On Sat, Mar 10, 2018 at 01:38:44AM -0800, Christoph Hellwig wrote: > On Sat, Mar 10, 2018 at 11:58:50AM +1100, Dave Chinner wrote: > > > > 3. Add an XFS-specific implementation of swapfile_activate. > > > > > > Ugh no. > > > > What we want is an iomap-based re-implementation of > > generic_swap_activate(). One of the ways to plumb that in is to > > use ->swapfile_activate() like so: > > Hmm. Fundamentally swap is the same problem as the pNFS block layout > or get_user_pages on DAX mappings - we want to get a 'lease' on the > current block mapping, and make sure it stays that way as the external > user (the swap code in this case) uses it. The twist for the swap code > is mostly that it never wants to break the least but instead disallow > any external operation, but that's not really such a big difference. True. > So maybe we want a layout based swap code instead of reinventing it, > with the slight twist to the layout break code to never try a lease > break and just return an error for the IS_SWAPFILE case. Hmmm - won't that change user visible behaviour on swapfiles? Not that it would be a bad thing to reject read/write from root on swap files, but it would make XFS different to everything else. Speaking of which - we probably need to spend some time at LSFMM in the fs track talking about the iomap infrastructure and long term plans to migrate the major filesystems to it.... Cheers, Dave. -- Dave Chinner david@xxxxxxxxxxxxx
