Consistent DNS resolution failure

[fatanugraha]

Description

I'm using Lima for work and I noticed that if I:

1. Run Lima VM
2. inside the VM: curl www.google.com works fine.
3. Connect to my work VPN
4. inside the VM: curl www.google.com failed due to DNS resolution.

As it turns out after I connected to my VPN:

1. the previous nameservers in /etc/resolv.conf before I connected to my VPN are no longer reachable
2. My VPN updated the nameservers list in /etc/resolv.conf
3. gvisor-tap-vsock are still using the nameserver list on step (1) hence causing the DNS resolution to fail.

systemd-resolved logs:

lima@colima:/$ sudo journalctl -u systemd-resolved -f
Jan 07 15:45:19 colima systemd-resolved[286]: Grace period over, resuming full feature set (UDP+EDNS0) for DNS server 192.168.5.2.
Jan 07 15:45:19 colima systemd-resolved[286]: Using degraded feature set UDP instead of UDP+EDNS0 for DNS server 192.168.5.2.
Jan 07 16:18:02 colima systemd-resolved[286]: Clock change detected. Flushing caches.

limactl usernet log

time="2025-01-07T16:57:43+08:00" level=error msg="Error during DNS Exchange: read udp <$corp-ip>:52316->192.168.50.1:53: i/o timeout"

I'm raising PR here to fix the issue in upstream. Creating this issue to ensure we're using the latest version once it's merged.

[fatanugraha]

Note that the fix in the upstream only fixes this issue for linux and darwin as I don't have access to windows machine.

The release will be tracked in this issue instead: #3132