dynamic configuration of wired network interfaces looks quite broken

[Michal Jaegermann]

Before I will waste more time on that does anybody know what broke
dhclient on _wired_ interfaces?

I was away for quite a while now and upon return I found that although
I can still use DHCP to configure wireless interfaces of my laptops
this is totally different story for _wired_ ones.  In the first case I
see in logs:

dhclient[2304]: DHCPDISCOVER on ath0 to 255.255.255.255 port 67 interval 20
 dhclient[2304]: DHCPOFFER from 192.168.xx.yy
 dhclient[2304]: DHCPREQUEST on ath0 to 255.255.255.255 port 67
 dhclient[2304]: DHCPACK from 192.168.xx.yy

and an interfaces comes up.  With a wired one I get:

dhclient[2135]: DHCPDISCOVER on eth0 to 255.255.255.255 port 67 interval 4
 dhclient[2135]: DHCPDISCOVER on eth0 to 255.255.255.255 port 67 interval 6
 dhclient[2135]: DHCPDISCOVER on eth0 to 255.255.255.255 port 67 interval 12
 dhclient[2135]: DHCPDISCOVER on eth0 to 255.255.255.255 port 67 interval 13
 dhclient[2135]: DHCPDISCOVER on eth0 to 255.255.255.255 port 67 interval 14
 dhclient[2135]: DHCPDISCOVER on eth0 to 255.255.255.255 port 67 interval 12
 dhclient[2135]: No DHCPOFFERS received.

and no networking.  This happens with my test rawhide installation,
both before and after I applied a big pile of pending updates, and
with Fedora 12 installations with all available updates applied.

In all cases dhclient here talks to the same DHCP server, which
happens to be some version of dnsmasq running on a WRTG router, and on
the same network but only through different network interfaces.  My
test machine does not have a wireless one but only a wired which up to
now was configured to use DHCP, and which I would prefer to keep that
way for various reasons, so you can see the trouble.

Asigning static IPs on wired interfaces makes a network to go "alive"
again but this is quite bad option for laptops which travel from time
to time with "foreign" networks not always wireless and a dynamic
configuration should work anyway.  There is still an option of
wireless breaking the same way on the next update.  Luckily my other
not running Feodra "wired DHCP clients", like a DHCP _requiring_
network interface of my printer - for example, still work just fine as
before (and this shows that my DHCP server in use is fine).

This does not look as a directly dhclient issue as reverting that to
older version did not change anything (and it seems that I went back
far enough).  Also a status of selinux, on which I practically gave up
anyway, is irrelevant.  A search through bugzilla, quite possibly as
unreliable as usual, also did not make me much wiser.  My suspicion
would be some other library doing some "creative things" with IPv6 but
I do not have a smoking gun.  I would file a bugzilla report but I
have not a clue against what.  Maybe this is already a well know
problem?   Hm, iptables?  I did not try that yet.

Again - the breakage may show up any time in the last two months and
I had no way to notice it but only now.

Michal

[Richard Ryniker]

The most simple (and perhaps too obvious) explanation would be your DHCP
server is not configured to respond to requests from the wired network
interace, but responds appropriately to requests from the wireless
network interface.  It this used to work, perhaps something changed your
DHCP server's configuration file (or the location of the configuration
file).  Check to verify whether the log file on the server host shows
receipt of the wireless network request, but no receipt of the wired
network request.

I use iptables to limit local network activity, and find it useful on
occasion to replace my regular rules by a completely permissive
configuration to test whether some error or unexpected rule consequence
has interfered with my desired network behavior.  If a failing application
works with the relaxed rules, I know to examine my usual rule set to
learn why it prevented delivery of a desired message.
> From your description, it does not appear likely to be your problem, but

on more than one occasion I found DHCP servers unexpectedly active on two
or more of my hosts.  This is a wonderful formula for confusion and
embarassment: no amount of scrutiny of a correct configuration file will
explain incorrect results delivered by a different server.  I sometimes
want multiple DHCP servers, but configuration to limit them to disjoint
sets of clients or networks is more difficult than with only one server.

[Michal Jaegermann]

On Mon, Jul 12, 2010 at 02:18:00PM -0400, Richard Ryniker wrote:
> The most simple (and perhaps too obvious) explanation would be your DHCP
> server is not configured to respond to requests from the wired network
> interace, but responds appropriately to requests from the wireless
> network interface.

As I wrote - all interfaces are on the same subnetwork and a DHCP
server is not likely "to know" from what kind of hardware a request
is coming.  Again - no problems with DHCPOFFER on wireless
interfaces.
> It this used to work, perhaps something changed your
> DHCP server's configuration file (or the location of the configuration
> file).

Well, no.  Although it is not very likely that something like that
will change by itself on a router this was one of the first things I
checked.  And no, a router software was NOT updated in the meantime
and non-Fedora clients work like worked before.
> Check to verify whether the log file on the server host shows
> receipt of the wireless network request,

Unfortunately there is not too much of log I can find there.
> From your description, it does not appear likely to be your problem, but
> on more than one occasion I found DHCP servers unexpectedly active on two
> or more of my hosts.

This is not the case here.  At least in this moment.  In any case
this server is marked as "dhcp-authoritative" and you may have more
that one active DHCP server on a network if they are properly
configured.

Michal

[Jóhann B. Guðmundsson]

Did you try older kernel and have you tried to wireshark the interface 
to see if it actually sends the dhcp request?

Does setting static IP address on the interface work?

JBG

dynamic configuration of wired network interfaces looks quite broken (looks like solved)

[Michal Jaegermann]

On Mon, Jul 12, 2010 at 09:55:40PM +0000, "Jóhann B. Guðmundsson" wrote:
> Does setting static IP address on the interface work?

It does.  I wrote that in the original posting and this was making
things more puzzling.  Yesterday I applied over 1 Gig of updates to
a rawhide installation using static settings.

Still I think that I figured it out.  It appears that a powered
switch on my network decided to misbehave.  After resetting it I
started to see wired interfaces dynamically configured again.  A
mystery remains how my printer got its IP number.  It is hanging
from the same switch.  Really weird ...

An implicit information that I am the only one with such issue was
of help.  Thanks!

Michal

dynamic configuration of wired network interfaces looks quite broken (looks like solved)

[cornel panceac]

> Still I think that I figured it out.  It appears that a powered
> switch on my network decided to misbehave.  After resetting it I
> started to see wired interfaces dynamically configured again.  A
> mystery remains how my printer got its IP number.  It is hanging
> from the same switch.  Really weird ...

could be only certain port in the switch. tried another? (like switch
between printer's and computer's)

dynamic configuration of wired network interfaces looks quite broken (looks like solved)

[Michal Jaegermann]

On Tue, Jul 13, 2010 at 08:34:44AM +0300, cornel panceac wrote:
> mystery remains how my printer got its IP number.  It is hanging
> from the same switch.  Really weird ...
>
> could be only certain port in the switch. tried another? (like switch between
> printer's and computer's)

At least four different ports on a switch were affected.  I know
that not from swapping cables but from trying other ends.  Now it is
too late to check although it is difficult to imagine why troubles
would be so selective but maybe.  That static IP assignments worked
just fine did not help in a diagnosis.  I realized eventually that
over-the-air packets do not travel through this piece of hardware.
Too late now.  After a switch reset everything works as expected.

Michal

dynamic configuration of wired network interfaces looks quite broken (looks like solved)

[cornel panceac]

> At least four different ports on a switch were affected.  I know
> that not from swapping cables but from trying other ends.  Now it is
> too late to check although it is difficult to imagine why troubles
> would be so selective but maybe.  That static IP assignments worked
> just fine did not help in a diagnosis.  I realized eventually that
> over-the-air packets do not travel through this piece of hardware.
> Too late now.  After a switch reset everything works as expected.

ok

[Antonio Olivares]

--- On Mon, 7/12/10, Michal Jaegermann  wrote:
> From: Michal Jaegermann 
> Subject: Re: dynamic configuration of wired network interfaces looks quite broken
> To: "For testers of Fedora development releases" 
> Date: Monday, July 12, 2010, 1:13 PM
> xuTMoÛ8½óWré�c'è¡ð!hÑÝ`]
> M¤è™–F×GKRQõïû†²�‹½‚,¾y_ÃÇ@*ú<zº¹­èvs³!›is»½y¿Ýl¾>Ðjón³©èÉÕ��

=ÍÁ9Ò%óÖÜÑsÇÔKÊ”?x¦?lhhàØÙ!Q!Ù¿8Ó[⃷Áf'�&}C{¦YÆHþýé+ Ç@»DA2ÕZw#7€¡Èi
> —ÇGN9Q¥§Œù“Ó¯çIâ@.dŽ¶æŠöc>Md‡!Ê�ÍìçÿGòœ@NpX´kc>&Ú-ÒiEÖû×?�™ Lù$ÛãgÜŸ!Ô»¨ü]£‡— rUÇ
> ÓÕ"iêÁÑá�´¤®OŠm!¤70¨–Þ…Ã5Ñǃud‚Äí=÷   
> vä®L{¼¿ÿëII©AEVqFÅ$¨¹£]_À�iqYý«.é%é‡!ûpÀ'šŽi^§´Þ(—%¨%ÙÖi   
> $#¼ÔË[Qc~û@úñ[0ùÎÞWÐ ‚|îd<täò¹   
> (ÅL'«²zóJLßÙf39ıð¤ýŒÓ‰}«Úᛌèât²   
> /‘ÔB¨uÕ-"­©;®�Ü(

¸¤z=�¤Í%EøòøLãÐ H
*P¤õlCv=
;HXÝs#ÑRí”U­-"ÊÎí¹•¨¥º£O:V
> ¥®�iêà{aÌäåPlºtkY’®¬'“V5rÍnÈgQç´/>§2æ[ÀÐ<bË
> `Ÿ]Ö]íǺSº£ÚŒ†%³0½×•+;Ûpª#†b›+ͪ^ö[Æ6^ò’Ë–ŸúY–¤M

Miichal,

Sorry, but I don't understand this.  The message does not appear right.  Which language is it set?

Am I the only one who sees it/saw it this way?

I have gotten many replies of excellent quality, but this one does not appear correctly :(

Thanks and sorry for not understanding your post.

Regards,

Antonio

[Ed Greshko]

On 07/13/2010 11:51 AM, Antonio Olivares wrote:
> ¸¤z=�¤Í%EøòøLãÐ H
> *P¤õlCv=
> ;HXÝs#ÑRí”U­-"ÊÎí¹•¨¥º£O:V
> > ¥®�iêà{aÌäåPlºtkY’®¬'“V5rÍnÈgQç´/>§2æ[ÀÐ<bË
> > `Ÿ]Ö]íǺSº£ÚŒ†%³0½×•+;Ûpª#†b›+ͪ^ö[Æ6^ò’Ë–ŸúY–¤M
>
> Miichal,
>
> Sorry, but I don't understand this.  The message does not appear right.  Which language is it set?
>
> Am I the only one who sees it/saw it this way?

I think you may be the only one seeing it this way.  The message was
sent as Content-Type: text/plain; charset="us-ascii".

Check the archives to see how it should look...

[Antonio Olivares]

> > Am I the only one who sees it/saw it this way?
>
> I think you may be the only one seeing it this way. 
> The message was
> sent as Content-Type: text/plain; charset="us-ascii".
>
> Check the archives to see how it should look...
>
> --

http://lists.fedoraproject.org/pipermail/test/2010-July/092051.html

His posts on the archives look fine, but this one ^ looks the same way I saw it.  What could have been the problem?

Thanks,

Antonio

mail text encoding (was: dynamic configuration .... )

[Michal Jaegermann]

On Tue, Jul 13, 2010 at 08:34:20AM -0700, Antonio Olivares wrote:
> > > Am I the only one who sees it/saw it this way?
> >
> > I think you may be the only one seeing it this way. 
> > The message was
> > sent as Content-Type: text/plain; charset="us-ascii".
> >
> > Check the archives to see how it should look...
>
> http://lists.fedoraproject.org/pipermail/test/2010-July/092051.html
>
> His posts on the archives look fine, but this one ^ looks the same
> way I saw it.

You posted it so it includes all garbled content you have quoted.
> What could have been the problem?

It appears that your mail reader is using some hardwired encoding
instead of paying attention to a charset declaration in what was
posted.  Or maybe some intermediate mail server is "kind enough" for
you to do conversions with some random settings.

BTW - this particular message is sent with an utf-8 encoding.

Michał

mail text encoding (was: dynamic configuration .... )

[Antonio Olivares]

--- On Tue, 7/13/10, Michal Jaegermann  wrote:
> From: Michal Jaegermann 
> Subject: Re: mail text encoding (was: dynamic configuration .... )
> To: "For testers of Fedora development releases" 
> Date: Tuesday, July 13, 2010, 10:18 AM
> On Tue, Jul 13, 2010 at 08:34:20AM
> -0700, Antonio Olivares wrote:
> > > > Am I the only one who sees it/saw it this
>
> way?
> > > I think you may be the only one seeing it this
>
> way.
> > > The message was
> > > sent as Content-Type: text/plain;
>
> charset="us-ascii".
> > > Check the archives to see how it should look...
> >
> > http://lists.fedoraproject.org/pipermail/test/2010-July/092051.html
> >
> > His posts on the archives look fine, but this one ^
>
> looks the same
> > way I saw it.
>
> You posted it so it includes all garbled content you have
> quoted.

I did not do it on purpose, that garbled content was what I saw, and I responded/asked is it only me?
> > What could have been the problem?
>
> It appears that your mail reader is using some hardwired
> encoding
> instead of paying attention to a charset declaration in
> what was
> posted.  Or maybe some intermediate mail server is
> "kind enough" for
> you to do conversions with some random settings.

Regular yahoo mail, using konqueror web browser.
> BTW - this particular message is sent with an utf-8
> encoding.
> Michał

I am not using any special mail, just basic yahoo mail.  There appears to be no problem with this one.  What could it have been?

I normally don't see these kinds of messages, but this one came like that. I ask myself what could it have been?

Thanks and sorry for asking.

Regards,

Antonio