On Fri, Feb 12, 2021 at 4:14 PM Alexander V. Chernikov <melif...@freebsd.org> wrote: > The slightly different approach here: https://reviews.freebsd.org/D28629 > We indeed are running under epoch, so that prevents _immediate_ ifa > destruction. > However, we still can run into the situation when > * in thread 1 we drop to 0 refcount for ifa and schedule its deletion. > * in thread 2 we use this ifa and reference it > * destroy callout kicks in > * unhappy user reports bug > The current approach minimises this possibility by taking an ifa refcount > early. > More general approach would probably be in introducing `ifa_try_ref()` based > on refcount_acquire_if_not_zero() but that's something that needs a bit more > effort.
Oops, you're completely correct. Do you think that we should put a KASSERT in ifa_ref() that ifa_refcnt > 0 to catch such a bug if somebody ever manages to introduce one? Of course it would not fire except in the case where we already lost the race, but at least it would make it easier to debug. _______________________________________________ freebsd-net@freebsd.org mailing list https://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to "freebsd-net-unsubscr...@freebsd.org"
