Loading...
1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/*
3 * ocfs2.h
4 *
5 * Defines macros and structures used in OCFS2
6 *
7 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
8 */
9
10#ifndef OCFS2_H
11#define OCFS2_H
12
13#include <linux/spinlock.h>
14#include <linux/sched.h>
15#include <linux/wait.h>
16#include <linux/list.h>
17#include <linux/llist.h>
18#include <linux/rbtree.h>
19#include <linux/workqueue.h>
20#include <linux/kref.h>
21#include <linux/mutex.h>
22#include <linux/lockdep.h>
23#include <linux/jbd2.h>
24
25/* For union ocfs2_dlm_lksb */
26#include "stackglue.h"
27
28#include "ocfs2_fs.h"
29#include "ocfs2_lockid.h"
30#include "ocfs2_ioctl.h"
31
32/* For struct ocfs2_blockcheck_stats */
33#include "blockcheck.h"
34
35#include "reservations.h"
36
37#include "filecheck.h"
38
39/* Caching of metadata buffers */
40
41/* Most user visible OCFS2 inodes will have very few pieces of
42 * metadata, but larger files (including bitmaps, etc) must be taken
43 * into account when designing an access scheme. We allow a small
44 * amount of inlined blocks to be stored on an array and grow the
45 * structure into a rb tree when necessary. */
46#define OCFS2_CACHE_INFO_MAX_ARRAY 2
47
48/* Flags for ocfs2_caching_info */
49
50enum ocfs2_caching_info_flags {
51 /* Indicates that the metadata cache is using the inline array */
52 OCFS2_CACHE_FL_INLINE = 1<<1,
53};
54
55struct ocfs2_caching_operations;
56struct ocfs2_caching_info {
57 /*
58 * The parent structure provides the locks, but because the
59 * parent structure can differ, it provides locking operations
60 * to struct ocfs2_caching_info.
61 */
62 const struct ocfs2_caching_operations *ci_ops;
63
64 /* next two are protected by trans_inc_lock */
65 /* which transaction were we created on? Zero if none. */
66 unsigned long ci_created_trans;
67 /* last transaction we were a part of. */
68 unsigned long ci_last_trans;
69
70 /* Cache structures */
71 unsigned int ci_flags;
72 unsigned int ci_num_cached;
73 union {
74 sector_t ci_array[OCFS2_CACHE_INFO_MAX_ARRAY];
75 struct rb_root ci_tree;
76 } ci_cache;
77};
78/*
79 * Need this prototype here instead of in uptodate.h because journal.h
80 * uses it.
81 */
82struct super_block *ocfs2_metadata_cache_get_super(struct ocfs2_caching_info *ci);
83
84/* this limits us to 256 nodes
85 * if we need more, we can do a kmalloc for the map */
86#define OCFS2_NODE_MAP_MAX_NODES 256
87struct ocfs2_node_map {
88 u16 num_nodes;
89 unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)];
90};
91
92enum ocfs2_ast_action {
93 OCFS2_AST_INVALID = 0,
94 OCFS2_AST_ATTACH,
95 OCFS2_AST_CONVERT,
96 OCFS2_AST_DOWNCONVERT,
97};
98
99/* actions for an unlockast function to take. */
100enum ocfs2_unlock_action {
101 OCFS2_UNLOCK_INVALID = 0,
102 OCFS2_UNLOCK_CANCEL_CONVERT,
103 OCFS2_UNLOCK_DROP_LOCK,
104};
105
106/* ocfs2_lock_res->l_flags flags. */
107#define OCFS2_LOCK_ATTACHED (0x00000001) /* we have initialized
108 * the lvb */
109#define OCFS2_LOCK_BUSY (0x00000002) /* we are currently in
110 * dlm_lock */
111#define OCFS2_LOCK_BLOCKED (0x00000004) /* blocked waiting to
112 * downconvert*/
113#define OCFS2_LOCK_LOCAL (0x00000008) /* newly created inode */
114#define OCFS2_LOCK_NEEDS_REFRESH (0x00000010)
115#define OCFS2_LOCK_REFRESHING (0x00000020)
116#define OCFS2_LOCK_INITIALIZED (0x00000040) /* track initialization
117 * for shutdown paths */
118#define OCFS2_LOCK_FREEING (0x00000080) /* help dlmglue track
119 * when to skip queueing
120 * a lock because it's
121 * about to be
122 * dropped. */
123#define OCFS2_LOCK_QUEUED (0x00000100) /* queued for downconvert */
124#define OCFS2_LOCK_NOCACHE (0x00000200) /* don't use a holder count */
125#define OCFS2_LOCK_PENDING (0x00000400) /* This lockres is pending a
126 call to dlm_lock. Only
127 exists with BUSY set. */
128#define OCFS2_LOCK_UPCONVERT_FINISHING (0x00000800) /* blocks the dc thread
129 * from downconverting
130 * before the upconvert
131 * has completed */
132
133#define OCFS2_LOCK_NONBLOCK_FINISHED (0x00001000) /* NONBLOCK cluster
134 * lock has already
135 * returned, do not block
136 * dc thread from
137 * downconverting */
138
139struct ocfs2_lock_res_ops;
140
141typedef void (*ocfs2_lock_callback)(int status, unsigned long data);
142
143#ifdef CONFIG_OCFS2_FS_STATS
144struct ocfs2_lock_stats {
145 u64 ls_total; /* Total wait in NSEC */
146 u32 ls_gets; /* Num acquires */
147 u32 ls_fail; /* Num failed acquires */
148
149 /* Storing max wait in usecs saves 24 bytes per inode */
150 u32 ls_max; /* Max wait in USEC */
151 u64 ls_last; /* Last unlock time in USEC */
152};
153#endif
154
155struct ocfs2_lock_res {
156 void *l_priv;
157 struct ocfs2_lock_res_ops *l_ops;
158
159
160 struct list_head l_blocked_list;
161 struct list_head l_mask_waiters;
162 struct list_head l_holders;
163
164 unsigned long l_flags;
165 char l_name[OCFS2_LOCK_ID_MAX_LEN];
166 unsigned int l_ro_holders;
167 unsigned int l_ex_holders;
168 signed char l_level;
169 signed char l_requested;
170 signed char l_blocking;
171
172 /* Data packed - type enum ocfs2_lock_type */
173 unsigned char l_type;
174
175 /* used from AST/BAST funcs. */
176 /* Data packed - enum type ocfs2_ast_action */
177 unsigned char l_action;
178 /* Data packed - enum type ocfs2_unlock_action */
179 unsigned char l_unlock_action;
180 unsigned int l_pending_gen;
181
182 spinlock_t l_lock;
183
184 struct ocfs2_dlm_lksb l_lksb;
185
186 wait_queue_head_t l_event;
187
188 struct list_head l_debug_list;
189
190#ifdef CONFIG_OCFS2_FS_STATS
191 struct ocfs2_lock_stats l_lock_prmode; /* PR mode stats */
192 u32 l_lock_refresh; /* Disk refreshes */
193 u64 l_lock_wait; /* First lock wait time */
194 struct ocfs2_lock_stats l_lock_exmode; /* EX mode stats */
195#endif
196#ifdef CONFIG_DEBUG_LOCK_ALLOC
197 struct lockdep_map l_lockdep_map;
198#endif
199};
200
201enum ocfs2_orphan_reco_type {
202 ORPHAN_NO_NEED_TRUNCATE = 0,
203 ORPHAN_NEED_TRUNCATE,
204};
205
206enum ocfs2_orphan_scan_state {
207 ORPHAN_SCAN_ACTIVE,
208 ORPHAN_SCAN_INACTIVE
209};
210
211struct ocfs2_orphan_scan {
212 struct mutex os_lock;
213 struct ocfs2_super *os_osb;
214 struct ocfs2_lock_res os_lockres; /* lock to synchronize scans */
215 struct delayed_work os_orphan_scan_work;
