I'd prefer to keep them separate, as they are not tiny patches (they are roughly +200/-150 each). And, they really are quite independent - at least in the sense that I can reorder them via rebase with no conflicts, and the code builds at each commit in either orientation. I think this implies they're easier to review separately, rather than squashed. I don't have a strong feeling about the order. I slightly prefer swapping them compared to this v4 series: first introduce minor faults, then introduce CONTINUE. Since Peter also has no strong opinion, and Hugh it sounds like you prefer it the other way around, I'll swap them as we had in some previous version of this series: first introduce minor faults, then introduce CONTINUE. On Tue, Apr 27, 2021 at 8:54 AM Peter Xu <peterx@xxxxxxxxxx> wrote: > > On Mon, Apr 26, 2021 at 07:19:58PM -0700, Hugh Dickins wrote: > > On Tue, 20 Apr 2021, Axel Rasmussen wrote: > > > > > With this change, userspace can resolve a minor fault within a > > > shmem-backed area with a UFFDIO_CONTINUE ioctl. The semantics for this > > > match those for hugetlbfs - we look up the existing page in the page > > > cache, and install a PTE for it. > > > > > > This commit introduces a new helper: mcopy_atomic_install_pte. > > > > > > Why handle UFFDIO_CONTINUE for shmem in mm/userfaultfd.c, instead of in > > > shmem.c? The existing userfault implementation only relies on shmem.c > > > for VM_SHARED VMAs. However, minor fault handling / CONTINUE work just > > > fine for !VM_SHARED VMAs as well. We'd prefer to handle CONTINUE for > > > shmem in one place, regardless of shared/private (to reduce code > > > duplication). > > > > > > Why add a new mcopy_atomic_install_pte helper? A problem we have with > > > continue is that shmem_mcopy_atomic_pte() and mcopy_atomic_pte() are > > > *close* to what we want, but not exactly. We do want to setup the PTEs > > > in a CONTINUE operation, but we don't want to e.g. allocate a new page, > > > charge it (e.g. to the shmem inode), manipulate various flags, etc. Also > > > we have the problem stated above: shmem_mcopy_atomic_pte() and > > > mcopy_atomic_pte() both handle one-half of the problem (shared / > > > private) continue cares about. So, introduce mcontinue_atomic_pte(), to > > > handle all of the shmem continue cases. Introduce the helper so it > > > doesn't duplicate code with mcopy_atomic_pte(). > > > > > > In a future commit, shmem_mcopy_atomic_pte() will also be modified to > > > use this new helper. However, since this is a bigger refactor, it seems > > > most clear to do it as a separate change. > > > > > > Signed-off-by: Axel Rasmussen <axelrasmussen@xxxxxxxxxx> > > > > If this "03/10" had been numbered 04/10, I would have said > > Acked-by: Hugh Dickins <hughd@xxxxxxxxxx> > > > > But I find this new ordering incomprehensible - I'm surprised that it > > even builds this way around (if it does): this patch is so much about > > what has been enabled in "04/10" (references to UFFDIO_CONTINUE shmem > > VMAs etc). > > > > Does Peter still think this way round is better? If he does, then we > > shall have to compromise by asking you just to squash the two together. > > Hi, Hugh, Axel, > > I have no strong opinion. To me, UFFDIO_CONTINUE can be introduced earlier like > this. As long as we don't enable the feature (which is done in the next patch), > no one will be able to call it, then it looks clean. Merging them also looks > good to me. > > Thanks, > > -- > Peter Xu >
