Whamcloud - gitweb
LU-1943 filter: remove obdfilter from tree
authorMikhail Pershin <tappro@whamcloud.com>
Sat, 6 Oct 2012 16:02:17 +0000 (20:02 +0400)
committerOleg Drokin <green@whamcloud.com>
Mon, 8 Oct 2012 07:34:28 +0000 (03:34 -0400)
Remove obdfilter and keep only OFD

Signed-off-by: Mikhail Pershin <tappro@whamcloud.com>
Change-Id: If72fe59238325cc63f804f84804b180156c94ea9
Reviewed-on: http://review.whamcloud.com/4209
Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Tested-by: Hudson
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com>
16 files changed:
lustre/Makefile.in
lustre/autoMakefile.am
lustre/autoconf/lustre-core.m4
lustre/obdclass/genops.c
lustre/obdfilter/Makefile.in [deleted file]
lustre/obdfilter/autoMakefile.am [deleted file]
lustre/obdfilter/filter.c [deleted file]
lustre/obdfilter/filter_capa.c [deleted file]
lustre/obdfilter/filter_internal.h [deleted file]
lustre/obdfilter/filter_io.c [deleted file]
lustre/obdfilter/filter_io_26.c [deleted file]
lustre/obdfilter/filter_log.c [deleted file]
lustre/obdfilter/filter_lvb.c [deleted file]
lustre/obdfilter/lproc_obdfilter.c [deleted file]
lustre/tests/conf-sanity.sh
lustre/tests/test-framework.sh

index 1405a62..b3f6f60 100644 (file)
@@ -7,7 +7,7 @@ subdir-m += mgc
 
 @SERVER_TRUE@subdir-m += mds ost mgs mdt cmm mdd ofd quota osp lod
 @CLIENT_TRUE@subdir-m += lov osc mdc lmv llite fld
-@LDISKFS_ENABLED_TRUE@subdir-m += obdfilter osd-ldiskfs
+@LDISKFS_ENABLED_TRUE@subdir-m += osd-ldiskfs
 @ZFS_ENABLED_TRUE@subdir-m += osd-zfs
 @OSDADDON@
 
index 779fe6f..57029c8 100644 (file)
@@ -42,7 +42,7 @@ AUTOMAKE_OPTIONS = foreign
 ALWAYS_SUBDIRS = include lvfs obdclass ldlm ptlrpc obdecho \
        mgc fid fld doc utils tests scripts autoconf contrib conf
 
-SERVER_SUBDIRS = obdfilter ost mds mgs mdt cmm mdd ofd osd-zfs osd-ldiskfs \
+SERVER_SUBDIRS = ost mds mgs mdt cmm mdd ofd osd-zfs osd-ldiskfs \
        quota osp lod
 
 CLIENT_SUBDIRS = mdc lmv llite lclient lov osc
index b319312..c8cc45f 100644 (file)
@@ -2507,8 +2507,6 @@ lustre/obdclass/autoMakefile
 lustre/obdclass/linux/Makefile
 lustre/obdecho/Makefile
 lustre/obdecho/autoMakefile
-lustre/obdfilter/Makefile
-lustre/obdfilter/autoMakefile
 lustre/ofd/Makefile
 lustre/ofd/autoMakefile
 lustre/osc/Makefile
index 680a00d..6fd2624 100644 (file)
@@ -121,6 +121,10 @@ struct obd_type *class_get_type(const char *name)
 #ifdef HAVE_MODULE_LOADING_SUPPORT
         if (!type) {
                 const char *modname = name;
+
+               if (strcmp(modname, "obdfilter") == 0)
+                       modname = "ofd";
+
                 if (!cfs_request_module("%s", modname)) {
                         CDEBUG(D_INFO, "Loaded module '%s'\n", modname);
                         type = class_search_type(name);
diff --git a/lustre/obdfilter/Makefile.in b/lustre/obdfilter/Makefile.in
deleted file mode 100644 (file)
index c1ebc18..0000000
+++ /dev/null
@@ -1,9 +0,0 @@
-MODULES := obdfilter
-
-obdfilter-objs := filter.o filter_io.o filter_log.o
-obdfilter-objs += lproc_obdfilter.o filter_lvb.o filter_capa.o
-obdfilter-objs += filter_io_26.o
-
-EXTRA_DIST = $(obdfilter-objs:%.o=%.c) filter_io_26.c filter_internal.h
-
-@INCLUDE_RULES@
diff --git a/lustre/obdfilter/autoMakefile.am b/lustre/obdfilter/autoMakefile.am
deleted file mode 100644 (file)
index 3896641..0000000
+++ /dev/null
@@ -1,43 +0,0 @@
-#
-# GPL HEADER START
-#
-# DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
-#
-# This program is free software; you can redistribute it and/or modify
-# it under the terms of the GNU General Public License version 2 only,
-# as published by the Free Software Foundation.
-#
-# This program is distributed in the hope that it will be useful, but
-# WITHOUT ANY WARRANTY; without even the implied warranty of
-# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-# General Public License version 2 for more details (a copy is included
-# in the LICENSE file that accompanied this code).
-#
-# You should have received a copy of the GNU General Public License
-# version 2 along with this program; If not, see
-# http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
-#
-# Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
-# CA 95054 USA or visit www.sun.com if you need additional information or
-# have any questions.
-#
-# GPL HEADER END
-#
-
-#
-# Copyright (c) 2004, 2010, Oracle and/or its affiliates. All rights reserved.
-# Use is subject to license terms.
-#
-
-#
-# This file is part of Lustre, http://www.lustre.org/
-# Lustre is a trademark of Sun Microsystems, Inc.
-#
-
-if MODULES
-if LDISKFS_ENABLED
-modulefs_DATA = obdfilter$(KMODEXT)
-endif
-endif
-
-MOSTLYCLEANFILES := @MOSTLYCLEANFILES@ 
diff --git a/lustre/obdfilter/filter.c b/lustre/obdfilter/filter.c
deleted file mode 100644 (file)
index e213d3e..0000000
+++ /dev/null
@@ -1,4908 +0,0 @@
-/*
- * GPL HEADER START
- *
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License version 2 only,
- * as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License version 2 for more details (a copy is included
- * in the LICENSE file that accompanied this code).
- *
- * You should have received a copy of the GNU General Public License
- * version 2 along with this program; If not, see
- * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
- *
- * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
- * CA 95054 USA or visit www.sun.com if you need additional information or
- * have any questions.
- *
- * GPL HEADER END
- */
-/*
- * Copyright (c) 2001, 2010, Oracle and/or its affiliates. All rights reserved.
- * Use is subject to license terms.
- *
- * Copyright (c) 2011, 2012, Whamcloud, Inc.
- */
-/*
- * This file is part of Lustre, http://www.lustre.org/
- * Lustre is a trademark of Sun Microsystems, Inc.
- *
- * lustre/obdfilter/filter.c
- *
- * Author: Peter Braam <braam@clusterfs.com>
- * Author: Andreas Dilger <adilger@clusterfs.com>
- */
-
-/*
- * Invariant: Get O/R i_mutex for lookup, if needed, before any journal ops
- *            (which need to get journal_lock, may block if journal full).
- *
- * Invariant: Call filter_start_transno() before any journal ops to avoid the
- *            same deadlock problem.  We can (and want) to get rid of the
- *            transno sem in favour of the dir/inode i_mutex to avoid single
- *            threaded operation on the OST.
- */
-
-#define DEBUG_SUBSYSTEM S_FILTER
-
-#include <linux/module.h>
-#include <linux/fs.h>
-#include <linux/dcache.h>
-#include <linux/init.h>
-#include <linux/version.h>
-#include <linux/sched.h>
-#include <linux/mount.h>
-#include <linux/buffer_head.h>
-
-#include <obd_cksum.h>
-#include <obd_class.h>
-#include <obd_lov.h>
-#include <lustre_dlm.h>
-#include <lustre_fsfilt.h>
-#include <lprocfs_status.h>
-#include <lustre_log.h>
-#include <libcfs/list.h>
-#include <lustre_disk.h>
-#include <linux/slab.h>
-#include <lustre_param.h>
-#include <lustre/ll_fiemap.h>
-
-#include "filter_internal.h"
-
-static struct lvfs_callback_ops filter_lvfs_ops;
-cfs_mem_cache_t *ll_fmd_cachep;
-
-static void filter_commit_cb(struct obd_device *obd, __u64 transno,
-                             void *cb_data, int error)
-{
-        struct obd_export *exp = cb_data;
-        LASSERT(exp->exp_obd == obd);
-        obd_transno_commit_cb(obd, transno, exp, error);
-        class_export_cb_put(exp);
-}
-
-int filter_version_get_check(struct obd_export *exp,
-                             struct obd_trans_info *oti, struct inode *inode)
-{
-        __u64 curr_version;
-
-        if (inode == NULL || oti == NULL)
-                RETURN(0);
-
-        curr_version = fsfilt_get_version(exp->exp_obd, inode);
-        if ((__s64)curr_version == -EOPNOTSUPP)
-                RETURN(0);
-        /* VBR: version is checked always because costs nothing */
-        if (oti->oti_pre_version != 0 &&
-            oti->oti_pre_version != curr_version) {
-                CDEBUG(D_INODE, "Version mismatch "LPX64" != "LPX64"\n",
-                       oti->oti_pre_version, curr_version);
-                cfs_spin_lock(&exp->exp_lock);
-                exp->exp_vbr_failed = 1;
-                cfs_spin_unlock(&exp->exp_lock);
-                RETURN (-EOVERFLOW);
-        }
-        oti->oti_pre_version = curr_version;
-        RETURN(0);
-}
-
-/* Assumes caller has already pushed us into the kernel context. */
-int filter_finish_transno(struct obd_export *exp, struct inode *inode,
-                          struct obd_trans_info *oti, int rc, int force_sync)
-{
-        struct obd_device_target *obt = &exp->exp_obd->u.obt;
-        struct tg_export_data *ted = &exp->exp_target_data;
-        struct lr_server_data *lsd = class_server_data(exp->exp_obd);
-        struct lsd_client_data *lcd;
-        __u64 last_rcvd;
-        loff_t off;
-        int err, log_pri = D_RPCTRACE;
-
-        /* Propagate error code. */
-        if (rc)
-                RETURN(rc);
-
-        if (!exp->exp_obd->obd_replayable || oti == NULL)
-                RETURN(rc);
-
-        cfs_mutex_lock(&ted->ted_lcd_lock);
-        lcd = ted->ted_lcd;
-        /* if the export has already been disconnected, we have no last_rcvd slot,
-         * update server data with latest transno then */
-        if (lcd == NULL) {
-                cfs_mutex_unlock(&ted->ted_lcd_lock);
-                CWARN("commit transaction for disconnected client %s: rc %d\n",
-                      exp->exp_client_uuid.uuid, rc);
-                err = filter_update_server_data(exp->exp_obd);
-                RETURN(err);
-        }
-
-        /* we don't allocate new transnos for replayed requests */
-        cfs_spin_lock(&obt->obt_lut->lut_translock);
-        if (oti->oti_transno == 0) {
-                last_rcvd = le64_to_cpu(lsd->lsd_last_transno) + 1;
-                lsd->lsd_last_transno = cpu_to_le64(last_rcvd);
-                LASSERT(last_rcvd >= le64_to_cpu(lcd->lcd_last_transno));
-        } else {
-                last_rcvd = oti->oti_transno;
-                if (last_rcvd > le64_to_cpu(lsd->lsd_last_transno))
-                        lsd->lsd_last_transno = cpu_to_le64(last_rcvd);
-                if (unlikely(last_rcvd < le64_to_cpu(lcd->lcd_last_transno))) {
-                        CERROR("Trying to overwrite bigger transno, on-disk: "
-                               LPU64", new: "LPU64"\n",
-                               le64_to_cpu(lcd->lcd_last_transno), last_rcvd);
-                        cfs_spin_lock(&exp->exp_lock);
-                        exp->exp_vbr_failed = 1;
-                        cfs_spin_unlock(&exp->exp_lock);
-                        cfs_spin_unlock(&obt->obt_lut->lut_translock);
-                        cfs_mutex_unlock(&ted->ted_lcd_lock);
-                        RETURN(-EOVERFLOW);
-                }
-        }
-        oti->oti_transno = last_rcvd;
-
-        lcd->lcd_last_transno = cpu_to_le64(last_rcvd);
-        lcd->lcd_pre_versions[0] = cpu_to_le64(oti->oti_pre_version);
-        lcd->lcd_last_xid = cpu_to_le64(oti->oti_xid);
-        cfs_spin_unlock(&obt->obt_lut->lut_translock);
-
-        if (inode)
-                fsfilt_set_version(exp->exp_obd, inode, last_rcvd);
-
-        off = ted->ted_lr_off;
-        if (off <= 0) {
-                CERROR("%s: client idx %d is %lld\n", exp->exp_obd->obd_name,
-                       ted->ted_lr_idx, ted->ted_lr_off);
-                err = -EINVAL;
-        } else {
-                class_export_cb_get(exp); /* released when the cb is called */
-                if (!force_sync)
-                        force_sync = fsfilt_add_journal_cb(exp->exp_obd,
-                                                           last_rcvd,
-                                                           oti->oti_handle,
-                                                           filter_commit_cb,
-                                                           exp);
-
-                err = fsfilt_write_record(exp->exp_obd, obt->obt_rcvd_filp,
-                                          lcd, sizeof(*lcd), &off,
-                                          force_sync | exp->exp_need_sync);
-                if (force_sync)
-                        filter_commit_cb(exp->exp_obd, last_rcvd, exp, err);
-        }
-        if (err) {
-                log_pri = D_ERROR;
-                if (rc == 0)
-                        rc = err;
-        }
-
-        CDEBUG(log_pri, "wrote trans "LPU64" for client %s at #%d: err = %d\n",
-               last_rcvd, lcd->lcd_uuid, ted->ted_lr_idx, err);
-        cfs_mutex_unlock(&ted->ted_lcd_lock);
-        RETURN(rc);
-}
-
-void f_dput(struct dentry *dentry)
-{
-        /* Can't go inside filter_ddelete because it can block */
-        CDEBUG(D_INODE, "putting %s: %p, count = %d\n",
-               dentry->d_name.name, dentry, atomic_read(&dentry->d_count) - 1);
-        LASSERT(atomic_read(&dentry->d_count) > 0);
-
-        dput(dentry);
-}
-
-static void init_brw_stats(struct brw_stats *brw_stats)
-{
-        int i;
-        for (i = 0; i < BRW_LAST; i++)
-                cfs_spin_lock_init(&brw_stats->hist[i].oh_lock);
-}
-
-static int lprocfs_init_rw_stats(struct obd_device *obd,
-                                 struct lprocfs_stats **stats)
-{
-        int num_stats;
-
-        num_stats = (sizeof(*obd->obd_type->typ_dt_ops) / sizeof(void *)) +
-                                                        LPROC_FILTER_LAST - 1;
-        *stats = lprocfs_alloc_stats(num_stats, LPROCFS_STATS_FLAG_NOPERCPU);
-        if (*stats == NULL)
-                return -ENOMEM;
-
-        lprocfs_init_ops_stats(LPROC_FILTER_LAST, *stats);
-        lprocfs_counter_init(*stats, LPROC_FILTER_READ_BYTES,
-                             LPROCFS_CNTR_AVGMINMAX, "read_bytes", "bytes");
-        lprocfs_counter_init(*stats, LPROC_FILTER_WRITE_BYTES,
-                             LPROCFS_CNTR_AVGMINMAX, "write_bytes", "bytes");
-
-        return(0);
-}
-
-/* brw_stats are 2128, ops are 3916, ldlm are 204, so 6248 bytes per client,
-   plus the procfs overhead :( */
-static int filter_export_stats_init(struct obd_device *obd,
-                                    struct obd_export *exp,
-                                    void *client_nid)
-{
-        int rc, newnid = 0;
-        ENTRY;
-
-        if (obd_uuid_equals(&exp->exp_client_uuid, &obd->obd_uuid))
-                /* Self-export gets no proc entry */
-                RETURN(0);
-
-        rc = lprocfs_exp_setup(exp, client_nid, &newnid);
-        if (rc) {
-                /* Mask error for already created
-                 * /proc entries */
-                if (rc == -EALREADY)
-                        rc = 0;
-                RETURN(rc);
-        }
-
-        if (newnid) {
-                struct nid_stat *tmp = exp->exp_nid_stats;
-                LASSERT(tmp != NULL);
-
-                OBD_ALLOC(tmp->nid_brw_stats, sizeof(struct brw_stats));
-                if (tmp->nid_brw_stats == NULL)
-                        GOTO(clean, rc = -ENOMEM);
-
-                init_brw_stats(tmp->nid_brw_stats);
-                rc = lprocfs_seq_create(exp->exp_nid_stats->nid_proc, "brw_stats",
-                                        0644, &filter_per_nid_stats_fops,
-                                        exp->exp_nid_stats);
-                if (rc)
-                        CWARN("Error adding the brw_stats file\n");
-
-                rc = lprocfs_init_rw_stats(obd, &exp->exp_nid_stats->nid_stats);
-                if (rc)
-                        GOTO(clean, rc);
-
-                rc = lprocfs_register_stats(tmp->nid_proc, "stats",
-                                            tmp->nid_stats);
-                if (rc)
-                        GOTO(clean, rc);
-                rc = lprocfs_nid_ldlm_stats_init(tmp);
-                if (rc)
-                        GOTO(clean, rc);
-        }
-
-        RETURN(0);
- clean:
-        return rc;
-}
-
-/* Add client data to the FILTER.  We use a bitmap to locate a free space
- * in the last_rcvd file if cl_idx is -1 (i.e. a new client).
- * Otherwise, we have just read the data from the last_rcvd file and
- * we know its offset. */
-static int filter_client_add(struct obd_device *obd, struct obd_export *exp,
-                             int cl_idx)
-{
-        struct obd_device_target *obt = &obd->u.obt;
-        struct tg_export_data *ted = &exp->exp_target_data;
-        struct lr_server_data *lsd = class_server_data(obd);
-        unsigned long *bitmap = obt->obt_lut->lut_client_bitmap;
-        int new_client = (cl_idx == -1);
-
-        ENTRY;
-
-        LASSERT(bitmap != NULL);
-        LASSERTF(cl_idx > -2, "%d\n", cl_idx);
-
-        /* Self-export */
-        if (strcmp(ted->ted_lcd->lcd_uuid, obd->obd_uuid.uuid) == 0)
-                RETURN(0);
-
-        /* the bitmap operations can handle cl_idx > sizeof(long) * 8, so
-         * there's no need for extra complication here
-         */
-        if (new_client) {
-                cl_idx = cfs_find_first_zero_bit(bitmap, LR_MAX_CLIENTS);
-        repeat:
-                if (cl_idx >= LR_MAX_CLIENTS) {
-                        CERROR("no room for %u client - fix LR_MAX_CLIENTS\n",
-                               cl_idx);
-                        RETURN(-EOVERFLOW);
-                }
-                if (cfs_test_and_set_bit(cl_idx, bitmap)) {
-                        cl_idx = cfs_find_next_zero_bit(bitmap, LR_MAX_CLIENTS,
-                                                        cl_idx);
-                        goto repeat;
-                }
-        } else {
-                if (cfs_test_and_set_bit(cl_idx, bitmap)) {
-                        CERROR("FILTER client %d: bit already set in bitmap!\n",
-                               cl_idx);
-                        LBUG();
-                }
-        }
-
-        ted->ted_lr_idx = cl_idx;
-        ted->ted_lr_off = le32_to_cpu(lsd->lsd_client_start) +
-                          cl_idx * le16_to_cpu(lsd->lsd_client_size);
-        cfs_mutex_init(&ted->ted_lcd_lock);
-        LASSERTF(ted->ted_lr_off > 0, "ted_lr_off = %llu\n", ted->ted_lr_off);
-
-        CDEBUG(D_INFO, "client at index %d (%llu) with UUID '%s' added\n",
-               ted->ted_lr_idx, ted->ted_lr_off, ted->ted_lcd->lcd_uuid);
-
-        if (new_client) {
-                struct lvfs_run_ctxt saved;
-                loff_t off = ted->ted_lr_off;
-                int rc;
-                void *handle;
-
-                CDEBUG(D_INFO, "writing client lcd at idx %u (%llu) (len %u)\n",
-                       ted->ted_lr_idx,off,(unsigned int)sizeof(*ted->ted_lcd));
-
-                if (OBD_FAIL_CHECK(OBD_FAIL_TGT_CLIENT_ADD))
-                        RETURN(-ENOSPC);
-
-                push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-                /* Transaction needed to fix bug 1403 */
-                handle = fsfilt_start(obd,
-                                      obt->obt_rcvd_filp->f_dentry->d_inode,
-                                      FSFILT_OP_SETATTR, NULL);
-                if (IS_ERR(handle)) {
-                        rc = PTR_ERR(handle);
-                        CERROR("unable to start transaction: rc %d\n", rc);
-                } else {
-                        ted->ted_lcd->lcd_last_epoch = lsd->lsd_start_epoch;
-                        exp->exp_last_request_time = cfs_time_current_sec();
-                        rc = fsfilt_add_journal_cb(obd, 0, handle,
-                                                   target_client_add_cb,
-                                                   class_export_cb_get(exp));
-                        if (rc == 0) {
-                                cfs_spin_lock(&exp->exp_lock);
-                                exp->exp_need_sync = 1;
-                                cfs_spin_unlock(&exp->exp_lock);
-                        }
-                        rc = fsfilt_write_record(obd, obt->obt_rcvd_filp,
-                                                 ted->ted_lcd,
-                                                 sizeof(*ted->ted_lcd),
-                                                 &off, rc /* sync if no cb */);
-                        fsfilt_commit(obd,
-                                      obt->obt_rcvd_filp->f_dentry->d_inode,
-                                      handle, 0);
-                }
-                pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-
-                if (rc) {
-                        CERROR("error writing %s client idx %u: rc %d\n",
-                               LAST_RCVD, ted->ted_lr_idx, rc);
-                        RETURN(rc);
-                }
-        }
-        RETURN(0);
-}
-
-static int filter_client_del(struct obd_export *exp)
-{
-        struct tg_export_data *ted = &exp->exp_target_data;
-        struct obd_device_target *obt = &exp->exp_obd->u.obt;
-        struct lvfs_run_ctxt saved;
-        int rc;
-        loff_t off;
-        ENTRY;
-
-        if (ted->ted_lcd == NULL)
-                RETURN(0);
-
-        /* XXX if lcd_uuid were a real obd_uuid, I could use obd_uuid_equals */
-        if (strcmp(ted->ted_lcd->lcd_uuid, exp->exp_obd->obd_uuid.uuid ) == 0)
-                GOTO(free, 0);
-
-        LASSERT(obt->obt_lut->lut_client_bitmap != NULL);
-
-        off = ted->ted_lr_off;
-
-        CDEBUG(D_INFO, "freeing client at idx %u, offset %lld with UUID '%s'\n",
-               ted->ted_lr_idx, ted->ted_lr_off, ted->ted_lcd->lcd_uuid);
-
-        /* Don't clear ted_lr_idx here as it is likely also unset.  At worst
-         * we leak a client slot that will be cleaned on the next recovery. */
-        if (off <= 0) {
-                CERROR("%s: client idx %d has med_off %lld\n",
-                       exp->exp_obd->obd_name, ted->ted_lr_idx, off);
-                GOTO(free, rc = -EINVAL);
-        }
-
-        /* Clear the bit _after_ zeroing out the client so we don't
-           race with filter_client_add and zero out new clients.*/
-        if (!cfs_test_bit(ted->ted_lr_idx, obt->obt_lut->lut_client_bitmap)) {
-                CERROR("FILTER client %u: bit already clear in bitmap!!\n",
-                       ted->ted_lr_idx);
-                LBUG();
-        }
-
-        push_ctxt(&saved, &exp->exp_obd->obd_lvfs_ctxt, NULL);
-        /* Make sure the server's last_transno is up to date.
-         * This should be done before zeroing client slot so last_transno will
-         * be in server data or in client data in case of failure */
-        filter_update_server_data(exp->exp_obd);
-
-        cfs_mutex_lock(&ted->ted_lcd_lock);
-        memset(ted->ted_lcd->lcd_uuid, 0, sizeof ted->ted_lcd->lcd_uuid);
-        rc = fsfilt_write_record(exp->exp_obd, obt->obt_rcvd_filp,
-                                 ted->ted_lcd,
-                                 sizeof(*ted->ted_lcd), &off, 0);
-        cfs_mutex_unlock(&ted->ted_lcd_lock);
-        pop_ctxt(&saved, &exp->exp_obd->obd_lvfs_ctxt, NULL);
-
-        CDEBUG(rc == 0 ? D_INFO : D_ERROR,
-               "zero out client %s at idx %u/%llu in %s, rc %d\n",
-               ted->ted_lcd->lcd_uuid, ted->ted_lr_idx, ted->ted_lr_off,
-               LAST_RCVD, rc);
-        EXIT;
-free:
-        return 0;
-}
-
-/* drop fmd reference, free it if last ref. must be called with fed_lock held.*/
-static inline void filter_fmd_put_nolock(struct filter_export_data *fed,
-                                         struct filter_mod_data *fmd)
-{
-        LASSERT_SPIN_LOCKED(&fed->fed_lock);
-        if (--fmd->fmd_refcount == 0) {
-                /* XXX when we have persistent reservations and the handle
-                 * is stored herein we need to drop it here. */
-                fed->fed_mod_count--;
-                cfs_list_del(&fmd->fmd_list);
-                OBD_SLAB_FREE(fmd, ll_fmd_cachep, sizeof(*fmd));
-        }
-}
-
-/* drop fmd reference, free it if last ref */
-void filter_fmd_put(struct obd_export *exp, struct filter_mod_data *fmd)
-{
-        struct filter_export_data *fed;
-
-        if (fmd == NULL)
-                return;
-
-        fed = &exp->exp_filter_data;
-        cfs_spin_lock(&fed->fed_lock);
-        filter_fmd_put_nolock(fed, fmd); /* caller reference */
-        cfs_spin_unlock(&fed->fed_lock);
-}
-
-/* expire entries from the end of the list if there are too many
- * or they are too old */
-static void filter_fmd_expire_nolock(struct filter_obd *filter,
-                                     struct filter_export_data *fed,
-                                     struct filter_mod_data *keep)
-{
-        struct filter_mod_data *fmd, *tmp;
-
-        cfs_list_for_each_entry_safe(fmd, tmp, &fed->fed_mod_list, fmd_list) {
-                if (fmd == keep)
-                        break;
-
-                if (cfs_time_before(jiffies, fmd->fmd_expire) &&
-                    fed->fed_mod_count < filter->fo_fmd_max_num)
-                        break;
-
-                cfs_list_del_init(&fmd->fmd_list);
-                filter_fmd_put_nolock(fed, fmd); /* list reference */
-        }
-}
-
-void filter_fmd_expire(struct obd_export *exp)
-{
-        cfs_spin_lock(&exp->exp_filter_data.fed_lock);
-        filter_fmd_expire_nolock(&exp->exp_obd->u.filter,
-                                 &exp->exp_filter_data, NULL);
-        cfs_spin_unlock(&exp->exp_filter_data.fed_lock);
-}
-
-/* find specified objid, group in export fmd list.
- * caller must hold fed_lock and take fmd reference itself */
-static struct filter_mod_data *filter_fmd_find_nolock(struct filter_obd *filter,
-                                                struct filter_export_data *fed,
-                                                obd_id objid, obd_seq group)
-{
-        struct filter_mod_data *found = NULL, *fmd;
-
-        LASSERT_SPIN_LOCKED(&fed->fed_lock);
-
-        cfs_list_for_each_entry_reverse(fmd, &fed->fed_mod_list, fmd_list) {
-                if (fmd->fmd_id == objid && fmd->fmd_gr == group) {
-                        found = fmd;
-                        cfs_list_del(&fmd->fmd_list);
-                        cfs_list_add_tail(&fmd->fmd_list, &fed->fed_mod_list);
-                        fmd->fmd_expire = jiffies + filter->fo_fmd_max_age;
-                        break;
-                }
-        }
-
-        filter_fmd_expire_nolock(filter, fed, found);
-
-        return found;
-}
-
-/* Find fmd based on objid and group, or return NULL if not found. */
-struct filter_mod_data *filter_fmd_find(struct obd_export *exp,
-                                        obd_id objid, obd_seq group)
-{
-        struct filter_mod_data *fmd;
-
-        cfs_spin_lock(&exp->exp_filter_data.fed_lock);
-        fmd = filter_fmd_find_nolock(&exp->exp_obd->u.filter,
-                                     &exp->exp_filter_data, objid, group);
-        if (fmd)
-                fmd->fmd_refcount++;    /* caller reference */
-        cfs_spin_unlock(&exp->exp_filter_data.fed_lock);
-
-        return fmd;
-}
-
-/* Find fmd based on objid and group, or create a new one if none is found.
- * It is possible for this function to return NULL under memory pressure,
- * or if objid = 0 is passed (which will only cause old entries to expire).
- * Currently this is not fatal because any fmd state is transient and
- * may also be freed when it gets sufficiently old. */
-struct filter_mod_data *filter_fmd_get(struct obd_export *exp,
-                                       obd_id objid, obd_seq group)
-{
-        struct filter_export_data *fed = &exp->exp_filter_data;
-        struct filter_mod_data *found = NULL, *fmd_new = NULL;
-
-        OBD_SLAB_ALLOC_PTR_GFP(fmd_new, ll_fmd_cachep, CFS_ALLOC_IO);
-
-        cfs_spin_lock(&fed->fed_lock);
-        found = filter_fmd_find_nolock(&exp->exp_obd->u.filter,fed,objid,group);
-        if (fmd_new) {
-                if (found == NULL) {
-                        cfs_list_add_tail(&fmd_new->fmd_list,
-                                          &fed->fed_mod_list);
-                        fmd_new->fmd_id = objid;
-                        fmd_new->fmd_gr = group;
-                        fmd_new->fmd_refcount++;   /* list reference */
-                        found = fmd_new;
-                        fed->fed_mod_count++;
-                } else {
-                        OBD_SLAB_FREE(fmd_new, ll_fmd_cachep, sizeof(*fmd_new));
-                }
-        }
-        if (found) {
-                found->fmd_refcount++;          /* caller reference */
-                found->fmd_expire = jiffies +
-                        exp->exp_obd->u.filter.fo_fmd_max_age;
-        }
-
-        cfs_spin_unlock(&fed->fed_lock);
-
-        return found;
-}
-
-#ifdef DO_FMD_DROP
-/* drop fmd list reference so it will disappear when last reference is put.
- * This isn't so critical because it would in fact only affect the one client
- * that is doing the unlink and at worst we have an stale entry referencing
- * an object that should never be used again. */
-static void filter_fmd_drop(struct obd_export *exp, obd_id objid, obd_seq group)
-{
-        struct filter_mod_data *found = NULL;
-
-        cfs_spin_lock(&exp->exp_filter_data.fed_lock);
-        found = filter_fmd_find_nolock(&exp->exp_filter_data, objid, group);
-        if (found) {
-                cfs_list_del_init(&found->fmd_list);
-                filter_fmd_put_nolock(&exp->exp_filter_data, found);
-        }
-        cfs_spin_unlock(&exp->exp_filter_data.fed_lock);
-}
-#else
-#define filter_fmd_drop(exp, objid, group)
-#endif
-
-/* remove all entries from fmd list */
-static void filter_fmd_cleanup(struct obd_export *exp)
-{
-        struct filter_export_data *fed = &exp->exp_filter_data;
-        struct filter_mod_data *fmd = NULL, *tmp;
-
-        cfs_spin_lock(&fed->fed_lock);
-        cfs_list_for_each_entry_safe(fmd, tmp, &fed->fed_mod_list, fmd_list) {
-                cfs_list_del_init(&fmd->fmd_list);
-                filter_fmd_put_nolock(fed, fmd);
-        }
-        cfs_spin_unlock(&fed->fed_lock);
-}
-
-static int filter_init_export(struct obd_export *exp)
-{
-        int rc;
-        ENTRY;
-
-        cfs_spin_lock_init(&exp->exp_filter_data.fed_lock);
-        CFS_INIT_LIST_HEAD(&exp->exp_filter_data.fed_mod_list);
-
-        cfs_spin_lock(&exp->exp_lock);
-        exp->exp_connecting = 1;
-        cfs_spin_unlock(&exp->exp_lock);
-
-        /* self-export doesn't need client data and ldlm initialization */
-        if (unlikely(obd_uuid_equals(&exp->exp_obd->obd_uuid,
-                                     &exp->exp_client_uuid)))
-                RETURN(0);
-
-        rc = lut_client_alloc(exp);
-        if (rc == 0)
-                rc = ldlm_init_export(exp);
-        if (rc)
-                CERROR("%s: Can't initialize export: rc %d\n",
-                       exp->exp_obd->obd_name, rc);
-
-        RETURN(rc);
-}
-
-static int filter_free_server_data(struct obd_device_target *obt)
-{
-        lut_fini(NULL, obt->obt_lut);
-        OBD_FREE_PTR(obt->obt_lut);
-        return 0;
-}
-
-/* assumes caller is already in kernel ctxt */
-int filter_update_server_data(struct obd_device *obd)
-{
-        struct file *filp = obd->u.obt.obt_rcvd_filp;
-        struct lr_server_data *lsd = class_server_data(obd);
-        loff_t off = 0;
-        int rc;
-        ENTRY;
-
-        CDEBUG(D_INODE, "server uuid      : %s\n", lsd->lsd_uuid);
-        CDEBUG(D_INODE, "server last_rcvd : "LPU64"\n",
-               le64_to_cpu(lsd->lsd_last_transno));
-        CDEBUG(D_INODE, "server last_mount: "LPU64"\n",
-               le64_to_cpu(lsd->lsd_mount_count));
-
-        rc = fsfilt_write_record(obd, filp, lsd, sizeof(*lsd), &off, 0);
-        if (rc)
-                CERROR("error writing lr_server_data: rc = %d\n", rc);
-
-        RETURN(rc);
-}
-
-int filter_update_last_objid(struct obd_device *obd, obd_seq group,
-                             int force_sync)
-{
-        struct filter_obd *filter = &obd->u.filter;
-        __u64 tmp;
-        loff_t off = 0;
-        int rc;
-        ENTRY;
-
-        if (filter->fo_last_objid_files[group] == NULL) {
-                CERROR("Object seq "LPU64" not fully setup; not updating "
-                       "last_objid\n", group);
-                RETURN(-EINVAL);
-        }
-
-        CDEBUG(D_INODE, "%s: server last_objid for "POSTID"\n",
-               obd->obd_name, filter->fo_last_objids[group], group);
-
-        tmp = cpu_to_le64(filter->fo_last_objids[group]);
-        rc = fsfilt_write_record(obd, filter->fo_last_objid_files[group],
-                                 &tmp, sizeof(tmp), &off, force_sync);
-        if (rc)
-                CERROR("error writing seq "LPU64" last objid: rc = %d\n",
-                       group, rc);
-        RETURN(rc);
-}
-extern int ost_handle(struct ptlrpc_request *req);
-/* assumes caller has already in kernel ctxt */
-static int filter_init_server_data(struct obd_device *obd, struct file * filp)
-{
-        struct filter_obd *filter = &obd->u.filter;
-        struct lr_server_data *lsd;
-        struct lsd_client_data *lcd = NULL;
-        struct inode *inode = filp->f_dentry->d_inode;
-        unsigned long last_rcvd_size = i_size_read(inode);
-        struct lu_target *lut;
-        __u64 mount_count;
-        __u32 start_epoch;
-        int cl_idx;
-        loff_t off = 0;
-        int rc;
-
-        /* ensure padding in the struct is the correct size */
-        CLASSERT (offsetof(struct lr_server_data, lsd_padding) +
-                 sizeof(lsd->lsd_padding) == LR_SERVER_SIZE);
-        CLASSERT (offsetof(struct lsd_client_data, lcd_padding) +
-                 sizeof(lcd->lcd_padding) == LR_CLIENT_SIZE);
-
-        /* allocate and initialize lu_target */
-        OBD_ALLOC_PTR(lut);
-        if (lut == NULL)
-                RETURN(-ENOMEM);
-        rc = lut_init(NULL, lut, obd, NULL);
-        if (rc)
-                GOTO(err_lut, rc);
-        lsd = class_server_data(obd);
-        if (last_rcvd_size == 0) {
-                LCONSOLE_WARN("%s: new disk, initializing\n", obd->obd_name);
-
-                memcpy(lsd->lsd_uuid, obd->obd_uuid.uuid,sizeof(lsd->lsd_uuid));
-                lsd->lsd_last_transno = 0;
-                mount_count = lsd->lsd_mount_count = 0;
-                lsd->lsd_server_size = cpu_to_le32(LR_SERVER_SIZE);
-                lsd->lsd_client_start = cpu_to_le32(LR_CLIENT_START);
-                lsd->lsd_client_size = cpu_to_le16(LR_CLIENT_SIZE);
-                lsd->lsd_subdir_count = cpu_to_le16(FILTER_SUBDIR_COUNT);
-                filter->fo_subdir_count = FILTER_SUBDIR_COUNT;
-                /* OBD_COMPAT_OST is set in filter_connect_internal when the
-                 * MDS first connects and assigns the OST index number. */
-                lsd->lsd_feature_incompat = cpu_to_le32(OBD_INCOMPAT_COMMON_LR|
-                                                        OBD_INCOMPAT_OST);
-        } else {
-                rc = fsfilt_read_record(obd, filp, lsd, sizeof(*lsd), &off);
-                if (rc) {
-                        CDEBUG(D_INODE,"OBD filter: error reading %s: rc %d\n",
-                               LAST_RCVD, rc);
-                        GOTO(err_lut, rc);
-                }
-                if (strcmp(lsd->lsd_uuid, obd->obd_uuid.uuid) != 0) {
-                        LCONSOLE_ERROR_MSG(0x134, "Trying to start OBD %s "
-                                           "using the wrong disk %s. Were the "
-                                           "/dev/ assignments rearranged?\n",
-                                           obd->obd_uuid.uuid, lsd->lsd_uuid);
-                        GOTO(err_lut, rc = -EINVAL);
-                }
-                mount_count = le64_to_cpu(lsd->lsd_mount_count);
-                filter->fo_subdir_count = le16_to_cpu(lsd->lsd_subdir_count);
-                /* COMPAT_146 */
-                /* Assume old last_rcvd format unless I_C_LR is set */
-                if (!(lsd->lsd_feature_incompat &
-                      cpu_to_le32(OBD_INCOMPAT_COMMON_LR)))
-                        lsd->lsd_last_transno = lsd->lsd_compat14;
-                /* end COMPAT_146 */
-                /* OBD_COMPAT_OST is set in filter_connect_internal when the
-                 * MDS first connects and assigns the OST index number. */
-                lsd->lsd_feature_incompat |= cpu_to_le32(OBD_INCOMPAT_COMMON_LR|
-                                                         OBD_INCOMPAT_OST);
-        }
-
-        if (lsd->lsd_feature_incompat & ~cpu_to_le32(FILTER_INCOMPAT_SUPP)) {
-                CERROR("%s: unsupported incompat filesystem feature(s) %x\n",
-                       obd->obd_name, le32_to_cpu(lsd->lsd_feature_incompat) &
-                       ~FILTER_INCOMPAT_SUPP);
-                GOTO(err_lut, rc = -EINVAL);
-        }
-        if (lsd->lsd_feature_rocompat & ~cpu_to_le32(FILTER_ROCOMPAT_SUPP)) {
-                CERROR("%s: unsupported read-only filesystem feature(s) %x\n",
-                       obd->obd_name, le32_to_cpu(lsd->lsd_feature_rocompat) &
-                       ~FILTER_ROCOMPAT_SUPP);
-                /* Do something like remount filesystem read-only */
-                GOTO(err_lut, rc = -EINVAL);
-        }
-
-        start_epoch = le32_to_cpu(lsd->lsd_start_epoch);
-
-        CDEBUG(D_INODE, "%s: server start_epoch : %#x\n",
-               obd->obd_name, start_epoch);
-        CDEBUG(D_INODE, "%s: server last_transno : "LPX64"\n",
-               obd->obd_name, le64_to_cpu(lsd->lsd_last_transno));
-        CDEBUG(D_INODE, "%s: server mount_count: "LPU64"\n",
-               obd->obd_name, mount_count + 1);
-        CDEBUG(D_INODE, "%s: server data size: %u\n",
-               obd->obd_name, le32_to_cpu(lsd->lsd_server_size));
-        CDEBUG(D_INODE, "%s: per-client data start: %u\n",
-               obd->obd_name, le32_to_cpu(lsd->lsd_client_start));
-        CDEBUG(D_INODE, "%s: per-client data size: %u\n",
-               obd->obd_name, le32_to_cpu(lsd->lsd_client_size));
-        CDEBUG(D_INODE, "%s: server subdir_count: %u\n",
-               obd->obd_name, le16_to_cpu(lsd->lsd_subdir_count));
-        CDEBUG(D_INODE, "%s: last_rcvd clients: %lu\n", obd->obd_name,
-               last_rcvd_size <= le32_to_cpu(lsd->lsd_client_start) ? 0 :
-               (last_rcvd_size - le32_to_cpu(lsd->lsd_client_start)) /
-                le16_to_cpu(lsd->lsd_client_size));
-
-        if (!obd->obd_replayable) {
-                CWARN("%s: recovery support OFF\n", obd->obd_name);
-                GOTO(out, rc = 0);
-        }
-
-        OBD_ALLOC_PTR(lcd);
-        if (!lcd)
-                GOTO(err_client, rc = -ENOMEM);
-
-        for (cl_idx = 0, off = le32_to_cpu(lsd->lsd_client_start);
-             off < last_rcvd_size; cl_idx++) {
-                __u64 last_rcvd;
-                struct obd_export *exp;
-                struct filter_export_data *fed;
-
-                /* Don't assume off is incremented properly by
-                 * fsfilt_read_record(), in case sizeof(*lcd)
-                 * isn't the same as lsd->lsd_client_size.  */
-                off = le32_to_cpu(lsd->lsd_client_start) +
-                        cl_idx * le16_to_cpu(lsd->lsd_client_size);
-                rc = fsfilt_read_record(obd, filp, lcd, sizeof(*lcd), &off);
-                if (rc) {
-                        CERROR("error reading FILT %s idx %d off %llu: rc %d\n",
-                               LAST_RCVD, cl_idx, off, rc);
-                        break; /* read error shouldn't cause startup to fail */
-                }
-
-                if (lcd->lcd_uuid[0] == '\0') {
-                        CDEBUG(D_INFO, "skipping zeroed client at offset %d\n",
-                               cl_idx);
-                        continue;
-                }
-
-                check_lcd(obd->obd_name, cl_idx, lcd);
-
-                last_rcvd = le64_to_cpu(lcd->lcd_last_transno);
-
-                CDEBUG(D_HA, "RCVRNG CLIENT uuid: %s idx: %d lr: "LPU64
-                       " srv lr: "LPU64"\n", lcd->lcd_uuid, cl_idx,
-                       last_rcvd, le64_to_cpu(lsd->lsd_last_transno));
-
-                /* These exports are cleaned up by filter_disconnect(), so they
-                 * need to be set up like real exports as filter_connect() does.
