Whamcloud - gitweb
Land b1_8_gate onto b1_8 (20081218_1708)
[fs/lustre-release.git] / lustre / ldlm / ldlm_request.c
index 75963f9..e318e88 100644 (file)
@@ -1,25 +1,37 @@
 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
  * vim:expandtab:shiftwidth=8:tabstop=8:
  *
- *  Copyright (C) 2002, 2003 Cluster File Systems, Inc.
+ * GPL HEADER START
  *
- *   This file is part of the Lustre file system, http://www.lustre.org
- *   Lustre is a trademark of Cluster File Systems, Inc.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
- *   You may have signed or agreed to another license before downloading
- *   this software.  If so, you are bound by the terms and conditions
- *   of that agreement, and the following does not apply to you.  See the
- *   LICENSE file included with this distribution for more information.
+ * 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.
  *
- *   If you did not agree to a different license, then this copy of Lustre
- *   is open source software; you can redistribute it and/or modify it
- *   under the terms of version 2 of the GNU General Public License 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).
  *
- *   In either case, Lustre 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
- *   license text for more details.
+ * 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  2008 Sun Microsystems, Inc. All rights reserved
+ * Use is subject to license terms.
+ */
+/*
+ * This file is part of Lustre, http://www.lustre.org/
+ * Lustre is a trademark of Sun Microsystems, Inc.
  */
 
 #define DEBUG_SUBSYSTEM S_LDLM
@@ -102,6 +114,7 @@ int ldlm_get_enq_timeout(struct ldlm_lock *lock)
         timeout = timeout + (timeout >> 1); /* 150% */
         return max(timeout, ldlm_enqueue_min);
 }
+EXPORT_SYMBOL(ldlm_get_enq_timeout);
 
 static int is_granted_or_cancelled(struct ldlm_lock *lock)
 {
@@ -175,8 +188,17 @@ noreproc:
                 spin_unlock(&imp->imp_lock);
         }
 
-        /* Go to sleep until the lock is granted or cancelled. */
-        rc = l_wait_event(lock->l_waitq, is_granted_or_cancelled(lock), &lwi);
+        if (ns_is_client(lock->l_resource->lr_namespace) && 
+            lock->l_resource->lr_type == LDLM_EXTENT &&
+            OBD_FAIL_CHECK(OBD_FAIL_LDLM_INTR_CP_AST | OBD_FAIL_ONCE)) {
+                obd_fail_loc = OBD_FAIL_LDLM_CP_BL_RACE | OBD_FAIL_ONCE;
+                lock->l_flags |= LDLM_FL_FAIL_LOC;
+                rc = -EINTR;
+        } else {
+                /* Go to sleep until the lock is granted or cancelled. */
+                rc = l_wait_event(lock->l_waitq, 
+                                  is_granted_or_cancelled(lock), &lwi);
+        }
 
         if (lock->l_destroyed || lock->l_flags & LDLM_FL_FAILED) {
                 LDLM_DEBUG(lock, "client-side enqueue waking up: destroyed");
@@ -337,13 +359,31 @@ static void failed_lock_cleanup(struct ldlm_namespace *ns,
                                 struct ldlm_lock *lock,
                                 struct lustre_handle *lockh, int mode)
 {
+        int need_cancel = 0;
+
         /* Set a flag to prevent us from sending a CANCEL (bug 407) */
         lock_res_and_lock(lock);
-        lock->l_flags |= LDLM_FL_LOCAL_ONLY;
+        /* Check that lock is not granted or failed, we might race. */
+        if ((lock->l_req_mode != lock->l_granted_mode) && 
+            !(lock->l_flags & LDLM_FL_FAILED)) {
+                /* Make sure that this lock will not be found by raced
+                 * bl_ast and -EINVAL reply is sent to server anyways. 
+                 * bug 17645 */
+                lock->l_flags |= LDLM_FL_LOCAL_ONLY | LDLM_FL_FAILED | 
+                                 LDLM_FL_ATOMIC_CB;
+                need_cancel = 1;
+        }
         unlock_res_and_lock(lock);
-        LDLM_DEBUG(lock, "setting FL_LOCAL_ONLY");
-
-        ldlm_lock_decref_and_cancel(lockh, mode);
+  
+        if (need_cancel) {
+                LDLM_DEBUG(lock, 
+                           "setting FL_LOCAL_ONLY | LDLM_FL_FAILED | " 
+                           "LDLM_FL_ATOMIC_CB");
+                ldlm_lock_decref_and_cancel(lockh, mode);
+        } else {
+                LDLM_DEBUG(lock, "lock was granted or failed in race");
+                ldlm_lock_decref(lockh, mode);
+        }
 
