From: Eric Dumazet <dada1@cosmosbay•com>
To: David Miller <davem@davemloft•net>
Cc: kchang@athenacr•com, netdev@vger•kernel.org, cl@linux-foundation•org
Subject: Re: Multicast packet loss
Date: Wed, 04 Mar 2009 09:36:57 +0100 [thread overview]
Message-ID: <49AE3DA9.2020103@cosmosbay.com> (raw)
In-Reply-To: <20090304.001646.100690134.davem@davemloft.net>
David Miller a écrit :
> From: Eric Dumazet <dada1@cosmosbay•com>
> Date: Sat, 28 Feb 2009 09:51:11 +0100
>
>> David, this is a preliminary work, not meant for inclusion as is,
>> comments are welcome.
>>
>> [PATCH] net: sk_forward_alloc becomes an atomic_t
>>
>> Commit 95766fff6b9a78d11fc2d3812dd035381690b55d
>> (UDP: Add memory accounting) introduced a regression for high rate UDP flows,
>> because of extra lock_sock() in udp_recvmsg()
>>
>> In order to reduce need for lock_sock() in UDP receive path, we might need
>> to declare sk_forward_alloc as an atomic_t.
>>
>> udp_recvmsg() can avoid a lock_sock()/release_sock() pair.
>>
>> Signed-off-by: Eric Dumazet <dada1@cosmosbay•com>
>
> This adds new overhead for TCP which has to hold the socket
> lock for other reasons in these paths.
>
> I don't get how an atomic_t operation is cheaper than a
> lock_sock/release_sock. Is it the case that in many
> executions of these paths only atomic_read()'s are necessary?
>
> I actually think this scheme is racy. There is a reason we
> have to hold the socket lock when doing memory scheduling.
> Two threads can get in there and say "hey I have enough space
> already" even though only enough space is allocated for one
> of their requests.
>
> What did I miss? :)
>
I believe you are right, and in fact was about to post a "dont look at this patch"
since it doesnt help the multicast reception at all, I redone tests more carefuly
and got nothing but noise.
We have a cache line ping pong mess here, and need more thinking.
I rewrote Kenny prog to use non blocking sockets.
Receivers are doing :
int delay = 50;
fcntl(s, F_SETFL, O_NDELAY);
while(1)
{
struct sockaddr_in from;
socklen_t fromlen = sizeof(from);
res = recvfrom(s, buf, 1000, 0, (struct sockaddr*)&from, &fromlen);
if (res == -1) {
delay++;
usleep(delay);
continue;
}
if (delay > 40)
delay--;
++npackets;
With this litle user space change and 8 receivers on my dual quad core, softirqd
only takes 8% of one cpu and no drops at all (instead of 100% cpu and 30% drops)
So this is definitly a problem mixing scheduler cache line ping pongs with network
stack cache line ping pongs.
We could reorder fields so that fewer cache lines are touched by the softirq processing,
I tried this but still got packet drops.
next prev parent reply other threads:[~2009-03-04 8:37 UTC|newest]
Thread overview: 70+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-01-30 17:49 Multicast packet loss Kenny Chang
2009-01-30 19:04 ` Eric Dumazet
2009-01-30 19:17 ` Denys Fedoryschenko
2009-01-30 20:03 ` Neil Horman
2009-01-30 22:29 ` Kenny Chang
2009-01-30 22:41 ` Eric Dumazet
2009-01-31 16:03 ` Neil Horman
2009-02-02 16:13 ` Kenny Chang
2009-02-02 16:48 ` Kenny Chang
2009-02-03 11:55 ` Neil Horman
2009-02-03 15:20 ` Kenny Chang
2009-02-04 1:15 ` Neil Horman
2009-02-04 16:07 ` Kenny Chang
2009-02-04 16:46 ` Wesley Chow
2009-02-04 18:11 ` Eric Dumazet
2009-02-05 13:33 ` Neil Horman
2009-02-05 13:46 ` Wesley Chow
2009-02-05 13:29 ` Neil Horman
2009-02-01 12:40 ` Eric Dumazet
2009-02-02 13:45 ` Neil Horman
2009-02-02 16:57 ` Eric Dumazet
2009-02-02 18:22 ` Neil Horman
2009-02-02 19:51 ` Wes Chow
2009-02-02 20:29 ` Eric Dumazet
2009-02-02 21:09 ` Wes Chow
2009-02-02 21:31 ` Eric Dumazet
2009-02-03 17:34 ` Kenny Chang
2009-02-04 1:21 ` Neil Horman
2009-02-26 17:15 ` Kenny Chang
2009-02-28 8:51 ` Eric Dumazet
2009-03-01 17:03 ` Eric Dumazet
2009-03-04 8:16 ` David Miller
2009-03-04 8:36 ` Eric Dumazet [this message]
2009-03-07 7:46 ` Eric Dumazet
2009-03-08 16:46 ` Eric Dumazet
2009-03-09 2:49 ` David Miller
2009-03-09 6:36 ` Eric Dumazet
2009-03-13 21:51 ` David Miller
2009-03-13 22:30 ` Eric Dumazet
2009-03-13 22:38 ` David Miller
2009-03-13 22:45 ` Eric Dumazet
2009-03-14 9:03 ` [PATCH] net: reorder fields of struct socket Eric Dumazet
2009-03-16 2:59 ` David Miller
2009-03-16 22:22 ` Multicast packet loss Eric Dumazet
2009-03-17 10:11 ` Peter Zijlstra
2009-03-17 11:08 ` Eric Dumazet
2009-03-17 11:57 ` Peter Zijlstra
2009-03-17 15:00 ` Brian Bloniarz
2009-03-17 15:16 ` Eric Dumazet
2009-03-17 19:39 ` David Stevens
2009-03-17 21:19 ` Eric Dumazet
2009-04-03 19:28 ` Brian Bloniarz
2009-04-05 13:49 ` Eric Dumazet
2009-04-06 21:53 ` Brian Bloniarz
2009-04-06 22:12 ` Brian Bloniarz
2009-04-07 20:08 ` Brian Bloniarz
2009-04-08 8:12 ` Eric Dumazet
2009-03-09 22:56 ` Brian Bloniarz
2009-03-10 5:28 ` Eric Dumazet
2009-03-10 23:22 ` Brian Bloniarz
2009-03-11 3:00 ` Eric Dumazet
2009-03-12 15:47 ` Brian Bloniarz
2009-03-12 16:34 ` Eric Dumazet
2009-02-27 18:40 ` Christoph Lameter
2009-02-27 18:56 ` Eric Dumazet
2009-02-27 19:45 ` Christoph Lameter
2009-02-27 20:12 ` Eric Dumazet
2009-02-27 21:36 ` Eric Dumazet
2009-02-02 13:53 ` Eric Dumazet
-- strict thread matches above, loose matches on Subject: below --
2009-04-05 14:42 bmb
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=49AE3DA9.2020103@cosmosbay.com \
--to=dada1@cosmosbay$(echo .)com \
--cc=cl@linux-foundation$(echo .)org \
--cc=davem@davemloft$(echo .)net \
--cc=kchang@athenacr$(echo .)com \
--cc=netdev@vger$(echo .)kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
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