From mboxrd@z Thu Jan 1 00:00:00 1970 From: Alan Cox Subject: Re: [PATCH] Re: libata: PIO bounce copies to avoid long IRQ off runs Date: Tue, 29 Apr 2008 23:01:05 +0100 Message-ID: <20080429230105.710cd464@core> References: <20080429144754.7d2952c8@core> <20080429220128.GA25818@havoc.gtf.org> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Return-path: Received: from outpipe-village-512-1.bc.nu ([81.2.110.250]:45800 "EHLO lxorguk.ukuu.org.uk" rhost-flags-OK-FAIL-OK-FAIL) by vger.kernel.org with ESMTP id S1755906AbYD2WIh (ORCPT ); Tue, 29 Apr 2008 18:08:37 -0400 In-Reply-To: <20080429220128.GA25818@havoc.gtf.org> Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: Jeff Garzik Cc: linux-ide@vger.kernel.org On Tue, 29 Apr 2008 18:01:28 -0400 Jeff Garzik wrote: > I would tend to prefer this version... > > - allocate buffer at init time > - use lesser-used KM_IRQ1 Seems reasonable enough. You actually want to deal with the buffer at run time however because you care about two cases 1. Box with no highmem - easy to do compile time 2. Box with no high pages allocated - kind of hairy with gcc as the crystal ball feature doesn't work.