-                 */
-                exp = class_new_export(obd, (struct obd_uuid *)lcd->lcd_uuid);
-                if (IS_ERR(exp)) {
-                        if (PTR_ERR(exp) == -EALREADY) {
-                                /* export already exists, zero out this one */
-                                CERROR("Duplicate export %s!\n", lcd->lcd_uuid);
-                                continue;
-                        }
-                        OBD_FREE_PTR(lcd);
-                        GOTO(err_client, rc = PTR_ERR(exp));
-                }
-
-                fed = &exp->exp_filter_data;
-                *fed->fed_ted.ted_lcd = *lcd;
-                fed->fed_group = 0; /* will be assigned at connect */
-                filter_export_stats_init(obd, exp, NULL);
-                rc = filter_client_add(obd, exp, cl_idx);
-                /* can't fail for existing client */
-                LASSERTF(rc == 0, "rc = %d\n", rc);
-
-                /* VBR: set export last committed */
-                exp->exp_last_committed = last_rcvd;
-                cfs_spin_lock(&exp->exp_lock);
-                exp->exp_connecting = 0;
-                exp->exp_in_recovery = 0;
-                cfs_spin_unlock(&exp->exp_lock);
-                obd->obd_max_recoverable_clients++;
-                class_export_put(exp);
-
-                if (last_rcvd > le64_to_cpu(lsd->lsd_last_transno))
-                        lsd->lsd_last_transno = cpu_to_le64(last_rcvd);
-        }
-        OBD_FREE_PTR(lcd);
-
-        obd->obd_last_committed = le64_to_cpu(lsd->lsd_last_transno);
-out:
-        obd->u.obt.obt_mount_count = mount_count + 1;
-        obd->u.obt.obt_instance = (__u32)obd->u.obt.obt_mount_count;
-        lsd->lsd_mount_count = cpu_to_le64(obd->u.obt.obt_mount_count);
-
-        /* save it, so mount count and last_transno is current */
-        rc = filter_update_server_data(obd);
-        if (rc)
-                GOTO(err_client, rc);
-
-        RETURN(0);
-
-err_client:
-        class_disconnect_exports(obd);
-err_lut:
-        filter_free_server_data(&obd->u.obt);
-        RETURN(rc);
-}
-
-static int filter_cleanup_groups(struct obd_device *obd)
-{
-        struct filter_obd *filter = &obd->u.filter;
-        struct file *filp;
-        struct dentry *dentry;
-        int i, j;
-        ENTRY;
-
-        if (filter->fo_dentry_O_groups != NULL) {
-                for (i = 0; i < filter->fo_group_count; i++) {
-                        dentry = filter->fo_dentry_O_groups[i];
-                        if (dentry != NULL)
-                                f_dput(dentry);
-                }
-                OBD_FREE(filter->fo_dentry_O_groups,
-                         filter->fo_group_count *
-                         sizeof(*filter->fo_dentry_O_groups));
-                filter->fo_dentry_O_groups = NULL;
-        }
-        if (filter->fo_last_objid_files != NULL) {
-                for (i = 0; i < filter->fo_group_count; i++) {
-                        filp = filter->fo_last_objid_files[i];
-                        if (filp != NULL)
-                                filp_close(filp, 0);
-                }
-                OBD_FREE(filter->fo_last_objid_files,
-                         filter->fo_group_count *
-                         sizeof(*filter->fo_last_objid_files));
-                filter->fo_last_objid_files = NULL;
-        }
-        if (filter->fo_dentry_O_sub != NULL) {
-                for (i = 0; i < filter->fo_group_count; i++) {
-                        for (j = 0; j < filter->fo_subdir_count; j++) {
-                                dentry = filter->fo_dentry_O_sub[i].dentry[j];
-                                if (dentry != NULL)
-                                        f_dput(dentry);
-                        }
-                }
-                OBD_FREE(filter->fo_dentry_O_sub,
-                         filter->fo_group_count *
-                         sizeof(*filter->fo_dentry_O_sub));
-                filter->fo_dentry_O_sub = NULL;
-        }
-        if (filter->fo_last_objids != NULL) {
-                OBD_FREE(filter->fo_last_objids,
-                         filter->fo_group_count *
-                         sizeof(*filter->fo_last_objids));
-                filter->fo_last_objids = NULL;
-        }
-        if (filter->fo_dentry_O != NULL) {
-                f_dput(filter->fo_dentry_O);
-                filter->fo_dentry_O = NULL;
-        }
-        RETURN(0);
-}
-
-static int filter_update_last_group(struct obd_device *obd, int group)
-{
-        struct filter_obd *filter = &obd->u.filter;
-        struct file *filp = NULL;
-        int last_group = 0, rc;
-        loff_t off = 0;
-        ENTRY;
-
-        if (group <= filter->fo_committed_group)
-                RETURN(0);
-
-        filp = filp_open("LAST_GROUP", O_RDWR, 0700);
-        if (IS_ERR(filp)) {
-                rc = PTR_ERR(filp);
-                filp = NULL;
-                CERROR("cannot open LAST_GROUP: rc = %d\n", rc);
-                GOTO(cleanup, rc);
-        }
-
-        rc = fsfilt_read_record(obd, filp, &last_group, sizeof(__u32), &off);
-        if (rc) {
-                CDEBUG(D_INODE, "error reading LAST_GROUP: rc %d\n",rc);
-                GOTO(cleanup, rc);
-        }
-
-        CDEBUG(D_INODE, "%s: previous %d, new %d\n",
-               obd->obd_name, last_group, group);
-
-        off = 0;
-        last_group = group;
-        /* must be sync: bXXXX */
-        rc = fsfilt_write_record(obd, filp, &last_group, sizeof(__u32), &off, 1);
-        if (rc) {
-                CDEBUG(D_INODE, "error updating LAST_GROUP: rc %d\n", rc);
-                GOTO(cleanup, rc);
-        }
-
-        filter->fo_committed_group = group;
-cleanup:
-        if (filp)
-                filp_close(filp, 0);
-        RETURN(rc);
-}
-
-static int filter_read_group_internal(struct obd_device *obd, int group,
-                                      int create)
-{
-        struct filter_obd *filter = &obd->u.filter;
-        __u64 *new_objids = NULL;
-        struct filter_subdirs *new_subdirs = NULL, *tmp_subdirs = NULL;
-        struct dentry **new_groups = NULL;
-        struct file **new_files = NULL;
-        struct dentry *dentry;
-        struct file *filp;
-        int old_count = filter->fo_group_count, rc, stage = 0, i;
-        char name[25];
-        __u64 last_objid;
-        loff_t off = 0;
-        int len = group + 1;
-
-        snprintf(name, 24, "%d", group);
-        name[24] = '\0';
-
-        if (!create) {
-                dentry = ll_lookup_one_len(name, filter->fo_dentry_O,
-                                           strlen(name));
-                if (IS_ERR(dentry)) {
-                        CERROR("Cannot lookup expected object group %d: %ld\n",
-                               group, PTR_ERR(dentry));
-                        RETURN(PTR_ERR(dentry));
-                }
-        } else {
-                dentry = simple_mkdir(filter->fo_dentry_O,
-                                      obd->u.obt.obt_vfsmnt, name, 0700, 1);
-                if (IS_ERR(dentry)) {
-                        CERROR("cannot lookup/create O/%s: rc = %ld\n", name,
-                               PTR_ERR(dentry));
-                        RETURN(PTR_ERR(dentry));
-                }
-        }
-        stage = 1;
-
-        snprintf(name, 24, "O/%d/LAST_ID", group);
-        name[24] = '\0';
-        filp = filp_open(name, O_CREAT | O_RDWR, 0700);
-        if (IS_ERR(filp)) {
-                CERROR("cannot create %s: rc = %ld\n", name, PTR_ERR(filp));
-                GOTO(cleanup, rc = PTR_ERR(filp));
-        }
-        stage = 2;
-
-        rc = fsfilt_read_record(obd, filp, &last_objid, sizeof(__u64), &off);
-        if (rc) {
-                CDEBUG(D_INODE, "error reading %s: rc %d\n", name, rc);
-                GOTO(cleanup, rc);
-        }
-
-        if (filter->fo_subdir_count && fid_seq_is_mdt(group)) {
-                OBD_ALLOC(tmp_subdirs, sizeof(*tmp_subdirs));
-                if (tmp_subdirs == NULL)
-                        GOTO(cleanup, rc = -ENOMEM);
-                stage = 3;
-
-                for (i = 0; i < filter->fo_subdir_count; i++) {
-                        char dir[20];
-                        snprintf(dir, sizeof(dir), "d%u", i);
-
-                        tmp_subdirs->dentry[i] = simple_mkdir(dentry,
-                                                              obd->u.obt.obt_vfsmnt,
-                                                              dir, 0700, 1);
-                        if (IS_ERR(tmp_subdirs->dentry[i])) {
-                                rc = PTR_ERR(tmp_subdirs->dentry[i]);
-                                CERROR("can't lookup/create O/%d/%s: rc = %d\n",
-                                       group, dir, rc);
-                                GOTO(cleanup, rc);
-                        }
-
-                        CDEBUG(D_INODE, "got/created O/%d/%s: %p\n", group, dir,
-                               tmp_subdirs->dentry[i]);
-                }
-        }
-
-        /* 'group' is an index; we need an array of length 'group + 1' */
-        if (group + 1 > old_count) {
-                OBD_ALLOC(new_objids, len * sizeof(*new_objids));
-                OBD_ALLOC(new_subdirs, len * sizeof(*new_subdirs));
-                OBD_ALLOC(new_groups, len * sizeof(*new_groups));
-                OBD_ALLOC(new_files, len * sizeof(*new_files));
-                stage = 4;
-                if (new_objids == NULL || new_subdirs == NULL ||
-                    new_groups == NULL || new_files == NULL)
-                        GOTO(cleanup, rc = -ENOMEM);
-
-                if (old_count) {
-                        memcpy(new_objids, filter->fo_last_objids,
-                               old_count * sizeof(*new_objids));
-                        memcpy(new_subdirs, filter->fo_dentry_O_sub,
-                               old_count * sizeof(*new_subdirs));
-                        memcpy(new_groups, filter->fo_dentry_O_groups,
-                               old_count * sizeof(*new_groups));
-                        memcpy(new_files, filter->fo_last_objid_files,
-                               old_count * sizeof(*new_files));
-
-                        OBD_FREE(filter->fo_last_objids,
-                                 old_count * sizeof(*new_objids));
-                        OBD_FREE(filter->fo_dentry_O_sub,
-                                 old_count * sizeof(*new_subdirs));
-                        OBD_FREE(filter->fo_dentry_O_groups,
-                                 old_count * sizeof(*new_groups));
-                        OBD_FREE(filter->fo_last_objid_files,
-                                 old_count * sizeof(*new_files));
-                }
-                filter->fo_last_objids = new_objids;
-                filter->fo_dentry_O_sub = new_subdirs;
-                filter->fo_dentry_O_groups = new_groups;
-                filter->fo_last_objid_files = new_files;
-                filter->fo_group_count = len;
-        }
-
-        filter->fo_dentry_O_groups[group] = dentry;
-        filter->fo_last_objid_files[group] = filp;
-        if (filter->fo_subdir_count && fid_seq_is_mdt(group)) {
-                filter->fo_dentry_O_sub[group] = *tmp_subdirs;
-                OBD_FREE(tmp_subdirs, sizeof(*tmp_subdirs));
-        }
-
-        filter_update_last_group(obd, group);
-
-        if (i_size_read(filp->f_dentry->d_inode) == 0) {
-                filter->fo_last_objids[group] = FILTER_INIT_OBJID;
-                rc = filter_update_last_objid(obd, group, 1);
-                RETURN(rc);
-        }
-
-        filter->fo_last_objids[group] = le64_to_cpu(last_objid);
-        CDEBUG(D_INODE, "%s: server last_objid group %d: "LPU64"\n",
-               obd->obd_name, group, last_objid);
-        RETURN(0);
- cleanup:
-        switch (stage) {
-        case 4:
-                if (new_objids != NULL)
-                        OBD_FREE(new_objids, len * sizeof(*new_objids));
-                if (new_subdirs != NULL)
-                        OBD_FREE(new_subdirs, len * sizeof(*new_subdirs));
-                if (new_groups != NULL)
-                        OBD_FREE(new_groups, len * sizeof(*new_groups));
-                if (new_files != NULL)
-                        OBD_FREE(new_files, len * sizeof(*new_files));
-        case 3:
-                if (filter->fo_subdir_count && fid_seq_is_mdt(group)) {
-                        for (i = 0; i < filter->fo_subdir_count; i++) {
-                                if (tmp_subdirs->dentry[i] != NULL)
-                                        dput(tmp_subdirs->dentry[i]);
-                        }
-                        OBD_FREE(tmp_subdirs, sizeof(*tmp_subdirs));
-                }
-        case 2:
-                filp_close(filp, 0);
-        case 1:
-                dput(dentry);
-        }
-        RETURN(rc);
-}
-
-static int filter_read_groups(struct obd_device *obd, int last_group,
-                              int create)
-{
-        struct filter_obd *filter = &obd->u.filter;
-        int old_count, group, rc = 0;
-
-        cfs_mutex_lock(&filter->fo_init_lock);
-        old_count = filter->fo_group_count;
-        for (group = old_count; group <= last_group; group++) {
-                rc = filter_read_group_internal(obd, group, create);
-                if (rc != 0)
-                        break;
-        }
-        cfs_mutex_unlock(&filter->fo_init_lock);
-        return rc;
-}
-
-/* FIXME: object groups */
-static int filter_prep_groups(struct obd_device *obd)
-{
-        struct filter_obd *filter = &obd->u.filter;
-        struct dentry *O_dentry;
-        struct file *filp;
-        int    last_group, rc = 0, cleanup_phase = 0;
-        loff_t off = 0;
-        ENTRY;
-
-        O_dentry = simple_mkdir(cfs_fs_pwd(current->fs), obd->u.obt.obt_vfsmnt,
-                                "O", 0700, 1);
-        CDEBUG(D_INODE, "%s: got/created O: %p\n", obd->obd_name, O_dentry);
-        if (IS_ERR(O_dentry)) {
-                rc = PTR_ERR(O_dentry);
-                CERROR("%s: cannot open/create O: rc = %d\n", obd->obd_name,rc);
-                GOTO(cleanup, rc);
-        }
-        filter->fo_dentry_O = O_dentry;
-        cleanup_phase = 1; /* O_dentry */
-
-        /* we have to initialize all groups before first connections from
-         * clients because they may send create/destroy for any group -bzzz */
-        filp = filp_open("LAST_GROUP", O_CREAT | O_RDWR, 0700);
-        if (IS_ERR(filp)) {
-                CERROR("%s: cannot create LAST_GROUP: rc = %ld\n",
-                       obd->obd_name, PTR_ERR(filp));
-                GOTO(cleanup, rc = PTR_ERR(filp));
-        }
-        cleanup_phase = 2; /* filp */
-
-        rc = fsfilt_read_record(obd, filp, &last_group, sizeof(__u32), &off);
-        if (rc) {
-                CERROR("%s: error reading LAST_GROUP: rc %d\n",
-                       obd->obd_name, rc);
-                GOTO(cleanup, rc);
-        }
-
-        if (off == 0)
-                last_group = FID_SEQ_OST_MDT0;
-
-        CDEBUG(D_INODE, "%s: initialize group %u (max %u)\n", obd->obd_name,
-               FID_SEQ_OST_MDT0, last_group);
-        filter->fo_committed_group = last_group;
-        rc = filter_read_groups(obd, last_group, 1);
-        if (rc)
-                GOTO(cleanup, rc);
-
-        filp_close(filp, 0);
-        RETURN(0);
-
- cleanup:
-        switch (cleanup_phase) {
-        case 2:
-                filp_close(filp, 0);
-        case 1:
-                filter_cleanup_groups(obd);
-                f_dput(filter->fo_dentry_O);
-                filter->fo_dentry_O = NULL;
-        default:
-                break;
-        }
-        return rc;
-
-}
-
-/* setup the object store with correct subdirectories */
-static int filter_prep(struct obd_device *obd)
-{
-        struct lvfs_run_ctxt saved;
-        struct filter_obd *filter = &obd->u.filter;
-        struct file *file;
-        struct inode *inode;
-        int rc = 0;
-        ENTRY;
-
-        push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-        file = filp_open(LAST_RCVD, O_RDWR | O_CREAT | O_LARGEFILE, 0700);
-        if (!file || IS_ERR(file)) {
-                rc = PTR_ERR(file);
-                CERROR("OBD filter: cannot open/create %s: rc = %d\n",
-                       LAST_RCVD, rc);
-                GOTO(out, rc);
-        }
-        obd->u.obt.obt_rcvd_filp = file;
-        if (!S_ISREG(file->f_dentry->d_inode->i_mode)) {
-                CERROR("%s is not a regular file!: mode = %o\n", LAST_RCVD,
-                       file->f_dentry->d_inode->i_mode);
-                GOTO(err_filp, rc = -ENOENT);
-        }
-
-        inode = file->f_dentry->d_parent->d_inode;
-        /* We use i_op->unlink directly in filter_vfs_unlink() */
-        if (!inode->i_op || !inode->i_op->create || !inode->i_op->unlink) {
-                CERROR("%s: filesystem does not support create/unlink ops\n",
-                       obd->obd_name);
-                GOTO(err_filp, rc = -EOPNOTSUPP);
-        }
-
-        rc = filter_init_server_data(obd, file);
-        if (rc) {
-                CERROR("cannot read %s: rc = %d\n", LAST_RCVD, rc);
-                GOTO(err_filp, rc);
-        }
-        LASSERT(obd->u.obt.obt_lut);
-        target_recovery_init(obd->u.obt.obt_lut, ost_handle);
-
-        /* open and create health check io file*/
-        file = filp_open(HEALTH_CHECK, O_RDWR | O_CREAT, 0644);
-        if (IS_ERR(file)) {
-                rc = PTR_ERR(file);
-                CERROR("OBD filter: cannot open/create %s rc = %d\n",
-                       HEALTH_CHECK, rc);
-                GOTO(err_server_data, rc);
-        }
-        filter->fo_obt.obt_health_check_filp = file;
-        if (!S_ISREG(file->f_dentry->d_inode->i_mode)) {
-                CERROR("%s is not a regular file!: mode = %o\n", HEALTH_CHECK,
-                       file->f_dentry->d_inode->i_mode);
-                GOTO(err_health_check, rc = -ENOENT);
-        }
-        rc = lvfs_check_io_health(obd, file);
-        if (rc)
-                GOTO(err_health_check, rc);
-
-        rc = filter_prep_groups(obd);
-        if (rc)
-                GOTO(err_health_check, rc);
-out:
-        pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-
-        return(rc);
-
-err_health_check:
-        if (filp_close(filter->fo_obt.obt_health_check_filp, 0))
-                CERROR("can't close %s after error\n", HEALTH_CHECK);
-        filter->fo_obt.obt_health_check_filp = NULL;
-err_server_data:
-        target_recovery_fini(obd);
-        filter_free_server_data(&obd->u.obt);
-err_filp:
-        if (filp_close(obd->u.obt.obt_rcvd_filp, 0))
-                CERROR("can't close %s after error\n", LAST_RCVD);
-        obd->u.obt.obt_rcvd_filp = NULL;
-        goto out;
-}
-
-/* cleanup the filter: write last used object id to status file */
-static void filter_post(struct obd_device *obd)
-{
-        struct lvfs_run_ctxt saved;
-        struct filter_obd *filter = &obd->u.filter;
-        int rc, i;
-
-        /* XXX: filter_update_lastobjid used to call fsync_dev.  It might be
-         * best to start a transaction with h_sync, because we removed this
-         * from lastobjid */
-
-        push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-        rc = filter_update_server_data(obd);
-        if (rc)
-                CERROR("error writing server data: rc = %d\n", rc);
-
-        for (i = 0; i < filter->fo_group_count; i++) {
-                rc = filter_update_last_objid(obd, i,
-                                (i == filter->fo_group_count - 1));
-                if (rc)
-                        CERROR("error writing group %d lastobjid: rc = %d\n",
-                               i, rc);
-        }
-
-        rc = filp_close(obd->u.obt.obt_rcvd_filp, 0);
-        obd->u.obt.obt_rcvd_filp = NULL;
-        if (rc)
-                CERROR("error closing %s: rc = %d\n", LAST_RCVD, rc);
-
-        rc = filp_close(filter->fo_obt.obt_health_check_filp, 0);
-        filter->fo_obt.obt_health_check_filp = NULL;
-        if (rc)
-                CERROR("error closing %s: rc = %d\n", HEALTH_CHECK, rc);
-
-        filter_cleanup_groups(obd);
-        filter_free_server_data(&obd->u.obt);
-        pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-
-        filter_free_capa_keys(filter);
-        cleanup_capa_hash(filter->fo_capa_hash);
-}
-
-static void filter_set_last_id(struct filter_obd *filter,
-                               obd_id id, obd_seq group)
-{
-        LASSERT(group <= filter->fo_group_count);
-
-        cfs_spin_lock(&filter->fo_objidlock);
-        filter->fo_last_objids[group] = id;
-        cfs_spin_unlock(&filter->fo_objidlock);
-}
-
-obd_id filter_last_id(struct filter_obd *filter, obd_seq group)
-{
-        obd_id id;
-        LASSERT(group <= filter->fo_group_count);
-        LASSERT(filter->fo_last_objids != NULL);
-
-        /* FIXME: object groups */
-        cfs_spin_lock(&filter->fo_objidlock);
-        id = filter->fo_last_objids[group];
-        cfs_spin_unlock(&filter->fo_objidlock);
-        return id;
-}
-
-static int filter_lock_dentry(struct obd_device *obd, struct dentry *dparent)
-{
-       mutex_lock_nested(&dparent->d_inode->i_mutex, I_MUTEX_PARENT);
-       return 0;
-}
-
-/* We never dget the object parent, so DON'T dput it either */
-struct dentry *filter_parent(struct obd_device *obd, obd_seq group, obd_id objid)
-{
-        struct filter_obd *filter = &obd->u.filter;
-        struct filter_subdirs *subdirs;
-
-        if (group >= filter->fo_group_count) /* FIXME: object groups */
-                return ERR_PTR(-EBADF);
-
-        if (!fid_seq_is_mdt(group) || filter->fo_subdir_count == 0)
-                return filter->fo_dentry_O_groups[group];
-
-        subdirs = &filter->fo_dentry_O_sub[group];
-        return subdirs->dentry[objid & (filter->fo_subdir_count - 1)];
-}
-
-/* We never dget the object parent, so DON'T dput it either */
-struct dentry *filter_parent_lock(struct obd_device *obd, obd_seq group,
-                                  obd_id objid)
-{
-        unsigned long now = jiffies;
-        struct dentry *dparent = filter_parent(obd, group, objid);
-        int rc;
-
-        if (IS_ERR(dparent))
-                return dparent;
-        if (dparent == NULL)
-                return ERR_PTR(-ENOENT);
-
-        rc = filter_lock_dentry(obd, dparent);
-        fsfilt_check_slow(obd, now, "parent lock");
-        return rc ? ERR_PTR(rc) : dparent;
-}
-
-/* We never dget the object parent, so DON'T dput it either */
-static void filter_parent_unlock(struct dentry *dparent)
-{
-       mutex_unlock(&dparent->d_inode->i_mutex);
-}
-
-/* How to get files, dentries, inodes from object id's.
- *
- * If dir_dentry is passed, the caller has already locked the parent
- * appropriately for this operation (normally a write lock).  If
- * dir_dentry is NULL, we do a read lock while we do the lookup to
- * avoid races with create/destroy and such changing the directory
- * internal to the filesystem code. */
-struct dentry *filter_fid2dentry(struct obd_device *obd,
-                                 struct dentry *dir_dentry,
-                                 obd_seq group, obd_id id)
-{
-        struct dentry *dparent = dir_dentry;
-        struct dentry *dchild;
-        char name[32];
-        int len;
-        ENTRY;
-
-        if (OBD_FAIL_CHECK(OBD_FAIL_OST_ENOENT) &&
-            obd->u.filter.fo_destroys_in_progress == 0) {
-                /* don't fail lookups for orphan recovery, it causes
-                 * later LBUGs when objects still exist during precreate */
-                CDEBUG(D_INFO, "*** cfs_fail_loc=%x ***\n",OBD_FAIL_OST_ENOENT);
-                RETURN(ERR_PTR(-ENOENT));
-        }
-        if (id == 0) {
-                CERROR("fatal: invalid object id 0\n");
-                RETURN(ERR_PTR(-ESTALE));
-        }
-
-        len = sprintf(name, LPU64, id);
-        if (dir_dentry == NULL) {
-                dparent = filter_parent_lock(obd, group, id);
-                if (IS_ERR(dparent)) {
-                        CERROR("%s: error getting object "POSTID
-                               " parent: rc %ld\n", obd->obd_name,
-                               id, group, PTR_ERR(dparent));
-                        RETURN(dparent);
-                }
-        }
-        CDEBUG(D_INODE, "looking up object O/%.*s/%s\n",
-               dparent->d_name.len, dparent->d_name.name, name);
-        /* dparent is already locked here, so we cannot use ll_lookup_one_len() */
-        dchild = lookup_one_len(name, dparent, len);
-        if (dir_dentry == NULL)
-                filter_parent_unlock(dparent);
-        if (IS_ERR(dchild)) {
-                CERROR("%s: object "LPU64":"LPU64" lookup error: rc %ld\n",
-                       obd->obd_name, id, group, PTR_ERR(dchild));
-                RETURN(dchild);
-        }
-
-        if (dchild->d_inode != NULL && is_bad_inode(dchild->d_inode)) {
-                CERROR("%s: got bad object "LPU64" inode %lu\n",
-                       obd->obd_name, id, dchild->d_inode->i_ino);
-                f_dput(dchild);
-                RETURN(ERR_PTR(-ENOENT));
-        }
-
-        CDEBUG(D_INODE, "got child objid %s: %p, count = %d\n",
-               name, dchild, atomic_read(&dchild->d_count));
-
-        LASSERT(atomic_read(&dchild->d_count) > 0);
-
-        RETURN(dchild);
-}
-
-static int filter_prepare_destroy(struct obd_device *obd, obd_id objid,
-                                  obd_id group, struct lustre_handle *lockh)
-{
-        int flags = LDLM_AST_DISCARD_DATA, rc;
-        struct ldlm_res_id res_id;
-        ldlm_policy_data_t policy = { .l_extent = { 0, OBD_OBJECT_EOF } };
-        ENTRY;
-
-        osc_build_res_name(objid, group, &res_id);
-        /* Tell the clients that the object is gone now and that they should
-         * throw away any cached pages. */
-        rc = ldlm_cli_enqueue_local(obd->obd_namespace, &res_id, LDLM_EXTENT,
-                                    &policy, LCK_PW, &flags, ldlm_blocking_ast,
-                                    ldlm_completion_ast, NULL, NULL, 0, NULL,
-                                    lockh);
-        if (rc != ELDLM_OK)
-                lockh->cookie = 0;
-        RETURN(rc);
-}
-
-static void filter_fini_destroy(struct obd_device *obd,
-                                struct lustre_handle *lockh)
-{
-        if (lustre_handle_is_used(lockh))
-                ldlm_lock_decref(lockh, LCK_PW);
-}
-
-/* This is vfs_unlink() without down(i_sem).  If we call regular vfs_unlink()
- * we have 2.6 lock ordering issues with filter_commitrw_write() as it takes
- * i_sem before starting a handle, while filter_destroy() + vfs_unlink do the
- * reverse.  Caller must take i_sem before starting the transaction and we
- * drop it here before the inode is removed from the dentry.  bug 4180/6984 */
-int filter_vfs_unlink(struct inode *dir, struct dentry *dentry,
-                      struct vfsmount *mnt)
-{
-        int rc;
-        ENTRY;
-
-        /* don't need dir->i_zombie for 2.4, it is for rename/unlink of dir
-         * itself we already hold dir->i_mutex for child create/unlink ops */
-       LASSERT(dentry->d_inode != NULL);
-       LASSERT(mutex_trylock(&dir->i_mutex) == 0);
-       LASSERT(mutex_trylock(&dentry->d_inode->i_mutex) == 0);
-
-
-        /* may_delete() */
-        if (/*!dentry->d_inode ||*/dentry->d_parent->d_inode != dir)
-                GOTO(out, rc = -ENOENT);
-
-        rc = ll_permission(dir, MAY_WRITE | MAY_EXEC, NULL);
-        if (rc)
-                GOTO(out, rc);
-
-        if (IS_APPEND(dir))
-                GOTO(out, rc = -EPERM);
-
-        /* check_sticky() */
-        if ((dentry->d_inode->i_uid != cfs_curproc_fsuid() &&
-             !cfs_capable(CFS_CAP_FOWNER)) || IS_APPEND(dentry->d_inode) ||
-            IS_IMMUTABLE(dentry->d_inode))
-                GOTO(out, rc = -EPERM);
-
-        /* Locking order: i_mutex -> journal_lock -> dqptr_sem. LU-952 */
-        ll_vfs_dq_init(dir);
-
-        rc = ll_security_inode_unlink(dir, dentry, mnt);
-        if (rc)
-                GOTO(out, rc);
-
-        rc = dir->i_op->unlink(dir, dentry);
-out:
-       /* need to drop i_mutex before we lose inode reference */
-       mutex_unlock(&dentry->d_inode->i_mutex);
-       if (rc == 0)
-               d_delete(dentry);
-
-       RETURN(rc);
-}
-
-/* Caller must hold LCK_PW on parent and push us into kernel context.
- * Caller must hold child i_mutex, we drop it always.
- * Caller is also required to ensure that dchild->d_inode exists. */
-static int filter_destroy_internal(struct obd_device *obd, obd_id objid,
-                                   obd_seq group, struct dentry *dparent,
-                                   struct dentry *dchild)
-{
-        struct inode *inode = dchild->d_inode;
-        int rc;
-
-        /* There should be 2 references to the inode:
-         *  1) taken by filter_prepare_destroy
-         *  2) taken by filter_destroy */
-        if (inode->i_nlink != 1 || atomic_read(&inode->i_count) != 2) {
-                CERROR("destroying objid %.*s ino %lu nlink %lu count %d\n",
-                       dchild->d_name.len, dchild->d_name.name, inode->i_ino,
-                       (unsigned long)inode->i_nlink,
-                       atomic_read(&inode->i_count));
-        }
-
-        rc = filter_vfs_unlink(dparent->d_inode, dchild, obd->u.obt.obt_vfsmnt);
-        if (rc)
-                CERROR("error unlinking objid %.*s: rc %d\n",
-                       dchild->d_name.len, dchild->d_name.name, rc);
-        return(rc);
-}
-
-struct filter_intent_args {
-        struct ldlm_lock **victim;
-        __u64 size;
-        int *liblustre;
-};
-
-static enum interval_iter filter_intent_cb(struct interval_node *n,
-                                           void *args)
-{
-        struct ldlm_interval *node = (struct ldlm_interval *)n;
-        struct filter_intent_args *arg = (struct filter_intent_args*)args;
-        __u64 size = arg->size;
-        struct ldlm_lock **v = arg->victim;
-        struct ldlm_lock *lck;
-
-        /* If the interval is lower than the current file size,
-         * just break. */
-        if (interval_high(n) <= size)
-                return INTERVAL_ITER_STOP;
-
-        cfs_list_for_each_entry(lck, &node->li_group, l_sl_policy) {
-                /* Don't send glimpse ASTs to liblustre clients.
-                 * They aren't listening for them, and they do
-                 * entirely synchronous I/O anyways. */
-                if (lck->l_export == NULL ||
-                    lck->l_export->exp_libclient == 1)
-                        continue;
-
-                if (*arg->liblustre)
-                        *arg->liblustre = 0;
-
-                if (*v == NULL) {
-                        *v = LDLM_LOCK_GET(lck);
-                } else if ((*v)->l_policy_data.l_extent.start <
-                           lck->l_policy_data.l_extent.start) {
-                        LDLM_LOCK_RELEASE(*v);
-                        *v = LDLM_LOCK_GET(lck);
-                }
-
-                /* the same policy group - every lock has the
-                 * same extent, so needn't do it any more */
-                break;
-        }
-
-        return INTERVAL_ITER_CONT;
-}
-
-static int filter_intent_policy(struct ldlm_namespace *ns,
-                                struct ldlm_lock **lockp, void *req_cookie,
-                                ldlm_mode_t mode, int flags, void *data)
-{
-        struct ptlrpc_request *req = req_cookie;
-        struct ldlm_lock *lock = *lockp, *l = NULL;
-        struct ldlm_resource *res = lock->l_resource;
-        ldlm_processing_policy policy;
-        struct ost_lvb *res_lvb, *reply_lvb;
-        struct ldlm_reply *rep;
-        ldlm_error_t err;
-        int idx, rc, tmpflags = 0, only_liblustre = 1;
-        struct ldlm_interval_tree *tree;
-        struct filter_intent_args arg;
-        __u32 repsize[3] = { [MSG_PTLRPC_BODY_OFF] = sizeof(struct ptlrpc_body),
-                           [DLM_LOCKREPLY_OFF]   = sizeof(*rep),
-                           [DLM_REPLY_REC_OFF]   = sizeof(*reply_lvb) };
-       struct ldlm_glimpse_work        gl_work;
-       CFS_LIST_HEAD(gl_list);
-        ENTRY;
-
-        policy = ldlm_get_processing_policy(res);
-        LASSERT(policy != NULL);
-        LASSERT(req != NULL);
-
-        rc = lustre_pack_reply(req, 3, repsize, NULL);
-        if (rc)
-                RETURN(req->rq_status = rc);
-
-        rep = lustre_msg_buf(req->rq_repmsg, DLM_LOCKREPLY_OFF, sizeof(*rep));
-        LASSERT(rep != NULL);
-
-        reply_lvb = lustre_msg_buf(req->rq_repmsg, DLM_REPLY_REC_OFF,
-                                   sizeof(*reply_lvb));
-        LASSERT(reply_lvb != NULL);
-
-        //fixup_handle_for_resent_req(req, lock, &lockh);
-
-        /* Call the extent policy function to see if our request can be
-         * granted, or is blocked.
