mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* [PATCH net v2 0/2] tcp: roccet: fix idle restart on RTO
@ 2026-11-18  2:28 Zhiqi Zhang
  2026-11-18  2:28 ` [PATCH net v2 1/2] tcp: roccet: do not restart from idle on RTO retransmission Zhiqi Zhang
                   ` (2 more replies)
  0 siblings, 3 replies; 6+ messages in thread
From: Zhiqi Zhang @ 2026-11-18  2:28 UTC (permalink / raw)
  To: Eric Dumazet, Neal Cardwell, David S . Miller, Jakub Kicinski,
	Paolo Abeni
  Cc: Kuniyuki Iwashima, Simon Horman, Shuah Khan, Tim Fuechsel,
	Lukas Prause, netdev, linux-kselftest, linux-kernel

Patch 1/2 skips the idle restart when packets are outstanding, so an
RTO retransmission no longer overwrites the snd_ssthresh preserved by
roccet_state(TCP_CA_Loss).  Patch 2/2 adds a packetdrill test.

v1:
https://lore.kernel.org/all/20261114141315.19007-1-zhiqizhangg@gmail.com/

Eric's review:
https://lore.kernel.org/all/CAL4WiiohBWs_BteL4o__osD6cPTXrBBPa-P7faV=acTaF_4g3w@mail.gmail.com/

Changes since v1:
- Make the changelog precise about when the idle restart is reached:
  with TLP the first RTO is affected; without TLP the first RTO can
  escape and the second RTO is always affected.
- Limit the claimed effect to tp->snd_ssthresh.  roccet_state() calls
  roccet_reset(), so ca->state is already LAUNCH and last_max_cwnd is
  already 0 when roccet_cwnd_event_tx_start() runs; the idle restart
  only overwrites snd_ssthresh.
- Shorten the comment to the style used in the file.
- Add the packetdrill test from Eric as 2/2, and enable
  CONFIG_TCP_CONG_ADVANCED plus CONFIG_TCP_CONG_ROCCET in the
  packetdrill config (ROCCET is only visible with advanced congestion
  control enabled).

Patch 2/2 is based on a test written by Eric Dumazet.  A variant with
tcp_early_retrans=0 (no TLP) fails at the second RTO only.

Testing: Eric ran the packetdrill test in 2/2 (it fails at the first RTO
without 1/2 and passes with it).  The netem reproducer from v1 still
shows snd_ssthresh being preserved during the RTO with this series.

AI assistance:

- DeepSeek (deepseek-flash) generated the code change in 1/2.
- I wrote the netem reproducer and the documentation, and reviewed the
  code.  The packetdrill test in 2/2 was written by Eric Dumazet.
- DeepSeek was used to polish the wording.

Thanks Eric for the review and the packetdrill test.


Zhiqi Zhang (2):
  tcp: roccet: do not restart from idle on RTO retransmission
  selftests: net: packetdrill: add ROCCET RTO test

 net/ipv4/tcp_roccet.c                         |  8 ++-
 .../testing/selftests/net/packetdrill/config  |  2 +
 .../net/packetdrill/tcp_roccet_rto.pkt        | 63 +++++++++++++++++++
 3 files changed, 71 insertions(+), 2 deletions(-)
 create mode 100644 tools/testing/selftests/net/packetdrill/tcp_roccet_rto.pkt

-- 
2.54.0 (Apple Git-157)



^ permalink raw reply	[flat|nested] 6+ messages in thread

* [PATCH net v2 1/2] tcp: roccet: do not restart from idle on RTO retransmission
  2026-11-18  2:28 [PATCH net v2 0/2] tcp: roccet: fix idle restart on RTO Zhiqi Zhang
@ 2026-11-18  2:28 ` Zhiqi Zhang
  2026-11-18 10:07   ` Eric Dumazet
  2026-11-18  2:28 ` [PATCH net v2 2/2] selftests: net: packetdrill: add ROCCET RTO test Zhiqi Zhang
  2026-11-21 22:40 ` [PATCH net v2 0/2] tcp: roccet: fix idle restart on RTO patchwork-bot+netdevbpf
  2 siblings, 1 reply; 6+ messages in thread
From: Zhiqi Zhang @ 2026-11-18  2:28 UTC (permalink / raw)
  To: Eric Dumazet, Neal Cardwell, David S . Miller, Jakub Kicinski,
	Paolo Abeni
  Cc: Kuniyuki Iwashima, Simon Horman, Shuah Khan, Tim Fuechsel,
	Lukas Prause, netdev, linux-kselftest, linux-kernel

roccet sets .cong_control, so tcp_slow_start_after_idle_check() returns
early and roccet implements the idle restart in cwnd_event_tx_start().

An RTO retransmission also raises CA_EVENT_TX_START: tcp_enter_loss()
marks the outstanding packets lost, which drains the in-flight count
while tp->packets_out stays non-zero.  If delta >= icsk_rto,
roccet_cwnd_event_tx_start() then takes the idle restart and overwrites
tp->snd_ssthresh with tcp_snd_cwnd() == TCP_INIT_CWND.

When a TLP probe was sent, the RTO timer is rearmed for a full icsk_rto
at that point and the probe is the last transmit, so the first RTO is
affected.  Without TLP, the first RTO can escape: the timer is armed
when the first packet of the flight is sent, and pacing spreads the
following ones, so now - tp->lsndtime < icsk_rto.  The second RTO is
always affected, since it fires one backed-off icsk_rto after the first
RTO retransmit.

roccet_state(TCP_CA_Loss) calls roccet_reset(), which clears the private
state; ca->state is already LAUNCH and last_max_cwnd is already 0 when
roccet_cwnd_event_tx_start() runs, with or without this fix.  The only
field the idle restart breaks is tp->snd_ssthresh: it overwrites the
value preserved by roccet_state() with TCP_INIT_CWND.

Skip the idle restart while packets are outstanding.  The packetdrill
test in the next patch fails at the first RTO without this fix
(snd_ssthresh == 10) and passes with it.

Fixes: 819d59638796 ("tcp: Add TCP ROCCET congestion control module.")
Link: https://lore.kernel.org/all/179179785248.434549.16829837649351382748@kernel.org/
Signed-off-by: Zhiqi Zhang <zhiqizhangg@gmail.com>
---
 net/ipv4/tcp_roccet.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/net/ipv4/tcp_roccet.c b/net/ipv4/tcp_roccet.c
index 04b9e52466..1769e4bbe3 100644
--- a/net/ipv4/tcp_roccet.c
+++ b/net/ipv4/tcp_roccet.c
@@ -1144,8 +1144,12 @@ static void roccet_cwnd_event_tx_start(struct sock *sk)
 			ca->epoch_start = now;
 	}
 
-	/* After idle periods, go into LAUNCH with slow start */
-	if (READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_slow_start_after_idle) &&
+	/* After idle periods, go into LAUNCH with slow start.
+	 * We are not idle if packets are outstanding (e.g. RTO retransmit),
+	 * see tcp_slow_start_after_idle_check().
+	 */
+	if (!tp->packets_out &&
+	    READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_slow_start_after_idle) &&
 	    delta >= inet_csk(sk)->icsk_rto) {
 		tp->snd_ssthresh = tcp_snd_cwnd(tp);
 		tcp_snd_cwnd_set(tp, min(TCP_INIT_CWND, tp->snd_cwnd_clamp));
-- 
2.54.0 (Apple Git-157)



^ permalink raw reply	[flat|nested] 6+ messages in thread

* [PATCH net v2 2/2] selftests: net: packetdrill: add ROCCET RTO test
  2026-11-18  2:28 [PATCH net v2 0/2] tcp: roccet: fix idle restart on RTO Zhiqi Zhang
  2026-11-18  2:28 ` [PATCH net v2 1/2] tcp: roccet: do not restart from idle on RTO retransmission Zhiqi Zhang
@ 2026-11-18  2:28 ` Zhiqi Zhang
  2026-11-18 10:05   ` Eric Dumazet
  2026-11-21 22:40 ` [PATCH net v2 0/2] tcp: roccet: fix idle restart on RTO patchwork-bot+netdevbpf
  2 siblings, 1 reply; 6+ messages in thread
From: Zhiqi Zhang @ 2026-11-18  2:28 UTC (permalink / raw)
  To: Eric Dumazet, Neal Cardwell, David S . Miller, Jakub Kicinski,
	Paolo Abeni
  Cc: Kuniyuki Iwashima, Simon Horman, Shuah Khan, Tim Fuechsel,
	Lukas Prause, netdev, linux-kselftest, linux-kernel

Add a packetdrill test for the ROCCET idle restart on RTO.  TLP is
enabled, so the first RTO is affected; the test asserts that
tp->snd_ssthresh stays at the prior cwnd (20) across the first and
second RTOs.

The test fails at the first RTO without the preceding fix
(snd_ssthresh == 10) and passes with it.  A variant with
tcp_early_retrans=0 (no TLP) fails at the second RTO only.

CONFIG_TCP_CONG_ADVANCED and CONFIG_TCP_CONG_ROCCET are added to the
packetdrill config; ROCCET is only visible when advanced congestion
control is enabled.

Test written by Eric Dumazet.

Suggested-by: Eric Dumazet <edumazet@kernel.org>
Signed-off-by: Zhiqi Zhang <zhiqizhangg@gmail.com>
---
 .../testing/selftests/net/packetdrill/config  |  2 +
 .../net/packetdrill/tcp_roccet_rto.pkt        | 63 +++++++++++++++++++
 2 files changed, 65 insertions(+)
 create mode 100644 tools/testing/selftests/net/packetdrill/tcp_roccet_rto.pkt

diff --git a/tools/testing/selftests/net/packetdrill/config b/tools/testing/selftests/net/packetdrill/config
index b7df8bf309..6e012141d5 100644
--- a/tools/testing/selftests/net/packetdrill/config
+++ b/tools/testing/selftests/net/packetdrill/config
@@ -6,6 +6,8 @@ CONFIG_NET_SCH_FIFO=y
 CONFIG_NET_SCH_FQ=y
 CONFIG_SYN_COOKIES=y
 CONFIG_SYSCTL=y
+CONFIG_TCP_CONG_ADVANCED=y
 CONFIG_TCP_CONG_CUBIC=y
+CONFIG_TCP_CONG_ROCCET=y
 CONFIG_TCP_MD5SIG=y
 CONFIG_TUN=y
diff --git a/tools/testing/selftests/net/packetdrill/tcp_roccet_rto.pkt b/tools/testing/selftests/net/packetdrill/tcp_roccet_rto.pkt
new file mode 100644
index 0000000000..dabaf07f32
--- /dev/null
+++ b/tools/testing/selftests/net/packetdrill/tcp_roccet_rto.pkt
@@ -0,0 +1,63 @@
+// SPDX-License-Identifier: GPL-2.0
+// Test that ROCCET does not take its "restart after idle" path on RTO.
+//
+// An RTO retransmit raises CA_EVENT_TX_START, because all outstanding
+// packets were just marked lost (tcp_packets_in_flight() == 0) while
+// tp->packets_out is not zero. If (now - tp->lsndtime) >= icsk_rto,
+// roccet_cwnd_event_tx_start() used to overwrite the snd_ssthresh
+// that tcp_enter_loss() had just set with TCP_INIT_CWND.
+//
+// TLP is enabled here. The RTO timer is rearmed for a full icsk_rto when
+// the loss probe is sent, and the probe is the last transmit, so the
+// very first RTO is affected.
+
+`./defaults.sh
+./set_sysctls.py /proc/sys/net/ipv4/tcp_slow_start_after_idle=1`
+
+    0 socket(..., SOCK_STREAM, IPPROTO_TCP) = 3
+   +0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+   +0 setsockopt(3, SOL_TCP, TCP_CONGESTION, "roccet", 6) = 0
+   +0 bind(3, ..., ...) = 0
+   +0 listen(3, 1) = 0
+
+   +0 < S 0:0(0) win 65535 <mss 1000,sackOK,nop,nop,nop,wscale 7>
+   +0 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale 8>
+  +.1 < . 1:1(0) ack 1 win 511
+   +0 accept(3, ..., ...) = 4
+
+// ROCCET paces its packets: do not expect precise outgoing packets,
+// rely on TCP_INFO instead.
+   +0 write(4, ..., 10000) = 10000
+  +.1 < . 1:1(0) ack 10001 win 511
+   +0 %{ assert tcpi_snd_cwnd == 20, tcpi_snd_cwnd }%
+
+// RTT is 100ms, RTO is ~300ms. Nothing is acked from now on.
+   +0 write(4, ..., 20000) = 20000
+
+// TLP fires ~250ms after the write.
+  +.4 %{
+assert tcpi_ca_state == TCP_CA_Open, tcpi_ca_state
+assert tcpi_total_retrans == 1, tcpi_total_retrans
+}%
+
+// First RTO, ~300ms after the TLP probe.
+// ROCCET is still in LAUNCH state: ssthresh is set to the prior cwnd.
+  +.3 %{
+assert tcpi_ca_state == TCP_CA_Loss, tcpi_ca_state
+assert tcpi_backoff == 1, tcpi_backoff
+assert tcpi_total_retrans == 2, tcpi_total_retrans
+assert tcpi_snd_cwnd == 10, tcpi_snd_cwnd
+assert tcpi_snd_ssthresh == 20, tcpi_snd_ssthresh
+}%
+
+// Second RTO, ~600ms after the first one.
+  +.6 %{
+assert tcpi_ca_state == TCP_CA_Loss, tcpi_ca_state
+assert tcpi_backoff == 2, tcpi_backoff
+assert tcpi_total_retrans == 3, tcpi_total_retrans
+assert tcpi_snd_cwnd == 10, tcpi_snd_cwnd
+assert tcpi_snd_ssthresh == 20, tcpi_snd_ssthresh
+}%
+
+// Reset sysctls
+`/tmp/sysctl_restore_${PPID}.sh`
-- 
2.54.0 (Apple Git-157)



