blob: 0462a7f4e21b398300f3ed1bf8b80b6b7d30ce07 [file] [log] [blame]
Mark Fashehccd979b2005-12-15 14:31:24 -08001/* -*- 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>
Arjan van de Venc74ec2f2006-01-13 21:54:23 -080036#include <linux/mutex.h>
Mark Fashehccd979b2005-12-15 14:31:24 -080037
38#include "cluster/nodemanager.h"
39#include "cluster/heartbeat.h"
40#include "cluster/tcp.h"
41
42#include "dlm/dlmapi.h"
43
44#include "ocfs2_fs.h"
45#include "endian.h"
46#include "ocfs2_lockid.h"
47
48struct ocfs2_extent_map {
49 u32 em_clusters;
50 struct rb_root em_extents;
51};
52
53/* Most user visible OCFS2 inodes will have very few pieces of
54 * metadata, but larger files (including bitmaps, etc) must be taken
55 * into account when designing an access scheme. We allow a small
56 * amount of inlined blocks to be stored on an array and grow the
57 * structure into a rb tree when necessary. */
58#define OCFS2_INODE_MAX_CACHE_ARRAY 2
59
60struct ocfs2_caching_info {
61 unsigned int ci_num_cached;
62 union {
63 sector_t ci_array[OCFS2_INODE_MAX_CACHE_ARRAY];
64 struct rb_root ci_tree;
65 } ci_cache;
66};
67
68/* this limits us to 256 nodes
69 * if we need more, we can do a kmalloc for the map */
70#define OCFS2_NODE_MAP_MAX_NODES 256
71struct ocfs2_node_map {
72 u16 num_nodes;
73 unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)];
74};
75
76enum ocfs2_ast_action {
77 OCFS2_AST_INVALID = 0,
78 OCFS2_AST_ATTACH,
79 OCFS2_AST_CONVERT,
80 OCFS2_AST_DOWNCONVERT,
81};
82
83/* actions for an unlockast function to take. */
84enum ocfs2_unlock_action {
85 OCFS2_UNLOCK_INVALID = 0,
86 OCFS2_UNLOCK_CANCEL_CONVERT,
87 OCFS2_UNLOCK_DROP_LOCK,
88};
89
90/* ocfs2_lock_res->l_flags flags. */
91#define OCFS2_LOCK_ATTACHED (0x00000001) /* have we initialized
92 * the lvb */
93#define OCFS2_LOCK_BUSY (0x00000002) /* we are currently in
94 * dlm_lock */
95#define OCFS2_LOCK_BLOCKED (0x00000004) /* blocked waiting to
96 * downconvert*/
97#define OCFS2_LOCK_LOCAL (0x00000008) /* newly created inode */
98#define OCFS2_LOCK_NEEDS_REFRESH (0x00000010)
99#define OCFS2_LOCK_REFRESHING (0x00000020)
100#define OCFS2_LOCK_INITIALIZED (0x00000040) /* track initialization
101 * for shutdown paths */
102#define OCFS2_LOCK_FREEING (0x00000080) /* help dlmglue track
103 * when to skip queueing
104 * a lock because it's
105 * about to be
106 * dropped. */
107#define OCFS2_LOCK_QUEUED (0x00000100) /* queued for downconvert */
108
109struct ocfs2_lock_res_ops;
110
111typedef void (*ocfs2_lock_callback)(int status, unsigned long data);
112
113struct ocfs2_lock_res {
114 void *l_priv;
115 struct ocfs2_lock_res_ops *l_ops;
116 spinlock_t l_lock;
117
118 struct list_head l_blocked_list;
119 struct list_head l_mask_waiters;
120
121 enum ocfs2_lock_type l_type;
122 unsigned long l_flags;
123 char l_name[OCFS2_LOCK_ID_MAX_LEN];
124 int l_level;
125 unsigned int l_ro_holders;
126 unsigned int l_ex_holders;
127 struct dlm_lockstatus l_lksb;
128
129 /* used from AST/BAST funcs. */
130 enum ocfs2_ast_action l_action;
131 enum ocfs2_unlock_action l_unlock_action;
132 int l_requested;
133 int l_blocking;
134
135 wait_queue_head_t l_event;
136
137 struct list_head l_debug_list;
138};
139
140struct ocfs2_dlm_debug {
141 struct kref d_refcnt;
142 struct dentry *d_locking_state;
143 struct list_head d_lockres_tracking;
144};
145
146enum ocfs2_vol_state
147{
148 VOLUME_INIT = 0,
149 VOLUME_MOUNTED,
150 VOLUME_DISMOUNTED,
151 VOLUME_DISABLED
152};
153
154struct ocfs2_alloc_stats
155{
156 atomic_t moves;
157 atomic_t local_data;
158 atomic_t bitmap_data;
159 atomic_t bg_allocs;
160 atomic_t bg_extends;
161};
162
163enum ocfs2_local_alloc_state
164{
165 OCFS2_LA_UNUSED = 0,
166 OCFS2_LA_ENABLED,
167 OCFS2_LA_DISABLED
168};
169
170enum ocfs2_mount_options
171{
172 OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Heartbeat started in local mode */
173 OCFS2_MOUNT_BARRIER = 1 << 1, /* Use block barriers */
174 OCFS2_MOUNT_NOINTR = 1 << 2, /* Don't catch signals */
175 OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */
176 OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */
Mark Fashehccd979b2005-12-15 14:31:24 -0800177};
178
