GFS2: Optimise glock lru and end of life inodes
[deliverable/linux.git] / fs / gfs2 / incore.h
1 /*
2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
3 * Copyright (C) 2004-2008 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 version 2.
8 */
9
10 #ifndef __INCORE_DOT_H__
11 #define __INCORE_DOT_H__
12
13 #include <linux/fs.h>
14 #include <linux/kobject.h>
15 #include <linux/workqueue.h>
16 #include <linux/dlm.h>
17 #include <linux/buffer_head.h>
18 #include <linux/rcupdate.h>
19 #include <linux/rculist_bl.h>
20
21 #define DIO_WAIT 0x00000010
22 #define DIO_METADATA 0x00000020
23 #define DIO_ALL 0x00000100
24
25 struct gfs2_log_operations;
26 struct gfs2_log_element;
27 struct gfs2_holder;
28 struct gfs2_glock;
29 struct gfs2_quota_data;
30 struct gfs2_trans;
31 struct gfs2_ail;
32 struct gfs2_jdesc;
33 struct gfs2_sbd;
34 struct lm_lockops;
35
36 typedef void (*gfs2_glop_bh_t) (struct gfs2_glock *gl, unsigned int ret);
37
38 struct gfs2_log_header_host {
39 u64 lh_sequence; /* Sequence number of this transaction */
40 u32 lh_flags; /* GFS2_LOG_HEAD_... */
41 u32 lh_tail; /* Block number of log tail */
42 u32 lh_blkno;
43 u32 lh_hash;
44 };
45
46 /*
47 * Structure of operations that are associated with each
48 * type of element in the log.
49 */
50
51 struct gfs2_log_operations {
52 void (*lo_add) (struct gfs2_sbd *sdp, struct gfs2_log_element *le);
53 void (*lo_before_commit) (struct gfs2_sbd *sdp);
54 void (*lo_after_commit) (struct gfs2_sbd *sdp, struct gfs2_ail *ai);
55 void (*lo_before_scan) (struct gfs2_jdesc *jd,
56 struct gfs2_log_header_host *head, int pass);
57 int (*lo_scan_elements) (struct gfs2_jdesc *jd, unsigned int start,
58 struct gfs2_log_descriptor *ld, __be64 *ptr,
59 int pass);
60 void (*lo_after_scan) (struct gfs2_jdesc *jd, int error, int pass);
61 const char *lo_name;
62 };
63
64 struct gfs2_log_element {
65 struct list_head le_list;
66 const struct gfs2_log_operations *le_ops;
67 };
68
69 #define GBF_FULL 1
70
71 struct gfs2_bitmap {
72 struct buffer_head *bi_bh;
73 char *bi_clone;
74 unsigned long bi_flags;
75 u32 bi_offset;
76 u32 bi_start;
77 u32 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 gfs2_glock *rd_gl; /* Glock for this rgrp */
84 u64 rd_addr; /* grp block disk address */
85 u64 rd_data0; /* first data location */
86 u32 rd_length; /* length of rgrp header in fs blocks */
87 u32 rd_data; /* num of data blocks in rgrp */
88 u32 rd_bitbytes; /* number of bytes in data bitmaps */
89 u32 rd_free;
90 u32 rd_free_clone;
91 u32 rd_dinodes;
92 u64 rd_igeneration;
93 struct gfs2_bitmap *rd_bits;
94 struct mutex rd_mutex;
95 struct gfs2_log_element rd_le;
96 struct gfs2_sbd *rd_sbd;
97 unsigned int rd_bh_count;
98 u32 rd_last_alloc;
99 u32 rd_flags;
100 #define GFS2_RDF_CHECK 0x10000000 /* check for unlinked inodes */
101 #define GFS2_RDF_UPTODATE 0x20000000 /* rg is up to date */
102 #define GFS2_RDF_ERROR 0x40000000 /* error in rg */
103 #define GFS2_RDF_MASK 0xf0000000 /* mask for internal flags */
104 };
105
106 enum gfs2_state_bits {
107 BH_Pinned = BH_PrivateStart,
108 BH_Escaped = BH_PrivateStart + 1,
109 };
110
111 BUFFER_FNS(Pinned, pinned)
112 TAS_BUFFER_FNS(Pinned, pinned)
113 BUFFER_FNS(Escaped, escaped)
114 TAS_BUFFER_FNS(Escaped, escaped)
115
116 struct gfs2_bufdata {
117 struct buffer_head *bd_bh;
118 struct gfs2_glock *bd_gl;
119
120 union {
121 struct list_head list_tr;
122 u64 blkno;
123 } u;
124 #define bd_list_tr u.list_tr
125 #define bd_blkno u.blkno
126
127 struct gfs2_log_element bd_le;
128
129 struct gfs2_ail *bd_ail;
130 struct list_head bd_ail_st_list;
131 struct list_head bd_ail_gl_list;
132 };
133
134 /*
135 * Internally, we prefix things with gdlm_ and GDLM_ (for gfs-dlm) since a
136 * prefix of lock_dlm_ gets awkward.