^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH net v2 2/2] selftests: net: packetdrill: add ROCCET RTO test
  2026-11-18  2:28 ` [PATCH net v2 2/2] selftests: net: packetdrill: add ROCCET RTO test Zhiqi Zhang
@ 2026-11-18 10:05   ` Eric Dumazet
  0 siblings, 0 replies; 6+ messages in thread
From: Eric Dumazet @ 2026-11-18 10:05 UTC (permalink / raw)
  To: Zhiqi Zhang
  Cc: Neal Cardwell, David S . Miller, Jakub Kicinski, Paolo Abeni,
	Kuniyuki Iwashima, Simon Horman, Shuah Khan, Tim Fuechsel,
	Lukas Prause, netdev, linux-kselftest, linux-kernel

Le mer. 18 nov. 2026 à 03:28, Zhiqi Zhang <zhiqizhangg@gmail.com> a écrit :
>
> Add a packetdrill test for the ROCCET idle restart on RTO.  TLP is
> enabled, so the first RTO is affected; the test asserts that
> tp->snd_ssthresh stays at the prior cwnd (20) across the first and
> second RTOs.
>
> The test fails at the first RTO without the preceding fix
> (snd_ssthresh == 10) and passes with it.  A variant with
> tcp_early_retrans=0 (no TLP) fails at the second RTO only.
>
> CONFIG_TCP_CONG_ADVANCED and CONFIG_TCP_CONG_ROCCET are added to the
> packetdrill config; ROCCET is only visible when advanced congestion
> control is enabled.
>
> Test written by Eric Dumazet.
>
> Suggested-by: Eric Dumazet <edumazet@kernel.org>
> Signed-off-by: Zhiqi Zhang <zhiqizhangg@gmail.com>

Reviewed-by: Eric Dumazet <edumazet@kernel.org>

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH net v2 1/2] tcp: roccet: do not restart from idle on RTO retransmission
  2026-11-18  2:28 ` [PATCH net v2 1/2] tcp: roccet: do not restart from idle on RTO retransmission Zhiqi Zhang
@ 2026-11-18 10:07   ` Eric Dumazet
  0 siblings, 0 replies; 6+ messages in thread
From: Eric Dumazet @ 2026-11-18 10:07 UTC (permalink / raw)
  To: Zhiqi Zhang
  Cc: Neal Cardwell, David S . Miller, Jakub Kicinski, Paolo Abeni,
	Kuniyuki Iwashima, Simon Horman, Shuah Khan, Tim Fuechsel,
	Lukas Prause, netdev, linux-kselftest, linux-kernel

