Major Dávid <major.david@xxxxxxxxxx> wrote: > Hi all, Hi Pablo, > > Following comments on bug: https://bugzilla.netfilter.org/show_bug.cgi?id=1662 > > I researched a little bit further. I started to use older kernels from > Ubuntu mainline PPA to get a rough estimate about affected kernel versions. > At kernel version v3.18.140 still getting a soft lockup, I spun up a good > oldie' Ubuntu Precise VM environment to test older kernels. No nftables > support in Precise so I ported my test ruleset to iptables: > > iptables -t mangle -F > iptables -t mangle -X Thanks, this is a bug in nft_tproxy.c. Can you test following fix? Thanks! Subject: netfilter: tproxy: fix deadlock due to missing BH disable The xtables packet traverser performs an unconditional local_bh_disable(), but the nf_tables evaluation loop does not. Functions that are called from either xtables or nftables must assume that they can be called in process context. inet_twsk_deschedule_put() assumes that no softirq interrupt can occur. If tproxy is used from nf_tables its possible that we'll deadlock trying to aquire a lock already held in process context. diff --git a/include/net/netfilter/nf_tproxy.h b/include/net/netfilter/nf_tproxy.h --- a/include/net/netfilter/nf_tproxy.h +++ b/include/net/netfilter/nf_tproxy.h @@ -17,6 +17,13 @@ static inline bool nf_tproxy_sk_is_transparent(struct sock *sk) return false; } +static inline void nf_tproxy_twsk_deschedule_put(struct inet_timewait_sock *tw) +{ + local_bh_disable(); + inet_twsk_deschedule_put(tw); + local_bh_enable(); +} + /* assign a socket to the skb -- consumes sk */ static inline void nf_tproxy_assign_sock(struct sk_buff *skb, struct sock *sk) { diff --git a/net/ipv4/netfilter/nf_tproxy_ipv4.c b/net/ipv4/netfilter/nf_tproxy_ipv4.c --- a/net/ipv4/netfilter/nf_tproxy_ipv4.c +++ b/net/ipv4/netfilter/nf_tproxy_ipv4.c @@ -38,7 +38,7 @@ nf_tproxy_handle_time_wait4(struct net *net, struct sk_buff *skb, hp->source, lport ? lport : hp->dest, skb->dev, NF_TPROXY_LOOKUP_LISTENER); if (sk2) { - inet_twsk_deschedule_put(inet_twsk(sk)); + nf_tproxy_twsk_deschedule_put(inet_twsk(sk)); sk = sk2; } } diff --git a/net/ipv6/netfilter/nf_tproxy_ipv6.c b/net/ipv6/netfilter/nf_tproxy_ipv6.c --- a/net/ipv6/netfilter/nf_tproxy_ipv6.c +++ b/net/ipv6/netfilter/nf_tproxy_ipv6.c @@ -63,7 +63,7 @@ nf_tproxy_handle_time_wait6(struct sk_buff *skb, int tproto, int thoff, lport ? lport : hp->dest, skb->dev, NF_TPROXY_LOOKUP_LISTENER); if (sk2) { - inet_twsk_deschedule_put(inet_twsk(sk)); + nf_tproxy_twsk_deschedule_put(inet_twsk(sk)); sk = sk2; } }
