On Mon, Jul 20, 2026 at 11:49:00AM -0400, Gregory Price wrote: > On Mon, Jul 20, 2026 at 03:38:27PM +0100, Lorenzo Stoakes (ARM) wrote: > > > > The virtual page offset of a VMA is equal to vma->vm_start >> PAGE_SHIFT if > > they are unfaulted or were not remapped, otherwise it is equal to this > > value at the point of first fault. > > > > The latter half of this description is a bit confusing. > > Why would the value differ from (vma->vm_start >> PAGE_SHIFT)? > (because it's file backed?
Because the VMA gets remapped. If it's remapped when it's unfaulted, then virtual page offset gets updated, if it's faulted then not. That's because when faulted folios may reference the original offset :) Scalable CoW will likely change this but that's a forthcoming series of many series :P > > It sounds like this value can change in response to an event, which > makes me thing a reader could see two values across two calls? Remapping :) and no a reader cannot (locking etc. :P) > > Apologies if i'm fundamentally misunderstanding something. No that's a sign that probably I'm not being clear enough! Can try to clarify on next respin. > > ~Gregory Cheers, Lorenzo
