ocfs2: Create the lock status block union.
[pandora-kernel.git] / fs / ocfs2 / ocfs2.h
1 /* -*- mode: c; c-basic-offset: 8; -*-
2  * vim: noexpandtab sw=8 ts=8 sts=0:
3  *
4  * ocfs2.h
5  *
6  * Defines macros and structures used in OCFS2
7  *
8  * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
9  *
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.
14  *
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.
19  *
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.
24  */
25
26 #ifndef OCFS2_H
27 #define OCFS2_H
28
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>
38
39 #include "cluster/nodemanager.h"
40 #include "cluster/heartbeat.h"
41 #include "cluster/tcp.h"
42
43 /* For union ocfs2_dlm_lksb */
44 #include "stackglue.h"
45
46 #include "ocfs2_fs.h"
47 #include "ocfs2_lockid.h"
48
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
55
56 struct ocfs2_caching_info {
57         unsigned int            ci_num_cached;
58         union {
59                 sector_t        ci_array[OCFS2_INODE_MAX_CACHE_ARRAY];
60                 struct rb_root  ci_tree;
61         } ci_cache;
62 };
63
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 {
68         u16 num_nodes;
69         unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)];
70 };
71
72 enum ocfs2_ast_action {
73         OCFS2_AST_INVALID = 0,
74         OCFS2_AST_ATTACH,
75         OCFS2_AST_CONVERT,
76         OCFS2_AST_DOWNCONVERT,
77 };
78
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,
84 };
85
86 /* ocfs2_lock_res->l_flags flags. */
87 #define OCFS2_LOCK_ATTACHED      (0x00000001) /* have we initialized
88                                                * the lvb */
89 #define OCFS2_LOCK_BUSY          (0x00000002) /* we are currently in
90                                                * dlm_lock */
91 #define OCFS2_LOCK_BLOCKED       (0x00000004) /* blocked waiting to
92                                                * downconvert*/
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
101                                                * about to be
102                                                * dropped. */
103 #define OCFS2_LOCK_QUEUED        (0x00000100) /* queued for downconvert */
104 #define OCFS2_LOCK_NOCACHE       (0x00000200) /* don't use a holder count */
105
106 struct ocfs2_lock_res_ops;
107
108 typedef void (*ocfs2_lock_callback)(int status, unsigned long data);
109
110 struct ocfs2_lock_res {
111         void                    *l_priv;
112         struct ocfs2_lock_res_ops *l_ops;
113         spinlock_t               l_lock;
114
115         struct list_head         l_blocked_list;
116         struct list_head         l_mask_waiters;
117
118         enum ocfs2_lock_type     l_type;
119         unsigned long            l_flags;
120         char                     l_name[OCFS2_LOCK_ID_MAX_LEN];
121         int                      l_level;
122         unsigned int             l_ro_holders;
123         unsigned int             l_ex_holders;
124         union ocfs2_dlm_lksb     l_lksb;
125
126         /* used from AST/BAST funcs. */
127         enum ocfs2_ast_action    l_action;
128         enum ocfs2_unlock_action l_unlock_action;
129         int                      l_requested;
130         int                      l_blocking;
131
132         wait_queue_head_t        l_event;
133
134         struct list_head         l_debug_list;
135 };
136
137 struct ocfs2_dlm_debug {
138         struct kref d_refcnt;
139         struct dentry *d_locking_state;
140         struct list_head d_lockres_tracking;
141 };
142
143 enum ocfs2_vol_state
144 {
145         VOLUME_INIT = 0,
146         VOLUME_MOUNTED,
147         VOLUME_DISMOUNTED,
148         VOLUME_DISABLED
149 };
150
151 struct ocfs2_alloc_stats
152 {
153         atomic_t moves;
154         atomic_t local_data;
155         atomic_t bitmap_data;
156         atomic_t bg_allocs;
157         atomic_t bg_extends;
158 };
159
160 enum ocfs2_local_alloc_state
161 {
162         OCFS2_LA_UNUSED = 0,
163         OCFS2_LA_ENABLED,
164         OCFS2_LA_DISABLED
165 };
166
167 enum ocfs2_mount_options
168 {
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 */
175 };
176
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
181
182 struct ocfs2_journal;
183 struct ocfs2_slot_info;
184 struct ocfs2_recovery_map;
185 struct ocfs2_super
186 {
187         struct task_struct *commit_task;
188         struct super_block *sb;
189         struct inode *root_inode;
190         struct inode *sys_root_inode;
191         struct inode *system_inodes[NUM_SYSTEM_INODES];
192
193         struct ocfs2_slot_info *slot_info;
194
195         spinlock_t node_map_lock;
196
197         u64 root_blkno;
198         u64 system_dir_blkno;
199         u64 bitmap_blkno;
200         u32 bitmap_cpg;
201         u8 *uuid;
202         char *uuid_str;
203         u8 *vol_label;
204         u64 first_cluster_group_blkno;
205         u32 fs_generation;
206
207         u32 s_feature_compat;
208         u32 s_feature_incompat;
209         u32 s_feature_ro_compat;
210
211         /* Protects s_next_generaion, osb_flags. Could protect more on
212          * osb as it's very short lived. */
213         spinlock_t osb_lock;
214         u32 s_next_generation;
215         unsigned long osb_flags;
216
217         unsigned long s_mount_opt;
218         unsigned int s_atime_quantum;
219
220         unsigned int max_slots;
221         s16 node_num;
222         int slot_num;
223         int preferred_slot;
224         int s_sectsize_bits;
225         int s_clustersize;
226         int s_clustersize_bits;
227
228         atomic_t vol_state;
229         struct mutex recovery_lock;
230         struct ocfs2_recovery_map *recovery_map;
231         struct task_struct *recovery_thread_task;
232         int disable_recovery;
233         wait_queue_head_t checkpoint_event;
234         atomic_t needs_checkpoint;
235         struct ocfs2_journal *journal;
236         unsigned long osb_commit_interval;
237
238         int local_alloc_size;
239         enum ocfs2_local_alloc_state local_alloc_state;
240         struct buffer_head *local_alloc_bh;
241         u64 la_last_gd;
242
243         /* Next two fields are for local node slot recovery during
244          * mount. */
245         int dirty;
246         struct ocfs2_dinode *local_alloc_copy;
247
248         struct ocfs2_alloc_stats alloc_stats;
249         char dev_str[20];               /* "major,minor" of the device */
250
251         struct dlm_ctxt *dlm;
252         struct ocfs2_lock_res osb_super_lockres;
253         struct ocfs2_lock_res osb_rename_lockres;
254         struct dlm_eviction_cb osb_eviction_cb;
255         struct ocfs2_dlm_debug *osb_dlm_debug;
256         struct dlm_protocol_version osb_locking_proto;
257
258         struct dentry *osb_debug_root;
259
260         wait_queue_head_t recovery_event;
261
262         spinlock_t dc_task_lock;
263         struct task_struct *dc_task;
264         wait_queue_head_t dc_event;
265         unsigned long dc_wake_sequence;
266         unsigned long dc_work_sequence;
267
268         /*
269          * Any thread can add locks to the list, but the downconvert
270          * thread is the only one allowed to remove locks. Any change
271          * to this rule requires updating
272          * ocfs2_downconvert_thread_do_work().
273          */
274         struct list_head blocked_lock_list;
275         unsigned long blocked_lock_count;
276
277         wait_queue_head_t               osb_mount_event;
278
279         /* Truncate log info */
280         struct inode                    *osb_tl_inode;
281         struct buffer_head              *osb_tl_bh;
282         struct delayed_work             osb_truncate_log_wq;
283
284         struct ocfs2_node_map           osb_recovering_orphan_dirs;
285         unsigned int                    *osb_orphan_wipes;
286         wait_queue_head_t               osb_wipe_event;
287 };
288
289 #define OCFS2_SB(sb)        ((struct ocfs2_super *)(sb)->s_fs_info)
290
291 static inline int ocfs2_should_order_data(struct inode *inode)
292 {
293         if (!S_ISREG(inode->i_mode))
294                 return 0;
295         if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
296                 return 0;
297         return 1;
298 }
299
300 static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb)
301 {
302         if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC)
303                 return 1;
304         return 0;
305 }
306
307 static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb)
308 {
309         /*
310          * Support for sparse files is a pre-requisite
311          */
312         if (!ocfs2_sparse_alloc(osb))
313                 return 0;
314
315         if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN)
316                 return 1;
317         return 0;
318 }
319
320 static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb)
321 {
322         if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA)
323                 return 1;
324         return 0;
325 }
326
327 /* set / clear functions because cluster events can make these happen
328  * in parallel so we want the transitions to be atomic. this also
329  * means that any future flags osb_flags must be protected by spinlock
330  * too! */
331 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
332                                       unsigned long flag)
333 {
334         spin_lock(&osb->osb_lock);
335         osb->osb_flags |= flag;
336         spin_unlock(&osb->osb_lock);
337 }
338
339 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
340                                      int hard)
341 {
342         spin_lock(&osb->osb_lock);
343         osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
344         if (hard)
345                 osb->osb_flags |= OCFS2_OSB_HARD_RO;
346         else
347                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
348         spin_unlock(&osb->osb_lock);
349 }
350
351 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
352 {
353         int ret;
354
355         spin_lock(&osb->osb_lock);
356         ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
357         spin_unlock(&osb->osb_lock);
358
359         return ret;
360 }
361
362 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
363 {
364         int ret;
365
366         spin_lock(&osb->osb_lock);
367         ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
368         spin_unlock(&osb->osb_lock);
369
370         return ret;
371 }
372
373 static inline int ocfs2_mount_local(struct ocfs2_super *osb)
374 {
375         return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT);
376 }
377
378 static inline int ocfs2_uses_extended_slot_map(struct ocfs2_super *osb)
379 {
380         return (osb->s_feature_incompat &
381                 OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP);
382 }
383
384
385 #define OCFS2_IS_VALID_DINODE(ptr)                                      \
386         (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
387
388 #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di)  do {                    \
389         typeof(__di) ____di = (__di);                                   \
390         ocfs2_error((__sb),                                             \
391                 "Dinode # %llu has bad signature %.*s",                 \
392                 (unsigned long long)le64_to_cpu((____di)->i_blkno), 7,  \
393                 (____di)->i_signature);                                 \
394 } while (0)
395
396 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr)                                \
397         (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
398
399 #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb)    do {            \
400         typeof(__eb) ____eb = (__eb);                                   \
401         ocfs2_error((__sb),                                             \
402                 "Extent Block # %llu has bad signature %.*s",           \
403                 (unsigned long long)le64_to_cpu((____eb)->h_blkno), 7,  \
404                 (____eb)->h_signature);                                 \
405 } while (0)
406
407 #define OCFS2_IS_VALID_GROUP_DESC(ptr)                                  \
408         (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
409
410 #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd)      do {            \
411         typeof(__gd) ____gd = (__gd);                                   \
412                 ocfs2_error((__sb),                                     \
413                 "Group Descriptor # %llu has bad signature %.*s",       \
414                 (unsigned long long)le64_to_cpu((____gd)->bg_blkno), 7, \
415                 (____gd)->bg_signature);                                \
416 } while (0)
417
418 static inline unsigned long ino_from_blkno(struct super_block *sb,
419                                            u64 blkno)
420 {
421         return (unsigned long)(blkno & (u64)ULONG_MAX);
422 }
423
424 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
425                                            u32 clusters)
426 {
427         int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
428                 sb->s_blocksize_bits;
429
430         return (u64)clusters << c_to_b_bits;
431 }
432
433 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
434                                            u64 blocks)
435 {
436         int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
437                 sb->s_blocksize_bits;
438
439         return (u32)(blocks >> b_to_c_bits);
440 }
441
442 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
443                                                     u64 bytes)
444 {
445         int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
446         unsigned int clusters;
447
448         bytes += OCFS2_SB(sb)->s_clustersize - 1;
449         /* OCFS2 just cannot have enough clusters to overflow this */
450         clusters = (unsigned int)(bytes >> cl_bits);
451
452         return clusters;
453 }
454
455 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
456                                          u64 bytes)
457 {
458         bytes += sb->s_blocksize - 1;
459         return bytes >> sb->s_blocksize_bits;
460 }
461
462 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
463                                           u32 clusters)
464 {
465         return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
466 }
467
468 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
469                                                 u64 bytes)
470 {
471         int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
472         unsigned int clusters;
473
474         clusters = ocfs2_clusters_for_bytes(sb, bytes);
475         return (u64)clusters << cl_bits;
476 }
477
478 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
479                                               u64 bytes)
480 {
481         u64 blocks;
482
483         blocks = ocfs2_blocks_for_bytes(sb, bytes);
484         return blocks << sb->s_blocksize_bits;
485 }
486
487 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
488 {
489         return (unsigned long)((bytes + 511) >> 9);
490 }
491
492 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb,
493                                                         unsigned long pg_index)
494 {
495         u32 clusters = pg_index;
496         unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
497
498         if (unlikely(PAGE_CACHE_SHIFT > cbits))
499                 clusters = pg_index << (PAGE_CACHE_SHIFT - cbits);
500         else if (PAGE_CACHE_SHIFT < cbits)
501                 clusters = pg_index >> (cbits - PAGE_CACHE_SHIFT);
502
503         return clusters;
504 }
505
506 /*
507  * Find the 1st page index which covers the given clusters.
508  */
509 static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb,
510                                                         u32 clusters)
511 {
512         unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
513         pgoff_t index = clusters;
514
515         if (PAGE_CACHE_SHIFT > cbits) {
516                 index = (pgoff_t)clusters >> (PAGE_CACHE_SHIFT - cbits);
517         } else if (PAGE_CACHE_SHIFT < cbits) {
518                 index = (pgoff_t)clusters << (cbits - PAGE_CACHE_SHIFT);
519         }
520
521         return index;
522 }
523
524 static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb)
525 {
526         unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
527         unsigned int pages_per_cluster = 1;
528
529         if (PAGE_CACHE_SHIFT < cbits)
530                 pages_per_cluster = 1 << (cbits - PAGE_CACHE_SHIFT);
531
532         return pages_per_cluster;
533 }
534
535 #define ocfs2_set_bit ext2_set_bit
536 #define ocfs2_clear_bit ext2_clear_bit
537 #define ocfs2_test_bit ext2_test_bit
538 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit
539 #endif  /* OCFS2_H */
540