Netdev List
 help / color / mirror / Atom feed
From: Eric Dumazet <edumazet@kernel.org>
To: "David S . Miller" <davem@davemloft.net>,
	Jakub Kicinski <kuba@kernel.org>, Paolo Abeni <pabeni@redhat.com>,
	Neal Cardwell <ncardwell@google.com>
Cc: Kuniyuki Iwashima <kuniyu@google.com>,
	Simon Horman <horms@kernel.org>,
	Flavius-Romeo Mazilu <flavius.mazilu@stud.acs.upb.ro>,
	Michael Tuexen <tuexen@fh-muenster.de>,
	netdev@vger.kernel.org, Eric Dumazet <edumazet@kernel.org>
Subject: [PATCH net-next 1/2] tcp: do not accept ACK of our ISS on active opens
Date: Wed, 18 Nov 2026 18:26:53 +0100	[thread overview]
Message-ID: <20261118172654.1267071-2-edumazet@kernel.org> (raw)
In-Reply-To: <20261118172654.1267071-1-edumazet@kernel.org>

Michael Tuexen asked on the tcpm list [1] whether, after a completed
handshake, a segment with SEG.ACK == ISS is a duplicate ACK or a ghost
ACK. Flavius-Romeo Mazilu answered that Linux behaves differently
depending on which side opened the connection:

- passive opens drop it and send a challenge ACK,
- active opens accept it as an old duplicate ACK.

This comes from commit 3d501dd326fb ("tcp: do not accept ACK of bytes
we never sent"), which bounds the range of acceptable old ACKs
by tp->bytes_acked.

On active opens (including Fast Open and simultaneous opens),
tp->snd_una starts at ISS and the ACK of our SYN goes through
tcp_snd_una_update(), so tp->bytes_acked counts the SYN.
On passive opens, tp->snd_una starts at ISS + 1 and the SYN-ACK
is not counted.

The SYN is not a payload byte, and no valid ACK ever carried ISS:
the lowest ACK a peer can send on a synchronized connection is ISS + 1.

Counting the SYN widens the acceptable range by one sequence number.

tp->bytes_acked is exported to user space (tcpi_bytes_acked,
bpf_tcp_sock, bpf_sock_ops) and tools/testing/selftests/bpf expects
the SYN to be counted, so leave it unchanged.

Instead, remember in a new bit that our SYN was counted,
and subtract it in the tcp_ack() check.

Mazilu's observation was confirmed by code inspection and with
packetdrill on an unmodified kernel: after an active open, a segment
with SEG.ACK == ISS is silently accepted, where a passive open
answers with a challenge ACK.

Tested with the packetdrill tests added in the following patch,
which fail on the active side before this patch, and with the other
tests in tools/testing/selftests/net/packetdrill.

[1] https://mailarchive.ietf.org/arch/msg/tcpm/ZbjDTFQSHeKSVHAToyAnKBr3TpU/

Fixes: 3d501dd326fb ("tcp: do not accept ACK of bytes we never sent")
Reported-by: Flavius-Romeo Mazilu <flavius.mazilu@stud.acs.upb.ro>
Closes: https://mailarchive.ietf.org/arch/msg/tcpm/ym6lb-UkrW2298LRnTLxCysoo0g/
Signed-off-by: Eric Dumazet <edumazet@kernel.org>
---
 Documentation/networking/net_cachelines/tcp_sock.rst |  1 +
 include/linux/tcp.h                                  |  3 ++-
 net/ipv4/tcp.c                                       |  1 +
 net/ipv4/tcp_input.c                                 | 10 ++++++++--
 4 files changed, 12 insertions(+), 3 deletions(-)

diff --git a/Documentation/networking/net_cachelines/tcp_sock.rst b/Documentation/networking/net_cachelines/tcp_sock.rst
index 554e8c14a401f27ec5a6c109877bb122942bade9..78f74eb7699063022adc5d4524a82466c82e4758 100644
--- a/Documentation/networking/net_cachelines/tcp_sock.rst
+++ b/Documentation/networking/net_cachelines/tcp_sock.rst
@@ -59,6 +59,7 @@ u8:2                          fastopen_client_fail
 u8:4                          nonagle                 read_write                              tcp_skb_entail,tcp_push_pending_frames
 u8:1                          thin_lto
 u8:1                          recvmsg_inq                                 read_mostly         tcp_recvmsg
+u8:1                          bytes_acked_syn                             read_mostly         tcp_ack
 u8:1                          repair                  read_mostly                             tcp_write_xmit
 u8:1                          frto
 u8                            repair_queue
diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index e6b79cf4798f497ea43bb987431da2c797ee3086..13a192cc3c62e219235e0610924fe460148be19c 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -235,7 +235,8 @@ struct tcp_sock {
 		tcp_usec_ts : 1, /* TSval values in usec */
 		is_sack_reneg:1,    /* in recovery from loss with SACK reneg? */
 		is_cwnd_limited:1,/* forward progress limited by snd_cwnd? */
-		recvmsg_inq : 1;/* Indicate # of bytes in queue upon recvmsg */
+		recvmsg_inq : 1,/* Indicate # of bytes in queue upon recvmsg */
+		bytes_acked_syn:1;/* bytes_acked includes our SYN */
 #ifdef CONFIG_BPF
 	u32	bpf_tcp_ops_flags;
 #define BPF_TCP_OPS_TEST_FLAG(TP, ARG) \
diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index d510e982dcea9830442d4c7afedb67deb3c28115..498889efe2e3323484b9d30842d9f16475165e4a 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c
@@ -3472,6 +3472,7 @@ int tcp_disconnect(struct sock *sk, int flags)
 	tp->segs_out = 0;
 	tp->bytes_sent = 0;
 	tp->bytes_acked = 0;
+	tp->bytes_acked_syn = 0;
 	tp->bytes_received = 0;
 	tp->bytes_retrans = 0;
 	tp->data_segs_in = 0;
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index d35048c363be5cb985ba1efb9dab24c9aaab4fda..6cf6cb3336b23b065b93d747e0565475ce4c3d0c 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -3712,6 +3712,8 @@ static int tcp_clean_rtx_queue(struct sock *sk, const struct sk_buff *ack_skb,
 		} else {
 			flag |= FLAG_SYN_ACKED;
 			tp->retrans_stamp = 0;
+			/* tcp_snd_una_update() counted our SYN. */
+			tp->bytes_acked_syn = 1;
 		}
 
 		if (!fully_acked)
@@ -4333,8 +4335,12 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
 	if (before(ack, prior_snd_una)) {
 		u32 max_window;
 
-		/* do not accept ACK for bytes we never sent. */
-		max_window = min_t(u64, tp->max_window, tp->bytes_acked);
+		/* Do not accept ACK for bytes we never sent.
+		 * Our SYN, if counted in bytes_acked, is not one of them:
+		 * no valid ACK ever carried ISS.
+		 */
+		max_window = min_t(u64, tp->max_window,
+				   tp->bytes_acked - tp->bytes_acked_syn);
 		/* RFC 5961 5.2 [Blind Data Injection Attack].[Mitigation] */
 		if (before(ack, prior_snd_una - max_window)) {
 			if (!(flag & FLAG_NO_CHALLENGE_ACK))
-- 
2.53.0


  reply	other threads:[~2026-11-18 17:27 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-18 17:26 [PATCH net-next 0/2] tcp: do not accept ACK of our ISS on active opens Eric Dumazet
2026-11-18 17:26 ` Eric Dumazet [this message]
2026-11-18 17:26 ` [PATCH net-next 2/2] selftests/net: packetdrill: add tests for ACK of ISS Eric Dumazet

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20261118172654.1267071-2-edumazet@kernel.org \
    --to=edumazet@kernel.org \
    --cc=davem@davemloft.net \
    --cc=flavius.mazilu@stud.acs.upb.ro \
    --cc=horms@kernel.org \
    --cc=kuba@kernel.org \
    --cc=kuniyu@google.com \
    --cc=ncardwell@google.com \
    --cc=netdev@vger.kernel.org \
    --cc=pabeni@redhat.com \
    --cc=tuexen@fh-muenster.de \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox