refcount_release_take_##lock

[John Baldwin]

On Tuesday, October 28, 2014 3:34:04 pm Mateusz Guzik wrote:
> On Tue, Oct 28, 2014 at 02:13:58PM -0400, John Baldwin wrote:
> > On Tuesday, October 28, 2014 1:44:28 pm Mateusz Guzik wrote:
> > > diff --git a/sys/kern/kern_jail.c b/sys/kern/kern_jail.c
> > > index f8ae0e6..e94ccde 100644
> > > --- a/sys/kern/kern_jail.c
> > > +++ b/sys/kern/kern_jail.c
> > 
> > The diff looks good to me.  Just need to update refcount.9 as well.
> > 
> 
> diff --git a/share/man/man9/refcount.9 b/share/man/man9/refcount.9
> index e7702a2..61b9b51 100644

FYI, it's often easier to review manpage changes if a rendered version is
included.

> --- a/share/man/man9/refcount.9
> +++ b/share/man/man9/refcount.9
> @@ -26,7 +26,7 @@
>  .\"
>  .\" $FreeBSD$
>  .\"
> -.Dd January 20, 2009
> +.Dd October 28, 2014
>  .Dt REFCOUNT 9
>  .Os
>  .Sh NAME

Need to add the new functions to the .Nm list

> @@ -44,6 +44,15 @@
>  .Fn refcount_acquire "volatile u_int *count"
>  .Ft int
>  .Fn refcount_release "volatile u_int *count"
> +.In sys/mutex.h
> +.Fn refcount_release_lock_mtx "volatile u_int *count, struct mtx *lock"
> +.In sys/rmlock.h
> +.Fn refcount_release_lock_rmlock "volatile u_int *count, struct rmlock *lock"
> +.In sys/rwlock.h
> +.Fn refcount_release_lock_rwlock "volatile u_int *count, struct rwlock *lock"
> +.In sys/lock.h
> +.In sys/sx.h
> +.Fn refcount_release_lock_sx "volatile u_int *count, struct sx *lock"

Hmm, <sys/lock.h> should also be required for the other three and not just sx?

>  .Sh DESCRIPTION
>  The
>  .Nm
> @@ -77,6 +86,13 @@ The function returns a non-zero value if the reference being released was
>  the last reference;
>  otherwise, it returns zero.
>  .Pp
> +.Fn refcount_release_lock_*

Even though it is tedious, most manpages tend to spell out the list of functions
(e.g. I did so in callout recently with all the callout_init_*() variants, etc.).
I believe I broke this rule in atomic(9), but I think if the list of functions
isn't too unreasonable, it's better to explicitly list them (and I think 4 functions
isn't unreasonable).

> +functions release an existing reference holding the lock if it is the last

Perhaps:

functions release an existing reference similar to
.Fn refcount_release .
If the reference being released is the last reference,
the
.Fa lock
is exclusively acquired before the reference is dropped and remains held when
the function returns.
These functions return a non-zero value if the reference being released was the
last reference;
otherwise,
they return zero.

> @@ -91,6 +107,18 @@ The
>  .Nm refcount_release
>  function returns non-zero when releasing the last reference and zero when
>  releasing any other reference.
> +.Pp
> +.Nm refcount_release_lock_*
> +functions return with the lock held and non-zero value when releasing the last
> +reference, zero without the lock held when releasing any other reference.

This this is just about return values, I would instead just update the previous
sentence to add the additional functions.  I.e.:

The
.Fn refcount_release ,   # this is probably my fault, should be Fn
.Fn refcount_release_lock_mtx ,
...
and
.Fn refcount_release_lock sx
functions return non-zero ...

>  .Sh HISTORY
> -These functions were introduced in

The
.Fn refcount_init ,
...

> +.Fn refcount_init ,
> +.Fn refcount_acquire
> +and
> +.Fn refcount_release
> +functions were introduced in
>  .Fx 6.0 .
> +.Pp
> +.Fn refcount_release_lock_*
> +functions were introduced in
> +.Fx 10.2 .

.Pp
The
.Fn refcount_release_lock_mtx ,
.Fn refcount_release_lock_rmlock ,
...

-- 
John Baldwin