On Tue, Aug 11, 2020 at 1:19 AM Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote: > On Mon, Aug 10, 2020 at 2:57 PM Peter Xu <peterx@xxxxxxxxxx> wrote: > > > > Yeah, that's why I totally agree we need to do enforced COW even for a read gup > > as long as the page can be further referenced (GET|PIN). However frankly > > speaking I didn't follow the rest on what's wrong with "Userfaultfd-wp should > > not care about this because it's not a write operation" that I mentiond. Is > > that the major part of the objection? > > You didn't _explain_ why it's ok. > > You said "it's only reading". > > And I told you why "only reading" is not an argument against COW. The way I understand Peter, he doesn't want to avoid doing COW; he wants to decouple userfaultfd-WP's fault handling from COW, so that userfaultfd-wp notifies only when a previously-write-protected page is actually written to. In other words, he wants the COW to basically happen as it happens now, but it should only create a readonly PTE; and if someone later triggers a real write fault, the fault handling path would run again, and this time userfaultfd-wp would be notified before that readonly PTE is turned into a writable one. The FOLL flag would only be generated by the GUP path, not passed in by any callers. It would cause the PTEs generated by breaking COW to be read-only (I think this part is missing from the patch?), and it would cause userfaultfd-WP to not synchronously block the fault. I hope I summarized Peter's idea correctly?
