blockdev: handle ZFS dataset sources in list_dev_by_dir

[hanthor]

ZFS filesystems report their mount source as a dataset name (e.g. rpool/root) rather than a block device path. Passing this to lsblk causes it to fail with not a block device, breaking bootc status and bootc upgrade on ZFS-rooted systems.

Add list_dev_for_zfs_dataset() which extracts the pool name and queries zpool list -H -v -P <pool> to find the underlying block device path, then delegates to list_dev() as normal.

Detect ZFS sources in list_dev_by_dir() via fstype == "zfs" or by a non-absolute path containing / (dataset name form).

Testing

Tested on an Ubuntu 26.04 system installed to ZFS via bootc install to-filesystem. The installed system uses rpool/root as the rootfs.

• bootc status — previously failed with lsblk: rpool/root: not a block device, now returns full status output ✅
• bootc upgrade --check — previously failed at the same point, now successfully queries the registry ✅

Reproduced against v1.15.0 and verified fix against current main.

Fixes: #1240
Assisted-by: GitHub Copilot (claude-sonnet-4-6)

[gemini-code-assist]

Code Review

This pull request adds support for identifying block devices backing ZFS datasets by introducing the list_dev_for_zfs_dataset function and updating list_dev_by_dir. The review feedback suggests improving the robustness of the ZFS pool name extraction and narrowing the filesystem detection heuristic to prevent incorrect matches with network filesystems like NFS.

[gemini-code-assist]

The split('/').next() call on a string will always return Some (even for an empty string), making the ok_or_else block unreachable. Furthermore, if the dataset string is empty or only contains the ZFS= prefix, pool will be an empty string, which will cause the zpool command to fail later. It's safer to filter out empty strings to ensure a valid pool name is extracted.

Suggested change

	let pool = dataset
	.split('/')
	.next()
	.ok_or_else(|| anyhow!("Invalid ZFS dataset: {dataset}"))?;
	let pool = dataset
	.split('/')
	.next()
	.filter(|s| !s.is_empty())
	.ok_or_else(|| anyhow!("Invalid ZFS dataset: {dataset}"))?;

[gemini-code-assist]

The heuristic (!source.starts_with('/') && source.contains('/')) is too broad and will incorrectly match network filesystems like NFS (e.g., server:/export). This would cause bootc to attempt running zpool on non-ZFS systems, which might fail if the zpool command is missing or the source is not a valid pool. It is recommended to rely primarily on fsinfo.fstype == "zfs" or a more specific prefix like ZFS=.

Suggested change

	if fsinfo.fstype == "zfs" || (!source.starts_with('/') && source.contains('/')) {
	if fsinfo.fstype == "zfs" || source.starts_with("ZFS=") {

[hanthor]

I don't know if this is wanted since ZFS doesn't support fs very as-is but it is a fun way to maintain an is lifecycle without it still. ZFS on root is fun and cool, but I'm more working on this as a proof of concept Ubuntu 26.04 Desktop