Re: PROBLEM: reproducible crash KVM+nf_conntrack all recent 2.6kernels

[Jon Masters]

Hi,

So I did some poking (still trying to figure out netfilter a little
internally) and looked over the handling of connection tracking. The
oops reports I have been getting generally lie in __nf_conntrack_find,
specifically within a hlist iterator that looks up the information for
the current connection in a per-net namespace hashtable (under RCU, it's
been locked already by the time we get in here). Here's the piece:

        hlist_nulls_for_each_entry_rcu(h, n, &net->ct.hash[hash],
hnnode) {
                if (nf_ct_tuple_equal(tuple, &h->tuple)) {
                        NF_CT_STAT_INC(net, found);
                        local_bh_enable();
                        return h;
                }
                NF_CT_STAT_INC(net, searched);
        }

Instrumenting the kernel at the moment and then setting up more of a
debugging environment to poke at what goes wrong here. Perhaps there's
some broken RCU assumption - I just spent the last few hours reading
over netfilter source and Paul's RCU docs again to brush up.

Perhaps you netdev folks can let me know if there's a handy netfilter
debugging guide somewhere.

Jon.


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/