Re: Oops from local semaphore race condition

[Bob Miller]

On Tue, May 21, 2002 at 03:41:19PM -0400, Ron Niles wrote:
> 
> 
> [e-mail] said:
> >> Then I realized it can possibly go corrupt, due to a race condition
> >> which lets down() continue before up() is complete:
> 
> From: David Woodhouse [mailto:[e-mail]]
> 
> >This is what completions were added for.
> 
> Thanks, struct completion is the best way; it's gonna be tough to maintain
> backward compatibility though.
> 
> One comment; it looks like the implementation in sched.c should more
> properly be using wq_write_lock_irqsave on the lock.
> 
I sent patches to Linus to fix this back in February.  Dave Jones picked
them up and they are still in his tree.  I don't know when/if he is going
to forward them onto Linus for inclusion into his tree.

-- 
Bob Miller					Email: [e-mail]
Open Source Development Lab			Phone: [phone] Ext. 17