Do I need to invalidate caches before enabling (on ARMv7)?

[Uwe Kleine-König]

Hello,

On Sun, Dec 07, 2014 at 02:30:31AM +0530, Afzal Mohammed wrote:
> On Thu, Dec 04, 2014 at 11:03:57AM +0100, Uwe Kleine-König wrote:
> 
> > The breakage I'm currently seeing in barebox might well be explained by
> > stale I-cache entries and barebox (as of now) doesn't invalidate the
> > i-cache before enabling it. Looking into how Linux enables the I-cache
> > in the decompressor for v7[1] revealed that the caches are not cleaned
> > there either. (So my plan to copy from Linux failed :-)
> > 
> > Now I wonder if that is only an unlikely (or even theoretical) issue
> > that wasn't noticed up to now or if I'm missing something.
> 
> > So stale entries in the cache might even hurt before the cache is
> > enabled?! This would mean that you want to invalidate/flush the cache at
> > disable-time. Still I think doing it before enabling it in Linux would
> > be a good idea. And if it's only because bootloaders and (maybe worse)
> > boot roms cannot be trusted in this area.
> 
> Are you sure that stale I-cache is causing the issue ?, but
No, at least it's not the only problem. I found another cache related
problem, similar to yours. Still I think it's the right thing to do (at
least in the bootloader) to invalidate the cache before enabling.

> D-cache should not have stale data - it is a pre-requisite for
> booting the Kernel [1,2] (though not in Booting documentation)
> 
> We were troubled by this issue when Kernel was loaded directly w/o
> bootloader, since D-cache was not invalidated, upon enabling,
> due to stale data Kernel was crashing randomly
> 
> Did want to patch it for long time, but then with cash invalidation
What is cash invalidation? Do you rupture bank notes? (SCNR)

> and all it never happened ;)
> 
> Regards
> Afzal
> 
> [1] http://www.arm.linux.org.uk/developer/booting.php
> [2] http://comments.gmane.org/gmane.linux.ports.arm.kernel/77718
> 
> ---------------------------------8<--------------------------------------
> diff --git a/Documentation/arm/Booting b/Documentation/arm/Booting
> index 371814a36719..c4c423164d5e 100644
> --- a/Documentation/arm/Booting
> +++ b/Documentation/arm/Booting
> @@ -193,6 +193,7 @@ In any case, the following conditions must be met:
>    The MMU must be off.
>    Instruction cache may be on or off.
>    Data cache must be off.
> +  Data cache should be invalidated.
That would match http://www.arm.linux.org.uk/developer/booting.php which
is good. I'd prefer if Linux handled stale data in the cache though for
two reasons:
 - Usually bootloaders are not perfect in adhering to requirements that
   are not obvious from testing.
 - Stale caches introduce problems that are hard to debug.

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |