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

Re: [Xen-devel] RFC: [0/2] Remove netloop by lazy copying in netback

On Tue, Mar 27, 2007 at 01:33:58AM +0100, Keir Fraser wrote:
> >> Can't we wrap the 'swap around' critical section in an irq-safe spinlock?
> >> All we'd need to do from the page-fault handler then is a barrier on that
> >> spinlock (i.e, wait for it to be released). Netback can simply copy the 
> >> page
> >> to new memory frame, unmap the grant, then relocate the new memory frame's
> >> pseudophysical address.
> > 
> > That works fine for the x86 case.  But when it's auto-translated,
> > you won't even get a page fault in the guest because the guest PTE
> > is unchanged and completely valid.
> How about you invalidate the PTE for the duration of the critical section.
> It's a bit skanky, but would work around this issue quite nicely!

Maybe I am missing something.  Because I thought we've agreed that
we can't (or don't want to) do the PTE trick because of the large
pages (16MB at least) used for the identity mapping.

Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <herbert@xxxxxxxxxxxxxxxxxxx>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Xen-devel mailing list



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