Re: [PATCH 5/5] futex: fix miss ordered wakeups

[Daniel Walker <dwalker@mvista.com>]

On Fri, 2008-06-13 at 00:43 +0200, Thomas Gleixner wrote:
> On Thu, 12 Jun 2008, Daniel Walker wrote:
> 
> > 
> > On Thu, 2008-06-12 at 21:55 +0200, Thomas Gleixner wrote:
> > > Also your interpretation of the POSIX requirement is very
> > > questionable:
> > > 
> > >  "If there are threads blocked on the mutex object referenced by mutex
> > >  when pthread_mutex_unlock() is called, resulting in the mutex
> > >  becoming available, the scheduling policy shall determine which
> > >  thread shall acquire the mutex."
> > 
> > The key is "scheduling policy" .. When the mutex is un-blocked the next
> > task to run is the same as if the scheduler was selecting tasks from the
> > list of blocked tasks .. For Linux, that means the highest priority
> > tasks should be selected.. So it's no more acceptable for the scheduler
> > to priority invert some tasks than it is for the futex to do it.
> 
> Sigh, when do you actually get a gripe that the default futex
> implementation does not and can not guarantee that at all and therefor
> your "correctness" patch is as important as a bag of rice which
> toopled over in China ?

Well, the last email I got from Arjan said this,

".. Don't look at the release path... look at the acquire path.
If a thread sees the futex is free, it'll take it, without even going
to the kernel at all."

And yes, I understand that fully.

> Provide answers to the real questions I asked more than once: 
> 
> What's the real world problem ? Who cares about that - except you ?

Any application which starts a thread, and later changes the priority
can observe the miss-ordering.. That's pretty common..

Daniel