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_before_commit) (struct gfs2_sbd *sdp);
56 void (*lo_after_commit) (struct gfs2_sbd *sdp, struct gfs2_ail *ai);
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,
62 void (*lo_after_scan) (struct gfs2_jdesc *jd, int error, int pass);
69 struct buffer_head *bi_bh;
71 unsigned long bi_flags;
78 struct rb_node rd_node; /* Link with superblock */
79 struct gfs2_glock *rd_gl; /* Glock for this rgrp */
80 u64 rd_addr; /* grp block disk address */
81 u64 rd_data0; /* first data location */
82 u32 rd_length; /* length of rgrp header in fs blocks */
83 u32 rd_data; /* num of data blocks in rgrp */
84 u32 rd_bitbytes; /* number of bytes in data bitmaps */
86 u32 rd_reserved; /* number of blocks reserved */
90 struct gfs2_bitmap *rd_bits;
91 struct gfs2_sbd *rd_sbd;
92 struct gfs2_rgrp_lvb *rd_rgl;
95 #define GFS2_RDF_CHECK 0x10000000 /* check for unlinked inodes */
96 #define GFS2_RDF_UPTODATE 0x20000000 /* rg is up to date */
97 #define GFS2_RDF_ERROR 0x40000000 /* error in rg */
98 #define GFS2_RDF_MASK 0xf0000000 /* mask for internal flags */
99 spinlock_t rd_rsspin; /* protects reservation related vars */
100 struct rb_root rd_rstree; /* multi-block reservation tree */
104 struct gfs2_rgrpd *rgd;
105 struct gfs2_bitmap *bi; /* Bitmap must belong to the rgd */
106 u32 offset; /* The offset is bitmap relative */
109 static inline u64 gfs2_rbm_to_block(const struct gfs2_rbm *rbm)
111 return rbm->rgd->rd_data0 + (rbm->bi->bi_start * GFS2_NBBY) + rbm->offset;
114 static inline bool gfs2_rbm_eq(const struct gfs2_rbm *rbm1,
115 const struct gfs2_rbm *rbm2)
117 return (rbm1->rgd == rbm2->rgd) && (rbm1->bi == rbm2->bi) &&
118 (rbm1->offset == rbm2->offset);
121 enum gfs2_state_bits {
122 BH_Pinned = BH_PrivateStart,
123 BH_Escaped = BH_PrivateStart + 1,
124 BH_Zeronew = BH_PrivateStart + 2,
127 BUFFER_FNS(Pinned, pinned)
128 TAS_BUFFER_FNS(Pinned, pinned)
129 BUFFER_FNS(Escaped, escaped)
130 TAS_BUFFER_FNS(Escaped, escaped)
131 BUFFER_FNS(Zeronew, zeronew)
132 TAS_BUFFER_FNS(Zeronew, zeronew)
134 struct gfs2_bufdata {
135 struct buffer_head *bd_bh;
136 struct gfs2_glock *bd_gl;
139 struct list_head bd_list;
140 const struct gfs2_log_operations *bd_ops;
142 struct gfs2_ail *bd_ail;
143 struct list_head bd_ail_st_list;
144 struct list_head bd_ail_gl_list;
148 * Internally, we prefix things with gdlm_ and GDLM_ (for gfs-dlm) since a
149 * prefix of lock_dlm_ gets awkward.
152 #define GDLM_STRNAME_BYTES 25
153 #define GDLM_LVB_SIZE 32
158 * DFL_BLOCK_LOCKS: dlm is in recovery and will grant locks that had been
159 * held by failed nodes whose journals need recovery. Those locks should
160 * only be used for journal recovery until the journal recovery is done.
161 * This is set by the dlm recover_prep callback and cleared by the
162 * gfs2_control thread when journal recovery is complete. To avoid
163 * races between recover_prep setting and gfs2_control clearing, recover_spin
164 * is held while changing this bit and reading/writing recover_block
167 * DFL_NO_DLM_OPS: dlm lockspace ops/callbacks are not being used.
169 * DFL_FIRST_MOUNT: this node is the first to mount this fs and is doing
170 * recovery of all journals before allowing other nodes to mount the fs.
171 * This is cleared when FIRST_MOUNT_DONE is set.
173 * DFL_FIRST_MOUNT_DONE: this node was the first mounter, and has finished
174 * recovery of all journals, and now allows other nodes to mount the fs.
176 * DFL_MOUNT_DONE: gdlm_mount has completed successfully and cleared
177 * BLOCK_LOCKS for the first time. The gfs2_control thread should now
178 * control clearing BLOCK_LOCKS for further recoveries.
180 * DFL_UNMOUNT: gdlm_unmount sets to keep sdp off gfs2_control_wq.
182 * DFL_DLM_RECOVERY: set while dlm is in recovery, between recover_prep()
183 * and recover_done(), i.e. set while recover_block == recover_start.
190 DFL_FIRST_MOUNT_DONE = 3,
193 DFL_DLM_RECOVERY = 6,
198 unsigned int ln_type;
201 #define lm_name_equal(name1, name2) \
202 (((name1)->ln_number == (name2)->ln_number) && \
203 ((name1)->ln_type == (name2)->ln_type))
206 struct gfs2_glock_operations {
207 void (*go_sync) (struct gfs2_glock *gl);
208 int (*go_xmote_bh) (struct gfs2_glock *gl, struct gfs2_holder *gh);
209 void (*go_inval) (struct gfs2_glock *gl, int flags);
210 int (*go_demote_ok) (const struct gfs2_glock *gl);
211 int (*go_lock) (struct gfs2_holder *gh);
212 void (*go_unlock) (struct gfs2_holder *gh);
213 int (*go_dump)(struct seq_file *seq, const struct gfs2_glock *gl);
214 void (*go_callback) (struct gfs2_glock *gl);
216 const unsigned long go_flags;
217 #define GLOF_ASPACE 1
222 GFS2_LKS_SRTT = 0, /* Non blocking smoothed round trip time */
223 GFS2_LKS_SRTTVAR = 1, /* Non blocking smoothed variance */
224 GFS2_LKS_SRTTB = 2, /* Blocking smoothed round trip time */
225 GFS2_LKS_SRTTVARB = 3, /* Blocking smoothed variance */
226 GFS2_LKS_SIRT = 4, /* Smoothed Inter-request time */
227 GFS2_LKS_SIRTVAR = 5, /* Smoothed Inter-request variance */
228 GFS2_LKS_DCOUNT = 6, /* Count of dlm requests */
229 GFS2_LKS_QCOUNT = 7, /* Count of gfs2_holder queues */
233 struct gfs2_lkstats {
234 s64 stats[GFS2_NR_LKSTATS];
239 HIF_HOLDER = 6, /* Set for gh that "holds" the glock */
245 struct list_head gh_list;
247 struct gfs2_glock *gh_gl;
248 struct pid *gh_owner_pid;
249 unsigned int gh_state;
253 unsigned long gh_iflags; /* HIF_... */
257 /* Resource group multi-block reservation, in order of appearance:
259 Step 1. Function prepares to write, allocates a mb, sets the size hint.
260 Step 2. User calls inplace_reserve to target an rgrp, sets the rgrp info
261 Step 3. Function get_local_rgrp locks the rgrp, determines which bits to use
262 Step 4. Bits are assigned from the rgrp based on either the reservation
266 struct gfs2_blkreserv {
267 /* components used during write (step 1): */
268 atomic_t rs_sizehint; /* hint of the write size */
270 struct gfs2_holder rs_rgd_gh; /* Filled in by get_local_rgrp */
271 struct rb_node rs_node; /* link to other block reservations */
272 struct gfs2_rbm rs_rbm; /* Start of reservation */
273 u32 rs_free; /* how many blocks are still free */
274 u64 rs_inum; /* Inode number for reservation */
276 /* ancillary quota stuff */
277 struct gfs2_quota_data *rs_qa_qd[2 * MAXQUOTAS];
278 struct gfs2_holder rs_qa_qd_ghs[2 * MAXQUOTAS];
279 unsigned int rs_qa_qd_num;
285 GLF_PENDING_DEMOTE = 4,
286 GLF_DEMOTE_IN_PROGRESS = 5,
289 GLF_INVALIDATE_IN_PROGRESS = 8,
290 GLF_REPLY_PENDING = 9,
295 GLF_OBJECT = 14, /* Used only for tracing */
300 struct hlist_bl_node gl_list;
301 struct gfs2_sbd *gl_sbd;
302 unsigned long gl_flags; /* GLF_... */
303 struct lm_lockname gl_name;
308 /* State fields protected by gl_spin */
309 unsigned int gl_state:2, /* Current state */
310 gl_target:2, /* Target state */
311 gl_demote_state:2, /* State requested by remote node */
312 gl_req:2, /* State in last dlm request */
313 gl_reply:8; /* Last reply from the dlm */
315 unsigned int gl_hash;
316 unsigned long gl_demote_time; /* time of first demote request */
318 struct list_head gl_holders;
320 const struct gfs2_glock_operations *gl_ops;
322 struct gfs2_lkstats gl_stats;
323 struct dlm_lksb gl_lksb;
324 unsigned long gl_tchange;
327 struct list_head gl_lru;
328 struct list_head gl_ail_list;
329 atomic_t gl_ail_count;
331 struct delayed_work gl_work;
332 struct work_struct gl_delete;
333 struct rcu_head gl_rcu;
336 #define GFS2_MIN_LVB_SIZE 32 /* Min size of LVB that gfs2 supports */
341 GIF_ALLOC_FAILED = 2,
346 struct inode i_inode;
351 unsigned long i_flags; /* GIF_... */
352 struct gfs2_glock *i_gl; /* Move into i_gh? */
353 struct gfs2_holder i_iopen_gh;
354 struct gfs2_holder i_gh; /* for prepare/commit_write only */
355 struct gfs2_blkreserv *i_res; /* rgrp multi-block reservation */
356 struct gfs2_rgrpd *i_rgd;
357 u64 i_goal; /* goal block for allocations */
358 struct rw_semaphore i_rw_mutex;
359 struct list_head i_trunc_list;
360 __be64 *i_hash_cache;
368 * Since i_inode is the first element of struct gfs2_inode,
369 * this is effectively a cast.
371 static inline struct gfs2_inode *GFS2_I(struct inode *inode)
373 return container_of(inode, struct gfs2_inode, i_inode);
376 static inline struct gfs2_sbd *GFS2_SB(const struct inode *inode)
378 return inode->i_sb->s_fs_info;
382 struct mutex f_fl_mutex;
383 struct gfs2_holder f_fl_gh;
386 struct gfs2_revoke_replay {
387 struct list_head rr_list;
389 unsigned int rr_where;
399 struct gfs2_quota_data {
400 struct list_head qd_list;
401 struct list_head qd_reclaim;
406 unsigned long qd_flags; /* QDF_... */
411 unsigned int qd_slot;
412 unsigned int qd_slot_count;
414 struct buffer_head *qd_bh;
