1 /******************************************************************************
2 *******************************************************************************
4 ** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
5 ** Copyright (C) 2004-2011 Red Hat, Inc. All rights reserved.
7 ** This copyrighted material is made available to anyone wishing to use,
8 ** modify, copy, or redistribute it subject to the terms and conditions
9 ** of the GNU General Public License v.2.
11 *******************************************************************************
12 ******************************************************************************/
14 #ifndef __DLM_INTERNAL_DOT_H__
15 #define __DLM_INTERNAL_DOT_H__
18 * This is the main header file to be included in each DLM source file.
21 #include <linux/module.h>
22 #include <linux/slab.h>
23 #include <linux/sched.h>
24 #include <linux/types.h>
25 #include <linux/ctype.h>
26 #include <linux/spinlock.h>
27 #include <linux/vmalloc.h>
28 #include <linux/list.h>
29 #include <linux/errno.h>
30 #include <linux/random.h>
31 #include <linux/delay.h>
32 #include <linux/socket.h>
33 #include <linux/kthread.h>
34 #include <linux/kobject.h>
35 #include <linux/kref.h>
36 #include <linux/kernel.h>
37 #include <linux/jhash.h>
38 #include <linux/miscdevice.h>
39 #include <linux/mutex.h>
40 #include <linux/idr.h>
41 #include <linux/ratelimit.h>
42 #include <asm/uaccess.h>
44 #include <linux/dlm.h>
47 /* Size of the temp buffer midcomms allocates on the stack.
48 We try to make this large enough so most messages fit.
49 FIXME: should sctp make this unnecessary? */
51 #define DLM_INBUF_LEN 148
64 #define log_print(fmt, args...) \
65 printk(KERN_ERR "dlm: "fmt"\n" , ##args)
66 #define log_error(ls, fmt, args...) \
67 printk(KERN_ERR "dlm: %s: " fmt "\n", (ls)->ls_name , ##args)
69 #define log_debug(ls, fmt, args...) \
71 if (dlm_config.ci_log_debug) \
72 printk(KERN_DEBUG "dlm: %s: " fmt "\n", \
73 (ls)->ls_name , ##args); \
76 #define log_limit(ls, fmt, args...) \
78 if (dlm_config.ci_log_debug) \
79 printk_ratelimited(KERN_DEBUG "dlm: %s: " fmt "\n", \
80 (ls)->ls_name , ##args); \
83 #define DLM_ASSERT(x, do) \
87 printk(KERN_ERR "\nDLM: Assertion failed on line %d of file %s\n" \
88 "DLM: assertion: \"%s\"\n" \
89 "DLM: time = %lu\n", \
90 __LINE__, __FILE__, #x, jiffies); \
94 panic("DLM: Record message above and reboot.\n"); \
107 * Lockspace member (per node in a ls)
111 struct list_head list;
121 * Save and manage recovery state for a lockspace.
125 struct list_head list;
126 struct dlm_config_node *nodes;
132 * Pass input args to second stage locking function.
137 void (*astfn) (void *astparam);
139 void (*bastfn) (void *astparam, int mode);
141 struct dlm_lksb *lksb;
142 unsigned long timeout;
149 * A lock can be one of three types:
151 * local copy lock is mastered locally
152 * (lkb_nodeid is zero and DLM_LKF_MSTCPY is not set)
153 * process copy lock is mastered on a remote node
154 * (lkb_nodeid is non-zero and DLM_LKF_MSTCPY is not set)
155 * master copy master node's copy of a lock owned by remote node
156 * (lkb_nodeid is non-zero and DLM_LKF_MSTCPY is set)
158 * lkb_exflags: a copy of the most recent flags arg provided to dlm_lock or
159 * dlm_unlock. The dlm does not modify these or use any private flags in
160 * this field; it only contains DLM_LKF_ flags from dlm.h. These flags
161 * are sent as-is to the remote master when the lock is remote.
