From mboxrd@z Thu Jan 1 00:00:00 1970 From: Stephen Hemminger Subject: Re: sky2 panic in 2.6.32.1 under load (new oops) Date: Thu, 24 Dec 2009 14:27:12 -0800 Message-ID: <20091224142712.1e5c46d2@nehalam> References: <4B300A2A.8040305@gmail.com> <4B300E30.9090707@majjas.com> <4B3114E3.1070602@majjas.com> <4B329FA3.9090904@majjas.com> <20091223230102.4bb0100e.akpm@linux-foundation.org> <4B33BE74.3030704@majjas.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Cc: Andrew Morton , "Berck E. Nash" , "linux-kernel@vger.kernel.org" , netdev@vger.kernel.org To: Michael Breuer Return-path: Received: from mail.vyatta.com ([76.74.103.46]:54452 "EHLO mail.vyatta.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755046AbZLXW1e (ORCPT ); Thu, 24 Dec 2009 17:27:34 -0500 In-Reply-To: <4B33BE74.3030704@majjas.com> Sender: netdev-owner@vger.kernel.org List-ID: On Thu, 24 Dec 2009 14:18:12 -0500 Michael Breuer wrote: > I'll be the first to admit I don't know much about network drivers... > but I am wondering why __free_page(frag->page); is called in > skb_put_frags when the fragment length is zero. I don't see why that > page pointer wouldn't potentially be reused afterwords. When using Jumbo frames, the sky2 driver allocates pages to hold the larger part of the frame. With 9K MTU it will allocate 2 * 4K pages and then 1K for start of frame. Then if frame is <= 1K it will drop the extra pages. Not sure why this broke now, I am not a VM expert. --