From: "David S. Miller" <davem@davemloft.net>
To: Andrew Grover <andy.grover@gmail.com>
Cc: ak@suse.de, anton@samba.org, netdev@oss.sgi.com
Subject: Re: bad TSO performance in 2.6.9-rc2-BK
Date: Wed, 22 Sep 2004 13:39:06 -0700 [thread overview]
Message-ID: <20040922133906.7d57ef49.davem@davemloft.net> (raw)
In-Reply-To: <c0a09e5c04092213126ba5f29d@mail.gmail.com>
On Wed, 22 Sep 2004 13:12:47 -0700
Andrew Grover <andy.grover@gmail.com> wrote:
> I think this is caused by the congestion changes added between
> 2.6.9-rc1 and rc2. I am seeing good performance with rc1 and tso on
> (or off), but bad performance with rc2 and bk-latest only if tso is
> on.
Yes, we know it is the packet counting change but those
are pretty much necessary else TSO violates congestion window
rules.
There is a slight bug somewhere limiting performance, but we'll
find it.
The thing to watch if you're debugging this is what tp->packets_out
is set to, and what happens in each tcp_snd_test() call. Also, watch
tcp_cong_avoid() to make sure that tp->snd_cwnd is incrementing
on every ACK and that tp->snd_cwnd_clamp is not some silly small
value.
Tracking tcp_snd_test() should tell you everything, watch what
happens to the values of:
tcp_packets_in_flight(tp);
'pkts' aka. TCP_SKB_CB(skb)->tso_factor after the
possible tcp_set_skb_tso_factor() call
tp->snd_cwnd
One thing that could be biting us is the nagle check which
probably needs to be adjusted to use the standard MSS not
the TSO one... perhaps play with this patch:
===== include/net/tcp.h 1.88 vs edited =====
--- 1.88/include/net/tcp.h 2004-09-14 13:57:07 -07:00
+++ edited/include/net/tcp.h 2004-09-22 13:18:43 -07:00
@@ -1505,7 +1505,7 @@
* final FIN frame. -DaveM
*/
return (((nonagle&TCP_NAGLE_PUSH) || tp->urg_mode
- || !tcp_nagle_check(tp, skb, cur_mss, nonagle)) &&
+ || !tcp_nagle_check(tp, skb, tp->mss_cache_std, nonagle)) &&
(((tcp_packets_in_flight(tp) + (pkts-1)) < tp->snd_cwnd) ||
(TCP_SKB_CB(skb)->flags & TCPCB_FLAG_FIN)) &&
!after(TCP_SKB_CB(skb)->end_seq, tp->snd_una + tp->snd_wnd));
next prev parent reply other threads:[~2004-09-22 20:39 UTC|newest]
Thread overview: 97+ messages / expand[flat|nested] mbox.gz Atom feed top
2004-09-20 6:30 bad TSO performance in 2.6.9-rc2-BK Anton Blanchard
2004-09-20 15:54 ` Nivedita Singhvi
2004-09-21 15:55 ` Anton Blanchard
2004-09-20 20:30 ` Andi Kleen
2004-09-21 22:58 ` David S. Miller
2004-09-22 14:00 ` Andi Kleen
2004-09-22 18:12 ` David S. Miller
2004-09-22 19:55 ` Andi Kleen
2004-09-22 20:07 ` Nivedita Singhvi
2004-09-22 20:30 ` David S. Miller
2004-09-22 20:56 ` Nivedita Singhvi
2004-09-22 21:56 ` Andi Kleen
2004-09-22 22:04 ` David S. Miller
2004-09-22 20:12 ` Andrew Grover
2004-09-22 20:39 ` David S. Miller [this message]
2004-09-22 22:06 ` Andi Kleen
2004-09-22 22:25 ` David S. Miller
2004-09-22 22:47 ` Andi Kleen
2004-09-22 22:50 ` David S. Miller
2004-09-23 23:11 ` David S. Miller
2004-09-23 23:41 ` Herbert Xu
2004-09-23 23:41 ` David S. Miller
2004-09-24 0:12 ` Herbert Xu
2004-09-24 0:40 ` Herbert Xu
2004-09-24 1:07 ` Herbert Xu
2004-09-24 1:17 ` David S. Miller
2004-09-27 1:27 ` Herbert Xu
2004-09-27 2:50 ` Herbert Xu
2004-09-27 4:00 ` David S. Miller
2004-09-27 5:45 ` Herbert Xu
2004-09-27 19:01 ` David S. Miller
2004-09-27 21:32 ` Herbert Xu
2004-09-28 21:10 ` David S. Miller
2004-09-28 21:34 ` Andi Kleen
2004-09-28 21:53 ` David S. Miller
2004-09-28 22:33 ` Andi Kleen
2004-09-28 22:57 ` David S. Miller
2004-09-28 23:27 ` Andi Kleen
2004-09-28 23:35 ` David S. Miller
2004-09-28 23:55 ` Andi Kleen
2004-09-29 0:04 ` David S. Miller
2004-09-29 20:58 ` John Heffner
2004-09-29 21:10 ` Nivedita Singhvi
2004-09-29 21:50 ` David S. Miller
2004-09-29 21:56 ` Andi Kleen
2004-09-29 23:29 ` David S. Miller
2004-09-29 23:51 ` John Heffner
2004-09-30 0:03 ` David S. Miller
2004-09-30 0:10 ` Herbert Xu
2004-10-01 0:34 ` David S. Miller
2004-10-01 1:12 ` David S. Miller
2004-10-01 3:40 ` David S. Miller
2004-10-01 10:35 ` Andi Kleen
2004-10-01 10:23 ` Andi Kleen
2004-09-30 0:10 ` John Heffner
2004-09-30 17:25 ` John Heffner
2004-09-30 20:23 ` David S. Miller
2004-09-30 0:05 ` Herbert Xu
2004-09-30 4:33 ` David S. Miller
2004-09-30 5:47 ` Herbert Xu
2004-09-30 7:39 ` David S. Miller
2004-09-30 8:09 ` Herbert Xu
2004-09-30 9:29 ` Andi Kleen
2004-09-30 20:20 ` David S. Miller
2004-09-29 3:27 ` John Heffner
2004-09-29 9:01 ` Andi Kleen
2004-09-29 19:56 ` David S. Miller
2004-09-29 20:56 ` Andi Kleen
2004-09-29 21:17 ` David S. Miller
2004-09-29 21:00 ` David S. Miller
2004-09-29 21:16 ` Nivedita Singhvi
2004-09-29 21:22 ` David S. Miller
2004-09-29 21:43 ` Andi Kleen
2004-09-29 21:51 ` John Heffner
2004-09-29 21:52 ` David S. Miller
2004-09-24 8:30 ` Andi Kleen
2004-09-27 22:38 ` John Heffner
2004-09-27 23:04 ` David S. Miller
2004-09-27 23:25 ` Andi Kleen
2004-09-27 23:37 ` David S. Miller
2004-09-27 23:51 ` Andi Kleen
2004-09-28 0:15 ` David S. Miller
2004-09-27 23:36 ` Herbert Xu
2004-09-28 0:13 ` David S. Miller
2004-09-28 0:34 ` Herbert Xu
2004-09-28 4:59 ` David S. Miller
2004-09-28 5:15 ` Herbert Xu
2004-09-28 5:58 ` David S. Miller
2004-09-28 6:45 ` Nivedita Singhvi
2004-09-28 7:20 ` Nivedita Singhvi
2004-09-28 20:38 ` David S. Miller
2004-09-28 7:23 ` Nivedita Singhvi
2004-09-28 8:23 ` Herbert Xu
2004-09-28 12:53 ` John Heffner
2004-09-22 20:28 ` David S. Miller
[not found] <Pine.NEB.4.33.0409301625560.13549-100000@dexter.psc.edu>
2004-10-02 1:32 ` John Heffner
2004-10-04 20:07 ` David S. Miller
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20040922133906.7d57ef49.davem@davemloft.net \
--to=davem@davemloft.net \
--cc=ak@suse.de \
--cc=andy.grover@gmail.com \
--cc=anton@samba.org \
--cc=netdev@oss.sgi.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).