[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20121015224044.GA9788@linux.vnet.ibm.com>
Date: Mon, 15 Oct 2012 15:40:44 -0700
From: "Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>
To: Lai Jiangshan <laijs@...fujitsu.com>
Cc: linux-kernel@...r.kernel.org
Subject: Re: [PATCH 3/4] srcu: add DEFINE_SRCU()
On Mon, Oct 15, 2012 at 02:56:07PM -0700, Paul E. McKenney wrote:
> On Sat, Oct 13, 2012 at 01:14:16AM +0800, Lai Jiangshan wrote:
> > In old days, we have two different API sets for dynamic-allocated per_cpu data
> > and DEFINE_PER_CPU()-defined per_cpu data, and since we used
> > dynamic-allocated per_cpu data, we can't use DEFINE_PER_CPU()-defined
> > per_cpu data(otherwise we will introduce a lot of duplicated code.
> >
> > In new days, we have only one API sets for both type of per_cpu data,
> > so we can use DEFINE_PER_CPU() for DEFINE_SRCU() which allows us
> > define and init srcu struct in build time and allows us use srcu APIs
> > in very early boot time.
> >
> > We also provide DEFINE_STATIC_SRCU() which defines an internal srcu struct
> > inside a single *.c.
> >
> > Signed-off-by: Lai Jiangshan <laijs@...fujitsu.com>
> > ---
> > include/linux/srcu.h | 30 ++++++++++++++++++++++++++++++
> > 1 files changed, 30 insertions(+), 0 deletions(-)
> >
> > diff --git a/include/linux/srcu.h b/include/linux/srcu.h
> > index 5cce128..f986df1 100644
> > --- a/include/linux/srcu.h
> > +++ b/include/linux/srcu.h
> > @@ -42,6 +42,8 @@ struct rcu_batch {
> > struct rcu_head *head, **tail;
> > };
> >
> > +#define RCU_BATCH_INIT(name) { NULL, &(name.head) }
> > +
> > struct srcu_struct {
> > unsigned completed;
> > struct srcu_struct_array __percpu *per_cpu_ref;
> > @@ -72,14 +74,42 @@ int __init_srcu_struct(struct srcu_struct *sp, const char *name,
> > __init_srcu_struct((sp), #sp, &__srcu_key); \
> > })
> >
> > +#define __SRCU_DEP_MAP_INIT(srcu_name) .dep_map = { .name = #srcu_name },
> > #else /* #ifdef CONFIG_DEBUG_LOCK_ALLOC */
> >
> > int init_srcu_struct(struct srcu_struct *sp);
> >
> > +#define __SRCU_DEP_MAP_INIT(srcu_name)
> > #endif /* #else #ifdef CONFIG_DEBUG_LOCK_ALLOC */
> >
> > void process_srcu(struct work_struct *work);
> >
> > +#define __SRCU_STRUCT_INIT(name) \
> > + { \
> > + .completed = -300, \
> > + .per_cpu_ref = &name##_srcu_array, \
> > + .queue_lock = __SPIN_LOCK_UNLOCKED(name.queue_lock), \
> > + .running = false, \
> > + .batch_queue = RCU_BATCH_INIT(name.batch_queue), \
> > + .batch_check0 = RCU_BATCH_INIT(name.batch_check0), \
> > + .batch_check1 = RCU_BATCH_INIT(name.batch_check1), \
> > + .batch_done = RCU_BATCH_INIT(name.batch_done), \
> > + .work = __DELAYED_WORK_INITIALIZER(name.work, process_srcu),\
> > + __SRCU_DEP_MAP_INIT(name) \
> > + }
> > +
> > +/*
> > + * define and init a srcu struct at build time.
> > + * dont't call init_srcu_struct() nor cleanup_srcu_struct() on it.
> > + */
> > +#define DEFINE_SRCU(name) \
> > + static DEFINE_PER_CPU(struct srcu_struct_array, name##_srcu_array);\
> > + struct srcu_struct name = __SRCU_STRUCT_INIT(name);
> > +
> > +#define DEFINE_STATIC_SRCU(name) \
> > + static DEFINE_PER_CPU(struct srcu_struct_array, name##_srcu_array);\
> > + static struct srcu_struct name = __SRCU_STRUCT_INIT(name);
> > +
>
> Why not collapse DEFINE_SRCU() and DEFINE_STATIC_SRCU() as follows?
>
> +#define DEFINE_SRCU(name) \
> + struct srcu_struct name = __SRCU_STRUCT_INIT(name); \
> + static DEFINE_PER_CPU(struct srcu_struct_array, name##_srcu_array);
>
> Then "static DEFINE_SRCU(foo)" would create a static srcu_struct without
> the need for a separate DEFINE_STATIC_SRCU().
Never mind!!! This of course results in a forward reference and a
compiler warning...
Thanx, Paul
> > /**
> > * call_srcu() - Queue a callback for invocation after an SRCU grace period
> > * @sp: srcu_struct in queue the callback
> > --
> > 1.7.7.6
> >
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists