On 20.12.2022 11:26, Bryan O'Donoghue wrote: > On 20/12/2022 10:21, Konrad Dybcio wrote: >> The point of this patchset is to remove logic like this, as it's >> very repetetive and all it does is read fuses with a ton of magic >> offsets. Dmitry pushes that to DT with the generic nvmem API, so >> that instead of these magic &s and <<s, one is simply supposed to >> define QFPROM fuses at the correct offset and with a matching bits=<> >> property. This does not remove any functionality, it just changes >> how the fuses are accessed and makes the code more generic. >> >> Konrad > > Hmm but then there's extra work to get this working again on 8939 right? > > Seems very dogmatic to drop working code for want of sending a dtsi out. > > Certainly my preference is to keep exiting working code and just complete landing the relevant dtsi, rather than eject working code and have to do the same work all over again. Well, I don't really know how to say it in a way that wouldn't make it sound mean, but trust me I don't want it to be mean.. Mainline does not and will not (for the most part) care about out of tree code, so cleanups of parts like this with no users are wholly expected if your DT hasn't landed upstream (or has been stuck in review for a long long time like it is the case with various parts of 8939).. Keeping this old iteration is blocking progress, as the other similar ones (that *do* have mainline users) are left in place just to be backwards compatible with old DTs that may have been pulled from torvalds/linux by third party projects, like U-Boot, *BSDs or something. Trimming away this now-duplicated code will shrink the driver, reducing bloat for everyone that compiles it in and doesn't use the 8939-specific path. Konrad > > --- > bod
