Whamcloud - gitweb
LU-7524 fld: fld_clientlookup retries next target
[fs/lustre-release.git] / lustre / fld / fld_request.c
index 1ec2b9c..9354922 100644 (file)
-/* -*- MODE: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
- * 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).
  *
- *  lustre/fld/fld_request.c
- *  FLD (Fids Location Database)
+ * 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
  *
- *  Copyright (C) 2006 Cluster File Systems, Inc.
- *   Author: Yury Umanets <umka@clusterfs.com>
+ * 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.
  *
- *   This file is part of the Lustre file system, http://www.lustre.org
- *   Lustre is a trademark of Cluster File Systems, Inc.
+ * Copyright (c) 2011, 2015, Intel Corporation.
+ */
+/*
+ * This file is part of Lustre, http://www.lustre.org/
+ * Lustre is a trademark of Sun Microsystems, Inc.
  *
- *   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.
+ * lustre/fld/fld_request.c
  *
- *   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.
+ * FLD (Fids Location Database)
  *
- *   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.
+ * Author: Yury Umanets <umka@clusterfs.com>
  */
-#ifndef EXPORT_SYMTAB
-# define EXPORT_SYMTAB
-#endif
+
 #define DEBUG_SUBSYSTEM S_FLD
 
-#ifdef __KERNEL__
-# include <libcfs/libcfs.h>
-# include <linux/module.h>
-# include <linux/jbd.h>
-# include <asm/div64.h>
-#else /* __KERNEL__ */
-# include <liblustre.h>
-# include <libcfs/list.h>
-#endif
+#include <libcfs/libcfs.h>
+#include <linux/module.h>
+#include <linux/math64.h>
 
 #include <obd.h>
 #include <obd_class.h>
-#include <lustre_ver.h>
 #include <obd_support.h>
 #include <lprocfs_status.h>
-
-#include <dt_object.h>
-#include <md_object.h>
 #include <lustre_req_layout.h>
 #include <lustre_fld.h>
 #include <lustre_mdc.h>
 #include "fld_internal.h"
 
-/* TODO: these 3 functions are copies of flow-control code from mdc_lib.c 
- * It should be common thing. The same about mdc RPC lock */
-static int fld_req_avail(struct client_obd *cli, struct mdc_cache_waiter *mcw)
+static int fld_rrb_hash(struct lu_client_fld *fld, u64 seq)
 {
-        int rc;
-        ENTRY;
-        spin_lock(&cli->cl_loi_list_lock);
-        rc = list_empty(&mcw->mcw_entry);
-        spin_unlock(&cli->cl_loi_list_lock);
-        RETURN(rc);
-};
-
-static void fld_enter_request(struct client_obd *cli)
-{
-        struct mdc_cache_waiter mcw;
-        struct l_wait_info lwi = { 0 };
-
-        spin_lock(&cli->cl_loi_list_lock);
-        if (cli->cl_r_in_flight >= cli->cl_max_rpcs_in_flight) {
-                list_add_tail(&mcw.mcw_entry, &cli->cl_cache_waiters);
-                cfs_waitq_init(&mcw.mcw_waitq);
-                spin_unlock(&cli->cl_loi_list_lock);
-                l_wait_event(mcw.mcw_waitq, fld_req_avail(cli, &mcw), &lwi);
-        } else {
-                cli->cl_r_in_flight++;
-                spin_unlock(&cli->cl_loi_list_lock);
-        }
-}
-
-static void fld_exit_request(struct client_obd *cli)
-{
-        struct list_head *l, *tmp;
-        struct mdc_cache_waiter *mcw;
-
-        spin_lock(&cli->cl_loi_list_lock);
-        cli->cl_r_in_flight--;
-        list_for_each_safe(l, tmp, &cli->cl_cache_waiters) {
-                
-                if (cli->cl_r_in_flight >= cli->cl_max_rpcs_in_flight) {
-                        /* No free request slots anymore */
-                        break;
-                }
-
-                mcw = list_entry(l, struct mdc_cache_waiter, mcw_entry);
-                list_del_init(&mcw->mcw_entry);
-                cli->cl_r_in_flight++;
-                cfs_waitq_signal(&mcw->mcw_waitq);
-        }
-        spin_unlock(&cli->cl_loi_list_lock);
-}
-
-static int fld_rrb_hash(struct lu_client_fld *fld,
-                        seqno_t seq)
-{
-        LASSERT(fld->lcf_count > 0);
-        return do_div(seq, fld->lcf_count);
+       LASSERT(fld->lcf_count > 0);
+       return do_div(seq, fld->lcf_count);
 }
 
 static struct lu_fld_target *
-fld_rrb_scan(struct lu_client_fld *fld, seqno_t seq)
+fld_rrb_scan(struct lu_client_fld *fld, u64 seq)
 {
         struct lu_fld_target *target;
         int hash;
         ENTRY;
 
-        hash = fld_rrb_hash(fld, seq);
-
-        list_for_each_entry(target, &fld->lcf_targets, ft_chain) {
+       /* Because almost all of special sequence located in MDT0,
+        * it should go to index 0 directly, instead of calculating
+        * hash again, and also if other MDTs is not being connected,
+        * the fld lookup requests(for seq on MDT0) should not be
+        * blocked because of other MDTs */
+       if (fid_seq_is_norm(seq))
+               hash = fld_rrb_hash(fld, seq);
+       else
+               hash = 0;
+
+again:
+       list_for_each_entry(target, &fld->lcf_targets, ft_chain) {
                 if (target->ft_idx == hash)
                         RETURN(target);
         }
 
+       if (hash != 0) {
+               /* It is possible the remote target(MDT) are not connected to
+                * with client yet, so we will refer this to MDT0, which should
+                * be connected during mount */
+               hash = 0;
+               goto again;
+       }
+
         CERROR("%s: Can't find target by hash %d (seq "LPX64"). "
                "Targets (%d):\n", fld->lcf_name, hash, seq,
                fld->lcf_count);
 
-        list_for_each_entry(target, &fld->lcf_targets, ft_chain) {
+       list_for_each_entry(target, &fld->lcf_targets, ft_chain) {
                 const char *srv_name = target->ft_srv != NULL  ?
                         target->ft_srv->lsf_name : "<null>";
                 const char *exp_name = target->ft_exp != NULL ?
@@ -149,48 +116,28 @@ fld_rrb_scan(struct lu_client_fld *fld, seqno_t seq)
         RETURN(NULL);
 }
 
-static int fld_dht_hash(struct lu_client_fld *fld,
-                        seqno_t seq)
-{
-        /* XXX: here should be DHT hash */
-        return fld_rrb_hash(fld, seq);
-}
-
-static struct lu_fld_target *
-fld_dht_scan(struct lu_client_fld *fld, seqno_t seq)
-{
-        /* XXX: here should be DHT scan code */
-        return fld_rrb_scan(fld, seq);
-}
-
-struct lu_fld_hash fld_hash[3] = {
-        {
-                .fh_name = "DHT",
-                .fh_hash_func = fld_dht_hash,
-                .fh_scan_func = fld_dht_scan
-        },
+struct lu_fld_hash fld_hash[] = {
         {
                 .fh_name = "RRB",
                 .fh_hash_func = fld_rrb_hash,
                 .fh_scan_func = fld_rrb_scan
         },
         {
-                0,
+               NULL,
         }
 };
 
 static struct lu_fld_target *
-fld_client_get_target(struct lu_client_fld *fld,
-                      seqno_t seq)
+fld_client_get_target(struct lu_client_fld *fld, u64 seq)
 {
-        struct lu_fld_target *target;
-        ENTRY;
+       struct lu_fld_target *target;
+       ENTRY;
 
-        LASSERT(fld->lcf_hash != NULL);
+       LASSERT(fld->lcf_hash != NULL);
 
-        spin_lock(&fld->lcf_lock);
-        target = fld->lcf_hash->fh_scan_func(fld, seq);
-        spin_unlock(&fld->lcf_lock);
+       spin_lock(&fld->lcf_lock);
+       target = fld->lcf_hash->fh_scan_func(fld, seq);
+       spin_unlock(&fld->lcf_lock);
 
         if (target != NULL) {
                 CDEBUG(D_INFO, "%s: Found target (idx "LPU64
@@ -208,32 +155,26 @@ fld_client_get_target(struct lu_client_fld *fld,
 int fld_client_add_target(struct lu_client_fld *fld,
                           struct lu_fld_target *tar)
 {
-        const char *name = fld_target_name(tar);
+       const char *name;
         struct lu_fld_target *target, *tmp;
         ENTRY;
 
         LASSERT(tar != NULL);
+       name = fld_target_name(tar);
         LASSERT(name != NULL);
         LASSERT(tar->ft_srv != NULL || tar->ft_exp != NULL);
 
-        if (fld->lcf_flags != LUSTRE_FLD_INIT) {
-                CERROR("%s: Attempt to add target %s (idx "LPU64") "
-                       "on fly - skip it\n", fld->lcf_name, name,
-                       tar->ft_idx);
-                RETURN(0);
-        } else {
-                CDEBUG(D_INFO, "%s: Adding target %s (idx "
-                       LPU64")\n", fld->lcf_name, name, tar->ft_idx);
-        }
+       CDEBUG(D_INFO, "%s: Adding target %s (idx "LPU64")\n", fld->lcf_name,
+              name, tar->ft_idx);
 
         OBD_ALLOC_PTR(target);
         if (target == NULL)
                 RETURN(-ENOMEM);
 
-        spin_lock(&fld->lcf_lock);
-        list_for_each_entry(tmp, &fld->lcf_targets, ft_chain) {
-                if (tmp->ft_idx == tar->ft_idx) {
-                        spin_unlock(&fld->lcf_lock);
+       spin_lock(&fld->lcf_lock);
+       list_for_each_entry(tmp, &fld->lcf_targets, ft_chain) {
+               if (tmp->ft_idx == tar->ft_idx) {
+                       spin_unlock(&fld->lcf_lock);
                         OBD_FREE_PTR(target);
                         CERROR("Target %s exists in FLD and known as %s:#"LPU64"\n",
                                name, fld_target_name(tmp), tmp->ft_idx);
@@ -247,30 +188,27 @@ int fld_client_add_target(struct lu_client_fld *fld,
         target->ft_srv = tar->ft_srv;
         target->ft_idx = tar->ft_idx;
 
-        list_add_tail(&target->ft_chain,
-                      &fld->lcf_targets);
+       list_add_tail(&target->ft_chain, &fld->lcf_targets);
 
         fld->lcf_count++;
-        spin_unlock(&fld->lcf_lock);
+       spin_unlock(&fld->lcf_lock);
 
-        RETURN(0);
+       RETURN(0);
 }
 EXPORT_SYMBOL(fld_client_add_target);
 
 /* Remove export from FLD */
-int fld_client_del_target(struct lu_client_fld *fld,
-                          __u64 idx)
+int fld_client_del_target(struct lu_client_fld *fld, __u64 idx)
 {
-        struct lu_fld_target *target, *tmp;
-        ENTRY;
+       struct lu_fld_target *target, *tmp;
+       ENTRY;
 
-        spin_lock(&fld->lcf_lock);
-        list_for_each_entry_safe(target, tmp,
-                                 &fld->lcf_targets, ft_chain) {
-                if (target->ft_idx == idx) {
-                        fld->lcf_count--;
-                        list_del(&target->ft_chain);
-                        spin_unlock(&fld->lcf_lock);
+       spin_lock(&fld->lcf_lock);
+       list_for_each_entry_safe(target, tmp, &fld->lcf_targets, ft_chain) {
+               if (target->ft_idx == idx) {
+                       fld->lcf_count--;
+                       list_del(&target->ft_chain);
+                       spin_unlock(&fld->lcf_lock);
 
                         if (target->ft_exp != NULL)
                                 class_export_put(target->ft_exp);
@@ -279,46 +217,40 @@ int fld_client_del_target(struct lu_client_fld *fld,
                         RETURN(0);
                 }
         }
-        spin_unlock(&fld->lcf_lock);
-        RETURN(-ENOENT);
+       spin_unlock(&fld->lcf_lock);
+       RETURN(-ENOENT);
 }
-EXPORT_SYMBOL(fld_client_del_target);
-
-static void fld_client_proc_fini(struct lu_client_fld *fld);
 
-#ifdef LPROCFS
+#ifdef CONFIG_PROC_FS
 static int fld_client_proc_init(struct lu_client_fld *fld)
 {
-        int rc;
-        ENTRY;
-
-        fld->lcf_proc_dir = lprocfs_register(fld->lcf_name,
-                                             fld_type_proc_dir,
-                                             NULL, NULL);
-
-        if (IS_ERR(fld->lcf_proc_dir)) {
-                CERROR("%s: LProcFS failed in fld-init\n",
-                       fld->lcf_name);
-                rc = PTR_ERR(fld->lcf_proc_dir);
-                RETURN(rc);
-        }
-
-        rc = lprocfs_add_vars(fld->lcf_proc_dir,
-                              fld_client_proc_list, fld);
-        if (rc) {
-                CERROR("%s: Can't init FLD proc, rc %d\n",
-                       fld->lcf_name, rc);
-                GOTO(out_cleanup, rc);
-        }
-
-        RETURN(0);
+       int rc;
+       ENTRY;
+
+       fld->lcf_proc_dir = lprocfs_register(fld->lcf_name, fld_type_proc_dir,
+                                            NULL, NULL);
+       if (IS_ERR(fld->lcf_proc_dir)) {
+               CERROR("%s: LProcFS failed in fld-init\n",
+                      fld->lcf_name);
+               rc = PTR_ERR(fld->lcf_proc_dir);
+               RETURN(rc);
+       }
+
+       rc = lprocfs_add_vars(fld->lcf_proc_dir, fld_client_proc_list, fld);
+       if (rc) {
+               CERROR("%s: Can't init FLD proc, rc %d\n",
+                      fld->lcf_name, rc);
+               GOTO(out_cleanup, rc);
+       }
+
+       RETURN(0);
 
 out_cleanup:
-        fld_client_proc_fini(fld);
-        return rc;
+       fld_client_proc_fini(fld);
+       return rc;
 }
 
-static void fld_client_proc_fini(struct lu_client_fld *fld)
+void fld_client_proc_fini(struct lu_client_fld *fld)
 {
         ENTRY;
         if (fld->lcf_proc_dir) {
@@ -328,17 +260,19 @@ static void fld_client_proc_fini(struct lu_client_fld *fld)
         }
         EXIT;
 }
-#else
+#else /* !CONFIG_PROC_FS */
 static int fld_client_proc_init(struct lu_client_fld *fld)
 {
         return 0;
 }
 
-static void fld_client_proc_fini(struct lu_client_fld *fld)
+void fld_client_proc_fini(struct lu_client_fld *fld)
 {
         return;
 }
-#endif
+#endif /* CONFIG_PROC_FS */
+
+EXPORT_SYMBOL(fld_client_proc_fini);
 
 static inline int hash_is_sane(int hash)
 {
@@ -348,9 +282,7 @@ static inline int hash_is_sane(int hash)
 int fld_client_init(struct lu_client_fld *fld,
                     const char *prefix, int hash)
 {
-#ifdef __KERNEL__
         int cache_size, cache_threshold;
-#endif
         int rc;
         ENTRY;
 
@@ -365,13 +297,11 @@ int fld_client_init(struct lu_client_fld *fld,
                 RETURN(-EINVAL);
         }
 
-        fld->lcf_count = 0;
-        spin_lock_init(&fld->lcf_lock);
-        fld->lcf_hash = &fld_hash[hash];
-        fld->lcf_flags = LUSTRE_FLD_INIT;
-        INIT_LIST_HEAD(&fld->lcf_targets);
+       fld->lcf_count = 0;
+       spin_lock_init(&fld->lcf_lock);
+       fld->lcf_hash = &fld_hash[hash];
+       INIT_LIST_HEAD(&fld->lcf_targets);
 
-#ifdef __KERNEL__
         cache_size = FLD_CLIENT_CACHE_SIZE /
                 sizeof(struct fld_cache_entry);
 
