From mboxrd@z Thu Jan 1 00:00:00 1970 From: Christoph Hellwig Subject: Re: SCSI eats error from flush failure during hot plug Date: Wed, 11 Jun 2014 06:37:36 -0700 Message-ID: <20140611133736.GA12240@infradead.org> References: <1402169384.2236.18.camel@dabdike.int.hansenpartnership.com> <1402334946.2197.17.camel@dabdike.int.hansenpartnership.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from bombadil.infradead.org ([198.137.202.9]:55543 "EHLO bombadil.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752127AbaFKNhi (ORCPT ); Wed, 11 Jun 2014 09:37:38 -0400 Content-Disposition: inline In-Reply-To: <1402334946.2197.17.camel@dabdike.int.hansenpartnership.com> Sender: linux-scsi-owner@vger.kernel.org List-Id: linux-scsi@vger.kernel.org To: James Bottomley Cc: Steven Haber , linux-scsi@vger.kernel.org, Jens Axboe On Mon, Jun 09, 2014 at 10:29:06AM -0700, James Bottomley wrote: > I'll do it as a bug fix, but I do need Jens to make sure nothing else > breaks first. Best I can tell, the state model for compound commands > like flush doesn't expect us to change the request type ... nothing puts > it back to REQ_TYPE_FS. In your case, the flush is the last command > sent, so there's no problem ... I just worry we will get an obscure > problem later on from something that does a BLOCK_PC prepared first > command. Yes, I don't think resetting cmd_type is a good idea. I'd much rather see a special case for rq->cmd_flags & REQ_FLUSH in the completion handler - we already treat it special during setup anyway.