Re: [RFC, PATCH] Slab counter troubles with swap prefetch?

[Christoph Lameter]

On Fri, 11 Nov 2005, Con Kolivas wrote:

> One last thing. Swap prefetch works off the accounting of total memory and is 
> only a single kernel thread rather than a thread per cpu or per pgdat unlike 
> kswapd. Currently it just cares about total slab data and total ram. 
> Depending on where this thread is scheduled (which node) your accounting 
> change will alter the behaviour of it. Does this affect the relevance of this 
> patch to you?

Yes, if its a truly global value then we would not need the patch. 
But then the prefetch code would have to add up all the nr_slab field for 
all processors and use that result for comparison. If you do this in a 
node specific fashion then the problem comes up again.



-
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/