1219974Smav/*- 2219974Smav * Copyright (c) 2011 Alexander Motin <mav@FreeBSD.org> 3234601Smav * Copyright (c) 2000 - 2008 S��ren Schmidt <sos@FreeBSD.org> 4219974Smav * All rights reserved. 5219974Smav * 6219974Smav * Redistribution and use in source and binary forms, with or without 7219974Smav * modification, are permitted provided that the following conditions 8219974Smav * are met: 9219974Smav * 1. Redistributions of source code must retain the above copyright 10219974Smav * notice, this list of conditions and the following disclaimer. 11219974Smav * 2. Redistributions in binary form must reproduce the above copyright 12219974Smav * notice, this list of conditions and the following disclaimer in the 13219974Smav * documentation and/or other materials provided with the distribution. 14219974Smav * 15219974Smav * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND 16219974Smav * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 17219974Smav * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 18219974Smav * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE 19219974Smav * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 20219974Smav * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 21219974Smav * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 22219974Smav * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 23219974Smav * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 24219974Smav * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 25219974Smav * SUCH DAMAGE. 26219974Smav */ 27219974Smav 28219974Smav#include <sys/cdefs.h> 29219974Smav__FBSDID("$FreeBSD$"); 30219974Smav 31219974Smav#include <sys/param.h> 32219974Smav#include <sys/bio.h> 33219974Smav#include <sys/endian.h> 34219974Smav#include <sys/kernel.h> 35219974Smav#include <sys/kobj.h> 36219974Smav#include <sys/limits.h> 37219974Smav#include <sys/lock.h> 38219974Smav#include <sys/malloc.h> 39219974Smav#include <sys/mutex.h> 40219974Smav#include <sys/systm.h> 41219974Smav#include <sys/taskqueue.h> 42219974Smav#include <geom/geom.h> 43219974Smav#include "geom/raid/g_raid.h" 44219974Smav#include "g_raid_md_if.h" 45219974Smav 46219974Smavstatic MALLOC_DEFINE(M_MD_SII, "md_sii_data", "GEOM_RAID SiI metadata"); 47219974Smav 48219974Smavstruct sii_raid_conf { 49219974Smav uint16_t ata_params_00_53[54]; 50219974Smav uint64_t total_sectors; /* 54 - 57 */ 51219974Smav uint16_t ata_params_58_81[72]; 52219974Smav uint16_t product_id; /* 130 */ 53219974Smav uint16_t vendor_id; /* 131 */ 54219974Smav uint16_t version_minor; /* 132 */ 55219974Smav uint16_t version_major; /* 133 */ 56219974Smav uint8_t timestamp[6]; /* 134 - 136 */ 57219974Smav uint16_t strip_sectors; /* 137 */ 58219974Smav uint16_t dummy_2; 59219974Smav uint8_t disk_number; /* 139 */ 60219974Smav uint8_t type; 61219974Smav#define SII_T_RAID0 0x00 62219974Smav#define SII_T_RAID1 0x01 63219974Smav#define SII_T_RAID01 0x02 64219974Smav#define SII_T_SPARE 0x03 65219974Smav#define SII_T_CONCAT 0x04 66219974Smav#define SII_T_RAID5 0x10 67219974Smav#define SII_T_RESERVED 0xfd 68219974Smav#define SII_T_JBOD 0xff 69219974Smav 70219974Smav uint8_t raid0_disks; /* 140 */ 71219974Smav uint8_t raid0_ident; 72219974Smav uint8_t raid1_disks; /* 141 */ 73219974Smav uint8_t raid1_ident; 74219974Smav uint64_t rebuild_lba; /* 142 - 145 */ 75219974Smav uint32_t generation; /* 146 - 147 */ 76219974Smav uint8_t disk_status; /* 148 */ 77219974Smav#define SII_S_CURRENT 0x01 78219974Smav#define SII_S_REBUILD 0x02 79219974Smav#define SII_S_DROPPED 0x03 80219974Smav#define SII_S_REMOVED 0x04 81219974Smav 82219974Smav uint8_t raid_status; 83219974Smav#define SII_S_ONLINE 0x01 84219974Smav#define SII_S_AVAILABLE 0x02 85219974Smav 86219974Smav uint8_t raid_location; /* 149 */ 87219974Smav uint8_t disk_location; 88219974Smav uint8_t auto_rebuild; /* 150 */ 89219974Smav#define SII_R_REBUILD 0x00 90219974Smav#define SII_R_NOREBUILD 0xff 91219974Smav 92219974Smav uint8_t dummy_3; 93219974Smav uint8_t name[16]; /* 151 - 158 */ 94219974Smav uint16_t checksum; /* 159 */ 95219974Smav uint16_t ata_params_160_255[96]; 96219974Smav} __packed; 97219974Smav 98219974Smavstruct g_raid_md_sii_perdisk { 99219974Smav struct sii_raid_conf *pd_meta; 100219974Smav int pd_disk_pos; 101219974Smav off_t pd_disk_size; 102219974Smav}; 103219974Smav 104219974Smavstruct g_raid_md_sii_object { 105219974Smav struct g_raid_md_object mdio_base; 106219974Smav uint8_t mdio_timestamp[6]; 107219974Smav uint8_t mdio_location; 108219974Smav uint32_t mdio_generation; 109219974Smav struct sii_raid_conf *mdio_meta; 110219974Smav struct callout mdio_start_co; /* STARTING state timer. */ 111219974Smav int mdio_total_disks; 112219974Smav int mdio_disks_present; 113219974Smav int mdio_started; 114219974Smav int mdio_incomplete; 115219974Smav struct root_hold_token *mdio_rootmount; /* Root mount delay token. */ 116219974Smav}; 117219974Smav 118219974Smavstatic g_raid_md_create_t g_raid_md_create_sii; 119219974Smavstatic g_raid_md_taste_t g_raid_md_taste_sii; 120219974Smavstatic g_raid_md_event_t g_raid_md_event_sii; 121219974Smavstatic g_raid_md_ctl_t g_raid_md_ctl_sii; 122219974Smavstatic g_raid_md_write_t g_raid_md_write_sii; 123219974Smavstatic g_raid_md_fail_disk_t g_raid_md_fail_disk_sii; 124219974Smavstatic g_raid_md_free_disk_t g_raid_md_free_disk_sii; 125219974Smavstatic g_raid_md_free_t g_raid_md_free_sii; 126219974Smav 127219974Smavstatic kobj_method_t g_raid_md_sii_methods[] = { 128219974Smav KOBJMETHOD(g_raid_md_create, g_raid_md_create_sii), 129219974Smav KOBJMETHOD(g_raid_md_taste, g_raid_md_taste_sii), 130219974Smav KOBJMETHOD(g_raid_md_event, g_raid_md_event_sii), 131219974Smav KOBJMETHOD(g_raid_md_ctl, g_raid_md_ctl_sii), 132219974Smav KOBJMETHOD(g_raid_md_write, g_raid_md_write_sii), 133219974Smav KOBJMETHOD(g_raid_md_fail_disk, g_raid_md_fail_disk_sii), 134219974Smav KOBJMETHOD(g_raid_md_free_disk, g_raid_md_free_disk_sii), 135219974Smav KOBJMETHOD(g_raid_md_free, g_raid_md_free_sii), 136219974Smav { 0, 0 } 137219974Smav}; 138219974Smav 139219974Smavstatic struct g_raid_md_class g_raid_md_sii_class = { 140219974Smav "SiI", 141219974Smav g_raid_md_sii_methods, 142219974Smav sizeof(struct g_raid_md_sii_object), 143240465Smav .mdc_enable = 1, 144219974Smav .mdc_priority = 100 145219974Smav}; 146219974Smav 147219974Smavstatic void 148219974Smavg_raid_md_sii_print(struct sii_raid_conf *meta) 149219974Smav{ 150219974Smav 151219974Smav if (g_raid_debug < 1) 152219974Smav return; 153219974Smav 154219974Smav printf("********* ATA SiI RAID Metadata *********\n"); 155219974Smav printf("total_sectors %llu\n", 156219974Smav (long long unsigned)meta->total_sectors); 157219974Smav printf("product_id 0x%04x\n", meta->product_id); 158219974Smav printf("vendor_id 0x%04x\n", meta->vendor_id); 159219974Smav printf("version_minor 0x%04x\n", meta->version_minor); 160219974Smav printf("version_major 0x%04x\n", meta->version_major); 161219974Smav printf("timestamp 0x%02x%02x%02x%02x%02x%02x\n", 162219974Smav meta->timestamp[5], meta->timestamp[4], meta->timestamp[3], 163219974Smav meta->timestamp[2], meta->timestamp[1], meta->timestamp[0]); 164219974Smav printf("strip_sectors %d\n", meta->strip_sectors); 165219974Smav printf("disk_number %d\n", meta->disk_number); 166219974Smav printf("type 0x%02x\n", meta->type); 167219974Smav printf("raid0_disks %d\n", meta->raid0_disks); 168219974Smav printf("raid0_ident %d\n", meta->raid0_ident); 169219974Smav printf("raid1_disks %d\n", meta->raid1_disks); 170219974Smav printf("raid1_ident %d\n", meta->raid1_ident); 171219974Smav printf("rebuild_lba %llu\n", 172219974Smav (long long unsigned)meta->rebuild_lba); 173219974Smav printf("generation %d\n", meta->generation); 174219974Smav printf("disk_status %d\n", meta->disk_status); 175219974Smav printf("raid_status %d\n", meta->raid_status); 176219974Smav printf("raid_location %d\n", meta->raid_location); 177219974Smav printf("disk_location %d\n", meta->disk_location); 178219974Smav printf("auto_rebuild %d\n", meta->auto_rebuild); 179219974Smav printf("name <%.16s>\n", meta->name); 180219974Smav printf("checksum 0x%04x\n", meta->checksum); 181219974Smav printf("=================================================\n"); 182219974Smav} 183219974Smav 184219974Smavstatic struct sii_raid_conf * 185219974Smavsii_meta_copy(struct sii_raid_conf *meta) 186219974Smav{ 187219974Smav struct sii_raid_conf *nmeta; 188219974Smav 189219974Smav nmeta = malloc(sizeof(*meta), M_MD_SII, M_WAITOK); 190219974Smav memcpy(nmeta, meta, sizeof(*meta)); 191219974Smav return (nmeta); 192219974Smav} 193219974Smav 194219974Smavstatic int 195219974Smavsii_meta_total_disks(struct sii_raid_conf *meta) 196219974Smav{ 197219974Smav 198219974Smav switch (meta->type) { 199219974Smav case SII_T_RAID0: 200219974Smav case SII_T_RAID5: 201219974Smav case SII_T_CONCAT: 202219974Smav return (meta->raid0_disks); 203219974Smav case SII_T_RAID1: 204219974Smav return (meta->raid1_disks); 205219974Smav case SII_T_RAID01: 206219974Smav return (meta->raid0_disks * meta->raid1_disks); 207219974Smav case SII_T_SPARE: 208219974Smav case SII_T_JBOD: 209219974Smav return (1); 210219974Smav } 211219974Smav return (0); 212219974Smav} 213219974Smav 214219974Smavstatic int 215219974Smavsii_meta_disk_pos(struct sii_raid_conf *meta, struct sii_raid_conf *pdmeta) 216219974Smav{ 217219974Smav 218219974Smav if (pdmeta->type == SII_T_SPARE) 219219974Smav return (-3); 220219974Smav 221219974Smav if (memcmp(&meta->timestamp, &pdmeta->timestamp, 6) != 0) 222219974Smav return (-1); 223219974Smav 224219974Smav switch (pdmeta->type) { 225219974Smav case SII_T_RAID0: 226219974Smav case SII_T_RAID1: 227219974Smav case SII_T_RAID5: 228219974Smav case SII_T_CONCAT: 229219974Smav return (pdmeta->disk_number); 230219974Smav case SII_T_RAID01: 231219974Smav return (pdmeta->raid1_ident * pdmeta->raid1_disks + 232219974Smav pdmeta->raid0_ident); 233219974Smav case SII_T_JBOD: 234219974Smav return (0); 235219974Smav } 236219974Smav return (-1); 237219974Smav} 238219974Smav 239219974Smavstatic void 240219974Smavsii_meta_get_name(struct sii_raid_conf *meta, char *buf) 241219974Smav{ 242219974Smav int i; 243219974Smav 244219974Smav strncpy(buf, meta->name, 16); 245219974Smav buf[16] = 0; 246219974Smav for (i = 15; i >= 0; i--) { 247219974Smav if (buf[i] > 0x20) 248219974Smav break; 249219974Smav buf[i] = 0; 250219974Smav } 251219974Smav} 252219974Smav 253219974Smavstatic void 254219974Smavsii_meta_put_name(struct sii_raid_conf *meta, char *buf) 255219974Smav{ 256219974Smav 257219974Smav memset(meta->name, 0x20, 16); 258219974Smav memcpy(meta->name, buf, MIN(strlen(buf), 16)); 259219974Smav} 260219974Smav 261219974Smavstatic struct sii_raid_conf * 262219974Smavsii_meta_read(struct g_consumer *cp) 263219974Smav{ 264219974Smav struct g_provider *pp; 265219974Smav struct sii_raid_conf *meta; 266219974Smav char *buf; 267219974Smav int error, i; 268219974Smav uint16_t checksum, *ptr; 269219974Smav 270219974Smav pp = cp->provider; 271219974Smav 272219974Smav /* Read the anchor sector. */ 273219974Smav buf = g_read_data(cp, 274219974Smav pp->mediasize - pp->sectorsize, pp->sectorsize, &error); 275219974Smav if (buf == NULL) { 276219974Smav G_RAID_DEBUG(1, "Cannot read metadata from %s (error=%d).", 277219974Smav pp->name, error); 278219974Smav return (NULL); 279219974Smav } 280219974Smav meta = malloc(sizeof(*meta), M_MD_SII, M_WAITOK); 281219974Smav memcpy(meta, buf, min(sizeof(*meta), pp->sectorsize)); 282219974Smav g_free(buf); 283219974Smav 284219974Smav /* Check vendor ID. */ 285219974Smav if (meta->vendor_id != 0x1095) { 286219974Smav G_RAID_DEBUG(1, "SiI vendor ID check failed on %s (0x%04x)", 287219974Smav pp->name, meta->vendor_id); 288219974Smav free(meta, M_MD_SII); 289219974Smav return (NULL); 290219974Smav } 291219974Smav 292219974Smav /* Check metadata major version. */ 293219974Smav if (meta->version_major != 2) { 294219974Smav G_RAID_DEBUG(1, "SiI version check failed on %s (%d.%d)", 295219974Smav pp->name, meta->version_major, meta->version_minor); 296219974Smav free(meta, M_MD_SII); 297219974Smav return (NULL); 298219974Smav } 299219974Smav 300219974Smav /* Check metadata checksum. */ 301219974Smav for (checksum = 0, ptr = (uint16_t *)meta, i = 0; i <= 159; i++) 302219974Smav checksum += *ptr++; 303219974Smav if (checksum != 0) { 304219974Smav G_RAID_DEBUG(1, "SiI checksum check failed on %s", pp->name); 305219974Smav free(meta, M_MD_SII); 306219974Smav return (NULL); 307219974Smav } 308219974Smav 309219974Smav /* Check raid type. */ 310219974Smav if (meta->type != SII_T_RAID0 && meta->type != SII_T_RAID1 && 311219974Smav meta->type != SII_T_RAID01 && meta->type != SII_T_SPARE && 312219974Smav meta->type != SII_T_RAID5 && meta->type != SII_T_CONCAT && 313219974Smav meta->type != SII_T_JBOD) { 314219974Smav G_RAID_DEBUG(1, "SiI unknown RAID level on %s (0x%02x)", 315219974Smav pp->name, meta->type); 316219974Smav free(meta, M_MD_SII); 317219974Smav return (NULL); 318219974Smav } 319219974Smav 320219974Smav return (meta); 321219974Smav} 322219974Smav 323219974Smavstatic int 324219974Smavsii_meta_write(struct g_consumer *cp, struct sii_raid_conf *meta) 325219974Smav{ 326219974Smav struct g_provider *pp; 327219974Smav char *buf; 328219974Smav int error, i; 329219974Smav uint16_t checksum, *ptr; 330219974Smav 331219974Smav pp = cp->provider; 332219974Smav 333219974Smav /* Recalculate checksum for case if metadata were changed. */ 334219974Smav meta->checksum = 0; 335219974Smav for (checksum = 0, ptr = (uint16_t *)meta, i = 0; i < 159; i++) 336219974Smav checksum += *ptr++; 337219974Smav meta->checksum -= checksum; 338219974Smav 339219974Smav /* Create and fill buffer. */ 340219974Smav buf = malloc(pp->sectorsize, M_MD_SII, M_WAITOK | M_ZERO); 341219974Smav memcpy(buf, meta, sizeof(*meta)); 342219974Smav 343219974Smav /* Write 4 copies of metadata. */ 344219974Smav for (i = 0; i < 4; i++) { 345219974Smav error = g_write_data(cp, 346219974Smav pp->mediasize - (pp->sectorsize * (1 + 0x200 * i)), 347219974Smav buf, pp->sectorsize); 348219974Smav if (error != 0) { 349219974Smav G_RAID_DEBUG(1, "Cannot write metadata to %s (error=%d).", 350219974Smav pp->name, error); 351219974Smav break; 352219974Smav } 353219974Smav } 354219974Smav 355219974Smav free(buf, M_MD_SII); 356219974Smav return (error); 357219974Smav} 358219974Smav 359219974Smavstatic int 360219974Smavsii_meta_erase(struct g_consumer *cp) 361219974Smav{ 362219974Smav struct g_provider *pp; 363219974Smav char *buf; 364219974Smav int error, i; 365219974Smav 366219974Smav pp = cp->provider; 367219974Smav buf = malloc(pp->sectorsize, M_MD_SII, M_WAITOK | M_ZERO); 368219974Smav /* Write 4 copies of metadata. */ 369219974Smav for (i = 0; i < 4; i++) { 370219974Smav error = g_write_data(cp, 371219974Smav pp->mediasize - (pp->sectorsize * (1 + 0x200 * i)), 372219974Smav buf, pp->sectorsize); 373219974Smav if (error != 0) { 374219974Smav G_RAID_DEBUG(1, "Cannot erase metadata on %s (error=%d).", 375219974Smav pp->name, error); 376219974Smav } 377219974Smav } 378219974Smav free(buf, M_MD_SII); 379219974Smav return (error); 380219974Smav} 381219974Smav 382219974Smavstatic int 383219974Smavsii_meta_write_spare(struct g_consumer *cp) 384219974Smav{ 385219974Smav struct sii_raid_conf *meta; 386219974Smav int error; 387219974Smav 388219974Smav meta = malloc(sizeof(*meta), M_MD_SII, M_WAITOK | M_ZERO); 389219974Smav meta->total_sectors = cp->provider->mediasize / 390219974Smav cp->provider->sectorsize - 0x800; 391219974Smav meta->vendor_id = 0x1095; 392219974Smav meta->version_minor = 0; 393219974Smav meta->version_major = 2; 394219974Smav meta->timestamp[0] = arc4random(); 395219974Smav meta->timestamp[1] = arc4random(); 396219974Smav meta->timestamp[2] = arc4random(); 397219974Smav meta->timestamp[3] = arc4random(); 398219974Smav meta->timestamp[4] = arc4random(); 399219974Smav meta->timestamp[5] = arc4random(); 400219974Smav meta->type = SII_T_SPARE; 401219974Smav meta->generation = 1; 402219974Smav meta->raid1_ident = 0xff; 403219974Smav meta->raid_location = arc4random(); 404219974Smav error = sii_meta_write(cp, meta); 405219974Smav free(meta, M_MD_SII); 406219974Smav return (error); 407219974Smav} 408219974Smav 409219974Smavstatic struct g_raid_disk * 410219974Smavg_raid_md_sii_get_disk(struct g_raid_softc *sc, int id) 411219974Smav{ 412219974Smav struct g_raid_disk *disk; 413219974Smav struct g_raid_md_sii_perdisk *pd; 414219974Smav 415219974Smav TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 416219974Smav pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data; 417219974Smav if (pd->pd_disk_pos == id) 418219974Smav break; 419219974Smav } 420219974Smav return (disk); 421219974Smav} 422219974Smav 423219974Smavstatic int 424219974Smavg_raid_md_sii_supported(int level, int qual, int disks, int force) 425219974Smav{ 426219974Smav 427219974Smav if (disks > 8) 428219974Smav return (0); 429219974Smav switch (level) { 430219974Smav case G_RAID_VOLUME_RL_RAID0: 431219974Smav if (disks < 1) 432219974Smav return (0); 433219974Smav if (!force && (disks < 2 || disks > 6)) 434219974Smav return (0); 435219974Smav break; 436219974Smav case G_RAID_VOLUME_RL_RAID1: 437219974Smav if (disks < 1) 438219974Smav return (0); 439219974Smav if (!force && (disks != 2)) 440219974Smav return (0); 441219974Smav break; 442219974Smav case G_RAID_VOLUME_RL_RAID1E: 443219974Smav if (disks < 2) 444219974Smav return (0); 445219974Smav if (disks % 2 != 0) 446219974Smav return (0); 447219974Smav if (!force && (disks < 4)) 448219974Smav return (0); 449219974Smav break; 450219974Smav case G_RAID_VOLUME_RL_SINGLE: 451219974Smav if (disks != 1) 452219974Smav return (0); 453219974Smav break; 454219974Smav case G_RAID_VOLUME_RL_CONCAT: 455219974Smav if (disks < 2) 456219974Smav return (0); 457219974Smav break; 458219974Smav case G_RAID_VOLUME_RL_RAID5: 459219974Smav if (disks < 3) 460219974Smav return (0); 461234458Smav if (qual != G_RAID_VOLUME_RLQ_R5LS) 462234458Smav return (0); 463219974Smav break; 464219974Smav default: 465219974Smav return (0); 466219974Smav } 467234458Smav if (level != G_RAID_VOLUME_RL_RAID5 && qual != G_RAID_VOLUME_RLQ_NONE) 468219974Smav return (0); 469219974Smav return (1); 470219974Smav} 471219974Smav 472219974Smavstatic int 473219974Smavg_raid_md_sii_start_disk(struct g_raid_disk *disk) 474219974Smav{ 475219974Smav struct g_raid_softc *sc; 476219974Smav struct g_raid_subdisk *sd, *tmpsd; 477219974Smav struct g_raid_disk *olddisk, *tmpdisk; 478219974Smav struct g_raid_md_object *md; 479219974Smav struct g_raid_md_sii_object *mdi; 480219974Smav struct g_raid_md_sii_perdisk *pd, *oldpd; 481219974Smav struct sii_raid_conf *meta; 482219974Smav int disk_pos, resurrection = 0; 483219974Smav 484219974Smav sc = disk->d_softc; 485219974Smav md = sc->sc_md; 486219974Smav mdi = (struct g_raid_md_sii_object *)md; 487219974Smav meta = mdi->mdio_meta; 488219974Smav pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data; 489219974Smav olddisk = NULL; 490219974Smav 491219974Smav /* Find disk position in metadata by it's serial. */ 492219974Smav if (pd->pd_meta != NULL) 493219974Smav disk_pos = sii_meta_disk_pos(meta, pd->pd_meta); 494219974Smav else 495219974Smav disk_pos = -3; 496219974Smav if (disk_pos < 0) { 497219974Smav G_RAID_DEBUG1(1, sc, "Unknown, probably new or stale disk"); 498219974Smav /* If we are in the start process, that's all for now. */ 499219974Smav if (!mdi->mdio_started) 500219974Smav goto nofit; 501219974Smav /* 502219974Smav * If we have already started - try to get use of the disk. 503219974Smav * Try to replace OFFLINE disks first, then FAILED. 504219974Smav */ 505219974Smav TAILQ_FOREACH(tmpdisk, &sc->sc_disks, d_next) { 506219974Smav if (tmpdisk->d_state != G_RAID_DISK_S_OFFLINE && 507219974Smav tmpdisk->d_state != G_RAID_DISK_S_FAILED) 508219974Smav continue; 509219974Smav /* Make sure this disk is big enough. */ 510219974Smav TAILQ_FOREACH(sd, &tmpdisk->d_subdisks, sd_next) { 511219974Smav if (sd->sd_offset + sd->sd_size + 512 > 512219974Smav pd->pd_disk_size) { 513219974Smav G_RAID_DEBUG1(1, sc, 514219974Smav "Disk too small (%ju < %ju)", 515219974Smav pd->pd_disk_size, 516219974Smav sd->sd_offset + sd->sd_size + 512); 517219974Smav break; 518219974Smav } 519219974Smav } 520219974Smav if (sd != NULL) 521219974Smav continue; 522219974Smav if (tmpdisk->d_state == G_RAID_DISK_S_OFFLINE) { 523219974Smav olddisk = tmpdisk; 524219974Smav break; 525219974Smav } else if (olddisk == NULL) 526219974Smav olddisk = tmpdisk; 527219974Smav } 528219974Smav if (olddisk == NULL) { 529219974Smavnofit: 530219974Smav if (disk_pos == -3 || pd->pd_disk_pos == -3) { 531219974Smav g_raid_change_disk_state(disk, 532219974Smav G_RAID_DISK_S_SPARE); 533219974Smav return (1); 534219974Smav } else { 535219974Smav g_raid_change_disk_state(disk, 536219974Smav G_RAID_DISK_S_STALE); 537219974Smav return (0); 538219974Smav } 539219974Smav } 540219974Smav oldpd = (struct g_raid_md_sii_perdisk *)olddisk->d_md_data; 541219974Smav disk_pos = oldpd->pd_disk_pos; 542219974Smav resurrection = 1; 543219974Smav } 544219974Smav 545219974Smav if (olddisk == NULL) { 546219974Smav /* Find placeholder by position. */ 547219974Smav olddisk = g_raid_md_sii_get_disk(sc, disk_pos); 548219974Smav if (olddisk == NULL) 549219974Smav panic("No disk at position %d!", disk_pos); 550219974Smav if (olddisk->d_state != G_RAID_DISK_S_OFFLINE) { 551219974Smav G_RAID_DEBUG1(1, sc, "More then one disk for pos %d", 552219974Smav disk_pos); 553219974Smav g_raid_change_disk_state(disk, G_RAID_DISK_S_STALE); 554219974Smav return (0); 555219974Smav } 556219974Smav oldpd = (struct g_raid_md_sii_perdisk *)olddisk->d_md_data; 557219974Smav } 558219974Smav 559219974Smav /* Replace failed disk or placeholder with new disk. */ 560219974Smav TAILQ_FOREACH_SAFE(sd, &olddisk->d_subdisks, sd_next, tmpsd) { 561219974Smav TAILQ_REMOVE(&olddisk->d_subdisks, sd, sd_next); 562219974Smav TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next); 563219974Smav sd->sd_disk = disk; 564219974Smav } 565219974Smav oldpd->pd_disk_pos = -2; 566219974Smav pd->pd_disk_pos = disk_pos; 567219974Smav 568219974Smav /* If it was placeholder -- destroy it. */ 569219974Smav if (olddisk->d_state == G_RAID_DISK_S_OFFLINE) { 570219974Smav g_raid_destroy_disk(olddisk); 571219974Smav } else { 572219974Smav /* Otherwise, make it STALE_FAILED. */ 573219974Smav g_raid_change_disk_state(olddisk, G_RAID_DISK_S_STALE_FAILED); 574219974Smav } 575219974Smav 576219974Smav /* Welcome the new disk. */ 577219974Smav if (resurrection) 578219974Smav g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE); 579219974Smav else if (pd->pd_meta->disk_status == SII_S_CURRENT || 580219974Smav pd->pd_meta->disk_status == SII_S_REBUILD) 581219974Smav g_raid_change_disk_state(disk, G_RAID_DISK_S_ACTIVE); 582219974Smav else 583219974Smav g_raid_change_disk_state(disk, G_RAID_DISK_S_FAILED); 584219974Smav TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) { 585219974Smav 586219974Smav /* 587219974Smav * Different disks may have different sizes, 588219974Smav * in concat mode. Update from real disk size. 589219974Smav */ 590219974Smav if (meta->type == SII_T_CONCAT || meta->type == SII_T_JBOD) 591219974Smav sd->sd_size = pd->pd_disk_size - 0x800 * 512; 592219974Smav 593219974Smav if (resurrection) { 594219974Smav /* New or ex-spare disk. */ 595219974Smav g_raid_change_subdisk_state(sd, 596219974Smav G_RAID_SUBDISK_S_NEW); 597219974Smav } else if (pd->pd_meta->disk_status == SII_S_REBUILD) { 598219974Smav /* Rebuilding disk. */ 599219974Smav g_raid_change_subdisk_state(sd, 600219974Smav G_RAID_SUBDISK_S_REBUILD); 601219974Smav if (pd->pd_meta->generation == meta->generation) 602219974Smav sd->sd_rebuild_pos = pd->pd_meta->rebuild_lba * 512; 603219974Smav else 604219974Smav sd->sd_rebuild_pos = 0; 605219974Smav } else if (pd->pd_meta->disk_status == SII_S_CURRENT) { 606219974Smav if (pd->pd_meta->raid_status == SII_S_ONLINE || 607219974Smav pd->pd_meta->generation != meta->generation) { 608219974Smav /* Dirty or resyncing disk. */ 609219974Smav g_raid_change_subdisk_state(sd, 610219974Smav G_RAID_SUBDISK_S_STALE); 611219974Smav } else { 612219974Smav /* Up to date disk. */ 613219974Smav g_raid_change_subdisk_state(sd, 614219974Smav G_RAID_SUBDISK_S_ACTIVE); 615219974Smav } 616219974Smav } else { 617219974Smav g_raid_change_subdisk_state(sd, 618219974Smav G_RAID_SUBDISK_S_FAILED); 619219974Smav } 620219974Smav g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW, 621219974Smav G_RAID_EVENT_SUBDISK); 622219974Smav } 623219974Smav 624219974Smav /* Update status of our need for spare. */ 625219974Smav if (mdi->mdio_started) { 626219974Smav mdi->mdio_incomplete = 627219974Smav (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) < 628219974Smav mdi->mdio_total_disks); 629219974Smav } 630219974Smav 631219974Smav return (resurrection); 632219974Smav} 633219974Smav 634219974Smavstatic void 635219974Smavg_disk_md_sii_retaste(void *arg, int pending) 636219974Smav{ 637219974Smav 638219974Smav G_RAID_DEBUG(1, "Array is not complete, trying to retaste."); 639219974Smav g_retaste(&g_raid_class); 640219974Smav free(arg, M_MD_SII); 641219974Smav} 642219974Smav 643219974Smavstatic void 644219974Smavg_raid_md_sii_refill(struct g_raid_softc *sc) 645219974Smav{ 646219974Smav struct g_raid_md_object *md; 647219974Smav struct g_raid_md_sii_object *mdi; 648219974Smav struct g_raid_disk *disk; 649219974Smav struct task *task; 650219974Smav int update, na; 651219974Smav 652219974Smav md = sc->sc_md; 653219974Smav mdi = (struct g_raid_md_sii_object *)md; 654219974Smav update = 0; 655219974Smav do { 656219974Smav /* Make sure we miss anything. */ 657219974Smav na = g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE); 658219974Smav if (na == mdi->mdio_total_disks) 659219974Smav break; 660219974Smav 661219974Smav G_RAID_DEBUG1(1, md->mdo_softc, 662219974Smav "Array is not complete (%d of %d), " 663219974Smav "trying to refill.", na, mdi->mdio_total_disks); 664219974Smav 665219974Smav /* Try to get use some of STALE disks. */ 666219974Smav TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 667219974Smav if (disk->d_state == G_RAID_DISK_S_STALE) { 668219974Smav update += g_raid_md_sii_start_disk(disk); 669219974Smav if (disk->d_state == G_RAID_DISK_S_ACTIVE) 670219974Smav break; 671219974Smav } 672219974Smav } 673219974Smav if (disk != NULL) 674219974Smav continue; 675219974Smav 676219974Smav /* Try to get use some of SPARE disks. */ 677219974Smav TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 678219974Smav if (disk->d_state == G_RAID_DISK_S_SPARE) { 679219974Smav update += g_raid_md_sii_start_disk(disk); 680219974Smav if (disk->d_state == G_RAID_DISK_S_ACTIVE) 681219974Smav break; 682219974Smav } 683219974Smav } 684219974Smav } while (disk != NULL); 685219974Smav 686219974Smav /* Write new metadata if we changed something. */ 687220210Smav if (update) 688219974Smav g_raid_md_write_sii(md, NULL, NULL, NULL); 689219974Smav 690219974Smav /* Update status of our need for spare. */ 691219974Smav mdi->mdio_incomplete = (g_raid_ndisks(sc, G_RAID_DISK_S_ACTIVE) < 692219974Smav mdi->mdio_total_disks); 693219974Smav 694219974Smav /* Request retaste hoping to find spare. */ 695219974Smav if (mdi->mdio_incomplete) { 696219974Smav task = malloc(sizeof(struct task), 697219974Smav M_MD_SII, M_WAITOK | M_ZERO); 698219974Smav TASK_INIT(task, 0, g_disk_md_sii_retaste, task); 699219974Smav taskqueue_enqueue(taskqueue_swi, task); 700219974Smav } 701219974Smav} 702219974Smav 703219974Smavstatic void 704219974Smavg_raid_md_sii_start(struct g_raid_softc *sc) 705219974Smav{ 706219974Smav struct g_raid_md_object *md; 707219974Smav struct g_raid_md_sii_object *mdi; 708219974Smav struct g_raid_md_sii_perdisk *pd; 709219974Smav struct sii_raid_conf *meta; 710219974Smav struct g_raid_volume *vol; 711219974Smav struct g_raid_subdisk *sd; 712219974Smav struct g_raid_disk *disk, *best; 713219974Smav off_t size; 714219974Smav int j, disk_pos; 715219974Smav uint32_t gendiff, bestgendiff; 716219974Smav char buf[17]; 717219974Smav 718219974Smav md = sc->sc_md; 719219974Smav mdi = (struct g_raid_md_sii_object *)md; 720219974Smav meta = mdi->mdio_meta; 721219974Smav 722219974Smav /* Create volumes and subdisks. */ 723219974Smav sii_meta_get_name(meta, buf); 724219974Smav vol = g_raid_create_volume(sc, buf, -1); 725219974Smav vol->v_mediasize = (off_t)meta->total_sectors * 512; 726234458Smav vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_NONE; 727219974Smav if (meta->type == SII_T_RAID0) { 728219974Smav vol->v_raid_level = G_RAID_VOLUME_RL_RAID0; 729219974Smav size = vol->v_mediasize / mdi->mdio_total_disks; 730219974Smav } else if (meta->type == SII_T_RAID1) { 731219974Smav vol->v_raid_level = G_RAID_VOLUME_RL_RAID1; 732219974Smav size = vol->v_mediasize; 733219974Smav } else if (meta->type == SII_T_RAID01) { 734219974Smav vol->v_raid_level = G_RAID_VOLUME_RL_RAID1E; 735219974Smav size = vol->v_mediasize / (mdi->mdio_total_disks / 2); 736219974Smav } else if (meta->type == SII_T_CONCAT) { 737219974Smav if (mdi->mdio_total_disks == 1) 738219974Smav vol->v_raid_level = G_RAID_VOLUME_RL_SINGLE; 739219974Smav else 740219974Smav vol->v_raid_level = G_RAID_VOLUME_RL_CONCAT; 741219974Smav size = 0; 742219974Smav } else if (meta->type == SII_T_RAID5) { 743219974Smav vol->v_raid_level = G_RAID_VOLUME_RL_RAID5; 744234458Smav vol->v_raid_level_qualifier = G_RAID_VOLUME_RLQ_R5LS; 745219974Smav size = vol->v_mediasize / (mdi->mdio_total_disks - 1); 746219974Smav } else if (meta->type == SII_T_JBOD) { 747219974Smav vol->v_raid_level = G_RAID_VOLUME_RL_SINGLE; 748219974Smav size = 0; 749219974Smav } else { 750219974Smav vol->v_raid_level = G_RAID_VOLUME_RL_UNKNOWN; 751219974Smav size = 0; 752219974Smav } 753219974Smav vol->v_strip_size = meta->strip_sectors * 512; //ZZZ 754219974Smav vol->v_disks_count = mdi->mdio_total_disks; 755219974Smav vol->v_sectorsize = 512; //ZZZ 756219974Smav for (j = 0; j < vol->v_disks_count; j++) { 757219974Smav sd = &vol->v_subdisks[j]; 758219974Smav sd->sd_offset = 0; 759219974Smav sd->sd_size = size; 760219974Smav } 761219974Smav g_raid_start_volume(vol); 762219974Smav 763219974Smav /* Create disk placeholders to store data for later writing. */ 764219974Smav for (disk_pos = 0; disk_pos < mdi->mdio_total_disks; disk_pos++) { 765219974Smav pd = malloc(sizeof(*pd), M_MD_SII, M_WAITOK | M_ZERO); 766219974Smav pd->pd_disk_pos = disk_pos; 767219974Smav disk = g_raid_create_disk(sc); 768219974Smav disk->d_md_data = (void *)pd; 769219974Smav disk->d_state = G_RAID_DISK_S_OFFLINE; 770219974Smav sd = &vol->v_subdisks[disk_pos]; 771219974Smav sd->sd_disk = disk; 772219974Smav TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next); 773219974Smav } 774219974Smav 775219974Smav /* 776219974Smav * Make all disks found till the moment take their places 777219974Smav * in order of their generation numbers. 778219974Smav */ 779219974Smav do { 780219974Smav best = NULL; 781219974Smav bestgendiff = 0xffffffff; 782219974Smav TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 783219974Smav if (disk->d_state != G_RAID_DISK_S_NONE) 784219974Smav continue; 785219974Smav pd = disk->d_md_data; 786219974Smav if (pd->pd_meta == NULL) 787219974Smav gendiff = 0xfffffffe; 788219974Smav else 789219974Smav gendiff = meta->generation - 790219974Smav pd->pd_meta->generation; 791219974Smav if (gendiff < bestgendiff) { 792219974Smav best = disk; 793219974Smav bestgendiff = gendiff; 794219974Smav } 795219974Smav } 796219974Smav if (best != NULL) 797219974Smav g_raid_md_sii_start_disk(best); 798219974Smav } while (best != NULL); 799219974Smav 800219974Smav mdi->mdio_started = 1; 801219974Smav G_RAID_DEBUG1(0, sc, "Array started."); 802219974Smav g_raid_md_write_sii(md, NULL, NULL, NULL); 803219974Smav 804219974Smav /* Pickup any STALE/SPARE disks to refill array if needed. */ 805219974Smav g_raid_md_sii_refill(sc); 806219974Smav 807219974Smav g_raid_event_send(vol, G_RAID_VOLUME_E_START, G_RAID_EVENT_VOLUME); 808219974Smav 809219974Smav callout_stop(&mdi->mdio_start_co); 810219974Smav G_RAID_DEBUG1(1, sc, "root_mount_rel %p", mdi->mdio_rootmount); 811219974Smav root_mount_rel(mdi->mdio_rootmount); 812219974Smav mdi->mdio_rootmount = NULL; 813219974Smav} 814219974Smav 815219974Smavstatic void 816219974Smavg_raid_md_sii_new_disk(struct g_raid_disk *disk) 817219974Smav{ 818219974Smav struct g_raid_softc *sc; 819219974Smav struct g_raid_md_object *md; 820219974Smav struct g_raid_md_sii_object *mdi; 821219974Smav struct sii_raid_conf *pdmeta; 822219974Smav struct g_raid_md_sii_perdisk *pd; 823219974Smav 824219974Smav sc = disk->d_softc; 825219974Smav md = sc->sc_md; 826219974Smav mdi = (struct g_raid_md_sii_object *)md; 827219974Smav pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data; 828219974Smav pdmeta = pd->pd_meta; 829219974Smav 830219974Smav if (mdi->mdio_started) { 831219974Smav if (g_raid_md_sii_start_disk(disk)) 832219974Smav g_raid_md_write_sii(md, NULL, NULL, NULL); 833219974Smav } else { 834219974Smav if (mdi->mdio_meta == NULL || 835219974Smav ((int32_t)(pdmeta->generation - mdi->mdio_generation)) > 0) { 836219974Smav G_RAID_DEBUG1(1, sc, "Newer disk"); 837219974Smav if (mdi->mdio_meta != NULL) 838219974Smav free(mdi->mdio_meta, M_MD_SII); 839219974Smav mdi->mdio_meta = sii_meta_copy(pdmeta); 840219974Smav mdi->mdio_generation = mdi->mdio_meta->generation; 841219974Smav mdi->mdio_total_disks = sii_meta_total_disks(pdmeta); 842219974Smav mdi->mdio_disks_present = 1; 843219974Smav } else if (pdmeta->generation == mdi->mdio_generation) { 844219974Smav mdi->mdio_disks_present++; 845219974Smav G_RAID_DEBUG1(1, sc, "Matching disk (%d of %d up)", 846219974Smav mdi->mdio_disks_present, 847219974Smav mdi->mdio_total_disks); 848219974Smav } else { 849219974Smav G_RAID_DEBUG1(1, sc, "Older disk"); 850219974Smav } 851219974Smav 852219974Smav /* If we collected all needed disks - start array. */ 853219974Smav if (mdi->mdio_disks_present == mdi->mdio_total_disks) 854219974Smav g_raid_md_sii_start(sc); 855219974Smav } 856219974Smav} 857219974Smav 858219974Smavstatic void 859219974Smavg_raid_sii_go(void *arg) 860219974Smav{ 861219974Smav struct g_raid_softc *sc; 862219974Smav struct g_raid_md_object *md; 863219974Smav struct g_raid_md_sii_object *mdi; 864219974Smav 865219974Smav sc = arg; 866219974Smav md = sc->sc_md; 867219974Smav mdi = (struct g_raid_md_sii_object *)md; 868219974Smav if (!mdi->mdio_started) { 869219974Smav G_RAID_DEBUG1(0, sc, "Force array start due to timeout."); 870219974Smav g_raid_event_send(sc, G_RAID_NODE_E_START, 0); 871219974Smav } 872219974Smav} 873219974Smav 874219974Smavstatic int 875219974Smavg_raid_md_create_sii(struct g_raid_md_object *md, struct g_class *mp, 876219974Smav struct g_geom **gp) 877219974Smav{ 878219974Smav struct g_raid_softc *sc; 879219974Smav struct g_raid_md_sii_object *mdi; 880219974Smav char name[32]; 881219974Smav 882219974Smav mdi = (struct g_raid_md_sii_object *)md; 883219974Smav mdi->mdio_timestamp[5] = arc4random(); 884219974Smav mdi->mdio_timestamp[4] = arc4random(); 885219974Smav mdi->mdio_timestamp[3] = arc4random(); 886219974Smav mdi->mdio_timestamp[2] = arc4random(); 887219974Smav mdi->mdio_timestamp[1] = arc4random(); 888219974Smav mdi->mdio_timestamp[0] = arc4random(); 889219974Smav mdi->mdio_location = arc4random(); 890219974Smav mdi->mdio_generation = 0; 891219974Smav snprintf(name, sizeof(name), "SiI-%02x%02x%02x%02x%02x%02x", 892219974Smav mdi->mdio_timestamp[5], mdi->mdio_timestamp[4], 893219974Smav mdi->mdio_timestamp[3], mdi->mdio_timestamp[2], 894219974Smav mdi->mdio_timestamp[1], mdi->mdio_timestamp[0]); 895219974Smav sc = g_raid_create_node(mp, name, md); 896219974Smav if (sc == NULL) 897219974Smav return (G_RAID_MD_TASTE_FAIL); 898219974Smav md->mdo_softc = sc; 899219974Smav *gp = sc->sc_geom; 900219974Smav return (G_RAID_MD_TASTE_NEW); 901219974Smav} 902219974Smav 903219974Smavstatic int 904219974Smavg_raid_md_taste_sii(struct g_raid_md_object *md, struct g_class *mp, 905219974Smav struct g_consumer *cp, struct g_geom **gp) 906219974Smav{ 907219974Smav struct g_consumer *rcp; 908219974Smav struct g_provider *pp; 909219974Smav struct g_raid_md_sii_object *mdi, *mdi1; 910219974Smav struct g_raid_softc *sc; 911219974Smav struct g_raid_disk *disk; 912219974Smav struct sii_raid_conf *meta; 913219974Smav struct g_raid_md_sii_perdisk *pd; 914219974Smav struct g_geom *geom; 915242323Smav int disk_pos, result, spare, len; 916219974Smav char name[32]; 917219974Smav uint16_t vendor; 918219974Smav 919219974Smav G_RAID_DEBUG(1, "Tasting SiI on %s", cp->provider->name); 920219974Smav mdi = (struct g_raid_md_sii_object *)md; 921219974Smav pp = cp->provider; 922219974Smav 923219974Smav /* Read metadata from device. */ 924219974Smav meta = NULL; 925219974Smav vendor = 0xffff; 926219974Smav g_topology_unlock(); 927219974Smav len = 2; 928219974Smav if (pp->geom->rank == 1) 929219974Smav g_io_getattr("GEOM::hba_vendor", cp, &len, &vendor); 930219974Smav meta = sii_meta_read(cp); 931219974Smav g_topology_lock(); 932219974Smav if (meta == NULL) { 933219974Smav if (g_raid_aggressive_spare) { 934219974Smav if (vendor == 0x1095) { 935219974Smav G_RAID_DEBUG(1, 936219974Smav "No SiI metadata, forcing spare."); 937219974Smav spare = 2; 938219974Smav goto search; 939219974Smav } else { 940219974Smav G_RAID_DEBUG(1, 941219974Smav "SiI vendor mismatch 0x%04x != 0x1095", 942219974Smav vendor); 943219974Smav } 944219974Smav } 945219974Smav return (G_RAID_MD_TASTE_FAIL); 946219974Smav } 947219974Smav 948219974Smav /* Check this disk position in obtained metadata. */ 949219974Smav disk_pos = sii_meta_disk_pos(meta, meta); 950219974Smav if (disk_pos == -1) { 951219974Smav G_RAID_DEBUG(1, "SiI disk position not found"); 952219974Smav goto fail1; 953219974Smav } 954219974Smav 955219974Smav /* Metadata valid. Print it. */ 956219974Smav g_raid_md_sii_print(meta); 957219974Smav G_RAID_DEBUG(1, "SiI disk position %d", disk_pos); 958219974Smav spare = (meta->type == SII_T_SPARE) ? 1 : 0; 959219974Smav 960219974Smavsearch: 961219974Smav /* Search for matching node. */ 962219974Smav sc = NULL; 963219974Smav mdi1 = NULL; 964219974Smav LIST_FOREACH(geom, &mp->geom, geom) { 965219974Smav sc = geom->softc; 966219974Smav if (sc == NULL) 967219974Smav continue; 968219974Smav if (sc->sc_stopping != 0) 969219974Smav continue; 970219974Smav if (sc->sc_md->mdo_class != md->mdo_class) 971219974Smav continue; 972219974Smav mdi1 = (struct g_raid_md_sii_object *)sc->sc_md; 973219974Smav if (spare) { 974219974Smav if (mdi1->mdio_incomplete) 975219974Smav break; 976219974Smav } else { 977219974Smav if (mdi1->mdio_location == meta->raid_location && 978219974Smav memcmp(&mdi1->mdio_timestamp, 979219974Smav &meta->timestamp, 6) == 0) 980219974Smav break; 981219974Smav } 982219974Smav } 983219974Smav 984219974Smav /* Found matching node. */ 985219974Smav if (geom != NULL) { 986219974Smav G_RAID_DEBUG(1, "Found matching array %s", sc->sc_name); 987219974Smav result = G_RAID_MD_TASTE_EXISTING; 988219974Smav 989219974Smav } else if (spare) { /* Not found needy node -- left for later. */ 990219974Smav G_RAID_DEBUG(1, "Spare is not needed at this time"); 991219974Smav goto fail1; 992219974Smav 993219974Smav } else { /* Not found matching node -- create one. */ 994219974Smav result = G_RAID_MD_TASTE_NEW; 995219974Smav memcpy(&mdi->mdio_timestamp, &meta->timestamp, 6); 996219974Smav mdi->mdio_location = meta->raid_location; 997219974Smav snprintf(name, sizeof(name), "SiI-%02x%02x%02x%02x%02x%02x", 998219974Smav mdi->mdio_timestamp[5], mdi->mdio_timestamp[4], 999219974Smav mdi->mdio_timestamp[3], mdi->mdio_timestamp[2], 1000219974Smav mdi->mdio_timestamp[1], mdi->mdio_timestamp[0]); 1001219974Smav sc = g_raid_create_node(mp, name, md); 1002219974Smav md->mdo_softc = sc; 1003219974Smav geom = sc->sc_geom; 1004219974Smav callout_init(&mdi->mdio_start_co, 1); 1005219974Smav callout_reset(&mdi->mdio_start_co, g_raid_start_timeout * hz, 1006219974Smav g_raid_sii_go, sc); 1007219974Smav mdi->mdio_rootmount = root_mount_hold("GRAID-SiI"); 1008219974Smav G_RAID_DEBUG1(1, sc, "root_mount_hold %p", mdi->mdio_rootmount); 1009219974Smav } 1010219974Smav 1011265669Smav /* There is no return after this point, so we close passed consumer. */ 1012265669Smav g_access(cp, -1, 0, 0); 1013265669Smav 1014219974Smav rcp = g_new_consumer(geom); 1015260385Sscottl rcp->flags |= G_CF_DIRECT_RECEIVE; 1016219974Smav g_attach(rcp, pp); 1017219974Smav if (g_access(rcp, 1, 1, 1) != 0) 1018219974Smav ; //goto fail1; 1019219974Smav 1020219974Smav g_topology_unlock(); 1021219974Smav sx_xlock(&sc->sc_lock); 1022219974Smav 1023219974Smav pd = malloc(sizeof(*pd), M_MD_SII, M_WAITOK | M_ZERO); 1024219974Smav pd->pd_meta = meta; 1025219974Smav if (spare == 2) { 1026219974Smav pd->pd_disk_pos = -3; 1027219974Smav } else { 1028219974Smav pd->pd_disk_pos = -1; 1029219974Smav } 1030219974Smav pd->pd_disk_size = pp->mediasize; 1031219974Smav disk = g_raid_create_disk(sc); 1032219974Smav disk->d_md_data = (void *)pd; 1033219974Smav disk->d_consumer = rcp; 1034219974Smav rcp->private = disk; 1035219974Smav 1036242323Smav g_raid_get_disk_info(disk); 1037219974Smav 1038219974Smav g_raid_md_sii_new_disk(disk); 1039219974Smav 1040219974Smav sx_xunlock(&sc->sc_lock); 1041219974Smav g_topology_lock(); 1042219974Smav *gp = geom; 1043219974Smav return (result); 1044219974Smavfail1: 1045219974Smav free(meta, M_MD_SII); 1046219974Smav return (G_RAID_MD_TASTE_FAIL); 1047219974Smav} 1048219974Smav 1049219974Smavstatic int 1050219974Smavg_raid_md_event_sii(struct g_raid_md_object *md, 1051219974Smav struct g_raid_disk *disk, u_int event) 1052219974Smav{ 1053219974Smav struct g_raid_softc *sc; 1054219974Smav struct g_raid_subdisk *sd; 1055219974Smav struct g_raid_md_sii_object *mdi; 1056219974Smav struct g_raid_md_sii_perdisk *pd; 1057219974Smav 1058219974Smav sc = md->mdo_softc; 1059219974Smav mdi = (struct g_raid_md_sii_object *)md; 1060219974Smav if (disk == NULL) { 1061219974Smav switch (event) { 1062219974Smav case G_RAID_NODE_E_START: 1063219974Smav if (!mdi->mdio_started) 1064219974Smav g_raid_md_sii_start(sc); 1065219974Smav return (0); 1066219974Smav } 1067219974Smav return (-1); 1068219974Smav } 1069219974Smav pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data; 1070219974Smav switch (event) { 1071219974Smav case G_RAID_DISK_E_DISCONNECTED: 1072219974Smav /* If disk was assigned, just update statuses. */ 1073219974Smav if (pd->pd_disk_pos >= 0) { 1074219974Smav g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE); 1075219974Smav if (disk->d_consumer) { 1076219974Smav g_raid_kill_consumer(sc, disk->d_consumer); 1077219974Smav disk->d_consumer = NULL; 1078219974Smav } 1079219974Smav TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) { 1080219974Smav g_raid_change_subdisk_state(sd, 1081219974Smav G_RAID_SUBDISK_S_NONE); 1082219974Smav g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED, 1083219974Smav G_RAID_EVENT_SUBDISK); 1084219974Smav } 1085219974Smav } else { 1086219974Smav /* Otherwise -- delete. */ 1087219974Smav g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE); 1088219974Smav g_raid_destroy_disk(disk); 1089219974Smav } 1090219974Smav 1091219974Smav /* Write updated metadata to all disks. */ 1092219974Smav g_raid_md_write_sii(md, NULL, NULL, NULL); 1093219974Smav 1094219974Smav /* Check if anything left except placeholders. */ 1095219974Smav if (g_raid_ndisks(sc, -1) == 1096219974Smav g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE)) 1097219974Smav g_raid_destroy_node(sc, 0); 1098219974Smav else 1099219974Smav g_raid_md_sii_refill(sc); 1100219974Smav return (0); 1101219974Smav } 1102219974Smav return (-2); 1103219974Smav} 1104219974Smav 1105219974Smavstatic int 1106219974Smavg_raid_md_ctl_sii(struct g_raid_md_object *md, 1107219974Smav struct gctl_req *req) 1108219974Smav{ 1109219974Smav struct g_raid_softc *sc; 1110219974Smav struct g_raid_volume *vol; 1111219974Smav struct g_raid_subdisk *sd; 1112219974Smav struct g_raid_disk *disk; 1113219974Smav struct g_raid_md_sii_object *mdi; 1114219974Smav struct g_raid_md_sii_perdisk *pd; 1115219974Smav struct g_consumer *cp; 1116219974Smav struct g_provider *pp; 1117219974Smav char arg[16]; 1118219974Smav const char *verb, *volname, *levelname, *diskname; 1119219974Smav int *nargs, *force; 1120219974Smav off_t size, sectorsize, strip; 1121219974Smav intmax_t *sizearg, *striparg; 1122219974Smav int numdisks, i, len, level, qual, update; 1123219974Smav int error; 1124219974Smav 1125219974Smav sc = md->mdo_softc; 1126219974Smav mdi = (struct g_raid_md_sii_object *)md; 1127219974Smav verb = gctl_get_param(req, "verb", NULL); 1128219974Smav nargs = gctl_get_paraml(req, "nargs", sizeof(*nargs)); 1129219974Smav error = 0; 1130219974Smav if (strcmp(verb, "label") == 0) { 1131219974Smav 1132219974Smav if (*nargs < 4) { 1133219974Smav gctl_error(req, "Invalid number of arguments."); 1134219974Smav return (-1); 1135219974Smav } 1136219974Smav volname = gctl_get_asciiparam(req, "arg1"); 1137219974Smav if (volname == NULL) { 1138219974Smav gctl_error(req, "No volume name."); 1139219974Smav return (-2); 1140219974Smav } 1141219974Smav levelname = gctl_get_asciiparam(req, "arg2"); 1142219974Smav if (levelname == NULL) { 1143219974Smav gctl_error(req, "No RAID level."); 1144219974Smav return (-3); 1145219974Smav } 1146234458Smav if (strcasecmp(levelname, "RAID5") == 0) 1147234727Smav levelname = "RAID5-LS"; 1148219974Smav if (g_raid_volume_str2level(levelname, &level, &qual)) { 1149219974Smav gctl_error(req, "Unknown RAID level '%s'.", levelname); 1150219974Smav return (-4); 1151219974Smav } 1152219974Smav numdisks = *nargs - 3; 1153219974Smav force = gctl_get_paraml(req, "force", sizeof(*force)); 1154219974Smav if (!g_raid_md_sii_supported(level, qual, numdisks, 1155219974Smav force ? *force : 0)) { 1156219974Smav gctl_error(req, "Unsupported RAID level " 1157219974Smav "(0x%02x/0x%02x), or number of disks (%d).", 1158219974Smav level, qual, numdisks); 1159219974Smav return (-5); 1160219974Smav } 1161219974Smav 1162219974Smav /* Search for disks, connect them and probe. */ 1163219974Smav size = 0x7fffffffffffffffllu; 1164219974Smav sectorsize = 0; 1165219974Smav for (i = 0; i < numdisks; i++) { 1166219974Smav snprintf(arg, sizeof(arg), "arg%d", i + 3); 1167219974Smav diskname = gctl_get_asciiparam(req, arg); 1168219974Smav if (diskname == NULL) { 1169219974Smav gctl_error(req, "No disk name (%s).", arg); 1170219974Smav error = -6; 1171219974Smav break; 1172219974Smav } 1173219974Smav if (strcmp(diskname, "NONE") == 0) { 1174219974Smav cp = NULL; 1175219974Smav pp = NULL; 1176219974Smav } else { 1177219974Smav g_topology_lock(); 1178219974Smav cp = g_raid_open_consumer(sc, diskname); 1179219974Smav if (cp == NULL) { 1180219974Smav gctl_error(req, "Can't open '%s'.", 1181219974Smav diskname); 1182219974Smav g_topology_unlock(); 1183219974Smav error = -7; 1184219974Smav break; 1185219974Smav } 1186219974Smav pp = cp->provider; 1187219974Smav } 1188219974Smav pd = malloc(sizeof(*pd), M_MD_SII, M_WAITOK | M_ZERO); 1189219974Smav pd->pd_disk_pos = i; 1190219974Smav disk = g_raid_create_disk(sc); 1191219974Smav disk->d_md_data = (void *)pd; 1192219974Smav disk->d_consumer = cp; 1193219974Smav if (cp == NULL) 1194219974Smav continue; 1195219974Smav cp->private = disk; 1196219974Smav g_topology_unlock(); 1197219974Smav 1198242323Smav g_raid_get_disk_info(disk); 1199219974Smav 1200219974Smav pd->pd_disk_size = pp->mediasize; 1201219974Smav if (size > pp->mediasize) 1202219974Smav size = pp->mediasize; 1203219974Smav if (sectorsize < pp->sectorsize) 1204219974Smav sectorsize = pp->sectorsize; 1205219974Smav } 1206219974Smav if (error != 0) 1207219974Smav return (error); 1208219974Smav 1209220210Smav if (sectorsize <= 0) { 1210220210Smav gctl_error(req, "Can't get sector size."); 1211220210Smav return (-8); 1212220210Smav } 1213220210Smav 1214219974Smav /* Reserve space for metadata. */ 1215219974Smav size -= 0x800 * sectorsize; 1216219974Smav 1217219974Smav /* Handle size argument. */ 1218219974Smav len = sizeof(*sizearg); 1219219974Smav sizearg = gctl_get_param(req, "size", &len); 1220219974Smav if (sizearg != NULL && len == sizeof(*sizearg) && 1221219974Smav *sizearg > 0) { 1222219974Smav if (*sizearg > size) { 1223219974Smav gctl_error(req, "Size too big %lld > %lld.", 1224219974Smav (long long)*sizearg, (long long)size); 1225219974Smav return (-9); 1226219974Smav } 1227219974Smav size = *sizearg; 1228219974Smav } 1229219974Smav 1230219974Smav /* Handle strip argument. */ 1231219974Smav strip = 131072; 1232219974Smav len = sizeof(*striparg); 1233219974Smav striparg = gctl_get_param(req, "strip", &len); 1234219974Smav if (striparg != NULL && len == sizeof(*striparg) && 1235219974Smav *striparg > 0) { 1236219974Smav if (*striparg < sectorsize) { 1237219974Smav gctl_error(req, "Strip size too small."); 1238219974Smav return (-10); 1239219974Smav } 1240219974Smav if (*striparg % sectorsize != 0) { 1241219974Smav gctl_error(req, "Incorrect strip size."); 1242219974Smav return (-11); 1243219974Smav } 1244219974Smav if (strip > 65535 * sectorsize) { 1245219974Smav gctl_error(req, "Strip size too big."); 1246219974Smav return (-12); 1247219974Smav } 1248219974Smav strip = *striparg; 1249219974Smav } 1250219974Smav 1251219974Smav /* Round size down to strip or sector. */ 1252219974Smav if (level == G_RAID_VOLUME_RL_RAID1) 1253219974Smav size -= (size % sectorsize); 1254219974Smav else if (level == G_RAID_VOLUME_RL_RAID1E && 1255219974Smav (numdisks & 1) != 0) 1256219974Smav size -= (size % (2 * strip)); 1257219974Smav else 1258219974Smav size -= (size % strip); 1259219974Smav if (size <= 0) { 1260219974Smav gctl_error(req, "Size too small."); 1261219974Smav return (-13); 1262219974Smav } 1263219974Smav if (size > 0xffffffffffffllu * sectorsize) { 1264219974Smav gctl_error(req, "Size too big."); 1265219974Smav return (-14); 1266219974Smav } 1267219974Smav 1268219974Smav /* We have all we need, create things: volume, ... */ 1269219974Smav mdi->mdio_total_disks = numdisks; 1270219974Smav mdi->mdio_started = 1; 1271219974Smav vol = g_raid_create_volume(sc, volname, -1); 1272219974Smav vol->v_md_data = (void *)(intptr_t)0; 1273219974Smav vol->v_raid_level = level; 1274234458Smav vol->v_raid_level_qualifier = qual; 1275219974Smav vol->v_strip_size = strip; 1276219974Smav vol->v_disks_count = numdisks; 1277219974Smav if (level == G_RAID_VOLUME_RL_RAID0 || 1278219974Smav level == G_RAID_VOLUME_RL_CONCAT || 1279219974Smav level == G_RAID_VOLUME_RL_SINGLE) 1280219974Smav vol->v_mediasize = size * numdisks; 1281219974Smav else if (level == G_RAID_VOLUME_RL_RAID1) 1282219974Smav vol->v_mediasize = size; 1283219974Smav else if (level == G_RAID_VOLUME_RL_RAID5) 1284219974Smav vol->v_mediasize = size * (numdisks - 1); 1285219974Smav else { /* RAID1E */ 1286219974Smav vol->v_mediasize = ((size * numdisks) / strip / 2) * 1287219974Smav strip; 1288219974Smav } 1289219974Smav vol->v_sectorsize = sectorsize; 1290219974Smav g_raid_start_volume(vol); 1291219974Smav 1292219974Smav /* , and subdisks. */ 1293219974Smav TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 1294219974Smav pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data; 1295219974Smav sd = &vol->v_subdisks[pd->pd_disk_pos]; 1296219974Smav sd->sd_disk = disk; 1297219974Smav sd->sd_offset = 0; 1298219974Smav sd->sd_size = size; 1299219974Smav TAILQ_INSERT_TAIL(&disk->d_subdisks, sd, sd_next); 1300219974Smav if (sd->sd_disk->d_consumer != NULL) { 1301219974Smav g_raid_change_disk_state(disk, 1302219974Smav G_RAID_DISK_S_ACTIVE); 1303219974Smav g_raid_change_subdisk_state(sd, 1304219974Smav G_RAID_SUBDISK_S_ACTIVE); 1305219974Smav g_raid_event_send(sd, G_RAID_SUBDISK_E_NEW, 1306219974Smav G_RAID_EVENT_SUBDISK); 1307219974Smav } else { 1308219974Smav g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE); 1309219974Smav } 1310219974Smav } 1311219974Smav 1312219974Smav /* Write metadata based on created entities. */ 1313219974Smav G_RAID_DEBUG1(0, sc, "Array started."); 1314219974Smav g_raid_md_write_sii(md, NULL, NULL, NULL); 1315219974Smav 1316219974Smav /* Pickup any STALE/SPARE disks to refill array if needed. */ 1317219974Smav g_raid_md_sii_refill(sc); 1318219974Smav 1319219974Smav g_raid_event_send(vol, G_RAID_VOLUME_E_START, 1320219974Smav G_RAID_EVENT_VOLUME); 1321219974Smav return (0); 1322219974Smav } 1323219974Smav if (strcmp(verb, "delete") == 0) { 1324219974Smav 1325219974Smav /* Check if some volume is still open. */ 1326219974Smav force = gctl_get_paraml(req, "force", sizeof(*force)); 1327219974Smav if (force != NULL && *force == 0 && 1328219974Smav g_raid_nopens(sc) != 0) { 1329219974Smav gctl_error(req, "Some volume is still open."); 1330219974Smav return (-4); 1331219974Smav } 1332219974Smav 1333219974Smav TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 1334219974Smav if (disk->d_consumer) 1335219974Smav sii_meta_erase(disk->d_consumer); 1336219974Smav } 1337219974Smav g_raid_destroy_node(sc, 0); 1338219974Smav return (0); 1339219974Smav } 1340219974Smav if (strcmp(verb, "remove") == 0 || 1341219974Smav strcmp(verb, "fail") == 0) { 1342219974Smav if (*nargs < 2) { 1343219974Smav gctl_error(req, "Invalid number of arguments."); 1344219974Smav return (-1); 1345219974Smav } 1346219974Smav for (i = 1; i < *nargs; i++) { 1347219974Smav snprintf(arg, sizeof(arg), "arg%d", i); 1348219974Smav diskname = gctl_get_asciiparam(req, arg); 1349219974Smav if (diskname == NULL) { 1350219974Smav gctl_error(req, "No disk name (%s).", arg); 1351219974Smav error = -2; 1352219974Smav break; 1353219974Smav } 1354219974Smav if (strncmp(diskname, "/dev/", 5) == 0) 1355219974Smav diskname += 5; 1356219974Smav 1357219974Smav TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 1358219974Smav if (disk->d_consumer != NULL && 1359219974Smav disk->d_consumer->provider != NULL && 1360219974Smav strcmp(disk->d_consumer->provider->name, 1361219974Smav diskname) == 0) 1362219974Smav break; 1363219974Smav } 1364219974Smav if (disk == NULL) { 1365219974Smav gctl_error(req, "Disk '%s' not found.", 1366219974Smav diskname); 1367219974Smav error = -3; 1368219974Smav break; 1369219974Smav } 1370219974Smav 1371219974Smav if (strcmp(verb, "fail") == 0) { 1372219974Smav g_raid_md_fail_disk_sii(md, NULL, disk); 1373219974Smav continue; 1374219974Smav } 1375219974Smav 1376219974Smav pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data; 1377219974Smav 1378219974Smav /* Erase metadata on deleting disk. */ 1379219974Smav sii_meta_erase(disk->d_consumer); 1380219974Smav 1381219974Smav /* If disk was assigned, just update statuses. */ 1382219974Smav if (pd->pd_disk_pos >= 0) { 1383219974Smav g_raid_change_disk_state(disk, G_RAID_DISK_S_OFFLINE); 1384220209Smav g_raid_kill_consumer(sc, disk->d_consumer); 1385220209Smav disk->d_consumer = NULL; 1386219974Smav TAILQ_FOREACH(sd, &disk->d_subdisks, sd_next) { 1387219974Smav g_raid_change_subdisk_state(sd, 1388219974Smav G_RAID_SUBDISK_S_NONE); 1389219974Smav g_raid_event_send(sd, G_RAID_SUBDISK_E_DISCONNECTED, 1390219974Smav G_RAID_EVENT_SUBDISK); 1391219974Smav } 1392219974Smav } else { 1393219974Smav /* Otherwise -- delete. */ 1394219974Smav g_raid_change_disk_state(disk, G_RAID_DISK_S_NONE); 1395219974Smav g_raid_destroy_disk(disk); 1396219974Smav } 1397219974Smav } 1398219974Smav 1399219974Smav /* Write updated metadata to remaining disks. */ 1400219974Smav g_raid_md_write_sii(md, NULL, NULL, NULL); 1401219974Smav 1402219974Smav /* Check if anything left except placeholders. */ 1403219974Smav if (g_raid_ndisks(sc, -1) == 1404219974Smav g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE)) 1405219974Smav g_raid_destroy_node(sc, 0); 1406219974Smav else 1407219974Smav g_raid_md_sii_refill(sc); 1408219974Smav return (error); 1409219974Smav } 1410219974Smav if (strcmp(verb, "insert") == 0) { 1411219974Smav if (*nargs < 2) { 1412219974Smav gctl_error(req, "Invalid number of arguments."); 1413219974Smav return (-1); 1414219974Smav } 1415219974Smav update = 0; 1416219974Smav for (i = 1; i < *nargs; i++) { 1417219974Smav /* Get disk name. */ 1418219974Smav snprintf(arg, sizeof(arg), "arg%d", i); 1419219974Smav diskname = gctl_get_asciiparam(req, arg); 1420219974Smav if (diskname == NULL) { 1421219974Smav gctl_error(req, "No disk name (%s).", arg); 1422219974Smav error = -3; 1423219974Smav break; 1424219974Smav } 1425219974Smav 1426219974Smav /* Try to find provider with specified name. */ 1427219974Smav g_topology_lock(); 1428219974Smav cp = g_raid_open_consumer(sc, diskname); 1429219974Smav if (cp == NULL) { 1430219974Smav gctl_error(req, "Can't open disk '%s'.", 1431219974Smav diskname); 1432219974Smav g_topology_unlock(); 1433219974Smav error = -4; 1434219974Smav break; 1435219974Smav } 1436219974Smav pp = cp->provider; 1437219974Smav 1438219974Smav pd = malloc(sizeof(*pd), M_MD_SII, M_WAITOK | M_ZERO); 1439219974Smav pd->pd_disk_pos = -3; 1440219974Smav pd->pd_disk_size = pp->mediasize; 1441219974Smav 1442219974Smav disk = g_raid_create_disk(sc); 1443219974Smav disk->d_consumer = cp; 1444219974Smav disk->d_md_data = (void *)pd; 1445219974Smav cp->private = disk; 1446219974Smav g_topology_unlock(); 1447219974Smav 1448242323Smav g_raid_get_disk_info(disk); 1449219974Smav 1450219974Smav /* Welcome the "new" disk. */ 1451219974Smav update += g_raid_md_sii_start_disk(disk); 1452219974Smav if (disk->d_state == G_RAID_DISK_S_SPARE) { 1453219974Smav sii_meta_write_spare(cp); 1454219974Smav g_raid_destroy_disk(disk); 1455219974Smav } else if (disk->d_state != G_RAID_DISK_S_ACTIVE) { 1456219974Smav gctl_error(req, "Disk '%s' doesn't fit.", 1457219974Smav diskname); 1458219974Smav g_raid_destroy_disk(disk); 1459219974Smav error = -8; 1460219974Smav break; 1461219974Smav } 1462219974Smav } 1463219974Smav 1464219974Smav /* Write new metadata if we changed something. */ 1465219974Smav if (update) 1466219974Smav g_raid_md_write_sii(md, NULL, NULL, NULL); 1467219974Smav return (error); 1468219974Smav } 1469219974Smav gctl_error(req, "Command '%s' is not supported.", verb); 1470219974Smav return (-100); 1471219974Smav} 1472219974Smav 1473219974Smavstatic int 1474219974Smavg_raid_md_write_sii(struct g_raid_md_object *md, struct g_raid_volume *tvol, 1475219974Smav struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk) 1476219974Smav{ 1477219974Smav struct g_raid_softc *sc; 1478219974Smav struct g_raid_volume *vol; 1479219974Smav struct g_raid_subdisk *sd; 1480219974Smav struct g_raid_disk *disk; 1481219974Smav struct g_raid_md_sii_object *mdi; 1482219974Smav struct g_raid_md_sii_perdisk *pd; 1483219974Smav struct sii_raid_conf *meta; 1484219974Smav int i; 1485219974Smav 1486219974Smav sc = md->mdo_softc; 1487219974Smav mdi = (struct g_raid_md_sii_object *)md; 1488219974Smav 1489219974Smav if (sc->sc_stopping == G_RAID_DESTROY_HARD) 1490219974Smav return (0); 1491219974Smav 1492219974Smav /* Bump generation. Newly written metadata may differ from previous. */ 1493219974Smav mdi->mdio_generation++; 1494219974Smav 1495219974Smav /* There is only one volume. */ 1496219974Smav vol = TAILQ_FIRST(&sc->sc_volumes); 1497219974Smav 1498219974Smav /* Fill global fields. */ 1499219974Smav meta = malloc(sizeof(*meta), M_MD_SII, M_WAITOK | M_ZERO); 1500219974Smav if (mdi->mdio_meta) 1501219974Smav memcpy(meta, mdi->mdio_meta, sizeof(*meta)); 1502219974Smav meta->total_sectors = vol->v_mediasize / vol->v_sectorsize; 1503219974Smav meta->vendor_id = 0x1095; 1504219974Smav meta->version_minor = 0; 1505219974Smav meta->version_major = 2; 1506219974Smav memcpy(&meta->timestamp, &mdi->mdio_timestamp, 6); 1507219974Smav meta->strip_sectors = vol->v_strip_size / vol->v_sectorsize; 1508219974Smav if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID0) { 1509219974Smav meta->type = SII_T_RAID0; 1510219974Smav meta->raid0_disks = vol->v_disks_count; 1511219974Smav meta->raid1_disks = 0xff; 1512219974Smav } else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1) { 1513219974Smav meta->type = SII_T_RAID1; 1514219974Smav meta->raid0_disks = 0xff; 1515219974Smav meta->raid1_disks = vol->v_disks_count; 1516219974Smav } else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E) { 1517219974Smav meta->type = SII_T_RAID01; 1518219974Smav meta->raid0_disks = vol->v_disks_count / 2; 1519219974Smav meta->raid1_disks = 2; 1520219974Smav } else if (vol->v_raid_level == G_RAID_VOLUME_RL_CONCAT || 1521219974Smav vol->v_raid_level == G_RAID_VOLUME_RL_SINGLE) { 1522219974Smav meta->type = SII_T_JBOD; 1523219974Smav meta->raid0_disks = vol->v_disks_count; 1524219974Smav meta->raid1_disks = 0xff; 1525219974Smav } else { 1526219974Smav meta->type = SII_T_RAID5; 1527219974Smav meta->raid0_disks = vol->v_disks_count; 1528219974Smav meta->raid1_disks = 0xff; 1529219974Smav } 1530219974Smav meta->generation = mdi->mdio_generation; 1531219974Smav meta->raid_status = vol->v_dirty ? SII_S_ONLINE : SII_S_AVAILABLE; 1532219974Smav for (i = 0; i < vol->v_disks_count; i++) { 1533219974Smav sd = &vol->v_subdisks[i]; 1534219974Smav if (sd->sd_state == G_RAID_SUBDISK_S_STALE || 1535219974Smav sd->sd_state == G_RAID_SUBDISK_S_RESYNC) 1536219974Smav meta->raid_status = SII_S_ONLINE; 1537219974Smav } 1538219974Smav meta->raid_location = mdi->mdio_location; 1539219974Smav sii_meta_put_name(meta, vol->v_name); 1540219974Smav 1541219974Smav /* We are done. Print meta data and store them to disks. */ 1542219974Smav if (mdi->mdio_meta != NULL) 1543219974Smav free(mdi->mdio_meta, M_MD_SII); 1544219974Smav mdi->mdio_meta = meta; 1545219974Smav TAILQ_FOREACH(disk, &sc->sc_disks, d_next) { 1546219974Smav pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data; 1547219974Smav if (disk->d_state != G_RAID_DISK_S_ACTIVE) 1548219974Smav continue; 1549219974Smav if (pd->pd_meta != NULL) { 1550219974Smav free(pd->pd_meta, M_MD_SII); 1551219974Smav pd->pd_meta = NULL; 1552219974Smav } 1553219974Smav pd->pd_meta = sii_meta_copy(meta); 1554219974Smav if ((sd = TAILQ_FIRST(&disk->d_subdisks)) != NULL) { 1555219974Smav if (sd->sd_state < G_RAID_SUBDISK_S_NEW) 1556219974Smav pd->pd_meta->disk_status = SII_S_DROPPED; 1557219974Smav else if (sd->sd_state < G_RAID_SUBDISK_S_STALE) { 1558219974Smav pd->pd_meta->disk_status = SII_S_REBUILD; 1559219974Smav pd->pd_meta->rebuild_lba = 1560219974Smav sd->sd_rebuild_pos / vol->v_sectorsize; 1561219974Smav } else 1562219974Smav pd->pd_meta->disk_status = SII_S_CURRENT; 1563219974Smav if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1) { 1564219974Smav pd->pd_meta->disk_number = sd->sd_pos; 1565219974Smav pd->pd_meta->raid0_ident = 0xff; 1566219974Smav pd->pd_meta->raid1_ident = 0; 1567219974Smav } else if (vol->v_raid_level == G_RAID_VOLUME_RL_RAID1E) { 1568219974Smav pd->pd_meta->disk_number = sd->sd_pos / meta->raid1_disks; 1569219974Smav pd->pd_meta->raid0_ident = sd->sd_pos % meta->raid1_disks; 1570219974Smav pd->pd_meta->raid1_ident = sd->sd_pos / meta->raid1_disks; 1571219974Smav } else { 1572219974Smav pd->pd_meta->disk_number = sd->sd_pos; 1573219974Smav pd->pd_meta->raid0_ident = 0; 1574219974Smav pd->pd_meta->raid1_ident = 0xff; 1575219974Smav } 1576219974Smav } 1577219974Smav G_RAID_DEBUG(1, "Writing SiI metadata to %s", 1578219974Smav g_raid_get_diskname(disk)); 1579219974Smav g_raid_md_sii_print(pd->pd_meta); 1580219974Smav sii_meta_write(disk->d_consumer, pd->pd_meta); 1581219974Smav } 1582219974Smav return (0); 1583219974Smav} 1584219974Smav 1585219974Smavstatic int 1586219974Smavg_raid_md_fail_disk_sii(struct g_raid_md_object *md, 1587219974Smav struct g_raid_subdisk *tsd, struct g_raid_disk *tdisk) 1588219974Smav{ 1589219974Smav struct g_raid_softc *sc; 1590219974Smav struct g_raid_md_sii_perdisk *pd; 1591219974Smav struct g_raid_subdisk *sd; 1592219974Smav 1593219974Smav sc = md->mdo_softc; 1594219974Smav pd = (struct g_raid_md_sii_perdisk *)tdisk->d_md_data; 1595219974Smav 1596219974Smav /* We can't fail disk that is not a part of array now. */ 1597219974Smav if (pd->pd_disk_pos < 0) 1598219974Smav return (-1); 1599219974Smav 1600219974Smav /* 1601219974Smav * Mark disk as failed in metadata and try to write that metadata 1602219974Smav * to the disk itself to prevent it's later resurrection as STALE. 1603219974Smav */ 1604219974Smav if (tdisk->d_consumer) { 1605219974Smav if (pd->pd_meta) { 1606219974Smav pd->pd_meta->disk_status = SII_S_REMOVED; 1607219974Smav sii_meta_write(tdisk->d_consumer, pd->pd_meta); 1608219974Smav } else 1609219974Smav sii_meta_erase(tdisk->d_consumer); 1610219974Smav } 1611219974Smav 1612219974Smav /* Change states. */ 1613219974Smav g_raid_change_disk_state(tdisk, G_RAID_DISK_S_FAILED); 1614219974Smav TAILQ_FOREACH(sd, &tdisk->d_subdisks, sd_next) { 1615219974Smav g_raid_change_subdisk_state(sd, 1616219974Smav G_RAID_SUBDISK_S_FAILED); 1617219974Smav g_raid_event_send(sd, G_RAID_SUBDISK_E_FAILED, 1618219974Smav G_RAID_EVENT_SUBDISK); 1619219974Smav } 1620219974Smav 1621219974Smav /* Write updated metadata to remaining disks. */ 1622219974Smav g_raid_md_write_sii(md, NULL, NULL, tdisk); 1623219974Smav 1624219974Smav /* Check if anything left except placeholders. */ 1625219974Smav if (g_raid_ndisks(sc, -1) == 1626219974Smav g_raid_ndisks(sc, G_RAID_DISK_S_OFFLINE)) 1627219974Smav g_raid_destroy_node(sc, 0); 1628219974Smav else 1629219974Smav g_raid_md_sii_refill(sc); 1630219974Smav return (0); 1631219974Smav} 1632219974Smav 1633219974Smavstatic int 1634219974Smavg_raid_md_free_disk_sii(struct g_raid_md_object *md, 1635219974Smav struct g_raid_disk *disk) 1636219974Smav{ 1637219974Smav struct g_raid_md_sii_perdisk *pd; 1638219974Smav 1639219974Smav pd = (struct g_raid_md_sii_perdisk *)disk->d_md_data; 1640219974Smav if (pd->pd_meta != NULL) { 1641219974Smav free(pd->pd_meta, M_MD_SII); 1642219974Smav pd->pd_meta = NULL; 1643219974Smav } 1644219974Smav free(pd, M_MD_SII); 1645219974Smav disk->d_md_data = NULL; 1646219974Smav return (0); 1647219974Smav} 1648219974Smav 1649219974Smavstatic int 1650219974Smavg_raid_md_free_sii(struct g_raid_md_object *md) 1651219974Smav{ 1652219974Smav struct g_raid_md_sii_object *mdi; 1653219974Smav 1654219974Smav mdi = (struct g_raid_md_sii_object *)md; 1655219974Smav if (!mdi->mdio_started) { 1656219974Smav mdi->mdio_started = 0; 1657219974Smav callout_stop(&mdi->mdio_start_co); 1658219974Smav G_RAID_DEBUG1(1, md->mdo_softc, 1659219974Smav "root_mount_rel %p", mdi->mdio_rootmount); 1660219974Smav root_mount_rel(mdi->mdio_rootmount); 1661219974Smav mdi->mdio_rootmount = NULL; 1662219974Smav } 1663219974Smav if (mdi->mdio_meta != NULL) { 1664219974Smav free(mdi->mdio_meta, M_MD_SII); 1665219974Smav mdi->mdio_meta = NULL; 1666219974Smav } 1667219974Smav return (0); 1668219974Smav} 1669219974Smav 1670240465SmavG_RAID_MD_DECLARE(sii, "SiI"); 1671