On Tue Oct 21, 2025 at 4:04 PM CEST, Michael Köppl wrote: > Tested this in both the GUI and TUI installers and the autoinstaller. > Tried the following: > > - Pinning names for multiple interfaces > - Pinning names for interface with invalid MAC in autoinstaller > Noticed that in this case, the default of nic0 is used since the map > from MACs to interface names is pre-populated with the default > values. So if I use a MAC address in my answer file that does not > exist, the interface name for any interface not mentioned in the > answer file will be set to nic0, since if > network.interface-name-pinning.enabled is set to true, the pinning > is done in any case. I suppose this is on purpose, since there's no > way to check for the existence of the MAC address beforehand and > once the installation is running, the .link file has already been > written. Just wanted to mention it nonetheless.
Yeah, it's intended. In the auto-installer case, we basically have to trust the administrator to do things (more) diligently, as checking whether MAC addresses actually exist can only be done at installation time, really. Although; I'll add a warning when an invalid/unknown MAC address is encountered at the start of the installation, so that there is at least _some_ notice. > - Testing invalid interface names in GUI, TUI, autoinstaller > Noticed that in the GUI, the error dialog will show up behind the > form where users enter the interface name. The form cannot really be > interacted with and the window has to be moved to the side to reveal > the error message. Oh, that's interesting. I'll look into it, thanks! > - Explicitly test the case of pinning a fully numeric interface, since > it is allowed by the installers, but systemd does not allow this [0]. > The result of this is that the interface that's affected does not > come up on first boot, since /etc/network/interfaces configures the > numeric-only interface which does not exist. Yep, going to add some additional validation for interface names. > - Checked that disabling pinning actually does not set any interface > names > > Other than the above, this seems to work as advertised. Nice work! > > Also had a look at the code and left a few comments on the individual > patches, but did not notice any real problems with it. A single patch > seems to add code that is not correctly formatted and other than that > just added some suggestions. > > With the comments on the individual patches addressed, consider this: > Tested-by: Michael Köppl <[email protected]> > Reviewed-by: Michael Köppl <[email protected]> Thanks! _______________________________________________ pve-devel mailing list [email protected] https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel
