Hi,
I'm using a KVM virtual machine running NetBSD 11.0 (on amd64). On it, I'm attempting to use Restic to perform backups and related operations over Tailscale. When I do, I consistently get errors saying "write: address family not supported by protocol family". This seems to happen regardless of whether the machine has IPv6 networking or not (just mentioning that since that's not the default network setup for qemu). For example, this is during a "restic copy" operation where I copy from the NetBSD system to a Synology NAS on my Tailscale network: repository daa3e6f6 opened (version 2, compression level auto) repository 83ee0a1d opened (version 2, compression level auto) [0:00] 100.00% 26 / 26 index files loaded [0:01] 100.00% 36 / 36 index files loaded snapshot 2b5075cd of [/home/myself] at 2026-10-16 09:39:16.78280866 +0200 CEST by myself@mymachine copy started, this may take a while... Save(<data/86d936aed2>) returned error, retrying after 590.629461ms: Post "http://:***@archiveserver.domain:9123/mymachine-archive/data/86d936aed2bc9fd0baed01c41b3d15813a85d07e3229a3fdbe0f2679a0c29d2d": readfrom tcp 100.104.233.114:65406->100.102.136.126:9123: write tcp 100.104.233.114:65406->100.102.136.126:9123: write: address family not supported by protocol family Save(<data/86d936aed2>) returned error, retrying after 2.704821272s: Post "http://:***@archiveserver.domain:9123/mymachine-archive/data/86d936aed2bc9fd0baed01c41b3d15813a85d07e3229a3fdbe0f2679a0c29d2d": readfrom tcp 100.104.233.114:65399->100.102.136.126:9123: write tcp 100.104.233.114:65399->100.102.136.126:9123: write: address family not supported by protocol family Save(<data/774763c0f9>) returned error, retrying after 1.352934404s: Post "http://:***@archiveserver.domain:9123/mymachine-archive/data/774763c0f91f212b51064ed602f6254a09248cac008901ad7093b0a7f429bfbe": readfrom tcp 100.104.233.114:65400->100.102.136.126:9123: write tcp 100.104.233.114:65400->100.102.136.126:9123: write: address family not supported by protocol family I have similar issue with "restic prune". Restricting the number of concurrent connections by means of "-o rest.connections=1" appears to help a bit, but I still get these errors. The operation ends up writing *something* to the destination, but it's treated as junk and removed when I run "restic prune" from another system later. I don't know whether this is a NetBSD networking issue, some bug in Tailscale or Restic that only surfaces on NetBSD (I use the same setup on Alpine Linux, Debian, and macOS), or whether it's something in a GO language dependency that is buggy on this platform. Has anyone seen this, or something similar? Regards, Andreas
