Loading...
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#include <linux/completion.h>
21#include <linux/rbtree.h>
22#include <linux/ktime.h>
23#include <linux/percpu.h>
24#include <linux/lockref.h>
25#include <linux/rhashtable.h>
26
27#define DIO_WAIT 0x00000010
28#define DIO_METADATA 0x00000020
29
30struct gfs2_log_operations;
31struct gfs2_bufdata;
32struct gfs2_holder;
33struct gfs2_glock;
34struct gfs2_quota_data;
35struct gfs2_trans;
36struct gfs2_jdesc;
37struct gfs2_sbd;
38struct lm_lockops;
39
40typedef void (*gfs2_glop_bh_t) (struct gfs2_glock *gl, unsigned int ret);
41
42struct gfs2_log_header_host {
43 u64 lh_sequence; /* Sequence number of this transaction */
44 u32 lh_flags; /* GFS2_LOG_HEAD_... */
45 u32 lh_tail; /* Block number of log tail */
46 u32 lh_blkno;
47};
48
49/*
50 * Structure of operations that are associated with each
51 * type of element in the log.
52 */
53
54struct gfs2_log_operations {
55 void (*lo_before_commit) (struct gfs2_sbd *sdp, struct gfs2_trans *tr);
56 void (*lo_after_commit) (struct gfs2_sbd *sdp, struct gfs2_trans *tr);
57 void (*lo_before_scan) (struct gfs2_jdesc *jd,
58 struct gfs2_log_header_host *head, int pass);
59 int (*lo_scan_elements) (struct gfs2_jdesc *jd, unsigned int start,
60 struct gfs2_log_descriptor *ld, __be64 *ptr,
61 int pass);
62 void (*lo_after_scan) (struct gfs2_jdesc *jd, int error, int pass);
63 const char *lo_name;
64};
65
66#define GBF_FULL 1
67
68struct gfs2_bitmap {
69 struct buffer_head *bi_bh;
70 char *bi_clone;
71 unsigned long bi_flags;
72 u32 bi_offset;
73 u32 bi_start;
74 u32 bi_len;
75 u32 bi_blocks;
76};
77
78struct gfs2_rgrpd {
79 struct rb_node rd_node; /* Link with superblock */
80 struct gfs2_glock *rd_gl; /* Glock for this rgrp */
81 u64 rd_addr; /* grp block disk address */
82 u64 rd_data0; /* first data location */
83 u32 rd_length; /* length of rgrp header in fs blocks */
84 u32 rd_data; /* num of data blocks in rgrp */
85 u32 rd_bitbytes; /* number of bytes in data bitmaps */
86 u32 rd_free;
87 u32 rd_reserved; /* number of blocks reserved */
88 u32 rd_free_clone;
89 u32 rd_dinodes;
90 u64 rd_igeneration;
91 struct gfs2_bitmap *rd_bits;
92 struct gfs2_sbd *rd_sbd;
93 struct gfs2_rgrp_lvb *rd_rgl;
94 u32 rd_last_alloc;
95 u32 rd_flags;
96 u32 rd_extfail_pt; /* extent failure point */
97#define GFS2_RDF_CHECK 0x10000000 /* check for unlinked inodes */
98#define GFS2_RDF_UPTODATE 0x20000000 /* rg is up to date */
99#define GFS2_RDF_ERROR 0x40000000 /* error in rg */
100#define GFS2_RDF_PREFERRED 0x80000000 /* This rgrp is preferred */
101#define GFS2_RDF_MASK 0xf0000000 /* mask for internal flags */
102 spinlock_t rd_rsspin; /* protects reservation related vars */
103 struct rb_root rd_rstree; /* multi-block reservation tree */
104};
105
106struct gfs2_rbm {
107 struct gfs2_rgrpd *rgd;
108 u32 offset; /* The offset is bitmap relative */
109 int bii; /* Bitmap index */
110};
111
112static inline struct gfs2_bitmap *rbm_bi(const struct gfs2_rbm *rbm)
113{
114 return rbm->rgd->rd_bits + rbm->bii;
115}
116
117static inline u64 gfs2_rbm_to_block(const struct gfs2_rbm *rbm)
118{
119 return rbm->rgd->rd_data0 + (rbm_bi(rbm)->bi_start * GFS2_NBBY) +
120 rbm->offset;
121}
122
123static inline bool gfs2_rbm_eq(const struct gfs2_rbm *rbm1,
124 const struct gfs2_rbm *rbm2)
125{
126 return (rbm1->rgd == rbm2->rgd) && (rbm1->bii == rbm2->bii) &&
127 (rbm1->offset == rbm2->offset);
128}
129
130enum gfs2_state_bits {
131 BH_Pinned = BH_PrivateStart,
132 BH_Escaped = BH_PrivateStart + 1,
133};
134
135BUFFER_FNS(Pinned, pinned)
136TAS_BUFFER_FNS(Pinned, pinned)
137BUFFER_FNS(Escaped, escaped)
138TAS_BUFFER_FNS(Escaped, escaped)
139
140struct gfs2_bufdata {
141 struct buffer_head *bd_bh;
142 struct gfs2_glock *bd_gl;
143 u64 bd_blkno;
144
145 struct list_head bd_list;
146 const struct gfs2_log_operations *bd_ops;
147
148 struct gfs2_trans *bd_tr;
149 struct list_head bd_ail_st_list;
150 struct list_head bd_ail_gl_list;
151};
152
153/*
154 * Internally, we prefix things with gdlm_ and GDLM_ (for gfs-dlm) since a
155 * prefix of lock_dlm_ gets awkward.
156 */
157
158#define GDLM_STRNAME_BYTES 25
159#define GDLM_LVB_SIZE 32
160
161/*
162 * ls_recover_flags:
163 *
164 * DFL_BLOCK_LOCKS: dlm is in recovery and will grant locks that had been
165 * held by failed nodes whose journals need recovery. Those locks should
166 * only be used for journal recovery until the journal recovery is done.
167 * This is set by the dlm recover_prep callback and cleared by the
168 * gfs2_control thread when journal recovery is complete. To avoid
169 * races between recover_prep setting and gfs2_control clearing, recover_spin
170 * is held while changing this bit and reading/writing recover_block
171 * and recover_start.
172 *
173 * DFL_NO_DLM_OPS: dlm lockspace ops/callbacks are not being used.
174 *
175 * DFL_FIRST_MOUNT: this node is the first to mount this fs and is doing
176 * recovery of all journals before allowing other nodes to mount the fs.
177 * This is cleared when FIRST_MOUNT_DONE is set.
178 *
179 * DFL_FIRST_MOUNT_DONE: this node was the first mounter, and has finished
180 * recovery of all journals, and now allows other nodes to mount the fs.
181 *
182 * DFL_MOUNT_DONE: gdlm_mount has completed successfully and cleared
183 * BLOCK_LOCKS for the first time. The gfs2_control thread should now
184 * control clearing BLOCK_LOCKS for further recoveries.
185 *
186 * DFL_UNMOUNT: gdlm_unmount sets to keep sdp off gfs2_control_wq.
187 *
188 * DFL_DLM_RECOVERY: set while dlm is in recovery, between recover_prep()
189 * and recover_done(), i.e. set while recover_block == recover_start.
190 */
191
192enum {
193 DFL_BLOCK_LOCKS = 0,
194 DFL_NO_DLM_OPS = 1,
195 DFL_FIRST_MOUNT = 2,
196 DFL_FIRST_MOUNT_DONE = 3,
197 DFL_MOUNT_DONE = 4,
198 DFL_UNMOUNT = 5,
199 DFL_DLM_RECOVERY = 6,
200};
201
202/*
203 * We are using struct lm_lockname as an rhashtable key. Avoid holes within
204 * the struct; padding at the end is fine.
205 */
206struct lm_lockname {
207 u64 ln_number;
208 struct gfs2_sbd *ln_sbd;
209 unsigned int ln_type;
210};
211
212#define lm_name_equal(name1, name2) \
213 (((name1)->ln_number == (name2)->ln_number) && \
214 ((name1)->ln_type == (name2)->ln_type) && \
215 ((name1)->ln_sbd == (name2)->ln_sbd))
216
217
218struct gfs2_glock_operations {
219 void (*go_sync) (struct gfs2_glock *gl);
220 int (*go_xmote_bh) (struct gfs2_glock *gl, struct gfs2_holder *gh);
221 void (*go_inval) (struct gfs2_glock *gl, int flags);
222 int (*go_demote_ok) (const struct gfs2_glock *gl);
223 int (*go_lock) (struct gfs2_holder *gh);
224 void (*go_unlock) (struct gfs2_holder *gh);
225 void (*go_dump)(struct seq_file *seq, const struct gfs2_glock *gl);
226 void (*go_callback)(struct gfs2_glock *gl, bool remote);
227 const int go_type;
228 const unsigned long go_flags;
229#define GLOF_ASPACE 1
230#define GLOF_LVB 2
231#define GLOF_LRU 4
232};
233
234enum {
235 GFS2_LKS_SRTT = 0, /* Non blocking smoothed round trip time */
236 GFS2_LKS_SRTTVAR = 1, /* Non blocking smoothed variance */
237 GFS2_LKS_SRTTB = 2, /* Blocking smoothed round trip time */
238 GFS2_LKS_SRTTVARB = 3, /* Blocking smoothed variance */
239 GFS2_LKS_SIRT = 4, /* Smoothed Inter-request time */
240 GFS2_LKS_SIRTVAR = 5, /* Smoothed Inter-request variance */
241 GFS2_LKS_DCOUNT = 6, /* Count of dlm requests */
242 GFS2_LKS_QCOUNT = 7, /* Count of gfs2_holder queues */
243 GFS2_NR_LKSTATS
244};
245
246struct gfs2_lkstats {
247 u64 stats[GFS2_NR_LKSTATS];
248};
249
250enum {
251 /* States */
252 HIF_HOLDER = 6, /* Set for gh that "holds" the glock */
253 HIF_FIRST = 7,
254 HIF_WAIT = 10,
255};
256
257struct gfs2_holder {
258 struct list_head gh_list;
259
260 struct gfs2_glock *gh_gl;
261 struct pid *gh_owner_pid;
262 u16 gh_flags;
263 u16 gh_state;
264
265 int gh_error;
266 unsigned long gh_iflags; /* HIF_... */
267 unsigned long gh_ip;
268};
269
270/* Number of quota types we support */
271#define GFS2_MAXQUOTAS 2
272
273struct gfs2_qadata { /* quota allocation data */
274 /* Quota stuff */
275 struct gfs2_quota_data *qa_qd[2 * GFS2_MAXQUOTAS];
276 struct gfs2_holder qa_qd_ghs[2 * GFS2_MAXQUOTAS];
277 unsigned int qa_qd_num;
278};
279
280/* Resource group multi-block reservation, in order of appearance:
281
282 Step 1. Function prepares to write, allocates a mb, sets the size hint.
283 Step 2. User calls inplace_reserve to target an rgrp, sets the rgrp info
284 Step 3. Function get_local_rgrp locks the rgrp, determines which bits to use
285 Step 4. Bits are assigned from the rgrp based on either the reservation
286 or wherever it can.
287*/
288
289struct gfs2_blkreserv {
290 /* components used during write (step 1): */
291 atomic_t rs_sizehint; /* hint of the write size */
292
293 struct gfs2_holder rs_rgd_gh; /* Filled in by get_local_rgrp */
294 struct rb_node rs_node; /* link to other block reservations */
295 struct gfs2_rbm rs_rbm; /* Start of reservation */
296 u32 rs_free; /* how many blocks are still free */
297 u64 rs_inum; /* Inode number for reservation */
298};
299
300/*
301 * Allocation parameters
302 * @target: The number of blocks we'd ideally like to allocate
303 * @aflags: The flags (e.g. Orlov flag)
304 *
305 * The intent is to gradually expand this structure over time in
306 * order to give more information, e.g. alignment, min extent size
307 * to the allocation code.
308 */
309struct gfs2_alloc_parms {
310 u64 target;
311 u32 min_target;
312 u32 aflags;
313 u64 allowed;
314};
315
316enum {
317 GLF_LOCK = 1,
318 GLF_DEMOTE = 3,
319 GLF_PENDING_DEMOTE = 4,
320 GLF_DEMOTE_IN_PROGRESS = 5,
321 GLF_DIRTY = 6,
322 GLF_LFLUSH = 7,
323 GLF_INVALIDATE_IN_PROGRESS = 8,
324 GLF_REPLY_PENDING = 9,
325 GLF_INITIAL = 10,
326 GLF_FROZEN = 11,
327 GLF_QUEUED = 12,
328 GLF_LRU = 13,
329 GLF_OBJECT = 14, /* Used only for tracing */
330 GLF_BLOCKING = 15,
331 GLF_INODE_CREATING = 16, /* Inode creation occurring */
332};
333
334struct gfs2_glock {
335 unsigned long gl_flags; /* GLF_... */
336 struct lm_lockname gl_name;
337
338 struct lockref gl_lockref;
339
340 /* State fields protected by gl_lockref.lock */
341 unsigned int gl_state:2, /* Current state */
342 gl_target:2, /* Target state */
343 gl_demote_state:2, /* State requested by remote node */
344 gl_req:2, /* State in last dlm request */
345 gl_reply:8; /* Last reply from the dlm */
346
347 unsigned long gl_demote_time; /* time of first demote request */
348 long gl_hold_time;
349 struct list_head gl_holders;
350
351 const struct gfs2_glock_operations *gl_ops;
352 ktime_t gl_dstamp;
353 struct gfs2_lkstats gl_stats;
354 struct dlm_lksb gl_lksb;
355 unsigned long gl_tchange;
356 void *gl_object;
357
358 struct list_head gl_lru;
359 struct list_head gl_ail_list;
360 atomic_t gl_ail_count;
361 atomic_t gl_revokes;
362 struct delayed_work gl_work;
363 union {
364 /* For inode and iopen glocks only */
365 struct work_struct gl_delete;
366 /* For rgrp glocks only */
367 struct {
368 loff_t start;
369 loff_t end;
370 } gl_vm;
371 };
372 struct rcu_head gl_rcu;
373 struct rhash_head gl_node;
374};
375
376#define GFS2_MIN_LVB_SIZE 32 /* Min size of LVB that gfs2 supports */
377
378enum {
379 GIF_INVALID = 0,
380 GIF_QD_LOCKED = 1,
381 GIF_ALLOC_FAILED = 2,
382 GIF_SW_PAGED = 3,
383 GIF_ORDERED = 4,
384 GIF_FREE_VFS_INODE = 5,
385 GIF_GLOP_PENDING = 6,
386};
387
388struct gfs2_inode {
389 struct inode i_inode;
390 u64 i_no_addr;
391 u64 i_no_formal_ino;
392 u64 i_generation;
393 u64 i_eattr;
394 unsigned long i_flags; /* GIF_... */
395 struct gfs2_glock *i_gl; /* Move into i_gh? */
396 struct gfs2_holder i_iopen_gh;
397 struct gfs2_holder i_gh; /* for prepare/commit_write only */
398 struct gfs2_qadata *i_qadata; /* quota allocation data */
399 struct gfs2_blkreserv i_res; /* rgrp multi-block reservation */
400 struct gfs2_rgrpd *i_rgd;
401 u64 i_goal; /* goal block for allocations */
402 struct rw_semaphore i_rw_mutex;
403 struct list_head i_ordered;
404 struct list_head i_trunc_list;
405 __be64 *i_hash_cache;
406 u32 i_entries;
407 u32 i_diskflags;
408 u8 i_height;
409 u8 i_depth;
410 u16 i_rahead;
411};
412
413/*
414 * Since i_inode is the first element of struct gfs2_inode,
415 * this is effectively a cast.
