xeon phi: move HW MODEL flag to public header as DMA library also needs
it

Signed-off-by: Roni Häcki <roni.haecki@inf.ethz.ch>

xeonphi: towards support for model based allocations

Signed-off-by: Reto Achermann <reto.achermann@inf.ethz.ch>

Xeon Phi boot: getting rid of the single boot image and dependencies

Adapting the boot protocol of the Xeon Phi cards to dynamically
load the individual modules at run time rather than compile time.

The following sources can be used to load the modules
- in ramfs of host machine
- from a TFTP server (using the --tftp flag)
- from NFS (using the --nfs flag)

The modules to be loaded are specified in a supplied menu.lst file.
The parameter --modlist can be used to specify that file.

The boot driver will handle the generation of the needed data structures
and headers when the card boots.

Signed-off-by: Reto Achermann <reto.achermann@inf.ethz.ch>

Xeon Phi Driver: switched to the new Xeon Phi Client, added open call to interphi interface.

adapted the K1OM boot to hand over the message frame size to xeon phi.

Addition of checks if running on xeon phi or host

Bugfix: Xeon Phi ID

Moved the xeon phi id to another part of the struct to make it work.
Removed ifdefed out code form xeon phi driver.

Kernel: Added a new field to the dispatcher struct for K1OM architecutre

Addition of xeon_phi_id information to the dispatcher generic struct
Adding of a new function to obtain this structure
Adding of new fields to multiboot and coredata structures
(xeon_phi_info)
Adapted lib spawndomain to forward the own xeon phi id when spawning
kernel: shows message on which Xeon Phi we are running on.

Xeon Phi DMA: Service stubs

Added some stubs for the service interface of the Xeon Phi DMA server.

Xeon Phi DMA

initial design and layout of the Xeon Phi DMA functionality.

Exchange of base address for messaging bootstrap

When the host starts the Xeon Phi it writes the address of the messaging
frame into the bootparams and this address is passed to the xeon phi
cardside driver which contacts the host and uses this frame to establish
more messaging channels.

Added K1OM architecture specific extensions that pass certain caps / values to the Xeon Phi manager (card side)

- Passing of cap for the MMIO space and the System Memory range
- passing of base address for the host messaging listener
- Removed IO cap

Adapted xeon phi boot params and passing of cmd lines.

added entries to gitignore

Addition of a second supercn.

It turned out that running Barrelfish on machines with lots of RAM will
results in a error during booting that the supercn has no more space
left. As soon as supercn0 has no space left, the allocator will switch
to the second supercn1 to store the capabilities.

Also memserver will get both endpoints and switches automatically to the
next supercn while building up the memory allocator. The maximum size
bits are extended to 40.

The offsets are adapted to support 1024GB physical memory.

The implementation has been tested on qemu prior commit.

Added: Xeon Phi Manager

Responsible for assigning a unique ID to the Xeon Phi drivers and
exchanging the irefs of xeon phi drivers such that they can connect
to eachother