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 /* For struct ocfs2_blockcheck_stats */
51 #include "blockcheck.h"
53 /* Most user visible OCFS2 inodes will have very few pieces of
54 * metadata, but larger files (including bitmaps, etc) must be taken
55 * into account when designing an access scheme. We allow a small
56 * amount of inlined blocks to be stored on an array and grow the
57 * structure into a rb tree when necessary. */
58 #define OCFS2_INODE_MAX_CACHE_ARRAY 2
60 struct ocfs2_caching_info {
61 unsigned int ci_num_cached;
63 sector_t ci_array[OCFS2_INODE_MAX_CACHE_ARRAY];
64 struct rb_root ci_tree;
68 /* this limits us to 256 nodes
69 * if we need more, we can do a kmalloc for the map */
70 #define OCFS2_NODE_MAP_MAX_NODES 256
71 struct ocfs2_node_map {
73 unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)];
76 enum ocfs2_ast_action {
77 OCFS2_AST_INVALID = 0,
80 OCFS2_AST_DOWNCONVERT,
83 /* actions for an unlockast function to take. */
84 enum ocfs2_unlock_action {
85 OCFS2_UNLOCK_INVALID = 0,
86 OCFS2_UNLOCK_CANCEL_CONVERT,
87 OCFS2_UNLOCK_DROP_LOCK,
90 /* ocfs2_lock_res->l_flags flags. */
91 #define OCFS2_LOCK_ATTACHED (0x00000001) /* we have initialized
93 #define OCFS2_LOCK_BUSY (0x00000002) /* we are currently in
95 #define OCFS2_LOCK_BLOCKED (0x00000004) /* blocked waiting to
97 #define OCFS2_LOCK_LOCAL (0x00000008) /* newly created inode */
98 #define OCFS2_LOCK_NEEDS_REFRESH (0x00000010)
99 #define OCFS2_LOCK_REFRESHING (0x00000020)
100 #define OCFS2_LOCK_INITIALIZED (0x00000040) /* track initialization
101 * for shutdown paths */
102 #define OCFS2_LOCK_FREEING (0x00000080) /* help dlmglue track
103 * when to skip queueing
104 * a lock because it's
107 #define OCFS2_LOCK_QUEUED (0x00000100) /* queued for downconvert */
108 #define OCFS2_LOCK_NOCACHE (0x00000200) /* don't use a holder count */
109 #define OCFS2_LOCK_PENDING (0x00000400) /* This lockres is pending a
110 call to dlm_lock. Only
111 exists with BUSY set. */
113 struct ocfs2_lock_res_ops;
115 typedef void (*ocfs2_lock_callback)(int status, unsigned long data);
117 struct ocfs2_lock_res {
119 struct ocfs2_lock_res_ops *l_ops;
122 struct list_head l_blocked_list;
123 struct list_head l_mask_waiters;
125 enum ocfs2_lock_type l_type;
126 unsigned long l_flags;
127 char l_name[OCFS2_LOCK_ID_MAX_LEN];
129 unsigned int l_ro_holders;
130 unsigned int l_ex_holders;
131 union ocfs2_dlm_lksb l_lksb;
133 /* used from AST/BAST funcs. */
134 enum ocfs2_ast_action l_action;
135 enum ocfs2_unlock_action l_unlock_action;
138 unsigned int l_pending_gen;
140 wait_queue_head_t l_event;
142 struct list_head l_debug_list;
144 #ifdef CONFIG_OCFS2_FS_STATS
145 unsigned long long l_lock_num_prmode; /* PR acquires */
146 unsigned long long l_lock_num_exmode; /* EX acquires */
147 unsigned int l_lock_num_prmode_failed; /* Failed PR gets */
148 unsigned int l_lock_num_exmode_failed; /* Failed EX gets */
149 unsigned long long l_lock_total_prmode; /* Tot wait for PR */
150 unsigned long long l_lock_total_exmode; /* Tot wait for EX */
151 unsigned int l_lock_max_prmode; /* Max wait for PR */
152 unsigned int l_lock_max_exmode; /* Max wait for EX */
153 unsigned int l_lock_refresh; /* Disk refreshes */
157 struct ocfs2_orphan_scan {
158 struct mutex os_lock;
159 struct ocfs2_super *os_osb;
160 struct ocfs2_lock_res os_lockres; /* lock to synchronize scans */
161 struct delayed_work os_orphan_scan_work;
162 struct timespec os_scantime; /* time this node ran the scan */
163 u32 os_count; /* tracks node specific scans */
164 u32 os_seqno; /* tracks cluster wide scans */
167 struct ocfs2_dlm_debug {
168 struct kref d_refcnt;
169 struct dentry *d_locking_state;
170 struct list_head d_lockres_tracking;
177 VOLUME_MOUNTED_QUOTAS,
182 struct ocfs2_alloc_stats
186 atomic_t bitmap_data;
191 enum ocfs2_local_alloc_state
193 OCFS2_LA_UNUSED = 0, /* Local alloc will never be used for
194 * this mountpoint. */
195 OCFS2_LA_ENABLED, /* Local alloc is in use. */
196 OCFS2_LA_THROTTLED, /* Local alloc is in use, but number
197 * of bits has been reduced. */
198 OCFS2_LA_DISABLED /* Local alloc has temporarily been
202 enum ocfs2_mount_options
204 OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Heartbeat started in local mode */
205 OCFS2_MOUNT_BARRIER = 1 << 1, /* Use block barriers */
206 OCFS2_MOUNT_NOINTR = 1 << 2, /* Don't catch signals */
207 OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */
208 OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */
209 OCFS2_MOUNT_LOCALFLOCKS = 1 << 5, /* No cluster aware user file locks */
210 OCFS2_MOUNT_NOUSERXATTR = 1 << 6, /* No user xattr */
211 OCFS2_MOUNT_INODE64 = 1 << 7, /* Allow inode numbers > 2^32 */
212 OCFS2_MOUNT_POSIX_ACL = 1 << 8, /* POSIX access control lists */
213 OCFS2_MOUNT_USRQUOTA = 1 << 9, /* We support user quotas */
214 OCFS2_MOUNT_GRPQUOTA = 1 << 10, /* We support group quotas */
217 #define OCFS2_OSB_SOFT_RO 0x0001
218 #define OCFS2_OSB_HARD_RO 0x0002
219 #define OCFS2_OSB_ERROR_FS 0x0004
220 #define OCFS2_DEFAULT_ATIME_QUANTUM 60
222 struct ocfs2_journal;
223 struct ocfs2_slot_info;
224 struct ocfs2_recovery_map;
225 struct ocfs2_replay_map;
226 struct ocfs2_quota_recovery;
227 struct ocfs2_dentry_lock;
230 struct task_struct *commit_task;
231 struct super_block *sb;
232 struct inode *root_inode;
233 struct inode *sys_root_inode;
234 struct inode *system_inodes[NUM_SYSTEM_INODES];
236 struct ocfs2_slot_info *slot_info;
238 u32 *slot_recovery_generations;
240 spinlock_t node_map_lock;
243 u64 system_dir_blkno;
250 u64 first_cluster_group_blkno;
253 u32 s_feature_compat;
254 u32 s_feature_incompat;
255 u32 s_feature_ro_compat;
257 /* Protects s_next_generation, osb_flags and s_inode_steal_slot.
258 * Could protect more on osb as it's very short lived.
261 u32 s_next_generation;
262 unsigned long osb_flags;
263 s16 s_inode_steal_slot;
264 atomic_t s_num_inodes_stolen;
266 unsigned long s_mount_opt;
267 unsigned int s_atime_quantum;
269 unsigned int max_slots;
270 unsigned int node_num;
275 int s_clustersize_bits;
276 unsigned int s_xattr_inline_size;
279 struct mutex recovery_lock;
280 struct ocfs2_recovery_map *recovery_map;
281 struct ocfs2_replay_map *replay_map;
282 struct task_struct *recovery_thread_task;
283 int disable_recovery;
284 wait_queue_head_t checkpoint_event;
285 atomic_t needs_checkpoint;
286 struct ocfs2_journal *journal;
287 unsigned long osb_commit_interval;
289 struct delayed_work la_enable_wq;
292 * Must hold local alloc i_mutex and osb->osb_lock to change
293 * local_alloc_bits. Reads can be done under either lock.
295 unsigned int local_alloc_bits;
296 unsigned int local_alloc_default_bits;
298 enum ocfs2_local_alloc_state local_alloc_state; /* protected
301 struct buffer_head *local_alloc_bh;
305 /* Next three fields are for local node slot recovery during
308 struct ocfs2_dinode *local_alloc_copy;
309 struct ocfs2_quota_recovery *quota_rec;
311 struct ocfs2_blockcheck_stats osb_ecc_stats;
312 struct ocfs2_alloc_stats alloc_stats;
313 char dev_str[20]; /* "major,minor" of the device */
315 char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
316 struct ocfs2_cluster_connection *cconn;
317 struct ocfs2_lock_res osb_super_lockres;
318 struct ocfs2_lock_res osb_rename_lockres;
319 struct ocfs2_lock_res osb_nfs_sync_lockres;
320 struct ocfs2_dlm_debug *osb_dlm_debug;
322 struct dentry *osb_debug_root;
323 struct dentry *osb_ctxt;
325 wait_queue_head_t recovery_event;
327 spinlock_t dc_task_lock;
328 struct task_struct *dc_task;
329 wait_queue_head_t dc_event;
330 unsigned long dc_wake_sequence;
331 unsigned long dc_work_sequence;
334 * Any thread can add locks to the list, but the downconvert
335 * thread is the only one allowed to remove locks. Any change
336 * to this rule requires updating
337 * ocfs2_downconvert_thread_do_work().
339 struct list_head blocked_lock_list;
340 unsigned long blocked_lock_count;
342 /* List of dentry locks to release. Anyone can add locks to
343 * the list, ocfs2_wq processes the list */
344 struct ocfs2_dentry_lock *dentry_lock_list;
345 struct work_struct dentry_lock_work;
347 wait_queue_head_t osb_mount_event;
349 /* Truncate log info */
350 struct inode *osb_tl_inode;
351 struct buffer_head *osb_tl_bh;
352 struct delayed_work osb_truncate_log_wq;
354 struct ocfs2_node_map osb_recovering_orphan_dirs;
355 unsigned int *osb_orphan_wipes;
356 wait_queue_head_t osb_wipe_event;
358 struct ocfs2_orphan_scan osb_orphan_scan;
360 /* used to protect metaecc calculation check of xattr. */
361 spinlock_t osb_xattr_lock;
363 unsigned int osb_dx_mask;
366 /* the group we used to allocate inodes. */
367 u64 osb_inode_alloc_group;
370 #define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info)
372 /* Useful typedef for passing around journal access functions */
373 typedef int (*ocfs2_journal_access_func)(handle_t *handle, struct inode *inode,
374 struct buffer_head *bh, int type);
376 static inline int ocfs2_should_order_data(struct inode *inode)
378 if (!S_ISREG(inode->i_mode))
380 if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
385 static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb)
387 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC)
392 static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb)
395 * Support for sparse files is a pre-requisite
397 if (!ocfs2_sparse_alloc(osb))
400 if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN)
405 static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb)
407 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA)
412 static inline int ocfs2_supports_xattr(struct ocfs2_super *osb)
414 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR)
419 static inline int ocfs2_meta_ecc(struct ocfs2_super *osb)
421 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_META_ECC)
426 static inline int ocfs2_supports_indexed_dirs(struct ocfs2_super *osb)
428 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INDEXED_DIRS)
433 static inline unsigned int ocfs2_link_max(struct ocfs2_super *osb)
435 if (ocfs2_supports_indexed_dirs(osb))
436 return OCFS2_DX_LINK_MAX;
437 return OCFS2_LINK_MAX;
440 static inline unsigned int ocfs2_read_links_count(struct ocfs2_dinode *di)
442 u32 nlink = le16_to_cpu(di->i_links_count);
443 u32 hi = le16_to_cpu(di->i_links_count_hi);
445 if (di->i_dyn_features & cpu_to_le16(OCFS2_INDEXED_DIR_FL))
446 nlink |= (hi << OCFS2_LINKS_HI_SHIFT);
451 static inline void ocfs2_set_links_count(struct ocfs2_dinode *di, u32 nlink)
456 hi = nlink >> OCFS2_LINKS_HI_SHIFT;
458 di->i_links_count = cpu_to_le16(lo);
459 di->i_links_count_hi = cpu_to_le16(hi);
462 static inline void ocfs2_add_links_count(struct ocfs2_dinode *di, int n)
464 u32 links = ocfs2_read_links_count(di);
468 ocfs2_set_links_count(di, links);
471 /* set / clear functions because cluster events can make these happen
472 * in parallel so we want the transitions to be atomic. this also
473 * means that any future flags osb_flags must be protected by spinlock
475 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
478 spin_lock(&osb->osb_lock);
479 osb->osb_flags |= flag;
480 spin_unlock(&osb->osb_lock);
483 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
486 spin_lock(&osb->osb_lock);
487 osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
489 osb->osb_flags |= OCFS2_OSB_HARD_RO;
491 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
492 spin_unlock(&osb->osb_lock);
495 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
499 spin_lock(&osb->osb_lock);
500 ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
501 spin_unlock(&osb->osb_lock);
506 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
510 spin_lock(&osb->osb_lock);
511 ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
512 spin_unlock(&osb->osb_lock);
517 static inline int ocfs2_userspace_stack(struct ocfs2_super *osb)
519 return (osb->s_feature_incompat &
520 OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK);
523 static inline int ocfs2_mount_local(struct ocfs2_super *osb)
525 return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT);
528 static inline int ocfs2_uses_extended_slot_map(struct ocfs2_super *osb)
530 return (osb->s_feature_incompat &
531 OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP);
535 #define OCFS2_IS_VALID_DINODE(ptr) \
536 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