137 */
138
139 #define GDLM_STRNAME_BYTES 25
140 #define GDLM_LVB_SIZE 32
141
142 enum {
143 DFL_BLOCK_LOCKS = 0,
144 };
145
146 struct lm_lockname {
147 u64 ln_number;
148 unsigned int ln_type;
149 };
150
151 #define lm_name_equal(name1, name2) \
152 (((name1)->ln_number == (name2)->ln_number) && \
153 ((name1)->ln_type == (name2)->ln_type))
154
155
156 struct gfs2_glock_operations {
157 void (*go_xmote_th) (struct gfs2_glock *gl);
158 int (*go_xmote_bh) (struct gfs2_glock *gl, struct gfs2_holder *gh);
159 void (*go_inval) (struct gfs2_glock *gl, int flags);
160 int (*go_demote_ok) (const struct gfs2_glock *gl);
161 int (*go_lock) (struct gfs2_holder *gh);
162 void (*go_unlock) (struct gfs2_holder *gh);
163 int (*go_dump)(struct seq_file *seq, const struct gfs2_glock *gl);
164 void (*go_callback) (struct gfs2_glock *gl);
165 const int go_type;
166 const unsigned long go_min_hold_time;
167 const unsigned long go_flags;
168 #define GLOF_ASPACE 1
169 };
170
171 enum {
172 /* States */
173 HIF_HOLDER = 6, /* Set for gh that "holds" the glock */
174 HIF_FIRST = 7,
175 HIF_WAIT = 10,
176 };
177
178 struct gfs2_holder {
179 struct list_head gh_list;
180
181 struct gfs2_glock *gh_gl;
182 struct pid *gh_owner_pid;
183 unsigned int gh_state;
184 unsigned gh_flags;
185
186 int gh_error;
187 unsigned long gh_iflags; /* HIF_... */
188 unsigned long gh_ip;
189 };
190
191 enum {
192 GLF_LOCK = 1,
193 GLF_DEMOTE = 3,
194 GLF_PENDING_DEMOTE = 4,
195 GLF_DEMOTE_IN_PROGRESS = 5,
196 GLF_DIRTY = 6,
197 GLF_LFLUSH = 7,
198 GLF_INVALIDATE_IN_PROGRESS = 8,
199 GLF_REPLY_PENDING = 9,
200 GLF_INITIAL = 10,
201 GLF_FROZEN = 11,
202 GLF_QUEUED = 12,
203 GLF_LRU = 13,
204 GLF_OBJECT = 14, /* Used only for tracing */
205 };
206
207 struct gfs2_glock {
208 struct hlist_bl_node gl_list;
209 unsigned long gl_flags; /* GLF_... */
210 struct lm_lockname gl_name;
211 atomic_t gl_ref;
212
213 spinlock_t gl_spin;
214
215 /* State fields protected by gl_spin */
216 unsigned int gl_state:2, /* Current state */
217 gl_target:2, /* Target state */
218 gl_demote_state:2, /* State requested by remote node */
219 gl_req:2, /* State in last dlm request */
220 gl_reply:8; /* Last reply from the dlm */
221
222 unsigned int gl_hash;
223 unsigned long gl_demote_time; /* time of first demote request */
224 struct list_head gl_holders;
225
226 const struct gfs2_glock_operations *gl_ops;
227 char gl_strname[GDLM_STRNAME_BYTES];
228 struct dlm_lksb gl_lksb;
229 char gl_lvb[32];
230 unsigned long gl_tchange;
231 void *gl_object;
232
233 struct list_head gl_lru;
234
235 struct gfs2_sbd *gl_sbd;
236
237 struct list_head gl_ail_list;
238 atomic_t gl_ail_count;
239 atomic_t gl_revokes;
240 struct delayed_work gl_work;
241 struct work_struct gl_delete;
242 struct rcu_head gl_rcu;
243 };
244
245 #define GFS2_MIN_LVB_SIZE 32 /* Min size of LVB that gfs2 supports */
246
247 struct gfs2_alloc {
248 /* Quota stuff */
249
250 struct gfs2_quota_data *al_qd[2*MAXQUOTAS];
251 struct gfs2_holder al_qd_ghs[2*MAXQUOTAS];
252 unsigned int al_qd_num;
253
254 u32 al_requested; /* Filled in by caller of gfs2_inplace_reserve() */
255 u32 al_alloced; /* Filled in by gfs2_alloc_*() */
256
257 /* Filled in by gfs2_inplace_reserve() */
258
259 unsigned int al_line;
260 char *al_file;
261 struct gfs2_holder al_ri_gh;
262 struct gfs2_holder al_rgd_gh;
263 struct gfs2_rgrpd *al_rgd;
264
265 };
266
267 enum {
268 GIF_INVALID = 0,
269 GIF_QD_LOCKED = 1,
270 GIF_SW_PAGED = 3,
271 };
272
273
274 struct gfs2_inode {
275 struct inode i_inode;
276 u64 i_no_addr;
277 u64 i_no_formal_ino;
278 u64 i_generation;
279 u64 i_eattr;
280 unsigned long i_flags; /* GIF_... */
281 struct gfs2_glock *i_gl; /* Move into i_gh? */
282 struct gfs2_holder i_iopen_gh;
283 struct gfs2_holder i_gh; /* for prepare/commit_write only */
284 struct gfs2_alloc *i_alloc;
285 u64 i_goal; /* goal block for allocations */
286 struct rw_semaphore i_rw_mutex;
287 struct list_head i_trunc_list;
288 u32 i_entries;
289 u32 i_diskflags;
290 u8 i_height;
291 u8 i_depth;
292 };
293
294 /*
295 * Since i_inode is the first element of struct gfs2_inode,
296 * this is effectively a cast.
297 */
298 static inline struct gfs2_inode *GFS2_I(struct inode *inode)
299 {
300 return container_of(inode, struct gfs2_inode, i_inode);
301 }
302
303 static inline struct gfs2_sbd *GFS2_SB(const struct inode *inode)
304 {
305 return inode->i_sb->s_fs_info;
306 }
307
308 struct gfs2_file {
309 struct mutex f_fl_mutex;
310 struct gfs2_holder f_fl_gh;
311 };
312
313 struct gfs2_revoke_replay {
314 struct list_head rr_list;
315 u64 rr_blkno;
316 unsigned int rr_where;
317 };
318
319 enum {
320 QDF_USER = 0,
321 QDF_CHANGE = 1,
322 QDF_LOCKED = 2,
323 QDF_REFRESH = 3,
324 };
325
326 struct gfs2_quota_data {
327 struct list_head qd_list;
328 struct list_head qd_reclaim;
329
330 atomic_t qd_count;
331
332 u32 qd_id;
333 unsigned long qd_flags; /* QDF_... */
334
335 s64 qd_change;
336 s64 qd_change_sync;
337
338 unsigned int qd_slot;
339 unsigned int qd_slot_count;
340
341 struct buffer_head *qd_bh;
342 struct gfs2_quota_change *qd_bh_qc;
343 unsigned int qd_bh_count;
344
345 struct gfs2_glock *qd_gl;
346 struct gfs2_quota_lvb qd_qb;
347
348 u64 qd_sync_gen;
349 unsigned long qd_last_warn;
350 };
351
352 struct gfs2_trans {
353 unsigned long tr_ip;
354
355 unsigned int tr_blocks;
356 unsigned int tr_revokes;
357 unsigned int tr_reserved;
358
359 struct gfs2_holder tr_t_gh;
360
361 int tr_touched;
362
363 unsigned int tr_num_buf;
364 unsigned int tr_num_buf_new;
365 unsigned int tr_num_databuf_new;
366 unsigned int tr_num_buf_rm;
367 unsigned int tr_num_databuf_rm;
368 struct list_head tr_list_buf;
369
370 unsigned int tr_num_revoke;
371 unsigned int tr_num_revoke_rm;
372 };
373
374 struct gfs2_ail {
375 struct list_head ai_list;
376
377 unsigned int ai_first;
378 struct list_head ai_ail1_list;
379 struct list_head ai_ail2_list;
380
381 u64 ai_sync_gen;
382 };
383
384 struct gfs2_journal_extent {
385 struct list_head extent_list;
386
387 unsigned int lblock; /* First logical block */
388 u64 dblock; /* First disk block */
389 u64 blocks;
390 };
391
392 struct gfs2_jdesc {
393 struct list_head jd_list;
394 struct list_head extent_list;
395 struct work_struct jd_work;
396 struct inode *jd_inode;
397 unsigned long jd_flags;
398 #define JDF_RECOVERY 1
399 unsigned int jd_jid;
400 unsigned int jd_blocks;
401 };
402
403 struct gfs2_statfs_change_host {
404 s64 sc_total;
405 s64 sc_free;
406 s64 sc_dinodes;
407 };
408
409 #define GFS2_QUOTA_DEFAULT GFS2_QUOTA_OFF
410 #define GFS2_QUOTA_OFF 0
411 #define GFS2_QUOTA_ACCOUNT 1
412 #define GFS2_QUOTA_ON 2
413
414 #define GFS2_DATA_DEFAULT GFS2_DATA_ORDERED
415 #define GFS2_DATA_WRITEBACK 1
416 #define GFS2_DATA_ORDERED 2
417
418 #define GFS2_ERRORS_DEFAULT GFS2_ERRORS_WITHDRAW
419 #define GFS2_ERRORS_WITHDRAW 0
420 #define GFS2_ERRORS_CONTINUE 1 /* place holder for future feature */
421 #define GFS2_ERRORS_RO 2 /* place holder for future feature */
422 #define GFS2_ERRORS_PANIC 3
423
424 struct gfs2_args {
425 char ar_lockproto[GFS2_LOCKNAME_LEN]; /* Name of the Lock Protocol */
426 char ar_locktable[GFS2_LOCKNAME_LEN]; /* Name of the Lock Table */
427 char ar_hostdata[GFS2_LOCKNAME_LEN]; /* Host specific data */
428 unsigned int ar_spectator:1; /* Don't get a journal */
429 unsigned int ar_localflocks:1; /* Let the VFS do flock|fcntl */
430 unsigned int ar_debug:1; /* Oops on errors */
431 unsigned int ar_posix_acl:1; /* Enable posix acls */
432 unsigned int ar_quota:2; /* off/account/on */
433 unsigned int ar_suiddir:1; /* suiddir support */
434 unsigned int ar_data:2; /* ordered/writeback */
435 unsigned int ar_meta:1; /* mount metafs */
436 unsigned int ar_discard:1; /* discard requests */
437 unsigned int ar_errors:2; /* errors=withdraw | panic */
438 unsigned int ar_nobarrier:1; /* do not send barriers */
439 int ar_commit; /* Commit interval */
440 int ar_statfs_quantum; /* The fast statfs interval */
441 int ar_quota_quantum; /* The quota interval */
442 int ar_statfs_percent; /* The % change to force sync */
443 };
444
445 struct gfs2_tune {
446 spinlock_t gt_spin;
447
448 unsigned int gt_logd_secs;
449
450 unsigned int gt_quota_simul_sync; /* Max quotavals to sync at once */
451 unsigned int gt_quota_warn_period; /* Secs between quota warn msgs */
452 unsigned int gt_quota_scale_num; /* Numerator */
453 unsigned int gt_quota_scale_den; /* Denominator */
454 unsigned int gt_quota_quantum; /* Secs between syncs to quota file */
455 unsigned int gt_new_files_jdata;
456 unsigned int gt_max_readahead; /* Max bytes to read-ahead from disk */
457 unsigned int gt_complain_secs;
458 unsigned int gt_statfs_quantum;
459 unsigned int gt_statfs_slow;
460 };
461
462 enum {
463 SDF_JOURNAL_CHECKED = 0,
464 SDF_JOURNAL_LIVE = 1,
465 SDF_SHUTDOWN = 2,
466 SDF_NOBARRIERS = 3,
467 SDF_NORECOVERY = 4,
468 SDF_DEMOTE = 5,
469 SDF_NOJOURNALID = 6,
470 };
471
472 #define GFS2_FSNAME_LEN 256
473
474 struct gfs2_inum_host {
475 u64 no_formal_ino;
476 u64 no_addr;
477 };
478
479 struct gfs2_sb_host {
480 u32 sb_magic;
481 u32 sb_type;
482 u32 sb_format;
483
484 u32 sb_fs_format;
485 u32 sb_multihost_format;
486 u32 sb_bsize;
487 u32 sb_bsize_shift;
488
489 struct gfs2_inum_host sb_master_dir;
490 struct gfs2_inum_host sb_root_dir;
491
492 char sb_lockproto[GFS2_LOCKNAME_LEN];
493 char sb_locktable[GFS2_LOCKNAME_LEN];
494 u8 sb_uuid[16];
495 };
496
497 /*
498 * lm_mount() return values
499 *
500 * ls_jid - the journal ID this node should use
501 * ls_first - this node is the first to mount the file system
502 * ls_lockspace - lock module's context for this file system
503 * ls_ops - lock module's functions
504 */
505
506 struct lm_lockstruct {
507 int ls_jid;
508 unsigned int ls_first;
509 unsigned int ls_first_done;
510 unsigned int ls_nodir;
511 const struct lm_lockops *ls_ops;
512 unsigned long ls_flags;
513 dlm_lockspace_t *ls_dlm;
514
515 int ls_recover_jid_done;
516 int ls_recover_jid_status;
517 };
518
519 struct gfs2_sbd {
520 struct super_block *sd_vfs;
521 struct kobject sd_kobj;
522 unsigned long sd_flags; /* SDF_... */
523 struct gfs2_sb_host sd_sb;
524
525 /* Constants computed on mount */
526
527 u32 sd_fsb2bb;
528 u32 sd_fsb2bb_shift;
529 u32 sd_diptrs; /* Number of pointers in a dinode */
530 u32 sd_inptrs; /* Number of pointers in a indirect block */
531 u32 sd_jbsize; /* Size of a journaled data block */
532 u32 sd_hash_bsize; /* sizeof(exhash block) */
533 u32 sd_hash_bsize_shift;
534 u32 sd_hash_ptrs; /* Number of pointers in a hash block */
535 u32 sd_qc_per_block;
536 u32 sd_max_dirres; /* Max blocks needed to add a directory entry */
537 u32 sd_max_height; /* Max height of a file's metadata tree */
538 u64 sd_heightsize[GFS2_MAX_META_HEIGHT + 1];
539 u32 sd_max_jheight; /* Max height of journaled file's meta tree */
540 u64 sd_jheightsize[GFS2_MAX_META_HEIGHT + 1];
541
542 struct gfs2_args sd_args; /* Mount arguments */
543 struct gfs2_tune sd_tune; /* Filesystem tuning structure */
544
545 /* Lock Stuff */
546
547 struct lm_lockstruct sd_lockstruct;
548 struct gfs2_holder sd_live_gh;
549 struct gfs2_glock *sd_rename_gl;
550 struct gfs2_glock *sd_trans_gl;
551 wait_queue_head_t sd_glock_wait;
552 atomic_t sd_glock_disposal;
553
554 /* Inode Stuff */
555
556 struct dentry *sd_master_dir;
557 struct dentry *sd_root_dir;
558
559 struct inode *sd_jindex;
560 struct inode *sd_statfs_inode;
561 struct inode *sd_sc_inode;
562 struct inode *sd_qc_inode;
563 struct inode *sd_rindex;
564 struct inode *sd_quota_inode;
565
566 /* StatFS stuff */
567
568 spinlock_t sd_statfs_spin;
569 struct gfs2_statfs_change_host sd_statfs_master;
570 struct gfs2_statfs_change_host sd_statfs_local;
571 int sd_statfs_force_sync;
572
573 /* Resource group stuff */
574
575 int sd_rindex_uptodate;
576 spinlock_t sd_rindex_spin;
577 struct mutex sd_rindex_mutex;
