Garrett D'Amore wrote:
> Steven Stallion wrote:
>> Garrett D'Amore wrote:
>>  
>>> If the PHY isn't initialized (not plumbed) you should report "default"
>>> values indicating no speed or duplex, IMO.
>>>
>>> But it isn't that big a deal as long as the STATE isn't misreported.
>>>     
>>
>> Those were my thoughts as well. It's odd to see an unplumbed interface
>> report duplex and speed, even if the link state is set to unknown.
>>
>> So, lets say I defer PHY initialization until m_start is called; should
>> the adv_* and en_* properties reflect a best guess prior to the plumb?
>> It wouldn't take much work to check the chipset as a part of attach and
>> set the appropriate flags, but other drivers (i.e. afe) don't seem to do
>> this and instead opt to do everything as a part of the PHY init.
>>   
> Best guess is probably wrong approach.  In some of these cases what the
> value will depend on the peer or which xcvr is selected (for NICs with
> multiple xcvrs, you might not know this until you try to initialize the
> PHY.)
> 
> Again, it probably doesn't matter that much, but I'd opt for consistency
> where possible.

Gotcha.

Thanks!

-- 
Yet magic and hierarchy
arise from the same source,
and this source has a null pointer.

Reference the NULL within NULL,
it is the gateway to all wizardry.
_______________________________________________
driver-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/driver-discuss

Reply via email to