1/*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 2004-2006 Pawel Jakub Dawidek <pjd@FreeBSD.org> 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND 17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE 20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26 * SUCH DAMAGE. 27 * 28 * $FreeBSD$ 29 */ 30 31#ifndef _G_RAID3_H_ 32#define _G_RAID3_H_ 33 34#include <sys/endian.h> 35#include <sys/md5.h> 36 37#define G_RAID3_CLASS_NAME "RAID3" 38 39#define G_RAID3_MAGIC "GEOM::RAID3" 40/* 41 * Version history: 42 * 0 - Initial version number. 43 * 1 - Added 'round-robin reading' algorithm. 44 * 2 - Added 'verify reading' algorithm. 45 * 3 - Added md_genid field to metadata. 46 * 4 - Added md_provsize field to metadata. 47 * 5 - Added 'no failure synchronization' flag. 48 */ 49#define G_RAID3_VERSION 5 50 51#define G_RAID3_DISK_FLAG_DIRTY 0x0000000000000001ULL 52#define G_RAID3_DISK_FLAG_SYNCHRONIZING 0x0000000000000002ULL 53#define G_RAID3_DISK_FLAG_FORCE_SYNC 0x0000000000000004ULL 54#define G_RAID3_DISK_FLAG_HARDCODED 0x0000000000000008ULL 55#define G_RAID3_DISK_FLAG_BROKEN 0x0000000000000010ULL 56#define G_RAID3_DISK_FLAG_MASK (G_RAID3_DISK_FLAG_DIRTY | \ 57 G_RAID3_DISK_FLAG_SYNCHRONIZING | \ 58 G_RAID3_DISK_FLAG_FORCE_SYNC) 59 60#define G_RAID3_DEVICE_FLAG_NOAUTOSYNC 0x0000000000000001ULL 61#define G_RAID3_DEVICE_FLAG_ROUND_ROBIN 0x0000000000000002ULL 62#define G_RAID3_DEVICE_FLAG_VERIFY 0x0000000000000004ULL 63#define G_RAID3_DEVICE_FLAG_NOFAILSYNC 0x0000000000000008ULL 64#define G_RAID3_DEVICE_FLAG_MASK (G_RAID3_DEVICE_FLAG_NOAUTOSYNC | \ 65 G_RAID3_DEVICE_FLAG_ROUND_ROBIN | \ 66 G_RAID3_DEVICE_FLAG_VERIFY | \ 67 G_RAID3_DEVICE_FLAG_NOFAILSYNC) 68 69#ifdef _KERNEL 70extern u_int g_raid3_debug; 71 72#define G_RAID3_DEBUG(lvl, ...) do { \ 73 if (g_raid3_debug >= (lvl)) { \ 74 printf("GEOM_RAID3"); \ 75 if (g_raid3_debug > 0) \ 76 printf("[%u]", lvl); \ 77 printf(": "); \ 78 printf(__VA_ARGS__); \ 79 printf("\n"); \ 80 } \ 81} while (0) 82#define G_RAID3_LOGREQ(lvl, bp, ...) do { \ 83 if (g_raid3_debug >= (lvl)) { \ 84 printf("GEOM_RAID3"); \ 85 if (g_raid3_debug > 0) \ 86 printf("[%u]", lvl); \ 87 printf(": "); \ 88 printf(__VA_ARGS__); \ 89 printf(" "); \ 90 g_print_bio(bp); \ 91 printf("\n"); \ 92 } \ 93} while (0) 94 95#define G_RAID3_BIO_CFLAG_REGULAR 0x01 96#define G_RAID3_BIO_CFLAG_SYNC 0x02 97#define G_RAID3_BIO_CFLAG_PARITY 0x04 98#define G_RAID3_BIO_CFLAG_NODISK 0x08 99#define G_RAID3_BIO_CFLAG_REGSYNC 0x10 100#define G_RAID3_BIO_CFLAG_MASK (G_RAID3_BIO_CFLAG_REGULAR | \ 101 G_RAID3_BIO_CFLAG_SYNC | \ 102 G_RAID3_BIO_CFLAG_PARITY | \ 103 G_RAID3_BIO_CFLAG_NODISK | \ 104 G_RAID3_BIO_CFLAG_REGSYNC) 105 106#define G_RAID3_BIO_PFLAG_DEGRADED 0x01 107#define G_RAID3_BIO_PFLAG_NOPARITY 0x02 108#define G_RAID3_BIO_PFLAG_VERIFY 0x04 109#define G_RAID3_BIO_PFLAG_MASK (G_RAID3_BIO_PFLAG_DEGRADED | \ 110 G_RAID3_BIO_PFLAG_NOPARITY | \ 111 G_RAID3_BIO_PFLAG_VERIFY) 112 113/* 114 * Informations needed for synchronization. 115 */ 116struct g_raid3_disk_sync { 117 struct g_consumer *ds_consumer; /* Consumer connected to our device. */ 118 off_t ds_offset; /* Offset of next request to send. */ 119 off_t ds_offset_done; /* Offset of already synchronized 120 region. */ 121 off_t ds_resync; /* Resynchronize from this offset. */ 122 u_int ds_syncid; /* Disk's synchronization ID. */ 123 u_int ds_inflight; /* Number of in-flight sync requests. */ 124 struct bio **ds_bios; /* BIOs for synchronization I/O. */ 125}; 126 127/* 128 * Informations needed for synchronization. 129 */ 130struct g_raid3_device_sync { 131 struct g_geom *ds_geom; /* Synchronization geom. */ 132}; 133 134#define G_RAID3_DISK_STATE_NODISK 0 135#define G_RAID3_DISK_STATE_NONE 1 136#define G_RAID3_DISK_STATE_NEW 2 137#define G_RAID3_DISK_STATE_ACTIVE 3 138#define G_RAID3_DISK_STATE_STALE 4 139#define G_RAID3_DISK_STATE_SYNCHRONIZING 5 140#define G_RAID3_DISK_STATE_DISCONNECTED 6 141#define G_RAID3_DISK_STATE_DESTROY 7 142struct g_raid3_disk { 143 u_int d_no; /* Disk number. */ 144 struct g_consumer *d_consumer; /* Consumer. */ 145 struct g_raid3_softc *d_softc; /* Back-pointer to softc. */ 146 int d_state; /* Disk state. */ 147 uint64_t d_flags; /* Additional flags. */ 148 u_int d_genid; /* Disk's generation ID. */ 149 struct g_raid3_disk_sync d_sync; /* Sync information. */ 150 LIST_ENTRY(g_raid3_disk) d_next; 151}; 152#define d_name d_consumer->provider->name 153 154#define G_RAID3_EVENT_DONTWAIT 0x1 155#define G_RAID3_EVENT_WAIT 0x2 156#define G_RAID3_EVENT_DEVICE 0x4 157#define G_RAID3_EVENT_DONE 0x8 158struct g_raid3_event { 159 struct g_raid3_disk *e_disk; 160 int e_state; 161 int e_flags; 162 int e_error; 163 TAILQ_ENTRY(g_raid3_event) e_next; 164}; 165 166#define G_RAID3_DEVICE_FLAG_DESTROY 0x0100000000000000ULL 167#define G_RAID3_DEVICE_FLAG_WAIT 0x0200000000000000ULL 168#define G_RAID3_DEVICE_FLAG_DESTROYING 0x0400000000000000ULL 169 170#define G_RAID3_DEVICE_STATE_STARTING 0 171#define G_RAID3_DEVICE_STATE_DEGRADED 1 172#define G_RAID3_DEVICE_STATE_COMPLETE 2 173 174/* Bump syncid on first write. */ 175#define G_RAID3_BUMP_SYNCID 0x1 176/* Bump genid immediately. */ 177#define G_RAID3_BUMP_GENID 0x2 178 179enum g_raid3_zones { 180 G_RAID3_ZONE_64K, 181 G_RAID3_ZONE_16K, 182 G_RAID3_ZONE_4K, 183 G_RAID3_NUM_ZONES 184}; 185 186static __inline enum g_raid3_zones 187g_raid3_zone(size_t nbytes) { 188 if (nbytes > 65536) 189 return (G_RAID3_NUM_ZONES); 190 else if (nbytes > 16384) 191 return (G_RAID3_ZONE_64K); 192 else if (nbytes > 4096) 193 return (G_RAID3_ZONE_16K); 194 else 195 return (G_RAID3_ZONE_4K); 196}; 197 198struct g_raid3_softc { 199 u_int sc_state; /* Device state. */ 200 uint64_t sc_mediasize; /* Device size. */ 201 uint32_t sc_sectorsize; /* Sector size. */ 202 uint64_t sc_flags; /* Additional flags. */ 203 204 struct g_geom *sc_geom; 205 struct g_provider *sc_provider; 206 207 uint32_t