[openib-general] Re: [PATCH] IPOIB: initialization patch in ipoib

[Roland Dreier]

    Michael> If the order of these operations really matters, wont you
    Michael> need some kind of memory barrier in between?

No, no CPU is so insane that you need a barrier between

	x->a->b = c;

and

	kfree(x);

And it is always wrong to do

	kfree(x);
	x->a->b = c;

 - Roland