xref: /linux/fs/gfs2/incore.h (revision f79e4d5f92a129a1159c973735007d4ddc8541f3)
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 
30 struct gfs2_log_operations;
31 struct gfs2_bufdata;
32 struct gfs2_holder;
33 struct gfs2_glock;
34 struct gfs2_quota_data;
35 struct gfs2_trans;
36 struct gfs2_jdesc;
37 struct gfs2_sbd;
38 struct lm_lockops;
39 
40 typedef void (*gfs2_glop_bh_t) (struct gfs2_glock *gl, unsigned int ret);
41 
42 struct 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 
54 struct 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 
68 struct 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 
78 struct 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 
106 struct gfs2_rbm {
107 	struct gfs2_rgrpd *rgd;
108 	u32 offset;		/* The offset is bitmap relative */
109 	int bii;		/* Bitmap index */
110 };
111 
112 static inline struct gfs2_bitmap *rbm_bi(const struct gfs2_rbm *rbm)
113 {
114 	return rbm->rgd->rd_bits + rbm->bii;
115 }
116 
117 static inline u64 gfs2_rbm_to_block(const struct gfs2_rbm *rbm)
118 {
119 	BUG_ON(rbm->offset >= rbm->rgd->rd_data);
120 	return rbm->rgd->rd_data0 + (rbm_bi(rbm)->bi_start * GFS2_NBBY) +
121 		rbm->offset;
122 }
123 
124 static inline bool gfs2_rbm_eq(const struct gfs2_rbm *rbm1,
125 			       const struct gfs2_rbm *rbm2)
126 {
127 	return (rbm1->rgd == rbm2->rgd) && (rbm1->bii == rbm2->bii) &&
128 	       (rbm1->offset == rbm2->offset);
129 }
130 
131 enum gfs2_state_bits {
132 	BH_Pinned = BH_PrivateStart,
133 	BH_Escaped = BH_PrivateStart + 1,
134 };
135 
136 BUFFER_FNS(Pinned, pinned)
137 TAS_BUFFER_FNS(Pinned, pinned)
138 BUFFER_FNS(Escaped, escaped)
139 TAS_BUFFER_FNS(Escaped, escaped)
140 
141 struct gfs2_bufdata {
142 	struct buffer_head *bd_bh;
143 	struct gfs2_glock *bd_gl;
144 	u64 bd_blkno;
145 
146 	struct list_head bd_list;
147 	const struct gfs2_log_operations *bd_ops;
148 
149 	struct gfs2_trans *bd_tr;
150 	struct list_head bd_ail_st_list;
151 	struct list_head bd_ail_gl_list;
152 };
153 
154 /*
155  * Internally, we prefix things with gdlm_ and GDLM_ (for gfs-dlm) since a
156  * prefix of lock_dlm_ gets awkward.
157  */
158 
159 #define GDLM_STRNAME_BYTES	25
160 #define GDLM_LVB_SIZE		32
161 
162 /*
163  * ls_recover_flags:
164  *
165  * DFL_BLOCK_LOCKS: dlm is in recovery and will grant locks that had been
166  * held by failed nodes whose journals need recovery.  Those locks should
167  * only be used for journal recovery until the journal recovery is done.
168  * This is set by the dlm recover_prep callback and cleared by the
169  * gfs2_control thread when journal recovery is complete.  To avoid
170  * races between recover_prep setting and gfs2_control clearing, recover_spin
171  * is held while changing this bit and reading/writing recover_block
172  * and recover_start.
173  *
174  * DFL_NO_DLM_OPS: dlm lockspace ops/callbacks are not being used.
175  *
176  * DFL_FIRST_MOUNT: this node is the first to mount this fs and is doing
177  * recovery of all journals before allowing other nodes to mount the fs.
178  * This is cleared when FIRST_MOUNT_DONE is set.
179  *
180  * DFL_FIRST_MOUNT_DONE: this node was the first mounter, and has finished
181  * recovery of all journals, and now allows other nodes to mount the fs.
182  *
183  * DFL_MOUNT_DONE: gdlm_mount has completed successfully and cleared
184  * BLOCK_LOCKS for the first time.  The gfs2_control thread should now
185  * control clearing BLOCK_LOCKS for further recoveries.
186  *
187  * DFL_UNMOUNT: gdlm_unmount sets to keep sdp off gfs2_control_wq.
188  *
189  * DFL_DLM_RECOVERY: set while dlm is in recovery, between recover_prep()
190  * and recover_done(), i.e. set while recover_block == recover_start.
191  */
192 
193 enum {
194 	DFL_BLOCK_LOCKS		= 0,
195 	DFL_NO_DLM_OPS		= 1,
196 	DFL_FIRST_MOUNT		= 2,
197 	DFL_FIRST_MOUNT_DONE	= 3,
198 	DFL_MOUNT_DONE		= 4,
199 	DFL_UNMOUNT		= 5,
200 	DFL_DLM_RECOVERY	= 6,
201 };
202 
203 /*
204  * We are using struct lm_lockname as an rhashtable key.  Avoid holes within
205  * the struct; padding at the end is fine.
206  */
207 struct lm_lockname {
208 	u64 ln_number;
209 	struct gfs2_sbd *ln_sbd;
210 	unsigned int ln_type;
211 };
212 
213 #define lm_name_equal(name1, name2) \
214         (((name1)->ln_number == (name2)->ln_number) &&	\
215 	 ((name1)->ln_type == (name2)->ln_type) &&	\
216 	 ((name1)->ln_sbd == (name2)->ln_sbd))
217 
218 
219 struct gfs2_glock_operations {
220 	void (*go_sync) (struct gfs2_glock *gl);
221 	int (*go_xmote_bh) (struct gfs2_glock *gl, struct gfs2_holder *gh);
222 	void (*go_inval) (struct gfs2_glock *gl, int flags);
223 	int (*go_demote_ok) (const struct gfs2_glock *gl);
224 	int (*go_lock) (struct gfs2_holder *gh);
225 	void (*go_unlock) (struct gfs2_holder *gh);
226 	void (*go_dump)(struct seq_file *seq, const struct gfs2_glock *gl);
227 	void (*go_callback)(struct gfs2_glock *gl, bool remote);
228 	const int go_type;
229 	const unsigned long go_flags;
230 #define GLOF_ASPACE 1
231 #define GLOF_LVB    2
232 #define GLOF_LRU    4
233 };
234 
235 enum {
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 
247 struct gfs2_lkstats {
248 	u64 stats[GFS2_NR_LKSTATS];
249 };
250 
251 enum {
252 	/* States */
253 	HIF_HOLDER		= 6,  /* Set for gh that "holds" the glock */
254 	HIF_FIRST		= 7,
255 	HIF_WAIT		= 10,
256 };
257 
258 struct gfs2_holder {
259 	struct list_head gh_list;
260 
261 	struct gfs2_glock *gh_gl;
262 	struct pid *gh_owner_pid;
263 	u16 gh_flags;
264 	u16 gh_state;
265 
266 	int gh_error;
267 	unsigned long gh_iflags; /* HIF_... */
268 	unsigned long gh_ip;
269 };
270 
271 /* Number of quota types we support */
272 #define GFS2_MAXQUOTAS 2
273 
274 struct gfs2_qadata { /* quota allocation data */
275 	/* Quota stuff */
276 	struct gfs2_quota_data *qa_qd[2 * GFS2_MAXQUOTAS];
277 	struct gfs2_holder qa_qd_ghs[2 * GFS2_MAXQUOTAS];
278 	unsigned int qa_qd_num;
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 
290 struct gfs2_blkreserv {
291 	/* components used during write (step 1): */
292 	atomic_t rs_sizehint;         /* hint of the write size */
293 
294 	struct gfs2_holder rs_rgd_gh; /* Filled in by get_local_rgrp */
295 	struct rb_node rs_node;       /* link to other block reservations */
296 	struct gfs2_rbm rs_rbm;       /* Start of reservation */
297 	u32 rs_free;                  /* how many blocks are still free */
298 	u64 rs_inum;                  /* Inode number for reservation */
299 };
300 
301 /*
302  * Allocation parameters
303  * @target: The number of blocks we'd ideally like to allocate
304  * @aflags: The flags (e.g. Orlov flag)
305  *
306  * The intent is to gradually expand this structure over time in
307  * order to give more information, e.g. alignment, min extent size
308  * to the allocation code.
309  */
310 struct gfs2_alloc_parms {
311 	u64 target;
312 	u32 min_target;
313 	u32 aflags;
314 	u64 allowed;
315 };
316 
317 enum {
318 	GLF_LOCK			= 1,
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_QUEUED			= 12,
329 	GLF_LRU				= 13,
330 	GLF_OBJECT			= 14, /* Used only for tracing */
331 	GLF_BLOCKING			= 15,
332 	GLF_INODE_CREATING		= 16, /* Inode creation occurring */
333 };
334 
335 struct 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 	union {
365 		/* For inode and iopen glocks only */
366 		struct work_struct gl_delete;
367 		/* For rgrp glocks only */
368 		struct {
369 			loff_t start;
370 			loff_t end;
371 		} gl_vm;
372 	};
373 	struct rcu_head gl_rcu;
374 	struct rhash_head gl_node;
375 };
376 
377 #define GFS2_MIN_LVB_SIZE 32	/* Min size of LVB that gfs2 supports */
378 
379 enum {
380 	GIF_INVALID		= 0,
381 	GIF_QD_LOCKED		= 1,
382 	GIF_ALLOC_FAILED	= 2,
383 	GIF_SW_PAGED		= 3,
384 	GIF_ORDERED		= 4,
385 	GIF_FREE_VFS_INODE      = 5,
386 	GIF_GLOP_PENDING	= 6,
387 };
388 
389 struct gfs2_inode {
390 	struct inode i_inode;
391 	u64 i_no_addr;
392 	u64 i_no_formal_ino;
393 	u64 i_generation;
394 	u64 i_eattr;
395 	unsigned long i_flags;		/* GIF_... */
396 	struct gfs2_glock *i_gl; /* Move into i_gh? */
397 	struct gfs2_holder i_iopen_gh;
398 	struct gfs2_holder i_gh; /* for prepare/commit_write only */
399 	struct gfs2_qadata *i_qadata; /* quota allocation data */
400 	struct gfs2_blkreserv i_res; /* rgrp multi-block reservation */
401 	struct gfs2_rgrpd *i_rgd;
402 	u64 i_goal;	/* goal block for allocations */
403 	struct rw_semaphore i_rw_mutex;
404 	struct list_head i_ordered;
405 	struct list_head i_trunc_list;
406 	__be64 *i_hash_cache;
407 	u32 i_entries;
408 	u32 i_diskflags;
409 	u8 i_height;
410 	u8 i_depth;
411 	u16 i_rahead;
412 };
413 
414 /*
415  * Since i_inode is the first element of struct gfs2_inode,
416  * this is effectively a cast.
417  */
418 static inline struct gfs2_inode *GFS2_I(struct inode *inode)
419 {
420 	return container_of(inode, struct gfs2_inode, i_inode);
421 }
422 
423 static inline struct gfs2_sbd *GFS2_SB(const struct inode *inode)
424 {
425 	return inode->i_sb->s_fs_info;
426 }
427 
428 struct gfs2_file {
429 	struct mutex f_fl_mutex;
430 	struct gfs2_holder f_fl_gh;
431 };
432 
433 struct gfs2_revoke_replay {
434 	struct list_head rr_list;
435 	u64 rr_blkno;
436 	unsigned int rr_where;
437 };
438 
439 enum {
440 	QDF_CHANGE		= 1,
441 	QDF_LOCKED		= 2,
442 	QDF_REFRESH		= 3,
443 	QDF_QMSG_QUIET          = 4,
444 };
445 
446 struct gfs2_quota_data {
447 	struct hlist_bl_node qd_hlist;
448 	struct list_head qd_list;
449 	struct kqid qd_id;
450 	struct gfs2_sbd *qd_sbd;
451 	struct lockref qd_lockref;
452 	struct list_head qd_lru;
453 	unsigned qd_hash;
454 
455 	unsigned long qd_flags;		/* QDF_... */
456 
457 	s64 qd_change;
458 	s64 qd_change_sync;
459 
460 	unsigned int qd_slot;
461 	unsigned int qd_slot_count;
462 
463 	struct buffer_head *qd_bh;
464 	struct gfs2_quota_change *qd_bh_qc;
465 	unsigned int qd_bh_count;
466 
467 	struct gfs2_glock *qd_gl;
468 	struct gfs2_quota_lvb qd_qb;
469 
470 	u64 qd_sync_gen;
471 	unsigned long qd_last_warn;
472 	struct rcu_head qd_rcu;
473 };
474 
475 enum {
476 	TR_TOUCHED = 1,
477 	TR_ATTACHED = 2,
478 	TR_ALLOCED = 3,
479 };
480 
481 struct gfs2_trans {
482 	unsigned long tr_ip;
483 
484 	unsigned int tr_blocks;
485 	unsigned int tr_revokes;
486 	unsigned int tr_reserved;
487 	unsigned long tr_flags;
488 
489 	unsigned int tr_num_buf_new;
490 	unsigned int tr_num_databuf_new;
491 	unsigned int tr_num_buf_rm;
492 	unsigned int tr_num_databuf_rm;
493 	unsigned int tr_num_revoke;
494 	unsigned int tr_num_revoke_rm;
495 
496 	struct list_head tr_list;
497 	struct list_head tr_databuf;
498 	struct list_head tr_buf;
499 
500 	unsigned int tr_first;
501 	struct list_head tr_ail1_list;
502 	struct list_head tr_ail2_list;
503 };
504 
505 struct gfs2_journal_extent {
506 	struct list_head list;
507 
508 	unsigned int lblock; /* First logical block */
509 	u64 dblock; /* First disk block */
510 	u64 blocks;
511 };
512 
513 struct gfs2_jdesc {
514 	struct list_head jd_list;
515 	struct list_head extent_list;
516 	unsigned int nr_extents;
517 	struct work_struct jd_work;
518 	struct inode *jd_inode;
519 	unsigned long jd_flags;
520 #define JDF_RECOVERY 1
521 	unsigned int jd_jid;
522 	unsigned int jd_blocks;
523 	int jd_recover_error;
524 	/* Replay stuff */
525 
526 	unsigned int jd_found_blocks;
527 	unsigned int jd_found_revokes;
528 	unsigned int jd_replayed_blocks;
529 
530 	struct list_head jd_revoke_list;
531 	unsigned int jd_replay_tail;
532 
533 };
534 
535 struct gfs2_statfs_change_host {
536 	s64 sc_total;
537 	s64 sc_free;
538 	s64 sc_dinodes;
539 };
540 
541 #define GFS2_QUOTA_DEFAULT	GFS2_QUOTA_OFF
542 #define GFS2_QUOTA_OFF		0
