* [PATCH] RT: update rcurefs for RT
@ 2005-09-27 18:32 Daniel Walker
2005-09-28 9:32 ` Ingo Molnar
2005-09-29 11:42 ` Ingo Molnar
0 siblings, 2 replies; 7+ messages in thread
From: Daniel Walker @ 2005-09-27 18:32 UTC (permalink / raw)
To: mingo; +Cc: linux-kernel
Make rcurefs compatible with RT w/o cmpxchg() .
Signed-Off-By: Daniel Walker <dwalker@mvista.com>
Index: linux-2.6.13/kernel/rcupdate.c
===================================================================
--- linux-2.6.13.orig/kernel/rcupdate.c
+++ linux-2.6.13/kernel/rcupdate.c
@@ -96,6 +96,25 @@ static void rcu_torture_init(void);
static inline void rcu_torture_init(void) { }
#endif
+#ifndef __HAVE_ARCH_CMPXCHG
+/*
+ * We use an array of spinlocks for the rcurefs -- similar to ones in sparc
+ * 32 bit atomic_t implementations, and a hash function similar to that
+ * for our refcounting needs.
+ * Can't help multiprocessors which donot have cmpxchg :(
+ */
+spinlock_t __rcuref_hash[RCUREF_HASH_SIZE];
+
+static inline void init_rcurefs(void)
+{
+ int i;
+ for (i=0; i < RCUREF_HASH_SIZE; i++)
+ SPIN_LOCK_UNLOCKED(__rcuref_hash[i]);
+}
+#else
+#define init_rcurefs() do { } while (0)
+#endif
+
#ifndef CONFIG_PREEMPT_RCU
/* Definition for rcupdate control block. */
@@ -123,18 +142,6 @@ DEFINE_PER_CPU(struct rcu_data, rcu_bh_d
static DEFINE_PER_CPU(struct tasklet_struct, rcu_tasklet) = {NULL};
static int maxbatch = 10;
-#ifndef __HAVE_ARCH_CMPXCHG
-/*
- * We use an array of spinlocks for the rcurefs -- similar to ones in sparc
- * 32 bit atomic_t implementations, and a hash function similar to that
- * for our refcounting needs.
- * Can't help multiprocessors which donot have cmpxchg :(
- */
-
-spinlock_t __rcuref_hash[RCUREF_HASH_SIZE] = {
- [0 ... (RCUREF_HASH_SIZE-1)] = SPIN_LOCK_UNLOCKED
-};
-#endif
/**
* call_rcu - Queue an RCU callback for invocation after a grace period.
@@ -487,6 +494,7 @@ static struct notifier_block __devinitda
*/
void __init rcu_init(void)
{
+ init_rcurefs();
rcu_torture_init();
rcu_cpu_notify(&rcu_nb, CPU_UP_PREPARE,
(void *)(long)smp_processor_id());
@@ -824,6 +832,7 @@ rcu_pending(int cpu)
void __init rcu_init(void)
{
+ init_rcurefs();
rcu_torture_init();
/*&&&&*/printk("WARNING: experimental RCU implementation.\n");
spin_lock_init(&rcu_data.lock);
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] RT: update rcurefs for RT
2005-09-27 18:32 [PATCH] RT: update rcurefs for RT Daniel Walker
@ 2005-09-28 9:32 ` Ingo Molnar
2005-09-29 11:42 ` Ingo Molnar
1 sibling, 0 replies; 7+ messages in thread
From: Ingo Molnar @ 2005-09-28 9:32 UTC (permalink / raw)
To: Daniel Walker; +Cc: linux-kernel
* Daniel Walker <dwalker@mvista.com> wrote:
> Make rcurefs compatible with RT w/o cmpxchg() .
thanks, applied.
Ingo
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] RT: update rcurefs for RT
2005-09-27 18:32 [PATCH] RT: update rcurefs for RT Daniel Walker
2005-09-28 9:32 ` Ingo Molnar
@ 2005-09-29 11:42 ` Ingo Molnar
2005-09-29 15:20 ` Daniel Walker
1 sibling, 1 reply; 7+ messages in thread
From: Ingo Molnar @ 2005-09-29 11:42 UTC (permalink / raw)
To: Daniel Walker; +Cc: linux-kernel
* Daniel Walker <dwalker@mvista.com> wrote:
> Make rcurefs compatible with RT w/o cmpxchg() .
> +static inline void init_rcurefs(void)
> +{
> + int i;
> + for (i=0; i < RCUREF_HASH_SIZE; i++)
> + SPIN_LOCK_UNLOCKED(__rcuref_hash[i]);
what the heck is this doing??? Patch reverted.
Ingo
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] RT: update rcurefs for RT
2005-09-29 11:42 ` Ingo Molnar
@ 2005-09-29 15:20 ` Daniel Walker
2005-09-30 0:55 ` Thomas Gleixner
0 siblings, 1 reply; 7+ messages in thread
From: Daniel Walker @ 2005-09-29 15:20 UTC (permalink / raw)
To: Ingo Molnar; +Cc: linux-kernel
On Thu, 2005-09-29 at 13:42 +0200, Ingo Molnar wrote:
> > + SPIN_LOCK_UNLOCKED(__rcuref_hash[i]);
>
> what the heck is this doing??? Patch reverted.
How is this ?
Index: linux-2.6.13/kernel/rcupdate.c
===================================================================
--- linux-2.6.13.orig/kernel/rcupdate.c
+++ linux-2.6.13/kernel/rcupdate.c
@@ -96,6 +96,25 @@ static void rcu_torture_init(void);
static inline void rcu_torture_init(void) { }
#endif
+#ifndef __HAVE_ARCH_CMPXCHG
+/*
+ * We use an array of spinlocks for the rcurefs -- similar to ones in sparc
+ * 32 bit atomic_t implementations, and a hash function similar to that
+ * for our refcounting needs.
+ * Can't help multiprocessors which donot have cmpxchg :(
+ */
+spinlock_t __rcuref_hash[RCUREF_HASH_SIZE];
+
+static inline void init_rcurefs(void)
+{
+ int i;
+ for (i=0; i < RCUREF_HASH_SIZE; i++)
+ __rcuref_hash[i] = SPIN_LOCK_UNLOCKED(__rcuref_hash[i]);
+}
+#else
+#define init_rcurefs() do { } while (0)
+#endif
+
#ifndef CONFIG_PREEMPT_RCU
/* Definition for rcupdate control block. */
@@ -123,18 +142,6 @@ DEFINE_PER_CPU(struct rcu_data, rcu_bh_d
static DEFINE_PER_CPU(struct tasklet_struct, rcu_tasklet) = {NULL};
static int maxbatch = 10;
-#ifndef __HAVE_ARCH_CMPXCHG
-/*
- * We use an array of spinlocks for the rcurefs -- similar to ones in sparc
- * 32 bit atomic_t implementations, and a hash function similar to that
- * for our refcounting needs.
- * Can't help multiprocessors which donot have cmpxchg :(
- */
-
-spinlock_t __rcuref_hash[RCUREF_HASH_SIZE] = {
- [0 ... (RCUREF_HASH_SIZE-1)] = SPIN_LOCK_UNLOCKED
-};
-#endif
/**
* call_rcu - Queue an RCU callback for invocation after a grace period.
@@ -487,6 +494,7 @@ static struct notifier_block __devinitda
*/
void __init rcu_init(void)
{
+ init_rcurefs();
rcu_torture_init();
rcu_cpu_notify(&rcu_nb, CPU_UP_PREPARE,
(void *)(long)smp_processor_id());
@@ -824,6 +832,7 @@ rcu_pending(int cpu)
void __init rcu_init(void)
{
+ init_rcurefs();
rcu_torture_init();
/*&&&&*/printk("WARNING: experimental RCU implementation.\n");
spin_lock_init(&rcu_data.lock);
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] RT: update rcurefs for RT
2005-09-29 15:20 ` Daniel Walker
@ 2005-09-30 0:55 ` Thomas Gleixner
2005-09-30 8:34 ` Ingo Molnar
0 siblings, 1 reply; 7+ messages in thread
From: Thomas Gleixner @ 2005-09-30 0:55 UTC (permalink / raw)
To: Daniel Walker; +Cc: Ingo Molnar, linux-kernel
On Thu, 2005-09-29 at 08:20 -0700, Daniel Walker wrote:
> +static inline void init_rcurefs(void)
> +{
> + int i;
> + for (i=0; i < RCUREF_HASH_SIZE; i++)
> + __rcuref_hash[i] = SPIN_LOCK_UNLOCKED(__rcuref_hash[i]);
Maybe a simple
spin_lock_init(&__rcuref_hash[i]);
would work all over tha place ?
tglx
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] RT: update rcurefs for RT
2005-09-30 0:55 ` Thomas Gleixner
@ 2005-09-30 8:34 ` Ingo Molnar
2005-09-30 14:17 ` Daniel Walker
0 siblings, 1 reply; 7+ messages in thread
From: Ingo Molnar @ 2005-09-30 8:34 UTC (permalink / raw)
To: Thomas Gleixner; +Cc: Daniel Walker, linux-kernel
* Thomas Gleixner <tglx@linutronix.de> wrote:
> On Thu, 2005-09-29 at 08:20 -0700, Daniel Walker wrote:
> > +static inline void init_rcurefs(void)
> > +{
> > + int i;
> > + for (i=0; i < RCUREF_HASH_SIZE; i++)
> > + __rcuref_hash[i] = SPIN_LOCK_UNLOCKED(__rcuref_hash[i]);
>
> Maybe a simple
>
> spin_lock_init(&__rcuref_hash[i]);
>
> would work all over the place ?
yep.
Ingo
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] RT: update rcurefs for RT
2005-09-30 8:34 ` Ingo Molnar
@ 2005-09-30 14:17 ` Daniel Walker
0 siblings, 0 replies; 7+ messages in thread
From: Daniel Walker @ 2005-09-30 14:17 UTC (permalink / raw)
To: Ingo Molnar; +Cc: Thomas Gleixner, linux-kernel
On Fri, 2005-09-30 at 10:34 +0200, Ingo Molnar wrote:
> * Thomas Gleixner <tglx@linutronix.de> wrote:
>
> > On Thu, 2005-09-29 at 08:20 -0700, Daniel Walker wrote:
> > > +static inline void init_rcurefs(void)
> > > +{
> > > + int i;
> > > + for (i=0; i < RCUREF_HASH_SIZE; i++)
> > > + __rcuref_hash[i] = SPIN_LOCK_UNLOCKED(__rcuref_hash[i]);
> >
> > Maybe a simple
> >
> > spin_lock_init(&__rcuref_hash[i]);
> >
> > would work all over the place ?
>
> yep.
The patch isn't needed if you accept Nick's cmpxchg() patches ..
Daniel
^ permalink raw reply [flat|nested] 7+ messages in thread
end of thread, other threads:[~2005-09-30 14:17 UTC | newest]
Thread overview: 7+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2005-09-27 18:32 [PATCH] RT: update rcurefs for RT Daniel Walker
2005-09-28 9:32 ` Ingo Molnar
2005-09-29 11:42 ` Ingo Molnar
2005-09-29 15:20 ` Daniel Walker
2005-09-30 0:55 ` Thomas Gleixner
2005-09-30 8:34 ` Ingo Molnar
2005-09-30 14:17 ` Daniel Walker
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox