Re: [PATCH] pids: simplify do_each_task_pid/while_each_task_pid

[Oleg Nesterov <oleg@tv-sign.ru>]

On 04/13, Christoph Hellwig wrote:
>
> On Thu, Apr 13, 2006 at 09:54:31PM +0400, Oleg Nesterov wrote:
> > > 
> > > #define for_each_task_pid(task, pid, type, pos) \
> > > 	hlist_for_each_entry_rcu((task), (pos),  \
> > > 		(&(pid))->tasks[type], pids[type].node) {
> > > 
> > > and move the find_pid to the caller?  That would make the code a whole lot
> > > more readable.
> > 
> > Then the caller should check find_pid() doesn't return NULL. But yes,
> > we can hide this check inside for_each_task_pid().
> > 
> > But what about current users of do_each_task_pid ? We can't just remove
> > these macros.
> 
> They'd have to switch over to the new variant.  There's just 18 callers
> ayway, currently, and with a patch like the one below that number firther
> decreases :)

Ok, In such a case we should first

#define NEW_IMPROVED_HLIST_FOR_EACH_ENTRY_RCU_WHICH_DOESNT_NEED_EXTRA_PARM(pos, head, member)	\
	for (pos = hlist_entry((head)->first, typeof(*(pos)), member);			\
		rcu_dereference(pos) != hlist_entry(NULL, typeof(*(pos)), member)	\
			&& ({ prefetch((pos)->member.next); 1; });			\
		(pos) = hlist_entry((pos)->member.next, typeof(*(pos)), member))

What do you think? What should be the name for it?

Oleg.