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;
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 */
101 enum gfs2_state_bits {
102 BH_Pinned = BH_PrivateStart,
103 BH_Escaped = BH_PrivateStart + 1,
104 BH_Zeronew = BH_PrivateStart + 2,
107 BUFFER_FNS(Pinned, pinned)
108 TAS_BUFFER_FNS(Pinned, pinned)
109 BUFFER_FNS(Escaped, escaped)
110 TAS_BUFFER_FNS(Escaped, escaped)
111 BUFFER_FNS(Zeronew, zeronew)
112 TAS_BUFFER_FNS(Zeronew, zeronew)
114 struct gfs2_bufdata {
115 struct buffer_head *bd_bh;
116 struct gfs2_glock *bd_gl;
119 struct list_head bd_list;
120 const struct gfs2_log_operations *bd_ops;
122 struct gfs2_ail *bd_ail;
123 struct list_head bd_ail_st_list;
124 struct list_head bd_ail_gl_list;
128 * Internally, we prefix things with gdlm_ and GDLM_ (for gfs-dlm) since a
129 * prefix of lock_dlm_ gets awkward.
132 #define GDLM_STRNAME_BYTES 25
133 #define GDLM_LVB_SIZE 32
138 * DFL_BLOCK_LOCKS: dlm is in recovery and will grant locks that had been
139 * held by failed nodes whose journals need recovery. Those locks should
140 * only be used for journal recovery until the journal recovery is done.
141 * This is set by the dlm recover_prep callback and cleared by the
142 * gfs2_control thread when journal recovery is complete. To avoid
143 * races between recover_prep setting and gfs2_control clearing, recover_spin
144 * is held while changing this bit and reading/writing recover_block
147 * DFL_NO_DLM_OPS: dlm lockspace ops/callbacks are not being used.
149 * DFL_FIRST_MOUNT: this node is the first to mount this fs and is doing
150 * recovery of all journals before allowing other nodes to mount the fs.
151 * This is cleared when FIRST_MOUNT_DONE is set.
153 * DFL_FIRST_MOUNT_DONE: this node was the first mounter, and has finished
154 * recovery of all journals, and now allows other nodes to mount the fs.
156 * DFL_MOUNT_DONE: gdlm_mount has completed successfully and cleared
157 * BLOCK_LOCKS for the first time. The gfs2_control thread should now
158 * control clearing BLOCK_LOCKS for further recoveries.
160 * DFL_UNMOUNT: gdlm_unmount sets to keep sdp off gfs2_control_wq.
162 * DFL_DLM_RECOVERY: set while dlm is in recovery, between recover_prep()
163 * and recover_done(), i.e. set while recover_block == recover_start.
170 DFL_FIRST_MOUNT_DONE = 3,
173 DFL_DLM_RECOVERY = 6,
178 unsigned int ln_type;
181 #define lm_name_equal(name1, name2) \
182 (((name1)->ln_number == (name2)->ln_number) && \
183 ((name1)->ln_type == (name2)->ln_type))
186 struct gfs2_glock_operations {
187 void (*go_xmote_th) (struct gfs2_glock *gl);
188 int (*go_xmote_bh) (struct gfs2_glock *gl, struct gfs2_holder *gh);
189 void (*go_inval) (struct gfs2_glock *gl, int flags);
190 int (*go_demote_ok) (const struct gfs2_glock *gl);
191 int (*go_lock) (struct gfs2_holder *gh);
192 void (*go_unlock) (struct gfs2_holder *gh);
193 int (*go_dump)(struct seq_file *seq, const struct gfs2_glock *gl);
