alistair23-linux/drivers/staging/lustre/lustre/mgc/mgc_request.c

/*
 * 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) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
 * Use is subject to license terms.
 *
 * Copyright (c) 2011, 2012, Intel Corporation.
 */
/*
 * This file is part of Lustre, http://www.lustre.org/
 * Lustre is a trademark of Sun Microsystems, Inc.
 *
 * lustre/mgc/mgc_request.c
 *
 * Author: Nathan Rutman <nathan@clusterfs.com>
 */

#define DEBUG_SUBSYSTEM S_MGC
#define D_MGC D_CONFIG /*|D_WARNING*/

# include <linux/module.h>
# include <linux/pagemap.h>
# include <linux/miscdevice.h>
# include <linux/init.h>

#include <obd_class.h>
#include <lustre_dlm.h>
#include <lprocfs_status.h>
#include <lustre_log.h>
#include <lustre_fsfilt.h>
#include <lustre_disk.h>
#include "mgc_internal.h"

static int mgc_name2resid(char *name, int len, struct ldlm_res_id *res_id,
			  int type)
{
	__u64 resname = 0;

	if (len > 8) {
		CERROR("name too long: %s\n", name);
		return -EINVAL;
	}
	if (len <= 0) {
		CERROR("missing name: %s\n", name);
		return -EINVAL;
	}
	memcpy(&resname, name, len);

	/* Always use the same endianness for the resid */
	memset(res_id, 0, sizeof(*res_id));
	res_id->name[0] = cpu_to_le64(resname);
	/* XXX: unfortunately, sptlprc and config llog share one lock */
	switch(type) {
	case CONFIG_T_CONFIG:
	case CONFIG_T_SPTLRPC:
		resname = 0;
		break;
	case CONFIG_T_RECOVER:
		resname = type;
		break;
	default:
		LBUG();
	}
	res_id->name[1] = cpu_to_le64(resname);
	CDEBUG(D_MGC, "log %s to resid "LPX64"/"LPX64" (%.8s)\n", name,
	       res_id->name[0], res_id->name[1], (char *)&res_id->name[0]);
	return 0;
}

int mgc_fsname2resid(char *fsname, struct ldlm_res_id *res_id, int type)
{
	/* fsname is at most 8 chars long, maybe contain "-".
	 * e.g. "lustre", "SUN-000" */
	return mgc_name2resid(fsname, strlen(fsname), res_id, type);
}
EXPORT_SYMBOL(mgc_fsname2resid);

int mgc_logname2resid(char *logname, struct ldlm_res_id *res_id, int type)
{
	char *name_end;
	int len;

	/* logname consists of "fsname-nodetype".
	 * e.g. "lustre-MDT0001", "SUN-000-client" */
	name_end = strrchr(logname, '-');
	LASSERT(name_end);
	len = name_end - logname;
	return mgc_name2resid(logname, len, res_id, type);
}

/********************** config llog list **********************/
static LIST_HEAD(config_llog_list);
static DEFINE_SPINLOCK(config_list_lock);

/* Take a reference to a config log */
static int config_log_get(struct config_llog_data *cld)
{
	ENTRY;
	atomic_inc(&cld->cld_refcount);
	CDEBUG(D_INFO, "log %s refs %d\n", cld->cld_logname,
	       atomic_read(&cld->cld_refcount));
	RETURN(0);
}

/* Drop a reference to a config log.  When no longer referenced,
   we can free the config log data */
static void config_log_put(struct config_llog_data *cld)
{
	ENTRY;

	CDEBUG(D_INFO, "log %s refs %d\n", cld->cld_logname,
	       atomic_read(&cld->cld_refcount));
	LASSERT(atomic_read(&cld->cld_refcount) > 0);

	/* spinlock to make sure no item with 0 refcount in the list */
	if (atomic_dec_and_lock(&cld->cld_refcount, &config_list_lock)) {
		list_del(&cld->cld_list_chain);
		spin_unlock(&config_list_lock);

		CDEBUG(D_MGC, "dropping config log %s\n", cld->cld_logname);

		if (cld->cld_recover)
			config_log_put(cld->cld_recover);
		if (cld->cld_sptlrpc)
			config_log_put(cld->cld_sptlrpc);
		if (cld_is_sptlrpc(cld))
			sptlrpc_conf_log_stop(cld->cld_logname);

		class_export_put(cld->cld_mgcexp);
		OBD_FREE(cld, sizeof(*cld) + strlen(cld->cld_logname) + 1);
	}

	EXIT;
}

/* Find a config log by name */
static
struct config_llog_data *config_log_find(char *logname,
					 struct config_llog_instance *cfg)
{
	struct config_llog_data *cld;
	struct config_llog_data *found = NULL;
	void *		   instance;
	ENTRY;

	LASSERT(logname != NULL);

	instance = cfg ? cfg->cfg_instance : NULL;
	spin_lock(&config_list_lock);
	list_for_each_entry(cld, &config_llog_list, cld_list_chain) {
		/* check if instance equals */
		if (instance != cld->cld_cfg.cfg_instance)
			continue;

		/* instance may be NULL, should check name */
		if (strcmp(logname, cld->cld_logname) == 0) {
			found = cld;
			break;
		}
	}
	if (found) {
		atomic_inc(&found->cld_refcount);
		LASSERT(found->cld_stopping == 0 || cld_is_sptlrpc(found) == 0);
	}
	spin_unlock(&config_list_lock);
	RETURN(found);
}

static
struct config_llog_data *do_config_log_add(struct obd_device *obd,
					   char *logname,
					   int type,
					   struct config_llog_instance *cfg,
					   struct super_block *sb)
{
	struct config_llog_data *cld;
	int		      rc;
	ENTRY;

	CDEBUG(D_MGC, "do adding config log %s:%p\n", logname,
	       cfg ? cfg->cfg_instance : 0);

	OBD_ALLOC(cld, sizeof(*cld) + strlen(logname) + 1);
	if (!cld)
		RETURN(ERR_PTR(-ENOMEM));

	strcpy(cld->cld_logname, logname);
	if (cfg)
		cld->cld_cfg = *cfg;
	else
		cld->cld_cfg.cfg_callback = class_config_llog_handler;
	mutex_init(&cld->cld_lock);
	cld->cld_cfg.cfg_last_idx = 0;
	cld->cld_cfg.cfg_flags = 0;
	cld->cld_cfg.cfg_sb = sb;
	cld->cld_type = type;
	atomic_set(&cld->cld_refcount, 1);

	/* Keep the mgc around until we are done */
	cld->cld_mgcexp = class_export_get(obd->obd_self_export);

	if (cld_is_sptlrpc(cld)) {
		sptlrpc_conf_log_start(logname);
		cld->cld_cfg.cfg_obdname = obd->obd_name;
	}

	rc = mgc_logname2resid(logname, &cld->cld_resid, type);

	spin_lock(&config_list_lock);
	list_add(&cld->cld_list_chain, &config_llog_list);
	spin_unlock(&config_list_lock);

	if (rc) {
		config_log_put(cld);
		RETURN(ERR_PTR(rc));
	}

	if (cld_is_sptlrpc(cld)) {
		rc = mgc_process_log(obd, cld);
		if (rc && rc != -ENOENT)
			CERROR("failed processing sptlrpc log: %d\n", rc);
	}

	RETURN(cld);
}

static struct config_llog_data *config_recover_log_add(struct obd_device *obd,
	char *fsname,
	struct config_llog_instance *cfg,
	struct super_block *sb)
{
	struct config_llog_instance lcfg = *cfg;
	struct lustre_sb_info *lsi = s2lsi(sb);
	struct config_llog_data *cld;
	char logname[32];

	if (IS_OST(lsi))
		return NULL;

	/* for osp-on-ost, see lustre_start_osp() */
	if (IS_MDT(lsi) && lcfg.cfg_instance)
		return NULL;

