1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
37 #ifndef _LUSTRE_DLM_H__
38 #define _LUSTRE_DLM_H__
40 #if defined(__linux__)
41 #include <linux/lustre_dlm.h>
42 #elif defined(__APPLE__)
43 #include <darwin/lustre_dlm.h>
44 #elif defined(__WINNT__)
45 #include <winnt/lustre_dlm.h>
47 #error Unsupported operating system.
50 #include <lustre_lib.h>
51 #include <lustre_net.h>
52 #include <lustre_import.h>
53 #include <lustre_handles.h>
54 #include <lustre_export.h> /* for obd_export, for LDLM_DEBUG */
55 #include <interval_tree.h> /* for interval_node{}, ldlm_extent */
60 #define OBD_LDLM_DEVICENAME "ldlm"
62 #ifdef HAVE_BGL_SUPPORT
63 /* 1.5 times the maximum 128 tasks available in VN mode */
64 #define LDLM_DEFAULT_LRU_SIZE 196
66 #define LDLM_DEFAULT_LRU_SIZE (100 * num_online_cpus())
68 #define LDLM_DEFAULT_MAX_ALIVE (cfs_time_seconds(36000))
73 ELDLM_LOCK_CHANGED = 300,
74 ELDLM_LOCK_ABORTED = 301,
75 ELDLM_LOCK_REPLACED = 302,
76 ELDLM_NO_LOCK_DATA = 303,
78 ELDLM_NAMESPACE_EXISTS = 400,
79 ELDLM_BAD_NAMESPACE = 401
83 LDLM_NAMESPACE_SERVER = 1 << 0,
84 LDLM_NAMESPACE_CLIENT = 1 << 1
87 #define LDLM_FL_LOCK_CHANGED 0x000001 /* extent, mode, or resource changed */
89 /* If the server returns one of these flags, then the lock was put on that list.
90 * If the client sends one of these flags (during recovery ONLY!), it wants the
91 * lock added to the specified list, no questions asked. -p */
92 #define LDLM_FL_BLOCK_GRANTED 0x000002
93 #define LDLM_FL_BLOCK_CONV 0x000004
94 #define LDLM_FL_BLOCK_WAIT 0x000008
96 #define LDLM_FL_CBPENDING 0x000010 /* this lock is being destroyed */
97 #define LDLM_FL_AST_SENT 0x000020 /* blocking or cancel packet was sent */
98 #define LDLM_FL_WAIT_NOREPROC 0x000040 /* not a real flag, not saved in lock */
99 #define LDLM_FL_CANCEL 0x000080 /* cancellation callback already run */
101 /* Lock is being replayed. This could probably be implied by the fact that one
102 * of BLOCK_{GRANTED,CONV,WAIT} is set, but that is pretty dangerous. */
103 #define LDLM_FL_REPLAY 0x000100
105 #define LDLM_FL_INTENT_ONLY 0x000200 /* don't grant lock, just do intent */
106 #define LDLM_FL_LOCAL_ONLY 0x000400 /* see ldlm_cli_cancel_unused */
108 /* don't run the cancel callback under ldlm_cli_cancel_unused */
109 #define LDLM_FL_FAILED 0x000800
111 #define LDLM_FL_HAS_INTENT 0x001000 /* lock request has intent */
112 #define LDLM_FL_CANCELING 0x002000 /* lock cancel has already been sent */
113 #define LDLM_FL_LOCAL 0x004000 /* local lock (ie, no srv/cli split) */
114 #define LDLM_FL_WARN 0x008000 /* see ldlm_cli_cancel_unused */
115 #define LDLM_FL_DISCARD_DATA 0x010000 /* discard (no writeback) on cancel */
117 #define LDLM_FL_NO_TIMEOUT 0x020000 /* Blocked by group lock - wait
120 /* file & record locking */
121 #define LDLM_FL_BLOCK_NOWAIT 0x040000 // server told not to wait if blocked
122 #define LDLM_FL_TEST_LOCK 0x080000 // return blocking lock
124 /* XXX FIXME: This is being added to b_size as a low-risk fix to the fact that
125 * the LVB filling happens _after_ the lock has been granted, so another thread
126 * can match`t before the LVB has been updated. As a dirty hack, we set
127 * LDLM_FL_LVB_READY only after we've done the LVB poop.
128 * this is only needed on lov/osc now, where lvb is actually used and callers
129 * must set it in input flags.
131 * The proper fix is to do the granting inside of the completion AST, which can
132 * be replaced with a LVB-aware wrapping function for OSC locks. That change is
133 * pretty high-risk, though, and would need a lot more testing. */
135 #define LDLM_FL_LVB_READY 0x100000
137 /* A lock contributes to the kms calculation until it has finished the part
138 * of it's cancelation that performs write back on its dirty pages. It
139 * can remain on the granted list during this whole time. Threads racing
140 * to update the kms after performing their writeback need to know to
141 * exclude each others locks from the calculation as they walk the granted
143 #define LDLM_FL_KMS_IGNORE 0x200000
145 /* Don't drop lock covering mmapped file in LRU */
146 #define LDLM_FL_NO_LRU 0x400000
148 /* Immediatelly cancel such locks when they block some other locks. Send
149 cancel notification to original lock holder, but expect no reply. */
150 #define LDLM_FL_CANCEL_ON_BLOCK 0x800000
152 /* Flags flags inherited from parent lock when doing intents. */
153 #define LDLM_INHERIT_FLAGS (LDLM_FL_CANCEL_ON_BLOCK)
155 /* completion ast to be executed */
156 #define LDLM_FL_CP_REQD 0x1000000
158 /* cleanup_resource has already handled the lock */
159 #define LDLM_FL_CLEANED 0x2000000
161 /* optimization hint: LDLM can run blocking callback from current context
162 * w/o involving separate thread. in order to decrease cs rate */
163 #define LDLM_FL_ATOMIC_CB 0x4000000
165 /* It may happen that a client initiate 2 operations, e.g. unlink and mkdir,
166 * such that server send blocking ast for conflict locks to this client for
167 * the 1st operation, whereas the 2nd operation has canceled this lock and
168 * is waiting for rpc_lock which is taken by the 1st operation.
169 * LDLM_FL_BL_AST is to be set by ldlm_callback_handler() to the lock not allow
170 * ELC code to cancel it.
171 * LDLM_FL_BL_DONE is to be set by ldlm_cancel_callback() when lock cache is
172 * droped to let ldlm_callback_handler() return EINVAL to the server. It is
173 * used when ELC rpc is already prepared and is waiting for rpc_lock, too late
174 * to send a separate CANCEL rpc. */
175 #define LDLM_FL_BL_AST 0x10000000
176 #define LDLM_FL_BL_DONE 0x20000000
178 /* measure lock contention and return -EUSERS if locking contention is high */
179 #define LDLM_FL_DENY_ON_CONTENTION 0x40000000
181 /* These are flags that are mapped into the flags and ASTs of blocking locks */
182 #define LDLM_AST_DISCARD_DATA 0x80000000 /* Add FL_DISCARD to blocking ASTs */
184 /* Flags sent in AST lock_flags to be mapped into the receiving lock. */
185 #define LDLM_AST_FLAGS (LDLM_FL_DISCARD_DATA)
188 * --------------------------------------------------------------------------
189 * NOTE! Starting from this point, that is, LDLM_FL_* flags with values above
190 * 0x80000000 will not be sent over the wire.
191 * --------------------------------------------------------------------------
194 /* Used for marking lock as an target for -EINTR while cp_ast sleep
195 * emulation + race with upcoming bl_ast. */
196 #define LDLM_FL_FAIL_LOC 0x100000000ULL
198 /* The blocking callback is overloaded to perform two functions. These flags
199 * indicate which operation should be performed. */
200 #define LDLM_CB_BLOCKING 1
201 #define LDLM_CB_CANCELING 2
203 /* compatibility matrix */
204 #define LCK_COMPAT_EX LCK_NL
205 #define LCK_COMPAT_PW (LCK_COMPAT_EX | LCK_CR)
206 #define LCK_COMPAT_PR (LCK_COMPAT_PW | LCK_PR)
207 #define LCK_COMPAT_CW (LCK_COMPAT_PW | LCK_CW)
208 #define LCK_COMPAT_CR (LCK_COMPAT_CW | LCK_PR | LCK_PW)
209 #define LCK_COMPAT_NL (LCK_COMPAT_CR | LCK_EX | LCK_GROUP)
210 #define LCK_COMPAT_GROUP (LCK_GROUP | LCK_NL)
212 extern ldlm_mode_t lck_compat_array[];
214 static inline void lockmode_verify(ldlm_mode_t mode)
216 LASSERT(mode > LCK_MINMODE && mode < LCK_MAXMODE);
219 static inline int lockmode_compat(ldlm_mode_t exist, ldlm_mode_t new)
221 return (lck_compat_array[exist] & new);
226 * cluster name spaces
230 #define DLM_OST_NAMESPACE 1
231 #define DLM_MDS_NAMESPACE 2
234 - do we just separate this by security domains and use a prefix for
235 multiple namespaces in the same domain?
245 * waiting_locks_spinlock
260 struct ldlm_resource;
261 struct ldlm_namespace;
263 struct ldlm_pool_ops {
264 int (*po_recalc)(struct ldlm_pool *pl);
265 int (*po_shrink)(struct ldlm_pool *pl, int nr,
266 unsigned int gfp_mask);
267 int (*po_setup)(struct ldlm_pool *pl, int limit);
271 * One second for pools thread check interval. Each pool has own period.
273 #define LDLM_POOLS_THREAD_PERIOD (1)
276 * 5% margin for modest pools. See ldlm_pool.c for details.
278 #define LDLM_POOLS_MODEST_MARGIN (5)
281 * Default recalc period for server side pools in sec.
283 #define LDLM_POOL_SRV_DEF_RECALC_PERIOD (1)
286 * Default recalc period for client side pools in sec.
288 #define LDLM_POOL_CLI_DEF_RECALC_PERIOD (10)
292 * Pool proc directory.
294 cfs_proc_dir_entry_t *pl_proc_dir;
296 * Pool name, should be long enough to contain compound proc entry name.
300 * Lock for protecting slv/clv updates.
304 * Number of allowed locks in in pool, both, client and server side.
308 * Number of granted locks in
314 atomic_t pl_grant_rate;
318 atomic_t pl_cancel_rate;
320 * Grant speed (GR-CR) per T.
322 atomic_t pl_grant_speed;
324 * Server lock volume. Protected by pl_lock.
326 __u64 pl_server_lock_volume;
328 * Current biggest client lock volume. Protected by pl_lock.
330 __u64 pl_client_lock_volume;
332 * Lock volume factor. SLV on client is calculated as following:
333 * server_slv * lock_volume_factor.
335 atomic_t pl_lock_volume_factor;
337 * Time when last slv from server was obtained.
339 time_t pl_recalc_time;
341 * Recalc period for pool.
343 time_t pl_recalc_period;
345 * Recalc and shrink ops.
347 struct ldlm_pool_ops *pl_ops;
349 * Number of planned locks for next period.
355 struct lprocfs_stats *pl_stats;
357 typedef int (*ldlm_res_policy)(struct ldlm_namespace *, struct ldlm_lock **,
358 void *req_cookie, ldlm_mode_t mode, int flags,
361 struct ldlm_valblock_ops {
362 int (*lvbo_init)(struct ldlm_resource *res);
363 int (*lvbo_update)(struct ldlm_resource *res, struct ptlrpc_request *r,
364 int buf_idx, int increase);
368 LDLM_NAMESPACE_GREEDY = 1 << 0,
369 LDLM_NAMESPACE_MODEST = 1 << 1
372 /* default values for the "max_nolock_size", "contention_time"
373 * and "contended_locks" namespace tunables */
374 #define NS_DEFAULT_MAX_NOLOCK_BYTES 0
375 #define NS_DEFAULT_CONTENTION_SECONDS 2
376 #define NS_DEFAULT_CONTENDED_LOCKS 32
378 struct ldlm_namespace {
380 ldlm_side_t ns_client; /* is this a client-side lock tree? */
381 __u64 ns_connect_flags; /* ns connect flags supported
382 * by server (may be changed via proc,
383 * lru resize may be disabled/enabled) */
384 __u64 ns_orig_connect_flags; /* client side orig connect
385 * flags supported by server */
386 struct list_head *ns_hash; /* hash table for ns */
387 spinlock_t ns_hash_lock;
388 __u32 ns_refcount; /* count of resources in the hash */
389 struct list_head ns_root_list; /* all root resources in ns */
390 struct list_head ns_list_chain; /* position in global NS list */
392 struct list_head ns_unused_list; /* all root resources in ns */
394 spinlock_t ns_unused_lock;
396 unsigned int ns_max_unused;
397 unsigned int ns_max_age;
398 unsigned int ns_timeouts;
400 cfs_time_t ns_next_dump; /* next debug dump, jiffies */
404 ldlm_res_policy ns_policy;
405 struct ldlm_valblock_ops *ns_lvbo;
407 cfs_waitq_t ns_waitq;
408 struct ldlm_pool ns_pool;
409 ldlm_appetite_t ns_appetite;
411 /* if more than @ns_contented_locks found, the resource considered
413 unsigned ns_contended_locks;
414 /* the resource remembers contended state during @ns_contention_time,
416 unsigned ns_contention_time;
417 /* limit size of nolock requests, in bytes */
418 unsigned ns_max_nolock_size;
420 struct adaptive_timeout ns_at_estimate;/* estimated lock callback time*/
421 /* backward link to obd, required for ldlm pool to store new SLV. */
422 struct obd_device *ns_obd;
425 static inline int ns_is_client(struct ldlm_namespace *ns)
428 LASSERT(!(ns->ns_client & ~(LDLM_NAMESPACE_CLIENT |
429 LDLM_NAMESPACE_SERVER)));
430 LASSERT(ns->ns_client == LDLM_NAMESPACE_CLIENT ||
431 ns->ns_client == LDLM_NAMESPACE_SERVER);
432 return ns->ns_client == LDLM_NAMESPACE_CLIENT;
435 static inline int ns_is_server(struct ldlm_namespace *ns)
438 LASSERT(!(ns->ns_client & ~(LDLM_NAMESPACE_CLIENT |
439 LDLM_NAMESPACE_SERVER)));
440 LASSERT(ns->ns_client == LDLM_NAMESPACE_CLIENT ||
441 ns->ns_client == LDLM_NAMESPACE_SERVER);
442 return ns->ns_client == LDLM_NAMESPACE_SERVER;
445 static inline int ns_connect_lru_resize(struct ldlm_namespace *ns)
448 return !!(ns->ns_connect_flags & OBD_CONNECT_LRU_RESIZE);
453 * Resource hash table
457 #define RES_HASH_BITS 10
458 #define RES_HASH_SIZE (1UL << RES_HASH_BITS)
459 #define RES_HASH_MASK (RES_HASH_SIZE - 1)
463 typedef int (*ldlm_blocking_callback)(struct ldlm_lock *lock,
464 struct ldlm_lock_desc *new, void *data,
466 typedef int (*ldlm_completion_callback)(struct ldlm_lock *lock, int flags,
468 typedef int (*ldlm_glimpse_callback)(struct ldlm_lock *lock, void *data);
470 /* Interval node data for each LDLM_EXTENT lock */
471 struct ldlm_interval {
472 struct interval_node li_node; /* node for tree mgmt */
473 struct list_head li_group; /* the locks which have the same
474 * policy - group of the policy */
476 #define to_ldlm_interval(n) container_of(n, struct ldlm_interval, li_node)
478 /* the interval tree must be accessed inside the resource lock. */
479 struct ldlm_interval_tree {
480 /* tree size, this variable is used to count
481 * granted PW locks in ldlm_extent_policy()*/
483 ldlm_mode_t lit_mode; /* lock mode */
484 struct interval_node *lit_root; /* actually ldlm_interval */
488 struct portals_handle l_handle; // must be first in the structure
491 /* internal spinlock protects l_resource. we should hold this lock
492 * first before grabbing res_lock.*/
495 /* ldlm_lock_change_resource() can change this */
496 struct ldlm_resource *l_resource;
498 /* protected by ns_hash_lock. FIXME */
499 struct list_head l_lru;
501 /* protected by lr_lock, linkage to resource's lock queues */
502 struct list_head l_res_link;
504 struct ldlm_interval *l_tree_node; /* tree node for ldlm_extent */
506 /* protected by per-bucket exp->exp_lock_hash locks */
507 struct hlist_node l_exp_hash; /* per export hash of locks */
509 /* protected by lr_lock */
510 ldlm_mode_t l_req_mode;
511 ldlm_mode_t l_granted_mode;
513 ldlm_completion_callback l_completion_ast;
514 ldlm_blocking_callback l_blocking_ast;
515 ldlm_glimpse_callback l_glimpse_ast;
517 struct obd_export *l_export;
518 struct obd_export *l_conn_export;
520 struct lustre_handle l_remote_handle;
521 ldlm_policy_data_t l_policy_data;
523 /* protected by lr_lock */
529 /* If the lock is granted, a process sleeps on this waitq to learn when
530 * it's no longer in use. If the lock is not granted, a process sleeps
531 * on this waitq to learn when it becomes granted. */
534 cfs_time_t l_last_activity; /* seconds */
535 cfs_time_t l_last_used; /* jiffies */
536 struct ldlm_extent l_req_extent;
538 /* Client-side-only members */
539 __u32 l_lvb_len; /* temporary storage for */
540 void *l_lvb_data; /* an LVB received during */
541 void *l_lvb_swabber; /* an enqueue */
543 spinlock_t l_extents_list_lock;
544 struct list_head l_extents_list;
546 struct list_head l_cache_locks_list;
548 /* Server-side-only members */
550 /* protected by elt_lock */
551 struct list_head l_pending_chain; /* callbacks pending */
552 cfs_time_t l_callback_timeout; /* jiffies */
554 __u32 l_pid; /* pid which created this lock */
556 /* for ldlm_add_ast_work_item() */
557 struct list_head l_bl_ast;
558 struct list_head l_cp_ast;
559 struct ldlm_lock *l_blocking_lock;
562 /* protected by lr_lock, linkages to "skip lists" */
563 struct list_head l_sl_mode;
564 struct list_head l_sl_policy;
567 struct ldlm_resource {
568 struct ldlm_namespace *lr_namespace;
570 /* protected by ns_hash_lock */
571 struct list_head lr_hash;
572 struct ldlm_resource *lr_parent; /* 0 for a root resource */
573 struct list_head lr_children; /* list head for child resources */
574 struct list_head lr_childof; /* part of ns_root_list if root res,
575 * part of lr_children if child */
578 /* protected by lr_lock */
579 struct list_head lr_granted;
580 struct list_head lr_converting;
581 struct list_head lr_waiting;
582 ldlm_mode_t lr_most_restr;
583 ldlm_type_t lr_type; /* LDLM_{PLAIN,EXTENT,FLOCK} */
584 struct ldlm_res_id lr_name;
585 atomic_t lr_refcount;
587 struct ldlm_interval_tree lr_itree[LCK_MODE_NUM]; /* interval trees*/
589 /* Server-side-only lock value block elements */
590 struct semaphore lr_lvb_sem;
594 /* when the resource was considered as contended */
595 cfs_time_t lr_contention_time;
598 struct ldlm_ast_work {
599 struct ldlm_lock *w_lock;
601 struct ldlm_lock_desc w_desc;
602 struct list_head w_list;
608 /* ldlm_enqueue parameters common */
609 struct ldlm_enqueue_info {
610 __u32 ei_type; /* Type of the lock being enqueued. */
611 __u32 ei_mode; /* Mode of the lock being enqueued. */
612 void *ei_cb_bl; /* Different callbacks for lock handling (blocking, */
613 void *ei_cb_cp; /* completion, glimpse) */
615 void *ei_cbdata; /* Data to be passed into callbacks. */
618 extern struct obd_ops ldlm_obd_ops;
620 extern char *ldlm_lockname[];
621 extern char *ldlm_typename[];
622 extern char *ldlm_it2str(int it);
624 #define ldlm_lock_debug(cdls, level, lock, file, func, line, fmt, a...) do { \
627 if (((level) & D_CANTMASK) != 0 || \
628 ((libcfs_debug & (level)) != 0 && \
629 (libcfs_subsystem_debug & DEBUG_SUBSYSTEM) != 0)) { \
630 static struct libcfs_debug_msg_data _ldlm_dbg_data = \
631 DEBUG_MSG_DATA_INIT(cdls, DEBUG_SUBSYSTEM, \
633 _ldlm_lock_debug(lock, level, &_ldlm_dbg_data, fmt, \
638 void _ldlm_lock_debug(struct ldlm_lock *lock, __u32 mask,
639 struct libcfs_debug_msg_data *data, const char *fmt,
641 __attribute__ ((format (printf, 4, 5)));
643 #define LDLM_DEBUG_LIMIT(mask, lock, fmt, a...) do { \
644 static cfs_debug_limit_state_t _ldlm_cdls; \
645 ldlm_lock_debug(&_ldlm_cdls, mask, lock, \
646 __FILE__, __FUNCTION__, __LINE__, \
650 #define LDLM_ERROR(lock, fmt, a...) LDLM_DEBUG_LIMIT(D_ERROR, lock, fmt, ## a)
651 #define LDLM_WARN(lock, fmt, a...) LDLM_DEBUG_LIMIT(D_WARNING, lock, fmt, ## a)
653 #define LDLM_DEBUG(lock, fmt, a...) do { \
654 ldlm_lock_debug(NULL, D_DLMTRACE, lock, \
655 __FILE__, __FUNCTION__, __LINE__, \
659 #define LDLM_DEBUG_NOLOCK(format, a...) \
660 CDEBUG(D_DLMTRACE, "### " format "\n" , ##a)
662 typedef int (*ldlm_processing_policy)(struct ldlm_lock *lock, int *flags,
663 int first_enq, ldlm_error_t *err,
664 struct list_head *work_list);
670 #define LDLM_ITER_CONTINUE 1 /* keep iterating */
671 #define LDLM_ITER_STOP 2 /* stop iterating */
673 typedef int (*ldlm_iterator_t)(struct ldlm_lock *, void *);
674 typedef int (*ldlm_res_iterator_t)(struct ldlm_resource *, void *);
676 int ldlm_resource_foreach(struct ldlm_resource *res, ldlm_iterator_t iter,
678 int ldlm_namespace_foreach(struct ldlm_namespace *ns, ldlm_iterator_t iter,
680 int ldlm_namespace_foreach_res(struct ldlm_namespace *ns,
681 ldlm_res_iterator_t iter, void *closure);
683 int ldlm_replay_locks(struct obd_import *imp);
684 void ldlm_resource_iterate(struct ldlm_namespace *, struct ldlm_res_id *,
685 ldlm_iterator_t iter, void *data);
688 int ldlm_flock_completion_ast(struct ldlm_lock *lock, int flags, void *data);
691 __u64 ldlm_extent_shift_kms(struct ldlm_lock *lock, __u64 old_kms);
695 int ldlm_server_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
696 void *data, int flag);
697 int ldlm_server_completion_ast(struct ldlm_lock *lock, int flags, void *data);
698 int ldlm_server_glimpse_ast(struct ldlm_lock *lock, void *data);
699 int ldlm_handle_enqueue(struct ptlrpc_request *req, ldlm_completion_callback,
700 ldlm_blocking_callback, ldlm_glimpse_callback);
701 int ldlm_handle_convert(struct ptlrpc_request *req);
702 int ldlm_handle_cancel(struct ptlrpc_request *req);
703 int ldlm_request_cancel(struct ptlrpc_request *req,
704 struct ldlm_request *dlm_req, int first);
705 int ldlm_del_waiting_lock(struct ldlm_lock *lock);
706 int ldlm_refresh_waiting_lock(struct ldlm_lock *lock, int timeout);
707 int ldlm_get_ref(void);
708 void ldlm_put_ref(void);
709 int ldlm_init_export(struct obd_export *exp);
710 void ldlm_destroy_export(struct obd_export *exp);
713 ldlm_processing_policy ldlm_get_processing_policy(struct ldlm_resource *res);
714 void ldlm_register_intent(struct ldlm_namespace *ns, ldlm_res_policy arg);
715 void ldlm_lock2handle(struct ldlm_lock *lock, struct lustre_handle *lockh);
716 struct ldlm_lock *__ldlm_handle2lock(struct lustre_handle *, int flags);
717 void ldlm_cancel_callback(struct ldlm_lock *);
718 int ldlm_lock_set_data(struct lustre_handle *, void *data);
719 int ldlm_lock_remove_from_lru(struct ldlm_lock *);
720 struct ldlm_lock *ldlm_handle2lock_ns(struct ldlm_namespace *,
721 struct lustre_handle *);
723 static inline struct ldlm_lock *ldlm_handle2lock(struct lustre_handle *h)
725 return __ldlm_handle2lock(h, 0);
728 static inline int ldlm_res_lvbo_update(struct ldlm_resource *res,
729 struct ptlrpc_request *r, int buf_idx,
732 if (res->lr_namespace->ns_lvbo &&
733 res->lr_namespace->ns_lvbo->lvbo_update) {
734 return res->lr_namespace->ns_lvbo->lvbo_update(res, r, buf_idx,
740 #define LDLM_LOCK_PUT(lock) \
742 /*LDLM_DEBUG((lock), "put");*/ \
743 ldlm_lock_put(lock); \
746 #define LDLM_LOCK_GET(lock) \
748 ldlm_lock_get(lock); \
749 /*LDLM_DEBUG((lock), "get");*/ \
753 #define ldlm_lock_list_put(head, member, count) \
755 struct ldlm_lock *_lock, *_next; \
757 list_for_each_entry_safe(_lock, _next, head, member) { \
760 list_del_init(&_lock->member); \
761 LDLM_LOCK_PUT(_lock); \
766 struct ldlm_lock *ldlm_lock_get(struct ldlm_lock *lock);
767 void ldlm_lock_put(struct ldlm_lock *lock);
768 void ldlm_lock_destroy(struct ldlm_lock *lock);
769 void ldlm_lock2desc(struct ldlm_lock *lock, struct ldlm_lock_desc *desc);
770 void ldlm_lock_addref(struct lustre_handle *lockh, __u32 mode);
771 void ldlm_lock_decref(struct lustre_handle *lockh, __u32 mode);
772 void ldlm_lock_decref_and_cancel(struct lustre_handle *lockh, __u32 mode);
773 void ldlm_lock_allow_match(struct ldlm_lock *lock);
774 int ldlm_lock_fast_match(struct ldlm_lock *, int, obd_off, obd_off, void **);
775 void ldlm_lock_fast_release(void *, int);
776 ldlm_mode_t ldlm_lock_match(struct ldlm_namespace *ns, int flags,
777 struct ldlm_res_id *, ldlm_type_t type,
778 ldlm_policy_data_t *, ldlm_mode_t mode,
779 struct lustre_handle *);
780 struct ldlm_resource *ldlm_lock_convert(struct ldlm_lock *lock, int new_mode,
782 void ldlm_lock_cancel(struct ldlm_lock *lock);
783 void ldlm_cancel_locks_for_export(struct obd_export *export);
784 void ldlm_reprocess_all(struct ldlm_resource *res);
785 void ldlm_reprocess_all_ns(struct ldlm_namespace *ns);
786 void ldlm_lock_dump(int level, struct ldlm_lock *lock, int pos);
787 void ldlm_lock_dump_handle(int level, struct lustre_handle *);
788 void ldlm_unlink_lock_skiplist(struct ldlm_lock *req);
791 struct ldlm_namespace *
792 ldlm_namespace_new(struct obd_device *obd, char *name,
793 ldlm_side_t client, ldlm_appetite_t apt);
794 int ldlm_namespace_cleanup(struct ldlm_namespace *ns, int flags);
795 void ldlm_namespace_free(struct ldlm_namespace *ns,
796 struct obd_import *imp, int force);
797 void ldlm_namespace_register(struct ldlm_namespace *ns, ldlm_side_t client);
798 void ldlm_namespace_unregister(struct ldlm_namespace *ns, ldlm_side_t client);
799 void ldlm_namespace_move_locked(struct ldlm_namespace *ns, ldlm_side_t client);
800 struct ldlm_namespace *ldlm_namespace_first_locked(ldlm_side_t client);
801 void ldlm_namespace_get_locked(struct ldlm_namespace *ns);
802 void ldlm_namespace_put_locked(struct ldlm_namespace *ns, int wakeup);
803 void ldlm_namespace_get(struct ldlm_namespace *ns);
804 void ldlm_namespace_put(struct ldlm_namespace *ns, int wakeup);
805 int ldlm_proc_setup(void);
807 void ldlm_proc_cleanup(void);
809 static inline void ldlm_proc_cleanup(void) {}
812 /* resource.c - internal */
813 struct ldlm_resource *ldlm_resource_get(struct ldlm_namespace *ns,
814 struct ldlm_resource *parent,
815 struct ldlm_res_id, ldlm_type_t type,
817 struct ldlm_resource *ldlm_resource_getref(struct ldlm_resource *res);
818 int ldlm_resource_putref(struct ldlm_resource *res);
819 void ldlm_resource_add_lock(struct ldlm_resource *res, struct list_head *head,
820 struct ldlm_lock *lock);
821 void ldlm_resource_unlink_lock(struct ldlm_lock *lock);
822 void ldlm_res2desc(struct ldlm_resource *res, struct ldlm_resource_desc *desc);
823 void ldlm_dump_all_namespaces(ldlm_side_t client, int level);
824 void ldlm_namespace_dump(int level, struct ldlm_namespace *);
825 void ldlm_resource_dump(int level, struct ldlm_resource *);
826 int ldlm_lock_change_resource(struct ldlm_namespace *, struct ldlm_lock *,
830 int ldlm_expired_completion_wait(void *data);
831 int ldlm_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc,
832 void *data, int flag);
833 int ldlm_glimpse_ast(struct ldlm_lock *lock, void *reqp);
834 int ldlm_completion_ast(struct ldlm_lock *lock, int flags, void *data);
835 int ldlm_cli_enqueue(struct obd_export *exp, struct ptlrpc_request **req,
836 struct ldlm_enqueue_info *einfo, struct ldlm_res_id res_id,
837 ldlm_policy_data_t *policy, int *flags,
838 void *lvb, __u32 lvb_len, void *lvb_swabber,
839 struct lustre_handle *lockh, int async);
840 struct ptlrpc_request *ldlm_prep_enqueue_req(struct obd_export *exp,
841 int bufcount, __u32 *size,
842 struct list_head *head, int count);
843 struct ptlrpc_request *ldlm_prep_elc_req(struct obd_export *exp, int version,
844 int opc, int bufcount, __u32 *size,
845 int bufoff, int canceloff,
846 struct list_head *cancels, int count);
847 int ldlm_cli_enqueue_fini(struct obd_export *exp, struct ptlrpc_request *req,
848 ldlm_type_t type, __u8 with_policy, ldlm_mode_t mode,
849 int *flags, void *lvb, __u32 lvb_len,
850 void *lvb_swabber, struct lustre_handle *lockh,
852 int ldlm_cli_enqueue_local(struct ldlm_namespace *ns,
853 struct ldlm_res_id *res_id,
854 ldlm_type_t type, ldlm_policy_data_t *policy,
855 ldlm_mode_t mode, int *flags,
856 ldlm_blocking_callback blocking,
857 ldlm_completion_callback completion,
858 ldlm_glimpse_callback glimpse,
859 void *data, __u32 lvb_len, void *lvb_swabber,
860 struct lustre_handle *lockh);
861 int ldlm_server_ast(struct lustre_handle *lockh, struct ldlm_lock_desc *new,
862 void *data, __u32 data_len);
863 int ldlm_cli_convert(struct lustre_handle *, int new_mode, __u32 *flags);
864 int ldlm_cli_update_pool(struct ptlrpc_request *req);
865 int ldlm_cli_cancel(struct lustre_handle *lockh);
866 int ldlm_cli_cancel_unused(struct ldlm_namespace *, struct ldlm_res_id *,
867 int flags, void *opaque);
868 int ldlm_cli_cancel_req(struct obd_export *exp,
869 struct list_head *head, int count);
870 int ldlm_cli_join_lru(struct ldlm_namespace *, struct ldlm_res_id *, int join);
871 int ldlm_cancel_resource_local(struct ldlm_resource *res,
872 struct list_head *cancels,
873 ldlm_policy_data_t *policy, ldlm_mode_t mode,
874 int lock_flags, int cancel_flags, void *opaque);
875 int ldlm_cli_cancel_list(struct list_head *head, int count,
876 struct ptlrpc_request *req, int off);
878 /* ioctls for trying requests */
879 #define IOC_LDLM_TYPE 'f'
880 #define IOC_LDLM_MIN_NR 40
882 #define IOC_LDLM_TEST _IOWR('f', 40, long)
883 #define IOC_LDLM_DUMP _IOWR('f', 41, long)
884 #define IOC_LDLM_REGRESS_START _IOWR('f', 42, long)
885 #define IOC_LDLM_REGRESS_STOP _IOWR('f', 43, long)
886 #define IOC_LDLM_MAX_NR 43
888 static inline void lock_res(struct ldlm_resource *res)
890 spin_lock(&res->lr_lock);
893 static inline void unlock_res(struct ldlm_resource *res)
895 spin_unlock(&res->lr_lock);
898 static inline void check_res_locked(struct ldlm_resource *res)
900 LASSERT_SPIN_LOCKED(&res->lr_lock);
903 struct ldlm_resource * lock_res_and_lock(struct ldlm_lock *lock);
904 void unlock_res_and_lock(struct ldlm_lock *lock);
907 void ldlm_pools_recalc(ldlm_side_t client);
908 int ldlm_pools_init(void);
909 void ldlm_pools_fini(void);
911 int ldlm_pool_init(struct ldlm_pool *pl, struct ldlm_namespace *ns,
912 int idx, ldlm_side_t client);
913 int ldlm_pool_shrink(struct ldlm_pool *pl, int nr,
914 unsigned int gfp_mask);
915 void ldlm_pool_fini(struct ldlm_pool *pl);
916 int ldlm_pool_setup(struct ldlm_pool *pl, int limit);
917 int ldlm_pool_recalc(struct ldlm_pool *pl);
918 __u32 ldlm_pool_get_lvf(struct ldlm_pool *pl);
919 __u64 ldlm_pool_get_slv(struct ldlm_pool *pl);
920 __u64 ldlm_pool_get_clv(struct ldlm_pool *pl);
921 __u32 ldlm_pool_get_limit(struct ldlm_pool *pl);
922 void ldlm_pool_set_slv(struct ldlm_pool *pl, __u64 slv);
923 void ldlm_pool_set_clv(struct ldlm_pool *pl, __u64 clv);
924 void ldlm_pool_set_limit(struct ldlm_pool *pl, __u32 limit);
925 void ldlm_pool_add(struct ldlm_pool *pl, struct ldlm_lock *lock);
926 void ldlm_pool_del(struct ldlm_pool *pl, struct ldlm_lock *lock);