-         * If the OST lock has LDLM_FL_HAS_INTENT set, it means a glimpse
-         * lock, and should not be granted if the lock will be blocked.
-         */
-
-        if (flags & LDLM_FL_BLOCK_NOWAIT) {
-                OBD_FAIL_TIMEOUT(OBD_FAIL_LDLM_AGL_DELAY, 5);
-
-                if (OBD_FAIL_CHECK(OBD_FAIL_LDLM_AGL_NOLOCK))
-                        RETURN(ELDLM_LOCK_ABORTED);
-        }
-
-        LASSERT(ns == ldlm_res_to_ns(res));
-        lock_res(res);
-        rc = policy(lock, &tmpflags, 0, &err, NULL);
-        check_res_locked(res);
-
-        /* The lock met with no resistance; we're finished. */
-        if (rc == LDLM_ITER_CONTINUE) {
-                /* do not grant locks to the liblustre clients: they cannot
-                 * handle ASTs robustly.  We need to do this while still
-                 * holding lr_lock to avoid the lock remaining on the res_link
-                 * list (and potentially being added to l_pending_list by an
-                 * AST) when we are going to drop this lock ASAP. */
-                if (lock->l_export->exp_libclient ||
-                    OBD_FAIL_TIMEOUT(OBD_FAIL_LDLM_GLIMPSE, 2)) {
-                        ldlm_resource_unlink_lock(lock);
-                        err = ELDLM_LOCK_ABORTED;
-                } else {
-                        err = ELDLM_LOCK_REPLACED;
-                }
-                unlock_res(res);
-                RETURN(err);
-        } else if (flags & LDLM_FL_BLOCK_NOWAIT) {
-                /* LDLM_FL_BLOCK_NOWAIT means it is for AGL. Do not send glimpse
-                 * callback for glimpse size. The real size user will trigger
-                 * the glimpse callback when necessary. */
-                unlock_res(res);
-                RETURN(ELDLM_LOCK_ABORTED);
-        }
-
-        /* Do not grant any lock, but instead send GL callbacks.  The extent
-         * policy nicely created a list of all PW locks for us.  We will choose
-         * the highest of those which are larger than the size in the LVB, if
-         * any, and perform a glimpse callback. */
-        res_lvb = res->lr_lvb_data;
-        LASSERT(res_lvb != NULL);
-        *reply_lvb = *res_lvb;
-
-        /*
-         * lr_lock guarantees that no new locks are granted, and,
-         * therefore, that res->lr_lvb_data cannot increase beyond the
-         * end of already granted lock. As a result, it is safe to
-         * check against "stale" reply_lvb->lvb_size value without
-         * res->lr_lvb_mutex.
-         */
-        arg.size = reply_lvb->lvb_size;
-        arg.victim = &l;
-        arg.liblustre = &only_liblustre;
-        for (idx = 0; idx < LCK_MODE_NUM; idx++) {
-                tree = &res->lr_itree[idx];
-                if (tree->lit_mode == LCK_PR)
-                        continue;
-
-                interval_iterate_reverse(tree->lit_root,
-                                         filter_intent_cb, &arg);
-        }
-        unlock_res(res);
-
-        /* There were no PW locks beyond the size in the LVB; finished. */
-        if (l == NULL) {
-                if (only_liblustre) {
-                        /* If we discovered a liblustre client with a PW lock,
-                         * however, the LVB may be out of date!  The LVB is
-                         * updated only on glimpse (which we don't do for
-                         * liblustre clients) and cancel (which the client
-                         * obviously has not yet done).  So if it has written
-                         * data but kept the lock, the LVB is stale and needs
-                         * to be updated from disk.
-                         *
-                         * Of course, this will all disappear when we switch to
-                         * taking liblustre locks on the OST. */
-                        ldlm_res_lvbo_update(res, NULL, 1);
-                }
-                RETURN(ELDLM_LOCK_ABORTED);
-        }
-
-        /*
-         * This check is for lock taken in filter_prepare_destroy() that does
-         * not have l_glimpse_ast set. So the logic is: if there is a lock
-         * with no l_glimpse_ast set, this object is being destroyed already.
-         *
-         * Hence, if you are grabbing DLM locks on the server, always set
-         * non-NULL glimpse_ast (e.g., ldlm_request.c:ldlm_glimpse_ast()).
-         */
-        if (l->l_glimpse_ast == NULL) {
-                /* We are racing with unlink(); just return -ENOENT */
-                rep->lock_policy_res1 = -ENOENT;
-                goto out;
-        }
-
-        LASSERTF(l->l_glimpse_ast != NULL, "l == %p", l);
-
-       /* Populate the gl_work structure.
-        * Grab additional reference on the lock which will be released in
-        * ldlm_work_gl_ast_lock() */
-       gl_work.gl_lock = LDLM_LOCK_GET(l);
-       /* The glimpse callback is sent to one single extent lock. As a result,
-        * the gl_work list is just composed of one element */
-       cfs_list_add_tail(&gl_work.gl_list, &gl_list);
-       /* There is actually no need for a glimpse descriptor when glimpsing
-        * extent locks */
-       gl_work.gl_desc = NULL;
-       /* the ldlm_glimpse_work structure is allocated on the stack */
-       gl_work.gl_flags = LDLM_GL_WORK_NOFREE;
-
-       rc = ldlm_glimpse_locks(res, &gl_list); /* this will update the LVB */
-
-       if (!cfs_list_empty(&gl_list))
-               LDLM_LOCK_RELEASE(l);
-
-        lock_res(res);
-        *reply_lvb = *res_lvb;
-        unlock_res(res);
-
- out:
-        LDLM_LOCK_RELEASE(l);
-
-        RETURN(ELDLM_LOCK_ABORTED);
-}
-
-/*
- * per-obd_device iobuf pool.
- *
- * To avoid memory deadlocks in low-memory setups, amount of dynamic
- * allocations in write-path has to be minimized (see bug 5137).
- *
- * Pages, niobuf_local's and niobuf_remote's are pre-allocated and attached to
- * OST threads (see ost_thread_{init,done}()).
- *
- * "iobuf's" used by filter cannot be attached to OST thread, however, because
- * at the OST layer there are only (potentially) multiple obd_device of type
- * unknown at the time of OST thread creation.
- *
- * We create a cfs_hash for struct filter_obd (->fo_iobuf_hash field) on
- * initializing, each OST thread will create it's own iobuf on the first
- * access and insert it into ->fo_iobuf_hash with thread ID as key,
- * so the iobuf can be found again by thread ID.
- *
- * Functions below
- *
- *     filter_iobuf_pool_init()
- *
- *     filter_iobuf_pool_done()
- *
- *     filter_iobuf_get()
- *
- * operate on this array. They are "generic" in a sense that they don't depend
- * on actual type of iobuf's (the latter depending on Linux kernel version).
- */
-
-/*
- * destroy pool created by filter_iobuf_pool_init
- */
-static void filter_iobuf_pool_done(struct filter_obd *filter)
-{
-       ENTRY;
-
-       if (filter->fo_iobuf_hash != NULL) {
-               cfs_hash_putref(filter->fo_iobuf_hash);
-               filter->fo_iobuf_hash = NULL;
-       }
-       EXIT;
-}
-
-static int filter_adapt_sptlrpc_conf(struct obd_device *obd, int initial)
-{
-        struct filter_obd       *filter = &obd->u.filter;
-        struct sptlrpc_rule_set  tmp_rset;
-        int                      rc;
-
-        sptlrpc_rule_set_init(&tmp_rset);
-        rc = sptlrpc_conf_target_get_rules(obd, &tmp_rset, initial);
-        if (rc) {
-                CERROR("obd %s: failed get sptlrpc rules: %d\n",
-                       obd->obd_name, rc);
-                return rc;
-        }
-
-        sptlrpc_target_update_exp_flavor(obd, &tmp_rset);
-
-        cfs_write_lock(&filter->fo_sptlrpc_lock);
-        sptlrpc_rule_set_free(&filter->fo_sptlrpc_rset);
-        filter->fo_sptlrpc_rset = tmp_rset;
-        cfs_write_unlock(&filter->fo_sptlrpc_lock);
-
-        return 0;
-}
-
-static unsigned
-filter_iobuf_hop_hash(cfs_hash_t *hs, const void *key, unsigned mask)
-{
-       __u64   val = *((__u64 *)key);
-
-       return cfs_hash_long(val, hs->hs_cur_bits);
-}
-
-static void *
-filter_iobuf_hop_key(cfs_hlist_node_t *hnode)
-{
-       struct filter_iobuf     *pool;
-
-       pool = cfs_hlist_entry(hnode, struct filter_iobuf, dr_hlist);
-       return &pool->dr_hkey;
-}
-
-static int
-filter_iobuf_hop_keycmp(const void *key, cfs_hlist_node_t *hnode)
-{
-       struct filter_iobuf     *pool;
-
-       pool = cfs_hlist_entry(hnode, struct filter_iobuf, dr_hlist);
-       return pool->dr_hkey == *((__u64 *)key);
-}
-
-static void *
-filter_iobuf_hop_object(cfs_hlist_node_t *hnode)
-{
-       return cfs_hlist_entry(hnode, struct filter_iobuf, dr_hlist);
-}
-
-static void
-filter_iobuf_hop_get(cfs_hash_t *hs, cfs_hlist_node_t *hnode)
-{
-       /* dummy, required by cfs_hash */
-}
-
-static void
-filter_iobuf_hop_put_locked(cfs_hash_t *hs, cfs_hlist_node_t *hnode)
-{
-       /* dummy, required by cfs_hash */
-}
-
-static void
-filter_iobuf_hop_exit(cfs_hash_t *hs, cfs_hlist_node_t *hnode)
-{
-       struct filter_iobuf     *pool;
-
-       pool = cfs_hlist_entry(hnode, struct filter_iobuf, dr_hlist);
-       filter_free_iobuf(pool);
-}
-
-static struct cfs_hash_ops filter_iobuf_hops = {
-       .hs_hash        = filter_iobuf_hop_hash,
-       .hs_key         = filter_iobuf_hop_key,
-       .hs_keycmp      = filter_iobuf_hop_keycmp,
-       .hs_object      = filter_iobuf_hop_object,
-       .hs_get         = filter_iobuf_hop_get,
-       .hs_put_locked  = filter_iobuf_hop_put_locked,
-       .hs_exit        = filter_iobuf_hop_exit
-};
-
-#define FILTER_IOBUF_HASH_BITS 9
-#define FILTER_IOBUF_HBKT_BITS 4
-
-/*
- * pre-allocate pool of iobuf's to be used by filter_{prep,commit}rw_write().
- */
-static int filter_iobuf_pool_init(struct filter_obd *filter)
-{
-       filter->fo_iobuf_hash = cfs_hash_create("filter_iobuf",
-                                               FILTER_IOBUF_HASH_BITS,
-                                               FILTER_IOBUF_HASH_BITS,
-                                               FILTER_IOBUF_HBKT_BITS, 0,
-                                               CFS_HASH_MIN_THETA,
-                                               CFS_HASH_MAX_THETA,
-                                               &filter_iobuf_hops,
-                                               CFS_HASH_RW_BKTLOCK |
-                                               CFS_HASH_NO_ITEMREF);
-
-       return filter->fo_iobuf_hash != NULL ? 0 : -ENOMEM;
-}
-
-/* Return iobuf allocated for @thread_id.
- * If we haven't allocated a pool entry for this thread before, do so now and
- * insert it into fo_iobuf_hash, otherwise we can find it from fo_iobuf_hash */
-void *filter_iobuf_get(struct filter_obd *filter, struct obd_trans_info *oti)
-{
-       struct filter_iobuf     *pool = NULL;
-       __u64                   key = 0;
-       int                     thread_id;
-       int                     rc;
-
-       thread_id = (oti && oti->oti_thread) ? oti->oti_thread->t_id : -1;
-       if (thread_id >= 0) {
-               struct ptlrpc_service_part *svcpt;
-
-               svcpt = oti->oti_thread->t_svcpt;
-               LASSERT(svcpt != NULL);
-
-               key = (__u64)(svcpt->scp_cpt) << 32 | thread_id;
-               pool = cfs_hash_lookup(filter->fo_iobuf_hash, &key);
-               if (pool != NULL)
-                       return pool;
-       }
-
-       pool = filter_alloc_iobuf(filter, OBD_BRW_WRITE, PTLRPC_MAX_BRW_PAGES);
-       if (pool == NULL)
-               return NULL;
-
-       if (thread_id >= 0) {
-               pool->dr_hkey = key;
-               rc = cfs_hash_add_unique(filter->fo_iobuf_hash,
-                                        &key, &pool->dr_hlist);
-               /* ptlrpc service thould guarantee thread ID is unique */
-               LASSERT(rc != -EALREADY);
-       }
-
-       return pool;
-}
-
-/* mount the file system (secretly).  lustre_cfg parameters are:
- * 1 = device
- * 2 = fstype
- * 3 = flags: failover=f, failout=n
- * 4 = mount options
- */
-int filter_common_setup(struct obd_device *obd, struct lustre_cfg* lcfg,
-                        void *option)
-{
-        struct filter_obd *filter = &obd->u.filter;
-        struct vfsmount *mnt;
-        struct file_system_type *type;
-        struct lustre_mount_info *lmi;
-        struct obd_uuid uuid;
-        __u8 *uuid_ptr;
-        char *str, *label;
-        char ns_name[48];
-        struct request_queue *q;
-        int rc, i;
-        ENTRY;
-
-        if (lcfg->lcfg_bufcount < 3 ||
-            LUSTRE_CFG_BUFLEN(lcfg, 1) < 1 ||
-            LUSTRE_CFG_BUFLEN(lcfg, 2) < 1)
-                RETURN(-EINVAL);
-
-        lmi = server_get_mount(obd->obd_name);
-        if (lmi) {
-                /* We already mounted in lustre_fill_super.
-                   lcfg bufs 1, 2, 4 (device, fstype, mount opts) are ignored.*/
-                struct lustre_sb_info *lsi = s2lsi(lmi->lmi_sb);
-                mnt = lmi->lmi_mnt;
-               obd->obd_fsops = fsfilt_get_ops(lsi->lsi_fstype);
-        } else {
-                /* old path - used by lctl */
-                CERROR("Using old MDS mount method\n");
-                type = get_fs_type(lustre_cfg_string(lcfg, 2));
-                if (!type) {
-                        CERROR("get_fs_type failed\n");
-                        RETURN(-ENODEV);
-                }
-                mnt = vfs_kern_mount(type, MS_NOATIME|MS_NODIRATIME,
-                                     lustre_cfg_string(lcfg, 1), option);
-                cfs_module_put(type->owner);
-                if (IS_ERR(mnt)) {
-                        rc = PTR_ERR(mnt);
-                        LCONSOLE_ERROR_MSG(0x135, "Can't mount disk %s (%d)\n",
-                                           lustre_cfg_string(lcfg, 1), rc);
-                        RETURN(rc);
-                }
-
-                obd->obd_fsops = fsfilt_get_ops(lustre_cfg_string(lcfg, 2));
-        }
-        if (IS_ERR(obd->obd_fsops))
-                GOTO(err_mntput, rc = PTR_ERR(obd->obd_fsops));
-
-        rc = filter_iobuf_pool_init(filter);
-        if (rc != 0)
-                GOTO(err_ops, rc);
-
-        if (lvfs_check_rdonly(lvfs_sbdev(mnt->mnt_sb))) {
-                CERROR("%s: Underlying device is marked as read-only. "
-                       "Setup failed\n", obd->obd_name);
-                GOTO(err_ops, rc = -EROFS);
-        }
-
-        /* failover is the default */
-        obd->obd_replayable = 1;
-
-        /* disable connection until configuration finishes */
-        obd->obd_no_conn = 1;
-
-        if (lcfg->lcfg_bufcount > 3 && LUSTRE_CFG_BUFLEN(lcfg, 3) > 0) {
-                str = lustre_cfg_string(lcfg, 3);
-                if (strchr(str, 'n')) {
-                        CWARN("%s: recovery disabled\n", obd->obd_name);
-                        obd->obd_replayable = 0;
-                }
-        }
-
-        obd->u.obt.obt_magic = OBT_MAGIC;
-        obd->u.obt.obt_vfsmnt = mnt;
-        obd->u.obt.obt_sb = mnt->mnt_sb;
-        filter->fo_fstype = mnt->mnt_sb->s_type->name;
-        CDEBUG(D_SUPER, "%s: mnt = %p\n", filter->fo_fstype, mnt);
-
-        rc = fsfilt_setup(obd, obd->u.obt.obt_sb);
-        if (rc)
-                GOTO(err_ops, rc);
-
-        OBD_SET_CTXT_MAGIC(&obd->obd_lvfs_ctxt);
-        obd->obd_lvfs_ctxt.pwdmnt = mnt;
-        obd->obd_lvfs_ctxt.pwd = mnt->mnt_root;
-        obd->obd_lvfs_ctxt.fs = get_ds();
-        obd->obd_lvfs_ctxt.cb_ops = filter_lvfs_ops;
-
-        cfs_mutex_init(&filter->fo_init_lock);
-        filter->fo_committed_group = 0;
-        filter->fo_destroys_in_progress = 0;
-        for (i = 0; i < 32; i++)
-                cfs_mutex_init(&filter->fo_create_locks[i]);
-
-        cfs_spin_lock_init(&filter->fo_objidlock);
-        CFS_INIT_LIST_HEAD(&filter->fo_export_list);
-        cfs_mutex_init(&filter->fo_alloc_lock);
-        init_brw_stats(&filter->fo_filter_stats);
-        cfs_spin_lock_init(&filter->fo_flags_lock);
-        filter->fo_read_cache = 1; /* enable read-only cache by default */
-        filter->fo_writethrough_cache = 1; /* enable writethrough cache */
-        filter->fo_readcache_max_filesize = FILTER_MAX_CACHE_SIZE;
-        filter->fo_fmd_max_num = FILTER_FMD_MAX_NUM_DEFAULT;
-        filter->fo_fmd_max_age = FILTER_FMD_MAX_AGE_DEFAULT;
-        filter->fo_syncjournal = 0; /* Don't sync journals on i/o by default */
-        filter_slc_set(filter); /* initialize sync on lock cancel */
-
-        rc = filter_prep(obd);
-        if (rc)
-                GOTO(err_ops, rc);
-
-        CFS_INIT_LIST_HEAD(&filter->fo_llog_list);
-        cfs_spin_lock_init(&filter->fo_llog_list_lock);
-
-        filter->fo_fl_oss_capa = 1;
-
-        CFS_INIT_LIST_HEAD(&filter->fo_capa_keys);
-        filter->fo_capa_hash = init_capa_hash();
-        if (filter->fo_capa_hash == NULL)
-                GOTO(err_post, rc = -ENOMEM);
-
-        sprintf(ns_name, "filter-%s", obd->obd_uuid.uuid);
-        obd->obd_namespace = ldlm_namespace_new(obd, ns_name,
-                                                LDLM_NAMESPACE_SERVER,
-                                                LDLM_NAMESPACE_GREEDY,
-                                                LDLM_NS_TYPE_OST);
-        if (obd->obd_namespace == NULL)
-                GOTO(err_post, rc = -ENOMEM);
-        obd->obd_namespace->ns_lvbp = obd;
-        obd->obd_namespace->ns_lvbo = &filter_lvbo;
-        ldlm_register_intent(obd->obd_namespace, filter_intent_policy);
-
-        ptlrpc_init_client(LDLM_CB_REQUEST_PORTAL, LDLM_CB_REPLY_PORTAL,
-                           "filter_ldlm_cb_client", &obd->obd_ldlm_client);
-
-        rc = obd_llog_init(obd, &obd->obd_olg, obd, NULL);
-        if (rc) {
-                CERROR("failed to setup llogging subsystems\n");
-                GOTO(err_post, rc);
-        }
-
-        cfs_rwlock_init(&filter->fo_sptlrpc_lock);
-        sptlrpc_rule_set_init(&filter->fo_sptlrpc_rset);
-        /* do this after llog being initialized */
-        filter_adapt_sptlrpc_conf(obd, 1);
-
-        q = bdev_get_queue(mnt->mnt_sb->s_bdev);
-        if (queue_max_sectors(q) < queue_max_hw_sectors(q) &&
-            queue_max_sectors(q) < PTLRPC_MAX_BRW_SIZE >> 9)
-                LCONSOLE_INFO("%s: underlying device %s should be tuned "
-                              "for larger I/O requests: max_sectors = %u "
-                              "could be up to max_hw_sectors=%u\n",
-                              obd->obd_name, mnt->mnt_sb->s_id,
-                              queue_max_sectors(q), queue_max_hw_sectors(q));
-
-        uuid_ptr = fsfilt_uuid(obd, obd->u.obt.obt_sb);
-        if (uuid_ptr != NULL) {
-                class_uuid_unparse(uuid_ptr, &uuid);
-                str = uuid.uuid;
-        } else {
-                str = "no UUID";
-        }
-
-        label = fsfilt_get_label(obd, obd->u.obt.obt_sb);
-        LCONSOLE_INFO("%s: Now serving %s %s%s with recovery %s\n",
-                      obd->obd_name, label ?: str, lmi ? "on " : "",
-                      lmi ? s2lsi(lmi->lmi_sb)->lsi_lmd->lmd_dev : "",
-                      obd->obd_replayable ? "enabled" : "disabled");
-
-        RETURN(0);
-
-err_post:
-        filter_post(obd);
-err_ops:
-        fsfilt_put_ops(obd->obd_fsops);
-        filter_iobuf_pool_done(filter);
-err_mntput:
-        server_put_mount(obd->obd_name, mnt);
-        obd->u.obt.obt_sb = 0;
-        return rc;
-}
-
-static int filter_setup(struct obd_device *obd, struct lustre_cfg* lcfg)
-{
-        struct lprocfs_static_vars lvars;
-        cfs_proc_dir_entry_t *entry;
-        unsigned long addr;
-        struct page *page;
-        int rc;
-        ENTRY;
-
-        CLASSERT(offsetof(struct obd_device, u.obt) ==
-                 offsetof(struct obd_device, u.filter.fo_obt));
-
-        if (!LUSTRE_CFG_BUFLEN(lcfg, 1) || !LUSTRE_CFG_BUFLEN(lcfg, 2))
-                RETURN(-EINVAL);
-
-        /* lprocfs must be setup before the filter so state can be safely added
-         * to /proc incrementally as the filter is setup */
-        lprocfs_filter_init_vars(&lvars);
-        rc = lprocfs_obd_setup(obd, lvars.obd_vars);
-        if (rc) {
-                CERROR("%s: lprocfs_obd_setup failed: %d.\n",
-                       obd->obd_name, rc);
-                RETURN(rc);
-        }
-
-        rc = lprocfs_alloc_obd_stats(obd, LPROC_FILTER_LAST);
-        if (rc) {
-                CERROR("%s: lprocfs_alloc_obd_stats failed: %d.\n",
-                       obd->obd_name, rc);
-                GOTO(obd_cleanup, rc);
-        }
-
-        /* Init obdfilter private stats here */
-        lprocfs_counter_init(obd->obd_stats, LPROC_FILTER_READ_BYTES,
-                             LPROCFS_CNTR_AVGMINMAX, "read_bytes", "bytes");
-        lprocfs_counter_init(obd->obd_stats, LPROC_FILTER_WRITE_BYTES,
-                             LPROCFS_CNTR_AVGMINMAX, "write_bytes", "bytes");
-        lprocfs_counter_init(obd->obd_stats, LPROC_FILTER_GET_PAGE,
-                             LPROCFS_CNTR_AVGMINMAX|LPROCFS_CNTR_STDDEV,
-                             "get_page", "usec");
-        lprocfs_counter_init(obd->obd_stats, LPROC_FILTER_NO_PAGE,
-                             LPROCFS_CNTR_AVGMINMAX, "get_page_failures", "num");
-        lprocfs_counter_init(obd->obd_stats, LPROC_FILTER_CACHE_ACCESS,
-                             LPROCFS_CNTR_AVGMINMAX, "cache_access", "pages");
-        lprocfs_counter_init(obd->obd_stats, LPROC_FILTER_CACHE_HIT,
-                             LPROCFS_CNTR_AVGMINMAX, "cache_hit", "pages");
-        lprocfs_counter_init(obd->obd_stats, LPROC_FILTER_CACHE_MISS,
-                             LPROCFS_CNTR_AVGMINMAX, "cache_miss", "pages");
-
-        rc = lproc_filter_attach_seqstat(obd);
-        if (rc) {
-                CERROR("%s: create seqstat failed: %d.\n", obd->obd_name, rc);
-                GOTO(free_obd_stats, rc);
-        }
-
-        entry = lprocfs_register("exports", obd->obd_proc_entry, NULL, NULL);
-        if (IS_ERR(entry)) {
-                rc = PTR_ERR(entry);
-                CERROR("%s: error %d setting up lprocfs for %s\n",
-                       obd->obd_name, rc, "exports");
-                GOTO(free_obd_stats, rc);
-        }
-        obd->obd_proc_exports_entry = entry;
-
-        entry = lprocfs_add_simple(obd->obd_proc_exports_entry, "clear",
-                                   lprocfs_nid_stats_clear_read,
-                                   lprocfs_nid_stats_clear_write, obd, NULL);
-        if (IS_ERR(entry)) {
-                rc = PTR_ERR(entry);
-                CERROR("%s: add proc entry 'clear' failed: %d.\n",
-                       obd->obd_name, rc);
-                GOTO(free_obd_stats, rc);
-        }
-
-       rc = lprocfs_job_stats_init(obd, LPROC_FILTER_STATS_LAST,
-                                   filter_stats_counter_init);
-       if (rc)
-               GOTO(remove_entry_clear, rc);
-
-        /* 2.6.9 selinux wants a full option page for do_kern_mount (bug6471) */
-        OBD_PAGE_ALLOC(page, CFS_ALLOC_STD);
-        if (!page)
-               GOTO(job_stats_fini, rc = -ENOMEM);
-        addr = (unsigned long)cfs_page_address(page);
-        clear_page((void *)addr);
-        memcpy((void *)addr, lustre_cfg_buf(lcfg, 4),
-               LUSTRE_CFG_BUFLEN(lcfg, 4));
-        rc = filter_common_setup(obd, lcfg, (void *)addr);
-        OBD_PAGE_FREE(page);
-        if (rc) {
-                CERROR("%s: filter_common_setup failed: %d.\n",
-                       obd->obd_name, rc);
-               GOTO(job_stats_fini, rc);
-        }
-
-        RETURN(0);
-
-job_stats_fini:
-       lprocfs_job_stats_fini(obd);
-remove_entry_clear:
-        lprocfs_remove_proc_entry("clear", obd->obd_proc_exports_entry);
-free_obd_stats:
-        lprocfs_free_obd_stats(obd);
-obd_cleanup:
-        lprocfs_obd_cleanup(obd);
-        return rc;
-}
-
-static struct llog_operations filter_mds_ost_repl_logops;
-
-static struct llog_operations filter_size_orig_logops = {};
-
-static int filter_olg_fini(struct obd_llog_group *olg)
-{
-       struct llog_ctxt *ctxt;
-
-       ENTRY;
-
-       ctxt = llog_group_get_ctxt(olg, LLOG_MDS_OST_REPL_CTXT);
-       if (ctxt)
-               llog_cleanup(NULL, ctxt);
-
-       ctxt = llog_group_get_ctxt(olg, LLOG_SIZE_ORIG_CTXT);
-       if (ctxt)
-               llog_cleanup(NULL, ctxt);
-
-       ctxt = llog_group_get_ctxt(olg, LLOG_CONFIG_ORIG_CTXT);
-       if (ctxt)
-               llog_cleanup(NULL, ctxt);
-
-       RETURN(0);
-}
-
-static int
-filter_olg_init(struct obd_device *obd, struct obd_llog_group *olg,
-                struct obd_device *tgt)
-{
-       struct llog_ctxt        *ctxt = NULL;
-       int                      rc;
-
-       ENTRY;
-
-       filter_mds_ost_repl_logops = llog_client_ops;
-       filter_mds_ost_repl_logops.lop_cancel = llog_obd_repl_cancel;
-       filter_mds_ost_repl_logops.lop_connect = llog_obd_repl_connect;
-       filter_mds_ost_repl_logops.lop_sync = llog_obd_repl_sync;
-
-       rc = llog_setup(NULL, obd, olg, LLOG_MDS_OST_REPL_CTXT, tgt,
-                       &filter_mds_ost_repl_logops);
-       if (rc)
-               RETURN(rc);
-
-       rc = llog_setup(NULL, obd, olg, LLOG_SIZE_ORIG_CTXT, tgt,
-                       &filter_size_orig_logops);
-       if (rc)
-               GOTO(out, rc);
-
-       RETURN(0);
-out:
-       ctxt = llog_group_get_ctxt(olg, LLOG_MDS_OST_REPL_CTXT);
-       llog_cleanup(NULL, ctxt);
-       return rc;
-}
-
-/**
- * Init the default olg, which is embeded in the obd_device, for filter.
- */
-static int
-filter_default_olg_init(struct obd_device *obd, struct obd_llog_group *olg,
-                        struct obd_device *tgt)
-{
-        struct filter_obd *filter = &obd->u.filter;
-        struct llog_ctxt *ctxt;
-        int rc;
-        ENTRY;
-
-        filter->fo_lcm = llog_recov_thread_init(obd->obd_name);
-        if (!filter->fo_lcm)
-                RETURN(-ENOMEM);
-
-        rc = filter_olg_init(obd, olg, tgt);
-        if (rc)
-                GOTO(cleanup_lcm, rc);
-
-       rc = llog_setup(NULL, obd, olg, LLOG_CONFIG_ORIG_CTXT, tgt,
-                       &llog_lvfs_ops);
-        if (rc)
-                GOTO(cleanup_olg, rc);
-
-        ctxt = llog_group_get_ctxt(olg, LLOG_MDS_OST_REPL_CTXT);
-        if (!ctxt) {
-                CERROR("Can't get ctxt for %p:%x\n", olg,
-                       LLOG_MDS_OST_REPL_CTXT);
-                GOTO(cleanup_olg, rc = -ENODEV);
-        }
-        ctxt->loc_lcm = lcm_get(filter->fo_lcm);
-        ctxt->llog_proc_cb = filter_recov_log_mds_ost_cb;
-        llog_ctxt_put(ctxt);
-
-        RETURN(0);
-cleanup_olg:
-        filter_olg_fini(olg);
-cleanup_lcm:
-        llog_recov_thread_fini(filter->fo_lcm, 1);
-        filter->fo_lcm = NULL;
-        return rc;
-}
-
-static int
-filter_llog_init(struct obd_device *obd, struct obd_llog_group *olg,
-                 struct obd_device *tgt, int *index)
-{
-        struct filter_obd *filter = &obd->u.filter;
-        struct llog_ctxt *ctxt;
-        int rc;
-        ENTRY;
-
-        LASSERT(olg != NULL);
-        if (olg == &obd->obd_olg)
-                return filter_default_olg_init(obd, olg, tgt);
-
-        LASSERT(filter->fo_lcm != NULL);
-        rc = filter_olg_init(obd, olg, tgt);
-        if (rc)
-                RETURN(rc);
-        ctxt = llog_group_get_ctxt(olg, LLOG_MDS_OST_REPL_CTXT);
-        if (!ctxt) {
-                CERROR("Can't get ctxt for %p:%x\n", olg,
-                       LLOG_MDS_OST_REPL_CTXT);
-                filter_olg_fini(olg);
-                RETURN(-ENODEV);
-        }
-        ctxt->llog_proc_cb = filter_recov_log_mds_ost_cb;
-        ctxt->loc_lcm = lcm_get(filter->fo_lcm);
-        llog_ctxt_put(ctxt);
-        RETURN(rc);
-}
-
-static int filter_llog_finish(struct obd_device *obd, int count)
-{
-        struct filter_obd *filter = &obd->u.filter;
-        struct llog_ctxt *ctxt;
-        ENTRY;
-
-        ctxt = llog_group_get_ctxt(&obd->obd_olg, LLOG_MDS_OST_REPL_CTXT);
-        if (ctxt) {
-                /*
-                 * Make sure that no cached llcds left in recov_thread.
-                 * We actually do sync in disconnect time, but disconnect
-                 * may not come being marked rq_no_resend = 1.
-                 */
-               llog_sync(ctxt, NULL, OBD_LLOG_FL_EXIT);
-
-                /*
-                 * Balance class_import_get() in llog_receptor_accept().
-                 * This is safe to do, as llog is already synchronized
-                 * and its import may go.
-                 */
-                cfs_mutex_lock(&ctxt->loc_mutex);
-                if (ctxt->loc_imp) {
-                        class_import_put(ctxt->loc_imp);
-                        ctxt->loc_imp = NULL;
-                }
-
-               if (filter->fo_lcm) {
-                       llog_recov_thread_fini(filter->fo_lcm, obd->obd_force);
-                       filter->fo_lcm = NULL;
-               }
-
-                cfs_mutex_unlock(&ctxt->loc_mutex);
-                llog_ctxt_put(ctxt);
-        }
-
-        RETURN(filter_olg_fini(&obd->obd_olg));
-}
-
-/**
- * Find the group llog according to group index in the llog group list.
- */
-static struct obd_llog_group *
-filter_find_olg_internal(struct filter_obd *filter, int group)
-{
-        struct obd_llog_group *olg;
-
-        LASSERT_SPIN_LOCKED(&filter->fo_llog_list_lock);
-        cfs_list_for_each_entry(olg, &filter->fo_llog_list, olg_list) {
-                if (olg->olg_seq == group)
-                        RETURN(olg);
-        }
-        RETURN(NULL);
-}
-
-/**
- * Find the group llog according to group index on the filter
- */
-struct obd_llog_group *filter_find_olg(struct obd_device *obd, int group)
-{
-        struct obd_llog_group *olg = NULL;
-        struct filter_obd *filter;
-
-        filter = &obd->u.filter;
-
-        if (group == FID_SEQ_LLOG)
-                RETURN(&obd->obd_olg);
-
-        cfs_spin_lock(&filter->fo_llog_list_lock);
-        olg = filter_find_olg_internal(filter, group);
-        cfs_spin_unlock(&filter->fo_llog_list_lock);
-
-        RETURN(olg);
-}
-/**
- * Find the llog_group of the filter according to the group. If it can not
- * find, create the llog_group, which only happens when mds is being synced
- * with OST.
- */
-struct obd_llog_group *filter_find_create_olg(struct obd_device *obd, int group)
-{
-        struct obd_llog_group *olg = NULL, *olg_new = NULL;
-        struct filter_obd *filter;
-        int rc;
-
-        filter = &obd->u.filter;
-
-        if (group == FID_SEQ_LLOG)
-                RETURN(&obd->obd_olg);
-
-        OBD_ALLOC_PTR(olg_new);
-        if (olg_new == NULL)
-               RETURN(ERR_PTR(-ENOMEM));
-
-        cfs_spin_lock(&filter->fo_llog_list_lock);
-        olg = filter_find_olg_internal(filter, group);
-        if (olg) {
-                if (olg->olg_initializing) {
-                        GOTO(out_unlock, olg = ERR_PTR(-EBUSY));
-                } else {
-                        GOTO(out_unlock, olg);
-                }
-        } else {
-                /* set as the newly allocated one */
-                olg = olg_new;
-                olg_new = NULL;
-        }
-
-        llog_group_init(olg, group);
-        cfs_list_add(&olg->olg_list, &filter->fo_llog_list);
-        olg->olg_initializing = 1;
-        cfs_spin_unlock(&filter->fo_llog_list_lock);
-
-        rc = obd_llog_init(obd, olg, obd, NULL);
-        if (rc) {
-               cfs_spin_lock(&filter->fo_llog_list_lock);
-               cfs_list_del(&olg->olg_list);
-               cfs_spin_unlock(&filter->fo_llog_list_lock);
-               OBD_FREE_PTR(olg);
-               GOTO(out, olg = ERR_PTR(-ENOMEM));
-        }
-        cfs_spin_lock(&filter->fo_llog_list_lock);
-        olg->olg_initializing = 0;
-        cfs_spin_unlock(&filter->fo_llog_list_lock);
-        CDEBUG(D_OTHER, "%s: new llog group %u (0x%p)\n",
-              obd->obd_name, group, olg);
-out:
-        RETURN(olg);
-
-out_unlock:
-        cfs_spin_unlock(&filter->fo_llog_list_lock);
-        if (olg_new)
-               OBD_FREE_PTR(olg_new);
-        goto out;
-}
-
-static int filter_llog_connect(struct obd_export *exp,
-                               struct llogd_conn_body *body)
-{
-        struct obd_device *obd = exp->exp_obd;
-        struct llog_ctxt *ctxt;
-        struct obd_llog_group *olg;
-        int rc;
-        ENTRY;
-
-        CDEBUG(D_OTHER, "%s: LLog connect for: "LPX64"/"LPX64":%x\n",
-               obd->obd_name, body->lgdc_logid.lgl_oid,
-               body->lgdc_logid.lgl_oseq, body->lgdc_logid.lgl_ogen);
-
-        olg = filter_find_olg(obd, body->lgdc_logid.lgl_oseq);
-        if (!olg) {
-                CERROR(" %s: can not find olg of group %d\n",
-                       obd->obd_name, (int)body->lgdc_logid.lgl_oseq);
-                RETURN(-ENOENT);
-        }
-        llog_group_set_export(olg, exp);
-
-        ctxt = llog_group_get_ctxt(olg, body->lgdc_ctxt_idx);
-        LASSERTF(ctxt != NULL, "ctxt is not null, ctxt idx %d \n",
-                 body->lgdc_ctxt_idx);
-
-        CDEBUG(D_HA, "%s: Recovery from log "LPX64"/"LPX64":%x\n",
-               obd->obd_name, body->lgdc_logid.lgl_oid,
-               body->lgdc_logid.lgl_oseq, body->lgdc_logid.lgl_ogen);
-
-        cfs_spin_lock(&obd->u.filter.fo_flags_lock);
-        obd->u.filter.fo_mds_ost_sync = 1;
-        cfs_spin_unlock(&obd->u.filter.fo_flags_lock);
-        rc = llog_connect(ctxt, &body->lgdc_logid,
-                          &body->lgdc_gen, NULL);
-        llog_ctxt_put(ctxt);
-        if (rc != 0)
-                CERROR("failed to connect rc %d idx %d\n", rc,
-                                body->lgdc_ctxt_idx);
-
-        RETURN(rc);
-}
-
-static int filter_llog_preclean(struct obd_device *obd)
-{
-        struct obd_llog_group *olg, *tmp;
-        struct filter_obd *filter;
-        cfs_list_t  remove_list;
-        int rc = 0;
-        ENTRY;
-
-        rc = obd_llog_finish(obd, 0);
-        if (rc)
-                CERROR("failed to cleanup llogging subsystem\n");
-
-        filter = &obd->u.filter;
-        CFS_INIT_LIST_HEAD(&remove_list);
-
-        cfs_spin_lock(&filter->fo_llog_list_lock);
-        while (!cfs_list_empty(&filter->fo_llog_list)) {
-                olg = cfs_list_entry(filter->fo_llog_list.next,
-                                     struct obd_llog_group, olg_list);
-                cfs_list_del(&olg->olg_list);
-                cfs_list_add(&olg->olg_list, &remove_list);
-        }
-        cfs_spin_unlock(&filter->fo_llog_list_lock);
-
-        cfs_list_for_each_entry_safe(olg, tmp, &remove_list, olg_list) {
-                cfs_list_del_init(&olg->olg_list);
-                rc = filter_olg_fini(olg);
-                if (rc)
-                        CERROR("failed to cleanup llogging subsystem for %u\n",
-                               olg->olg_seq);
-                OBD_FREE_PTR(olg);
-        }
-
-        RETURN(rc);
-}
-
-static int filter_precleanup(struct obd_device *obd,
-                             enum obd_cleanup_stage stage)
-{
-        int rc = 0;
-        ENTRY;
-
-        switch(stage) {
-        case OBD_CLEANUP_EARLY:
-                break;
-        case OBD_CLEANUP_EXPORTS:
-                /* Stop recovery before namespace cleanup. */
-                target_recovery_fini(obd);
-
-                obd_exports_barrier(obd);
-                obd_zombie_barrier();
-
-                rc = filter_llog_preclean(obd);
-               lprocfs_job_stats_fini(obd);
-                lprocfs_remove_proc_entry("clear", obd->obd_proc_exports_entry);
-                lprocfs_free_per_client_stats(obd);
-                lprocfs_obd_cleanup(obd);
-                lprocfs_free_obd_stats(obd);
-                break;
-        }
-        RETURN(rc);
-}
-
-static int filter_cleanup(struct obd_device *obd)
-{
-        struct filter_obd *filter = &obd->u.filter;
-        ENTRY;
-
-        if (obd->obd_fail)
-                LCONSOLE_WARN("%s: shutting down for failover; client state "
-                              "will be preserved.\n", obd->obd_name);
-
-        ldlm_namespace_free(obd->obd_namespace, NULL, obd->obd_force);
-        obd->obd_namespace = NULL;
-
-        sptlrpc_rule_set_free(&filter->fo_sptlrpc_rset);
-
-        if (obd->u.obt.obt_sb == NULL)
-                RETURN(0);
-
-        filter_post(obd);
-
-        ll_vfs_dq_off(obd->u.obt.obt_sb, 0);
-        shrink_dcache_sb(obd->u.obt.obt_sb);
-
-        server_put_mount(obd->obd_name, obd->u.obt.obt_vfsmnt);
-        obd->u.obt.obt_sb = NULL;
-
-        fsfilt_put_ops(obd->obd_fsops);
-
-        filter_iobuf_pool_done(filter);
-
-        LCONSOLE_INFO("OST %s has stopped.\n", obd->obd_name);
-
-        RETURN(0);
-}
-
-static int filter_connect_internal(struct obd_export *exp,
-                                   struct obd_connect_data *data,
-                                   int reconnect)
-{
-        struct filter_export_data *fed = &exp->exp_filter_data;
-
-        if (!data)
-                RETURN(0);
-
-        CDEBUG(D_RPCTRACE, "%s: cli %s/%p ocd_connect_flags: "LPX64
-               " ocd_version: %x ocd_grant: %d ocd_index: %u\n",
-               exp->exp_obd->obd_name, exp->exp_client_uuid.uuid, exp,
-               data->ocd_connect_flags, data->ocd_version,
-               data->ocd_grant, data->ocd_index);
-
-        if (fed->fed_group != 0 && fed->fed_group != data->ocd_group) {
-                CWARN("!!! This export (nid %s) used object group %d "
-                       "earlier; now it's trying to use group %d!  This could "
-                       "be a bug in the MDS. Please report to "
-                       "http://bugs.whamcloud.com/\n",
-                       obd_export_nid2str(exp), fed->fed_group,data->ocd_group);
-                RETURN(-EPROTO);
-        }
-        fed->fed_group = data->ocd_group;
-
-        data->ocd_connect_flags &= OST_CONNECT_SUPPORTED;
-        exp->exp_connect_flags = data->ocd_connect_flags;
-        data->ocd_version = LUSTRE_VERSION_CODE;
-
-        /* Kindly make sure the SKIP_ORPHAN flag is from MDS. */
-        if (data->ocd_connect_flags & OBD_CONNECT_MDS)
-                CDEBUG(D_HA, "%s: Received MDS connection for group %u\n",
-                       exp->exp_obd->obd_name, data->ocd_group);
-        else if (data->ocd_connect_flags & OBD_CONNECT_SKIP_ORPHAN)
-                RETURN(-EPROTO);
-
-        if (exp->exp_connect_flags & OBD_CONNECT_GRANT) {
-                struct filter_obd *filter = &exp->exp_obd->u.filter;
-                obd_size left, want;
-
-                cfs_spin_lock(&exp->exp_obd->obd_osfs_lock);
-                left = filter_grant_space_left(exp);
-                want = data->ocd_grant;
-                filter_grant(exp, fed->fed_grant, want, left, (reconnect == 0));
-                data->ocd_grant = fed->fed_grant;
-                cfs_spin_unlock(&exp->exp_obd->obd_osfs_lock);
-
-                CDEBUG(D_CACHE, "%s: cli %s/%p ocd_grant: %d want: "
-                       LPU64" left: "LPU64"\n", exp->exp_obd->obd_name,
-                       exp->exp_client_uuid.uuid, exp,
-                       data->ocd_grant, want, left);
-
-                filter->fo_tot_granted_clients ++;
-        }
-
-        if (data->ocd_connect_flags & OBD_CONNECT_INDEX) {
-                struct lr_server_data *lsd = class_server_data(exp->exp_obd);
-                int index = le32_to_cpu(lsd->lsd_ost_index);
-
-                if (!(lsd->lsd_feature_compat &
-                      cpu_to_le32(OBD_COMPAT_OST))) {
-                        /* this will only happen on the first connect */
-                        lsd->lsd_ost_index = cpu_to_le32(data->ocd_index);
-                        lsd->lsd_feature_compat |= cpu_to_le32(OBD_COMPAT_OST);
-                        /* sync is not needed here as filter_client_add will
-                         * set exp_need_sync flag */
-                        filter_update_server_data(exp->exp_obd);
-                } else if (index != data->ocd_index) {
-                        LCONSOLE_ERROR_MSG(0x136, "Connection from %s to index"
-                                           " %u doesn't match actual OST index"
-                                           " %u in last_rcvd file, bad "
-                                           "configuration?\n",
-                                           obd_export_nid2str(exp), index,
-                                           data->ocd_index);
-                        RETURN(-EBADF);
-                }
-                /* FIXME: Do the same with the MDS UUID and lsd_peeruuid.
-                 * FIXME: We don't strictly need the COMPAT flag for that,
-                 * FIXME: as lsd_peeruuid[0] will tell us if that is set.
-                 * FIXME: We needed it for the index, as index 0 is valid. */
-        }
-
-        if (OBD_FAIL_CHECK(OBD_FAIL_OST_BRW_SIZE)) {
-                data->ocd_brw_size = 65536;
-        } else if (data->ocd_connect_flags & OBD_CONNECT_BRW_SIZE) {
-                data->ocd_brw_size = min(data->ocd_brw_size,
-                               (__u32)(PTLRPC_MAX_BRW_PAGES << CFS_PAGE_SHIFT));
-                if (data->ocd_brw_size == 0) {
-                        CERROR("%s: cli %s/%p ocd_connect_flags: "LPX64
-                               " ocd_version: %x ocd_grant: %d ocd_index: %u "
-                               "ocd_brw_size is unexpectedly zero, "
-                               "network data corruption?"
-                               "Refusing connection of this client\n",
-                                exp->exp_obd->obd_name,
-                                exp->exp_client_uuid.uuid,
-                                exp, data->ocd_connect_flags, data->ocd_version,
-                                data->ocd_grant, data->ocd_index);
-                        RETURN(-EPROTO);
-                }
-        }
-
-        if (data->ocd_connect_flags & OBD_CONNECT_CKSUM) {
-                __u32 cksum_types = data->ocd_cksum_types;
-
-                /* The client set in ocd_cksum_types the checksum types it
-                 * supports. We have to mask off the algorithms that we don't
-                 * support */
-               data->ocd_cksum_types &= cksum_types_supported_server();
-
-               /* 1.6.4 clients are not supported any more */
-
-                CDEBUG(D_RPCTRACE, "%s: cli %s supports cksum type %x, return "
-                                   "%x\n", exp->exp_obd->obd_name,
-                                   obd_export_nid2str(exp), cksum_types,
-                                   data->ocd_cksum_types);
-        } else {
-                /* This client does not support OBD_CONNECT_CKSUM
-                 * fall back to CRC32 */
-                CDEBUG(D_RPCTRACE, "%s: cli %s does not support "
-                                   "OBD_CONNECT_CKSUM, CRC32 will be used\n",
-                                   exp->exp_obd->obd_name,
-                                   obd_export_nid2str(exp));
-        }
-
-        if (data->ocd_connect_flags & OBD_CONNECT_MAXBYTES)
-                data->ocd_maxbytes = exp->exp_obd->u.obt.obt_sb->s_maxbytes;
-
-        RETURN(0);
-}
-
-static int filter_reconnect(const struct lu_env *env,
-                            struct obd_export *exp, struct obd_device *obd,
-                            struct obd_uuid *cluuid,
-                            struct obd_connect_data *data,
-                            void *localdata)
-{
-        int rc;
-        ENTRY;
-
-        if (exp == NULL || obd == NULL || cluuid == NULL)
-                RETURN(-EINVAL);
-
-        rc = filter_connect_internal(exp, data, 1);
-        if (rc == 0)
-                filter_export_stats_init(obd, exp, localdata);
-
-        RETURN(rc);
-}
-
-static int filter_connect(const struct lu_env *env,
-                          struct obd_export **exp, struct obd_device *obd,
-                          struct obd_uuid *cluuid,
-                          struct obd_connect_data *data, void *localdata)
-{
-        struct lvfs_run_ctxt saved;
-        struct lustre_handle conn = { 0 };
-        struct obd_export *lexp;
-        int rc;
-        ENTRY;
-
-        if (exp == NULL || obd == NULL || cluuid == NULL)
-                RETURN(-EINVAL);
-
-        rc = class_connect(&conn, obd, cluuid);
-        if (rc)
-                RETURN(rc);
-        lexp = class_conn2export(&conn);
-        LASSERT(lexp != NULL);
-
-        rc = filter_connect_internal(lexp, data, 0);
-        if (rc)
-                GOTO(cleanup, rc);
-
-        filter_export_stats_init(obd, lexp, localdata);
-        if (obd->obd_replayable) {
-                struct lsd_client_data *lcd = lexp->exp_target_data.ted_lcd;
-                LASSERT(lcd);
-                memcpy(lcd->lcd_uuid, cluuid, sizeof(lcd->lcd_uuid));
-                rc = filter_client_add(obd, lexp, -1);
-                if (rc)
-                        GOTO(cleanup, rc);
-        }
-
-        push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-        rc = filter_read_groups(obd, data->ocd_group, 1);
-        pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-        if (rc != 0) {
-                CERROR("can't read group %u\n", data->ocd_group);
-                GOTO(cleanup, rc);
-        }
-
-        GOTO(cleanup, rc);
-
-cleanup:
-        if (rc) {
-                class_disconnect(lexp);
-                *exp = NULL;
-        } else {
-                *exp = lexp;
-        }
-
-        RETURN(rc);
-}
-
-/* Do extra sanity checks for grant accounting.  We do this at connect,
- * disconnect, and statfs RPC time, so it shouldn't be too bad.  We can
- * always get rid of it or turn it off when we know accounting is good. */
-static void filter_grant_sanity_check(struct obd_device *obd, const char *func)
-{
-        struct filter_export_data *fed;
-        struct obd_export *exp;
-        obd_size maxsize = obd->obd_osfs.os_blocks * obd->obd_osfs.os_bsize;
-        obd_size tot_dirty = 0, tot_pending = 0, tot_granted = 0;
-        obd_size fo_tot_dirty, fo_tot_pending, fo_tot_granted;
-
-        if (cfs_list_empty(&obd->obd_exports))
-                return;
-
-        /* We don't want to do this for large machines that do lots of
-           mounts or unmounts.  It burns... */
-        if (obd->obd_num_exports > 100)
-                return;
-
-        cfs_spin_lock(&obd->obd_osfs_lock);
-        cfs_spin_lock(&obd->obd_dev_lock);
-        cfs_list_for_each_entry(exp, &obd->obd_exports, exp_obd_chain) {
-                int error = 0;
-                fed = &exp->exp_filter_data;
-                if (fed->fed_grant < 0 || fed->fed_pending < 0 ||
-                    fed->fed_dirty < 0)
-                        error = 1;
-                if (maxsize > 0) { /* we may not have done a statfs yet */
-                        LASSERTF(fed->fed_grant + fed->fed_pending <= maxsize,
-                                 "%s: cli %s/%p %ld+%ld > "LPU64"\n", func,
-                                 exp->exp_client_uuid.uuid, exp,
-                                 fed->fed_grant, fed->fed_pending, maxsize);
-                        LASSERTF(fed->fed_dirty <= maxsize,
-                                 "%s: cli %s/%p %ld > "LPU64"\n", func,
-                                 exp->exp_client_uuid.uuid, exp,
-                                 fed->fed_dirty, maxsize);
-                }
-                if (error)
-                        CERROR("%s: cli %s/%p dirty %ld pend %ld grant %ld\n",
-                               obd->obd_name, exp->exp_client_uuid.uuid, exp,
-                               fed->fed_dirty, fed->fed_pending,fed->fed_grant);
-                else
-                        CDEBUG(D_CACHE, "%s: cli %s/%p dirty %ld pend %ld grant %ld\n",
-                               obd->obd_name, exp->exp_client_uuid.uuid, exp,
-                               fed->fed_dirty, fed->fed_pending,fed->fed_grant);
-                tot_granted += fed->fed_grant + fed->fed_pending;
-                tot_pending += fed->fed_pending;
-                tot_dirty += fed->fed_dirty;
-        }
-        fo_tot_granted = obd->u.filter.fo_tot_granted;
-        fo_tot_pending = obd->u.filter.fo_tot_pending;
-        fo_tot_dirty = obd->u.filter.fo_tot_dirty;
-        cfs_spin_unlock(&obd->obd_dev_lock);
-        cfs_spin_unlock(&obd->obd_osfs_lock);
-
-        /* Do these assertions outside the spinlocks so we don't kill system */
-        if (tot_granted != fo_tot_granted)
-                CERROR("%s: tot_granted "LPU64" != fo_tot_granted "LPU64"\n",
-                       func, tot_granted, fo_tot_granted);
-        if (tot_pending != fo_tot_pending)
-                CERROR("%s: tot_pending "LPU64" != fo_tot_pending "LPU64"\n",
-                       func, tot_pending, fo_tot_pending);
-        if (tot_dirty != fo_tot_dirty)
-                CERROR("%s: tot_dirty "LPU64" != fo_tot_dirty "LPU64"\n",
-                       func, tot_dirty, fo_tot_dirty);
-        if (tot_pending > tot_granted)
-                CERROR("%s: tot_pending "LPU64" > tot_granted "LPU64"\n",
-                       func, tot_pending, tot_granted);
-        if (tot_granted > maxsize)
-                CERROR("%s: tot_granted "LPU64" > maxsize "LPU64"\n",
-                       func, tot_granted, maxsize);
-        if (tot_dirty > maxsize)
-                CERROR("%s: tot_dirty "LPU64" > maxsize "LPU64"\n",
-                       func, tot_dirty, maxsize);
-}
-
-/* Remove this client from the grant accounting totals.  We also remove
- * the export from the obd device under the osfs and dev locks to ensure
- * that the filter_grant_sanity_check() calculations are always valid.
- * The client should do something similar when it invalidates its import. */
-static void filter_grant_discard(struct obd_export *exp)
-{
-        struct obd_device *obd = exp->exp_obd;
-        struct filter_obd *filter = &obd->u.filter;
-        struct filter_export_data *fed = &exp->exp_filter_data;
-
-        cfs_spin_lock(&obd->obd_osfs_lock);
-        LASSERTF(filter->fo_tot_granted >= fed->fed_grant,
-                 "%s: tot_granted "LPU64" cli %s/%p fed_grant %ld\n",
-                 obd->obd_name, filter->fo_tot_granted,
-                 exp->exp_client_uuid.uuid, exp, fed->fed_grant);
-        filter->fo_tot_granted -= fed->fed_grant;
-        LASSERTF(filter->fo_tot_pending >= fed->fed_pending,
-                 "%s: tot_pending "LPU64" cli %s/%p fed_pending %ld\n",
-                 obd->obd_name, filter->fo_tot_pending,
-                 exp->exp_client_uuid.uuid, exp, fed->fed_pending);
-        /* fo_tot_pending is handled in filter_grant_commit as bulk finishes */
-        LASSERTF(filter->fo_tot_dirty >= fed->fed_dirty,
-                 "%s: tot_dirty "LPU64" cli %s/%p fed_dirty %ld\n",
-                 obd->obd_name, filter->fo_tot_dirty,
-                 exp->exp_client_uuid.uuid, exp, fed->fed_dirty);
-        filter->fo_tot_dirty -= fed->fed_dirty;
-        fed->fed_dirty = 0;
-        fed->fed_grant = 0;
-
-        cfs_spin_unlock(&obd->obd_osfs_lock);
-}
-
-static int filter_destroy_export(struct obd_export *exp)
-{
-        struct filter_export_data *fed = &exp->exp_filter_data;
-        ENTRY;
-
-        if (fed->fed_pending)
-                CERROR("%s: cli %s/%p has %lu pending on destroyed export\n",
-                       exp->exp_obd->obd_name, exp->exp_client_uuid.uuid,
-                       exp, fed->fed_pending);
-
-        target_destroy_export(exp);
-
-        if (unlikely(obd_uuid_equals(&exp->exp_obd->obd_uuid,
-                                     &exp->exp_client_uuid)))
-               RETURN(0);
-
-        ldlm_destroy_export(exp);
-        lut_client_free(exp);
-
-        if (!exp->exp_obd->obd_replayable)
-                fsfilt_sync(exp->exp_obd, exp->exp_obd->u.obt.obt_sb);
-
-        filter_grant_discard(exp);
-        filter_fmd_cleanup(exp);
-
-        if (exp->exp_connect_flags & OBD_CONNECT_GRANT_SHRINK) {
-                struct filter_obd *filter = &exp->exp_obd->u.filter;
-                if (filter->fo_tot_granted_clients > 0)
-                        filter->fo_tot_granted_clients --;
-        }
-
-        if (!(exp->exp_flags & OBD_OPT_FORCE))
-                filter_grant_sanity_check(exp->exp_obd, __func__);
-
-        RETURN(0);
-}
-
-static void filter_sync_llogs(struct obd_device *obd, struct obd_export *dexp)
-{
-        struct obd_llog_group *olg_min, *olg;
-        struct filter_obd *filter;
-        int worked = -1, group;
-        struct llog_ctxt *ctxt;
-        ENTRY;
-
-        filter = &obd->u.filter;
-
-        /* we can't sync log holding spinlock. also, we do not want to get
-         * into livelock. so we do following: loop over MDS's exports in
-         * group order and skip already synced llogs -bzzz */
-        do {
-                /* look for group with min. number, but > worked */
-                olg_min = NULL;
-                group = 1 << 30;
-                cfs_spin_lock(&filter->fo_llog_list_lock);
-                cfs_list_for_each_entry(olg, &filter->fo_llog_list, olg_list) {
-                        if (olg->olg_seq <= worked) {
-                                /* this group is already synced */
-                                continue;
-                        }
-                        if (group < olg->olg_seq) {
-                                /* we have group with smaller number to sync */
-                                continue;
-                        }
-                        /* store current minimal group */
-                        olg_min = olg;
-                        group = olg->olg_seq;
-                }
-                cfs_spin_unlock(&filter->fo_llog_list_lock);
-
-                if (olg_min == NULL)
-                        break;
-
-                worked = olg_min->olg_seq;
-                if (olg_min->olg_exp &&
-                    (dexp == olg_min->olg_exp || dexp == NULL)) {
-                        int err;
-                        ctxt = llog_group_get_ctxt(olg_min,
-                                                   LLOG_MDS_OST_REPL_CTXT);
-                        if (ctxt) {
-                               err = llog_sync(ctxt, olg_min->olg_exp, 0);
-                                llog_ctxt_put(ctxt);
-                                if (err) {
-                                        CERROR("error flushing logs to MDS: "
-                                               "rc %d\n", err);
-                                }
-                        }
-                }
-        } while (olg_min != NULL);
-}
-
-/* Also incredibly similar to mds_disconnect */
-static int filter_disconnect(struct obd_export *exp)
-{
-        struct obd_device *obd = exp->exp_obd;
-        int rc;
-        ENTRY;
-
-        LASSERT(exp);
-        class_export_get(exp);
-
-        if (!(exp->exp_flags & OBD_OPT_FORCE))
-                filter_grant_sanity_check(obd, __func__);
-        filter_grant_discard(exp);
-
-        /* Flush any remaining cancel messages out to the target */
-        filter_sync_llogs(obd, exp);
-
-        rc = server_disconnect_export(exp);
-
-        /* Do not erase record for recoverable client. */
-        if (obd->obd_replayable && (!obd->obd_fail || exp->exp_failed))
-                filter_client_del(exp);
-        else
-                fsfilt_sync(obd, obd->u.obt.obt_sb);
-
-        class_export_put(exp);
-        RETURN(rc);
-}
-
-/* reverse import is changed, sync all cancels */
-static void filter_revimp_update(struct obd_export *exp)
-{
-        ENTRY;
-
-        LASSERT(exp);
-        class_export_get(exp);
-
-        /* flush any remaining cancel messages out to the target */
-        filter_sync_llogs(exp->exp_obd, exp);
-        class_export_put(exp);
-        EXIT;
-}
-
-static int filter_ping(const struct lu_env *env, struct obd_export *exp)
-{
-        filter_fmd_expire(exp);
-        return 0;
-}
-
-struct dentry *__filter_oa2dentry(struct obd_device *obd, struct ost_id *ostid,
-                                  const char *what, int quiet)
-{
-        struct dentry *dchild = NULL;
-
-        dchild = filter_fid2dentry(obd, NULL,  ostid->oi_seq, ostid->oi_id);
-
-        if (IS_ERR(dchild)) {
-                CERROR("%s error looking up object: "POSTID"\n",
-                       what, ostid->oi_id, ostid->oi_seq);
-                RETURN(dchild);
-        }
-
-        if (dchild->d_inode == NULL) {
-                if (!quiet)
-                        CERROR("%s: %s on non-existent object: "POSTID" \n",
-                               obd->obd_name, what, ostid->oi_id,ostid->oi_seq);
-                f_dput(dchild);
-                RETURN(ERR_PTR(-ENOENT));
-        }
-
-#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 7, 50, 0)
-        /* Try to correct for a bug in 2.1.0 (LU-221) that caused negative
-         * timestamps to appear to be in the far future, due old timestamp
-         * being stored on disk as an unsigned value.  This fixes up any
-         * bad values stored on disk before returning them to the client,
-         * and ensures any timestamp updates are correct.  LU-1042 */
-        if (unlikely(LTIME_S(dchild->d_inode->i_atime) == LU221_BAD_TIME))
-                LTIME_S(dchild->d_inode->i_atime) = 0;
-        if (unlikely(LTIME_S(dchild->d_inode->i_mtime) == LU221_BAD_TIME))
-                LTIME_S(dchild->d_inode->i_mtime) = 0;
-        if (unlikely(LTIME_S(dchild->d_inode->i_ctime) == LU221_BAD_TIME))
-                LTIME_S(dchild->d_inode->i_ctime) = 0;
-#else
-#warning "remove old LU-221/LU-1042 workaround code"
-#endif
-
-        return dchild;
-}
-
-static int filter_getattr(const struct lu_env *env, struct obd_export *exp,
-                          struct obd_info *oinfo)
-{
-        struct dentry *dentry = NULL;
-        struct obd_device *obd;
-        __u64 curr_version;
-        int rc = 0;
-        ENTRY;
-
-        rc = filter_auth_capa(exp, NULL, oinfo->oi_oa->o_seq,
-                              oinfo_capa(oinfo), CAPA_OPC_META_READ);
-        if (rc)
-                RETURN(rc);
-
-        obd = class_exp2obd(exp);
-        if (obd == NULL) {
-                CDEBUG(D_IOCTL, "invalid client export %p\n", exp);
-                RETURN(-EINVAL);
-        }
-
-        dentry = filter_oa2dentry(obd, &oinfo->oi_oa->o_oi);
-        if (IS_ERR(dentry))
-                RETURN(PTR_ERR(dentry));
-
-        /* Limit the valid bits in the return data to what we actually use */
-        oinfo->oi_oa->o_valid = OBD_MD_FLID;
-        obdo_from_inode(oinfo->oi_oa, dentry->d_inode, FILTER_VALID_FLAGS);
-
-        /* Store inode version in reply */
-        curr_version = fsfilt_get_version(exp->exp_obd, dentry->d_inode);
-        if ((__s64)curr_version != -EOPNOTSUPP) {
-                oinfo->oi_oa->o_valid |= OBD_MD_FLDATAVERSION;
-                oinfo->oi_oa->o_data_version = curr_version;
-        }
-
-        f_dput(dentry);
-        RETURN(rc);
-}
-
-/* this should be enabled/disabled in condition to enabled/disabled large
- * inodes (fast EAs) in backing store FS. */
-int filter_update_fidea(struct obd_export *exp, struct inode *inode,
-                        void *handle, struct obdo *oa)
-{
-        struct obd_device *obd = exp->exp_obd;
-        int rc = 0;
-        ENTRY;
-
-        if (oa->o_valid & OBD_MD_FLFID) {
-                struct filter_fid ff;
-
-                if (!(oa->o_valid & OBD_MD_FLGROUP))
-                        oa->o_seq = 0;
-                /* packing fid and converting it to LE for storing into EA.
-                 * Here ->o_stripe_idx should be filled by LOV and rest of
-                 * fields - by client. */
-                ff.ff_parent.f_seq = cpu_to_le64(oa->o_parent_seq);
-                ff.ff_parent.f_oid = cpu_to_le32(oa->o_parent_oid);
-                /* XXX: we are ignoring o_parent_ver here, since this should
-                 *      be the same for all objects in this fileset. */
-                ff.ff_parent.f_ver = cpu_to_le32(oa->o_stripe_idx);
-                ff.ff_objid = cpu_to_le64(oa->o_id);
-                ff.ff_seq = cpu_to_le64(oa->o_seq);
-
-                CDEBUG(D_INODE, "storing filter fid EA (parent "DFID" "
-                       LPU64"/"LPU64")\n", PFID(&ff.ff_parent), oa->o_id,
-                       oa->o_seq);
-
-                rc = fsfilt_set_md(obd, inode, handle, &ff, sizeof(ff), "fid");
-                if (rc)
-                        CERROR("store fid in object failed! rc: %d\n", rc);
-        } else {
-                CDEBUG(D_HA, "OSS object without fid info!\n");
-        }
-
-        RETURN(rc);
-}
-
-/* this is called from filter_truncate() until we have filter_punch() */
-int filter_setattr_internal(struct obd_export *exp, struct dentry *dentry,
-                            struct obdo *oa, struct obd_trans_info *oti)
-{
-        struct llog_cookie *fcc = NULL;
-        struct filter_obd *filter;
-        int rc, err, sync = 0;
-        loff_t old_size = 0;
-        unsigned int ia_valid;
-        struct inode *inode;
-        struct page *page = NULL;
-        struct iattr iattr;
-        void *handle;
-        ENTRY;
-
-        LASSERT(dentry != NULL);
-        LASSERT(!IS_ERR(dentry));
-
-        inode = dentry->d_inode;
-        LASSERT(inode != NULL);
-
-        filter = &exp->exp_obd->u.filter;
-        iattr_from_obdo(&iattr, oa, oa->o_valid);
-        ia_valid = iattr.ia_valid;
-
-        if (oa->o_valid & OBD_MD_FLCOOKIE) {
-                OBD_ALLOC(fcc, sizeof(*fcc));
-                if (fcc != NULL)
-                        *fcc = oa->o_lcookie;
-        }
-       if (ia_valid & (ATTR_SIZE | ATTR_UID | ATTR_GID)) {
-               unsigned long now = jiffies;
-               /* Filter truncates and writes are serialized by
-                * i_alloc_sem, see the comment in
-                * filter_preprw_write.*/
-               if (ia_valid & ATTR_SIZE)
-                       down_write(&inode->i_alloc_sem);
-               mutex_lock(&inode->i_mutex);
-               fsfilt_check_slow(exp->exp_obd, now, "i_alloc_sem and i_mutex");
-               old_size = i_size_read(inode);
-       }
-
-        /* VBR: version recovery check */
-        rc = filter_version_get_check(exp, oti, inode);
-        if (rc)
-                GOTO(out_unlock, rc);
-
-        /* Let's pin the last page so that ldiskfs_truncate
-         * should not start GFP_FS allocation. */
-        if (ia_valid & ATTR_SIZE) {
-                page = grab_cache_page(inode->i_mapping,
-                                       iattr.ia_size >> PAGE_CACHE_SHIFT);
-                if (page == NULL)
-                        GOTO(out_unlock, rc = -ENOMEM);
-
-                unlock_page(page);
-        }
-
-        /* If the inode still has SUID+SGID bits set (see filter_precreate())
-         * then we will accept the UID+GID sent by the client during write for
-         * initializing the ownership of this inode.  We only allow this to
-         * happen once so clear these bits in setattr. In 2.6 kernels it is
-         * possible to get ATTR_UID and ATTR_GID separately, so we only clear
-         * the flags that are actually being set. */
-        if (ia_valid & (ATTR_UID | ATTR_GID)) {
-                CDEBUG(D_INODE, "update UID/GID to %lu/%lu\n",
-                       (unsigned long)oa->o_uid, (unsigned long)oa->o_gid);
-
-                if ((inode->i_mode & S_ISUID) && (ia_valid & ATTR_UID)) {
-                        if (!(ia_valid & ATTR_MODE)) {
-                                iattr.ia_mode = inode->i_mode;
-                                iattr.ia_valid |= ATTR_MODE;
-                        }
-                        iattr.ia_mode &= ~S_ISUID;
-                }
-                if ((inode->i_mode & S_ISGID) && (ia_valid & ATTR_GID)) {
-                        if (!(iattr.ia_valid & ATTR_MODE)) {
-                                iattr.ia_mode = inode->i_mode;
-                                iattr.ia_valid |= ATTR_MODE;
-                        }
-                        iattr.ia_mode &= ~S_ISGID;
-                }
-
-                handle = fsfilt_start_log(exp->exp_obd, inode,
-                                          FSFILT_OP_SETATTR, oti, 1);
-                if (IS_ERR(handle))
-                        GOTO(out_unlock, rc = PTR_ERR(handle));
-
-                /* update inode EA only once when inode is suid bit marked. As
-                 * on 2.6.x UID and GID may be set separately, we check here
-                 * only one of them to avoid double setting. */
-                if (inode->i_mode & S_ISUID)
-                        filter_update_fidea(exp, inode, handle, oa);
-        } else {
-                handle = fsfilt_start(exp->exp_obd, inode,
-                                      FSFILT_OP_SETATTR, oti);
-                if (IS_ERR(handle))
-                        GOTO(out_unlock, rc = PTR_ERR(handle));
-        }
-
-        /* Locking order: i_mutex -> journal_lock -> dqptr_sem. LU-952 */
-        if (ia_valid & (ATTR_SIZE | ATTR_UID | ATTR_GID))
-                ll_vfs_dq_init(inode);
-
-        if (oa->o_valid & OBD_MD_FLFLAGS) {
-                rc = fsfilt_iocontrol(exp->exp_obd, dentry,
-                                      FSFILT_IOC_SETFLAGS, (long)&oa->o_flags);
-        } else {
-                rc = fsfilt_setattr(exp->exp_obd, dentry, handle, &iattr, 1);
-                if (fcc != NULL)
-                        /* set cancel cookie callback function */
-                        sync = fsfilt_add_journal_cb(exp->exp_obd, 0, handle,
-                                                     filter_cancel_cookies_cb,
-                                                     fcc);
-        }
-
-        if (OBD_FAIL_CHECK(OBD_FAIL_OST_SETATTR_CREDITS))
-                fsfilt_extend(exp->exp_obd, inode, 0, handle);
-
-       /* The truncate might have used up our transaction credits.  Make sure
-        * we have two left for the last_rcvd and VBR inode version updates. */
-        err = fsfilt_extend(exp->exp_obd, inode, 2, handle);
-
-        /* Update inode version only if data has changed => size has changed */
-        rc = filter_finish_transno(exp, ia_valid & ATTR_SIZE ? inode : NULL,
-                                   oti, rc, sync);
-
-        if (sync) {
-                filter_cancel_cookies_cb(exp->exp_obd, 0, fcc, rc);
-                fcc = NULL;
-        }
-
-        err = fsfilt_commit(exp->exp_obd, inode, handle, 0);
-        if (err) {
-                CERROR("error on commit, err = %d\n", err);
-                if (!rc)
-                        rc = err;
-        } else {
-                fcc = NULL;
-        }
-
-        /* For a partial-page truncate flush the page to disk immediately
-         * to avoid data corruption during direct disk write. b=17397 */
-        if (!sync && (iattr.ia_valid & ATTR_SIZE) &&
-            old_size != iattr.ia_size && (iattr.ia_size & ~CFS_PAGE_MASK)) {
-                err = filemap_fdatawrite_range(inode->i_mapping, iattr.ia_size,
-                                               iattr.ia_size + 1);
-                if (!rc)
-                        rc = err;
-        }
-
-        EXIT;
-
-out_unlock:
-        if (page)
-                page_cache_release(page);
-
-       if (ia_valid & (ATTR_SIZE | ATTR_UID | ATTR_GID))
-               mutex_unlock(&inode->i_mutex);
-       if (ia_valid & ATTR_SIZE)
-               up_write(&inode->i_alloc_sem);
-       if (fcc)
-               OBD_FREE(fcc, sizeof(*fcc));
-        return rc;
-}
-
-/* this is called from filter_truncate() until we have filter_punch() */
-int filter_setattr(const struct lu_env *env, struct obd_export *exp,
-                   struct obd_info *oinfo, struct obd_trans_info *oti)
-{
-        struct obdo *oa = oinfo->oi_oa;
-        struct lustre_capa *capa = oinfo_capa(oinfo);
-        struct ldlm_res_id res_id;
-        struct filter_mod_data *fmd;
-        struct lvfs_run_ctxt saved;
-        struct filter_obd *filter;
-        struct ldlm_resource *res;
-        struct dentry *dentry;
-        __u64 opc = CAPA_OPC_META_WRITE;
-        int rc;
-        ENTRY;
-
-        if (oinfo->oi_flags & OBD_FL_PUNCH)
-                opc |= CAPA_OPC_OSS_TRUNC;
-
-        rc = filter_auth_capa(exp, NULL, oa->o_seq, capa, opc);
-        if (rc)
-                RETURN(rc);
-
-        if (oa->o_valid & (OBD_MD_FLUID | OBD_MD_FLGID)) {
-                rc = filter_capa_fixoa(exp, oa, oa->o_seq, capa);
-                if (rc)
-                        RETURN(rc);
-        }
-
-        osc_build_res_name(oa->o_id, oa->o_seq, &res_id);
-        /* This would be very bad - accidentally truncating a file when
-         * changing the time or similar - bug 12203. */
-        if (oa->o_valid & OBD_MD_FLSIZE &&
-            oinfo->oi_policy.l_extent.end != OBD_OBJECT_EOF) {
-                static char mdsinum[48];
-
-                if (oa->o_valid & OBD_MD_FLFID)
-                        snprintf(mdsinum, sizeof(mdsinum) - 1, " of inode "DFID,
-                                 oa->o_parent_seq, oa->o_parent_oid,
-                                 oa->o_parent_ver);
-                else
-                        mdsinum[0] = '\0';
-
-                CERROR("%s: setattr from %s trying to truncate objid "POSTID
-                       "%s\n", exp->exp_obd->obd_name, obd_export_nid2str(exp),
-                       oa->o_id, oa->o_seq, mdsinum);
-                RETURN(-EPERM);
-        }
-
-        dentry = __filter_oa2dentry(exp->exp_obd, &oinfo->oi_oa->o_oi, __func__, 1);
-        if (IS_ERR(dentry))
-                RETURN(PTR_ERR(dentry));
-
-        filter = &exp->exp_obd->u.filter;
-        push_ctxt(&saved, &exp->exp_obd->obd_lvfs_ctxt, NULL);
-
-        /*
-         * We need to be atomic against a concurrent write
-         * (which takes the semaphore for reading). fmd_mactime_xid
-         * checks will have no effect if a write request with lower
-         * xid starts just before a setattr and finishes later than
-         * the setattr (see bug 21489, comment 27).
-         */
-        if (oa->o_valid &
-            (OBD_MD_FLMTIME | OBD_MD_FLATIME | OBD_MD_FLCTIME)) {
-                unsigned long now = jiffies;
-                down_write(&dentry->d_inode->i_alloc_sem);
-                fsfilt_check_slow(exp->exp_obd, now, "i_alloc_sem");
-                fmd = filter_fmd_get(exp, oa->o_id, oa->o_seq);
-                if (fmd && fmd->fmd_mactime_xid < oti->oti_xid)
-                        fmd->fmd_mactime_xid = oti->oti_xid;
-                filter_fmd_put(exp, fmd);
-                up_write(&dentry->d_inode->i_alloc_sem);
-        }
-
-        /* setting objects attributes (including owner/group) */
-        rc = filter_setattr_internal(exp, dentry, oa, oti);
-        if (rc)
-                GOTO(out_unlock, rc);
-
-        res = ldlm_resource_get(exp->exp_obd->obd_namespace, NULL,
-                                &res_id, LDLM_EXTENT, 0);
-
-        if (res != NULL) {
-                LDLM_RESOURCE_ADDREF(res);
-                rc = ldlm_res_lvbo_update(res, NULL, 0);
-                LDLM_RESOURCE_DELREF(res);
-                ldlm_resource_putref(res);
-        }
-
-        oa->o_valid = OBD_MD_FLID;
-
-        /* Quota release need uid/gid info */
-        obdo_from_inode(oa, dentry->d_inode,
-                        FILTER_VALID_FLAGS | OBD_MD_FLUID | OBD_MD_FLGID);
-
-       filter_counter_incr(exp, LPROC_FILTER_STATS_SETATTR,
-                           oti ? oti->oti_jobid : NULL, 1);
-        EXIT;
-out_unlock:
-        f_dput(dentry);
-        pop_ctxt(&saved, &exp->exp_obd->obd_lvfs_ctxt, NULL);
-        return rc;
-}
-
-/* XXX identical to osc_unpackmd */
-static int filter_unpackmd(struct obd_export *exp, struct lov_stripe_md **lsmp,
-                           struct lov_mds_md *lmm, int lmm_bytes)
-{
-        int lsm_size;
-        ENTRY;
-
-        if (lmm != NULL) {
-                if (lmm_bytes < sizeof (*lmm)) {
-                        CERROR("lov_mds_md too small: %d, need %d\n",
-                               lmm_bytes, (int)sizeof(*lmm));
-                        RETURN(-EINVAL);
-                }
-                /* XXX LOV_MAGIC etc check? */
-
-                if (lmm->lmm_object_id == cpu_to_le64(0)) {
-                        CERROR("lov_mds_md: zero lmm_object_id\n");
-                        RETURN(-EINVAL);
-                }
-        }
-
-        lsm_size = lov_stripe_md_size(1);
-        if (lsmp == NULL)
-                RETURN(lsm_size);
-
-        if (*lsmp != NULL && lmm == NULL) {
-                OBD_FREE((*lsmp)->lsm_oinfo[0], sizeof(struct lov_oinfo));
-                OBD_FREE(*lsmp, lsm_size);
-                *lsmp = NULL;
-                RETURN(0);
-        }
-
-        if (*lsmp == NULL) {
-                OBD_ALLOC(*lsmp, lsm_size);
-                if (*lsmp == NULL)
-                        RETURN(-ENOMEM);
-
-                OBD_ALLOC((*lsmp)->lsm_oinfo[0], sizeof(struct lov_oinfo));
-                if ((*lsmp)->lsm_oinfo[0] == NULL) {
-                        OBD_FREE(*lsmp, lsm_size);
-                        RETURN(-ENOMEM);
-                }
-                loi_init((*lsmp)->lsm_oinfo[0]);
-        }
-
-        if (lmm != NULL) {
-                /* XXX zero *lsmp? */
-                (*lsmp)->lsm_object_id = le64_to_cpu (lmm->lmm_object_id);
-                LASSERT((*lsmp)->lsm_object_id);
-        }
-
-        (*lsmp)->lsm_maxbytes = exp->exp_obd->u.obt.obt_sb->s_maxbytes;
-
-        RETURN(lsm_size);
-}
-
-/* caller must hold fo_create_locks[oa->o_seq] */
-static int filter_destroy_precreated(struct obd_export *exp, struct obdo *oa,
-                                     struct filter_obd *filter)
-{
-        struct obdo doa = { 0 }; /* XXX obdo on stack */
-        obd_id last, id;
-        int rc = 0;
-        int skip_orphan;
-        ENTRY;
-
-        LASSERT_MUTEX_LOCKED(&filter->fo_create_locks[oa->o_seq]);
-
-        memset(&doa, 0, sizeof(doa));
-
-        doa.o_valid |= OBD_MD_FLGROUP;
-        doa.o_seq = oa->o_seq;
-        doa.o_mode = S_IFREG;
-
-        if (!cfs_test_bit(doa.o_seq, &filter->fo_destroys_in_progress)) {
-                CERROR("%s:["LPU64"] destroys_in_progress already cleared\n",
-                       exp->exp_obd->obd_name, doa.o_seq);
-                RETURN(0);
-        }
-
-        last = filter_last_id(filter, doa.o_seq);
-
-        skip_orphan = !!(exp->exp_connect_flags & OBD_CONNECT_SKIP_ORPHAN);
-
-        CDEBUG(D_HA, "%s: deleting orphan objects from "LPU64" to "LPU64"%s\n",
-               exp->exp_obd->obd_name, oa->o_id + 1, last,
-               skip_orphan ? ", orphan objids won't be reused any more." : ".");
-
-        for (id = last; id > oa->o_id; id--) {
-                doa.o_id = id;
-                rc = filter_destroy(NULL, exp, &doa, NULL, NULL, NULL, NULL);
-                if (rc && rc != -ENOENT) /* this is pretty fatal... */
-                        CEMERG("error destroying precreate objid "LPU64": %d\n",
-                               id, rc);
-
-                /* update last_id on disk periodically so that if we restart
-                 * we don't need to re-scan all of the just-deleted objects. */
-                if ((id & 511) == 0 && !skip_orphan) {
-                        filter_set_last_id(filter, id - 1, doa.o_seq);
-                        filter_update_last_objid(exp->exp_obd, doa.o_seq, 0);
-                }
-        }
-
-        CDEBUG(D_HA, "%s: after destroy: set last_objids["LPU64"] = "LPU64"\n",
-               exp->exp_obd->obd_name, doa.o_seq, oa->o_id);
-
-        if (!skip_orphan) {
-                filter_set_last_id(filter, id, doa.o_seq);
-                rc = filter_update_last_objid(exp->exp_obd, doa.o_seq, 1);
-        } else {
-                /*
-                 * We have destroyed orphan objects, but don't want to reuse
-                 * them. Therefore we don't reset last_id to the last created
-                 * objects. Instead, we report back to the MDS the object id
-                 * of the last orphan, so that the MDS can restart allocating
-                 * objects from this id + 1 and thus skip the whole orphan
-                 * object id range
-                 */
-                oa->o_id = last;
-                rc = 0;
-        }
-        cfs_clear_bit(doa.o_seq, &filter->fo_destroys_in_progress);
-
-        RETURN(rc);
-}
-
-static int filter_precreate(struct obd_device *obd, struct obdo *oa,
-                            obd_seq group, int *num);
-/* returns a negative error or a nonnegative number of files to create */
-static int filter_handle_precreate(struct obd_export *exp, struct obdo *oa,
-                                   obd_seq group, struct obd_trans_info *oti)
-{
-        struct obd_device *obd = exp->exp_obd;
-        struct filter_obd *filter = &obd->u.filter;
-        int diff, rc;
-        ENTRY;
-
-        /* delete orphans request */
-        if ((oa->o_valid & OBD_MD_FLFLAGS) && (oa->o_flags & OBD_FL_DELORPHAN)){
-                obd_id last = filter_last_id(filter, group);
-
-                if (oti->oti_conn_cnt < exp->exp_conn_cnt) {
-                        CERROR("%s: dropping old orphan cleanup request\n",
-                               obd->obd_name);
-                        RETURN(0);
-                }
-                /* This causes inflight precreates to abort and drop lock */
-                cfs_set_bit(group, &filter->fo_destroys_in_progress);
-                cfs_mutex_lock(&filter->fo_create_locks[group]);
-                if (!cfs_test_bit(group, &filter->fo_destroys_in_progress)) {
-                        CERROR("%s:["LPU64"] destroys_in_progress already cleared\n",
-                               exp->exp_obd->obd_name, group);
-                        cfs_mutex_unlock(&filter->fo_create_locks[group]);
-                        RETURN(0);
-                }
-                diff = oa->o_id - last;
-                CDEBUG(D_HA, "filter_last_id() = "LPU64" -> diff = %d\n",
-                       last, diff);
-
-                if (-diff > (OST_MAX_PRECREATE * 3) / 2) {
-                        CERROR("%s: ignoring bogus orphan destroy request: "
-                               "obdid "LPU64" last_id "LPU64"\n", obd->obd_name,
-                               oa->o_id, last);
-                        /* FIXME: should reset precreate_next_id on MDS */
-                        GOTO(out, rc = -EINVAL);
-                }
-                if (diff < 0) {
-                        rc = filter_destroy_precreated(exp, oa, filter);
-                        if (rc)
-                                CERROR("%s: unable to write lastobjid, but "
-                                       "orphans were deleted\n", obd->obd_name);
-                        GOTO(out, rc);
-                } else {
-                        /* XXX: Used by MDS for the first time! */
-                        cfs_clear_bit(group, &filter->fo_destroys_in_progress);
-                }
-        } else {
-                cfs_mutex_lock(&filter->fo_create_locks[group]);
-                if (oti->oti_conn_cnt < exp->exp_conn_cnt) {
-                        CERROR("%s: dropping old precreate request\n",
-                               obd->obd_name);
-                        GOTO(out, rc = 0);
-                }
-                /* only precreate if group == 0 and o_id is specfied */
-                if (!fid_seq_is_mdt(group) || oa->o_id == 0)
-                        diff = 1;
-                else
-                        diff = oa->o_id - filter_last_id(filter, group);
-                CDEBUG(D_RPCTRACE, "filter_last_id() = "LPU64" -> diff = %d\n",
-                       filter_last_id(filter, group), diff);
-
-               /*
-                * Check obd->obd_recovering to handle the race condition
-                * while recreating missing precreated objects through
-                * filter_preprw_write() and mds_lov_clear_orphans()
-                * at the same time.
-                */
-               LASSERTF(ergo(!obd->obd_recovering, diff >= 0),
-                        "%s: "LPU64" - "LPU64" = %d\n", obd->obd_name,
-                        oa->o_id, filter_last_id(filter, group), diff);
-        }
-
-        if (diff > 0) {
-                oa->o_id = filter_last_id(&obd->u.filter, group);
-                rc = filter_precreate(obd, oa, group, &diff);
-                oa->o_id = filter_last_id(&obd->u.filter, group);
-                oa->o_seq = group;
-                oa->o_valid |= (OBD_MD_FLID | OBD_MD_FLGROUP);
-                GOTO(out, rc);
-        }
-        /* else diff == 0 */
-        GOTO(out, rc = 0);
-out:
-        cfs_mutex_unlock(&filter->fo_create_locks[group]);
-        return rc;
-}
-
-static int filter_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 filter_obd *filter = &obd->u.filter;
-        int blockbits = obd->u.obt.obt_sb->s_blocksize_bits;
-        struct lr_server_data *lsd = class_server_data(obd);
-        int rc;
-        ENTRY;
-
-        /* at least try to account for cached pages.  its still racey and
-         * might be under-reporting if clients haven't announced their
-         * caches with brw recently */
-        cfs_spin_lock(&obd->obd_osfs_lock);
-        rc = fsfilt_statfs(obd, obd->u.obt.obt_sb, max_age);
-        memcpy(osfs, &obd->obd_osfs, sizeof(*osfs));
-        cfs_spin_unlock(&obd->obd_osfs_lock);
-
-        CDEBUG(D_SUPER | D_CACHE, "blocks cached "LPU64" granted "LPU64
-               " pending "LPU64" free "LPU64" avail "LPU64"\n",
-               filter->fo_tot_dirty, filter->fo_tot_granted,
-               filter->fo_tot_pending,
-               osfs->os_bfree << blockbits, osfs->os_bavail << blockbits);
-
-        filter_grant_sanity_check(obd, __func__);
-
-        osfs->os_bavail -= min(osfs->os_bavail, GRANT_FOR_LLOG(obd) +
-                               ((filter->fo_tot_dirty + filter->fo_tot_pending +
-                                 osfs->os_bsize - 1) >> blockbits));
-
-        if (OBD_FAIL_CHECK_VALUE(OBD_FAIL_OST_ENOSPC,
-                                 le32_to_cpu(lsd->lsd_ost_index)))
-                osfs->os_bfree = osfs->os_bavail = 2;
-
-        if (OBD_FAIL_CHECK_VALUE(OBD_FAIL_OST_ENOINO,
-                                 le32_to_cpu(lsd->lsd_ost_index)))
-                osfs->os_ffree = 0;
-
-        /* set EROFS to state field if FS is mounted as RDONLY. The goal is to
-         * stop creating files on MDS if OST is not good shape to create
-         * objects.*/
-        osfs->os_state = 0;
-
-        if (filter->fo_obt.obt_sb->s_flags & MS_RDONLY)
-                osfs->os_state = OS_STATE_READONLY;
-
-        if (filter->fo_raid_degraded)
-                osfs->os_state |= OS_STATE_DEGRADED;
-        RETURN(rc);
-}
-
-static int filter_use_existing_obj(struct obd_device *obd,
-                                   struct dentry *dchild, void **handle,
-                                   int *cleanup_phase)
-{
-        struct inode *inode = dchild->d_inode;
-        struct iattr iattr;
-        int rc;
-
-        if ((inode->i_mode & (S_ISUID | S_ISGID)) == (S_ISUID|S_ISGID))
-                return 0;
-
-        *handle = fsfilt_start_log(obd, inode, FSFILT_OP_SETATTR, NULL, 1);
-        if (IS_ERR(*handle))
-                return PTR_ERR(*handle);
-
-        iattr.ia_valid = ATTR_MODE;
-        iattr.ia_mode = S_ISUID | S_ISGID |0666;
-        rc = fsfilt_setattr(obd, dchild, *handle, &iattr, 1);
-        if (rc == 0)
-                *cleanup_phase = 3;
-
-        return rc;
-}
-
-static __u64 filter_calc_free_inodes(struct obd_device *obd)
-{
-        int rc;
-        __u64 os_ffree = -1;
-
-        cfs_spin_lock(&obd->obd_osfs_lock);
-        rc = fsfilt_statfs(obd, obd->u.obt.obt_sb, cfs_time_shift_64(1));
-        if (rc == 0)
-                os_ffree = obd->obd_osfs.os_ffree;
-        cfs_spin_unlock(&obd->obd_osfs_lock);
-
-        return os_ffree;
-}
-
-/* We rely on the fact that only one thread will be creating files in a given
- * group at a time, which is why we don't need an atomic filter_get_new_id.
- * Even if we had that atomic function, the following race would exist:
- *
- * thread 1: gets id x from filter_next_id
- * thread 2: gets id (x + 1) from filter_next_id
- * thread 2: creates object (x + 1)
- * thread 1: tries to create object x, gets -ENOSPC
- *
- * Caller must hold fo_create_locks[group]
- */
-static int filter_precreate(struct obd_device *obd, struct obdo *oa,
-                            obd_seq group, int *num)
-{
-        struct dentry *dchild = NULL, *dparent = NULL;
-        struct filter_obd *filter;
-        struct obd_statfs *osfs;
-        struct iattr iattr;
-        int err = 0, rc = 0, recreate_obj = 0, i;
-        cfs_time_t enough_time = cfs_time_shift(DISK_TIMEOUT/2);
-        __u64 os_ffree;
-        obd_id next_id;
-        void *handle = NULL;
-        ENTRY;
-
-        filter = &obd->u.filter;
-
-        LASSERT_MUTEX_LOCKED(&filter->fo_create_locks[group]);
-
-        OBD_FAIL_TIMEOUT(OBD_FAIL_TGT_DELAY_PRECREATE, obd_timeout / 2);
-
-        if ((oa->o_valid & OBD_MD_FLFLAGS) &&
-            (oa->o_flags & OBD_FL_RECREATE_OBJS)) {
-                recreate_obj = 1;
-        } else {
-                OBD_ALLOC(osfs, sizeof(*osfs));
-                if (osfs == NULL)
-                        RETURN(-ENOMEM);
-                rc = filter_statfs(NULL, obd->obd_self_export, osfs,
-                                   cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS),
-                                   0);
-                if (rc == 0 && osfs->os_bavail < (osfs->os_blocks >> 10)) {
-                        CDEBUG(D_RPCTRACE,"%s: not enough space for create "
-                               LPU64"\n", obd->obd_name, osfs->os_bavail <<
-                               obd->u.obt.obt_vfsmnt->mnt_sb->s_blocksize_bits);
-                        *num = 0;
-                        if (oa->o_valid & OBD_MD_FLFLAGS)
-                                oa->o_flags |= OBD_FL_NOSPC_BLK;
-                        else {
-                                oa->o_valid |= OBD_MD_FLFLAGS;
-                                oa->o_flags = OBD_FL_NOSPC_BLK;
-                        }
-
-                        rc = -ENOSPC;
-                }
-                OBD_FREE(osfs, sizeof(*osfs));
-                if (rc)
-                        RETURN(rc);
-        }
-
-        CDEBUG(D_RPCTRACE, "%s: precreating %d objects in group "LPU64
-               " at "LPU64"\n", obd->obd_name, *num, group, oa->o_id);
-
-        for (i = 0; i < *num && err == 0; i++) {
-                int cleanup_phase = 0;
-
-                if (cfs_test_bit(group, &filter->fo_destroys_in_progress)) {
-                        CWARN("%s: create aborted by destroy\n",
-                              obd->obd_name);
-                        rc = -EAGAIN;
-                        break;
-                }
-
-                if (recreate_obj) {
-                        __u64 last_id;
-                        next_id = oa->o_id;
-                        last_id = filter_last_id(filter, group);
-                        if (next_id > last_id) {
-                                CERROR("Error: Trying to recreate obj greater"
-                                       "than last id "LPD64" > "LPD64"\n",
-                                       next_id, last_id);
-                                GOTO(cleanup, rc = -EINVAL);
-                        }
-                } else
-                        next_id = filter_last_id(filter, group) + 1;
-
-                /* Don't create objects beyond the valid range for this SEQ */
-                if (unlikely(fid_seq_is_mdt0(group) &&
-                            next_id >= IDIF_MAX_OID)) {
-                        CERROR("%s:"POSTID" hit the IDIF_MAX_OID (1<<48)!\n",
-                                obd->obd_name, next_id, group);
-                        GOTO(cleanup, rc = -ENOSPC);
-               } else if (unlikely(!fid_seq_is_mdt0(group) &&
-                                   next_id >= OBIF_MAX_OID)) {
-                        CERROR("%s:"POSTID" hit the OBIF_MAX_OID (1<<32)!\n",
-                                obd->obd_name, next_id, group);
-                        GOTO(cleanup, rc = -ENOSPC);
-                }
-
-                dparent = filter_parent_lock(obd, group, next_id);
-                if (IS_ERR(dparent))
-                        GOTO(cleanup, rc = PTR_ERR(dparent));
-                cleanup_phase = 1; /* filter_parent_unlock(dparent) */
-
-                dchild = filter_fid2dentry(obd, dparent, group, next_id);
-                if (IS_ERR(dchild))
-                        GOTO(cleanup, rc = PTR_ERR(dchild));
-                cleanup_phase = 2;  /* f_dput(dchild) */
-
-                if (dchild->d_inode != NULL) {
-                        /* This would only happen if lastobjid was bad on disk*/
-                        /* Could also happen if recreating missing obj but it
-                         * already exists. */
-                        if (recreate_obj) {
-                                CERROR("%s: recreating existing object %.*s?\n",
-                                       obd->obd_name, dchild->d_name.len,
-                                       dchild->d_name.name);
-                        } else {
-                                /* Use these existing objects if they are
-                                 * zero length. */
-                                if (i_size_read(dchild->d_inode) == 0) {
-                                        rc = filter_use_existing_obj(obd,dchild,
-                                                      &handle, &cleanup_phase);
-                                        if (rc == 0)
-                                                goto set_last_id;
-                                        else
-                                                GOTO(cleanup, rc);
-                                }
-
-                                CERROR("%s: Serious error: objid %.*s already "
-                                       "exists; is this filesystem corrupt?\n",
-                                       obd->obd_name, dchild->d_name.len,
-                                       dchild->d_name.name);
-                                LBUG();
-                        }
-                        GOTO(cleanup, rc = -EEXIST);
-                }
-
-                handle = fsfilt_start_log(obd, dparent->d_inode,
-                                          FSFILT_OP_CREATE, NULL, 1);
-                if (IS_ERR(handle))
-                        GOTO(cleanup, rc = PTR_ERR(handle));
-                cleanup_phase = 3;
-
-                CDEBUG(D_INODE, "%s: filter_precreate(od->o_seq="LPU64
-                       ",od->o_id="LPU64")\n", obd->obd_name, group,
-                       next_id);
-
-                /* We mark object SUID+SGID to flag it for accepting UID+GID
-                 * from client on first write.  Currently the permission bits
-                 * on the OST are never used, so this is OK. */
-                rc = ll_vfs_create(dparent->d_inode, dchild,
-                                   S_IFREG |  S_ISUID | S_ISGID | 0666, NULL);
-                if (rc) {
-                        CWARN("%s: create failed: rc = %d\n", obd->obd_name,rc);
-                        if (rc == -ENOSPC) {
-                                os_ffree = filter_calc_free_inodes(obd);
-                                if (os_ffree == -1)
-                                        GOTO(cleanup, rc);
-
-                                if (obd->obd_osfs.os_bavail <
-                                    (obd->obd_osfs.os_blocks >> 10)) {
-                                        if (oa->o_valid & OBD_MD_FLFLAGS) {
-                                                oa->o_flags |= OBD_FL_NOSPC_BLK;
-                                        } else {
-                                                oa->o_valid |= OBD_MD_FLFLAGS;
-                                                oa->o_flags = OBD_FL_NOSPC_BLK;
-                                        }
-
-                                        CWARN("%s: free inode "LPU64"\n",
-                                              obd->obd_name, os_ffree);
-                                }
-                        }
-                        GOTO(cleanup, rc);
-                }
-
-                if (dchild->d_inode)
-                        CDEBUG(D_INFO, "objid "LPU64" got inum %lu\n", next_id,
-                                       dchild->d_inode->i_ino);
-
-set_last_id:
-                /* Initialize a/c/m time so any client timestamp will always
-                 * be newer and update the inode. ctime = 0 is also handled
-                 * specially in fsfilt_ext3_setattr(). See LU-221, LU-1042 */
-                iattr.ia_valid = ATTR_ATIME | ATTR_MTIME | ATTR_CTIME;
-                LTIME_S(iattr.ia_atime) = 0;
-                LTIME_S(iattr.ia_mtime) = 0;
-                LTIME_S(iattr.ia_ctime) = 0;
-                err = fsfilt_setattr(obd, dchild, handle, &iattr, 0);
-                 if (err)
-                        CWARN("%s: unable to initialize a/c/m time of newly "
-                              "created object %.*s: rc = %d\n",
-                              obd->obd_name, dchild->d_name.len,
-                              dchild->d_name.name, err);
-
-                if (!recreate_obj) {
-                        filter_set_last_id(filter, next_id, group);
-                        err = filter_update_last_objid(obd, group, 0);
-                        if (err)
-                                CERROR("%s: unable to write lastobjid "
-                                       "but file created: rc = %d\n",
-                                       obd->obd_name, err);
-                }
-
-        cleanup:
-                switch(cleanup_phase) {
-                case 3:
-                        err = fsfilt_commit(obd, dparent->d_inode, handle, 0);
-                        if (err) {
-                                CERROR("error on commit, err = %d\n", err);
-                                if (!rc)
-                                        rc = err;
-                        }
-                case 2:
-                        f_dput(dchild);
-                case 1:
-                        filter_parent_unlock(dparent);
-                case 0:
-                        break;
-                }
-
-                if (rc)
-                        break;
-                if (cfs_time_after(jiffies, enough_time)) {
-                        i++;
-                        CDEBUG(D_RPCTRACE,
-                               "%s: precreate slow - want %d got %d \n",
-                               obd->obd_name, *num, i);
-                        break;
-                }
-        }
-        *num = i;
-
-        CDEBUG(D_RPCTRACE,
-               "%s: created %d objects for group "POSTID" rc %d\n",
-               obd->obd_name, i, filter->fo_last_objids[group], group, rc);
-
-        RETURN(rc);
-}
-
-int filter_create(const struct lu_env *env, struct obd_export *exp,
-                  struct obdo *oa, struct lov_stripe_md **ea,
-                  struct obd_trans_info *oti)
-{
-        struct obd_device *obd = exp->exp_obd;
-        struct filter_export_data *fed;
-        struct filter_obd *filter;
-        struct lvfs_run_ctxt saved;
-        struct lov_stripe_md *lsm = NULL;
-        int rc = 0, diff;
-        ENTRY;
-
-        CDEBUG(D_INODE, "%s: filter_create(group="LPU64",id="
-               LPU64")\n", obd->obd_name, oa->o_seq, oa->o_id);
-
-        fed = &exp->exp_filter_data;
-        filter = &obd->u.filter;
-
-        /* 1.8 client doesn't carry the ocd_group with connect request,
-         * so the fed_group will always be zero for 1.8 client. */
-        if (!(exp->exp_connect_flags & OBD_CONNECT_FULL20)) {
-                if (oa->o_seq != FID_SEQ_OST_MDT0 &&
-                    oa->o_seq != FID_SEQ_LLOG &&
-                    oa->o_seq != FID_SEQ_ECHO) {
-                        CERROR("The request from older client has invalid"
-                               " group "LPU64"!\n", oa->o_seq);
-                        RETURN(-EINVAL);
-                }
-        } else if (fed->fed_group != oa->o_seq) {
-                CERROR("%s: this export (nid %s) used object group %d "
-                        "earlier; now it's trying to use group "LPU64"!"
-                        " This could be a bug in the MDS. Please report to "
-                        "http://bugzilla.lustre.org/\n", obd->obd_name,
-                        obd_export_nid2str(exp), fed->fed_group, oa->o_seq);
-                RETURN(-ENOTUNIQ);
-        }
-
-        if (ea != NULL) {
-                lsm = *ea;
-                if (lsm == NULL) {
-                        rc = obd_alloc_memmd(exp, &lsm);
-                        if (rc < 0)
-                                RETURN(rc);
-                }
-        }
-
-        obd = exp->exp_obd;
-        push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-
-        if ((oa->o_valid & OBD_MD_FLFLAGS) &&
-            (oa->o_flags & OBD_FL_RECREATE_OBJS)) {
-                if (!obd->obd_recovering ||
-                    oa->o_id > filter_last_id(filter, oa->o_seq)) {
-                        CERROR("recreate objid "LPU64" > last id "LPU64"\n",
-                               oa->o_id, filter_last_id(filter, oa->o_seq));
-                        rc = -EINVAL;
-                } else {
-                        diff = 1;
-                        cfs_mutex_lock(&filter->fo_create_locks[oa->o_seq]);
-                        rc = filter_precreate(obd, oa, oa->o_seq, &diff);
-                        cfs_mutex_unlock(&filter->fo_create_locks[oa->o_seq]);
-                }
-        } else {
-                rc = filter_handle_precreate(exp, oa, oa->o_seq, oti);
-        }
-
-        pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-        if (rc && ea != NULL && *ea != lsm) {
-                obd_free_memmd(exp, &lsm);
-        } else if (rc == 0 && ea != NULL) {
-                /* XXX LOV STACKING: the lsm that is passed to us from
-                 * LOV does not have valid lsm_oinfo data structs, so
-                 * don't go touching that.  This needs to be fixed in a
-                 * big way. */
-                lsm->lsm_object_id = oa->o_id;
-                *ea = lsm;
-        }
-
-        RETURN(rc);
-}
-
-int filter_destroy(const struct lu_env *env, struct obd_export *exp,
-                   struct obdo *oa, struct lov_stripe_md *md,
-                   struct obd_trans_info *oti, struct obd_export *md_exp,
-                   void *capa)
-{
-        struct obd_device *obd;
-        struct filter_obd *filter;
-        struct dentry *dchild = NULL, *dparent = NULL;
-        struct lustre_handle lockh = { 0 };
-        struct lvfs_run_ctxt saved;
-        void *handle = NULL;
-        struct llog_cookie *fcc = NULL;
-        int rc, rc2, cleanup_phase = 0, sync = 0;
-        struct iattr iattr;
-        unsigned long now;
-        ENTRY;
-
-        rc = filter_auth_capa(exp, NULL, oa->o_seq,
-                              (struct lustre_capa *)capa, CAPA_OPC_OSS_DESTROY);
-        if (rc)
-                RETURN(rc);
-
-        obd = exp->exp_obd;
-        filter = &obd->u.filter;
-
-        push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-        cleanup_phase = 1;
-
-        CDEBUG(D_INODE, "%s: filter_destroy(group="LPU64",oid="
-               LPU64")\n", obd->obd_name, oa->o_seq, oa->o_id);
-
-        dchild = filter_fid2dentry(obd, NULL, oa->o_seq, oa->o_id);
-        if (IS_ERR(dchild))
-                GOTO(cleanup, rc = PTR_ERR(dchild));
-        cleanup_phase = 2;
-
-        if (dchild->d_inode == NULL) {
-                CDEBUG(D_INODE, "destroying non-existent object "POSTID"\n",
-                       oa->o_id, oa->o_seq);
-                /* If object already gone, cancel cookie right now */
-                if (oa->o_valid & OBD_MD_FLCOOKIE) {
-                        struct llog_ctxt *ctxt;
-                        struct obd_llog_group *olg;
-
-                        olg = filter_find_olg(obd, oa->o_seq);
-                        if (!olg) {
-                               CERROR(" %s: can not find olg of group %d\n",
-                                      obd->obd_name, (int)oa->o_seq);
-                               GOTO(cleanup, rc = PTR_ERR(olg));
-                        }
-                        fcc = &oa->o_lcookie;
-                        ctxt = llog_group_get_ctxt(olg, fcc->lgc_subsys + 1);
-                       llog_cancel(NULL, ctxt, NULL, 1, fcc, 0);
-                        llog_ctxt_put(ctxt);
-                        fcc = NULL; /* we didn't allocate fcc, don't free it */
-                }
-                GOTO(cleanup, rc = -ENOENT);
-        }
-
-        rc = filter_prepare_destroy(obd, oa->o_id, oa->o_seq, &lockh);
-        if (rc)
-                GOTO(cleanup, rc);
-
-        /* Our MDC connection is established by the MDS to us */
-        if (oa->o_valid & OBD_MD_FLCOOKIE) {
-                OBD_ALLOC(fcc, sizeof(*fcc));
-                if (fcc != NULL)
-                        *fcc = oa->o_lcookie;
-        }
-
-        /* we're gonna truncate it first in order to avoid possible deadlock:
-         *      P1                      P2
-         * open trasaction      open transaction
-         * down(i_zombie)       down(i_zombie)
-         *                      restart transaction
-         * (see BUG 4180) -bzzz
-         *
-         * take i_alloc_sem too to prevent other threads from writing to the
-         * file while we are truncating it. This can cause lock ordering issue
-         * between page lock, i_mutex & starting new journal handle.
-         * (see bug 20321) -johann
-         */
-       now = jiffies;
-       down_write(&dchild->d_inode->i_alloc_sem);
-       mutex_lock(&dchild->d_inode->i_mutex);
-       fsfilt_check_slow(exp->exp_obd, now, "i_alloc_sem and i_mutex");
-
-       /* VBR: version recovery check */
-       rc = filter_version_get_check(exp, oti, dchild->d_inode);
-       if (rc) {
-               mutex_unlock(&dchild->d_inode->i_mutex);
-               up_write(&dchild->d_inode->i_alloc_sem);
-               GOTO(cleanup, rc);
-       }
-
-       handle = fsfilt_start_log(obd, dchild->d_inode, FSFILT_OP_SETATTR,
-                                 NULL, 1);
-       if (IS_ERR(handle)) {
-               mutex_unlock(&dchild->d_inode->i_mutex);
-               up_write(&dchild->d_inode->i_alloc_sem);
-               GOTO(cleanup, rc = PTR_ERR(handle));
-       }
-
-       /* Locking order: i_mutex -> journal_lock -> dqptr_sem. LU-952 */
-       ll_vfs_dq_init(dchild->d_inode);
-
-       iattr.ia_valid = ATTR_SIZE;
-       iattr.ia_size = 0;
-       rc = fsfilt_setattr(obd, dchild, handle, &iattr, 1);
-       rc2 = fsfilt_commit(obd, dchild->d_inode, handle, 0);
-       mutex_unlock(&dchild->d_inode->i_mutex);
-       up_write(&dchild->d_inode->i_alloc_sem);
-       if (rc)
-               GOTO(cleanup, rc);
-       if (rc2)
-               GOTO(cleanup, rc = rc2);
-
-        /* We don't actually need to lock the parent until we are unlinking
-         * here, and not while truncating above.  That avoids holding the
-         * parent lock for a long time during truncate, which can block other
-         * threads from doing anything to objects in that directory. bug 7171 */
-        dparent = filter_parent_lock(obd, oa->o_seq, oa->o_id);
-        if (IS_ERR(dparent))
-                GOTO(cleanup, rc = PTR_ERR(dparent));
-        cleanup_phase = 3; /* filter_parent_unlock */
-
-       mutex_lock(&dchild->d_inode->i_mutex);
-       handle = fsfilt_start_log(obd, dparent->d_inode,
-                                 FSFILT_OP_UNLINK, oti, 1);
-       if (IS_ERR(handle)) {
-               mutex_unlock(&dchild->d_inode->i_mutex);
-               GOTO(cleanup, rc = PTR_ERR(handle));
-       }
-       cleanup_phase = 4; /* fsfilt_commit */
-
-        /* Quota release need uid/gid of inode */
-        obdo_from_inode(oa, dchild->d_inode, OBD_MD_FLUID | OBD_MD_FLGID);
-
-        filter_fmd_drop(exp, oa->o_id, oa->o_seq);
-
-        /* this drops dchild->d_inode->i_mutex unconditionally */
-        rc = filter_destroy_internal(obd, oa->o_id, oa->o_seq, dparent, dchild);
-
-        EXIT;
-cleanup:
-        switch(cleanup_phase) {
-        case 4:
-                if (fcc != NULL)
-                        sync = fsfilt_add_journal_cb(obd, 0, oti ?
-                                                     oti->oti_handle : handle,
-                                                     filter_cancel_cookies_cb,
-                                                     fcc);
-                /* If add_journal_cb failed, then filter_finish_transno
-                 * will commit the handle and we will do a sync
-                 * on commit. then we call callback directly to free
-                 * the fcc.
