Re: [PATCH 4/4] i2c: octeon: thunderx: Add I2C_CLASS_HWMON

[Jan Glauber <jan.glauber@caviumnetworks.com>]

On Thu, Apr 20, 2017 at 05:55:29PM +0200, Wolfram Sang wrote:
> Hi Jan,
> 
> > We do want auto-detection, so either this text is wrong or setting
> > I2C_CLASS_HWMON in our i2c adapter is correct. Am I missing
> > something?
> 
> I was assuming the pci-driver could ACPI for device instantiation, and
> the platdrv could use DT. Then, the adapter class usage raised my
> eyebrow.
> 
> I just saw that the pci-driver does not have ACPI support, yet. So, the
> class could be used. Not sure about the platdrv case? But even then, the
> adapter will probe EVERY hwmon device as soon as its i2c client driver
> loads. Do you really want that? This may cost boot time, has unneeded
> traffic on the bus, etc.

What is missing in the pci-driver for ACPI support? We already use ACPI
to detect the sclk setting.

The automatic probing is requested by distributions. As we have to support
both ACPI and DT there I thought of letting ipmi-ssif use the smbios/dmi
information that is already there and usable regardless of ACPI / DT.

For servers I don't think the probing overhead is an issue, the firmware
only exposes the BMC device there. On the "embedded" systems I've not
seen a BMC yet and anyone attempting do minimize boot time can disable
the automatic probing.

So, if you're ok with this I'll re-phrase the commit message and
re-submit the patch.

Gruß,
Jan

> Regards,
> 
>    Wolfram
>