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/geom_dbg.h>
47 #include "geom/raid/g_raid.h"
48 #include "g_raid_md_if.h"
50 static MALLOC_DEFINE(M_MD_INTEL, "md_intel_data", "GEOM_RAID Intel metadata");
52 struct intel_raid_map {
54 uint32_t disk_sectors;
55 uint32_t stripe_count;
56 uint16_t strip_sectors;
58 #define INTEL_S_READY 0x00
59 #define INTEL_S_UNINITIALIZED 0x01
60 #define INTEL_S_DEGRADED 0x02
61 #define INTEL_S_FAILURE 0x03
64 #define INTEL_T_RAID0 0x00
65 #define INTEL_T_RAID1 0x01
66 #define INTEL_T_RAID5 0x05
69 uint8_t total_domains;
70 uint8_t failed_disk_num;
73 uint32_t disk_sectors_hi;
74 uint32_t stripe_count_hi;
76 uint32_t disk_idx[1]; /* total_disks entries. */
77 #define INTEL_DI_IDX 0x00ffffff
78 #define INTEL_DI_RBLD 0x01000000
81 struct intel_raid_vol {
83 u_int64_t total_sectors __packed;
85 #define INTEL_ST_BOOTABLE 0x00000001
86 #define INTEL_ST_BOOT_DEVICE 0x00000002
87 #define INTEL_ST_READ_COALESCING 0x00000004
88 #define INTEL_ST_WRITE_COALESCING 0x00000008
89 #define INTEL_ST_LAST_SHUTDOWN_DIRTY 0x00000010
90 #define INTEL_ST_HIDDEN_AT_BOOT 0x00000020
91 #define INTEL_ST_CURRENTLY_HIDDEN 0x00000040
92 #define INTEL_ST_VERIFY_AND_FIX 0x00000080
93 #define INTEL_ST_MAP_STATE_UNINIT 0x00000100
94 #define INTEL_ST_NO_AUTO_RECOVERY 0x00000200
95 #define INTEL_ST_CLONE_N_GO 0x00000400
96 #define INTEL_ST_CLONE_MAN_SYNC 0x00000800
97 #define INTEL_ST_CNG_MASTER_DISK_NUM 0x00001000
99 uint8_t migr_priority;
100 uint8_t num_sub_vols;
102 uint8_t cng_master_disk;
103 uint16_t cache_policy;
105 #define INTEL_CNGST_UPDATED 0
106 #define INTEL_CNGST_NEEDS_UPDATE 1
107 #define INTEL_CNGST_MASTER_MISSING 2
108 uint8_t cng_sub_state;
109 uint32_t filler_0[10];
111 uint32_t curr_migr_unit;
112 uint32_t checkpoint_id;
115 #define INTEL_MT_INIT 0
116 #define INTEL_MT_REBUILD 1
117 #define INTEL_MT_VERIFY 2
118 #define INTEL_MT_GEN_MIGR 3
119 #define INTEL_MT_STATE_CHANGE 4
120 #define INTEL_MT_REPAIR 5
123 uint16_t verify_errors;
125 uint32_t curr_migr_unit_hi;
126 uint32_t filler_1[3];
127 struct intel_raid_map map[1]; /* 2 entries if migr_state != 0. */
130 struct intel_raid_disk {
131 #define INTEL_SERIAL_LEN 16
132 uint8_t serial[INTEL_SERIAL_LEN];
136 #define INTEL_F_SPARE 0x01
137 #define INTEL_F_ASSIGNED 0x02
138 #define INTEL_F_FAILED 0x04
139 #define INTEL_F_ONLINE 0x08
140 #define INTEL_F_DISABLED 0x80
141 uint32_t owner_cfg_num;
146 struct intel_raid_conf {
147 uint8_t intel_id[24];
148 #define INTEL_MAGIC "Intel Raid ISM Cfg Sig. "
151 #define INTEL_VERSION_1000 "1.0.00" /* RAID0 */
152 #define INTEL_VERSION_1100 "1.1.00" /* RAID1 */
153 #define INTEL_VERSION_1200 "1.2.00" /* Many volumes */
154 #define INTEL_VERSION_1201 "1.2.01" /* 3 or 4 disks */
155 #define INTEL_VERSION_1202 "1.2.02" /* RAID5 */
156 #define INTEL_VERSION_1204 "1.2.04" /* 5 or 6 disks */
157 #define INTEL_VERSION_1206 "1.2.06" /* CNG */
158 #define INTEL_VERSION_1300 "1.3.00" /* Attributes */
162 uint32_t config_size;
165 uint32_t error_log_size;
167 #define INTEL_ATTR_RAID0 0x00000001
168 #define INTEL_ATTR_RAID1 0x00000002
169 #define INTEL_ATTR_RAID10 0x00000004
170 #define INTEL_ATTR_RAID1E 0x00000008
171 #define INTEL_ATTR_RAID5 0x00000010
172 #define INTEL_ATTR_RAIDCNG 0x00000020
173 #define INTEL_ATTR_EXT_STRIP 0x00000040
174 #define INTEL_ATTR_NVM_CACHE 0x02000000
175 #define INTEL_ATTR_2TB_DISK 0x04000000
176 #define INTEL_ATTR_BBM 0x08000000
177 #define INTEL_ATTR_NVM_CACHE2 0x10000000
178 #define INTEL_ATTR_2TB 0x20000000
179 #define INTEL_ATTR_PM 0x40000000
180 #define INTEL_ATTR_CHECKSUM 0x80000000
183 uint8_t total_volumes;
184 uint8_t error_log_pos;
187 uint32_t orig_config_id;
188 uint32_t pwr_cycle_count;
189 uint32_t bbm_log_size;
190 uint32_t filler_0[35];
191 struct intel_raid_disk disk[1]; /* total_disks entries. */
192 /* Here goes total_volumes of struct intel_raid_vol. */
195 #define INTEL_ATTR_SUPPORTED ( INTEL_ATTR_RAID0 | INTEL_ATTR_RAID1 | \
196 INTEL_ATTR_RAID10 | INTEL_ATTR_RAID1E | INTEL_ATTR_RAID5 | \
197 INTEL_ATTR_RAIDCNG | INTEL_ATTR_EXT_STRIP | INTEL_ATTR_2TB_DISK | \
198 INTEL_ATTR_2TB | INTEL_ATTR_PM | INTEL_ATTR_CHECKSUM )
200 #define INTEL_MAX_MD_SIZE(ndisks) \
201 (sizeof(struct intel_raid_conf) + \
202 sizeof(struct intel_raid_disk) * (ndisks - 1) + \
203 sizeof(struct intel_raid_vol) * 2 + \
204 sizeof(struct intel_raid_map) * 2 + \
205 sizeof(uint32_t) * (ndisks - 1) * 4)
207 struct g_raid_md_intel_perdisk {
208 struct intel_raid_conf *pd_meta;
210 struct intel_raid_disk pd_disk_meta;
213 struct g_raid_md_intel_pervolume {
217 int pv_cng_master_disk;
220 struct g_raid_md_intel_object {
221 struct g_raid_md_object mdio_base;
222 uint32_t mdio_config_id;
223 uint32_t mdio_orig_config_id;
224 uint32_t mdio_generation;
225 struct intel_raid_conf *mdio_meta;
226 struct callout mdio_start_co; /* STARTING state timer. */
227 int mdio_disks_present;
230 struct root_hold_token *mdio_rootmount; /* Root mount delay token. */
233 static g_raid_md_create_t g_raid_md_create_intel;
234 static g_raid_md_taste_t g_raid_md_taste_intel;
235 static g_raid_md_event_t g_raid_md_event_intel;
236 static g_raid_md_ctl_t g_raid_md_ctl_intel;
237 static g_raid_md_write_t g_raid_md_write_intel;
238 static g_raid_md_fail_disk_t g_raid_md_fail_disk_intel;
239 static g_raid_md_free_disk_t g_raid_md_free_disk_intel;
240 static g_raid_md_free_volume_t g_raid_md_free_volume_intel;
241 static g_raid_md_free_t g_raid_md_free_intel;
243 static kobj_method_t g_raid_md_intel_methods[] = {
244 KOBJMETHOD(g_raid_md_create, g_raid_md_create_intel),
245 KOBJMETHOD(g_raid_md_taste, g_raid_md_taste_intel),
246 KOBJMETHOD(g_raid_md_event, g_raid_md_event_intel),
247 KOBJMETHOD(g_raid_md_ctl, g_raid_md_ctl_intel),
248 KOBJMETHOD(g_raid_md_write, g_raid_md_write_intel),
249 KOBJMETHOD(g_raid_md_fail_disk, g_raid_md_fail_disk_intel),
250 KOBJMETHOD(g_raid_md_free_disk, g_raid_md_free_disk_intel),
251 KOBJMETHOD(g_raid_md_free_volume, g_raid_md_free_volume_intel),
252 KOBJMETHOD(g_raid_md_free, g_raid_md_free_intel),
256 static struct g_raid_md_class g_raid_md_intel_class = {
258 g_raid_md_intel_methods,
259 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;
596 if (pp->sectorsize < sizeof(*meta))
598 /* Read the anchor sector. */
599 buf = g_read_data(cp,
600 pp->mediasize - pp->sectorsize * 2, pp->sectorsize, &error);
602 G_RAID_DEBUG(1, "Cannot read metadata from %s (error=%d).",
606 meta = (struct intel_raid_conf *)buf;
608 /* Check if this is an Intel RAID struct */
609 if (strncmp(meta->intel_id, INTEL_MAGIC, strlen(INTEL_MAGIC))) {
610 G_RAID_DEBUG(1, "Intel signature check failed on %s", pp->name);
614 if (meta->config_size > 65536 ||
615 meta->config_size < sizeof(struct intel_raid_conf)) {
616 G_RAID_DEBUG(1, "Intel metadata size looks wrong: %d",
621 size = meta->config_size;
622 meta = malloc(size, M_MD_INTEL, M_WAITOK);
623 memcpy(meta, buf, min(size, pp->sectorsize));
626 /* Read all the rest, if needed. */
627 if (meta->config_size > pp->sectorsize) {
628 left = (meta->config_size - 1) / pp->sectorsize;
629 buf = g_read_data(cp,
630 pp->mediasize - pp->sectorsize * (2 + left),
631 pp->sectorsize * left, &error);
633 G_RAID_DEBUG(1, "Cannot read remaining metadata"
634 " part from %s (error=%d).",
636 free(meta, M_MD_INTEL);
639 memcpy(((char *)meta) + pp->sectorsize, buf,
