xref: /openbmc/linux/fs/gfs2/incore.h (revision dbf563ee)
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 
24 #define DIO_WAIT	0x00000010
25 #define DIO_METADATA	0x00000020
26 
27 struct gfs2_log_operations;
28 struct gfs2_bufdata;
29 struct gfs2_holder;
30 struct gfs2_glock;
31 struct gfs2_quota_data;
32 struct gfs2_trans;
33 struct gfs2_jdesc;
34 struct gfs2_sbd;
35 struct lm_lockops;
36 
37 typedef void (*gfs2_glop_bh_t) (struct gfs2_glock *gl, unsigned int ret);
38 
39 struct gfs2_log_header_host {
40 	u64 lh_sequence;	/* Sequence number of this transaction */
41 	u32 lh_flags;		/* GFS2_LOG_HEAD_... */
42 	u32 lh_tail;		/* Block number of log tail */
43 	u32 lh_blkno;
44 };
45 
46 /*
47  * Structure of operations that are associated with each
48  * type of element in the log.
49  */
50 
51 struct gfs2_log_operations {
52 	void (*lo_before_commit) (struct gfs2_sbd *sdp, struct gfs2_trans *tr);
53 	void (*lo_after_commit) (struct gfs2_sbd *sdp, struct gfs2_trans *tr);
54 	void (*lo_before_scan) (struct gfs2_jdesc *jd,
55 				struct gfs2_log_header_host *head, int pass);
56 	int (*lo_scan_elements) (struct gfs2_jdesc *jd, unsigned int start,
57 				 struct gfs2_log_descriptor *ld, __be64 *ptr,
58 				 int pass);
59 	void (*lo_after_scan) (struct gfs2_jdesc *jd, int error, int pass);
60 	const char *lo_name;
61 };
62 
63 #define GBF_FULL 1
64 
65 /**
66  * Clone bitmaps (bi_clone):
67  *
68  * - When a block is freed, we remember the previous state of the block in the
69  *   clone bitmap, and only mark the block as free in the real bitmap.
70  *
71  * - When looking for a block to allocate, we check for a free block in the
72  *   clone bitmap, and if no clone bitmap exists, in the real bitmap.
73  *
74  * - For allocating a block, we mark it as allocated in the real bitmap, and if
75  *   a clone bitmap exists, also in the clone bitmap.
76  *
77  * - At the end of a log_flush, we copy the real bitmap into the clone bitmap
78  *   to make the clone bitmap reflect the current allocation state.
79  *   (Alternatively, we could remove the clone bitmap.)
80  *
81  * The clone bitmaps are in-core only, and is never written to disk.
82  *
83  * These steps ensure that blocks which have been freed in a transaction cannot
84  * be reallocated in that same transaction.
85  */
86 struct gfs2_bitmap {
87 	struct buffer_head *bi_bh;
88 	char *bi_clone;
89 	unsigned long bi_flags;
90 	u32 bi_offset;
91 	u32 bi_start;
92 	u32 bi_bytes;
93 	u32 bi_blocks;
94 };
95 
96 struct gfs2_rgrpd {
97 	struct rb_node rd_node;		/* Link with superblock */
98 	struct gfs2_glock *rd_gl;	/* Glock for this rgrp */
99 	u64 rd_addr;			/* grp block disk address */
100 	u64 rd_data0;			/* first data location */
101 	u32 rd_length;			/* length of rgrp header in fs blocks */
102 	u32 rd_data;			/* num of data blocks in rgrp */
103 	u32 rd_bitbytes;		/* number of bytes in data bitmaps */
104 	u32 rd_free;
105 	u32 rd_reserved;                /* number of blocks reserved */
106 	u32 rd_free_clone;
107 	u32 rd_dinodes;
108 	u64 rd_igeneration;
109 	struct gfs2_bitmap *rd_bits;
110 	struct gfs2_sbd *rd_sbd;
111 	struct gfs2_rgrp_lvb *rd_rgl;
112 	u32 rd_last_alloc;
113 	u32 rd_flags;
114 	u32 rd_extfail_pt;		/* extent failure point */
115 #define GFS2_RDF_CHECK		0x10000000 /* check for unlinked inodes */
116 #define GFS2_RDF_UPTODATE	0x20000000 /* rg is up to date */
117 #define GFS2_RDF_ERROR		0x40000000 /* error in rg */
118 #define GFS2_RDF_PREFERRED	0x80000000 /* This rgrp is preferred */
