On Tue, Mar 04, 2025 at 05:49:51AM -0800, Christoph Hellwig wrote: > On Mon, Mar 03, 2025 at 10:03:42PM +0100, Mikulas Patocka wrote: > > Swapfile does ahead of time mapping. > > It does. But it is: > > a) protected against modification by the S_SWAPFILE flag and checked > for full allocation first > b) something we want to get rid of because even with the above it is > rather problematic > > > And I just looked at what swapfile > > does and copied the logic into dm-loop. If swapfile is not broken, how > > could dm-loop be broken? > > As said above, swapfile works around the brokenness in ways that you > can't. And just blindly copying old code without understanding it is > never a good idea. > > > > > > > Would Jens Axboe agree to merge the dm-loop logic into the existing loop > > > > driver? > > > > > > What logic? > > > > The ahead-of-time mapping. > > As said multiple times you can't do that. The block mapping is > file system private information. We might be able to provide an API to _request_ a stable mapping to a file - with proper synchronization, of course. I don't recall anyone ever trying that, it'd replace all the weird IF_SWAPFILE() hacks and be a safe way to do these kinds of performance optimizations.
