Whamcloud - gitweb
mgs_handler
authorlincent <lincent>
Mon, 15 Aug 2005 18:32:59 +0000 (18:32 +0000)
committerlincent <lincent>
Mon, 15 Aug 2005 18:32:59 +0000 (18:32 +0000)
 change the name from mgs_hander.c to mgs_handler.c

lustre/mgs/mgs_handler.c [new file with mode: 0644]

diff --git a/lustre/mgs/mgs_handler.c b/lustre/mgs/mgs_handler.c
new file mode 100644 (file)
index 0000000..21ae4fb
--- /dev/null
@@ -0,0 +1,812 @@
+/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
+ * vim:expandtab:shiftwidth=8:tabstop=8:
+ *
+ *  lustre/mgs/mgs_handler.c
+ *  Lustre Management Server (mgs) request handler
+ *
+ *  Copyright (C) 2001-2005 Cluster File Systems, Inc.
+ *   Author Nathan <nathan@clusterfs.com>
+ *   Author LinSongTao <lincent@clusterfs.com>
+ *
+ *   This file is part of Lustre, http://www.lustre.org.
+ *
+ *   Lustre is free software; you can redistribute it and/or
+ *   modify it under the terms of version 2 of the GNU General Public
+ *   License as published by the Free Software Foundation.
+ *
+ *   Lustre is distributed in the hope that it will be useful,
+ *   but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *   GNU General Public License for more details.
+ *
+ *   You should have received a copy of the GNU General Public License
+ *   along with Lustre; if not, write to the Free Software
+ *   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#ifndef EXPORT_SYMTAB
+# define EXPORT_SYMTAB
+#endif
+#define DEBUG_SUBSYSTEM S_MGS
+
+#ifdef __KERNEL__
+# include <linux/module.h>
+# include <linux/pagemap.h>
+# include <linux/miscdevice.h>
+# include <linux/init.h>
+#else
+# include <liblustre.h>
+#endif
+
+#include <linux/obd_class.h>
+#include <linux/lustre_dlm.h>
+#include <linux/lustre_mgs.h>
+#include <linux/lprocfs_status.h>
+#include <linux/lustre_fsfilt.h>
+#include <linux/lustre_commit_confd.h>
+#include "mgs_internal.h"
+
+static int mgs_postsetup(struct obd_device *obd);
+static int mgs_cleanup(struct obd_device *obd);
+
+/* Establish a connection to the MGS.*/
+static int mgs_connect(struct lustre_handle *conn, struct obd_device *obd,
+                       struct obd_uuid *cluuid, struct obd_connect_data *data)
+{
+        struct obd_export *exp;
+        struct mgs_export_data *med;
+        struct mgs_client_data *mcd;
+        int rc, abort_recovery;
+        ENTRY;
+
+        if (!conn || !obd || !cluuid)
+                RETURN(-EINVAL);
+
+        /* Check for aborted recovery. */
+        spin_lock_bh(&obd->obd_processing_task_lock);
+        abort_recovery = obd->obd_abort_recovery;
+        spin_unlock_bh(&obd->obd_processing_task_lock);
+        if (abort_recovery)
+                target_abort_recovery(obd);
+
+        rc = class_connect(conn, obd, cluuid);
+        if (rc)
+                RETURN(rc);
+        exp = class_conn2export(conn);
+        LASSERT(exp);
+        med = &exp->exp_mgs_data;
+
+        if (data != NULL) {
+                data->ocd_connect_flags &= MGS_CONNECT_SUPPORTED;
+                exp->exp_connect_flags = data->ocd_connect_flags;
+        }
+
+        OBD_ALLOC(mcd, sizeof(*mcd));
+        if (!mcd) {
+                CERROR("mgs: out of memory for client data\n");
+                GOTO(out, rc = -ENOMEM);
+        }
+
+        memcpy(mcd->mcd_uuid, cluuid, sizeof(mcd->mcd_uuid));
+        med->med_mcd = mcd;
+
+out:
+        if (rc) {
+                if (mcd) {
+                        OBD_FREE(mcd, sizeof(*mcd));
+                        med->med_mcd = NULL;
+                }
+                class_disconnect(exp);
+        } else {
+                class_export_put(exp);
+        }
+
+        RETURN(rc);
+}
+
+static int mgs_init_export(struct obd_export *exp)
+{
+        struct mgs_export_data *med = &exp->exp_mgs_data;
+
+        INIT_LIST_HEAD(&med->med_open_head);
+        spin_lock_init(&med->med_open_lock);
+        RETURN(0);
+}
+
+static int mgs_disconnect(struct obd_export *exp)
+{
+        unsigned long irqflags;
+        int rc;
+        ENTRY;
+
+        LASSERT(exp);
+        class_export_get(exp);
+
+        /* Disconnect early so that clients can't keep using export */
+        rc = class_disconnect(exp);
+
+        /* complete all outstanding replies */
+        spin_lock_irqsave(&exp->exp_lock, irqflags);
+        while (!list_empty(&exp->exp_outstanding_replies)) {
+                struct ptlrpc_reply_state *rs =
+                        list_entry(exp->exp_outstanding_replies.next,
+                                   struct ptlrpc_reply_state, rs_exp_list);
+                struct ptlrpc_service *svc = rs->rs_service;
+
+                spin_lock(&svc->srv_lock);
+                list_del_init(&rs->rs_exp_list);
+                ptlrpc_schedule_difficult_reply(rs);
+                spin_unlock(&svc->srv_lock);
+        }
+        spin_unlock_irqrestore(&exp->exp_lock, irqflags);
+
+        class_export_put(exp);
+        RETURN(rc);
+}
+
+/* mount the file system (secretly) */
+static int mgs_setup(struct obd_device *obd, obd_count len, void *buf)
+{
+        struct lprocfs_static_vars lvars;
+        struct lustre_cfg* lcfg = buf;
+        char *options = NULL;
+        struct mgs_obd *mgs = &obd->u.mgs;
+        struct vfsmount *mnt;
+        unsigned long page;
+        int rc = 0;
+        ENTRY;
+
+        /* setup 1:/dev/loop/0 2:ext3 3:mgs 4:errors=remount-ro,iopen_nopriv*/
+
+        if (lcfg->lcfg_bufcount < 3)
+                RETURN(rc = -EINVAL);
+
+        if (LUSTRE_CFG_BUFLEN(lcfg, 1) == 0 || LUSTRE_CFG_BUFLEN(lcfg, 2) == 0)
+                RETURN(rc = -EINVAL);
+
+        obd->obd_fsops = fsfilt_get_ops(lustre_cfg_string(lcfg, 2));
+        if (IS_ERR(obd->obd_fsops))
+                RETURN(rc = PTR_ERR(obd->obd_fsops));
+
+        page = __get_free_page(GFP_KERNEL);
+        if (!page)
+                RETURN(-ENOMEM);
+
+        options = (char *)page;
+        memset(options, 0, PAGE_SIZE);
+
+        if (LUSTRE_CFG_BUFLEN(lcfg, 4) > 0 && lustre_cfg_buf(lcfg, 4))
+                sprintf(options , ",%s", lustre_cfg_string(lcfg, 4));
+
+        //FIXME mount was already done in lustre_fill_super,
+        //we just need to access it
+        mnt = do_kern_mount(lustre_cfg_string(lcfg, 2), 0,
+                            lustre_cfg_string(lcfg, 1), (void *)options);
+        free_page(page);
+        if (IS_ERR(mnt)) {
+                rc = PTR_ERR(mnt);
+                CERROR("do_kern_mount failed: rc = %d\n", rc);
+                GOTO(err_ops, rc);
+        }
+
+        CDEBUG(D_SUPER, "%s: mnt = %p\n", lustre_cfg_string(lcfg, 1), mnt);
+
+        LASSERT(!lvfs_check_rdonly(lvfs_sbdev(mnt->mnt_sb)));
+
+        rc = mgs_fs_setup(obd, mnt);
+        if (rc) {
+                CERROR("%s: MGS filesystem method init failed: rc = %d\n",
+                       obd->obd_name, rc);
+                GOTO(err_put, rc);
+        }
+
+        rc = llog_start_commit_thread();
+        if (rc < 0)
+                GOTO(err_fs, rc);
+#if 0  
+        //FIXME: no LDLM support for llog now
+        ptlrpc_init_client(LDLM_CB_REQUEST_PORTAL, LDLM_CB_REPLY_PORTAL,
+                           "mgs_ldlm_client", &obd->obd_ldlm_client);
+#endif
+        obd->obd_replayable = 1;
+
+        rc = mgs_postsetup(obd);
+        if (rc)
+                GOTO(err_fs, rc);
+
+        lprocfs_init_vars(mgs, &lvars);
+        lprocfs_obd_setup(obd, lvars.obd_vars);
+
+        if (obd->obd_recovering) {
+                LCONSOLE_WARN("MGT %s now serving %s, but will be in recovery "
+                              "until %d %s reconnect, or if no clients "
+                              "reconnect for %d:%.02d; during that time new "
+                              "clients will not be allowed to connect. "
+                              "Recovery progress can be monitored by watching "
+                              "/proc/fs/lustre/mgs/%s/recovery_status.\n",
+                              obd->obd_name,
+                              lustre_cfg_string(lcfg, 1),
+                              obd->obd_recoverable_clients,
+                              (obd->obd_recoverable_clients == 1) 
+                              ? "client" : "clients",
+                              (int)(OBD_RECOVERY_TIMEOUT / HZ) / 60,
+                              (int)(OBD_RECOVERY_TIMEOUT / HZ) % 60,
+                              obd->obd_name);
+        } else {
+                LCONSOLE_INFO("MGT %s now serving %s with recovery %s.\n",
+                              obd->obd_name,
+                              lustre_cfg_string(lcfg, 1),
+                              obd->obd_replayable ? "enabled" : "disabled");
+        }
+//FIXME: no ldlm support now
+        ldlm_timeout = 6;
+        ping_evictor_start();
+
+        RETURN(0);
+
+err_fs:
+        /* No extra cleanup needed for llog_init_commit_thread() */
+        mgs_fs_cleanup(obd);
+err_put:
+        unlock_kernel();
+        mntput(mgs->mgs_vfsmnt);
+        mgs->mgs_sb = 0;
+        lock_kernel();
+err_ops:
+        fsfilt_put_ops(obd->obd_fsops);
+        return rc;
+}
+
+static int mgs_postsetup(struct obd_device *obd)
+{
+        int rc = 0;
+        ENTRY;
+
+        rc = llog_setup(obd, LLOG_CONFIG_ORIG_CTXT, obd, 0, NULL,
+                        &llog_lvfs_ops);
+        RETURN(rc);
+}
+
+static int mgs_precleanup(struct obd_device *obd, int stage)
+{
+        int rc = 0;
+        ENTRY;
+
+        llog_cleanup(llog_get_context(obd, LLOG_CONFIG_ORIG_CTXT));
+        rc = obd_llog_finish(obd, 0);
+        RETURN(rc);
+}
+
+static int mgs_cleanup(struct obd_device *obd)
+{
+        struct mgs_obd *mgs = &obd->u.mgs;
+        lvfs_sbdev_type save_dev;
+        int must_relock = 0;
+        ENTRY;
+
+        ping_evictor_stop();
+
+        if (mgs->mgs_sb == NULL)
+                RETURN(0);
+        save_dev = lvfs_sbdev(mgs->mgs_sb);
+
+        lprocfs_obd_cleanup(obd);
+
+        mgs_update_server_data(obd, 1);
+
+        mgs_fs_cleanup(obd);
+
+        if (atomic_read(&obd->u.mgs.mgs_vfsmnt->mnt_count) > 2)
+                CERROR("%s: mount busy, mnt_count %d != 2\n", obd->obd_name,
+                       atomic_read(&obd->u.mgs.mgs_vfsmnt->mnt_count));
+
+        /* We can only unlock kernel if we are in the context of sys_ioctl,
+           otherwise we never called lock_kernel */
+        if (kernel_locked()) {
+                unlock_kernel();
+                must_relock++;
+        }
+
+        mntput(mgs->mgs_vfsmnt);
+        mgs->mgs_sb = NULL;
+
+        spin_lock_bh(&obd->obd_processing_task_lock);
+        if (obd->obd_recovering) {
+                target_cancel_recovery_timer(obd);
+                obd->obd_recovering = 0;
+        }
+        spin_unlock_bh(&obd->obd_processing_task_lock);
+
+        lvfs_clear_rdonly(save_dev);
+
+        if (must_relock)
+                lock_kernel();
+
+        fsfilt_put_ops(obd->obd_fsops);
+
+        LCONSOLE_INFO("MDT %s has stopped.\n", obd->obd_name);
+
+        RETURN(0);
+}
+
+/* Look up an entry by inode number. */
+/* this function ONLY returns valid dget'd dentries with an initialized inode
+   or errors */
+struct dentry *mgs_fid2dentry(struct mgs_obd *mgs, struct ll_fid *fid,
+                              struct vfsmount **mnt)
+{
+        unsigned long ino = fid->id;
+        __u32 generation = fid->generation;
+        struct mgs_open_llogs *mollog, *n;
+        struct list_head *llog_list = &mgs->mgs_open_llogs;
+        struct inode *inode;
+        struct dentry *result = NULL;
+
+        if (ino == 0)
+                RETURN(ERR_PTR(-ESTALE));
+
+
+        CDEBUG(D_DENTRY, "--> mgs_fid2dentry: ino/gen %lu/%u, sb %p\n",
+               ino, generation, mgs->mgs_sb);
+
+        list_for_each_entry_safe(mollog, n, llog_list, mol_list) {
+                if (mollog->mod_id == ino) {
+                        result = mollog->mol_dentry;
+                        dget(result);
+                }
+        }
+
+        if (!result)
+                RETURN(NULL);
+
+        inode = result->d_inode;
+        if (!inode)
+                RETURN(ERR_PTR(-ENOENT));
+
+        if (generation && inode->i_generation != generation) {
+                /* we didn't find the right inode.. */
+                CERROR("bad inode %lu, link: %lu ct: %d or generation %u/%u\n",
+                       inode->i_ino, (unsigned long)inode->i_nlink,
+                       atomic_read(&inode->i_count), inode->i_generation,
+                       generation);
+                dput(result);
+                RETURN(ERR_PTR(-ENOENT));
+        }
+
+        if (mnt) {
+                *mnt = mgs->mgs_vfsmnt;
+                mntget(*mnt);
+        }
+
+        RETURN(result);
+}
+
+static struct dentry *mgs_lvfs_fid2dentry(__u64 id, __u32 gen, __u64 gr,
+                                          void *data)
+{
+        struct obd_device *obd = data;
+        struct ll_fid fid;
+        fid.id = id;
+        fid.generation = gen;
+        return mgs_fid2dentry(&obd->u.mgs, &fid, NULL);
+}
+
+
+int mgs_iocontrol(unsigned int cmd, struct obd_export *exp, int len,
+                  void *karg, void *uarg)
+{
+        static struct obd_uuid cfg_uuid = { .uuid = "config_uuid" };
+        struct obd_device *obd = exp->exp_obd;
+        struct mgs_obd *mgs = &obd->u.mgs;
+        struct obd_ioctl_data *data = karg;
+        struct lvfs_run_ctxt saved;
+        int rc = 0;
+
+        ENTRY;
+        CDEBUG(D_IOCTL, "handling ioctl cmd %#x\n", cmd);
+
+        switch (cmd) {
+        case OBD_IOC_RECORD: {
+                char *name = data->ioc_inlbuf1;
+                if (mgs->mgs_cfg_llh)
+                        RETURN(-EBUSY);
+
+                push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
+                rc = llog_create(llog_get_context(obd, LLOG_CONFIG_ORIG_CTXT),
+                                 &mgs->mgs_cfg_llh, NULL, name);
+                if (rc == 0)
+                        llog_init_handle(mgs->mgs_cfg_llh, LLOG_F_IS_PLAIN,
+                                         &cfg_uuid);
+                else
+                        mgs->mgs_cfg_llh = NULL;
+                pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
+
+                RETURN(rc);
+        }
+
+        case OBD_IOC_ENDRECORD: {
+                if (!mgs->mgs_cfg_llh)
+                        RETURN(-EBADF);
+
+                push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
+                rc = llog_close(mgs->mgs_cfg_llh);
+                pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
+
+                mgs->mgs_cfg_llh = NULL;
+                RETURN(rc);
+        }
+
+        case OBD_IOC_CLEAR_LOG: {
+                char *name = data->ioc_inlbuf1;
+                if (mgs->mgs_cfg_llh)
+                        RETURN(-EBUSY);
+
+                push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
+                rc = llog_create(llog_get_context(obd, LLOG_CONFIG_ORIG_CTXT),
+                                 &mgs->mgs_cfg_llh, NULL, name);
+                if (rc == 0) {
+                        llog_init_handle(mgs->mgs_cfg_llh, LLOG_F_IS_PLAIN,
+                                         NULL);
+
+                        rc = llog_destroy(mgs->mgs_cfg_llh);
+                        llog_free_handle(mgs->mgs_cfg_llh);
+                }
+                pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
+
+                mgs->mgs_cfg_llh = NULL;
+                RETURN(rc);
+        }
+
+        case OBD_IOC_DORECORD: {
+                char *cfg_buf;
+                struct llog_rec_hdr rec;
+                if (!mgs->mgs_cfg_llh)
+                        RETURN(-EBADF);
+
+                rec.lrh_len = llog_data_len(data->ioc_plen1);
+
+                if (data->ioc_type == LUSTRE_CFG_TYPE) {
+                        rec.lrh_type = OBD_CFG_REC;
+                } else {
+                        CERROR("unknown cfg record type:%d \n", data->ioc_type);
+                        RETURN(-EINVAL);
+                }
+
+                OBD_ALLOC(cfg_buf, data->ioc_plen1);
+                if (cfg_buf == NULL)
+                        RETURN(-EINVAL);
+                rc = copy_from_user(cfg_buf, data->ioc_pbuf1, data->ioc_plen1);
+                if (rc) {
+                        OBD_FREE(cfg_buf, data->ioc_plen1);
+                        RETURN(rc);
+                }
+
+                push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
+                rc = llog_write_rec(mgs->mgs_cfg_llh, &rec, NULL, 0,
+                                    cfg_buf, -1);
+                pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
+
+                OBD_FREE(cfg_buf, data->ioc_plen1);
+                RETURN(rc);
+        }
+
+        case OBD_IOC_PARSE: {
+                struct llog_ctxt *ctxt =
+                        llog_get_context(obd, LLOG_CONFIG_ORIG_CTXT);
+                push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
+                rc = class_config_parse_llog(ctxt, data->ioc_inlbuf1, NULL);
+                pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
+                if (rc)
+                        RETURN(rc);
+
+                RETURN(rc);
+        }
+
+        case OBD_IOC_DUMP_LOG: {
+                struct llog_ctxt *ctxt =
+                        llog_get_context(obd, LLOG_CONFIG_ORIG_CTXT);
+                push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
+                rc = class_config_dump_llog(ctxt, data->ioc_inlbuf1, NULL);
+                pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
+                if (rc)
+                        RETURN(rc);
+
+                RETURN(rc);
+        }
+
+        case OBD_IOC_SYNC: {
+                CDEBUG(D_HA, "syncing mgs %s\n", obd->obd_name);
+                rc = fsfilt_sync(obd, obd->u.mgs.mgs_sb);
+                RETURN(rc);
+        }
+
+        case OBD_IOC_SET_READONLY: {
+                void *handle;
+                struct inode *inode = obd->u.mgs.mgs_sb->s_root->d_inode;
+                BDEVNAME_DECLARE_STORAGE(tmp);
+                CERROR("*** setting device %s read-only ***\n",
+                       ll_bdevname(obd->u.mgs.mgs_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 mgs %s\n", obd->obd_name);
+                rc = fsfilt_sync(obd, obd->u.mgs.mgs_sb);
+
+                lvfs_set_rdonly(lvfs_sbdev(obd->u.mgs.mgs_sb));
+                RETURN(0);
+        }
+
+
+        case OBD_IOC_LLOG_CHECK:
+        case OBD_IOC_LLOG_CANCEL:
+        case OBD_IOC_LLOG_REMOVE: {
+                struct llog_ctxt *ctxt =
+                        llog_get_context(obd, LLOG_CONFIG_ORIG_CTXT);
+
+                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);
+        }
+        case OBD_IOC_LLOG_INFO:
+        case OBD_IOC_LLOG_PRINT: {
+                struct llog_ctxt *ctxt =
+                        llog_get_context(obd, LLOG_CONFIG_ORIG_CTXT);
+
+                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);
+        }
+
+        case OBD_IOC_ABORT_RECOVERY:
+                CERROR("aborting recovery for device %s\n", obd->obd_name);
+                target_abort_recovery(obd);
+                RETURN(0);
+
+        default:
+                CDEBUG(D_INFO, "unknown command %x\n", cmd);
+                RETURN(-EINVAL);
+        }
+        RETURN(0);
+
+}
+
+
+int mgs_handle(struct ptlrpc_request *req)
+{
+        int should_process, fail = OBD_FAIL_MGS_ALL_REPLY_NET;
+        int rc = 0;
+        struct mgs_obd *mgs = NULL; /* quell gcc overwarning */
+        struct obd_device *obd = NULL;
+        ENTRY;
+
+        OBD_FAIL_RETURN(OBD_FAIL_MGS_ALL_REQUEST_NET | OBD_FAIL_ONCE, 0);
+
+        LASSERT(current->journal_info == NULL);
+        /* XXX identical to MDS */
+        if (req->rq_reqmsg->opc != MGS_CONNECT) {
+                struct mgs_export_data *med;
+                int abort_recovery;
+
+                if (req->rq_export == NULL) {
+                        CERROR("lustre_mgs: operation %d on unconnected MGS\n",
+                               req->rq_reqmsg->opc);
+                        req->rq_status = -ENOTCONN;
+                        GOTO(out, rc = -ENOTCONN);
+                }
+
+                med = &req->rq_export->exp_mgs_data;
+                obd = req->rq_export->exp_obd;
+                mgs = &obd->u.mgs;
+
+                /* sanity check: if the xid matches, the request must
+                 * be marked as a resent or replayed */
+                if (req->rq_xid == med->med_mcd->mcd_last_xid)
+                        LASSERTF(lustre_msg_get_flags(req->rq_reqmsg) &
+                                 (MSG_RESENT | MSG_REPLAY),
+                                 "rq_xid "LPU64" matches last_xid, "
+                                 "expected RESENT flag\n",
+                                 req->rq_xid);
+                /* else: note the opposite is not always true; a
+                 * RESENT req after a failover will usually not match
+                 * the last_xid, since it was likely never
+                 * committed. A REPLAYed request will almost never
+                 * match the last xid, however it could for a
+                 * committed, but still retained, open. */
+
+                /* Check for aborted recovery. */
+                spin_lock_bh(&obd->obd_processing_task_lock);
+                abort_recovery = obd->obd_abort_recovery;
+                spin_unlock_bh(&obd->obd_processing_task_lock);
+                if (abort_recovery) {
+                        target_abort_recovery(obd);
+                } 
+        }
+
+        switch (req->rq_reqmsg->opc) {
+        case MGS_CONNECT:
+                DEBUG_REQ(D_INODE, req, "connect");
+                OBD_FAIL_RETURN(OBD_FAIL_MGS_CONNECT_NET, 0);
+                rc = target_handle_connect(req, mgs_handle);
+                if (!rc) {
+                        /* Now that we have an export, set mgs. */
+                        obd = req->rq_export->exp_obd;
+                        mgs = mgs_req2mgs(req);
+                }
+                break;
+
+        case MGS_DISCONNECT:
+                DEBUG_REQ(D_INODE, req, "disconnect");
+                OBD_FAIL_RETURN(OBD_FAIL_MGS_DISCONNECT_NET, 0);
+                rc = target_handle_disconnect(req);
+                req->rq_status = rc;            /* superfluous? */
+                break;
+
+        case OBD_PING:
+                DEBUG_REQ(D_INODE, req, "ping");
+                rc = target_handle_ping(req);
+                break;
+
+        case OBD_LOG_CANCEL:
+                CDEBUG(D_INODE, "log cancel\n");
+                OBD_FAIL_RETURN(OBD_FAIL_OBD_LOG_CANCEL_NET, 0);
+                rc = -ENOTSUPP; /* la la la */
+                break;
+
+        case LLOG_ORIGIN_HANDLE_CREATE:
+                DEBUG_REQ(D_INODE, req, "llog_init");
+                OBD_FAIL_RETURN(OBD_FAIL_OBD_LOGD_NET, 0);
+                rc = llog_origin_handle_create(req);
+                break;
+        case LLOG_ORIGIN_HANDLE_NEXT_BLOCK:
+                DEBUG_REQ(D_INODE, req, "llog next block");
+                OBD_FAIL_RETURN(OBD_FAIL_OBD_LOGD_NET, 0);
+                rc = llog_origin_handle_next_block(req);
+                break;
+        case LLOG_ORIGIN_HANDLE_READ_HEADER:
+                DEBUG_REQ(D_INODE, req, "llog read header");
+                OBD_FAIL_RETURN(OBD_FAIL_OBD_LOGD_NET, 0);
+                rc = llog_origin_handle_read_header(req);
+                break;
+        case LLOG_ORIGIN_HANDLE_CLOSE:
+                DEBUG_REQ(D_INODE, req, "llog close");
+                OBD_FAIL_RETURN(OBD_FAIL_OBD_LOGD_NET, 0);
+                rc = llog_origin_handle_close(req);
+                break;
+        case LLOG_CATINFO:
+                DEBUG_REQ(D_INODE, req, "llog catinfo");
+                OBD_FAIL_RETURN(OBD_FAIL_OBD_LOGD_NET, 0);
+                rc = llog_catinfo(req);
+                break;
+        default:
+                req->rq_status = -ENOTSUPP;
+                rc = ptlrpc_error(req);
+                RETURN(rc);
+        }
+
+        LASSERT(current->journal_info == NULL);
+
+        /* If we're DISCONNECTing, the mgs_export_data is already freed */
+        if (!rc && req->rq_reqmsg->opc != MGS_DISCONNECT) {
+                struct mgs_export_data *med = &req->rq_export->exp_mgs_data;
