[PATCH] msm: fix debug-macro.S build failure

[Russell King - ARM Linux]

On Thu, Oct 28, 2010 at 02:43:16PM -0400, Nicolas Pitre wrote:
> On Thu, 28 Oct 2010, Russell King - ARM Linux wrote:
> 
> > On Thu, Oct 28, 2010 at 02:24:33PM -0400, Nicolas Pitre wrote:
> > > On Thu, 28 Oct 2010, Daniel Walker wrote:
> > > 
> > > > This is what the function currently has,
> > > > 
> > > >         .macro  addruart, rp, rv
> > > >         ldr     \rp, =MSM_DEBUG_UART_PHYS
> > > >         ldr     \rv, =MSM_DEBUG_UART_BASE
> > > >         .endm
> > > > 
> > > > So if we have a MSM_DEBUG_UART_PHYS and MSM_DEBUG_UART_BASE we're
> > > > returning it. We don't actually have those values for all the boards
> > > > tho. My understanding was that there are some generic arm changes
> > > > needed, but I need to confirm that.
> > > 
> > > Just return 0 in both registers when you have nothing better to return.
> > 
> > That's not a good idea - it'll cause 512MB of 1:1 mappings to be setup
> > at virtual location 0 using the IO flags, which may conflict on ARMv6+.
> > A better idea would be to return 0xfff00000, which'll cause it to only
> > populate the top-most 1MB.
> 
> Given that this a phony address, better test for 0 explicitly and skip 
> the mapping as well as bailing out early from putchar, etc.

That could be 0 phys, which given there is no defined memory layout on
ARM, I would not put it past someone to put a UART at phys location 0
one day.

As you can wire ARM CPUs to boot with highvecs enabled, this is something
that could realistically happen, with the boot rom at the high address.