Merge git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-nmw
[linux-2.6.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 #include "dlm/dlmapi.h"
44
45 #include "ocfs2_fs.h"
46 #include "ocfs2_lockid.h"
47
48 /* Most user visible OCFS2 inodes will have very few pieces of
49  * metadata, but larger files (including bitmaps, etc) must be taken
50  * into account when designing an access scheme. We allow a small
51  * amount of inlined blocks to be stored on an array and grow the
52  * structure into a rb tree when necessary. */
53 #define OCFS2_INODE_MAX_CACHE_ARRAY 2
54
55 struct ocfs2_caching_info {
56         unsigned int            ci_num_cached;
57         union {
58                 sector_t        ci_array[OCFS2_INODE_MAX_CACHE_ARRAY];
59                 struct rb_root  ci_tree;
60         } ci_cache;
61 };
62
63 /* this limits us to 256 nodes
64  * if we need more, we can do a kmalloc for the map */
65 #define OCFS2_NODE_MAP_MAX_NODES    256
66 struct ocfs2_node_map {
67         u16 num_nodes;
68         unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)];
69 };
70
71 enum ocfs2_ast_action {
72         OCFS2_AST_INVALID = 0,
73         OCFS2_AST_ATTACH,
74         OCFS2_AST_CONVERT,
75         OCFS2_AST_DOWNCONVERT,
76 };
77
78 /* actions for an unlockast function to take. */
79 enum ocfs2_unlock_action {
80         OCFS2_UNLOCK_INVALID = 0,
81         OCFS2_UNLOCK_CANCEL_CONVERT,
82         OCFS2_UNLOCK_DROP_LOCK,
83 };
84
85 /* ocfs2_lock_res->l_flags flags. */
86 #define OCFS2_LOCK_ATTACHED      (0x00000001) /* have we initialized
87                                                * the lvb */
88 #define OCFS2_LOCK_BUSY          (0x00000002) /* we are currently in
89                                                * dlm_lock */
90 #define OCFS2_LOCK_BLOCKED       (0x00000004) /* blocked waiting to
91                                                * downconvert*/
92 #define OCFS2_LOCK_LOCAL         (0x00000008) /* newly created inode */
93 #define OCFS2_LOCK_NEEDS_REFRESH (0x00000010)
94 #define OCFS2_LOCK_REFRESHING    (0x00000020)
95 #define OCFS2_LOCK_INITIALIZED   (0x00000040) /* track initialization
96                                                * for shutdown paths */
97 #define OCFS2_LOCK_FREEING       (0x00000080) /* help dlmglue track
98                                                * when to skip queueing
99                                                * a lock because it's
100                                                * about to be
101                                                * dropped. */
102 #define OCFS2_LOCK_QUEUED        (0x00000100) /* queued for downconvert */
103 #define OCFS2_LOCK_NOCACHE       (0x00000200) /* don't use a holder count */
104
105 struct ocfs2_lock_res_ops;
106
107 typedef void (*ocfs2_lock_callback)(int status, unsigned long data);
108
109 struct ocfs2_lock_res {
110         void                    *l_priv;
111         struct ocfs2_lock_res_ops *l_ops;
112         spinlock_t               l_lock;
113
114         struct list_head         l_blocked_list;
115         struct list_head         l_mask_waiters;
116
117         enum ocfs2_lock_type     l_type;
118         unsigned long            l_flags;
119         char                     l_name[OCFS2_LOCK_ID_MAX_LEN];
120         int                      l_level;
121         unsigned int             l_ro_holders;
122         unsigned int             l_ex_holders;
123         struct dlm_lockstatus    l_lksb;
124
125         /* used from AST/BAST funcs. */
126         enum ocfs2_ast_action    l_action;
127         enum ocfs2_unlock_action l_unlock_action;
128         int                      l_requested;
129         int                      l_blocking;
130
131         wait_queue_head_t        l_event;
132
133         struct list_head         l_debug_list;
134 };
135
136 struct ocfs2_dlm_debug {
137         struct kref d_refcnt;
138         struct dentry *d_locking_state;
139         struct list_head d_lockres_tracking;
140 };
141
142 enum ocfs2_vol_state
143 {
144         VOLUME_INIT = 0,
145         VOLUME_MOUNTED,
146         VOLUME_DISMOUNTED,
147         VOLUME_DISABLED
148 };
149
150 struct ocfs2_alloc_stats
151 {
152         atomic_t moves;
153         atomic_t local_data;
154         atomic_t bitmap_data;
155         atomic_t bg_allocs;
156         atomic_t bg_extends;
157 };
158
159 enum ocfs2_local_alloc_state
160 {
161         OCFS2_LA_UNUSED = 0,
162         OCFS2_LA_ENABLED,
163         OCFS2_LA_DISABLED
164 };
165
166 enum ocfs2_mount_options
167 {
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 */
173         OCFS2_MOUNT_LOCALFLOCKS = 1 << 5, /* No cluster aware user file locks */
174 };
175
176 #define OCFS2_OSB_SOFT_RO       0x0001
177 #define OCFS2_OSB_HARD_RO       0x0002
178 #define OCFS2_OSB_ERROR_FS      0x0004
179 #define OCFS2_DEFAULT_ATIME_QUANTUM     60
180
181 struct ocfs2_journal;
182 struct ocfs2_super
183 {
184         struct task_struct *commit_task;
185         struct super_block *sb;
186         struct inode *root_inode;
187         struct inode *sys_root_inode;
188         struct inode *system_inodes[NUM_SYSTEM_INODES];
189
190         struct ocfs2_slot_info *slot_info;
191
192         spinlock_t node_map_lock;
193         struct ocfs2_node_map recovery_map;
194
195         u64 root_blkno;
196         u64 system_dir_blkno;
197         u64 bitmap_blkno;
198         u32 bitmap_cpg;
199         u8 *uuid;
200         char *uuid_str;
201         u8 *vol_label;
202         u64 first_cluster_group_blkno;
203         u32 fs_generation;
204
205         u32 s_feature_compat;
206         u32 s_feature_incompat;
207         u32 s_feature_ro_compat;
208
209         /* Protects s_next_generaion, osb_flags. Could protect more on
210          * osb as it's very short lived. */
211         spinlock_t osb_lock;
212         u32 s_next_generation;
213         unsigned long osb_flags;
214
215         unsigned long s_mount_opt;
216         unsigned int s_atime_quantum;
217
218         u16 max_slots;
219         s16 node_num;
220         s16 slot_num;
221         s16 preferred_slot;
222         int s_sectsize_bits;
223         int s_clustersize;
224         int s_clustersize_bits;
225
226         atomic_t vol_state;
227         struct mutex recovery_lock;
228         struct task_struct *recovery_thread_task;
229         int disable_recovery;
230         wait_queue_head_t checkpoint_event;
231         atomic_t needs_checkpoint;
232         struct ocfs2_journal *journal;
233         unsigned long osb_commit_interval;
234
235         int local_alloc_size;
236         enum ocfs2_local_alloc_state local_alloc_state;
237         struct buffer_head *local_alloc_bh;
238         u64 la_last_gd;
239
240         /* Next two fields are for local node slot recovery during
241          * mount. */
242         int dirty;
243         struct ocfs2_dinode *local_alloc_copy;
244
245         struct ocfs2_alloc_stats alloc_stats;
246         char dev_str[20];               /* "major,minor" of the device */
247
248         struct dlm_ctxt *dlm;
249         struct ocfs2_lock_res osb_super_lockres;
250         struct ocfs2_lock_res osb_rename_lockres;
251         struct dlm_eviction_cb osb_eviction_cb;
252         struct ocfs2_dlm_debug *osb_dlm_debug;
253         struct dlm_protocol_version osb_locking_proto;
254
255         struct dentry *osb_debug_root;
256
257         wait_queue_head_t recovery_event;
258
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;
264
265         /*
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().
270          */
271         struct list_head blocked_lock_list;
272         unsigned long blocked_lock_count;
273
274         wait_queue_head_t               osb_mount_event;
275
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;
280
281         struct ocfs2_node_map           osb_recovering_orphan_dirs;
282         unsigned int                    *osb_orphan_wipes;
283         wait_queue_head_t               osb_wipe_event;
284 };
285
286 #define OCFS2_SB(sb)        ((struct ocfs2_super *)(sb)->s_fs_info)
287
288 static inline int ocfs2_should_order_data(struct inode *inode)
289 {
290         if (!S_ISREG(inode->i_mode))
291                 return 0;
292         if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
293                 return 0;
294         return 1;
295 }
296
297 static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb)
298 {
299         if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC)
300                 return 1;
301         return 0;
302 }
303
304 static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb)
305 {
306         /*
307          * Support for sparse files is a pre-requisite
308          */
309         if (!ocfs2_sparse_alloc(osb))
310                 return 0;
311
312         if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN)
313                 return 1;
314         return 0;
315 }
316
317 static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb)
318 {
319         if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA)
320                 return 1;
321         return 0;
322 }
323
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
327  * too! */
328 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
329                                       unsigned long flag)
330 {
331         spin_lock(&osb->osb_lock);
332         osb->osb_flags |= flag;
333         spin_unlock(&osb->osb_lock);
334 }
335
336 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
337                                      int hard)
338 {
339         spin_lock(&osb->osb_lock);
340         osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
341         if (hard)
342                 osb->osb_flags |= OCFS2_OSB_HARD_RO;
343         else
344                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
345         spin_unlock(&osb->osb_lock);
346 }
347
348 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
349 {
350         int ret;
351
352         spin_lock(&osb->osb_lock);
353         ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
354         spin_unlock(&osb->osb_lock);
355
356         return ret;
357 }
358
359 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
360 {
361         int ret;
362
363         spin_lock(&osb->osb_lock);
364         ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
365         spin_unlock(&osb->osb_lock);
366
367         return ret;
368 }
369
370 static inline int ocfs2_mount_local(struct ocfs2_super *osb)
371 {
372         return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT);
373 }
374
375 #define OCFS2_IS_VALID_DINODE(ptr)                                      \
376         (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
377
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);                                 \
384 } while (0)
385
386 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr)                                \
387         (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
388
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);                                 \
395 } while (0)
396
397 #define OCFS2_IS_VALID_GROUP_DESC(ptr)                                  \
398         (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
399
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);                                \
406 } while (0)
407
408 static inline unsigned long ino_from_blkno(struct super_block *sb,
409                                            u64 blkno)
410 {
411         return (unsigned long)(blkno & (u64)ULONG_MAX);
412 }
413
414 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
415                                            u32 clusters)
416 {
417         int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
418                 sb->s_blocksize_bits;
419
420         return (u64)clusters << c_to_b_bits;
421 }
422
423 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
424                                            u64 blocks)
425 {
426         int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
427                 sb->s_blocksize_bits;
428
429         return (u32)(blocks >> b_to_c_bits);
430 }
431
432 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
433                                                     u64 bytes)
434 {
435         int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
436         unsigned int clusters;
437
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);
441
442         return clusters;
443 }
444
445 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
446                                          u64 bytes)
447 {
448         bytes += sb->s_blocksize - 1;
449         return bytes >> sb->s_blocksize_bits;
450 }
451
452 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
453                                           u32 clusters)
454 {
455         return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
456 }
457
458 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
459                                                 u64 bytes)
460 {
461         int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
462         unsigned int clusters;
463
464         clusters = ocfs2_clusters_for_bytes(sb, bytes);
465         return (u64)clusters << cl_bits;
466 }
467
468 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
469                                               u64 bytes)
470 {
471         u64 blocks;
472
473         blocks = ocfs2_blocks_for_bytes(sb, bytes);
474         return blocks << sb->s_blocksize_bits;
475 }
476
477 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
478 {
479         return (unsigned long)((bytes + 511) >> 9);
480 }
481
482 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb,
483                                                         unsigned long pg_index)
484 {
485         u32 clusters = pg_index;
486         unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
487
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);
492
493         return clusters;
494 }
495
496 /*
497  * Find the 1st page index which covers the given clusters.
498  */
499 static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb,
500                                                         u32 clusters)
501 {
502         unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
503         pgoff_t index = clusters;
504
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);
509         }
510
511         return index;
512 }
513
514 static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb)
515 {
516         unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
517         unsigned int pages_per_cluster = 1;
518
519         if (PAGE_CACHE_SHIFT < cbits)
520                 pages_per_cluster = 1 << (cbits - PAGE_CACHE_SHIFT);
521
522         return pages_per_cluster;
523 }
524
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
529 #endif  /* OCFS2_H */
530