mara/hyperhive
mara/hyperhive
Watch
0
0
Fork
You've already forked hyperhive
 0
Code Activity

allow erofs as the idmapped store filesystem

Browse source 
nspawn idmaps the store on erofs, which the VM test's store uses; the
allowlist from the store check refused it and failed the namespaced-agent
subtest in vm-nightly #7084. vm-nightly #7030 ran the same agent on the
same erofs store green. The magic is EROFS_SUPER_MAGIC_V1 from the
kernel's linux/magic.h; libc has no constant.

Refs #5888
This commit is contained in:
atlas 2026-11-06 15:45:01 +01:00 • committed by mara
commit a71e7cea75
2 changed files with 15 additions and 11 deletions
Download patch file Download diff file Expand all files Collapse all files

11
docs/agent-lifecycle/persistence.md
View file

@ -488,13 +488,14 @@ hive-c0re can't read, or none, leaves the agent as it is.
user. hive-priv mounts the agent's `state/`, `harness/`, `work/` and
`claude/` dirs and its socket dir with `owneridmap`, and every other bind too
(config, `/knowledge`, the runtime dir), whose owner shows as root inside.
- **`/nix/store` has to be on ext4, btrfs or xfs.** nspawn mounts the store
into a namespaced container idmapped, and the kernel refuses that on
- **`/nix/store` has to be on ext4, btrfs, xfs or erofs.** nspawn mounts the
store into a namespaced container idmapped, and the kernel refuses that on
overlayfs and on some other filesystems. `WriteNspawnFlags` for an opted-in
agent checks the `statfs` type of `/nix/store` against an allowlist of those
three. The check refuses every other filesystem, zfs and tmpfs included, even
where a newer kernel could idmap it, because the gate exists to stop a
container that would not boot. The error names the type and the fix, and
four; the VM test's erofs store idmaps on the hive's kernel. The check
refuses every other filesystem, zfs and tmpfs included, even where a newer
kernel could idmap it, because the gate exists to stop a container that would
not boot. The error names the type and the fix, and
comes before anything moves. Check a host with
`findmnt -no FSTYPE /nix/store`.
- **The container starts with `--private-users=pick --private-users-ownership=map`.**

15
hive-priv/src/userns.rs
View file

@ -51,8 +51,11 @@ const STORE: &str = "/nix/store";
type FsType = libc::__fsword_t;
/// `f_type` of the filesystems that support idmapped mounts and that a store
/// can sit on: ext4, btrfs and xfs. Nothing else is accepted.
const IDMAP_FS: [FsType; 3] = [0xEF53, 0x9123_683E, 0x5846_5342];
/// can sit on: ext4, btrfs, xfs and erofs (`EROFS_SUPER_MAGIC_V1` in the
/// kernel's `linux/magic.h`; `libc` has no constant for it). Nothing else is
/// accepted. A namespaced agent ran idmapped binds on the VM test's erofs
/// store, so erofs idmaps on our kernel.
const IDMAP_FS: [FsType; 4] = [0xEF53, 0x9123_683E, 0x5846_5342, 0xE0F5_E1E2];
/// The dirs under `<AGENT_STATE_ROOT>/<agent>` the agent writes into. `config`
/// is not one of them: hive-core writes it and the agent only reads it. `state`
@ -111,7 +114,7 @@ fn fs_type(path: &Path) -> std::io::Result<FsType> {
Ok(buf.f_type)
}
/// Refuse a store whose filesystem is not ext4, btrfs or xfs. nspawn clones a
/// Refuse a store whose filesystem is not ext4, btrfs, xfs or erofs. nspawn clones a
/// namespaced container's `/nix/store` bind with an idmap, which the kernel
/// refuses on other filesystems, so the container would not boot. Others are
/// refused even where a newer kernel could idmap them.
@ -122,7 +125,7 @@ fn check_store(agent: &str, f_type: std::io::Result<FsType>) -> Result<()> {
}
bail!(
"{agent} sets privateUsers, but {STORE} is on a filesystem (f_type {f_type:#x}) other \
than ext4, btrfs or xfs, the only ones accepted; move {STORE} to one of them, or leave \
than ext4, btrfs, xfs or erofs, the only ones accepted; move {STORE} to one of them, or leave \
privateUsers off for this agent"
)
}
@ -686,14 +689,14 @@ mod tests {
/// looks.
#[test]
fn the_store_has_to_be_on_a_filesystem_that_can_be_idmapped() {
for t in [0xEF53, 0x9123_683E, 0x5846_5342] {
for t in [0xEF53, 0x9123_683E, 0x5846_5342, 0xE0F5_E1E2] {
check_store("a", Ok(t)).unwrap();
}
// overlayfs and tmpfs
for t in [0x794c_7630, 0x0102_1994] {
let msg = format!("{:#}", check_store("a", Ok(t)).unwrap_err());
assert!(
msg.contains(&format!("{t:#x}")) && msg.contains("ext4, btrfs or xfs"),
msg.contains(&format!("{t:#x}")) && msg.contains("ext4, btrfs, xfs or erofs"),
"{msg}"
);
}