sc_id; /* Device unique ID. */ 208 209 struct sx sc_lock; 210 struct bio_queue_head sc_queue; 211 struct mtx sc_queue_mtx; 212 struct proc *sc_worker; 213 struct bio_queue_head sc_regular_delayed; /* Delayed I/O requests due 214 collision with sync 215 requests. */ 216 struct bio_queue_head sc_inflight; /* In-flight regular write 217 requests. */ 218 struct bio_queue_head sc_sync_delayed; /* Delayed sync requests due 219 collision with regular 220 requests. */ 221 222 struct g_raid3_disk *sc_disks; 223 u_int sc_ndisks; /* Number of disks. */ 224 u_int sc_round_robin; 225 struct g_raid3_disk *sc_syncdisk; 226 227 struct g_raid3_zone { 228 uma_zone_t sz_zone; 229 size_t sz_inuse; 230 size_t sz_max; 231 u_int sz_requested; 232 u_int sz_failed; 233 } sc_zones[G_RAID3_NUM_ZONES]; 234 235 u_int sc_genid; /* Generation ID. */ 236 u_int sc_syncid; /* Synchronization ID. */ 237 int sc_bump_id; 238 struct g_raid3_device_sync sc_sync; 239 int sc_idle; /* DIRTY flags removed. */ 240 time_t sc_last_write; 241 u_int sc_writes; 242 243 TAILQ_HEAD(, g_raid3_event) sc_events; 244 struct mtx sc_events_mtx; 245 246 struct callout sc_callout; 247 248 struct root_hold_token *sc_rootmount; 249}; 250#define sc_name sc_geom->name 251 252const char *g_raid3_get_diskname(struct g_raid3_disk *disk); 253u_int g_raid3_ndisks(struct g_raid3_softc *sc, int state); 254#define G_RAID3_DESTROY_SOFT 0 255#define G_RAID3_DESTROY_DELAYED 1 256#define G_RAID3_DESTROY_HARD 2 257int g_raid3_destroy(struct g_raid3_softc *sc, int how); 258int g_raid3_event_send(void *arg, int state, int flags); 259struct g_raid3_metadata; 260int g_raid3_add_disk(struct g_raid3_softc *sc, struct g_provider *pp, 261 struct g_raid3_metadata *md); 262int g_raid3_read_metadata(struct g_consumer *cp, struct g_raid3_metadata *md); 263void g_raid3_fill_metadata(struct g_raid3_disk *disk, 264 struct g_raid3_metadata *md); 265int g_raid3_clear_metadata(struct g_raid3_disk *disk); 266void g_raid3_update_metadata(struct g_raid3_disk *disk); 267 268g_ctl_req_t g_raid3_config; 269#endif /* _KERNEL */ 270 271struct g_raid3_metadata { 272 char md_magic[16]; /* Magic value. */ 273 uint32_t md_version; /* Version number. */ 274 char md_name[16]; /* Device name. */ 275 uint32_t md_id; /* Device unique ID. */ 276 uint16_t md_no; /* Component number. */ 277 uint16_t md_all; /* Number of disks in device. */ 278 uint32_t md_genid; /* Generation ID. */ 279 uint32_t md_syncid; /* Synchronization ID. */ 280 uint64_t md_mediasize; /* Size of whole device. */ 281 uint32_t md_sectorsize; /* Sector size. */ 282 uint64_t md_sync_offset; /* Synchronized offset. */ 283 uint64_t md_mflags; /* Additional device flags. */ 284 uint64_t md_dflags; /* Additional disk flags. */ 285 char md_provider[16]; /* Hardcoded provider. */ 286 uint64_t md_provsize; /* Provider's size. */ 287 u_char md_hash[16]; /* MD5 hash. */ 288}; 289static __inline void 290raid3_metadata_encode(struct g_raid3_metadata *md, u_char *data) 291{ 292 MD5_CTX ctx; 293 294 bcopy(md->md_magic, data, 16); 295 le32enc(data + 16, md->md_version); 296 bcopy(md->md_name, data + 20, 16); 297 le32enc(data + 36, md->md_id); 298 le16enc(data + 40, md->md_no); 299 le16enc(data + 42, md->md_all); 300 le32enc(data + 44, md->md_genid); 301 le32enc(data + 48, md->md_syncid); 302 le64enc(data + 52, md->md_mediasize); 303 le32enc(data + 60, md->md_sectorsize); 304 le64enc(data + 64, md->md_sync_offset); 305 le64enc(data + 72, md->md_mflags); 306 le64enc(data + 80, md->md_dflags); 307 bcopy(md->md_provider, data + 88, 16); 308 le64enc(data + 104, md->md_provsize); 309 MD5Init(&ctx); 310 MD5Update(&ctx, data, 112); 311 MD5Final(md->md_hash, &ctx); 312 bcopy(md->md_hash, data + 112, 16); 313} 314static __inline int 315raid3_metadata_decode_v0v1v2(const u_char *data, struct g_raid3_metadata *md) 316{ 317 MD5_CTX ctx; 318 319 bcopy(data + 20, md->md_name, 16); 320 md->md_id = le32dec(data + 36); 321 md->md_no = le16dec(data + 40); 322 md->md_all = le16dec(data + 42); 323 md->md_syncid = le32dec(data + 44); 324 md->md_mediasize = le64dec(data + 48); 325 md->md_sectorsize = le32dec(data + 56); 326 md->md_sync_offset = le64dec(data + 60); 327 md->md_mflags = le64dec(data + 68); 328 md->md_dflags = le64dec(data + 76); 329 bcopy(data + 84, md->md_provider, 16); 330 bcopy(data + 100, md->md_hash, 16); 331 MD5Init(&ctx); 332 MD5Update(&ctx, data, 100); 333 MD5Final(md->md_hash, &ctx); 334 if (bcmp(md->md_hash, data + 100, 16) != 0) 335 return (EINVAL); 336 337 /* New fields. */ 338 md->md_genid = 0; 339 md->md_provsize = 0; 340 341 return (0); 342} 343static __inline int 344raid3_metadata_decode_v3(const u_char *data, struct g_raid3_metadata *md) 345{ 346 MD5_CTX ctx; 347 348 bcopy(data + 20, md->md_name, 16); 349 md->md_id = le32dec(data + 36); 350 md->md_no = le16dec(data + 40); 351 md->md_all = le16dec(data + 42); 352 md->md_genid = le32dec(data + 44); 353 md->md_syncid = le32dec(data + 48); 354 md->md_mediasize = le64dec(data + 52); 355 md->md_sectorsize = le32dec(data + 60); 356 md->md_sync_offset = le64dec(data + 64); 357 md->md_mflags = le64dec(data + 72); 358 md->md_dflags = le64dec(data + 80); 359 bcopy(data + 88, md->md_provider, 16); 360 bcopy(data + 104, md->md_hash, 16); 361 MD5Init(&ctx); 362 MD5Update(&ctx, data, 104); 363 MD5Final(md->md_hash, &ctx); 364 if (bcmp(md->md_hash, data + 104, 16) != 0) 365 return (EINVAL); 366 367 /* New fields. */ 368 md->md_provsize = 0; 369 370 return (0); 371} 372static __inline int 373raid3_metadata_decode_v4v5(const u_char *data, struct g_raid3_metadata *md) 374{ 375 MD5_CTX ctx; 376 377 bcopy(data + 20, md->md_name, 16); 378 md->md_id = le32dec(data + 36); 379 md->md_no = le16dec(data + 40); 380 md->md_all = le16dec(data + 42); 381 md->md_genid = le32dec(data + 44); 382 md->md_syncid = le32dec(data + 48); 383 md->md_mediasize = le64dec(data + 52); 384 md->md_sectorsize = le32dec(data + 60); 385 md->md_sync_offset = le64dec(data + 64); 386 md->md_mflags = le64dec(data + 72); 387 md->md_dflags = le64dec(data + 80); 388 bcopy(data + 88, md->md_provider, 16); 389 md->md_provsize = le64dec(data + 104); 390 bcopy(data + 112, md->md_hash, 16); 391 MD5Init(&ctx); 392 MD5Update(&ctx, data, 112); 393 MD5Final(md->md_hash, &ctx); 394 if (bcmp(md->md_hash, data + 112, 16) != 0) 395 return (EINVAL); 396 return (0); 397} 398static __inline int 399raid3_metadata_decode(const u_char *data, struct g_raid3_metadata *md) 400{ 401 int error; 402 403 bcopy(data, md->md_magic, 16); 404 md->md_version = le32dec(data + 16); 405 switch (md->md_version) { 406 case 0: 407 case 1: 408 case 2: 409 error = raid3_metadata_decode_v0v1v2(data, md); 410 break; 411 case 3: 412 error = raid3_metadata_decode_v3(data, md); 413 break; 414 case 4: 415 case 5: 416 error = raid3_metadata_decode_v4v5(data, md); 417 break; 418 default: 419 error = EINVAL; 420 break; 421 } 422 return (error); 423} 424 425static __inline void 426raid3_metadata_dump(const struct g_raid3_metadata *md) 427{ 428 static const char hex[] = "0123456789abcdef"; 429 char hash[16 * 2 + 1]; 430 u_int i; 431 432 printf(" magic: %s\n", md->md_magic); 433 printf(" version: %u\n", (u_int)md->md_version); 434 printf(" name: %s\n", md->md_name); 435 printf(" id: %u\n", (u_int)md->md_id); 436 printf(" no: %u\n", (u_int)md->md_no); 437 printf(" all: %u\n", (u_int)md->md_all); 438 printf(" genid: %u\n", (u_int)md->md_genid); 439 printf(" syncid: %u\n", (u_int)md->md_syncid); 440 printf(" mediasize: %jd\n", (intmax_t)md->md_mediasize); 441 printf("sectorsize: %u\n", (u_int)md->md_sectorsize); 442 printf("syncoffset: %jd\n", (intmax_t)md->md_sync_offset); 443 printf(" mflags:"); 444 if (md->md_mflags == 0) 445 printf(" NONE"); 446 else { 447 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_NOAUTOSYNC) != 0) 448 printf(" NOAUTOSYNC"); 449 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_ROUND_ROBIN) != 0) 450 printf(" ROUND-ROBIN"); 451 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_VERIFY) != 0) 452 printf(" VERIFY"); 453 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_NOFAILSYNC) != 0) 454 printf(" NOFAILSYNC"); 455 } 456 printf("\n"); 457 printf(" dflags:"); 458 if (md->md_dflags == 0) 459 printf(" NONE"); 460 else { 461 if ((md->md_dflags & G_RAID3_DISK_FLAG_DIRTY) != 0) 462 printf(" DIRTY"); 463 if ((md->md_dflags & G_RAID3_DISK_FLAG_SYNCHRONIZING) != 0) 464 printf(" SYNCHRONIZING"); 465 if ((md->md_dflags & G_RAID3_DISK_FLAG_FORCE_SYNC) != 0) 466 printf(" FORCE_SYNC"); 467 } 468 printf("\n"); 469 printf("hcprovider: %s\n", md->md_provider); 470 printf(" provsize: %ju\n", (uintmax_t)md->md_provsize); 471 bzero(hash, sizeof(hash)); 472 for (i = 0; i < 16; i++) { 473 hash[i * 2] = hex[md->md_hash[i] >> 4]; 474 hash[i * 2 + 1] = hex[md->md_hash[i] & 0x0f]; 475 } 476 printf(" MD5 hash: %s\n", hash); 477} 478#endif /* !_G_RAID3_H_ */ 479