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>
45 #include <geom/geom.h>
46 #include "geom/raid/g_raid.h"
47 #include "g_raid_md_if.h"
49 static MALLOC_DEFINE(M_MD_INTEL, "md_intel_data", "GEOM_RAID Intel metadata");
51 struct intel_raid_map {
53 uint32_t disk_sectors;
54 uint32_t stripe_count;
55 uint16_t strip_sectors;
57 #define INTEL_S_READY 0x00
58 #define INTEL_S_UNINITIALIZED 0x01
59 #define INTEL_S_DEGRADED 0x02
60 #define INTEL_S_FAILURE 0x03
63 #define INTEL_T_RAID0 0x00
64 #define INTEL_T_RAID1 0x01
65 #define INTEL_T_RAID5 0x05
68 uint8_t total_domains;
69 uint8_t failed_disk_num;
72 uint32_t disk_sectors_hi;
73 uint32_t stripe_count_hi;
75 uint32_t disk_idx[1]; /* total_disks entries. */
76 #define INTEL_DI_IDX 0x00ffffff
77 #define INTEL_DI_RBLD 0x01000000
80 struct intel_raid_vol {
82 u_int64_t total_sectors __packed;
84 #define INTEL_ST_BOOTABLE 0x00000001
85 #define INTEL_ST_BOOT_DEVICE 0x00000002
86 #define INTEL_ST_READ_COALESCING 0x00000004
87 #define INTEL_ST_WRITE_COALESCING 0x00000008
88 #define INTEL_ST_LAST_SHUTDOWN_DIRTY 0x00000010
89 #define INTEL_ST_HIDDEN_AT_BOOT 0x00000020
90 #define INTEL_ST_CURRENTLY_HIDDEN 0x00000040
91 #define INTEL_ST_VERIFY_AND_FIX 0x00000080
92 #define INTEL_ST_MAP_STATE_UNINIT 0x00000100
93 #define INTEL_ST_NO_AUTO_RECOVERY 0x00000200
94 #define INTEL_ST_CLONE_N_GO 0x00000400
95 #define INTEL_ST_CLONE_MAN_SYNC 0x00000800
96 #define INTEL_ST_CNG_MASTER_DISK_NUM 0x00001000
98 uint8_t migr_priority;
101 uint8_t cng_master_disk;
102 uint16_t cache_policy;
104 #define INTEL_CNGST_UPDATED 0
105 #define INTEL_CNGST_NEEDS_UPDATE 1
106 #define INTEL_CNGST_MASTER_MISSING 2
107 uint8_t cng_sub_state;
108 uint32_t filler_0[10];
110 uint32_t curr_migr_unit;
111 uint32_t checkpoint_id;
114 #define INTEL_MT_INIT 0
115 #define INTEL_MT_REBUILD 1
116 #define INTEL_MT_VERIFY 2
117 #define INTEL_MT_GEN_MIGR 3
118 #define INTEL_MT_STATE_CHANGE 4
119 #define INTEL_MT_REPAIR 5
122 uint16_t verify_errors;
124 uint32_t curr_migr_unit_hi;
125 uint32_t filler_1[3];
126 struct intel_raid_map map[1]; /* 2 entries if migr_state != 0. */
129 struct intel_raid_disk {
130 #define INTEL_SERIAL_LEN 16
131 uint8_t serial[INTEL_SERIAL_LEN];
135 #define INTEL_F_SPARE 0x01
136 #define INTEL_F_ASSIGNED 0x02
137 #define INTEL_F_FAILED 0x04
138 #define INTEL_F_ONLINE 0x08
139 #define INTEL_F_DISABLED 0x80
140 uint32_t owner_cfg_num;
145 struct intel_raid_conf {
146 uint8_t intel_id[24];
147 #define INTEL_MAGIC "Intel Raid ISM Cfg Sig. "
150 #define INTEL_VERSION_1000 "1.0.00" /* RAID0 */
151 #define INTEL_VERSION_1100 "1.1.00" /* RAID1 */
152 #define INTEL_VERSION_1200 "1.2.00" /* Many volumes */
153 #define INTEL_VERSION_1201 "1.2.01" /* 3 or 4 disks */
154 #define INTEL_VERSION_1202 "1.2.02" /* RAID5 */
155 #define INTEL_VERSION_1204 "1.2.04" /* 5 or 6 disks */
156 #define INTEL_VERSION_1206 "1.2.06" /* CNG */
157 #define INTEL_VERSION_1300 "1.3.00" /* Attributes */
161 uint32_t config_size;
164 uint32_t error_log_size;
166 #define INTEL_ATTR_RAID0 0x00000001
167 #define INTEL_ATTR_RAID1 0x00000002
168 #define INTEL_ATTR_RAID10 0x00000004
169 #define INTEL_ATTR_RAID1E 0x00000008
170 #define INTEL_ATTR_RAID5 0x00000010
171 #define INTEL_ATTR_RAIDCNG 0x00000020
172 #define INTEL_ATTR_EXT_STRIP 0x00000040
173 #define INTEL_ATTR_NVM_CACHE 0x02000000
174 #define INTEL_ATTR_2TB_DISK 0x04000000
175 #define INTEL_ATTR_BBM 0x08000000
176 #define INTEL_ATTR_NVM_CACHE2 0x10000000
177 #define INTEL_ATTR_2TB 0x20000000
178 #define INTEL_ATTR_PM 0x40000000
179 #define INTEL_ATTR_CHECKSUM 0x80000000
182 uint8_t total_volumes;
183 uint8_t error_log_pos;
186 uint32_t orig_config_id;
187 uint32_t pwr_cycle_count;
188 uint32_t bbm_log_size;
189 uint32_t filler_0[35];
190 struct intel_raid_disk disk[1]; /* total_disks entries. */
191 /* Here goes total_volumes of struct intel_raid_vol. */
194 #define INTEL_ATTR_SUPPORTED ( INTEL_ATTR_RAID0 | INTEL_ATTR_RAID1 | \
195 INTEL_ATTR_RAID10 | INTEL_ATTR_RAID1E | INTEL_ATTR_RAID5 | \
196 INTEL_ATTR_RAIDCNG | INTEL_ATTR_EXT_STRIP | INTEL_ATTR_2TB_DISK | \
197 INTEL_ATTR_2TB | INTEL_ATTR_PM | INTEL_ATTR_CHECKSUM )
199 #define INTEL_MAX_MD_SIZE(ndisks) \
200 (sizeof(struct intel_raid_conf) + \
201 sizeof(struct intel_raid_disk) * (ndisks - 1) + \
202 sizeof(struct intel_raid_vol) * 2 + \
203 sizeof(struct intel_raid_map) * 2 + \
204 sizeof(uint32_t) * (ndisks - 1) * 4)
206 struct g_raid_md_intel_perdisk {
207 struct intel_raid_conf *pd_meta;
209 struct intel_raid_disk pd_disk_meta;
212 struct g_raid_md_intel_pervolume {
216 int pv_cng_master_disk;
219 struct g_raid_md_intel_object {
220 struct g_raid_md_object mdio_base;
221 uint32_t mdio_config_id;
222 uint32_t mdio_orig_config_id;
223 uint32_t mdio_generation;
224 struct intel_raid_conf *mdio_meta;
225 struct callout mdio_start_co; /* STARTING state timer. */
226 int mdio_disks_present;
229 struct root_hold_token *mdio_rootmount; /* Root mount delay token. */
232 static g_raid_md_create_t g_raid_md_create_intel;
233 static g_raid_md_taste_t g_raid_md_taste_intel;
234 static g_raid_md_event_t g_raid_md_event_intel;
235 static g_raid_md_ctl_t g_raid_md_ctl_intel;
236 static g_raid_md_write_t g_raid_md_write_intel;
237 static g_raid_md_fail_disk_t g_raid_md_fail_disk_intel;
238 static g_raid_md_free_disk_t g_raid_md_free_disk_intel;
239 static g_raid_md_free_volume_t g_raid_md_free_volume_intel;
240 static g_raid_md_free_t g_raid_md_free_intel;
242 static kobj_method_t g_raid_md_intel_methods[] = {
243 KOBJMETHOD(g_raid_md_create, g_raid_md_create_intel),
244 KOBJMETHOD(g_raid_md_taste, g_raid_md_taste_intel),
245 KOBJMETHOD(g_raid_md_event, g_raid_md_event_intel),
246 KOBJMETHOD(g_raid_md_ctl, g_raid_md_ctl_intel),
247 KOBJMETHOD(g_raid_md_write, g_raid_md_write_intel),
248 KOBJMETHOD(g_raid_md_fail_disk, g_raid_md_fail_disk_intel),
249 KOBJMETHOD(g_raid_md_free_disk, g_raid_md_free_disk_intel),
250 KOBJMETHOD(g_raid_md_free_volume, g_raid_md_free_volume_intel),
251 KOBJMETHOD(g_raid_md_free, g_raid_md_free_intel),
255 static struct g_raid_md_class g_raid_md_intel_class = {
257 g_raid_md_intel_methods,
258 sizeof(struct g_raid_md_intel_object),
264 static struct intel_raid_map *
265 intel_get_map(struct intel_raid_vol *mvol, int i)
267 struct intel_raid_map *mmap;
269 if (i > (mvol->migr_state ? 1 : 0))
271 mmap = &mvol->map[0];
273 mmap = (struct intel_raid_map *)
274 &mmap->disk_idx[mmap->total_disks];
276 return ((struct intel_raid_map *)mmap);
279 static struct intel_raid_vol *
280 intel_get_volume(struct intel_raid_conf *meta, int i)
282 struct intel_raid_vol *mvol;
283 struct intel_raid_map *mmap;
287 mvol = (struct intel_raid_vol *)&meta->disk[meta->total_disks];
289 mmap = intel_get_map(mvol, mvol->migr_state ? 1 : 0);
290 mvol = (struct intel_raid_vol *)
291 &mmap->disk_idx[mmap->total_disks];
297 intel_get_map_offset(struct intel_raid_map *mmap)
299 off_t offset = (off_t)mmap->offset_hi << 32;
301 offset += mmap->offset;
306 intel_set_map_offset(struct intel_raid_map *mmap, off_t offset)
309 mmap->offset = offset & 0xffffffff;
310 mmap->offset_hi = offset >> 32;
314 intel_get_map_disk_sectors(struct intel_raid_map *mmap)
316 off_t disk_sectors = (off_t)mmap->disk_sectors_hi << 32;
318 disk_sectors += mmap->disk_sectors;
319 return (disk_sectors);
323 intel_set_map_disk_sectors(struct intel_raid_map *mmap, off_t disk_sectors)
326 mmap->disk_sectors = disk_sectors & 0xffffffff;
327 mmap->disk_sectors_hi = disk_sectors >> 32;
331 intel_set_map_stripe_count(struct intel_raid_map *mmap, off_t stripe_count)
334 mmap->stripe_count = stripe_count & 0xffffffff;
335 mmap->stripe_count_hi = stripe_count >> 32;
339 intel_get_disk_sectors(struct intel_raid_disk *disk)
341 off_t sectors = (off_t)disk->sectors_hi << 32;
343 sectors += disk->sectors;
348 intel_set_disk_sectors(struct intel_raid_disk *disk, off_t sectors)
351 disk->sectors = sectors & 0xffffffff;
352 disk->sectors_hi = sectors >> 32;
356 intel_get_vol_curr_migr_unit(struct intel_raid_vol *vol)
358 off_t curr_migr_unit = (off_t)vol->curr_migr_unit_hi << 32;
360 curr_migr_unit += vol->curr_migr_unit;
361 return (curr_migr_unit);
365 intel_set_vol_curr_migr_unit(struct intel_raid_vol *vol, off_t curr_migr_unit)
368 vol->curr_migr_unit = curr_migr_unit & 0xffffffff;
369 vol->curr_migr_unit_hi = curr_migr_unit >> 32;
373 intel_status2str(int status)
379 case INTEL_S_UNINITIALIZED:
380 return ("UNINITIALIZED");
381 case INTEL_S_DEGRADED:
383 case INTEL_S_FAILURE:
391 intel_type2str(int type)
407 intel_cngst2str(int cng_state)
411 case INTEL_CNGST_UPDATED:
413 case INTEL_CNGST_NEEDS_UPDATE:
414 return ("NEEDS_UPDATE");
415 case INTEL_CNGST_MASTER_MISSING:
416 return ("MASTER_MISSING");
423 intel_mt2str(int type)
429 case INTEL_MT_REBUILD:
431 case INTEL_MT_VERIFY:
433 case INTEL_MT_GEN_MIGR:
435 case INTEL_MT_STATE_CHANGE:
436 return ("STATE_CHANGE");
437 case INTEL_MT_REPAIR:
445 g_raid_md_intel_print(struct intel_raid_conf *meta)
447 struct intel_raid_vol *mvol;
448 struct intel_raid_map *mmap;
451 if (g_raid_debug < 1)
454 printf("********* ATA Intel MatrixRAID Metadata *********\n");
455 printf("intel_id <%.24s>\n", meta->intel_id);
456 printf("version <%.6s>\n", meta->version);
457 printf("checksum 0x%08x\n", meta->checksum);
458 printf("config_size 0x%08x\n", meta->config_size);
459 printf("config_id 0x%08x\n", meta->config_id);
460 printf("generation 0x%08x\n", meta->generation);
461 printf("error_log_size %d\n", meta->error_log_size);
462 printf("attributes 0x%b\n", meta->attributes,
478 printf("total_disks %u\n", meta->total_disks);
479 printf("total_volumes %u\n", meta->total_volumes);
480 printf("error_log_pos %u\n", meta->error_log_pos);
481 printf("cache_size %u\n", meta->cache_size);
482 printf("orig_config_id 0x%08x\n", meta->orig_config_id);
483 printf("pwr_cycle_count %u\n", meta->pwr_cycle_count);
484 printf("bbm_log_size %u\n", meta->bbm_log_size);
485 printf("Flags: S - Spare, A - Assigned, F - Failed, O - Online, D - Disabled\n");
486 printf("DISK# serial disk_sectors disk_sectors_hi disk_id flags owner\n");
487 for (i = 0; i < meta->total_disks; i++ ) {
488 printf(" %d <%.16s> %u %u 0x%08x 0x%b %08x\n", i,
489 meta->disk[i].serial, meta->disk[i].sectors,
490 meta->disk[i].sectors_hi, meta->disk[i].id,
491 meta->disk[i].flags, "\20\01S\02A\03F\04O\05D",
492 meta->disk[i].owner_cfg_num);
494 for (i = 0; i < meta->total_volumes; i++) {
495 mvol = intel_get_volume(meta, i);
496 printf(" ****** Volume %d ******\n", i);
497 printf(" name %.16s\n", mvol->name);
498 printf(" total_sectors %ju\n", mvol->total_sectors);
499 printf(" state 0x%b\n", mvol->state,
503 "\003READ_COALESCING"
504 "\004WRITE_COALESCING"
505 "\005LAST_SHUTDOWN_DIRTY"
507 "\007CURRENTLY_HIDDEN"
509 "\011MAP_STATE_UNINIT"
510 "\012NO_AUTO_RECOVERY"
513 "\015CNG_MASTER_DISK_NUM");
514 printf(" reserved %u\n", mvol->reserved);
515 printf(" migr_priority %u\n", mvol->migr_priority);
516 printf(" num_sub_vols %u\n", mvol->num_sub_vols);
517 printf(" tid %u\n", mvol->tid);
518 printf(" cng_master_disk %u\n", mvol->cng_master_disk);
519 printf(" cache_policy %u\n", mvol->cache_policy);
520 printf(" cng_state %u (%s)\n", mvol->cng_state,
521 intel_cngst2str(mvol->cng_state));
522 printf(" cng_sub_state %u\n", mvol->cng_sub_state);
523 printf(" curr_migr_unit %u\n", mvol->curr_migr_unit);
524 printf(" curr_migr_unit_hi %u\n", mvol->curr_migr_unit_hi);
525 printf(" checkpoint_id %u\n", mvol->checkpoint_id);
526 printf(" migr_state %u\n", mvol->migr_state);
527 printf(" migr_type %u (%s)\n", mvol->migr_type,
528 intel_mt2str(mvol->migr_type));
529 printf(" dirty %u\n", mvol->dirty);
530 printf(" fs_state %u\n", mvol->fs_state);
531 printf(" verify_errors %u\n", mvol->verify_errors);
532 printf(" bad_blocks %u\n", mvol->bad_blocks);
534 for (j = 0; j < (mvol->migr_state ? 2 : 1); j++) {
535 printf(" *** Map %d ***\n", j);
536 mmap = intel_get_map(mvol, j);
537 printf(" offset %u\n", mmap->offset);
538 printf(" offset_hi %u\n", mmap->offset_hi);
539 printf(" disk_sectors %u\n", mmap->disk_sectors);
540 printf(" disk_sectors_hi %u\n", mmap->disk_sectors_hi);
541 printf(" stripe_count %u\n", mmap->stripe_count);
542 printf(" stripe_count_hi %u\n", mmap->stripe_count_hi);
543 printf(" strip_sectors %u\n", mmap->strip_sectors);
544 printf(" status %u (%s)\n", mmap->status,
545 intel_status2str(mmap->status));
546 printf(" type %u (%s)\n", mmap->type,
547 intel_type2str(mmap->type));
548 printf(" total_disks %u\n", mmap->total_disks);
549 printf(" total_domains %u\n", mmap->total_domains);
550 printf(" failed_disk_num %u\n", mmap->failed_disk_num);
551 printf(" ddf %u\n", mmap->ddf);
552 printf(" disk_idx ");
553 for (k = 0; k < mmap->total_disks; k++)
554 printf(" 0x%08x", mmap->disk_idx[k]);
558 printf("=================================================\n");
561 static struct intel_raid_conf *
562 intel_meta_copy(struct intel_raid_conf *meta)
564 struct intel_raid_conf *nmeta;
566 nmeta = malloc(meta->config_size, M_MD_INTEL, M_WAITOK);
567 memcpy(nmeta, meta, meta->config_size);
572 intel_meta_find_disk(struct intel_raid_conf *meta, char *serial)
576 for (pos = 0; pos < meta->total_disks; pos++) {
577 if (strncmp(meta->disk[pos].serial,
578 serial, INTEL_SERIAL_LEN) == 0)
584 static struct intel_raid_conf *
585 intel_meta_read(struct g_consumer *cp)
587 struct g_provider *pp;
588 struct intel_raid_conf *meta;
589 struct intel_raid_vol *mvol;
590 struct intel_raid_map *mmap, *mmap1;
592 int error, i, j, k, left, size;
593 uint32_t checksum, *ptr;
597 /* Read the anchor sector. */
598 buf = g_read_data(cp,
599 pp->mediasize - pp->sectorsize * 2, pp->sectorsize, &error);
601 G_RAID_DEBUG(1, "Cannot read metadata from %s (error=%d).",
605 meta = (struct intel_raid_conf *)buf;
607 /* Check if this is an Intel RAID struct */
608 if (strncmp(meta->intel_id, INTEL_MAGIC, strlen(INTEL_MAGIC))) {
609 G_RAID_DEBUG(1, "Intel signature check failed on %s", pp->name);
613 if (meta->config_size > 65536 ||
614 meta->config_size < sizeof(struct intel_raid_conf)) {
615 G_RAID_DEBUG(1, "Intel metadata size looks wrong: %d",
620 size = meta->config_size;
621 meta = malloc(size, M_MD_INTEL, M_WAITOK);
622 memcpy(meta, buf, min(size, pp->sectorsize));
625 /* Read all the rest, if needed. */
626 if (meta->config_size > pp->sectorsize) {
627 left = (meta->config_size - 1) / pp->sectorsize;
628 buf = g_read_data(cp,
629 pp->mediasize - pp->sectorsize * (2 + left),
630 pp->sectorsize * left, &error);
632 G_RAID_DEBUG(1, "Cannot read remaining metadata"
633 " part from %s (error=%d).",
635 free(meta, M_MD_INTEL);
638 memcpy(((char *)meta) + pp->sectorsize, buf,
639 pp->sectorsize * left);
643 /* Check metadata checksum. */
644 for (checksum = 0, ptr = (uint32_t *)meta, i = 0;
645 i < (meta->config_size / sizeof(uint32_t)); i++) {
648 checksum -= meta->checksum;
649 if (checksum != meta->checksum) {
650 G_RAID_DEBUG(1, "Intel checksum check failed on %s", pp->name);
651 free(meta, M_MD_INTEL);
655 /* Validate metadata size. */
656 size = sizeof(struct intel_raid_conf) +
657 sizeof(struct intel_raid_disk) * (meta->total_disks - 1) +
658 sizeof(struct intel_raid_vol) * meta->total_volumes;
659 if (size > meta->config_size) {
661 G_RAID_DEBUG(1, "Intel metadata size incorrect %d < %d",
662 meta->config_size, size);
663 free(meta, M_MD_INTEL);
666 for (i = 0; i < meta->total_volumes; i++) {
667 mvol = intel_get_volume(meta, i);
668 mmap = intel_get_map(mvol, 0);
669 size += 4 * (mmap->total_disks - 1);
670 if (size > meta->config_size)
672 if (mvol->migr_state) {
673 size += sizeof(struct intel_raid_map);
674 if (size > meta->config_size)
676 mmap = intel_get_map(mvol, 1);
677 size += 4 * (mmap->total_disks - 1);
678 if (size > meta->config_size)
683 g_raid_md_intel_print(meta);
685 if (strncmp(meta->version, INTEL_VERSION_1300, 6) > 0) {
686 G_RAID_DEBUG(1, "Intel unsupported version: '%.6s'",
688 free(meta, M_MD_INTEL);
692 if (strncmp(meta->version, INTEL_VERSION_1300, 6) >= 0 &&
693 (meta->attributes & ~INTEL_ATTR_SUPPORTED) != 0) {
694 G_RAID_DEBUG(1, "Intel unsupported attributes: 0x%08x",
695 meta->attributes & ~INTEL_ATTR_SUPPORTED);
696 free(meta, M_MD_INTEL);
700 /* Validate disk indexes. */
701 for (i = 0; i < meta->total_volumes; i++) {
702 mvol = intel_get_volume(meta, i);
703 for (j = 0; j < (mvol->migr_state ? 2 : 1); j++) {
704 mmap = intel_get_map(mvol, j);
705 for (k = 0; k < mmap->total_disks; k++) {
706 if ((mmap->disk_idx[k] & INTEL_DI_IDX) >
708 G_RAID_DEBUG(1, "Intel metadata disk"
709 " index %d too big (>%d)",
710 mmap->disk_idx[k] & INTEL_DI_IDX,
712 free(meta, M_MD_INTEL);
719 /* Validate migration types. */
720 for (i = 0; i < meta->total_volumes; i++) {
721 mvol = intel_get_volume(meta, i);
722 /* Deny unknown migration types. */
723 if (mvol->migr_state &&
724 mvol->migr_type != INTEL_MT_INIT &&
725 mvol->migr_type != INTEL_MT_REBUILD &&
726 mvol->migr_type != INTEL_MT_VERIFY &&
727 mvol->migr_type != INTEL_MT_GEN_MIGR &&
728 mvol->migr_type != INTEL_MT_REPAIR) {
729 G_RAID_DEBUG(1, "Intel metadata has unsupported"
730 " migration type %d", mvol->migr_type);
731 free(meta, M_MD_INTEL);
734 /* Deny general migrations except SINGLE->RAID1. */
735 if (mvol->migr_state &&
736 mvol->migr_type == INTEL_MT_GEN_MIGR) {
737 mmap = intel_get_map(mvol, 0);
738 mmap1 = intel_get_map(mvol, 1);
739 if (mmap1->total_disks != 1 ||
740 mmap->type != INTEL_T_RAID1 ||
741 mmap->total_disks != 2 ||
742 mmap->offset != mmap1->offset ||
743 mmap->disk_sectors != mmap1->disk_sectors ||
744 mmap->total_domains != mmap->total_disks ||