Le mer. 18 nov. 2026 à 03:28, Zhiqi Zhang <zhiqizhangg@gmail.com> a écrit :
>
> roccet sets .cong_control, so tcp_slow_start_after_idle_check() returns
> early and roccet implements the idle restart in cwnd_event_tx_start().
>
> An RTO retransmission also raises CA_EVENT_TX_START: tcp_enter_loss()
> marks the outstanding packets lost, which drains the in-flight count
> while tp->packets_out stays non-zero.  If delta >= icsk_rto,
> roccet_cwnd_event_tx_start() then takes the idle restart and overwrites
> tp->snd_ssthresh with tcp_snd_cwnd() == TCP_INIT_CWND.
>
> When a TLP probe was sent, the RTO timer is rearmed for a full icsk_rto
> at that point and the probe is the last transmit, so the first RTO is
> affected.  Without TLP, the first RTO can escape: the timer is armed
> when the first packet of the flight is sent, and pacing spreads the
> following ones, so now - tp->lsndtime < icsk_rto.  The second RTO is
> always affected, since it fires one backed-off icsk_rto after the first
> RTO retransmit.
>
> roccet_state(TCP_CA_Loss) calls roccet_reset(), which clears the private
> state; ca->state is already LAUNCH and last_max_cwnd is already 0 when
> roccet_cwnd_event_tx_start() runs, with or without this fix.  The only
> field the idle restart breaks is tp->snd_ssthresh: it overwrites the
> value preserved by roccet_state() with TCP_INIT_CWND.
>
> Skip the idle restart while packets are outstanding.  The packetdrill
> test in the next patch fails at the first RTO without this fix
> (snd_ssthresh == 10) and passes with it.
>
> Fixes: 819d59638796 ("tcp: Add TCP ROCCET congestion control module.")
> Link: https://lore.kernel.org/all/179179785248.434549.16829837649351382748@kernel.org/
> Signed-off-by: Zhiqi Zhang <zhiqizhangg@gmail.com>
> ---

Reviewed-by: Eric Dumazet <edumazet@kernel.org>

Thanks.

^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH net v2 0/2] tcp: roccet: fix idle restart on RTO
  2026-11-18  2:28 [PATCH net v2 0/2] tcp: roccet: fix idle restart on RTO Zhiqi Zhang
  2026-11-18  2:28 ` [PATCH net v2 1/2] tcp: roccet: do not restart from idle on RTO retransmission Zhiqi Zhang
  2026-11-18  2:28 ` [PATCH net v2 2/2] selftests: net: packetdrill: add ROCCET RTO test Zhiqi Zhang
@ 2026-11-21 22:40 ` patchwork-bot+netdevbpf
  2 siblings, 0 replies; 6+ messages in thread
From: patchwork-bot+netdevbpf @ 2026-11-21 22:40 UTC (permalink / raw)
  To: Zhiqi Zhang
  Cc: edumazet, ncardwell, davem, kuba, pabeni, kuniyu, horms, shuah,
	t.fuechsel, lukas.prause, netdev, linux-kselftest, linux-kernel

Hello:

This series was applied to netdev/net-next.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Wed, 18 Nov 2026 10:28:11 +0800 you wrote:
> Patch 1/2 skips the idle restart when packets are outstanding, so an
> RTO retransmission no longer overwrites the snd_ssthresh preserved by
> roccet_state(TCP_CA_Loss).  Patch 2/2 adds a packetdrill test.
> 
> v1:
> https://lore.kernel.org/all/20261114141315.19007-1-zhiqizhangg@gmail.com/
> 
> [...]

Here is the summary with links:
  - [net,v2,1/2] tcp: roccet: do not restart from idle on RTO retransmission
    https://git.kernel.org/netdev/net-next/c/19764eae5794
  - [net,v2,2/2] selftests: net: packetdrill: add ROCCET RTO test
    https://git.kernel.org/netdev/net-next/c/295aaa8398ce

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html



^ permalink raw reply	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2026-11-21 22:40 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-11-18  2:28 [PATCH net v2 0/2] tcp: roccet: fix idle restart on RTO Zhiqi Zhang
2026-11-18  2:28 ` [PATCH net v2 1/2] tcp: roccet: do not restart from idle on RTO retransmission Zhiqi Zhang
2026-11-18 10:07   ` Eric Dumazet
2026-11-18  2:28 ` [PATCH net v2 2/2] selftests: net: packetdrill: add ROCCET RTO test Zhiqi Zhang
2026-11-18 10:05   ` Eric Dumazet
2026-11-21 22:40 ` [PATCH net v2 0/2] tcp: roccet: fix idle restart on RTO patchwork-bot+netdevbpf

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®