1 files changed, 34 insertions, 13 deletions

diff --git a/fs/xfs/xfs_log.c b/fs/xfs/xfs_log.c
index 184c68ea62bb..160b8bb7ee60 100644
--- a/fs/xfs/xfs_log.c
+++ b/fs/xfs/xfs_log.c
@@ -787,6 +787,7 @@ xlog_force_iclog(
 	struct xlog_in_core	*iclog)
 {
 	atomic_inc(&iclog->ic_refcnt);
+	iclog->ic_flags |= XLOG_ICL_NEED_FLUSH | XLOG_ICL_NEED_FUA;
 	if (iclog->ic_state == XLOG_STATE_ACTIVE)
 		xlog_state_switch_iclogs(iclog->ic_log, iclog, 0);
 	return xlog_state_release_iclog(iclog->ic_log, iclog, 0);
@@ -877,7 +878,6 @@ out_err:
 
 	spin_lock(&log->l_icloglock);
 	iclog = log->l_iclog;
-	iclog->ic_flags |= (XLOG_ICL_NEED_FLUSH | XLOG_ICL_NEED_FUA);
 	error = xlog_force_iclog(iclog);
 	xlog_wait_on_iclog(iclog);
 
@@ -3214,22 +3214,23 @@ xfs_log_force(
 				goto out_unlock;
 		} else {
 			/*
-			 * Someone else is writing to this iclog.
-			 *
-			 * Use its call to flush out the data.  However, the
-			 * other thread may not force out this LR, so we mark
-			 * it WANT_SYNC.
+			 * Someone else is still writing to this iclog, so we
+			 * need to ensure that when they release the iclog it
+			 * gets synced immediately as we may be waiting on it.
 			 */
 			xlog_state_switch_iclogs(log, iclog, 0);
 		}
-	} else {
-		/*
-		 * If the head iclog is not active nor dirty, we just attach
-		 * ourselves to the head and go to sleep if necessary.
-		 */
-		;
 	}
 
+	/*
+	 * The iclog we are about to wait on may contain the checkpoint pushed
+	 * by the above xlog_cil_force() call, but it may not have been pushed
+	 * to disk yet. Like the ACTIVE case above, we need to make sure caches
+	 * are flushed when this iclog is written.
+	 */
+	if (iclog->ic_state == XLOG_STATE_WANT_SYNC)
+		iclog->ic_flags |= XLOG_ICL_NEED_FLUSH | XLOG_ICL_NEED_FUA;
+
 	if (flags & XFS_LOG_SYNC)
 		return xlog_wait_on_iclog(iclog);
 out_unlock:
@@ -3262,7 +3263,8 @@ xlog_force_lsn(
 			goto out_unlock;
 	}
 
-	if (iclog->ic_state == XLOG_STATE_ACTIVE) {
+	switch (iclog->ic_state) {
+	case XLOG_STATE_ACTIVE:
 		/*
 		 * We sleep here if we haven't already slept (e.g. this is the
 		 * first time we've looked at the correct iclog buf) and the
@@ -3289,6 +3291,25 @@ xlog_force_lsn(
 			goto out_error;
 		if (log_flushed)
 			*log_flushed = 1;
+		break;
+	case XLOG_STATE_WANT_SYNC:
+		/*
+		 * This iclog may contain the checkpoint pushed by the
+		 * xlog_cil_force_seq() call, but there are other writers still
+		 * accessing it so it hasn't been pushed to disk yet. Like the
+		 * ACTIVE case above, we need to make sure caches are flushed
+		 * when this iclog is written.
+		 */
+		iclog->ic_flags |= XLOG_ICL_NEED_FLUSH | XLOG_ICL_NEED_FUA;
+		break;
+	default:
+		/*
+		 * The entire checkpoint was written by the CIL force and is on
+		 * its way to disk already. It will be stable when it
+		 * completes, so we don't need to manipulate caches here at all.
+		 * We just need to wait for completion if necessary.
+		 */
+		break;
 	}
 
 	if (flags & XFS_LOG_SYNC)