From mboxrd@z Thu Jan 1 00:00:00 1970 From: James Bottomley Subject: Re: [PATCH 1/2] scsi-mq: Only show the CDB if available Date: Tue, 05 Dec 2017 08:43:49 -0800 Message-ID: <1512492229.3019.25.camel@linux.vnet.ibm.com> References: <20171205003809.1705-1-bart.vanassche@wdc.com> <20171205003809.1705-2-bart.vanassche@wdc.com> <20171205011536.GC8365@ming.t460p> <1512439190.2739.0.camel@wdc.com> <20171205050008.GA9989@ming.t460p> <1512490952.2660.18.camel@wdc.com> <20171205163824.GB23788@ming.t460p> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 8bit Return-path: In-Reply-To: <20171205163824.GB23788@ming.t460p> Sender: linux-block-owner@vger.kernel.org To: Ming Lei , Bart Van Assche Cc: "linux-block@vger.kernel.org" , "jthumshirn@suse.de" , "hch@lst.de" , "martin.petersen@oracle.com" , "axboe@kernel.dk" , "linux-scsi@vger.kernel.org" , "hare@suse.com" List-Id: linux-scsi@vger.kernel.org On Wed, 2017-12-06 at 00:38 +0800, Ming Lei wrote: > On Tue, Dec 05, 2017 at 04:22:33PM +0000, Bart Van Assche wrote: > > > > On Tue, 2017-12-05 at 13:00 +0800, Ming Lei wrote: > > > > > > No, do not mix two different things in one patch, especially the > > > fix part need to be backported to stable. > > > > > > The fix part should aim at V4.15, and the other part can be a > > > V4.16 stuff. > > > > Does this mean that you do not plan to post a v5 of your patch and > > that you want me to rework this patch series? I can do that. > > I believe V4 has been OK for merge, actually the only concern from > James is that 'set the cmnd to NULL *before* calling free so we > narrow the race window.', but that isn't required as I explained, > even though you don't do that in this patch too. > > https://marc.info/?t=151030464300003&r=1&w=2 > > I will work on V5 if Martin/James thinks it is needed. I don't buy that it isn't needed.  The point (and the pattern) is for a destructive action set the signal *before* you execute the action not after.  The reason should be obvious: if you set it after you invite a race where the check says OK but the object has gone.  Even if the race is highly unlikely, the pattern point still holds. James