On Wed 17-10-18 09:30:58, Dan Williams wrote: > On Wed, Oct 17, 2018 at 1:18 AM Michal Hocko <mhocko@xxxxxxxxxx> wrote: [...] > > > Again, devm_memremap_pagex() exposes and relies upon core kernel > > > internal assumptions and will continue to evolve along with 'struct > > > page', memory hotplug, and support for new memory types / topologies. > > > Only an in-kernel GPL-only driver is expected to keep up with this > > > ongoing evolution. This interface, and functionality derived from this > > > interface, is not suitable for kernel-external drivers. > > > > I do not follow this line of argumentation though. We generally do not > > care about out-of-tree modules and breaking them if the interface has to > > be updated. > > Exactly right. The EXPORT_SYMBOL_GPL is there to say that this api has > deep enough ties into the core kernel to lower the confidence that the > API will stay stable from one kernel revision to the next. It's also > an api that is attracting widening and varied reuse and the long term > health of the implementation depends on being able to peer deeply into > its users and refactor the interface and the core kernel as a result. I am sorry I do not follow. For in-tree modules you have to update users whether the export is GPL or not and we do not care _at all_ about out of tree because we do not guarantee _any_ kABI/API stability (Documentation/process/stable-api-nonsense.rst). Anyway, I do not really care much, but I find the way of the argumentation dubious. I can clearly understand a simple line "me as the author want this GPL - live with that". -- Michal Hocko SUSE Labs
