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 * 5 - Added 'no failure synchronization' flag. 46 */ 47#define G_RAID3_VERSION 5 48 49#define G_RAID3_DISK_FLAG_DIRTY 0x0000000000000001ULL 50#define G_RAID3_DISK_FLAG_SYNCHRONIZING 0x0000000000000002ULL 51#define G_RAID3_DISK_FLAG_FORCE_SYNC 0x0000000000000004ULL 52#define G_RAID3_DISK_FLAG_HARDCODED 0x0000000000000008ULL 53#define G_RAID3_DISK_FLAG_BROKEN 0x0000000000000010ULL 54#define G_RAID3_DISK_FLAG_MASK (G_RAID3_DISK_FLAG_DIRTY | \ 55 G_RAID3_DISK_FLAG_SYNCHRONIZING | \ 56 G_RAID3_DISK_FLAG_FORCE_SYNC) 57 58#define G_RAID3_DEVICE_FLAG_NOAUTOSYNC 0x0000000000000001ULL 59#define G_RAID3_DEVICE_FLAG_ROUND_ROBIN 0x0000000000000002ULL 60#define G_RAID3_DEVICE_FLAG_VERIFY 0x0000000000000004ULL 61#define G_RAID3_DEVICE_FLAG_NOFAILSYNC 0x0000000000000008ULL 62#define G_RAID3_DEVICE_FLAG_MASK (G_RAID3_DEVICE_FLAG_NOAUTOSYNC | \ 63 G_RAID3_DEVICE_FLAG_ROUND_ROBIN | \ 64 G_RAID3_DEVICE_FLAG_VERIFY | \ 65 G_RAID3_DEVICE_FLAG_NOFAILSYNC) 66 67#ifdef _KERNEL 68extern u_int g_raid3_debug; 69 70#define G_RAID3_DEBUG(lvl, ...) do { \ 71 if (g_raid3_debug >= (lvl)) { \ 72 printf("GEOM_RAID3"); \ 73 if (g_raid3_debug > 0) \ 74 printf("[%u]", lvl); \ 75 printf(": "); \ 76 printf(__VA_ARGS__); \ 77 printf("\n"); \ 78 } \ 79} while (0) 80#define G_RAID3_LOGREQ(lvl, bp, ...) do { \ 81 if (g_raid3_debug >= (lvl)) { \ 82 printf("GEOM_RAID3"); \ 83 if (g_raid3_debug > 0) \ 84 printf("[%u]", lvl); \ 85 printf(": "); \ 86 printf(__VA_ARGS__); \ 87 printf(" "); \ 88 g_print_bio(bp); \ 89 printf("\n"); \ 90 } \ 91} while (0) 92 93#define G_RAID3_BIO_CFLAG_REGULAR 0x01 94#define G_RAID3_BIO_CFLAG_SYNC 0x02 95#define G_RAID3_BIO_CFLAG_PARITY 0x04 96#define G_RAID3_BIO_CFLAG_NODISK 0x08 97#define G_RAID3_BIO_CFLAG_REGSYNC 0x10 98#define G_RAID3_BIO_CFLAG_MASK (G_RAID3_BIO_CFLAG_REGULAR | \ 99 G_RAID3_BIO_CFLAG_SYNC | \ 100 G_RAID3_BIO_CFLAG_PARITY | \ 101 G_RAID3_BIO_CFLAG_NODISK | \ 102 G_RAID3_BIO_CFLAG_REGSYNC) 103 104#define G_RAID3_BIO_PFLAG_DEGRADED 0x01 105#define G_RAID3_BIO_PFLAG_NOPARITY 0x02 106#define G_RAID3_BIO_PFLAG_VERIFY 0x04 107#define G_RAID3_BIO_PFLAG_MASK (G_RAID3_BIO_PFLAG_DEGRADED | \ 108 G_RAID3_BIO_PFLAG_NOPARITY | \ 109 G_RAID3_BIO_PFLAG_VERIFY) 110 111/* 112 * Informations needed for synchronization. 113 */ 114struct g_raid3_disk_sync { 115 struct g_consumer *ds_consumer; /* Consumer connected to our device. */ 116 off_t ds_offset; /* Offset of next request to send. */ 117 off_t ds_offset_done; /* Offset of already synchronized 118 region. */ 119 off_t ds_resync; /* Resynchronize from this offset. */ 120 u_int ds_syncid; /* Disk's synchronization ID. */ 121 u_int ds_inflight; /* Number of in-flight sync requests. */ 122 struct bio **ds_bios; /* BIOs for synchronization I/O. */ 123}; 124 125/* 126 * Informations needed for synchronization. 127 */ 128struct g_raid3_device_sync { 129 struct g_geom *ds_geom; /* Synchronization geom. */ 130}; 131 132#define G_RAID3_DISK_STATE_NODISK 0 133#define G_RAID3_DISK_STATE_NONE 1 134#define G_RAID3_DISK_STATE_NEW 2 135#define G_RAID3_DISK_STATE_ACTIVE 3 136#define G_RAID3_DISK_STATE_STALE 4 137#define G_RAID3_DISK_STATE_SYNCHRONIZING 5 138#define G_RAID3_DISK_STATE_DISCONNECTED 6 139#define G_RAID3_DISK_STATE_DESTROY 7 140struct g_raid3_disk { 141 u_int d_no; /* Disk number. */ 142 struct g_consumer *d_consumer; /* Consumer. */ 143 struct g_raid3_softc *d_softc; /* Back-pointer to softc. */ 144 int d_state; /* Disk state. */ 145 uint64_t d_flags; /* Additional flags. */ 146 u_int d_genid; /* Disk's generation ID. */ 147 struct g_raid3_disk_sync d_sync; /* Sync information. */ 148 LIST_ENTRY(g_raid3_disk) d_next; 149}; 150#define d_name d_consumer->provider->name 151 152#define G_RAID3_EVENT_DONTWAIT 0x1 153#define G_RAID3_EVENT_WAIT 0x2 154#define G_RAID3_EVENT_DEVICE 0x4 155#define G_RAID3_EVENT_DONE 0x8 156struct g_raid3_event { 157 struct g_raid3_disk *e_disk; 158 int e_state; 159 int e_flags; 160 int e_error; 161 TAILQ_ENTRY(g_raid3_event) e_next; 162}; 163 164#define G_RAID3_DEVICE_FLAG_DESTROY 0x0100000000000000ULL 165#define G_RAID3_DEVICE_FLAG_WAIT 0x0200000000000000ULL 166#define G_RAID3_DEVICE_FLAG_DESTROYING 0x0400000000000000ULL 167 168#define G_RAID3_DEVICE_STATE_STARTING 0 169#define G_RAID3_DEVICE_STATE_DEGRADED 1 170#define G_RAID3_DEVICE_STATE_COMPLETE 2 171 172/* Bump syncid on first write. */ 173#define G_RAID3_BUMP_SYNCID 0x1 174/* Bump genid immediately. */ 175#define G_RAID3_BUMP_GENID 0x2 176 177enum g_raid3_zones { 178 G_RAID3_ZONE_64K, 179 G_RAID3_ZONE_16K, 180 G_RAID3_ZONE_4K, 181 G_RAID3_NUM_ZONES 182}; 183 184static __inline enum g_raid3_zones 185g_raid3_zone(size_t nbytes) { 186 if (nbytes > 65536) 187 return (G_RAID3_NUM_ZONES); 188 else if (nbytes > 16384) 189 return (G_RAID3_ZONE_64K); 190 else if (nbytes > 4096) 191 return (G_RAID3_ZONE_16K); 192 else 193 return (G_RAID3_ZONE_4K); 194}; 195 196struct g_raid3_softc { 197 u_int sc_state; /* Device state. */ 198 uint64_t sc_mediasize; /* Device size. */ 199 uint32_t sc_sectorsize; /* Sector size. */ 200 uint64_t sc_flags; /* Additional flags. */ 201 202 struct g_geom *sc_geom; 203 struct g_provider *sc_provider; 204 205 uint32_t sc_id; /* Device unique ID. */ 206 207 struct sx sc_lock; 208 struct bio_queue_head sc_queue; 209 struct mtx sc_queue_mtx; 210 struct proc *sc_worker; 211 struct bio_queue_head sc_regular_delayed; /* Delayed I/O requests due 212 collision with sync 213 requests. */ 214 struct bio_queue_head sc_inflight; /* In-flight regular write 215 requests. */ 216 struct bio_queue_head sc_sync_delayed; /* Delayed sync requests due 217 collision with regular 218 requests. */ 219 220 struct g_raid3_disk *sc_disks; 221 u_int sc_ndisks; /* Number of disks. */ 222 u_int sc_round_robin; 223 struct g_raid3_disk *sc_syncdisk; 224 225 struct g_raid3_zone { 226 uma_zone_t sz_zone; 227 size_t sz_inuse; 228 size_t sz_max; 229 u_int sz_requested; 230 u_int sz_failed; 231 } sc_zones[G_RAID3_NUM_ZONES]; 232 233 u_int sc_genid; /* Generation ID. */ 234 u_int sc_syncid; /* Synchronization ID. */ 235 int sc_bump_id; 236 struct g_raid3_device_sync sc_sync; 237 int sc_idle; /* DIRTY flags removed. */ 238 time_t sc_last_write; 239 u_int sc_writes; 240 241 TAILQ_HEAD(, g_raid3_event) sc_events; 242 struct mtx sc_events_mtx; 243 244 struct callout sc_callout; 245 246 struct root_hold_token *sc_rootmount; 247}; 248#define sc_name sc_geom->name 249 250const char *g_raid3_get_diskname(struct g_raid3_disk *disk); 251u_int g_raid3_ndisks(struct g_raid3_softc *sc, int state); 252#define G_RAID3_DESTROY_SOFT 0 253#define G_RAID3_DESTROY_DELAYED 1 254#define G_RAID3_DESTROY_HARD 2 255int g_raid3_destroy(struct g_raid3_softc *sc, int how); 256int g_raid3_event_send(void *arg, int state, int flags); 257struct g_raid3_metadata; 258int g_raid3_add_disk(struct g_raid3_softc *sc, struct g_provider *pp, 259 struct g_raid3_metadata *md); 260int g_raid3_read_metadata(struct g_consumer *cp, struct g_raid3_metadata *md); 261void g_raid3_fill_metadata(struct g_raid3_disk *disk, 262 struct g_raid3_metadata *md); 263int g_raid3_clear_metadata(struct g_raid3_disk *disk); 264void g_raid3_update_metadata(struct g_raid3_disk *disk); 265 266g_ctl_req_t g_raid3_config; 267#endif /* _KERNEL */ 268 269struct g_raid3_metadata { 270 char md_magic[16]; /* Magic value. */ 271 uint32_t md_version; /* Version number. */ 272 char md_name[16]; /* Device name. */ 273 uint32_t md_id; /* Device unique ID. */ 274 uint16_t md_no; /* Component number. */ 275 uint16_t md_all; /* Number of disks in device. */ 276 uint32_t md_genid; /* Generation ID. */ 277 uint32_t md_syncid; /* Synchronization ID. */ 278 uint64_t md_mediasize; /* Size of whole device. */ 279 uint32_t md_sectorsize; /* Sector size. */ 280 uint64_t md_sync_offset; /* Synchronized offset. */ 281 uint64_t md_mflags; /* Additional device flags. */ 282 uint64_t md_dflags; /* Additional disk flags. */ 283 char md_provider[16]; /* Hardcoded provider. */ 284 uint64_t md_provsize; /* Provider's size. */ 285 u_char md_hash[16]; /* MD5 hash. */ 286}; 287static __inline void 288raid3_metadata_encode(struct g_raid3_metadata *md, u_char *data) 289{ 290 MD5_CTX ctx; 291 292 bcopy(md->md_magic, data, 16); 293 le32enc(data + 16, md->md_version); 294 bcopy(md->md_name, data + 20, 16); 295 le32enc(data + 36, md->md_id); 296 le16enc(data + 40, md->md_no); 297 le16enc(data + 42, md->md_all); 298 le32enc(data + 44, md->md_genid); 299 le32enc(data + 48, md->md_syncid); 300 le64enc(data + 52, md->md_mediasize); 301 le32enc(data + 60, md->md_sectorsize); 302 le64enc(data + 64, md->md_sync_offset); 303 le64enc(data + 72, md->md_mflags); 304 le64enc(data + 80, md->md_dflags); 305 bcopy(md->md_provider, data + 88, 16); 306 le64enc(data + 104, md->md_provsize); 307 MD5Init(&ctx); 308 MD5Update(&ctx, data, 112); 309 MD5Final(md->md_hash, &ctx); 310 bcopy(md->md_hash, data + 112, 16); 311} 312static __inline int 313raid3_metadata_decode_v0v1v2(const u_char *data, struct g_raid3_metadata *md) 314{ 315 MD5_CTX ctx; 316 317 bcopy(data + 20, md->md_name, 16); 318 md->md_id = le32dec(data + 36); 319 md->md_no = le16dec(data + 40); 320 md->md_all = le16dec(data + 42); 321 md->md_syncid = le32dec(data + 44); 322 md->md_mediasize = le64dec(data + 48); 323 md->md_sectorsize = le32dec(data + 56); 324 md->md_sync_offset = le64dec(data + 60); 325 md->md_mflags = le64dec(data + 68); 326 md->md_dflags = le64dec(data + 76); 327 bcopy(data + 84, md->md_provider, 16); 328 bcopy(data + 100, md->md_hash, 16); 329 MD5Init(&ctx); 330 MD5Update(&ctx, data, 100); 331 MD5Final(md->md_hash, &ctx); 332 if (bcmp(md->md_hash, data + 100, 16) != 0) 333 return (EINVAL); 334 335 /* New fields. */ 336 md->md_genid = 0; 337 md->md_provsize = 0; 338 339 return (0); 340} 341static __inline int 342raid3_metadata_decode_v3(const u_char *data, struct g_raid3_metadata *md) 343{ 344 MD5_CTX ctx; 345 346 bcopy(data + 20, md->md_name, 16); 347 md->md_id = le32dec(data + 36); 348 md->md_no = le16dec(data + 40); 349 md->md_all = le16dec(data + 42); 350 md->md_genid = le32dec(data + 44); 351 md->md_syncid = le32dec(data + 48); 352 md->md_mediasize = le64dec(data + 52); 353 md->md_sectorsize = le32dec(data + 60); 354 md->md_sync_offset = le64dec(data + 64); 355 md->md_mflags = le64dec(data + 72); 356 md->md_dflags = le64dec(data + 80); 357 bcopy(data + 88, md->md_provider, 16); 358 bcopy(data + 104, md->md_hash, 16); 359 MD5Init(&ctx); 360 MD5Update(&ctx, data, 104); 361 MD5Final(md->md_hash, &ctx); 362 if (bcmp(md->md_hash, data + 104, 16) != 0) 363 return (EINVAL); 364 365 /* New fields. */ 366 md->md_provsize = 0; 367 368 return (0); 369} 370static __inline int 371raid3_metadata_decode_v4v5(const u_char *data, struct g_raid3_metadata *md) 372{ 373 MD5_CTX ctx; 374 375 bcopy(data + 20, md->md_name, 16); 376 md->md_id = le32dec(data + 36); 377 md->md_no = le16dec(data + 40); 378 md->md_all = le16dec(data + 42); 379 md->md_genid = le32dec(data + 44); 380 md->md_syncid = le32dec(data + 48); 381 md->md_mediasize = le64dec(data + 52); 382 md->md_sectorsize = le32dec(data + 60); 383 md->md_sync_offset = le64dec(data + 64); 384 md->md_mflags = le64dec(data + 72); 385 md->md_dflags = le64dec(data + 80); 386 bcopy(data + 88, md->md_provider, 16); 387 md->md_provsize = le64dec(data + 104); 388 bcopy(data + 112, md->md_hash, 16); 389 MD5Init(&ctx); 390 MD5Update(&ctx, data, 112); 391 MD5Final(md->md_hash, &ctx); 392 if (bcmp(md->md_hash, data + 112, 16) != 0) 393 return (EINVAL); 394 return (0); 395} 396static __inline int 397raid3_metadata_decode(const u_char *data, struct g_raid3_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 case 2: 407 error = raid3_metadata_decode_v0v1v2(data, md); 408 break; 409 case 3: 410 error = raid3_metadata_decode_v3(data, md); 411 break; 412 case 4: 413 case 5: 414 error = raid3_metadata_decode_v4v5(data, md); 415 break; 416 default: 417 error = EINVAL; 418 break; 419 } 420 return (error); 421} 422 423static __inline void 424raid3_metadata_dump(const struct g_raid3_metadata *md) 425{ 426 static const char hex[] = "0123456789abcdef"; 427 char hash[16 * 2 + 1]; 428 u_int i; 429 430 printf(" magic: %s\n", md->md_magic); 431 printf(" version: %u\n", (u_int)md->md_version); 432 printf(" name: %s\n", md->md_name); 433 printf(" id: %u\n", (u_int)md->md_id); 434 printf(" no: %u\n", (u_int)md->md_no); 435 printf(" all: %u\n", (u_int)md->md_all); 436 printf(" genid: %u\n", (u_int)md->md_genid); 437 printf(" syncid: %u\n", (u_int)md->md_syncid); 438 printf(" mediasize: %jd\n", (intmax_t)md->md_mediasize); 439 printf("sectorsize: %u\n", (u_int)md->md_sectorsize); 440 printf("syncoffset: %jd\n", (intmax_t)md->md_sync_offset); 441 printf(" mflags:"); 442 if (md->md_mflags == 0) 443 printf(" NONE"); 444 else { 445 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_NOAUTOSYNC) != 0) 446 printf(" NOAUTOSYNC"); 447 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_ROUND_ROBIN) != 0) 448 printf(" ROUND-ROBIN"); 449 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_VERIFY) != 0) 450 printf(" VERIFY"); 451 if ((md->md_mflags & G_RAID3_DEVICE_FLAG_NOFAILSYNC) != 0) 452 printf(" NOFAILSYNC"); 453 } 454 printf("\n"); 455 printf(" dflags:"); 456 if (md->md_dflags == 0) 457 printf(" NONE"); 458 else { 459 if ((md->md_dflags & G_RAID3_DISK_FLAG_DIRTY) != 0) 460 printf(" DIRTY"); 461 if ((md->md_dflags & G_RAID3_DISK_FLAG_SYNCHRONIZING) != 0) 462 printf(" SYNCHRONIZING"); 463 if ((md->md_dflags & G_RAID3_DISK_FLAG_FORCE_SYNC) != 0) 464 printf(" FORCE_SYNC"); 465 } 466 printf("\n"); 467 printf("hcprovider: %s\n", md->md_provider); 468 printf(" provsize: %ju\n", (uintmax_t)md->md_provsize); 469 bzero(hash, sizeof(hash)); 470 for (i = 0; i < 16; i++) { 471 hash[i * 2] = hex[md->md_hash[i] >> 4]; 472 hash[i * 2 + 1] = hex[md->md_hash[i] & 0x0f]; 473 } 474 printf(" MD5 hash: %s\n", hash); 475} 476#endif /* !_G_RAID3_H_ */ 477