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 struct ocfs2_extent_map {
  50         u32             em_clusters;
  51         struct rb_root  em_extents;
  52 };
  53
  54 /* Most user visible OCFS2 inodes will have very few pieces of
  55  * metadata, but larger files (including bitmaps, etc) must be taken
  56  * into account when designing an access scheme. We allow a small
  57  * amount of inlined blocks to be stored on an array and grow the
  58  * structure into a rb tree when necessary. */
  59 #define OCFS2_INODE_MAX_CACHE_ARRAY 2
  60
  61 struct ocfs2_caching_info {
  62         unsigned int            ci_num_cached;
  63         union {
  64                 sector_t        ci_array[OCFS2_INODE_MAX_CACHE_ARRAY];
  65                 struct rb_root  ci_tree;
  66         } ci_cache;
  67 };
  68
  69 /* this limits us to 256 nodes
  70  * if we need more, we can do a kmalloc for the map */
  71 #define OCFS2_NODE_MAP_MAX_NODES    256
  72 struct ocfs2_node_map {
  73         u16 num_nodes;
  74         unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)];
  75 };
  76
  77 enum ocfs2_ast_action {
  78         OCFS2_AST_INVALID = 0,
  79         OCFS2_AST_ATTACH,
  80         OCFS2_AST_CONVERT,
  81         OCFS2_AST_DOWNCONVERT,
  82 };
  83
  84 /* actions for an unlockast function to take. */
  85 enum ocfs2_unlock_action {
  86         OCFS2_UNLOCK_INVALID = 0,
  87         OCFS2_UNLOCK_CANCEL_CONVERT,
  88         OCFS2_UNLOCK_DROP_LOCK,
  89 };
  90
  91 /* ocfs2_lock_res->l_flags flags. */
  92 #define OCFS2_LOCK_ATTACHED      (0x00000001) /* have we initialized
  93                                                * the lvb */
  94 #define OCFS2_LOCK_BUSY          (0x00000002) /* we are currently in
  95                                                * dlm_lock */
  96 #define OCFS2_LOCK_BLOCKED       (0x00000004) /* blocked waiting to
  97                                                * downconvert*/
  98 #define OCFS2_LOCK_LOCAL         (0x00000008) /* newly created inode */
  99 #define OCFS2_LOCK_NEEDS_REFRESH (0x00000010)
 100 #define OCFS2_LOCK_REFRESHING    (0x00000020)
 101 #define OCFS2_LOCK_INITIALIZED   (0x00000040) /* track initialization
 102                                                * for shutdown paths */
 103 #define OCFS2_LOCK_FREEING       (0x00000080) /* help dlmglue track
 104                                                * when to skip queueing
 105                                                * a lock because it's
 106                                                * about to be
 107                                                * dropped. */
 108 #define OCFS2_LOCK_QUEUED        (0x00000100) /* queued for downconvert */
 109
 110 struct ocfs2_lock_res_ops;
 111
 112 typedef void (*ocfs2_lock_callback)(int status, unsigned long data);
 113
 114 struct ocfs2_lock_res {
 115         void                    *l_priv;
 116         struct ocfs2_lock_res_ops *l_ops;
 117         spinlock_t               l_lock;
 118
 119         struct list_head         l_blocked_list;
 120         struct list_head         l_mask_waiters;
 121
 122         enum ocfs2_lock_type     l_type;
 123         unsigned long            l_flags;
 124         char                     l_name[OCFS2_LOCK_ID_MAX_LEN];
 125         int                      l_level;
 126         unsigned int             l_ro_holders;
 127         unsigned int             l_ex_holders;
 128         struct dlm_lockstatus    l_lksb;
 129
 130         /* used from AST/BAST funcs. */
 131         enum ocfs2_ast_action    l_action;
 132         enum ocfs2_unlock_action l_unlock_action;
 133         int                      l_requested;
 134         int                      l_blocking;
 135
 136         wait_queue_head_t        l_event;
 137
 138         struct list_head         l_debug_list;
 139 };
 140
 141 struct ocfs2_dlm_debug {
 142         struct kref d_refcnt;
 143         struct dentry *d_locking_state;
 144         struct list_head d_lockres_tracking;
 145 };
 146
 147 enum ocfs2_vol_state
 148 {
 149         VOLUME_INIT = 0,
 150         VOLUME_MOUNTED,
 151         VOLUME_DISMOUNTED,
 152         VOLUME_DISABLED
 153 };
 154
 155 struct ocfs2_alloc_stats
 156 {
 157         atomic_t moves;
 158         atomic_t local_data;
 159         atomic_t bitmap_data;
 160         atomic_t bg_allocs;
 161         atomic_t bg_extends;
 162 };
 163
 164 enum ocfs2_local_alloc_state
 165 {
 166         OCFS2_LA_UNUSED = 0,
 167         OCFS2_LA_ENABLED,
 168         OCFS2_LA_DISABLED
 169 };
 170
 171 enum ocfs2_mount_options
 172 {
 173         OCFS2_MOUNT_HB_LOCAL   = 1 << 0, /* Heartbeat started in local mode */
 174         OCFS2_MOUNT_BARRIER = 1 << 1,   /* Use block barriers */
 175         OCFS2_MOUNT_NOINTR  = 1 << 2,   /* Don't catch signals */
 176         OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */
 177         OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */
 178         OCFS2_MOUNT_TAGGED = 1 << 8, /* use tagging */
 179 };
 180
 181 #define OCFS2_OSB_SOFT_RO       0x0001
 182 #define OCFS2_OSB_HARD_RO       0x0002
 183 #define OCFS2_OSB_ERROR_FS      0x0004
 184 #define OCFS2_DEFAULT_ATIME_QUANTUM     60
 185
 186 struct ocfs2_journal;
 187 struct ocfs2_super
 188 {
 189         struct task_struct *commit_task;
 190         struct super_block *sb;
 191         struct inode *root_inode;
 192         struct inode *sys_root_inode;
 193         struct inode *system_inodes[NUM_SYSTEM_INODES];
 194
 195         struct ocfs2_slot_info *slot_info;
 196
 197         spinlock_t node_map_lock;
 198         struct ocfs2_node_map mounted_map;
 199         struct ocfs2_node_map recovery_map;
 200         struct ocfs2_node_map umount_map;
 201
 202         u64 root_blkno;
 203         u64 system_dir_blkno;
 204         u64 bitmap_blkno;
 205         u32 bitmap_cpg;
 206         u8 *uuid;
 207         char *uuid_str;
 208         u8 *vol_label;
 209         u64 first_cluster_group_blkno;
 210         u32 fs_generation;
 211
 212         u32 s_feature_compat;
 213         u32 s_feature_incompat;
 214         u32 s_feature_ro_compat;
 215
 216         /* Protects s_next_generaion, osb_flags. Could protect more on
 217          * osb as it's very short lived. */
 218         spinlock_t osb_lock;
 219         u32 s_next_generation;
 220         unsigned long osb_flags;
 221
 222         unsigned long s_mount_opt;
 223         unsigned int s_atime_quantum;
 224
 225         u16 max_slots;
 226         s16 node_num;
 227         s16 slot_num;
 228         int s_sectsize_bits;
 229         int s_clustersize;
 230         int s_clustersize_bits;
 231
 232         atomic_t vol_state;
 233         struct mutex recovery_lock;
 234         struct task_struct *recovery_thread_task;
 235         int disable_recovery;
 236         wait_queue_head_t checkpoint_event;
 237         atomic_t needs_checkpoint;
 238         struct ocfs2_journal *journal;
 239
 240         enum ocfs2_local_alloc_state local_alloc_state;
 241         struct buffer_head *local_alloc_bh;
 242         u64 la_last_gd;
 243
 244         /* Next two fields are for local node slot recovery during
 245          * mount. */
 246         int dirty;
 247         struct ocfs2_dinode *local_alloc_copy;
 248
 249         struct ocfs2_alloc_stats alloc_stats;
 250         char dev_str[20];               /* "major,minor" of the device */
 251
 252         struct dlm_ctxt *dlm;
 253         struct ocfs2_lock_res osb_super_lockres;
 254         struct ocfs2_lock_res osb_rename_lockres;
 255         struct dlm_eviction_cb osb_eviction_cb;
 256         struct ocfs2_dlm_debug *osb_dlm_debug;
 257
 258         struct dentry *osb_debug_root;
 259
 260         wait_queue_head_t recovery_event;
 261
 262         spinlock_t vote_task_lock;
 263         struct task_struct *vote_task;
 264         wait_queue_head_t vote_event;
 265         unsigned long vote_wake_sequence;
 266         unsigned long vote_work_sequence;
 267
 268         struct list_head blocked_lock_list;
 269         unsigned long blocked_lock_count;
 270
 271         struct list_head vote_list;
 272         int vote_count;
 273
 274         u32 net_key;
 275         spinlock_t net_response_lock;
 276         unsigned int net_response_ids;
 277         struct list_head net_response_list;
 278
 279         struct o2hb_callback_func osb_hb_up;
 280         struct o2hb_callback_func osb_hb_down;
 281
 282         struct list_head        osb_net_handlers;
 283
 284         wait_queue_head_t               osb_mount_event;
 285
 286         /* Truncate log info */
 287         struct inode                    *osb_tl_inode;
 288         struct buffer_head              *osb_tl_bh;
 289         struct delayed_work             osb_truncate_log_wq;
 290
 291         struct ocfs2_node_map           osb_recovering_orphan_dirs;
 292         unsigned int                    *osb_orphan_wipes;
 293         wait_queue_head_t               osb_wipe_event;
 294 };
 295
 296 #define OCFS2_SB(sb)        ((struct ocfs2_super *)(sb)->s_fs_info)
 297
 298 static inline int ocfs2_should_order_data(struct inode *inode)
 299 {
 300         if (!S_ISREG(inode->i_mode))
 301                 return 0;
 302         if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
 303                 return 0;
 304         return 1;
 305 }
 306
 307 /* set / clear functions because cluster events can make these happen
 308  * in parallel so we want the transitions to be atomic. this also
 309  * means that any future flags osb_flags must be protected by spinlock
 310  * too! */
 311 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
 312                                       unsigned long flag)
 313 {
 314         spin_lock(&osb->osb_lock);
 315         osb->osb_flags |= flag;
 316         spin_unlock(&osb->osb_lock);
 317 }
 318
 319 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
 320                                      int hard)
 321 {
 322         spin_lock(&osb->osb_lock);
 323         osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
 324         if (hard)
 325                 osb->osb_flags |= OCFS2_OSB_HARD_RO;
 326         else
 327                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 328         spin_unlock(&osb->osb_lock);
 329 }
 330
 331 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
 332 {
 333         int ret;
 334
 335         spin_lock(&osb->osb_lock);
 336         ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
