From 4c05ec47384ab3627b62814e8f886e90cc38ce15 Mon Sep 17 00:00:00 2001 From: Taehee Yoo Date: Mon, 26 Nov 2018 20:03:30 +0900 Subject: netfilter: nf_tables: fix suspicious RCU usage in nft_chain_stats_replace() basechain->stats is rcu protected data which is updated from nft_chain_stats_replace(). This function is executed from the commit phase which holds the pernet nf_tables commit mutex - not the global nfnetlink subsystem mutex. Test commands to reproduce the problem are: %iptables-nft -I INPUT %iptables-nft -Z %iptables-nft -Z This patch uses RCU calls to handle basechain->stats updates to fix a splat that looks like: [89279.358755] ============================= [89279.363656] WARNING: suspicious RCU usage [89279.368458] 4.20.0-rc2+ #44 Tainted: G W L [89279.374661] ----------------------------- [89279.379542] net/netfilter/nf_tables_api.c:1404 suspicious rcu_dereference_protected() usage! [...] [89279.406556] 1 lock held by iptables-nft/5225: [89279.411728] #0: 00000000bf45a000 (&net->nft.commit_mutex){+.+.}, at: nf_tables_valid_genid+0x1f/0x70 [nf_tables] [89279.424022] stack backtrace: [89279.429236] CPU: 0 PID: 5225 Comm: iptables-nft Tainted: G W L 4.20.0-rc2+ #44 [89279.430135] Call Trace: [89279.430135] dump_stack+0xc9/0x16b [89279.430135] ? show_regs_print_info+0x5/0x5 [89279.430135] ? lockdep_rcu_suspicious+0x117/0x160 [89279.430135] nft_chain_commit_update+0x4ea/0x640 [nf_tables] [89279.430135] ? sched_clock_local+0xd4/0x140 [89279.430135] ? check_flags.part.35+0x440/0x440 [89279.430135] ? __rhashtable_remove_fast.constprop.67+0xec0/0xec0 [nf_tables] [89279.430135] ? sched_clock_cpu+0x126/0x170 [89279.430135] ? find_held_lock+0x39/0x1c0 [89279.430135] ? hlock_class+0x140/0x140 [89279.430135] ? is_bpf_text_address+0x5/0xf0 [89279.430135] ? check_flags.part.35+0x440/0x440 [89279.430135] ? __lock_is_held+0xb4/0x140 [89279.430135] nf_tables_commit+0x2555/0x39c0 [nf_tables] Fixes: f102d66b335a4 ("netfilter: nf_tables: use dedicated mutex to guard transactions") Signed-off-by: Taehee Yoo Signed-off-by: Pablo Neira Ayuso --- net/netfilter/nf_tables_api.c | 21 +++++++++++++-------- net/netfilter/nf_tables_core.c | 2 +- 2 files changed, 14 insertions(+), 9 deletions(-) (limited to 'net') diff --git a/net/netfilter/nf_tables_api.c b/net/netfilter/nf_tables_api.c index 2e61aab6ed73..6e548d7c9f67 100644 --- a/net/netfilter/nf_tables_api.c +++ b/net/netfilter/nf_tables_api.c @@ -1216,7 +1216,8 @@ static int nf_tables_fill_chain_info(struct sk_buff *skb, struct net *net, if (nla_put_string(skb, NFTA_CHAIN_TYPE, basechain->type->name)) goto nla_put_failure; - if (basechain->stats && nft_dump_stats(skb, basechain->stats)) + if (rcu_access_pointer(basechain->stats) && + nft_dump_stats(skb, rcu_dereference(basechain->stats))) goto nla_put_failure; } @@ -1392,7 +1393,8 @@ static struct nft_stats __percpu *nft_stats_alloc(const struct nlattr *attr) return newstats; } -static void nft_chain_stats_replace(struct nft_base_chain *chain, +static void nft_chain_stats_replace(struct net *net, + struct nft_base_chain *chain, struct nft_stats __percpu *newstats) { struct nft_stats __percpu *oldstats; @@ -1400,8 +1402,9 @@ static void nft_chain_stats_replace(struct nft_base_chain *chain, if (newstats == NULL) return; - if (chain->stats) { - oldstats = nfnl_dereference(chain->stats, NFNL_SUBSYS_NFTABLES); + if (rcu_access_pointer(chain->stats)) { + oldstats = rcu_dereference_protected(chain->stats, + lockdep_commit_lock_is_held(net)); rcu_assign_pointer(chain->stats, newstats); synchronize_rcu(); free_percpu(oldstats); @@ -1439,9 +1442,10 @@ static void nf_tables_chain_destroy(struct nft_ctx *ctx) struct nft_base_chain *basechain = nft_base_chain(chain); module_put(basechain->type->owner); - free_percpu(basechain->stats); - if (basechain->stats) + if (rcu_access_pointer(basechain->stats)) { static_branch_dec(&nft_counters_enabled); + free_percpu(rcu_dereference_raw(basechain->stats)); + } kfree(chain->name); kfree(basechain); } else { @@ -1590,7 +1594,7 @@ static int nf_tables_addchain(struct nft_ctx *ctx, u8 family, u8 genmask, kfree(basechain); return PTR_ERR(stats); } - basechain->stats = stats; + rcu_assign_pointer(basechain->stats, stats); static_branch_inc(&nft_counters_enabled); } @@ -6180,7 +6184,8 @@ static void nft_chain_commit_update(struct nft_trans *trans) return; basechain = nft_base_chain(trans->ctx.chain); - nft_chain_stats_replace(basechain, nft_trans_chain_stats(trans)); + nft_chain_stats_replace(trans->ctx.net, basechain, + nft_trans_chain_stats(trans)); switch (nft_trans_chain_policy(trans)) { case NF_DROP: diff --git a/net/netfilter/nf_tables_core.c b/net/netfilter/nf_tables_core.c index 3fbce3b9c5ec..a50500232b0a 100644 --- a/net/netfilter/nf_tables_core.c +++ b/net/netfilter/nf_tables_core.c @@ -101,7 +101,7 @@ static noinline void nft_update_chain_stats(const struct nft_chain *chain, struct nft_stats *stats; base_chain = nft_base_chain(chain); - if (!base_chain->stats) + if (!rcu_access_pointer(base_chain->stats)) return; local_bh_disable(); -- cgit v1.2.3 From 530aad77010b81526586dfc09130ec875cd084e4 Mon Sep 17 00:00:00 2001 From: Florian Westphal Date: Wed, 5 Dec 2018 14:12:19 +0100 Subject: netfilter: seqadj: re-load tcp header pointer after possible head reallocation When adjusting sack block sequence numbers, skb_make_writable() gets called to make sure tcp options are all in the linear area, and buffer is not shared. This can cause tcp header pointer to get reallocated, so we must reaload it to avoid memory corruption. This bug pre-dates git history. Reported-by: Neel Mehta Reported-by: Shane Huntley Reported-by: Heather Adkins Signed-off-by: Florian Westphal Signed-off-by: Pablo Neira Ayuso --- net/netfilter/nf_conntrack_seqadj.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'net') diff --git a/net/netfilter/nf_conntrack_seqadj.c b/net/netfilter/nf_conntrack_seqadj.c index a975efd6b8c3..9da303461069 100644 --- a/net/netfilter/nf_conntrack_seqadj.c +++ b/net/netfilter/nf_conntrack_seqadj.c @@ -115,12 +115,12 @@ static void nf_ct_sack_block_adjust(struct sk_buff *skb, /* TCP SACK sequence number adjustment */ static unsigned int nf_ct_sack_adjust(struct sk_buff *skb, unsigned int protoff, - struct tcphdr *tcph, struct nf_conn *ct, enum ip_conntrack_info ctinfo) { - unsigned int dir, optoff, optend; + struct tcphdr *tcph = (void *)skb->data + protoff; struct nf_conn_seqadj *seqadj = nfct_seqadj(ct); + unsigned int dir, optoff, optend; optoff = protoff + sizeof(struct tcphdr); optend = protoff + tcph->doff * 4; @@ -128,6 +128,7 @@ static unsigned int nf_ct_sack_adjust(struct sk_buff *skb, if (!skb_make_writable(skb, optend)) return 0; + tcph = (void *)skb->data + protoff; dir = CTINFO2DIR(ctinfo); while (optoff < optend) { @@ -207,7 +208,7 @@ int nf_ct_seq_adjust(struct sk_buff *skb, ntohl(newack)); tcph->ack_seq = newack; - res = nf_ct_sack_adjust(skb, protoff, tcph, ct, ctinfo); + res = nf_ct_sack_adjust(skb, protoff, ct, ctinfo); out: spin_unlock_bh(&ct->lock); -- cgit v1.2.3 From 708abf74dd87f8640871b814faa195fb5970b0e3 Mon Sep 17 00:00:00 2001 From: Pan Bian Date: Mon, 10 Dec 2018 14:39:37 +0100 Subject: netfilter: ipset: do not call ipset_nest_end after nla_nest_cancel In the error handling block, nla_nest_cancel(skb, atd) is called to cancel the nest operation. But then, ipset_nest_end(skb, atd) is unexpected called to end the nest operation. This patch calls the ipset_nest_end only on the branch that nla_nest_cancel is not called. Fixes: 45040978c899 ("netfilter: ipset: Fix set:list type crash when flush/dump set in parallel") Signed-off-by: Pan Bian Signed-off-by: Jozsef Kadlecsik Signed-off-by: Pablo Neira Ayuso --- net/netfilter/ipset/ip_set_list_set.