]>
Commit | Line | Data |
---|---|---|
ccd979bd MF |
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> | |
c74ec2f7 | 36 | #include <linux/mutex.h> |
1fabe148 | 37 | #include <linux/jbd.h> |
ccd979bd MF |
38 | |
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 | ||
ccd979bd MF |
49 | /* Most user visible OCFS2 inodes will have very few pieces of |
50 | * metadata, but larger files (including bitmaps, etc) must be taken | |
51 | * into account when designing an access scheme. We allow a small | |
52 | * amount of inlined blocks to be stored on an array and grow the | |
53 | * structure into a rb tree when necessary. */ | |
54 | #define OCFS2_INODE_MAX_CACHE_ARRAY 2 | |
55 | ||
56 | struct ocfs2_caching_info { | |
57 | unsigned int ci_num_cached; | |
58 | union { | |
59 | sector_t ci_array[OCFS2_INODE_MAX_CACHE_ARRAY]; | |
60 | struct rb_root ci_tree; | |
61 | } ci_cache; | |
62 | }; | |
63 | ||
64 | /* this limits us to 256 nodes | |
65 | * if we need more, we can do a kmalloc for the map */ | |
66 | #define OCFS2_NODE_MAP_MAX_NODES 256 | |
67 | struct ocfs2_node_map { | |
68 | u16 num_nodes; | |
69 | unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)]; | |
70 | }; | |
71 | ||
72 | enum ocfs2_ast_action { | |
73 | OCFS2_AST_INVALID = 0, | |
74 | OCFS2_AST_ATTACH, | |
75 | OCFS2_AST_CONVERT, | |
76 | OCFS2_AST_DOWNCONVERT, | |
77 | }; | |
78 | ||
79 | /* actions for an unlockast function to take. */ | |
80 | enum ocfs2_unlock_action { | |
81 | OCFS2_UNLOCK_INVALID = 0, | |
82 | OCFS2_UNLOCK_CANCEL_CONVERT, | |
83 | OCFS2_UNLOCK_DROP_LOCK, | |
84 | }; | |
85 | ||
86 | /* ocfs2_lock_res->l_flags flags. */ | |
87 | #define OCFS2_LOCK_ATTACHED (0x00000001) /* have we initialized | |
88 | * the lvb */ | |
89 | #define OCFS2_LOCK_BUSY (0x00000002) /* we are currently in | |
90 | * dlm_lock */ | |
91 | #define OCFS2_LOCK_BLOCKED (0x00000004) /* blocked waiting to | |
92 | * downconvert*/ | |
93 | #define OCFS2_LOCK_LOCAL (0x00000008) /* newly created inode */ | |
94 | #define OCFS2_LOCK_NEEDS_REFRESH (0x00000010) | |
95 | #define OCFS2_LOCK_REFRESHING (0x00000020) | |
96 | #define OCFS2_LOCK_INITIALIZED (0x00000040) /* track initialization | |
97 | * for shutdown paths */ | |
98 | #define OCFS2_LOCK_FREEING (0x00000080) /* help dlmglue track | |
99 | * when to skip queueing | |
100 | * a lock because it's | |
101 | * about to be | |
102 | * dropped. */ | |
103 | #define OCFS2_LOCK_QUEUED (0x00000100) /* queued for downconvert */ | |
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 | struct dlm_lockstatus 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 | ||
131 | wait_queue_head_t l_event; | |
132 | ||
133 | struct list_head l_debug_list; | |
134 | }; | |
135 | ||
136 | struct ocfs2_dlm_debug { | |
137 | struct kref d_refcnt; | |
138 | struct dentry *d_locking_state; | |
139 | struct list_head d_lockres_tracking; | |
140 | }; | |
141 | ||
142 | enum ocfs2_vol_state | |
143 | { | |
144 | VOLUME_INIT = 0, | |
145 | VOLUME_MOUNTED, | |
146 | VOLUME_DISMOUNTED, | |
147 | VOLUME_DISABLED | |
148 | }; | |
149 | ||
150 | struct ocfs2_alloc_stats | |
151 | { | |
152 | atomic_t moves; | |
153 | atomic_t local_data; | |
154 | atomic_t bitmap_data; | |
155 | atomic_t bg_allocs; | |
156 | atomic_t bg_extends; | |
157 | }; | |
158 | ||
159 | enum ocfs2_local_alloc_state | |
160 | { | |
161 | OCFS2_LA_UNUSED = 0, | |
162 | OCFS2_LA_ENABLED, | |
163 | OCFS2_LA_DISABLED | |
164 | }; | |
165 | ||
166 | enum ocfs2_mount_options | |
167 | { | |
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 */ | |
ccd979bd MF |
173 | }; |
174 | ||
175 | #define OCFS2_OSB_SOFT_RO 0x0001 | |
176 | #define OCFS2_OSB_HARD_RO 0x0002 | |
177 | #define OCFS2_OSB_ERROR_FS 0x0004 | |
7f1a37e3 | 178 | #define OCFS2_DEFAULT_ATIME_QUANTUM 60 |
ccd979bd MF |
179 | |
180 | struct ocfs2_journal; | |
ccd979bd MF |
181 | struct ocfs2_super |
182 | { | |
ccd979bd MF |
183 | struct task_struct *commit_task; |
184 | struct super_block *sb; | |
185 | struct inode *root_inode; | |
186 | struct inode *sys_root_inode; | |
187 | struct inode *system_inodes[NUM_SYSTEM_INODES]; | |
188 | ||
189 | struct ocfs2_slot_info *slot_info; | |
190 | ||
191 | spinlock_t node_map_lock; | |
192 | struct ocfs2_node_map mounted_map; | |
193 | struct ocfs2_node_map recovery_map; | |
194 | struct ocfs2_node_map umount_map; | |
195 | ||
ccd979bd MF |
196 | u64 root_blkno; |
197 | u64 system_dir_blkno; | |
198 | u64 bitmap_blkno; | |
199 | u32 bitmap_cpg; | |
200 | u8 *uuid; | |
201 | char *uuid_str; | |
202 | u8 *vol_label; | |
203 | u64 first_cluster_group_blkno; | |
204 | u32 fs_generation; | |
205 | ||
206 | u32 s_feature_compat; | |
207 | u32 s_feature_incompat; | |
208 | u32 s_feature_ro_compat; | |
209 | ||
210 | /* Protects s_next_generaion, osb_flags. Could protect more on | |
211 | * osb as it's very short lived. */ | |
212 | spinlock_t osb_lock; | |
213 | u32 s_next_generation; | |
214 | unsigned long osb_flags; | |
215 | ||
216 | unsigned long s_mount_opt; | |
7f1a37e3 | 217 | unsigned int s_atime_quantum; |
ccd979bd MF |
218 | |
219 | u16 max_slots; | |
ccd979bd MF |
220 | s16 node_num; |
221 | s16 slot_num; | |
222 | int s_sectsize_bits; | |
223 | int s_clustersize; | |
224 | int s_clustersize_bits; | |
ccd979bd MF |
225 | |
226 | atomic_t vol_state; | |
c74ec2f7 | 227 | struct mutex recovery_lock; |
ccd979bd MF |
228 | struct task_struct *recovery_thread_task; |
229 | int disable_recovery; | |
230 | wait_queue_head_t checkpoint_event; | |
231 | atomic_t needs_checkpoint; | |
232 | struct ocfs2_journal *journal; | |
233 | ||
234 | enum ocfs2_local_alloc_state local_alloc_state; | |
235 | struct buffer_head *local_alloc_bh; | |
883d4cae | 236 | u64 la_last_gd; |
ccd979bd MF |
237 | |
238 | /* Next two fields are for local node slot recovery during | |
239 | * mount. */ | |
240 | int dirty; | |
241 | struct ocfs2_dinode *local_alloc_copy; | |
242 | ||
243 | struct ocfs2_alloc_stats alloc_stats; | |
244 | char dev_str[20]; /* "major,minor" of the device */ | |
245 | ||
246 | struct dlm_ctxt *dlm; | |
247 | struct ocfs2_lock_res osb_super_lockres; | |
248 | struct ocfs2_lock_res osb_rename_lockres; | |
249 | struct dlm_eviction_cb osb_eviction_cb; | |
250 | struct ocfs2_dlm_debug *osb_dlm_debug; | |
251 | ||
252 | struct dentry *osb_debug_root; | |
253 | ||
254 | wait_queue_head_t recovery_event; | |
255 | ||
256 | spinlock_t vote_task_lock; | |
257 | struct task_struct *vote_task; | |
258 | wait_queue_head_t vote_event; | |
259 | unsigned long vote_wake_sequence; | |
260 | unsigned long vote_work_sequence; | |
261 | ||
262 | struct list_head blocked_lock_list; | |
263 | unsigned long blocked_lock_count; | |
264 | ||
265 | struct list_head vote_list; | |
266 | int vote_count; | |
267 | ||
268 | u32 net_key; | |
269 | spinlock_t net_response_lock; | |
270 | unsigned int net_response_ids; | |
271 | struct list_head net_response_list; | |
272 | ||
273 | struct o2hb_callback_func osb_hb_up; | |
274 | struct o2hb_callback_func osb_hb_down; | |
275 | ||
276 | struct list_head osb_net_handlers; | |
277 | ||
278 | wait_queue_head_t osb_mount_event; | |
279 | ||
280 | /* Truncate log info */ | |
281 | struct inode *osb_tl_inode; | |
282 | struct buffer_head *osb_tl_bh; | |
c4028958 | 283 | struct delayed_work osb_truncate_log_wq; |
b4df6ed8 MF |
284 | |
285 | struct ocfs2_node_map osb_recovering_orphan_dirs; | |
286 | unsigned int *osb_orphan_wipes; | |
287 | wait_queue_head_t osb_wipe_event; | |
ccd979bd MF |
288 | }; |
289 | ||
290 | #define OCFS2_SB(sb) ((struct ocfs2_super *)(sb)->s_fs_info) | |
ccd979bd MF |
291 | |
292 | static inline int ocfs2_should_order_data(struct inode *inode) | |
293 | { | |
294 | if (!S_ISREG(inode->i_mode)) | |
295 | return 0; | |
296 | if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) | |
297 | return 0; | |
298 | return 1; | |
299 | } | |
300 | ||
dcd0538f MF |
301 | static inline int ocfs2_sparse_alloc(struct ocfs2_super *osb) |
302 | { | |
303 | if (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_SPARSE_ALLOC) | |
304 | return 1; | |
305 | return 0; | |
306 | } | |
307 | ||
ccd979bd MF |
308 | /* set / clear functions because cluster events can make these happen |
309 | * in parallel so we want the transitions to be atomic. this also | |
310 | * means that any future flags osb_flags must be protected by spinlock | |
311 | * too! */ | |
312 | static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb, | |
313 | unsigned long flag) | |
314 | { | |
315 | spin_lock(&osb->osb_lock); | |
316 | osb->osb_flags |= flag; | |
317 | spin_unlock(&osb->osb_lock); | |
318 | } | |
319 | ||
320 | static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb, | |
321 | int hard) | |
322 | { | |
323 | spin_lock(&osb->osb_lock); | |
324 | osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO); | |
325 | if (hard) | |
326 | osb->osb_flags |= OCFS2_OSB_HARD_RO; | |
327 | else | |
328 | osb->osb_flags |= OCFS2_OSB_SOFT_RO; | |
329 | spin_unlock(&osb->osb_lock); | |
330 | } | |
331 | ||
332 | static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb) | |
333 | { | |
334 | int ret; | |
335 | ||
336 | spin_lock(&osb->osb_lock); | |
337 | ret = osb->osb_flags & OCFS2_OSB_HARD_RO; | |
338 | spin_unlock(&osb->osb_lock); | |
339 | ||
340 | return ret; | |
341 | } | |
342 | ||
343 | static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb) | |
344 | { | |
345 | int ret; | |
346 | ||
347 | spin_lock(&osb->osb_lock); | |
348 | ret = osb->osb_flags & OCFS2_OSB_SOFT_RO; | |
349 | spin_unlock(&osb->osb_lock); | |
350 | ||
351 | return ret; | |
352 | } | |
353 | ||
c271c5c2 SM |
354 | static inline int ocfs2_mount_local(struct ocfs2_super *osb) |
355 | { | |
356 | return (osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT); | |
357 | } | |
358 | ||
ccd979bd MF |
359 | #define OCFS2_IS_VALID_DINODE(ptr) \ |
360 | (!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE)) | |
361 | ||
362 | #define OCFS2_RO_ON_INVALID_DINODE(__sb, __di) do { \ | |
363 | typeof(__di) ____di = (__di); \ | |
364 | ocfs2_error((__sb), \ | |
b0697053 MF |
365 | "Dinode # %llu has bad signature %.*s", \ |
366 | (unsigned long long)(____di)->i_blkno, 7, \ | |
ccd979bd MF |
367 | (____di)->i_signature); \ |
368 | } while (0); | |
369 | ||
370 | #define OCFS2_IS_VALID_EXTENT_BLOCK(ptr) \ | |
371 | (!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE)) | |
372 | ||
373 | #define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb) do { \ | |
374 | typeof(__eb) ____eb = (__eb); \ | |
375 | ocfs2_error((__sb), \ | |
b0697053 MF |
376 | "Extent Block # %llu has bad signature %.*s", \ |
377 | (unsigned long long)(____eb)->h_blkno, 7, \ | |
ccd979bd MF |
378 | (____eb)->h_signature); \ |
379 | } while (0); | |
380 | ||
381 | #define OCFS2_IS_VALID_GROUP_DESC(ptr) \ | |
382 | (!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE)) | |
383 | ||
384 | #define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd) do { \ | |
385 | typeof(__gd) ____gd = (__gd); \ | |
386 | ocfs2_error((__sb), \ | |
b0697053 MF |
387 | "Group Descriptor # %llu has bad signature %.*s", \ |
388 | (unsigned long long)(____gd)->bg_blkno, 7, \ | |
ccd979bd MF |
389 | (____gd)->bg_signature); \ |
390 | } while (0); | |
391 | ||
392 | static inline unsigned long ino_from_blkno(struct super_block *sb, | |
393 | u64 blkno) | |
394 | { | |
395 | return (unsigned long)(blkno & (u64)ULONG_MAX); | |
396 | } | |
397 | ||
398 | static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb, | |
399 | u32 clusters) | |
400 | { | |
401 | int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits - | |
402 | sb->s_blocksize_bits; | |
403 | ||
404 | return (u64)clusters << c_to_b_bits; | |
405 | } | |
406 | ||
407 | static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb, | |
408 | u64 blocks) | |
409 | { | |
410 | int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits - | |
411 | sb->s_blocksize_bits; | |
412 | ||
413 | return (u32)(blocks >> b_to_c_bits); | |
414 | } | |
415 | ||
416 | static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb, | |
417 | u64 bytes) | |
418 | { | |
419 | int cl_bits = OCFS2_SB(sb)->s_clustersize_bits; | |
420 | unsigned int clusters; | |
421 | ||
422 | bytes += OCFS2_SB(sb)->s_clustersize - 1; | |
423 | /* OCFS2 just cannot have enough clusters to overflow this */ | |
424 | clusters = (unsigned int)(bytes >> cl_bits); | |
425 | ||
426 | return clusters; | |
427 | } | |
428 | ||
429 | static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb, | |
430 | u64 bytes) | |
431 | { | |
432 | bytes += sb->s_blocksize - 1; | |
433 | return bytes >> sb->s_blocksize_bits; | |
434 | } | |
435 | ||
436 | static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb, | |
437 | u32 clusters) | |
438 | { | |
439 | return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits; | |
440 | } | |
441 | ||
442 | static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb, | |
443 | u64 bytes) | |
444 | { | |
445 | int cl_bits = OCFS2_SB(sb)->s_clustersize_bits; | |
446 | unsigned int clusters; | |
447 | ||
448 | clusters = ocfs2_clusters_for_bytes(sb, bytes); | |
449 | return (u64)clusters << cl_bits; | |
450 | } | |
451 | ||
452 | static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb, | |
453 | u64 bytes) | |
454 | { | |
455 | u64 blocks; | |
456 | ||
457 | blocks = ocfs2_blocks_for_bytes(sb, bytes); | |
458 | return blocks << sb->s_blocksize_bits; | |
459 | } | |
460 | ||
461 | static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes) | |
462 | { | |
463 | return (unsigned long)((bytes + 511) >> 9); | |
464 | } | |
465 | ||
9517bac6 MF |
466 | static inline unsigned int ocfs2_page_index_to_clusters(struct super_block *sb, |
467 | unsigned long pg_index) | |
468 | { | |
469 | u32 clusters = pg_index; | |
470 | unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; | |
471 | ||
472 | if (unlikely(PAGE_CACHE_SHIFT > cbits)) | |
473 | clusters = pg_index << (PAGE_CACHE_SHIFT - cbits); | |
474 | else if (PAGE_CACHE_SHIFT < cbits) | |
475 | clusters = pg_index >> (cbits - PAGE_CACHE_SHIFT); | |
476 | ||
477 | return clusters; | |
478 | } | |
479 | ||
480 | /* | |
481 | * Find the 1st page index which covers the given clusters. | |
482 | */ | |
483 | static inline unsigned long ocfs2_align_clusters_to_page_index(struct super_block *sb, | |
484 | u32 clusters) | |
485 | { | |
486 | unsigned int cbits = OCFS2_SB(sb)->s_clustersize_bits; | |
487 | unsigned long index = clusters; | |
488 | ||
489 | if (PAGE_CACHE_SHIFT > cbits) { | |
490 | index = clusters >> (PAGE_CACHE_SHIFT - cbits); | |
491 | } else if (PAGE_CACHE_SHIFT < cbits) { | |
492 | index = clusters << (cbits - PAGE_CACHE_SHIFT); | |
493 | } | |
494 | ||
495 | return index; | |
496 | } | |
497 | ||
ccd979bd MF |
498 | #define ocfs2_set_bit ext2_set_bit |
499 | #define ocfs2_clear_bit ext2_clear_bit | |
500 | #define ocfs2_test_bit ext2_test_bit | |
501 | #define ocfs2_find_next_zero_bit ext2_find_next_zero_bit | |
502 | #endif /* OCFS2_H */ | |
503 |