As I experienced invalid header length e.g. worst-case zero triggers an
endless loop inside the receive handling, because we don't increment any
consumed bytes. This patch will add also a check for the minimum size of
a dlm message inside the dlm header length field.

Signed-off-by: Alexander Aring <[email protected]>
---
 fs/dlm/midcomms.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/fs/dlm/midcomms.c b/fs/dlm/midcomms.c
index b146842be54a..139d87f1ba17 100644
--- a/fs/dlm/midcomms.c
+++ b/fs/dlm/midcomms.c
@@ -95,7 +95,8 @@ int dlm_process_incoming_buffer(int nodeid, unsigned char 
*buf, int len)
                 * cannot deliver this message to upper layers
                 */
                msglen = le16_to_cpu(hd->h_length);
-               if (msglen > DEFAULT_BUFFER_SIZE) {
+               if (msglen > DEFAULT_BUFFER_SIZE ||
+                   msglen < sizeof(struct dlm_header)) {
                        log_print("received invalid length header: %u, will 
abort message parsing",
                                  msglen);
                        return -EBADMSG;
-- 
2.26.2

Reply via email to