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 #include "cluster/nodemanager.h"
40 #include "cluster/heartbeat.h"
41 #include "cluster/tcp.h"
43 #include "dlm/dlmapi.h"
47 #include "ocfs2_lockid.h"
49 /* Most user visible OCFS2 inodes will have very few pieces of
50 * metadata, but larger files (including bitmaps, etc) must be taken
51 * into account when designing an access scheme. We allow a small
52 * amount of inlined blocks to be stored on an array and grow the
53 * structure into a rb tree when necessary. */
54 #define OCFS2_INODE_MAX_CACHE_ARRAY 2
56 struct ocfs2_caching_info
{
57 unsigned int ci_num_cached
;
59 sector_t ci_array
[OCFS2_INODE_MAX_CACHE_ARRAY
];
60 struct rb_root ci_tree
;
64 /* this limits us to 256 nodes
65 * if we need more, we can do a kmalloc for the map */
66 #define OCFS2_NODE_MAP_MAX_NODES 256
67 struct ocfs2_node_map
{
69 unsigned long map
[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES
)];
72 enum ocfs2_ast_action
{
73 OCFS2_AST_INVALID
= 0,
76 OCFS2_AST_DOWNCONVERT
,
79 /* actions for an unlockast function to take. */
80 enum ocfs2_unlock_action
{
81 OCFS2_UNLOCK_INVALID
= 0,
82 OCFS2_UNLOCK_CANCEL_CONVERT
,
83 OCFS2_UNLOCK_DROP_LOCK
,
86 /* ocfs2_lock_res->l_flags flags. */
87 #define OCFS2_LOCK_ATTACHED (0x00000001) /* have we initialized
89 #define OCFS2_LOCK_BUSY (0x00000002) /* we are currently in
91 #define OCFS2_LOCK_BLOCKED (0x00000004) /* blocked waiting to
93 #define OCFS2_LOCK_LOCAL (0x00000008) /* newly created inode */
94 #define OCFS2_LOCK_NEEDS_REFRESH (0x00000010)
95 #define OCFS2_LOCK_REFRESHING (0x00000020)
96 #define OCFS2_LOCK_INITIALIZED (0x00000040) /* track initialization
97 * for shutdown paths */
98 #define OCFS2_LOCK_FREEING (0x00000080) /* help dlmglue track
99 * when to skip queueing
100 * a lock because it's
103 #define OCFS2_LOCK_QUEUED (0x00000100) /* queued for downconvert */
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 struct dlm_lockstatus l_lksb
;
125 /* used from AST/BAST funcs. */
126 enum ocfs2_ast_action l_action
;
127 enum ocfs2_unlock_action l_unlock_action
;
131 wait_queue_head_t l_event
;
133 struct list_head l_debug_list
;
136 struct ocfs2_dlm_debug
{
137 struct kref d_refcnt
;
138 struct dentry
*d_locking_state
;
139 struct list_head d_lockres_tracking
;
150 struct ocfs2_alloc_stats
154 atomic_t bitmap_data
;
159 enum ocfs2_local_alloc_state
166 enum ocfs2_mount_options
168 OCFS2_MOUNT_HB_LOCAL
= 1 << 0, /* Heartbeat started in local mode */
169 OCFS2_MOUNT_BARRIER
= 1 << 1, /* Use block barriers */
170 OCFS2_MOUNT_NOINTR
= 1 << 2, /* Don't catch signals */
171 OCFS2_MOUNT_ERRORS_PANIC
= 1 << 3, /* Panic on errors */
172 OCFS2_MOUNT_DATA_WRITEBACK
= 1 << 4, /* No data ordering */
175 #define OCFS2_OSB_SOFT_RO 0x0001
176 #define OCFS2_OSB_HARD_RO 0x0002
177 #define OCFS2_OSB_ERROR_FS 0x0004
178 #define OCFS2_DEFAULT_ATIME_QUANTUM 60
180 struct ocfs2_journal
;
183 struct task_struct
*commit_task
;
184 struct super_block
*sb
;
185 struct inode
*root_inode
;
186 struct inode
*sys_root_inode
;
187 struct inode
*system_inodes
[NUM_SYSTEM_INODES
];
189 struct ocfs2_slot_info
*slot_info
;
191 spinlock_t node_map_lock
;
192 struct ocfs2_node_map recovery_map
;
195 u64 system_dir_blkno
;
201 u64 first_cluster_group_blkno
;
204 u32 s_feature_compat
;
205 u32 s_feature_incompat
;
206 u32 s_feature_ro_compat
;
208 /* Protects s_next_generaion, osb_flags. Could protect more on
209 * osb as it's very short lived. */
211 u32 s_next_generation
;
212 unsigned long osb_flags
;
214 unsigned long s_mount_opt
;
215 unsigned int s_atime_quantum
;
223 int s_clustersize_bits
;
226 struct mutex recovery_lock
;
227 struct task_struct
*recovery_thread_task
;
228 int disable_recovery
;