163 * lkb_flags: internal dlm flags (DLM_IFL_ prefix) from dlm_internal.h.
164 * Some internal flags are shared between the master and process nodes;
165 * these shared flags are kept in the lower two bytes. One of these
166 * flags set on the master copy will be propagated to the process copy
167 * and v.v. Other internal flags are private to the master or process
168 * node (e.g. DLM_IFL_MSTCPY). These are kept in the high two bytes.
170 * lkb_sbflags: status block flags. These flags are copied directly into
171 * the caller's lksb.sb_flags prior to the dlm_lock/dlm_unlock completion
172 * ast. All defined in dlm.h with DLM_SBF_ prefix.
174 * lkb_status: the lock status indicates which rsb queue the lock is
175 * on, grant, convert, or wait. DLM_LKSTS_ WAITING/GRANTED/CONVERT
177 * lkb_wait_type: the dlm message type (DLM_MSG_ prefix) for which a
178 * reply is needed. Only set when the lkb is on the lockspace waiters
179 * list awaiting a reply from a remote node.
181 * lkb_nodeid: when the lkb is a local copy, nodeid is 0; when the lkb
182 * is a master copy, nodeid specifies the remote lock holder, when the
183 * lkb is a process copy, the nodeid specifies the lock master.
188 #define DLM_LKSTS_WAITING 1
189 #define DLM_LKSTS_GRANTED 2
190 #define DLM_LKSTS_CONVERT 3
194 #define DLM_IFL_MSTCPY 0x00010000
195 #define DLM_IFL_RESEND 0x00020000
196 #define DLM_IFL_DEAD 0x00040000
197 #define DLM_IFL_OVERLAP_UNLOCK 0x00080000
198 #define DLM_IFL_OVERLAP_CANCEL 0x00100000
199 #define DLM_IFL_ENDOFLIFE 0x00200000
200 #define DLM_IFL_WATCH_TIMEWARN 0x00400000
201 #define DLM_IFL_TIMEOUT_CANCEL 0x00800000
202 #define DLM_IFL_DEADLOCK_CANCEL 0x01000000
203 #define DLM_IFL_STUB_MS 0x02000000 /* magic number for m_flags */
204 #define DLM_IFL_USER 0x00000001
205 #define DLM_IFL_ORPHAN 0x00000002
207 #define DLM_CALLBACKS_SIZE 6
209 #define DLM_CB_CAST 0x00000001
210 #define DLM_CB_BAST 0x00000002
211 #define DLM_CB_SKIP 0x00000004
213 struct dlm_callback {
215 uint32_t flags; /* DLM_CBF_ */
216 int sb_status; /* copy to lksb status */
217 uint8_t sb_flags; /* copy to lksb flags */
218 int8_t mode; /* rq mode of bast, gr mode of cast */
222 struct dlm_rsb *lkb_resource; /* the rsb */
224 int lkb_nodeid; /* copied from rsb */
225 int lkb_ownpid; /* pid of lock owner */
226 uint32_t lkb_id; /* our lock ID */
227 uint32_t lkb_remid; /* lock ID on remote partner */
228 uint32_t lkb_exflags; /* external flags from caller */
229 uint32_t lkb_sbflags; /* lksb flags */
230 uint32_t lkb_flags; /* internal flags */
231 uint32_t lkb_lvbseq; /* lvb sequence number */
233 int8_t lkb_status; /* granted, waiting, convert */
234 int8_t lkb_rqmode; /* requested lock mode */
235 int8_t lkb_grmode; /* granted lock mode */
236 int8_t lkb_highbast; /* highest mode bast sent for */
238 int8_t lkb_wait_type; /* type of reply waiting for */
239 int8_t lkb_wait_count;
240 int lkb_wait_nodeid; /* for debugging */
242 struct list_head lkb_statequeue; /* rsb g/c/w list */
243 struct list_head lkb_rsb_lookup; /* waiting for rsb lookup */
244 struct list_head lkb_wait_reply; /* waiting for remote reply */
245 struct list_head lkb_ownqueue; /* list of locks for a process */
246 struct list_head lkb_time_list;
247 ktime_t lkb_timestamp;
248 ktime_t lkb_wait_time;
249 unsigned long lkb_timeout_cs;
251 struct mutex lkb_cb_mutex;
252 struct work_struct lkb_cb_work;
253 struct list_head lkb_cb_list; /* for ls_cb_delay or proc->asts */
254 struct dlm_callback lkb_callbacks[DLM_CALLBACKS_SIZE];
255 struct dlm_callback lkb_last_cast;
256 struct dlm_callback lkb_last_bast;
257 ktime_t lkb_last_cast_time; /* for debugging */
258 ktime_t lkb_last_bast_time; /* for debugging */
260 uint64_t lkb_recover_seq; /* from ls_recover_seq */
263 struct dlm_lksb *lkb_lksb; /* caller's status block */
264 void (*lkb_astfn) (void *astparam);
265 void (*lkb_bastfn) (void *astparam, int mode);
267 void *lkb_astparam; /* caller's ast arg */
268 struct dlm_user_args *lkb_ua;
273 * res_master_nodeid is "normal": 0 is unset/invalid, non-zero is the real
274 * nodeid, even when nodeid is our_nodeid.
276 * res_nodeid is "odd": -1 is unset/invalid, zero means our_nodeid,
277 * greater than zero when another nodeid.
279 * (TODO: remove res_nodeid and only use res_master_nodeid)
283 struct dlm_ls *res_ls; /* the lockspace */
285 struct mutex res_mutex;
286 unsigned long res_flags;
287 int res_length; /* length of rsb name */
289 int res_master_nodeid;
291 int res_id; /* for ls_recover_idr */
294 uint32_t res_bucket; /* rsbtbl */
295 unsigned long res_toss_time;
296 uint32_t res_first_lkid;
297 struct list_head res_lookup; /* lkbs waiting on first */
299 struct list_head res_hashchain;
300 struct rb_node res_hashnode; /* rsbtbl */
302 struct list_head res_grantqueue;
303 struct list_head res_convertqueue;
304 struct list_head res_waitqueue;
306 struct list_head res_root_list; /* used for recovery */
307 struct list_head res_recover_list; /* used for recovery */
308 int res_recover_locks_count;
311 char res_name[DLM_RESNAME_MAXLEN+1];
314 /* dlm_master_lookup() flags */
316 #define DLM_LU_RECOVER_DIR 1
317 #define DLM_LU_RECOVER_MASTER 2
319 /* dlm_master_lookup() results */
321 #define DLM_LU_MATCH 1
324 /* find_rsb() flags */
326 #define R_REQUEST 0x00000001
327 #define R_RECEIVE_REQUEST 0x00000002
328 #define R_RECEIVE_RECOVER 0x00000004
333 RSB_MASTER_UNCERTAIN,
335 RSB_VALNOTVALID_PREV,
340 RSB_RECOVER_LVB_INVAL,
343 static inline void rsb_set_flag(struct dlm_rsb *r, enum rsb_flags flag)
345 __set_bit(flag, &r->res_flags);
348 static inline void rsb_clear_flag(struct dlm_rsb *r, enum rsb_flags flag)
350 __clear_bit(flag, &r->res_flags);
353 static inline int rsb_flag(struct dlm_rsb *r, enum rsb_flags flag)
355 return test_bit(flag, &r->res_flags);
359 /* dlm_header is first element of all structs sent between nodes */
361 #define DLM_HEADER_MAJOR 0x00030000
362 #define DLM_HEADER_MINOR 0x00000001
364 #define DLM_HEADER_SLOTS 0x00000001
371 uint32_t h_lockspace;
372 uint32_t h_nodeid; /* nodeid of sender */
374 uint8_t h_cmd; /* DLM_MSG, DLM_RCOM */
379 #define DLM_MSG_REQUEST 1
380 #define DLM_MSG_CONVERT 2
381 #define DLM_MSG_UNLOCK 3
382 #define DLM_MSG_CANCEL 4
383 #define DLM_MSG_REQUEST_REPLY 5
384 #define DLM_MSG_CONVERT_REPLY 6
385 #define DLM_MSG_UNLOCK_REPLY 7
386 #define DLM_MSG_CANCEL_REPLY 8
387 #define DLM_MSG_GRANT 9
388 #define DLM_MSG_BAST 10
389 #define DLM_MSG_LOOKUP 11
390 #define DLM_MSG_REMOVE 12
391 #define DLM_MSG_LOOKUP_REPLY 13
392 #define DLM_MSG_PURGE 14
395 struct dlm_header m_header;
396 uint32_t m_type; /* DLM_MSG_ */
399 uint32_t m_lkid; /* lkid on sender */
400 uint32_t m_remid; /* lkid on receiver */
401 uint32_t m_parent_lkid;
402 uint32_t m_parent_remid;
413 int m_result; /* 0 or -EXXX */
414 char m_extra[0]; /* name or lvb */
418 #define DLM_RS_NODES 0x00000001
419 #define DLM_RS_NODES_ALL 0x00000002
420 #define DLM_RS_DIR 0x00000004
421 #define DLM_RS_DIR_ALL 0x00000008
422 #define DLM_RS_LOCKS 0x00000010
423 #define DLM_RS_LOCKS_ALL 0x00000020
424 #define DLM_RS_DONE 0x00000040
425 #define DLM_RS_DONE_ALL 0x00000080
427 #define DLM_RCOM_STATUS 1
428 #define DLM_RCOM_NAMES 2
429 #define DLM_RCOM_LOOKUP 3
430 #define DLM_RCOM_LOCK 4
431 #define DLM_RCOM_STATUS_REPLY 5
432 #define DLM_RCOM_NAMES_REPLY 6
433 #define DLM_RCOM_LOOKUP_REPLY 7
434 #define DLM_RCOM_LOCK_REPLY 8
437 struct dlm_header rc_header;
438 uint32_t rc_type; /* DLM_RCOM_ */
439 int rc_result; /* multi-purpose */
440 uint64_t rc_id; /* match reply with request */
441 uint64_t rc_seq; /* sender's ls_recover_seq */
442 uint64_t rc_seq_reply; /* remote ls_recover_seq */
447 struct dlm_header header; /* common to other two */
448 struct dlm_message message;
449 struct dlm_rcom rcom;
452 #define DLM_RSF_NEED_SLOTS 0x00000001
454 /* RCOM_STATUS data */
461 /* RCOM_STATUS_REPLY data */
466 /* DLM_HEADER_SLOTS adds: */
470 __le32 rf_generation;
486 __le32 rl_parent_lkid;
487 __le32 rl_parent_remid;
498 char rl_name[DLM_RESNAME_MAXLEN];
503 * The max number of resources per rsbtbl bucket that shrink will attempt
504 * to remove in each iteration.
