On Sun, 2014-01-12 at 21:50 +0400, Andrey Vagin wrote: > Lets look at destroy_conntrack: > > hlist_nulls_del_rcu(&ct->tuplehash[IP_CT_DIR_ORIGINAL].hnnode); > ... > nf_conntrack_free(ct) > kmem_cache_free(net->ct.nf_conntrack_cachep, ct); > > net->ct.nf_conntrack_cachep is created with SLAB_DESTROY_BY_RCU. > > The hash is protected by rcu, so readers look up conntracks without > locks. > A conntrack is removed from the hash, but in this moment a few readers > still can use the conntrack. Then this conntrack is released and another > thread creates conntrack with the same address and the equal tuple. > After this a reader starts to validate the conntrack: > * It's not dying, because a new conntrack was created > * nf_ct_tuple_equal() returns true. ... > v2: move nf_ct_is_confirmed into the unlikely() annotation > v3: Eric suggested to fix refcnt, so that it becomes zero before adding > in a hash, but we can't find a way how to do that. Another way is to > interpret the confirm bit as part of a search key and check it in > ____nf_conntrack_find() too. > > Cc: Eric Dumazet <eric.dumazet@xxxxxxxxx> > Cc: Florian Westphal <fw@xxxxxxxxx> > Cc: Pablo Neira Ayuso <pablo@xxxxxxxxxxxxx> > Cc: Patrick McHardy <kaber@xxxxxxxxx> > Cc: Jozsef Kadlecsik <kadlec@xxxxxxxxxxxxxxxxx> > Cc: "David S. Miller" <davem@xxxxxxxxxxxxx> > Cc: Cyrill Gorcunov <gorcunov@xxxxxxxxxx> > Signed-off-by: Andrey Vagin <avagin@xxxxxxxxxx> > --- Acked-by: Eric Dumazet <edumazet@xxxxxxxxxx> Thanks Andrey ! -- To unsubscribe from this list: send the line "unsubscribe netfilter" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
