Why can flock(1) do inter-process lock?

Question

flock is associated with open-file-description, and open(2) creates new open-file-description. So invoking flock(1) in two separate shells will open the lock file separately and result in two open-file-descriptions. How could flock(1) know the lock status associated with each open-file-description? I am confused... Also, does flock(1) treat path name and file descriptor differently?

UPDATE:

The flock I am talking about is the one under Linux.

In my understanding, the struct representing flock is system-wide unique. One or more open-file-descriptions may reference to one flock struct (like the relationship between open-file-description and i-node). Is my 'guess' correct?

Answer 1

No, flock, which is a wrapper around the system call flock(), locks on files, not file descriptors.

The OpenBSD manual for flock() says (my emphasis):

flock() applies or removes an advisory lock on the file associated with the file descriptor fd.

[...]

Locks are on files, not file descriptors. That is, file descriptors duplicated through dup(2) or fork(2) do not result in multiple instances of a lock, but rather multiple references to a single lock. If a process holding a lock on a file forks and the child explicitly unlocks the file, the parent will lose its lock.

flock() is a system call; the kernel keeps track of locks.