Re: Memory barrier question.

[Tetsuo Handa]

Hello.

Eric Dumazet wrote:
> Le samedi 30 octobre 2010 a 14:48 +0900, Tetsuo Handa a ecrit :
> > Then, what about list heads that are dynamically allocated/initialized?
> > 
> > sruct hashed_word {
> > 	struct list_head list;
> > 	void *key;
> > 	char *buf;
> > };
> > static struct list_head tables[1u << TABLEBIT];
> > static void (*some_callback) (void *);
> > 
> 
> typedef void (*some_callback_t) (void *);
> 
> static some_callback_t __rcu *some_callback;
> 
> > static void init(void)
> > {
> > 	int i;
> > 	/* Initialize the table. */
> > 	for (i = 0; i < (1u << TABLEBIT); i++)
> > 		INIT_LIST_HEAD(&tables[i]);
> > 	smp_wmb();
> > 	/* Allow other CPUs to access the table. */
> > 	some_callback = func;
> 
> rcu_assign_pointer(some_callback, func);
> 
This one is OK. But a bit different from what I want to do.

> > }
> > 
> > /*
> >  * This function is called by
> >  *
> >  *	if (some_callback)
> >  *		some_callback(some_ptr);
> >  *
> >  * rather than
> >  *
> >  *	func(some_ptr);
> 
> well no, see later.
> 
I want to split built-in part and module part. I'm happy to embed

	if (some_callback)
		some_callback(some_ptr);

into built-in part but I'm not happy to expose tables to built-in part.
I'm using function pointer to realize it (like LSM).

> >  *
> >  * .
> >  */
> > static void func(void *some_ptr)
> > {
> > 	struct hashed_word *ptr;
> > 	struct list_head *table = &tables[hash_ptr(some_ptr, TABLEBIT)];
> > 	/* We must make sure that readers see table->next != NULL. */
> 
> 
> > 	smp_rmb();
> Nope.. delete this
> 
> > 	rcu_read_lock();
> > 	list_for_each_entry_rcu(ptr, table, list) {
> > 		if (ptr->key != some_ptr)
> > 			continue;
> > 		printk("%s\n", ptr->buf);
> > 		break;
> > 	}
> > 	rcu_read_unlock();
> > }
> > 
> > How can I make sure INIT_LIST_HEAD() in init() takes effect?
> > Is smp_rmb() appropriate?
> 
> Please avoid smp_rmb()/smb_wmb() and use RCU api only.
> 
> some_callback_t *ptr;
> 
> rcu_read_lock();
> ptr = rcu_dereference(some_callback);
> if (ptr) {

I want to know how I can guarantee that "table->next != NULL"
by exposing only "some_callback" to built-in part.
Once the result of INIT_LIST_HEAD() became visible to readers,
readers no longer need to use RCU for guaranteeing table->next != NULL.

> 	list_for_each_entry_rcu(aux, table, list) {
>  		if (aux->key != ptr)
>  			continue;
>  		printk("%s\n", aux->buf);
>  		break;
>  	}
>  }
> rcu_read_unlock();

Regards.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/