179#define OCFS2_OSB_SOFT_RO 0x0001
180#define OCFS2_OSB_HARD_RO 0x0002
181#define OCFS2_OSB_ERROR_FS 0x0004
182
183struct ocfs2_journal;
184struct ocfs2_journal_handle;
185struct ocfs2_super
186{
Mark Fashehccd979b2005-12-15 14:31:24 -0800187 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 struct ocfs2_node_map mounted_map;
197 struct ocfs2_node_map recovery_map;
198 struct ocfs2_node_map umount_map;
199
Mark Fashehccd979b2005-12-15 14:31:24 -0800200 u64 root_blkno;
201 u64 system_dir_blkno;
202 u64 bitmap_blkno;
203 u32 bitmap_cpg;
204 u8 *uuid;
205 char *uuid_str;
206 u8 *vol_label;
207 u64 first_cluster_group_blkno;
208 u32 fs_generation;
209
210 u32 s_feature_compat;
211 u32 s_feature_incompat;
212 u32 s_feature_ro_compat;
213
214 /* Protects s_next_generaion, osb_flags. Could protect more on
215 * osb as it's very short lived. */
216 spinlock_t osb_lock;
217 u32 s_next_generation;
218 unsigned long osb_flags;
219
220 unsigned long s_mount_opt;
221
222 u16 max_slots;
Mark Fashehccd979b2005-12-15 14:31:24 -0800223 s16 node_num;
224 s16 slot_num;
225 int s_sectsize_bits;
226 int s_clustersize;
227 int s_clustersize_bits;
Mark Fashehccd979b2005-12-15 14:31:24 -0800228
229 atomic_t vol_state;
Arjan van de Venc74ec2f2006-01-13 21:54:23 -0800230 struct mutex recovery_lock;
Mark Fashehccd979b2005-12-15 14:31:24 -0800231 struct task_struct *recovery_thread_task;
232 int disable_recovery;
233 wait_queue_head_t checkpoint_event;
234 atomic_t needs_checkpoint;
235 struct ocfs2_journal *journal;
236
237 enum ocfs2_local_alloc_state local_alloc_state;
238 struct buffer_head *local_alloc_bh;
Mark Fasheh883d4ca2006-06-05 16:41:00 -0400239 u64 la_last_gd;
Mark Fashehccd979b2005-12-15 14:31:24 -0800240
241 /* Next two fields are for local node slot recovery during
242 * mount. */
243 int dirty;
244 struct ocfs2_dinode *local_alloc_copy;
245
246 struct ocfs2_alloc_stats alloc_stats;
247 char dev_str[20]; /* "major,minor" of the device */
248
249 struct dlm_ctxt *dlm;
250 struct ocfs2_lock_res osb_super_lockres;
251 struct ocfs2_lock_res osb_rename_lockres;
252 struct dlm_eviction_cb osb_eviction_cb;
253 struct ocfs2_dlm_debug *osb_dlm_debug;
254
255 struct dentry *osb_debug_root;
256
257 wait_queue_head_t recovery_event;
258
259 spinlock_t vote_task_lock;
260 struct task_struct *vote_task;
261 wait_queue_head_t vote_event;
262 unsigned long vote_wake_sequence;
263 unsigned long vote_work_sequence;
264
265 struct list_head blocked_lock_list;
266 unsigned long blocked_lock_count;
267
268 struct list_head vote_list;
269 int vote_count;
270
271 u32 net_key;
272 spinlock_t net_response_lock;
273 unsigned int net_response_ids;
274 struct list_head net_response_list;
275
276 struct o2hb_callback_func osb_hb_up;
277 struct o2hb_callback_func osb_hb_down;
278
279 struct list_head osb_net_handlers;
280
281 wait_queue_head_t osb_mount_event;
282
283 /* Truncate log info */
284 struct inode *osb_tl_inode;
285 struct buffer_head *osb_tl_bh;
286 struct work_struct osb_truncate_log_wq;
Mark Fashehb4df6ed2006-02-22 17:35:08 -0800287
288 struct ocfs2_node_map osb_recovering_orphan_dirs;
289 unsigned int *osb_orphan_wipes;
290 wait_queue_head_t osb_wipe_event;
Mark Fashehccd979b2005-12-15 14:31:24 -0800291};
292
293#define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info)
Mark Fashehccd979b2005-12-15 14:31:24 -0800294
295static inline int ocfs2_should_order_data(struct inode *inode)
296{
297 if (!S_ISREG(inode->i_mode))
298 return 0;
299 if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
300 return 0;
301 return 1;
302}
303
304/* set / clear functions because cluster events can make these happen
305 * in parallel so we want the transitions to be atomic. this also
306 * means that any future flags osb_flags must be protected by spinlock
307 * too! */
308static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
309 unsigned long flag)
310{
311 spin_lock(&osb->osb_lock);
312 osb->osb_flags |= flag;
313 spin_unlock(&osb->osb_lock);
314}
315
316static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
317 int hard)
318{
319 spin_lock(&osb->osb_lock);
320 osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
321 if (hard)
322 osb->osb_flags |= OCFS2_OSB_HARD_RO;
323 else
324 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
325 spin_unlock(&osb->osb_lock);
326}
327
328static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
329{
330 int ret;
331
332 spin_lock(&osb->osb_lock);
333 ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
334 spin_unlock(&osb->osb_lock);
335
336 return ret;
337}
338
339static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
340{
341 int ret;
342
343 spin_lock(&osb->osb_lock);
344 ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
345 spin_unlock(&osb->osb_lock);
346
347 return ret;
348}
349
350#define OCFS2_IS_VALID_DINODE(ptr) \
351 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
352
353#define OCFS2_RO_ON_INVALID_DINODE(__sb, __di) do { \
354 typeof(__di) ____di = (__di); \
355 ocfs2_error((__sb), \
Mark Fashehb0697052006-03-03 10:24:33 -0800356 "Dinode # %llu has bad signature %.*s", \
357 (unsigned long long)(____di)->i_blkno, 7, \
Mark Fashehccd979b2005-12-15 14:31:24 -0800358 (____di)->i_signature); \
359} while (0);
360
361#define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \
362 (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
363
364#define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb) do { \
365 typeof(__eb) ____eb = (__eb); \
366 ocfs2_error((__sb), \
Mark Fashehb0697052006-03-03 10:24:33 -0800367 "Extent Block # %llu has bad signature %.*s", \
368 (unsigned long long)(____eb)->h_blkno, 7, \
Mark Fashehccd979b2005-12-15 14:31:24 -0800369 (____eb)->h_signature); \
370} while (0);
371
372#define OCFS2_IS_VALID_GROUP_DESC(ptr) \
373 (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
374
375#define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd) do { \
376 typeof(__gd) ____gd = (__gd); \
377 ocfs2_error((__sb), \
Mark Fashehb0697052006-03-03 10:24:33 -0800378 "Group Descriptor # %llu has bad signature %.*s", \
379 (unsigned long long)(____gd)->bg_blkno, 7, \
Mark Fashehccd979b2005-12-15 14:31:24 -0800380 (____gd)->bg_signature); \
381} while (0);
382
383static inline unsigned long ino_from_blkno(struct super_block *sb,
384 u64 blkno)
385{
386 return (unsigned long)(blkno & (u64)ULONG_MAX);
387}
388
389static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
390 u32 clusters)
391{
392 int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
393 sb->s_blocksize_bits;
394
395 return (u64)clusters << c_to_b_bits;
396}
397
398static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
399 u64 blocks)
400{
401 int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
402 sb->s_blocksize_bits;
403
404 return (u32)(blocks >> b_to_c_bits);
405}
406
407static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
408 u64 bytes)
409{
410 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
411 unsigned int clusters;
412
413 bytes += OCFS2_SB(sb)->s_clustersize - 1;
414 /* OCFS2 just cannot have enough clusters to overflow this */
415 clusters = (unsigned int)(bytes >> cl_bits);
416
417 return clusters;
418}
419
420static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
421 u64 bytes)
422{
423 bytes += sb->s_blocksize - 1;
424 return bytes >> sb->s_blocksize_bits;
425}
426
427static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
428 u32 clusters)
429{
430 return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
431}
432
433static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
434 u64 bytes)
435{
436 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
437 unsigned int clusters;
438
439 clusters = ocfs2_clusters_for_bytes(sb, bytes);
440 return (u64)clusters << cl_bits;
441}
442
443static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
444 u64 bytes)
445{
446 u64 blocks;
447
448 blocks = ocfs2_blocks_for_bytes(sb, bytes);
449 return blocks << sb->s_blocksize_bits;
450}
451
452static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
453{
454 return (unsigned long)((bytes + 511) >> 9);
455}
456
457#define ocfs2_set_bit ext2_set_bit
458#define ocfs2_clear_bit ext2_clear_bit
459#define ocfs2_test_bit ext2_test_bit
460#define ocfs2_find_next_zero_bit ext2_find_next_zero_bit
461#endif /* OCFS2_H */
462