1 /* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
6 * Defines macros and structures used in OCFS2
8 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
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.
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.
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.
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 #ifndef CONFIG_OCFS2_COMPAT_JBD
38 # include <linux/jbd2.h>
40 # include <linux/jbd.h>
41 # include "ocfs2_jbd_compat.h"
44 /* For union ocfs2_dlm_lksb */
45 #include "stackglue.h"
48 #include "ocfs2_lockid.h"
50 /* Most user visible OCFS2 inodes will have very few pieces of
51 * metadata, but larger files (including bitmaps, etc) must be taken
52 * into account when designing an access scheme. We allow a small
53 * amount of inlined blocks to be stored on an array and grow the
54 * structure into a rb tree when necessary. */
55 #define OCFS2_INODE_MAX_CACHE_ARRAY 2
57 struct ocfs2_caching_info {
58 unsigned int ci_num_cached;
60 sector_t ci_array[OCFS2_INODE_MAX_CACHE_ARRAY];
61 struct rb_root ci_tree;
65 /* this limits us to 256 nodes
66 * if we need more, we can do a kmalloc for the map */
67 #define OCFS2_NODE_MAP_MAX_NODES 256
68 struct ocfs2_node_map {
70 unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)];
73 enum ocfs2_ast_action {
74 OCFS2_AST_INVALID = 0,
77 OCFS2_AST_DOWNCONVERT,
80 /* actions for an unlockast function to take. */
81 enum ocfs2_unlock_action {
82 OCFS2_UNLOCK_INVALID = 0,
83 OCFS2_UNLOCK_CANCEL_CONVERT,
84 OCFS2_UNLOCK_DROP_LOCK,
87 /* ocfs2_lock_res->l_flags flags. */
88 #define OCFS2_LOCK_ATTACHED (0x00000001) /* we have initialized
90 #define OCFS2_LOCK_BUSY (0x00000002) /* we are currently in
92 #define OCFS2_LOCK_BLOCKED (0x00000004) /* blocked waiting to
94 #define OCFS2_LOCK_LOCAL (0x00000008) /* newly created inode */
95 #define OCFS2_LOCK_NEEDS_REFRESH (0x00000010)
96 #define OCFS2_LOCK_REFRESHING (0x00000020)
97 #define OCFS2_LOCK_INITIALIZED (0x00000040) /* track initialization
98 * for shutdown paths */
99 #define OCFS2_LOCK_FREEING (0x00000080) /* help dlmglue track
100 * when to skip queueing
101 * a lock because it's
104 #define OCFS2_LOCK_QUEUED (0x00000100) /* queued for downconvert */
105 #define OCFS2_LOCK_NOCACHE (0x00000200) /* don't use a holder count */
106 #define OCFS2_LOCK_PENDING (0x00000400) /* This lockres is pending a
107 call to dlm_lock. Only
108 exists with BUSY set. */
110 struct ocfs2_lock_res_ops;
112 typedef void (*ocfs2_lock_callback)(int status, unsigned long data);
114 struct ocfs2_lock_res {
116 struct ocfs2_lock_res_ops *l_ops;
119 struct list_head l_blocked_list;
120 struct list_head l_mask_waiters;
122 enum ocfs2_lock_type l_type;
123 unsigned long l_flags;
124 char l_name[OCFS2_LOCK_ID_MAX_LEN];
126 unsigned int l_ro_holders;
127 unsigned int l_ex_holders;
128 union ocfs2_dlm_lksb l_lksb;
130 /* used from AST/BAST funcs. */
131 enum ocfs2_ast_action l_action;
132 enum ocfs2_unlock_action l_unlock_action;
135 unsigned int l_pending_gen;
137 wait_queue_head_t l_event;
139 struct list_head l_debug_list;
141 #ifdef CONFIG_OCFS2_FS_STATS
142 unsigned long long l_lock_num_prmode; /* PR acquires */
143 unsigned long long l_lock_num_exmode; /* EX acquires */
144 unsigned int l_lock_num_prmode_failed; /* Failed PR gets */
145 unsigned int l_lock_num_exmode_failed; /* Failed EX gets */
146 unsigned long long l_lock_total_prmode; /* Tot wait for PR */
147 unsigned long long l_lock_total_exmode; /* Tot wait for EX */
148 unsigned int l_lock_max_prmode; /* Max wait for PR */
149 unsigned int l_lock_max_exmode; /* Max wait for EX */
150 unsigned int l_lock_refresh; /* Disk refreshes */
154 struct ocfs2_dlm_debug {
155 struct kref d_refcnt;
156 struct dentry *d_locking_state;
157 struct list_head d_lockres_tracking;
164 VOLUME_MOUNTED_QUOTAS,
169 struct ocfs2_alloc_stats
173 atomic_t bitmap_data;
178 enum ocfs2_local_alloc_state
180 OCFS2_LA_UNUSED = 0, /* Local alloc will never be used for
181 * this mountpoint. */
182 OCFS2_LA_ENABLED, /* Local alloc is in use. */
183 OCFS2_LA_THROTTLED, /* Local alloc is in use, but number
184 * of bits has been reduced. */
185 OCFS2_LA_DISABLED /* Local alloc has temporarily been
189 enum ocfs2_mount_options
191 OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Heartbeat started in local mode */
192 OCFS2_MOUNT_BARRIER = 1 << 1, /* Use block barriers */
193 OCFS2_MOUNT_NOINTR = 1 << 2, /* Don't catch signals */
194 OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */
195 OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */
196 OCFS2_MOUNT_LOCALFLOCKS = 1 << 5, /* No cluster aware user file locks */
197 OCFS2_MOUNT_NOUSERXATTR = 1 << 6, /* No user xattr */
198 OCFS2_MOUNT_INODE64 = 1 << 7, /* Allow inode numbers > 2^32 */
199 OCFS2_MOUNT_POSIX_ACL = 1 << 8, /* POSIX access control lists */
200 OCFS2_MOUNT_USRQUOTA = 1 << 9, /* We support user quotas */
201 OCFS2_MOUNT_GRPQUOTA = 1 << 10, /* We support group quotas */
204 #define OCFS2_OSB_SOFT_RO 0x0001
205 #define OCFS2_OSB_HARD_RO 0x0002
206 #define OCFS2_OSB_ERROR_FS 0x0004
207 #define OCFS2_DEFAULT_ATIME_QUANTUM 60
209 struct ocfs2_journal;
210 struct ocfs2_slot_info;
211 struct ocfs2_recovery_map;
212 struct ocfs2_replay_map;
213 struct ocfs2_quota_recovery;
214 struct ocfs2_dentry_lock;
217 struct task_struct *commit_task;
218 struct super_block *sb;
219 struct inode *root_inode;
220 struct inode *sys_root_inode;
221 struct inode *system_inodes[NUM_SYSTEM_INODES];
223 struct ocfs2_slot_info *slot_info;
225 u32 *slot_recovery_generations;
227 spinlock_t node_map_lock;
230 u64 system_dir_blkno;
237 u64 first_cluster_group_blkno;
240 u32 s_feature_compat;
241 u32 s_feature_incompat;
242 u32 s_feature_ro_compat;
244 /* Protects s_next_generation, osb_flags and s_inode_steal_slot.
245 * Could protect more on osb as it's very short lived.
248 u32 s_next_generation;
249 unsigned long osb_flags;
250 s16 s_inode_steal_slot;
251 atomic_t s_num_inodes_stolen;
253 unsigned long s_mount_opt;
254 unsigned int s_atime_quantum;
256 unsigned int max_slots;
257 unsigned int node_num;
262 int s_clustersize_bits;
263 unsigned int s_xattr_inline_size;
266 struct mutex recovery_lock;
267 struct ocfs2_recovery_map *recovery_map;
268 struct ocfs2_replay_map *replay_map;
269 struct task_struct *recovery_thread_task;
270 int disable_recovery;
271 wait_queue_head_t checkpoint_event;
272 atomic_t needs_checkpoint;
273 struct ocfs2_journal *journal;
274 unsigned long osb_commit_interval;
276 struct delayed_work la_enable_wq;
279 * Must hold local alloc i_mutex and osb->osb_lock to change
280 * local_alloc_bits. Reads can be done under either lock.
282 unsigned int local_alloc_bits;
283 unsigned int local_alloc_default_bits;
285 enum ocfs2_local_alloc_state local_alloc_state; /* protected
288 struct buffer_head *local_alloc_bh;
292 /* Next three fields are for local node slot recovery during
295 struct ocfs2_dinode *local_alloc_copy;
296 struct ocfs2_quota_recovery *quota_rec;
298 struct ocfs2_alloc_stats alloc_stats;
299 char dev_str[20]; /* "major,minor" of the device */
301 char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
302 struct ocfs2_cluster_connection *cconn;
303 struct ocfs2_lock_res osb_super_lockres;
304 struct ocfs2_lock_res osb_rename_lockres;
305 struct ocfs2_lock_res osb_nfs_sync_lockres;
306 struct ocfs2_dlm_debug *osb_dlm_debug;
308 struct dentry *osb_debug_root;
309 struct dentry *osb_ctxt;
311 wait_queue_head_t recovery_event;
313 spinlock_t dc_task_lock;
314 struct task_struct *dc_task;
315 wait_queue_head_t dc_event;
316 unsigned long dc_wake_sequence;
317 unsigned long dc_work_sequence;
320 * Any thread can add locks to the list, but the downconvert
321 * thread is the only one allowed to remove locks. Any change
322 * to this rule requires updating
323 * ocfs2_downconvert_thread_do_work().
325 struct list_head blocked_lock_list;
326 unsigned long blocked_lock_count;
328 /* List of dentry locks to release. Anyone can add locks to
329 * the list, ocfs2_wq processes the list */
330 struct ocfs2_dentry_lock *dentry_lock_list;
331 struct work_struct dentry_lock_work;
333 wait_queue_head_t osb_mount_event;
335 /* Truncate log info */
336 struct inode *osb_tl_inode;
337 struct buffer_head *osb_tl_bh;
338 struct delayed_work osb_truncate_log_wq;
340 struct ocfs2_node_map osb_recovering_orphan_dirs;
341 unsigned int *osb_orphan_wipes;
342 wait_queue_head_t osb_wipe_event;
344 /* used to protect metaecc calculation check of xattr. */
345 spinlock_t osb_xattr_lock;
347 unsigned int osb_dx_mask;
350 /* the group we used to allocate inodes. */
351 u64 osb_inode_alloc_group;
354 #define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info)
356 /* Useful typedef for passing around journal access functions */
357 typedef int (*ocfs2_journal_access_func)(handle_t *handle, struct inode *inode,
358 struct buffer_head *bh, int type);
360 static inline int ocfs2_should_order_data(struct inode *inode)
362 if (!S_ISREG(inode->i_mode))
364 if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
369 static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb)
371 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC)
376 static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb)
379 * Support for sparse files is a pre-requisite
381 if (!ocfs2_sparse_alloc(osb))
384 if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN)
389 static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb)
391 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA)
396 static inline int ocfs2_supports_xattr(struct ocfs2_super *osb)
398 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR)
403 static inline int ocfs2_meta_ecc(struct ocfs2_super *osb)
405 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_META_ECC)
410 static inline int ocfs2_supports_indexed_dirs(struct ocfs2_super *osb)
412 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INDEXED_DIRS)
417 static inline unsigned int ocfs2_link_max(struct ocfs2_super *osb)
419 if (ocfs2_supports_indexed_dirs(osb))
420 return OCFS2_DX_LINK_MAX;
421 return OCFS2_LINK_MAX;
424 static inline unsigned int ocfs2_read_links_count(struct ocfs2_dinode *di)
426 u32 nlink = le16_to_cpu(di->i_links_count);
427 u32 hi = le16_to_cpu(di->i_links_count_hi);
429 if (di->i_dyn_features & cpu_to_le16(OCFS2_INDEXED_DIR_FL))
430 nlink |= (hi << OCFS2_LINKS_HI_SHIFT);
435 static inline void ocfs2_set_links_count(struct ocfs2_dinode *di, u32 nlink)
440 hi = nlink >> OCFS2_LINKS_HI_SHIFT;
442 di->i_links_count = cpu_to_le16(lo);
443 di->i_links_count_hi = cpu_to_le16(hi);
446 static inline void ocfs2_add_links_count(struct ocfs2_dinode *di, int n)
448 u32 links = ocfs2_read_links_count(di);
452 ocfs2_set_links_count(di, links);
455 /* set / clear functions because cluster events can make these happen
456 * in parallel so we want the transitions to be atomic. this also
457 * means that any future flags osb_flags must be protected by spinlock
459 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
462 spin_lock(&osb->osb_lock);
463 osb->osb_flags |= flag;
464 spin_unlock(&osb->osb_lock);
467 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
470 spin_lock(&osb->osb_lock);
471 osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
473 osb->osb_flags |= OCFS2_OSB_HARD_RO;
475 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
476 spin_unlock(&osb->osb_lock);
479 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
483 spin_lock(&osb->osb_lock);
484 ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
485 spin_unlock(&osb->osb_lock);
490 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
494 spin_lock(&osb->osb_lock);
495 ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
496 spin_unlock(&osb->osb_lock);
501 static inline int ocfs2_userspace_stack(struct ocfs2_super *osb)
503 return (osb->s_feature_incompat &
504 OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK);
507 static inline int ocfs2_mount_local(struct ocfs2_super *osb)
509 return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT);
512 static inline int ocfs2_uses_extended_slot_map(struct ocfs2_super *osb)
514 return (osb->s_feature_incompat &
515 OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP);
519 #define OCFS2_IS_VALID_DINODE(ptr) \
520 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
