Loading...
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_demote_ok) (const struct gfs2_glock *gl);
222 int (*go_instantiate) (struct gfs2_glock *gl);
223 int (*go_held)(struct gfs2_holder *gh);
224 void (*go_dump)(struct seq_file *seq, struct gfs2_glock *gl,
225 const char *fs_id_buf);
226 void (*go_callback)(struct gfs2_glock *gl, bool remote);
227 void (*go_free)(struct gfs2_glock *gl);
228 const int go_subclass;
229 const int go_type;
230 const unsigned long go_flags;
231#define GLOF_ASPACE 1 /* address space attached */
232#define GLOF_LVB 2 /* Lock Value Block attached */
233#define GLOF_LRU 4 /* LRU managed */
234#define GLOF_NONDISK 8 /* not I/O related */
235};
236
237enum {
238 GFS2_LKS_SRTT = 0, /* Non blocking smoothed round trip time */
239 GFS2_LKS_SRTTVAR = 1, /* Non blocking smoothed variance */
240 GFS2_LKS_SRTTB = 2, /* Blocking smoothed round trip time */
241 GFS2_LKS_SRTTVARB = 3, /* Blocking smoothed variance */
242 GFS2_LKS_SIRT = 4, /* Smoothed Inter-request time */
243 GFS2_LKS_SIRTVAR = 5, /* Smoothed Inter-request variance */
244 GFS2_LKS_DCOUNT = 6, /* Count of dlm requests */
245 GFS2_LKS_QCOUNT = 7, /* Count of gfs2_holder queues */
246 GFS2_NR_LKSTATS
247};
248
249struct gfs2_lkstats {
250 u64 stats[GFS2_NR_LKSTATS];
251};
252
253enum {
254 /* States */
255 HIF_HOLDER = 6, /* Set for gh that "holds" the glock */
256 HIF_WAIT = 10,
257};
258
259struct gfs2_holder {
260 struct list_head gh_list;
261
262 struct gfs2_glock *gh_gl;
263 struct pid *gh_owner_pid;
264 u16 gh_flags;
265 u16 gh_state;
266
267 int gh_error;
268 unsigned long gh_iflags; /* HIF_... */
269 unsigned long gh_ip;
270};
271
272/* Number of quota types we support */
273#define GFS2_MAXQUOTAS 2
274
275struct gfs2_qadata { /* quota allocation data */
276 /* Quota stuff */
277 struct gfs2_quota_data *qa_qd[2 * GFS2_MAXQUOTAS];
278 struct gfs2_holder qa_qd_ghs[2 * GFS2_MAXQUOTAS];
279 unsigned int qa_qd_num;
280 int qa_ref;
281};
282
283/* Resource group multi-block reservation, in order of appearance:
284
285 Step 1. Function prepares to write, allocates a mb, sets the size hint.
286 Step 2. User calls inplace_reserve to target an rgrp, sets the rgrp info
287 Step 3. Function get_local_rgrp locks the rgrp, determines which bits to use
288 Step 4. Bits are assigned from the rgrp based on either the reservation
289 or wherever it can.
290*/
291
292struct gfs2_blkreserv {
293 struct rb_node rs_node; /* node within rd_rstree */
294 struct gfs2_rgrpd *rs_rgd;
295 u64 rs_start;
296 u32 rs_requested;
297 u32 rs_reserved; /* number of reserved blocks */
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_INSTANTIATE_NEEDED = 2, /* needs instantiate */
319 GLF_DEMOTE = 3,
320 GLF_PENDING_DEMOTE = 4,
321 GLF_DEMOTE_IN_PROGRESS = 5,
322 GLF_DIRTY = 6,
323 GLF_LFLUSH = 7,
324 GLF_INVALIDATE_IN_PROGRESS = 8,
325 GLF_REPLY_PENDING = 9,
326 GLF_INITIAL = 10,
327 GLF_FROZEN = 11,
328 GLF_INSTANTIATE_IN_PROG = 12, /* instantiate happening now */
329 GLF_LRU = 13,
330 GLF_OBJECT = 14, /* Used only for tracing */
331 GLF_BLOCKING = 15,
332 GLF_PENDING_DELETE = 17,
333 GLF_FREEING = 18, /* Wait for glock to be freed */
334};
335
336struct gfs2_glock {
337 unsigned long gl_flags; /* GLF_... */
338 struct lm_lockname gl_name;
339
340 struct lockref gl_lockref;
341
342 /* State fields protected by gl_lockref.lock */
343 unsigned int gl_state:2, /* Current state */
344 gl_target:2, /* Target state */
345 gl_demote_state:2, /* State requested by remote node */
346 gl_req:2, /* State in last dlm request */
347 gl_reply:8; /* Last reply from the dlm */
348
349 unsigned long gl_demote_time; /* time of first demote request */
350 long gl_hold_time;
351 struct list_head gl_holders;
352
353 const struct gfs2_glock_operations *gl_ops;
354 ktime_t gl_dstamp;
355 struct gfs2_lkstats gl_stats;
356 struct dlm_lksb gl_lksb;
357 unsigned long gl_tchange;
358 void *gl_object;
359
360 struct list_head gl_lru;
361 struct list_head gl_ail_list;
362 atomic_t gl_ail_count;
363 atomic_t gl_revokes;
364 struct delayed_work gl_work;
365 /* For iopen glocks only */
366 struct {
367 struct delayed_work gl_delete;
368 u64 gl_no_formal_ino;
369 };
370 struct rcu_head gl_rcu;
371 struct rhash_head gl_node;
372};
373
374enum {
375 GIF_QD_LOCKED = 1,
376 GIF_ALLOC_FAILED = 2,
377 GIF_SW_PAGED = 3,
378 GIF_FREE_VFS_INODE = 5,
379 GIF_GLOP_PENDING = 6,
380 GIF_DEFERRED_DELETE = 7,
381};
382
383struct gfs2_inode {
384 struct inode i_inode;
385 u64 i_no_addr;
386 u64 i_no_formal_ino;
387 u64 i_generation;
388 u64 i_eattr;
389 unsigned long i_flags; /* GIF_... */
390 struct gfs2_glock *i_gl;
391 struct gfs2_holder i_iopen_gh;
392 struct gfs2_qadata *i_qadata; /* quota allocation data */
393 struct gfs2_holder i_rgd_gh;
394 struct gfs2_blkreserv i_res; /* rgrp multi-block reservation */
395 u64 i_goal; /* goal block for allocations */
396 atomic_t i_sizehint; /* hint of the write size */
397 struct rw_semaphore i_rw_mutex;
398 struct list_head i_ordered;
399 __be64 *i_hash_cache;
400 u32 i_entries;
401 u32 i_diskflags;
402 u8 i_height;
403 u8 i_depth;
404 u16 i_rahead;
405};
406
407/*
408 * Since i_inode is the first element of struct gfs2_inode,
409 * this is effectively a cast.
410 */
411static inline struct gfs2_inode *GFS2_I(struct inode *inode)
412{
413 return container_of(inode, struct gfs2_inode, i_inode);
414}
415
416static inline struct gfs2_sbd *GFS2_SB(const struct inode *inode)
417{
418 return inode->i_sb->s_fs_info;
419}
420
421struct gfs2_file {
422 struct mutex f_fl_mutex;
423 struct gfs2_holder f_fl_gh;
424};
425
426struct gfs2_revoke_replay {
427 struct list_head rr_list;
428 u64 rr_blkno;
429 unsigned int rr_where;
430};
431
432enum {
433 QDF_CHANGE = 1,
434 QDF_LOCKED = 2,
435 QDF_REFRESH = 3,
436 QDF_QMSG_QUIET = 4,
437};
438
439struct gfs2_quota_data {
440 struct hlist_bl_node qd_hlist;
441 struct list_head qd_list;
442 struct kqid qd_id;
443 struct gfs2_sbd *qd_sbd;
444 struct lockref qd_lockref;
445 struct list_head qd_lru;
446 unsigned qd_hash;
447
448 unsigned long qd_flags; /* QDF_... */
449
450 s64 qd_change;
451 s64 qd_change_sync;
452
453 unsigned int qd_slot;
454 unsigned int qd_slot_count;
455
456 struct buffer_head *qd_bh;
457 struct gfs2_quota_change *qd_bh_qc;
458 unsigned int qd_bh_count;
459
460 struct gfs2_glock *qd_gl;
461 struct gfs2_quota_lvb qd_qb;
462
463 u64 qd_sync_gen;
464 unsigned long qd_last_warn;
465 struct rcu_head qd_rcu;
466};
467
468enum {
469 TR_TOUCHED = 1,
470 TR_ATTACHED = 2,
471 TR_ONSTACK = 3,
472};
473
474struct gfs2_trans {
475 unsigned long tr_ip;
476
477 unsigned int tr_blocks;
478 unsigned int tr_revokes;
479 unsigned int tr_reserved;
480 unsigned long tr_flags;
481
482 unsigned int tr_num_buf_new;
483 unsigned int tr_num_databuf_new;
484 unsigned int tr_num_buf_rm;
485 unsigned int tr_num_databuf_rm;
486 unsigned int tr_num_revoke;
487
488 struct list_head tr_list;
489 struct list_head tr_databuf;
490 struct list_head tr_buf;
491
492 unsigned int tr_first;
493 struct list_head tr_ail1_list;
494 struct list_head tr_ail2_list;
495};
496
497struct gfs2_journal_extent {
498 struct list_head list;
499
500 unsigned int lblock; /* First logical block */
501 u64 dblock; /* First disk block */
502 u64 blocks;
503};
504
505struct gfs2_jdesc {
506 struct list_head jd_list;
507 struct list_head extent_list;
508 unsigned int nr_extents;
509 struct work_struct jd_work;
510 struct inode *jd_inode;
511 struct bio *jd_log_bio;
512 unsigned long jd_flags;
513#define JDF_RECOVERY 1
514 unsigned int jd_jid;
515 u32 jd_blocks;
516 int jd_recover_error;
517 /* Replay stuff */
518
519 unsigned int jd_found_blocks;
520 unsigned int jd_found_revokes;
521 unsigned int jd_replayed_blocks;
522
523 struct list_head jd_revoke_list;
524 unsigned int jd_replay_tail;
525
526 u64 jd_no_addr;
527};
528
529struct gfs2_statfs_change_host {
530 s64 sc_total;
531 s64 sc_free;
532 s64 sc_dinodes;
533};
534
535#define GFS2_QUOTA_DEFAULT GFS2_QUOTA_OFF
536#define GFS2_QUOTA_OFF 0
537#define GFS2_QUOTA_ACCOUNT 1
538#define GFS2_QUOTA_ON 2
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_FS_FROZEN = 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};
609
610enum gfs2_freeze_state {
611 SFS_UNFROZEN = 0,
612 SFS_STARTING_FREEZE = 1,
613 SFS_FROZEN = 2,
614};
615
616#define GFS2_FSNAME_LEN 256
617
618struct gfs2_inum_host {
619 u64 no_formal_ino;
620 u64 no_addr;
621};
622
623struct gfs2_sb_host {
624 u32 sb_magic;
625 u32 sb_type;
626
627 u32 sb_fs_format;
628 u32 sb_multihost_format;
629 u32 sb_bsize;
630 u32 sb_bsize_shift;
631
632 struct gfs2_inum_host sb_master_dir;
633 struct gfs2_inum_host sb_root_dir;
634
635 char sb_lockproto[GFS2_LOCKNAME_LEN];
636 char sb_locktable[GFS2_LOCKNAME_LEN];
637};
638
639/*
640 * lm_mount() return values
641 *
642 * ls_jid - the journal ID this node should use
643 * ls_first - this node is the first to mount the file system
644 * ls_lockspace - lock module's context for this file system
645 * ls_ops - lock module's functions
646 */
647
648struct lm_lockstruct {
649 int ls_jid;
650 unsigned int ls_first;
651 const struct lm_lockops *ls_ops;
652 dlm_lockspace_t *ls_dlm;
653
654 int ls_recover_jid_done; /* These two are deprecated, */
