Debian Bug report logs - #245435
ifplugd.hotplug: Please ifrename the i'face first

Toggle useless messages

Message #5 received at submit@bugs.debian.org (full text, mbox, reply):

From: "J.D. Thomas Hood" <jdthood@yahoo.co.uk>

To: submit@bugs.debian.org

Subject: ifrename shouldn't be in net.agent

Date: Fri, 23 Apr 2004 08:19:26 +0100 (BST)

Package: hotplug
Version: 0.0.20040329-8

/etc/hotplug/net.agent calls ifrename to rename the hot plugged
interface.  It then calls ifup using the new name.  This is all
well and good, but fails to account for the fact that hook scripts
in /etc/hotplug.d/net/ have already processed the interface under
the old name.

If ifrename is to be integrated into hotplug then it has to be done
in /sbin/hotplug itself so that the latter can change the name of
the interface before any hook scripts are run and so that the new
name can be passed along in INTERFACE instead of the old name.

It might be best if /sbin/hotplug were modified so that in addition
to what it does now it _sourced_ all files from some directory
(e.g., /etc/hotplug/hook/).  Then ifrename could install a file in
that directory that did the rename and changed the INTERFACE
variable.  Other packages could add other hook scripts to do
similar things without /sbin/hotplug having to be changed again.

Obviously, this issue has to be addressed in coordination with the
ifrename maintainer.
--
Thomas Hood


	
	
		
____________________________________________________________
Yahoo! Messenger - Communicate instantly..."Ping" 
your friends today! Download Messenger Now 
http://uk.messenger.yahoo.com/download/index.html

Message #14 received at 245435@bugs.debian.org (full text, mbox, reply):

From: Jean Tourrilhes <jt@bougret.hpl.hp.com>

To: 245435@bugs.debian.org, Thomas Hood <jdthood@yahoo.co.uk>

Subject: I don't see it...

Date: Mon, 26 Apr 2004 19:01:56 -0700

	Hi,

	I may bit a bit slow today, but I can't see the problem.
	For consistency purpose, any script that acts on the network
interface should run from ifup, so that manually changing the state of
the interface does the right thing (ifdown/ifup). That way, the whole
interface configuration is neatly in /etc/network/interfaces.
	Before ifrename, the old interface name is valid. So, if a
hook script does something on the old interface name, it will succeed
as expected. This script will just have to take into account that it's
using the old iface name. If the script require the new interface
name, it might as well run from ifup. But in a sense, that's a
feature, because if you want to do a script that use the old interface
name, you can hook it in the hotplug subsystem.
	If you look at Red-Hat hotplug scripts, the will rename
interfaces as well, but even later (they do it in the ifup script
itself).
	Personally, I don't see why you would require the new iface
name in hotplug hook scripts that should only concern themselves with
hotplug matters, not network matters.

	Have fun...

	Jean

Message #19 received at 245435@bugs.debian.org (full text, mbox, reply):

From: Thomas Hood <jdthood@yahoo.co.uk>

To: jt@hpl.hp.com

Cc: 245435@bugs.debian.org

Subject: Re: I don't see it...

Date: Tue, 27 Apr 2004 20:05:09 +0200

On Tue, 2004-04-27 at 04:01, Jean Tourrilhes wrote:
> 	I may bit a bit slow today, but I can't see the problem.
> 	For consistency purpose, any script that acts on the network
> interface should run from ifup, so that manually changing the state of
> the interface does the right thing (ifdown/ifup). That way, the whole
> interface configuration is neatly in /etc/network/interfaces.

I agree that lots of things should be run from ifup.  Not all, though.
See below.


> 	Before ifrename, the old interface name is valid. So, if a
> hook script does something on the old interface name, it will succeed
> as expected. This script will just have to take into account that it's
> using the old iface name.

The reasons for renaming interfaces for the sake of ifup are
also reasons for renaming them for the sake of other programs
that want to hook hotplug events.  I am thinking of ifplugd and
waproamd here.



> If the script require the new interface name, it might as well run
> from ifup.

ifplugd and waproamd can't run from ifup because they are the ones
calling ifup in the first place (on systems where they are given
the job of doing that).
--
Thomas

Message #24 received at 245435@bugs.debian.org (full text, mbox, reply):

From: Jean Tourrilhes <jt@bougret.hpl.hp.com>

To: Thomas Hood <jdthood@yahoo.co.uk>

Cc: 245435@bugs.debian.org

Subject: Re: I don't see it...

Date: Tue, 27 Apr 2004 11:52:12 -0700

On Tue, Apr 27, 2004 at 08:05:09PM +0200, Thomas Hood wrote:
> On Tue, 2004-04-27 at 04:01, Jean Tourrilhes wrote:
> > 	I may bit a bit slow today, but I can't see the problem.
> > 	For consistency purpose, any script that acts on the network
> > interface should run from ifup, so that manually changing the state of
> > the interface does the right thing (ifdown/ifup). That way, the whole
> > interface configuration is neatly in /etc/network/interfaces.
> 
> I agree that lots of things should be run from ifup.  Not all, though.
> See below.
> 
> 
> > 	Before ifrename, the old interface name is valid. So, if a
> > hook script does something on the old interface name, it will succeed
> > as expected. This script will just have to take into account that it's
> > using the old iface name.
> 
> The reasons for renaming interfaces for the sake of ifup are
> also reasons for renaming them for the sake of other programs
> that want to hook hotplug events.  I am thinking of ifplugd and
> waproamd here.

	Are they triggered on hotplug events or RTnetlink events ? I
was under the impression that they were only processing RTnetlink
events, because they also need to pick wireless events and IP address
changes. And a quick look at their source do confirm that.
	If they are only processing netlink events, they should be
fine. Well, almost, they need to also process the event resulting from
the interface rename, which is a RTnetlink event with change == 0 and
a name attribute (been there, done that). Remember that RTnetlink
application need to refer to interfaces by ifindex, so are somewhat
more resistant to name changes.
	Personally, I don't believe anybody sane would use hotplug
events for serious networking stuff, because it's so coarse and may
not always be enabled.

> > If the script require the new interface name, it might as well run
> > from ifup.
> 
> ifplugd and waproamd can't run from ifup because they are the ones
> calling ifup in the first place (on systems where they are given
> the job of doing that).

	Actually, they also are the one that should call ifrename as
appropriate, because they attempt to bypass the natural flow where
ifrename would be called.

> Thomas

	Regards,

	Jean

Message #29 received at 245435@bugs.debian.org (full text, mbox, reply):

From: Thomas Hood <jdthood@yahoo.co.uk>

To: 245435@bugs.debian.org

Subject: hotplug + ifrename + ifplugd

Date: Tue, 27 Apr 2004 21:30:51 +0200

reassign 245435 ifplugd
retitle 245435 ifplugd.hotplug: Please ifrename the i'face first
severity 245435 wishlist
thanks

On Tue, 2004-04-27 at 20:52, Jean Tourrilhes wrote:
> 	Are they triggered on hotplug events or RTnetlink events ?

ifplugd is started by a hotplug event, whereafter it checks the
link status using one of several different methods, depending on
which method is available.


> 	Actually, they also are the one that should call ifrename as
> appropriate, because they attempt to bypass the natural flow where
> ifrename would be called.

They aren't bypassing a natural flow.  Installing hook scripts under
/etc/hotplug.d/ is the only way a package can hook hotplug events
without asking the hotplug maintainers to make changes to net.agent.

However you do have a point when you say that ifplugd and waproamd
can call ifrename themselves.  Reassigning to ifplugd.
--
Thomas

Message #40 received at 245435@bugs.debian.org (full text, mbox, reply):

From: Thomas Hood <jdthood@aglu.demon.nl>

To: 245435@bugs.debian.org

Subject: dontfix

Date: Mon, 14 Jun 2004 13:01:30 +0200

tag 245435 wontfix
thanks

I submitted this but I am marking it 'wontfix' because I have just
been reading linux-hotplug-devel and there are signs that net dev
renaming will be handed by udev and hotplug.
--
Thomas Hood

Message #45 received at 245435@bugs.debian.org (full text, mbox, reply):

From: martin f krafft <madduck@debian.org>

To: 245435@bugs.debian.org

Cc: udev@packages.qa.debian.org, ifrename@packages.qa.debian.org, ifplugd@packages.qa.debian.org, hotplug@packages.qa.debian.org

Subject: ifplugd should just run later

Date: Sun, 26 Dec 2004 15:20:40 +0100

[Message part 1 (text/plain, inline)]

Sorry for the excessive CC'ing. I think I found a solution to this
problem, and it affects hotplug, udev, ifplugd, and ifrename.

Please read the bug report for background information. Here are my
comments:

Even if renaming is to be handled by udev from now on, the problem
persists, since /etc/hotplug.d/default/10-udev.hotplug is run after
/etc/hotplug.d/net/ifplugd.hotplug. Therefore, if ifplugd brings up
the interface already before udev (or ifrename) had a chance to
rename it, the rename will fail because the interface is in use.

The proper solution to me seems to be:

  1. ifplugd moves to /etc/hotplug/net and is thus invoked by
     net.agent and long after ifrename/udev get a chance to rename
     the interface.

  2. the use of /etc/hotplug.d/net should be discouraged, or udev
     should be run from this directory as one of first hooks, in
     addition to the /etc/hotplug.d/default location. Since udev
     effectively replaces the current hotplug invocation with a new
     invocation after renaming the interface, this would not be
     a problem and default/udev would not be called after the
     rename. udev itself then calls the appropriate agent directly
     via the hotplug.d/default/default hook.

  3. ifrename itself gets called from /etc/hotplug.d/net and does
     the same as udev in /etc/dev.d/net/hotplug.dev does, after
     renaming the interface:
    
       export INTERFACE=$NEWNAME
       exec /etc/hotplug.d/default/default.hotplug net

     further execution of hooks will stop since the device is gone
     after the rename.

If you like this suggestion, please clone this bug against udev and
ifrename. Once all three implement the proposed solution, another
bug needs to be filed against hotplug to remove the ifrename
invocation from net.agent.

Thanks,

-- 
 .''`.     martin f. krafft <madduck@debian.org>
: :'  :    proud Debian developer, admin, user, and author
`. `'`
  `-  Debian - when you have better things to do than fixing a system
 
Invalid/expired PGP subkeys? Use subkeys.pgp.net as keyserver!

[signature.asc (application/pgp-signature, inline)]

Message #48 received at 245435@bugs.debian.org (full text, mbox, reply):

From: martin f krafft <madduck@debian.org>

To: 245435@bugs.debian.org

Cc: udev@packages.qa.debian.org, ifrename@packages.qa.debian.org

Subject: Re: ifplugd should just run later

Date: Sun, 26 Dec 2004 15:26:32 +0100

[Message part 1 (text/plain, inline)]

... and another reason why we should not forget about ifrename and
assume that udev will take over anyway: udev cannot rename
interfaces that are not hotpluggable, such as ISA cards. For those,
ifrename is still needed.

-- 
 .''`.     martin f. krafft <madduck@debian.org>
: :'  :    proud Debian developer, admin, user, and author
`. `'`
  `-  Debian - when you have better things to do than fixing a system
 
Invalid/expired PGP subkeys? Use subkeys.pgp.net as keyserver!

[signature.asc (application/pgp-signature, inline)]

Message #53 received at 245435@bugs.debian.org (full text, mbox, reply):

From: md@Linux.IT (Marco d'Itri)

To: martin f krafft <madduck@debian.org>

Cc: 245435@bugs.debian.org, udev@packages.qa.debian.org, ifrename@packages.qa.debian.org, ifplugd@packages.qa.debian.org, hotplug@packages.qa.debian.org

Subject: Re: ifplugd should just run later

Date: Sun, 26 Dec 2004 15:30:27 +0100

[Message part 1 (text/plain, inline)]

On Dec 26, martin f krafft <madduck@debian.org> wrote:

>   2. the use of /etc/hotplug.d/net should be discouraged, or udev
>      should be run from this directory as one of first hooks, in
At some undefined point in the future, udev will take over as the
hotplug.d multiplex in place of /sbin/hotplug, so it will always
"magically" be run before everything else.
There is no need to open bugs about this, it will happen at the right
time. If you want to test this:

echo /sbin/udevsend > /proc/sys/kernel/hotplug

-- 
ciao,
Marco

[signature.asc (application/pgp-signature, inline)]

Message #56 received at 245435@bugs.debian.org (full text, mbox, reply):

From: martin f krafft <madduck@debian.org>

To: 245435@bugs.debian.org

Cc: udev@packages.qa.debian.org, ifrename@packages.qa.debian.org, hotplug@packages.qa.debian.org

Subject: Re: ifplugd should just run later

Date: Sun, 26 Dec 2004 16:13:55 +0100

[Message part 1 (text/plain, inline)]

[removed ifplugd@packages.qa.d.o from CC list, as they receive the
bug report stuff anyway].

also sprach Marco d'Itri <md@Linux.IT> [2004.12.26.1530 +0100]:
> At some undefined point in the future, udev will take over as the
> hotplug.d multiplex in place of /sbin/hotplug, so it will always

I seem to recall it was you who told me that udev does not intend to
replace hotplug, it only deals with device nodes (and network
interfaces).

Are we now to expect udev to become the all-in-one management tool
for devices?

-- 
 .''`.     martin f. krafft <madduck@debian.org>
: :'  :    proud Debian developer, admin, user, and author
`. `'`
  `-  Debian - when you have better things to do than fixing a system
 
Invalid/expired PGP subkeys? Use subkeys.pgp.net as keyserver!

[signature.asc (application/pgp-signature, inline)]

Message #61 received at 245435@bugs.debian.org (full text, mbox, reply):

From: md@Linux.IT (Marco d'Itri)

To: martin f krafft <madduck@debian.org>

Cc: 245435@bugs.debian.org, udev@packages.debian.org, ifrename@packages.debian.org, hotplug@packages.debian.org

Subject: Re: ifplugd should just run later

Date: Sun, 26 Dec 2004 16:21:33 +0100

[Message part 1 (text/plain, inline)]

On Dec 26, martin f krafft <madduck@debian.org> wrote:

> > At some undefined point in the future, udev will take over as the
> > hotplug.d multiplex in place of /sbin/hotplug, so it will always
> I seem to recall it was you who told me that udev does not intend to
> replace hotplug, it only deals with device nodes (and network
Indeed, it does not. It only (optionally: now it does not do it)
replaces the /sbin/hotplug script.

-- 
ciao,
Marco

[signature.asc (application/pgp-signature, inline)]

Message #64 received at 245435@bugs.debian.org (full text, mbox, reply):

From: martin f krafft <madduck@debian.org>

To: Marco d'Itri <md@Linux.IT>

Cc: 245435@bugs.debian.org, udev@packages.debian.org, ifrename@packages.debian.org, hotplug@packages.debian.org

Subject: Re: ifplugd should just run later

Date: Sun, 26 Dec 2004 16:28:51 +0100

[Message part 1 (text/plain, inline)]

also sprach Marco d'Itri <md@Linux.IT> [2004.12.26.1621 +0100]:
> Indeed, it does not. It only (optionally: now it does not do it)
> replaces the /sbin/hotplug script.

What is the purpose of this change? Just to let udev run before
hotplug.d? Will hotplug.d continue to be used[0]? Will udev become
mandatory?

0. GregKH just recently told me to use /etc/hotplug.d instead of
/etc/hotplug to change the permissions on a USB device upon
insertion. 

-- 
 .''`.     martin f. krafft <madduck@debian.org>
: :'  :    proud Debian developer, admin, user, and author
`. `'`
  `-  Debian - when you have better things to do than fixing a system
 
Invalid/expired PGP subkeys? Use subkeys.pgp.net as keyserver!

[signature.asc (application/pgp-signature, inline)]

Message #69 received at 245435@bugs.debian.org (full text, mbox, reply):

From: md@Linux.IT (Marco d'Itri)

To: martin f krafft <madduck@debian.org>

Cc: 245435@bugs.debian.org, udev@packages.debian.org, ifrename@packages.debian.org, hotplug@packages.debian.org

Subject: Re: ifplugd should just run later

Date: Sun, 26 Dec 2004 16:30:41 +0100

[Message part 1 (text/plain, inline)]

On Dec 26, martin f krafft <madduck@debian.org> wrote:

> also sprach Marco d'Itri <md@Linux.IT> [2004.12.26.1621 +0100]:
> > Indeed, it does not. It only (optionally: now it does not do it)
> > replaces the /sbin/hotplug script.
> What is the purpose of this change? Just to let udev run before
> hotplug.d? Will hotplug.d continue to be used[0]? Will udev become
> mandatory?
Yes (mostly), yes, no (but I'm working hard to make it be).

-- 
ciao,
Marco

[signature.asc (application/pgp-signature, inline)]

Message #74 received at 245435@bugs.debian.org (full text, mbox, reply):

From: "Perez-Gonzalez, Inaky" <inaky.perez-gonzalez@intel.com>

To: <245435@bugs.debian.org>

Subject: Another possible sollution to integrate with udev

Date: Tue, 28 Dec 2004 14:58:45 -0800

Hi All

I just hit this bug and found a more or less "workaround" that
seems to fix the problem, although it still needs some tweaking:

I moved ifplugd.hotplug to /etc/dev.d/ifplugd.dev and added
a line saying INTERFACE=$DEVNAME around the beginning. This
seems to take care of most issues.

Would it work?

Inaky -- not speaking for Intel; my opinions are my own and my fault

Message #83 received at 295514-close@bugs.debian.org (full text, mbox, reply):

From: Oliver Kurth <oku@debian.org>

To: 295514-close@bugs.debian.org

Subject: Bug#295514: fixed in ifplugd 0.26-2

Date: Wed, 16 Mar 2005 01:32:16 -0500

Source: ifplugd
Source-Version: 0.26-2

We believe that the bug you reported is fixed in the latest version of
ifplugd, which is due to be installed in the Debian FTP archive:

ifplugd_0.26-2.diff.gz
  to pool/main/i/ifplugd/ifplugd_0.26-2.diff.gz
ifplugd_0.26-2.dsc
  to pool/main/i/ifplugd/ifplugd_0.26-2.dsc
ifplugd_0.26-2_i386.deb
  to pool/main/i/ifplugd/ifplugd_0.26-2_i386.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 295514@bugs.debian.org,
and the maintainer will reopen the bug report if appropriate.

Debian distribution maintenance software
pp.
Oliver Kurth <oku@debian.org> (supplier of updated ifplugd 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: Mon, 14 Mar 2005 13:33:13 -0800
Source: ifplugd
Binary: ifplugd
Architecture: source i386
Version: 0.26-2
Distribution: unstable
Urgency: low
Maintainer: Oliver Kurth <oku@debian.org>
Changed-By: Oliver Kurth <oku@debian.org>
Description: 
 ifplugd    - A configuration daemon for ethernet devices
Closes: 236856 245435 251860 287513 291111 294180 295514 298589 299428
Changes: 
 ifplugd (0.26-2) unstable; urgency=low
 .
   * clarified /etc/default/ifplugd docs (closes: #287513). Thanks to
     Marc Haber.
   * remove list of allowed ifaces in hotplug script (closes: #251860,
     #295514, #298589)
   * call ifrename in hotplug script (closes: #294180, #291111, #245435)
   * po-debconf (closes: #236856). Thanks to Lucas Wall for support.
   * added pt_BR debconf templates (closes: #299428). Thanks to
     Felipe Augusto van de Wiel.
   * added po files for cs, da, es, fr, zh_TW. Thanks to the submitters.
Files: 
 71abed9b7e48cfb10b146291fa8ee8b2 609 net optional ifplugd_0.26-2.dsc
 c73388d28f228d06e974e24b4f4ddb64 38806 net optional ifplugd_0.26-2.diff.gz
 afed995ce2720e6183d003f22be39068 55928 net optional ifplugd_0.26-2_i386.deb
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (GNU/Linux)

