Skip to contentSkip to conversation

udp_tunnel: GRO optimization follow-up

PATCH In review Paolo Abeni 4 participants 19 messages Mar 21, 2025 – Mar 25, 2025

Subsystems (3)
lore
Share & tools
lore

Paolo Abeni

author maintainer
#0 lore raw

v2 cover udp_tunnel: GRO optimization follow-up

Syzkaller and Nathan reported a few issues on the reference changeset. This series tries to address them all and additionally does some little cleanup.

See the individual patches for the details.
---
Should this prove to be too invasive, too late or ineffective, I'll be ok with a revert before the upcoming net-next PR.

Paolo Abeni (5):
udp_tunnel: properly deal with xfrm gro encap.
udp_tunnel: fix compile warning
udp_tunnel: fix UaF in GRO accounting
udp_tunnel: avoid inconsistent local variables usage
udp_tunnel: prevent GRO lookup optimization for user-space sockets

 include/net/udp_tunnel.h   | 4 ----
 net/ipv4/udp.c             | 5 +++++
 net/ipv4/udp_offload.c     | 9 ++++++---
 net/ipv4/udp_tunnel_core.c | 7 ++++---
 4 files changed, 15 insertions(+), 10 deletions(-)
Signature · 1 line
2.48.1
author maintainer
#1 lore raw

PATCH v2 1/5 udp_tunnel: properly deal with xfrm gro encap.

The blamed commit below does not take in account that xfrm can enable GRO over UDP encapsulation without going through setup_udp_tunnel_sock().

At deletion time such socket will still go through udp_tunnel_cleanup_gro(), and the failed GRO type lookup will trigger the reported warning.

Add the GRO accounting for XFRM tunnel when GRO is enabled, and adjust the known gro types accordingly.

Note that we can't use setup_udp_tunnel_sock() here, as the xfrm tunnel setup can be "incremental" - e.g. the encapsulation is created first and GRO is enabled later.

Also we can not allow GRO sk lookup optimization for XFRM tunnels, as the socket could match the selection criteria at enable time, and later on the user-space could disconnect/bind it breaking such criteria.

Reported-by [email protected]
Fixes 311b36574ceac ("udp_tunnel: use static call for GRO hooks when possible")

---
v1 -> v2:
- do proper account for xfrm, retain the warning

---
 net/ipv4/udp.c         | 5 +++++
 net/ipv4/udp_offload.c | 4 +++-
 2 files changed, 8 insertions(+), 1 deletion(-)
net/ipv4/udp.c +5 -0
index db606f7e41638..79efbf465fb04 100644
@@ -2903,10 +2903,15 @@ static void set_xfrm_gro_udp_encap_rcv(__u16 encap_type, unsigned short family,
 {
 #ifdef CONFIG_XFRM
 	if (udp_test_bit(GRO_ENABLED, sk) && encap_type == UDP_ENCAP_ESPINUDP) {
+		bool old_enabled = !!udp_sk(sk)->gro_receive;
+
 		if (family == AF_INET)
 			WRITE_ONCE(udp_sk(sk)->gro_receive, xfrm4_gro_udp_encap_rcv);
 		else if (IS_ENABLED(CONFIG_IPV6) && family == AF_INET6)
 			WRITE_ONCE(udp_sk(sk)->gro_receive, ipv6_stub->xfrm6_gro_udp_encap_rcv);
+
+		if (!old_enabled && udp_sk(sk)->gro_receive)
+			udp_tunnel_update_gro_rcv(sk, true);
 	}
 #endif
 }
net/ipv4/udp_offload.c +3 -1
index 088aa8cb8ac0c..02365b818f1af 100644
@@ -37,9 +37,11 @@ struct udp_tunnel_type_entry {
 	refcount_t count;
 };
 
+/* vxlan, fou and xfrm have 2 different gro_receive hooks each */
 #define UDP_MAX_TUNNEL_TYPES (IS_ENABLED(CONFIG_GENEVE) + \
 			      IS_ENABLED(CONFIG_VXLAN) * 2 + \
-			      IS_ENABLED(CONFIG_NET_FOU) * 2)
+			      IS_ENABLED(CONFIG_NET_FOU) * 2 + \
+			      IS_ENABLED(CONFIG_XFRM) * 2)
 
 DEFINE_STATIC_CALL(udp_tunnel_gro_rcv, dummy_gro_rcv);
 static DEFINE_STATIC_KEY_FALSE(udp_tunnel_static_call);
Signature · 1 line
2.48.1
maintainer
#2 lore raw

Re: [PATCH net-next v2 1/5] udp_tunnel: properly deal with xfrm gro encap.

Quoted 39 lines · Paolo Abeni

The blamed commit below does not take in account that xfrm can enable GRO over UDP encapsulation without going through setup_udp_tunnel_sock().

At deletion time such socket will still go through udp_tunnel_cleanup_gro(), and the failed GRO type lookup will trigger the reported warning.

Add the GRO accounting for XFRM tunnel when GRO is enabled, and adjust the known gro types accordingly.

Note that we can't use setup_udp_tunnel_sock() here, as the xfrm tunnel setup can be "incremental" - e.g. the encapsulation is created first and GRO is enabled later.

Also we can not allow GRO sk lookup optimization for XFRM tunnels, as the socket could match the selection criteria at enable time, and later on the user-space could disconnect/bind it breaking such criteria.

Reported-by [email protected]
Fixes 311b36574ceac ("udp_tunnel: use static call for GRO hooks when possible")

---
v1 -> v2:
- do proper account for xfrm, retain the warning

---
 net/ipv4/udp.c         | 5 +++++
 net/ipv4/udp_offload.c | 4 +++-
 2 files changed, 8 insertions(+), 1 deletion(-)
net/ipv4/udp.c +0 -0
index db606f7e41638..79efbf465fb04 100644
@@ -2903,10 +2903,15 @@ static void set_xfrm_gro_udp_encap_rcv(__u16 encap_type, unsigned short family,
 {
 #ifdef CONFIG_XFRM

net/ipv4/udp.c static void set_xfrm_gro_udp_encap_rcv(__u16 encap_type, unsigned short family, · line 2914

Quoted 10 lines
net/ipv4/udp.c +5 -0
 	if (udp_test_bit(GRO_ENABLED, sk) && encap_type == UDP_ENCAP_ESPINUDP) {
+		bool old_enabled = !!udp_sk(sk)->gro_receive;
+
 		if (family == AF_INET)
 			WRITE_ONCE(udp_sk(sk)->gro_receive, xfrm4_gro_udp_encap_rcv);
 		else if (IS_ENABLED(CONFIG_IPV6) && family == AF_INET6)
 			WRITE_ONCE(udp_sk(sk)->gro_receive, ipv6_stub->xfrm6_gro_udp_encap_rcv);
+
+		if (!old_enabled && udp_sk(sk)->gro_receive)
+			udp_tunnel_update_gro_rcv(sk, true);

The second part of the condition is always true right?

author maintainer
#3 lore raw

Re: [PATCH net-next v2 1/5] udp_tunnel: properly deal with xfrm gro encap.

Hi,

Quoted 44 lines · Willem de Bruijn
Quoted 43 lines · Paolo Abeni

The blamed commit below does not take in account that xfrm can enable GRO over UDP encapsulation without going through setup_udp_tunnel_sock().

At deletion time such socket will still go through udp_tunnel_cleanup_gro(), and the failed GRO type lookup will trigger the reported warning.

Add the GRO accounting for XFRM tunnel when GRO is enabled, and adjust the known gro types accordingly.

Note that we can't use setup_udp_tunnel_sock() here, as the xfrm tunnel setup can be "incremental" - e.g. the encapsulation is created first and GRO is enabled later.

Also we can not allow GRO sk lookup optimization for XFRM tunnels, as the socket could match the selection criteria at enable time, and later on the user-space could disconnect/bind it breaking such criteria.

Reported-by [email protected]
Fixes 311b36574ceac ("udp_tunnel: use static call for GRO hooks when possible")

---
v1 -> v2:
- do proper account for xfrm, retain the warning

---
 net/ipv4/udp.c         | 5 +++++
 net/ipv4/udp_offload.c | 4 +++-
 2 files changed, 8 insertions(+), 1 deletion(-)
net/ipv4/udp.c +2 -0
index db606f7e41638..79efbf465fb04 100644
@@ -2903,10 +2903,15 @@ static void set_xfrm_gro_udp_encap_rcv(__u16 encap_type, unsigned short family,
 {
 #ifdef CONFIG_XFRM
 	if (udp_test_bit(GRO_ENABLED, sk) && encap_type == UDP_ENCAP_ESPINUDP) {
+		bool old_enabled = !!udp_sk(sk)->gro_receive;
+
 		if (family == AF_INET)
Quoted 8 lines

net/ipv4/udp.c static void set_xfrm_gro_udp_encap_rcv(__u16 encap_type, unsigned short family, · line 2914

Quoted 6 lines
net/ipv4/udp.c +3 -0
 			WRITE_ONCE(udp_sk(sk)->gro_receive, xfrm4_gro_udp_encap_rcv);
 		else if (IS_ENABLED(CONFIG_IPV6) && family == AF_INET6)
 			WRITE_ONCE(udp_sk(sk)->gro_receive, ipv6_stub->xfrm6_gro_udp_encap_rcv);
+
+		if (!old_enabled && udp_sk(sk)->gro_receive)
+			udp_tunnel_update_gro_rcv(sk, true);

The second part of the condition is always true right?

Jakub noted my initial reply did not land on the ML, sorry.

Yes, AFAICS the second part of the condition should be always true, or at least I fail to see how to reach there otherwise.

Still syzkaller is too good to prove me wrong, I guess it's good to err on the pedantic/safe side here - checking that condition explicitly. It's also IMHO more readable.

Cheers,

Paolo

#4 lore raw

Re: [PATCH net-next v2 1/5] udp_tunnel: properly deal with xfrm gro encap.

In reply to Paolo Abeni
Quoted 39 lines · Paolo Abeni

The blamed commit below does not take in account that xfrm can enable GRO over UDP encapsulation without going through setup_udp_tunnel_sock().

At deletion time such socket will still go through udp_tunnel_cleanup_gro(), and the failed GRO type lookup will trigger the reported warning.

Add the GRO accounting for XFRM tunnel when GRO is enabled, and adjust the known gro types accordingly.

Note that we can't use setup_udp_tunnel_sock() here, as the xfrm tunnel setup can be "incremental" - e.g. the encapsulation is created first and GRO is enabled later.

Also we can not allow GRO sk lookup optimization for XFRM tunnels, as the socket could match the selection criteria at enable time, and later on the user-space could disconnect/bind it breaking such criteria.

Reported-by [email protected]
Fixes 311b36574ceac ("udp_tunnel: use static call for GRO hooks when possible")

---
v1 -> v2:
- do proper account for xfrm, retain the warning

---
 net/ipv4/udp.c         | 5 +++++
 net/ipv4/udp_offload.c | 4 +++-
 2 files changed, 8 insertions(+), 1 deletion(-)
net/ipv4/udp.c +0 -0
index db606f7e41638..79efbf465fb04 100644
@@ -2903,10 +2903,15 @@ static void set_xfrm_gro_udp_encap_rcv(__u16 encap_type, unsigned short family,
 {
 #ifdef CONFIG_XFRM

net/ipv4/udp.c static void set_xfrm_gro_udp_encap_rcv(__u16 encap_type, unsigned short family, · line 2914

Quoted 10 lines
net/ipv4/udp.c +5 -0
 	if (udp_test_bit(GRO_ENABLED, sk) && encap_type == UDP_ENCAP_ESPINUDP) {
+		bool old_enabled = !!udp_sk(sk)->gro_receive;
+
 		if (family == AF_INET)
 			WRITE_ONCE(udp_sk(sk)->gro_receive, xfrm4_gro_udp_encap_rcv);
 		else if (IS_ENABLED(CONFIG_IPV6) && family == AF_INET6)
 			WRITE_ONCE(udp_sk(sk)->gro_receive, ipv6_stub->xfrm6_gro_udp_encap_rcv);
+
+		if (!old_enabled && udp_sk(sk)->gro_receive)
+			udp_tunnel_update_gro_rcv(sk, true);

We're not under any lock at this point, so this is a bit racy. I think we'll "only" end up leaking a ref on cur->count if this happens. Not ideal, but much better than the current situation.

Quoted 24 lines
net/ipv4/udp.c +0 -0
 	}
 #endif
 }
net/ipv4/udp_offload.c +3 -1
index 088aa8cb8ac0c..02365b818f1af 100644
@@ -37,9 +37,11 @@ struct udp_tunnel_type_entry {
 	refcount_t count;
 };
 
+/* vxlan, fou and xfrm have 2 different gro_receive hooks each */
 #define UDP_MAX_TUNNEL_TYPES (IS_ENABLED(CONFIG_GENEVE) + \
 			      IS_ENABLED(CONFIG_VXLAN) * 2 + \
-			      IS_ENABLED(CONFIG_NET_FOU) * 2)
+			      IS_ENABLED(CONFIG_NET_FOU) * 2 + \
+			      IS_ENABLED(CONFIG_XFRM) * 2)
 
 DEFINE_STATIC_CALL(udp_tunnel_gro_rcv, dummy_gro_rcv);
 static DEFINE_STATIC_KEY_FALSE(udp_tunnel_static_call);
Signature · 1 line
2.48.1
Signature · 1 line
Sabrina
author maintainer
#5 lore raw

PATCH v2 2/5 udp_tunnel: fix compile warning

In reply to Paolo Abeni

Nathan reported that the compiler is not happy to use a zero size udp_tunnel_gro_types array:

  net/ipv4/udp_offload.c:130:8: warning: array index 0 is past the end of the array (that has type 'struct udp_tunnel_type_entry[0]') [-Warray-bounds]
    130 |                                    udp_tunnel_gro_types[0].gro_receive);
        |                                    ^                    ~
  include/linux/static_call.h:154:42: note: expanded from macro 'static_call_update'
    154 |         typeof(&STATIC_CALL_TRAMP(name)) __F = (func);                  \
        |                                                 ^~~~
  net/ipv4/udp_offload.c:47:1: note: array 'udp_tunnel_gro_types' declared here
     47 | static struct udp_tunnel_type_entry udp_tunnel_gro_types[UDP_MAX_TUNNEL_TYPES];
        | ^
  1 warning generated.

In such (unusual) configuration we should skip entirely the static call optimization accounting.

Reported-by [email protected]
Fixes 311b36574ceac ("udp_tunnel: use static call for GRO hooks when possible")
---
 net/ipv4/udp_offload.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/ipv4/udp_offload.c +1 -1
index 02365b818f1af..fd2b8e3830beb 100644
@@ -83,7 +83,7 @@ void udp_tunnel_update_gro_rcv(struct sock *sk, bool add)
 	struct udp_sock *up = udp_sk(sk);
 	int i, old_gro_type_nr;
 
-	if (!up->gro_receive)
+	if (!UDP_MAX_TUNNEL_TYPES || !up->gro_receive)
 		return;
 
 	mutex_lock(&udp_tunnel_gro_type_lock);
Signature · 1 line
2.48.1
maintainer
#6 lore raw

Re: [PATCH net-next v2 2/5] udp_tunnel: fix compile warning

Quoted 14 lines · Paolo Abeni

Nathan reported that the compiler is not happy to use a zero size udp_tunnel_gro_types array:

  net/ipv4/udp_offload.c:130:8: warning: array index 0 is past the end of the array (that has type 'struct udp_tunnel_type_entry[0]') [-Warray-bounds]
    130 |                                    udp_tunnel_gro_types[0].gro_receive);
        |                                    ^                    ~
  include/linux/static_call.h:154:42: note: expanded from macro 'static_call_update'
    154 |         typeof(&STATIC_CALL_TRAMP(name)) __F = (func);                  \
        |                                                 ^~~~
  net/ipv4/udp_offload.c:47:1: note: array 'udp_tunnel_gro_types' declared here
     47 | static struct udp_tunnel_type_entry udp_tunnel_gro_types[UDP_MAX_TUNNEL_TYPES];
        | ^
  1 warning generated.
Quoted 7 lines

In such (unusual) configuration we should skip entirely the static call optimization accounting.

Reported-by [email protected]
Fixes 311b36574ceac ("udp_tunnel: use static call for GRO hooks when possible")

Should CONFIG_NET_UDP_TUNNEL just not be user selectable and only enabled by implementations of UDP tunnels like vxlan and geneve?

Quoted 9 lines
---
 net/ipv4/udp_offload.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/ipv4/udp_offload.c +0 -0
index 02365b818f1af..fd2b8e3830beb 100644

net/ipv4/udp_offload.c void udp_tunnel_update_gro_rcv(struct sock *sk, bool add) · line 87

Quoted 6 lines
net/ipv4/udp_offload.c +1 -1
@@ -83,7 +83,7 @@ void udp_tunnel_update_gro_rcv(struct sock *sk, bool add)
 	struct udp_sock *up = udp_sk(sk);
 	int i, old_gro_type_nr;
 
-	if (!up->gro_receive)
+	if (!UDP_MAX_TUNNEL_TYPES || !up->gro_receive)
 		return;

If that is too risky, I suppose this workaround is sufficient. But having a zero length array seems a bit odd.

#7 lore raw

Re: [PATCH net-next v2 2/5] udp_tunnel: fix compile warning

Quoted 18 lines · Willem de Bruijn
Quoted 17 lines · Paolo Abeni

Nathan reported that the compiler is not happy to use a zero size udp_tunnel_gro_types array:

  net/ipv4/udp_offload.c:130:8: warning: array index 0 is past the end of the array (that has type 'struct udp_tunnel_type_entry[0]') [-Warray-bounds]
    130 |                                    udp_tunnel_gro_types[0].gro_receive);
        |                                    ^                    ~
  include/linux/static_call.h:154:42: note: expanded from macro 'static_call_update'
    154 |         typeof(&STATIC_CALL_TRAMP(name)) __F = (func);                  \
        |                                                 ^~~~
  net/ipv4/udp_offload.c:47:1: note: array 'udp_tunnel_gro_types' declared here
     47 | static struct udp_tunnel_type_entry udp_tunnel_gro_types[UDP_MAX_TUNNEL_TYPES];
        | ^
  1 warning generated.

In such (unusual) configuration we should skip entirely the static call optimization accounting.

Quoted 7 lines
Quoted 4 lines
Reported-by [email protected]
Fixes 311b36574ceac ("udp_tunnel: use static call for GRO hooks when possible")

Should CONFIG_NET_UDP_TUNNEL just not be user selectable and only enabled by implementations of UDP tunnels like vxlan and geneve?

It's already not user selectable?

config NET_UDP_TUNNEL
tristate

(no string after bool/tristate, so no manual config)

But there are tunnels that don't do GRO, so they're not counted in UDP_MAX_TUNNEL_TYPES, and if only those types are enabled, we'll have CONFIG_NET_UDP_TUNNEL=y with UDP_MAX_TUNNEL_TYPES == 0.

Quoted 12 lines
Quoted 12 lines
---
 net/ipv4/udp_offload.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/ipv4/udp_offload.c +0 -0
index 02365b818f1af..fd2b8e3830beb 100644
@@ -83,7 +83,7 @@ void udp_tunnel_update_gro_rcv(struct sock *sk, bool add)
 	struct udp_sock *up = udp_sk(sk);
 	int i, old_gro_type_nr;
 
Quoted 6 lines

net/ipv4/udp_offload.c void udp_tunnel_update_gro_rcv(struct sock *sk, bool add) · line 87

Quoted 3 lines
net/ipv4/udp_offload.c +1 -1
-	if (!up->gro_receive)
+	if (!UDP_MAX_TUNNEL_TYPES || !up->gro_receive)
 		return;

If that is too risky, I suppose this workaround is sufficient. But having a zero length array seems a bit odd.

I think the alternative would be to add

    #if UDP_MAX_TUNNEL_TYPES > 0

around some of this code.

Signature · 1 line
Sabrina
Reviewed-by on v2 2/5
#8 lore raw

Re: [PATCH net-next v2 2/5] udp_tunnel: fix compile warning

In reply to Paolo Abeni
Quoted 14 lines · Paolo Abeni

Nathan reported that the compiler is not happy to use a zero size udp_tunnel_gro_types array:

  net/ipv4/udp_offload.c:130:8: warning: array index 0 is past the end of the array (that has type 'struct udp_tunnel_type_entry[0]') [-Warray-bounds]
    130 |                                    udp_tunnel_gro_types[0].gro_receive);
        |                                    ^                    ~
  include/linux/static_call.h:154:42: note: expanded from macro 'static_call_update'
    154 |         typeof(&STATIC_CALL_TRAMP(name)) __F = (func);                  \
        |                                                 ^~~~
  net/ipv4/udp_offload.c:47:1: note: array 'udp_tunnel_gro_types' declared here
     47 | static struct udp_tunnel_type_entry udp_tunnel_gro_types[UDP_MAX_TUNNEL_TYPES];
        | ^
  1 warning generated.
Quoted 7 lines

In such (unusual) configuration we should skip entirely the static call optimization accounting.

Reported-by [email protected]
Fixes 311b36574ceac ("udp_tunnel: use static call for GRO hooks when possible")
Signature · 1 line
Sabrina
author maintainer
#9 lore raw

PATCH v2 3/5 udp_tunnel: fix UaF in GRO accounting

In reply to Paolo Abeni

Siyzkaller reported a race in UDP tunnel GRO accounting, leading to UaF:

BUG: KASAN: slab-use-after-free in udp_tunnel_update_gro_lookup+0x23c/0x2c0 net/ipv4/udp_offload.c:65
Read of size 8 at addr ffff88801235ebe8 by task syz.2.655/7921

CPU: 1 UID: 0 PID: 7921 Comm: syz.2.655 Not tainted 6.14.0-rc6-syzkaller-01313-g23c9ff659140 #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 02/12/2025
Call Trace:
 <TASK>
 __dump_stack lib/dump_stack.c:94 [inline]
 dump_stack_lvl+0x241/0x360 lib/dump_stack.c:120
 print_address_description mm/kasan/report.c:408 [inline]
 print_report+0x16e/0x5b0 mm/kasan/report.c:521
 kasan_report+0x143/0x180 mm/kasan/report.c:634
 udp_tunnel_update_gro_lookup+0x23c/0x2c0 net/ipv4/udp_offload.c:65
 sk_common_release+0x71/0x2e0 net/core/sock.c:3896
 inet_release+0x17d/0x200 net/ipv4/af_inet.c:435
 __sock_release net/socket.c:647 [inline]
 sock_release+0x82/0x150 net/socket.c:675
 sock_free drivers/net/wireguard/socket.c:339 [inline]
 wg_socket_reinit+0x215/0x380 drivers/net/wireguard/socket.c:435
 wg_stop+0x59f/0x600 drivers/net/wireguard/device.c:133
 __dev_close_many+0x3a6/0x700 net/core/dev.c:1717
 dev_close_many+0x24e/0x4c0 net/core/dev.c:1742
 unregister_netdevice_many_notify+0x629/0x24f0 net/core/dev.c:11923
 rtnl_delete_link net/core/rtnetlink.c:3512 [inline]
 rtnl_dellink+0x526/0x8c0 net/core/rtnetlink.c:3554
 rtnetlink_rcv_msg+0x791/0xcf0 net/core/rtnetlink.c:6945
 netlink_rcv_skb+0x206/0x480 net/netlink/af_netlink.c:2534
 netlink_unicast_kernel net/netlink/af_netlink.c:1313 [inline]
 netlink_unicast+0x7f6/0x990 net/netlink/af_netlink.c:1339
 netlink_sendmsg+0x8de/0xcb0 net/netlink/af_netlink.c:1883
 sock_sendmsg_nosec net/socket.c:709 [inline]
 __sock_sendmsg+0x221/0x270 net/socket.c:724
 ____sys_sendmsg+0x53a/0x860 net/socket.c:2564
 ___sys_sendmsg net/socket.c:2618 [inline]
 __sys_sendmsg+0x269/0x350 net/socket.c:2650
 do_syscall_x64 arch/x86/entry/common.c:52 [inline]
 do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83
 entry_SYSCALL_64_after_hwframe+0x77/0x7f
RIP: 0033:0x7f35ab38d169
Code: ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 a8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007f35ac28f038 EFLAGS: 00000246 ORIG_RAX: 000000000000002e
RAX: ffffffffffffffda RBX: 00007f35ab5a6160 RCX: 00007f35ab38d169
RDX: 0000000000000000 RSI: 0000400000000000 RDI: 0000000000000004
RBP: 00007f35ab40e2a0 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 0000000000000001 R14: 00007f35ab5a6160 R15: 00007ffdddd781b8
 </TASK>

Allocated by task 7770:
 kasan_save_stack mm/kasan/common.c:47 [inline]
 kasan_save_track+0x3f/0x80 mm/kasan/common.c:68
 unpoison_slab_object mm/kasan/common.c:319 [inline]
 __kasan_slab_alloc+0x66/0x80 mm/kasan/common.c:345
 kasan_slab_alloc include/linux/kasan.h:250 [inline]
 slab_post_alloc_hook mm/slub.c:4115 [inline]
 slab_alloc_node mm/slub.c:4164 [inline]
 kmem_cache_alloc_noprof+0x1d9/0x380 mm/slub.c:4171
 sk_prot_alloc+0x58/0x210 net/core/sock.c:2190
 sk_alloc+0x3e/0x370 net/core/sock.c:2249
 inet_create+0x648/0xea0 net/ipv4/af_inet.c:326
 __sock_create+0x4c0/0xa30 net/socket.c:1539
 sock_create net/socket.c:1597 [inline]
 __sys_socket_create net/socket.c:1634 [inline]
 __sys_socket+0x150/0x3c0 net/socket.c:1681
 __do_sys_socket net/socket.c:1695 [inline]
 __se_sys_socket net/socket.c:1693 [inline]
 __x64_sys_socket+0x7a/0x90 net/socket.c:1693
 do_syscall_x64 arch/x86/entry/common.c:52 [inline]
 do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83
 entry_SYSCALL_64_after_hwframe+0x77/0x7f

Freed by task 7768:
 kasan_save_stack mm/kasan/common.c:47 [inline]
 kasan_save_track+0x3f/0x80 mm/kasan/common.c:68
 kasan_save_free_info+0x40/0x50 mm/kasan/generic.c:576
 poison_slab_object mm/kasan/common.c:247 [inline]
 __kasan_slab_free+0x59/0x70 mm/kasan/common.c:264
 kasan_slab_free include/linux/kasan.h:233 [inline]
 slab_free_hook mm/slub.c:2353 [inline]
 slab_free mm/slub.c:4609 [inline]
 kmem_cache_free+0x195/0x410 mm/slub.c:4711
 sk_prot_free net/core/sock.c:2230 [inline]
 __sk_destruct+0x4fd/0x690 net/core/sock.c:2327
 inet_release+0x17d/0x200 net/ipv4/af_inet.c:435
 __sock_release net/socket.c:647 [inline]
 sock_close+0xbc/0x240 net/socket.c:1389
 __fput+0x3e9/0x9f0 fs/file_table.c:464
 task_work_run+0x24f/0x310 kernel/task_work.c:227
 resume_user_mode_work include/linux/resume_user_mode.h:50 [inline]
 exit_to_user_mode_loop kernel/entry/common.c:114 [inline]
 exit_to_user_mode_prepare include/linux/entry-common.h:329 [inline]
 __syscall_exit_to_user_mode_work kernel/entry/common.c:207 [inline]
 syscall_exit_to_user_mode+0x13f/0x340 kernel/entry/common.c:218
 do_syscall_64+0x100/0x230 arch/x86/entry/common.c:89
 entry_SYSCALL_64_after_hwframe+0x77/0x7f

The buggy address belongs to the object at ffff88801235e4c0
 which belongs to the cache UDP of size 1856

The buggy address is located 1832 bytes inside of
freed 1856-byte region [ffff88801235e4c0, ffff88801235ec00)

At disposal time, to avoid unconditionally acquiring a spin lock, UDP tunnel sockets are conditionally removed from the known tunnels list only if the socket is actually present in such a list.

Such check happens outside the socket lock scope: the current CPU could observe an uninitialized list entry even if the tunnel has been actually registered by a different core.

Address the issue moving the blamed check under the relevant list spin lock.

Reported-by [email protected]
Fixes 8d4880db37835 ("udp_tunnel: create a fastpath GRO lookup.")
---
 include/net/udp_tunnel.h | 4 ----
 net/ipv4/udp_offload.c   | 3 ++-
 2 files changed, 2 insertions(+), 5 deletions(-)
include/net/udp_tunnel.h +0 -4
index a7b230867eb14..13b54e6856414 100644
@@ -214,14 +214,10 @@ static inline void udp_tunnel_update_gro_rcv(struct sock *sk, bool add) {}
 
 static inline void udp_tunnel_cleanup_gro(struct sock *sk)
 {
-	struct udp_sock *up = udp_sk(sk);
 	struct net *net = sock_net(sk);
 
 	udp_tunnel_update_gro_rcv(sk, false);
 
-	if (!up->tunnel_list.pprev)
-		return;
-
 	udp_tunnel_update_gro_lookup(net, sk, false);
 }
 
net/ipv4/udp_offload.c +2 -1
index fd2b8e3830beb..b124355a36aee 100644
@@ -57,10 +57,11 @@ void udp_tunnel_update_gro_lookup(struct net *net, struct sock *sk, bool add)
 	struct udp_tunnel_gro *udp_tunnel_gro;
 
 	spin_lock(&udp_tunnel_gro_lock);
+
 	udp_tunnel_gro = &net->ipv4.udp_tunnel_gro[is_ipv6];
 	if (add)
 		hlist_add_head(&up->tunnel_list, &udp_tunnel_gro->list);
-	else
+	else if (up->tunnel_list.pprev)
 		hlist_del_init(&up->tunnel_list);
 
 	if (udp_tunnel_gro->list.first &&
Signature · 1 line
2.48.1
maintainer Reviewed-by on v2 3/5
#10 lore raw

Re: [PATCH net-next v2 3/5] udp_tunnel: fix UaF in GRO accounting

Quoted 112 lines · Paolo Abeni

Siyzkaller reported a race in UDP tunnel GRO accounting, leading to UaF:

BUG: KASAN: slab-use-after-free in udp_tunnel_update_gro_lookup+0x23c/0x2c0 net/ipv4/udp_offload.c:65
Read of size 8 at addr ffff88801235ebe8 by task syz.2.655/7921

