1 /*- 2 * Copyright (c) 2004-2005 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 */ 45 #define G_MIRROR_VERSION 3 46 47 #define G_MIRROR_BALANCE_NONE 0 48 #define G_MIRROR_BALANCE_ROUND_ROBIN 1 49 #define G_MIRROR_BALANCE_LOAD 2 50 #define G_MIRROR_BALANCE_SPLIT 3 51 #define G_MIRROR_BALANCE_PREFER 4 52 #define G_MIRROR_BALANCE_MIN G_MIRROR_BALANCE_NONE 53 #define G_MIRROR_BALANCE_MAX G_MIRROR_BALANCE_PREFER 54 55 #define G_MIRROR_DISK_FLAG_DIRTY 0x0000000000000001ULL 56 #define G_MIRROR_DISK_FLAG_SYNCHRONIZING 0x0000000000000002ULL 57 #define G_MIRROR_DISK_FLAG_FORCE_SYNC 0x0000000000000004ULL 58 #define G_MIRROR_DISK_FLAG_INACTIVE 0x0000000000000008ULL 59 #define G_MIRROR_DISK_FLAG_HARDCODED 0x0000000000000010ULL 60 #define G_MIRROR_DISK_FLAG_MASK (G_MIRROR_DISK_FLAG_DIRTY | \ 61 G_MIRROR_DISK_FLAG_SYNCHRONIZING | \ 62 G_MIRROR_DISK_FLAG_FORCE_SYNC | \ 63 G_MIRROR_DISK_FLAG_INACTIVE) 64 65 #define G_MIRROR_DEVICE_FLAG_NOAUTOSYNC 0x0000000000000001ULL 66 #define G_MIRROR_DEVICE_FLAG_MASK (G_MIRROR_DEVICE_FLAG_NOAUTOSYNC) 67 68 #ifdef _KERNEL 69 extern u_int g_mirror_debug; 70 71 #define G_MIRROR_DEBUG(lvl, ...) do { \ 72 if (g_mirror_debug >= (lvl)) { \ 73 printf("GEOM_MIRROR"); \ 74 if (g_mirror_debug > 0) \ 75 printf("[%u]", lvl); \ 76 printf(": "); \ 77 printf(__VA_ARGS__); \ 78 printf("\n"); \ 79 } \ 80 } while (0) 81 #define G_MIRROR_LOGREQ(lvl, bp, ...) do { \ 82 if (g_mirror_debug >= (lvl)) { \ 83 printf("GEOM_MIRROR"); \ 84 if (g_mirror_debug > 0) \ 85 printf("[%u]", lvl); \ 86 printf(": "); \ 87 printf(__VA_ARGS__); \ 88 printf(" "); \ 89 g_print_bio(bp); \ 90 printf("\n"); \ 91 } \ 92 } while (0) 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 off_t ds_resync; /* Resynchronize from this offset. */ 106 u_int ds_syncid; /* Disk's synchronization ID. */ 107 u_char *ds_data; 108 }; 109 110 /* 111 * Informations needed for synchronization. 112 */ 113 struct g_mirror_device_sync { 114 struct g_geom *ds_geom; /* Synchronization geom. */ 115 u_int ds_ndisks; /* Number of disks in SYNCHRONIZING 116 state. */ 117 }; 118 119 #define G_MIRROR_DISK_STATE_NONE 0 120 #define G_MIRROR_DISK_STATE_NEW 1 121 #define G_MIRROR_DISK_STATE_ACTIVE 2 122 #define G_MIRROR_DISK_STATE_STALE 3 123 #define G_MIRROR_DISK_STATE_SYNCHRONIZING 4 124 #define G_MIRROR_DISK_STATE_DISCONNECTED 5 125 #define G_MIRROR_DISK_STATE_DESTROY 6 126 struct g_mirror_disk { 127 uint32_t d_id; /* Disk ID. */ 128 struct g_consumer *d_consumer; /* Consumer. */ 129 struct g_mirror_softc *d_softc; /* Back-pointer to softc. */ 130 int d_state; /* Disk state. */ 131 u_int d_priority; /* Disk priority. */ 132 struct bintime d_delay; /* Disk delay. */ 133 struct bintime d_last_used; /* When disk was last used. */ 134 uint64_t d_flags; /* Additional flags. */ 135 u_int d_genid; /* Disk's generation ID. */ 136 struct g_mirror_disk_sync d_sync;/* Sync information. */ 137 LIST_ENTRY(g_mirror_disk) d_next; 138 }; 139 #define d_name d_consumer->provider->name 140 141 #define G_MIRROR_EVENT_DONTWAIT 0x1 142 #define G_MIRROR_EVENT_WAIT 0x2 143 #define G_MIRROR_EVENT_DEVICE 0x4 144 #define G_MIRROR_EVENT_DONE 0x8 145 struct g_mirror_event { 146 struct g_mirror_disk *e_disk; 147 int e_state; 148 int e_flags; 149 int e_error; 150 TAILQ_ENTRY(g_mirror_event) e_next; 151 }; 152 153 #define G_MIRROR_DEVICE_FLAG_DESTROY 0x0100000000000000ULL 154 #define G_MIRROR_DEVICE_FLAG_WAIT 0x0200000000000000ULL 155 156 #define G_MIRROR_DEVICE_STATE_STARTING 0 157 #define G_MIRROR_DEVICE_STATE_RUNNING 1 158 159 /* Bump syncid on first write. */ 160 #define G_MIRROR_BUMP_SYNCID 0x1 161 /* Bump genid immediately. */ 162 #define G_MIRROR_BUMP_GENID 0x2 163 struct g_mirror_softc { 164 u_int sc_state; /* Device state. */ 165 uint32_t sc_slice; /* Slice size. */ 166 uint8_t sc_balance; /* Balance algorithm. */ 167 uint64_t sc_mediasize; /* Device size. */ 168 uint32_t sc_sectorsize; /* Sector size. */ 169 uint64_t sc_flags; /* Additional flags. */ 170 171 struct g_geom *sc_geom; 172 struct g_provider *sc_provider; 173 174 uint32_t sc_id; /* Mirror unique ID. */ 175 176 struct bio_queue_head sc_queue; 177 struct mtx sc_queue_mtx; 178 struct proc *sc_worker; 179 180 LIST_HEAD(, g_mirror_disk) sc_disks; 181 u_int sc_ndisks; /* Number of disks. */ 182 struct g_mirror_disk *sc_hint; 183 184 u_int sc_genid; /* Generation ID. */ 185 u_int sc_syncid; /* Synchronization ID. */ 186 int sc_bump_id; 187 struct g_mirror_device_sync sc_sync; 188 int sc_idle; /* DIRTY flags removed. */ 189 190 TAILQ_HEAD(, g_mirror_event) sc_events; 191 struct mtx sc_events_mtx; 192 193 struct callout sc_callout; 194 195 struct root_hold_token *sc_rootmount; 196 }; 197 #define sc_name sc_geom->name 198 199 u_int g_mirror_ndisks(struct g_mirror_softc *sc, int state); 200 int g_mirror_destroy(struct g_mirror_softc *sc, boolean_t force); 201 int g_mirror_event_send(void *arg, int state, int flags); 202 struct g_mirror_metadata; 203 int g_mirror_add_disk(struct g_mirror_softc *sc, struct g_provider *pp, 204 struct g_mirror_metadata *md); 205 int g_mirror_read_metadata(struct g_consumer *cp, struct g_mirror_metadata *md); 206 void g_mirror_fill_metadata(struct g_mirror_softc *sc, 207 struct g_mirror_disk *disk, struct g_mirror_metadata *md); 208 void g_mirror_update_metadata(struct g_mirror_disk *disk); 209 210 g_ctl_req_t g_mirror_config; 211 #endif /* _KERNEL */ 212 213 struct g_mirror_metadata { 214 char md_magic[16]; /* Magic value. */ 215 uint32_t md_version; /* Version number. */ 216 char md_name[16]; /* Mirror name. */ 217 uint32_t md_mid; /* Mirror unique ID. */ 218 uint32_t md_did; /* Disk unique ID. */ 219 uint8_t md_all; /* Number of disks in mirror. */ 220 uint32_t md_genid; /* Generation ID. */ 221 uint32_t md_syncid; /* Synchronization ID. */ 222 uint8_t md_priority; /* Disk priority. */ 223 uint32_t md_slice; /* Slice size. */ 224 uint8_t md_balance; /* Balance type. */ 225 uint64_t md_mediasize; /* Size of the smallest 226 disk in mirror. */ 227 uint32_t md_sectorsize; /* Sector size. */ 228 uint64_t md_sync_offset; /* Synchronized offset. */ 229 uint64_t md_mflags; /* Additional mirror flags. */ 230 uint64_t md_dflags; /* Additional disk flags. */ 231 char md_provider[16]; /* Hardcoded provider. */ 232 uint64_t md_provsize; /* Provider's size. */ 233 u_char md_hash[16]; /* MD5 hash. */ 234 }; 235 static __inline void 236 mirror_metadata_encode(struct g_mirror_metadata *md, u_char *data) 237 { 238 MD5_CTX ctx; 239 240 bcopy(md->md_magic, data, 16); 241 le32enc(data + 16, md->md_version); 242 bcopy(md->md_name, data + 20, 16); 243 le32enc(data + 36, md->md_mid); 244 le32enc(data + 40, md->md_did); 245 *(data + 44) = md->md_all; 246 le32enc(data + 45, md->md_genid); 247 le32enc(data + 49, md->md_syncid); 248 *(data + 53) = md->md_priority; 249 le32enc(data + 54, md->md_slice); 250 *(data + 58) = md->md_balance; 251 le64enc(data + 59, md->md_mediasize); 252 le32enc(data + 67, md->md_sectorsize); 253 le64enc(data + 71, md->md_sync_offset); 254 le64enc(data + 79, md->md_mflags); 255 le64enc(data + 87, md->md_dflags); 256 bcopy(md->md_provider, data + 95, 16); 257 le64enc(data + 111, md->md_provsize); 258 MD5Init(&ctx); 259 MD5Update(&ctx, data, 119); 260 MD5Final(md->md_hash, &ctx); 261 bcopy(md->md_hash, data + 119, 16); 262 } 263 static __inline int 264 mirror_metadata_decode_v0v1(const u_char *data, struct g_mirror_metadata *md) 265 { 266 MD5_CTX ctx; 267 268 bcopy(data + 20, md->md_name, 16); 269 md->md_mid = le32dec(data + 36); 270 md->md_did = le32dec(data + 40); 271 md->md_all = *(data + 44); 272 md->md_syncid = le32dec(data + 45); 273 md->md_priority = *(data + 49); 274 md->md_slice = le32dec(data + 50); 275 md->md_balance = *(data + 54); 276 md->md_mediasize = le64dec(data + 55); 277 md->md_sectorsize = le32dec(data + 63); 278 md->md_sync_offset = le64dec(data + 67); 279 md->md_mflags = le64dec(data + 75); 280 md->md_dflags = le64dec(data + 83); 281 bcopy(data + 91, md->md_provider, 16); 282 bcopy(data + 107, md->md_hash, 16); 283 MD5Init(&ctx); 284 MD5Update(&ctx, data, 107); 285 MD5Final(md->md_hash, &ctx); 286 if (bcmp(md->md_hash, data + 107, 16) != 0) 287 return (EINVAL); 288 289 /* New fields. */ 290 md->md_genid = 0; 