On 1/17/17 1:38 PM, David Miller wrote: > But, that being said, I don't think we can legitimately just remove the > autoload functionality. It's been there for close to a full year and > I guarantee people will get burned if we take it away. > > We have to find a way to fix this.
I have patch that removes the dev from build_state. Nothing about an encap state should require a device and none of the current encaps use it so it can be removed without a problem. After that the only reference I have noted is to a table and the only table that can be deleted - and invalidating the pointer - is the local table and that only happens once on fib_unmerge. In short seems like removing the dev + the current patch dropping the lock fixes the current deadlock problem and should be fine. Robert: do you agree?