655 int ls_recover_jid_status; /* used previously by gfs_controld */
656
657 struct dlm_lksb ls_mounted_lksb; /* mounted_lock */
658 struct dlm_lksb ls_control_lksb; /* control_lock */
659 char ls_control_lvb[GDLM_LVB_SIZE]; /* control_lock lvb */
660 struct completion ls_sync_wait; /* {control,mounted}_{lock,unlock} */
661 char *ls_lvb_bits;
662
663 spinlock_t ls_recover_spin; /* protects following fields */
664 unsigned long ls_recover_flags; /* DFL_ */
665 uint32_t ls_recover_mount; /* gen in first recover_done cb */
666 uint32_t ls_recover_start; /* gen in last recover_done cb */
667 uint32_t ls_recover_block; /* copy recover_start in last recover_prep */
668 uint32_t ls_recover_size; /* size of recover_submit, recover_result */
669 uint32_t *ls_recover_submit; /* gen in last recover_slot cb per jid */
670 uint32_t *ls_recover_result; /* result of last jid recovery */
671};
672
673struct gfs2_pcpu_lkstats {
674 /* One struct for each glock type */
675 struct gfs2_lkstats lkstats[10];
676};
677
678/* List of local (per node) statfs inodes */
679struct local_statfs_inode {
680 struct list_head si_list;
681 struct inode *si_sc_inode;
682 unsigned int si_jid; /* journal id this statfs inode corresponds to */
683};
684
685struct gfs2_sbd {
686 struct super_block *sd_vfs;
687 struct gfs2_pcpu_lkstats __percpu *sd_lkstats;
688 struct kobject sd_kobj;
689 struct completion sd_kobj_unregister;
690 unsigned long sd_flags; /* SDF_... */
691 struct gfs2_sb_host sd_sb;
692
693 /* Constants computed on mount */
694
695 u32 sd_fsb2bb;
696 u32 sd_fsb2bb_shift;
697 u32 sd_diptrs; /* Number of pointers in a dinode */
698 u32 sd_inptrs; /* Number of pointers in a indirect block */
699 u32 sd_ldptrs; /* Number of pointers in a log descriptor block */
700 u32 sd_jbsize; /* Size of a journaled data block */
701 u32 sd_hash_bsize; /* sizeof(exhash block) */
702 u32 sd_hash_bsize_shift;
703 u32 sd_hash_ptrs; /* Number of pointers in a hash block */
704 u32 sd_qc_per_block;
705 u32 sd_blocks_per_bitmap;
706 u32 sd_max_dirres; /* Max blocks needed to add a directory entry */
707 u32 sd_max_height; /* Max height of a file's metadata tree */
708 u64 sd_heightsize[GFS2_MAX_META_HEIGHT + 1];
709 u32 sd_max_dents_per_leaf; /* Max number of dirents in a leaf block */
710
711 struct gfs2_args sd_args; /* Mount arguments */
712 struct gfs2_tune sd_tune; /* Filesystem tuning structure */
713
714 /* Lock Stuff */
715
716 struct lm_lockstruct sd_lockstruct;
717 struct gfs2_holder sd_live_gh;
718 struct gfs2_glock *sd_rename_gl;
719 struct gfs2_glock *sd_freeze_gl;
720 struct work_struct sd_freeze_work;
721 wait_queue_head_t sd_glock_wait;
722 wait_queue_head_t sd_async_glock_wait;
723 atomic_t sd_glock_disposal;
724 struct completion sd_locking_init;
725 struct completion sd_wdack;
726 struct delayed_work sd_control_work;
727
728 /* Inode Stuff */
729
730 struct dentry *sd_master_dir;
731 struct dentry *sd_root_dir;
732
733 struct inode *sd_jindex;
734 struct inode *sd_statfs_inode;
735 struct inode *sd_sc_inode;
736 struct list_head sd_sc_inodes_list;
737 struct inode *sd_qc_inode;
738 struct inode *sd_rindex;
739 struct inode *sd_quota_inode;
740
741 /* StatFS stuff */
742
743 spinlock_t sd_statfs_spin;
744 struct gfs2_statfs_change_host sd_statfs_master;
745 struct gfs2_statfs_change_host sd_statfs_local;
746 int sd_statfs_force_sync;
747
748 /* Resource group stuff */
749
750 int sd_rindex_uptodate;
751 spinlock_t sd_rindex_spin;
752 struct rb_root sd_rindex_tree;
753 unsigned int sd_rgrps;
754 unsigned int sd_max_rg_data;
755
756 /* Journal index stuff */
757
758 struct list_head sd_jindex_list;
759 spinlock_t sd_jindex_spin;
760 struct mutex sd_jindex_mutex;
761 unsigned int sd_journals;
762
763 struct gfs2_jdesc *sd_jdesc;
764 struct gfs2_holder sd_journal_gh;
765 struct gfs2_holder sd_jinode_gh;
766 struct gfs2_glock *sd_jinode_gl;
767
768 struct gfs2_holder sd_sc_gh;
769 struct buffer_head *sd_sc_bh;
770 struct gfs2_holder sd_qc_gh;
771
772 struct completion sd_journal_ready;
773
774 /* Daemon stuff */
775
776 struct task_struct *sd_logd_process;
777 struct task_struct *sd_quotad_process;
778
779 /* Quota stuff */
780
781 struct list_head sd_quota_list;
782 atomic_t sd_quota_count;
783 struct mutex sd_quota_mutex;
784 struct mutex sd_quota_sync_mutex;
785 wait_queue_head_t sd_quota_wait;
786
787 unsigned int sd_quota_slots;
788 unsigned long *sd_quota_bitmap;
789 spinlock_t sd_bitmap_lock;
790
791 u64 sd_quota_sync_gen;
792
793 /* Log stuff */
794
795 struct address_space sd_aspace;
796
797 spinlock_t sd_log_lock;
798
799 struct gfs2_trans *sd_log_tr;
800 unsigned int sd_log_blks_reserved;
801
802 atomic_t sd_log_pinned;
803 unsigned int sd_log_num_revoke;
804
805 struct list_head sd_log_revokes;
806 struct list_head sd_log_ordered;
807 spinlock_t sd_ordered_lock;
808
809 atomic_t sd_log_thresh1;
810 atomic_t sd_log_thresh2;
811 atomic_t sd_log_blks_free;
812 atomic_t sd_log_blks_needed;
813 atomic_t sd_log_revokes_available;
814 wait_queue_head_t sd_log_waitq;
815 wait_queue_head_t sd_logd_waitq;
816
817 u64 sd_log_sequence;
818 int sd_log_idle;
819
820 struct rw_semaphore sd_log_flush_lock;
821 atomic_t sd_log_in_flight;
822 wait_queue_head_t sd_log_flush_wait;
823 int sd_log_error; /* First log error */
824 wait_queue_head_t sd_withdraw_wait;
825
826 unsigned int sd_log_tail;
827 unsigned int sd_log_flush_tail;
828 unsigned int sd_log_head;
829 unsigned int sd_log_flush_head;
830
831 spinlock_t sd_ail_lock;
832 struct list_head sd_ail1_list;
833 struct list_head sd_ail2_list;
834
835 /* For quiescing the filesystem */
836 struct gfs2_holder sd_freeze_gh;
837 atomic_t sd_freeze_state;
838 struct mutex sd_freeze_mutex;
839
840 char sd_fsname[GFS2_FSNAME_LEN + 3 * sizeof(int) + 2];
841 char sd_table_name[GFS2_FSNAME_LEN];
842 char sd_proto_name[GFS2_FSNAME_LEN];
843
844 /* Debugging crud */
845
846 unsigned long sd_last_warning;
847 struct dentry *debugfs_dir; /* debugfs directory */
848 unsigned long sd_glock_dqs_held;
849};
850
851static inline void gfs2_glstats_inc(struct gfs2_glock *gl, int which)
852{
853 gl->gl_stats.stats[which]++;
854}
855
856static inline void gfs2_sbstats_inc(const struct gfs2_glock *gl, int which)
857{
858 const struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
859 preempt_disable();
860 this_cpu_ptr(sdp->sd_lkstats)->lkstats[gl->gl_name.ln_type].stats[which]++;
861 preempt_enable();
862}
863
864extern struct gfs2_rgrpd *gfs2_glock2rgrp(struct gfs2_glock *gl);
865
866static inline unsigned gfs2_max_stuffed_size(const struct gfs2_inode *ip)
867{
868 return GFS2_SB(&ip->i_inode)->sd_sb.sb_bsize - sizeof(struct gfs2_dinode);
869}
870
871#endif /* __INCORE_DOT_H__ */
872
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_demote_ok) (const struct gfs2_glock *gl);
222 int (*go_instantiate) (struct gfs2_glock *gl);
223 int (*go_held)(struct gfs2_holder *gh);
224 void (*go_dump)(struct seq_file *seq, const struct gfs2_glock *gl,
225 const char *fs_id_buf);
226 void (*go_callback)(struct gfs2_glock *gl, bool remote);
227 void (*go_free)(struct gfs2_glock *gl);
228 const int go_subclass;
229 const int go_type;
230 const unsigned long go_flags;
231#define GLOF_ASPACE 1 /* address space attached */
232#define GLOF_LVB 2 /* Lock Value Block attached */
233#define GLOF_LRU 4 /* LRU managed */
234#define GLOF_NONDISK 8 /* not I/O related */
235};
236
237enum {
238 GFS2_LKS_SRTT = 0, /* Non blocking smoothed round trip time */
239 GFS2_LKS_SRTTVAR = 1, /* Non blocking smoothed variance */
240 GFS2_LKS_SRTTB = 2, /* Blocking smoothed round trip time */
241 GFS2_LKS_SRTTVARB = 3, /* Blocking smoothed variance */
242 GFS2_LKS_SIRT = 4, /* Smoothed Inter-request time */
243 GFS2_LKS_SIRTVAR = 5, /* Smoothed Inter-request variance */
244 GFS2_LKS_DCOUNT = 6, /* Count of dlm requests */
245 GFS2_LKS_QCOUNT = 7, /* Count of gfs2_holder queues */
246 GFS2_NR_LKSTATS
247};
248
249struct gfs2_lkstats {
250 u64 stats[GFS2_NR_LKSTATS];
251};
252
253enum {
