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
  37 #include "cluster/nodemanager.h"
  38 #include "cluster/heartbeat.h"
  39 #include "cluster/tcp.h"
  40
  41 #include "dlm/dlmapi.h"
  42
  43 #include "ocfs2_fs.h"
  44 #include "endian.h"
  45 #include "ocfs2_lockid.h"
  46
  47 struct ocfs2_extent_map {
  48         u32             em_clusters;
  49         struct rb_root  em_extents;
  50 };
  51
  52 /* Most user visible OCFS2 inodes will have very few pieces of
  53  * metadata, but larger files (including bitmaps, etc) must be taken
  54  * into account when designing an access scheme. We allow a small
  55  * amount of inlined blocks to be stored on an array and grow the
  56  * structure into a rb tree when necessary. */
  57 #define OCFS2_INODE_MAX_CACHE_ARRAY 2
  58
  59 struct ocfs2_caching_info {
  60         unsigned int            ci_num_cached;
  61         union {
  62                 sector_t        ci_array[OCFS2_INODE_MAX_CACHE_ARRAY];
  63                 struct rb_root  ci_tree;
  64         } ci_cache;
  65 };
  66
  67 /* this limits us to 256 nodes
  68  * if we need more, we can do a kmalloc for the map */
  69 #define OCFS2_NODE_MAP_MAX_NODES    256
  70 struct ocfs2_node_map {
  71         u16 num_nodes;
  72         unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)];
  73 };
  74
  75 enum ocfs2_ast_action {
  76         OCFS2_AST_INVALID = 0,
  77         OCFS2_AST_ATTACH,
  78         OCFS2_AST_CONVERT,
  79         OCFS2_AST_DOWNCONVERT,
  80 };
  81
  82 /* actions for an unlockast function to take. */
  83 enum ocfs2_unlock_action {
  84         OCFS2_UNLOCK_INVALID = 0,
  85         OCFS2_UNLOCK_CANCEL_CONVERT,
  86         OCFS2_UNLOCK_DROP_LOCK,
  87 };
  88
  89 /* ocfs2_lock_res->l_flags flags. */
  90 #define OCFS2_LOCK_ATTACHED      (0x00000001) /* have we initialized
  91                                                * the lvb */
  92 #define OCFS2_LOCK_BUSY          (0x00000002) /* we are currently in
  93                                                * dlm_lock */
  94 #define OCFS2_LOCK_BLOCKED       (0x00000004) /* blocked waiting to
  95                                                * downconvert*/
  96 #define OCFS2_LOCK_LOCAL         (0x00000008) /* newly created inode */
  97 #define OCFS2_LOCK_NEEDS_REFRESH (0x00000010)
  98 #define OCFS2_LOCK_REFRESHING    (0x00000020)
  99 #define OCFS2_LOCK_INITIALIZED   (0x00000040) /* track initialization
 100                                                * for shutdown paths */
 101 #define OCFS2_LOCK_FREEING       (0x00000080) /* help dlmglue track
 102                                                * when to skip queueing
 103                                                * a lock because it's
 104                                                * about to be
 105                                                * dropped. */
 106 #define OCFS2_LOCK_QUEUED        (0x00000100) /* queued for downconvert */
 107
 108 struct ocfs2_lock_res_ops;
 109
 110 typedef void (*ocfs2_lock_callback)(int status, unsigned long data);
 111
 112 struct ocfs2_lock_res {
 113         void                    *l_priv;
 114         struct ocfs2_lock_res_ops *l_ops;
 115         spinlock_t               l_lock;
 116
 117         struct list_head         l_blocked_list;
 118         struct list_head         l_mask_waiters;
 119
 120         enum ocfs2_lock_type     l_type;
 121         unsigned long            l_flags;
 122         char                     l_name[OCFS2_LOCK_ID_MAX_LEN];
 123         int                      l_level;
 124         unsigned int             l_ro_holders;
 125         unsigned int             l_ex_holders;
 126         struct dlm_lockstatus    l_lksb;
 127
 128         /* used from AST/BAST funcs. */
 129         enum ocfs2_ast_action    l_action;
 130         enum ocfs2_unlock_action l_unlock_action;
 131         int                      l_requested;
 132         int                      l_blocking;
 133
 134         wait_queue_head_t        l_event;
 135
 136         struct list_head         l_debug_list;
 137 };
 138
 139 struct ocfs2_dlm_debug {
 140         struct kref d_refcnt;
 141         struct dentry *d_locking_state;
 142         struct list_head d_lockres_tracking;
 143 };
 144
 145 enum ocfs2_vol_state
 146 {
 147         VOLUME_INIT = 0,
 148         VOLUME_MOUNTED,
 149         VOLUME_DISMOUNTED,
 150         VOLUME_DISABLED
 151 };
 152
 153 struct ocfs2_alloc_stats
 154 {
 155         atomic_t moves;
 156         atomic_t local_data;
 157         atomic_t bitmap_data;
 158         atomic_t bg_allocs;
 159         atomic_t bg_extends;
 160 };
 161
 162 enum ocfs2_local_alloc_state
 163 {
 164         OCFS2_LA_UNUSED = 0,
 165         OCFS2_LA_ENABLED,
 166         OCFS2_LA_DISABLED
 167 };
 168
 169 enum ocfs2_mount_options
 170 {
 171         OCFS2_MOUNT_HB_LOCAL   = 1 << 0, /* Heartbeat started in local mode */
 172         OCFS2_MOUNT_BARRIER = 1 << 1,   /* Use block barriers */
 173         OCFS2_MOUNT_NOINTR  = 1 << 2,   /* Don't catch signals */
 174         OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */
 175         OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */
 176 #ifdef OCFS2_ORACORE_WORKAROUNDS
 177         OCFS2_MOUNT_COMPAT_OCFS = 1 << 30, /* ocfs1 compatibility mode */
 178 #endif
 179 };
 180
 181 #define OCFS2_OSB_SOFT_RO       0x0001
 182 #define OCFS2_OSB_HARD_RO       0x0002
 183 #define OCFS2_OSB_ERROR_FS      0x0004
 184
 185 struct ocfs2_journal;
 186 struct ocfs2_journal_handle;
 187 struct ocfs2_super
 188 {
 189         u32 osb_id;             /* id used by the proc interface */
 190         struct task_struct *commit_task;
 191         struct super_block *sb;
 192         struct inode *root_inode;
 193         struct inode *sys_root_inode;
 194         struct inode *system_inodes[NUM_SYSTEM_INODES];
 195
 196         struct ocfs2_slot_info *slot_info;
 197
 198         spinlock_t node_map_lock;
 199         struct ocfs2_node_map mounted_map;
 200         struct ocfs2_node_map recovery_map;
 201         struct ocfs2_node_map umount_map;
 202
 203         u32 num_clusters;
 204         u64 root_blkno;
 205         u64 system_dir_blkno;
 206         u64 bitmap_blkno;
 207         u32 bitmap_cpg;
 208         u8 *uuid;
 209         char *uuid_str;
 210         u8 *vol_label;
 211         u64 first_cluster_group_blkno;
 212         u32 fs_generation;
 213
 214         u32 s_feature_compat;
 215         u32 s_feature_incompat;
 216         u32 s_feature_ro_compat;
 217
 218         /* Protects s_next_generaion, osb_flags. Could protect more on
 219          * osb as it's very short lived. */
 220         spinlock_t osb_lock;
 221         u32 s_next_generation;
 222         unsigned long osb_flags;
 223
 224         unsigned long s_mount_opt;
 225
 226         u16 max_slots;
 227         u16 num_nodes;
 228         s16 node_num;
 229         s16 slot_num;
 230         int s_sectsize_bits;
 231         int s_clustersize;
 232         int s_clustersize_bits;
 233         struct proc_dir_entry *proc_sub_dir; /* points to /proc/fs/ocfs2/<maj_min> */
 234
 235         atomic_t vol_state;
 236         struct semaphore recovery_lock;
 237         struct task_struct *recovery_thread_task;
 238         int disable_recovery;
 239         wait_queue_head_t checkpoint_event;
 240         atomic_t needs_checkpoint;
 241         struct ocfs2_journal *journal;
 242
 243         enum ocfs2_local_alloc_state local_alloc_state;
 244         struct buffer_head *local_alloc_bh;
 245
 246         /* Next two fields are for local node slot recovery during
 247          * mount. */
 248         int dirty;
 249         struct ocfs2_dinode *local_alloc_copy;
 250
 251         struct ocfs2_alloc_stats alloc_stats;
 252         char dev_str[20];               /* "major,minor" of the device */
 253
 254         struct dlm_ctxt *dlm;
 255         struct ocfs2_lock_res osb_super_lockres;
 256         struct ocfs2_lock_res osb_rename_lockres;
 257         struct dlm_eviction_cb osb_eviction_cb;
 258         struct ocfs2_dlm_debug *osb_dlm_debug;
 259
 260         struct dentry *osb_debug_root;
 261
 262         wait_queue_head_t recovery_event;
 263
 264         spinlock_t vote_task_lock;
 265         struct task_struct *vote_task;
 266         wait_queue_head_t vote_event;
 267         unsigned long vote_wake_sequence;
 268         unsigned long vote_work_sequence;
 269
 270         struct list_head blocked_lock_list;
 271         unsigned long blocked_lock_count;
 272
 273         struct list_head vote_list;
 274         int vote_count;
 275
 276         u32 net_key;
 277         spinlock_t net_response_lock;
 278         unsigned int net_response_ids;
 279         struct list_head net_response_list;
 280
 281         struct o2hb_callback_func osb_hb_up;
 282         struct o2hb_callback_func osb_hb_down;
 283
 284         struct list_head        osb_net_handlers;
 285
 286         wait_queue_head_t               osb_mount_event;
 287
 288         /* Truncate log info */
 289         struct inode                    *osb_tl_inode;
 290         struct buffer_head              *osb_tl_bh;
 291         struct work_struct              osb_truncate_log_wq;
 292 };
 293
 294 #define OCFS2_SB(sb)        ((struct ocfs2_super *)(sb)->s_fs_info)
 295 #define OCFS2_MAX_OSB_ID             65536
 296
 297 static inline int ocfs2_should_order_data(struct inode *inode)
 298 {
 299         if (!S_ISREG(inode->i_mode))
 300                 return 0;
 301         if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
 302                 return 0;
 303         return 1;
 304 }
 305
 306 /* set / clear functions because cluster events can make these happen
 307  * in parallel so we want the transitions to be atomic. this also
 308  * means that any future flags osb_flags must be protected by spinlock
 309  * too! */
 310 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
 311                                       unsigned long flag)
 312 {
 313         spin_lock(&osb->osb_lock);
 314         osb->osb_flags |= flag;
 315         spin_unlock(&osb->osb_lock);
 316 }
 317
 318 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
 319                                      int hard)
 320 {
 321         spin_lock(&osb->osb_lock);
 322         osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
 323         if (hard)
 324                 osb->osb_flags |= OCFS2_OSB_HARD_RO;
 325         else
 326                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 327         spin_unlock(&osb->osb_lock);