745 mmap->offset_hi != mmap1->offset_hi ||
746 mmap->disk_sectors_hi != mmap1->disk_sectors_hi ||
747 (mmap->disk_idx[0] != mmap1->disk_idx[0] &&
748 mmap->disk_idx[0] != mmap1->disk_idx[1])) {
749 G_RAID_DEBUG(1, "Intel metadata has unsupported"
750 " variant of general migration");
751 free(meta, M_MD_INTEL);
761 intel_meta_write(struct g_consumer *cp, struct intel_raid_conf *meta)
763 struct g_provider *pp;
765 int error, i, sectors;
766 uint32_t checksum, *ptr;
770 /* Recalculate checksum for case if metadata were changed. */
772 for (checksum = 0, ptr = (uint32_t *)meta, i = 0;
773 i < (meta->config_size / sizeof(uint32_t)); i++) {
776 meta->checksum = checksum;
778 /* Create and fill buffer. */
779 sectors = howmany(meta->config_size, pp->sectorsize);
780 buf = malloc(sectors * pp->sectorsize, M_MD_INTEL, M_WAITOK | M_ZERO);
782 memcpy(buf, ((char *)meta) + pp->sectorsize,
783 (sectors - 1) * pp->sectorsize);
785 memcpy(buf + (sectors - 1) * pp->sectorsize, meta, pp->sectorsize);
787 error = g_write_data(cp,
788 pp->mediasize - pp->sectorsize * (1 + sectors),
789 buf, pp->sectorsize * sectors);
791 G_RAID_DEBUG(1, "Cannot write metadata to %s (error=%d).",
795 free(buf, M_MD_INTEL);
800 intel_meta_erase(struct g_consumer *cp)
802 struct g_provider *pp;
807 buf = malloc(pp->sectorsize, M_MD_INTEL, M_WAITOK | M_ZERO);
808 error = g_write_data(cp,
809 pp->mediasize - 2 * pp->sectorsize,
810 buf, pp->sectorsize);
812 G_RAID_DEBUG(1, "Cannot erase metadata on %s (error=%d).",
815 free(buf, M_MD_INTEL);
820 intel_meta_write_spare(struct g_consumer *cp, struct intel_raid_disk *d)
822 struct intel_raid_conf *meta;
825 /* Fill anchor and single disk. */
826 meta = malloc(INTEL_MAX_MD_SIZE(1), M_MD_INTEL, M_WAITOK | M_ZERO);
827 memcpy(&meta->intel_id[0], INTEL_MAGIC, sizeof(INTEL_MAGIC) - 1);
828 memcpy(&meta->version[0], INTEL_VERSION_1000,
829 sizeof(INTEL_VERSION_1000) - 1);
830 meta->config_size = INTEL_MAX_MD_SIZE(1);
831 meta->config_id = meta->orig_config_id = arc4random();
832 meta->generation = 1;
833 meta->total_disks = 1;
835 error = intel_meta_write(cp, meta);
836 free(meta, M_MD_INTEL);
840 static struct g_raid_disk *
841 g_raid_md_intel_get_disk(struct g_raid_softc *sc, int id)
843 struct g_raid_disk *disk;
844 struct g_raid_md_intel_perdisk *pd;
846 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
847 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
848 if (pd->pd_disk_pos == id)
855 g_raid_md_intel_supported(int level, int qual, int disks, int force)
859 case G_RAID_VOLUME_RL_RAID0:
862 if (!force && (disks < 2 || disks > 6))
865 case G_RAID_VOLUME_RL_RAID1:
868 if (!force && (disks != 2))
871 case G_RAID_VOLUME_RL_RAID1E:
874 if (!force && (disks != 4))
877 case G_RAID_VOLUME_RL_RAID5:
880 if (!force && disks > 6)
882 if (qual != G_RAID_VOLUME_RLQ_R5LA)
888 if (level != G_RAID_VOLUME_RL_RAID5 && qual != G_RAID_VOLUME_RLQ_NONE)
893 static struct g_raid_volume *
894 g_raid_md_intel_get_volume(struct g_raid_softc *sc, int id)
896 struct g_raid_volume *mvol;
897 struct g_raid_md_intel_pervolume *pv;
899 TAILQ_FOREACH(mvol, &sc->sc_volumes, v_next) {
900 pv = mvol->v_md_data;
901 if (pv->pv_volume_pos == id)
908 g_raid_md_intel_start_disk(struct g_raid_disk *disk)
910 struct g_raid_softc *sc;
911 struct g_raid_subdisk *sd, *tmpsd;
912 struct g_raid_disk *olddisk, *tmpdisk;
913 struct g_raid_md_object *md;
914 struct g_raid_md_intel_object *mdi;
915 struct g_raid_md_intel_pervolume *pv;
916 struct g_raid_md_intel_perdisk *pd, *oldpd;
917 struct intel_raid_conf *meta;
918 struct intel_raid_vol *mvol;
919 struct intel_raid_map *mmap0, *mmap1;
920 int disk_pos, resurrection = 0, migr_global, i;
924 mdi = (struct g_raid_md_intel_object *)md;
925 meta = mdi->mdio_meta;
926 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
929 /* Find disk position in metadata by its serial. */
930 disk_pos = intel_meta_find_disk(meta, pd->pd_disk_meta.serial);
932 G_RAID_DEBUG1(1, sc, "Unknown, probably new or stale disk");
933 /* Failed stale disk is useless for us. */
934 if ((pd->pd_disk_meta.flags & INTEL_F_FAILED) &&
935 !(pd->pd_disk_meta.flags & INTEL_F_DISABLED)) {
936 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE_FAILED);
939 /* If we are in the start process, that's all for now. */
940 if (!mdi->mdio_started)
943 * If we have already started - try to get use of the disk.
944 * Try to replace OFFLINE disks first, then FAILED.
946 TAILQ_FOREACH(tmpdisk, &sc->sc_disks, d_next) {
947 if (tmpdisk->d_state != G_RAID_DISK_S_OFFLINE &&
948 tmpdisk->d_state != G_RAID_DISK_S_FAILED)
950 /* Make sure this disk is big enough. */
951 TAILQ_FOREACH(sd, &tmpdisk->d_subdisks, sd_next) {
953 intel_get_disk_sectors(&pd->pd_disk_meta);
955 if (sd->sd_offset + sd->sd_size + 4096 >
956 disk_sectors * 512) {
958 "Disk too small (%llu < %llu)",
962 sd->sd_offset + sd->sd_size + 4096);
968 if (tmpdisk->d_state == G_RAID_DISK_S_OFFLINE) {
971 } else if (olddisk == NULL)
974 if (olddisk == NULL) {
976 if (pd->pd_disk_meta.flags & INTEL_F_SPARE) {
977 g_raid_change_disk_state(disk,
978 G_RAID_DISK_S_SPARE);
981 g_raid_change_disk_state(disk,
982 G_RAID_DISK_S_STALE);
986 oldpd = (struct g_raid_md_intel_perdisk *)olddisk->d_md_data;
987 disk_pos = oldpd->pd_disk_pos;
991 if (olddisk == NULL) {
992 /* Find placeholder by position. */
993 olddisk = g_raid_md_intel_get_disk(sc, disk_pos);
995 panic("No disk at position %d!", disk_pos);
996 if (olddisk->d_state != G_RAID_DISK_S_OFFLINE) {
997 G_RAID_DEBUG1(1, sc, "More than one disk for pos %d",
999 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE);
1002 oldpd = (struct g_raid_md_intel_perdisk *)olddisk->d_md_data;
1005 /* Replace failed disk or placeholder with new disk. */
1006 TAILQ_FOREACH_SAFE(sd, &olddisk->d_subdisks, sd_next, tmpsd) {
1007 TAILQ_REMOVE(&olddisk->d_subdisks, sd, sd_next);
1008 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1011 oldpd->pd_disk_pos = -2;
1012 pd->pd_disk_pos = disk_pos;
1014 /* If it was placeholder -- destroy it. */
1015 if (olddisk->d_state == G_RAID_DISK_S_OFFLINE) {
1016 g_raid_destroy_disk(olddisk);
1018 /* Otherwise, make it STALE_FAILED. */
1019 g_raid_change_disk_state(olddisk, G_RAID_DISK_S_STALE_FAILED);
1020 /* Update global metadata just in case. */
1021 memcpy(&meta->disk[disk_pos], &pd->pd_disk_meta,
1022 sizeof(struct intel_raid_disk));
1025 /* Welcome the new disk. */
1026 if ((meta->disk[disk_pos].flags & INTEL_F_DISABLED) &&
1027 !(pd->pd_disk_meta.flags & INTEL_F_SPARE))
1028 g_raid_change_disk_state(disk, G_RAID_DISK_S_DISABLED);
1029 else if (resurrection)
1030 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
1031 else if (meta->disk[disk_pos].flags & INTEL_F_FAILED)
1032 g_raid_change_disk_state(disk, G_RAID_DISK_S_FAILED);
1033 else if (meta->disk[disk_pos].flags & INTEL_F_SPARE)
1034 g_raid_change_disk_state(disk, G_RAID_DISK_S_SPARE);
1036 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
1037 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
1038 pv = sd->sd_volume->v_md_data;
1039 mvol = intel_get_volume(meta, pv->pv_volume_pos);
1040 mmap0 = intel_get_map(mvol, 0);
1041 if (mvol->migr_state)
1042 mmap1 = intel_get_map(mvol, 1);
1047 for (i = 0; i < mmap0->total_disks; i++) {
1048 if ((mmap0->disk_idx[i] & INTEL_DI_RBLD) == 0 &&
1049 (mmap1->disk_idx[i] & INTEL_DI_RBLD) != 0)
1053 if ((meta->disk[disk_pos].flags & INTEL_F_DISABLED) &&
1054 !(pd->pd_disk_meta.flags & INTEL_F_SPARE)) {
1055 /* Disabled disk, useless. */
1056 g_raid_change_subdisk_state(sd,
1057 G_RAID_SUBDISK_S_NONE);
1058 } else if (resurrection) {
1059 /* Stale disk, almost same as new. */
1060 g_raid_change_subdisk_state(sd,
1061 G_RAID_SUBDISK_S_NEW);
1062 } else if (meta->disk[disk_pos].flags & INTEL_F_FAILED) {
1063 /* Failed disk, almost useless. */
1064 g_raid_change_subdisk_state(sd,
1065 G_RAID_SUBDISK_S_FAILED);
1066 } else if (mvol->migr_state == 0) {
1067 if (mmap0->status == INTEL_S_UNINITIALIZED &&
1068 (!pv->pv_cng || pv->pv_cng_master_disk != disk_pos)) {
1069 /* Freshly created uninitialized volume. */
1070 g_raid_change_subdisk_state(sd,
1071 G_RAID_SUBDISK_S_UNINITIALIZED);
1072 } else if (mmap0->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) {
1073 /* Freshly inserted disk. */
1074 g_raid_change_subdisk_state(sd,
1075 G_RAID_SUBDISK_S_NEW);
1076 } else if (mvol->dirty && (!pv->pv_cng ||
1077 pv->pv_cng_master_disk != disk_pos)) {
1078 /* Dirty volume (unclean shutdown). */
1079 g_raid_change_subdisk_state(sd,
1080 G_RAID_SUBDISK_S_STALE);
1082 /* Up to date disk. */
1083 g_raid_change_subdisk_state(sd,
1084 G_RAID_SUBDISK_S_ACTIVE);
1086 } else if (mvol->migr_type == INTEL_MT_INIT ||
1087 mvol->migr_type == INTEL_MT_REBUILD) {
1088 if (mmap0->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) {
1089 /* Freshly inserted disk. */
1090 g_raid_change_subdisk_state(sd,
1091 G_RAID_SUBDISK_S_NEW);
1092 } else if (mmap1->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) {
1093 /* Rebuilding disk. */
1094 g_raid_change_subdisk_state(sd,
1095 G_RAID_SUBDISK_S_REBUILD);
1097 sd->sd_rebuild_pos = 0;
1099 sd->sd_rebuild_pos =
1100 intel_get_vol_curr_migr_unit(mvol) *
1101 sd->sd_volume->v_strip_size *
1102 mmap0->total_domains;
1104 } else if (mvol->migr_type == INTEL_MT_INIT &&
1106 /* Freshly created uninitialized volume. */
1107 g_raid_change_subdisk_state(sd,
1108 G_RAID_SUBDISK_S_UNINITIALIZED);
1109 } else if (mvol->dirty && (!pv->pv_cng ||
1110 pv->pv_cng_master_disk != disk_pos)) {
1111 /* Dirty volume (unclean shutdown). */
1112 g_raid_change_subdisk_state(sd,
1113 G_RAID_SUBDISK_S_STALE);
1115 /* Up to date disk. */
1116 g_raid_change_subdisk_state(sd,
1117 G_RAID_SUBDISK_S_ACTIVE);
1119 } else if (mvol->migr_type == INTEL_MT_VERIFY ||
1120 mvol->migr_type == INTEL_MT_REPAIR) {
1121 if (mmap0->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) {
1122 /* Freshly inserted disk. */
1123 g_raid_change_subdisk_state(sd,
1124 G_RAID_SUBDISK_S_NEW);
1125 } else if ((mmap1->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) ||
1127 /* Resyncing disk. */
1128 g_raid_change_subdisk_state(sd,
1129 G_RAID_SUBDISK_S_RESYNC);
1131 sd->sd_rebuild_pos = 0;
1133 sd->sd_rebuild_pos =
1134 intel_get_vol_curr_migr_unit(mvol) *
1135 sd->sd_volume->v_strip_size *
1136 mmap0->total_domains;
1138 } else if (mvol->dirty) {
1139 /* Dirty volume (unclean shutdown). */
1140 g_raid_change_subdisk_state(sd,
1141 G_RAID_SUBDISK_S_STALE);
1143 /* Up to date disk. */
1144 g_raid_change_subdisk_state(sd,
1145 G_RAID_SUBDISK_S_ACTIVE);
1147 } else if (mvol->migr_type == INTEL_MT_GEN_MIGR) {
1148 if ((mmap1->disk_idx[0] & INTEL_DI_IDX) != disk_pos) {
1149 /* Freshly inserted disk. */
1150 g_raid_change_subdisk_state(sd,
1151 G_RAID_SUBDISK_S_NEW);
1153 /* Up to date disk. */
1154 g_raid_change_subdisk_state(sd,
1155 G_RAID_SUBDISK_S_ACTIVE);
1158 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
1159 G_RAID_EVENT_SUBDISK);
1162 /* Update status of our need for spare. */
1163 if (mdi->mdio_started) {
1164 mdi->mdio_incomplete =
1165 (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) +
1166 g_raid_ndisks(sc, G_RAID_DISK_S_DISABLED) <
1170 return (resurrection);
1174 g_disk_md_intel_retaste(void *arg, int pending)
1177 G_RAID_DEBUG(1, "Array is not complete, trying to retaste.");
1178 g_retaste(&g_raid_class);
1179 free(arg, M_MD_INTEL);
1183 g_raid_md_intel_refill(struct g_raid_softc *sc)
1185 struct g_raid_md_object *md;
1186 struct g_raid_md_intel_object *mdi;
1187 struct intel_raid_conf *meta;
1188 struct g_raid_disk *disk;
1193 mdi = (struct g_raid_md_intel_object *)md;
1194 meta = mdi->mdio_meta;
1197 /* Make sure we miss anything. */
1198 na = g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) +
1199 g_raid_ndisks(sc, G_RAID_DISK_S_DISABLED);
1200 if (na == meta->total_disks)
1203 G_RAID_DEBUG1(1, md->mdo_softc,
1204 "Array is not complete (%d of %d), "
1205 "trying to refill.", na, meta->total_disks);
1207 /* Try to get use some of STALE disks. */
1208 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1209 if (disk->d_state == G_RAID_DISK_S_STALE) {
1210 update += g_raid_md_intel_start_disk(disk);
1211 if (disk->d_state == G_RAID_DISK_S_ACTIVE ||
1212 disk->d_state == G_RAID_DISK_S_DISABLED)
1219 /* Try to get use some of SPARE disks. */
1220 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1221 if (disk->d_state == G_RAID_DISK_S_SPARE) {
1222 update += g_raid_md_intel_start_disk(disk);
1223 if (disk->d_state == G_RAID_DISK_S_ACTIVE)
1227 } while (disk != NULL);
1229 /* Write new metadata if we changed something. */
1231 g_raid_md_write_intel(md, NULL, NULL, NULL);
1232 meta = mdi->mdio_meta;
1235 /* Update status of our need for spare. */
1236 mdi->mdio_incomplete = (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) +
1237 g_raid_ndisks(sc, G_RAID_DISK_S_DISABLED) < meta->total_disks);
1239 /* Request retaste hoping to find spare. */
1240 if (mdi->mdio_incomplete) {
1241 task = malloc(sizeof(struct task),
1242 M_MD_INTEL, M_WAITOK | M_ZERO);
1243 TASK_INIT(task, 0, g_disk_md_intel_retaste, task);
1244 taskqueue_enqueue(taskqueue_swi, task);
1249 g_raid_md_intel_start(struct g_raid_softc *sc)
1251 struct g_raid_md_object *md;
1252 struct g_raid_md_intel_object *mdi;
1253 struct g_raid_md_intel_pervolume *pv;
1254 struct g_raid_md_intel_perdisk *pd;
1255 struct intel_raid_conf *meta;
1256 struct intel_raid_vol *mvol;
1257 struct intel_raid_map *mmap;
1258 struct g_raid_volume *vol;
1259 struct g_raid_subdisk *sd;
1260 struct g_raid_disk *disk;
1264 mdi = (struct g_raid_md_intel_object *)md;
1265 meta = mdi->mdio_meta;
1267 /* Create volumes and subdisks. */
1268 for (i = 0; i < meta->total_volumes; i++) {
1269 mvol = intel_get_volume(meta, i);
1270 mmap = intel_get_map(mvol, 0);
1271 vol = g_raid_create_volume(sc, mvol->name, mvol->tid - 1);
1272 pv = malloc(sizeof(*pv), M_MD_INTEL, M_WAITOK | M_ZERO);
1273 pv->pv_volume_pos = i;
1274 pv->pv_cng = (mvol->state & INTEL_ST_CLONE_N_GO) != 0;
1275 pv->pv_cng_man_sync = (mvol->state & INTEL_ST_CLONE_MAN_SYNC) != 0;
1276 if (mvol->cng_master_disk < mmap->total_disks)
1277 pv->pv_cng_master_disk = mvol->cng_master_disk;
1278 vol->v_md_data = pv;
1279 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE;
1280 if (mmap->type == INTEL_T_RAID0)
1281 vol->v_raid_level = G_RAID_VOLUME_RL_RAID0;
1282 else if (mmap->type == INTEL_T_RAID1 &&
1283 mmap->total_domains >= 2 &&
1284 mmap->total_domains <= mmap->total_disks) {
1285 /* Assume total_domains is correct. */
1286 if (mmap->total_domains == mmap->total_disks)
1287 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1;
1289 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E;
1290 } else if (mmap->type == INTEL_T_RAID1) {
1291 /* total_domains looks wrong. */
1292 if (mmap->total_disks <= 2)
1293 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1;
1295 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E;
1296 } else if (mmap->type == INTEL_T_RAID5) {
1297 vol->v_raid_level = G_RAID_VOLUME_RL_RAID5;
1298 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_R5LA;
1300 vol->v_raid_level = G_RAID_VOLUME_RL_UNKNOWN;
1301 vol->v_strip_size = (u_int)mmap->strip_sectors * 512; //ZZZ
1302 vol->v_disks_count = mmap->total_disks;
1303 vol->v_mediasize = (off_t)mvol->total_sectors * 512; //ZZZ
1304 vol->v_sectorsize = 512; //ZZZ
1305 for (j = 0; j < vol->v_disks_count; j++) {
1306 sd = &vol->v_subdisks[j];
1307 sd->sd_offset = intel_get_map_offset(mmap) * 512; //ZZZ
1308 sd->sd_size = intel_get_map_disk_sectors(mmap) * 512; //ZZZ
1310 g_raid_start_volume(vol);
1313 /* Create disk placeholders to store data for later writing. */
1314 for (disk_pos = 0; disk_pos < meta->total_disks; disk_pos++) {
1315 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO);
1316 pd->pd_disk_pos = disk_pos;
1317 pd->pd_disk_meta = meta->disk[disk_pos];
1318 disk = g_raid_create_disk(sc);
1319 disk->d_md_data = (void *)pd;
1320 disk->d_state = G_RAID_DISK_S_OFFLINE;
1321 for (i = 0; i < meta->total_volumes; i++) {
1322 mvol = intel_get_volume(meta, i);
1323 mmap = intel_get_map(mvol, 0);
1324 for (j = 0; j < mmap->total_disks; j++) {
1325 if ((mmap->disk_idx[j] & INTEL_DI_IDX) == disk_pos)
1328 if (j == mmap->total_disks)
1330 vol = g_raid_md_intel_get_volume(sc, i);
1331 sd = &vol->v_subdisks[j];
1333 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1337 /* Make all disks found till the moment take their places. */
1339 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1340 if (disk->d_state == G_RAID_DISK_S_NONE) {
1341 g_raid_md_intel_start_disk(disk);
1345 } while (disk != NULL);
1347 mdi->mdio_started = 1;
1348 G_RAID_DEBUG1(0, sc, "Array started.");
1349 g_raid_md_write_intel(md, NULL, NULL, NULL);
1351 /* Pickup any STALE/SPARE disks to refill array if needed. */
1352 g_raid_md_intel_refill(sc);
1354 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
1355 g_raid_event_send(vol, G_RAID_VOLUME_E_START,
1356 G_RAID_EVENT_VOLUME);
1359 callout_stop(&mdi->mdio_start_co);
1360 G_RAID_DEBUG1(1, sc, "root_mount_rel %p", mdi->mdio_rootmount);
1361 root_mount_rel(mdi->mdio_rootmount);
1362 mdi->mdio_rootmount = NULL;
1366 g_raid_md_intel_new_disk(struct g_raid_disk *disk)
1368 struct g_raid_softc *sc;
1369 struct g_raid_md_object *md;
1370 struct g_raid_md_intel_object *mdi;
1371 struct intel_raid_conf *pdmeta;
1372 struct g_raid_md_intel_perdisk *pd;
1376 mdi = (struct g_raid_md_intel_object *)md;
1377 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
1378 pdmeta = pd->pd_meta;
1380 if (mdi->mdio_started) {
1381 if (g_raid_md_intel_start_disk(disk))
1382 g_raid_md_write_intel(md, NULL, NULL, NULL);
1384 /* If we haven't started yet - check metadata freshness. */
1385 if (mdi->mdio_meta == NULL ||
1386 ((int32_t)(pdmeta->generation - mdi->mdio_generation)) > 0) {
1387 G_RAID_DEBUG1(1, sc, "Newer disk");
1388 if (mdi->mdio_meta != NULL)
1389 free(mdi->mdio_meta, M_MD_INTEL);
1390 mdi->mdio_meta = intel_meta_copy(pdmeta);
1391 mdi->mdio_generation = mdi->mdio_meta->generation;
1392 mdi->mdio_disks_present = 1;
1393 } else if (pdmeta->generation == mdi->mdio_generation) {
1394 mdi->mdio_disks_present++;
1395 G_RAID_DEBUG1(1, sc, "Matching disk (%d of %d up)",
1396 mdi->mdio_disks_present,
1397 mdi->mdio_meta->total_disks);
1399 G_RAID_DEBUG1(1, sc, "Older disk");
1401 /* If we collected all needed disks - start array. */
1402 if (mdi->mdio_disks_present == mdi->mdio_meta->total_disks)
1403 g_raid_md_intel_start(sc);
1408 g_raid_intel_go(void *arg)
1410 struct g_raid_softc *sc;
1411 struct g_raid_md_object *md;
1412 struct g_raid_md_intel_object *mdi;
1416 mdi = (struct g_raid_md_intel_object *)md;
1417 if (!mdi->mdio_started) {
1418 G_RAID_DEBUG1(0, sc, "Force array start due to timeout.");
1419 g_raid_event_send(sc, G_RAID_NODE_E_START, 0);
1424 g_raid_md_create_intel(struct g_raid_md_object *md, struct g_class *mp,
1427 struct g_raid_softc *sc;
1428 struct g_raid_md_intel_object *mdi;
1431 mdi = (struct g_raid_md_intel_object *)md;
1432 mdi->mdio_config_id = mdi->mdio_orig_config_id = arc4random();
1433 mdi->mdio_generation = 0;
1434 snprintf(name, sizeof(name), "Intel-%08x", mdi->mdio_config_id);
1435 sc = g_raid_create_node(mp, name, md);
1437 return (G_RAID_MD_TASTE_FAIL);
1440 return (G_RAID_MD_TASTE_NEW);
1444 * Return the last N characters of the serial label. The Linux and
1445 * ataraid(7) code always uses the last 16 characters of the label to
1446 * store into the Intel meta format. Generalize this to N characters
1447 * since that's easy. Labels can be up to 20 characters for SATA drives
1448 * and up 251 characters for SAS drives. Since intel controllers don't
1449 * support SAS drives, just stick with the SATA limits for stack friendliness.
1452 g_raid_md_get_label(struct g_consumer *cp, char *serial, int serlen)
1454 char serial_buffer[DISK_IDENT_SIZE];
1457 len = sizeof(serial_buffer);
1458 error = g_io_getattr("GEOM::ident", cp, &len, serial_buffer);
1461 len = strlen(serial_buffer);
1466 strncpy(serial, serial_buffer + len, serlen);
1471 g_raid_md_taste_intel(struct g_raid_md_object *md, struct g_class *mp,
1472 struct g_consumer *cp, struct g_geom **gp)
1474 struct g_consumer *rcp;
1475 struct g_provider *pp;
1476 struct g_raid_md_intel_object *mdi, *mdi1;
1477 struct g_raid_softc *sc;
1478 struct g_raid_disk *disk;
1479 struct intel_raid_conf *meta;
1480 struct g_raid_md_intel_perdisk *pd;
1481 struct g_geom *geom;
1482 int error, disk_pos, result, spare, len;
1483 char serial[INTEL_SERIAL_LEN];
1487 G_RAID_DEBUG(1, "Tasting Intel on %s", cp->provider->name);
1488 mdi = (struct g_raid_md_intel_object *)md;
1491 /* Read metadata from device. */
1494 g_topology_unlock();
1495 error = g_raid_md_get_label(cp, serial, sizeof(serial));
1497 G_RAID_DEBUG(1, "Cannot get serial number from %s (error=%d).",
1502 len = sizeof(vendor);
1503 if (pp->geom->rank == 1)
1504 g_io_getattr("GEOM::hba_vendor", cp, &len, &vendor);
1505 meta = intel_meta_read(cp);
1508 if (g_raid_aggressive_spare) {
1509 if (vendor != 0x8086) {
1511 "Intel vendor mismatch 0x%04x != 0x8086",
1515 "No Intel metadata, forcing spare.");
1520 return (G_RAID_MD_TASTE_FAIL);
1523 /* Check this disk position in obtained metadata. */
1524 disk_pos = intel_meta_find_disk(meta, serial);
1526 G_RAID_DEBUG(1, "Intel serial '%s' not found", serial);
1529 if (intel_get_disk_sectors(&meta->disk[disk_pos]) !=
1530 (pp->mediasize / pp->sectorsize)) {
1531 G_RAID_DEBUG(1, "Intel size mismatch %ju != %ju",
1532 intel_get_disk_sectors(&meta->disk[disk_pos]),
1533 (off_t)(pp->mediasize / pp->sectorsize));
1537 G_RAID_DEBUG(1, "Intel disk position %d", disk_pos);
1538 spare = meta->disk[disk_pos].flags & INTEL_F_SPARE;
1541 /* Search for matching node. */
1544 LIST_FOREACH(geom, &mp->geom, geom) {
1548 if (sc->sc_stopping != 0)
1550 if (sc->sc_md->mdo_class != md->mdo_class)
1552 mdi1 = (struct g_raid_md_intel_object *)sc->sc_md;
1554 if (mdi1->mdio_incomplete)
1557 if (mdi1->mdio_config_id == meta->config_id)
1562 /* Found matching node. */
1564 G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name);
1565 result = G_RAID_MD_TASTE_EXISTING;
1567 } else if (spare) { /* Not found needy node -- left for later. */
1568 G_RAID_DEBUG(1, "Spare is not needed at this time");
1571 } else { /* Not found matching node -- create one. */
1572 result = G_RAID_MD_TASTE_NEW;
1573 mdi->mdio_config_id = meta->config_id;
1574 mdi->mdio_orig_config_id = meta->orig_config_id;
1575 snprintf(name, sizeof(name), "Intel-%08x", meta->config_id);
1576 sc = g_raid_create_node(mp, name, md);
1579 callout_init(&mdi->mdio_start_co, 1);
1580 callout_reset(&mdi->mdio_start_co, g_raid_start_timeout * hz,
1581 g_raid_intel_go, sc);
1582 mdi->mdio_rootmount = root_mount_hold("GRAID-Intel");
1583 G_RAID_DEBUG1(1, sc, "root_mount_hold %p", mdi->mdio_rootmount);
1586 /* There is no return after this point, so we close passed consumer. */
1587 g_access(cp, -1, 0, 0);
1589 rcp = g_new_consumer(geom);
1590 rcp->flags |= G_CF_DIRECT_RECEIVE;
1592 if (g_access(rcp, 1, 1, 1) != 0)
1595 g_topology_unlock();
1596 sx_xlock(&sc->sc_lock);
1598 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO);
1600 pd->pd_disk_pos = -1;
1602 memcpy(&pd->pd_disk_meta.serial[0], serial, INTEL_SERIAL_LEN);
1603 intel_set_disk_sectors(&pd->pd_disk_meta,
1604 pp->mediasize / pp->sectorsize);
1605 pd->pd_disk_meta.id = 0;
1606 pd->pd_disk_meta.flags = INTEL_F_SPARE;
1608 pd->pd_disk_meta = meta->disk[disk_pos];
1610 disk = g_raid_create_disk(sc);
1611 disk->d_md_data = (void *)pd;
1612 disk->d_consumer = rcp;
1613 rcp->private = disk;
1615 g_raid_get_disk_info(disk);
1617 g_raid_md_intel_new_disk(disk);
1619 sx_xunlock(&sc->sc_lock);
1626 free(meta, M_MD_INTEL);
1627 return (G_RAID_MD_TASTE_FAIL);
1631 g_raid_md_event_intel(struct g_raid_md_object *md,
1632 struct g_raid_disk *disk, u_int event)
1634 struct g_raid_softc *sc;
1635 struct g_raid_subdisk *sd;
1636 struct g_raid_md_intel_object *mdi;
1637 struct g_raid_md_intel_perdisk *pd;
1640 mdi = (struct g_raid_md_intel_object *)md;
1643 case G_RAID_NODE_E_START:
1644 if (!mdi->mdio_started)
1645 g_raid_md_intel_start(sc);
1650 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
1652 case G_RAID_DISK_E_DISCONNECTED:
1653 /* If disk was assigned, just update statuses. */
1654 if (pd->pd_disk_pos >= 0) {
1655 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1656 if (disk->d_consumer) {
1657 g_raid_kill_consumer(sc, disk->d_consumer);
1658 disk->d_consumer = NULL;
1660 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
1661 g_raid_change_subdisk_state(sd,
1662 G_RAID_SUBDISK_S_NONE);
1663 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
1664 G_RAID_EVENT_SUBDISK);
1667 /* Otherwise -- delete. */
1668 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
1669 g_raid_destroy_disk(disk);
1672 /* Write updated metadata to all disks. */
1673 g_raid_md_write_intel(md, NULL, NULL, NULL);
1675 /* Check if anything left except placeholders. */
1676 if (g_raid_ndisks(sc, -1) ==
1677 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1678 g_raid_destroy_node(sc, 0);
1680 g_raid_md_intel_refill(sc);
1687 g_raid_md_ctl_intel(struct g_raid_md_object *md,
1688 struct gctl_req *req)
1690 struct g_raid_softc *sc;
1691 struct g_raid_volume *vol, *vol1;
1692 struct g_raid_subdisk *sd;
1693 struct g_raid_disk *disk;
1694 struct g_raid_md_intel_object *mdi;
1695 struct g_raid_md_intel_pervolume *pv;
1696 struct g_raid_md_intel_perdisk *pd;
1697 struct g_consumer *cp;
1698 struct g_provider *pp;
1699 char arg[16], serial[INTEL_SERIAL_LEN];
1700 const char *nodename, *verb, *volname, *levelname, *diskname;
1703 off_t off, size, sectorsize, strip, disk_sectors;
1704 intmax_t *sizearg, *striparg;
1705 int numdisks, i, len, level, qual, update;
1709 mdi = (struct g_raid_md_intel_object *)md;
1710 verb = gctl_get_param(req, "verb", NULL);
1711 nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
1713 if (strcmp(verb, "label") == 0) {
1716 gctl_error(req, "Invalid number of arguments.");
1719 volname = gctl_get_asciiparam(req, "arg1");
1720 if (volname == NULL) {
1721 gctl_error(req, "No volume name.");
1724 levelname = gctl_get_asciiparam(req, "arg2");
1725 if (levelname == NULL) {
1726 gctl_error(req, "No RAID level.");
1729 if (strcasecmp(levelname, "RAID5") == 0)
1730 levelname = "RAID5-LA";
