I've committed the arm64 fix in rS365578.

I have no idea, although if we remove it I think it should be a second change that could be easily reverted if needed.

Only support aout.c on i386 & amd64

Why is it in the pci code rather then iommu? I could imagine broken non-pci devices could also need a DMA alias.

Normally non-WB memory would be accessed via bus_space. KCSan will ignore these.

Remove a gpiokeys change

• Style fixes
• Split out the ACPI attachment
• Add an FDT attachment

Grab as I have style fixes & FDT support

Invalidate the tlb after calling set_ttbr0

Remove an unneeded dsb

• Store the ttbr address so we don't need to walk the page tables
• Free pages in pmap_release
• Update the wording of a comment in pmap_pinit_stage

• Remove "All rights reserved"
• Fix spelling
• Use the argument to vector_tail

Not yet, I'm planning on that soon.

I expect to commit this in multiple commits, but need to figure out where the best places to split are.

You should use get_kernel_reg to read the ID registers when we need a common version over all CPUs. It's not going to affect it here, but when use later in the boot it can as on a big.LITTLE system these ID registers can sometimes be different across the different core types.

We only map physical memory that exists, so this shouldn't cause a problem on existing systems.

In D26108#579497, @gbe wrote:

Good to go from manpages!

It would maybe feasible to include a notice about incompatibles about other kernel santinzing options.
I remember that I tried to build a kernel once with following options, which was not possible:

include GENERIC
ident GENERIC-SANTI
options		COVERAGE
options		KCOV
options		KCSAN

In D26108#579767, @tuexen wrote:

Maybe you can mention that it is only supported on amd64 and arm64, possibly in the BUGS section.

On arm64 we have the following in fork_trampoline to stop unwinding past the top of the stack.

mov     fp, #0  /* Stack traceback stops here. */

Committed in rS363636

Do any of the new caps require kernel support before they can be used? If so we shouldn't report them to the user until that support is in place.

arm64 enables interrupts for userspace exceptions in do_el0_sync. It also needs to enable them in data_abort when not in the kernel debugger. See rS287961 and rS353920 for details.

Always return RPI_FW_GPIO_DEFAULT_CAPS from getcaps.

Switch to being simplebus based. Depends on D25925.