[Lucid] SRU: [PATCH] UBUNTU: Make ahci ignore the controller for MBP7, 1 and fall back to ata_generic. Patch based on Tejun Heo's submitted upstream.

Andy Whitcroft apw at canonical.com
Fri Jul 2 08:17:22 UTC 2010


On Fri, Jul 02, 2010 at 09:43:29AM +0200, Stefan Bader wrote:
> As of today, this change is upstream but as two separate patches. I took the
> liberty to do provide them as backported in your name.
> Both have the necessary tags to make them go upstream stable and should come in
> one of the next stable releases. But due to the fact that an installation is not
> possible without that changes on the affected hw, I think it is sensible to take
> them pre-stable. Changes are limited to specific device, so regression potential
> is low.
> 
> -Stefan
> 
> For the split up version of patches.
> Acked-by: Stefan Bader <stefan.bader at canonical.com>
> 
> On 07/02/2010 01:37 AM, Dave Walker wrote:
> > The Linux kernel doesn't currently support the latest generation of the
> > Macbook Pro (7,1), and also seems to have issues with the MacBook Mini.
> >  The issue presents itself as SATA support being unavailable once the
> > initrd tries to reset the SATA device.
> > 
> > There is a high priority open bug[0] regarding this, and at time of
> > writing 106 Launchpad users have shared that the bug affects them (aka
> > "me too")
> > 
> > The patch attached is heavily based on the work of Tejun Heo who has
> > submitted it upstream[1] (and submitted to stable).  The changes to this
> > patch include some slight refactoring to which make it compatiable with
> > the 2.6.32 code, and a manual refresh to apply cleanly to the lucid git
> > tree.
> > 
> > Based on looking what the patch achieves, i might suggest that the
> > regression potential is low as it essentially blacklists if the
> > following conditions are met:
> > - The vendor of the SATA device is nvidia
> > - The device is MCP89
> > - The hardware is Apple
> > - Particular subsystem id
> > 
> > If the blacklist is achieved, then ahci ignores the controller and
> > allows fall back to ata_generic.
> > 
> > Whilst it hasn't yet been applied to Linus' git tree (so can't send a
> > stable pull request), I would personally consider it paramount that this
> > patch is applied soon to the Lucid kernel with the hope of landing on
> > the 10.04.1 cd image.
> > 
> > As it stands, Ubuntu users with this hardware are unable to install from
> > the 10.04 cd image, and are currently dependent on using an community
> > built unofficial iso to be able to install Ubuntu.
> > 
> > Based on the high "me too" on Launchpad, I feel that many users are
> > highly dependent on this landing in an official cd spin, and am
> > therefore hoping that this patch is included and an SRU is created in
> > time for the 10.04.1 release.
> > 
> > The original author of this patch described it as:
> > "For yet unknown reason, MCP89 on MBP 7,1 doesn't work w/ ahci under
> > linux but the controller doesn't require explicit mode setting and
> > works fine with ata_generic.  Make ahci ignore the controller on MBP
> > 7,1 and let ata_generic take it for now." -- Tejun Heo
> > 
> > Thanks
> > 
> > [0] http://bugs.launchpad.net/bugs/576601
> > [1] http://article.gmane.org/gmane.linux.ide/46514

> From b09867f655ad766cf2aa8032d75423d681530cea Mon Sep 17 00:00:00 2001
> From: Tejun Heo <tj at kernel.org>
> Date: Thu, 17 Jun 2010 11:42:22 +0200
> Subject: [PATCH 1/2] (pre-stable) ahci,ata_generic: let ata_generic handle new MBP w/ MCP89
> MIME-Version: 1.0
> Content-Type: text/plain; charset=UTF-8
> Content-Transfer-Encoding: 8bit
> 
> BugLink: http://bugs.launchpad.net/bugs/576601
> 
> For yet unknown reason, MCP89 on MBP 7,1 doesn't work w/ ahci under
> linux but the controller doesn't require explicit mode setting and
> works fine with ata_generic.  Make ahci ignore the controller on MBP
> 7,1 and let ata_generic take it for now.
> 
> Reported in bko#15923.
> 
>   https://bugzilla.kernel.org/show_bug.cgi?id=15923
> 
> NVIDIA is investigating why ahci mode doesn't work.
> 
> Signed-off-by: Tejun Heo <tj at kernel.org>
> Cc: Peer Chen <pchen at nvidia.com>
> Cc: stable at kernel.org
> Reported-by: Anders ??sthus <grapz666 at gmail.com>
> Reported-by: Andreas Graf <andreas_graf at csgraf.de>
> Reported-by: Benoit Gschwind <gschwind at gnu-log.net>
> Reported-by: Damien Cassou <damien.cassou at gmail.com>
> Reported-by: tixetsal at juno.com
> Signed-off-by: Jeff Garzik <jgarzik at redhat.com>
> (backported from commit b09867f655ad766cf2aa8032d75423d681530cea upstream)
> Signed-off-by: Dave Walker (Daviey) <DaveWalker at ubuntu.com>
> ---
>  drivers/ata/ahci.c        |   10 ++++++++++
>  drivers/ata/ata_generic.c |    6 ++++++
>  include/linux/pci_ids.h   |    1 +
>  3 files changed, 17 insertions(+), 0 deletions(-)
> 
> diff --git a/drivers/ata/ahci.c b/drivers/ata/ahci.c
> index cb05205..e3d9816 100644
> --- a/drivers/ata/ahci.c
> +++ b/drivers/ata/ahci.c
> @@ -3037,6 +3037,16 @@ static int ahci_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
>  	if (pdev->vendor == PCI_VENDOR_ID_MARVELL && !marvell_enable)
>  		return -ENODEV;
>  
> +	/*
> +	 * For some reason, MCP89 on MacBook 7,1 doesn't work with
> +	 * ahci, use ata_generic instead.
> +	 */
> +	if (pdev->vendor == PCI_VENDOR_ID_NVIDIA &&
> +	    pdev->device == PCI_DEVICE_ID_NVIDIA_NFORCE_MCP89_SATA &&
> +	    pdev->subsystem_vendor == PCI_VENDOR_ID_APPLE &&
> +	    pdev->subsystem_device == 0xcb89)
> +		return -ENODEV;
> +
>  	/* acquire resources */
>  	rc = pcim_enable_device(pdev);
>  	if (rc)
> diff --git a/drivers/ata/ata_generic.c b/drivers/ata/ata_generic.c
> index ecfd22b..b5aa93c 100644
> --- a/drivers/ata/ata_generic.c
> +++ b/drivers/ata/ata_generic.c
> @@ -168,6 +168,12 @@ static struct pci_device_id ata_generic[] = {
>  	{ PCI_DEVICE(PCI_VENDOR_ID_VIA,    PCI_DEVICE_ID_VIA_82C561), },
>  	{ PCI_DEVICE(PCI_VENDOR_ID_OPTI,   PCI_DEVICE_ID_OPTI_82C558), },
>  	{ PCI_DEVICE(PCI_VENDOR_ID_CENATEK,PCI_DEVICE_ID_CENATEK_IDE), },
> +	/*
> +	 * For some reason, MCP89 on MacBook 7,1 doesn't work with
> +	 * ahci, use ata_generic instead.
> +	 */
> +	{ PCI_VENDOR_ID_NVIDIA, PCI_DEVICE_ID_NVIDIA_NFORCE_MCP89_SATA,
> +	  PCI_VENDOR_ID_APPLE, 0xcb89, },
>  	{ PCI_DEVICE(PCI_VENDOR_ID_TOSHIBA,PCI_DEVICE_ID_TOSHIBA_PICCOLO), },
>  	{ PCI_DEVICE(PCI_VENDOR_ID_TOSHIBA,PCI_DEVICE_ID_TOSHIBA_PICCOLO_1), },
>  	{ PCI_DEVICE(PCI_VENDOR_ID_TOSHIBA,PCI_DEVICE_ID_TOSHIBA_PICCOLO_2),  },
> diff --git a/include/linux/pci_ids.h b/include/linux/pci_ids.h
> index 63adefe..84f60fc 100644
> --- a/include/linux/pci_ids.h
> +++ b/include/linux/pci_ids.h
> @@ -1262,6 +1262,7 @@
>  #define PCI_DEVICE_ID_NVIDIA_NFORCE_MCP77_IDE       0x0759
>  #define PCI_DEVICE_ID_NVIDIA_NFORCE_MCP73_SMBUS     0x07D8
>  #define PCI_DEVICE_ID_NVIDIA_NFORCE_MCP79_SMBUS     0x0AA2
> +#define PCI_DEVICE_ID_NVIDIA_NFORCE_MCP89_SATA	    0x0D85
>  
>  #define PCI_VENDOR_ID_IMS		0x10e0
>  #define PCI_DEVICE_ID_IMS_TT128		0x9128
> -- 
> 1.7.0.4
> 

> From 281f43a7aa754ad6a4eda9077548a532c147a246 Mon Sep 17 00:00:00 2001
> From: Tejun Heo <tj at kernel.org>
> Date: Tue, 22 Jun 2010 12:27:26 +0200
> Subject: [PATCH 2/2] (pre-stable) ata_generic: implement ATA_GEN_* flags and force enable DMA on MBP 7,1
> MIME-Version: 1.0
> Content-Type: text/plain; charset=UTF-8
> Content-Transfer-Encoding: 8bit
> 
> BugLink: http://bugs.launchpad.net/bugs/576601
> 
> IDE mode of MCP89 on MBP 7,1 doesn't set DMA enable bits in the BMDMA
> status register.  Make the following changes to work around the problem.
> 
> * Instead of using hard coded 1 in id->driver_data as class code
>   match, use ATA_GEN_CLASS_MATCH and carry the matched id in
>   host->private_data.
> 
> * Instead of matching PCI_VENDOR_ID_CENATEK, use ATA_GEN_FORCE_DMA
>   flag in id instead.
> 
> * Add ATA_GEN_FORCE_DMA to the id entry of MBP 7,1.
> 
> Signed-off-by: Tejun Heo <tj at kernel.org>
> Cc: Peer Chen <pchen at nvidia.com>
> Cc: stable at kernel.org
> Reported-by: Anders ??sthus <grapz666 at gmail.com>
> Reported-by: Andreas Graf <andreas_graf at csgraf.de>
> Reported-by: Benoit Gschwind <gschwind at gnu-log.net>
> Reported-by: Damien Cassou <damien.cassou at gmail.com>
> Reported-by: tixetsal at juno.com
> Signed-off-by: Jeff Garzik <jgarzik at redhat.com>
> (backported from commit 281f43a7aa754ad6a4eda9077548a532c147a246 upstream)
> Signed-off-by: Dave Walker (Daviey) <DaveWalker at ubuntu.com>
> ---
>  drivers/ata/ata_generic.c |   26 +++++++++++++++++++-------
>  1 files changed, 19 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/ata/ata_generic.c b/drivers/ata/ata_generic.c
> index b5aa93c..99e7196 100644
> --- a/drivers/ata/ata_generic.c
> +++ b/drivers/ata/ata_generic.c
> @@ -32,6 +32,11 @@
>   *	A generic parallel ATA driver using libata
>   */
>  
> +enum {
> +	ATA_GEN_CLASS_MATCH		= (1 << 0),
> +	ATA_GEN_FORCE_DMA		= (1 << 1),
> +};
> +
>  /**
>   *	generic_set_mode	-	mode setting
>   *	@link: link to set up
> @@ -46,13 +51,17 @@
>  static int generic_set_mode(struct ata_link *link, struct ata_device **unused)
>  {
>  	struct ata_port *ap = link->ap;
> +	const struct pci_device_id *id = ap->host->private_data;
>  	int dma_enabled = 0;
>  	struct ata_device *dev;
>  	struct pci_dev *pdev = to_pci_dev(ap->host->dev);
>  
> -	/* Bits 5 and 6 indicate if DMA is active on master/slave */
> -	if (ap->ioaddr.bmdma_addr)
> +	if (id->driver_data & ATA_GEN_FORCE_DMA) {
> +		dma_enabled = 0xff;
> +	} else if (ap->ioaddr.bmdma_addr) {
> +		/* Bits 5 and 6 indicate if DMA is active on master/slave */
>  		dma_enabled = ioread8(ap->ioaddr.bmdma_addr + ATA_DMA_STATUS);
> +	}
>  
>  	if (pdev->vendor == PCI_VENDOR_ID_CENATEK)
>  		dma_enabled = 0xFF;

I think this patch should have removed the conditional above as it is
now encoded into the driver_data field for the Cenatek devices.  Though
it remaining is benign.  Tejun you probabally want to kill that for the
upstream version.

> @@ -126,7 +135,7 @@ static int ata_generic_init_one(struct pci_dev *dev, const struct pci_device_id
>  	const struct ata_port_info *ppi[] = { &info, NULL };
>  
>  	/* Don't use the generic entry unless instructed to do so */
> -	if (id->driver_data == 1 && all_generic_ide == 0)
> +	if ((id->driver_data & ATA_GEN_CLASS_MATCH) && all_generic_ide == 0)
>  		return -ENODEV;
>  
>  	/* Devices that need care */
> @@ -155,7 +164,7 @@ static int ata_generic_init_one(struct pci_dev *dev, const struct pci_device_id
>  			return rc;
>  		pcim_pin_device(dev);
>  	}
> -	return ata_pci_sff_init_one(dev, ppi, &generic_sht, NULL);
> +	return ata_pci_sff_init_one(dev, ppi, &generic_sht, (void *)id);
>  }
>  
>  static struct pci_device_id ata_generic[] = {
> @@ -167,18 +176,21 @@ static struct pci_device_id ata_generic[] = {
>  	{ PCI_DEVICE(PCI_VENDOR_ID_HINT,   PCI_DEVICE_ID_HINT_VXPROII_IDE), },
>  	{ PCI_DEVICE(PCI_VENDOR_ID_VIA,    PCI_DEVICE_ID_VIA_82C561), },
>  	{ PCI_DEVICE(PCI_VENDOR_ID_OPTI,   PCI_DEVICE_ID_OPTI_82C558), },
> -	{ PCI_DEVICE(PCI_VENDOR_ID_CENATEK,PCI_DEVICE_ID_CENATEK_IDE), },
> +	{ PCI_DEVICE(PCI_VENDOR_ID_CENATEK,PCI_DEVICE_ID_CENATEK_IDE),
> +	  .driver_data = ATA_GEN_FORCE_DMA },
>  	/*
>  	 * For some reason, MCP89 on MacBook 7,1 doesn't work with
>  	 * ahci, use ata_generic instead.
>  	 */
>  	{ PCI_VENDOR_ID_NVIDIA, PCI_DEVICE_ID_NVIDIA_NFORCE_MCP89_SATA,
> -	  PCI_VENDOR_ID_APPLE, 0xcb89, },
> +	  PCI_VENDOR_ID_APPLE, 0xcb89,
> +	  .driver_data = ATA_GEN_FORCE_DMA },
>  	{ PCI_DEVICE(PCI_VENDOR_ID_TOSHIBA,PCI_DEVICE_ID_TOSHIBA_PICCOLO), },
>  	{ PCI_DEVICE(PCI_VENDOR_ID_TOSHIBA,PCI_DEVICE_ID_TOSHIBA_PICCOLO_1), },
>  	{ PCI_DEVICE(PCI_VENDOR_ID_TOSHIBA,PCI_DEVICE_ID_TOSHIBA_PICCOLO_2),  },
>  	/* Must come last. If you add entries adjust this table appropriately */
> -	{ PCI_ANY_ID,		PCI_ANY_ID,			   PCI_ANY_ID, PCI_ANY_ID, PCI_CLASS_STORAGE_IDE << 8, 0xFFFFFF00UL, 1},
> +	{ PCI_DEVICE_CLASS(PCI_CLASS_STORAGE_IDE << 8, 0xFFFFFF00UL),
> +	  .driver_data = ATA_GEN_CLASS_MATCH },
>  	{ 0, },
>  };

Looks good for Ubuntu.

Acked-by: Andy Whitcroft <apw at canonical.com>

-apw




More information about the kernel-team mailing list