1 /* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
6 * Defines macros and structures used in OCFS2
8 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
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.
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.
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.
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>
39 #include "cluster/nodemanager.h"
40 #include "cluster/heartbeat.h"
41 #include "cluster/tcp.h"
43 #include "dlm/dlmapi.h"
46 #include "ocfs2_lockid.h"
48 /* Most user visible OCFS2 inodes will have very few pieces of
49 * metadata, but larger files (including bitmaps, etc) must be taken
50 * into account when designing an access scheme. We allow a small
51 * amount of inlined blocks to be stored on an array and grow the
52 * structure into a rb tree when necessary. */
53 #define OCFS2_INODE_MAX_CACHE_ARRAY 2
55 struct ocfs2_caching_info {
56 unsigned int ci_num_cached;
58 sector_t ci_array[OCFS2_INODE_MAX_CACHE_ARRAY];
59 struct rb_root ci_tree;
63 /* this limits us to 256 nodes
64 * if we need more, we can do a kmalloc for the map */
65 #define OCFS2_NODE_MAP_MAX_NODES 256
66 struct ocfs2_node_map {
68 unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)];
71 enum ocfs2_ast_action {
72 OCFS2_AST_INVALID = 0,
75 OCFS2_AST_DOWNCONVERT,
78 /* actions for an unlockast function to take. */
79 enum ocfs2_unlock_action {
80 OCFS2_UNLOCK_INVALID = 0,
81 OCFS2_UNLOCK_CANCEL_CONVERT,
82 OCFS2_UNLOCK_DROP_LOCK,
85 /* ocfs2_lock_res->l_flags flags. */
86 #define OCFS2_LOCK_ATTACHED (0x00000001) /* have we initialized
88 #define OCFS2_LOCK_BUSY (0x00000002) /* we are currently in
90 #define OCFS2_LOCK_BLOCKED (0x00000004) /* blocked waiting to
92 #define OCFS2_LOCK_LOCAL (0x00000008) /* newly created inode */
93 #define OCFS2_LOCK_NEEDS_REFRESH (0x00000010)
94 #define OCFS2_LOCK_REFRESHING (0x00000020)
95 #define OCFS2_LOCK_INITIALIZED (0x00000040) /* track initialization
96 * for shutdown paths */
97 #define OCFS2_LOCK_FREEING (0x00000080) /* help dlmglue track
98 * when to skip queueing
102 #define OCFS2_LOCK_QUEUED (0x00000100) /* queued for downconvert */
103 #define OCFS2_LOCK_NOCACHE (0x00000200) /* don't use a holder count */
105 struct ocfs2_lock_res_ops;
107 typedef void (*ocfs2_lock_callback)(int status, unsigned long data);
109 struct ocfs2_lock_res {
111 struct ocfs2_lock_res_ops *l_ops;
114 struct list_head l_blocked_list;
115 struct list_head l_mask_waiters;
117 enum ocfs2_lock_type l_type;
118 unsigned long l_flags;
119 char l_name[OCFS2_LOCK_ID_MAX_LEN];
121 unsigned int l_ro_holders;
122 unsigned int l_ex_holders;
123 struct dlm_lockstatus l_lksb;
125 /* used from AST/BAST funcs. */
126 enum ocfs2_ast_action l_action;
127 enum ocfs2_unlock_action l_unlock_action;
131 wait_queue_head_t l_event;
133 struct list_head l_debug_list;
136 struct ocfs2_dlm_debug {
137 struct kref d_refcnt;
138 struct dentry *d_locking_state;
139 struct list_head d_lockres_tracking;
150 struct ocfs2_alloc_stats
154 atomic_t bitmap_data;
159 enum ocfs2_local_alloc_state
166 enum ocfs2_mount_options
168 OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Heartbeat started in local mode */
169 OCFS2_MOUNT_BARRIER = 1 << 1, /* Use block barriers */
170 OCFS2_MOUNT_NOINTR = 1 << 2, /* Don't catch signals */
171 OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */
172 OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */
173 OCFS2_MOUNT_LOCALFLOCKS = 1 << 5, /* No cluster aware user file locks */
176 #define OCFS2_OSB_SOFT_RO 0x0001
177 #define OCFS2_OSB_HARD_RO 0x0002
178 #define OCFS2_OSB_ERROR_FS 0x0004
179 #define OCFS2_DEFAULT_ATIME_QUANTUM 60
181 struct ocfs2_journal;
182 struct ocfs2_slot_info;
183 struct ocfs2_recovery_map;
186 struct task_struct *commit_task;
187 struct super_block *sb;
188 struct inode *root_inode;
189 struct inode *sys_root_inode;
190 struct inode *system_inodes[NUM_SYSTEM_INODES];
192 struct ocfs2_slot_info *slot_info;
194 spinlock_t node_map_lock;
197 u64 system_dir_blkno;
203 u64 first_cluster_group_blkno;
206 u32 s_feature_compat;
207 u32 s_feature_incompat;
208 u32 s_feature_ro_compat;
210 /* Protects s_next_generaion, osb_flags. Could protect more on
211 * osb as it's very short lived. */
213 u32 s_next_generation;
214 unsigned long osb_flags;
216 unsigned long s_mount_opt;
217 unsigned int s_atime_quantum;
219 unsigned int max_slots;
225 int s_clustersize_bits;
228 struct mutex recovery_lock;
229 struct ocfs2_recovery_map *recovery_map;
230 struct task_struct *recovery_thread_task;
231 int disable_recovery;
232 wait_queue_head_t checkpoint_event;
233 atomic_t needs_checkpoint;
234 struct ocfs2_journal *journal;
235 unsigned long osb_commit_interval;
237 int local_alloc_size;
238 enum ocfs2_local_alloc_state local_alloc_state;
239 struct buffer_head *local_alloc_bh;
242 /* Next two fields are for local node slot recovery during
245 struct ocfs2_dinode *local_alloc_copy;
247 struct ocfs2_alloc_stats alloc_stats;
248 char dev_str[20]; /* "major,minor" of the device */
250 struct dlm_ctxt *dlm;
251 struct ocfs2_lock_res osb_super_lockres;
252 struct ocfs2_lock_res osb_rename_lockres;
253 struct dlm_eviction_cb osb_eviction_cb;
254 struct ocfs2_dlm_debug *osb_dlm_debug;
255 struct dlm_protocol_version osb_locking_proto;
257 struct dentry *osb_debug_root;
259 wait_queue_head_t recovery_event;
261 spinlock_t dc_task_lock;
262 struct task_struct *dc_task;
263 wait_queue_head_t dc_event;
264 unsigned long dc_wake_sequence;
265 unsigned long dc_work_sequence;
268 * Any thread can add locks to the list, but the downconvert
269 * thread is the only one allowed to remove locks. Any change
270 * to this rule requires updating
271 * ocfs2_downconvert_thread_do_work().
273 struct list_head blocked_lock_list;
274 unsigned long blocked_lock_count;
276 wait_queue_head_t osb_mount_event;
278 /* Truncate log info */
279 struct inode *osb_tl_inode;
280 struct buffer_head *osb_tl_bh;
281 struct delayed_work osb_truncate_log_wq;
283 struct ocfs2_node_map osb_recovering_orphan_dirs;
284 unsigned int *osb_orphan_wipes;
285 wait_queue_head_t osb_wipe_event;
288 #define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info)
290 static inline int ocfs2_should_order_data(struct inode *inode)
292 if (!S_ISREG(inode->i_mode))
294 if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
299 static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb)
301 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC)
306 static inline int ocfs2_writes_unwritten_extents(struct ocfs2_super *osb)
309 * Support for sparse files is a pre-requisite
311 if (!ocfs2_sparse_alloc(osb))
314 if (osb->s_feature_ro_compat & OCFS2_FEATURE_RO_COMPAT_UNWRITTEN)
319 static inline int ocfs2_supports_inline_data(struct ocfs2_super *osb)
321 if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_INLINE_DATA)
326 /* set / clear functions because cluster events can make these happen
327 * in parallel so we want the transitions to be atomic. this also
328 * means that any future flags osb_flags must be protected by spinlock
330 static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
333 spin_lock(&osb->osb_lock);
334 osb->osb_flags |= flag;
335 spin_unlock(&osb->osb_lock);
338 static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
341 spin_lock(&osb->osb_lock);
342 osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
344 osb->osb_flags |= OCFS2_OSB_HARD_RO;
346 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
347 spin_unlock(&osb->osb_lock);
350 static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
354 spin_lock(&osb->osb_lock);
355 ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
356 spin_unlock(&osb->osb_lock);
361 static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
365 spin_lock(&osb->osb_lock);
366 ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
367 spin_unlock(&osb->osb_lock);
372 static inline int ocfs2_mount_local(struct ocfs2_super *osb)
374 return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT);
377 #define OCFS2_IS_VALID_DINODE(ptr) \
378 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
380 #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di) do { \
381 typeof(__di) ____di = (__di); \
382 ocfs2_error((__sb), \
383 "Dinode # %llu has bad signature %.*s", \
384 (unsigned long long)le64_to_cpu((____di)->i_blkno), 7, \
385 (____di)->i_signature); \
388 #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \
389 (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
391 #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb) do { \
392 typeof(__eb) ____eb = (__eb); \
393 ocfs2_error((__sb), \
394 "Extent Block # %llu has bad signature %.*s", \
395 (unsigned long long)le64_to_cpu((____eb)->h_blkno), 7, \
396 (____eb)->h_signature); \
399 #define OCFS2_IS_VALID_GROUP_DESC(ptr) \
400 (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
402 #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd) do { \
403 typeof(__gd) ____gd = (__gd); \
404 ocfs2_error((__sb), \
405 "Group Descriptor # %llu has bad signature %.*s", \
406 (unsigned long long)le64_to_cpu((____gd)->bg_blkno), 7, \
407 (____gd)->bg_signature); \
410 static inline unsigned long ino_from_blkno(struct super_block *sb,
413 return (unsigned long)(blkno & (u64)ULONG_MAX);
416 static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
419 int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
420 sb->s_blocksize_bits;
422 return (u64)clusters << c_to_b_bits;
425 static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
428 int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
429 sb->s_blocksize_bits;
431 return (u32)(blocks >> b_to_c_bits);
434 static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
437 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
438 unsigned int clusters;
440 bytes += OCFS2_SB(sb)->s_clustersize - 1;
441 /* OCFS2 just cannot have enough clusters to overflow this */
442 clusters = (unsigned int)(bytes >> cl_bits);
447 static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
450 bytes += sb->s_blocksize - 1;
451 return bytes >> sb->s_blocksize_bits;
454 static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
457 return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
460 static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
463 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
464 unsigned int clusters;
466 clusters = ocfs2_clusters_for_bytes(sb, bytes);
467 return (u64)clusters << cl_bits;
470 static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
475 blocks = ocfs2_blocks_for_bytes(sb, bytes);
476 return blocks << sb->s_blocksize_bits;
479 static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
481 return (unsigned long)((bytes + 511) >> 9);
484 static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb,
485 unsigned long pg_index)
487 u32 clusters = pg_index;
488 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
490 if (unlikely(PAGE_CACHE_SHIFT > cbits))
491 clusters = pg_index << (PAGE_CACHE_SHIFT - cbits);
492 else if (PAGE_CACHE_SHIFT < cbits)
493 clusters = pg_index >> (cbits - PAGE_CACHE_SHIFT);
499 * Find the 1st page index which covers the given clusters.
501 static inline pgoff_t ocfs2_align_clusters_to_page_index(struct super_block *sb,
504 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
505 pgoff_t index = clusters;
507 if (PAGE_CACHE_SHIFT > cbits) {
508 index = (pgoff_t)clusters >> (PAGE_CACHE_SHIFT - cbits);
509 } else if (PAGE_CACHE_SHIFT < cbits) {
510 index = (pgoff_t)clusters << (cbits - PAGE_CACHE_SHIFT);
516 static inline unsigned int ocfs2_pages_per_cluster(struct super_block *sb)
518 unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits;
519 unsigned int pages_per_cluster = 1;
521 if (PAGE_CACHE_SHIFT < cbits)
522 pages_per_cluster = 1 << (cbits - PAGE_CACHE_SHIFT);
524 return pages_per_cluster;
527 #define ocfs2_set_bit ext2_set_bit
528 #define ocfs2_clear_bit ext2_clear_bit
529 #define ocfs2_test_bit ext2_test_bit
530 #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit