Re: [PATCH] Add kerneldoc for flush_scheduled_work()

[James Bottomley <James.Bottomley@HansenPartnership.com>]

On Wed, 2009-08-12 at 10:54 -0400, Alan Stern wrote:
> > > + * Consider using cancel_work_sync() or cancel_delayed_work_sync() instead.
> > > + * They don't do the same thing (they cancel the work instead of waiting
> > > + * for it to complete), but in most cases they will suffice.
> > > + */
> > 
> > And this is wrong advice.  If you've violated the entangled deadlock
> > rules, the cancel functions will deadlock on you as well if the work is
> > still pending.
> 
> No they won't.  They will remove the work item from the workqueue right
> away, without blocking, assuming it hasn't started yet (i.e., is still 
> pending).  This is a large part of their raison d'etre.

Yes, they will ... you advised the _sync function which waits if the
work is in progress and hence induces the entanglement.  You can get
away with this if you don't use _sync (but then you won't know when the
queue is safely not touching any module code before removal).

James