xref: /openbmc/linux/fs/ocfs2/ocfs2.h (revision f42b3800)
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 /* For union ocfs2_dlm_lksb */
40 #include "stackglue.h"
41 
42 #include "ocfs2_fs.h"
43 #include "ocfs2_lockid.h"
44 
45 /* Most user visible OCFS2 inodes will have very few pieces of
46  * metadata, but larger files (including bitmaps, etc) must be taken
47  * into account when designing an access scheme. We allow a small
48  * amount of inlined blocks to be stored on an array and grow the
49  * structure into a rb tree when necessary. */
50 #define OCFS2_INODE_MAX_CACHE_ARRAY 2
51 
52 struct ocfs2_caching_info {
53 	unsigned int		ci_num_cached;
54 	union {
55 		sector_t	ci_array[OCFS2_INODE_MAX_CACHE_ARRAY];
56 		struct rb_root	ci_tree;
57 	} ci_cache;
58 };
59 
60 /* this limits us to 256 nodes
61  * if we need more, we can do a kmalloc for the map */
62 #define OCFS2_NODE_MAP_MAX_NODES    256
63 struct ocfs2_node_map {
64 	u16 num_nodes;
65 	unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)];
66 };
67 
68 enum ocfs2_ast_action {
69 	OCFS2_AST_INVALID = 0,
70 	OCFS2_AST_ATTACH,
71 	OCFS2_AST_CONVERT,
72 	OCFS2_AST_DOWNCONVERT,
73 };
74 
75 /* actions for an unlockast function to take. */
76 enum ocfs2_unlock_action {
77 	OCFS2_UNLOCK_INVALID = 0,
78 	OCFS2_UNLOCK_CANCEL_CONVERT,
79 	OCFS2_UNLOCK_DROP_LOCK,
80 };
81 
82 /* ocfs2_lock_res->l_flags flags. */
83 #define OCFS2_LOCK_ATTACHED      (0x00000001) /* have we initialized
84 					       * the lvb */
85 #define OCFS2_LOCK_BUSY          (0x00000002) /* we are currently in
86 					       * dlm_lock */
87 #define OCFS2_LOCK_BLOCKED       (0x00000004) /* blocked waiting to
88 					       * downconvert*/
89 #define OCFS2_LOCK_LOCAL         (0x00000008) /* newly created inode */
90 #define OCFS2_LOCK_NEEDS_REFRESH (0x00000010)
91 #define OCFS2_LOCK_REFRESHING    (0x00000020)
92 #define OCFS2_LOCK_INITIALIZED   (0x00000040) /* track initialization
93 					       * for shutdown paths */
94 #define OCFS2_LOCK_FREEING       (0x00000080) /* help dlmglue track
95 					       * when to skip queueing
96 					       * a lock because it's
97 					       * about to be
98 					       * dropped. */
99 #define OCFS2_LOCK_QUEUED        (0x00000100) /* queued for downconvert */
100 #define OCFS2_LOCK_NOCACHE       (0x00000200) /* don't use a holder count */
101 #define OCFS2_LOCK_PENDING       (0x00000400) /* This lockres is pending a
102 						 call to dlm_lock.  Only
103 						 exists with BUSY set. */
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 	union ocfs2_dlm_lksb     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 	unsigned int             l_pending_gen;
131 
132 	wait_queue_head_t        l_event;
133 
134 	struct list_head         l_debug_list;
135 };
136 
137 struct ocfs2_dlm_debug {
138 	struct kref d_refcnt;
139 	struct dentry *d_locking_state;
140 	struct list_head d_lockres_tracking;
141 };
142 
143 enum ocfs2_vol_state
144 {
145 	VOLUME_INIT = 0,
146 	VOLUME_MOUNTED,
147 	VOLUME_DISMOUNTED,
148 	VOLUME_DISABLED
149 };
150 
151 struct ocfs2_alloc_stats
152 {
153 	atomic_t moves;
154 	atomic_t local_data;
155 	atomic_t bitmap_data;
156 	atomic_t bg_allocs;
157 	atomic_t bg_extends;
158 };
159 
160 enum ocfs2_local_alloc_state
161 {
162 	OCFS2_LA_UNUSED = 0,
163 	OCFS2_LA_ENABLED,
164 	OCFS2_LA_DISABLED
