Re: Sleeping function called from illegal context...

[Andrew Morton]

John Levon wrote:
> 
> On Fri, Sep 27, 2002 at 08:51:30PM -0400, Robert Love wrote:
> 
> > Note this has nothing to do with kernel preemption.  IDE explicitly
> > sleeps while purposely holding a lock.
> >
> > It is just we do not have the ability to measure atomicity w/o
> > preemption enabled - e.g. the debugging only works when it is enabled.
> 
> Would it be particularly difficult to separate this debug tool from the
> feature ? Surely we could make it so that CONFIG_PREEMPT depends on
> CONFIG_MIGHT_SLEEP or whatever, and just adds the actual ability to
> reschedule.

We need a standalone CONFIG_MIGHT_SLEEP.  I sinfully hooked it
to CONFIG_DEBUG_KERNEL (it's not obvious why CONFIG_DEBUG_KERNEL
exists actually).

So yes, you could make CONFIG_MIGHT_SLEEP mutually exclusive
with CONFIG_OPROFILE. But that would make people look at you
suspiciously.

> I have a bit of a problem with __might_sleep because I call sleepable
> stuff holding a spinlock (yes, it is justified, and yes, it is safe
> afaics, at least with PREEMPT=n)

I'm looking at you suspiciously.  How come?