Re: [Xen-devel] [PATCH] Provide support for multiple frame buffers in Xen

[Robert Phillips <robert.phillips@xxxxxxxxxx>]

Tim,
I will be submitting a patch in a few minutes that addresses both of these 
concerns.
-- rsp

> -----Original Message-----
> From: Tim Deegan [mailto:tim@xxxxxxx]
> Sent: Thursday, January 10, 2013 7:29 AM
> To: Robert Phillips
> Cc: Jan Beulich; xen-devel@xxxxxxxxxxxxx
> Subject: Re: [Xen-devel] [PATCH] Provide support for multiple frame buffers
> in Xen
> 
> Hi,
> 
> I was really hoping to commit this today, but unfortunately, I've
> spotted two things:
> 
> 1. There are a few uses of *_to_virt() to get persistent mappings
>    of arbitrary addresses (PTEs, and pages of linked-list entries).
>    That's presumably going to interact badly with Jan's work to support
>    very large machines, which will reintroduce [un]map_domain_page()
>    for 64-bit x86.
> 
>    I wouldn't hold this patch back just for that, specially since the
>    code it replaces already uses maddr_to_virt(), but since it needs a
>    respin for the other thing I'd like Jan's opinion.
> 
> 2. This code still allocates user-controlled amounts of stack memory:
> 
> > +    {
> > +        int size = (nr + BITS_PER_LONG - 1) / BITS_PER_LONG;
> > +        unsigned long dirty_bitmap[size];
> > +
> > +        memset(dirty_bitmap, 0x00, size * BYTES_PER_LONG);
> > +
> > +        flush_tlb |= shadow_scan_dirty_flags(d, range,
> (uint8_t*)dirty_bitmap);
> > +
> > +        rc = -EFAULT;
> > +        if ( copy_to_guest(guest_dirty_bitmap,
> > +                           (uint8_t*)dirty_bitmap,
> > +                           size * BYTES_PER_LONG) == 0 )
> > +            rc = 0;
> > +    }
> 
> Cheers,
> 
> Tim.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel