* [PATCH v1 net 0/8] tcp: Misc fixes for tcp_v[46]_err().
@ 2026-11-17 5:35 Kuniyuki Iwashima
2026-11-17 5:35 ` [PATCH v1 net 1/8] tcp: Don't accept ICMP for SND.NXT in TCP_SYN_SENT Kuniyuki Iwashima
` (7 more replies)
0 siblings, 8 replies; 19+ messages in thread
From: Kuniyuki Iwashima @ 2026-11-17 5:35 UTC (permalink / raw)
To: Eric Dumazet, Neal Cardwell, David S . Miller, Jakub Kicinski,
Paolo Abeni
Cc: Simon Horman, Kuniyuki Iwashima, Kuniyuki Iwashima, netdev
A packetdrill-equipped LLM found a bunch of bugs with repro.
This is the first series fixing bugs in tcp_v4_err() and
tcp_v6_err().
Patches 1-5, 7, and 8 are fixes, and Patch 6 is a selftest
for the bugs fixed by Patches 1-5.
Kuniyuki Iwashima (8):
tcp: Don't accept ICMP for SND.NXT in TCP_SYN_SENT.
tcp: Don't use inner IP TTL for IP_MINTTL in tcp_v4_err().
tcp: Don't bypass GTSM check for reqsk.
tcp: Fix reqsk abort condition on ICMP errors.
tcp: Check ip_sk_accept_pmtu() in tcp_v4_err().
selftest: net: packetdrill: Add ICMP and ICMPv6 tests.
tcp: Pass inet_sdif() to __inet_lookup_established() in tcp_v4_err().
tcp: Pass inet6_iif() to __inet6_lookup_established() in tcp_v6_err().
net/ipv4/tcp_ipv4.c | 42 ++++---
net/ipv6/tcp_ipv6.c | 33 ++++--
.../net/packetdrill/tcp_icmp_ipv4.pkt | 98 ++++++++++++++++
.../net/packetdrill/tcp_icmp_ipv6.pkt | 111 ++++++++++++++++++
4 files changed, 253 insertions(+), 31 deletions(-)
create mode 100644 tools/testing/selftests/net/packetdrill/tcp_icmp_ipv4.pkt
create mode 100644 tools/testing/selftests/net/packetdrill/tcp_icmp_ipv6.pkt
--
2.56.0.679.gc527c65306-goog
^ permalink raw reply [flat|nested] 19+ messages in thread
* [PATCH v1 net 1/8] tcp: Don't accept ICMP for SND.NXT in TCP_SYN_SENT.
2026-11-17 5:35 [PATCH v1 net 0/8] tcp: Misc fixes for tcp_v[46]_err() Kuniyuki Iwashima
@ 2026-11-17 5:35 ` Kuniyuki Iwashima
2026-11-17 8:35 ` Eric Dumazet
2026-11-17 5:35 ` [PATCH v1 net 2/8] tcp: Don't use inner IP TTL for IP_MINTTL in tcp_v4_err() Kuniyuki Iwashima
` (6 subsequent siblings)
7 siblings, 1 reply; 19+ messages in thread
From: Kuniyuki Iwashima @ 2026-11-17 5:35 UTC (permalink / raw)
To: Eric Dumazet, Neal Cardwell, David S . Miller, Jakub Kicinski,
Paolo Abeni
Cc: Simon Horman, Kuniyuki Iwashima, Kuniyuki Iwashima, netdev
RFC 5927 states that the acceptable ICMP has seq in
SND.UNA =< SEG.SEQ < SND.NXT:
These implementations check that the TCP sequence number
contained in the payload of the ICMP error message is within
the range SND.UNA =< SEG.SEQ < SND.NXT.
tcp_v4_err() and tcp_v6_err() accept packets if seq meets
between(seq, snd_una, tp->snd_nxt)
-> tp->snd_nxt - snd_una >= seq - snd_una
, which includes both ends.
This is fine once connection is established because
tcp_acceptable_seq() uses SND_NXT for ACK and RST.
However, we should not accept ICMP to SND_NXT for SYN_SENT.
The minimal repro is:
--ip_version=ipv4
0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+0 fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
+0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
+0 > S 0:0(0) <...>
+0 < icmp unreachable host_unreachable [1:1(0)]
+0 %{assert tcpi_state == TCP_SYN_SENT}%
Let's fix off-by-one.
Note that we need to check seq != snd_una because tp->snd_nxt
is only advanced to tp->write_seq after tcp_transmit_skb().
Otherwise, ICMP for SYN could be dropped before tcp_connect()
advances tp->snd_nxt.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
net/ipv4/tcp_ipv4.c | 9 ++++++---
net/ipv6/tcp_ipv6.c | 9 ++++++---
2 files changed, 12 insertions(+), 6 deletions(-)
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 2fc92a7884fc..7d58f522068d 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -499,8 +499,8 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
const int type = icmp_hdr(skb)->type;
const int code = icmp_hdr(skb)->code;
struct request_sock *fastopen;
+ u32 seq, snd_una, snd_nxt;
struct tcp_sock *tp;
- u32 seq, snd_una;
struct sock *sk;
int err;
@@ -555,9 +555,12 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
tp = tcp_sk(sk);
/* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */
fastopen = rcu_dereference(tp->fastopen_rsk);
- snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : tp->snd_una;
+ snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : READ_ONCE(tp->snd_una);
+ snd_nxt = READ_ONCE(tp->snd_nxt);
if (sk->sk_state != TCP_LISTEN &&
- !between(seq, snd_una, tp->snd_nxt)) {
+ (!between(seq, snd_una, snd_nxt) ||
+ (sk->sk_state == TCP_SYN_SENT &&
+ seq != snd_una && seq == snd_nxt))) {
__NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
goto out;
}
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index 5c7ed38f1f56..8691d220cf94 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -410,9 +410,9 @@ static int tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
struct net *net = dev_net_rcu(skb->dev);
struct request_sock *fastopen;
+ __u32 seq, snd_una, snd_nxt;
struct ipv6_pinfo *np;
struct tcp_sock *tp;
- __u32 seq, snd_una;
struct sock *sk;
bool fatal;
int err;
@@ -463,9 +463,12 @@ static int tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
tp = tcp_sk(sk);
/* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */
fastopen = rcu_dereference(tp->fastopen_rsk);
- snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : tp->snd_una;
+ snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : READ_ONCE(tp->snd_una);
+ snd_nxt = READ_ONCE(tp->snd_nxt);
if (sk->sk_state != TCP_LISTEN &&
- !between(seq, snd_una, tp->snd_nxt)) {
+ (!between(seq, snd_una, snd_nxt) ||
+ (sk->sk_state == TCP_SYN_SENT &&
+ seq != snd_una && seq == snd_nxt))) {
__NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
goto out;
}
--
2.56.0.679.gc527c65306-goog
^ permalink raw reply related [flat|nested] 19+ messages in thread
* [PATCH v1 net 2/8] tcp: Don't use inner IP TTL for IP_MINTTL in tcp_v4_err().
2026-11-17 5:35 [PATCH v1 net 0/8] tcp: Misc fixes for tcp_v[46]_err() Kuniyuki Iwashima
2026-11-17 5:35 ` [PATCH v1 net 1/8] tcp: Don't accept ICMP for SND.NXT in TCP_SYN_SENT Kuniyuki Iwashima
@ 2026-11-17 5:35 ` Kuniyuki Iwashima
2026-11-17 8:42 ` Eric Dumazet
2026-11-17 5:35 ` [PATCH v1 net 3/8] tcp: Don't bypass GTSM check for reqsk Kuniyuki Iwashima
` (5 subsequent siblings)
7 siblings, 1 reply; 19+ messages in thread
From: Kuniyuki Iwashima @ 2026-11-17 5:35 UTC (permalink / raw)
To: Eric Dumazet, Neal Cardwell, David S . Miller, Jakub Kicinski,
Paolo Abeni
Cc: Simon Horman, Kuniyuki Iwashima, Kuniyuki Iwashima, netdev,
Stephen Hemminger, stephen hemminger
RFC 5082 introduced Generalized TTL Security Mechanism (GTSM),
and IP_MINTTL and IPV6_MINHOPCOUNT follow that.
If an incoming IPv4 packet has TTL less than inet_sk(sk)->min_ttl
configured via IP_MINTTL, the packet is dropped.
However, tcp_v4_err() checks the TTL extracted from the ICMP
payload, not the one in the outer IP header.
The minimal repro is:
--ip_version=ipv4
0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+0 setsockopt(3, SOL_IP, 21, [250], 4) = 0
+0 fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
+0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
+0 > S 0:0(0) <...>
+0 < (ttl 10) icmp unreachable host_unreachable [0:0(0)]
+0 %{assert tcpi_state == TCP_SYN_SENT}%
Let's check ip_hdr(skb)->ttl.
Now GTSM is applied to locally generated ICMP errors, but this
behaviour matches tcp_v6_err().
Note that tcp_v6_err() properly checks ipv6_hdr(skb)->hop_limit.
Fixes: 97e3ecd112ba ("TCP: check min TTL on received ICMP packets")
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
Cc: Stephen Hemminger <stephen@networkplumber.org>
Cc: stephen hemminger <shemminger@vyatta.com>
---
net/ipv4/tcp_ipv4.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 7d58f522068d..bf8542368032 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -546,7 +546,7 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
if (static_branch_unlikely(&ip4_min_ttl)) {
/* min_ttl can be changed concurrently from do_ip_setsockopt() */
- if (unlikely(iph->ttl < READ_ONCE(inet_sk(sk)->min_ttl))) {
+ if (unlikely(ip_hdr(skb)->ttl < READ_ONCE(inet_sk(sk)->min_ttl))) {
__NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
goto out;
}
--
2.56.0.679.gc527c65306-goog
^ permalink raw reply related [flat|nested] 19+ messages in thread
* [PATCH v1 net 3/8] tcp: Don't bypass GTSM check for reqsk.
2026-11-17 5:35 [PATCH v1 net 0/8] tcp: Misc fixes for tcp_v[46]_err() Kuniyuki Iwashima
2026-11-17 5:35 ` [PATCH v1 net 1/8] tcp: Don't accept ICMP for SND.NXT in TCP_SYN_SENT Kuniyuki Iwashima
2026-11-17 5:35 ` [PATCH v1 net 2/8] tcp: Don't use inner IP TTL for IP_MINTTL in tcp_v4_err() Kuniyuki Iwashima
@ 2026-11-17 5:35 ` Kuniyuki Iwashima
2026-11-17 8:45 ` Eric Dumazet
2026-11-17 5:35 ` [PATCH v1 net 4/8] tcp: Fix reqsk abort condition on ICMP errors Kuniyuki Iwashima
` (4 subsequent siblings)
7 siblings, 1 reply; 19+ messages in thread
From: Kuniyuki Iwashima @ 2026-11-17 5:35 UTC (permalink / raw)
To: Eric Dumazet, Neal Cardwell, David S . Miller, Jakub Kicinski,
Paolo Abeni
Cc: Simon Horman, Kuniyuki Iwashima, Kuniyuki Iwashima, netdev
Since cited commits moved up TCP_NEW_SYN_RECV handling in
tcp_v4_err() and tcp_v6_err(), ICMP messages bypass IP_MINTTL
and IPV6_MINHOPCOUNT checks.
The minimal repro is:
--ip_version=ipv6
0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+0 setsockopt(3, SOL_IPV6, 73, [255], 4) = 0
+0 bind(3, ..., ...) = 0
+0 listen(3, 10) = 0
+0 < (ttl 255) S 0:0(0) win 65535 <mss 1440,nop,nop,sackOK>
+0 > S. 0:0(0) ack 1 <mss 1440,nop,nop,sackOK>
+0 < (ttl 1) icmp unreachable port_unreachable [0:0(0)]
+0 < (ttl 255) . 1:1(0) ack 1 win 65535
+0 fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
+0 accept(3, ..., ...) = 4
+0 %{assert tcpi_state == TCP_ESTABLISHED}%
Let's restore IP_MINTTL/IPV6_MINHOPCOUNT for reqsk.
Fixes: 26e3736090e1 ("ipv4: tcp: handle ICMP messages on TCP_NEW_SYN_RECV request sockets")
Fixes: 2215089b2244 ("ipv6: tcp: handle ICMP messages on TCP_NEW_SYN_RECV request sockets")
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
net/ipv4/tcp_ipv4.c | 19 +++++++++++--------
net/ipv6/tcp_ipv6.c | 19 +++++++++++--------
2 files changed, 22 insertions(+), 16 deletions(-)
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index bf8542368032..fd9588bba967 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -516,6 +516,17 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
inet_twsk_put(inet_twsk(sk));
return 0;
}
+
+ if (static_branch_unlikely(&ip4_min_ttl)) {
+ /* min_ttl can be changed concurrently from do_ip_setsockopt() */
+ if (unlikely(ip_hdr(skb)->ttl <
+ READ_ONCE(inet_sk(sk_to_full_sk(sk))->min_ttl))) {
+ __NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
+ sock_gen_put(sk);
+ return 0;
+ }
+ }
+
seq = ntohl(th->seq);
if (sk->sk_state == TCP_NEW_SYN_RECV) {
tcp_req_err(sk, seq, type == ICMP_PARAMETERPROB ||
@@ -544,14 +555,6 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
if (sk->sk_state == TCP_CLOSE || inet_test_bit(DEFER_CONNECT, sk))
goto out;
- if (static_branch_unlikely(&ip4_min_ttl)) {
- /* min_ttl can be changed concurrently from do_ip_setsockopt() */
- if (unlikely(ip_hdr(skb)->ttl < READ_ONCE(inet_sk(sk)->min_ttl))) {
- __NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
- goto out;
- }
- }
-
tp = tcp_sk(sk);
/* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */
fastopen = rcu_dereference(tp->fastopen_rsk);
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index 8691d220cf94..0f1523fff939 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -433,6 +433,17 @@ static int tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
inet_twsk_put(inet_twsk(sk));
return 0;
}
+
+ if (static_branch_unlikely(&ip6_min_hopcount)) {
+ /* min_hopcount can be changed concurrently from do_ipv6_setsockopt() */
+ if (ipv6_hdr(skb)->hop_limit <
+ READ_ONCE(tcp_inet6_sk(sk_to_full_sk(sk))->min_hopcount)) {
+ __NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
+ sock_gen_put(sk);
+ return 0;
+ }
+ }
+
seq = ntohl(th->seq);
fatal = icmpv6_err_convert(type, code, &err);
if (sk->sk_state == TCP_NEW_SYN_RECV) {
@@ -452,14 +463,6 @@ static int tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
if (sk->sk_state == TCP_CLOSE || inet_test_bit(DEFER_CONNECT, sk))
goto out;
- if (static_branch_unlikely(&ip6_min_hopcount)) {
- /* min_hopcount can be changed concurrently from do_ipv6_setsockopt() */
- if (ipv6_hdr(skb)->hop_limit < READ_ONCE(tcp_inet6_sk(sk)->min_hopcount)) {
- __NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
- goto out;
- }
- }
-
tp = tcp_sk(sk);
/* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */
fastopen = rcu_dereference(tp->fastopen_rsk);
--
2.56.0.679.gc527c65306-goog
^ permalink raw reply related [flat|nested] 19+ messages in thread
* [PATCH v1 net 4/8] tcp: Fix reqsk abort condition on ICMP errors.
2026-11-17 5:35 [PATCH v1 net 0/8] tcp: Misc fixes for tcp_v[46]_err() Kuniyuki Iwashima
` (2 preceding siblings ...)
2026-11-17 5:35 ` [PATCH v1 net 3/8] tcp: Don't bypass GTSM check for reqsk Kuniyuki Iwashima
@ 2026-11-17 5:35 ` Kuniyuki Iwashima
2026-11-17 8:49 ` Eric Dumazet
2026-11-17 5:35 ` [PATCH v1 net 5/8] tcp: Check ip_sk_accept_pmtu() in tcp_v4_err() Kuniyuki Iwashima
` (3 subsequent siblings)
7 siblings, 1 reply; 19+ messages in thread
From: Kuniyuki Iwashima @ 2026-11-17 5:35 UTC (permalink / raw)
To: Eric Dumazet, Neal Cardwell, David S . Miller, Jakub Kicinski,
Paolo Abeni
Cc: Simon Horman, Kuniyuki Iwashima, Kuniyuki Iwashima, netdev
RFC 1122 Section 4.2.3.9 states the following soft errors
must not abort connections:
o Destination Unreachable -- codes 0, 1, 5
Since these Unreachable messages indicate soft error
conditions, TCP MUST NOT abort the connection, and it
SHOULD make the information available to the
application.
o Time Exceeded -- codes 0, 1
o Parameter Problem
This should be handled the same way as Destination
Unreachable codes 0, 1, 5 (see above).
Commit 9cf7490360bf ("tcp: do not drop syn_recv on all icmp
reports") introduced the abort arg to tcp_req_err() to avoid
aborting connection on ICMP_REDIRECT.
However, tcp_v4_err() still abort on soft errors (ICMP_NET_UNREACH
(0), ICMP_HOST_UNREACH (1)) and ignore hard errors.
For example, we can see SYN+ACK retransmitted after receiving
hard error:
--ip_version=ipv4
0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+0 bind(3, ..., ...) = 0
+0 listen(3, 10) = 0
+0 < S 0:0(0) win 65535 <mss 1460,nop,nop,sackOK>
+0 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK>
+0 < icmp unreachable port_unreachable [0:0(0)]
+1 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK>
Also, tcp_v6_err() uses icmpv6_err_convert(), which handles
ICMPV6_PARAMPROB messages as hard error while RFC 5461 defines
as soft error.
But we should handle ICMPV6_UNK_NEXTHDR as hard error because
it is used as IPv6 version ICMP_PROT_UNREACH.
Let's correct the abort condition in tcp_v4_err() and
tcp_v6_err().
Note that icmpv6_err_convert() is used by other protocols.
Fixes: 9cf7490360bf ("tcp: do not drop syn_recv on all icmp reports")
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
net/ipv4/tcp_ipv4.c | 7 ++-----
net/ipv6/tcp_ipv6.c | 3 ++-
2 files changed, 4 insertions(+), 6 deletions(-)
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index fd9588bba967..f11e090d04fb 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -529,11 +529,8 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
seq = ntohl(th->seq);
if (sk->sk_state == TCP_NEW_SYN_RECV) {
- tcp_req_err(sk, seq, type == ICMP_PARAMETERPROB ||
- type == ICMP_TIME_EXCEEDED ||
- (type == ICMP_DEST_UNREACH &&
- (code == ICMP_NET_UNREACH ||
- code == ICMP_HOST_UNREACH)));
+ tcp_req_err(sk, seq, (type == ICMP_DEST_UNREACH &&
+ icmp_err_convert[code].fatal));
return 0;
}
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index 0f1523fff939..29fd0a084c91 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -447,7 +447,8 @@ static int tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
seq = ntohl(th->seq);
fatal = icmpv6_err_convert(type, code, &err);
if (sk->sk_state == TCP_NEW_SYN_RECV) {
- tcp_req_err(sk, seq, fatal);
+ tcp_req_err(sk, seq, type == ICMPV6_PARAMPROB ?
+ code == ICMPV6_UNK_NEXTHDR : fatal);
return 0;
}
--
2.56.0.679.gc527c65306-goog
^ permalink raw reply related [flat|nested] 19+ messages in thread
* [PATCH v1 net 5/8] tcp: Check ip_sk_accept_pmtu() in tcp_v4_err().
2026-11-17 5:35 [PATCH v1 net 0/8] tcp: Misc fixes for tcp_v[46]_err() Kuniyuki Iwashima
` (3 preceding siblings ...)
2026-11-17 5:35 ` [PATCH v1 net 4/8] tcp: Fix reqsk abort condition on ICMP errors Kuniyuki Iwashima
@ 2026-11-17 5:35 ` Kuniyuki Iwashima
2026-11-17 8:54 ` Eric Dumazet
2026-11-17 5:35 ` [PATCH v1 net 6/8] selftest: net: packetdrill: Add ICMP and ICMPv6 tests Kuniyuki Iwashima
` (2 subsequent siblings)
7 siblings, 1 reply; 19+ messages in thread
From: Kuniyuki Iwashima @ 2026-11-17 5:35 UTC (permalink / raw)
To: Eric Dumazet, Neal Cardwell, David S . Miller, Jakub Kicinski,
Paolo Abeni
Cc: Simon Horman, Kuniyuki Iwashima, Kuniyuki Iwashima, netdev,
Hannes Frederic Sowa
Currently, tcp_v4_mtu_reduced() checks ip_sk_accept_pmtu()
after calling inet_csk_update_pmtu(), which is too late.
Even if we configure a socket not to perform Path MTU
Discovery with IP_PMTUDISC_INTERFACE, ICMP_FRAG_NEEDED
can overwrite MTU.
The last getsockopt() returns 1200 MTU:
--ip_version=ipv4
0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+0 bind(3, ..., ...) = 0
+0 listen(3, 1) = 0
+0 < S 0:0(0) win 65535 <mss 1460,sackOK,nop,nop,nop,wscale 8>
+0 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 9>
+0 < . 1:1(0) ack 1 win 257
+0 accept(3, ..., ...) = 4
+0 setsockopt(4, SOL_IP, IP_MTU_DISCOVER, [4], 4) = 0
+0 getsockopt(4, SOL_IP, IP_MTU, [1500], [4]) = 0
+0 write(4, ..., 1000) = 1000
+0 > P. 1:1001(1000) ack 1
+0 < icmp unreachable frag_needed mtu 1200 [1:1001(1000)]
+0 getsockopt(4, SOL_IP, IP_MTU, [1500], [4]) = 0
Let's check ip_sk_accept_pmtu() in tcp_v4_err() as done
in tcp_v6_err().
Note that inet_sk(sk)->pmtudisc is configured locklessly,
so we do not recheck it in tcp_v4_mtu_reduced(). This is
same with tcp_v6_mtu_reduced() too.
Fixes: 482fc6094afa ("ipv4: introduce new IP_MTU_DISCOVER mode IP_PMTUDISC_INTERFACE")
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
Cc: Hannes Frederic Sowa <hannes@stressinduktion.org>
---
net/ipv4/tcp_ipv4.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index f11e090d04fb..bc909fedda3a 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -391,7 +391,6 @@ void tcp_v4_mtu_reduced(struct sock *sk)
WRITE_ONCE(sk->sk_err_soft, EMSGSIZE);
if (inet->pmtudisc != IP_PMTUDISC_DONT &&
- ip_sk_accept_pmtu(sk) &&
inet_csk(sk)->icsk_pmtu_cookie > dmtu) {
tcp_sync_mss(sk, dmtu);
@@ -588,6 +587,9 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
if (sk->sk_state == TCP_LISTEN)
goto out;
+ if (!ip_sk_accept_pmtu(sk))
+ goto out;
+
WRITE_ONCE(tp->mtu_info, info);
if (!sock_owned_by_user(sk)) {
tcp_v4_mtu_reduced(sk);
--
2.56.0.679.gc527c65306-goog
^ permalink raw reply related [flat|nested] 19+ messages in thread
* [PATCH v1 net 6/8] selftest: net: packetdrill: Add ICMP and ICMPv6 tests.
2026-11-17 5:35 [PATCH v1 net 0/8] tcp: Misc fixes for tcp_v[46]_err() Kuniyuki Iwashima
` (4 preceding siblings ...)
2026-11-17 5:35 ` [PATCH v1 net 5/8] tcp: Check ip_sk_accept_pmtu() in tcp_v4_err() Kuniyuki Iwashima
@ 2026-11-17 5:35 ` Kuniyuki Iwashima
2026-11-17 8:58 ` Eric Dumazet
2026-11-17 5:35 ` [PATCH v1 net 7/8] tcp: Pass inet_sdif() to __inet_lookup_established() in tcp_v4_err() Kuniyuki Iwashima
2026-11-17 5:35 ` [PATCH v1 net 8/8] tcp: Pass inet6_iif() to __inet6_lookup_established() in tcp_v6_err() Kuniyuki Iwashima
7 siblings, 1 reply; 19+ messages in thread
From: Kuniyuki Iwashima @ 2026-11-17 5:35 UTC (permalink / raw)
To: Eric Dumazet, Neal Cardwell, David S . Miller, Jakub Kicinski,
Paolo Abeni
Cc: Simon Horman, Kuniyuki Iwashima, Kuniyuki Iwashima, netdev
tcp_icmp_ipv4.pkt and tcp_icmp_ipv6.pkt cover the scenarios
fixed by the preceding patches.
tcp_v6_err() already checks the outer hop limit and
ip6_sk_accept_pmtu(), so Test 2 and 5 in tcp_icmp_ipv6.pkt
pass without the series.
ICMP and ICMPv6 behaviour are unlikely to change, so splitting
into multiple files per test case is not worth it.
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
.../net/packetdrill/tcp_icmp_ipv4.pkt | 98 ++++++++++++++++
.../net/packetdrill/tcp_icmp_ipv6.pkt | 111 ++++++++++++++++++
2 files changed, 209 insertions(+)
create mode 100644 tools/testing/selftests/net/packetdrill/tcp_icmp_ipv4.pkt
create mode 100644 tools/testing/selftests/net/packetdrill/tcp_icmp_ipv6.pkt
diff --git a/tools/testing/selftests/net/packetdrill/tcp_icmp_ipv4.pkt b/tools/testing/selftests/net/packetdrill/tcp_icmp_ipv4.pkt
new file mode 100644
index 000000000000..122bf9a14862
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_icmp_ipv4.pkt
@@ -0,0 +1,98 @@
+// SPDX-License-Identifier: GPL-2.0
+--ip_version=ipv4
+
+`./defaults.sh`
+
+// Test 1: Client in TCP_SYN_SENT rejects ICMP quoting seq == snd_nxt (RFC 5927 4.1)
+// In SYN_SENT, SND.UNA = ISN, SND.NXT = ISN + 1.
+// The sequence number in ICMP must satisfy SND.UNA <= SEG.SEQ < SND.NXT.
+// An ICMP message quoting seq == 1 (snd_nxt) must be dropped, while seq == 0 (snd_una) aborts.
+ 0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+ +0 fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
+ +0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
+ +0 > S 0:0(0) <...>
+ // Out-of-window ICMP quoting seq == 1 (snd_nxt) must be dropped
+ +0 < icmp unreachable host_unreachable [1:1(0)]
+ // Socket stays in TCP_SYN_SENT, so SYN retransmission occurs after 1s
+ +1 > S 0:0(0) <...>
+ // In-window ICMP quoting seq == 0 (snd_una) is accepted and aborts connection
+ +0 < icmp unreachable host_unreachable [0:0(0)]
+ +0 %{ assert tcpi_state == TCP_CLOSE }%
+ +0 getsockopt(3, SOL_SOCKET, SO_ERROR, [EHOSTUNREACH], [4]) = 0
+ +0 close(3) = 0
+
+// Test 2: Client with IP_MINTTL (21) checks outer IP TTL in tcp_v4_err()
+// An ICMP with outer TTL < min_ttl (250) must be dropped, while outer TTL >= min_ttl
+// must be accepted even though the quoted inner IP header has lower TTL (64).
+ +0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+ +0 setsockopt(3, SOL_IP, 21, [250], 4) = 0
+ +0 fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
+ +0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
+ +0 > S 0:0(0) <...>
+ // Outer TTL 10 < min_ttl (250): dropped, socket stays in TCP_SYN_SENT
+ +0 < (ttl 10) icmp unreachable host_unreachable [0:0(0)]
+ +1 > S 0:0(0) <...>
+ // Outer TTL 255 >= min_ttl (250): accepted, socket aborts
+ +0 < (ttl 255) icmp unreachable host_unreachable [0:0(0)]
+ +0 %{ assert tcpi_state == TCP_CLOSE }%
+ +0 getsockopt(3, SOL_SOCKET, SO_ERROR, [EHOSTUNREACH], [4]) = 0
+ +0 close(3) = 0
+
+// Test 3: Listener with IP_MINTTL (21) checks TTL before reqsk handling
+// An incoming fatal ICMP with outer TTL < min_ttl (255) must not abort TCP_NEW_SYN_RECV,
+// while outer TTL >= min_ttl (255) must abort it.
+ +0 socket(..., SOCK_STREAM|SOCK_NONBLOCK, IPPROTO_TCP) = 3
+ +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+ +0 setsockopt(3, SOL_IP, 21, [255], 4) = 0
+ +0 bind(3, ..., ...) = 0
+ +0 listen(3, 10) = 0
+ +0 < (ttl 255) S 0:0(0) win 65535 <mss 1460,nop,nop,sackOK>
+ +0 > (ttl 64) S. 0:0(0) ack 1 <...>
+ // Outer TTL 10 < min_ttl (255): dropped, reqsk survives
+ +0 < (ttl 10) icmp unreachable port_unreachable [0:0(0)]
+ +1 > (ttl 64) S. 0:0(0) ack 1 <...>
+ // Outer TTL 255 >= min_ttl (255): accepted, fatal ICMP aborts reqsk
+ +0 < (ttl 255) icmp unreachable port_unreachable [0:0(0)]
+ +0 < (ttl 255) . 1:1(0) ack 1 win 65535
+ +0 > R 1:1(0)
+ +0 close(3) = 0
+
+// Test 4: Server in TCP_NEW_SYN_RECV does not abort on soft errors but aborts on fatal error
+ +0 socket(..., SOCK_STREAM|SOCK_NONBLOCK, IPPROTO_TCP) = 3
+ +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+ +0 bind(3, ..., ...) = 0
+ +0 listen(3, 10) = 0
+ +0 < S 0:0(0) win 65535 <mss 1460,nop,nop,sackOK>
+ +0 > S. 0:0(0) ack 1 <...>
+ // 4a: Soft errors must not abort reqsk
+ +0 < icmp unreachable host_unreachable [0:0(0)]
+ +0 < icmp unreachable net_unreachable [0:0(0)]
+ +0 < icmp time_exceeded [0:0(0)]
+ +0 < icmp parameter_problem [0:0(0)]
+ +1 > S. 0:0(0) ack 1 <...>
+ // 4b: Fatal error (port_unreachable) must abort reqsk
+ +0 < icmp unreachable port_unreachable [0:0(0)]
+ // Remote sends 3rd ACK: listener responds with RST because reqsk was aborted
+ +0 < . 1:1(0) ack 1 win 65535
+ +0 > R 1:1(0)
+ +0 close(3) = 0
+
+// Test 5: Socket with IP_PMTUDISC_INTERFACE ignores ICMP_FRAG_NEEDED
+// Socket configured with IP_PMTUDISC_INTERFACE (4) must not overwrite MTU
+// upon receiving ICMP_DEST_UNREACH with code ICMP_FRAG_NEEDED.
+ +0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+ +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+ +0 bind(3, ..., ...) = 0
+ +0 listen(3, 1) = 0
+ +0 < S 0:0(0) win 65535 <mss 1460,sackOK,nop,nop,nop,wscale 8>
+ +0 > S. 0:0(0) ack 1 <...>
+ +0 < . 1:1(0) ack 1 win 257
+ +0 accept(3, ..., ...) = 4
+ +0 setsockopt(4, SOL_IP, IP_MTU_DISCOVER, [4], 4) = 0
+ +0 getsockopt(4, SOL_IP, IP_MTU, [1500], [4]) = 0
+ +0 write(4, ..., 1000) = 1000
+ +0 > P. 1:1001(1000) ack 1 <...>
+ +0 < icmp unreachable frag_needed mtu 1200 [1:1001(1000)]
+ +0 getsockopt(4, SOL_IP, IP_MTU, [1500], [4]) = 0
+ +0 close(4) = 0
+ +0 close(3) = 0
diff --git a/tools/testing/selftests/net/packetdrill/tcp_icmp_ipv6.pkt b/tools/testing/selftests/net/packetdrill/tcp_icmp_ipv6.pkt
new file mode 100644
index 000000000000..bd0cc365a088
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_icmp_ipv6.pkt
@@ -0,0 +1,111 @@
+// SPDX-License-Identifier: GPL-2.0
+--ip_version=ipv6
+
+`./defaults.sh`
+
+// Test 1: Client in TCP_SYN_SENT rejects ICMPv6 quoting seq == snd_nxt (RFC 5927 4.1)
+// In SYN_SENT, SND.UNA = ISN, SND.NXT = ISN + 1.
+// The sequence number in ICMP must satisfy SND.UNA <= SEG.SEQ < SND.NXT.
+// An ICMP message quoting seq == 1 (snd_nxt) must be dropped, while seq == 0 (snd_una) aborts.
+ 0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+ +0 fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
+ +0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
+ +0 > S 0:0(0) <...>
+ // Out-of-window ICMPv6 quoting seq == 1 (snd_nxt) must be dropped
+ +0 < icmp unreachable address_unreachable [1:1(0)]
+ // Socket stays in TCP_SYN_SENT, so SYN retransmission occurs after 1s
+ +1 > S 0:0(0) <...>
+ // In-window ICMPv6 quoting seq == 0 (snd_una) is accepted and aborts connection
+ +0 < icmp unreachable address_unreachable [0:0(0)]
+ +0 %{ assert tcpi_state == TCP_CLOSE }%
+ +0 getsockopt(3, SOL_SOCKET, SO_ERROR, [EHOSTUNREACH], [4]) = 0
+ +0 close(3) = 0
+
+// Test 2: Client with IPV6_MINHOPCOUNT (73) checks outer IPv6 hop_limit in tcp_v6_err()
+// An ICMPv6 with outer hop_limit < min_hopcount (250) must be dropped, while outer
+// hop_limit >= min_hopcount must be accepted even though the quoted inner hop_limit is 64.
+ +0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+ +0 setsockopt(3, SOL_IPV6, 73, [250], 4) = 0
+ +0 fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
+ +0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
+ +0 > S 0:0(0) <...>
+ // Outer hop_limit 10 < min_hopcount (250): dropped, socket stays in TCP_SYN_SENT
+ +0 < (ttl 10) icmp unreachable address_unreachable [0:0(0)]
+ +1 > S 0:0(0) <...>
+ // Outer hop_limit 255 >= min_hopcount (250): accepted, socket aborts
+ +0 < (ttl 255) icmp unreachable address_unreachable [0:0(0)]
+ +0 %{ assert tcpi_state == TCP_CLOSE }%
+ +0 getsockopt(3, SOL_SOCKET, SO_ERROR, [EHOSTUNREACH], [4]) = 0
+ +0 close(3) = 0
+
+// Test 3: Listener with IPV6_MINHOPCOUNT (73) checks hop_limit before reqsk handling
+// An incoming fatal ICMPv6 with outer hop_limit < min_hopcount (255) must not abort
+// TCP_NEW_SYN_RECV, while outer hop_limit >= min_hopcount (255) must abort it.
+ +0 socket(..., SOCK_STREAM|SOCK_NONBLOCK, IPPROTO_TCP) = 3
+ +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+ +0 setsockopt(3, SOL_IPV6, 73, [255], 4) = 0
+ +0 bind(3, ..., ...) = 0
+ +0 listen(3, 10) = 0
+ +0 < (ttl 255) S 0:0(0) win 65535 <mss 1440,nop,nop,sackOK>
+ +0 > (ttl 64) S. 0:0(0) ack 1 <...>
+ // Outer hop_limit 10 < min_hopcount (255): dropped, reqsk survives
+ +0 < (ttl 10) icmp unreachable port_unreachable [0:0(0)]
+ +1 > (ttl 64) S. 0:0(0) ack 1 <...>
+ // Outer hop_limit 255 >= min_hopcount (255): accepted, fatal ICMPv6 aborts reqsk
+ +0 < (ttl 255) icmp unreachable port_unreachable [0:0(0)]
+ +0 < (ttl 255) . 1:1(0) ack 1 win 65535
+ +0 > R 1:1(0)
+ +0 close(3) = 0
+
+// Test 4: Server in TCP_NEW_SYN_RECV does not abort on soft error or
+// non-UNK_NEXTHDR ICMPV6_PARAMPROB, but aborts on fatal error or
+// ICMPV6_UNK_NEXTHDR.
