I'm filing this primarily to complement bug 913413. I recently set up two installations that use btrfs, one booting with the "intel-uc.img" initramfs image and the other with no initramfs at all. The latter reported the following: [ 0.854360] BTRFS: device label system devid 1 transid 86094 /dev/root scanned by swapper/0 (1) [ 0.854708] BTRFS info (device sda4): first mount of filesystem 71c15220-db0b-42e6-92eb-7aea008d23eb [ 0.854737] BTRFS info (device sda4): using crc32c (crc32c-intel) checksum algorithm [ 0.854745] BTRFS info (device sda4): using free space tree [ 0.864016] VFS: Mounted root (btrfs filesystem) readonly on device 0:16. [ 0.864054] devtmpfs: mounted Then, a moment later: [ 1.020388] systemd-gpt-auto-generator[220]: File system behind root file system is reported by btrfs to be backed by pseudo-device /dev/root, which is not a valid userspace accessible device node. Cannot determine correct backing block device. [ 1.022138] (sd-exec-[213]: /usr/lib/systemd/system-generators/systemd-gpt-auto-generator failed with exit status 1. It proceeds to boot successfully but that's probably because I have a fully populated /etc/fstab; the error message implies that systemd would (at least) not be able to support the Discoverable Partitions Specification under the present circumstances. Echoing Hadrien's comments in the xfsprogs bug, I don't think that systemd is necessarily in the wrong here. Still, there remains the broader issue of initramfs-less systems being increasingly undermined by the status quo.
(bug 438380 as a tracker isn't quite a good fit but it's close enough as a way to group related-ish bugs.)
There's nothing for Gentoo to do on this. Please work with the kernel or systemd folks if you want to resolve it.
