[PATCH] ARM: fix amba_get_enable_vcore()'s handling of probe deferral

[Domenico Andreoli]

On Tue, Apr 10, 2012 at 10:09:28AM +0100, Mark Brown wrote:
> On Tue, Apr 10, 2012 at 10:48:33AM +0200, Domenico Andreoli wrote:
> > From: Domenico Andreoli <domenico.andreoli at linux.com>
> 
> > With v3.4-rc1 my board's console started to have problems. earlyprintk
> > revealed that the PL011 serial driver deferred the initialization and
> > never attempted it again. This resulted in console not being bound to
> > any physical device.
> 
> > The error in probing the serial was due to amba_get_enable_vcore()
> > not ignoring the -EPROBE_DEFER got from regulator_get(), which has
> > been recently updated to ask for probe deferral in case the regulator
> > is not available (yet).
> 
> > This fix makes amba_get_enable_vcore() consistently handle the new
> > -EPROBE_DEFER error case.
> 
> As discussed in the other thread where Shawn proposed the same fix this
> is not a good fix (and the original code is not a good idea either), it
> makes things very dependant on probe ordering which is fragile and prone
> to error.  With the code as-is platforms not using this feature really
> should be stubbing out the regulator rather than ignoring failures to
> acquire the regulator.
> 
> Russell just applied a patch removing this feature entirely (in favour
> of implementing it in power domains on platforms that want it) which
> avoids these issues.

Yes, I should have spent few minutes reading the ml instead of digging
for a night into the code - but my board had also other issues ;)

thanks,
Domenico