Whamcloud - gitweb
LU-1061 agl: cl_locks_prune() waits for the last user
[fs/lustre-release.git] / lustre / include / lustre_dlm.h
index cbf2b59..bd4d87d 100644 (file)
@@ -1,11 +1,50 @@
 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
- * (visit-tags-table FILE)
  * vim:expandtab:shiftwidth=8:tabstop=8:
+ *
+ * GPL HEADER START
+ *
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 only,
+ * as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License version 2 for more details (a copy is included
+ * in the LICENSE file that accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License
+ * version 2 along with this program; If not, see
+ * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
+ *
+ * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
+ * CA 95054 USA or visit www.sun.com if you need additional information or
+ * have any questions.
+ *
+ * GPL HEADER END
+ */
+/*
+ * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Use is subject to license terms.
+ *
+ * Copyright (c) 2010 Whamcloud, Inc
+ *
+ */
+/*
+ * This file is part of Lustre, http://www.lustre.org/
+ * Lustre is a trademark of Sun Microsystems, Inc.
  */
 
 #ifndef _LUSTRE_DLM_H__
 #define _LUSTRE_DLM_H__
 
+/** \defgroup ldlm ldlm
+ *
+ * @{
+ */
+
 #if defined(__linux__)
 #include <linux/lustre_dlm.h>
 #elif defined(__APPLE__)
 #include <lustre_net.h>
 #include <lustre_import.h>
 #include <lustre_handles.h>
-#include <lustre_export.h> /* for obd_export, for LDLM_DEBUG */
+#include <interval_tree.h> /* for interval_node{}, ldlm_extent */
+#include <lu_ref.h>
 
 struct obd_ops;
 struct obd_device;
 
 #define OBD_LDLM_DEVICENAME  "ldlm"
 
-#define LDLM_DEFAULT_LRU_SIZE (100 * smp_num_cpus)
+#ifdef HAVE_BGL_SUPPORT
+/* 1.5 times the maximum 128 tasks available in VN mode */
+#define LDLM_DEFAULT_LRU_SIZE 196
+#else
+#define LDLM_DEFAULT_LRU_SIZE (100 * cfs_num_online_cpus())
+#endif
 #define LDLM_DEFAULT_MAX_ALIVE (cfs_time_seconds(36000))
+#define LDLM_CTIME_AGE_LIMIT (10)
+#define LDLM_DEFAULT_PARALLEL_AST_LIMIT 1024
 
 typedef enum {
         ELDLM_OK = 0,
@@ -42,8 +89,10 @@ typedef enum {
         ELDLM_BAD_NAMESPACE    = 401
 } ldlm_error_t;
 
-#define LDLM_NAMESPACE_SERVER 0
-#define LDLM_NAMESPACE_CLIENT 1
+typedef enum {
+        LDLM_NAMESPACE_SERVER = 1 << 0,
+        LDLM_NAMESPACE_CLIENT = 1 << 1
+} ldlm_side_t;
 
 #define LDLM_FL_LOCK_CHANGED   0x000001 /* extent, mode, or resource changed */
 
@@ -55,7 +104,8 @@ typedef enum {
 #define LDLM_FL_BLOCK_WAIT     0x000008
 
 #define LDLM_FL_CBPENDING      0x000010 /* this lock is being destroyed */
-#define LDLM_FL_AST_SENT       0x000020 /* blocking or cancel packet was sent */
+#define LDLM_FL_AST_SENT       0x000020 /* blocking or cancel packet was
+                                         * queued for sending. */
 #define LDLM_FL_WAIT_NOREPROC  0x000040 /* not a real flag, not saved in lock */
 #define LDLM_FL_CANCEL         0x000080 /* cancellation callback already run */
 
@@ -64,7 +114,7 @@ typedef enum {
 #define LDLM_FL_REPLAY         0x000100
 
 #define LDLM_FL_INTENT_ONLY    0x000200 /* don't grant lock, just do intent */
-#define LDLM_FL_LOCAL_ONLY     0x000400 /* see ldlm_cli_cancel_unused */
+#define LDLM_FL_LOCAL_ONLY     0x000400
 
 /* don't run the cancel callback under ldlm_cli_cancel_unused */
 #define LDLM_FL_FAILED         0x000800
@@ -72,14 +122,15 @@ typedef enum {
 #define LDLM_FL_HAS_INTENT     0x001000 /* lock request has intent */
 #define LDLM_FL_CANCELING      0x002000 /* lock cancel has already been sent */
 #define LDLM_FL_LOCAL          0x004000 /* local lock (ie, no srv/cli split) */
-#define LDLM_FL_WARN           0x008000 /* see ldlm_cli_cancel_unused */
 #define LDLM_FL_DISCARD_DATA   0x010000 /* discard (no writeback) on cancel */
 
 #define LDLM_FL_NO_TIMEOUT     0x020000 /* Blocked by group lock - wait
                                          * indefinitely */
 
 /* file & record locking */
-#define LDLM_FL_BLOCK_NOWAIT   0x040000 // server told not to wait if blocked
+#define LDLM_FL_BLOCK_NOWAIT   0x040000 /* server told not to wait if blocked.
+                                         * For AGL, OST will not send glimpse
+                                         * callback. */
 #define LDLM_FL_TEST_LOCK      0x080000 // return blocking lock
 
 /* XXX FIXME: This is being added to b_size as a low-risk fix to the fact that
@@ -103,21 +154,19 @@ typedef enum {
  * list. */
 #define LDLM_FL_KMS_IGNORE     0x200000
 
-/* Don't drop lock covering mmapped file in LRU */
+/* Don't put lock into the LRU list, so that it is not canceled due to aging.
+ * Used by MGC locks, they are cancelled only at unmount or by callback. */
 #define LDLM_FL_NO_LRU         0x400000
 
 /* Immediatelly cancel such locks when they block some other locks. Send
-   cancel notification to original lock holder, but expect no reply. */
+ * cancel notification to original lock holder, but expect no reply. This is
+ * for clients (like liblustre) that cannot be expected to reliably response
+ * to blocking ast. */
 #define LDLM_FL_CANCEL_ON_BLOCK 0x800000
 
 /* Flags flags inherited from parent lock when doing intents. */
 #define LDLM_INHERIT_FLAGS     (LDLM_FL_CANCEL_ON_BLOCK)
 
-/* These are flags that are mapped into the flags and ASTs of blocking locks */
-#define LDLM_AST_DISCARD_DATA  0x80000000 /* Add FL_DISCARD to blocking ASTs */
-/* Flags sent in AST lock_flags to be mapped into the receiving lock. */
-#define LDLM_AST_FLAGS         (LDLM_FL_DISCARD_DATA)
-
 /* completion ast to be executed */
 #define LDLM_FL_CP_REQD        0x1000000
 
@@ -128,20 +177,48 @@ typedef enum {
  * w/o involving separate thread. in order to decrease cs rate */
 #define LDLM_FL_ATOMIC_CB      0x4000000
 
-/* Cancel lock asynchronously. See ldlm_cli_cancel_unused_resource. */
-#define LDLM_FL_ASYNC           0x20000000
+/* It may happen that a client initiate 2 operations, e.g. unlink and mkdir,
+ * such that server send blocking ast for conflict locks to this client for
+ * the 1st operation, whereas the 2nd operation has canceled this lock and
+ * is waiting for rpc_lock which is taken by the 1st operation.
+ * LDLM_FL_BL_AST is to be set by ldlm_callback_handler() to the lock not allow
+ * ELC code to cancel it.
+ * LDLM_FL_BL_DONE is to be set by ldlm_cancel_callback() when lock cache is
+ * droped to let ldlm_callback_handler() return EINVAL to the server. It is
+ * used when ELC rpc is already prepared and is waiting for rpc_lock, too late
+ * to send a separate CANCEL rpc. */
+#define LDLM_FL_BL_AST          0x10000000
+#define LDLM_FL_BL_DONE         0x20000000
+
+/* measure lock contention and return -EUSERS if locking contention is high */
+#define LDLM_FL_DENY_ON_CONTENTION 0x40000000
+
+/* These are flags that are mapped into the flags and ASTs of blocking locks */
+#define LDLM_AST_DISCARD_DATA  0x80000000 /* Add FL_DISCARD to blocking ASTs */
+
+/* Flags sent in AST lock_flags to be mapped into the receiving lock. */
+#define LDLM_AST_FLAGS         (LDLM_FL_DISCARD_DATA)
+
+/* 
+ * --------------------------------------------------------------------------
+ * NOTE! Starting from this point, that is, LDLM_FL_* flags with values above
+ * 0x80000000 will not be sent over the wire.
+ * --------------------------------------------------------------------------
+ */
+
+/* Used for marking lock as an target for -EINTR while cp_ast sleep
+ * emulation + race with upcoming bl_ast.  */
+#define LDLM_FL_FAIL_LOC       0x100000000ULL
+
+/* Used while processing the unused list to know that we have already
+ * handled this lock and decided to skip it */
+#define LDLM_FL_SKIPPED        0x200000000ULL
 
 /* The blocking callback is overloaded to perform two functions.  These flags
  * indicate which operation should be performed. */
 #define LDLM_CB_BLOCKING    1
 #define LDLM_CB_CANCELING   2
 
-/* position flag of skip list pointers */
-#define LDLM_SL_HEAD(skip_list)   ((skip_list)->next != NULL)
-#define LDLM_SL_TAIL(skip_list)   ((skip_list)->prev != NULL)
-#define LDLM_SL_EMPTY(skip_list)  ((skip_list)->next == NULL && \
-                                   (skip_list)->prev == NULL)
-
 /* compatibility matrix */
 #define LCK_COMPAT_EX  LCK_NL
 #define LCK_COMPAT_PW  (LCK_COMPAT_EX | LCK_CR)
@@ -150,6 +227,7 @@ typedef enum {
 #define LCK_COMPAT_CR  (LCK_COMPAT_CW | LCK_PR | LCK_PW)
 #define LCK_COMPAT_NL  (LCK_COMPAT_CR | LCK_EX | LCK_GROUP)
 #define LCK_COMPAT_GROUP  (LCK_GROUP | LCK_NL)
+#define LCK_COMPAT_COS (LCK_COS)
 
 extern ldlm_mode_t lck_compat_array[];
 
@@ -158,9 +236,9 @@ static inline void lockmode_verify(ldlm_mode_t mode)
        LASSERT(mode > LCK_MINMODE && mode < LCK_MAXMODE);
 }
 
-static inline int lockmode_compat(ldlm_mode_t exist, ldlm_mode_t new)
+static inline int lockmode_compat(ldlm_mode_t exist_mode, ldlm_mode_t new_mode)
 {
-       return (lck_compat_array[exist] & new);
+       return (lck_compat_array[exist_mode] & new_mode);
 }
 
 /*
@@ -190,61 +268,298 @@ static inline int lockmode_compat(ldlm_mode_t exist, ldlm_mode_t new)
  *     led_lock
  *
  * lr_lock
- *     ns_unused_lock
+ *     ns_lock
  *
  * lr_lvb_sem
  *     lr_lock
  *
  */
 
+struct ldlm_pool;
 struct ldlm_lock;
 struct ldlm_resource;
 struct ldlm_namespace;
 
+struct ldlm_pool_ops {
+        int (*po_recalc)(struct ldlm_pool *pl);
+        int (*po_shrink)(struct ldlm_pool *pl, int nr,
+                         unsigned int gfp_mask);
+        int (*po_setup)(struct ldlm_pool *pl, int limit);
+};
+
+/**
+ * One second for pools thread check interval. Each pool has own period.
+ */
+#define LDLM_POOLS_THREAD_PERIOD (1)
+
+/**
+ * ~6% margin for modest pools. See ldlm_pool.c for details.
+ */
+#define LDLM_POOLS_MODEST_MARGIN_SHIFT (4)
+
+/**
+ * Default recalc period for server side pools in sec.
+ */
+#define LDLM_POOL_SRV_DEF_RECALC_PERIOD (1)
+
+/**
+ * Default recalc period for client side pools in sec.
+ */
+#define LDLM_POOL_CLI_DEF_RECALC_PERIOD (10)
+
+struct ldlm_pool {
+        /**
+         * Pool proc directory.
+         */
+        cfs_proc_dir_entry_t  *pl_proc_dir;
+        /**
+         * Pool name, should be long enough to contain compound proc entry name.
+         */
+        char                   pl_name[100];
+        /**
+         * Lock for protecting slv/clv updates.
+         */
+        cfs_spinlock_t         pl_lock;
+        /**
+         * Number of allowed locks in in pool, both, client and server side.
+         */
+        cfs_atomic_t           pl_limit;
+        /**
+         * Number of granted locks in
+         */
+        cfs_atomic_t           pl_granted;
+        /**
+         * Grant rate per T.
+         */
+        cfs_atomic_t           pl_grant_rate;
+        /**
+         * Cancel rate per T.
+         */
+        cfs_atomic_t           pl_cancel_rate;
+        /**
+         * Server lock volume. Protected by pl_lock.
+         */
+        __u64                  pl_server_lock_volume;
+        /**
+         * Current biggest client lock volume. Protected by pl_lock.
+         */
+        __u64                  pl_client_lock_volume;
+        /**
+         * Lock volume factor. SLV on client is calculated as following:
+         * server_slv * lock_volume_factor.
+         */
+        cfs_atomic_t           pl_lock_volume_factor;
+        /**
+         * Time when last slv from server was obtained.
+         */
+        time_t                 pl_recalc_time;
+        /**
+          * Recalc period for pool.
+          */
+        time_t                 pl_recalc_period;
+        /**
+         * Recalc and shrink ops.
+         */
+        struct ldlm_pool_ops  *pl_ops;
+        /**
+         * Number of planned locks for next period.
+         */
+        int                    pl_grant_plan;
+        /**
+         * Pool statistics.
+         */
+        struct lprocfs_stats  *pl_stats;
+};
+
 typedef int (*ldlm_res_policy)(struct ldlm_namespace *, struct ldlm_lock **,
                                void *req_cookie, ldlm_mode_t mode, int flags,
                                void *data);
 