578 struct list_head sd_rindex_list;
579 struct list_head sd_rindex_mru_list;
580 struct gfs2_rgrpd *sd_rindex_forward;
581 unsigned int sd_rgrps;
582 unsigned int sd_max_rg_data;
583
584 /* Journal index stuff */
585
586 struct list_head sd_jindex_list;
587 spinlock_t sd_jindex_spin;
588 struct mutex sd_jindex_mutex;
589 unsigned int sd_journals;
590
591 struct gfs2_jdesc *sd_jdesc;
592 struct gfs2_holder sd_journal_gh;
593 struct gfs2_holder sd_jinode_gh;
594
595 struct gfs2_holder sd_sc_gh;
596 struct gfs2_holder sd_qc_gh;
597
598 /* Daemon stuff */
599
600 struct task_struct *sd_logd_process;
601 struct task_struct *sd_quotad_process;
602
603 /* Quota stuff */
604
605 struct list_head sd_quota_list;
606 atomic_t sd_quota_count;
607 struct mutex sd_quota_mutex;
608 wait_queue_head_t sd_quota_wait;
609 struct list_head sd_trunc_list;
610 spinlock_t sd_trunc_lock;
611
612 unsigned int sd_quota_slots;
613 unsigned int sd_quota_chunks;
614 unsigned char **sd_quota_bitmap;
615
616 u64 sd_quota_sync_gen;
617
618 /* Log stuff */
619
620 spinlock_t sd_log_lock;
621
622 unsigned int sd_log_blks_reserved;
623 unsigned int sd_log_commited_buf;
624 unsigned int sd_log_commited_databuf;
625 int sd_log_commited_revoke;
626
627 atomic_t sd_log_pinned;
628 unsigned int sd_log_num_buf;
629 unsigned int sd_log_num_revoke;
630 unsigned int sd_log_num_rg;
631 unsigned int sd_log_num_databuf;
632
633 struct list_head sd_log_le_buf;
634 struct list_head sd_log_le_revoke;
635 struct list_head sd_log_le_rg;
636 struct list_head sd_log_le_databuf;
637 struct list_head sd_log_le_ordered;
638
639 atomic_t sd_log_thresh1;
640 atomic_t sd_log_thresh2;
641 atomic_t sd_log_blks_free;
642 wait_queue_head_t sd_log_waitq;
643 wait_queue_head_t sd_logd_waitq;
644
645 u64 sd_log_sequence;
646 unsigned int sd_log_head;
647 unsigned int sd_log_tail;
648 int sd_log_idle;
649
650 struct rw_semaphore sd_log_flush_lock;
651 atomic_t sd_log_in_flight;
652 wait_queue_head_t sd_log_flush_wait;
653
654 unsigned int sd_log_flush_head;
655 u64 sd_log_flush_wrapped;
656
657 spinlock_t sd_ail_lock;
658 struct list_head sd_ail1_list;
659 struct list_head sd_ail2_list;
660 u64 sd_ail_sync_gen;
661
662 /* Replay stuff */
663
664 struct list_head sd_revoke_list;
665 unsigned int sd_replay_tail;
666
667 unsigned int sd_found_blocks;
668 unsigned int sd_found_revokes;
669 unsigned int sd_replayed_blocks;
670
671 /* For quiescing the filesystem */
672
673 struct gfs2_holder sd_freeze_gh;
674 struct mutex sd_freeze_lock;
675 unsigned int sd_freeze_count;
676
677 char sd_fsname[GFS2_FSNAME_LEN];
678 char sd_table_name[GFS2_FSNAME_LEN];
679 char sd_proto_name[GFS2_FSNAME_LEN];
680
681 /* Debugging crud */
682
683 unsigned long sd_last_warning;
684 struct dentry *debugfs_dir; /* debugfs directory */
685 struct dentry *debugfs_dentry_glocks; /* for debugfs */
686 };
687
688 #endif /* __INCORE_DOT_H__ */
689
This page took 0.046007 seconds and 5 git commands to generate.