Xen devel - Nov 2006 - Why "Don''t use shadow linear maps in sh_set_toplevel_shadow"?

It''s changeset 12399, but can not understand, can you pls explain?
With this change, kernel build of FC5 changed from around 370s to 420s,
over 10% gap.
Thanks
-Xin

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

Is it harmful if we use the heuristics? Can we fallback to search the
hash?
-Xin
>It''s changeset 12399, but can not understand, can you pls explain?
>With this change, kernel build of FC5 changed from around 370s 
>to 420s, over 10% gap.
>Thanks
>-Xin
>
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

At 23:13 +0800 on 17 Nov (1163805188), Li, Xin B wrote:
> It''s changeset 12399, but can not understand, can you pls explain?
Because we can''t use the shadow linear maps before we build them. 
The top-level shadow entry is explicitly null during that routine, and
the writeable-mappings heuristics use the shadow linear map, so it''s
not safe to use them there. 

Cheers,

Tim.

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

>At 23:13 +0800 on 17 Nov (1163805188), Li, Xin B wrote:
>> It''s changeset 12399, but can not understand, can you pls
explain?
>
>Because we can''t use the shadow linear maps before we build them. 
>The top-level shadow entry is explicitly null during that routine, and
>the writeable-mappings heuristics use the shadow linear map, so
it''s
>not safe to use them there. 
>
Not understand.
writing a new cr3 means a context switch, guest kernel address space
should be still accessible during the switching, so it seems OK to me to
use the heuristics.
And what do you mean by "safe"?
thanks
-Xin

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