Skip to contentSkip to conversation

netfilter: nf_tables: fix oops during rule dump

PATCH Applied Florian Westphal 3 participants 3 messages Apr 30, 2019 – May 20, 2019

Subsystems (2)
lore
Share & tools
lore

Florian Westphal

author maintainer
#0 lore raw

PATCH 1/1 netfilter: nf_tables: fix oops during rule dump

We can oops in nf_tables_fill_rule_info().

Its not possible to fetch previous element in rcu-protected lists when deletions are not prevented somehow: list_del_rcu poisons the ->prev pointer value.

Before rcu-conversion this was safe as dump operations did hold nfnetlink mutex.

Pass previous rule as argument, obtained by keeping a pointer to the previous rule during traversal.

Fixes d9adf22a291883 ("netfilter: nf_tables: use call_rcu in netlink dumps")
---
 net/netfilter/nf_tables_api.c | 20 +++++++++++---------
 1 file changed, 11 insertions(+), 9 deletions(-)
net/netfilter/nf_tables_api.c +11 -9
index aa5e7b00a581..0969f9647927 100644
@@ -2261,13 +2261,13 @@ static int nf_tables_fill_rule_info(struct sk_buff *skb, struct net *net,
 				    u32 flags, int family,
 				    const struct nft_table *table,
 				    const struct nft_chain *chain,
-				    const struct nft_rule *rule)
+				    const struct nft_rule *rule,
+				    const struct nft_rule *prule)
 {
 	struct nlmsghdr *nlh;
 	struct nfgenmsg *nfmsg;
 	const struct nft_expr *expr, *next;
 	struct nlattr *list;
-	const struct nft_rule *prule;
 	u16 type = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, event);
 
 	nlh = nlmsg_put(skb, portid, seq, type, sizeof(struct nfgenmsg), flags);
@@ -2287,8 +2287,7 @@ static int nf_tables_fill_rule_info(struct sk_buff *skb, struct net *net,
 			 NFTA_RULE_PAD))
 		goto nla_put_failure;
 
-	if ((event != NFT_MSG_DELRULE) && (rule->list.prev != &chain->rules)) {
-		prule = list_prev_entry(rule, list);
+	if (event != NFT_MSG_DELRULE && prule) {
 		if (nla_put_be64(skb, NFTA_RULE_POSITION,
 				 cpu_to_be64(prule->handle),
 				 NFTA_RULE_PAD))
@@ -2335,7 +2334,7 @@ static void nf_tables_rule_notify(const struct nft_ctx *ctx,
 
 	err = nf_tables_fill_rule_info(skb, ctx->net, ctx->portid, ctx->seq,
 				       event, 0, ctx->family, ctx->table,
-				       ctx->chain, rule);
+				       ctx->chain, rule, NULL);
 	if (err < 0) {
 		kfree_skb(skb);
 		goto err;
@@ -2360,12 +2359,13 @@ static int __nf_tables_dump_rules(struct sk_buff *skb,
 				  const struct nft_chain *chain)
 {
 	struct net *net = sock_net(skb->sk);
+	const struct nft_rule *rule, *prule;
 	unsigned int s_idx = cb->args[0];
-	const struct nft_rule *rule;
 
+	prule = NULL;
 	list_for_each_entry_rcu(rule, &chain->rules, list) {
 		if (!nft_is_active(net, rule))
-			goto cont;
+			goto cont_skip;
 		if (*idx < s_idx)
 			goto cont;
 		if (*idx > s_idx) {
@@ -2377,11 +2377,13 @@ static int __nf_tables_dump_rules(struct sk_buff *skb,
 					NFT_MSG_NEWRULE,
 					NLM_F_MULTI | NLM_F_APPEND,
 					table->family,
-					table, chain, rule) < 0)
+					table, chain, rule, prule) < 0)
 			return 1;
 
 		nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 cont:
+		prule = rule;
+cont_skip:
 		(*idx)++;
 	}
 	return 0;
@@ -2537,7 +2539,7 @@ static int nf_tables_getrule(struct net *net, struct sock *nlsk,
 
 	err = nf_tables_fill_rule_info(skb2, net, NETLINK_CB(skb).portid,
 				       nlh->nlmsg_seq, NFT_MSG_NEWRULE, 0,
-				       family, table, chain, rule);
+				       family, table, chain, rule, NULL);
 	if (err < 0)
 		goto err;
 
Signature · 1 line
2.21.0
#1 lore raw
Quoted 7 lines · Florian Westphal

We can oops in nf_tables_fill_rule_info().

Its not possible to fetch previous element in rcu-protected lists when deletions are not prevented somehow: list_del_rcu poisons the ->prev pointer value.

Quoted 6 lines

Before rcu-conversion this was safe as dump operations did hold nfnetlink mutex.

Pass previous rule as argument, obtained by keeping a pointer to the previous rule during traversal.

Hi Florian,
I have tested this patch and I think this patch works well. I would like to share my test method.

SHELL#1
while :
do
nft flush ruleset
nft -f test.nft
done

SHELL#2
while :
do
nft list ruleset -a
iptables-nft -vnL
done

This test method could make panic.

[ 1887.092089] general protection fault: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI
[ 1887.100127] CPU: 0 PID: 1652 Comm: nft Not tainted 5.1.0-rc5+ #76
[ 1887.106982] Hardware name: To be filled by O.E.M. To be filled by
O.E.M./Aptio CRB, BIOS 5.6.5 07/08/2015
[ 1887.117769] RIP: 0010:nf_tables_fill_rule_info.isra.59+0x362/0x860
[nf_tables]
[ 1887.125892] Code: 8b 84 24 d8 00 00 00 4d 8b 65 08 48 83 c0 10 49
39 c4 74 5a 49 8d 7c 24 10 48 b8 00 00 00 00 00 fc ff df 48 89 fa 48
c1 ea 03 <80> 3c 02 00 0f 85 9c 04 00 00 48 b8 ff ff ff ff ff 03 00 00
49 23
[ 1887.146996] RSP: 0018:ffff88810c0872a0 EFLAGS: 00010206
[ 1887.152878] RAX: dffffc0000000000 RBX: ffff88810ea0b1c0 RCX: 0000000000000000
[ 1887.160894] RDX: 1bd5a00000000042 RSI: ffff88810c0872f8 RDI: dead000000000210
[ 1887.168909] RBP: ffff8880b4b78000 R08: ffffed101696f008 R09: ffffed101696f008
[ 1887.176933] R10: 0000000000000002 R11: ffffed101696f007 R12: dead000000000200
[ 1887.184957] R13: ffff8880b8142688 R14: ffff8880b8142698 R15: ffff88810c0872f0
[ 1887.192973] FS:  00007fb475d8a740(0000) GS:ffff888116600000(0000)
knlGS:0000000000000000
[ 1887.202064] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 1887.208527] CR2: 0000557d208c4088 CR3: 00000001102de000 CR4: 00000000001006f0
[ 1887.216548] Call Trace:
[ 1887.219344]  ? nft_expr_dump+0x4c0/0x4c0 [nf_tables]
[ 1887.224947]  ? debug_show_all_locks+0x2d0/0x2d0
[ 1887.230055]  ? rcu_read_lock_sched_held+0x114/0x130
[ 1887.235573]  __nf_tables_dump_rules+0x27d/0x620 [nf_tables]
[ 1887.241892]  nf_tables_dump_rules+0x4c4/0xc80 [nf_tables]
[ 1887.247967]  ? rcu_read_lock_sched_held+0x114/0x130
[ 1887.253466]  ? __kmalloc_node_track_caller+0x321/0x360
[ 1887.259285]  ? __nf_tables_dump_rules+0x620/0x620 [nf_tables]
[ 1887.265759]  ? __alloc_skb+0x314/0x500
[ 1887.269990]  ? sock_spd_release+0xf0/0xf0
[ 1887.274514]  ? check_flags.part.41+0x440/0x440
[ 1887.279532]  netlink_dump+0x476/0xea0
[ 1887.283669]  ? __netlink_sendskb+0xc0/0xc0
[ 1887.288283]  ? __netlink_dump_start+0x165/0x750
[ 1887.293408]  __netlink_dump_start+0x537/0x750
[ 1887.298345]  nft_netlink_dump_start_rcu.constprop.71+0x97/0x180 [nf_tables]
[ 1887.306205]  nf_tables_getrule+0x3b6/0x620 [nf_tables]
[ 1887.312019]  ? nf_tables_rule_notify+0x3e0/0x3e0 [nf_tables]
[ 1887.318407]  ? nf_tables_dump_obj_start+0x230/0x230 [nf_tables]
[ 1887.325086]  ? __nf_tables_dump_rules+0x620/0x620 [nf_tables]
[ 1887.331569]  ? nf_tables_dump_sets_done+0x40/0x40 [nf_tables]
[ 1887.338045]  ? __nla_parse+0x34/0x310
[ 1887.342198]  ? nf_tables_rule_notify+0x3e0/0x3e0 [nf_tables]
[ 1887.348576]  nfnetlink_rcv_msg+0x3f0/0xd0b [nfnetlink]
[ 1887.354366]  ? kernel_text_address+0x111/0x120
[ 1887.359391]  ? nfnetlink_bind+0x200/0x200 [nfnetlink]
[ 1887.365097]  ? sched_clock_cpu+0x18/0x170
[ 1887.369616]  ? __sys_sendto+0x263/0x300
[ 1887.373929]  ? sched_clock_cpu+0x18/0x170
[ 1887.378447]  ? do_syscall_64+0x9c/0x3e0
[ 1887.382779]  ? __lock_acquire+0xe7c/0x3bd0
[ 1887.387403]  ? check_flags.part.41+0x440/0x440
[ 1887.392416]  netlink_rcv_skb+0x112/0x330
[ 1887.396841]  ? nfnetlink_bind+0x200/0x200 [nfnetlink]
[ 1887.402538]  ? netlink_ack+0x940/0x940
[ 1887.406776]  ? ns_capable_common+0x5c/0xd0
[ 1887.411405]  nfnetlink_rcv+0x161/0x320 [nfnetlink]
[ 1887.416809]  ? nfnetlink_rcv_batch+0x1280/0x1280 [nfnetlink]
[ 1887.423200]  netlink_unicast+0x3ee/0x5a0

After this patch, I couldn't see this panic message.

