Re: prepare_to_wait/waitqueue_active issues in 2.6

From: Anton Blanchard
Date: Sun Dec 14 2003 - 00:28:50 EST



Hi,

> Pardon my French, but this patch sure looks like crap.

I agree. Your patch wins, and that comment would even pass akpm best
practices.

> So my preference would be to add the barrier into prepare_to_wait(), along
> with a comment on why it is sometimes needed. Something like the
> appended.. (which just uses "set_current_state()", since that's what it
> exists for).

And thats pretty much how 2.4 handled the problem (set_task_state sits
between the waitqueue addition and the test). That still leaves kswapd
with problems, but in low memory conditions we'll be calling it often so
one lost wakeup here and there shouldnt matter.

Anton
-
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/