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
;
248 unsigned int s_xattr_inline_size
;
251 struct mutex recovery_lock
;
252 struct ocfs2_recovery_map
*recovery_map
;
253 struct task_struct
*recovery_thread_task
;
254 int disable_recovery
;
255 wait_queue_head_t checkpoint_event
;
256 atomic_t needs_checkpoint
;
257 struct ocfs2_journal
*journal
;
258 unsigned long osb_commit_interval
;
260 struct delayed_work la_enable_wq
;
263 * Must hold local alloc i_mutex and osb->osb_lock to change
264 * local_alloc_bits. Reads can be done under either lock.
266 unsigned int local_alloc_bits
;
267 unsigned int local_alloc_default_bits
;
269 enum ocfs2_local_alloc_state local_alloc_state
; /* protected
272 struct buffer_head
*local_alloc_bh
;
276 #ifdef CONFIG_OCFS2_FS_STATS
277 struct dentry
*local_alloc_debug
;
278 char *local_alloc_debug_buf
;
281 /* Next two fields are for local node slot recovery during
284 struct ocfs2_dinode
*local_alloc_copy
;
286 struct ocfs2_alloc_stats alloc_stats
;
287 char dev_str
[20]; /* "major,minor" of the device */
289 char osb_cluster_stack
[OCFS2_STACK_LABEL_LEN
+ 1];
290 struct ocfs2_cluster_connection
*cconn
;
291 struct ocfs2_lock_res osb_super_lockres
;
292 struct ocfs2_lock_res osb_rename_lockres
;
293 struct ocfs2_dlm_debug
*osb_dlm_debug
;
295 struct dentry
*osb_debug_root
;
297 wait_queue_head_t recovery_event
;
299 spinlock_t dc_task_lock
;
300 struct task_struct
*dc_task
;
301 wait_queue_head_t dc_event
;
302 unsigned long dc_wake_sequence
;
303 unsigned long dc_work_sequence
;
306 * Any thread can add locks to the list, but the downconvert
307 * thread is the only one allowed to remove locks. Any change
308 * to this rule requires updating
309 * ocfs2_downconvert_thread_do_work().
311 struct list_head blocked_lock_list
;
312 unsigned long blocked_lock_count
;
314 wait_queue_head_t osb_mount_event
;
316 /* Truncate log info */
317 struct inode
*osb_tl_inode
;
318 struct buffer_head
*osb_tl_bh
;
319 struct delayed_work osb_truncate_log_wq
;
321 struct ocfs2_node_map osb_recovering_orphan_dirs
;
322 unsigned int *osb_orphan_wipes
;
323 wait_queue_head_t osb_wipe_event
;
326 #define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info)
328 static inline int ocfs2_should_order_data(struct inode
*inode
)
330 if (!S_ISREG(inode
->i_mode
))
332 if (OCFS2_SB(inode
->i_sb
)->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
)
337 static inline int ocfs2_sparse_alloc(struct ocfs2_super
*osb
)
339 if (osb
->s_feature_incompat
& OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC
)
344 static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super
*osb
)
347 * Support for sparse files is a pre-requisite
349 if (!ocfs2_sparse_alloc(osb
))
352 if (osb
->s_feature_ro_compat
& OCFS2_FEATURE_RO_COMPAT_UNWRITTEN
)
357 static inline int ocfs2_supports_inline_data(struct ocfs2_super
*osb
)
359 if (osb
->s_feature_incompat
& OCFS2_FEATURE_INCOMPAT_INLINE_DATA
)
364 /* set / clear functions because cluster events can make these happen
365 * in parallel so we want the transitions to be atomic. this also
366 * means that any future flags osb_flags must be protected by spinlock
368 static inline void ocfs2_set_osb_flag(struct ocfs2_super
*osb
,
371 spin_lock(&osb
->osb_lock
);
372 osb
->osb_flags
|= flag
;
373 spin_unlock(&osb
->osb_lock
);
376 static inline void ocfs2_set_ro_flag(struct ocfs2_super
*osb
,
379 spin_lock(&osb
->osb_lock
);
380 osb
->osb_flags
&= ~(OCFS2_OSB_SOFT_RO
|OCFS2_OSB_HARD_RO
);
382 osb
->osb_flags
|= OCFS2_OSB_HARD_RO
;
384 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
385 spin_unlock(&osb
->osb_lock
);
388 static inline int ocfs2_is_hard_readonly(struct ocfs2_super
*osb
)
392 spin_lock(&osb
->osb_lock
);
393 ret
= osb
->osb_flags
& OCFS2_OSB_HARD_RO
;
394 spin_unlock(&osb
->osb_lock
);
399 static inline int ocfs2_is_soft_readonly(struct ocfs2_super
*osb
)
403 spin_lock(&osb
->osb_lock
);
404 ret
= osb
->osb_flags
& OCFS2_OSB_SOFT_RO
;
405 spin_unlock(&osb
->osb_lock
);
410 static inline int ocfs2_userspace_stack(struct ocfs2_super
*osb
)
412 return (osb
->s_feature_incompat
&
413 OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK
);
416 static inline int ocfs2_mount_local(struct ocfs2_super
*osb
)
418 return (osb
->s_feature_incompat
& OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT
);
421 static inline int ocfs2_uses_extended_slot_map(struct ocfs2_super
*osb
)
423 return (osb
->s_feature_incompat
&
424 OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP
);
428 #define OCFS2_IS_VALID_DINODE(ptr) \
429 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
431 #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di) do { \
432 typeof(__di) ____di = (__di); \
433 ocfs2_error((__sb), \
434 "Dinode # %llu has bad signature %.*s", \
435 (unsigned long long)le64_to_cpu((____di)->i_blkno), 7, \
436 (____di)->i_signature); \
439 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \
440 (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
442 #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb) do { \
443 typeof(__eb) ____eb = (__eb); \
444 ocfs2_error((__sb), \
445 "Extent Block # %llu has bad signature %.*s", \
446 (unsigned long long)le64_to_cpu((____eb)->h_blkno), 7, \
447 (____eb)->h_signature); \
450 #define OCFS2_IS_VALID_GROUP_DESC(ptr) \
451 (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
453 #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd) do { \
454 typeof(__gd) ____gd = (__gd); \
455 ocfs2_error((__sb), \
456 "Group Descriptor # %llu has bad signature %.*s", \
457 (unsigned long long)le64_to_cpu((____gd)->bg_blkno), 7, \
458 (____gd)->bg_signature); \
461 static inline unsigned long ino_from_blkno(struct super_block
*sb
,
464 return (unsigned long)(blkno
& (u64
)ULONG_MAX
);
467 static inline u64
ocfs2_clusters_to_blocks(struct super_block
*sb
,
470 int c_to_b_bits
= OCFS2_SB(sb
)->s_clustersize_bits
-
471 sb
->s_blocksize_bits
;
473 return (u64
)clusters
<< c_to_b_bits
;
476 static inline u32
ocfs2_blocks_to_clusters(struct super_block
*sb
,
479 int b_to_c_bits
= OCFS2_SB(sb
)->s_clustersize_bits
-
480 sb
->s_blocksize_bits
;
482 return (u32
)(blocks
>> b_to_c_bits
);
485 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block
*sb
,
488 int cl_bits
= OCFS2_SB(sb
