smb: client: introduce close_cached_dir_locked()

[smfrench]

Replace close_cached_dir() calls under cfid_list_lock with a new close_cached_dir_locked() variant that uses kref_put() instead of kref_put_lock() to avoid recursive locking when dropping references.

While the existing code works if the refcount >= 2 invariant holds, this area has proven error-prone. Make deadlocks impossible and WARN on invariant violations.

Cc: [email protected]

[Copilot]

Pull Request Overview

This PR introduces a new close_cached_dir_locked() function to safely drop cached directory references while holding the cfid_list_lock spinlock. The change prevents potential recursive locking deadlocks by using kref_put() instead of kref_put_lock() when the lock is already held.

• Adds close_cached_dir_locked() variant that expects cfid_list_lock to be held and validates refcount >= 2 before decrementing
• Updates two call sites to use the new locked variant where the spinlock is already held
• Adds comprehensive documentation and lockdep assertions to prevent misuse

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The documentation states "Must not be called with cfid_list_lock held; use close_cached_dir_locked() called instead." The word "called" appears to be redundant or incorrectly placed. It should be either "use close_cached_dir_locked() instead" or "call close_cached_dir_locked() instead".

Suggested change

	* close_cached_dir_locked() called instead.
	* close_cached_dir_locked() instead.

[Copilot]

[nitpick] The documentation could be clearer about the safety mechanism. While it mentions that calling with refcount < 2 would cause a deadlock, it should explicitly state that when refcount >= 2, calling kref_put() is safe because the refcount will remain >= 1 after decrement, ensuring smb2_close_cached_fid() (the release function) is not invoked while holding the lock.

Suggested change

	*
	*
	* When refcount >= 2, calling kref_put() is safe because the refcount will remain
	* >= 1 after decrement, ensuring smb2_close_cached_fid() (the release function)
	* is not invoked while holding the lock.
	*