X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fptlrpc%2Fpinger.c;h=af1924a09a5c7fcb36848afabc1662f1b7da1fa9;hp=77507ca0508b993e13f6bc7663c43ee811cba362;hb=49f3a017b8425a3e03d0f7dd5d8dd27fa69f7c9f;hpb=859678cc6b075f7c81903e44b99bdbd18c635cbb diff --git a/lustre/ptlrpc/pinger.c b/lustre/ptlrpc/pinger.c index 77507ca..af1924a 100644 --- a/lustre/ptlrpc/pinger.c +++ b/lustre/ptlrpc/pinger.c @@ -1,29 +1,41 @@ /* -*- 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. + * + * 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. * - * Copyright (c) 2003 Cluster File Systems, Inc. - * Authors: Phil Schwan - * Mike Shaver + * 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). * - * This file is part of the Lustre file system, http://www.lustre.org - * Lustre is a trademark of Cluster File Systems, Inc. + * 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 * - * 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. + * 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 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. * - * 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. + * lustre/ptlrpc/pinger.c * - * 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. + * Portal-RPC reconnection and replay operations, for use in recovery. */ #ifndef __KERNEL__ @@ -37,37 +49,74 @@ #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; - int rc = 0; ENTRY; - req = ptlrpc_prep_req(imp, LUSTRE_OBD_VERSION, OBD_PING, 1, NULL, NULL); - if (req) { - 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_req_set_repsize(req, 1, NULL); - ptlrpcd_add_req(req); - } else { + req = ptlrpc_prep_ping(imp); + if (req == NULL) { CERROR("OOM trying to ping %s->%s\n", imp->imp_obd->obd_uuid.uuid, obd2cli_tgt(imp->imp_obd)); - rc = -ENOMEM; + RETURN(-ENOMEM); } - RETURN(rc); + DEBUG_REQ(D_INFO, req, "pinging %s->%s", + imp->imp_obd->obd_uuid.uuid, obd2cli_tgt(imp->imp_obd)); + ptlrpcd_add_req(req, PSCOPE_OTHER); + + RETURN(0); } -static void ptlrpc_update_next_ping(struct obd_import *imp) +void ptlrpc_update_next_ping(struct obd_import *imp) { - imp->imp_next_ping = cfs_time_shift( - (imp->imp_state == LUSTRE_IMP_DISCON ? - RECONNECT_INTERVAL : PING_INTERVAL)); +#ifdef ENABLE_PINGER + 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 */ } void ptlrpc_ping_import_soon(struct obd_import *imp) @@ -75,7 +124,141 @@ void ptlrpc_ping_import_soon(struct obd_import *imp) 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) + return cfs_time_shift(obd_timeout / 2); + else + return cfs_time_shift(obd_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 + +cfs_time_t ptlrpc_suspend_wakeup_time(void) +{ + return suspend_wakeup_time; +} + +void ptlrpc_deactivate_timeouts(struct obd_import *imp) +{ + /*XXX: disabled for now, will be replaced by adaptive timeouts */ +#if 0 + if (imp->imp_no_timeout) + return; + imp->imp_no_timeout = 1; + atomic_inc(&suspend_timeouts); + CDEBUG(D_HA|D_WARNING, "deactivate timeouts %u\n", atomic_read(&suspend_timeouts)); +#endif +} + +void ptlrpc_activate_timeouts(struct obd_import *imp) +{ + /*XXX: disabled for now, will be replaced by adaptive timeouts */ +#if 0 + if (!imp->imp_no_timeout) + return; + imp->imp_no_timeout = 0; + LASSERT(atomic_read(&suspend_timeouts) > 0); + if (atomic_dec_and_test(&suspend_timeouts)) { + suspend_wakeup_time = cfs_time_current(); + wake_up(&suspend_timeouts_waitq); + } + CDEBUG(D_HA|D_WARNING, "activate timeouts %u\n", atomic_read(&suspend_timeouts)); +#endif +} + +int ptlrpc_check_suspend(void) +{ + if (atomic_read(&suspend_timeouts)) + return 1; + return 0; +} + +int ptlrpc_check_and_wait_suspend(struct ptlrpc_request *req) +{ + struct l_wait_info lwi; + + if (atomic_read(&suspend_timeouts)) { + DEBUG_REQ(D_NET, req, "-- suspend %d regular timeout", + atomic_read(&suspend_timeouts)); + lwi = LWI_INTR(NULL, NULL); + l_wait_event(suspend_timeouts_waitq, + atomic_read(&suspend_timeouts) == 0, &lwi); + DEBUG_REQ(D_NET, req, "-- recharge regular timeout"); + return 1; + } + return 0; +} + +#ifdef __KERNEL__ + +static void ptlrpc_pinger_process_import(struct obd_import *imp, + unsigned long this_ping) +{ + int force, level; + + spin_lock(&imp->imp_lock); + level = imp->imp_state; + force = imp->imp_force_verify; + if (force) + imp->imp_force_verify = 0; + spin_unlock(&imp->imp_lock); + + CDEBUG(level == LUSTRE_IMP_FULL ? D_INFO : D_HA, + "level %s/%u force %u deactive %u pingable %u\n", + ptlrpc_import_state_name(level), level, + force, imp->imp_deactive, imp->imp_pingable); + + if (cfs_time_aftereq(imp->imp_next_ping - 5 * CFS_TICK, + this_ping) && force == 0) + return; + + 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 || + imp->imp_obd->obd_no_recov || + imp_is_deactive(imp)) { + CDEBUG(D_HA, "not pinging %s (in recovery " + " or recovery disabled: %s)\n", + obd2cli_tgt(imp->imp_obd), + ptlrpc_import_state_name(level)); + } else if (imp->imp_pingable || force) { + ptlrpc_ping(imp); + } +} + static int ptlrpc_pinger_main(void *arg) { struct ptlrpc_svc_data *data = (struct ptlrpc_svc_data *)arg; @@ -92,86 +275,45 @@ static int ptlrpc_pinger_main(void *arg) 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, imp_pinger_chain); - int force, level; - - spin_lock(&imp->imp_lock); - level = imp->imp_state; - force = imp->imp_force_verify; - imp->imp_force_verify = 0; - spin_unlock(&imp->imp_lock); - - CDEBUG(level == LUSTRE_IMP_FULL ? D_INFO : D_HA, - "level %s/%u force %u deactive %u pingable %u\n", - ptlrpc_import_state_name(level), level, - force, imp->imp_deactive, imp->imp_pingable); - - if (force || - /* if the next ping is within, say, 5 jiffies from - now, go ahead and ping. See note below. */ - cfs_time_aftereq(this_ping, - imp->imp_next_ping - 5 * CFS_TICK)) { - if (level == LUSTRE_IMP_DISCON && - !imp->imp_deactive) { - /* wait at least a timeout before - trying recovery again. */ - imp->imp_next_ping = cfs_time_shift(obd_timeout); - ptlrpc_initiate_recovery(imp); - } else if (level != LUSTRE_IMP_FULL || - imp->imp_obd->obd_no_recov || - imp->imp_deactive) { - CDEBUG(D_HA, "not pinging %s " - "(in recovery: %s or recovery " - "disabled: %u/%u)\n", - obd2cli_tgt(imp->imp_obd), - ptlrpc_import_state_name(level), - imp->imp_deactive, - imp->imp_obd->obd_no_recov); - } else if (imp->imp_pingable || force) { - ptlrpc_ping(imp); - } - } else { - if (!imp->imp_pingable) - continue; - CDEBUG(D_INFO, - "don't need to ping %s ("CFS_TIME_T - " > "CFS_TIME_T")\n", - obd2cli_tgt(imp->imp_obd), - imp->imp_next_ping, this_ping); - } + ptlrpc_pinger_process_import(imp, this_ping); /* obd_timeout might have changed */ - if (cfs_time_after(imp->imp_next_ping, - cfs_time_add(this_ping, + if (imp->imp_pingable && imp->imp_next_ping && + cfs_time_after(imp->imp_next_ping, + cfs_time_add(this_ping, cfs_time_seconds(PING_INTERVAL)))) ptlrpc_update_next_ping(imp); } mutex_up(&pinger_sem); + /* update memory usage info */ + 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); @@ -208,10 +350,11 @@ int ptlrpc_start_pinger(void) if (pinger_thread != NULL) RETURN(-EALREADY); - OBD_ALLOC(pinger_thread, sizeof(*pinger_thread)); + OBD_ALLOC_PTR(pinger_thread); if (pinger_thread == NULL) RETURN(-ENOMEM); cfs_waitq_init(&pinger_thread->t_ctl_waitq); + cfs_waitq_init(&suspend_timeouts_waitq); d.name = "ll_ping"; d.thread = pinger_thread; @@ -231,6 +374,8 @@ int ptlrpc_start_pinger(void) RETURN(0); } +int ptlrpc_pinger_remove_timeouts(void); + int ptlrpc_stop_pinger(void) { struct l_wait_info lwi = { 0 }; @@ -242,6 +387,8 @@ int ptlrpc_stop_pinger(void) 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); @@ -250,7 +397,7 @@ int ptlrpc_stop_pinger(void) l_wait_event(pinger_thread->t_ctl_waitq, (pinger_thread->t_flags & SVC_STOPPED), &lwi); - OBD_FREE(pinger_thread, sizeof(*pinger_thread)); + OBD_FREE_PTR(pinger_thread); pinger_thread = NULL; RETURN(rc); } @@ -269,6 +416,8 @@ int ptlrpc_pinger_add_import(struct obd_import *imp) mutex_down(&pinger_sem); CDEBUG(D_HA, "adding pingable import %s->%s\n", imp->imp_obd->obd_uuid.uuid, obd2cli_tgt(imp->imp_obd)); + /* if we add to pinger we want recovery on this import */ + imp->imp_obd->obd_no_recov = 0; ptlrpc_update_next_ping(imp); /* XXX sort, blah blah */ list_add_tail(&imp->imp_pinger_chain, &pinger_imports); @@ -290,11 +439,128 @@ int ptlrpc_pinger_del_import(struct obd_import *imp) list_del_init(&imp->imp_pinger_chain); CDEBUG(D_HA, "removing pingable import %s->%s\n", imp->imp_obd->obd_uuid.uuid, obd2cli_tgt(imp->imp_obd)); + /* if we remove from pinger we don't want recovery on this import */ + imp->imp_obd->obd_no_recov = 1; class_import_put(imp); mutex_up(&pinger_sem); 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 @@ -316,7 +582,7 @@ static spinlock_t pet_lock = SPIN_LOCK_UNLOCKED; int ping_evictor_wake(struct obd_export *exp) { spin_lock(&pet_lock); - if (pet_exp) { + if (pet_exp || (pet_state != PET_READY)) { /* eventually the new obd will call here again. */ spin_unlock(&pet_lock); return 1; @@ -359,7 +625,7 @@ static int ping_evictor_main(void *arg) obd = pet_exp->exp_obd; spin_unlock(&pet_lock); - expire_time = CURRENT_SECONDS - (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); @@ -375,14 +641,19 @@ static int ping_evictor_main(void *arg) if (expire_time > exp->exp_last_request_time) { class_export_get(exp); spin_unlock(&obd->obd_dev_lock); - LCONSOLE_WARN("%s: haven't heard from client %s" + LCONSOLE_WARN("%s: haven't heard from client %s" " (at %s) in %ld seconds. I think" " it's dead, and I am evicting" - " it.\n", obd->obd_name, + " it. exp %p, cur %ld expire %ld" + " last %ld\n", + obd->obd_name, obd_uuid2str(&exp->exp_client_uuid), obd_export_nid2str(exp), - (long)(CURRENT_SECONDS - - exp->exp_last_request_time)); + (long)(cfs_time_current_sec() - + exp->exp_last_request_time), + exp, (long)cfs_time_current_sec(), + (long)expire_time, + (long)exp->exp_last_request_time); CDEBUG(D_HA, "Last request was at %ld\n", exp->exp_last_request_time); class_fail_export(exp); @@ -452,6 +723,7 @@ static int pinger_check_rpcs(void *arg) 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; @@ -489,7 +761,7 @@ static int pinger_check_rpcs(void *arg) 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); @@ -504,21 +776,25 @@ static int pinger_check_rpcs(void *arg) continue; } - req = ptlrpc_prep_req(imp, LUSTRE_OBD_VERSION, OBD_PING, - 1, NULL, NULL); - if (!req) { - CERROR("out of memory\n"); + req = ptlrpc_request_alloc_pack(imp, &RQF_OBD_PING, + LUSTRE_OBD_VERSION, + OBD_PING); + if (req == NULL) { + CERROR("OOM trying to ping %s->%s\n", + imp->imp_obd->obd_uuid.uuid, + obd2cli_tgt(imp->imp_obd)); break; } + req->rq_no_resend = 1; - ptlrpc_req_set_repsize(req, 1, NULL); + 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 { - CDEBUG(D_HA, "don't need to ping %s ("CFS_TIME_T" > " - CFS_TIME_T")\n", obd2cli_tgt(imp->imp_obd), + CDEBUG(D_INFO, "don't need to ping %s ("CFS_TIME_T + " > "CFS_TIME_T")\n", obd2cli_tgt(imp->imp_obd), imp->imp_next_ping, pd->pd_this_ping); } } @@ -527,26 +803,25 @@ static int pinger_check_rpcs(void *arg) /* Might be empty, that's OK. */ if (set->set_remaining == 0) - CDEBUG(D_HA, "nothing to ping\n"); + CDEBUG(D_RPCTRACE, "nothing to ping\n"); list_for_each(iter, &set->set_requests) { struct ptlrpc_request *req = list_entry(iter, struct ptlrpc_request, rq_set_chain); - DEBUG_REQ(D_HA, req, "pinging %s->%s", + DEBUG_REQ(D_RPCTRACE, req, "pinging %s->%s", req->rq_import->imp_obd->obd_uuid.uuid, obd2cli_tgt(req->rq_import->imp_obd)); (void)ptl_send_rpc(req, 0); } 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, - cfs_time_seconds(PING_INTERVAL)))) { - CDEBUG(D_HA, "not finished, but also not expired\n"); + if (!rc && cfs_time_before(curtime, cfs_time_add(pd->pd_this_ping, + cfs_time_seconds(PING_INTERVAL)))) { + CDEBUG(D_RPCTRACE, "not finished, but also not expired\n"); pd->pd_recursion--; return 0; } @@ -557,20 +832,26 @@ do_check_set: req = list_entry(iter, struct ptlrpc_request, rq_set_chain); - if (req->rq_replied) + 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); + + /* This will also unregister reply. */ + ptlrpc_expire_one_request(req, 0); - CDEBUG(D_HA, "pinger initiate expire_one_request\n"); - ptlrpc_expire_one_request(req); + /* 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); @@ -582,7 +863,7 @@ out: cfs_time_seconds(PING_INTERVAL)); pd->pd_this_ping = 0; /* XXX for debug */ - CDEBUG(D_HA, "finished a round ping\n"); + CDEBUG(D_INFO, "finished a round ping\n"); pd->pd_recursion--; return 0; } @@ -625,6 +906,19 @@ void ptlrpc_pinger_sending_on_import(struct obd_import *imp) #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; @@ -661,19 +955,20 @@ int ptlrpc_pinger_del_import(struct obd_import *imp) void ptlrpc_pinger_wake_up() { #ifdef ENABLE_PINGER - ENTRY; /* XXX force pinger to run, if needed */ struct obd_import *imp; + ENTRY; list_for_each_entry(imp, &pinger_imports, imp_pinger_chain) { - CDEBUG(D_HA, "Checking that we need to do anything about import" - " %s->%s\n", imp->imp_obd->obd_uuid.uuid, - obd2cli_tgt(imp->imp_obd)); + 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) @@ -681,9 +976,9 @@ void ptlrpc_pinger_wake_up() "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)); } -#endif EXIT; +#endif } #endif /* !__KERNEL__ */