| 6 Mar 2025 |
 ElvishJerricco | apparently I don't understand socket units | 19:54:46 |
 raitobezarius | systemd-ssh-generator is relevant for the server usecase (the one you have in the VM) IIRC, what I dumped is required for the client usecase (the one you have on the hypervisor) if I'm not saying dumb shit | 19:54:48 |
 gdamjan | note this is an Accept=yes socket unit, which means pid1 accepts the connection and starts the sshd with the connection file descriptors | 19:56:36 |
| ElvishJerricco | oh that makes a lot more sense | 19:57:25 |
| ElvishJerricco | so, because the host makes the connection, the socket unit starts the service unit? | 19:58:01 |
| ElvishJerricco | do I get that right? | 19:58:09 |
| gdamjan | raitobezarius: note 258 will have one more vsock implementation vsock-mux/
https://github.com/systemd/systemd/pull/32941 | 19:58:12 |
| gdamjan | or was that in 257? :D | 19:59:03 |
| ElvishJerricco | ok a lot more stuff is making a lot more sense | 19:59:07 |
| ElvishJerricco | still need to look at the ssh generator pr | 19:59:27 |
| gdamjan | yes, one per connection | 20:00:00 |
| gdamjan | (in classic inetd mode, so to speak) | 20:00:09 |
| ElvishJerricco | thanks | 20:00:21 |
| ElvishJerricco | I have a lot to learn about vsock I think | 20:00:29 |
| gdamjan | man 7 vsock :D | 20:00:42 |
| gdamjan | ugh, I remember it, the man page, was longer | 20:01:13 |
| ElvishJerricco | "No manual entry for vsock in section 7" I'm sure it's out there :P | 20:01:13 |
| gdamjan | raitobezarius: I think you DO need the vscok module for the AF_VSOCK | 20:02:04 |
| gdamjan | * raitobezarius: I think you DO need the vsock module for the AF_VSOCK | 20:02:17 |
| ElvishJerricco | this is a tangent but I'm interested in what you mean | 20:03:00 |
| ElvishJerricco | I quite like generators | 20:03:12 |
| gdamjan | but python -c 'import socket;socket.socket(socket.AF_VSOCK, socket.SOCK_STREAM, 0)' did load the modules automatically | 20:03:26 |
| ElvishJerricco | specifically because they don't need nixos | 20:03:26 |
| gdamjan | right, and they take the "live" environment into consideration | 20:04:08 |
| gdamjan | so if I move the same nixos image from bare metal to a vm vsock will start working | 20:04:49 |
| ElvishJerricco | yea it's "discoverable" in a way | 20:05:35 |
| ElvishJerricco | I just had a nice conversation with Kent Overstreet about systemd generators | 20:07:27 |
| raitobezarius | In reply to @gdamjan:spodeli.org
raitobezarius: note 258 will have one more vsock implementation vsock-mux/
https://github.com/systemd/systemd/pull/32941 oh gosh i needed that so hard | 20:07:38 |
| raitobezarius | i just fixed it in my thing | 20:07:45 |
| ElvishJerricco | which is unusual. He's normally quite difficuult | 20:08:04 |
