Regression: not booting anymore on OMAP35xx

[Luca Ceresoli]

Sascha Hauer wrote:
> On Tue, Jun 22, 2010 at 04:04:52PM +0200, Luca Ceresoli wrote:
> > Sascha Hauer wrote:
> > > Hi Luca,
> > > 
> > > On Tue, Jun 22, 2010 at 02:26:11PM +0200, Luca Ceresoli wrote:
> > > > Hi,
> > > > 
> > > 
> > > Seems that I lost the call to arch_init_lowlevel. Does the following
> > > patch help?
> > 
> > I'm afraid it doesn't seem to.
> > It fails with the same symptoms.
> > 
> > Luca
> > 
> > > 
> > > 
> > > diff --git a/arch/arm/cpu/start.c b/arch/arm/cpu/start.c
> > > index eea7dcf..91d2ccd 100644
> > > --- a/arch/arm/cpu/start.c
> > > +++ b/arch/arm/cpu/start.c
> > > @@ -50,6 +50,8 @@ void __naked __bare_init reset(void)
> > >  {
> > >  	uint32_t r;
> > >  
> > > +	arch_init_lowlevel();
> > > +
> 
> Ok, this has to be done *after* switching the cpu to SVC32 mode.
> 
> I tested this on my beagle board and now it works again. I pushed
> a patch to the master repository.

Sascha, thank you for the prompt response.
Unfortunately the new version works only partially with my Beagle
Board (rev. C4).

I tested commit:
 46976f7 OMAP: fix compilation

- Test 1: the board is always powered.
  Press RESET with the USER key pressed.
  10 tests, 10 passed.

- Test 2: the board is not powered.
  Turn turn it on with the USER key pressed.
  Turn turn it off again at the end of the test.
  10 tests, 5 passed, 5 failed.

During all of the 20 tests, barebox.bin was always the same.
I didn't recompile it.

OTOH the commit before the "reimplement in C" passes also test 2.

I hope this gives enough hints to solve the problem.

Luca