[PATCH 6.1.y] net: Fix icmp host relookup triggering ip_rt_bug

[PATCH 6.1.y] net: Fix icmp host relookup triggering ip_rt_bug

[Jiayuan Chen]

From: Dong Chenchen <dongchenchen2@huawei.com>

[ Upstream commit c44daa7e3c73229f7ac74985acb8c7fb909c4e0a ]

arp link failure may trigger ip_rt_bug while xfrm enabled, call trace is:

WARNING: CPU: 0 PID: 0 at net/ipv4/route.c:1241 ip_rt_bug+0x14/0x20
Modules linked in:
CPU: 0 UID: 0 PID: 0 Comm: swapper/0 Not tainted 6.12.0-rc6-00077-g2e1b3cc9d7f7
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996),
BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
RIP: 0010:ip_rt_bug+0x14/0x20
Call Trace:
 <IRQ>
 ip_send_skb+0x14/0x40
 __icmp_send+0x42d/0x6a0
 ipv4_link_failure+0xe2/0x1d0
 arp_error_report+0x3c/0x50
 neigh_invalidate+0x8d/0x100
 neigh_timer_handler+0x2e1/0x330
 call_timer_fn+0x21/0x120
 __run_timer_base.part.0+0x1c9/0x270
 run_timer_softirq+0x4c/0x80
 handle_softirqs+0xac/0x280
 irq_exit_rcu+0x62/0x80
 sysvec_apic_timer_interrupt+0x77/0x90

The script below reproduces this scenario:
ip xfrm policy add src 0.0.0.0/0 dst 0.0.0.0/0 \
	dir out priority 0 ptype main flag localok icmp
ip l a veth1 type veth
ip a a 192.168.141.111/24 dev veth0
ip l s veth0 up
ping 192.168.141.155 -c 1

icmp_route_lookup() create input routes for locally generated packets
while xfrm relookup ICMP traffic.Then it will set input route
(dst->out = ip_rt_bug) to skb for DESTUNREACH.

For ICMP err triggered by locally generated packets, dst->dev of output
route is loopback. Generally, xfrm relookup verification is not required
on loopback interfaces (net.ipv4.conf.lo.disable_xfrm = 1).

Skip icmp relookup for locally generated packets to fix it.

Fixes: 8b7817f3a959 ("[IPSEC]: Add ICMP host relookup support")
Signed-off-by: Dong Chenchen <dongchenchen2@huawei.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20241127040850.1513135-1-dongchenchen2@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Jiayuan Chen <jiayuan.chen@linux.dev>

---
failed backport
https://lore.kernel.org/stable/20250207161555-b1a8749027831a1a@stable.kernel.org/T/#m0c880c1f04f7211aea9b7f6b4de0b64aa1726417
---
 net/ipv4/icmp.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/net/ipv4/icmp.c b/net/ipv4/icmp.c
index d5d745c3e345..737e6caad716 100644
--- a/net/ipv4/icmp.c
+++ b/net/ipv4/icmp.c
@@ -509,6 +509,9 @@ static struct rtable *icmp_route_lookup(struct net *net,
 	if (!IS_ERR(rt)) {
 		if (rt != rt2)
 			return rt;
+		if (inet_addr_type_dev_table(net, route_lookup_dev,
+					     fl4->daddr) == RTN_LOCAL)
+			return rt;
 	} else if (PTR_ERR(rt) == -EPERM) {
 		rt = NULL;
 	} else
-- 
2.43.0

Re: [PATCH 6.1.y] net: Fix icmp host relookup triggering ip_rt_bug

[Salvatore Bonaccorso]

Hi,

On Wed, May 06, 2026 at 09:20:57AM +0800, Jiayuan Chen wrote:
> From: Dong Chenchen <dongchenchen2@huawei.com>
> 
> [ Upstream commit c44daa7e3c73229f7ac74985acb8c7fb909c4e0a ]
> 
> arp link failure may trigger ip_rt_bug while xfrm enabled, call trace is:
> 
> WARNING: CPU: 0 PID: 0 at net/ipv4/route.c:1241 ip_rt_bug+0x14/0x20
> Modules linked in:
> CPU: 0 UID: 0 PID: 0 Comm: swapper/0 Not tainted 6.12.0-rc6-00077-g2e1b3cc9d7f7
> Hardware name: QEMU Standard PC (i440FX + PIIX, 1996),
> BIOS rel-1.13.0-0-gf21b5a4aeb02-prebuilt.qemu.org 04/01/2014
> RIP: 0010:ip_rt_bug+0x14/0x20
> Call Trace:
>  <IRQ>
>  ip_send_skb+0x14/0x40
>  __icmp_send+0x42d/0x6a0
>  ipv4_link_failure+0xe2/0x1d0
>  arp_error_report+0x3c/0x50
>  neigh_invalidate+0x8d/0x100
>  neigh_timer_handler+0x2e1/0x330
>  call_timer_fn+0x21/0x120
>  __run_timer_base.part.0+0x1c9/0x270
>  run_timer_softirq+0x4c/0x80
>  handle_softirqs+0xac/0x280
>  irq_exit_rcu+0x62/0x80
>  sysvec_apic_timer_interrupt+0x77/0x90
> 
> The script below reproduces this scenario:
> ip xfrm policy add src 0.0.0.0/0 dst 0.0.0.0/0 \
> 	dir out priority 0 ptype main flag localok icmp
> ip l a veth1 type veth
> ip a a 192.168.141.111/24 dev veth0
> ip l s veth0 up
> ping 192.168.141.155 -c 1
> 
> icmp_route_lookup() create input routes for locally generated packets
> while xfrm relookup ICMP traffic.Then it will set input route
> (dst->out = ip_rt_bug) to skb for DESTUNREACH.
> 
> For ICMP err triggered by locally generated packets, dst->dev of output
> route is loopback. Generally, xfrm relookup verification is not required
> on loopback interfaces (net.ipv4.conf.lo.disable_xfrm = 1).
> 
> Skip icmp relookup for locally generated packets to fix it.
> 
> Fixes: 8b7817f3a959 ("[IPSEC]: Add ICMP host relookup support")
> Signed-off-by: Dong Chenchen <dongchenchen2@huawei.com>
> Reviewed-by: David Ahern <dsahern@kernel.org>
> Reviewed-by: Eric Dumazet <edumazet@google.com>
> Link: https://patch.msgid.link/20241127040850.1513135-1-dongchenchen2@huawei.com
> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
> Signed-off-by: Jiayuan Chen <jiayuan.chen@linux.dev>
> 
> ---
> failed backport
> https://lore.kernel.org/stable/20250207161555-b1a8749027831a1a@stable.kernel.org/T/#m0c880c1f04f7211aea9b7f6b4de0b64aa1726417
> ---
>  net/ipv4/icmp.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/net/ipv4/icmp.c b/net/ipv4/icmp.c
> index d5d745c3e345..737e6caad716 100644
> --- a/net/ipv4/icmp.c
> +++ b/net/ipv4/icmp.c
> @@ -509,6 +509,9 @@ static struct rtable *icmp_route_lookup(struct net *net,
>  	if (!IS_ERR(rt)) {
>  		if (rt != rt2)
>  			return rt;
> +		if (inet_addr_type_dev_table(net, route_lookup_dev,
> +					     fl4->daddr) == RTN_LOCAL)
> +			return rt;
>  	} else if (PTR_ERR(rt) == -EPERM) {
>  		rt = NULL;
>  	} else
> -- 
> 2.43.0

This fixes the problem reported in Debian as
https://bugs.debian.org/1135514

Tested-by: Salvatore Bonaccorso <carnil@debian.org>

Regards,
Salvatore

Re: [PATCH 6.1.y] net: Fix icmp host relookup triggering ip_rt_bug

[Sasha Levin]

> [PATCH 6.1.y] net: Fix icmp host relookup triggering ip_rt_bug

Queued for 6.1.y, thanks.

--
Sasha