)->s_clustersize_bits
;
489 unsigned int clusters
;
491 bytes
+= OCFS2_SB(sb
)->s_clustersize
- 1;
492 /* OCFS2 just cannot have enough clusters to overflow this */
493 clusters
= (unsigned int)(bytes
>> cl_bits
);
498 static inline u64
ocfs2_blocks_for_bytes(struct super_block
*sb
,
501 bytes
+= sb
->s_blocksize
- 1;
502 return bytes
>> sb
->s_blocksize_bits
;
505 static inline u64
ocfs2_clusters_to_bytes(struct super_block
*sb
,
508 return (u64
)clusters
<< OCFS2_SB(sb
)->s_clustersize_bits
;
511 static inline u64
ocfs2_align_bytes_to_clusters(struct super_block
*sb
,
514 int cl_bits
= OCFS2_SB(sb
)->s_clustersize_bits
;
515 unsigned int clusters
;
517 clusters
= ocfs2_clusters_for_bytes(sb
, bytes
);
518 return (u64
)clusters
<< cl_bits
;
521 static inline u64
ocfs2_align_bytes_to_blocks(struct super_block
*sb
,
526 blocks
= ocfs2_blocks_for_bytes(sb
, bytes
);
527 return blocks
<< sb
->s_blocksize_bits
;
530 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes
)
532 return (unsigned long)((bytes
+ 511) >> 9);
535 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block
*sb
,
536 unsigned long pg_index
)
538 u32 clusters
= pg_index
;
539 unsigned int cbits
= OCFS2_SB(sb
)->s_clustersize_bits
;
541 if (unlikely(PAGE_CACHE_SHIFT
> cbits
))
542 clusters
= pg_index
<< (PAGE_CACHE_SHIFT
- cbits
);
543 else if (PAGE_CACHE_SHIFT
< cbits
)
544 clusters
= pg_index
>> (cbits
- PAGE_CACHE_SHIFT
);
550 * Find the 1st page index which covers the given clusters.
552 static inline pgoff_t
ocfs2_align_clusters_to_page_index(struct super_block
*sb
,
555 unsigned int cbits
= OCFS2_SB(sb
)->s_clustersize_bits
;
556 pgoff_t index
= clusters
;
558 if (PAGE_CACHE_SHIFT
> cbits
) {
559 index
= (pgoff_t
)clusters
>> (PAGE_CACHE_SHIFT
- cbits
);
560 } else if (PAGE_CACHE_SHIFT
< cbits
) {
561 index
= (pgoff_t
)clusters
<< (cbits
- PAGE_CACHE_SHIFT
);
567 static inline unsigned int ocfs2_pages_per_cluster(struct super_block
*sb
)
569 unsigned int cbits
= OCFS2_SB(sb
)->s_clustersize_bits
;
570 unsigned int pages_per_cluster
= 1;
572 if (PAGE_CACHE_SHIFT
< cbits
)
573 pages_per_cluster
= 1 << (cbits
- PAGE_CACHE_SHIFT
);
575 return pages_per_cluster
;
578 static inline unsigned int ocfs2_megabytes_to_clusters(struct super_block
*sb
,
581 BUILD_BUG_ON(OCFS2_MAX_CLUSTERSIZE
> 1048576);
583 return megs
<< (20 - OCFS2_SB(sb
)->s_clustersize_bits
);
586 static inline void ocfs2_init_inode_steal_slot(struct ocfs2_super
*osb
)
588 spin_lock(&osb
->osb_lock
);
589 osb
->s_inode_steal_slot
= OCFS2_INVALID_SLOT
;
590 spin_unlock(&osb
->osb_lock
);
591 atomic_set(&osb
->s_num_inodes_stolen
, 0);
594 static inline void ocfs2_set_inode_steal_slot(struct ocfs2_super
*osb
,
597 spin_lock(&osb
->osb_lock
);
598 osb
->s_inode_steal_slot
= slot
;
599 spin_unlock(&osb
->osb_lock
);
602 static inline s16
ocfs2_get_inode_steal_slot(struct ocfs2_super
*osb
)
606 spin_lock(&osb
->osb_lock
);
607 slot
= osb
->s_inode_steal_slot
;
608 spin_unlock(&osb
->osb_lock
);
613 #define ocfs2_set_bit ext2_set_bit
614 #define ocfs2_clear_bit ext2_clear_bit
615 #define ocfs2_test_bit ext2_test_bit
616 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit