arm and patch phys offset

Nicolas Pitre nico at fluxnic.net
Mon Dec 12 17:56:38 EST 2011


On Mon, 12 Dec 2011, Russell King - ARM Linux wrote:

> On Mon, Dec 12, 2011 at 11:09:00PM +0100, Michael Walle wrote:
> > Am Montag 12 Dezember 2011, 23:02:47 schrieb Russell King - ARM Linux:
> > > > > Are you running a Thumb-2 kernel?  Which kernel are you running?
> > > > 
> > > > what do you mean by which kernel?
> > > > linus' master from yesterday, ARCH_KIRKWOOD=y
> > > > CONFIG_THUMB2_KERNEL is not set
> > > 
> > > Yes, that's what I mean.
> > > 
> > > Right, so the problem you're describing is _impossible_ - there is no way
> > > the fixup function can fix only some instructions and skip over others in
> > > a properly functioning system.
> > > 
> > > The only options are that either the CPU is not executing the instructions
> > > we're giving it (unlikely as - I assume - your hardware executes older
> > > kernels fine), or for some reason your kernel is being called with caches
> > > still enabled, violating the long-standing kernel's calling requirements.
> > > 
> > > So, some more questions to try to narrow this down:
> > > 
> > > 1. What boot loader are you using?
> > u-boot 2011.09 with own bsp support.
> > 
> > > 2. What file are you taking from the kernel build in order to boot?
> > generated uImage
> 
> Okay, so this should mean that the kernel's own decompressor has run,
> which should turn on/off the mmu and caches, cleaning and invalidating
> them, which will take the boot loader completely out of the picture at
> this stage.
> 
> I think I'll leave this in the hands of Nicolas - from what I remember,
> kirkwood is a feroceon based system, and I have little working knowledge
> of that ARM derivative.

I just compiled vanilla v3.2-rc5 using kirkwood_defconfig and 
successfully booted it on a SheevaPlug and a DB-88F6281-BP development 
board.  And I do have CONFIG_ARM_PATCH_PHYS_VIRT=y.  Both platforms are 
using u-Boot which booted the default uImage produced by the kernel 
build system.

So at this point I wouldn't think the issue is Kirkwood related as I 
can't reproduce it on two different Kirkwood targets.


Nicolas



More information about the linux-arm-kernel mailing list