[PATCH review for 4.9 35/50] ceph: clean up unsafe d_parent accesses in build_dentry_path

From: Levin, Alexander (Sasha Levin)
Date: Sun Sep 24 2017 - 21:27:02 EST


From: Jeff Layton <jlayton@xxxxxxxxxx>

[ Upstream commit c6b0b656ca24ede6657abb4a2cd910fa9c1879ba ]

While we hold a reference to the dentry when build_dentry_path is
called, we could end up racing with a rename that changes d_parent.
Handle that situation correctly, by using the rcu_read_lock to
ensure that the parent dentry and inode stick around long enough
to safely check ceph_snap and ceph_ino.

Link: http://tracker.ceph.com/issues/18148
Signed-off-by: Jeff Layton <jlayton@xxxxxxxxxx>
Reviewed-by: Yan, Zheng <zyan@xxxxxxxxxx>
Signed-off-by: Ilya Dryomov <idryomov@xxxxxxxxx>
Signed-off-by: Sasha Levin <alexander.levin@xxxxxxxxxxx>
---
fs/ceph/mds_client.c | 9 +++++++--
1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index e3e1a80b351e..c0f52c443c34 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -1782,13 +1782,18 @@ static int build_dentry_path(struct dentry *dentry,
int *pfreepath)
{
char *path;
+ struct inode *dir;

- if (ceph_snap(d_inode(dentry->d_parent)) == CEPH_NOSNAP) {
- *pino = ceph_ino(d_inode(dentry->d_parent));
+ rcu_read_lock();
+ dir = d_inode_rcu(dentry->d_parent);
+ if (dir && ceph_snap(dir) == CEPH_NOSNAP) {
+ *pino = ceph_ino(dir);
+ rcu_read_unlock();
*ppath = dentry->d_name.name;
*ppathlen = dentry->d_name.len;
return 0;
}
+ rcu_read_unlock();
path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
if (IS_ERR(path))
return PTR_ERR(path);
--
2.11.0