On Mon, Apr 24, 2023 at 09:28:07AM -0300, Jason Gunthorpe wrote: > On Mon, Apr 24, 2023 at 11:17:55AM +0100, Lorenzo Stoakes wrote: > > On Mon, Apr 24, 2023 at 02:43:56AM -0700, Christoph Hellwig wrote: > > > I'm pretty sure DIRECT I/O reads that write into file backed mappings > > > are out there in the wild. > > I wonder if that is really the case? I know people tried this with > RDMA and it didn't get very far before testing uncovered data > corruption and kernel crashes.. Maybe O_DIRECT has a much smaller race > window so people can get away with it? It absolutely causes all kinds of issues even with O_DIRECT. I'd be all for trying to disallow it as it simplies a lot of things, but I fear it's not going to stick. > So, my suggestion was to mark the places where we want to allow this, > eg O_DIRECT, and block everwhere else. Lorenzo, I would significantly > par back the list you have. I think an opt-in is a good idea no matter how many places end up needing it. I'd prefer a less dramatic name and a better explanation on why it should only be set when needed. > I also suggest we force block it at some kernel lockdown level.. > > Alternatively, perhaps we abuse FOLL_LONGTERM and prevent it from > working with filebacked pages since, I think, the ease of triggering a > bug goes up the longer the pages are pinned. FOLL_LONGTERM on file backed pages is a nightmare. If you think you can get away with prohibiting it for RDMA, and KVM doesn't need it I'd be all for not allowing that at all.
