Bug 616765

The check in /sbin/ifup-dhcp is not sufficient with newer kernels.
Going to fix it.

Created attachment 375233 [details]
Patch fixing the ipv6 support check in ifup-dhcp

Note,

that when ipv6 is disabled for the interface only, for example using:

   net.ipv6.conf.eth0.disable_ipv6 = 1

in /etc/sysconfig/network/ifsysctl, this check will not work because
of kernel bug #559170. But it works fine, when it is disabled at all
(net.ipv6.conf.all.disable_ipv6 = 1) as done by yast2 network.

Fixed wrong reassing to Michal instead of needinfo...

Michal, does yast2 network restart the network (completely) or inform/ask
the user to reboot when disable_ipv6 setting has been changed?

BTW:
Another way to fix the problem is to fix the interface configuration,
that is the /etc/sysconfig/network/ifcfg-<interface> file, and use a
BOOTPROTO=dhcp4 and not BOOTPROTO=dhcp (start both, v4 + v6 clients).

Submitted the fix to OBS/Base:System/sysconfig and openSUSE:Factory:

 42834  State:new     By:mtomaschewski When:2010-07-12T13:27:52
        submit:       Base:System/sysconfig  ->  openSUSE:Factory       
        Descr: Bug 616765 - IPv6 not enabled in Yast, yet dhclient spams logs
               with ipv6 failures

http://download.opensuse.org/repositories/Base:/System/openSUSE_Factory/

sysconfig gets quite some update usually, perhaps we should collect some issues first.

(does not look that worrysome of an issue).

Only for the next sysconfig update.

(In reply to comment #8)
(In reply to comment #9)

OK, thanks!

Workaround is to change BOOTPROTO=dhcp to BOOTPROTO=dhcp4 in ifcfg-<ifname>
files -- the user disabled ipv6 anyway.

Marius, sorry for no response (vacation). YaST informs user that reboot is needed to correctly apply changes.

*** Bug 620298 has been marked as a duplicate of this bug. ***

Why should a *reboot* be necessary? 
That seems rather silly, when a change to the ifcfg file is sufficient.

In my experience (without reboot) some applications like Firefox, Thunderbird could randomly crash.

Is that still true? I thought that particular issue (the IPv6-goes-away and firefox-crashes bug) was fixed.

IN any case, I'd much rather see "you may have to restart some networking applications if you are disabling IPv6" instead of "reboot your computer".

It's not like we run Windows or anything.

(In reply to comment #15)
> Is that still true? I thought that particular issue (the IPv6-goes-away and
> firefox-crashes bug) was fixed.
> 
> IN any case, I'd much rather see "you may have to restart some networking
> applications if you are disabling IPv6" instead of "reboot your computer".

Short:
 The problem is, that "some networking applications" is an insufficient
 action description and may be also technically not sufficient.

Generally, it should be possible to handle it completely manually... it
should be not even strictly required to restart the network when the
disable_ipv6 way is effective (as on 11.3), the kernel will then remove
all ipv6 addresses itself...
But on older systems/kernels, ipv6 may be a module and yast2 disables the
ipv6 module loading. It is not possible to unload the module -- only a
reboot works here.

It is not easy even for advanced user with disable_ipv6, because there
are several side effects / corner cases.

There are not only GUI applications like firefox, but also several services,
possibly inclusive of X11, that may have bound ipv6 ports or even handle
ipv4 connections through the ipv6 socket and at least side effects have to
be taken into account.

Last but not least: the above dependencies for the applications may be
also true for things running inside of the kernel. It may include also
another type of problems: see bug 559170.

Another issue is: disabling ipv6 usually introduces more problems than
it solves and just delays the correct solutions. It was never required
for me to disable ipv6 -- except for tests.


Anyway: fixed for me, fix will be picked up in the next update.

Submitted the fix to the 11.3 update test queue:

 43469  State:new     By:mtomaschewski When:2010-07-20T11:13:19
        submit:       home:mtomaschewski:branches:openSUSE:11.3:Update:Test/sysconfig
 -> openSUSE:11.3:Update:Test   
        Descr: fix for bug 616765 (next update)

test packages / previews will appear at:

http://download.opensuse.org/repositories/home:/mtomaschewski:/branches:/openSUSE:/11.3:/Update:/Test/openSUSE_11.3_Update_Test

=> Technically fixed.

Update released for: sysconfig, sysconfig-debuginfo, sysconfig-debugsource
Products:
openSUSE 11.3 (debug, i586, x86_64)

fix was incorrect.

Created attachment 399687 [details]
Correction of the fix.

The result was always HAVE_IPv6=no after the previous patch.

digging for solutions to my bug,

"libcanberra.so failing to connect to pulseaudio results in menu hangs in Firefox"
 https://bugzilla.novell.com/show_bug.cgi?id=632530

which has 

  --   kernel /vmlinuz ...
  ++   kernel /vmlinuz ... ipv6.disable=1

as a workaound, I found your comments:

 (In reply to comment #16)
 ...
 > It is not easy even for advanced user with disable_ipv6, because there
 > are several side effects / corner cases.
 > 
 > There are not only GUI applications like firefox, but also several services,
 > possibly inclusive of X11, that may have bound ipv6 ports or even handle
 > ipv4 connections through the ipv6 socket and at least side effects have to
 > be taken into account.
 ...

_are_ these issues, possibly, related?  tbh, i'm not at all clear ...

(In reply to comment #21)
> _are_ these issues, possibly, related?  tbh, i'm not at all clear ...

No.
The mentioned comment was about changing to "disable_ipv6=1" at runtime
and the problems that may happen when you don't reboot after.

Mr Maintenance,

we have a fix of the fix (see comment 20). For next update?

I've submitted it already to openSUSE:11.3:Update:Test -- currently there
are following fixes / submissions in the queue:

IBS Request #9421:
  submit:   home:mtomaschewski:branches:SUSE:SLE-11-SP1:Update:Test/sysconfig
            ->  SUSE:SLE-11-SP1:Update:Test   
  message:  bug 637183,616765,644738,638508

OBS Request #54941:
  submit:   home:mtomaschewski:branches:openSUSE:11.3:Update:Test/sysconfig
            -> openSUSE:11.3:Update:Test/sysconfig
 message:   bug 616765,644738,637183

OBS Request #53146:
  submit:   home:mtomaschewski:branches:openSUSE:11.2:Update:Test/sysconfig
            -> openSUSE:11.2:Update:Test/sysconfig
 message:   next update, bnc#637183

opensuse maintenance is public, un-privatized previous comment.


I think we could do a sysconfig update now for openSUSE *  +1

The SWAMPID for this issue is 37696.
This issue was rated as low.
Please submit fixed packages until 2011-01-05.
Also create a patchinfo file using this link:
https://swamp.suse.de/webswamp/wf/37696

+1 ... update started.

and accepted

Update released for: sysconfig, sysconfig-debuginfo, sysconfig-debugsource
Products:
openSUSE 11.3 (debug, i586, x86_64)

Update released for: sysconfig, sysconfig-debuginfo, sysconfig-debugsource
Products:
SLE-DEBUGINFO 11-SP1 (i386, ia64, ppc64, s390x, x86_64)
SLE-DESKTOP 11-SP1 (i386, x86_64)
SLE-SERVER 11-SP1 (i386, ia64, ppc64, s390x, x86_64)
SLES4VMWARE 11-SP1 (i386, x86_64)

This is an autogenerated message for OBS integration:
This bug (616765) was mentioned in
https://build.opensuse.org/request/show/52837 Factory / sysconfig