Report forwarded to debian-bugs-dist@lists.debian.org, Anthony Towns <ajt@debian.org>: Bug#296640; Package netbase.
(full text, mbox, link).
Acknowledgement sent to Hadmut Danisch <hadmut@danisch.de>:
New Bug report received and forwarded. Copy sent to Anthony Towns <ajt@debian.org>.
(full text, mbox, link).
To: Debian Bug Tracking System <submit@bugs.debian.org>
Subject: netbase: /etc/rcS.d script order collision with hotplug
Date: Wed, 23 Feb 2005 19:12:22 +0100
Package: netbase
Version: 4.18
Severity: normal
Hi,
just to let you know about a bug report I've sent to
the hotplug package:
I was just trying to figure out why my notebook hangs when booting with the
WLAN card inserted. I finally found the problem:
The reason is that /etc/rcS.d/S40hotplug is executed before
/etc/rcS.d/S40networking.
hotplug detects the presence of the network card and runs ifup for that
device. ifup runs some if-pre-up and if-up scripts. Some of the programs make
use of the loopback device 127.0.0.1. Unfortunately, at this time
S40networking has not yet been run and therefore the lo device is not yet up.
127.0.0.1 is not reachable at this time.
Maybe it would be a good idea to have a separate rcS.d script which
ifconfigs lo up (hardcoded without /etc/network/interfaces) much earlier,
at a very early booting stage.
I guess that only very few debian users will have dedicated hardware
equipment for the loopback device. Most users will use the software
emulation only. So in most cases the loopback should be enabled before
the hotplug daemon starts to work. ;-)
regards
Hadmut
-- System Information:
Debian Release: 3.1
APT prefers unstable
APT policy: (500, 'unstable')
Architecture: i386 (i686)
Kernel: Linux 2.6.9-danisch-p4-intel
Locale: LANG=C, LC_CTYPE=de_DE (charmap=ISO-8859-1)
Versions of packages netbase depends on:
ii debconf 1.4.39 Debian configuration management sy
ii ifupdown 0.6.4-4.10 High level tools to configure netw
ii iputils-ping [ping] 3:20020927-2 Tools to test the reachability of
ii netkit-inetd 0.10-10 The Internet Superserver
ii tcpd 7.6.dbs-6 Wietse Venema's TCP wrapper utilit
-- debconf information excluded
Message sent on to Hadmut Danisch <hadmut@danisch.de>:
Bug#296640.
(full text, mbox, link).
Subject: Re: [hadmut@danisch.de: Bug#296640: netbase: /etc/rcS.d script
order collision with hotplug]
Date: Wed, 23 Feb 2005 21:02:51 +0100
> > hotplug detects the presence of the network card and runs ifup for that
> > device. ifup runs some if-pre-up and if-up scripts. Some of the programs make
> > use of the loopback device 127.0.0.1. Unfortunately, at this time
> > S40networking has not yet been run and therefore the lo device is not yet up.
> > 127.0.0.1 is not reachable at this time.
Well, one could say "Don't do that, then: Don't do anything requiring
the loopback interface in an up script of a logical interface that could
be applied to a physical interface brought up by hotplug at
S:S40hotplug!" But that wouldn't be very helpful.
Given that S:S40networking is where networking is supposed to be enabled
(see /etc/rcS.d/README), I think that the "correct" solution is to do
: > /etc/hotplug/.run/net.enable
at S:S41hotplug-ifup rather than at S:S40hotplug so that
hotplug-generated ifup commands execute after S:S40networking's ifup
commands.
This means adding another initscript to the hotplug package. The main
hotplug initscript still has to run before S:S40networking because it
handles the initialization of some hardware that is needed for
networking.
> > Maybe it would be a good idea to have a separate rcS.d script which
> > ifconfigs lo up (hardcoded without /etc/network/interfaces) much earlier,
> > at a very early booting stage.
That might be another solution. It would mean adding another initscript
to the netbase package. If ifup were used instead of ifconfig (which
seems cleaner to me) then the initscript would have to be at
S:S39networking-lo (_after_ S39ifupdown).
--
Thomas Hood <jdthood@aglu.demon.nl>
Information stored: Bug#296640; Package netbase.
(full text, mbox, link).
Acknowledgement sent to Hadmut Danisch <hadmut@danisch.de>:
Extra info received and filed, but not forwarded.
(full text, mbox, link).
To: Thomas Hood <jdthood@aglu.demon.nl>, 296640-quiet@bugs.debian.org
Subject: Re: Bug#296640: [hadmut@danisch.de: Bug#296640: netbase: /etc/rcS.d
script order collision with hotplug]
Date: Wed, 23 Feb 2005 21:41:58 +0100
Thomas Hood wrote:
>
> Well, one could say "Don't do that, then: Don't do anything requiring
> the loopback interface in an up script of a logical interface that could
> be applied to a physical interface brought up by hotplug at
> S:S40hotplug!" But that wouldn't be very helpful.
That's absolutely not a good idea. When configuring complex
network devices such as wlan devices, one needs advanced
scripts to configure daemons etc. E.g. I do reconfigure
daemons like a web proxy depending on the network connection.
I'll do it in /etc/network/interfaces. It must still work when
the card is plugged in at boot time.
In general, any program must be available in the up/down scripts.
And how should someone tell reliably, which programs do use
the loopback? All it takes is a query to a DNS cache to
make use of the proxy.
lo is not just a network devices like the others. It has a
standard configuration and is required on all systems by default,
even those without network connection.
So I do consider it as wrong that the lo configuration is in
/etc/network/interfaces at all. From my point of view,
interfaces, ifup, ifdown should deal with "real" interfaces only.
The loopback device should not depend on the configuration, but
be brought up always and as early as possible.
> Given that S:S40networking is where networking is supposed to be enabled
> (see /etc/rcS.d/README), I think that the "correct" solution is to do
>
> : > /etc/hotplug/.run/net.enable
It is not up to the hotplug deamon to prepare the operating system.
loopback is not a hotplug issue, it should be there in time even without
hotplugging. Even if there is no /etc/network/interfaces at all.
> This means adding another initscript to the hotplug package. The main
> hotplug initscript still has to run before S:S40networking because it
> handles the initialization of some hardware that is needed for
> networking.
I believe that it is bad design to have loopback configured by the
hotplug. netbase is the place to do this.
>>>Maybe it would be a good idea to have a separate rcS.d script which
>>>ifconfigs lo up (hardcoded without /etc/network/interfaces) much earlier,
>>>at a very early booting stage.
>
>
>
> That might be another solution. It would mean adding another initscript
> to the netbase package. If ifup were used instead of ifconfig (which
> seems cleaner to me) then the initscript would have to be at
> S:S39networking-lo (_after_ S39ifupdown).
On the contrary. My proposal is to do it _before_ S39ifupdown and
without ifup. loopback should be present before anything happens with
all that ifup/ifdown stuff. It should be completely separated from that
mechanisms.
My proposal is to have a init script with just
ifconfig lo 127.0.0.1 up
at the earliest possible step in boot sequence. before S39ifupdown.
regards
Hadmut
Information stored: Bug#296640; Package netbase.
(full text, mbox, link).
Acknowledgement sent to Thomas Hood <jdthood@aglu.demon.nl>:
Extra info received and filed, but not forwarded.
(full text, mbox, link).
Cc: 296640-quiet@bugs.debian.org, Marco d'Itri <md@Linux.IT>
Subject: Re: Bug#296640: [hadmut@danisch.de: Bug#296640: netbase:
/etc/rcS.d script order collision with hotplug]
Date: Wed, 23 Feb 2005 22:05:55 +0100
On Wed, 2005-02-23 at 21:41 +0100, Hadmut Danisch wrote:
> > Given that S:S40networking is where networking is supposed to be enabled
> > (see /etc/rcS.d/README), I think that the "correct" solution is to do
> >
> > : > /etc/hotplug/.run/net.enable
>
> It is not up to the hotplug deamon to prepare the operating system.
> loopback is not a hotplug issue, it should be there in time even without
> hotplugging.
Your reply makes no sense because you don't know how hotplug works.
The proposal is to let S:S40networking bring up lo the way it always
has and to delay hotplug-generated ifup commands until after
S:S40networking.
--
Thomas Hood <jdthood@aglu.demon.nl>
Information stored: Bug#296640; Package netbase.
(full text, mbox, link).
Acknowledgement sent to Hadmut Danisch <hadmut@danisch.de>:
Extra info received and filed, but not forwarded.
(full text, mbox, link).
Subject: Re: Bug#296640: [hadmut@danisch.de: Bug#296640: netbase: /etc/rcS.d script order collision with hotplug]
Date: Wed, 23 Feb 2005 22:24:39 +0100
On Wed, Feb 23, 2005 at 10:05:55PM +0100, Thomas Hood wrote:
>
> The proposal is to let S:S40networking bring up lo the way it always
> has and to delay hotplug-generated ifup commands until after
> S:S40networking.
Sure, why make it simple, if it can be done complicated and
error-prone?
Please discuss this with the hotplug package maintainer. His
point of view is that hotplug needs to run before networking, since
networking could depend on hotplug.
Just deal with him direclty. I'm not going to play the discussion
proxy between both of you.
regards
Hadmut
Information stored: Bug#296640; Package netbase.
(full text, mbox, link).
Acknowledgement sent to jdthood@aglu.demon.nl:
Extra info received and filed, but not forwarded.
(full text, mbox, link).
Subject: configuring network interfaces in the right order
Date: Thu, 24 Feb 2005 14:33:57 GMT
> Please discuss this with the hotplug package maintainer. His
> point of view is that hotplug needs to run before networking, since
> networking could depend on hotplug.
That is right. "/etc/init.d/hotplug start" needs to run before "/etc/init.d/networking
start" in the init sequence because it
initializes some networking hardware. My proposal is to add an
/etc/init.d/hotplug-net script (the name doesn't matter) which
would be run _after_ "/etc/init.d/networking start" and which
would just trigger the configuration of network devices over
which hotplug has control. This triggering consists in doing the
following:
: > /etc/hotplug/.run/net.enable
which allows waiting net.ifup processes to continue. Currently
the triggering is done by /etc/init.d/hotplug (via
/etc/hotplug/net.rc, I believe) but your bug report indicates
that this is too early.
The result would be:
S40hotplug # configures early hot plugged network hardware
S40networking # ifup lo and other auto i'faces
S41hotplug-net # ifup hot plugged interfaces
Delaying the triggering a bit should be a safe change because what
are triggered (viz., ifup commands for hot plugged interfaces) are
events that happen asynchronously with rcS.d scripts.
--
Thomas
Hood
Information stored: Bug#296640; Package netbase.
(full text, mbox, link).
Acknowledgement sent to Hadmut Danisch <hadmut@danisch.de>:
Extra info received and filed, but not forwarded.
(full text, mbox, link).
Subject: Re: configuring network interfaces in the right order
Date: Thu, 24 Feb 2005 15:43:12 +0100
On Thu, Feb 24, 2005 at 02:33:57PM +0000, jdthood@aglu.demon.nl wrote:
>
> The result would be:
>
> S40hotplug # configures early hot plugged network hardware
> S40networking # ifup lo and other auto i'faces
> S41hotplug-net # ifup hot plugged interfaces
Still wrong.
What if S40hotplug calls any script/subprogram which requires
loopback? E.g. asking a DNS cache?
You should understand that this problem might not be related to
hotplug only, it could just be the point where the problem occured
in my case. Configuring the loopback too late could also cause trouble
with other packages.
Furthermore, I do consider your proposal as too complicated and
too error-prone and thus not advisable for most debian users.
Again: Don't play these extra-complicated games. Presence of loopback
is a precaution for many programs, not just hotplug.
So I stress again: Instead of fiddling around with funny scripts and
hidden directories starting with a dot in their name, just do it the
simple way: Just configure lo as early and as simple as possible.
Please don't construct traps and bug-generators like this.
regards
Hadmut
Information stored: Bug#296640; Package netbase.
(full text, mbox, link).
Acknowledgement sent to jdthood@aglu.demon.nl:
Extra info received and filed, but not forwarded.
(full text, mbox, link).
Subject: Re: configuring network interfaces in the right order
Date: Thu, 24 Feb 2005 15:04:24 GMT
> Instead of fiddling around with funny scripts and
> hidden directories starting with a dot in their name, just do it the
> simple way:
It only seems "funny" because you don't know how hotplug currently
handles hot plugged network interfaces. Go and read /etc/hotplug/net.agent.
Delaying the creation of net.enable is the more conservative solution
because it doesn't change the order of things done in rcS.d.
--
Thomas
Hood
Information stored: Bug#296640; Package netbase.
(full text, mbox, link).
Acknowledgement sent to Hadmut Danisch <hadmut@danisch.de>:
Extra info received and filed, but not forwarded.
(full text, mbox, link).
Subject: Re: configuring network interfaces in the right order
Date: Thu, 24 Feb 2005 16:25:53 +0100
On Thu, Feb 24, 2005 at 03:04:24PM +0000, jdthood@aglu.demon.nl wrote:
>
> It only seems "funny" because you don't know how hotplug currently
> handles hot plugged network interfaces. Go and read
> /etc/hotplug/net.agent.
I do consider it as highly insulting that you assume that everyone
who doesn't agree with your personal opinion simply has no clue about
what is going on. And it doesn't get any better if you repeat it.
Stop insulting other people. Fix the problem. And see whether your
particular way to cope with the problem will cause more bug reports
or not. If you assume that I hadn't read net.agent, then what do
you think how many user have read it?
> Delaying the creation of net.enable is the more conservative solution
> because it doesn't change the order of things done in rcS.d.
Sure. Being conservative and not changing the things which caused
the problem is always the best way to fix a problem.
Stop trying to teach me anything. Fix the package, not me.
Hadmut
Information stored: Bug#296640; Package netbase.
(full text, mbox, link).
Acknowledgement sent to jdthood@aglu.demon.nl:
Extra info received and filed, but not forwarded.
(full text, mbox, link).
To: Hadmut Danisch <hadmut@danisch.de>, jdthood@aglu.demon.nl, 296640-quiet@bugs.debian.org, "Marco d'Itri" <md@Linux.IT>
Subject: Re: configuring network interfaces in the right order
Date: Thu, 24 Feb 2005 15:58:13 GMT
> I do consider it as highly insulting that you assume that everyone
> who doesn't agree with your personal opinion simply has no clue about
> what is going on.
I do not "assume" that "everyone" has no clue about what is going on
just become some one person disagrees with me.
But if some person makes a mistaken statement then I may reasonably
draw the conclusion that that person is missing some relevant
information.
> Stop insulting other people. Fix the problem.
I have made a proposal. The decision about whether or not to regard
the issue as a problem, and if so, about what to do about it, lies
with the package maintainer.
--
Thomas
Hood
Information stored: Bug#296640; Package netbase.
(full text, mbox, link).
Acknowledgement sent to jdthood@aglu.demon.nl:
Extra info received and filed, but not forwarded.
(full text, mbox, link).
reassign 296630 hotplug
severity 296630 important
thanks
So I propose fixing this in the hotplug package by delaying
: > /etc/hotplug/.run/net.enable
until after S:S40networking. The required changes should be
obvious but I will provide patches if asked. Besides
fixing #296640 it will also eliminate the phenomenon of
S:S40networking and net.ifup both running ifup commands at the
same time, with chance determining which of them happens first.
ifup serializes these commands so nothing is actually corrupted,
but it will be better if we let S:S40networking finish doing its
"ifup -a" before allowing net.ifup to do ifups.
--
Thomas
Bug reassigned from package `netbase' to `hotplug'.
Request was from jdthood@aglu.demon.nl
to control@bugs.debian.org.
(full text, mbox, link).
Severity set to `important'.
Request was from jdthood@aglu.demon.nl
to control@bugs.debian.org.
(full text, mbox, link).
Reply sent to Marco d'Itri <md@linux.it>:
You have taken responsibility.
(full text, mbox, link).
Notification sent to Hadmut Danisch <hadmut@danisch.de>:
Bug acknowledged by developer.
(full text, mbox, link).
Subject: Bug#296640: fixed in hotplug 0.0.20040329-20
Date: Sun, 20 Mar 2005 09:32:10 -0500
Source: hotplug
Source-Version: 0.0.20040329-20
We believe that the bug you reported is fixed in the latest version of
hotplug, which is due to be installed in the Debian FTP archive:
hotplug_0.0.20040329-20.diff.gz
to pool/main/h/hotplug/hotplug_0.0.20040329-20.diff.gz
hotplug_0.0.20040329-20.dsc
to pool/main/h/hotplug/hotplug_0.0.20040329-20.dsc
hotplug_0.0.20040329-20_all.deb
to pool/main/h/hotplug/hotplug_0.0.20040329-20_all.deb
A summary of the changes between this version and the previous one is
attached.
Thank you for reporting the bug, which will now be closed. If you
have further comments please address them to 296640@bugs.debian.org,
and the maintainer will reopen the bug report if appropriate.
Debian distribution maintenance software
pp.
Marco d'Itri <md@linux.it> (supplier of updated hotplug package)
(This message was generated automatically at their request; if you
believe that there is a problem with it please contact the archive
administrators by mailing ftpmaster@debian.org)
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Format: 1.7
Date: Sun, 20 Mar 2005 15:04:42 +0100
Source: hotplug
Binary: hotplug
Architecture: source all
Version: 0.0.20040329-20
Distribution: unstable
Urgency: medium
Maintainer: Fumitoshi UKAI <ukai@debian.or.jp>
Changed-By: Marco d'Itri <md@linux.it>
Description:
hotplug - Linux Hotplug Scripts
Closes: 272947295738296640297226297705298353299215299946300113
Changes:
hotplug (0.0.20040329-20) unstable; urgency=medium
.
* Fixed postrm failure when removing already removed directories.
(Closes: #297705)
* Sort the modules in the debconf question, patch by Sean Finney.
(Closes: #295738)
* Added patch scsi.agent_cleanup implementing support for loading the sg
driver and the blacklist check. (Closes: #272947)
Also added a scsi.rc script, by Jeff Bailey of Ubuntu.
* Added versioned dependencies on bash and sed. (Closes: #300113, #299946)
* Removed patch pci_agent_removers, because $DEVPATH does not exist
anymore after a device has been removed. (Closes: #298353)
(Actually this was a patch to usb.agent.)
* Consolidated the patches 010_load_drivers_module_name,
013_load_drivers_check_module_dup, 014_load_drivers_blacklist_dir and
052_module_name_for_blacklist in the new patch 010_functions_misc.
This also fixes another -/_ mismatch in modules names. (Closes: #299215)
* Removed patch 007_more_logging which has been merged in
015_load_drivers_mesg_format.
* Added de4x5 and am53c974 to the blacklist. (See: #294867, #296199)
* Modified postinst to add an example of the QUIET variable to
/etc/default/hotplug if not present. (Closes: #297226)
* Fixed an init script bug which made it ignore scripts not in $RC_ORDER.
* Moved net.rc to a new init script hotplug-net run at S41.
(Closes: #296640)
* Removed the useless workaround introduced in -18.
Files:
e96913e4bad1b78d991bd32ba97aa1b9 668 admin standard hotplug_0.0.20040329-20.dsc
d86d4092cd7a20e83ad019d19b5f459a 48963 admin standard hotplug_0.0.20040329-20.diff.gz
b744d1521fc5a04aa19022084ac9f3d7 67594 admin standard hotplug_0.0.20040329-20_all.deb
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (GNU/Linux)
iD8DBQFCPYUDFGfw2OHuP7ERAmp5AJ9j2MR1XX5mUZCRz3clWtUmLROGNQCfT5WZ
KT2Hmkr7AVqC7WiGKEwvEtI=
=Huc6
-----END PGP SIGNATURE-----
Information forwarded to debian-bugs-dist@lists.debian.org, Fumitoshi UKAI <ukai@debian.or.jp>: Bug#296640; Package hotplug.
(full text, mbox, link).
Acknowledgement sent to Thomas Hood <jdthood@aglu.demon.nl>:
Extra info received and forwarded to list. Copy sent to Fumitoshi UKAI <ukai@debian.or.jp>.
(full text, mbox, link).
Subject: Re: Bug#299215 acknowledged by developer (Bug#299215: fixed in
hotplug 0.0.20040329-20)
Date: Sun, 20 Mar 2005 16:48:57 +0100
On Sun, 2005-03-20 at 06:50 -0800, Debian Bug Tracking System wrote:
> * Moved net.rc to a new init script hotplug-net run at S41.
> (Closes: #296640)
I just thought I would point out that although /etc/hotplug/net.rc has
disappeared from the package, people who upgrade the package from an
earlier version in unstable/testing still have the script, and so in
their case hotplug will still create net.enable at S:S40hotplug.
The bug should be fixed for fresh installs, though.
I'd say that what you have done is arguably good enough. You could
disable net.rc, but then there may be people who don't want it disabled
and then you have to get into asking debconf questions. Bah.
Thanks for the new release.
--
Thomas Hood <jdthood@aglu.demon.nl>
Information forwarded to debian-bugs-dist@lists.debian.org, Fumitoshi UKAI <ukai@debian.or.jp>: Bug#296640; Package hotplug.
(full text, mbox, link).
Acknowledgement sent to md@Linux.IT (Marco d'Itri):
Extra info received and forwarded to list. Copy sent to Fumitoshi UKAI <ukai@debian.or.jp>.
(full text, mbox, link).
On Mar 20, Thomas Hood <jdthood@aglu.demon.nl> wrote:
> I just thought I would point out that although /etc/hotplug/net.rc has
> disappeared from the package, people who upgrade the package from an
> earlier version in unstable/testing still have the script, and so in
If they have modified it... or else why should dpkg not remove it?
--
ciao,
Marco
Subject: Re: Bug#296640: Bug#299215 acknowledged by developer (Bug#299215:
fixed in hotplug 0.0.20040329-20)
Date: Sun, 20 Mar 2005 17:31:28 +0100
On Sun, 2005-03-20 at 16:57 +0100, Marco d'Itri wrote:
> If they have modified it... or else why should dpkg not remove it?
dpkg never removes a conffile just because it disappears from the
package. See #108587 for a discussion of this.
--
Thomas Hood <jdthood@aglu.demon.nl>
Information forwarded to debian-bugs-dist@lists.debian.org, Fumitoshi UKAI <ukai@debian.or.jp>: Bug#296640; Package hotplug.
(full text, mbox, link).
Acknowledgement sent to md@Linux.IT (Marco d'Itri):
Extra info received and forwarded to list. Copy sent to Fumitoshi UKAI <ukai@debian.or.jp>.
(full text, mbox, link).
On Mar 20, Thomas Hood <jdthood@aglu.demon.nl> wrote:
> I'd say that what you have done is arguably good enough. You could
> disable net.rc, but then there may be people who don't want it disabled
> and then you have to get into asking debconf questions. Bah.
There is no point in leaving it there, so the next release will delete
it if present.
--
ciao,
Marco
Debbugs is free software and licensed under the terms of the GNU General
Public License version 2. The current version can be obtained
from https://bugs.debian.org/debbugs-source/.