fix(namespace): make recursive propagation changes atomic (#2112)

Prepare recursive propagation changes as an explicit transaction while holding the mount lifecycle lock. Snapshot the target subtree and all affected peer, master, and slave relationships, reserve fallible resources up front, simulate Linux propagation semantics privately, and publish only fully prepared final state.

Match Linux 6.6 behavior for DFS mount traversal, next-peer ring selection, exact mount-root identity, and slave-list rewiring. Keep group ID release allocation-free with IDA-backed hole tracking and avoid full-graph scans by maintaining direct slave adjacency.

Add deterministic failure-path coverage and DragonOS guest tests for recursive propagation semantics, atomic mountinfo snapshots, namespace copying, and concurrent topology changes.

Fixes #2098

Signed-off-by: longjin <longjin@dragonos.org>

Author: fslongjin

kernel/src/filesystem/vfs/mount/mod.rs

@@ -54,6 +54,11 @@ use system_error::SystemError;
 
 /// Serializes mount pin admission against multi-mount busy preflight and
 /// detach, including propagation peers in other namespaces.
+///
+/// Mount topology and propagation code acquires locks in this order:
+/// lifecycle -> namespace -> dentry mount gate -> parent mountpoints -> peer
+/// registry -> one mount's propagation state -> propagation group allocator.
+/// A lower layer must never acquire the lifecycle/topology layers in reverse.
 pub(crate) static MOUNT_LIFECYCLE_LOCK: Mutex<()> = Mutex::new(());
 
 lazy_static! {