8d8e4779df92ba7ffe84f293d17cddb9b7afb6f3
[sfrench/cifs-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
38 #include "cluster/nodemanager.h"
39 #include "cluster/heartbeat.h"
40 #include "cluster/tcp.h"
41
42 #include "dlm/dlmapi.h"
43
44 #include "ocfs2_fs.h"
45 #include "endian.h"
46 #include "ocfs2_lockid.h"
47
48 struct ocfs2_extent_map {
49         u32             em_clusters;
50         struct rb_root  em_extents;
51 };
52
53 /* Most user visible OCFS2 inodes will have very few pieces of
54  * metadata, but larger files (including bitmaps, etc) must be taken
55  * into account when designing an access scheme. We allow a small
56  * amount of inlined blocks to be stored on an array and grow the
57  * structure into a rb tree when necessary. */
58 #define OCFS2_INODE_MAX_CACHE_ARRAY 2
59
60 struct ocfs2_caching_info {
61         unsigned int            ci_num_cached;
62         union {
63                 sector_t        ci_array[OCFS2_INODE_MAX_CACHE_ARRAY];
64                 struct rb_root  ci_tree;
65         } ci_cache;
66 };
67
68 /* this limits us to 256 nodes
69  * if we need more, we can do a kmalloc for the map */
70 #define OCFS2_NODE_MAP_MAX_NODES    256
71 struct ocfs2_node_map {
72         u16 num_nodes;
73         unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)];
74 };
75
76 enum ocfs2_ast_action {
77         OCFS2_AST_INVALID = 0,
78         OCFS2_AST_ATTACH,
79         OCFS2_AST_CONVERT,
80         OCFS2_AST_DOWNCONVERT,
81 };
82
83 /* actions for an unlockast function to take. */
84 enum ocfs2_unlock_action {
85         OCFS2_UNLOCK_INVALID = 0,
86         OCFS2_UNLOCK_CANCEL_CONVERT,
87         OCFS2_UNLOCK_DROP_LOCK,
88 };
89
90 /* ocfs2_lock_res->l_flags flags. */
91 #define OCFS2_LOCK_ATTACHED      (0x00000001) /* have we initialized
92                                                * the lvb */
93 #define OCFS2_LOCK_BUSY          (0x00000002) /* we are currently in
94                                                * dlm_lock */
95 #define OCFS2_LOCK_BLOCKED       (0x00000004) /* blocked waiting to
96                                                * downconvert*/
97 #define OCFS2_LOCK_LOCAL         (0x00000008) /* newly created inode */
98 #define OCFS2_LOCK_NEEDS_REFRESH (0x00000010)
99 #define OCFS2_LOCK_REFRESHING    (0x00000020)
100 #define OCFS2_LOCK_INITIALIZED   (0x00000040) /* track initialization
101                                                * for shutdown paths */
102 #define OCFS2_LOCK_FREEING       (0x00000080) /* help dlmglue track
103                                                * when to skip queueing
104                                                * a lock because it's
105                                                * about to be
106                                                * dropped. */
107 #define OCFS2_LOCK_QUEUED        (0x00000100) /* queued for downconvert */
108
109 struct ocfs2_lock_res_ops;
110
111 typedef void (*ocfs2_lock_callback)(int status, unsigned long data);
112
113 struct ocfs2_lock_res {
114         void                    *l_priv;
115         struct ocfs2_lock_res_ops *l_ops;
116         spinlock_t               l_lock;
117
118         struct list_head         l_blocked_list;
119         struct list_head         l_mask_waiters;
120
121         enum ocfs2_lock_type     l_type;
122         unsigned long            l_flags;
123         char                     l_name[OCFS2_LOCK_ID_MAX_LEN];
124         int                      l_level;
125         unsigned int             l_ro_holders;
126         unsigned int             l_ex_holders;
127         struct dlm_lockstatus    l_lksb;
128
129         /* used from AST/BAST funcs. */
130         enum ocfs2_ast_action    l_action;
131         enum ocfs2_unlock_action l_unlock_action;
132         int                      l_requested;
133         int                      l_blocking;
134
135         wait_queue_head_t        l_event;
136
137         struct list_head         l_debug_list;
138 };
139
140 struct ocfs2_dlm_debug {
141         struct kref d_refcnt;
142         struct dentry *d_locking_state;
143         struct list_head d_lockres_tracking;
144 };
145
146 enum ocfs2_vol_state
147 {
148         VOLUME_INIT = 0,
149         VOLUME_MOUNTED,
150         VOLUME_DISMOUNTED,
151         VOLUME_DISABLED
152 };
153
154 struct ocfs2_alloc_stats
155 {
156         atomic_t moves;
157         atomic_t local_data;
158         atomic_t bitmap_data;
159         atomic_t bg_allocs;
160         atomic_t bg_extends;
161 };
162
163 enum ocfs2_local_alloc_state
164 {
165         OCFS2_LA_UNUSED = 0,
166         OCFS2_LA_ENABLED,
167         OCFS2_LA_DISABLED
168 };
169
170 enum ocfs2_mount_options
171 {
172         OCFS2_MOUNT_HB_LOCAL   = 1 << 0, /* Heartbeat started in local mode */
173         OCFS2_MOUNT_BARRIER = 1 << 1,   /* Use block barriers */
174         OCFS2_MOUNT_NOINTR  = 1 << 2,   /* Don't catch signals */
175         OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */
176         OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */
177 #ifdef OCFS2_ORACORE_WORKAROUNDS
178         OCFS2_MOUNT_COMPAT_OCFS = 1 << 30, /* ocfs1 compatibility mode */
179 #endif
180 };
181
182 #define OCFS2_OSB_SOFT_RO       0x0001
183 #define OCFS2_OSB_HARD_RO       0x0002
184 #define OCFS2_OSB_ERROR_FS      0x0004
185
186 struct ocfs2_journal;
187 struct ocfs2_journal_handle;
188 struct ocfs2_super
189 {
190         u32 osb_id;             /* id used by the proc interface */
191         struct task_struct *commit_task;
192         struct super_block *sb;
193         struct inode *root_inode;
194         struct inode *sys_root_inode;
195         struct inode *system_inodes[NUM_SYSTEM_INODES];
196
197         struct ocfs2_slot_info *slot_info;
198
199         spinlock_t node_map_lock;
200         struct ocfs2_node_map mounted_map;
201         struct ocfs2_node_map recovery_map;
202         struct ocfs2_node_map umount_map;
203
204         u32 num_clusters;
205         u64 root_blkno;
206         u64 system_dir_blkno;
207         u64 bitmap_blkno;
208         u32 bitmap_cpg;
209         u8 *uuid;
210         char *uuid_str;
211         u8 *vol_label;
212         u64 first_cluster_group_blkno;
213         u32 fs_generation;
214
215         u32 s_feature_compat;
216         u32 s_feature_incompat;
217         u32 s_feature_ro_compat;
218
219         /* Protects s_next_generaion, osb_flags. Could protect more on
220          * osb as it's very short lived. */
221         spinlock_t osb_lock;
222         u32 s_next_generation;
223         unsigned long osb_flags;
224
225         unsigned long s_mount_opt;
226
227         u16 max_slots;
228         u16 num_nodes;
229         s16 node_num;
230         s16 slot_num;
231         int s_sectsize_bits;
232         int s_clustersize;
233         int s_clustersize_bits;
234         struct proc_dir_entry *proc_sub_dir; /* points to /proc/fs/ocfs2/<maj_min> */
235
236         atomic_t vol_state;
237         struct mutex recovery_lock;
238         struct task_struct *recovery_thread_task;
239         int disable_recovery;
240         wait_queue_head_t checkpoint_event;
241         atomic_t needs_checkpoint;
242         struct ocfs2_journal *journal;
243
244         enum ocfs2_local_alloc_state local_alloc_state;
245         struct buffer_head *local_alloc_bh;
246
247         /* Next two fields are for local node slot recovery during
248          * mount. */
249         int dirty;
250         struct ocfs2_dinode *local_alloc_copy;
251
252         struct ocfs2_alloc_stats alloc_stats;
253         char dev_str[20];               /* "major,minor" of the device */
254
255         struct dlm_ctxt *dlm;
256         struct ocfs2_lock_res osb_super_lockres;
257         struct ocfs2_lock_res osb_rename_lockres;
258         struct dlm_eviction_cb osb_eviction_cb;
259         struct ocfs2_dlm_debug *osb_dlm_debug;
260
261         struct dentry *osb_debug_root;
262
263         wait_queue_head_t recovery_event;
264
265         spinlock_t vote_task_lock;
266         struct task_struct *vote_task;
267         wait_queue_head_t vote_event;
268         unsigned long vote_wake_sequence;
269         unsigned long vote_work_sequence;
270
271         struct list_head blocked_lock_list;
272         unsigned long blocked_lock_count;
273
274         struct list_head vote_list;
275         int vote_count;
276
277         u32 net_key;
278         spinlock_t net_response_lock;
279         unsigned int net_response_ids;
280         struct list_head net_response_list;
281
282         struct o2hb_callback_func osb_hb_up;
283         struct o2hb_callback_func osb_hb_down;
284
285         struct list_head        osb_net_handlers;
286
287         wait_queue_head_t               osb_mount_event;
288
289         /* Truncate log info */
290         struct inode                    *osb_tl_inode;
291         struct buffer_head              *osb_tl_bh;
292         struct work_struct              osb_truncate_log_wq;
293 };
294
295 #define OCFS2_SB(sb)        ((struct ocfs2_super *)(sb)->s_fs_info)
296 #define OCFS2_MAX_OSB_ID             65536
297
298 static inline int ocfs2_should_order_data(struct inode *inode)
299 {
300         if (!S_ISREG(inode->i_mode))
301                 return 0;
302         if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
303                 return 0;
304         return 1;
305 }
306
307 /* set / clear functions because cluster events can make these happen
308  * in parallel so we want the transitions to be atomic. this also
309  * means that any future flags osb_flags must be protected by spinlock
310  * too! */
311 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
312                                       unsigned long flag)
313 {
314         spin_lock(&osb->osb_lock);
315         osb->osb_flags |= flag;
316         spin_unlock(&osb->osb_lock);
317 }
318
319 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
320                                      int hard)
321 {
322         spin_lock(&osb->osb_lock);
323         osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
324         if (hard)
325                 osb->osb_flags |= OCFS2_OSB_HARD_RO;
326         else
327                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
328         spin_unlock(&osb->osb_lock);
329 }
330
331 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
332 {
333         int ret;
334
335         spin_lock(&osb->osb_lock);
336         ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
337         spin_unlock(&osb->osb_lock);
338
339         return ret;
340 }
341
342 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
343 {
344         int ret;
345
346         spin_lock(&osb->osb_lock);
347         ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
348         spin_unlock(&osb->osb_lock);
349
350         return ret;
351 }
352
353 #define OCFS2_IS_VALID_DINODE(ptr)                                      \
354         (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
355
356 #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di)  do {                    \
357         typeof(__di) ____di = (__di);                                   \
358         ocfs2_error((__sb),                                             \
359                 "Dinode # %"MLFu64" has bad signature %.*s",            \
360                 (____di)->i_blkno, 7,                                   \
361                 (____di)->i_signature);                                 \
362 } while (0);
363
364 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr)                                \
365         (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
366
367 #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb)    do {            \
368         typeof(__eb) ____eb = (__eb);                                   \
369         ocfs2_error((__sb),                                             \
370                 "Extent Block # %"MLFu64" has bad signature %.*s",      \
371                 (____eb)->h_blkno, 7,                                   \
372                 (____eb)->h_signature);                                 \
373 } while (0);
374
375 #define OCFS2_IS_VALID_GROUP_DESC(ptr)                                  \
376         (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
377
378 #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd)      do {            \
379         typeof(__gd) ____gd = (__gd);                                   \
380                 ocfs2_error((__sb),                                     \
381                 "Group Descriptor # %"MLFu64" has bad signature %.*s",  \
382                 (____gd)->bg_blkno, 7,                                  \
383                 (____gd)->bg_signature);                                \
384 } while (0);
385
386 static inline unsigned long ino_from_blkno(struct super_block *sb,
387                                            u64 blkno)
388 {
389         return (unsigned long)(blkno & (u64)ULONG_MAX);
390 }
391
392 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
393                                            u32 clusters)
394 {
395         int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
396                 sb->s_blocksize_bits;
397
398         return (u64)clusters << c_to_b_bits;
399 }
400
401 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
402                                            u64 blocks)
403 {
404         int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
405                 sb->s_blocksize_bits;
406
407         return (u32)(blocks >> b_to_c_bits);
408 }
409
410 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
411                                                     u64 bytes)
412 {
413         int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
414         unsigned int clusters;
415
416         bytes += OCFS2_SB(sb)->s_clustersize - 1;
417         /* OCFS2 just cannot have enough clusters to overflow this */
418         clusters = (unsigned int)(bytes >> cl_bits);
419
420         return clusters;
421 }
422
423 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
424                                          u64 bytes)
425 {
426         bytes += sb->s_blocksize - 1;
427         return bytes >> sb->s_blocksize_bits;
428 }
429
430 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
431                                           u32 clusters)
432 {
433         return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
434 }
435
436 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
437                                                 u64 bytes)
438 {
439         int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
440         unsigned int clusters;
441
442         clusters = ocfs2_clusters_for_bytes(sb, bytes);
443         return (u64)clusters << cl_bits;
444 }
445
446 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
447                                               u64 bytes)
448 {
449         u64 blocks;
450
451         blocks = ocfs2_blocks_for_bytes(sb, bytes);
452         return blocks << sb->s_blocksize_bits;
453 }
454
455 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
456 {
457         return (unsigned long)((bytes + 511) >> 9);
458 }
459
460 #define ocfs2_set_bit ext2_set_bit
461 #define ocfs2_clear_bit ext2_clear_bit
462 #define ocfs2_test_bit ext2_test_bit
463 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit
464 #endif  /* OCFS2_H */
465