From mboxrd@z Thu Jan 1 00:00:00 1970 From: Mark Lord Subject: Re: sata_mv query Date: Tue, 07 Aug 2012 14:20:48 -0400 Message-ID: <50215C80.8080601@teksavvy.com> References: <006c01cd74ac$a06edf30$e14c9d90$@sturgeon@perpetual-data.com> <50214D56.9090907@teksavvy.com> <50215BC9.2020709@teksavvy.com> Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="------------050308030001050803010407" Return-path: Received: from ironport2-out.teksavvy.com ([206.248.154.182]:38027 "EHLO ironport2-out.teksavvy.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753016Ab2HGSUt (ORCPT ); Tue, 7 Aug 2012 14:20:49 -0400 In-Reply-To: <50215BC9.2020709@teksavvy.com> Sender: linux-ide-owner@vger.kernel.org List-Id: linux-ide@vger.kernel.org To: Barry J Sturgeon Cc: linux-ide@vger.kernel.org This is a multi-part message in MIME format. --------------050308030001050803010407 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit On 12-08-07 02:17 PM, Mark Lord wrote: .. > Say, here's a thought: We could get rid of (or relocate to upper layers) > the busy-wait by having something similar get called from ata_qc_defer(). > > If you are feeling adventurous, here is a 100% untested patch to do just that. > It looks correct to me, it compiles, but that's all I can say. Whoops.. return codes were reversed. Here's the corrected patch. The attached copy is the better one to use -- my mailer mangles inline patches. --- linux-3.4.4/drivers/ata/sata_mv.c 2012-06-22 14:37:50.000000000 -0400 +++ linux/drivers/ata/sata_mv.c 2012-08-07 14:14:43.554503157 -0400 @@ -1388,6 +1388,17 @@ } } +static int mv_check_for_edma_empty_idle(struct ata_port *ap) +{ + void __iomem *port_mmio = mv_ap_base(ap); + const u32 empty_idle = (EDMA_STATUS_CACHE_EMPTY | EDMA_STATUS_IDLE); + + u32 edma_stat = readl(port_mmio + EDMA_STATUS); + if ((edma_stat & empty_idle) == empty_idle) + return 0; + return ATA_DEFER_PORT; +} + static int mv_qc_defer(struct ata_queued_cmd *qc) { struct ata_link *link = qc->dev->link; @@ -1423,7 +1434,7 @@ * If the port is completely idle, then allow the new qc. */ if (ap->nr_active_links == 0) - return 0; + return mv_check_for_edma_empty_idle(ap); /* * The port is operating in host queuing mode (EDMA) with NCQ --------------050308030001050803010407 Content-Type: text/x-patch; name="xx_sata_mv_move_busywait.patch" Content-Transfer-Encoding: 7bit Content-Disposition: attachment; filename="xx_sata_mv_move_busywait.patch" --- linux-3.4.4/drivers/ata/sata_mv.c 2012-06-22 14:37:50.000000000 -0400 +++ linux/drivers/ata/sata_mv.c 2012-08-07 14:14:43.554503157 -0400 @@ -1388,6 +1388,17 @@ } } +static int mv_check_for_edma_empty_idle(struct ata_port *ap) +{ + void __iomem *port_mmio = mv_ap_base(ap); + const u32 empty_idle = (EDMA_STATUS_CACHE_EMPTY | EDMA_STATUS_IDLE); + + u32 edma_stat = readl(port_mmio + EDMA_STATUS); + if ((edma_stat & empty_idle) == empty_idle) + return 0; + return ATA_DEFER_PORT; +} + static int mv_qc_defer(struct ata_queued_cmd *qc) { struct ata_link *link = qc->dev->link; @@ -1423,7 +1434,7 @@ * If the port is completely idle, then allow the new qc. */ if (ap->nr_active_links == 0) - return 0; + return mv_check_for_edma_empty_idle(ap); /* * The port is operating in host queuing mode (EDMA) with NCQ --------------050308030001050803010407--