outputs all CPUs described in the DT to the elfloader
header and also includes any devices in the seL4,elfoader-devices
property.

When the FPU state is loaded from a thread for the first time the
uninitialised floating point exception register is 0 leading to the FPU
being disabled until an instruction fault occurs in that thread upon
re-entry.

Clearly, if we are loading the FPU state for a thread we want the FPU to
be enabled so we enable it whenever it is loaded.

This change broadens the caching of the FPU enable/disable bit from arm
to all platforms to ensure that on x86 time is not wasted on the
fastpath disabling or enabling an already disabled/enabled FPU.

The bootinfo extra size region is rounded up to a power of 2 size before
being mapped into the initial threads address space.  This wasn't being
handled correctly which lead to a situation where if the bootinfo
finishes on the last address of a page table but the extra region size
is not a power of 2 size then there won't be a page directory allocated
where one needs to be and the kernel would crash.

This function returns the size_bits for an allocation covering the size
of the extra boot info region.

Previously it was incorrectly handling rounding up the result of CLZL.

The elfloader is going to start using devices based on the device tree.
Add an output method for hardware_gen.py that generates header files
that the elfloader can use. Currently they contain an array of
"devices", where each device has a compatiblity string and a series of
regions (represented simply as the physical address of the region).

For now, the elfloader only uses the serial device specified by the
stdout-path property in the DTB.

With the prescaler set the frequency would be 24MHZ/2.
In order to synchronise with the timer in platsupport for mcs,
both require a 24MHZ freq. Hence this commit turns prescaling
off in the kernel. Kernel timers should accurately timeout test
1 second now. Previously it was running too fast, around half a second.

Unrelated to the rest of the changeset on this branch but a good
opportunity to make the code quality checks happy.

This change introduces a `gcc` hunting helper function to `gcc.cmake` in
order to help us find an appropriately prefixed `gcc` for the target
being built for. We use this helper to find 32b ARM cross-compiling
`gcc`s for both Debian and Red Hat based distros.

Relates to #168. The definition for `seL4_UntypedRetypeMaxObjects` lives
in the UAPI `types.h` but appears to have no link to the value actually
used by the kernel, which is configurable. This change sets the UAPI
definition to the generated definition from the kernel config steps and
defaults to the previous fixed value if, for some reason, the configured
definition is not available.

Recast introductory paragraphs to better describe the document and
characterise the intended audience.

Move information intended for document maintainers and kernel engineers
to a comment.  Add guidance for engineers so they better know when and
how to update this document.  (Thanks to Kent for the discussion!)

The `docs/sel4_release` page is not a set of release notes per se, more
like a feature grid and release history; describe it differently.

Style document title as a top-level heading.

These changes need to be reverted until we can properly test the changes
with all of our hardware online (currently we cannot test on x86
hardware).

This reverts commit a0aeec3b304d16909619bc847197df221c73a0c4.
This reverts commit af0fdc82c23f1e7e3f1e568f6987f5afd0d2150a.
This reverts commit a190a14ecaae688a6b40dd0a699b8f49c2c0631e.

When the FPU state is loaded from a thread for the first time the
uninitialised floating point exception register is 0 leading to the FPU
being disabled until an instruction fault occurs in that thread upon
re-entry.

Clearly, if we are loading the FPU state for a thread we want the FPU to
be enabled so we enable it whenever it is loaded.

This change broadens the caching of the FPU enable/disable bit from arm
to all platforms to ensure that on x86 time is not wasted on the
fastpath disabling or enabling an already disabled/enabled FPU.

Support Rocketchip SoC maps to Xilinx ZC706 board and ZCU102 board

Hifive is the only platform that supports kernel devices currently.

Ariane platform doesn't have any kernel devices. But the DTS parser
doesn't accept empty properties. Removing it completely to avoid DTS
parsing error.

Setting the environment variable GCC_COLORS or CMake option
FORCE_COLORED_OUTPUT will cause GCC's colour messages to be outputted.

This was using the hardware irq number instead of the irq number seL4
uses. This change affects MCS preemption point handling.

The numberings are changed from 1,2,... to 0,1,...

This is the virtual address for the start of the kernel device mapping
region.

Update VERSION
Update CHANGES

Try to add all missing changes that affect the seL4 interface.

Move TICKS_PER_MS out of ifdef guard,
so that the gicv3 can find it.
Turn of buildsystem check for mcs.
Add magic constants to config.

Added hypervisor support on the qemu-arm-virt platform. Required a few
changes in the gic_v2 driver to not reroute interrupts to group 1. Set
secure mode to 'off' in qemu for the elf-loader to work.

Added a check for the user's qemu version, generates a warning if less
than 3.1. This is important because the DTS QEMU generates for the virt
platform is slightly different between versions and can (will) clash
with our overlays.

On SMP configurations, the VSpace may be in use on other cores and so a
TLB shoot down has to be performed unconditionally.

Currently on x86 device untypeds are generated by passing the entire
address space minus any parts that are reserved by the kernel or that
are "real" memory (e.g. kernel image, physical RAM).

On ARM and RISC-V, device untypeds were generated at compile-time from
a device tree. This patch moves ARM and RISC-V to use the same approach
as x86, and moves the code from x86 into a common location that's
shared between the three architectures.

Co-Authored-By: Anna Lyons <anna@gh.st>

expose the maximum physical address that can be given to userspace
for use in hardware_gen.py

the intc region contains a timer that userspace uses on the MCS kernel.

hardware_gen assumes this prop is empty if it isn't specified.

The two interrupt-related 'macros' are rather confusingly different.
Explain them, rename them and simplify interrupt specification while
we're at it.

this is now implicitly true. MMIO regions should always be executeNever.