* [PATCH 1/2] sched: Introduce APIs for waiting multi events
@ 2009-04-15 3:18 Lai Jiangshan
2009-04-17 2:59 ` Paul E. McKenney
0 siblings, 1 reply; 2+ messages in thread
From: Lai Jiangshan @ 2009-04-15 3:18 UTC (permalink / raw)
To: Ingo Molnar, Paul E. McKenney, LKML
Impact: make kernel has APIs for waiting multi events
Based on code in the RCU, the simple waiting-multi-events
algorithm may do not serve RCU only.
Introduce 5 APIs for waiting multi events
start scheduling events: ref_completion_get_init()
schedule one event: ref_completion_get()
end scheduling events: ref_completion_put_init()
any event is finished: ref_completion_put()
waiting scheduled events: ref_completion_wait()
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
---
diff --git a/include/linux/completion.h b/include/linux/completion.h
index 4a6b604..86c8236 100644
--- a/include/linux/completion.h
+++ b/include/linux/completion.h
@@ -8,6 +8,7 @@
* See kernel/sched.c for details.
*/
+#include <asm/atomic.h>
#include <linux/wait.h>
/**
@@ -98,5 +99,44 @@ extern void complete_all(struct completion *);
*/
#define INIT_COMPLETION(x) ((x).done = 0)
+/*
+ * wait all references are put, or wait multi events finished in other words
+ * ref_completion_get_init() called before the struct is used or reused.
+ * ref_completion_put_init() called after all ref_completion_get()s called.
+ * ref_completion_wait() waiting until all ref_completion_put()s and
+ * the ref_completion_put_init() are called.
+ */
+struct ref_completion {
+ atomic_t ref;
+ struct completion wait;
+};
+
+static inline void ref_completion_get_init(struct ref_completion *rc)
+{
+ atomic_set(&rc->ref, 1);
+ init_completion(&rc->wait);
+}
+
+static inline void ref_completion_get(struct ref_completion *rc)
+{
+ atomic_inc(&rc->ref);
+}
+
+static inline void ref_completion_put(struct ref_completion *rc)
+{
+ if (atomic_dec_and_test(&rc->ref))
+ complete_all(&rc->wait);
+}
+
+static inline void ref_completion_put_init(struct ref_completion *rc)
+{
+ if (atomic_dec_and_test(&rc->ref))
+ complete_all(&rc->wait);
+}
+
+static inline void ref_completion_wait(struct ref_completion *rc)
+{
+ wait_for_completion(&rc->wait);
+}
#endif
^ permalink raw reply related [flat|nested] 2+ messages in thread
* Re: [PATCH 1/2] sched: Introduce APIs for waiting multi events
2009-04-15 3:18 [PATCH 1/2] sched: Introduce APIs for waiting multi events Lai Jiangshan
@ 2009-04-17 2:59 ` Paul E. McKenney
0 siblings, 0 replies; 2+ messages in thread
From: Paul E. McKenney @ 2009-04-17 2:59 UTC (permalink / raw)
To: Lai Jiangshan; +Cc: Ingo Molnar, LKML
On Wed, Apr 15, 2009 at 11:18:48AM +0800, Lai Jiangshan wrote:
>
> Impact: make kernel has APIs for waiting multi events
>
> Based on code in the RCU, the simple waiting-multi-events
> algorithm may do not serve RCU only.
>
> Introduce 5 APIs for waiting multi events
>
> start scheduling events: ref_completion_get_init()
> schedule one event: ref_completion_get()
> end scheduling events: ref_completion_put_init()
>
> any event is finished: ref_completion_put()
>
> waiting scheduled events: ref_completion_wait()
Interesting API! What other uses for it have you found?
One question down below.
Thanx, Paul
> Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
> ---
> diff --git a/include/linux/completion.h b/include/linux/completion.h
> index 4a6b604..86c8236 100644
> --- a/include/linux/completion.h
> +++ b/include/linux/completion.h
> @@ -8,6 +8,7 @@
> * See kernel/sched.c for details.
> */
>
> +#include <asm/atomic.h>
> #include <linux/wait.h>
>
> /**
> @@ -98,5 +99,44 @@ extern void complete_all(struct completion *);
> */
> #define INIT_COMPLETION(x) ((x).done = 0)
>
> +/*
> + * wait all references are put, or wait multi events finished in other words
> + * ref_completion_get_init() called before the struct is used or reused.
> + * ref_completion_put_init() called after all ref_completion_get()s called.
> + * ref_completion_wait() waiting until all ref_completion_put()s and
> + * the ref_completion_put_init() are called.
> + */
> +struct ref_completion {
> + atomic_t ref;
> + struct completion wait;
> +};
> +
> +static inline void ref_completion_get_init(struct ref_completion *rc)
> +{
> + atomic_set(&rc->ref, 1);
> + init_completion(&rc->wait);
> +}
> +
> +static inline void ref_completion_get(struct ref_completion *rc)
> +{
> + atomic_inc(&rc->ref);
> +}
> +
> +static inline void ref_completion_put(struct ref_completion *rc)
> +{
> + if (atomic_dec_and_test(&rc->ref))
> + complete_all(&rc->wait);
> +}
> +
> +static inline void ref_completion_put_init(struct ref_completion *rc)
> +{
> + if (atomic_dec_and_test(&rc->ref))
> + complete_all(&rc->wait);
> +}
Looks to me that ref_completion_put() and ref_completion_put_init() are
identical. What is the story with that?
> +static inline void ref_completion_wait(struct ref_completion *rc)
> +{
> + wait_for_completion(&rc->wait);
> +}
>
> #endif
>
>
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2009-04-17 3:00 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2009-04-15 3:18 [PATCH 1/2] sched: Introduce APIs for waiting multi events Lai Jiangshan
2009-04-17 2:59 ` Paul E. McKenney
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox