/* -*- mode: c; c-basic-offset: 8; -*- * vim: noexpandtab sw=8 ts=8 sts=0: * * dlmrecovery.c * * recovery stuff * * Copyright (C) 2004 Oracle. All rights reserved. * * This program is free software; you can redistribute it and/or * modify it under the terms of the GNU General Public * License as published by the Free Software Foundation; either * version 2 of the License, or (at your option) any later version. * * 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 for more details. * * You should have received a copy of the GNU General Public * License along with this program; if not, write to the * Free Software Foundation, Inc., 59 Temple Place - Suite 330, * Boston, MA 021110-1307, USA. * */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "cluster/heartbeat.h" #include "cluster/nodemanager.h" #include "cluster/tcp.h" #include "dlmapi.h" #include "dlmcommon.h" #include "dlmdomain.h" #define MLOG_MASK_PREFIX (ML_DLM|ML_DLM_RECOVERY) #include "cluster/masklog.h" static void dlm_do_local_recovery_cleanup(struct dlm_ctxt *dlm, u8 dead_node); static int dlm_recovery_thread(void *data); void dlm_complete_recovery_thread(struct dlm_ctxt *dlm); int dlm_launch_recovery_thread(struct dlm_ctxt *dlm); void dlm_kick_recovery_thread(struct dlm_ctxt *dlm); static int dlm_do_recovery(struct dlm_ctxt *dlm); static int dlm_pick_recovery_master(struct dlm_ctxt *dlm); static int dlm_remaster_locks(struct dlm_ctxt *dlm, u8 dead_node); static int dlm_init_recovery_area(struct dlm_ctxt *dlm, u8 dead_node); static int dlm_request_all_locks(struct dlm_ctxt *dlm, u8 request_from, u8 dead_node); static void dlm_destroy_recovery_area(struct dlm_ctxt *dlm, u8 dead_node); static inline int dlm_num_locks_in_lockres(struct dlm_lock_resource *res); static void dlm_init_migratable_lockres(struct dlm_migratable_lockres *mres, const char *lockname, int namelen, int total_locks, u64 cookie, u8 flags, u8 master); static int dlm_send_mig_lockres_msg(struct dlm_ctxt *dlm, struct dlm_migratable_lockres *mres, u8 send_to, struct dlm_lock_resource *res, int total_locks); static int dlm_process_recovery_data(struct dlm_ctxt *dlm, struct dlm_lock_resource *res, struct dlm_migratable_lockres *mres); static int dlm_send_finalize_reco_message(struct dlm_ctxt *dlm); static int dlm_send_all_done_msg(struct dlm_ctxt *dlm, u8 dead_node, u8 send_to); static int dlm_send_begin_reco_message(struct dlm_ctxt *dlm, u8 dead_node); static void dlm_move_reco_locks_to_list(struct dlm_ctxt *dlm, struct list_head *list, u8 dead_node); static void dlm_finish_local_lockres_recovery(struct dlm_ctxt *dlm, u8 dead_node, u8 new_master); static void dlm_reco_ast(void *astdata); static void dlm_reco_bast(void *astdata, int blocked_type); static void dlm_reco_unlock_ast(void *astdata, enum dlm_status st); static void dlm_request_all_locks_worker(struct dlm_work_item *item, void *data); static void dlm_mig_lockres_worker(struct dlm_work_item *item, void *data); static u64 dlm_get_next_mig_cookie(void); static spinlock_t dlm_reco_state_lock = SPIN_LOCK_UNLOCKED; static spinlock_t dlm_mig_cookie_lock = SPIN_LOCK_UNLOCKED; static u64 dlm_mig_cookie = 1; static u64 dlm_get_next_mig_cookie(void) { u64 c; spin_lock(&dlm_mig_cookie_lock); c = dlm_mig_cookie; if (dlm_mig_cookie == (~0ULL)) dlm_mig_cookie = 1; else dlm_mig_cookie++; spin_unlock(&dlm_mig_cookie_lock); return c; } static inline void dlm_reset_recovery(struct dlm_ctxt *dlm) { spin_lock(&dlm->spinlock); clear_bit(dlm->reco.dead_node, dlm->recovery_map); dlm->reco.dead_node = O2NM_INVALID_NODE_NUM; dlm->reco.new_master = O2NM_INVALID_NODE_NUM; spin_unlock(&dlm->spinlock); } /* Worker function used during recovery. */ void dlm_dispatch_work(void *data) { struct dlm_ctxt *dlm = (struct dlm_ctxt *)data; LIST_HEAD(tmp_list); struct list_head *iter, *iter2; struct dlm_work_item *item; dlm_workfunc_t *workfunc; spin_lock(&dlm->work_lock); list_splice_init(&dlm->work_list, &tmp_list); spin_unlock(&dlm->work_lock); list_for_each_safe(iter, iter2, &tmp_list) { item = list_entry(iter, struct dlm_work_item, list); workfunc = item->func; list_del_init(&item->list); /* already have ref on dlm to avoid having * it disappear. just double-check. */ BUG_ON(item->dlm != dlm); /* this is allowed to sleep and * call network stuff */ workfunc(item, item->data); dlm_put(dlm); kfree(item); } } /* * RECOVERY THREAD */ void dlm_kick_recovery_thread(struct dlm_ctxt *dlm) { /* wake the recovery thread * this will wake the reco thread in one of three places * 1) sleeping with no recovery happening * 2) sleeping with recovery mastered elsewhere * 3) recovery mastered here, waiting on reco data */ wake_up(&dlm->dlm_reco_thread_wq); } /* Launch the recovery thread */ int dlm_launch_recovery_thread(struct dlm_ctxt *dlm) { mlog(0, "starting dlm recovery thread...\n"); dlm->dlm_reco_thread_task = kthread_run(dlm_recovery_thread, dlm, "dlm_reco_thread"); if (IS_ERR(dlm->dlm_reco_thread_task)) { mlog_errno(PTR_ERR(dlm->dlm_reco_thread_task)); dlm->dlm_reco_thread_task = NULL; return -EINVAL; } return 0; } void dlm_complete_recovery_thread(struct dlm_ctxt *dlm) { if (dlm->dlm_reco_thread_task) { mlog(0, "waiting for dlm recovery thread to exit\n"); kthread_stop(dlm->dlm_reco_thread_task); dlm->dlm_reco_thread_task = NULL; } } /* * this is lame, but here's how recovery works... * 1) all recovery threads cluster wide will work on recovering * ONE node at a time * 2) negotiate who will take over all the locks for the dead node. * thats right... ALL the locks. * 3) once a new master is chosen, everyone scans all locks * and moves aside those mastered by the dead guy * 4) each of these locks should be locked until recovery is done * 5) the new master collects up all of secondary lock queue info * one lock at a time, forcing each node to communicate back * before continuing * 6) each secondary lock queue responds with the full known lock info * 7) once the new master has run all its locks, it sends a ALLDONE! * message to everyone * 8) upon receiving this message, the secondary queue node unlocks * and responds to the ALLDONE * 9) once the new master gets responses from everyone, he unlocks * everything and recovery for this dead node is done *10) go back to 2) while there are still dead nodes * */ #define DLM_RECO_THREAD_TIMEOUT_MS (5 * 1000) static int dlm_recovery_thread(void *data) { int status; struct dlm_ctxt *dlm = data; unsigned long timeout = msecs_to_jiffies(DLM_RECO_THREAD_TIMEOUT_MS); mlog(0, "dlm thread running for %s...\n", dlm->name); while (!kthread_should_stop()) { if (dlm_joined(dlm)) { status = dlm_do_recovery(dlm); if (status == -EAGAIN) { /* do not sleep, recheck immediately. */ continue; } if (status < 0) mlog_errno(status); } wait_event_interruptible_timeout(dlm->dlm_reco_thread_wq, kthread_should_stop(), timeout); } mlog(0, "quitting DLM recovery thread\n"); return 0; } /* returns true when the recovery master has contacted us */ static int dlm_reco_master_ready(struct dlm_ctxt *dlm) { int ready; spin_lock(&dlm->spinlock); ready = (dlm->reco.new_master != O2NM_INVALID_NODE_NUM); spin_unlock(&dlm->spinlock); return ready; } /* returns true if node is no longer in the domain * could be dead or just not joined */ int dlm_is_node_dead(struct dlm_ctxt *dlm, u8 node) { int dead; spin_lock(&dlm->spinlock); dead = test_bit(node, dlm->domain_map); spin_unlock(&dlm->spinlock); return dead; } int dlm_wait_for_node_death(struct dlm_ctxt *dlm, u8 node, int timeout) { if (timeout) { mlog(ML_NOTICE, "%s: waiting %dms for notification of " "death of node %u\n", dlm->name, timeout, node); wait_event_timeout(dlm->dlm_reco_thread_wq, dlm_is_node_dead(dlm, node), msecs_to_jiffies(timeout)); } else { mlog(ML_NOTICE, "%s: waiting indefinitely for notification " "of death of node %u\n", dlm->name, node); wait_event(dlm->dlm_reco_thread_wq, dlm_is_node_dead(dlm, node)); } /* for now, return 0 */ return 0; } /* callers of the top-level api calls (dlmlock/dlmunlock) should * block on the dlm->reco.event when recovery is in progress. * the dlm recovery thread will set this state when it begins * recovering a dead node (as the new master or not) and clear * the state and wake as soon as all affected lock resources have * been marked with the RECOVERY flag */ static int dlm_in_recovery(struct dlm_ctxt *dlm) { int in_recovery; spin_lock(&dlm->spinlock); in_recovery = !!(dlm->reco.state & DLM_RECO_STATE_ACTIVE); spin_unlock(&dlm->spinlock); return in_recovery; } void dlm_wait_for_recovery(struct dlm_ctxt *dlm) { wait_event(dlm->reco.event, !dlm_in_recovery(dlm)); } static void dlm_begin_recovery(struct dlm_ctxt *dlm) { spin_lock(&dlm->spinlock); BUG_ON(dlm->reco.state & DLM_RECO_STATE_ACTIVE); dlm->reco.state |= DLM_RECO_STATE_ACTIVE; spin_unlock(&dlm->spinlock); } static void dlm_end_recovery(struct dlm_ctxt *dlm) { spin_lock(&dlm->spinlock); BUG_ON(!(dlm->reco.state & DLM_RECO_STATE_ACTIVE)); dlm->reco.state &= ~DLM_RECO_STATE_ACTIVE; spin_unlock(&dlm->spinlock); wake_up(&dlm->reco.event); } static int dlm_do_recovery(struct dlm_ctxt *dlm) { int status = 0; int ret; spin_lock(&dlm->spinlock); /* check to see if the new master has died */ if (dlm->reco.new_master != O2NM_INVALID_NODE_NUM && test_bit(dlm->reco.new_master, dlm->recovery_map)) { mlog(0, "new master %u died while recovering %u!\n", dlm->reco.new_master, dlm->reco.dead_node); /* unset the new_master, leave dead_node */ dlm->reco.new_master = O2NM_INVALID_NODE_NUM; } /* select a target to recover */ if (dlm->reco.dead_node == O2NM_INVALID_NODE_NUM) { int bit; bit = find_next_bit (dlm->recovery_map, O2NM_MAX_NODES+1, 0); if (bit >= O2NM_MAX_NODES || bit < 0) dlm->reco.dead_node = O2NM_INVALID_NODE_NUM; else dlm->reco.dead_node = bit; } else if (!test_bit(dlm->reco.dead_node, dlm->recovery_map)) { /* BUG? */ mlog(ML_ERROR, "dead_node %u no longer in recovery map!\n", dlm->reco.dead_node); dlm->reco.dead_node = O2NM_INVALID_NODE_NUM; } if (dlm->reco.dead_node == O2NM_INVALID_NODE_NUM) { // mlog(0, "nothing to recover! sleeping now!\n"); spin_unlock(&dlm->spinlock); /* return to main thread loop and sleep. */ return 0; } mlog(0, "recovery thread found node %u in the recovery map!\n", dlm->reco.dead_node); spin_unlock(&dlm->spinlock); /* take write barrier */ /* (stops the list reshuffling thread, proxy ast handling) */ dlm_begin_recovery(dlm); if (dlm->reco.new_master == dlm->node_num) goto master_here; if (dlm->reco.new_master == O2NM_INVALID_NODE_NUM) { /* choose a new master, returns 0 if this node * is the master, -EEXIST if it's another node. * this does not return until a new master is chosen * or recovery completes entirely. */ ret = dlm_pick_recovery_master(dlm); if (!ret) { /* already notified everyone. go. */ goto master_here; } mlog(0, "another node will master this recovery session.\n"); } mlog(0, "dlm=%s, new_master=%u, this node=%u, dead_node=%u\n", dlm->name, dlm->reco.new_master, dlm->node_num, dlm->reco.dead_node); /* it is safe to start everything back up here * because all of the dead node's lock resources * have been marked as in-recovery */ dlm_end_recovery(dlm); /* sleep out in main dlm_recovery_thread loop. */ return 0; master_here: mlog(0, "mastering recovery of %s:%u here(this=%u)!\n", dlm->name, dlm->reco.dead_node, dlm->node_num); status = dlm_remaster_locks(dlm, dlm->reco.dead_node); if (status < 0) { mlog(ML_ERROR, "error %d remastering locks for node %u, " "retrying.\n", status, dlm->reco.dead_node); /* yield a bit to allow any final network messages * to get handled on remaining nodes */ msleep(100); } else { /* success! see if any other nodes need recovery */ mlog(0, "DONE mastering recovery of %s:%u here(this=%u)!\n", dlm->name, dlm->reco.dead_node, dlm->node_num); dlm_reset_recovery(dlm); } dlm_end_recovery(dlm); /* continue and look for another dead node */ return -EAGAIN; } static int dlm_remaster_locks(struct dlm_ctxt *dlm, u8 dead_node) { int status = 0; struct dlm_reco_node_data *ndata; struct list_head *iter; int all_nodes_done; int destroy = 0; int pass = 0; status = dlm_init_recovery_area(dlm, dead_node); if (status < 0) goto leave; /* safe to access the node data list without a lock, since this * process is the only one to change the list */ list_for_each(iter, &dlm->reco.node_data) { ndata = list_entry (iter, struct dlm_reco_node_data, list); BUG_ON(ndata->state != DLM_RECO_NODE_DATA_INIT); ndata->state = DLM_RECO_NODE_DATA_REQUESTING; mlog(0, "requesting lock info from node %u\n", ndata->node_num); if (ndata->node_num == dlm->node_num) { ndata->state = DLM_RECO_NODE_DATA_DONE; continue; } status = dlm_request_all_locks(dlm, ndata->node_num, dead_node); if (status < 0) { mlog_errno(status); if (dlm_is_host_down(status)) ndata->state = DLM_RECO_NODE_DATA_DEAD; else { destroy = 1; goto leave; } } switch (ndata->state) { case DLM_RECO_NODE_DATA_INIT: case DLM_RECO_NODE_DATA_FINALIZE_SENT: case DLM_RECO_NODE_DATA_REQUESTED: BUG(); break; case DLM_RECO_NODE_DATA_DEAD: mlog(0, "node %u died after requesting " "recovery info for node %u\n", ndata->node_num, dead_node); // start all over destroy = 1; status = -EAGAIN; goto leave; case DLM_RECO_NODE_DATA_REQUESTING: ndata->state = DLM_RECO_NODE_DATA_REQUESTED; mlog(0, "now receiving recovery data from " "node %u for dead node %u\n", ndata->node_num, dead_node); break; case DLM_RECO_NODE_DATA_RECEIVING: mlog(0, "already receiving recovery data from " "node %u for dead node %u\n", ndata->node_num, dead_node); break; case DLM_RECO_NODE_DATA_DONE: mlog(0, "already DONE receiving recovery data " "from node %u for dead node %u\n", ndata->node_num, dead_node); break; } } mlog(0, "done requesting all lock info\n"); /* nodes should be sending reco data now * just need to wait */ while (1) { /* check all the nodes now to see if we are * done, or if anyone died */ all_nodes_done = 1; spin_lock(&dlm_reco_state_lock); list_for_each(iter, &dlm->reco.node_data) { ndata = list_entry (iter, struct dlm_reco_node_data, list); mlog(0, "checking recovery state of node %u\n", ndata->node_num); switch (ndata->state) { case DLM_RECO_NODE_DATA_INIT: case DLM_RECO_NODE_DATA_REQUESTING: mlog(ML_ERROR, "bad ndata state for " "node %u: state=%d\n", ndata->node_num, ndata->state); BUG(); break; case DLM_RECO_NODE_DATA_DEAD: mlog(ML_NOTICE, "node %u died after " "requesting recovery info for " "node %u\n", ndata->node_num, dead_node); spin_unlock(&dlm_reco_state_lock); // start all over destroy = 1; status = -EAGAIN; /* instead of spinning like crazy here, * wait for the domain map to catch up * with the network state. otherwise this * can be hit hundreds of times before * the node is really seen as dead. */ wait_event_timeout(dlm->dlm_reco_thread_wq, dlm_is_node_dead(dlm, ndata->node_num), msecs_to_jiffies(1000)); mlog(0, "waited 1 sec for %u, " "dead? %s\n", ndata->node_num, dlm_is_node_dead(dlm, ndata->node_num) ? "yes" : "no"); goto leave; case DLM_RECO_NODE_DATA_RECEIVING: case DLM_RECO_NODE_DATA_REQUESTED: all_nodes_done = 0; break; case DLM_RECO_NODE_DATA_DONE: break; case DLM_RECO_NODE_DATA_FINALIZE_SENT: break; } } spin_unlock(&dlm_reco_state_lock); mlog(0, "pass #%d, all_nodes_done?: %s\n", ++pass, all_nodes_done?"yes":"no"); if (all_nodes_done) { int ret; /* all nodes are now in DLM_RECO_NODE_DATA_DONE state * just send a finalize message to everyone and * clean up */ mlog(0, "all nodes are done! send finalize\n"); ret = dlm_send_finalize_reco_message(dlm); if (ret < 0) mlog_errno(ret); spin_lock(&dlm->spinlock); dlm_finish_local_lockres_recovery(dlm, dead_node, dlm->node_num); spin_unlock(&dlm->spinlock); mlog(0, "should be done with recovery!\n"); mlog(0, "finishing recovery of %s at %lu, " "dead=%u, this=%u, new=%u\n", dlm->name, jiffies, dlm->reco.dead_node, dlm->node_num, dlm->reco.new_master); destroy = 1; status = ret; /* rescan everything marked dirty along the way */ dlm_kick_thread(dlm, NULL); break; } /* wait to be signalled, with periodic timeout * to check for node death */ wait_event_interruptible_timeout(dlm->dlm_reco_thread_wq, kthread_should_stop(), msecs_to_jiffies(DLM_RECO_THREAD_TIMEOUT_MS)); } leave: if (destroy) dlm_destroy_recovery_area(dlm, dead_node); mlog_exit(status); return status; } static int dlm_init_recovery_area(struct dlm_ctxt *dlm, u8 dead_node) { int num=0; struct dlm_reco_node_data *ndata; spin_lock(&dlm->spinlock); memcpy(dlm->reco.node_map, dlm->domain_map, sizeof(dlm->domain_map)); /* nodes can only be removed (by dying) after dropping * this lock, and death will be trapped later, so this should do */ spin_unlock(&dlm->spinlock); while (1) { num = find_next_bit (dlm->reco.node_map, O2NM_MAX_NODES, num); if (num >= O2NM_MAX_NODES) { break; } BUG_ON(num == dead_node); ndata = kcalloc(1, sizeof(*ndata), GFP_KERNEL); if (!ndata) { dlm_destroy_recovery_area(dlm, dead_node); return -ENOMEM; } ndata->node_num = num; ndata->state = DLM_RECO_NODE_DATA_INIT; spin_lock(&dlm_reco_state_lock); list_add_tail(&ndata->list, &dlm->reco.node_data); spin_unlock(&dlm_reco_state_lock); num++; } return 0; } static void dlm_destroy_recovery_area(struct dlm_ctxt *dlm, u8 dead_node) { struct list_head *iter, *iter2; struct dlm_reco_node_data *ndata; LIST_HEAD(tmplist); spin_lock(&dlm_reco_state_lock); list_splice_init(&dlm->reco.node_data, &tmplist); spin_unlock(&dlm_reco_state_lock); list_for_each_safe(iter, iter2, &tmplist) { ndata = list_entry (iter, struct dlm_reco_node_data, list); list_del_init(&ndata->list); kfree(ndata); } } static int dlm_request_all_locks(struct dlm_ctxt *dlm, u8 request_from, u8 dead_node) { struct dlm_lock_request lr; enum dlm_status ret; mlog(0, "\n"); mlog(0, "dlm_request_all_locks: dead node is %u, sending request " "to %u\n", dead_node, request_from); memset(&lr, 0, sizeof(lr)); lr.node_idx = dlm->node_num; lr.dead_node = dead_node; // send message ret = DLM_NOLOCKMGR; ret = o2net_send_message(DLM_LOCK_REQUEST_MSG, dlm->key, &lr, sizeof(lr), request_from, NULL); /* negative status is handled by caller */ if (ret < 0) mlog_errno(ret); // return from here, then // sleep until all received or error return ret; } int dlm_request_all_locks_handler(struct o2net_msg *msg, u32 len, void *data) { struct dlm_ctxt *dlm = data; struct dlm_lock_request *lr = (struct dlm_lock_request *)msg->buf; char *buf = NULL; struct dlm_work_item *item = NULL; if (!dlm_grab(dlm)) return -EINVAL; BUG_ON(lr->dead_node != dlm->reco.dead_node); item = kcalloc(1, sizeof(*item), GFP_KERNEL); if (!item) { dlm_put(dlm); return -ENOMEM; } /* this will get freed by dlm_request_all_locks_worker */ buf = (char *) __get_free_page(GFP_KERNEL); if (!buf) { kfree(item); dlm_put(dlm); return -ENOMEM; } /* queue up work for dlm_request_all_locks_worker */ dlm_grab(dlm); /* get an extra ref for the work item */ dlm_init_work_item(dlm, item, dlm_request_all_locks_worker, buf); item->u.ral.reco_master = lr->node_idx; item->u.ral.dead_node = lr->dead_node; spin_lock(&dlm->work_lock); list_add_tail(&item->list, &dlm->work_list); spin_unlock(&dlm->work_lock); schedule_work(&dlm->dispatched_work); dlm_put(dlm); return 0; } static void dlm_request_all_locks_worker(struct dlm_work_item *item, void *data) { struct dlm_migratable_lockres *mres; struct dlm_lock_resource *res; struct dlm_ctxt *dlm; LIST_HEAD(resources); struct list_head *iter; int ret; u8 dead_node, reco_master; dlm = item->dlm; dead_node = item->u.ral.dead_node; reco_master = item->u.ral.reco_master; mres = (struct dlm_migratable_lockres *)data; if (dead_node != dlm->reco.dead_node || reco_master != dlm->reco.new_master) { /* show extra debug info if the recovery state is messed */ mlog(ML_ERROR, "%s: bad reco state: reco(dead=%u, master=%u), " "request(dead=%u, master=%u)\n", dlm->name, dlm->reco.dead_node, dlm->reco.new_master, dead_node, reco_master); mlog(ML_ERROR, "%s: name=%.*s master=%u locks=%u/%u flags=%u " "entry[0]={c=%u:%llu,l=%u,f=%u,t=%d,ct=%d,hb=%d,n=%u}\n", dlm->name, mres->lockname_len, mres->lockname, mres->master, mres->num_locks, mres->total_locks, mres->flags, dlm_get_lock_cookie_node(mres->ml[0].cookie), dlm_get_lock_cookie_seq(mres->ml[0].cookie), mres->ml[0].list, mres->ml[0].flags, mres->ml[0].type, mres->ml[0].convert_type, mres->ml[0].highest_blocked, mres->ml[0].node); BUG(); } BUG_ON(dead_node != dlm->reco.dead_node); BUG_ON(reco_master != dlm->reco.new_master); /* lock resources should have already been moved to the * dlm->reco.resources list. now move items from that list * to a temp list if the dead owner matches. note that the * whole cluster recovers only one node at a time, so we * can safely move UNKNOWN lock resources for each recovery * session. */ dlm_move_reco_locks_to_list(dlm, &resources, dead_node); /* now we can begin blasting lockreses without the dlm lock */ list_for_each(iter, &resources) { res = list_entry (iter, struct dlm_lock_resource, recovering); ret = dlm_send_one_lockres(dlm, res, mres, reco_master, DLM_MRES_RECOVERY); if (ret < 0) mlog_errno(ret); } /* move the resources back to the list */ spin_lock(&dlm->spinlock); list_splice_init(&resources, &dlm->reco.resources); spin_unlock(&dlm->spinlock); ret = dlm_send_all_done_msg(dlm, dead_node, reco_master); if (ret < 0) mlog_errno(ret); free_page((unsigned long)data); } static int dlm_send_all_done_msg(struct dlm_ctxt *dlm, u8 dead_node, u8 send_to) { int ret, tmpret; struct dlm_reco_data_done done_msg; memset(&done_msg, 0, sizeof(done_msg)); done_msg.node_idx = dlm->node_num; done_msg.dead_node = dead_node; mlog(0, "sending DATA DONE message to %u, " "my node=%u, dead node=%u\n", send_to, done_msg.node_idx, done_msg.dead_node); ret = o2net_send_message(DLM_RECO_DATA_DONE_MSG, dlm->key, &done_msg, sizeof(done_msg), send_to, &tmpret); /* negative status is ignored by the caller */ if (ret >= 0) ret = tmpret; return ret; } int dlm_reco_data_done_handler(struct o2net_msg *msg, u32 len, void *data) { struct dlm_ctxt *dlm = data; struct dlm_reco_data_done *done = (struct dlm_reco_data_done *)msg->buf; struct list_head *iter; struct dlm_reco_node_data *ndata = NULL; int ret = -EINVAL; if (!dlm_grab(dlm)) return -EINVAL; mlog(0, "got DATA DONE: dead_node=%u, reco.dead_node=%u, " "node_idx=%u, this node=%u\n", done->dead_node, dlm->reco.dead_node, done->node_idx, dlm->node_num); BUG_ON(done->dead_node != dlm->reco.dead_node); spin_lock(&dlm_reco_state_lock); list_for_each(iter, &dlm->reco.node_data) { ndata = list_entry (iter, struct dlm_reco_node_data, list); if (ndata->node_num != done->node_idx) continue; switch (ndata->state) { /* should have moved beyond INIT but not to FINALIZE yet */ case DLM_RECO_NODE_DATA_INIT: case DLM_RECO_NODE_DATA_DEAD: case DLM_RECO_NODE_DATA_FINALIZE_SENT: mlog(ML_ERROR, "bad ndata state for node %u:" " state=%d\n", ndata->node_num, ndata->state); BUG(); break; /* these states are possible at this point, anywhere along * the line of recovery */ case DLM_RECO_NODE_DATA_DONE: case DLM_RECO_NODE_DATA_RECEIVING: case DLM_RECO_NODE_DATA_REQUESTED: case DLM_RECO_NODE_DATA_REQUESTING: mlog(0, "node %u is DONE sending " "recovery data!\n", ndata->node_num); ndata->state = DLM_RECO_NODE_DATA_DONE; ret = 0; break; } } spin_unlock(&dlm_reco_state_lock); /* wake the recovery thread, some node is done */ if (!ret) dlm_kick_recovery_thread(dlm); if (ret < 0) mlog(ML_ERROR, "failed to find recovery node data for node " "%u\n", done->node_idx); dlm_put(dlm); mlog(0, "leaving reco data done handler, ret=%d\n", ret); return ret; } static void dlm_move_reco_locks_to_list(struct dlm_ctxt *dlm, struct list_head *list, u8 dead_node) { struct dlm_lock_resource *res; struct list_head *iter, *iter2; struct dlm_lock *lock; spin_lock(&dlm->spinlock); list_for_each_safe(iter, iter2, &dlm->reco.resources) { res = list_entry (iter, struct dlm_lock_resource, recovering); /* always prune any $RECOVERY entries for dead nodes, * otherwise hangs can occur during later recovery */ if (dlm_is_recovery_lock(res->lockname.name, res->lockname.len)) { spin_lock(&res->spinlock); list_for_each_entry(lock, &res->granted, list) { if (lock->ml.node == dead_node) { mlog(0, "AHA! there was " "a $RECOVERY lock for dead " "node %u (%s)!