@@ -379,14 +309,12 @@ int fld_client_init(struct lu_client_fld *fld,
                 FLD_CLIENT_CACHE_THRESHOLD / 100;
 
         fld->lcf_cache = fld_cache_init(fld->lcf_name,
-                                        FLD_CLIENT_HTABLE_SIZE,
                                         cache_size, cache_threshold);
         if (IS_ERR(fld->lcf_cache)) {
                 rc = PTR_ERR(fld->lcf_cache);
                 fld->lcf_cache = NULL;
                 GOTO(out, rc);
         }
-#endif
 
         rc = fld_client_proc_init(fld);
         if (rc)
@@ -404,221 +332,237 @@ EXPORT_SYMBOL(fld_client_init);
 
 void fld_client_fini(struct lu_client_fld *fld)
 {
-        struct lu_fld_target *target, *tmp;
-        ENTRY;
-
-        fld_client_proc_fini(fld);
+       struct lu_fld_target *target, *tmp;
+       ENTRY;
 
-        spin_lock(&fld->lcf_lock);
-        list_for_each_entry_safe(target, tmp,
-                                 &fld->lcf_targets, ft_chain) {
+       spin_lock(&fld->lcf_lock);
+       list_for_each_entry_safe(target, tmp, &fld->lcf_targets, ft_chain) {
                 fld->lcf_count--;
-                list_del(&target->ft_chain);
+               list_del(&target->ft_chain);
                 if (target->ft_exp != NULL)
                         class_export_put(target->ft_exp);
                 OBD_FREE_PTR(target);
         }
-        spin_unlock(&fld->lcf_lock);
+       spin_unlock(&fld->lcf_lock);
 
-#ifdef __KERNEL__
         if (fld->lcf_cache != NULL) {
                 if (!IS_ERR(fld->lcf_cache))
                         fld_cache_fini(fld->lcf_cache);
                 fld->lcf_cache = NULL;
         }
-#endif
 
         EXIT;
 }
 EXPORT_SYMBOL(fld_client_fini);
 
-static int fld_client_rpc(struct obd_export *exp,
-                          struct md_fld *mf, __u32 fld_op)
+int fld_client_rpc(struct obd_export *exp,
+                  struct lu_seq_range *range, __u32 fld_op,
+                  struct ptlrpc_request **reqp)
 {
-        struct ptlrpc_request *req;
-        struct md_fld         *pmf;
-        __u32                 *op;
-        int                    rc;
-        ENTRY;
-
-        LASSERT(exp != NULL);
-
-        req = ptlrpc_request_alloc_pack(class_exp2cliimp(exp), &RQF_FLD_QUERY,
-                                        LUSTRE_MDS_VERSION, FLD_QUERY);
-        if (req == NULL)
-                RETURN(-ENOMEM);
-
-        op = req_capsule_client_get(&req->rq_pill, &RMF_FLD_OPC);
-        *op = fld_op;
-
-        pmf = req_capsule_client_get(&req->rq_pill, &RMF_FLD_MDFLD);
-        *pmf = *mf;
-
-        ptlrpc_request_set_replen(req);
+       struct ptlrpc_request *req = NULL;
+       struct lu_seq_range   *prange;
+       __u32                 *op;
+       int                    rc = 0;
+       struct obd_import     *imp;
+       ENTRY;
+
+       LASSERT(exp != NULL);
+
+again:
+       imp = class_exp2cliimp(exp);
+       switch (fld_op) {
+       case FLD_QUERY:
+               req = ptlrpc_request_alloc_pack(imp, &RQF_FLD_QUERY,
+                                               LUSTRE_MDS_VERSION, FLD_QUERY);
+               if (req == NULL)
+                       RETURN(-ENOMEM);
+
+               /* XXX: only needed when talking to old server(< 2.6), it should
+                * be removed when < 2.6 server is not supported */
+               op = req_capsule_client_get(&req->rq_pill, &RMF_FLD_OPC);
+               *op = FLD_LOOKUP;
+
+               /* For MDS_MDS seq lookup, it will always use LWP connection,
+                * but LWP will be evicted after restart, so cause the error.
+                * so we will set no_delay for seq lookup request, once the
+                * request fails because of the eviction. always retry here */
+               if (imp->imp_connect_flags_orig & OBD_CONNECT_MDS_MDS) {
+                       req->rq_allow_replay = 1;
+                       req->rq_no_delay = 1;
+               }
+               break;
+       case FLD_READ:
+               req = ptlrpc_request_alloc_pack(imp, &RQF_FLD_READ,
+                                               LUSTRE_MDS_VERSION, FLD_READ);
+               if (req == NULL)
+                       RETURN(-ENOMEM);
+
+               req_capsule_set_size(&req->rq_pill, &RMF_GENERIC_DATA,
+                                    RCL_SERVER, PAGE_CACHE_SIZE);
+               break;
+       default:
+               rc = -EINVAL;
+               break;
+       }
+
+       if (rc != 0)
+               RETURN(rc);
+
+       prange = req_capsule_client_get(&req->rq_pill, &RMF_FLD_MDFLD);
+       *prange = *range;
+       ptlrpc_request_set_replen(req);
         req->rq_request_portal = FLD_REQUEST_PORTAL;
-
-        if (fld_op != FLD_LOOKUP)
-                mdc_get_rpc_lock(exp->exp_obd->u.cli.cl_rpc_lock, NULL);
-        fld_enter_request(&exp->exp_obd->u.cli);
-        rc = ptlrpc_queue_wait(req);
-        fld_exit_request(&exp->exp_obd->u.cli);
-        if (fld_op != FLD_LOOKUP)
-                mdc_put_rpc_lock(exp->exp_obd->u.cli.cl_rpc_lock, NULL);
-        if (rc)
-                GOTO(out_req, rc);
-
-        pmf = req_capsule_server_get(&req->rq_pill, &RMF_FLD_MDFLD);
-        if (pmf == NULL)
-                GOTO(out_req, rc = -EFAULT);
-        *mf = *pmf;
-        EXIT;
+       req->rq_reply_portal = MDC_REPLY_PORTAL;
+        ptlrpc_at_set_req_timeout(req);
+
+       obd_get_request_slot(&exp->exp_obd->u.cli);
+       rc = ptlrpc_queue_wait(req);
+       obd_put_request_slot(&exp->exp_obd->u.cli);
+
+       if (rc == -ENOENT) {
+               /* Don't loop forever on non-existing FID sequences. */
+               GOTO(out_req, rc);
+       }
+
+       if (rc != 0) {
+               if (imp->imp_state != LUSTRE_IMP_CLOSED && !imp->imp_deactive) {
+                       /* Since LWP is not replayable, so it will keep
+                        * trying unless umount happens, otherwise it would
+                        * cause unecessary failure of the application. */
+                       ptlrpc_req_finished(req);
+                       rc = 0;
+                       goto again;
+               }
+               GOTO(out_req, rc);
+       }
+
+       if (fld_op == FLD_QUERY) {
+               prange = req_capsule_server_get(&req->rq_pill,
+                                               &RMF_FLD_MDFLD);
+               if (prange == NULL)
+                       GOTO(out_req, rc = -EFAULT);
+               *range = *prange;
+       }
+
+       EXIT;
 out_req:
-        ptlrpc_req_finished(req);
-        return rc;
-}
+       if (rc != 0 || reqp == NULL) {
+               ptlrpc_req_finished(req);
+               req = NULL;
+       }
 
-int fld_client_create(struct lu_client_fld *fld,
-                      seqno_t seq, mdsno_t mds,
-                      const struct lu_env *env)
-{
-        struct md_fld md_fld = { .mf_seq = seq, .mf_mds = mds };
-        struct lu_fld_target *target;
-        int rc;
-        ENTRY;
+       if (reqp != NULL)
+               *reqp = req;
 
