Xen devel - Mar 2011 - Questions for the shadow page code

Hi, guys
Lately I''ve been researching the implementation of the shadow page and
I didn''t really understand the mechanism. For instance, the function
shadow_get_and_create_l*e in x86/mm/shadow/multi.c
If the GuestOS is 32-bit, then shadow_get_and_create_l1e should execute
{
    *sl2mfn = pagetable_get_mfn(v->arch.shadow_table[0]);


    (void) shadow_l2_index(sl2mfn, guest_l2_table_offset(gw->va));


    return sh_linear_l2_table(v) + shadow_l2_linear_offset(gw->va);
}
1.How does the Xen organize the spt? it looks like xen used 36-bit pae for spt,
but how does it correspond to the guest 2 level page table?
2.What''s the function of sh_linear_l2_table & the whole return
sentence? It''s the same situation in the shadow_get_and_create_l2e. 
There are some assemble code and it screw me up :(
3.Are there any docs to help me to figure it out?




Regards,
Yao

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

Hi, 

At 02:35 +0000 on 11 Mar (1299810912), Yao wrote:
> If the GuestOS is 32-bit, then shadow_get_and_create_l1e should execute
> {
>     *sl2mfn = pagetable_get_mfn(v->arch.shadow_table[0]);
> 
>     (void) shadow_l2_index(sl2mfn, guest_l2_table_offset(gw->va));
> 
>     return sh_linear_l2_table(v) + shadow_l2_linear_offset(gw->va);
> }
> 1.How does the Xen organize the spt? it looks like xen used 36-bit pae
> for spt, but how does it correspond to the guest 2 level page table?
Yes.  The shadow page tables _must_ be PAE or 64-bit in order to be able
to use machine memory >4GiB for VMs.  From the comments in shadow/common.c:

 * A 32-bit guest l1 table covers 4MB
 * of virtual address space, and needs to be shadowed by two PAE/64-bit
 * l1 tables (covering 2MB of virtual address space each).  Similarly, a
 * 32-bit guest l2 table (4GB va) needs to be shadowed by four
 * PAE/64-bit l2 tables (1GB va each).  These multi-page shadows are
 * not contiguous in memory; functions for handling offsets into them are
 * defined in shadow/multi.c (shadow_l1_index() etc.)
> 2.What''s the function of sh_linear_l2_table & the whole return
> sentence? It''s the same situation in the shadow_get_and_cr
eate_l2e.
sh_linear_l2_table is a linear mapping of the _shadow_ pagetables.
Xen''s pagetables have two linear mappings in them - one of the
pagetables you''re running on (__linear_l1_table &c in page.h) and
one of
the current VCPU''s shadow pagetables (__sh_linear_l1_table &c in
types.h).  sh_linear_l*_table is a macro that mixes those appropriately
to generate the address where a shadow PTE will be visible.
> There are some assemble code and it screw me up :(
>
> 3.Are there any docs to help me to figure it out?
Sorry, there''s only the code and the mailing list archives. 

Cheers,

Tim.

-- 
Tim Deegan <Tim.Deegan@citrix.com>
Principal Software Engineer, Xen Platform Team
Citrix Systems UK Ltd.  (Company #02937203, SL9 0BG)

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