[PATCH v4 1/8] of: Add NVIDIA Tegra SATA controller binding

[Hans de Goede]

Hi,

On 07/17/2014 01:42 PM, Hans de Goede wrote:
> Hi,
> 
> On 07/17/2014 12:52 PM, Thierry Reding wrote:
>> On Thu, Jul 17, 2014 at 12:23:47PM +0200, Hans de Goede wrote:
> 
> <snip>
> 
>>> The libahci_platform.c code / ahci_platform.c code is also used for
>>> devices going way back who may not yet be using the new clk framework,
>>> so where we need to use clk_get(dev, NULL); quoting from libahci_platform.c :
>>>
>>>         for (i = 0; i < AHCI_MAX_CLKS; i++) {
>>>                 /*
>>>                  * For now we must use clk_get(dev, NULL) for the first clock,
>>>                  * because some platforms (da850, spear13xx) are not yet
>>>                  * converted to use devicetree for clocks.  For new platforms
>>>                  * this is equivalent to of_clk_get(dev->of_node, 0).
>>>                  */
>>>                 if (i == 0)
>>>                         clk = clk_get(dev, NULL);
>>>                 else
>>>                         clk = of_clk_get(dev->of_node, i);
>>>
>>>                 if (IS_ERR(clk)) {
>>>                         rc = PTR_ERR(clk);
>>>                         if (rc == -EPROBE_DEFER)
>>>                                 goto err_out;
>>>                         break;
>>>                 }
>>>                 hpriv->clks[i] = clk;
>>>         }
>>>
>>> And there is no devm variant of that, nor is there one to get clocks by index.
>>> Note that we also need ahci_platform_put_resources for runtime pm support, so
>>> that one is going to stay around anyways and thus there is not that much value
>>> in fixing this.
>>>
>>> So although I like Thierry's idea, if we go this way (which sounds good), we
>>> should add support for taking a NULL ahci_platform_resources argument and in
>>> that case behave as before, esp. because of the platforms needing the old
>>> style clock handling. An advantage of doing this, is that we can simply patch
>>> all existing users to pass NULL.
>>
>> Isn't the "legacy" case really just this:
>>
>> 	static const char *const legacy_ahci_clocks[] = {
>> 		NULL
>> 	};
>>
>> 	static const struct ahci_platform_resources legacy_ahci_resources = {
>> 		.num_clocks = ARRAY_SIZE(legacy_ahci_clocks),
>> 		.clocks = legacy_ahci_clocks,
>> 	};
>>
>> ?
> 
> Ah yes that would work for the really legacy ones, as well as less legacy
> (full dts) ones with only one clk, we need to check if there are current
> users which use more then one clk (yes there are which is why MAX_CLKS
> was 3) and fixup those to pass in a correct ahci_platform_resources struct
> then.
> 
> The checking + fixing up will be a bit of extra work, but I think the end result
> will be quite nice. so I'm all in favor of this.

Correction, this is not going to work I'm afraid, as not all current dts files
set clock-names. So we need a fallback to get clks by index for compatibility
with old dts files.

At least: arch/arm/boot/dts/sun4i-a10.dtsi and arch/arm/boot/dts/sun7i-a20.dtsi
are affected. Note in this case the dts file is typically not burned into
a ROM or some such, so IMHO we could get away with requiring a new dts file.

So it might be worthwhile to still do a full check if all affected SoCs and
see if we can move over to using clock-names for all platforms with
more then 1 ahci/sata clk.

FWIW I've also just checked imx6q.dts which is the one which has 3 clocks, and
that one does define clk names.

Regards,

Hans