From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Miller Subject: Re: [PATCH] block: fix q->max_segment_size checking in blk_recalc_rq_segments about VMERGE Date: Thu, 24 Jul 2008 14:53:36 -0700 (PDT) Message-ID: <20080724.145336.41899163.davem@davemloft.net> References: <1216918371.4524.38.camel@localhost.localdomain> Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Content-Transfer-Encoding: 7bit Return-path: Received: from 74-93-104-97-Washington.hfc.comcastbusiness.net ([74.93.104.97]:36592 "EHLO sunset.davemloft.net" rhost-flags-OK-FAIL-OK-OK) by vger.kernel.org with ESMTP id S1751552AbYGXVxg (ORCPT ); Thu, 24 Jul 2008 17:53:36 -0400 In-Reply-To: Sender: linux-scsi-owner@vger.kernel.org List-Id: linux-scsi@vger.kernel.org To: mpatocka@redhat.com Cc: James.Bottomley@HansenPartnership.com, fujita.tomonori@lab.ntt.co.jp, jens.axboe@oracle.com, linux-kernel@vger.kernel.org, linux-scsi@vger.kernel.org, linux-parisc@vger.kernel.org From: Mikulas Patocka Date: Thu, 24 Jul 2008 17:49:14 -0400 (EDT) > * it is prone to bugs and hard to maintain, because the same value must be > calculated in blk-merge.c and in architectural iommu functions --- if the > value differs, you create too long request, corrupt kernel memory and > crash (happened on sparc64). Anyone changing blk-merge in the future will > risk breaking something on the architectures that use BIO_VMERGE_BOUNDARY > --- and because these architectures are so rare, the bug will go unnoticed > for long time --- like in the case of sparc64. I completely agree with this point. This VMERGE stuff is now a non-trivial maintainence burdon because anyone who wants to hack on the block layer has to be mindful of VMERGE but is very unlikely to have access to a system that it can even be tested on. And the answer isn't "James Bottomly will test your changes for you", because that simply doesn't scale. I still say we should definitely remove the VMERGE code. It's not worth the maintainence hassle just for some SG chaining test rig on some obscure platform. I really only hear one person who really wants this code around any more. Is that the Linux way? :-) Can't he patch it into his tree when he needs it or write an alternative way to stress the SG chaining code? He has the source, right? :-)))