On Wed, May 04, 2016 at 12:15:09PM -0600, Jason Gunthorpe wrote: > On Wed, May 04, 2016 at 08:36:21AM -0400, Dennis Dalessandro wrote: > > I think it's slightly more complicated than that. There are three options > > really: core, driver, or get it out of the kernel. > > From a core maintenance perspective, I think it is very simple, if > someone wishes to add code to their driver to manipulate the EEPROM, > then at this point the work to make a common uAPI falls on to their > shoulders. Agree, And before rushing to code this functionality, please double check that there is no already implemented such similar common functionality in other subsystems. I bet that you already did it. > > I'd also make a very clear message to driver submitters: Do not > include uAPIs in your initial driver patch set. Those should follow on > as dedicated well identified patches so that they attracted the proper > review. It is hard to agree with this point. It contradicts to development model of submitting whole feature at once and not unconnected piece of code, which someone will be needed to maintain without any real user behind it. > > Jason
Attachment:
signature.asc
Description: Digital signature