Re: virtio hard disk on qemu
Mark Cave-Ayland <[email protected]> Wed, 25 Jan 2023 16:10:28 +0000
| Newsgroups | gmane.os.netbsd.ports.sparc64 |
|---|---|
| Message-ID | <[email protected]> |
On 25/01/2023 13:41, Jason Thorpe wrote: >> On Jan 24, 2023, at 9:20 PM, Tyler Mitchell <[email protected]> wrote: >> >> Hi, >> >> Has anyone tried using virtio hard disks on NetBSD/sparc64 on QEMU? >> >> I gave it a go today, but the kernel didn't finish booting. From >> what I can gather, it hung while trying to read the disklabel (on a >> blank, unformatted disk, in my case). > > Last I checked, there is a bug (or maybe it’s intentional, I dunno) in Qemu where the Virtio code assumes that DMA addresses == CPU addresses for RAM. This would impact sparc64 and alpha (which is where I noticed it). I haven’t looked for a while now, but I guess this hasn’t been fixed yet (it also might be the case that the Qemu maintainers don’t consider it a bug.) There was actually a follow-up to this here: https://lists.nongnu.org/archive/html/qemu-devel/2021-10/msg01031.html. In short, it looks like in order to use a different address space for DMA the virtio guest in NetBSD should be negotiating VIRTIO_F_ACCESS_PLATFORM. ATB, Mark.