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) /* have we 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;
168 struct ocfs2_alloc_stats
172 atomic_t bitmap_data;
177 enum ocfs2_local_alloc_state
179 OCFS2_LA_UNUSED = 0, /* Local alloc will never be used for
180 * this mountpoint. */
181 OCFS2_LA_ENABLED, /* Local alloc is in use. */
182 OCFS2_LA_THROTTLED, /* Local alloc is in use, but number
183 * of bits has been reduced. */
184 OCFS2_LA_DISABLED /* Local alloc has temporarily been
188 enum ocfs2_mount_options
190 OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Heartbeat started in local mode */
191 OCFS2_MOUNT_BARRIER = 1 << 1, /* Use block barriers */
192 OCFS2_MOUNT_NOINTR = 1 << 2, /* Don't catch signals */
193 OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */
194 OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */
195 OCFS2_MOUNT_LOCALFLOCKS = 1 << 5, /* No cluster aware user file locks */
196 OCFS2_MOUNT_NOUSERXATTR = 1 << 6, /* No user xattr */
197 OCFS2_MOUNT_INODE64 = 1 << 7, /* Allow inode numbers > 2^32 */
200 #define OCFS2_OSB_SOFT_RO 0x0001
201 #define OCFS2_OSB_HARD_RO 0x0002
202 #define OCFS2_OSB_ERROR_FS 0x0004
203 #define OCFS2_DEFAULT_ATIME_QUANTUM 60
205 struct ocfs2_journal;
206 struct ocfs2_slot_info;
207 struct ocfs2_recovery_map;
210 struct task_struct *commit_task;
211 struct super_block *sb;
212 struct inode *root_inode;
213 struct inode *sys_root_inode;
214 struct inode *system_inodes[NUM_SYSTEM_INODES];
216 struct ocfs2_slot_info *slot_info;
218 u32 *slot_recovery_generations;
220 spinlock_t node_map_lock;
223 u64 system_dir_blkno;
230 u64 first_cluster_group_blkno;
233 u32 s_feature_compat;
234 u32 s_feature_incompat;
235 u32 s_feature_ro_compat;
237 /* Protects s_next_generation, osb_flags and s_inode_steal_slot.
238 * Could protect more on osb as it's very short lived.
241 u32 s_next_generation;
242 unsigned long osb_flags;
243 s16 s_inode_steal_slot;
244 atomic_t s_num_inodes_stolen;
246 unsigned long s_mount_opt;
247 unsigned int s_atime_quantum;
249 unsigned int max_slots;
250 unsigned int node_num;
255 int s_clustersize_bits;
256 unsigned int s_xattr_inline_size;
259 struct mutex recovery_lock;
260 struct ocfs2_recovery_map *recovery_map;
261 struct task_struct *recovery_thread_task;
262 int disable_recovery;
263 wait_queue_head_t checkpoint_event;
264 atomic_t needs_checkpoint;
265 struct ocfs2_journal *journal;
266 unsigned long osb_commit_interval;
268 struct delayed_work la_enable_wq;
271 * Must hold local alloc i_mutex and osb->osb_lock to change
272 * local_alloc_bits. Reads can be done under either lock.
274 unsigned int local_alloc_bits;
275 unsigned int local_alloc_default_bits;
277 enum ocfs2_local_alloc_state local_alloc_state; /* protected
280 struct buffer_head *local_alloc_bh;
284 #ifdef CONFIG_OCFS2_FS_STATS
285 struct dentry *local_alloc_debug;
286 char *local_alloc_debug_buf;
289 /* Next two fields are for local node slot recovery during
292 struct ocfs2_dinode *local_alloc_copy;
294 struct ocfs2_alloc_stats alloc_stats;
295 char dev_str[20]; /* "major,minor" of the device */
297 char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
298 struct ocfs2_cluster_connection *cconn;
299 struct ocfs2_lock_res osb_super_lockres;
300 struct ocfs2_lock_res osb_rename_lockres;
301 struct ocfs2_dlm_debug *osb_dlm_debug;
303 struct dentry *osb_debug_root;
305 wait_queue_head_t recovery_event;
307 spinlock_t dc_task_lock;
308 struct task_struct *dc_task;
309 wait_queue_head_t dc_event;
310 unsigned long dc_wake_sequence;
311 unsigned long dc_work_sequence;
314 * Any thread can add locks to the list, but the downconvert
315 * thread is the only one allowed to remove locks. Any change
316 * to this rule requires updating
317 * ocfs2_downconvert_thread_do_work().
