RE: [Xen-users] Debian Sid amd64 libc6-xen

["Petersson, Mats" <Mats.Petersson@xxxxxxx>]

> -----Original Message-----
> From: xen-users-bounces@xxxxxxxxxxxxxxxxxxx 
> [mailto:xen-users-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of 
> Konstantinos Pachopoulos
> Sent: 31 August 2006 11:59
> To: xen-users@xxxxxxxxxxxxxxxxxxx
> Subject: [Xen-users] Debian Sid amd64 libc6-xen
> 
> Hi,
> libc6-xen exists only for i386 Sid, not for amd64. Is
> this not necessary for the system- hypervisor?

64-bit doesn't need special treatment to deal with libc, since it's not
using the same method as i386 for protecting the Xen kernel memory from
Linux, and thus it's fine to use negative segment offsets to access
thread-local storage without having any penalty.

In 32-bit, segment limis are used to prevent the Linux kernel from
accidentally (or indeed intentionally) overwriting the hypervisor. Since
there are quite a few processors available that doesn't support segment
limits in 64-bit (only recent AMD processors DO support this, AFAIK),
it's not usable to prevent such overwriting, and thus another method is
used [I think it involves using multiple page-tables, but I've not
looked at it]. 

And why is segment limits and negative segment offsets incompatible?
Well, it causes problems because the negative offset is actually just a
big unsigned number that happens to overflow when it's added to the base
from the segment. But with a limit that isn't 4GB, the overflow
calculation doesn't work right (as in, it traps and the fault handler
gets to work out what actually was meant to happen - which is MUCH
slower...). 

--
Mats
> 
> Thanks
> 
> 
>               
> ___________________________________________________________ 
> Try the all-new Yahoo! Mail. "The New Version is radically 
> easier to use" - The Wall Street Journal 
> http://uk.docs.yahoo.com/nowyoucan.html
> 
> _______________________________________________
> Xen-users mailing list
> Xen-users@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-users
> 
> 
> 



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