CPU: 1 UID: 0 PID: 7921 Comm: syz.2.655 Not tainted 6.14.0-rc6-syzkaller-01313-g23c9ff659140 #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 02/12/2025
Call Trace:
 <TASK>
 __dump_stack lib/dump_stack.c:94 [inline]
 dump_stack_lvl+0x241/0x360 lib/dump_stack.c:120
 print_address_description mm/kasan/report.c:408 [inline]
 print_report+0x16e/0x5b0 mm/kasan/report.c:521
 kasan_report+0x143/0x180 mm/kasan/report.c:634
 udp_tunnel_update_gro_lookup+0x23c/0x2c0 net/ipv4/udp_offload.c:65
 sk_common_release+0x71/0x2e0 net/core/sock.c:3896
 inet_release+0x17d/0x200 net/ipv4/af_inet.c:435
 __sock_release net/socket.c:647 [inline]
 sock_release+0x82/0x150 net/socket.c:675
 sock_free drivers/net/wireguard/socket.c:339 [inline]
 wg_socket_reinit+0x215/0x380 drivers/net/wireguard/socket.c:435
 wg_stop+0x59f/0x600 drivers/net/wireguard/device.c:133
 __dev_close_many+0x3a6/0x700 net/core/dev.c:1717
 dev_close_many+0x24e/0x4c0 net/core/dev.c:1742
 unregister_netdevice_many_notify+0x629/0x24f0 net/core/dev.c:11923
 rtnl_delete_link net/core/rtnetlink.c:3512 [inline]
 rtnl_dellink+0x526/0x8c0 net/core/rtnetlink.c:3554
 rtnetlink_rcv_msg+0x791/0xcf0 net/core/rtnetlink.c:6945
 netlink_rcv_skb+0x206/0x480 net/netlink/af_netlink.c:2534
 netlink_unicast_kernel net/netlink/af_netlink.c:1313 [inline]
 netlink_unicast+0x7f6/0x990 net/netlink/af_netlink.c:1339
 netlink_sendmsg+0x8de/0xcb0 net/netlink/af_netlink.c:1883
 sock_sendmsg_nosec net/socket.c:709 [inline]
 __sock_sendmsg+0x221/0x270 net/socket.c:724
 ____sys_sendmsg+0x53a/0x860 net/socket.c:2564
 ___sys_sendmsg net/socket.c:2618 [inline]
 __sys_sendmsg+0x269/0x350 net/socket.c:2650
 do_syscall_x64 arch/x86/entry/common.c:52 [inline]
 do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83
 entry_SYSCALL_64_after_hwframe+0x77/0x7f
RIP: 0033:0x7f35ab38d169
Code: ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 a8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007f35ac28f038 EFLAGS: 00000246 ORIG_RAX: 000000000000002e
RAX: ffffffffffffffda RBX: 00007f35ab5a6160 RCX: 00007f35ab38d169
RDX: 0000000000000000 RSI: 0000400000000000 RDI: 0000000000000004
RBP: 00007f35ab40e2a0 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 0000000000000001 R14: 00007f35ab5a6160 R15: 00007ffdddd781b8
 </TASK>

Allocated by task 7770:
 kasan_save_stack mm/kasan/common.c:47 [inline]
 kasan_save_track+0x3f/0x80 mm/kasan/common.c:68
 unpoison_slab_object mm/kasan/common.c:319 [inline]
 __kasan_slab_alloc+0x66/0x80 mm/kasan/common.c:345
 kasan_slab_alloc include/linux/kasan.h:250 [inline]
 slab_post_alloc_hook mm/slub.c:4115 [inline]
 slab_alloc_node mm/slub.c:4164 [inline]
 kmem_cache_alloc_noprof+0x1d9/0x380 mm/slub.c:4171
 sk_prot_alloc+0x58/0x210 net/core/sock.c:2190
 sk_alloc+0x3e/0x370 net/core/sock.c:2249
 inet_create+0x648/0xea0 net/ipv4/af_inet.c:326
 __sock_create+0x4c0/0xa30 net/socket.c:1539
 sock_create net/socket.c:1597 [inline]
 __sys_socket_create net/socket.c:1634 [inline]
 __sys_socket+0x150/0x3c0 net/socket.c:1681
 __do_sys_socket net/socket.c:1695 [inline]
 __se_sys_socket net/socket.c:1693 [inline]
 __x64_sys_socket+0x7a/0x90 net/socket.c:1693
 do_syscall_x64 arch/x86/entry/common.c:52 [inline]
 do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83
 entry_SYSCALL_64_after_hwframe+0x77/0x7f

Freed by task 7768:
 kasan_save_stack mm/kasan/common.c:47 [inline]
 kasan_save_track+0x3f/0x80 mm/kasan/common.c:68
 kasan_save_free_info+0x40/0x50 mm/kasan/generic.c:576
 poison_slab_object mm/kasan/common.c:247 [inline]
 __kasan_slab_free+0x59/0x70 mm/kasan/common.c:264
 kasan_slab_free include/linux/kasan.h:233 [inline]
 slab_free_hook mm/slub.c:2353 [inline]
 slab_free mm/slub.c:4609 [inline]
 kmem_cache_free+0x195/0x410 mm/slub.c:4711
 sk_prot_free net/core/sock.c:2230 [inline]
 __sk_destruct+0x4fd/0x690 net/core/sock.c:2327
 inet_release+0x17d/0x200 net/ipv4/af_inet.c:435
 __sock_release net/socket.c:647 [inline]
 sock_close+0xbc/0x240 net/socket.c:1389
 __fput+0x3e9/0x9f0 fs/file_table.c:464
 task_work_run+0x24f/0x310 kernel/task_work.c:227
 resume_user_mode_work include/linux/resume_user_mode.h:50 [inline]
 exit_to_user_mode_loop kernel/entry/common.c:114 [inline]
 exit_to_user_mode_prepare include/linux/entry-common.h:329 [inline]
 __syscall_exit_to_user_mode_work kernel/entry/common.c:207 [inline]
 syscall_exit_to_user_mode+0x13f/0x340 kernel/entry/common.c:218
 do_syscall_64+0x100/0x230 arch/x86/entry/common.c:89
 entry_SYSCALL_64_after_hwframe+0x77/0x7f

The buggy address belongs to the object at ffff88801235e4c0
 which belongs to the cache UDP of size 1856

The buggy address is located 1832 bytes inside of
freed 1856-byte region [ffff88801235e4c0, ffff88801235ec00)

At disposal time, to avoid unconditionally acquiring a spin lock, UDP tunnel sockets are conditionally removed from the known tunnels list only if the socket is actually present in such a list.

Such check happens outside the socket lock scope: the current CPU could observe an uninitialized list entry even if the tunnel has been actually registered by a different core.

Quoted 7 lines

Address the issue moving the blamed check under the relevant list spin lock.

Reported-by [email protected]
Fixes 8d4880db37835 ("udp_tunnel: create a fastpath GRO lookup.")
Reviewed-by on v2 3/5
#11 lore raw

Re: [PATCH net-next v2 3/5] udp_tunnel: fix UaF in GRO accounting

In reply to Paolo Abeni
Quoted 112 lines · Paolo Abeni

Siyzkaller reported a race in UDP tunnel GRO accounting, leading to UaF:

BUG: KASAN: slab-use-after-free in udp_tunnel_update_gro_lookup+0x23c/0x2c0 net/ipv4/udp_offload.c:65
Read of size 8 at addr ffff88801235ebe8 by task syz.2.655/7921

CPU: 1 UID: 0 PID: 7921 Comm: syz.2.655 Not tainted 6.14.0-rc6-syzkaller-01313-g23c9ff659140 #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 02/12/2025
Call Trace:
 <TASK>
 __dump_stack lib/dump_stack.c:94 [inline]
 dump_stack_lvl+0x241/0x360 lib/dump_stack.c:120
 print_address_description mm/kasan/report.c:408 [inline]
 print_report+0x16e/0x5b0 mm/kasan/report.c:521
 kasan_report+0x143/0x180 mm/kasan/report.c:634
 udp_tunnel_update_gro_lookup+0x23c/0x2c0 net/ipv4/udp_offload.c:65
 sk_common_release+0x71/0x2e0 net/core/sock.c:3896
 inet_release+0x17d/0x200 net/ipv4/af_inet.c:435
 __sock_release net/socket.c:647 [inline]
 sock_release+0x82/0x150 net/socket.c:675
 sock_free drivers/net/wireguard/socket.c:339 [inline]
 wg_socket_reinit+0x215/0x380 drivers/net/wireguard/socket.c:435
 wg_stop+0x59f/0x600 drivers/net/wireguard/device.c:133
 __dev_close_many+0x3a6/0x700 net/core/dev.c:1717
 dev_close_many+0x24e/0x4c0 net/core/dev.c:1742
 unregister_netdevice_many_notify+0x629/0x24f0 net/core/dev.c:11923
 rtnl_delete_link net/core/rtnetlink.c:3512 [inline]
 rtnl_dellink+0x526/0x8c0 net/core/rtnetlink.c:3554
 rtnetlink_rcv_msg+0x791/0xcf0 net/core/rtnetlink.c:6945
 netlink_rcv_skb+0x206/0x480 net/netlink/af_netlink.c:2534
 netlink_unicast_kernel net/netlink/af_netlink.c:1313 [inline]
 netlink_unicast+0x7f6/0x990 net/netlink/af_netlink.c:1339
 netlink_sendmsg+0x8de/0xcb0 net/netlink/af_netlink.c:1883
 sock_sendmsg_nosec net/socket.c:709 [inline]
 __sock_sendmsg+0x221/0x270 net/socket.c:724
 ____sys_sendmsg+0x53a/0x860 net/socket.c:2564
 ___sys_sendmsg net/socket.c:2618 [inline]
 __sys_sendmsg+0x269/0x350 net/socket.c:2650
 do_syscall_x64 arch/x86/entry/common.c:52 [inline]
 do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83
 entry_SYSCALL_64_after_hwframe+0x77/0x7f
RIP: 0033:0x7f35ab38d169
Code: ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 a8 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007f35ac28f038 EFLAGS: 00000246 ORIG_RAX: 000000000000002e
RAX: ffffffffffffffda RBX: 00007f35ab5a6160 RCX: 00007f35ab38d169
RDX: 0000000000000000 RSI: 0000400000000000 RDI: 0000000000000004
RBP: 00007f35ab40e2a0 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 0000000000000001 R14: 00007f35ab5a6160 R15: 00007ffdddd781b8
 </TASK>

Allocated by task 7770:
 kasan_save_stack mm/kasan/common.c:47 [inline]
 kasan_save_track+0x3f/0x80 mm/kasan/common.c:68
 unpoison_slab_object mm/kasan/common.c:319 [inline]
 __kasan_slab_alloc+0x66/0x80 mm/kasan/common.c:345
 kasan_slab_alloc include/linux/kasan.h:250 [inline]
 slab_post_alloc_hook mm/slub.c:4115 [inline]
 slab_alloc_node mm/slub.c:4164 [inline]
 kmem_cache_alloc_noprof+0x1d9/0x380 mm/slub.c:4171
 sk_prot_alloc+0x58/0x210 net/core/sock.c:2190
 sk_alloc+0x3e/0x370 net/core/sock.c:2249
 inet_create+0x648/0xea0 net/ipv4/af_inet.c:326
 __sock_create+0x4c0/0xa30 net/socket.c:1539
 sock_create net/socket.c:1597 [inline]
 __sys_socket_create net/socket.c:1634 [inline]
 __sys_socket+0x150/0x3c0 net/socket.c:1681
 __do_sys_socket net/socket.c:1695 [inline]
 __se_sys_socket net/socket.c:1693 [inline]
 __x64_sys_socket+0x7a/0x90 net/socket.c:1693
 do_syscall_x64 arch/x86/entry/common.c:52 [inline]
 do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83
 entry_SYSCALL_64_after_hwframe+0x77/0x7f

Freed by task 7768:
 kasan_save_stack mm/kasan/common.c:47 [inline]
 kasan_save_track+0x3f/0x80 mm/kasan/common.c:68
 kasan_save_free_info+0x40/0x50 mm/kasan/generic.c:576
 poison_slab_object mm/kasan/common.c:247 [inline]
 __kasan_slab_free+0x59/0x70 mm/kasan/common.c:264
 kasan_slab_free include/linux/kasan.h:233 [inline]
 slab_free_hook mm/slub.c:2353 [inline]
 slab_free mm/slub.c:4609 [inline]
 kmem_cache_free+0x195/0x410 mm/slub.c:4711
 sk_prot_free net/core/sock.c:2230 [inline]
 __sk_destruct+0x4fd/0x690 net/core/sock.c:2327
 inet_release+0x17d/0x200 net/ipv4/af_inet.c:435
 __sock_release net/socket.c:647 [inline]
 sock_close+0xbc/0x240 net/socket.c:1389
 __fput+0x3e9/0x9f0 fs/file_table.c:464
 task_work_run+0x24f/0x310 kernel/task_work.c:227
 resume_user_mode_work include/linux/resume_user_mode.h:50 [inline]
 exit_to_user_mode_loop kernel/entry/common.c:114 [inline]
 exit_to_user_mode_prepare include/linux/entry-common.h:329 [inline]
 __syscall_exit_to_user_mode_work kernel/entry/common.c:207 [inline]
 syscall_exit_to_user_mode+0x13f/0x340 kernel/entry/common.c:218
 do_syscall_64+0x100/0x230 arch/x86/entry/common.c:89
 entry_SYSCALL_64_after_hwframe+0x77/0x7f

The buggy address belongs to the object at ffff88801235e4c0
 which belongs to the cache UDP of size 1856

The buggy address is located 1832 bytes inside of
freed 1856-byte region [ffff88801235e4c0, ffff88801235ec00)

At disposal time, to avoid unconditionally acquiring a spin lock, UDP tunnel sockets are conditionally removed from the known tunnels list only if the socket is actually present in such a list.

Such check happens outside the socket lock scope: the current CPU could observe an uninitialized list entry even if the tunnel has been actually registered by a different core.

Quoted 7 lines

Address the issue moving the blamed check under the relevant list spin lock.

Reported-by [email protected]
Fixes 8d4880db37835 ("udp_tunnel: create a fastpath GRO lookup.")
Signature · 1 line
Sabrina
author maintainer
#12 lore raw

PATCH v2 4/5 udp_tunnel: avoid inconsistent local variables usage

In reply to Paolo Abeni

In setup_udp_tunnel_sock() 'sk' and 'sock->sk' are alias. The code I introduced there uses alternatively both variables, for no good reasons.

Stick to 'sk' usage, to be consistent with the prior code.

---
 net/ipv4/udp_tunnel_core.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
net/ipv4/udp_tunnel_core.c +3 -3
index c49fceea83139..aa9016619c25a 100644
@@ -90,10 +90,10 @@ void setup_udp_tunnel_sock(struct net *net, struct socket *sock,
 
 	udp_tunnel_encap_enable(sk);
 
-	udp_tunnel_update_gro_rcv(sock->sk, true);
+	udp_tunnel_update_gro_rcv(sk, true);
 
-	if (!sk->sk_dport && !sk->sk_bound_dev_if && sk_saddr_any(sock->sk))
-		udp_tunnel_update_gro_lookup(net, sock->sk, true);
+	if (!sk->sk_dport && !sk->sk_bound_dev_if && sk_saddr_any(sk))
+		udp_tunnel_update_gro_lookup(net, sk, true);
 }
 EXPORT_SYMBOL_GPL(setup_udp_tunnel_sock);
 
Signature · 1 line
2.48.1
maintainer Reviewed-by on v2 4/5
#13 lore raw

Re: [PATCH net-next v2 4/5] udp_tunnel: avoid inconsistent local variables usage

Quoted 7 lines · Paolo Abeni

In setup_udp_tunnel_sock() 'sk' and 'sock->sk' are alias. The code I introduced there uses alternatively both variables, for no good reasons.

Stick to 'sk' usage, to be consistent with the prior code.

Reviewed-by on v2 4/5
#14 lore raw

Re: [PATCH net-next v2 4/5] udp_tunnel: avoid inconsistent local variables usage

In reply to Paolo Abeni
Quoted 7 lines · Paolo Abeni

In setup_udp_tunnel_sock() 'sk' and 'sock->sk' are alias. The code I introduced there uses alternatively both variables, for no good reasons.

Stick to 'sk' usage, to be consistent with the prior code.

Signature · 1 line
Sabrina
author maintainer
#15 lore raw

PATCH v2 5/5 udp_tunnel: prevent GRO lookup optimization for user-space sockets

In reply to Paolo Abeni

UDP tunnel sockets owned by the kernel are never disconnected/rebound after setup_udp_tunnel_sock(), but sockets owned by the user-space could go through such changes after being matching the criteria to enable the GRO lookup optimization, breaking them.

Explicitly prevent user-space owned sockets from leveraging such optimization.

---
 net/ipv4/udp_tunnel_core.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
net/ipv4/udp_tunnel_core.c +2 -1
index aa9016619c25a..2326548997d3f 100644
@@ -92,7 +92,8 @@ void setup_udp_tunnel_sock(struct net *net, struct socket *sock,
 
 	udp_tunnel_update_gro_rcv(sk, true);
 
-	if (!sk->sk_dport && !sk->sk_bound_dev_if && sk_saddr_any(sk))
+	if (!sk->sk_dport && !sk->sk_bound_dev_if && sk_saddr_any(sk) &&
+	    sk->sk_kern_sock)
 		udp_tunnel_update_gro_lookup(net, sk, true);
 }
 EXPORT_SYMBOL_GPL(setup_udp_tunnel_sock);
Signature · 1 line
2.48.1
maintainer Reviewed-by on v2 5/5
#16 lore raw

Re: [PATCH net-next v2 5/5] udp_tunnel: prevent GRO lookup optimization for user-space sockets

Quoted 7 lines · Paolo Abeni

UDP tunnel sockets owned by the kernel are never disconnected/rebound after setup_udp_tunnel_sock(), but sockets owned by the user-space could go through such changes after being matching the criteria to enable the GRO lookup optimization, breaking them.

Explicitly prevent user-space owned sockets from leveraging such optimization.

UDP tunnel sockets are generally initialized with udp_sock_create, which calls sock_create_kern.

Quoted 1 line
Reviewed-by on v2 5/5
#17 lore raw

Re: [PATCH net-next v2 5/5] udp_tunnel: prevent GRO lookup optimization for user-space sockets

In reply to Paolo Abeni
Quoted 9 lines · Paolo Abeni

UDP tunnel sockets owned by the kernel are never disconnected/rebound after setup_udp_tunnel_sock(), but sockets owned by the user-space could go through such changes after being matching the criteria to enable the GRO lookup optimization, breaking them.

Explicitly prevent user-space owned sockets from leveraging such optimization.

Signature · 1 line
Sabrina
maintainer
#18 lore raw
In reply to Paolo Abeni
Quoted 2 lines · On Fri, 21 Mar 2025 12:52:51 +0100 Paolo Abeni wrote:

Should this prove to be too invasive, too late or ineffective, I'll be ok with a revert before the upcoming net-next PR.

Looks like the conversation isn't dying down and the patches were merged quite recently, so revert isn't too risky. I'll go ahead and revert.

Signature · 1 line
pw-bot: nap