[GFS2] Fix lock ordering bug in page fault path
[deliverable/linux.git] / fs / gfs2 / incore.h
1 /*
2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
3 * Copyright (C) 2004-2006 Red Hat, Inc. All rights reserved.
4 *
5 * This copyrighted material is made available to anyone wishing to use,
6 * modify, copy, or redistribute it subject to the terms and conditions
7 * of the GNU General Public License v.2.
8 */
9
10 #ifndef __INCORE_DOT_H__
11 #define __INCORE_DOT_H__
12
13 #define DIO_FORCE 0x00000001
14 #define DIO_CLEAN 0x00000002
15 #define DIO_DIRTY 0x00000004
16 #define DIO_START 0x00000008
17 #define DIO_WAIT 0x00000010
18 #define DIO_METADATA 0x00000020
19 #define DIO_DATA 0x00000040
20 #define DIO_RELEASE 0x00000080
21 #define DIO_ALL 0x00000100
22
23 struct gfs2_log_operations;
24 struct gfs2_log_element;
25 struct gfs2_bitmap;
26 struct gfs2_rgrpd;
27 struct gfs2_bufdata;
28 struct gfs2_glock_operations;
29 struct gfs2_holder;
30 struct gfs2_glock;
31 struct gfs2_alloc;
32 struct gfs2_inode;
33 struct gfs2_file;
34 struct gfs2_revoke;
35 struct gfs2_revoke_replay;
36 struct gfs2_quota_data;
37 struct gfs2_log_buf;
38 struct gfs2_trans;
39 struct gfs2_ail;
40 struct gfs2_jdesc;
41 struct gfs2_args;
42 struct gfs2_tune;
43 struct gfs2_gl_hash_bucket;
44 struct gfs2_sbd;
45
46 typedef void (*gfs2_glop_bh_t) (struct gfs2_glock *gl, unsigned int ret);
47
48 /*
49 * Structure of operations that are associated with each
50 * type of element in the log.
51 */
52
53 struct gfs2_log_operations {
54 void (*lo_add) (struct gfs2_sbd *sdp, struct gfs2_log_element *le);
55 void (*lo_incore_commit) (struct gfs2_sbd *sdp, struct gfs2_trans *tr);
56 void (*lo_before_commit) (struct gfs2_sbd *sdp);
57 void (*lo_after_commit) (struct gfs2_sbd *sdp, struct gfs2_ail *ai);
58 void (*lo_before_scan) (struct gfs2_jdesc *jd,
59 struct gfs2_log_header *head, int pass);
60 int (*lo_scan_elements) (struct gfs2_jdesc *jd, unsigned int start,
61 struct gfs2_log_descriptor *ld, __be64 *ptr,
62 int pass);
63 void (*lo_after_scan) (struct gfs2_jdesc *jd, int error, int pass);
64 const char *lo_name;
65 };
66
67 struct gfs2_log_element {
68 struct list_head le_list;
69 const struct gfs2_log_operations *le_ops;
70 };
71
72 struct gfs2_bitmap {
73 struct buffer_head *bi_bh;
74 char *bi_clone;
75 uint32_t bi_offset;
76 uint32_t bi_start;
77 uint32_t bi_len;
78 };
79
80 struct gfs2_rgrpd {
81 struct list_head rd_list; /* Link with superblock */
82 struct list_head rd_list_mru;
83 struct list_head rd_recent; /* Recently used rgrps */
84 struct gfs2_glock *rd_gl; /* Glock for this rgrp */
85 struct gfs2_rindex rd_ri;
86 struct gfs2_rgrp rd_rg;
87 uint64_t rd_rg_vn;
88 struct gfs2_bitmap *rd_bits;
89 unsigned int rd_bh_count;
90 struct mutex rd_mutex;
91 uint32_t rd_free_clone;
92 struct gfs2_log_element rd_le;
93 uint32_t rd_last_alloc_data;
94 uint32_t rd_last_alloc_meta;
95 struct gfs2_sbd *rd_sbd;
96 };
97
98 enum gfs2_state_bits {
99 BH_Pinned = BH_PrivateStart,
100 BH_Escaped = BH_PrivateStart + 1,
101 };
102
103 BUFFER_FNS(Pinned, pinned)
