11:43 <qyliss> Well, I managed to get a single ping packet in each direction over vvu

12:56 <MichaelRaskin> Is it again «but only the first one?»

13:00 <qyliss> yeah

13:17 <bqv> Has this for the most part stalled?

13:18 <qyliss> what do you mean by "this"?

13:18 <bqv> I see you have funding, but there's also been no substantial changes for a while

13:18 <qyliss> have you read the weekly updates?

13:19 <qyliss> the reason for no substantial code changes for a while is that the recent work has mostly not been code, but research and investigation

13:19 <bqv> Nope, latest looks relevant, but that's not all that obvious to passers by I'd say

13:20 <bqv> I see, fair enough

13:20 <qyliss> the weekly updates are linked from the home page, but I'm open to suggestions to make them more obvious

13:31 <qyliss> Progress is certainly not going as fast as I'd like, but we are getting there, and I work on Spectrum for hours almost every day.

13:32 <qyliss> the last big win was ~3 weeks ago, see https://spectrum-os.org/lists/archives/spectrum-discuss/8736654ij2.fsf@alyssa.is/

13:34 <qyliss> I realise it's not clear in that email what it actually means to somebody who hasn't been following the work, but essentially it means that it is now possible to run Wayland applications in one VM, and display them on a compositor in another VM, without overhead from e.g. tunneling Wayland over a network connection.

13:36 <bqv> I'm not sure. My impression was just built from seeing a fairly minimal page (not a problem) that as I saw, just linked to the repos, which were fairly stale. I definitely wouldn't have thought to check a months old blog post for confirmation that it's still active

13:36 <qyliss> I was referring to " For more frequent, detailed updates, a "This Week in Spectrum" message is posted on the “discuss” mailing list every week. "

13:37 <qyliss> I suppose I could mention that first, above the blog.

13:37 <bqv> Oh

13:37 <bqv> Yeah I missed that

18:31 <qyliss> It's looking like maybe vvu can do outbound traffic, but not inbound, or something like that

19:35 <qyliss> Somewhat good news

19:35 <qyliss> I looked at the source code and it turns out inbound traffic is not actually implemented in the vvu driver yet

19:36 <qyliss> (Why is this good news? Because it presumably will be at some point, and this means that I can just note that and move on.)

19:36 <qyliss> (Because it's not my bug)

19:37 <qyliss> Compare vvu_process_txq (inbound traffic) to vvu_process_rxq here and you'll see what I mean :) https://github.com/ndragazis/dpdk-next-virtio/blob/virtio-vhost-user/drivers/virtio_vhost_user/trans_virtio_vhost_user.c#L379

19:37 <qyliss> So I can move on to trying to do vhost-user from crosvm

20:21 <qyliss> VirtioDevice in cloud-hypervisor seems to be much closer to crosvm's than I recall Firecracker's being, which is nice for porting the vhost-user code.

22:03 <qyliss> Oh wow cloud-hypervisor actually uses the rust-vmm crates