RE: grant table anxiety [was RE: [Xen-devel] GNTTABOP_unmap_grant_ref]

["King, Steven R" <steven.r.king@xxxxxxxxx>]

Per Mark's pointers to LWN articles, Linux's page remapping scheme is
churning.  Once that settles, is a reasonable test of "sensibleness"
that paravirtual Linux be able to map shared pages in the same manner as
traditional Linux maps device based memory?

-----Original Message-----
From: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
[mailto:xen-devel-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of Ian Pratt
Sent: Wednesday, January 25, 2006 2:37 PM
To: King, Steven R; Keir Fraser; Mark Williamson; Rusty Russell
Cc: xen-devel
Subject: RE: grant table anxiety [was RE: [Xen-devel]
GNTTABOP_unmap_grant_ref]

> I've aired grievances against grant tables in other threads, but now 
> I'm getting worried all over again.  Your suggestions give a hint of 
> the intricate handling required to manage shared pages.  IMHO, this 
> pushes paravirtual guests to unnatural lengths, not to mention 
> prospects in pure-virtual environments.

Allowing user space to get access to shared pages by allowing them to
map pages in a special device seems pretty sensible, and is pretty
trivial to implement in the kernel.

> Could we dispense with these complexities if a 3rd party
> (Xen?) owned shared pages?  

No, the accounting issue is totally orthogonal. 

Ian

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel