* [PATCH net-next] tcp: make tcp_rcv_state_process() drop monitor friendly
@ 2022-03-23 13:04 Jianguo Wu
2022-03-23 13:40 ` Eric Dumazet
0 siblings, 1 reply; 3+ messages in thread
From: Jianguo Wu @ 2022-03-23 13:04 UTC (permalink / raw)
To: dsahern, Jakub Kicinski, menglong8.dong; +Cc: netdev, linux-kernel, edumazet
From: Jianguo Wu <wujianguo@chinatelecom.cn>
In tcp_rcv_state_process(), should not call tcp_drop() for same case,
like after process ACK packet in TCP_LAST_ACK state, it should call
consume_skb() instead of tcp_drop() to be drop monitor friendly,
otherwise every last ack will be report as dropped packet by drop monitor.
Signed-off-by: Jianguo Wu <wujianguo@chinatelecom.cn>
---
net/ipv4/tcp_input.c | 9 ++++++---
1 file changed, 6 insertions(+), 3 deletions(-)
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index 2088f93..feb6f83 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -6574,7 +6574,8 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb)
inet_csk_reset_keepalive_timer(sk, tmo);
} else {
tcp_time_wait(sk, TCP_FIN_WAIT2, tmo);
- goto discard;
+ consume_skb(skb);
+ return 0;
}
break;
}
@@ -6582,7 +6583,8 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb)
case TCP_CLOSING:
if (tp->snd_una == tp->write_seq) {
tcp_time_wait(sk, TCP_TIME_WAIT, 0);
- goto discard;
+ consume_skb(skb);
+ return 0;
}
break;
@@ -6590,7 +6592,8 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb)
if (tp->snd_una == tp->write_seq) {
tcp_update_metrics(sk);
tcp_done(sk);
- goto discard;
+ consume_skb(skb);
+ return 0;
}
break;
}
--
1.8.3.1
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [PATCH net-next] tcp: make tcp_rcv_state_process() drop monitor friendly
2022-03-23 13:04 [PATCH net-next] tcp: make tcp_rcv_state_process() drop monitor friendly Jianguo Wu
@ 2022-03-23 13:40 ` Eric Dumazet
2022-03-24 2:33 ` Jianguo Wu
0 siblings, 1 reply; 3+ messages in thread
From: Eric Dumazet @ 2022-03-23 13:40 UTC (permalink / raw)
To: Jianguo Wu; +Cc: David Ahern, Jakub Kicinski, Menglong Dong, netdev, LKML
On Wed, Mar 23, 2022 at 6:05 AM Jianguo Wu <wujianguo106@163.com> wrote:
>
> From: Jianguo Wu <wujianguo@chinatelecom.cn>
>
> In tcp_rcv_state_process(), should not call tcp_drop() for same case,
> like after process ACK packet in TCP_LAST_ACK state, it should call
> consume_skb() instead of tcp_drop() to be drop monitor friendly,
> otherwise every last ack will be report as dropped packet by drop monitor.
>
> Signed-off-by: Jianguo Wu <wujianguo@chinatelecom.cn>
> ---
1) net-next is closed
2) Same remarks as for the other patch.
You mark the packet as consumed, while maybe we had to throw away
some payload from it ?
You will have to wait for net-next being open,
then send patches with one change at a time, with clear explanations
and possibly packetdrill tests.
I am concerned about all these patches making future backports
difficult because of merge conflicts.
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH net-next] tcp: make tcp_rcv_state_process() drop monitor friendly
2022-03-23 13:40 ` Eric Dumazet
@ 2022-03-24 2:33 ` Jianguo Wu
0 siblings, 0 replies; 3+ messages in thread
From: Jianguo Wu @ 2022-03-24 2:33 UTC (permalink / raw)
To: Eric Dumazet; +Cc: David Ahern, Jakub Kicinski, Menglong Dong, netdev, LKML
Hi,
Thanks for your reply. This is more complicated than I thought, i will do some more dig.
在 2022/3/23 21:40, Eric Dumazet 写道:
> On Wed, Mar 23, 2022 at 6:05 AM Jianguo Wu <wujianguo106@163.com> wrote:
>>
>> From: Jianguo Wu <wujianguo@chinatelecom.cn>
>>
>> In tcp_rcv_state_process(), should not call tcp_drop() for same case,
>> like after process ACK packet in TCP_LAST_ACK state, it should call
>> consume_skb() instead of tcp_drop() to be drop monitor friendly,
>> otherwise every last ack will be report as dropped packet by drop monitor.
>>
>> Signed-off-by: Jianguo Wu <wujianguo@chinatelecom.cn>
>> ---
>
> 1) net-next is closed
>
> 2) Same remarks as for the other patch.
> You mark the packet as consumed, while maybe we had to throw away
> some payload from it ?
>
> You will have to wait for net-next being open,
> then send patches with one change at a time, with clear explanations
> and possibly packetdrill tests.
>
> I am concerned about all these patches making future backports
> difficult because of merge conflicts.
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2022-03-24 2:35 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-03-23 13:04 [PATCH net-next] tcp: make tcp_rcv_state_process() drop monitor friendly Jianguo Wu
2022-03-23 13:40 ` Eric Dumazet
2022-03-24 2:33 ` Jianguo Wu
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).