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.
10 #ifndef __INCORE_DOT_H__
11 #define __INCORE_DOT_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>
25 #define DIO_WAIT 0x00000010
26 #define DIO_METADATA 0x00000020
28 struct gfs2_log_operations
;
32 struct gfs2_quota_data
;
39 typedef void (*gfs2_glop_bh_t
) (struct gfs2_glock
*gl
, unsigned int ret
);
41 struct gfs2_log_header_host
{
42 u64 lh_sequence
; /* Sequence number of this transaction */
43 u32 lh_flags
; /* GFS2_LOG_HEAD_... */
44 u32 lh_tail
; /* Block number of log tail */
50 * Structure of operations that are associated with each
51 * type of element in the log.
54 struct gfs2_log_operations
{
55 void (*lo_add
) (struct gfs2_sbd
*sdp
, struct gfs2_bufdata
*bd
);
56 void (*lo_before_commit
) (struct gfs2_sbd
*sdp
);
57 void (*lo_after_commit
) (struct gfs2_sbd
*sdp
, struct gfs2_ail
*ai
);
58 void (*lo_before_scan
) (struct gfs2_jdesc
*jd
,
59 struct gfs2_log_header_host
*head
, int pass
);
60 int (*lo_scan_elements
) (struct gfs2_jdesc
*jd
, unsigned int start
,
61 struct gfs2_log_descriptor
*ld
, __be64
*ptr
,
63 void (*lo_after_scan
) (struct gfs2_jdesc
*jd
, int error
, int pass
);
70 struct buffer_head
*bi_bh
;
72 unsigned long bi_flags
;
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 */
90 struct gfs2_bitmap
*rd_bits
;
91 struct gfs2_sbd
*rd_sbd
;
94 #define GFS2_RDF_CHECK 0x10000000 /* check for unlinked inodes */
95 #define GFS2_RDF_UPTODATE 0x20000000 /* rg is up to date */
96 #define GFS2_RDF_ERROR 0x40000000 /* error in rg */
97 #define GFS2_RDF_MASK 0xf0000000 /* mask for internal flags */
100 enum gfs2_state_bits
{
101 BH_Pinned
= BH_PrivateStart
,
102 BH_Escaped
= BH_PrivateStart
+ 1,
103 BH_Zeronew
= BH_PrivateStart
+ 2,
106 BUFFER_FNS(Pinned
, pinned
)
107 TAS_BUFFER_FNS(Pinned
, pinned
)
108 BUFFER_FNS(Escaped
, escaped
)
109 TAS_BUFFER_FNS(Escaped
, escaped
)
110 BUFFER_FNS(Zeronew
, zeronew
)
111 TAS_BUFFER_FNS(Zeronew
, zeronew
)
113 struct gfs2_bufdata
{
114 struct buffer_head
*bd_bh
;
115 struct gfs2_glock
*bd_gl
;
118 struct list_head bd_list
;
119 const struct gfs2_log_operations
*bd_ops
;
121 struct gfs2_ail
*bd_ail
;
122 struct list_head bd_ail_st_list
;
123 struct list_head bd_ail_gl_list
;
127 * Internally, we prefix things with gdlm_ and GDLM_ (for gfs-dlm) since a
128 * prefix of lock_dlm_ gets awkward.
131 #define GDLM_STRNAME_BYTES 25
132 #define GDLM_LVB_SIZE 32
137 * DFL_BLOCK_LOCKS: dlm is in recovery and will grant locks that had been
138 * held by failed nodes whose journals need recovery. Those locks should
139 * only be used for journal recovery until the journal recovery is done.
140 * This is set by the dlm recover_prep callback and cleared by the
141 * gfs2_control thread when journal recovery is complete. To avoid
142 * races between recover_prep setting and gfs2_control clearing, recover_spin
143 * is held while changing this bit and reading/writing recover_block
146 * DFL_NO_DLM_OPS: dlm lockspace ops/callbacks are not being used.
148 * DFL_FIRST_MOUNT: this node is the first to mount this fs and is doing
149 * recovery of all journals before allowing other nodes to mount the fs.
150 * This is cleared when FIRST_MOUNT_DONE is set.
152 * DFL_FIRST_MOUNT_DONE: this node was the first mounter, and has finished
153 * recovery of all journals, and now allows other nodes to mount the fs.
155 * DFL_MOUNT_DONE: gdlm_mount has completed successfully and cleared
156 * BLOCK_LOCKS for the first time. The gfs2_control thread should now
157 * control clearing BLOCK_LOCKS for further recoveries.
159 * DFL_UNMOUNT: gdlm_unmount sets to keep sdp off gfs2_control_wq.
161 * DFL_DLM_RECOVERY: set while dlm is in recovery, between recover_prep()
162 * and recover_done(), i.e. set while recover_block == recover_start.
169 DFL_FIRST_MOUNT_DONE
= 3,
172 DFL_DLM_RECOVERY
= 6,
177 unsigned int ln_type
;
180 #define lm_name_equal(name1, name2) \
181 (((name1)->ln_number == (name2)->ln_number) && \
182 ((name1)->ln_type == (name2)->ln_type))
185 struct gfs2_glock_operations
{
186 void (*go_xmote_th
) (struct gfs2_glock
*gl
);
187 int (*go_xmote_bh
) (struct gfs2_glock
*gl
, struct gfs2_holder
*gh
);
188 void (*go_inval
) (struct gfs2_glock
*gl
, int flags
);
189 int (*go_demote_ok
) (const struct gfs2_glock
*gl
);
190 int (*go_lock
) (struct gfs2_holder
*gh
);
191 void (*go_unlock
) (struct gfs2_holder
*gh
);
192 int (*go_dump
)(struct seq_file
*seq
, const struct gfs2_glock
*gl
);
193 void (*go_callback
) (struct gfs2_glock
*gl
);
195 const unsigned long go_flags
;
196 #define GLOF_ASPACE 1
200 GFS2_LKS_SRTT
= 0, /* Non blocking smoothed round trip time */
201 GFS2_LKS_SRTTVAR
= 1, /* Non blocking smoothed variance */
202 GFS2_LKS_SRTTB
= 2, /* Blocking smoothed round trip time */
203 GFS2_LKS_SRTTVARB
= 3, /* Blocking smoothed variance */
204 GFS2_LKS_SIRT
= 4, /* Smoothed Inter-request time */
205 GFS2_LKS_SIRTVAR
= 5, /* Smoothed Inter-request variance */
206 GFS2_LKS_DCOUNT
= 6, /* Count of dlm requests */
207 GFS2_LKS_QCOUNT
= 7, /* Count of gfs2_holder queues */
211 struct gfs2_lkstats
{
212 s64 stats
[GFS2_NR_LKSTATS
];
217 HIF_HOLDER
= 6, /* Set for gh that "holds" the glock */
223 struct list_head gh_list
;
225 struct gfs2_glock
*gh_gl
;
226 struct pid
*gh_owner_pid
;
227 unsigned int gh_state
;
231 unsigned long gh_iflags
; /* HIF_... */
238 GLF_PENDING_DEMOTE
= 4,
239 GLF_DEMOTE_IN_PROGRESS
= 5,
242 GLF_INVALIDATE_IN_PROGRESS
= 8,
243 GLF_REPLY_PENDING
= 9,
248 GLF_OBJECT
= 14, /* Used only for tracing */
253 struct hlist_bl_node gl_list
;
254 struct gfs2_sbd
*gl_sbd
;
255 unsigned long gl_flags
; /* GLF_... */
256 struct lm_lockname gl_name
;
261 /* State fields protected by gl_spin */
262 unsigned int gl_state
:2, /* Current state */
263 gl_target
:2, /* Target state */
264 gl_demote_state
:2, /* State requested by remote node */
265 gl_req
:2, /* State in last dlm request */
266 gl_reply
:8; /* Last reply from the dlm */
268 unsigned int gl_hash
;
269 unsigned long gl_demote_time
; /* time of first demote request */
271 struct list_head gl_holders
;
273 const struct gfs2_glock_operations