416 */
417static inline struct gfs2_inode *GFS2_I(struct inode *inode)
418{
419 return container_of(inode, struct gfs2_inode, i_inode);
420}
421
422static inline struct gfs2_sbd *GFS2_SB(const struct inode *inode)
423{
424 return inode->i_sb->s_fs_info;
425}
426
427struct gfs2_file {
428 struct mutex f_fl_mutex;
429 struct gfs2_holder f_fl_gh;
430};
431
432struct gfs2_revoke_replay {
433 struct list_head rr_list;
434 u64 rr_blkno;
435 unsigned int rr_where;
436};
437
438enum {
439 QDF_CHANGE = 1,
440 QDF_LOCKED = 2,
441 QDF_REFRESH = 3,
442 QDF_QMSG_QUIET = 4,
443};
444
445struct gfs2_quota_data {
446 struct hlist_bl_node qd_hlist;
447 struct list_head qd_list;
448 struct kqid qd_id;
449 struct gfs2_sbd *qd_sbd;
450 struct lockref qd_lockref;
451 struct list_head qd_lru;
452 unsigned qd_hash;
453
454 unsigned long qd_flags; /* QDF_... */
455
456 s64 qd_change;
457 s64 qd_change_sync;
458
459 unsigned int qd_slot;
460 unsigned int qd_slot_count;
461
462 struct buffer_head *qd_bh;
463 struct gfs2_quota_change *qd_bh_qc;
464 unsigned int qd_bh_count;
465
466 struct gfs2_glock *qd_gl;
467 struct gfs2_quota_lvb qd_qb;
468
469 u64 qd_sync_gen;
470 unsigned long qd_last_warn;
471 struct rcu_head qd_rcu;
472};
473
474enum {
475 TR_TOUCHED = 1,
476 TR_ATTACHED = 2,
477 TR_ALLOCED = 3,
478};
479
480struct gfs2_trans {
481 unsigned long tr_ip;
482
483 unsigned int tr_blocks;
484 unsigned int tr_revokes;
485 unsigned int tr_reserved;
486 unsigned long tr_flags;
487
488 unsigned int tr_num_buf_new;
489 unsigned int tr_num_databuf_new;
490 unsigned int tr_num_buf_rm;
491 unsigned int tr_num_databuf_rm;
492 unsigned int tr_num_revoke;
493 unsigned int tr_num_revoke_rm;
494
495 struct list_head tr_list;
496 struct list_head tr_databuf;
497 struct list_head tr_buf;
498
499 unsigned int tr_first;
500 struct list_head tr_ail1_list;
501 struct list_head tr_ail2_list;
502};
503
504struct gfs2_journal_extent {
505 struct list_head list;
506
507 unsigned int lblock; /* First logical block */
508 u64 dblock; /* First disk block */
509 u64 blocks;
510};
511
512struct gfs2_jdesc {
513 struct list_head jd_list;
514 struct list_head extent_list;
515 unsigned int nr_extents;
516 struct work_struct jd_work;
517 struct inode *jd_inode;
518 unsigned long jd_flags;
519#define JDF_RECOVERY 1
520 unsigned int jd_jid;
521 unsigned int jd_blocks;
522 int jd_recover_error;
523 /* Replay stuff */
524
525 unsigned int jd_found_blocks;
526 unsigned int jd_found_revokes;
527 unsigned int jd_replayed_blocks;
528
529 struct list_head jd_revoke_list;
530 unsigned int jd_replay_tail;
531
532};
533
534struct gfs2_statfs_change_host {
535 s64 sc_total;
536 s64 sc_free;
537 s64 sc_dinodes;
538};
539
540#define GFS2_QUOTA_DEFAULT GFS2_QUOTA_OFF
541#define GFS2_QUOTA_OFF 0
542#define GFS2_QUOTA_ACCOUNT 1
543#define GFS2_QUOTA_ON 2
544
545#define GFS2_DATA_DEFAULT GFS2_DATA_ORDERED
546#define GFS2_DATA_WRITEBACK 1
547#define GFS2_DATA_ORDERED 2
548
549#define GFS2_ERRORS_DEFAULT GFS2_ERRORS_WITHDRAW
550#define GFS2_ERRORS_WITHDRAW 0
551#define GFS2_ERRORS_CONTINUE 1 /* place holder for future feature */
552#define GFS2_ERRORS_RO 2 /* place holder for future feature */
553#define GFS2_ERRORS_PANIC 3
554
555struct gfs2_args {
556 char ar_lockproto[GFS2_LOCKNAME_LEN]; /* Name of the Lock Protocol */
557 char ar_locktable[GFS2_LOCKNAME_LEN]; /* Name of the Lock Table */
558 char ar_hostdata[GFS2_LOCKNAME_LEN]; /* Host specific data */
559 unsigned int ar_spectator:1; /* Don't get a journal */
560 unsigned int ar_localflocks:1; /* Let the VFS do flock|fcntl */
561 unsigned int ar_debug:1; /* Oops on errors */
562 unsigned int ar_posix_acl:1; /* Enable posix acls */
563 unsigned int ar_quota:2; /* off/account/on */
564 unsigned int ar_suiddir:1; /* suiddir support */
565 unsigned int ar_data:2; /* ordered/writeback */
566 unsigned int ar_meta:1; /* mount metafs */
567 unsigned int ar_discard:1; /* discard requests */
568 unsigned int ar_errors:2; /* errors=withdraw | panic */
569 unsigned int ar_nobarrier:1; /* do not send barriers */
570 unsigned int ar_rgrplvb:1; /* use lvbs for rgrp info */
571 unsigned int ar_loccookie:1; /* use location based readdir
572 cookies */
573 int ar_commit; /* Commit interval */
574 int ar_statfs_quantum; /* The fast statfs interval */
575 int ar_quota_quantum; /* The quota interval */
576 int ar_statfs_percent; /* The % change to force sync */
577};
578
579struct gfs2_tune {
580 spinlock_t gt_spin;
581
582 unsigned int gt_logd_secs;
583
584 unsigned int gt_quota_warn_period; /* Secs between quota warn msgs */
585 unsigned int gt_quota_scale_num; /* Numerator */
586 unsigned int gt_quota_scale_den; /* Denominator */
587 unsigned int gt_quota_quantum; /* Secs between syncs to quota file */
588 unsigned int gt_new_files_jdata;
589 unsigned int gt_max_readahead; /* Max bytes to read-ahead from disk */
590 unsigned int gt_complain_secs;
591 unsigned int gt_statfs_quantum;
592 unsigned int gt_statfs_slow;
593};
594
595enum {
596 SDF_JOURNAL_CHECKED = 0,
597 SDF_JOURNAL_LIVE = 1,
598 SDF_SHUTDOWN = 2,
599 SDF_NOBARRIERS = 3,
600 SDF_NORECOVERY = 4,
601 SDF_DEMOTE = 5,
602 SDF_NOJOURNALID = 6,
603 SDF_RORECOVERY = 7, /* read only recovery */
604 SDF_SKIP_DLM_UNLOCK = 8,
605 SDF_FORCE_AIL_FLUSH = 9,
606};
607
608enum gfs2_freeze_state {
609 SFS_UNFROZEN = 0,
610 SFS_STARTING_FREEZE = 1,
611 SFS_FROZEN = 2,
612};
613
614#define GFS2_FSNAME_LEN 256
615
616struct gfs2_inum_host {
617 u64 no_formal_ino;
618 u64 no_addr;
619};
620
621struct gfs2_sb_host {
622 u32 sb_magic;
623 u32 sb_type;
624 u32 sb_format;
625
626 u32 sb_fs_format;
627 u32 sb_multihost_format;
628 u32 sb_bsize;
629 u32 sb_bsize_shift;
630
631 struct gfs2_inum_host sb_master_dir;
632 struct gfs2_inum_host sb_root_dir;
633
634 char sb_lockproto[GFS2_LOCKNAME_LEN];
635 char sb_locktable[GFS2_LOCKNAME_LEN];
636};
637
638/*
639 * lm_mount() return values
640 *
641 * ls_jid - the journal ID this node should use
642 * ls_first - this node is the first to mount the file system
643 * ls_lockspace - lock module's context for this file system
644 * ls_ops - lock module's functions
645 */
646
647struct lm_lockstruct {
648 int ls_jid;
649 unsigned int ls_first;
650 const struct lm_lockops *ls_ops;
651 dlm_lockspace_t *ls_dlm;
652
653 int ls_recover_jid_done; /* These two are deprecated, */
654 int ls_recover_jid_status; /* used previously by gfs_controld */
655
656 struct dlm_lksb ls_mounted_lksb; /* mounted_lock */
657 struct dlm_lksb ls_control_lksb; /* control_lock */
658 char ls_control_lvb[GDLM_LVB_SIZE]; /* control_lock lvb */
659 struct completion ls_sync_wait; /* {control,mounted}_{lock,unlock} */
660 char *ls_lvb_bits;
661
662 spinlock_t ls_recover_spin; /* protects following fields */
663 unsigned long ls_recover_flags; /* DFL_ */
664 uint32_t ls_recover_mount; /* gen in first recover_done cb */
665 uint32_t ls_recover_start; /* gen in last recover_done cb */
666 uint32_t ls_recover_block; /* copy recover_start in last recover_prep */
667 uint32_t ls_recover_size; /* size of recover_submit, recover_result */
668 uint32_t *ls_recover_submit; /* gen in last recover_slot cb per jid */
669 uint32_t *ls_recover_result; /* result of last jid recovery */
670};
671
672struct gfs2_pcpu_lkstats {
673 /* One struct for each glock type */
674 struct gfs2_lkstats lkstats[10];
675};
676
677struct gfs2_sbd {
678 struct super_block *sd_vfs;
679 struct gfs2_pcpu_lkstats __percpu *sd_lkstats;
680 struct kobject sd_kobj;
681 unsigned long sd_flags; /* SDF_... */
682 struct gfs2_sb_host sd_sb;
683
684 /* Constants computed on mount */
685
686 u32 sd_fsb2bb;
687 u32 sd_fsb2bb_shift;
688 u32 sd_diptrs; /* Number of pointers in a dinode */
689 u32 sd_inptrs; /* Number of pointers in a indirect block */
690 u32 sd_jbsize; /* Size of a journaled data block */
691 u32 sd_hash_bsize; /* sizeof(exhash block) */
692 u32 sd_hash_bsize_shift;
693 u32 sd_hash_ptrs; /* Number of pointers in a hash block */
694 u32 sd_qc_per_block;
695 u32 sd_blocks_per_bitmap;
696 u32 sd_max_dirres; /* Max blocks needed to add a directory entry */
697 u32 sd_max_height; /* Max height of a file's metadata tree */
698 u64 sd_heightsize[GFS2_MAX_META_HEIGHT + 1];
699 u32 sd_max_jheight; /* Max height of journaled file's meta tree */
700 u64 sd_jheightsize[GFS2_MAX_META_HEIGHT + 1];
701 u32 sd_max_dents_per_leaf; /* Max number of dirents in a leaf block */
702
703 struct gfs2_args sd_args; /* Mount arguments */
704 struct gfs2_tune sd_tune; /* Filesystem tuning structure */
705
706 /* Lock Stuff */
707
708 struct lm_lockstruct sd_lockstruct;
709 struct gfs2_holder sd_live_gh;
710 struct gfs2_glock *sd_rename_gl;
711 struct gfs2_glock *sd_freeze_gl;
712 struct work_struct sd_freeze_work;
713 wait_queue_head_t sd_glock_wait;
714 atomic_t sd_glock_disposal;
715 struct completion sd_locking_init;
716 struct completion sd_wdack;
717 struct delayed_work sd_control_work;
718
719 /* Inode Stuff */
720
721 struct dentry *sd_master_dir;
722 struct dentry *sd_root_dir;
723
724 struct inode *sd_jindex;
725 struct inode *sd_statfs_inode;
726 struct inode *sd_sc_inode;
727 struct inode *sd_qc_inode;
728 struct inode *sd_rindex;
729 struct inode *sd_quota_inode;
730
731 /* StatFS stuff */
732
733 spinlock_t sd_statfs_spin;
734 struct gfs2_statfs_change_host sd_statfs_master;
735 struct gfs2_statfs_change_host sd_statfs_local;
736 int sd_statfs_force_sync;
737
738 /* Resource group stuff */
739
740 int sd_rindex_uptodate;
741 spinlock_t sd_rindex_spin;
742 struct rb_root sd_rindex_tree;
743 unsigned int sd_rgrps;
744 unsigned int sd_max_rg_data;
745
746 /* Journal index stuff */
747
748 struct list_head sd_jindex_list;
749 spinlock_t sd_jindex_spin;
750 struct mutex sd_jindex_mutex;
751 unsigned int sd_journals;
752
753 struct gfs2_jdesc *sd_jdesc;
754 struct gfs2_holder sd_journal_gh;
755 struct gfs2_holder sd_jinode_gh;
756
757 struct gfs2_holder sd_sc_gh;
758 struct gfs2_holder sd_qc_gh;
759
760 struct completion sd_journal_ready;
761
762 /* Daemon stuff */
763
764 struct task_struct *sd_logd_process;
765 struct task_struct *sd_quotad_process;
766
767 /* Quota stuff */
768
769 struct list_head sd_quota_list;
770 atomic_t sd_quota_count;
771 struct mutex sd_quota_mutex;
772 struct mutex sd_quota_sync_mutex;
773 wait_queue_head_t sd_quota_wait;
774 struct list_head sd_trunc_list;
775 spinlock_t sd_trunc_lock;
776
777 unsigned int sd_quota_slots;
778 unsigned long *sd_quota_bitmap;
779 spinlock_t sd_bitmap_lock;
780
781 u64 sd_quota_sync_gen;
782
783 /* Log stuff */
784
785 struct address_space sd_aspace;
786
787 spinlock_t sd_log_lock;
788
789 struct gfs2_trans *sd_log_tr;
790 unsigned int sd_log_blks_reserved;
791 int sd_log_commited_revoke;
792
793 atomic_t sd_log_pinned;
794 unsigned int sd_log_num_revoke;
795
796 struct list_head sd_log_le_revoke;
797 struct list_head sd_log_le_ordered;
798 spinlock_t sd_ordered_lock;
799
800 atomic_t sd_log_thresh1;
801 atomic_t sd_log_thresh2;
802 atomic_t sd_log_blks_free;
803 atomic_t sd_log_blks_needed;
804 wait_queue_head_t sd_log_waitq;
805 wait_queue_head_t sd_logd_waitq;
806
807 u64 sd_log_sequence;
808 unsigned int sd_log_head;
809 unsigned int sd_log_tail;
810 int sd_log_idle;
811
812 struct rw_semaphore sd_log_flush_lock;
813 atomic_t sd_log_in_flight;
814 struct bio *sd_log_bio;
815 wait_queue_head_t sd_log_flush_wait;
816 int sd_log_error;
817
818 atomic_t sd_reserving_log;
819 wait_queue_head_t sd_reserving_log_wait;
820
821 unsigned int sd_log_flush_head;
822
823 spinlock_t sd_ail_lock;
824 struct list_head sd_ail1_list;
825 struct list_head sd_ail2_list;
826
827 /* For quiescing the filesystem */
828 struct gfs2_holder sd_freeze_gh;
829 atomic_t sd_freeze_state;
830 struct mutex sd_freeze_mutex;
831
832 char sd_fsname[GFS2_FSNAME_LEN + 3 * sizeof(int) + 2];
833 char sd_table_name[GFS2_FSNAME_LEN];
834 char sd_proto_name[GFS2_FSNAME_LEN];
835
836 /* Debugging crud */
837
838 unsigned long sd_last_warning;
839 struct dentry *debugfs_dir; /* debugfs directory */
840 struct dentry *debugfs_dentry_glocks;
841 struct dentry *debugfs_dentry_glstats;
842 struct dentry *debugfs_dentry_sbstats;
843};
844
845static inline void gfs2_glstats_inc(struct gfs2_glock *gl, int which)
846{
847 gl->gl_stats.stats[which]++;
848}
849
850static inline void gfs2_sbstats_inc(const struct gfs2_glock *gl, int which)
851{
852 const struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
853 preempt_disable();
854 this_cpu_ptr(sdp->sd_lkstats)->lkstats[gl->gl_name.ln_type].stats[which]++;
855 preempt_enable();
856}
857
858extern struct gfs2_rgrpd *gfs2_glock2rgrp(struct gfs2_glock *gl);
859
860static inline unsigned gfs2_max_stuffed_size(const struct gfs2_inode *ip)
861{
862 return GFS2_SB(&ip->i_inode)->sd_sb.sb_bsize - sizeof(struct gfs2_dinode);
863}
864
865#endif /* __INCORE_DOT_H__ */
866
1/* SPDX-License-Identifier: GPL-2.0-only */
2/*
3 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
4 * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
5 */
6
7#ifndef __INCORE_DOT_H__
8#define __INCORE_DOT_H__
9
10#include <linux/fs.h>
11#include <linux/kobject.h>
12#include <linux/workqueue.h>
13#include <linux/dlm.h>
14#include <linux/buffer_head.h>
15#include <linux/rcupdate.h>
16#include <linux/rculist_bl.h>
17#include <linux/completion.h>
18#include <linux/rbtree.h>
19#include <linux/ktime.h>
20#include <linux/percpu.h>
21#include <linux/lockref.h>
22#include <linux/rhashtable.h>
23#include <linux/mutex.h>
24
25#define DIO_WAIT 0x00000010
26#define DIO_METADATA 0x00000020
27
28struct gfs2_log_operations;
29struct gfs2_bufdata;
30struct gfs2_holder;
31struct gfs2_glock;
32struct gfs2_quota_data;
33struct gfs2_trans;
34struct gfs2_jdesc;
35struct gfs2_sbd;
36struct lm_lockops;
37
38typedef void (*gfs2_glop_bh_t) (struct gfs2_glock *gl, unsigned int ret);
39
40struct gfs2_log_header_host {
41 u64 lh_sequence; /* Sequence number of this transaction */
42 u32 lh_flags; /* GFS2_LOG_HEAD_... */
43 u32 lh_tail; /* Block number of log tail */
44 u32 lh_blkno;
45
46 s64 lh_local_total;
47 s64 lh_local_free;
48 s64 lh_local_dinodes;
49};
50
51/*
52 * Structure of operations that are associated with each
53 * type of element in the log.
54 */
55
56struct gfs2_log_operations {
57 void (*lo_before_commit) (struct gfs2_sbd *sdp, struct gfs2_trans *tr);
58 void (*lo_after_commit) (struct gfs2_sbd *sdp, struct gfs2_trans *tr);
59 void (*lo_before_scan) (struct gfs2_jdesc *jd,
60 struct gfs2_log_header_host *head, int pass);
61 int (*lo_scan_elements) (struct gfs2_jdesc *jd, unsigned int start,
62 struct gfs2_log_descriptor *ld, __be64 *ptr,
63 int pass);
64 void (*lo_after_scan) (struct gfs2_jdesc *jd, int error, int pass);
65 const char *lo_name;
66};
67
68#define GBF_FULL 1
69
70/**
71 * Clone bitmaps (bi_clone):
72 *
73 * - When a block is freed, we remember the previous state of the block in the
74 * clone bitmap, and only mark the block as free in the real bitmap.
75 *
76 * - When looking for a block to allocate, we check for a free block in the
77 * clone bitmap, and if no clone bitmap exists, in the real bitmap.
78 *
79 * - For allocating a block, we mark it as allocated in the real bitmap, and if
80 * a clone bitmap exists, also in the clone bitmap.
81 *
82 * - At the end of a log_flush, we copy the real bitmap into the clone bitmap
83 * to make the clone bitmap reflect the current allocation state.
84 * (Alternatively, we could remove the clone bitmap.)
85 *
86 * The clone bitmaps are in-core only, and is never written to disk.
87 *
88 * These steps ensure that blocks which have been freed in a transaction cannot
89 * be reallocated in that same transaction.
90 */
91struct gfs2_bitmap {
92 struct buffer_head *bi_bh;
93 char *bi_clone;
94 unsigned long bi_flags;
95 u32 bi_offset;
96 u32 bi_start;
97 u32 bi_bytes;
98 u32 bi_blocks;
99};
100
101struct gfs2_rgrpd {
102 struct rb_node rd_node; /* Link with superblock */
103 struct gfs2_glock *rd_gl; /* Glock for this rgrp */
104 u64 rd_addr; /* grp block disk address */
105 u64 rd_data0; /* first data location */
106 u32 rd_length; /* length of rgrp header in fs blocks */
107 u32 rd_data; /* num of data blocks in rgrp */
108 u32 rd_bitbytes; /* number of bytes in data bitmaps */
109 u32 rd_free;
110 u32 rd_requested; /* number of blocks in rd_rstree */
111 u32 rd_reserved; /* number of reserved blocks */
112 u32 rd_free_clone;
113 u32 rd_dinodes;
114 u64 rd_igeneration;
115 struct gfs2_bitmap *rd_bits;
116 struct gfs2_sbd *rd_sbd;
117 struct gfs2_rgrp_lvb *rd_rgl;
118 u32 rd_last_alloc;
119 u32 rd_flags;
120 u32 rd_extfail_pt; /* extent failure point */
121#define GFS2_RDF_CHECK 0x10000000 /* check for unlinked inodes */
122#define GFS2_RDF_ERROR 0x40000000 /* error in rg */
123#define GFS2_RDF_PREFERRED 0x80000000 /* This rgrp is preferred */
124#define GFS2_RDF_MASK 0xf0000000 /* mask for internal flags */
125 spinlock_t rd_rsspin; /* protects reservation related vars */
126 struct mutex rd_mutex;
127 struct rb_root rd_rstree; /* multi-block reservation tree */
128};
129
130enum gfs2_state_bits {
131 BH_Pinned = BH_PrivateStart,
132 BH_Escaped = BH_PrivateStart + 1,
133};
134
135BUFFER_FNS(Pinned, pinned)
136TAS_BUFFER_FNS(Pinned, pinned)
137BUFFER_FNS(Escaped, escaped)
138TAS_BUFFER_FNS(Escaped, escaped)
139
140struct gfs2_bufdata {
141 struct buffer_head *bd_bh;
142 struct gfs2_glock *bd_gl;
143 u64 bd_blkno;
144
145 struct list_head bd_list;
146
147 struct gfs2_trans *bd_tr;
148 struct list_head bd_ail_st_list;
149 struct list_head bd_ail_gl_list;
150};
151
152/*
153 * Internally, we prefix things with gdlm_ and GDLM_ (for gfs-dlm) since a
154 * prefix of lock_dlm_ gets awkward.
155 */
156
157#define GDLM_STRNAME_BYTES 25
158#define GDLM_LVB_SIZE 32
159
160/*
161 * ls_recover_flags:
162 *
163 * DFL_BLOCK_LOCKS: dlm is in recovery and will grant locks that had been
164 * held by failed nodes whose journals need recovery. Those locks should
165 * only be used for journal recovery until the journal recovery is done.
166 * This is set by the dlm recover_prep callback and cleared by the
167 * gfs2_control thread when journal recovery is complete. To avoid
168 * races between recover_prep setting and gfs2_control clearing, recover_spin
169 * is held while changing this bit and reading/writing recover_block
170 * and recover_start.
171 *
172 * DFL_NO_DLM_OPS: dlm lockspace ops/callbacks are not being used.
173 *
174 * DFL_FIRST_MOUNT: this node is the first to mount this fs and is doing
175 * recovery of all journals before allowing other nodes to mount the fs.
176 * This is cleared when FIRST_MOUNT_DONE is set.
177 *
178 * DFL_FIRST_MOUNT_DONE: this node was the first mounter, and has finished
179 * recovery of all journals, and now allows other nodes to mount the fs.
180 *
181 * DFL_MOUNT_DONE: gdlm_mount has completed successfully and cleared
182 * BLOCK_LOCKS for the first time. The gfs2_control thread should now
183 * control clearing BLOCK_LOCKS for further recoveries.
184 *
185 * DFL_UNMOUNT: gdlm_unmount sets to keep sdp off gfs2_control_wq.
186 *
187 * DFL_DLM_RECOVERY: set while dlm is in recovery, between recover_prep()
188 * and recover_done(), i.e. set while recover_block == recover_start.
189 */
190
191enum {
192 DFL_BLOCK_LOCKS = 0,
193 DFL_NO_DLM_OPS = 1,
194 DFL_FIRST_MOUNT = 2,
195 DFL_FIRST_MOUNT_DONE = 3,
196 DFL_MOUNT_DONE = 4,
197 DFL_UNMOUNT = 5,
198 DFL_DLM_RECOVERY = 6,
199};
200
201/*
202 * We are using struct lm_lockname as an rhashtable key. Avoid holes within
203 * the struct; padding at the end is fine.
204 */
205struct lm_lockname {
206 u64 ln_number;
207 struct gfs2_sbd *ln_sbd;
208 unsigned int ln_type;
209};
210
211#define lm_name_equal(name1, name2) \
212 (((name1)->ln_number == (name2)->ln_number) && \
213 ((name1)->ln_type == (name2)->ln_type) && \
214 ((name1)->ln_sbd == (name2)->ln_sbd))
215
216
217struct gfs2_glock_operations {
218 int (*go_sync) (struct gfs2_glock *gl);
219 int (*go_xmote_bh)(struct gfs2_glock *gl);
220 void (*go_inval) (struct gfs2_glock *gl, int flags);
221 int (*go_instantiate) (struct gfs2_glock *gl);
222 int (*go_held)(struct gfs2_holder *gh);
223 void (*go_dump)(struct seq_file *seq, const struct gfs2_glock *gl,
224 const char *fs_id_buf);
225 void (*go_callback)(struct gfs2_glock *gl, bool remote);
226 void (*go_unlocked)(struct gfs2_glock *gl);
227 const int go_subclass;
228 const int go_type;
229 const unsigned long go_flags;
230#define GLOF_ASPACE 1 /* address space attached */
231#define GLOF_LVB 2 /* Lock Value Block attached */
232#define GLOF_NONDISK 8 /* not I/O related */
233};
234
235enum {
236 GFS2_LKS_SRTT = 0, /* Non blocking smoothed round trip time */
237 GFS2_LKS_SRTTVAR = 1, /* Non blocking smoothed variance */
238 GFS2_LKS_SRTTB = 2, /* Blocking smoothed round trip time */
239 GFS2_LKS_SRTTVARB = 3, /* Blocking smoothed variance */
240 GFS2_LKS_SIRT = 4, /* Smoothed Inter-request time */
241 GFS2_LKS_SIRTVAR = 5, /* Smoothed Inter-request variance */
242 GFS2_LKS_DCOUNT = 6, /* Count of dlm requests */
243 GFS2_LKS_QCOUNT = 7, /* Count of gfs2_holder queues */
244 GFS2_NR_LKSTATS
245};
246
247struct gfs2_lkstats {
248 u64 stats[GFS2_NR_LKSTATS];
249};
250
251enum {
252 /* States */
253 HIF_HOLDER = 6, /* Set for gh that "holds" the glock */
254 HIF_WAIT = 10,
255};
256
257struct gfs2_holder {
258 struct list_head gh_list;
259
260 struct gfs2_glock *gh_gl;
261 struct pid *gh_owner_pid;
262 u16 gh_flags;
263 u16 gh_state;
264
265 int gh_error;
266 unsigned long gh_iflags; /* HIF_... */
267 unsigned long gh_ip;
268};
269
270/* Number of quota types we support */
271#define GFS2_MAXQUOTAS 2
272
273struct gfs2_qadata { /* quota allocation data */
274 /* Quota stuff */
275 struct gfs2_quota_data *qa_qd[2 * GFS2_MAXQUOTAS];
276 struct gfs2_holder qa_qd_ghs[2 * GFS2_MAXQUOTAS];
277 unsigned int qa_qd_num;
278 int qa_ref;
279};
280
281/* Resource group multi-block reservation, in order of appearance:
282
283 Step 1. Function prepares to write, allocates a mb, sets the size hint.
284 Step 2. User calls inplace_reserve to target an rgrp, sets the rgrp info
285 Step 3. Function get_local_rgrp locks the rgrp, determines which bits to use
286 Step 4. Bits are assigned from the rgrp based on either the reservation
287 or wherever it can.
288*/
289
290struct gfs2_blkreserv {
291 struct rb_node rs_node; /* node within rd_rstree */
292 struct gfs2_rgrpd *rs_rgd;
293 u64 rs_start;
294 u32 rs_requested;
295 u32 rs_reserved; /* number of reserved blocks */
296};
297
298/*
299 * Allocation parameters
300 * @target: The number of blocks we'd ideally like to allocate
301 * @aflags: The flags (e.g. Orlov flag)
302 *
303 * The intent is to gradually expand this structure over time in
304 * order to give more information, e.g. alignment, min extent size
305 * to the allocation code.
306 */
307struct gfs2_alloc_parms {
308 u64 target;
309 u32 min_target;
310 u32 aflags;
311 u64 allowed;
312};
313
314enum {
315 GLF_LOCK = 1,
316 GLF_INSTANTIATE_NEEDED = 2, /* needs instantiate */
317 GLF_DEMOTE = 3,
318 GLF_PENDING_DEMOTE = 4,
319 GLF_DEMOTE_IN_PROGRESS = 5,
320 GLF_DIRTY = 6,
321 GLF_LFLUSH = 7,
322 GLF_INVALIDATE_IN_PROGRESS = 8,
323 GLF_HAVE_REPLY = 9,
324 GLF_INITIAL = 10,
325 GLF_HAVE_FROZEN_REPLY = 11,
326 GLF_INSTANTIATE_IN_PROG = 12, /* instantiate happening now */
327 GLF_LRU = 13,
328 GLF_OBJECT = 14, /* Used only for tracing */
329 GLF_BLOCKING = 15,
330 GLF_UNLOCKED = 16, /* Wait for glock to be unlocked */
331 GLF_TRY_TO_EVICT = 17, /* iopen glocks only */
332 GLF_VERIFY_DELETE = 18, /* iopen glocks only */
333};
334
335struct gfs2_glock {
336 unsigned long gl_flags; /* GLF_... */
337 struct lm_lockname gl_name;
338
339 struct lockref gl_lockref;
340
341 /* State fields protected by gl_lockref.lock */
342 unsigned int gl_state:2, /* Current state */
343 gl_target:2, /* Target state */
344 gl_demote_state:2, /* State requested by remote node */
345 gl_req:2, /* State in last dlm request */
346 gl_reply:8; /* Last reply from the dlm */
347
348 unsigned long gl_demote_time; /* time of first demote request */
349 long gl_hold_time;
350 struct list_head gl_holders;
351
352 const struct gfs2_glock_operations *gl_ops;
353 ktime_t gl_dstamp;
354 struct gfs2_lkstats gl_stats;
355 struct dlm_lksb gl_lksb;
356 unsigned long gl_tchange;
357 void *gl_object;
358
359 struct list_head gl_lru;
360 struct list_head gl_ail_list;
361 atomic_t gl_ail_count;
362 atomic_t gl_revokes;
363 struct delayed_work gl_work;
364 /* For iopen glocks only */
365 struct {
366 struct delayed_work gl_delete;
367 u64 gl_no_formal_ino;
368 };
369 struct rcu_head gl_rcu;
370 struct rhash_head gl_node;
371};
372
373enum {
374 GIF_QD_LOCKED = 1,
375 GIF_ALLOC_FAILED = 2,
376 GIF_SW_PAGED = 3,
377 GIF_FREE_VFS_INODE = 5,
378 GIF_GLOP_PENDING = 6,
379 GIF_DEFER_DELETE = 7,
380};
381
382struct gfs2_inode {
383 struct inode i_inode;
384 u64 i_no_addr;
385 u64 i_no_formal_ino;
386 u64 i_generation;
387 u64 i_eattr;
388 unsigned long i_flags; /* GIF_... */
389 struct gfs2_glock *i_gl;
390 struct gfs2_holder i_iopen_gh;
391 struct gfs2_qadata *i_qadata; /* quota allocation data */
392 struct gfs2_holder i_rgd_gh;
393 struct gfs2_blkreserv i_res; /* rgrp multi-block reservation */
394 u64 i_goal; /* goal block for allocations */
395 atomic_t i_sizehint; /* hint of the write size */
396 struct rw_semaphore i_rw_mutex;
397 struct list_head i_ordered;
398 __be64 *i_hash_cache;
399 u32 i_entries;
400 u32 i_diskflags;
401 u8 i_height;
402 u8 i_depth;
403 u16 i_rahead;
404};
405
406/*
407 * Since i_inode is the first element of struct gfs2_inode,
408 * this is effectively a cast.