507 #define DLM_REMOVE_NAMES_MAX 8
510 struct list_head ls_list; /* list of lockspaces */
511 dlm_lockspace_t *ls_local_handle;
512 uint32_t ls_global_id; /* global unique lockspace ID */
513 uint32_t ls_generation;
516 int ls_count; /* refcount of processes in
517 the dlm using this ls */
518 int ls_create_count; /* create/release refcount */
519 unsigned long ls_flags; /* LSFL_ */
520 unsigned long ls_scan_time;
521 struct kobject ls_kobj;
523 struct idr ls_lkbidr;
524 spinlock_t ls_lkbidr_spin;
526 struct dlm_rsbtable *ls_rsbtbl;
527 uint32_t ls_rsbtbl_size;
529 struct mutex ls_waiters_mutex;
530 struct list_head ls_waiters; /* lkbs needing a reply */
532 struct mutex ls_orphans_mutex;
533 struct list_head ls_orphans;
535 struct mutex ls_timeout_mutex;
536 struct list_head ls_timeout;
538 spinlock_t ls_new_rsb_spin;
539 int ls_new_rsb_count;
540 struct list_head ls_new_rsb; /* new rsb structs */
542 spinlock_t ls_remove_spin;
543 char ls_remove_name[DLM_RESNAME_MAXLEN+1];
544 char *ls_remove_names[DLM_REMOVE_NAMES_MAX];
546 int ls_remove_lens[DLM_REMOVE_NAMES_MAX];
548 struct list_head ls_nodes; /* current nodes in ls */
549 struct list_head ls_nodes_gone; /* dead node list, recovery */
550 int ls_num_nodes; /* number of nodes in ls */
558 struct dlm_slot *ls_slots;
560 struct dlm_rsb ls_stub_rsb; /* for returning errors */
561 struct dlm_lkb ls_stub_lkb; /* for returning errors */
562 struct dlm_message ls_stub_ms; /* for faking a reply */
564 struct dentry *ls_debug_rsb_dentry; /* debugfs */
565 struct dentry *ls_debug_waiters_dentry; /* debugfs */
566 struct dentry *ls_debug_locks_dentry; /* debugfs */
567 struct dentry *ls_debug_all_dentry; /* debugfs */
568 struct dentry *ls_debug_toss_dentry; /* debugfs */
570 wait_queue_head_t ls_uevent_wait; /* user part of join/leave */
571 int ls_uevent_result;
572 struct completion ls_members_done;
573 int ls_members_result;
575 struct miscdevice ls_device;
577 struct workqueue_struct *ls_callback_wq;
579 /* recovery related */
581 struct mutex ls_cb_mutex;
582 struct list_head ls_cb_delay; /* save for queue_work later */
583 struct timer_list ls_timer;
584 struct task_struct *ls_recoverd_task;
585 struct mutex ls_recoverd_active;
586 spinlock_t ls_recover_lock;
587 unsigned long ls_recover_begin; /* jiffies timestamp */
588 uint32_t ls_recover_status; /* DLM_RS_ */
589 uint64_t ls_recover_seq;
590 struct dlm_recover *ls_recover_args;
591 struct rw_semaphore ls_in_recovery; /* block local requests */
592 struct rw_semaphore ls_recv_active; /* block dlm_recv */
593 struct list_head ls_requestqueue;/* queue remote requests */
594 struct mutex ls_requestqueue_mutex;
595 struct dlm_rcom *ls_recover_buf;
596 int ls_recover_nodeid; /* for debugging */
597 unsigned int ls_recover_dir_sent_res; /* for log info */
598 unsigned int ls_recover_dir_sent_msg; /* for log info */
599 unsigned int ls_recover_locks_in; /* for log info */
600 uint64_t ls_rcom_seq;
601 spinlock_t ls_rcom_spin;
602 struct list_head ls_recover_list;
603 spinlock_t ls_recover_list_lock;
604 int ls_recover_list_count;
605 struct idr ls_recover_idr;
606 spinlock_t ls_recover_idr_lock;
607 wait_queue_head_t ls_wait_general;
608 wait_queue_head_t ls_recover_lock_wait;
609 struct mutex ls_clear_proc_locks;
611 struct list_head ls_root_list; /* root resources */
612 struct rw_semaphore ls_root_sem; /* protect root_list */
614 const struct dlm_lockspace_ops *ls_ops;
622 * LSFL_RECOVER_STOP - dlm_ls_stop() sets this to tell dlm recovery routines
623 * that they should abort what they're doing so new recovery can be started.
625 * LSFL_RECOVER_DOWN - dlm_ls_stop() sets this to tell dlm_recoverd that it
626 * should do down_write() on the in_recovery rw_semaphore. (doing down_write
627 * within dlm_ls_stop causes complaints about the lock acquired/released
628 * in different contexts.)
630 * LSFL_RECOVER_LOCK - dlm_recoverd holds the in_recovery rw_semaphore.
631 * It sets this after it is done with down_write() on the in_recovery
632 * rw_semaphore and clears it after it has released the rw_semaphore.
634 * LSFL_RECOVER_WORK - dlm_ls_start() sets this to tell dlm_recoverd that it
635 * should begin recovery of the lockspace.
637 * LSFL_RUNNING - set when normal locking activity is enabled.
638 * dlm_ls_stop() clears this to tell dlm locking routines that they should
639 * quit what they are doing so recovery can run. dlm_recoverd sets
640 * this after recovery is finished.
643 #define LSFL_RECOVER_STOP 0
644 #define LSFL_RECOVER_DOWN 1
645 #define LSFL_RECOVER_LOCK 2
646 #define LSFL_RECOVER_WORK 3
647 #define LSFL_RUNNING 4
649 #define LSFL_RCOM_READY 5
650 #define LSFL_RCOM_WAIT 6
651 #define LSFL_UEVENT_WAIT 7
652 #define LSFL_TIMEWARN 8
653 #define LSFL_CB_DELAY 9
654 #define LSFL_NODIR 10
656 /* much of this is just saving user space pointers associated with the
657 lock that we pass back to the user lib with an ast */
659 struct dlm_user_args {
660 struct dlm_user_proc *proc; /* each process that opens the lockspace
661 device has private data
662 (dlm_user_proc) on the struct file,
663 the process's locks point back to it*/
664 struct dlm_lksb lksb;
665 struct dlm_lksb __user *user_lksb;
666 void __user *castparam;
667 void __user *castaddr;
668 void __user *bastparam;
669 void __user *bastaddr;
673 #define DLM_PROC_FLAGS_CLOSING 1
674 #define DLM_PROC_FLAGS_COMPAT 2
676 /* locks list is kept so we can remove all a process's locks when it
677 exits (or orphan those that are persistent) */
679 struct dlm_user_proc {
680 dlm_lockspace_t *lockspace;
681 unsigned long flags; /* DLM_PROC_FLAGS */
682 struct list_head asts;
683 spinlock_t asts_spin;
684 struct list_head locks;
685 spinlock_t locks_spin;
686 struct list_head unlocking;
687 wait_queue_head_t wait;
690 static inline int dlm_locking_stopped(struct dlm_ls *ls)
692 return !test_bit(LSFL_RUNNING, &ls->ls_flags);
695 static inline int dlm_recovery_stopped(struct dlm_ls *ls)
697 return test_bit(LSFL_RECOVER_STOP, &ls->ls_flags);
700 static inline int dlm_no_directory(struct dlm_ls *ls)
702 return test_bit(LSFL_NODIR, &ls->ls_flags);
705 int dlm_netlink_init(void);
706 void dlm_netlink_exit(void);
707 void dlm_timeout_warn(struct dlm_lkb *lkb);
708 int dlm_plock_init(void);
709 void dlm_plock_exit(void);
711 #ifdef CONFIG_DLM_DEBUG
712 int dlm_register_debugfs(void);
713 void dlm_unregister_debugfs(void);
714 int dlm_create_debug_file(struct dlm_ls *ls);
715 void dlm_delete_debug_file(struct dlm_ls *ls);
717 static inline int dlm_register_debugfs(void) { return 0; }
718 static inline void dlm_unregister_debugfs(void) { }
719 static inline int dlm_create_debug_file(struct dlm_ls *ls) { return 0; }
720 static inline void dlm_delete_debug_file(struct dlm_ls *ls) { }
723 #endif /* __DLM_INTERNAL_DOT_H__ */