Crest @crest@bsd.network

@lw For SR-IOV interfaces you can lock them to a single MAC address and then use MAC based VLANs if your switch supports that, but it's a really naughty hack.

@jpmens DNSSEC bei wish.com oder ChatGPT bestellt?

@feld This is about your article https://blog.feld.me/posts/2026/02/using-new-bridges-freebsd-15/ which scared some people with the implied inability to put IP addresses on bridge interfaces (instead of on the member interfaces).

@feld The new FreeBSD 15+ bridge driver will still allow you to put IP addresses on the bridge itself and on VLAN interfaces created on top of the bridge e.g. bridge0.1. There is and never has been a need to use an epair for non-vnet jails.

@cks: If you want hear something truly cursed as progress bar I can offer this gem: a one line high VT100 scroll region that's updated with a new line every 100ms.

@mwl Apple should be liable for bricking customer devices and data that are tied to the software and service they're capricously withholding.

@FiLiS I've done the same thing 100s of timeson my home lab just today which is why I would be surprised if this was a filesystem regression, but it's hard to proof a negative.

@FiLiS And in my jail automation development I should've run into this already if it's easy to reproduce.

Create a new filesystem and copy 100s to 1000s of of files into it describes my automation pretty well from the kernel's point of view and mv does the same (before it deletes the source if crossing file systems).

Do you have block cloning enabled?

@FiLiS Is the pool on remote block storage e.g. iSCSI, because that sounds like a hanging process which would be a massive bug in FreeBSD 15.0's OpenZFS port if it was that trival to reproduce.

@dexter @zolotkey @lw I can't just use mount program because mount(8) checks that the mountpoint directory exists **before** running the mount program.

@dexter @zolotkey @lw My use case is the following: I want to have a special map that knows the jails on my FreeBSD system and compares the origins of cloned ZFS filesystems to the configured origin snapshot. If the read-only clone is stale it should be destroyed and replaced with a fresh clone.

@dexter @zolotkey @lw It looks like FreeBSD's autofs may be heavily inspired by Solaris it doesn't truly support nested mount points aka multi-maps.

If you create one direct map entry for each mountpoint mounting the nested autofs implicitly traverses the earlier autofs mountpoint which triggers the mount, but now filesystems are kept mounted if they have a child mountpoint.

A dirty workaround would be writing a mount program that mounts a whole subtree of filesystems with the automounted flag.

@rmi These fucking idiots should be punished by having to work on a 1.66GHz single-core (with SMT to reveal race-conditions), 1GB RAM, a 160GB 4200rpm HDD and a 56k modem for a year.