-                 */
-                rc = filter_finish_transno(exp, NULL, oti, rc, sync);
-                if (sync) {
-                        filter_cancel_cookies_cb(obd, 0, fcc, rc);
-                        fcc = NULL;
-                }
-                rc2 = fsfilt_commit(obd, dparent->d_inode, handle, 0);
-                if (rc2) {
-                        CERROR("error on commit, err = %d\n", rc2);
-                        if (!rc)
-                                rc = rc2;
-                } else {
-                        fcc = NULL;
-                }
-        case 3:
-                filter_parent_unlock(dparent);
-        case 2:
-                filter_fini_destroy(obd, &lockh);
-
-                f_dput(dchild);
-                if (fcc != NULL)
-                        OBD_FREE(fcc, sizeof(*fcc));
-        case 1:
-                pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-                break;
-        default:
-                CERROR("invalid cleanup_phase %d\n", cleanup_phase);
-                LBUG();
-        }
-
-        return rc;
-}
-
-/* NB start and end are used for punch, but not truncate */
-static int filter_truncate(const struct lu_env *env, struct obd_export *exp,
-                           struct obd_info *oinfo, struct obd_trans_info *oti,
-                           struct ptlrpc_request_set *rqset)
-{
-        int rc;
-        ENTRY;
-
-        if (oinfo->oi_policy.l_extent.end != OBD_OBJECT_EOF) {
-                CERROR("PUNCH not supported, only truncate: end = "LPX64"\n",
-                       oinfo->oi_policy.l_extent.end);
-                RETURN(-EFAULT);
-        }
-
-        CDEBUG(D_INODE, "calling truncate for object "LPU64", valid = "LPX64
-               ", o_size = "LPD64"\n", oinfo->oi_oa->o_id,oinfo->oi_oa->o_valid,
-                oinfo->oi_policy.l_extent.start);
-
-        oinfo->oi_oa->o_size = oinfo->oi_policy.l_extent.start;
-        rc = filter_setattr(env, exp, oinfo, oti);
-        RETURN(rc);
-}
-
-static int filter_sync(const struct lu_env *env, struct obd_export *exp,
-                       struct obd_info *oinfo, obd_off start, obd_off end,
-                       struct ptlrpc_request_set *set)
-{
-        struct lvfs_run_ctxt saved;
-        struct obd_device_target *obt;
-        struct dentry *dentry;
-        int rc, rc2;
-        ENTRY;
-
-        rc = filter_auth_capa(exp, NULL, oinfo->oi_oa->o_seq,
-                              (struct lustre_capa *)oinfo->oi_capa,
-                              CAPA_OPC_OSS_WRITE);
-        if (rc)
-                RETURN(rc);
-
-        obt = &exp->exp_obd->u.obt;
-
-        /* An objid of zero is taken to mean "sync whole filesystem" */
-        if (!oinfo->oi_oa || !(oinfo->oi_oa->o_valid & OBD_MD_FLID)) {
-                rc = fsfilt_sync(exp->exp_obd, obt->obt_sb);
-                /* Flush any remaining cancel messages out to the target */
-                filter_sync_llogs(exp->exp_obd, exp);
-                RETURN(rc);
-        }
-
-        dentry = filter_oa2dentry(exp->exp_obd, &oinfo->oi_oa->o_oi);
-        if (IS_ERR(dentry))
-                RETURN(PTR_ERR(dentry));
-
-        push_ctxt(&saved, &exp->exp_obd->obd_lvfs_ctxt, NULL);
-
-       mutex_lock(&dentry->d_inode->i_mutex);
-
-       rc = filemap_fdatawrite(dentry->d_inode->i_mapping);
-       if (rc == 0) {
-               /* just any file to grab fsync method - "file" arg unused */
-               struct file *file = obt->obt_rcvd_filp;
-
-               if (file->f_op && file->f_op->fsync)
-                       rc = file->f_op->fsync(NULL, dentry, 1);
-
-               rc2 = filemap_fdatawait(dentry->d_inode->i_mapping);
-               if (!rc)
-                       rc = rc2;
-       }
-       mutex_unlock(&dentry->d_inode->i_mutex);
-
-        oinfo->oi_oa->o_valid = OBD_MD_FLID;
-        obdo_from_inode(oinfo->oi_oa, dentry->d_inode, FILTER_VALID_FLAGS);
-
-        pop_ctxt(&saved, &exp->exp_obd->obd_lvfs_ctxt, NULL);
-
-       filter_counter_incr(exp, LPROC_FILTER_STATS_SYNC, oinfo->oi_jobid, 1);
-        f_dput(dentry);
-        RETURN(rc);
-}
-
-static int filter_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;
-        ENTRY;
-
-        obd = class_exp2obd(exp);
-        if (obd == NULL) {
-                CDEBUG(D_IOCTL, "invalid client export %p\n", exp);
-                RETURN(-EINVAL);
-        }
-
-        if (KEY_IS(KEY_BLOCKSIZE)) {
-                __u32 *blocksize = val;
-                if (blocksize) {
-                        if (*vallen < sizeof(*blocksize))
-                                RETURN(-EOVERFLOW);
-                        *blocksize = obd->u.obt.obt_sb->s_blocksize;
-                }
-                *vallen = sizeof(*blocksize);
-                RETURN(0);
-        }
-
-        if (KEY_IS(KEY_BLOCKSIZE_BITS)) {
-                __u32 *blocksize_bits = val;
-                if (blocksize_bits) {
-                        if (*vallen < sizeof(*blocksize_bits))
-                                RETURN(-EOVERFLOW);
-                        *blocksize_bits = obd->u.obt.obt_sb->s_blocksize_bits;
-                }
-                *vallen = sizeof(*blocksize_bits);
-                RETURN(0);
-        }
-
-        if (KEY_IS(KEY_LAST_ID)) {
-                obd_id *last_id = val;
-                /* FIXME: object groups */
-                if (last_id) {
-                        if (*vallen < sizeof(*last_id))
-                                RETURN(-EOVERFLOW);
-                        *last_id = filter_last_id(&obd->u.filter,
-                                                  exp->exp_filter_data.fed_group);
-                }
-                *vallen = sizeof(*last_id);
-                RETURN(0);
-        }
-
-        if (KEY_IS(KEY_FIEMAP)) {
-                struct ll_fiemap_info_key *fm_key = key;
-                struct dentry *dentry;
-                struct ll_user_fiemap *fiemap = val;
-                struct lvfs_run_ctxt saved;
-                int rc;
-
-                if (fiemap == NULL) {
-                        *vallen = fiemap_count_to_size(
-                                                fm_key->fiemap.fm_extent_count);
-                        RETURN(0);
-                }
-
-                dentry = __filter_oa2dentry(exp->exp_obd, &fm_key->oa.o_oi,
-                                            __func__, 1);
-                if (IS_ERR(dentry))
-                        RETURN(PTR_ERR(dentry));
-
-                memcpy(fiemap, &fm_key->fiemap, sizeof(*fiemap));
-                push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-                rc = fsfilt_iocontrol(obd, dentry, FSFILT_IOC_FIEMAP,
-                                      (long)fiemap);
-                pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-
-                f_dput(dentry);
-                RETURN(rc);
-        }
-
-        if (KEY_IS(KEY_SYNC_LOCK_CANCEL)) {
-                *((__u32 *) val) = obd->u.filter.fo_sync_lock_cancel;
-                *vallen = sizeof(__u32);
-                RETURN(0);
-        }
-
-        CDEBUG(D_IOCTL, "invalid key\n");
-        RETURN(-EINVAL);
-}
-
-static inline int filter_setup_llog_group(struct obd_export *exp,
-                                          struct obd_device *obd,
-                                           int group)
-{
-        struct obd_llog_group *olg;
-        struct llog_ctxt *ctxt;
-        int rc;
-
-        olg = filter_find_create_olg(obd, group);
-        if (IS_ERR(olg))
-                RETURN(PTR_ERR(olg));
-
-        llog_group_set_export(olg, exp);
-
-        ctxt = llog_group_get_ctxt(olg, LLOG_MDS_OST_REPL_CTXT);
-        LASSERTF(ctxt != NULL, "ctxt is null\n");
-
-        rc = llog_receptor_accept(ctxt, exp->exp_imp_reverse);
-        llog_ctxt_put(ctxt);
-        return rc;
-}
-
-static int filter_set_grant_shrink(struct obd_export *exp,
-                                   struct ost_body *body)
-{
-        /* handle shrink grant */
-        cfs_spin_lock(&exp->exp_obd->obd_osfs_lock);
-        filter_grant_incoming(exp, &body->oa);
-        cfs_spin_unlock(&exp->exp_obd->obd_osfs_lock);
-
-        RETURN(0);
-
-}
-
-static int filter_set_mds_conn(struct obd_export *exp, void *val)
-{
-        struct obd_device *obd;
-        int rc = 0, group;
-        ENTRY;
-
-        obd = exp->exp_obd;
-        if (obd == NULL) {
-                CDEBUG(D_IOCTL, "invalid export %p\n", exp);
-                RETURN(-EINVAL);
-        }
-
-        LCONSOLE_WARN("%s: received MDS connection from %s\n", obd->obd_name,
-                      obd_export_nid2str(exp));
-        obd->u.filter.fo_mdc_conn.cookie = exp->exp_handle.h_cookie;
-
-        /* setup llog imports */
-        if (val != NULL)
-                group = (int)(*(__u32 *)val);
-        else
-                group = 0; /* default value */
-
-        LASSERT_SEQ_IS_MDT(group);
-        rc = filter_setup_llog_group(exp, obd, group);
-        if (rc)
-                goto out;
-
-        if (group == FID_SEQ_OST_MDT0) {
-                /* setup llog group 1 for interop */
-                filter_setup_llog_group(exp, obd, FID_SEQ_LLOG);
-        }
-out:
-        RETURN(rc);
-}
-
-static int filter_set_info_async(const struct lu_env *env,
-                                 struct obd_export *exp, __u32 keylen,
-                                 void *key, __u32 vallen, void *val,
-                                 struct ptlrpc_request_set *set)
-{
-        struct obd_device *obd;
-        ENTRY;
-
-        obd = exp->exp_obd;
-        if (obd == NULL) {
-                CDEBUG(D_IOCTL, "invalid export %p\n", exp);
-                RETURN(-EINVAL);
-        }
-
-        if (KEY_IS(KEY_CAPA_KEY)) {
-                int rc;
-                rc = filter_update_capa_key(obd, (struct lustre_capa_key *)val);
-                if (rc)
-                        CERROR("filter update capability key failed: %d\n", rc);
-                RETURN(rc);
-        }
-
-        if (KEY_IS(KEY_REVIMP_UPD)) {
-                filter_revimp_update(exp);
-                RETURN(0);
-        }
-
-        if (KEY_IS(KEY_SPTLRPC_CONF)) {
-                filter_adapt_sptlrpc_conf(obd, 0);
-                RETURN(0);
-        }
-
-        if (KEY_IS(KEY_MDS_CONN))
-                RETURN(filter_set_mds_conn(exp, val));
-
-        if (KEY_IS(KEY_GRANT_SHRINK))
-                RETURN(filter_set_grant_shrink(exp, val));
-
-        RETURN(-EINVAL);
-}
-
-int filter_iocontrol(unsigned int cmd, struct obd_export *exp,
-                     int len, void *karg, void *uarg)
-{
-        struct obd_device *obd = exp->exp_obd;
-        struct obd_ioctl_data *data = karg;
-        int rc = 0;
-
-        switch (cmd) {
-        case OBD_IOC_ABORT_RECOVERY: {
-                LCONSOLE_WARN("%s: Aborting recovery.\n", obd->obd_name);
-                target_stop_recovery_thread(obd);
-                RETURN(0);
-        }
-
-        case OBD_IOC_SYNC: {
-                CDEBUG(D_RPCTRACE, "syncing ost %s\n", obd->obd_name);
-                rc = fsfilt_sync(obd, obd->u.obt.obt_sb);
-                RETURN(rc);
-        }
-
-        case OBD_IOC_SET_READONLY: {
-                void *handle;
-                struct super_block *sb = obd->u.obt.obt_sb;
-                struct inode *inode = sb->s_root->d_inode;
-                BDEVNAME_DECLARE_STORAGE(tmp);
-                CERROR("*** setting device %s read-only ***\n",
-                       ll_bdevname(sb, tmp));
-
-                handle = fsfilt_start(obd, inode, FSFILT_OP_MKNOD, NULL);
-                if (!IS_ERR(handle))
-                        rc = fsfilt_commit(obd, inode, handle, 1);
-
-                CDEBUG(D_HA, "syncing ost %s\n", obd->obd_name);
-                rc = fsfilt_sync(obd, obd->u.obt.obt_sb);
-
-                rc = lvfs_set_rdonly(obd, obd->u.obt.obt_sb);
-                RETURN(rc);
-        }
-
-        case OBD_IOC_CATLOGLIST: {
-                rc = llog_catalog_list(obd, 1, data);
-                RETURN(rc);
-        }
-
-        case OBD_IOC_LLOG_CANCEL:
-        case OBD_IOC_LLOG_REMOVE:
-        case OBD_IOC_LLOG_INFO:
-        case OBD_IOC_LLOG_PRINT: {
-                /* FIXME to be finished */
-                RETURN(-EOPNOTSUPP);
-/*
-                struct llog_ctxt *ctxt = NULL;
-
-                push_ctxt(&saved, &ctxt->loc_exp->exp_obd->obd_lvfs_ctxt, NULL);
-                rc = llog_ioctl(ctxt, cmd, data);
-                pop_ctxt(&saved, &ctxt->loc_exp->exp_obd->obd_lvfs_ctxt, NULL);
-
-                RETURN(rc);
-*/
-        }
-
-
-        default:
-                RETURN(-EINVAL);
-        }
-        RETURN(0);
-}
-
-static int filter_health_check(const struct lu_env *env, struct obd_device *obd)
-{
-#ifdef USE_HEALTH_CHECK_WRITE
-        struct filter_obd *filter = &obd->u.filter;
-#endif
-        int rc = 0;
-
-        /*
-         * health_check to return 0 on healthy
-         * and 1 on unhealthy.
-         */
-        if (obd->u.obt.obt_sb->s_flags & MS_RDONLY)
-                rc = 1;
-
-#ifdef USE_HEALTH_CHECK_WRITE
-        LASSERT(filter->fo_obt.obt_health_check_filp != NULL);
-        rc |= !!lvfs_check_io_health(obd, filter->fo_obt.obt_health_check_filp);
-#endif
-        return rc;
-}
-
-static struct dentry *filter_lvfs_fid2dentry(__u64 id, __u32 gen, __u64 gr,
-                                             void *data)
-{
-        return filter_fid2dentry(data, NULL, gr, id);
-}
-
-static int filter_process_config(struct obd_device *obd, obd_count len,
-                                 void *buf)
-{
-        struct lustre_cfg *lcfg = buf;
-        struct lprocfs_static_vars lvars;
-        int rc = 0;
-
-        switch (lcfg->lcfg_command) {
-        default:
-                lprocfs_filter_init_vars(&lvars);
-
-                rc = class_process_proc_param(PARAM_OST, lvars.obd_vars,
-                                              lcfg, obd);
-                if (rc > 0)
-                        rc = 0;
-                break;
-        }
-
-        return rc;
-}
-
-static int filter_notify(struct obd_device *obd,
-                         struct obd_device *unused,
-                         enum obd_notify_event ev, void *data)
-{
-        switch (ev) {
-        case OBD_NOTIFY_CONFIG:
-                LASSERT(obd->obd_no_conn);
-                cfs_spin_lock(&obd->obd_dev_lock);
-                obd->obd_no_conn = 0;
-                cfs_spin_unlock(&obd->obd_dev_lock);
-                break;
-        default:
-                CDEBUG(D_INFO, "%s: Unhandled notification %#x\n",
-                       obd->obd_name, ev);
-        }
-        return 0;
-}
-
-static struct lvfs_callback_ops filter_lvfs_ops = {
-        l_fid2dentry:     filter_lvfs_fid2dentry,
-};
-
-static struct obd_ops filter_obd_ops = {
-        .o_owner          = THIS_MODULE,
-        .o_get_info       = filter_get_info,
-        .o_set_info_async = filter_set_info_async,
-        .o_setup          = filter_setup,
-        .o_precleanup     = filter_precleanup,
-        .o_cleanup        = filter_cleanup,
-        .o_connect        = filter_connect,
-        .o_reconnect      = filter_reconnect,
-        .o_disconnect     = filter_disconnect,
-        .o_ping           = filter_ping,
-        .o_init_export    = filter_init_export,
-        .o_destroy_export = filter_destroy_export,
-        .o_statfs         = filter_statfs,
-        .o_getattr        = filter_getattr,
-        .o_unpackmd       = filter_unpackmd,
-        .o_create         = filter_create,
-        .o_setattr        = filter_setattr,
-        .o_destroy        = filter_destroy,
-        .o_punch          = filter_truncate,
-        .o_sync           = filter_sync,
-        .o_preprw         = filter_preprw,
-        .o_commitrw       = filter_commitrw,
-        .o_llog_init      = filter_llog_init,
-        .o_llog_connect   = filter_llog_connect,
-        .o_llog_finish    = filter_llog_finish,
-        .o_iocontrol      = filter_iocontrol,
-        .o_health_check   = filter_health_check,
-        .o_process_config = filter_process_config,
-        .o_notify         = filter_notify,
-};
-
-static int __init obdfilter_init(void)
-{
-        struct lprocfs_static_vars lvars;
-        int rc, i;
-
-        /** sanity check for group<->mdsno conversion */
-        for (i = 0; i < MAX_MDT_COUNT; i++)
-                 LASSERT(objseq_to_mdsno(mdt_to_obd_objseq(i)) == i);
-
-        lprocfs_filter_init_vars(&lvars);
-
-        OBD_ALLOC(obdfilter_created_scratchpad,
-                  OBDFILTER_CREATED_SCRATCHPAD_ENTRIES *
-                  sizeof(*obdfilter_created_scratchpad));
-        if (obdfilter_created_scratchpad == NULL)
-                return -ENOMEM;
-
-        ll_fmd_cachep = cfs_mem_cache_create("ll_fmd_cache",
-                                             sizeof(struct filter_mod_data),
-                                             0, 0);
-        if (!ll_fmd_cachep)
-                GOTO(out, rc = -ENOMEM);
-
-        rc = class_register_type(&filter_obd_ops, NULL, lvars.module_vars,
-                                 LUSTRE_OST_NAME, NULL);
-        if (rc) {
-                int err;
-
-                err = cfs_mem_cache_destroy(ll_fmd_cachep);
-                LASSERTF(err == 0, "Cannot destroy ll_fmd_cachep: rc %d\n",err);
-                ll_fmd_cachep = NULL;
-out:
-                OBD_FREE(obdfilter_created_scratchpad,
-                         OBDFILTER_CREATED_SCRATCHPAD_ENTRIES *
-                         sizeof(*obdfilter_created_scratchpad));
-        }
-
-        return rc;
-}
-
-static void __exit obdfilter_exit(void)
-{
-        if (ll_fmd_cachep) {
-                int rc = cfs_mem_cache_destroy(ll_fmd_cachep);
-                LASSERTF(rc == 0, "Cannot destroy ll_fmd_cachep: rc %d\n", rc);
-                ll_fmd_cachep = NULL;
-        }
-
-        class_unregister_type(LUSTRE_OST_NAME);
-        OBD_FREE(obdfilter_created_scratchpad,
-                 OBDFILTER_CREATED_SCRATCHPAD_ENTRIES *
-                 sizeof(*obdfilter_created_scratchpad));
-}
-
-MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
-MODULE_DESCRIPTION("Lustre Filtering OBD driver");
-MODULE_LICENSE("GPL");
-
-module_init(obdfilter_init);
-module_exit(obdfilter_exit);
diff --git a/lustre/obdfilter/filter_capa.c b/lustre/obdfilter/filter_capa.c
deleted file mode 100644 (file)
index 0e23110..0000000
+++ /dev/null
@@ -1,306 +0,0 @@
-/*
- * GPL HEADER START
- *
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License version 2 only,
- * as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License version 2 for more details (a copy is included
- * in the LICENSE file that accompanied this code).
- *
- * You should have received a copy of the GNU General Public License
- * version 2 along with this program; If not, see
- * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
- *
- * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
- * CA 95054 USA or visit www.sun.com if you need additional information or
- * have any questions.
- *
- * GPL HEADER END
- */
-/*
- * Copyright (c) 2005, 2010, Oracle and/or its affiliates. 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.
- *
- * lustre/obdfilter/filter_capa.c
- *
- * Author: Lai Siyao <lsy@clusterfs.com>
- */
-
-#define DEBUG_SUBSYSTEM S_FILTER
-
-#include <linux/fs.h>
-#include <linux/version.h>
-#include <asm/uaccess.h>
-#include <linux/file.h>
-#include <linux/kmod.h>
-
-#include <lustre_fsfilt.h>
-#include <lustre_capa.h>
-
-#include "filter_internal.h"
-
-static inline __u32 filter_ck_keyid(struct filter_capa_key *key)
-{
-        return key->k_key.lk_keyid;
-}
-
-int filter_update_capa_key(struct obd_device *obd, struct lustre_capa_key *new)
-{
-        struct filter_obd *filter = &obd->u.filter;
-        struct filter_capa_key *k, *keys[2] = { NULL, NULL };
-        int i;
-
-        cfs_spin_lock(&capa_lock);
-        cfs_list_for_each_entry(k, &filter->fo_capa_keys, k_list) {
-                if (k->k_key.lk_seq != new->lk_seq)
-                        continue;
-
-                if (keys[0]) {
-                        keys[1] = k;
-                        if (filter_ck_keyid(keys[1]) > filter_ck_keyid(keys[0]))
-                                keys[1] = keys[0], keys[0] = k;
-                } else {
-                        keys[0] = k;
-                }
-        }
-        cfs_spin_unlock(&capa_lock);
-
-        for (i = 0; i < 2; i++) {
-                if (!keys[i])
-                        continue;
-                if (filter_ck_keyid(keys[i]) != new->lk_keyid)
-                        continue;
-                /* maybe because of recovery or other reasons, MDS sent the
-                 * the old capability key again.
-                 */
-                cfs_spin_lock(&capa_lock);
-                keys[i]->k_key = *new;
-                cfs_spin_unlock(&capa_lock);
-
-                RETURN(0);
-        }
-
-        if (keys[1]) {
-                /* if OSS already have two keys, update the old one */
-                k = keys[1];
-        } else {
-                OBD_ALLOC_PTR(k);
-                if (!k)
-                        RETURN(-ENOMEM);
-                CFS_INIT_LIST_HEAD(&k->k_list);
-        }
-
-        cfs_spin_lock(&capa_lock);
-        k->k_key = *new;
-        if (cfs_list_empty(&k->k_list))
-                cfs_list_add(&k->k_list, &filter->fo_capa_keys);
-        cfs_spin_unlock(&capa_lock);
-
-        DEBUG_CAPA_KEY(D_SEC, new, "new");
-        RETURN(0);
-}
-
-int filter_auth_capa(struct obd_export *exp,
-                    struct lu_fid *fid, obd_seq seq,
-                    struct lustre_capa *capa, __u64 opc)
-{
-        struct obd_device *obd = exp->exp_obd;
-        struct filter_obd *filter = &obd->u.filter;
-        struct filter_capa_key *k;
-        struct lustre_capa_key key;
-        struct obd_capa *oc;
-        __u8 *hmac;
-        int keys_ready = 0, key_found = 0, rc = 0;
-        ENTRY;
-
-        /* skip capa check for llog and obdecho */
-        if (!fid_seq_is_mdt(seq))
-                RETURN(0);
-
-       if (capa == BYPASS_CAPA)
-               RETURN(0);
-
-        /* capability is disabled */
-        if (!filter->fo_fl_oss_capa)
-                RETURN(0);
-
-        if (!(exp->exp_connect_flags & OBD_CONNECT_OSS_CAPA))
-                RETURN(0);
-
-        if (capa == NULL) {
-                if (fid)
-                        CERROR("seq/fid/opc "LPU64"/"DFID"/"LPX64
-                               ": no capability has been passed\n",
-                               seq, PFID(fid), opc);
-                else
-                        CERROR("seq/opc "LPU64"/"LPX64
-                               ": no capability has been passed\n",
-                               seq, opc);
-                RETURN(-EACCES);
-        }
-
-        if (opc == CAPA_OPC_OSS_READ) {
-                if (!(capa->lc_opc & CAPA_OPC_OSS_RW))
-                        rc = -EACCES;
-        } else if (!capa_opc_supported(capa, opc)) {
-                rc = -EACCES;
-        }
-        if (rc) {
-                DEBUG_CAPA(D_ERROR, capa, "opc "LPX64" not supported by", opc);
-                RETURN(rc);
-        }
-
-        oc = capa_lookup(filter->fo_capa_hash, capa, 0);
-        if (oc) {
-                cfs_spin_lock(&oc->c_lock);
-                if (capa_is_expired(oc)) {
-                        DEBUG_CAPA(D_ERROR, capa, "expired");
-                        rc = -ESTALE;
-                }
-                cfs_spin_unlock(&oc->c_lock);
-
-                capa_put(oc);
-                RETURN(rc);
-        }
-
-        if (capa_is_expired_sec(capa)) {
-                DEBUG_CAPA(D_ERROR, capa, "expired");
-                RETURN(-ESTALE);
-        }
-
-        cfs_spin_lock(&capa_lock);
-        cfs_list_for_each_entry(k, &filter->fo_capa_keys, k_list) {
-                if (k->k_key.lk_seq == seq) {
-                        keys_ready = 1;
-                        if (k->k_key.lk_keyid == capa_keyid(capa)) {
-                                key = k->k_key;
-                                key_found = 1;
-                                break;
-                        }
-                }
-        }
-        cfs_spin_unlock(&capa_lock);
-
-        if (!keys_ready) {
-                CDEBUG(D_SEC, "MDS hasn't propagated capability keys yet, "
-                       "ignore check!\n");
-                RETURN(0);
-        }
-
-       if (!key_found) {
-                DEBUG_CAPA(D_ERROR, capa, "no matched capability key for");
-                RETURN(-ESTALE);
-        }
-
-        OBD_ALLOC(hmac, CAPA_HMAC_MAX_LEN);
-        if (hmac == NULL)
-                RETURN(-ENOMEM);
-
-        rc = capa_hmac(hmac, capa, key.lk_key);
-        if (rc) {
-                DEBUG_CAPA(D_ERROR, capa, "HMAC failed: rc %d", rc);
-                OBD_FREE(hmac, CAPA_HMAC_MAX_LEN);
-                RETURN(rc);
-        }
-
-        rc = memcmp(hmac, capa->lc_hmac, CAPA_HMAC_MAX_LEN);
-        OBD_FREE(hmac, CAPA_HMAC_MAX_LEN);
-        if (rc) {
-                DEBUG_CAPA_KEY(D_ERROR, &key, "calculate HMAC with ");
-                DEBUG_CAPA(D_ERROR, capa, "HMAC mismatch");
-                RETURN(-EACCES);
-        }
-
-        /* store in capa hash */
-        oc = capa_add(filter->fo_capa_hash, capa);
-        capa_put(oc);
-        RETURN(0);
-}
-
-int filter_capa_fixoa(struct obd_export *exp, struct obdo *oa, obd_seq seq,
-                      struct lustre_capa *capa)
-{
-       struct obd_device       *obd = exp->exp_obd;
-       struct filter_obd       *filter = &obd->u.filter;
-       int                     rc = 0;
-       ENTRY;
-
-        /* skip capa check for llog and obdecho */
-        if (!fid_seq_is_mdt(seq))
-                RETURN(0);
-
-        if (!(exp->exp_connect_flags & OBD_CONNECT_OSS_CAPA))
-                RETURN(0);
-
-       /* capability is disabled */
-       if (!filter->fo_fl_oss_capa)
-               RETURN(0);
-
-        if (unlikely(!capa))
-                RETURN(-EACCES);
-
-        if (capa_flags(capa) == LC_ID_CONVERT) {
-                struct filter_capa_key *k;
-                int found = 0;
-
-                cfs_spin_lock(&capa_lock);
-                cfs_list_for_each_entry(k, &filter->fo_capa_keys, k_list) {
-                        if (k->k_key.lk_seq == seq &&
-                            k->k_key.lk_keyid == capa_keyid(capa)) {
-                                found = 1;
-                                break;
-                        }
-                }
-                cfs_spin_unlock(&capa_lock);
-
-                if (found) {
-                        union {
-                                __u64 id64;
-                                __u32 id32[2];
-                        } uid, gid;
-                        __u32 d[4], s[4];
-
-                        uid.id64 = capa_uid(capa);
-                        gid.id64 = capa_gid(capa);
-                        s[0] = uid.id32[0];
-                        s[1] = uid.id32[1];
-                        s[2] = gid.id32[0];
-                        s[3] = gid.id32[1];
-
-                        rc = capa_decrypt_id(d, s, k->k_key.lk_key,
-                                             CAPA_HMAC_KEY_MAX_LEN);
-                        if (unlikely(rc))
-                                RETURN(rc);
-
-                        oa->o_uid = d[0];
-                        oa->o_gid = d[2];
-                } else {
-                        DEBUG_CAPA(D_ERROR, capa, "no matched capability key for");
-                        rc = -ESTALE;
-                }
-        }
-
-        RETURN(rc);
-}
-
-void filter_free_capa_keys(struct filter_obd *filter)
-{
-        struct filter_capa_key *key, *n;
-
-        cfs_spin_lock(&capa_lock);
-        cfs_list_for_each_entry_safe(key, n, &filter->fo_capa_keys, k_list) {
-                cfs_list_del_init(&key->k_list);
-                OBD_FREE(key, sizeof(*key));
-        }
-        cfs_spin_unlock(&capa_lock);
-}
diff --git a/lustre/obdfilter/filter_internal.h b/lustre/obdfilter/filter_internal.h
deleted file mode 100644 (file)
index fb51922..0000000
+++ /dev/null
@@ -1,291 +0,0 @@
-/*
- * GPL HEADER START
- *
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License version 2 only,
- * as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License version 2 for more details (a copy is included
- * in the LICENSE file that accompanied this code).
- *
- * You should have received a copy of the GNU General Public License
- * version 2 along with this program; If not, see
- * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
- *
- * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
- * CA 95054 USA or visit www.sun.com if you need additional information or
- * have any questions.
- *
- * GPL HEADER END
- */
-/*
- * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
- * Use is subject to license terms.
- *
- * Copyright (c) 2011, Whamcloud, Inc.
- */
-/*
- * This file is part of Lustre, http://www.lustre.org/
- * Lustre is a trademark of Sun Microsystems, Inc.
- */
-
-#ifndef _FILTER_INTERNAL_H
-#define _FILTER_INTERNAL_H
-
-#ifdef __KERNEL__
-# include <linux/spinlock.h>
-#endif
-#include <lustre_handles.h>
-#include <lustre_debug.h>
-#include <obd.h>
-#include <obd_class.h>
-#include <lprocfs_status.h>
-
-#define FILTER_LAYOUT_VERSION "2"
-
-#define FILTER_INIT_OBJID 0
-
-#define FILTER_SUBDIR_COUNT 32 /* set to zero for no subdirs */
-#define FILTER_GROUPS        3 /* must be at least 3; not dynamic yet */
-
-#define FILTER_ROCOMPAT_SUPP (0)
-
-#define FILTER_INCOMPAT_SUPP (OBD_INCOMPAT_GROUPS | OBD_INCOMPAT_OST | \
-                              OBD_INCOMPAT_COMMON_LR)
-
-#define FILTER_GRANT_CHUNK (2ULL * PTLRPC_MAX_BRW_SIZE)
-#define FILTER_GRANT_SHRINK_LIMIT (16ULL * FILTER_GRANT_CHUNK)
-#define GRANT_FOR_LLOG(obd) 16
-
-extern struct file_operations filter_per_export_stats_fops;
-extern struct file_operations filter_per_nid_stats_fops;
-
-/* Limit the returned fields marked valid to those that we actually might set */
-#define FILTER_VALID_FLAGS (OBD_MD_FLTYPE | OBD_MD_FLMODE | OBD_MD_FLGENER  |\
-                            OBD_MD_FLSIZE | OBD_MD_FLBLOCKS | OBD_MD_FLBLKSZ|\
-                            OBD_MD_FLATIME | OBD_MD_FLMTIME | OBD_MD_FLCTIME)
-
-/* per-client-per-object persistent state (LRU) */
-struct filter_mod_data {
-        cfs_list_t       fmd_list;       /* linked to fed_mod_list */
-        __u64            fmd_id;         /* object being written to */
-        __u64            fmd_gr;         /* group being written to */
-        __u64            fmd_mactime_xid;/* xid highest {m,a,c}time
-                                              * setattr */
-        unsigned long    fmd_expire;   /* jiffies when it should expire */
-        int              fmd_refcount; /* reference counter, list holds 1 */
-};
-
-#ifdef HAVE_BGL_SUPPORT
-#define FILTER_FMD_MAX_NUM_DEFAULT 128 /* many active files per client on BGL */
-#else
-#define FILTER_FMD_MAX_NUM_DEFAULT  32
-#endif
-/* Client cache seconds */
-#define FILTER_FMD_MAX_AGE_DEFAULT ((obd_timeout + 10) * CFS_HZ)
-
-#ifndef HAVE_PAGE_CONSTANT
-#define mapping_cap_page_constant_write(mapping) 0
-#define SetPageConstant(page) do {} while (0)
-#define ClearPageConstant(page) do {} while (0)
-#endif
-
-struct filter_mod_data *filter_fmd_find(struct obd_export *exp,
-                                        obd_id objid, obd_seq seq);
-struct filter_mod_data *filter_fmd_get(struct obd_export *exp,
-                                       obd_id objid, obd_seq seq);
-void filter_fmd_put(struct obd_export *exp, struct filter_mod_data *fmd);
-void filter_fmd_expire(struct obd_export *exp);
-
-enum {
-        LPROC_FILTER_READ_BYTES = 0,
-        LPROC_FILTER_WRITE_BYTES = 1,
-        LPROC_FILTER_GET_PAGE = 2,
-        LPROC_FILTER_NO_PAGE = 3,
-        LPROC_FILTER_CACHE_ACCESS = 4,
-        LPROC_FILTER_CACHE_HIT = 5,
-        LPROC_FILTER_CACHE_MISS = 6,
-        LPROC_FILTER_LAST,
-};
-
-/* for job stats */
-enum {
-       LPROC_FILTER_STATS_READ = 0,
-       LPROC_FILTER_STATS_WRITE = 1,
-       LPROC_FILTER_STATS_SETATTR = 2,
-       LPROC_FILTER_STATS_PUNCH = 3,
-       LPROC_FILTER_STATS_SYNC = 4,
-       LPROC_FILTER_STATS_LAST,
-};
-
-static inline void filter_counter_incr(struct obd_export *exp, int opcode,
-                                      char *jobid, long amount)
-{
-       if (exp->exp_obd && exp->exp_obd->u.obt.obt_jobstats.ojs_hash &&
-           (exp->exp_connect_flags & OBD_CONNECT_JOBSTATS))
-               lprocfs_job_stats_log(exp->exp_obd, jobid, opcode, amount);
-}
-
-//#define FILTER_MAX_CACHE_SIZE (32 * 1024 * 1024) /* was OBD_OBJECT_EOF */
-#define FILTER_MAX_CACHE_SIZE OBD_OBJECT_EOF
-
-/* We have to pass a 'created' array to fsfilt_map_inode_pages() which we
- * then ignore.  So we pre-allocate one that everyone can use... */
-#define OBDFILTER_CREATED_SCRATCHPAD_ENTRIES 1024
-extern int *obdfilter_created_scratchpad;
-
-extern void target_recovery_fini(struct obd_device *obd);
-extern void target_recovery_init(struct lu_target *lut,
-                                 svc_handler_t handler);
-
-/* filter.c */
-void f_dput(struct dentry *);
-struct dentry *filter_fid2dentry(struct obd_device *, struct dentry *dir,
-                                 obd_seq seq, obd_id id);
-struct dentry *__filter_oa2dentry(struct obd_device *obd, struct ost_id *ostid,
-                                  const char *what, int quiet);
-#define filter_oa2dentry(obd, ostid) __filter_oa2dentry(obd, ostid,     \
-                                                        __func__, 0)
-
-int filter_finish_transno(struct obd_export *, struct inode *,
-                          struct obd_trans_info *, int rc, int force_sync);
-__u64 filter_next_id(struct filter_obd *, struct obdo *);
-__u64 filter_last_id(struct filter_obd *, obd_seq seq);
-int filter_update_fidea(struct obd_export *exp, struct inode *inode,
-                        void *handle, struct obdo *oa);
-int filter_update_server_data(struct obd_device *);
-int filter_update_last_objid(struct obd_device *, obd_seq, int force_sync);
-int filter_common_setup(struct obd_device *, struct lustre_cfg *lcfg,
-                        void *option);
-int filter_destroy(const struct lu_env *env, struct obd_export *exp,
-                   struct obdo *oa, struct lov_stripe_md *md,
-                   struct obd_trans_info *, struct obd_export *, void *);
-int filter_setattr_internal(struct obd_export *exp, struct dentry *dentry,
-                            struct obdo *oa, struct obd_trans_info *oti);
-int filter_setattr(const struct lu_env *env, struct obd_export *exp,
-                   struct obd_info *oinfo, struct obd_trans_info *oti);
-
-int filter_create(const struct lu_env *env, struct obd_export *exp,
-                  struct obdo *oa, struct lov_stripe_md **ea,
-                  struct obd_trans_info *oti);
-
-struct obd_llog_group *filter_find_olg(struct obd_device *obd, int seq);
-
-/* filter_lvb.c */
-extern struct ldlm_valblock_ops filter_lvbo;
-
-
-/* filter_io.c */
-int filter_preprw(const struct lu_env *env, int cmd, struct obd_export *,
-                  struct obdo *, int objcount,
-                  struct obd_ioobj *, struct niobuf_remote *,
-                  int *, struct niobuf_local *, struct obd_trans_info *,
-                  struct lustre_capa *);
-int filter_commitrw(const struct lu_env *, int cmd, struct obd_export *,
-                    struct obdo *, int objcount,
-                    struct obd_ioobj *, struct niobuf_remote *,  int,
-                    struct niobuf_local *, struct obd_trans_info *, int rc);
-void filter_release_cache(struct obd_device *, struct obd_ioobj *,
-                          struct niobuf_remote *, struct inode *);
-
-/* filter_io_*.c */
-struct filter_iobuf {
-       cfs_hlist_node_t        dr_hlist;
-       __u64                   dr_hkey;
-       /* number of reqs being processed */
-       cfs_atomic_t            dr_numreqs;
-       cfs_waitq_t             dr_wait;
-       int                     dr_max_pages;
-       int                     dr_npages;
-       int                     dr_error;
-       unsigned int            dr_ignore_quota:1;
-       struct page             **dr_pages;
-       unsigned long           *dr_blocks;
-       struct filter_obd       *dr_filter;
-};
-
-int filter_commitrw_write(struct obd_export *exp, struct obdo *oa, int objcount,
-                          struct obd_ioobj *obj, struct niobuf_remote *, int,
-                          struct niobuf_local *res, struct obd_trans_info *oti,
-                          int rc);
-obd_size filter_grant_space_left(struct obd_export *exp);
-long filter_grant(struct obd_export *exp, obd_size current_grant,
-                  obd_size want, obd_size fs_space_left, int conservative);
-void filter_grant_commit(struct obd_export *exp, int niocount,
-                         struct niobuf_local *res);
-void filter_grant_incoming(struct obd_export *exp, struct obdo *oa);
-struct filter_iobuf *filter_alloc_iobuf(struct filter_obd *, int rw,
-                                        int num_pages);
-void filter_free_iobuf(struct filter_iobuf *iobuf);
-int filter_iobuf_add_page(struct obd_device *obd, struct filter_iobuf *iobuf,
-                          struct inode *inode, struct page *page);
-void *filter_iobuf_get(struct filter_obd *filter, struct obd_trans_info *oti);
-void filter_iobuf_put(struct filter_obd *filter, struct filter_iobuf *iobuf,
-                      struct obd_trans_info *oti);
-int filter_direct_io(int rw, struct dentry *dchild, struct filter_iobuf *iobuf,
-                     struct obd_export *exp, struct iattr *attr,
-                     struct obd_trans_info *oti, void **wait_handle);
-int filter_clear_truncated_page(struct inode *inode);
-
-/* filter_log.c */
-struct ost_filterdata {
-        __u32  ofd_epoch;
-};
-int filter_log_sz_change(struct llog_handle *cathandle,
-                         struct ll_fid *mds_fid,
-                         __u32 ioepoch,
-                         struct llog_cookie *logcookie,
-                         struct inode *inode);
-//int filter_get_catalog(struct obd_device *);
-void filter_cancel_cookies_cb(struct obd_device *obd, __u64 transno,
-                              void *cb_data, int error);
-int filter_recov_log_mds_ost_cb(const struct lu_env *env,
-                               struct llog_handle *llh,
-                               struct llog_rec_hdr *rec, void *data);
-
-#ifdef LPROCFS
-void filter_tally(struct obd_export *exp, struct page **pages, int nr_pages,
-                  unsigned long *blocks, int blocks_per_page, int wr);
-int lproc_filter_attach_seqstat(struct obd_device *dev);
-void lprocfs_filter_init_vars(struct lprocfs_static_vars *lvars);
-void filter_stats_counter_init(struct lprocfs_stats *stats);
-#else
-static inline void filter_tally(struct obd_export *exp, struct page **pages,
-                                int nr_pages, unsigned long *blocks,
-                                int blocks_per_page, int wr) {}
-static inline int lproc_filter_attach_seqstat(struct obd_device *dev) {}
-static void lprocfs_filter_init_vars(struct lprocfs_static_vars *lvars)
-{
-        memset(lvars, 0, sizeof(*lvars));
-}
-static inline void filter_stats_counter_init(struct lprocfs_stats *stats) {}
-#endif
-
-int filter_update_capa_key(struct obd_device *obd, struct lustre_capa_key *key);
-int filter_auth_capa(struct obd_export *exp, struct lu_fid *fid, obd_seq seq,
-                     struct lustre_capa *capa, __u64 opc);
-int filter_capa_fixoa(struct obd_export *exp, struct obdo *oa, obd_seq seq,
-                      struct lustre_capa *capa);
-void filter_free_capa_keys(struct filter_obd *filter);
-
-void blacklist_add(uid_t uid);
-void blacklist_del(uid_t uid);
-int blacklist_display(char *buf, int bufsize);
-
-/* sync on lock cancel is useless when we force a journal flush,
- * and if we enable async journal commit, we should also turn on
- * sync on lock cancel if it is not enabled already. */
-static inline void filter_slc_set(struct filter_obd *filter)
-{
-        if (filter->fo_syncjournal == 1)
-                filter->fo_sync_lock_cancel = NEVER_SYNC_ON_CANCEL;
-        else if (filter->fo_sync_lock_cancel == NEVER_SYNC_ON_CANCEL)
-                filter->fo_sync_lock_cancel = ALWAYS_SYNC_ON_CANCEL;
-}
-
-#endif /* _FILTER_INTERNAL_H */
diff --git a/lustre/obdfilter/filter_io.c b/lustre/obdfilter/filter_io.c
deleted file mode 100644 (file)
index eefb0c6..0000000
+++ /dev/null
@@ -1,1036 +0,0 @@
-/*
- * GPL HEADER START
- *
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License version 2 only,
- * as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License version 2 for more details (a copy is included
- * in the LICENSE file that accompanied this code).
- *
- * You should have received a copy of the GNU General Public License
- * version 2 along with this program; If not, see
- * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
- *
- * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
- * CA 95054 USA or visit www.sun.com if you need additional information or
- * have any questions.
- *
- * GPL HEADER END
- */
-/*
- * Copyright (c) 2003, 2010, Oracle and/or its affiliates. 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.
- *
- * lustre/obdfilter/filter_io.c
- *
- * Author: Peter Braam <braam@clusterfs.com>
- * Author: Andreas Dilger <adilger@clusterfs.com>
- * Author: Phil Schwan <phil@clusterfs.com>
- */
-
-#define DEBUG_SUBSYSTEM S_FILTER
-
-#include <linux/module.h>
-#include <linux/pagemap.h> // XXX kill me soon
-#include <linux/version.h>
-
-#include <obd_class.h>
-#include <obd_ost.h>
-#include <lustre_fsfilt.h>
-#include "filter_internal.h"
-
-int *obdfilter_created_scratchpad;
-
-/* Grab the dirty and seen grant announcements from the incoming obdo.
- * We will later calculate the clients new grant and return it.
- * Caller must hold osfs lock */
-void filter_grant_incoming(struct obd_export *exp, struct obdo *oa)
-{
-        struct filter_export_data *fed;
-        struct obd_device *obd = exp->exp_obd;
-        ENTRY;
-
-        LASSERT_SPIN_LOCKED(&obd->obd_osfs_lock);
-
-        if ((oa->o_valid & (OBD_MD_FLBLOCKS|OBD_MD_FLGRANT)) !=
-                                        (OBD_MD_FLBLOCKS|OBD_MD_FLGRANT)) {
-                oa->o_valid &= ~OBD_MD_FLGRANT;
-                EXIT;
-                return;
-        }
-
-        fed = &exp->exp_filter_data;
-
-        /* Add some margin, since there is a small race if other RPCs arrive
-         * out-or-order and have already consumed some grant.  We want to
-         * leave this here in case there is a large error in accounting. */
-        CDEBUG(D_CACHE,
-               "%s: cli %s/%p reports grant: "LPU64" dropped: %u, local: %lu\n",
-               obd->obd_name, exp->exp_client_uuid.uuid, exp, oa->o_grant,
-               oa->o_dropped, fed->fed_grant);
-
-        /* Update our accounting now so that statfs takes it into account.
-         * Note that fed_dirty is only approximate and can become incorrect
-         * if RPCs arrive out-of-order.  No important calculations depend
-         * on fed_dirty however, but we must check sanity to not assert. */
-        if ((long long)oa->o_dirty < 0)
-                oa->o_dirty = 0;
-        else if (oa->o_dirty > fed->fed_grant + 4 * FILTER_GRANT_CHUNK)
-                oa->o_dirty = fed->fed_grant + 4 * FILTER_GRANT_CHUNK;
-        obd->u.filter.fo_tot_dirty += oa->o_dirty - fed->fed_dirty;
-        if (fed->fed_grant < oa->o_dropped) {
-                CDEBUG(D_CACHE,"%s: cli %s/%p reports %u dropped > grant %lu\n",
-                       obd->obd_name, exp->exp_client_uuid.uuid, exp,
-                       oa->o_dropped, fed->fed_grant);
-                oa->o_dropped = 0;
-        }
-        if (obd->u.filter.fo_tot_granted < oa->o_dropped) {
-                CERROR("%s: cli %s/%p reports %u dropped > tot_grant "LPU64"\n",
-                       obd->obd_name, exp->exp_client_uuid.uuid, exp,
-                       oa->o_dropped, obd->u.filter.fo_tot_granted);
-                oa->o_dropped = 0;
-        }
-        obd->u.filter.fo_tot_granted -= oa->o_dropped;
-        fed->fed_grant -= oa->o_dropped;
-        fed->fed_dirty = oa->o_dirty;
-
-        if (oa->o_valid & OBD_MD_FLFLAGS && oa->o_flags & OBD_FL_SHRINK_GRANT) {
-                obd_size left_space = filter_grant_space_left(exp);
-                struct filter_obd *filter = &exp->exp_obd->u.filter;
-
-                /*Only if left_space < fo_tot_clients * 32M,
-                 *then the grant space could be shrinked */
-                if (left_space < filter->fo_tot_granted_clients *
-                                 FILTER_GRANT_SHRINK_LIMIT) {
-                        fed->fed_grant -= oa->o_grant;
-                        filter->fo_tot_granted -= oa->o_grant;
-                        CDEBUG(D_CACHE, "%s: cli %s/%p shrink "LPU64
-                               "fed_grant %ld total "LPU64"\n",
-                               obd->obd_name, exp->exp_client_uuid.uuid,
-                               exp, oa->o_grant, fed->fed_grant,
-                               filter->fo_tot_granted);
-                        oa->o_grant = 0;
-                }
-        }
-
-        if (fed->fed_dirty < 0 || fed->fed_grant < 0 || fed->fed_pending < 0) {
-                CERROR("%s: cli %s/%p dirty %ld pend %ld grant %ld\n",
-                       obd->obd_name, exp->exp_client_uuid.uuid, exp,
-                       fed->fed_dirty, fed->fed_pending, fed->fed_grant);
-                cfs_spin_unlock(&obd->obd_osfs_lock);
-                LBUG();
-        }
-        EXIT;
-}
-
-/* Figure out how much space is available between what we've granted
- * and what remains in the filesystem.  Compensate for ext3 indirect
- * block overhead when computing how much free space is left ungranted.
- *
- * Caller must hold obd_osfs_lock. */
-obd_size filter_grant_space_left(struct obd_export *exp)
-{
-        struct obd_device *obd = exp->exp_obd;
-        int blockbits = obd->u.obt.obt_sb->s_blocksize_bits;
-        obd_size tot_granted = obd->u.filter.fo_tot_granted, avail, left = 0;
-        int rc, statfs_done = 0;
-
-        LASSERT_SPIN_LOCKED(&obd->obd_osfs_lock);
-
-        if (cfs_time_before_64(obd->obd_osfs_age,
-                               cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS))) {
-restat:
-                rc = fsfilt_statfs(obd, obd->u.obt.obt_sb,
-                                   cfs_time_shift_64(OBD_STATFS_CACHE_SECONDS));
-                if (rc) /* N.B. statfs can't really fail */
-                        RETURN(0);
-                statfs_done = 1;
-        }
-
-        avail = obd->obd_osfs.os_bavail;
-        left = avail - (avail >> (blockbits - 3)); /* (d)indirect */
-        if (left > GRANT_FOR_LLOG(obd)) {
-                left = (left - GRANT_FOR_LLOG(obd)) << blockbits;
-        } else {
-                left = 0 /* << blockbits */;
-        }
-
-        if (!statfs_done && left < 32 * FILTER_GRANT_CHUNK + tot_granted) {
-                CDEBUG(D_CACHE, "fs has no space left and statfs too old\n");
-                goto restat;
-        }
-
-        if (left >= tot_granted) {
-                left -= tot_granted;
-        } else {
-                if (left < tot_granted - obd->u.filter.fo_tot_pending) {
-                        CERROR("%s: cli %s/%p grant "LPU64" > available "
-                               LPU64" and pending "LPU64"\n", obd->obd_name,
-                               exp->exp_client_uuid.uuid, exp, tot_granted,
-                               left, obd->u.filter.fo_tot_pending);
-                }
-                left = 0;
-        }
-
-        CDEBUG(D_CACHE, "%s: cli %s/%p free: "LPU64" avail: "LPU64" grant "LPU64
-               " left: "LPU64" pending: "LPU64"\n", obd->obd_name,
-               exp->exp_client_uuid.uuid, exp,
-               obd->obd_osfs.os_bfree << blockbits, avail << blockbits,
-               tot_granted, left, obd->u.filter.fo_tot_pending);
-
-        return left;
-}
-
-/* Calculate how much grant space to allocate to this client, based on how
- * much space is currently free and how much of that is already granted.
- *
- * if @conservative != 0, we limit the maximum grant to FILTER_GRANT_CHUNK;
- * otherwise we'll satisfy the requested amount as possible as we can, this
- * is usually due to client reconnect.
- *
- * Caller must hold obd_osfs_lock. */
-long filter_grant(struct obd_export *exp, obd_size current_grant,
-                  obd_size want, obd_size fs_space_left, int conservative)
-{
-        struct obd_device *obd = exp->exp_obd;
-        struct filter_export_data *fed = &exp->exp_filter_data;
-        int blockbits = obd->u.obt.obt_sb->s_blocksize_bits;
-        __u64 grant = 0;
-
-        LASSERT_SPIN_LOCKED(&obd->obd_osfs_lock);
-
-        /* Grant some fraction of the client's requested grant space so that
-         * they are not always waiting for write credits (not all of it to
-         * avoid overgranting in face of multiple RPCs in flight).  This
-         * essentially will be able to control the OSC_MAX_RIF for a client.
-         *
-         * If we do have a large disparity between what the client thinks it
-         * has and what we think it has, don't grant very much and let the
-         * client consume its grant first.  Either it just has lots of RPCs
-         * in flight, or it was evicted and its grants will soon be used up. */
-        if (want > 0x7fffffff) {
-                CERROR("%s: client %s/%p requesting > 2GB grant "LPU64"\n",
-                       obd->obd_name, exp->exp_client_uuid.uuid, exp, want);
-        } else if (current_grant < want &&
-                   current_grant < fed->fed_grant + FILTER_GRANT_CHUNK) {
-                grant = min(want + (1 << blockbits) - 1, fs_space_left / 8);
-                grant &= ~((1ULL << blockbits) - 1);
-
-                if (grant) {
-                        if (grant > FILTER_GRANT_CHUNK && conservative)
-                                grant = FILTER_GRANT_CHUNK;
-
-                        obd->u.filter.fo_tot_granted += grant;
-                        fed->fed_grant += grant;
-                        if (fed->fed_grant < 0) {
-                                CERROR("%s: cli %s/%p grant %ld want "LPU64
-                                       "current"LPU64"\n",
-                                       obd->obd_name, exp->exp_client_uuid.uuid,
-                                       exp, fed->fed_grant, want,current_grant);
-                                cfs_spin_unlock(&obd->obd_osfs_lock);
-                                LBUG();
-                        }
-                }
-        }
-
-        CDEBUG(D_CACHE,
-               "%s: cli %s/%p wants: "LPU64" current grant "LPU64
-               " granting: "LPU64"\n", obd->obd_name, exp->exp_client_uuid.uuid,
-               exp, want, current_grant, grant);
-        CDEBUG(D_CACHE,
-               "%s: cli %s/%p tot cached:"LPU64" granted:"LPU64
-               " num_exports: %d\n", obd->obd_name, exp->exp_client_uuid.uuid,
-               exp, obd->u.filter.fo_tot_dirty,
-               obd->u.filter.fo_tot_granted, obd->obd_num_exports);
-
-        return grant;
-}
-
-/*
- * the routine is used to request pages from pagecache
- *
- * use GFP_NOFS for requests from a local client not allowing to enter FS
- * as we might end up waiting on a page he sent in the request we're serving.
- * use __GFP_HIGHMEM so that the pages can use all of the available memory
- * on 32-bit machines
- * use more aggressive GFP_HIGHUSER flags from non-local clients to be able to
- * generate more memory pressure.
- *
- * See Bug 19529 and Bug 19917 for details.
- */
-static struct page *filter_get_page(struct obd_device *obd,
-                                    struct inode *inode,
-                                    obd_off offset,
-                                    int localreq)
-{
-        struct page *page;
-
-        page = find_or_create_page(inode->i_mapping, offset >> CFS_PAGE_SHIFT,
-                                   (localreq ? (GFP_NOFS | __GFP_HIGHMEM)
-                                             : GFP_HIGHUSER));
-        if (unlikely(page == NULL))
-                lprocfs_counter_add(obd->obd_stats, LPROC_FILTER_NO_PAGE, 1);
-
-        return page;
-}
-
-/*
- * the routine initializes array of local_niobuf from remote_niobuf
- */
-static int filter_map_remote_to_local(int objcount, struct obd_ioobj *obj,
-                                      struct niobuf_remote *nb,
-                                      int *nrpages, struct niobuf_local *res)
-{
-        struct niobuf_remote *rnb;
-        struct niobuf_local *lnb;
-        int i, max;
-        ENTRY;
-
-        /* we don't support multiobject RPC yet
-         * ost_brw_read() and ost_brw_write() check this */
-        LASSERT(objcount == 1);
-
-        max = *nrpages;
-        *nrpages = 0;
-        for (i = 0, rnb = nb, lnb = res; i < obj->ioo_bufcnt; i++, rnb++) {
-                obd_off offset = rnb->offset;
-                unsigned int len = rnb->len;
-
-                while (len > 0) {
-                        int poff = offset & (CFS_PAGE_SIZE - 1);
-                        int plen = CFS_PAGE_SIZE - poff;
-
-                        if (*nrpages >= max) {
-                                CERROR("small array of local bufs: %d\n", max);
-                                RETURN(-EINVAL);
-                        }
-
-                        if (plen > len)
-                                plen = len;
-                       lnb->lnb_file_offset = offset;
-                       lnb->lnb_page_offset = poff;
-                        lnb->len = plen;
-                        lnb->flags = rnb->flags;
-                        lnb->page = NULL;
-                        lnb->rc = 0;
-                        lnb->lnb_grant_used = 0;
-
-                        LASSERTF(plen <= len, "plen %u, len %u\n", plen, len);
-                        offset += plen;
-                        len -= plen;
-                        lnb++;
-                        (*nrpages)++;
-                }
-        }
-        RETURN(0);
-}
-
-/*
- * the invalidate above doesn't work during read because lnet pins pages.
- * The truncate is used here instead to drop pages from cache
- */
-void filter_release_cache(struct obd_device *obd, struct obd_ioobj *obj,
-                          struct niobuf_remote *rnb, struct inode *inode)
-{
-        int i;
-
-        LASSERT(inode != NULL);
-        for (i = 0; i < obj->ioo_bufcnt; i++, rnb++) {
-                /* remove pages in which range is fit */
-                truncate_inode_pages_range(inode->i_mapping,
-                                           rnb->offset & CFS_PAGE_MASK,
-                                           (rnb->offset + rnb->len - 1) |
-                                           ~CFS_PAGE_MASK);
-        }
-}
-
-static int filter_preprw_read(int cmd, struct obd_export *exp, struct obdo *oa,
-                              int objcount, struct obd_ioobj *obj,
-                              struct niobuf_remote *nb,
-                              int *npages, struct niobuf_local *res,
-                              struct obd_trans_info *oti,
-                              struct lustre_capa *capa)
-{
-        struct obd_device *obd = exp->exp_obd;
-        struct timeval start, end;
-        struct lvfs_run_ctxt saved;
-        struct niobuf_local *lnb;
-        struct dentry *dentry = NULL;
-        struct inode *inode = NULL;
-        void *iobuf = NULL;
-        int rc = 0, i, tot_bytes = 0;
-        unsigned long now = jiffies;
-        long timediff;
-        loff_t isize;
-        ENTRY;
-
-        /* We are currently not supporting multi-obj BRW_READ RPCS at all.
-         * When we do this function's dentry cleanup will need to be fixed.
-         * These values are verified in ost_brw_write() from the wire. */
-        LASSERTF(objcount == 1, "%d\n", objcount);
-        LASSERTF(obj->ioo_bufcnt > 0, "%d\n", obj->ioo_bufcnt);
-
-        rc = filter_auth_capa(exp, NULL, oa->o_seq, capa,
-                              CAPA_OPC_OSS_READ);
-        if (rc)
-                RETURN(rc);
-
-        if (oa && oa->o_valid & OBD_MD_FLGRANT) {
-                cfs_spin_lock(&obd->obd_osfs_lock);
-                filter_grant_incoming(exp, oa);
-
-                if (!(oa->o_valid & OBD_MD_FLFLAGS) ||
-                    !(oa->o_flags & OBD_FL_SHRINK_GRANT))
-                        oa->o_grant = 0;
-                cfs_spin_unlock(&obd->obd_osfs_lock);
-        }
-
-        iobuf = filter_iobuf_get(&obd->u.filter, oti);
-        if (IS_ERR(iobuf))
-                RETURN(PTR_ERR(iobuf));
-
-        push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-        dentry = filter_oa2dentry(obd, &oa->o_oi);
-        if (IS_ERR(dentry)) {
-                rc = PTR_ERR(dentry);
-                dentry = NULL;
-                GOTO(cleanup, rc);
-        }
-
-        inode = dentry->d_inode;
-        isize = i_size_read(inode);
-
-        obdo_to_inode(inode, oa, OBD_MD_FLATIME);
-
-        rc = filter_map_remote_to_local(objcount, obj, nb, npages, res);
-        if (rc)
-                GOTO(cleanup, rc);
-
-        fsfilt_check_slow(obd, now, "preprw_read setup");
-
-        /* find pages for all segments, fill array with them */
-        cfs_gettimeofday(&start);
-        for (i = 0, lnb = res; i < *npages; i++, lnb++) {
-
-                lnb->dentry = dentry;
-
-               if (isize <= lnb->lnb_file_offset)
-                       /* If there's no more data, abort early.  lnb->rc == 0,
-                        * so it's easy to detect later. */
-                       break;
-
-               lnb->page = filter_get_page(obd, inode, lnb->lnb_file_offset,
-                                           0);
-               if (lnb->page == NULL)
-                       GOTO(cleanup, rc = -ENOMEM);
-
-               lprocfs_counter_add(obd->obd_stats, LPROC_FILTER_CACHE_ACCESS, 1);
-
-               if (isize < lnb->lnb_file_offset + lnb->len - 1)
-                       lnb->rc = isize - lnb->lnb_file_offset;
-                else
-                        lnb->rc = lnb->len;
-
-                tot_bytes += lnb->rc;
-
-                if (PageUptodate(lnb->page)) {
-                        lprocfs_counter_add(obd->obd_stats,
-                                            LPROC_FILTER_CACHE_HIT, 1);
-                        continue;
-                }
-
-                lprocfs_counter_add(obd->obd_stats, LPROC_FILTER_CACHE_MISS, 1);
-                filter_iobuf_add_page(obd, iobuf, inode, lnb->page);
-        }
-        cfs_gettimeofday(&end);
-        timediff = cfs_timeval_sub(&end, &start, NULL);
-        lprocfs_counter_add(obd->obd_stats, LPROC_FILTER_GET_PAGE, timediff);
-
-        if (OBD_FAIL_CHECK(OBD_FAIL_OST_NOMEM))
-                GOTO(cleanup, rc = -ENOMEM);
-
-        fsfilt_check_slow(obd, now, "start_page_read");
-
-        rc = filter_direct_io(OBD_BRW_READ, dentry, iobuf,
-                              exp, NULL, NULL, NULL);
-        if (rc)
-                GOTO(cleanup, rc);
-
-        lprocfs_counter_add(obd->obd_stats, LPROC_FILTER_READ_BYTES, tot_bytes);
-
-        if (exp->exp_nid_stats && exp->exp_nid_stats->nid_stats)
-                lprocfs_counter_add(exp->exp_nid_stats->nid_stats,
-                                    LPROC_FILTER_READ_BYTES, tot_bytes);
-       filter_counter_incr(exp, LPROC_FILTER_STATS_READ,
-                           oti ? oti->oti_jobid : NULL, tot_bytes);
-        EXIT;
-
- cleanup:
-        /* unlock pages to allow access from concurrent OST_READ */
-        for (i = 0, lnb = res; i < *npages; i++, lnb++) {
-                if (lnb->page) {
-                        LASSERT(PageLocked(lnb->page));
-                        unlock_page(lnb->page);
-
-                        if (rc) {
-                                page_cache_release(lnb->page);
-                                lnb->page = NULL;
-                        }
-                }
-        }
-
-        if (rc != 0) {
-                if (dentry != NULL)
-                        f_dput(dentry);
-        }
-
-        filter_iobuf_put(&obd->u.filter, iobuf, oti);
-
-        pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-        if (rc)
-                CERROR("io error %d\n", rc);
-
-        return rc;
-}
-
-/* When clients have dirtied as much space as they've been granted they
- * fall through to sync writes.  These sync writes haven't been expressed
- * in grants and need to error with ENOSPC when there isn't room in the
- * filesystem for them after grants are taken into account.  However,
- * writeback of the dirty data that was already granted space can write
- * right on through.
- *
- * Caller must hold obd_osfs_lock. */
-static int filter_grant_check(struct obd_export *exp, struct obdo *oa,
-                              int objcount, struct fsfilt_objinfo *fso,
-                              int niocount, struct niobuf_local *lnb,
-                              obd_size *left, struct inode *inode)
-{
-        struct filter_export_data *fed = &exp->exp_filter_data;
-        int blocksize = exp->exp_obd->u.obt.obt_sb->s_blocksize;
-        unsigned long used = 0, ungranted = 0, using;
-        int i, rc = -ENOSPC, obj, n = 0;
-        int resend = 0;
-
-        if ((oa->o_valid & OBD_MD_FLFLAGS) &&
-            (oa->o_flags & OBD_FL_RECOV_RESEND)) {
-                resend = 1;
-                CDEBUG(D_CACHE, "Recoverable resend arrived, skipping "
-                       "accounting\n");
-        }
-
-        LASSERT_SPIN_LOCKED(&exp->exp_obd->obd_osfs_lock);
-
-        for (obj = 0; obj < objcount; obj++) {
-                for (i = 0; i < fso[obj].fso_bufcnt; i++, n++) {
-                        int tmp, bytes;
-
-                        /* should match the code in osc_exit_cache */
-                        bytes = lnb[n].len;
-                       bytes += lnb[n].lnb_file_offset & (blocksize - 1);
-                       tmp = (lnb[n].lnb_file_offset + lnb[n].len) &
-                             (blocksize - 1);
-                        if (tmp)
-                                bytes += blocksize - tmp;
-
-                       lnb[n].rc = 0;
-                        if ((lnb[n].flags & OBD_BRW_FROM_GRANT) &&
-                            (oa->o_valid & OBD_MD_FLGRANT)) {
-                                if (resend) {
-                                        /* this is a recoverable resent */
-                                        lnb[n].flags |= OBD_BRW_GRANTED;
-                                        rc = 0;
-                                        continue;
-                                } else if (fed->fed_grant < used + bytes) {
-                                        CDEBUG(D_CACHE,
-                                               "%s: cli %s/%p claims %ld+%d "
-                                               "GRANT, real grant %lu idx %d\n",
-                                               exp->exp_obd->obd_name,
-                                               exp->exp_client_uuid.uuid, exp,
-                                               used, bytes, fed->fed_grant, n);
-                                } else {
-                                        used += bytes;
-                                        lnb[n].flags |= OBD_BRW_GRANTED;
-                                        lnb[n].lnb_grant_used = bytes;
-                                        CDEBUG(0, "idx %d used=%lu\n", n, used);
-                                        rc = 0;
-                                        continue;
-                                }
-                        }
-                        if (*left > ungranted + bytes) {
-                                /* if enough space, pretend it was granted */
-                                ungranted += bytes;
-                                lnb[n].flags |= OBD_BRW_GRANTED;
-                                lnb[n].lnb_grant_used = bytes;
-                                CDEBUG(0, "idx %d ungranted=%lu\n",n,ungranted);
-                                rc = 0;
-                                continue;
-                        }
-
-                        /* We can't check for already-mapped blocks here, as
-                         * it requires dropping the osfs lock to do the bmap.
-                         * Instead, we return ENOSPC and in that case we need
-                         * to go through and verify if all of the blocks not
-                         * marked BRW_GRANTED are already mapped and we can
-                         * ignore this error. */
-                        lnb[n].rc = -ENOSPC;
-                        lnb[n].flags &= ~OBD_BRW_GRANTED;
-                        CDEBUG(D_CACHE,"%s: cli %s/%p idx %d no space for %d\n",
-                               exp->exp_obd->obd_name,
-                               exp->exp_client_uuid.uuid, exp, n, bytes);
-                }
-        }
-
-        /* Now substract what client have used already.  We don't subtract
-         * this from the tot_granted yet, so that other client's can't grab
-         * that space before we have actually allocated our blocks.  That
-         * happens in filter_grant_commit() after the writes are done. */
-        *left -= ungranted;
-        fed->fed_grant -= used;
-        fed->fed_pending += used + ungranted;
-        exp->exp_obd->u.filter.fo_tot_granted += ungranted;
-        exp->exp_obd->u.filter.fo_tot_pending += used + ungranted;
-
-        CDEBUG(D_CACHE,
-               "%s: cli %s/%p used: %lu ungranted: %lu grant: %lu dirty: %lu\n",
-               exp->exp_obd->obd_name, exp->exp_client_uuid.uuid, exp, used,
-               ungranted, fed->fed_grant, fed->fed_dirty);
-
-        /* Rough calc in case we don't refresh cached statfs data */
-        using = (used + ungranted + 1 ) >>
-                exp->exp_obd->u.obt.obt_sb->s_blocksize_bits;
-        if (exp->exp_obd->obd_osfs.os_bavail > using)
-                exp->exp_obd->obd_osfs.os_bavail -= using;
-        else
-                exp->exp_obd->obd_osfs.os_bavail = 0;
-
-        if (fed->fed_dirty < used) {
-                CERROR("%s: cli %s/%p claims used %lu > fed_dirty %lu\n",
-                       exp->exp_obd->obd_name, exp->exp_client_uuid.uuid, exp,
-                       used, fed->fed_dirty);
-                used = fed->fed_dirty;
-        }
-        exp->exp_obd->u.filter.fo_tot_dirty -= used;
-        fed->fed_dirty -= used;
-
-        if (fed->fed_dirty < 0 || fed->fed_grant < 0 || fed->fed_pending < 0) {
-                CERROR("%s: cli %s/%p dirty %ld pend %ld grant %ld\n",
-                       exp->exp_obd->obd_name, exp->exp_client_uuid.uuid, exp,
-                       fed->fed_dirty, fed->fed_pending, fed->fed_grant);
-                cfs_spin_unlock(&exp->exp_obd->obd_osfs_lock);
-                LBUG();
-        }
-        return rc;
-}
-
-/* If we ever start to support multi-object BRW RPCs, we will need to get locks
- * on mulitple inodes.  That isn't all, because there still exists the
- * possibility of a truncate starting a new transaction while holding the ext3
- * rwsem = write while some writes (which have started their transactions here)
- * blocking on the ext3 rwsem = read => lock inversion.
- *
- * The handling gets very ugly when dealing with locked pages.  It may be easier
- * to just get rid of the locked page code (which has problems of its own) and
- * either discover we do not need it anymore (i.e. it was a symptom of another
- * bug) or ensure we get the page locks in an appropriate order. */
-static int filter_preprw_write(int cmd, struct obd_export *exp, struct obdo *oa,
-                               int objcount, struct obd_ioobj *obj,
-                               struct niobuf_remote *nb, int *npages,
-                               struct niobuf_local *res,
-                               struct obd_trans_info *oti,
-                               struct lustre_capa *capa)
-{
-        struct obd_device *obd = exp->exp_obd;
-        struct timeval start, end;
-        struct lvfs_run_ctxt saved;
-        struct niobuf_local *lnb = res;
-        struct fsfilt_objinfo fso;
-        struct filter_mod_data *fmd;
-        struct dentry *dentry = NULL;
-        void *iobuf;
-        obd_size left;
-        unsigned long now = jiffies, timediff;
-        int rc = 0, i, tot_bytes = 0, cleanup_phase = 0, localreq = 0;
-       int retries = 0;
-        ENTRY;
-        LASSERT(objcount == 1);
-        LASSERT(obj->ioo_bufcnt > 0);
-
-        rc = filter_auth_capa(exp, NULL, oa->o_seq, capa,
-                              CAPA_OPC_OSS_WRITE);
-        if (rc)
-                RETURN(rc);
-
-        if (exp->exp_connection &&
-            exp->exp_connection->c_peer.nid == exp->exp_connection->c_self)
-                localreq = 1;
-
-        push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-        iobuf = filter_iobuf_get(&obd->u.filter, oti);
-        if (IS_ERR(iobuf))
-                GOTO(cleanup, rc = PTR_ERR(iobuf));
-        cleanup_phase = 1;
-
-        dentry = filter_fid2dentry(obd, NULL, obj->ioo_seq,
-                                   obj->ioo_id);
-        if (IS_ERR(dentry))
-                GOTO(cleanup, rc = PTR_ERR(dentry));
-        cleanup_phase = 2;
-
-        if (dentry->d_inode == NULL) {
-                if (exp->exp_obd->obd_recovering) {
-                        struct obdo *noa = oa;
-
-                        if (oa == NULL) {
-                                OBDO_ALLOC(noa);
-                                if (noa == NULL)
-                                        GOTO(recreate_out, rc = -ENOMEM);
-                                noa->o_id = obj->ioo_id;
-                                noa->o_valid = OBD_MD_FLID;
-                        }
-
-                        if (filter_create(NULL, exp, noa, NULL, oti) == 0) {
-                                f_dput(dentry);
-                                dentry = filter_fid2dentry(exp->exp_obd, NULL,
-                                                           obj->ioo_seq,
-                                                           obj->ioo_id);
-                        }
-                        if (oa == NULL)
-                                OBDO_FREE(noa);
-                }
-    recreate_out:
-                if (IS_ERR(dentry) || dentry->d_inode == NULL) {
-                        CERROR("%s: BRW to missing obj "LPU64"/"LPU64":rc %d\n",
-                               exp->exp_obd->obd_name,
-                               obj->ioo_id, obj->ioo_seq,
-                               IS_ERR(dentry) ? (int)PTR_ERR(dentry) : -ENOENT);
-                        if (IS_ERR(dentry))
-                                cleanup_phase = 1;
-                        GOTO(cleanup, rc = -ENOENT);
-                }
-        }
-
-        if (oa->o_valid & (OBD_MD_FLUID | OBD_MD_FLGID) &&
-            dentry->d_inode->i_mode & (S_ISUID | S_ISGID)) {
-                rc = filter_capa_fixoa(exp, oa, oa->o_seq, capa);
-                if (rc)
-                        GOTO(cleanup, rc);
-        }
-
-        rc = filter_map_remote_to_local(objcount, obj, nb, npages, res);
-        if (rc)
-                GOTO(cleanup, rc);
-
-        fsfilt_check_slow(obd, now, "preprw_write setup");
-
-        /* Filter truncate first locks i_mutex then partially truncated
-         * page, filter write code first locks pages then take
-         * i_mutex.  To avoid a deadlock in case of concurrent
-         * punch/write requests from one client, filter writes and
-         * filter truncates are serialized by i_alloc_sem, allowing
-         * multiple writes or single truncate. */
-        down_read(&dentry->d_inode->i_alloc_sem);
-        fsfilt_check_slow(obd, now, "i_alloc_sem");
-
-        /* Don't update inode timestamps if this write is older than a
-         * setattr which modifies the timestamps. b=10150 */
-        /* XXX when we start having persistent reservations this needs to
-         * be changed to filter_fmd_get() to create the fmd if it doesn't
-         * already exist so we can store the reservation handle there. */
-        fmd = filter_fmd_find(exp, obj->ioo_id, obj->ioo_seq);
-
-       LASSERT(oa != NULL);
-retry:
-       cfs_spin_lock(&obd->obd_osfs_lock);
-       if (retries == 0)
-               filter_grant_incoming(exp, oa);
-       if (fmd && fmd->fmd_mactime_xid > oti->oti_xid)
-               oa->o_valid &= ~(OBD_MD_FLMTIME | OBD_MD_FLCTIME |
-                                OBD_MD_FLATIME);
-       else
-               obdo_to_inode(dentry->d_inode, oa, OBD_MD_FLATIME |
-                             OBD_MD_FLMTIME | OBD_MD_FLCTIME);
-       cleanup_phase = 3;
-
-       left = filter_grant_space_left(exp);
-
-       fso.fso_dentry = dentry;
-       fso.fso_bufcnt = *npages;
-
-       rc = filter_grant_check(exp, oa, objcount, &fso, *npages, res,
-                               &left, dentry->d_inode);
-
-       /* do not zero out oa->o_valid as it is used in filter_commitrw_write()
-        * for setting UID/GID and fid EA in first write time. */
-       /* If OBD_FL_SHRINK_GRANT is set, the client just returned us some grant
-        * so no sense in allocating it some more. We either return the grant
-        * back to the client if we have plenty of space or we don't return
-        * anything if we are short. This was decided in filter_grant_incoming*/
-       if ((retries == 0) && (oa->o_valid & OBD_MD_FLGRANT) &&
-           (!(oa->o_valid & OBD_MD_FLFLAGS) ||
-            !(oa->o_flags & OBD_FL_SHRINK_GRANT)))
-               oa->o_grant = filter_grant(exp, oa->o_grant, oa->o_undirty,
-                                          left, 1);
-
-       cfs_spin_unlock(&obd->obd_osfs_lock);
-
-       if (rc == -ENOSPC && retries == 0) {
-               void *handle = NULL;
-
-               CDEBUG(D_INODE, "retry after commit pending journals");
-
-               retries = 1;
-               handle = fsfilt_start(obd, dentry->d_inode,
-                                     FSFILT_OP_SETATTR, NULL);
-               if (handle != NULL &&
-                   fsfilt_commit(obd, dentry->d_inode, handle, 1) == 0)
-                       goto retry;
-       }
-
-        filter_fmd_put(exp, fmd);
-
-        OBD_FAIL_TIMEOUT(OBD_FAIL_OST_BRW_PAUSE_BULK2, (obd_timeout + 1) / 4);
-
-        if (rc)
-                GOTO(cleanup, rc);
-        cleanup_phase = 4;
-
-        cfs_gettimeofday(&start);
-        for (i = 0, lnb = res; i < *npages; i++, lnb++) {
-
-                /* We still set up for ungranted pages so that granted pages
-                 * can be written to disk as they were promised, and portals
-                 * needs to keep the pages all aligned properly. */
-                lnb->dentry = dentry;
-
-               lnb->page = filter_get_page(obd, dentry->d_inode,
-                                           lnb->lnb_file_offset, localreq);
-                if (lnb->page == NULL)
-                        GOTO(cleanup, rc = -ENOMEM);
-
-                /* DLM locking protects us from write and truncate competing
-                 * for same region, but truncate can leave dirty page in the
-                 * cache. it's possible the writeout on a such a page is in
-                 * progress when we access it. it's also possible that during
-                 * this writeout we put new (partial) data, but then won't
-                 * be able to proceed in filter_commitrw_write(). thus let's
-                 * just wait for writeout completion, should be rare enough.
