[OpenAFS-devel] Re: Change in openafs[master]: cmdebug -dcache

Simon Wilkinson sxw@inf.ed.ac.uk
Mon, 24 Jan 2011 20:17:25 +0000


On 24 Jan 2011, at 20:09, Harald Barth wrote:
> I would suggest to bind it in a way that it can be accessed by
> localhost only.

Having cmdebug output accessible outside of a particular machine can be =
of use for both debugging, and administration - it's making it available =
anonymously that is less than ideal, and we should probably consider =
changing that. However, until we have better RX security classes, there =
is no easy way for the cache manager to authenticate incoming =
connections.

> That would make it kindof compatible (because then the
> binary and the CM are of course of the same architcture)

You can't assume that. We have platforms where you can have a 64bit =
userland, and a 32 bit kernel space, and vice versa.

I'm in favour of Jeff's proposal that we stop treating cmdebug as a =
"standard" interface. However, that means that we need to move the RPCs =
used by cmdebug out of the callback service, and into their own RX =
service. =46rom memory, I think this has the added advantage that it =
should be possible to build a cmdebug that returns results, even when =
the callback service thread is blocked.