1731 if (g_raid_volume_str2level(levelname, &level, &qual)) {
1732 gctl_error(req, "Unknown RAID level '%s'.", levelname);
1735 numdisks = *nargs - 3;
1736 force = gctl_get_paraml(req, "force", sizeof(*force));
1737 if (!g_raid_md_intel_supported(level, qual, numdisks,
1738 force ? *force : 0)) {
1739 gctl_error(req, "Unsupported RAID level "
1740 "(0x%02x/0x%02x), or number of disks (%d).",
1741 level, qual, numdisks);
1745 /* Search for disks, connect them and probe. */
1746 size = 0x7fffffffffffffffllu;
1748 for (i = 0; i < numdisks; i++) {
1749 snprintf(arg, sizeof(arg), "arg%d", i + 3);
1750 diskname = gctl_get_asciiparam(req, arg);
1751 if (diskname == NULL) {
1752 gctl_error(req, "No disk name (%s).", arg);
1756 if (strcmp(diskname, "NONE") == 0) {
1761 cp = g_raid_open_consumer(sc, diskname);
1763 gctl_error(req, "Can't open disk '%s'.",
1765 g_topology_unlock();
1771 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO);
1772 pd->pd_disk_pos = i;
1773 disk = g_raid_create_disk(sc);
1774 disk->d_md_data = (void *)pd;
1775 disk->d_consumer = cp;
1777 strcpy(&pd->pd_disk_meta.serial[0], "NONE");
1778 pd->pd_disk_meta.id = 0xffffffff;
1779 pd->pd_disk_meta.flags = INTEL_F_ASSIGNED;
1783 g_topology_unlock();
1785 error = g_raid_md_get_label(cp,
1786 &pd->pd_disk_meta.serial[0], INTEL_SERIAL_LEN);
1789 "Can't get serial for provider '%s'.",
1795 g_raid_get_disk_info(disk);
1797 intel_set_disk_sectors(&pd->pd_disk_meta,
1798 pp->mediasize / pp->sectorsize);
1799 if (size > pp->mediasize)
1800 size = pp->mediasize;
1801 if (sectorsize < pp->sectorsize)
1802 sectorsize = pp->sectorsize;
1803 pd->pd_disk_meta.id = 0;
1804 pd->pd_disk_meta.flags = INTEL_F_ASSIGNED | INTEL_F_ONLINE;
1809 if (sectorsize <= 0) {
1810 gctl_error(req, "Can't get sector size.");
1814 /* Reserve some space for metadata. */
1815 size -= ((4096 + sectorsize - 1) / sectorsize) * sectorsize;
1817 /* Handle size argument. */
1818 len = sizeof(*sizearg);
1819 sizearg = gctl_get_param(req, "size", &len);
1820 if (sizearg != NULL && len == sizeof(*sizearg) &&
1822 if (*sizearg > size) {
1823 gctl_error(req, "Size too big %lld > %lld.",
1824 (long long)*sizearg, (long long)size);
1830 /* Handle strip argument. */
1832 len = sizeof(*striparg);
1833 striparg = gctl_get_param(req, "strip", &len);
1834 if (striparg != NULL && len == sizeof(*striparg) &&
1836 if (*striparg < sectorsize) {
1837 gctl_error(req, "Strip size too small.");
1840 if (*striparg % sectorsize != 0) {
1841 gctl_error(req, "Incorrect strip size.");
1844 if (strip > 65535 * sectorsize) {
1845 gctl_error(req, "Strip size too big.");
1851 /* Round size down to strip or sector. */
1852 if (level == G_RAID_VOLUME_RL_RAID1)
1853 size -= (size % sectorsize);
1854 else if (level == G_RAID_VOLUME_RL_RAID1E &&
1855 (numdisks & 1) != 0)
1856 size -= (size % (2 * strip));
1858 size -= (size % strip);
1860 gctl_error(req, "Size too small.");
1864 /* We have all we need, create things: volume, ... */
1865 mdi->mdio_started = 1;
1866 vol = g_raid_create_volume(sc, volname, -1);
1867 pv = malloc(sizeof(*pv), M_MD_INTEL, M_WAITOK | M_ZERO);
1868 pv->pv_volume_pos = 0;
1869 vol->v_md_data = pv;
1870 vol->v_raid_level = level;
1871 vol->v_raid_level_qualifier = qual;
1872 vol->v_strip_size = strip;
1873 vol->v_disks_count = numdisks;
1874 if (level == G_RAID_VOLUME_RL_RAID0)
1875 vol->v_mediasize = size * numdisks;
1876 else if (level == G_RAID_VOLUME_RL_RAID1)
1877 vol->v_mediasize = size;
1878 else if (level == G_RAID_VOLUME_RL_RAID5)
1879 vol->v_mediasize = size * (numdisks - 1);
1881 vol->v_mediasize = ((size * numdisks) / strip / 2) *
1884 vol->v_sectorsize = sectorsize;
1885 g_raid_start_volume(vol);
1887 /* , and subdisks. */
1888 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1889 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
1890 sd = &vol->v_subdisks[pd->pd_disk_pos];
1894 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1895 if (sd->sd_disk->d_consumer != NULL) {
1896 g_raid_change_disk_state(disk,
1897 G_RAID_DISK_S_ACTIVE);
1898 if (level == G_RAID_VOLUME_RL_RAID5)
1899 g_raid_change_subdisk_state(sd,
1900 G_RAID_SUBDISK_S_UNINITIALIZED);
1902 g_raid_change_subdisk_state(sd,
1903 G_RAID_SUBDISK_S_ACTIVE);
1904 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
1905 G_RAID_EVENT_SUBDISK);
1907 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1911 /* Write metadata based on created entities. */
1912 G_RAID_DEBUG1(0, sc, "Array started.");
1913 g_raid_md_write_intel(md, NULL, NULL, NULL);
1915 /* Pickup any STALE/SPARE disks to refill array if needed. */
1916 g_raid_md_intel_refill(sc);
1918 g_raid_event_send(vol, G_RAID_VOLUME_E_START,
1919 G_RAID_EVENT_VOLUME);
1922 if (strcmp(verb, "add") == 0) {
1925 gctl_error(req, "Invalid number of arguments.");
1928 volname = gctl_get_asciiparam(req, "arg1");
1929 if (volname == NULL) {
1930 gctl_error(req, "No volume name.");
1933 levelname = gctl_get_asciiparam(req, "arg2");
1934 if (levelname == NULL) {
1935 gctl_error(req, "No RAID level.");
1938 if (strcasecmp(levelname, "RAID5") == 0)
1939 levelname = "RAID5-LA";
1940 if (g_raid_volume_str2level(levelname, &level, &qual)) {
1941 gctl_error(req, "Unknown RAID level '%s'.", levelname);
1945 /* Look for existing volumes. */
1948 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
1953 gctl_error(req, "Maximum two volumes supported.");
1957 gctl_error(req, "At least one volume must exist.");
1961 numdisks = vol1->v_disks_count;
1962 force = gctl_get_paraml(req, "force", sizeof(*force));
1963 if (!g_raid_md_intel_supported(level, qual, numdisks,
1964 force ? *force : 0)) {
1965 gctl_error(req, "Unsupported RAID level "
1966 "(0x%02x/0x%02x), or number of disks (%d).",
1967 level, qual, numdisks);
1971 /* Collect info about present disks. */
1972 size = 0x7fffffffffffffffllu;
1974 for (i = 0; i < numdisks; i++) {
1975 disk = vol1->v_subdisks[i].sd_disk;
1976 pd = (struct g_raid_md_intel_perdisk *)
1979 intel_get_disk_sectors(&pd->pd_disk_meta);
1981 if (disk_sectors * 512 < size)
1982 size = disk_sectors * 512;
1983 if (disk->d_consumer != NULL &&
1984 disk->d_consumer->provider != NULL &&
1985 disk->d_consumer->provider->sectorsize >
1988 disk->d_consumer->provider->sectorsize;
1992 /* Reserve some space for metadata. */
1993 size -= ((4096 + sectorsize - 1) / sectorsize) * sectorsize;
1995 /* Decide insert before or after. */
1996 sd = &vol1->v_subdisks[0];
1998 size - (sd->sd_offset + sd->sd_size)) {
2000 size = sd->sd_offset;
2002 off = sd->sd_offset + sd->sd_size;
2003 size = size - (sd->sd_offset + sd->sd_size);
2006 /* Handle strip argument. */
2008 len = sizeof(*striparg);
2009 striparg = gctl_get_param(req, "strip", &len);
2010 if (striparg != NULL && len == sizeof(*striparg) &&
2012 if (*striparg < sectorsize) {
2013 gctl_error(req, "Strip size too small.");
2016 if (*striparg % sectorsize != 0) {
2017 gctl_error(req, "Incorrect strip size.");
2020 if (strip > 65535 * sectorsize) {
2021 gctl_error(req, "Strip size too big.");
2027 /* Round offset up to strip. */
2028 if (off % strip != 0) {
2029 size -= strip - off % strip;
2030 off += strip - off % strip;
2033 /* Handle size argument. */
2034 len = sizeof(*sizearg);
2035 sizearg = gctl_get_param(req, "size", &len);
2036 if (sizearg != NULL && len == sizeof(*sizearg) &&
2038 if (*sizearg > size) {
2039 gctl_error(req, "Size too big %lld > %lld.",
2040 (long long)*sizearg, (long long)size);
2046 /* Round size down to strip or sector. */
2047 if (level == G_RAID_VOLUME_RL_RAID1)
2048 size -= (size % sectorsize);
2050 size -= (size % strip);
2052 gctl_error(req, "Size too small.");
2055 if (size > 0xffffffffllu * sectorsize) {
2056 gctl_error(req, "Size too big.");
2060 /* We have all we need, create things: volume, ... */
2061 vol = g_raid_create_volume(sc, volname, -1);
2062 pv = malloc(sizeof(*pv), M_MD_INTEL, M_WAITOK | M_ZERO);
2063 pv->pv_volume_pos = i;
2064 vol->v_md_data = pv;
2065 vol->v_raid_level = level;
2066 vol->v_raid_level_qualifier = qual;
2067 vol->v_strip_size = strip;
2068 vol->v_disks_count = numdisks;
2069 if (level == G_RAID_VOLUME_RL_RAID0)
2070 vol->v_mediasize = size * numdisks;
2071 else if (level == G_RAID_VOLUME_RL_RAID1)
2072 vol->v_mediasize = size;
2073 else if (level == G_RAID_VOLUME_RL_RAID5)
2074 vol->v_mediasize = size * (numdisks - 1);
2076 vol->v_mediasize = ((size * numdisks) / strip / 2) *
2079 vol->v_sectorsize = sectorsize;
2080 g_raid_start_volume(vol);
2082 /* , and subdisks. */
2083 for (i = 0; i < numdisks; i++) {
2084 disk = vol1->v_subdisks[i].sd_disk;
2085 sd = &vol->v_subdisks[i];
2087 sd->sd_offset = off;
2089 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
2090 if (disk->d_state == G_RAID_DISK_S_ACTIVE) {
2091 if (level == G_RAID_VOLUME_RL_RAID5)
2092 g_raid_change_subdisk_state(sd,
2093 G_RAID_SUBDISK_S_UNINITIALIZED);
2095 g_raid_change_subdisk_state(sd,
2096 G_RAID_SUBDISK_S_ACTIVE);
2097 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
2098 G_RAID_EVENT_SUBDISK);
2102 /* Write metadata based on created entities. */
2103 g_raid_md_write_intel(md, NULL, NULL, NULL);
2105 g_raid_event_send(vol, G_RAID_VOLUME_E_START,
2106 G_RAID_EVENT_VOLUME);
2109 if (strcmp(verb, "delete") == 0) {
2111 nodename = gctl_get_asciiparam(req, "arg0");
2112 if (nodename != NULL && strcasecmp(sc->sc_name, nodename) != 0)
2115 /* Full node destruction. */
2116 if (*nargs == 1 && nodename != NULL) {
2117 /* Check if some volume is still open. */
2118 force = gctl_get_paraml(req, "force", sizeof(*force));
2119 if (force != NULL && *force == 0 &&
2120 g_raid_nopens(sc) != 0) {
2121 gctl_error(req, "Some volume is still open.");
2125 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2126 if (disk->d_consumer)
2127 intel_meta_erase(disk->d_consumer);
2129 g_raid_destroy_node(sc, 0);
2133 /* Destroy specified volume. If it was last - all node. */
2135 gctl_error(req, "Invalid number of arguments.");
2138 volname = gctl_get_asciiparam(req,
2139 nodename != NULL ? "arg1" : "arg0");
2140 if (volname == NULL) {
2141 gctl_error(req, "No volume name.");
2145 /* Search for volume. */
2146 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2147 if (strcmp(vol->v_name, volname) == 0)
2149 pp = vol->v_provider;
2152 if (strcmp(pp->name, volname) == 0)
2154 if (strncmp(pp->name, "raid/", 5) == 0 &&
2155 strcmp(pp->name + 5, volname) == 0)
2159 i = strtol(volname, &tmp, 10);
2160 if (verb != volname && tmp[0] == 0) {
2161 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2162 if (vol->v_global_id == i)
2168 gctl_error(req, "Volume '%s' not found.", volname);
2172 /* Check if volume is still open. */
2173 force = gctl_get_paraml(req, "force", sizeof(*force));
2174 if (force != NULL && *force == 0 &&
2175 vol->v_provider_open != 0) {
2176 gctl_error(req, "Volume is still open.");
2180 /* Destroy volume and potentially node. */
2182 TAILQ_FOREACH(vol1, &sc->sc_volumes, v_next)
2185 g_raid_destroy_volume(vol);
2186 g_raid_md_write_intel(md, NULL, NULL, NULL);
2188 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2189 if (disk->d_consumer)
2190 intel_meta_erase(disk->d_consumer);
2192 g_raid_destroy_node(sc, 0);
2196 if (strcmp(verb, "remove") == 0 ||
2197 strcmp(verb, "fail") == 0) {
2199 gctl_error(req, "Invalid number of arguments.");
2202 for (i = 1; i < *nargs; i++) {
2203 snprintf(arg, sizeof(arg), "arg%d", i);
2204 diskname = gctl_get_asciiparam(req, arg);
2205 if (diskname == NULL) {
2206 gctl_error(req, "No disk name (%s).", arg);
2210 if (strncmp(diskname, "/dev/", 5) == 0)
2213 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2214 if (disk->d_consumer != NULL &&
2215 disk->d_consumer->provider != NULL &&
2216 strcmp(disk->d_consumer->provider->name,
2221 gctl_error(req, "Disk '%s' not found.",
2227 if (strcmp(verb, "fail") == 0) {
2228 g_raid_md_fail_disk_intel(md, NULL, disk);
2232 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2234 /* Erase metadata on deleting disk. */
2235 intel_meta_erase(disk->d_consumer);
2237 /* If disk was assigned, just update statuses. */
2238 if (pd->pd_disk_pos >= 0) {
2239 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
2240 g_raid_kill_consumer(sc, disk->d_consumer);
2241 disk->d_consumer = NULL;
2242 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
2243 g_raid_change_subdisk_state(sd,
2244 G_RAID_SUBDISK_S_NONE);
2245 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
2246 G_RAID_EVENT_SUBDISK);
2249 /* Otherwise -- delete. */
2250 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
2251 g_raid_destroy_disk(disk);
2255 /* Write updated metadata to remaining disks. */
2256 g_raid_md_write_intel(md, NULL, NULL, NULL);
2258 /* Check if anything left except placeholders. */
2259 if (g_raid_ndisks(sc, -1) ==
2260 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
2261 g_raid_destroy_node(sc, 0);
2263 g_raid_md_intel_refill(sc);
2266 if (strcmp(verb, "insert") == 0) {
2268 gctl_error(req, "Invalid number of arguments.");
2272 for (i = 1; i < *nargs; i++) {
2273 /* Get disk name. */
2274 snprintf(arg, sizeof(arg), "arg%d", i);
2275 diskname = gctl_get_asciiparam(req, arg);
2276 if (diskname == NULL) {
2277 gctl_error(req, "No disk name (%s).", arg);
2282 /* Try to find provider with specified name. */
2284 cp = g_raid_open_consumer(sc, diskname);
2286 gctl_error(req, "Can't open disk '%s'.",
2288 g_topology_unlock();
2293 g_topology_unlock();
2295 /* Read disk serial. */
2296 error = g_raid_md_get_label(cp,
2297 &serial[0], INTEL_SERIAL_LEN);
2300 "Can't get serial for provider '%s'.",
2302 g_raid_kill_consumer(sc, cp);
2307 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO);
2308 pd->pd_disk_pos = -1;
2310 disk = g_raid_create_disk(sc);
2311 disk->d_consumer = cp;
2312 disk->d_md_data = (void *)pd;
2315 g_raid_get_disk_info(disk);
2317 memcpy(&pd->pd_disk_meta.serial[0], &serial[0],
2319 intel_set_disk_sectors(&pd->pd_disk_meta,
2320 pp->mediasize / pp->sectorsize);
2321 pd->pd_disk_meta.id = 0;
2322 pd->pd_disk_meta.flags = INTEL_F_SPARE;
2324 /* Welcome the "new" disk. */
2325 update += g_raid_md_intel_start_disk(disk);
2326 if (disk->d_state == G_RAID_DISK_S_SPARE) {
2327 intel_meta_write_spare(cp, &pd->pd_disk_meta);
2328 g_raid_destroy_disk(disk);
2329 } else if (disk->d_state != G_RAID_DISK_S_ACTIVE) {
2330 gctl_error(req, "Disk '%s' doesn't fit.",
2332 g_raid_destroy_disk(disk);
2338 /* Write new metadata if we changed something. */
2340 g_raid_md_write_intel(md, NULL, NULL, NULL);
2347 g_raid_md_write_intel(struct g_raid_md_object *md, struct g_raid_volume *tvol,
2348 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
2350 struct g_raid_softc *sc;
2351 struct g_raid_volume *vol;
2352 struct g_raid_subdisk *sd;
2353 struct g_raid_disk *disk;
2354 struct g_raid_md_intel_object *mdi;
2355 struct g_raid_md_intel_pervolume *pv;
2356 struct g_raid_md_intel_perdisk *pd;
2357 struct intel_raid_conf *meta;
2358 struct intel_raid_vol *mvol;
2359 struct intel_raid_map *mmap0, *mmap1;
2360 off_t sectorsize = 512, pos;
2361 const char *version, *cv;
2362 int vi, sdi, numdisks, len, state, stale;
2365 mdi = (struct g_raid_md_intel_object *)md;
2367 if (sc->sc_stopping == G_RAID_DESTROY_HARD)
2370 /* Bump generation. Newly written metadata may differ from previous. */
2371 mdi->mdio_generation++;
2373 /* Count number of disks. */
2375 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2376 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2377 if (pd->pd_disk_pos < 0)
2380 if (disk->d_state == G_RAID_DISK_S_ACTIVE) {
2381 pd->pd_disk_meta.flags =
2382 INTEL_F_ONLINE | INTEL_F_ASSIGNED;
2383 } else if (disk->d_state == G_RAID_DISK_S_FAILED) {
2384 pd->pd_disk_meta.flags = INTEL_F_FAILED |
2386 } else if (disk->d_state == G_RAID_DISK_S_DISABLED) {
2387 pd->pd_disk_meta.flags = INTEL_F_FAILED |
2388 INTEL_F_ASSIGNED | INTEL_F_DISABLED;
2390 if (!(pd->pd_disk_meta.flags & INTEL_F_DISABLED))
2391 pd->pd_disk_meta.flags = INTEL_F_ASSIGNED;
2392 if (pd->pd_disk_meta.id != 0xffffffff) {
2393 pd->pd_disk_meta.id = 0xffffffff;
2394 len = strlen(pd->pd_disk_meta.serial);
2395 len = min(len, INTEL_SERIAL_LEN - 3);
2396 strcpy(pd->pd_disk_meta.serial + len, ":0");
2401 /* Fill anchor and disks. */
2402 meta = malloc(INTEL_MAX_MD_SIZE(numdisks),
2403 M_MD_INTEL, M_WAITOK | M_ZERO);
2404 memcpy(&meta->intel_id[0], INTEL_MAGIC, sizeof(INTEL_MAGIC) - 1);
2405 meta->config_size = INTEL_MAX_MD_SIZE(numdisks);
2406 meta->config_id = mdi->mdio_config_id;
2407 meta->orig_config_id = mdi->mdio_orig_config_id;
2408 meta->generation = mdi->mdio_generation;
2409 meta->attributes = INTEL_ATTR_CHECKSUM;
2410 meta->total_disks = numdisks;
2411 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2412 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2413 if (pd->pd_disk_pos < 0)
2415 meta->disk[pd->pd_disk_pos] = pd->pd_disk_meta;
2416 if (pd->pd_disk_meta.sectors_hi != 0)
2417 meta->attributes |= INTEL_ATTR_2TB_DISK;
2420 /* Fill volumes and maps. */
2422 version = INTEL_VERSION_1000;
2423 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2424 pv = vol->v_md_data;
2425 if (vol->v_stopping)
2427 mvol = intel_get_volume(meta, vi);
2429 /* New metadata may have different volumes order. */
2430 pv->pv_volume_pos = vi;
2432 for (sdi = 0; sdi < vol->v_disks_count; sdi++) {
2433 sd = &vol->v_subdisks[sdi];
2434 if (sd->sd_disk != NULL)
2437 if (sdi >= vol->v_disks_count)
2438 panic("No any filled subdisk in volume");
2439 if (vol->v_mediasize >= 0x20000000000llu)
2440 meta->attributes |= INTEL_ATTR_2TB;
2441 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0)
