[Ksummit-2008-discuss] topic proposal

[Andi Kleen]

Daniel Walker wrote:
> Semaphores do we keep them or do we remove them?
> 
> I've had enough semaphore removal patches accepted 

My understanding on the area is: It's generally agreed that semaphores
with mutex semantics should replaced with mutexes, unless there's another
strong reason against it (which applies to ACPI IMHO). That is because
mutexes have better behaviour and can be lockdep checked and support PI
in RT kernels.

What I think is not agreed is that semaphores with non mutex semantics
should be forcibly converted to something else. None of the earlier
arguments apply in this case. e.g. for your completion case:
neither is there lockdep for completions, nor do they support PI, nor
do they have better scheduling behaviour to my knowledge.

To be honest I don't think the topic needs much discussion.

-Andi