fs/block_dev.c: fix performance regression in O_DIRECT|O_SYNC writes to block devices
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / gfs2 / incore.h
blob3aac46f6853e620f818a86ec3022ac846bc90515
1 /*
2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
3 * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
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 */
10 #ifndef __INCORE_DOT_H__
11 #define __INCORE_DOT_H__
13 #include <linux/fs.h>
14 #include <linux/workqueue.h>
15 #include <linux/slow-work.h>
16 #include <linux/dlm.h>
17 #include <linux/buffer_head.h>
19 #define DIO_WAIT 0x00000010
20 #define DIO_METADATA 0x00000020
21 #define DIO_ALL 0x00000100
23 struct gfs2_log_operations;
24 struct gfs2_log_element;
25 struct gfs2_holder;
26 struct gfs2_glock;
27 struct gfs2_quota_data;
28 struct gfs2_trans;
29 struct gfs2_ail;
30 struct gfs2_jdesc;
31 struct gfs2_sbd;
32 struct lm_lockops;
34 typedef void (*gfs2_glop_bh_t) (struct gfs2_glock *gl, unsigned int ret);
36 struct gfs2_log_header_host {
37 u64 lh_sequence; /* Sequence number of this transaction */
38 u32 lh_flags; /* GFS2_LOG_HEAD_... */
39 u32 lh_tail; /* Block number of log tail */
40 u32 lh_blkno;
41 u32 lh_hash;
45 * Structure of operations that are associated with each
46 * type of element in the log.
49 struct gfs2_log_operations {
50 void (*lo_add) (struct gfs2_sbd *sdp, struct gfs2_log_element *le);
51 void (*lo_before_commit) (struct gfs2_sbd *sdp);
52 void (*lo_after_commit) (struct gfs2_sbd *sdp, struct gfs2_ail *ai);
53 void (*lo_before_scan) (struct gfs2_jdesc *jd,
54 struct gfs2_log_header_host *head, int pass);
55 int (*lo_scan_elements) (struct gfs2_jdesc *jd, unsigned int start,
56 struct gfs2_log_descriptor *ld, __be64 *ptr,
57 int pass);
58 void (*lo_after_scan) (struct gfs2_jdesc *jd, int error, int pass);
59 const char *lo_name;
62 struct gfs2_log_element {
63 struct list_head le_list;
64 const struct gfs2_log_operations *le_ops;
67 #define GBF_FULL 1
69 struct gfs2_bitmap {
70 struct buffer_head *bi_bh;
71 char *bi_clone;
72 unsigned long bi_flags;
73 u32 bi_offset;
74 u32 bi_start;
75 u32 bi_len;
78 struct gfs2_rgrpd {
79 struct list_head rd_list; /* Link with superblock */
80 struct list_head rd_list_mru;
81 struct gfs2_glock *rd_gl; /* Glock for this rgrp */
82 u64 rd_addr; /* grp block disk address */
83 u64 rd_data0; /* first data location */
84 u32 rd_length; /* length of rgrp header in fs blocks */
85 u32 rd_data; /* num of data blocks in rgrp */
86 u32 rd_bitbytes; /* number of bytes in data bitmaps */
87 u32 rd_free;
88 u32 rd_free_clone;
89 u32 rd_dinodes;
90 u64 rd_igeneration;
91 struct gfs2_bitmap *rd_bits;
92 struct mutex rd_mutex;
93 struct gfs2_log_element rd_le;
94 struct gfs2_sbd *rd_sbd;
95 unsigned int rd_bh_count;
96 u32 rd_last_alloc;
97 u32 rd_flags;
98 #define GFS2_RDF_CHECK 0x10000000 /* check for unlinked inodes */
99 #define GFS2_RDF_UPTODATE 0x20000000 /* rg is up to date */
100 #define GFS2_RDF_ERROR 0x40000000 /* error in rg */
101 #define GFS2_RDF_MASK 0xf0000000 /* mask for internal flags */
104 enum gfs2_state_bits {
105 BH_Pinned = BH_PrivateStart,
106 BH_Escaped = BH_PrivateStart + 1,
109 BUFFER_FNS(Pinned, pinned)
110 TAS_BUFFER_FNS(Pinned, pinned)
111 BUFFER_FNS(Escaped, escaped)
112 TAS_BUFFER_FNS(Escaped, escaped)
114 struct gfs2_bufdata {
115 struct buffer_head *bd_bh;
116 struct gfs2_glock *bd_gl;
118 union {
119 struct list_head list_tr;
120 u64 blkno;
121 } u;
122 #define bd_list_tr u.list_tr
123 #define bd_blkno u.blkno
125 struct gfs2_log_element bd_le;
127 struct gfs2_ail *bd_ail;
128 struct list_head bd_ail_st_list;
129 struct list_head bd_ail_gl_list;
133 * Internally, we prefix things with gdlm_ and GDLM_ (for gfs-dlm) since a
134 * prefix of lock_dlm_ gets awkward.
137 #define GDLM_STRNAME_BYTES 25
138 #define GDLM_LVB_SIZE 32
140 enum {
141 DFL_BLOCK_LOCKS = 0,
144 struct lm_lockname {
145 u64 ln_number;
146 unsigned int ln_type;
149 #define lm_name_equal(name1, name2) \
150 (((name1)->ln_number == (name2)->ln_number) && \
151 ((name1)->ln_type == (name2)->ln_type))
154 struct gfs2_glock_operations {
155 void (*go_xmote_th) (struct gfs2_glock *gl);
156 int (*go_xmote_bh) (struct gfs2_glock *gl, struct gfs2_holder *gh);
157 void (*go_inval) (struct gfs2_glock *gl, int flags);
158 int (*go_demote_ok) (const struct gfs2_glock *gl);
159 int (*go_lock) (struct gfs2_holder *gh);
160 void (*go_unlock) (struct gfs2_holder *gh);
161 int (*go_dump)(struct seq_file *seq, const struct gfs2_glock *gl);
162 void (*go_callback) (struct gfs2_glock *gl);
163 const int go_type;
164 const unsigned long go_min_hold_time;
165 const unsigned long go_flags;
166 #define GLOF_ASPACE 1
169 enum {
170 /* States */
171 HIF_HOLDER = 6, /* Set for gh that "holds" the glock */
172 HIF_FIRST = 7,
173 HIF_WAIT = 10,
176 struct gfs2_holder {
177 struct list_head gh_list;
179 struct gfs2_glock *gh_gl;
180 struct pid *gh_owner_pid;
181 unsigned int gh_state;
182 unsigned gh_flags;
184 int gh_error;
185 unsigned long gh_iflags; /* HIF_... */
186 unsigned long gh_ip;
189 enum {
190 GLF_LOCK = 1,
191 GLF_DEMOTE = 3,
192 GLF_PENDING_DEMOTE = 4,
193 GLF_DEMOTE_IN_PROGRESS = 5,
194 GLF_DIRTY = 6,
195 GLF_LFLUSH = 7,
196 GLF_INVALIDATE_IN_PROGRESS = 8,
197 GLF_REPLY_PENDING = 9,
198 GLF_INITIAL = 10,
199 GLF_FROZEN = 11,
202 struct gfs2_glock {
203 struct hlist_node gl_list;
204 unsigned long gl_flags; /* GLF_... */
205 struct lm_lockname gl_name;
206 atomic_t gl_ref;
208 spinlock_t gl_spin;
210 unsigned int gl_state;
211 unsigned int gl_target;
212 unsigned int gl_reply;
213 unsigned int gl_hash;
214 unsigned int gl_req;
215 unsigned int gl_demote_state; /* state requested by remote node */
216 unsigned long gl_demote_time; /* time of first demote request */
217 struct list_head gl_holders;
219 const struct gfs2_glock_operations *gl_ops;
220 char gl_strname[GDLM_STRNAME_BYTES];
221 struct dlm_lksb gl_lksb;
222 char gl_lvb[32];
223 unsigned long gl_tchange;
224 void *gl_object;
226 struct list_head gl_lru;
228 struct gfs2_sbd *gl_sbd;
230 struct list_head gl_ail_list;
231 atomic_t gl_ail_count;
232 struct delayed_work gl_work;
233 struct work_struct gl_delete;
236 #define GFS2_MIN_LVB_SIZE 32 /* Min size of LVB that gfs2 supports */
238 struct gfs2_alloc {
239 /* Quota stuff */
241 struct gfs2_quota_data *al_qd[2*MAXQUOTAS];
242 struct gfs2_holder al_qd_ghs[2*MAXQUOTAS];
243 unsigned int al_qd_num;
245 u32 al_requested; /* Filled in by caller of gfs2_inplace_reserve() */
246 u32 al_alloced; /* Filled in by gfs2_alloc_*() */