*gl_ops
;
275 struct gfs2_lkstats gl_stats
;
276 struct dlm_lksb gl_lksb
;
278 unsigned long gl_tchange
;
281 struct list_head gl_lru
;
282 struct list_head gl_ail_list
;
283 atomic_t gl_ail_count
;
285 struct delayed_work gl_work
;
286 struct work_struct gl_delete
;
287 struct rcu_head gl_rcu
;
290 #define GFS2_MIN_LVB_SIZE 32 /* Min size of LVB that gfs2 supports */
292 struct gfs2_blkreserv
{
293 u32 rs_requested
; /* Filled in by caller of gfs2_inplace_reserve() */
294 struct gfs2_holder rs_rgd_gh
; /* Filled in by gfs2_inplace_reserve() */
296 /* ancillary quota stuff */
297 struct gfs2_quota_data
*rs_qa_qd
[2 * MAXQUOTAS
];
298 struct gfs2_holder rs_qa_qd_ghs
[2 * MAXQUOTAS
];
299 unsigned int rs_qa_qd_num
;
305 GIF_ALLOC_FAILED
= 2,
311 struct inode i_inode
;
316 unsigned long i_flags
; /* GIF_... */
317 struct gfs2_glock
*i_gl
; /* Move into i_gh? */
318 struct gfs2_holder i_iopen_gh
;
319 struct gfs2_holder i_gh
; /* for prepare/commit_write only */
320 struct gfs2_blkreserv
*i_res
; /* resource group block reservation */
321 struct gfs2_rgrpd
*i_rgd
;
322 u64 i_goal
; /* goal block for allocations */
323 struct rw_semaphore i_rw_mutex
;
324 struct list_head i_trunc_list
;
325 __be64
*i_hash_cache
;
333 * Since i_inode is the first element of struct gfs2_inode,
334 * this is effectively a cast.
336 static inline struct gfs2_inode
*GFS2_I(struct inode
*inode
)
338 return container_of(inode
, struct gfs2_inode
, i_inode
);
341 static inline struct gfs2_sbd
*GFS2_SB(const struct inode
*inode
)
343 return inode
->i_sb
->s_fs_info
;
347 struct mutex f_fl_mutex
;
348 struct gfs2_holder f_fl_gh
;
351 struct gfs2_revoke_replay
{
352 struct list_head rr_list
;
354 unsigned int rr_where
;
364 struct gfs2_quota_data
{
365 struct list_head qd_list
;
366 struct list_head qd_reclaim
;
371 unsigned long qd_flags
; /* QDF_... */
376 unsigned int qd_slot
;
377 unsigned int qd_slot_count
;
379 struct buffer_head
*qd_bh
;
380 struct gfs2_quota_change
*qd_bh_qc
;
381 unsigned int qd_bh_count
;
383 struct gfs2_glock
*qd_gl
;
384 struct gfs2_quota_lvb qd_qb
;
387 unsigned long qd_last_warn
;
393 unsigned int tr_blocks
;
394 unsigned int tr_revokes
;
395 unsigned int tr_reserved
;
397 struct gfs2_holder tr_t_gh
;
401 unsigned int tr_num_buf_new
;
402 unsigned int tr_num_databuf_new
;
403 unsigned int tr_num_buf_rm
;
404 unsigned int tr_num_databuf_rm
;
405 unsigned int tr_num_revoke
;
406 unsigned int tr_num_revoke_rm
;
410 struct list_head ai_list
;
412 unsigned int ai_first
;
413 struct list_head ai_ail1_list
;
414 struct list_head ai_ail2_list
;
417 struct gfs2_journal_extent
{
418 struct list_head extent_list
;
420 unsigned int lblock
; /* First logical block */
421 u64 dblock
; /* First disk block */
426 struct list_head jd_list
;
427 struct list_head extent_list
;
428 struct work_struct jd_work
;
429 struct inode
*jd_inode
;
430 unsigned long jd_flags
;
431 #define JDF_RECOVERY 1
433 unsigned int jd_blocks
;
434 int jd_recover_error
;
437 struct gfs2_statfs_change_host
{
443 #define GFS2_QUOTA_DEFAULT GFS2_QUOTA_OFF
444 #define GFS2_QUOTA_OFF 0
445 #define GFS2_QUOTA_ACCOUNT 1
446 #define GFS2_QUOTA_ON 2
448 #define GFS2_DATA_DEFAULT GFS2_DATA_ORDERED
449 #define GFS2_DATA_WRITEBACK 1
450 #define GFS2_DATA_ORDERED 2
452 #define GFS2_ERRORS_DEFAULT GFS2_ERRORS_WITHDRAW
453 #define GFS2_ERRORS_WITHDRAW 0
454 #define GFS2_ERRORS_CONTINUE 1 /* place holder for future feature */
455 #define GFS2_ERRORS_RO 2 /* place holder for future feature */
456 #define GFS2_ERRORS_PANIC 3
459 char ar_lockproto
[GFS2_LOCKNAME_LEN
]; /* Name of the Lock Protocol */
460 char ar_locktable
[GFS2_LOCKNAME_LEN
]; /* Name of the Lock Table */
461 char ar_hostdata
[GFS2_LOCKNAME_LEN
]; /* Host specific data */
462 unsigned int ar_spectator
:1; /* Don't get a journal */
463 unsigned int ar_localflocks
:1; /* Let the VFS do flock|fcntl */
464 unsigned int ar_debug
:1; /* Oops on errors */
465 unsigned int ar_posix_acl
:1; /* Enable posix acls */
466 unsigned int ar_quota
:2; /* off/account/on */
467 unsigned int ar_suiddir
:1; /* suiddir support */
468 unsigned int ar_data
:2; /* ordered/writeback */
469 unsigned int ar_meta
:1; /* mount metafs */
470 unsigned int ar_discard
:1; /* discard requests */
471 unsigned int ar_errors
:2; /* errors=withdraw | panic */
472 unsigned int ar_nobarrier
:1; /* do not send barriers */
473 int ar_commit
; /* Commit interval */
474 int ar_statfs_quantum
; /* The fast statfs interval */
475 int ar_quota_quantum
; /* The quota interval */
476 int ar_statfs_percent
; /* The % change to force sync */
482 unsigned int gt_logd_secs
;
484 unsigned int gt_quota_simul_sync
; /* Max quotavals to sync at once */
485 unsigned int gt_quota_warn_period
; /* Secs between quota warn msgs */
486 unsigned int gt_quota_scale_num
; /* Numerator */
487 unsigned int gt_quota_scale_den
; /* Denominator */
488 unsigned int gt_quota_quantum
; /* Secs between syncs to quota file */
489 unsigned int gt_new_files_jdata
;
490 unsigned int gt_max_readahead
; /* Max bytes to read-ahead from disk */
491 unsigned int gt_complain_secs
;
492 unsigned int gt_statfs_quantum
;
493 unsigned int gt_statfs_slow
;
497 SDF_JOURNAL_CHECKED
= 0,
498 SDF_JOURNAL_LIVE
= 1,
504 SDF_RORECOVERY
= 7, /* read only recovery */
507 #define GFS2_FSNAME_LEN 256
509 struct gfs2_inum_host
{
514 struct gfs2_sb_host
{
520 u32 sb_multihost_format
;
524 struct gfs2_inum_host sb_master_dir
;
525 struct gfs2_inum_host sb_root_dir
;
527 char sb_lockproto
[GFS2_LOCKNAME_LEN
];
528 char sb_locktable
[GFS2_LOCKNAME_LEN
];
532 * lm_mount() return values
534 * ls_jid - the journal ID this node should use
535 * ls_first - this node is the first to mount the file system
536 * ls_lockspace - lock module's context for this file system
537 * ls_ops - lock module's functions
540 struct lm_lockstruct
{
542 unsigned int ls_first
;
543 const struct lm_lockops
*ls_ops
;
544 dlm_lockspace_t
*ls_dlm
;
546 int ls_recover_jid_done
; /* These two are deprecated, */
547 int ls_recover_jid_status
; /* used previously by gfs_controld */
549 struct dlm_lksb ls_mounted_lksb
; /* mounted_lock */
550 struct dlm_lksb ls_control_lksb
; /* control_lock */
551 char ls_control_lvb
[GDLM_LVB_SIZE
]; /* control_lock lvb */
552 struct completion ls_sync_wait
; /* {control,mounted}_{lock,unlock} */
554 spinlock_t ls_recover_spin
; /* protects following fields */
555 unsigned long ls_recover_flags
; /* DFL_ */
556 uint32_t ls_recover_mount
; /* gen in first recover_done cb */
557 uint32_t ls_recover_start
; /* gen in last recover_done cb */
558 uint32_t ls_recover_block
; /* copy recover_start in last recover_prep */
559 uint32_t ls_recover_size
; /* size of recover_submit, recover_result */
560 uint32_t *ls_recover_submit
; /* gen in last recover_slot cb per jid */
561 uint32_t *ls_recover_result
; /* result of last jid recovery */
564 struct gfs2_pcpu_lkstats
{
565 /* One struct for each glock type */
566 struct gfs2_lkstats lkstats
[10];
570 struct super_block
*sd_vfs
;
571 struct gfs2_pcpu_lkstats __percpu
*sd_lkstats
;
572 struct kobject sd_kobj
;
573 unsigned long sd_flags
; /* SDF_... */
574 struct gfs2_sb_host sd_sb
;
576 /* Constants computed on mount */
580 u32 sd_diptrs
; /* Number of pointers in a dinode */
581 u32 sd_inptrs
; /* Number of pointers in a indirect block */
582 u32 sd_jbsize
; /* Size of a journaled data block */
583 u32 sd_hash_bsize
; /* sizeof(exhash block) */
584 u32 sd_hash_bsize_shift
;
585 u32 sd_hash_ptrs
; /* Number of pointers in a hash block */
587 u32 sd_max_dirres
; /* Max blocks needed to add a directory entry */
588 u32 sd_max_height
; /* Max height of a file's metadata tree */
589 u64 sd_heightsize
[GFS2_MAX_META_HEIGHT
+ 1];
590 u32 sd_max_jheight
; /* Max height of journaled file's meta tree */
591 u64 sd_jheightsize
[GFS2_MAX_META_HEIGHT
+ 1];
593 struct gfs2_args sd_args
; /* Mount arguments */
594 struct gfs2_tune sd_tune
; /* Filesystem tuning structure */
598 struct lm_lockstruct sd_lockstruct
;
599 struct gfs2_holder sd_live_gh
;
600 struct gfs2_glock
*sd_rename_gl
;
601 struct gfs2_glock
*sd_trans_gl
;
602 wait_queue_head_t sd_glock_wait
;
603 atomic_t sd_glock_disposal
;
604 struct completion sd_locking_init
;
605 struct delayed_work sd_control_work
;
609 struct dentry
*sd_master_dir
;
610 struct dentry
*sd_root_dir
;
612 struct inode
*sd_jindex
;
613 struct inode
*sd_statfs_inode
;
614 struct inode
*sd_sc_inode
;
615 struct inode
*sd_qc_inode
;
616 struct inode
*sd_rindex
;
617 struct inode
*sd_quota_inode
;
621 spinlock_t sd_statfs_spin
;
622 struct gfs2_statfs_change_host sd_statfs_master
;
623 struct gfs2_statfs_change_host sd_statfs_local
;
624 int sd_statfs_force_sync
;
626 /* Resource group stuff */
628 int sd_rindex_uptodate
;
629 spinlock_t sd_rindex_spin
;
630 struct rb_root sd_rindex_tree
;
631 unsigned int sd_rgrps
;
632 unsigned int sd_max_rg_data
;
634 /* Journal index stuff */
636 struct list_head sd_jindex_list
;
637 spinlock_t sd_jindex_spin
;
638 struct mutex sd_jindex_mutex
;
639 unsigned int sd_journals
;
641 struct gfs2_jdesc
*sd_jdesc
;
642 struct gfs2_holder sd_journal_gh
;
643 struct gfs2_holder sd_jinode_gh
;
645 struct gfs2_holder sd_sc_gh
;
646 struct gfs2_holder sd_qc_gh
;
650 struct task_struct
*sd_logd_process
;
651 struct task_struct
*sd_quotad_process
;
655 struct list_head sd_quota_list
;
656 atomic_t sd_quota_count
;
657 struct mutex sd_quota_mutex
;
658 wait_queue_head_t sd_quota_wait
;
659 struct list_head sd_trunc_list
;
660 spinlock_t sd_trunc_lock
;
662 unsigned int sd_quota_slots
;
663 unsigned int sd_quota_chunks
;
664 unsigned char **sd_quota_bitmap
;
666 u64 sd_quota_sync_gen
;
670 spinlock_t sd_log_lock
;
672 unsigned int sd_log_blks_reserved
;
673 unsigned int sd_log_commited_buf
;
674 unsigned int sd_log_commited_databuf
;
675 int sd_log_commited_revoke
;
677 atomic_t sd_log_pinned
;
678 unsigned int sd_log_num_buf
;
679 unsigned int sd_log_num_revoke
;
680 unsigned int sd_log_num_rg
;
681 unsigned int sd_log_num_databuf
;
683 struct list_head sd_log_le_buf
;
684 struct list_head sd_log_le_revoke
;
685 struct list_head sd_log_le_databuf
;
686 struct list_head sd_log_le_ordered
;
688 atomic_t sd_log_thresh1
;
689 atomic_t sd_log_thresh2
;
690 atomic_t sd_log_blks_free
;
691 wait_queue_head_t sd_log_waitq
;
692 wait_queue_head_t sd_logd_waitq
;
695 unsigned int sd_log_head
;
696 unsigned int sd_log_tail
;
699 struct rw_semaphore sd_log_flush_lock
;
700 atomic_t sd_log_in_flight
;
701 struct bio
*sd_log_bio
;
702 wait_queue_head_t sd_log_flush_wait
;
705 unsigned int sd_log_flush_head
;
706 u64 sd_log_flush_wrapped
;
708 spinlock_t sd_ail_lock
;
709 struct list_head sd_ail1_list
;
710 struct list_head sd_ail2_list
;
714 struct list_head sd_revoke_list
;
715 unsigned int sd_replay_tail
;
717 unsigned int sd_found_blocks
;
718 unsigned int sd_found_revokes
;
719 unsigned int sd_replayed_blocks
;
721 /* For quiescing the filesystem */
723 struct gfs2_holder sd_freeze_gh
;
724 struct mutex sd_freeze_lock
;
725 unsigned int sd_freeze_count
;
727 char sd_fsname
[GFS2_FSNAME_LEN
];
728 char sd_table_name
[GFS2_FSNAME_LEN
];
729 char sd_proto_name
[GFS2_FSNAME_LEN
];
733 unsigned long sd_last_warning
;
734 struct dentry
*debugfs_dir
; /* debugfs directory */
735 struct dentry
*debugfs_dentry_glocks
;
736 struct dentry
*debugfs_dentry_glstats
;
737 struct dentry
*debugfs_dentry_sbstats
;
740 static inline void gfs2_glstats_inc(struct gfs2_glock
*gl
, int which
)
742 gl
->gl_stats
.stats
[which
]++;
745 static inline void gfs2_sbstats_inc(const struct gfs2_glock
*gl
, int which
)
747 const struct gfs2_sbd
*sdp
= gl
->gl_sbd
;
749 this_cpu_ptr(sdp
->sd_lkstats
)->lkstats
[gl
->gl_name
.ln_type
].stats
[which
]++;
753 #endif /* __INCORE_DOT_H__ */