> If so, please make it clear that this is only used for physical links.

Okay.

> This leads to another two questions:
> 
> a. Do you expect legacy physical drivers to use this interface, and how?

Nope.

> b. Can we remove the second m_instance argument? As I said, it is impossible 
> for a driver to specify it if it is different from the instance number of the 
> dip.

During design review, we agreed to do it this way just to be consistent 
with mac_register(). I would just leave it there, why split hairs when 
we can not to.

>> aspect that bothers me is interaction of dlmgmtd with partially attached
>> devices - i.e. the attach(9E)->mac_prop_init->mac_prop_load->... chain.
>>   Do you know of any existing kernel functions to translate macnames
>> into devnames and vice versa?
>>
> No, there is no existing functions can do this. Especially if it is before 
> _attach(), when a mac is not even registered, and mac name is not determined 
> yet.

Exactly. That's the reason I'm using macname, because at this early 
stage it's the only thing available for sure. Otherwise we get into a 
deadlock.

> I understand that this RFE is for this purpose. But I have incorrectly 
> thought 
> that datalink autopush must be configured before plumbing.
> 
> But note mac_start() will be moved to be part of DL_BIND_REQ processing soon 
> (as part of the Clearview softmac fast-path work). In my opinion, mac_start() 
> should have been called as the result of DL_BIND_REQ, as only from then on, 
> the driver is allowed to transfer data.
> 
> So my question changes to: Is it possible to move mac_prop_load() call from 
> mac_start() to mac_open()?

Given that mac_open happens before mac_start, I think yes, this should 
be possible. I'll give it a try.

-Artem
_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to