[PATCH] ksmbd: fix use-after-free of lease table in smb2_lease_break_noti()

Abdifatah Suruur <[email protected]>
Newsgroups org.kernel.vger.linux-cifs,org.kernel.vger.linux-kernel
Message-ID <[email protected]>
smb2_lease_break_noti() redirects v2 lease break notifications to the
client lease channel by reading lease->l_lb->conn.  The lease table is
linked into the global lease_table_list and destroyed by
destroy_lease_table() on connection teardown, which unlinks it and
kfree()s it under write_lock(&lease_list_lock).

The break path dereferences lease->l_lb and lease->l_lb->conn without
holding lease_list_lock, so when the owning connection is torn down
concurrently (e.g. the client disconnects while another connection's
open triggers a lease break), the table can be freed between the load
and the dereference.  This is a use-after-free read of struct
lease_table; the stale conn pointer is then passed to
ksmbd_conn_releasing() and ksmbd_conn_get(), corrupting a refcount on
memory that may have been reallocated.

Take a conn reference and perform the lease-table lookup under
read_lock(&lease_list_lock), matching the lock discipline of
find_same_lease_key() and lookup_lease_in_table(), and transfer the
reference to the async work item instead of re-taking it.

Fixes: 2145945feb2c2 ("ksmbd: route v2 lease breaks on the client lease channel")
Cc: [email protected]
Signed-off-by: Abdifatah Suruur <[email protected]>

---
 fs/smb/server/oplock.c | 16 +++++++++-------
 1 file changed, 9 insertions(+), 7 deletions(-)

diff --git a/fs/smb/server/oplock.c b/fs/smb/server/oplock.c
index 79787099afdc6..ac21583cb6222 100644
--- a/fs/smb/server/oplock.c
+++ b/fs/smb/server/oplock.c
@@ -1007,20 +1007,27 @@ static int smb2_lease_break_noti(struct oplock_info *opinfo, bool wait_ack,
 	struct lease *lease = opinfo->o_lease;
 	int ret = 0;

-	conn = READ_ONCE(opinfo->conn);
+	conn = ksmbd_conn_get(READ_ONCE(opinfo->conn));
+	read_lock(&lease_list_lock);
 	if (lease->version == 2 && lease->l_lb && lease->l_lb->conn &&
-	    !ksmbd_conn_releasing(lease->l_lb->conn))
-		conn = lease->l_lb->conn;
+	    !ksmbd_conn_releasing(lease->l_lb->conn)) {
+		ksmbd_conn_put(conn);
+		conn = ksmbd_conn_get(lease->l_lb->conn);
+	}
+	read_unlock(&lease_list_lock);
 	if (!conn)
 		return ksmbd_invalidate_durable_fd(opinfo->fid);

 	work = ksmbd_alloc_work_struct();
-	if (!work)
+	if (!work) {
+		ksmbd_conn_put(conn);
 		return -ENOMEM;
+	}

 	br_info = kmalloc_obj(struct lease_break_info, KSMBD_DEFAULT_GFP);
 	if (!br_info) {
 		ksmbd_free_work_struct(work);
+		ksmbd_conn_put(conn);
 		return -ENOMEM;
 	}

@@ -1036,7 +1043,7 @@ static int smb2_lease_break_noti(struct oplock_info *opinfo, bool wait_ack,
 	memcpy(br_info->lease_key, lease->lease_key, SMB2_LEASE_KEY_SIZE);

 	work->request_buf = (char *)br_info;
-	work->conn = ksmbd_conn_get(conn);
+	work->conn = conn;
 	work->sess = opinfo->sess;

 	ksmbd_conn_r_count_inc(conn);
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.