From mboxrd@z Thu Jan 1 00:00:00 1970 From: Evgeniy Polyakov Subject: Re: skb_shared_info() Date: Thu, 10 Aug 2006 23:49:14 +0400 Message-ID: <20060810194914.GA29976@2ka.mipt.ru> References: <20060808.163915.03600382.davem@davemloft.net> Mime-Version: 1.0 Content-Type: text/plain; charset=koi8-r Cc: netdev@vger.kernel.org, herbert@gondor.apana.org.au Return-path: Received: from relay.2ka.mipt.ru ([194.85.82.65]:8851 "EHLO 2ka.mipt.ru") by vger.kernel.org with ESMTP id S932299AbWHJTuJ (ORCPT ); Thu, 10 Aug 2006 15:50:09 -0400 To: David Miller Content-Disposition: inline In-Reply-To: <20060808.163915.03600382.davem@davemloft.net> Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On Tue, Aug 08, 2006 at 04:39:15PM -0700, David Miller (davem@davemloft.net) wrote: > > I'm beginning to think that where we store the > skb_shared_info() is a weakness of the SKB design. Food for thoughts - unix sockets can use PAGE_SIZEd chunks of memory (and they do it almost always), which are aligned to 2*PAGE_SIZE due to alignment issues with skb_shared_info, so unix sockets waste 3.5 kb of memory on each skb. I think it is time to resurrect idea of placing shared_info inside skb and allow to allocate it from own cache for special cases, what do you think? -- Evgeniy Polyakov