         /* XXX - HACK because we shouldn't call ldlm_lock_destroy()
          *       from llite/file.c/ll_file_flock(). */
@@ -359,6 +399,7 @@ int ldlm_cli_enqueue_fini(struct obd_export *exp, struct ptlrpc_request *req,
 {
         struct ldlm_namespace *ns = exp->exp_obd->obd_namespace;
         int is_replay = *flags & LDLM_FL_REPLAY;
+        struct lustre_handle old_hash_key;
         struct ldlm_lock *lock;
         struct ldlm_reply *reply;
         int cleanup_phase = 1;
@@ -410,7 +451,15 @@ int ldlm_cli_enqueue_fini(struct obd_export *exp, struct ptlrpc_request *req,
         cleanup_phase = 0;
 
         lock_res_and_lock(lock);
+        old_hash_key = lock->l_remote_handle;
         lock->l_remote_handle = reply->lock_handle;
+
+        /* Key change rehash lock in per-export hash with new key */
+        if (exp->exp_lock_hash)
+                lustre_hash_rehash_key(exp->exp_lock_hash, &old_hash_key,
+                                       &lock->l_remote_handle,
+                                       &lock->l_exp_hash);
+
         *flags = reply->lock_flags;
         lock->l_flags |= reply->lock_flags & LDLM_INHERIT_FLAGS;
         /* move NO_TIMEOUT flag to the lock to force ldlm_lock_match()
@@ -433,12 +482,18 @@ int ldlm_cli_enqueue_fini(struct obd_export *exp, struct ptlrpc_request *req,
                         lock->l_req_mode = newmode;
                 }
 
-                if (reply->lock_desc.l_resource.lr_name.name[0] !=
-                    lock->l_resource->lr_name.name[0]) {
-                        CDEBUG(D_INFO, "remote intent success, locking %ld "
-                               "instead of %ld\n",
-                              (long)reply->lock_desc.l_resource.lr_name.name[0],
-                               (long)lock->l_resource->lr_name.name[0]);
+                if (memcmp(reply->lock_desc.l_resource.lr_name.name,
+                          lock->l_resource->lr_name.name,
+                          sizeof(struct ldlm_res_id))) {
+                        CDEBUG(D_INFO, "remote intent success, locking "
+                                        "("LPU64"/"LPU64"/"LPU64") instead of "
+                                        "("LPU64"/"LPU64"/"LPU64")\n",
+                               reply->lock_desc.l_resource.lr_name.name[0],
+                               reply->lock_desc.l_resource.lr_name.name[1],
+                               reply->lock_desc.l_resource.lr_name.name[2],
+                               lock->l_resource->lr_name.name[0],
+                               lock->l_resource->lr_name.name[1],
+                               lock->l_resource->lr_name.name[2]);
 
                         rc = ldlm_lock_change_resource(ns, lock,
                                            reply->lock_desc.l_resource.lr_name);
@@ -509,13 +564,16 @@ cleanup:
  * a single page on the send/receive side. XXX: 512 should be changed
  * to more adequate value. */
 static inline int ldlm_req_handles_avail(struct obd_export *exp,
-                                         int *size, int bufcount, int off)
+                                         __u32 *size, int bufcount, int off)
 {
         int avail = min_t(int, LDLM_MAXREQSIZE, CFS_PAGE_SIZE - 512);
 
         avail -= lustre_msg_size(class_exp2cliimp(exp)->imp_msg_magic,
                                  bufcount, size);
-        avail /= sizeof(struct lustre_handle);
+        if (likely(avail >= 0))
+                avail /= (int)sizeof(struct lustre_handle);
+        else
+                avail = 0;
         avail += LDLM_LOCKREQ_HANDLES - off;
 
         return avail;
@@ -523,7 +581,7 @@ static inline int ldlm_req_handles_avail(struct obd_export *exp,
 
 static inline int ldlm_cancel_handles_avail(struct obd_export *exp)
 {
-        int size[2] = { sizeof(struct ptlrpc_body),
+        __u32 size[2] = { sizeof(struct ptlrpc_body),
                         sizeof(struct ldlm_request) };
         return ldlm_req_handles_avail(exp, size, 2, 0);
 }
@@ -531,7 +589,7 @@ static inline int ldlm_cancel_handles_avail(struct obd_export *exp)
 /* Cancel lru locks and pack them into the enqueue request. Pack there the given
  * @count locks in @cancels. */
 struct ptlrpc_request *ldlm_prep_elc_req(struct obd_export *exp, int version,
-                                         int opc, int bufcount, int *size,
+                                         int opc, int bufcount, __u32 *size,
                                          int bufoff, int canceloff,
                                          struct list_head *cancels, int count)
 {
@@ -549,12 +607,12 @@ struct ptlrpc_request *ldlm_prep_elc_req(struct obd_export *exp, int version,
                 LASSERT(bufoff < bufcount);
 
                 avail = ldlm_req_handles_avail(exp, size, bufcount, canceloff);
-                flags = ns_connect_lru_resize(ns) ? 
+                flags = ns_connect_lru_resize(ns) ?
                         LDLM_CANCEL_LRUR : LDLM_CANCEL_AGED;
                 to_free = !ns_connect_lru_resize(ns) &&
                           opc == LDLM_ENQUEUE ? 1 : 0;
 
-                /* Cancel lru locks here _only_ if the server supports 
+                /* Cancel lru locks here _only_ if the server supports
                  * EARLY_CANCEL. Otherwise we have to send extra CANCEL
                  * rpc, what will make us slower. */
                 if (avail > count)
@@ -566,8 +624,10 @@ struct ptlrpc_request *ldlm_prep_elc_req(struct obd_export *exp, int version,
                         pack = avail;
                 size[bufoff] = ldlm_request_bufsize(pack, opc);
         }
+
         req = ptlrpc_prep_req(class_exp2cliimp(exp), version,
                               opc, bufcount, size, NULL);
+        req->rq_export = class_export_get(exp);
         if (exp_connect_cancelset(exp) && req) {
                 if (canceloff) {
                         dlm = lustre_msg_buf(req->rq_reqmsg, bufoff,
@@ -589,7 +649,7 @@ struct ptlrpc_request *ldlm_prep_elc_req(struct obd_export *exp, int version,
 }
 
 struct ptlrpc_request *ldlm_prep_enqueue_req(struct obd_export *exp,
-                                             int bufcount, int *size,
+                                             int bufcount, __u32 *size,
                                              struct list_head *cancels,
                                              int count)
 {
@@ -614,9 +674,10 @@ int ldlm_cli_enqueue(struct obd_export *exp, struct ptlrpc_request **reqp,
         struct ldlm_lock *lock;
         struct ldlm_request *body;
         struct ldlm_reply *reply;
-        int size[3] = { [MSG_PTLRPC_BODY_OFF] = sizeof(struct ptlrpc_body),
+        __u32 size[3] = { [MSG_PTLRPC_BODY_OFF] = sizeof(struct ptlrpc_body),
                         [DLM_LOCKREQ_OFF]     = sizeof(*body),
-                        [DLM_REPLY_REC_OFF]   = lvb_len };
+                        [DLM_REPLY_REC_OFF]   = lvb_len ? lvb_len :
+                                                sizeof(struct ost_lvb) };
         int is_replay = *flags & LDLM_FL_REPLAY;
         int req_passed_in = 1, rc, err;
         struct ptlrpc_request *req;
@@ -696,7 +757,7 @@ int ldlm_cli_enqueue(struct obd_export *exp, struct ptlrpc_request **reqp,
         /* Continue as normal. */
         if (!req_passed_in) {
                 size[DLM_LOCKREPLY_OFF] = sizeof(*reply);
-                ptlrpc_req_set_repsize(req, 2 + (lvb_len > 0), size);
+                ptlrpc_req_set_repsize(req, 3, size);
         }
 
         /*
@@ -769,7 +830,7 @@ int ldlm_cli_convert(struct lustre_handle *lockh, int new_mode, __u32 *flags)
         struct ldlm_lock *lock;
         struct ldlm_resource *res;
         struct ptlrpc_request *req;
-        int size[2] = { [MSG_PTLRPC_BODY_OFF] = sizeof(struct ptlrpc_body),
+        __u32 size[2] = { [MSG_PTLRPC_BODY_OFF] = sizeof(struct ptlrpc_body),
                         [DLM_LOCKREQ_OFF]     = sizeof(*body) };
         int rc;
         ENTRY;
@@ -922,7 +983,7 @@ int ldlm_cli_cancel_req(struct obd_export *exp,
 {
         struct ptlrpc_request *req = NULL;
         struct ldlm_request *body;
-        int size[2] = { [MSG_PTLRPC_BODY_OFF] = sizeof(struct ptlrpc_body),
+        __u32 size[2] = { [MSG_PTLRPC_BODY_OFF] = sizeof(struct ptlrpc_body),
                         [DLM_LOCKREQ_OFF]     = sizeof(*body) };
         struct obd_import *imp;
         int free, sent = 0;
@@ -1006,28 +1067,28 @@ int ldlm_cli_update_pool(struct ptlrpc_request *req)
         __u64 old_slv, new_slv;
         __u32 new_limit;
         ENTRY;
-    
-        if (unlikely(!req->rq_import || !req->rq_import->imp_obd || 
+
+        if (unlikely(!req->rq_import || !req->rq_import->imp_obd ||
                      !imp_connect_lru_resize(req->rq_import)))
         {
-                /* 
-                 * Do nothing for corner cases. 
+                /*
+                 * Do nothing for corner cases.
                  */
                 RETURN(0);
         }
 
-        /* 
-         * In some cases RPC may contain slv and limit zeroed out. This is 
+        /*
+         * In some cases RPC may contain slv and limit zeroed out. This is
          * the case when server does not support lru resize feature. This is
          * also possible in some recovery cases when server side reqs have no
-         * ref to obd export and thus access to server side namespace is no 
-         * possible. 
+         * ref to obd export and thus access to server side namespace is no
+         * possible.
          */
-        if (lustre_msg_get_slv(req->rq_repmsg) == 0 || 
+        if (lustre_msg_get_slv(req->rq_repmsg) == 0 ||
             lustre_msg_get_limit(req->rq_repmsg) == 0) {
                 DEBUG_REQ(D_HA, req, "Zero SLV or Limit found "
-                          "(SLV: "LPU64", Limit: %u)", 
-                          lustre_msg_get_slv(req->rq_repmsg), 
+                          "(SLV: "LPU64", Limit: %u)",
+                          lustre_msg_get_slv(req->rq_repmsg),
                           lustre_msg_get_limit(req->rq_repmsg));
                 RETURN(0);
         }
@@ -1036,12 +1097,12 @@ int ldlm_cli_update_pool(struct ptlrpc_request *req)
         new_slv = lustre_msg_get_slv(req->rq_repmsg);
         obd = req->rq_import->imp_obd;
 
-        /* 
-         * Set new SLV and Limit to obd fields to make accessible for pool 
+        /*
+         * Set new SLV and Limit to obd fields to make accessible for pool
          * thread. We do not access obd_namespace and pool directly here
          * as there is no reliable way to make sure that they are still
          * alive in cleanup time. Evil races are possible which may cause
-         * oops in that time. 
+         * oops in that time.
          */
         write_lock(&obd->obd_pool_lock);
         old_slv = obd->obd_pool_slv;
@@ -1049,26 +1110,6 @@ int ldlm_cli_update_pool(struct ptlrpc_request *req)
         obd->obd_pool_limit = new_limit;
         write_unlock(&obd->obd_pool_lock);
 
