On Sat, Jun 17, 2017 at 08:09:38AM -0700, Christoph Hellwig wrote: > On Sat, Jun 17, 2017 at 08:42:54AM -0400, Olga Kornievskaia wrote: > > > > > On Jun 17, 2017, at 6:03 AM, Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote: > > > > > > On Fri, Jun 16, 2017 at 01:36:30PM -0400, J. Bruce Fields wrote: > > >> It makes a difference in the clone case, doesn't it? (Though I'd think > > >> best there would be a special "copy to end of file" value.) > > > > > > Clones uses a 0 length as "whole" file. Both for the NFS operation > > > and the Linux ioctl. > > > > Does that actually work? There is check vfs_copy_file_range() > > Please re-read the above sentence. I'm talking about NFS clone and > IOC_CLONE_RANGE, not copy_file_range. copy_file_range had to follow > the NFS semantics that don't have the special 0 case. Nope, COPY has it: https://tools.ietf.org/html/rfc7862#section-15.2.3 A count of 0 (zero) requests that all bytes from ca_src_offset through EOF be copied to the destination. I think leaving it out of copy_file_range was just an oversight. --b.
