Re: [PATCH] tcp: fix TCP_USER_TIMEOUT with zero window

Hi, Neal:

Thank you for your detailed analysis and your help in coming up with the
right fix. After going through multiple iterations of fixes and discussions,
we are converging to using the timestamp for measuring the elapsed time.

-- Enke

On Wed, Jan 13, 2021 at 04:07:00PM -0500, Neal Cardwell wrote:

Hi Enke,

Sorry I was not clear. :-) I'm trying to convey that there is a functional
difference between the probes_nzw and  icsk_probes_start versions of the
patch.

The functional difference is the one Eric just mentioned, and for which
Eric provided the script that we see behaving in a way that illustrates the
functional difference. The script Eric provided misbehaves for both the
probes_nzw patch and the patch that reverts 9721e709fa68 ("tcp: simplify
window probe aborting on USER_TIMEOUT"). Here is a summary of how the
various approaches behave when run with this test:

o for the probes_nzw patch, the connection times out due to USER_TIMEOUT
too soon (e.g. with a TCP_USER_TIMEOUT of 30 secs the connection times out
after roughly 4 secs)

o for the revert of 9721e709fa68 ("tcp: simplify window probe aborting on
USER_TIMEOUT") the connection never times out due to USER_TIMEOUT

o for the icsk_probes_start version the connection times out due
to USER_TIMEOUT at the appropriate time

As Eric noted, the issue in the probes_nzw case is that the probes_nzw
patch relies on tcp_model_timeout(), which assumes exponential backoff, and
exponential backoff does not happen in the tcp_send_probe0() code path that
sets timeout = TCP_RESOURCE_PROBE_INTERVAL.

best,
neal


On Wed, Jan 13, 2021 at 3:49 PM Eric Dumazet [off-list ref] wrote:

quoted

On Wed, Jan 13, 2021 at 9:12 PM Enke Chen [off-list ref] wrote:

quoted

From: Enke Chen <redacted>

The TCP session does not terminate with TCP_USER_TIMEOUT when data
remain untransmitted due to zero window.

The number of unanswered zero-window probes (tcp_probes_out) is
reset to zero with incoming acks irrespective of the window size,
as described in tcp_probe_timer():

    RFC 1122 4.2.2.17 requires the sender to stay open indefinitely
    as long as the receiver continues to respond probes. We support
    this by default and reset icsk_probes_out with incoming ACKs.

This counter, however, is the wrong one to be used in calculating the
duration that the window remains closed and data remain untransmitted.
Thanks to Jonathan Maxwell [off-list ref] for diagnosing the
actual issue.

In this patch a separate counter is introduced to track the number of
zero-window probes that are not answered with any non-zero window ack.
This new counter is used in determining when to abort the session with
TCP_USER_TIMEOUT.

I think one possible issue would be that local congestion (full qdisc)
would abort early,
because tcp_model_timeout() assumes linear backoff.

Neal or Yuchung can further comment on that, it is late for me in France.

packetdrill test would be :

   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 0 <mss 1460>
   +0 > S. 0:0(0) ack 1 <mss 1460>

  +.1 < . 1:1(0) ack 1 win 65530
   +0 accept(3, ..., ...) = 4

   +0 setsockopt(4, SOL_TCP, TCP_USER_TIMEOUT, [3000], 4) = 0
   +0 write(4, ..., 24) = 24
   +0 > P. 1:25(24) ack 1
   +.1 < . 1:1(0) ack 25 win 65530
   +0 %{ assert tcpi_probes == 0, tcpi_probes; \
         assert tcpi_backoff == 0, tcpi_backoff }%

// install a qdisc dropping all packets
   +0 `tc qdisc delete dev tun0 root 2>/dev/null ; tc qdisc add dev
tun0 root pfifo limit 0`
   +0 write(4, ..., 24) = 24
   // When qdisc is congested we retry every 500ms therefore in theory
   // we'd retry 6 times before hitting 3s timeout. However, since we
   // estimate the elapsed time based on exp backoff of actual RTO (300ms),
   // we'd bail earlier with only 3 probes.
   +2.1 write(4, ..., 24) = -1
   +0 %{ assert tcpi_probes == 3, tcpi_probes; \
         assert tcpi_backoff == 0, tcpi_backoff }%
   +0 close(4) = 0

quoted

Cc: stable@vger.kernel.org
Fixes: 9721e709fa68 ("tcp: simplify window probe aborting on

USER_TIMEOUT")

quoted

Reported-by: William McCall <redacted>
Signed-off-by: Enke Chen <redacted>
---
 include/linux/tcp.h   | 5 +++++
 net/ipv4/tcp.c        | 1 +
 net/ipv4/tcp_input.c  | 3 ++-
 net/ipv4/tcp_output.c | 2 ++
 net/ipv4/tcp_timer.c  | 5 +++--
 5 files changed, 13 insertions(+), 3 deletions(-)

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 2f87377e9af7..c9415b30fa67 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h

@@ -352,6 +352,11 @@ struct tcp_sock {

        int                     linger2;

+       /* While icsk_probes_out is for unanswered 0 window probes, this
+        * counter is for 0-window probes that are not answered with any
+        * non-zero window (nzw) acks.
+        */
+       u8      probes_nzw;

 /* Sock_ops bpf program related variables */
 #ifdef CONFIG_BPF

diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c
index ed42d2193c5c..af6a41a5a5ac 100644
--- a/net/ipv4/tcp.c
+++ b/net/ipv4/tcp.c

@@ -2940,6 +2940,7 @@ int tcp_disconnect(struct sock *sk, int flags)
        icsk->icsk_rto = TCP_TIMEOUT_INIT;
        icsk->icsk_rto_min = TCP_RTO_MIN;
        icsk->icsk_delack_max = TCP_DELACK_MAX;
+       tp->probes_nzw = 0;
        tp->snd_ssthresh = TCP_INFINITE_SSTHRESH;
        tp->snd_cwnd = TCP_INIT_CWND;
        tp->snd_cwnd_cnt = 0;

diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index c7e16b0ed791..4812a969c18a 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c

@@ -3377,13 +3377,14 @@ static void tcp_ack_probe(struct sock *sk)
 {
        struct inet_connection_sock *icsk = inet_csk(sk);
        struct sk_buff *head = tcp_send_head(sk);
-       const struct tcp_sock *tp = tcp_sk(sk);
+       struct tcp_sock *tp = tcp_sk(sk);

        /* Was it a usable window open? */
        if (!head)
                return;
        if (!after(TCP_SKB_CB(head)->end_seq, tcp_wnd_end(tp))) {
                icsk->icsk_backoff = 0;
+               tp->probes_nzw = 0;
                inet_csk_clear_xmit_timer(sk, ICSK_TIME_PROBE0);
                /* Socket must be waked up by subsequent

tcp_data_snd_check().

quoted

                 * This function is not for random using!

diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index f322e798a351..1b64cdabc299 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c

@@ -4084,10 +4084,12 @@ void tcp_send_probe0(struct sock *sk)
                /* Cancel probe timer, if it is not required. */
                icsk->icsk_probes_out = 0;
                icsk->icsk_backoff = 0;
+               tp->probes_nzw = 0;
                return;
        }

        icsk->icsk_probes_out++;
+       tp->probes_nzw++;
        if (err <= 0) {
                if (icsk->icsk_backoff < net->ipv4.sysctl_tcp_retries2)
                        icsk->icsk_backoff++;

diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c
index 6c62b9ea1320..87e9f5998b8e 100644
--- a/net/ipv4/tcp_timer.c
+++ b/net/ipv4/tcp_timer.c

@@ -349,6 +349,7 @@ static void tcp_probe_timer(struct sock *sk)

        if (tp->packets_out || !skb) {
                icsk->icsk_probes_out = 0;
+               tp->probes_nzw = 0;
                return;
        }

@@ -360,8 +361,8 @@ static void tcp_probe_timer(struct sock *sk)
         * corresponding system limit. We also implement similar policy

when

quoted

         * we use RTO to probe window in tcp_retransmit_timer().
         */
-       if (icsk->icsk_user_timeout) {
-               u32 elapsed = tcp_model_timeout(sk,

icsk->icsk_probes_out,

quoted

+       if (icsk->icsk_user_timeout && tp->probes_nzw) {
+               u32 elapsed = tcp_model_timeout(sk, tp->probes_nzw,
                                                tcp_probe0_base(sk));

                if (elapsed >= icsk->icsk_user_timeout)
--
2.29.2