ok,
ill check it in in a few hours.
by the way, do you have access to ndmp captures you can share and upload to the wiki page?
we do not have any public sample captures of NDMP yet.
On 6/30/06, Aaron Christensen <aaronmf@xxxxxxxxx> wrote:
Hi, Ronnie,
I initally checked this against the specification for NDMPv2 and v4 to make sure. In going back, I double checked with v2, and then check against v3, v4, and v5. It looks like the error field has been in there all along for NDMP_CONFIG_GET_AUTH_ATTR. I admit, I only have an implementation for NDMPv4, and since the specification for this message is unchanged it seems to me that other versions should be the same implementation-wise.
Regards,
~Aaron
On 6/29/06, ronnie sahlberg <
ronniesahlberg@xxxxxxxxx> wrote:
Can you check ndmp versions 2 to 5 that this is true for all versions of ndmp?
if not you have to add a switch statement to manage the difference between versions.
Added the dissection of a field that wasn't being picked up during NDMP MD5 authentication.
The NDMP_CONFIG_GET_AUTH_ATTR message response was using the function dissect_auth_attr_msg() which dissects an auth_attr struct. However, the specification includes and error field previous to that struct. Since that function was being called directly, the error field was being picked up as part of the auth_attr, and then leaving some extra data at the end.
I added a new function, dissect_ndmp_config_get_auth_attr_reply(), which first calls dissect_error() and then dissect_auth_attr_msg(). I didn't change dissect_auth_attr_msg() because the NDMP_CONNECT_SERVER_AUTH request uses it directly. After testing it, it appears to be picking things up correctly, now.
Regards,
~Aaron Christensen