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 #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 #define OCFS2_LOCK_NOCACHE       (0x00000200) /* don't use a holder count */
 105
 106 struct ocfs2_lock_res_ops;
 107
 108 typedef void (*ocfs2_lock_callback)(int status, unsigned long data);
 109
 110 struct ocfs2_lock_res {
 111         void                    *l_priv;
 112         struct ocfs2_lock_res_ops *l_ops;
 113         spinlock_t               l_lock;
 114
 115         struct list_head         l_blocked_list;
 116         struct list_head         l_mask_waiters;
 117
 118         enum ocfs2_lock_type     l_type;
 119         unsigned long            l_flags;
 120         char                     l_name[OCFS2_LOCK_ID_MAX_LEN];
 121         int                      l_level;
 122         unsigned int             l_ro_holders;
 123         unsigned int             l_ex_holders;
 124         struct dlm_lockstatus    l_lksb;
 125
 126         /* used from AST/BAST funcs. */
 127         enum ocfs2_ast_action    l_action;
 128         enum ocfs2_unlock_action l_unlock_action;
 129         int                      l_requested;
 130         int                      l_blocking;
 131
 132         wait_queue_head_t        l_event;
 133
 134         struct list_head         l_debug_list;
 135 };
 136
 137 struct ocfs2_dlm_debug {
 138         struct kref d_refcnt;
 139         struct dentry *d_locking_state;
 140         struct list_head d_lockres_tracking;
 141 };
 142
 143 enum ocfs2_vol_state
 144 {
 145         VOLUME_INIT = 0,
 146         VOLUME_MOUNTED,
 147         VOLUME_DISMOUNTED,
 148         VOLUME_DISABLED
 149 };
 150
 151 struct ocfs2_alloc_stats
 152 {
 153         atomic_t moves;
 154         atomic_t local_data;
 155         atomic_t bitmap_data;
 156         atomic_t bg_allocs;
 157         atomic_t bg_extends;
 158 };
 159
 160 enum ocfs2_local_alloc_state
 161 {
 162         OCFS2_LA_UNUSED = 0,
 163         OCFS2_LA_ENABLED,
 164         OCFS2_LA_DISABLED
 165 };
 166
 167 enum ocfs2_mount_options
 168 {
 169         OCFS2_MOUNT_HB_LOCAL   = 1 << 0, /* Heartbeat started in local mode */
 170         OCFS2_MOUNT_BARRIER = 1 << 1,   /* Use block barriers */
 171         OCFS2_MOUNT_NOINTR  = 1 << 2,   /* Don't catch signals */
 172         OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */
 173         OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */
 174         OCFS2_MOUNT_LOCALFLOCKS = 1 << 5, /* No cluster aware user file locks */
 175 };
 176
 177 #define OCFS2_OSB_SOFT_RO       0x0001
 178 #define OCFS2_OSB_HARD_RO       0x0002
 179 #define OCFS2_OSB_ERROR_FS      0x0004
 180 #define OCFS2_DEFAULT_ATIME_QUANTUM     60
 181
 182 struct ocfs2_journal;
 183 struct ocfs2_super
 184 {
 185         struct task_struct *commit_task;
 186         struct super_block *sb;
 187         struct inode *root_inode;
 188         struct inode *sys_root_inode;
 189         struct inode *system_inodes[NUM_SYSTEM_INODES];
 190
 191         struct ocfs2_slot_info *slot_info;
 192
 193         spinlock_t node_map_lock;
 194         struct ocfs2_node_map recovery_map;
 195
 196         u64 root_blkno;
 197         u64 system_dir_blkno;
 198         u64 bitmap_blkno;
 199         u32 bitmap_cpg;
 200         u8 *uuid;
 201         char *uuid_str;
 202         u8 *vol_label;
 203         u64 first_cluster_group_blkno;
 204         u32 fs_generation;
 205
 206         u32 s_feature_compat;
 207         u32 s_feature_incompat;
 208         u32 s_feature_ro_compat;
 209
 210         /* Protects s_next_generaion, osb_flags. Could protect more on
 211          * osb as it's very short lived. */
 212         spinlock_t osb_lock;
 213         u32 s_next_generation;
 214         unsigned long osb_flags;
 215
 216         unsigned long s_mount_opt;
 217         unsigned int s_atime_quantum;
 218
 219         u16 max_slots;
 220         s16 node_num;
 221         s16 slot_num;
 222         s16 preferred_slot;
 223         int s_sectsize_bits;
 224         int s_clustersize;
 225         int s_clustersize_bits;
 226
 227         atomic_t vol_state;
 228         struct mutex recovery_lock;
 229         struct task_struct *recovery_thread_task;
 230         int disable_recovery;
 231         wait_queue_head_t checkpoint_event;
 232         atomic_t needs_checkpoint;
 233         struct ocfs2_journal *journal;
 234         unsigned long osb_commit_interval;
 235
 236         int local_alloc_size;
 237         enum ocfs2_local_alloc_state local_alloc_state;
 238         struct buffer_head *local_alloc_bh;
 239         u64 la_last_gd;
 240
 241         /* Next two fields are for local node slot recovery during
 242          * mount. */
 243         int dirty;
 244         struct ocfs2_dinode *local_alloc_copy;
 245
 246         struct ocfs2_alloc_stats alloc_stats;
 247         char dev_str[20];               /* "major,minor" of the device */
 248
 249         struct dlm_ctxt *dlm;
 250         struct ocfs2_lock_res osb_super_lockres;
 251         struct ocfs2_lock_res osb_rename_lockres;
 252         struct dlm_eviction_cb osb_eviction_cb;
 253         struct ocfs2_dlm_debug *osb_dlm_debug;
