ARM realview breakages w/ 3.7 under qemu

[Russell King - ARM Linux]

On Fri, Dec 14, 2012 at 11:37:04AM +0100, Florian Fainelli wrote:
> Le 12/14/12 10:40, Russell King - ARM Linux a écrit :
>> On Thu, Dec 13, 2012 at 05:48:17PM +0000, Will Deacon wrote:
>>> On Thu, Dec 13, 2012 at 05:26:25PM +0000, Florian Fainelli wrote:
>>>> - PBX A9: booting with less than 1024MBytes of RAM causes this:
>>>> [    0.000000] Truncating RAM at 20000000-3fffffff to -3f7fffff (vmalloc
>>>> region overlap).
>>>> [    0.000000] Ignoring RAM at 80000000-8fffffff (vmalloc region overlap).
>>>>
>>>> Do these issues sounds familiar to any of you? The kernel config is
>>>> attached in case that helps.
>> I don't have the original email to reply to (so I don't have the .config
>> either) but my guess for the above would be that HIGHMEM is disabled.
> The original mail containing the .config file was actually rejected by  
> the mailing-list manager. I have uploaded it here:  
> http://alphacore.org/~florian/openwrt/realview-eb-config
>
> You are right, HIGHMEM is disabled. Even though enabling HIGHMEM does  
> not allow me booting either without at least 1Gbyte of RAM:
>
> [    0.000000] Memory policy: ECC disabled, Data cache writealloc
> and we are stuck.

Beware of the version of qemu you're using; older qemu had a bug in the
handling of the SP804 timers which later kernels broke (it looked to me
like qemu's SP804 code was written with assumptions about how the kernel
worked rather than how the hardware works.)  This breakage means that
the timer won't run, which in turn means that the kernel will get stuck
trying to calibrate its delay loop.

However, this does seem to be rather too early for that - and is around
the memory map re-initialization minefield.  How are you getting these
messages out of your kernel?  If you're stopping at that point you
can't be using the standard ttyAMA driver.