Re: [PATCH, RFC] xfs: remove i_iolock and use i_rwsem in the VFS inode instead

[Christoph Hellwig <hch@xxxxxx>]

On Thu, Sep 08, 2016 at 07:45:36AM +1000, Dave Chinner wrote:
> It's a semaphore, not a mutex. Semaphore locking is independent of
> task context, the lock follows the object it protects, not the task
> that took the lock. i.e. Lockdep is wrong to assume the "owner" of a
> rw_sem will not change between lock and unlock.

That's not the case - rw_semaphores had strict owner semanics for a
long time (although I wish we could get rid of that for a different
reason..).

The problem here is not that we have different tasks acquire and release
the lock - it's always the same.

The "problem" is that that we hand off work to a different task inbetween
and that task asserts that the lock is held.  With the old mrlock hack
our islocked macros would return true as long as _someone_ holds the
lock, while lockdep is generally more strict and wants the current
process to hold the lock.