194 void (*go_callback) (struct gfs2_glock *gl);
196 const unsigned long go_flags;
197 #define GLOF_ASPACE 1
201 GFS2_LKS_SRTT = 0, /* Non blocking smoothed round trip time */
202 GFS2_LKS_SRTTVAR = 1, /* Non blocking smoothed variance */
203 GFS2_LKS_SRTTB = 2, /* Blocking smoothed round trip time */
204 GFS2_LKS_SRTTVARB = 3, /* Blocking smoothed variance */
205 GFS2_LKS_SIRT = 4, /* Smoothed Inter-request time */
206 GFS2_LKS_SIRTVAR = 5, /* Smoothed Inter-request variance */
207 GFS2_LKS_DCOUNT = 6, /* Count of dlm requests */
208 GFS2_LKS_QCOUNT = 7, /* Count of gfs2_holder queues */
212 struct gfs2_lkstats {
213 s64 stats[GFS2_NR_LKSTATS];
218 HIF_HOLDER = 6, /* Set for gh that "holds" the glock */
224 struct list_head gh_list;
226 struct gfs2_glock *gh_gl;
227 struct pid *gh_owner_pid;
228 unsigned int gh_state;
232 unsigned long gh_iflags; /* HIF_... */
239 GLF_PENDING_DEMOTE = 4,
240 GLF_DEMOTE_IN_PROGRESS = 5,
243 GLF_INVALIDATE_IN_PROGRESS = 8,
244 GLF_REPLY_PENDING = 9,
249 GLF_OBJECT = 14, /* Used only for tracing */
254 struct hlist_bl_node gl_list;
255 struct gfs2_sbd *gl_sbd;
256 unsigned long gl_flags; /* GLF_... */
257 struct lm_lockname gl_name;
262 /* State fields protected by gl_spin */
263 unsigned int gl_state:2, /* Current state */
264 gl_target:2, /* Target state */
265 gl_demote_state:2, /* State requested by remote node */
266 gl_req:2, /* State in last dlm request */
267 gl_reply:8; /* Last reply from the dlm */
269 unsigned int gl_hash;
270 unsigned long gl_demote_time; /* time of first demote request */
272 struct list_head gl_holders;
274 const struct gfs2_glock_operations *gl_ops;
276 struct gfs2_lkstats gl_stats;
277 struct dlm_lksb gl_lksb;
279 unsigned long gl_tchange;
282 struct list_head gl_lru;
283 struct list_head gl_ail_list;
284 atomic_t gl_ail_count;
286 struct delayed_work gl_work;
287 struct work_struct gl_delete;
288 struct rcu_head gl_rcu;
291 #define GFS2_MIN_LVB_SIZE 32 /* Min size of LVB that gfs2 supports */
293 struct gfs2_blkreserv {
294 u32 rs_requested; /* Filled in by caller of gfs2_inplace_reserve() */
295 struct gfs2_holder rs_rgd_gh; /* Filled in by gfs2_inplace_reserve() */
297 /* ancillary quota stuff */
298 struct gfs2_quota_data *rs_qa_qd[2 * MAXQUOTAS];
299 struct gfs2_holder rs_qa_qd_ghs[2 * MAXQUOTAS];
300 unsigned int rs_qa_qd_num;
306 GIF_ALLOC_FAILED = 2,
312 struct inode i_inode;
317 unsigned long i_flags; /* GIF_... */
318 struct gfs2_glock *i_gl; /* Move into i_gh? */
319 struct gfs2_holder i_iopen_gh;
320 struct gfs2_holder i_gh; /* for prepare/commit_write only */
321 struct gfs2_blkreserv *i_res; /* resource group block reservation */
322 struct gfs2_rgrpd *i_rgd;
323 u64 i_goal; /* goal block for allocations */
324 struct rw_semaphore i_rw_mutex;
325 struct list_head i_trunc_list;
326 __be64 *i_hash_cache;
334 * Since i_inode is the first element of struct gfs2_inode,
335 * this is effectively a cast.
337 static inline struct gfs2_inode *GFS2_I(struct inode *inode)
339 return container_of(inode, struct gfs2_inode, i_inode);
342 static inline struct gfs2_sbd *GFS2_SB(const struct inode *inode)
344 return inode->i_sb->s_fs_info;
348 struct mutex f_fl_mutex;
349 struct gfs2_holder f_fl_gh;
352 struct gfs2_revoke_replay {
353 struct list_head rr_list;
355 unsigned int rr_where;
365 struct gfs2_quota_data {
366 struct list_head qd_list;
367 struct list_head qd_reclaim;
372 unsigned long qd_flags; /* QDF_... */
377 unsigned int qd_slot;
378 unsigned int qd_slot_count;
380 struct buffer_head *qd_bh;
381 struct gfs2_quota_change *qd_bh_qc;
382 unsigned int qd_bh_count;
384 struct gfs2_glock *qd_gl;
385 struct gfs2_quota_lvb qd_qb;
388 unsigned long qd_last_warn;
394 unsigned int tr_blocks;
395 unsigned int tr_revokes;
396 unsigned int tr_reserved;
398 struct gfs2_holder tr_t_gh;
402 unsigned int tr_num_buf_new;
403 unsigned int tr_num_databuf_new;
404 unsigned int tr_num_buf_rm;
405 unsigned int tr_num_databuf_rm;
406 unsigned int tr_num_revoke;
407 unsigned int tr_num_revoke_rm;
411 struct list_head ai_list;
413 unsigned int ai_first;
414 struct list_head ai_ail1_list;
415 struct list_head ai_ail2_list;
418 struct gfs2_journal_extent {
419 struct list_head extent_list;
421 unsigned int lblock; /* First logical block */
422 u64 dblock; /* First disk block */
427 struct list_head jd_list;
428 struct list_head extent_list;
429 struct work_struct jd_work;
430 struct inode *jd_inode;
431 unsigned long jd_flags;
432 #define JDF_RECOVERY 1
434 unsigned int jd_blocks;
435 int jd_recover_error;
438 struct gfs2_statfs_change_host {
444 #define GFS2_QUOTA_DEFAULT GFS2_QUOTA_OFF
445 #define GFS2_QUOTA_OFF 0
446 #define GFS2_QUOTA_ACCOUNT 1
447 #define GFS2_QUOTA_ON 2
449 #define GFS2_DATA_DEFAULT GFS2_DATA_ORDERED
450 #define GFS2_DATA_WRITEBACK 1
451 #define GFS2_DATA_ORDERED 2
453 #define GFS2_ERRORS_DEFAULT GFS2_ERRORS_WITHDRAW
454 #define GFS2_ERRORS_WITHDRAW 0
455 #define GFS2_ERRORS_CONTINUE 1 /* place holder for future feature */
456 #define GFS2_ERRORS_RO 2 /* place holder for future feature */
457 #define GFS2_ERRORS_PANIC 3
460 char ar_lockproto[GFS2_LOCKNAME_LEN]; /* Name of the Lock Protocol */
461 char ar_locktable[GFS2_LOCKNAME_LEN]; /* Name of the Lock Table */
462 char ar_hostdata[GFS2_LOCKNAME_LEN]; /* Host specific data */
463 unsigned int ar_spectator:1; /* Don't get a journal */
464 unsigned int ar_localflocks:1; /* Let the VFS do flock|fcntl */
465 unsigned int ar_debug:1; /* Oops on errors */
466 unsigned int ar_posix_acl:1; /* Enable posix acls */
467 unsigned int ar_quota:2; /* off/account/on */
468 unsigned int ar_suiddir:1; /* suiddir support */
469 unsigned int ar_data:2; /* ordered/writeback */
470 unsigned int ar_meta:1; /* mount metafs */
471 unsigned int ar_discard:1; /* discard requests */
472 unsigned int ar_errors:2; /* errors=withdraw | panic */
473 unsigned int ar_nobarrier:1; /* do not send barriers */
474 unsigned int ar_rgrplvb:1; /* use lvbs for rgrp info */
475 int ar_commit; /* Commit interval */
476 int ar_statfs_quantum; /* The fast statfs interval */
477 int ar_quota_quantum; /* The quota interval */
478 int ar_statfs_percent; /* The % change to force sync */
484 unsigned int gt_logd_secs;
486 unsigned int gt_quota_simul_sync; /* Max quotavals to sync at once */
487 unsigned int gt_quota_warn_period; /* Secs between quota warn msgs */
488 unsigned int gt_quota_scale_num; /* Numerator */
489 unsigned int gt_quota_scale_den; /* Denominator */
490 unsigned int gt_quota_quantum; /* Secs between syncs to quota file */
491 unsigned int gt_new_files_jdata;
492 unsigned int gt_max_readahead; /* Max bytes to read-ahead from disk */
493 unsigned int gt_complain_secs;
494 unsigned int gt_statfs_quantum;
495 unsigned int gt_statfs_slow;
499 SDF_JOURNAL_CHECKED = 0,
500 SDF_JOURNAL_LIVE = 1,
506 SDF_RORECOVERY = 7, /* read only recovery */
509 #define GFS2_FSNAME_LEN 256
511 struct gfs2_inum_host {
516 struct gfs2_sb_host {
522 u32 sb_multihost_format;
526 struct gfs2_inum_host sb_master_dir;
527 struct gfs2_inum_host sb_root_dir;
529 char sb_lockproto[GFS2_LOCKNAME_LEN];
530 char sb_locktable[GFS2_LOCKNAME_LEN];
534 * lm_mount() return values
536 * ls_jid - the journal ID this node should use
537 * ls_first - this node is the first to mount the file system
538 * ls_lockspace - lock module's context for this file system
539 * ls_ops - lock module's functions
542 struct lm_lockstruct {
544 unsigned int ls_first;
545 const struct lm_lockops *ls_ops;
546 dlm_lockspace_t *ls_dlm;
548 int ls_recover_jid_done; /* These two are deprecated, */
549 int ls_recover_jid_status; /* used previously by gfs_controld */
551 struct dlm_lksb ls_mounted_lksb; /* mounted_lock */
552 struct dlm_lksb ls_control_lksb; /* control_lock */
553 char ls_control_lvb[GDLM_LVB_SIZE]; /* control_lock lvb */
554 struct completion ls_sync_wait; /* {control,mounted}_{lock,unlock} */
556 spinlock_t ls_recover_spin; /* protects following fields */
557 unsigned long ls_recover_flags; /* DFL_ */
558 uint32_t ls_recover_mount; /* gen in first recover_done cb */
559 uint32_t ls_recover_start; /* gen in last recover_done cb */
560 uint32_t ls_recover_block; /* copy recover_start in last recover_prep */
561 uint32_t ls_recover_size; /* size of recover_submit, recover_result */
562 uint32_t *ls_recover_submit; /* gen in last recover_slot cb per jid */
563 uint32_t *ls_recover_result; /* result of last jid recovery */
566 struct gfs2_pcpu_lkstats {
567 /* One struct for each glock type */
568 struct gfs2_lkstats lkstats[10];
572 struct super_block *sd_vfs;
573 struct gfs2_pcpu_lkstats __percpu *sd_lkstats;
574 struct kobject sd_kobj;
575 unsigned long sd_flags; /* SDF_... */
576 struct gfs2_sb_host sd_sb;
578 /* Constants computed on mount */
582 u32 sd_diptrs; /* Number of pointers in a dinode */
583 u32 sd_inptrs; /* Number of pointers in a indirect block */
584 u32 sd_jbsize; /* Size of a journaled data block */
585 u32 sd_hash_bsize; /* sizeof(exhash block) */
586 u32 sd_hash_bsize_shift;
587 u32 sd_hash_ptrs; /* Number of pointers in a hash block */
589 u32 sd_max_dirres; /* Max blocks needed to add a directory entry */
590 u32 sd_max_height; /* Max height of a file's metadata tree */
591 u64 sd_heightsize[GFS2_MAX_META_HEIGHT + 1];
592 u32 sd_max_jheight; /* Max height of journaled file's meta tree */
593 u64 sd_jheightsize[GFS2_MAX_META_HEIGHT + 1];
595 struct gfs2_args sd_args; /* Mount arguments */
596 struct gfs2_tune sd_tune; /* Filesystem tuning structure */
600 struct lm_lockstruct sd_lockstruct;
601 struct gfs2_holder sd_live_gh;
602 struct gfs2_glock *sd_rename_gl;
603 struct gfs2_glock *sd_trans_gl;
604 wait_queue_head_t sd_glock_wait;
605 atomic_t sd_glock_disposal;
606 struct completion sd_locking_init;
