RE: [PATCH v7 RESEND 2/3] thermal: exynos_tmu: Support new hardware and update TMU interface

[손신 <shin.son@samsung.com>]

Hello, Tudor Ambarus.

> -----Original Message-----
> From: Tudor Ambarus [mailto:tudor.ambarus@linaro.org]
> Sent: Monday, November 24, 2025 7:43 PM
> To: 손신 <shin.son@samsung.com>; 'Daniel Lezcano'
> <daniel.lezcano@linaro.org>; 'Bartlomiej Zolnierkiewicz'
> <bzolnier@gmail.com>; 'Krzysztof Kozlowski' <krzk@kernel.org>; 'Rafael J .
> Wysocki' <rafael@kernel.org>; 'Zhang Rui' <rui.zhang@intel.com>; 'Lukasz
> Luba' <lukasz.luba@arm.com>; 'Rob Herring' <robh@kernel.org>; 'Conor
> Dooley' <conor+dt@kernel.org>; 'Alim Akhtar' <alim.akhtar@samsung.com>
> Cc: 'Henrik Grimler' <henrik@grimler.se>; linux-pm@vger.kernel.org; linux-
> samsung-soc@vger.kernel.org; devicetree@vger.kernel.org; linux-arm-
> kernel@lists.infradead.org; linux-kernel@vger.kernel.org; Peter Griffin
> <peter.griffin@linaro.org>; André Draszik <andre.draszik@linaro.org>;
> William McVicker <willmcvicker@google.com>; jyescas@google.com
> Subject: Re: [PATCH v7 RESEND 2/3] thermal: exynos_tmu: Support new
> hardware and update TMU interface
> 
> Hi, Shin,
> 
> On 11/24/25 12:06 PM, 손신 wrote:
> >>> +static void update_con_reg(struct exynos_tmu_data *data) {
> >>> +	u32 val, t_buf_vref_sel, t_buf_slope_sel;
> >>> +
> >>> +	val = readl(data->base + EXYNOS_TMU_REG_TRIMINFO);
> >>> +	t_buf_vref_sel = (val >> EXYNOSAUTOV920_TMU_T_BUF_VREF_SEL_SHIFT)
> >>> +				& EXYNOSAUTOV920_TMU_T_BUF_VREF_SEL_MASK;
> >>> +	t_buf_slope_sel = (val >> EXYNOSAUTOV920_TMU_T_BUF_SLOPE_SEL_SHIFT)
> >>> +				& EXYNOSAUTOV920_TMU_T_BUF_SLOPE_SEL_MASK;
> >>> +
> >>> +	val = readl(data->base +  EXYNOSAUTOV920_TMU_REG_CONTROL);
> >>> +	val &= ~(EXYNOS_TMU_REF_VOLTAGE_MASK <<
> >> EXYNOS_TMU_REF_VOLTAGE_SHIFT);
> >>> +	val |= (t_buf_vref_sel << EXYNOS_TMU_REF_VOLTAGE_SHIFT);
> >>> +	val &= ~(EXYNOS_TMU_BUF_SLOPE_SEL_MASK <<
> >> EXYNOS_TMU_BUF_SLOPE_SEL_SHIFT);
> >>> +	val |= (t_buf_slope_sel << EXYNOS_TMU_BUF_SLOPE_SEL_SHIFT);
> >>> +	writel(val, data->base + EXYNOSAUTOV920_TMU_REG_CONTROL);
> >>> +
> >>> +	val = readl(data->base + EXYNOSAUTOV920_TMU_REG_CONTROL1);
> >>> +	val &= ~(EXYNOSAUTOV920_TMU_NUM_PROBE_MASK <<
> >> EXYNOSAUTOV920_TMU_NUM_PROBE_SHIFT);
> >>> +	val &= ~(EXYNOSAUTOV920_TMU_LPI_MODE_MASK <<
> >> EXYNOSAUTOV920_TMU_LPI_MODE_SHIFT);
> >>> +	val |= (data->sensor_count << EXYNOSAUTOV920_TMU_NUM_PROBE_SHIFT);
> >>> +	writel(val, data->base + EXYNOSAUTOV920_TMU_REG_CONTROL1);
> >>> +
> >>> +	writel(1, data->base + EXYNOSAUTOV920_TMU_SAMPLING_INTERVAL);
> >>> +	writel(EXYNOSAUTOV920_TMU_AVG_CON_UPDATE, data->base +
> >> EXYNOSAUTOV920_TMU_REG_AVG_CONTROL);
> >>> +	writel(EXYNOSAUTOV920_TMU_COUNTER_VALUE0_UPDATE,
> >>> +	       data->base + EXYNOSAUTOV920_TMU_REG_COUNTER_VALUE0);
> >>> +	writel(EXYNOSAUTOV920_TMU_COUNTER_VALUE1_UPDATE,
> >>> +	       data->base + EXYNOSAUTOV920_TMU_REG_COUNTER_VALUE1);
> >>> +}
> >>> +
> >> This is unreadable; please make it understandable for those who don’t
> >> have the documentation (explicit static inline functions, comments,
> etc ...).
> > I'll restructure this code by introducing explicit static inline helper
> functions and proper comments to improve readability.
> 
> We likely shouldn't use inlines here, see Linus's reply from 2006:
> https://lore.kernel.org/all/Pine.LNX.4.64.0601021105000.3668@g5.osdl.org/T
> /#u
> 
> I guess you can make this easier to read if you use FIELD_GET/SET from
> bitfield.h. Other improvement would be using the regmap api.
> 
> Shin, a bit unrelated with the patch, but I wanted to let you know that I
> started looking at the GS101 TMU. I assume it's very similar with the TMU
> on exynosautov920. Do you know if they share the same IP version?
> 
> I noticed GS101 uses ACPM calls to communicate with the TMU. Why did you
> choose to not use ACPM for exynosautov920 TMU?
> 
> Thanks!
> ta

I will adopt FIELD_GET/FIELD_PREP to simplify the bitfield handling, and I will avoid using inline functions excessively as suggested.

Regarding ACPM, I did not introduce it earlier because I was trying to align the implementation with the existing framework.
However, if we move toward a separate driver, I will reconsider whether ACPM integration makes sense there.
Would it be possible to get your feedback again when I prepare the next revision of the driver?

Plus, the GS101 TMU driver isn't upstream yet, right?
Could you share where I can find the example code you mentioned? Thank you in advance.

Best regards,
Shin