Re: xfs vs. lockdep

[Eric Sandeen <sandeen@xxxxxxxxxxx>]

David Chinner wrote:

> FWIW, we call mrfree() on both the ilock and the iolock, but these are
> #defined to null statements. If there is a destructor for the underlying
> lock type, we probably should call that in mrfree() so the debugging code
> can catch these probelms that only trigger in debug code.

From a quick look I don't see those destructors, might be good to write some 
though.

It'd also be good to turn our spinlock_destroy(lock) into a 
WARN_ON(spin_is_locked(lock)) or something...

it'd be extra nice if lockdep could grok that the lock it's looking at is full 
of free poison, print a warning, take it off the list and move on... this was a 
bear to figure out which one was causing the problem.

I have a few changes to get more desructors called for some of the newer locks 
that are missing them (think agirotor_lock...) that I'll send when I get a moment.

-Eric

> Cheers,
>
> Dave.