254 /* States */
255 HIF_HOLDER = 6, /* Set for gh that "holds" the glock */
256 HIF_WAIT = 10,
257};
258
259struct gfs2_holder {
260 struct list_head gh_list;
261
262 struct gfs2_glock *gh_gl;
263 struct pid *gh_owner_pid;
264 u16 gh_flags;
265 u16 gh_state;
266
267 int gh_error;
268 unsigned long gh_iflags; /* HIF_... */
269 unsigned long gh_ip;
270};
271
272/* Number of quota types we support */
273#define GFS2_MAXQUOTAS 2
274
275struct gfs2_qadata { /* quota allocation data */
276 /* Quota stuff */
277 struct gfs2_quota_data *qa_qd[2 * GFS2_MAXQUOTAS];
278 struct gfs2_holder qa_qd_ghs[2 * GFS2_MAXQUOTAS];
279 unsigned int qa_qd_num;
280 int qa_ref;
281};
282
283/* Resource group multi-block reservation, in order of appearance:
284
285 Step 1. Function prepares to write, allocates a mb, sets the size hint.
286 Step 2. User calls inplace_reserve to target an rgrp, sets the rgrp info
287 Step 3. Function get_local_rgrp locks the rgrp, determines which bits to use
288 Step 4. Bits are assigned from the rgrp based on either the reservation
289 or wherever it can.
290*/
291
292struct gfs2_blkreserv {
293 struct rb_node rs_node; /* node within rd_rstree */
294 struct gfs2_rgrpd *rs_rgd;
295 u64 rs_start;
296 u32 rs_requested;
297 u32 rs_reserved; /* number of reserved blocks */
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_INSTANTIATE_NEEDED = 2, /* needs instantiate */
319 GLF_DEMOTE = 3,
320 GLF_PENDING_DEMOTE = 4,
321 GLF_DEMOTE_IN_PROGRESS = 5,
322 GLF_DIRTY = 6,
323 GLF_LFLUSH = 7,
324 GLF_INVALIDATE_IN_PROGRESS = 8,
325 GLF_REPLY_PENDING = 9,
326 GLF_INITIAL = 10,
327 GLF_FROZEN = 11,
328 GLF_INSTANTIATE_IN_PROG = 12, /* instantiate happening now */
329 GLF_LRU = 13,
330 GLF_OBJECT = 14, /* Used only for tracing */
331 GLF_BLOCKING = 15,
332 GLF_FREEING = 16, /* Wait for glock to be freed */
333 GLF_TRY_TO_EVICT = 17, /* iopen glocks only */
334 GLF_VERIFY_EVICT = 18, /* iopen glocks only */
335};
336
337struct gfs2_glock {
338 unsigned long gl_flags; /* GLF_... */
339 struct lm_lockname gl_name;
340
341 struct lockref gl_lockref;
342
343 /* State fields protected by gl_lockref.lock */
344 unsigned int gl_state:2, /* Current state */
345 gl_target:2, /* Target state */
346 gl_demote_state:2, /* State requested by remote node */
347 gl_req:2, /* State in last dlm request */
348 gl_reply:8; /* Last reply from the dlm */
349
350 unsigned long gl_demote_time; /* time of first demote request */
351 long gl_hold_time;
352 struct list_head gl_holders;
353
354 const struct gfs2_glock_operations *gl_ops;
355 ktime_t gl_dstamp;
356 struct gfs2_lkstats gl_stats;
357 struct dlm_lksb gl_lksb;
358 unsigned long gl_tchange;
359 void *gl_object;
360
361 struct list_head gl_lru;
362 struct list_head gl_ail_list;
363 atomic_t gl_ail_count;
364 atomic_t gl_revokes;
365 struct delayed_work gl_work;
366 /* For iopen glocks only */
367 struct {
368 struct delayed_work gl_delete;
369 u64 gl_no_formal_ino;
370 };
371 struct rcu_head gl_rcu;
372 struct rhash_head gl_node;
373};
374
375enum {
376 GIF_QD_LOCKED = 1,
377 GIF_ALLOC_FAILED = 2,
378 GIF_SW_PAGED = 3,
379 GIF_FREE_VFS_INODE = 5,
380 GIF_GLOP_PENDING = 6,
381 GIF_DEFERRED_DELETE = 7,
382};
383
384struct gfs2_inode {
385 struct inode i_inode;
386 u64 i_no_addr;
387 u64 i_no_formal_ino;
388 u64 i_generation;
389 u64 i_eattr;
390 unsigned long i_flags; /* GIF_... */
391 struct gfs2_glock *i_gl;
392 struct gfs2_holder i_iopen_gh;
393 struct gfs2_qadata *i_qadata; /* quota allocation data */
394 struct gfs2_holder i_rgd_gh;
395 struct gfs2_blkreserv i_res; /* rgrp multi-block reservation */
396 u64 i_goal; /* goal block for allocations */
397 atomic_t i_sizehint; /* hint of the write size */
398 struct rw_semaphore i_rw_mutex;
399 struct list_head i_ordered;
400 __be64 *i_hash_cache;
401 u32 i_entries;
402 u32 i_diskflags;
403 u8 i_height;
404 u8 i_depth;
405 u16 i_rahead;
406};
407
408/*
409 * Since i_inode is the first element of struct gfs2_inode,
410 * this is effectively a cast.
411 */
412static inline struct gfs2_inode *GFS2_I(struct inode *inode)
413{
414 return container_of(inode, struct gfs2_inode, i_inode);
415}
416
417static inline struct gfs2_sbd *GFS2_SB(const struct inode *inode)
418{
419 return inode->i_sb->s_fs_info;
420}
421
422struct gfs2_file {
423 struct mutex f_fl_mutex;
424 struct gfs2_holder f_fl_gh;
425};
426
427struct gfs2_revoke_replay {
428 struct list_head rr_list;
429 u64 rr_blkno;
430 unsigned int rr_where;
431};
432
433enum {
434 QDF_CHANGE = 1,
435 QDF_LOCKED = 2,
436 QDF_REFRESH = 3,
437 QDF_QMSG_QUIET = 4,
438};
439
440struct gfs2_quota_data {
441 struct hlist_bl_node qd_hlist;
442 struct list_head qd_list;
443 struct kqid qd_id;
444 struct gfs2_sbd *qd_sbd;
445 struct lockref qd_lockref;
446 struct list_head qd_lru;
447 unsigned qd_hash;
448
449 unsigned long qd_flags; /* QDF_... */
450
451 s64 qd_change;
452 s64 qd_change_sync;
453
454 unsigned int qd_slot;
455 unsigned int qd_slot_ref;
456
457 struct buffer_head *qd_bh;
458 struct gfs2_quota_change *qd_bh_qc;
459 unsigned int qd_bh_count;
460
461 struct gfs2_glock *qd_gl;
462 struct gfs2_quota_lvb qd_qb;
463
464 u64 qd_sync_gen;
465 unsigned long qd_last_warn;
466 struct rcu_head qd_rcu;
467};
468
469enum {
470 TR_TOUCHED = 1,
471 TR_ATTACHED = 2,
472 TR_ONSTACK = 3,
473};
474
475struct gfs2_trans {
476 unsigned long tr_ip;
477
478 unsigned int tr_blocks;
479 unsigned int tr_revokes;
480 unsigned int tr_reserved;
481 unsigned long tr_flags;
482
483 unsigned int tr_num_buf_new;
484 unsigned int tr_num_databuf_new;
485 unsigned int tr_num_buf_rm;
486 unsigned int tr_num_databuf_rm;
487 unsigned int tr_num_revoke;
488
489 struct list_head tr_list;
490 struct list_head tr_databuf;
491 struct list_head tr_buf;
492
493 unsigned int tr_first;
494 struct list_head tr_ail1_list;
495 struct list_head tr_ail2_list;
496};
497
498struct gfs2_journal_extent {
499 struct list_head list;
500
501 unsigned int lblock; /* First logical block */
502 u64 dblock; /* First disk block */
503 u64 blocks;
504};
505
506struct gfs2_jdesc {
507 struct list_head jd_list;
508 struct list_head extent_list;
509 unsigned int nr_extents;
510 struct work_struct jd_work;
511 struct inode *jd_inode;
512 struct bio *jd_log_bio;
513 unsigned long jd_flags;
514#define JDF_RECOVERY 1
515 unsigned int jd_jid;
516 u32 jd_blocks;
517 int jd_recover_error;
518 /* Replay stuff */
519
520 unsigned int jd_found_blocks;
521 unsigned int jd_found_revokes;
522 unsigned int jd_replayed_blocks;
523
524 struct list_head jd_revoke_list;
525 unsigned int jd_replay_tail;
526
527 u64 jd_no_addr;
528};
529
530struct gfs2_statfs_change_host {
531 s64 sc_total;
532 s64 sc_free;
533 s64 sc_dinodes;
534};
535
536#define GFS2_QUOTA_DEFAULT GFS2_QUOTA_OFF
537#define GFS2_QUOTA_OFF 0
538#define GFS2_QUOTA_ACCOUNT 1
539#define GFS2_QUOTA_ON 2
540#define GFS2_QUOTA_QUIET 3 /* on but not complaining */
541
542#define GFS2_DATA_DEFAULT GFS2_DATA_ORDERED
543#define GFS2_DATA_WRITEBACK 1
544#define GFS2_DATA_ORDERED 2
545
546#define GFS2_ERRORS_DEFAULT GFS2_ERRORS_WITHDRAW
547#define GFS2_ERRORS_WITHDRAW 0
548#define GFS2_ERRORS_CONTINUE 1 /* place holder for future feature */
549#define GFS2_ERRORS_RO 2 /* place holder for future feature */
550#define GFS2_ERRORS_PANIC 3
551
552struct gfs2_args {
553 char ar_lockproto[GFS2_LOCKNAME_LEN]; /* Name of the Lock Protocol */
554 char ar_locktable[GFS2_LOCKNAME_LEN]; /* Name of the Lock Table */
555 char ar_hostdata[GFS2_LOCKNAME_LEN]; /* Host specific data */
556 unsigned int ar_spectator:1; /* Don't get a journal */
557 unsigned int ar_localflocks:1; /* Let the VFS do flock|fcntl */
558 unsigned int ar_debug:1; /* Oops on errors */
559 unsigned int ar_posix_acl:1; /* Enable posix acls */
560 unsigned int ar_quota:2; /* off/account/on */
561 unsigned int ar_suiddir:1; /* suiddir support */
562 unsigned int ar_data:2; /* ordered/writeback */
