On Fri, 9 Oct 2020 18:59:59 +0200 Paolo Abeni wrote: > pktdrill pointed-out we currently don't handle properly some > fallback scenario for MP_JOIN subflows > > The first patch addresses such issue. > > Patch 2/2 fixes a related pre-existing issue that is more > evident after 1/2: we could keep using for MPTCP signaling > closed subflows.
Applied, thanks Paolo. You already have a few of those in the code, but: + if (... && + schedule_work(&mptcp_sk(sk)->work)) + sock_hold(sk); isn't this a fairly questionable construct? You take a reference for the async work to release _after_ you scheduled the async work?