119 #define GFS2_RDF_MASK		0xf0000000 /* mask for internal flags */
120 	spinlock_t rd_rsspin;           /* protects reservation related vars */
121 	struct rb_root rd_rstree;       /* multi-block reservation tree */
122 };
123 
124 struct gfs2_rbm {
125 	struct gfs2_rgrpd *rgd;
126 	u32 offset;		/* The offset is bitmap relative */
127 	int bii;		/* Bitmap index */
128 };
129 
130 static inline struct gfs2_bitmap *rbm_bi(const struct gfs2_rbm *rbm)
131 {
132 	return rbm->rgd->rd_bits + rbm->bii;
133 }
134 
135 static inline u64 gfs2_rbm_to_block(const struct gfs2_rbm *rbm)
136 {
137 	BUG_ON(rbm->offset >= rbm->rgd->rd_data);
138 	return rbm->rgd->rd_data0 + (rbm_bi(rbm)->bi_start * GFS2_NBBY) +
139 		rbm->offset;
140 }
141 
142 static inline bool gfs2_rbm_eq(const struct gfs2_rbm *rbm1,
143 			       const struct gfs2_rbm *rbm2)
144 {
145 	return (rbm1->rgd == rbm2->rgd) && (rbm1->bii == rbm2->bii) &&
146 	       (rbm1->offset == rbm2->offset);
147 }
148 
149 enum gfs2_state_bits {
150 	BH_Pinned = BH_PrivateStart,
151 	BH_Escaped = BH_PrivateStart + 1,
152 };
153 
154 BUFFER_FNS(Pinned, pinned)
155 TAS_BUFFER_FNS(Pinned, pinned)
156 BUFFER_FNS(Escaped, escaped)
157 TAS_BUFFER_FNS(Escaped, escaped)
158 
159 struct gfs2_bufdata {
160 	struct buffer_head *bd_bh;
161 	struct gfs2_glock *bd_gl;
162 	u64 bd_blkno;
163 
164 	struct list_head bd_list;
165 
166 	struct gfs2_trans *bd_tr;
167 	struct list_head bd_ail_st_list;
168 	struct list_head bd_ail_gl_list;
169 };
170 
171 /*
172  * Internally, we prefix things with gdlm_ and GDLM_ (for gfs-dlm) since a
173  * prefix of lock_dlm_ gets awkward.
174  */
175 
176 #define GDLM_STRNAME_BYTES	25
177 #define GDLM_LVB_SIZE		32
178 
179 /*
180  * ls_recover_flags:
181  *
182  * DFL_BLOCK_LOCKS: dlm is in recovery and will grant locks that had been
183  * held by failed nodes whose journals need recovery.  Those locks should
184  * only be used for journal recovery until the journal recovery is done.
185  * This is set by the dlm recover_prep callback and cleared by the
186  * gfs2_control thread when journal recovery is complete.  To avoid
187  * races between recover_prep setting and gfs2_control clearing, recover_spin
188  * is held while changing this bit and reading/writing recover_block
189  * and recover_start.
190  *
191  * DFL_NO_DLM_OPS: dlm lockspace ops/callbacks are not being used.
192  *
193  * DFL_FIRST_MOUNT: this node is the first to mount this fs and is doing
194  * recovery of all journals before allowing other nodes to mount the fs.
195  * This is cleared when FIRST_MOUNT_DONE is set.
196  *
197  * DFL_FIRST_MOUNT_DONE: this node was the first mounter, and has finished
198  * recovery of all journals, and now allows other nodes to mount the fs.
199  *
200  * DFL_MOUNT_DONE: gdlm_mount has completed successfully and cleared
201  * BLOCK_LOCKS for the first time.  The gfs2_control thread should now
202  * control clearing BLOCK_LOCKS for further recoveries.
203  *
204  * DFL_UNMOUNT: gdlm_unmount sets to keep sdp off gfs2_control_wq.
205  *
206  * DFL_DLM_RECOVERY: set while dlm is in recovery, between recover_prep()
207  * and recover_done(), i.e. set while recover_block == recover_start.
208  */
209 
210 enum {
211 	DFL_BLOCK_LOCKS		= 0,
212 	DFL_NO_DLM_OPS		= 1,
213 	DFL_FIRST_MOUNT		= 2,
214 	DFL_FIRST_MOUNT_DONE	= 3,
215 	DFL_MOUNT_DONE		= 4,
216 	DFL_UNMOUNT		= 5,
217 	DFL_DLM_RECOVERY	= 6,
218 };
219 
220 /*
221  * We are using struct lm_lockname as an rhashtable key.  Avoid holes within
222  * the struct; padding at the end is fine.
223  */
224 struct lm_lockname {
225 	u64 ln_number;
226 	struct gfs2_sbd *ln_sbd;
227 	unsigned int ln_type;
228 };
229 
230 #define lm_name_equal(name1, name2) \
231         (((name1)->ln_number == (name2)->ln_number) &&	\
232 	 ((name1)->ln_type == (name2)->ln_type) &&	\
233 	 ((name1)->ln_sbd == (name2)->ln_sbd))
234 
235 
236 struct gfs2_glock_operations {
237 	int (*go_sync) (struct gfs2_glock *gl);
238 	int (*go_xmote_bh) (struct gfs2_glock *gl, struct gfs2_holder *gh);
239 	void (*go_inval) (struct gfs2_glock *gl, int flags);
240 	int (*go_demote_ok) (const struct gfs2_glock *gl);
241 	int (*go_lock) (struct gfs2_holder *gh);
242 	void (*go_dump)(struct seq_file *seq, struct gfs2_glock *gl,
243 			const char *fs_id_buf);
244 	void (*go_callback)(struct gfs2_glock *gl, bool remote);
245 	void (*go_free)(struct gfs2_glock *gl);
246 	const int go_type;
247 	const unsigned long go_flags;
248 #define GLOF_ASPACE 1 /* address space attached */
249 #define GLOF_LVB    2 /* Lock Value Block attached */
250 #define GLOF_LRU    4 /* LRU managed */
251 #define GLOF_NONDISK   8 /* not I/O related */
252 };
253 
254 enum {
255 	GFS2_LKS_SRTT = 0,	/* Non blocking smoothed round trip time */
256 	GFS2_LKS_SRTTVAR = 1,	/* Non blocking smoothed variance */
257 	GFS2_LKS_SRTTB = 2,	/* Blocking smoothed round trip time */
258 	GFS2_LKS_SRTTVARB = 3,	/* Blocking smoothed variance */
259 	GFS2_LKS_SIRT = 4,	/* Smoothed Inter-request time */
260 	GFS2_LKS_SIRTVAR = 5,	/* Smoothed Inter-request variance */
261 	GFS2_LKS_DCOUNT = 6,	/* Count of dlm requests */
262 	GFS2_LKS_QCOUNT = 7,	/* Count of gfs2_holder queues */
263 	GFS2_NR_LKSTATS
264 };
265 
266 struct gfs2_lkstats {
267 	u64 stats[GFS2_NR_LKSTATS];
268 };
269 
270 enum {
271 	/* States */
272 	HIF_HOLDER		= 6,  /* Set for gh that "holds" the glock */
273 	HIF_FIRST		= 7,
274 	HIF_WAIT		= 10,
275 };
276 
277 struct gfs2_holder {
278 	struct list_head gh_list;
279 
280 	struct gfs2_glock *gh_gl;
281 	struct pid *gh_owner_pid;
282 	u16 gh_flags;
283 	u16 gh_state;
284 
285 	int gh_error;
286 	unsigned long gh_iflags; /* HIF_... */
287 	unsigned long gh_ip;
288 };
289 
290 /* Number of quota types we support */
291 #define GFS2_MAXQUOTAS 2
292 
293 struct gfs2_qadata { /* quota allocation data */
294 	/* Quota stuff */
295 	struct gfs2_quota_data *qa_qd[2 * GFS2_MAXQUOTAS];
296 	struct gfs2_holder qa_qd_ghs[2 * GFS2_MAXQUOTAS];
297 	unsigned int qa_qd_num;
298 	int qa_ref;
299 };
300 
301 /* Resource group multi-block reservation, in order of appearance:
302 
303    Step 1. Function prepares to write, allocates a mb, sets the size hint.
304    Step 2. User calls inplace_reserve to target an rgrp, sets the rgrp info
305    Step 3. Function get_local_rgrp locks the rgrp, determines which bits to use
306    Step 4. Bits are assigned from the rgrp based on either the reservation
307            or wherever it can.
308 */
309 
310 struct gfs2_blkreserv {
311 	struct rb_node rs_node;       /* link to other block reservations */
312 	struct gfs2_rbm rs_rbm;       /* Start of reservation */
313 	u32 rs_free;                  /* how many blocks are still free */
314 };
315 
316 /*
317  * Allocation parameters
318  * @target: The number of blocks we'd ideally like to allocate
319  * @aflags: The flags (e.g. Orlov flag)
320  *
321  * The intent is to gradually expand this structure over time in
322  * order to give more information, e.g. alignment, min extent size
323  * to the allocation code.
