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_quota_recovery;
215 struct task_struct *commit_task;
216 struct super_block *sb;
217 struct inode *root_inode;
218 struct inode *sys_root_inode;
219 struct inode *system_inodes[NUM_SYSTEM_INODES];
221 struct ocfs2_slot_info *slot_info;
223 u32 *slot_recovery_generations;
225 spinlock_t node_map_lock;
228 u64 system_dir_blkno;
235 u64 first_cluster_group_blkno;
238 u32 s_feature_compat;
239 u32 s_feature_incompat;
240 u32 s_feature_ro_compat;
242 /* Protects s_next_generation, osb_flags and s_inode_steal_slot.
243 * Could protect more on osb as it's very short lived.
246 u32 s_next_generation;
247 unsigned long osb_flags;
248 s16 s_inode_steal_slot;
249 atomic_t s_num_inodes_stolen;
251 unsigned long s_mount_opt;
252 unsigned int s_atime_quantum;
254 unsigned int max_slots;
255 unsigned int node_num;
260 int s_clustersize_bits;
261 unsigned int s_xattr_inline_size;
264 struct mutex recovery_lock;
265 struct ocfs2_recovery_map *recovery_map;
266 struct task_struct *recovery_thread_task;
267 int disable_recovery;
268 wait_queue_head_t checkpoint_event;
269 atomic_t needs_checkpoint;
270 struct ocfs2_journal *journal;
271 unsigned long osb_commit_interval;
273 struct delayed_work la_enable_wq;
276 * Must hold local alloc i_mutex and osb->osb_lock to change
277 * local_alloc_bits. Reads can be done under either lock.
279 unsigned int local_alloc_bits;
280 unsigned int local_alloc_default_bits;
282 enum ocfs2_local_alloc_state local_alloc_state; /* protected
285 struct buffer_head *local_alloc_bh;
289 #ifdef CONFIG_OCFS2_FS_STATS
290 struct dentry *local_alloc_debug;
291 char *local_alloc_debug_buf;
294 /* Next three fields are for local node slot recovery during
297 struct ocfs2_dinode *local_alloc_copy;
298 struct ocfs2_quota_recovery *quota_rec;
300 struct ocfs2_alloc_stats alloc_stats;
301 char dev_str[20]; /* "major,minor" of the device */
303 char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
304 struct ocfs2_cluster_connection *cconn;
305 struct ocfs2_lock_res osb_super_lockres;
306 struct ocfs2_lock_res osb_rename_lockres;
307 struct ocfs2_dlm_debug *osb_dlm_debug;
309 struct dentry *osb_debug_root;
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 wait_queue_head_t osb_mount_event;
330 /* Truncate log info */
331 struct inode *osb_tl_inode;
332 struct buffer_head *osb_tl_bh;
333 struct delayed_work osb_truncate_log_wq;
335 struct ocfs2_node_map osb_recovering_orphan_dirs;
336 unsigned int *osb_orphan_wipes;
337 wait_queue_head_t osb_wipe_event;
340 #define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info)
342 /* Useful typedef for passing around journal access functions */
343 typedef int (*ocfs2_journal_access_func)(handle_t *handle, struct inode *inode,
344 struct buffer_head *bh, int type);
346 static inline int ocfs2_should_order_data(struct inode *inode)
348 if (!S_ISREG(inode->i_mode))
350 if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
355 static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb)
357 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC)
362 static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb)
365 * Support for sparse files is a pre-requisite
367 if (!ocfs2_sparse_alloc(osb))
370 if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN)
375 static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb)
377 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA)
382 static inline int ocfs2_supports_xattr(struct ocfs2_super *osb)
384 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR)
389 static inline int ocfs2_meta_ecc(struct ocfs2_super *osb)
391 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_META_ECC)
396 /* set / clear functions because cluster events can make these happen
397 * in parallel so we want the transitions to be atomic. this also
398 * means that any future flags osb_flags must be protected by spinlock
400 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
403 spin_lock(&osb->osb_lock);
404 osb->osb_flags |= flag;
405 spin_unlock(&osb->osb_lock);
408 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
411 spin_lock(&osb->osb_lock);
412 osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
414 osb->osb_flags |= OCFS2_OSB_HARD_RO;
416 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
417 spin_unlock(&osb->osb_lock);
420 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
424 spin_lock(&osb->osb_lock);
425 ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
426 spin_unlock(&osb->osb_lock);
431 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
435 spin_lock(&osb->osb_lock);
436 ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
437 spin_unlock(&osb->osb_lock);
442 static inline int ocfs2_userspace_stack(struct ocfs2_super *osb)
444 return (osb->s_feature_incompat &
445 OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK);
448 static inline int ocfs2_mount_local(struct ocfs2_super *osb)
450 return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT);
453 static inline int ocfs2_uses_extended_slot_map(struct ocfs2_super *osb)
455 return (osb->s_feature_incompat &
456 OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP);
460 #define OCFS2_IS_VALID_DINODE(ptr) \
461 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
463 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \
464 (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
466 #define OCFS2_IS_VALID_GROUP_DESC(ptr) \
467 (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
470 #define OCFS2_IS_VALID_XATTR_BLOCK(ptr) \
471 (!strcmp((ptr)->xb_signature, OCFS2_XATTR_BLOCK_SIGNATURE))
473 #define OCFS2_IS_VALID_DIR_TRAILER(ptr) \
474 (!strcmp((ptr)->db_signature, OCFS2_DIR_TRAILER_SIGNATURE))
476 static inline unsigned long ino_from_blkno(struct super_block *sb,
479 return (unsigned long)(blkno & (u64)ULONG_MAX);
482 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
485 int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
486 sb->s_blocksize_bits;
488 return (u64)clusters << c_to_b_bits;
491 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
494 int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
495 sb->s_blocksize_bits;
497 return (u32)(blocks >> b_to_c_bits);
500 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
503 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
504 unsigned int clusters;
506 bytes += OCFS2_SB(sb)->s_clustersize - 1;
507 /* OCFS2 just cannot have enough clusters to overflow this */
508 clusters = (unsigned int)(bytes >> cl_bits);
513 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
516 bytes += sb->s_blocksize - 1;
517 return bytes >> sb->s_blocksize_bits;
520 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
523 return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
526 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
529 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
530 unsigned int clusters;
532 clusters = ocfs2_clusters_for_bytes(sb, bytes);
533 return (u64)clusters << cl_bits;
536 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
541 blocks = ocfs2_blocks_for_bytes(sb, bytes);
542 return blocks << sb->s_blocksize_bits;
545 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
547 return (unsigned long)((bytes + 511) >> 9);
550 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb,
551 unsigned long pg_index)
553 u32 clusters = pg_index;
554 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
556 if (unlikely(PAGE_CACHE_SHIFT > cbits))
557 clusters = pg_index << (PAGE_CACHE_SHIFT - cbits);
558 else if (PAGE_CACHE_SHIFT < cbits)
559 clusters = pg_index >> (cbits - PAGE_CACHE_SHIFT);
565 * Find the 1st page index which covers the given clusters.
567 static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb,
570 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
571 pgoff_t index = clusters;
573 if (PAGE_CACHE_SHIFT > cbits) {
574 index = (pgoff_t)clusters >> (PAGE_CACHE_SHIFT - cbits);
575 } else if (PAGE_CACHE_SHIFT < cbits) {
576 index = (pgoff_t)clusters << (cbits - PAGE_CACHE_SHIFT);
582 static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb)
584 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
585 unsigned int pages_per_cluster = 1;
587 if (PAGE_CACHE_SHIFT < cbits)
588 pages_per_cluster = 1 << (cbits - PAGE_CACHE_SHIFT);
590 return pages_per_cluster;
593 static inline unsigned int ocfs2_megabytes_to_clusters(struct super_block *sb,
596 BUILD_BUG_ON(OCFS2_MAX_CLUSTERSIZE > 1048576);
598 return megs << (20 - OCFS2_SB(sb)->s_clustersize_bits);
601 static inline void ocfs2_init_inode_steal_slot(struct ocfs2_super *osb)
603 spin_lock(&osb->osb_lock);
604 osb->s_inode_steal_slot = OCFS2_INVALID_SLOT;
605 spin_unlock(&osb->osb_lock);
606 atomic_set(&osb->s_num_inodes_stolen, 0);
609 static inline void ocfs2_set_inode_steal_slot(struct ocfs2_super *osb,
612 spin_lock(&osb->osb_lock);
613 osb->s_inode_steal_slot = slot;
614 spin_unlock(&osb->osb_lock);
617 static inline s16 ocfs2_get_inode_steal_slot(struct ocfs2_super *osb)
621 spin_lock(&osb->osb_lock);
622 slot = osb->s_inode_steal_slot;
623 spin_unlock(&osb->osb_lock);
628 #define ocfs2_set_bit ext2_set_bit
629 #define ocfs2_clear_bit ext2_clear_bit
630 #define ocfs2_test_bit ext2_test_bit
631 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit
632 #define ocfs2_find_next_bit ext2_find_next_bit