On Wed, Feb 12, 2025 at 02:45:04PM +1100, NeilBrown wrote: > On Wed, 12 Feb 2025, Al Viro wrote: > I do see some value in the simplicity of this approach, though maybe not > as much value as you see. But the above uses inode_lock_share(), rather > than the nested version, so lockdep will complain. IDGI... It doesn't grab any ->i_rwsem inside that one, so what's there to complain about? And in that case it returns with no locks held, so...
