On Fri, Jul 08 2022, Peter Xu <peterx@xxxxxxxxxx> wrote: > On Fri, Jul 08, 2022 at 03:03:34PM +0200, Cornelia Huck wrote: >> I was thinking about a new flag that implies "copy metadata"; not sure >> how we would get the same atomicity with a separate ioctl. I've only >> just started looking at userfaultfd, though, and I might be on a wrong >> track... One thing I'd like to avoid is having something that is too >> ARM-specific, I think there are other architecture features that might >> have similar issues. > > Agreed, to propose such an interface we'd better make sure it'll be easily > applicable to other similar memory protection mechanisms elsewhere. There's storage keys on s390, although I believe they are considered legacy by now. I dimly recall something in x86 land. > >> >> Maybe someone more familiar with uffd and/or postcopy can chime in? > > Hanving UFFDIO_COPY provide a new flag sounds reasonable to me. I'm > curious what's the maximum possible size of the tags and whether they can > be embeded already into struct uffdio_copy somehow. Each tag is four bits and covers 16 bytes (also see the defs in arch/arm64/include/asm/mte-def.h). _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
