Re: [hybrid]: code review for function mapping pfn to foreign mfn

[Tim Deegan <tim@xen.org>]

At 18:37 -0700 on 23 Apr (1335206229), Mukesh Rathor wrote:
> On Thu, 19 Apr 2012 15:15:27 +0100
> Tim Deegan <tim@xen.org> wrote:
> 
> > At 18:29 -0700 on 13 Apr (1334341792), Mukesh Rathor wrote:
> >  - AFAICT you're using set_mmio_p2m_entry and adding a new unmap
> >    operation just to avoid having the m2p updated.  Since you can't
> > rely on the unmap always happening through the new call (and you don't
> >    enforce it anywhere), it would be better to add a new p2m_type
> >    just for non-grant foreign mappings.  Then you can gate the m2p
> >    updates in the existing code on the map being normal RAM, as is
> >    already done for p2m_is_grant().
>  
> Hi Tim,
> 
> The variants of get_page* are confusing me, so wanna double check with 
> you.  I should be able to do something like following, right?
> 

[...]

>     if ( (rc=get_page_and_type_from_pagenr(mfn, PGT_writable_page,fdom,0,0)) ) {
>         put_pg_owner(fdom);
>         return rc;
>     }

Yes, but: 

 - You should use get_page_from_pagenr() if fdom is paging_mode_external()
   and reference/copy the comment in get_page_from_l1e() to explain why:

    /* Foreign mappings into guests in shadow external mode don't       
     * contribute to writeable mapping refcounts.  (This allows the
     * qemu-dm helper process in dom0 to map the domain's memory without
     * messing up the count of "real" writable mappings.) */

 - You should drop the refcount (and typecount, if you took one) if the 
   mapping fails.

 - You need to make sure that _any_ path that removes the mapping drops
   the ref/type (_after_ any TLB flushes have happened, please!)  Maybe
   the best way to do that is in ept_set_entry() for EPT and
   paging_write_p2m_entry() for NPT/shadow. 

 - You need to handle the p2m teardown path as well.  I think the best
   way to do that is to hoist the relinquish_shared_pages() loop 
   up into a new function in p2m.c, and add your put_page[_and_type] 
   calls in there. 

Cheers,

Tim.