415 struct gfs2_quota_change *qd_bh_qc;
416 unsigned int qd_bh_count;
418 struct gfs2_glock *qd_gl;
419 struct gfs2_quota_lvb qd_qb;
422 unsigned long qd_last_warn;
428 unsigned int tr_blocks;
429 unsigned int tr_revokes;
430 unsigned int tr_reserved;
432 struct gfs2_holder tr_t_gh;
436 unsigned int tr_num_buf_new;
437 unsigned int tr_num_databuf_new;
438 unsigned int tr_num_buf_rm;
439 unsigned int tr_num_databuf_rm;
440 unsigned int tr_num_revoke;
441 unsigned int tr_num_revoke_rm;
445 struct list_head ai_list;
447 unsigned int ai_first;
448 struct list_head ai_ail1_list;
449 struct list_head ai_ail2_list;
452 struct gfs2_journal_extent {
453 struct list_head extent_list;
455 unsigned int lblock; /* First logical block */
456 u64 dblock; /* First disk block */
461 struct list_head jd_list;
462 struct list_head extent_list;
463 struct work_struct jd_work;
464 struct inode *jd_inode;
465 unsigned long jd_flags;
466 #define JDF_RECOVERY 1
468 unsigned int jd_blocks;
469 int jd_recover_error;
472 struct gfs2_statfs_change_host {
478 #define GFS2_QUOTA_DEFAULT GFS2_QUOTA_OFF
479 #define GFS2_QUOTA_OFF 0
480 #define GFS2_QUOTA_ACCOUNT 1
481 #define GFS2_QUOTA_ON 2
483 #define GFS2_DATA_DEFAULT GFS2_DATA_ORDERED
484 #define GFS2_DATA_WRITEBACK 1
485 #define GFS2_DATA_ORDERED 2
487 #define GFS2_ERRORS_DEFAULT GFS2_ERRORS_WITHDRAW
488 #define GFS2_ERRORS_WITHDRAW 0
489 #define GFS2_ERRORS_CONTINUE 1 /* place holder for future feature */
490 #define GFS2_ERRORS_RO 2 /* place holder for future feature */
491 #define GFS2_ERRORS_PANIC 3
494 char ar_lockproto[GFS2_LOCKNAME_LEN]; /* Name of the Lock Protocol */
495 char ar_locktable[GFS2_LOCKNAME_LEN]; /* Name of the Lock Table */
496 char ar_hostdata[GFS2_LOCKNAME_LEN]; /* Host specific data */
497 unsigned int ar_spectator:1; /* Don't get a journal */
498 unsigned int ar_localflocks:1; /* Let the VFS do flock|fcntl */
499 unsigned int ar_debug:1; /* Oops on errors */
500 unsigned int ar_posix_acl:1; /* Enable posix acls */
501 unsigned int ar_quota:2; /* off/account/on */
502 unsigned int ar_suiddir:1; /* suiddir support */
503 unsigned int ar_data:2; /* ordered/writeback */
504 unsigned int ar_meta:1; /* mount metafs */
505 unsigned int ar_discard:1; /* discard requests */
506 unsigned int ar_errors:2; /* errors=withdraw | panic */
507 unsigned int ar_nobarrier:1; /* do not send barriers */
508 unsigned int ar_rgrplvb:1; /* use lvbs for rgrp info */
509 int ar_commit; /* Commit interval */
510 int ar_statfs_quantum; /* The fast statfs interval */
511 int ar_quota_quantum; /* The quota interval */
512 int ar_statfs_percent; /* The % change to force sync */
518 unsigned int gt_logd_secs;
520 unsigned int gt_quota_simul_sync; /* Max quotavals to sync at once */
521 unsigned int gt_quota_warn_period; /* Secs between quota warn msgs */
522 unsigned int gt_quota_scale_num; /* Numerator */
523 unsigned int gt_quota_scale_den; /* Denominator */
524 unsigned int gt_quota_quantum; /* Secs between syncs to quota file */
525 unsigned int gt_new_files_jdata;
526 unsigned int gt_max_readahead; /* Max bytes to read-ahead from disk */
527 unsigned int gt_complain_secs;
528 unsigned int gt_statfs_quantum;
529 unsigned int gt_statfs_slow;
533 SDF_JOURNAL_CHECKED = 0,
534 SDF_JOURNAL_LIVE = 1,
540 SDF_RORECOVERY = 7, /* read only recovery */
541 SDF_SKIP_DLM_UNLOCK = 8,
544 #define GFS2_FSNAME_LEN 256
546 struct gfs2_inum_host {
551 struct gfs2_sb_host {
557 u32 sb_multihost_format;
561 struct gfs2_inum_host sb_master_dir;
562 struct gfs2_inum_host sb_root_dir;
564 char sb_lockproto[GFS2_LOCKNAME_LEN];
565 char sb_locktable[GFS2_LOCKNAME_LEN];
569 * lm_mount() return values
571 * ls_jid - the journal ID this node should use
572 * ls_first - this node is the first to mount the file system
573 * ls_lockspace - lock module's context for this file system
574 * ls_ops - lock module's functions
577 struct lm_lockstruct {
579 unsigned int ls_first;
580 const struct lm_lockops *ls_ops;
581 dlm_lockspace_t *ls_dlm;
583 int ls_recover_jid_done; /* These two are deprecated, */
584 int ls_recover_jid_status; /* used previously by gfs_controld */
586 struct dlm_lksb ls_mounted_lksb; /* mounted_lock */
587 struct dlm_lksb ls_control_lksb; /* control_lock */
588 char ls_control_lvb[GDLM_LVB_SIZE]; /* control_lock lvb */
589 struct completion ls_sync_wait; /* {control,mounted}_{lock,unlock} */
591 spinlock_t ls_recover_spin; /* protects following fields */
592 unsigned long ls_recover_flags; /* DFL_ */
593 uint32_t ls_recover_mount; /* gen in first recover_done cb */
594 uint32_t ls_recover_start; /* gen in last recover_done cb */
595 uint32_t ls_recover_block; /* copy recover_start in last recover_prep */
596 uint32_t ls_recover_size; /* size of recover_submit, recover_result */
597 uint32_t *ls_recover_submit; /* gen in last recover_slot cb per jid */
598 uint32_t *ls_recover_result; /* result of last jid recovery */
601 struct gfs2_pcpu_lkstats {
602 /* One struct for each glock type */
603 struct gfs2_lkstats lkstats[10];
607 struct super_block *sd_vfs;
608 struct gfs2_pcpu_lkstats __percpu *sd_lkstats;
609 struct kobject sd_kobj;
610 unsigned long sd_flags; /* SDF_... */
611 struct gfs2_sb_host sd_sb;
613 /* Constants computed on mount */
617 u32 sd_diptrs; /* Number of pointers in a dinode */
618 u32 sd_inptrs; /* Number of pointers in a indirect block */
619 u32 sd_jbsize; /* Size of a journaled data block */
620 u32 sd_hash_bsize; /* sizeof(exhash block) */
621 u32 sd_hash_bsize_shift;
622 u32 sd_hash_ptrs; /* Number of pointers in a hash block */
624 u32 sd_blocks_per_bitmap;
625 u32 sd_max_dirres; /* Max blocks needed to add a directory entry */
626 u32 sd_max_height; /* Max height of a file's metadata tree */
627 u64 sd_heightsize[GFS2_MAX_META_HEIGHT + 1];
628 u32 sd_max_jheight; /* Max height of journaled file's meta tree */
629 u64 sd_jheightsize[GFS2_MAX_META_HEIGHT + 1];
631 struct gfs2_args sd_args; /* Mount arguments */
632 struct gfs2_tune sd_tune; /* Filesystem tuning structure */
636 struct lm_lockstruct sd_lockstruct;
637 struct gfs2_holder sd_live_gh;
638 struct gfs2_glock *sd_rename_gl;
639 struct gfs2_glock *sd_trans_gl;
640 wait_queue_head_t sd_glock_wait;
641 atomic_t sd_glock_disposal;
642 struct completion sd_locking_init;
643 struct delayed_work sd_control_work;
647 struct dentry *sd_master_dir;
648 struct dentry *sd_root_dir;
650 struct inode *sd_jindex;
651 struct inode *sd_statfs_inode;
652 struct inode *sd_sc_inode;
653 struct inode *sd_qc_inode;
654 struct inode *sd_rindex;
655 struct inode *sd_quota_inode;
659 spinlock_t sd_statfs_spin;
660 struct gfs2_statfs_change_host sd_statfs_master;
661 struct gfs2_statfs_change_host sd_statfs_local;
662 int sd_statfs_force_sync;
664 /* Resource group stuff */
666 int sd_rindex_uptodate;
667 spinlock_t sd_rindex_spin;
668 struct rb_root sd_rindex_tree;
669 unsigned int sd_rgrps;
670 unsigned int sd_max_rg_data;
672 /* Journal index stuff */
674 struct list_head sd_jindex_list;
675 spinlock_t sd_jindex_spin;
676 struct mutex sd_jindex_mutex;
677 unsigned int sd_journals;
679 struct gfs2_jdesc *sd_jdesc;
680 struct gfs2_holder sd_journal_gh;
681 struct gfs2_holder sd_jinode_gh;
683 struct gfs2_holder sd_sc_gh;
684 struct gfs2_holder sd_qc_gh;
688 struct task_struct *sd_logd_process;
689 struct task_struct *sd_quotad_process;
693 struct list_head sd_quota_list;
694 atomic_t sd_quota_count;
695 struct mutex sd_quota_mutex;
696 wait_queue_head_t sd_quota_wait;
697 struct list_head sd_trunc_list;
698 spinlock_t sd_trunc_lock;
700 unsigned int sd_quota_slots;
701 unsigned int sd_quota_chunks;
702 unsigned char **sd_quota_bitmap;
704 u64 sd_quota_sync_gen;
708 spinlock_t sd_log_lock;
710 unsigned int sd_log_blks_reserved;
711 unsigned int sd_log_commited_buf;
712 unsigned int sd_log_commited_databuf;
713 int sd_log_commited_revoke;
715 atomic_t sd_log_pinned;
716 unsigned int sd_log_num_buf;
717 unsigned int sd_log_num_revoke;
718 unsigned int sd_log_num_rg;
719 unsigned int sd_log_num_databuf;
721 struct list_head sd_log_le_buf;
722 struct list_head sd_log_le_revoke;
723 struct list_head sd_log_le_databuf;
724 struct list_head sd_log_le_ordered;
726 atomic_t sd_log_thresh1;
727 atomic_t sd_log_thresh2;
728 atomic_t sd_log_blks_free;
729 wait_queue_head_t sd_log_waitq;
730 wait_queue_head_t sd_logd_waitq;
733 unsigned int sd_log_head;
734 unsigned int sd_log_tail;
737 struct rw_semaphore sd_log_flush_lock;
738 atomic_t sd_log_in_flight;
739 struct bio *sd_log_bio;
740 wait_queue_head_t sd_log_flush_wait;
743 unsigned int sd_log_flush_head;
744 u64 sd_log_flush_wrapped;
746 spinlock_t sd_ail_lock;
747 struct list_head sd_ail1_list;
748 struct list_head sd_ail2_list;
752 struct list_head sd_revoke_list;
753 unsigned int sd_replay_tail;
755 unsigned int sd_found_blocks;
756 unsigned int sd_found_revokes;
757 unsigned int sd_replayed_blocks;
759 /* For quiescing the filesystem */
761 struct gfs2_holder sd_freeze_gh;
762 struct mutex sd_freeze_lock;
763 unsigned int sd_freeze_count;
765 char sd_fsname[GFS2_FSNAME_LEN];
766 char sd_table_name[GFS2_FSNAME_LEN];
767 char sd_proto_name[GFS2_FSNAME_LEN];
771 unsigned long sd_last_warning;
772 struct dentry *debugfs_dir; /* debugfs directory */
773 struct dentry *debugfs_dentry_glocks;
774 struct dentry *debugfs_dentry_glstats;
775 struct dentry *debugfs_dentry_sbstats;
778 static inline void gfs2_glstats_inc(struct gfs2_glock *gl, int which)
780 gl->gl_stats.stats[which]++;
783 static inline void gfs2_sbstats_inc(const struct gfs2_glock *gl, int which)
785 const struct gfs2_sbd *sdp = gl->gl_sbd;
787 this_cpu_ptr(sdp->sd_lkstats)->lkstats[gl->gl_name.ln_type].stats[which]++;
791 #endif /* __INCORE_DOT_H__ */