1 /*- 2 * Copyright (c) 2004-2006 Pawel Jakub Dawidek <pjd@FreeBSD.org> 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 14 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND 15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE 18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24 * SUCH DAMAGE. 25 * 26 * $FreeBSD$ 27 */ 28 29 #ifndef _G_MIRROR_H_ 30 #define _G_MIRROR_H_ 31 32 #include <sys/endian.h> 33 #include <sys/md5.h> 34 35 #define G_MIRROR_CLASS_NAME "MIRROR" 36 37 #define G_MIRROR_MAGIC "GEOM::MIRROR" 38 /* 39 * Version history: 40 * 0 - Initial version number. 41 * 1 - Added 'prefer' balance algorithm. 42 * 2 - Added md_genid field to metadata. 43 * 3 - Added md_provsize field to metadata. 44 * 4 - Added 'no failure synchronization' flag. 45 */ 46 #define G_MIRROR_VERSION 4 47 48 #define G_MIRROR_BALANCE_NONE 0 49 #define G_MIRROR_BALANCE_ROUND_ROBIN 1 50 #define G_MIRROR_BALANCE_LOAD 2 51 #define G_MIRROR_BALANCE_SPLIT 3 52 #define G_MIRROR_BALANCE_PREFER 4 53 #define G_MIRROR_BALANCE_MIN G_MIRROR_BALANCE_NONE 54 #define G_MIRROR_BALANCE_MAX G_MIRROR_BALANCE_PREFER 55 56 #define G_MIRROR_DISK_FLAG_DIRTY 0x0000000000000001ULL 57 #define G_MIRROR_DISK_FLAG_SYNCHRONIZING 0x0000000000000002ULL 58 #define G_MIRROR_DISK_FLAG_FORCE_SYNC 0x0000000000000004ULL 59 #define G_MIRROR_DISK_FLAG_INACTIVE 0x0000000000000008ULL 60 #define G_MIRROR_DISK_FLAG_HARDCODED 0x0000000000000010ULL 61 #define G_MIRROR_DISK_FLAG_BROKEN 0x0000000000000020ULL 62 #define G_MIRROR_DISK_FLAG_MASK (G_MIRROR_DISK_FLAG_DIRTY | \ 63 G_MIRROR_DISK_FLAG_SYNCHRONIZING | \ 64 G_MIRROR_DISK_FLAG_FORCE_SYNC | \ 65 G_MIRROR_DISK_FLAG_INACTIVE) 66 67 #define G_MIRROR_DEVICE_FLAG_NOAUTOSYNC 0x0000000000000001ULL 68 #define G_MIRROR_DEVICE_FLAG_NOFAILSYNC 0x0000000000000002ULL 69 #define G_MIRROR_DEVICE_FLAG_MASK (G_MIRROR_DEVICE_FLAG_NOAUTOSYNC | \ 70 G_MIRROR_DEVICE_FLAG_NOFAILSYNC) 71 72 #ifdef _KERNEL 73 extern u_int g_mirror_debug; 74 75 #define G_MIRROR_DEBUG(lvl, ...) do { \ 76 if (g_mirror_debug >= (lvl)) { \ 77 printf("GEOM_MIRROR"); \ 78 if (g_mirror_debug > 0) \ 79 printf("[%u]", lvl); \ 80 printf(": "); \ 81 printf(__VA_ARGS__); \ 82 printf("\n"); \ 83 } \ 84 } while (0) 85 #define G_MIRROR_LOGREQ(lvl, bp, ...) do { \ 86 if (g_mirror_debug >= (lvl)) { \ 87 printf("GEOM_MIRROR"); \ 88 if (g_mirror_debug > 0) \ 89 printf("[%u]", lvl); \ 90 printf(": "); \ 91 printf(__VA_ARGS__); \ 92 printf(" "); \ 93 g_print_bio(bp); \ 94 printf("\n"); \ 95 } \ 96 } while (0) 97 98 #define G_MIRROR_BIO_FLAG_REGULAR 0x01 99 #define G_MIRROR_BIO_FLAG_SYNC 0x02 100 101 /* 102 * Informations needed for synchronization. 103 */ 104 struct g_mirror_disk_sync { 105 struct g_consumer *ds_consumer; /* Consumer connected to our mirror. */ 106 off_t ds_offset; /* Offset of next request to send. */ 107 off_t ds_offset_done; /* Offset of already synchronized 108 region. */ 109 u_int ds_syncid; /* Disk's synchronization ID. */ 110 u_int ds_inflight; /* Number of in-flight sync requests. */ 111 struct bio **ds_bios; /* BIOs for synchronization I/O. */ 112 }; 113 114 /* 115 * Informations needed for synchronization. 116 */ 117 struct g_mirror_device_sync { 118 struct g_geom *ds_geom; /* Synchronization geom. */ 119 u_int ds_ndisks; /* Number of disks in SYNCHRONIZING 120 state. */ 121 }; 122 123 #define G_MIRROR_DISK_STATE_NONE 0 124 #define G_MIRROR_DISK_STATE_NEW 1 125 #define G_MIRROR_DISK_STATE_ACTIVE 2 126 #define G_MIRROR_DISK_STATE_STALE 3 127 #define G_MIRROR_DISK_STATE_SYNCHRONIZING 4 128 #define G_MIRROR_DISK_STATE_DISCONNECTED 5 129 #define G_MIRROR_DISK_STATE_DESTROY 6 130 struct g_mirror_disk { 131 uint32_t d_id; /* Disk ID. */ 132 struct g_consumer *d_consumer; /* Consumer. */ 133 struct g_mirror_softc *d_softc; /* Back-pointer to softc. */ 134 int d_state; /* Disk state. */ 135 u_int d_priority; /* Disk priority. */ 136 struct bintime d_delay; /* Disk delay. */ 137 struct bintime d_last_used; /* When disk was last used. */ 138 uint64_t d_flags; /* Additional flags. */ 139 u_int d_genid; /* Disk's generation ID. */ 140 struct g_mirror_disk_sync d_sync;/* Sync information. */ 141 LIST_ENTRY(g_mirror_disk) d_next; 142 }; 143 #define d_name d_consumer->provider->name 144 145 #define G_MIRROR_EVENT_DONTWAIT 0x1 146 #define G_MIRROR_EVENT_WAIT 0x2 147 #define G_MIRROR_EVENT_DEVICE 0x4 148 #define G_MIRROR_EVENT_DONE 0x8 149 struct g_mirror_event { 150 struct g_mirror_disk *e_disk; 151 int e_state; 152 int e_flags; 153 int e_error; 154 TAILQ_ENTRY(g_mirror_event) e_next; 155 }; 156 157 #define G_MIRROR_DEVICE_FLAG_DESTROY 0x0100000000000000ULL 158 #define G_MIRROR_DEVICE_FLAG_WAIT 0x0200000000000000ULL 159 #define G_MIRROR_DEVICE_FLAG_DESTROYING 0x0400000000000000ULL 160 161 #define G_MIRROR_DEVICE_STATE_STARTING 0 162 #define G_MIRROR_DEVICE_STATE_RUNNING 1 163 164 /* Bump syncid on first write. */ 165 #define G_MIRROR_BUMP_SYNCID 0x1 166 /* Bump genid immediately. */ 167 #define G_MIRROR_BUMP_GENID 0x2 168 struct g_mirror_softc { 169 u_int sc_state; /* Device state. */ 170 uint32_t sc_slice; /* Slice size. */ 171 uint8_t sc_balance; /* Balance algorithm. */ 172 uint64_t sc_mediasize; /* Device size. */ 173 uint32_t sc_sectorsize; /* Sector size. */ 174 uint64_t sc_flags; /* Additional flags. */ 175 176 struct g_geom *sc_geom; 177 struct g_provider *sc_provider; 178 179 uint32_t sc_id; /* Mirror unique ID. */ 180 181 struct sx sc_lock; 182 struct bio_queue_head sc_queue; 183 struct mtx sc_queue_mtx; 184 struct proc *sc_worker; 185 struct bio_queue_head sc_regular_delayed; /* Delayed I/O requests due 186 collision with sync 187 requests. */ 188 struct bio_queue_head sc_inflight; /* In-flight regular write 189 requests. */ 190 struct bio_queue_head sc_sync_delayed; /* Delayed sync requests due 191 collision with regular 192 requests. */ 193 194 LIST_HEAD(, g_mirror_disk) sc_disks; 195 u_int sc_ndisks; /* Number of disks. */ 196 struct g_mirror_disk *sc_hint; 197 198 u_int sc_genid; /* Generation ID. */ 199 u_int sc_syncid; /* Synchronization ID. */ 200 int sc_bump_id; 201 struct g_mirror_device_sync sc_sync; 202 int sc_idle; /* DIRTY flags removed. */ 203 time_t sc_last_write; 204 u_int sc_writes; 205 206 TAILQ_HEAD(, g_mirror_event) sc_events; 207 struct mtx sc_events_mtx; 208 209 struct callout sc_callout; 210 211 struct root_hold_token *sc_rootmount; 212 }; 213 #define sc_name sc_geom->name 214 215 u_int g_mirror_ndisks(struct g_mirror_softc *sc, int state); 216 #define G_MIRROR_DESTROY_SOFT 0 217 #define G_MIRROR_DESTROY_DELAYED 1 218 #define G_MIRROR_DESTROY_HARD 2 219 int g_mirror_destroy(struct g_mirror_softc *sc, int how); 220 int g_mirror_event_send(void *arg, int state, int flags); 221 struct g_mirror_metadata; 222 int g_mirror_add_disk(struct g_mirror_softc *sc, struct g_provider *pp, 223 struct g_mirror_metadata *md); 224 int g_mirror_read_metadata(struct g_consumer *cp, struct g_mirror_metadata *md); 225 void g_mirror_fill_metadata(struct g_mirror_softc *sc, 226 struct g_mirror_disk *disk, struct g_mirror_metadata *md); 227 void g_mirror_update_metadata(struct g_mirror_disk *disk); 228 229 g_ctl_req_t g_mirror_config; 230 #endif /* _KERNEL */ 231 232 struct g_mirror_metadata { 233 char md_magic[16]; /* Magic value. */ 234 uint32_t md_version; /* Version number. */ 235 char md_name[16]; /* Mirror name. */ 236 uint32_t md_mid; /* Mirror unique ID. */ 237 uint32_t md_did; /* Disk unique ID. */ 238 uint8_t md_all; /* Number of disks in mirror. */ 239 uint32_t md_genid; /* Generation ID. */ 240 uint32_t md_syncid; /* Synchronization ID. */ 241 uint8_t md_priority; /* Disk priority. */ 242 uint32_t md_slice; /* Slice size. */ 243 uint8_t md_balance; /* Balance type. */ 244 uint64_t md_mediasize; /* Size of the smallest 245 disk in mirror. */ 246 uint32_t md_sectorsize; /* Sector size. */ 247 uint64_t md_sync_offset; /* Synchronized offset. */ 248 uint64_t md_mflags; /* Additional mirror flags. */ 249 uint64_t md_dflags; /* Additional disk flags. */ 250 char md_provider[16]; /* Hardcoded provider. */ 251 uint64_t md_provsize; /* Provider's size. */ 252 u_char md_hash[16]; /* MD5 hash. */ 253 }; 254 static __inline void 255 mirror_metadata_encode(struct g_mirror_metadata *md, u_char *data) 256 { 257 MD5_CTX ctx; 258 259 bcopy(md->md_magic, data, 16); 260 le32enc(data + 16, md->md_version); 261 bcopy(md->md_name, data + 20, 16); 262 le32enc(data + 36, md->md_mid); 263 le32enc(data + 40, md->md_did); 264 *(data + 44) = md->md_all; 265 le32enc(data + 45, md->md_genid); 266 le32enc(data + 49, md->md_syncid); 267 *(data + 53) = md->md_priority; 268 le32enc(data + 54, md->md_slice); 269 *(data + 58) = md->md_balance; 270 le64enc(data + 59, md->md_mediasize); 271 le32enc(data + 67, md->md_sectorsize); 272 le64enc(data + 71, md->md_sync_offset); 273 le64enc(data + 79, md->md_mflags); 274 le64enc(data + 87, md->md_dflags); 275 bcopy(md->md_provider, data + 95, 16); 276 le64enc(data + 111, md->md_provsize); 277 MD5Init(&ctx); 278 MD5Update(&ctx, data, 119); 279 MD5Final(md->md_hash, &ctx); 280 bcopy(md->md_hash, data + 119, 16); 281 } 282 static __inline int 283 mirror_metadata_decode_v0v1(const u_char *data, struct g_mirror_metadata *md) 284 { 285 MD5_CTX ctx; 286 287 bcopy(data + 20, md->md_name, 16); 288 md->md_mid = le32dec(data + 36); 289 md->md_did = le32dec(data + 40); 290 md->md_all = *(data + 44); 291 md->md_syncid = le32dec(data + 45); 292 md->md_priority = *(data + 49); 293 md->md_slice = le32dec(data + 50); 294 md->md_balance = *(data + 54); 295 md->md_mediasize = le64dec(data + 55); 296 md->md_sectorsize = le32dec(data + 63); 297 md->md_sync_offset = le64dec(data + 67); 298 md->md_mflags = le64dec(data + 75); 299 md->md_dflags = le64dec(data + 83); 300 bcopy(data + 91, md->md_provider, 16); 301 bcopy(data + 107, md->md_hash, 16); 302 MD5Init(&ctx); 303 MD5Update(&ctx, data, 107); 304 MD5Final(md->md_hash, &ctx); 305 if (bcmp(md->md_hash, data + 107, 16) != 0) 306 return (EINVAL); 307 308 /* New fields. */ 309 md->md_genid = 0; 310 md->md_provsize = 0; 311 312 return (0); 313 } 314 static __inline int 315 mirror_metadata_decode_v2(const u_char *data, struct g_mirror_metadata *md) 316 { 317 MD5_CTX ctx; 318 319 bcopy(data + 20, md->md_name, 16); 320 md->md_mid = le32dec(data + 36); 321 md->md_did = le32dec(data + 40); 322 md->md_all = *(data + 44); 323 md->md_genid = le32dec(data + 45); 324 md->md_syncid = le32dec(data + 49); 325 md->md_priority = *(data + 53); 326 md->md_slice = le32dec(data + 54); 327 md->md_balance = *(data + 58); 328 md->md_mediasize = le64dec(data + 59); 329 md->md_sectorsize = le32dec(data + 67); 330 md->md_sync_offset = le64dec(data + 71); 331 md->md_mflags = le64dec(data + 79); 332 md->md_dflags = le64dec(data + 87); 333 bcopy(data + 95, md->md_provider, 16); 334 bcopy(data + 111, md->md_hash, 16); 335 MD5Init(&ctx); 336 MD5Update(&ctx, data, 111); 337 MD5Final(md->md_hash, &ctx); 338 if (bcmp(md->md_hash, data + 111, 16) != 0) 339 return (EINVAL); 340 341 /* New fields. */ 342 md->md_provsize = 0; 343 344 return (0); 345 } 346 static __inline int 347 mirror_metadata_decode_v3v4(const u_char *data, struct g_mirror_metadata *md) 348 { 349 MD5_CTX ctx; 350 351 bcopy(data + 20, md->md_name, 16); 352 md->md_mid = le32dec(data + 36); 353 md->md_did = le32dec(data + 40); 354 md->md_all = *(data + 44); 355 md->md_genid = le32dec(data + 45); 356 md->md_syncid = le32dec(data + 49); 357 md->md_priority = *(data + 53); 358 md->md_slice = le32dec(data + 54); 359 md->md_balance = *(data + 58); 360 md->md_mediasize = le64dec(data + 59); 361 md->md_sectorsize = le32dec(data + 67); 362 md->md_sync_offset = le64dec(data + 71); 363 md->md_mflags = le64dec(data + 79); 364 md->md_dflags = le64dec(data + 87); 365 bcopy(data + 95, md->md_provider, 16); 366 md->md_provsize = le64dec(data + 111); 367 bcopy(data + 119, md->md_hash, 16); 368 MD5Init(&ctx); 369 MD5Update(&ctx, data, 119); 370 MD5Final(md->md_hash, &ctx); 371 if (bcmp(md->md_hash, data + 119, 16) != 0) 372 return (EINVAL); 373 return (0); 374 } 375 static __inline int 376 mirror_metadata_decode(const u_char *data, struct g_mirror_metadata *md) 377 { 378 int error; 379 380 bcopy(data, md->md_magic, 16); 381 md->md_version = le32dec(data + 16); 382 switch (md->md_version) { 383 case 0: 384 case 1: 385 error = mirror_metadata_decode_v0v1(data, md); 386 break; 387 case 2: 388 error = mirror_metadata_decode_v2(data, md); 389 break; 390 case 3: 391 case 4: 392 error = mirror_metadata_decode_v3v4(data, md); 393 break; 394 default: 395 error = EINVAL; 396 break; 397 } 398 return (error); 399 } 400 401 static __inline const char * 402 balance_name(u_int balance) 403 { 404 static const char *algorithms[] = { 405 [G_MIRROR_BALANCE_NONE] = "none", 406 [G_MIRROR_BALANCE_ROUND_ROBIN] = "round-robin", 407 [G_MIRROR_BALANCE_LOAD] = "load", 408 [G_MIRROR_BALANCE_SPLIT] = "split", 409 [G_MIRROR_BALANCE_PREFER] = "prefer", 410 [G_MIRROR_BALANCE_MAX + 1] = "unknown" 411 }; 412 413 if (balance > G_MIRROR_BALANCE_MAX) 414 balance = G_MIRROR_BALANCE_MAX + 1; 415 416 return (algorithms[balance]); 417 } 418 419 static __inline int 420 balance_id(const char *name) 421 { 422 static const char *algorithms[] = { 423 [G_MIRROR_BALANCE_NONE] = "none", 424 [G_MIRROR_BALANCE_ROUND_ROBIN] = "round-robin", 425 [G_MIRROR_BALANCE_LOAD] = "load", 426 [G_MIRROR_BALANCE_SPLIT] = "split", 427 [G_MIRROR_BALANCE_PREFER] = "prefer" 428 }; 429 int n; 430 431 for (n = G_MIRROR_BALANCE_MIN; n <= G_MIRROR_BALANCE_MAX; n++) { 432 if (strcmp(name, algorithms[n]) == 0) 433 return (n); 434 } 435 return (-1); 436 } 437 438 static __inline void 439 mirror_metadata_dump(const struct g_mirror_metadata *md) 440 { 441 static const char hex[] = "0123456789abcdef"; 442 char hash[16 * 2 + 1]; 443 u_int i; 444 445 printf(" magic: %s\n", md->md_magic); 446 printf(" version: %u\n", (u_int)md->md_version); 447 printf(" name: %s\n", md->md_name); 448 printf(" mid: %u\n", (u_int)md->md_mid); 449 printf(" did: %u\n", (u_int)md->md_did); 450 printf(" all: %u\n", (u_int)md->md_all); 451 printf(" genid: %u\n", (u_int)md->md_genid); 452 printf(" syncid: %u\n", (u_int)md->md_syncid); 453 printf(" priority: %u\n", (u_int)md->md_priority); 454 printf(" slice: %u\n", (u_int)md->md_slice); 455 printf(" balance: %s\n", balance_name((u_int)md->md_balance)); 456 printf(" mediasize: %jd\n", (intmax_t)md->md_mediasize); 457 printf("sectorsize: %u\n", (u_int)md->md_sectorsize); 458 printf("syncoffset: %jd\n", (intmax_t)md->md_sync_offset); 459 printf(" mflags:"); 460 if (md->md_mflags == 0) 461 printf(" NONE"); 462 else { 463 if ((md->md_mflags & G_MIRROR_DEVICE_FLAG_NOFAILSYNC) != 0) 464 printf(" NOFAILSYNC"); 465 if ((md->md_mflags & G_MIRROR_DEVICE_FLAG_NOAUTOSYNC) != 0) 466 printf(" NOAUTOSYNC"); 467 } 468 printf("\n"); 469 printf(" dflags:"); 470 if (md->md_dflags == 0) 471 printf(" NONE"); 472 else { 473 if ((md->md_dflags & G_MIRROR_DISK_FLAG_DIRTY) != 0) 474 printf(" DIRTY"); 475 if ((md->md_dflags & G_MIRROR_DISK_FLAG_SYNCHRONIZING) != 0) 476 printf(" SYNCHRONIZING"); 477 if ((md->md_dflags & G_MIRROR_DISK_FLAG_FORCE_SYNC) != 0) 478 printf(" FORCE_SYNC"); 479 if ((md->md_dflags & G_MIRROR_DISK_FLAG_INACTIVE) != 0) 480 printf(" INACTIVE"); 481 } 482 printf("\n"); 483 printf("hcprovider: %s\n", md->md_provider); 484 printf(" provsize: %ju\n", (uintmax_t)md->md_provsize); 485 bzero(hash, sizeof(hash)); 486 for (i = 0; i < 16; i++) { 487 hash[i * 2] = hex[md->md_hash[i] >> 4]; 488 hash[i * 2 + 1] = hex[md->md_hash[i] & 0x0f]; 489 } 490 printf(" MD5 hash: %s\n", hash); 491 } 492 #endif /* !_G_MIRROR_H_ */ 493