On Sun, 16 Jul 2023 19:30:23 -0300 Jason Gunthorpe <jgg@xxxxxxxxxx> wrote: > On Sun, Jul 16, 2023 at 08:09:02AM -0700, Catalin Marinas wrote: > > > In terms of security for arm64 at least, Device vs Normal NC (or nc vs > > wc in Linux terminology) doesn't make much difference with the former > > occasionally being worse. The kernel would probably trust the DPDK code > > if it allows direct device access. > > RDMA and DRM already allow device drivers to map WC to userspace on > demand, we expect the platform to support this. > > > > So the userspace component needs to be responsible for selecting the > > > mapping, the same way using the PCI sysfs resource files today allows > > > to do that by selecting the _wc variant. > > > > I guess the sysfs interface is just trying to work around the VFIO > > limitations. > > I think just nobody has ever asked for VFIO WC support. The main > non-VM user is DPDK and none of the NIC drivers have wanted this (DPDK > applications areis more of throughput than latency focused typically) Yes, QEMU can't know whether the device or driver want a WC BAR mapping, so we've left it for KVM manipulation relative to VM use cases. Nobody has followed through with a complete proposal to enable it otherwise for direct userspace driver access, but I don't think there's opposition to providing such a thing. Thanks, Alex > > > This is particularly suited for the case (which used to exist, I don't > > > know if it still does) where the buffer that wants write combining > > > reside in the same BAR as registers that otherwise don't. > > > > IIUC that's still the case for some devices (I think Jason mentioned > > some Mellanox cards). > > Right, VFIO will have to allow it page-by-page > > > I think this interface would help KVM when we'll need a cacheable > > mapping. For WC, we are ok without any VFIO changes. > > Yes, it may be interesting to map cachable CXL memory as NORMAL_NC > into userspace for similar reasons. > > Jason >
