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 };
204
205 struct gfs2_glock {
206 struct hlist_bl_node gl_list;
207 unsigned long gl_flags; /* GLF_... */
208 struct lm_lockname gl_name;
209 atomic_t gl_ref;
210
211 spinlock_t gl_spin;
212
213 /* State fields protected by gl_spin */
214 unsigned int gl_state:2, /* Current state */
215 gl_target:2, /* Target state */
216 gl_demote_state:2, /* State requested by remote node */
217 gl_req:2, /* State in last dlm request */
218 gl_reply:8; /* Last reply from the dlm */
219
220 unsigned int gl_hash;
221 unsigned long gl_demote_time; /* time of first demote request */
222 struct list_head gl_holders;
223
224 const struct gfs2_glock_operations *gl_ops;
225 char gl_strname[GDLM_STRNAME_BYTES];
226 struct dlm_lksb gl_lksb;
227 char gl_lvb[32];
228 unsigned long gl_tchange;
229 void *gl_object;
230
231 struct list_head gl_lru;
232
233 struct gfs2_sbd *gl_sbd;
234
235 struct list_head gl_ail_list;
236 atomic_t gl_ail_count;
237 struct delayed_work gl_work;
238 struct work_struct gl_delete;
239 struct rcu_head gl_rcu;
240 };
241
242 #define GFS2_MIN_LVB_SIZE 32 /* Min size of LVB that gfs2 supports */
243
244 struct gfs2_alloc {
245 /* Quota stuff */
246
247 struct gfs2_quota_data *al_qd[2*MAXQUOTAS];
248 struct gfs2_holder al_qd_ghs[2*MAXQUOTAS];
249 unsigned int al_qd_num;
250
251 u32 al_requested; /* Filled in by caller of gfs2_inplace_reserve() */
252 u32 al_alloced; /* Filled in by gfs2_alloc_*() */
253
254 /* Filled in by gfs2_inplace_reserve() */
255
256 unsigned int al_line;
257 char *al_file;
258 struct gfs2_holder al_ri_gh;
259 struct gfs2_holder al_rgd_gh;
260 struct gfs2_rgrpd *al_rgd;
261
262 };
263
264 enum {
265 GIF_INVALID = 0,
266 GIF_QD_LOCKED = 1,
267 GIF_SW_PAGED = 3,
268 };
269
270
271 struct gfs2_inode {
272 struct inode i_inode;
273 u64 i_no_addr;
274 u64 i_no_formal_ino;
275 u64 i_generation;
276 u64 i_eattr;
277 unsigned long i_flags; /* GIF_... */
278 struct gfs2_glock *i_gl; /* Move into i_gh? */
279 struct gfs2_holder i_iopen_gh;
280 struct gfs2_holder i_gh; /* for prepare/commit_write only */
281 struct gfs2_alloc *i_alloc;
282 u64 i_goal; /* goal block for allocations */
283 struct rw_semaphore i_rw_mutex;
284 struct list_head i_trunc_list;
285 u32 i_entries;
286 u32 i_diskflags;
287 u8 i_height;
288 u8 i_depth;
289 };
290
291 /*
292 * Since i_inode is the first element of struct gfs2_inode,
293 * this is effectively a cast.
294 */
GFS2_I(struct inode * inode)295 static inline struct gfs2_inode *GFS2_I(struct inode *inode)
296 {
297 return container_of(inode, struct gfs2_inode, i_inode);
298 }
299
GFS2_SB(const struct inode * inode)300 static inline struct gfs2_sbd *GFS2_SB(const struct inode *inode)
301 {
302 return inode->i_sb->s_fs_info;
303 }
304
305 struct gfs2_file {
306 struct mutex f_fl_mutex;
307 struct gfs2_holder f_fl_gh;
308 };
309
310 struct gfs2_revoke_replay {
311 struct list_head rr_list;
312 u64 rr_blkno;
313 unsigned int rr_where;
314 };
315
316 enum {
317 QDF_USER = 0,
318 QDF_CHANGE = 1,
319 QDF_LOCKED = 2,
320 QDF_REFRESH = 3,
321 };
322
323 struct gfs2_quota_data {
324 struct list_head qd_list;
325 struct list_head qd_reclaim;
326
327 atomic_t qd_count;
328
329 u32 qd_id;
330 unsigned long qd_flags; /* QDF_... */
331
332 s64 qd_change;
333 s64 qd_change_sync;
334
335 unsigned int qd_slot;
336 unsigned int qd_slot_count;
337
338 struct buffer_head *qd_bh;
339 struct gfs2_quota_change *qd_bh_qc;
340 unsigned int qd_bh_count;
341
342 struct gfs2_glock *qd_gl;
343 struct gfs2_quota_lvb qd_qb;
344
345 u64 qd_sync_gen;
346 unsigned long qd_last_warn;
347 };
348
349 struct gfs2_trans {
350 unsigned long tr_ip;
351
352 unsigned int tr_blocks;
353 unsigned int tr_revokes;
354 unsigned int tr_reserved;
355
356 struct gfs2_holder tr_t_gh;
357
358 int tr_touched;
359
360 unsigned int tr_num_buf;
361 unsigned int tr_num_buf_new;
362 unsigned int tr_num_databuf_new;
363 unsigned int tr_num_buf_rm;
364 unsigned int tr_num_databuf_rm;
365 struct list_head tr_list_buf;
366
367 unsigned int tr_num_revoke;
368 unsigned int tr_num_revoke_rm;
369 };
370
371 struct gfs2_ail {
372 struct list_head ai_list;
373
374 unsigned int ai_first;
375 struct list_head ai_ail1_list;
376 struct list_head ai_ail2_list;
377
378 u64 ai_sync_gen;
379 };
380
381 struct gfs2_journal_extent {
382 struct list_head extent_list;
383
384 unsigned int lblock; /* First logical block */
385 u64 dblock; /* First disk block */
386 u64 blocks;
387 };
388
389 struct gfs2_jdesc {
390 struct list_head jd_list;
391 struct list_head extent_list;
392 struct work_struct jd_work;
393 struct inode *jd_inode;
394 unsigned long jd_flags;
395 #define JDF_RECOVERY 1
396 unsigned int jd_jid;
397 unsigned int jd_blocks;
398 };
399
400 struct gfs2_statfs_change_host {
401 s64 sc_total;
402 s64 sc_free;
403 s64 sc_dinodes;
404 };
405
406 #define GFS2_QUOTA_DEFAULT GFS2_QUOTA_OFF
407 #define GFS2_QUOTA_OFF 0
408 #define GFS2_QUOTA_ACCOUNT 1
409 #define GFS2_QUOTA_ON 2
410
411 #define GFS2_DATA_DEFAULT GFS2_DATA_ORDERED
412 #define GFS2_DATA_WRITEBACK 1
413 #define GFS2_DATA_ORDERED 2
414
415 #define GFS2_ERRORS_DEFAULT GFS2_ERRORS_WITHDRAW
416 #define GFS2_ERRORS_WITHDRAW 0
417 #define GFS2_ERRORS_CONTINUE 1 /* place holder for future feature */
418 #define GFS2_ERRORS_RO 2 /* place holder for future feature */
419 #define GFS2_ERRORS_PANIC 3
420
421 struct gfs2_args {
422 char ar_lockproto[GFS2_LOCKNAME_LEN]; /* Name of the Lock Protocol */
423 char ar_locktable[GFS2_LOCKNAME_LEN]; /* Name of the Lock Table */
424 char ar_hostdata[GFS2_LOCKNAME_LEN]; /* Host specific data */
425 unsigned int ar_spectator:1; /* Don't get a journal */
426 unsigned int ar_localflocks:1; /* Let the VFS do flock|fcntl */
427 unsigned int ar_debug:1; /* Oops on errors */
428 unsigned int ar_posix_acl:1; /* Enable posix acls */
429 unsigned int ar_quota:2; /* off/account/on */
430 unsigned int ar_suiddir:1; /* suiddir support */
431 unsigned int ar_data:2; /* ordered/writeback */