165 };
166 
167 enum ocfs2_mount_options
168 {
169 	OCFS2_MOUNT_HB_LOCAL   = 1 << 0, /* Heartbeat started in local mode */
170 	OCFS2_MOUNT_BARRIER = 1 << 1,	/* Use block barriers */
171 	OCFS2_MOUNT_NOINTR  = 1 << 2,   /* Don't catch signals */
172 	OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */
173 	OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */
174 	OCFS2_MOUNT_LOCALFLOCKS = 1 << 5, /* No cluster aware user file locks */
175 };
176 
177 #define OCFS2_OSB_SOFT_RO	0x0001
178 #define OCFS2_OSB_HARD_RO	0x0002
179 #define OCFS2_OSB_ERROR_FS	0x0004
180 #define OCFS2_DEFAULT_ATIME_QUANTUM	60
181 
182 struct ocfs2_journal;
183 struct ocfs2_slot_info;
184 struct ocfs2_recovery_map;
185 struct ocfs2_super
186 {
187 	struct task_struct *commit_task;
188 	struct super_block *sb;
189 	struct inode *root_inode;
190 	struct inode *sys_root_inode;
191 	struct inode *system_inodes[NUM_SYSTEM_INODES];
192 
193 	struct ocfs2_slot_info *slot_info;
194 
195 	spinlock_t node_map_lock;
196 
197 	u64 root_blkno;
198 	u64 system_dir_blkno;
199 	u64 bitmap_blkno;
200 	u32 bitmap_cpg;
201 	u8 *uuid;
202 	char *uuid_str;
203 	u8 *vol_label;
204 	u64 first_cluster_group_blkno;
205 	u32 fs_generation;
206 
207 	u32 s_feature_compat;
208 	u32 s_feature_incompat;
209 	u32 s_feature_ro_compat;
210 
211 	/* Protects s_next_generation, osb_flags and s_inode_steal_slot.
212 	 * Could protect more on osb as it's very short lived.
213 	 */
214 	spinlock_t osb_lock;
215 	u32 s_next_generation;
216 	unsigned long osb_flags;
217 	s16 s_inode_steal_slot;
218 	atomic_t s_num_inodes_stolen;
219 
220 	unsigned long s_mount_opt;
221 	unsigned int s_atime_quantum;
222 
223 	unsigned int max_slots;
224 	unsigned int node_num;
225 	int slot_num;
226 	int preferred_slot;
227 	int s_sectsize_bits;
228 	int s_clustersize;
229 	int s_clustersize_bits;
230 
231 	atomic_t vol_state;
232 	struct mutex recovery_lock;
233 	struct ocfs2_recovery_map *recovery_map;
234 	struct task_struct *recovery_thread_task;
235 	int disable_recovery;
236 	wait_queue_head_t checkpoint_event;
237 	atomic_t needs_checkpoint;
238 	struct ocfs2_journal *journal;
239 	unsigned long osb_commit_interval;
240 
241 	int local_alloc_size;
242 	enum ocfs2_local_alloc_state local_alloc_state;
243 	struct buffer_head *local_alloc_bh;
244 	u64 la_last_gd;
245 
246 	/* Next two fields are for local node slot recovery during
247 	 * mount. */
248 	int dirty;
249 	struct ocfs2_dinode *local_alloc_copy;
250 
251 	struct ocfs2_alloc_stats alloc_stats;
252 	char dev_str[20];		/* "major,minor" of the device */
253 
254 	char osb_cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
255 	struct ocfs2_cluster_connection *cconn;
256 	struct ocfs2_lock_res osb_super_lockres;
257 	struct ocfs2_lock_res osb_rename_lockres;
258 	struct ocfs2_dlm_debug *osb_dlm_debug;
259 
260 	struct dentry *osb_debug_root;
261 
262 	wait_queue_head_t recovery_event;
263 
264 	spinlock_t dc_task_lock;
265 	struct task_struct *dc_task;
266 	wait_queue_head_t dc_event;
267 	unsigned long dc_wake_sequence;
268 	unsigned long dc_work_sequence;
269 
270 	/*
271 	 * Any thread can add locks to the list, but the downconvert
272 	 * thread is the only one allowed to remove locks. Any change
273 	 * to this rule requires updating
274 	 * ocfs2_downconvert_thread_do_work().
275 	 */
276 	struct list_head blocked_lock_list;
277 	unsigned long blocked_lock_count;
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 static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb)
310 {
311 	/*
312 	 * Support for sparse files is a pre-requisite
313 	 */
314 	if (!ocfs2_sparse_alloc(osb))
315 		return 0;
316 
317 	if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN)
318 		return 1;
319 	return 0;
320 }
321 
322 static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb)
323 {
324 	if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA)
325 		return 1;
326 	return 0;
327 }
328 
329 /* set / clear functions because cluster events can make these happen
330  * in parallel so we want the transitions to be atomic. this also
331  * means that any future flags osb_flags must be protected by spinlock
332  * too! */
333 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
334 				      unsigned long flag)
335 {
336 	spin_lock(&osb->osb_lock);
337 	osb->osb_flags |= flag;
338 	spin_unlock(&osb->osb_lock);
339 }
340 
341 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
342 				     int hard)
343 {
344 	spin_lock(&osb->osb_lock);
345 	osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
346 	if (hard)
347 		osb->osb_flags |= OCFS2_OSB_HARD_RO;
348 	else
349 		osb->osb_flags |= OCFS2_OSB_SOFT_RO;
350 	spin_unlock(&osb->osb_lock);
351 }
352 
353 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
354 {
355 	int ret;
356 
357 	spin_lock(&osb->osb_lock);
358 	ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
359 	spin_unlock(&osb->osb_lock);
360 
361 	return ret;
362 }
363 
364 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
365 {
366 	int ret;
367 
368 	spin_lock(&osb->osb_lock);
369 	ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
370 	spin_unlock(&osb->osb_lock);
371 
372 	return ret;
373 }
374 
375 static inline int ocfs2_userspace_stack(struct ocfs2_super *osb)
376 {
377 	return (osb->s_feature_incompat &
378 		OCFS2_FEATURE_INCOMPAT_USERSPACE_STACK);
379 }
380 
381 static inline int ocfs2_mount_local(struct ocfs2_super *osb)
382 {
383 	return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT);
384 }
385 
386 static inline int ocfs2_uses_extended_slot_map(struct ocfs2_super *osb)
387 {
388 	return (osb->s_feature_incompat &
389 		OCFS2_FEATURE_INCOMPAT_EXTENDED_SLOT_MAP);
390 }
391 
392 
393 #define OCFS2_IS_VALID_DINODE(ptr)					\
394 	(!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
395 
396 #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di)	do {			\
397 	typeof(__di) ____di = (__di);					\
398 	ocfs2_error((__sb), 						\
399 		"Dinode # %llu has bad signature %.*s",			\
400 		(unsigned long long)le64_to_cpu((____di)->i_blkno), 7, 	\
401 		(____di)->i_signature);					\
402 } while (0)
403 
404 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr)				\
405 	(!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
406 
407 #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb)	do {		\
408 	typeof(__eb) ____eb = (__eb);					\
409 	ocfs2_error((__sb), 						\
410 		"Extent Block # %llu has bad signature %.*s",		\
411 		(unsigned long long)le64_to_cpu((____eb)->h_blkno), 7,	\
412 		(____eb)->h_signature);					\
413 } while (0)
414 
415 #define OCFS2_IS_VALID_GROUP_DESC(ptr)					\
416 	(!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
417 
418 #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd)	do {		\
419 	typeof(__gd) ____gd = (__gd);					\
420 		ocfs2_error((__sb),					\
421 		"Group Descriptor # %llu has bad signature %.*s",	\
422 		(unsigned long long)le64_to_cpu((____gd)->bg_blkno), 7, \
423 		(____gd)->bg_signature);				\
424 } while (0)
425 
426 static inline unsigned long ino_from_blkno(struct super_block *sb,
427 					   u64 blkno)
428 {
429 	return (unsigned long)(blkno & (u64)ULONG_MAX);
430 }
431 
432 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
433 					   u32 clusters)
434 {
435 	int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
436 		sb->s_blocksize_bits;
437 
438 	return (u64)clusters << c_to_b_bits;
439 }
440 
441 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
442 					   u64 blocks)
443 {
444 	int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
445 		sb->s_blocksize_bits;
446 
447 	return (u32)(blocks >> b_to_c_bits);
448 }
449 
450 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
451 						    u64 bytes)
452 {
453 	int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
454 	unsigned int clusters;
455 
456 	bytes += OCFS2_SB(sb)->s_clustersize - 1;
457 	/* OCFS2 just cannot have enough clusters to overflow this */
458 	clusters = (unsigned int)(bytes >> cl_bits);
459 
460 	return clusters;
461 }
462 
463 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
464 					 u64 bytes)
465 {
466 	bytes += sb->s_blocksize - 1;
467 	return bytes >> sb->s_blocksize_bits;
468 }
469 
470 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
471 					  u32 clusters)
472 {
473 	return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
474 }
475 
476 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
477 						u64 bytes)
478 {
479 	int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
480 	unsigned int clusters;
481 
482 	clusters = ocfs2_clusters_for_bytes(sb, bytes);
483 	return (u64)clusters << cl_bits;
484 }
485 
486 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
487 					      u64 bytes)
488 {
489 	u64 blocks;
490 
491         blocks = ocfs2_blocks_for_bytes(sb, bytes);
492 	return blocks << sb->s_blocksize_bits;
493 }
494 
495 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
496 {
497 	return (unsigned long)((bytes + 511) >> 9);
498 }
499 
500 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb,
501 							unsigned long pg_index)
502 {
503 	u32 clusters = pg_index;
504 	unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
505 
506 	if (unlikely(PAGE_CACHE_SHIFT > cbits))
507 		clusters = pg_index << (PAGE_CACHE_SHIFT - cbits);
508 	else if (PAGE_CACHE_SHIFT < cbits)
509 		clusters = pg_index >> (cbits - PAGE_CACHE_SHIFT);
510 
511 	return clusters;
512 }
513 
514 /*
515  * Find the 1st page index which covers the given clusters.
516  */
517 static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb,
518 							u32 clusters)
519 {
520 	unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
521         pgoff_t index = clusters;
522 
523 	if (PAGE_CACHE_SHIFT > cbits) {
524 		index = (pgoff_t)clusters >> (PAGE_CACHE_SHIFT - cbits);
525 	} else if (PAGE_CACHE_SHIFT < cbits) {
526 		index = (pgoff_t)clusters << (cbits - PAGE_CACHE_SHIFT);
527 	}
528 
529 	return index;
530 }
531 
532 static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb)
533 {
534 	unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
535 	unsigned int pages_per_cluster = 1;
536 
537 	if (PAGE_CACHE_SHIFT < cbits)
538 		pages_per_cluster = 1 << (cbits - PAGE_CACHE_SHIFT);
539 
540 	return pages_per_cluster;
541 }
542 
543 static inline void ocfs2_init_inode_steal_slot(struct ocfs2_super *osb)
544 {
545 	spin_lock(&osb->osb_lock);
546 	osb->s_inode_steal_slot = OCFS2_INVALID_SLOT;
547 	spin_unlock(&osb->osb_lock);
548 	atomic_set(&osb->s_num_inodes_stolen, 0);
549 }
550 
551 static inline void ocfs2_set_inode_steal_slot(struct ocfs2_super *osb,
552 					      s16 slot)
553 {
554 	spin_lock(&osb->osb_lock);
555 	osb->s_inode_steal_slot = slot;
556 	spin_unlock(&osb->osb_lock);
557 }
558 
559 static inline s16 ocfs2_get_inode_steal_slot(struct ocfs2_super *osb)
560 {
561 	s16 slot;
562 
563 	spin_lock(&osb->osb_lock);
564 	slot = osb->s_inode_steal_slot;
565 	spin_unlock(&osb->osb_lock);
566 
567 	return slot;
568 }
569 
570 #define ocfs2_set_bit ext2_set_bit
571 #define ocfs2_clear_bit ext2_clear_bit
572 #define ocfs2_test_bit ext2_test_bit
573 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit
574 #endif  /* OCFS2_H */
575 
576