Re: [PATCH] cpuset semaphore depth check deadlock fix

[Paul Jackson <pj@sgi.com>]

Linus wrote:
> We _really_ don't want to have function names like "cs_up()" 

I thoroughly agree with your attention to naming, and spent more time
than I will admit in public futzing over this detail.

I wrote the code using cpuset_lock(void) and cpuset_unlock(void), for
reasons such as you state, and out of personnal instinct.

But then I noticed that I wanted these routines to replace up(&sem) and
down(&sem) (in kernel/cpuset.c), so changed them to cpuset_up(&sem) and
cpuset_down(&sem), adding in the explicitly passed argument.

But then I noticed that these names looked "too global" to me, and
intentionally changed that to cs_up(&sem) and cs_down(&sem).  I tend
to intentionally choose shorter names for more local stuff, especially
inlines and such that won't even show up on a stack trace.

 1) Is cpuset_up(&sem) and cpuset_down(&sem) ok by you?  I would like
    to have the up/down in there somewhere.

 2) How the heck do I make this change:
     - Send another patch from scratch, ignoring the first one I sent.
     - Send a second patch that layers on the first.
     - Let you do the edit.
     - ??

-- 
                  I won't rest till it's the best ...
                  Programmer, Linux Scalability
                  Paul Jackson <pj@sgi.com> 1.925.600.0401