Re: [PATCH 3/5] ata: add Broadcom AHCI SATA3 driver for STB chips

[Florian Fainelli <f.fainelli@gmail.com>]

On 18/03/15 18:23, Brian Norris wrote:
> Pretty straightforward driver, using the nice library-ization of the
> generic ahci_platform driver.
> 
> Signed-off-by: Brian Norris <computersforpeace@gmail.com>
> ---
>  drivers/ata/Kconfig        |   9 +++
>  drivers/ata/Makefile       |   1 +
>  drivers/ata/sata_brcmstb.c | 148 +++++++++++++++++++++++++++++++++++++++++++++
>  3 files changed, 158 insertions(+)
>  create mode 100644 drivers/ata/sata_brcmstb.c
> 
> diff --git a/drivers/ata/Kconfig b/drivers/ata/Kconfig
> index 5f601553b9b0..33d4b3031705 100644
> --- a/drivers/ata/Kconfig
> +++ b/drivers/ata/Kconfig
> @@ -98,6 +98,15 @@ config SATA_AHCI_PLATFORM
>  
>  	  If unsure, say N.
>  
> +config SATA_BRCMSTB
> +	tristate "Broadcom STB AHCI SATA support"
> +	depends on ARCH_BRCMSTB

We would probably want a select PHY_BRCMSTB_SATA here?

[snip]

> +
> +static void brcm_sata3_init_config(struct brcm_ahci_priv *priv)
> +{
> +	/* Configure endianness */
> +	writel((DATA_ENDIAN << 4) | (DATA_ENDIAN << 2) | (MMIO_ENDIAN << 0),
> +		priv->top_ctrl + SATA_TOP_CTRL_BUS_CTRL);

AFAIR, this portion of the initialization must be done in the host-CPU
native endianness, so __raw_writel() would be more appropriate, or we
could use Kevin's conditional I/O accessors and do either ioread32() or
ioread32be() based on the absence/presence of the "big-endian" property?


[snip]

> +
> +static const struct of_device_id ahci_of_match[] = {
> +	{.compatible = "brcm,sata3-ahci"},

The binding specifies brcm,bcm7445-ahci as a valid compatible string,
such that we would probably want to match it here for consistency.
-- 
Florian