On Wed, May 18, 2005 at 06:26:50PM -0700, Dave Hansen wrote:
> There appears to still be some serious issues in the patch with respect
> to per-zone accounting.  There is only accounting in each ckrm_mem_res
> for each *kind* of zone, not each zone.

In the absense of NUMA/DISCONTIGMEM, isn't 'kind of zone' and 'zone'
the same ? Correct me if this assumption is wrong.

> 
> For instance, the accounting for a page appears to be the same no matter
> which zone it came from, just which kind of zone
> 
> Then, when it comes to actually use some of the information, the kswapd
> wakeup just throws a completely unrelated number into wakeup_kswapd().
> ZONE_DMA (zone 0) tends to be *MUCH* smaller than ZONE_HIGHMEM, for
> instance.  It doesn't make a whole lot of logical sense to me to be
> waking up kswapd for a possibly 16GB zone with data from a 16MB zone.

When control goes into wakeup_kswapd(), it looks for the over_limit list and
works only on the classes, and completely ignores the arguments to 
wakeup_kswapd().

I did it this way(instead of having my own logic) to use existing code.
> 
> +       for_each_zone(zone) {
> +               /* This is just a number to get to wakeup kswapd */
> +               order = cls->pg_total[0] -
> +                       ((ckrm_mem_shrink_to * cls->pg_limit) / 100);
> +               wakeup_kswapd(zone, order);
> +               break; /* only once is enough */
> +       }
> 
> If the number doesn't matter, why not just pass 0 into it?

Yes, i could. will do it.
> 
> Could you explain what advantages keeping a per-zone-type count has over
> actually doing one count for each zone?  Also, why bother tracking it
> per-zone-type anyway?  Would a single count work the same way

fits the NUMA/DISCONTIGMEM issue discussed above.

> 
> -- Dave
> 

-- 

----------------------------------------------------------------------
    Chandra Seetharaman               | Be careful what you choose....
              - [EMAIL PROTECTED]   |      .......you may get it.
----------------------------------------------------------------------


-------------------------------------------------------
This SF.Net email is sponsored by Oracle Space Sweepstakes
Want to be the first software developer in space?
Enter now for the Oracle Space Sweepstakes!
http://ads.osdn.com/?ad_id=7412&alloc_id=16344&op=click
_______________________________________________
ckrm-tech mailing list
https://lists.sourceforge.net/lists/listinfo/ckrm-tech

Reply via email to