2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2010 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/raid/g_raid.h"
46 #include "g_raid_md_if.h"
48 static MALLOC_DEFINE(M_MD_INTEL, "md_intel_data", "GEOM_RAID Intel metadata");
50 struct intel_raid_map {
52 uint32_t disk_sectors;
53 uint32_t stripe_count;
54 uint16_t strip_sectors;
56 #define INTEL_S_READY 0x00
57 #define INTEL_S_UNINITIALIZED 0x01
58 #define INTEL_S_DEGRADED 0x02
59 #define INTEL_S_FAILURE 0x03
62 #define INTEL_T_RAID0 0x00
63 #define INTEL_T_RAID1 0x01
64 #define INTEL_T_RAID5 0x05
67 uint8_t total_domains;
68 uint8_t failed_disk_num;
71 uint32_t disk_sectors_hi;
72 uint32_t stripe_count_hi;
74 uint32_t disk_idx[1]; /* total_disks entries. */
75 #define INTEL_DI_IDX 0x00ffffff
76 #define INTEL_DI_RBLD 0x01000000
79 struct intel_raid_vol {
81 u_int64_t total_sectors __packed;
83 #define INTEL_ST_BOOTABLE 0x00000001
84 #define INTEL_ST_BOOT_DEVICE 0x00000002
85 #define INTEL_ST_READ_COALESCING 0x00000004
86 #define INTEL_ST_WRITE_COALESCING 0x00000008
87 #define INTEL_ST_LAST_SHUTDOWN_DIRTY 0x00000010
88 #define INTEL_ST_HIDDEN_AT_BOOT 0x00000020
89 #define INTEL_ST_CURRENTLY_HIDDEN 0x00000040
90 #define INTEL_ST_VERIFY_AND_FIX 0x00000080
91 #define INTEL_ST_MAP_STATE_UNINIT 0x00000100
92 #define INTEL_ST_NO_AUTO_RECOVERY 0x00000200
93 #define INTEL_ST_CLONE_N_GO 0x00000400
94 #define INTEL_ST_CLONE_MAN_SYNC 0x00000800
95 #define INTEL_ST_CNG_MASTER_DISK_NUM 0x00001000
97 uint8_t migr_priority;
100 uint8_t cng_master_disk;
101 uint16_t cache_policy;
103 #define INTEL_CNGST_UPDATED 0
104 #define INTEL_CNGST_NEEDS_UPDATE 1
105 #define INTEL_CNGST_MASTER_MISSING 2
106 uint8_t cng_sub_state;
107 uint32_t filler_0[10];
109 uint32_t curr_migr_unit;
110 uint32_t checkpoint_id;
113 #define INTEL_MT_INIT 0
114 #define INTEL_MT_REBUILD 1
115 #define INTEL_MT_VERIFY 2
116 #define INTEL_MT_GEN_MIGR 3
117 #define INTEL_MT_STATE_CHANGE 4
118 #define INTEL_MT_REPAIR 5
121 uint16_t verify_errors;
123 uint32_t curr_migr_unit_hi;
124 uint32_t filler_1[3];
125 struct intel_raid_map map[1]; /* 2 entries if migr_state != 0. */
128 struct intel_raid_disk {
129 #define INTEL_SERIAL_LEN 16
130 uint8_t serial[INTEL_SERIAL_LEN];
134 #define INTEL_F_SPARE 0x01
135 #define INTEL_F_ASSIGNED 0x02
136 #define INTEL_F_FAILED 0x04
137 #define INTEL_F_ONLINE 0x08
138 #define INTEL_F_DISABLED 0x80
139 uint32_t owner_cfg_num;
144 struct intel_raid_conf {
145 uint8_t intel_id[24];
146 #define INTEL_MAGIC "Intel Raid ISM Cfg Sig. "
149 #define INTEL_VERSION_1000 "1.0.00" /* RAID0 */
150 #define INTEL_VERSION_1100 "1.1.00" /* RAID1 */
151 #define INTEL_VERSION_1200 "1.2.00" /* Many volumes */
152 #define INTEL_VERSION_1201 "1.2.01" /* 3 or 4 disks */
153 #define INTEL_VERSION_1202 "1.2.02" /* RAID5 */
154 #define INTEL_VERSION_1204 "1.2.04" /* 5 or 6 disks */
155 #define INTEL_VERSION_1206 "1.2.06" /* CNG */
156 #define INTEL_VERSION_1300 "1.3.00" /* Attributes */
160 uint32_t config_size;
163 uint32_t error_log_size;
165 #define INTEL_ATTR_RAID0 0x00000001
166 #define INTEL_ATTR_RAID1 0x00000002
167 #define INTEL_ATTR_RAID10 0x00000004
168 #define INTEL_ATTR_RAID1E 0x00000008
169 #define INTEL_ATTR_RAID5 0x00000010
170 #define INTEL_ATTR_RAIDCNG 0x00000020
171 #define INTEL_ATTR_EXT_STRIP 0x00000040
172 #define INTEL_ATTR_NVM_CACHE 0x02000000
173 #define INTEL_ATTR_2TB_DISK 0x04000000
174 #define INTEL_ATTR_BBM 0x08000000
175 #define INTEL_ATTR_NVM_CACHE2 0x10000000
176 #define INTEL_ATTR_2TB 0x20000000
177 #define INTEL_ATTR_PM 0x40000000
178 #define INTEL_ATTR_CHECKSUM 0x80000000
181 uint8_t total_volumes;
182 uint8_t error_log_pos;
185 uint32_t orig_config_id;
186 uint32_t pwr_cycle_count;
187 uint32_t bbm_log_size;
188 uint32_t filler_0[35];
189 struct intel_raid_disk disk[1]; /* total_disks entries. */
190 /* Here goes total_volumes of struct intel_raid_vol. */
193 #define INTEL_ATTR_SUPPORTED ( INTEL_ATTR_RAID0 | INTEL_ATTR_RAID1 | \
194 INTEL_ATTR_RAID10 | INTEL_ATTR_RAID1E | INTEL_ATTR_RAID5 | \
195 INTEL_ATTR_RAIDCNG | INTEL_ATTR_EXT_STRIP | INTEL_ATTR_2TB_DISK | \
196 INTEL_ATTR_2TB | INTEL_ATTR_PM | INTEL_ATTR_CHECKSUM )
198 #define INTEL_MAX_MD_SIZE(ndisks) \
199 (sizeof(struct intel_raid_conf) + \
200 sizeof(struct intel_raid_disk) * (ndisks - 1) + \
201 sizeof(struct intel_raid_vol) * 2 + \
202 sizeof(struct intel_raid_map) * 2 + \
203 sizeof(uint32_t) * (ndisks - 1) * 4)
205 struct g_raid_md_intel_perdisk {
206 struct intel_raid_conf *pd_meta;
208 struct intel_raid_disk pd_disk_meta;
211 struct g_raid_md_intel_pervolume {
215 int pv_cng_master_disk;
218 struct g_raid_md_intel_object {
219 struct g_raid_md_object mdio_base;
220 uint32_t mdio_config_id;
221 uint32_t mdio_orig_config_id;
222 uint32_t mdio_generation;
223 struct intel_raid_conf *mdio_meta;
224 struct callout mdio_start_co; /* STARTING state timer. */
225 int mdio_disks_present;
228 struct root_hold_token *mdio_rootmount; /* Root mount delay token. */
231 static g_raid_md_create_t g_raid_md_create_intel;
232 static g_raid_md_taste_t g_raid_md_taste_intel;
233 static g_raid_md_event_t g_raid_md_event_intel;
234 static g_raid_md_ctl_t g_raid_md_ctl_intel;
235 static g_raid_md_write_t g_raid_md_write_intel;
236 static g_raid_md_fail_disk_t g_raid_md_fail_disk_intel;
237 static g_raid_md_free_disk_t g_raid_md_free_disk_intel;
238 static g_raid_md_free_volume_t g_raid_md_free_volume_intel;
239 static g_raid_md_free_t g_raid_md_free_intel;
241 static kobj_method_t g_raid_md_intel_methods[] = {
242 KOBJMETHOD(g_raid_md_create, g_raid_md_create_intel),
243 KOBJMETHOD(g_raid_md_taste, g_raid_md_taste_intel),
244 KOBJMETHOD(g_raid_md_event, g_raid_md_event_intel),
245 KOBJMETHOD(g_raid_md_ctl, g_raid_md_ctl_intel),
246 KOBJMETHOD(g_raid_md_write, g_raid_md_write_intel),
247 KOBJMETHOD(g_raid_md_fail_disk, g_raid_md_fail_disk_intel),
248 KOBJMETHOD(g_raid_md_free_disk, g_raid_md_free_disk_intel),
249 KOBJMETHOD(g_raid_md_free_volume, g_raid_md_free_volume_intel),
250 KOBJMETHOD(g_raid_md_free, g_raid_md_free_intel),
254 static struct g_raid_md_class g_raid_md_intel_class = {
256 g_raid_md_intel_methods,
257 sizeof(struct g_raid_md_intel_object),
263 static struct intel_raid_map *
264 intel_get_map(struct intel_raid_vol *mvol, int i)
266 struct intel_raid_map *mmap;
268 if (i > (mvol->migr_state ? 1 : 0))
270 mmap = &mvol->map[0];
272 mmap = (struct intel_raid_map *)
273 &mmap->disk_idx[mmap->total_disks];
275 return ((struct intel_raid_map *)mmap);
278 static struct intel_raid_vol *
279 intel_get_volume(struct intel_raid_conf *meta, int i)
281 struct intel_raid_vol *mvol;
282 struct intel_raid_map *mmap;
286 mvol = (struct intel_raid_vol *)&meta->disk[meta->total_disks];
288 mmap = intel_get_map(mvol, mvol->migr_state ? 1 : 0);
289 mvol = (struct intel_raid_vol *)
290 &mmap->disk_idx[mmap->total_disks];
296 intel_get_map_offset(struct intel_raid_map *mmap)
298 off_t offset = (off_t)mmap->offset_hi << 32;
300 offset += mmap->offset;
305 intel_set_map_offset(struct intel_raid_map *mmap, off_t offset)
308 mmap->offset = offset & 0xffffffff;
309 mmap->offset_hi = offset >> 32;
313 intel_get_map_disk_sectors(struct intel_raid_map *mmap)
315 off_t disk_sectors = (off_t)mmap->disk_sectors_hi << 32;
317 disk_sectors += mmap->disk_sectors;
318 return (disk_sectors);
322 intel_set_map_disk_sectors(struct intel_raid_map *mmap, off_t disk_sectors)
325 mmap->disk_sectors = disk_sectors & 0xffffffff;
326 mmap->disk_sectors_hi = disk_sectors >> 32;
330 intel_set_map_stripe_count(struct intel_raid_map *mmap, off_t stripe_count)
333 mmap->stripe_count = stripe_count & 0xffffffff;
334 mmap->stripe_count_hi = stripe_count >> 32;
338 intel_get_disk_sectors(struct intel_raid_disk *disk)
340 off_t sectors = (off_t)disk->sectors_hi << 32;
342 sectors += disk->sectors;
347 intel_set_disk_sectors(struct intel_raid_disk *disk, off_t sectors)
350 disk->sectors = sectors & 0xffffffff;
351 disk->sectors_hi = sectors >> 32;
355 intel_get_vol_curr_migr_unit(struct intel_raid_vol *vol)
357 off_t curr_migr_unit = (off_t)vol->curr_migr_unit_hi << 32;
359 curr_migr_unit += vol->curr_migr_unit;
360 return (curr_migr_unit);
364 intel_set_vol_curr_migr_unit(struct intel_raid_vol *vol, off_t curr_migr_unit)
367 vol->curr_migr_unit = curr_migr_unit & 0xffffffff;
368 vol->curr_migr_unit_hi = curr_migr_unit >> 32;
372 intel_status2str(int status)
378 case INTEL_S_UNINITIALIZED:
379 return ("UNINITIALIZED");
380 case INTEL_S_DEGRADED:
382 case INTEL_S_FAILURE:
390 intel_type2str(int type)
406 intel_cngst2str(int cng_state)
410 case INTEL_CNGST_UPDATED:
412 case INTEL_CNGST_NEEDS_UPDATE:
413 return ("NEEDS_UPDATE");
414 case INTEL_CNGST_MASTER_MISSING:
415 return ("MASTER_MISSING");
422 intel_mt2str(int type)
428 case INTEL_MT_REBUILD:
430 case INTEL_MT_VERIFY:
432 case INTEL_MT_GEN_MIGR:
434 case INTEL_MT_STATE_CHANGE:
435 return ("STATE_CHANGE");
436 case INTEL_MT_REPAIR:
444 g_raid_md_intel_print(struct intel_raid_conf *meta)
446 struct intel_raid_vol *mvol;
447 struct intel_raid_map *mmap;
450 if (g_raid_debug < 1)
453 printf("********* ATA Intel MatrixRAID Metadata *********\n");
454 printf("intel_id <%.24s>\n", meta->intel_id);
455 printf("version <%.6s>\n", meta->version);
456 printf("checksum 0x%08x\n", meta->checksum);
457 printf("config_size 0x%08x\n", meta->config_size);
458 printf("config_id 0x%08x\n", meta->config_id);
459 printf("generation 0x%08x\n", meta->generation);
460 printf("error_log_size %d\n", meta->error_log_size);
461 printf("attributes 0x%b\n", meta->attributes,
477 printf("total_disks %u\n", meta->total_disks);
478 printf("total_volumes %u\n", meta->total_volumes);
479 printf("error_log_pos %u\n", meta->error_log_pos);
480 printf("cache_size %u\n", meta->cache_size);
481 printf("orig_config_id 0x%08x\n", meta->orig_config_id);
482 printf("pwr_cycle_count %u\n", meta->pwr_cycle_count);
483 printf("bbm_log_size %u\n", meta->bbm_log_size);
484 printf("Flags: S - Spare, A - Assigned, F - Failed, O - Online, D - Disabled\n");
485 printf("DISK# serial disk_sectors disk_sectors_hi disk_id flags owner\n");
486 for (i = 0; i < meta->total_disks; i++ ) {
487 printf(" %d <%.16s> %u %u 0x%08x 0x%b %08x\n", i,
488 meta->disk[i].serial, meta->disk[i].sectors,
489 meta->disk[i].sectors_hi, meta->disk[i].id,
490 meta->disk[i].flags, "\20\01S\02A\03F\04O\05D",
491 meta->disk[i].owner_cfg_num);
493 for (i = 0; i < meta->total_volumes; i++) {
494 mvol = intel_get_volume(meta, i);
495 printf(" ****** Volume %d ******\n", i);
496 printf(" name %.16s\n", mvol->name);
497 printf(" total_sectors %ju\n", mvol->total_sectors);
498 printf(" state 0x%b\n", mvol->state,
502 "\003READ_COALESCING"
503 "\004WRITE_COALESCING"
504 "\005LAST_SHUTDOWN_DIRTY"
506 "\007CURRENTLY_HIDDEN"
508 "\011MAP_STATE_UNINIT"
509 "\012NO_AUTO_RECOVERY"
512 "\015CNG_MASTER_DISK_NUM");
513 printf(" reserved %u\n", mvol->reserved);
514 printf(" migr_priority %u\n", mvol->migr_priority);
515 printf(" num_sub_vols %u\n", mvol->num_sub_vols);
516 printf(" tid %u\n", mvol->tid);
517 printf(" cng_master_disk %u\n", mvol->cng_master_disk);
518 printf(" cache_policy %u\n", mvol->cache_policy);
519 printf(" cng_state %u (%s)\n", mvol->cng_state,
520 intel_cngst2str(mvol->cng_state));
521 printf(" cng_sub_state %u\n", mvol->cng_sub_state);
522 printf(" curr_migr_unit %u\n", mvol->curr_migr_unit);
523 printf(" curr_migr_unit_hi %u\n", mvol->curr_migr_unit_hi);
524 printf(" checkpoint_id %u\n", mvol->checkpoint_id);
525 printf(" migr_state %u\n", mvol->migr_state);
526 printf(" migr_type %u (%s)\n", mvol->migr_type,
527 intel_mt2str(mvol->migr_type));
528 printf(" dirty %u\n", mvol->dirty);
529 printf(" fs_state %u\n", mvol->fs_state);
530 printf(" verify_errors %u\n", mvol->verify_errors);
531 printf(" bad_blocks %u\n", mvol->bad_blocks);
533 for (j = 0; j < (mvol->migr_state ? 2 : 1); j++) {
534 printf(" *** Map %d ***\n", j);
535 mmap = intel_get_map(mvol, j);
536 printf(" offset %u\n", mmap->offset);
537 printf(" offset_hi %u\n", mmap->offset_hi);
538 printf(" disk_sectors %u\n", mmap->disk_sectors);
539 printf(" disk_sectors_hi %u\n", mmap->disk_sectors_hi);
540 printf(" stripe_count %u\n", mmap->stripe_count);
541 printf(" stripe_count_hi %u\n", mmap->stripe_count_hi);
542 printf(" strip_sectors %u\n", mmap->strip_sectors);
543 printf(" status %u (%s)\n", mmap->status,
544 intel_status2str(mmap->status));
545 printf(" type %u (%s)\n", mmap->type,
546 intel_type2str(mmap->type));
547 printf(" total_disks %u\n", mmap->total_disks);
548 printf(" total_domains %u\n", mmap->total_domains);
549 printf(" failed_disk_num %u\n", mmap->failed_disk_num);
550 printf(" ddf %u\n", mmap->ddf);
551 printf(" disk_idx ");
552 for (k = 0; k < mmap->total_disks; k++)
553 printf(" 0x%08x", mmap->disk_idx[k]);
557 printf("=================================================\n");
560 static struct intel_raid_conf *
561 intel_meta_copy(struct intel_raid_conf *meta)
563 struct intel_raid_conf *nmeta;
565 nmeta = malloc(meta->config_size, M_MD_INTEL, M_WAITOK);
566 memcpy(nmeta, meta, meta->config_size);
571 intel_meta_find_disk(struct intel_raid_conf *meta, char *serial)
575 for (pos = 0; pos < meta->total_disks; pos++) {
576 if (strncmp(meta->disk[pos].serial,
577 serial, INTEL_SERIAL_LEN) == 0)
583 static struct intel_raid_conf *
584 intel_meta_read(struct g_consumer *cp)
586 struct g_provider *pp;
587 struct intel_raid_conf *meta;
588 struct intel_raid_vol *mvol;
589 struct intel_raid_map *mmap, *mmap1;
591 int error, i, j, k, left, size;
592 uint32_t checksum, *ptr;
596 /* Read the anchor sector. */
597 buf = g_read_data(cp,
598 pp->mediasize - pp->sectorsize * 2, pp->sectorsize, &error);
600 G_RAID_DEBUG(1, "Cannot read metadata from %s (error=%d).",
604 meta = (struct intel_raid_conf *)buf;
606 /* Check if this is an Intel RAID struct */
607 if (strncmp(meta->intel_id, INTEL_MAGIC, strlen(INTEL_MAGIC))) {
608 G_RAID_DEBUG(1, "Intel signature check failed on %s", pp->name);
612 if (meta->config_size > 65536 ||
613 meta->config_size < sizeof(struct intel_raid_conf)) {
614 G_RAID_DEBUG(1, "Intel metadata size looks wrong: %d",
619 size = meta->config_size;
620 meta = malloc(size, M_MD_INTEL, M_WAITOK);
621 memcpy(meta, buf, min(size, pp->sectorsize));
624 /* Read all the rest, if needed. */
625 if (meta->config_size > pp->sectorsize) {
626 left = (meta->config_size - 1) / pp->sectorsize;
627 buf = g_read_data(cp,
628 pp->mediasize - pp->sectorsize * (2 + left),
629 pp->sectorsize * left, &error);
631 G_RAID_DEBUG(1, "Cannot read remaining metadata"
632 " part from %s (error=%d).",
634 free(meta, M_MD_INTEL);
637 memcpy(((char *)meta) + pp->sectorsize, buf,
638 pp->sectorsize * left);
642 /* Check metadata checksum. */
643 for (checksum = 0, ptr = (uint32_t *)meta, i = 0;
644 i < (meta->config_size / sizeof(uint32_t)); i++) {
647 checksum -= meta->checksum;
648 if (checksum != meta->checksum) {
649 G_RAID_DEBUG(1, "Intel checksum check failed on %s", pp->name);
650 free(meta, M_MD_INTEL);
654 /* Validate metadata size. */
655 size = sizeof(struct intel_raid_conf) +
656 sizeof(struct intel_raid_disk) * (meta->total_disks - 1) +
657 sizeof(struct intel_raid_vol) * meta->total_volumes;
658 if (size > meta->config_size) {
660 G_RAID_DEBUG(1, "Intel metadata size incorrect %d < %d",
661 meta->config_size, size);
662 free(meta, M_MD_INTEL);
665 for (i = 0; i < meta->total_volumes; i++) {
666 mvol = intel_get_volume(meta, i);
667 mmap = intel_get_map(mvol, 0);
668 size += 4 * (mmap->total_disks - 1);
669 if (size > meta->config_size)
671 if (mvol->migr_state) {
672 size += sizeof(struct intel_raid_map);
673 if (size > meta->config_size)
675 mmap = intel_get_map(mvol, 1);
676 size += 4 * (mmap->total_disks - 1);
677 if (size > meta->config_size)
682 g_raid_md_intel_print(meta);
684 if (strncmp(meta->version, INTEL_VERSION_1300, 6) > 0) {
685 G_RAID_DEBUG(1, "Intel unsupported version: '%.6s'",
687 free(meta, M_MD_INTEL);
691 if (strncmp(meta->version, INTEL_VERSION_1300, 6) >= 0 &&
692 (meta->attributes & ~INTEL_ATTR_SUPPORTED) != 0) {
693 G_RAID_DEBUG(1, "Intel unsupported attributes: 0x%08x",
694 meta->attributes & ~INTEL_ATTR_SUPPORTED);
695 free(meta, M_MD_INTEL);
699 /* Validate disk indexes. */
700 for (i = 0; i < meta->total_volumes; i++) {
701 mvol = intel_get_volume(meta, i);
702 for (j = 0; j < (mvol->migr_state ? 2 : 1); j++) {
703 mmap = intel_get_map(mvol, j);
704 for (k = 0; k < mmap->total_disks; k++) {
705 if ((mmap->disk_idx[k] & INTEL_DI_IDX) >
707 G_RAID_DEBUG(1, "Intel metadata disk"
708 " index %d too big (>%d)",
709 mmap->disk_idx[k] & INTEL_DI_IDX,
711 free(meta, M_MD_INTEL);
718 /* Validate migration types. */
719 for (i = 0; i < meta->total_volumes; i++) {
720 mvol = intel_get_volume(meta, i);
721 /* Deny unknown migration types. */
722 if (mvol->migr_state &&
723 mvol->migr_type != INTEL_MT_INIT &&
724 mvol->migr_type != INTEL_MT_REBUILD &&
725 mvol->migr_type != INTEL_MT_VERIFY &&
726 mvol->migr_type != INTEL_MT_GEN_MIGR &&
727 mvol->migr_type != INTEL_MT_REPAIR) {
728 G_RAID_DEBUG(1, "Intel metadata has unsupported"
729 " migration type %d", mvol->migr_type);
730 free(meta, M_MD_INTEL);
733 /* Deny general migrations except SINGLE->RAID1. */
734 if (mvol->migr_state &&
735 mvol->migr_type == INTEL_MT_GEN_MIGR) {
736 mmap = intel_get_map(mvol, 0);
737 mmap1 = intel_get_map(mvol, 1);
738 if (mmap1->total_disks != 1 ||
739 mmap->type != INTEL_T_RAID1 ||
740 mmap->total_disks != 2 ||
741 mmap->offset != mmap1->offset ||
742 mmap->disk_sectors != mmap1->disk_sectors ||
743 mmap->total_domains != mmap->total_disks ||
744 mmap->offset_hi != mmap1->offset_hi ||
745 mmap->disk_sectors_hi != mmap1->disk_sectors_hi ||
746 (mmap->disk_idx[0] != mmap1->disk_idx[0] &&
747 mmap->disk_idx[0] != mmap1->disk_idx[1])) {
748 G_RAID_DEBUG(1, "Intel metadata has unsupported"
749 " variant of general migration");
750 free(meta, M_MD_INTEL);
760 intel_meta_write(struct g_consumer *cp, struct intel_raid_conf *meta)
762 struct g_provider *pp;
764 int error, i, sectors;
765 uint32_t checksum, *ptr;
769 /* Recalculate checksum for case if metadata were changed. */
771 for (checksum = 0, ptr = (uint32_t *)meta, i = 0;
772 i < (meta->config_size / sizeof(uint32_t)); i++) {
775 meta->checksum = checksum;
777 /* Create and fill buffer. */
778 sectors = howmany(meta->config_size, pp->sectorsize);
779 buf = malloc(sectors * pp->sectorsize, M_MD_INTEL, M_WAITOK | M_ZERO);
781 memcpy(buf, ((char *)meta) + pp->sectorsize,
782 (sectors - 1) * pp->sectorsize);
784 memcpy(buf + (sectors - 1) * pp->sectorsize, meta, pp->sectorsize);
786 error = g_write_data(cp,
787 pp->mediasize - pp->sectorsize * (1 + sectors),
788 buf, pp->sectorsize * sectors);
790 G_RAID_DEBUG(1, "Cannot write metadata to %s (error=%d).",
794 free(buf, M_MD_INTEL);
799 intel_meta_erase(struct g_consumer *cp)
801 struct g_provider *pp;
806 buf = malloc(pp->sectorsize, M_MD_INTEL, M_WAITOK | M_ZERO);
807 error = g_write_data(cp,
808 pp->mediasize - 2 * pp->sectorsize,
809 buf, pp->sectorsize);
811 G_RAID_DEBUG(1, "Cannot erase metadata on %s (error=%d).",
814 free(buf, M_MD_INTEL);
819 intel_meta_write_spare(struct g_consumer *cp, struct intel_raid_disk *d)
821 struct intel_raid_conf *meta;
824 /* Fill anchor and single disk. */
825 meta = malloc(INTEL_MAX_MD_SIZE(1), M_MD_INTEL, M_WAITOK | M_ZERO);
826 memcpy(&meta->intel_id[0], INTEL_MAGIC, sizeof(INTEL_MAGIC) - 1);
827 memcpy(&meta->version[0], INTEL_VERSION_1000,
828 sizeof(INTEL_VERSION_1000) - 1);
829 meta->config_size = INTEL_MAX_MD_SIZE(1);
830 meta->config_id = meta->orig_config_id = arc4random();
831 meta->generation = 1;
832 meta->total_disks = 1;
834 error = intel_meta_write(cp, meta);
835 free(meta, M_MD_INTEL);
839 static struct g_raid_disk *
840 g_raid_md_intel_get_disk(struct g_raid_softc *sc, int id)
842 struct g_raid_disk *disk;
843 struct g_raid_md_intel_perdisk *pd;
845 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
846 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
847 if (pd->pd_disk_pos == id)
854 g_raid_md_intel_supported(int level, int qual, int disks, int force)
858 case G_RAID_VOLUME_RL_RAID0:
861 if (!force && (disks < 2 || disks > 6))
864 case G_RAID_VOLUME_RL_RAID1:
867 if (!force && (disks != 2))
870 case G_RAID_VOLUME_RL_RAID1E:
873 if (!force && (disks != 4))
876 case G_RAID_VOLUME_RL_RAID5:
879 if (!force && disks > 6)
881 if (qual != G_RAID_VOLUME_RLQ_R5LA)
887 if (level != G_RAID_VOLUME_RL_RAID5 && qual != G_RAID_VOLUME_RLQ_NONE)
892 static struct g_raid_volume *
893 g_raid_md_intel_get_volume(struct g_raid_softc *sc, int id)
895 struct g_raid_volume *mvol;
896 struct g_raid_md_intel_pervolume *pv;
898 TAILQ_FOREACH(mvol, &sc->sc_volumes, v_next) {
899 pv = mvol->v_md_data;
900 if (pv->pv_volume_pos == id)
907 g_raid_md_intel_start_disk(struct g_raid_disk *disk)
909 struct g_raid_softc *sc;
910 struct g_raid_subdisk *sd, *tmpsd;
911 struct g_raid_disk *olddisk, *tmpdisk;
912 struct g_raid_md_object *md;
913 struct g_raid_md_intel_object *mdi;
914 struct g_raid_md_intel_pervolume *pv;
915 struct g_raid_md_intel_perdisk *pd, *oldpd;
916 struct intel_raid_conf *meta;
917 struct intel_raid_vol *mvol;
918 struct intel_raid_map *mmap0, *mmap1;
919 int disk_pos, resurrection = 0, migr_global, i;
923 mdi = (struct g_raid_md_intel_object *)md;
924 meta = mdi->mdio_meta;
925 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
928 /* Find disk position in metadata by its serial. */
929 disk_pos = intel_meta_find_disk(meta, pd->pd_disk_meta.serial);
931 G_RAID_DEBUG1(1, sc, "Unknown, probably new or stale disk");
932 /* Failed stale disk is useless for us. */
933 if ((pd->pd_disk_meta.flags & INTEL_F_FAILED) &&
934 !(pd->pd_disk_meta.flags & INTEL_F_DISABLED)) {
935 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE_FAILED);
938 /* If we are in the start process, that's all for now. */
939 if (!mdi->mdio_started)
942 * If we have already started - try to get use of the disk.
943 * Try to replace OFFLINE disks first, then FAILED.
945 TAILQ_FOREACH(tmpdisk, &sc->sc_disks, d_next) {
946 if (tmpdisk->d_state != G_RAID_DISK_S_OFFLINE &&
947 tmpdisk->d_state != G_RAID_DISK_S_FAILED)
949 /* Make sure this disk is big enough. */
950 TAILQ_FOREACH(sd, &tmpdisk->d_subdisks, sd_next) {
952 intel_get_disk_sectors(&pd->pd_disk_meta);
954 if (sd->sd_offset + sd->sd_size + 4096 >
955 disk_sectors * 512) {
957 "Disk too small (%llu < %llu)",
961 sd->sd_offset + sd->sd_size + 4096);
967 if (tmpdisk->d_state == G_RAID_DISK_S_OFFLINE) {
970 } else if (olddisk == NULL)
973 if (olddisk == NULL) {
975 if (pd->pd_disk_meta.flags & INTEL_F_SPARE) {
976 g_raid_change_disk_state(disk,
977 G_RAID_DISK_S_SPARE);
980 g_raid_change_disk_state(disk,
981 G_RAID_DISK_S_STALE);
985 oldpd = (struct g_raid_md_intel_perdisk *)olddisk->d_md_data;
986 disk_pos = oldpd->pd_disk_pos;
990 if (olddisk == NULL) {
991 /* Find placeholder by position. */
992 olddisk = g_raid_md_intel_get_disk(sc, disk_pos);
994 panic("No disk at position %d!", disk_pos);
995 if (olddisk->d_state != G_RAID_DISK_S_OFFLINE) {
996 G_RAID_DEBUG1(1, sc, "More than one disk for pos %d",
998 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE);
1001 oldpd = (struct g_raid_md_intel_perdisk *)olddisk->d_md_data;
1004 /* Replace failed disk or placeholder with new disk. */
1005 TAILQ_FOREACH_SAFE(sd, &olddisk->d_subdisks, sd_next, tmpsd) {
1006 TAILQ_REMOVE(&olddisk->d_subdisks, sd, sd_next);
1007 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1010 oldpd->pd_disk_pos = -2;
1011 pd->pd_disk_pos = disk_pos;
1013 /* If it was placeholder -- destroy it. */
1014 if (olddisk->d_state == G_RAID_DISK_S_OFFLINE) {
1015 g_raid_destroy_disk(olddisk);
1017 /* Otherwise, make it STALE_FAILED. */
1018 g_raid_change_disk_state(olddisk, G_RAID_DISK_S_STALE_FAILED);
1019 /* Update global metadata just in case. */
1020 memcpy(&meta->disk[disk_pos], &pd->pd_disk_meta,
1021 sizeof(struct intel_raid_disk));
1024 /* Welcome the new disk. */
1025 if ((meta->disk[disk_pos].flags & INTEL_F_DISABLED) &&
1026 !(pd->pd_disk_meta.flags & INTEL_F_SPARE))
1027 g_raid_change_disk_state(disk, G_RAID_DISK_S_DISABLED);
1028 else if (resurrection)
1029 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
1030 else if (meta->disk[disk_pos].flags & INTEL_F_FAILED)
1031 g_raid_change_disk_state(disk, G_RAID_DISK_S_FAILED);
1032 else if (meta->disk[disk_pos].flags & INTEL_F_SPARE)
1033 g_raid_change_disk_state(disk, G_RAID_DISK_S_SPARE);
1035 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
1036 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
1037 pv = sd->sd_volume->v_md_data;
1038 mvol = intel_get_volume(meta, pv->pv_volume_pos);
1039 mmap0 = intel_get_map(mvol, 0);
1040 if (mvol->migr_state)
1041 mmap1 = intel_get_map(mvol, 1);
1046 for (i = 0; i < mmap0->total_disks; i++) {
1047 if ((mmap0->disk_idx[i] & INTEL_DI_RBLD) == 0 &&
1048 (mmap1->disk_idx[i] & INTEL_DI_RBLD) != 0)
1052 if ((meta->disk[disk_pos].flags & INTEL_F_DISABLED) &&
1053 !(pd->pd_disk_meta.flags & INTEL_F_SPARE)) {
1054 /* Disabled disk, useless. */
1055 g_raid_change_subdisk_state(sd,
1056 G_RAID_SUBDISK_S_NONE);
1057 } else if (resurrection) {
1058 /* Stale disk, almost same as new. */
1059 g_raid_change_subdisk_state(sd,
1060 G_RAID_SUBDISK_S_NEW);
1061 } else if (meta->disk[disk_pos].flags & INTEL_F_FAILED) {
1062 /* Failed disk, almost useless. */
1063 g_raid_change_subdisk_state(sd,
1064 G_RAID_SUBDISK_S_FAILED);
1065 } else if (mvol->migr_state == 0) {
1066 if (mmap0->status == INTEL_S_UNINITIALIZED &&
1067 (!pv->pv_cng || pv->pv_cng_master_disk != disk_pos)) {
1068 /* Freshly created uninitialized volume. */
1069 g_raid_change_subdisk_state(sd,
1070 G_RAID_SUBDISK_S_UNINITIALIZED);
1071 } else if (mmap0->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) {
1072 /* Freshly inserted disk. */
1073 g_raid_change_subdisk_state(sd,
1074 G_RAID_SUBDISK_S_NEW);
1075 } else if (mvol->dirty && (!pv->pv_cng ||
1076 pv->pv_cng_master_disk != disk_pos)) {
1077 /* Dirty volume (unclean shutdown). */
1078 g_raid_change_subdisk_state(sd,
1079 G_RAID_SUBDISK_S_STALE);
1081 /* Up to date disk. */
1082 g_raid_change_subdisk_state(sd,
1083 G_RAID_SUBDISK_S_ACTIVE);
1085 } else if (mvol->migr_type == INTEL_MT_INIT ||
1086 mvol->migr_type == INTEL_MT_REBUILD) {
1087 if (mmap0->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) {
1088 /* Freshly inserted disk. */
1089 g_raid_change_subdisk_state(sd,
1090 G_RAID_SUBDISK_S_NEW);
1091 } else if (mmap1->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) {
1092 /* Rebuilding disk. */
1093 g_raid_change_subdisk_state(sd,
1094 G_RAID_SUBDISK_S_REBUILD);
1096 sd->sd_rebuild_pos = 0;
1098 sd->sd_rebuild_pos =
1099 intel_get_vol_curr_migr_unit(mvol) *
1100 sd->sd_volume->v_strip_size *
1101 mmap0->total_domains;
1103 } else if (mvol->migr_type == INTEL_MT_INIT &&
1105 /* Freshly created uninitialized volume. */
1106 g_raid_change_subdisk_state(sd,
1107 G_RAID_SUBDISK_S_UNINITIALIZED);
1108 } else if (mvol->dirty && (!pv->pv_cng ||
1109 pv->pv_cng_master_disk != disk_pos)) {
1110 /* Dirty volume (unclean shutdown). */
1111 g_raid_change_subdisk_state(sd,
1112 G_RAID_SUBDISK_S_STALE);
1114 /* Up to date disk. */
1115 g_raid_change_subdisk_state(sd,
1116 G_RAID_SUBDISK_S_ACTIVE);
1118 } else if (mvol->migr_type == INTEL_MT_VERIFY ||
1119 mvol->migr_type == INTEL_MT_REPAIR) {
1120 if (mmap0->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) {
1121 /* Freshly inserted disk. */
1122 g_raid_change_subdisk_state(sd,
1123 G_RAID_SUBDISK_S_NEW);
1124 } else if ((mmap1->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) ||
1126 /* Resyncing disk. */
1127 g_raid_change_subdisk_state(sd,
1128 G_RAID_SUBDISK_S_RESYNC);
1130 sd->sd_rebuild_pos = 0;
1132 sd->sd_rebuild_pos =
1133 intel_get_vol_curr_migr_unit(mvol) *
1134 sd->sd_volume->v_strip_size *
1135 mmap0->total_domains;
1137 } else if (mvol->dirty) {
1138 /* Dirty volume (unclean shutdown). */
1139 g_raid_change_subdisk_state(sd,
1140 G_RAID_SUBDISK_S_STALE);
1142 /* Up to date disk. */
1143 g_raid_change_subdisk_state(sd,
1144 G_RAID_SUBDISK_S_ACTIVE);
1146 } else if (mvol->migr_type == INTEL_MT_GEN_MIGR) {
1147 if ((mmap1->disk_idx[0] & INTEL_DI_IDX) != disk_pos) {
1148 /* Freshly inserted disk. */
1149 g_raid_change_subdisk_state(sd,
1150 G_RAID_SUBDISK_S_NEW);
1152 /* Up to date disk. */
1153 g_raid_change_subdisk_state(sd,
1154 G_RAID_SUBDISK_S_ACTIVE);
1157 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
1158 G_RAID_EVENT_SUBDISK);
1161 /* Update status of our need for spare. */
1162 if (mdi->mdio_started) {
1163 mdi->mdio_incomplete =
1164 (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) +
1165 g_raid_ndisks(sc, G_RAID_DISK_S_DISABLED) <
1169 return (resurrection);
1173 g_disk_md_intel_retaste(void *arg, int pending)
1176 G_RAID_DEBUG(1, "Array is not complete, trying to retaste.");
1177 g_retaste(&g_raid_class);
1178 free(arg, M_MD_INTEL);
1182 g_raid_md_intel_refill(struct g_raid_softc *sc)
1184 struct g_raid_md_object *md;
1185 struct g_raid_md_intel_object *mdi;
1186 struct intel_raid_conf *meta;
1187 struct g_raid_disk *disk;
1192 mdi = (struct g_raid_md_intel_object *)md;
1193 meta = mdi->mdio_meta;
1196 /* Make sure we miss anything. */
1197 na = g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) +
1198 g_raid_ndisks(sc, G_RAID_DISK_S_DISABLED);
1199 if (na == meta->total_disks)
1202 G_RAID_DEBUG1(1, md->mdo_softc,
1203 "Array is not complete (%d of %d), "
1204 "trying to refill.", na, meta->total_disks);
1206 /* Try to get use some of STALE disks. */
1207 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1208 if (disk->d_state == G_RAID_DISK_S_STALE) {
1209 update += g_raid_md_intel_start_disk(disk);
1210 if (disk->d_state == G_RAID_DISK_S_ACTIVE ||
1211 disk->d_state == G_RAID_DISK_S_DISABLED)
1218 /* Try to get use some of SPARE disks. */
1219 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1220 if (disk->d_state == G_RAID_DISK_S_SPARE) {
1221 update += g_raid_md_intel_start_disk(disk);
1222 if (disk->d_state == G_RAID_DISK_S_ACTIVE)
1226 } while (disk != NULL);
1228 /* Write new metadata if we changed something. */
1230 g_raid_md_write_intel(md, NULL, NULL, NULL);
1231 meta = mdi->mdio_meta;
1234 /* Update status of our need for spare. */
1235 mdi->mdio_incomplete = (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) +
1236 g_raid_ndisks(sc, G_RAID_DISK_S_DISABLED) < meta->total_disks);
1238 /* Request retaste hoping to find spare. */
1239 if (mdi->mdio_incomplete) {
1240 task = malloc(sizeof(struct task),
1241 M_MD_INTEL, M_WAITOK | M_ZERO);
1242 TASK_INIT(task, 0, g_disk_md_intel_retaste, task);
1243 taskqueue_enqueue(taskqueue_swi, task);
1248 g_raid_md_intel_start(struct g_raid_softc *sc)
1250 struct g_raid_md_object *md;
1251 struct g_raid_md_intel_object *mdi;
1252 struct g_raid_md_intel_pervolume *pv;
1253 struct g_raid_md_intel_perdisk *pd;
1254 struct intel_raid_conf *meta;
1255 struct intel_raid_vol *mvol;
1256 struct intel_raid_map *mmap;
1257 struct g_raid_volume *vol;
1258 struct g_raid_subdisk *sd;
1259 struct g_raid_disk *disk;
1263 mdi = (struct g_raid_md_intel_object *)md;
1264 meta = mdi->mdio_meta;
1266 /* Create volumes and subdisks. */
1267 for (i = 0; i < meta->total_volumes; i++) {
1268 mvol = intel_get_volume(meta, i);
1269 mmap = intel_get_map(mvol, 0);
1270 vol = g_raid_create_volume(sc, mvol->name, mvol->tid - 1);
1271 pv = malloc(sizeof(*pv), M_MD_INTEL, M_WAITOK | M_ZERO);
1272 pv->pv_volume_pos = i;
1273 pv->pv_cng = (mvol->state & INTEL_ST_CLONE_N_GO) != 0;
1274 pv->pv_cng_man_sync = (mvol->state & INTEL_ST_CLONE_MAN_SYNC) != 0;
1275 if (mvol->cng_master_disk < mmap->total_disks)
1276 pv->pv_cng_master_disk = mvol->cng_master_disk;
1277 vol->v_md_data = pv;
1278 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE;
1279 if (mmap->type == INTEL_T_RAID0)
1280 vol->v_raid_level = G_RAID_VOLUME_RL_RAID0;
1281 else if (mmap->type == INTEL_T_RAID1 &&
1282 mmap->total_domains >= 2 &&
1283 mmap->total_domains <= mmap->total_disks) {
1284 /* Assume total_domains is correct. */
1285 if (mmap->total_domains == mmap->total_disks)
1286 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1;
1288 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E;
1289 } else if (mmap->type == INTEL_T_RAID1) {
1290 /* total_domains looks wrong. */
1291 if (mmap->total_disks <= 2)
1292 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1;
1294 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E;
1295 } else if (mmap->type == INTEL_T_RAID5) {
1296 vol->v_raid_level = G_RAID_VOLUME_RL_RAID5;
1297 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_R5LA;
1299 vol->v_raid_level = G_RAID_VOLUME_RL_UNKNOWN;
1300 vol->v_strip_size = (u_int)mmap->strip_sectors * 512; //ZZZ
1301 vol->v_disks_count = mmap->total_disks;
1302 vol->v_mediasize = (off_t)mvol->total_sectors * 512; //ZZZ
1303 vol->v_sectorsize = 512; //ZZZ
1304 for (j = 0; j < vol->v_disks_count; j++) {
1305 sd = &vol->v_subdisks[j];
1306 sd->sd_offset = intel_get_map_offset(mmap) * 512; //ZZZ
1307 sd->sd_size = intel_get_map_disk_sectors(mmap) * 512; //ZZZ
1309 g_raid_start_volume(vol);
1312 /* Create disk placeholders to store data for later writing. */
1313 for (disk_pos = 0; disk_pos < meta->total_disks; disk_pos++) {
1314 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO);
1315 pd->pd_disk_pos = disk_pos;
1316 pd->pd_disk_meta = meta->disk[disk_pos];
1317 disk = g_raid_create_disk(sc);
1318 disk->d_md_data = (void *)pd;
1319 disk->d_state = G_RAID_DISK_S_OFFLINE;
1320 for (i = 0; i < meta->total_volumes; i++) {
1321 mvol = intel_get_volume(meta, i);
1322 mmap = intel_get_map(mvol, 0);
1323 for (j = 0; j < mmap->total_disks; j++) {
1324 if ((mmap->disk_idx[j] & INTEL_DI_IDX) == disk_pos)
1327 if (j == mmap->total_disks)
1329 vol = g_raid_md_intel_get_volume(sc, i);
1330 sd = &vol->v_subdisks[j];
1332 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1336 /* Make all disks found till the moment take their places. */
1338 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1339 if (disk->d_state == G_RAID_DISK_S_NONE) {
1340 g_raid_md_intel_start_disk(disk);
1344 } while (disk != NULL);
1346 mdi->mdio_started = 1;
1347 G_RAID_DEBUG1(0, sc, "Array started.");
1348 g_raid_md_write_intel(md, NULL, NULL, NULL);
1350 /* Pickup any STALE/SPARE disks to refill array if needed. */
1351 g_raid_md_intel_refill(sc);
1353 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
1354 g_raid_event_send(vol, G_RAID_VOLUME_E_START,
1355 G_RAID_EVENT_VOLUME);
1358 callout_stop(&mdi->mdio_start_co);
1359 G_RAID_DEBUG1(1, sc, "root_mount_rel %p", mdi->mdio_rootmount);
1360 root_mount_rel(mdi->mdio_rootmount);
1361 mdi->mdio_rootmount = NULL;
1365 g_raid_md_intel_new_disk(struct g_raid_disk *disk)
1367 struct g_raid_softc *sc;
1368 struct g_raid_md_object *md;
1369 struct g_raid_md_intel_object *mdi;
1370 struct intel_raid_conf *pdmeta;
1371 struct g_raid_md_intel_perdisk *pd;
1375 mdi = (struct g_raid_md_intel_object *)md;
1376 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
1377 pdmeta = pd->pd_meta;
1379 if (mdi->mdio_started) {
1380 if (g_raid_md_intel_start_disk(disk))
1381 g_raid_md_write_intel(md, NULL, NULL, NULL);
1383 /* If we haven't started yet - check metadata freshness. */
1384 if (mdi->mdio_meta == NULL ||
1385 ((int32_t)(pdmeta->generation - mdi->mdio_generation)) > 0) {
1386 G_RAID_DEBUG1(1, sc, "Newer disk");
1387 if (mdi->mdio_meta != NULL)
1388 free(mdi->mdio_meta, M_MD_INTEL);
1389 mdi->mdio_meta = intel_meta_copy(pdmeta);
1390 mdi->mdio_generation = mdi->mdio_meta->generation;
1391 mdi->mdio_disks_present = 1;
1392 } else if (pdmeta->generation == mdi->mdio_generation) {
1393 mdi->mdio_disks_present++;
1394 G_RAID_DEBUG1(1, sc, "Matching disk (%d of %d up)",
1395 mdi->mdio_disks_present,
1396 mdi->mdio_meta->total_disks);
1398 G_RAID_DEBUG1(1, sc, "Older disk");
1400 /* If we collected all needed disks - start array. */
1401 if (mdi->mdio_disks_present == mdi->mdio_meta->total_disks)
1402 g_raid_md_intel_start(sc);
1407 g_raid_intel_go(void *arg)
1409 struct g_raid_softc *sc;
1410 struct g_raid_md_object *md;
1411 struct g_raid_md_intel_object *mdi;
1415 mdi = (struct g_raid_md_intel_object *)md;
1416 if (!mdi->mdio_started) {
1417 G_RAID_DEBUG1(0, sc, "Force array start due to timeout.");
1418 g_raid_event_send(sc, G_RAID_NODE_E_START, 0);
1423 g_raid_md_create_intel(struct g_raid_md_object *md, struct g_class *mp,
1426 struct g_raid_softc *sc;
1427 struct g_raid_md_intel_object *mdi;
1430 mdi = (struct g_raid_md_intel_object *)md;
1431 mdi->mdio_config_id = mdi->mdio_orig_config_id = arc4random();
1432 mdi->mdio_generation = 0;
1433 snprintf(name, sizeof(name), "Intel-%08x", mdi->mdio_config_id);
1434 sc = g_raid_create_node(mp, name, md);
1436 return (G_RAID_MD_TASTE_FAIL);
1439 return (G_RAID_MD_TASTE_NEW);
1443 * Return the last N characters of the serial label. The Linux and
1444 * ataraid(7) code always uses the last 16 characters of the label to
1445 * store into the Intel meta format. Generalize this to N characters
1446 * since that's easy. Labels can be up to 20 characters for SATA drives
1447 * and up 251 characters for SAS drives. Since intel controllers don't
1448 * support SAS drives, just stick with the SATA limits for stack friendliness.
1451 g_raid_md_get_label(struct g_consumer *cp, char *serial, int serlen)
1453 char serial_buffer[24];
1456 len = sizeof(serial_buffer);
1457 error = g_io_getattr("GEOM::ident", cp, &len, serial_buffer);
1460 len = strlen(serial_buffer);
1465 strncpy(serial, serial_buffer + len, serlen);
1470 g_raid_md_taste_intel(struct g_raid_md_object *md, struct g_class *mp,
1471 struct g_consumer *cp, struct g_geom **gp)
1473 struct g_consumer *rcp;
1474 struct g_provider *pp;
1475 struct g_raid_md_intel_object *mdi, *mdi1;
1476 struct g_raid_softc *sc;
1477 struct g_raid_disk *disk;
1478 struct intel_raid_conf *meta;
1479 struct g_raid_md_intel_perdisk *pd;
1480 struct g_geom *geom;
1481 int error, disk_pos, result, spare, len;
1482 char serial[INTEL_SERIAL_LEN];
1486 G_RAID_DEBUG(1, "Tasting Intel on %s", cp->provider->name);
1487 mdi = (struct g_raid_md_intel_object *)md;
1490 /* Read metadata from device. */
1493 g_topology_unlock();
1494 error = g_raid_md_get_label(cp, serial, sizeof(serial));
1496 G_RAID_DEBUG(1, "Cannot get serial number from %s (error=%d).",
1501 len = sizeof(vendor);
1502 if (pp->geom->rank == 1)
1503 g_io_getattr("GEOM::hba_vendor", cp, &len, &vendor);
1504 meta = intel_meta_read(cp);
1507 if (g_raid_aggressive_spare) {
1508 if (vendor != 0x8086) {
1510 "Intel vendor mismatch 0x%04x != 0x8086",
1514 "No Intel metadata, forcing spare.");
1519 return (G_RAID_MD_TASTE_FAIL);
1522 /* Check this disk position in obtained metadata. */
1523 disk_pos = intel_meta_find_disk(meta, serial);
1525 G_RAID_DEBUG(1, "Intel serial '%s' not found", serial);
1528 if (intel_get_disk_sectors(&meta->disk[disk_pos]) !=
1529 (pp->mediasize / pp->sectorsize)) {
1530 G_RAID_DEBUG(1, "Intel size mismatch %ju != %ju",
1531 intel_get_disk_sectors(&meta->disk[disk_pos]),
1532 (off_t)(pp->mediasize / pp->sectorsize));
1536 G_RAID_DEBUG(1, "Intel disk position %d", disk_pos);
1537 spare = meta->disk[disk_pos].flags & INTEL_F_SPARE;
1540 /* Search for matching node. */
1543 LIST_FOREACH(geom, &mp->geom, geom) {
1547 if (sc->sc_stopping != 0)
1549 if (sc->sc_md->mdo_class != md->mdo_class)
1551 mdi1 = (struct g_raid_md_intel_object *)sc->sc_md;
1553 if (mdi1->mdio_incomplete)
1556 if (mdi1->mdio_config_id == meta->config_id)
1561 /* Found matching node. */
1563 G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name);
1564 result = G_RAID_MD_TASTE_EXISTING;
1566 } else if (spare) { /* Not found needy node -- left for later. */
1567 G_RAID_DEBUG(1, "Spare is not needed at this time");
1570 } else { /* Not found matching node -- create one. */
1571 result = G_RAID_MD_TASTE_NEW;
1572 mdi->mdio_config_id = meta->config_id;
1573 mdi->mdio_orig_config_id = meta->orig_config_id;
1574 snprintf(name, sizeof(name), "Intel-%08x", meta->config_id);
1575 sc = g_raid_create_node(mp, name, md);
1578 callout_init(&mdi->mdio_start_co, 1);
1579 callout_reset(&mdi->mdio_start_co, g_raid_start_timeout * hz,
1580 g_raid_intel_go, sc);
1581 mdi->mdio_rootmount = root_mount_hold("GRAID-Intel");
1582 G_RAID_DEBUG1(1, sc, "root_mount_hold %p", mdi->mdio_rootmount);
1585 /* There is no return after this point, so we close passed consumer. */
1586 g_access(cp, -1, 0, 0);
1588 rcp = g_new_consumer(geom);
1589 rcp->flags |= G_CF_DIRECT_RECEIVE;
1591 if (g_access(rcp, 1, 1, 1) != 0)
1594 g_topology_unlock();
1595 sx_xlock(&sc->sc_lock);
1597 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO);
1599 pd->pd_disk_pos = -1;
1601 memcpy(&pd->pd_disk_meta.serial[0], serial, INTEL_SERIAL_LEN);
1602 intel_set_disk_sectors(&pd->pd_disk_meta,
1603 pp->mediasize / pp->sectorsize);
1604 pd->pd_disk_meta.id = 0;
1605 pd->pd_disk_meta.flags = INTEL_F_SPARE;
1607 pd->pd_disk_meta = meta->disk[disk_pos];
1609 disk = g_raid_create_disk(sc);
1610 disk->d_md_data = (void *)pd;
1611 disk->d_consumer = rcp;
1612 rcp->private = disk;
1614 g_raid_get_disk_info(disk);
1616 g_raid_md_intel_new_disk(disk);
1618 sx_xunlock(&sc->sc_lock);
1625 free(meta, M_MD_INTEL);
1626 return (G_RAID_MD_TASTE_FAIL);
1630 g_raid_md_event_intel(struct g_raid_md_object *md,
1631 struct g_raid_disk *disk, u_int event)
1633 struct g_raid_softc *sc;
1634 struct g_raid_subdisk *sd;
1635 struct g_raid_md_intel_object *mdi;
1636 struct g_raid_md_intel_perdisk *pd;
1639 mdi = (struct g_raid_md_intel_object *)md;
1642 case G_RAID_NODE_E_START:
1643 if (!mdi->mdio_started)
1644 g_raid_md_intel_start(sc);
1649 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
1651 case G_RAID_DISK_E_DISCONNECTED:
1652 /* If disk was assigned, just update statuses. */
1653 if (pd->pd_disk_pos >= 0) {
1654 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1655 if (disk->d_consumer) {
1656 g_raid_kill_consumer(sc, disk->d_consumer);
1657 disk->d_consumer = NULL;
1659 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
1660 g_raid_change_subdisk_state(sd,
1661 G_RAID_SUBDISK_S_NONE);
1662 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
1663 G_RAID_EVENT_SUBDISK);
1666 /* Otherwise -- delete. */
1667 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
1668 g_raid_destroy_disk(disk);
1671 /* Write updated metadata to all disks. */
1672 g_raid_md_write_intel(md, NULL, NULL, NULL);
1674 /* Check if anything left except placeholders. */
1675 if (g_raid_ndisks(sc, -1) ==
1676 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1677 g_raid_destroy_node(sc, 0);
1679 g_raid_md_intel_refill(sc);
1686 g_raid_md_ctl_intel(struct g_raid_md_object *md,
1687 struct gctl_req *req)
1689 struct g_raid_softc *sc;
1690 struct g_raid_volume *vol, *vol1;
1691 struct g_raid_subdisk *sd;
1692 struct g_raid_disk *disk;
1693 struct g_raid_md_intel_object *mdi;
1694 struct g_raid_md_intel_pervolume *pv;
1695 struct g_raid_md_intel_perdisk *pd;
1696 struct g_consumer *cp;
1697 struct g_provider *pp;
1698 char arg[16], serial[INTEL_SERIAL_LEN];
1699 const char *nodename, *verb, *volname, *levelname, *diskname;
1702 off_t off, size, sectorsize, strip, disk_sectors;
1703 intmax_t *sizearg, *striparg;
1704 int numdisks, i, len, level, qual, update;
1708 mdi = (struct g_raid_md_intel_object *)md;
1709 verb = gctl_get_param(req, "verb", NULL);
1710 nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
1712 if (strcmp(verb, "label") == 0) {
1715 gctl_error(req, "Invalid number of arguments.");
1718 volname = gctl_get_asciiparam(req, "arg1");
1719 if (volname == NULL) {
1720 gctl_error(req, "No volume name.");
1723 levelname = gctl_get_asciiparam(req, "arg2");
1724 if (levelname == NULL) {
1725 gctl_error(req, "No RAID level.");
1728 if (strcasecmp(levelname, "RAID5") == 0)
1729 levelname = "RAID5-LA";
1730 if (g_raid_volume_str2level(levelname, &level, &qual)) {
1731 gctl_error(req, "Unknown RAID level '%s'.", levelname);
1734 numdisks = *nargs - 3;
1735 force = gctl_get_paraml(req, "force", sizeof(*force));
1736 if (!g_raid_md_intel_supported(level, qual, numdisks,
1737 force ? *force : 0)) {
1738 gctl_error(req, "Unsupported RAID level "
1739 "(0x%02x/0x%02x), or number of disks (%d).",
1740 level, qual, numdisks);
1744 /* Search for disks, connect them and probe. */
1745 size = 0x7fffffffffffffffllu;
1747 for (i = 0; i < numdisks; i++) {
1748 snprintf(arg, sizeof(arg), "arg%d", i + 3);
1749 diskname = gctl_get_asciiparam(req, arg);
1750 if (diskname == NULL) {
1751 gctl_error(req, "No disk name (%s).", arg);
1755 if (strcmp(diskname, "NONE") == 0) {
1760 cp = g_raid_open_consumer(sc, diskname);
1762 gctl_error(req, "Can't open disk '%s'.",
1764 g_topology_unlock();
1770 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO);
1771 pd->pd_disk_pos = i;
1772 disk = g_raid_create_disk(sc);
1773 disk->d_md_data = (void *)pd;
1774 disk->d_consumer = cp;
1776 strcpy(&pd->pd_disk_meta.serial[0], "NONE");
1777 pd->pd_disk_meta.id = 0xffffffff;
1778 pd->pd_disk_meta.flags = INTEL_F_ASSIGNED;
1782 g_topology_unlock();
1784 error = g_raid_md_get_label(cp,
1785 &pd->pd_disk_meta.serial[0], INTEL_SERIAL_LEN);
1788 "Can't get serial for provider '%s'.",
1794 g_raid_get_disk_info(disk);
1796 intel_set_disk_sectors(&pd->pd_disk_meta,
1797 pp->mediasize / pp->sectorsize);
1798 if (size > pp->mediasize)
1799 size = pp->mediasize;
1800 if (sectorsize < pp->sectorsize)
1801 sectorsize = pp->sectorsize;
1802 pd->pd_disk_meta.id = 0;
1803 pd->pd_disk_meta.flags = INTEL_F_ASSIGNED | INTEL_F_ONLINE;
1808 if (sectorsize <= 0) {
1809 gctl_error(req, "Can't get sector size.");
1813 /* Reserve some space for metadata. */
1814 size -= ((4096 + sectorsize - 1) / sectorsize) * sectorsize;
1816 /* Handle size argument. */
1817 len = sizeof(*sizearg);
1818 sizearg = gctl_get_param(req, "size", &len);
1819 if (sizearg != NULL && len == sizeof(*sizearg) &&
1821 if (*sizearg > size) {
1822 gctl_error(req, "Size too big %lld > %lld.",
1823 (long long)*sizearg, (long long)size);
1829 /* Handle strip argument. */
1831 len = sizeof(*striparg);
1832 striparg = gctl_get_param(req, "strip", &len);
1833 if (striparg != NULL && len == sizeof(*striparg) &&
1835 if (*striparg < sectorsize) {
1836 gctl_error(req, "Strip size too small.");
1839 if (*striparg % sectorsize != 0) {
1840 gctl_error(req, "Incorrect strip size.");
1843 if (strip > 65535 * sectorsize) {
1844 gctl_error(req, "Strip size too big.");
1850 /* Round size down to strip or sector. */
1851 if (level == G_RAID_VOLUME_RL_RAID1)
1852 size -= (size % sectorsize);
1853 else if (level == G_RAID_VOLUME_RL_RAID1E &&
1854 (numdisks & 1) != 0)
1855 size -= (size % (2 * strip));
1857 size -= (size % strip);
1859 gctl_error(req, "Size too small.");
1863 /* We have all we need, create things: volume, ... */
1864 mdi->mdio_started = 1;
1865 vol = g_raid_create_volume(sc, volname, -1);
1866 pv = malloc(sizeof(*pv), M_MD_INTEL, M_WAITOK | M_ZERO);
1867 pv->pv_volume_pos = 0;
1868 vol->v_md_data = pv;
1869 vol->v_raid_level = level;
1870 vol->v_raid_level_qualifier = qual;
1871 vol->v_strip_size = strip;
1872 vol->v_disks_count = numdisks;
1873 if (level == G_RAID_VOLUME_RL_RAID0)
1874 vol->v_mediasize = size * numdisks;
1875 else if (level == G_RAID_VOLUME_RL_RAID1)
1876 vol->v_mediasize = size;
1877 else if (level == G_RAID_VOLUME_RL_RAID5)
1878 vol->v_mediasize = size * (numdisks - 1);
1880 vol->v_mediasize = ((size * numdisks) / strip / 2) *
1883 vol->v_sectorsize = sectorsize;
1884 g_raid_start_volume(vol);
1886 /* , and subdisks. */
1887 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1888 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
1889 sd = &vol->v_subdisks[pd->pd_disk_pos];
1893 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1894 if (sd->sd_disk->d_consumer != NULL) {
1895 g_raid_change_disk_state(disk,
1896 G_RAID_DISK_S_ACTIVE);
1897 if (level == G_RAID_VOLUME_RL_RAID5)
1898 g_raid_change_subdisk_state(sd,
1899 G_RAID_SUBDISK_S_UNINITIALIZED);
1901 g_raid_change_subdisk_state(sd,
1902 G_RAID_SUBDISK_S_ACTIVE);
1903 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
1904 G_RAID_EVENT_SUBDISK);
1906 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1910 /* Write metadata based on created entities. */
1911 G_RAID_DEBUG1(0, sc, "Array started.");
1912 g_raid_md_write_intel(md, NULL, NULL, NULL);
1914 /* Pickup any STALE/SPARE disks to refill array if needed. */
1915 g_raid_md_intel_refill(sc);
1917 g_raid_event_send(vol, G_RAID_VOLUME_E_START,
1918 G_RAID_EVENT_VOLUME);
1921 if (strcmp(verb, "add") == 0) {
1924 gctl_error(req, "Invalid number of arguments.");
1927 volname = gctl_get_asciiparam(req, "arg1");
1928 if (volname == NULL) {
1929 gctl_error(req, "No volume name.");
1932 levelname = gctl_get_asciiparam(req, "arg2");
1933 if (levelname == NULL) {
1934 gctl_error(req, "No RAID level.");
1937 if (strcasecmp(levelname, "RAID5") == 0)
1938 levelname = "RAID5-LA";
1939 if (g_raid_volume_str2level(levelname, &level, &qual)) {
1940 gctl_error(req, "Unknown RAID level '%s'.", levelname);
1944 /* Look for existing volumes. */
1947 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
1952 gctl_error(req, "Maximum two volumes supported.");
1956 gctl_error(req, "At least one volume must exist.");
1960 numdisks = vol1->v_disks_count;
1961 force = gctl_get_paraml(req, "force", sizeof(*force));
1962 if (!g_raid_md_intel_supported(level, qual, numdisks,
1963 force ? *force : 0)) {
1964 gctl_error(req, "Unsupported RAID level "
1965 "(0x%02x/0x%02x), or number of disks (%d).",
1966 level, qual, numdisks);
1970 /* Collect info about present disks. */
1971 size = 0x7fffffffffffffffllu;
1973 for (i = 0; i < numdisks; i++) {
1974 disk = vol1->v_subdisks[i].sd_disk;
1975 pd = (struct g_raid_md_intel_perdisk *)
1978 intel_get_disk_sectors(&pd->pd_disk_meta);
1980 if (disk_sectors * 512 < size)
1981 size = disk_sectors * 512;
1982 if (disk->d_consumer != NULL &&
1983 disk->d_consumer->provider != NULL &&
1984 disk->d_consumer->provider->sectorsize >
1987 disk->d_consumer->provider->sectorsize;
1991 /* Reserve some space for metadata. */
1992 size -= ((4096 + sectorsize - 1) / sectorsize) * sectorsize;
1994 /* Decide insert before or after. */
1995 sd = &vol1->v_subdisks[0];
1997 size - (sd->sd_offset + sd->sd_size)) {
1999 size = sd->sd_offset;
2001 off = sd->sd_offset + sd->sd_size;
2002 size = size - (sd->sd_offset + sd->sd_size);
2005 /* Handle strip argument. */
2007 len = sizeof(*striparg);
2008 striparg = gctl_get_param(req, "strip", &len);
2009 if (striparg != NULL && len == sizeof(*striparg) &&
2011 if (*striparg < sectorsize) {
2012 gctl_error(req, "Strip size too small.");
2015 if (*striparg % sectorsize != 0) {
2016 gctl_error(req, "Incorrect strip size.");
2019 if (strip > 65535 * sectorsize) {
2020 gctl_error(req, "Strip size too big.");
2026 /* Round offset up to strip. */
2027 if (off % strip != 0) {
2028 size -= strip - off % strip;
2029 off += strip - off % strip;
2032 /* Handle size argument. */
2033 len = sizeof(*sizearg);
2034 sizearg = gctl_get_param(req, "size", &len);
2035 if (sizearg != NULL && len == sizeof(*sizearg) &&
2037 if (*sizearg > size) {
2038 gctl_error(req, "Size too big %lld > %lld.",
2039 (long long)*sizearg, (long long)size);
2045 /* Round size down to strip or sector. */
2046 if (level == G_RAID_VOLUME_RL_RAID1)
2047 size -= (size % sectorsize);
2049 size -= (size % strip);
2051 gctl_error(req, "Size too small.");
2054 if (size > 0xffffffffllu * sectorsize) {
2055 gctl_error(req, "Size too big.");
2059 /* We have all we need, create things: volume, ... */
2060 vol = g_raid_create_volume(sc, volname, -1);
2061 pv = malloc(sizeof(*pv), M_MD_INTEL, M_WAITOK | M_ZERO);
2062 pv->pv_volume_pos = i;
2063 vol->v_md_data = pv;
2064 vol->v_raid_level = level;
2065 vol->v_raid_level_qualifier = qual;
2066 vol->v_strip_size = strip;
2067 vol->v_disks_count = numdisks;
2068 if (level == G_RAID_VOLUME_RL_RAID0)
2069 vol->v_mediasize = size * numdisks;
2070 else if (level == G_RAID_VOLUME_RL_RAID1)
2071 vol->v_mediasize = size;
2072 else if (level == G_RAID_VOLUME_RL_RAID5)
2073 vol->v_mediasize = size * (numdisks - 1);
2075 vol->v_mediasize = ((size * numdisks) / strip / 2) *
2078 vol->v_sectorsize = sectorsize;
2079 g_raid_start_volume(vol);
2081 /* , and subdisks. */
2082 for (i = 0; i < numdisks; i++) {
2083 disk = vol1->v_subdisks[i].sd_disk;
2084 sd = &vol->v_subdisks[i];
2086 sd->sd_offset = off;
2088 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
2089 if (disk->d_state == G_RAID_DISK_S_ACTIVE) {
2090 if (level == G_RAID_VOLUME_RL_RAID5)
2091 g_raid_change_subdisk_state(sd,