216 time64_t os_scantime; /* time this node ran the scan */
217 u32 os_count; /* tracks node specific scans */
218 u32 os_seqno; /* tracks cluster wide scans */
219 atomic_t os_state; /* ACTIVE or INACTIVE */
220};
221
222struct ocfs2_dlm_debug {
223 struct kref d_refcnt;
224 u32 d_filter_secs;
225 struct list_head d_lockres_tracking;
226};
227
228enum ocfs2_vol_state
229{
230 VOLUME_INIT = 0,
231 VOLUME_MOUNTED,
232 VOLUME_MOUNTED_QUOTAS,
233 VOLUME_DISMOUNTED,
234 VOLUME_DISABLED
235};
236
237struct ocfs2_alloc_stats
238{
239 atomic_t moves;
240 atomic_t local_data;
241 atomic_t bitmap_data;
242 atomic_t bg_allocs;
243 atomic_t bg_extends;
244};
245
246enum ocfs2_local_alloc_state
247{
248 OCFS2_LA_UNUSED = 0, /* Local alloc will never be used for
249 * this mountpoint. */
250 OCFS2_LA_ENABLED, /* Local alloc is in use. */
251 OCFS2_LA_THROTTLED, /* Local alloc is in use, but number
252 * of bits has been reduced. */
253 OCFS2_LA_DISABLED /* Local alloc has temporarily been
254 * disabled. */
255};
256
257enum ocfs2_mount_options
258{
259 OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Local heartbeat */
260 OCFS2_MOUNT_BARRIER = 1 << 1, /* Use block barriers */
261 OCFS2_MOUNT_NOINTR = 1 << 2, /* Don't catch signals */
262 OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */
263 OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */
264 OCFS2_MOUNT_LOCALFLOCKS = 1 << 5, /* No cluster aware user file locks */
265 OCFS2_MOUNT_NOUSERXATTR = 1 << 6, /* No user xattr */
266 OCFS2_MOUNT_INODE64 = 1 << 7, /* Allow inode numbers > 2^32 */
267 OCFS2_MOUNT_POSIX_ACL = 1 << 8, /* Force POSIX access control lists */
268 OCFS2_MOUNT_NO_POSIX_ACL = 1 << 9, /* Disable POSIX access
269 control lists */
270 OCFS2_MOUNT_USRQUOTA = 1 << 10, /* We support user quotas */
271 OCFS2_MOUNT_GRPQUOTA = 1 << 11, /* We support group quotas */
272 OCFS2_MOUNT_COHERENCY_BUFFERED = 1 << 12, /* Allow concurrent O_DIRECT
273 writes */
274 OCFS2_MOUNT_HB_NONE = 1 << 13, /* No heartbeat */
275 OCFS2_MOUNT_HB_GLOBAL = 1 << 14, /* Global heartbeat */
276
277 OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT = 1 << 15, /* Journal Async Commit */
278 OCFS2_MOUNT_ERRORS_CONT = 1 << 16, /* Return EIO to the calling process on error */
279 OCFS2_MOUNT_ERRORS_ROFS = 1 << 17, /* Change filesystem to read-only on error */
280};
281
282#define OCFS2_OSB_SOFT_RO 0x0001
283#define OCFS2_OSB_HARD_RO 0x0002
284#define OCFS2_OSB_ERROR_FS 0x0004
285#define OCFS2_DEFAULT_ATIME_QUANTUM 60
286
287struct ocfs2_journal;
288struct ocfs2_slot_info;
289struct ocfs2_recovery_map;
290struct ocfs2_replay_map;
291struct ocfs2_quota_recovery;
292struct ocfs2_super
293{
294 struct task_struct *commit_task;
295 struct super_block *sb;
296 struct inode *root_inode;
297 struct inode *sys_root_inode;
298 struct inode *global_system_inodes[NUM_GLOBAL_SYSTEM_INODES];
299 struct inode **local_system_inodes;
300
301 struct ocfs2_slot_info *slot_info;
302
303 u32 *slot_recovery_generations;
304
305 spinlock_t node_map_lock;
306
307 u64 root_blkno;
308 u64 system_dir_blkno;
309 u64 bitmap_blkno;
310 u32 bitmap_cpg;
311 char *uuid_str;
312 u32 uuid_hash;
313 u8 *vol_label;
314 u64 first_cluster_group_blkno;
315 u32 fs_generation;
316
317 u32 s_feature_compat;
318 u32 s_feature_incompat;
319 u32 s_feature_ro_compat;
320
321 /* Protects s_next_generation, osb_flags and s_inode_steal_slot.
322 * Could protect more on osb as it's very short lived.
323 */
324 spinlock_t osb_lock;
325 u32 s_next_generation;
326 unsigned long osb_flags;
327 u16 s_inode_steal_slot;
328 u16 s_meta_steal_slot;
329 atomic_t s_num_inodes_stolen;
330 atomic_t s_num_meta_stolen;
331
332 unsigned long s_mount_opt;
333 unsigned int s_atime_quantum;
334
335 unsigned int max_slots;
336 unsigned int node_num;
337 int slot_num;
338 int preferred_slot;
339 int s_sectsize_bits;
340 int s_clustersize;
341 int s_clustersize_bits;
342 unsigned int s_xattr_inline_size;
343
344 atomic_t vol_state;
345 struct mutex recovery_lock;
346 struct ocfs2_recovery_map *recovery_map;
347 struct ocfs2_replay_map *replay_map;
348 struct task_struct *recovery_thread_task;
349 int disable_recovery;
350 wait_queue_head_t checkpoint_event;
351 struct ocfs2_journal *journal;
352 unsigned long osb_commit_interval;
353
354 struct delayed_work la_enable_wq;
355
356 /*
357 * Must hold local alloc i_rwsem and osb->osb_lock to change
358 * local_alloc_bits. Reads can be done under either lock.
359 */
360 unsigned int local_alloc_bits;
361 unsigned int local_alloc_default_bits;
362 /* osb_clusters_at_boot can become stale! Do not trust it to
363 * be up to date. */
364 unsigned int osb_clusters_at_boot;
365
366 enum ocfs2_local_alloc_state local_alloc_state; /* protected
367 * by osb_lock */
368
369 struct buffer_head *local_alloc_bh;
370
371 u64 la_last_gd;
372
373 struct ocfs2_reservation_map osb_la_resmap;
374
375 unsigned int osb_resv_level;
376 unsigned int osb_dir_resv_level;
377
378 /* Next two fields are for local node slot recovery during
379 * mount. */
380 struct ocfs2_dinode *local_alloc_copy;
381 struct ocfs2_quota_recovery *quota_rec;
382
383 struct ocfs2_blockcheck_stats osb_ecc_stats;
384 struct ocfs2_alloc_stats alloc_stats;
385 char dev_str[20]; /* "major,minor" of the device */
386
387 u8 osb_stackflags;
388
389 char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
390 char osb_cluster_name[OCFS2_CLUSTER_NAME_LEN + 1];
391 struct ocfs2_cluster_connection *cconn;
392 struct ocfs2_lock_res osb_super_lockres;
393 struct ocfs2_lock_res osb_rename_lockres;
394 struct ocfs2_lock_res osb_nfs_sync_lockres;
395 struct rw_semaphore nfs_sync_rwlock;
396 struct ocfs2_lock_res osb_trim_fs_lockres;
397 struct mutex obs_trim_fs_mutex;
398 struct ocfs2_dlm_debug *osb_dlm_debug;
399
400 struct dentry *osb_debug_root;
401
402 wait_queue_head_t recovery_event;
403
404 spinlock_t dc_task_lock;
405 struct task_struct *dc_task;
406 wait_queue_head_t dc_event;
407 unsigned long dc_wake_sequence;
408 unsigned long dc_work_sequence;
409
410 /*
411 * Any thread can add locks to the list, but the downconvert
412 * thread is the only one allowed to remove locks. Any change
413 * to this rule requires updating
414 * ocfs2_downconvert_thread_do_work().
