From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Miller Subject: Re: [PANIC] lro + iscsi or lro + skb text search causes panic Date: Sun, 25 Jan 2009 21:34:36 -0800 (PST) Message-ID: <20090125.213436.268287293.davem@davemloft.net> References: <4978F804.3060502@cs.wisc.edu> <49794612.9010204@cs.wisc.edu> Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Cc: jesse.brandeburg@intel.com, netdev@vger.kernel.org, olaf.kirch@oracle.com, tgraf@suug.ch, kkeil@suse.de, herbert@gondor.apana.org.au To: michaelc@cs.wisc.edu Return-path: Received: from 74-93-104-97-Washington.hfc.comcastbusiness.net ([74.93.104.97]:51176 "EHLO sunset.davemloft.net" rhost-flags-OK-FAIL-OK-OK) by vger.kernel.org with ESMTP id S1750753AbZAZFed (ORCPT ); Mon, 26 Jan 2009 00:34:33 -0500 In-Reply-To: <49794612.9010204@cs.wisc.edu> Sender: netdev-owner@vger.kernel.org List-ID: From: Mike Christie Date: Thu, 22 Jan 2009 22:22:42 -0600 > Attached is a patch made against the scsi maintainer's tree (I think > it should also apply to linus's) that converts iscsi to use > skb_copy_bits. It is lightly tested. If there is no benefit in > having skb_find_text use skb_seq_read maybe we can just kill it, so > people do not have to maintain two helpers that provide similar > functionality. Regardless of what we decide to do with iSCSI we have to make this function work properly as long as there is a user in the tree. This iterator can in fact be more efficient than using copy bits since copy bits doesn't remember any state from previous invocations whereas the iterator does. So if you call it multiple times on the same SKB that's a lot of wasted work. In fact I'm pretty sure that's why we added it for textsearch, because in that application we're constantly beating through the same SKB via the testsearch state machine. Therefore, keeping it's use for iSCSI would be beneficial.