[DLM] block dlm_recv in recovery transition

Introduce a per-lockspace rwsem that's held in read mode by dlm_recv
threads while working in the dlm.  This allows dlm_recv activity to be
suspended when the lockspace transitions to, from and between recovery
cycles.

The specific bug prompting this change is one where an in-progress
recovery cycle is aborted by a new recovery cycle.  While dlm_recv was
processing a recovery message, the recovery cycle was aborted and
dlm_recoverd began cleaning up.  dlm_recv decremented recover_locks_count
on an rsb after dlm_recoverd had reset it to zero.  This is fixed by
suspending dlm_recv (taking write lock on the rwsem) before aborting the
current recovery.

The transitions to/from normal and recovery modes are simplified by using
this new ability to block dlm_recv.  The switch from normal to recovery
mode means dlm_recv goes from processing locking messages, to saving them
for later, and vice versa.  Races are avoided by blocking dlm_recv when
setting the flag that switches between modes.

Signed-off-by: David Teigland <teigland@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

fs/dlm/dlm_internal.h

@@ -491,6 +491,7 @@ struct dlm_ls {
 	uint64_t		ls_recover_seq;
 	struct dlm_recover	*ls_recover_args;
 	struct rw_semaphore	ls_in_recovery;	/* block local requests */
+	struct rw_semaphore	ls_recv_active;	/* block dlm_recv */
 	struct list_head	ls_requestqueue;/* queue remote requests */
 	struct mutex		ls_requestqueue_mutex;
 	char			*ls_recover_buf;

fs/dlm/lock.c

@@ -3638,55 +3638,8 @@ static void receive_lookup_reply(struct dlm_ls *ls, struct dlm_message *ms)
 	dlm_put_lkb(lkb);
 }
 
-int dlm_receive_message(struct dlm_header *hd, int nodeid, int recovery)
+static void _receive_message(struct dlm_ls *ls, struct dlm_message *ms)
 {
-	struct dlm_message *ms = (struct dlm_message *) hd;
-	struct dlm_ls *ls;
-	int error = 0;
-
-	if (!recovery)
-		dlm_message_in(ms);
-
-	ls = dlm_find_lockspace_global(hd->h_lockspace);
-	if (!ls) {
-		log_print("drop message %d from %d for unknown lockspace %d",
-			  ms->m_type, nodeid, hd->h_lockspace);
-		return -EINVAL;
-	}
-
-	/* recovery may have just ended leaving a bunch of backed-up requests
-	   in the requestqueue; wait while dlm_recoverd clears them */
-
-	if (!recovery)
-		dlm_wait_requestqueue(ls);
-
-	/* recovery may have just started while there were a bunch of
-	   in-flight requests -- save them in requestqueue to be processed
-	   after recovery.  we can't let dlm_recvd block on the recovery
-	   lock.  if dlm_recoverd is calling this function to clear the
-	   requestqueue, it needs to be interrupted (-EINTR) if another
-	   recovery operation is starting. */
-
-	while (1) {
-		if (dlm_locking_stopped(ls)) {
-			if (recovery) {
-				error = -EINTR;
-				goto out;
-			}
-			error = dlm_add_requestqueue(ls, nodeid, hd);
-			if (error == -EAGAIN)
-				continue;
-			else {
-				error = -EINTR;
-				goto out;
-			}
-		}
-
-		if (dlm_lock_recovery_try(ls))
-			break;
-		schedule();
-	}
-
 	switch (ms->m_type) {
 
 	/* messages sent to a master node */
@@ -3761,17 +3714,90 @@ int dlm_receive_message(struct dlm_header *hd, int nodeid, int recovery)
 		log_error(ls, "unknown message type %d", ms->m_type);
 	}
 
-	dlm_unlock_recovery(ls);
- out:
-	dlm_put_lockspace(ls);
 	dlm_astd_wake();
-	return error;
 }
 
+/* If the lockspace is in recovery mode (locking stopped), then normal
+   messages are saved on the requestqueue for processing after recovery is
+   done.  When not in recovery mode, we wait for dlm_recoverd to drain saved
+   messages off the requestqueue before we process new ones. This occurs right
+   after recovery completes when we transition from saving all messages on
+   requestqueue, to processing all the saved messages, to processing new
+   messages as they arrive. */
 
-/*
+static void dlm_receive_message(struct dlm_ls *ls, struct dlm_message *ms,
- * Recovery related
+				int nodeid)
- */
+{
+	if (dlm_locking_stopped(ls)) {
+		dlm_add_requestqueue(ls, nodeid, (struct dlm_header *) ms);
+	} else {
+		dlm_wait_requestqueue(ls);
+		_receive_message(ls, ms);
+	}
+}
+
+/* This is called by dlm_recoverd to process messages that were saved on
+   the requestqueue. */
+
+void dlm_receive_message_saved(struct dlm_ls *ls, struct dlm_message *ms)
+{
+	_receive_message(ls, ms);
+}
+
+/* This is called by the midcomms layer when something is received for
+   the lockspace.  It could be either a MSG (normal message sent as part of
+   standard locking activity) or an RCOM (recovery message sent as part of
+   lockspace recovery). */
+
+void dlm_receive_buffer(struct dlm_header *hd, int nodeid)
+{
+	struct dlm_message *ms = (struct dlm_message *) hd;
+	struct dlm_rcom *rc = (struct dlm_rcom *) hd;
+	struct dlm_ls *ls;
+	int type = 0;
+
+	switch (hd->h_cmd) {
+	case DLM_MSG:
+		dlm_message_in(ms);
+		type = ms->m_type;
+		break;
+	case DLM_RCOM:
+		dlm_rcom_in(rc);
+		type = rc->rc_type;
+		break;
+	default:
+		log_print("invalid h_cmd %d from %u", hd->h_cmd, nodeid);
+		return;
+	}
+
+	if (hd->h_nodeid != nodeid) {
+		log_print("invalid h_nodeid %d from %d lockspace %x",
+			  hd->h_nodeid, nodeid, hd->h_lockspace);
+		return;
+	}
+
+	ls = dlm_find_lockspace_global(hd->h_lockspace);
+	if (!ls) {
+		log_print("invalid h_lockspace %x from %d cmd %d type %d",
+			  hd->h_lockspace, nodeid, hd->h_cmd, type);
+
+		if (hd->h_cmd == DLM_RCOM && type == DLM_RCOM_STATUS)
+			dlm_send_ls_not_ready(nodeid, rc);
+		return;
+	}
+
+	/* this rwsem allows dlm_ls_stop() to wait for all dlm_recv threads to
+	   be inactive (in this ls) before transitioning to recovery mode */
+
+	down_read(&ls->ls_recv_active);
+	if (hd->h_cmd == DLM_MSG)
+		dlm_receive_message(ls, ms, nodeid);
+	else
+		dlm_receive_rcom(ls, rc, nodeid);
+	up_read(&ls->ls_recv_active);
+
+	dlm_put_lockspace(ls);
+}
 
 static void recover_convert_waiter(struct dlm_ls *ls, struct dlm_lkb *lkb)
 {

fs/dlm/lock.h

@@ -16,7 +16,8 @@
 void dlm_print_rsb(struct dlm_rsb *r);
 void dlm_dump_rsb(struct dlm_rsb *r);
 void dlm_print_lkb(struct dlm_lkb *lkb);
-int dlm_receive_message(struct dlm_header *hd, int nodeid, int recovery);
+void dlm_receive_message_saved(struct dlm_ls *ls, struct dlm_message *ms);
+void dlm_receive_buffer(struct dlm_header *hd, int nodeid);
 int dlm_modes_compat(int mode1, int mode2);
 int dlm_find_rsb(struct dlm_ls *ls, char *name, int namelen,
 	unsigned int flags, struct dlm_rsb **r_ret);

fs/dlm/lockspace.c

@@ -519,6 +519,7 @@ static int new_lockspace(char *name, int namelen, void **lockspace,
 	ls->ls_recover_seq = 0;
 	ls->ls_recover_args = NULL;
 	init_rwsem(&ls->ls_in_recovery);
+	init_rwsem(&ls->ls_recv_active);
 	INIT_LIST_HEAD(&ls->ls_requestqueue);
 	mutex_init(&ls->ls_requestqueue_mutex);
 	mutex_init(&ls->ls_clear_proc_locks);

fs/dlm/member.c

@@ -18,10 +18,6 @@
 #include "rcom.h"
 #include "config.h"
 
-/*
- * Following called by dlm_recoverd thread
- */
-
 static void add_ordered_member(struct dlm_ls *ls, struct dlm_member *new)
 {
 	struct dlm_member *memb = NULL;
@@ -250,18 +246,30 @@ int dlm_recover_members(struct dlm_ls *ls, struct dlm_recover *rv, int *neg_out)
 	return error;
 }
 
-/*
+/* Userspace guarantees that dlm_ls_stop() has completed on all nodes before
- * Following called from lockspace.c
+   dlm_ls_start() is called on any of them to start the new recovery. */
- */
 
 int dlm_ls_stop(struct dlm_ls *ls)
 {
 	int new;
 
 	/*
-	 * A stop cancels any recovery that's in progress (see RECOVERY_STOP,
+	 * Prevent dlm_recv from being in the middle of something when we do
-	 * dlm_recovery_stopped()) and prevents any new locks from being
+	 * the stop.  This includes ensuring dlm_recv isn't processing a
-	 * processed (see RUNNING, dlm_locking_stopped()).
+	 * recovery message (rcom), while dlm_recoverd is aborting and
+	 * resetting things from an in-progress recovery.  i.e. we want
+	 * dlm_recoverd to abort its recovery without worrying about dlm_recv
+	 * processing an rcom at the same time.  Stopping dlm_recv also makes
+	 * it easy for dlm_receive_message() to check locking stopped and add a
+	 * message to the requestqueue without races.
+	 */
+
+	down_write(&ls->ls_recv_active);
+
+	/*
+	 * Abort any recovery that's in progress (see RECOVERY_STOP,
+	 * dlm_recovery_stopped()) and tell any other threads running in the
+	 * dlm to quit any processing (see RUNNING, dlm_locking_stopped()).
 	 */
 
 	spin_lock(&ls->ls_recover_lock);
@@ -270,9 +278,15 @@ int dlm_ls_stop(struct dlm_ls *ls)
 	ls->ls_recover_seq++;
 	spin_unlock(&ls->ls_recover_lock);
 
+	/*
+	 * Let dlm_recv run again, now any normal messages will be saved on the
+	 * requestqueue for later.
+	 */
+
+	up_write(&ls->ls_recv_active);
+
 	/*
 	 * This in_recovery lock does two things:
-	 *
 	 * 1) Keeps this function from returning until all threads are out
 	 *    of locking routines and locking is truely stopped.
 	 * 2) Keeps any new requests from being processed until it's unlocked
@@ -284,9 +298,8 @@ int dlm_ls_stop(struct dlm_ls *ls)
 
 	/*
 	 * The recoverd suspend/resume makes sure that dlm_recoverd (if
-	 * running) has noticed the clearing of RUNNING above and quit
+	 * running) has noticed RECOVERY_STOP above and quit processing the
-	 * processing the previous recovery.  This will be true for all nodes
+	 * previous recovery.
-	 * before any nodes start the new recovery.
 	 */
 
 	dlm_recoverd_suspend(ls);

fs/dlm/midcomms.c

@@ -2,7 +2,7 @@
 *******************************************************************************
 **
 **  Copyright (C) Sistina Software, Inc.  1997-2003  All rights reserved.
-**  Copyright (C) 2004-2005 Red Hat, Inc.  All rights reserved.
+**  Copyright (C) 2004-2007 Red Hat, Inc.  All rights reserved.
 **
 **  This copyrighted material is made available to anyone wishing to use,
 **  modify, copy, or redistribute it subject to the terms and conditions
@@ -27,7 +27,6 @@
 #include "dlm_internal.h"
 #include "lowcomms.h"
 #include "config.h"
-#include "rcom.h"
 #include "lock.h"
 #include "midcomms.h"
 
@@ -117,19 +116,7 @@ int dlm_process_incoming_buffer(int nodeid, const void *base,
 		offset &= (limit - 1);
 		len -= msglen;
 
-		switch (msg->h_cmd) {
+		dlm_receive_buffer(msg, nodeid);
-		case DLM_MSG:
-			dlm_receive_message(msg, nodeid, 0);
-			break;
-
-		case DLM_RCOM:
-			dlm_receive_rcom(msg, nodeid);
-			break;
-
-		default:
-			log_print("unknown msg type %x from %u: %u %u %u %u",
-				  msg->h_cmd, nodeid, msglen, len, offset, ret);
-		}
 	}
 
 	if (msg != (struct dlm_header *) __tmp)

fs/dlm/rcom.c

@@ -2,7 +2,7 @@
 *******************************************************************************
 **
 **  Copyright (C) Sistina Software, Inc.  1997-2003  All rights reserved.
-**  Copyright (C) 2005 Red Hat, Inc.  All rights reserved.
+**  Copyright (C) 2005-2007 Red Hat, Inc.  All rights reserved.
 **
 **  This copyrighted material is made available to anyone wishing to use,
 **  modify, copy, or redistribute it subject to the terms and conditions
@@ -386,7 +386,10 @@ static void receive_rcom_lock_reply(struct dlm_ls *ls, struct dlm_rcom *rc_in)
 	dlm_recover_process_copy(ls, rc_in);
 }
 
-static int send_ls_not_ready(int nodeid, struct dlm_rcom *rc_in)
+/* If the lockspace doesn't exist then still send a status message
+   back; it's possible that it just doesn't have its global_id yet. */
+
+int dlm_send_ls_not_ready(int nodeid, struct dlm_rcom *rc_in)
 {
 	struct dlm_rcom *rc;
 	struct rcom_config *rf;
@@ -446,28 +449,11 @@ static int is_old_reply(struct dlm_ls *ls, struct dlm_rcom *rc)
 	return rv;
 }
 
-/* Called by dlm_recvd; corresponds to dlm_receive_message() but special
+/* Called by dlm_recv; corresponds to dlm_receive_message() but special
    recovery-only comms are sent through here. */
 
-void dlm_receive_rcom(struct dlm_header *hd, int nodeid)
+void dlm_receive_rcom(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid)
 {
-	struct dlm_rcom *rc = (struct dlm_rcom *) hd;
-	struct dlm_ls *ls;
-
-	dlm_rcom_in(rc);
-
-	/* If the lockspace doesn't exist then still send a status message
-	   back; it's possible that it just doesn't have its global_id yet. */
-
-	ls = dlm_find_lockspace_global(hd->h_lockspace);
-	if (!ls) {
-		log_print("lockspace %x from %d type %x not found",
-			  hd->h_lockspace, nodeid, rc->rc_type);
-		if (rc->rc_type == DLM_RCOM_STATUS)
-			send_ls_not_ready(nodeid, rc);
-		return;
-	}
-
 	if (dlm_recovery_stopped(ls) && (rc->rc_type != DLM_RCOM_STATUS)) {
 		log_debug(ls, "ignoring recovery message %x from %d",
 			  rc->rc_type, nodeid);
@@ -477,12 +463,6 @@ void dlm_receive_rcom(struct dlm_header *hd, int nodeid)
 	if (is_old_reply(ls, rc))
 		goto out;
 
-	if (nodeid != rc->rc_header.h_nodeid) {
-		log_error(ls, "bad rcom nodeid %d from %d",
-			  rc->rc_header.h_nodeid, nodeid);
-		goto out;
-	}
-
 	switch (rc->rc_type) {
 	case DLM_RCOM_STATUS:
 		receive_rcom_status(ls, rc);
@@ -520,6 +500,6 @@ void dlm_receive_rcom(struct dlm_header *hd, int nodeid)
 		DLM_ASSERT(0, printk("rc_type=%x\n", rc->rc_type););
 	}
  out:
-	dlm_put_lockspace(ls);
+	return;
 }
 

