->sectorsize);
359 G_RAID_DEBUG(1, "Cannot erase metadata on %s (error=%d).",
362 free(buf, M_MD_NVIDIA);
366 static struct g_raid_disk *
367 g_raid_md_nvidia_get_disk(struct g_raid_softc *sc, int id)
369 struct g_raid_disk *disk;
370 struct g_raid_md_nvidia_perdisk *pd;
372 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
373 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
374 if (pd->pd_disk_pos == id)
381 g_raid_md_nvidia_supported(int level, int qual, int disks, int force)
385 case G_RAID_VOLUME_RL_RAID0:
388 if (!force && (disks < 2 || disks > 6))
391 case G_RAID_VOLUME_RL_RAID1:
394 if (!force && (disks != 2))
397 case G_RAID_VOLUME_RL_RAID1E:
402 if (!force && (disks < 4))
405 case G_RAID_VOLUME_RL_SINGLE:
409 case G_RAID_VOLUME_RL_CONCAT:
413 case G_RAID_VOLUME_RL_RAID5:
416 if (qual != G_RAID_VOLUME_RLQ_R5LA &&
417 qual != G_RAID_VOLUME_RLQ_R5LS)
423 if (level != G_RAID_VOLUME_RL_RAID5 && qual != G_RAID_VOLUME_RLQ_NONE)
429 g_raid_md_nvidia_start_disk(struct g_raid_disk *disk)
431 struct g_raid_softc *sc;
432 struct g_raid_subdisk *sd, *tmpsd;
433 struct g_raid_disk *olddisk, *tmpdisk;
434 struct g_raid_md_object *md;
435 struct g_raid_md_nvidia_object *mdi;
436 struct g_raid_md_nvidia_perdisk *pd, *oldpd;
437 struct nvidia_raid_conf *meta;
438 int disk_pos, resurrection = 0;
442 mdi = (struct g_raid_md_nvidia_object *)md;
443 meta = mdi->mdio_meta;
444 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
447 /* Find disk position in metadata by its serial. */
448 if (pd->pd_meta != NULL) {
449 disk_pos = pd->pd_meta->disk_number;
450 if (disk_pos >= meta->total_disks || mdi->mdio_started)
454 /* For RAID0+1 we need to translate order. */
455 disk_pos = nvidia_meta_translate_disk(meta, disk_pos);
457 G_RAID_DEBUG1(1, sc, "Unknown, probably new or stale disk");
458 /* If we are in the start process, that's all for now. */
459 if (!mdi->mdio_started)
462 * If we have already started - try to get use of the disk.
463 * Try to replace OFFLINE disks first, then FAILED.
465 TAILQ_FOREACH(tmpdisk, &sc->sc_disks, d_next) {
466 if (tmpdisk->d_state != G_RAID_DISK_S_OFFLINE &&
467 tmpdisk->d_state != G_RAID_DISK_S_FAILED)
469 /* Make sure this disk is big enough. */
470 TAILQ_FOREACH(sd, &tmpdisk->d_subdisks, sd_next) {
471 if (sd->sd_offset + sd->sd_size + 2 * 512 >
474 "Disk too small (%ju < %ju)",
476 sd->sd_offset + sd->sd_size + 512);
482 if (tmpdisk->d_state == G_RAID_DISK_S_OFFLINE) {
485 } else if (olddisk == NULL)
488 if (olddisk == NULL) {
490 g_raid_change_disk_state(disk, G_RAID_DISK_S_SPARE);
493 oldpd = (struct g_raid_md_nvidia_perdisk *)olddisk->d_md_data;
494 disk_pos = oldpd->pd_disk_pos;
498 if (olddisk == NULL) {
499 /* Find placeholder by position. */
500 olddisk = g_raid_md_nvidia_get_disk(sc, disk_pos);
502 panic("No disk at position %d!", disk_pos);
503 if (olddisk->d_state != G_RAID_DISK_S_OFFLINE) {
504 G_RAID_DEBUG1(1, sc, "More than one disk for pos %d",
506 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE);
509 oldpd = (struct g_raid_md_nvidia_perdisk *)olddisk->d_md_data;
512 /* Replace failed disk or placeholder with new disk. */
513 TAILQ_FOREACH_SAFE(sd, &olddisk->d_subdisks, sd_next, tmpsd) {
514 TAILQ_REMOVE(&olddisk->d_subdisks, sd, sd_next);
515 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
518 oldpd->pd_disk_pos = -2;
519 pd->pd_disk_pos = disk_pos;
521 /* If it was placeholder -- destroy it. */
522 if (olddisk->d_state == G_RAID_DISK_S_OFFLINE) {
523 g_raid_destroy_disk(olddisk);
525 /* Otherwise, make it STALE_FAILED. */
526 g_raid_change_disk_state(olddisk, G_RAID_DISK_S_STALE_FAILED);
529 /* Welcome the new disk. */
531 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
532 else// if (pd->pd_meta->disk_status == NVIDIA_S_CURRENT ||
533 //pd->pd_meta->disk_status == NVIDIA_S_REBUILD)
534 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
536 // g_raid_change_disk_state(disk, G_RAID_DISK_S_FAILED);
537 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
540 * Different disks may have different sizes,
541 * in concat mode. Update from real disk size.
543 if (meta->type == NVIDIA_T_CONCAT)
544 sd->sd_size = pd->pd_disk_size - 0x800 * 512;
547 /* New or ex-spare disk. */
548 g_raid_change_subdisk_state(sd,
549 G_RAID_SUBDISK_S_NEW);
550 } else if (meta->state == NVIDIA_S_REBUILD &&
551 (pd->pd_meta->disk_status & 0x100)) {
552 /* Rebuilding disk. */
553 g_raid_change_subdisk_state(sd,
554 G_RAID_SUBDISK_S_REBUILD);
555 sd->sd_rebuild_pos = (off_t)pd->pd_meta->rebuild_lba /
556 meta->array_width * pd->pd_meta->sector_size;
557 } else if (meta->state == NVIDIA_S_SYNC) {
558 /* Resyncing/dirty disk. */
559 g_raid_change_subdisk_state(sd,
560 G_RAID_SUBDISK_S_RESYNC);
561 sd->sd_rebuild_pos = (off_t)pd->pd_meta->rebuild_lba /
562 meta->array_width * pd->pd_meta->sector_size;
564 /* Up to date disk. */
565 g_raid_change_subdisk_state(sd,
566 G_RAID_SUBDISK_S_ACTIVE);
568 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
569 G_RAID_EVENT_SUBDISK);
572 /* Update status of our need for spare. */
573 if (mdi->mdio_started) {
574 mdi->mdio_incomplete =
575 (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) <
576 mdi->mdio_total_disks);
579 return (resurrection);
583 g_disk_md_nvidia_retaste(void *arg, int pending)
586 G_RAID_DEBUG(1, "Array is not complete, trying to retaste.");
587 g_retaste(&g_raid_class);
588 free(arg, M_MD_NVIDIA);
592 g_raid_md_nvidia_refill(struct g_raid_softc *sc)
594 struct g_raid_md_object *md;
595 struct g_raid_md_nvidia_object *mdi;
596 struct g_raid_disk *disk;
601 mdi = (struct g_raid_md_nvidia_object *)md;
604 /* Make sure we miss anything. */
605 na = g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE);
606 if (na == mdi->mdio_total_disks)
609 G_RAID_DEBUG1(1, md->mdo_softc,
610 "Array is not complete (%d of %d), "
611 "trying to refill.", na, mdi->mdio_total_disks);
613 /* Try to get use some of STALE disks. */
614 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
615 if (disk->d_state == G_RAID_DISK_S_STALE) {
616 update += g_raid_md_nvidia_start_disk(disk);
617 if (disk->d_state == G_RAID_DISK_S_ACTIVE)
624 /* Try to get use some of SPARE disks. */
625 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
626 if (disk->d_state == G_RAID_DISK_S_SPARE) {
627 update += g_raid_md_nvidia_start_disk(disk);
628 if (disk->d_state == G_RAID_DISK_S_ACTIVE)
632 } while (disk != NULL);
634 /* Write new metadata if we changed something. */
636 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
638 /* Update status of our need for spare. */
639 mdi->mdio_incomplete = (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) <
640 mdi->mdio_total_disks);
642 /* Request retaste hoping to find spare. */
643 if (mdi->mdio_incomplete) {
644 task = malloc(sizeof(struct task),
645 M_MD_NVIDIA, M_WAITOK | M_ZERO);
646 TASK_INIT(task, 0, g_disk_md_nvidia_retaste, task);
647 taskqueue_enqueue(taskqueue_swi, task);
652 g_raid_md_nvidia_start(struct g_raid_softc *sc)
654 struct g_raid_md_object *md;
655 struct g_raid_md_nvidia_object *mdi;
656 struct g_raid_md_nvidia_perdisk *pd;
657 struct nvidia_raid_conf *meta;
658 struct g_raid_volume *vol;
659 struct g_raid_subdisk *sd;
660 struct g_raid_disk *disk;
666 mdi = (struct g_raid_md_nvidia_object *)md;
667 meta = mdi->mdio_meta;
669 /* Create volumes and subdisks. */
670 nvidia_meta_get_name(meta, buf);
671 vol = g_raid_create_volume(sc, buf, -1);
672 vol->v_mediasize = (off_t)meta->total_sectors * 512;
673 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE;
674 if (meta->type == NVIDIA_T_RAID0) {
675 vol->v_raid_level = G_RAID_VOLUME_RL_RAID0;
676 size = vol->v_mediasize / mdi->mdio_total_disks;
677 } else if (meta->type == NVIDIA_T_RAID1) {
678 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1;
679 size = vol->v_mediasize;
680 } else if (meta->type == NVIDIA_T_RAID01) {
681 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E;
682 size = vol->v_mediasize / (mdi->mdio_total_disks / 2);
683 } else if (meta->type == NVIDIA_T_CONCAT) {
684 if (mdi->mdio_total_disks == 1)
685 vol->v_raid_level = G_RAID_VOLUME_RL_SINGLE;
687 vol->v_raid_level = G_RAID_VOLUME_RL_CONCAT;
689 } else if (meta->type == NVIDIA_T_RAID5) {
690 vol->v_raid_level = G_RAID_VOLUME_RL_RAID5;
691 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_R5LA;
692 size = vol->v_mediasize / (mdi->mdio_total_disks - 1);
693 } else if (meta->type == NVIDIA_T_RAID5_SYM) {
694 vol->v_raid_level = G_RAID_VOLUME_RL_RAID5;
695 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_R5LS;
696 size = vol->v_mediasize / (mdi->mdio_total_disks - 1);
698 vol->v_raid_level = G_RAID_VOLUME_RL_UNKNOWN;
701 vol->v_strip_size = meta->strip_sectors * 512; //ZZZ
702 vol->v_disks_count = mdi->mdio_total_disks;
703 vol->v_sectorsize = 512; //ZZZ
704 for (j = 0; j < vol->v_disks_count; j++) {
705 sd = &vol->v_subdisks[j];
709 g_raid_start_volume(vol);
711 /* Create disk placeholders to store data for later writing. */
712 for (disk_pos = 0; disk_pos < mdi->mdio_total_disks; disk_pos++) {
713 pd = malloc(sizeof(*pd), M_MD_NVIDIA, M_WAITOK | M_ZERO);
714 pd->pd_disk_pos = disk_pos;
715 disk = g_raid_create_disk(sc);
716 disk->d_md_data = (void *)pd;
717 disk->d_state = G_RAID_DISK_S_OFFLINE;
718 sd = &vol->v_subdisks[disk_pos];
720 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
723 /* Make all disks found till the moment take their places. */
725 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
726 if (disk->d_state == G_RAID_DISK_S_NONE) {
727 g_raid_md_nvidia_start_disk(disk);
731 } while (disk != NULL);
733 mdi->mdio_started = 1;
734 G_RAID_DEBUG1(0, sc, "Array started.");
735 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
737 /* Pickup any STALE/SPARE disks to refill array if needed. */
738 g_raid_md_nvidia_refill(sc);
740 g_raid_event_send(vol, G_RAID_VOLUME_E_START, G_RAID_EVENT_VOLUME);
742 callout_stop(&mdi->mdio_start_co);
743 G_RAID_DEBUG1(1, sc, "root_mount_rel %p", mdi->mdio_rootmount);
744 root_mount_rel(mdi->mdio_rootmount);
745 mdi->mdio_rootmount = NULL;
749 g_raid_md_nvidia_new_disk(struct g_raid_disk *disk)
751 struct g_raid_softc *sc;
752 struct g_raid_md_object *md;
753 struct g_raid_md_nvidia_object *mdi;
754 struct nvidia_raid_conf *pdmeta;
755 struct g_raid_md_nvidia_perdisk *pd;
759 mdi = (struct g_raid_md_nvidia_object *)md;
760 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
761 pdmeta = pd->pd_meta;
763 if (mdi->mdio_started) {
764 if (g_raid_md_nvidia_start_disk(disk))
765 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
767 if (mdi->mdio_meta == NULL ||
768 mdi->mdio_meta->disk_number >= mdi->mdio_meta->total_disks) {
769 G_RAID_DEBUG1(1, sc, "Newer disk");
770 if (mdi->mdio_meta != NULL)
771 free(mdi->mdio_meta, M_MD_NVIDIA);
772 mdi->mdio_meta = nvidia_meta_copy(pdmeta);
773 mdi->mdio_total_disks = pdmeta->total_disks;
774 mdi->mdio_disks_present = 1;
775 } else if (pdmeta->disk_number < mdi->mdio_meta->total_disks) {
776 mdi->mdio_disks_present++;
777 G_RAID_DEBUG1(1, sc, "Matching disk (%d of %d up)",
778 mdi->mdio_disks_present,
779 mdi->mdio_total_disks);
781 G_RAID_DEBUG1(1, sc, "Spare disk");
783 /* If we collected all needed disks - start array. */
784 if (mdi->mdio_disks_present == mdi->mdio_total_disks)
785 g_raid_md_nvidia_start(sc);
790 g_raid_nvidia_go(void *arg)
792 struct g_raid_softc *sc;
793 struct g_raid_md_object *md;
794 struct g_raid_md_nvidia_object *mdi;
798 mdi = (struct g_raid_md_nvidia_object *)md;
799 if (!mdi->mdio_started) {
800 G_RAID_DEBUG1(0, sc, "Force array start due to timeout.");
801 g_raid_event_send(sc, G_RAID_NODE_E_START, 0);
806 g_raid_md_create_nvidia(struct g_raid_md_object *md, struct g_class *mp,
809 struct g_raid_softc *sc;
810 struct g_raid_md_nvidia_object *mdi;
813 mdi = (struct g_raid_md_nvidia_object *)md;
814 arc4rand(&mdi->mdio_volume_id, 16, 0);
815 snprintf(name, sizeof(name), "NVIDIA-%d",
816 atomic_fetchadd_int(&NVIDIANodeID, 1));
817 sc = g_raid_create_node(mp, name, md);
819 return (G_RAID_MD_TASTE_FAIL);
822 return (G_RAID_MD_TASTE_NEW);
826 g_raid_md_taste_nvidia(struct g_raid_md_object *md, struct g_class *mp,
827 struct g_consumer *cp, struct g_geom **gp)
829 struct g_consumer *rcp;
830 struct g_provider *pp;
831 struct g_raid_md_nvidia_object *mdi, *mdi1;
832 struct g_raid_softc *sc;
833 struct g_raid_disk *disk;
834 struct nvidia_raid_conf *meta;
835 struct g_raid_md_nvidia_perdisk *pd;
837 int result, spare, len;
841 G_RAID_DEBUG(1, "Tasting NVIDIA on %s", cp->provider->name);
842 mdi = (struct g_raid_md_nvidia_object *)md;
845 /* Read metadata from device. */
849 len = sizeof(vendor);
850 if (pp->geom->rank == 1)
851 g_io_getattr("GEOM::hba_vendor", cp, &len, &vendor);
852 meta = nvidia_meta_read(cp);
855 if (g_raid_aggressive_spare) {
856 if (vendor == 0x10de) {
858 "No NVIDIA metadata, forcing spare.");
863 "NVIDIA vendor mismatch 0x%04x != 0x10de",
867 return (G_RAID_MD_TASTE_FAIL);
870 /* Metadata valid. Print it. */
871 g_raid_md_nvidia_print(meta);
872 G_RAID_DEBUG(1, "NVIDIA disk position %d", meta->disk_number);
873 spare = 0;//(meta->type == NVIDIA_T_SPARE) ? 1 : 0;
876 /* Search for matching node. */
879 LIST_FOREACH(geom, &mp->geom, geom) {
883 if (sc->sc_stopping != 0)
885 if (sc->sc_md->mdo_class != md->mdo_class)
887 mdi1 = (struct g_raid_md_nvidia_object *)sc->sc_md;
889 if (mdi1->mdio_incomplete)
892 if (memcmp(&mdi1->mdio_volume_id,
893 &meta->volume_id, 16) == 0)
898 /* Found matching node. */
900 G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name);
901 result = G_RAID_MD_TASTE_EXISTING;
903 } else if (spare) { /* Not found needy node -- left for later. */
904 G_RAID_DEBUG(1, "Spare is not needed at this time");
907 } else { /* Not found matching node -- create one. */
908 result = G_RAID_MD_TASTE_NEW;
909 memcpy(&mdi->mdio_volume_id, &meta->volume_id, 16);
910 snprintf(name, sizeof(name), "NVIDIA-%d",
911 atomic_fetchadd_int(&NVIDIANodeID, 1));
912 sc = g_raid_create_node(mp, name, md);
915 callout_init(&mdi->mdio_start_co, 1);
916 callout_reset(&mdi->mdio_start_co, g_raid_start_timeout * hz,
917 g_raid_nvidia_go, sc);
918 mdi->mdio_rootmount = root_mount_hold("GRAID-NVIDIA");
919 G_RAID_DEBUG1(1, sc, "root_mount_hold %p", mdi->mdio_rootmount);
922 /* There is no return after this point, so we close passed consumer. */
923 g_access(cp, -1, 0, 0);
925 rcp = g_new_consumer(geom);
926 rcp->flags |= G_CF_DIRECT_RECEIVE;
928 if (g_access(rcp, 1, 1, 1) != 0)
932 sx_xlock(&sc->sc_lock);
934 pd = malloc(sizeof(*pd), M_MD_NVIDIA, M_WAITOK | M_ZERO);
937 pd->pd_disk_pos = -3;
939 pd->pd_disk_pos = -1;
941 pd->pd_disk_size = pp->mediasize;
942 disk = g_raid_create_disk(sc);
943 disk->d_md_data = (void *)pd;
944 disk->d_consumer = rcp;
947 g_raid_get_disk_info(disk);
949 g_raid_md_nvidia_new_disk(disk);
951 sx_xunlock(&sc->sc_lock);
956 free(meta, M_MD_NVIDIA);
957 return (G_RAID_MD_TASTE_FAIL);
961 g_raid_md_event_nvidia(struct g_raid_md_object *md,
962 struct g_raid_disk *disk, u_int event)
964 struct g_raid_softc *sc;
965 struct g_raid_subdisk *sd;
966 struct g_raid_md_nvidia_object *mdi;
967 struct g_raid_md_nvidia_perdisk *pd;
970 mdi = (struct g_raid_md_nvidia_object *)md;
973 case G_RAID_NODE_E_START:
974 if (!mdi->mdio_started) {
975 /* Bump volume ID to drop missing disks. */
976 arc4rand(&mdi->mdio_volume_id, 16, 0);
977 g_raid_md_nvidia_start(sc);
983 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
985 case G_RAID_DISK_E_DISCONNECTED:
986 /* If disk was assigned, just update statuses. */
987 if (pd->pd_disk_pos >= 0) {
988 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
989 if (disk->d_consumer) {
990 g_raid_kill_consumer(sc, disk->d_consumer);
991 disk->d_consumer = NULL;
993 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
994 g_raid_change_subdisk_state(sd,
995 G_RAID_SUBDISK_S_NONE);
996 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
997 G_RAID_EVENT_SUBDISK);
1000 /* Otherwise -- delete. */
1001 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
1002 g_raid_destroy_disk(disk);
1005 if (mdi->mdio_started) {
1006 /* Bump volume ID to prevent disk resurrection. */
1007 if (pd->pd_disk_pos >= 0)
1008 arc4rand(&mdi->mdio_volume_id, 16, 0);
1010 /* Write updated metadata to all disks. */
1011 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
1014 /* Check if anything left except placeholders. */
1015 if (g_raid_ndisks(sc, -1) ==
1016 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1017 g_raid_destroy_node(sc, 0);
1019 g_raid_md_nvidia_refill(sc);
1026 g_raid_md_ctl_nvidia(struct g_raid_md_object *md,
1027 struct gctl_req *req)
1029 struct g_raid_softc *sc;
1030 struct g_raid_volume *vol;
1031 struct g_raid_subdisk *sd;
1032 struct g_raid_disk *disk;
1033 struct g_raid_md_nvidia_object *mdi;
1034 struct g_raid_md_nvidia_perdisk *pd;
1035 struct g_consumer *cp;
1036 struct g_provider *pp;
1038 const char *verb, *volname, *levelname, *diskname;
1040 off_t size, sectorsize, strip, volsize;
1041 intmax_t *sizearg, *striparg;
1042 int numdisks, i, len, level, qual, update;
1046 mdi = (struct g_raid_md_nvidia_object *)md;
1047 verb = gctl_get_param(req, "verb", NULL);
1048 nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
1050 if (strcmp(verb, "label") == 0) {
1053 gctl_error(req, "Invalid number of arguments.");
1056 volname = gctl_get_asciiparam(req, "arg1");
1057 if (volname == NULL) {
1058 gctl_error(req, "No volume name.");
1061 levelname = gctl_get_asciiparam(req, "arg2");
1062 if (levelname == NULL) {
1063 gctl_error(req, "No RAID level.");
1066 if (strcasecmp(levelname, "RAID5") == 0)
1067 levelname = "RAID5-LS";
1068 if (g_raid_volume_str2level(levelname, &level, &qual)) {
1069 gctl_error(req, "Unknown RAID level '%s'.", levelname);
1072 numdisks = *nargs - 3;
1073 force = gctl_get_paraml(req, "force", sizeof(*force));
1074 if (!g_raid_md_nvidia_supported(level, qual, numdisks,
1075 force ? *force : 0)) {
1076 gctl_error(req, "Unsupported RAID level "
1077 "(0x%02x/0x%02x), or number of disks (%d).",
1078 level, qual, numdisks);
1082 /* Search for disks, connect them and probe. */
1083 size = 0x7fffffffffffffffllu;
1085 for (i = 0; i < numdisks; i++) {
1086 snprintf(arg, sizeof(arg), "arg%d", i + 3);
1087 diskname = gctl_get_asciiparam(req, arg);
1088 if (diskname == NULL) {
1089 gctl_error(req, "No disk name (%s).", arg);
1093 if (strcmp(diskname, "NONE") == 0) {
1098 cp = g_raid_open_consumer(sc, diskname);
1100 gctl_error(req, "Can't open '%s'.",
1102 g_topology_unlock();
1108 pd = malloc(sizeof(*pd), M_MD_NVIDIA, M_WAITOK | M_ZERO);
1109 pd->pd_disk_pos = i;
1110 disk = g_raid_create_disk(sc);
1111 disk->d_md_data = (void *)pd;
1112 disk->d_consumer = cp;
1116 g_topology_unlock();
1118 g_raid_get_disk_info(disk);
1120 pd->pd_disk_size = pp->mediasize;
1121 if (size > pp->mediasize)
1122 size = pp->mediasize;
1123 if (sectorsize < pp->sectorsize)
1124 sectorsize = pp->sectorsize;
1129 if (sectorsize <= 0) {
1130 gctl_error(req, "Can't get sector size.");
1134 /* Reserve space for metadata. */
1135 size -= 2 * sectorsize;
1137 /* Handle size argument. */
1138 len = sizeof(*sizearg);
1139 sizearg = gctl_get_param(req, "size", &len);
1140 if (sizearg != NULL && len == sizeof(*sizearg) &&
1142 if (*sizearg > size) {
1143 gctl_error(req, "Size too big %lld > %lld.",
1144 (long long)*sizearg, (long long)size);
1150 /* Handle strip argument. */
1152 len = sizeof(*striparg);
1153 striparg = gctl_get_param(req, "strip", &len);
1154 if (striparg != NULL && len == sizeof(*striparg) &&
1156 if (*striparg < sectorsize) {
1157 gctl_error(req, "Strip size too small.");
1160 if (*striparg % sectorsize != 0) {
1161 gctl_error(req, "Incorrect strip size.");
1164 if (strip > 65535 * sectorsize) {
1165 gctl_error(req, "Strip size too big.");
1171 /* Round size down to strip or sector. */
1172 if (level == G_RAID_VOLUME_RL_RAID1)
1173 size -= (size % sectorsize);
1174 else if (level == G_RAID_VOLUME_RL_RAID1E &&
1175 (numdisks & 1) != 0)
1176 size -= (size % (2 * strip));
1178 size -= (size % strip);
1180 gctl_error(req, "Size too small.");
1184 if (level == G_RAID_VOLUME_RL_RAID0 ||
1185 level == G_RAID_VOLUME_RL_CONCAT ||
1186 level == G_RAID_VOLUME_RL_SINGLE)
1187 volsize = size * numdisks;
1188 else if (level == G_RAID_VOLUME_RL_RAID1)
1190 else if (level == G_RAID_VOLUME_RL_RAID5)
1191 volsize = size * (numdisks - 1);
1193 volsize = ((size * numdisks) / strip / 2) *
1196 if (volsize > 0xffffffffllu * sectorsize) {
1197 gctl_error(req, "Size too big.");
1201 /* We have all we need, create things: volume, ... */
1202 mdi->mdio_total_disks = numdisks;
1203 mdi->mdio_started = 1;
1204 vol = g_raid_create_volume(sc, volname, -1);
1205 vol->v_md_data = (void *)(intptr_t)0;
1206 vol->v_raid_level = level;
1207 vol->v_raid_level_qualifier = qual;
1208 vol->v_strip_size = strip;
1209 vol->v_disks_count = numdisks;
1210 vol->v_mediasize = volsize;
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 g_raid_get_disk_info(disk);
1372 /* Welcome the "new" disk. */
1373 update += g_raid_md_nvidia_start_disk(disk);
1374 if (disk->d_state != G_RAID_DISK_S_SPARE &&
1375 disk->d_state != G_RAID_DISK_S_ACTIVE) {
1376 gctl_error(req, "Disk '%s' doesn't fit.",
1378 g_raid_destroy_disk(disk);
1384 /* Write new metadata if we changed something. */
1386 g_raid_md_write_nvidia(md, NULL, NULL, NULL);
1389 gctl_error(req, "Command '%s' is not supported.", verb);
1394 g_raid_md_write_nvidia(struct g_raid_md_object *md, struct g_raid_volume *tvol,
1395 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
1397 struct g_raid_softc *sc;
1398 struct g_raid_volume *vol;
1399 struct g_raid_subdisk *sd;
1400 struct g_raid_disk *disk;
1401 struct g_raid_md_nvidia_object *mdi;
1402 struct g_raid_md_nvidia_perdisk *pd;
1403 struct nvidia_raid_conf *meta;
1407 mdi = (struct g_raid_md_nvidia_object *)md;
1409 if (sc->sc_stopping == G_RAID_DESTROY_HARD)
1412 /* There is only one volume. */
1413 vol = TAILQ_FIRST(&sc->sc_volumes);
1415 /* Fill global fields. */
1416 meta = malloc(sizeof(*meta), M_MD_NVIDIA, M_WAITOK | M_ZERO);
1418 memcpy(meta, mdi->mdio_meta, sizeof(*meta));
1419 memcpy(meta->nvidia_id, NVIDIA_MAGIC, sizeof(NVIDIA_MAGIC) - 1);
1420 meta->config_size = 30;
1421 meta->version = 0x0064;
1422 meta->total_sectors = vol->v_mediasize / vol->v_sectorsize;
1423 meta->sector_size = vol->v_sectorsize;
1424 nvidia_meta_put_name(meta, vol->v_name);
1425 meta->magic_0 = NVIDIA_MAGIC0;
1426 memcpy(&meta->volume_id, &mdi->mdio_volume_id, 16);
1427 meta->state = NVIDIA_S_IDLE;
1428 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
1429 meta->array_width = 1;
1430 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E)
1431 meta->array_width = vol->v_disks_count / 2;
1432 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID5)
1433 meta->array_width = vol->v_disks_count - 1;
1435 meta->array_width = vol->v_disks_count;
1436 meta->total_disks = vol->v_disks_count;
1437 meta->orig_array_width = meta->array_width;
1438 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0)
1439 meta->type = NVIDIA_T_RAID0;
1440 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
1441 meta->type = NVIDIA_T_RAID1;
1442 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E)
1443 meta->type = NVIDIA_T_RAID01;
1444 else if (vol->v_raid_level == G_RAID_VOLUME_RL_CONCAT ||
1445 vol->v_raid_level == G_RAID_VOLUME_RL_SINGLE)
1446 meta->type = NVIDIA_T_CONCAT;
1447 else if (vol->v_raid_level_qualifier == G_RAID_VOLUME_RLQ_R5LA)
1448 meta->type = NVIDIA_T_RAID5;
1450 meta->type = NVIDIA_T_RAID5_SYM;
1451 meta->strip_sectors = vol->v_strip_size / vol->v_sectorsize;
1452 meta->strip_bytes = vol->v_strip_size;
1453 meta->strip_shift = ffs(meta->strip_sectors) - 1;
1454 meta->strip_mask = meta->strip_sectors - 1;
1455 meta->stripe_sectors = meta->strip_sectors * meta->orig_array_width;
1456 meta->stripe_bytes = meta->stripe_sectors * vol->v_sectorsize;
1457 meta->rebuild_lba = 0;
1458 meta->orig_type = meta->type;
1459 meta->orig_total_sectors = meta->total_sectors;
1462 for (i = 0; i < vol->v_disks_count; i++) {
1463 sd = &vol->v_subdisks[i];
1464 if ((sd->sd_state == G_RAID_SUBDISK_S_STALE ||
1465 sd->sd_state == G_RAID_SUBDISK_S_RESYNC ||
1467 meta->state != NVIDIA_S_REBUILD)
1468 meta->state = NVIDIA_S_SYNC;
1469 else if (sd->sd_state == G_RAID_SUBDISK_S_NEW ||
1470 sd->sd_state == G_RAID_SUBDISK_S_REBUILD)
1471 meta->state = NVIDIA_S_REBUILD;
1474 /* We are done. Print meta data and store them to disks. */
1475 if (mdi->mdio_meta != NULL)
1476 free(mdi->mdio_meta, M_MD_NVIDIA);
1477 mdi->mdio_meta = meta;
1479 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1480 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
1481 if (disk->d_state != G_RAID_DISK_S_ACTIVE &&
1482 disk->d_state != G_RAID_DISK_S_SPARE)
1484 if (pd->pd_meta != NULL) {
1485 free(pd->pd_meta, M_MD_NVIDIA);
1488 pd->pd_meta = nvidia_meta_copy(meta);
1489 if ((sd = TAILQ_FIRST(&disk->d_subdisks)) != NULL) {
1490 /* For RAID0+1 we need to translate order. */
1491 pd->pd_meta->disk_number =
1492 nvidia_meta_translate_disk(meta, sd->sd_pos);
1493 if (sd->sd_state != G_RAID_SUBDISK_S_ACTIVE) {
1494 pd->pd_meta->disk_status = 0x100;
1495 pd->pd_meta->rebuild_lba =
1496 sd->sd_rebuild_pos / vol->v_sectorsize *
1500 pd->pd_meta->disk_number = meta->total_disks + spares++;
1501 G_RAID_DEBUG(1, "Writing NVIDIA metadata to %s",
1502 g_raid_get_diskname(disk));
1503 g_raid_md_nvidia_print(pd->pd_meta);
1504 nvidia_meta_write(disk->d_consumer, pd->pd_meta);
1510 g_raid_md_fail_disk_nvidia(struct g_raid_md_object *md,
1511 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
1513 struct g_raid_softc *sc;
1514 struct g_raid_md_nvidia_perdisk *pd;
1515 struct g_raid_subdisk *sd;
1518 pd = (struct g_raid_md_nvidia_perdisk *)tdisk->d_md_data;
1520 /* We can't fail disk that is not a part of array now. */
1521 if (pd->pd_disk_pos < 0)
1524 /* Erase metadata to prevent disks's later resurrection. */
1525 if (tdisk->d_consumer)
1526 nvidia_meta_erase(tdisk->d_consumer);
1528 /* Change states. */
1529 g_raid_change_disk_state(tdisk, G_RAID_DISK_S_FAILED);
1530 TAILQ_FOREACH(sd, &tdisk->d_subdisks, sd_next) {
1531 g_raid_change_subdisk_state(sd,
1532 G_RAID_SUBDISK_S_FAILED);
1533 g_raid_event_send(sd, G_RAID_SUBDISK_E_FAILED,
1534 G_RAID_EVENT_SUBDISK);
1537 /* Write updated metadata to remaining disks. */
1538 g_raid_md_write_nvidia(md, NULL, NULL, tdisk);
1540 /* Check if anything left except placeholders. */
1541 if (g_raid_ndisks(sc, -1) ==
1542 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1543 g_raid_destroy_node(sc, 0);
1545 g_raid_md_nvidia_refill(sc);
1550 g_raid_md_free_disk_nvidia(struct g_raid_md_object *md,
1551 struct g_raid_disk *disk)
1553 struct g_raid_md_nvidia_perdisk *pd;
1555 pd = (struct g_raid_md_nvidia_perdisk *)disk->d_md_data;
1556 if (pd->pd_meta != NULL) {
1557 free(pd->pd_meta, M_MD_NVIDIA);
1560 free(pd, M_MD_NVIDIA);
1561 disk->d_md_data = NULL;
1566 g_raid_md_free_nvidia(struct g_raid_md_object *md)
1568 struct g_raid_md_nvidia_object *mdi;
1570 mdi = (struct g_raid_md_nvidia_object *)md;
1571 if (!mdi->mdio_started) {
1572 mdi->mdio_started = 0;
1573 callout_stop(&mdi->mdio_start_co);
1574 G_RAID_DEBUG1(1, md->mdo_softc,
1575 "root_mount_rel %p", mdi->mdio_rootmount);
1576 root_mount_rel(mdi->mdio_rootmount);
1577 mdi->mdio_rootmount = NULL;
1579 if (mdi->mdio_meta != NULL) {
1580 free(mdi->mdio_meta, M_MD_NVIDIA);
1581 mdi->mdio_meta = NULL;
1586 G_RAID_MD_DECLARE(nvidia, "NVIDIA");