blob: b767fd7da6ebaeb44ed58a888b80cb2dd99ce545 [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 Fasheh1fabe142006-10-09 18:11:45 -070037#include <linux/jbd.h>
Mark Fashehccd979b2005-12-15 14:31:24 -080038
39#include "cluster/nodemanager.h"
40#include "cluster/heartbeat.h"
41#include "cluster/tcp.h"
42
43#include "dlm/dlmapi.h"
44
45#include "ocfs2_fs.h"
46#include "endian.h"
47#include "ocfs2_lockid.h"
48
49struct ocfs2_extent_map {
50 u32 em_clusters;
51 struct rb_root em_extents;
52};
53
54/* Most user visible OCFS2 inodes will have very few pieces of
55 * metadata, but larger files (including bitmaps, etc) must be taken
56 * into account when designing an access scheme. We allow a small
57 * amount of inlined blocks to be stored on an array and grow the
58 * structure into a rb tree when necessary. */
59#define OCFS2_INODE_MAX_CACHE_ARRAY 2
60
61struct ocfs2_caching_info {
62 unsigned int ci_num_cached;
63 union {
64 sector_t ci_array[OCFS2_INODE_MAX_CACHE_ARRAY];
65 struct rb_root ci_tree;
66 } ci_cache;
67};
68
69/* this limits us to 256 nodes
70 * if we need more, we can do a kmalloc for the map */
71#define OCFS2_NODE_MAP_MAX_NODES 256
72struct ocfs2_node_map {
73 u16 num_nodes;
74 unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)];
75};
76
77enum ocfs2_ast_action {
78 OCFS2_AST_INVALID = 0,
79 OCFS2_AST_ATTACH,
80 OCFS2_AST_CONVERT,
81 OCFS2_AST_DOWNCONVERT,
82};
83
84/* actions for an unlockast function to take. */
85enum ocfs2_unlock_action {
86 OCFS2_UNLOCK_INVALID = 0,
87 OCFS2_UNLOCK_CANCEL_CONVERT,
88 OCFS2_UNLOCK_DROP_LOCK,
89};
90
91/* ocfs2_lock_res->l_flags flags. */
92#define OCFS2_LOCK_ATTACHED (0x00000001) /* have we initialized
93 * the lvb */
94#define OCFS2_LOCK_BUSY (0x00000002) /* we are currently in
95 * dlm_lock */
96#define OCFS2_LOCK_BLOCKED (0x00000004) /* blocked waiting to
97 * downconvert*/
98#define OCFS2_LOCK_LOCAL (0x00000008) /* newly created inode */
99#define OCFS2_LOCK_NEEDS_REFRESH (0x00000010)
100#define OCFS2_LOCK_REFRESHING (0x00000020)
101#define OCFS2_LOCK_INITIALIZED (0x00000040) /* track initialization
102 * for shutdown paths */
103#define OCFS2_LOCK_FREEING (0x00000080) /* help dlmglue track
104 * when to skip queueing
105 * a lock because it's
106 * about to be
107 * dropped. */
108#define OCFS2_LOCK_QUEUED (0x00000100) /* queued for downconvert */
109
110struct ocfs2_lock_res_ops;
111
112typedef void (*ocfs2_lock_callback)(int status, unsigned long data);
113
114struct ocfs2_lock_res {
115 void *l_priv;
116 struct ocfs2_lock_res_ops *l_ops;
117 spinlock_t l_lock;
118
119 struct list_head l_blocked_list;
120 struct list_head l_mask_waiters;
121
122 enum ocfs2_lock_type l_type;
123 unsigned long l_flags;
124 char l_name[OCFS2_LOCK_ID_MAX_LEN];
125 int l_level;
126 unsigned int l_ro_holders;
127 unsigned int l_ex_holders;
128 struct dlm_lockstatus l_lksb;
129
130 /* used from AST/BAST funcs. */
131 enum ocfs2_ast_action l_action;
132 enum ocfs2_unlock_action l_unlock_action;
133 int l_requested;
134 int l_blocking;
135
136 wait_queue_head_t l_event;
137
138 struct list_head l_debug_list;
139};
140
141struct ocfs2_dlm_debug {
142 struct kref d_refcnt;
143 struct dentry *d_locking_state;
144 struct list_head d_lockres_tracking;
145};
146
147enum ocfs2_vol_state
148{
149 VOLUME_INIT = 0,
150 VOLUME_MOUNTED,
151 VOLUME_DISMOUNTED,
152 VOLUME_DISABLED
153};
154
155struct ocfs2_alloc_stats
156{
157 atomic_t moves;
158 atomic_t local_data;
159 atomic_t bitmap_data;
160 atomic_t bg_allocs;
161 atomic_t bg_extends;
162};
163
164enum ocfs2_local_alloc_state
165{
166 OCFS2_LA_UNUSED = 0,
167 OCFS2_LA_ENABLED,
168 OCFS2_LA_DISABLED
169};
170
171enum ocfs2_mount_options
172{
173 OCFS2_MOUNT_HB_LOCAL = 1 << 0, /* Heartbeat started in local mode */
174 OCFS2_MOUNT_BARRIER = 1 << 1, /* Use block barriers */
175 OCFS2_MOUNT_NOINTR = 1 << 2, /* Don't catch signals */
176 OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */
177 OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */
Mark Fashehccd979b2005-12-15 14:31:24 -0800178};
179
180#define OCFS2_OSB_SOFT_RO 0x0001
181#define OCFS2_OSB_HARD_RO 0x0002
182#define OCFS2_OSB_ERROR_FS 0x0004
Tiger Yang7f1a37e2006-11-15 15:48:42 +0800183#define OCFS2_DEFAULT_ATIME_QUANTUM 60
Mark Fashehccd979b2005-12-15 14:31:24 -0800184
185struct ocfs2_journal;
Mark Fashehccd979b2005-12-15 14:31:24 -0800186struct ocfs2_super
187{
Mark Fashehccd979b2005-12-15 14:31:24 -0800188 struct task_struct *commit_task;
189 struct super_block *sb;
190 struct inode *root_inode;
191 struct inode *sys_root_inode;
192 struct inode *system_inodes[NUM_SYSTEM_INODES];
193
194 struct ocfs2_slot_info *slot_info;
195
196 spinlock_t node_map_lock;
197 struct ocfs2_node_map mounted_map;
198 struct ocfs2_node_map recovery_map;
199 struct ocfs2_node_map umount_map;
200
Mark Fashehccd979b2005-12-15 14:31:24 -0800201 u64 root_blkno;
202 u64 system_dir_blkno;
203 u64 bitmap_blkno;
204 u32 bitmap_cpg;
205 u8 *uuid;
206 char *uuid_str;
207 u8 *vol_label;
208 u64 first_cluster_group_blkno;
209 u32 fs_generation;
210
211 u32 s_feature_compat;
212 u32 s_feature_incompat;
213 u32 s_feature_ro_compat;
214
215 /* Protects s_next_generaion, osb_flags. Could protect more on
216 * osb as it's very short lived. */
217 spinlock_t osb_lock;
218 u32 s_next_generation;
219 unsigned long osb_flags;
220
221 unsigned long s_mount_opt;
Tiger Yang7f1a37e2006-11-15 15:48:42 +0800222 unsigned int s_atime_quantum;
Mark Fashehccd979b2005-12-15 14:31:24 -0800223
224 u16 max_slots;
Mark Fashehccd979b2005-12-15 14:31:24 -0800225 s16 node_num;
226 s16 slot_num;
227 int s_sectsize_bits;
228 int s_clustersize;
229 int s_clustersize_bits;
Mark Fashehccd979b2005-12-15 14:31:24 -0800230
231 atomic_t vol_state;
Arjan van de Venc74ec2f2006-01-13 21:54:23 -0800232 struct mutex recovery_lock;
Mark Fashehccd979b2005-12-15 14:31:24 -0800233 struct task_struct *recovery_thread_task;
234 int disable_recovery;
235 wait_queue_head_t checkpoint_event;
236 atomic_t needs_checkpoint;
237 struct ocfs2_journal *journal;
238
239 enum ocfs2_local_alloc_state local_alloc_state;
240 struct buffer_head *local_alloc_bh;
Mark Fasheh883d4ca2006-06-05 16:41:00 -0400241 u64 la_last_gd;
Mark Fashehccd979b2005-12-15 14:31:24 -0800242
243 /* Next two fields are for local node slot recovery during
244 * mount. */
245 int dirty;
246 struct ocfs2_dinode *local_alloc_copy;
247
248 struct ocfs2_alloc_stats alloc_stats;
249 char dev_str[20]; /* "major,minor" of the device */
250
251 struct dlm_ctxt *dlm;
252 struct ocfs2_lock_res osb_super_lockres;
253 struct ocfs2_lock_res osb_rename_lockres;
254 struct dlm_eviction_cb osb_eviction_cb;
255 struct ocfs2_dlm_debug *osb_dlm_debug;
256
257 struct dentry *osb_debug_root;
258
259 wait_queue_head_t recovery_event;
260
261 spinlock_t vote_task_lock;
262 struct task_struct *vote_task;
263 wait_queue_head_t vote_event;
264 unsigned long vote_wake_sequence;
265 unsigned long vote_work_sequence;
266
267 struct list_head blocked_lock_list;
268 unsigned long blocked_lock_count;
269
270 struct list_head vote_list;
271 int vote_count;
272
273 u32 net_key;
274 spinlock_t net_response_lock;
275 unsigned int net_response_ids;
276 struct list_head net_response_list;
277
278 struct o2hb_callback_func osb_hb_up;
279 struct o2hb_callback_func osb_hb_down;
280
281 struct list_head osb_net_handlers;
282
283 wait_queue_head_t osb_mount_event;
284
285 /* Truncate log info */
286 struct inode *osb_tl_inode;
287 struct buffer_head *osb_tl_bh;
David Howellsc4028952006-11-22 14:57:56 +0000288 struct delayed_work osb_truncate_log_wq;
Mark Fashehb4df6ed2006-02-22 17:35:08 -0800289
290 struct ocfs2_node_map osb_recovering_orphan_dirs;
291 unsigned int *osb_orphan_wipes;
292 wait_queue_head_t osb_wipe_event;
Mark Fashehccd979b2005-12-15 14:31:24 -0800293};
294
295#define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info)
Mark Fashehccd979b2005-12-15 14:31:24 -0800296
297static inline int ocfs2_should_order_data(struct inode *inode)
298{
299 if (!S_ISREG(inode->i_mode))
300 return 0;
301 if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
302 return 0;
303 return 1;
304}
305
306/* set / clear functions because cluster events can make these happen
307 * in parallel so we want the transitions to be atomic. this also
308 * means that any future flags osb_flags must be protected by spinlock
309 * too! */
310static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
311 unsigned long flag)
312{
313 spin_lock(&osb->osb_lock);
314 osb->osb_flags |= flag;
315 spin_unlock(&osb->osb_lock);
316}
317
318static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
319 int hard)
320{
321 spin_lock(&osb->osb_lock);
322 osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
323 if (hard)
324 osb->osb_flags |= OCFS2_OSB_HARD_RO;
325 else
326 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
327 spin_unlock(&osb->osb_lock);
328}
329
330static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
331{
332 int ret;
333
334 spin_lock(&osb->osb_lock);
335 ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
336 spin_unlock(&osb->osb_lock);
337
338 return ret;
339}
340
341static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
342{
343 int ret;
344
345 spin_lock(&osb->osb_lock);
346 ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
347 spin_unlock(&osb->osb_lock);
348
349 return ret;
350}
351
352#define OCFS2_IS_VALID_DINODE(ptr) \
353 (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
354
355#define OCFS2_RO_ON_INVALID_DINODE(__sb, __di) do { \
356 typeof(__di) ____di = (__di); \
357 ocfs2_error((__sb), \
Mark Fashehb06970532006-03-03 10:24:33 -0800358 "Dinode # %llu has bad signature %.*s", \
359 (unsigned long long)(____di)->i_blkno, 7, \
Mark Fashehccd979b2005-12-15 14:31:24 -0800360 (____di)->i_signature); \
361} while (0);
362
363#define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \
364 (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
365
366#define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb) do { \
367 typeof(__eb) ____eb = (__eb); \
368 ocfs2_error((__sb), \
Mark Fashehb06970532006-03-03 10:24:33 -0800369 "Extent Block # %llu has bad signature %.*s", \
370 (unsigned long long)(____eb)->h_blkno, 7, \
Mark Fashehccd979b2005-12-15 14:31:24 -0800371 (____eb)->h_signature); \
372} while (0);
373
374#define OCFS2_IS_VALID_GROUP_DESC(ptr) \
375 (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
376
377#define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd) do { \
378 typeof(__gd) ____gd = (__gd); \
379 ocfs2_error((__sb), \
Mark Fashehb06970532006-03-03 10:24:33 -0800380 "Group Descriptor # %llu has bad signature %.*s", \
381 (unsigned long long)(____gd)->bg_blkno, 7, \
Mark Fashehccd979b2005-12-15 14:31:24 -0800382 (____gd)->bg_signature); \
383} while (0);
384
385static inline unsigned long ino_from_blkno(struct super_block *sb,
386 u64 blkno)
387{
388 return (unsigned long)(blkno & (u64)ULONG_MAX);
389}
390
391static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
392 u32 clusters)
393{
394 int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
395 sb->s_blocksize_bits;
396
397 return (u64)clusters << c_to_b_bits;
398}
399
400static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
401 u64 blocks)
402{
403 int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
404 sb->s_blocksize_bits;
405
406 return (u32)(blocks >> b_to_c_bits);
407}
408
409static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
410 u64 bytes)
411{
412 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
413 unsigned int clusters;
414
415 bytes += OCFS2_SB(sb)->s_clustersize - 1;
416 /* OCFS2 just cannot have enough clusters to overflow this */
417 clusters = (unsigned int)(bytes >> cl_bits);
418
419 return clusters;
420}
421
422static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
423 u64 bytes)
424{
425 bytes += sb->s_blocksize - 1;
426 return bytes >> sb->s_blocksize_bits;
427}
428
429static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
430 u32 clusters)
431{
432 return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
433}
434
435static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
436 u64 bytes)
437{
438 int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
439 unsigned int clusters;
440
441 clusters = ocfs2_clusters_for_bytes(sb, bytes);
442 return (u64)clusters << cl_bits;
443}
444
445static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
446 u64 bytes)
447{
448 u64 blocks;
449
450 blocks = ocfs2_blocks_for_bytes(sb, bytes);
451 return blocks << sb->s_blocksize_bits;
452}
453
454static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
455{
456 return (unsigned long)((bytes + 511) >> 9);
457}
458
459#define ocfs2_set_bit ext2_set_bit
460#define ocfs2_clear_bit ext2_clear_bit
461#define ocfs2_test_bit ext2_test_bit
462#define ocfs2_find_next_zero_bit ext2_find_next_zero_bit
463#endif /* OCFS2_H */
464