432 unsigned int ar_meta:1; /* mount metafs */
433 unsigned int ar_discard:1; /* discard requests */
434 unsigned int ar_errors:2; /* errors=withdraw | panic */
435 unsigned int ar_nobarrier:1; /* do not send barriers */
436 int ar_commit; /* Commit interval */
437 int ar_statfs_quantum; /* The fast statfs interval */
438 int ar_quota_quantum; /* The quota interval */
439 int ar_statfs_percent; /* The % change to force sync */
440 };
441
442 struct gfs2_tune {
443 spinlock_t gt_spin;
444
445 unsigned int gt_logd_secs;
446
447 unsigned int gt_quota_simul_sync; /* Max quotavals to sync at once */
448 unsigned int gt_quota_warn_period; /* Secs between quota warn msgs */
449 unsigned int gt_quota_scale_num; /* Numerator */
450 unsigned int gt_quota_scale_den; /* Denominator */
451 unsigned int gt_quota_quantum; /* Secs between syncs to quota file */
452 unsigned int gt_new_files_jdata;
453 unsigned int gt_max_readahead; /* Max bytes to read-ahead from disk */
454 unsigned int gt_complain_secs;
455 unsigned int gt_statfs_quantum;
456 unsigned int gt_statfs_slow;
457 };
458
459 enum {
460 SDF_JOURNAL_CHECKED = 0,
461 SDF_JOURNAL_LIVE = 1,
462 SDF_SHUTDOWN = 2,
463 SDF_NOBARRIERS = 3,
464 SDF_NORECOVERY = 4,
465 SDF_DEMOTE = 5,
466 SDF_NOJOURNALID = 6,
467 };
468
469 #define GFS2_FSNAME_LEN 256
470
471 struct gfs2_inum_host {
472 u64 no_formal_ino;
473 u64 no_addr;
474 };
475
476 struct gfs2_sb_host {
477 u32 sb_magic;
478 u32 sb_type;
479 u32 sb_format;
480
481 u32 sb_fs_format;
482 u32 sb_multihost_format;
483 u32 sb_bsize;
484 u32 sb_bsize_shift;
485
486 struct gfs2_inum_host sb_master_dir;
487 struct gfs2_inum_host sb_root_dir;
488
489 char sb_lockproto[GFS2_LOCKNAME_LEN];
490 char sb_locktable[GFS2_LOCKNAME_LEN];
491 u8 sb_uuid[16];
492 };
493
494 /*
495 * lm_mount() return values
496 *
497 * ls_jid - the journal ID this node should use
498 * ls_first - this node is the first to mount the file system
499 * ls_lockspace - lock module's context for this file system
500 * ls_ops - lock module's functions
501 */
502
503 struct lm_lockstruct {
504 int ls_jid;
505 unsigned int ls_first;
506 unsigned int ls_first_done;
507 unsigned int ls_nodir;
508 const struct lm_lockops *ls_ops;
509 unsigned long ls_flags;
510 dlm_lockspace_t *ls_dlm;
511
512 int ls_recover_jid_done;
513 int ls_recover_jid_status;
514 };
515
516 struct gfs2_sbd {
517 struct super_block *sd_vfs;
518 struct kobject sd_kobj;
519 unsigned long sd_flags; /* SDF_... */
520 struct gfs2_sb_host sd_sb;
521
522 /* Constants computed on mount */
523
524 u32 sd_fsb2bb;
525 u32 sd_fsb2bb_shift;
526 u32 sd_diptrs; /* Number of pointers in a dinode */
527 u32 sd_inptrs; /* Number of pointers in a indirect block */
528 u32 sd_jbsize; /* Size of a journaled data block */
529 u32 sd_hash_bsize; /* sizeof(exhash block) */
530 u32 sd_hash_bsize_shift;
531 u32 sd_hash_ptrs; /* Number of pointers in a hash block */
532 u32 sd_qc_per_block;
533 u32 sd_max_dirres; /* Max blocks needed to add a directory entry */
534 u32 sd_max_height; /* Max height of a file's metadata tree */
535 u64 sd_heightsize[GFS2_MAX_META_HEIGHT + 1];
536 u32 sd_max_jheight; /* Max height of journaled file's meta tree */
537 u64 sd_jheightsize[GFS2_MAX_META_HEIGHT + 1];
538
539 struct gfs2_args sd_args; /* Mount arguments */
540 struct gfs2_tune sd_tune; /* Filesystem tuning structure */
541
542 /* Lock Stuff */
543
544 struct lm_lockstruct sd_lockstruct;
545 struct gfs2_holder sd_live_gh;
546 struct gfs2_glock *sd_rename_gl;
547 struct gfs2_glock *sd_trans_gl;
548 wait_queue_head_t sd_glock_wait;
549 atomic_t sd_glock_disposal;
550
551 /* Inode Stuff */
552
553 struct dentry *sd_master_dir;
554 struct dentry *sd_root_dir;
555
556 struct inode *sd_jindex;
557 struct inode *sd_statfs_inode;
558 struct inode *sd_sc_inode;
559 struct inode *sd_qc_inode;
560 struct inode *sd_rindex;
561 struct inode *sd_quota_inode;
562
563 /* StatFS stuff */
564
565 spinlock_t sd_statfs_spin;
566 struct gfs2_statfs_change_host sd_statfs_master;
567 struct gfs2_statfs_change_host sd_statfs_local;
568 int sd_statfs_force_sync;
569
570 /* Resource group stuff */
571
572 int sd_rindex_uptodate;
573 spinlock_t sd_rindex_spin;
574 struct mutex sd_rindex_mutex;
575 struct list_head sd_rindex_list;
576 struct list_head sd_rindex_mru_list;
577 struct gfs2_rgrpd *sd_rindex_forward;
578 unsigned int sd_rgrps;
579 unsigned int sd_max_rg_data;
580
581 /* Journal index stuff */
582
583 struct list_head sd_jindex_list;
584 spinlock_t sd_jindex_spin;
585 struct mutex sd_jindex_mutex;
586 unsigned int sd_journals;
587
588 struct gfs2_jdesc *sd_jdesc;
589 struct gfs2_holder sd_journal_gh;
590 struct gfs2_holder sd_jinode_gh;
591
592 struct gfs2_holder sd_sc_gh;
593 struct gfs2_holder sd_qc_gh;
594
595 /* Daemon stuff */
596
597 struct task_struct *sd_logd_process;
598 struct task_struct *sd_quotad_process;
599
600 /* Quota stuff */
601
602 struct list_head sd_quota_list;
603 atomic_t sd_quota_count;
604 struct mutex sd_quota_mutex;
605 wait_queue_head_t sd_quota_wait;
606 struct list_head sd_trunc_list;
607 spinlock_t sd_trunc_lock;
608
609 unsigned int sd_quota_slots;
610 unsigned int sd_quota_chunks;
611 unsigned char **sd_quota_bitmap;
612
613 u64 sd_quota_sync_gen;
614
615 /* Log stuff */
616
617 spinlock_t sd_log_lock;
618
619 unsigned int sd_log_blks_reserved;
620 unsigned int sd_log_commited_buf;
621 unsigned int sd_log_commited_databuf;
622 int sd_log_commited_revoke;
623
624 atomic_t sd_log_pinned;
625 unsigned int sd_log_num_buf;
626 unsigned int sd_log_num_revoke;
627 unsigned int sd_log_num_rg;
628 unsigned int sd_log_num_databuf;
629
630 struct list_head sd_log_le_buf;
631 struct list_head sd_log_le_revoke;
632 struct list_head sd_log_le_rg;
633 struct list_head sd_log_le_databuf;
634 struct list_head sd_log_le_ordered;
635
636 atomic_t sd_log_thresh1;
637 atomic_t sd_log_thresh2;
638 atomic_t sd_log_blks_free;
639 wait_queue_head_t sd_log_waitq;
640 wait_queue_head_t sd_logd_waitq;
641
642 u64 sd_log_sequence;
643 unsigned int sd_log_head;
644 unsigned int sd_log_tail;
645 int sd_log_idle;
646
647 struct rw_semaphore sd_log_flush_lock;
648 atomic_t sd_log_in_flight;
649 wait_queue_head_t sd_log_flush_wait;
650
651 unsigned int sd_log_flush_head;
652 u64 sd_log_flush_wrapped;
653
654 spinlock_t sd_ail_lock;
655 struct list_head sd_ail1_list;
656 struct list_head sd_ail2_list;
657 u64 sd_ail_sync_gen;
658
659 /* Replay stuff */
660
661 struct list_head sd_revoke_list;
662 unsigned int sd_replay_tail;
663
664 unsigned int sd_found_blocks;
665 unsigned int sd_found_revokes;
666 unsigned int sd_replayed_blocks;
667
668 /* For quiescing the filesystem */
669
670 struct gfs2_holder sd_freeze_gh;
671 struct mutex sd_freeze_lock;
672 unsigned int sd_freeze_count;
673
674 char sd_fsname[GFS2_FSNAME_LEN];
675 char sd_table_name[GFS2_FSNAME_LEN];
676 char sd_proto_name[GFS2_FSNAME_LEN];
677
678 /* Debugging crud */
679
680 unsigned long sd_last_warning;
681 struct dentry *debugfs_dir; /* debugfs directory */
682 struct dentry *debugfs_dentry_glocks; /* for debugfs */
683 };
684
685 #endif /* __INCORE_DOT_H__ */
686
687