On 06/27/2012 11:48 PM, David Rientjes wrote:

On Wed, 27 Jun 2012, Glauber Costa wrote:

quoted

quoted

@@ -2206,7 +2214,7 @@ static int mem_cgroup_do_charge(struct mem_cgroup

*memcg, gfp_t gfp_mask,

quoted

 	 * unlikely to succeed so close to the limit, and we fall back
 	 * to regular pages anyway in case of failure.
 	 */
-	if (nr_pages == 1 && ret)
+	if (nr_pages <= NR_PAGES_TO_RETRY && ret)
 		return CHARGE_RETRY;

Changed to costly order.

1 << PAGE_ALLOC_COSTLY_ORDER was the suggestion.

That is what I meant - to the costly order suggestion - , should have
been more explicit.

quoted

One more thing. The original version of this patch included
a cond_resched() here, that was also removed. From my re-reading
of the code in page_alloc.c and vmscan.c now, I tend to think
this is indeed not needed, since any cond_resched()s that might
be needed to ensure the safety of the code will be properly
inserted by the reclaim code itself, so there is no need for us
to include any when we signal that a retry is needed.

For __GFP_WAIT, that sounds like a safe guarantee.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>