Re: [patch 10/15] Generic Mutex Subsystem, mutex-migration-helper-core.patch

[Ingo Molnar <mingo@elte.hu>]

* Christoph Hellwig <hch@infradead.org> wrote:

> On Mon, Dec 19, 2005 at 02:38:37AM +0100, Ingo Molnar wrote:
> > 
> > introduce the mutex_debug type, and switch the semaphore APIs to it in a 
> > type-sensitive way. Plain semaphores will still use the proper 
> > arch-semaphore calls.
> 
> I think we shouldn't introduce this one.  It just encourages people to do
> really things.
        ^---("weird" ?)

ok. I can keep it a separate patch - it's needed for -rt, and it's also 
a good debugging tool to test-drive the mutex implementation.  

historically, i found 90% of the mutex.c bugs by stress-testing it from 
userspace, but a crutial 10% did come from the DEBUG_MUTEX_FULL mode, 
where almost all of the 7000+ semaphore users are switched over to 
mutexes.

but in terms of keeping the upstream kernel code base clean, i agree 
with a clean and separate set of mutex APIs that are introduced into 
subsystems gradually.

>  Everything else in the patchseries looks sensible to me.

great!

	Ingo