<qyliss>
maybe they're balancing out for never releasing 80
<qyliss>
which IIRC also happened
<qyliss>
I've fixed the "Failed to send message to virtwl: Success" log spam you'd get when running spectrum-vm
<qyliss>
(Or rather, I've applied the upstream patch that fixes it)
<qyliss>
Sommelier had its ioctl check backwards, so you'd in fact *only* get an error message on success
<qyliss>
if it failed you'd get nothing
<qyliss>
current status: got a program inside the VM to connect to a socket outside the VM over virtio_wl
<qyliss>
(This is in addition to the normal Wayland socket hooked up to a compositor)
<qyliss>
Doing this requires using special virtio_wl ioctls, that are a bit but not entirely like the normal unix socket interface
<qyliss>
So connecting to the socket works (good start), and next I need to get sending some data to the socket working
<qyliss>
And now that works too!
<qyliss>
Now lets try to receive a reply
<qyliss>
Think this would be a cool demo of using virtio_wl to put in the manual once it all works
cole-h has joined #spectrum
<MichaelRaskin>
So you had to add another argument to allow multiple virtio_wl sokets?
<qyliss>
no, that's upstream!
<qyliss>
they use virtio_wl as a generic communication channel too
<MichaelRaskin>
And the VM is always connecting to some listening socket in their current model, so what you had to do is to allow a VM to listen?
<qyliss>
All I've done so far is write a program in the guest that connects to this host socket
<hyperfekt>
i'm getting the feeling we should rename virtio_wl while we still can, this is terrible
<hyperfekt>
maybe it's already too late
<qyliss>
there's been talk of upstreaming it
<qyliss>
and you can bet there's no way the virtio spec people will want a generic message transport called "virtio_wl"
<hyperfekt>
nice, that should do it
<qyliss>
Wow, writing spectrum-vm was so worth it
<qyliss>
has made my life so much better
<Profpatsch>
I haven’t been paying attention, where do I get an overview of what spectrum-vm does?
<qyliss>
All it does is run crosvm with a predefined kernel/rootfs/etc
<qyliss>
It's literally just a wrapper around crosvm with some default values
<qyliss>
But it means if all I want to test is a custom kernel, I can do spectrum-vm -k /path/to/bzImage rather than having to construct a whole crosvm command line
<qyliss>
Anyway, for everybody interested, here's a simple demo of communicating with a host socket over virtio_wl: https://0x0.st/iLup.c
<qyliss>
MichaelRaskin: ^
<qyliss>
I haven't actually tested that file descriptors work, but they should
<qyliss>
Running that requires Chromium OS Linux HEAD, not the linux_cros we currently have packaged, btw.
<qyliss>
Because our version doesn't let support multiple virtio_wl sockets
<qyliss>
*yet
<qyliss>
I'll either backport that or just wait for the next CrOS release; haven't decided
<qyliss>
Next step is to have crosvm listen on the host end of this socket for a message asking for memory, and then reply with a memfd
<qyliss>
but jumping into crosvm is quite the context switch so I shall take a break :)
<MichaelRaskin>
wl0 means that one only gets one socket, which is eaither for Wayland or for one arbitrary socket?
<qyliss>
No
<qyliss>
/dev/wl0 is like the pty master
<MichaelRaskin>
Ah
<qyliss>
Note that I do the VIRTWL_IOCTL_NEW ioctl on it, and it gives me back a new fd
<qyliss>
You can call virtio_wl_connect as many times as you want with different names, and you'll get a different fd back each time
<MichaelRaskin>
Nice.
<qyliss>
To actually run this program, you do crosvm run --wayland-sock /path/to/s.sock,name=whatever
<qyliss>
And then "whatever" is the name you pass to virtio_wl_connect
<MichaelRaskin>
And it works with multiple invocations? Ohh, nice indeed
<qyliss>
Yep
<MichaelRaskin>
rust-9p + virtio-wl with no network stack involved sounds like a good idea now…
<Shell>
alj[m]: the librem 5 is built on a set of components decidedly not built to be put in a phone, with power management and heat properties that make them unsuitable for the task. the company has been very poor at communicating setbacks, very likely as a result of being low on funds and concerned about customers demanding their prepayments back.
<qyliss>
The one Purism product I did find interesting was the Librem mini, because coreboot support is much harder to come by on desktop-ish devices, and it looked like it would actually fill a niche
<Shell>
honestly speaking, they're *trying*, it just turns out to be... way more difficult than they expected and I expect to see Purism declare bankruptcy if they cannot make many more sales than are currently forecasted, tbh.
<qyliss>
But then I saw how much money they want for them, and I realised I could probably put together one of the corebootable supermicro servers for the same price
<qyliss>
The really good thing coming out of Purism is that they've hired a few people to work on important stuff, like Wayland and Coreboot
<Shell>
also a lot of the GNOME stuff is now in a much better shape for if a decent Linux phone/tablet *does* happen
<qyliss>
I have a sort of decent Linux sort of tablet!
<qyliss>
it's a google pixelbook
<qyliss>
(it's a 2-in-1, but a very pleasant tablet experience)
<Shell>
nice:)
<qyliss>
I recently discovered that Chromebooks are actually really cool
* Shell
recently acquired some more ram for her x200 by mentioning that she only had 2gb in some random channel
<Shell>
someone was like "that's not much" and posted me some more :p
<qyliss>
like, this thing comes from the factory running coreboot
<qyliss>
(The List will be expanded, but I figured sticking a few things I could think of off the top of my head at first and actually getting the thing on the website was a good start :))
<cole-h>
Hey, I was just about to annoy^W ping you about that :D
<qyliss>
I have some more things for the list in PMs with MichaelRaskin that I shall dig out after some sleeping
<cole-h>
🎉
<qyliss>
But the important thing is that there is now a list
<cole-h>
Now adding to it is much easier
<qyliss>
Indeed!
<tazjin>
qyliss: s/appears/appeals/
<tazjin>
(sorry I spot these automatically :p)
<cole-h>
Get that one character patch in, c'mon tazjin!
<tazjin>
cole-h: I'd need to double-check if I can contribute to docs with that license
<cole-h>
:(
<qyliss>
CC-BY-SA??
<tazjin>
checked and the answer is yes
<cole-h>
:D
<qyliss>
:3
<MichaelRaskin>
Yay, The List!
<MichaelRaskin>
Clearly incomplete, but on the other hand, can it ever be complete as long as the project is moving forward?…