+ +0 socket(..., SOCK_STREAM|SOCK_NONBLOCK, IPPROTO_TCP) = 3
+ +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+ +0 bind(3, ..., ...) = 0
+ +0 listen(3, 10) = 0
+ +0 < S 0:0(0) win 65535 <mss 1440,nop,nop,sackOK>
+ +0 > S. 0:0(0) ack 1 <...>
+ // 4a: Soft error (address_unreachable) must not abort reqsk
+ +0 < icmp unreachable address_unreachable [0:0(0)]
+ +1 > S. 0:0(0) ack 1 <...>
+ // 4b: Non-UNK_NEXTHDR parameter problem (header_field) must not abort reqsk
+ +0 < icmp parameter_problem header_field [0:0(0)]
+ +2 > S. 0:0(0) ack 1 <...>
+ // 4c: ICMPV6_UNK_NEXTHDR parameter problem (unknown_next_header) must abort reqsk
+ +0 < icmp parameter_problem unknown_next_header [0:0(0)]
+ // Remote sends 3rd ACK: listener responds with RST because reqsk was aborted
+ +0 < . 1:1(0) ack 1 win 65535
+ +0 > R 1:1(0)
+ +0 close(3) = 0
+ // 4d: Fatal unreachable error (port_unreachable) must abort reqsk
+ +0 socket(..., SOCK_STREAM|SOCK_NONBLOCK, IPPROTO_TCP) = 3
+ +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+ +0 bind(3, ..., ...) = 0
+ +0 listen(3, 10) = 0
+ +0 < S 0:0(0) win 65535 <mss 1440,nop,nop,sackOK>
+ +0 > S. 0:0(0) ack 1 <...>
+ +0 < icmp unreachable port_unreachable [0:0(0)]
+ +0 < . 1:1(0) ack 1 win 65535
+ +0 > R 1:1(0)
+ +0 close(3) = 0
+
+// Test 5: Socket with IPV6_PMTUDISC_INTERFACE ignores ICMPV6_PKT_TOOBIG
+// Socket configured with IPV6_PMTUDISC_INTERFACE (4) must not overwrite MTU
+// upon receiving ICMPV6_PKT_TOOBIG.
+ +0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+ +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+ +0 bind(3, ..., ...) = 0
+ +0 listen(3, 1) = 0
+ +0 < S 0:0(0) win 65535 <mss 1440,sackOK,nop,nop,nop,wscale 8>
+ +0 > S. 0:0(0) ack 1 <...>
+ +0 < . 1:1(0) ack 1 win 257
+ +0 accept(3, ..., ...) = 4
+ +0 setsockopt(4, SOL_IPV6, IPV6_MTU_DISCOVER, [4], 4) = 0
+ +0 getsockopt(4, SOL_IPV6, IPV6_MTU, [1520], [4]) = 0
+ +0 write(4, ..., 1000) = 1000
+ +0 > P. 1:1001(1000) ack 1 <...>
+ +0 < icmp packet_too_big mtu 1280 [1:1001(1000)]
+ +0 getsockopt(4, SOL_IPV6, IPV6_MTU, [1520], [4]) = 0
+ +0 close(4) = 0
+ +0 close(3) = 0
--
2.56.0.679.gc527c65306-goog
^ permalink raw reply related [flat|nested] 19+ messages in thread
* [PATCH v1 net 7/8] tcp: Pass inet_sdif() to __inet_lookup_established() in tcp_v4_err().
2026-11-17 5:35 [PATCH v1 net 0/8] tcp: Misc fixes for tcp_v[46]_err() Kuniyuki Iwashima
` (5 preceding siblings ...)
2026-11-17 5:35 ` [PATCH v1 net 6/8] selftest: net: packetdrill: Add ICMP and ICMPv6 tests Kuniyuki Iwashima
@ 2026-11-17 5:35 ` Kuniyuki Iwashima
2026-11-17 9:01 ` Eric Dumazet
2026-11-17 5:35 ` [PATCH v1 net 8/8] tcp: Pass inet6_iif() to __inet6_lookup_established() in tcp_v6_err() Kuniyuki Iwashima
7 siblings, 1 reply; 19+ messages in thread
From: Kuniyuki Iwashima @ 2026-11-17 5:35 UTC (permalink / raw)
To: Eric Dumazet, Neal Cardwell, David S . Miller, Jakub Kicinski,
Paolo Abeni
Cc: Simon Horman, Kuniyuki Iwashima, Kuniyuki Iwashima, netdev,
David Ahern, David Ahern
tcp_v4_err() passes 0 sdif to __inet_lookup_established().
So, ICMP error coming from VRF can be leaked to a socket not
bound to the device.
For example, this test fails at the last assert because the
socket receives ICMP error:
--ip_version=ipv4
0 socket(..., SOCK_STREAM|SOCK_NONBLOCK, IPPROTO_TCP) = 3
+0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
+0 > S 0:0(0) <...>
+0 `ip link add vrf0 type vrf table 10`
+0 `ip link set vrf0 up`
+0 `ip link set dev tun0 master vrf0`
+0 `ip route add 192.0.2.0/24 dev tun0 table 10`
+0 < icmp unreachable port_unreachable [0:0(0)]
+0 %{ assert tcpi_state == TCP_SYN_SENT, tcpi_state }%
Let's pass inet_sdif() to __inet_lookup_established().
Fixes: 3fa6f616a7a4 ("net: ipv4: add second dif to inet socket lookups")
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
Cc: David Ahern <dsahern@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
---
net/ipv4/tcp_ipv4.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index bc909fedda3a..35a1309f8b21 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -504,7 +504,8 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
int err;
sk = __inet_lookup_established(net, iph->daddr, th->dest, iph->saddr,
- ntohs(th->source), inet_iif(skb), 0);
+ ntohs(th->source), inet_iif(skb),
+ inet_sdif(skb));
if (!sk) {
__ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
return -ENOENT;
--
2.56.0.679.gc527c65306-goog
^ permalink raw reply related [flat|nested] 19+ messages in thread
* [PATCH v1 net 8/8] tcp: Pass inet6_iif() to __inet6_lookup_established() in tcp_v6_err().
2026-11-17 5:35 [PATCH v1 net 0/8] tcp: Misc fixes for tcp_v[46]_err() Kuniyuki Iwashima
` (6 preceding siblings ...)
2026-11-17 5:35 ` [PATCH v1 net 7/8] tcp: Pass inet_sdif() to __inet_lookup_established() in tcp_v4_err() Kuniyuki Iwashima
@ 2026-11-17 5:35 ` Kuniyuki Iwashima
2026-11-17 9:03 ` Eric Dumazet
7 siblings, 1 reply; 19+ messages in thread
From: Kuniyuki Iwashima @ 2026-11-17 5:35 UTC (permalink / raw)
To: Eric Dumazet, Neal Cardwell, David S . Miller, Jakub Kicinski,
Paolo Abeni
Cc: Simon Horman, Kuniyuki Iwashima, Kuniyuki Iwashima, netdev
tcp_v6_err() passes skb->dev->ifindex to __inet6_lookup_established().
When ICMP is sent to a local address not on loopback device,
ip6_rt_get_dev_rcu() sets dst->dev to lo and ip6_rt_copy_init()
sets rt6i_idev to the actual device.
In ip6_rcv_core(), skb->dev is still lo, but IP6CB(skb)->iif
is set to ip6_dst_idev(skb_dst(skb))->dev->ifindex.
So skb->dev->ifindex is 1 while inet6_iif(skb) points to the
actual device.
Since tcp_v6_err() passes the wrong ifindex, such an ICMP error
is not delivered.
For example, this script creates a socket bound to tun0 and
generates ICMPv6 Port Unreachable to it, but the socket cannot
receive it and getsockopt() fails:
--ip_version=ipv6
0 `ip link set lo up`
+0 `ip6tables -A OUTPUT -p tcp -j REJECT`
+0 socket(..., SOCK_STREAM|SOCK_NONBLOCK, IPPROTO_TCP) = 3
+0 setsockopt(3, SOL_SOCKET, SO_BINDTODEVICE, "tun0", 4) = 0
+0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
+0 getsockopt(3, SOL_SOCKET, SO_ERROR, [ECONNREFUSED], [4]) = 0
Let's pass inet6_iif() to __inet6_lookup_established().
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
net/ipv6/tcp_ipv6.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index 29fd0a084c91..ab0eff884d3d 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -419,7 +419,7 @@ static int tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
sk = __inet6_lookup_established(net, &hdr->daddr, th->dest,
&hdr->saddr, ntohs(th->source),
- skb->dev->ifindex, inet6_sdif(skb));
+ inet6_iif(skb), inet6_sdif(skb));
if (!sk) {
__ICMP6_INC_STATS(net, __in6_dev_get(skb->dev),
--
2.56.0.679.gc527c65306-goog
^ permalink raw reply related [flat|nested] 19+ messages in thread
* Re: [PATCH v1 net 1/8] tcp: Don't accept ICMP for SND.NXT in TCP_SYN_SENT.
2026-11-17 5:35 ` [PATCH v1 net 1/8] tcp: Don't accept ICMP for SND.NXT in TCP_SYN_SENT Kuniyuki Iwashima
@ 2026-11-17 8:35 ` Eric Dumazet
2026-11-17 23:30 ` Kuniyuki Iwashima
0 siblings, 1 reply; 19+ messages in thread
From: Eric Dumazet @ 2026-11-17 8:35 UTC (permalink / raw)
To: Kuniyuki Iwashima
Cc: Neal Cardwell, David S . Miller, Jakub Kicinski, Paolo Abeni,
Simon Horman, Kuniyuki Iwashima, netdev
Le mar. 17 nov. 2026 à 06:35, Kuniyuki Iwashima <kuniyu@google.com> a écrit :
>
> RFC 5927 states that the acceptable ICMP has seq in
> SND.UNA =< SEG.SEQ < SND.NXT:
>
> These implementations check that the TCP sequence number
> contained in the payload of the ICMP error message is within
> the range SND.UNA =< SEG.SEQ < SND.NXT.
>
> tcp_v4_err() and tcp_v6_err() accept packets if seq meets
>
> between(seq, snd_una, tp->snd_nxt)
> -> tp->snd_nxt - snd_una >= seq - snd_una
>
> , which includes both ends.
>
> This is fine once connection is established because
> tcp_acceptable_seq() uses SND_NXT for ACK and RST.
>
> However, we should not accept ICMP to SND_NXT for SYN_SENT.
>
> The minimal repro is:
>
> --ip_version=ipv4
> 0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
> +0 fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
> +0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
> +0 > S 0:0(0) <...>
> +0 < icmp unreachable host_unreachable [1:1(0)]
> +0 %{assert tcpi_state == TCP_SYN_SENT}%
>
> Let's fix off-by-one.
>
> Note that we need to check seq != snd_una because tp->snd_nxt
> is only advanced to tp->write_seq after tcp_transmit_skb().
> Otherwise, ICMP for SYN could be dropped before tcp_connect()
> advances tp->snd_nxt.
This hints at a separate issue.
An ICMP can only be processed in this window if it was generated
locally and looped back while tcp_transmit_skb() is running
(netfilter REJECT in OUTPUT for instance).
At this point connect() owns the socket, so we end up in :
if (!sock_owned_by_user(sk))
tcp_done_with_error(sk, err);
else
WRITE_ONCE(sk->sk_err_soft, err);
The socket stays in TCP_SYN_SENT, SYN is retransmitted until
tcp_write_err(), and a non blocking user gets no EPOLLERR.
I think the repro in patch 8 only passes because SO_ERROR falls back
to sk_err_soft. Could you check tcpi_state there ?
Also, given an attacker has to guess the ISN anyway, I am not sure
this one is net material.
>
> Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
> ---
> net/ipv4/tcp_ipv4.c | 9 ++++++---
> net/ipv6/tcp_ipv6.c | 9 ++++++---
> 2 files changed, 12 insertions(+), 6 deletions(-)
>
> diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
> index 2fc92a7884fc..7d58f522068d 100644
> --- a/net/ipv4/tcp_ipv4.c
> +++ b/net/ipv4/tcp_ipv4.c
> @@ -499,8 +499,8 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
> const int type = icmp_hdr(skb)->type;
> const int code = icmp_hdr(skb)->code;
> struct request_sock *fastopen;
> + u32 seq, snd_una, snd_nxt;
> struct tcp_sock *tp;
> - u32 seq, snd_una;
> struct sock *sk;
> int err;
>
> @@ -555,9 +555,12 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
> tp = tcp_sk(sk);
> /* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */
> fastopen = rcu_dereference(tp->fastopen_rsk);
> - snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : tp->snd_una;
> + snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : READ_ONCE(tp->snd_una);
> + snd_nxt = READ_ONCE(tp->snd_nxt);
> if (sk->sk_state != TCP_LISTEN &&
> - !between(seq, snd_una, tp->snd_nxt)) {
> + (!between(seq, snd_una, snd_nxt) ||
> + (sk->sk_state == TCP_SYN_SENT &&
> + seq != snd_una && seq == snd_nxt))) {
In TCP_SYN_SENT the only sequence we ever put on the wire is the ISN,
including the Fast Open case (SYN+data is sent with seq == ISN, and the
RTO retransmits the bare SYN).
