From mboxrd@z Thu Jan 1 00:00:00 1970 From: Roland Dreier Subject: Re: Some fixes for the SRP driver Date: Wed, 17 May 2006 07:54:13 -0700 Message-ID: References: <20060513044445.GQ12272@parisc-linux.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Received: from sj-iport-5.cisco.com ([171.68.10.87]:64554 "EHLO sj-iport-5.cisco.com") by vger.kernel.org with ESMTP id S1750809AbWEQOyQ (ORCPT ); Wed, 17 May 2006 10:54:16 -0400 In-Reply-To: <20060513044445.GQ12272@parisc-linux.org> (Matthew Wilcox's message of "Fri, 12 May 2006 22:44:45 -0600") Sender: linux-scsi-owner@vger.kernel.org List-Id: linux-scsi@vger.kernel.org To: Matthew Wilcox Cc: linux-scsi@vger.kernel.org Thanks, I queued up most of this for 2.6.18. > - You might consider moving this driver to drivers/scsi. It's actually > fairly small compared to most scsi drivers, and basically you've got > two files plus Kbuild/Kconfig machinery. Seems a bit silly, plus > people would notice this scsi driver more readily. Yeah, it's a valid point. On the other hand, it's also an infiniband driver. It's kind of like the ieee1394 sbp2 driver -- both fish and fowl ;) > - I'm not convinced you need the target list/lock I mention above. > The Scsi_Host structure has a list of associated targets, although > I don't see a good iterator for them right now. It's a little different than that. Confusingly enough, the SRP driver uses a scsi_host for each target port it connects to, and then has a higher-level private structure for each host adapter with a list of target ports. That list is what's being protected. The reason for using a full scsi_host for a target port is so that we can limit the number of commands sent to a target port from the midlayer (because the number of outstanding work requests on a given connection is what is limited). - R.