Whamcloud - gitweb
LU-1445 fid: Add DATA fid type in fid_request.
[fs/lustre-release.git] / lustre / lmv / lmv_obd.c
index 5346d49..e8044f9 100644 (file)
@@ -1,6 +1,4 @@
-/* -*- 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.
  * GPL HEADER END
  */
 /*
- * Copyright  2008 Sun Microsystems, Inc. All rights reserved
+ * Copyright (c) 2004, 2010, Oracle and/or its affiliates. All rights reserved.
  * Use is subject to license terms.
+ *
+ * Copyright (c) 2011, 2012, Intel Corporation.
  */
 /*
  * This file is part of Lustre, http://www.lustre.org/
  * Lustre is a trademark of Sun Microsystems, Inc.
  */
 
-#ifndef EXPORT_SYMTAB
-# define EXPORT_SYMTAB
-#endif
 #define DEBUG_SUBSYSTEM S_LMV
 #ifdef __KERNEL__
 #include <linux/slab.h>
@@ -53,7 +50,6 @@
 #endif
 
 #include <lustre/lustre_idl.h>
-#include <lustre_log.h>
 #include <obd_support.h>
 #include <lustre_lib.h>
 #include <lustre_net.h>
@@ -97,7 +93,7 @@ static int lmv_set_mdc_active(struct lmv_obd *lmv, struct obd_uuid *uuid,
         CDEBUG(D_INFO, "Searching in lmv %p for uuid %s (activate=%d)\n",
                lmv, uuid->uuid, activate);
 
-        cfs_spin_lock(&lmv->lmv_lock);
+       spin_lock(&lmv->lmv_lock);
         for (i = 0, tgt = lmv->tgts; i < lmv->desc.ld_tgt_count; i++, tgt++) {
                 if (tgt->ltd_exp == NULL)
                         continue;
@@ -133,31 +129,31 @@ static int lmv_set_mdc_active(struct lmv_obd *lmv, struct obd_uuid *uuid,
         EXIT;
 
  out_lmv_lock:
-        cfs_spin_unlock(&lmv->lmv_lock);
-        return rc;
+       spin_unlock(&lmv->lmv_lock);
+       return rc;
 }
 
 static int lmv_set_mdc_data(struct lmv_obd *lmv, struct obd_uuid *uuid,
-                            struct obd_connect_data *data)
+                           struct obd_connect_data *data)
 {
-        struct lmv_tgt_desc    *tgt;
-        int                     i;
-        ENTRY;
+       struct lmv_tgt_desc    *tgt;
+       int                     i;
+       ENTRY;
 
-        LASSERT(data != NULL);
+       LASSERT(data != NULL);
 
-        cfs_spin_lock(&lmv->lmv_lock);
-        for (i = 0, tgt = lmv->tgts; i < lmv->desc.ld_tgt_count; i++, tgt++) {
-                if (tgt->ltd_exp == NULL)
-                        continue;
+       spin_lock(&lmv->lmv_lock);
+       for (i = 0, tgt = lmv->tgts; i < lmv->desc.ld_tgt_count; i++, tgt++) {
+               if (tgt->ltd_exp == NULL)
+                       continue;
 
-                if (obd_uuid_equals(uuid, &tgt->ltd_uuid)) {
-                        lmv->datas[tgt->ltd_idx] = *data;
-                        break;
-                }
-        }
-        cfs_spin_unlock(&lmv->lmv_lock);
-        RETURN(0);
+               if (obd_uuid_equals(uuid, &tgt->ltd_uuid)) {
+                       lmv->datas[tgt->ltd_idx] = *data;
+                       break;
+               }
+       }
+       spin_unlock(&lmv->lmv_lock);
+       RETURN(0);
 }
 
 struct obd_uuid *lmv_get_uuid(struct obd_export *exp) {
@@ -323,7 +319,7 @@ static void lmv_set_timeouts(struct obd_device *obd)
                 if (tgts->ltd_exp == NULL)
                         continue;
 
-                obd_set_info_async(tgts->ltd_exp, sizeof(KEY_INTERMDS),
+                obd_set_info_async(NULL, tgts->ltd_exp, sizeof(KEY_INTERMDS),
                                    KEY_INTERMDS, 0, NULL, NULL);
         }
 }
@@ -417,7 +413,7 @@ int lmv_connect_mdc(struct obd_device *obd, struct lmv_tgt_desc *tgt)
         /*
          * Init fid sequence client for this mdc and add new fld target.
          */
-        rc = obd_fid_init(mdc_exp);
+       rc = obd_fid_init(mdc_exp, LUSTRE_SEQ_METADATA);
         if (rc)
                 RETURN(rc);
 
