[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20210311203506.3450792-2-eric.dumazet@gmail.com>
Date: Thu, 11 Mar 2021 12:35:04 -0800
From: Eric Dumazet <eric.dumazet@...il.com>
To: "David S . Miller" <davem@...emloft.net>,
Jakub Kicinski <kuba@...nel.org>
Cc: netdev <netdev@...r.kernel.org>,
Eric Dumazet <edumazet@...gle.com>,
Eric Dumazet <eric.dumazet@...il.com>,
Neal Cardwell <ncardwell@...gle.com>,
Yuchung Cheng <ycheng@...gle.com>,
Neil Spring <ntspring@...com>
Subject: [PATCH net-next 1/3] tcp: plug skb_still_in_host_queue() to TSQ
From: Eric Dumazet <edumazet@...gle.com>
Jakub and Neil reported an increase of RTO timers whenever
TX completions are delayed a bit more (by increasing
NIC TX coalescing parameters)
Main issue is that TCP stack has a logic preventing a packet
being retransmit if the prior clone has not yet been
orphaned or freed.
This logic came with commit 1f3279ae0c13 ("tcp: avoid
retransmits of TCP packets hanging in host queues")
Thankfully, in the case skb_still_in_host_queue() detects
the initial clone is still in flight, it can use TSQ logic
that will eventually retry later, at the moment the clone
is freed or orphaned.
Signed-off-by: Eric Dumazet <edumazet@...gle.com>
Reported-by: Neil Spring <ntspring@...com>
Reported-by: Jakub Kicinski <kuba@...nel.org>
Cc: Neal Cardwell <ncardwell@...gle.com>
Cc: Yuchung Cheng <ycheng@...gle.com>
---
include/linux/skbuff.h | 2 +-
net/ipv4/tcp_output.c | 12 ++++++++----
2 files changed, 9 insertions(+), 5 deletions(-)
diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index 0503c917d77301f433122bf34a659bb855763144..483e89348f78b48235748de37ae3ea7ec9450491 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -1140,7 +1140,7 @@ static inline bool skb_fclone_busy(const struct sock *sk,
return skb->fclone == SKB_FCLONE_ORIG &&
refcount_read(&fclones->fclone_ref) > 1 &&
- fclones->skb2.sk == sk;
+ READ_ONCE(fclones->skb2.sk) == sk;
}
/**
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index fbf140a770d8e21b936369b79abbe9857537acd8..0dbf208a4f2f17c630084e87f4a9a2ad0dc24168 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -2775,13 +2775,17 @@ bool tcp_schedule_loss_probe(struct sock *sk, bool advancing_rto)
* a packet is still in a qdisc or driver queue.
* In this case, there is very little point doing a retransmit !
*/
-static bool skb_still_in_host_queue(const struct sock *sk,
+static bool skb_still_in_host_queue(struct sock *sk,
const struct sk_buff *skb)
{
if (unlikely(skb_fclone_busy(sk, skb))) {
- NET_INC_STATS(sock_net(sk),
- LINUX_MIB_TCPSPURIOUS_RTX_HOSTQUEUES);
- return true;
+ set_bit(TSQ_THROTTLED, &sk->sk_tsq_flags);
+ smp_mb__after_atomic();
+ if (skb_fclone_busy(sk, skb)) {
+ NET_INC_STATS(sock_net(sk),
+ LINUX_MIB_TCPSPURIOUS_RTX_HOSTQUEUES);
+ return true;
+ }
}
return false;
}
--
2.31.0.rc2.261.g7f71774620-goog
Powered by blists - more mailing lists