538 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \
539 (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
541 #define OCFS2_IS_VALID_GROUP_DESC(ptr) \
542 (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
545 #define OCFS2_IS_VALID_XATTR_BLOCK(ptr) \
546 (!strcmp((ptr)->xb_signature, OCFS2_XATTR_BLOCK_SIGNATURE))
548 #define OCFS2_IS_VALID_DIR_TRAILER(ptr) \
549 (!strcmp((ptr)->db_signature, OCFS2_DIR_TRAILER_SIGNATURE))
551 #define OCFS2_IS_VALID_DX_ROOT(ptr) \
552 (!strcmp((ptr)->dr_signature, OCFS2_DX_ROOT_SIGNATURE))
554 #define OCFS2_IS_VALID_DX_LEAF(ptr) \
555 (!strcmp((ptr)->dl_signature, OCFS2_DX_LEAF_SIGNATURE))
557 static inline unsigned long ino_from_blkno(struct super_block *sb,
560 return (unsigned long)(blkno & (u64)ULONG_MAX);
563 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
566 int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
567 sb->s_blocksize_bits;
569 return (u64)clusters << c_to_b_bits;
572 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
575 int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
576 sb->s_blocksize_bits;
578 return (u32)(blocks >> b_to_c_bits);
581 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
584 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
585 unsigned int clusters;
587 bytes += OCFS2_SB(sb)->s_clustersize - 1;
588 /* OCFS2 just cannot have enough clusters to overflow this */
589 clusters = (unsigned int)(bytes >> cl_bits);
594 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
597 bytes += sb->s_blocksize - 1;
598 return bytes >> sb->s_blocksize_bits;
601 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
604 return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
607 static inline u64 ocfs2_block_to_cluster_start(struct super_block *sb,
610 int bits = OCFS2_SB(sb)->s_clustersize_bits - sb->s_blocksize_bits;
611 unsigned int clusters;
613 clusters = ocfs2_blocks_to_clusters(sb, blocks);
614 return (u64)clusters << bits;
617 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
620 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
621 unsigned int clusters;
623 clusters = ocfs2_clusters_for_bytes(sb, bytes);
624 return (u64)clusters << cl_bits;
627 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
632 blocks = ocfs2_blocks_for_bytes(sb, bytes);
633 return blocks << sb->s_blocksize_bits;
636 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
638 return (unsigned long)((bytes + 511) >> 9);
641 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb,
642 unsigned long pg_index)
644 u32 clusters = pg_index;
645 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
647 if (unlikely(PAGE_CACHE_SHIFT > cbits))
648 clusters = pg_index << (PAGE_CACHE_SHIFT - cbits);
649 else if (PAGE_CACHE_SHIFT < cbits)
650 clusters = pg_index >> (cbits - PAGE_CACHE_SHIFT);
656 * Find the 1st page index which covers the given clusters.
658 static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb,
661 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
662 pgoff_t index = clusters;
664 if (PAGE_CACHE_SHIFT > cbits) {
665 index = (pgoff_t)clusters >> (PAGE_CACHE_SHIFT - cbits);
666 } else if (PAGE_CACHE_SHIFT < cbits) {
667 index = (pgoff_t)clusters << (cbits - PAGE_CACHE_SHIFT);
673 static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb)
675 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
676 unsigned int pages_per_cluster = 1;
678 if (PAGE_CACHE_SHIFT < cbits)
679 pages_per_cluster = 1 << (cbits - PAGE_CACHE_SHIFT);
681 return pages_per_cluster;
684 static inline unsigned int ocfs2_megabytes_to_clusters(struct super_block *sb,
687 BUILD_BUG_ON(OCFS2_MAX_CLUSTERSIZE > 1048576);
689 return megs << (20 - OCFS2_SB(sb)->s_clustersize_bits);
692 static inline void ocfs2_init_inode_steal_slot(struct ocfs2_super *osb)
694 spin_lock(&osb->osb_lock);
695 osb->s_inode_steal_slot = OCFS2_INVALID_SLOT;
696 spin_unlock(&osb->osb_lock);
697 atomic_set(&osb->s_num_inodes_stolen, 0);
700 static inline void ocfs2_set_inode_steal_slot(struct ocfs2_super *osb,
703 spin_lock(&osb->osb_lock);
704 osb->s_inode_steal_slot = slot;
705 spin_unlock(&osb->osb_lock);
708 static inline s16 ocfs2_get_inode_steal_slot(struct ocfs2_super *osb)
712 spin_lock(&osb->osb_lock);
713 slot = osb->s_inode_steal_slot;
714 spin_unlock(&osb->osb_lock);
719 #define ocfs2_set_bit ext2_set_bit
720 #define ocfs2_clear_bit ext2_clear_bit
721 #define ocfs2_test_bit ext2_test_bit
722 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit
723 #define ocfs2_find_next_bit ext2_find_next_bit