	/* we have to use different llog for clients and mdts for cmd
	 * where only clients are notified if one of cmd server restarts */
	LASSERT(strlen(fsname) < sizeof(logname) / 2);
	strcpy(logname, fsname);
	if (IS_SERVER(lsi)) { /* mdt */
		LASSERT(lcfg.cfg_instance == NULL);
		lcfg.cfg_instance = sb;
		strcat(logname, "-mdtir");
	} else {
		LASSERT(lcfg.cfg_instance != NULL);
		strcat(logname, "-cliir");
	}

	cld = do_config_log_add(obd, logname, CONFIG_T_RECOVER, &lcfg, sb);
	return cld;
}


/** Add this log to the list of active logs watched by an MGC.
 * Active means we're watching for updates.
 * We have one active log per "mount" - client instance or servername.
 * Each instance may be at a different point in the log.
 */
static int config_log_add(struct obd_device *obd, char *logname,
			  struct config_llog_instance *cfg,
			  struct super_block *sb)
{
	struct lustre_sb_info *lsi = s2lsi(sb);
	struct config_llog_data *cld;
	struct config_llog_data *sptlrpc_cld;
	char		     seclogname[32];
	char		    *ptr;
	ENTRY;

	CDEBUG(D_MGC, "adding config log %s:%p\n", logname, cfg->cfg_instance);

	/*
	 * for each regular log, the depended sptlrpc log name is
	 * <fsname>-sptlrpc. multiple regular logs may share one sptlrpc log.
	 */
	ptr = strrchr(logname, '-');
	if (ptr == NULL || ptr - logname > 8) {
		CERROR("logname %s is too long\n", logname);
		RETURN(-EINVAL);
	}

	memcpy(seclogname, logname, ptr - logname);
	strcpy(seclogname + (ptr - logname), "-sptlrpc");

	sptlrpc_cld = config_log_find(seclogname, NULL);
	if (sptlrpc_cld == NULL) {
		sptlrpc_cld = do_config_log_add(obd, seclogname,
						CONFIG_T_SPTLRPC, NULL, NULL);
		if (IS_ERR(sptlrpc_cld)) {
			CERROR("can't create sptlrpc log: %s\n", seclogname);
			RETURN(PTR_ERR(sptlrpc_cld));
		}
	}

	cld = do_config_log_add(obd, logname, CONFIG_T_CONFIG, cfg, sb);
	if (IS_ERR(cld)) {
		CERROR("can't create log: %s\n", logname);
		config_log_put(sptlrpc_cld);
		RETURN(PTR_ERR(cld));
	}

	cld->cld_sptlrpc = sptlrpc_cld;

	LASSERT(lsi->lsi_lmd);
	if (!(lsi->lsi_lmd->lmd_flags & LMD_FLG_NOIR)) {
		struct config_llog_data *recover_cld;
		*strrchr(seclogname, '-') = 0;
		recover_cld = config_recover_log_add(obd, seclogname, cfg, sb);
		if (IS_ERR(recover_cld)) {
			config_log_put(cld);
			RETURN(PTR_ERR(recover_cld));
		}
		cld->cld_recover = recover_cld;
	}

	RETURN(0);
}

DEFINE_MUTEX(llog_process_lock);

/** Stop watching for updates on this log.
 */
static int config_log_end(char *logname, struct config_llog_instance *cfg)
{
	struct config_llog_data *cld;
	struct config_llog_data *cld_sptlrpc = NULL;
	struct config_llog_data *cld_recover = NULL;
	int rc = 0;
	ENTRY;

	cld = config_log_find(logname, cfg);
	if (cld == NULL)
		RETURN(-ENOENT);

	mutex_lock(&cld->cld_lock);
	/*
	 * if cld_stopping is set, it means we didn't start the log thus
	 * not owning the start ref. this can happen after previous umount:
	 * the cld still hanging there waiting for lock cancel, and we
	 * remount again but failed in the middle and call log_end without
	 * calling start_log.
	 */
	if (unlikely(cld->cld_stopping)) {
		mutex_unlock(&cld->cld_lock);
		/* drop the ref from the find */
		config_log_put(cld);
		RETURN(rc);
	}

	cld->cld_stopping = 1;

	cld_recover = cld->cld_recover;
	cld->cld_recover = NULL;
	mutex_unlock(&cld->cld_lock);

	if (cld_recover) {
		mutex_lock(&cld_recover->cld_lock);
		cld_recover->cld_stopping = 1;
		mutex_unlock(&cld_recover->cld_lock);
		config_log_put(cld_recover);
	}

	spin_lock(&config_list_lock);
	cld_sptlrpc = cld->cld_sptlrpc;
	cld->cld_sptlrpc = NULL;
	spin_unlock(&config_list_lock);

	if (cld_sptlrpc)
		config_log_put(cld_sptlrpc);

	/* drop the ref from the find */
	config_log_put(cld);
	/* drop the start ref */
	config_log_put(cld);

	CDEBUG(D_MGC, "end config log %s (%d)\n", logname ? logname : "client",
	       rc);
	RETURN(rc);
}

int lprocfs_mgc_rd_ir_state(char *page, char **start, off_t off,
			    int count, int *eof, void *data)
{
	struct obd_device       *obd = data;
	struct obd_import       *imp = obd->u.cli.cl_import;
	struct obd_connect_data *ocd = &imp->imp_connect_data;
	struct config_llog_data *cld;
	int rc = 0;
	ENTRY;

	rc = snprintf(page, count, "imperative_recovery: %s\n",
		      OCD_HAS_FLAG(ocd, IMP_RECOV) ? "ENABLED" : "DISABLED");
	rc += snprintf(page + rc, count - rc, "client_state:\n");

	spin_lock(&config_list_lock);
	list_for_each_entry(cld, &config_llog_list, cld_list_chain) {
		if (cld->cld_recover == NULL)
			continue;
		rc += snprintf(page + rc, count - rc,
			       "    - { client: %s, nidtbl_version: %u }\n",
			       cld->cld_logname,
			       cld->cld_recover->cld_cfg.cfg_last_idx);
	}
	spin_unlock(&config_list_lock);

	RETURN(rc);
}

/* reenqueue any lost locks */
#define RQ_RUNNING 0x1
#define RQ_NOW     0x2
#define RQ_LATER   0x4
#define RQ_STOP    0x8
static int		    rq_state = 0;
static wait_queue_head_t	    rq_waitq;
static DECLARE_COMPLETION(rq_exit);

static void do_requeue(struct config_llog_data *cld)
{
	ENTRY;
	LASSERT(atomic_read(&cld->cld_refcount) > 0);

	/* Do not run mgc_process_log on a disconnected export or an
	   export which is being disconnected. Take the client
	   semaphore to make the check non-racy. */
	down_read(&cld->cld_mgcexp->exp_obd->u.cli.cl_sem);
	if (cld->cld_mgcexp->exp_obd->u.cli.cl_conn_count != 0) {
		CDEBUG(D_MGC, "updating log %s\n", cld->cld_logname);
		mgc_process_log(cld->cld_mgcexp->exp_obd, cld);
	} else {
		CDEBUG(D_MGC, "disconnecting, won't update log %s\n",
		       cld->cld_logname);
	}
	up_read(&cld->cld_mgcexp->exp_obd->u.cli.cl_sem);

	EXIT;
}

/* this timeout represents how many seconds MGC should wait before
 * requeue config and recover lock to the MGS. We need to randomize this
 * in order to not flood the MGS.
 */
#define MGC_TIMEOUT_MIN_SECONDS   5
#define MGC_TIMEOUT_RAND_CENTISEC 0x1ff /* ~500 */

static int mgc_requeue_thread(void *data)
{
	int rc = 0;
	ENTRY;

	CDEBUG(D_MGC, "Starting requeue thread\n");

	/* Keep trying failed locks periodically */
	spin_lock(&config_list_lock);
	rq_state |= RQ_RUNNING;
	while (1) {
		struct l_wait_info lwi;
		struct config_llog_data *cld, *cld_prev;
		int rand = cfs_rand() & MGC_TIMEOUT_RAND_CENTISEC;
		int stopped = !!(rq_state & RQ_STOP);
		int to;

		/* Any new or requeued lostlocks will change the state */
		rq_state &= ~(RQ_NOW | RQ_LATER);
		spin_unlock(&config_list_lock);

		/* Always wait a few seconds to allow the server who
		   caused the lock revocation to finish its setup, plus some
		   random so everyone doesn't try to reconnect at once. */
		to = MGC_TIMEOUT_MIN_SECONDS * HZ;
		to += rand * HZ / 100; /* rand is centi-seconds */
		lwi = LWI_TIMEOUT(to, NULL, NULL);
		l_wait_event(rq_waitq, rq_state & RQ_STOP, &lwi);

		/*
		 * iterate & processing through the list. for each cld, process
		 * its depending sptlrpc cld firstly (if any) and then itself.
		 *
		 * it's guaranteed any item in the list must have
		 * reference > 0; and if cld_lostlock is set, at
		 * least one reference is taken by the previous enqueue.
		 */
		cld_prev = NULL;

		spin_lock(&config_list_lock);
		list_for_each_entry(cld, &config_llog_list,
					cld_list_chain) {
			if (!cld->cld_lostlock)
				continue;

			spin_unlock(&config_list_lock);

			LASSERT(atomic_read(&cld->cld_refcount) > 0);

			/* Whether we enqueued again or not in mgc_process_log,
			 * we're done with the ref from the old enqueue */
			if (cld_prev)
				config_log_put(cld_prev);
			cld_prev = cld;

			cld->cld_lostlock = 0;
			if (likely(!stopped))
				do_requeue(cld);

			spin_lock(&config_list_lock);
		}
		spin_unlock(&config_list_lock);
		if (cld_prev)
			config_log_put(cld_prev);

		/* break after scanning the list so that we can drop
		 * refcount to losing lock clds */
		if (unlikely(stopped)) {
			spin_lock(&config_list_lock);
			break;
		}

		/* Wait a bit to see if anyone else needs a requeue */
		lwi = (struct l_wait_info) { 0 };
		l_wait_event(rq_waitq, rq_state & (RQ_NOW | RQ_STOP),
			     &lwi);
		spin_lock(&config_list_lock);
	}
	/* spinlock and while guarantee RQ_NOW and RQ_LATER are not set */
	rq_state &= ~RQ_RUNNING;
	spin_unlock(&config_list_lock);

	complete(&rq_exit);

	CDEBUG(D_MGC, "Ending requeue thread\n");
	RETURN(rc);
}

/* Add a cld to the list to requeue.  Start the requeue thread if needed.
   We are responsible for dropping the config log reference from here on out. */
static void mgc_requeue_add(struct config_llog_data *cld)
{
	ENTRY;

	CDEBUG(D_INFO, "log %s: requeue (r=%d sp=%d st=%x)\n",
	       cld->cld_logname, atomic_read(&cld->cld_refcount),
	       cld->cld_stopping, rq_state);
	LASSERT(atomic_read(&cld->cld_refcount) > 0);

	mutex_lock(&cld->cld_lock);
	if (cld->cld_stopping || cld->cld_lostlock) {
		mutex_unlock(&cld->cld_lock);
		RETURN_EXIT;
	}
	/* this refcount will be released in mgc_requeue_thread. */
	config_log_get(cld);
	cld->cld_lostlock = 1;
	mutex_unlock(&cld->cld_lock);

	/* Hold lock for rq_state */
	spin_lock(&config_list_lock);
	if (rq_state & RQ_STOP) {
		spin_unlock(&config_list_lock);
		cld->cld_lostlock = 0;
		config_log_put(cld);
	} else {
		rq_state |= RQ_NOW;
		spin_unlock(&config_list_lock);
		wake_up(&rq_waitq);
	}
	EXIT;
}

/********************** class fns **********************/

static int mgc_fs_setup(struct obd_device *obd, struct super_block *sb,
			struct vfsmount *mnt)
{
	struct lvfs_run_ctxt saved;
	struct lustre_sb_info *lsi = s2lsi(sb);
	struct client_obd *cli = &obd->u.cli;
	struct dentry *dentry;
	char *label;
	int err = 0;
	ENTRY;

	LASSERT(lsi);
	LASSERT(lsi->lsi_srv_mnt == mnt);

	/* The mgc fs exclusion sem. Only one fs can be setup at a time. */
	down(&cli->cl_mgc_sem);

	cfs_cleanup_group_info();

	obd->obd_fsops = fsfilt_get_ops(lsi->lsi_fstype);
	if (IS_ERR(obd->obd_fsops)) {
		up(&cli->cl_mgc_sem);
		CERROR("%s: No fstype %s: rc = %ld\n", lsi->lsi_fstype,
		       obd->obd_name, PTR_ERR(obd->obd_fsops));
		RETURN(PTR_ERR(obd->obd_fsops));
	}

	cli->cl_mgc_vfsmnt = mnt;
	err = fsfilt_setup(obd, mnt->mnt_sb);
	if (err)
		GOTO(err_ops, err);

	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();

	push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
	dentry = ll_lookup_one_len(MOUNT_CONFIGS_DIR, cfs_fs_pwd(current->fs),
				   strlen(MOUNT_CONFIGS_DIR));
	pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
	if (IS_ERR(dentry)) {
		err = PTR_ERR(dentry);
		CERROR("cannot lookup %s directory: rc = %d\n",
		       MOUNT_CONFIGS_DIR, err);
		GOTO(err_ops, err);
	}
	cli->cl_mgc_configs_dir = dentry;

	/* We take an obd ref to insure that we can't get to mgc_cleanup
	   without calling mgc_fs_cleanup first. */
	class_incref(obd, "mgc_fs", obd);

	label = fsfilt_get_label(obd, mnt->mnt_sb);
	if (label)
		CDEBUG(D_MGC, "MGC using disk labelled=%s\n", label);

	/* We keep the cl_mgc_sem until mgc_fs_cleanup */
	RETURN(0);

err_ops:
	fsfilt_put_ops(obd->obd_fsops);
	obd->obd_fsops = NULL;
	cli->cl_mgc_vfsmnt = NULL;
	up(&cli->cl_mgc_sem);
	RETURN(err);
}

static int mgc_fs_cleanup(struct obd_device *obd)
{
	struct client_obd *cli = &obd->u.cli;
	int rc = 0;
	ENTRY;

	LASSERT(cli->cl_mgc_vfsmnt != NULL);

	if (cli->cl_mgc_configs_dir != NULL) {
		struct lvfs_run_ctxt saved;
		push_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
		l_dput(cli->cl_mgc_configs_dir);
		cli->cl_mgc_configs_dir = NULL;
		pop_ctxt(&saved, &obd->obd_lvfs_ctxt, NULL);
		class_decref(obd, "mgc_fs", obd);
	}

	cli->cl_mgc_vfsmnt = NULL;
	if (obd->obd_fsops)
		fsfilt_put_ops(obd->obd_fsops);

	up(&cli->cl_mgc_sem);

	RETURN(rc);
}

static atomic_t mgc_count = ATOMIC_INIT(0);
static int mgc_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:
		if (atomic_dec_and_test(&mgc_count)) {
			int running;
			/* stop requeue thread */
			spin_lock(&config_list_lock);
			running = rq_state & RQ_RUNNING;
			if (running)
				rq_state |= RQ_STOP;
			spin_unlock(&config_list_lock);
			if (running) {
				wake_up(&rq_waitq);
				wait_for_completion(&rq_exit);
			}
		}
		obd_cleanup_client_import(obd);
		rc = obd_llog_finish(obd, 0);
		if (rc != 0)
			CERROR("failed to cleanup llogging subsystems\n");
		break;
	}
	RETURN(rc);
}

static int mgc_cleanup(struct obd_device *obd)
{
	struct client_obd *cli = &obd->u.cli;
	int rc;
	ENTRY;

	LASSERT(cli->cl_mgc_vfsmnt == NULL);

	/* COMPAT_146 - old config logs may have added profiles we don't
	   know about */
	if (obd->obd_type->typ_refcnt <= 1)
		/* Only for the last mgc */
		class_del_profiles();

	lprocfs_obd_cleanup(obd);
	ptlrpcd_decref();

	rc = client_obd_cleanup(obd);
	RETURN(rc);
}

static int mgc_setup(struct obd_device *obd, struct lustre_cfg *lcfg)
{
	struct lprocfs_static_vars lvars;
	int rc;
	ENTRY;

	ptlrpcd_addref();

	rc = client_obd_setup(obd, lcfg);
	if (rc)
		GOTO(err_decref, rc);

	rc = obd_llog_init(obd, &obd->obd_olg, obd, NULL);
	if (rc) {
		CERROR("failed to setup llogging subsystems\n");
		GOTO(err_cleanup, rc);
	}

	lprocfs_mgc_init_vars(&lvars);
	lprocfs_obd_setup(obd, lvars.obd_vars);
	sptlrpc_lprocfs_cliobd_attach(obd);

	if (atomic_inc_return(&mgc_count) == 1) {
		rq_state = 0;
		init_waitqueue_head(&rq_waitq);

		/* start requeue thread */
		rc = PTR_ERR(kthread_run(mgc_requeue_thread, NULL,
					     "ll_cfg_requeue"));
		if (IS_ERR_VALUE(rc)) {
			CERROR("%s: Cannot start requeue thread (%d),"
			       "no more log updates!\n",
			       obd->obd_name, rc);
			GOTO(err_cleanup, rc);
		}
		/* rc is the task_struct pointer of mgc_requeue_thread. */
		rc = 0;
	}

	RETURN(rc);

err_cleanup:
	client_obd_cleanup(obd);
err_decref:
	ptlrpcd_decref();
	RETURN(rc);
}

/* based on ll_mdc_blocking_ast */
static int mgc_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc,
			    void *data, int flag)
{
	struct lustre_handle lockh;
	struct config_llog_data *cld = (struct config_llog_data *)data;
	int rc = 0;
	ENTRY;

	switch (flag) {
	case LDLM_CB_BLOCKING:
		/* mgs wants the lock, give it up... */
		LDLM_DEBUG(lock, "MGC blocking CB");
		ldlm_lock2handle(lock, &lockh);
		rc = ldlm_cli_cancel(&lockh, LCF_ASYNC);
		break;
	case LDLM_CB_CANCELING:
		/* We've given up the lock, prepare ourselves to update. */
		LDLM_DEBUG(lock, "MGC cancel CB");

		CDEBUG(D_MGC, "Lock res "LPX64" (%.8s)\n",
		       lock->l_resource->lr_name.name[0],
		       (char *)&lock->l_resource->lr_name.name[0]);

		if (!cld) {
			CDEBUG(D_INFO, "missing data, won't requeue\n");
			break;
		}

		/* held at mgc_process_log(). */
		LASSERT(atomic_read(&cld->cld_refcount) > 0);
		/* Are we done with this log? */
		if (cld->cld_stopping) {
			CDEBUG(D_MGC, "log %s: stopping, won't requeue\n",
			       cld->cld_logname);
			config_log_put(cld);
			break;
		}
		/* Make sure not to re-enqueue when the mgc is stopping
		   (we get called from client_disconnect_export) */
		if (!lock->l_conn_export ||
		    !lock->l_conn_export->exp_obd->u.cli.cl_conn_count) {
			CDEBUG(D_MGC, "log %.8s: disconnecting, won't requeue\n",
			       cld->cld_logname);
			config_log_put(cld);
			break;
		}

		/* Re-enqueue now */
		mgc_requeue_add(cld);
		config_log_put(cld);
		break;
	default:
		LBUG();
	}

	RETURN(rc);
}

/* Not sure where this should go... */
#define  MGC_ENQUEUE_LIMIT 50
#define  MGC_TARGET_REG_LIMIT 10
#define  MGC_SEND_PARAM_LIMIT 10

/* Send parameter to MGS*/
static int mgc_set_mgs_param(struct obd_export *exp,
			     struct mgs_send_param *msp)
{
	struct ptlrpc_request *req;
	struct mgs_send_param *req_msp, *rep_msp;
	int rc;
	ENTRY;

	req = ptlrpc_request_alloc_pack(class_exp2cliimp(exp),
					&RQF_MGS_SET_INFO, LUSTRE_MGS_VERSION,
					MGS_SET_INFO);
	if (!req)
		RETURN(-ENOMEM);

	req_msp = req_capsule_client_get(&req->rq_pill, &RMF_MGS_SEND_PARAM);
	if (!req_msp) {
		ptlrpc_req_finished(req);
		RETURN(-ENOMEM);
	}

	memcpy(req_msp, msp, sizeof(*req_msp));
	ptlrpc_request_set_replen(req);

	/* Limit how long we will wait for the enqueue to complete */
	req->rq_delay_limit = MGC_SEND_PARAM_LIMIT;
	rc = ptlrpc_queue_wait(req);
	if (!rc) {
		rep_msp = req_capsule_server_get(&req->rq_pill, &RMF_MGS_SEND_PARAM);
		memcpy(msp, rep_msp, sizeof(*rep_msp));
	}

	ptlrpc_req_finished(req);

	RETURN(rc);
}

/* Take a config lock so we can get cancel notifications */
static int mgc_enqueue(struct obd_export *exp, struct lov_stripe_md *lsm,
		       __u32 type, ldlm_policy_data_t *policy, __u32 mode,
		       __u64 *flags, void *bl_cb, void *cp_cb, void *gl_cb,
		       void *data, __u32 lvb_len, void *lvb_swabber,
		       struct lustre_handle *lockh)
{
	struct config_llog_data *cld = (struct config_llog_data *)data;
	struct ldlm_enqueue_info einfo = { type, mode, mgc_blocking_ast,
			 ldlm_completion_ast, NULL, NULL, NULL };
	struct ptlrpc_request *req;
	int short_limit = cld_is_sptlrpc(cld);
	int rc;
	ENTRY;

	CDEBUG(D_MGC, "Enqueue for %s (res "LPX64")\n", cld->cld_logname,
	       cld->cld_resid.name[0]);

	/* We need a callback for every lockholder, so don't try to
	   ldlm_lock_match (see rev 1.1.2.11.2.47) */
	req = ptlrpc_request_alloc_pack(class_exp2cliimp(exp),
					&RQF_LDLM_ENQUEUE, LUSTRE_DLM_VERSION,
					LDLM_ENQUEUE);
	if (req == NULL)
		RETURN(-ENOMEM);

	req_capsule_set_size(&req->rq_pill, &RMF_DLM_LVB, RCL_SERVER, 0);
	ptlrpc_request_set_replen(req);

	/* check if this is server or client */
	if (cld->cld_cfg.cfg_sb) {
		struct lustre_sb_info *lsi = s2lsi(cld->cld_cfg.cfg_sb);
		if (lsi && IS_SERVER(lsi))
			short_limit = 1;
	}
	/* Limit how long we will wait for the enqueue to complete */
	req->rq_delay_limit = short_limit ? 5 : MGC_ENQUEUE_LIMIT;
	rc = ldlm_cli_enqueue(exp, &req, &einfo, &cld->cld_resid, NULL, flags,
			      NULL, 0, LVB_T_NONE, lockh, 0);
	/* A failed enqueue should still call the mgc_blocking_ast,
	   where it will be requeued if needed ("grant failed"). */
	ptlrpc_req_finished(req);
	RETURN(rc);
}

static int mgc_cancel(struct obd_export *exp, struct lov_stripe_md *md,
		      __u32 mode, struct lustre_handle *lockh)
{
	ENTRY;

	ldlm_lock_decref(lockh, mode);

	RETURN(0);
}

static void mgc_notify_active(struct obd_device *unused)
{
	/* wakeup mgc_requeue_thread to requeue mgc lock */
	spin_lock(&config_list_lock);
	rq_state |= RQ_NOW;
	spin_unlock(&config_list_lock);
	wake_up(&rq_waitq);

