Re: [PATCH 6/6] mfd: da9063: Add DA9063L support

[Marek Vasut <marek.vasut@gmail.com>]

On 05/24/2018 07:30 PM, Steve Twiss wrote:
> On 24 May 2018 15:51 Marek Vasut wrote:
> 
> Hi Marek,
> 
>> To: Steve Twiss <stwiss.opensource@diasemi.com>; linux-kernel@vger.kernel.org
>> Cc: Marek Vasut <marek.vasut+renesas@gmail.com>; Geert Uytterhoeven
>> <geert+renesas@glider.be>; Lee Jones <lee.jones@linaro.org>; Mark Brown
>> <broonie@kernel.org>; Wolfram Sang <wsa+renesas@sang-engineering.com>;
>> linux-renesas-soc@vger.kernel.org
>> Subject: Re: [PATCH 6/6] mfd: da9063: Add DA9063L support
>>
>> On 05/24/2018 02:32 PM, Steve Twiss wrote:
>>> On 24 May 2018 @ 12:49 Steve Twiss wrote:
>>>>> On 23 May 2018 12:43 Marek Vasut wrote,
>>>>>
>>>>> To: linux-kernel@vger.kernel.org
>>>>> Subject: [PATCH 6/6] mfd: da9063: Add DA9063L support
>>>>>
>>>>> Add support for DA9063L, which is a reduced variant of the DA9063 with less regulators and without RTC.
>>>>>
>>>>
>>>> There's potentially more to this file. Without an RTC the regmap
>>>> access tables would change and the usual DA9063 (BB silicon) tables would become invalid.
>>>> The tables for da9063_bb_readable_ranges, da9063_bb_writeable_ranges,
>>>> da9063_bb_volatile_ranges, would need to be updated for DA9063L, if a new chip model was needed.
>>>>
>>>> The new ranges would be this (see below), and would remove any RTC accesses in the new chip model.
>>>>
>>>> static const struct regmap_range da9063l_bb_readable_ranges[] = {
>>>> 	{
>>>> 		.range_min = DA9063_REG_PAGE_CON,
>>>> 		.range_max = DA9063_REG_MON_A10_RES,
>>>> 	}, {
>>>> 		.range_min = DA9063_REG_SEQ,
>>>> 		.range_max = DA9063_REG_ID_32_31,
>>>> 	}, {
>>>> 		.range_min = DA9063_REG_SEQ_A,
>>>> 		.range_max = DA9063_REG_AUTO3_LOW,
>>>> 	}, {
>>>> 		.range_min = DA9063_REG_T_OFFSET,
>>>> 		.range_max = DA9063_BB_REG_GP_ID_19,
>>>> 	}, {
>>>> 		.range_min = DA9063_REG_CHIP_ID,
>>>> 		.range_max = DA9063_REG_CHIP_VARIANT,
>>>> 	},
>>>> };
>>>>
>>>> static const struct regmap_range da9063l_bb_writeable_ranges[] = {
>>>> 	{
>>>> 		.range_min = DA9063_REG_PAGE_CON,
>>>> 		.range_max = DA9063_REG_PAGE_CON,
>>>> 	}, {
>>>> 		.range_min = DA9063_REG_FAULT_LOG,
>>>> 		.range_max = DA9063_REG_VSYS_MON,
>>>> 	}, {
>>>> 		.range_min = DA9063_REG_SEQ,
>>>> 		.range_max = DA9063_REG_ID_32_31,
>>>> 	}, {
>>>> 		.range_min = DA9063_REG_SEQ_A,
>>>> 		.range_max = DA9063_REG_AUTO3_LOW,
>>>> 	}, {
>>>> 		.range_min = DA9063_REG_CONFIG_I,
>>>> 		.range_max = DA9063_BB_REG_MON_REG_4,
>>>> 	}, {
>>>> 		.range_min = DA9063_BB_REG_GP_ID_0,
>>>> 		.range_max = DA9063_BB_REG_GP_ID_19,
>>>> 	},
>>>> };
>>>>
>>>> static const struct regmap_range da9063l_bb_volatile_ranges[] = {
>>>> 	{
>>>> 		.range_min = DA9063_REG_PAGE_CON,
>>>> 		.range_max = DA9063_REG_EVENT_D,
>>>> 	}, {
>>>> 		.range_min = DA9063_REG_CONTROL_A,
>>>> 		.range_max = DA9063_REG_CONTROL_B,
>>>> 	}, {
>>>> 		.range_min = DA9063_REG_CONTROL_E,
>>>> 		.range_max = DA9063_REG_CONTROL_F,
>>>> 	}, {
>>>> 		.range_min = DA9063_REG_BCORE2_CONT,
>>>> 		.range_max = DA9063_REG_LDO11_CONT,
>>>> 	}, {
>>>> 		.range_min = DA9063_REG_DVC_1,
>>>> 		.range_max = DA9063_REG_ADC_MAN,
>>>> 	}, {
>>>> 		.range_min = DA9063_REG_ADC_RES_L,
>>>> 		.range_max = DA9063_REG_MON_A10_RES,
>>>> 	}, {
>>>> 		.range_min = DA9063_REG_SEQ,
>>>> 		.range_max = DA9063_REG_SEQ,
>>>> 	}, {
>>>> 		.range_min = DA9063_REG_EN_32K,
>>>> 		.range_max = DA9063_REG_EN_32K,
>>>> 	}, {
>>>> 		.range_min = DA9063_BB_REG_MON_REG_5,
>>>> 		.range_max = DA9063_BB_REG_MON_REG_6,
>>>> 	},
>>>> };
>>>>
>>>> However this is a larger and more wide-ranging change compared to the
>>>> one proposed by Marek, and would require other alterations to fit
>>>> this in. Also I'm undecided to what it would really add apart from a
>>>> new chip model: I have been told accessing the DA9063 RTC register locations
>>>> has no effect in the DA9063L.
>>>
>>> Looking at this further, there is also a new IRQ regmap.
>>> Again this comes down to whether a full chip model is needed or not.
>>> If not, then the IRQ map does not need to be changed as given. Otherwise the
>>> removal of the following:
>>>
>>> 	[DA9063_IRQ_ALARM] = {
>>> 		.reg_offset = DA9063_REG_EVENT_A_OFFSET,
>>> 		.mask = DA9063_M_ALARM,
>>> 	},
>>> 	[DA9063_IRQ_TICK] = {
>>> 		.reg_offset = DA9063_REG_EVENT_A_OFFSET,
>>> 		.mask = DA9063_M_TICK,
>>> 	},
>>>
>>> prior to registering the IRQs in the chip model would be needed.
>>> The new regmap_irq would be:
>>>
>>> static const struct regmap_irq da9063l_irqs[] = {
>>> 	/* DA9063 event A register */
>>> 	[DA9063L_IRQ_ONKEY] = {
>>> 		.reg_offset = DA9063_REG_EVENT_A_OFFSET,
>>> 		.mask = DA9063_M_ONKEY,
>>> 	},
>>> 	[DA9063L_IRQ_ADC_RDY] = {
>>> 		.reg_offset = DA9063_REG_EVENT_A_OFFSET,
>>> 		.mask = DA9063_M_ADC_RDY,
>>> 	},
>>> 	[DA9063L_IRQ_SEQ_RDY] = {
>>> 		.reg_offset = DA9063_REG_EVENT_A_OFFSET,
>>> 		.mask = DA9063_M_SEQ_RDY,
>>> 	},
>>> 	/* DA9063 event B register */
>>> 	[DA9063L_IRQ_WAKE] = {
>>> 		.reg_offset = DA9063_REG_EVENT_B_OFFSET,
>>> 		.mask = DA9063_M_WAKE,
>>> 	},
>>> 	[DA9063L_IRQ_TEMP] = {
>>> 		.reg_offset = DA9063_REG_EVENT_B_OFFSET,
>>> 		.mask = DA9063_M_TEMP,
>>> 	},
>>> 	[DA9063L_IRQ_COMP_1V2] = {
>>> 		.reg_offset = DA9063_REG_EVENT_B_OFFSET,
>>> 		.mask = DA9063_M_COMP_1V2,
>>> 	},
>>> 	[DA9063L_IRQ_LDO_LIM] = {
>>> 		.reg_offset = DA9063_REG_EVENT_B_OFFSET,
>>> 		.mask = DA9063_M_LDO_LIM,
>>> 	},
>>> 	[DA9063L_IRQ_REG_UVOV] = {
>>> 		.reg_offset = DA9063_REG_EVENT_B_OFFSET,
>>> 		.mask = DA9063_M_UVOV,
>>> 	},
>>> 	[DA9063L_IRQ_DVC_RDY] = {
>>> 		.reg_offset = DA9063_REG_EVENT_B_OFFSET,
>>> 		.mask = DA9063_M_DVC_RDY,
>>> 	},
>>> 	[DA9063L_IRQ_VDD_MON] = {
>>> 		.reg_offset = DA9063_REG_EVENT_B_OFFSET,
>>> 		.mask = DA9063_M_VDD_MON,
>>> 	},
>>> 	[DA9063L_IRQ_WARN] = {
>>> 		.reg_offset = DA9063_REG_EVENT_B_OFFSET,
>>> 		.mask = DA9063_M_VDD_WARN,
>>> 	},
>>> 	/* DA9063 event C register */
>>> 	[DA9063L_IRQ_GPI0] = {
>>> 		.reg_offset = DA9063_REG_EVENT_C_OFFSET,
>>> 		.mask = DA9063_M_GPI0,
>>> 	},
>>> 	[DA9063L_IRQ_GPI1] = {
>>> 		.reg_offset = DA9063_REG_EVENT_C_OFFSET,
>>> 		.mask = DA9063_M_GPI1,
>>> 	},
>>> 	[DA9063L_IRQ_GPI2] = {
>>> 		.reg_offset = DA9063_REG_EVENT_C_OFFSET,
>>> 		.mask = DA9063_M_GPI2,
>>> 	},
>>> 	[DA9063L_IRQ_GPI3] = {
>>> 		.reg_offset = DA9063_REG_EVENT_C_OFFSET,
>>> 		.mask = DA9063_M_GPI3,
>>> 	},
>>> 	[DA9063L_IRQ_GPI4] = {
>>> 		.reg_offset = DA9063_REG_EVENT_C_OFFSET,
>>> 		.mask = DA9063_M_GPI4,
>>> 	},
>>> 	[DA9063L_IRQ_GPI5] = {
>>> 		.reg_offset = DA9063_REG_EVENT_C_OFFSET,
>>> 		.mask = DA9063_M_GPI5,
>>> 	},
>>> 	[DA9063L_IRQ_GPI6] = {
>>> 		.reg_offset = DA9063_REG_EVENT_C_OFFSET,
>>> 		.mask = DA9063_M_GPI6,
>>> 	},
>>> 	[DA9063L_IRQ_GPI7] = {
>>> 		.reg_offset = DA9063_REG_EVENT_C_OFFSET,
>>> 		.mask = DA9063_M_GPI7,
>>> 	},
>>> 	/* DA9063 event D register */
>>> 	[DA9063L_IRQ_GPI8] = {
>>> 		.reg_offset = DA9063_REG_EVENT_D_OFFSET,
>>> 		.mask = DA9063_M_GPI8,
>>> 	},
>>> 	[DA9063L_IRQ_GPI9] = {
>>> 		.reg_offset = DA9063_REG_EVENT_D_OFFSET,
>>> 		.mask = DA9063_M_GPI9,
>>> 	},
>>> 	[DA9063L_IRQ_GPI10] = {
>>> 		.reg_offset = DA9063_REG_EVENT_D_OFFSET,
>>> 		.mask = DA9063_M_GPI10,
>>> 	},
>>> 	[DA9063L_IRQ_GPI11] = {
>>> 		.reg_offset = DA9063_REG_EVENT_D_OFFSET,
>>> 		.mask = DA9063_M_GPI11,
>>> 	},
>>> 	[DA9063L_IRQ_GPI12] = {
>>> 		.reg_offset = DA9063_REG_EVENT_D_OFFSET,
>>> 		.mask = DA9063_M_GPI12,
>>> 	},
>>> 	[DA9063L_IRQ_GPI13] = {
>>> 		.reg_offset = DA9063_REG_EVENT_D_OFFSET,
>>> 		.mask = DA9063_M_GPI13,
>>> 	},
>>> 	[DA9063L_IRQ_GPI14] = {
>>> 		.reg_offset = DA9063_REG_EVENT_D_OFFSET,
>>> 		.mask = DA9063_M_GPI14,
>>> 	},
>>> 	[DA9063L_IRQ_GPI15] = {
>>> 		.reg_offset = DA9063_REG_EVENT_D_OFFSET,
>>> 		.mask = DA9063_M_GPI15,
>>> 	},
>>> };
>>
>> We can probably do the same trick with the regmaps and irqmaps as with the
>> rest, that is, reorder them and register only a smaller portion ?
> 
> I like the "reorder and only register a smaller portion" trick. But it wouldn't work
> with what I gave earlier today, without some modification.
> For instance, the first register readable entry range in the  DA9063 BB is:
> 
> static const struct regmap_range da9063_bb_readable_ranges[] = {
> 	{
> 		.range_min = DA9063_REG_PAGE_CON,
> 		.range_max = DA9063_BB_REG_SECOND_D,
> 	}, {
> 
> But for the DA9063L, this first range entry would be changed, not removed:
> 
> static const struct regmap_range da9063l_bb_readable_ranges[] = {
> 	{
> 		.range_min = DA9063_REG_PAGE_CON,
> 		.range_max = DA9063_REG_MON_A10_RES,
> 	}, {
> 
> So it's not all-or-nothing. But possibly it could be made to work if those ranges were split
> into two pieces.
> 
> However, it might get messy to maintain in future -- sometimes register ranges need to be
> updated with new components or if a new feature is added -- usually I need to work it
> all out on paper with the full register map. Splitting up ranges might make it a little
> messier. But, it's not impossible.
> 
> For the DA9062 and DA9061 this was done using separate ranges and using the macro
> regmap_reg_range(). It's not that messy to read, e.g. 
> https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/tree/drivers/mfd/da9062-core.c?h=next-20180517#n367

Hum, can you point me to the datasheet sections so I can check this
difference please ? I think I have the rest of the feedback addressed,
so I want to check this one before submitting the next version.

-- 
Best regards,
Marek Vasut