-        /* 
-         * Check if we need to wakeup pools thread for fast SLV change. 
-         * This is only done when threads period is noticably long like 
-         * 10s or more. 
-         */
-#if defined(__KERNEL__) && (LDLM_POOLS_THREAD_PERIOD >= 10)
-        if (old_slv > 0) {
-                __u64 fast_change = old_slv * LDLM_POOLS_FAST_SLV_CHANGE;
-                do_div(fast_change, 100);
-
-                /* 
-                 * Wake up pools thread only if SLV has changed more than 
-                 * 50% since last update. In this case we want to react asap. 
-                 * Otherwise it is no sense to wake up pools as they are 
-                 * re-calculated every LDLM_POOLS_THREAD_PERIOD anyways. 
-                 */
-                if (old_slv > new_slv && old_slv - new_slv > fast_change)
-                        ldlm_pools_wakeup();
-        }
-#endif
         RETURN(0);
 }
 EXPORT_SYMBOL(ldlm_cli_update_pool);
@@ -1087,7 +1128,7 @@ int ldlm_cli_cancel(struct lustre_handle *lockh)
                 LDLM_DEBUG_NOLOCK("lock is already being destroyed\n");
                 RETURN(0);
         }
-        
+
         rc = ldlm_cli_cancel_local(lock);
         if (rc < 0 || rc == LDLM_FL_LOCAL_ONLY) {
                 LDLM_LOCK_PUT(lock);
@@ -1155,17 +1196,17 @@ static int ldlm_cancel_list(struct list_head *cancels, int count, int flags)
         RETURN(count);
 }
 
-/* Return 1 to stop lru processing and keep current lock cached. Return zero 
+/* Return 1 to stop lru processing and keep current lock cached. Return zero
  * otherwise. */
 static ldlm_policy_res_t ldlm_cancel_shrink_policy(struct ldlm_namespace *ns,
                                                    struct ldlm_lock *lock,
-                                                   int unused, int added, 
+                                                   int unused, int added,
                                                    int count)
 {
         int lock_cost;
         __u64 page_nr;
 
-        /* Stop lru processing when we reached passed @count or checked all 
+        /* Stop lru processing when we reached passed @count or checked all
          * locks in lru. */
         if (count && added >= count)
                 return LDLM_POLICY_KEEP_LOCK;
@@ -1182,7 +1223,7 @@ static ldlm_policy_res_t ldlm_cancel_shrink_policy(struct ldlm_namespace *ns,
 #ifdef __KERNEL__
                 /* XXX: In fact this is evil hack, we can't access inode
                  * here. For doing it right we need somehow to have number
-                 * of covered by lock. This should be fixed later when 10718 
+                 * of covered by lock. This should be fixed later when 10718
                  * is landed. */
                 if (lock->l_ast_data != NULL) {
                         struct inode *inode = lock->l_ast_data;
@@ -1199,15 +1240,15 @@ static ldlm_policy_res_t ldlm_cancel_shrink_policy(struct ldlm_namespace *ns,
         /* Keep all expensive locks in lru for the memory pressure time
          * cancel policy. They anyways may be canceled by lru resize
          * pplicy if they have not small enough CLV. */
-        return lock_cost > ns->ns_shrink_thumb ? 
+        return lock_cost > ns->ns_shrink_thumb ?
                 LDLM_POLICY_KEEP_LOCK : LDLM_POLICY_CANCEL_LOCK;
 }
 
-/* Return 1 to stop lru processing and keep current lock cached. Return zero 
+/* Return 1 to stop lru processing and keep current lock cached. Return zero
  * otherwise. */
 static ldlm_policy_res_t ldlm_cancel_lrur_policy(struct ldlm_namespace *ns,
-                                                 struct ldlm_lock *lock, 
-                                                 int unused, int added, 
+                                                 struct ldlm_lock *lock,
+                                                 int unused, int added,
                                                  int count)
 {
         cfs_time_t cur = cfs_time_current();
@@ -1215,7 +1256,7 @@ static ldlm_policy_res_t ldlm_cancel_lrur_policy(struct ldlm_namespace *ns,
         __u64 slv, lvf, lv;
         cfs_time_t la;
 
