Re: [PATCH 3/4 RFC] x86/p2m: use large pages for MMIO mappings

[Julien Grall <julien.grall@citrix.com>]

Hi Jan,

Sorry I hadn't much time to look at it until now.

On 15/09/15 08:34, Jan Beulich wrote:
> When mapping large BARs (e.g. the frame buffer of a graphics card) the
> overhead or establishing such mappings using onle 4k pages has,
> particularly after the XSA-125 fix, become unacceptable. Alter the
> XEN_DOMCTL_memory_mapping semantics once again, so that there's no
> longer a fixed amount of guest frames that represents the upper limit
> of what a single invocation can map. Instead bound execution time by
> limiting the number of iterations (regardless of page size).
> 
> Signed-off-by: Jan Beulich <jbeulich@suse.com>
> ---
> RFC reasons:
> - ARM side unimplemented (and hence libxc for now made cope with both
>   models), the main issue (besides my inability to test any change
>   there) being the many internal uses of map_mmio_regions())

map_mmio_regions is used in ARM to map all the device memory in a guest.
We expect this function to map everything at once when called during
DOM0 build and/or when a guest is created (used to map the interrupt
controller).

I would rather prefer to avoid introducing specific helpers with
slightly different behavior (i.e one is only mapping N page, the other
everything).

What about extending map_mmio_regions to take a parameter telling if we
want to limit the number of mapping in a single invocation?

Regards,

-- 
Julien Grall