	/* TODO: Help the MGS rebuild nidtbl. -jay */
}

/* Send target_reg message to MGS */
static int mgc_target_register(struct obd_export *exp,
			       struct mgs_target_info *mti)
{
	struct ptlrpc_request  *req;
	struct mgs_target_info *req_mti, *rep_mti;
	int		     rc;
	ENTRY;

	req = ptlrpc_request_alloc_pack(class_exp2cliimp(exp),
					&RQF_MGS_TARGET_REG, LUSTRE_MGS_VERSION,
					MGS_TARGET_REG);
	if (req == NULL)
		RETURN(-ENOMEM);

	req_mti = req_capsule_client_get(&req->rq_pill, &RMF_MGS_TARGET_INFO);
	if (!req_mti) {
		ptlrpc_req_finished(req);
		RETURN(-ENOMEM);
	}

	memcpy(req_mti, mti, sizeof(*req_mti));
	ptlrpc_request_set_replen(req);
	CDEBUG(D_MGC, "register %s\n", mti->mti_svname);
	/* Limit how long we will wait for the enqueue to complete */
	req->rq_delay_limit = MGC_TARGET_REG_LIMIT;

	rc = ptlrpc_queue_wait(req);
	if (!rc) {
		rep_mti = req_capsule_server_get(&req->rq_pill,
						 &RMF_MGS_TARGET_INFO);
		memcpy(mti, rep_mti, sizeof(*rep_mti));
		CDEBUG(D_MGC, "register %s got index = %d\n",
		       mti->mti_svname, mti->mti_stripe_index);
	}
	ptlrpc_req_finished(req);

	RETURN(rc);
}

int mgc_set_info_async(const struct lu_env *env, struct obd_export *exp,
		       obd_count keylen, void *key, obd_count vallen,
		       void *val, struct ptlrpc_request_set *set)
{
	int rc = -EINVAL;
	ENTRY;

	/* Turn off initial_recov after we try all backup servers once */
	if (KEY_IS(KEY_INIT_RECOV_BACKUP)) {
		struct obd_import *imp = class_exp2cliimp(exp);
		int value;
		if (vallen != sizeof(int))
			RETURN(-EINVAL);
		value = *(int *)val;
		CDEBUG(D_MGC, "InitRecov %s %d/d%d:i%d:r%d:or%d:%s\n",
		       imp->imp_obd->obd_name, value,
		       imp->imp_deactive, imp->imp_invalid,
		       imp->imp_replayable, imp->imp_obd->obd_replayable,
		       ptlrpc_import_state_name(imp->imp_state));
		/* Resurrect if we previously died */
		if ((imp->imp_state != LUSTRE_IMP_FULL &&
		     imp->imp_state != LUSTRE_IMP_NEW) || value > 1)
			ptlrpc_reconnect_import(imp);
		RETURN(0);
	}
	/* FIXME move this to mgc_process_config */
	if (KEY_IS(KEY_REGISTER_TARGET)) {
		struct mgs_target_info *mti;
		if (vallen != sizeof(struct mgs_target_info))
			RETURN(-EINVAL);
		mti = (struct mgs_target_info *)val;
		CDEBUG(D_MGC, "register_target %s %#x\n",
		       mti->mti_svname, mti->mti_flags);
		rc =  mgc_target_register(exp, mti);
		RETURN(rc);
	}
	if (KEY_IS(KEY_SET_FS)) {
		struct super_block *sb = (struct super_block *)val;
		struct lustre_sb_info *lsi;
		if (vallen != sizeof(struct super_block))
			RETURN(-EINVAL);
		lsi = s2lsi(sb);
		rc = mgc_fs_setup(exp->exp_obd, sb, lsi->lsi_srv_mnt);
		if (rc) {
			CERROR("set_fs got %d\n", rc);
		}
		RETURN(rc);
	}
	if (KEY_IS(KEY_CLEAR_FS)) {
		if (vallen != 0)
			RETURN(-EINVAL);
		rc = mgc_fs_cleanup(exp->exp_obd);
		if (rc) {
			CERROR("clear_fs got %d\n", rc);
		}
		RETURN(rc);
	}
	if (KEY_IS(KEY_SET_INFO)) {
		struct mgs_send_param *msp;

		msp = (struct mgs_send_param *)val;
		rc =  mgc_set_mgs_param(exp, msp);
		RETURN(rc);
	}
	if (KEY_IS(KEY_MGSSEC)) {
		struct client_obd     *cli = &exp->exp_obd->u.cli;
		struct sptlrpc_flavor  flvr;

		/*
		 * empty string means using current flavor, if which haven't
		 * been set yet, set it as null.
		 *
		 * if flavor has been set previously, check the asking flavor
		 * must match the existing one.
		 */
		if (vallen == 0) {
			if (cli->cl_flvr_mgc.sf_rpc != SPTLRPC_FLVR_INVALID)
				RETURN(0);
			val = "null";
			vallen = 4;
		}

		rc = sptlrpc_parse_flavor(val, &flvr);
		if (rc) {
			CERROR("invalid sptlrpc flavor %s to MGS\n",
			       (char *) val);
			RETURN(rc);
		}

		/*
		 * caller already hold a mutex
		 */
		if (cli->cl_flvr_mgc.sf_rpc == SPTLRPC_FLVR_INVALID) {
			cli->cl_flvr_mgc = flvr;
		} else if (memcmp(&cli->cl_flvr_mgc, &flvr,
				  sizeof(flvr)) != 0) {
			char    str[20];

			sptlrpc_flavor2name(&cli->cl_flvr_mgc,
					    str, sizeof(str));
			LCONSOLE_ERROR("asking sptlrpc flavor %s to MGS but "
				       "currently %s is in use\n",
				       (char *) val, str);
			rc = -EPERM;
		}
		RETURN(rc);
	}

	RETURN(rc);
}

static int mgc_get_info(const struct lu_env *env, struct obd_export *exp,
			__u32 keylen, void *key, __u32 *vallen, void *val,
			struct lov_stripe_md *unused)
{
	int rc = -EINVAL;

	if (KEY_IS(KEY_CONN_DATA)) {
		struct obd_import *imp = class_exp2cliimp(exp);
		struct obd_connect_data *data = val;

		if (*vallen == sizeof(*data)) {
			*data = imp->imp_connect_data;
			rc = 0;
		}
	}

	return rc;
}

static int mgc_import_event(struct obd_device *obd,
			    struct obd_import *imp,
			    enum obd_import_event event)
{
	int rc = 0;

	LASSERT(imp->imp_obd == obd);
	CDEBUG(D_MGC, "import event %#x\n", event);

	switch (event) {
	case IMP_EVENT_DISCON:
		/* MGC imports should not wait for recovery */
		if (OCD_HAS_FLAG(&imp->imp_connect_data, IMP_RECOV))
			ptlrpc_pinger_ir_down();
		break;
	case IMP_EVENT_INACTIVE:
		break;
	case IMP_EVENT_INVALIDATE: {
		struct ldlm_namespace *ns = obd->obd_namespace;
		ldlm_namespace_cleanup(ns, LDLM_FL_LOCAL_ONLY);
		break;
	}
	case IMP_EVENT_ACTIVE:
		CDEBUG(D_INFO, "%s: Reactivating import\n", obd->obd_name);
		/* Clearing obd_no_recov allows us to continue pinging */
		obd->obd_no_recov = 0;
		mgc_notify_active(obd);
		if (OCD_HAS_FLAG(&imp->imp_connect_data, IMP_RECOV))
			ptlrpc_pinger_ir_up();
		break;
	case IMP_EVENT_OCD:
		break;
	case IMP_EVENT_DEACTIVATE:
	case IMP_EVENT_ACTIVATE:
		break;
	default:
		CERROR("Unknown import event %#x\n", event);
		LBUG();
	}
	RETURN(rc);
}

static int mgc_llog_init(struct obd_device *obd, struct obd_llog_group *olg,
			 struct obd_device *tgt, int *index)
{
	struct llog_ctxt *ctxt;
	int rc;
	ENTRY;

	LASSERT(olg == &obd->obd_olg);


	rc = llog_setup(NULL, obd, olg, LLOG_CONFIG_REPL_CTXT, tgt,
			&llog_client_ops);
	if (rc)
		GOTO(out, rc);

	ctxt = llog_group_get_ctxt(olg, LLOG_CONFIG_REPL_CTXT);
	if (!ctxt)
		GOTO(out, rc = -ENODEV);

	llog_initiator_connect(ctxt);
	llog_ctxt_put(ctxt);

	RETURN(0);
out:
	ctxt = llog_get_context(obd, LLOG_CONFIG_ORIG_CTXT);
	if (ctxt)
		llog_cleanup(NULL, ctxt);
	RETURN(rc);
}

static int mgc_llog_finish(struct obd_device *obd, int count)
{
	struct llog_ctxt *ctxt;

	ENTRY;

	ctxt = llog_get_context(obd, LLOG_CONFIG_REPL_CTXT);
	if (ctxt)
		llog_cleanup(NULL, ctxt);

	ctxt = llog_get_context(obd, LLOG_CONFIG_ORIG_CTXT);
	if (ctxt)
		llog_cleanup(NULL, ctxt);
	RETURN(0);
}

enum {
	CONFIG_READ_NRPAGES_INIT = 1 << (20 - PAGE_CACHE_SHIFT),
	CONFIG_READ_NRPAGES      = 4
};

static int mgc_apply_recover_logs(struct obd_device *mgc,
				  struct config_llog_data *cld,
				  __u64 max_version,
				  void *data, int datalen, bool mne_swab)
{
	struct config_llog_instance *cfg = &cld->cld_cfg;
	struct lustre_sb_info       *lsi = s2lsi(cfg->cfg_sb);
	struct mgs_nidtbl_entry *entry;
	struct lustre_cfg       *lcfg;
	struct lustre_cfg_bufs   bufs;
	u64   prev_version = 0;
	char *inst;
	char *buf;
	int   bufsz;
	int   pos;
	int   rc  = 0;
	int   off = 0;
	ENTRY;

	LASSERT(cfg->cfg_instance != NULL);
	LASSERT(cfg->cfg_sb == cfg->cfg_instance);

	OBD_ALLOC(inst, PAGE_CACHE_SIZE);
	if (inst == NULL)
		RETURN(-ENOMEM);

	if (!IS_SERVER(lsi)) {
		pos = snprintf(inst, PAGE_CACHE_SIZE, "%p", cfg->cfg_instance);
		if (pos >= PAGE_CACHE_SIZE) {
			OBD_FREE(inst, PAGE_CACHE_SIZE);
			return -E2BIG;
		}
	} else {
		LASSERT(IS_MDT(lsi));
		rc = server_name2svname(lsi->lsi_svname, inst, NULL,
					PAGE_CACHE_SIZE);
		if (rc) {
			OBD_FREE(inst, PAGE_CACHE_SIZE);
			RETURN(-EINVAL);
		}
		pos = strlen(inst);
	}

	++pos;
	buf   = inst + pos;
	bufsz = PAGE_CACHE_SIZE - pos;

	while (datalen > 0) {
		int   entry_len = sizeof(*entry);
		int   is_ost;
		struct obd_device *obd;
		char *obdname;
		char *cname;
		char *params;
		char *uuid;

		rc = -EINVAL;
		if (datalen < sizeof(*entry))
			break;

		entry = (typeof(entry))(data + off);

		/* sanity check */
		if (entry->mne_nid_type != 0) /* only support type 0 for ipv4 */
			break;
		if (entry->mne_nid_count == 0) /* at least one nid entry */
			break;
		if (entry->mne_nid_size != sizeof(lnet_nid_t))
			break;

		entry_len += entry->mne_nid_count * entry->mne_nid_size;
		if (datalen < entry_len) /* must have entry_len at least */
			break;

		/* Keep this swab for normal mixed endian handling. LU-1644 */
		if (mne_swab)
			lustre_swab_mgs_nidtbl_entry(entry);
		if (entry->mne_length > PAGE_CACHE_SIZE) {
			CERROR("MNE too large (%u)\n", entry->mne_length);
			break;
		}

		if (entry->mne_length < entry_len)
			break;

		off     += entry->mne_length;
		datalen -= entry->mne_length;
		if (datalen < 0)
			break;

		if (entry->mne_version > max_version) {
			CERROR("entry index(%lld) is over max_index(%lld)\n",
			       entry->mne_version, max_version);
			break;
		}

		if (prev_version >= entry->mne_version) {
			CERROR("index unsorted, prev %lld, now %lld\n",
			       prev_version, entry->mne_version);
			break;
		}
		prev_version = entry->mne_version;

		/*
		 * Write a string with format "nid::instance" to
		 * lustre/<osc|mdc>/<target>-<osc|mdc>-<instance>/import.
		 */

		is_ost = entry->mne_type == LDD_F_SV_TYPE_OST;
		memset(buf, 0, bufsz);
		obdname = buf;
		pos = 0;

		/* lustre-OST0001-osc-<instance #> */
		strcpy(obdname, cld->cld_logname);
		cname = strrchr(obdname, '-');
		if (cname == NULL) {
			CERROR("mgc %s: invalid logname %s\n",
			       mgc->obd_name, obdname);
			break;
		}

		pos = cname - obdname;
		obdname[pos] = 0;
		pos += sprintf(obdname + pos, "-%s%04x",
				  is_ost ? "OST" : "MDT", entry->mne_index);

		cname = is_ost ? "osc" : "mdc",
		pos += sprintf(obdname + pos, "-%s-%s", cname, inst);
		lustre_cfg_bufs_reset(&bufs, obdname);

		/* find the obd by obdname */
		obd = class_name2obd(obdname);
		if (obd == NULL) {
			CDEBUG(D_INFO, "mgc %s: cannot find obdname %s\n",
			       mgc->obd_name, obdname);
			rc = 0;
			/* this is a safe race, when the ost is starting up...*/
			continue;
		}

		/* osc.import = "connection=<Conn UUID>::<target instance>" */
		++pos;
		params = buf + pos;
		pos += sprintf(params, "%s.import=%s", cname, "connection=");
		uuid = buf + pos;

		down_read(&obd->u.cli.cl_sem);
		if (obd->u.cli.cl_import == NULL) {
			/* client does not connect to the OST yet */
			up_read(&obd->u.cli.cl_sem);
			rc = 0;
			continue;
		}

		/* TODO: iterate all nids to find one */
		/* find uuid by nid */
		rc = client_import_find_conn(obd->u.cli.cl_import,
					     entry->u.nids[0],
					     (struct obd_uuid *)uuid);
		up_read(&obd->u.cli.cl_sem);
		if (rc < 0) {
			CERROR("mgc: cannot find uuid by nid %s\n",
			       libcfs_nid2str(entry->u.nids[0]));
			break;
		}

		CDEBUG(D_INFO, "Find uuid %s by nid %s\n",
		       uuid, libcfs_nid2str(entry->u.nids[0]));

		pos += strlen(uuid);
		pos += sprintf(buf + pos, "::%u", entry->mne_instance);
		LASSERT(pos < bufsz);

		lustre_cfg_bufs_set_string(&bufs, 1, params);

		rc = -ENOMEM;
		lcfg = lustre_cfg_new(LCFG_PARAM, &bufs);
		if (lcfg == NULL) {
			CERROR("mgc: cannot allocate memory\n");
			break;
		}

		CDEBUG(D_INFO, "ir apply logs "LPD64"/"LPD64" for %s -> %s\n",
		       prev_version, max_version, obdname, params);

		rc = class_process_config(lcfg);
		lustre_cfg_free(lcfg);
		if (rc)
			CDEBUG(D_INFO, "process config for %s error %d\n",
			       obdname, rc);