fs/dlm/rcom.h

@@ -2,7 +2,7 @@
 *******************************************************************************
 **
 **  Copyright (C) Sistina Software, Inc.  1997-2003  All rights reserved.
-**  Copyright (C) 2005 Red Hat, Inc.  All rights reserved.
+**  Copyright (C) 2005-2007 Red Hat, Inc.  All rights reserved.
 **
 **  This copyrighted material is made available to anyone wishing to use,
 **  modify, copy, or redistribute it subject to the terms and conditions
@@ -18,7 +18,8 @@ int dlm_rcom_status(struct dlm_ls *ls, int nodeid);
 int dlm_rcom_names(struct dlm_ls *ls, int nodeid, char *last_name,int last_len);
 int dlm_send_rcom_lookup(struct dlm_rsb *r, int dir_nodeid);
 int dlm_send_rcom_lock(struct dlm_rsb *r, struct dlm_lkb *lkb);
-void dlm_receive_rcom(struct dlm_header *hd, int nodeid);
+void dlm_receive_rcom(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid);
+int dlm_send_ls_not_ready(int nodeid, struct dlm_rcom *rc_in);
 
 #endif
 

fs/dlm/recoverd.c

@@ -24,19 +24,28 @@
 
 
 /* If the start for which we're re-enabling locking (seq) has been superseded
-   by a newer stop (ls_recover_seq), we need to leave locking disabled. */
+   by a newer stop (ls_recover_seq), we need to leave locking disabled.
+
+   We suspend dlm_recv threads here to avoid the race where dlm_recv a) sees
+   locking stopped and b) adds a message to the requestqueue, but dlm_recoverd
+   enables locking and clears the requestqueue between a and b. */
 
 static int enable_locking(struct dlm_ls *ls, uint64_t seq)
 {
 	int error = -EINTR;
 
+	down_write(&ls->ls_recv_active);
+
 	spin_lock(&ls->ls_recover_lock);
 	if (ls->ls_recover_seq == seq) {
 		set_bit(LSFL_RUNNING, &ls->ls_flags);
+		/* unblocks processes waiting to enter the dlm */
 		up_write(&ls->ls_in_recovery);
 		error = 0;
 	}
 	spin_unlock(&ls->ls_recover_lock);
+
+	up_write(&ls->ls_recv_active);
 	return error;
 }
 

fs/dlm/requestqueue.c

@@ -1,7 +1,7 @@
 /******************************************************************************
 *******************************************************************************
 **
-**  Copyright (C) 2005 Red Hat, Inc.  All rights reserved.
+**  Copyright (C) 2005-2007 Red Hat, Inc.  All rights reserved.
 **
 **  This copyrighted material is made available to anyone wishing to use,
 **  modify, copy, or redistribute it subject to the terms and conditions
@@ -20,7 +20,7 @@
 struct rq_entry {
 	struct list_head list;
 	int nodeid;
-	char request[1];
+	char request[0];
 };
 
 /*
@@ -30,42 +30,39 @@ struct rq_entry {
  * lockspace is enabled on some while still suspended on others.
  */
 
-int dlm_add_requestqueue(struct dlm_ls *ls, int nodeid, struct dlm_header *hd)
+void dlm_add_requestqueue(struct dlm_ls *ls, int nodeid, struct dlm_header *hd)
 {
 	struct rq_entry *e;
 	int length = hd->h_length;
-	int rv = 0;
 
 	e = kmalloc(sizeof(struct rq_entry) + length, GFP_KERNEL);
 	if (!e) {
-		log_print("dlm_add_requestqueue: out of memory\n");
+		log_print("dlm_add_requestqueue: out of memory len %d", length);
-		return 0;
+		return;
 	}
 
 	e->nodeid = nodeid;
 	memcpy(e->request, hd, length);
 
-	/* We need to check dlm_locking_stopped() after taking the mutex to
-	   avoid a race where dlm_recoverd enables locking and runs
-	   process_requestqueue between our earlier dlm_locking_stopped check
-	   and this addition to the requestqueue. */
-
 	mutex_lock(&ls->ls_requestqueue_mutex);
-	if (dlm_locking_stopped(ls))
 	list_add_tail(&e->list, &ls->ls_requestqueue);
-	else {
-		log_debug(ls, "dlm_add_requestqueue skip from %d", nodeid);
-		kfree(e);
-		rv = -EAGAIN;
-	}
 	mutex_unlock(&ls->ls_requestqueue_mutex);
-	return rv;
 }
 
+/*
+ * Called by dlm_recoverd to process normal messages saved while recovery was
+ * happening.  Normal locking has been enabled before this is called.  dlm_recv
+ * upon receiving a message, will wait for all saved messages to be drained
+ * here before processing the message it got.  If a new dlm_ls_stop() arrives
+ * while we're processing these saved messages, it may block trying to suspend
+ * dlm_recv if dlm_recv is waiting for us in dlm_wait_requestqueue.  In that
+ * case, we don't abort since locking_stopped is still 0.  If dlm_recv is not
+ * waiting for us, then this processing may be aborted due to locking_stopped.
+ */
+
 int dlm_process_requestqueue(struct dlm_ls *ls)
 {
 	struct rq_entry *e;
-	struct dlm_header *hd;
 	int error = 0;
 
 	mutex_lock(&ls->ls_requestqueue_mutex);
@@ -79,14 +76,7 @@ int dlm_process_requestqueue(struct dlm_ls *ls)
 		e = list_entry(ls->ls_requestqueue.next, struct rq_entry, list);
 		mutex_unlock(&ls->ls_requestqueue_mutex);
 
-		hd = (struct dlm_header *) e->request;
+		dlm_receive_message_saved(ls, (struct dlm_message *)e->request);
-		error = dlm_receive_message(hd, e->nodeid, 1);
-
-		if (error == -EINTR) {
-			/* entry is left on requestqueue */
-			log_debug(ls, "process_requestqueue abort eintr");
-			break;
-		}
 
 		mutex_lock(&ls->ls_requestqueue_mutex);
 		list_del(&e->list);
@@ -106,10 +96,12 @@ int dlm_process_requestqueue(struct dlm_ls *ls)
 
 /*
  * After recovery is done, locking is resumed and dlm_recoverd takes all the
- * saved requests and processes them as they would have been by dlm_recvd.  At
+ * saved requests and processes them as they would have been by dlm_recv.  At
- * the same time, dlm_recvd will start receiving new requests from remote
+ * the same time, dlm_recv will start receiving new requests from remote nodes.
- * nodes.  We want to delay dlm_recvd processing new requests until
+ * We want to delay dlm_recv processing new requests until dlm_recoverd has
- * dlm_recoverd has finished processing the old saved requests.
+ * finished processing the old saved requests.  We don't check for locking
+ * stopped here because dlm_ls_stop won't stop locking until it's suspended us
+ * (dlm_recv).
  */
 
 void dlm_wait_requestqueue(struct dlm_ls *ls)
@@ -118,8 +110,6 @@ void dlm_wait_requestqueue(struct dlm_ls *ls)
 		mutex_lock(&ls->ls_requestqueue_mutex);
 		if (list_empty(&ls->ls_requestqueue))
 			break;
-		if (dlm_locking_stopped(ls))
-			break;
 		mutex_unlock(&ls->ls_requestqueue_mutex);
 		schedule();
 	}

fs/dlm/requestqueue.h

@@ -1,7 +1,7 @@
 /******************************************************************************
 *******************************************************************************
 **
-**  Copyright (C) 2005 Red Hat, Inc.  All rights reserved.
+**  Copyright (C) 2005-2007 Red Hat, Inc.  All rights reserved.
 **
 **  This copyrighted material is made available to anyone wishing to use,
 **  modify, copy, or redistribute it subject to the terms and conditions
@@ -13,7 +13,7 @@
 #ifndef __REQUESTQUEUE_DOT_H__
 #define __REQUESTQUEUE_DOT_H__
 
-int dlm_add_requestqueue(struct dlm_ls *ls, int nodeid, struct dlm_header *hd);
+void dlm_add_requestqueue(struct dlm_ls *ls, int nodeid, struct dlm_header *hd);
 int dlm_process_requestqueue(struct dlm_ls *ls);
 void dlm_wait_requestqueue(struct dlm_ls *ls);
 void dlm_purge_requestqueue(struct dlm_ls *ls);