[PATCH openEuler-1.0-LTS] net:fix tcp timeout retransmits are always missing 2 times
mainline inclusion from mainline-v5.3.0 commit 3256a2d6ab1f71f9a1bd2d7f6f18eb8108c48d17 category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I4AFRJ?from=project-issue CVE: NA --------------------------------------------------------------- tcp: adjust rto_base in retransmits_timed_out() The cited commit exposed an old retransmits_timed_out() bug which assumed it could call tcp_model_timeout() with TCP_RTO_MIN as rto_base for all states. But flows in SYN_SENT or SYN_RECV state uses a different RTO base (1 sec instead of 200 ms, unless BPF choses another value) This caused a reduction of SYN retransmits from 6 to 4 with the default /proc/sys/net/ipv4/tcp_syn_retries value. Fixes: a41e8a8 ("tcp: better handle TCP_USER_TIMEOUT in SYN_SENT state") Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Yuchung Cheng <ycheng@google.com> Cc: Marek Majkowski <marek@cloudflare.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: jiazhenyuan <jiazhenyuan@uniontech.com> --- net/ipv4/tcp_timer.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c index 681882a40968..81a47e87c35d 100644 --- a/net/ipv4/tcp_timer.c +++ b/net/ipv4/tcp_timer.c @@ -190,7 +190,7 @@ static bool retransmits_timed_out(struct sock *sk, unsigned int boundary, unsigned int timeout) { - const unsigned int rto_base = TCP_RTO_MIN; + unsigned int rto_base = TCP_RTO_MIN; unsigned int linear_backoff_thresh, start_ts; if (!inet_csk(sk)->icsk_retransmits) @@ -201,6 +201,9 @@ static bool retransmits_timed_out(struct sock *sk, return false; if (likely(timeout == 0)) { + if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) + rto_base = tcp_timeout_init(sk); + linear_backoff_thresh = ilog2(TCP_RTO_MAX/rto_base); if (boundary <= linear_backoff_thresh) -- 2.27.0
Hi jiazhenyuan, The original commit 3256a2d6ab1f relay on commits 01a523b07161, 7ae189759cc4 and 9efdda4e3abe. The commits 7ae189759cc4 and 9efdda4e3abe solve the problems which are valuable. I think it is better to make a patchset including above commits to solve the problem. And, that can let openeuler codes has less difference with linux-stable as possible. Thank you!
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit?id... 这个补丁和社区原版补丁的标题不一样，而且作者也被更改 On 2021/9/17 21:51, jiazhenyuan wrote:
mainline inclusion from mainline-v5.3.0 commit 3256a2d6ab1f71f9a1bd2d7f6f18eb8108c48d17 category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I4AFRJ?from=project-issue CVE: NA
---------------------------------------------------------------
tcp: adjust rto_base in retransmits_timed_out() The cited commit exposed an old retransmits_timed_out() bug which assumed it could call tcp_model_timeout() with TCP_RTO_MIN as rto_base for all states.
But flows in SYN_SENT or SYN_RECV state uses a different RTO base (1 sec instead of 200 ms, unless BPF choses another value)
This caused a reduction of SYN retransmits from 6 to 4 with the default /proc/sys/net/ipv4/tcp_syn_retries value.
Fixes: a41e8a8 ("tcp: better handle TCP_USER_TIMEOUT in SYN_SENT state") Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Yuchung Cheng <ycheng@google.com> Cc: Marek Majkowski <marek@cloudflare.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: jiazhenyuan <jiazhenyuan@uniontech.com> --- net/ipv4/tcp_timer.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c index 681882a40968..81a47e87c35d 100644 --- a/net/ipv4/tcp_timer.c +++ b/net/ipv4/tcp_timer.c @@ -190,7 +190,7 @@ static bool retransmits_timed_out(struct sock *sk, unsigned int boundary, unsigned int timeout) { - const unsigned int rto_base = TCP_RTO_MIN; + unsigned int rto_base = TCP_RTO_MIN; unsigned int linear_backoff_thresh, start_ts;
if (!inet_csk(sk)->icsk_retransmits) @@ -201,6 +201,9 @@ static bool retransmits_timed_out(struct sock *sk, return false;
if (likely(timeout == 0)) { + if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) + rto_base = tcp_timeout_init(sk); + linear_backoff_thresh = ilog2(TCP_RTO_MAX/rto_base);
if (boundary <= linear_backoff_thresh)
participants (3)
-
jiazhenyuan -
Yang Yingliang -
Ziyang Xuan (William)