Re: [QUESTION] How to use interruptible_sleep_on() without races ?

[torvalds@transmeta.com (Linus Torvalds)]

In article <20020404190848.C27209@bougret.hpl.hp.com>,
Jean Tourrilhes  <jt@bougret.hpl.hp.com> wrote:
>On Fri, Apr 05, 2002 at 04:20:04AM +0100, Alan Cox wrote:
>> > 
>> > 	I looked at it in every possible way, and I don't see how it
>> > is possible to use safely interruptible_sleep_on(). And I wonder :
>> 
>> It isnt for interrupt stuff - its going back to the old kernel behaviour
>> when it used to be usable
>
>	So, maybe it would be a nice idea to remove it from the 2.5.X
>kernel to force a "spring cleanup" of the old code. If it's no longer
>usable and only confusing, it should be purged...

It's still usable, but under rather specific conditions, namely:
 - both sleeper and waker in process context and with BKL held.
 - OR if missing a wakeup isn't a horrible problem.

And there does seem to be a lot of legacy users out there.

I wouldn't mind a spring cleaning, but the fact is that right now in
2.5.x I'd rather have driver writers wake up to the fact that we had a
spring cleaning in the block layer several months ago, rather than
introduce a new one ;)

		Linus