[BusyBox] dhcpc: broadcast flag

Brenda Butler bbutler at symbium.com
Fri Dec 10 03:12:06 UTC 2004


 
Sorry for the delay in updates, we had a few fires to fight.

It turns out the new busybox does not fix the problem.
Neither does setting the broadcast bit in the DHCP packet, 
nor setting the hardware type to 6 (instead of the default 
1), nor doing all these things at once.

Now the principal investigator has seen what I have seen, 
that is that although udhcp reports on the console that it is 
sending DISCOVER packets, only the 5th or 6th one actually 
shows up on the network.  He used to say that all the packets 
went out, but the managed switch wasn't responding due to the 
spanning tree algorithm back-off timeouts, and/or the packets 
were corrupted.  I think that was when ARMmon was doing the 
networking, rather than when busybox was doing it.

I forgot to answer the question "is udhcp in busybox 
1.00-pre3 the same version as the one in busybox-1.00" in my 
last reply.  The answer is yes, it is the same as far as I can tell.

About the link negotiation, we're having trouble with that... 
We still don't know why.  But I wouldn't have thought that 
would cause problems with the DISCOVER packets, they don't go 
out till the link negotiation is done with (whatever it turns 
out to come to).

Oh well.

cheerio,
bjb           bbutler at symbium.com 

> -----Original Message-----
> From: Brenda Butler 
> Sent: Friday, November 26, 2004 2:25 PM
> To: busybox at mail.codepoet.org
> Subject: Re: [BusyBox] dhcpc: broadcast flag
> 
> 
> 
> Thanks for the long response and references to standards,
> much appreciated.
> 
> We haven't even tried any busybox changes yet, the
> latest theory is that the Dell switch on which we
> see these problems is misconfigured to use spanning
> tree on end-points like our device (as opposed to
> other managed switches).
> 
> I'll keep you-all posted.
> 
> I'm having the principal investigator try out busybox
> 1.00 anyway :-)  I just now made up a rootfs with it and
> delivered it to him.
> 
> 
> 
> On Fri, Nov 26, 2004 at 08:31:38AM -0800, Doug Kehn wrote:
> > --- Brenda Butler wrote:
> > > 
> > > 
> > > We've been having some flaky networking happening.
> > > 
> > > We are using Busybox 1.00-pre3 but I see the code
> > > for
> > > creating the DISCOVER packets (init_header) is
> > > pretty
> > > much the same for busybox 1.00.
> > > 
> > 
> > Does BusyBox 1.00-pre3 also include udhcp 0.9.9-pre or
> > an earlier udhcp version?
> > 
> > > The flaky networking consists of:
> > > 
> > > software sends out DHCP DISCOVER
> > >    - sometimes it doesn't get to the network
> > > (observed
> > >         from another machine on hub)
> > 
> > Is the interface up?  If you are not seeing any
> > DISCOVER packets a quick ifconfig check will tell you
> > if the interface is up.  Also is udhcp outputing
> > 'Sending discover...'?
> > 
> > >    - sometimes it gets to the network (observed from
> > >         another machine on hub) and isn't answered
> > >         until the 5th or 6th or later try
> > 
> > I would check the sever logs to see why it isn't
> > replying.
> > 
> > >    - sometimes it gets to the server and a reply
> > > (DHCPOFFER)
> > >         is sent first try, and networking proceeds
> > > normally.
> > > 
> > 
> > I take it that nothing has changed.  For some reason
> > it just decided to work?
> > 
> > > 
> > > Usually when it fails, it consistently fails.  But
> > > another
> > > card in the same slot (same environment) will
> > > succeed.
> > > The same card in other environment usually succeeds.
> > > But same card back in same slot still fails.
> > > 
> > > The suggestions we've had so far are:
> > > 
> > > - make sure the broadcast bit is set
> > 
> > The broadcast bit tells the _server_ to broadcast
> > replies back to the client.  This isn't the problem
> > because, as you said above, dhcp does work (just not
> > consistently).
> > 
> > > - Set the htype field in the dhcp packet to 6
> > > (ARPHRD_IEEE802)
> > 
> > Hm.  Are you running an 802 (i.e. 802.3 frames)
> > network or an Ethernet (i.e. Ethernet frames) network?
> >  I would leave this field alone.
> > 
> > > - Make sure the xid field is really random (it seems
> > > to be 0 always)
> > 
> > Make sure you are looking at the correct field.  The
> > field before (hops) and after (seconds elapsed) the
> > xid field are 0.  I never observe and xid = 0.
> > 
> > > 
> > > I'm looking for comments, advice and stories of
> > > similar cases,
> > > thanks!  Advice of the sort "upgrade to busybox
> > > 1.00" would have
> > > to be accompanied with specific info about what
> > > changed, as
> > 
> > In the future I'd be less demanding.  You'll get more
> > responses.  So what if someone only says "upgrade to
> > busybox 1.00".  Someone else might be more helpful.
> > 
> > > I don't see that these specific fields are any
> > > different in
> > > 1.00 from 1.00-pre3.  I'd love to have an excuse to
> > > force
> > > the upgrade now, so if I missed something please do
> > > let me
> > > know.
> > > 
> > 
> > Why not just try it out?  Just because you try it out
> > doesn't mean you have to upgrade.
> > 
> > > I'll be trying out the above suggestions but I'm
> > > interested in
> > > any gotchas that you'd know about.  I really know
> > > nothing
> > > about the htype field...  
> > > 
> > 
> > IMHO the above suggestions won't do any good.
> > 
> > Please read RFC2131, RFC2132, and
> > http://www.iana.org/assignments/arp-parameters.
> > 
> > > It is being suggested to change that from 1 to 6
> > > because the
> > > link negotiation is being held at 10 MB (rather than
> > > going to
> > > 100 MB where it should be) and someone noticed that
> > > field is
> > > called 10MB.  Personally, I would have thought that
> > > link
> > > negotiation would have been done by the time dhcp
> > > starts,
> > > but I'm a newb.
> > > 
> > 
> > No.  Use htype = 1 if the network is Ethernet.  Don't
> > worry about the rate.
> > 
> > Regards,
> > ...doug
> > 
> > 
> > __________________________________________________
> > Do You Yahoo!?
> > Tired of spam?  Yahoo! Mail has the best spam protection around 
> > http://mail.yahoo.com 
> 
> > _______________________________________________
> > busybox mailing list
> > busybox at mail.busybox.net
> > http://busybox.net/mailman/listinfo/busybox
> 
> ---end quoted text---
> 




More information about the busybox mailing list