Function try_flock 

pub fn try_flock<P: AsRef<Path>>(
    path: P,
    mode: FlockMode,
) -> Result<Option<OwnedFd>>

Open a lock file and attempt flock with LOCK_NB.

Creates the file with mode 0o666 if absent. Returns Ok(Some(fd)) on successful acquire, Ok(None) on EWOULDBLOCK (peer already holds an incompatible lock), and propagates other errors. The returned fd owns the open-file description; dropping it closes the fd AND releases the kernel flock (the kernel releases flock(2) only when the last fd referring to its OFD closes — OwnedFd::drop is what makes that work).

O_CLOEXEC is mandatory: a leaked fd across exec(2) (cargo subcommand, build-pipeline subprocess, initramfs compressor) would keep the lock alive in the child process after the parent’s OwnedFd::drop runs, producing phantom holders the next acquirer would blame on the wrong pid.

Calls super::fs_filter::reject_remote_fs before the open to fail-fast on NFS / CIFS / SMB2 / CEPH / AFS / FUSE — see the module-level rationale.

Accepts any AsRef<Path> so &str, &Path, &PathBuf, and String callers all work without string-ifying round trips. LLC lockfile paths are built as String via format! and cache lockfile paths are built as PathBuf via Path::join — both pass straight through.