Hello Eric,

On Thu, Aug 1, 2024 at 2:58 PM Eric Dumazet [off-list ref] wrote:

On Wed, Jul 31, 2024 at 2:10 PM Jason Xing [off-list ref] wrote:

quoted

From: Jason Xing <kernelxing@tencent.com>

Whether user sets TCP_USER_TIMEOUT option or not, when we find there
is no left chance to proceed, we will send an RST to the other side.

Not sure why you mention TCP_USER_TIMEOUT here.

quoted

Signed-off-by: Jason Xing <kernelxing@tencent.com>
---
v2
Link: https://lore.kernel.org/all/CAL+tcoB-12pUS0adK8M_=C97aXewYYmDA6rJKLXvAXEHvEsWjA@mail.gmail.com/ (local)
1. correct the comment and changelog
---
 include/net/rstreason.h | 8 ++++++++
 net/ipv4/tcp_timer.c    | 2 +-
 2 files changed, 9 insertions(+), 1 deletion(-)

diff --git a/include/net/rstreason.h b/include/net/rstreason.h
index bbf20d0bbde7..739ad1db4212 100644
--- a/include/net/rstreason.h
+++ b/include/net/rstreason.h

@@ -21,6 +21,7 @@
        FN(TCP_ABORT_ON_LINGER)         \
        FN(TCP_ABORT_ON_MEMORY)         \
        FN(TCP_STATE)                   \
+       FN(TCP_TIMEOUT)                 \
        FN(MPTCP_RST_EUNSPEC)           \
        FN(MPTCP_RST_EMPTCP)            \
        FN(MPTCP_RST_ERESOURCE)         \

@@ -108,6 +109,13 @@ enum sk_rst_reason {
         * Please see RFC 9293 for all possible reset conditions
         */
        SK_RST_REASON_TCP_STATE,
+       /**
+        * @SK_RST_REASON_TCP_TIMEOUT: time to timeout
+        * Whether user sets TCP_USER_TIMEOUT options or not, when we
+        * have already run out of all the chances, we have to reset the
+        * connection
+        */
+       SK_RST_REASON_TCP_TIMEOUT,

        /* Copy from include/uapi/linux/mptcp.h.
         * These reset fields will not be changed since they adhere to

diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c
index 3910f6d8614e..bd403300e4c4 100644
--- a/net/ipv4/tcp_timer.c
+++ b/net/ipv4/tcp_timer.c

@@ -807,7 +807,7 @@ static void tcp_keepalive_timer (struct timer_list *t)
                    (user_timeout == 0 &&
                    icsk->icsk_probes_out >= keepalive_probes(tp))) {
                        tcp_send_active_reset(sk, GFP_ATOMIC,
-                                             SK_RST_REASON_NOT_SPECIFIED);
+                                             SK_RST_REASON_TCP_TIMEOUT);
                        tcp_write_err(sk);
                        goto out;
                }
--

2.37.3

This is more about keepalive really. You should use a better name
reflecting that it is a keepalive timeout.

I think you're right. Let me use 'TCP_KEEPALIVE_TIMEOUT' then.

Thanks,
Jason