Re: [PATCH v7 1/2] hypervisor: XENMEM_claim_pages (subop of existing) hypercall

[Dan Magenheimer <dan.magenheimer@oracle.com>]

> From: Jan Beulich [mailto:JBeulich@suse.com]
> Subject: Re: [PATCH v7 1/2] hypervisor: XENMEM_claim_pages (subop of existing) hypercall
> 
> >>> On 26.11.12 at 19:11, Dan Magenheimer <dan.magenheimer@oracle.com> wrote:
> > +unsigned long domain_increase_tot_pages(struct domain *d, unsigned long pages)
> > +{
> > +    long dom_before, dom_after, dom_claimed, sys_before, sys_after;
> > +
> > +    ASSERT(spin_is_locked(&d->page_alloc_lock));
> > +    if ( !d->unclaimed_pages )
> > +        return d->tot_pages += pages;
> 
> So after the conditional adjustment above, I fail to see where
> d->tot_pages would get adjusted in the code below. Which gets
> me to ask ...

Yep.  Looks like I accidentally lost a line of code somewhere.
 
> > +    spin_lock(&heap_lock);
> > +    dom_before = d->unclaimed_pages;
> > +    dom_after = dom_before - pages;
> > +    if ( (dom_before > 0) && (dom_after < 0) )
> > +        dom_claimed = 0;
> > +    else
> > +        dom_claimed = dom_after;
> > +    sys_before = total_unclaimed_pages;
> > +    sys_after = sys_before - (dom_before - dom_claimed);
> > +    BUG_ON( (sys_before > 0) && (sys_after < 0) );
> > +    total_unclaimed_pages = sys_after;
> > +    d->unclaimed_pages = dom_claimed;
> > +    spin_unlock(&heap_lock);
> > +    return d->tot_pages;
> > +}
> 
> ...: Is there actually a strong reason why there needs to be an
> "increase" and a "decrease" version of this, rather than just
> having an "adjust" operation with a signed page count
> parameter?

Good point.  A good part of the logic seems different for the
two cases but the special cases in the increase case can
never occur in the decrease case and the additional checks
are harmless, so increase and decrease can be combined.

Spinning v8....

Thanks,
Dan