248 /* Filled in by gfs2_inplace_reserve() */
250 unsigned int al_line;
251 char *al_file;
252 struct gfs2_holder al_ri_gh;
253 struct gfs2_holder al_rgd_gh;
254 struct gfs2_rgrpd *al_rgd;
258 enum {
259 GIF_INVALID = 0,
260 GIF_QD_LOCKED = 1,
261 GIF_SW_PAGED = 3,
265 struct gfs2_inode {
266 struct inode i_inode;
267 u64 i_no_addr;
268 u64 i_no_formal_ino;
269 u64 i_generation;
270 u64 i_eattr;
271 loff_t i_disksize;
272 unsigned long i_flags; /* GIF_... */
273 struct gfs2_glock *i_gl; /* Move into i_gh? */
274 struct gfs2_holder i_iopen_gh;
275 struct gfs2_holder i_gh; /* for prepare/commit_write only */
276 struct gfs2_alloc *i_alloc;
277 u64 i_goal; /* goal block for allocations */
278 struct rw_semaphore i_rw_mutex;
279 struct list_head i_trunc_list;
280 u32 i_entries;
281 u32 i_diskflags;
282 u8 i_height;
283 u8 i_depth;
287 * Since i_inode is the first element of struct gfs2_inode,
288 * this is effectively a cast.
290 static inline struct gfs2_inode *GFS2_I(struct inode *inode)
292 return container_of(inode, struct gfs2_inode, i_inode);
295 static inline struct gfs2_sbd *GFS2_SB(const struct inode *inode)
297 return inode->i_sb->s_fs_info;
300 struct gfs2_file {
301 struct mutex f_fl_mutex;
302 struct gfs2_holder f_fl_gh;
305 struct gfs2_revoke_replay {
306 struct list_head rr_list;
307 u64 rr_blkno;
308 unsigned int rr_where;
311 enum {
312 QDF_USER = 0,
313 QDF_CHANGE = 1,
314 QDF_LOCKED = 2,
317 struct gfs2_quota_data {
318 struct list_head qd_list;
319 struct list_head qd_reclaim;
321 atomic_t qd_count;
323 u32 qd_id;
324 unsigned long qd_flags; /* QDF_... */
326 s64 qd_change;
327 s64 qd_change_sync;
329 unsigned int qd_slot;
330 unsigned int qd_slot_count;
332 struct buffer_head *qd_bh;
333 struct gfs2_quota_change *qd_bh_qc;
334 unsigned int qd_bh_count;
336 struct gfs2_glock *qd_gl;
337 struct gfs2_quota_lvb qd_qb;
339 u64 qd_sync_gen;
340 unsigned long qd_last_warn;
343 struct gfs2_trans {
344 unsigned long tr_ip;
346 unsigned int tr_blocks;
347 unsigned int tr_revokes;
348 unsigned int tr_reserved;
350 struct gfs2_holder tr_t_gh;
352 int tr_touched;
354 unsigned int tr_num_buf;
355 unsigned int tr_num_buf_new;
356 unsigned int tr_num_databuf_new;
357 unsigned int tr_num_buf_rm;
358 unsigned int tr_num_databuf_rm;
359 struct list_head tr_list_buf;
361 unsigned int tr_num_revoke;
362 unsigned int tr_num_revoke_rm;
365 struct gfs2_ail {
366 struct list_head ai_list;
368 unsigned int ai_first;
369 struct list_head ai_ail1_list;
370 struct list_head ai_ail2_list;
372 u64 ai_sync_gen;
375 struct gfs2_journal_extent {
376 struct list_head extent_list;
378 unsigned int lblock; /* First logical block */
379 u64 dblock; /* First disk block */
380 u64 blocks;
383 struct gfs2_jdesc {
384 struct list_head jd_list;
385 struct list_head extent_list;
386 struct slow_work jd_work;
387 struct inode *jd_inode;
388 unsigned long jd_flags;
389 #define JDF_RECOVERY 1
390 unsigned int jd_jid;
391 unsigned int jd_blocks;
394 struct gfs2_statfs_change_host {
395 s64 sc_total;
396 s64 sc_free;
397 s64 sc_dinodes;
400 #define GFS2_QUOTA_DEFAULT GFS2_QUOTA_OFF
401 #define GFS2_QUOTA_OFF 0
402 #define GFS2_QUOTA_ACCOUNT 1
403 #define GFS2_QUOTA_ON 2
405 #define GFS2_DATA_DEFAULT GFS2_DATA_ORDERED
406 #define GFS2_DATA_WRITEBACK 1
407 #define GFS2_DATA_ORDERED 2
409 #define GFS2_ERRORS_DEFAULT GFS2_ERRORS_WITHDRAW
410 #define GFS2_ERRORS_WITHDRAW 0
411 #define GFS2_ERRORS_CONTINUE 1 /* place holder for future feature */
412 #define GFS2_ERRORS_RO 2 /* place holder for future feature */
413 #define GFS2_ERRORS_PANIC 3
415 struct gfs2_args {
416 char ar_lockproto[GFS2_LOCKNAME_LEN]; /* Name of the Lock Protocol */
417 char ar_locktable[GFS2_LOCKNAME_LEN]; /* Name of the Lock Table */
418 char ar_hostdata[GFS2_LOCKNAME_LEN]; /* Host specific data */
419 unsigned int ar_spectator:1; /* Don't get a journal */
420 unsigned int ar_ignore_local_fs:1; /* Ignore optimisations */
421 unsigned int ar_localflocks:1; /* Let the VFS do flock|fcntl */
422 unsigned int ar_localcaching:1; /* Local caching */
423 unsigned int ar_debug:1; /* Oops on errors */
424 unsigned int ar_upgrade:1; /* Upgrade ondisk format */
425 unsigned int ar_posix_acl:1; /* Enable posix acls */
426 unsigned int ar_quota:2; /* off/account/on */
427 unsigned int ar_suiddir:1; /* suiddir support */
428 unsigned int ar_data:2; /* ordered/writeback */
429 unsigned int ar_meta:1; /* mount metafs */
430 unsigned int ar_discard:1; /* discard requests */
431 unsigned int ar_errors:2; /* errors=withdraw | panic */
432 unsigned int ar_nobarrier:1; /* do not send barriers */
433 int ar_commit; /* Commit interval */
434 int ar_statfs_quantum; /* The fast statfs interval */
435 int ar_quota_quantum; /* The quota interval */
436 int ar_statfs_percent; /* The % change to force sync */
439 struct gfs2_tune {
440 spinlock_t gt_spin;
442 unsigned int gt_incore_log_blocks;
443 unsigned int gt_log_flush_secs;
445 unsigned int gt_logd_secs;
447 unsigned int gt_quota_simul_sync; /* Max quotavals to sync at once */
448 unsigned int gt_quota_warn_period; /* Secs between quota warn msgs */
449 unsigned int gt_quota_scale_num; /* Numerator */
450 unsigned int gt_quota_scale_den; /* Denominator */
451 unsigned int gt_quota_quantum; /* Secs between syncs to quota file */
452 unsigned int gt_new_files_jdata;
453 unsigned int gt_max_readahead; /* Max bytes to read-ahead from disk */
454 unsigned int gt_complain_secs;
455 unsigned int gt_statfs_quantum;
456 unsigned int gt_statfs_slow;
459 enum {
460 SDF_JOURNAL_CHECKED = 0,
461 SDF_JOURNAL_LIVE = 1,
462 SDF_SHUTDOWN = 2,
463 SDF_NOBARRIERS = 3,
464 SDF_NORECOVERY = 4,
467 #define GFS2_FSNAME_LEN 256
469 struct gfs2_inum_host {
470 u64 no_formal_ino;
471 u64 no_addr;
474 struct gfs2_sb_host {
475 u32 sb_magic;
476 u32 sb_type;
477 u32 sb_format;
479 u32 sb_fs_format;
480 u32 sb_multihost_format;
481 u32 sb_bsize;
482 u32 sb_bsize_shift;
484 struct gfs2_inum_host sb_master_dir;
485 struct gfs2_inum_host sb_root_dir;
487 char sb_lockproto[GFS2_LOCKNAME_LEN];
488 char sb_locktable[GFS2_LOCKNAME_LEN];
489 u8 sb_uuid[16];
493 * lm_mount() return values
495 * ls_jid - the journal ID this node should use
496 * ls_first - this node is the first to mount the file system
497 * ls_lockspace - lock module's context for this file system
498 * ls_ops - lock module's functions
501 struct lm_lockstruct {
502 unsigned int ls_jid;
503 unsigned int ls_first;
504 unsigned int ls_first_done;
505 unsigned int ls_nodir;
506 const struct lm_lockops *ls_ops;
507 unsigned long ls_flags;
508 dlm_lockspace_t *ls_dlm;
510 int ls_recover_jid_done;
511 int ls_recover_jid_status;
514 struct gfs2_sbd {
515 struct super_block *sd_vfs;
516 struct kobject sd_kobj;
517 unsigned long sd_flags; /* SDF_... */
518 struct gfs2_sb_host sd_sb;
520 /* Constants computed on mount */
522 u32 sd_fsb2bb;
523 u32 sd_fsb2bb_shift;
524 u32 sd_diptrs; /* Number of pointers in a dinode */
525 u32 sd_inptrs; /* Number of pointers in a indirect block */
526 u32 sd_jbsize; /* Size of a journaled data block */
527 u32 sd_hash_bsize; /* sizeof(exhash block) */
528 u32 sd_hash_bsize_shift;
529 u32 sd_hash_ptrs; /* Number of pointers in a hash block */
530 u32 sd_qc_per_block;
531 u32 sd_max_dirres; /* Max blocks needed to add a directory entry */
532 u32 sd_max_height; /* Max height of a file's metadata tree */
533 u64 sd_heightsize[GFS2_MAX_META_HEIGHT + 1];
534 u32 sd_max_jheight; /* Max height of journaled file's meta tree */
535 u64 sd_jheightsize[GFS2_MAX_META_HEIGHT + 1];
537 struct gfs2_args sd_args; /* Mount arguments */
538 struct gfs2_tune sd_tune; /* Filesystem tuning structure */
540 /* Lock Stuff */
542 struct lm_lockstruct sd_lockstruct;
543 struct gfs2_holder sd_live_gh;
544 struct gfs2_glock *sd_rename_gl;
545 struct gfs2_glock *sd_trans_gl;
546 wait_queue_head_t sd_glock_wait;
547 atomic_t sd_glock_disposal;
549 /* Inode Stuff */
551 struct dentry *sd_master_dir;
552 struct dentry *sd_root_dir;
554 struct inode *sd_jindex;
555 struct inode *sd_statfs_inode;
556 struct inode *sd_sc_inode;
557 struct inode *sd_qc_inode;
558 struct inode *sd_rindex;
559 struct inode *sd_quota_inode;
561 /* StatFS stuff */
563 spinlock_t sd_statfs_spin;
564 struct gfs2_statfs_change_host sd_statfs_master;
565 struct gfs2_statfs_change_host sd_statfs_local;
566 int sd_statfs_force_sync;
568 /* Resource group stuff */
570 int sd_rindex_uptodate;
571 spinlock_t sd_rindex_spin;
572 struct mutex sd_rindex_mutex;
573 struct list_head sd_rindex_list;
574 struct list_head sd_rindex_mru_list;
575 struct gfs2_rgrpd *sd_rindex_forward;
576 unsigned int sd_rgrps;
578 /* Journal index stuff */
580 struct list_head sd_jindex_list;
581 spinlock_t sd_jindex_spin;
582 struct mutex sd_jindex_mutex;
583 unsigned int sd_journals;
585 struct gfs2_jdesc *sd_jdesc;
586 struct gfs2_holder sd_journal_gh;
587 struct gfs2_holder sd_jinode_gh;
589 struct gfs2_holder sd_sc_gh;
590 struct gfs2_holder sd_qc_gh;
592 /* Daemon stuff */
594 struct task_struct *sd_logd_process;
595 struct task_struct *sd_quotad_process;
597 /* Quota stuff */
599 struct list_head sd_quota_list;
600 atomic_t sd_quota_count;
601 struct mutex sd_quota_mutex;
602 wait_queue_head_t sd_quota_wait;
603 struct list_head sd_trunc_list;
604 spinlock_t sd_trunc_lock;
606 unsigned int sd_quota_slots;
607 unsigned int sd_quota_chunks;
608 unsigned char **sd_quota_bitmap;
610 u64 sd_quota_sync_gen;
612 /* Log stuff */
614 spinlock_t sd_log_lock;
616 unsigned int sd_log_blks_reserved;
617 unsigned int sd_log_commited_buf;
618 unsigned int sd_log_commited_databuf;
619 int sd_log_commited_revoke;
621 unsigned int sd_log_num_buf;
622 unsigned int sd_log_num_revoke;
623 unsigned int sd_log_num_rg;
624 unsigned int sd_log_num_databuf;
626 struct list_head sd_log_le_buf;
627 struct list_head sd_log_le_revoke;
628 struct list_head sd_log_le_rg;
629 struct list_head sd_log_le_databuf;
630 struct list_head sd_log_le_ordered;
632 atomic_t sd_log_blks_free;
633 struct mutex sd_log_reserve_mutex;
635 u64 sd_log_sequence;
636 unsigned int sd_log_head;
637 unsigned int sd_log_tail;
638 int sd_log_idle;
640 unsigned long sd_log_flush_time;
641 struct rw_semaphore sd_log_flush_lock;
642 atomic_t sd_log_in_flight;
643 wait_queue_head_t sd_log_flush_wait;
645 unsigned int sd_log_flush_head;
646 u64 sd_log_flush_wrapped;
648 struct list_head sd_ail1_list;
649 struct list_head sd_ail2_list;
650 u64 sd_ail_sync_gen;
652 /* Replay stuff */
654 struct list_head sd_revoke_list;
655 unsigned int sd_replay_tail;
657 unsigned int sd_found_blocks;
658 unsigned int sd_found_revokes;
659 unsigned int sd_replayed_blocks;
661 /* For quiescing the filesystem */
663 struct gfs2_holder sd_freeze_gh;
664 struct mutex sd_freeze_lock;
665 unsigned int sd_freeze_count;
667 char sd_fsname[GFS2_FSNAME_LEN];
668 char sd_table_name[GFS2_FSNAME_LEN];
669 char sd_proto_name[GFS2_FSNAME_LEN];
671 /* Debugging crud */
673 unsigned long sd_last_warning;
674 struct dentry *debugfs_dir; /* debugfs directory */
675 struct dentry *debugfs_dentry_glocks; /* for debugfs */
678 #endif /* __INCORE_DOT_H__ */