From: Eric Dumazet <dada1@cosmosbay.com>
To: Dave Boutcher <daveboutcher@gmail.com>
Cc: netdev@vger.kernel.org
Subject: Re: IGMP Join dropping multicast packets
Date: Sun, 15 Mar 2009 03:37:09 +0100 [thread overview]
Message-ID: <49BC69D5.5000002@cosmosbay.com> (raw)
In-Reply-To: <91bdcedb0903141316j2dbf4160wb348a5a9e3bde8ad@mail.gmail.com>
Dave Boutcher a écrit :
> I'm running into an interesting problem with joining multiple
> multicast feeds. If you join multiple multicast feeds using
> setsockopt(...,IP_ADD_MEMBERSHIP...) it causes packets on UNRELATED
> multicast feeds to get dropped. We have a multicast feed on a rock
> solid network, and we were very surprised to see dropped packets. The
> cause was a different process/program being run by a different user
> joining a bunch of mulitcast feeds.
>
> I can recreate this with a fairly simple testcase (attached below.)
> The problem doesn't happen with unicast UDP data, and it doesn't
> happen with loopback, so you need at least two systems to run this
> (and what subscriber to netdev doesn't have at least two systems.) To
> recreate, run "receiver" on one system, "sender", on another, and then
> "joiner" on the receiving system. You should see a message pop out
> saying that packets have been dropped. I've recreated this on a few
> different kernel versions (the latest being 2.6.28) and a few
> different sets off hardware. I HAVEN"T recreated it if the system
> doing the IP_ADD_MEMBERSHIP specifies a specific interface rather than
> INADDR_ANY. I'm not sure if that is core to the issue or not. You
> may also need to bump the value in
> /proc/sys/net/ipv4/igmp_max_memberships (though that hasn't seemed
> necessary for me.)
>
> I poked around in igmp.c, but its mojo exceeds my threshold. If
> anyone has any ideas or questions I'd be happy to hear them.
>
I could not reproduce the problem on my machines (bnx2 adapter), even if changing
NUMSOCK from 55 to 200 in joiner.c
Is your network a 100Mb one or Gigabit ?
Try to slow down your joiner ?
(Could be a flood of IGMP messages your router/switch cannot cope with)
Please describe your "rock solid" network setup (kind of network adapters you have, kind of router...)
Each time an address is added, NIC driver have to reprogram mcfilter of
the device. Maybe some NIC can drop some packets at this moment...
If using tcpdump to force promiscuous mode on the device also triggers packet losses ?
(see also ifconfig ethX promisc|allmulti)
next prev parent reply other threads:[~2009-03-15 2:37 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-03-14 20:16 IGMP Join dropping multicast packets Dave Boutcher
2009-03-15 2:37 ` Eric Dumazet [this message]
2009-03-16 2:04 ` Dave Boutcher
2009-03-16 19:01 ` Eric Dumazet
2009-03-17 7:08 ` Eric Dumazet
2009-03-18 3:50 ` Dave Boutcher
2009-03-18 7:38 ` Eric Dumazet
2009-03-18 17:24 ` Brandeburg, Jesse
2009-03-19 1:48 ` Dave Boutcher
2009-03-19 1:51 ` Dave Boutcher
2009-03-20 20:36 ` Brandeburg, Jesse
2009-03-19 5:46 ` David 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=49BC69D5.5000002@cosmosbay.com \
--to=dada1@cosmosbay.com \
--cc=daveboutcher@gmail.com \
--cc=netdev@vger.kernel.org \
/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).