607 struct delayed_work sd_control_work;
611 struct dentry *sd_master_dir;
612 struct dentry *sd_root_dir;
614 struct inode *sd_jindex;
615 struct inode *sd_statfs_inode;
616 struct inode *sd_sc_inode;
617 struct inode *sd_qc_inode;
618 struct inode *sd_rindex;
619 struct inode *sd_quota_inode;
623 spinlock_t sd_statfs_spin;
624 struct gfs2_statfs_change_host sd_statfs_master;
625 struct gfs2_statfs_change_host sd_statfs_local;
626 int sd_statfs_force_sync;
628 /* Resource group stuff */
630 int sd_rindex_uptodate;
631 spinlock_t sd_rindex_spin;
632 struct rb_root sd_rindex_tree;
633 unsigned int sd_rgrps;
634 unsigned int sd_max_rg_data;
636 /* Journal index stuff */
638 struct list_head sd_jindex_list;
639 spinlock_t sd_jindex_spin;
640 struct mutex sd_jindex_mutex;
641 unsigned int sd_journals;
643 struct gfs2_jdesc *sd_jdesc;
644 struct gfs2_holder sd_journal_gh;
645 struct gfs2_holder sd_jinode_gh;
647 struct gfs2_holder sd_sc_gh;
648 struct gfs2_holder sd_qc_gh;
652 struct task_struct *sd_logd_process;
653 struct task_struct *sd_quotad_process;
657 struct list_head sd_quota_list;
658 atomic_t sd_quota_count;
659 struct mutex sd_quota_mutex;
660 wait_queue_head_t sd_quota_wait;
661 struct list_head sd_trunc_list;
662 spinlock_t sd_trunc_lock;
664 unsigned int sd_quota_slots;
665 unsigned int sd_quota_chunks;
666 unsigned char **sd_quota_bitmap;
668 u64 sd_quota_sync_gen;
672 spinlock_t sd_log_lock;
674 unsigned int sd_log_blks_reserved;
675 unsigned int sd_log_commited_buf;
676 unsigned int sd_log_commited_databuf;
677 int sd_log_commited_revoke;
679 atomic_t sd_log_pinned;
680 unsigned int sd_log_num_buf;
681 unsigned int sd_log_num_revoke;
682 unsigned int sd_log_num_rg;
683 unsigned int sd_log_num_databuf;
685 struct list_head sd_log_le_buf;
686 struct list_head sd_log_le_revoke;
687 struct list_head sd_log_le_databuf;
688 struct list_head sd_log_le_ordered;
690 atomic_t sd_log_thresh1;
691 atomic_t sd_log_thresh2;
692 atomic_t sd_log_blks_free;
693 wait_queue_head_t sd_log_waitq;
694 wait_queue_head_t sd_logd_waitq;
697 unsigned int sd_log_head;
698 unsigned int sd_log_tail;
701 struct rw_semaphore sd_log_flush_lock;
702 atomic_t sd_log_in_flight;
703 struct bio *sd_log_bio;
704 wait_queue_head_t sd_log_flush_wait;
707 unsigned int sd_log_flush_head;
708 u64 sd_log_flush_wrapped;
710 spinlock_t sd_ail_lock;
711 struct list_head sd_ail1_list;
712 struct list_head sd_ail2_list;
716 struct list_head sd_revoke_list;
717 unsigned int sd_replay_tail;
719 unsigned int sd_found_blocks;
720 unsigned int sd_found_revokes;
721 unsigned int sd_replayed_blocks;
723 /* For quiescing the filesystem */
725 struct gfs2_holder sd_freeze_gh;
726 struct mutex sd_freeze_lock;
727 unsigned int sd_freeze_count;
729 char sd_fsname[GFS2_FSNAME_LEN];
730 char sd_table_name[GFS2_FSNAME_LEN];
731 char sd_proto_name[GFS2_FSNAME_LEN];
735 unsigned long sd_last_warning;
736 struct dentry *debugfs_dir; /* debugfs directory */
737 struct dentry *debugfs_dentry_glocks;
738 struct dentry *debugfs_dentry_glstats;
739 struct dentry *debugfs_dentry_sbstats;
742 static inline void gfs2_glstats_inc(struct gfs2_glock *gl, int which)
744 gl->gl_stats.stats[which]++;
747 static inline void gfs2_sbstats_inc(const struct gfs2_glock *gl, int which)
749 const struct gfs2_sbd *sdp = gl->gl_sbd;
751 this_cpu_ptr(sdp->sd_lkstats)->lkstats[gl->gl_name.ln_type].stats[which]++;
755 #endif /* __INCORE_DOT_H__ */