So this could simply be :
if (sk->sk_state != TCP_LISTEN &&
(!between(seq, snd_una, snd_nxt) ||
(sk->sk_state == TCP_SYN_SENT && seq != snd_una))) {
This mirrors the (seq != tcp_rsk(req)->snt_isn) test in tcp_req_err(),
and does not need the special case for the window where snd_nxt has
not been advanced yet.
> __NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
> goto out;
> }
> diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
> index 5c7ed38f1f56..8691d220cf94 100644
> --- a/net/ipv6/tcp_ipv6.c
> +++ b/net/ipv6/tcp_ipv6.c
> @@ -410,9 +410,9 @@ static int tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
> const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
> struct net *net = dev_net_rcu(skb->dev);
> struct request_sock *fastopen;
> + __u32 seq, snd_una, snd_nxt;
> struct ipv6_pinfo *np;
> struct tcp_sock *tp;
> - __u32 seq, snd_una;
> struct sock *sk;
> bool fatal;
> int err;
> @@ -463,9 +463,12 @@ static int tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
> tp = tcp_sk(sk);
> /* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */
> fastopen = rcu_dereference(tp->fastopen_rsk);
> - snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : tp->snd_una;
> + snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : READ_ONCE(tp->snd_una);
> + snd_nxt = READ_ONCE(tp->snd_nxt);
> if (sk->sk_state != TCP_LISTEN &&
> - !between(seq, snd_una, tp->snd_nxt)) {
> + (!between(seq, snd_una, snd_nxt) ||
> + (sk->sk_state == TCP_SYN_SENT &&
> + seq != snd_una && seq == snd_nxt))) {
> __NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
> goto out;
> }
> --
> 2.56.0.679.gc527c65306-goog
>
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH v1 net 2/8] tcp: Don't use inner IP TTL for IP_MINTTL in tcp_v4_err().
2026-11-17 5:35 ` [PATCH v1 net 2/8] tcp: Don't use inner IP TTL for IP_MINTTL in tcp_v4_err() Kuniyuki Iwashima
@ 2026-11-17 8:42 ` Eric Dumazet
2026-11-17 23:35 ` Kuniyuki Iwashima
0 siblings, 1 reply; 19+ messages in thread
From: Eric Dumazet @ 2026-11-17 8:42 UTC (permalink / raw)
To: Kuniyuki Iwashima
Cc: Neal Cardwell, David S . Miller, Jakub Kicinski, Paolo Abeni,
Simon Horman, Kuniyuki Iwashima, netdev, Stephen Hemminger,
stephen hemminger
Le mar. 17 nov. 2026 à 06:35, Kuniyuki Iwashima <kuniyu@google.com> a écrit :
>
> RFC 5082 introduced Generalized TTL Security Mechanism (GTSM),
> and IP_MINTTL and IPV6_MINHOPCOUNT follow that.
>
> If an incoming IPv4 packet has TTL less than inet_sk(sk)->min_ttl
> configured via IP_MINTTL, the packet is dropped.
>
> However, tcp_v4_err() checks the TTL extracted from the ICMP
> payload, not the one in the outer IP header.
>
> The minimal repro is:
>
> --ip_version=ipv4
> 0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
> +0 setsockopt(3, SOL_IP, 21, [250], 4) = 0
> +0 fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
> +0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
> +0 > S 0:0(0) <...>
> +0 < (ttl 10) icmp unreachable host_unreachable [0:0(0)]
> +0 %{assert tcpi_state == TCP_SYN_SENT}%
>
> Let's check ip_hdr(skb)->ttl.
>
> Now GTSM is applied to locally generated ICMP errors, but this
> behaviour matches tcp_v6_err().
>
> Note that tcp_v6_err() properly checks ipv6_hdr(skb)->hop_limit.
>
> Fixes: 97e3ecd112ba ("TCP: check min TTL on received ICMP packets")
> Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
> ---
> Cc: Stephen Hemminger <stephen@networkplumber.org>
> Cc: stephen hemminger <shemminger@vyatta.com>
> ---
> net/ipv4/tcp_ipv4.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
> index 7d58f522068d..bf8542368032 100644
> --- a/net/ipv4/tcp_ipv4.c
> +++ b/net/ipv4/tcp_ipv4.c
> @@ -546,7 +546,7 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
>
> if (static_branch_unlikely(&ip4_min_ttl)) {
> /* min_ttl can be changed concurrently from do_ip_setsockopt() */
> - if (unlikely(iph->ttl < READ_ONCE(inet_sk(sk)->min_ttl))) {
> + if (unlikely(ip_hdr(skb)->ttl < READ_ONCE(inet_sk(sk)->min_ttl))) {
> __NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
> goto out;
> }
Linux sends ICMP errors with the default TTL (64), so locally
generated errors are now dropped for any min_ttl above 64.
For the strict GTSM case (min_ttl == 255) there is no router between
the peers, so nothing else changes.
But IP_MINTTL is also used with lower values for multi-hop sessions
("ttl-security hops N"). There, an ICMP from an intermediate router
using an initial TTL of 64 was accepted before this patch (the quoted
TTL is the one we sent, minus the hops to that router), and is dropped
after.
I think this is the right thing to do, since the inner TTL is under
the sender's control, but I would mention this in the changelog.
Reviewed-by: Eric Dumazet <edumazet@kernel.org>
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH v1 net 3/8] tcp: Don't bypass GTSM check for reqsk.
2026-11-17 5:35 ` [PATCH v1 net 3/8] tcp: Don't bypass GTSM check for reqsk Kuniyuki Iwashima
@ 2026-11-17 8:45 ` Eric Dumazet
0 siblings, 0 replies; 19+ messages in thread
From: Eric Dumazet @ 2026-11-17 8:45 UTC (permalink / raw)
To: Kuniyuki Iwashima
Cc: Neal Cardwell, David S . Miller, Jakub Kicinski, Paolo Abeni,
Simon Horman, Kuniyuki Iwashima, netdev
Le mar. 17 nov. 2026 à 06:35, Kuniyuki Iwashima <kuniyu@google.com> a écrit :
>
> Since cited commits moved up TCP_NEW_SYN_RECV handling in
> tcp_v4_err() and tcp_v6_err(), ICMP messages bypass IP_MINTTL
> and IPV6_MINHOPCOUNT checks.
>
> The minimal repro is:
>
> --ip_version=ipv6
> 0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
> +0 setsockopt(3, SOL_IPV6, 73, [255], 4) = 0
> +0 bind(3, ..., ...) = 0
> +0 listen(3, 10) = 0
> +0 < (ttl 255) S 0:0(0) win 65535 <mss 1440,nop,nop,sackOK>
> +0 > S. 0:0(0) ack 1 <mss 1440,nop,nop,sackOK>
> +0 < (ttl 1) icmp unreachable port_unreachable [0:0(0)]
> +0 < (ttl 255) . 1:1(0) ack 1 win 65535
> +0 fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
> +0 accept(3, ..., ...) = 4
> +0 %{assert tcpi_state == TCP_ESTABLISHED}%
>
> Let's restore IP_MINTTL/IPV6_MINHOPCOUNT for reqsk.
>
> Fixes: 26e3736090e1 ("ipv4: tcp: handle ICMP messages on TCP_NEW_SYN_RECV request sockets")
> Fixes: 2215089b2244 ("ipv6: tcp: handle ICMP messages on TCP_NEW_SYN_RECV request sockets")
> Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
> ---
Note the same bypass exists in the receive path : in tcp_v4_rcv() and
tcp_v6_rcv(), the TCP_NEW_SYN_RECV branch calls tcp_check_req() /
tcp_child_process() and returns before reaching the min_ttl /
min_hopcount check done after the process: label.
Reviewed-by: Eric Dumazet <edumazet@kernel.org>
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH v1 net 4/8] tcp: Fix reqsk abort condition on ICMP errors.
2026-11-17 5:35 ` [PATCH v1 net 4/8] tcp: Fix reqsk abort condition on ICMP errors Kuniyuki Iwashima
@ 2026-11-17 8:49 ` Eric Dumazet
0 siblings, 0 replies; 19+ messages in thread
From: Eric Dumazet @ 2026-11-17 8:49 UTC (permalink / raw)
To: Kuniyuki Iwashima
Cc: Neal Cardwell, David S . Miller, Jakub Kicinski, Paolo Abeni,
Simon Horman, Kuniyuki Iwashima, netdev
Le mar. 17 nov. 2026 à 06:35, Kuniyuki Iwashima <kuniyu@google.com> a écrit :
>
> RFC 1122 Section 4.2.3.9 states the following soft errors
> must not abort connections:
>
Linux has never followed this for non-synchronized states.
tcp_v4_err() and tcp_v6_err() call tcp_done_with_error() for any error
in TCP_SYN_SENT and TCP_SYN_RECV, soft or hard. RFC 5461 section 4.1
documents this, and names Linux.
Tests 1 and 2 in patch 6 actually depend on it : host_unreachable
aborts the TCP_SYN_SENT socket.
> o Destination Unreachable -- codes 0, 1, 5
>
> Since these Unreachable messages indicate soft error
> conditions, TCP MUST NOT abort the connection, and it
> SHOULD make the information available to the
> application.
>
> o Time Exceeded -- codes 0, 1
> o Parameter Problem
>
> This should be handled the same way as Destination
> Unreachable codes 0, 1, 5 (see above).
>
> Commit 9cf7490360bf ("tcp: do not drop syn_recv on all icmp
> reports") introduced the abort arg to tcp_req_err() to avoid
> aborting connection on ICMP_REDIRECT.
>
> However, tcp_v4_err() still abort on soft errors (ICMP_NET_UNREACH
> (0), ICMP_HOST_UNREACH (1)) and ignore hard errors.
>
> For example, we can see SYN+ACK retransmitted after receiving
> hard error:
>
> --ip_version=ipv4
> 0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
> +0 bind(3, ..., ...) = 0
> +0 listen(3, 10) = 0
>
> +0 < S 0:0(0) win 65535 <mss 1460,nop,nop,sackOK>
> +0 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK>
> +0 < icmp unreachable port_unreachable [0:0(0)]
> +1 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK>
>
> Also, tcp_v6_err() uses icmpv6_err_convert(), which handles
> ICMPV6_PARAMPROB messages as hard error while RFC 5461 defines
> as soft error.
>
> But we should handle ICMPV6_UNK_NEXTHDR as hard error because
> it is used as IPv6 version ICMP_PROT_UNREACH.
>
> Let's correct the abort condition in tcp_v4_err() and
> tcp_v6_err().
>
> Note that icmpv6_err_convert() is used by other protocols.
>
> Fixes: 9cf7490360bf ("tcp: do not drop syn_recv on all icmp reports")
> Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
> ---
> net/ipv4/tcp_ipv4.c | 7 ++-----
> net/ipv6/tcp_ipv6.c | 3 ++-
> 2 files changed, 4 insertions(+), 6 deletions(-)
>
> diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
> index fd9588bba967..f11e090d04fb 100644
> --- a/net/ipv4/tcp_ipv4.c
> +++ b/net/ipv4/tcp_ipv4.c
> @@ -529,11 +529,8 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
>
> seq = ntohl(th->seq);
> if (sk->sk_state == TCP_NEW_SYN_RECV) {
> - tcp_req_err(sk, seq, type == ICMP_PARAMETERPROB ||
> - type == ICMP_TIME_EXCEEDED ||
> - (type == ICMP_DEST_UNREACH &&
> - (code == ICMP_NET_UNREACH ||
> - code == ICMP_HOST_UNREACH)));
> + tcp_req_err(sk, seq, (type == ICMP_DEST_UNREACH &&
> + icmp_err_convert[code].fatal));
After this patch, a TCP_SYN_SENT socket and a Fast Open TCP_SYN_RECV
child abort on net/host unreachable, but a request socket does not.
We also lose the ability to get rid of a request for an unreachable
client, and keep retransmitting SYNACK instead.
If the problem is that hard errors are ignored for IPv4 request
sockets, I would rather add them than remove the soft ones :
abort = type == ICMP_PARAMETERPROB ||
type == ICMP_TIME_EXCEEDED ||
(type == ICMP_DEST_UNREACH && code != ICMP_FRAG_NEEDED);
This matches what we do for TCP_SYN_SENT. The IPv6 side would then
need the same treatment instead of using 'fatal'.
> return 0;
> }
>
> diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
> index 0f1523fff939..29fd0a084c91 100644
> --- a/net/ipv6/tcp_ipv6.c
> +++ b/net/ipv6/tcp_ipv6.c
> @@ -447,7 +447,8 @@ static int tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
> seq = ntohl(th->seq);
> fatal = icmpv6_err_convert(type, code, &err);
> if (sk->sk_state == TCP_NEW_SYN_RECV) {
> - tcp_req_err(sk, seq, fatal);
> + tcp_req_err(sk, seq, type == ICMPV6_PARAMPROB ?
> + code == ICMPV6_UNK_NEXTHDR : fatal);
> return 0;
> }
>
> --
> 2.56.0.679.gc527c65306-goog
>
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH v1 net 5/8] tcp: Check ip_sk_accept_pmtu() in tcp_v4_err().
2026-11-17 5:35 ` [PATCH v1 net 5/8] tcp: Check ip_sk_accept_pmtu() in tcp_v4_err() Kuniyuki Iwashima
@ 2026-11-17 8:54 ` Eric Dumazet
0 siblings, 0 replies; 19+ messages in thread
From: Eric Dumazet @ 2026-11-17 8:54 UTC (permalink / raw)
To: Kuniyuki Iwashima
Cc: Neal Cardwell, David S . Miller, Jakub Kicinski, Paolo Abeni,
Simon Horman, Kuniyuki Iwashima, netdev, Hannes Frederic Sowa
Le mar. 17 nov. 2026 à 06:35, Kuniyuki Iwashima <kuniyu@google.com> a écrit :
>
> Currently, tcp_v4_mtu_reduced() checks ip_sk_accept_pmtu()
> after calling inet_csk_update_pmtu(), which is too late.
>
> Even if we configure a socket not to perform Path MTU
> Discovery with IP_PMTUDISC_INTERFACE, ICMP_FRAG_NEEDED
> can overwrite MTU.
>
> The last getsockopt() returns 1200 MTU:
>
> --ip_version=ipv4
> 0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
> +0 bind(3, ..., ...) = 0
> +0 listen(3, 1) = 0
> +0 < S 0:0(0) win 65535 <mss 1460,sackOK,nop,nop,nop,wscale 8>
> +0 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 9>
> +0 < . 1:1(0) ack 1 win 257
> +0 accept(3, ..., ...) = 4
> +0 setsockopt(4, SOL_IP, IP_MTU_DISCOVER, [4], 4) = 0
> +0 getsockopt(4, SOL_IP, IP_MTU, [1500], [4]) = 0
> +0 write(4, ..., 1000) = 1000
> +0 > P. 1:1001(1000) ack 1
> +0 < icmp unreachable frag_needed mtu 1200 [1:1001(1000)]
> +0 getsockopt(4, SOL_IP, IP_MTU, [1500], [4]) = 0
>
> Let's check ip_sk_accept_pmtu() in tcp_v4_err() as done
> in tcp_v6_err().
>
> Note that inet_sk(sk)->pmtudisc is configured locklessly,
> so we do not recheck it in tcp_v4_mtu_reduced(). This is
> same with tcp_v6_mtu_reduced() too.
>
> Fixes: 482fc6094afa ("ipv4: introduce new IP_MTU_DISCOVER mode IP_PMTUDISC_INTERFACE")
> Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Reviewed-by: Eric Dumazet <edumazet@kernel.org>
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH v1 net 6/8] selftest: net: packetdrill: Add ICMP and ICMPv6 tests.
2026-11-17 5:35 ` [PATCH v1 net 6/8] selftest: net: packetdrill: Add ICMP and ICMPv6 tests Kuniyuki Iwashima
@ 2026-11-17 8:58 ` Eric Dumazet
0 siblings, 0 replies; 19+ messages in thread
From: Eric Dumazet @ 2026-11-17 8:58 UTC (permalink / raw)
To: Kuniyuki Iwashima
Cc: Neal Cardwell, David S . Miller, Jakub Kicinski, Paolo Abeni,
Simon Horman, Kuniyuki Iwashima, netdev
Le mar. 17 nov. 2026 à 06:35, Kuniyuki Iwashima <kuniyu@google.com> a écrit :
>
> tcp_icmp_ipv4.pkt and tcp_icmp_ipv6.pkt cover the scenarios
> fixed by the preceding patches.
>
> tcp_v6_err() already checks the outer hop limit and
> ip6_sk_accept_pmtu(), so Test 2 and 5 in tcp_icmp_ipv6.pkt
> pass without the series.
>
> ICMP and ICMPv6 behaviour are unlikely to change, so splitting
> into multiple files per test case is not worth it.
>
> Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
> ---
> .../net/packetdrill/tcp_icmp_ipv4.pkt | 98 ++++++++++++++++
> .../net/packetdrill/tcp_icmp_ipv6.pkt | 111 ++++++++++++++++++
> 2 files changed, 209 insertions(+)
> create mode 100644 tools/testing/selftests/net/packetdrill/tcp_icmp_ipv4.pkt
> create mode 100644 tools/testing/selftests/net/packetdrill/tcp_icmp_ipv6.pkt
>
> diff --git a/tools/testing/selftests/net/packetdrill/tcp_icmp_ipv4.pkt b/tools/testing/selftests/net/packetdrill/tcp_icmp_ipv4.pkt
> new file mode 100644
> index 000000000000..122bf9a14862
> --- /dev/null
> +++ b/tools/testing/selftests/net/packetdrill/tcp_icmp_ipv4.pkt
> @@ -0,0 +1,98 @@
> +// SPDX-License-Identifier: GPL-2.0
> +--ip_version=ipv4
> +
> +`./defaults.sh`
> +
> +// Test 1: Client in TCP_SYN_SENT rejects ICMP quoting seq == snd_nxt (RFC 5927 4.1)
> +// In SYN_SENT, SND.UNA = ISN, SND.NXT = ISN + 1.
> +// The sequence number in ICMP must satisfy SND.UNA <= SEG.SEQ < SND.NXT.
> +// An ICMP message quoting seq == 1 (snd_nxt) must be dropped, while seq == 0 (snd_una) aborts.
> + 0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
> + +0 fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
> + +0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
> + +0 > S 0:0(0) <...>
> + // Out-of-window ICMP quoting seq == 1 (snd_nxt) must be dropped
> + +0 < icmp unreachable host_unreachable [1:1(0)]
> + // Socket stays in TCP_SYN_SENT, so SYN retransmission occurs after 1s
> + +1 > S 0:0(0) <...>
> + // In-window ICMP quoting seq == 0 (snd_una) is accepted and aborts connection
> + +0 < icmp unreachable host_unreachable [0:0(0)]
> + +0 %{ assert tcpi_state == TCP_CLOSE }%
> + +0 getsockopt(3, SOL_SOCKET, SO_ERROR, [EHOSTUNREACH], [4]) = 0
> + +0 close(3) = 0
> +
> +// Test 2: Client with IP_MINTTL (21) checks outer IP TTL in tcp_v4_err()
> +// An ICMP with outer TTL < min_ttl (250) must be dropped, while outer TTL >= min_ttl
> +// must be accepted even though the quoted inner IP header has lower TTL (64).
> + +0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
> + +0 setsockopt(3, SOL_IP, 21, [250], 4) = 0
Ah.... could we instead update packedrill, adding IP_MINTTL ?
Same for IPV6_MINHOPCOUNT
> + +0 fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
> + +0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
> + +0 > S 0:0(0) <...>
> + // Outer TTL 10 < min_ttl (250): dropped, socket stays in TCP_SYN_SENT
> + +0 < (ttl 10) icmp unreachable host_unreachable [0:0(0)]
> + +1 > S 0:0(0) <...>
> + // Outer TTL 255 >= min_ttl (250): accepted, socket aborts
> + +0 < (ttl 255) icmp unreachable host_unreachable [0:0(0)]
> + +0 %{ assert tcpi_state == TCP_CLOSE }%
> + +0 getsockopt(3, SOL_SOCKET, SO_ERROR, [EHOSTUNREACH], [4]) = 0
> + +0 close(3) = 0
> +
> +// Test 3: Listener with IP_MINTTL (21) checks TTL before reqsk handling
> +// An incoming fatal ICMP with outer TTL < min_ttl (255) must not abort TCP_NEW_SYN_RECV,
> +// while outer TTL >= min_ttl (255) must abort it.
> + +0 socket(..., SOCK_STREAM|SOCK_NONBLOCK, IPPROTO_TCP) = 3
> + +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
> + +0 setsockopt(3, SOL_IP, 21, [255], 4) = 0
> + +0 bind(3, ..., ...) = 0
> + +0 listen(3, 10) = 0
> + +0 < (ttl 255) S 0:0(0) win 65535 <mss 1460,nop,nop,sackOK>
> + +0 > (ttl 64) S. 0:0(0) ack 1 <...>
> + // Outer TTL 10 < min_ttl (255): dropped, reqsk survives
> + +0 < (ttl 10) icmp unreachable port_unreachable [0:0(0)]
> + +1 > (ttl 64) S. 0:0(0) ack 1 <...>
> + // Outer TTL 255 >= min_ttl (255): accepted, fatal ICMP aborts reqsk
> + +0 < (ttl 255) icmp unreachable port_unreachable [0:0(0)]
> + +0 < (ttl 255) . 1:1(0) ack 1 win 65535
> + +0 > R 1:1(0)
> + +0 close(3) = 0
> +
> +// Test 4: Server in TCP_NEW_SYN_RECV does not abort on soft errors but aborts on fatal error
> + +0 socket(..., SOCK_STREAM|SOCK_NONBLOCK, IPPROTO_TCP) = 3
> + +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
> + +0 bind(3, ..., ...) = 0
> + +0 listen(3, 10) = 0
> + +0 < S 0:0(0) win 65535 <mss 1460,nop,nop,sackOK>
> + +0 > S. 0:0(0) ack 1 <...>
> + // 4a: Soft errors must not abort reqsk
> + +0 < icmp unreachable host_unreachable [0:0(0)]
> + +0 < icmp unreachable net_unreachable [0:0(0)]
> + +0 < icmp time_exceeded [0:0(0)]
> + +0 < icmp parameter_problem [0:0(0)]
> + +1 > S. 0:0(0) ack 1 <...>
> + // 4b: Fatal error (port_unreachable) must abort reqsk
> + +0 < icmp unreachable port_unreachable [0:0(0)]
> + // Remote sends 3rd ACK: listener responds with RST because reqsk was aborted
> + +0 < . 1:1(0) ack 1 win 65535
> + +0 > R 1:1(0)
> + +0 close(3) = 0
> +
> +// Test 5: Socket with IP_PMTUDISC_INTERFACE ignores ICMP_FRAG_NEEDED
> +// Socket configured with IP_PMTUDISC_INTERFACE (4) must not overwrite MTU
> +// upon receiving ICMP_DEST_UNREACH with code ICMP_FRAG_NEEDED.
> + +0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
> + +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
> + +0 bind(3, ..., ...) = 0
> + +0 listen(3, 1) = 0
> + +0 < S 0:0(0) win 65535 <mss 1460,sackOK,nop,nop,nop,wscale 8>
> + +0 > S. 0:0(0) ack 1 <...>
> + +0 < . 1:1(0) ack 1 win 257
> + +0 accept(3, ..., ...) = 4
> + +0 setsockopt(4, SOL_IP, IP_MTU_DISCOVER, [4], 4) = 0
> + +0 getsockopt(4, SOL_IP, IP_MTU, [1500], [4]) = 0
> + +0 write(4, ..., 1000) = 1000
> + +0 > P. 1:1001(1000) ack 1 <...>
> + +0 < icmp unreachable frag_needed mtu 1200 [1:1001(1000)]
> + +0 getsockopt(4, SOL_IP, IP_MTU, [1500], [4]) = 0
> + +0 close(4) = 0
> + +0 close(3) = 0
> diff --git a/tools/testing/selftests/net/packetdrill/tcp_icmp_ipv6.pkt b/tools/testing/selftests/net/packetdrill/tcp_icmp_ipv6.pkt
> new file mode 100644
> index 000000000000..bd0cc365a088
> --- /dev/null
> +++ b/tools/testing/selftests/net/packetdrill/tcp_icmp_ipv6.pkt
> @@ -0,0 +1,111 @@
> +// SPDX-License-Identifier: GPL-2.0
> +--ip_version=ipv6
> +
> +`./defaults.sh`
> +
> +// Test 1: Client in TCP_SYN_SENT rejects ICMPv6 quoting seq == snd_nxt (RFC 5927 4.1)
> +// In SYN_SENT, SND.UNA = ISN, SND.NXT = ISN + 1.
> +// The sequence number in ICMP must satisfy SND.UNA <= SEG.SEQ < SND.NXT.
> +// An ICMP message quoting seq == 1 (snd_nxt) must be dropped, while seq == 0 (snd_una) aborts.
> + 0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
> + +0 fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
> + +0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
> + +0 > S 0:0(0) <...>
> + // Out-of-window ICMPv6 quoting seq == 1 (snd_nxt) must be dropped
> + +0 < icmp unreachable address_unreachable [1:1(0)]
> + // Socket stays in TCP_SYN_SENT, so SYN retransmission occurs after 1s
> + +1 > S 0:0(0) <...>
> + // In-window ICMPv6 quoting seq == 0 (snd_una) is accepted and aborts connection
> + +0 < icmp unreachable address_unreachable [0:0(0)]
> + +0 %{ assert tcpi_state == TCP_CLOSE }%
> + +0 getsockopt(3, SOL_SOCKET, SO_ERROR, [EHOSTUNREACH], [4]) = 0
> + +0 close(3) = 0
> +
> +// Test 2: Client with IPV6_MINHOPCOUNT (73) checks outer IPv6 hop_limit in tcp_v6_err()
> +// An ICMPv6 with outer hop_limit < min_hopcount (250) must be dropped, while outer
> +// hop_limit >= min_hopcount must be accepted even though the quoted inner hop_limit is 64.
> + +0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
> + +0 setsockopt(3, SOL_IPV6, 73, [250], 4) = 0
> + +0 fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
> + +0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
> + +0 > S 0:0(0) <...>
> + // Outer hop_limit 10 < min_hopcount (250): dropped, socket stays in TCP_SYN_SENT
> + +0 < (ttl 10) icmp unreachable address_unreachable [0:0(0)]
> + +1 > S 0:0(0) <...>
> + // Outer hop_limit 255 >= min_hopcount (250): accepted, socket aborts
> + +0 < (ttl 255) icmp unreachable address_unreachable [0:0(0)]
> + +0 %{ assert tcpi_state == TCP_CLOSE }%
> + +0 getsockopt(3, SOL_SOCKET, SO_ERROR, [EHOSTUNREACH], [4]) = 0
> + +0 close(3) = 0
> +
> +// Test 3: Listener with IPV6_MINHOPCOUNT (73) checks hop_limit before reqsk handling
> +// An incoming fatal ICMPv6 with outer hop_limit < min_hopcount (255) must not abort
> +// TCP_NEW_SYN_RECV, while outer hop_limit >= min_hopcount (255) must abort it.
> + +0 socket(..., SOCK_STREAM|SOCK_NONBLOCK, IPPROTO_TCP) = 3
> + +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
> + +0 setsockopt(3, SOL_IPV6, 73, [255], 4) = 0
> + +0 bind(3, ..., ...) = 0
> + +0 listen(3, 10) = 0
> + +0 < (ttl 255) S 0:0(0) win 65535 <mss 1440,nop,nop,sackOK>
> + +0 > (ttl 64) S. 0:0(0) ack 1 <...>
> + // Outer hop_limit 10 < min_hopcount (255): dropped, reqsk survives
> + +0 < (ttl 10) icmp unreachable port_unreachable [0:0(0)]
> + +1 > (ttl 64) S. 0:0(0) ack 1 <...>
> + // Outer hop_limit 255 >= min_hopcount (255): accepted, fatal ICMPv6 aborts reqsk
> + +0 < (ttl 255) icmp unreachable port_unreachable [0:0(0)]
> + +0 < (ttl 255) . 1:1(0) ack 1 win 65535
> + +0 > R 1:1(0)
> + +0 close(3) = 0
> +
> +// Test 4: Server in TCP_NEW_SYN_RECV does not abort on soft error or
> +// non-UNK_NEXTHDR ICMPV6_PARAMPROB, but aborts on fatal error or
> +// ICMPV6_UNK_NEXTHDR.
> + +0 socket(..., SOCK_STREAM|SOCK_NONBLOCK, IPPROTO_TCP) = 3
> + +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
> + +0 bind(3, ..., ...) = 0
> + +0 listen(3, 10) = 0
> + +0 < S 0:0(0) win 65535 <mss 1440,nop,nop,sackOK>
> + +0 > S. 0:0(0) ack 1 <...>
> + // 4a: Soft error (address_unreachable) must not abort reqsk
> + +0 < icmp unreachable address_unreachable [0:0(0)]
> + +1 > S. 0:0(0) ack 1 <...>
> + // 4b: Non-UNK_NEXTHDR parameter problem (header_field) must not abort reqsk
> + +0 < icmp parameter_problem header_field [0:0(0)]
> + +2 > S. 0:0(0) ack 1 <...>
> + // 4c: ICMPV6_UNK_NEXTHDR parameter problem (unknown_next_header) must abort reqsk
> + +0 < icmp parameter_problem unknown_next_header [0:0(0)]
> + // Remote sends 3rd ACK: listener responds with RST because reqsk was aborted
> + +0 < . 1:1(0) ack 1 win 65535
> + +0 > R 1:1(0)
> + +0 close(3) = 0
> + // 4d: Fatal unreachable error (port_unreachable) must abort reqsk
> + +0 socket(..., SOCK_STREAM|SOCK_NONBLOCK, IPPROTO_TCP) = 3
> + +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
> + +0 bind(3, ..., ...) = 0
> + +0 listen(3, 10) = 0
> + +0 < S 0:0(0) win 65535 <mss 1440,nop,nop,sackOK>
> + +0 > S. 0:0(0) ack 1 <...>
> + +0 < icmp unreachable port_unreachable [0:0(0)]
> + +0 < . 1:1(0) ack 1 win 65535
> + +0 > R 1:1(0)
> + +0 close(3) = 0
> +
> +// Test 5: Socket with IPV6_PMTUDISC_INTERFACE ignores ICMPV6_PKT_TOOBIG
> +// Socket configured with IPV6_PMTUDISC_INTERFACE (4) must not overwrite MTU
> +// upon receiving ICMPV6_PKT_TOOBIG.
> + +0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
> + +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
> + +0 bind(3, ..., ...) = 0
> + +0 listen(3, 1) = 0
> + +0 < S 0:0(0) win 65535 <mss 1440,sackOK,nop,nop,nop,wscale 8>
> + +0 > S. 0:0(0) ack 1 <...>
> + +0 < . 1:1(0) ack 1 win 257
> + +0 accept(3, ..., ...) = 4
> + +0 setsockopt(4, SOL_IPV6, IPV6_MTU_DISCOVER, [4], 4) = 0
> + +0 getsockopt(4, SOL_IPV6, IPV6_MTU, [1520], [4]) = 0
> + +0 write(4, ..., 1000) = 1000
> + +0 > P. 1:1001(1000) ack 1 <...>
> + +0 < icmp packet_too_big mtu 1280 [1:1001(1000)]
> + +0 getsockopt(4, SOL_IPV6, IPV6_MTU, [1520], [4]) = 0
> + +0 close(4) = 0
> + +0 close(3) = 0
> --
> 2.56.0.679.gc527c65306-goog
>
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH v1 net 7/8] tcp: Pass inet_sdif() to __inet_lookup_established() in tcp_v4_err().
2026-11-17 5:35 ` [PATCH v1 net 7/8] tcp: Pass inet_sdif() to __inet_lookup_established() in tcp_v4_err() Kuniyuki Iwashima
@ 2026-11-17 9:01 ` Eric Dumazet
0 siblings, 0 replies; 19+ messages in thread
From: Eric Dumazet @ 2026-11-17 9:01 UTC (permalink / raw)
To: Kuniyuki Iwashima
Cc: Neal Cardwell, David S . Miller, Jakub Kicinski, Paolo Abeni,
Simon Horman, Kuniyuki Iwashima, netdev, David Ahern, David Ahern
Le mar. 17 nov. 2026 à 06:35, Kuniyuki Iwashima <kuniyu@google.com> a écrit :
>
> tcp_v4_err() passes 0 sdif to __inet_lookup_established().
>
> So, ICMP error coming from VRF can be leaked to a socket not
> bound to the device.
>
> For example, this test fails at the last assert because the
> socket receives ICMP error:
>
> --ip_version=ipv4
> 0 socket(..., SOCK_STREAM|SOCK_NONBLOCK, IPPROTO_TCP) = 3
> +0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
> +0 > S 0:0(0) <...>
> +0 `ip link add vrf0 type vrf table 10`
> +0 `ip link set vrf0 up`
> +0 `ip link set dev tun0 master vrf0`
> +0 `ip route add 192.0.2.0/24 dev tun0 table 10`
> +0 < icmp unreachable port_unreachable [0:0(0)]
> +0 %{ assert tcpi_state == TCP_SYN_SENT, tcpi_state }%
>
> Let's pass inet_sdif() to __inet_lookup_established().
>
> Fixes: 3fa6f616a7a4 ("net: ipv4: add second dif to inet socket lookups")
> Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Reviewed-by: Eric Dumazet <edumazet@kernel.org>
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH v1 net 8/8] tcp: Pass inet6_iif() to __inet6_lookup_established() in tcp_v6_err().
2026-11-17 5:35 ` [PATCH v1 net 8/8] tcp: Pass inet6_iif() to __inet6_lookup_established() in tcp_v6_err() Kuniyuki Iwashima
@ 2026-11-17 9:03 ` Eric Dumazet
0 siblings, 0 replies; 19+ messages in thread
From: Eric Dumazet @ 2026-11-17 9:03 UTC (permalink / raw)
To: Kuniyuki Iwashima
Cc: Neal Cardwell, David S . Miller, Jakub Kicinski, Paolo Abeni,
Simon Horman, Kuniyuki Iwashima, netdev
Le mar. 17 nov. 2026 à 06:35, Kuniyuki Iwashima <kuniyu@google.com> a écrit :
>
> tcp_v6_err() passes skb->dev->ifindex to __inet6_lookup_established().
>
> When ICMP is sent to a local address not on loopback device,
> ip6_rt_get_dev_rcu() sets dst->dev to lo and ip6_rt_copy_init()
> sets rt6i_idev to the actual device.
>
> In ip6_rcv_core(), skb->dev is still lo, but IP6CB(skb)->iif
> is set to ip6_dst_idev(skb_dst(skb))->dev->ifindex.
>
> So skb->dev->ifindex is 1 while inet6_iif(skb) points to the
> actual device.
>
> Since tcp_v6_err() passes the wrong ifindex, such an ICMP error
> is not delivered.
>
> For example, this script creates a socket bound to tun0 and
> generates ICMPv6 Port Unreachable to it, but the socket cannot
> receive it and getsockopt() fails:
>
> --ip_version=ipv6
> 0 `ip link set lo up`
> +0 `ip6tables -A OUTPUT -p tcp -j REJECT`
> +0 socket(..., SOCK_STREAM|SOCK_NONBLOCK, IPPROTO_TCP) = 3
> +0 setsockopt(3, SOL_SOCKET, SO_BINDTODEVICE, "tun0", 4) = 0
> +0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
> +0 getsockopt(3, SOL_SOCKET, SO_ERROR, [ECONNREFUSED], [4]) = 0
>
> Let's pass inet6_iif() to __inet6_lookup_established().
>
> Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Reviewed-by: Eric Dumazet <edumazet@kernel.org>
I wonder if you could send (maybe later in net-next) packetdrill tests
for 7 and 8 ?
Thanks.
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH v1 net 1/8] tcp: Don't accept ICMP for SND.NXT in TCP_SYN_SENT.
2026-11-17 8:35 ` Eric Dumazet
@ 2026-11-17 23:30 ` Kuniyuki Iwashima
0 siblings, 0 replies; 19+ messages in thread
From: Kuniyuki Iwashima @ 2026-11-17 23:30 UTC (permalink / raw)
To: Eric Dumazet
Cc: Neal Cardwell, David S . Miller, Jakub Kicinski, Paolo Abeni,
Simon Horman, Kuniyuki Iwashima, netdev
On Tue, 17 Nov 2026 09:35:41 +0100, Eric Dumazet wrote:
> Le mar. 17 nov. 2026 à 06:35, Kuniyuki Iwashima <kuniyu@google.com> a écrit :
> >
> > RFC 5927 states that the acceptable ICMP has seq in
> > SND.UNA =< SEG.SEQ < SND.NXT:
> >
> > These implementations check that the TCP sequence number
> > contained in the payload of the ICMP error message is within
> > the range SND.UNA =< SEG.SEQ < SND.NXT.
> >
> > tcp_v4_err() and tcp_v6_err() accept packets if seq meets
> >
> > between(seq, snd_una, tp->snd_nxt)
> > -> tp->snd_nxt - snd_una >= seq - snd_una
> >
> > , which includes both ends.
