[Intel-wired-lan] [PATCH iwl-next v2] ixgbe: add support for thermal sensor event reception

[Jagielski, Jedrzej]

From: Andrew Lunn <andrew at lunn.ch> 
Sent: Thursday, February 6, 2025 2:59 PM
>On Thu, Feb 06, 2025 at 01:05:27PM +0000, Jagielski, Jedrzej wrote:
>> From: Andrew Lunn <andrew at lunn.ch> 
>> Sent: Tuesday, February 4, 2025 2:09 PM
>> >On Tue, Feb 04, 2025 at 08:17:00AM +0100, Jedrzej Jagielski wrote:
>> >> E610 NICs unlike the previous devices utilising ixgbe driver
>> >> are notified in the case of overheatning by the FW ACI event.
>> >> 
>> >> In event of overheat when treshold is exceeded, FW suspends all
>> >> traffic and sends overtemp event to the driver. Then driver
>> >> logs appropriate message and closes the adapter instance.
>> >> The card remains in that state until the platform is rebooted.
>> >
>> >There is also an HWMON temp[1-*]_emergency_alarm you can set. I
>> >_think_ that should also cause a udev event, so user space knows the
>> >print^h^h^h^h^hnetwork is on fire.
>> >
>> >	Andrew
>> 
>> I am not sure whether HWMON is applicable in that case.
>> Driver receives an async notification from the FW that an overheating
>> occurred, so has to handle it. In that case - by printing msg
>> and making the interface disabled for the user.
>> FW is responsible for monitoring temperature itself.
>> There's even no possibility to read temperature by the driver
>
>https://elixir.bootlin.com/linux/v6.13.1/source/drivers/net/ethernet/intel/ixgbe/ixgbe_sysfs.c#L27
>
>ixgbe_hwmon_show_temp() is some other temperature sensor? Which you do
>have HWMON support for?

This feature is not supported for E610 which has no support for reading
temperature

hw->mac.ops.get_thermal_sensor_data() callback used in
ixgbe_hwmon_show_temp has no implementation for E610, as there is no
such support from the FW side

>
>Or is the E610 not really an ixgbe, it has a different architecture,

ixgbe is used by several adapters, each is slightly different
in this case monitoring stuff is pushed into FW

>more stuff pushed into firmware, less visibility from the kernel, no
>temperature monitoring, just a NIC on fire indication?

yeah, right

Jedrek