Subject: Re: CVS commit: src/sys/arch/xen
To: None <>
From: YAMAMOTO Takashi <>
List: source-changes
Date: 08/22/2005 21:33:08
> > - there is no recovery in the case of failure of MEMOP_increase_reservation.
> >   is this intended?
> There's no recovery needed. We lost these pages, they are marked as
> allocated in UVM so that's OK. We could eventually reallocate the pages one by
> one, but I'm not sure it's worth it as it's not something that should
> happen anyway.

assuming that xen's buddy allocator has the same characteristics as
linux's, "high-order" allocation failure is something that likely
happens on long run.

> >   i think it's better to do MEMOP_decrease_reservation
> >   only when MEMOP_increase_reservation succeeded.
> >   MEMOP_decrease_reservation unlikely helps MEMOP_increase_reservation
> >   anyway.
> MEMOP_increase_reservation won't work if we didn't give enouth page
> back first, because the hypervisor won't allow a domain to increase
> its number of pages past the configured limit (and a domain starts with
> this number of pages allocated).

ok, i overlooked the limit.  thanks.