104 TAS_BUFFER_FNS(Pinned, pinned)
105 BUFFER_FNS(Escaped, escaped)
106 TAS_BUFFER_FNS(Escaped, escaped)
107
108 struct gfs2_bufdata {
109 struct buffer_head *bd_bh;
110 struct gfs2_glock *bd_gl;
111
112 struct list_head bd_list_tr;
113 struct gfs2_log_element bd_le;
114
115 struct gfs2_ail *bd_ail;
116 struct list_head bd_ail_st_list;
117 struct list_head bd_ail_gl_list;
118 };
119
120 struct gfs2_glock_operations {
121 void (*go_xmote_th) (struct gfs2_glock * gl, unsigned int state,
122 int flags);
123 void (*go_xmote_bh) (struct gfs2_glock * gl);
124 void (*go_drop_th) (struct gfs2_glock * gl);
125 void (*go_drop_bh) (struct gfs2_glock * gl);
126 void (*go_sync) (struct gfs2_glock * gl, int flags);
127 void (*go_inval) (struct gfs2_glock * gl, int flags);
128 int (*go_demote_ok) (struct gfs2_glock * gl);
129 int (*go_lock) (struct gfs2_holder * gh);
130 void (*go_unlock) (struct gfs2_holder * gh);
131 void (*go_callback) (struct gfs2_glock * gl, unsigned int state);
132 void (*go_greedy) (struct gfs2_glock * gl);
133 int go_type;
134 };
135
136 enum {
137 /* Actions */
138 HIF_MUTEX = 0,
139 HIF_PROMOTE = 1,
140 HIF_DEMOTE = 2,
141 HIF_GREEDY = 3,
142
143 /* States */
144 HIF_ALLOCED = 4,
145 HIF_DEALLOC = 5,
146 HIF_HOLDER = 6,
147 HIF_FIRST = 7,
148 HIF_ABORTED = 9,
149 };
150
151 struct gfs2_holder {
152 struct list_head gh_list;
153
154 struct gfs2_glock *gh_gl;
155 struct task_struct *gh_owner;
156 unsigned int gh_state;
157 unsigned gh_flags;
158
159 int gh_error;
160 unsigned long gh_iflags;
161 struct completion gh_wait;
162 unsigned long gh_ip;
163 };
164
165 enum {
166 GLF_PLUG = 0,
167 GLF_LOCK = 1,
168 GLF_STICKY = 2,
169 GLF_PREFETCH = 3,
170 GLF_SYNC = 4,
171 GLF_DIRTY = 5,
172 GLF_SKIP_WAITERS2 = 6,
173 GLF_GREEDY = 7,
174 };
175
176 struct gfs2_glock {
177 struct list_head gl_list;
178 unsigned long gl_flags; /* GLF_... */
179 struct lm_lockname gl_name;
180 struct kref gl_ref;
181
182 spinlock_t gl_spin;
183
184 unsigned int gl_state;
185 struct task_struct *gl_owner;
186 unsigned long gl_ip;
187 struct list_head gl_holders;
188 struct list_head gl_waiters1; /* HIF_MUTEX */
189 struct list_head gl_waiters2; /* HIF_DEMOTE, HIF_GREEDY */
190 struct list_head gl_waiters3; /* HIF_PROMOTE */
191
192 struct gfs2_glock_operations *gl_ops;
193
194 struct gfs2_holder *gl_req_gh;
195 gfs2_glop_bh_t gl_req_bh;
196
197 lm_lock_t *gl_lock;
198 char *gl_lvb;
199 atomic_t gl_lvb_count;
200
201 uint64_t gl_vn;
202 unsigned long gl_stamp;
203 void *gl_object;
204
205 struct gfs2_gl_hash_bucket *gl_bucket;
206 struct list_head gl_reclaim;
207
208 struct gfs2_sbd *gl_sbd;
209
210 struct inode *gl_aspace;
211 struct gfs2_log_element gl_le;
212 struct list_head gl_ail_list;
213 atomic_t gl_ail_count;
214 };
215
216 struct gfs2_alloc {
217 /* Quota stuff */
218
219 struct gfs2_quota_data *al_qd[4];
220 struct gfs2_holder al_qd_ghs[4];
221 unsigned int al_qd_num;
222
223 u32 al_requested; /* Filled in by caller of gfs2_inplace_reserve() */
224 u32 al_alloced; /* Filled in by gfs2_alloc_*() */
225
226 /* Filled in by gfs2_inplace_reserve() */
227
228 unsigned int al_line;
229 char *al_file;
230 struct gfs2_holder al_ri_gh;
231 struct gfs2_holder al_rgd_gh;
232 struct gfs2_rgrpd *al_rgd;
233
234 };
235
236 enum {
237 GIF_QD_LOCKED = 1,
238 GIF_PAGED = 2,
239 GIF_SW_PAGED = 3,
240 };
241
242 struct gfs2_inode {
243 struct inode i_inode;
244 struct gfs2_inum i_num;
245
246 unsigned long i_flags; /* GIF_... */
247
248 uint64_t i_vn;
249 struct gfs2_dinode i_di; /* To be replaced by ref to block */
250
251 struct gfs2_glock *i_gl; /* Move into i_gh? */
252 struct gfs2_holder i_iopen_gh;
253 struct gfs2_holder i_gh; /* for prepare/commit_write only */
254 struct gfs2_alloc i_alloc;
255 uint64_t i_last_rg_alloc;
256
257 spinlock_t i_spin;
258 struct rw_semaphore i_rw_mutex;
259 unsigned int i_greedy;
260 unsigned long i_last_pfault;
261
262 struct buffer_head *i_cache[GFS2_MAX_META_HEIGHT];
263 };
264
265 /*
266 * Since i_inode is the first element of struct gfs2_inode,
267 * this is effectively a cast.
268 */
269 static inline struct gfs2_inode *GFS2_I(struct inode *inode)
270 {
271 return container_of(inode, struct gfs2_inode, i_inode);
272 }
273
274 /* To be removed? */
275 static inline struct gfs2_sbd *GFS2_SB(struct inode *inode)
276 {
277 return inode->i_sb->s_fs_info;
278 }
279
280 enum {
281 GFF_DID_DIRECT_ALLOC = 0,
282 GFF_EXLOCK = 1,
283 };
284
285 struct gfs2_file {
286 unsigned long f_flags; /* GFF_... */
287 struct mutex f_fl_mutex;
288 struct gfs2_holder f_fl_gh;
289 };
290
291 struct gfs2_revoke {
292 struct gfs2_log_element rv_le;
293 uint64_t rv_blkno;
294 };
295
296 struct gfs2_revoke_replay {
297 struct list_head rr_list;
298 uint64_t rr_blkno;
299 unsigned int rr_where;
300 };
301
302 enum {
303 QDF_USER = 0,
304 QDF_CHANGE = 1,
305 QDF_LOCKED = 2,
306 };
307
308 struct gfs2_quota_lvb {
309 uint32_t qb_magic;
310 uint32_t __pad;
311 uint64_t qb_limit; /* Hard limit of # blocks to alloc */
312 uint64_t qb_warn; /* Warn user when alloc is above this # */
313 int64_t qb_value; /* Current # blocks allocated */
314 };
315
316 struct gfs2_quota_data {
317 struct list_head qd_list;
318 unsigned int qd_count;
319
320 uint32_t qd_id;
321 unsigned long qd_flags; /* QDF_... */
322
323 int64_t qd_change;
324 int64_t qd_change_sync;
325
326 unsigned int qd_slot;
327 unsigned int qd_slot_count;
328
329 struct buffer_head *qd_bh;
330 struct gfs2_quota_change *qd_bh_qc;
331 unsigned int qd_bh_count;
332
333 struct gfs2_glock *qd_gl;
334 struct gfs2_quota_lvb qd_qb;
335
336 uint64_t qd_sync_gen;
337 unsigned long qd_last_warn;
338 unsigned long qd_last_touched;
339 };
340
341 struct gfs2_log_buf {
342 struct list_head lb_list;
343 struct buffer_head *lb_bh;
344 struct buffer_head *lb_real;
345 };
346
347 struct gfs2_trans {
348 unsigned long tr_ip;
349
350 unsigned int tr_blocks;
351 unsigned int tr_revokes;
352 unsigned int tr_reserved;
353
354 struct gfs2_holder tr_t_gh;
355
356 int tr_touched;
357
358 unsigned int tr_num_buf;
359 unsigned int tr_num_buf_new;
360 unsigned int tr_num_buf_rm;
361 struct list_head tr_list_buf;
362
363 unsigned int tr_num_revoke;
364 unsigned int tr_num_revoke_rm;
365 };
366
367 struct gfs2_ail {
368 struct list_head ai_list;
369
370 unsigned int ai_first;
371 struct list_head ai_ail1_list;
372 struct list_head ai_ail2_list;
373
374 uint64_t ai_sync_gen;
375 };
376
377 struct gfs2_jdesc {
378 struct list_head jd_list;
379
380 struct inode *jd_inode;
381 unsigned int jd_jid;
382 int jd_dirty;
383
384 unsigned int jd_blocks;
385 };
386
387 #define GFS2_GLOCKD_DEFAULT 1
388 #define GFS2_GLOCKD_MAX 16
389
390 #define GFS2_QUOTA_DEFAULT GFS2_QUOTA_OFF
391 #define GFS2_QUOTA_OFF 0
392 #define GFS2_QUOTA_ACCOUNT 1
393 #define GFS2_QUOTA_ON 2
394
395 #define GFS2_DATA_DEFAULT GFS2_DATA_ORDERED
396 #define GFS2_DATA_WRITEBACK 1
397 #define GFS2_DATA_ORDERED 2
398
399 struct gfs2_args {
400 char ar_lockproto[GFS2_LOCKNAME_LEN]; /* Name of the Lock Protocol */
401 char ar_locktable[GFS2_LOCKNAME_LEN]; /* Name of the Lock Table */
402 char ar_hostdata[GFS2_LOCKNAME_LEN]; /* Host specific data */
403 int ar_spectator; /* Don't get a journal because we're always RO */
404 int ar_ignore_local_fs; /* Don't optimize even if local_fs is 1 */
405 int ar_localflocks; /* Let the VFS do flock|fcntl locks for us */
406 int ar_localcaching; /* Local-style caching (dangerous on multihost) */
407 int ar_debug; /* Oops on errors instead of trying to be graceful */
408 int ar_upgrade; /* Upgrade ondisk/multihost format */
409 unsigned int ar_num_glockd; /* Number of glockd threads */
410 int ar_posix_acl; /* Enable posix acls */
411 int ar_quota; /* off/account/on */
412 int ar_suiddir; /* suiddir support */
413 int ar_data; /* ordered/writeback */
414 };
415
416 struct gfs2_tune {
417 spinlock_t gt_spin;
418
419 unsigned int gt_ilimit;
420 unsigned int gt_ilimit_tries;
421 unsigned int gt_ilimit_min;
422 unsigned int gt_demote_secs; /* Cache retention for unheld glock */
423 unsigned int gt_incore_log_blocks;
424 unsigned int gt_log_flush_secs;
425 unsigned int gt_jindex_refresh_secs; /* Check for new journal index */
426
427 unsigned int gt_scand_secs;
428 unsigned int gt_recoverd_secs;
429 unsigned int gt_logd_secs;
430 unsigned int gt_quotad_secs;
431
432 unsigned int gt_quota_simul_sync; /* Max quotavals to sync at once */
433 unsigned int gt_quota_warn_period; /* Secs between quota warn msgs */
434 unsigned int gt_quota_scale_num; /* Numerator */
435 unsigned int gt_quota_scale_den; /* Denominator */
436 unsigned int gt_quota_cache_secs;
437 unsigned int gt_quota_quantum; /* Secs between syncs to quota file */
438 unsigned int gt_atime_quantum; /* Min secs between atime updates */
439 unsigned int gt_new_files_jdata;
440 unsigned int gt_new_files_directio;
441 unsigned int gt_max_atomic_write; /* Split big writes into this size */
442 unsigned int gt_max_readahead; /* Max bytes to read-ahead from disk */
443 unsigned int gt_lockdump_size;
444 unsigned int gt_stall_secs; /* Detects trouble! */
445 unsigned int gt_complain_secs;
446 unsigned int gt_reclaim_limit; /* Max num of glocks in reclaim list */
447 unsigned int gt_entries_per_readdir;
448 unsigned int gt_prefetch_secs; /* Usage window for prefetched glocks */
449 unsigned int gt_greedy_default;
450 unsigned int gt_greedy_quantum;
451 unsigned int gt_greedy_max;
452 unsigned int gt_statfs_quantum;
453 unsigned int gt_statfs_slow;
454 };
455
456 struct gfs2_gl_hash_bucket {
457 rwlock_t hb_lock;
458 struct list_head hb_list;
459 };
460
461 enum {
462 SDF_JOURNAL_CHECKED = 0,
463 SDF_JOURNAL_LIVE = 1,
464 SDF_SHUTDOWN = 2,
465 SDF_NOATIME = 3,
466 };
467
468 #define GFS2_GL_HASH_SHIFT 13
469 #define GFS2_GL_HASH_SIZE (1 << GFS2_GL_HASH_SHIFT)
470 #define GFS2_GL_HASH_MASK (GFS2_GL_HASH_SIZE - 1)
471 #define GFS2_FSNAME_LEN 256
472
473 struct gfs2_sbd {
474 struct super_block *sd_vfs;
475 struct kobject sd_kobj;
476 unsigned long sd_flags; /* SDF_... */
477 struct gfs2_sb sd_sb;
478
479 /* Constants computed on mount */
480
481 uint32_t sd_fsb2bb;
482 uint32_t sd_fsb2bb_shift;
483 uint32_t sd_diptrs; /* Number of pointers in a dinode */
484 uint32_t sd_inptrs; /* Number of pointers in a indirect block */
485 uint32_t sd_jbsize; /* Size of a journaled data block */
486 uint32_t sd_hash_bsize; /* sizeof(exhash block) */
487 uint32_t sd_hash_bsize_shift;
488 uint32_t sd_hash_ptrs; /* Number of pointers in a hash block */
489 uint32_t sd_qc_per_block;
490 uint32_t sd_max_dirres; /* Max blocks needed to add a directory entry */
491 uint32_t sd_max_height; /* Max height of a file's metadata tree */
492 uint64_t sd_heightsize[GFS2_MAX_META_HEIGHT];
493 uint32_t sd_max_jheight; /* Max height of journaled file's meta tree */
494 uint64_t sd_jheightsize[GFS2_MAX_META_HEIGHT];
495
496 struct gfs2_args sd_args; /* Mount arguments */
497 struct gfs2_tune sd_tune; /* Filesystem tuning structure */
498
499 /* Lock Stuff */
500
501 struct lm_lockstruct sd_lockstruct;
502 struct gfs2_gl_hash_bucket sd_gl_hash[GFS2_GL_HASH_SIZE];
503 struct list_head sd_reclaim_list;
504 spinlock_t sd_reclaim_lock;
505 wait_queue_head_t sd_reclaim_wq;
506 atomic_t sd_reclaim_count;
507 struct gfs2_holder sd_live_gh;
508 struct gfs2_glock *sd_rename_gl;
509 struct gfs2_glock *sd_trans_gl;
510 struct mutex sd_invalidate_inodes_mutex;
511
512 /* Inode Stuff */
513
514 struct inode *sd_master_dir;
515 struct inode *sd_jindex;
516 struct inode *sd_inum_inode;
517 struct inode *sd_statfs_inode;
518 struct inode *sd_ir_inode;
519 struct inode *sd_sc_inode;
520 struct inode *sd_qc_inode;
521 struct inode *sd_rindex;
522 struct inode *sd_quota_inode;
523
524 /* Inum stuff */
525
526 struct mutex sd_inum_mutex;
527
528 /* StatFS stuff */
529
530 spinlock_t sd_statfs_spin;
531 struct mutex sd_statfs_mutex;
532 struct gfs2_statfs_change sd_statfs_master;
533 struct gfs2_statfs_change sd_statfs_local;
534 unsigned long sd_statfs_sync_time;
535
536 /* Resource group stuff */
537
538 uint64_t sd_rindex_vn;
539 spinlock_t sd_rindex_spin;
540 struct mutex sd_rindex_mutex;
541 struct list_head sd_rindex_list;
542 struct list_head sd_rindex_mru_list;
543 struct list_head sd_rindex_recent_list;
544 struct gfs2_rgrpd *sd_rindex_forward;
545 unsigned int sd_rgrps;
546
547 /* Journal index stuff */
548
549 struct list_head sd_jindex_list;
550 spinlock_t sd_jindex_spin;
551 struct mutex sd_jindex_mutex;
552 unsigned int sd_journals;
553 unsigned long sd_jindex_refresh_time;
554
555 struct gfs2_jdesc *sd_jdesc;
556 struct gfs2_holder sd_journal_gh;
557 struct gfs2_holder sd_jinode_gh;
558
559 struct gfs2_holder sd_ir_gh;
560 struct gfs2_holder sd_sc_gh;
561 struct gfs2_holder sd_qc_gh;
562
563 /* Daemon stuff */
564
565 struct task_struct *sd_scand_process;
566 struct task_struct *sd_recoverd_process;
567 struct task_struct *sd_logd_process;
568 struct task_struct *sd_quotad_process;
569 struct task_struct *sd_glockd_process[GFS2_GLOCKD_MAX];
570 unsigned int sd_glockd_num;
571
572 /* Quota stuff */
573
574 struct list_head sd_quota_list;
575 atomic_t sd_quota_count;
576 spinlock_t sd_quota_spin;
577 struct mutex sd_quota_mutex;
578
579 unsigned int sd_quota_slots;
580 unsigned int sd_quota_chunks;
581 unsigned char **sd_quota_bitmap;
582
583 uint64_t sd_quota_sync_gen;
584 unsigned long sd_quota_sync_time;
585
586 /* Log stuff */
587
588 spinlock_t sd_log_lock;
589
590 unsigned int sd_log_blks_reserved;
591 unsigned int sd_log_commited_buf;
592 unsigned int sd_log_commited_revoke;
593
594 unsigned int sd_log_num_gl;
595 unsigned int sd_log_num_buf;
596 unsigned int sd_log_num_revoke;
597 unsigned int sd_log_num_rg;
598 unsigned int sd_log_num_databuf;
599 unsigned int sd_log_num_jdata;
600 unsigned int sd_log_num_hdrs;
601
602 struct list_head sd_log_le_gl;
603 struct list_head sd_log_le_buf;
604 struct list_head sd_log_le_revoke;
605 struct list_head sd_log_le_rg;
606 struct list_head sd_log_le_databuf;
607
608 unsigned int sd_log_blks_free;
609 struct mutex sd_log_reserve_mutex;
610
611 uint64_t sd_log_sequence;
612 unsigned int sd_log_head;
613 unsigned int sd_log_tail;
614 int sd_log_idle;
615
616 unsigned long sd_log_flush_time;
617 struct rw_semaphore sd_log_flush_lock;
618 struct list_head sd_log_flush_list;
619
620 unsigned int sd_log_flush_head;
621 uint64_t sd_log_flush_wrapped;
622
623 struct list_head sd_ail1_list;
624 struct list_head sd_ail2_list;
625 uint64_t sd_ail_sync_gen;
626
627 /* Replay stuff */
628
629 struct list_head sd_revoke_list;
630 unsigned int sd_replay_tail;
631
632 unsigned int sd_found_blocks;
633 unsigned int sd_found_revokes;
634 unsigned int sd_replayed_blocks;
635
636 /* For quiescing the filesystem */
637
638 struct gfs2_holder sd_freeze_gh;
639 struct mutex sd_freeze_lock;
640 unsigned int sd_freeze_count;
641
642 /* Counters */
643
644 atomic_t sd_glock_count;
645 atomic_t sd_glock_held_count;
646 atomic_t sd_inode_count;
647 atomic_t sd_reclaimed;
648
649 char sd_fsname[GFS2_FSNAME_LEN];
650 char sd_table_name[GFS2_FSNAME_LEN];
651 char sd_proto_name[GFS2_FSNAME_LEN];
652
653 /* Debugging crud */
654
655 unsigned long sd_last_warning;
656 };
657
658 #endif /* __INCORE_DOT_H__ */
659
This page took 0.226302 seconds and 6 git commands to generate.