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_RAID3_H_ 30 #define _G_RAID3_H_ 31 32 #include <sys/endian.h> 33 #include <sys/md5.h> 34 35 #define G_RAID3_CLASS_NAME "RAID3" 36 37 #define G_RAID3_MAGIC "GEOM::RAID3" 38 /* 39 * Version history: 40 * 0 - Initial version number. 41 * 1 - Added 'round-robin reading' algorithm. 42 * 2 - Added 'verify reading' algorithm. 43 * 3 - Added md_genid field to metadata. 44 * 4 - Added md_provsize field to metadata. 45 */ 46 #define G_RAID3_VERSION 4 47 48 #define G_RAID3_DISK_FLAG_DIRTY 0x0000000000000001ULL 49 #define G_RAID3_DISK_FLAG_SYNCHRONIZING 0x0000000000000002ULL 50 #define G_RAID3_DISK_FLAG_FORCE_SYNC 0x0000000000000004ULL 51 #define G_RAID3_DISK_FLAG_HARDCODED 0x0000000000000008ULL 52 #define G_RAID3_DISK_FLAG_BROKEN 0x0000000000000010ULL 53 #define G_RAID3_DISK_FLAG_MASK (G_RAID3_DISK_FLAG_DIRTY | \ 54 G_RAID3_DISK_FLAG_SYNCHRONIZING | \ 55 G_RAID3_DISK_FLAG_FORCE_SYNC) 56 57 #define G_RAID3_DEVICE_FLAG_NOAUTOSYNC 0x0000000000000001ULL 58 #define G_RAID3_DEVICE_FLAG_ROUND_ROBIN 0x0000000000000002ULL 59 #define G_RAID3_DEVICE_FLAG_VERIFY 0x0000000000000004ULL 60 #define G_RAID3_DEVICE_FLAG_MASK (G_RAID3_DEVICE_FLAG_NOAUTOSYNC | \ 61 G_RAID3_DEVICE_FLAG_ROUND_ROBIN | \ 62 G_RAID3_DEVICE_FLAG_VERIFY) 63 64 #ifdef _KERNEL 65 extern u_int g_raid3_debug; 66 67 #define G_RAID3_DEBUG(lvl, ...) do { \ 68 if (g_raid3_debug >= (lvl)) { \ 69 printf("GEOM_RAID3"); \ 70 if (g_raid3_debug > 0) \ 71 printf("[%u]", lvl); \ 72 printf(": "); \ 73 printf(__VA_ARGS__); \ 74 printf("\n"); \ 75 } \ 76 } while (0) 77 #define G_RAID3_LOGREQ(lvl, bp, ...) do { \ 78 if (g_raid3_debug >= (lvl)) { \ 79 printf("GEOM_RAID3"); \ 80 if (g_raid3_debug > 0) \ 81 printf("[%u]", lvl); \ 82 printf(": "); \ 83 printf(__VA_ARGS__); \ 84 printf(" "); \ 85 g_print_bio(bp); \ 86 printf("\n"); \ 87 } \ 88 } while (0) 89 90 #define G_RAID3_BIO_CFLAG_REGULAR 0x01 91 #define G_RAID3_BIO_CFLAG_SYNC 0x02 92 #define G_RAID3_BIO_CFLAG_PARITY 0x04 93 #define G_RAID3_BIO_CFLAG_NODISK 0x08 94 #define G_RAID3_BIO_CFLAG_REGSYNC 0x10 95 #define G_RAID3_BIO_CFLAG_MASK (G_RAID3_BIO_CFLAG_REGULAR | \ 96 G_RAID3_BIO_CFLAG_SYNC | \ 97 G_RAID3_BIO_CFLAG_PARITY | \ 98 G_RAID3_BIO_CFLAG_NODISK | \ 99 G_RAID3_BIO_CFLAG_REGSYNC) 100 101 #define G_RAID3_BIO_PFLAG_DEGRADED 0x01 102 #define G_RAID3_BIO_PFLAG_NOPARITY 0x02 103 #define G_RAID3_BIO_PFLAG_VERIFY 0x04 104 #define G_RAID3_BIO_PFLAG_MASK (G_RAID3_BIO_PFLAG_DEGRADED | \ 105 G_RAID3_BIO_PFLAG_NOPARITY | \ 106 G_RAID3_BIO_PFLAG_VERIFY) 107 108 /* 109 * Informations needed for synchronization. 110 */ 111 struct g_raid3_disk_sync { 112 struct g_consumer *ds_consumer; /* Consumer connected to our device. */ 113 off_t ds_offset; /* Offset of next request to send. */ 114 off_t ds_offset_done; /* Offset of already synchronized 115 region. */ 116 off_t ds_resync; /* Resynchronize from this offset. */ 117 u_int ds_syncid; /* Disk's synchronization ID. */ 118 u_int ds_inflight; /* Number of in-flight sync requests. */ 119 struct bio **ds_bios; /* BIOs for synchronization I/O. */ 120 }; 121 122 /* 123 * Informations needed for synchronization. 124 */ 125 struct g_raid3_device_sync { 126 struct g_geom *ds_geom; /* Synchronization geom. */ 127 }; 128 129 #define G_RAID3_DISK_STATE_NODISK 0 130 #define G_RAID3_DISK_STATE_NONE 1 131 #define G_RAID3_DISK_STATE_NEW 2 132 #define G_RAID3_DISK_STATE_ACTIVE 3 133 #define G_RAID3_DISK_STATE_STALE 4 134 #define G_RAID3_DISK_STATE_SYNCHRONIZING 5 135 #define G_RAID3_DISK_STATE_DISCONNECTED 6 136 #define G_RAID3_DISK_STATE_DESTROY 7 137 struct g_raid3_disk { 138 u_int d_no; /* Disk number. */ 139 struct g_consumer *d_consumer; /* Consumer. */ 140 struct g_raid3_softc *d_softc; /* Back-pointer to softc. */ 141 int d_state; /* Disk state. */ 142 uint64_t d_flags; /* Additional flags. */ 143 u_int d_genid; /* Disk's generation ID. */ 144 struct g_raid3_disk_sync d_sync; /* Sync information. */ 145 LIST_ENTRY(g_raid3_disk) d_next; 146 }; 147 #define d_name d_consumer->provider->name 148 149 #define G_RAID3_EVENT_DONTWAIT 0x1 150 #define G_RAID3_EVENT_WAIT 0x2 151 #define G_RAID3_EVENT_DEVICE 0x4 152 #define G_RAID3_EVENT_DONE 0x8 153 struct g_raid3_event { 154 struct g_raid3_disk *e_disk; 155 int e_state; 156 int e_flags; 157 int e_error; 158 TAILQ_ENTRY(g_raid3_event) e_next; 159 }; 160 161 #define G_RAID3_DEVICE_FLAG_DESTROY 0x0100000000000000ULL 162 #define G_RAID3_DEVICE_FLAG_WAIT 0x0200000000000000ULL 163 #define G_RAID3_DEVICE_FLAG_DESTROYING 0x0400000000000000ULL 164 165 #define G_RAID3_DEVICE_STATE_STARTING 0 166 #define G_RAID3_DEVICE_STATE_DEGRADED 1 167 #define G_RAID3_DEVICE_STATE_COMPLETE 2 168 169 /* Bump syncid on first write. */ 170 #define G_RAID3_BUMP_SYNCID 0x1 171 /* Bump genid immediately. */ 172 #define G_RAID3_BUMP_GENID 0x2 173 174 enum g_raid3_zones { 175 G_RAID3_ZONE_64K, 176 G_RAID3_ZONE_16K, 177 G_RAID3_ZONE_4K, 178 G_RAID3_NUM_ZONES 179 }; 180 181 static __inline enum g_raid3_zones 182 g_raid3_zone(size_t nbytes) { 183 if (nbytes > 16384) 184 return (G_RAID3_ZONE_64K); 185 else if (nbytes > 4096) 186 return (G_RAID3_ZONE_16K); 187 else 188 return (G_RAID3_ZONE_4K); 189 }; 190 191 struct g_raid3_softc { 192 u_int sc_state; /* Device state. */ 193 uint64_t sc_mediasize; /* Device size. */ 194 uint32_t sc_sectorsize; /* Sector size. */ 195 uint64_t sc_flags; /* Additional flags. */ 196 197 struct g_geom *sc_geom; 198 struct g_provider *sc_provider; 199 200 uint32_t sc_id; /* Device unique ID. */ 201 202 struct sx sc_lock; 203 struct bio_queue_head sc_queue; 204 struct mtx sc_queue_mtx; 205 struct proc *sc_worker; 206 struct bio_queue_head sc_regular_delayed; /* Delayed I/O requests due 207 collision with sync 208 requests. */ 209 struct bio_queue_head sc_inflight; /* In-flight regular write 210 requests. */ 211 struct bio_queue_head sc_sync_delayed; /* Delayed sync requests due 212 collision with regular 213 requests. */ 214 215 struct g_raid3_disk *sc_disks; 216 u_int sc_ndisks; /* Number of disks. */ 217 u_int sc_round_robin; 218 struct g_raid3_disk *sc_syncdisk; 219 220 struct g_raid3_zone { 221 uma_zone_t sz_zone; 222 size_t sz_inuse; 223 size_t sz_max; 224 u_int sz_requested; 225 u_int sz_failed; 226 } sc_zones[G_RAID3_NUM_ZONES]; 227 228 u_int sc_genid; /* Generation ID. */ 229 u_int sc_syncid; /* Synchronization ID. */ 230 int sc_bump_id; 231 struct g_raid3_device_sync sc_sync; 232 int sc_idle; /* DIRTY flags removed. */ 233 time_t sc_last_write; 234 u_int sc_writes; 235 236 TAILQ_HEAD(, g_raid3_event) sc_events; 237 struct mtx sc_events_mtx; 238 239 struct callout sc_callout; 240 241 struct root_hold_token *sc_rootmount; 242 }; 243 #define sc_name sc_geom->name 244 245 const char *g_raid3_get_diskname(struct g_raid3_disk *disk); 246 u_int g_raid3_ndisks(struct g_raid3_softc *sc, int state); 247 #define G_RAID3_DESTROY_SOFT 0 248 #define G_RAID3_DESTROY_DELAYED 1 249 #define G_RAID3_DESTROY_HARD 2 250 int g_raid3_destroy(struct g_raid3_softc *sc, int how); 251 int g_raid3_event_send(void *arg, int state, int flags); 252 struct g_raid3_metadata; 253 int g_raid3_add_disk(struct g_raid3_softc *sc, struct g_provider *pp, 254 struct g_raid3_metadata *md); 255 int g_raid3_read_metadata(struct g_consumer *cp, struct g_raid3_metadata *md); 256 void g_raid3_fill_metadata(struct g_raid3_disk *disk, 257 struct g_raid3_metadata *md); 258 int g_raid3_clear_metadata(struct g_raid3_disk *disk); 259 void g_raid3_update_metadata(struct g_raid3_disk *disk); 260 261 g_ctl_req_t g_raid3_config; 262 #endif /* _KERNEL */ 263 264 struct g_raid3_metadata { 265 char md_magic[16]; /* Magic value. */ 266 uint32_t md_version; /* Version number. */ 267 char md_name[16]; /* Device name. */ 268 uint32_t md_id; /* Device unique ID. */ 269 uint16_t md_no; /* Component number. */ 270 uint16_t md_all; /* Number of disks in device. */ 271 uint32_t md_genid; /* Generation ID. */ 272 uint32_t md_syncid; /* Synchronization ID. */ 273 uint64_t md_mediasize; /* Size of whole device. */ 274 uint32_t md_sectorsize; /* Sector size. */ 275 uint64_t md_sync_offset; /* Synchronized offset. */ 276 uint64_t md_mflags; /* Additional device flags. */ 277 uint64_t md_dflags; /* Additional disk flags. */ 278 char md_provider[16]; /* Hardcoded provider. */ 279 uint64_t md_provsize; /* Provider's size. */ 280 u_char md_hash[16]; /* MD5 hash. */ 281 }; 282 static __inline void 283 raid3_metadata_encode(struct g_raid3_metadata *md, u_char *data) 284 { 285 MD5_CTX ctx; 286 287 bcopy(md->md_magic, data, 16); 288 le32enc(data + 16, md->md_version); 289 bcopy(md->md_name, data + 20, 16); 290 le32enc(data + 36, md->md_id); 291 le16enc(data + 40, md->md_no); 292 le16enc(data + 42, md->md_all); 293 le32enc(data + 44, md->md_genid); 294 le32enc(data + 48, md->md_syncid); 295 le64enc(data + 52, md->md_mediasize); 296 le32enc(data + 60, md->md_sectorsize); 297 le64enc(data + 64, md->md_sync_offset); 298 le64enc(data + 72, md->md_mflags); 299 le64enc(data + 80, md->md_dflags); 300 bcopy(md->md_provider, data + 88, 16); 301 le64enc(data + 104, md->md_provsize); 302 MD5Init(&ctx); 303 MD5Update(&ctx, data, 112); 304 MD5Final(md->md_hash, &ctx); 305 bcopy(md->md_hash, data + 112, 16); 306 } 307 static __inline int 308 raid3_metadata_decode_v0v1v2(const u_char *data, struct g_raid3_metadata *md) 309 { 310 MD5_CTX ctx; 311 312 bcopy(data + 20, md->md_name, 16); 313 md->md_id = le32dec(data + 36); 314 md->md_no = le16dec(data + 40); 315 md->md_all = le16dec(data + 42); 316 md->md_syncid = le32dec(data + 44); 317 md->md_mediasize = le64dec(data + 48); 318 md->md_sectorsize = le32dec(data + 56); 319 md->md_sync_offset = le64dec(data + 60); 320 md->md_mflags = le64dec(data + 68); 321 md->md_dflags = le64dec(data + 76); 322 bcopy(data + 84, md->md_provider, 16); 323 bcopy(data + 100, md->md_hash, 16); 324 MD5Init(&ctx); 325 MD5Update(&ctx, data, 100); 326 MD5Final(md->md_hash, &ctx); 327 if (bcmp(md->md_hash, data + 100, 16) != 0) 328 return (EINVAL); 329 330 /* New fields. */ 331 md->md_genid = 0; 332 md->md_provsize = 0; 333 334 return (0); 335 } 336 static __inline int 337 raid3_metadata_decode_v3(const u_char *data, struct g_raid3_metadata *md) 338 { 339 MD5_CTX ctx; 340 341 bcopy(data + 20, md->md_name, 16); 342 md->md_id = le32dec(data + 36); 343 md->md_no = le16dec(data + 40); 344 md->md_all = le16dec(data + 42); 345 md->md_genid = le32dec(data + 44); 346 md->md_syncid = le32dec(data + 48); 347 md->md_mediasize = le64dec(data + 52); 348 md->md_sectorsize = le32dec(data + 60); 349 md->md_sync_offset = le64dec(data + 64); 350 md->md_mflags = le64dec(data + 72); 351 md->md_dflags = le64dec(data + 80); 352 bcopy(data + 88, md->md_provider, 16); 353 bcopy(data + 104, md->md_hash, 16); 354 MD5Init(&ctx); 355 MD5Update(&ctx, data, 104); 356 MD5Final(md->md_hash, &ctx); 357 if (bcmp(md->md_hash, data + 104, 16) != 0) 358 return (EINVAL); 359 360 /* New fields. */ 361 md->md_provsize = 0; 362 363 return (0); 364 } 365 static __inline int 366 raid3_metadata_decode_v4(const u_char *data, struct g_raid3_metadata *md) 367 { 368 MD5_CTX ctx; 369 370 bcopy(data + 20, md->md_name, 16); 371 md->md_id = le32dec(data + 36); 372 md->md_no = le16dec(data + 40); 373 md->md_all = le16dec(data + 42); 374 md->md_genid = le32dec(data + 44); 375 md->md_syncid = le32dec(data + 48); 376 md->md_mediasize = le64dec(data + 52); 377 md->md_sectorsize = le32dec(data + 60); 378 md->md_sync_offset = le64dec(data + 64); 379 md->md_mflags = le64dec(data + 72); 380 md->md_dflags = le64dec(data + 80); 381 bcopy(data + 88, md->md_provider, 16); 382 md->md_provsize = le64dec(data + 104); 383 bcopy(data + 112, md->md_hash, 16); 384 MD5Init(&ctx); 385 MD5Update(&ctx, data, 112); 386 MD5Final(md->md_hash, &ctx); 387 if (bcmp(md->md_hash, data + 112, 16) != 0) 388 return (EINVAL); 389 return (0); 390 } 391 static __inline int 392 raid3_metadata_decode(const u_char *data, struct g_raid3_metadata *md) 393 { 394 int error; 395 396 bcopy(data, md->md_magic, 16); 397 md->md_version = le32dec(data + 16); 398 switch (md->md_version) { 399 case 0: 400 case 1: 401 case 2: 402 error = raid3_metadata_decode_v0v1v2(data, md); 403 break; 404 case 3: 405 error = raid3_metadata_decode_v3(data, md); 406 break; 407 case 4: 408 error = raid3_metadata_decode_v4(data, md); 409 break; 410 default: 411 error = EINVAL; 412 break; 413 } 414 return (error); 415 } 416 417 static __inline void 418 raid3_metadata_dump(const struct g_raid3_metadata *md) 419 { 420 static const char hex[] = "0123456789abcdef"; 421 char hash[16 * 2 + 1]; 422 u_int i; 423 424 printf(" magic: %s\n", md->md_magic); 425 printf(" version: %u\n", (u_int)md->md_version); 426 printf(" name: %s\n", md->md_name); 427 printf(" id: %u\n", (u_int)md->md_id); 428 printf(" no: %u\n", (u_int)md->md_no); 429 printf(" all: %u\n", (u_int)md->md_all); 430 printf(" genid: %u\n", (u_int)md->md_genid); 431 printf(" syncid: %u\n", (u_int)md->md_syncid); 432 printf(" mediasize: %jd\n", (intmax_t)md->md_mediasize); 433 printf("sectorsize: %u\n", (u_int)md->md_sectorsize); 434 printf("syncoffset: %jd\n", (intmax_t)md->md_sync_offset); 435 printf(" mflags:"); 436 if (md->md_mflags == 0) 437 printf(" NONE"); 438 else { 439 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_NOAUTOSYNC) != 0) 440 printf(" NOAUTOSYNC"); 441 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_ROUND_ROBIN) != 0) 442 printf(" ROUND-ROBIN"); 443 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_VERIFY) != 0) 444 printf(" VERIFY"); 445 } 446 printf("\n"); 447 printf(" dflags:"); 448 if (md->md_dflags == 0) 449 printf(" NONE"); 450 else { 451 if ((md->md_dflags & G_RAID3_DISK_FLAG_DIRTY) != 0) 452 printf(" DIRTY"); 453 if ((md->md_dflags & G_RAID3_DISK_FLAG_SYNCHRONIZING) != 0) 454 printf(" SYNCHRONIZING"); 455 if ((md->md_dflags & G_RAID3_DISK_FLAG_FORCE_SYNC) != 0) 456 printf(" FORCE_SYNC"); 457 } 458 printf("\n"); 459 printf("hcprovider: %s\n", md->md_provider); 460 printf(" provsize: %ju\n", (uintmax_t)md->md_provsize); 461 bzero(hash, sizeof(hash)); 462 for (i = 0; i < 16; i++) { 463 hash[i * 2] = hex[md->md_hash[i] >> 4]; 464 hash[i * 2 + 1] = hex[md->md_hash[i] & 0x0f]; 465 } 466 printf(" MD5 hash: %s\n", hash); 467 } 468 #endif /* !_G_RAID3_H_ */ 469