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'net') diff --git a/net/netfilter/ipset/ip_set_list_set.c b/net/netfilter/ipset/ip_set_list_set.c index 4eef55da0878..8da228da53ae 100644 --- a/net/netfilter/ipset/ip_set_list_set.c +++ b/net/netfilter/ipset/ip_set_list_set.c @@ -531,8 +531,8 @@ nla_put_failure: ret = -EMSGSIZE; } else { cb->args[IPSET_CB_ARG0] = i; + ipset_nest_end(skb, atd); } - ipset_nest_end(skb, atd); out: rcu_read_unlock(); return ret; -- cgit v1.2.3 From 542fbda0f08f1cbbc250f9e59f7537649651d0c8 Mon Sep 17 00:00:00 2001 From: Florian Westphal Date: Tue, 11 Dec 2018 07:45:29 +0100 Subject: netfilter: nat: can't use dst_hold on noref dst The dst entry might already have a zero refcount, waiting on rcu list to be free'd. Using dst_hold() transitions its reference count to 1, and next dst release will try to free it again -- resulting in a double free: WARNING: CPU: 1 PID: 0 at include/net/dst.h:239 nf_xfrm_me_harder+0xe7/0x130 [nf_nat] RIP: 0010:nf_xfrm_me_harder+0xe7/0x130 [nf_nat] Code: 48 8b 5c 24 60 65 48 33 1c 25 28 00 00 00 75 53 48 83 c4 68 5b 5d 41 5c c3 85 c0 74 0d 8d 48 01 f0 0f b1 0a 74 86 85 c0 75 f3 <0f> 0b e9 7b ff ff ff 29 c6 31 d2 b9 20 00 48 00 4c 89 e7 e8 31 27 Call Trace: nf_nat_ipv4_out+0x78/0x90 [nf_nat_ipv4] nf_hook_slow+0x36/0xd0 ip_output+0x9f/0xd0 ip_forward+0x328/0x440 ip_rcv+0x8a/0xb0 Use dst_hold_safe instead and bail out if we cannot take a reference. Fixes: a4c2fd7f7891 ("net: remove DST_NOCACHE flag") Reported-by: Martin Zaharinov Signed-off-by: Florian Westphal Signed-off-by: Pablo Neira Ayuso --- net/netfilter/nf_nat_core.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'net') diff --git a/net/netfilter/nf_nat_core.c b/net/netfilter/nf_nat_core.c index e2b196054dfc..2268b10a9dcf 100644 --- a/net/netfilter/nf_nat_core.c +++ b/net/netfilter/nf_nat_core.c @@ -117,7 +117,8 @@ int nf_xfrm_me_harder(struct net *net, struct sk_buff *skb, unsigned int family) dst = skb_dst(skb); if (dst->xfrm) dst = ((struct xfrm_dst *)dst)->route; - dst_hold(dst); + if (!dst_hold_safe(dst)) + return -EHOSTUNREACH; if (sk && !net_eq(net, sock_net(sk))) sk = NULL; -- cgit v1.2.3 From d4e7df16567b80836a78d31b42f1a9355a636d67 Mon Sep 17 00:00:00 2001 From: Taehee Yoo Date: Sat, 8 Dec 2018 11:03:01 +0900 Subject: netfilter: nf_conncount: use rb_link_node_rcu() instead of rb_link_node() rbnode in insert_tree() is rcu protected pointer. So, in order to handle this pointer, _rcu function should be used. rb_link_node_rcu() is a rcu version of rb_link_node(). Fixes: 34848d5c896e ("netfilter: nf_conncount: Split insert and traversal") Signed-off-by: Taehee Yoo Signed-off-by: Pablo Neira Ayuso --- net/netfilter/nf_conncount.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'net') diff --git a/net/netfilter/nf_conncount.c b/net/netfilter/nf_conncount.c index b6d0f6deea86..9cd180bda092 100644 --- a/net/netfilter/nf_conncount.c +++ b/net/netfilter/nf_conncount.c @@ -427,7 +427,7 @@ insert_tree(struct net *net, count = 1; rbconn->list.count = count; - rb_link_node(&rbconn->node, parent, rbnode); + rb_link_node_rcu(&rbconn->node, parent, rbnode); rb_insert_color(&rbconn->node, root); out_unlock: spin_unlock_bh(&nf_conncount_locks[hash % CONNCOUNT_LOCK_SLOTS]); -- cgit v1.2.3