net/sched: act_nat: validate the control action inside init()

the following script:

 # tc qdisc add dev crash0 clsact
 # tc filter add dev crash0 egress matchall \
 > action nat ingress 1.18.1.1 1.18.2.2 pass index 90
 # tc actions replace action nat \
 > ingress 1.18.1.1 1.18.2.2 goto chain 42 index 90 cookie c1a0c1a0
 # tc actions show action nat

had the following output:

 Error: Failed to init TC action chain.
 We have an error talking to the kernel
 total acts 1

         action order 0:  nat ingress 1.18.1.1/32 1.18.2.2 goto chain 42
          index 90 ref 2 bind 1
         cookie c1a0c1a0

Then, the first packet transmitted by crash0 made the kernel crash:

 BUG: unable to handle kernel NULL pointer dereference at 0000000000000000
 #PF error: [normal kernel read fault]
 PGD 800000002d180067 P4D 800000002d180067 PUD 7cb8b067 PMD 0
 Oops: 0000 [#1] SMP PTI
 CPU: 3 PID: 164 Comm: kworker/3:1 Not tainted 5.0.0-rc4.gotochain_crash+ #533
 Hardware name: Red Hat KVM, BIOS 0.5.1 01/01/2011
 Workqueue: ipv6_addrconf addrconf_dad_work
 RIP: 0010:tcf_action_exec+0xb8/0x100
 Code: 00 00 00 20 74 1d 83 f8 03 75 09 49 83 c4 08 4d 39 ec 75 bc 48 83 c4 10 5b 5d 41 5c 41 5d 41 5e 41 5f c3 49 8b 97 a8 00 00 00 <48> 8b 12 48 89 55 00 48 83 c4 10 5b 5d 41 5c 41 5d 41 5e 41 5f c3
 RSP: 0018:ffffae4500e2fad0 EFLAGS: 00010246
 RAX: 000000002000002a RBX: ffff9fa52e28c800 RCX: 0000000001011201
 RDX: 0000000000000000 RSI: 0000000000000056 RDI: ffff9fa52ca12800
 RBP: ffffae4500e2fb70 R08: 0000000000000022 R09: 000000000000000e
 R10: 00000000ffffffff R11: 0000000001011201 R12: ffff9fa52cbc9c00
 R13: ffff9fa52cbc9c08 R14: 0000000000000001 R15: ffff9fa52ca12780
 FS:  0000000000000000(0000) GS:ffff9fa57db80000(0000) knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: 0000000000000000 CR3: 0000000073f8c004 CR4: 00000000001606e0
 Call Trace:
  tcf_classify+0x58/0x120
  __dev_queue_xmit+0x40a/0x890
  ? ndisc_next_option+0x50/0x50
  ? ___neigh_create+0x4d5/0x680
  ? ip6_finish_output2+0x1b5/0x590
  ip6_finish_output2+0x1b5/0x590
  ? ip6_output+0x68/0x110
  ip6_output+0x68/0x110
  ? nf_hook.constprop.28+0x79/0xc0
  ndisc_send_skb+0x248/0x2e0
  ndisc_send_ns+0xf8/0x200
  ? addrconf_dad_work+0x389/0x4b0
  addrconf_dad_work+0x389/0x4b0
  ? __switch_to_asm+0x34/0x70
  ? process_one_work+0x195/0x380
  ? addrconf_dad_completed+0x370/0x370
  process_one_work+0x195/0x380
  worker_thread+0x30/0x390
  ? process_one_work+0x380/0x380
  kthread+0x113/0x130
  ? kthread_park+0x90/0x90
  ret_from_fork+0x35/0x40
 Modules linked in: act_nat veth ip6table_filter ip6_tables iptable_filter binfmt_misc ext4 crct10dif_pclmul crc32_pclmul ghash_clmulni_intel mbcache jbd2 snd_hda_codec_generic snd_hda_intel snd_hda_codec snd_hwdep snd_hda_core snd_seq snd_seq_device snd_pcm aesni_intel crypto_simd cryptd glue_helper snd_timer snd joydev virtio_balloon pcspkr soundcore i2c_piix4 nfsd auth_rpcgss nfs_acl lockd grace sunrpc ip_tables xfs qxl ata_generic pata_acpi drm_kms_helper syscopyarea sysfillrect sysimgblt fb_sys_fops ttm virtio_net virtio_blk net_failover failover virtio_console drm crc32c_intel floppy ata_piix libata virtio_pci virtio_ring virtio serio_raw dm_mirror dm_region_hash dm_log dm_mod
 CR2: 0000000000000000

Validating the control action within tcf_nat_init() proved to fix the
above issue. A TDC selftest is added to verify the correct behavior.

Fixes: db50514f ("net: sched: add termination action to allow goto chain")
Fixes: 97763dc0 ("net_sched: reject unknown tcfa_action values")
Signed-off-by: Davide Caratti <dcaratti@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/sched/act_nat.c

@@ -21,6 +21,7 @@
 #include <linux/string.h>
 #include <linux/tc_act/tc_nat.h>
 #include <net/act_api.h>
+#include <net/pkt_cls.h>
 #include <net/icmp.h>
 #include <net/ip.h>
 #include <net/netlink.h>
@@ -43,6 +44,7 @@ static int tcf_nat_init(struct net *net, struct nlattr *nla, struct nlattr *est,
 {
 	struct tc_action_net *tn = net_generic(net, nat_net_id);
 	struct nlattr *tb[TCA_NAT_MAX + 1];
+	struct tcf_chain *goto_ch = NULL;
 	struct tc_nat *parm;
 	int ret = 0, err;
 	struct tcf_nat *p;
@@ -77,6 +79,9 @@ static int tcf_nat_init(struct net *net, struct nlattr *nla, struct nlattr *est,
 	} else {
 		return err;
 	}
+	err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
+	if (err < 0)
+		goto release_idr;
 	p = to_tcf_nat(*a);
 
 	spin_lock_bh(&p->tcf_lock);
@@ -85,13 +90,18 @@ static int tcf_nat_init(struct net *net, struct nlattr *nla, struct nlattr *est,
 	p->mask = parm->mask;
 	p->flags = parm->flags;
 
-	p->tcf_action = parm->action;
+	goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
 	spin_unlock_bh(&p->tcf_lock);
+	if (goto_ch)
+		tcf_chain_put_by_act(goto_ch);
 
 	if (ret == ACT_P_CREATED)
 		tcf_idr_insert(tn, *a);
 
 	return ret;
+release_idr:
+	tcf_idr_release(*a, bind);
+	return err;
 }
 
 static int tcf_nat_act(struct sk_buff *skb, const struct tc_action *a,

tools/testing/selftests/tc-testing/tc-tests/actions/nat.json

@@ -589,5 +589,30 @@
         "teardown": [
             "$TC actions flush action nat"
         ]
+    },
+    {
+        "id": "4b12",
+        "name": "Replace nat action with invalid goto chain control",
+        "category": [
+            "actions",
+            "nat"
+        ],
+        "setup": [
+            [
+                "$TC actions flush action nat",
+                0,
+                1,
+                255
+            ],
+            "$TC actions add action nat ingress 1.18.1.1 1.18.2.2 drop index 90"
+        ],
+        "cmdUnderTest": "$TC actions replace action nat ingress 1.18.1.1 1.18.2.2 goto chain 42 index 90 cookie c1a0c1a0",
+        "expExitCode": "255",
+        "verifyCmd": "$TC actions get action nat index 90",
+        "matchPattern": "action order [0-9]+:  nat ingress 1.18.1.1/32 1.18.2.2 drop.*index 90 ref",
+        "matchCount": "1",
+        "teardown": [
+            "$TC actions flush action nat"
+        ]
     }
 ]