From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Dumazet Subject: RE: [PATCH] net-next: ipv6: ndisc: allocate a ndisc socket per inet6_dev Date: Mon, 25 Jun 2012 14:02:37 +0200 Message-ID: <1340625757.10893.40.camel@edumazet-glaptop> References: <1340620878.10893.26.camel@edumazet-glaptop> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org To: Menny_Hamburger@Dell.com Return-path: Received: from mail-bk0-f46.google.com ([209.85.214.46]:40005 "EHLO mail-bk0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752048Ab2FYMCm (ORCPT ); Mon, 25 Jun 2012 08:02:42 -0400 Received: by bkcji2 with SMTP id ji2so3028276bkc.19 for ; Mon, 25 Jun 2012 05:02:41 -0700 (PDT) In-Reply-To: Sender: netdev-owner@vger.kernel.org List-ID: On Mon, 2012-06-25 at 12:08 +0100, Menny_Hamburger@Dell.com wrote: > I'm sorry for not responding on your post. ... > I really want to understand how this fixes our problem. > This fix will make the skb allocations succeed, but what mechanism releases the stuck socket associated with the bad device? There is no stuck socket, I don't know why you believe this. And I don't want to spend time writing a changelog if the patch doesn't solve the problem, especially knowing you didn't even test it. net/ipv4/arp.c arp_create() doesn't use a 'global socket' to attach skbs on a socket. This kind of protection makes sense for UDP / ICMP sockets, not for ARP or ND resolution. If you don't understand my patch, don't post your own, thats really lost time for everyone. We don't want adding yet another per device stuff for ipv6. We already have too big memory footprint. Once I am sure patch fixes the problem, I'll make an official submission with all credits and changelog.