g_mirror.h revision 235599
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: head/sys/geom/mirror/g_mirror.h 235599 2012-05-18 09:19:07Z ae $ 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 73extern 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 */ 104struct 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 */ 117struct 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 130struct 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 u_int load; /* Averaged queue length */ 137 off_t d_last_offset; /* Last read offset */ 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 149struct 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#define G_MIRROR_DEVICE_FLAG_TASTING 0x0800000000000000ULL 161 162#define G_MIRROR_DEVICE_STATE_STARTING 0 163#define G_MIRROR_DEVICE_STATE_RUNNING 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 169struct g_mirror_softc { 170 u_int sc_state; /* Device state. */ 171 uint32_t sc_slice; /* Slice size. */ 172 uint8_t sc_balance; /* Balance algorithm. */ 173 uint64_t sc_mediasize; /* Device size. */ 174 uint32_t sc_sectorsize; /* Sector size. */ 175 uint64_t sc_flags; /* Additional flags. */ 176 177 struct g_geom *sc_geom; 178 struct g_provider *sc_provider; 179 180 uint32_t sc_id; /* Mirror unique ID. */ 181 182 struct sx sc_lock; 183 struct bio_queue_head sc_queue; 184 struct mtx sc_queue_mtx; 185 struct proc *sc_worker; 186 struct bio_queue_head sc_regular_delayed; /* Delayed I/O requests due 187 collision with sync 188 requests. */ 189 struct bio_queue_head sc_inflight; /* In-flight regular write 190 requests. */ 191 struct bio_queue_head sc_sync_delayed; /* Delayed sync requests due 192 collision with regular 193 requests. */ 194 195 LIST_HEAD(, g_mirror_disk) sc_disks; 196 u_int sc_ndisks; /* Number of disks. */ 197 struct g_mirror_disk *sc_hint; 198 199 u_int sc_genid; /* Generation ID. */ 200 u_int sc_syncid; /* Synchronization ID. */ 201 int sc_bump_id; 202 struct g_mirror_device_sync sc_sync; 203 int sc_idle; /* DIRTY flags removed. */ 204 time_t sc_last_write; 205 u_int sc_writes; 206 207 TAILQ_HEAD(, g_mirror_event) sc_events; 208 struct mtx sc_events_mtx; 209 210 struct callout sc_callout; 211 212 struct root_hold_token *sc_rootmount; 213}; 214#define sc_name sc_geom->name 215 216u_int g_mirror_ndisks(struct g_mirror_softc *sc, int state); 217#define G_MIRROR_DESTROY_SOFT 0 218#define G_MIRROR_DESTROY_DELAYED 1 219#define G_MIRROR_DESTROY_HARD 2 220int g_mirror_destroy(struct g_mirror_softc *sc, int how); 221int g_mirror_event_send(void *arg, int state, int flags); 222struct g_mirror_metadata; 223int g_mirror_add_disk(struct g_mirror_softc *sc, struct g_provider *pp, 224 struct g_mirror_metadata *md); 225int g_mirror_read_metadata(struct g_consumer *cp, struct g_mirror_metadata *md); 226void g_mirror_fill_metadata(struct g_mirror_softc *sc, 227 struct g_mirror_disk *disk, struct g_mirror_metadata *md); 228void g_mirror_update_metadata(struct g_mirror_disk *disk); 229 230g_ctl_req_t g_mirror_config; 231#endif /* _KERNEL */ 232 233struct g_mirror_metadata { 234 char md_magic[16]; /* Magic value. */ 235 uint32_t md_version; /* Version number. */ 236 char md_name[16]; /* Mirror name. */ 237 uint32_t md_mid; /* Mirror unique ID. */ 238 uint32_t md_did; /* Disk unique ID. */ 239 uint8_t md_all; /* Number of disks in mirror. */ 240 uint32_t md_genid; /* Generation ID. */ 241 uint32_t md_syncid; /* Synchronization ID. */ 242 uint8_t md_priority; /* Disk priority. */ 243 uint32_t md_slice; /* Slice size. */ 244 uint8_t md_balance; /* Balance type. */ 245 uint64_t md_mediasize; /* Size of the smallest 246 disk in mirror. */ 247 uint32_t md_sectorsize; /* Sector size. */ 248 uint64_t md_sync_offset; /* Synchronized offset. */ 249 uint64_t md_mflags; /* Additional mirror flags. */ 