Keir Fraser wrote: > If the guest gives explicit hints, and the extra branch on set_pte > does not hurt, then I think it makes sense to do straightforward > explicit batching. Providing a PT page hint sounds like it could be > ambiguous in some contexts too (e.g., the fork loop modifies two PT > pages at a time). Yes, that is why I shyed away from passing PT hints - then you need to deal with the double hint case, and I think unhooking and revalidating two page tables probably does not make sense based on the UP writeable page table numbers. > >> The explicit batching does have one disadvantage without writable >> page tables, which is a potential long term maintenance / correctness >> issue - you must remove read hazards from these encapsulated paths. >> That is not so hard to do, and not a large general problem, because >> the batching is explicit rather than implicit, so you can pick paths >> to batch that are small, compact, and easy to reason about. But >> nevertheless, a point I would like to make sure you are comfortable >> with before we all decide these hooks will work for everyone. > > Yes, that's why we moved away from this approach before. But > previously we did it for *all* pagetable updates, which was a pain. > Doing it just for a few important cases, and having the hooks > maintained in upstream Linux, makes this rather less of a headache. Cool. It sounds like the lazy mode hooks are exactly what you want then? Cheers, Zach
