Getting Samba 3.6.12 and possibly 4.0.x to properly handle an OpLock Break in a compound request

[Richard Sharpe]

Hi folks,

The following ugly patch seems to allow Samba 3.6.12 to handle an
OpLock break in a compound request.

Next up I have to see if not responding with a STATUS_PENDING response matters.

-bash-4.0$ diff -up
build/cloudcc/build_x86_64/samba-3.6.12/source3/smbd/smb2_server.c
smb2_server.c
--- build/cloudcc/build_x86_64/samba-3.6.12/source3/smbd/smb2_server.c
 2013-03-13 21:18:24.839321028 -0700
+++ smb2_server.c       2013-03-13 21:14:09.827919858 -0700
@@ -930,6 +930,7 @@ NTSTATUS smbd_smb2_request_pending_queue
        struct iovec *outvec = NULL;

        if (!tevent_req_is_in_progress(subreq)) {
+               DEBUG(10, ("Already in progress\n"));
                return NT_STATUS_OK;
        }

@@ -937,10 +938,17 @@ NTSTATUS smbd_smb2_request_pending_queue
        subreq = NULL;

        if (req->async) {
+               DEBUG(10, ("Already async"));
                /* We're already async. */
                return NT_STATUS_OK;
        }

+       if (req->smb1req && open_was_deferred_smb2(req->smb1req->sconn,
+                                               req->smb1req->mid)) {
+               DEBUG(10, ("Open Request was deferred ... doing nothing\n"));
+               return NT_STATUS_OK;
+       }
+
        if (req->in.vector_count > i + 3) {
                /*
                 * We're trying to go async in a compound


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