543 #define GFS2_QUOTA_ACCOUNT	1
544 #define GFS2_QUOTA_ON		2
545 
546 #define GFS2_DATA_DEFAULT	GFS2_DATA_ORDERED
547 #define GFS2_DATA_WRITEBACK	1
548 #define GFS2_DATA_ORDERED	2
549 
550 #define GFS2_ERRORS_DEFAULT     GFS2_ERRORS_WITHDRAW
551 #define GFS2_ERRORS_WITHDRAW    0
552 #define GFS2_ERRORS_CONTINUE    1 /* place holder for future feature */
553 #define GFS2_ERRORS_RO          2 /* place holder for future feature */
554 #define GFS2_ERRORS_PANIC       3
555 
556 struct gfs2_args {
557 	char ar_lockproto[GFS2_LOCKNAME_LEN];	/* Name of the Lock Protocol */
558 	char ar_locktable[GFS2_LOCKNAME_LEN];	/* Name of the Lock Table */
559 	char ar_hostdata[GFS2_LOCKNAME_LEN];	/* Host specific data */
560 	unsigned int ar_spectator:1;		/* Don't get a journal */
561 	unsigned int ar_localflocks:1;		/* Let the VFS do flock|fcntl */
562 	unsigned int ar_debug:1;		/* Oops on errors */
563 	unsigned int ar_posix_acl:1;		/* Enable posix acls */
564 	unsigned int ar_quota:2;		/* off/account/on */
565 	unsigned int ar_suiddir:1;		/* suiddir support */
566 	unsigned int ar_data:2;			/* ordered/writeback */
567 	unsigned int ar_meta:1;			/* mount metafs */
568 	unsigned int ar_discard:1;		/* discard requests */
569 	unsigned int ar_errors:2;               /* errors=withdraw | panic */
570 	unsigned int ar_nobarrier:1;            /* do not send barriers */
571 	unsigned int ar_rgrplvb:1;		/* use lvbs for rgrp info */
572 	unsigned int ar_loccookie:1;		/* use location based readdir
573 						   cookies */
574 	int ar_commit;				/* Commit interval */
575 	int ar_statfs_quantum;			/* The fast statfs interval */
576 	int ar_quota_quantum;			/* The quota interval */
577 	int ar_statfs_percent;			/* The % change to force sync */
578 };
579 
580 struct gfs2_tune {
581 	spinlock_t gt_spin;
582 
583 	unsigned int gt_logd_secs;
584 
585 	unsigned int gt_quota_warn_period; /* Secs between quota warn msgs */
586 	unsigned int gt_quota_scale_num; /* Numerator */
587 	unsigned int gt_quota_scale_den; /* Denominator */
588 	unsigned int gt_quota_quantum; /* Secs between syncs to quota file */
589 	unsigned int gt_new_files_jdata;
590 	unsigned int gt_max_readahead; /* Max bytes to read-ahead from disk */
591 	unsigned int gt_complain_secs;
592 	unsigned int gt_statfs_quantum;
593 	unsigned int gt_statfs_slow;
594 };
595 
596 enum {
597 	SDF_JOURNAL_CHECKED	= 0,
598 	SDF_JOURNAL_LIVE	= 1,
599 	SDF_SHUTDOWN		= 2,
600 	SDF_NOBARRIERS		= 3,
601 	SDF_NORECOVERY		= 4,
602 	SDF_DEMOTE		= 5,
603 	SDF_NOJOURNALID		= 6,
604 	SDF_RORECOVERY		= 7, /* read only recovery */
605 	SDF_SKIP_DLM_UNLOCK	= 8,
606 	SDF_FORCE_AIL_FLUSH     = 9,
607 };
608 
609 enum gfs2_freeze_state {
610 	SFS_UNFROZEN		= 0,
611 	SFS_STARTING_FREEZE	= 1,
612 	SFS_FROZEN		= 2,
613 };
614 
615 #define GFS2_FSNAME_LEN		256
616 
617 struct gfs2_inum_host {
618 	u64 no_formal_ino;
619 	u64 no_addr;
620 };
621 
622 struct gfs2_sb_host {
623 	u32 sb_magic;
624 	u32 sb_type;
625 	u32 sb_format;
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 
648 struct 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 
673 struct gfs2_pcpu_lkstats {
674 	/* One struct for each glock type */
675 	struct gfs2_lkstats lkstats[10];
676 };
677 
678 struct gfs2_sbd {
679 	struct super_block *sd_vfs;
680 	struct gfs2_pcpu_lkstats __percpu *sd_lkstats;
681 	struct kobject sd_kobj;
682 	unsigned long sd_flags;	/* SDF_... */
683 	struct gfs2_sb_host sd_sb;
684 
685 	/* Constants computed on mount */
686 
687 	u32 sd_fsb2bb;
688 	u32 sd_fsb2bb_shift;
689 	u32 sd_diptrs;	/* Number of pointers in a dinode */
690 	u32 sd_inptrs;	/* Number of pointers in a indirect block */
691 	u32 sd_jbsize;	/* Size of a journaled data block */
692 	u32 sd_hash_bsize;	/* sizeof(exhash block) */
693 	u32 sd_hash_bsize_shift;
694 	u32 sd_hash_ptrs;	/* Number of pointers in a hash block */
695 	u32 sd_qc_per_block;
696 	u32 sd_blocks_per_bitmap;
697 	u32 sd_max_dirres;	/* Max blocks needed to add a directory entry */
698 	u32 sd_max_height;	/* Max height of a file's metadata tree */
699 	u64 sd_heightsize[GFS2_MAX_META_HEIGHT + 1];
700 	u32 sd_max_dents_per_leaf; /* Max number of dirents in a leaf block */
701 
702 	struct gfs2_args sd_args;	/* Mount arguments */
703 	struct gfs2_tune sd_tune;	/* Filesystem tuning structure */
704 
705 	/* Lock Stuff */
706 
707 	struct lm_lockstruct sd_lockstruct;
708 	struct gfs2_holder sd_live_gh;
709 	struct gfs2_glock *sd_rename_gl;
710 	struct gfs2_glock *sd_freeze_gl;
711 	struct work_struct sd_freeze_work;
712 	wait_queue_head_t sd_glock_wait;
713 	atomic_t sd_glock_disposal;
714 	struct completion sd_locking_init;
715 	struct completion sd_wdack;
716 	struct delayed_work sd_control_work;
717 
718 	/* Inode Stuff */
719 
720 	struct dentry *sd_master_dir;
721 	struct dentry *sd_root_dir;
722 
723 	struct inode *sd_jindex;
724 	struct inode *sd_statfs_inode;
725 	struct inode *sd_sc_inode;
726 	struct inode *sd_qc_inode;
727 	struct inode *sd_rindex;
728 	struct inode *sd_quota_inode;
729 
730 	/* StatFS stuff */
731 
732 	spinlock_t sd_statfs_spin;
733 	struct gfs2_statfs_change_host sd_statfs_master;
734 	struct gfs2_statfs_change_host sd_statfs_local;
735 	int sd_statfs_force_sync;
736 
737 	/* Resource group stuff */
738 
739 	int sd_rindex_uptodate;
740 	spinlock_t sd_rindex_spin;
741 	struct rb_root sd_rindex_tree;
742 	unsigned int sd_rgrps;
743 	unsigned int sd_max_rg_data;
744 
745 	/* Journal index stuff */
746 
747 	struct list_head sd_jindex_list;
748 	spinlock_t sd_jindex_spin;
749 	struct mutex sd_jindex_mutex;
750 	unsigned int sd_journals;
751 
752 	struct gfs2_jdesc *sd_jdesc;
753 	struct gfs2_holder sd_journal_gh;
754 	struct gfs2_holder sd_jinode_gh;
755 
756 	struct gfs2_holder sd_sc_gh;
757 	struct gfs2_holder sd_qc_gh;
758 
759 	struct completion sd_journal_ready;
760 
761 	/* Daemon stuff */
762 
763 	struct task_struct *sd_logd_process;
764 	struct task_struct *sd_quotad_process;
765 
766 	/* Quota stuff */
767 
768 	struct list_head sd_quota_list;
769 	atomic_t sd_quota_count;
770 	struct mutex sd_quota_mutex;
771 	struct mutex sd_quota_sync_mutex;
772 	wait_queue_head_t sd_quota_wait;
773 	struct list_head sd_trunc_list;
774 	spinlock_t sd_trunc_lock;
775 
776 	unsigned int sd_quota_slots;
777 	unsigned long *sd_quota_bitmap;
778 	spinlock_t sd_bitmap_lock;
779 
780 	u64 sd_quota_sync_gen;
781 
782 	/* Log stuff */
783 
784 	struct address_space sd_aspace;
785 
786 	spinlock_t sd_log_lock;
787 
788 	struct gfs2_trans *sd_log_tr;
789 	unsigned int sd_log_blks_reserved;
790 	int sd_log_commited_revoke;
791 
792 	atomic_t sd_log_pinned;
793 	unsigned int sd_log_num_revoke;
794 
795 	struct list_head sd_log_le_revoke;
796 	struct list_head sd_log_le_ordered;
797 	spinlock_t sd_ordered_lock;
798 
799 	atomic_t sd_log_thresh1;
800 	atomic_t sd_log_thresh2;
801 	atomic_t sd_log_blks_free;
802 	atomic_t sd_log_blks_needed;
803 	wait_queue_head_t sd_log_waitq;
804 	wait_queue_head_t sd_logd_waitq;
805 
806 	u64 sd_log_sequence;
807 	unsigned int sd_log_head;
808 	unsigned int sd_log_tail;
809 	int sd_log_idle;
810 
811 	struct rw_semaphore sd_log_flush_lock;
812 	atomic_t sd_log_in_flight;
813 	struct bio *sd_log_bio;
814 	wait_queue_head_t sd_log_flush_wait;
815 	int sd_log_error;
816 
817 	atomic_t sd_reserving_log;
818 	wait_queue_head_t sd_reserving_log_wait;
819 
820 	unsigned int sd_log_flush_head;
821 
822 	spinlock_t sd_ail_lock;
823 	struct list_head sd_ail1_list;
824 	struct list_head sd_ail2_list;
825 
826 	/* For quiescing the filesystem */
827 	struct gfs2_holder sd_freeze_gh;
828 	atomic_t sd_freeze_state;
829 	struct mutex sd_freeze_mutex;
830 
831 	char sd_fsname[GFS2_FSNAME_LEN + 3 * sizeof(int) + 2];
832 	char sd_table_name[GFS2_FSNAME_LEN];
833 	char sd_proto_name[GFS2_FSNAME_LEN];
834 
835 	/* Debugging crud */
836 
837 	unsigned long sd_last_warning;
838 	struct dentry *debugfs_dir;    /* debugfs directory */
839 	struct dentry *debugfs_dentry_glocks;
840 	struct dentry *debugfs_dentry_glstats;
841 	struct dentry *debugfs_dentry_sbstats;
842 };
843 
844 static inline void gfs2_glstats_inc(struct gfs2_glock *gl, int which)
845 {
846 	gl->gl_stats.stats[which]++;
847 }
848 
849 static inline void gfs2_sbstats_inc(const struct gfs2_glock *gl, int which)
850 {
851 	const struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
852 	preempt_disable();
853 	this_cpu_ptr(sdp->sd_lkstats)->lkstats[gl->gl_name.ln_type].stats[which]++;
854 	preempt_enable();
855 }
856 
857 extern struct gfs2_rgrpd *gfs2_glock2rgrp(struct gfs2_glock *gl);
858 
859 static inline unsigned gfs2_max_stuffed_size(const struct gfs2_inode *ip)
860 {
861 	return GFS2_SB(&ip->i_inode)->sd_sb.sb_bsize - sizeof(struct gfs2_dinode);
862 }
863 
864 #endif /* __INCORE_DOT_H__ */
865 
866