@@ -518,25 +514,19 @@ int lmv_add_target(struct obd_device *obd, struct obd_uuid *tgt_uuid)
                         CERROR("Target %s not attached\n", tgt_uuid->uuid);
                         RETURN(-EINVAL);
                 }
-
-                rc = obd_llog_init(obd, &obd->obd_olg, mdc_obd, NULL);
-                if (rc) {
-                        lmv_init_unlock(lmv);
-                        CERROR("lmv failed to setup llogging subsystems\n");
-                }
         }
-        cfs_spin_lock(&lmv->lmv_lock);
-        tgt = lmv->tgts + lmv->desc.ld_tgt_count++;
-        tgt->ltd_uuid = *tgt_uuid;
-        cfs_spin_unlock(&lmv->lmv_lock);
-
-        if (lmv->connected) {
-                rc = lmv_connect_mdc(obd, tgt);
-                if (rc) {
-                        cfs_spin_lock(&lmv->lmv_lock);
-                        lmv->desc.ld_tgt_count--;
-                        memset(tgt, 0, sizeof(*tgt));
-                        cfs_spin_unlock(&lmv->lmv_lock);
+       spin_lock(&lmv->lmv_lock);
+       tgt = lmv->tgts + lmv->desc.ld_tgt_count++;
+       tgt->ltd_uuid = *tgt_uuid;
+       spin_unlock(&lmv->lmv_lock);
+
+       if (lmv->connected) {
+               rc = lmv_connect_mdc(obd, tgt);
+               if (rc) {
+                       spin_lock(&lmv->lmv_lock);
+                       lmv->desc.ld_tgt_count--;
+                       memset(tgt, 0, sizeof(*tgt));
+                       spin_unlock(&lmv->lmv_lock);
                 } else {
                         int easize = sizeof(struct lmv_stripe_md) +
                                      lmv->desc.ld_tgt_count *
@@ -568,6 +558,7 @@ int lmv_check_connect(struct obd_device *obd)
         }
 
         if (lmv->desc.ld_tgt_count == 0) {
+                lmv_init_unlock(lmv);
                 CERROR("%s: no targets configured.\n", obd->obd_name);
                 RETURN(-EINVAL);
         }
@@ -626,8 +617,11 @@ static int lmv_disconnect_mdc(struct obd_device *obd, struct lmv_tgt_desc *tgt)
 
         mdc_obd = class_exp2obd(tgt->ltd_exp);
 
-        if (mdc_obd)
+        if (mdc_obd) {
+                mdc_obd->obd_force = obd->obd_force;
+                mdc_obd->obd_fail = obd->obd_fail;
                 mdc_obd->obd_no_recov = obd->obd_no_recov;
+        }
 
 #ifdef __KERNEL__
         lmv_proc_dir = lprocfs_srch(obd->obd_proc_entry, "target_obds");
@@ -754,8 +748,9 @@ static int lmv_iocontrol(unsigned int cmd, struct obd_export *exp,
                                          (int) sizeof(struct obd_uuid))))
                         RETURN(-EFAULT);
 
-                rc = obd_statfs(mdc_obd, &stat_buf,
-                                cfs_time_current_64() - CFS_HZ, 0);
+                rc = obd_statfs(NULL, lmv->tgts[index].ltd_exp, &stat_buf,
+                                cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS),
+                                0);
                 if (rc)
                         RETURN(rc);
                 if (cfs_copy_to_user(data->ioc_pbuf1, &stat_buf,
@@ -829,10 +824,14 @@ static int lmv_iocontrol(unsigned int cmd, struct obd_export *exp,
         default : {
                 for (i = 0; i < count; i++) {
                         int err;
+                        struct obd_device *mdc_obd;
 
                         if (lmv->tgts[i].ltd_exp == NULL)
                                 continue;
-
+                        /* ll_umount_begin() sets force flag but for lmv, not
+                         * mdc. Let's pass it through */
+                        mdc_obd = class_exp2obd(lmv->tgts[i].ltd_exp);
+                        mdc_obd->obd_force = obddev->obd_force;
                         err = obd_iocontrol(cmd, lmv->tgts[i].ltd_exp, len,
                                             karg, uarg);
                         if (err == -ENODATA && cmd == OBD_IOC_POLL_QUOTACHECK) {
@@ -982,7 +981,7 @@ int __lmv_fid_alloc(struct lmv_obd *lmv, struct lu_fid *fid,
          * New seq alloc and FLD setup should be atomic. Otherwise we may find
          * on server that seq in new allocated fid is not yet known.
          */
-        cfs_down(&tgt->ltd_fid_sem);
+       mutex_lock(&tgt->ltd_fid_mutex);
 
         if (!tgt->ltd_active)
                 GOTO(out, rc = -ENODEV);
@@ -998,7 +997,7 @@ int __lmv_fid_alloc(struct lmv_obd *lmv, struct lu_fid *fid,
 
         EXIT;
 out:
-        cfs_up(&tgt->ltd_fid_sem);
+       mutex_unlock(&tgt->ltd_fid_mutex);
         return rc;
 }
 