229 wait_queue_head_t checkpoint_event
;
230 atomic_t needs_checkpoint
;
231 struct ocfs2_journal
*journal
;
232 unsigned long osb_commit_interval
;
234 int local_alloc_size
;
235 enum ocfs2_local_alloc_state local_alloc_state
;
236 struct buffer_head
*local_alloc_bh
;
239 /* Next two fields are for local node slot recovery during
242 struct ocfs2_dinode
*local_alloc_copy
;
244 struct ocfs2_alloc_stats alloc_stats
;
245 char dev_str
[20]; /* "major,minor" of the device */
247 struct dlm_ctxt
*dlm
;
248 struct ocfs2_lock_res osb_super_lockres
;
249 struct ocfs2_lock_res osb_rename_lockres
;
250 struct dlm_eviction_cb osb_eviction_cb
;
251 struct ocfs2_dlm_debug
*osb_dlm_debug
;
253 struct dentry
*osb_debug_root
;
255 wait_queue_head_t recovery_event
;
257 spinlock_t dc_task_lock
;
258 struct task_struct
*dc_task
;
259 wait_queue_head_t dc_event
;
260 unsigned long dc_wake_sequence
;
261 unsigned long dc_work_sequence
;
263 struct list_head blocked_lock_list
;
264 unsigned long blocked_lock_count
;
266 wait_queue_head_t osb_mount_event
;
268 /* Truncate log info */
269 struct inode
*osb_tl_inode
;
270 struct buffer_head
*osb_tl_bh
;
271 struct delayed_work osb_truncate_log_wq
;
273 struct ocfs2_node_map osb_recovering_orphan_dirs
;
274 unsigned int *osb_orphan_wipes
;
275 wait_queue_head_t osb_wipe_event
;
278 #define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info)
280 static inline int ocfs2_should_order_data(struct inode
*inode
)
282 if (!S_ISREG(inode
->i_mode
))
284 if (OCFS2_SB(inode
->i_sb
)->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
)
289 static inline int ocfs2_sparse_alloc(struct ocfs2_super
*osb
)
291 if (osb
->s_feature_incompat
& OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC
)
296 static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super
*osb
)
299 * Support for sparse files is a pre-requisite
301 if (!ocfs2_sparse_alloc(osb
))
304 if (osb
->s_feature_ro_compat
& OCFS2_FEATURE_RO_COMPAT_UNWRITTEN
)
309 static inline int ocfs2_supports_inline_data(struct ocfs2_super
*osb
)
311 if (osb
->s_feature_incompat
& OCFS2_FEATURE_INCOMPAT_INLINE_DATA
)
316 /* set / clear functions because cluster events can make these happen
317 * in parallel so we want the transitions to be atomic. this also
318 * means that any future flags osb_flags must be protected by spinlock
320 static inline void ocfs2_set_osb_flag(struct ocfs2_super
*osb
,
323 spin_lock(&osb
->osb_lock
);
324 osb
->osb_flags
|= flag
;
325 spin_unlock(&osb
->osb_lock
);
328 static inline void ocfs2_set_ro_flag(struct ocfs2_super
*osb
,
331 spin_lock(&osb
->osb_lock
);
332 osb
->osb_flags
&= ~(OCFS2_OSB_SOFT_RO
|OCFS2_OSB_HARD_RO
);
334 osb
->osb_flags
|= OCFS2_OSB_HARD_RO
;
336 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
337 spin_unlock(&osb
->osb_lock
);
340 static inline int ocfs2_is_hard_readonly(struct ocfs2_super
*osb
)
344 spin_lock(&osb
->osb_lock
);
345 ret
= osb
->osb_flags
& OCFS2_OSB_HARD_RO
;
346 spin_unlock(&osb
->osb_lock
);
351 static inline int ocfs2_is_soft_readonly(struct ocfs2_super
*osb
)
355 spin_lock(&osb
->osb_lock
);
356 ret
= osb
->osb_flags
& OCFS2_OSB_SOFT_RO
;
357 spin_unlock(&osb
->osb_lock
);
362 static inline int ocfs2_mount_local(struct ocfs2_super
*osb
)
364 return (osb
->s_feature_incompat
& OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT
);
367 #define OCFS2_IS_VALID_DINODE(ptr) \
368 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
370 #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di) do { \
371 typeof(__di) ____di = (__di); \
372 ocfs2_error((__sb), \
373 "Dinode # %llu has bad signature %.*s", \
374 (unsigned long long)le64_to_cpu((____di)->i_blkno), 7, \
375 (____di)->i_signature); \
378 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \
379 (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
381 #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb) do { \
382 typeof(__eb) ____eb = (__eb); \
383 ocfs2_error((__sb), \
384 "Extent Block # %llu has bad signature %.*s", \
385 (unsigned long long)le64_to_cpu((____eb)->h_blkno), 7, \
386 (____eb)->h_signature); \
389 #define OCFS2_IS_VALID_GROUP_DESC(ptr) \
390 (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
392 #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd) do { \
393 typeof(__gd) ____gd = (__gd); \
394 ocfs2_error((__sb), \
395 "Group Descriptor # %llu has bad signature %.*s", \
396 (unsigned long long)le64_to_cpu((____gd)->bg_blkno), 7, \
397 (____gd)->bg_signature); \
400 static inline unsigned long ino_from_blkno(struct super_block
*sb
,
403 return (unsigned long)(blkno
& (u64
)ULONG_MAX
);
406 static inline u64
ocfs2_clusters_to_blocks(struct super_block
*sb
,
409 int c_to_b_bits
= OCFS2_SB(sb
)->s_clustersize_bits
-
410 sb
->s_blocksize_bits
;
412 return (u64
)clusters
<< c_to_b_bits
;
415 static inline u32
ocfs2_blocks_to_clusters(struct super_block
*sb
,
418 int b_to_c_bits
= OCFS2_SB(sb
)->s_clustersize_bits
-
419 sb
->s_blocksize_bits
;
421 return (u32
)(blocks
>> b_to_c_bits
);
424 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block
*sb
,
427 int cl_bits
= OCFS2_SB(sb
)->s_clustersize_bits
;
428 unsigned int clusters
;
430 bytes
+= OCFS2_SB(sb
)->s_clustersize
- 1;
431 /* OCFS2 just cannot have enough clusters to overflow this */
432 clusters
= (unsigned int)(bytes
>> cl_bits
);
437 static inline u64
ocfs2_blocks_for_bytes(struct super_block
*sb
,
440 bytes
+= sb
->s_blocksize
- 1;
441 return bytes
>> sb
->s_blocksize_bits
;
444 static inline u64
ocfs2_clusters_to_bytes(struct super_block
*sb
,
447 return (u64
)clusters
<< OCFS2_SB(sb
)->s_clustersize_bits
;
450 static inline u64
ocfs2_align_bytes_to_clusters(struct super_block
*sb
,
453 int cl_bits
= OCFS2_SB(sb
)->s_clustersize_bits
;
454 unsigned int clusters
;
456 clusters
= ocfs2_clusters_for_bytes(sb
, bytes
);
457 return (u64
)clusters
<< cl_bits
;
460 static inline u64
ocfs2_align_bytes_to_blocks(struct super_block
*sb
,
465 blocks
= ocfs2_blocks_for_bytes(sb
, bytes
);
466 return blocks
<< sb
->s_blocksize_bits
;
469 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes
)
471 return (unsigned long)((bytes
+ 511) >> 9);
474 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block
*sb
,
475 unsigned long pg_index
)
477 u32 clusters
= pg_index
;
478 unsigned int cbits
= OCFS2_SB(sb
)->s_clustersize_bits
;
480 if (unlikely(PAGE_CACHE_SHIFT
> cbits
))
481 clusters
= pg_index
<< (PAGE_CACHE_SHIFT
- cbits
);
482 else if (PAGE_CACHE_SHIFT
< cbits
)
483 clusters
= pg_index
>> (cbits
- PAGE_CACHE_SHIFT
);
489 * Find the 1st page index which covers the given clusters.
491 static inline pgoff_t
ocfs2_align_clusters_to_page_index(struct super_block
*sb
,
494 unsigned int cbits
= OCFS2_SB(sb
)->s_clustersize_bits
;
495 pgoff_t index
= clusters
;
497 if (PAGE_CACHE_SHIFT
> cbits
) {
498 index
= (pgoff_t
)clusters
>> (PAGE_CACHE_SHIFT
- cbits
);
499 } else if (PAGE_CACHE_SHIFT
< cbits
) {
500 index
= (pgoff_t
)clusters
<< (cbits
- PAGE_CACHE_SHIFT
);
506 static inline unsigned int ocfs2_pages_per_cluster(struct super_block
*sb
)
508 unsigned int cbits
= OCFS2_SB(sb
)->s_clustersize_bits
;
509 unsigned int pages_per_cluster
= 1;
511 if (PAGE_CACHE_SHIFT
< cbits
)
512 pages_per_cluster
= 1 << (cbits
- PAGE_CACHE_SHIFT
);
514 return pages_per_cluster
;
517 #define ocfs2_set_bit ext2_set_bit
518 #define ocfs2_clear_bit ext2_clear_bit
519 #define ocfs2_test_bit ext2_test_bit
520 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit