[v2] ceph: fix use-after-free bug for inodes when flushing capsnaps

Commit Message

Xiubo Li June 1, 2023, 1:24 a.m. UTC

From: Xiubo Li <[e-mail]>

There is racy between capsnaps flush and removing the inode from
'mdsc->snap_flush_list' list:

   == Thread A ==                     == Thread B ==
ceph_queue_cap_snap()
 -> allocate 'capsnapA'
 ->ihold('&ci->vfs_inode')
 ->add 'capsnapA' to 'ci->i_cap_snaps'
 ->add 'ci' to 'mdsc->snap_flush_list'
    ...
   == Thread C ==
ceph_flush_snaps()
 ->__ceph_flush_snaps()
  ->__send_flush_snap()
                                handle_cap_flushsnap_ack()
                                 ->iput('&ci->vfs_inode')
                                   this also will release 'ci'
                                    ...
				      == Thread D ==
                                ceph_handle_snap()
                                 ->flush_snaps()
                                  ->iterate 'mdsc->snap_flush_list'
                                   ->get the stale 'ci'
 ->remove 'ci' from                ->ihold(&ci->vfs_inode) this
   'mdsc->snap_flush_list'           will WARNING

To fix this we will increase the ihold the inode when adding 'ci'
to the 'mdsc->snap_flush_list' list.

Cc: [e-mail]
URL: https://bugzilla.redhat.com/show_bug.cgi?id=2209299
Signed-off-by: Xiubo Li <[e-mail]>
---

V2:
- Increase the i_count reference instead to fix this issue


 fs/ceph/caps.c | 6 ++++++
 fs/ceph/snap.c | 4 +++-
 2 files changed, 9 insertions(+), 1 deletion(-)

Patch

diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c
index feabf4cc0c4f..7c2cb813aba4 100644
--- a/fs/ceph/caps.c
+++ b/fs/ceph/caps.c
@@ -1684,6 +1684,7 @@  void ceph_flush_snaps(struct ceph_inode_info *ci,
 	struct inode *inode = &ci->netfs.inode;
 	struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
 	struct ceph_mds_session *session = NULL;
+	int put = 0;
 	int mds;
 
 	dout("ceph_flush_snaps %p\n", inode);
@@ -1728,8 +1729,13 @@  void ceph_flush_snaps(struct ceph_inode_info *ci,
 		ceph_put_mds_session(session);
 	/* we flushed them all; remove this inode from the queue */
 	spin_lock(&mdsc->snap_flush_lock);
+	if (!list_empty(&ci->i_snap_flush_item))
+		put++;
 	list_del_init(&ci->i_snap_flush_item);
 	spin_unlock(&mdsc->snap_flush_lock);
+
+	if (put)
+		iput(inode);
 }
 
 /*
diff --git a/fs/ceph/snap.c b/fs/ceph/snap.c
index 5a4bf0201737..d5ad10d94424 100644
--- a/fs/ceph/snap.c
+++ b/fs/ceph/snap.c
@@ -697,8 +697,10 @@  int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
 	     capsnap->size);
 
 	spin_lock(&mdsc->snap_flush_lock);
-	if (list_empty(&ci->i_snap_flush_item))
+	if (list_empty(&ci->i_snap_flush_item)) {
+		ihold(inode);
 		list_add_tail(&ci->i_snap_flush_item, &mdsc->snap_flush_list);
+	}
 	spin_unlock(&mdsc->snap_flush_lock);
 	return 1;  /* caller may want to ceph_flush_snaps */
 }