Re: [PATCH v6 4/5] clk: qcom: rpmh: Add support for SM8550 rpmh clocks

[Krzysztof Kozlowski]

On 06/01/2023 08:45, Krzysztof Kozlowski wrote:
> On 28/12/2022 19:59, Dmitry Baryshkov wrote:
>> On 28/12/2022 20:52, Bjorn Andersson wrote:
>>> On Wed, Dec 14, 2022 at 06:25:01PM +0200, Dmitry Baryshkov wrote:
>>>> On 07/12/2022 00:45, Abel Vesa wrote:
>>>>> Adds the RPMH clocks present in SM8550 SoC.
>>>>>
>>>>> Signed-off-by: Abel Vesa <abel.vesa@xxxxxxxxxx>
>>>>> Reviewed-by: Konrad Dybcio <konrad.dybcio@xxxxxxxxxx>
>>>>> ---
>>>>>    drivers/clk/qcom/clk-rpmh.c | 110 +++++++++++++++++++++++++++++-------
>>>>>    1 file changed, 90 insertions(+), 20 deletions(-)
>>>>>
>>>>> diff --git a/drivers/clk/qcom/clk-rpmh.c b/drivers/clk/qcom/clk-rpmh.c
>>>>> index 2c2ef4b6d130..ce81c76ed0fd 100644
>>>>> --- a/drivers/clk/qcom/clk-rpmh.c
>>>>> +++ b/drivers/clk/qcom/clk-rpmh.c
>>>>> @@ -130,6 +130,34 @@ static DEFINE_MUTEX(rpmh_clk_lock);
>>>>>    		},							\
>>>>>    	}
>>>>> +#define DEFINE_CLK_FIXED_FACTOR(_name, _parent_name, _div)		\
>>>>> +	static struct clk_fixed_factor clk_fixed_factor##_##_name = {	\
>>>>> +		.mult = 1,						\
>>>>> +		.div = _div,						\
>>>>> +		.hw.init = &(struct clk_init_data){			\
>>>>> +			.ops = &clk_fixed_factor_ops,			\
>>>>> +			.name = #_name,					\
>>>>> +			.parent_data =  &(const struct clk_parent_data){ \
>>>>> +				.fw_name = #_parent_name,		\
>>>>> +				.name = #_parent_name,			\
>>>>> +			},						\
>>>>> +			.num_parents = 1,				\
>>>>> +		},							\
>>>>> +	};								\
>>>>> +	static struct clk_fixed_factor clk_fixed_factor##_##_name##_ao = { \
>>>>> +		.mult = 1,						\
>>>>> +		.div = _div,						\
>>>>> +		.hw.init = &(struct clk_init_data){			\
>>>>> +			.ops = &clk_fixed_factor_ops,			\
>>>>> +			.name = #_name "_ao",				\
>>>>> +			.parent_data =  &(const struct clk_parent_data){ \
>>>>> +				.fw_name = #_parent_name "_ao",		\
>>>>> +				.name = #_parent_name "_ao",		\
>>>>> +			},						\
>>>>> +			.num_parents = 1,				\
>>>>> +		},							\
>>>>> +	}
>>>>> +
>>>>>    static inline struct clk_rpmh *to_clk_rpmh(struct clk_hw *_hw)
>>>>>    {
>>>>>    	return container_of(_hw, struct clk_rpmh, hw);
>>>>> @@ -345,6 +373,8 @@ DEFINE_CLK_RPMH_ARC(bi_tcxo, "xo.lvl", 0x3, 2);
>>>>>    DEFINE_CLK_RPMH_ARC(bi_tcxo, "xo.lvl", 0x3, 4);
>>>>>    DEFINE_CLK_RPMH_ARC(qlink, "qphy.lvl", 0x1, 4);
>>>>> +DEFINE_CLK_FIXED_FACTOR(bi_tcxo_div2, bi_tcxo, 2);
>>>>> +
>>>>>    DEFINE_CLK_RPMH_VRM(ln_bb_clk1, _a2, "lnbclka1", 2);
>>>>>    DEFINE_CLK_RPMH_VRM(ln_bb_clk2, _a2, "lnbclka2", 2);
>>>>>    DEFINE_CLK_RPMH_VRM(ln_bb_clk3, _a2, "lnbclka3", 2);
>>>>> @@ -366,6 +396,16 @@ DEFINE_CLK_RPMH_VRM(rf_clk2, _d, "rfclkd2", 1);
>>>>>    DEFINE_CLK_RPMH_VRM(rf_clk3, _d, "rfclkd3", 1);
>>>>>    DEFINE_CLK_RPMH_VRM(rf_clk4, _d, "rfclkd4", 1);
>>>>> +DEFINE_CLK_RPMH_VRM(clk1, _a1, "clka1", 1);
>>>>> +DEFINE_CLK_RPMH_VRM(clk2, _a1, "clka2", 1);
>>>>> +DEFINE_CLK_RPMH_VRM(clk3, _a1, "clka3", 1);
>>>>> +DEFINE_CLK_RPMH_VRM(clk4, _a1, "clka4", 1);
>>>>> +DEFINE_CLK_RPMH_VRM(clk5, _a1, "clka5", 1);
>>>>> +
>>>>> +DEFINE_CLK_RPMH_VRM(clk6, _a2, "clka6", 2);
>>>>> +DEFINE_CLK_RPMH_VRM(clk7, _a2, "clka7", 2);
>>>>> +DEFINE_CLK_RPMH_VRM(clk8, _a2, "clka8", 2);
>>>>> +
>>>>>    DEFINE_CLK_RPMH_VRM(div_clk1, _div2, "divclka1", 2);
>>>>>    DEFINE_CLK_RPMH_BCM(ce, "CE0");
>>>>> @@ -576,6 +616,33 @@ static const struct clk_rpmh_desc clk_rpmh_sm8450 = {
>>>>>    	.num_clks = ARRAY_SIZE(sm8450_rpmh_clocks),
>>>>>    };
>>>>> +static struct clk_hw *sm8550_rpmh_clocks[] = {
>>>>> +	[RPMH_CXO_PAD_CLK]      = &clk_rpmh_bi_tcxo_div2.hw,
>>>>> +	[RPMH_CXO_PAD_CLK_A]    = &clk_rpmh_bi_tcxo_div2_ao.hw,
>>>>> +	[RPMH_CXO_CLK]		= &clk_fixed_factor_bi_tcxo_div2.hw,
>>>>> +	[RPMH_CXO_CLK_A]	= &clk_fixed_factor_bi_tcxo_div2_ao.hw,
>>>>> +	[RPMH_LN_BB_CLK1]	= &clk_rpmh_clk6_a2.hw,
>>>>> +	[RPMH_LN_BB_CLK1_A]	= &clk_rpmh_clk6_a2_ao.hw,
>>>>> +	[RPMH_LN_BB_CLK2]	= &clk_rpmh_clk7_a2.hw,
>>>>> +	[RPMH_LN_BB_CLK2_A]	= &clk_rpmh_clk7_a2_ao.hw,
>>>>> +	[RPMH_LN_BB_CLK3]	= &clk_rpmh_clk8_a2.hw,
>>>>> +	[RPMH_LN_BB_CLK3_A]	= &clk_rpmh_clk8_a2_ao.hw,
>>>>> +	[RPMH_RF_CLK1]		= &clk_rpmh_clk1_a1.hw,
>>>>> +	[RPMH_RF_CLK1_A]	= &clk_rpmh_clk1_a1_ao.hw,
>>>>> +	[RPMH_RF_CLK2]		= &clk_rpmh_clk2_a1.hw,
>>>>> +	[RPMH_RF_CLK2_A]	= &clk_rpmh_clk2_a1_ao.hw,
>>>>> +	[RPMH_RF_CLK3]		= &clk_rpmh_clk3_a1.hw,
>>>>> +	[RPMH_RF_CLK3_A]	= &clk_rpmh_clk3_a1_ao.hw,
>>>>> +	[RPMH_RF_CLK4]		= &clk_rpmh_clk4_a1.hw,
>>>>> +	[RPMH_RF_CLK4_A]	= &clk_rpmh_clk4_a1_ao.hw,
>>>>> +	[RPMH_IPA_CLK]		= &clk_rpmh_ipa.hw,
>>>>> +};
>>>>> +
>>>>> +static const struct clk_rpmh_desc clk_rpmh_sm8550 = {
>>>>> +	.clks = sm8550_rpmh_clocks,
>>>>> +	.num_clks = ARRAY_SIZE(sm8550_rpmh_clocks),
>>>>> +};
>>>>> +
>>>>>    static struct clk_hw *sc7280_rpmh_clocks[] = {
>>>>>    	[RPMH_CXO_CLK]      = &clk_rpmh_bi_tcxo_div4.hw,
>>>>>    	[RPMH_CXO_CLK_A]    = &clk_rpmh_bi_tcxo_div4_ao.hw,
>>>>> @@ -683,29 +750,31 @@ static int clk_rpmh_probe(struct platform_device *pdev)
>>>>>    		name = hw_clks[i]->init->name;
>>>>> -		rpmh_clk = to_clk_rpmh(hw_clks[i]);
>>>>> -		res_addr = cmd_db_read_addr(rpmh_clk->res_name);
>>>>> -		if (!res_addr) {
>>>>> -			dev_err(&pdev->dev, "missing RPMh resource address for %s\n",
>>>>> -				rpmh_clk->res_name);
>>>>> -			return -ENODEV;
>>>>> -		}
>>>>> +		if (hw_clks[i]->init->ops != &clk_fixed_factor_ops) {
>>>>
>>>> We discussed this separately, the fixed factor clocks will be moved to the
>>>> child nodes, leaving rpmhcc with only cmd-db based clocks.
>>>>
>>>
>>> Are you saying that you will represent bi_tcxo as a fixed-factor-clock
>>> under /clocks with RPMH_CXO_PAD_CLK as parent and a clock-div = <2>; ?
>>
>> Yes, this was the idea. The rpmhcc driver is pretty much centric around 
>> the cmd-db clocks. Adding a fixed-factor clock results either in a 
>> horrible hacks or in a significant code refactoring. However we already 
>> have an existing way to fixed-factor clocks: DT nodes. Adding support 
>> for such nodes to rpmhcc driver requires just a single additional API 
>> call: devm_of_platform_populate().
> 
> Please no. DT is not to solve driver issues, skip some code or make
> things simpler for driver developers. If everyone - U-boot, *BSD,
> firmwares - pushes to DT stuff like this, because this makes their
> driver development easier, you would have total mess. Linux does not
> have any priorities here in this approach.

Assuming we talk about Abel's implementation of putting these nodes in
rpmhcc, because you wrote here devm_of_platform_populate()...

Best regards,
Krzysztof