Re: [fuse-devel] Posix locking

[Dominick L. <dom...@gm...>]

Let me ask another question about traditional, Posix-style record locks.

For a F_SETLK, the PID of the process requesting the lock is passed to the 
fuse_operations.lock() function in the l_pid field of the struct flock.

My problem is that for both a regular F_UNLCK call from fcntl(), *and* for the 
F_UNLCK triggered by a close(), the l_pid field is zero.

i.e. I don't get any process ID information for an unlock.

Is this a bug?  Or are unlocks supposed to work that way?

In order to propagate the locks across the network, I need to keep track of both 
PID and node information in order to identify locks uniquely: I can't allow an 
unlock from node A to remove locks from node B.

-- Dominick