Skip to contentSkip to conversation

ipv6: ndisc: fix NULL deref in accept_untracked_na()

PATCH Applied Weiming Shi 4 participants 7 messages Jun 17 – Jun 21

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ipv6: ndisc: fix NULL deref in accept_untracked_na()

accept_untracked_na() re-fetches the inet6_dev with __in6_dev_get(dev) and dereferences idev->cnf.accept_untracked_na without a NULL check, even though its only caller ndisc_recv_na() already fetched and NULL-checked idev for the same device.

Both reads of dev->ip6_ptr run in the same RCU read-side critical section, but a concurrent addrconf_ifdown() can clear dev->ip6_ptr between them: lowering the MTU below IPV6_MIN_MTU calls addrconf_ifdown() without the synchronize_net() that orders the unregister path, so the re-fetch returns NULL and oopses:

 BUG: KASAN: null-ptr-deref in ndisc_recv_na (net/ipv6/ndisc.c:974)
 Read of size 4 at addr 0000000000000364
 Call Trace:
  <IRQ>
  ndisc_recv_na (net/ipv6/ndisc.c:974)
  icmpv6_rcv (net/ipv6/icmp.c:1193)
  ip6_protocol_deliver_rcu (net/ipv6/ip6_input.c:479)
  ip6_input_finish (net/ipv6/ip6_input.c:534)
  ip6_input (net/ipv6/ip6_input.c:545)
  ip6_mc_input (net/ipv6/ip6_input.c:635)
  ipv6_rcv (net/ipv6/ip6_input.c:351)
  </IRQ>

It is reachable by an unprivileged user via a network namespace.

Pass the caller's already validated idev instead of re-fetching it; the idev stays alive for the whole RCU critical section, so it is safe even after dev->ip6_ptr has been cleared.

Fixes aaa5f515b16b ("net: ipv6: new accept_untracked_na option to accept na only if in-network")
Assisted-by Claude:claude-opus-4-8
---
 net/ipv6/ndisc.c | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)
net/ipv6/ndisc.c +3 -5
index e7ad13c5bd267..f867ec8d3d905 100644
@@ -967,10 +967,8 @@ static enum skb_drop_reason ndisc_recv_ns(struct sk_buff *skb)
 	return reason;
 }
 
-static int accept_untracked_na(struct net_device *dev, struct in6_addr *saddr)
+static int accept_untracked_na(struct inet6_dev *idev, struct in6_addr *saddr)
 {
-	struct inet6_dev *idev = __in6_dev_get(dev);
-
 	switch (READ_ONCE(idev->cnf.accept_untracked_na)) {
 	case 0: /* Don't accept untracked na (absent in neighbor cache) */
 		return 0;
@@ -980,7 +978,7 @@ static int accept_untracked_na(struct net_device *dev, struct in6_addr *saddr)
 		 * same subnet as an address configured on the interface that
 		 * received the na
 		 */
-		return !!ipv6_chk_prefix(saddr, dev);
+		return !!ipv6_chk_prefix(saddr, idev->dev);
 	default:
 		return 0;
 	}
@@ -1078,7 +1076,7 @@ static enum skb_drop_reason ndisc_recv_na(struct sk_buff *skb)
 	 */
 	new_state = msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE;
 	if (!neigh && lladdr && idev && READ_ONCE(idev->cnf.forwarding)) {
-		if (accept_untracked_na(dev, saddr)) {
+		if (accept_untracked_na(idev, saddr)) {
 			neigh = neigh_create(&nd_tbl, &msg->target, dev);
 			new_state = NUD_STALE;
 		}
Signature · 1 line
2.43.0
#1 lore raw
Quoted 2 lines · Weiming Shi

accept_untracked_na() re-fetches the inet6_dev with __in6_dev_get(dev) and dereferences idev->cnf.accept_untracked_na without a NULL check,

Does ipv6_rpl_srh_rcv have same problem?

#2 lore raw
Quoted 7 lines · Jiayuan Chen
Quoted 2 lines · Weiming Shi

accept_untracked_na() re-fetches the inet6_dev with __in6_dev_get(dev) and dereferences idev->cnf.accept_untracked_na without a NULL check,

Does ipv6_rpl_srh_rcv have same problem?

Hi,

Yes, ipv6_rpl_srh_rcv() has the same missing check. It reads idev->cnf.rpl_seg_enabled right after __in6_dev_get(skb->dev) with no NULL check, while seg6 and ioam6 in the same file both check it.

But I tried to trigger it and couldn't. With a guard added as an instrument, idev never came back NULL over tens of millions of RPL packets while flapping the MTU, so I can't say it's actually reachable.

Still, it's the only one of the three without the check. Want me to send a patch adding it there too, for consistency?

Thanks,
Weiming Shi

#3 lore raw
Quoted 8 lines · Weiming Shi
Quoted 5 lines · Jiayuan Chen
Quoted 2 lines · Weiming Shi

accept_untracked_na() re-fetches the inet6_dev with __in6_dev_get(dev) and dereferences idev->cnf.accept_untracked_na without a NULL check,

Does ipv6_rpl_srh_rcv have same problem?

Hi,

Quoted 7 lines

Yes, ipv6_rpl_srh_rcv() has the same missing check. It reads idev->cnf.rpl_seg_enabled right after __in6_dev_get(skb->dev) with no NULL check, while seg6 and ioam6 in the same file both check it.

But I tried to trigger it and couldn't. With a guard added as an instrument, idev never came back NULL over tens of millions of RPL packets while flapping the MTU, so I can't say it's actually reachable.

Can you need to add mdelay to enlarge the race window to reproduce it?

I believe we need more precise traffic and timing control, instead of aggressively ramping up traffic and load in an attempt to reproduce the issue.

#4 lore raw
Quoted 15 lines · Jiayuan Chen
Quoted 12 lines · Weiming Shi
Quoted 5 lines · Jiayuan Chen
Quoted 2 lines · Weiming Shi

accept_untracked_na() re-fetches the inet6_dev with __in6_dev_get(dev) and dereferences idev->cnf.accept_untracked_na without a NULL check,

Does ipv6_rpl_srh_rcv have same problem?

Hi,

Yes, ipv6_rpl_srh_rcv() has the same missing check. It reads idev->cnf.rpl_seg_enabled right after __in6_dev_get(skb->dev) with no NULL check, while seg6 and ioam6 in the same file both check it.

Quoted 7 lines
Quoted 3 lines

But I tried to trigger it and couldn't. With a guard added as an instrument, idev never came back NULL over tens of millions of RPL packets while flapping the MTU, so I can't say it's actually reachable.

Can you need to add mdelay to enlarge the race window to reproduce it?

The delay works; we insert one at

@@ -1078,6 +1079,7 @@ static enum skb_drop_reason ndisc_recv_na(struct sk_buff *skb)

         */
        new_state = msg->icmph.icmp6_solicited ? NUD_REACHABLE : NUD_STALE;
        if (!neigh && lladdr && idev && READ_ONCE(idev->cnf.forwarding)) {
+               mdelay(100); /* AI-POC: widen race window vs
concurrent addrconf_ifdown() NULLing dev->ip6_ptr */
                if (accept_untracked_na(dev, saddr)) {
                        neigh = neigh_create(&nd_tbl, &msg->target, dev);
                        new_state = NUD_STALE;

And our poc triggers:

```
[    0.887292] Oops: general protection fault, probably for
non-canonical address 0xdffffc0000000068: 0000 [#1] SMP KASAN NOPTI
[    0.887791] KASAN: null-ptr-deref in range
[0x0000000000000340-0x0000000000000347]
[    0.888101] CPU: 1 UID: 0 PID: 146 Comm: exploit Not tainted 7.1.0+
#9 PREEMPTLAZY
[    0.888419] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996),
BIOS rel-1.17.0-0-gb52ca86e094d-prebuilt.qemu.org 04/01/4
[    0.888880] RIP: 0010:ndisc_recv_na+0x67b/0xf30
[    0.889072] Code: 03 80 3c 02 00 0f 85 5f 07 00 00 49 8b af c0 00
00 00 48 b8 00 00 00 00 00 fc ff df 48 8d bd 40 03 00 00 e
[    0.889829] RSP: 0018:ffff8881029084f8 EFLAGS: 00010202
[    0.890043] RAX: dffffc0000000000 RBX: 1ffff110205210a8 RCX: 0000000000000001
[    0.890340] RDX: 0000000000000068 RSI: 0000000000004d9a RDI: 0000000000000340
[    0.890633] RBP: 0000000000000000 R08: 0000000000000004 R09: ffff888102908570
[    0.890921] R10: ffff88800ee7f792 R11: ffff888102908580 R12: ffff88800ee7f780
[    0.891207] R13: ffff88800ee7f778 R14: ffff88800ee7f792 R15: ffff88800f2d6000
[    0.891511] FS:  00007b8516aec6c0(0000) GS:ffff888177d3d000(0000)
knlGS:0000000000000000
[    0.891853] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[    0.892094] CR2: 00007b8516aebb30 CR3: 000000000f164006 CR4: 0000000000772ef0
[    0.892402] PKRU: 55555554
[    0.892517] Call Trace:
[    0.892627]  <IRQ>
[    0.892718]  ? stack_trace_save+0x93/0xd0
[    0.892895]  ? __pfx_ndisc_recv_na+0x10/0x10
[    0.893087]  ? _raw_spin_unlock_irqrestore+0xe/0x40
[    0.893295]  ? stack_depot_save_flags+0x488/0x810
[    0.893506]  ? fib6_rule_lookup+0x40a/0x5a0
[    0.893689]  ? __napi_poll+0xa1/0x560
[    0.893852]  ? net_rx_action+0x401/0xd80
[    0.894025]  icmpv6_rcv+0x10a4/0x14e0
[    0.894181]  ip6_protocol_deliver_rcu+0xbd4/0x1340
[    0.894391]  ip6_input_finish+0x1d1/0x3a0
[    0.894559]  ip6_input+0x195/0x1d0
[    0.894709]  ? __pfx_ip6_input+0x10/0x10
[    0.894873]  ? __asan_memset+0x23/0x50
[    0.895034]  ip6_mc_input+0x26c/0x3a0
[    0.895188]  ipv6_rcv+0x31b/0x390
[    0.895340]  ? __pfx_ipv6_rcv+0x10/0x10
[    0.895501]  ? get_stack_info+0x2f/0x90
[    0.895665]  ? stack_access_ok+0xde/0x200
[    0.895833]  ? get_stack_info_noinstr+0x18/0x110
[    0.896036]  __netif_receive_skb_core.constprop.0+0xd78/0x2f00
[    0.896286]  ? common_startup_64+0x13e/0x151
[    0.896475]  ? common_startup_64+0x13e/0x151
[    0.896655]  ? __pfx_stack_trace_consume_entry+0x10/0x10
[    0.896875]  ? arch_stack_walk+0x99/0x100
[    0.897049]  ? __pfx___netif_receive_skb_core.constprop.0+0x10/0x10
[    0.897310]  ? common_startup_64+0x13e/0x151
[    0.897498]  ? stack_trace_save+0x93/0xd0
[    0.897668]  ? __pfx_stack_trace_save+0x10/0x10
[    0.897855]  ? stack_depot_save_flags+0x29/0x810
[    0.898046]  ? kasan_save_stack+0x33/0x60
[    0.898213]  ? kasan_save_track+0x14/0x30
[    0.898389]  ? kasan_save_free_info+0x3b/0x60
...
```