409 */
410static inline struct gfs2_inode *GFS2_I(struct inode *inode)
411{
412 return container_of(inode, struct gfs2_inode, i_inode);
413}
414
415static inline struct gfs2_sbd *GFS2_SB(const struct inode *inode)
416{
417 return inode->i_sb->s_fs_info;
418}
419
420struct gfs2_file {
421 struct mutex f_fl_mutex;
422 struct gfs2_holder f_fl_gh;
423};
424
425struct gfs2_revoke_replay {
426 struct list_head rr_list;
427 u64 rr_blkno;
428 unsigned int rr_where;
429};
430
431enum {
432 QDF_CHANGE = 1,
433 QDF_LOCKED = 2,
434 QDF_REFRESH = 3,
435 QDF_QMSG_QUIET = 4,
436};
437
438struct gfs2_quota_data {
439 struct hlist_bl_node qd_hlist;
440 struct list_head qd_list;
441 struct kqid qd_id;
442 struct gfs2_sbd *qd_sbd;
443 struct lockref qd_lockref;
444 struct list_head qd_lru;
445 unsigned qd_hash;
446
447 unsigned long qd_flags; /* QDF_... */
448
449 s64 qd_change;
450 s64 qd_change_sync;
451
452 unsigned int qd_slot;
453 unsigned int qd_slot_ref;
454
455 struct buffer_head *qd_bh;
456 struct gfs2_quota_change *qd_bh_qc;
457 unsigned int qd_bh_count;
458
459 struct gfs2_glock *qd_gl;
460 struct gfs2_quota_lvb qd_qb;
461
462 u64 qd_sync_gen;
463 unsigned long qd_last_warn;
464 struct rcu_head qd_rcu;
465};
466
467enum {
468 TR_TOUCHED = 1,
469 TR_ATTACHED = 2,
470 TR_ONSTACK = 3,
471};
472
473struct gfs2_trans {
474 unsigned long tr_ip;
475
476 unsigned int tr_blocks;
477 unsigned int tr_revokes;
478 unsigned int tr_reserved;
479 unsigned long tr_flags;
480
481 unsigned int tr_num_buf_new;
482 unsigned int tr_num_databuf_new;
483 unsigned int tr_num_buf_rm;
484 unsigned int tr_num_databuf_rm;
485 unsigned int tr_num_revoke;
486
487 struct list_head tr_list;
488 struct list_head tr_databuf;
489 struct list_head tr_buf;
490
491 unsigned int tr_first;
492 struct list_head tr_ail1_list;
493 struct list_head tr_ail2_list;
494};
495
496struct gfs2_journal_extent {
497 struct list_head list;
498
499 unsigned int lblock; /* First logical block */
500 u64 dblock; /* First disk block */
501 u64 blocks;
502};
503
504struct gfs2_jdesc {
505 struct list_head jd_list;
506 struct list_head extent_list;
507 unsigned int nr_extents;
508 struct work_struct jd_work;
509 struct inode *jd_inode;
510 struct bio *jd_log_bio;
511 unsigned long jd_flags;
512#define JDF_RECOVERY 1
513 unsigned int jd_jid;
514 u32 jd_blocks;
515 int jd_recover_error;
516 /* Replay stuff */
517
518 unsigned int jd_found_blocks;
519 unsigned int jd_found_revokes;
520 unsigned int jd_replayed_blocks;
521
522 struct list_head jd_revoke_list;
523 unsigned int jd_replay_tail;
524
525 u64 jd_no_addr;
526};
527
528struct gfs2_statfs_change_host {
529 s64 sc_total;
530 s64 sc_free;
531 s64 sc_dinodes;
532};
533
534#define GFS2_QUOTA_DEFAULT GFS2_QUOTA_OFF
535#define GFS2_QUOTA_OFF 0
536#define GFS2_QUOTA_ACCOUNT 1
537#define GFS2_QUOTA_ON 2
538#define GFS2_QUOTA_QUIET 3 /* on but not complaining */
539
540#define GFS2_DATA_DEFAULT GFS2_DATA_ORDERED
541#define GFS2_DATA_WRITEBACK 1
542#define GFS2_DATA_ORDERED 2
543
544#define GFS2_ERRORS_DEFAULT GFS2_ERRORS_WITHDRAW
545#define GFS2_ERRORS_WITHDRAW 0
546#define GFS2_ERRORS_CONTINUE 1 /* place holder for future feature */
547#define GFS2_ERRORS_RO 2 /* place holder for future feature */
548#define GFS2_ERRORS_PANIC 3
549
550struct gfs2_args {
551 char ar_lockproto[GFS2_LOCKNAME_LEN]; /* Name of the Lock Protocol */
552 char ar_locktable[GFS2_LOCKNAME_LEN]; /* Name of the Lock Table */
553 char ar_hostdata[GFS2_LOCKNAME_LEN]; /* Host specific data */
554 unsigned int ar_spectator:1; /* Don't get a journal */
555 unsigned int ar_localflocks:1; /* Let the VFS do flock|fcntl */
556 unsigned int ar_debug:1; /* Oops on errors */
557 unsigned int ar_posix_acl:1; /* Enable posix acls */
558 unsigned int ar_quota:2; /* off/account/on */
559 unsigned int ar_suiddir:1; /* suiddir support */
560 unsigned int ar_data:2; /* ordered/writeback */
561 unsigned int ar_meta:1; /* mount metafs */
562 unsigned int ar_discard:1; /* discard requests */
563 unsigned int ar_errors:2; /* errors=withdraw | panic */
564 unsigned int ar_nobarrier:1; /* do not send barriers */
565 unsigned int ar_rgrplvb:1; /* use lvbs for rgrp info */
566 unsigned int ar_got_rgrplvb:1; /* Was the rgrplvb opt given? */
567 unsigned int ar_loccookie:1; /* use location based readdir
568 cookies */
569 s32 ar_commit; /* Commit interval */
570 s32 ar_statfs_quantum; /* The fast statfs interval */
571 s32 ar_quota_quantum; /* The quota interval */
572 s32 ar_statfs_percent; /* The % change to force sync */
573};
574
575struct gfs2_tune {
576 spinlock_t gt_spin;
577
578 unsigned int gt_logd_secs;
579
580 unsigned int gt_quota_warn_period; /* Secs between quota warn msgs */
581 unsigned int gt_quota_scale_num; /* Numerator */
582 unsigned int gt_quota_scale_den; /* Denominator */
583 unsigned int gt_quota_quantum; /* Secs between syncs to quota file */
584 unsigned int gt_new_files_jdata;
585 unsigned int gt_max_readahead; /* Max bytes to read-ahead from disk */
586 unsigned int gt_complain_secs;
587 unsigned int gt_statfs_quantum;
588 unsigned int gt_statfs_slow;
589};
590
591enum {
592 SDF_JOURNAL_CHECKED = 0,
593 SDF_JOURNAL_LIVE = 1,
594 SDF_WITHDRAWN = 2,
595 SDF_NOBARRIERS = 3,
596 SDF_NORECOVERY = 4,
597 SDF_DEMOTE = 5,
598 SDF_NOJOURNALID = 6,
599 SDF_RORECOVERY = 7, /* read only recovery */
600 SDF_SKIP_DLM_UNLOCK = 8,
601 SDF_FORCE_AIL_FLUSH = 9,
602 SDF_FREEZE_INITIATOR = 10,
603 SDF_WITHDRAWING = 11, /* Will withdraw eventually */
604 SDF_WITHDRAW_IN_PROG = 12, /* Withdraw is in progress */
605 SDF_REMOTE_WITHDRAW = 13, /* Performing remote recovery */
606 SDF_WITHDRAW_RECOVERY = 14, /* Wait for journal recovery when we are
607 withdrawing */
608 SDF_KILL = 15,
609 SDF_EVICTING = 16,
610 SDF_FROZEN = 17,
611};
612
613#define GFS2_FSNAME_LEN 256
614
615struct gfs2_inum_host {
616 u64 no_formal_ino;
617 u64 no_addr;
618};
619
620struct gfs2_sb_host {
621 u32 sb_magic;
622 u32 sb_type;
623
624 u32 sb_fs_format;
625 u32 sb_multihost_format;
626 u32 sb_bsize;
627 u32 sb_bsize_shift;
628
629 struct gfs2_inum_host sb_master_dir;
630 struct gfs2_inum_host sb_root_dir;
631
632 char sb_lockproto[GFS2_LOCKNAME_LEN];
633 char sb_locktable[GFS2_LOCKNAME_LEN];
634};
635
636/*
637 * lm_mount() return values
638 *
639 * ls_jid - the journal ID this node should use
640 * ls_first - this node is the first to mount the file system
641 * ls_lockspace - lock module's context for this file system
642 * ls_ops - lock module's functions
643 */
644
645struct lm_lockstruct {
646 int ls_jid;
647 unsigned int ls_first;
648 const struct lm_lockops *ls_ops;
649 dlm_lockspace_t *ls_dlm;
650
651 int ls_recover_jid_done; /* These two are deprecated, */
652 int ls_recover_jid_status; /* used previously by gfs_controld */
653
654 struct dlm_lksb ls_mounted_lksb; /* mounted_lock */
655 struct dlm_lksb ls_control_lksb; /* control_lock */
656 char ls_control_lvb[GDLM_LVB_SIZE]; /* control_lock lvb */
657 struct completion ls_sync_wait; /* {control,mounted}_{lock,unlock} */
658 char *ls_lvb_bits;
659
660 spinlock_t ls_recover_spin; /* protects following fields */
661 unsigned long ls_recover_flags; /* DFL_ */
662 uint32_t ls_recover_mount; /* gen in first recover_done cb */
663 uint32_t ls_recover_start; /* gen in last recover_done cb */
664 uint32_t ls_recover_block; /* copy recover_start in last recover_prep */
665 uint32_t ls_recover_size; /* size of recover_submit, recover_result */
666 uint32_t *ls_recover_submit; /* gen in last recover_slot cb per jid */
667 uint32_t *ls_recover_result; /* result of last jid recovery */
668};
669
670struct gfs2_pcpu_lkstats {
671 /* One struct for each glock type */
672 struct gfs2_lkstats lkstats[10];
673};
674
675/* List of local (per node) statfs inodes */
676struct local_statfs_inode {
677 struct list_head si_list;
678 struct inode *si_sc_inode;
679 unsigned int si_jid; /* journal id this statfs inode corresponds to */
680};
681
682struct gfs2_sbd {
683 struct super_block *sd_vfs;
684 struct gfs2_pcpu_lkstats __percpu *sd_lkstats;
685 struct kobject sd_kobj;
686 struct completion sd_kobj_unregister;
687 unsigned long sd_flags; /* SDF_... */
688 struct gfs2_sb_host sd_sb;
689
690 /* Constants computed on mount */
691
692 u32 sd_fsb2bb;
693 u32 sd_fsb2bb_shift;
694 u32 sd_diptrs; /* Number of pointers in a dinode */
695 u32 sd_inptrs; /* Number of pointers in a indirect block */
696 u32 sd_ldptrs; /* Number of pointers in a log descriptor block */
697 u32 sd_jbsize; /* Size of a journaled data block */
698 u32 sd_hash_bsize; /* sizeof(exhash block) */
699 u32 sd_hash_bsize_shift;
700 u32 sd_hash_ptrs; /* Number of pointers in a hash block */
701 u32 sd_qc_per_block;
702 u32 sd_blocks_per_bitmap;
703 u32 sd_max_dirres; /* Max blocks needed to add a directory entry */
704 u32 sd_max_height; /* Max height of a file's metadata tree */
705 u64 sd_heightsize[GFS2_MAX_META_HEIGHT + 1];
706 u32 sd_max_dents_per_leaf; /* Max number of dirents in a leaf block */
707
708 struct gfs2_args sd_args; /* Mount arguments */
709 struct gfs2_tune sd_tune; /* Filesystem tuning structure */
710
711 /* Lock Stuff */
712
713 struct lm_lockstruct sd_lockstruct;
714 struct gfs2_holder sd_live_gh;
715 struct gfs2_glock *sd_rename_gl;
716 struct gfs2_glock *sd_freeze_gl;
717 struct work_struct sd_freeze_work;
718 wait_queue_head_t sd_kill_wait;
719 wait_queue_head_t sd_async_glock_wait;
720 atomic_t sd_glock_disposal;
721 struct completion sd_locking_init;
722 struct completion sd_wdack;
723 struct delayed_work sd_control_work;
724
725 /* Inode Stuff */
726
727 struct dentry *sd_master_dir;
728 struct dentry *sd_root_dir;
729
730 struct inode *sd_jindex;
731 struct inode *sd_statfs_inode;
732 struct inode *sd_sc_inode;
733 struct list_head sd_sc_inodes_list;
734 struct inode *sd_qc_inode;
735 struct inode *sd_rindex;
736 struct inode *sd_quota_inode;
737
738 /* StatFS stuff */
739
740 spinlock_t sd_statfs_spin;
741 struct gfs2_statfs_change_host sd_statfs_master;
742 struct gfs2_statfs_change_host sd_statfs_local;
743 int sd_statfs_force_sync;
744
745 /* Resource group stuff */
746
747 int sd_rindex_uptodate;
748 spinlock_t sd_rindex_spin;
749 struct rb_root sd_rindex_tree;
750 unsigned int sd_rgrps;
751 unsigned int sd_max_rg_data;
752
753 /* Journal index stuff */
754
755 struct list_head sd_jindex_list;
756 spinlock_t sd_jindex_spin;
757 struct mutex sd_jindex_mutex;
758 unsigned int sd_journals;
759
760 struct gfs2_jdesc *sd_jdesc;
761 struct gfs2_holder sd_journal_gh;
762 struct gfs2_holder sd_jinode_gh;
763 struct gfs2_glock *sd_jinode_gl;
764
765 struct gfs2_holder sd_sc_gh;
766 struct buffer_head *sd_sc_bh;
767 struct gfs2_holder sd_qc_gh;
768
769 struct completion sd_journal_ready;
770
771 /* Workqueue stuff */
772
773 struct workqueue_struct *sd_glock_wq;
774 struct workqueue_struct *sd_delete_wq;
775
776 /* Daemon stuff */
777
778 struct task_struct *sd_logd_process;
779 struct task_struct *sd_quotad_process;
780
781 /* Quota stuff */
782
783 struct list_head sd_quota_list;
784 atomic_t sd_quota_count;
785 struct mutex sd_quota_sync_mutex;
786 wait_queue_head_t sd_quota_wait;
787
788 unsigned int sd_quota_slots;
789 unsigned long *sd_quota_bitmap;
790 spinlock_t sd_bitmap_lock;
791
792 u64 sd_quota_sync_gen;
793
794 /* Log stuff */
795
796 struct address_space sd_aspace;
797
798 spinlock_t sd_log_lock;
799
800 struct gfs2_trans *sd_log_tr;
801 unsigned int sd_log_blks_reserved;
802
803 atomic_t sd_log_pinned;
804 unsigned int sd_log_num_revoke;
805
806 struct list_head sd_log_revokes;
807 struct list_head sd_log_ordered;
808 spinlock_t sd_ordered_lock;
809
810 atomic_t sd_log_thresh1;
811 atomic_t sd_log_thresh2;
812 atomic_t sd_log_blks_free;
813 atomic_t sd_log_blks_needed;
814 atomic_t sd_log_revokes_available;
815 wait_queue_head_t sd_log_waitq;
816 wait_queue_head_t sd_logd_waitq;
817
818 u64 sd_log_sequence;
819 int sd_log_idle;
820
821 struct rw_semaphore sd_log_flush_lock;
822 atomic_t sd_log_in_flight;
823 wait_queue_head_t sd_log_flush_wait;
824 int sd_log_error; /* First log error */
825 wait_queue_head_t sd_withdraw_wait;
826
827 unsigned int sd_log_tail;
828 unsigned int sd_log_flush_tail;
829 unsigned int sd_log_head;
830 unsigned int sd_log_flush_head;
831
832 spinlock_t sd_ail_lock;
833 struct list_head sd_ail1_list;
834 struct list_head sd_ail2_list;
835
836 /* For quiescing the filesystem */
837 struct gfs2_holder sd_freeze_gh;
838 struct mutex sd_freeze_mutex;
839 struct list_head sd_dead_glocks;
840
841 char sd_fsname[GFS2_FSNAME_LEN + 3 * sizeof(int) + 2];
842 char sd_table_name[GFS2_FSNAME_LEN];
843 char sd_proto_name[GFS2_FSNAME_LEN];
844
845 /* Debugging crud */
846
847 unsigned long sd_last_warning;
848 struct dentry *debugfs_dir; /* debugfs directory */
849 unsigned long sd_glock_dqs_held;
850};
851
852static inline void gfs2_glstats_inc(struct gfs2_glock *gl, int which)
853{
854 gl->gl_stats.stats[which]++;
855}
856
857static inline void gfs2_sbstats_inc(const struct gfs2_glock *gl, int which)
858{
859 const struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
860 preempt_disable();
861 this_cpu_ptr(sdp->sd_lkstats)->lkstats[gl->gl_name.ln_type].stats[which]++;
862 preempt_enable();
863}
864
865struct gfs2_rgrpd *gfs2_glock2rgrp(struct gfs2_glock *gl);
866
867static inline unsigned gfs2_max_stuffed_size(const struct gfs2_inode *ip)
868{
869 return GFS2_SB(&ip->i_inode)->sd_sb.sb_bsize - sizeof(struct gfs2_dinode);
870}
871
872#endif /* __INCORE_DOT_H__ */
873