[OpenAFS] FIX: broken hyperlinks to MS docs on AFS

Chaskiel M Grundman cg2v@andrew.cmu.edu
Mon, 08 Sep 2003 13:05:44 -0400

--On Monday, September 08, 2003 08:43:49 -0700 Noel Burton-Krahn
<noel@bkbox.com> wrote:

> I'm running a MS Loopback Adapter and OpenAFS client 1.2.10. 
Does your loopback adapter have a static IP address? if not, it will have
to time out DHCP before being given an autonet (169.254.x.x.) address. the
"Loopback adapter" is _not_ where is implemented. is
not associated with any interface, it's hardcoded into the tcpip stack.

> Jeffrey, I've been meaning to redo the OpenAFS install as an MSI with the
> loopback adapter and some other bug fixes like this one.  Do you know
> anyone else going down that path?

I currently have an MSI installer that CMU is using internally.
Constructing the MSI is currently not automated, since I'm using Visual
Studio to produce a base MSI, and then patching it using MSITran (from the
platform SDK) to add functionality that Visual Studio is not yet capable of
dealing with. (win32 service setup, setting up the afscreds start-menu item
the way I want it to. something else I'm forgetting). Additionally, the
patching process is fragile. Sometimes the VS generated file will be too
different from a previous run and the patch will fail to apply, which will
require that the changes be redone from scratch.

My current loopback-management code is very different from anyone else's.
1) identifies the loopback adapter to bind to by putting the interface's
GUID into the TransarcAfsDaemon services' Parameter key, rather than trying
MAC matching, which turns out not to work in all cases.
2) does the adapter install/uninstall from within afsd_service. I ended up
doing this because nothing else that I tried worked when my MSI was
installed non-interactively on domain machines using GPO. The hideous code
that does this can be found at
http://www.contrib.andrew.cmu.edu/~cg2v/afsd_loop.c if anyone wants to look
at it, but I do not recommend that anyone use it.