> > The issue I have is that the i2c device emulates several other devices > > with existing drivers (pca953x, ds1672, at24) and those drivers don't > > have any retry mechanism in place for a retry. > > > > Maybe if I converted those drivers to use regmap I could implement a > > regmap with retries in the mfd driver for my device? > > Wolfram: what do you think? What NAK means is device specific IMO, so to do it generally at regmap level is convenient but wrong. at24 already has retry support because it may be accessed while in an erase cycle. I don't know the other devices. You need to discuss with the driver authors/maintainers if a NAK can generally mean "let's retry" or if you need a seperate i2c_device_id for your variant which then handles the NAK differently.
Attachment:
signature.asc
Description: PGP signature
