Currently the threshold limits are updated in 2 stages, once for all
software trigger levels and again for hardware trip point.
While updating the software trigger levels, it overwrites the threshold
limit for hardware trip point thereby forcing the Exynos core to issue
an emergency shutdown.
Updating only the required fields in threshold register fixes this issue.
Signed-off-by: Tushar Behera <[email protected]>
---
Based on v3.15-rc1.
drivers/thermal/samsung/exynos_tmu.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/drivers/thermal/samsung/exynos_tmu.c b/drivers/thermal/samsung/exynos_tmu.c
index 0d96a51..ffccc89 100644
--- a/drivers/thermal/samsung/exynos_tmu.c
+++ b/drivers/thermal/samsung/exynos_tmu.c
@@ -225,6 +225,8 @@ skip_calib_data:
trigger_levs++;
}
+ rising_threshold = readl(data->base + reg->threshold_th0);
+
if (data->soc == SOC_ARCH_EXYNOS4210) {
/* Write temperature code for threshold */
threshold_code = temp_to_code(data, pdata->threshold);
@@ -249,6 +251,7 @@ skip_calib_data:
ret = threshold_code;
goto out;
}
+ rising_threshold &= ~(0xff << 8 * i);
rising_threshold |= threshold_code << 8 * i;
if (pdata->threshold_falling) {
threshold_code = temp_to_code(data,
@@ -281,6 +284,7 @@ skip_calib_data:
}
if (i == EXYNOS_MAX_TRIGGER_PER_REG - 1) {
/* 1-4 level to be assigned in th0 reg */
+ rising_threshold &= ~(0xff << 8 * i);
rising_threshold |= threshold_code << 8 * i;
writel(rising_threshold,
data->base + reg->threshold_th0);
--
1.7.9.5
On 04/14/2014 11:08 AM, Tushar Behera wrote:
> Currently the threshold limits are updated in 2 stages, once for all
> software trigger levels and again for hardware trip point.
>
> While updating the software trigger levels, it overwrites the threshold
> limit for hardware trip point thereby forcing the Exynos core to issue
> an emergency shutdown.
>
> Updating only the required fields in threshold register fixes this issue.
>
> Signed-off-by: Tushar Behera <[email protected]>
> ---
> Based on v3.15-rc1.
>
> drivers/thermal/samsung/exynos_tmu.c | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/drivers/thermal/samsung/exynos_tmu.c b/drivers/thermal/samsung/exynos_tmu.c
> index 0d96a51..ffccc89 100644
> --- a/drivers/thermal/samsung/exynos_tmu.c
> +++ b/drivers/thermal/samsung/exynos_tmu.c
> @@ -225,6 +225,8 @@ skip_calib_data:
> trigger_levs++;
> }
>
> + rising_threshold = readl(data->base + reg->threshold_th0);
> +
> if (data->soc == SOC_ARCH_EXYNOS4210) {
> /* Write temperature code for threshold */
> threshold_code = temp_to_code(data, pdata->threshold);
> @@ -249,6 +251,7 @@ skip_calib_data:
> ret = threshold_code;
> goto out;
> }
> + rising_threshold &= ~(0xff << 8 * i);
> rising_threshold |= threshold_code << 8 * i;
> if (pdata->threshold_falling) {
> threshold_code = temp_to_code(data,
> @@ -281,6 +284,7 @@ skip_calib_data:
> }
> if (i == EXYNOS_MAX_TRIGGER_PER_REG - 1) {
> /* 1-4 level to be assigned in th0 reg */
> + rising_threshold &= ~(0xff << 8 * i);
> rising_threshold |= threshold_code << 8 * i;
> writel(rising_threshold,
> data->base + reg->threshold_th0);
>
Adding Amit to get Samsung specific review.
--
Tushar Behera
On 4/14/14, Tushar Behera <[email protected]> wrote:
> Currently the threshold limits are updated in 2 stages, once for all
> software trigger levels and again for hardware trip point.
I guess the first stage is bootloader as could not find this in this file.
Anyways the changes looks fine to me.
Acked-by: Amit Daniel Kachhap <[email protected]>
>
> While updating the software trigger levels, it overwrites the threshold
> limit for hardware trip point thereby forcing the Exynos core to issue
> an emergency shutdown.
>
> Updating only the required fields in threshold register fixes this issue.
>
> Signed-off-by: Tushar Behera <[email protected]>
> ---
> Based on v3.15-rc1.
>
> drivers/thermal/samsung/exynos_tmu.c | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/drivers/thermal/samsung/exynos_tmu.c
> b/drivers/thermal/samsung/exynos_tmu.c
> index 0d96a51..ffccc89 100644
> --- a/drivers/thermal/samsung/exynos_tmu.c
> +++ b/drivers/thermal/samsung/exynos_tmu.c
> @@ -225,6 +225,8 @@ skip_calib_data:
> trigger_levs++;
> }
>
> + rising_threshold = readl(data->base + reg->threshold_th0);
> +
> if (data->soc == SOC_ARCH_EXYNOS4210) {
> /* Write temperature code for threshold */
> threshold_code = temp_to_code(data, pdata->threshold);
> @@ -249,6 +251,7 @@ skip_calib_data:
> ret = threshold_code;
> goto out;
> }
> + rising_threshold &= ~(0xff << 8 * i);
> rising_threshold |= threshold_code << 8 * i;
> if (pdata->threshold_falling) {
> threshold_code = temp_to_code(data,
> @@ -281,6 +284,7 @@ skip_calib_data:
> }
> if (i == EXYNOS_MAX_TRIGGER_PER_REG - 1) {
> /* 1-4 level to be assigned in th0 reg */
> + rising_threshold &= ~(0xff << 8 * i);
> rising_threshold |= threshold_code << 8 * i;
> writel(rising_threshold,
> data->base + reg->threshold_th0);
> --
> 1.7.9.5
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-samsung-soc"
> in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
Hi,
On Monday, April 14, 2014 11:08:15 AM Tushar Behera wrote:
> Currently the threshold limits are updated in 2 stages, once for all
> software trigger levels and again for hardware trip point.
>
> While updating the software trigger levels, it overwrites the threshold
> limit for hardware trip point thereby forcing the Exynos core to issue
> an emergency shutdown.
On what SoC type have you encountered this problem? It doesn't seem to
happen on older SoCs (at least Exynos4210 and Exynos4x12 ones).
> Updating only the required fields in threshold register fixes this issue.
With the current code there is indeed a time window during which threshold
limit for hardware trip point is set to zero so the fix is correct.
> Signed-off-by: Tushar Behera <[email protected]>
> ---
> Based on v3.15-rc1.
>
> drivers/thermal/samsung/exynos_tmu.c | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/drivers/thermal/samsung/exynos_tmu.c b/drivers/thermal/samsung/exynos_tmu.c
> index 0d96a51..ffccc89 100644
> --- a/drivers/thermal/samsung/exynos_tmu.c
> +++ b/drivers/thermal/samsung/exynos_tmu.c
> @@ -225,6 +225,8 @@ skip_calib_data:
> trigger_levs++;
> }
>
> + rising_threshold = readl(data->base + reg->threshold_th0);
You may move this inside "} else {" block as rising_threshold
is not used for data->soc == SOC_ARCH_EXYNOS4210 case.
Also rising_threshold initialization to zero at the beginning of
exynos_tmu_initialize() is not needed anylonger.
> +
> if (data->soc == SOC_ARCH_EXYNOS4210) {
> /* Write temperature code for threshold */
> threshold_code = temp_to_code(data, pdata->threshold);
> @@ -249,6 +251,7 @@ skip_calib_data:
> ret = threshold_code;
> goto out;
> }
> + rising_threshold &= ~(0xff << 8 * i);
> rising_threshold |= threshold_code << 8 * i;
> if (pdata->threshold_falling) {
> threshold_code = temp_to_code(data,
> @@ -281,6 +284,7 @@ skip_calib_data:
> }
> if (i == EXYNOS_MAX_TRIGGER_PER_REG - 1) {
> /* 1-4 level to be assigned in th0 reg */
> + rising_threshold &= ~(0xff << 8 * i);
> rising_threshold |= threshold_code << 8 * i;
> writel(rising_threshold,
> data->base + reg->threshold_th0);
Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics
On 04/24/2014 04:18 PM, Bartlomiej Zolnierkiewicz wrote:
>
> Hi,
>
> On Monday, April 14, 2014 11:08:15 AM Tushar Behera wrote:
>> Currently the threshold limits are updated in 2 stages, once for all
>> software trigger levels and again for hardware trip point.
>>
>> While updating the software trigger levels, it overwrites the threshold
>> limit for hardware trip point thereby forcing the Exynos core to issue
>> an emergency shutdown.
>
> On what SoC type have you encountered this problem? It doesn't seem to
> happen on older SoCs (at least Exynos4210 and Exynos4x12 ones).
>
I found this issue while testing on Exynos5420 with these patches [1].
[1] https://lkml.org/lkml/2013/8/1/38
>> Updating only the required fields in threshold register fixes this issue.
>
> With the current code there is indeed a time window during which threshold
> limit for hardware trip point is set to zero so the fix is correct.
>
--
Tushar Behera