+typedef int (*ldlm_cancel_for_recovery)(struct ldlm_lock *lock);
+
 struct ldlm_valblock_ops {
         int (*lvbo_init)(struct ldlm_resource *res);
-        int (*lvbo_update)(struct ldlm_resource *res, struct lustre_msg *m,
-                           int buf_idx, int increase);
+        int (*lvbo_update)(struct ldlm_resource *res,
+                           struct ptlrpc_request *r,
+                           int increase);
+        int (*lvbo_free)(struct ldlm_resource *res);
 };
 
+typedef enum {
+        LDLM_NAMESPACE_GREEDY = 1 << 0,
+        LDLM_NAMESPACE_MODEST = 1 << 1
+} ldlm_appetite_t;
+
+/*
+ * Default values for the "max_nolock_size", "contention_time" and
+ * "contended_locks" namespace tunables.
+ */
+#define NS_DEFAULT_MAX_NOLOCK_BYTES 0
+#define NS_DEFAULT_CONTENTION_SECONDS 2
+#define NS_DEFAULT_CONTENDED_LOCKS 32
+
+struct ldlm_ns_bucket {
+        /** refer back */
+        struct ldlm_namespace      *nsb_namespace;
+        /** estimated lock callback time */
+        struct adaptive_timeout     nsb_at_estimate;
+};
+
+enum {
+        /** ldlm namespace lock stats */
+        LDLM_NSS_LOCKS          = 0,
+        LDLM_NSS_LAST
+};
+
+typedef enum {
+        /** invalide type */
+        LDLM_NS_TYPE_UNKNOWN    = 0,
+        /** mdc namespace */
+        LDLM_NS_TYPE_MDC,
+        /** mds namespace */
+        LDLM_NS_TYPE_MDT,
+        /** osc namespace */
+        LDLM_NS_TYPE_OSC,
+        /** ost namespace */
+        LDLM_NS_TYPE_OST,
+        /** mgc namespace */
+        LDLM_NS_TYPE_MGC,
+        /** mgs namespace */
+        LDLM_NS_TYPE_MGT,
+} ldlm_ns_type_t;
+
 struct ldlm_namespace {
-        char                  *ns_name;
-        __u32                  ns_client; /* is this a client-side lock tree? */
-        struct list_head      *ns_hash; /* hash table for ns */
-        spinlock_t             ns_hash_lock;
-        __u32                  ns_refcount; /* count of resources in the hash */
-        struct list_head       ns_root_list; /* all root resources in ns */
-        struct list_head       ns_list_chain; /* position in global NS list */
-
-        struct list_head       ns_unused_list; /* all root resources in ns */
+        /**
+         * Backward link to obd, required for ldlm pool to store new SLV.
+         */
+        struct obd_device     *ns_obd;
+
+        /**
+         * Is this a client-side lock tree?
+         */
+        ldlm_side_t            ns_client;
+
+        /**
+         * resource hash
+         */
+        cfs_hash_t            *ns_rs_hash;
+
+        /**
+         * serialize
+         */
+        cfs_spinlock_t         ns_lock;
+
+        /**
+         * big refcount (by bucket)
+         */
+        cfs_atomic_t           ns_bref;
+
+        /**
+         * Namespce connect flags supported by server (may be changed via proc,
+         * lru resize may be disabled/enabled).
+         */
+        __u64                  ns_connect_flags;
+
+         /**
+          * Client side orig connect flags supported by server.
+          */
+        __u64                  ns_orig_connect_flags;
+
+        /**
+         * Position in global namespace list.
+         */
+        cfs_list_t             ns_list_chain;
+
+        /**
+         * All root resources in namespace.
+         */
+        cfs_list_t             ns_unused_list;
         int                    ns_nr_unused;
-        spinlock_t             ns_unused_lock;
 
         unsigned int           ns_max_unused;
         unsigned int           ns_max_age;
-        cfs_time_t             ns_next_dump;   /* next debug dump, jiffies */
+        unsigned int           ns_timeouts;
+         /**
+          * Seconds.
+          */
+        unsigned int           ns_ctime_age_limit;
+
+        /**
+         * Next debug dump, jiffies.
+         */
+        cfs_time_t             ns_next_dump;
 
-        atomic_t               ns_locks;
-        __u64                  ns_resources;
         ldlm_res_policy        ns_policy;
         struct ldlm_valblock_ops *ns_lvbo;
         void                  *ns_lvbp;
         cfs_waitq_t            ns_waitq;
+        struct ldlm_pool       ns_pool;
+        ldlm_appetite_t        ns_appetite;
+
+        /**
+         * If more than \a ns_contended_locks found, the resource is considered
+         * to be contended.
+         */
+        unsigned               ns_contended_locks;
+
+        /**
+         * The resource remembers contended state during \a ns_contention_time,
+         * in seconds.
+         */
+        unsigned               ns_contention_time;
+
+        /**
+         * Limit size of nolock requests, in bytes.
+         */
+        unsigned               ns_max_nolock_size;
+
+        /**
+         * Limit of parallel AST RPC count.
+         */
+        unsigned               ns_max_parallel_ast;
+
+        /* callback to cancel locks before replaying it during recovery */
+        ldlm_cancel_for_recovery ns_cancel_for_recovery;
+        /**
+         * ldlm lock stats
+         */
+        struct lprocfs_stats  *ns_stats;
+
+        unsigned               ns_stopping:1;   /* namespace cleanup */
 };
 
-/*
- *
- * Resource hash table
- *
- */
+static inline int ns_is_client(struct ldlm_namespace *ns)
+{
+        LASSERT(ns != NULL);
+        LASSERT(!(ns->ns_client & ~(LDLM_NAMESPACE_CLIENT |
+                                    LDLM_NAMESPACE_SERVER)));
+        LASSERT(ns->ns_client == LDLM_NAMESPACE_CLIENT ||
+                ns->ns_client == LDLM_NAMESPACE_SERVER);
+        return ns->ns_client == LDLM_NAMESPACE_CLIENT;
+}
 
-#define RES_HASH_BITS 10
-#define RES_HASH_SIZE (1UL << RES_HASH_BITS)
-#define RES_HASH_MASK (RES_HASH_SIZE - 1)
+static inline int ns_is_server(struct ldlm_namespace *ns)
+{
+        LASSERT(ns != NULL);
+        LASSERT(!(ns->ns_client & ~(LDLM_NAMESPACE_CLIENT |
+                                    LDLM_NAMESPACE_SERVER)));
+        LASSERT(ns->ns_client == LDLM_NAMESPACE_CLIENT ||
+                ns->ns_client == LDLM_NAMESPACE_SERVER);
+        return ns->ns_client == LDLM_NAMESPACE_SERVER;
+}
+
+static inline int ns_connect_lru_resize(struct ldlm_namespace *ns)
+{
+        LASSERT(ns != NULL);
+        return !!(ns->ns_connect_flags & OBD_CONNECT_LRU_RESIZE);
+}
+
+static inline void ns_register_cancel(struct ldlm_namespace *ns,
+                                      ldlm_cancel_for_recovery arg)
+{
+        LASSERT(ns != NULL);
+        ns->ns_cancel_for_recovery = arg;
+}
 
 struct ldlm_lock;
 
@@ -254,114 +569,324 @@ typedef int (*ldlm_blocking_callback)(struct ldlm_lock *lock,
 typedef int (*ldlm_completion_callback)(struct ldlm_lock *lock, int flags,
                                         void *data);
 typedef int (*ldlm_glimpse_callback)(struct ldlm_lock *lock, void *data);
+typedef unsigned long (*ldlm_weigh_callback)(struct ldlm_lock *lock);
 
-struct ldlm_lock {
-        struct portals_handle l_handle; // must be first in the structure
-        atomic_t              l_refc;
-
-        /* internal spinlock protects l_resource.  we should hold this lock 
-         * first before grabbing res_lock.*/
-        spinlock_t            l_lock;
-
-        /* ldlm_lock_change_resource() can change this */
-        struct ldlm_resource *l_resource;
-
-        /* protected by ns_hash_lock. FIXME */
-        struct list_head      l_lru;
+/* Interval node data for each LDLM_EXTENT lock */
+struct ldlm_interval {
+        struct interval_node li_node;   /* node for tree mgmt */
+        cfs_list_t           li_group;  /* the locks which have the same
+                                         * policy - group of the policy */
+};
+#define to_ldlm_interval(n) container_of(n, struct ldlm_interval, li_node)
+
+/* the interval tree must be accessed inside the resource lock. */
+struct ldlm_interval_tree {
+        /* tree size, this variable is used to count
+         * granted PW locks in ldlm_extent_policy()*/
+        int                   lit_size;
+        ldlm_mode_t           lit_mode; /* lock mode */
+        struct interval_node *lit_root; /* actually ldlm_interval */
+};
 
-        /* protected by lr_lock */
-        struct list_head      l_res_link; // position in one of three res lists
+#define LUSTRE_TRACKS_LOCK_EXP_REFS (1)
 
-        struct list_head      l_sl_mode;        // skip pointer for request mode
-        struct list_head      l_sl_policy;      // skip pointer for inodebits
+/* Cancel flag. */
+typedef enum {
+        LCF_ASYNC      = 0x1, /* Cancel locks asynchronously. */
+        LCF_LOCAL      = 0x2, /* Cancel locks locally, not notifing server */
+        LCF_BL_AST     = 0x4, /* Cancel locks marked as LDLM_FL_BL_AST
+                               * in the same RPC */
+} ldlm_cancel_flags_t;
+
+struct ldlm_flock {
+        __u64 start;
+        __u64 end;
+        __u64 owner;
+        __u64 blocking_owner;
+        void *blocking_export;
+        __u32 pid;
+};
 
-        /* protected by led_lock */
-        struct list_head      l_export_chain; // per-export chain of locks
+typedef union {
+        struct ldlm_extent l_extent;
+        struct ldlm_flock l_flock;
+        struct ldlm_inodebits l_inodebits;
+} ldlm_policy_data_t;
 
-        /* protected by lr_lock */
-        ldlm_mode_t           l_req_mode;
-        ldlm_mode_t           l_granted_mode;
+void ldlm_convert_policy_to_wire(ldlm_type_t type,
+                                 const ldlm_policy_data_t *lpolicy,
+                                 ldlm_wire_policy_data_t *wpolicy);
+void ldlm_convert_policy_to_local(ldlm_type_t type,
+                                  const ldlm_wire_policy_data_t *wpolicy,
+                                  ldlm_policy_data_t *lpolicy);
 
+struct ldlm_lock {
+        /**
+         * Must be first in the structure.
+         */
+        struct portals_handle    l_handle;
+        /**
+         * Lock reference count.
+         */
+        cfs_atomic_t             l_refc;
+        /**
+         * Internal spinlock protects l_resource.  we should hold this lock
+         * first before grabbing res_lock.
+         */
+        cfs_spinlock_t           l_lock;
+        /**
+         * ldlm_lock_change_resource() can change this.
+         */
+        struct ldlm_resource    *l_resource;
+        /**
+         * Protected by ns_hash_lock. List item for client side lru list.
+         */
+        cfs_list_t               l_lru;
+        /**
+         * Protected by lr_lock, linkage to resource's lock queues.
+         */
+        cfs_list_t               l_res_link;
+        /**
+         * Tree node for ldlm_extent.
+         */
+        struct ldlm_interval    *l_tree_node;
+        /**
+         * Protected by per-bucket exp->exp_lock_hash locks. Per export hash
+         * of locks.
+         */
+        cfs_hlist_node_t         l_exp_hash;
+        /**
+         * Protected by lr_lock. Requested mode.
+         */
+        ldlm_mode_t              l_req_mode;
+        /**
+         * Granted mode, also protected by lr_lock.
+         */
+        ldlm_mode_t              l_granted_mode;
+        /**
+         * Lock enqueue completion handler.
+         */
         ldlm_completion_callback l_completion_ast;
+        /**
+         * Lock blocking ast handler.
+         */
         ldlm_blocking_callback   l_blocking_ast;
+        /**
+         * Lock glimpse handler.
+         */
         ldlm_glimpse_callback    l_glimpse_ast;
-
-        struct obd_export    *l_export;
-        struct obd_export    *l_conn_export;
-
-        struct lustre_handle  l_remote_handle;
-        ldlm_policy_data_t    l_policy_data;
-
-        /* protected by lr_lock */
-        __u32                 l_flags;
+        ldlm_weigh_callback      l_weigh_ast;
+
+        /**
+         * Lock export.
+         */
+        struct obd_export       *l_export;
+        /**
+         * Lock connection export.
+         */
+        struct obd_export       *l_conn_export;
+
+        /**
+         * Remote lock handle.
+         */
+        struct lustre_handle     l_remote_handle;
+
+        ldlm_policy_data_t       l_policy_data;
+
+        /*
+         * Protected by lr_lock. Various counters: readers, writers, etc.
+         */
+        __u64                 l_flags;
         __u32                 l_readers;
         __u32                 l_writers;
-        __u8                  l_destroyed;
-
-        /* If the lock is granted, a process sleeps on this waitq to learn when
+        /**
+         * If the lock is granted, a process sleeps on this waitq to learn when
          * it's no longer in use.  If the lock is not granted, a process sleeps
-         * on this waitq to learn when it becomes granted. */
+         * on this waitq to learn when it becomes granted.
+         */
         cfs_waitq_t           l_waitq;
-        struct timeval        l_enqueued_time;
 
-        cfs_time_t            l_last_used;      /* jiffies */
+        /** 
+         * Seconds. it will be updated if there is any activity related to 
+         * the lock, e.g. enqueue the lock or send block AST.
+         */
+        cfs_time_t            l_last_activity;
+
+        /**
+         * Jiffies. Should be converted to time if needed.
+         */
+        cfs_time_t            l_last_used;
+
         struct ldlm_extent    l_req_extent;
 
-        /* Client-side-only members */
-        __u32                 l_lvb_len;        /* temporary storage for */
-        void                 *l_lvb_data;       /* an LVB received during */
-        void                 *l_lvb_swabber;    /* an enqueue */
+        unsigned int          l_failed:1,
+        /*
+         * Set for locks that were removed from class hash table and will be
+         * destroyed when last reference to them is released. Set by
+         * ldlm_lock_destroy_internal().
+         *
+         * Protected by lock and resource locks.
+         */
+                              l_destroyed:1,
+        /**
+         * flag whether this is a server namespace lock.
+         */
+                              l_ns_srv:1;
+
+        /*
+         * Client-side-only members.
+         */
+
+        /**
+         * Temporary storage for an LVB received during an enqueue operation.
+         */
+        __u32                 l_lvb_len;
+        void                 *l_lvb_data;
+
         void                 *l_ast_data;
 
-        /* Server-side-only members */
+        /*
+         * Server-side-only members.
+         */
 
-        /* protected by elt_lock */
-        struct list_head      l_pending_chain;  /* callbacks pending */
-        cfs_time_t            l_callback_timeout; /* jiffies */
+        /** connection cookie for the client originated the operation. */
+        __u64                 l_client_cookie;
 
-        __u32                 l_pid;            /* pid which created this lock */
+        /**
+         * Protected by elt_lock. Callbacks pending.
+         */
+        cfs_list_t            l_pending_chain;
+
+        cfs_time_t            l_callback_timeout;
+
+        /**
+         * Pid which created this lock.
+         */
+        __u32                 l_pid;
 
-        /* for ldlm_add_ast_work_item() */
-        struct list_head      l_bl_ast;
-        struct list_head      l_cp_ast;
-        struct ldlm_lock     *l_blocking_lock;
         int                   l_bl_ast_run;
+        /**
+         * For ldlm_add_ast_work_item().
+         */
+        cfs_list_t            l_bl_ast;
+        /**
+         * For ldlm_add_ast_work_item().
+         */
+        cfs_list_t            l_cp_ast;
+        /**
+         * For ldlm_add_ast_work_item().
+         */
+        cfs_list_t            l_rk_ast;
+
+        struct ldlm_lock     *l_blocking_lock;
+
+        /**
+         * Protected by lr_lock, linkages to "skip lists".
+         */
+        cfs_list_t            l_sl_mode;
+        cfs_list_t            l_sl_policy;
+        struct lu_ref         l_reference;
+#if LUSTRE_TRACKS_LOCK_EXP_REFS
+        /* Debugging stuff for bug 20498, for tracking export
+           references. */
+        /** number of export references taken */
+        int                   l_exp_refs_nr;
+        /** link all locks referencing one export */
+        cfs_list_t            l_exp_refs_link;
+        /** referenced export object */
+        struct obd_export    *l_exp_refs_target;
+#endif
+        /** export blocking dlm lock list, protected by
+         * l_export->exp_bl_list_lock.
+         * Lock order of waiting_lists_spinlock, exp_bl_list_lock and res lock
+         * is: res lock -> exp_bl_list_lock -> wanting_lists_spinlock. */
+        cfs_list_t            l_exp_list;
 };
 
 struct ldlm_resource {
-        struct ldlm_namespace *lr_namespace;
+        struct ldlm_ns_bucket *lr_ns_bucket;
 
         /* protected by ns_hash_lock */
-        struct list_head       lr_hash;
-        struct ldlm_resource  *lr_parent;   /* 0 for a root resource */
-        struct list_head       lr_children; /* list head for child resources */
-        struct list_head       lr_childof;  /* part of ns_root_list if root res,
-                                             * part of lr_children if child */
-        spinlock_t             lr_lock;
+        cfs_hlist_node_t       lr_hash;
+        cfs_spinlock_t         lr_lock;
 
         /* protected by lr_lock */
-        struct list_head       lr_granted;
-        struct list_head       lr_converting;
-        struct list_head       lr_waiting;
+        cfs_list_t             lr_granted;
+        cfs_list_t             lr_converting;
+        cfs_list_t             lr_waiting;
         ldlm_mode_t            lr_most_restr;
         ldlm_type_t            lr_type; /* LDLM_{PLAIN,EXTENT,FLOCK} */
         struct ldlm_res_id     lr_name;
-        atomic_t               lr_refcount;
+        cfs_atomic_t           lr_refcount;
+
+        struct ldlm_interval_tree lr_itree[LCK_MODE_NUM];  /* interval trees*/
 
         /* Server-side-only lock value block elements */
-        struct semaphore       lr_lvb_sem;
+        /** to serialize lvbo_init */
+        cfs_semaphore_t        lr_lvb_sem;
         __u32                  lr_lvb_len;
+        /** protect by lr_lock */
         void                  *lr_lvb_data;
+
+        /* when the resource was considered as contended */
+        cfs_time_t             lr_contention_time;
+        /**
+         * List of references to this resource. For debugging.
+         */
+        struct lu_ref          lr_reference;
+
+        struct inode          *lr_lvb_inode;
 };
 
+static inline char *
+ldlm_ns_name(struct ldlm_namespace *ns)
+{
+        return ns->ns_rs_hash->hs_name;
+}
+
+static inline struct ldlm_namespace *
+ldlm_res_to_ns(struct ldlm_resource *res)
+{
+        return res->lr_ns_bucket->nsb_namespace;
+}
+
+static inline struct ldlm_namespace *
+ldlm_lock_to_ns(struct ldlm_lock *lock)
+{
+        return ldlm_res_to_ns(lock->l_resource);
+}
+
+static inline char *
+ldlm_lock_to_ns_name(struct ldlm_lock *lock)
+{
+        return ldlm_ns_name(ldlm_lock_to_ns(lock));
+}
+
+static inline struct adaptive_timeout *
+ldlm_lock_to_ns_at(struct ldlm_lock *lock)
+{
+        return &lock->l_resource->lr_ns_bucket->nsb_at_estimate;
+}
+
 struct ldlm_ast_work {
-        struct ldlm_lock *w_lock;
-        int               w_blocking;
-        struct ldlm_lock_desc w_desc;
-        struct list_head   w_list;
-        int w_flags;
-        void *w_data;
-        int w_datalen;
+        struct ldlm_lock      *w_lock;
+        int                    w_blocking;
+        struct ldlm_lock_desc  w_desc;
+        cfs_list_t             w_list;
+        int                    w_flags;
+        void                  *w_data;
+        int                    w_datalen;
+};
+
+/* ldlm_enqueue parameters common */
+struct ldlm_enqueue_info {
+        __u32 ei_type;   /* Type of the lock being enqueued. */
+        __u32 ei_mode;   /* Mode of the lock being enqueued. */
+        void *ei_cb_bl;  /* blocking lock callback */
+        void *ei_cb_cp;  /* lock completion callback */
+        void *ei_cb_gl;  /* lock glimpse callback */
+        void *ei_cb_wg;  /* lock weigh callback */
+        void *ei_cbdata; /* Data to be passed into callbacks. */
 };
 
 extern struct obd_ops ldlm_obd_ops;
@@ -371,7 +896,7 @@ extern char *ldlm_typename[];
 extern char *ldlm_it2str(int it);
 #ifdef LIBCFS_DEBUG
 #define ldlm_lock_debug(cdls, level, lock, file, func, line, fmt, a...) do { \
-        CHECK_STACK();                                                  \
+        CFS_CHECK_STACK();                                              \
                                                                         \
         if (((level) & D_CANTMASK) != 0 ||                              \
             ((libcfs_debug & (level)) != 0 &&                           \
@@ -389,21 +914,26 @@ void _ldlm_lock_debug(struct ldlm_lock *lock, __u32 mask,
                       ...)
         __attribute__ ((format (printf, 4, 5)));
 
-#define LDLM_ERROR(lock, fmt, a...) do {                                \
+#define LDLM_DEBUG_LIMIT(mask, lock, fmt, a...) do {                    \
         static cfs_debug_limit_state_t _ldlm_cdls;                      \
-        ldlm_lock_debug(&_ldlm_cdls, D_ERROR, lock,                     \
+        ldlm_lock_debug(&_ldlm_cdls, mask, lock,                        \
                         __FILE__, __FUNCTION__, __LINE__,               \
                         "### " fmt , ##a);                              \
 } while (0)
 
+#define LDLM_ERROR(lock, fmt, a...) LDLM_DEBUG_LIMIT(D_ERROR, lock, fmt, ## a)
+#define LDLM_WARN(lock, fmt, a...)  LDLM_DEBUG_LIMIT(D_WARNING, lock, fmt, ## a)
+
 #define LDLM_DEBUG(lock, fmt, a...)   do {                              \
         ldlm_lock_debug(NULL, D_DLMTRACE, lock,                         \
                         __FILE__, __FUNCTION__, __LINE__,               \
                          "### " fmt , ##a);                             \
 } while (0)
-#else
-#define LDLM_DEBUG(lock, fmt, a...) ((void)0)
-#define LDLM_ERROR(lock, fmt, a...) ((void)0)
+#else /* !LIBCFS_DEBUG */
+# define LDLM_DEBUG(lock, fmt, a...) ((void)0)
+# define LDLM_ERROR(lock, fmt, a...) ((void)0)
+# define ldlm_lock_debuf(cdls, level, lock, file, func, line, fmt, a...) \
+         ((void)0)
 #endif
 
 #define LDLM_DEBUG_NOLOCK(format, a...)                 \
@@ -411,7 +941,7 @@ void _ldlm_lock_debug(struct ldlm_lock *lock, __u32 mask,
 
 typedef int (*ldlm_processing_policy)(struct ldlm_lock *lock, int *flags,
                                       int first_enq, ldlm_error_t *err,
-                                      struct list_head *work_list);
+                                      cfs_list_t *work_list);
 
 /*
  * Iterators.
@@ -425,16 +955,13 @@ typedef int (*ldlm_res_iterator_t)(struct ldlm_resource *, void *);
 
 int ldlm_resource_foreach(struct ldlm_resource *res, ldlm_iterator_t iter,
                           void *closure);
-int ldlm_namespace_foreach(struct ldlm_namespace *ns, ldlm_iterator_t iter,
-                           void *closure);
-int ldlm_namespace_foreach_res(struct ldlm_namespace *ns,
-                               ldlm_res_iterator_t iter, void *closure);
+void ldlm_namespace_foreach(struct ldlm_namespace *ns, ldlm_iterator_t iter,
+                            void *closure);
 
 int ldlm_replay_locks(struct obd_import *imp);
-void ldlm_resource_iterate(struct ldlm_namespace *, const struct ldlm_res_id *,
+int ldlm_resource_iterate(struct ldlm_namespace *, const struct ldlm_res_id *,
                            ldlm_iterator_t iter, void *data);
 
-
 /* ldlm_flock.c */
 int ldlm_flock_completion_ast(struct ldlm_lock *lock, int flags, void *data);
 
@@ -454,10 +981,12 @@ int ldlm_handle_cancel(struct ptlrpc_request *req);
 int ldlm_request_cancel(struct ptlrpc_request *req,
                         const struct ldlm_request *dlm_req, int first);
 int ldlm_del_waiting_lock(struct ldlm_lock *lock);
-int ldlm_refresh_waiting_lock(struct ldlm_lock *lock);
+int ldlm_refresh_waiting_lock(struct ldlm_lock *lock, int timeout);
 void ldlm_revoke_export_locks(struct obd_export *exp);
 int ldlm_get_ref(void);
-void ldlm_put_ref(int force);
+void ldlm_put_ref(void);
+int ldlm_init_export(struct obd_export *exp);
+void ldlm_destroy_export(struct obd_export *exp);
 
 /* ldlm_lock.c */
 ldlm_processing_policy ldlm_get_processing_policy(struct ldlm_resource *res);
@@ -466,18 +995,63 @@ void ldlm_lock2handle(const struct ldlm_lock *lock,
                       struct lustre_handle *lockh);
 struct ldlm_lock *__ldlm_handle2lock(const struct lustre_handle *, int flags);
 void ldlm_cancel_callback(struct ldlm_lock *);
-int ldlm_lock_set_data(struct lustre_handle *, void *data);
 int ldlm_lock_remove_from_lru(struct ldlm_lock *);
-struct ldlm_lock *ldlm_handle2lock_ns(struct ldlm_namespace *,
-                                      const struct lustre_handle *);
+int ldlm_lock_set_data(struct lustre_handle *, void *);
 
 static inline struct ldlm_lock *ldlm_handle2lock(const struct lustre_handle *h)
 {
         return __ldlm_handle2lock(h, 0);
 }
 
+#define LDLM_LOCK_REF_DEL(lock) \
+        lu_ref_del(&lock->l_reference, "handle", cfs_current())
+
+static inline struct ldlm_lock *
+ldlm_handle2lock_long(const struct lustre_handle *h, int flags)
+{
+        struct ldlm_lock *lock;
+
+        lock = __ldlm_handle2lock(h, flags);
+        if (lock != NULL)
+                LDLM_LOCK_REF_DEL(lock);
+        return lock;
+}
+
+static inline int ldlm_res_lvbo_update(struct ldlm_resource *res,
+                                       struct ptlrpc_request *r, int increase)
+{
+        if (ldlm_res_to_ns(res)->ns_lvbo &&
+            ldlm_res_to_ns(res)->ns_lvbo->lvbo_update) {
+                return ldlm_res_to_ns(res)->ns_lvbo->lvbo_update(res, r,
+                                                                 increase);
+        }
+        return 0;
+}
+
+int ldlm_error2errno(ldlm_error_t error);
+ldlm_error_t ldlm_errno2error(int err_no); /* don't call it `errno': this
+                                            * confuses user-space. */
+#if LUSTRE_TRACKS_LOCK_EXP_REFS
+void ldlm_dump_export_locks(struct obd_export *exp);
+#endif
+
+/**
+ * Release a temporary lock reference obtained by ldlm_handle2lock() or
+ * __ldlm_handle2lock().
+ */
 #define LDLM_LOCK_PUT(lock)                     \
 do {                                            \
+        LDLM_LOCK_REF_DEL(lock);                \
+        /*LDLM_DEBUG((lock), "put");*/          \
+        ldlm_lock_put(lock);                    \
+} while (0)
+
+/**
+ * Release a lock reference obtained by some other means (see
+ * LDLM_LOCK_PUT()).
+ */
+#define LDLM_LOCK_RELEASE(lock)                 \
+do {                                            \
         /*LDLM_DEBUG((lock), "put");*/          \
         ldlm_lock_put(lock);                    \
 } while (0)
@@ -489,16 +1063,17 @@ do {                                            \
         lock;                                   \
 })
 
