/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
* vim:expandtab:shiftwidth=8:tabstop=8:
*
- * Portal-RPC reconnection and replay operations, for use in recovery.
+ * GPL HEADER START
+ *
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
- * Copyright (c) 2003 Cluster File Systems, Inc.
- * Authors: Phil Schwan <phil@clusterfs.com>
- * Mike Shaver <shaver@clusterfs.com>
+ * 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 file is part of the Lustre file system, http://www.lustre.org
- * Lustre is a trademark of Cluster File Systems, Inc.
+ * 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 may have signed or agreed to another license before downloading
- * this software. If so, you are bound by the terms and conditions
- * of that agreement, and the following does not apply to you. See the
- * LICENSE file included with this distribution for more information.
+ * 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
*
- * If you did not agree to a different license, then this copy of Lustre
- * is open source software; you can redistribute it and/or modify it
- * under the terms of version 2 of the GNU General Public License as
- * published by the Free Software Foundation.
+ * 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.
*
- * In either case, Lustre is distributed in the hope that it will be
- * useful, but WITHOUT ANY WARRANTY; without even the implied warranty
- * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
- * license text for more details.
+ * GPL HEADER END
+ */
+/*
+ * Copyright 2008 Sun Microsystems, Inc. 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/ptlrpc/pinger.c
+ *
+ * Portal-RPC reconnection and replay operations, for use in recovery.
*/
#ifndef __KERNEL__
#include "ptlrpc_internal.h"
struct semaphore pinger_sem;
-static struct list_head pinger_imports = CFS_LIST_HEAD_INIT(pinger_imports);
+static CFS_LIST_HEAD(pinger_imports);
+static struct list_head timeout_list = CFS_LIST_HEAD_INIT(timeout_list);
+struct ptlrpc_request *
+ptlrpc_prep_ping(struct obd_import *imp)
+{
+ struct ptlrpc_request *req;
+
+ req = ptlrpc_request_alloc_pack(imp, &RQF_OBD_PING,
+ LUSTRE_OBD_VERSION, OBD_PING);
+ if (req) {
+ ptlrpc_request_set_replen(req);
+ req->rq_no_resend = req->rq_no_delay = 1;
+ }
+ return req;
+}
+
+int ptlrpc_obd_ping(struct obd_device *obd)
+{
+ int rc;
+ struct ptlrpc_request *req;
+ ENTRY;
+
+ req = ptlrpc_prep_ping(obd->u.cli.cl_import);
+ if (req == NULL)
+ RETURN(-ENOMEM);
+
+ req->rq_send_state = LUSTRE_IMP_FULL;
+
+ rc = ptlrpc_queue_wait(req);
+
+ ptlrpc_req_finished(req);
+
+ RETURN(rc);
+}
+EXPORT_SYMBOL(ptlrpc_obd_ping);
int ptlrpc_ping(struct obd_import *imp)
{
struct ptlrpc_request *req;
ENTRY;
- req = ptlrpc_request_alloc_pack(imp, &RQF_OBD_PING, LUSTRE_OBD_VERSION,
- OBD_PING);
+ req = ptlrpc_prep_ping(imp);
if (req == NULL) {
CERROR("OOM trying to ping %s->%s\n",
imp->imp_obd->obd_uuid.uuid,
DEBUG_REQ(D_INFO, req, "pinging %s->%s",
imp->imp_obd->obd_uuid.uuid, obd2cli_tgt(imp->imp_obd));
- req->rq_no_resend = req->rq_no_delay = 1;
- ptlrpc_request_set_replen(req);
- req->rq_timeout = PING_INTERVAL;
- ptlrpcd_add_req(req);
+ ptlrpcd_add_req(req, PSCOPE_OTHER);
RETURN(0);
}
void ptlrpc_update_next_ping(struct obd_import *imp)
{
#ifdef ENABLE_PINGER
- imp->imp_next_ping = cfs_time_shift(
- (imp->imp_state == LUSTRE_IMP_DISCON ?
- RECONNECT_INTERVAL : PING_INTERVAL));
+ int time = PING_INTERVAL;
+ if (imp->imp_state == LUSTRE_IMP_DISCON) {
+ int dtime = max_t(int, CONNECTION_SWITCH_MIN,
+ AT_OFF ? 0 :
+ at_get(&imp->imp_at.iat_net_latency));
+ time = min(time, dtime);
+ }
+ imp->imp_next_ping = cfs_time_shift(time);
#endif /* ENABLE_PINGER */
}
imp->imp_next_ping = cfs_time_current();
}
+static inline int imp_is_deactive(struct obd_import *imp)
+{
+ return (imp->imp_deactive ||
+ OBD_FAIL_CHECK(OBD_FAIL_PTLRPC_IMP_DEACTIVE));
+}
+
static inline int ptlrpc_next_reconnect(struct obd_import *imp)
{
if (imp->imp_server_timeout)
static atomic_t suspend_timeouts = ATOMIC_INIT(0);
static cfs_time_t suspend_wakeup_time = 0;
+cfs_duration_t pinger_check_timeout(cfs_time_t time)
+{
+ struct timeout_item *item;
+ cfs_time_t timeout = PING_INTERVAL;
+
+ /* The timeout list is a increase order sorted list */
+ mutex_down(&pinger_sem);
+ list_for_each_entry(item, &timeout_list, ti_chain) {
+ int ti_timeout = item->ti_timeout;
+ if (timeout > ti_timeout)
+ timeout = ti_timeout;
+ break;
+ }
+ mutex_up(&pinger_sem);
+
+ return cfs_time_sub(cfs_time_add(time, cfs_time_seconds(timeout)),
+ cfs_time_current());
+}
+
#ifdef __KERNEL__
static wait_queue_head_t suspend_timeouts_waitq;
#endif
this_ping) && force == 0)
return;
- if (level == LUSTRE_IMP_DISCON && !imp->imp_deactive) {
+ if (level == LUSTRE_IMP_DISCON && !imp_is_deactive(imp)) {
/* wait at least a timeout before trying recovery again */
imp->imp_next_ping = ptlrpc_next_reconnect(imp);
ptlrpc_initiate_recovery(imp);
- } else if (level != LUSTRE_IMP_FULL ||
+ } else if (level != LUSTRE_IMP_FULL ||
imp->imp_obd->obd_no_recov ||
- imp->imp_deactive) {
+ imp_is_deactive(imp)) {
CDEBUG(D_HA, "not pinging %s (in recovery "
" or recovery disabled: %s)\n",
obd2cli_tgt(imp->imp_obd),
while (1) {
cfs_time_t this_ping = cfs_time_current();
struct l_wait_info lwi;
- cfs_duration_t time_to_next_ping;
+ cfs_duration_t time_to_next_wake;
+ struct timeout_item *item;
struct list_head *iter;
mutex_down(&pinger_sem);
+ list_for_each_entry(item, &timeout_list, ti_chain) {
+ item->ti_cb(item, item->ti_cb_data);
+ }
list_for_each(iter, &pinger_imports) {
struct obd_import *imp =
list_entry(iter, struct obd_import,
ptlrpc_pinger_process_import(imp, this_ping);
/* obd_timeout might have changed */
if (imp->imp_pingable && imp->imp_next_ping &&
- cfs_time_after(imp->imp_next_ping,
- cfs_time_add(this_ping,
+ cfs_time_after(imp->imp_next_ping,
+ cfs_time_add(this_ping,
cfs_time_seconds(PING_INTERVAL))))
ptlrpc_update_next_ping(imp);
}
obd_update_maxusage();
/* Wait until the next ping time, or until we're stopped. */
- time_to_next_ping = cfs_time_sub(cfs_time_add(this_ping,
- cfs_time_seconds(PING_INTERVAL)),
- cfs_time_current());
-
+ time_to_next_wake = pinger_check_timeout(this_ping);
/* The ping sent by ptlrpc_send_rpc may get sent out
say .01 second after this.
- ptlrpc_pinger_eending_on_import will then set the
+ ptlrpc_pinger_sending_on_import will then set the
next ping time to next_ping + .01 sec, which means
we will SKIP the next ping at next_ping, and the
ping will get sent 2 timeouts from now! Beware. */
- CDEBUG(D_INFO, "next ping in "CFS_DURATION_T" ("CFS_TIME_T")\n",
- time_to_next_ping,
- cfs_time_add(this_ping, cfs_time_seconds(PING_INTERVAL)));
- if (time_to_next_ping > 0) {
- lwi = LWI_TIMEOUT(max_t(cfs_duration_t, time_to_next_ping, cfs_time_seconds(1)),
- NULL, NULL);
+ CDEBUG(D_INFO, "next wakeup in "CFS_DURATION_T" ("CFS_TIME_T")\n",
+ time_to_next_wake,
+ cfs_time_add(this_ping, cfs_time_seconds(PING_INTERVAL)));
+ if (time_to_next_wake > 0) {
+ lwi = LWI_TIMEOUT(max_t(cfs_duration_t, time_to_next_wake, cfs_time_seconds(1)),
+ NULL, NULL);
l_wait_event(thread->t_ctl_waitq,
thread->t_flags & (SVC_STOPPING|SVC_EVENT),
&lwi);
RETURN(-ENOMEM);
cfs_waitq_init(&pinger_thread->t_ctl_waitq);
cfs_waitq_init(&suspend_timeouts_waitq);
-
+
d.name = "ll_ping";
d.thread = pinger_thread;
RETURN(0);
}
+int ptlrpc_pinger_remove_timeouts(void);
+
int ptlrpc_stop_pinger(void)
{
struct l_wait_info lwi = { 0 };
if (pinger_thread == NULL)
RETURN(-EALREADY);
+
+ ptlrpc_pinger_remove_timeouts();
mutex_down(&pinger_sem);
pinger_thread->t_flags = SVC_STOPPING;
cfs_waitq_signal(&pinger_thread->t_ctl_waitq);
RETURN(0);
}
+/**
+ * Register a timeout callback to the pinger list, and the callback will
+ * be called when timeout happens.
+ */
+struct timeout_item* ptlrpc_new_timeout(int time, enum timeout_event event,
+ timeout_cb_t cb, void *data)
+{
+ struct timeout_item *ti;
+
+ OBD_ALLOC_PTR(ti);
+ if (!ti)
+ return(NULL);
+
+ CFS_INIT_LIST_HEAD(&ti->ti_obd_list);
+ CFS_INIT_LIST_HEAD(&ti->ti_chain);
+ ti->ti_timeout = time;
+ ti->ti_event = event;
+ ti->ti_cb = cb;
+ ti->ti_cb_data = data;
+
+ return ti;
+}
+
+/**
+ * Register timeout event on the the pinger thread.
+ * Note: the timeout list is an sorted list with increased timeout value.
+ */
+static struct timeout_item*
+ptlrpc_pinger_register_timeout(int time, enum timeout_event event,
+ timeout_cb_t cb, void *data)
+{
+ struct timeout_item *item, *tmp;
+
+ LASSERT_SEM_LOCKED(&pinger_sem);
+
+ list_for_each_entry(item, &timeout_list, ti_chain)
+ if (item->ti_event == event)
+ goto out;
+
+ item = ptlrpc_new_timeout(time, event, cb, data);
+ if (item) {
+ list_for_each_entry_reverse(tmp, &timeout_list, ti_chain) {
+ if (tmp->ti_timeout < time) {
+ list_add(&item->ti_chain, &tmp->ti_chain);
+ goto out;
+ }
+ }
+ list_add(&item->ti_chain, &timeout_list);
+ }
+out:
+ return item;
+}
+
+/* Add a client_obd to the timeout event list, when timeout(@time)
+ * happens, the callback(@cb) will be called.
+ */
+int ptlrpc_add_timeout_client(int time, enum timeout_event event,
+ timeout_cb_t cb, void *data,
+ struct list_head *obd_list)
+{
+ struct timeout_item *ti;
+
+ mutex_down(&pinger_sem);
+ ti = ptlrpc_pinger_register_timeout(time, event, cb, data);
+ if (!ti) {
+ mutex_up(&pinger_sem);
+ return (-EINVAL);
+ }
+ list_add(obd_list, &ti->ti_obd_list);
+ mutex_up(&pinger_sem);
+ return 0;
+}
+
+int ptlrpc_del_timeout_client(struct list_head *obd_list,
+ enum timeout_event event)
+{
+ struct timeout_item *ti = NULL, *item;
+
+ if (list_empty(obd_list))
+ return 0;
+ mutex_down(&pinger_sem);
+ list_del_init(obd_list);
+ /**
+ * If there are no obd attached to the timeout event
+ * list, remove this timeout event from the pinger
+ */
+ list_for_each_entry(item, &timeout_list, ti_chain) {
+ if (item->ti_event == event) {
+ ti = item;
+ break;
+ }
+ }
+ LASSERTF(ti != NULL, "ti is NULL ! \n");
+ if (list_empty(&ti->ti_obd_list)) {
+ list_del(&ti->ti_chain);
+ OBD_FREE_PTR(ti);
+ }
+ mutex_up(&pinger_sem);
+ return 0;
+}
+
+int ptlrpc_pinger_remove_timeouts(void)
+{
+ struct timeout_item *item, *tmp;
+
+ mutex_down(&pinger_sem);
+ list_for_each_entry_safe(item, tmp, &timeout_list, ti_chain) {
+ LASSERT(list_empty(&item->ti_obd_list));
+ list_del(&item->ti_chain);
+ OBD_FREE_PTR(item);
+ }
+ mutex_up(&pinger_sem);
+ return 0;
+}
+
void ptlrpc_pinger_wake_up()
{
#ifdef ENABLE_PINGER
obd = pet_exp->exp_obd;
spin_unlock(&pet_lock);
- expire_time = cfs_time_current_sec() - (3 * obd_timeout / 2);
+ expire_time = cfs_time_current_sec() - PING_EVICT_TIMEOUT;
CDEBUG(D_HA, "evicting all exports of obd %s older than %ld\n",
obd->obd_name, expire_time);
struct ptlrpc_request *req;
struct ptlrpc_request_set *set;
struct list_head *iter;
+ struct obd_import *imp;
struct pinger_data *pd = &pinger_args;
int rc;
list_entry(iter, struct obd_import, imp_pinger_chain);
int generation, level;
- if (cfs_time_aftereq(pd->pd_this_ping,
+ if (cfs_time_aftereq(pd->pd_this_ping,
imp->imp_next_ping - 5 * CFS_TICK)) {
/* Add a ping. */
spin_lock(&imp->imp_lock);
req->rq_no_resend = 1;
ptlrpc_request_set_replen(req);
req->rq_send_state = LUSTRE_IMP_FULL;
- req->rq_phase = RQ_PHASE_RPC;
+ ptlrpc_rqphase_move(req, RQ_PHASE_RPC);
req->rq_import_generation = generation;
ptlrpc_set_add_req(set, req);
} else {
}
do_check_set:
- rc = ptlrpc_check_set(set);
+ rc = ptlrpc_check_set(NULL, set);
/* not finished, and we are not expired, simply return */
if (!rc && cfs_time_before(curtime, cfs_time_add(pd->pd_this_ping,
if (req->rq_phase == RQ_PHASE_COMPLETE)
continue;
- req->rq_phase = RQ_PHASE_COMPLETE;
- atomic_dec(&req->rq_import->imp_inflight);
- set->set_remaining--;
- /* If it was disconnected, don't sweat it. */
- if (list_empty(&req->rq_import->imp_pinger_chain)) {
- ptlrpc_unregister_reply(req);
- continue;
- }
+ CDEBUG(D_RPCTRACE, "Pinger initiate expire request(%p)\n",
+ req);
- CDEBUG(D_RPCTRACE, "pinger initiate expire_one_request\n");
- ptlrpc_expire_one_request(req);
+ /* This will also unregister reply. */
+ ptlrpc_expire_one_request(req, 0);
+
+ /* We're done with this req, let's finally move it to complete
+ * phase and take care of inflights. */
+ ptlrpc_rqphase_move(req, RQ_PHASE_COMPLETE);
+ imp = req->rq_import;
+ spin_lock(&imp->imp_lock);
+ if (!list_empty(&req->rq_list)) {
+ list_del_init(&req->rq_list);
+ atomic_dec(&imp->imp_inflight);
+ }
+ spin_unlock(&imp->imp_lock);
+ set->set_remaining--;
}
mutex_up(&pinger_sem);
#endif
}
+int ptlrpc_add_timeout_client(int time, enum timeout_event event,
+ timeout_cb_t cb, void *data,
+ struct list_head *obd_list)
+{
+ return 0;
+}
+
+int ptlrpc_del_timeout_client(struct list_head *obd_list,
+ enum timeout_event event)
+{
+ return 0;
+}
+
int ptlrpc_pinger_add_import(struct obd_import *imp)
{
ENTRY;
CDEBUG(D_RPCTRACE, "checking import %s->%s\n",
imp->imp_obd->obd_uuid.uuid, obd2cli_tgt(imp->imp_obd));
#ifdef ENABLE_LIBLUSTRE_RECOVERY
- if (imp->imp_state == LUSTRE_IMP_DISCON && !imp->imp_deactive)
+ if (imp->imp_state == LUSTRE_IMP_DISCON &&
+ !imp_is_deactive(imp))
#else
/*XXX only recover for the initial connection */
if (!lustre_handle_is_used(&imp->imp_remote_handle) &&
- imp->imp_state == LUSTRE_IMP_DISCON && !imp->imp_deactive)
+ imp->imp_state == LUSTRE_IMP_DISCON &&
+ !imp_is_deactive(imp))
#endif
ptlrpc_initiate_recovery(imp);
else if (imp->imp_state != LUSTRE_IMP_FULL)
"state %d, deactive %d\n",
imp->imp_obd->obd_uuid.uuid,
obd2cli_tgt(imp->imp_obd), imp->imp_state,
- imp->imp_deactive);
+ imp_is_deactive(imp));
}
EXIT;
#endif