\n", dead_node, dlm->name); list_del_init(&lock->list); dlm_lock_put(lock); break; } } spin_unlock(&res->spinlock); continue; } if (res->owner == dead_node) { mlog(0, "found lockres owned by dead node while " "doing recovery for node %u. sending it.\n", dead_node); list_move_tail(&res->recovering, list); } else if (res->owner == DLM_LOCK_RES_OWNER_UNKNOWN) { mlog(0, "found UNKNOWN owner while doing recovery " "for node %u. sending it.\n", dead_node); list_move_tail(&res->recovering, list); } } spin_unlock(&dlm->spinlock); } static inline int dlm_num_locks_in_lockres(struct dlm_lock_resource *res) { int total_locks = 0; struct list_head *iter, *queue = &res->granted; int i; for (i=0; i<3; i++) { list_for_each(iter, queue) total_locks++; queue++; } return total_locks; } static int dlm_send_mig_lockres_msg(struct dlm_ctxt *dlm, struct dlm_migratable_lockres *mres, u8 send_to, struct dlm_lock_resource *res, int total_locks) { u64 mig_cookie = be64_to_cpu(mres->mig_cookie); int mres_total_locks = be32_to_cpu(mres->total_locks); int sz, ret = 0, status = 0; u8 orig_flags = mres->flags, orig_master = mres->master; BUG_ON(mres->num_locks > DLM_MAX_MIGRATABLE_LOCKS); if (!mres->num_locks) return 0; sz = sizeof(struct dlm_migratable_lockres) + (mres->num_locks * sizeof(struct dlm_migratable_lock)); /* add an all-done flag if we reached the last lock */ orig_flags = mres->flags; BUG_ON(total_locks > mres_total_locks); if (total_locks == mres_total_locks) mres->flags |= DLM_MRES_ALL_DONE; /* send it */ ret = o2net_send_message(DLM_MIG_LOCKRES_MSG, dlm->key, mres, sz, send_to, &status); if (ret < 0) { /* XXX: negative status is not handled. * this will end up killing this node. */ mlog_errno(ret); } else { /* might get an -ENOMEM back here */ ret = status; if (ret < 0) { mlog_errno(ret); if (ret == -EFAULT) { mlog(ML_ERROR, "node %u told me to kill " "myself!\n", send_to); BUG(); } } } /* zero and reinit the message buffer */ dlm_init_migratable_lockres(mres, res->lockname.name, res->lockname.len, mres_total_locks, mig_cookie, orig_flags, orig_master); return ret; } static void dlm_init_migratable_lockres(struct dlm_migratable_lockres *mres, const char *lockname, int namelen, int total_locks, u64 cookie, u8 flags, u8 master) { /* mres here is one full page */ memset(mres, 0, PAGE_SIZE); mres->lockname_len = namelen; memcpy(mres->lockname, lockname, namelen); mres->num_locks = 0; mres->total_locks = cpu_to_be32(total_locks); mres->mig_cookie = cpu_to_be64(cookie); mres->flags = flags; mres->master = master; } /* returns 1 if this lock fills the network structure, * 0 otherwise */ static int dlm_add_lock_to_array(struct dlm_lock *lock, struct dlm_migratable_lockres *mres, int queue) { struct dlm_migratable_lock *ml; int lock_num = mres->num_locks; ml = &(mres->ml[lock_num]); ml->cookie = lock->ml.cookie; ml->type = lock->ml.type; ml->convert_type = lock->ml.convert_type; ml->highest_blocked = lock->ml.highest_blocked; ml->list = queue; if (lock->lksb) { ml->flags = lock->lksb->flags; /* send our current lvb */ if (ml->type == LKM_EXMODE || ml->type == LKM_PRMODE) { /* if it is already set, this had better be a PR * and it has to match */ if (mres->lvb[0] && (ml->type == LKM_EXMODE || memcmp(mres->lvb, lock->lksb->lvb, DLM_LVB_LEN))) { mlog(ML_ERROR, "mismatched lvbs!\n"); __dlm_print_one_lock_resource(lock->lockres); BUG(); } memcpy(mres->lvb, lock->lksb->lvb, DLM_LVB_LEN); } } ml->node = lock->ml.node; mres->num_locks++; /* we reached the max, send this network message */ if (mres->num_locks == DLM_MAX_MIGRATABLE_LOCKS) return 1; return 0; } int dlm_send_one_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res, struct dlm_migratable_lockres *mres, u8 send_to, u8 flags) { struct list_head *queue, *iter; int total_locks, i; u64 mig_cookie = 0; struct dlm_lock *lock; int ret = 0; BUG_ON(!(flags & (DLM_MRES_RECOVERY|DLM_MRES_MIGRATION))); mlog(0, "sending to %u\n", send_to); total_locks = dlm_num_locks_in_lockres(res); if (total_locks > DLM_MAX_MIGRATABLE_LOCKS) { /* rare, but possible */ mlog(0, "argh. lockres has %d locks. this will " "require more than one network packet to " "migrate\n", total_locks); mig_cookie = dlm_get_next_mig_cookie(); } dlm_init_migratable_lockres(mres, res->lockname.name, res->lockname.len, total_locks, mig_cookie, flags, res->owner); total_locks = 0; for (i=DLM_GRANTED_LIST; i<=DLM_BLOCKED_LIST; i++) { queue = dlm_list_idx_to_ptr(res, i); list_for_each(iter, queue) { lock = list_entry (iter, struct dlm_lock, list); /* add another lock. */ total_locks++; if (!dlm_add_lock_to_array(lock, mres, i)) continue; /* this filled the lock message, * we must send it immediately. */ ret = dlm_send_mig_lockres_msg(dlm, mres, send_to, res, total_locks); if (ret < 0) { // TODO mlog(ML_ERROR, "dlm_send_mig_lockres_msg " "returned %d, TODO\n", ret); BUG(); } } } /* flush any remaining locks */ ret = dlm_send_mig_lockres_msg(dlm, mres, send_to, res, total_locks); if (ret < 0) { // TODO mlog(ML_ERROR, "dlm_send_mig_lockres_msg returned %d, " "TODO\n", ret); BUG(); } return ret; } /* * this message will contain no more than one page worth of * recovery data, and it will work on only one lockres. * there may be many locks in this page, and we may need to wait * for additional packets to complete all the locks (rare, but * possible). */ /* * NOTE: the allocation error cases here are scary * we really cannot afford to fail an alloc in recovery * do we spin? returning an error only delays the problem really */ int dlm_mig_lockres_handler(struct o2net_msg *msg, u32 len, void *data) { struct dlm_ctxt *dlm = data; struct dlm_migratable_lockres *mres = (struct dlm_migratable_lockres *)msg->buf; int ret = 0; u8 real_master; char *buf = NULL; struct dlm_work_item *item = NULL; struct dlm_lock_resource *res = NULL; if (!dlm_grab(dlm)) return -EINVAL; BUG_ON(!(mres->flags & (DLM_MRES_RECOVERY|DLM_MRES_MIGRATION))); real_master = mres->master; if (real_master == DLM_LOCK_RES_OWNER_UNKNOWN) { /* cannot migrate a lockres with no master */ BUG_ON(!(mres->flags & DLM_MRES_RECOVERY)); } mlog(0, "%s message received from node %u\n", (mres->flags & DLM_MRES_RECOVERY) ? "recovery" : "migration", mres->master); if (mres->flags & DLM_MRES_ALL_DONE) mlog(0, "all done flag. all lockres data received!\n"); ret = -ENOMEM; buf = kmalloc(be16_to_cpu(msg->data_len), GFP_KERNEL); item = kcalloc(1, sizeof(*item), GFP_KERNEL); if (!buf || !item) goto leave; /* lookup the lock to see if we have a secondary queue for this * already... just add the locks in and this will have its owner * and RECOVERY flag changed when it completes. */ res = dlm_lookup_lockres(dlm, mres->lockname, mres->lockname_len); if (res) { /* this will get a ref on res */ /* mark it as recovering/migrating and hash it */ spin_lock(&res->spinlock); if (mres->flags & DLM_MRES_RECOVERY) { res->state |= DLM_LOCK_RES_RECOVERING; } else { if (res->state & DLM_LOCK_RES_MIGRATING) { /* this is at least the second * lockres message */ mlog(0, "lock %.*s is already migrating\n", mres->lockname_len, mres->lockname); } else if (res->state & DLM_LOCK_RES_RECOVERING) { /* caller should BUG */ mlog(ML_ERROR, "node is attempting to migrate " "lock %.*s, but marked as recovering!\n", mres->lockname_len, mres->lockname); ret = -EFAULT; spin_unlock(&res->spinlock); goto leave; } res->state |= DLM_LOCK_RES_MIGRATING; } spin_unlock(&res->spinlock); } else { /* need to allocate, just like if it was * mastered here normally */ res = dlm_new_lockres(dlm, mres->lockname, mres->lockname_len); if (!res) goto leave; /* to match the ref that we would have gotten if * dlm_lookup_lockres had succeeded */ dlm_lockres_get(res); /* mark it as recovering/migrating and hash it */ if (mres->flags & DLM_MRES_RECOVERY) res->state |= DLM_LOCK_RES_RECOVERING; else res->state |= DLM_LOCK_RES_MIGRATING; spin_lock(&dlm->spinlock); __dlm_insert_lockres(dlm, res); spin_unlock(&dlm->spinlock); /* now that the new lockres is inserted, * make it usable by other processes */ spin_lock(&res->spinlock); res->state &= ~DLM_LOCK_RES_IN_PROGRESS; spin_unlock(&res->spinlock); /* add an extra ref for just-allocated lockres * otherwise the lockres will be purged immediately */ dlm_lockres_get(res); } /* at this point we have allocated everything we need, * and we have a hashed lockres with an extra ref and * the proper res->state flags. */ ret = 0; if (mres->master == DLM_LOCK_RES_OWNER_UNKNOWN) { /* migration cannot have an unknown master */ BUG_ON(!(mres->flags & DLM_MRES_RECOVERY)); mlog(0, "recovery has passed me a lockres with an " "unknown owner.. will need to requery: " "%.*s\n", mres->lockname_len, mres->lockname); } else { spin_lock(&res->spinlock); dlm_change_lockres_owner(dlm, res, dlm->node_num); spin_unlock(&res->spinlock); } /* queue up work for dlm_mig_lockres_worker */ dlm_grab(dlm); /* get an extra ref for the work item */ memcpy(buf, msg->buf, be16_to_cpu(msg->data_len)); /* copy the whole message */ dlm_init_work_item(dlm, item, dlm_mig_lockres_worker, buf); item->u.ml.lockres = res; /* already have a ref */ item->u.ml.real_master = real_master; spin_lock(&dlm->work_lock); list_add_tail(&item->list, &dlm->work_list); spin_unlock(&dlm->work_lock); schedule_work(&dlm->dispatched_work); leave: dlm_put(dlm); if (ret < 0) { if (buf) kfree(buf); if (item) kfree(item); } mlog_exit(ret); return ret; } static void dlm_mig_lockres_worker(struct dlm_work_item *item, void *data) { struct dlm_ctxt *dlm = data; struct dlm_migratable_lockres *mres; int ret = 0; struct dlm_lock_resource *res; u8 real_master; dlm = item->dlm; mres = (struct dlm_migratable_lockres *)data; res = item->u.ml.lockres; real_master = item->u.ml.real_master; if (real_master == DLM_LOCK_RES_OWNER_UNKNOWN) { /* this case is super-rare. only occurs if * node death happens during migration. */ again: ret = dlm_lockres_master_requery(dlm, res, &real_master); if (ret < 0) { mlog(0, "dlm_lockres_master_requery ret=%d\n", ret); goto again; } if (real_master == DLM_LOCK_RES_OWNER_UNKNOWN) { mlog(0, "lockres %.*s not claimed. " "this node will take it.\n", res->lockname.len, res->lockname.name); } else { mlog(0, "master needs to respond to sender " "that node %u still owns %.*s\n", real_master, res->lockname.len, res->lockname.name); /* cannot touch this lockres */ goto leave; } } ret = dlm_process_recovery_data(dlm, res, mres); if (ret < 0) mlog(0, "dlm_process_recovery_data returned %d\n", ret); else mlog(0, "dlm_process_recovery_data succeeded\n"); if ((mres->flags & (DLM_MRES_MIGRATION|DLM_MRES_ALL_DONE)) == (DLM_MRES_MIGRATION|DLM_MRES_ALL_DONE)) { ret = dlm_finish_migration(dlm, res, mres->master); if (ret < 0) mlog_errno(ret); } leave: kfree(data); mlog_exit(ret); } int dlm_lockres_master_requery(struct dlm_ctxt *dlm, struct dlm_lock_resource *res, u8 *real_master) { struct dlm_node_iter iter; int nodenum; int ret = 0; *real_master = DLM_LOCK_RES_OWNER_UNKNOWN; /* we only reach here if one of the two nodes in a * migration died while the migration was in progress. * at this point we need to requery the master. we * know that the new_master got as far as creating * an mle on at least one node, but we do not know * if any nodes had actually cleared the mle and set * the master to the new_master. the old master * is supposed to set the owner to UNKNOWN in the * event of a new_master death, so the only possible * responses that we can get from nodes here are * that the master is new_master, or that the master * is UNKNOWN. * if all nodes come back with UNKNOWN then we know * the lock needs remastering here. * if any node comes back with a valid master, check * to see if that master is the one that we are * recovering. if so, then the new_master died and * we need to remaster this lock. if not, then the * new_master survived and that node will respond to * other nodes about the owner. * if there is an owner, this node needs to dump this * lockres and alert the sender that this lockres * was rejected. */ spin_lock(&dlm->spinlock); dlm_node_iter_init(dlm->domain_map, &iter); spin_unlock(&dlm->spinlock); while ((nodenum = dlm_node_iter_next(&iter)) >= 0) { /* do not send to self */ if (nodenum == dlm->node_num) continue; ret = dlm_do_master_requery(dlm, res, nodenum, real_master); if (ret < 0) { mlog_errno(ret); if (!dlm_is_host_down(ret)) BUG(); /* host is down, so answer for that node would be * DLM_LOCK_RES_OWNER_UNKNOWN. continue. */ } if (*real_master != DLM_LOCK_RES_OWNER_UNKNOWN) { mlog(0, "lock master is %u\n", *real_master); break; } } return ret; } int dlm_do_master_requery(struct dlm_ctxt *dlm, struct dlm_lock_resource *res, u8 nodenum, u8 *real_master) { int ret = -EINVAL; struct dlm_master_requery req; int status = DLM_LOCK_RES_OWNER_UNKNOWN; memset(&req, 0, sizeof(req)); req.node_idx = dlm->node_num; req.namelen = res->lockname.len; memcpy(req.name, res->lockname.name, res->lockname.len); ret = o2net_send_message(DLM_MASTER_REQUERY_MSG, dlm->key, &req, sizeof(req), nodenum, &status); /* XXX: negative status not handled properly here. */ if (ret < 0) mlog_errno(ret); else { BUG_ON(status < 0); BUG_ON(status > DLM_LOCK_RES_OWNER_UNKNOWN); *real_master = (u8) (status & 0xff); mlog(0, "node %u responded to master requery with %u\n", nodenum, *real_master); ret = 0; } return ret; } /* this function cannot error, so unless the sending * or receiving of the message failed, the owner can * be trusted */ int dlm_master_requery_handler(struct o2net_msg *msg, u32 len, void *data) { struct dlm_ctxt *dlm = data; struct dlm_master_requery *req = (struct dlm_master_requery *)msg->buf; struct dlm_lock_resource *res = NULL; unsigned int hash; int master = DLM_LOCK_RES_OWNER_UNKNOWN; u32 flags = DLM_ASSERT_MASTER_REQUERY; if (!dlm_grab(dlm)) { /* since the domain has gone away on this * node, the proper response is UNKNOWN */ return master; } hash = dlm_lockid_hash(req->name, req->namelen); spin_lock(&dlm->spinlock); res = __dlm_lookup_lockres(dlm, req->name, req->namelen, hash); if (res) { spin_lock(&res->spinlock); master = res->owner; if (master == dlm->node_num) { int ret = dlm_dispatch_assert_master(dlm, res, 0, 0, flags); if (ret < 0) { mlog_errno(-ENOMEM); /* retry!? */ BUG(); } } spin_unlock(&res->spinlock); } spin_unlock(&dlm->spinlock); dlm_put(dlm); return master; } static inline struct list_head * dlm_list_num_to_pointer(struct dlm_lock_resource *res, int list_num) { struct list_head *ret; BUG_ON(list_num < 0); BUG_ON(list_num > 2); ret = &(res->granted); ret += list_num; return ret; } /* TODO: do ast flush business * TODO: do MIGRATING and RECOVERING spinning */ /* * NOTE about in-flight requests during migration: * * Before attempting the migrate, the master has marked the lockres as * MIGRATING and then flushed all of its pending ASTS. So any in-flight * requests either got queued before the MIGRATING flag got set, in which * case the lock data will reflect the change and a return message is on * the way, or the request failed to get in before MIGRATING got set. In * this case, the caller will be told to spin and wait for the MIGRATING * flag to be dropped, then recheck the master. * This holds true for the convert, cancel and unlock cases, and since lvb * updates are tied to these same messages, it applies to lvb updates as * well. For the lock case, there is no way a lock can be on the master * queue and not be on the secondary queue since the lock is always added * locally first. This means that the new target node will never be sent * a lock that he doesn't already have on the list. * In total, this means that the local lock is correct and should not be * updated to match the one sent by the master. Any messages sent back * from the master before the MIGRATING flag will bring the lock properly * up-to-date, and the change will be ordered properly for the waiter. * We will *not* attempt to modify the lock underneath the waiter. */ static int dlm_process_recovery_data(struct dlm_ctxt *dlm, struct dlm_lock_resource *res, struct dlm_migratable_lockres *mres) { struct dlm_migratable_lock *ml; struct list_head *queue; struct dlm_lock *newlock = NULL; struct dlm_lockstatus *lksb = NULL; int ret = 0; int i; struct list_head *iter; struct dlm_lock *lock = NULL; mlog(0, "running %d locks for this lockres\n", mres->num_locks); for (i=0; inum_locks; i++) { ml = &(mres->ml[i]); BUG_ON(ml->highest_blocked != LKM_IVMODE); newlock = NULL; lksb = NULL; queue = dlm_list_num_to_pointer(res, ml->list); /* if the lock is for the local node it needs to * be moved to the proper location within the queue. * do not allocate a new lock structure. */ if (ml->node == dlm->node_num) { /* MIGRATION ONLY! */ BUG_ON(!(mres->flags & DLM_MRES_MIGRATION)); spin_lock(&res->spinlock); list_for_each(iter, queue) { lock = list_entry (iter, struct dlm_lock, list); if (lock->ml.cookie != ml->cookie) lock = NULL; else break; } /* lock is always created locally first, and * destroyed locally last. it must be on the list */ if (!lock) { u64 c = ml->cookie; mlog(ML_ERROR, "could not find local lock " "with cookie %u:%llu!\n", dlm_get_lock_cookie_node(c), dlm_get_lock_cookie_seq(c)); BUG(); } BUG_ON(lock->ml.node != ml->node); /* see NOTE above about why we do not update * to match the master here */ /* move the lock to its proper place */ /* do not alter lock refcount. switching lists. */ list_move_tail(&lock->list, queue); spin_unlock(&res->spinlock); mlog(0, "just reordered a local lock!\n"); continue; } /* lock is for another node. */ newlock = dlm_new_lock(ml->type, ml->node, be64_to_cpu(ml->cookie), NULL); if (!newlock) { ret = -ENOMEM; goto leave; } lksb = newlock->lksb; dlm_lock_attach_lockres(newlock, res); if (ml->convert_type != LKM_IVMODE) { BUG_ON(queue != &res->converting); newlock->ml.convert_type = ml->convert_type; } lksb->flags |= (ml->flags & (DLM_LKSB_PUT_LVB|DLM_LKSB_GET_LVB)); if (mres->lvb[0]) { if (lksb->flags & DLM_LKSB_PUT_LVB) { /* other node was trying to update * lvb when node died. recreate the * lksb with the updated lvb. */ memcpy(lksb->lvb, mres->lvb, DLM_LVB_LEN); } else { /* otherwise, the node is sending its * most recent valid lvb info */ BUG_ON(ml->type != LKM_EXMODE && ml->type != LKM_PRMODE); if (res->lvb[0] && (ml->type == LKM_EXMODE || memcmp(res->lvb, mres->lvb, DLM_LVB_LEN))) { mlog(ML_ERROR, "received bad lvb!\n"); __dlm_print_one_lock_resource(res); BUG(); } memcpy(res->lvb, mres->lvb, DLM_LVB_LEN); } } /* NOTE: * wrt lock queue ordering and recovery: * 1. order of locks on granted queue is * meaningless. * 2. order of locks on converting queue is * LOST with the node death. sorry charlie. * 3. order of locks on the blocked queue is * also LOST. * order of locks does not affect integrity, it * just means that a lock request may get pushed * back in line as a result of the node death. * also note that for a given node the lock order * for its secondary queue locks is preserved * relative to each other, but clearly *not* * preserved relative to locks from other nodes. */ spin_lock(&res->spinlock); dlm_lock_get(newlock); list_add_tail(&newlock->list, queue); spin_unlock(&res->spinlock); } mlog(0, "done running all the locks\n"); leave: if (ret < 0) { mlog_errno(ret); if (newlock) dlm_lock_put(newlock); } mlog_exit(ret); return ret; } void dlm_move_lockres_to_recovery_list(struct dlm_ctxt *dlm, struct dlm_lock_resource *res) { int i; struct list_head *queue, *iter, *iter2; struct dlm_lock *lock; res->state |= DLM_LOCK_RES_RECOVERING; if (!list_empty(&res->recovering)) list_del_init(&res->recovering); list_add_tail(&res->recovering, &dlm->reco.resources); /* find any pending locks and put them back on proper list */ for (i=DLM_BLOCKED_LIST; i>=DLM_GRANTED_LIST; i--) { queue = dlm_list_idx_to_ptr(res, i); list_for_each_safe(iter, iter2, queue) { lock = list_entry (iter, struct dlm_lock, list); dlm_lock_get(lock); if (lock->convert_pending) { /* move converting lock back to granted */ BUG_ON(i != DLM_CONVERTING_LIST); mlog(0, "node died with convert pending " "on %.*s. move back to granted list.\n", res->lockname.len, res->lockname.name); dlm_revert_pending_convert(res, lock); lock->convert_pending = 0; } else if (lock->lock_pending) { /* remove pending lock requests completely */ BUG_ON(i != DLM_BLOCKED_LIST); mlog(0, "node died with lock pending " "on %.*s. remove from blocked list and skip.\n", res->lockname.len, res->lockname.name); /* lock will be floating until ref in * dlmlock_remote is freed after the network * call returns. ok for it to not be on any * list since no ast can be called * (the master is dead). */ dlm_revert_pending_lock(res, lock); lock->lock_pending = 0; } else if (lock->unlock_pending) { /* if an unlock was in progress, treat as * if this had completed successfully * before sending this lock state to the * new master. note that the dlm_unlock * call is still responsible for calling * the unlockast. that will happen after * the network call times out. for now, * just move lists to prepare the new * recovery master. */ BUG_ON(i != DLM_GRANTED_LIST); mlog(0, "node died with unlock pending " "on %.*s. remove from blocked list and skip.\n", res->lockname.len, res->lockname.name); dlm_commit_pending_unlock(res, lock); lock->unlock_pending = 0; } else if (lock->cancel_pending) { /* if a cancel was in progress, treat as * if this had completed successfully * before sending this lock state to the * new master */ BUG_ON(i != DLM_CONVERTING_LIST); mlog(0, "node died with cancel pending " "on %.*s. move back to granted list.\n", res->lockname.len, res->lockname.name); dlm_commit_pending_cancel(res, lock); lock->cancel_pending = 0; } dlm_lock_put(lock); } } } /* removes all recovered locks from the recovery list. * sets the res->owner to the new master. * unsets the RECOVERY flag and wakes waiters. */ static void dlm_finish_local_lockres_recovery(struct dlm_ctxt *dlm, u8 dead_node, u8 new_master) { int i; struct list_head *iter, *iter2; struct hlist_node *hash_iter; struct hlist_head *bucket; struct dlm_lock_resource *res; mlog_entry_void(); assert_spin_locked(&dlm->spinlock); list_for_each_safe(iter, iter2, &dlm->reco.resources) { res = list_entry (iter, struct dlm_lock_resource, recovering); if (res->owner == dead_node) { list_del_init(&res->recovering); spin_lock(&res->spinlock); dlm_change_lockres_owner(dlm, res, new_master); res->state &= ~DLM_LOCK_RES_RECOVERING; __dlm_dirty_lockres(dlm, res); spin_unlock(&res->spinlock); wake_up(&res->wq); } } /* this will become unnecessary eventually, but * for now we need to run the whole hash, clear * the RECOVERING state and set the owner * if necessary */ for (i = 0; i < DLM_HASH_BUCKETS; i++) { bucket = dlm_lockres_hash(dlm, i); hlist_for_each_entry(res, hash_iter, bucket, hash_node) { if (res->state & DLM_LOCK_RES_RECOVERING) { if (res->owner == dead_node) { mlog(0, "(this=%u) res %.*s owner=%u " "was not on recovering list, but " "clearing state anyway\n", dlm->node_num, res->lockname.len, res->lockname.name, new_master); } else if (res->owner == dlm->node_num) { mlog(0, "(this=%u) res %.*s owner=%u " "was not on recovering list, " "owner is THIS node, clearing\n", dlm->node_num, res->lockname.len, res->lockname.name, new_master); } else continue; if (!list_empty(&res->recovering)) { mlog(0, "%s:%.*s: lockres was " "marked RECOVERING, owner=%u\n", dlm->name, res->lockname.len, res->lockname.name, res->owner); list_del_init(&res->recovering); } spin_lock(&res->spinlock); dlm_change_lockres_owner(dlm, res, new_master); res->state &= ~DLM_LOCK_RES_RECOVERING; __dlm_dirty_lockres(dlm, res); spin_unlock(&res->spinlock); wake_up(&res->wq); } } } } static inline int dlm_lvb_needs_invalidation(struct dlm_lock *lock, int local) { if (local) { if (lock->ml.type != LKM_EXMODE && lock->ml.type != LKM_PRMODE) return 1; } else if (lock->ml.type == LKM_EXMODE) return 1; return 0; } static void dlm_revalidate_lvb(struct dlm_ctxt *dlm, struct dlm_lock_resource *res, u8 dead_node) { struct list_head *iter, *queue; struct dlm_lock *lock; int blank_lvb = 0, local = 0; int i; u8 search_node; assert_spin_locked(&dlm->spinlock); assert_spin_locked(&res->spinlock); if (res->owner == dlm->node_num) /* if this node owned the lockres, and if the dead node * had an EX when he died, blank out the lvb */ search_node = dead_node; else { /* if this is a secondary lockres, and we had no EX or PR * locks granted, we can no longer trust the lvb */ search_node = dlm->node_num; local = 1; /* check local state for valid lvb */ } for (i=DLM_GRANTED_LIST; i<=DLM_CONVERTING_LIST; i++) { queue = dlm_list_idx_to_ptr(res, i); list_for_each(iter, queue) { lock = list_entry (iter, struct dlm_lock, list); if (lock->ml.node == search_node) { if (dlm_lvb_needs_invalidation(lock, local)) { /* zero the lksb lvb and lockres lvb */ blank_lvb = 1; memset(lock->lksb->lvb, 0, DLM_LVB_LEN); } } } } if (blank_lvb) { mlog(0, "clearing %.*s lvb, dead node %u had EX\n", res->lockname.len, res->lockname.name, dead_node); memset(res->lvb, 0, DLM_LVB_LEN); } } static void dlm_free_dead_locks(struct dlm_ctxt *dlm, struct dlm_lock_resource *res, u8 dead_node) { struct list_head *iter, *tmpiter; struct dlm_lock *lock; /* this node is the lockres master: * 1) remove any stale locks for the dead node * 2) if the dead node had an EX when he died, blank out the lvb */ assert_spin_locked(&dlm->spinlock); assert_spin_locked(&res->spinlock); /* TODO: check pending_asts, pending_basts here */ list_for_each_safe(iter, tmpiter, &res->granted) { lock = list_entry (iter, struct dlm_lock, list); if (lock->ml.node == dead_node) { list_del_init(&lock->list); dlm_lock_put(lock); } } list_for_each_safe(iter, tmpiter, &res->converting) { lock = list_entry (iter, struct dlm_lock, list); if (lock->ml.node == dead_node) { list_del_init(&lock->list); dlm_lock_put(lock); } } list_for_each_safe(iter, tmpiter, &res->blocked) { lock = list_entry (iter, struct dlm_lock, list); if (lock->ml.node == dead_node) { list_del_init(&lock->list); dlm_lock_put(lock); } } /* do not kick thread yet */ __dlm_dirty_lockres(dlm, res); } /* if this node is the recovery master, and there are no * locks for a given lockres owned by this node that are in * either PR or EX mode, zero out the lvb before requesting. * */ static void dlm_do_local_recovery_cleanup(struct dlm_ctxt *dlm, u8 dead_node) { struct hlist_node *iter; struct dlm_lock_resource *res; int i; struct hlist_head *bucket; struct dlm_lock *lock; /* purge any stale mles */ dlm_clean_master_list(dlm, dead_node); /* * now clean up all lock resources. there are two rules: * * 1) if the dead node was the master, move the lockres * to the recovering list. set the RECOVERING flag. * this lockres needs to be cleaned up before it can * be used further. * * 2) if this node was the master, remove all locks from * each of the lockres queues that were owned by the * dead node. once recovery finishes, the dlm thread * can be kicked again to see if any ASTs or BASTs * need to be fired as a result. */ for (i = 0; i < DLM_HASH_BUCKETS; i++) { bucket = dlm_lockres_hash(dlm, i); hlist_for_each_entry(res, iter, bucket, hash_node) { /* always prune any $RECOVERY entries for dead nodes, * otherwise hangs can occur during later recovery */ if (dlm_is_recovery_lock(res->lockname.name, res->lockname.len)) { spin_lock(&res->spinlock); list_for_each_entry(lock, &res->granted, list) { if (lock->ml.node == dead_node) { mlog(0, "AHA! there was " "a $RECOVERY lock for dead " "node %u (%s)!\n", dead_node, dlm->name); list_del_init(&lock->list); dlm_lock_put(lock); break; } } spin_unlock(&res->spinlock); continue; } spin_lock(&res->spinlock); /* zero the lvb if necessary */ dlm_revalidate_lvb(dlm, res, dead_node); if (res->owner == dead_node) dlm_move_lockres_to_recovery_list(dlm, res); else if (res->owner == dlm->node_num) { dlm_free_dead_locks(dlm, res, dead_node); __dlm_lockres_calc_usage(dlm, res); } spin_unlock(&res->spinlock); } } } static void __dlm_hb_node_down(struct dlm_ctxt *dlm, int idx) { assert_spin_locked(&dlm->spinlock); /* check to see if the node is already considered dead */ if (!test_bit(idx, dlm->live_nodes_map)) { mlog(0, "for domain %s, node %d is already dead. " "another node likely did recovery already.\n", dlm->name, idx); return; } /* check to see if we do not care about this node */ if (!test_bit(idx, dlm->domain_map)) { /* This also catches the case that we get a node down * but haven't joined the domain yet. */ mlog(0, "node %u already removed from domain!\n", idx); return; } clear_bit(idx, dlm->live_nodes_map); /* Clean up join state on node death. */ if (dlm->joining_node == idx) { mlog(0, "Clearing join state for node %u\n", idx); __dlm_set_joining_node(dlm, DLM_LOCK_RES_OWNER_UNKNOWN); } /* make sure local cleanup occurs before the heartbeat events */ if (!test_bit(idx, dlm->recovery_map)) dlm_do_local_recovery_cleanup(dlm, idx); /* notify anything attached to the heartbeat events */ dlm_hb_event_notify_attached(dlm, idx, 0); mlog(0, "node %u being removed from domain map!\n", idx); clear_bit(idx, dlm->domain_map); /* wake up migration waiters if a node goes down. * perhaps later we can genericize this for other waiters. */ wake_up(&dlm->migration_wq); if (test_bit(idx, dlm->recovery_map)) mlog(0, "domain %s, node %u already added " "to recovery map!\n", dlm->name, idx); else set_bit(idx, dlm->recovery_map); } void dlm_hb_node_down_cb(struct o2nm_node *node, int idx, void *data) { struct dlm_ctxt *dlm = data; if (!dlm_grab(dlm)) return; spin_lock(&dlm->spinlock); __dlm_hb_node_down(dlm, idx); spin_unlock(&dlm->spinlock); dlm_put(dlm); } void dlm_hb_node_up_cb(struct o2nm_node *node, int idx, void *data) { struct dlm_ctxt *dlm = data; if (!dlm_grab(dlm)) return; spin_lock(&dlm->spinlock); set_bit(idx, dlm->live_nodes_map); /* do NOT notify mle attached to the heartbeat events. * new nodes are not interesting in mastery until joined. */ spin_unlock(&dlm->spinlock); dlm_put(dlm); } static void dlm_reco_ast(void *astdata) { struct dlm_ctxt *dlm = astdata; mlog(0, "ast for recovery lock fired!, this=%u, dlm=%s\n", dlm->node_num, dlm->name); } static void dlm_reco_bast(void *astdata, int blocked_type) { struct dlm_ctxt *dlm = astdata; mlog(0, "bast for recovery lock fired!, this=%u, dlm=%s\n", dlm->node_num, dlm->name); } static void dlm_reco_unlock_ast(void *astdata, enum dlm_status st) { mlog(0, "unlockast for recovery lock fired!\n"); } /* * dlm_pick_recovery_master will continually attempt to use * dlmlock() on the special "$RECOVERY" lockres with the * LKM_NOQUEUE flag to get an EX. every thread that enters * this function on each node racing to become the recovery * master will not stop attempting this until either: * a) this node gets the EX (and becomes the recovery master), * or b) dlm->reco.new_master gets set to some nodenum * != O2NM_INVALID_NODE_NUM (another node will do the reco). * so each time a recovery master is needed, the entire cluster * will sync at this point. if the new master dies, that will * be detected in dlm_do_recovery */ static int dlm_pick_recovery_master(struct dlm_ctxt *dlm) { enum dlm_status ret; struct dlm_lockstatus lksb; int status = -EINVAL; mlog(0, "starting recovery of %s at %lu, dead=%u, this=%u\n", dlm->name, jiffies, dlm->reco.dead_node, dlm->node_num); again: memset(&lksb, 0, sizeof(lksb)); ret = dlmlock(dlm, LKM_EXMODE, &lksb, LKM_NOQUEUE|LKM_RECOVERY, DLM_RECOVERY_LOCK_NAME, dlm_reco_ast, dlm, dlm_reco_bast); mlog(0, "%s: dlmlock($RECOVERY) returned %d, lksb=%d\n", dlm->name, ret, lksb.status); if (ret == DLM_NORMAL) { mlog(0, "dlm=%s dlmlock says I got it (this=%u)\n", dlm->name, dlm->node_num); /* got the EX lock. check to see if another node * just became the reco master */ if (dlm_reco_master_ready(dlm)) { mlog(0, "%s: got reco EX lock, but %u will " "do the recovery\n", dlm->name, dlm->reco.new_master); status = -EEXIST; } else { status = 0; /* see if recovery was already finished elsewhere */ spin_lock(&dlm->spinlock); if (dlm->reco.dead_node == O2NM_INVALID_NODE_NUM) { status = -EINVAL; mlog(0, "%s: got reco EX lock, but " "node got recovered already\n", dlm->name); if (dlm->reco.new_master != O2NM_INVALID_NODE_NUM) { mlog(ML_ERROR, "%s: new master is %u " "but no dead node!\n", dlm->name, dlm->reco.new_master); BUG(); } } spin_unlock(&dlm->spinlock); } /* if this node has actually become the recovery master, * set the master and send the messages to begin recovery */ if (!status) { mlog(0, "%s: dead=%u, this=%u, sending " "begin_reco now\n", dlm->name, dlm->reco.dead_node, dlm->node_num); status = dlm_send_begin_reco_message(dlm, dlm->reco.dead_node); /* this always succeeds */ BUG_ON(status); /* set the new_master to this node */ spin_lock(&dlm->spinlock); dlm->reco.new_master = dlm->node_num; spin_unlock(&dlm->spinlock); } /* recovery lock is a special case. ast will not get fired, * so just go ahead and unlock it. */ ret = dlmunlock(dlm, &lksb, 0, dlm_reco_unlock_ast, dlm); if (ret == DLM_DENIED) { mlog(0, "got DLM_DENIED, trying LKM_CANCEL\n"); ret = dlmunlock(dlm, &lksb, LKM_CANCEL, dlm_reco_unlock_ast, dlm); } if (ret != DLM_NORMAL) { /* this would really suck. this could only happen * if there was a network error during the unlock * because of node death. this means the unlock * is actually "done" and the lock structure is * even freed. we can continue, but only * because this specific lock name is special. */ mlog(ML_ERROR, "dlmunlock returned %d\n", ret); } } else if (ret == DLM_NOTQUEUED) { mlog(0, "dlm=%s dlmlock says another node got it (this=%u)\n", dlm->name, dlm->node_num); /* another node is master. wait on * reco.new_master != O2NM_INVALID_NODE_NUM * for at most one second */ wait_event_timeout(dlm->dlm_reco_thread_wq, dlm_reco_master_ready(dlm), msecs_to_jiffies(1000)); if (!dlm_reco_master_ready(dlm)) { mlog(0, "%s: reco master taking awhile\n", dlm->name); goto again; } /* another node has informed this one that it is reco master */ mlog(0, "%s: reco master %u is ready to recover %u\n", dlm->name, dlm->reco.new_master, dlm->reco.dead_node); status = -EEXIST; } else { struct dlm_lock_resource *res; /* dlmlock returned something other than NOTQUEUED or NORMAL */ mlog(ML_ERROR, "%s: got %s from dlmlock($RECOVERY), " "lksb.status=%s\n", dlm->name, dlm_errname(ret), dlm_errname(lksb.status)); res = dlm_lookup_lockres(dlm, DLM_RECOVERY_LOCK_NAME, DLM_RECOVERY_LOCK_NAME_LEN); if (res) { dlm_print_one_lock_resource(res); dlm_lockres_put(res); } else { mlog(ML_ERROR, "recovery lock not found\n"); } BUG(); } return status; } static int dlm_send_begin_reco_message(struct dlm_ctxt *dlm, u8 dead_node) { struct dlm_begin_reco br; int ret = 0; struct dlm_node_iter iter; int nodenum; int status; mlog_entry("%u\n", dead_node); mlog(0, "dead node is %u\n", dead_node); spin_lock(&dlm->spinlock); dlm_node_iter_init(dlm->domain_map, &iter); spin_unlock(&dlm->spinlock); clear_bit(dead_node, iter.node_map); memset(&br, 0, sizeof(br)); br.node_idx = dlm->node_num; br.dead_node = dead_node; while ((nodenum = dlm_node_iter_next(&iter)) >= 0) { ret = 0; if (nodenum == dead_node) { mlog(0, "not sending begin reco to dead node " "%u\n", dead_node); continue; } if (nodenum == dlm->node_num) { mlog(0, "not sending begin reco to self\n"); continue; } retry: ret = -EINVAL; mlog(0, "attempting to send begin reco msg to %d\n", nodenum); ret = o2net_send_message(DLM_BEGIN_RECO_MSG, dlm->key, &br, sizeof(br), nodenum, &status); /* negative status is handled ok by caller here */ if (ret >= 0) ret = status; if (dlm_is_host_down(ret)) { /* node is down. not involved in recovery * so just keep going */ mlog(0, "%s: node %u was down when sending " "begin reco msg (%d)\n", dlm->name, nodenum, ret); ret = 0; } if (ret < 0) { struct dlm_lock_resource *res; /* this is now a serious problem, possibly ENOMEM * in the network stack. must retry */ mlog_errno(ret); mlog(ML_ERROR, "begin reco of dlm %s to node %u " " returned %d\n", dlm->name, nodenum, ret); res = dlm_lookup_lockres(dlm, DLM_RECOVERY_LOCK_NAME, DLM_RECOVERY_LOCK_NAME_LEN); if (res) { dlm_print_one_lock_resource(res); dlm_lockres_put(res); } else { mlog(ML_ERROR, "recovery lock not found\n"); } /* sleep for a bit in hopes that we can avoid * another ENOMEM */ msleep(100); goto retry; } } return ret; } int dlm_begin_reco_handler(struct o2net_msg *msg, u32 len, void *data) { struct dlm_ctxt *dlm = data; struct dlm_begin_reco *br = (struct dlm_begin_reco *)msg->buf; /* ok to return 0, domain has gone away */ if (!dlm_grab(dlm)) return 0; mlog(0, "node %u wants to recover node %u\n", br->node_idx, br->dead_node); dlm_fire_domain_eviction_callbacks(dlm, br->dead_node); spin_lock(&dlm->spinlock); if (dlm->reco.new_master != O2NM_INVALID_NODE_NUM) { if (test_bit(dlm->reco.new_master, dlm->recovery_map)) { mlog(0, "%s: new_master %u died, changing " "to %u\n", dlm->name, dlm->reco.new_master, br->node_idx); } else { mlog(0, "%s: new_master %u NOT DEAD, changing " "to %u\n", dlm->name, dlm->reco.new_master, br->node_idx); /* may not have seen the new master as dead yet */ } } if (dlm->reco.dead_node != O2NM_INVALID_NODE_NUM) { mlog(ML_NOTICE, "%s: dead_node previously set to %u, " "node %u changing it to %u\n", dlm->name, dlm->reco.dead_node, br->node_idx, br->dead_node); } dlm->reco.new_master = br->node_idx; dlm->reco.dead_node = br->dead_node; if (!test_bit(br->dead_node, dlm->recovery_map)) { mlog(0, "recovery master %u sees %u as dead, but this " "node has not yet. marking %u as dead\n", br->node_idx, br->dead_node, br->dead_node); if (!test_bit(br->dead_node, dlm->domain_map) || !test_bit(br->dead_node, dlm->live_nodes_map)) mlog(0, "%u not in domain/live_nodes map " "so setting it in reco map manually\n", br->dead_node); /* force the recovery cleanup in __dlm_hb_node_down * both of these will be cleared in a moment */ set_bit(br->dead_node, dlm->domain_map); set_bit(br->dead_node, dlm->live_nodes_map); __dlm_hb_node_down(dlm, br->dead_node); } spin_unlock(&dlm->spinlock); dlm_kick_recovery_thread(dlm); dlm_put(dlm); return 0; } static int dlm_send_finalize_reco_message(struct dlm_ctxt *dlm) { int ret = 0; struct dlm_finalize_reco fr; struct dlm_node_iter iter; int nodenum; int status; mlog(0, "finishing recovery for node %s:%u\n", dlm->name, dlm->reco.dead_node); spin_lock(&dlm->spinlock); dlm_node_iter_init(dlm->domain_map, &iter); spin_unlock(&dlm->spinlock); memset(&fr, 0, sizeof(fr)); fr.node_idx = dlm->node_num; fr.dead_node = dlm->reco.dead_node; while ((nodenum = dlm_node_iter_next(&iter)) >= 0) { if (nodenum == dlm->node_num) continue; ret = o2net_send_message(DLM_FINALIZE_RECO_MSG, dlm->key, &fr, sizeof(fr), nodenum, &status); if (ret >= 0) { ret = status; if (dlm_is_host_down(ret)) { /* this has no effect on this recovery * session, so set the status to zero to * finish out the last recovery */ mlog(ML_ERROR, "node %u went down after this " "node finished recovery.\n", nodenum); ret = 0; } } if (ret < 0) { mlog_errno(ret); break; } } return ret; } int dlm_finalize_reco_handler(struct o2net_msg *msg, u32 len, void *data) { struct dlm_ctxt *dlm = data; struct dlm_finalize_reco *fr = (struct dlm_finalize_reco *)msg->buf; /* ok to return 0, domain has gone away */ if (!dlm_grab(dlm)) return 0; mlog(0, "node %u finalizing recovery of node %u\n", fr->node_idx, fr->dead_node); spin_lock(&dlm->spinlock); if (dlm->reco.new_master != fr->node_idx) { mlog(ML_ERROR, "node %u sent recovery finalize msg, but node " "%u is supposed to be the new master, dead=%u\n", fr->node_idx, dlm->reco.new_master, fr->dead_node); BUG(); } if (dlm->reco.dead_node != fr->dead_node) { mlog(ML_ERROR, "node %u sent recovery finalize msg for dead " "node %u, but node %u is supposed to be dead\n", fr->node_idx, fr->dead_node, dlm->reco.dead_node); BUG(); } dlm_finish_local_lockres_recovery(dlm, fr->dead_node, fr->node_idx); spin_unlock(&dlm->spinlock); dlm_reset_recovery(dlm); dlm_kick_recovery_thread(dlm); dlm_put(dlm); return 0; }