On Tue, 3 Dec 2019 16:38:50 +0000 Mark Brown <broonie@xxxxxxxxxx> wrote: > On Sun, Dec 01, 2019 at 11:40:32AM +0000, Jonathan Cameron wrote: > > > +CC Mark as we probably need a more general view point on > > the question of whether SPI mode should be enforced by binding > > or in the driver. > > Not sure I see the question here, I think I was missing a bit of > the conversation? It's perfectly fine for a driver to specify a > mode, if the hardware always uses some unusual mode then there's > no sense in forcing every single DT to set the same mode. On the > other hand if there's some configuration for the driver that was > handling some board specific configuration that there's already > some generic SPI support for setting then it seems odd to have a > custom driver specific configuration mechanism. > If the driver picks a mode because that's what it says on the datasheet it prevents odd board configurations from working. The question becomes whether it makes sense in general to assume those odd board conditions don't exist until we actually have one, or to assume that they might and push the burden on to all DT files. Traditionally in IIO at least we've mostly taken the view the DT should be right and complete and had bindings state what normal parameters must be for it to work (assuming no inverters etc) If we encode it in the driver, and we later meet such a board we end up with a custom dance to query the DT parameters again and only override if present. We can't rely on the core SPI handling because I don't think there is any means of specifying a default. We can adopt the view that in general these weird boards with inverters are weird and just handle them when they occur. Sounds like that is your preference, at least for new parts. For old ones we have no idea if there are boards out there using them with inverters so easiest is probably to just carry on putting them in the DT bindings. Jonathan