-                 * -bzzz */
-                wait_on_page_writeback(lnb->page);
-                BUG_ON(PageWriteback(lnb->page));
-
-                /* If the filter writes a partial page, then has the file
-                 * extended, the client will read in the whole page.  the
-                 * filter has to be careful to zero the rest of the partial
-                 * page on disk.  we do it by hand for partial extending
-                 * writes, send_bio() is responsible for zeroing pages when
-                 * asked to read unmapped blocks -- brw_kiovec() does this. */
-                if (lnb->len != CFS_PAGE_SIZE) {
-                        __s64 maxidx;
-
-                        maxidx = ((i_size_read(dentry->d_inode) +
-                                   CFS_PAGE_SIZE - 1) >> CFS_PAGE_SHIFT) - 1;
-                        if (maxidx >= lnb->page->index) {
-                                LL_CDEBUG_PAGE(D_PAGE, lnb->page, "write %u @ "
-                                               LPU64" flg %x before EOF %llu\n",
-                                              lnb->len, lnb->lnb_file_offset,
-                                              lnb->flags,
-                                               i_size_read(dentry->d_inode));
-                                filter_iobuf_add_page(obd, iobuf,
-                                                      dentry->d_inode,
-                                                      lnb->page);
-                        } else {
-                                long off;
-                                char *p = kmap(lnb->page);
-
-                                off = lnb->lnb_file_offset & ~CFS_PAGE_MASK;
-                                if (off)
-                                        memset(p, 0, off);
-                               off = (lnb->lnb_file_offset + lnb->len) &
-                                     ~CFS_PAGE_MASK;
-                                if (off)
-                                        memset(p + off, 0, CFS_PAGE_SIZE - off);
-                                kunmap(lnb->page);
-                        }
-                }
-                if (lnb->rc == 0)
-                        tot_bytes += lnb->len;
-        }
-        cfs_gettimeofday(&end);
-        timediff = cfs_timeval_sub(&end, &start, NULL);
-        lprocfs_counter_add(obd->obd_stats, LPROC_FILTER_GET_PAGE, timediff);
-
-        if (OBD_FAIL_CHECK(OBD_FAIL_OST_NOMEM))
-                GOTO(cleanup, rc = -ENOMEM);
-
-        /* don't unlock pages to prevent any access */
-        rc = filter_direct_io(OBD_BRW_READ, dentry, iobuf, exp,
-                              NULL, NULL, NULL);
-
-        fsfilt_check_slow(obd, now, "start_page_write");
-
-        lprocfs_counter_add(obd->obd_stats, LPROC_FILTER_WRITE_BYTES,
-                            tot_bytes);
-
-        if (exp->exp_nid_stats && exp->exp_nid_stats->nid_stats)
-                lprocfs_counter_add(exp->exp_nid_stats->nid_stats,
-                                    LPROC_FILTER_WRITE_BYTES, tot_bytes);
-       filter_counter_incr(exp, LPROC_FILTER_STATS_WRITE,
-                           oti ? oti->oti_jobid : NULL, tot_bytes);
-        EXIT;
-cleanup:
-        switch(cleanup_phase) {
-        case 4:
-                if (rc) {
-                        for (i = 0, lnb = res; i < *npages; i++, lnb++) {
-                                if (lnb->page != NULL) {
-                                        unlock_page(lnb->page);
-                                        page_cache_release(lnb->page);
-                                        lnb->page = NULL;
-                                }
-                        }
-                }
-        case 3:
-                if (rc)
-                        up_read(&dentry->d_inode->i_alloc_sem);
-
-                filter_iobuf_put(&obd->u.filter, iobuf, oti);
-        case 2:
-                pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-                if (rc)
-                        f_dput(dentry);
-                break;
-        case 1:
-                filter_iobuf_put(&obd->u.filter, iobuf, oti);
-        case 0:
-                cfs_spin_lock(&obd->obd_osfs_lock);
-                if (oa)
-                        filter_grant_incoming(exp, oa);
-                cfs_spin_unlock(&obd->obd_osfs_lock);
-                pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-                break;
-        default:;
-        }
-        return rc;
-}
-
-int filter_preprw(const struct lu_env *env, int cmd, struct obd_export *exp,
-                  struct obdo *oa, int objcount, struct obd_ioobj *obj,
-                  struct niobuf_remote *nb, int *npages,
-                  struct niobuf_local *res, struct obd_trans_info *oti,
-                  struct lustre_capa *capa)
-{
-        if (cmd == OBD_BRW_WRITE)
-                return filter_preprw_write(cmd, exp, oa, objcount, obj,
-                                           nb, npages, res, oti, capa);
-        if (cmd == OBD_BRW_READ)
-                return filter_preprw_read(cmd, exp, oa, objcount, obj,
-                                          nb, npages, res, oti, capa);
-        LBUG();
-        return -EPROTO;
-}
-
-static int filter_commitrw_read(struct obd_export *exp, struct obdo *oa,
-                                int objcount, struct obd_ioobj *obj,
-                                struct niobuf_remote *rnb,
-                                int npages, struct niobuf_local *res,
-                                struct obd_trans_info *oti, int rc)
-{
-        struct filter_obd *fo = &exp->exp_obd->u.filter;
-        struct inode *inode = NULL;
-        struct ldlm_res_id res_id;
-        struct ldlm_resource *resource = NULL;
-        struct ldlm_namespace *ns = exp->exp_obd->obd_namespace;
-        struct niobuf_local *lnb;
-        int i;
-        ENTRY;
-
-        osc_build_res_name(obj->ioo_id, obj->ioo_seq, &res_id);
-        /* If oa != NULL then filter_preprw_read updated the inode atime
-         * and we should update the lvb so that other glimpses will also
-         * get the updated value. bug 5972 */
-        if (oa && ns && ns->ns_lvbo && ns->ns_lvbo->lvbo_update) {
-                resource = ldlm_resource_get(ns, NULL, &res_id, LDLM_EXTENT, 0);
-
-                if (resource != NULL) {
-                        LDLM_RESOURCE_ADDREF(resource);
-                        ns->ns_lvbo->lvbo_update(resource, NULL, 1);
-                        LDLM_RESOURCE_DELREF(resource);
-                        ldlm_resource_putref(resource);
-                }
-        }
-
-        if (res->dentry != NULL)
-                inode = res->dentry->d_inode;
-
-        for (i = 0, lnb = res; i < npages; i++, lnb++) {
-                if (lnb->page != NULL) {
-                        page_cache_release(lnb->page);
-                        lnb->page = NULL;
-                }
-        }
-        if (inode && (fo->fo_read_cache == 0 ||
-                      i_size_read(inode) > fo->fo_readcache_max_filesize))
-                filter_release_cache(exp->exp_obd, obj, rnb, inode);
-
-        if (res->dentry != NULL)
-                f_dput(res->dentry);
-        RETURN(rc);
-}
-
-void filter_grant_commit(struct obd_export *exp, int niocount,
-                         struct niobuf_local *res)
-{
-        struct filter_obd *filter = &exp->exp_obd->u.filter;
-        struct niobuf_local *lnb = res;
-        unsigned long pending = 0;
-        int i;
-
-        cfs_spin_lock(&exp->exp_obd->obd_osfs_lock);
-        for (i = 0, lnb = res; i < niocount; i++, lnb++)
-                pending += lnb->lnb_grant_used;
-
-        LASSERTF(exp->exp_filter_data.fed_pending >= pending,
-                 "%s: cli %s/%p fed_pending: %lu grant_used: %lu\n",
-                 exp->exp_obd->obd_name, exp->exp_client_uuid.uuid, exp,
-                 exp->exp_filter_data.fed_pending, pending);
-        exp->exp_filter_data.fed_pending -= pending;
-        LASSERTF(filter->fo_tot_granted >= pending,
-                 "%s: cli %s/%p tot_granted: "LPU64" grant_used: %lu\n",
-                 exp->exp_obd->obd_name, exp->exp_client_uuid.uuid, exp,
-                 exp->exp_obd->u.filter.fo_tot_granted, pending);
-        filter->fo_tot_granted -= pending;
-        LASSERTF(filter->fo_tot_pending >= pending,
-                 "%s: cli %s/%p tot_pending: "LPU64" grant_used: %lu\n",
-                 exp->exp_obd->obd_name, exp->exp_client_uuid.uuid, exp,
-                 filter->fo_tot_pending, pending);
-        filter->fo_tot_pending -= pending;
-
-        cfs_spin_unlock(&exp->exp_obd->obd_osfs_lock);
-}
-
-int filter_commitrw(const struct lu_env *env, int cmd, struct obd_export *exp,
-                    struct obdo *oa, int objcount, struct obd_ioobj *obj,
-                    struct niobuf_remote *nb, int npages,
-                    struct niobuf_local *res, struct obd_trans_info *oti,
-                    int rc)
-{
-        if (cmd == OBD_BRW_WRITE)
-                return filter_commitrw_write(exp, oa, objcount, obj,
-                                             nb, npages, res, oti, rc);
-        if (cmd == OBD_BRW_READ)
-                return filter_commitrw_read(exp, oa, objcount, obj,
-                                            nb, npages, res, oti, rc);
-        LBUG();
-        return -EPROTO;
-}
diff --git a/lustre/obdfilter/filter_io_26.c b/lustre/obdfilter/filter_io_26.c
deleted file mode 100644 (file)
index f2becd9..0000000
+++ /dev/null
@@ -1,837 +0,0 @@
-/*
- * GPL HEADER START
- *
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License version 2 only,
- * as published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License version 2 for more details (a copy is included
- * in the LICENSE file that accompanied this code).
- *
- * You should have received a copy of the GNU General Public License
- * version 2 along with this program; If not, see
- * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
- *
- * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
- * CA 95054 USA or visit www.sun.com if you need additional information or
- * have any questions.
- *
- * GPL HEADER END
- */
-/*
- * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
- * Use is subject to license terms.
- *
- * Copyright (c) 2011, 2012, Whamcloud, Inc.
- */
-/*
- * This file is part of Lustre, http://www.lustre.org/
- * Lustre is a trademark of Sun Microsystems, Inc.
- *
- * lustre/obdfilter/filter_io_26.c
- *
- * Author: Peter Braam <braam@clusterfs.com>
- * Author: Andreas Dilger <adilger@clusterfs.com>
- * Author: Phil Schwan <phil@clusterfs.com>
- */
-
-#include <linux/module.h>
-#include <linux/pagemap.h> // XXX kill me soon
-#include <linux/version.h>
-#include <linux/buffer_head.h>
-
-#define DEBUG_SUBSYSTEM S_FILTER
-
-#include <obd_class.h>
-#include <lustre_fsfilt.h>
-#include "filter_internal.h"
-
-/* 512byte block min */
-#define MAX_BLOCKS_PER_PAGE (CFS_PAGE_SIZE / 512)
-
-static void record_start_io(struct filter_iobuf *iobuf, int rw, int size,
-                            struct obd_export *exp)
-{
-        struct filter_obd *filter = iobuf->dr_filter;
-
-        cfs_atomic_inc(&iobuf->dr_numreqs);
-
-        if (rw == OBD_BRW_READ) {
-                cfs_atomic_inc(&filter->fo_r_in_flight);
-                lprocfs_oh_tally(&filter->fo_filter_stats.hist[BRW_R_RPC_HIST],
-                                 cfs_atomic_read(&filter->fo_r_in_flight));
-                lprocfs_oh_tally_log2(&filter->
-                                       fo_filter_stats.hist[BRW_R_DISK_IOSIZE],
-                                      size);
-                if (exp->exp_nid_stats && exp->exp_nid_stats->nid_brw_stats) {
-                        lprocfs_oh_tally(&exp->exp_nid_stats->nid_brw_stats->
-                                         hist[BRW_R_RPC_HIST],
-                                         cfs_atomic_read(&filter-> \
-                                         fo_r_in_flight));
-                        lprocfs_oh_tally_log2(&exp->exp_nid_stats->
-                                         nid_brw_stats->hist[BRW_R_DISK_IOSIZE],
-                                              size);
-                }
-        } else {
-                cfs_atomic_inc(&filter->fo_w_in_flight);
-                lprocfs_oh_tally(&filter->fo_filter_stats.hist[BRW_W_RPC_HIST],
-                                 cfs_atomic_read(&filter->fo_w_in_flight));
-                lprocfs_oh_tally_log2(&filter->
-                                       fo_filter_stats.hist[BRW_W_DISK_IOSIZE],
-                                      size);
-                if (exp->exp_nid_stats && exp->exp_nid_stats->nid_brw_stats) {
-                        lprocfs_oh_tally(&exp->exp_nid_stats->nid_brw_stats->
-                                          hist[BRW_W_RPC_HIST],
-                                         cfs_atomic_read(&filter-> \
-                                         fo_w_in_flight));
-                        lprocfs_oh_tally_log2(&exp->exp_nid_stats->
-                                        nid_brw_stats->hist[BRW_W_DISK_IOSIZE],
-                                              size);
-                }
-        }
-}
-
-static void record_finish_io(struct filter_iobuf *iobuf, int rw, int rc)
-{
-        struct filter_obd *filter = iobuf->dr_filter;
-
-        /* CAVEAT EMPTOR: possibly in IRQ context
-         * DO NOT record procfs stats here!!! */
-
-        if (rw == OBD_BRW_READ)
-                cfs_atomic_dec(&filter->fo_r_in_flight);
-        else
-                cfs_atomic_dec(&filter->fo_w_in_flight);
-
-        if (cfs_atomic_dec_and_test(&iobuf->dr_numreqs))
-                cfs_waitq_signal(&iobuf->dr_wait);
-}
-
-#ifdef HAVE_BIO_ENDIO_2ARG
-#define DIO_RETURN(a)
-static void dio_complete_routine(struct bio *bio, int error)
-#else
-#define DIO_RETURN(a)   return(a)
-static int dio_complete_routine(struct bio *bio, unsigned int done, int error)
-#endif
-{
-        struct filter_iobuf *iobuf = bio->bi_private;
-        struct bio_vec *bvl;
-        int i;
-
-        /* CAVEAT EMPTOR: possibly in IRQ context
-         * DO NOT record procfs stats here!!! */
-
-#ifdef HAVE_BIO_ENDIO_2ARG
-       /* The "bi_size" check was needed for kernels < 2.6.24 in order to
-        * handle the case where a SCSI request error caused this callback
-        * to be called before all of the biovecs had been processed.
-        * Without this check the server thread will hang.  In newer kernels
-        * the bio_end_io routine is never called for partial completions,
-        * so this check is no longer needed. */
-        if (bio->bi_size)                      /* Not complete */
-                DIO_RETURN(1);
-#endif
-
-        if (unlikely(iobuf == NULL)) {
-                CERROR("***** bio->bi_private is NULL!  This should never "
-                       "happen.  Normally, I would crash here, but instead I "
-                       "will dump the bio contents to the console.  Please "
-                       "report this to <http://bugs.whamcloud.com/>, along "
-                       "with any interesting messages leading up to this point "
-                       "(like SCSI errors, perhaps).  Because bi_private is "
-                       "NULL, I can't wake up the thread that initiated this "
-                       "IO - you will probably have to reboot this node.\n");
-                CERROR("bi_next: %p, bi_flags: %lx, bi_rw: %lu, bi_vcnt: %d, "
-                       "bi_idx: %d, bi->size: %d, bi_end_io: %p, bi_cnt: %d, "
-                       "bi_private: %p\n", bio->bi_next, bio->bi_flags,
-                       bio->bi_rw, bio->bi_vcnt, bio->bi_idx, bio->bi_size,
-                       bio->bi_end_io, cfs_atomic_read(&bio->bi_cnt),
-                       bio->bi_private);
-                DIO_RETURN(0);
-        }
-
-        /* the check is outside of the cycle for performance reason -bzzz */
-        if (!cfs_test_bit(BIO_RW, &bio->bi_rw)) {
-                bio_for_each_segment(bvl, bio, i) {
-                        if (likely(error == 0))
-                                SetPageUptodate(bvl->bv_page);
-                        LASSERT(PageLocked(bvl->bv_page));
-                        ClearPageConstant(bvl->bv_page);
-                }
-                record_finish_io(iobuf, OBD_BRW_READ, error);
-        } else {
-                if (mapping_cap_page_constant_write(iobuf->dr_pages[0]->mapping)){
-                        bio_for_each_segment(bvl, bio, i) {
-                                ClearPageConstant(bvl->bv_page);
-                        }
-                }
-                record_finish_io(iobuf, OBD_BRW_WRITE, error);
-        }
-
-        /* any real error is good enough -bzzz */
-        if (error != 0 && iobuf->dr_error == 0)
-                iobuf->dr_error = error;
-
-        /* Completed bios used to be chained off iobuf->dr_bios and freed in
-         * filter_clear_dreq().  It was then possible to exhaust the biovec-256
-         * mempool when serious on-disk fragmentation was encountered,
-         * deadlocking the OST.  The bios are now released as soon as complete
-         * so the pool cannot be exhausted while IOs are competing. bug 10076 */
-        bio_put(bio);
-        DIO_RETURN(0);
-}
-
-static int can_be_merged(struct bio *bio, sector_t sector)
-{
-        unsigned int size;
-
-        if (!bio)
-                return 0;
-
-        size = bio->bi_size >> 9;
-        return bio->bi_sector + size == sector ? 1 : 0;
-}
-
-struct filter_iobuf *filter_alloc_iobuf(struct filter_obd *filter,
-                                        int rw, int num_pages)
-{
-        struct filter_iobuf *iobuf;
-
-        LASSERTF(rw == OBD_BRW_WRITE || rw == OBD_BRW_READ, "%x\n", rw);
-
-        OBD_ALLOC(iobuf, sizeof(*iobuf));
-        if (iobuf == NULL)
-                goto failed_0;
-
-        OBD_ALLOC(iobuf->dr_pages, num_pages * sizeof(*iobuf->dr_pages));
-        if (iobuf->dr_pages == NULL)
-                goto failed_1;
-
-        OBD_ALLOC(iobuf->dr_blocks,
-                  MAX_BLOCKS_PER_PAGE * num_pages * sizeof(*iobuf->dr_blocks));
-        if (iobuf->dr_blocks == NULL)
-                goto failed_2;
-
-       CFS_INIT_HLIST_NODE(&iobuf->dr_hlist);
-        iobuf->dr_filter = filter;
-        cfs_waitq_init(&iobuf->dr_wait);
-        cfs_atomic_set(&iobuf->dr_numreqs, 0);
-        iobuf->dr_max_pages = num_pages;
-        iobuf->dr_npages = 0;
-        iobuf->dr_error = 0;
-
-        RETURN(iobuf);
-
- failed_2:
-        OBD_FREE(iobuf->dr_pages,
-                 num_pages * sizeof(*iobuf->dr_pages));
- failed_1:
-        OBD_FREE(iobuf, sizeof(*iobuf));
- failed_0:
-        RETURN(ERR_PTR(-ENOMEM));
-}
-
-static void filter_clear_iobuf(struct filter_iobuf *iobuf)
-{
-        iobuf->dr_npages = 0;
-        iobuf->dr_error = 0;
-        cfs_atomic_set(&iobuf->dr_numreqs, 0);
-}
-
-void filter_free_iobuf(struct filter_iobuf *iobuf)
-{
-       int num_pages = iobuf->dr_max_pages;
-
-       filter_clear_iobuf(iobuf);
-
-       LASSERT(cfs_hlist_unhashed(&iobuf->dr_hlist));
-        OBD_FREE(iobuf->dr_blocks,
-                 MAX_BLOCKS_PER_PAGE * num_pages * sizeof(*iobuf->dr_blocks));
-        OBD_FREE(iobuf->dr_pages,
-                 num_pages * sizeof(*iobuf->dr_pages));
-        OBD_FREE_PTR(iobuf);
-}
-
-void filter_iobuf_put(struct filter_obd *filter, struct filter_iobuf *iobuf,
-                      struct obd_trans_info *oti)
-{
-        int thread_id = (oti && oti->oti_thread) ?
-                        oti->oti_thread->t_id : -1;
-
-        if (unlikely(thread_id < 0)) {
-                filter_free_iobuf(iobuf);
-                return;
-        }
-
-        filter_clear_iobuf(iobuf);
-}
-
-int filter_iobuf_add_page(struct obd_device *obd, struct filter_iobuf *iobuf,
-                          struct inode *inode, struct page *page)
-{
-        LASSERT(iobuf->dr_npages < iobuf->dr_max_pages);
-        iobuf->dr_pages[iobuf->dr_npages++] = page;
-
-        return 0;
-}
-
-int filter_do_bio(struct obd_export *exp, struct inode *inode,
-                  struct filter_iobuf *iobuf, int rw)
-{
-        struct obd_device *obd = exp->exp_obd;
-        int            blocks_per_page = CFS_PAGE_SIZE >> inode->i_blkbits;
-        struct page  **pages = iobuf->dr_pages;
-        int            npages = iobuf->dr_npages;
-        unsigned long *blocks = iobuf->dr_blocks;
-        int            total_blocks = npages * blocks_per_page;
-        int            sector_bits = inode->i_sb->s_blocksize_bits - 9;
-        unsigned int   blocksize = inode->i_sb->s_blocksize;
-        struct bio    *bio = NULL;
-        int            frags = 0;
-        unsigned long  start_time = jiffies;
-        struct page   *page;
-        unsigned int   page_offset;
-        sector_t       sector;
-        int            nblocks;
-        int            block_idx;
-        int            page_idx;
-        int            i;
-        int            rc = 0;
-        ENTRY;
-
-        LASSERT(iobuf->dr_npages == npages);
-        LASSERT(total_blocks <= OBDFILTER_CREATED_SCRATCHPAD_ENTRIES);
-
-        for (page_idx = 0, block_idx = 0;
-             page_idx < npages;
-             page_idx++, block_idx += blocks_per_page) {
-
-                page = pages[page_idx];
-                LASSERT (block_idx + blocks_per_page <= total_blocks);
-
-                for (i = 0, page_offset = 0;
-                     i < blocks_per_page;
-                     i += nblocks, page_offset += blocksize * nblocks) {
-
-                        nblocks = 1;
-
-                        if (blocks[block_idx + i] == 0) {  /* hole */
-                                LASSERT(rw == OBD_BRW_READ);
-                                memset(kmap(page) + page_offset, 0, blocksize);
-                                kunmap(page);
-                                continue;
-                        }
-
-                        sector = (sector_t)blocks[block_idx + i] << sector_bits;
-
-                        /* Additional contiguous file blocks? */
-                        while (i + nblocks < blocks_per_page &&
-                               (sector + (nblocks << sector_bits)) ==
-                               ((sector_t)blocks[block_idx + i + nblocks] <<
-                                sector_bits))
-                                nblocks++;
-
-                        /* I only set the page to be constant only if it
-                         * is mapped to a contiguous underlying disk block(s).
-                         * It will then make sure the corresponding device
-                         * cache of raid5 will be overwritten by this page.
-                         * - jay */
-                        if ((rw == OBD_BRW_WRITE) &&
-                            (nblocks == blocks_per_page) &&
-                            mapping_cap_page_constant_write(inode->i_mapping))
-                               SetPageConstant(page);
-
-                        if (bio != NULL &&
-                            can_be_merged(bio, sector) &&
-                            bio_add_page(bio, page,
-                                         blocksize * nblocks, page_offset) != 0)
-                                continue;       /* added this frag OK */
-
-                        if (bio != NULL) {
-                                struct request_queue *q =
-                                        bdev_get_queue(bio->bi_bdev);
-
-                                /* Dang! I have to fragment this I/O */
-                                CDEBUG(D_INODE, "bio++ sz %d vcnt %d(%d) "
-                                       "sectors %d(%d) psg %d(%d) hsg %d(%d) "
-                                       "sector %llu next %llu\n",
-                                       bio->bi_size,
-                                       bio->bi_vcnt, bio->bi_max_vecs,
-                                       bio->bi_size >> 9, queue_max_sectors(q),
-                                       bio_phys_segments(q, bio),
-                                       queue_max_phys_segments(q),
-                                       bio_hw_segments(q, bio),
-                                       queue_max_hw_segments(q),
-                                       (unsigned long long)bio->bi_sector,
-                                       (unsigned long long)sector);
-
-                                record_start_io(iobuf, rw, bio->bi_size, exp);
-                                rc = fsfilt_send_bio(rw, obd, inode, bio);
-                                if (rc < 0) {
-                                        CERROR("Can't send bio: %d\n", rc);
-                                        record_finish_io(iobuf, rw, rc);
-                                        goto out;
-                                }
-                                frags++;
-                        }
-
-                        /* allocate new bio, limited by max BIO size, b=9945 */
-                        bio = bio_alloc(GFP_NOIO, min(BIO_MAX_PAGES,
-                                                      (npages - page_idx) *
-                                                      blocks_per_page));
-                        if (bio == NULL) {
-                                CERROR("Can't allocate bio %u*%u = %u pages\n",
-                                       (npages - page_idx), blocks_per_page,
-                                       (npages - page_idx) * blocks_per_page);
-                                rc = -ENOMEM;
-                                goto out;
-                        }
-
-                        bio->bi_bdev = inode->i_sb->s_bdev;
-                        bio->bi_sector = sector;
-                       bio->bi_rw = (rw == OBD_BRW_READ ? READ : WRITE);
-                        bio->bi_end_io = dio_complete_routine;
-                        bio->bi_private = iobuf;
-
-                        rc = bio_add_page(bio, page,
-                                          blocksize * nblocks, page_offset);
-                        LASSERT (rc != 0);
-                }
-        }
-
-        if (bio != NULL) {
-                record_start_io(iobuf, rw, bio->bi_size, exp);
-                rc = fsfilt_send_bio(rw, obd, inode, bio);
-                if (rc >= 0) {
-                        frags++;
-                        rc = 0;
-                } else {
-                        CERROR("Can't send bio: %d\n", rc);
-                        record_finish_io(iobuf, rw, rc);
-                }
-        }
-
- out:
-        cfs_wait_event(iobuf->dr_wait,
-                       cfs_atomic_read(&iobuf->dr_numreqs) == 0);
-
-        if (rw == OBD_BRW_READ) {
-                lprocfs_oh_tally(&obd->u.filter.fo_filter_stats.
-                                  hist[BRW_R_DIO_FRAGS],
-                                 frags);
-                lprocfs_oh_tally_log2(&obd->u.filter.
-                                       fo_filter_stats.hist[BRW_R_IO_TIME],
-                                      jiffies - start_time);
-                if (exp->exp_nid_stats && exp->exp_nid_stats->nid_brw_stats) {
-                        lprocfs_oh_tally(&exp->exp_nid_stats->nid_brw_stats->
-                                          hist[BRW_R_DIO_FRAGS],
-                                         frags);
-                        lprocfs_oh_tally_log2(&exp->exp_nid_stats->
-                                             nid_brw_stats->hist[BRW_R_IO_TIME],
-                                              jiffies - start_time);
-                }
-        } else {
-                lprocfs_oh_tally(&obd->u.filter.fo_filter_stats.
-                                  hist[BRW_W_DIO_FRAGS], frags);
-                lprocfs_oh_tally_log2(&obd->u.filter.fo_filter_stats.
-                                       hist[BRW_W_IO_TIME],
-                                      jiffies - start_time);
-                if (exp->exp_nid_stats && exp->exp_nid_stats->nid_brw_stats) {
-                        lprocfs_oh_tally(&exp->exp_nid_stats->nid_brw_stats->
-                                          hist[BRW_W_DIO_FRAGS],
-                                         frags);
-                        lprocfs_oh_tally_log2(&exp->exp_nid_stats->
-                                             nid_brw_stats->hist[BRW_W_IO_TIME],
-                                              jiffies - start_time);
-                }
-        }
-
-        if (rc == 0)
-                rc = iobuf->dr_error;
-        RETURN(rc);
-}
-
-/* Must be called with i_mutex taken for writes; this will drop it */
-int filter_direct_io(int rw, struct dentry *dchild, struct filter_iobuf *iobuf,
-                     struct obd_export *exp, struct iattr *attr,
-                     struct obd_trans_info *oti, void **wait_handle)
-{
-        struct obd_device *obd = exp->exp_obd;
-        struct inode *inode = dchild->d_inode;
-        int blocks_per_page = CFS_PAGE_SIZE >> inode->i_blkbits;
-        int rc, rc2, create;
-        cfs_mutex_t *mutex;
-        ENTRY;
-
-        LASSERTF(iobuf->dr_npages <= iobuf->dr_max_pages, "%d,%d\n",
-                 iobuf->dr_npages, iobuf->dr_max_pages);
-        LASSERT(iobuf->dr_npages <= OBDFILTER_CREATED_SCRATCHPAD_ENTRIES);
-
-        if (rw == OBD_BRW_READ) {
-                if (iobuf->dr_npages == 0)
-                        RETURN(0);
-                create = 0;
-                mutex = NULL;
-        } else {
-                LASSERTF(rw == OBD_BRW_WRITE, "%x\n", rw);
-                LASSERT(iobuf->dr_npages > 0);
-                create = 1;
-                mutex = &obd->u.filter.fo_alloc_lock;
-        }
-
-        if (rw == OBD_BRW_WRITE &&
-            OBD_FAIL_CHECK(OBD_FAIL_OST_MAPBLK_ENOSPC)) {
-                rc = -ENOSPC;
-        } else {
-                rc = fsfilt_map_inode_pages(obd, inode, iobuf->dr_pages,
-                                    iobuf->dr_npages, iobuf->dr_blocks,
-                                    obdfilter_created_scratchpad, create, mutex);
-        }
-
-        if (rw == OBD_BRW_WRITE) {
-                if (rc == 0) {
-                        filter_tally(exp, iobuf->dr_pages,
-                                     iobuf->dr_npages, iobuf->dr_blocks,
-                                     blocks_per_page, 1);
-                        if (attr->ia_size > i_size_read(inode))
-                                attr->ia_valid |= ATTR_SIZE;
-                        rc = fsfilt_setattr(obd, dchild,
-                                            oti->oti_handle, attr, 0);
-                }
-
-               mutex_unlock(&inode->i_mutex);
-
-                /* Force commit to make the just-deleted blocks
-                 * reusable. LU-456 */
-                if (rc == -ENOSPC) {
-                        fsfilt_commit(obd, inode, oti->oti_handle, 1);
-                        RETURN(rc);
-                }
-
-                rc2 = filter_finish_transno(exp, inode, oti, 0, 0);
-                if (rc2 != 0) {
-                        CERROR("can't close transaction: %d\n", rc2);
-                        if (rc == 0)
-                                rc = rc2;
-                }
-
-                if (wait_handle)
-                        rc2 = fsfilt_commit_async(obd,inode,oti->oti_handle,
-                                                  wait_handle);
-                else
-                        rc2 = fsfilt_commit(obd, inode, oti->oti_handle, 0);
-                if (rc == 0)
-                        rc = rc2;
-                if (rc != 0)
-                        RETURN(rc);
-        } else if (rc == 0) {
-                filter_tally(exp, iobuf->dr_pages, iobuf->dr_npages,
-                             iobuf->dr_blocks, blocks_per_page, 0);
-        }
-
-        RETURN(filter_do_bio(exp, inode, iobuf, rw));
-}
-
-/* See if there are unallocated parts in given file region */
-static int filter_range_is_mapped(struct inode *inode, obd_size offset, int len)
-{
-        sector_t (*fs_bmap)(struct address_space *, sector_t) =
-                inode->i_mapping->a_ops->bmap;
-        int j;
-
-        /* We can't know if we are overwriting or not */
-        if (fs_bmap == NULL)
-                return 0;
-
-        offset >>= inode->i_blkbits;
-        len >>= inode->i_blkbits;
-
-        for (j = 0; j <= len; j++)
-                if (fs_bmap(inode->i_mapping, offset + j) == 0)
-                        return 0;
-
-        return 1;
-}
-
-/*
- * interesting use cases on how it interacts with VM:
- *
- * - vm writeout -- shouldn't see our pages as we don't mark them dirty
- *   though vm can find partial page left dirty by truncate. in this
- *   usual writeout is used unless our write rewrite that page - then we
- *   drop PG_dirty with PG_lock held.
- *
- * - else?
- *
- */
-int filter_commitrw_write(struct obd_export *exp, struct obdo *oa,
-                          int objcount, struct obd_ioobj *obj,
-                          struct niobuf_remote *nb, int niocount,
-                          struct niobuf_local *res, struct obd_trans_info *oti,
-                          int rc)
-{
-        struct niobuf_local *lnb;
-        struct filter_iobuf *iobuf = NULL;
-        struct lvfs_run_ctxt saved;
-        struct fsfilt_objinfo fso;
-        struct iattr iattr = { 0 };
-        struct inode *inode = res->dentry->d_inode;
-        unsigned long now = jiffies;
-        int i, err, cleanup_phase = 0;
-        struct obd_device *obd = exp->exp_obd;
-        struct filter_obd *fo = &obd->u.filter;
-        void *wait_handle = NULL;
-        int total_size = 0;
-        int sync_journal_commit = obd->u.filter.fo_syncjournal;
-        int retries = 0;
-        ENTRY;
-
-        LASSERT(oti != NULL);
-        LASSERT(objcount == 1);
-        LASSERT(current->journal_info == NULL);
-
-        if (rc != 0)
-                GOTO(cleanup, rc);
-
-        iobuf = filter_iobuf_get(&obd->u.filter, oti);
-        if (IS_ERR(iobuf))
-                GOTO(cleanup, rc = PTR_ERR(iobuf));
-        cleanup_phase = 1;
-
-        fso.fso_dentry = res->dentry;
-        fso.fso_bufcnt = obj->ioo_bufcnt;
-
-        iobuf->dr_ignore_quota = 0;
-        for (i = 0, lnb = res; i < niocount; i++, lnb++) {
-                loff_t this_size;
-                __u32 flags = lnb->flags;
-
-               if (filter_range_is_mapped(inode, lnb->lnb_file_offset,
-                                          lnb->len)) {
-                        /* If overwriting an existing block,
-                         * we don't need a grant */
-                        if (!(flags & OBD_BRW_GRANTED) && lnb->rc == -ENOSPC)
-                                lnb->rc = 0;
-                }
-
-                if (lnb->rc) { /* ENOSPC, network RPC error, etc. */
-                        CDEBUG(D_INODE, "Skipping [%d] == %d\n", i, lnb->rc);
-                        continue;
-                }
-
-                LASSERT(PageLocked(lnb->page));
-                LASSERT(!PageWriteback(lnb->page));
-
-                /* since write & truncate are serialized by the i_alloc_sem,
-                 * even partial truncate should not leave dirty pages in
-                 * the page cache */
-                LASSERT(!PageDirty(lnb->page));
-
-                SetPageUptodate(lnb->page);
-
-                err = filter_iobuf_add_page(obd, iobuf, inode, lnb->page);
-                LASSERT (err == 0);
-
-                total_size += lnb->len;
-
-                /* we expect these pages to be in offset order, but we'll
-                 * be forgiving */
-               this_size = lnb->lnb_file_offset + lnb->len;
-                if (this_size > iattr.ia_size)
-                        iattr.ia_size = this_size;
-
-                /* if one page is a write-back page from client cache and
-                 * not from direct_io, or it's written by root, then mark
-                 * the whole io request as ignore quota request, remote
-                 * client can not break through quota. */
-                if (exp_connect_rmtclient(exp))
-                        flags &= ~OBD_BRW_NOQUOTA;
-                if ((flags & OBD_BRW_NOQUOTA) ||
-                    (flags & (OBD_BRW_FROM_GRANT | OBD_BRW_SYNC)) ==
-                     OBD_BRW_FROM_GRANT)
-                        iobuf->dr_ignore_quota = 1;
-
-                if (!(lnb->flags & OBD_BRW_ASYNC)) {
-                        sync_journal_commit = 1;
-                }
-        }
-
-        if (OBD_FAIL_CHECK(OBD_FAIL_OST_DQACQ_NET))
-                GOTO(cleanup, rc = -EINPROGRESS);
-
-        push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
-        cleanup_phase = 2;
-
-retry:
-       mutex_lock(&inode->i_mutex);
-       fsfilt_check_slow(obd, now, "i_mutex");
-       oti->oti_handle = fsfilt_brw_start(obd, objcount, &fso, niocount, res,
-                                          oti);
-       if (IS_ERR(oti->oti_handle)) {
-               mutex_unlock(&inode->i_mutex);
-               rc = PTR_ERR(oti->oti_handle);
-               CDEBUG(rc == -ENOSPC ? D_INODE : D_ERROR,
-                      "error starting transaction: rc = %d\n", rc);
-               oti->oti_handle = NULL;
-               GOTO(cleanup, rc);
-       }
-       /* have to call fsfilt_commit() from this point on */
-
-        fsfilt_check_slow(obd, now, "brw_start");
-
-        /* Locking order: i_mutex -> journal_lock -> dqptr_sem. LU-952 */
-        ll_vfs_dq_init(inode);
-
-        i = OBD_MD_FLATIME | OBD_MD_FLMTIME | OBD_MD_FLCTIME;
-
-        /* If the inode still has SUID+SGID bits set (see filter_precreate())
-         * then we will accept the UID+GID if sent by the client for
-         * initializing the ownership of this inode.  We only allow this to
-         * happen once (so clear these bits) and later only allow setattr. */
-        if (inode->i_mode & S_ISUID)
-                i |= OBD_MD_FLUID;
-        if (inode->i_mode & S_ISGID)
-                i |= OBD_MD_FLGID;
-
-        iattr_from_obdo(&iattr, oa, i);
-        if (iattr.ia_valid & (ATTR_UID | ATTR_GID)) {
-                unsigned int save;
-
-                CDEBUG(D_INODE, "update UID/GID to %lu/%lu\n",
-                       (unsigned long)oa->o_uid, (unsigned long)oa->o_gid);
-
-                cfs_cap_raise(CFS_CAP_SYS_RESOURCE);
-
-                iattr.i