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_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 #define G_MIRROR_DISK_FLAG_MASK (G_MIRROR_DISK_FLAG_DIRTY | \ 66 G_MIRROR_DISK_FLAG_SYNCHRONIZING | \ 67 G_MIRROR_DISK_FLAG_FORCE_SYNC | \ 68 G_MIRROR_DISK_FLAG_INACTIVE | \ 69 G_MIRROR_DISK_FLAG_CANDELETE) 70 71 #define G_MIRROR_DEVICE_FLAG_NOAUTOSYNC 0x0000000000000001ULL 72 #define G_MIRROR_DEVICE_FLAG_NOFAILSYNC 0x0000000000000002ULL 73 #define G_MIRROR_DEVICE_FLAG_MASK (G_MIRROR_DEVICE_FLAG_NOAUTOSYNC | \ 74 G_MIRROR_DEVICE_FLAG_NOFAILSYNC) 75 76 #ifdef _KERNEL 77 extern int g_mirror_debug; 78 79 #define G_MIRROR_DEBUG(lvl, ...) do { \ 80 if (g_mirror_debug >= (lvl)) { \ 81 printf("GEOM_MIRROR"); \ 82 if (g_mirror_debug > 0) \ 83 printf("[%u]", lvl); \ 84 printf(": "); \ 85 printf(__VA_ARGS__); \ 86 printf("\n"); \ 87 } \ 88 } while (0) 89 #define G_MIRROR_LOGREQ(lvl, bp, ...) do { \ 90 if (g_mirror_debug >= (lvl)) { \ 91 printf("GEOM_MIRROR"); \ 92 if (g_mirror_debug > 0) \ 93 printf("[%u]", lvl); \ 94 printf(": "); \ 95 printf(__VA_ARGS__); \ 96 printf(" "); \ 97 g_print_bio(bp); \ 98 printf("\n"); \ 99 } \ 100 } while (0) 101 102 #define G_MIRROR_BIO_FLAG_REGULAR 0x01 103 #define G_MIRROR_BIO_FLAG_SYNC 0x02 104 105 /* 106 * Informations needed for synchronization. 107 */ 108 struct g_mirror_disk_sync { 109 struct g_consumer *ds_consumer; /* Consumer connected to our mirror. */ 110 off_t ds_offset; /* Offset of next request to send. */ 111 off_t ds_offset_done; /* Offset of already synchronized 112 region. */ 113 time_t ds_update_ts; /* Time of last metadata update. */ 114 u_int ds_syncid; /* Disk's synchronization ID. */ 115 u_int ds_inflight; /* Number of in-flight sync requests. */ 116 struct bio **ds_bios; /* BIOs for synchronization I/O. */ 117 }; 118 119 /* 120 * Informations needed for synchronization. 121 */ 122 struct g_mirror_device_sync { 123 struct g_geom *ds_geom; /* Synchronization geom. */ 124 u_int ds_ndisks; /* Number of disks in SYNCHRONIZING 125 state. */ 126 }; 127 128 #define G_MIRROR_DISK_STATE_NONE 0 129 #define G_MIRROR_DISK_STATE_NEW 1 130 #define G_MIRROR_DISK_STATE_ACTIVE 2 131 #define G_MIRROR_DISK_STATE_STALE 3 132 #define G_MIRROR_DISK_STATE_SYNCHRONIZING 4 133 #define G_MIRROR_DISK_STATE_DISCONNECTED 5 134 #define G_MIRROR_DISK_STATE_DESTROY 6 135 struct g_mirror_disk { 136 uint32_t d_id; /* Disk ID. */ 137 struct g_consumer *d_consumer; /* Consumer. */ 138 struct g_mirror_softc *d_softc; /* Back-pointer to softc. */ 139 int d_state; /* Disk state. */ 140 u_int d_priority; /* Disk priority. */ 141 u_int load; /* Averaged queue length */ 142 off_t d_last_offset; /* Last read offset */ 143 uint64_t d_flags; /* Additional flags. */ 144 u_int d_genid; /* Disk's generation ID. */ 145 struct g_mirror_disk_sync d_sync;/* Sync information. */ 146 LIST_ENTRY(g_mirror_disk) d_next; 147 }; 148 #define d_name d_consumer->provider->name 149 150 #define G_MIRROR_EVENT_DONTWAIT 0x1 151 #define G_MIRROR_EVENT_WAIT 0x2 152 #define G_MIRROR_EVENT_DEVICE 0x4 153 #define G_MIRROR_EVENT_DONE 0x8 154 struct g_mirror_event { 155 struct g_mirror_disk *e_disk; 156 int e_state; 157 int e_flags; 158 int e_error; 159 TAILQ_ENTRY(g_mirror_event) e_next; 160 }; 161 162 #define G_MIRROR_DEVICE_FLAG_DESTROY 0x0100000000000000ULL 163 #define G_MIRROR_DEVICE_FLAG_DRAIN 0x0200000000000000ULL 164 #define G_MIRROR_DEVICE_FLAG_CLOSEWAIT 0x0400000000000000ULL 165 #define G_MIRROR_DEVICE_FLAG_TASTING 0x0800000000000000ULL 166 #define G_MIRROR_DEVICE_FLAG_WIPE 0x1000000000000000ULL 167 168 #define G_MIRROR_DEVICE_STATE_STARTING 0 169 #define G_MIRROR_DEVICE_STATE_RUNNING 1 170 171 #define G_MIRROR_TYPE_MANUAL 0 172 #define G_MIRROR_TYPE_AUTOMATIC 1 173 174 /* Bump syncid on first write. */ 175 #define G_MIRROR_BUMP_SYNCID 0x1 176 /* Bump genid immediately. */ 177 #define G_MIRROR_BUMP_GENID 0x2 178 /* Bump syncid immediately. */ 179 #define G_MIRROR_BUMP_SYNCID_NOW 0x4 180 struct g_mirror_softc { 181 u_int sc_type; /* Device type (manual/automatic). */ 182 u_int sc_state; /* Device state. */ 183 uint32_t sc_slice; /* Slice size. */ 184 uint8_t sc_balance; /* Balance algorithm. */ 185 uint64_t sc_mediasize; /* Device size. */ 186 uint32_t sc_sectorsize; /* Sector size. */ 187 uint64_t sc_flags; /* Additional flags. */ 188 189 struct g_geom *sc_geom; 190 struct g_provider *sc_provider; 191 int sc_provider_open; 192 193 uint32_t sc_id; /* Mirror unique ID. */ 194 195 struct sx sc_lock; 196 struct bio_queue_head sc_queue; 197 struct mtx sc_queue_mtx; 198 struct proc *sc_worker; 199 struct bio_queue_head sc_regular_delayed; /* Delayed I/O requests due 200 collision with sync 201 requests. */ 202 struct bio_queue_head sc_inflight; /* In-flight regular write 203 requests. */ 204 struct bio_queue_head sc_sync_delayed; /* Delayed sync requests due 205 collision with regular 206 requests. */ 207 208 LIST_HEAD(, g_mirror_disk) sc_disks; 209 u_int sc_ndisks; /* Number of disks. */ 210 struct g_mirror_disk *sc_hint; 211 212 u_int sc_genid; /* Generation ID. */ 213 u_int sc_syncid; /* Synchronization ID. */ 214 int sc_bump_id; 215 struct g_mirror_device_sync sc_sync; 216 int sc_idle; /* DIRTY flags removed. */ 217 time_t sc_last_write; 218 u_int sc_writes; 219 u_int sc_refcnt; /* Number of softc references */ 220 221 TAILQ_HEAD(, g_mirror_event) sc_events; 222 struct mtx sc_events_mtx; 223 224 struct callout sc_callout; 225 226 struct root_hold_token *sc_rootmount; 227 228 struct mtx sc_done_mtx; 229 }; 230 #define sc_name sc_geom->name 231 232 struct g_mirror_metadata; 233 234 u_int g_mirror_ndisks(struct g_mirror_softc *sc, int state); 235 struct g_geom * g_mirror_create(struct g_class *mp, 236 const struct g_mirror_metadata *md, u_int type); 237 #define G_MIRROR_DESTROY_SOFT 0 238 #define G_MIRROR_DESTROY_DELAYED 1 239 #define G_MIRROR_DESTROY_HARD 2 240 int g_mirror_destroy(struct g_mirror_softc *sc, int how); 241 int g_mirror_event_send(void *arg, int state, int flags); 242 struct g_mirror_metadata; 243 int g_mirror_add_disk(struct g_mirror_softc *sc, struct g_provider *pp, 244 struct g_mirror_metadata *md); 245 int g_mirror_read_metadata(struct g_consumer *cp, struct g_mirror_metadata *md); 246 void g_mirror_fill_metadata(struct g_mirror_softc *sc, 247 struct g_mirror_disk *disk, struct g_mirror_metadata *md); 248 void g_mirror_update_metadata(struct g_mirror_disk *disk); 249 250 g_ctl_req_t g_mirror_config; 251 #endif /* _KERNEL */ 252 253 struct g_mirror_metadata { 254 char md_magic[16]; /* Magic value. */ 255 uint32_t md_version; /* Version number. */ 256 char md_name[16]; /* Mirror name. */ 257 uint32_t md_mid; /* Mirror unique ID. */ 258 uint32_t md_did; /* Disk unique ID. */ 259 uint8_t md_all; /* Number of disks in mirror. */ 260 uint32_t md_genid; /* Generation ID. */ 261 uint32_t md_syncid; /* Synchronization ID. */ 262 uint8_t md_priority; /* Disk priority. */ 263 uint32_t md_slice; /* Slice size. */ 264 uint8_t md_balance; /* Balance type. */ 265 uint64_t md_mediasize; /* Size of the smallest 266 disk in mirror. */ 267 uint32_t md_sectorsize; /* Sector size. */ 268 uint64_t md_sync_offset; /* Synchronized offset. */ 269 uint64_t md_mflags; /* Additional mirror flags. */ 270 uint64_t md_dflags; /* Additional disk flags. */ 271 char md_provider[16]; /* Hardcoded provider. */ 272 uint64_t md_provsize; /* Provider's size. */ 273 u_char md_hash[16]; /* MD5 hash. */ 274 }; 275 static __inline void 276 mirror_metadata_encode(struct g_mirror_metadata *md, u_char *data) 277 { 278 MD5_CTX ctx; 279 280 bcopy(md->md_magic, data, 16); 281 le32enc(data + 16, md->md_version); 282 bcopy(md->md_name, data + 20, 16); 283 le32enc(data + 36, md->md_mid); 284 le32enc(data + 40, md->md_did); 285 *(data + 44) = md->md_all; 286 le32enc(data + 45, md->md_genid); 287 le32enc(data + 49, md->md_syncid); 288 *(data + 53) = md->md_priority; 289 le32enc(data + 54, md->md_slice); 290 *(data + 58) = md->md_balance; 291 le64enc(data + 59, md->md_mediasize); 292 le32enc(data + 67, md->md_sectorsize); 293 le64enc(data + 71, md->md_sync_offset); 294 le64enc(data + 79, md->md_mflags); 295 le64enc(data + 87, md->md_dflags); 296 bcopy(md->md_provider, data + 95, 16); 297 le64enc(data + 111, md->md_provsize); 298 MD5Init(&ctx); 299 MD5Update(&ctx, data, 119); 300 MD5Final(md->md_hash, &ctx); 301 bcopy(md->md_hash, data + 119, 16); 302 } 303 static __inline int 304 mirror_metadata_decode_v0v1(const u_char *data, struct g_mirror_metadata *md) 305 { 306 MD5_CTX ctx; 307 308 bcopy(data + 20, md->md_name, 16); 309 md->md_mid = le32dec(data + 36); 310 md->md_did = le32dec(data + 40); 311 md->md_all = *(data + 44); 312 md->md_syncid = le32dec(data + 45); 313 md->md_priority = *(data + 49); 314 md->md_slice = le32dec(data + 50); 315 md->md_balance = *(data + 54); 316 md->md_mediasize = le64dec(data + 55); 317 md->md_sectorsize = le32dec(data + 63); 318 md->md_sync_offset = le64dec(data + 67); 319 md->md_mflags = le64dec(data + 75); 320 md->md_dflags = le64dec(data + 83); 321 bcopy(data + 91, md->md_provider, 16); 322 bcopy(data + 107, md->md_hash, 16); 323 MD5Init(&ctx); 324 MD5Update(&ctx, data, 107); 325 MD5Final(md->md_hash, &ctx); 326 if (bcmp(md->md_hash, data + 107, 16) != 0) 327 return (EINVAL); 328 329 /* New fields. */ 330 md->md_genid = 0; 331 md->md_provsize = 0; 332 333 return (0); 334 } 335 static __inline int 336 mirror_metadata_decode_v2(const u_char *data, struct g_mirror_metadata *md) 337 { 338 MD5_CTX ctx; 339 340 bcopy(data + 20, md->md_name, 16); 341 md->md_mid = le32dec(data + 36); 342 md->md_did = le32dec(data + 40); 343 md->md_all = *(data + 44); 344 md->md_genid = le32dec(data + 45); 345 md->md_syncid = le32dec(data + 49); 346 md->md_priority = *(data + 53); 347 md->md_slice = le32dec(data + 54); 348 md->md_balance = *(data + 58); 349 md->md_mediasize = le64dec(data + 59); 350 md->md_sectorsize = le32dec(data + 67); 351 md->md_sync_offset = le64dec(data + 71); 352 md->md_mflags = le64dec(data + 79); 353 md->md_dflags = le64dec(data + 87); 354 bcopy(data + 95, md->md_provider, 16); 355 bcopy(data + 111, md->md_hash, 16); 356 MD5Init(&ctx); 357 MD5Update(&ctx, data, 111); 358 MD5Final(md->md_hash, &ctx); 359 if (bcmp(md->md_hash, data + 111, 16) != 0) 360 return (EINVAL); 361 362 /* New fields. */ 363 md->md_provsize = 0; 364 365 return (0); 366 } 367 static __inline int 368 mirror_metadata_decode_v3v4(const u_char *data, struct g_mirror_metadata *md) 369 { 370 MD5_CTX ctx; 371 372 bcopy(data + 20, md->md_name, 16); 373 md->md_mid = le32dec(data + 36); 374 md->md_did = le32dec(data + 40); 375 md->md_all = *(data + 44); 376 md->md_genid = le32dec(data + 45); 377 md->md_syncid = le32dec(data + 49); 378 md->md_priority = *(data + 53); 379 md->md_slice = le32dec(data + 54); 380 md->md_balance = *(data + 58); 381 md->md_mediasize = le64dec(data + 59); 382 md->md_sectorsize = le32dec(data + 67); 383 md->md_sync_offset = le64dec(data + 71); 384 md->md_mflags = le64dec(data + 79); 385 md->md_dflags = le64dec(data + 87); 386 bcopy(data + 95, md->md_provider, 16); 387 md->md_provsize = le64dec(data + 111); 388 bcopy(data + 119, md->md_hash, 16); 389 MD5Init(&ctx); 390 MD5Update(&ctx, data, 119); 391 MD5Final(md->md_hash, &ctx); 392 if (bcmp(md->md_hash, data + 119, 16) != 0) 393 return (EINVAL); 394 return (0); 395 } 396 static __inline int 397 mirror_metadata_decode(const u_char *data, struct g_mirror_metadata *md) 398 { 399 int error; 400 401 bcopy(data, md->md_magic, 16); 402 md->md_version = le32dec(data + 16); 403 switch (md->md_version) { 404 case 0: 405 case 1: 406 error = mirror_metadata_decode_v0v1(data, md); 407 break; 408 case 2: 409 error = mirror_metadata_decode_v2(data, md); 410 break; 411 case 3: 412 case 4: 413 error = mirror_metadata_decode_v3v4(data, md); 414 break; 415 default: 416 error = EINVAL; 417 break; 418 } 419 return (error); 420 } 421 422 static __inline const char * 423 balance_name(u_int balance) 424 { 425 static const char *algorithms[] = { 426 [G_MIRROR_BALANCE_NONE] = "none", 427 [G_MIRROR_BALANCE_ROUND_ROBIN] = "round-robin", 428 [G_MIRROR_BALANCE_LOAD] = "load", 429 [G_MIRROR_BALANCE_SPLIT] = "split", 430 [G_MIRROR_BALANCE_PREFER] = "prefer", 431 [G_MIRROR_BALANCE_MAX + 1] = "unknown" 432 }; 433 434 if (balance > G_MIRROR_BALANCE_MAX) 435 balance = G_MIRROR_BALANCE_MAX + 1; 436 437 return (algorithms[balance]); 438 } 439 440 static __inline int 441 balance_id(const char *name) 442 { 443 static const char *algorithms[] = { 444 [G_MIRROR_BALANCE_NONE] = "none", 445 [G_MIRROR_BALANCE_ROUND_ROBIN] = "round-robin", 446 [G_MIRROR_BALANCE_LOAD] = "load", 447 [G_MIRROR_BALANCE_SPLIT] = "split", 448 [G_MIRROR_BALANCE_PREFER] = "prefer" 449 }; 450 int n; 451 452 for (n = G_MIRROR_BALANCE_MIN; n <= G_MIRROR_BALANCE_MAX; n++) { 453 if (strcmp(name, algorithms[n]) == 0) 454 return (n); 455 } 456 return (-1); 457 } 458 459 static __inline void 460 mirror_metadata_dump(const struct g_mirror_metadata *md) 461 { 462 static const char hex[] = "0123456789abcdef"; 463 char hash[16 * 2 + 1]; 464 u_int i; 465 466 printf(" magic: %s\n", md->md_magic); 467 printf(" version: %u\n", (u_int)md->md_version); 468 printf(" name: %s\n", md->md_name); 469 printf(" mid: %u\n", (u_int)md->md_mid); 470 printf(" did: %u\n", (u_int)md->md_did); 471 printf(" all: %u\n", (u_int)md->md_all); 472 printf(" genid: %u\n", (u_int)md->md_genid); 473 printf(" syncid: %u\n", (u_int)md->md_syncid); 474 printf(" priority: %u\n", (u_int)md->md_priority); 475 printf(" slice: %u\n", (u_int)md->md_slice); 476 printf(" balance: %s\n", balance_name((u_int)md->md_balance)); 477 printf(" mediasize: %jd\n", (intmax_t)md->md_mediasize); 478 printf("sectorsize: %u\n", (u_int)md->md_sectorsize); 479 printf("syncoffset: %jd\n", (intmax_t)md->md_sync_offset); 480 printf(" mflags:"); 481 if (md->md_mflags == 0) 482 printf(" NONE"); 483 else { 484 if ((md->md_mflags & G_MIRROR_DEVICE_FLAG_NOFAILSYNC) != 0) 485 printf(" NOFAILSYNC"); 486 if ((md->md_mflags & G_MIRROR_DEVICE_FLAG_NOAUTOSYNC) != 0) 487 printf(" NOAUTOSYNC"); 488 } 489 printf("\n"); 490 printf(" dflags:"); 491 if (md->md_dflags == 0) 492 printf(" NONE"); 493 else { 494 if ((md->md_dflags & G_MIRROR_DISK_FLAG_DIRTY) != 0) 495 printf(" DIRTY"); 496 if ((md->md_dflags & G_MIRROR_DISK_FLAG_SYNCHRONIZING) != 0) 497 printf(" SYNCHRONIZING"); 498 if ((md->md_dflags & G_MIRROR_DISK_FLAG_FORCE_SYNC) != 0) 499 printf(" FORCE_SYNC"); 500 if ((md->md_dflags & G_MIRROR_DISK_FLAG_INACTIVE) != 0) 501 printf(" INACTIVE"); 502 } 503 printf("\n"); 504 printf("hcprovider: %s\n", md->md_provider); 505 printf(" provsize: %ju\n", (uintmax_t)md->md_provsize); 506 bzero(hash, sizeof(hash)); 507 for (i = 0; i < 16; i++) { 508 hash[i * 2] = hex[md->md_hash[i] >> 4]; 509 hash[i * 2 + 1] = hex[md->md_hash[i] & 0x0f]; 510 } 511 printf(" MD5 hash: %s\n", hash); 512 } 513 #endif /* !_G_MIRROR_H_ */ 514