-        fld->lcf_flags |= LUSTRE_FLD_RUN;
-        target = fld_client_get_target(fld, seq);
-        LASSERT(target != NULL);
-
-        CDEBUG(D_INFO, "%s: Create fld entry (seq: "LPX64"; mds: "
-               LPU64") on target %s (idx "LPU64")\n", fld->lcf_name,
-               seq, mds, fld_target_name(target), target->ft_idx);
-
-#ifdef __KERNEL__
-        if (target->ft_srv != NULL) {
-                LASSERT(env != NULL);
-                rc = fld_server_create(target->ft_srv, env, seq, mds);
-        } else {
-#endif
-                rc = fld_client_rpc(target->ft_exp, &md_fld, FLD_CREATE);
-#ifdef __KERNEL__
-        }
-#endif
-
-        if (rc == 0) {
-                /*
-                 * Do not return result of calling fld_cache_insert()
-                 * here. First of all because it may return -EEXISTS. Another
-                 * reason is that, we do not want to stop proceeding because of
-                 * cache errors.
-                 */
-                fld_cache_insert(fld->lcf_cache, seq, mds);
-        } else {
-                CERROR("%s: Can't create FLD entry, rc %d\n",
-                       fld->lcf_name, rc);
-        }
-
-        RETURN(rc);
+       return rc;
 }
-EXPORT_SYMBOL(fld_client_create);
 
-int fld_client_delete(struct lu_client_fld *fld, seqno_t seq,
-                      const struct lu_env *env)
+int fld_client_lookup(struct lu_client_fld *fld, u64 seq, u32 *mds,
+                     __u32 flags, const struct lu_env *env)
 {
-        struct md_fld md_fld = { .mf_seq = seq, .mf_mds = 0 };
-        struct lu_fld_target *target;
-        int rc;
-        ENTRY;
-
-        fld->lcf_flags |= LUSTRE_FLD_RUN;
-        fld_cache_delete(fld->lcf_cache, seq);
+       struct lu_seq_range res = { 0 };
+       struct lu_fld_target *target;
+       struct lu_fld_target *origin;
+       int rc;
+       ENTRY;
+
+       rc = fld_cache_lookup(fld->lcf_cache, seq, &res);
+       if (rc == 0) {
+               *mds = res.lsr_index;
+               RETURN(0);
+       }
 
+        /* Can not find it in the cache */
         target = fld_client_get_target(fld, seq);
         LASSERT(target != NULL);
-
-        CDEBUG(D_INFO, "%s: Delete fld entry (seq: "LPX64") on "
+       origin = target;
+again:
+        CDEBUG(D_INFO, "%s: Lookup fld entry (seq: "LPX64") on "
                "target %s (idx "LPU64")\n", fld->lcf_name, seq,
                fld_target_name(target), target->ft_idx);
 
-#ifdef __KERNEL__
-        if (target->ft_srv != NULL) {
-                LASSERT(env != NULL);
-                rc = fld_server_delete(target->ft_srv,
-                                       env, seq);
-        } else {
-#endif
-                rc = fld_client_rpc(target->ft_exp,
-                                    &md_fld, FLD_DELETE);
-#ifdef __KERNEL__
-        }
-#endif
-
-        RETURN(rc);
+       res.lsr_start = seq;
+       fld_range_set_type(&res, flags);
+
+#ifdef HAVE_SERVER_SUPPORT
+       if (target->ft_srv != NULL) {
+               LASSERT(env != NULL);
+               rc = fld_server_lookup(env, target->ft_srv, seq, &res);
+       } else
+#endif /* HAVE_SERVER_SUPPORT */
+       {
+               rc = fld_client_rpc(target->ft_exp, &res, FLD_QUERY, NULL);
+       }
+
+       if (rc == -ESHUTDOWN) {
+               /* If fld lookup failed because the target has been shutdown,
+                * then try next target in the list, until trying all targets
+                * or fld lookup succeeds */
+               spin_lock(&fld->lcf_lock);
+
+               /* If the next entry in the list is the head of the list,
+                * move to the next entry after the head and retrieve
+                * the target. Else retreive the next target entry. */
+
+               if (target->ft_chain.next == &fld->lcf_targets)
+                       target = list_entry(target->ft_chain.next->next,
+                                           struct lu_fld_target, ft_chain);
+               else
+                       target = list_entry(target->ft_chain.next,
+                                                struct lu_fld_target,
+                                                ft_chain);
+               spin_unlock(&fld->lcf_lock);
+               if (target != origin)
+                       goto again;
+       }
+       if (rc == 0) {
+               *mds = res.lsr_index;
+               fld_cache_insert(fld->lcf_cache, &res);
+       }
+
+       RETURN(rc);
 }
-EXPORT_SYMBOL(fld_client_delete);
+EXPORT_SYMBOL(fld_client_lookup);
 
-int fld_client_lookup(struct lu_client_fld *fld,
-                      seqno_t seq, mdsno_t *mds,
-                      const struct lu_env *env)
+void fld_client_flush(struct lu_client_fld *fld)
 {
-        struct md_fld md_fld = { .mf_seq = seq, .mf_mds = 0 };
-        struct lu_fld_target *target;
-        int rc;
-        ENTRY;
+        fld_cache_flush(fld->lcf_cache);
+}
 
-        fld->lcf_flags |= LUSTRE_FLD_RUN;
 
-        rc = fld_cache_lookup(fld->lcf_cache, seq, mds);
-        if (rc == 0)
-                RETURN(0);
+struct proc_dir_entry *fld_type_proc_dir;
 
-        /* Can not find it in the cache */
-        target = fld_client_get_target(fld, seq);
-        LASSERT(target != NULL);
+static int __init fld_init(void)
+{
+       fld_type_proc_dir = lprocfs_register(LUSTRE_FLD_NAME,
+                                            proc_lustre_root,
+                                            NULL, NULL);
+       if (IS_ERR(fld_type_proc_dir))
+               return PTR_ERR(fld_type_proc_dir);
 
-        CDEBUG(D_INFO, "%s: Lookup fld entry (seq: "LPX64") on "
-               "target %s (idx "LPU64")\n", fld->lcf_name, seq,
-               fld_target_name(target), target->ft_idx);
+#ifdef HAVE_SERVER_SUPPORT
+       fld_server_mod_init();
+#endif /* HAVE_SERVER_SUPPORT */
 
-#ifdef __KERNEL__
-        if (target->ft_srv != NULL) {
-                LASSERT(env != NULL);
-                rc = fld_server_lookup(target->ft_srv,
-                                       env, seq, &md_fld.mf_mds);
-        } else {
-#endif
-                /*
-                 * insert the 'inflight' sequence. No need to protect that,
-                 * we are trying to reduce numbers of RPC but not restrict
-                 * to them exactly one 
-                 */
-                fld_cache_insert_inflight(fld->lcf_cache, seq);
-                rc = fld_client_rpc(target->ft_exp,
-                                    &md_fld, FLD_LOOKUP);
-#ifdef __KERNEL__
-        }
-#endif
-        if (rc == 0) {
-                *mds = md_fld.mf_mds;
-
-                /*
-                 * Do not return error here as well. See previous comment in
-                 * same situation in function fld_client_create().
-                 */
-                fld_cache_insert(fld->lcf_cache, seq, *mds);
-        } else {
-                /* remove 'inflight' seq if it exists */
-                fld_cache_delete(fld->lcf_cache, seq);
-        }
-        RETURN(rc);
+       return 0;
 }
-EXPORT_SYMBOL(fld_client_lookup);
 
-void fld_client_flush(struct lu_client_fld *fld)
+static void __exit fld_exit(void)
 {
-#ifdef __KERNEL__
-        fld_cache_flush(fld->lcf_cache);
-#endif
+#ifdef HAVE_SERVER_SUPPORT
+       fld_server_mod_exit();
+#endif /* HAVE_SERVER_SUPPORT */
+
+       if (fld_type_proc_dir != NULL && !IS_ERR(fld_type_proc_dir)) {
+               lprocfs_remove(&fld_type_proc_dir);
+               fld_type_proc_dir = NULL;
+       }
 }
-EXPORT_SYMBOL(fld_client_flush);
 
+MODULE_AUTHOR("OpenSFS, Inc. <http://www.lustre.org/>");
+MODULE_DESCRIPTION("Lustre FID Location Database");
+MODULE_VERSION(LUSTRE_VERSION_STRING);
+MODULE_LICENSE("GPL");
+
+module_init(fld_init);
+module_exit(fld_exit);