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 #include <linux/jbd.h>
39 /* For union ocfs2_dlm_lksb */
40 #include "stackglue.h"
43 #include "ocfs2_lockid.h"
45 /* Most user visible OCFS2 inodes will have very few pieces of
46 * metadata, but larger files (including bitmaps, etc) must be taken
47 * into account when designing an access scheme. We allow a small
48 * amount of inlined blocks to be stored on an array and grow the
49 * structure into a rb tree when necessary. */
50 #define OCFS2_INODE_MAX_CACHE_ARRAY 2
52 struct ocfs2_caching_info {
53 unsigned int ci_num_cached;
55 sector_t ci_array[OCFS2_INODE_MAX_CACHE_ARRAY];
56 struct rb_root ci_tree;
60 /* this limits us to 256 nodes
61 * if we need more, we can do a kmalloc for the map */
62 #define OCFS2_NODE_MAP_MAX_NODES 256
63 struct ocfs2_node_map {
65 unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)];
68 enum ocfs2_ast_action {
69 OCFS2_AST_INVALID = 0,
72 OCFS2_AST_DOWNCONVERT,
75 /* actions for an unlockast function to take. */
76 enum ocfs2_unlock_action {
77 OCFS2_UNLOCK_INVALID = 0,
78 OCFS2_UNLOCK_CANCEL_CONVERT,
79 OCFS2_UNLOCK_DROP_LOCK,
82 /* ocfs2_lock_res->l_flags flags. */
83 #define OCFS2_LOCK_ATTACHED (0x00000001) /* have we initialized
85 #define OCFS2_LOCK_BUSY (0x00000002) /* we are currently in
87 #define OCFS2_LOCK_BLOCKED (0x00000004) /* blocked waiting to
89 #define OCFS2_LOCK_LOCAL (0x00000008) /* newly created inode */
90 #define OCFS2_LOCK_NEEDS_REFRESH (0x00000010)
91 #define OCFS2_LOCK_REFRESHING (0x00000020)
92 #define OCFS2_LOCK_INITIALIZED (0x00000040) /* track initialization
93 * for shutdown paths */
94 #define OCFS2_LOCK_FREEING (0x00000080) /* help dlmglue track
95 * when to skip queueing
99 #define OCFS2_LOCK_QUEUED (0x00000100) /* queued for downconvert */
100 #define OCFS2_LOCK_NOCACHE (0x00000200) /* don't use a holder count */
101 #define OCFS2_LOCK_PENDING (0x00000400) /* This lockres is pending a
102 call to dlm_lock. Only
103 exists with BUSY set. */
105 struct ocfs2_lock_res_ops;
107 typedef void (*ocfs2_lock_callback)(int status, unsigned long data);
109 struct ocfs2_lock_res {
111 struct ocfs2_lock_res_ops *l_ops;
114 struct list_head l_blocked_list;
115 struct list_head l_mask_waiters;
117 enum ocfs2_lock_type l_type;
118 unsigned long l_flags;
119 char l_name[OCFS2_LOCK_ID_MAX_LEN];
121 unsigned int l_ro_holders;
122 unsigned int l_ex_holders;
123 union ocfs2_dlm_lksb l_lksb;
125 /* used from AST/BAST funcs. */
126 enum ocfs2_ast_action l_action;
127 enum ocfs2_unlock_action l_unlock_action;
130 unsigned int l_pending_gen;
132 wait_queue_head_t l_event;
134 struct list_head l_debug_list;
136 #ifdef CONFIG_OCFS2_FS_STATS
137 unsigned long long l_lock_num_prmode; /* PR acquires */
138 unsigned long long l_lock_num_exmode; /* EX acquires */
139 unsigned int l_lock_num_prmode_failed; /* Failed PR gets */
140 unsigned int l_lock_num_exmode_failed; /* Failed EX gets */
141 unsigned long long l_lock_total_prmode; /* Tot wait for PR */
142 unsigned long long l_lock_total_exmode; /* Tot wait for EX */
143 unsigned int l_lock_max_prmode; /* Max wait for PR */
144 unsigned int l_lock_max_exmode; /* Max wait for EX */
145 unsigned int l_lock_refresh; /* Disk refreshes */
149 struct ocfs2_dlm_debug {
150 struct kref d_refcnt;
151 struct dentry *d_locking_state;
152 struct list_head d_lockres_tracking;
163 struct ocfs2_alloc_stats
167 atomic_t bitmap_data;
172 enum ocfs2_local_alloc_state
174 OCFS2_LA_UNUSED = 0, /* Local alloc will never be used for
175 * this mountpoint. */
176 OCFS2_LA_ENABLED, /* Local alloc is in use. */
177 OCFS2_LA_THROTTLED, /* Local alloc is in use, but number
178 * of bits has been reduced. */
179 OCFS2_LA_DISABLED /* Local alloc has temporarily been
183 enum ocfs2_mount_options
185 OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Heartbeat started in local mode */
186 OCFS2_MOUNT_BARRIER = 1 << 1, /* Use block barriers */
187 OCFS2_MOUNT_NOINTR = 1 << 2, /* Don't catch signals */
188 OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */
189 OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */
190 OCFS2_MOUNT_LOCALFLOCKS = 1 << 5, /* No cluster aware user file locks */
193 #define OCFS2_OSB_SOFT_RO 0x0001
194 #define OCFS2_OSB_HARD_RO 0x0002
195 #define OCFS2_OSB_ERROR_FS 0x0004
196 #define OCFS2_DEFAULT_ATIME_QUANTUM 60
198 struct ocfs2_journal;
199 struct ocfs2_slot_info;
200 struct ocfs2_recovery_map;
203 struct task_struct *commit_task;
204 struct super_block *sb;
205 struct inode *root_inode;
206 struct inode *sys_root_inode;
207 struct inode *system_inodes[NUM_SYSTEM_INODES];
209 struct ocfs2_slot_info *slot_info;
211 u32 *slot_recovery_generations;
213 spinlock_t node_map_lock;
216 u64 system_dir_blkno;
222 u64 first_cluster_group_blkno;
225 u32 s_feature_compat;
226 u32 s_feature_incompat;
227 u32 s_feature_ro_compat;
229 /* Protects s_next_generation, osb_flags and s_inode_steal_slot.
230 * Could protect more on osb as it's very short lived.
233 u32 s_next_generation;
234 unsigned long osb_flags;
235 s16 s_inode_steal_slot;
236 atomic_t s_num_inodes_stolen;
238 unsigned long s_mount_opt;
239 unsigned int s_atime_quantum;
241 unsigned int max_slots;
242 unsigned int node_num;
247 int s_clustersize_bits;
250 struct mutex recovery_lock;
251 struct ocfs2_recovery_map *recovery_map;
252 struct task_struct *recovery_thread_task;
253 int disable_recovery;
254 wait_queue_head_t checkpoint_event;
255 atomic_t needs_checkpoint;
256 struct ocfs2_journal *journal;
257 unsigned long osb_commit_interval;
259 struct delayed_work la_enable_wq;
262 * Must hold local alloc i_mutex and osb->osb_lock to change
263 * local_alloc_bits. Reads can be done under either lock.
265 unsigned int local_alloc_bits;
266 unsigned int local_alloc_default_bits;
268 enum ocfs2_local_alloc_state local_alloc_state; /* protected
271 struct buffer_head *local_alloc_bh;
275 #ifdef CONFIG_OCFS2_FS_STATS
276 struct dentry *local_alloc_debug;
277 char *local_alloc_debug_buf;
280 /* Next two fields are for local node slot recovery during
283 struct ocfs2_dinode *local_alloc_copy;
285 struct ocfs2_alloc_stats alloc_stats;
286 char dev_str[20]; /* "major,minor" of the device */
288 char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
289 struct ocfs2_cluster_connection *cconn;
290 struct ocfs2_lock_res osb_super_lockres;
291 struct ocfs2_lock_res osb_rename_lockres;
292 struct ocfs2_dlm_debug *osb_dlm_debug;
294 struct dentry *osb_debug_root;
296 wait_queue_head_t recovery_event;
298 spinlock_t dc_task_lock;
299 struct task_struct *dc_task;
300 wait_queue_head_t dc_event;
301 unsigned long dc_wake_sequence;
302 unsigned long dc_work_sequence;
305 * Any thread can add locks to the list, but the downconvert
306 * thread is the only one allowed to remove locks. Any change
307 * to this rule requires updating
308 * ocfs2_downconvert_thread_do_work().
310 struct list_head blocked_lock_list;
311 unsigned long blocked_lock_count;
313 wait_queue_head_t osb_mount_event;
315 /* Truncate log info */
316 struct inode *osb_tl_inode;
317 struct buffer_head *osb_tl_bh;
318 struct delayed_work osb_truncate_log_wq;
320 struct ocfs2_node_map osb_recovering_orphan_dirs;
321 unsigned int *osb_orphan_wipes;
322 wait_queue_head_t osb_wipe_event;
325 #define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info)
327 static inline int ocfs2_should_order_data(struct inode *inode)
329 if (!S_ISREG(inode->i_mode))
331 if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
336 static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb)
338 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC)
343 static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb)
346 * Support for sparse files is a pre-requisite
348 if (!ocfs2_sparse_alloc(osb))
351 if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN)
356 static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb)
358 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA)
363 /* set / clear functions because cluster events can make these happen
364 * in parallel so we want the transitions to be atomic. this also
365 * means that any future flags osb_flags must be protected by spinlock
367 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
370 spin_lock(&osb->osb_lock);
371 osb->osb_flags |= flag;
372 spin_unlock(&osb->osb_lock);
375 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
378 spin_lock(&osb->osb_lock);
379 osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
381 osb->osb_flags |= OCFS2_OSB_HARD_RO;
383 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
384 spin_unlock(&osb->osb_lock);
387 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
391 spin_lock(&osb->osb_lock);
392 ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
393 spin_unlock(&osb->osb_lock);
398 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
402 spin_lock(&osb->osb_lock);
403 ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
404 spin_unlock(&osb->osb_lock);
409 static inline int ocfs2_userspace_stack(struct ocfs2_super *osb)
411 return (osb->s_feature_incompat &
412 OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK);
415 static inline int ocfs2_mount_local(struct ocfs2_super *osb)
417 return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT);
420 static inline int ocfs2_uses_extended_slot_map(struct ocfs2_super *osb)
422 return (osb->s_feature_incompat &
423 OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP);
427 #define OCFS2_IS_VALID_DINODE(ptr) \
428 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
430 #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di) do { \
431 typeof(__di) ____di = (__di); \
432 ocfs2_error((__sb), \
433 "Dinode # %llu has bad signature %.*s", \
434 (unsigned long long)le64_to_cpu((____di)->i_blkno), 7, \
435 (____di)->i_signature); \
438 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \
439 (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
441 #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb) do { \
442 typeof(__eb) ____eb = (__eb); \
443 ocfs2_error((__sb), \
444 "Extent Block # %llu has bad signature %.*s", \
445 (unsigned long long)le64_to_cpu((____eb)->h_blkno), 7, \
446 (____eb)->h_signature); \
449 #define OCFS2_IS_VALID_GROUP_DESC(ptr) \
450 (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
452 #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd) do { \
453 typeof(__gd) ____gd = (__gd); \
454 ocfs2_error((__sb), \
455 "Group Descriptor # %llu has bad signature %.*s", \
456 (unsigned long long)le64_to_cpu((____gd)->bg_blkno), 7, \
457 (____gd)->bg_signature); \
460 static inline unsigned long ino_from_blkno(struct super_block *sb,
463 return (unsigned long)(blkno & (u64)ULONG_MAX);
466 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
469 int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
470 sb->s_blocksize_bits;
472 return (u64)clusters << c_to_b_bits;
475 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
478 int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
479 sb->s_blocksize_bits;
481 return (u32)(blocks >> b_to_c_bits);
484 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
487 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
488 unsigned int clusters;
490 bytes += OCFS2_SB(sb)->s_clustersize - 1;
491 /* OCFS2 just cannot have enough clusters to overflow this */
492 clusters = (unsigned int)(bytes >> cl_bits);
497 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
500 bytes += sb->s_blocksize - 1;
501 return bytes >> sb->s_blocksize_bits;
504 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
507 return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
510 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
513 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
514 unsigned int clusters;
516 clusters = ocfs2_clusters_for_bytes(sb, bytes);
517 return (u64)clusters << cl_bits;
520 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
525 blocks = ocfs2_blocks_for_bytes(sb, bytes);
526 return blocks << sb->s_blocksize_bits;
529 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
531 return (unsigned long)((bytes + 511) >> 9);
534 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb,
535 unsigned long pg_index)
537 u32 clusters = pg_index;
538 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
540 if (unlikely(PAGE_CACHE_SHIFT > cbits))
541 clusters = pg_index << (PAGE_CACHE_SHIFT - cbits);
542 else if (PAGE_CACHE_SHIFT < cbits)
543 clusters = pg_index >> (cbits - PAGE_CACHE_SHIFT);
549 * Find the 1st page index which covers the given clusters.
551 static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb,
554 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
555 pgoff_t index = clusters;
557 if (PAGE_CACHE_SHIFT > cbits) {
558 index = (pgoff_t)clusters >> (PAGE_CACHE_SHIFT - cbits);
559 } else if (PAGE_CACHE_SHIFT < cbits) {
560 index = (pgoff_t)clusters << (cbits - PAGE_CACHE_SHIFT);
566 static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb)
568 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
569 unsigned int pages_per_cluster = 1;
571 if (PAGE_CACHE_SHIFT < cbits)
572 pages_per_cluster = 1 << (cbits - PAGE_CACHE_SHIFT);
574 return pages_per_cluster;
577 static inline unsigned int ocfs2_megabytes_to_clusters(struct super_block *sb,
580 BUILD_BUG_ON(OCFS2_MAX_CLUSTERSIZE > 1048576);
582 return megs << (20 - OCFS2_SB(sb)->s_clustersize_bits);
585 static inline void ocfs2_init_inode_steal_slot(struct ocfs2_super *osb)
587 spin_lock(&osb->osb_lock);
588 osb->s_inode_steal_slot = OCFS2_INVALID_SLOT;
589 spin_unlock(&osb->osb_lock);
590 atomic_set(&osb->s_num_inodes_stolen, 0);
593 static inline void ocfs2_set_inode_steal_slot(struct ocfs2_super *osb,
596 spin_lock(&osb->osb_lock);
597 osb->s_inode_steal_slot = slot;
598 spin_unlock(&osb->osb_lock);
601 static inline s16 ocfs2_get_inode_steal_slot(struct ocfs2_super *osb)
605 spin_lock(&osb->osb_lock);
606 slot = osb->s_inode_steal_slot;
607 spin_unlock(&osb->osb_lock);
612 #define ocfs2_set_bit ext2_set_bit
613 #define ocfs2_clear_bit ext2_clear_bit
614 #define ocfs2_test_bit ext2_test_bit
615 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit