From mboxrd@z Thu Jan 1 00:00:00 1970 From: James Bottomley Subject: RE: [PATCH] kill aac_io_done Date: 11 Oct 2004 11:55:44 -0500 Sender: linux-scsi-owner@vger.kernel.org Message-ID: <1097513750.2029.152.camel@mulgrave> References: <60807403EABEB443939A5A7AA8A7458B2FC33F@otce2k01.adaptec.com> Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Return-path: Received: from stat16.steeleye.com ([209.192.50.48]:18585 "EHLO hancock.sc.steeleye.com") by vger.kernel.org with ESMTP id S269177AbUJKQ4Q (ORCPT ); Mon, 11 Oct 2004 12:56:16 -0400 In-Reply-To: <60807403EABEB443939A5A7AA8A7458B2FC33F@otce2k01.adaptec.com> List-Id: linux-scsi@vger.kernel.org To: "Salyzyn, Mark" Cc: Mark Haverkamp , Christoph Hellwig , linux-scsi On Mon, 2004-10-11 at 11:07, Salyzyn, Mark wrote: > The host lock around scsi_done() was there *before* I started > maintaining this driver. I believe the reasoning could be as simple as > `we had an io_request_lock there, so we should need a host_lock for the > same reasons'. I believe this locking is a legacy of code prior to > 2.1.95 though... I think it was a 2.2 requirement; the then "new" SCSI code in 2.4 moved to completing requests in the bottom half handler which queue was protected by the scsi_bhqueue_lock in the scsi_done() routine and no longer needed the io_request lock. In 2.6 the BH handler became a softirq thanks to Mattew Wilcox. James