 254         struct dlm_protocol_version osb_locking_proto;
 255
 256         struct dentry *osb_debug_root;
 257
 258         wait_queue_head_t recovery_event;
 259
 260         spinlock_t dc_task_lock;
 261         struct task_struct *dc_task;
 262         wait_queue_head_t dc_event;
 263         unsigned long dc_wake_sequence;
 264         unsigned long dc_work_sequence;
 265
 266         /*
 267          * Any thread can add locks to the list, but the downconvert
 268          * thread is the only one allowed to remove locks. Any change
 269          * to this rule requires updating
 270          * ocfs2_downconvert_thread_do_work().
 271          */
 272         struct list_head blocked_lock_list;
 273         unsigned long blocked_lock_count;
 274
 275         wait_queue_head_t               osb_mount_event;
 276
 277         /* Truncate log info */
 278         struct inode                    *osb_tl_inode;
 279         struct buffer_head              *osb_tl_bh;
 280         struct delayed_work             osb_truncate_log_wq;
 281
 282         struct ocfs2_node_map           osb_recovering_orphan_dirs;
 283         unsigned int                    *osb_orphan_wipes;
 284         wait_queue_head_t               osb_wipe_event;
 285 };
 286
 287 #define OCFS2_SB(sb)        ((struct ocfs2_super *)(sb)->s_fs_info)
 288
 289 static inline int ocfs2_should_order_data(struct inode *inode)
 290 {
 291         if (!S_ISREG(inode->i_mode))
 292                 return 0;
 293         if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
 294                 return 0;
 295         return 1;
 296 }
 297
 298 static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb)
 299 {
 300         if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC)
 301                 return 1;
 302         return 0;
 303 }
 304
 305 static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb)
 306 {
 307         /*
 308          * Support for sparse files is a pre-requisite
 309          */
 310         if (!ocfs2_sparse_alloc(osb))
 311                 return 0;
 312
 313         if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN)
 314                 return 1;
 315         return 0;
 316 }
 317
 318 static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb)
 319 {
 320         if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA)
 321                 return 1;
 322         return 0;
 323 }
 324
 325 /* set / clear functions because cluster events can make these happen
 326  * in parallel so we want the transitions to be atomic. this also
 327  * means that any future flags osb_flags must be protected by spinlock
 328  * too! */
 329 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
 330                                       unsigned long flag)
 331 {
 332         spin_lock(&osb->osb_lock);
 333         osb->osb_flags |= flag;
 334         spin_unlock(&osb->osb_lock);
 335 }
 336
 337 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
 338                                      int hard)
 339 {
 340         spin_lock(&osb->osb_lock);
 341         osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
 342         if (hard)
 343                 osb->osb_flags |= OCFS2_OSB_HARD_RO;
 344         else
 345                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 346         spin_unlock(&osb->osb_lock);
 347 }
 348
 349 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
 350 {
 351         int ret;
 352
 353         spin_lock(&osb->osb_lock);
 354         ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
 355         spin_unlock(&osb->osb_lock);
 356
 357         return ret;
 358 }
 359
 360 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
 361 {
 362         int ret;
 363
 364         spin_lock(&osb->osb_lock);
 365         ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
 366         spin_unlock(&osb->osb_lock);
 367
 368         return ret;
 369 }
 370
 371 static inline int ocfs2_mount_local(struct ocfs2_super *osb)
 372 {
 373         return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT);
 374 }
 375
 376 #define OCFS2_IS_VALID_DINODE(ptr)                                      \
 377         (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
 378
 379 #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di)  do {                    \
 380         typeof(__di) ____di = (__di);                                   \
 381         ocfs2_error((__sb),                                             \
 382                 "Dinode # %llu has bad signature %.*s",                 \
 383                 (unsigned long long)le64_to_cpu((____di)->i_blkno), 7,  \
 384                 (____di)->i_signature);                                 \
 385 } while (0)
 386
 387 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr)                                \
 388         (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
 389
 390 #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb)    do {            \
 391         typeof(__eb) ____eb = (__eb);                                   \
 392         ocfs2_error((__sb),                                             \
 393                 "Extent Block # %llu has bad signature %.*s",           \
 394                 (unsigned long long)le64_to_cpu((____eb)->h_blkno), 7,  \
 395                 (____eb)->h_signature);                                 \
 396 } while (0)
 397
 398 #define OCFS2_IS_VALID_GROUP_DESC(ptr)                                  \
 399         (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
 400
 401 #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd)      do {            \
 402         typeof(__gd) ____gd = (__gd);                                   \
 403                 ocfs2_error((__sb),                                     \
 404                 "Group Descriptor # %llu has bad signature %.*s",       \
 405                 (unsigned long long)le64_to_cpu((____gd)->bg_blkno), 7, \
 406                 (____gd)->bg_signature);                                \
 407 } while (0)
 408
 409 static inline unsigned long ino_from_blkno(struct super_block *sb,
 410                                            u64 blkno)
 411 {
 412         return (unsigned long)(blkno & (u64)ULONG_MAX);
 413 }
 414
 415 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
 416                                            u32 clusters)
 417 {
 418         int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
 419                 sb->s_blocksize_bits;
 420
 421         return (u64)clusters << c_to_b_bits;
 422 }
 423
 424 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
 425                                            u64 blocks)
 426 {
 427         int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
 428                 sb->s_blocksize_bits;
 429
 430         return (u32)(blocks >> b_to_c_bits);
 431 }
 432
 433 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
 434                                                     u64 bytes)
 435 {
 436         int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
 437         unsigned int clusters;
 438
 439         bytes += OCFS2_SB(sb)->s_clustersize - 1;
 440         /* OCFS2 just cannot have enough clusters to overflow this */
 441         clusters = (unsigned int)(bytes >> cl_bits);
 442
 443         return clusters;
 444 }
 445
 446 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
 447                                          u64 bytes)
 448 {
 449         bytes += sb->s_blocksize - 1;
 450         return bytes >> sb->s_blocksize_bits;
 451 }
 452
 453 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
 454                                           u32 clusters)
 455 {
 456         return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
 457 }
 458
 459 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
 460                                                 u64 bytes)
 461 {
 462         int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
 463         unsigned int clusters;
 464
 465         clusters = ocfs2_clusters_for_bytes(sb, bytes);
 466         return (u64)clusters << cl_bits;
 467 }
 468
 469 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
 470                                               u64 bytes)
 471 {
 472         u64 blocks;
 473
 474         blocks = ocfs2_blocks_for_bytes(sb, bytes);
 475         return blocks << sb->s_blocksize_bits;
 476 }
 477
 478 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
 479 {
 480         return (unsigned long)((bytes + 511) >> 9);
 481 }
 482
 483 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb,
 484                                                         unsigned long pg_index)
 485 {
 486         u32 clusters = pg_index;
 487         unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
 488
 489         if (unlikely(PAGE_CACHE_SHIFT > cbits))
 490                 clusters = pg_index << (PAGE_CACHE_SHIFT - cbits);
 491         else if (PAGE_CACHE_SHIFT < cbits)
 492                 clusters = pg_index >> (cbits - PAGE_CACHE_SHIFT);
 493
 494         return clusters;
 495 }
 496
 497 /*
 498  * Find the 1st page index which covers the given clusters.
 499  */
 500 static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb,
 501                                                         u32 clusters)
 502 {
 503         unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
 504         pgoff_t index = clusters;
 505
 506         if (PAGE_CACHE_SHIFT > cbits) {
 507                 index = (pgoff_t)clusters >> (PAGE_CACHE_SHIFT - cbits);
 508         } else if (PAGE_CACHE_SHIFT < cbits) {
 509                 index = (pgoff_t)clusters << (cbits - PAGE_CACHE_SHIFT);
 510         }
 511
 512         return index;
 513 }
 514
 515 static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb)
 516 {
 517         unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
 518         unsigned int pages_per_cluster = 1;
 519
 520         if (PAGE_CACHE_SHIFT < cbits)
 521                 pages_per_cluster = 1 << (cbits - PAGE_CACHE_SHIFT);
 522
 523         return pages_per_cluster;
 524 }
 525
 526 #define ocfs2_set_bit ext2_set_bit
 527 #define ocfs2_clear_bit ext2_clear_bit
 528 #define ocfs2_test_bit ext2_test_bit
 529 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit
 530 #endif  /* OCFS2_H */
 531