 337         spin_unlock(&osb->osb_lock);
 338
 339         return ret;
 340 }
 341
 342 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
 343 {
 344         int ret;
 345
 346         spin_lock(&osb->osb_lock);
 347         ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
 348         spin_unlock(&osb->osb_lock);
 349
 350         return ret;
 351 }
 352
 353 static inline int ocfs2_mount_local(struct ocfs2_super *osb)
 354 {
 355         return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT);
 356 }
 357
 358 #define OCFS2_IS_VALID_DINODE(ptr)                                      \
 359         (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
 360
 361 #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di)  do {                    \
 362         typeof(__di) ____di = (__di);                                   \
 363         ocfs2_error((__sb),                                             \
 364                 "Dinode # %llu has bad signature %.*s",                 \
 365                 (unsigned long long)(____di)->i_blkno, 7,               \
 366                 (____di)->i_signature);                                 \
 367 } while (0);
 368
 369 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr)                                \
 370         (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
 371
 372 #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb)    do {            \
 373         typeof(__eb) ____eb = (__eb);                                   \
 374         ocfs2_error((__sb),                                             \
 375                 "Extent Block # %llu has bad signature %.*s",           \
 376                 (unsigned long long)(____eb)->h_blkno, 7,               \
 377                 (____eb)->h_signature);                                 \
 378 } while (0);
 379
 380 #define OCFS2_IS_VALID_GROUP_DESC(ptr)                                  \
 381         (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
 382
 383 #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd)      do {            \
 384         typeof(__gd) ____gd = (__gd);                                   \
 385                 ocfs2_error((__sb),                                     \
 386                 "Group Descriptor # %llu has bad signature %.*s",       \
 387                 (unsigned long long)(____gd)->bg_blkno, 7,              \
 388                 (____gd)->bg_signature);                                \
 389 } while (0);
 390
 391 static inline unsigned long ino_from_blkno(struct super_block *sb,
 392                                            u64 blkno)
 393 {
 394         return (unsigned long)(blkno & (u64)ULONG_MAX);
 395 }
 396
 397 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
 398                                            u32 clusters)
 399 {
 400         int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
 401                 sb->s_blocksize_bits;
 402
 403         return (u64)clusters << c_to_b_bits;
 404 }
 405
 406 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
 407                                            u64 blocks)
 408 {
 409         int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
 410                 sb->s_blocksize_bits;
 411
 412         return (u32)(blocks >> b_to_c_bits);
 413 }
 414
 415 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
 416                                                     u64 bytes)
 417 {
 418         int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
 419         unsigned int clusters;
 420
 421         bytes += OCFS2_SB(sb)->s_clustersize - 1;
 422         /* OCFS2 just cannot have enough clusters to overflow this */
 423         clusters = (unsigned int)(bytes >> cl_bits);
 424
 425         return clusters;
 426 }
 427
 428 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
 429                                          u64 bytes)
 430 {
 431         bytes += sb->s_blocksize - 1;
 432         return bytes >> sb->s_blocksize_bits;
 433 }
 434
 435 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
 436                                           u32 clusters)
 437 {
 438         return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
 439 }
 440
 441 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
 442                                                 u64 bytes)
 443 {
 444         int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
 445         unsigned int clusters;
 446
 447         clusters = ocfs2_clusters_for_bytes(sb, bytes);
 448         return (u64)clusters << cl_bits;
 449 }
 450
 451 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
 452                                               u64 bytes)
 453 {
 454         u64 blocks;
 455
 456         blocks = ocfs2_blocks_for_bytes(sb, bytes);
 457         return blocks << sb->s_blocksize_bits;
 458 }
 459
 460 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
 461 {
 462         return (unsigned long)((bytes + 511) >> 9);
 463 }
 464
 465 #define ocfs2_set_bit ext2_set_bit
 466 #define ocfs2_clear_bit ext2_clear_bit
 467 #define ocfs2_test_bit ext2_test_bit
 468 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit
 469 #endif  /* OCFS2_H */
 470