[PATCH] Fix bug 9722 - Samba does not properly handle Oplock breaks in compound requests

[Richard Sharpe]

On Fri, May 3, 2013 at 11:19 AM, Jeremy Allison <jra at samba.org> wrote:
> Here is a bugfix for master for Bug 9722, discovered by Richard.
>
> It includes a torture test (written by Richard) that
> reproduces the Windows client test case that discovered
> the bug.
>
> Passes all testing here (I've also tested with SMB3
> encryption turned on :-).
>
> Please review and push if you approve.

Is there a problem here?

diff --git a/source3/smbd/smb2_server.c b/source3/smbd/smb2_server.c
index 57e9c7b..9a55d6a 100644
--- a/source3/smbd/smb2_server.c
+++ b/source3/smbd/smb2_server.c
@@ -1599,6 +1599,14 @@ static NTSTATUS smbd_smb2_request_process_cancel(struct s
mbd_smb2_request *req)
 		uint64_t message_id;
 		uint64_t async_id;

+		if (cur->compound_related) {
+			/*
+			 * Never cancel anything in a compound request.
+			 * Way too hard to deal with the result.
+			 */
+			continue;
+		}
+
 		outhdr = SMBD_SMB2_OUT_HDR_PTR(cur);

 		message_id = BVAL(outhdr, SMB2_HDR_MESSAGE_ID);
-- 
1.8.1.2

because just before the for loop that this goes in, we do this:

        /*
         * we don't need the request anymore
         * cancel requests never have a response
         */
        DLIST_REMOVE(req->sconn->smb2.requests, req);
        TALLOC_FREE(req);

which seems to suggest that the request is now gone from our view.

-- 
Regards,
Richard Sharpe
(何以解憂？唯有杜康。--曹操)