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 */
87 u32 rd_reserved; /* number of blocks reserved */
91 struct gfs2_bitmap *rd_bits;
92 struct gfs2_sbd *rd_sbd;
93 struct gfs2_rgrp_lvb *rd_rgl;
96 #define GFS2_RDF_CHECK 0x10000000 /* check for unlinked inodes */
97 #define GFS2_RDF_UPTODATE 0x20000000 /* rg is up to date */
98 #define GFS2_RDF_ERROR 0x40000000 /* error in rg */
99 #define GFS2_RDF_MASK 0xf0000000 /* mask for internal flags */
100 spinlock_t rd_rsspin; /* protects reservation related vars */
101 struct rb_root rd_rstree; /* multi-block reservation tree */
102 u32 rd_rs_cnt; /* count of current reservations */
105 enum gfs2_state_bits {
106 BH_Pinned = BH_PrivateStart,
107 BH_Escaped = BH_PrivateStart + 1,
108 BH_Zeronew = BH_PrivateStart + 2,
111 BUFFER_FNS(Pinned, pinned)
112 TAS_BUFFER_FNS(Pinned, pinned)
113 BUFFER_FNS(Escaped, escaped)
114 TAS_BUFFER_FNS(Escaped, escaped)
115 BUFFER_FNS(Zeronew, zeronew)
116 TAS_BUFFER_FNS(Zeronew, zeronew)
118 struct gfs2_bufdata {
119 struct buffer_head *bd_bh;
120 struct gfs2_glock *bd_gl;
123 struct list_head bd_list;
124 const struct gfs2_log_operations *bd_ops;
126 struct gfs2_ail *bd_ail;
127 struct list_head bd_ail_st_list;
128 struct list_head bd_ail_gl_list;
132 * Internally, we prefix things with gdlm_ and GDLM_ (for gfs-dlm) since a
133 * prefix of lock_dlm_ gets awkward.
136 #define GDLM_STRNAME_BYTES 25
137 #define GDLM_LVB_SIZE 32
142 * DFL_BLOCK_LOCKS: dlm is in recovery and will grant locks that had been
143 * held by failed nodes whose journals need recovery. Those locks should
144 * only be used for journal recovery until the journal recovery is done.
145 * This is set by the dlm recover_prep callback and cleared by the
146 * gfs2_control thread when journal recovery is complete. To avoid
147 * races between recover_prep setting and gfs2_control clearing, recover_spin
148 * is held while changing this bit and reading/writing recover_block
151 * DFL_NO_DLM_OPS: dlm lockspace ops/callbacks are not being used.
153 * DFL_FIRST_MOUNT: this node is the first to mount this fs and is doing
154 * recovery of all journals before allowing other nodes to mount the fs.
155 * This is cleared when FIRST_MOUNT_DONE is set.
157 * DFL_FIRST_MOUNT_DONE: this node was the first mounter, and has finished
158 * recovery of all journals, and now allows other nodes to mount the fs.
160 * DFL_MOUNT_DONE: gdlm_mount has completed successfully and cleared
161 * BLOCK_LOCKS for the first time. The gfs2_control thread should now
162 * control clearing BLOCK_LOCKS for further recoveries.
164 * DFL_UNMOUNT: gdlm_unmount sets to keep sdp off gfs2_control_wq.
166 * DFL_DLM_RECOVERY: set while dlm is in recovery, between recover_prep()
167 * and recover_done(), i.e. set while recover_block == recover_start.
174 DFL_FIRST_MOUNT_DONE = 3,
177 DFL_DLM_RECOVERY = 6,
182 unsigned int ln_type;
185 #define lm_name_equal(name1, name2) \
186 (((name1)->ln_number == (name2)->ln_number) && \
187 ((name1)->ln_type == (name2)->ln_type))
190 struct gfs2_glock_operations {
191 void (*go_xmote_th) (struct gfs2_glock *gl);
192 int (*go_xmote_bh) (struct gfs2_glock *gl, struct gfs2_holder *gh);
193 void (*go_inval) (struct gfs2_glock *gl, int flags);
194 int (*go_demote_ok) (const struct gfs2_glock *gl);
195 int (*go_lock) (struct gfs2_holder *gh);
196 void (*go_unlock) (struct gfs2_holder *gh);
197 int (*go_dump)(struct seq_file *seq, const struct gfs2_glock *gl);
198 void (*go_callback) (struct gfs2_glock *gl);
200 const unsigned long go_flags;
201 #define GLOF_ASPACE 1
205 GFS2_LKS_SRTT = 0, /* Non blocking smoothed round trip time */
206 GFS2_LKS_SRTTVAR = 1, /* Non blocking smoothed variance */
207 GFS2_LKS_SRTTB = 2, /* Blocking smoothed round trip time */
208 GFS2_LKS_SRTTVARB = 3, /* Blocking smoothed variance */
209 GFS2_LKS_SIRT = 4, /* Smoothed Inter-request time */
210 GFS2_LKS_SIRTVAR = 5, /* Smoothed Inter-request variance */
211 GFS2_LKS_DCOUNT = 6, /* Count of dlm requests */
212 GFS2_LKS_QCOUNT = 7, /* Count of gfs2_holder queues */
216 struct gfs2_lkstats {
217 s64 stats[GFS2_NR_LKSTATS];
222 HIF_HOLDER = 6, /* Set for gh that "holds" the glock */
228 struct list_head gh_list;
230 struct gfs2_glock *gh_gl;
231 struct pid *gh_owner_pid;
232 unsigned int gh_state;
236 unsigned long gh_iflags; /* HIF_... */
240 /* Resource group multi-block reservation, in order of appearance:
242 Step 1. Function prepares to write, allocates a mb, sets the size hint.
243 Step 2. User calls inplace_reserve to target an rgrp, sets the rgrp info
244 Step 3. Function get_local_rgrp locks the rgrp, determines which bits to use
245 Step 4. Bits are assigned from the rgrp based on either the reservation
249 struct gfs2_blkreserv {
250 /* components used during write (step 1): */
251 atomic_t rs_sizehint; /* hint of the write size */
253 /* components used during inplace_reserve (step 2): */
254 u32 rs_requested; /* Filled in by caller of gfs2_inplace_reserve() */
256 /* components used during get_local_rgrp (step 3): */
257 struct gfs2_rgrpd *rs_rgd; /* pointer to the gfs2_rgrpd */
258 struct gfs2_holder rs_rgd_gh; /* Filled in by get_local_rgrp */
259 struct rb_node rs_node; /* link to other block reservations */
261 /* components used during block searches and assignments (step 4): */
262 struct gfs2_bitmap *rs_bi; /* bitmap for the current allocation */
263 u32 rs_biblk; /* start block relative to the bi */
264 u32 rs_free; /* how many blocks are still free */
266 /* ancillary quota stuff */
267 struct gfs2_quota_data *rs_qa_qd[2 * MAXQUOTAS];
268 struct gfs2_holder rs_qa_qd_ghs[2 * MAXQUOTAS];
269 unsigned int rs_qa_qd_num;
275 GLF_PENDING_DEMOTE = 4,
276 GLF_DEMOTE_IN_PROGRESS = 5,
279 GLF_INVALIDATE_IN_PROGRESS = 8,
280 GLF_REPLY_PENDING = 9,
285 GLF_OBJECT = 14, /* Used only for tracing */
290 struct hlist_bl_node gl_list;
291 struct gfs2_sbd *gl_sbd;
292 unsigned long gl_flags; /* GLF_... */
293 struct lm_lockname gl_name;
298 /* State fields protected by gl_spin */
299 unsigned int gl_state:2, /* Current state */
300 gl_target:2, /* Target state */
301 gl_demote_state:2, /* State requested by remote node */
302 gl_req:2, /* State in last dlm request */
303 gl_reply:8; /* Last reply from the dlm */
305 unsigned int gl_hash;
306 unsigned long gl_demote_time; /* time of first demote request */
308 struct list_head gl_holders;
310 const struct gfs2_glock_operations *gl_ops;
312 struct gfs2_lkstats gl_stats;
313 struct dlm_lksb gl_lksb;
315 unsigned long gl_tchange;
318 struct list_head gl_lru;
319 struct list_head gl_ail_list;
320 atomic_t gl_ail_count;
322 struct delayed_work gl_work;
323 struct work_struct gl_delete;
324 struct rcu_head gl_rcu;
327 #define GFS2_MIN_LVB_SIZE 32 /* Min size of LVB that gfs2 supports */
332 GIF_ALLOC_FAILED = 2,
337 struct inode i_inode;
342 unsigned long i_flags; /* GIF_... */
343 struct gfs2_glock *i_gl; /* Move into i_gh? */
344 struct gfs2_holder i_iopen_gh;
345 struct gfs2_holder i_gh; /* for prepare/commit_write only */
346 struct gfs2_blkreserv *i_res; /* rgrp multi-block reservation */
347 struct gfs2_rgrpd *i_rgd;
348 u64 i_goal; /* goal block for allocations */
349 struct rw_semaphore i_rw_mutex;
350 struct list_head i_trunc_list;
351 __be64 *i_hash_cache;
359 * Since i_inode is the first element of struct gfs2_inode,
360 * this is effectively a cast.
362 static inline struct gfs2_inode *GFS2_I(struct inode *inode)
364 return container_of(inode, struct gfs2_inode, i_inode);
367 static inline struct gfs2_sbd *GFS2_SB(const struct inode *inode)
369 return inode->i_sb->s_fs_info;
373 struct mutex f_fl_mutex;
374 struct gfs2_holder f_fl_gh;
377 struct gfs2_revoke_replay {
378 struct list_head rr_list;
380 unsigned int rr_where;
390 struct gfs2_quota_data {
391 struct list_head qd_list;
392 struct list_head qd_reclaim;
397 unsigned long qd_flags; /* QDF_... */
402 unsigned int qd_slot;
403 unsigned int qd_slot_count;
405 struct buffer_head *qd_bh;
406 struct gfs2_quota_change *qd_bh_qc;
407 unsigned int qd_bh_count;
409 struct gfs2_glock *qd_gl;
410 struct gfs2_quota_lvb qd_qb;
413 unsigned long qd_last_warn;
419 unsigned int tr_blocks;
420 unsigned int tr_revokes;
421 unsigned int tr_reserved;
423 struct gfs2_holder tr_t_gh;
427 unsigned int tr_num_buf_new;
428 unsigned int tr_num_databuf_new;
429 unsigned int tr_num_buf_rm;
430 unsigned int tr_num_databuf_rm;
431 unsigned int tr_num_revoke;
432 unsigned int tr_num_revoke_rm;
436 struct list_head ai_list;
438 unsigned int ai_first;
439 struct list_head ai_ail1_list;
440 struct list_head ai_ail2_list;
443 struct gfs2_journal_extent {
444 struct list_head extent_list;
446 unsigned int lblock; /* First logical block */
447 u64 dblock; /* First disk block */
452 struct list_head jd_list;
453 struct list_head extent_list;
454 struct work_struct jd_work;
455 struct inode *jd_inode;
456 unsigned long jd_flags;
457 #define JDF_RECOVERY 1
459 unsigned int jd_blocks;
460 int jd_recover_error;
463 struct gfs2_statfs_change_host {
469 #define GFS2_QUOTA_DEFAULT GFS2_QUOTA_OFF
470 #define GFS2_QUOTA_OFF 0
471 #define GFS2_QUOTA_ACCOUNT 1
472 #define GFS2_QUOTA_ON 2
474 #define GFS2_DATA_DEFAULT GFS2_DATA_ORDERED
475 #define GFS2_DATA_WRITEBACK 1
476 #define GFS2_DATA_ORDERED 2
478 #define GFS2_ERRORS_DEFAULT GFS2_ERRORS_WITHDRAW
479 #define GFS2_ERRORS_WITHDRAW 0
480 #define GFS2_ERRORS_CONTINUE 1 /* place holder for future feature */
481 #define GFS2_ERRORS_RO 2 /* place holder for future feature */
482 #define GFS2_ERRORS_PANIC 3
485 char ar_lockproto[GFS2_LOCKNAME_LEN]; /* Name of the Lock Protocol */
486 char ar_locktable[GFS2_LOCKNAME_LEN]; /* Name of the Lock Table */
487 char ar_hostdata[GFS2_LOCKNAME_LEN]; /* Host specific data */
488 unsigned int ar_spectator:1; /* Don't get a journal */
489 unsigned int ar_localflocks:1; /* Let the VFS do flock|fcntl */
490 unsigned int ar_debug:1; /* Oops on errors */
491 unsigned int ar_posix_acl:1; /* Enable posix acls */
492 unsigned int ar_quota:2; /* off/account/on */
493 unsigned int ar_suiddir:1; /* suiddir support */
494 unsigned int ar_data:2; /* ordered/writeback */
495 unsigned int ar_meta:1; /* mount metafs */
496 unsigned int ar_discard:1; /* discard requests */
497 unsigned int ar_errors:2; /* errors=withdraw | panic */
498 unsigned int ar_nobarrier:1; /* do not send barriers */
499 unsigned int ar_rgrplvb:1; /* use lvbs for rgrp info */
500 int ar_commit; /* Commit interval */
501 int ar_statfs_quantum; /* The fast statfs interval */
502 int ar_quota_quantum; /* The quota interval */
503 int ar_statfs_percent; /* The % change to force sync */
509 unsigned int gt_logd_secs;
511 unsigned int gt_quota_simul_sync; /* Max quotavals to sync at once */
512 unsigned int gt_quota_warn_period; /* Secs between quota warn msgs */
513 unsigned int gt_quota_scale_num; /* Numerator */
514 unsigned int gt_quota_scale_den; /* Denominator */
515 unsigned int gt_quota_quantum; /* Secs between syncs to quota file */
516 unsigned int gt_new_files_jdata;
517 unsigned int gt_max_readahead; /* Max bytes to read-ahead from disk */
518 unsigned int gt_complain_secs;
519 unsigned int gt_statfs_quantum;
520 unsigned int gt_statfs_slow;
524 SDF_JOURNAL_CHECKED = 0,
525 SDF_JOURNAL_LIVE = 1,
531 SDF_RORECOVERY = 7, /* read only recovery */
534 #define GFS2_FSNAME_LEN 256
536 struct gfs2_inum_host {
541 struct gfs2_sb_host {
547 u32 sb_multihost_format;
551 struct gfs2_inum_host sb_master_dir;
552 struct gfs2_inum_host sb_root_dir;
554 char sb_lockproto[GFS2_LOCKNAME_LEN];
555 char sb_locktable[GFS2_LOCKNAME_LEN];
559 * lm_mount() return values
561 * ls_jid - the journal ID this node should use
562 * ls_first - this node is the first to mount the file system
563 * ls_lockspace - lock module's context for this file system
564 * ls_ops - lock module's functions
567 struct lm_lockstruct {
569 unsigned int ls_first;
570 const struct lm_lockops *ls_ops;
571 dlm_lockspace_t *ls_dlm;
573 int ls_recover_jid_done; /* These two are deprecated, */
574 int ls_recover_jid_status; /* used previously by gfs_controld */
576 struct dlm_lksb ls_mounted_lksb; /* mounted_lock */
577 struct dlm_lksb ls_control_lksb; /* control_lock */
578 char ls_control_lvb[GDLM_LVB_SIZE]; /* control_lock lvb */
579 struct completion ls_sync_wait; /* {control,mounted}_{lock,unlock} */
581 spinlock_t ls_recover_spin; /* protects following fields */
582 unsigned long ls_recover_flags; /* DFL_ */
583 uint32_t ls_recover_mount; /* gen in first recover_done cb */
584 uint32_t ls_recover_start; /* gen in last recover_done cb */
585 uint32_t ls_recover_block; /* copy recover_start in last recover_prep */
586 uint32_t ls_recover_size; /* size of recover_submit, recover_result */
587 uint32_t *ls_recover_submit; /* gen in last recover_slot cb per jid */
588 uint32_t *ls_recover_result; /* result of last jid recovery */
591 struct gfs2_pcpu_lkstats {
592 /* One struct for each glock type */
593 struct gfs2_lkstats lkstats[10];
597 struct super_block *sd_vfs;
598 struct gfs2_pcpu_lkstats __percpu *sd_lkstats;
599 struct kobject sd_kobj;
600 unsigned long sd_flags; /* SDF_... */
601 struct gfs2_sb_host sd_sb;
603 /* Constants computed on mount */
607 u32 sd_diptrs; /* Number of pointers in a dinode */
608 u32 sd_inptrs; /* Number of pointers in a indirect block */
609 u32 sd_jbsize; /* Size of a journaled data block */
610 u32 sd_hash_bsize; /* sizeof(exhash block) */
611 u32 sd_hash_bsize_shift;
612 u32 sd_hash_ptrs; /* Number of pointers in a hash block */
614 u32 sd_max_dirres; /* Max blocks needed to add a directory entry */
615 u32 sd_max_height; /* Max height of a file's metadata tree */
616 u64 sd_heightsize[GFS2_MAX_META_HEIGHT + 1];
617 u32 sd_max_jheight; /* Max height of journaled file's meta tree */
618 u64 sd_jheightsize[GFS2_MAX_META_HEIGHT + 1];
620 struct gfs2_args sd_args; /* Mount arguments */
621 struct gfs2_tune sd_tune; /* Filesystem tuning structure */
625 struct lm_lockstruct sd_lockstruct;
626 struct gfs2_holder sd_live_gh;
627 struct gfs2_glock *sd_rename_gl;
628 struct gfs2_glock *sd_trans_gl;
629 wait_queue_head_t sd_glock_wait;
630 atomic_t sd_glock_disposal;
631 struct completion sd_locking_init;
632 struct delayed_work sd_control_work;
636 struct dentry *sd_master_dir;
637 struct dentry *sd_root_dir;
639 struct inode *sd_jindex;
640 struct inode *sd_statfs_inode;
641 struct inode *sd_sc_inode;
642 struct inode *sd_qc_inode;
643 struct inode *sd_rindex;
644 struct inode *sd_quota_inode;
648 spinlock_t sd_statfs_spin;
649 struct gfs2_statfs_change_host sd_statfs_master;
650 struct gfs2_statfs_change_host sd_statfs_local;
651 int sd_statfs_force_sync;
653 /* Resource group stuff */
655 int sd_rindex_uptodate;
656 spinlock_t sd_rindex_spin;
657 struct rb_root sd_rindex_tree;
658 unsigned int sd_rgrps;
659 unsigned int sd_max_rg_data;
661 /* Journal index stuff */
663 struct list_head sd_jindex_list;
664 spinlock_t sd_jindex_spin;
665 struct mutex sd_jindex_mutex;
666 unsigned int sd_journals;
668 struct gfs2_jdesc *sd_jdesc;
669 struct gfs2_holder sd_journal_gh;
670 struct gfs2_holder sd_jinode_gh;
672 struct gfs2_holder sd_sc_gh;
673 struct gfs2_holder sd_qc_gh;
677 struct task_struct *sd_logd_process;
678 struct task_struct *sd_quotad_process;
682 struct list_head sd_quota_list;
683 atomic_t sd_quota_count;
684 struct mutex sd_quota_mutex;
685 wait_queue_head_t sd_quota_wait;
686 struct list_head sd_trunc_list;
687 spinlock_t sd_trunc_lock;
689 unsigned int sd_quota_slots;
690 unsigned int sd_quota_chunks;
691 unsigned char **sd_quota_bitmap;
693 u64 sd_quota_sync_gen;
697 spinlock_t sd_log_lock;
699 unsigned int sd_log_blks_reserved;
700 unsigned int sd_log_commited_buf;
701 unsigned int sd_log_commited_databuf;
702 int sd_log_commited_revoke;
704 atomic_t sd_log_pinned;
705 unsigned int sd_log_num_buf;
706 unsigned int sd_log_num_revoke;
707 unsigned int sd_log_num_rg;
708 unsigned int sd_log_num_databuf;
710 struct list_head sd_log_le_buf;
711 struct list_head sd_log_le_revoke;
712 struct list_head sd_log_le_databuf;
713 struct list_head sd_log_le_ordered;
715 atomic_t sd_log_thresh1;
716 atomic_t sd_log_thresh2;
717 atomic_t sd_log_blks_free;
718 wait_queue_head_t sd_log_waitq;
719 wait_queue_head_t sd_logd_waitq;
722 unsigned int sd_log_head;
723 unsigned int sd_log_tail;
726 struct rw_semaphore sd_log_flush_lock;
727 atomic_t sd_log_in_flight;
728 struct bio *sd_log_bio;
729 wait_queue_head_t sd_log_flush_wait;
732 unsigned int sd_log_flush_head;
733 u64 sd_log_flush_wrapped;
735 spinlock_t sd_ail_lock;
736 struct list_head sd_ail1_list;
737 struct list_head sd_ail2_list;
741 struct list_head sd_revoke_list;
742 unsigned int sd_replay_tail;
744 unsigned int sd_found_blocks;
745 unsigned int sd_found_revokes;
746 unsigned int sd_replayed_blocks;
748 /* For quiescing the filesystem */
750 struct gfs2_holder sd_freeze_gh;
751 struct mutex sd_freeze_lock;
752 unsigned int sd_freeze_count;
754 char sd_fsname[GFS2_FSNAME_LEN];
755 char sd_table_name[GFS2_FSNAME_LEN];
756 char sd_proto_name[GFS2_FSNAME_LEN];
760 unsigned long sd_last_warning;
761 struct dentry *debugfs_dir; /* debugfs directory */
762 struct dentry *debugfs_dentry_glocks;
763 struct dentry *debugfs_dentry_glstats;
764 struct dentry *debugfs_dentry_sbstats;
767 static inline void gfs2_glstats_inc(struct gfs2_glock *gl, int which)
769 gl->gl_stats.stats[which]++;
772 static inline void gfs2_sbstats_inc(const struct gfs2_glock *gl, int which)
774 const struct gfs2_sbd *sdp = gl->gl_sbd;
776 this_cpu_ptr(sdp->sd_lkstats)->lkstats[gl->gl_name.ln_type].stats[which]++;
780 #endif /* __INCORE_DOT_H__ */