Loading...
1/* SPDX-License-Identifier: GPL-2.0-only */
2/******************************************************************************
3*******************************************************************************
4**
5** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
6** Copyright (C) 2004-2011 Red Hat, Inc. All rights reserved.
7**
8**
9*******************************************************************************
10******************************************************************************/
11
12#ifndef __DLM_INTERNAL_DOT_H__
13#define __DLM_INTERNAL_DOT_H__
14
15/*
16 * This is the main header file to be included in each DLM source file.
17 */
18
19#include <linux/slab.h>
20#include <linux/sched.h>
21#include <linux/types.h>
22#include <linux/ctype.h>
23#include <linux/spinlock.h>
24#include <linux/vmalloc.h>
25#include <linux/list.h>
26#include <linux/errno.h>
27#include <linux/random.h>
28#include <linux/delay.h>
29#include <linux/socket.h>
30#include <linux/kthread.h>
31#include <linux/kobject.h>
32#include <linux/kref.h>
33#include <linux/kernel.h>
34#include <linux/jhash.h>
35#include <linux/miscdevice.h>
36#include <linux/mutex.h>
37#include <linux/idr.h>
38#include <linux/ratelimit.h>
39#include <linux/uaccess.h>
40
41#include <linux/dlm.h>
42#include "config.h"
43
44struct dlm_ls;
45struct dlm_lkb;
46struct dlm_rsb;
47struct dlm_member;
48struct dlm_rsbtable;
49struct dlm_recover;
50struct dlm_header;
51struct dlm_message;
52struct dlm_rcom;
53struct dlm_mhandle;
54struct dlm_msg;
55
56#define log_print(fmt, args...) \
57 printk(KERN_ERR "dlm: "fmt"\n" , ##args)
58#define log_print_ratelimited(fmt, args...) \
59 printk_ratelimited(KERN_ERR "dlm: "fmt"\n", ##args)
60#define log_error(ls, fmt, args...) \
61 printk(KERN_ERR "dlm: %s: " fmt "\n", (ls)->ls_name , ##args)
62
63#define log_rinfo(ls, fmt, args...) \
64do { \
65 if (dlm_config.ci_log_info) \
66 printk(KERN_INFO "dlm: %s: " fmt "\n", \
67 (ls)->ls_name, ##args); \
68 else if (dlm_config.ci_log_debug) \
69 printk(KERN_DEBUG "dlm: %s: " fmt "\n", \
70 (ls)->ls_name , ##args); \
71} while (0)
72
73#define log_debug(ls, fmt, args...) \
74do { \
75 if (dlm_config.ci_log_debug) \
76 printk(KERN_DEBUG "dlm: %s: " fmt "\n", \
77 (ls)->ls_name , ##args); \
78} while (0)
79
80#define log_limit(ls, fmt, args...) \
81do { \
82 if (dlm_config.ci_log_debug) \
83 printk_ratelimited(KERN_DEBUG "dlm: %s: " fmt "\n", \
84 (ls)->ls_name , ##args); \
85} while (0)
86
87#define DLM_ASSERT(x, do) \
88{ \
89 if (!(x)) \
90 { \
91 printk(KERN_ERR "\nDLM: Assertion failed on line %d of file %s\n" \
92 "DLM: assertion: \"%s\"\n" \
93 "DLM: time = %lu\n", \
94 __LINE__, __FILE__, #x, jiffies); \
95 {do} \
96 printk("\n"); \
97 panic("DLM: Record message above and reboot.\n"); \
98 } \
99}
100
101
102#define DLM_RTF_SHRINK_BIT 0
103
104struct dlm_rsbtable {
105 struct rb_root keep;
106 struct rb_root toss;
107 spinlock_t lock;
108 unsigned long flags;
109};
110
111
112/*
113 * Lockspace member (per node in a ls)
114 */
115
116struct dlm_member {
117 struct list_head list;
118 int nodeid;
119 int weight;
120 int slot;
121 int slot_prev;
122 int comm_seq;
123 uint32_t generation;
124};
125
126/*
127 * Save and manage recovery state for a lockspace.
128 */
129
130struct dlm_recover {
131 struct list_head list;
132 struct dlm_config_node *nodes;
133 int nodes_count;
134 uint64_t seq;
135};
136
137/*
138 * Pass input args to second stage locking function.
139 */
140
141struct dlm_args {
142 uint32_t flags;
143 void (*astfn) (void *astparam);
144 void *astparam;
145 void (*bastfn) (void *astparam, int mode);
146 int mode;
147 struct dlm_lksb *lksb;
148};
149
150
151/*
152 * Lock block
153 *
154 * A lock can be one of three types:
155 *
156 * local copy lock is mastered locally
157 * (lkb_nodeid is zero and DLM_LKF_MSTCPY is not set)
158 * process copy lock is mastered on a remote node
159 * (lkb_nodeid is non-zero and DLM_LKF_MSTCPY is not set)
160 * master copy master node's copy of a lock owned by remote node
161 * (lkb_nodeid is non-zero and DLM_LKF_MSTCPY is set)
162 *
163 * lkb_exflags: a copy of the most recent flags arg provided to dlm_lock or
164 * dlm_unlock. The dlm does not modify these or use any private flags in
165 * this field; it only contains DLM_LKF_ flags from dlm.h. These flags
166 * are sent as-is to the remote master when the lock is remote.
167 *
168 * lkb_flags: internal dlm flags (DLM_IFL_ prefix) from dlm_internal.h.
169 * Some internal flags are shared between the master and process nodes;
170 * these shared flags are kept in the lower two bytes. One of these
171 * flags set on the master copy will be propagated to the process copy
172 * and v.v. Other internal flags are private to the master or process
173 * node (e.g. DLM_IFL_MSTCPY). These are kept in the high two bytes.
174 *
175 * lkb_sbflags: status block flags. These flags are copied directly into
176 * the caller's lksb.sb_flags prior to the dlm_lock/dlm_unlock completion
177 * ast. All defined in dlm.h with DLM_SBF_ prefix.
178 *
179 * lkb_status: the lock status indicates which rsb queue the lock is
180 * on, grant, convert, or wait. DLM_LKSTS_ WAITING/GRANTED/CONVERT
181 *
182 * lkb_wait_type: the dlm message type (DLM_MSG_ prefix) for which a
183 * reply is needed. Only set when the lkb is on the lockspace waiters
184 * list awaiting a reply from a remote node.
185 *
186 * lkb_nodeid: when the lkb is a local copy, nodeid is 0; when the lkb
187 * is a master copy, nodeid specifies the remote lock holder, when the
188 * lkb is a process copy, the nodeid specifies the lock master.
189 */
190
191/* lkb_status */
192
193#define DLM_LKSTS_WAITING 1
194#define DLM_LKSTS_GRANTED 2
195#define DLM_LKSTS_CONVERT 3
196
197/* lkb_iflags */
198
199#define DLM_IFL_MSTCPY_BIT 16
200#define __DLM_IFL_MIN_BIT DLM_IFL_MSTCPY_BIT
201#define DLM_IFL_RESEND_BIT 17
202#define DLM_IFL_DEAD_BIT 18
203#define DLM_IFL_OVERLAP_UNLOCK_BIT 19
204#define DLM_IFL_OVERLAP_CANCEL_BIT 20
205#define DLM_IFL_ENDOFLIFE_BIT 21
206#define DLM_IFL_DEADLOCK_CANCEL_BIT 24
207#define DLM_IFL_CB_PENDING_BIT 25
208#define __DLM_IFL_MAX_BIT DLM_IFL_CB_PENDING_BIT
209
210/* lkb_dflags */
211
212#define DLM_DFL_USER_BIT 0
213#define __DLM_DFL_MIN_BIT DLM_DFL_USER_BIT
214#define DLM_DFL_ORPHAN_BIT 1
215#define __DLM_DFL_MAX_BIT DLM_DFL_ORPHAN_BIT
216
217#define DLM_CB_CAST 0x00000001
218#define DLM_CB_BAST 0x00000002
219
220struct dlm_callback {
221 uint32_t flags; /* DLM_CBF_ */
222 int sb_status; /* copy to lksb status */
223 uint8_t sb_flags; /* copy to lksb flags */
224 int8_t mode; /* rq mode of bast, gr mode of cast */
225 int copy_lvb;
226
227 struct list_head list;
228 struct kref ref;
229};
230
231struct dlm_lkb {
232 struct dlm_rsb *lkb_resource; /* the rsb */
233 struct kref lkb_ref;
234 int lkb_nodeid; /* copied from rsb */
235 int lkb_ownpid; /* pid of lock owner */
236 uint32_t lkb_id; /* our lock ID */
237 uint32_t lkb_remid; /* lock ID on remote partner */
238 uint32_t lkb_exflags; /* external flags from caller */
239 unsigned long lkb_sbflags; /* lksb flags */
240 unsigned long lkb_dflags; /* distributed flags */
241 unsigned long lkb_iflags; /* internal flags */
242 uint32_t lkb_lvbseq; /* lvb sequence number */
243
244 int8_t lkb_status; /* granted, waiting, convert */
245 int8_t lkb_rqmode; /* requested lock mode */
246 int8_t lkb_grmode; /* granted lock mode */
247 int8_t lkb_highbast; /* highest mode bast sent for */
248
249 int8_t lkb_wait_type; /* type of reply waiting for */
250 int8_t lkb_wait_count;
251 int lkb_wait_nodeid; /* for debugging */
252
253 struct list_head lkb_statequeue; /* rsb g/c/w list */
254 struct list_head lkb_rsb_lookup; /* waiting for rsb lookup */
255 struct list_head lkb_wait_reply; /* waiting for remote reply */
256 struct list_head lkb_ownqueue; /* list of locks for a process */
257 ktime_t lkb_timestamp;
258
259 spinlock_t lkb_cb_lock;
260 struct work_struct lkb_cb_work;
261 struct list_head lkb_cb_list; /* for ls_cb_delay or proc->asts */
262 struct list_head lkb_callbacks;
263 struct dlm_callback *lkb_last_cast;
264 struct dlm_callback *lkb_last_cb;
265 int lkb_last_bast_mode;
266 ktime_t lkb_last_cast_time; /* for debugging */
267 ktime_t lkb_last_bast_time; /* for debugging */
268
269 uint64_t lkb_recover_seq; /* from ls_recover_seq */
270
271 char *lkb_lvbptr;
272 struct dlm_lksb *lkb_lksb; /* caller's status block */
273 void (*lkb_astfn) (void *astparam);
274 void (*lkb_bastfn) (void *astparam, int mode);
275 union {
276 void *lkb_astparam; /* caller's ast arg */
277 struct dlm_user_args *lkb_ua;
278 };
279};
280
281/*
282 * res_master_nodeid is "normal": 0 is unset/invalid, non-zero is the real
283 * nodeid, even when nodeid is our_nodeid.
284 *
285 * res_nodeid is "odd": -1 is unset/invalid, zero means our_nodeid,
286 * greater than zero when another nodeid.
287 *
288 * (TODO: remove res_nodeid and only use res_master_nodeid)
289 */
290
291struct dlm_rsb {
292 struct dlm_ls *res_ls; /* the lockspace */
293 struct kref res_ref;
294 struct mutex res_mutex;
295 unsigned long res_flags;
296 int res_length; /* length of rsb name */
297 int res_nodeid;
298 int res_master_nodeid;
299 int res_dir_nodeid;
300 int res_id; /* for ls_recover_idr */
301 uint32_t res_lvbseq;
302 uint32_t res_hash;
303 uint32_t res_bucket; /* rsbtbl */
304 unsigned long res_toss_time;
305 uint32_t res_first_lkid;
306 struct list_head res_lookup; /* lkbs waiting on first */
307 union {
308 struct list_head res_hashchain;
309 struct rb_node res_hashnode; /* rsbtbl */
310 };
311 struct list_head res_grantqueue;
312 struct list_head res_convertqueue;
313 struct list_head res_waitqueue;
314
315 struct list_head res_root_list; /* used for recovery */
316 struct list_head res_recover_list; /* used for recovery */
317 int res_recover_locks_count;
318
319 char *res_lvbptr;
320 char res_name[DLM_RESNAME_MAXLEN+1];
321};
322
323/* dlm_master_lookup() flags */
324
325#define DLM_LU_RECOVER_DIR 1
326#define DLM_LU_RECOVER_MASTER 2
327
328/* dlm_master_lookup() results */
329
330#define DLM_LU_MATCH 1
331#define DLM_LU_ADD 2
332
333/* find_rsb() flags */
334
335#define R_REQUEST 0x00000001
336#define R_RECEIVE_REQUEST 0x00000002
337#define R_RECEIVE_RECOVER 0x00000004
338
339/* rsb_flags */
340
341enum rsb_flags {
342 RSB_MASTER_UNCERTAIN,
343 RSB_VALNOTVALID,
344 RSB_VALNOTVALID_PREV,
345 RSB_NEW_MASTER,
346 RSB_NEW_MASTER2,
347 RSB_RECOVER_CONVERT,
348 RSB_RECOVER_GRANT,
349 RSB_RECOVER_LVB_INVAL,
350};
351
352static inline void rsb_set_flag(struct dlm_rsb *r, enum rsb_flags flag)
353{
354 __set_bit(flag, &r->res_flags);
355}
356
357static inline void rsb_clear_flag(struct dlm_rsb *r, enum rsb_flags flag)
358{
359 __clear_bit(flag, &r->res_flags);
360}
361
362static inline int rsb_flag(struct dlm_rsb *r, enum rsb_flags flag)
363{
364 return test_bit(flag, &r->res_flags);
365}
366
367
368/* dlm_header is first element of all structs sent between nodes */
369
370#define DLM_HEADER_MAJOR 0x00030000
371#define DLM_HEADER_MINOR 0x00000002
372
373#define DLM_VERSION_3_1 0x00030001
374#define DLM_VERSION_3_2 0x00030002
375
376#define DLM_HEADER_SLOTS 0x00000001
377
378#define DLM_MSG 1
379#define DLM_RCOM 2
380#define DLM_OPTS 3
381#define DLM_ACK 4
382#define DLM_FIN 5
383
384struct dlm_header {
385 __le32 h_version;
386 union {
387 /* for DLM_MSG and DLM_RCOM */
388 __le32 h_lockspace;
389 /* for DLM_ACK and DLM_OPTS */
390 __le32 h_seq;
391 } u;
392 __le32 h_nodeid; /* nodeid of sender */
393 __le16 h_length;
394 uint8_t h_cmd; /* DLM_MSG, DLM_RCOM */
395 uint8_t h_pad;
396};
397
398#define DLM_MSG_REQUEST 1
399#define DLM_MSG_CONVERT 2
400#define DLM_MSG_UNLOCK 3
401#define DLM_MSG_CANCEL 4
402#define DLM_MSG_REQUEST_REPLY 5
403#define DLM_MSG_CONVERT_REPLY 6
404#define DLM_MSG_UNLOCK_REPLY 7
405#define DLM_MSG_CANCEL_REPLY 8
406#define DLM_MSG_GRANT 9
407#define DLM_MSG_BAST 10
408#define DLM_MSG_LOOKUP 11
409#define DLM_MSG_REMOVE 12
410#define DLM_MSG_LOOKUP_REPLY 13
411#define DLM_MSG_PURGE 14
412
413struct dlm_message {
414 struct dlm_header m_header;
415 __le32 m_type; /* DLM_MSG_ */
416 __le32 m_nodeid;
417 __le32 m_pid;
418 __le32 m_lkid; /* lkid on sender */
419 __le32 m_remid; /* lkid on receiver */
420 __le32 m_parent_lkid;
421 __le32 m_parent_remid;
422 __le32 m_exflags;
423 __le32 m_sbflags;
424 __le32 m_flags;
425 __le32 m_lvbseq;
426 __le32 m_hash;
427 __le32 m_status;
428 __le32 m_grmode;
429 __le32 m_rqmode;
430 __le32 m_bastmode;
431 __le32 m_asts;
432 __le32 m_result; /* 0 or -EXXX */
433 char m_extra[]; /* name or lvb */
434};
435
436
437#define DLM_RS_NODES 0x00000001
438#define DLM_RS_NODES_ALL 0x00000002
439#define DLM_RS_DIR 0x00000004
440#define DLM_RS_DIR_ALL 0x00000008
441#define DLM_RS_LOCKS 0x00000010
442#define DLM_RS_LOCKS_ALL 0x00000020
443#define DLM_RS_DONE 0x00000040
444#define DLM_RS_DONE_ALL 0x00000080
445
446#define DLM_RCOM_STATUS 1
447#define DLM_RCOM_NAMES 2
448#define DLM_RCOM_LOOKUP 3
449#define DLM_RCOM_LOCK 4
450#define DLM_RCOM_STATUS_REPLY 5
451#define DLM_RCOM_NAMES_REPLY 6
452#define DLM_RCOM_LOOKUP_REPLY 7
453#define DLM_RCOM_LOCK_REPLY 8
454
455struct dlm_rcom {
456 struct dlm_header rc_header;
457 __le32 rc_type; /* DLM_RCOM_ */
458 __le32 rc_result; /* multi-purpose */
459 __le64 rc_id; /* match reply with request */
460 __le64 rc_seq; /* sender's ls_recover_seq */
461 __le64 rc_seq_reply; /* remote ls_recover_seq */
462 char rc_buf[];
463};
464
465struct dlm_opt_header {
466 __le16 t_type;
467 __le16 t_length;
468 __le32 t_pad;
469 /* need to be 8 byte aligned */
470 char t_value[];
471};
472
473/* encapsulation header */
474struct dlm_opts {
475 struct dlm_header o_header;
476 uint8_t o_nextcmd;
477 uint8_t o_pad;
478 __le16 o_optlen;
479 __le32 o_pad2;
480 char o_opts[];
481};
482
483union dlm_packet {
484 struct dlm_header header; /* common to other two */
485 struct dlm_message message;
486 struct dlm_rcom rcom;
487 struct dlm_opts opts;
488};
489
490#define DLM_RSF_NEED_SLOTS 0x00000001
491
492/* RCOM_STATUS data */
493struct rcom_status {
494 __le32 rs_flags;
495 __le32 rs_unused1;
496 __le64 rs_unused2;
497};
498
499/* RCOM_STATUS_REPLY data */
500struct rcom_config {
501 __le32 rf_lvblen;
502 __le32 rf_lsflags;
503
504 /* DLM_HEADER_SLOTS adds: */
505 __le32 rf_flags;
506 __le16 rf_our_slot;
507 __le16 rf_num_slots;
508 __le32 rf_generation;
509 __le32 rf_unused1;
510 __le64 rf_unused2;
511};
512
513struct rcom_slot {
514 __le32 ro_nodeid;
515 __le16 ro_slot;
516 __le16 ro_unused1;
517 __le64 ro_unused2;
518};
519
520struct rcom_lock {
521 __le32 rl_ownpid;
522 __le32 rl_lkid;
523 __le32 rl_remid;
524 __le32 rl_parent_lkid;
525 __le32 rl_parent_remid;
526 __le32 rl_exflags;
527 __le32 rl_flags;
528 __le32 rl_lvbseq;
529 __le32 rl_result;
530 int8_t rl_rqmode;
531 int8_t rl_grmode;
532 int8_t rl_status;
533 int8_t rl_asts;
534 __le16 rl_wait_type;
535 __le16 rl_namelen;
536 char rl_name[DLM_RESNAME_MAXLEN];
537 char rl_lvb[];
538};
539
540/*
541 * The max number of resources per rsbtbl bucket that shrink will attempt
542 * to remove in each iteration.
543 */
544
545#define DLM_REMOVE_NAMES_MAX 8
546
547struct dlm_ls {
548 struct list_head ls_list; /* list of lockspaces */
549 dlm_lockspace_t *ls_local_handle;
550 uint32_t ls_global_id; /* global unique lockspace ID */
551 uint32_t ls_generation;
552 uint32_t ls_exflags;
553 int ls_lvblen;
554 atomic_t ls_count; /* refcount of processes in
555 the dlm using this ls */
556 wait_queue_head_t ls_count_wait;
557 int ls_create_count; /* create/release refcount */
558 unsigned long ls_flags; /* LSFL_ */
559 unsigned long ls_scan_time;
560 struct kobject ls_kobj;
561
562 struct idr ls_lkbidr;
563 spinlock_t ls_lkbidr_spin;
564
565 struct dlm_rsbtable *ls_rsbtbl;
566 uint32_t ls_rsbtbl_size;
567
568 struct mutex ls_waiters_mutex;
569 struct list_head ls_waiters; /* lkbs needing a reply */
570
571 struct mutex ls_orphans_mutex;
572 struct list_head ls_orphans;
573
574 spinlock_t ls_new_rsb_spin;
575 int ls_new_rsb_count;
576 struct list_head ls_new_rsb; /* new rsb structs */
577
578 char *ls_remove_names[DLM_REMOVE_NAMES_MAX];
579 int ls_remove_lens[DLM_REMOVE_NAMES_MAX];
580
581 struct list_head ls_nodes; /* current nodes in ls */
582 struct list_head ls_nodes_gone; /* dead node list, recovery */
583 int ls_num_nodes; /* number of nodes in ls */
584 int ls_low_nodeid;
585 int ls_total_weight;
586 int *ls_node_array;
587
588 int ls_slot;
589 int ls_num_slots;
590 int ls_slots_size;
591 struct dlm_slot *ls_slots;
592
593 struct dlm_rsb ls_local_rsb; /* for returning errors */
594 struct dlm_lkb ls_local_lkb; /* for returning errors */
595 struct dlm_message ls_local_ms; /* for faking a reply */
596
597 struct dentry *ls_debug_rsb_dentry; /* debugfs */
598 struct dentry *ls_debug_waiters_dentry; /* debugfs */
599 struct dentry *ls_debug_locks_dentry; /* debugfs */
600 struct dentry *ls_debug_all_dentry; /* debugfs */
601 struct dentry *ls_debug_toss_dentry; /* debugfs */
602 struct dentry *ls_debug_queued_asts_dentry; /* debugfs */
603
604 wait_queue_head_t ls_uevent_wait; /* user part of join/leave */
605 int ls_uevent_result;
606 struct completion ls_recovery_done;
607 int ls_recovery_result;
608
609 struct miscdevice ls_device;
610
611 struct workqueue_struct *ls_callback_wq;
612
613 /* recovery related */
614
615 spinlock_t ls_cb_lock;
616 struct list_head ls_cb_delay; /* save for queue_work later */
617 struct timer_list ls_timer;
618 struct task_struct *ls_recoverd_task;
619 struct mutex ls_recoverd_active;
620 spinlock_t ls_recover_lock;
621 unsigned long ls_recover_begin; /* jiffies timestamp */
622 uint32_t ls_recover_status; /* DLM_RS_ */
623 uint64_t ls_recover_seq;
624 struct dlm_recover *ls_recover_args;
625 struct rw_semaphore ls_in_recovery; /* block local requests */
626 struct rw_semaphore ls_recv_active; /* block dlm_recv */
627 struct list_head ls_requestqueue;/* queue remote requests */
628 atomic_t ls_requestqueue_cnt;
629 wait_queue_head_t ls_requestqueue_wait;
630 struct mutex ls_requestqueue_mutex;
631 struct dlm_rcom *ls_recover_buf;
632 int ls_recover_nodeid; /* for debugging */
633 unsigned int ls_recover_dir_sent_res; /* for log info */
634 unsigned int ls_recover_dir_sent_msg; /* for log info */
635 unsigned int ls_recover_locks_in; /* for log info */
636 uint64_t ls_rcom_seq;
637 spinlock_t ls_rcom_spin;
638 struct list_head ls_recover_list;
639 spinlock_t ls_recover_list_lock;
640 int ls_recover_list_count;
641 struct idr ls_recover_idr;
642 spinlock_t ls_recover_idr_lock;
643 wait_queue_head_t ls_wait_general;
644 wait_queue_head_t ls_recover_lock_wait;
645 spinlock_t ls_clear_proc_locks;
646
647 struct list_head ls_root_list; /* root resources */
648 struct rw_semaphore ls_root_sem; /* protect root_list */
649
650 const struct dlm_lockspace_ops *ls_ops;
651 void *ls_ops_arg;
652
653 int ls_namelen;
654 char ls_name[DLM_LOCKSPACE_LEN + 1];
655};
656
657/*
658 * LSFL_RECOVER_STOP - dlm_ls_stop() sets this to tell dlm recovery routines
659 * that they should abort what they're doing so new recovery can be started.
660 *
661 * LSFL_RECOVER_DOWN - dlm_ls_stop() sets this to tell dlm_recoverd that it
662 * should do down_write() on the in_recovery rw_semaphore. (doing down_write
663 * within dlm_ls_stop causes complaints about the lock acquired/released
664 * in different contexts.)
665 *
666 * LSFL_RECOVER_LOCK - dlm_recoverd holds the in_recovery rw_semaphore.
667 * It sets this after it is done with down_write() on the in_recovery
668 * rw_semaphore and clears it after it has released the rw_semaphore.
669 *
670 * LSFL_RECOVER_WORK - dlm_ls_start() sets this to tell dlm_recoverd that it
671 * should begin recovery of the lockspace.
672 *
673 * LSFL_RUNNING - set when normal locking activity is enabled.
674 * dlm_ls_stop() clears this to tell dlm locking routines that they should
675 * quit what they are doing so recovery can run. dlm_recoverd sets
676 * this after recovery is finished.
677 */
678
679#define LSFL_RECOVER_STOP 0
680#define LSFL_RECOVER_DOWN 1
681#define LSFL_RECOVER_LOCK 2
682#define LSFL_RECOVER_WORK 3
683#define LSFL_RUNNING 4
684
685#define LSFL_RCOM_READY 5
686#define LSFL_RCOM_WAIT 6
687#define LSFL_UEVENT_WAIT 7
688#define LSFL_CB_DELAY 9
689#define LSFL_NODIR 10
690
691/* much of this is just saving user space pointers associated with the
692 lock that we pass back to the user lib with an ast */
693
694struct dlm_user_args {
695 struct dlm_user_proc *proc; /* each process that opens the lockspace
696 device has private data
697 (dlm_user_proc) on the struct file,
698 the process's locks point back to it*/
699 struct dlm_lksb lksb;
700 struct dlm_lksb __user *user_lksb;
701 void __user *castparam;
702 void __user *castaddr;
703 void __user *bastparam;
704 void __user *bastaddr;
705 uint64_t xid;
706};
707
708#define DLM_PROC_FLAGS_CLOSING 1
709#define DLM_PROC_FLAGS_COMPAT 2
710
711/* locks list is kept so we can remove all a process's locks when it
712 exits (or orphan those that are persistent) */
713
714struct dlm_user_proc {
715 dlm_lockspace_t *lockspace;
716 unsigned long flags; /* DLM_PROC_FLAGS */
717 struct list_head asts;
718 spinlock_t asts_spin;
719 struct list_head locks;
720 spinlock_t locks_spin;
721 struct list_head unlocking;
722 wait_queue_head_t wait;
723};
724
725static inline int dlm_locking_stopped(struct dlm_ls *ls)
726{
727 return !test_bit(LSFL_RUNNING, &ls->ls_flags);
728}
729
730static inline int dlm_recovery_stopped(struct dlm_ls *ls)
731{
732 return test_bit(LSFL_RECOVER_STOP, &ls->ls_flags);
733}
734
735static inline int dlm_no_directory(struct dlm_ls *ls)
736{
737 return test_bit(LSFL_NODIR, &ls->ls_flags);
738}
739
740/* takes a snapshot from dlm atomic flags */
741static inline uint32_t dlm_flags_val(const unsigned long *addr,
742 uint32_t min, uint32_t max)
743{
744 uint32_t bit = min, val = 0;
745
746 for_each_set_bit_from(bit, addr, max + 1) {
747 val |= BIT(bit);
748 }
749
750 return val;
751}
752
753static inline uint32_t dlm_iflags_val(const struct dlm_lkb *lkb)
754{
755 return dlm_flags_val(&lkb->lkb_iflags, __DLM_IFL_MIN_BIT,
756 __DLM_IFL_MAX_BIT);
757}
758
759static inline uint32_t dlm_dflags_val(const struct dlm_lkb *lkb)
760{
761 return dlm_flags_val(&lkb->lkb_dflags, __DLM_DFL_MIN_BIT,
762 __DLM_DFL_MAX_BIT);
763}
764
765/* coming from UAPI header
766 *
767 * TODO:
768 * Move this to UAPI header and let other values point to them and use BIT()
769 */
770#define DLM_SBF_DEMOTED_BIT 0
771#define __DLM_SBF_MIN_BIT DLM_SBF_DEMOTED_BIT
772#define DLM_SBF_VALNOTVALID_BIT 1
773#define DLM_SBF_ALTMODE_BIT 2
774#define __DLM_SBF_MAX_BIT DLM_SBF_ALTMODE_BIT
775
776static inline uint32_t dlm_sbflags_val(const struct dlm_lkb *lkb)
777{
778 /* be sure the next person updates this */
779 BUILD_BUG_ON(BIT(__DLM_SBF_MAX_BIT) != DLM_SBF_ALTMODE);
780
781 return dlm_flags_val(&lkb->lkb_sbflags, __DLM_SBF_MIN_BIT,
782 __DLM_SBF_MAX_BIT);
783}
784
785static inline void dlm_set_flags_val(unsigned long *addr, uint32_t val,
786 uint32_t min, uint32_t max)
787{
788 uint32_t bit;
789
790 for (bit = min; bit < (max + 1); bit++) {
791 if (val & BIT(bit))
792 set_bit(bit, addr);
793 else
794 clear_bit(bit, addr);
795 }
796}
797
798static inline void dlm_set_dflags_val(struct dlm_lkb *lkb, uint32_t val)
799{
800 dlm_set_flags_val(&lkb->lkb_dflags, val, __DLM_DFL_MIN_BIT,
801 __DLM_DFL_MAX_BIT);
802}
803
804static inline void dlm_set_sbflags_val(struct dlm_lkb *lkb, uint32_t val)
805{
806 dlm_set_flags_val(&lkb->lkb_sbflags, val, __DLM_SBF_MIN_BIT,
807 __DLM_SBF_MAX_BIT);
808}
809
810int dlm_plock_init(void);
811void dlm_plock_exit(void);
812
813#ifdef CONFIG_DLM_DEBUG
814void dlm_register_debugfs(void);
815void dlm_unregister_debugfs(void);
816void dlm_create_debug_file(struct dlm_ls *ls);
817void dlm_delete_debug_file(struct dlm_ls *ls);
818void *dlm_create_debug_comms_file(int nodeid, void *data);
819void dlm_delete_debug_comms_file(void *ctx);
820#else
821static inline void dlm_register_debugfs(void) { }
822static inline void dlm_unregister_debugfs(void) { }
823static inline void dlm_create_debug_file(struct dlm_ls *ls) { }
824static inline void dlm_delete_debug_file(struct dlm_ls *ls) { }
825static inline void *dlm_create_debug_comms_file(int nodeid, void *data) { return NULL; }
826static inline void dlm_delete_debug_comms_file(void *ctx) { }
827#endif
828
829#endif /* __DLM_INTERNAL_DOT_H__ */
830
1/* SPDX-License-Identifier: GPL-2.0-only */
2/******************************************************************************
3*******************************************************************************
4**
5** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
6** Copyright (C) 2004-2011 Red Hat, Inc. All rights reserved.
7**
8**
9*******************************************************************************
10******************************************************************************/
11
12#ifndef __DLM_INTERNAL_DOT_H__
13#define __DLM_INTERNAL_DOT_H__
14
15/*
16 * This is the main header file to be included in each DLM source file.
17 */
18
19#include <uapi/linux/dlm_device.h>
20#include <linux/slab.h>
21#include <linux/sched.h>
22#include <linux/types.h>
23#include <linux/ctype.h>
24#include <linux/spinlock.h>
25#include <linux/vmalloc.h>
26#include <linux/list.h>
27#include <linux/errno.h>
28#include <linux/random.h>
29#include <linux/delay.h>
30#include <linux/socket.h>
31#include <linux/kthread.h>
32#include <linux/kobject.h>
33#include <linux/kref.h>
34#include <linux/kernel.h>
35#include <linux/jhash.h>
36#include <linux/miscdevice.h>
37#include <linux/rhashtable.h>
38#include <linux/mutex.h>
39#include <linux/xarray.h>
40#include <linux/ratelimit.h>
41#include <linux/uaccess.h>
42
43#include <linux/dlm.h>
44#include "config.h"
45
46struct dlm_ls;
47struct dlm_lkb;
48struct dlm_rsb;
49struct dlm_member;
50struct dlm_rsbtable;
51struct dlm_recover;
52struct dlm_header;
53struct dlm_message;
54struct dlm_rcom;
55struct dlm_mhandle;
56struct dlm_msg;
57
58#define log_print(fmt, args...) \
59 printk(KERN_ERR "dlm: "fmt"\n" , ##args)
60#define log_print_ratelimited(fmt, args...) \
61 printk_ratelimited(KERN_ERR "dlm: "fmt"\n", ##args)
62#define log_error(ls, fmt, args...) \
63 printk(KERN_ERR "dlm: %s: " fmt "\n", (ls)->ls_name , ##args)
64
65#define log_rinfo(ls, fmt, args...) \
66do { \
67 if (dlm_config.ci_log_info) \
68 printk(KERN_INFO "dlm: %s: " fmt "\n", \
69 (ls)->ls_name, ##args); \
70 else if (dlm_config.ci_log_debug) \
71 printk(KERN_DEBUG "dlm: %s: " fmt "\n", \
72 (ls)->ls_name , ##args); \
73} while (0)
74
75#define log_debug(ls, fmt, args...) \
76do { \
77 if (dlm_config.ci_log_debug) \
78 printk(KERN_DEBUG "dlm: %s: " fmt "\n", \
79 (ls)->ls_name , ##args); \
80} while (0)
81
82#define log_limit(ls, fmt, args...) \
83do { \
84 if (dlm_config.ci_log_debug) \
85 printk_ratelimited(KERN_DEBUG "dlm: %s: " fmt "\n", \
86 (ls)->ls_name , ##args); \
87} while (0)
88
89#define DLM_ASSERT(x, do) \
90{ \
91 if (!(x)) \
92 { \
93 printk(KERN_ERR "\nDLM: Assertion failed on line %d of file %s\n" \
94 "DLM: assertion: \"%s\"\n" \
95 "DLM: time = %lu\n", \
96 __LINE__, __FILE__, #x, jiffies); \
97 {do} \
98 printk("\n"); \
99 panic("DLM: Record message above and reboot.\n"); \
100 } \
101}
102
103/*
104 * Lockspace member (per node in a ls)
105 */
106
107struct dlm_member {
108 struct list_head list;
109 int nodeid;
110 int weight;
111 int slot;
112 int slot_prev;
113 int comm_seq;
114 uint32_t generation;
115};
116
117/*
118 * Save and manage recovery state for a lockspace.
119 */
120
121struct dlm_recover {
122 struct list_head list;
123 struct dlm_config_node *nodes;
124 int nodes_count;
125 uint64_t seq;
126};
127
128/*
129 * Pass input args to second stage locking function.
130 */
131
132struct dlm_args {
133 uint32_t flags;
134 void (*astfn) (void *astparam);
135 void *astparam;
136 void (*bastfn) (void *astparam, int mode);
137 int mode;
138 struct dlm_lksb *lksb;
139};
140
141
142/*
143 * Lock block
144 *
145 * A lock can be one of three types:
146 *
147 * local copy lock is mastered locally
148 * (lkb_nodeid is zero and DLM_LKF_MSTCPY is not set)
149 * process copy lock is mastered on a remote node
150 * (lkb_nodeid is non-zero and DLM_LKF_MSTCPY is not set)
151 * master copy master node's copy of a lock owned by remote node
152 * (lkb_nodeid is non-zero and DLM_LKF_MSTCPY is set)
153 *
154 * lkb_exflags: a copy of the most recent flags arg provided to dlm_lock or
155 * dlm_unlock. The dlm does not modify these or use any private flags in
156 * this field; it only contains DLM_LKF_ flags from dlm.h. These flags
157 * are sent as-is to the remote master when the lock is remote.
158 *
159 * lkb_flags: internal dlm flags (DLM_IFL_ prefix) from dlm_internal.h.
160 * Some internal flags are shared between the master and process nodes;
161 * these shared flags are kept in the lower two bytes. One of these
162 * flags set on the master copy will be propagated to the process copy
163 * and v.v. Other internal flags are private to the master or process
164 * node (e.g. DLM_IFL_MSTCPY). These are kept in the high two bytes.
165 *
166 * lkb_sbflags: status block flags. These flags are copied directly into
167 * the caller's lksb.sb_flags prior to the dlm_lock/dlm_unlock completion
168 * ast. All defined in dlm.h with DLM_SBF_ prefix.
169 *
170 * lkb_status: the lock status indicates which rsb queue the lock is
171 * on, grant, convert, or wait. DLM_LKSTS_ WAITING/GRANTED/CONVERT
172 *
173 * lkb_wait_type: the dlm message type (DLM_MSG_ prefix) for which a
174 * reply is needed. Only set when the lkb is on the lockspace waiters
175 * list awaiting a reply from a remote node.
176 *
177 * lkb_nodeid: when the lkb is a local copy, nodeid is 0; when the lkb
178 * is a master copy, nodeid specifies the remote lock holder, when the
179 * lkb is a process copy, the nodeid specifies the lock master.
180 */
181
182/* lkb_status */
183
184#define DLM_LKSTS_WAITING 1
185#define DLM_LKSTS_GRANTED 2
186#define DLM_LKSTS_CONVERT 3
187
188/* lkb_iflags */
189
190#define DLM_IFL_MSTCPY_BIT 16
191#define __DLM_IFL_MIN_BIT DLM_IFL_MSTCPY_BIT
192#define DLM_IFL_RESEND_BIT 17
193#define DLM_IFL_DEAD_BIT 18
194#define DLM_IFL_OVERLAP_UNLOCK_BIT 19
195#define DLM_IFL_OVERLAP_CANCEL_BIT 20
196#define DLM_IFL_ENDOFLIFE_BIT 21
197#define DLM_IFL_DEADLOCK_CANCEL_BIT 24
198#define __DLM_IFL_MAX_BIT DLM_IFL_DEADLOCK_CANCEL_BIT
199
200/* lkb_dflags */
201
202#define DLM_DFL_USER_BIT 0
203#define __DLM_DFL_MIN_BIT DLM_DFL_USER_BIT
204#define DLM_DFL_ORPHAN_BIT 1
205#define __DLM_DFL_MAX_BIT DLM_DFL_ORPHAN_BIT
206
207#define DLM_CB_CAST 0x00000001
208#define DLM_CB_BAST 0x00000002
209
210/* much of this is just saving user space pointers associated with the
211 * lock that we pass back to the user lib with an ast
212 */
213
214struct dlm_user_args {
215 struct dlm_user_proc *proc; /* each process that opens the lockspace
216 * device has private data
217 * (dlm_user_proc) on the struct file,
218 * the process's locks point back to it
219 */
220 struct dlm_lksb lksb;
221 struct dlm_lksb __user *user_lksb;
222 void __user *castparam;
223 void __user *castaddr;
224 void __user *bastparam;
225 void __user *bastaddr;
226 uint64_t xid;
227};
228
229struct dlm_callback {
230 uint32_t flags; /* DLM_CBF_ */
231 int sb_status; /* copy to lksb status */
232 uint8_t sb_flags; /* copy to lksb flags */
233 int8_t mode; /* rq mode of bast, gr mode of cast */
234 bool copy_lvb;
235 struct dlm_lksb *lkb_lksb;
236 unsigned char lvbptr[DLM_USER_LVB_LEN];
237
238 union {
239 void *astparam; /* caller's ast arg */
240 struct dlm_user_args ua;
241 };
242 struct work_struct work;
243 void (*bastfn)(void *astparam, int mode);
244 void (*astfn)(void *astparam);
245 char res_name[DLM_RESNAME_MAXLEN];
246 size_t res_length;
247 uint32_t ls_id;
248 uint32_t lkb_id;
249
250 struct list_head list;
251};
252
253struct dlm_lkb {
254 struct dlm_rsb *lkb_resource; /* the rsb */
255 struct kref lkb_ref;
256 int lkb_nodeid; /* copied from rsb */
257 int lkb_ownpid; /* pid of lock owner */
258 uint32_t lkb_id; /* our lock ID */
259 uint32_t lkb_remid; /* lock ID on remote partner */
260 uint32_t lkb_exflags; /* external flags from caller */
261 unsigned long lkb_sbflags; /* lksb flags */
262 unsigned long lkb_dflags; /* distributed flags */
263 unsigned long lkb_iflags; /* internal flags */
264 uint32_t lkb_lvbseq; /* lvb sequence number */
265
266 int8_t lkb_status; /* granted, waiting, convert */
267 int8_t lkb_rqmode; /* requested lock mode */
268 int8_t lkb_grmode; /* granted lock mode */
269 int8_t lkb_highbast; /* highest mode bast sent for */
270
271 int8_t lkb_wait_type; /* type of reply waiting for */
272 int8_t lkb_wait_count;
273 int lkb_wait_nodeid; /* for debugging */
274
275 struct list_head lkb_statequeue; /* rsb g/c/w list */
276 struct list_head lkb_rsb_lookup; /* waiting for rsb lookup */
277 struct list_head lkb_wait_reply; /* waiting for remote reply */
278 struct list_head lkb_ownqueue; /* list of locks for a process */
279 ktime_t lkb_timestamp;
280
281 int8_t lkb_last_cast_cb_mode;
282 int8_t lkb_last_bast_cb_mode;
283 int8_t lkb_last_cb_mode;
284 uint8_t lkb_last_cb_flags;
285 ktime_t lkb_last_cast_time; /* for debugging */
286 ktime_t lkb_last_bast_time; /* for debugging */
287
288 uint64_t lkb_recover_seq; /* from ls_recover_seq */
289
290 char *lkb_lvbptr;
291 struct dlm_lksb *lkb_lksb; /* caller's status block */
292 void (*lkb_astfn) (void *astparam);
293 void (*lkb_bastfn) (void *astparam, int mode);
294 union {
295 void *lkb_astparam; /* caller's ast arg */
296 struct dlm_user_args *lkb_ua;
297 };
298 struct rcu_head rcu;
299};
300
301/*
302 * res_master_nodeid is "normal": 0 is unset/invalid, non-zero is the real
303 * nodeid, even when nodeid is our_nodeid.
304 *
305 * res_nodeid is "odd": -1 is unset/invalid, zero means our_nodeid,
306 * greater than zero when another nodeid.
307 *
308 * (TODO: remove res_nodeid and only use res_master_nodeid)
309 */
310
311struct dlm_rsb {
312 struct dlm_ls *res_ls; /* the lockspace */
313 struct kref res_ref;
314 spinlock_t res_lock;
315 unsigned long res_flags;
316 int res_length; /* length of rsb name */
317 int res_nodeid;
318 int res_master_nodeid;
319 int res_dir_nodeid;
320 unsigned long res_id; /* for ls_recover_xa */
321 uint32_t res_lvbseq;
322 uint32_t res_hash;
323 unsigned long res_toss_time;
324 uint32_t res_first_lkid;
325 struct list_head res_lookup; /* lkbs waiting on first */
326 struct rhash_head res_node; /* rsbtbl */
327 struct list_head res_grantqueue;
328 struct list_head res_convertqueue;
329 struct list_head res_waitqueue;
330
331 struct list_head res_slow_list; /* ls_slow_* */
332 struct list_head res_scan_list;
333 struct list_head res_root_list; /* used for recovery */
334 struct list_head res_masters_list; /* used for recovery */
335 struct list_head res_recover_list; /* used for recovery */
336 int res_recover_locks_count;
337 struct rcu_head rcu;
338
339 char *res_lvbptr;
340 char res_name[DLM_RESNAME_MAXLEN+1];
341};
342
343/* dlm_master_lookup() flags */
344
345#define DLM_LU_RECOVER_DIR 1
346#define DLM_LU_RECOVER_MASTER 2
347
348/* dlm_master_lookup() results */
349
350#define DLM_LU_MATCH 1
351#define DLM_LU_ADD 2
352
353/* find_rsb() flags */
354
355#define R_REQUEST 0x00000001
356#define R_RECEIVE_REQUEST 0x00000002
357#define R_RECEIVE_RECOVER 0x00000004
358
359/* rsb_flags */
360
361enum rsb_flags {
362 RSB_MASTER_UNCERTAIN,
363 RSB_VALNOTVALID,
364 RSB_VALNOTVALID_PREV,
365 RSB_NEW_MASTER,
366 RSB_NEW_MASTER2,
367 RSB_RECOVER_CONVERT,
368 RSB_RECOVER_GRANT,
369 RSB_RECOVER_LVB_INVAL,
370 RSB_INACTIVE,
371 RSB_HASHED, /* set while rsb is on ls_rsbtbl */
372};
373
374static inline void rsb_set_flag(struct dlm_rsb *r, enum rsb_flags flag)
375{
376 __set_bit(flag, &r->res_flags);
377}
378
379static inline void rsb_clear_flag(struct dlm_rsb *r, enum rsb_flags flag)
380{
381 __clear_bit(flag, &r->res_flags);
382}
383
384static inline int rsb_flag(struct dlm_rsb *r, enum rsb_flags flag)
385{
386 return test_bit(flag, &r->res_flags);
387}
388
389
390/* dlm_header is first element of all structs sent between nodes */
391
392#define DLM_HEADER_MAJOR 0x00030000
393#define DLM_HEADER_MINOR 0x00000002
394
395#define DLM_VERSION_3_1 0x00030001
396#define DLM_VERSION_3_2 0x00030002
397
398#define DLM_HEADER_SLOTS 0x00000001
399
400#define DLM_MSG 1
401#define DLM_RCOM 2
402#define DLM_OPTS 3
403#define DLM_ACK 4
404#define DLM_FIN 5
405
406struct dlm_header {
407 __le32 h_version;
408 union {
409 /* for DLM_MSG and DLM_RCOM */
410 __le32 h_lockspace;
411 /* for DLM_ACK and DLM_OPTS */
412 __le32 h_seq;
413 } u;
414 __le32 h_nodeid; /* nodeid of sender */
415 __le16 h_length;
416 uint8_t h_cmd; /* DLM_MSG, DLM_RCOM */
417 uint8_t h_pad;
418};
419
420#define DLM_MSG_REQUEST 1
421#define DLM_MSG_CONVERT 2
422#define DLM_MSG_UNLOCK 3
423#define DLM_MSG_CANCEL 4
424#define DLM_MSG_REQUEST_REPLY 5
425#define DLM_MSG_CONVERT_REPLY 6
426#define DLM_MSG_UNLOCK_REPLY 7
427#define DLM_MSG_CANCEL_REPLY 8
428#define DLM_MSG_GRANT 9
429#define DLM_MSG_BAST 10
430#define DLM_MSG_LOOKUP 11
431#define DLM_MSG_REMOVE 12
432#define DLM_MSG_LOOKUP_REPLY 13
433#define DLM_MSG_PURGE 14
434
435struct dlm_message {
436 struct dlm_header m_header;
437 __le32 m_type; /* DLM_MSG_ */
438 __le32 m_nodeid;
439 __le32 m_pid;
440 __le32 m_lkid; /* lkid on sender */
441 __le32 m_remid; /* lkid on receiver */
442 __le32 m_parent_lkid;
443 __le32 m_parent_remid;
444 __le32 m_exflags;
445 __le32 m_sbflags;
446 __le32 m_flags;
447 __le32 m_lvbseq;
448 __le32 m_hash;
449 __le32 m_status;
450 __le32 m_grmode;
451 __le32 m_rqmode;
452 __le32 m_bastmode;
453 __le32 m_asts;
454 __le32 m_result; /* 0 or -EXXX */
455 char m_extra[]; /* name or lvb */
456};
457
458
459#define DLM_RS_NODES 0x00000001
460#define DLM_RS_NODES_ALL 0x00000002
461#define DLM_RS_DIR 0x00000004
462#define DLM_RS_DIR_ALL 0x00000008
463#define DLM_RS_LOCKS 0x00000010
464#define DLM_RS_LOCKS_ALL 0x00000020
465#define DLM_RS_DONE 0x00000040
466#define DLM_RS_DONE_ALL 0x00000080
467
468#define DLM_RCOM_STATUS 1
469#define DLM_RCOM_NAMES 2
470#define DLM_RCOM_LOOKUP 3
471#define DLM_RCOM_LOCK 4
472#define DLM_RCOM_STATUS_REPLY 5
473#define DLM_RCOM_NAMES_REPLY 6
474#define DLM_RCOM_LOOKUP_REPLY 7
475#define DLM_RCOM_LOCK_REPLY 8
476
477struct dlm_rcom {
478 struct dlm_header rc_header;
479 __le32 rc_type; /* DLM_RCOM_ */
480 __le32 rc_result; /* multi-purpose */
481 __le64 rc_id; /* match reply with request */
482 __le64 rc_seq; /* sender's ls_recover_seq */
483 __le64 rc_seq_reply; /* remote ls_recover_seq */
484 char rc_buf[];
485};
486
487struct dlm_opt_header {
488 __le16 t_type;
489 __le16 t_length;
490 __le32 t_pad;
491 /* need to be 8 byte aligned */
492 char t_value[];
493};
494
495/* encapsulation header */
496struct dlm_opts {
497 struct dlm_header o_header;
498 uint8_t o_nextcmd;
499 uint8_t o_pad;
500 __le16 o_optlen;
501 __le32 o_pad2;
502 char o_opts[];
503};
504
505union dlm_packet {
506 struct dlm_header header; /* common to other two */
507 struct dlm_message message;
508 struct dlm_rcom rcom;
509 struct dlm_opts opts;
510};
511
512#define DLM_RSF_NEED_SLOTS 0x00000001
513
514/* RCOM_STATUS data */
515struct rcom_status {
516 __le32 rs_flags;
517 __le32 rs_unused1;
518 __le64 rs_unused2;
519};
520
521/* RCOM_STATUS_REPLY data */
522struct rcom_config {
523 __le32 rf_lvblen;
524 __le32 rf_lsflags;
525
526 /* DLM_HEADER_SLOTS adds: */
527 __le32 rf_flags;
528 __le16 rf_our_slot;
529 __le16 rf_num_slots;
530 __le32 rf_generation;
531 __le32 rf_unused1;
532 __le64 rf_unused2;
533};
534
535struct rcom_slot {
536 __le32 ro_nodeid;
537 __le16 ro_slot;
538 __le16 ro_unused1;
539 __le64 ro_unused2;
540};
541
542struct rcom_lock {
543 __le32 rl_ownpid;
544 __le32 rl_lkid;
545 __le32 rl_remid;
546 __le32 rl_parent_lkid;
547 __le32 rl_parent_remid;
548 __le32 rl_exflags;
549 __le32 rl_flags;
550 __le32 rl_lvbseq;
551 __le32 rl_result;
552 int8_t rl_rqmode;
553 int8_t rl_grmode;
554 int8_t rl_status;
555 int8_t rl_asts;
556 __le16 rl_wait_type;
557 __le16 rl_namelen;
558 char rl_name[DLM_RESNAME_MAXLEN];
559 char rl_lvb[];
560};
561
562struct dlm_ls {
563 struct list_head ls_list; /* list of lockspaces */
564 uint32_t ls_global_id; /* global unique lockspace ID */
565 uint32_t ls_generation;
566 uint32_t ls_exflags;
567 int ls_lvblen;
568 atomic_t ls_count; /* refcount of processes in
569 the dlm using this ls */
570 wait_queue_head_t ls_count_wait;
571 int ls_create_count; /* create/release refcount */
572 unsigned long ls_flags; /* LSFL_ */
573 struct kobject ls_kobj;
574
575 struct xarray ls_lkbxa;
576 rwlock_t ls_lkbxa_lock;
577
578 /* an rsb is on rsbtl for primary locking functions,
579 and on a slow list for recovery/dump iteration */
580 struct rhashtable ls_rsbtbl;
581 rwlock_t ls_rsbtbl_lock; /* for ls_rsbtbl and ls_slow */
582 struct list_head ls_slow_inactive; /* to iterate rsbtbl */
583 struct list_head ls_slow_active; /* to iterate rsbtbl */
584
585 struct timer_list ls_scan_timer; /* based on first scan_list rsb toss_time */
586 struct list_head ls_scan_list; /* rsbs ordered by res_toss_time */
587 spinlock_t ls_scan_lock;
588
589 spinlock_t ls_waiters_lock;
590 struct list_head ls_waiters; /* lkbs needing a reply */
591
592 spinlock_t ls_orphans_lock;
593 struct list_head ls_orphans;
594
595 struct list_head ls_nodes; /* current nodes in ls */
596 struct list_head ls_nodes_gone; /* dead node list, recovery */
597 int ls_num_nodes; /* number of nodes in ls */
598 int ls_low_nodeid;
599 int ls_total_weight;
600 int *ls_node_array;
601
602 int ls_slot;
603 int ls_num_slots;
604 int ls_slots_size;
605 struct dlm_slot *ls_slots;
606
607 struct dlm_rsb ls_local_rsb; /* for returning errors */
608 struct dlm_lkb ls_local_lkb; /* for returning errors */
609 struct dlm_message ls_local_ms; /* for faking a reply */
610
611 struct dentry *ls_debug_rsb_dentry; /* debugfs */
612 struct dentry *ls_debug_waiters_dentry; /* debugfs */
613 struct dentry *ls_debug_locks_dentry; /* debugfs */
614 struct dentry *ls_debug_all_dentry; /* debugfs */
615 struct dentry *ls_debug_toss_dentry; /* debugfs */
616 struct dentry *ls_debug_queued_asts_dentry; /* debugfs */
617
618 wait_queue_head_t ls_uevent_wait; /* user part of join/leave */
619 int ls_uevent_result;
620 struct completion ls_recovery_done;
621 int ls_recovery_result;
622
623 struct miscdevice ls_device;
624
625 struct workqueue_struct *ls_callback_wq;
626
627 /* recovery related */
628
629 spinlock_t ls_cb_lock;
630 struct list_head ls_cb_delay; /* save for queue_work later */
631 struct task_struct *ls_recoverd_task;
632 struct mutex ls_recoverd_active;
633 spinlock_t ls_recover_lock;
634 unsigned long ls_recover_begin; /* jiffies timestamp */
635 uint32_t ls_recover_status; /* DLM_RS_ */
636 uint64_t ls_recover_seq;
637 struct dlm_recover *ls_recover_args;
638 struct rw_semaphore ls_in_recovery; /* block local requests */
639 rwlock_t ls_recv_active; /* block dlm_recv */
640 struct list_head ls_requestqueue;/* queue remote requests */
641 rwlock_t ls_requestqueue_lock;
642 struct dlm_rcom *ls_recover_buf;
643 int ls_recover_nodeid; /* for debugging */
644 unsigned int ls_recover_locks_in; /* for log info */
645 uint64_t ls_rcom_seq;
646 spinlock_t ls_rcom_spin;
647 struct list_head ls_recover_list;
648 spinlock_t ls_recover_list_lock;
649 int ls_recover_list_count;
650 struct xarray ls_recover_xa;
651 spinlock_t ls_recover_xa_lock;
652 wait_queue_head_t ls_wait_general;
653 wait_queue_head_t ls_recover_lock_wait;
654 spinlock_t ls_clear_proc_locks;
655
656 struct list_head ls_masters_list; /* root resources */
657 rwlock_t ls_masters_lock; /* protect root_list */
658 struct list_head ls_dir_dump_list; /* root resources */
659 rwlock_t ls_dir_dump_lock; /* protect root_list */
660
661 const struct dlm_lockspace_ops *ls_ops;
662 void *ls_ops_arg;
663
664 struct work_struct ls_free_work;
665
666 int ls_namelen;
667 char ls_name[DLM_LOCKSPACE_LEN + 1];
668};
669
670/*
671 * LSFL_RECOVER_STOP - dlm_ls_stop() sets this to tell dlm recovery routines
672 * that they should abort what they're doing so new recovery can be started.
673 *
674 * LSFL_RECOVER_DOWN - dlm_ls_stop() sets this to tell dlm_recoverd that it
675 * should do down_write() on the in_recovery rw_semaphore. (doing down_write
676 * within dlm_ls_stop causes complaints about the lock acquired/released
677 * in different contexts.)
678 *
679 * LSFL_RECOVER_LOCK - dlm_recoverd holds the in_recovery rw_semaphore.
680 * It sets this after it is done with down_write() on the in_recovery
681 * rw_semaphore and clears it after it has released the rw_semaphore.
682 *
683 * LSFL_RECOVER_WORK - dlm_ls_start() sets this to tell dlm_recoverd that it
684 * should begin recovery of the lockspace.
685 *
686 * LSFL_RUNNING - set when normal locking activity is enabled.
687 * dlm_ls_stop() clears this to tell dlm locking routines that they should
688 * quit what they are doing so recovery can run. dlm_recoverd sets
689 * this after recovery is finished.
690 */
691
692#define LSFL_RECOVER_STOP 0
693#define LSFL_RECOVER_DOWN 1
694#define LSFL_RECOVER_LOCK 2
695#define LSFL_RECOVER_WORK 3
696#define LSFL_RUNNING 4
697
698#define LSFL_RCOM_READY 5
699#define LSFL_RCOM_WAIT 6
700#define LSFL_UEVENT_WAIT 7
701#define LSFL_CB_DELAY 9
702#define LSFL_NODIR 10
703#define LSFL_RECV_MSG_BLOCKED 11
704#define LSFL_FS 12
705#define LSFL_SOFTIRQ 13
706
707#define DLM_PROC_FLAGS_CLOSING 1
708#define DLM_PROC_FLAGS_COMPAT 2
709
710/* locks list is kept so we can remove all a process's locks when it
711 exits (or orphan those that are persistent) */
712
713struct dlm_user_proc {
714 dlm_lockspace_t *lockspace;
715 unsigned long flags; /* DLM_PROC_FLAGS */
716 struct list_head asts;
717 spinlock_t asts_spin;
718 struct list_head locks;
719 spinlock_t locks_spin;
720 struct list_head unlocking;
721 wait_queue_head_t wait;
722};
723
724static inline int dlm_locking_stopped(struct dlm_ls *ls)
725{
726 return !test_bit(LSFL_RUNNING, &ls->ls_flags);
727}
728
729static inline int dlm_recovery_stopped(struct dlm_ls *ls)
730{
731 return test_bit(LSFL_RECOVER_STOP, &ls->ls_flags);
732}
733
734static inline int dlm_no_directory(struct dlm_ls *ls)
735{
736 return test_bit(LSFL_NODIR, &ls->ls_flags);
737}
738
739/* takes a snapshot from dlm atomic flags */
740static inline uint32_t dlm_flags_val(const unsigned long *addr,
741 uint32_t min, uint32_t max)
742{
743 uint32_t bit = min, val = 0;
744
745 for_each_set_bit_from(bit, addr, max + 1) {
746 val |= BIT(bit);
747 }
748
749 return val;
750}
751
752static inline uint32_t dlm_iflags_val(const struct dlm_lkb *lkb)
753{
754 return dlm_flags_val(&lkb->lkb_iflags, __DLM_IFL_MIN_BIT,
755 __DLM_IFL_MAX_BIT);
756}
757
758static inline uint32_t dlm_dflags_val(const struct dlm_lkb *lkb)
759{
760 return dlm_flags_val(&lkb->lkb_dflags, __DLM_DFL_MIN_BIT,
761 __DLM_DFL_MAX_BIT);
762}
763
764/* coming from UAPI header
765 *
766 * TODO:
767 * Move this to UAPI header and let other values point to them and use BIT()
768 */
769#define DLM_SBF_DEMOTED_BIT 0
770#define __DLM_SBF_MIN_BIT DLM_SBF_DEMOTED_BIT
771#define DLM_SBF_VALNOTVALID_BIT 1
772#define DLM_SBF_ALTMODE_BIT 2
773#define __DLM_SBF_MAX_BIT DLM_SBF_ALTMODE_BIT
774
775static inline uint32_t dlm_sbflags_val(const struct dlm_lkb *lkb)
776{
777 /* be sure the next person updates this */
778 BUILD_BUG_ON(BIT(__DLM_SBF_MAX_BIT) != DLM_SBF_ALTMODE);
779
780 return dlm_flags_val(&lkb->lkb_sbflags, __DLM_SBF_MIN_BIT,
781 __DLM_SBF_MAX_BIT);
782}
783
784static inline void dlm_set_flags_val(unsigned long *addr, uint32_t val,
785 uint32_t min, uint32_t max)
786{
787 uint32_t bit;
788
789 for (bit = min; bit < (max + 1); bit++) {
790 if (val & BIT(bit))
791 set_bit(bit, addr);
792 else
793 clear_bit(bit, addr);
794 }
795}
796
797static inline void dlm_set_dflags_val(struct dlm_lkb *lkb, uint32_t val)
798{
799 dlm_set_flags_val(&lkb->lkb_dflags, val, __DLM_DFL_MIN_BIT,
800 __DLM_DFL_MAX_BIT);
801}
802
803static inline void dlm_set_sbflags_val(struct dlm_lkb *lkb, uint32_t val)
804{
805 dlm_set_flags_val(&lkb->lkb_sbflags, val, __DLM_SBF_MIN_BIT,
806 __DLM_SBF_MAX_BIT);
807}
808
809extern struct workqueue_struct *dlm_wq;
810
811int dlm_plock_init(void);
812void dlm_plock_exit(void);
813
814#ifdef CONFIG_DLM_DEBUG
815void dlm_register_debugfs(void);
816void dlm_unregister_debugfs(void);
817void dlm_create_debug_file(struct dlm_ls *ls);
818void dlm_delete_debug_file(struct dlm_ls *ls);
819void *dlm_create_debug_comms_file(int nodeid, void *data);
820void dlm_delete_debug_comms_file(void *ctx);
821#else
822static inline void dlm_register_debugfs(void) { }
823static inline void dlm_unregister_debugfs(void) { }
824static inline void dlm_create_debug_file(struct dlm_ls *ls) { }
825static inline void dlm_delete_debug_file(struct dlm_ls *ls) { }
826static inline void *dlm_create_debug_comms_file(int nodeid, void *data) { return NULL; }
827static inline void dlm_delete_debug_comms_file(void *ctx) { }
828#endif
829
830#endif /* __DLM_INTERNAL_DOT_H__ */
831