@@ -1007,7 +1006,7 @@ int lmv_fid_alloc(struct obd_export *exp, struct lu_fid *fid,
 {
         struct obd_device     *obd = class_exp2obd(exp);
         struct lmv_obd        *lmv = &obd->u.lmv;
-        mdsno_t                mds;
+        mdsno_t                mds = 0;
         int                    rc;
         ENTRY;
 
@@ -1030,17 +1029,6 @@ int lmv_fid_alloc(struct obd_export *exp, struct lu_fid *fid,
         RETURN(rc);
 }
 
-static int lmv_fid_delete(struct obd_export *exp, const struct lu_fid *fid)
-{
-        ENTRY;
-        LASSERT(exp != NULL && fid != NULL);
-        if (lmv_object_delete(exp, fid)) {
-                CDEBUG(D_INODE, "Object "DFID" is destroyed.\n",
-                       PFID(fid));
-        }
-        RETURN(0);
-}
-
 static int lmv_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
 {
         struct lmv_obd             *lmv = &obd->u.lmv;
@@ -1069,7 +1057,7 @@ static int lmv_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
                 RETURN(-ENOMEM);
 
         for (i = 0; i < LMV_MAX_TGT_COUNT; i++) {
-                cfs_sema_init(&lmv->tgts[i].ltd_fid_sem, 1);
+               mutex_init(&lmv->tgts[i].ltd_fid_mutex);
                 lmv->tgts[i].ltd_idx = i;
         }
 
@@ -1087,8 +1075,8 @@ static int lmv_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
         lmv->max_easize = 0;
         lmv->lmv_placement = PLACEMENT_CHAR_POLICY;
 
-        cfs_spin_lock_init(&lmv->lmv_lock);
-        cfs_sema_init(&lmv->init_sem, 1);
+       spin_lock_init(&lmv->lmv_lock);
+       mutex_init(&lmv->init_mutex);
 
         rc = lmv_object_setup(obd);
         if (rc) {
@@ -1100,10 +1088,11 @@ static int lmv_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
         lprocfs_obd_setup(obd, lvars.obd_vars);
 #ifdef LPROCFS
         {
-                rc = lprocfs_seq_create(obd->obd_proc_entry, "target_obd_status",
+                rc = lprocfs_seq_create(obd->obd_proc_entry, "target_obd",
                                         0444, &lmv_proc_target_fops, obd);
                 if (rc)
-                        CWARN("Error adding target_obd_stats file (%d)\n", rc);
+                        CWARN("%s: error adding LMV target_obd file: rc = %d\n",
+                               obd->obd_name, rc);
        }
 #endif
         rc = fld_client_init(&lmv->lmv_fld, obd->obd_name,
@@ -1130,7 +1119,6 @@ static int lmv_cleanup(struct obd_device *obd)
         ENTRY;
 
         fld_client_fini(&lmv->lmv_fld);
-        lprocfs_obd_cleanup(obd);
         lmv_object_cleanup(obd);
         OBD_FREE(lmv->datas, lmv->datas_size);
         OBD_FREE(lmv->tgts, lmv->tgts_size);
@@ -1162,9 +1150,10 @@ out:
         RETURN(rc);
 }
 
-static int lmv_statfs(struct obd_device *obd, struct obd_statfs *osfs,
-                      __u64 max_age, __u32 flags)
+static int lmv_statfs(const struct lu_env *env, struct obd_export *exp,
+                      struct obd_statfs *osfs, __u64 max_age, __u32 flags)
 {
+        struct obd_device     *obd = class_exp2obd(exp);
         struct lmv_obd        *lmv = &obd->u.lmv;
         struct obd_statfs     *temp;
         int                    rc = 0;
@@ -1183,7 +1172,7 @@ static int lmv_statfs(struct obd_device *obd, struct obd_statfs *osfs,
                 if (lmv->tgts[i].ltd_exp == NULL)
                         continue;
 
-                rc = obd_statfs(lmv->tgts[i].ltd_exp->exp_obd, temp,
+                rc = obd_statfs(env, lmv->tgts[i].ltd_exp, temp,
                                 max_age, flags);
                 if (rc) {
                         CERROR("can't stat MDS #%d (%s), error %d\n", i,
@@ -1295,10 +1284,10 @@ static int lmv_getattr(struct obd_export *exp, struct md_op_data *op_data,
         if (IS_ERR(tgt))
                 RETURN(PTR_ERR(tgt));
 
-        if (op_data->op_valid & OBD_MD_MDTIDX) {
-                op_data->op_mds = tgt->ltd_idx;
-                RETURN(0);
-        }
+       if (op_data->op_flags & MF_GET_MDT_IDX) {
+               op_data->op_mds = tgt->ltd_idx;
+               RETURN(0);
+       }
 
         rc = md_getattr(tgt->ltd_exp, op_data, request);
         if (rc)
@@ -1732,7 +1721,7 @@ static int
 lmv_enqueue(struct obd_export *exp, struct ldlm_enqueue_info *einfo,
             struct lookup_intent *it, struct md_op_data *op_data,
             struct lustre_handle *lockh, void *lmm, int lmmsize,
-            struct ptlrpc_request **req, int extra_lock_flags)
+           struct ptlrpc_request **req, __u64 extra_lock_flags)
 {
         struct obd_device        *obd = exp->exp_obd;
         struct lmv_obd           *lmv = &obd->u.lmv;
@@ -2064,7 +2053,6 @@ static int lmv_rename(struct obd_export *exp, struct md_op_data *op_data,
         struct obd_device       *obd = exp->exp_obd;
         struct lmv_obd          *lmv = &obd->u.lmv;
         struct lmv_tgt_desc     *src_tgt;
-        struct lmv_tgt_desc     *tgt_tgt;
         int                      rc;
         int                      sidx;
         int                      loop = 0;
@@ -2124,7 +2112,6 @@ repeat:
         op_data->op_cap = cfs_curproc_cap_pack();
 
         src_tgt = lmv_get_target(lmv, mds1);
-        tgt_tgt = lmv_get_target(lmv, mds2);
 
         /*
          * LOOKUP lock on src child (fid3) should also be cancelled for
@@ -2313,7 +2300,7 @@ static void lmv_hash_adjust(__u64 *hash, __u64 hash_adj)
         val = le64_to_cpu(*hash);
         if (val < hash_adj)
                 val += MAX_HASH_SIZE;
-        if (val != DIR_END_OFF)
+        if (val != MDS_DIR_END_OFF)
                 *hash = cpu_to_le64(val - hash_adj);
 }
 
@@ -2334,15 +2321,14 @@ static __u32 lmv_node_rank(struct obd_export *exp, const struct lu_fid *fid)
         return id ^ (id >> 32);
 }
 
-static int lmv_readpage(struct obd_export *exp, const struct lu_fid *fid,
-                        struct obd_capa *oc, __u64 offset64, struct page *page,
-                        struct ptlrpc_request **request)
+static int lmv_readpage(struct obd_export *exp, struct md_op_data *op_data,
+                        struct page **pages, struct ptlrpc_request **request)
 {
         struct obd_device       *obd = exp->exp_obd;
         struct lmv_obd          *lmv = &obd->u.lmv;
-        struct lu_fid            rid = *fid;
         struct lmv_object       *obj;
-        __u64                    offset;
+        struct lu_fid            rid = op_data->op_fid1;
+        __u64                    offset = op_data->op_offset;
         __u64                    hash_adj = 0;
         __u32                    rank = 0;
         __u64                    seg_size = 0;
@@ -2351,14 +2337,17 @@ static int lmv_readpage(struct obd_export *exp, const struct lu_fid *fid,
         int                      tgt0_idx = 0;
         int                      rc;
         int                      nr = 0;
+        int                      i;
+        /* number of pages read, in CFS_PAGE_SIZE */
+        int                      nrdpgs;
+        /* number of pages transferred in LU_PAGE_SIZE */
+        int                      nlupgs;
         struct lmv_stripe       *los;
         struct lmv_tgt_desc     *tgt;
         struct lu_dirpage       *dp;
         struct lu_dirent        *ent;
         ENTRY;
 
-        offset = offset64;
-
         rc = lmv_check_connect(obd);
         if (rc)
                 RETURN(rc);
@@ -2385,7 +2374,7 @@ static int lmv_readpage(struct obd_export *exp, const struct lu_fid *fid,
          * [R*MAX_HASH/N ... (R + 1)*MAX_HASH/N] there for we do hash_adj
          * on hash  values that we get.
          */
-        obj = lmv_object_find_lock(obd, fid);
+        obj = lmv_object_find_lock(obd, &rid);
         if (obj) {
                 nr       = obj->lo_objcount;
                 LASSERT(nr > 0);
@@ -2393,7 +2382,7 @@ static int lmv_readpage(struct obd_export *exp, const struct lu_fid *fid,
                 do_div(seg_size, nr);
                 los      = obj->lo_stripes;
                 tgt      = lmv_get_target(lmv, los[0].ls_mds);
-                rank     = lmv_node_rank(tgt->ltd_exp, fid) % nr;
+                rank     = lmv_node_rank(tgt->ltd_exp, &rid) % nr;
                 tgt_tmp  = offset;
                 do_div(tgt_tmp, seg_size);
                 tgt0_idx = do_div(tgt_tmp,  nr);
@@ -2428,34 +2417,103 @@ static int lmv_readpage(struct obd_export *exp, const struct lu_fid *fid,
         if (IS_ERR(tgt))
                 GOTO(cleanup, rc = PTR_ERR(tgt));
 
