Bug#969599: systemd: Valid IPv6 addresses and routes discarded erroneously under certain conditions.

2020-09-07 Thread Michael Biebl
Am 07.09.20 um 20:52 schrieb Colm Buckley:
> Package: systemd
> Version: 246.4-1~bpo10+1
> Followup-For: Bug #969599
> 
> Hey folks -
> 
> I had a productive conversation with one of the upstream authors at 
> https://github.com/systemd/systemd/issues/16719;
> we think we have found the root cause of this issue. A new version of the 
> upstream PR is being merged and (we hope)
> will be backported to 246-stable. Hopefully Debian can then integrate it.
> 
> If it is causing significant issues in the meantime, the commits badd492, 
> 99a2878, 501b09d and 5055072 referenced at
> https://github.com/systemd/systemd/pull/16725 patch cleanly against the 
> current debian-bpo source and result in a
> corrected package.

Awesome. Thanks a lot for your efforts, very much appreciated!
I'll see that those commits are applied to unstable and bpo in a timely
manner once they have been merged upstream.

Regards,
Michael



signature.asc
Description: OpenPGP digital signature


Bug#969599: systemd: Valid IPv6 addresses and routes discarded erroneously under certain conditions.

2020-09-07 Thread Colm Buckley
Package: systemd
Version: 246.4-1~bpo10+1
Followup-For: Bug #969599

Hey folks -

I had a productive conversation with one of the upstream authors at 
https://github.com/systemd/systemd/issues/16719;
we think we have found the root cause of this issue. A new version of the 
upstream PR is being merged and (we hope)
will be backported to 246-stable. Hopefully Debian can then integrate it.

If it is causing significant issues in the meantime, the commits badd492, 
99a2878, 501b09d and 5055072 referenced at
https://github.com/systemd/systemd/pull/16725 patch cleanly against the current 
debian-bpo source and result in a
corrected package.

Thanks for your help!

Colm



Bug#969599: Info received (Bug#969599: systemd: Valid IPv6 addresses and routes discarded erroneously under certain conditions.)

2020-09-05 Thread Colm Buckley
The PR referenced earlier does ameliorate the issue; however I don't think
it's a complete fix, as there are worrying messages in the debug log, and
there is still nonzero packet loss. I will follow up with upstream.

I would encourage Debian users to remain with systemd 245, and not to
migrate this version to stable, until this issue is resolved.

Colm


Bug#969599: systemd: Valid IPv6 addresses and routes discarded erroneously under certain conditions.

2020-09-05 Thread Colm Buckley
That PR patches cleanly against the Debian source; so I'm building a local
package version now to test.

Will follow up here and with upstream.

Colm


On Sat, 5 Sep 2020 at 20:48, Michael Biebl  wrote:

> Am 05.09.20 um 21:31 schrieb Colm Buckley:
> > Package: systemd
> > Version: 246.4-1~bpo10+1
> > Severity: important
> > Tags: ipv6
> >
> > Dear Maintainer,
>
>
>
> >
>
>
>
> > The current BPO release of systemd includes a serious regression; valid
> SLAAC v6 addresses and routes
>
>
> > are discarded under certain circumstances. I believe this to be an
> instance of
>
>
> > https://github.com/systemd/systemd/issues/16719 which is possibly fixed
> by https://github.com/systemd/systemd/pull/16725
>
>
> >
>
>
>
> > The user-visible symptom is that the systems stop responding on their
> SLAAC-configured v6 addresses until the link
>
>
> > is next reconfigured; some time later they drop these addresses again.
>
>
>
> >
>
>
>
> > An extract from the debug log from systemd-networkd follows: note in
> particular the "Removing address" and "Forgetting address"
>
>
> > lines in the log.
>
>
>
> >
>
>
>
> > I think that pull/16725 should be cherry-picked into Debian as soon as
> possible.
>
>
>
> Thanks for the bug report.
> Could you test the pull request and verify that this fixes your issue
> (and ideally also report back to the upstream bug report).
> Once the PR is merged upstream, we can consider cherry-picking it.
>
> Michael
>
>

-- 
Colm Buckley | c...@tuatha.org


Bug#969599: systemd: Valid IPv6 addresses and routes discarded erroneously under certain conditions.

2020-09-05 Thread Michael Biebl
Am 05.09.20 um 21:31 schrieb Colm Buckley:
> Package: systemd
> Version: 246.4-1~bpo10+1
> Severity: important
> Tags: ipv6
> 
> Dear Maintainer,  
>   
>   
>   
>   
>   
> The current BPO release of systemd includes a serious regression; valid SLAAC 
> v6 addresses and routes   
>   
> are discarded under certain circumstances. I believe this to be an instance 
> of
> 
> https://github.com/systemd/systemd/issues/16719 which is possibly fixed by 
> https://github.com/systemd/systemd/pull/16725 
>  
>   
>   
>   
> The user-visible symptom is that the systems stop responding on their 
> SLAAC-configured v6 addresses until the link  
>   
> is next reconfigured; some time later they drop these addresses again.
>   
>   
>   
>   
>   
> An extract from the debug log from systemd-networkd follows: note in 
> particular the "Removing address" and "Forgetting address"
>   
>  
> lines in the log. 
>   
>   
>   
>   
>   
> I think that pull/16725 should be cherry-picked into Debian as soon as 
> possible. 
>  

Thanks for the bug report.
Could you test the pull request and verify that this fixes your issue
(and ideally also report back to the upstream bug report).
Once the PR is merged upstream, we can consider cherry-picking it.

Michael



signature.asc
Description: OpenPGP digital signature


Bug#969599: systemd: Valid IPv6 addresses and routes discarded erroneously under certain conditions.

2020-09-05 Thread Colm Buckley
Package: systemd
Version: 246.4-1~bpo10+1
Severity: important
Tags: ipv6

Dear Maintainer,

  


  
The current BPO release of systemd includes a serious regression; valid SLAAC 
v6 addresses and routes 

are discarded under certain circumstances. I believe this to be an instance of  

  
https://github.com/systemd/systemd/issues/16719 which is possibly fixed by 
https://github.com/systemd/systemd/pull/16725   
   


  
The user-visible symptom is that the systems stop responding on their 
SLAAC-configured v6 addresses until the link

is next reconfigured; some time later they drop these addresses again.  

  


  
An extract from the debug log from systemd-networkd follows: note in particular 
the "Removing address" and "Forgetting address" 
  
lines in the log.   

  


  
I think that pull/16725 should be cherry-picked into Debian as soon as 
possible.   
   


  
Thanks, 

  


  
 Colm   

  


  


  
Sep 05 20:23:13 lugh systemd-networkd[1148]: NDISC: Received Router 
Advertisement: flags none preference medium lifetime 0 sec  
  
Sep 05 20:23:13 lugh systemd-networkd[1148]: NDISC: Invoking callback for 
'router' event. 

Sep 05 20:23:13 lugh systemd-networkd[1148]: int0: Configuring route: dst: 
fd79:b3fc:4a5b:1::/64, src: n/a, gw: