From mboxrd@z Thu Jan 1 00:00:00 1970 From: Christoph Hellwig Subject: Re: [PATCH] scsi/arcmsr: Add timeout module parameter Date: Tue, 19 Aug 2014 11:07:54 -0700 Message-ID: <20140819180754.GB16742@infradead.org> References: <20140804111552.GA2494@infradead.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from bombadil.infradead.org ([198.137.202.9]:41678 "EHLO bombadil.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751496AbaHSSH4 (ORCPT ); Tue, 19 Aug 2014 14:07:56 -0400 Content-Disposition: inline In-Reply-To: Sender: linux-scsi-owner@vger.kernel.org List-Id: linux-scsi@vger.kernel.org To: Ari Sundholm Cc: linux-scsi@vger.kernel.org, Ching Huang On Tue, Aug 05, 2014 at 04:10:08PM +0300, Ari Sundholm wrote: > > On Mon, 4 Aug 2014, Christoph Hellwig wrote: > > > To modify the timeout on a queue please use blk_queue_rq_timeout in > > the slave_configure method instead of poking directly into the block > > timer, which won't work e.g. for the blk-mq path. > > I see. It seems the way the driver Areca offers on its download page does > this is wrong, then, as all I did was I ported the feature from there. > > > But this really needs an explanation on why you'd need a configurable > > timeout to start with. > > A colleague of mine has experienced problems with disks not supporting > TLER performing lengthy error recovery, hitting a timeout. He has > mitigated this problem by configuring the timeout to be long enough to > allow for the error recovery to finish. And again, the driver offered by > Areca supports this feature, so it seems they have their reasons as well. Can you test this patch: http://git.infradead.org/users/hch/scsi-queue.git/commitdiff/3a7fcf9e5f57469fab78d6b6db81fbef6fa48a0d which is part of the latest Areca code drop? I don't really like the timeout parameter, but if there's no better work around for you we'll probablyneed to merge something. In that case I'd love a patch ontop of the branch that the above patches is in, which used the timeout manipulation method I mentioned in my previous mail.