Re: [PATCH/RFC] ahci: add support for VIA VT8251

[Sergey Vlasov <vsu@altlinux.ru>]

[-- Attachment #1: Type: text/plain, Size: 4144 bytes --]

On Tue, 11 Apr 2006 19:15:45 +0200 Bastiaan Jacques wrote:

> This patch adds AHCI support for the VIA VT8251 chipset. The patch does so by:
> 	1) Adding the PCI device ID.
> 	2) Adding a workaround in ahci_probe_reset() for a VIA hardware bug.
> 
> Signed-off-by: Bastiaan Jacques <b.jacques@planet.nl>
> ---
> 
> The patch is based upon the patch[1] provided by VIA. With the softreset and 
> probing work that has been merged, this patch has become much simpler.
> 
> I am happy to report that without the workaround in ahci_probe_reset(), the 
> chipset works quite well; both my SATA drives are recognised. However, 
> softreset does not work (but hardreset does), and ata_std_probeinit delays, 
> for my primary SATA drive/port,  due to the "busy" bug. A Command List 
> Override will unbusy the drive/port.
> 
> Since the CLO sequence also appears in ahci_softreset(), I have moved that 
> code into a new ahci_clo() function.
> 
> This patch is against 2.6.17-rc1 (and -mm2).
> 
> This patch has only been tested by yours truly. Whilst I am waiting for 
> feedback from other people, I am sending this patch to solicit comments.
> 
> NOTE: My email client cuts off some lines (of context) in my patch because 
> they are longer than 78 characters. Therefore, I am making the patch 
> available from my web server [2], as well.
> 
> Bastiaan
> 
> [1] http://lkml.org/lkml/2005/12/11/204
> [2] http://mirrors.pagefault.net/linux-via-ahci-latest.patch
> 
> --- drivers/scsi/ahci.c.orig	2006-04-11 18:58:32.000000000 +0200
> +++ drivers/scsi/ahci.c	2006-04-11 12:44:41.000000000 +0200
> @@ -297,6 +297,8 @@
>  	  board_ahci }, /* ATI SB600 non-raid */
>  	{ PCI_VENDOR_ID_ATI, 0x4381, PCI_ANY_ID, PCI_ANY_ID, 0, 0,
>  	  board_ahci }, /* ATI SB600 raid */
> +	{ PCI_VENDOR_ID_VIA, 0x3349, PCI_ANY_ID, PCI_ANY_ID, 0, 0,
> +	  board_ahci }, /* VT8251 */

You can add a separate host type (e.g., board_via_vt8251_ahci), and then
add an entry for this type to ahci_port_info[].  The entry would have a
special flag in .host_flags, but the same .port_ops, so there will not
be excessive duplication like in the original vendor patch.

Look like other drivers which need workarounds for hardware bugs (e.g.,
sata_sil) do this.

>  	{ }	/* terminate list */
>  };
> 
> @@ -535,6 +537,17 @@
>  	return -1;
>  }
> 
> +static void ahci_clo(struct ata_port *ap)
> +{
> +	void __iomem *port_mmio = (void __iomem *) ap->ioaddr.cmd_addr;
> +	u32 tmp;
> +
> +	tmp = readl(port_mmio + PORT_CMD);
> +	tmp |= PORT_CMD_CLO;
> +	writel(tmp, port_mmio + PORT_CMD);
> +	readl(port_mmio + PORT_CMD); /* flush */
> +}
> +
>  static int ahci_softreset(struct ata_port *ap, int verbose, unsigned int 
> *class)
>  {
>  	struct ahci_host_priv *hpriv = ap->host_set->private_data;
> @@ -565,7 +578,6 @@
>  	/* check BUSY/DRQ, perform Command List Override if necessary */
>  	ahci_tf_read(ap, &tf);
>  	if (tf.command & (ATA_BUSY | ATA_DRQ)) {
> -		u32 tmp;
> 

The following empty line should probably be deleted too.

>  		if (!(hpriv->cap & HOST_CAP_CLO)) {
>  			rc = -EIO;
> @@ -573,10 +585,7 @@
>  			goto fail_restart;
>  		}
> 
> -		tmp = readl(port_mmio + PORT_CMD);
> -		tmp |= PORT_CMD_CLO;
> -		writel(tmp, port_mmio + PORT_CMD);
> -		readl(port_mmio + PORT_CMD); /* flush */
> +		ahci_clo(ap);
> 
>  		if (ahci_poll_register(port_mmio + PORT_CMD, PORT_CMD_CLO, 0x0,
>  				       1, 500)) {
> @@ -695,6 +704,16 @@
> 
>  static int ahci_probe_reset(struct ata_port *ap, unsigned int *classes)
>  {
> +	struct pci_dev *pdev = to_pci_dev(ap->host_set->dev);
> +
> +	/* Workaround for VIA VT8251 "busy" bug */
> +	if (pdev->device == 0x3349) {

After adding a separate host type and a host flag this will be:

	if (ap->flags & AHCI_FLAG_RESET_NEEDS_CLO) {

> +		if (ata_busy_wait(ap, ATA_BUSY, 1000) & ATA_BUSY) {
> +			/* ATA_BUSY hasn't cleared, so send a CLO */
> +			ahci_clo(ap);
> +		}
> +	}
> +
>  	return ata_drive_probe_reset(ap, ata_std_probeinit,
>  				     ahci_softreset, ahci_hardreset,
>  				     ahci_postreset, classes);

[-- Attachment #2: Type: application/pgp-signature, Size: 190 bytes --]