-        rc = md_readpage(tgt->ltd_exp, &rid, oc, offset, page, request);
+        op_data->op_fid1 = rid;
+        rc = md_readpage(tgt->ltd_exp, op_data, pages, request);
         if (rc)
                 GOTO(cleanup, rc);
-        if (obj) {
-                dp = cfs_kmap(page);
 
-                lmv_hash_adjust(&dp->ldp_hash_start, hash_adj);
-                lmv_hash_adjust(&dp->ldp_hash_end,   hash_adj);
-                LASSERT(le64_to_cpu(dp->ldp_hash_start) <= offset64);
+        nrdpgs = ((*request)->rq_bulk->bd_nob_transferred + CFS_PAGE_SIZE - 1)
+                 >> CFS_PAGE_SHIFT;
+        nlupgs = (*request)->rq_bulk->bd_nob_transferred >> LU_PAGE_SHIFT;
+        LASSERT(!((*request)->rq_bulk->bd_nob_transferred & ~LU_PAGE_MASK));
+        LASSERT(nrdpgs > 0 && nrdpgs <= op_data->op_npages);
 
-                for (ent = lu_dirent_start(dp); ent != NULL;
-                     ent = lu_dirent_next(ent))
-                        lmv_hash_adjust(&ent->lde_hash, hash_adj);
+        CDEBUG(D_INODE, "read %d(%d)/%d pages\n", nrdpgs, nlupgs,
+               op_data->op_npages);
 
-                if (tgt0_idx != nr - 1) {
-                        __u64 end;
+        for (i = 0; i < nrdpgs; i++) {
+#if CFS_PAGE_SIZE > LU_PAGE_SIZE
+                struct lu_dirpage *first;
+                __u64 hash_end = 0;
+                __u32 flags = 0;
+#endif
+                struct lu_dirent *tmp = NULL;
 
-                        end = le64_to_cpu(dp->ldp_hash_end);
-                        if (end == DIR_END_OFF) {
+                dp = cfs_kmap(pages[i]);
+                if (obj) {
+                        lmv_hash_adjust(&dp->ldp_hash_start, hash_adj);
+                        lmv_hash_adjust(&dp->ldp_hash_end,   hash_adj);
+                        LASSERT(le64_to_cpu(dp->ldp_hash_start) <=
+                                op_data->op_offset);
+
+                        if ((tgt0_idx != nr - 1) &&
+                            (le64_to_cpu(dp->ldp_hash_end) == MDS_DIR_END_OFF))
+                        {
                                 dp->ldp_hash_end = cpu_to_le32(seg_size *
                                                                (tgt0_idx + 1));
                                 CDEBUG(D_INODE,
                                        ""DFID" reset end "LPX64" tgt %d\n",
                                        PFID(&rid),
-                                       (__u64)le64_to_cpu(dp->ldp_hash_end), tgt_idx);
+                                       (__u64)le64_to_cpu(dp->ldp_hash_end),
+                                       tgt_idx);
                         }
                 }
-                cfs_kunmap(page);
+
+                ent = lu_dirent_start(dp);
+#if CFS_PAGE_SIZE > LU_PAGE_SIZE
+                first = dp;
+                hash_end = dp->ldp_hash_end;
+repeat:
+#endif
+                nlupgs--;
+                for (tmp = ent; ent != NULL;
+                     tmp = ent, ent = lu_dirent_next(ent)) {
+                        if (obj)
+                                lmv_hash_adjust(&ent->lde_hash, hash_adj);
+                }
+
+#if CFS_PAGE_SIZE > LU_PAGE_SIZE
+                dp = (struct lu_dirpage *)((char *)dp + LU_PAGE_SIZE);
+                if (((unsigned long)dp & ~CFS_PAGE_MASK) && nlupgs > 0) {
+                        ent = lu_dirent_start(dp);
+
+                        if (obj) {
+                                lmv_hash_adjust(&dp->ldp_hash_end, hash_adj);
+                                if ((tgt0_idx != nr - 1) &&
+                                    (le64_to_cpu(dp->ldp_hash_end) ==
+                                     MDS_DIR_END_OFF)) {
+                                        hash_end = cpu_to_le32(seg_size *
+                                                               (tgt0_idx + 1));
+                                        CDEBUG(D_INODE,
+                                            ""DFID" reset end "LPX64" tgt %d\n",
+                                            PFID(&rid),
+                                            (__u64)le64_to_cpu(hash_end),
+                                            tgt_idx);
+                                }
+                        }
+                        hash_end = dp->ldp_hash_end;
+                        flags = dp->ldp_flags;
+
+                        if (tmp) {
+                                /* enlarge the end entry lde_reclen from 0 to
+                                 * first entry of next lu_dirpage, in this way
+                                 * several lu_dirpages can be stored into one
+                                 * client page on client. */
+                                tmp = ((void *)tmp) +
+                                      le16_to_cpu(tmp->lde_reclen);
+                                tmp->lde_reclen =
+                                        cpu_to_le16((char *)(dp->ldp_entries) -
+                                                    (char *)tmp);
+                                goto repeat;
+                        }
+                }
+                first->ldp_hash_end = hash_end;
+                first->ldp_flags &= ~cpu_to_le32(LDF_COLLIDE);
+                first->ldp_flags |= flags & cpu_to_le32(LDF_COLLIDE);
+#else
+                SET_BUT_UNUSED(tmp);
+#endif
+                cfs_kunmap(pages[i]);
         }
         EXIT;
 cleanup:
@@ -2550,7 +2608,8 @@ repeat:
 
