From: "Yan, Zheng" <zyan@redhat.com>
Date: Wed, 6 Sep 2017 10:15:16 +0800
Subject: ceph: stop on-going cached readdir if mds revokes FILE_SHARED cap
Git-commit: 15b51bd6badbb373c723aa019cf530c8263efd7e
Patch-mainline: v4.14-rc1
References: FATE#324714

If directory's FILE_SHARED cap get revoked, dentry in the directory
can get spliced into other directory (Eg, other client move the
dentry into directory B, then we do readdir on directory B). So we
should stop on-going cached readdir. this can be achieved by marking
dir not complete, because __dcache_readdir() checks dir completeness
before emitting each dentry.

Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Acked-by: Luis Henriques <lhenriques@suse.com>
---
 fs/ceph/caps.c | 13 +++++++------
 1 file changed, 7 insertions(+), 6 deletions(-)

diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c
index 7a7945032802..157fe59fbabe 100644
--- a/fs/ceph/caps.c
+++ b/fs/ceph/caps.c
@@ -490,13 +490,14 @@ static void __check_cap_issue(struct ceph_inode_info *ci, struct ceph_cap *cap,
 	}
 
 	/*
-	 * if we are newly issued FILE_SHARED, mark dir not complete; we
-	 * don't know what happened to this directory while we didn't
-	 * have the cap.
+	 * If FILE_SHARED is newly issued, mark dir not complete. We don't
+	 * know what happened to this directory while we didn't have the cap.
+	 * If FILE_SHARED is being revoked, also mark dir not complete. It
+	 * stops on-going cached readdir.
 	 */
-	if ((issued & CEPH_CAP_FILE_SHARED) &&
-	    (had & CEPH_CAP_FILE_SHARED) == 0) {
-		ci->i_shared_gen++;
+	if ((issued & CEPH_CAP_FILE_SHARED) != (had & CEPH_CAP_FILE_SHARED)) {
+		if (issued & CEPH_CAP_FILE_SHARED)
+			ci->i_shared_gen++;
 		if (S_ISDIR(ci->vfs_inode.i_mode)) {
 			dout(" marking %p NOT complete\n", &ci->vfs_inode);
 			__ceph_dir_clear_complete(ci);