Re: [PATCH v2] Revert "PCI: imx6: Add support for active-low reset GPIO"

[Lucas Stach <l.stach@pengutronix.de>]

Am Donnerstag, den 31.03.2016, 07:07 +0200 schrieb Krzysztof Hałasa:
> Fabio Estevam <festevam@gmail.com> writes:
> 
> > In order to make the Toradex board to work without breaking old dtb's
> > is to introduce a property like 'phy-reset-active-high' and handle it
> > in the driver.
> 
> Well, I don't know. Most DTBs are distributed along the kernel on the
> same boot medium. Both ways are far from the ideal. Reverting for now is
> ok, but long-term I'd rather fix the buggy DTSs and remove extra stuff.
> 
> Old DTBs aren't compatible with new kernels anyway and I guess it will
> be the case in the future again (e.g. IIRC pre-4.2 IMX6 DTBs don't boot
> on 4.2+ - or was it 4.1?).

Sorry, but I strongly object to the argument of "DTs are not stable
anyways, so we might just continue to break them". We are trying to keep
them stable.

I think you are referring to the GPC interrupt domain change with your
above statement about DT stability breaking on 4.1/4.2. This is not the
case. While I only catched the boot regression in the -rc phase of this
kernel, the final released kernel will boot just fine with an old DTB.

The i.MX platform tries to keep DTs stable. No exceptions.

Regards,
Lucas