1 files changed, 142 insertions, 52 deletions

diff --git a/fs/dlm/recover.c b/fs/dlm/recover.c
index 3c025fe49ad..eaea789bf97 100644
--- a/fs/dlm/recover.c
+++ b/fs/dlm/recover.c
@@ -36,30 +36,23 @@
  * (LS_RECOVERY_STOP set due to failure of a node in ls_nodes).  When another
  * function thinks it could have completed the waited-on task, they should wake
  * up ls_wait_general to get an immediate response rather than waiting for the
- * timer to detect the result.  A timer wakes us up periodically while waiting
- * to see if we should abort due to a node failure.  This should only be called
- * by the dlm_recoverd thread.
+ * timeout.  This uses a timeout so it can check periodically if the wait
+ * should abort due to node failure (which doesn't cause a wake_up).
+ * This should only be called by the dlm_recoverd thread.
  */
 
-static void dlm_wait_timer_fn(unsigned long data)
-{
-	struct dlm_ls *ls = (struct dlm_ls *) data;
-	mod_timer(&ls->ls_timer, jiffies + (dlm_config.ci_recover_timer * HZ));
-	wake_up(&ls->ls_wait_general);
-}
-
 int dlm_wait_function(struct dlm_ls *ls, int (*testfn) (struct dlm_ls *ls))
 {
 	int error = 0;
+	int rv;
 
-	init_timer(&ls->ls_timer);
-	ls->ls_timer.function = dlm_wait_timer_fn;
-	ls->ls_timer.data = (long) ls;
-	ls->ls_timer.expires = jiffies + (dlm_config.ci_recover_timer * HZ);
-	add_timer(&ls->ls_timer);
-
-	wait_event(ls->ls_wait_general, testfn(ls) || dlm_recovery_stopped(ls));
-	del_timer_sync(&ls->ls_timer);
+	while (1) {
+		rv = wait_event_timeout(ls->ls_wait_general,
+					testfn(ls) || dlm_recovery_stopped(ls),
+					dlm_config.ci_recover_timer * HZ);
+		if (rv)
+			break;
+	}
 
 	if (dlm_recovery_stopped(ls)) {
 		log_debug(ls, "dlm_wait_function aborted");
@@ -277,32 +270,100 @@ static void recover_list_del(struct dlm_rsb *r)
 	dlm_put_rsb(r);
 }
 
-static struct dlm_rsb *recover_list_find(struct dlm_ls *ls, uint64_t id)
+static void recover_list_clear(struct dlm_ls *ls)
 {
-	struct dlm_rsb *r = NULL;
+	struct dlm_rsb *r, *s;
 
 	spin_lock(&ls->ls_recover_list_lock);
+	list_for_each_entry_safe(r, s, &ls->ls_recover_list, res_recover_list) {
+		list_del_init(&r->res_recover_list);
+		r->res_recover_locks_count = 0;
+		dlm_put_rsb(r);
+		ls->ls_recover_list_count--;
+	}
 
-	list_for_each_entry(r, &ls->ls_recover_list, res_recover_list) {
-		if (id == (unsigned long) r)
-			goto out;
+	if (ls->ls_recover_list_count != 0) {
+		log_error(ls, "warning: recover_list_count %d",
+			  ls->ls_recover_list_count);
+		ls->ls_recover_list_count = 0;
 	}
-	r = NULL;
- out:
 	spin_unlock(&ls->ls_recover_list_lock);
+}
+
+static int recover_idr_empty(struct dlm_ls *ls)
+{
+	int empty = 1;
+
+	spin_lock(&ls->ls_recover_idr_lock);
+	if (ls->ls_recover_list_count)
+		empty = 0;
+	spin_unlock(&ls->ls_recover_idr_lock);
+
+	return empty;
+}
+
+static int recover_idr_add(struct dlm_rsb *r)
+{
+	struct dlm_ls *ls = r->res_ls;
+	int rv;
+
+	idr_preload(GFP_NOFS);
+	spin_lock(&ls->ls_recover_idr_lock);
+	if (r->res_id) {
+		rv = -1;
+		goto out_unlock;
+	}
+	rv = idr_alloc(&ls->ls_recover_idr, r, 1, 0, GFP_NOWAIT);
+	if (rv < 0)
+		goto out_unlock;
+
+	r->res_id = rv;
+	ls->ls_recover_list_count++;
+	dlm_hold_rsb(r);
+	rv = 0;
+out_unlock:
+	spin_unlock(&ls->ls_recover_idr_lock);
+	idr_preload_end();
+	return rv;
+}
+
+static void recover_idr_del(struct dlm_rsb *r)
+{
+	struct dlm_ls *ls = r->res_ls;
+
+	spin_lock(&ls->ls_recover_idr_lock);
+	idr_remove(&ls->ls_recover_idr, r->res_id);
+	r->res_id = 0;
+	ls->ls_recover_list_count--;
+	spin_unlock(&ls->ls_recover_idr_lock);
+
+	dlm_put_rsb(r);
+}
+
+static struct dlm_rsb *recover_idr_find(struct dlm_ls *ls, uint64_t id)
+{
+	struct dlm_rsb *r;
+
+	spin_lock(&ls->ls_recover_idr_lock);
+	r = idr_find(&ls->ls_recover_idr, (int)id);
+	spin_unlock(&ls->ls_recover_idr_lock);
 	return r;
 }
 
-static void recover_list_clear(struct dlm_ls *ls)
+static void recover_idr_clear(struct dlm_ls *ls)
 {
-	struct dlm_rsb *r, *s;
+	struct dlm_rsb *r;
+	int id;
 
-	spin_lock(&ls->ls_recover_list_lock);
-	list_for_each_entry_safe(r, s, &ls->ls_recover_list, res_recover_list) {
-		list_del_init(&r->res_recover_list);
+	spin_lock(&ls->ls_recover_idr_lock);
+
+	idr_for_each_entry(&ls->ls_recover_idr, r, id) {
+		idr_remove(&ls->ls_recover_idr, id);
+		r->res_id = 0;
 		r->res_recover_locks_count = 0;
-		dlm_put_rsb(r);
 		ls->ls_recover_list_count--;
+
+		dlm_put_rsb(r);
 	}
 
 	if (ls->ls_recover_list_count != 0) {
@@ -310,7 +371,7 @@ static void recover_list_clear(struct dlm_ls *ls)
 			  ls->ls_recover_list_count);
 		ls->ls_recover_list_count = 0;
 	}
-	spin_unlock(&ls->ls_recover_list_lock);
+	spin_unlock(&ls->ls_recover_idr_lock);
 }
 
 
@@ -408,7 +469,7 @@ static int recover_master(struct dlm_rsb *r, unsigned int *count)
 		set_new_master(r);
 		error = 0;
 	} else {
-		recover_list_add(r);
+		recover_idr_add(r);
 		error = dlm_send_rcom_lookup(r, dir_nodeid);
 	}
 
@@ -465,7 +526,7 @@ int dlm_recover_masters(struct dlm_ls *ls)
 	int nodir = dlm_no_directory(ls);
 	int error;
 
-	log_debug(ls, "dlm_recover_masters");
+	log_rinfo(ls, "dlm_recover_masters");
 
 	down_read(&ls->ls_root_sem);
 	list_for_each_entry(r, &ls->ls_root_list, res_root_list) {
@@ -491,12 +552,12 @@ int dlm_recover_masters(struct dlm_ls *ls)
 	}
 	up_read(&ls->ls_root_sem);
 
-	log_debug(ls, "dlm_recover_masters %u of %u", count, total);
+	log_rinfo(ls, "dlm_recover_masters %u of %u", count, total);
 
-	error = dlm_wait_function(ls, &recover_list_empty);
+	error = dlm_wait_function(ls, &recover_idr_empty);
  out:
 	if (error)
-		recover_list_clear(ls);
+		recover_idr_clear(ls);
 	return error;
 }
 
@@ -505,7 +566,7 @@ int dlm_recover_master_reply(struct dlm_ls *ls, struct dlm_rcom *rc)
 	struct dlm_rsb *r;
 	int ret_nodeid, new_master;
 
-	r = recover_list_find(ls, rc->rc_id);
+	r = recover_idr_find(ls, rc->rc_id);
 	if (!r) {
 		log_error(ls, "dlm_recover_master_reply no id %llx",
 			  (unsigned long long)rc->rc_id);
@@ -524,9 +585,9 @@ int dlm_recover_master_reply(struct dlm_ls *ls, struct dlm_rcom *rc)
 	r->res_nodeid = new_master;
 	set_new_master(r);
 	unlock_rsb(r);
-	recover_list_del(r);
+	recover_idr_del(r);
 
-	if (recover_list_empty(ls))
+	if (recover_idr_empty(ls))
 		wake_up(&ls->ls_wait_general);
  out:
 	return 0;
@@ -624,7 +685,7 @@ int dlm_recover_locks(struct dlm_ls *ls)
 	}
 	up_read(&ls->ls_root_sem);
 
-	log_debug(ls, "dlm_recover_locks %d out", count);
+	log_rinfo(ls, "dlm_recover_locks %d out", count);
 
 	error = dlm_wait_function(ls, &recover_list_empty);
  out:
@@ -652,8 +713,14 @@ void dlm_recovered_lock(struct dlm_rsb *r)
  * the VALNOTVALID flag if necessary, and determining the correct lvb contents
  * based on the lvb's of the locks held on the rsb.
  *
- * RSB_VALNOTVALID is set if there are only NL/CR locks on the rsb.  If it
- * was already set prior to recovery, it's not cleared, regardless of locks.
+ * RSB_VALNOTVALID is set in two cases:
+ *
+ * 1. we are master, but not new, and we purged an EX/PW lock held by a
+ * failed node (in dlm_recover_purge which set RSB_RECOVER_LVB_INVAL)
+ *
+ * 2. we are a new master, and there are only NL/CR locks left.
+ * (We could probably improve this by only invaliding in this way when
+ * the previous master left uncleanly.  VMS docs mention that.)
  *
  * The LVB contents are only considered for changing when this is a new master
  * of the rsb (NEW_MASTER2).  Then, the rsb's lvb is taken from any lkb with
@@ -669,6 +736,19 @@ static void recover_lvb(struct dlm_rsb *r)
 	int big_lock_exists = 0;
 	int lvblen = r->res_ls->ls_lvblen;
 
+	if (!rsb_flag(r, RSB_NEW_MASTER2) &&
+	    rsb_flag(r, RSB_RECOVER_LVB_INVAL)) {
+		/* case 1 above */
+		rsb_set_flag(r, RSB_VALNOTVALID);
+		return;
+	}
+
+	if (!rsb_flag(r, RSB_NEW_MASTER2))
+		return;
+
+	/* we are the new master, so figure out if VALNOTVALID should
+	   be set, and set the rsb lvb from the best lkb available. */
+
 	list_for_each_entry(lkb, &r->res_grantqueue, lkb_statequeue) {
 		if (!(lkb->lkb_exflags & DLM_LKF_VALBLK))
 			continue;
@@ -707,13 +787,10 @@ static void recover_lvb(struct dlm_rsb *r)
 	if (!lock_lvb_exists)
 		goto out;
 
+	/* lvb is invalidated if only NL/CR locks remain */
 	if (!big_lock_exists)
 		rsb_set_flag(r, RSB_VALNOTVALID);
 
-	/* don't mess with the lvb unless we're the new master */
-	if (!rsb_flag(r, RSB_NEW_MASTER2))
-		goto out;
-
 	if (!r->res_lvbptr) {
 		r->res_lvbptr = dlm_allocate_lvb(r->res_ls);
 		if (!r->res_lvbptr)
@@ -739,6 +816,7 @@ static void recover_lvb(struct dlm_rsb *r)
 
 static void recover_conversion(struct dlm_rsb *r)
 {
+	struct dlm_ls *ls = r->res_ls;
 	struct dlm_lkb *lkb;
 	int grmode = -1;
 
@@ -753,10 +831,15 @@ static void recover_conversion(struct dlm_rsb *r)
 	list_for_each_entry(lkb, &r->res_convertqueue, lkb_statequeue) {
 		if (lkb->lkb_grmode != DLM_LOCK_IV)
 			continue;
-		if (grmode == -1)
+		if (grmode == -1) {
+			log_debug(ls, "recover_conversion %x set gr to rq %d",
+				  lkb->lkb_id, lkb->lkb_rqmode);
 			lkb->lkb_grmode = lkb->lkb_rqmode;
-		else
+		} else {
+			log_debug(ls, "recover_conversion %x set gr %d",
+				  lkb->lkb_id, grmode);
 			lkb->lkb_grmode = grmode;
+		}
 	}
 }
 
@@ -781,19 +864,26 @@ void dlm_recover_rsbs(struct dlm_ls *ls)
 		if (is_master(r)) {
 			if (rsb_flag(r, RSB_RECOVER_CONVERT))
 				recover_conversion(r);
+
+			/* recover lvb before granting locks so the updated
+			   lvb/VALNOTVALID is presented in the completion */
+			recover_lvb(r);
+
 			if (rsb_flag(r, RSB_NEW_MASTER2))
 				recover_grant(r);
-			recover_lvb(r);
 			count++;
+		} else {
+			rsb_clear_flag(r, RSB_VALNOTVALID);
 		}
 		rsb_clear_flag(r, RSB_RECOVER_CONVERT);
+		rsb_clear_flag(r, RSB_RECOVER_LVB_INVAL);
 		rsb_clear_flag(r, RSB_NEW_MASTER2);
 		unlock_rsb(r);
 	}
 	up_read(&ls->ls_root_sem);
 
 	if (count)
-		log_debug(ls, "dlm_recover_rsbs %d done", count);
+		log_rinfo(ls, "dlm_recover_rsbs %d done", count);
 }
 
 /* Create a single list of all root rsb's to be used during recovery */
@@ -860,6 +950,6 @@ void dlm_clear_toss(struct dlm_ls *ls)
 	}
 
 	if (count)
-		log_debug(ls, "dlm_clear_toss %u done", count);
+		log_rinfo(ls, "dlm_clear_toss %u done", count);
 }