1 /* -*- mode: c; c-basic-offset: 8; -*- 2 * vim: noexpandtab sw=8 ts=8 sts=0: 3 * 4 * ocfs2.h 5 * 6 * Defines macros and structures used in OCFS2 7 * 8 * Copyright (C) 2002, 2004 Oracle. All rights reserved. 9 * 10 * This program is free software; you can redistribute it and/or 11 * modify it under the terms of the GNU General Public 12 * License as published by the Free Software Foundation; either 13 * version 2 of the License, or (at your option) any later version. 14 * 15 * This program is distributed in the hope that it will be useful, 16 * but WITHOUT ANY WARRANTY; without even the implied warranty of 17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 18 * General Public License for more details. 19 * 20 * You should have received a copy of the GNU General Public 21 * License along with this program; if not, write to the 22 * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 23 * Boston, MA 021110-1307, USA. 24 */ 25 26 #ifndef OCFS2_H 27 #define OCFS2_H 28 29 #include <linux/spinlock.h> 30 #include <linux/sched.h> 31 #include <linux/wait.h> 32 #include <linux/list.h> 33 #include <linux/rbtree.h> 34 #include <linux/workqueue.h> 35 #include <linux/kref.h> 36 #include <linux/mutex.h> 37 #include <linux/jbd.h> 38 39 #include "cluster/nodemanager.h" 40 #include "cluster/heartbeat.h" 41 #include "cluster/tcp.h" 42 43 #include "dlm/dlmapi.h" 44 45 #include "ocfs2_fs.h" 46 #include "endian.h" 47 #include "ocfs2_lockid.h" 48 49 /* Most user visible OCFS2 inodes will have very few pieces of 50 * metadata, but larger files (including bitmaps, etc) must be taken 51 * into account when designing an access scheme. We allow a small 52 * amount of inlined blocks to be stored on an array and grow the 53 * structure into a rb tree when necessary. */ 54 #define OCFS2_INODE_MAX_CACHE_ARRAY 2 55 56 struct ocfs2_caching_info { 57 unsigned int ci_num_cached; 58 union { 59 sector_t ci_array[OCFS2_INODE_MAX_CACHE_ARRAY]; 60 struct rb_root ci_tree; 61 } ci_cache; 62 }; 63 64 /* this limits us to 256 nodes 65 * if we need more, we can do a kmalloc for the map */ 66 #define OCFS2_NODE_MAP_MAX_NODES 256 67 struct ocfs2_node_map { 68 u16 num_nodes; 69 unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)]; 70 }; 71 72 enum ocfs2_ast_action { 73 OCFS2_AST_INVALID = 0, 74 OCFS2_AST_ATTACH, 75 OCFS2_AST_CONVERT, 76 OCFS2_AST_DOWNCONVERT, 77 }; 78 79 /* actions for an unlockast function to take. */ 80 enum ocfs2_unlock_action { 81 OCFS2_UNLOCK_INVALID = 0, 82 OCFS2_UNLOCK_CANCEL_CONVERT, 83 OCFS2_UNLOCK_DROP_LOCK, 84 }; 85 86 /* ocfs2_lock_res->l_flags flags. */ 87 #define OCFS2_LOCK_ATTACHED (0x00000001) /* have we initialized 88 * the lvb */ 89 #define OCFS2_LOCK_BUSY (0x00000002) /* we are currently in 90 * dlm_lock */ 91 #define OCFS2_LOCK_BLOCKED (0x00000004) /* blocked waiting to 92 * downconvert*/ 93 #define OCFS2_LOCK_LOCAL (0x00000008) /* newly created inode */ 94 #define OCFS2_LOCK_NEEDS_REFRESH (0x00000010) 95 #define OCFS2_LOCK_REFRESHING (0x00000020) 96 #define OCFS2_LOCK_INITIALIZED (0x00000040) /* track initialization 97 * for shutdown paths */ 98 #define OCFS2_LOCK_FREEING (0x00000080) /* help dlmglue track 99 * when to skip queueing 100 * a lock because it's 101 * about to be 102 * dropped. */ 103 #define OCFS2_LOCK_QUEUED (0x00000100) /* queued for downconvert */ 104 105 struct ocfs2_lock_res_ops; 106 107 typedef void (*ocfs2_lock_callback)(int status, unsigned long data); 108 109 struct ocfs2_lock_res { 110 void *l_priv; 111 struct ocfs2_lock_res_ops *l_ops; 112 spinlock_t l_lock; 113 114 struct list_head l_blocked_list; 115 struct list_head l_mask_waiters; 116 117 enum ocfs2_lock_type l_type; 118 unsigned long l_flags; 119 char l_name[OCFS2_LOCK_ID_MAX_LEN]; 120 int l_level; 121 unsigned int l_ro_holders; 122 unsigned int l_ex_holders; 123 struct dlm_lockstatus l_lksb; 124 125 /* used from AST/BAST funcs. */ 126 enum ocfs2_ast_action l_action; 127 enum ocfs2_unlock_action l_unlock_action; 128 int l_requested; 129 int l_blocking; 130 131 wait_queue_head_t l_event; 132 133 struct list_head l_debug_list; 134 }; 135 136 struct ocfs2_dlm_debug { 137 struct kref d_refcnt; 138 struct dentry *d_locking_state; 139 struct list_head d_lockres_tracking; 140 }; 141 142 enum ocfs2_vol_state 143 { 144 VOLUME_INIT = 0, 145 VOLUME_MOUNTED, 146 VOLUME_DISMOUNTED, 147 VOLUME_DISABLED 148 }; 149 150 struct ocfs2_alloc_stats 151 { 152 atomic_t moves; 153 atomic_t local_data; 154 atomic_t bitmap_data; 155 atomic_t bg_allocs; 156 atomic_t bg_extends; 157 }; 158 159 enum ocfs2_local_alloc_state 160 { 161 OCFS2_LA_UNUSED = 0, 162 OCFS2_LA_ENABLED, 163 OCFS2_LA_DISABLED 164 }; 165 166 enum ocfs2_mount_options 167 { 168 OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Heartbeat started in local mode */ 169 OCFS2_MOUNT_BARRIER = 1 << 1, /* Use block barriers */ 170 OCFS2_MOUNT_NOINTR = 1 << 2, /* Don't catch signals */ 171 OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */ 172 OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */ 173 }; 174 175 #define OCFS2_OSB_SOFT_RO 0x0001 176 #define OCFS2_OSB_HARD_RO 0x0002 177 #define OCFS2_OSB_ERROR_FS 0x0004 178 #define OCFS2_DEFAULT_ATIME_QUANTUM 60 179 180 struct ocfs2_journal; 181 struct ocfs2_super 182 { 183 struct task_struct *commit_task; 184 struct super_block *sb; 185 struct inode *root_inode; 186 struct inode *sys_root_inode; 187 struct inode *system_inodes[NUM_SYSTEM_INODES]; 188 189 struct ocfs2_slot_info *slot_info; 190 191 spinlock_t node_map_lock; 192 struct ocfs2_node_map recovery_map; 193 194 u64 root_blkno; 195 u64 system_dir_blkno; 196 u64 bitmap_blkno; 197 u32 bitmap_cpg; 198 u8 *uuid; 199 char *uuid_str; 200 u8 *vol_label; 201 u64 first_cluster_group_blkno; 202 u32 fs_generation; 203 204 u32 s_feature_compat; 205 u32 s_feature_incompat; 206 u32 s_feature_ro_compat; 207 208 /* Protects s_next_generaion, osb_flags. Could protect more on 209 * osb as it's very short lived. */ 210 spinlock_t osb_lock; 211 u32 s_next_generation; 212 unsigned long osb_flags; 213 214 unsigned long s_mount_opt; 215 unsigned int s_atime_quantum; 216 217 u16 max_slots; 218 s16 node_num; 219 s16 slot_num; 220 s16 preferred_slot; 221 int s_sectsize_bits; 222 int s_clustersize; 223 int s_clustersize_bits; 224 225 atomic_t vol_state; 226 struct mutex recovery_lock; 227 struct task_struct *recovery_thread_task; 228 int disable_recovery; 229 wait_queue_head_t checkpoint_event; 230 atomic_t needs_checkpoint; 231 struct ocfs2_journal *journal; 232 233 enum ocfs2_local_alloc_state local_alloc_state; 234 struct buffer_head *local_alloc_bh; 235 u64 la_last_gd; 236 237 /* Next two fields are for local node slot recovery during 238 * mount. */ 239 int dirty; 240 struct ocfs2_dinode *local_alloc_copy; 241 242 struct ocfs2_alloc_stats alloc_stats; 243 char dev_str[20]; /* "major,minor" of the device */ 244 245 struct dlm_ctxt *dlm; 246 struct ocfs2_lock_res osb_super_lockres; 247 struct ocfs2_lock_res osb_rename_lockres; 248 struct dlm_eviction_cb osb_eviction_cb; 249 struct ocfs2_dlm_debug *osb_dlm_debug; 250 251 struct dentry *osb_debug_root; 252 253 wait_queue_head_t recovery_event; 254 255 spinlock_t dc_task_lock; 256 struct task_struct *dc_task; 257 wait_queue_head_t dc_event; 258 unsigned long dc_wake_sequence; 259 unsigned long dc_work_sequence; 260 261 struct list_head blocked_lock_list; 262 unsigned long blocked_lock_count; 263 264 wait_queue_head_t osb_mount_event; 265 266 /* Truncate log info */ 267 struct inode *osb_tl_inode; 268 struct buffer_head *osb_tl_bh; 269 struct delayed_work osb_truncate_log_wq; 270 271 struct ocfs2_node_map osb_recovering_orphan_dirs; 272 unsigned int *osb_orphan_wipes; 273 wait_queue_head_t osb_wipe_event; 274 }; 275 276 #define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info) 277 278 static inline int ocfs2_should_order_data(struct inode *inode) 279 { 280 if (!S_ISREG(inode->i_mode)) 281 return 0; 282 if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) 283 return 0; 284 return 1; 285 } 286 287 static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb) 288 { 289 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC) 290 return 1; 291 return 0; 292 } 293 294 static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb) 295 { 296 /* 297 * Support for sparse files is a pre-requisite 298 */ 299 if (!ocfs2_sparse_alloc(osb)) 300 return 0; 301 302 if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN) 303 return 1; 304 return 0; 305 } 306 307 static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb) 308 { 309 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA) 310 return 1; 311 return 0; 312 } 313 314 /* set / clear functions because cluster events can make these happen 315 * in parallel so we want the transitions to be atomic. this also 316 * means that any future flags osb_flags must be protected by spinlock 317 * too! */ 318 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb, 319 unsigned long flag) 320 { 321 spin_lock(&osb->osb_lock); 322 osb->osb_flags |= flag; 323 spin_unlock(&osb->osb_lock); 324 } 325 326 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb, 327 int hard) 328 { 329 spin_lock(&osb->osb_lock); 330 osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO); 331 if (hard) 332 osb->osb_flags |= OCFS2_OSB_HARD_RO; 333 else 334 osb->osb_flags |= OCFS2_OSB_SOFT_RO; 335 spin_unlock(&osb->osb_lock); 336 } 337 338 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb) 339 { 340 int ret; 341 342 spin_lock(&osb->osb_lock); 343 ret = osb->osb_flags & OCFS2_OSB_HARD_RO; 344 spin_unlock(&osb->osb_lock); 345 346 return ret; 347 } 348 349 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb) 350 { 351 int ret; 352 353 spin_lock(&osb->osb_lock); 354 ret = osb->osb_flags & OCFS2_OSB_SOFT_RO; 355 spin_unlock(&osb->osb_lock); 356 357 return ret; 358 } 359 360 static inline int ocfs2_mount_local(struct ocfs2_super *osb) 361 { 362 return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT); 363 } 364 365 #define OCFS2_IS_VALID_DINODE(ptr) \ 366 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE)) 367 368 #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di) do { \ 369 typeof(__di) ____di = (__di); \ 370 ocfs2_error((__sb), \ 371 "Dinode # %llu has bad signature %.*s", \ 372 (unsigned long long)le64_to_cpu((____di)->i_blkno), 7, \ 373 (____di)->i_signature); \ 374 } while (0) 375 376 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \ 377 (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE)) 378 379 #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb) do { \ 380 typeof(__eb) ____eb = (__eb); \ 381 ocfs2_error((__sb), \ 382 "Extent Block # %llu has bad signature %.*s", \ 383 (unsigned long long)le64_to_cpu((____eb)->h_blkno), 7, \ 384 (____eb)->h_signature); \ 385 } while (0) 386 387 #define OCFS2_IS_VALID_GROUP_DESC(ptr) \ 388 (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE)) 389 390 #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd) do { \ 391 typeof(__gd) ____gd = (__gd); \ 392 ocfs2_error((__sb), \ 393 "Group Descriptor # %llu has bad signature %.*s", \ 394 (unsigned long long)le64_to_cpu((____gd)->bg_blkno), 7, \ 395 (____gd)->bg_signature); \ 396 } while (0) 397 398 static inline unsigned long ino_from_blkno(struct super_block *sb, 399 u64 blkno) 400 { 401 return (unsigned long)(blkno & (u64)ULONG_MAX); 402 } 403 404 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb, 405 u32 clusters) 406 { 407 int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits - 408 sb->s_blocksize_bits; 409 410 return (u64)clusters << c_to_b_bits; 411 } 412 413 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb, 414 u64 blocks) 415 { 416 int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits - 417 sb->s_blocksize_bits; 418 419 return (u32)(blocks >> b_to_c_bits); 420 } 421 422 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb, 423 u64 bytes) 424 { 425 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits; 426 unsigned int clusters; 427 428 bytes += OCFS2_SB(sb)->s_clustersize - 1; 429 /* OCFS2 just cannot have enough clusters to overflow this */ 430 clusters = (unsigned int)(bytes >> cl_bits); 431 432 return clusters; 433 } 434 435 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb, 436 u64 bytes) 437 { 438 bytes += sb->s_blocksize - 1; 439 return bytes >> sb->s_blocksize_bits; 440 } 441 442 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb, 443 u32 clusters) 444 { 445 return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits; 446 } 447 448 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb, 449 u64 bytes) 450 { 451 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits; 452 unsigned int clusters; 453 454 clusters = ocfs2_clusters_for_bytes(sb, bytes); 455 return (u64)clusters << cl_bits; 456 } 457 458 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb, 459 u64 bytes) 460 { 461 u64 blocks; 462 463 blocks = ocfs2_blocks_for_bytes(sb, bytes); 464 return blocks << sb->s_blocksize_bits; 465 } 466 467 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes) 468 { 469 return (unsigned long)((bytes + 511) >> 9); 470 } 471 472 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb, 473 unsigned long pg_index) 474 { 475 u32 clusters = pg_index; 476 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; 477 478 if (unlikely(PAGE_CACHE_SHIFT > cbits)) 479 clusters = pg_index << (PAGE_CACHE_SHIFT - cbits); 480 else if (PAGE_CACHE_SHIFT < cbits) 481 clusters = pg_index >> (cbits - PAGE_CACHE_SHIFT); 482 483 return clusters; 484 } 485 486 /* 487 * Find the 1st page index which covers the given clusters. 488 */ 489 static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb, 490 u32 clusters) 491 { 492 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; 493 pgoff_t index = clusters; 494 495 if (PAGE_CACHE_SHIFT > cbits) { 496 index = (pgoff_t)clusters >> (PAGE_CACHE_SHIFT - cbits); 497 } else if (PAGE_CACHE_SHIFT < cbits) { 498 index = (pgoff_t)clusters << (cbits - PAGE_CACHE_SHIFT); 499 } 500 501 return index; 502 } 503 504 static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb) 505 { 506 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; 507 unsigned int pages_per_cluster = 1; 508 509 if (PAGE_CACHE_SHIFT < cbits) 510 pages_per_cluster = 1 << (cbits - PAGE_CACHE_SHIFT); 511 512 return pages_per_cluster; 513 } 514 515 #define ocfs2_set_bit ext2_set_bit 516 #define ocfs2_clear_bit ext2_clear_bit 517 #define ocfs2_test_bit ext2_test_bit 518 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit 519 #endif /* OCFS2_H */ 520 521