1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause 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_MIRROR_H_ 32 #define _G_MIRROR_H_ 33 34 #include <sys/endian.h> 35 #include <sys/md5.h> 36 37 #define G_MIRROR_CLASS_NAME "MIRROR" 38 39 #define G_MIRROR_MAGIC "GEOM::MIRROR" 40 /* 41 * Version history: 42 * 0 - Initial version number. 43 * 1 - Added 'prefer' balance algorithm. 44 * 2 - Added md_genid field to metadata. 45 * 3 - Added md_provsize field to metadata. 46 * 4 - Added 'no failure synchronization' flag. 47 */ 48 #define G_MIRROR_VERSION 4 49 50 #define G_MIRROR_BALANCE_NONE 0 51 #define G_MIRROR_BALANCE_ROUND_ROBIN 1 52 #define G_MIRROR_BALANCE_LOAD 2 53 #define G_MIRROR_BALANCE_SPLIT 3 54 #define G_MIRROR_BALANCE_PREFER 4 55 #define G_MIRROR_BALANCE_MIN G_MIRROR_BALANCE_NONE 56 #define G_MIRROR_BALANCE_MAX G_MIRROR_BALANCE_PREFER 57 58 #define G_MIRROR_DISK_FLAG_DIRTY 0x0000000000000001ULL 59 #define G_MIRROR_DISK_FLAG_SYNCHRONIZING 0x0000000000000002ULL 60 #define G_MIRROR_DISK_FLAG_FORCE_SYNC 0x0000000000000004ULL 61 #define G_MIRROR_DISK_FLAG_INACTIVE 0x0000000000000008ULL 62 #define G_MIRROR_DISK_FLAG_HARDCODED 0x0000000000000010ULL 63 #define G_MIRROR_DISK_FLAG_BROKEN 0x0000000000000020ULL 64 #define G_MIRROR_DISK_FLAG_CANDELETE 0x0000000000000040ULL 65 66 /* Per-disk flags which are recorded in on-disk metadata. */ 67 #define G_MIRROR_DISK_FLAG_MASK (G_MIRROR_DISK_FLAG_DIRTY | \ 68 G_MIRROR_DISK_FLAG_SYNCHRONIZING | \ 69 G_MIRROR_DISK_FLAG_FORCE_SYNC | \ 70 G_MIRROR_DISK_FLAG_INACTIVE | \ 71 G_MIRROR_DISK_FLAG_CANDELETE) 72 73 #define G_MIRROR_DEVICE_FLAG_NOAUTOSYNC 0x0000000000000001ULL 74 #define G_MIRROR_DEVICE_FLAG_NOFAILSYNC 0x0000000000000002ULL 75 76 /* Mirror flags which are recorded in on-disk metadata. */ 77 #define G_MIRROR_DEVICE_FLAG_MASK (G_MIRROR_DEVICE_FLAG_NOAUTOSYNC | \ 78 G_MIRROR_DEVICE_FLAG_NOFAILSYNC) 79 80 #ifdef _KERNEL 81 #define G_MIRROR_DEVICE_FLAG_DESTROY 0x0100000000000000ULL 82 #define G_MIRROR_DEVICE_FLAG_DRAIN 0x0200000000000000ULL 83 #define G_MIRROR_DEVICE_FLAG_CLOSEWAIT 0x0400000000000000ULL 84 #define G_MIRROR_DEVICE_FLAG_TASTING 0x0800000000000000ULL 85 #define G_MIRROR_DEVICE_FLAG_WIPE 0x1000000000000000ULL 86 87 extern int g_mirror_debug; 88 89 #define G_MIRROR_DEBUG(lvl, ...) \ 90 _GEOM_DEBUG("GEOM_MIRROR", g_mirror_debug, (lvl), NULL, __VA_ARGS__) 91 #define G_MIRROR_LOGREQ(lvl, bp, ...) \ 92 _GEOM_DEBUG("GEOM_MIRROR", g_mirror_debug, (lvl), (bp), __VA_ARGS__) 93 94 #define G_MIRROR_BIO_FLAG_REGULAR 0x01 95 #define G_MIRROR_BIO_FLAG_SYNC 0x02 96 97 /* 98 * Informations needed for synchronization. 99 */ 100 struct g_mirror_disk_sync { 101 struct g_consumer *ds_consumer; /* Consumer connected to our mirror. */ 102 off_t ds_offset; /* Offset of next request to send. */ 103 off_t ds_offset_done; /* Offset of already synchronized 104 region. */ 105 time_t ds_update_ts; /* Time of last metadata update. */ 106 u_int ds_syncid; /* Disk's synchronization ID. */ 107 u_int ds_inflight; /* Number of in-flight sync requests. */ 108 struct bio **ds_bios; /* BIOs for synchronization I/O. */ 109 }; 110 111 /* 112 * Informations needed for synchronization. 113 */ 114 struct g_mirror_device_sync { 115 struct g_geom *ds_geom; /* Synchronization geom. */ 116 u_int ds_ndisks; /* Number of disks in SYNCHRONIZING 117 state. */ 118 }; 119 120 #define G_MIRROR_DISK_STATE_NONE 0 121 #define G_MIRROR_DISK_STATE_NEW 1 122 #define G_MIRROR_DISK_STATE_ACTIVE 2 123 #define G_MIRROR_DISK_STATE_STALE 3 124 #define G_MIRROR_DISK_STATE_SYNCHRONIZING 4 125 #define G_MIRROR_DISK_STATE_DISCONNECTED 5 126 #define G_MIRROR_DISK_STATE_DESTROY 6 127 struct g_mirror_disk { 128 uint32_t d_id; /* Disk ID. */ 129 struct g_consumer *d_consumer; /* Consumer. */ 130 struct g_mirror_softc *d_softc; /* Back-pointer to softc. */ 131 int d_state; /* Disk state. */ 132 u_int d_priority; /* Disk priority. */ 133 u_int load; /* Averaged queue length */ 134 off_t d_last_offset; /* Last read offset */ 135 uint64_t d_flags; /* Additional flags. */ 136 u_int d_genid; /* Disk's generation ID. */ 137 struct g_mirror_disk_sync d_sync;/* Sync information. */ 138 LIST_ENTRY(g_mirror_disk) d_next; 139 u_int d_init_ndisks; /* Initial number of mirror components */ 140 uint32_t d_init_slice; /* Initial slice size */ 141 uint8_t d_init_balance;/* Initial balance */ 142 uint16_t d_rotation_rate;/* Disk's rotation rate */ 143 uint64_t d_init_mediasize;/* Initial mediasize */ 144 }; 145 #define d_name d_consumer->provider->name 146 147 #define G_MIRROR_EVENT_DONTWAIT 0x1 148 #define G_MIRROR_EVENT_WAIT 0x2 149 #define G_MIRROR_EVENT_DEVICE 0x4 150 #define G_MIRROR_EVENT_DONE 0x8 151 struct g_mirror_event { 152 struct g_mirror_disk *e_disk; 153 int e_state; 154 int e_flags; 155 int e_error; 156 TAILQ_ENTRY(g_mirror_event) e_next; 157 }; 158 159 #define G_MIRROR_DEVICE_STATE_STARTING 0 160 #define G_MIRROR_DEVICE_STATE_RUNNING 1 161 162 #define G_MIRROR_TYPE_MANUAL 0 163 #define G_MIRROR_TYPE_AUTOMATIC 1 164 165 /* Bump syncid on first write. */ 166 #define G_MIRROR_BUMP_SYNCID 0x1 167 /* Bump genid immediately. */ 168 #define G_MIRROR_BUMP_GENID 0x2 169 /* Bump syncid immediately. */ 170 #define G_MIRROR_BUMP_SYNCID_NOW 0x4 171 struct g_mirror_softc { 172 u_int sc_type; /* Device type (manual/automatic). */ 173 u_int sc_state; /* Device state. */ 174 uint32_t sc_slice; /* Slice size. */ 175 uint8_t sc_balance; /* Balance algorithm. */ 176 uint64_t sc_mediasize; /* Device size. */ 177 uint32_t sc_sectorsize; /* Sector size. */ 178 uint64_t sc_flags; /* Additional flags. */ 179 180 struct g_geom *sc_geom; 181 struct g_provider *sc_provider; 182 int sc_provider_open; 183 184 uint32_t sc_id; /* Mirror unique ID. */ 185 186 struct sx sc_lock; 187 struct bio_queue sc_queue; 188 struct mtx sc_queue_mtx; 189 struct proc *sc_worker; 190 struct bio_queue sc_inflight; /* In-flight regular write requests. */ 191 struct bio_queue sc_regular_delayed; /* Delayed I/O requests due to 192 collision with sync requests. */ 193 struct bio_queue sc_sync_delayed; /* Delayed sync requests due to 194 collision with regular requests. */ 195 196 LIST_HEAD(, g_mirror_disk) sc_disks; 197 u_int sc_ndisks; /* Number of disks. */ 198 struct g_mirror_disk *sc_hint; 199 200 u_int sc_genid; /* Generation ID. */ 201 u_int sc_syncid; /* Synchronization ID. */ 202 int sc_bump_id; 203 struct g_mirror_device_sync sc_sync; 204 int sc_idle; /* DIRTY flags removed. */ 205 time_t sc_last_write; 206 u_int sc_writes; 207 u_int sc_refcnt; /* Number of softc references */ 208 209 TAILQ_HEAD(, g_mirror_event) sc_events; 210 struct mtx sc_events_mtx; 211 struct g_mirror_event *sc_timeout_event; 212 213 struct callout sc_callout; 214 215 struct root_hold_token *sc_rootmount; 216 217 struct mtx sc_done_mtx; 218 }; 219 #define sc_name sc_geom->name 220 221 struct g_mirror_metadata; 222 223 u_int g_mirror_ndisks(struct g_mirror_softc *sc, int state); 224 struct g_geom * g_mirror_create(struct g_class *mp, 225 const struct g_mirror_metadata *md, u_int type); 226 #define G_MIRROR_DESTROY_SOFT 0 227 #define G_MIRROR_DESTROY_DELAYED 1 228 #define G_MIRROR_DESTROY_HARD 2 229 int g_mirror_destroy(struct g_mirror_softc *sc, int how); 230 int g_mirror_event_send(void *arg, int state, int flags); 231 struct g_mirror_metadata; 232 int g_mirror_add_disk(struct g_mirror_softc *sc, struct g_provider *pp, 233 struct g_mirror_metadata *md); 234 int g_mirror_read_metadata(struct g_consumer *cp, struct g_mirror_metadata *md); 235 void g_mirror_fill_metadata(struct g_mirror_softc *sc, 236 struct g_mirror_disk *disk, struct g_mirror_metadata *md); 237 void g_mirror_update_metadata(struct g_mirror_disk *disk); 238 239 g_ctl_req_t g_mirror_config; 240 #endif /* _KERNEL */ 241 242 struct g_mirror_metadata { 243 char md_magic[16]; /* Magic value. */ 244 uint32_t md_version; /* Version number. */ 245 char md_name[16]; /* Mirror name. */ 246 uint32_t md_mid; /* Mirror unique ID. */ 247 uint32_t md_did; /* Disk unique ID. */ 248 uint8_t md_all; /* Number of disks in mirror. */ 249 uint32_t md_genid; /* Generation ID. */ 250 uint32_t md_syncid; /* Synchronization ID. */ 251 uint8_t md_priority; /* Disk priority. */ 252 uint32_t md_slice; /* Slice size. */ 253 uint8_t md_balance; /* Balance type. */ 254 uint64_t md_mediasize; /* Size of the smallest 255 disk in mirror. */ 256 uint32_t md_sectorsize; /* Sector size. */ 257 uint64_t md_sync_offset; /* Synchronized offset. */ 258 uint64_t md_mflags; /* Additional mirror flags. */ 259 uint64_t md_dflags; /* Additional disk flags. */ 260 char md_provider[16]; /* Hardcoded provider. */ 261 uint64_t md_provsize; /* Provider's size. */ 262 u_char md_hash[16]; /* MD5 hash. */ 263 }; 264 static __inline void 265 mirror_metadata_encode(struct g_mirror_metadata *md, u_char *data) 266 { 267 MD5_CTX ctx; 268 269 bcopy(md->md_magic, data, 16); 270 le32enc(data + 16, md->md_version); 271 bcopy(md->md_name, data + 20, 16); 272 le32enc(data + 36, md->md_mid); 273 le32enc(data + 40, md->md_did); 274 *(data + 44) = md->md_all; 275 le32enc(data + 45, md->md_genid); 276 le32enc(data + 49, md->md_syncid); 277 *(data + 53) = md->md_priority; 278 le32enc(data + 54, md->md_slice); 279 *(data + 58) = md->md_balance; 280 le64enc(data + 59, md->md_mediasize); 281 le32enc(data + 67, md->md_sectorsize); 282 le64enc(data + 71, md->md_sync_offset); 283 le64enc(data + 79, md->md_mflags); 284 le64enc(data + 87, md->md_dflags); 285 bcopy(md->md_provider, data + 95, 16); 286 le64enc(data + 111, md->md_provsize); 287 MD5Init(&ctx); 288 MD5Update(&ctx, data, 119); 289 MD5Final(md->md_hash, &ctx); 290 bcopy(md->md_hash, data + 119, 16); 291 } 292 static __inline int 293 mirror_metadata_decode_v0v1(const u_char *data, struct g_mirror_metadata *md) 294 { 295 MD5_CTX ctx; 296 297 bcopy(data + 20, md->md_name, 16); 298 md->md_mid = le32dec(data + 36); 299 md->md_did = le32dec(data + 40); 300 md->md_all = *(data + 44); 301 md->md_syncid = le32dec(data + 45); 302 md->md_priority = *(data + 49); 303 md->md_slice = le32dec(data + 50); 304 md->md_balance = *(data + 54); 305 md->md_mediasize = le64dec(data + 55); 306 md->md_sectorsize = le32dec(data + 63); 307 md->md_sync_offset = le64dec(data + 67); 308 md->md_mflags = le64dec(data + 75); 309 md->md_dflags = le64dec(data + 83); 310 bcopy(data + 91, md->md_provider, 16); 311 bcopy(data + 107, md->md_hash, 16); 312 MD5Init(&ctx); 313 MD5Update(&ctx, data, 107); 314 MD5Final(md->md_hash, &ctx); 315 if (bcmp(md->md_hash, data + 107, 16) != 0) 316 return (EINVAL); 317 318 /* New fields. */ 319 md->md_genid = 0; 320 md->md_provsize = 0; 321 322 return (0); 323 } 324 static __inline int 325 mirror_metadata_decode_v2(const u_char *data, struct g_mirror_metadata *md) 326 { 327 MD5_CTX ctx; 328 329 bcopy(data + 20, md->md_name, 16); 330 md->md_mid = le32dec(data + 36); 331 md->md_did = le32dec(data + 40); 332 md->md_all = *(data + 44); 333 md->md_genid = le32dec(data + 45); 334 md->md_syncid = le32dec(data + 49); 335 md->md_priority = *(data + 53); 336 md->md_slice = le32dec(data + 54); 337 md->md_balance = *(data + 58); 338 md->md_mediasize = le64dec(data + 59); 339 md->md_sectorsize = le32dec(data + 67); 340 md->md_sync_offset = le64dec(data + 71); 341 md->md_mflags = le64dec(data + 79); 342 md->md_dflags = le64dec(data + 87); 343 bcopy(data + 95, md->md_provider, 16); 344 bcopy(data + 111, md->md_hash, 16); 345 MD5Init(&ctx); 346 MD5Update(&ctx, data, 111); 347 MD5Final(md->md_hash, &ctx); 348 if (bcmp(md->md_hash, data + 111, 16) != 0) 349 return (EINVAL); 350 351 /* New fields. */ 352 md->md_provsize = 0; 353 354 return (0); 355 } 356 static __inline int 357 mirror_metadata_decode_v3v4(const u_char *data, struct g_mirror_metadata *md) 358 { 359 MD5_CTX ctx; 360 361 bcopy(data + 20, md->md_name, 16); 362 md->md_mid = le32dec(data + 36); 363 md->md_did = le32dec(data + 40); 364 md->md_all = *(data + 44); 365 md->md_genid = le32dec(data + 45); 366 md->md_syncid = le32dec(data + 49); 367 md->md_priority = *(data + 53); 368 md->md_slice = le32dec(data + 54); 369 md->md_balance = *(data + 58); 370 md->md_mediasize = le64dec(data + 59); 371 md->md_sectorsize = le32dec(data + 67); 372 md->md_sync_offset = le64dec(data + 71); 373 md->md_mflags = le64dec(data + 79); 374 md->md_dflags = le64dec(data + 87); 375 bcopy(data + 95, md->md_provider, 16); 376 md->md_provsize = le64dec(data + 111); 377 bcopy(data + 119, md->md_hash, 16); 378 MD5Init(&ctx); 379 MD5Update(&ctx, data, 119); 380 MD5Final(md->md_hash, &ctx); 381 if (bcmp(md->md_hash, data + 119, 16) != 0) 382 return (EINVAL); 383 return (0); 384 } 385 static __inline int 386 mirror_metadata_decode(const u_char *data, struct g_mirror_metadata *md) 387 { 388 int error; 389 390 bcopy(data, md->md_magic, 16); 391 md->md_version = le32dec(data + 16); 392 switch (md->md_version) { 393 case 0: 394 case 1: 395 error = mirror_metadata_decode_v0v1(data, md); 396 break; 397 case 2: 398 error = mirror_metadata_decode_v2(data, md); 399 break; 400 case 3: 401 case 4: 402 error = mirror_metadata_decode_v3v4(data, md); 403 break; 404 default: 405 error = EINVAL; 406 break; 407 } 408 return (error); 409 } 410 411 static __inline const char * 412 balance_name(u_int balance) 413 { 414 static const char *algorithms[] = { 415 [G_MIRROR_BALANCE_NONE] = "none", 416 [G_MIRROR_BALANCE_ROUND_ROBIN] = "round-robin", 417 [G_MIRROR_BALANCE_LOAD] = "load", 418 [G_MIRROR_BALANCE_SPLIT] = "split", 419 [G_MIRROR_BALANCE_PREFER] = "prefer", 420 [G_MIRROR_BALANCE_MAX + 1] = "unknown" 421 }; 422 423 if (balance > G_MIRROR_BALANCE_MAX) 424 balance = G_MIRROR_BALANCE_MAX + 1; 425 426 return (algorithms[balance]); 427 } 428 429 static __inline int 430 balance_id(const char *name) 431 { 432 static const char *algorithms[] = { 433 [G_MIRROR_BALANCE_NONE] = "none", 434 [G_MIRROR_BALANCE_ROUND_ROBIN] = "round-robin", 435 [G_MIRROR_BALANCE_LOAD] = "load", 436 [G_MIRROR_BALANCE_SPLIT] = "split", 437 [G_MIRROR_BALANCE_PREFER] = "prefer" 438 }; 439 int n; 440 441 for (n = G_MIRROR_BALANCE_MIN; n <= G_MIRROR_BALANCE_MAX; n++) { 442 if (strcmp(name, algorithms[n]) == 0) 443 return (n); 444 } 445 return (-1); 446 } 447 448 static __inline void 449 mirror_metadata_dump(const struct g_mirror_metadata *md) 450 { 451 static const char hex[] = "0123456789abcdef"; 452 char hash[16 * 2 + 1]; 453 u_int i; 454 455 printf(" magic: %s\n", md->md_magic); 456 printf(" version: %u\n", (u_int)md->md_version); 457 printf(" name: %s\n", md->md_name); 458 printf(" mid: %u\n", (u_int)md->md_mid); 459 printf(" did: %u\n", (u_int)md->md_did); 460 printf(" all: %u\n", (u_int)md->md_all); 461 printf(" genid: %u\n", (u_int)md->md_genid); 462 printf(" syncid: %u\n", (u_int)md->md_syncid); 463 printf(" priority: %u\n", (u_int)md->md_priority); 464 printf(" slice: %u\n", (u_int)md->md_slice); 465 printf(" balance: %s\n", balance_name((u_int)md->md_balance)); 466 printf(" mediasize: %jd\n", (intmax_t)md->md_mediasize); 467 printf("sectorsize: %u\n", (u_int)md->md_sectorsize); 468 printf("syncoffset: %jd\n", (intmax_t)md->md_sync_offset); 469 printf(" mflags:"); 470 if (md->md_mflags == 0) 471 printf(" NONE"); 472 else { 473 if ((md->md_mflags & G_MIRROR_DEVICE_FLAG_NOFAILSYNC) != 0) 474 printf(" NOFAILSYNC"); 475 if ((md->md_mflags & G_MIRROR_DEVICE_FLAG_NOAUTOSYNC) != 0) 476 printf(" NOAUTOSYNC"); 477 } 478 printf("\n"); 479 printf(" dflags:"); 480 if (md->md_dflags == 0) 481 printf(" NONE"); 482 else { 483 if ((md->md_dflags & G_MIRROR_DISK_FLAG_DIRTY) != 0) 484 printf(" DIRTY"); 485 if ((md->md_dflags & G_MIRROR_DISK_FLAG_SYNCHRONIZING) != 0) 486 printf(" SYNCHRONIZING"); 487 if ((md->md_dflags & G_MIRROR_DISK_FLAG_FORCE_SYNC) != 0) 488 printf(" FORCE_SYNC"); 489 if ((md->md_dflags & G_MIRROR_DISK_FLAG_INACTIVE) != 0) 490 printf(" INACTIVE"); 491 } 492 printf("\n"); 493 printf("hcprovider: %s\n", md->md_provider); 494 printf(" provsize: %ju\n", (uintmax_t)md->md_provsize); 495 bzero(hash, sizeof(hash)); 496 for (i = 0; i < 16; i++) { 497 hash[i * 2] = hex[md->md_hash[i] >> 4]; 498 hash[i * 2 + 1] = hex[md->md_hash[i] & 0x0f]; 499 } 500 printf(" MD5 hash: %s\n", hash); 501 } 502 #endif /* !_G_MIRROR_H_ */ 503