svn commit: r207140 - in user/kmacy/head_page_lock_2/sys: amd64/amd64 kern sys

[John Baldwin]

On Tuesday 27 April 2010 3:29:06 pm K. Macy wrote:
> On Tue, Apr 27, 2010 at 10:06 AM, John Baldwin <jhb at freebsd.org> wrote:
> > On Saturday 24 April 2010 1:03:42 am Kip Macy wrote:
> >> Author: kmacy
> >> Date: Sat Apr 24 05:03:42 2010
> >> New Revision: 207140
> >> URL: http://svn.freebsd.org/changeset/base/207140
> >>
> >> Log:
> >>   generalize lock stack operations and move to subr_lock.c
> >>
> >> Modified:
> >>   user/kmacy/head_page_lock_2/sys/amd64/amd64/pmap.c
> >>   user/kmacy/head_page_lock_2/sys/kern/kern_mutex.c
> >>   user/kmacy/head_page_lock_2/sys/kern/subr_lock.c
> >>   user/kmacy/head_page_lock_2/sys/sys/lock.h
> >>
> >> Modified: user/kmacy/head_page_lock_2/sys/kern/kern_mutex.c
> >>
> > ==============================================================================
> >> --- user/kmacy/head_page_lock_2/sys/kern/kern_mutex.c Sat Apr 24 03:11:35
> > 2010    (r207139)
> >> +++ user/kmacy/head_page_lock_2/sys/kern/kern_mutex.c Sat Apr 24 05:03:42
> > 2010    (r207140)
> >> @@ -160,7 +185,7 @@ unlock_mtx(struct lock_object *lock)
> >>       struct mtx *m;
> >>
> >>       m = (struct mtx *)lock;
> >> -     mtx_assert(m, MA_OWNED | MA_NOTRECURSED);
> >> +     mtx_assert(m, MA_OWNED);
> >>       mtx_unlock(m);
> >>       return (0);
> >>  }
> >
> > Note that this breaks the assertion used to prevent passing a recursed mutex
> > to mtx_sleep() or cv_wait().
> >
> 
> Thanks. MA_NOTRECURSED breaks its use in lock_stack, I'll add another
> function for it too.

Yes, the fault may be that lc_unlock and lc_lock are not truly generic.  They
were only really designed to handle unlock/lock pairs in mtx_sleep() and friends.

-- 
John Baldwin