Re: BKL removal

[William Lee Irwin III]

On Wed, Jul 10, 2002 at 12:03:08PM +0200, Marco Colombo wrote:
>> Larry, there's something I've always wanted to ask you about your
>> idea of the "locking cliff": when you're counting the number of locks,
>> are you looking at the running image of an OS or at the source? 

On Wed, Jul 10, 2002 at 03:40:03PM +0100, Matthew Wilcox wrote:
> Larry normally talks about the number of conceptual locks.  So in order
> to manipulate a `struct file', it really doesn't matter whether you have
> to grab the BKL, the files_struct lock or the filp->lock.  There's a big
> difference if you have to grab the filp->pos_lock, the filp->ra_lock and
> the filp->iobuf_lock.  You'd have to know what order to grab them in,
> for a start.

This is called "lock depth" and is not related to the total number of
locks declared in the source. AFAIK no one wants to increase lock depth.

Cheers,
Bill