-#define ldlm_lock_list_put(head, member, count)                 \
-({                                                              \
-        struct ldlm_lock *_lock, *_next;                        \
-        int c = count;                                          \
-        list_for_each_entry_safe(_lock, _next, head, member) {  \
-                list_del_init(&_lock->member);                  \
-                LDLM_LOCK_PUT(_lock);                           \
-                if (--c == 0)                                   \
-                        break;                                  \
-        }                                                       \
+#define ldlm_lock_list_put(head, member, count)                     \
+({                                                                  \
+        struct ldlm_lock *_lock, *_next;                            \
+        int c = count;                                              \
+        cfs_list_for_each_entry_safe(_lock, _next, head, member) {  \
+                if (c-- == 0)                                       \
+                        break;                                      \
+                cfs_list_del_init(&_lock->member);                  \
+                LDLM_LOCK_RELEASE(_lock);                           \
+        }                                                           \
+        LASSERT(c <= 0);                                            \
 })
 
 struct ldlm_lock *ldlm_lock_get(struct ldlm_lock *lock);
@@ -506,17 +1081,23 @@ void ldlm_lock_put(struct ldlm_lock *lock);
 void ldlm_lock_destroy(struct ldlm_lock *lock);
 void ldlm_lock2desc(struct ldlm_lock *lock, struct ldlm_lock_desc *desc);
 void ldlm_lock_addref(struct lustre_handle *lockh, __u32 mode);
+int  ldlm_lock_addref_try(struct lustre_handle *lockh, __u32 mode);
 void ldlm_lock_decref(struct lustre_handle *lockh, __u32 mode);
 void ldlm_lock_decref_and_cancel(struct lustre_handle *lockh, __u32 mode);
+void ldlm_lock_fail_match_locked(struct ldlm_lock *lock);
+void ldlm_lock_fail_match(struct ldlm_lock *lock);
 void ldlm_lock_allow_match(struct ldlm_lock *lock);
-int ldlm_lock_match(struct ldlm_namespace *ns, int flags,
-                    const struct ldlm_res_id *,
-                    ldlm_type_t type, ldlm_policy_data_t *, ldlm_mode_t mode,
-                    struct lustre_handle *);
+void ldlm_lock_allow_match_locked(struct ldlm_lock *lock);
+ldlm_mode_t ldlm_lock_match(struct ldlm_namespace *ns, int flags,
+                            const struct ldlm_res_id *, ldlm_type_t type,
+                            ldlm_policy_data_t *, ldlm_mode_t mode,
+                            struct lustre_handle *, int unref);
+ldlm_mode_t ldlm_revalidate_lock_handle(struct lustre_handle *lockh,
+                                        __u64 *bits);
 struct ldlm_resource *ldlm_lock_convert(struct ldlm_lock *lock, int new_mode,
-                                        int *flags);
+                                        __u32 *flags);
+void ldlm_lock_downgrade(struct ldlm_lock *lock, int new_mode);
 void ldlm_lock_cancel(struct ldlm_lock *lock);
-void ldlm_cancel_locks_for_export(struct obd_export *export);
 void ldlm_reprocess_all(struct ldlm_resource *res);
 void ldlm_reprocess_all_ns(struct ldlm_namespace *ns);
 void ldlm_lock_dump(int level, struct ldlm_lock *lock, int pos);
@@ -524,9 +1105,19 @@ void ldlm_lock_dump_handle(int level, struct lustre_handle *);
 void ldlm_unlink_lock_skiplist(struct ldlm_lock *req);
 
 /* resource.c */
-struct ldlm_namespace *ldlm_namespace_new(char *name, __u32 local);
+struct ldlm_namespace *
+ldlm_namespace_new(struct obd_device *obd, char *name,
+                   ldlm_side_t client, ldlm_appetite_t apt,
+                   ldlm_ns_type_t ns_type);
 int ldlm_namespace_cleanup(struct ldlm_namespace *ns, int flags);
-int ldlm_namespace_free(struct ldlm_namespace *ns, int force);
+void ldlm_namespace_free(struct ldlm_namespace *ns,
+                         struct obd_import *imp, int force);
+void ldlm_namespace_register(struct ldlm_namespace *ns, ldlm_side_t client);
+void ldlm_namespace_unregister(struct ldlm_namespace *ns, ldlm_side_t client);
+void ldlm_namespace_move_locked(struct ldlm_namespace *ns, ldlm_side_t client);
+struct ldlm_namespace *ldlm_namespace_first_locked(ldlm_side_t client);
+void ldlm_namespace_get(struct ldlm_namespace *ns);
+void ldlm_namespace_put(struct ldlm_namespace *ns);
 int ldlm_proc_setup(void);
 #ifdef LPROCFS
 void ldlm_proc_cleanup(void);
@@ -541,48 +1132,61 @@ struct ldlm_resource *ldlm_resource_get(struct ldlm_namespace *ns,
                                         ldlm_type_t type, int create);
 struct ldlm_resource *ldlm_resource_getref(struct ldlm_resource *res);
 int ldlm_resource_putref(struct ldlm_resource *res);
-void ldlm_resource_add_lock(struct ldlm_resource *res, struct list_head *head,
+void ldlm_resource_add_lock(struct ldlm_resource *res,
+                            cfs_list_t *head,
                             struct ldlm_lock *lock);
 void ldlm_resource_unlink_lock(struct ldlm_lock *lock);
 void ldlm_res2desc(struct ldlm_resource *res, struct ldlm_resource_desc *desc);
-void ldlm_dump_all_namespaces(int level);
+void ldlm_dump_all_namespaces(ldlm_side_t client, int level);
 void ldlm_namespace_dump(int level, struct ldlm_namespace *);
 void ldlm_resource_dump(int level, struct ldlm_resource *);
 int ldlm_lock_change_resource(struct ldlm_namespace *, struct ldlm_lock *,
                               const struct ldlm_res_id *);
 
+#define LDLM_RESOURCE_ADDREF(res) do {                                  \
+        lu_ref_add_atomic(&(res)->lr_reference, __FUNCTION__, cfs_current());  \
+} while (0)
+
+#define LDLM_RESOURCE_DELREF(res) do {                                  \
+        lu_ref_del(&(res)->lr_reference, __FUNCTION__, cfs_current());  \
+} while (0)
+
 struct ldlm_callback_suite {
         ldlm_completion_callback lcs_completion;
         ldlm_blocking_callback   lcs_blocking;
         ldlm_glimpse_callback    lcs_glimpse;
+        ldlm_weigh_callback      lcs_weigh;
 };
 
 /* ldlm_request.c */
 int ldlm_expired_completion_wait(void *data);
+int ldlm_blocking_ast_nocheck(struct ldlm_lock *lock);
 int ldlm_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc,
                       void *data, int flag);
 int ldlm_glimpse_ast(struct ldlm_lock *lock, void *reqp);
+int ldlm_completion_ast_async(struct ldlm_lock *lock, int flags, void *data);
 int ldlm_completion_ast(struct ldlm_lock *lock, int flags, void *data);
 int ldlm_cli_enqueue(struct obd_export *exp, struct ptlrpc_request **reqp,
+                     struct ldlm_enqueue_info *einfo,
                      const struct ldlm_res_id *res_id,
-                     ldlm_type_t type, ldlm_policy_data_t *policy,
-                     ldlm_mode_t mode, int *flags,
-                     ldlm_blocking_callback blocking,
-                     ldlm_completion_callback completion,
-                     ldlm_glimpse_callback glimpse,
-                     void *data, void *lvb, __u32 lvb_len, void *lvb_swabber,
-                     struct lustre_handle *lockh, int async);
-struct ptlrpc_request *ldlm_prep_enqueue_req(struct obd_export *exp,
-                                             int bufcount, int *size,
-                                             struct list_head *head, int count);
+                     ldlm_policy_data_t const *policy, int *flags,
+                     void *lvb, __u32 lvb_len, struct lustre_handle *lockh,
+                     int async);
+int ldlm_prep_enqueue_req(struct obd_export *exp,
+                          struct ptlrpc_request *req,
+                          cfs_list_t *cancels,
+                          int count);
+int ldlm_prep_elc_req(struct obd_export *exp,
+                      struct ptlrpc_request *req,
+                      int version, int opc, int canceloff,
+                      cfs_list_t *cancels, int count);
 int ldlm_handle_enqueue0(struct ldlm_namespace *ns, struct ptlrpc_request *req,
                          const struct ldlm_request *dlm_req,
                          const struct ldlm_callback_suite *cbs);
 int ldlm_cli_enqueue_fini(struct obd_export *exp, struct ptlrpc_request *req,
                           ldlm_type_t type, __u8 with_policy, ldlm_mode_t mode,
                           int *flags, void *lvb, __u32 lvb_len,
-                          void *lvb_swabber, struct lustre_handle *lockh,
-                          int rc);
+                          struct lustre_handle *lockh, int rc);
 int ldlm_cli_enqueue_local(struct ldlm_namespace *ns,
                            const struct ldlm_res_id *res_id,
                            ldlm_type_t type, ldlm_policy_data_t *policy,
@@ -590,32 +1194,36 @@ int ldlm_cli_enqueue_local(struct ldlm_namespace *ns,
                            ldlm_blocking_callback blocking,
                            ldlm_completion_callback completion,
                            ldlm_glimpse_callback glimpse,
-                           void *data, __u32 lvb_len, void *lvb_swabber,
+                           void *data, __u32 lvb_len,
+                           const __u64 *client_cookie,
                            struct lustre_handle *lockh);
 int ldlm_server_ast(struct lustre_handle *lockh, struct ldlm_lock_desc *new,
                     void *data, __u32 data_len);
-int ldlm_cli_convert(struct lustre_handle *, int new_mode, int *flags);
+int ldlm_cli_convert(struct lustre_handle *, int new_mode, __u32 *flags);
+int ldlm_cli_update_pool(struct ptlrpc_request *req);
 int ldlm_handle_convert0(struct ptlrpc_request *req,
                          const struct ldlm_request *dlm_req);
 int ldlm_cli_cancel(struct lustre_handle *lockh);
 int ldlm_cli_cancel_unused(struct ldlm_namespace *, const struct ldlm_res_id *,
-                           int flags, void *opaque);
+                           ldlm_cancel_flags_t flags, void *opaque);
 int ldlm_cli_cancel_unused_resource(struct ldlm_namespace *ns,
                                     const struct ldlm_res_id *res_id,
                                     ldlm_policy_data_t *policy,
-                                    int mode, int flags, void *opaque);
-int ldlm_cli_cancel_req(struct obd_export *exp, struct list_head *head,
-                        int count, int flags);
-int ldlm_cli_join_lru(struct ldlm_namespace *,
-                      const struct ldlm_res_id *, int join);
+                                    ldlm_mode_t mode,
+                                    ldlm_cancel_flags_t flags,
+                                    void *opaque);
+int ldlm_cli_cancel_req(struct obd_export *exp, cfs_list_t *head,
+                        int count, ldlm_cancel_flags_t flags);
 int ldlm_cancel_resource_local(struct ldlm_resource *res,
-                               struct list_head *cancels,
+                               cfs_list_t *cancels,
                                ldlm_policy_data_t *policy,
                                ldlm_mode_t mode, int lock_flags,
-                               int flags, void *opaque);
-int ldlm_cli_cancel_list(struct list_head *head, int count,
-                         struct ptlrpc_request *req, int off, int flags);
+                               ldlm_cancel_flags_t cancel_flags, void *opaque);
+int ldlm_cli_cancel_list_local(cfs_list_t *cancels, int count,
+                               ldlm_cancel_flags_t flags);
+int ldlm_cli_cancel_list(cfs_list_t *head, int count,
+                         struct ptlrpc_request *req, ldlm_cancel_flags_t flags);
+
 /* mds/handler.c */
 /* This has to be here because recursive inclusion sucks. */
 int intent_disposition(struct ldlm_reply *rep, int flag);
@@ -632,14 +1240,30 @@ void intent_set_disposition(struct ldlm_reply *rep, int flag);
 #define IOC_LDLM_REGRESS_STOP           _IOWR('f', 43, long)
 #define IOC_LDLM_MAX_NR                 43
 
+/**
+ * "Modes" of acquiring lock_res, necessary to tell lockdep that taking more
+ * than one lock_res is dead-lock safe.
+ */
+enum lock_res_type {
+        LRT_NORMAL,
+        LRT_NEW
+};
+
 static inline void lock_res(struct ldlm_resource *res)
 {
-        spin_lock(&res->lr_lock);
+        cfs_spin_lock(&res->lr_lock);
 }
 
+static inline void lock_res_nested(struct ldlm_resource *res,
+                                   enum lock_res_type mode)
+{
+        cfs_spin_lock_nested(&res->lr_lock, mode);
+}
+
+
 static inline void unlock_res(struct ldlm_resource *res)
 {
-        spin_unlock(&res->lr_lock);
+        cfs_spin_unlock(&res->lr_lock);
 }
 
 static inline void check_res_locked(struct ldlm_resource *res)
@@ -650,4 +1274,28 @@ static inline void check_res_locked(struct ldlm_resource *res)
 struct ldlm_resource * lock_res_and_lock(struct ldlm_lock *lock);
 void unlock_res_and_lock(struct ldlm_lock *lock);
 
+/* ldlm_pool.c */
+void ldlm_pools_recalc(ldlm_side_t client);
+int ldlm_pools_init(void);
+void ldlm_pools_fini(void);
+
+int ldlm_pool_init(struct ldlm_pool *pl, struct ldlm_namespace *ns,
+                   int idx, ldlm_side_t client);
+int ldlm_pool_shrink(struct ldlm_pool *pl, int nr,
+                     unsigned int gfp_mask);
+void ldlm_pool_fini(struct ldlm_pool *pl);
+int ldlm_pool_setup(struct ldlm_pool *pl, int limit);
+int ldlm_pool_recalc(struct ldlm_pool *pl);
+__u32 ldlm_pool_get_lvf(struct ldlm_pool *pl);
+__u64 ldlm_pool_get_slv(struct ldlm_pool *pl);
+__u64 ldlm_pool_get_clv(struct ldlm_pool *pl);
+__u32 ldlm_pool_get_limit(struct ldlm_pool *pl);
+void ldlm_pool_set_slv(struct ldlm_pool *pl, __u64 slv);
+void ldlm_pool_set_clv(struct ldlm_pool *pl, __u64 clv);
+void ldlm_pool_set_limit(struct ldlm_pool *pl, __u32 limit);
+void ldlm_pool_add(struct ldlm_pool *pl, struct ldlm_lock *lock);
+void ldlm_pool_del(struct ldlm_pool *pl, struct ldlm_lock *lock);
+
+/** @} ldlm */
+
 #endif