From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S933256Ab2GYOhw (ORCPT ); Wed, 25 Jul 2012 10:37:52 -0400 Received: from natasha.panasas.com ([67.152.220.90]:43496 "EHLO natasha.panasas.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S933148Ab2GYOhv (ORCPT ); Wed, 25 Jul 2012 10:37:51 -0400 Message-ID: <5010047D.6070807@panasas.com> Date: Wed, 25 Jul 2012 17:36:45 +0300 From: Boaz Harrosh User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:8.0) Gecko/20111113 Thunderbird/8.0 MIME-Version: 1.0 To: Paolo Bonzini CC: Wang Sen , , , , , Subject: Re: [PATCH] scsi: virtio-scsi: Fix address translation failure of HighMem pages used by sg list References: <1343204966-23560-1-git-send-email-senwang@linux.vnet.ibm.com> <500FB1DE.1000100@redhat.com> <500FBAE8.2050107@panasas.com> <500FBF37.50603@redhat.com> <500FE7D2.7070101@panasas.com> <500FEB63.3000709@redhat.com> <500FF412.3090600@panasas.com> <500FF656.6000203@redhat.com> In-Reply-To: <500FF656.6000203@redhat.com> Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 07/25/2012 04:36 PM, Paolo Bonzini wrote: > Il 25/07/2012 15:26, Boaz Harrosh ha scritto: >> On 07/25/2012 03:49 PM, Paolo Bonzini wrote: >> >>> >>> Except here the destination array has to be given to virtio, which >>> doesn't (yet) understand chaining. I'm using for_each_sg rather than a >>> simple memcpy exactly because I want to flatten the input scatterlist >>> onto consecutive scatterlist entries, which is what virtio expects (and >>> what I'll change when I get to it). >>> >>> for_each_sg guarantees that I get non-chain scatterlists only, so it is >>> okay to value-assign them to sg[]. >> >> So if the virtio does not understand chaining at all then surly it will >> not understand the 2-bit end marker and will get a wrong page pointer >> with the 1st bit set. > > It doesn't understand chaining, but it does use sg_phys(x) so it will > not get a wrong page pointer for the end marker. > >> Fine then your code is now a crash because the terminating bit was just >> copied over, which it was not before. > > I did test the patch with value-assignment. > Still you should use the sg_set_page()!! 1. It is not allowed to directly manipulate sg entries. One should always use the proper accessor. Even if open coding does work and is not a bug it should not be used anyway! 2. Future code that will support chaining will need to do as I say so why change it then, again? Please don't change two things in one patch. The fix is for high-pages please fix only that here. You can blasphemy open-code the sg manipulation in a separate patch. Please Boaz >> Lets separate the two topics from now on. Send me one mail concerning >> the proper above patch, And a different mail for how to support chaining. > > Ok, and I'll change the topic. > > Paolo