xfs: create a polled function to force inode inactivationdeferred-inactivation_2019-08-16

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>

5 files changed, 53 insertions, 2 deletions

diff --git a/fs/xfs/xfs_icache.c b/fs/xfs/xfs_icache.c
index 13e7b806d9a1..168c3c71887c 100644
--- a/fs/xfs/xfs_icache.c
+++ b/fs/xfs/xfs_icache.c
@@ -24,6 +24,7 @@
 #include "xfs_reflink.h"
 
 #include <linux/iversion.h>
+#include <linux/nmi.h>
 
 static void xfs_perag_set_inactive_tag(struct xfs_perag *pag);
 static void xfs_perag_clear_inactive_tag(struct xfs_perag *pag);
@@ -2139,6 +2140,8 @@ xfs_inactive_inodes_pag(
 	if (inctx.kick_reclaim)
 		xfs_reclaim_work_queue(pag->pag_mount);
 
+	wake_up(&pag->pag_mount->m_inactive_wait);
+
 	return error;
 }
 
@@ -2162,6 +2165,8 @@ xfs_inactive_inodes(
 	if (inctx.kick_reclaim)
 		xfs_reclaim_work_queue(mp);
 
+	wake_up(&mp->m_inactive_wait);
+
 	return error;
 }
 
@@ -2273,3 +2278,40 @@ xfs_inactive_schedule_work(
 		xfs_perag_put(pag);
 	}
 }
+
+/* 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		ag = 0;
+	bool			ret = false;
+
+	while (!ret && (pag = xfs_perag_get_tag(mp, ag, XFS_ICI_RECLAIM_TAG))) {
+		ag = 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_work(mp, 0);
+	while (wait_event_timeout(mp->m_inactive_wait,
+			xfs_inactive_pending(mp) == false, HZ / 10) == 0) {
+		touch_softlockup_watchdog();
+	}
+}
diff --git a/fs/xfs/xfs_icache.h b/fs/xfs/xfs_icache.h
index f0a18238bc86..77422b2be148 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_work(struct xfs_mount *mp, unsigned long delay);
 
+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 15cab53341bc..f962c6d753fc 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -1036,7 +1036,7 @@ 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_stop_block_reaping(mp);
 	xfs_fs_unreserve_ag_blocks(mp);
diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
index cd8d4b9ca551..2a5f3f903802 100644
--- a/fs/xfs/xfs_mount.h
+++ b/fs/xfs/xfs_mount.h
@@ -221,6 +221,12 @@ 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 915a7bfa3777..63d2b4056920 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -1401,7 +1401,7 @@ xfs_fs_remount(
 		 * inodes.  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);
@@ -1642,6 +1642,7 @@ xfs_mount_alloc(
 	INIT_DELAYED_WORK(&mp->m_eofblocks_work, xfs_eofblocks_worker);
 	INIT_DELAYED_WORK(&mp->m_cowblocks_work, xfs_cowblocks_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