On Thu, Mar 25, 2021 at 09:58:12AM +1100, Balbir Singh wrote: > On Wed, Mar 24, 2021 at 12:24:12PM -0700, Andrew Morton wrote: > > > > > Let's remove /dev/kmem, which is unused and obsolete. > > > > I grabbed these. Silently - the cc list is amazing ;) > > > > I was wondering if it would be better to permanently disable /dev/kmem > > in Kconfig along with a comment "if you really want this thing then > > email peeps@places with a very good reason why". Let that ride for a > > year or three then blam. > > > > But this is so much more attractive, and it certainly sounds like it's > > worth any damage it might cause. > > > > We do tend to think about distros. I bet there are a number of weird > > embedded type systems using /dev/kmem - it's amazing what sorts of > > hacks those people will put up with the get something out the door. > > But those systems tend to carry a lot of specialized changes anyway, so > > they can just add "revert David's patch" to their pile. > > > > > I wonder if we should have the opposite of driver/staging and call it > outgoing, with a big thank you to the users and developers and also > to indicate this feature will be removed in the next (few) merge(s) > cycles. I guess not all code can be accumulated under a single > hierarchy. May not be worth the effort, just thinking out loud. That is exactly what drivers/staging/ is being used for, for stuff on the way out of the kernel. wimax just left the kernel that way a week or so ago, we've been doing this for many years now, the fact that no one has noticed is good as that means that no one has needed the code removed that way :) thanks, greg k-h
