s3: smbd: Ensure check_fsp_ntquota_handle() doesn't send SMB1 error packets.

check_fsp_ntquota_handle() is called from SMB2 codepaths as
well as from SMB1. Even in the SMB1 cases the callers of
check_fsp_ntquota_handle() handle sendng the error packet when
check_fsp_ntquota_handle returns false so on a 'return false'
we'd end up sending an error packet twice.

The SMB2 callers of check_fsp_ntquota_handle()
already check that fsp is valid, so there's
no danger of us sending an SMB1 error packet
over the SMB2 stream (so I'm not classing
this as a bug to be back-ported).

Fix check_fsp_ntquota_handle() by inlineing
the check_fsp_open() functionality without
the reply_nterror() calls.

This will allow the next commit to move check_fsp_open()
with the implicit reply_nterror() and also check_fsp()
(which calls check_fsp_open()) into the SMB1 smb1_reply.c
file as SMB1-only code.

Signed-off-by: Jeremy Allison <jra@samba.org>
Reviewed-by: Volker Lendecke <vl@samba.org>

diff --git a/source3/smbd/smb2_reply.c b/source3/smbd/smb2_reply.c
index 90aa92193b954cf9d28984a9699ef05f533e41a1..8b2eca20cb1b5b8c5b2c0c2de839dfc97f909d39 100644 (file)
--- a/source3/smbd/smb2_reply.c
+++ b/source3/smbd/smb2_reply.c
@@ -572,7 +572,11 @@ bool check_fsp(connection_struct *conn, struct smb_request *req,
 bool check_fsp_ntquota_handle(connection_struct *conn, struct smb_request *req,
                              files_struct *fsp)
 {
-       if (!check_fsp_open(conn, req, fsp)) {
+       if ((fsp == NULL) || (conn == NULL)) {
+               return false;
+       }
+
+       if ((conn != fsp->conn) || (req->vuid != fsp->vuid)) {
                return false;
        }