On Fri, Oct 27, 2023 at 09:11:57AM +0300, Amir Goldstein wrote: > On Fri, Oct 27, 2023 at 9:01 AM Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote: > > > > On Thu, Oct 26, 2023 at 10:46:06PM +0300, Amir Goldstein wrote: > > > I would much rather turn EXPORTFS into a bool config > > > and avoid the unneeded build test matrix. > > > > Yes. Especially given that the defaul on open by handle syscalls > > require it anyway. > > Note that those syscalls depend on CONFIG_FHANDLE and the latter > selects EXPORTFS. Yes, this means that for all somewhat sane configfs exportfs if always built in anyway. And for the ones where it isn't because people are concerned about micro-optimizing kernel size, nfsd is unlikely to be built in either. > The bigger issue is that so many of the filesystems that use the > generic export ops do not select EXPORTFS, so it's easier to > leave the generic helper in libfs.c as Arnd suggested. Agreed.
