Re: [Xen-devel] [RFC/PATCH] XENMEM_claim_pages (subop of an existing) hypercall

[Dan Magenheimer <dan.magenheimer@xxxxxxxxxx>]

> From: Jan Beulich [mailto:JBeulich@xxxxxxxx]
> Subject: Re: [RFC/PATCH] XENMEM_claim_pages (subop of an existing) hypercall

Hi Jan --

Excellent feedback, thanks.  You probably saved me hours
of debugging... always nice to have another pair of eyes
to catch stupid mistakes that I could start at for a long
time without noticing the insanity. :-)

I'll start working on v2 incorporating your suggestions
(and get some runtime testing completed).

One thing:

> Finally, assuming we really want/need this, shouldn't there be
> a way for claims to expire (e.g. in case of tool stack problems),
> so they don't indefinitely prevent (perhaps large) portions of
> memory to be used for other purpose (e.g. when a domain can't
> get fully cleaned up)?

I was envisioning (and thanks for catching that I had neglected
it in the patch) that the reset_claim call would be done very
early in the domain death sequence.

But maybe that's a good reason to add back in the "manual"
unclaim hypercall/subop, so a toolstack can force the
claim to zero.  I realized I can easily add that by
allowing an existing claim to be "overridden" by a
claim of zero pages, so another subop isn't even needed.

Dan

.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel