All of lore.kernel.org
 help / color / mirror / Atom feed
From: Ben Greear <greearb@candelatech.com>
To: Lk Overrun <lkoverrun@yahoo.com>
Cc: linux-kernel@vger.kernel.org
Subject: Re: Brust data send problem on gigabit NIC on Linux
Date: Tue, 22 Oct 2002 19:48:54 -0700	[thread overview]
Message-ID: <3DB60E16.2060401@candelatech.com> (raw)
In-Reply-To: 20021023015905.63415.qmail@web21501.mail.yahoo.com

Lk Overrun wrote:
> Hi, I am seeking advice on how to best send out huge
> number of packets on a gigabit ethernet interface.  I
> am using kernel 2.4.19.  I try to send out as many as
> possible 15Kbyte-long ethernet packets to try to
> utilize the giga-bit/sec bandwidth.  My CPU is really
> fast (2 GHz) amd I dump the packets to the interface
> in a  tight loop in user space.  However, I can only
> reach around 400 Mbits/sec before the packets get
> dropped.  The queue discipline (qdisc) seems to be
> responsible because the queue length (txqueuelen) is
> only 100 by default, and the queue just cannot store
> so many packets at once.  I can eliminate the packet
> drop by raising the queue length to somewhere like
> 60000 but that is not practical because it uses too
> much memory. It seems I need some delay between
> sending packets but I cannot sleep for less than 10 ms
> (1/Hz) in user space and 10 ms is too long.
> 
> I am using raw socket bypassing the IP stack and my
> NIC is the Intel Pro1000 (using the e1000.o driver).
> 
> What is the best way to send raw ethernet packets,
> reaching gigabit range withuut packet drop on Linux? 
> Thanks for any advice.

Make sure the e1000 driver is tuned, try insmodding it with:
TxDescriptors=1024 RxDescriptors=4096

Also, in your user-space app, check the return code of your
packet-sending call.  That can let you know that the kernel
dropped it, and that you need to re-send.

How do you know you are dropping packets?  (Ie, are you also
reading on another machine?)  Usually it's read that drops
more packets than write.

Good luck,
Ben

> 
>  
> 
> __________________________________________________
> Do you Yahoo!?
> Y! Web Hosting - Let the expert host your web site
> http://webhosting.yahoo.com/
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 


-- 
Ben Greear <greearb@candelatech.com>       <Ben_Greear AT excite.com>
President of Candela Technologies Inc      http://www.candelatech.com
ScryMUD:  http://scry.wanfear.com     http://scry.wanfear.com/~greear



  reply	other threads:[~2002-10-23  2:42 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2002-10-23  1:59 Brust data send problem on gigabit NIC on Linux Lk Overrun
2002-10-23  2:48 ` Ben Greear [this message]
  -- strict thread matches above, loose matches on Subject: below --
2002-10-24 16:39 Jon Burgess

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=3DB60E16.2060401@candelatech.com \
    --to=greearb@candelatech.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=lkoverrun@yahoo.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.