Weiming's patch works with a delay. The poc doesn't crash the kernel. I noticed that you mentioned ipv6_rpl_srh_rcv. We actually have sent one, but we later found that someone (Andrea Mayer) had sent the issue earlier, so we just let them fix it:

https://lore.kernel.org/all/[email protected]/

Xiang

Quoted 4 lines

I believe we need more precise traffic and timing control, instead of aggressively ramping up traffic and load in an attempt to reproduce the issue.

Reviewed-by on v1 1/1
#5 lore raw
In reply to Weiming Shi
Quoted 26 lines · Weiming Shi

accept_untracked_na() re-fetches the inet6_dev with __in6_dev_get(dev) and dereferences idev->cnf.accept_untracked_na without a NULL check, even though its only caller ndisc_recv_na() already fetched and NULL-checked idev for the same device.

Both reads of dev->ip6_ptr run in the same RCU read-side critical section, but a concurrent addrconf_ifdown() can clear dev->ip6_ptr between them: lowering the MTU below IPV6_MIN_MTU calls addrconf_ifdown() without the synchronize_net() that orders the unregister path, so the re-fetch returns NULL and oopses:

  BUG: KASAN: null-ptr-deref in ndisc_recv_na (net/ipv6/ndisc.c:974)
  Read of size 4 at addr 0000000000000364
  Call Trace:
   <IRQ>
   ndisc_recv_na (net/ipv6/ndisc.c:974)
   icmpv6_rcv (net/ipv6/icmp.c:1193)
   ip6_protocol_deliver_rcu (net/ipv6/ip6_input.c:479)
   ip6_input_finish (net/ipv6/ip6_input.c:534)
   ip6_input (net/ipv6/ip6_input.c:545)
   ip6_mc_input (net/ipv6/ip6_input.c:635)
   ipv6_rcv (net/ipv6/ip6_input.c:351)
   </IRQ>

It is reachable by an unprivileged user via a network namespace.

Quoted 8 lines

Pass the caller's already validated idev instead of re-fetching it; the idev stays alive for the whole RCU critical section, so it is safe even after dev->ip6_ptr has been cleared.

Fixes aaa5f515b16b ("net: ipv6: new accept_untracked_na option to accept na only if in-network")
Assisted-by Claude:claude-opus-4-8
#6 lore raw
In reply to Weiming Shi
This patch was applied to netdev/net.git (main) by Jakub Kicinski <[email protected]>:

Hello:

This patch was applied to netdev/net.git (main)
by Jakub Kicinski <[email protected]>:

Quoted 5 lines · On Wed, 17 Jun 2026 14:55:13 +0800 you wrote:

accept_untracked_na() re-fetches the inet6_dev with __in6_dev_get(dev) and dereferences idev->cnf.accept_untracked_na without a NULL check, even though its only caller ndisc_recv_na() already fetched and NULL-checked idev for the same device.

Quoted 7 lines

Both reads of dev->ip6_ptr run in the same RCU read-side critical section, but a concurrent addrconf_ifdown() can clear dev->ip6_ptr between them: lowering the MTU below IPV6_MIN_MTU calls addrconf_ifdown() without the synchronize_net() that orders the unregister path, so the re-fetch returns NULL and oopses:

[...]

Here is the summary with links:
- [net] ipv6: ndisc: fix NULL deref in accept_untracked_na()

    https://git.kernel.org/netdev/net/c/d186e942365a

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html