Dear Niklas-san
On 2018/11/02 18:31, Niklas Söderlund wrote:
Hi Hoan-san,
Thanks for your patch.
On 2018-10-30 18:10:01 +0900, Nguyen An Hoan wrote:
From: Hoan Nguyen An <na-hoan@xxxxxxxxxxx>
This patch fixes thermal interrupts that did not happen when temprature changed.
Add the function rcar_gen3_thermal_update_threshold(), this function is used to
calculate the value written to the threshold registers REG_GEN3_IRQTEMP1 and
REG_GEN3_IRQTEMP2.
Is it really a error if the IRQ don't happen and the thermal framework
use the .set_trip() callback to move the window due to the polling
instead of a IRQ due to a rapid thermal increase? I think not. I do not
like this patch but I'm not expert on thermal subsystem so I'm open to
be proven wrong :-)
I'm investigating more about this issue!
This patch is based on renesas-bsp/rcar-3.5.4.rc2 !
Signed-off-by: Hoan Nguyen An <na-hoan@xxxxxxxxxxx>
---
drivers/thermal/rcar_gen3_thermal.c | 45 +++++++++++++++++++++----------------
1 file changed, 26 insertions(+), 19 deletions(-)
diff --git a/drivers/thermal/rcar_gen3_thermal.c b/drivers/thermal/rcar_gen3_thermal.c
index 7aed533..009c0db 100644
--- a/drivers/thermal/rcar_gen3_thermal.c
+++ b/drivers/thermal/rcar_gen3_thermal.c
@@ -185,37 +185,39 @@ static int rcar_gen3_thermal_get_temp(void *devdata, int *temp)
static int rcar_gen3_thermal_mcelsius_to_temp(struct rcar_gen3_thermal_tsc *tsc,
int mcelsius)
{
- int celsius, val1, val2;
+ int val1, val2;
- celsius = DIV_ROUND_CLOSEST(mcelsius, 1000);
- val1 = celsius * tsc->coef.a1 + tsc->coef.b1;
- val2 = celsius * tsc->coef.a2 + tsc->coef.b2;
+ val1 = (mcelsius * tsc->coef.a1)/1000 + tsc->coef.b1;
+ val2 = (mcelsius * tsc->coef.a2)/1000 + tsc->coef.b2;
This won't work I'm afraid as it can overflow. As the thcodes and ptat
values can be read from hardware there theoretical max size is 0xfff
according to the documentation. Those values are then used to calculate
a1, b1, a2 and b2 constants. If you can prove that no overflow can
happen I'm happy for this change as it increases accuracy. If so this
should be posted as a separate patch.
With this patch, when rcar_gen3_thermal_mcelsius_to_temp () is called,
since the mcelsius value was passed is calculated from the previous formula:
rcar_gen3_thermal_get_temp():
int mcelsius, val1, val2;
u32 reg;
/* Read register and convert to mili Celsius */
reg = rcar_gen3_thermal_read(tsc, REG_GEN3_TEMP) & CTEMP_MASK;
val1 = FIXPT_DIV(FIXPT_INT(reg) - tsc->coef.b1, tsc->coef.a1);
val2 = FIXPT_DIV(FIXPT_INT(reg) - tsc->coef.b2, tsc->coef.a2);
mcelsius = FIXPT_TO_MCELSIUS((val1 + val2) / 2);
/* Make sure we are inside specifications */
if ((mcelsius < MCELSIUS(-40)) || (mcelsius > MCELSIUS(125)))
return -EIO;
From int mcelsius values, val1, val2;
The reverse calculation I do not think will happen the overflow!
With the current situation, I am not clear about the high, low values
But with (a1, b1), (a2, b2)
tsc1->coef.a1 965 tsc1->coef.a2 953 tsc1->coef.b1 323853 tsc1->coef.b2
324268
tsc2->coef.a1 965 tsc2->coef.a2 954 tsc2->coef.b1 323213 tsc2->coef.b2
323640
tsc3->coef.a1 947 tsc3->coef.a2 932 tsc3->coef.b1 325163 tsc3->coef.b2
325680
with all tscX, aX <=1000.
If the temperature is 1000celsius = 1000000 mcelsius.
val1 <1000000 * 1000 + 323853 = 1 000 323 853 <2 147 483 647 (max int value)
(In fact the temperature of the board could not get to 1000 celsius
degrees )
Here the current constant values are used.
But at future THCODE and PLAT values are read from registers
to calculate aX, bX, I do not know whether there really will be overflow
or not.
If you use this patches without this part.
I have received incorrect interrupts with H3ES1.1.
So I need to integrate this part to ensure correctly in testing.
return INT_FIXPT((val1 + val2) / 2);
}
-static int rcar_gen3_thermal_set_trips(void *devdata, int low, int high)
+static int rcar_gen3_thermal_update_threshold(struct rcar_gen3_thermal_tsc *tsc)
{
- struct rcar_gen3_thermal_tsc *tsc = devdata;
-
- low = clamp_val(low, -40000, 120000);
- high = clamp_val(high, -40000, 120000);
-
- rcar_gen3_thermal_write(tsc, REG_GEN3_IRQTEMP1,
- rcar_gen3_thermal_mcelsius_to_temp(tsc, low));
+ u32 ctemp;
+ int temp_code;
+ int mcelsius, val1, val2;
- rcar_gen3_thermal_write(tsc, REG_GEN3_IRQTEMP2,
- rcar_gen3_thermal_mcelsius_to_temp(tsc, high));
+ ctemp = rcar_gen3_thermal_read(tsc, REG_GEN3_TEMP) & CTEMP_MASK;
+ val1 = FIXPT_DIV(FIXPT_INT(ctemp) - tsc->coef.b1, tsc->coef.a1);
+ val2 = FIXPT_DIV(FIXPT_INT(ctemp) - tsc->coef.b2, tsc->coef.a2);
+ mcelsius = FIXPT_TO_MCELSIUS((val1 + val2) / 2);
- tsc->low = low;
- tsc->high = high;
+ /* set the interrupts to exceed the temperature */
+ temp_code = rcar_gen3_thermal_mcelsius_to_temp(tsc,
+ mcelsius + MCELSIUS(1));
+ rcar_gen3_thermal_write(tsc, REG_GEN3_IRQTEMP1, temp_code);
+ /* set the interrupts to fall below the temperature */
+ temp_code = rcar_gen3_thermal_mcelsius_to_temp(tsc,
+ mcelsius - MCELSIUS(1));
+ rcar_gen3_thermal_write(tsc, REG_GEN3_IRQTEMP2, temp_code);
return 0;
}
static const struct thermal_zone_of_device_ops rcar_gen3_tz_of_ops = {
.get_temp = rcar_gen3_thermal_get_temp,
- .set_trips = rcar_gen3_thermal_set_trips,
My understanding of the thermal framework is that implementing
.set_trips() is correct and removing it here and rewriting is just
wrong. The driver should use the framework not do it's own thing, that
way all the DT properties set for hysteresis and such are used in a
correct way.
I'm investigating more about this issue!
};
static void rcar_thermal_irq_set(struct rcar_gen3_thermal_priv *priv, bool on)
@@ -256,8 +258,11 @@ static irqreturn_t rcar_gen3_thermal_irq_thread(int irq, void *data)
int i;
for (i = 0; i < priv->num_tscs; i++)
+ {
+ rcar_gen3_thermal_update_threshold(priv->tscs[i]);
thermal_zone_device_update(priv->tscs[i]->zone,
THERMAL_EVENT_UNSPECIFIED);
+ }
spin_lock_irqsave(&priv->lock, flags);
rcar_thermal_irq_set(priv, true);
@@ -306,7 +311,7 @@ static void rcar_gen3_thermal_init(struct rcar_gen3_thermal_tsc *tsc)
usleep_range(1000, 2000);
- rcar_gen3_thermal_write(tsc, REG_GEN3_IRQCTL, 0x3F);
+ rcar_gen3_thermal_write(tsc, REG_GEN3_IRQCTL, 0);
This looks correct according to the new datasheets but should be posted
in a separate patch.
I have created this patch separately before.
"https://patchwork.kernel.org/patch/10655351/";
If possible Niklas-san could give me Signed-of-by
or Acked? Thank you!
Thank you for the review
Hoan.
rcar_gen3_thermal_write(tsc, REG_GEN3_IRQMSK, 0);
rcar_gen3_thermal_write(tsc, REG_GEN3_IRQEN, IRQ_TEMPD1 | IRQ_TEMP2);
@@ -414,6 +419,8 @@ static int rcar_gen3_thermal_probe(struct platform_device *pdev)
priv->thermal_init(tsc);
rcar_gen3_thermal_calc_coefs(&tsc->coef, ptat, thcode[i]);
+ rcar_gen3_thermal_update_threshold(tsc);
+
zone = devm_thermal_zone_of_sensor_register(dev, i, tsc,
&rcar_gen3_tz_of_ops);
if (IS_ERR(zone)) {
@@ -465,7 +472,7 @@ static int __maybe_unused rcar_gen3_thermal_resume(struct device *dev)
struct rcar_gen3_thermal_tsc *tsc = priv->tscs[i];
priv->thermal_init(tsc);
- rcar_gen3_thermal_set_trips(tsc, tsc->low, tsc->high);
+ rcar_gen3_thermal_update_threshold(tsc);
}
rcar_thermal_irq_set(priv, true);
--
2.7.4