Thanks!

Quoted 88 lines
Fixes d9adf22a291883 ("netfilter: nf_tables: use call_rcu in netlink dumps")
---
 net/netfilter/nf_tables_api.c | 20 +++++++++++---------
 1 file changed, 11 insertions(+), 9 deletions(-)
net/netfilter/nf_tables_api.c +11 -10
index aa5e7b00a581..0969f9647927 100644
@@ -2261,13 +2261,13 @@ static int nf_tables_fill_rule_info(struct sk_buff *skb, struct net *net,
                                    u32 flags, int family,
                                    const struct nft_table *table,
                                    const struct nft_chain *chain,
-                                   const struct nft_rule *rule)
+                                   const struct nft_rule *rule,
+                                   const struct nft_rule *prule)
 {
        struct nlmsghdr *nlh;
        struct nfgenmsg *nfmsg;
        const struct nft_expr *expr, *next;
        struct nlattr *list;
-       const struct nft_rule *prule;
        u16 type = nfnl_msg_type(NFNL_SUBSYS_NFTABLES, event);
 
        nlh = nlmsg_put(skb, portid, seq, type, sizeof(struct nfgenmsg), flags);
@@ -2287,8 +2287,7 @@ static int nf_tables_fill_rule_info(struct sk_buff *skb, struct net *net,
                         NFTA_RULE_PAD))
                goto nla_put_failure;
 
-       if ((event != NFT_MSG_DELRULE) && (rule->list.prev != &chain->rules)) {
-               prule = list_prev_entry(rule, list);
+       if (event != NFT_MSG_DELRULE && prule) {
                if (nla_put_be64(skb, NFTA_RULE_POSITION,
                                 cpu_to_be64(prule->handle),
                                 NFTA_RULE_PAD))
@@ -2335,7 +2334,7 @@ static void nf_tables_rule_notify(const struct nft_ctx *ctx,
 
        err = nf_tables_fill_rule_info(skb, ctx->net, ctx->portid, ctx->seq,
                                       event, 0, ctx->family, ctx->table,
-                                      ctx->chain, rule);
+                                      ctx->chain, rule, NULL);
        if (err < 0) {
                kfree_skb(skb);
                goto err;
@@ -2360,12 +2359,13 @@ static int __nf_tables_dump_rules(struct sk_buff *skb,
                                  const struct nft_chain *chain)
 {
        struct net *net = sock_net(skb->sk);
+       const struct nft_rule *rule, *prule;
        unsigned int s_idx = cb->args[0];
-       const struct nft_rule *rule;
 
+       prule = NULL;
        list_for_each_entry_rcu(rule, &chain->rules, list) {
                if (!nft_is_active(net, rule))
-                       goto cont;
+                       goto cont_skip;
                if (*idx < s_idx)
                        goto cont;
                if (*idx > s_idx) {
@@ -2377,11 +2377,13 @@ static int __nf_tables_dump_rules(struct sk_buff *skb,
                                        NFT_MSG_NEWRULE,
                                        NLM_F_MULTI | NLM_F_APPEND,
                                        table->family,
-                                       table, chain, rule) < 0)
+                                       table, chain, rule, prule) < 0)
                        return 1;
 
                nl_dump_check_consistent(cb, nlmsg_hdr(skb));
 cont:
+               prule = rule;
+cont_skip:
                (*idx)++;
        }
        return 0;
@@ -2537,7 +2539,7 @@ static int nf_tables_getrule(struct net *net, struct sock *nlsk,
 
        err = nf_tables_fill_rule_info(skb2, net, NETLINK_CB(skb).portid,
                                       nlh->nlmsg_seq, NFT_MSG_NEWRULE, 0,
-                                      family, table, chain, rule);
+                                      family, table, chain, rule, NULL);
        if (err < 0)
                goto err;
 
--

2.21.0

maintainer Applied
#2 lore raw
In reply to Florian Westphal
Quoted 11 lines · Florian Westphal

We can oops in nf_tables_fill_rule_info().

Its not possible to fetch previous element in rcu-protected lists when deletions are not prevented somehow: list_del_rcu poisons the ->prev pointer value.

Before rcu-conversion this was safe as dump operations did hold nfnetlink mutex.

Pass previous rule as argument, obtained by keeping a pointer to the previous rule during traversal.

Applied, thanks.