-        /* Stop lru processing when we reached passed @count or checked all 
+        /* Stop lru processing when we reached passed @count or checked all
          * locks in lru. */
         if (count && added >= count)
                 return LDLM_POLICY_KEEP_LOCK;
@@ -1223,63 +1264,63 @@ static ldlm_policy_res_t ldlm_cancel_lrur_policy(struct ldlm_namespace *ns,
         slv = ldlm_pool_get_slv(pl);
         lvf = ldlm_pool_get_lvf(pl);
 
-        la = cfs_duration_sec(cfs_time_sub(cur, 
+        la = cfs_duration_sec(cfs_time_sub(cur,
                               lock->l_last_used));
 
-        /* Stop when slv is not yet come from server or 
+        /* Stop when slv is not yet come from server or
          * lv is smaller than it is. */
         lv = lvf * la * unused;
 
         /* Inform pool about current CLV to see it via proc. */
         ldlm_pool_set_clv(pl, lv);
-        return (slv == 1 || lv < slv) ? 
+        return (slv == 1 || lv < slv) ?
                 LDLM_POLICY_KEEP_LOCK : LDLM_POLICY_CANCEL_LOCK;
 }
 
-/* Return 1 to stop lru processing and keep current lock cached. Return zero 
+/* Return 1 to stop lru processing and keep current lock cached. Return zero
  * otherwise. */
 static ldlm_policy_res_t ldlm_cancel_passed_policy(struct ldlm_namespace *ns,
-                                                   struct ldlm_lock *lock, 
+                                                   struct ldlm_lock *lock,
                                                    int unused, int added,
                                                    int count)
 {
-        /* Stop lru processing when we reached passed @count or checked all 
+        /* Stop lru processing when we reached passed @count or checked all
          * locks in lru. */
-        return (added >= count) ? 
+        return (added >= count) ?
                 LDLM_POLICY_KEEP_LOCK : LDLM_POLICY_CANCEL_LOCK;
 }
 
-/* Return 1 to stop lru processing and keep current lock cached. Return zero 
+/* Return 1 to stop lru processing and keep current lock cached. Return zero
  * otherwise. */
 static ldlm_policy_res_t ldlm_cancel_aged_policy(struct ldlm_namespace *ns,
-                                                 struct ldlm_lock *lock, 
+                                                 struct ldlm_lock *lock,
                                                  int unused, int added,
                                                  int count)
 {
-        /* Stop lru processing if young lock is found and we reached passed 
+        /* Stop lru processing if young lock is found and we reached passed
          * @count. */
-        return ((added >= count) && 
+        return ((added >= count) &&
                 cfs_time_before(cfs_time_current(),
                                 cfs_time_add(lock->l_last_used,
-                                             ns->ns_max_age))) ? 
+                                             ns->ns_max_age))) ?
                 LDLM_POLICY_KEEP_LOCK : LDLM_POLICY_CANCEL_LOCK;
 }
 
-/* Return 1 to stop lru processing and keep current lock cached. Return zero 
+/* Return 1 to stop lru processing and keep current lock cached. Return zero
  * otherwise. */
 static ldlm_policy_res_t ldlm_cancel_default_policy(struct ldlm_namespace *ns,
-                                                    struct ldlm_lock *lock, 
+                                                    struct ldlm_lock *lock,
                                                     int unused, int added,
                                                     int count)
 {
-        /* Stop lru processing when we reached passed @count or checked all 
+        /* Stop lru processing when we reached passed @count or checked all
          * locks in lru. */
-        return (added >= count) ? 
+        return (added >= count) ?
                 LDLM_POLICY_KEEP_LOCK : LDLM_POLICY_CANCEL_LOCK;
 }
 
-typedef ldlm_policy_res_t (*ldlm_cancel_lru_policy_t)(struct ldlm_namespace *, 
-                                                      struct ldlm_lock *, int, 
+typedef ldlm_policy_res_t (*ldlm_cancel_lru_policy_t)(struct ldlm_namespace *,
+                                                      struct ldlm_lock *, int,
                                                       int, int);
 
 static ldlm_cancel_lru_policy_t
@@ -1296,10 +1337,10 @@ ldlm_cancel_lru_policy(struct ldlm_namespace *ns, int flags)
                 if (flags & LDLM_CANCEL_AGED)
                         return ldlm_cancel_aged_policy;
         }
-        
+
         return ldlm_cancel_default_policy;
 }
+
 /* - Free space in lru for @count new locks,
  *   redundant unused locks are canceled locally;
  * - also cancel locally unused aged locks;
@@ -1342,7 +1383,7 @@ int ldlm_cancel_lru_local(struct ldlm_namespace *ns, struct list_head *cancels,
 
