[PATCH] mmc: sunxi: Add support to the Allwinner A83T

[PATCH] mmc: sunxi: Add support to the Allwinner A83T

[Jean-Francois Moine]

The A83T has different clock delays.
The values have been adapted from the Banana Pi M3 driver.

Signed-off-by: Jean-Francois Moine <moinejf@free.fr>
---
 Documentation/devicetree/bindings/mmc/sunxi-mmc.txt |  3 ++-
 drivers/mmc/host/sunxi-mmc.c                        | 12 +++++++++++-
 2 files changed, 13 insertions(+), 2 deletions(-)

diff --git a/Documentation/devicetree/bindings/mmc/sunxi-mmc.txt b/Documentation/devicetree/bindings/mmc/sunxi-mmc.txt
index 4bf41d8..45b8520 100644
--- a/Documentation/devicetree/bindings/mmc/sunxi-mmc.txt
+++ b/Documentation/devicetree/bindings/mmc/sunxi-mmc.txt
@@ -8,7 +8,8 @@ as the speed of SD standard 3.0.
 Absolute maximum transfer rate is 200MB/s
 
 Required properties:
- - compatible : "allwinner,sun4i-a10-mmc" or "allwinner,sun5i-a13-mmc"
+ - compatible : "allwinner,sun4i-a10-mmc", "allwinner,sun5i-a13-mmc",
+	"allwinner,sun8i-a83t-mmc" or "allwinner,sun9i-a80-mmc"
  - reg : mmc controller base registers
  - clocks : a list with 4 phandle + clock specifier pairs
  - clock-names : must contain "ahb", "mmc", "output" and "sample"
diff --git a/drivers/mmc/host/sunxi-mmc.c b/drivers/mmc/host/sunxi-mmc.c
index 7fc8b7a..707e705 100644
--- a/drivers/mmc/host/sunxi-mmc.c
+++ b/drivers/mmc/host/sunxi-mmc.c
@@ -941,6 +941,7 @@ static int sunxi_mmc_card_busy(struct mmc_host *mmc)
 static const struct of_device_id sunxi_mmc_of_match[] = {
 	{ .compatible = "allwinner,sun4i-a10-mmc", },
 	{ .compatible = "allwinner,sun5i-a13-mmc", },
+	{ .compatible = "allwinner,sun8i-a83t-mmc", },
 	{ .compatible = "allwinner,sun9i-a80-mmc", },
 	{ /* sentinel */ }
 };
@@ -962,10 +963,17 @@ static const struct sunxi_mmc_clk_delay sunxi_mmc_clk_delays[] = {
 	[SDXC_CLK_25M]		= { .output = 180, .sample =  75 },
 	[SDXC_CLK_50M]		= { .output =  90, .sample = 120 },
 	[SDXC_CLK_50M_DDR]	= { .output =  60, .sample = 120 },
-	/* Value from A83T "new timing mode". Works but might not be right. */
 	[SDXC_CLK_50M_DDR_8BIT]	= { .output =  90, .sample = 180 },
 };
 
+static const struct sunxi_mmc_clk_delay sun8i_a83t_mmc_clk_delays[] = {
+	[SDXC_CLK_400K]		= { .output = 180, .sample = 180 },
+	[SDXC_CLK_25M]		= { .output = 180, .sample =  50 },
+	[SDXC_CLK_50M]		= { .output =  60, .sample =  50 },
+	[SDXC_CLK_50M_DDR]	= { .output =  180, .sample = 90 },
+	[SDXC_CLK_50M_DDR_8BIT]	= { .output =  180, .sample = 90 },
+};
+
 static const struct sunxi_mmc_clk_delay sun9i_mmc_clk_delays[] = {
 	[SDXC_CLK_400K]		= { .output = 180, .sample = 180 },
 	[SDXC_CLK_25M]		= { .output = 180, .sample =  75 },
@@ -987,6 +995,8 @@ static int sunxi_mmc_resource_request(struct sunxi_mmc_host *host,
 
 	if (of_device_is_compatible(np, "allwinner,sun9i-a80-mmc"))
 		host->clk_delays = sun9i_mmc_clk_delays;
+	else if (of_device_is_compatible(np, "allwinner,sun8i-a83t-mmc"))
+		host->clk_delays = sun8i_a83t_mmc_clk_delays;
 	else
 		host->clk_delays = sunxi_mmc_clk_delays;
 
-- 
2.8.3

[PATCH] mmc: sunxi: Add support to the Allwinner A83T

[Chen-Yu Tsai]

Hi,

On Sat, Jun 4, 2016 at 6:01 PM, Jean-Francois Moine <moinejf@free.fr> wrote:
> The A83T has different clock delays.
> The values have been adapted from the Banana Pi M3 driver.
>
> Signed-off-by: Jean-Francois Moine <moinejf@free.fr>
> ---
>  Documentation/devicetree/bindings/mmc/sunxi-mmc.txt |  3 ++-
>  drivers/mmc/host/sunxi-mmc.c                        | 12 +++++++++++-
>  2 files changed, 13 insertions(+), 2 deletions(-)
>
> diff --git a/Documentation/devicetree/bindings/mmc/sunxi-mmc.txt b/Documentation/devicetree/bindings/mmc/sunxi-mmc.txt
> index 4bf41d8..45b8520 100644
> --- a/Documentation/devicetree/bindings/mmc/sunxi-mmc.txt
> +++ b/Documentation/devicetree/bindings/mmc/sunxi-mmc.txt
> @@ -8,7 +8,8 @@ as the speed of SD standard 3.0.
>  Absolute maximum transfer rate is 200MB/s
>
>  Required properties:
> - - compatible : "allwinner,sun4i-a10-mmc" or "allwinner,sun5i-a13-mmc"
> + - compatible : "allwinner,sun4i-a10-mmc", "allwinner,sun5i-a13-mmc",
> +       "allwinner,sun8i-a83t-mmc" or "allwinner,sun9i-a80-mmc"
>   - reg : mmc controller base registers
>   - clocks : a list with 4 phandle + clock specifier pairs
>   - clock-names : must contain "ahb", "mmc", "output" and "sample"
> diff --git a/drivers/mmc/host/sunxi-mmc.c b/drivers/mmc/host/sunxi-mmc.c
> index 7fc8b7a..707e705 100644
> --- a/drivers/mmc/host/sunxi-mmc.c
> +++ b/drivers/mmc/host/sunxi-mmc.c
> @@ -941,6 +941,7 @@ static int sunxi_mmc_card_busy(struct mmc_host *mmc)
>  static const struct of_device_id sunxi_mmc_of_match[] = {
>         { .compatible = "allwinner,sun4i-a10-mmc", },
>         { .compatible = "allwinner,sun5i-a13-mmc", },
> +       { .compatible = "allwinner,sun8i-a83t-mmc", },
>         { .compatible = "allwinner,sun9i-a80-mmc", },
>         { /* sentinel */ }
>  };
> @@ -962,10 +963,17 @@ static const struct sunxi_mmc_clk_delay sunxi_mmc_clk_delays[] = {
>         [SDXC_CLK_25M]          = { .output = 180, .sample =  75 },
>         [SDXC_CLK_50M]          = { .output =  90, .sample = 120 },
>         [SDXC_CLK_50M_DDR]      = { .output =  60, .sample = 120 },
> -       /* Value from A83T "new timing mode". Works but might not be right. */

Please don't remove this. It was how I found a working value. Until we figure
out an actual correct value, the warning should stay there.

>         [SDXC_CLK_50M_DDR_8BIT] = { .output =  90, .sample = 180 },
>  };
>
> +static const struct sunxi_mmc_clk_delay sun8i_a83t_mmc_clk_delays[] = {
> +       [SDXC_CLK_400K]         = { .output = 180, .sample = 180 },
> +       [SDXC_CLK_25M]          = { .output = 180, .sample =  50 },
> +       [SDXC_CLK_50M]          = { .output =  60, .sample =  50 },

These 2 don't look right.

> +       [SDXC_CLK_50M_DDR]      = { .output =  180, .sample = 90 },
> +       [SDXC_CLK_50M_DDR_8BIT] = { .output =  180, .sample = 90 },

Nor does this one. As explained in the comment from the clock driver:

    We can only outphase the clocks by multiple of the PLL's period.

    Since the MMC clock in only a divider, and the formula to get the
    outphasing in degrees is deg = 360 * delta / period

Now, for the fishy ones I mentioned, MMC clock = card clock (or card * 2
for DDR 8 bit), and parent clock is PLL6 @ 600 MHz. A step would be 15,
30 or 60 degrees. Thus you cannot get an outphasing of 50 degrees, or 90
degrees for 50M DDR 8bit.

Could you provide a link to the original driver? Or perhaps just the
original delay values?

The values do work better on my Cubietruck plus, giving a throughput
of 36 MB/s read, instead of 17 MB/s with the original values. However
that is still only half of what it should be capable of. I will test
my BPI M3 tomorrow.


Regards
ChenYu

> +};
> +
>  static const struct sunxi_mmc_clk_delay sun9i_mmc_clk_delays[] = {
>         [SDXC_CLK_400K]         = { .output = 180, .sample = 180 },
>         [SDXC_CLK_25M]          = { .output = 180, .sample =  75 },
> @@ -987,6 +995,8 @@ static int sunxi_mmc_resource_request(struct sunxi_mmc_host *host,
>
>         if (of_device_is_compatible(np, "allwinner,sun9i-a80-mmc"))
>                 host->clk_delays = sun9i_mmc_clk_delays;
> +       else if (of_device_is_compatible(np, "allwinner,sun8i-a83t-mmc"))
> +               host->clk_delays = sun8i_a83t_mmc_clk_delays;
>         else
>                 host->clk_delays = sunxi_mmc_clk_delays;
>
> --
> 2.8.3
>

[PATCH] mmc: sunxi: Add support to the Allwinner A83T

[Jean-Francois Moine]

On Sun, 5 Jun 2016 20:58:42 +0800
Chen-Yu Tsai <wens@csie.org> wrote:

> > +static const struct sunxi_mmc_clk_delay sun8i_a83t_mmc_clk_delays[] = {
> > +       [SDXC_CLK_400K]         = { .output = 180, .sample = 180 },
> > +       [SDXC_CLK_25M]          = { .output = 180, .sample =  50 },
> > +       [SDXC_CLK_50M]          = { .output =  60, .sample =  50 },
> 
> These 2 don't look right.
> 
> > +       [SDXC_CLK_50M_DDR]      = { .output =  180, .sample = 90 },
> > +       [SDXC_CLK_50M_DDR_8BIT] = { .output =  180, .sample = 90 },
> 
> Nor does this one. As explained in the comment from the clock driver:
> 
>     We can only outphase the clocks by multiple of the PLL's period.
> 
>     Since the MMC clock in only a divider, and the formula to get the
>     outphasing in degrees is deg = 360 * delta / period
> 
> Now, for the fishy ones I mentioned, MMC clock = card clock (or card * 2
> for DDR 8 bit), and parent clock is PLL6 @ 600 MHz. A step would be 15,
> 30 or 60 degrees. Thus you cannot get an outphasing of 50 degrees, or 90
> degrees for 50M DDR 8bit.
> 
> Could you provide a link to the original driver? Or perhaps just the
> original delay values?
> 
> The values do work better on my Cubietruck plus, giving a throughput
> of 36 MB/s read, instead of 17 MB/s with the original values. However
> that is still only half of what it should be capable of. I will test
> my BPI M3 tomorrow.

Hi,

Yes, you are right, my values are wrong. Please, forget about this patch.
I did not look far enough in the clock code.

Anyway, I had timeouts in accessing the MMC because the BPI-M3 legacy
u-boot sets the pll-periph rate to 1.2GHz.
Then, according to the comment in the BPI-M3 original driver
(https://github.com/BPI-SINOVOIP/BPI-M3-bsp/blob/master/linux-sunxi/drivers/mmc/host/sunxi-mci.c),
if "//in 1.2GHz pll_periph only [MMC_CLK-50M] is correct",
setting the clock delays to 6 and 7 would be, for us:

	[SDXC_CLK_50M]          = { .output =  90, .sample =  105 },

-- 
Ken ar c'henta?	|	      ** Breizh ha Linux atav! **
Jef		|		http://moinejf.free.fr/