virtio — archive

[Rusty Russell]

RAUCHFUSS Holm <holm.rauchfuss@huawei.com> writes: >> From: Rusty Russell [ mailto:rusty@au1.ibm.com ] >> Paul Mundt <Paul.Mundt@huawei.com> writes: >> > Add some clarification to the head of the virtqueue section outlining >> > how the vring/virtio_ring terminology is used throughout the >> > specification. >> >> Here's an alternative, which simply uses consistent naming. >> >> This is slightly cleaner of course, but is a bit of a pain for those >> used to the legacy names. So it needs addition of the legacy sections >> mentioning that, but I'll wait until MST merges his stuff into master. > > I think this approach does more harm than good. I suggest to rather use > the patch from Paul. It is less invasive to the spec. I understand your concern, but the pain of changing the spec is transitory. The pain of having an unclear spec is more permanent. > Otherwise, we would > also have to change the virtio_ring.h in the appendix (which the patch > from Rusty did not modify) and we could not link to the Linux version > of it as it uses the wrong naming The Linux headers will need their own set of fixes anyway, for handling both legacy and non-legacy cases. I already did this in my experimental Linux patchset. > (btw, path in the text is pointing > to the wrong virtio_ring.h, trivial patch to fix this is attached). Thanks, I've applied that immediately. Cheers, Rusty.