563 unsigned int ar_meta:1; /* mount metafs */
564 unsigned int ar_discard:1; /* discard requests */
565 unsigned int ar_errors:2; /* errors=withdraw | panic */
566 unsigned int ar_nobarrier:1; /* do not send barriers */
567 unsigned int ar_rgrplvb:1; /* use lvbs for rgrp info */
568 unsigned int ar_got_rgrplvb:1; /* Was the rgrplvb opt given? */
569 unsigned int ar_loccookie:1; /* use location based readdir
570 cookies */
571 s32 ar_commit; /* Commit interval */
572 s32 ar_statfs_quantum; /* The fast statfs interval */
573 s32 ar_quota_quantum; /* The quota interval */
574 s32 ar_statfs_percent; /* The % change to force sync */
575};
576
577struct gfs2_tune {
578 spinlock_t gt_spin;
579
580 unsigned int gt_logd_secs;
581
582 unsigned int gt_quota_warn_period; /* Secs between quota warn msgs */
583 unsigned int gt_quota_scale_num; /* Numerator */
584 unsigned int gt_quota_scale_den; /* Denominator */
585 unsigned int gt_quota_quantum; /* Secs between syncs to quota file */
586 unsigned int gt_new_files_jdata;
587 unsigned int gt_max_readahead; /* Max bytes to read-ahead from disk */
588 unsigned int gt_complain_secs;
589 unsigned int gt_statfs_quantum;
590 unsigned int gt_statfs_slow;
591};
592
593enum {
594 SDF_JOURNAL_CHECKED = 0,
595 SDF_JOURNAL_LIVE = 1,
596 SDF_WITHDRAWN = 2,
597 SDF_NOBARRIERS = 3,
598 SDF_NORECOVERY = 4,
599 SDF_DEMOTE = 5,
600 SDF_NOJOURNALID = 6,
601 SDF_RORECOVERY = 7, /* read only recovery */
602 SDF_SKIP_DLM_UNLOCK = 8,
603 SDF_FORCE_AIL_FLUSH = 9,
604 SDF_FREEZE_INITIATOR = 10,
605 SDF_WITHDRAWING = 11, /* Will withdraw eventually */
606 SDF_WITHDRAW_IN_PROG = 12, /* Withdraw is in progress */
607 SDF_REMOTE_WITHDRAW = 13, /* Performing remote recovery */
608 SDF_WITHDRAW_RECOVERY = 14, /* Wait for journal recovery when we are
609 withdrawing */
610 SDF_KILL = 15,
611 SDF_EVICTING = 16,
612 SDF_FROZEN = 17,
613};
614
615#define GFS2_FSNAME_LEN 256
616
617struct gfs2_inum_host {
618 u64 no_formal_ino;
619 u64 no_addr;
620};
621
622struct gfs2_sb_host {
623 u32 sb_magic;
624 u32 sb_type;
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
677/* List of local (per node) statfs inodes */
678struct local_statfs_inode {
679 struct list_head si_list;
680 struct inode *si_sc_inode;
681 unsigned int si_jid; /* journal id this statfs inode corresponds to */
682};
683
684struct gfs2_sbd {
685 struct super_block *sd_vfs;
686 struct gfs2_pcpu_lkstats __percpu *sd_lkstats;
687 struct kobject sd_kobj;
688 struct completion sd_kobj_unregister;
689 unsigned long sd_flags; /* SDF_... */
690 struct gfs2_sb_host sd_sb;
691
692 /* Constants computed on mount */
693
694 u32 sd_fsb2bb;
695 u32 sd_fsb2bb_shift;
696 u32 sd_diptrs; /* Number of pointers in a dinode */
697 u32 sd_inptrs; /* Number of pointers in a indirect block */
698 u32 sd_ldptrs; /* Number of pointers in a log descriptor block */
699 u32 sd_jbsize; /* Size of a journaled data block */
700 u32 sd_hash_bsize; /* sizeof(exhash block) */
701 u32 sd_hash_bsize_shift;
702 u32 sd_hash_ptrs; /* Number of pointers in a hash block */
703 u32 sd_qc_per_block;
704 u32 sd_blocks_per_bitmap;
705 u32 sd_max_dirres; /* Max blocks needed to add a directory entry */
706 u32 sd_max_height; /* Max height of a file's metadata tree */
707 u64 sd_heightsize[GFS2_MAX_META_HEIGHT + 1];
708 u32 sd_max_dents_per_leaf; /* Max number of dirents in a leaf block */
709
710 struct gfs2_args sd_args; /* Mount arguments */
711 struct gfs2_tune sd_tune; /* Filesystem tuning structure */
712
713 /* Lock Stuff */
714
715 struct lm_lockstruct sd_lockstruct;
716 struct gfs2_holder sd_live_gh;
717 struct gfs2_glock *sd_rename_gl;
718 struct gfs2_glock *sd_freeze_gl;
719 struct work_struct sd_freeze_work;
720 wait_queue_head_t sd_kill_wait;
721 wait_queue_head_t sd_async_glock_wait;
722 atomic_t sd_glock_disposal;
723 struct completion sd_locking_init;
724 struct completion sd_wdack;
725 struct delayed_work sd_control_work;
