s3: libcli: smb1: Ensure we correctly finish a tevent req if the writev fails in...

We haven't added the req to the pending array yet, as SMB1 requests can
be one-way (no reply needed). So if we error out after the writev, but
before we add to the pending array we must rember to terminate our current
tevent req with a tevent_req_nterror call.

Fixes bug 11173 - SMB1 Server disconnect can cause timeout on client write error.

https://bugzilla.samba.org/show_bug.cgi?id=11173

Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Michael Adam <obnox@samba.org>
Autobuild-User(master): Michael Adam <obnox@samba.org>
Autobuild-Date(master): Fri Mar 20 05:08:25 CET 2015 on sn-devel-104

(cherry picked from commit 00d92f59257ce8fe1729becd72e13a42dfff0589)

diff --git a/libcli/smb/smbXcli_base.c b/libcli/smb/smbXcli_base.c
index 2fe6a5a7a92187ae5033c29f47fb69a3adb2307a..742caf1e620dc37ec1f1b0de1c6f8212e3a98656 100644 (file)
--- a/libcli/smb/smbXcli_base.c
+++ b/libcli/smb/smbXcli_base.c
@@ -1592,6 +1592,7 @@ static void smb1cli_req_writev_done(struct tevent_req *subreq)
        if (nwritten == -1) {
                NTSTATUS status = map_nt_error_from_unix_common(err);
                smbXcli_conn_disconnect(state->conn, status);
+               tevent_req_nterror(req, status);
                return;
        }