Re: [PATCH] ipc: use list_for_each_entry for list traversing

[Andrew Morton <akpm@linux-foundation.org>]

On Tue, 09 Apr 2013 11:39:07 +0200 Nikola Pajkovsky <npajkovs@redhat.com> wrote:

> Andrew Morton <akpm@linux-foundation.org> writes:
> 
> > urgh, that code is sick.  What's it doing poking around in the
> > list_head internals?
> 
> No idea, it there from beginning of first kernel importation into
> git. Where is history before git?

The old bitkeeper repo is at
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/old-2.6-bkcvs.git
and goes back to the 2.4->2.5 split iirc.  But this code predates that
- my trusty CVS tree goes back to linux-2.4.2-pre2 (November 2000) and
the list.next hackery is in there.

> >> @@ -253,14 +248,9 @@ static void ss_wakeup(struct list_head *h, int kill)
> >>  
> >>  static void expunge_all(struct msg_queue *msq, int res)
> >>  {
> >> -	struct list_head *tmp;
> >> -
> >> -	tmp = msq->q_receivers.next;
> >> -	while (tmp != &msq->q_receivers) {
> >> -		struct msg_receiver *msr;
> >> +	struct msg_receiver *msr, *t;
> >>  
> >> -		msr = list_entry(tmp, struct msg_receiver, r_list);
> >> -		tmp = tmp->next;
> >> +	list_for_each_entry_safe(msr, t, &msq->q_receivers, r_list) {
> >>  		msr->r_msg = NULL;
> >>  		wake_up_process(msr->r_tsk);
> >>  		smp_mb();
> >
> > I think list_for_each_entry() would suffice here.
> 
> I don't know, I found wake_up_sem_queue_do in sem.c and it looks almost same
> except preempt stuff. I'll be dancing around ipc/

wake_up_sem_queue_do() is wrong ;)