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

RE: [Xen-devel] Re: user/hypervisor address space solution

On Fri, 2006-02-17 at 17:05 +0800, Tian, Kevin wrote:
> >From: Hollis Blanchard
> >Sent: 2006å2æ14æ 19:13
> >
> >I don't know what implications this could have for ia64, since they
> suck
> >in so much Linux code... but perhaps that Linux code isn't the stuff
> >that deals with userland, so might not have uaccess users.
> Regarding this area to access domain parameters, IA64 should be OK
> since xen/ia64 shares same virtual address space (i.e same ASN) with
> domain, thus able to access pointer passed in directly. Linux style
> uaccess interfaces can easily work for xen/ia64 in this aspect.

Actually I was referring to the API rename, not the underlying
implementation. But since then I've changed my mind, and I'm fine with
leaving e.g. "put_user" around, as long as nobody calls it from common
code. So if there are put_user() callers in the Linux code you suck in,
it will still be ok.

But while we're on the subject, I recall hearing that IA64 has an issue
with TLB faults when accessing domain virtual addresses. Could you
elaborate on that? I guess you're using a software-controlled TLB, and
when a fault occurs you need to call up to Linux to handle it?

Hollis Blanchard
IBM Linux Technology Center

Xen-devel mailing list



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