On Mon, Mar 30, 2020 at 8:43 PM Richard Laager <rlaa...@wiktel.com> wrote:
> On 3/30/20 10:27 PM, Matthew Ahrens via openzfs-developer wrote: > > On Mon, Mar 30, 2020 at 7:08 PM Richard Laager <rlaa...@wiktel.com> wrote: > >> My only personal interest in O_DIRECT is for KVM qemu virtualization. It >> sounds like I will probably need to set direct=disabled. Alternatively, >> if I could get all the writes to be 4K-aligned (e.g. by making all the >> virtual disks 4Kn?), then ZFS's O_DIRECT would work. >> > > We were thinking that qemu *would* be able to use O_DIRECT, or at least it > wouldn't need direct=disabled. But I think your assessment implies that > qemu usually uses O_DIRECT i/o that is not page (4K) aligned > > Yes, that was my assumption. Imagine the (likely still typical) case of > 512B virtual disks. If the guest does a 512B write, is KVM really doing RMW > to make that 4K? I'm assuming not. This very well may be a faulty > assumption. I'm not a qemu developer. > QEMU could try to do the 512b O_DIRECT write, get an error, and then fall back on 512b non-DIRECT write. Or tell you to change your config to not use O_DIRECT. I can't imagine what else would work, given other filesystems' implementations of O_DIRECT. Given that we are proposing ZFS have the same semantics, and we haven't heard of anyone having to tell ext4 to ignore O_DIRECT, I don't imagine you will need to tell ZFS to ignore O_DIRECT (with direct=disabled) either. --matt ------------------------------------------ openzfs: openzfs-developer Permalink: https://openzfs.topicbox.com/groups/developer/T950b02acdf392290-Mf69b8eb6c10e94b9b6b8592e Delivery options: https://openzfs.topicbox.com/groups/developer/subscription
