[ofa-general] Re: bonding & ipoib patches to support bonding of ipoib devices

[Or Gerlitz]

Michael S. Tsirkin wrote:
>> Quoting Or Gerlitz <ogerlitz at voltaire.com>:
>>> 2.Ah kept in the ipoib_neigh might belong to a device which is different
>>>  from the one start_xmit is called at.
>> Michael, Roland,
>>
>> Following the high-availability/bonding session at Sonoma, I'd like to 
>> have a BOF to discuss the issues which from your perspective should be 
>> addressed before the patch set is merged upstream. Will you be around?
> 
> I won't be able to make it unfortunately.

OK, I see. I assume that this means we would have to continue at least 
this part of the discussion (down the road there are also configuration 
issues) over the list, let it be.

>> I just need a clarification on (2), can you educate me how can it 
>> happen?
> 
> Consider a specific neighbour. Can the following happen? If no, why not?
> - a packet is sent through ib0, ipoib_neigh and ->ah are created
> - bonding switches to ib1 for some reason
> - before ipoib_neigh and ah are flushed, a new skb arrives. It is posted on
>   ib1 with ah which belongs to ib0

oh well, this is very rare but possible. IB wise, what would happen on 
this case to the device (ib1 in your example) QP? will this wrong ah 
usage in the WR cause it get into the error case?

Or.