		/* continue, even one with error */
	}

	OBD_FREE(inst, PAGE_CACHE_SIZE);
	RETURN(rc);
}

/**
 * This function is called if this client was notified for target restarting
 * by the MGS. A CONFIG_READ RPC is going to send to fetch recovery logs.
 */
static int mgc_process_recover_log(struct obd_device *obd,
				   struct config_llog_data *cld)
{
	struct ptlrpc_request *req = NULL;
	struct config_llog_instance *cfg = &cld->cld_cfg;
	struct mgs_config_body *body;
	struct mgs_config_res  *res;
	struct ptlrpc_bulk_desc *desc;
	struct page **pages;
	int nrpages;
	bool eof = true;
	bool mne_swab = false;
	int i;
	int ealen;
	int rc;
	ENTRY;

	/* allocate buffer for bulk transfer.
	 * if this is the first time for this mgs to read logs,
	 * CONFIG_READ_NRPAGES_INIT will be used since it will read all logs
	 * once; otherwise, it only reads increment of logs, this should be
	 * small and CONFIG_READ_NRPAGES will be used.
	 */
	nrpages = CONFIG_READ_NRPAGES;
	if (cfg->cfg_last_idx == 0) /* the first time */
		nrpages = CONFIG_READ_NRPAGES_INIT;

	OBD_ALLOC(pages, sizeof(*pages) * nrpages);
	if (pages == NULL)
		GOTO(out, rc = -ENOMEM);

	for (i = 0; i < nrpages; i++) {
		pages[i] = alloc_page(GFP_IOFS);
		if (pages[i] == NULL)
			GOTO(out, rc = -ENOMEM);
	}

again:
	LASSERT(cld_is_recover(cld));
	LASSERT(mutex_is_locked(&cld->cld_lock));
	req = ptlrpc_request_alloc(class_exp2cliimp(cld->cld_mgcexp),
				   &RQF_MGS_CONFIG_READ);
	if (req == NULL)
		GOTO(out, rc = -ENOMEM);

	rc = ptlrpc_request_pack(req, LUSTRE_MGS_VERSION, MGS_CONFIG_READ);
	if (rc)
		GOTO(out, rc);

	/* pack request */
	body = req_capsule_client_get(&req->rq_pill, &RMF_MGS_CONFIG_BODY);
	LASSERT(body != NULL);
	LASSERT(sizeof(body->mcb_name) > strlen(cld->cld_logname));
	if (strlcpy(body->mcb_name, cld->cld_logname, sizeof(body->mcb_name))
	    >= sizeof(body->mcb_name))
		GOTO(out, rc = -E2BIG);
	body->mcb_offset = cfg->cfg_last_idx + 1;
	body->mcb_type   = cld->cld_type;
	body->mcb_bits   = PAGE_CACHE_SHIFT;
	body->mcb_units  = nrpages;

	/* allocate bulk transfer descriptor */
	desc = ptlrpc_prep_bulk_imp(req, nrpages, 1, BULK_PUT_SINK,
				    MGS_BULK_PORTAL);
	if (desc == NULL)
		GOTO(out, rc = -ENOMEM);

	for (i = 0; i < nrpages; i++)
		ptlrpc_prep_bulk_page_pin(desc, pages[i], 0, PAGE_CACHE_SIZE);

	ptlrpc_request_set_replen(req);
	rc = ptlrpc_queue_wait(req);
	if (rc)
		GOTO(out, rc);

	res = req_capsule_server_get(&req->rq_pill, &RMF_MGS_CONFIG_RES);
	if (res->mcr_size < res->mcr_offset)
		GOTO(out, rc = -EINVAL);

	/* always update the index even though it might have errors with
	 * handling the recover logs */
	cfg->cfg_last_idx = res->mcr_offset;
	eof = res->mcr_offset == res->mcr_size;

	CDEBUG(D_INFO, "Latest version "LPD64", more %d.\n",
	       res->mcr_offset, eof == false);

	ealen = sptlrpc_cli_unwrap_bulk_read(req, req->rq_bulk, 0);
	if (ealen < 0)
		GOTO(out, rc = ealen);

	if (ealen > nrpages << PAGE_CACHE_SHIFT)
		GOTO(out, rc = -EINVAL);

	if (ealen == 0) { /* no logs transferred */
		if (!eof)
			rc = -EINVAL;
		GOTO(out, rc);
	}

	mne_swab = !!ptlrpc_rep_need_swab(req);
#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(3, 2, 50, 0)
	/* This import flag means the server did an extra swab of IR MNE
	 * records (fixed in LU-1252), reverse it here if needed. LU-1644 */
	if (unlikely(req->rq_import->imp_need_mne_swab))
		mne_swab = !mne_swab;
#else
#warning "LU-1644: Remove old OBD_CONNECT_MNE_SWAB fixup and imp_need_mne_swab"
#endif

	for (i = 0; i < nrpages && ealen > 0; i++) {
		int rc2;
		void *ptr;

		ptr = kmap(pages[i]);
		rc2 = mgc_apply_recover_logs(obd, cld, res->mcr_offset, ptr,
					     min_t(int, ealen, PAGE_CACHE_SIZE),
					     mne_swab);
		kunmap(pages[i]);
		if (rc2 < 0) {
			CWARN("Process recover log %s error %d\n",
			      cld->cld_logname, rc2);
			break;
		}

		ealen -= PAGE_CACHE_SIZE;
	}

out:
	if (req)
		ptlrpc_req_finished(req);

	if (rc == 0 && !eof)
		goto again;

	if (pages) {
		for (i = 0; i < nrpages; i++) {
			if (pages[i] == NULL)
				break;
			__free_page(pages[i]);
		}
		OBD_FREE(pages, sizeof(*pages) * nrpages);
	}
	return rc;
}


/* local_only means it cannot get remote llogs */
static int mgc_process_cfg_log(struct obd_device *mgc,
			       struct config_llog_data *cld,
			       int local_only)
{
	struct llog_ctxt *ctxt, *lctxt = NULL;
	struct lvfs_run_ctxt *saved_ctxt;
	struct lustre_sb_info *lsi = NULL;
	int rc = 0, must_pop = 0;
	bool sptlrpc_started = false;

	ENTRY;

	LASSERT(cld);
	LASSERT(mutex_is_locked(&cld->cld_lock));

	/*
	 * local copy of sptlrpc log is controlled elsewhere, don't try to
	 * read it up here.
	 */
	if (cld_is_sptlrpc(cld) && local_only)
		RETURN(0);

	if (cld->cld_cfg.cfg_sb)
		lsi = s2lsi(cld->cld_cfg.cfg_sb);

	ctxt = llog_get_context(mgc, LLOG_CONFIG_REPL_CTXT);
	if (!ctxt) {
		CERROR("missing llog context\n");
		RETURN(-EINVAL);
	}

	OBD_ALLOC_PTR(saved_ctxt);
	if (saved_ctxt == NULL)
		RETURN(-ENOMEM);

	lctxt = llog_get_context(mgc, LLOG_CONFIG_ORIG_CTXT);

		if (local_only) { /* no local log at client side */
		GOTO(out_pop, rc = -EIO);
	}

	if (cld_is_sptlrpc(cld)) {
		sptlrpc_conf_log_update_begin(cld->cld_logname);
		sptlrpc_started = true;
	}

	/* logname and instance info should be the same, so use our
	   copy of the instance for the update.  The cfg_last_idx will
	   be updated here. */
	rc = class_config_parse_llog(NULL, ctxt, cld->cld_logname,
				     &cld->cld_cfg);
	EXIT;

out_pop:
	llog_ctxt_put(ctxt);
	if (lctxt)
		llog_ctxt_put(lctxt);
	if (must_pop)
		pop_ctxt(saved_ctxt, &mgc->obd_lvfs_ctxt, NULL);

	OBD_FREE_PTR(saved_ctxt);
	/*
	 * update settings on existing OBDs. doing it inside
	 * of llog_process_lock so no device is attaching/detaching
	 * in parallel.
	 * the logname must be <fsname>-sptlrpc
	 */
	if (sptlrpc_started) {
		LASSERT(cld_is_sptlrpc(cld));
		sptlrpc_conf_log_update_end(cld->cld_logname);
		class_notify_sptlrpc_conf(cld->cld_logname,
					  strlen(cld->cld_logname) -
					  strlen("-sptlrpc"));
	}

	RETURN(rc);
}

/** Get a config log from the MGS and process it.
 * This func is called for both clients and servers.
 * Copy the log locally before parsing it if appropriate (non-MGS server)
 */
int mgc_process_log(struct obd_device *mgc, struct config_llog_data *cld)
{
	struct lustre_handle lockh = { 0 };
	__u64 flags = LDLM_FL_NO_LRU;
	int rc = 0, rcl;
	ENTRY;

	LASSERT(cld);

	/* I don't want multiple processes running process_log at once --
	   sounds like badness.  It actually might be fine, as long as
	   we're not trying to update from the same log
	   simultaneously (in which case we should use a per-log sem.) */
	mutex_lock(&cld->cld_lock);
	if (cld->cld_stopping) {
		mutex_unlock(&cld->cld_lock);
		RETURN(0);
	}

	OBD_FAIL_TIMEOUT(OBD_FAIL_MGC_PAUSE_PROCESS_LOG, 20);

	CDEBUG(D_MGC, "Process log %s:%p from %d\n", cld->cld_logname,
	       cld->cld_cfg.cfg_instance, cld->cld_cfg.cfg_last_idx + 1);

	/* Get the cfg lock on the llog */
	rcl = mgc_enqueue(mgc->u.cli.cl_mgc_mgsexp, NULL, LDLM_PLAIN, NULL,
			  LCK_CR, &flags, NULL, NULL, NULL,
			  cld, 0, NULL, &lockh);
	if (rcl == 0) {
		/* Get the cld, it will be released in mgc_blocking_ast. */
		config_log_get(cld);
		rc = ldlm_lock_set_data(&lockh, (void *)cld);
		LASSERT(rc == 0);
	} else {
		CDEBUG(D_MGC, "Can't get cfg lock: %d\n", rcl);

		/* mark cld_lostlock so that it will requeue
		 * after MGC becomes available. */
		cld->cld_lostlock = 1;
		/* Get extra reference, it will be put in requeue thread */
		config_log_get(cld);
	}


	if (cld_is_recover(cld)) {
		rc = 0; /* this is not a fatal error for recover log */
		if (rcl == 0)
			rc = mgc_process_recover_log(mgc, cld);
	} else {
		rc = mgc_process_cfg_log(mgc, cld, rcl != 0);
	}

	CDEBUG(D_MGC, "%s: configuration from log '%s' %sed (%d).\n",
	       mgc->obd_name, cld->cld_logname, rc ? "fail" : "succeed", rc);

	mutex_unlock(&cld->cld_lock);

	/* Now drop the lock so MGS can revoke it */
	if (!rcl) {
		rcl = mgc_cancel(mgc->u.cli.cl_mgc_mgsexp, NULL,
				 LCK_CR, &lockh);
		if (rcl)
			CERROR("Can't drop cfg lock: %d\n", rcl);
	}

	RETURN(rc);
}


/** Called from lustre_process_log.
 * LCFG_LOG_START gets the config log from the MGS, processes it to start
 * any services, and adds it to the list logs to watch (follow).
 */
static int mgc_process_config(struct obd_device *obd, obd_count len, void *buf)
{
	struct lustre_cfg *lcfg = buf;
	struct config_llog_instance *cfg = NULL;
	char *logname;
	int rc = 0;
	ENTRY;

	switch(lcfg->lcfg_command) {
	case LCFG_LOV_ADD_OBD: {
		/* Overloading this cfg command: register a new target */
		struct mgs_target_info *mti;

		if (LUSTRE_CFG_BUFLEN(lcfg, 1) !=
		    sizeof(struct mgs_target_info))
			GOTO(out, rc = -EINVAL);

		mti = (struct mgs_target_info *)lustre_cfg_buf(lcfg, 1);
		CDEBUG(D_MGC, "add_target %s %#x\n",
		       mti->mti_svname, mti->mti_flags);
		rc = mgc_target_register(obd->u.cli.cl_mgc_mgsexp, mti);
		break;
	}
	case LCFG_LOV_DEL_OBD:
		/* Unregister has no meaning at the moment. */
		CERROR("lov_del_obd unimplemented\n");
		rc = -ENOSYS;
		break;
	case LCFG_SPTLRPC_CONF: {
		rc = sptlrpc_process_config(lcfg);
		break;
	}
	case LCFG_LOG_START: {
		struct config_llog_data *cld;
		struct super_block *sb;

		logname = lustre_cfg_string(lcfg, 1);
		cfg = (struct config_llog_instance *)lustre_cfg_buf(lcfg, 2);
		sb = *(struct super_block **)lustre_cfg_buf(lcfg, 3);

		CDEBUG(D_MGC, "parse_log %s from %d\n", logname,
		       cfg->cfg_last_idx);

		/* We're only called through here on the initial mount */
		rc = config_log_add(obd, logname, cfg, sb);
		if (rc)
			break;
		cld = config_log_find(logname, cfg);
		if (cld == NULL) {
			rc = -ENOENT;
			break;
		}

		/* COMPAT_146 */
		/* FIXME only set this for old logs!  Right now this forces
		   us to always skip the "inside markers" check */
		cld->cld_cfg.cfg_flags |= CFG_F_COMPAT146;

		rc = mgc_process_log(obd, cld);
		if (rc == 0 && cld->cld_recover != NULL) {
			if (OCD_HAS_FLAG(&obd->u.cli.cl_import->
					 imp_connect_data, IMP_RECOV)) {
				rc = mgc_process_log(obd, cld->cld_recover);
			} else {
				struct config_llog_data *cir = cld->cld_recover;
				cld->cld_recover = NULL;
				config_log_put(cir);
			}
			if (rc)
				CERROR("Cannot process recover llog %d\n", rc);
		}
		config_log_put(cld);

		break;
	}
	case LCFG_LOG_END: {
		logname = lustre_cfg_string(lcfg, 1);

		if (lcfg->lcfg_bufcount >= 2)
			cfg = (struct config_llog_instance *)lustre_cfg_buf(
				lcfg, 2);
		rc = config_log_end(logname, cfg);
		break;
	}
	default: {
		CERROR("Unknown command: %d\n", lcfg->lcfg_command);
		GOTO(out, rc = -EINVAL);

	}
	}
out:
	RETURN(rc);
}

struct obd_ops mgc_obd_ops = {
	.o_owner	= THIS_MODULE,
	.o_setup	= mgc_setup,
	.o_precleanup   = mgc_precleanup,
	.o_cleanup      = mgc_cleanup,
	.o_add_conn     = client_import_add_conn,
	.o_del_conn     = client_import_del_conn,
	.o_connect      = client_connect_import,
	.o_disconnect   = client_disconnect_export,
	//.o_enqueue      = mgc_enqueue,
	.o_cancel       = mgc_cancel,
	//.o_iocontrol    = mgc_iocontrol,
	.o_set_info_async = mgc_set_info_async,
	.o_get_info       = mgc_get_info,
	.o_import_event = mgc_import_event,
	.o_llog_init    = mgc_llog_init,
	.o_llog_finish  = mgc_llog_finish,
	.o_process_config = mgc_process_config,
};

int __init mgc_init(void)
{
	return class_register_type(&mgc_obd_ops, NULL, NULL,
				   LUSTRE_MGC_NAME, NULL);
}

static void /*__exit*/ mgc_exit(void)
{
	class_unregister_type(LUSTRE_MGC_NAME);
}

MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
MODULE_DESCRIPTION("Lustre Management Client");
MODULE_LICENSE("GPL");

module_init(mgc_init);
module_exit(mgc_exit);