522 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \
523 (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
525 #define OCFS2_IS_VALID_GROUP_DESC(ptr) \
526 (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
529 #define OCFS2_IS_VALID_XATTR_BLOCK(ptr) \
530 (!strcmp((ptr)->xb_signature, OCFS2_XATTR_BLOCK_SIGNATURE))
532 #define OCFS2_IS_VALID_DIR_TRAILER(ptr) \
533 (!strcmp((ptr)->db_signature, OCFS2_DIR_TRAILER_SIGNATURE))
535 #define OCFS2_IS_VALID_DX_ROOT(ptr) \
536 (!strcmp((ptr)->dr_signature, OCFS2_DX_ROOT_SIGNATURE))
538 #define OCFS2_IS_VALID_DX_LEAF(ptr) \
539 (!strcmp((ptr)->dl_signature, OCFS2_DX_LEAF_SIGNATURE))
541 static inline unsigned long ino_from_blkno(struct super_block *sb,
544 return (unsigned long)(blkno & (u64)ULONG_MAX);
547 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
550 int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
551 sb->s_blocksize_bits;
553 return (u64)clusters << c_to_b_bits;
556 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
559 int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
560 sb->s_blocksize_bits;
562 return (u32)(blocks >> b_to_c_bits);
565 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
568 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
569 unsigned int clusters;
571 bytes += OCFS2_SB(sb)->s_clustersize - 1;
572 /* OCFS2 just cannot have enough clusters to overflow this */
573 clusters = (unsigned int)(bytes >> cl_bits);
578 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
581 bytes += sb->s_blocksize - 1;
582 return bytes >> sb->s_blocksize_bits;
585 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
588 return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
591 static inline u64 ocfs2_block_to_cluster_start(struct super_block *sb,
594 int bits = OCFS2_SB(sb)->s_clustersize_bits - sb->s_blocksize_bits;
595 unsigned int clusters;
597 clusters = ocfs2_blocks_to_clusters(sb, blocks);
598 return (u64)clusters << bits;
601 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
604 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
605 unsigned int clusters;
607 clusters = ocfs2_clusters_for_bytes(sb, bytes);
608 return (u64)clusters << cl_bits;
611 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
616 blocks = ocfs2_blocks_for_bytes(sb, bytes);
617 return blocks << sb->s_blocksize_bits;
620 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
622 return (unsigned long)((bytes + 511) >> 9);
625 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb,
626 unsigned long pg_index)
628 u32 clusters = pg_index;
629 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
631 if (unlikely(PAGE_CACHE_SHIFT > cbits))
632 clusters = pg_index << (PAGE_CACHE_SHIFT - cbits);
633 else if (PAGE_CACHE_SHIFT < cbits)
634 clusters = pg_index >> (cbits - PAGE_CACHE_SHIFT);
640 * Find the 1st page index which covers the given clusters.
642 static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb,
645 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
646 pgoff_t index = clusters;
648 if (PAGE_CACHE_SHIFT > cbits) {
649 index = (pgoff_t)clusters >> (PAGE_CACHE_SHIFT - cbits);
650 } else if (PAGE_CACHE_SHIFT < cbits) {
651 index = (pgoff_t)clusters << (cbits - PAGE_CACHE_SHIFT);
657 static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb)
659 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
660 unsigned int pages_per_cluster = 1;
662 if (PAGE_CACHE_SHIFT < cbits)
663 pages_per_cluster = 1 << (cbits - PAGE_CACHE_SHIFT);
665 return pages_per_cluster;
668 static inline unsigned int ocfs2_megabytes_to_clusters(struct super_block *sb,
671 BUILD_BUG_ON(OCFS2_MAX_CLUSTERSIZE > 1048576);
673 return megs << (20 - OCFS2_SB(sb)->s_clustersize_bits);
676 static inline void ocfs2_init_inode_steal_slot(struct ocfs2_super *osb)
678 spin_lock(&osb->osb_lock);
679 osb->s_inode_steal_slot = OCFS2_INVALID_SLOT;
680 spin_unlock(&osb->osb_lock);
681 atomic_set(&osb->s_num_inodes_stolen, 0);
684 static inline void ocfs2_set_inode_steal_slot(struct ocfs2_super *osb,
687 spin_lock(&osb->osb_lock);
688 osb->s_inode_steal_slot = slot;
689 spin_unlock(&osb->osb_lock);
692 static inline s16 ocfs2_get_inode_steal_slot(struct ocfs2_super *osb)
696 spin_lock(&osb->osb_lock);
697 slot = osb->s_inode_steal_slot;
698 spin_unlock(&osb->osb_lock);
703 #define ocfs2_set_bit ext2_set_bit
704 #define ocfs2_clear_bit ext2_clear_bit
705 #define ocfs2_test_bit ext2_test_bit
706 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit
707 #define ocfs2_find_next_bit ext2_find_next_bit