415 */
416 struct list_head blocked_lock_list;
417 unsigned long blocked_lock_count;
418
419 /* List of dquot structures to drop last reference to */
420 struct llist_head dquot_drop_list;
421 struct work_struct dquot_drop_work;
422
423 wait_queue_head_t osb_mount_event;
424
425 /* Truncate log info */
426 struct inode *osb_tl_inode;
427 struct buffer_head *osb_tl_bh;
428 struct delayed_work osb_truncate_log_wq;
429 atomic_t osb_tl_disable;
430 /*
431 * How many clusters in our truncate log.
432 * It must be protected by osb_tl_inode->i_rwsem.
433 */
434 unsigned int truncated_clusters;
435
436 struct ocfs2_node_map osb_recovering_orphan_dirs;
437 unsigned int *osb_orphan_wipes;
438 wait_queue_head_t osb_wipe_event;
439
440 struct ocfs2_orphan_scan osb_orphan_scan;
441
442 /* used to protect metaecc calculation check of xattr. */
443 spinlock_t osb_xattr_lock;
444
445 unsigned int osb_dx_mask;
446 u32 osb_dx_seed[4];
447
448 /* the group we used to allocate inodes. */
449 u64 osb_inode_alloc_group;
450
451 /* rb tree root for refcount lock. */
452 struct rb_root osb_rf_lock_tree;
453 struct ocfs2_refcount_tree *osb_ref_tree_lru;
454
455 struct mutex system_file_mutex;
456
457 /*
458 * OCFS2 needs to schedule several different types of work which
459 * require cluster locking, disk I/O, recovery waits, etc. Since these
460 * types of work tend to be heavy we avoid using the kernel events
461 * workqueue and schedule on our own.
462 */
463 struct workqueue_struct *ocfs2_wq;
464
465 /* sysfs directory per partition */
466 struct kset *osb_dev_kset;
467
468 /* file check related stuff */
469 struct ocfs2_filecheck_sysfs_entry osb_fc_ent;
470};
471
472#define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info)
473
474/* Useful typedef for passing around journal access functions */
475typedef int (*ocfs2_journal_access_func)(handle_t *handle,
476 struct ocfs2_caching_info *ci,
477 struct buffer_head *bh, int type);
478
479static inline int ocfs2_should_order_data(struct inode *inode)
480{
481 if (!S_ISREG(inode->i_mode))
482 return 0;
483 if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
484 return 0;
485 return 1;
486}
487
488static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb)
489{
490 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC)
491 return 1;
492 return 0;
493}
494
495static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb)
496{
497 /*
498 * Support for sparse files is a pre-requisite
499 */
500 if (!ocfs2_sparse_alloc(osb))
501 return 0;
502
503 if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN)
504 return 1;
505 return 0;
506}
507
508static inline int ocfs2_supports_append_dio(struct ocfs2_super *osb)
509{
510 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_APPEND_DIO)
511 return 1;
512 return 0;
513}
514
515
516static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb)
517{
518 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA)
519 return 1;
520 return 0;
521}
522
523static inline int ocfs2_supports_xattr(struct ocfs2_super *osb)
524{
525 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR)
526 return 1;
527 return 0;
528}
529
530static inline int ocfs2_meta_ecc(struct ocfs2_super *osb)
531{
532 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_META_ECC)
533 return 1;
534 return 0;
535}
536
537static inline int ocfs2_supports_indexed_dirs(struct ocfs2_super *osb)
538{
539 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INDEXED_DIRS)
540 return 1;
541 return 0;
542}
543
544static inline int ocfs2_supports_discontig_bg(struct ocfs2_super *osb)
545{
546 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG)
547 return 1;
548 return 0;
549}
550
551static inline unsigned int ocfs2_link_max(struct ocfs2_super *osb)
552{
553 if (ocfs2_supports_indexed_dirs(osb))
554 return OCFS2_DX_LINK_MAX;
555 return OCFS2_LINK_MAX;
556}
557
558static inline unsigned int ocfs2_read_links_count(struct ocfs2_dinode *di)
559{
560 u32 nlink = le16_to_cpu(di->i_links_count);
561 u32 hi = le16_to_cpu(di->i_links_count_hi);
562
563 nlink |= (hi << OCFS2_LINKS_HI_SHIFT);
564
565 return nlink;
566}
567
568static inline void ocfs2_set_links_count(struct ocfs2_dinode *di, u32 nlink)
569{
570 u16 lo, hi;
571
572 lo = nlink;
573 hi = nlink >> OCFS2_LINKS_HI_SHIFT;
574
575 di->i_links_count = cpu_to_le16(lo);
576 di->i_links_count_hi = cpu_to_le16(hi);
577}
578
579static inline void ocfs2_add_links_count(struct ocfs2_dinode *di, int n)
580{
581 u32 links = ocfs2_read_links_count(di);
582
583 links += n;
584
585 ocfs2_set_links_count(di, links);
586}
587
588static inline int ocfs2_refcount_tree(struct ocfs2_super *osb)
589{
590 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE)
591 return 1;
592 return 0;
593}
594
595/* set / clear functions because cluster events can make these happen
596 * in parallel so we want the transitions to be atomic. this also
597 * means that any future flags osb_flags must be protected by spinlock
598 * too! */
599static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
600 unsigned long flag)
601{
602 spin_lock(&osb->osb_lock);
603 osb->osb_flags |= flag;
604 spin_unlock(&osb->osb_lock);
605}
606
607static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
608 int hard)
609{
610 spin_lock(&osb->osb_lock);
611 osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
612 if (hard)
613 osb->osb_flags |= OCFS2_OSB_HARD_RO;
614 else
615 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
616 spin_unlock(&osb->osb_lock);
617}
618
619static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
620{
621 int ret;
622
623 spin_lock(&osb->osb_lock);
624 ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
625 spin_unlock(&osb->osb_lock);
626
627 return ret;
628}
629
630static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
631{
632 int ret;
633
634 spin_lock(&osb->osb_lock);
635 ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
636 spin_unlock(&osb->osb_lock);
637
638 return ret;
639}
640
641static inline int ocfs2_clusterinfo_valid(struct ocfs2_super *osb)
642{
643 return (osb->s_feature_incompat &
644 (OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK |
645 OCFS2_FEATURE_INCOMPAT_CLUSTERINFO));
646}
647
648static inline int ocfs2_userspace_stack(struct ocfs2_super *osb)
649{
650 if (ocfs2_clusterinfo_valid(osb) &&
651 memcmp(osb->osb_cluster_stack, OCFS2_CLASSIC_CLUSTER_STACK,
652 OCFS2_STACK_LABEL_LEN))
653 return 1;
654 return 0;
655}
656
657static inline int ocfs2_o2cb_stack(struct ocfs2_super *osb)
658{
659 if (ocfs2_clusterinfo_valid(osb) &&
660 !memcmp(osb->osb_cluster_stack, OCFS2_CLASSIC_CLUSTER_STACK,
661 OCFS2_STACK_LABEL_LEN))
662 return 1;
663 return 0;
664}
665
666static inline int ocfs2_cluster_o2cb_global_heartbeat(struct ocfs2_super *osb)
667{
668 return ocfs2_o2cb_stack(osb) &&
669 (osb->osb_stackflags & OCFS2_CLUSTER_O2CB_GLOBAL_HEARTBEAT);
670}
671
672static inline int ocfs2_mount_local(struct ocfs2_super *osb)
673{
674 return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT);
675}
676
677static inline int ocfs2_uses_extended_slot_map(struct ocfs2_super *osb)
678{
679 return (osb->s_feature_incompat &
680 OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP);
681}
682
683
684#define OCFS2_IS_VALID_DINODE(ptr) \
685 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
686
687#define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \
688 (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
689
690#define OCFS2_IS_VALID_GROUP_DESC(ptr) \
691 (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
692
693
694#define OCFS2_IS_VALID_XATTR_BLOCK(ptr) \
695 (!strcmp((ptr)->xb_signature, OCFS2_XATTR_BLOCK_SIGNATURE))
696
697#define OCFS2_IS_VALID_DIR_TRAILER(ptr) \
698 (!strcmp((ptr)->db_signature, OCFS2_DIR_TRAILER_SIGNATURE))
699
700#define OCFS2_IS_VALID_DX_ROOT(ptr) \
701 (!strcmp((ptr)->dr_signature, OCFS2_DX_ROOT_SIGNATURE))
702
703#define OCFS2_IS_VALID_DX_LEAF(ptr) \
704 (!strcmp((ptr)->dl_signature, OCFS2_DX_LEAF_SIGNATURE))
705
706#define OCFS2_IS_VALID_REFCOUNT_BLOCK(ptr) \
707 (!strcmp((ptr)->rf_signature, OCFS2_REFCOUNT_BLOCK_SIGNATURE))
708
709static inline unsigned long ino_from_blkno(struct super_block *sb,
710 u64 blkno)
711{
712 return (unsigned long)(blkno & (u64)ULONG_MAX);
713}
714
715static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
716 u32 clusters)
717{
718 int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
719 sb->s_blocksize_bits;
720
721 return (u64)clusters << c_to_b_bits;
722}
723
724static inline u32 ocfs2_clusters_for_blocks(struct super_block *sb,
725 u64 blocks)
726{
727 int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
728 sb->s_blocksize_bits;
729
730 blocks += (1 << b_to_c_bits) - 1;
731 return (u32)(blocks >> b_to_c_bits);
732}
733
734static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
735 u64 blocks)
736{
737 int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
738 sb->s_blocksize_bits;
739
740 return (u32)(blocks >> b_to_c_bits);
741}
742
743static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
744 u64 bytes)
745{
746 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
747 unsigned int clusters;
748
749 bytes += OCFS2_SB(sb)->s_clustersize - 1;
750 /* OCFS2 just cannot have enough clusters to overflow this */
751 clusters = (unsigned int)(bytes >> cl_bits);
752
753 return clusters;
754}
755
756static inline unsigned int ocfs2_bytes_to_clusters(struct super_block *sb,
757 u64 bytes)
758{
759 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
760 unsigned int clusters;
761
762 clusters = (unsigned int)(bytes >> cl_bits);
763 return clusters;
764}
765
766static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
767 u64 bytes)
768{
769 bytes += sb->s_blocksize - 1;
770 return bytes >> sb->s_blocksize_bits;
771}
772
773static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
774 u32 clusters)
775{
776 return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
777}
778
779static inline u64 ocfs2_block_to_cluster_start(struct super_block *sb,
780 u64 blocks)
781{
782 int bits = OCFS2_SB(sb)->s_clustersize_bits - sb->s_blocksize_bits;
783 unsigned int clusters;
784
785 clusters = ocfs2_blocks_to_clusters(sb, blocks);
786 return (u64)clusters << bits;
787}
788
789static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
790 u64 bytes)
791{
792 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
793 unsigned int clusters;
794
795 clusters = ocfs2_clusters_for_bytes(sb, bytes);
796 return (u64)clusters << cl_bits;
797}
798
799static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
800 u64 bytes)
801{
802 u64 blocks;
803
804 blocks = ocfs2_blocks_for_bytes(sb, bytes);
805 return blocks << sb->s_blocksize_bits;
806}
807
808static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
809{
810 return (unsigned long)((bytes + 511) >> 9);
811}
812
813static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb,
814 unsigned long pg_index)
815{
816 u32 clusters = pg_index;
817 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
818
819 if (unlikely(PAGE_SHIFT > cbits))
820 clusters = pg_index << (PAGE_SHIFT - cbits);
821 else if (PAGE_SHIFT < cbits)
822 clusters = pg_index >> (cbits - PAGE_SHIFT);
823
824 return clusters;
825}
826
827/*
828 * Find the 1st page index which covers the given clusters.
829 */
830static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb,
831 u32 clusters)
832{
833 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
834 pgoff_t index = clusters;
835
836 if (PAGE_SHIFT > cbits) {
837 index = (pgoff_t)clusters >> (PAGE_SHIFT - cbits);
838 } else if (PAGE_SHIFT < cbits) {
839 index = (pgoff_t)clusters << (cbits - PAGE_SHIFT);
840 }
841
842 return index;
843}
844
845static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb)
846{
847 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
848 unsigned int pages_per_cluster = 1;
849
850 if (PAGE_SHIFT < cbits)
851 pages_per_cluster = 1 << (cbits - PAGE_SHIFT);
852
853 return pages_per_cluster;
854}
855
856static inline unsigned int ocfs2_megabytes_to_clusters(struct super_block *sb,
857 unsigned int megs)
858{
859 BUILD_BUG_ON(OCFS2_MAX_CLUSTERSIZE > 1048576);
860
861 return megs << (20 - OCFS2_SB(sb)->s_clustersize_bits);
862}
863
864static inline unsigned int ocfs2_clusters_to_megabytes(struct super_block *sb,
865 unsigned int clusters)
866{
867 return clusters >> (20 - OCFS2_SB(sb)->s_clustersize_bits);
868}
869
870static inline void _ocfs2_set_bit(unsigned int bit, unsigned long *bitmap)
871{
872 __set_bit_le(bit, bitmap);
873}
874#define ocfs2_set_bit(bit, addr) _ocfs2_set_bit((bit), (unsigned long *)(addr))
875
876static inline void _ocfs2_clear_bit(unsigned int bit, unsigned long *bitmap)
877{
878 __clear_bit_le(bit, bitmap);
879}
880#define ocfs2_clear_bit(bit, addr) _ocfs2_clear_bit((bit), (unsigned long *)(addr))
881
882#define ocfs2_test_bit test_bit_le
883#define ocfs2_find_next_zero_bit find_next_zero_bit_le
884#define ocfs2_find_next_bit find_next_bit_le
885
886static inline void *correct_addr_and_bit_unaligned(int *bit, void *addr)
887{
888#if BITS_PER_LONG == 64
889 *bit += ((unsigned long) addr & 7UL) << 3;
890 addr = (void *) ((unsigned long) addr & ~7UL);
891#elif BITS_PER_LONG == 32
892 *bit += ((unsigned long) addr & 3UL) << 3;
893 addr = (void *) ((unsigned long) addr & ~3UL);
894#else
895#error "how many bits you are?!"
