* [PATCH] tcp: tcp_mark_head_lost is only valid for sack-tcp
@ 2020-05-07 3:08 zhang kai
2020-05-07 13:58 ` Neal Cardwell
2020-05-08 0:58 ` David Miller
0 siblings, 2 replies; 4+ messages in thread
From: zhang kai @ 2020-05-07 3:08 UTC (permalink / raw)
To: ycheng, edumazet, ncardwell, davem, kuznet, yoshfuji, kuba, netdev
so tcp_is_sack/reno checks are removed from tcp_mark_head_lost.
Signed-off-by: zhang kai <zhangkaiheb@126.com>
---
net/ipv4/tcp_input.c | 32 +++++++-------------------------
1 file changed, 7 insertions(+), 25 deletions(-)
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index b996dc106..c306becf6 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -2183,8 +2183,7 @@ static bool tcp_time_to_recover(struct sock *sk, int flag)
}
/* Detect loss in event "A" above by marking head of queue up as lost.
- * For non-SACK(Reno) senders, the first "packets" number of segments
- * are considered lost. For RFC3517 SACK, a segment is considered lost if it
+ * For RFC3517 SACK, a segment is considered lost if it
* has at least tp->reordering SACKed seqments above it; "packets" refers to
* the maximum SACKed segments to pass before reaching this limit.
*/
@@ -2192,10 +2191,9 @@ static void tcp_mark_head_lost(struct sock *sk, int packets, int mark_head)
{
struct tcp_sock *tp = tcp_sk(sk);
struct sk_buff *skb;
- int cnt, oldcnt, lost;
- unsigned int mss;
+ int cnt;
/* Use SACK to deduce losses of new sequences sent during recovery */
- const u32 loss_high = tcp_is_sack(tp) ? tp->snd_nxt : tp->high_seq;
+ const u32 loss_high = tp->snd_nxt;
WARN_ON(packets > tp->packets_out);
skb = tp->lost_skb_hint;
@@ -2218,26 +2216,11 @@ static void tcp_mark_head_lost(struct sock *sk, int packets, int mark_head)
if (after(TCP_SKB_CB(skb)->end_seq, loss_high))
break;
- oldcnt = cnt;
- if (tcp_is_reno(tp) ||
- (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED))
+ if (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED)
cnt += tcp_skb_pcount(skb);
- if (cnt > packets) {
- if (tcp_is_sack(tp) ||
- (TCP_SKB_CB(skb)->sacked & TCPCB_SACKED_ACKED) ||
- (oldcnt >= packets))
- break;
-
- mss = tcp_skb_mss(skb);
- /* If needed, chop off the prefix to mark as lost. */
- lost = (packets - oldcnt) * mss;
- if (lost < skb->len &&
- tcp_fragment(sk, TCP_FRAG_IN_RTX_QUEUE, skb,
- lost, mss, GFP_ATOMIC) < 0)
- break;
- cnt = packets;
- }
+ if (cnt > packets)
+ break;
tcp_skb_mark_lost(tp, skb);
@@ -2849,8 +2832,7 @@ static void tcp_fastretrans_alert(struct sock *sk, const u32 prior_snd_una,
if (tcp_try_undo_partial(sk, prior_snd_una))
return;
/* Partial ACK arrived. Force fast retransmit. */
- do_lost = tcp_is_reno(tp) ||
- tcp_force_fast_retransmit(sk);
+ do_lost = tcp_force_fast_retransmit(sk);
}
if (tcp_try_undo_dsack(sk)) {
tcp_try_keep_open(sk);
--
2.17.1
^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [PATCH] tcp: tcp_mark_head_lost is only valid for sack-tcp
2020-05-07 3:08 [PATCH] tcp: tcp_mark_head_lost is only valid for sack-tcp zhang kai
@ 2020-05-07 13:58 ` Neal Cardwell
2020-05-07 14:14 ` Eric Dumazet
2020-05-08 0:58 ` David Miller
1 sibling, 1 reply; 4+ messages in thread
From: Neal Cardwell @ 2020-05-07 13:58 UTC (permalink / raw)
To: zhang kai
Cc: Yuchung Cheng, Eric Dumazet, David Miller, Alexey Kuznetsov,
Hideaki YOSHIFUJI, Jakub Kicinski, Netdev
On Wed, May 6, 2020 at 11:09 PM zhang kai <zhangkaiheb@126.com> wrote:
>
> so tcp_is_sack/reno checks are removed from tcp_mark_head_lost.
>
> Signed-off-by: zhang kai <zhangkaiheb@126.com>
> ---
Nice clean-up, thanks.
Acked-by: Neal Cardwell <ncardwell@google.com>
As Eric noted in previous threads, this clean-up is now possible after:
6ac06ecd3a5d1dd1aaea5c2a8f6d6e4c81d5de6a ("tcp: simpler NewReno implementation")
As Yuchung noted, the tcp_mark_head_lost() function is not used today
when RACK is enabled (which is by
default).
neal
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] tcp: tcp_mark_head_lost is only valid for sack-tcp
2020-05-07 13:58 ` Neal Cardwell
@ 2020-05-07 14:14 ` Eric Dumazet
0 siblings, 0 replies; 4+ messages in thread
From: Eric Dumazet @ 2020-05-07 14:14 UTC (permalink / raw)
To: Neal Cardwell
Cc: zhang kai, Yuchung Cheng, David Miller, Alexey Kuznetsov,
Hideaki YOSHIFUJI, Jakub Kicinski, Netdev
On Thu, May 7, 2020 at 6:58 AM Neal Cardwell <ncardwell@google.com> wrote:
>
> On Wed, May 6, 2020 at 11:09 PM zhang kai <zhangkaiheb@126.com> wrote:
> >
> > so tcp_is_sack/reno checks are removed from tcp_mark_head_lost.
> >
> > Signed-off-by: zhang kai <zhangkaiheb@126.com>
> > ---
>
> Nice clean-up, thanks.
>
> Acked-by: Neal Cardwell <ncardwell@google.com>
>
> As Eric noted in previous threads, this clean-up is now possible after:
>
> 6ac06ecd3a5d1dd1aaea5c2a8f6d6e4c81d5de6a ("tcp: simpler NewReno implementation")
>
> As Yuchung noted, the tcp_mark_head_lost() function is not used today
> when RACK is enabled (which is by
> default).
>
> neal
Yes, I was hoping for a bit more detailed changelog, it always helps
for future reference.
Thank you
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH] tcp: tcp_mark_head_lost is only valid for sack-tcp
2020-05-07 3:08 [PATCH] tcp: tcp_mark_head_lost is only valid for sack-tcp zhang kai
2020-05-07 13:58 ` Neal Cardwell
@ 2020-05-08 0:58 ` David Miller
1 sibling, 0 replies; 4+ messages in thread
From: David Miller @ 2020-05-08 0:58 UTC (permalink / raw)
To: zhangkaiheb; +Cc: ycheng, edumazet, ncardwell, kuznet, yoshfuji, kuba, netdev
From: zhang kai <zhangkaiheb@126.com>
Date: Thu, 7 May 2020 11:08:30 +0800
> so tcp_is_sack/reno checks are removed from tcp_mark_head_lost.
>
> Signed-off-by: zhang kai <zhangkaiheb@126.com>
Applied to net-next, thanks.
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2020-05-08 0:58 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-05-07 3:08 [PATCH] tcp: tcp_mark_head_lost is only valid for sack-tcp zhang kai
2020-05-07 13:58 ` Neal Cardwell
2020-05-07 14:14 ` Eric Dumazet
2020-05-08 0:58 ` David Miller
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).