726
727 /* Inode Stuff */
728
729 struct dentry *sd_master_dir;
730 struct dentry *sd_root_dir;
731
732 struct inode *sd_jindex;
733 struct inode *sd_statfs_inode;
734 struct inode *sd_sc_inode;
735 struct list_head sd_sc_inodes_list;
736 struct inode *sd_qc_inode;
737 struct inode *sd_rindex;
738 struct inode *sd_quota_inode;
739
740 /* StatFS stuff */
741
742 spinlock_t sd_statfs_spin;
743 struct gfs2_statfs_change_host sd_statfs_master;
744 struct gfs2_statfs_change_host sd_statfs_local;
745 int sd_statfs_force_sync;
746
747 /* Resource group stuff */
748
749 int sd_rindex_uptodate;
750 spinlock_t sd_rindex_spin;
751 struct rb_root sd_rindex_tree;
752 unsigned int sd_rgrps;
753 unsigned int sd_max_rg_data;
754
755 /* Journal index stuff */
756
757 struct list_head sd_jindex_list;
758 spinlock_t sd_jindex_spin;
759 struct mutex sd_jindex_mutex;
760 unsigned int sd_journals;
761
762 struct gfs2_jdesc *sd_jdesc;
763 struct gfs2_holder sd_journal_gh;
764 struct gfs2_holder sd_jinode_gh;
765 struct gfs2_glock *sd_jinode_gl;
766
767 struct gfs2_holder sd_sc_gh;
768 struct buffer_head *sd_sc_bh;
769 struct gfs2_holder sd_qc_gh;
770
771 struct completion sd_journal_ready;
772
773 /* Workqueue stuff */
774
775 struct workqueue_struct *sd_delete_wq;
776
777 /* Daemon stuff */
778
779 struct task_struct *sd_logd_process;
780 struct task_struct *sd_quotad_process;
781
782 /* Quota stuff */
783
784 struct list_head sd_quota_list;
785 atomic_t sd_quota_count;
786 struct mutex sd_quota_mutex;
787 struct mutex sd_quota_sync_mutex;
788 wait_queue_head_t sd_quota_wait;
789
790 unsigned int sd_quota_slots;
791 unsigned long *sd_quota_bitmap;
792 spinlock_t sd_bitmap_lock;
793
794 u64 sd_quota_sync_gen;
795
796 /* Log stuff */
797
798 struct address_space sd_aspace;
799
800 spinlock_t sd_log_lock;
801
802 struct gfs2_trans *sd_log_tr;
803 unsigned int sd_log_blks_reserved;
804
805 atomic_t sd_log_pinned;
806 unsigned int sd_log_num_revoke;
807
808 struct list_head sd_log_revokes;
809 struct list_head sd_log_ordered;
810 spinlock_t sd_ordered_lock;
811
812 atomic_t sd_log_thresh1;
813 atomic_t sd_log_thresh2;
814 atomic_t sd_log_blks_free;
815 atomic_t sd_log_blks_needed;
816 atomic_t sd_log_revokes_available;
817 wait_queue_head_t sd_log_waitq;
818 wait_queue_head_t sd_logd_waitq;
819
820 u64 sd_log_sequence;
821 int sd_log_idle;
822
823 struct rw_semaphore sd_log_flush_lock;
824 atomic_t sd_log_in_flight;
825 wait_queue_head_t sd_log_flush_wait;
826 int sd_log_error; /* First log error */
827 wait_queue_head_t sd_withdraw_wait;
828
829 unsigned int sd_log_tail;
830 unsigned int sd_log_flush_tail;
831 unsigned int sd_log_head;
832 unsigned int sd_log_flush_head;
833
834 spinlock_t sd_ail_lock;
835 struct list_head sd_ail1_list;
836 struct list_head sd_ail2_list;
837
838 /* For quiescing the filesystem */
839 struct gfs2_holder sd_freeze_gh;
840 struct mutex sd_freeze_mutex;
841
842 char sd_fsname[GFS2_FSNAME_LEN + 3 * sizeof(int) + 2];
843 char sd_table_name[GFS2_FSNAME_LEN];
844 char sd_proto_name[GFS2_FSNAME_LEN];
845
846 /* Debugging crud */
847
848 unsigned long sd_last_warning;
849 struct dentry *debugfs_dir; /* debugfs directory */
850 unsigned long sd_glock_dqs_held;
851};
852
853static inline void gfs2_glstats_inc(struct gfs2_glock *gl, int which)
854{
855 gl->gl_stats.stats[which]++;
856}
857
858static inline void gfs2_sbstats_inc(const struct gfs2_glock *gl, int which)
859{
860 const struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
861 preempt_disable();
862 this_cpu_ptr(sdp->sd_lkstats)->lkstats[gl->gl_name.ln_type].stats[which]++;
863 preempt_enable();
864}
865
866struct gfs2_rgrpd *gfs2_glock2rgrp(struct gfs2_glock *gl);
867
868static inline unsigned gfs2_max_stuffed_size(const struct gfs2_inode *ip)
869{
870 return GFS2_SB(&ip->i_inode)->sd_sb.sb_bsize - sizeof(struct gfs2_dinode);
871}
872
873#endif /* __INCORE_DOT_H__ */
874