250 uint64_t md_dflags; /* Additional disk flags. */ 251 char md_provider[16]; /* Hardcoded provider. */ 252 uint64_t md_provsize; /* Provider's size. */ 253 u_char md_hash[16]; /* MD5 hash. */ 254}; 255static __inline void 256mirror_metadata_encode(struct g_mirror_metadata *md, u_char *data) 257{ 258 MD5_CTX ctx; 259 260 bcopy(md->md_magic, data, 16); 261 le32enc(data + 16, md->md_version); 262 bcopy(md->md_name, data + 20, 16); 263 le32enc(data + 36, md->md_mid); 264 le32enc(data + 40, md->md_did); 265 *(data + 44) = md->md_all; 266 le32enc(data + 45, md->md_genid); 267 le32enc(data + 49, md->md_syncid); 268 *(data + 53) = md->md_priority; 269 le32enc(data + 54, md->md_slice); 270 *(data + 58) = md->md_balance; 271 le64enc(data + 59, md->md_mediasize); 272 le32enc(data + 67, md->md_sectorsize); 273 le64enc(data + 71, md->md_sync_offset); 274 le64enc(data + 79, md->md_mflags); 275 le64enc(data + 87, md->md_dflags); 276 bcopy(md->md_provider, data + 95, 16); 277 le64enc(data + 111, md->md_provsize); 278 MD5Init(&ctx); 279 MD5Update(&ctx, data, 119); 280 MD5Final(md->md_hash, &ctx); 281 bcopy(md->md_hash, data + 119, 16); 282} 283static __inline int 284mirror_metadata_decode_v0v1(const u_char *data, struct g_mirror_metadata *md) 285{ 286 MD5_CTX ctx; 287 288 bcopy(data + 20, md->md_name, 16); 289 md->md_mid = le32dec(data + 36); 290 md->md_did = le32dec(data + 40); 291 md->md_all = *(data + 44); 292 md->md_syncid = le32dec(data + 45); 293 md->md_priority = *(data + 49); 294 md->md_slice = le32dec(data + 50); 295 md->md_balance = *(data + 54); 296 md->md_mediasize = le64dec(data + 55); 297 md->md_sectorsize = le32dec(data + 63); 298 md->md_sync_offset = le64dec(data + 67); 299 md->md_mflags = le64dec(data + 75); 300 md->md_dflags = le64dec(data + 83); 301 bcopy(data + 91, md->md_provider, 16); 302 bcopy(data + 107, md->md_hash, 16); 303 MD5Init(&ctx); 304 MD5Update(&ctx, data, 107); 305 MD5Final(md->md_hash, &ctx); 306 if (bcmp(md->md_hash, data + 107, 16) != 0) 307 return (EINVAL); 308 309 /* New fields. */ 310 md->md_genid = 0; 311 md->md_provsize = 0; 312 313 return (0); 314} 315static __inline int 316mirror_metadata_decode_v2(const u_char *data, struct g_mirror_metadata *md) 317{ 318 MD5_CTX ctx; 319 320 bcopy(data + 20, md->md_name, 16); 321 md->md_mid = le32dec(data + 36); 322 md->md_did = le32dec(data + 40); 323 md->md_all = *(data + 44); 324 md->md_genid = le32dec(data + 45); 325 md->md_syncid = le32dec(data + 49); 326 md->md_priority = *(data + 53); 327 md->md_slice = le32dec(data + 54); 328 md->md_balance = *(data + 58); 329 md->md_mediasize = le64dec(data + 59); 330 md->md_sectorsize = le32dec(data + 67); 331 md->md_sync_offset = le64dec(data + 71); 332 md->md_mflags = le64dec(data + 79); 333 md->md_dflags = le64dec(data + 87); 334 bcopy(data + 95, md->md_provider, 16); 335 bcopy(data + 111, md->md_hash, 16); 336 MD5Init(&ctx); 337 MD5Update(&ctx, data, 111); 338 MD5Final(md->md_hash, &ctx); 339 if (bcmp(md->md_hash, data + 111, 16) != 0) 340 return (EINVAL); 341 342 /* New fields. */ 343 md->md_provsize = 0; 344 345 return (0); 346} 347static __inline int 348mirror_metadata_decode_v3v4(const u_char *data, struct g_mirror_metadata *md) 349{ 350 MD5_CTX ctx; 351 352 bcopy(data + 20, md->md_name, 16); 353 md->md_mid = le32dec(data + 36); 354 md->md_did = le32dec(data + 40); 355 md->md_all = *(data + 44); 356 md->md_genid = le32dec(data + 45); 357 md->md_syncid = le32dec(data + 49); 358 md->md_priority = *(data + 53); 359 md->md_slice = le32dec(data + 54); 360 md->md_balance = *(data + 58); 361 md->md_mediasize = le64dec(data + 59); 362 md->md_sectorsize = le32dec(data + 67); 363 md->md_sync_offset = le64dec(data + 71); 364 md->md_mflags = le64dec(data + 79); 365 md->md_dflags = le64dec(data + 87); 366 bcopy(data + 95, md->md_provider, 16); 367 md->md_provsize = le64dec(data + 111); 368 bcopy(data + 119, md->md_hash, 16); 369 MD5Init(&ctx); 370 MD5Update(&ctx, data, 119); 371 MD5Final(md->md_hash, &ctx); 372 if (bcmp(md->md_hash, data + 119, 16) != 0) 373 return (EINVAL); 374 return (0); 375} 376static __inline int 377mirror_metadata_decode(const u_char *data, struct g_mirror_metadata *md) 378{ 379 int error; 380 381 bcopy(data, md->md_magic, 16); 382 md->md_version = le32dec(data + 16); 383 switch (md->md_version) { 384 case 0: 385 case 1: 386 error = mirror_metadata_decode_v0v1(data, md); 387 break; 388 case 2: 389 error = mirror_metadata_decode_v2(data, md); 390 break; 391 case 3: 392 case 4: 393 error = mirror_metadata_decode_v3v4(data, md); 394 break; 395 default: 396 error = EINVAL; 397 break; 398 } 399 return (error); 400} 401 402static __inline const char * 403balance_name(u_int balance) 404{ 405 static const char *algorithms[] = { 406 [G_MIRROR_BALANCE_NONE] = "none", 407 [G_MIRROR_BALANCE_ROUND_ROBIN] = "round-robin", 408 [G_MIRROR_BALANCE_LOAD] = "load", 409 [G_MIRROR_BALANCE_SPLIT] = "split", 410 [G_MIRROR_BALANCE_PREFER] = "prefer", 411 [G_MIRROR_BALANCE_MAX + 1] = "unknown" 412 }; 413 414 if (balance > G_MIRROR_BALANCE_MAX) 415 balance = G_MIRROR_BALANCE_MAX + 1; 416 417 return (algorithms[balance]); 418} 419 420static __inline int 421balance_id(const char *name) 422{ 423 static const char *algorithms[] = { 424 [G_MIRROR_BALANCE_NONE] = "none", 425 [G_MIRROR_BALANCE_ROUND_ROBIN] = "round-robin", 426 [G_MIRROR_BALANCE_LOAD] = "load", 427 [G_MIRROR_BALANCE_SPLIT] = "split", 428 [G_MIRROR_BALANCE_PREFER] = "prefer" 429 }; 430 int n; 431 432 for (n = G_MIRROR_BALANCE_MIN; n <= G_MIRROR_BALANCE_MAX; n++) { 433 if (strcmp(name, algorithms[n]) == 0) 434 return (n); 435 } 436 return (-1); 437} 438 439static __inline void 440mirror_metadata_dump(const struct g_mirror_metadata *md) 441{ 442 static const char hex[] = "0123456789abcdef"; 443 char hash[16 * 2 + 1]; 444 u_int i; 445 446 printf(" magic: %s\n", md->md_magic); 447 printf(" version: %u\n", (u_int)md->md_version); 448 printf(" name: %s\n", md->md_name); 449 printf(" mid: %u\n", (u_int)md->md_mid); 450 printf(" did: %u\n", (u_int)md->md_did); 451 printf(" all: %u\n", (u_int)md->md_all); 452 printf(" genid: %u\n", (u_int)md->md_genid); 453 printf(" syncid: %u\n", (u_int)md->md_syncid); 454 printf(" priority: %u\n", (u_int)md->md_priority); 455 printf(" slice: %u\n", (u_int)md->md_slice); 456 printf(" balance: %s\n", balance_name((u_int)md->md_balance)); 457 printf(" mediasize: %jd\n", (intmax_t)md->md_mediasize); 458 printf("sectorsize: %u\n", (u_int)md->md_sectorsize); 459 printf("syncoffset: %jd\n", (intmax_t)md->md_sync_offset); 460 printf(" mflags:"); 461 if (md->md_mflags == 0) 462 printf(" NONE"); 463 else { 464 if ((md->md_mflags & G_MIRROR_DEVICE_FLAG_NOFAILSYNC) != 0) 465 printf(" NOFAILSYNC"); 466 if ((md->md_mflags & G_MIRROR_DEVICE_FLAG_NOAUTOSYNC) != 0) 467 printf(" NOAUTOSYNC"); 468 } 469 printf("\n"); 470 printf(" dflags:"); 471 if (md->md_dflags == 0) 472 printf(" NONE"); 473 else { 474 if ((md->md_dflags & G_MIRROR_DISK_FLAG_DIRTY) != 0) 475 printf(" DIRTY"); 476 if ((md->md_dflags & G_MIRROR_DISK_FLAG_SYNCHRONIZING) != 0) 477 printf(" SYNCHRONIZING"); 478 if ((md->md_dflags & G_MIRROR_DISK_FLAG_FORCE_SYNC) != 0) 479 printf(" FORCE_SYNC"); 480 if ((md->md_dflags & G_MIRROR_DISK_FLAG_INACTIVE) != 0) 481 printf(" INACTIVE"); 482 } 483 printf("\n"); 484 printf("hcprovider: %s\n", md->md_provider); 485 printf(" provsize: %ju\n", (uintmax_t)md->md_provsize); 486 bzero(hash, sizeof(hash)); 487 for (i = 0; i < 16; i++) { 488 hash[i * 2] = hex[md->md_hash[i] >> 4]; 489 hash[i * 2 + 1] = hex[md->md_hash[i] & 0x0f]; 490 } 491 printf(" MD5 hash: %s\n", hash); 492} 493#endif /* !_G_MIRROR_H_ */ 494