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 */
104 #define OCFS2_LOCK_NOCACHE (0x00000200) /* don't use a holder count */
106 struct ocfs2_lock_res_ops
;
108 typedef void (*ocfs2_lock_callback
)(int status
, unsigned long data
);
110 struct ocfs2_lock_res
{
112 struct ocfs2_lock_res_ops
*l_ops
;
115 struct list_head l_blocked_list
;
116 struct list_head l_mask_waiters
;
118 enum ocfs2_lock_type l_type
;
119 unsigned long l_flags
;
120 char l_name
[OCFS2_LOCK_ID_MAX_LEN
];
122 unsigned int l_ro_holders
;
123 unsigned int l_ex_holders
;
124 struct dlm_lockstatus l_lksb
;
126 /* used from AST/BAST funcs. */
127 enum ocfs2_ast_action l_action
;
128 enum ocfs2_unlock_action l_unlock_action
;
132 wait_queue_head_t l_event
;
134 struct list_head l_debug_list
;
137 struct ocfs2_dlm_debug
{
138 struct kref d_refcnt
;
139 struct dentry
*d_locking_state
;
140 struct list_head d_lockres_tracking
;
151 struct ocfs2_alloc_stats
155 atomic_t bitmap_data
;
160 enum ocfs2_local_alloc_state
167 enum ocfs2_mount_options
169 OCFS2_MOUNT_HB_LOCAL
= 1 << 0, /* Heartbeat started in local mode */
170 OCFS2_MOUNT_BARRIER
= 1 << 1, /* Use block barriers */
171 OCFS2_MOUNT_NOINTR
= 1 << 2, /* Don't catch signals */
172 OCFS2_MOUNT_ERRORS_PANIC
= 1 << 3, /* Panic on errors */
173 OCFS2_MOUNT_DATA_WRITEBACK
= 1 << 4, /* No data ordering */
174 OCFS2_MOUNT_LOCALFLOCKS
= 1 << 5, /* No cluster aware user file locks */
177 #define OCFS2_OSB_SOFT_RO 0x0001
178 #define OCFS2_OSB_HARD_RO 0x0002
179 #define OCFS2_OSB_ERROR_FS 0x0004
180 #define OCFS2_DEFAULT_ATIME_QUANTUM 60
182 struct ocfs2_journal
;
185 struct task_struct
*commit_task
;
186 struct super_block
*sb
;
187 struct inode
*root_inode
;
188 struct inode
*sys_root_inode
;
189 struct inode
*system_inodes
[NUM_SYSTEM_INODES
];
191 struct ocfs2_slot_info
*slot_info
;
193 spinlock_t node_map_lock
;
194 struct ocfs2_node_map recovery_map
;
197 u64 system_dir_blkno
;
203 u64 first_cluster_group_blkno
;
206 u32 s_feature_compat
;
207 u32 s_feature_incompat
;
208 u32 s_feature_ro_compat
;
210 /* Protects s_next_generaion, osb_flags. Could protect more on
211 * osb as it's very short lived. */
213 u32 s_next_generation
;
214 unsigned long osb_flags
;
216 unsigned long s_mount_opt
;
217 unsigned int s_atime_quantum
;
225 int s_clustersize_bits
;
228 struct mutex recovery_lock
;
229 struct task_struct
*recovery_thread_task
;
230 int disable_recovery
;
231 wait_queue_head_t checkpoint_event
;
232 atomic_t needs_checkpoint
;
233 struct ocfs2_journal
*journal
;
234 unsigned long osb_commit_interval
;
236 int local_alloc_size
;
237 enum ocfs2_local_alloc_state local_alloc_state
;
238 struct buffer_head
*local_alloc_bh
;
241 /* Next two fields are for local node slot recovery during
244 struct ocfs2_dinode
*local_alloc_copy
;
246 struct ocfs2_alloc_stats alloc_stats
;
247 char dev_str
[20]; /* "major,minor" of the device */
249 struct dlm_ctxt
*dlm
;
250 struct ocfs2_lock_res osb_super_lockres
;
251 struct ocfs2_lock_res osb_rename_lockres
;
252 struct dlm_eviction_cb osb_eviction_cb
;
253 struct ocfs2_dlm_debug
*osb_dlm_debug
;
255 struct dentry
*osb_debug_root
;
257 wait_queue_head_t recovery_event
;
259 spinlock_t dc_task_lock
;
260 struct task_struct
*dc_task
;
261 wait_queue_head_t dc_event
;
262 unsigned long dc_wake_sequence
;
263 unsigned long dc_work_sequence
;
266 * Any thread can add locks to the list, but the downconvert
267 * thread is the only one allowed to remove locks. Any change
268 * to this rule requires updating
269 * ocfs2_downconvert_thread_do_work().
271 struct list_head blocked_lock_list
;
272 unsigned long blocked_lock_count
;
274 wait_queue_head_t osb_mount_event
;
276 /* Truncate log info */
277 struct inode
*osb_tl_inode
;
278 struct buffer_head
*osb_tl_bh
;
279 struct delayed_work osb_truncate_log_wq
;
281 struct ocfs2_node_map osb_recovering_orphan_dirs
;
282 unsigned int *osb_orphan_wipes
;
283 wait_queue_head_t osb_wipe_event
;
286 #define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info)
288 static inline int ocfs2_should_order_data(struct inode
*inode
)
290 if (!S_ISREG(inode
->i_mode
))
292 if (OCFS2_SB(inode
->i_sb
)->s_mount_opt
& OCFS2_MOUNT_DATA_WRITEBACK
)
297 static inline int ocfs2_sparse_alloc(struct ocfs2_super
*osb
)
299 if (osb
->s_feature_incompat
& OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC
)
304 static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super
*osb
)
307 * Support for sparse files is a pre-requisite
309 if (!ocfs2_sparse_alloc(osb
))
312 if (osb
->s_feature_ro_compat
& OCFS2_FEATURE_RO_COMPAT_UNWRITTEN
)
317 static inline int ocfs2_supports_inline_data(struct ocfs2_super
*osb
)
319 if (osb
->s_feature_incompat
& OCFS2_FEATURE_INCOMPAT_INLINE_DATA
)
324 /* set / clear functions because cluster events can make these happen
325 * in parallel so we want the transitions to be atomic. this also
326 * means that any future flags osb_flags must be protected by spinlock
328 static inline void ocfs2_set_osb_flag(struct ocfs2_super
*osb
,
331 spin_lock(&osb
->osb_lock
);
332 osb
->osb_flags
|= flag
;
333 spin_unlock(&osb
->osb_lock
);
336 static inline void ocfs2_set_ro_flag(struct ocfs2_super
*osb
,
339 spin_lock(&osb
->osb_lock
);
340 osb
->osb_flags
&= ~(OCFS2_OSB_SOFT_RO
|OCFS2_OSB_HARD_RO
);
342 osb
->osb_flags
|= OCFS2_OSB_HARD_RO
;
344 osb
->osb_flags
|= OCFS2_OSB_SOFT_RO
;
345 spin_unlock(&osb
->osb_lock
);
348 static inline int ocfs2_is_hard_readonly(struct ocfs2_super
*osb
)
352 spin_lock(&osb
->osb_lock
);
353 ret
= osb
->osb_flags
& OCFS2_OSB_HARD_RO
;
354 spin_unlock(&osb
->osb_lock
);
359 static inline int ocfs2_is_soft_readonly(struct ocfs2_super
*osb
)
363 spin_lock(&osb
->osb_lock
);
364 ret
= osb
->osb_flags
& OCFS2_OSB_SOFT_RO
;
365 spin_unlock(&osb
->osb_lock
);
370 static inline int ocfs2_mount_local(struct ocfs2_super
*osb
)
372 return (osb
->s_feature_incompat
& OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT
);
375 #define OCFS2_IS_VALID_DINODE(ptr) \
376 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
378 #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di) do { \
379 typeof(__di) ____di = (__di); \
380 ocfs2_error((__sb), \
381 "Dinode # %llu has bad signature %.*s", \
382 (unsigned long long)le64_to_cpu((____di)->i_blkno), 7, \
383 (____di)->i_signature); \
386 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \
387 (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
389 #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb) do { \
390 typeof(__eb) ____eb = (__eb); \
391 ocfs2_error((__sb), \
392 "Extent Block # %llu has bad signature %.*s", \
393 (unsigned long long)le64_to_cpu((____eb)->h_blkno), 7, \
394 (____eb)->h_signature); \
397 #define OCFS2_IS_VALID_GROUP_DESC(ptr) \
398 (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
400 #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd) do { \
401 typeof(__gd) ____gd = (__gd); \
402 ocfs2_error((__sb), \
403 "Group Descriptor # %llu has bad signature %.*s", \
404 (unsigned long long)le64_to_cpu((____gd)->bg_blkno), 7, \
405 (____gd)->bg_signature); \
408 static inline unsigned long ino_from_blkno(struct super_block
*sb
,
411 return (unsigned long)(blkno
& (u64
)ULONG_MAX
);
414 static inline u64
ocfs2_clusters_to_blocks(struct super_block
*sb
,
417 int c_to_b_bits
= OCFS2_SB(sb
)->s_clustersize_bits
-
418 sb
->s_blocksize_bits
;
420 return (u64
)clusters
<< c_to_b_bits
;
423 static inline u32
ocfs2_blocks_to_clusters(struct super_block
*sb
,
426 int b_to_c_bits
= OCFS2_SB(sb
)->s_clustersize_bits
-
427 sb
->s_blocksize_bits
;
429 return (u32
)(blocks
>> b_to_c_bits
);
432 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block
*sb
,
435 int cl_bits
= OCFS2_SB(sb
)->s_clustersize_bits
;
436 unsigned int clusters
;
438 bytes
+= OCFS2_SB(sb
)->s_clustersize
- 1;
439 /* OCFS2 just cannot have enough clusters to overflow this */
440 clusters
= (unsigned int)(bytes
>> cl_bits
);
445 static inline u64
ocfs2_blocks_for_bytes(struct super_block
*sb
,
448 bytes
+= sb
->s_blocksize
- 1;
449 return bytes
>> sb
->s_blocksize_bits
;
452 static inline u64
ocfs2_clusters_to_bytes(struct super_block
*sb
,
455 return (u64
)clusters
<< OCFS2_SB(sb
)->s_clustersize_bits
;
458 static inline u64
ocfs2_align_bytes_to_clusters(struct super_block
*sb
,
461 int cl_bits
= OCFS2_SB(sb
)->s_clustersize_bits
;
462 unsigned int clusters
;
464 clusters
= ocfs2_clusters_for_bytes(sb
, bytes
);
465 return (u64
)clusters
<< cl_bits
;
468 static inline u64
ocfs2_align_bytes_to_blocks(struct super_block
*sb
,
473 blocks
= ocfs2_blocks_for_bytes(sb
, bytes
);
474 return blocks
<< sb
->s_blocksize_bits
;
477 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes
)
479 return (unsigned long)((bytes
+ 511) >> 9);
482 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block
*sb
,
483 unsigned long pg_index
)
485 u32 clusters
= pg_index
;
486 unsigned int cbits
= OCFS2_SB(sb
)->s_clustersize_bits
;
488 if (unlikely(PAGE_CACHE_SHIFT
> cbits
))
489 clusters
= pg_index
<< (PAGE_CACHE_SHIFT
- cbits
);
490 else if (PAGE_CACHE_SHIFT
< cbits
)
491 clusters
= pg_index
>> (cbits
- PAGE_CACHE_SHIFT
);
497 * Find the 1st page index which covers the given clusters.
499 static inline pgoff_t
ocfs2_align_clusters_to_page_index(struct super_block
*sb
,
502 unsigned int cbits
= OCFS2_SB(sb
)->s_clustersize_bits
;
503 pgoff_t index
= clusters
;
505 if (PAGE_CACHE_SHIFT
> cbits
) {
506 index
= (pgoff_t
)clusters
>> (PAGE_CACHE_SHIFT
- cbits
);
507 } else if (PAGE_CACHE_SHIFT
< cbits
) {
508 index
= (pgoff_t
)clusters
<< (cbits
- PAGE_CACHE_SHIFT
);
514 static inline unsigned int ocfs2_pages_per_cluster(struct super_block
*sb
)
516 unsigned int cbits
= OCFS2_SB(sb
)->s_clustersize_bits
;
517 unsigned int pages_per_cluster
= 1;
519 if (PAGE_CACHE_SHIFT
< cbits
)
520 pages_per_cluster
= 1 << (cbits
- PAGE_CACHE_SHIFT
);
522 return pages_per_cluster
;
525 #define ocfs2_set_bit ext2_set_bit
526 #define ocfs2_clear_bit ext2_clear_bit
527 #define ocfs2_test_bit ext2_test_bit
528 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit