From mboxrd@z Thu Jan 1 00:00:00 1970 From: Christoph Hellwig Subject: Re: smp dead lock of io_request_lock/queue_lock patch Date: Sat, 17 Jan 2004 20:54:50 +0000 Sender: linux-scsi-owner@vger.kernel.org Message-ID: <20040117205450.A7305@infradead.org> References: <20040112194829.A7078@infradead.org> <1073937102.3114.300.camel@compaq.xsintricity.com> <1074345000.13198.25.camel@compaq.xsintricity.com> <20040117165828.A4977@infradead.org> <1074366452.13198.48.camel@compaq.xsintricity.com> <20040117191704.A6344@infradead.org> <1074367303.13198.52.camel@compaq.xsintricity.com> <20040117192946.A6479@infradead.org> <1074371793.13198.58.camel@compaq.xsintricity.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from phoenix.infradead.org ([213.86.99.234]:60175 "EHLO phoenix.infradead.org") by vger.kernel.org with ESMTP id S266145AbUAQUyv (ORCPT ); Sat, 17 Jan 2004 15:54:51 -0500 Content-Disposition: inline In-Reply-To: <1074371793.13198.58.camel@compaq.xsintricity.com>; from dledford@redhat.com on Sat, Jan 17, 2004 at 03:36:33PM -0500 List-Id: linux-scsi@vger.kernel.org To: Doug Ledford Cc: linux-scsi mailing list On Sat, Jan 17, 2004 at 03:36:33PM -0500, Doug Ledford wrote: > I'm exactly the opposite. In my opinion, the mid layer should be using > the device locks for its internal operations well, I strongly disagree with that. > and calling into the low > level drivers with no locks held. The fact that the mid layer still > tries to lock drivers for the drivers is a bug IMO. Every time I see a > driver do spin_unlock_irq(host->host_lock); do driver work; > spin_lock_irq(host->host_lock); return; it really points out that the > mid layer has no business locking down drivers for them. I completely agree with that, though. I didn't have time fixing up all drivers before we got 2.6, though so this won't happen before 2.7. Btw, I just noticed how little this has to do with the original discussion, Cc list thus cut down quite a lot.