640 pp->sectorsize * left);
644 /* Check metadata checksum. */
645 for (checksum = 0, ptr = (uint32_t *)meta, i = 0;
646 i < (meta->config_size / sizeof(uint32_t)); i++) {
649 checksum -= meta->checksum;
650 if (checksum != meta->checksum) {
651 G_RAID_DEBUG(1, "Intel checksum check failed on %s", pp->name);
652 free(meta, M_MD_INTEL);
656 /* Validate metadata size. */
657 size = sizeof(struct intel_raid_conf) +
658 sizeof(struct intel_raid_disk) * (meta->total_disks - 1) +
659 sizeof(struct intel_raid_vol) * meta->total_volumes;
660 if (size > meta->config_size) {
662 G_RAID_DEBUG(1, "Intel metadata size incorrect %d < %d",
663 meta->config_size, size);
664 free(meta, M_MD_INTEL);
667 for (i = 0; i < meta->total_volumes; i++) {
668 mvol = intel_get_volume(meta, i);
669 mmap = intel_get_map(mvol, 0);
670 size += 4 * (mmap->total_disks - 1);
671 if (size > meta->config_size)
673 if (mvol->migr_state) {
674 size += sizeof(struct intel_raid_map);
675 if (size > meta->config_size)
677 mmap = intel_get_map(mvol, 1);
678 size += 4 * (mmap->total_disks - 1);
679 if (size > meta->config_size)
684 g_raid_md_intel_print(meta);
686 if (strncmp(meta->version, INTEL_VERSION_1300, 6) > 0) {
687 G_RAID_DEBUG(1, "Intel unsupported version: '%.6s'",
689 free(meta, M_MD_INTEL);
693 if (strncmp(meta->version, INTEL_VERSION_1300, 6) >= 0 &&
694 (meta->attributes & ~INTEL_ATTR_SUPPORTED) != 0) {
695 G_RAID_DEBUG(1, "Intel unsupported attributes: 0x%08x",
696 meta->attributes & ~INTEL_ATTR_SUPPORTED);
697 free(meta, M_MD_INTEL);
701 /* Validate disk indexes. */
702 for (i = 0; i < meta->total_volumes; i++) {
703 mvol = intel_get_volume(meta, i);
704 for (j = 0; j < (mvol->migr_state ? 2 : 1); j++) {
705 mmap = intel_get_map(mvol, j);
706 for (k = 0; k < mmap->total_disks; k++) {
707 if ((mmap->disk_idx[k] & INTEL_DI_IDX) >
709 G_RAID_DEBUG(1, "Intel metadata disk"
710 " index %d too big (>%d)",
711 mmap->disk_idx[k] & INTEL_DI_IDX,
713 free(meta, M_MD_INTEL);
720 /* Validate migration types. */
721 for (i = 0; i < meta->total_volumes; i++) {
722 mvol = intel_get_volume(meta, i);
723 /* Deny unknown migration types. */
724 if (mvol->migr_state &&
725 mvol->migr_type != INTEL_MT_INIT &&
726 mvol->migr_type != INTEL_MT_REBUILD &&
727 mvol->migr_type != INTEL_MT_VERIFY &&
728 mvol->migr_type != INTEL_MT_GEN_MIGR &&
729 mvol->migr_type != INTEL_MT_REPAIR) {
730 G_RAID_DEBUG(1, "Intel metadata has unsupported"
731 " migration type %d", mvol->migr_type);
732 free(meta, M_MD_INTEL);
735 /* Deny general migrations except SINGLE->RAID1. */
736 if (mvol->migr_state &&
737 mvol->migr_type == INTEL_MT_GEN_MIGR) {
738 mmap = intel_get_map(mvol, 0);
739 mmap1 = intel_get_map(mvol, 1);
740 if (mmap1->total_disks != 1 ||
741 mmap->type != INTEL_T_RAID1 ||
742 mmap->total_disks != 2 ||
743 mmap->offset != mmap1->offset ||
744 mmap->disk_sectors != mmap1->disk_sectors ||
745 mmap->total_domains != mmap->total_disks ||
746 mmap->offset_hi != mmap1->offset_hi ||
747 mmap->disk_sectors_hi != mmap1->disk_sectors_hi ||
748 (mmap->disk_idx[0] != mmap1->disk_idx[0] &&
749 mmap->disk_idx[0] != mmap1->disk_idx[1])) {
750 G_RAID_DEBUG(1, "Intel metadata has unsupported"
751 " variant of general migration");
752 free(meta, M_MD_INTEL);
762 intel_meta_write(struct g_consumer *cp, struct intel_raid_conf *meta)
764 struct g_provider *pp;
766 int error, i, sectors;
767 uint32_t checksum, *ptr;
771 /* Recalculate checksum for case if metadata were changed. */
773 for (checksum = 0, ptr = (uint32_t *)meta, i = 0;
774 i < (meta->config_size / sizeof(uint32_t)); i++) {
777 meta->checksum = checksum;
779 /* Create and fill buffer. */
780 sectors = howmany(meta->config_size, pp->sectorsize);
781 buf = malloc(sectors * pp->sectorsize, M_MD_INTEL, M_WAITOK | M_ZERO);
783 memcpy(buf, ((char *)meta) + pp->sectorsize,
784 (sectors - 1) * pp->sectorsize);
786 memcpy(buf + (sectors - 1) * pp->sectorsize, meta, pp->sectorsize);
788 error = g_write_data(cp,
789 pp->mediasize - pp->sectorsize * (1 + sectors),
790 buf, pp->sectorsize * sectors);
792 G_RAID_DEBUG(1, "Cannot write metadata to %s (error=%d).",
796 free(buf, M_MD_INTEL);
801 intel_meta_erase(struct g_consumer *cp)
803 struct g_provider *pp;
808 buf = malloc(pp->sectorsize, M_MD_INTEL, M_WAITOK | M_ZERO);
809 error = g_write_data(cp,
810 pp->mediasize - 2 * pp->sectorsize,
811 buf, pp->sectorsize);
813 G_RAID_DEBUG(1, "Cannot erase metadata on %s (error=%d).",
816 free(buf, M_MD_INTEL);
821 intel_meta_write_spare(struct g_consumer *cp, struct intel_raid_disk *d)
823 struct intel_raid_conf *meta;
826 /* Fill anchor and single disk. */
827 meta = malloc(INTEL_MAX_MD_SIZE(1), M_MD_INTEL, M_WAITOK | M_ZERO);
828 memcpy(&meta->intel_id[0], INTEL_MAGIC, sizeof(INTEL_MAGIC) - 1);
829 memcpy(&meta->version[0], INTEL_VERSION_1000,
830 sizeof(INTEL_VERSION_1000) - 1);
831 meta->config_size = INTEL_MAX_MD_SIZE(1);
832 meta->config_id = meta->orig_config_id = arc4random();
833 meta->generation = 1;
834 meta->total_disks = 1;
836 error = intel_meta_write(cp, meta);
837 free(meta, M_MD_INTEL);
841 static struct g_raid_disk *
842 g_raid_md_intel_get_disk(struct g_raid_softc *sc, int id)
844 struct g_raid_disk *disk;
845 struct g_raid_md_intel_perdisk *pd;
847 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
848 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
849 if (pd->pd_disk_pos == id)
856 g_raid_md_intel_supported(int level, int qual, int disks, int force)
860 case G_RAID_VOLUME_RL_RAID0:
863 if (!force && (disks < 2 || disks > 6))
866 case G_RAID_VOLUME_RL_RAID1:
869 if (!force && (disks != 2))
872 case G_RAID_VOLUME_RL_RAID1E:
875 if (!force && (disks != 4))
878 case G_RAID_VOLUME_RL_RAID5:
881 if (!force && disks > 6)
883 if (qual != G_RAID_VOLUME_RLQ_R5LA)
889 if (level != G_RAID_VOLUME_RL_RAID5 && qual != G_RAID_VOLUME_RLQ_NONE)
894 static struct g_raid_volume *
895 g_raid_md_intel_get_volume(struct g_raid_softc *sc, int id)
897 struct g_raid_volume *mvol;
898 struct g_raid_md_intel_pervolume *pv;
900 TAILQ_FOREACH(mvol, &sc->sc_volumes, v_next) {
901 pv = mvol->v_md_data;
902 if (pv->pv_volume_pos == id)
909 g_raid_md_intel_start_disk(struct g_raid_disk *disk)
911 struct g_raid_softc *sc;
912 struct g_raid_subdisk *sd, *tmpsd;
913 struct g_raid_disk *olddisk, *tmpdisk;
914 struct g_raid_md_object *md;
915 struct g_raid_md_intel_object *mdi;
916 struct g_raid_md_intel_pervolume *pv;
917 struct g_raid_md_intel_perdisk *pd, *oldpd;
918 struct intel_raid_conf *meta;
919 struct intel_raid_vol *mvol;
920 struct intel_raid_map *mmap0, *mmap1;
921 int disk_pos, resurrection = 0, migr_global, i;
925 mdi = (struct g_raid_md_intel_object *)md;
926 meta = mdi->mdio_meta;
927 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
930 /* Find disk position in metadata by its serial. */
931 disk_pos = intel_meta_find_disk(meta, pd->pd_disk_meta.serial);
933 G_RAID_DEBUG1(1, sc, "Unknown, probably new or stale disk");
934 /* Failed stale disk is useless for us. */
935 if ((pd->pd_disk_meta.flags & INTEL_F_FAILED) &&
936 !(pd->pd_disk_meta.flags & INTEL_F_DISABLED)) {
937 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE_FAILED);
940 /* If we are in the start process, that's all for now. */
941 if (!mdi->mdio_started)
944 * If we have already started - try to get use of the disk.
945 * Try to replace OFFLINE disks first, then FAILED.
947 TAILQ_FOREACH(tmpdisk, &sc->sc_disks, d_next) {
948 if (tmpdisk->d_state != G_RAID_DISK_S_OFFLINE &&
949 tmpdisk->d_state != G_RAID_DISK_S_FAILED)
951 /* Make sure this disk is big enough. */
952 TAILQ_FOREACH(sd, &tmpdisk->d_subdisks, sd_next) {
954 intel_get_disk_sectors(&pd->pd_disk_meta);
956 if (sd->sd_offset + sd->sd_size + 4096 >
957 disk_sectors * 512) {
959 "Disk too small (%llu < %llu)",
963 sd->sd_offset + sd->sd_size + 4096);
969 if (tmpdisk->d_state == G_RAID_DISK_S_OFFLINE) {
972 } else if (olddisk == NULL)
975 if (olddisk == NULL) {
977 if (pd->pd_disk_meta.flags & INTEL_F_SPARE) {
978 g_raid_change_disk_state(disk,
979 G_RAID_DISK_S_SPARE);
982 g_raid_change_disk_state(disk,
983 G_RAID_DISK_S_STALE);
987 oldpd = (struct g_raid_md_intel_perdisk *)olddisk->d_md_data;
988 disk_pos = oldpd->pd_disk_pos;
992 if (olddisk == NULL) {
993 /* Find placeholder by position. */
994 olddisk = g_raid_md_intel_get_disk(sc, disk_pos);
996 panic("No disk at position %d!", disk_pos);
997 if (olddisk->d_state != G_RAID_DISK_S_OFFLINE) {
998 G_RAID_DEBUG1(1, sc, "More than one disk for pos %d",
1000 g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE);
1003 oldpd = (struct g_raid_md_intel_perdisk *)olddisk->d_md_data;
1006 /* Replace failed disk or placeholder with new disk. */
1007 TAILQ_FOREACH_SAFE(sd, &olddisk->d_subdisks, sd_next, tmpsd) {
1008 TAILQ_REMOVE(&olddisk->d_subdisks, sd, sd_next);
1009 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1012 oldpd->pd_disk_pos = -2;
1013 pd->pd_disk_pos = disk_pos;
1015 /* If it was placeholder -- destroy it. */
1016 if (olddisk->d_state == G_RAID_DISK_S_OFFLINE) {
1017 g_raid_destroy_disk(olddisk);
1019 /* Otherwise, make it STALE_FAILED. */
1020 g_raid_change_disk_state(olddisk, G_RAID_DISK_S_STALE_FAILED);
1021 /* Update global metadata just in case. */
1022 memcpy(&meta->disk[disk_pos], &pd->pd_disk_meta,
1023 sizeof(struct intel_raid_disk));
1026 /* Welcome the new disk. */
1027 if ((meta->disk[disk_pos].flags & INTEL_F_DISABLED) &&
1028 !(pd->pd_disk_meta.flags & INTEL_F_SPARE))
1029 g_raid_change_disk_state(disk, G_RAID_DISK_S_DISABLED);
1030 else if (resurrection)
1031 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
1032 else if (meta->disk[disk_pos].flags & INTEL_F_FAILED)
1033 g_raid_change_disk_state(disk, G_RAID_DISK_S_FAILED);
1034 else if (meta->disk[disk_pos].flags & INTEL_F_SPARE)
1035 g_raid_change_disk_state(disk, G_RAID_DISK_S_SPARE);
1037 g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE);
1038 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
1039 pv = sd->sd_volume->v_md_data;
1040 mvol = intel_get_volume(meta, pv->pv_volume_pos);
1041 mmap0 = intel_get_map(mvol, 0);
1042 if (mvol->migr_state)
1043 mmap1 = intel_get_map(mvol, 1);
1048 for (i = 0; i < mmap0->total_disks; i++) {
1049 if ((mmap0->disk_idx[i] & INTEL_DI_RBLD) == 0 &&
1050 (mmap1->disk_idx[i] & INTEL_DI_RBLD) != 0)
1054 if ((meta->disk[disk_pos].flags & INTEL_F_DISABLED) &&
1055 !(pd->pd_disk_meta.flags & INTEL_F_SPARE)) {
1056 /* Disabled disk, useless. */
1057 g_raid_change_subdisk_state(sd,
1058 G_RAID_SUBDISK_S_NONE);
1059 } else if (resurrection) {
1060 /* Stale disk, almost same as new. */
1061 g_raid_change_subdisk_state(sd,
1062 G_RAID_SUBDISK_S_NEW);
1063 } else if (meta->disk[disk_pos].flags & INTEL_F_FAILED) {
1064 /* Failed disk, almost useless. */
1065 g_raid_change_subdisk_state(sd,
1066 G_RAID_SUBDISK_S_FAILED);
1067 } else if (mvol->migr_state == 0) {
1068 if (mmap0->status == INTEL_S_UNINITIALIZED &&
1069 (!pv->pv_cng || pv->pv_cng_master_disk != disk_pos)) {
1070 /* Freshly created uninitialized volume. */
1071 g_raid_change_subdisk_state(sd,
1072 G_RAID_SUBDISK_S_UNINITIALIZED);
1073 } else if (mmap0->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) {
1074 /* Freshly inserted disk. */
1075 g_raid_change_subdisk_state(sd,
1076 G_RAID_SUBDISK_S_NEW);
1077 } else if (mvol->dirty && (!pv->pv_cng ||
1078 pv->pv_cng_master_disk != disk_pos)) {
1079 /* Dirty volume (unclean shutdown). */
1080 g_raid_change_subdisk_state(sd,
1081 G_RAID_SUBDISK_S_STALE);
1083 /* Up to date disk. */
1084 g_raid_change_subdisk_state(sd,
1085 G_RAID_SUBDISK_S_ACTIVE);
1087 } else if (mvol->migr_type == INTEL_MT_INIT ||
1088 mvol->migr_type == INTEL_MT_REBUILD) {
1089 if (mmap0->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) {
1090 /* Freshly inserted disk. */
1091 g_raid_change_subdisk_state(sd,
1092 G_RAID_SUBDISK_S_NEW);
1093 } else if (mmap1->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) {
1094 /* Rebuilding disk. */
1095 g_raid_change_subdisk_state(sd,
1096 G_RAID_SUBDISK_S_REBUILD);
1098 sd->sd_rebuild_pos = 0;
1100 sd->sd_rebuild_pos =
1101 intel_get_vol_curr_migr_unit(mvol) *
1102 sd->sd_volume->v_strip_size *
1103 mmap0->total_domains;
1105 } else if (mvol->migr_type == INTEL_MT_INIT &&
1107 /* Freshly created uninitialized volume. */
1108 g_raid_change_subdisk_state(sd,
1109 G_RAID_SUBDISK_S_UNINITIALIZED);
1110 } else if (mvol->dirty && (!pv->pv_cng ||
1111 pv->pv_cng_master_disk != disk_pos)) {
1112 /* Dirty volume (unclean shutdown). */
1113 g_raid_change_subdisk_state(sd,
1114 G_RAID_SUBDISK_S_STALE);
1116 /* Up to date disk. */
1117 g_raid_change_subdisk_state(sd,
1118 G_RAID_SUBDISK_S_ACTIVE);
1120 } else if (mvol->migr_type == INTEL_MT_VERIFY ||
1121 mvol->migr_type == INTEL_MT_REPAIR) {
1122 if (mmap0->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) {
1123 /* Freshly inserted disk. */
1124 g_raid_change_subdisk_state(sd,
1125 G_RAID_SUBDISK_S_NEW);
1126 } else if ((mmap1->disk_idx[sd->sd_pos] & INTEL_DI_RBLD) ||
1128 /* Resyncing disk. */
1129 g_raid_change_subdisk_state(sd,
1130 G_RAID_SUBDISK_S_RESYNC);
1132 sd->sd_rebuild_pos = 0;
1134 sd->sd_rebuild_pos =
1135 intel_get_vol_curr_migr_unit(mvol) *
1136 sd->sd_volume->v_strip_size *
1137 mmap0->total_domains;
1139 } else if (mvol->dirty) {
1140 /* Dirty volume (unclean shutdown). */
1141 g_raid_change_subdisk_state(sd,
1142 G_RAID_SUBDISK_S_STALE);
1144 /* Up to date disk. */
1145 g_raid_change_subdisk_state(sd,
1146 G_RAID_SUBDISK_S_ACTIVE);
1148 } else if (mvol->migr_type == INTEL_MT_GEN_MIGR) {
1149 if ((mmap1->disk_idx[0] & INTEL_DI_IDX) != disk_pos) {
1150 /* Freshly inserted disk. */
1151 g_raid_change_subdisk_state(sd,
1152 G_RAID_SUBDISK_S_NEW);
1154 /* Up to date disk. */
1155 g_raid_change_subdisk_state(sd,
1156 G_RAID_SUBDISK_S_ACTIVE);
1159 g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW,
1160 G_RAID_EVENT_SUBDISK);
1163 /* Update status of our need for spare. */
1164 if (mdi->mdio_started) {
1165 mdi->mdio_incomplete =
1166 (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) +
1167 g_raid_ndisks(sc, G_RAID_DISK_S_DISABLED) <
1171 return (resurrection);
1175 g_disk_md_intel_retaste(void *arg, int pending)
1178 G_RAID_DEBUG(1, "Array is not complete, trying to retaste.");
1179 g_retaste(&g_raid_class);
1180 free(arg, M_MD_INTEL);
1184 g_raid_md_intel_refill(struct g_raid_softc *sc)
1186 struct g_raid_md_object *md;
1187 struct g_raid_md_intel_object *mdi;
1188 struct intel_raid_conf *meta;
1189 struct g_raid_disk *disk;
1194 mdi = (struct g_raid_md_intel_object *)md;
1195 meta = mdi->mdio_meta;
1198 /* Make sure we miss anything. */
1199 na = g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) +
1200 g_raid_ndisks(sc, G_RAID_DISK_S_DISABLED);
1201 if (na == meta->total_disks)
1204 G_RAID_DEBUG1(1, md->mdo_softc,
1205 "Array is not complete (%d of %d), "
1206 "trying to refill.", na, meta->total_disks);
1208 /* Try to get use some of STALE disks. */
1209 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1210 if (disk->d_state == G_RAID_DISK_S_STALE) {
1211 update += g_raid_md_intel_start_disk(disk);
1212 if (disk->d_state == G_RAID_DISK_S_ACTIVE ||
1213 disk->d_state == G_RAID_DISK_S_DISABLED)
1220 /* Try to get use some of SPARE disks. */
1221 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1222 if (disk->d_state == G_RAID_DISK_S_SPARE) {
1223 update += g_raid_md_intel_start_disk(disk);
1224 if (disk->d_state == G_RAID_DISK_S_ACTIVE)
1228 } while (disk != NULL);
1230 /* Write new metadata if we changed something. */
1232 g_raid_md_write_intel(md, NULL, NULL, NULL);
1233 meta = mdi->mdio_meta;
1236 /* Update status of our need for spare. */
1237 mdi->mdio_incomplete = (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) +
1238 g_raid_ndisks(sc, G_RAID_DISK_S_DISABLED) < meta->total_disks);
1240 /* Request retaste hoping to find spare. */
1241 if (mdi->mdio_incomplete) {
1242 task = malloc(sizeof(struct task),
1243 M_MD_INTEL, M_WAITOK | M_ZERO);
1244 TASK_INIT(task, 0, g_disk_md_intel_retaste, task);
1245 taskqueue_enqueue(taskqueue_swi, task);
1250 g_raid_md_intel_start(struct g_raid_softc *sc)
1252 struct g_raid_md_object *md;
1253 struct g_raid_md_intel_object *mdi;
1254 struct g_raid_md_intel_pervolume *pv;
1255 struct g_raid_md_intel_perdisk *pd;
1256 struct intel_raid_conf *meta;
1257 struct intel_raid_vol *mvol;
1258 struct intel_raid_map *mmap;
1259 struct g_raid_volume *vol;
1260 struct g_raid_subdisk *sd;
1261 struct g_raid_disk *disk;
1265 mdi = (struct g_raid_md_intel_object *)md;
1266 meta = mdi->mdio_meta;
1268 /* Create volumes and subdisks. */
1269 for (i = 0; i < meta->total_volumes; i++) {
1270 mvol = intel_get_volume(meta, i);
1271 mmap = intel_get_map(mvol, 0);
1272 vol = g_raid_create_volume(sc, mvol->name, mvol->tid - 1);
1273 pv = malloc(sizeof(*pv), M_MD_INTEL, M_WAITOK | M_ZERO);
1274 pv->pv_volume_pos = i;
1275 pv->pv_cng = (mvol->state & INTEL_ST_CLONE_N_GO) != 0;
1276 pv->pv_cng_man_sync = (mvol->state & INTEL_ST_CLONE_MAN_SYNC) != 0;
1277 if (mvol->cng_master_disk < mmap->total_disks)
1278 pv->pv_cng_master_disk = mvol->cng_master_disk;
1279 vol->v_md_data = pv;
1280 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE;
1281 if (mmap->type == INTEL_T_RAID0)
1282 vol->v_raid_level = G_RAID_VOLUME_RL_RAID0;
1283 else if (mmap->type == INTEL_T_RAID1 &&
1284 mmap->total_domains >= 2 &&
1285 mmap->total_domains <= mmap->total_disks) {
1286 /* Assume total_domains is correct. */
1287 if (mmap->total_domains == mmap->total_disks)
1288 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1;
1290 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E;
1291 } else if (mmap->type == INTEL_T_RAID1) {
1292 /* total_domains looks wrong. */
1293 if (mmap->total_disks <= 2)
1294 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1;
1296 vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E;
1297 } else if (mmap->type == INTEL_T_RAID5) {
1298 vol->v_raid_level = G_RAID_VOLUME_RL_RAID5;
1299 vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_R5LA;
1301 vol->v_raid_level = G_RAID_VOLUME_RL_UNKNOWN;
1302 vol->v_strip_size = (u_int)mmap->strip_sectors * 512; //ZZZ
1303 vol->v_disks_count = mmap->total_disks;
1304 vol->v_mediasize = (off_t)mvol->total_sectors * 512; //ZZZ
1305 vol->v_sectorsize = 512; //ZZZ
1306 for (j = 0; j < vol->v_disks_count; j++) {
1307 sd = &vol->v_subdisks[j];
1308 sd->sd_offset = intel_get_map_offset(mmap) * 512; //ZZZ
1309 sd->sd_size = intel_get_map_disk_sectors(mmap) * 512; //ZZZ
1311 g_raid_start_volume(vol);
1314 /* Create disk placeholders to store data for later writing. */
1315 for (disk_pos = 0; disk_pos < meta->total_disks; disk_pos++) {
1316 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO);
1317 pd->pd_disk_pos = disk_pos;
1318 pd->pd_disk_meta = meta->disk[disk_pos];
1319 disk = g_raid_create_disk(sc);
1320 disk->d_md_data = (void *)pd;
1321 disk->d_state = G_RAID_DISK_S_OFFLINE;
1322 for (i = 0; i < meta->total_volumes; i++) {
1323 mvol = intel_get_volume(meta, i);
1324 mmap = intel_get_map(mvol, 0);
1325 for (j = 0; j < mmap->total_disks; j++) {
1326 if ((mmap->disk_idx[j] & INTEL_DI_IDX) == disk_pos)
1329 if (j == mmap->total_disks)
1331 vol = g_raid_md_intel_get_volume(sc, i);
1332 sd = &vol->v_subdisks[j];
1334 TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next);
1338 /* Make all disks found till the moment take their places. */
1340 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
1341 if (disk->d_state == G_RAID_DISK_S_NONE) {
1342 g_raid_md_intel_start_disk(disk);
1346 } while (disk != NULL);
1348 mdi->mdio_started = 1;
1349 G_RAID_DEBUG1(0, sc, "Array started.");
1350 g_raid_md_write_intel(md, NULL, NULL, NULL);
1352 /* Pickup any STALE/SPARE disks to refill array if needed. */
1353 g_raid_md_intel_refill(sc);
1355 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
1356 g_raid_event_send(vol, G_RAID_VOLUME_E_START,
1357 G_RAID_EVENT_VOLUME);
1360 callout_stop(&mdi->mdio_start_co);
1361 G_RAID_DEBUG1(1, sc, "root_mount_rel %p", mdi->mdio_rootmount);
1362 root_mount_rel(mdi->mdio_rootmount);
1363 mdi->mdio_rootmount = NULL;
1367 g_raid_md_intel_new_disk(struct g_raid_disk *disk)
1369 struct g_raid_softc *sc;
1370 struct g_raid_md_object *md;
1371 struct g_raid_md_intel_object *mdi;
1372 struct intel_raid_conf *pdmeta;
1373 struct g_raid_md_intel_perdisk *pd;
1377 mdi = (struct g_raid_md_intel_object *)md;
1378 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
1379 pdmeta = pd->pd_meta;
1381 if (mdi->mdio_started) {
1382 if (g_raid_md_intel_start_disk(disk))
1383 g_raid_md_write_intel(md, NULL, NULL, NULL);
1385 /* If we haven't started yet - check metadata freshness. */
1386 if (mdi->mdio_meta == NULL ||
1387 ((int32_t)(pdmeta->generation - mdi->mdio_generation)) > 0) {
1388 G_RAID_DEBUG1(1, sc, "Newer disk");
1389 if (mdi->mdio_meta != NULL)
1390 free(mdi->mdio_meta, M_MD_INTEL);
1391 mdi->mdio_meta = intel_meta_copy(pdmeta);
1392 mdi->mdio_generation = mdi->mdio_meta->generation;
1393 mdi->mdio_disks_present = 1;
1394 } else if (pdmeta->generation == mdi->mdio_generation) {
1395 mdi->mdio_disks_present++;
1396 G_RAID_DEBUG1(1, sc, "Matching disk (%d of %d up)",
1397 mdi->mdio_disks_present,
1398 mdi->mdio_meta->total_disks);
1400 G_RAID_DEBUG1(1, sc, "Older disk");
1402 /* If we collected all needed disks - start array. */
1403 if (mdi->mdio_disks_present == mdi->mdio_meta->total_disks)
1404 g_raid_md_intel_start(sc);
1409 g_raid_intel_go(void *arg)
1411 struct g_raid_softc *sc;
1412 struct g_raid_md_object *md;
1413 struct g_raid_md_intel_object *mdi;
1417 mdi = (struct g_raid_md_intel_object *)md;
1418 if (!mdi->mdio_started) {
1419 G_RAID_DEBUG1(0, sc, "Force array start due to timeout.");
1420 g_raid_event_send(sc, G_RAID_NODE_E_START, 0);
1425 g_raid_md_create_intel(struct g_raid_md_object *md, struct g_class *mp,
1428 struct g_raid_softc *sc;
1429 struct g_raid_md_intel_object *mdi;
1432 mdi = (struct g_raid_md_intel_object *)md;
1433 mdi->mdio_config_id = mdi->mdio_orig_config_id = arc4random();
1434 mdi->mdio_generation = 0;
1435 snprintf(name, sizeof(name), "Intel-%08x", mdi->mdio_config_id);
1436 sc = g_raid_create_node(mp, name, md);
1438 return (G_RAID_MD_TASTE_FAIL);
1441 return (G_RAID_MD_TASTE_NEW);
1445 * Return the last N characters of the serial label. The Linux and
1446 * ataraid(7) code always uses the last 16 characters of the label to
1447 * store into the Intel meta format. Generalize this to N characters
1448 * since that's easy. Labels can be up to 20 characters for SATA drives
1449 * and up 251 characters for SAS drives. Since intel controllers don't
1450 * support SAS drives, just stick with the SATA limits for stack friendliness.
1453 g_raid_md_get_label(struct g_consumer *cp, char *serial, int serlen)
1455 char serial_buffer[DISK_IDENT_SIZE];
1458 len = sizeof(serial_buffer);
1459 error = g_io_getattr("GEOM::ident", cp, &len, serial_buffer);
1462 len = strlen(serial_buffer);
1467 strncpy(serial, serial_buffer + len, serlen);
1472 g_raid_md_taste_intel(struct g_raid_md_object *md, struct g_class *mp,
1473 struct g_consumer *cp, struct g_geom **gp)
1475 struct g_consumer *rcp;
1476 struct g_provider *pp;
1477 struct g_raid_md_intel_object *mdi, *mdi1;
1478 struct g_raid_softc *sc;
1479 struct g_raid_disk *disk;
1480 struct intel_raid_conf *meta;
1481 struct g_raid_md_intel_perdisk *pd;
1482 struct g_geom *geom;
1483 int error, disk_pos, result, spare, len;
1484 char serial[INTEL_SERIAL_LEN];
1488 G_RAID_DEBUG(1, "Tasting Intel on %s", cp->provider->name);
1489 mdi = (struct g_raid_md_intel_object *)md;
1492 /* Read metadata from device. */
1495 g_topology_unlock();
1496 error = g_raid_md_get_label(cp, serial, sizeof(serial));
1498 G_RAID_DEBUG(1, "Cannot get serial number from %s (error=%d).",
1503 len = sizeof(vendor);
1504 if (pp->geom->rank == 1)
1505 g_io_getattr("GEOM::hba_vendor", cp, &len, &vendor);
1506 meta = intel_meta_read(cp);
1509 if (g_raid_aggressive_spare) {
1510 if (vendor != 0x8086) {
1512 "Intel vendor mismatch 0x%04x != 0x8086",
1516 "No Intel metadata, forcing spare.");
1521 return (G_RAID_MD_TASTE_FAIL);
1524 /* Check this disk position in obtained metadata. */
1525 disk_pos = intel_meta_find_disk(meta, serial);
1527 G_RAID_DEBUG(1, "Intel serial '%s' not found", serial);
1530 if (intel_get_disk_sectors(&meta->disk[disk_pos]) !=
1531 (pp->mediasize / pp->sectorsize)) {
1532 G_RAID_DEBUG(1, "Intel size mismatch %ju != %ju",
1533 intel_get_disk_sectors(&meta->disk[disk_pos]),
1534 (off_t)(pp->mediasize / pp->sectorsize));
1538 G_RAID_DEBUG(1, "Intel disk position %d", disk_pos);
1539 spare = meta->disk[disk_pos].flags & INTEL_F_SPARE;
1542 /* Search for matching node. */
1545 LIST_FOREACH(geom, &mp->geom, geom) {
1549 if (sc->sc_stopping != 0)
1551 if (sc->sc_md->mdo_class != md->mdo_class)
1553 mdi1 = (struct g_raid_md_intel_object *)sc->sc_md;
1555 if (mdi1->mdio_incomplete)
1558 if (mdi1->mdio_config_id == meta->config_id)
1563 /* Found matching node. */
1565 G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name);
1566 result = G_RAID_MD_TASTE_EXISTING;
1568 } else if (spare) { /* Not found needy node -- left for later. */
1569 G_RAID_DEBUG(1, "Spare is not needed at this time");
1572 } else { /* Not found matching node -- create one. */
1573 result = G_RAID_MD_TASTE_NEW;
1574 mdi->mdio_config_id = meta->config_id;
1575 mdi->mdio_orig_config_id = meta->orig_config_id;
1576 snprintf(name, sizeof(name), "Intel-%08x", meta->config_id);
1577 sc = g_raid_create_node(mp, name, md);
1580 callout_init(&mdi->mdio_start_co, 1);
1581 callout_reset(&mdi->mdio_start_co, g_raid_start_timeout * hz,
1582 g_raid_intel_go, sc);
1583 mdi->mdio_rootmount = root_mount_hold("GRAID-Intel");
1584 G_RAID_DEBUG1(1, sc, "root_mount_hold %p", mdi->mdio_rootmount);
1587 /* There is no return after this point, so we close passed consumer. */
1588 g_access(cp, -1, 0, 0);
1590 rcp = g_new_consumer(geom);
1591 rcp->flags |= G_CF_DIRECT_RECEIVE;
1593 if (g_access(rcp, 1, 1, 1) != 0)
1596 g_topology_unlock();
1597 sx_xlock(&sc->sc_lock);
1599 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO);
1601 pd->pd_disk_pos = -1;
1603 memcpy(&pd->pd_disk_meta.serial[0], serial, INTEL_SERIAL_LEN);
1604 intel_set_disk_sectors(&pd->pd_disk_meta,
1605 pp->mediasize / pp->sectorsize);
1606 pd->pd_disk_meta.id = 0;
1607 pd->pd_disk_meta.flags = INTEL_F_SPARE;
1609 pd->pd_disk_meta = meta->disk[disk_pos];
1611 disk = g_raid_create_disk(sc);
1612 disk->d_md_data = (void *)pd;
1613 disk->d_consumer = rcp;
1614 rcp->private = disk;
1616 g_raid_get_disk_info(disk);
1618 g_raid_md_intel_new_disk(disk);
1620 sx_xunlock(&sc->sc_lock);
1627 free(meta, M_MD_INTEL);
1628 return (G_RAID_MD_TASTE_FAIL);
1632 g_raid_md_event_intel(struct g_raid_md_object *md,
1633 struct g_raid_disk *disk, u_int event)
1635 struct g_raid_softc *sc;
1636 struct g_raid_subdisk *sd;
1637 struct g_raid_md_intel_object *mdi;
1638 struct g_raid_md_intel_perdisk *pd;
1641 mdi = (struct g_raid_md_intel_object *)md;
1644 case G_RAID_NODE_E_START:
1645 if (!mdi->mdio_started)
1646 g_raid_md_intel_start(sc);
1651 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
1653 case G_RAID_DISK_E_DISCONNECTED:
1654 /* If disk was assigned, just update statuses. */
1655 if (pd->pd_disk_pos >= 0) {
1656 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
1657 if (disk->d_consumer) {
1658 g_raid_kill_consumer(sc, disk->d_consumer);
1659 disk->d_consumer = NULL;
1661 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
1662 g_raid_change_subdisk_state(sd,
1663 G_RAID_SUBDISK_S_NONE);
1664 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
1665 G_RAID_EVENT_SUBDISK);
1668 /* Otherwise -- delete. */
1669 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
1670 g_raid_destroy_disk(disk);
1673 /* Write updated metadata to all disks. */
1674 g_raid_md_write_intel(md, NULL, NULL, NULL);
1676 /* Check if anything left except placeholders. */
1677 if (g_raid_ndisks(sc, -1) ==
1678 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
1679 g_raid_destroy_node(sc, 0);
1681 g_raid_md_intel_refill(sc);
1688 g_raid_md_ctl_intel(struct g_raid_md_object *md,
1689 struct gctl_req *req)
1691 struct g_raid_softc *sc;
1692 struct g_raid_volume *vol, *vol1;
1693 struct g_raid_subdisk *sd;
1694 struct g_raid_disk *disk;
1695 struct g_raid_md_intel_object *mdi;
1696 struct g_raid_md_intel_pervolume *pv;
1697 struct g_raid_md_intel_perdisk *pd;
1698 struct g_consumer *cp;
1699 struct g_provider *pp;
1700 char arg[16], serial[INTEL_SERIAL_LEN];
1701 const char *nodename, *verb, *volname, *levelname, *diskname;
1704 off_t off, size, sectorsize, strip, disk_sectors;
1705 intmax_t *sizearg, *striparg;
1706 int numdisks, i, len, level, qual, update;
1710 mdi = (struct g_raid_md_intel_object *)md;
1711 verb = gctl_get_param(req, "verb", NULL);
1712 nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs));
1714 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) {
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) {
2109 nodename = gctl_get_asciiparam(req, "arg0");
2110 if (nodename != NULL && strcasecmp(sc->sc_name, nodename) != 0)
2113 /* Full node destruction. */
2114 if (*nargs == 1 && nodename != NULL) {
2115 /* Check if some volume is still open. */
2116 force = gctl_get_paraml(req, "force", sizeof(*force));
2117 if (force != NULL && *force == 0 &&
2118 g_raid_nopens(sc) != 0) {
2119 gctl_error(req, "Some volume is still open.");
2123 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2124 if (disk->d_consumer)
2125 intel_meta_erase(disk->d_consumer);
2127 g_raid_destroy_node(sc, 0);
2131 /* Destroy specified volume. If it was last - all node. */
2133 gctl_error(req, "Invalid number of arguments.");
2136 volname = gctl_get_asciiparam(req,
2137 nodename != NULL ? "arg1" : "arg0");
2138 if (volname == NULL) {
2139 gctl_error(req, "No volume name.");
2143 /* Search for volume. */
2144 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2145 if (strcmp(vol->v_name, volname) == 0)
2147 pp = vol->v_provider;
2150 if (strcmp(pp->name, volname) == 0)
2152 if (strncmp(pp->name, "raid/", 5) == 0 &&
2153 strcmp(pp->name + 5, volname) == 0)
2157 i = strtol(volname, &tmp, 10);
2158 if (verb != volname && tmp[0] == 0) {
2159 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2160 if (vol->v_global_id == i)
2166 gctl_error(req, "Volume '%s' not found.", volname);
2170 /* Check if volume is still open. */
2171 force = gctl_get_paraml(req, "force", sizeof(*force));
2172 if (force != NULL && *force == 0 &&
2173 vol->v_provider_open != 0) {
2174 gctl_error(req, "Volume is still open.");
2178 /* Destroy volume and potentially node. */
2180 TAILQ_FOREACH(vol1, &sc->sc_volumes, v_next)
2183 g_raid_destroy_volume(vol);
2184 g_raid_md_write_intel(md, NULL, NULL, NULL);
2186 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2187 if (disk->d_consumer)
2188 intel_meta_erase(disk->d_consumer);
2190 g_raid_destroy_node(sc, 0);
2194 if (strcmp(verb, "remove") == 0 ||
2195 strcmp(verb, "fail") == 0) {
2197 gctl_error(req, "Invalid number of arguments.");
2200 for (i = 1; i < *nargs; i++) {
2201 snprintf(arg, sizeof(arg), "arg%d", i);
2202 diskname = gctl_get_asciiparam(req, arg);
2203 if (diskname == NULL) {
2204 gctl_error(req, "No disk name (%s).", arg);
2208 if (strncmp(diskname, _PATH_DEV, 5) == 0)
2211 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2212 if (disk->d_consumer != NULL &&
2213 disk->d_consumer->provider != NULL &&
2214 strcmp(disk->d_consumer->provider->name,
2219 gctl_error(req, "Disk '%s' not found.",
2225 if (strcmp(verb, "fail") == 0) {
2226 g_raid_md_fail_disk_intel(md, NULL, disk);
2230 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2232 /* Erase metadata on deleting disk. */
2233 intel_meta_erase(disk->d_consumer);
2235 /* If disk was assigned, just update statuses. */
2236 if (pd->pd_disk_pos >= 0) {
2237 g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE);
2238 g_raid_kill_consumer(sc, disk->d_consumer);
2239 disk->d_consumer = NULL;
2240 TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) {
2241 g_raid_change_subdisk_state(sd,
2242 G_RAID_SUBDISK_S_NONE);
2243 g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED,
2244 G_RAID_EVENT_SUBDISK);
2247 /* Otherwise -- delete. */
2248 g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE);
2249 g_raid_destroy_disk(disk);
2253 /* Write updated metadata to remaining disks. */
2254 g_raid_md_write_intel(md, NULL, NULL, NULL);
2256 /* Check if anything left except placeholders. */
2257 if (g_raid_ndisks(sc, -1) ==
2258 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
2259 g_raid_destroy_node(sc, 0);
2261 g_raid_md_intel_refill(sc);
2264 if (strcmp(verb, "insert") == 0) {
2266 gctl_error(req, "Invalid number of arguments.");
2270 for (i = 1; i < *nargs; i++) {
2271 /* Get disk name. */
2272 snprintf(arg, sizeof(arg), "arg%d", i);
2273 diskname = gctl_get_asciiparam(req, arg);
2274 if (diskname == NULL) {
2275 gctl_error(req, "No disk name (%s).", arg);
2280 /* Try to find provider with specified name. */
2282 cp = g_raid_open_consumer(sc, diskname);
2284 gctl_error(req, "Can't open disk '%s'.",
2286 g_topology_unlock();
2291 g_topology_unlock();
2293 /* Read disk serial. */
2294 error = g_raid_md_get_label(cp,
2295 &serial[0], INTEL_SERIAL_LEN);
2298 "Can't get serial for provider '%s'.",
2300 g_raid_kill_consumer(sc, cp);
2305 pd = malloc(sizeof(*pd), M_MD_INTEL, M_WAITOK | M_ZERO);
2306 pd->pd_disk_pos = -1;
2308 disk = g_raid_create_disk(sc);
2309 disk->d_consumer = cp;
2310 disk->d_md_data = (void *)pd;
2313 g_raid_get_disk_info(disk);
2315 memcpy(&pd->pd_disk_meta.serial[0], &serial[0],
2317 intel_set_disk_sectors(&pd->pd_disk_meta,
2318 pp->mediasize / pp->sectorsize);
2319 pd->pd_disk_meta.id = 0;
2320 pd->pd_disk_meta.flags = INTEL_F_SPARE;
2322 /* Welcome the "new" disk. */
2323 update += g_raid_md_intel_start_disk(disk);
2324 if (disk->d_state == G_RAID_DISK_S_SPARE) {
2325 intel_meta_write_spare(cp, &pd->pd_disk_meta);
2326 g_raid_destroy_disk(disk);
2327 } else if (disk->d_state != G_RAID_DISK_S_ACTIVE) {
2328 gctl_error(req, "Disk '%s' doesn't fit.",
2330 g_raid_destroy_disk(disk);
2336 /* Write new metadata if we changed something. */
2338 g_raid_md_write_intel(md, NULL, NULL, NULL);
2345 g_raid_md_write_intel(struct g_raid_md_object *md, struct g_raid_volume *tvol,
2346 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
2348 struct g_raid_softc *sc;
2349 struct g_raid_volume *vol;
2350 struct g_raid_subdisk *sd;
2351 struct g_raid_disk *disk;
2352 struct g_raid_md_intel_object *mdi;
2353 struct g_raid_md_intel_pervolume *pv;
2354 struct g_raid_md_intel_perdisk *pd;
2355 struct intel_raid_conf *meta;
2356 struct intel_raid_vol *mvol;
2357 struct intel_raid_map *mmap0, *mmap1;
2358 off_t sectorsize = 512, pos;
2359 const char *version, *cv;
2360 int vi, sdi, numdisks, len, state, stale;
2363 mdi = (struct g_raid_md_intel_object *)md;
2365 if (sc->sc_stopping == G_RAID_DESTROY_HARD)
2368 /* Bump generation. Newly written metadata may differ from previous. */
2369 mdi->mdio_generation++;
2371 /* Count number of disks. */
2373 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2374 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2375 if (pd->pd_disk_pos < 0)
2378 if (disk->d_state == G_RAID_DISK_S_ACTIVE) {
2379 pd->pd_disk_meta.flags =
2380 INTEL_F_ONLINE | INTEL_F_ASSIGNED;
2381 } else if (disk->d_state == G_RAID_DISK_S_FAILED) {
2382 pd->pd_disk_meta.flags = INTEL_F_FAILED |
2384 } else if (disk->d_state == G_RAID_DISK_S_DISABLED) {
2385 pd->pd_disk_meta.flags = INTEL_F_FAILED |
2386 INTEL_F_ASSIGNED | INTEL_F_DISABLED;
2388 if (!(pd->pd_disk_meta.flags & INTEL_F_DISABLED))
2389 pd->pd_disk_meta.flags = INTEL_F_ASSIGNED;
2390 if (pd->pd_disk_meta.id != 0xffffffff) {
2391 pd->pd_disk_meta.id = 0xffffffff;
2392 len = strlen(pd->pd_disk_meta.serial);
2393 len = min(len, INTEL_SERIAL_LEN - 3);
2394 strcpy(pd->pd_disk_meta.serial + len, ":0");
2399 /* Fill anchor and disks. */
2400 meta = malloc(INTEL_MAX_MD_SIZE(numdisks),
2401 M_MD_INTEL, M_WAITOK | M_ZERO);
2402 memcpy(&meta->intel_id[0], INTEL_MAGIC, sizeof(INTEL_MAGIC) - 1);
2403 meta->config_size = INTEL_MAX_MD_SIZE(numdisks);
2404 meta->config_id = mdi->mdio_config_id;
2405 meta->orig_config_id = mdi->mdio_orig_config_id;
2406 meta->generation = mdi->mdio_generation;
2407 meta->attributes = INTEL_ATTR_CHECKSUM;
2408 meta->total_disks = numdisks;
2409 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2410 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2411 if (pd->pd_disk_pos < 0)
2413 meta->disk[pd->pd_disk_pos] = pd->pd_disk_meta;
2414 if (pd->pd_disk_meta.sectors_hi != 0)
2415 meta->attributes |= INTEL_ATTR_2TB_DISK;
2418 /* Fill volumes and maps. */
2420 version = INTEL_VERSION_1000;
2421 TAILQ_FOREACH(vol, &sc->sc_volumes, v_next) {
2422 pv = vol->v_md_data;
2423 if (vol->v_stopping)
2425 mvol = intel_get_volume(meta, vi);
2427 /* New metadata may have different volumes order. */
2428 pv->pv_volume_pos = vi;
2430 for (sdi = 0; sdi < vol->v_disks_count; sdi++) {
2431 sd = &vol->v_subdisks[sdi];
2432 if (sd->sd_disk != NULL)
2435 if (sdi >= vol->v_disks_count)
2436 panic("No any filled subdisk in volume");
2437 if (vol->v_mediasize >= 0x20000000000llu)
2438 meta->attributes |= INTEL_ATTR_2TB;
2439 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0)
2440 meta->attributes |= INTEL_ATTR_RAID0;
2441 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
2442 meta->attributes |= INTEL_ATTR_RAID1;
2443 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID5)
2444 meta->attributes |= INTEL_ATTR_RAID5;
2445 else if ((vol->v_disks_count & 1) == 0)
2446 meta->attributes |= INTEL_ATTR_RAID10;
2448 meta->attributes |= INTEL_ATTR_RAID1E;
2450 meta->attributes |= INTEL_ATTR_RAIDCNG;
2451 if (vol->v_strip_size > 131072)
2452 meta->attributes |= INTEL_ATTR_EXT_STRIP;
2455 cv = INTEL_VERSION_1206;
2456 else if (vol->v_disks_count > 4)
2457 cv = INTEL_VERSION_1204;
2458 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID5)
2459 cv = INTEL_VERSION_1202;
2460 else if (vol->v_disks_count > 2)
2461 cv = INTEL_VERSION_1201;
2462 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
2463 cv = INTEL_VERSION_1100;
2465 cv = INTEL_VERSION_1000;
2466 if (strcmp(cv, version) > 0)
2469 strlcpy(&mvol->name[0], vol->v_name, sizeof(mvol->name));
2470 mvol->total_sectors = vol->v_mediasize / sectorsize;
2471 mvol->state = (INTEL_ST_READ_COALESCING |
2472 INTEL_ST_WRITE_COALESCING);
2473 mvol->tid = vol->v_global_id + 1;
2475 mvol->state |= INTEL_ST_CLONE_N_GO;
2476 if (pv->pv_cng_man_sync)
2477 mvol->state |= INTEL_ST_CLONE_MAN_SYNC;
2478 mvol->cng_master_disk = pv->pv_cng_master_disk;
2479 if (vol->v_subdisks[pv->pv_cng_master_disk].sd_state ==
2480 G_RAID_SUBDISK_S_NONE)
2481 mvol->cng_state = INTEL_CNGST_MASTER_MISSING;
2482 else if (vol->v_state != G_RAID_VOLUME_S_OPTIMAL)
2483 mvol->cng_state = INTEL_CNGST_NEEDS_UPDATE;
2485 mvol->cng_state = INTEL_CNGST_UPDATED;
2488 /* Check for any recovery in progress. */
2489 state = G_RAID_SUBDISK_S_ACTIVE;
2490 pos = 0x7fffffffffffffffllu;
2492 for (sdi = 0; sdi < vol->v_disks_count; sdi++) {
2493 sd = &vol->v_subdisks[sdi];
2494 if (sd->sd_state == G_RAID_SUBDISK_S_REBUILD)
2495 state = G_RAID_SUBDISK_S_REBUILD;
2496 else if (sd->sd_state == G_RAID_SUBDISK_S_RESYNC &&
2497 state != G_RAID_SUBDISK_S_REBUILD)
2498 state = G_RAID_SUBDISK_S_RESYNC;
2499 else if (sd->sd_state == G_RAID_SUBDISK_S_STALE)
2501 if ((sd->sd_state == G_RAID_SUBDISK_S_REBUILD ||
2502 sd->sd_state == G_RAID_SUBDISK_S_RESYNC) &&
2503 sd->sd_rebuild_pos < pos)
2504 pos = sd->sd_rebuild_pos;
2506 if (state == G_RAID_SUBDISK_S_REBUILD) {
2507 mvol->migr_state = 1;
2508 mvol->migr_type = INTEL_MT_REBUILD;
2509 } else if (state == G_RAID_SUBDISK_S_RESYNC) {
2510 mvol->migr_state = 1;
2511 /* mvol->migr_type = INTEL_MT_REPAIR; */
2512 mvol->migr_type = INTEL_MT_VERIFY;
2513 mvol->state |= INTEL_ST_VERIFY_AND_FIX;
2515 mvol->migr_state = 0;
2516 mvol->dirty = (vol->v_dirty || stale);
2518 mmap0 = intel_get_map(mvol, 0);
2520 /* Write map / common part of two maps. */
2521 intel_set_map_offset(mmap0, sd->sd_offset / sectorsize);
2522 intel_set_map_disk_sectors(mmap0, sd->sd_size / sectorsize);
2523 mmap0->strip_sectors = vol->v_strip_size / sectorsize;
2524 if (vol->v_state == G_RAID_VOLUME_S_BROKEN)
2525 mmap0->status = INTEL_S_FAILURE;
2526 else if (vol->v_state == G_RAID_VOLUME_S_DEGRADED)
2527 mmap0->status = INTEL_S_DEGRADED;
2528 else if (g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_UNINITIALIZED)
2529 == g_raid_nsubdisks(vol, -1))
2530 mmap0->status = INTEL_S_UNINITIALIZED;
2532 mmap0->status = INTEL_S_READY;
2533 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0)
2534 mmap0->type = INTEL_T_RAID0;
2535 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1 ||
2536 vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E)
2537 mmap0->type = INTEL_T_RAID1;
2539 mmap0->type = INTEL_T_RAID5;
2540 mmap0->total_disks = vol->v_disks_count;
2541 if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1)
2542 mmap0->total_domains = vol->v_disks_count;
2543 else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E)
2544 mmap0->total_domains = 2;
2546 mmap0->total_domains = 1;
2547 intel_set_map_stripe_count(mmap0,
2548 sd->sd_size / vol->v_strip_size / mmap0->total_domains);
2549 mmap0->failed_disk_num = 0xff;
2552 /* If there are two maps - copy common and update. */
2553 if (mvol->migr_state) {
2554 intel_set_vol_curr_migr_unit(mvol,
2555 pos / vol->v_strip_size / mmap0->total_domains);
2556 mmap1 = intel_get_map(mvol, 1);
2557 memcpy(mmap1, mmap0, sizeof(struct intel_raid_map));
2558 mmap0->status = INTEL_S_READY;
2562 /* Write disk indexes and put rebuild flags. */
2563 for (sdi = 0; sdi < vol->v_disks_count; sdi++) {
2564 sd = &vol->v_subdisks[sdi];
2565 pd = (struct g_raid_md_intel_perdisk *)
2566 sd->sd_disk->d_md_data;
2567 mmap0->disk_idx[sdi] = pd->pd_disk_pos;
2568 if (mvol->migr_state)
2569 mmap1->disk_idx[sdi] = pd->pd_disk_pos;
2570 if (sd->sd_state == G_RAID_SUBDISK_S_REBUILD ||
2571 sd->sd_state == G_RAID_SUBDISK_S_RESYNC) {
2572 mmap1->disk_idx[sdi] |= INTEL_DI_RBLD;
2573 } else if (sd->sd_state != G_RAID_SUBDISK_S_ACTIVE &&
2574 sd->sd_state != G_RAID_SUBDISK_S_STALE &&
2575 sd->sd_state != G_RAID_SUBDISK_S_UNINITIALIZED) {
2576 mmap0->disk_idx[sdi] |= INTEL_DI_RBLD;
2577 if (mvol->migr_state)
2578 mmap1->disk_idx[sdi] |= INTEL_DI_RBLD;
2580 if ((sd->sd_state == G_RAID_SUBDISK_S_NONE ||
2581 sd->sd_state == G_RAID_SUBDISK_S_FAILED ||
2582 sd->sd_state == G_RAID_SUBDISK_S_REBUILD) &&
2583 mmap0->failed_disk_num == 0xff) {
2584 mmap0->failed_disk_num = sdi;
2585 if (mvol->migr_state)
2586 mmap1->failed_disk_num = sdi;
2591 meta->total_volumes = vi;
2592 if (vi > 1 || meta->attributes &
2593 (INTEL_ATTR_EXT_STRIP | INTEL_ATTR_2TB_DISK | INTEL_ATTR_2TB))
2594 version = INTEL_VERSION_1300;
2595 if (strcmp(version, INTEL_VERSION_1300) < 0)
2596 meta->attributes &= INTEL_ATTR_CHECKSUM;
2597 memcpy(&meta->version[0], version, sizeof(INTEL_VERSION_1000) - 1);
2599 /* We are done. Print meta data and store them to disks. */
2600 g_raid_md_intel_print(meta);
2601 if (mdi->mdio_meta != NULL)
2602 free(mdi->mdio_meta, M_MD_INTEL);
2603 mdi->mdio_meta = meta;
2604 TAILQ_FOREACH(disk, &sc->sc_disks, d_next) {
2605 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2606 if (disk->d_state != G_RAID_DISK_S_ACTIVE)
2608 if (pd->pd_meta != NULL) {
2609 free(pd->pd_meta, M_MD_INTEL);
2612 pd->pd_meta = intel_meta_copy(meta);
2613 intel_meta_write(disk->d_consumer, meta);
2619 g_raid_md_fail_disk_intel(struct g_raid_md_object *md,
2620 struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk)
2622 struct g_raid_softc *sc;
2623 struct g_raid_md_intel_object *mdi;
2624 struct g_raid_md_intel_perdisk *pd;
2625 struct g_raid_subdisk *sd;
2628 mdi = (struct g_raid_md_intel_object *)md;
2629 pd = (struct g_raid_md_intel_perdisk *)tdisk->d_md_data;
2631 /* We can't fail disk that is not a part of array now. */
2632 if (pd->pd_disk_pos < 0)
2636 * Mark disk as failed in metadata and try to write that metadata
2637 * to the disk itself to prevent it's later resurrection as STALE.
2639 mdi->mdio_meta->disk[pd->pd_disk_pos].flags = INTEL_F_FAILED;
2640 pd->pd_disk_meta.flags = INTEL_F_FAILED;
2641 g_raid_md_intel_print(mdi->mdio_meta);
2642 if (tdisk->d_consumer)
2643 intel_meta_write(tdisk->d_consumer, mdi->mdio_meta);
2645 /* Change states. */
2646 g_raid_change_disk_state(tdisk, G_RAID_DISK_S_FAILED);
2647 TAILQ_FOREACH(sd, &tdisk->d_subdisks, sd_next) {
2648 g_raid_change_subdisk_state(sd,
2649 G_RAID_SUBDISK_S_FAILED);
2650 g_raid_event_send(sd, G_RAID_SUBDISK_E_FAILED,
2651 G_RAID_EVENT_SUBDISK);
2654 /* Write updated metadata to remaining disks. */
2655 g_raid_md_write_intel(md, NULL, NULL, tdisk);
2657 /* Check if anything left except placeholders. */
2658 if (g_raid_ndisks(sc, -1) ==
2659 g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE))
2660 g_raid_destroy_node(sc, 0);
2662 g_raid_md_intel_refill(sc);
2667 g_raid_md_free_disk_intel(struct g_raid_md_object *md,
2668 struct g_raid_disk *disk)
2670 struct g_raid_md_intel_perdisk *pd;
2672 pd = (struct g_raid_md_intel_perdisk *)disk->d_md_data;
2673 if (pd->pd_meta != NULL) {
2674 free(pd->pd_meta, M_MD_INTEL);
2677 free(pd, M_MD_INTEL);
2678 disk->d_md_data = NULL;
2683 g_raid_md_free_volume_intel(struct g_raid_md_object *md,
2684 struct g_raid_volume *vol)
2686 struct g_raid_md_intel_pervolume *pv;
2688 pv = (struct g_raid_md_intel_pervolume *)vol->v_md_data;
2689 free(pv, M_MD_INTEL);
2690 vol->v_md_data = NULL;
2695 g_raid_md_free_intel(struct g_raid_md_object *md)
2697 struct g_raid_md_intel_object *mdi;
2699 mdi = (struct g_raid_md_intel_object *)md;
2700 if (!mdi->mdio_started) {
2701 mdi->mdio_started = 0;
2702 callout_stop(&mdi->mdio_start_co);
2703 G_RAID_DEBUG1(1, md->mdo_softc,
2704 "root_mount_rel %p", mdi->mdio_rootmount);
2705 root_mount_rel(mdi->mdio_rootmount);
2706 mdi->mdio_rootmount = NULL;
2708 if (mdi->mdio_meta != NULL) {
2709 free(mdi->mdio_meta, M_MD_INTEL);
2710 mdi->mdio_meta = NULL;
2715 G_RAID_MD_DECLARE(intel, "Intel");