iD8DBQFCN9DCUmVSJkUeqxsRAsAMAJ0WiMD/Pe1XGUjZ5xTsn2jikFmKVQCgkKZI
XT+16JSRZrX6tLpt+JE4evg=
=zqz5
-----END PGP SIGNATURE-----

Message #84 received at 245435-close@bugs.debian.org (full text, mbox, reply):

From: Oliver Kurth <oku@debian.org>

To: 245435-close@bugs.debian.org

Subject: Bug#245435: fixed in ifplugd 0.26-2

Date: Wed, 16 Mar 2005 01:32:16 -0500

Source: ifplugd
Source-Version: 0.26-2

We believe that the bug you reported is fixed in the latest version of
ifplugd, which is due to be installed in the Debian FTP archive:

ifplugd_0.26-2.diff.gz
  to pool/main/i/ifplugd/ifplugd_0.26-2.diff.gz
ifplugd_0.26-2.dsc
  to pool/main/i/ifplugd/ifplugd_0.26-2.dsc
ifplugd_0.26-2_i386.deb
  to pool/main/i/ifplugd/ifplugd_0.26-2_i386.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 245435@bugs.debian.org,
and the maintainer will reopen the bug report if appropriate.

Debian distribution maintenance software
pp.
Oliver Kurth <oku@debian.org> (supplier of updated ifplugd 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: Mon, 14 Mar 2005 13:33:13 -0800
Source: ifplugd
Binary: ifplugd
Architecture: source i386
Version: 0.26-2
Distribution: unstable
Urgency: low
Maintainer: Oliver Kurth <oku@debian.org>
Changed-By: Oliver Kurth <oku@debian.org>
Description: 
 ifplugd    - A configuration daemon for ethernet devices
Closes: 236856 245435 251860 287513 291111 294180 295514 298589 299428
Changes: 
 ifplugd (0.26-2) unstable; urgency=low
 .
   * clarified /etc/default/ifplugd docs (closes: #287513). Thanks to
     Marc Haber.
   * remove list of allowed ifaces in hotplug script (closes: #251860,
     #295514, #298589)
   * call ifrename in hotplug script (closes: #294180, #291111, #245435)
   * po-debconf (closes: #236856). Thanks to Lucas Wall for support.
   * added pt_BR debconf templates (closes: #299428). Thanks to
     Felipe Augusto van de Wiel.
   * added po files for cs, da, es, fr, zh_TW. Thanks to the submitters.
Files: 
 71abed9b7e48cfb10b146291fa8ee8b2 609 net optional ifplugd_0.26-2.dsc
 c73388d28f228d06e974e24b4f4ddb64 38806 net optional ifplugd_0.26-2.diff.gz
 afed995ce2720e6183d003f22be39068 55928 net optional ifplugd_0.26-2_i386.deb
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (GNU/Linux)

iD8DBQFCN9DCUmVSJkUeqxsRAsAMAJ0WiMD/Pe1XGUjZ5xTsn2jikFmKVQCgkKZI
XT+16JSRZrX6tLpt+JE4evg=
=zqz5
-----END PGP SIGNATURE-----

Send a report that this bug log contains spam.