Quoting r. Roland Dreier <[EMAIL PROTECTED]>: > Subject: Re: IPoIB and lid change > > Michael> One simple way to address this would be to have a list of > Michael> all address handles per net device and kill them on an SM > Michael> change event. > > Seems reasonable. It seems a little painful to implement at a first > glance but I might be looking at it wrong.
Here's my plan (might not have the time to look at it before Sunday which is why I'm not just doing it and sending a patch): - start with ipoib_all_neigh_issues_2.patch - make the neigh list per-device - now in the destructor we'll walk the list of devices rather than the global neigh list - walk all neighbours and kill lids on client reregistration event We now get the destructor issue fixed in a slightly less ugly way, code will be ready for when destructor moves into the params structure, and the LID change will get fixed for simplest/common case. Does this look like a good plan? -- Michael S. Tsirkin Staff Engineer, Mellanox Technologies _______________________________________________ openib-general mailing list [email protected] http://openib.org/mailman/listinfo/openib-general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
