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

[Guenter Roeck]

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