> >
> > This is fine once connection is established because
> > tcp_acceptable_seq() uses SND_NXT for ACK and RST.
> >
> > However, we should not accept ICMP to SND_NXT for SYN_SENT.
> >
> > The minimal repro is:
> >
> > --ip_version=ipv4
> > 0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
> > +0 fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
> > +0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
> > +0 > S 0:0(0) <...>
> > +0 < icmp unreachable host_unreachable [1:1(0)]
> > +0 %{assert tcpi_state == TCP_SYN_SENT}%
> >
> > Let's fix off-by-one.
> >
> > Note that we need to check seq != snd_una because tp->snd_nxt
> > is only advanced to tp->write_seq after tcp_transmit_skb().
> > Otherwise, ICMP for SYN could be dropped before tcp_connect()
> > advances tp->snd_nxt.
>
> This hints at a separate issue.
>
> An ICMP can only be processed in this window if it was generated
> locally and looped back while tcp_transmit_skb() is running
> (netfilter REJECT in OUTPUT for instance).
>
> At this point connect() owns the socket, so we end up in :
>
> if (!sock_owned_by_user(sk))
> tcp_done_with_error(sk, err);
> else
> WRITE_ONCE(sk->sk_err_soft, err);
>
> The socket stays in TCP_SYN_SENT, SYN is retransmitted until
> tcp_write_err(), and a non blocking user gets no EPOLLERR.
>
> I think the repro in patch 8 only passes because SO_ERROR falls back
> to sk_err_soft. Could you check tcpi_state there ?
Ah yes, just after checking SO_ERROR, sk->sk_state is still at
TCP_SYN_SENT and it transitions to TCP_CLOSE only at SYN retrans.
--ip_version=ipv6
0 `ip link set lo up`
+0 `ip6tables -A OUTPUT -p tcp -j REJECT`
+0 socket(..., SOCK_STREAM|SOCK_NONBLOCK, IPPROTO_TCP) = 3
+0 setsockopt(3, SOL_SOCKET, SO_BINDTODEVICE, "tun0", 4) = 0
+0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
+0 getsockopt(3, SOL_SOCKET, SO_ERROR, [ECONNREFUSED], [4]) = 0
+0 %{ assert tcpi_state == TCP_SYN_SENT, tcpi_state }%
+1.2 %{ assert tcpi_state == TCP_CLOSE, tcpi_state }%
Maybe we could change TCP_MTU_REDUCED_DEFERRED to TCP_ICMP_DEFERRED
and have a new flag set for mtu etc in a hole in cold cache line ?
>
> Also, given an attacker has to guess the ISN anyway, I am not sure
> this one is net material.
Fair, I'll post this to net-next later.
>
> >
> > Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> > Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
> > ---
> > net/ipv4/tcp_ipv4.c | 9 ++++++---
> > net/ipv6/tcp_ipv6.c | 9 ++++++---
> > 2 files changed, 12 insertions(+), 6 deletions(-)
> >
> > diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
> > index 2fc92a7884fc..7d58f522068d 100644
> > --- a/net/ipv4/tcp_ipv4.c
> > +++ b/net/ipv4/tcp_ipv4.c
> > @@ -499,8 +499,8 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
> > const int type = icmp_hdr(skb)->type;
> > const int code = icmp_hdr(skb)->code;
> > struct request_sock *fastopen;
> > + u32 seq, snd_una, snd_nxt;
> > struct tcp_sock *tp;
> > - u32 seq, snd_una;
> > struct sock *sk;
> > int err;
> >
> > @@ -555,9 +555,12 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
> > tp = tcp_sk(sk);
> > /* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */
> > fastopen = rcu_dereference(tp->fastopen_rsk);
> > - snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : tp->snd_una;
> > + snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : READ_ONCE(tp->snd_una);
> > + snd_nxt = READ_ONCE(tp->snd_nxt);
> > if (sk->sk_state != TCP_LISTEN &&
> > - !between(seq, snd_una, tp->snd_nxt)) {
> > + (!between(seq, snd_una, snd_nxt) ||
> > + (sk->sk_state == TCP_SYN_SENT &&
> > + seq != snd_una && seq == snd_nxt))) {
>
>
> In TCP_SYN_SENT the only sequence we ever put on the wire is the ISN,
> including the Fast Open case (SYN+data is sent with seq == ISN, and the
> RTO retransmits the bare SYN).
>
> So this could simply be :
>
> if (sk->sk_state != TCP_LISTEN &&
> (!between(seq, snd_una, snd_nxt) ||
> (sk->sk_state == TCP_SYN_SENT && seq != snd_una))) {
>
> This mirrors the (seq != tcp_rsk(req)->snt_isn) test in tcp_req_err(),
> and does not need the special case for the window where snd_nxt has
> not been advanced yet.
Looks much better, I'll use it.
Thank you !
>
> > __NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
> > goto out;
> > }
> > diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
> > index 5c7ed38f1f56..8691d220cf94 100644
> > --- a/net/ipv6/tcp_ipv6.c
> > +++ b/net/ipv6/tcp_ipv6.c
> > @@ -410,9 +410,9 @@ static int tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
> > const struct tcphdr *th = (struct tcphdr *)(skb->data+offset);
> > struct net *net = dev_net_rcu(skb->dev);
> > struct request_sock *fastopen;
> > + __u32 seq, snd_una, snd_nxt;
> > struct ipv6_pinfo *np;
> > struct tcp_sock *tp;
> > - __u32 seq, snd_una;
> > struct sock *sk;
> > bool fatal;
> > int err;
> > @@ -463,9 +463,12 @@ static int tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
> > tp = tcp_sk(sk);
> > /* XXX (TFO) - tp->snd_una should be ISN (tcp_create_openreq_child() */
> > fastopen = rcu_dereference(tp->fastopen_rsk);
> > - snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : tp->snd_una;
> > + snd_una = fastopen ? tcp_rsk(fastopen)->snt_isn : READ_ONCE(tp->snd_una);
> > + snd_nxt = READ_ONCE(tp->snd_nxt);
> > if (sk->sk_state != TCP_LISTEN &&
> > - !between(seq, snd_una, tp->snd_nxt)) {
> > + (!between(seq, snd_una, snd_nxt) ||
> > + (sk->sk_state == TCP_SYN_SENT &&
> > + seq != snd_una && seq == snd_nxt))) {
> > __NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
> > goto out;
> > }
> > --
> > 2.56.0.679.gc527c65306-goog
> >
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH v1 net 2/8] tcp: Don't use inner IP TTL for IP_MINTTL in tcp_v4_err().
2026-11-17 8:42 ` Eric Dumazet
@ 2026-11-17 23:35 ` Kuniyuki Iwashima
0 siblings, 0 replies; 19+ messages in thread
From: Kuniyuki Iwashima @ 2026-11-17 23:35 UTC (permalink / raw)
To: Eric Dumazet
Cc: Neal Cardwell, David S . Miller, Jakub Kicinski, Paolo Abeni,
Simon Horman, Kuniyuki Iwashima, netdev, Stephen Hemminger,
stephen hemminger
On Tue, 17 Nov 2026 09:42:02 +0100, Eric Dumazet wrote:
> Le mar. 17 nov. 2026 à 06:35, Kuniyuki Iwashima <kuniyu@google.com> a écrit :
> >
> > RFC 5082 introduced Generalized TTL Security Mechanism (GTSM),
> > and IP_MINTTL and IPV6_MINHOPCOUNT follow that.
> >
> > If an incoming IPv4 packet has TTL less than inet_sk(sk)->min_ttl
> > configured via IP_MINTTL, the packet is dropped.
> >
> > However, tcp_v4_err() checks the TTL extracted from the ICMP
> > payload, not the one in the outer IP header.
> >
> > The minimal repro is:
> >
> > --ip_version=ipv4
> > 0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
> > +0 setsockopt(3, SOL_IP, 21, [250], 4) = 0
> > +0 fcntl(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
> > +0 connect(3, ..., ...) = -1 EINPROGRESS (Operation now in progress)
> > +0 > S 0:0(0) <...>
> > +0 < (ttl 10) icmp unreachable host_unreachable [0:0(0)]
> > +0 %{assert tcpi_state == TCP_SYN_SENT}%
> >
> > Let's check ip_hdr(skb)->ttl.
> >
> > Now GTSM is applied to locally generated ICMP errors, but this
> > behaviour matches tcp_v6_err().
> >
> > Note that tcp_v6_err() properly checks ipv6_hdr(skb)->hop_limit.
> >
> > Fixes: 97e3ecd112ba ("TCP: check min TTL on received ICMP packets")
> > Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
> > ---
> > Cc: Stephen Hemminger <stephen@networkplumber.org>
> > Cc: stephen hemminger <shemminger@vyatta.com>
> > ---
> > net/ipv4/tcp_ipv4.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
> > index 7d58f522068d..bf8542368032 100644
> > --- a/net/ipv4/tcp_ipv4.c
> > +++ b/net/ipv4/tcp_ipv4.c
> > @@ -546,7 +546,7 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
> >
> > if (static_branch_unlikely(&ip4_min_ttl)) {
> > /* min_ttl can be changed concurrently from do_ip_setsockopt() */
> > - if (unlikely(iph->ttl < READ_ONCE(inet_sk(sk)->min_ttl))) {
> > + if (unlikely(ip_hdr(skb)->ttl < READ_ONCE(inet_sk(sk)->min_ttl))) {
> > __NET_INC_STATS(net, LINUX_MIB_TCPMINTTLDROP);
> > goto out;
> > }
>
> Linux sends ICMP errors with the default TTL (64), so locally
> generated errors are now dropped for any min_ttl above 64.
>
> For the strict GTSM case (min_ttl == 255) there is no router between
> the peers, so nothing else changes.
>
> But IP_MINTTL is also used with lower values for multi-hop sessions
> ("ttl-security hops N"). There, an ICMP from an intermediate router
> using an initial TTL of 64 was accepted before this patch (the quoted
> TTL is the one we sent, minus the hops to that router), and is dropped
> after.
>
> I think this is the right thing to do, since the inner TTL is under
> the sender's control, but I would mention this in the changelog.
Will add it to changelog.
Thanks !
^ permalink raw reply [flat|nested] 19+ messages in thread
end of thread, other threads:[~2026-11-17 23:35 UTC | newest]
Thread overview: 19+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-17 5:35 [PATCH v1 net 0/8] tcp: Misc fixes for tcp_v[46]_err() Kuniyuki Iwashima
2026-11-17 5:35 ` [PATCH v1 net 1/8] tcp: Don't accept ICMP for SND.NXT in TCP_SYN_SENT Kuniyuki Iwashima
2026-11-17 8:35 ` Eric Dumazet
2026-11-17 23:30 ` Kuniyuki Iwashima
2026-11-17 5:35 ` [PATCH v1 net 2/8] tcp: Don't use inner IP TTL for IP_MINTTL in tcp_v4_err() Kuniyuki Iwashima
2026-11-17 8:42 ` Eric Dumazet
2026-11-17 23:35 ` Kuniyuki Iwashima
2026-11-17 5:35 ` [PATCH v1 net 3/8] tcp: Don't bypass GTSM check for reqsk Kuniyuki Iwashima
2026-11-17 8:45 ` Eric Dumazet
2026-11-17 5:35 ` [PATCH v1 net 4/8] tcp: Fix reqsk abort condition on ICMP errors Kuniyuki Iwashima
2026-11-17 8:49 ` Eric Dumazet
2026-11-17 5:35 ` [PATCH v1 net 5/8] tcp: Check ip_sk_accept_pmtu() in tcp_v4_err() Kuniyuki Iwashima
2026-11-17 8:54 ` Eric Dumazet
2026-11-17 5:35 ` [PATCH v1 net 6/8] selftest: net: packetdrill: Add ICMP and ICMPv6 tests Kuniyuki Iwashima
2026-11-17 8:58 ` Eric Dumazet
2026-11-17 5:35 ` [PATCH v1 net 7/8] tcp: Pass inet_sdif() to __inet_lookup_established() in tcp_v4_err() Kuniyuki Iwashima
2026-11-17 9:01 ` Eric Dumazet
2026-11-17 5:35 ` [PATCH v1 net 8/8] tcp: Pass inet6_iif() to __inet6_lookup_established() in tcp_v6_err() Kuniyuki Iwashima
2026-11-17 9:03 ` Eric Dumazet
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox