[net-next,v1] net: phy: nxp-tja11xx: log critical health state

TJA1102 provides interrupt notification for the critical health states
like overtemperature and undervoltage.

The overtemperature bit is set if package temperature is beyond 155C°.
This functionality was tested by heating the package up to 200C°

The undervoltage bit is set if supply voltage drops beyond some critical
threshold. Currently not tested.

In a typical use case, both of this events should be logged and stored
(or send to some remote system) for further investigations.

To make testing and troubleshooting more easier, other bits are exported
over the ethtool phy stats interface:
- level of overtemperature. There is no access to thermal sensor.
- power supply line where undervoltage was detected

Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
---
 drivers/net/phy/nxp-tja11xx.c | 19 +++++++++++++++++--
 1 file changed, 17 insertions(+), 2 deletions(-)

Hi Oleksij

> @@ -89,6 +91,12 @@ static struct tja11xx_phy_stats tja11xx_hw_stats[] = {
>  	{ "phy_polarity_detect", 25, 6, BIT(6) },
>  	{ "phy_open_detect", 25, 7, BIT(7) },
>  	{ "phy_short_detect", 25, 8, BIT(8) },
> +	{ "phy_temp_warn (temp > 155C°)", 25, 9, BIT(9) },
> +	{ "phy_temp_high (temp > 180C°)", 25, 10, BIT(10) },
> +	{ "phy_uv_vddio", 25, 11, BIT(11) },
> +	{ "phy_uv_vddd_1v8", 25, 13, BIT(13) },
> +	{ "phy_uv_vdda_3v3", 25, 14, BIT(14) },
> +	{ "phy_uv_vddd_3v3", 25, 15, BIT(15) },
>  	{ "phy_rem_rcvr_count", 26, 0, GENMASK(7, 0) },
>  	{ "phy_loc_rcvr_count", 26, 8, GENMASK(15, 8) },

I'm not so happy abusing the statistic counters like this. Especially
when we have a better API for temperature and voltage: hwmon.

phy_temp_warn maps to hwmon_temp_max_alarm. phy_temp_high maps to
either hwmon_temp_crit_alarm or hwmon_temp_emergency_alarm.

The under voltage maps to hwmon_in_lcrit_alarm.

> @@ -630,6 +640,11 @@ static irqreturn_t tja11xx_handle_interrupt(struct phy_device *phydev)
>  		return IRQ_NONE;
>  	}
>  
> +	if (irq_status & MII_INTSRC_TEMP_ERR)
> +		dev_err(dev, "Overtemperature error detected (temp > 155C°).\n");
> +	if (irq_status & MII_INTSRC_UV_ERR)
> +		dev_err(dev, "Undervoltage error detected.\n");
> +

These are not actual errors, in the linux sense. So dev_warn() or
maybe dev_info().

      Andrew

On 8/10/21 8:05 AM, Andrew Lunn wrote:
> Hi Oleksij
> 
>> @@ -89,6 +91,12 @@ static struct tja11xx_phy_stats tja11xx_hw_stats[] = {
>>   	{ "phy_polarity_detect", 25, 6, BIT(6) },
>>   	{ "phy_open_detect", 25, 7, BIT(7) },
>>   	{ "phy_short_detect", 25, 8, BIT(8) },
>> +	{ "phy_temp_warn (temp > 155C°)", 25, 9, BIT(9) },
>> +	{ "phy_temp_high (temp > 180C°)", 25, 10, BIT(10) },
>> +	{ "phy_uv_vddio", 25, 11, BIT(11) },
>> +	{ "phy_uv_vddd_1v8", 25, 13, BIT(13) },
>> +	{ "phy_uv_vdda_3v3", 25, 14, BIT(14) },
>> +	{ "phy_uv_vddd_3v3", 25, 15, BIT(15) },
>>   	{ "phy_rem_rcvr_count", 26, 0, GENMASK(7, 0) },
>>   	{ "phy_loc_rcvr_count", 26, 8, GENMASK(15, 8) },
> 
> I'm not so happy abusing the statistic counters like this. Especially
> when we have a better API for temperature and voltage: hwmon.
> 
> phy_temp_warn maps to hwmon_temp_max_alarm. phy_temp_high maps to
> either hwmon_temp_crit_alarm or hwmon_temp_emergency_alarm.
> 
> The under voltage maps to hwmon_in_lcrit_alarm.
> 

FWIW, the statistics counters in this driver are already abused
(phy_polarity_detect, phy_open_detect, phy_short_detect), so
I am not sure if adding more abuse makes a difference (and/or
if such abuse is common for phy drivers in general).

Guenter

>> @@ -630,6 +640,11 @@ static irqreturn_t tja11xx_handle_interrupt(struct phy_device *phydev)
>>   		return IRQ_NONE;
>>   	}
>>   
>> +	if (irq_status & MII_INTSRC_TEMP_ERR)
>> +		dev_err(dev, "Overtemperature error detected (temp > 155C°).\n");
>> +	if (irq_status & MII_INTSRC_UV_ERR)
>> +		dev_err(dev, "Undervoltage error detected.\n");
>> +
> 
> These are not actual errors, in the linux sense. So dev_warn() or
> maybe dev_info().
> 
>        Andrew
>

> > I'm not so happy abusing the statistic counters like this. Especially
> > when we have a better API for temperature and voltage: hwmon.
> > 
> > phy_temp_warn maps to hwmon_temp_max_alarm. phy_temp_high maps to
> > either hwmon_temp_crit_alarm or hwmon_temp_emergency_alarm.
> > 
> > The under voltage maps to hwmon_in_lcrit_alarm.
> > 
> 
> FWIW, the statistics counters in this driver are already abused
> (phy_polarity_detect, phy_open_detect, phy_short_detect), so
> I am not sure if adding more abuse makes a difference (and/or
> if such abuse is common for phy drivers in general).

Hi Guenter

Abuse is not common in general. I think this is the only driver
abusing stats to return flags.  At the time those where added, we did
not have phy cable test support. Now we do, i would also suggest that
the driver makes use of that infrastructure to issue a cable test
report. These 'stats' need to stay, since they are ABI, but we should
not add more.

That is also why i said "Especially when we have a better API".

     Andrew