+                req->rq_repmsg->last_xid =
+                        le64_to_cpu(med->med_mcd->mcd_last_xid);
+
+                target_committed_to_req(req);
+        }
+
+        EXIT;
+ out:
+
+        if (lustre_msg_get_flags(req->rq_reqmsg) & MSG_LAST_REPLAY) {
+                if (obd && obd->obd_recovering) {
+                        DEBUG_REQ(D_HA, req, "LAST_REPLAY, queuing reply");
+                        return target_queue_final_reply(req, rc);
+                }
+                /* Lost a race with recovery; let the error path DTRT. */
+                rc = req->rq_status = -ENOTCONN;
+        }
+
+        target_send_reply(req, rc, fail);
+        return 0;
+}
+
+static int mgt_setup(struct obd_device *obd, obd_count len, void *buf)
+{
+        struct mgs_obd *mgs = &obd->u.mgs;
+        struct lprocfs_static_vars lvars;
+        int rc = 0;
+        ENTRY;
+
+        lprocfs_init_vars(mgt, &lvars);
+        lprocfs_obd_setup(obd, lvars.obd_vars);
+
+        mgs->mgs_service =
+                ptlrpc_init_svc(MGS_NBUFS, MGS_BUFSIZE, MGS_MAXREQSIZE,
+                                MGS_REQUEST_PORTAL, MGC_REPLY_PORTAL,
+                                MGS_SERVICE_WATCHDOG_TIMEOUT,
+                                mgs_handle, "mgs", obd->obd_proc_entry, NULL);
+
+        if (!mgs->mgs_service) {
+                CERROR("failed to start service\n");
+                GOTO(err_lprocfs, rc = -ENOMEM);
+        }
+
+        rc = ptlrpc_start_n_threads(obd, mgs->mgs_service, MGT_NUM_THREADS,
+                                    "ll_mgt");
+        if (rc)
+                GOTO(err_thread, rc);
+
+        RETURN(0);
+
+err_thread:
+        ptlrpc_unregister_service(mgs->mgs_service);
+err_lprocfs:
+        lprocfs_obd_cleanup(obd);
+        return rc;
+}
+
+static int mgt_cleanup(struct obd_device *obd)
+{
+        struct mgs_obd *mgs = &obd->u.mgs;
+        ENTRY;
+
+        ptlrpc_unregister_service(mgs->mgs_service);
+
+        lprocfs_obd_cleanup(obd);
+
+        RETURN(0);
+}
+
+struct lvfs_callback_ops mgs_lvfs_ops = {
+        l_fid2dentry:     mgs_lvfs_fid2dentry,
+};
+
+/* use obd ops to offer management infrastructure */
+static struct obd_ops mgs_obd_ops = {
+        .o_owner           = THIS_MODULE,
+        .o_connect         = mgs_connect,
+        .o_init_export     = mgs_init_export,
+        .o_disconnect      = mgs_disconnect,
+        .o_setup           = mgs_setup,
+        .o_precleanup      = mgs_precleanup,
+        .o_cleanup         = mgs_cleanup,
+        .o_iocontrol       = mgs_iocontrol,
+};
+
+static struct obd_ops mgt_obd_ops = {
+        .o_owner           = THIS_MODULE,
+        .o_setup           = mgt_setup,
+        .o_cleanup         = mgt_cleanup,
+};
+
+static int __init mgs_init(void)
+{
+        int rc;
+        struct lprocfs_static_vars lvars;
+
+        lprocfs_init_vars(mgs, &lvars);
+        class_register_type(&mgs_obd_ops, lvars.module_vars, LUSTRE_MGS_NAME);
+        lprocfs_init_vars(mgt, &lvars);
+        class_register_type(&mgt_obd_ops, lvars.module_vars, LUSTRE_MGT_NAME);
+
+        return 0;
+}
+
+static void /*__exit*/ mgs_exit(void)
+{
+        class_unregister_type(LUSTRE_MGS_NAME);
+        class_unregister_type(LUSTRE_MGT_NAME);
+}
+
+MODULE_AUTHOR("Cluster File Systems, Inc. <info@clusterfs.com>");
+MODULE_DESCRIPTION("Lustre  Management Server (MGS)");
+MODULE_LICENSE("GPL");
+
+module_init(mgs_init);
+module_exit(mgs_exit);