896#endif
897 return addr;
898}
899
900static inline void ocfs2_set_bit_unaligned(int bit, void *bitmap)
901{
902 bitmap = correct_addr_and_bit_unaligned(&bit, bitmap);
903 ocfs2_set_bit(bit, bitmap);
904}
905
906static inline void ocfs2_clear_bit_unaligned(int bit, void *bitmap)
907{
908 bitmap = correct_addr_and_bit_unaligned(&bit, bitmap);
909 ocfs2_clear_bit(bit, bitmap);
910}
911
912static inline int ocfs2_test_bit_unaligned(int bit, void *bitmap)
913{
914 bitmap = correct_addr_and_bit_unaligned(&bit, bitmap);
915 return ocfs2_test_bit(bit, bitmap);
916}
917
918static inline int ocfs2_find_next_zero_bit_unaligned(void *bitmap, int max,
919 int start)
920{
921 int fix = 0, ret, tmpmax;
922 bitmap = correct_addr_and_bit_unaligned(&fix, bitmap);
923 tmpmax = max + fix;
924 start += fix;
925
926 ret = ocfs2_find_next_zero_bit(bitmap, tmpmax, start) - fix;
927 if (ret > max)
928 return max;
929 return ret;
930}
931
932#endif /* OCFS2_H */
933
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/llist.h>
34#include <linux/rbtree.h>
35#include <linux/workqueue.h>
36#include <linux/kref.h>
37#include <linux/mutex.h>
38#include <linux/lockdep.h>
39#include <linux/jbd2.h>
40
41/* For union ocfs2_dlm_lksb */
42#include "stackglue.h"
43
44#include "ocfs2_fs.h"
45#include "ocfs2_lockid.h"
46#include "ocfs2_ioctl.h"
47
48/* For struct ocfs2_blockcheck_stats */
49#include "blockcheck.h"
50
51#include "reservations.h"
52
53/* Caching of metadata buffers */
54
55/* Most user visible OCFS2 inodes will have very few pieces of
56 * metadata, but larger files (including bitmaps, etc) must be taken
57 * into account when designing an access scheme. We allow a small
58 * amount of inlined blocks to be stored on an array and grow the
59 * structure into a rb tree when necessary. */
60#define OCFS2_CACHE_INFO_MAX_ARRAY 2
61
62/* Flags for ocfs2_caching_info */
63
64enum ocfs2_caching_info_flags {
65 /* Indicates that the metadata cache is using the inline array */
66 OCFS2_CACHE_FL_INLINE = 1<<1,
67};
68
69struct ocfs2_caching_operations;
70struct ocfs2_caching_info {
71 /*
72 * The parent structure provides the locks, but because the
73 * parent structure can differ, it provides locking operations
74 * to struct ocfs2_caching_info.
75 */
76 const struct ocfs2_caching_operations *ci_ops;
77
78 /* next two are protected by trans_inc_lock */
79 /* which transaction were we created on? Zero if none. */
80 unsigned long ci_created_trans;
81 /* last transaction we were a part of. */
82 unsigned long ci_last_trans;
83
84 /* Cache structures */
85 unsigned int ci_flags;
86 unsigned int ci_num_cached;
87 union {
88 sector_t ci_array[OCFS2_CACHE_INFO_MAX_ARRAY];
89 struct rb_root ci_tree;
90 } ci_cache;
91};
92/*
93 * Need this prototype here instead of in uptodate.h because journal.h
94 * uses it.
95 */
96struct super_block *ocfs2_metadata_cache_get_super(struct ocfs2_caching_info *ci);
97
98/* this limits us to 256 nodes
99 * if we need more, we can do a kmalloc for the map */
100#define OCFS2_NODE_MAP_MAX_NODES 256
101struct ocfs2_node_map {
102 u16 num_nodes;
103 unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)];
104};
105
106enum ocfs2_ast_action {
107 OCFS2_AST_INVALID = 0,
108 OCFS2_AST_ATTACH,
109 OCFS2_AST_CONVERT,
110 OCFS2_AST_DOWNCONVERT,
111};
112
113/* actions for an unlockast function to take. */
114enum ocfs2_unlock_action {
115 OCFS2_UNLOCK_INVALID = 0,
116 OCFS2_UNLOCK_CANCEL_CONVERT,
117 OCFS2_UNLOCK_DROP_LOCK,
118};
119
120/* ocfs2_lock_res->l_flags flags. */
121#define OCFS2_LOCK_ATTACHED (0x00000001) /* we have initialized
122 * the lvb */
123#define OCFS2_LOCK_BUSY (0x00000002) /* we are currently in
124 * dlm_lock */
125#define OCFS2_LOCK_BLOCKED (0x00000004) /* blocked waiting to
126 * downconvert*/
127#define OCFS2_LOCK_LOCAL (0x00000008) /* newly created inode */
128#define OCFS2_LOCK_NEEDS_REFRESH (0x00000010)
129#define OCFS2_LOCK_REFRESHING (0x00000020)
130#define OCFS2_LOCK_INITIALIZED (0x00000040) /* track initialization
131 * for shutdown paths */
132#define OCFS2_LOCK_FREEING (0x00000080) /* help dlmglue track
133 * when to skip queueing
134 * a lock because it's
135 * about to be
136 * dropped. */
137#define OCFS2_LOCK_QUEUED (0x00000100) /* queued for downconvert */
138#define OCFS2_LOCK_NOCACHE (0x00000200) /* don't use a holder count */
139#define OCFS2_LOCK_PENDING (0x00000400) /* This lockres is pending a
140 call to dlm_lock. Only
141 exists with BUSY set. */
142#define OCFS2_LOCK_UPCONVERT_FINISHING (0x00000800) /* blocks the dc thread
143 * from downconverting
144 * before the upconvert
145 * has completed */
146
147#define OCFS2_LOCK_NONBLOCK_FINISHED (0x00001000) /* NONBLOCK cluster
148 * lock has already
149 * returned, do not block
150 * dc thread from
151 * downconverting */
152
153struct ocfs2_lock_res_ops;
154
155typedef void (*ocfs2_lock_callback)(int status, unsigned long data);
156
157#ifdef CONFIG_OCFS2_FS_STATS
158struct ocfs2_lock_stats {
159 u64 ls_total; /* Total wait in NSEC */
160 u32 ls_gets; /* Num acquires */
161 u32 ls_fail; /* Num failed acquires */
162
163 /* Storing max wait in usecs saves 24 bytes per inode */
164 u32 ls_max; /* Max wait in USEC */
165};
166#endif
167
168struct ocfs2_lock_res {
169 void *l_priv;
170 struct ocfs2_lock_res_ops *l_ops;
171
172
173 struct list_head l_blocked_list;
174 struct list_head l_mask_waiters;
175
176 unsigned long l_flags;
177 char l_name[OCFS2_LOCK_ID_MAX_LEN];
178 unsigned int l_ro_holders;
179 unsigned int l_ex_holders;
180 signed char l_level;
181 signed char l_requested;
182 signed char l_blocking;
183
184 /* Data packed - type enum ocfs2_lock_type */
185 unsigned char l_type;
186
187 /* used from AST/BAST funcs. */
188 /* Data packed - enum type ocfs2_ast_action */
189 unsigned char l_action;
190 /* Data packed - enum type ocfs2_unlock_action */
191 unsigned char l_unlock_action;
192 unsigned int l_pending_gen;
193
194 spinlock_t l_lock;
195
196 struct ocfs2_dlm_lksb l_lksb;
197
198 wait_queue_head_t l_event;
199
200 struct list_head l_debug_list;
201
202#ifdef CONFIG_OCFS2_FS_STATS
203 struct ocfs2_lock_stats l_lock_prmode; /* PR mode stats */
204 u32 l_lock_refresh; /* Disk refreshes */
205 struct ocfs2_lock_stats l_lock_exmode; /* EX mode stats */
206#endif
207#ifdef CONFIG_DEBUG_LOCK_ALLOC
208 struct lockdep_map l_lockdep_map;
209#endif
210};
211
212enum ocfs2_orphan_reco_type {
213 ORPHAN_NO_NEED_TRUNCATE = 0,
214 ORPHAN_NEED_TRUNCATE,
215};
216
217enum ocfs2_orphan_scan_state {
218 ORPHAN_SCAN_ACTIVE,
219 ORPHAN_SCAN_INACTIVE
220};
221
222struct ocfs2_orphan_scan {
223 struct mutex os_lock;
224 struct ocfs2_super *os_osb;
225 struct ocfs2_lock_res os_lockres; /* lock to synchronize scans */
226 struct delayed_work os_orphan_scan_work;
227 struct timespec os_scantime; /* time this node ran the scan */
228 u32 os_count; /* tracks node specific scans */
229 u32 os_seqno; /* tracks cluster wide scans */
230 atomic_t os_state; /* ACTIVE or INACTIVE */
231};
232
233struct ocfs2_dlm_debug {
234 struct kref d_refcnt;
235 struct dentry *d_locking_state;
236 struct list_head d_lockres_tracking;
237};
238
239enum ocfs2_vol_state
240{
241 VOLUME_INIT = 0,
242 VOLUME_MOUNTED,
243 VOLUME_MOUNTED_QUOTAS,
244 VOLUME_DISMOUNTED,
245 VOLUME_DISABLED
246};
247
248struct ocfs2_alloc_stats
249{
250 atomic_t moves;
251 atomic_t local_data;
252 atomic_t bitmap_data;
253 atomic_t bg_allocs;
254 atomic_t bg_extends;
255};
256
257enum ocfs2_local_alloc_state
258{
259 OCFS2_LA_UNUSED = 0, /* Local alloc will never be used for
260 * this mountpoint. */
261 OCFS2_LA_ENABLED, /* Local alloc is in use. */
262 OCFS2_LA_THROTTLED, /* Local alloc is in use, but number
263 * of bits has been reduced. */
264 OCFS2_LA_DISABLED /* Local alloc has temporarily been
265 * disabled. */
266};
267
268enum ocfs2_mount_options
269{
270 OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Local heartbeat */
271 OCFS2_MOUNT_BARRIER = 1 << 1, /* Use block barriers */
272 OCFS2_MOUNT_NOINTR = 1 << 2, /* Don't catch signals */
273 OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */
274 OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */
275 OCFS2_MOUNT_LOCALFLOCKS = 1 << 5, /* No cluster aware user file locks */
276 OCFS2_MOUNT_NOUSERXATTR = 1 << 6, /* No user xattr */
277 OCFS2_MOUNT_INODE64 = 1 << 7, /* Allow inode numbers > 2^32 */
278 OCFS2_MOUNT_POSIX_ACL = 1 << 8, /* Force POSIX access control lists */
279 OCFS2_MOUNT_NO_POSIX_ACL = 1 << 9, /* Disable POSIX access
280 control lists */
281 OCFS2_MOUNT_USRQUOTA = 1 << 10, /* We support user quotas */
282 OCFS2_MOUNT_GRPQUOTA = 1 << 11, /* We support group quotas */
283 OCFS2_MOUNT_COHERENCY_BUFFERED = 1 << 12, /* Allow concurrent O_DIRECT
284 writes */
285 OCFS2_MOUNT_HB_NONE = 1 << 13, /* No heartbeat */
286 OCFS2_MOUNT_HB_GLOBAL = 1 << 14, /* Global heartbeat */
287
288 OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT = 1 << 15, /* Journal Async Commit */
289 OCFS2_MOUNT_ERRORS_CONT = 1 << 16, /* Return EIO to the calling process on error */
290 OCFS2_MOUNT_ERRORS_ROFS = 1 << 17, /* Change filesystem to read-only on error */
291};
292
293#define OCFS2_OSB_SOFT_RO 0x0001
294#define OCFS2_OSB_HARD_RO 0x0002
295#define OCFS2_OSB_ERROR_FS 0x0004
296#define OCFS2_DEFAULT_ATIME_QUANTUM 60
297
298struct ocfs2_journal;
299struct ocfs2_slot_info;
300struct ocfs2_recovery_map;
301struct ocfs2_replay_map;
302struct ocfs2_quota_recovery;
303struct ocfs2_super
304{
305 struct task_struct *commit_task;
306 struct super_block *sb;
307 struct inode *root_inode;
308 struct inode *sys_root_inode;
309 struct inode *global_system_inodes[NUM_GLOBAL_SYSTEM_INODES];
310 struct inode **local_system_inodes;
311
312 struct ocfs2_slot_info *slot_info;
313
314 u32 *slot_recovery_generations;
315
316 spinlock_t node_map_lock;
317
318 u64 root_blkno;
319 u64 system_dir_blkno;
320 u64 bitmap_blkno;
321 u32 bitmap_cpg;
322 u8 *uuid;
323 char *uuid_str;
324 u32 uuid_hash;
325 u8 *vol_label;
326 u64 first_cluster_group_blkno;
327 u32 fs_generation;
328
329 u32 s_feature_compat;
330 u32 s_feature_incompat;
331 u32 s_feature_ro_compat;
332
333 /* Protects s_next_generation, osb_flags and s_inode_steal_slot.
334 * Could protect more on osb as it's very short lived.
335 */
336 spinlock_t osb_lock;
337 u32 s_next_generation;
338 unsigned long osb_flags;
339 s16 s_inode_steal_slot;
340 s16 s_meta_steal_slot;
341 atomic_t s_num_inodes_stolen;
342 atomic_t s_num_meta_stolen;
343
344 unsigned long s_mount_opt;
345 unsigned int s_atime_quantum;
346
347 unsigned int max_slots;
348 unsigned int node_num;
349 int slot_num;
350 int preferred_slot;
351 int s_sectsize_bits;
352 int s_clustersize;
353 int s_clustersize_bits;
354 unsigned int s_xattr_inline_size;
355
356 atomic_t vol_state;
357 struct mutex recovery_lock;
358 struct ocfs2_recovery_map *recovery_map;
359 struct ocfs2_replay_map *replay_map;
360 struct task_struct *recovery_thread_task;
361 int disable_recovery;
362 wait_queue_head_t checkpoint_event;
363 struct ocfs2_journal *journal;
364 unsigned long osb_commit_interval;
365
366 struct delayed_work la_enable_wq;
367
368 /*
369 * Must hold local alloc i_mutex and osb->osb_lock to change
370 * local_alloc_bits. Reads can be done under either lock.
371 */
372 unsigned int local_alloc_bits;
373 unsigned int local_alloc_default_bits;
374 /* osb_clusters_at_boot can become stale! Do not trust it to
375 * be up to date. */
376 unsigned int osb_clusters_at_boot;
377
378 enum ocfs2_local_alloc_state local_alloc_state; /* protected
379 * by osb_lock */
380
381 struct buffer_head *local_alloc_bh;
382
383 u64 la_last_gd;
384
385 struct ocfs2_reservation_map osb_la_resmap;
386
387 unsigned int osb_resv_level;
388 unsigned int osb_dir_resv_level;
389
390 /* Next three fields are for local node slot recovery during
391 * mount. */
392 int dirty;
393 struct ocfs2_dinode *local_alloc_copy;
394 struct ocfs2_quota_recovery *quota_rec;
395
396 struct ocfs2_blockcheck_stats osb_ecc_stats;
397 struct ocfs2_alloc_stats alloc_stats;
398 char dev_str[20]; /* "major,minor" of the device */
399
400 u8 osb_stackflags;
401
402 char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
403 char osb_cluster_name[OCFS2_CLUSTER_NAME_LEN + 1];
404 struct ocfs2_cluster_connection *cconn;
405 struct ocfs2_lock_res osb_super_lockres;
406 struct ocfs2_lock_res osb_rename_lockres;
407 struct ocfs2_lock_res osb_nfs_sync_lockres;
408 struct ocfs2_dlm_debug *osb_dlm_debug;
409
410 struct dentry *osb_debug_root;
411 struct dentry *osb_ctxt;
412
413 wait_queue_head_t recovery_event;
414
415 spinlock_t dc_task_lock;
416 struct task_struct *dc_task;
417 wait_queue_head_t dc_event;
418 unsigned long dc_wake_sequence;
419 unsigned long dc_work_sequence;
420
421 /*
422 * Any thread can add locks to the list, but the downconvert
423 * thread is the only one allowed to remove locks. Any change
424 * to this rule requires updating
425 * ocfs2_downconvert_thread_do_work().
426 */
427 struct list_head blocked_lock_list;
428 unsigned long blocked_lock_count;
429
430 /* List of dquot structures to drop last reference to */
431 struct llist_head dquot_drop_list;
432 struct work_struct dquot_drop_work;
433
434 wait_queue_head_t osb_mount_event;
435
436 /* Truncate log info */
437 struct inode *osb_tl_inode;
438 struct buffer_head *osb_tl_bh;
439 struct delayed_work osb_truncate_log_wq;
440 atomic_t osb_tl_disable;
441 /*
442 * How many clusters in our truncate log.
443 * It must be protected by osb_tl_inode->i_mutex.
444 */
445 unsigned int truncated_clusters;
446
447 struct ocfs2_node_map osb_recovering_orphan_dirs;
448 unsigned int *osb_orphan_wipes;
449 wait_queue_head_t osb_wipe_event;
450
451 struct ocfs2_orphan_scan osb_orphan_scan;
452
453 /* used to protect metaecc calculation check of xattr. */
454 spinlock_t osb_xattr_lock;
455
456 unsigned int osb_dx_mask;
457 u32 osb_dx_seed[4];
458
459 /* the group we used to allocate inodes. */
460 u64 osb_inode_alloc_group;
461
462 /* rb tree root for refcount lock. */
463 struct rb_root osb_rf_lock_tree;
464 struct ocfs2_refcount_tree *osb_ref_tree_lru;
465
466 struct mutex system_file_mutex;
467
468 /*
469 * OCFS2 needs to schedule several different types of work which
470 * require cluster locking, disk I/O, recovery waits, etc. Since these
471 * types of work tend to be heavy we avoid using the kernel events
472 * workqueue and schedule on our own.
473 */
474 struct workqueue_struct *ocfs2_wq;
475};
476
477#define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info)
478
479/* Useful typedef for passing around journal access functions */
480typedef int (*ocfs2_journal_access_func)(handle_t *handle,
481 struct ocfs2_caching_info *ci,
482 struct buffer_head *bh, int type);
483
484static inline int ocfs2_should_order_data(struct inode *inode)
485{
486 if (!S_ISREG(inode->i_mode))
487 return 0;
488 if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
489 return 0;
490 return 1;
491}
492
493static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb)
494{
495 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC)
496 return 1;
497 return 0;
498}
499
500static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb)
501{
502 /*
503 * Support for sparse files is a pre-requisite
504 */
505 if (!ocfs2_sparse_alloc(osb))
506 return 0;
507
508 if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN)
509 return 1;
510 return 0;
511}
512
513static inline int ocfs2_supports_append_dio(struct ocfs2_super *osb)
514{
515 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_APPEND_DIO)
516 return 1;
517 return 0;
518}
519
520
521static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb)
522{
523 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA)
524 return 1;
525 return 0;
526}
527
528static inline int ocfs2_supports_xattr(struct ocfs2_super *osb)
529{
530 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR)
531 return 1;
532 return 0;
533}
534
535static inline int ocfs2_meta_ecc(struct ocfs2_super *osb)
536{
537 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_META_ECC)
538 return 1;
539 return 0;
540}
541
542static inline int ocfs2_supports_indexed_dirs(struct ocfs2_super *osb)
543{
544 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INDEXED_DIRS)
545 return 1;
546 return 0;
547}
548
549static inline int ocfs2_supports_discontig_bg(struct ocfs2_super *osb)
550{
551 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_DISCONTIG_BG)
552 return 1;
553 return 0;
554}
555
556static inline unsigned int ocfs2_link_max(struct ocfs2_super *osb)
557{
558 if (ocfs2_supports_indexed_dirs(osb))
559 return OCFS2_DX_LINK_MAX;
560 return OCFS2_LINK_MAX;
561}
562
563static inline unsigned int ocfs2_read_links_count(struct ocfs2_dinode *di)
564{
565 u32 nlink = le16_to_cpu(di->i_links_count);
566 u32 hi = le16_to_cpu(di->i_links_count_hi);
567
568 if (di->i_dyn_features & cpu_to_le16(OCFS2_INDEXED_DIR_FL))
569 nlink |= (hi << OCFS2_LINKS_HI_SHIFT);
570
571 return nlink;
572}
573
574static inline void ocfs2_set_links_count(struct ocfs2_dinode *di, u32 nlink)
575{
576 u16 lo, hi;
577
578 lo = nlink;
579 hi = nlink >> OCFS2_LINKS_HI_SHIFT;
580
581 di->i_links_count = cpu_to_le16(lo);
582 di->i_links_count_hi = cpu_to_le16(hi);
583}
584
585static inline void ocfs2_add_links_count(struct ocfs2_dinode *di, int n)
586{
587 u32 links = ocfs2_read_links_count(di);
588
589 links += n;
590
591 ocfs2_set_links_count(di, links);
592}
593
594static inline int ocfs2_refcount_tree(struct ocfs2_super *osb)
595{
596 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_REFCOUNT_TREE)
597 return 1;
598 return 0;
599}
600
601/* set / clear functions because cluster events can make these happen
602 * in parallel so we want the transitions to be atomic. this also
603 * means that any future flags osb_flags must be protected by spinlock
604 * too! */
605static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
606 unsigned long flag)
607{
608 spin_lock(&osb->osb_lock);
609 osb->osb_flags |= flag;
610 spin_unlock(&osb->osb_lock);
611}
612
613static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
614 int hard)
615{
616 spin_lock(&osb->osb_lock);
617 osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
618 if (hard)
619 osb->osb_flags |= OCFS2_OSB_HARD_RO;
620 else
621 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
622 spin_unlock(&osb->osb_lock);
623}
624
625static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
626{
627 int ret;
628
629 spin_lock(&osb->osb_lock);
630 ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
631 spin_unlock(&osb->osb_lock);
632
633 return ret;
634}
635
636static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
637{
638 int ret;
639
640 spin_lock(&osb->osb_lock);
641 ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
642 spin_unlock(&osb->osb_lock);
643
644 return ret;
645}
646
647static inline int ocfs2_clusterinfo_valid(struct ocfs2_super *osb)
648{
649 return (osb->s_feature_incompat &
650 (OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK |
651 OCFS2_FEATURE_INCOMPAT_CLUSTERINFO));
652}
653
654static inline int ocfs2_userspace_stack(struct ocfs2_super *osb)
655{
656 if (ocfs2_clusterinfo_valid(osb) &&
657 memcmp(osb->osb_cluster_stack, OCFS2_CLASSIC_CLUSTER_STACK,
658 OCFS2_STACK_LABEL_LEN))
659 return 1;
660 return 0;
661}
662
663static inline int ocfs2_o2cb_stack(struct ocfs2_super *osb)
664{
665 if (ocfs2_clusterinfo_valid(osb) &&
666 !memcmp(osb->osb_cluster_stack, OCFS2_CLASSIC_CLUSTER_STACK,
667 OCFS2_STACK_LABEL_LEN))
668 return 1;
669 return 0;
670}
671
672static inline int ocfs2_cluster_o2cb_global_heartbeat(struct ocfs2_super *osb)
673{
674 return ocfs2_o2cb_stack(osb) &&
675 (osb->osb_stackflags & OCFS2_CLUSTER_O2CB_GLOBAL_HEARTBEAT);
676}
677
678static inline int ocfs2_mount_local(struct ocfs2_super *osb)
679{
680 return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT);
681}
682
683static inline int ocfs2_uses_extended_slot_map(struct ocfs2_super *osb)
684{
685 return (osb->s_feature_incompat &
686 OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP);
687}
688
689
690#define OCFS2_IS_VALID_DINODE(ptr) \
691 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
692
693#define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \
694 (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
695
696#define OCFS2_IS_VALID_GROUP_DESC(ptr) \
697 (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
698
699
700#define OCFS2_IS_VALID_XATTR_BLOCK(ptr) \
701 (!strcmp((ptr)->xb_signature, OCFS2_XATTR_BLOCK_SIGNATURE))
702
703#define OCFS2_IS_VALID_DIR_TRAILER(ptr) \
704 (!strcmp((ptr)->db_signature, OCFS2_DIR_TRAILER_SIGNATURE))
705
706#define OCFS2_IS_VALID_DX_ROOT(ptr) \
707 (!strcmp((ptr)->dr_signature, OCFS2_DX_ROOT_SIGNATURE))
708
709#define OCFS2_IS_VALID_DX_LEAF(ptr) \
710 (!strcmp((ptr)->dl_signature, OCFS2_DX_LEAF_SIGNATURE))
711
712#define OCFS2_IS_VALID_REFCOUNT_BLOCK(ptr) \
713 (!strcmp((ptr)->rf_signature, OCFS2_REFCOUNT_BLOCK_SIGNATURE))
714
715static inline unsigned long ino_from_blkno(struct super_block *sb,
716 u64 blkno)
717{
718 return (unsigned long)(blkno & (u64)ULONG_MAX);
719}
720
721static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
722 u32 clusters)
723{
724 int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
725 sb->s_blocksize_bits;
726
727 return (u64)clusters << c_to_b_bits;
728}
729
730static inline u32 ocfs2_clusters_for_blocks(struct super_block *sb,
731 u64 blocks)
732{
733 int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
734 sb->s_blocksize_bits;
735
736 blocks += (1 << b_to_c_bits) - 1;
737 return (u32)(blocks >> b_to_c_bits);
738}
739
740static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
741 u64 blocks)
742{
743 int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
744 sb->s_blocksize_bits;
745
746 return (u32)(blocks >> b_to_c_bits);
747}
748
749static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
750 u64 bytes)
751{
752 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
753 unsigned int clusters;
754
755 bytes += OCFS2_SB(sb)->s_clustersize - 1;
756 /* OCFS2 just cannot have enough clusters to overflow this */
757 clusters = (unsigned int)(bytes >> cl_bits);
758
759 return clusters;
760}
761
762static inline unsigned int ocfs2_bytes_to_clusters(struct super_block *sb,
763 u64 bytes)
764{
765 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
766 unsigned int clusters;
767
768 clusters = (unsigned int)(bytes >> cl_bits);
769 return clusters;
770}
771
772static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
773 u64 bytes)
774{
775 bytes += sb->s_blocksize - 1;
776 return bytes >> sb->s_blocksize_bits;
777}
778
779static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
780 u32 clusters)
781{
782 return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
783}
784
785static inline u64 ocfs2_block_to_cluster_start(struct super_block *sb,
786 u64 blocks)
787{
788 int bits = OCFS2_SB(sb)->s_clustersize_bits - sb->s_blocksize_bits;
789 unsigned int clusters;
790
791 clusters = ocfs2_blocks_to_clusters(sb, blocks);
792 return (u64)clusters << bits;
793}
794
795static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
796 u64 bytes)
797{
798 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
799 unsigned int clusters;
800
801 clusters = ocfs2_clusters_for_bytes(sb, bytes);
802 return (u64)clusters << cl_bits;
803}
804
805static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
806 u64 bytes)
807{
808 u64 blocks;
809
810 blocks = ocfs2_blocks_for_bytes(sb, bytes);
811 return blocks << sb->s_blocksize_bits;
812}
813
814static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
815{
816 return (unsigned long)((bytes + 511) >> 9);
817}
818
819static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb,
820 unsigned long pg_index)
821{
822 u32 clusters = pg_index;
823 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
824
825 if (unlikely(PAGE_SHIFT > cbits))
826 clusters = pg_index << (PAGE_SHIFT - cbits);
827 else if (PAGE_SHIFT < cbits)
828 clusters = pg_index >> (cbits - PAGE_SHIFT);
829
830 return clusters;
831}
832
833/*
834 * Find the 1st page index which covers the given clusters.
835 */
836static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb,
837 u32 clusters)
838{
839 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
840 pgoff_t index = clusters;
841
842 if (PAGE_SHIFT > cbits) {
843 index = (pgoff_t)clusters >> (PAGE_SHIFT - cbits);
844 } else if (PAGE_SHIFT < cbits) {
845 index = (pgoff_t)clusters << (cbits - PAGE_SHIFT);
846 }
847
848 return index;
849}
850
851static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb)
852{
853 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
854 unsigned int pages_per_cluster = 1;
855
856 if (PAGE_SHIFT < cbits)
857 pages_per_cluster = 1 << (cbits - PAGE_SHIFT);
858
859 return pages_per_cluster;
860}
861
862static inline unsigned int ocfs2_megabytes_to_clusters(struct super_block *sb,
863 unsigned int megs)
864{
865 BUILD_BUG_ON(OCFS2_MAX_CLUSTERSIZE > 1048576);
866
867 return megs << (20 - OCFS2_SB(sb)->s_clustersize_bits);
868}
869
870static inline unsigned int ocfs2_clusters_to_megabytes(struct super_block *sb,
871 unsigned int clusters)
872{
873 return clusters >> (20 - OCFS2_SB(sb)->s_clustersize_bits);
874}
875
876static inline void _ocfs2_set_bit(unsigned int bit, unsigned long *bitmap)
877{
878 __set_bit_le(bit, bitmap);
879}
880#define ocfs2_set_bit(bit, addr) _ocfs2_set_bit((bit), (unsigned long *)(addr))
881
882static inline void _ocfs2_clear_bit(unsigned int bit, unsigned long *bitmap)
883{
884 __clear_bit_le(bit, bitmap);
885}
886#define ocfs2_clear_bit(bit, addr) _ocfs2_clear_bit((bit), (unsigned long *)(addr))
887
888#define ocfs2_test_bit test_bit_le
889#define ocfs2_find_next_zero_bit find_next_zero_bit_le
890#define ocfs2_find_next_bit find_next_bit_le
891
892static inline void *correct_addr_and_bit_unaligned(int *bit, void *addr)
893{
894#if BITS_PER_LONG == 64
895 *bit += ((unsigned long) addr & 7UL) << 3;
896 addr = (void *) ((unsigned long) addr & ~7UL);
897#elif BITS_PER_LONG == 32
898 *bit += ((unsigned long) addr & 3UL) << 3;
899 addr = (void *) ((unsigned long) addr & ~3UL);
900#else
901#error "how many bits you are?!"
902#endif
903 return addr;
904}
905
906static inline void ocfs2_set_bit_unaligned(int bit, void *bitmap)
907{
908 bitmap = correct_addr_and_bit_unaligned(&bit, bitmap);
909 ocfs2_set_bit(bit, bitmap);
910}
911
912static inline void ocfs2_clear_bit_unaligned(int bit, void *bitmap)
913{
914 bitmap = correct_addr_and_bit_unaligned(&bit, bitmap);
915 ocfs2_clear_bit(bit, bitmap);
916}
917
918static inline int ocfs2_test_bit_unaligned(int bit, void *bitmap)
919{
920 bitmap = correct_addr_and_bit_unaligned(&bit, bitmap);
921 return ocfs2_test_bit(bit, bitmap);
922}
923
924static inline int ocfs2_find_next_zero_bit_unaligned(void *bitmap, int max,
925 int start)
926{
927 int fix = 0, ret, tmpmax;
928 bitmap = correct_addr_and_bit_unaligned(&fix, bitmap);
929 tmpmax = max + fix;
930 start += fix;
931
932 ret = ocfs2_find_next_zero_bit(bitmap, tmpmax, start) - fix;
933 if (ret > max)
934 return max;
935 return ret;
936}
937
938#endif /* OCFS2_H */
939