On Wednesday 28 May 2008 04:47, Stephen Hemminger wrote:

On Thu, 22 May 2008 05:57:44 -0400

Jeff Garzik [off-list ref] wrote:

quoted

quoted

 	for (i = 0; i < sky2->rx_nfrags; i++) {
-		struct page *page = alloc_page(GFP_ATOMIC);
+		struct page *page = alloc_page(GFP_ATOMIC | __GFP_NOWARN);

 		if (!page)
 			goto free_partial;

IMO it's inappropriate to add these warnings to net drivers that
properly check all return values.

This approach is too maintenance intensive, and winds up fixing the same
problem over and over again -- a hint that the fix is in the wrong place.

	Jeff

So the __GFP_NOWARN should go away and get replaced by GFP_WARN?

We actually still want to see the messages, regardless of what is causing
them, because they can indicate problems in the VM.

The best way is not to add __GFP_NOWARN at all, and just improve the
throttling/thresholding/reporting of the page allocation failure warnings.