324  */
325 struct gfs2_alloc_parms {
326 	u64 target;
327 	u32 min_target;
328 	u32 aflags;
329 	u64 allowed;
330 };
331 
332 enum {
333 	GLF_LOCK			= 1,
334 	GLF_DEMOTE			= 3,
335 	GLF_PENDING_DEMOTE		= 4,
336 	GLF_DEMOTE_IN_PROGRESS		= 5,
337 	GLF_DIRTY			= 6,
338 	GLF_LFLUSH			= 7,
339 	GLF_INVALIDATE_IN_PROGRESS	= 8,
340 	GLF_REPLY_PENDING		= 9,
341 	GLF_INITIAL			= 10,
342 	GLF_FROZEN			= 11,
343 	GLF_QUEUED			= 12,
344 	GLF_LRU				= 13,
345 	GLF_OBJECT			= 14, /* Used only for tracing */
346 	GLF_BLOCKING			= 15,
347 	GLF_INODE_CREATING		= 16, /* Inode creation occurring */
348 	GLF_PENDING_DELETE		= 17,
349 	GLF_FREEING			= 18, /* Wait for glock to be freed */
350 };
351 
352 struct gfs2_glock {
353 	unsigned long gl_flags;		/* GLF_... */
354 	struct lm_lockname gl_name;
355 
356 	struct lockref gl_lockref;
357 
358 	/* State fields protected by gl_lockref.lock */
359 	unsigned int gl_state:2,	/* Current state */
360 		     gl_target:2,	/* Target state */
361 		     gl_demote_state:2,	/* State requested by remote node */
362 		     gl_req:2,		/* State in last dlm request */
363 		     gl_reply:8;	/* Last reply from the dlm */
364 
365 	unsigned long gl_demote_time; /* time of first demote request */
366 	long gl_hold_time;
367 	struct list_head gl_holders;
368 
369 	const struct gfs2_glock_operations *gl_ops;
370 	ktime_t gl_dstamp;
371 	struct gfs2_lkstats gl_stats;
372 	struct dlm_lksb gl_lksb;
373 	unsigned long gl_tchange;
374 	void *gl_object;
375 
376 	struct list_head gl_lru;
377 	struct list_head gl_ail_list;
378 	atomic_t gl_ail_count;
379 	atomic_t gl_revokes;
380 	struct delayed_work gl_work;
381 	union {
382 		/* For iopen glocks only */
383 		struct {
384 			struct delayed_work gl_delete;
385 			u64 gl_no_formal_ino;
386 		};
387 		/* For rgrp glocks only */
388 		struct {
389 			loff_t start;
390 			loff_t end;
391 		} gl_vm;
392 	};
393 	struct rcu_head gl_rcu;
394 	struct rhash_head gl_node;
395 };
396 
397 enum {
398 	GIF_INVALID		= 0,
399 	GIF_QD_LOCKED		= 1,
400 	GIF_ALLOC_FAILED	= 2,
401 	GIF_SW_PAGED		= 3,
402 	GIF_FREE_VFS_INODE      = 5,
403 	GIF_GLOP_PENDING	= 6,
404 	GIF_DEFERRED_DELETE	= 7,
405 };
406 
407 struct gfs2_inode {
408 	struct inode i_inode;
409 	u64 i_no_addr;
410 	u64 i_no_formal_ino;
411 	u64 i_generation;
412 	u64 i_eattr;
413 	unsigned long i_flags;		/* GIF_... */
414 	struct gfs2_glock *i_gl; /* Move into i_gh? */
415 	struct gfs2_holder i_iopen_gh;
416 	struct gfs2_holder i_gh; /* for prepare/commit_write only */
417 	struct gfs2_qadata *i_qadata; /* quota allocation data */
418 	struct gfs2_holder i_rgd_gh;
419 	struct gfs2_blkreserv i_res; /* rgrp multi-block reservation */
420 	u64 i_goal;	/* goal block for allocations */
421 	atomic_t i_sizehint;  /* hint of the write size */
422 	struct rw_semaphore i_rw_mutex;
423 	struct list_head i_ordered;
424 	struct list_head i_trunc_list;
425 	__be64 *i_hash_cache;
426 	u32 i_entries;
427 	u32 i_diskflags;
428 	u8 i_height;
429 	u8 i_depth;
430 	u16 i_rahead;
431 };
432 
433 /*
434  * Since i_inode is the first element of struct gfs2_inode,
435  * this is effectively a cast.
436  */
437 static inline struct gfs2_inode *GFS2_I(struct inode *inode)
438 {
439 	return container_of(inode, struct gfs2_inode, i_inode);
440 }
441 
442 static inline struct gfs2_sbd *GFS2_SB(const struct inode *inode)
443 {
444 	return inode->i_sb->s_fs_info;
445 }
446 
447 struct gfs2_file {
448 	struct mutex f_fl_mutex;
449 	struct gfs2_holder f_fl_gh;
450 };
451 
452 struct gfs2_revoke_replay {
453 	struct list_head rr_list;
454 	u64 rr_blkno;
455 	unsigned int rr_where;
456 };
457 
458 enum {
459 	QDF_CHANGE		= 1,
460 	QDF_LOCKED		= 2,
461 	QDF_REFRESH		= 3,
462 	QDF_QMSG_QUIET          = 4,
463 };
464 
465 struct gfs2_quota_data {
466 	struct hlist_bl_node qd_hlist;
467 	struct list_head qd_list;
468 	struct kqid qd_id;
469 	struct gfs2_sbd *qd_sbd;
470 	struct lockref qd_lockref;
471 	struct list_head qd_lru;
472 	unsigned qd_hash;
473 
474 	unsigned long qd_flags;		/* QDF_... */
475 
476 	s64 qd_change;
477 	s64 qd_change_sync;
478 
479 	unsigned int qd_slot;
480 	unsigned int qd_slot_count;
481 
482 	struct buffer_head *qd_bh;
483 	struct gfs2_quota_change *qd_bh_qc;
484 	unsigned int qd_bh_count;
485 
486 	struct gfs2_glock *qd_gl;
487 	struct gfs2_quota_lvb qd_qb;
488 
489 	u64 qd_sync_gen;
490 	unsigned long qd_last_warn;
491 	struct rcu_head qd_rcu;
492 };
493 
494 enum {
495 	TR_TOUCHED = 1,
496 	TR_ATTACHED = 2,
497 	TR_ALLOCED = 3,
498 };
499 
500 struct gfs2_trans {
501 	unsigned long tr_ip;
502 
503 	unsigned int tr_blocks;
504 	unsigned int tr_revokes;
505 	unsigned int tr_reserved;
506 	unsigned long tr_flags;
507 
508 	unsigned int tr_num_buf_new;
509 	unsigned int tr_num_databuf_new;
510 	unsigned int tr_num_buf_rm;
511 	unsigned int tr_num_databuf_rm;
512 	unsigned int tr_num_revoke;
513 	unsigned int tr_num_revoke_rm;
514 
515 	struct list_head tr_list;
516 	struct list_head tr_databuf;
517 	struct list_head tr_buf;
518 
519 	unsigned int tr_first;
520 	struct list_head tr_ail1_list;
521 	struct list_head tr_ail2_list;
522 };
523 
524 struct gfs2_journal_extent {
525 	struct list_head list;
526 
527 	unsigned int lblock; /* First logical block */
528 	u64 dblock; /* First disk block */
529 	u64 blocks;
530 };
531 
532 struct gfs2_jdesc {
533 	struct list_head jd_list;
534 	struct list_head extent_list;
535 	unsigned int nr_extents;
536 	struct work_struct jd_work;
537 	struct inode *jd_inode;
538 	unsigned long jd_flags;
539 #define JDF_RECOVERY 1
540 	unsigned int jd_jid;
541 	u32 jd_blocks;
542 	int jd_recover_error;
543 	/* Replay stuff */
544 
545 	unsigned int jd_found_blocks;
546 	unsigned int jd_found_revokes;
547 	unsigned int jd_replayed_blocks;
548 
549 	struct list_head jd_revoke_list;
550 	unsigned int jd_replay_tail;
551 
552 	u64 jd_no_addr;
553 };
554 
555 struct gfs2_statfs_change_host {
556 	s64 sc_total;
557 	s64 sc_free;
558 	s64 sc_dinodes;
559 };
560 
561 #define GFS2_QUOTA_DEFAULT	GFS2_QUOTA_OFF
562 #define GFS2_QUOTA_OFF		0
563 #define GFS2_QUOTA_ACCOUNT	1
564 #define GFS2_QUOTA_ON		2
565 
566 #define GFS2_DATA_DEFAULT	GFS2_DATA_ORDERED
567 #define GFS2_DATA_WRITEBACK	1
568 #define GFS2_DATA_ORDERED	2
569 
570 #define GFS2_ERRORS_DEFAULT     GFS2_ERRORS_WITHDRAW
571 #define GFS2_ERRORS_WITHDRAW    0
572 #define GFS2_ERRORS_CONTINUE    1 /* place holder for future feature */
573 #define GFS2_ERRORS_RO          2 /* place holder for future feature */
574 #define GFS2_ERRORS_PANIC       3
575 
576 struct gfs2_args {
577 	char ar_lockproto[GFS2_LOCKNAME_LEN];	/* Name of the Lock Protocol */
578 	char ar_locktable[GFS2_LOCKNAME_LEN];	/* Name of the Lock Table */
579 	char ar_hostdata[GFS2_LOCKNAME_LEN];	/* Host specific data */
580 	unsigned int ar_spectator:1;		/* Don't get a journal */
581 	unsigned int ar_localflocks:1;		/* Let the VFS do flock|fcntl */
582 	unsigned int ar_debug:1;		/* Oops on errors */
583 	unsigned int ar_posix_acl:1;		/* Enable posix acls */
584 	unsigned int ar_quota:2;		/* off/account/on */
585 	unsigned int ar_suiddir:1;		/* suiddir support */
586 	unsigned int ar_data:2;			/* ordered/writeback */
587 	unsigned int ar_meta:1;			/* mount metafs */
588 	unsigned int ar_discard:1;		/* discard requests */
589 	unsigned int ar_errors:2;               /* errors=withdraw | panic */
590 	unsigned int ar_nobarrier:1;            /* do not send barriers */
591 	unsigned int ar_rgrplvb:1;		/* use lvbs for rgrp info */
592 	unsigned int ar_loccookie:1;		/* use location based readdir
593 						   cookies */
594 	s32 ar_commit;				/* Commit interval */
595 	s32 ar_statfs_quantum;			/* The fast statfs interval */
596 	s32 ar_quota_quantum;			/* The quota interval */
597 	s32 ar_statfs_percent;			/* The % change to force sync */
598 };
599 
600 struct gfs2_tune {
601 	spinlock_t gt_spin;
602 
603 	unsigned int gt_logd_secs;
604 
605 	unsigned int gt_quota_warn_period; /* Secs between quota warn msgs */
606 	unsigned int gt_quota_scale_num; /* Numerator */
607 	unsigned int gt_quota_scale_den; /* Denominator */
608 	unsigned int gt_quota_quantum; /* Secs between syncs to quota file */
609 	unsigned int gt_new_files_jdata;
610 	unsigned int gt_max_readahead; /* Max bytes to read-ahead from disk */
611 	unsigned int gt_complain_secs;
612 	unsigned int gt_statfs_quantum;
613 	unsigned int gt_statfs_slow;
614 };
615 
616 enum {
617 	SDF_JOURNAL_CHECKED	= 0,
618 	SDF_JOURNAL_LIVE	= 1,
619 	SDF_WITHDRAWN		= 2,
620 	SDF_NOBARRIERS		= 3,
621 	SDF_NORECOVERY		= 4,
622 	SDF_DEMOTE		= 5,
623 	SDF_NOJOURNALID		= 6,
624 	SDF_RORECOVERY		= 7, /* read only recovery */
625 	SDF_SKIP_DLM_UNLOCK	= 8,
626 	SDF_FORCE_AIL_FLUSH     = 9,
627 	SDF_FS_FROZEN           = 10,
628 	SDF_WITHDRAWING		= 11, /* Will withdraw eventually */
629 	SDF_WITHDRAW_IN_PROG	= 12, /* Withdraw is in progress */
630 	SDF_REMOTE_WITHDRAW	= 13, /* Performing remote recovery */
631 	SDF_WITHDRAW_RECOVERY	= 14, /* Wait for journal recovery when we are
632 					 withdrawing */
633 };
634 
635 enum gfs2_freeze_state {
636 	SFS_UNFROZEN		= 0,
637 	SFS_STARTING_FREEZE	= 1,
638 	SFS_FROZEN		= 2,
639 };
640 
641 #define GFS2_FSNAME_LEN		256
642 
643 struct gfs2_inum_host {
644 	u64 no_formal_ino;
645 	u64 no_addr;
646 };
647 
648 struct gfs2_sb_host {
649 	u32 sb_magic;
650 	u32 sb_type;
651 	u32 sb_format;
652 
653 	u32 sb_fs_format;
654 	u32 sb_multihost_format;
655 	u32 sb_bsize;
656 	u32 sb_bsize_shift;
657 
658 	struct gfs2_inum_host sb_master_dir;
659 	struct gfs2_inum_host sb_root_dir;
660 
661 	char sb_lockproto[GFS2_LOCKNAME_LEN];
662 	char sb_locktable[GFS2_LOCKNAME_LEN];
663 };
664 
665 /*
666  * lm_mount() return values
667  *
668  * ls_jid - the journal ID this node should use
669  * ls_first - this node is the first to mount the file system
670  * ls_lockspace - lock module's context for this file system
671  * ls_ops - lock module's functions
672  */
673 
674 struct lm_lockstruct {
675 	int ls_jid;
676 	unsigned int ls_first;
677 	const struct lm_lockops *ls_ops;
678 	dlm_lockspace_t *ls_dlm;
679 
680 	int ls_recover_jid_done;   /* These two are deprecated, */
681 	int ls_recover_jid_status; /* used previously by gfs_controld */
682 
683 	struct dlm_lksb ls_mounted_lksb; /* mounted_lock */
684 	struct dlm_lksb ls_control_lksb; /* control_lock */
685 	char ls_control_lvb[GDLM_LVB_SIZE]; /* control_lock lvb */
686 	struct completion ls_sync_wait; /* {control,mounted}_{lock,unlock} */
687 	char *ls_lvb_bits;
688 
689 	spinlock_t ls_recover_spin; /* protects following fields */
690 	unsigned long ls_recover_flags; /* DFL_ */
691 	uint32_t ls_recover_mount; /* gen in first recover_done cb */
692 	uint32_t ls_recover_start; /* gen in last recover_done cb */
693 	uint32_t ls_recover_block; /* copy recover_start in last recover_prep */
694 	uint32_t ls_recover_size; /* size of recover_submit, recover_result */
695 	uint32_t *ls_recover_submit; /* gen in last recover_slot cb per jid */
696 	uint32_t *ls_recover_result; /* result of last jid recovery */
697 };
698 
699 struct gfs2_pcpu_lkstats {
700 	/* One struct for each glock type */
701 	struct gfs2_lkstats lkstats[10];
702 };
703 
704 struct gfs2_sbd {
705 	struct super_block *sd_vfs;
706 	struct gfs2_pcpu_lkstats __percpu *sd_lkstats;
707 	struct kobject sd_kobj;
708 	unsigned long sd_flags;	/* SDF_... */
709 	struct gfs2_sb_host sd_sb;
710 
711 	/* Constants computed on mount */
712 
713 	u32 sd_fsb2bb;
714 	u32 sd_fsb2bb_shift;
715 	u32 sd_diptrs;	/* Number of pointers in a dinode */
716 	u32 sd_inptrs;	/* Number of pointers in a indirect block */
717 	u32 sd_ldptrs;  /* Number of pointers in a log descriptor block */
718 	u32 sd_jbsize;	/* Size of a journaled data block */
719 	u32 sd_hash_bsize;	/* sizeof(exhash block) */
720 	u32 sd_hash_bsize_shift;
721 	u32 sd_hash_ptrs;	/* Number of pointers in a hash block */
722 	u32 sd_qc_per_block;
723 	u32 sd_blocks_per_bitmap;
724 	u32 sd_max_dirres;	/* Max blocks needed to add a directory entry */
725 	u32 sd_max_height;	/* Max height of a file's metadata tree */
726 	u64 sd_heightsize[GFS2_MAX_META_HEIGHT + 1];
727 	u32 sd_max_dents_per_leaf; /* Max number of dirents in a leaf block */
728 
729 	struct gfs2_args sd_args;	/* Mount arguments */
730 	struct gfs2_tune sd_tune;	/* Filesystem tuning structure */
731 
732 	/* Lock Stuff */
733 
734 	struct lm_lockstruct sd_lockstruct;
735 	struct gfs2_holder sd_live_gh;
736 	struct gfs2_glock *sd_rename_gl;
737 	struct gfs2_glock *sd_freeze_gl;
738 	struct work_struct sd_freeze_work;
739 	wait_queue_head_t sd_glock_wait;
740 	wait_queue_head_t sd_async_glock_wait;
741 	atomic_t sd_glock_disposal;
742 	struct completion sd_locking_init;
743 	struct completion sd_wdack;
744 	struct delayed_work sd_control_work;
745 
746 	/* Inode Stuff */
747 
748 	struct dentry *sd_master_dir;
749 	struct dentry *sd_root_dir;
750 
751 	struct inode *sd_jindex;
752 	struct inode *sd_statfs_inode;
753 	struct inode *sd_sc_inode;
754 	struct inode *sd_qc_inode;
755 	struct inode *sd_rindex;
756 	struct inode *sd_quota_inode;
757 
758 	/* StatFS stuff */
759 
760 	spinlock_t sd_statfs_spin;
761 	struct gfs2_statfs_change_host sd_statfs_master;
762 	struct gfs2_statfs_change_host sd_statfs_local;
763 	int sd_statfs_force_sync;
764 
765 	/* Resource group stuff */
766 
767 	int sd_rindex_uptodate;
768 	spinlock_t sd_rindex_spin;
769 	struct rb_root sd_rindex_tree;
770 	unsigned int sd_rgrps;
771 	unsigned int sd_max_rg_data;
772 
773 	/* Journal index stuff */
774 
775 	struct list_head sd_jindex_list;
776 	spinlock_t sd_jindex_spin;
777 	struct mutex sd_jindex_mutex;
778 	unsigned int sd_journals;
779 
780 	struct gfs2_jdesc *sd_jdesc;
781 	struct gfs2_holder sd_journal_gh;
782 	struct gfs2_holder sd_jinode_gh;
783 	struct gfs2_glock *sd_jinode_gl;
784 
785 	struct gfs2_holder sd_sc_gh;
786 	struct gfs2_holder sd_qc_gh;
787 
788 	struct completion sd_journal_ready;
789 
790 	/* Daemon stuff */
791 
792 	struct task_struct *sd_logd_process;
793 	struct task_struct *sd_quotad_process;
794 
795 	/* Quota stuff */
796 
797 	struct list_head sd_quota_list;
798 	atomic_t sd_quota_count;
799 	struct mutex sd_quota_mutex;
800 	struct mutex sd_quota_sync_mutex;
801 	wait_queue_head_t sd_quota_wait;
802 	struct list_head sd_trunc_list;
803 	spinlock_t sd_trunc_lock;
804 
805 	unsigned int sd_quota_slots;
806 	unsigned long *sd_quota_bitmap;
807 	spinlock_t sd_bitmap_lock;
808 
809 	u64 sd_quota_sync_gen;
810 
811 	/* Log stuff */
812 
813 	struct address_space sd_aspace;
814 
815 	spinlock_t sd_log_lock;
816 
817 	struct gfs2_trans *sd_log_tr;
818 	unsigned int sd_log_blks_reserved;
819 	int sd_log_committed_revoke;
820 
821 	atomic_t sd_log_pinned;
822 	unsigned int sd_log_num_revoke;
823 
824 	struct list_head sd_log_revokes;
825 	struct list_head sd_log_ordered;
826 	spinlock_t sd_ordered_lock;
827 
828 	atomic_t sd_log_thresh1;
829 	atomic_t sd_log_thresh2;
830 	atomic_t sd_log_blks_free;
831 	atomic_t sd_log_blks_needed;
832 	wait_queue_head_t sd_log_waitq;
833 	wait_queue_head_t sd_logd_waitq;
834 
835 	u64 sd_log_sequence;
836 	unsigned int sd_log_head;
837 	unsigned int sd_log_tail;
838 	int sd_log_idle;
839 
840 	struct rw_semaphore sd_log_flush_lock;
841 	atomic_t sd_log_in_flight;
842 	struct bio *sd_log_bio;
843 	wait_queue_head_t sd_log_flush_wait;
844 	int sd_log_error; /* First log error */
845 	wait_queue_head_t sd_withdraw_wait;
846 
847 	atomic_t sd_reserving_log;
848 	wait_queue_head_t sd_reserving_log_wait;
849 
850 	unsigned int sd_log_flush_head;
851 
852 	spinlock_t sd_ail_lock;
853 	struct list_head sd_ail1_list;
854 	struct list_head sd_ail2_list;
855 
856 	/* For quiescing the filesystem */
857 	struct gfs2_holder sd_freeze_gh;
858 	atomic_t sd_freeze_state;
859 	struct mutex sd_freeze_mutex;
860 
861 	char sd_fsname[GFS2_FSNAME_LEN + 3 * sizeof(int) + 2];
862 	char sd_table_name[GFS2_FSNAME_LEN];
863 	char sd_proto_name[GFS2_FSNAME_LEN];
864 
865 	/* Debugging crud */
866 
867 	unsigned long sd_last_warning;
868 	struct dentry *debugfs_dir;    /* debugfs directory */
869 	unsigned long sd_glock_dqs_held;
870 };
871 
872 static inline void gfs2_glstats_inc(struct gfs2_glock *gl, int which)
873 {
874 	gl->gl_stats.stats[which]++;
875 }
876 
877 static inline void gfs2_sbstats_inc(const struct gfs2_glock *gl, int which)
878 {
879 	const struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
880 	preempt_disable();
881 	this_cpu_ptr(sdp->sd_lkstats)->lkstats[gl->gl_name.ln_type].stats[which]++;
882 	preempt_enable();
883 }
884 
885 extern struct gfs2_rgrpd *gfs2_glock2rgrp(struct gfs2_glock *gl);
886 
887 static inline unsigned gfs2_max_stuffed_size(const struct gfs2_inode *ip)
888 {
889 	return GFS2_SB(&ip->i_inode)->sd_sb.sb_bsize - sizeof(struct gfs2_dinode);
890 }
891 
892 #endif /* __INCORE_DOT_H__ */
893 
894