 328 }
 329
 330 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
 331 {
 332         int ret;
 333
 334         spin_lock(&osb->osb_lock);
 335         ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
 336         spin_unlock(&osb->osb_lock);
 337
 338         return ret;
 339 }
 340
 341 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
 342 {
 343         int ret;
 344
 345         spin_lock(&osb->osb_lock);
 346         ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
 347         spin_unlock(&osb->osb_lock);
 348
 349         return ret;
 350 }
 351
 352 #define OCFS2_IS_VALID_DINODE(ptr)                                      \
 353         (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
 354
 355 #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di)  do {                    \
 356         typeof(__di) ____di = (__di);                                   \
 357         ocfs2_error((__sb),                                             \
 358                 "Dinode # %"MLFu64" has bad signature %.*s",            \
 359                 (____di)->i_blkno, 7,                                   \
 360                 (____di)->i_signature);                                 \
 361 } while (0);
 362
 363 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr)                                \
 364         (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
 365
 366 #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb)    do {            \
 367         typeof(__eb) ____eb = (__eb);                                   \
 368         ocfs2_error((__sb),                                             \
 369                 "Extent Block # %"MLFu64" has bad signature %.*s",      \
 370                 (____eb)->h_blkno, 7,                                   \
 371                 (____eb)->h_signature);                                 \
 372 } while (0);
 373
 374 #define OCFS2_IS_VALID_GROUP_DESC(ptr)                                  \
 375         (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
 376
 377 #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd)      do {            \
 378         typeof(__gd) ____gd = (__gd);                                   \
 379                 ocfs2_error((__sb),                                     \
 380                 "Group Descriptor # %"MLFu64" has bad signature %.*s",  \
 381                 (____gd)->bg_blkno, 7,                                  \
 382                 (____gd)->bg_signature);                                \
 383 } while (0);
 384
 385 static inline unsigned long ino_from_blkno(struct super_block *sb,
 386                                            u64 blkno)
 387 {
 388         return (unsigned long)(blkno & (u64)ULONG_MAX);
 389 }
 390
 391 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
 392                                            u32 clusters)
 393 {
 394         int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
 395                 sb->s_blocksize_bits;
 396
 397         return (u64)clusters << c_to_b_bits;
 398 }
 399
 400 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
 401                                            u64 blocks)
 402 {
 403         int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
 404                 sb->s_blocksize_bits;
 405
 406         return (u32)(blocks >> b_to_c_bits);
 407 }
 408
 409 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
 410                                                     u64 bytes)
 411 {
 412         int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
 413         unsigned int clusters;
 414
 415         bytes += OCFS2_SB(sb)->s_clustersize - 1;
 416         /* OCFS2 just cannot have enough clusters to overflow this */
 417         clusters = (unsigned int)(bytes >> cl_bits);
 418
 419         return clusters;
 420 }
 421
 422 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
 423                                          u64 bytes)
 424 {
 425         bytes += sb->s_blocksize - 1;
 426         return bytes >> sb->s_blocksize_bits;
 427 }
 428
 429 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
 430                                           u32 clusters)
 431 {
 432         return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
 433 }
 434
 435 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
 436                                                 u64 bytes)
 437 {
 438         int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
 439         unsigned int clusters;
 440
 441         clusters = ocfs2_clusters_for_bytes(sb, bytes);
 442         return (u64)clusters << cl_bits;
 443 }
 444
 445 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
 446                                               u64 bytes)
 447 {
 448         u64 blocks;
 449
 450         blocks = ocfs2_blocks_for_bytes(sb, bytes);
 451         return blocks << sb->s_blocksize_bits;
 452 }
 453
 454 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
 455 {
 456         return (unsigned long)((bytes + 511) >> 9);
 457 }
 458
 459 #define ocfs2_set_bit ext2_set_bit
 460 #define ocfs2_clear_bit ext2_clear_bit
 461 #define ocfs2_test_bit ext2_test_bit
 462 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit
 463 #endif  /* OCFS2_H */
 464