From owner-svn-src-head@FreeBSD.ORG Sun Jun 6 14:09:48 2010 Return-Path: Delivered-To: svn-src-head@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id D7D401065679; Sun, 6 Jun 2010 14:09:48 +0000 (UTC) (envelope-from nwhitehorn@FreeBSD.org) Received: from svn.freebsd.org (unknown [IPv6:2001:4f8:fff6::2c]) by mx1.freebsd.org (Postfix) with ESMTP id BBDF98FC13; Sun, 6 Jun 2010 14:09:48 +0000 (UTC) Received: from svn.freebsd.org (localhost [127.0.0.1]) by svn.freebsd.org (8.14.3/8.14.3) with ESMTP id o56E9mf8063170; Sun, 6 Jun 2010 14:09:48 GMT (envelope-from nwhitehorn@svn.freebsd.org) Received: (from nwhitehorn@localhost) by svn.freebsd.org (8.14.3/8.14.3/Submit) id o56E9mWI063166; Sun, 6 Jun 2010 14:09:48 GMT (envelope-from nwhitehorn@svn.freebsd.org) Message-Id: <201006061409.o56E9mWI063166@svn.freebsd.org> From: Nathan Whitehorn Date: Sun, 6 Jun 2010 14:09:48 +0000 (UTC) To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org X-SVN-Group: head MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Cc: Subject: svn commit: r208870 - in head/sys/dev/ata: . chipsets X-BeenThere: svn-src-head@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: SVN commit messages for the src tree for head/-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 06 Jun 2010 14:09:49 -0000 Author: nwhitehorn Date: Sun Jun 6 14:09:48 2010 New Revision: 208870 URL: http://svn.freebsd.org/changeset/base/208870 Log: Some revisions of the Serverworks K2 SATA controller have a data corruption bug where if an ATA command is issued before DMA is started, data will become available to the controller before it knows what to do with it. This results in either data corruption or a controller crash. This patch remedies the problem by adopting the workaround employed by Linux and Darwin: starting the DMA engine prior to sending the ATA command. Observer on: Xserve G5 Reviewed by: mav MFC after: 1 week Modified: head/sys/dev/ata/ata-all.h head/sys/dev/ata/ata-lowlevel.c head/sys/dev/ata/chipsets/ata-serverworks.c Modified: head/sys/dev/ata/ata-all.h ============================================================================== --- head/sys/dev/ata/ata-all.h Sun Jun 6 13:08:36 2010 (r208869) +++ head/sys/dev/ata/ata-all.h Sun Jun 6 14:09:48 2010 (r208870) @@ -564,6 +564,7 @@ struct ata_channel { #define ATA_CHECKS_CABLE 0x20 #define ATA_NO_ATAPI_DMA 0x40 #define ATA_SATA 0x80 +#define ATA_DMA_BEFORE_CMD 0x100 int pm_level; /* power management level */ int devices; /* what is present */ Modified: head/sys/dev/ata/ata-lowlevel.c ============================================================================== --- head/sys/dev/ata/ata-lowlevel.c Sun Jun 6 13:08:36 2010 (r208869) +++ head/sys/dev/ata/ata-lowlevel.c Sun Jun 6 14:09:48 2010 (r208870) @@ -141,6 +141,14 @@ ata_begin_transaction(struct ata_request goto begin_finished; } + /* start DMA engine if necessary */ + if ((ch->flags & ATA_DMA_BEFORE_CMD) && + ch->dma.start && ch->dma.start(request)) { + device_printf(request->parent, "error starting DMA\n"); + request->result = EIO; + goto begin_finished; + } + /* issue command */ if (ch->hw.command(request)) { device_printf(request->parent, "error issuing %s command\n", @@ -150,7 +158,8 @@ ata_begin_transaction(struct ata_request } /* start DMA engine */ - if (ch->dma.start && ch->dma.start(request)) { + if (!(ch->flags & ATA_DMA_BEFORE_CMD) && + ch->dma.start && ch->dma.start(request)) { device_printf(request->parent, "error starting DMA\n"); request->result = EIO; goto begin_finished; Modified: head/sys/dev/ata/chipsets/ata-serverworks.c ============================================================================== --- head/sys/dev/ata/chipsets/ata-serverworks.c Sun Jun 6 13:08:36 2010 (r208869) +++ head/sys/dev/ata/chipsets/ata-serverworks.c Sun Jun 6 14:09:48 2010 (r208870) @@ -241,6 +241,16 @@ ata_serverworks_ch_attach(device_t dev) ATA_OUTL(ctlr->r_res2, ch_offset + 0x88, 0); ATA_OUTL(ctlr->r_res2, ch_offset + 0x80, ATA_INL(ctlr->r_res2, ch_offset + 0x80) & ~0x00040000); + + /* + * Some controllers have a bug where they will send the command + * to the drive before seeing a DMA start, and then can begin + * receiving data before the DMA start arrives. The controller + * will then become confused and either corrupt the data or crash. + * Remedy this by starting DMA before sending the drive command. + */ + + ch->flags |= ATA_DMA_BEFORE_CMD; } /* chip does not reliably do 64K DMA transfers */