ocfs2: Add "preferred slot" mount option
[linux-2.6] / 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 "endian.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
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 };
174
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
179
180 struct ocfs2_journal;
181 struct ocfs2_super
182 {
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];
188
189         struct ocfs2_slot_info *slot_info;
190
191         spinlock_t node_map_lock;
192         struct ocfs2_node_map mounted_map;
193         struct ocfs2_node_map recovery_map;
194         struct ocfs2_node_map umount_map;
195
196         u64 root_blkno;
197         u64 system_dir_blkno;
198         u64 bitmap_blkno;
199         u32 bitmap_cpg;
200         u8 *uuid;
201         char *uuid_str;
202         u8 *vol_label;
203         u64 first_cluster_group_blkno;
204         u32 fs_generation;
205
206         u32 s_feature_compat;
207         u32 s_feature_incompat;
208         u32 s_feature_ro_compat;
209
210         /* Protects s_next_generaion, osb_flags. Could protect more on
211          * osb as it's very short lived. */
212         spinlock_t osb_lock;
213         u32 s_next_generation;
214         unsigned long osb_flags;
215
216         unsigned long s_mount_opt;
217         unsigned int s_atime_quantum;
218
219         u16 max_slots;
220         s16 node_num;
221         s16 slot_num;
222         s16 preferred_slot;
223         int s_sectsize_bits;
224         int s_clustersize;
225         int s_clustersize_bits;
226
227         atomic_t vol_state;
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
235         enum ocfs2_local_alloc_state local_alloc_state;
236         struct buffer_head *local_alloc_bh;
237         u64 la_last_gd;
238
239         /* Next two fields are for local node slot recovery during
240          * mount. */
241         int dirty;
242         struct ocfs2_dinode *local_alloc_copy;
243
244         struct ocfs2_alloc_stats alloc_stats;
245         char dev_str[20];               /* "major,minor" of the device */
246
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;
252
253         struct dentry *osb_debug_root;
254
255         wait_queue_head_t recovery_event;
256
257         spinlock_t vote_task_lock;
258         struct task_struct *vote_task;
259         wait_queue_head_t vote_event;
260         unsigned long vote_wake_sequence;
261         unsigned long vote_work_sequence;
262
263         struct list_head blocked_lock_list;
264         unsigned long blocked_lock_count;
265
266         struct list_head vote_list;
267         int vote_count;
268
269         u32 net_key;
270         spinlock_t net_response_lock;
271         unsigned int net_response_ids;
272         struct list_head net_response_list;
273
274         struct o2hb_callback_func osb_hb_up;
275         struct o2hb_callback_func osb_hb_down;
276
277         struct list_head        osb_net_handlers;
278
279         wait_queue_head_t               osb_mount_event;
280
281         /* Truncate log info */
282         struct inode                    *osb_tl_inode;
283         struct buffer_head              *osb_tl_bh;
284         struct delayed_work             osb_truncate_log_wq;
285
286         struct ocfs2_node_map           osb_recovering_orphan_dirs;
287         unsigned int                    *osb_orphan_wipes;
288         wait_queue_head_t               osb_wipe_event;
289 };
290
291 #define OCFS2_SB(sb)        ((struct ocfs2_super *)(sb)->s_fs_info)
292
293 static inline int ocfs2_should_order_data(struct inode *inode)
294 {
295         if (!S_ISREG(inode->i_mode))
296                 return 0;
297         if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
298                 return 0;
299         return 1;
300 }
301
302 static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb)
303 {
304         if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC)
305                 return 1;
306         return 0;
307 }
308
309 /* set / clear functions because cluster events can make these happen
310  * in parallel so we want the transitions to be atomic. this also
311  * means that any future flags osb_flags must be protected by spinlock
312  * too! */
313 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
314                                       unsigned long flag)
315 {
316         spin_lock(&osb->osb_lock);
317         osb->osb_flags |= flag;
318         spin_unlock(&osb->osb_lock);
319 }
320
321 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
322                                      int hard)
323 {
324         spin_lock(&osb->osb_lock);
325         osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
326         if (hard)
327                 osb->osb_flags |= OCFS2_OSB_HARD_RO;
328         else
329                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
330         spin_unlock(&osb->osb_lock);
331 }
332
333 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
334 {
335         int ret;
336
337         spin_lock(&osb->osb_lock);
338         ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
339         spin_unlock(&osb->osb_lock);
340
341         return ret;
342 }
343
344 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
345 {
346         int ret;
347
348         spin_lock(&osb->osb_lock);
349         ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
350         spin_unlock(&osb->osb_lock);
351
352         return ret;
353 }
354
355 static inline int ocfs2_mount_local(struct ocfs2_super *osb)
356 {
357         return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT);
358 }
359
360 #define OCFS2_IS_VALID_DINODE(ptr)                                      \
361         (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
362
363 #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di)  do {                    \
364         typeof(__di) ____di = (__di);                                   \
365         ocfs2_error((__sb),                                             \
366                 "Dinode # %llu has bad signature %.*s",                 \
367                 (unsigned long long)le64_to_cpu((____di)->i_blkno), 7,  \
368                 (____di)->i_signature);                                 \
369 } while (0)
370
371 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr)                                \
372         (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
373
374 #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb)    do {            \
375         typeof(__eb) ____eb = (__eb);                                   \
376         ocfs2_error((__sb),                                             \
377                 "Extent Block # %llu has bad signature %.*s",           \
378                 (unsigned long long)le64_to_cpu((____eb)->h_blkno), 7,  \
379                 (____eb)->h_signature);                                 \
380 } while (0)
381
382 #define OCFS2_IS_VALID_GROUP_DESC(ptr)                                  \
383         (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
384
385 #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd)      do {            \
386         typeof(__gd) ____gd = (__gd);                                   \
387                 ocfs2_error((__sb),                                     \
388                 "Group Descriptor # %llu has bad signature %.*s",       \
389                 (unsigned long long)le64_to_cpu((____gd)->bg_blkno), 7, \
390                 (____gd)->bg_signature);                                \
391 } while (0)
392
393 static inline unsigned long ino_from_blkno(struct super_block *sb,
394                                            u64 blkno)
395 {
396         return (unsigned long)(blkno & (u64)ULONG_MAX);
397 }
398
399 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
400                                            u32 clusters)
401 {
402         int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
403                 sb->s_blocksize_bits;
404
405         return (u64)clusters << c_to_b_bits;
406 }
407
408 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
409                                            u64 blocks)
410 {
411         int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
412                 sb->s_blocksize_bits;
413
414         return (u32)(blocks >> b_to_c_bits);
415 }
416
417 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
418                                                     u64 bytes)
419 {
420         int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
421         unsigned int clusters;
422
423         bytes += OCFS2_SB(sb)->s_clustersize - 1;
424         /* OCFS2 just cannot have enough clusters to overflow this */
425         clusters = (unsigned int)(bytes >> cl_bits);
426
427         return clusters;
428 }
429
430 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
431                                          u64 bytes)
432 {
433         bytes += sb->s_blocksize - 1;
434         return bytes >> sb->s_blocksize_bits;
435 }
436
437 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
438                                           u32 clusters)
439 {
440         return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
441 }
442
443 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
444                                                 u64 bytes)
445 {
446         int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
447         unsigned int clusters;
448
449         clusters = ocfs2_clusters_for_bytes(sb, bytes);
450         return (u64)clusters << cl_bits;
451 }
452
453 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
454                                               u64 bytes)
455 {
456         u64 blocks;
457
458         blocks = ocfs2_blocks_for_bytes(sb, bytes);
459         return blocks << sb->s_blocksize_bits;
460 }
461
462 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
463 {
464         return (unsigned long)((bytes + 511) >> 9);
465 }
466
467 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb,
468                                                         unsigned long pg_index)
469 {
470         u32 clusters = pg_index;
471         unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
472
473         if (unlikely(PAGE_CACHE_SHIFT > cbits))
474                 clusters = pg_index << (PAGE_CACHE_SHIFT - cbits);
475         else if (PAGE_CACHE_SHIFT < cbits)
476                 clusters = pg_index >> (cbits - PAGE_CACHE_SHIFT);
477
478         return clusters;
479 }
480
481 /*
482  * Find the 1st page index which covers the given clusters.
483  */
484 static inline unsigned long ocfs2_align_clusters_to_page_index(struct super_block *sb,
485                                                         u32 clusters)
486 {
487         unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
488         unsigned long index = clusters;
489
490         if (PAGE_CACHE_SHIFT > cbits) {
491                 index = clusters >> (PAGE_CACHE_SHIFT - cbits);
492         } else if (PAGE_CACHE_SHIFT < cbits) {
493                 index = clusters << (cbits - PAGE_CACHE_SHIFT);
494         }
495
496         return index;
497 }
498
499 static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb)
500 {
501         unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
502         unsigned int pages_per_cluster = 1;
503
504         if (PAGE_CACHE_SHIFT < cbits)
505                 pages_per_cluster = 1 << (cbits - PAGE_CACHE_SHIFT);
506
507         return pages_per_cluster;
508 }
509
510 #define ocfs2_set_bit ext2_set_bit
511 #define ocfs2_clear_bit ext2_clear_bit
512 #define ocfs2_test_bit ext2_test_bit
513 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit
514 #endif  /* OCFS2_H */
515