2092 G_RAID_SUBDISK_S_UNINITIALIZED);
2094 g_raid_change_subdisk_state(sd,
2095 G_RAID_SUBDISK_S_ACTIVE);
2096 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
2097 G_RAID_EVENT_SUBDISK);
2101 /* Write metadata based on created entities. */
2102 g_raid_md_write_intel(md, NULL, NULL, NULL);
2104 g_raid_event_send(vol, G_RAID_VOLUME_E_START,
2105 G_RAID_EVENT_VOLUME);
2108 if (strcmp(verb, "delete") == 0) {
2110 nodename = gctl_get_asciiparam(req, "arg0");
2111 if (nodename != NULL && strcasecmp(sc->sc_name, nodename) != 0)
2114 /* Full node destruction. */
2115 if (*nargs == 1 && nodename != NULL) {
2116 /* Check if some volume is still open. */
2117 force = gctl_get_paraml(req, "force", sizeof(*force));
2118 if (force != NULL && *force == 0 &&
2119 g_raid_nopens(sc) != 0) {
2120 gctl_error(req, "Some volume is still open.");
2124 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2125 if (disk->d_consumer)
2126 intel_meta_erase(disk->d_consumer);
2128 g_raid_destroy_node(sc, 0);
2132 /* Destroy specified volume. If it was last - all node. */
2134 gctl_error(req, "Invalid number of arguments.");
2137 volname = gctl_get_asciiparam(req,
2138 nodename != NULL ? "arg1" : "arg0");
2139 if (volname == NULL) {
2140 gctl_error(req, "No volume name.");
2144 /* Search for volume. */
2145 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2146 if (strcmp(vol->v_name, volname) == 0)
2148 pp = vol->v_provider;
2151 if (strcmp(pp->name, volname) == 0)
2153 if (strncmp(pp->name, "raid/", 5) == 0 &&
2154 strcmp(pp->name + 5, volname) == 0)
2158 i = strtol(volname, &tmp, 10);
2159 if (verb != volname && tmp[0] == 0) {
2160 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2161 if (vol->v_global_id == i)
2167 gctl_error(req, "Volume '%s' not found.", volname);
2171 /* Check if volume is still open. */
2172 force = gctl_get_paraml(req, "force", sizeof(*force));
2173 if (force != NULL && *force == 0 &&
2174 vol->v_provider_open != 0) {
2175 gctl_error(req, "Volume is still open.");
2179 /* Destroy volume and potentially node. */
2181 TAILQ_FOREACH(vol1, &sc->sc_volumes, v_next)
2184 g_raid_destroy_volume(vol);
2185 g_raid_md_write_intel(md, NULL, NULL, NULL);
2187 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2188 if (disk->d_consumer)
2189 intel_meta_erase(disk->d_consumer);
2191 g_raid_destroy_node(sc, 0);
2195 if (strcmp(verb, "remove") == 0 ||
2196 strcmp(verb, "fail") == 0) {
2198 gctl_error(req, "Invalid number of arguments.");
2201 for (i = 1; i < *nargs; i++) {
2202 snprintf(arg, sizeof(arg), "arg%d", i);
2203 diskname = gctl_get_asciiparam(req, arg);
2204 if (diskname == NULL) {
2205 gctl_error(req, "No disk name (%s).", arg);
2209 if (strncmp(diskname, "/dev/", 5) == 0)
2212 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2213 if (disk->d_consumer != NULL &&
2214 disk->d_consumer->provider != NULL &&
2215 strcmp(disk->d_consumer->provider->name,
2220 gctl_error(req, "Disk '%s' not found.",
2226 if (strcmp(verb, "fail") == 0) {
2227 g_raid_md_fail_disk_intel(md, NULL, disk);
2231 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2233 /* Erase metadata on deleting disk. */
2234 intel_meta_erase(disk->d_consumer);
2236 /* If disk was assigned, just update statuses. */
2237 if (pd->pd_disk_pos >= 0) {
2238 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
2239 g_raid_kill_consumer(sc, disk->d_consumer);
2240 disk->d_consumer = NULL;
2241 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
2242 g_raid_change_subdisk_state(sd,
2243 G_RAID_SUBDISK_S_NONE);
2244 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
2245 G_RAID_EVENT_SUBDISK);
2248 /* Otherwise -- delete. */
2249 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
2250 g_raid_destroy_disk(disk);
2254 /* Write updated metadata to remaining disks. */
2255 g_raid_md_write_intel(md, NULL, NULL, NULL);
2257 /* Check if anything left except placeholders. */
2258 if (g_raid_ndisks(sc, -1) ==
2259 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
2260 g_raid_destroy_node(sc, 0);
2262 g_raid_md_intel_refill(sc);
2265 if (strcmp(verb, "insert") == 0) {
2267 gctl_error(req, "Invalid number of arguments.");
2271 for (i = 1; i < *nargs; i++) {
2272 /* Get disk name. */
2273 snprintf(arg, sizeof(arg), "arg%d", i);
2274 diskname = gctl_get_asciiparam(req, arg);
2275 if (diskname == NULL) {
2276 gctl_error(req, "No disk name (%s).", arg);
2281 /* Try to find provider with specified name. */
2283 cp = g_raid_open_consumer(sc, diskname);
2285 gctl_error(req, "Can't open disk '%s'.",
2287 g_topology_unlock();
2292 g_topology_unlock();
2294 /* Read disk serial. */
2295 error = g_raid_md_get_label(cp,
2296 &serial[0], INTEL_SERIAL_LEN);
2299 "Can't get serial for provider '%s'.",
2301 g_raid_kill_consumer(sc, cp);
2306 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO);
2307 pd->pd_disk_pos = -1;
2309 disk = g_raid_create_disk(sc);
2310 disk->d_consumer = cp;
2311 disk->d_md_data = (void *)pd;
2314 g_raid_get_disk_info(disk);
2316 memcpy(&pd->pd_disk_meta.serial[0], &serial[0],
2318 intel_set_disk_sectors(&pd->pd_disk_meta,
2319 pp->mediasize / pp->sectorsize);
2320 pd->pd_disk_meta.id = 0;
2321 pd->pd_disk_meta.flags = INTEL_F_SPARE;
2323 /* Welcome the "new" disk. */
2324 update += g_raid_md_intel_start_disk(disk);
2325 if (disk->d_state == G_RAID_DISK_S_SPARE) {
2326 intel_meta_write_spare(cp, &pd->pd_disk_meta);
2327 g_raid_destroy_disk(disk);
2328 } else if (disk->d_state != G_RAID_DISK_S_ACTIVE) {
2329 gctl_error(req, "Disk '%s' doesn't fit.",
2331 g_raid_destroy_disk(disk);
2337 /* Write new metadata if we changed something. */
2339 g_raid_md_write_intel(md, NULL, NULL, NULL);
2346 g_raid_md_write_intel(struct g_raid_md_object *md, struct g_raid_volume *tvol,
2347 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
2349 struct g_raid_softc *sc;
2350 struct g_raid_volume *vol;
2351 struct g_raid_subdisk *sd;
2352 struct g_raid_disk *disk;
2353 struct g_raid_md_intel_object *mdi;
2354 struct g_raid_md_intel_pervolume *pv;
2355 struct g_raid_md_intel_perdisk *pd;
2356 struct intel_raid_conf *meta;
2357 struct intel_raid_vol *mvol;
2358 struct intel_raid_map *mmap0, *mmap1;
2359 off_t sectorsize = 512, pos;
2360 const char *version, *cv;
2361 int vi, sdi, numdisks, len, state, stale;
2364 mdi = (struct g_raid_md_intel_object *)md;
2366 if (sc->sc_stopping == G_RAID_DESTROY_HARD)
2369 /* Bump generation. Newly written metadata may differ from previous. */
2370 mdi->mdio_generation++;
2372 /* Count number of disks. */
2374 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2375 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2376 if (pd->pd_disk_pos < 0)
2379 if (disk->d_state == G_RAID_DISK_S_ACTIVE) {
2380 pd->pd_disk_meta.flags =
2381 INTEL_F_ONLINE | INTEL_F_ASSIGNED;
2382 } else if (disk->d_state == G_RAID_DISK_S_FAILED) {
2383 pd->pd_disk_meta.flags = INTEL_F_FAILED |
2385 } else if (disk->d_state == G_RAID_DISK_S_DISABLED) {
2386 pd->pd_disk_meta.flags = INTEL_F_FAILED |
2387 INTEL_F_ASSIGNED | INTEL_F_DISABLED;
2389 if (!(pd->pd_disk_meta.flags & INTEL_F_DISABLED))
2390 pd->pd_disk_meta.flags = INTEL_F_ASSIGNED;
2391 if (pd->pd_disk_meta.id != 0xffffffff) {
2392 pd->pd_disk_meta.id = 0xffffffff;
2393 len = strlen(pd->pd_disk_meta.serial);
2394 len = min(len, INTEL_SERIAL_LEN - 3);
2395 strcpy(pd->pd_disk_meta.serial + len, ":0");
2400 /* Fill anchor and disks. */
2401 meta = malloc(INTEL_MAX_MD_SIZE(numdisks),
2402 M_MD_INTEL, M_WAITOK | M_ZERO);
2403 memcpy(&meta->intel_id[0], INTEL_MAGIC, sizeof(INTEL_MAGIC) - 1);
2404 meta->config_size = INTEL_MAX_MD_SIZE(numdisks);
2405 meta->config_id = mdi->mdio_config_id;
2406 meta->orig_config_id = mdi->mdio_orig_config_id;
2407 meta->generation = mdi->mdio_generation;
2408 meta->attributes = INTEL_ATTR_CHECKSUM;
2409 meta->total_disks = numdisks;
2410 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2411 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2412 if (pd->pd_disk_pos < 0)
2414 meta->disk[pd->pd_disk_pos] = pd->pd_disk_meta;
2415 if (pd->pd_disk_meta.sectors_hi != 0)
2416 meta->attributes |= INTEL_ATTR_2TB_DISK;
2419 /* Fill volumes and maps. */
2421 version = INTEL_VERSION_1000;
2422 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2423 pv = vol->v_md_data;
2424 if (vol->v_stopping)
2426 mvol = intel_get_volume(meta, vi);
2428 /* New metadata may have different volumes order. */
2429 pv->pv_volume_pos = vi;
2431 for (sdi = 0; sdi < vol->v_disks_count; sdi++) {
2432 sd = &vol->v_subdisks[sdi];
2433 if (sd->sd_disk != NULL)
2436 if (sdi >= vol->v_disks_count)
2437 panic("No any filled subdisk in volume");
2438 if (vol->v_mediasize >= 0x20000000000llu)
2439 meta->attributes |= INTEL_ATTR_2TB;
2440 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0)
2441 meta->attributes |= INTEL_ATTR_RAID0;
2442 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
2443 meta->attributes |= INTEL_ATTR_RAID1;
2444 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID5)
2445 meta->attributes |= INTEL_ATTR_RAID5;
2446 else if ((vol->v_disks_count & 1) == 0)
2447 meta->attributes |= INTEL_ATTR_RAID10;
2449 meta->attributes |= INTEL_ATTR_RAID1E;
2451 meta->attributes |= INTEL_ATTR_RAIDCNG;
2452 if (vol->v_strip_size > 131072)
2453 meta->attributes |= INTEL_ATTR_EXT_STRIP;
2456 cv = INTEL_VERSION_1206;
2457 else if (vol->v_disks_count > 4)
2458 cv = INTEL_VERSION_1204;
2459 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID5)
2460 cv = INTEL_VERSION_1202;
2461 else if (vol->v_disks_count > 2)
2462 cv = INTEL_VERSION_1201;
2463 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
2464 cv = INTEL_VERSION_1100;
2466 cv = INTEL_VERSION_1000;
2467 if (strcmp(cv, version) > 0)
2470 strlcpy(&mvol->name[0], vol->v_name, sizeof(mvol->name));
2471 mvol->total_sectors = vol->v_mediasize / sectorsize;
2472 mvol->state = (INTEL_ST_READ_COALESCING |
2473 INTEL_ST_WRITE_COALESCING);
2474 mvol->tid = vol->v_global_id + 1;
2476 mvol->state |= INTEL_ST_CLONE_N_GO;
2477 if (pv->pv_cng_man_sync)
2478 mvol->state |= INTEL_ST_CLONE_MAN_SYNC;
2479 mvol->cng_master_disk = pv->pv_cng_master_disk;
2480 if (vol->v_subdisks[pv->pv_cng_master_disk].sd_state ==
2481 G_RAID_SUBDISK_S_NONE)
2482 mvol->cng_state = INTEL_CNGST_MASTER_MISSING;
2483 else if (vol->v_state != G_RAID_VOLUME_S_OPTIMAL)
2484 mvol->cng_state = INTEL_CNGST_NEEDS_UPDATE;
2486 mvol->cng_state = INTEL_CNGST_UPDATED;
2489 /* Check for any recovery in progress. */
2490 state = G_RAID_SUBDISK_S_ACTIVE;
2491 pos = 0x7fffffffffffffffllu;
2493 for (sdi = 0; sdi < vol->v_disks_count; sdi++) {
2494 sd = &vol->v_subdisks[sdi];
2495 if (sd->sd_state == G_RAID_SUBDISK_S_REBUILD)
2496 state = G_RAID_SUBDISK_S_REBUILD;
2497 else if (sd->sd_state == G_RAID_SUBDISK_S_RESYNC &&
2498 state != G_RAID_SUBDISK_S_REBUILD)
2499 state = G_RAID_SUBDISK_S_RESYNC;
2500 else if (sd->sd_state == G_RAID_SUBDISK_S_STALE)
2502 if ((sd->sd_state == G_RAID_SUBDISK_S_REBUILD ||
2503 sd->sd_state == G_RAID_SUBDISK_S_RESYNC) &&
2504 sd->sd_rebuild_pos < pos)
2505 pos = sd->sd_rebuild_pos;
2507 if (state == G_RAID_SUBDISK_S_REBUILD) {
2508 mvol->migr_state = 1;
2509 mvol->migr_type = INTEL_MT_REBUILD;
2510 } else if (state == G_RAID_SUBDISK_S_RESYNC) {
2511 mvol->migr_state = 1;
2512 /* mvol->migr_type = INTEL_MT_REPAIR; */
2513 mvol->migr_type = INTEL_MT_VERIFY;
2514 mvol->state |= INTEL_ST_VERIFY_AND_FIX;
2516 mvol->migr_state = 0;
2517 mvol->dirty = (vol->v_dirty || stale);
2519 mmap0 = intel_get_map(mvol, 0);
2521 /* Write map / common part of two maps. */
2522 intel_set_map_offset(mmap0, sd->sd_offset / sectorsize);
2523 intel_set_map_disk_sectors(mmap0, sd->sd_size / sectorsize);
2524 mmap0->strip_sectors = vol->v_strip_size / sectorsize;
2525 if (vol->v_state == G_RAID_VOLUME_S_BROKEN)
2526 mmap0->status = INTEL_S_FAILURE;
2527 else if (vol->v_state == G_RAID_VOLUME_S_DEGRADED)
2528 mmap0->status = INTEL_S_DEGRADED;
2529 else if (g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_UNINITIALIZED)
2530 == g_raid_nsubdisks(vol, -1))
2531 mmap0->status = INTEL_S_UNINITIALIZED;
2533 mmap0->status = INTEL_S_READY;
2534 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0)
2535 mmap0->type = INTEL_T_RAID0;
2536 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1 ||
2537 vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E)
2538 mmap0->type = INTEL_T_RAID1;
2540 mmap0->type = INTEL_T_RAID5;
2541 mmap0->total_disks = vol->v_disks_count;
2542 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
2543 mmap0->total_domains = vol->v_disks_count;
2544 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E)
2545 mmap0->total_domains = 2;
2547 mmap0->total_domains = 1;
2548 intel_set_map_stripe_count(mmap0,
2549 sd->sd_size / vol->v_strip_size / mmap0->total_domains);
2550 mmap0->failed_disk_num = 0xff;
2553 /* If there are two maps - copy common and update. */
2554 if (mvol->migr_state) {
2555 intel_set_vol_curr_migr_unit(mvol,
2556 pos / vol->v_strip_size / mmap0->total_domains);
2557 mmap1 = intel_get_map(mvol, 1);
2558 memcpy(mmap1, mmap0, sizeof(struct intel_raid_map));
2559 mmap0->status = INTEL_S_READY;
2563 /* Write disk indexes and put rebuild flags. */
2564 for (sdi = 0; sdi < vol->v_disks_count; sdi++) {
2565 sd = &vol->v_subdisks[sdi];
2566 pd = (struct g_raid_md_intel_perdisk *)
2567 sd->sd_disk->d_md_data;
2568 mmap0->disk_idx[sdi] = pd->pd_disk_pos;
2569 if (mvol->migr_state)
2570 mmap1->disk_idx[sdi] = pd->pd_disk_pos;
2571 if (sd->sd_state == G_RAID_SUBDISK_S_REBUILD ||
2572 sd->sd_state == G_RAID_SUBDISK_S_RESYNC) {
2573 mmap1->disk_idx[sdi] |= INTEL_DI_RBLD;
2574 } else if (sd->sd_state != G_RAID_SUBDISK_S_ACTIVE &&
2575 sd->sd_state != G_RAID_SUBDISK_S_STALE &&
2576 sd->sd_state != G_RAID_SUBDISK_S_UNINITIALIZED) {
2577 mmap0->disk_idx[sdi] |= INTEL_DI_RBLD;
2578 if (mvol->migr_state)
2579 mmap1->disk_idx[sdi] |= INTEL_DI_RBLD;
2581 if ((sd->sd_state == G_RAID_SUBDISK_S_NONE ||
2582 sd->sd_state == G_RAID_SUBDISK_S_FAILED ||
2583 sd->sd_state == G_RAID_SUBDISK_S_REBUILD) &&
2584 mmap0->failed_disk_num == 0xff) {
2585 mmap0->failed_disk_num = sdi;
2586 if (mvol->migr_state)
2587 mmap1->failed_disk_num = sdi;
2592 meta->total_volumes = vi;
2593 if (vi > 1 || meta->attributes &
2594 (INTEL_ATTR_EXT_STRIP | INTEL_ATTR_2TB_DISK | INTEL_ATTR_2TB))
2595 version = INTEL_VERSION_1300;
2596 if (strcmp(version, INTEL_VERSION_1300) < 0)
2597 meta->attributes &= INTEL_ATTR_CHECKSUM;
2598 memcpy(&meta->version[0], version, sizeof(INTEL_VERSION_1000) - 1);
2600 /* We are done. Print meta data and store them to disks. */
2601 g_raid_md_intel_print(meta);
2602 if (mdi->mdio_meta != NULL)
2603 free(mdi->mdio_meta, M_MD_INTEL);
2604 mdi->mdio_meta = meta;
2605 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2606 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2607 if (disk->d_state != G_RAID_DISK_S_ACTIVE)
2609 if (pd->pd_meta != NULL) {
2610 free(pd->pd_meta, M_MD_INTEL);
2613 pd->pd_meta = intel_meta_copy(meta);
2614 intel_meta_write(disk->d_consumer, meta);
2620 g_raid_md_fail_disk_intel(struct g_raid_md_object *md,
2621 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
2623 struct g_raid_softc *sc;
2624 struct g_raid_md_intel_object *mdi;
2625 struct g_raid_md_intel_perdisk *pd;
2626 struct g_raid_subdisk *sd;
2629 mdi = (struct g_raid_md_intel_object *)md;
2630 pd = (struct g_raid_md_intel_perdisk *)tdisk->d_md_data;
2632 /* We can't fail disk that is not a part of array now. */
2633 if (pd->pd_disk_pos < 0)
2637 * Mark disk as failed in metadata and try to write that metadata
2638 * to the disk itself to prevent it's later resurrection as STALE.
2640 mdi->mdio_meta->disk[pd->pd_disk_pos].flags = INTEL_F_FAILED;
2641 pd->pd_disk_meta.flags = INTEL_F_FAILED;
2642 g_raid_md_intel_print(mdi->mdio_meta);
2643 if (tdisk->d_consumer)
2644 intel_meta_write(tdisk->d_consumer, mdi->mdio_meta);
2646 /* Change states. */
2647 g_raid_change_disk_state(tdisk, G_RAID_DISK_S_FAILED);
2648 TAILQ_FOREACH(sd, &tdisk->d_subdisks, sd_next) {
2649 g_raid_change_subdisk_state(sd,
2650 G_RAID_SUBDISK_S_FAILED);
2651 g_raid_event_send(sd, G_RAID_SUBDISK_E_FAILED,
2652 G_RAID_EVENT_SUBDISK);
2655 /* Write updated metadata to remaining disks. */
2656 g_raid_md_write_intel(md, NULL, NULL, tdisk);
2658 /* Check if anything left except placeholders. */
2659 if (g_raid_ndisks(sc, -1) ==
2660 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
2661 g_raid_destroy_node(sc, 0);
2663 g_raid_md_intel_refill(sc);
2668 g_raid_md_free_disk_intel(struct g_raid_md_object *md,
2669 struct g_raid_disk *disk)
2671 struct g_raid_md_intel_perdisk *pd;
2673 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2674 if (pd->pd_meta != NULL) {
2675 free(pd->pd_meta, M_MD_INTEL);
2678 free(pd, M_MD_INTEL);
2679 disk->d_md_data = NULL;
2684 g_raid_md_free_volume_intel(struct g_raid_md_object *md,
2685 struct g_raid_volume *vol)
2687 struct g_raid_md_intel_pervolume *pv;
2689 pv = (struct g_raid_md_intel_pervolume *)vol->v_md_data;
2690 free(pv, M_MD_INTEL);
2691 vol->v_md_data = NULL;
2696 g_raid_md_free_intel(struct g_raid_md_object *md)
2698 struct g_raid_md_intel_object *mdi;
2700 mdi = (struct g_raid_md_intel_object *)md;
2701 if (!mdi->mdio_started) {
2702 mdi->mdio_started = 0;
2703 callout_stop(&mdi->mdio_start_co);
2704 G_RAID_DEBUG1(1, md->mdo_softc,
2705 "root_mount_rel %p", mdi->mdio_rootmount);
2706 root_mount_rel(mdi->mdio_rootmount);
2707 mdi->mdio_rootmount = NULL;
2709 if (mdi->mdio_meta != NULL) {
2710 free(mdi->mdio_meta, M_MD_INTEL);
2711 mdi->mdio_meta = NULL;
2716 G_RAID_MD_DECLARE(intel, "Intel");