> On Jan 2, 2019, at 6:09 AM, Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote: > > I'm not sure I agree with that. It's going to depend on exactly what this > code is doing; I can definitely see there being places in the VM where we > care about how this page is currently mapped, but I think those places > are probably using the wrong interface (get_user_pages()) and should > really be using an interface which doesn't exist yet (get_user_sg()). Fair enough; I also agree the VM_BUG_ON for tail pages might be a good safety measure, at least to see if anyone ends up calling page_size() that way at present.