 static int lmv_precleanup(struct obd_device *obd, enum obd_cleanup_stage stage)
 {
-        int        rc = 0;
+        struct lmv_obd *lmv = &obd->u.lmv;
+        int rc = 0;
 
         switch (stage) {
         case OBD_CLEANUP_EARLY:
@@ -2558,9 +2617,8 @@ static int lmv_precleanup(struct obd_device *obd, enum obd_cleanup_stage stage)
                  * stack. */
                 break;
         case OBD_CLEANUP_EXPORTS:
-                rc = obd_llog_finish(obd, 0);
-                if (rc != 0)
-                        CERROR("failed to cleanup llogging subsystems\n");
+                fld_client_proc_fini(&lmv->lmv_fld);
+                lprocfs_obd_cleanup(obd);
                 break;
         default:
                 break;
@@ -2568,8 +2626,8 @@ static int lmv_precleanup(struct obd_device *obd, enum obd_cleanup_stage stage)
         RETURN(rc);
 }
 
-static int lmv_get_info(struct obd_export *exp, __u32 keylen,
-                        void *key, __u32 *vallen, void *val,
+static int lmv_get_info(const struct lu_env *env, struct obd_export *exp,
+                        __u32 keylen, void *key, __u32 *vallen, void *val,
                         struct lov_stripe_md *lsm)
 {
         struct obd_device       *obd;
@@ -2605,7 +2663,7 @@ static int lmv_get_info(struct obd_export *exp, __u32 keylen,
                                 continue;
                         }
 
-                        if (!obd_get_info(tgts->ltd_exp, keylen, key,
+                        if (!obd_get_info(env, tgts->ltd_exp, keylen, key,
                                           vallen, val, NULL))
                                 RETURN(0);
                 }
@@ -2619,22 +2677,25 @@ static int lmv_get_info(struct obd_export *exp, __u32 keylen,
                  * Forwarding this request to first MDS, it should know LOV
                  * desc.
                  */
-                rc = obd_get_info(lmv->tgts[0].ltd_exp, keylen, key,
+                rc = obd_get_info(env, lmv->tgts[0].ltd_exp, keylen, key,
                                   vallen, val, NULL);
                 if (!rc && KEY_IS(KEY_CONN_DATA)) {
                         exp->exp_connect_flags =
                         ((struct obd_connect_data *)val)->ocd_connect_flags;
                 }
                 RETURN(rc);
+        } else if (KEY_IS(KEY_TGT_COUNT)) {
+                *((int *)val) = lmv->desc.ld_tgt_count;
+                RETURN(0);
         }
 
         CDEBUG(D_IOCTL, "Invalid key\n");
         RETURN(-EINVAL);
 }
 
-int lmv_set_info_async(struct obd_export *exp, obd_count keylen,
-                       void *key, obd_count vallen, void *val,
-                       struct ptlrpc_request_set *set)
+int lmv_set_info_async(const struct lu_env *env, struct obd_export *exp,
+                       obd_count keylen, void *key, obd_count vallen,
+                       void *val, struct ptlrpc_request_set *set)
 {
         struct lmv_tgt_desc    *tgt;
         struct obd_device      *obd;
@@ -2659,7 +2720,7 @@ int lmv_set_info_async(struct obd_export *exp, obd_count keylen,
                         if (!tgt->ltd_exp)
                                 continue;
 
-                        err = obd_set_info_async(tgt->ltd_exp,
+                        err = obd_set_info_async(env, tgt->ltd_exp,
                                                  keylen, key, vallen, val, set);
                         if (err && rc == 0)
                                 rc = err;
@@ -2687,13 +2748,13 @@ int lmv_packmd(struct obd_export *exp, struct lov_mds_md **lmmp,
                 RETURN(mea_size);
 
         if (*lmmp && !lsm) {
-                OBD_FREE(*lmmp, mea_size);
+                OBD_FREE_LARGE(*lmmp, mea_size);
                 *lmmp = NULL;
                 RETURN(0);
         }
 
         if (*lmmp == NULL) {
-                OBD_ALLOC(*lmmp, mea_size);
+                OBD_ALLOC_LARGE(*lmmp, mea_size);
                 if (*lmmp == NULL)
                         RETURN(-ENOMEM);
         }
@@ -2737,14 +2798,14 @@ int lmv_unpackmd(struct obd_export *exp, struct lov_stripe_md **lsmp,
                 return mea_size;
 
         if (*lsmp != NULL && lmm == NULL) {
-                OBD_FREE(*tmea, mea_size);
+                OBD_FREE_LARGE(*tmea, mea_size);
                 *lsmp = NULL;
                 RETURN(0);
         }
 
         LASSERT(mea_size == lmm_size);
 
-        OBD_ALLOC(*tmea, mea_size);
+        OBD_ALLOC_LARGE(*tmea, mea_size);
         if (*tmea == NULL)
                 RETURN(-ENOMEM);
 
@@ -2801,7 +2862,7 @@ static int lmv_cancel_unused(struct obd_export *exp, const struct lu_fid *fid,
 }
 
 int lmv_set_lock_data(struct obd_export *exp, __u64 *lockh, void *data,
-                      __u32 *bits)
+                      __u64 *bits)
 {
         struct obd_device       *obd = exp->exp_obd;
         struct lmv_obd          *lmv = &obd->u.lmv;
@@ -2812,7 +2873,7 @@ int lmv_set_lock_data(struct obd_export *exp, __u64 *lockh, void *data,
         RETURN(rc);
 }
 
-ldlm_mode_t lmv_lock_match(struct obd_export *exp, int flags,
+ldlm_mode_t lmv_lock_match(struct obd_export *exp, __u64 flags,
                            const struct lu_fid *fid, ldlm_type_t type,
                            ldlm_policy_data_t *policy, ldlm_mode_t mode,
                            struct lustre_handle *lockh)
@@ -2959,7 +3020,7 @@ int lmv_intent_getattr_async(struct obd_export *exp,
         struct obd_device       *obd = exp->exp_obd;
         struct lmv_obd          *lmv = &obd->u.lmv;
         struct lmv_object       *obj;
-        struct lmv_tgt_desc     *tgt;
+        struct lmv_tgt_desc     *tgt = NULL;
         int                      rc;
         int                      sidx;
         ENTRY;
@@ -2968,36 +3029,21 @@ int lmv_intent_getattr_async(struct obd_export *exp,
         if (rc)
                 RETURN(rc);
 
-        if (!fid_is_sane(&op_data->op_fid2)) {
+        if (op_data->op_namelen) {
                 obj = lmv_object_find(obd, &op_data->op_fid1);
-                if (obj && op_data->op_namelen) {
-                        sidx = raw_name2idx(obj->lo_hashtype,
-                                            obj->lo_objcount,
+                if (obj) {
+                        sidx = raw_name2idx(obj->lo_hashtype, obj->lo_objcount,
                                             (char *)op_data->op_name,
                                             op_data->op_namelen);
                         op_data->op_fid1 = obj->lo_stripes[sidx].ls_fid;
-                        tgt = lmv_get_target(lmv,
-                                             obj->lo_stripes[sidx].ls_mds);
-                        CDEBUG(D_INODE,
-                               "Choose slave dir ("DFID") -> mds #%d\n",
-                               PFID(&op_data->op_fid1), tgt->ltd_idx);
-                } else {
-                        tgt = lmv_find_target(lmv, &op_data->op_fid1);
-                }
-                if (obj)
+                        tgt = lmv_get_target(lmv, obj->lo_stripes[sidx].ls_mds);
                         lmv_object_put(obj);
-        } else {
-                op_data->op_fid1 = op_data->op_fid2;
-                tgt = lmv_find_target(lmv, &op_data->op_fid2);
-                op_data->op_bias = MDS_CROSS_REF;
-                /*
-                 * Unfortunately, we have to lie to MDC/MDS to retrieve
-                 * attributes llite needs.
-                */
-                if (minfo->mi_it.it_op & IT_LOOKUP)
-                        minfo->mi_it.it_op = IT_GETATTR;
+                }
         }
 
+        if (tgt == NULL)
+                tgt = lmv_find_target(lmv, &op_data->op_fid1);
+
         if (IS_ERR(tgt))
                 RETURN(PTR_ERR(tgt));
 
@@ -3006,7 +3052,7 @@ int lmv_intent_getattr_async(struct obd_export *exp,
 }
 
 int lmv_revalidate_lock(struct obd_export *exp, struct lookup_intent *it,
-                        struct lu_fid *fid, __u32 *bits)
+                        struct lu_fid *fid, __u64 *bits)
 {
         struct obd_device       *obd = exp->exp_obd;
         struct lmv_obd          *lmv = &obd->u.lmv;
@@ -3026,6 +3072,83 @@ int lmv_revalidate_lock(struct obd_export *exp, struct lookup_intent *it,
         RETURN(rc);
 }
 
+/**
+ * For lmv, only need to send request to master MDT, and the master MDT will
+ * process with other slave MDTs. The only exception is Q_GETOQUOTA for which
+ * we directly fetch data from the slave MDTs.
+ */
+int lmv_quotactl(struct obd_device *unused, struct obd_export *exp,
+                 struct obd_quotactl *oqctl)
+{
+        struct obd_device   *obd = class_exp2obd(exp);
+        struct lmv_obd      *lmv = &obd->u.lmv;
+        struct lmv_tgt_desc *tgt = &lmv->tgts[0];
+        int                  rc = 0, i;
+        __u64                curspace, curinodes;
+        ENTRY;
+
+        if (!lmv->desc.ld_tgt_count || !tgt->ltd_active) {
+                CERROR("master lmv inactive\n");
+                RETURN(-EIO);
+        }
+
+        if (oqctl->qc_cmd != Q_GETOQUOTA) {
+                rc = obd_quotactl(tgt->ltd_exp, oqctl);
+                RETURN(rc);
+        }
+
+        curspace = curinodes = 0;
+        for (i = 0; i < lmv->desc.ld_tgt_count; i++) {
+                int err;
+                tgt = &lmv->tgts[i];
+
+                if (tgt->ltd_exp == NULL)
+                        continue;
+                if (!tgt->ltd_active) {
+                        CDEBUG(D_HA, "mdt %d is inactive.\n", i);
+                        continue;
+                }
+
+                err = obd_quotactl(tgt->ltd_exp, oqctl);
+                if (err) {
+                        CERROR("getquota on mdt %d failed. %d\n", i, err);
+                        if (!rc)
+                                rc = err;
+                } else {
+                        curspace += oqctl->qc_dqblk.dqb_curspace;
+                        curinodes += oqctl->qc_dqblk.dqb_curinodes;
+                }
+        }
+        oqctl->qc_dqblk.dqb_curspace = curspace;
+        oqctl->qc_dqblk.dqb_curinodes = curinodes;
+
+        RETURN(rc);
+}
+
+int lmv_quotacheck(struct obd_device *unused, struct obd_export *exp,
+                   struct obd_quotactl *oqctl)
+{
+        struct obd_device   *obd = class_exp2obd(exp);
+        struct lmv_obd      *lmv = &obd->u.lmv;
+        struct lmv_tgt_desc *tgt;
+        int                  i, rc = 0;
+        ENTRY;
+
+        for (i = 0, tgt = lmv->tgts; i < lmv->desc.ld_tgt_count; i++, tgt++) {
+                int err;
+
+                if (!tgt->ltd_active) {
+                        CERROR("lmv idx %d inactive\n", i);
+                        RETURN(-EIO);
+                }
+
+                err = obd_quotacheck(tgt->ltd_exp, oqctl);
+                if (err && !rc)
+                        rc = err;
+        }
+
+        RETURN(rc);
+}
 
 struct obd_ops lmv_obd_ops = {
         .o_owner                = THIS_MODULE,
@@ -3043,7 +3166,8 @@ struct obd_ops lmv_obd_ops = {
         .o_notify               = lmv_notify,
         .o_get_uuid             = lmv_get_uuid,
         .o_iocontrol            = lmv_iocontrol,
-        .o_fid_delete           = lmv_fid_delete
+        .o_quotacheck           = lmv_quotacheck,
+        .o_quotactl             = lmv_quotactl
 };
 
 struct md_ops lmv_md_ops = {
@@ -3080,9 +3204,6 @@ struct md_ops lmv_md_ops = {
         .m_revalidate_lock      = lmv_revalidate_lock
 };
 
-static quota_interface_t *quota_interface;
-extern quota_interface_t lmv_quota_interface;
-
 int __init lmv_init(void)
 {
         struct lprocfs_static_vars lvars;
@@ -3098,17 +3219,10 @@ int __init lmv_init(void)
 
         lprocfs_lmv_init_vars(&lvars);
 
-        cfs_request_module("lquota");
-        quota_interface = PORTAL_SYMBOL_GET(lmv_quota_interface);
-        init_obd_quota_ops(quota_interface, &lmv_obd_ops);
-
         rc = class_register_type(&lmv_obd_ops, &lmv_md_ops,
                                  lvars.module_vars, LUSTRE_LMV_NAME, NULL);
-        if (rc) {
-                if (quota_interface)
-                        PORTAL_SYMBOL_PUT(lmv_quota_interface);
+        if (rc)
                 cfs_mem_cache_destroy(lmv_object_cache);
-        }
 
         return rc;
 }
@@ -3116,9 +3230,6 @@ int __init lmv_init(void)
 #ifdef __KERNEL__
 static void lmv_exit(void)
 {
-        if (quota_interface)
-                PORTAL_SYMBOL_PUT(lmv_quota_interface);
-
         class_unregister_type(LUSTRE_LMV_NAME);
 
         LASSERTF(cfs_atomic_read(&lmv_object_count) == 0,