[PATCH v3 0/9] Migrate Tegra to common clock framework

[Stephen Warren]

On 01/04/2013 10:22 AM, Stephen Warren wrote:
> On 01/04/2013 02:40 AM, Prashant Gaikwad wrote:
>> This patchset does following:
>> 1. Decompose single tegra clock structure into multiple clocks.
>> 2. Try to use standard clock types supported by common clock framework.
>> 3. Use dynamic initialization.
>> 4. Move all clock code to drivers/clk/tegra from mach-tegra.
>> 5. Add device tree support for Tegra20 and Tegra30 clocks.
>> 6. Remove all legacy clock code from mach-tegra.
> 
> I think there are bugs here. I applied all your clock patches on top of
> Tegra's for-next (see list below), and found that the following don't
> work on Springbank:
> 
> * HDMI display
> * Audio playback
> * WiFi

(BTW, I stopped Cc'ing linux-kernel@, but added linux-tegra@ instead...)

Prashant, some updated testing results based off the "dev/ccf" branch
you sent me on our internal git server:

The following work, on both Tegra20/30 unless otherwise specified:

* MMC/SD, I2C, audio, HDMI display (Tegra20 only; on Springbank), USB
(Tegra20 only; not implemented on Tegra30).

I didn't test the following yet:

* LCD display on Tegra20.

The following are broken:

1)

SPI on either chip. This is because the SPI drivers call clk_get() with
a specific clock name, and the DT patches you created don't have a
clk-names property for any of the SPI nodes, hence this fails. However,
since there's only a single clock to the SPI controllers, I think the
drivers should be fixed not to request a specific clock name. The
following makes SPI work on TrimSlice (Tegra20) and Cardhu (Tegra30):

> diff --git a/drivers/spi/spi-tegra20-sflash.c b/drivers/spi/spi-tegra20-sflash.c
> index 02feaa5..e5dce91 100644
> --- a/drivers/spi/spi-tegra20-sflash.c
> +++ b/drivers/spi/spi-tegra20-sflash.c
> @@ -525,7 +525,7 @@ static int tegra_sflash_probe(struct platform_device *pdev)
>                 goto exit_free_master;
>         }
>  
> -       tsd->clk = devm_clk_get(&pdev->dev, "spi");
> +       tsd->clk = devm_clk_get(&pdev->dev, NULL);
>         if (IS_ERR(tsd->clk)) {
>                 dev_err(&pdev->dev, "can not get clock\n");
>                 ret = PTR_ERR(tsd->clk);
> diff --git a/drivers/spi/spi-tegra20-slink.c b/drivers/spi/spi-tegra20-slink.c
> index fa208a5..e255e7a 100644
> --- a/drivers/spi/spi-tegra20-slink.c
> +++ b/drivers/spi/spi-tegra20-slink.c
> @@ -1191,7 +1191,7 @@ static int tegra_slink_probe(struct platform_device *pdev)
>                 goto exit_free_master;
>         }
>  
> -       tspi->clk = devm_clk_get(&pdev->dev, "slink");
> +       tspi->clk = devm_clk_get(&pdev->dev, NULL);
>         if (IS_ERR(tspi->clk)) {
>                 dev_err(&pdev->dev, "can not get clock\n");
>                 ret = PTR_ERR(tspi->clk);

I think you should include that in your series before the patches which
add clocks into the DT.

2)

HDMI and LCD display are both broken on Cardhu/Tegra30.

I found that the DT lists the HDMI clock's parent as pll_d2_out0, yet
the clock driver lists pll_d2 as a valid parent. I think the following
patch fixes this correctly?

> diff --git a/drivers/clk/tegra/clk-tegra30.c b/drivers/clk/tegra/clk-tegra30.c
> index a52b80c..661b9e1 100644
> --- a/drivers/clk/tegra/clk-tegra30.c
> +++ b/drivers/clk/tegra/clk-tegra30.c
> @@ -1385,8 +1385,8 @@ static const char *mux_pllpc_clkm_clk32k[] = { "pll_p", "pll_c", "clk_m",
>  static const char *mux_pllmcpa[] = { "pll_m", "pll_c", "pll_p", "pll_a" };
>  static const char *mux_pllpdc_clkm[] = { "pll_p", "pll_d", "pll_c", "clk_m" };
>  static const char *mux_pllp_clkm[] = { "pll_p", "unused", "unused", "clk_m" };
> -static const char *mux_pllpmdacd2_clkm[] = { "pll_p", "pll_m", "pll_d",
> -                                            "pll_a", "pll_c", "pll_d2",
> +static const char *mux_pllpmdacd2_clkm[] = { "pll_p", "pll_m", "pll_d_out0",
> +                                            "pll_a", "pll_c", "pll_d2_out0",
>                                              "clk_m" };
>  static const char *mux_plla_clk32k_pllp_clkm_plle[] = { "pll_a_out0",
>                                                         "clk_32k", "pll_p",

(are there any other similar problems in the code?)

After that, if I comment out:

late_initcall(clk_disable_unused);

... then either HDMI or LCD output on Cardhu work OK (or rather in
HDMI's case, as well as can be expected due incorrect memory controller
programming and hence display underflow at the moment). I haven't had a
chance to investigate why that initcall is causing problems (the problem
being a HW hang during boot in the HDMI case, and simply no visible
output in the LCD case, but the system otherwise works OK).

For reference, the patches to enable LCD/HDMI on Cardhu can be found in:

git://nv-tegra.nvidia.com/user/swarren/linux-2.6 linux-next_common

3)

WiFi doesn't work on either chip (Springbank or Cardhu, but I expect
Ventana has the same issue). I haven't been able to investigate why yet.

4)

I noticed that the PCIe driver (on Trimslice/Tegra20) was complaining
about clk_get() failing or similar. I haven't been able to investigate
why yet.

For reference, all of these problems are all introduced by your commit
"arm: tegra: Migrate to new clock code"; everything works immediately
before that.

Let me know if you find any solutions to these; I guess just update your
branch. Otherwise, I'll keep investigating tomorrow.