2442 meta->attributes |= INTEL_ATTR_RAID0;
2443 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
2444 meta->attributes |= INTEL_ATTR_RAID1;
2445 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID5)
2446 meta->attributes |= INTEL_ATTR_RAID5;
2447 else if ((vol->v_disks_count & 1) == 0)
2448 meta->attributes |= INTEL_ATTR_RAID10;
2450 meta->attributes |= INTEL_ATTR_RAID1E;
2452 meta->attributes |= INTEL_ATTR_RAIDCNG;
2453 if (vol->v_strip_size > 131072)
2454 meta->attributes |= INTEL_ATTR_EXT_STRIP;
2457 cv = INTEL_VERSION_1206;
2458 else if (vol->v_disks_count > 4)
2459 cv = INTEL_VERSION_1204;
2460 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID5)
2461 cv = INTEL_VERSION_1202;
2462 else if (vol->v_disks_count > 2)
2463 cv = INTEL_VERSION_1201;
2464 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
2465 cv = INTEL_VERSION_1100;
2467 cv = INTEL_VERSION_1000;
2468 if (strcmp(cv, version) > 0)
2471 strlcpy(&mvol->name[0], vol->v_name, sizeof(mvol->name));
2472 mvol->total_sectors = vol->v_mediasize / sectorsize;
2473 mvol->state = (INTEL_ST_READ_COALESCING |
2474 INTEL_ST_WRITE_COALESCING);
2475 mvol->tid = vol->v_global_id + 1;
2477 mvol->state |= INTEL_ST_CLONE_N_GO;
2478 if (pv->pv_cng_man_sync)
2479 mvol->state |= INTEL_ST_CLONE_MAN_SYNC;
2480 mvol->cng_master_disk = pv->pv_cng_master_disk;
2481 if (vol->v_subdisks[pv->pv_cng_master_disk].sd_state ==
2482 G_RAID_SUBDISK_S_NONE)
2483 mvol->cng_state = INTEL_CNGST_MASTER_MISSING;
2484 else if (vol->v_state != G_RAID_VOLUME_S_OPTIMAL)
2485 mvol->cng_state = INTEL_CNGST_NEEDS_UPDATE;
2487 mvol->cng_state = INTEL_CNGST_UPDATED;
2490 /* Check for any recovery in progress. */
2491 state = G_RAID_SUBDISK_S_ACTIVE;
2492 pos = 0x7fffffffffffffffllu;
2494 for (sdi = 0; sdi < vol->v_disks_count; sdi++) {
2495 sd = &vol->v_subdisks[sdi];
2496 if (sd->sd_state == G_RAID_SUBDISK_S_REBUILD)
2497 state = G_RAID_SUBDISK_S_REBUILD;
2498 else if (sd->sd_state == G_RAID_SUBDISK_S_RESYNC &&
2499 state != G_RAID_SUBDISK_S_REBUILD)
2500 state = G_RAID_SUBDISK_S_RESYNC;
2501 else if (sd->sd_state == G_RAID_SUBDISK_S_STALE)
2503 if ((sd->sd_state == G_RAID_SUBDISK_S_REBUILD ||
2504 sd->sd_state == G_RAID_SUBDISK_S_RESYNC) &&
2505 sd->sd_rebuild_pos < pos)
2506 pos = sd->sd_rebuild_pos;
2508 if (state == G_RAID_SUBDISK_S_REBUILD) {
2509 mvol->migr_state = 1;
2510 mvol->migr_type = INTEL_MT_REBUILD;
2511 } else if (state == G_RAID_SUBDISK_S_RESYNC) {
2512 mvol->migr_state = 1;
2513 /* mvol->migr_type = INTEL_MT_REPAIR; */
2514 mvol->migr_type = INTEL_MT_VERIFY;
2515 mvol->state |= INTEL_ST_VERIFY_AND_FIX;
2517 mvol->migr_state = 0;
2518 mvol->dirty = (vol->v_dirty || stale);
2520 mmap0 = intel_get_map(mvol, 0);
2522 /* Write map / common part of two maps. */
2523 intel_set_map_offset(mmap0, sd->sd_offset / sectorsize);
2524 intel_set_map_disk_sectors(mmap0, sd->sd_size / sectorsize);
2525 mmap0->strip_sectors = vol->v_strip_size / sectorsize;
2526 if (vol->v_state == G_RAID_VOLUME_S_BROKEN)
2527 mmap0->status = INTEL_S_FAILURE;
2528 else if (vol->v_state == G_RAID_VOLUME_S_DEGRADED)
2529 mmap0->status = INTEL_S_DEGRADED;
2530 else if (g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_UNINITIALIZED)
2531 == g_raid_nsubdisks(vol, -1))
2532 mmap0->status = INTEL_S_UNINITIALIZED;
2534 mmap0->status = INTEL_S_READY;
2535 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0)
2536 mmap0->type = INTEL_T_RAID0;
2537 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1 ||
2538 vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E)
2539 mmap0->type = INTEL_T_RAID1;
2541 mmap0->type = INTEL_T_RAID5;
2542 mmap0->total_disks = vol->v_disks_count;
2543 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
2544 mmap0->total_domains = vol->v_disks_count;
2545 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E)
2546 mmap0->total_domains = 2;
2548 mmap0->total_domains = 1;
2549 intel_set_map_stripe_count(mmap0,
2550 sd->sd_size / vol->v_strip_size / mmap0->total_domains);
2551 mmap0->failed_disk_num = 0xff;
2554 /* If there are two maps - copy common and update. */
2555 if (mvol->migr_state) {
2556 intel_set_vol_curr_migr_unit(mvol,
2557 pos / vol->v_strip_size / mmap0->total_domains);
2558 mmap1 = intel_get_map(mvol, 1);
2559 memcpy(mmap1, mmap0, sizeof(struct intel_raid_map));
2560 mmap0->status = INTEL_S_READY;
2564 /* Write disk indexes and put rebuild flags. */
2565 for (sdi = 0; sdi < vol->v_disks_count; sdi++) {
2566 sd = &vol->v_subdisks[sdi];
2567 pd = (struct g_raid_md_intel_perdisk *)
2568 sd->sd_disk->d_md_data;
2569 mmap0->disk_idx[sdi] = pd->pd_disk_pos;
2570 if (mvol->migr_state)
2571 mmap1->disk_idx[sdi] = pd->pd_disk_pos;
2572 if (sd->sd_state == G_RAID_SUBDISK_S_REBUILD ||
2573 sd->sd_state == G_RAID_SUBDISK_S_RESYNC) {
2574 mmap1->disk_idx[sdi] |= INTEL_DI_RBLD;
2575 } else if (sd->sd_state != G_RAID_SUBDISK_S_ACTIVE &&
2576 sd->sd_state != G_RAID_SUBDISK_S_STALE &&
2577 sd->sd_state != G_RAID_SUBDISK_S_UNINITIALIZED) {
2578 mmap0->disk_idx[sdi] |= INTEL_DI_RBLD;
2579 if (mvol->migr_state)
2580 mmap1->disk_idx[sdi] |= INTEL_DI_RBLD;
2582 if ((sd->sd_state == G_RAID_SUBDISK_S_NONE ||
2583 sd->sd_state == G_RAID_SUBDISK_S_FAILED ||
2584 sd->sd_state == G_RAID_SUBDISK_S_REBUILD) &&
2585 mmap0->failed_disk_num == 0xff) {
2586 mmap0->failed_disk_num = sdi;
2587 if (mvol->migr_state)
2588 mmap1->failed_disk_num = sdi;
2593 meta->total_volumes = vi;
2594 if (vi > 1 || meta->attributes &
2595 (INTEL_ATTR_EXT_STRIP | INTEL_ATTR_2TB_DISK | INTEL_ATTR_2TB))
2596 version = INTEL_VERSION_1300;
2597 if (strcmp(version, INTEL_VERSION_1300) < 0)
2598 meta->attributes &= INTEL_ATTR_CHECKSUM;
2599 memcpy(&meta->version[0], version, sizeof(INTEL_VERSION_1000) - 1);
2601 /* We are done. Print meta data and store them to disks. */
2602 g_raid_md_intel_print(meta);
2603 if (mdi->mdio_meta != NULL)
2604 free(mdi->mdio_meta, M_MD_INTEL);
2605 mdi->mdio_meta = meta;
2606 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2607 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2608 if (disk->d_state != G_RAID_DISK_S_ACTIVE)
2610 if (pd->pd_meta != NULL) {
2611 free(pd->pd_meta, M_MD_INTEL);
2614 pd->pd_meta = intel_meta_copy(meta);
2615 intel_meta_write(disk->d_consumer, meta);
2621 g_raid_md_fail_disk_intel(struct g_raid_md_object *md,
2622 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
2624 struct g_raid_softc *sc;
2625 struct g_raid_md_intel_object *mdi;
2626 struct g_raid_md_intel_perdisk *pd;
2627 struct g_raid_subdisk *sd;
2630 mdi = (struct g_raid_md_intel_object *)md;
2631 pd = (struct g_raid_md_intel_perdisk *)tdisk->d_md_data;
2633 /* We can't fail disk that is not a part of array now. */
2634 if (pd->pd_disk_pos < 0)
2638 * Mark disk as failed in metadata and try to write that metadata
2639 * to the disk itself to prevent it's later resurrection as STALE.
2641 mdi->mdio_meta->disk[pd->pd_disk_pos].flags = INTEL_F_FAILED;
2642 pd->pd_disk_meta.flags = INTEL_F_FAILED;
2643 g_raid_md_intel_print(mdi->mdio_meta);
2644 if (tdisk->d_consumer)
2645 intel_meta_write(tdisk->d_consumer, mdi->mdio_meta);
2647 /* Change states. */
2648 g_raid_change_disk_state(tdisk, G_RAID_DISK_S_FAILED);
2649 TAILQ_FOREACH(sd, &tdisk->d_subdisks, sd_next) {
2650 g_raid_change_subdisk_state(sd,
2651 G_RAID_SUBDISK_S_FAILED);
2652 g_raid_event_send(sd, G_RAID_SUBDISK_E_FAILED,
2653 G_RAID_EVENT_SUBDISK);
2656 /* Write updated metadata to remaining disks. */
2657 g_raid_md_write_intel(md, NULL, NULL, tdisk);
2659 /* Check if anything left except placeholders. */
2660 if (g_raid_ndisks(sc, -1) ==
2661 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
2662 g_raid_destroy_node(sc, 0);
2664 g_raid_md_intel_refill(sc);
2669 g_raid_md_free_disk_intel(struct g_raid_md_object *md,
2670 struct g_raid_disk *disk)
2672 struct g_raid_md_intel_perdisk *pd;
2674 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2675 if (pd->pd_meta != NULL) {
2676 free(pd->pd_meta, M_MD_INTEL);
2679 free(pd, M_MD_INTEL);
2680 disk->d_md_data = NULL;
2685 g_raid_md_free_volume_intel(struct g_raid_md_object *md,
2686 struct g_raid_volume *vol)
2688 struct g_raid_md_intel_pervolume *pv;
2690 pv = (struct g_raid_md_intel_pervolume *)vol->v_md_data;
2691 free(pv, M_MD_INTEL);
2692 vol->v_md_data = NULL;
2697 g_raid_md_free_intel(struct g_raid_md_object *md)
2699 struct g_raid_md_intel_object *mdi;
2701 mdi = (struct g_raid_md_intel_object *)md;
2702 if (!mdi->mdio_started) {
2703 mdi->mdio_started = 0;
2704 callout_stop(&mdi->mdio_start_co);
2705 G_RAID_DEBUG1(1, md->mdo_softc,
2706 "root_mount_rel %p", mdi->mdio_rootmount);
2707 root_mount_rel(mdi->mdio_rootmount);
2708 mdi->mdio_rootmount = NULL;
2710 if (mdi->mdio_meta != NULL) {
2711 free(mdi->mdio_meta, M_MD_INTEL);
2712 mdi->mdio_meta = NULL;
2717 G_RAID_MD_DECLARE(intel, "Intel");