> On Aug 10, 2016, at 10:52 PM, Robert Mustacchi <[email protected]> wrote:
> 
> On 8/10/16 19:46 , Dale Ghent wrote:
>> The problem stems from the card initializing at power-on and going into LPLU 
>> mode immediately. When the kernel gets around to booting and the ixgbe 
>> driver loads, the ixgbe driver (traditionally) has always presumed that the 
>> card will, upon the earlier power-on initialization, sync up at the fastest 
>> speed it can determine... which for non-LPLU situations is usually 10Gb 
>> (peer and cabling permitting, of course)... so ixgbe never bothered forcing 
>> it to re-neg the link. This is why when you use dladm to diddle the autoneg 
>> speed settings later on after boot, it'll come up at 10Gb.
>> 
>> It surprises me that the X540 has a LPLU in a on-board form... I can't 
>> recollect that being mentioned in the docs. It just means we need to add 
>> X540 to the list of MACs to force a on-start autonegotiation in ixgbe.
> 
> It's there in the docs, hiding there a bit. What I'm going to try and
> code up is something that looks at the NVM bit to determine if it's
> there and also indicate the presence as a devinfo property. I'll go and
> see if there's a similar NVM bit on the X550 as well that we can leverage.

+1 on that strategy.

/dale

Attachment: signature.asc
Description: Message signed with OpenPGP using GPGMail




-------------------------------------------
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=25769125&id_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com

Reply via email to