[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Long mbuf chains
- To: freebsd-questions_(_at_)_freebsd_(_dot_)_org
- Subject: Long mbuf chains
- From: "Yony Yossef" <yonyossef_(_dot_)_lists_(_at_)_gmail_(_dot_)_com>
- Date: Wed, 6 Aug 2008 16:36:28 +0300
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:received:received:message-id:date:from:to :subject:mime-version:content-type; bh=xAxV9cvNfsobteCFnNAbZKYlrkQaIN5sSdKGiDIPD2U=; b=i5wkIlOQ1ZPsUy2y7jZZFvmRih3TUvdWfckvjDeClZiSPZySM1phrXs4ZcU1lEV5X5 Wpj1s9kyCi3ve6iERvp0PzB1WoKGBGiRBAa6FbJeYjTR/i9/o0RAzzwKcm4s5wYB7r/o 61HrJdgtsIm6j3/WHzlFp42vVLSujCkyj8I24=
Hi All
I'm working on an Ethernet driver for FreeBSD 7.0.
Taking network performance numbers I encountered very long mbuf chains on
the sender side.
The symptom is constant, always during iperf/netperf TCP stream tests with
message sizes of 128 bytes (>200 mbufs per chain),
1024 bytes (30-60 mbufs per chain) and 2048 bytes.
My problem is that long chains require some kind of defragmentation/cutting
before it can be properly DMAd.
This is pretty a expansive operation.
1.
Is there a way of tuning the OS for sending limited length mbuf chains? I
thought setting "net.inet.ip.maxfragsperpacket" would do it but it doesn't.
2.
Is there a better way of handling this issue?
Thanks,
Yony
_______________________________________________
freebsd-questions_(_at_)_freebsd_(_dot_)_org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to "freebsd-questions-unsubscribe_(_at_)_freebsd_(_dot_)_org"
Visit your host, monkey.org