[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH] This patch fixes several issues related to vmxassist

On 30 Mar 2006, at 18:02, Li, Xin B wrote:

Shouldn't we get the Xen portion of vmxassist help us with that, for
example by making the hidden descriptor info (base, limit, etc)
available to us? There's already a method for loading that
stuff out of
Xen, right?

Looks to me as though the kludge won't work if you unluckily load a
real-mode segment value that happens to also reference a 'big segment'
in the currently registered GDT.

Yes, we may have potential bug here, maybe we should hold this patch and try to find a cleaner way.

In fact, the existing implementation of address() is kludgy. It already does tests on the selector value to decide whether it is likely to refer to a protected-mode or real-mode segment. Unfortunately the test may sometimes yield false positives (selectors that look like they could be a valid protected-mode value, but actually it's some arbitrary real-mode selector).

I don't know the heritage of that code. I expect someone decided it was good enough to be getting on with but maybe now it is time to revisit and see if we can implement a watertight version which correctly uses hidden segment descriptor state which is readily available when running on VMX.

It might be worth pinging Leendert about this and see what he thinks.

 -- Keir

Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.