319 struct list_head blocked_lock_list;
320 unsigned long blocked_lock_count;
322 wait_queue_head_t osb_mount_event;
324 /* Truncate log info */
325 struct inode *osb_tl_inode;
326 struct buffer_head *osb_tl_bh;
327 struct delayed_work osb_truncate_log_wq;
329 struct ocfs2_node_map osb_recovering_orphan_dirs;
330 unsigned int *osb_orphan_wipes;
331 wait_queue_head_t osb_wipe_event;
334 #define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info)
336 static inline int ocfs2_should_order_data(struct inode *inode)
338 if (!S_ISREG(inode->i_mode))
340 if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
345 static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb)
347 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC)
352 static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb)
355 * Support for sparse files is a pre-requisite
357 if (!ocfs2_sparse_alloc(osb))
360 if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN)
365 static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb)
367 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA)
372 static inline int ocfs2_supports_xattr(struct ocfs2_super *osb)
374 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR)
379 /* set / clear functions because cluster events can make these happen
380 * in parallel so we want the transitions to be atomic. this also
381 * means that any future flags osb_flags must be protected by spinlock
383 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
386 spin_lock(&osb->osb_lock);
387 osb->osb_flags |= flag;
388 spin_unlock(&osb->osb_lock);
391 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
394 spin_lock(&osb->osb_lock);
395 osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
397 osb->osb_flags |= OCFS2_OSB_HARD_RO;
399 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
400 spin_unlock(&osb->osb_lock);
403 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
407 spin_lock(&osb->osb_lock);
408 ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
409 spin_unlock(&osb->osb_lock);
414 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
418 spin_lock(&osb->osb_lock);
419 ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
420 spin_unlock(&osb->osb_lock);
425 static inline int ocfs2_userspace_stack(struct ocfs2_super *osb)
427 return (osb->s_feature_incompat &
428 OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK);
431 static inline int ocfs2_mount_local(struct ocfs2_super *osb)
433 return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT);
436 static inline int ocfs2_uses_extended_slot_map(struct ocfs2_super *osb)
438 return (osb->s_feature_incompat &
439 OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP);
443 #define OCFS2_IS_VALID_DINODE(ptr) \
444 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
446 #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di) do { \
447 typeof(__di) ____di = (__di); \
448 ocfs2_error((__sb), \
449 "Dinode # %llu has bad signature %.*s", \
450 (unsigned long long)le64_to_cpu((____di)->i_blkno), 7, \
451 (____di)->i_signature); \
454 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \
455 (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
457 #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb) do { \
458 typeof(__eb) ____eb = (__eb); \
459 ocfs2_error((__sb), \
460 "Extent Block # %llu has bad signature %.*s", \
461 (unsigned long long)le64_to_cpu((____eb)->h_blkno), 7, \
462 (____eb)->h_signature); \
465 #define OCFS2_IS_VALID_GROUP_DESC(ptr) \
466 (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
468 #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd) do { \
469 typeof(__gd) ____gd = (__gd); \
470 ocfs2_error((__sb), \
471 "Group Descriptor # %llu has bad signature %.*s", \
472 (unsigned long long)le64_to_cpu((____gd)->bg_blkno), 7, \
473 (____gd)->bg_signature); \
476 #define OCFS2_IS_VALID_XATTR_BLOCK(ptr) \
477 (!strcmp((ptr)->xb_signature, OCFS2_XATTR_BLOCK_SIGNATURE))
479 static inline unsigned long ino_from_blkno(struct super_block *sb,
482 return (unsigned long)(blkno & (u64)ULONG_MAX);
485 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
488 int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
489 sb->s_blocksize_bits;
491 return (u64)clusters << c_to_b_bits;
494 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
497 int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
498 sb->s_blocksize_bits;
500 return (u32)(blocks >> b_to_c_bits);
503 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
506 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
507 unsigned int clusters;
509 bytes += OCFS2_SB(sb)->s_clustersize - 1;
510 /* OCFS2 just cannot have enough clusters to overflow this */
511 clusters = (unsigned int)(bytes >> cl_bits);
516 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
519 bytes += sb->s_blocksize - 1;
520 return bytes >> sb->s_blocksize_bits;
523 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
526 return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
529 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
532 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
533 unsigned int clusters;
535 clusters = ocfs2_clusters_for_bytes(sb, bytes);
536 return (u64)clusters << cl_bits;
539 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
544 blocks = ocfs2_blocks_for_bytes(sb, bytes);
545 return blocks << sb->s_blocksize_bits;
548 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
550 return (unsigned long)((bytes + 511) >> 9);
553 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb,
554 unsigned long pg_index)
556 u32 clusters = pg_index;
557 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
559 if (unlikely(PAGE_CACHE_SHIFT > cbits))
560 clusters = pg_index << (PAGE_CACHE_SHIFT - cbits);
561 else if (PAGE_CACHE_SHIFT < cbits)
562 clusters = pg_index >> (cbits - PAGE_CACHE_SHIFT);
568 * Find the 1st page index which covers the given clusters.
570 static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb,
573 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
574 pgoff_t index = clusters;
576 if (PAGE_CACHE_SHIFT > cbits) {
577 index = (pgoff_t)clusters >> (PAGE_CACHE_SHIFT - cbits);
578 } else if (PAGE_CACHE_SHIFT < cbits) {
579 index = (pgoff_t)clusters << (cbits - PAGE_CACHE_SHIFT);
585 static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb)
587 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
588 unsigned int pages_per_cluster = 1;
590 if (PAGE_CACHE_SHIFT < cbits)
591 pages_per_cluster = 1 << (cbits - PAGE_CACHE_SHIFT);
593 return pages_per_cluster;
596 static inline unsigned int ocfs2_megabytes_to_clusters(struct super_block *sb,
599 BUILD_BUG_ON(OCFS2_MAX_CLUSTERSIZE > 1048576);
601 return megs << (20 - OCFS2_SB(sb)->s_clustersize_bits);
604 static inline void ocfs2_init_inode_steal_slot(struct ocfs2_super *osb)
606 spin_lock(&osb->osb_lock);
607 osb->s_inode_steal_slot = OCFS2_INVALID_SLOT;
608 spin_unlock(&osb->osb_lock);
609 atomic_set(&osb->s_num_inodes_stolen, 0);
612 static inline void ocfs2_set_inode_steal_slot(struct ocfs2_super *osb,
615 spin_lock(&osb->osb_lock);
616 osb->s_inode_steal_slot = slot;
617 spin_unlock(&osb->osb_lock);
620 static inline s16 ocfs2_get_inode_steal_slot(struct ocfs2_super *osb)
624 spin_lock(&osb->osb_lock);
625 slot = osb->s_inode_steal_slot;
626 spin_unlock(&osb->osb_lock);
631 #define ocfs2_set_bit ext2_set_bit
632 #define ocfs2_clear_bit ext2_clear_bit
633 #define ocfs2_test_bit ext2_test_bit
634 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit