On 6/18/08, Roland Dreier <[EMAIL PROTECTED]> wrote:
>> just saving pointer to the netdevice seemed quite not elegant...

> To me using the pointer to the netdev seems much cleaner than doing a
>  string comparison on the name, but I'm probably missing something.  Why
>  didn't you like that idea?

I was thinking that I would need to go into reference counting / book
keeping on the netdevice, but as Steve noted, it seems doable, and I guess I
don't need to issue a dev_get() just for keeping the copy of neigh->dev,
correct? since its holds the rtnl lock, the netevent notififier should
provide a safe context for comparision event->dev to id->addr->dev.  If we
agree on this approach and you prefer it, I can change the patch.

Or.
_______________________________________________
general mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general

To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general

Reply via email to