[PATCH 10/10] xfs: create a polled function to force inode inactivation

["Darrick J. Wong" <darrick.wong@oracle.com>]

From: Darrick J. Wong <darrick.wong@oracle.com>

Create a polled version of xfs_inactive_force so that we can force
inactivation while holding a lock (usually the umount lock) without
tripping over the softlockup timer.  This is for callers that hold vfs
locks while calling inactivation, which is currently unmount, iunlink
processing during mount, and rw->ro remount.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 fs/xfs/xfs_icache.c |   52 +++++++++++++++++++++++++++++++++++++++++++++++++--
 fs/xfs/xfs_icache.h |    2 ++
 fs/xfs/xfs_mount.c  |    2 +-
 fs/xfs/xfs_mount.h  |    6 ++++++
 fs/xfs/xfs_super.c  |    3 ++-
 5 files changed, 61 insertions(+), 4 deletions(-)


diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c
index 5240e9e517d7..892bb789dcbf 100644
--- a/fs/xfs/xfs_icache.c
+++ b/fs/xfs/xfs_icache.c
@@ -25,6 +25,7 @@
 #include "xfs_health.h"
 
 #include <linux/iversion.h>
+#include <linux/nmi.h>
 
 STATIC int xfs_inode_free_eofblocks(struct xfs_inode *ip, struct xfs_perag *pag,
 		void *args);
@@ -2284,8 +2285,12 @@ static int
 xfs_inactive_inodes_pag(
 	struct xfs_perag	*pag)
 {
-	return xfs_ici_walk_ag(pag, &xfs_inactive_iwalk_ops, 0, NULL,
+	int			error;
+
+	error = xfs_ici_walk_ag(pag, &xfs_inactive_iwalk_ops, 0, NULL,
 			XFS_ICI_INACTIVE_TAG);
+	wake_up(&pag->pag_mount->m_inactive_wait);
+	return error;
 }
 
 /*
@@ -2298,8 +2303,12 @@ xfs_inactive_inodes(
 	struct xfs_mount	*mp,
 	struct xfs_eofblocks	*eofb)
 {
-	return xfs_ici_walk_fns(mp, &xfs_inactive_iwalk_ops, 0, eofb,
+	int			error;
+
+	error = xfs_ici_walk_fns(mp, &xfs_inactive_iwalk_ops, 0, eofb,
 			XFS_ICI_INACTIVE_TAG);
+	wake_up(&mp->m_inactive_wait);
+	return error;
 }
 
 /* Try to get inode inactivation moving. */
@@ -2406,3 +2415,42 @@ xfs_inactive_schedule_now(
 		spin_unlock(&pag->pag_ici_lock);
 	}
 }
+
+/* Return true if there are inodes still being inactivated. */
+static bool
+xfs_inactive_pending(
+	struct xfs_mount	*mp)
+{
+	struct xfs_perag	*pag;
+	xfs_agnumber_t		agno = 0;
+	bool			ret = false;
+
+	while (!ret &&
+	       (pag = xfs_perag_get_tag(mp, agno, XFS_ICI_INACTIVE_TAG))) {
+		agno = pag->pag_agno + 1;
+		spin_lock(&pag->pag_ici_lock);
+		if (pag->pag_ici_inactive)
+			ret = true;
+		spin_unlock(&pag->pag_ici_lock);
+		xfs_perag_put(pag);
+	}
+
+	return ret;
+}
+
+/*
+ * Flush all pending inactivation work and poll until finished.  This function
+ * is for callers that must flush with vfs locks held, such as unmount,
+ * remount, and iunlinks processing during mount.
+ */
+void
+xfs_inactive_force_poll(
+	struct xfs_mount	*mp)
+{
+	xfs_inactive_schedule_now(mp);
+
+	while (!wait_event_timeout(mp->m_inactive_wait,
+				   xfs_inactive_pending(mp) == false, HZ)) {
+		touch_softlockup_watchdog();
+	}
+}
diff --git a/fs/xfs/xfs_icache.h b/fs/xfs/xfs_icache.h
index a82b473b88a2..75332d4450ba 100644
--- a/fs/xfs/xfs_icache.h
+++ b/fs/xfs/xfs_icache.h
@@ -89,4 +89,6 @@ void xfs_inactive_shutdown(struct xfs_mount *mp);
 void xfs_inactive_cancel_work(struct xfs_mount *mp);
 void xfs_inactive_schedule_now(struct xfs_mount *mp);
 
+void xfs_inactive_force_poll(struct xfs_mount *mp);
+
 #endif
diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
index b9b37eff4063..5e2ce91f4ab8 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -1066,7 +1066,7 @@ xfs_unmountfs(
 	 * Since this can involve finobt updates, do it now before we lose the
 	 * per-AG space reservations.
 	 */
-	xfs_inactive_force(mp);
+	xfs_inactive_force_poll(mp);
 
 	xfs_blockgc_stop(mp);
 	xfs_fs_unreserve_ag_blocks(mp);
diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
index 87a62b0543ec..237a15a136c8 100644
--- a/fs/xfs/xfs_mount.h
+++ b/fs/xfs/xfs_mount.h
@@ -206,6 +206,12 @@ typedef struct xfs_mount {
 	 * into a single flush.
 	 */
 	struct work_struct	m_flush_inodes_work;
+
+	/*
+	 * Use this to wait for the inode inactivation workqueue to finish
+	 * inactivating all the inodes.
+	 */
+	struct wait_queue_head	m_inactive_wait;
 } xfs_mount_t;
 
 #define M_IGEO(mp)		(&(mp)->m_ino_geo)
diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index fced499ecdc9..af1fe32247cf 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -1736,7 +1736,7 @@ xfs_remount_ro(
 	 * Since this can involve finobt updates, do it now before we lose the
 	 * per-AG space reservations.
 	 */
-	xfs_inactive_force(mp);
+	xfs_inactive_force_poll(mp);
 
 	/* Free the per-AG metadata reservation pool. */
 	error = xfs_fs_unreserve_ag_blocks(mp);
@@ -1859,6 +1859,7 @@ static int xfs_init_fs_context(
 	INIT_WORK(&mp->m_flush_inodes_work, xfs_flush_inodes_worker);
 	INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker);
 	mp->m_kobj.kobject.kset = xfs_kset;
+	init_waitqueue_head(&mp->m_inactive_wait);
 	/*
 	 * We don't create the finobt per-ag space reservation until after log
 	 * recovery, so we must set this to true so that an ifree transaction