291 md->md_provsize = 0; 292 293 return (0); 294 } 295 static __inline int 296 mirror_metadata_decode_v2(const u_char *data, struct g_mirror_metadata *md) 297 { 298 MD5_CTX ctx; 299 300 bcopy(data + 20, md->md_name, 16); 301 md->md_mid = le32dec(data + 36); 302 md->md_did = le32dec(data + 40); 303 md->md_all = *(data + 44); 304 md->md_genid = le32dec(data + 45); 305 md->md_syncid = le32dec(data + 49); 306 md->md_priority = *(data + 53); 307 md->md_slice = le32dec(data + 54); 308 md->md_balance = *(data + 58); 309 md->md_mediasize = le64dec(data + 59); 310 md->md_sectorsize = le32dec(data + 67); 311 md->md_sync_offset = le64dec(data + 71); 312 md->md_mflags = le64dec(data + 79); 313 md->md_dflags = le64dec(data + 87); 314 bcopy(data + 95, md->md_provider, 16); 315 bcopy(data + 111, md->md_hash, 16); 316 MD5Init(&ctx); 317 MD5Update(&ctx, data, 111); 318 MD5Final(md->md_hash, &ctx); 319 if (bcmp(md->md_hash, data + 111, 16) != 0) 320 return (EINVAL); 321 322 /* New fields. */ 323 md->md_provsize = 0; 324 325 return (0); 326 } 327 static __inline int 328 mirror_metadata_decode_v3(const u_char *data, struct g_mirror_metadata *md) 329 { 330 MD5_CTX ctx; 331 332 bcopy(data + 20, md->md_name, 16); 333 md->md_mid = le32dec(data + 36); 334 md->md_did = le32dec(data + 40); 335 md->md_all = *(data + 44); 336 md->md_genid = le32dec(data + 45); 337 md->md_syncid = le32dec(data + 49); 338 md->md_priority = *(data + 53); 339 md->md_slice = le32dec(data + 54); 340 md->md_balance = *(data + 58); 341 md->md_mediasize = le64dec(data + 59); 342 md->md_sectorsize = le32dec(data + 67); 343 md->md_sync_offset = le64dec(data + 71); 344 md->md_mflags = le64dec(data + 79); 345 md->md_dflags = le64dec(data + 87); 346 bcopy(data + 95, md->md_provider, 16); 347 md->md_provsize = le64dec(data + 111); 348 bcopy(data + 119, md->md_hash, 16); 349 MD5Init(&ctx); 350 MD5Update(&ctx, data, 119); 351 MD5Final(md->md_hash, &ctx); 352 if (bcmp(md->md_hash, data + 119, 16) != 0) 353 return (EINVAL); 354 return (0); 355 } 356 static __inline int 357 mirror_metadata_decode(const u_char *data, struct g_mirror_metadata *md) 358 { 359 int error; 360 361 bcopy(data, md->md_magic, 16); 362 md->md_version = le32dec(data + 16); 363 switch (md->md_version) { 364 case 0: 365 case 1: 366 error = mirror_metadata_decode_v0v1(data, md); 367 break; 368 case 2: 369 error = mirror_metadata_decode_v2(data, md); 370 break; 371 case 3: 372 error = mirror_metadata_decode_v3(data, md); 373 break; 374 default: 375 error = EINVAL; 376 break; 377 } 378 return (error); 379 } 380 381 static __inline const char * 382 balance_name(u_int balance) 383 { 384 static const char *algorithms[] = { 385 [G_MIRROR_BALANCE_NONE] = "none", 386 [G_MIRROR_BALANCE_ROUND_ROBIN] = "round-robin", 387 [G_MIRROR_BALANCE_LOAD] = "load", 388 [G_MIRROR_BALANCE_SPLIT] = "split", 389 [G_MIRROR_BALANCE_PREFER] = "prefer", 390 [G_MIRROR_BALANCE_MAX + 1] = "unknown" 391 }; 392 393 if (balance > G_MIRROR_BALANCE_MAX) 394 balance = G_MIRROR_BALANCE_MAX + 1; 395 396 return (algorithms[balance]); 397 } 398 399 static __inline int 400 balance_id(const char *name) 401 { 402 static const char *algorithms[] = { 403 [G_MIRROR_BALANCE_NONE] = "none", 404 [G_MIRROR_BALANCE_ROUND_ROBIN] = "round-robin", 405 [G_MIRROR_BALANCE_LOAD] = "load", 406 [G_MIRROR_BALANCE_SPLIT] = "split", 407 [G_MIRROR_BALANCE_PREFER] = "prefer" 408 }; 409 int n; 410 411 for (n = G_MIRROR_BALANCE_MIN; n <= G_MIRROR_BALANCE_MAX; n++) { 412 if (strcmp(name, algorithms[n]) == 0) 413 return (n); 414 } 415 return (-1); 416 } 417 418 static __inline void 419 mirror_metadata_dump(const struct g_mirror_metadata *md) 420 { 421 static const char hex[] = "0123456789abcdef"; 422 char hash[16 * 2 + 1]; 423 u_int i; 424 425 printf(" magic: %s\n", md->md_magic); 426 printf(" version: %u\n", (u_int)md->md_version); 427 printf(" name: %s\n", md->md_name); 428 printf(" mid: %u\n", (u_int)md->md_mid); 429 printf(" did: %u\n", (u_int)md->md_did); 430 printf(" all: %u\n", (u_int)md->md_all); 431 printf(" genid: %u\n", (u_int)md->md_genid); 432 printf(" syncid: %u\n", (u_int)md->md_syncid); 433 printf(" priority: %u\n", (u_int)md->md_priority); 434 printf(" slice: %u\n", (u_int)md->md_slice); 435 printf(" balance: %s\n", balance_name((u_int)md->md_balance)); 436 printf(" mediasize: %jd\n", (intmax_t)md->md_mediasize); 437 printf("sectorsize: %u\n", (u_int)md->md_sectorsize); 438 printf("syncoffset: %jd\n", (intmax_t)md->md_sync_offset); 439 printf(" mflags:"); 440 if (md->md_mflags == 0) 441 printf(" NONE"); 442 else { 443 if ((md->md_mflags & G_MIRROR_DEVICE_FLAG_NOAUTOSYNC) != 0) 444 printf(" NOAUTOSYNC"); 445 } 446 printf("\n"); 447 printf(" dflags:"); 448 if (md->md_dflags == 0) 449 printf(" NONE"); 450 else { 451 if ((md->md_dflags & G_MIRROR_DISK_FLAG_DIRTY) != 0) 452 printf(" DIRTY"); 453 if ((md->md_dflags & G_MIRROR_DISK_FLAG_SYNCHRONIZING) != 0) 454 printf(" SYNCHRONIZING"); 455 if ((md->md_dflags & G_MIRROR_DISK_FLAG_FORCE_SYNC) != 0) 456 printf(" FORCE_SYNC"); 457 if ((md->md_dflags & G_MIRROR_DISK_FLAG_INACTIVE) != 0) 458 printf(" INACTIVE"); 459 } 460 printf("\n"); 461 printf("hcprovider: %s\n", md->md_provider); 462 printf(" provsize: %ju\n", (uintmax_t)md->md_provsize); 463 bzero(hash, sizeof(hash)); 464 for (i = 0; i < 16; i++) { 465 hash[i * 2] = hex[md->md_hash[i] >> 4]; 466 hash[i * 2 + 1] = hex[md->md_hash[i] & 0x0f]; 467 } 468 printf(" MD5 hash: %s\n", hash); 469 } 470 #endif /* !_G_MIRROR_H_ */ 471