         pf = ldlm_cancel_lru_policy(ns, flags);
         LASSERT(pf != NULL);
-        
+
         while (!list_empty(&ns->ns_unused_list)) {
                 /* For any flags, stop scanning if @max is reached. */
                 if (max && added >= max)
@@ -1369,11 +1410,11 @@ int ldlm_cancel_lru_local(struct ldlm_namespace *ns, struct list_head *cancels,
                  * we find a lock that should stay in the cache.
                  * We should take into account lock age anyway
                  * as new lock even if it is small of weight is
-                 * valuable resource. 
+                 * valuable resource.
                  *
                  * That is, for shrinker policy we drop only
                  * old locks, but additionally chose them by
-                 * their weight. Big extent locks will stay in 
+                 * their weight. Big extent locks will stay in
                  * the cache. */
                 if (pf(ns, lock, unused, added, count) == LDLM_POLICY_KEEP_LOCK)
                         break;
@@ -1399,8 +1440,8 @@ int ldlm_cancel_lru_local(struct ldlm_namespace *ns, struct list_head *cancels,
 
                 /* If we have chosen to cancel this lock voluntarily, we
                  * better send cancel notification to server, so that it
-                 * frees appropriate state. This might lead to a race 
-                 * where while we are doing cancel here, server is also 
+                 * frees appropriate state. This might lead to a race
+                 * where while we are doing cancel here, server is also
                  * silently cancelling this lock. */
                 lock->l_flags &= ~LDLM_FL_CANCEL_ON_BLOCK;
 
@@ -1429,7 +1470,7 @@ int ldlm_cancel_lru_local(struct ldlm_namespace *ns, struct list_head *cancels,
         RETURN(ldlm_cancel_list(cancels, added, cancel_flags));
 }
 
-/* Returns number of locks which could be canceled next time when 
+/* Returns number of locks which could be canceled next time when
  * ldlm_cancel_lru() is called. Used from locks pool shrinker. */
 int ldlm_cancel_lru_estimate(struct ldlm_namespace *ns,
                              int count, int max, int flags)
@@ -1450,10 +1491,10 @@ int ldlm_cancel_lru_estimate(struct ldlm_namespace *ns,
                         break;
 
                 /* Somebody is already doing CANCEL or there is a
-                 * blocking request will send cancel. Let's not count 
+                 * blocking request will send cancel. Let's not count
                  * this lock. */
                 if ((lock->l_flags & LDLM_FL_CANCELING) ||
-                    (lock->l_flags & LDLM_FL_BL_AST)) 
+                    (lock->l_flags & LDLM_FL_BL_AST))
                         continue;
 
                 /* Pass the lock through the policy filter and see if it
@@ -1472,7 +1513,7 @@ int ldlm_cancel_lru_estimate(struct ldlm_namespace *ns,
  * in a thread and this function will return after the thread has been
  * asked to call the callback.  when called with LDLM_SYNC the blocking
  * callback will be performed in this function. */
-int ldlm_cancel_lru(struct ldlm_namespace *ns, int nr, ldlm_sync_t sync, 
+int ldlm_cancel_lru(struct ldlm_namespace *ns, int nr, ldlm_sync_t sync,
                     int flags)
 {
         CFS_LIST_HEAD(cancels);
@@ -1527,7 +1568,7 @@ int ldlm_cancel_resource_local(struct ldlm_resource *res,
 
                 /* If somebody is already doing CANCEL, or blocking ast came,
                  * skip this lock. */
-                if (lock->l_flags & LDLM_FL_BL_AST || 
+                if (lock->l_flags & LDLM_FL_BL_AST ||
                     lock->l_flags & LDLM_FL_CANCELING)
                         continue;
 
@@ -1870,15 +1911,16 @@ static int ldlm_chain_lock_for_replay(struct ldlm_lock *lock, void *closure)
 static int replay_lock_interpret(struct ptlrpc_request *req,
                                  struct ldlm_async_args *aa, int rc)
 {
-        struct ldlm_lock *lock;
-        struct ldlm_reply *reply;
+        struct lustre_handle  old_hash_key;
+        struct ldlm_lock     *lock;
+        struct ldlm_reply    *reply;
+        struct obd_export    *exp;
 
         ENTRY;
         atomic_dec(&req->rq_import->imp_replay_inflight);
         if (rc != ELDLM_OK)
                 GOTO(out, rc);
 
-
         reply = lustre_swab_repbuf(req, DLM_LOCKREPLY_OFF, sizeof(*reply),
                                    lustre_swab_ldlm_reply);
         if (reply == NULL) {
@@ -1896,7 +1938,16 @@ static int replay_lock_interpret(struct ptlrpc_request *req,
                 GOTO(out, rc = -ESTALE);
         }
 
+        old_hash_key = lock->l_remote_handle;
         lock->l_remote_handle = reply->lock_handle;
+
+        /* Key change rehash lock in per-export hash with new key */
+       exp = req->rq_export;
+        if (exp && exp->exp_lock_hash)
+                lustre_hash_rehash_key(exp->exp_lock_hash, &old_hash_key,
+                                      &lock->l_remote_handle,
+                                       &lock->l_exp_hash);
+
         LDLM_DEBUG(lock, "replayed lock:");
         ptlrpc_import_recovery_state_machine(req->rq_import);
         LDLM_LOCK_PUT(lock);
@@ -1915,7 +1966,7 @@ static int replay_one_lock(struct obd_import *imp, struct ldlm_lock *lock)
         struct ldlm_reply *reply;
         struct ldlm_async_args *aa;
         int buffers = 2;
-        int size[3] = { sizeof(struct ptlrpc_body) };
+        __u32 size[3] = { sizeof(struct ptlrpc_body) };
         int flags;
         ENTRY;
 
@@ -1934,6 +1985,7 @@ static int replay_one_lock(struct obd_import *imp, struct ldlm_lock *lock)
                 ldlm_lock_cancel(lock);
                 RETURN(0);
         }
+
         /*
          * If granted mode matches the requested mode, this lock is granted.
          *
@@ -1982,7 +2034,7 @@ static int replay_one_lock(struct obd_import *imp, struct ldlm_lock *lock)
 
         atomic_inc(&req->rq_import->imp_replay_inflight);
         CLASSERT(sizeof(*aa) <= sizeof(req->rq_async_args));
-        aa = (struct ldlm_async_args *)&req->rq_async_args;
+        aa = ptlrpc_req_async_args(req);
         aa->lock_handle = body->lock_handle[0];
         req->rq_interpret_reply = replay_lock_interpret;
         ptlrpcd_add_req(req);
@@ -2005,15 +2057,22 @@ int ldlm_replay_locks(struct obd_import *imp)
         /* ensure this doesn't fall to 0 before all have been queued */
         atomic_inc(&imp->imp_replay_inflight);
 
-        (void)ldlm_namespace_foreach(ns, ldlm_chain_lock_for_replay, &list);
-
-        list_for_each_entry_safe(lock, next, &list, l_pending_chain) {
-                list_del_init(&lock->l_pending_chain);
-                if (rc)
-                        continue; /* or try to do the rest? */
-                rc = replay_one_lock(imp, lock);
+        if (imp->imp_no_lock_replay) {
+                /* VBR: locks should be cancelled here */
+                ldlm_namespace_cleanup(ns, LDLM_FL_LOCAL_ONLY);
+                spin_lock(&imp->imp_lock);
+                imp->imp_no_lock_replay = 0;
+                spin_unlock(&imp->imp_lock);
+        } else {
+                (void)ldlm_namespace_foreach(ns, ldlm_chain_lock_for_replay,
+                                             &list);
+                list_for_each_entry_safe(lock, next, &list, l_pending_chain) {
+                        list_del_init(&lock->l_pending_chain);
+                        if (rc)
+                                continue; /* or try to do the rest? */
+                        rc = replay_one_lock(imp, lock);
+                }
         }
-
         atomic_dec(&imp->imp_replay_inflight);
 
         RETURN(rc);