[RFC PATCH 1/5] gpio: gpiolib: Add core support for maintaining GPIO values on reset

Charles Keepax ckeepax at opensource.cirrus.com
Wed Oct 25 19:11:41 AEDT 2017

On Fri, Oct 20, 2017 at 07:02:27PM +1030, Andrew Jeffery wrote:
> On Fri, 2017-10-20 at 09:43 +0200, Linus Walleij wrote:
> > On Fri, Oct 20, 2017 at 9:17 AM, Linus Walleij <linus.walleij at linaro.org> wrote:
> > > > > On Fri, Oct 20, 2017 at 5:37 AM, Andrew Jeffery <andrew at aj.id.au> wrote:
> > > 
> > > > GPIO state reset tolerance is implemented in gpiolib through the
> > > > addition of a new pinconf parameter. With that, some renaming of helpers
> > > > is done to clarify the scope of the already existing
> > > > gpiochip_line_is_persistent(), as it's now ambiguous as to whether that
> > > > means on suspend, reset or both.
> > > 
> > > Isn't it most reasonable to say persistance covers both cases, reset
> > > and/or sleep? This seems a bit like overdefined.
> > 

Seems reasonable to me to just expand the existing stuff to cover
reset as well, I don't think that should cause any issues for the
Arizona stuff.

> > I should also add: right now persistance is defined in negative terms,
> > you can supply the flag "may lose value", which means the subsystem
> > by default, and driver by default, will try to keep values persistent across
> > sleep.
> > 
> > Then it is possible to opt in for not doing so. (Usually to save power I
> > think.)
> > 
> > I think that especially for userspace use cases, saving power should
> > not really be the concern, but correct me if I'm wrong. I am thinking
> > of a box with a DC plug wired up to a factory line here.
> > 
> > What we have in the Arizona driver is an opt-in where the DT can
> > say "don't preserve the value  this line during system sleep" i.e. "lay lose
> > value" and we can extend that flag to mean "don't preserve this line
> > during reset either" but by default assume that we should.
> Yeah, the preserve polarity was another thing I debated given the
> current example with the Arizona driver. Not preserving is the default
> for the Aspeed hardware, so that ended up influencing my choice. Not
> that implementation details should necessarily influence interface
> design, but it was at least more than a coin toss.

The way I think we ended up looking at this was when a user
requests a GPIO the least surprising thing for them is the value
maintains until they change it again. So it made sense to make
the allowing the value to be dropped the opt in side.

> I don't have anything specific against preserving by default, just my
> gut instinct and the hardware went the other way. As long as we expose
> the option to opt out, which the additions for the Arizona already do.
> Cheers,
> Andrew


More information about the Linux-aspeed mailing list