* [PATCH] thermal/drivers/rzg2l: Fix precision loss in temperature calculation
@ 2026-11-16 6:29 Nhat Tran
0 siblings, 0 replies; only message in thread
From: Nhat Tran @ 2026-11-16 6:29 UTC (permalink / raw)
To: Rafael J . Wysocki, Daniel Lezcano
Cc: Zhang Rui, Lukasz Luba, Biju Das, Lad Prabhakar, linux-pm,
linux-renesas-soc, linux-kernel, Nhat Tran
According to the RZ/G2L hardware user's manual (Rev.1.50, section
44.4.4 Procedure for Measuring the Temperature) the temperature is
computed as
Tj = (DSENSOR - OTPTSUTRIM1) * (165 / (OTPTSUTRIM0 - OTPTSUTRIM1)) - 40
The driver follows the same grouping with integer arithmetic, so the
fractional part of 165000 / (calib0 - calib1) is truncated before it is
multiplied by (dsensor - calib1). Above -40 °C this always makes the
driver report a lower value than the real one, and the error grows
linearly with the temperature, so trip points are crossed later than
configured.
On an RZ/G2L SMARC EVK with OTP calibration values calib0 = 2902 and
calib1 = 586, the slope 71.24 is truncated to 71. The computed value is
313 m°C too low at 52 °C and 351 m°C too low at 63 °C, e.g. 62382 m°C
instead of 62733 m°C, which is reported as 62500 instead of 63000 after
rounding. With the default calibration values (3148, 503) the
calculated error is about 1 °C at 125 °C.
Multiply before dividing, as rzg3s_thermal already does, and use signed
arithmetic so that temperatures below -40 °C are still handled. The
intermediate value is at most 4095 * 165000, which fits in a signed
32-bit integer.
Fixes: 673c68bd4839 ("thermal/drivers: Add TSU driver for RZ/G2L")
Signed-off-by: Nhat Tran <nhat.tranminhk2@gmail.com>
---
drivers/thermal/renesas/rzg2l_thermal.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/drivers/thermal/renesas/rzg2l_thermal.c b/drivers/thermal/renesas/rzg2l_thermal.c
index 90e2ba378a4b..3295eef12001 100644
--- a/drivers/thermal/renesas/rzg2l_thermal.c
+++ b/drivers/thermal/renesas/rzg2l_thermal.c
@@ -75,8 +75,8 @@ static inline void rzg2l_thermal_write(struct rzg2l_thermal_priv *priv, u32 reg,
static int rzg2l_thermal_get_temp(struct thermal_zone_device *tz, int *temp)
{
struct rzg2l_thermal_priv *priv = thermal_zone_device_priv(tz);
- u32 result = 0, dsensor, ts_code_ave;
- int val, i;
+ u32 result = 0, ts_code_ave;
+ int dsensor, val, i;
for (i = 0; i < TS_CODE_CAP_TIMES ; i++) {
/*
@@ -105,8 +105,8 @@ static int rzg2l_thermal_get_temp(struct thermal_zone_device *tz, int *temp)
* Tj = (dsensor − calib1) * 165/ (calib0 − calib1) − 40
* where calib0 and calib1 are the calibration values.
*/
- val = ((dsensor - priv->calib1) * (MCELSIUS(165) /
- (priv->calib0 - priv->calib1))) - MCELSIUS(40);
+ val = (dsensor - (int)priv->calib1) * MCELSIUS(165) /
+ (int)(priv->calib0 - priv->calib1) - MCELSIUS(40);
*temp = roundup(val, RZG2L_THERMAL_GRAN);
--
2.25.1
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2026-11-16 6:29 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-11-16 6:29 [PATCH] thermal/drivers/rzg2l: Fix precision loss in temperature calculation Nhat Tran
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®