Hi Maxime,

2015-05-23 10:28 GMT+02:00 Maxime Ripard [off-list ref]:

On Thu, May 21, 2015 at 10:10:35PM +0200, Maxime Coquelin wrote:

quoted

We have the same problem on stm32, as just discussed with Andreas [0].
I decided to patch the bootloader to deassert timers reset, after
discussions with Rob Herring and Arnd.

That seems like the best solution, but we don't have control over the
bootloader in most cases. We have an alternative implementation (ie
mainline u-boot) that is required for a few things already, but I'd
really like to still be able to at least boot the kernel on a "legacy"
bootloader (that is the one flashed on the device).

For sure it will have to remain compliant with the legacy bootloader.
It is easier for STM32 since it is a newly platform.

Regards,
Maxime C.