On 2016-07-14 13:28, Hans Dedecker wrote: > On Thu, Jul 14, 2016 at 12:01 PM, Yousong Zhou <yszhou4t...@gmail.com> wrote: >> On 14 July 2016 at 16:14, Hans Dedecker <dedec...@gmail.com> wrote: >>> Commit c6858766 adds teardown support when l3_dev link is lost especially >>> for shell protocols >>> that have no proto task. However shell protocols which have a proto task >>> running like ppp will >>> also be teared down which is not always the expected action. >>> As an example the PPP daemon can be put into persist state trying to >>> re-establish the link via >>> a hold off mechanism which is not possible when the daemon is terminated by >>> the proto shell >>> teardown. >>> Therefore restrict the teardown action for shell protocols having no proto >>> task. >>> >> >> How about adding an extra flag like managed-link, persistent-link, >> on-demand-link? It looks to me doing teardown at link-down is more >> common a case. > Initially I was thinking about adding another flag like you propose > but then I was doubting if the change in behavior for shell protocols > having a proto task task was on purpose or not. In case of PPP and > link failure you don't want an immediate restart by netifd in some > cases (see https://github.com/lede-project/source/pull/200) as PPP > daemon can take care of the link re-negotiation based on a holdoff > timeout. > Additionally if the wan link loses connectivity a link down > notification will be received on the main device which will teardown > the protocol. Anyway I'm open for suggestions which way to go forward. Yousong,
please provide some more details on where your commit c6858766 is actually needed/useful. In all the use cases I can think of, handling setup/teardown based on the l2 dev should be enough. - Felix _______________________________________________ Lede-dev mailing list Lede-dev@lists.infradead.org http://lists.infradead.org/mailman/listinfo/lede-dev
