fs/ocfs2/ocfs2.h

   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 /* For union ocfs2_dlm_lksb */
  40 #include "stackglue.h"
  41
  42 #include "ocfs2_fs.h"
  43 #include "ocfs2_lockid.h"
  44
  45 /* Most user visible OCFS2 inodes will have very few pieces of
  46  * metadata, but larger files (including bitmaps, etc) must be taken
  47  * into account when designing an access scheme. We allow a small
  48  * amount of inlined blocks to be stored on an array and grow the
  49  * structure into a rb tree when necessary. */
  50 #define OCFS2_INODE_MAX_CACHE_ARRAY 2
  51
  52 struct ocfs2_caching_info {
  53         unsigned int            ci_num_cached;
  54         union {
  55                 sector_t        ci_array[OCFS2_INODE_MAX_CACHE_ARRAY];
  56                 struct rb_root  ci_tree;
  57         } ci_cache;
  58 };
  59
  60 /* this limits us to 256 nodes
  61  * if we need more, we can do a kmalloc for the map */
  62 #define OCFS2_NODE_MAP_MAX_NODES    256
  63 struct ocfs2_node_map {
  64         u16 num_nodes;
  65         unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)];
  66 };
  67
  68 enum ocfs2_ast_action {
  69         OCFS2_AST_INVALID = 0,
  70         OCFS2_AST_ATTACH,
  71         OCFS2_AST_CONVERT,
  72         OCFS2_AST_DOWNCONVERT,
  73 };
  74
  75 /* actions for an unlockast function to take. */
  76 enum ocfs2_unlock_action {
  77         OCFS2_UNLOCK_INVALID = 0,
  78         OCFS2_UNLOCK_CANCEL_CONVERT,
  79         OCFS2_UNLOCK_DROP_LOCK,
  80 };
  81
  82 /* ocfs2_lock_res->l_flags flags. */
  83 #define OCFS2_LOCK_ATTACHED      (0x00000001) /* have we initialized
  84                                                * the lvb */
  85 #define OCFS2_LOCK_BUSY          (0x00000002) /* we are currently in
  86                                                * dlm_lock */
  87 #define OCFS2_LOCK_BLOCKED       (0x00000004) /* blocked waiting to
  88                                                * downconvert*/
  89 #define OCFS2_LOCK_LOCAL         (0x00000008) /* newly created inode */
  90 #define OCFS2_LOCK_NEEDS_REFRESH (0x00000010)
  91 #define OCFS2_LOCK_REFRESHING    (0x00000020)
  92 #define OCFS2_LOCK_INITIALIZED   (0x00000040) /* track initialization
  93                                                * for shutdown paths */
  94 #define OCFS2_LOCK_FREEING       (0x00000080) /* help dlmglue track
  95                                                * when to skip queueing
  96                                                * a lock because it's
  97                                                * about to be
  98                                                * dropped. */
  99 #define OCFS2_LOCK_QUEUED        (0x00000100) /* queued for downconvert */
 100 #define OCFS2_LOCK_NOCACHE       (0x00000200) /* don't use a holder count */
 101 #define OCFS2_LOCK_PENDING       (0x00000400) /* This lockres is pending a
 102                                                  call to dlm_lock.  Only
 103                                                  exists with BUSY set. */
 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         union ocfs2_dlm_lksb     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         unsigned int             l_pending_gen;
 131
 132         wait_queue_head_t        l_event;
 133
 134         struct list_head         l_debug_list;
 135
 136 #ifdef CONFIG_OCFS2_FS_STATS
 137         unsigned long long       l_lock_num_prmode;        /* PR acquires */
 138         unsigned long long       l_lock_num_exmode;        /* EX acquires */
 139         unsigned int             l_lock_num_prmode_failed; /* Failed PR gets */
 140         unsigned int             l_lock_num_exmode_failed; /* Failed EX gets */
 141         unsigned long long       l_lock_total_prmode;      /* Tot wait for PR */
 142         unsigned long long       l_lock_total_exmode;      /* Tot wait for EX */
 143         unsigned int             l_lock_max_prmode;        /* Max wait for PR */
 144         unsigned int             l_lock_max_exmode;        /* Max wait for EX */
 145         unsigned int             l_lock_refresh;           /* Disk refreshes */
 146 #endif
 147 };
 148
 149 struct ocfs2_dlm_debug {
 150         struct kref d_refcnt;
 151         struct dentry *d_locking_state;
 152         struct list_head d_lockres_tracking;
 153 };
 154
 155 enum ocfs2_vol_state
 156 {
 157         VOLUME_INIT = 0,
 158         VOLUME_MOUNTED,
 159         VOLUME_DISMOUNTED,
 160         VOLUME_DISABLED
 161 };
 162
 163 struct ocfs2_alloc_stats
 164 {
 165         atomic_t moves;
 166         atomic_t local_data;
 167         atomic_t bitmap_data;
 168         atomic_t bg_allocs;
 169         atomic_t bg_extends;
 170 };
 171
 172 enum ocfs2_local_alloc_state
 173 {
 174         OCFS2_LA_UNUSED = 0,
 175         OCFS2_LA_ENABLED,
 176         OCFS2_LA_DISABLED
 177 };
 178
 179 enum ocfs2_mount_options
 180 {
 181         OCFS2_MOUNT_HB_LOCAL   = 1 << 0, /* Heartbeat started in local mode */
 182         OCFS2_MOUNT_BARRIER = 1 << 1,   /* Use block barriers */
 183         OCFS2_MOUNT_NOINTR  = 1 << 2,   /* Don't catch signals */
 184         OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */
 185         OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */
 186         OCFS2_MOUNT_LOCALFLOCKS = 1 << 5, /* No cluster aware user file locks */
 187 };
 188
 189 #define OCFS2_OSB_SOFT_RO       0x0001
 190 #define OCFS2_OSB_HARD_RO       0x0002
 191 #define OCFS2_OSB_ERROR_FS      0x0004
 192 #define OCFS2_DEFAULT_ATIME_QUANTUM     60
 193
 194 struct ocfs2_journal;
 195 struct ocfs2_slot_info;
 196 struct ocfs2_recovery_map;
 197 struct ocfs2_super
 198 {
 199         struct task_struct *commit_task;
 200         struct super_block *sb;
 201         struct inode *root_inode;
 202         struct inode *sys_root_inode;
 203         struct inode *system_inodes[NUM_SYSTEM_INODES];
 204
 205         struct ocfs2_slot_info *slot_info;
 206
 207         spinlock_t node_map_lock;
 208
 209         u64 root_blkno;
 210         u64 system_dir_blkno;
 211         u64 bitmap_blkno;
 212         u32 bitmap_cpg;
 213         u8 *uuid;
 214         char *uuid_str;
 215         u8 *vol_label;
 216         u64 first_cluster_group_blkno;
 217         u32 fs_generation;
 218
 219         u32 s_feature_compat;
 220         u32 s_feature_incompat;
 221         u32 s_feature_ro_compat;
 222
 223         /* Protects s_next_generation, osb_flags and s_inode_steal_slot.
 224          * Could protect more on osb as it's very short lived.
 225          */
 226         spinlock_t osb_lock;
 227         u32 s_next_generation;
 228         unsigned long osb_flags;
 229         s16 s_inode_steal_slot;
 230         atomic_t s_num_inodes_stolen;
 231
 232         unsigned long s_mount_opt;
 233         unsigned int s_atime_quantum;
 234
 235         unsigned int max_slots;
 236         unsigned int node_num;
 237         int slot_num;
 238         int preferred_slot;
 239         int s_sectsize_bits;
 240         int s_clustersize;
 241         int s_clustersize_bits;
 242
 243         atomic_t vol_state;
 244         struct mutex recovery_lock;
 245         struct ocfs2_recovery_map *recovery_map;
 246         struct task_struct *recovery_thread_task;
 247         int disable_recovery;
 248         wait_queue_head_t checkpoint_event;
 249         atomic_t needs_checkpoint;
 250         struct ocfs2_journal *journal;
 251         unsigned long osb_commit_interval;
 252
 253         int local_alloc_size;
 254         enum ocfs2_local_alloc_state local_alloc_state;
 255         struct buffer_head *local_alloc_bh;
 256         u64 la_last_gd;
 257
 258         /* Next two fields are for local node slot recovery during
 259          * mount. */
 260         int dirty;
 261         struct ocfs2_dinode *local_alloc_copy;
 262
 263         struct ocfs2_alloc_stats alloc_stats;
 264         char dev_str[20];               /* "major,minor" of the device */
 265
 266         char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
 267         struct ocfs2_cluster_connection *cconn;
 268         struct ocfs2_lock_res osb_super_lockres;
 269         struct ocfs2_lock_res osb_rename_lockres;
 270         struct ocfs2_dlm_debug *osb_dlm_debug;
 271
 272         struct dentry *osb_debug_root;
 273
 274         wait_queue_head_t recovery_event;
 275
 276         spinlock_t dc_task_lock;
 277         struct task_struct *dc_task;
 278         wait_queue_head_t dc_event;
 279         unsigned long dc_wake_sequence;
 280         unsigned long dc_work_sequence;
 281
 282         /*
 283          * Any thread can add locks to the list, but the downconvert
 284          * thread is the only one allowed to remove locks. Any change
 285          * to this rule requires updating
 286          * ocfs2_downconvert_thread_do_work().
 287          */
 288         struct list_head blocked_lock_list;
 289         unsigned long blocked_lock_count;
 290
 291         wait_queue_head_t               osb_mount_event;
 292
 293         /* Truncate log info */
 294         struct inode                    *osb_tl_inode;
 295         struct buffer_head              *osb_tl_bh;
 296         struct delayed_work             osb_truncate_log_wq;
 297
 298         struct ocfs2_node_map           osb_recovering_orphan_dirs;
 299         unsigned int                    *osb_orphan_wipes;
 300         wait_queue_head_t               osb_wipe_event;
 301 };
 302
 303 #define OCFS2_SB(sb)        ((struct ocfs2_super *)(sb)->s_fs_info)
 304
 305 static inline int ocfs2_should_order_data(struct inode *inode)
 306 {
 307         if (!S_ISREG(inode->i_mode))
 308                 return 0;
 309         if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
 310                 return 0;
 311         return 1;
 312 }
 313
 314 static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb)
 315 {
 316         if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC)
 317                 return 1;
 318         return 0;
 319 }
 320
 321 static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb)
 322 {
 323         /*
 324          * Support for sparse files is a pre-requisite
 325          */
 326         if (!ocfs2_sparse_alloc(osb))
 327                 return 0;
 328
 329         if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN)
 330                 return 1;
 331         return 0;
 332 }
 333
 334 static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb)
 335 {
 336         if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA)
 337                 return 1;
 338         return 0;
 339 }
 340
 341 /* set / clear functions because cluster events can make these happen
 342  * in parallel so we want the transitions to be atomic. this also
 343  * means that any future flags osb_flags must be protected by spinlock
 344  * too! */
 345 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
 346                                       unsigned long flag)
 347 {
 348         spin_lock(&osb->osb_lock);
 349         osb->osb_flags |= flag;
 350         spin_unlock(&osb->osb_lock);
 351 }
 352
 353 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
 354                                      int hard)
 355 {
 356         spin_lock(&osb->osb_lock);
 357         osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
 358         if (hard)
 359                 osb->osb_flags |= OCFS2_OSB_HARD_RO;
 360         else
 361                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 362         spin_unlock(&osb->osb_lock);
 363 }
 364
 365 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
 366 {
 367         int ret;
 368
 369         spin_lock(&osb->osb_lock);
 370         ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
 371         spin_unlock(&osb->osb_lock);
 372
 373         return ret;
 374 }
 375
 376 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
 377 {
 378         int ret;
 379
 380         spin_lock(&osb->osb_lock);
 381         ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
 382         spin_unlock(&osb->osb_lock);
 383
 384         return ret;
 385 }
 386
 387 static inline int ocfs2_userspace_stack(struct ocfs2_super *osb)
 388 {
 389         return (osb->s_feature_incompat &
 390                 OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK);
 391 }
 392
 393 static inline int ocfs2_mount_local(struct ocfs2_super *osb)
 394 {
 395         return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT);
 396 }
 397
 398 static inline int ocfs2_uses_extended_slot_map(struct ocfs2_super *osb)
 399 {
 400         return (osb->s_feature_incompat &
 401                 OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP);
 402 }
 403
 404
 405 #define OCFS2_IS_VALID_DINODE(ptr)                                      \
 406         (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
 407
 408 #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di)  do {                    \
 409         typeof(__di) ____di = (__di);                                   \
 410         ocfs2_error((__sb),                                             \
 411                 "Dinode # %llu has bad signature %.*s",                 \
 412                 (unsigned long long)le64_to_cpu((____di)->i_blkno), 7,  \
 413                 (____di)->i_signature);                                 \
 414 } while (0)
 415
 416 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr)                                \
 417         (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
 418
 419 #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb)    do {            \
 420         typeof(__eb) ____eb = (__eb);                                   \
 421         ocfs2_error((__sb),                                             \
 422                 "Extent Block # %llu has bad signature %.*s",           \
 423                 (unsigned long long)le64_to_cpu((____eb)->h_blkno), 7,  \
 424                 (____eb)->h_signature);                                 \
 425 } while (0)
 426
 427 #define OCFS2_IS_VALID_GROUP_DESC(ptr)                                  \
 428         (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
 429
 430 #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd)      do {            \
 431         typeof(__gd) ____gd = (__gd);                                   \
 432                 ocfs2_error((__sb),                                     \
 433                 "Group Descriptor # %llu has bad signature %.*s",       \
 434                 (unsigned long long)le64_to_cpu((____gd)->bg_blkno), 7, \
 435                 (____gd)->bg_signature);                                \
 436 } while (0)
 437
 438 static inline unsigned long ino_from_blkno(struct super_block *sb,
 439                                            u64 blkno)
 440 {
 441         return (unsigned long)(blkno & (u64)ULONG_MAX);
 442 }
 443
 444 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
 445                                            u32 clusters)
 446 {
 447         int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
 448                 sb->s_blocksize_bits;
 449
 450         return (u64)clusters << c_to_b_bits;
 451 }
 452
 453 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
 454                                            u64 blocks)
 455 {
 456         int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
 457                 sb->s_blocksize_bits;
 458
 459         return (u32)(blocks >> b_to_c_bits);
 460 }
 461
 462 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
 463                                                     u64 bytes)
 464 {
 465         int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
 466         unsigned int clusters;
 467
 468         bytes += OCFS2_SB(sb)->s_clustersize - 1;
 469         /* OCFS2 just cannot have enough clusters to overflow this */
 470         clusters = (unsigned int)(bytes >> cl_bits);
 471
 472         return clusters;
 473 }
 474
 475 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
 476                                          u64 bytes)
 477 {
 478         bytes += sb->s_blocksize - 1;
 479         return bytes >> sb->s_blocksize_bits;
 480 }
 481
 482 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
 483                                           u32 clusters)
 484 {
 485         return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
 486 }
 487
 488 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
 489                                                 u64 bytes)
 490 {
 491         int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
 492         unsigned int clusters;
 493
 494         clusters = ocfs2_clusters_for_bytes(sb, bytes);
 495         return (u64)clusters << cl_bits;
 496 }
 497
 498 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
 499                                               u64 bytes)
 500 {
 501         u64 blocks;
 502
 503         blocks = ocfs2_blocks_for_bytes(sb, bytes);
 504         return blocks << sb->s_blocksize_bits;
 505 }
 506
 507 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
 508 {
 509         return (unsigned long)((bytes + 511) >> 9);
 510 }
 511
 512 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb,
 513                                                         unsigned long pg_index)
 514 {
 515         u32 clusters = pg_index;
 516         unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
 517
 518         if (unlikely(PAGE_CACHE_SHIFT > cbits))
 519                 clusters = pg_index << (PAGE_CACHE_SHIFT - cbits);
 520         else if (PAGE_CACHE_SHIFT < cbits)
 521                 clusters = pg_index >> (cbits - PAGE_CACHE_SHIFT);
 522
 523         return clusters;
 524 }
 525
 526 /*
 527  * Find the 1st page index which covers the given clusters.
 528  */
 529 static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb,
 530                                                         u32 clusters)
 531 {
 532         unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
 533         pgoff_t index = clusters;
 534
 535         if (PAGE_CACHE_SHIFT > cbits) {
 536                 index = (pgoff_t)clusters >> (PAGE_CACHE_SHIFT - cbits);
 537         } else if (PAGE_CACHE_SHIFT < cbits) {
 538                 index = (pgoff_t)clusters << (cbits - PAGE_CACHE_SHIFT);
 539         }
 540
 541         return index;
 542 }
 543
 544 static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb)
 545 {
 546         unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
 547         unsigned int pages_per_cluster = 1;
 548
 549         if (PAGE_CACHE_SHIFT < cbits)
 550                 pages_per_cluster = 1 << (cbits - PAGE_CACHE_SHIFT);
 551
 552         return pages_per_cluster;
 553 }
 554
 555 static inline void ocfs2_init_inode_steal_slot(struct ocfs2_super *osb)
 556 {
 557         spin_lock(&osb->osb_lock);
 558         osb->s_inode_steal_slot = OCFS2_INVALID_SLOT;
 559         spin_unlock(&osb->osb_lock);
 560         atomic_set(&osb->s_num_inodes_stolen, 0);
 561 }
 562
 563 static inline void ocfs2_set_inode_steal_slot(struct ocfs2_super *osb,
 564                                               s16 slot)
 565 {
 566         spin_lock(&osb->osb_lock);
 567         osb->s_inode_steal_slot = slot;
 568         spin_unlock(&osb->osb_lock);
 569 }
 570
 571 static inline s16 ocfs2_get_inode_steal_slot(struct ocfs2_super *osb)
 572 {
 573         s16 slot;
 574
 575         spin_lock(&osb->osb_lock);
 576         slot = osb->s_inode_steal_slot;
 577         spin_unlock(&osb->osb_lock);
 578
 579         return slot;
 580 }
 581
 582 #define ocfs2_set_bit ext2_set_bit
 583 #define ocfs2_clear_bit ext2_clear_bit
 584 #define ocfs2_test_bit ext2_test_bit
 585 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit
 586 #endif  /* OCFS2_H */
 587