On Sun, 23 Dec 2018 01:51:00 +0100 Wolfgang Walter
<wolfgang.wal...@stwm.de> wrote:
> Package:  udev
> Version: 240-1
> Severity: important
> 
> After upgrading from 239-15 to 240-1 or higher my network setup breaks.
> 
> The breakage is similar to the last one I reported (bug #904198) though 
> different in detail.
> 
> This times all my vlan-devices on a physical device netint are first renamed 
> to renameX and than udev tries to rename them again but this time it tries to 
> give them all the name netint - which of course fails.
> 
> A setup like the following one should be able to reproduce this problem:
> 
> 10-netint.link:
> [Match]
> MACAddress=00:01:2e:77:a5:45
> 
> [Link]
> Name=netint
> WakeOnLan=off
> 
> 
> netint.network:
> [Match]
> Name=netint
> 
> [Network]
> VLAN=kbs
> LinkLocalAddressing=no
> DHCP=no
> 
> 
> kbs.link
> [NetDev]
> Name=kbs
> Kind=vlan
> 
> [VLAN]
> Id=10
> 
> 
> 
> In the journal you should see something like
> 
> kernel: rename31: renamed from kbs
> ....
> systemd-networkd[852]: kbs: Interface name change detected, kbs has been 
> renamed to rename31
> ....
> [347]: kbs: Failed to rename network interface 31 from 'kbs' to 'netint': 
> Device or resource busy
> ....
> 
> 
> 
> Downgrading udev to 239-15 fixes it in principle that is this rename does not 
> happen.
> 
> systemd-networkd from 240-1 though crashes with
> 
>       Assertion 'IN_SET(link->state, LINK_STATE_CONFIGURING, 
> LINK_STATE_CONFIGURING, LINK_STATE_FAILED, LINK_STATE_LINGER)' failed at 
> ../src/network/network/networkd-link.c:934, function address_handler(). 
> Aborting.
> 
> Therefor I also had to downgrade systemd back to 239-15.
> 

Someone mentioned on IRC:

<grawity> hmm, 917128 actually comes pre-broken
<grawity> the reporter is applying renaming based purely on MACAddress,
and that's *already* a problem when VLANs are involved
<grawity> because they all have the same MAC address as the underlying
physical interface
<grawity> (which I found out the hard way years ago, with just udev
.rules too)
<grawity> so the only change is that it was a race previously and a
crash now

Attachment: signature.asc
Description: OpenPGP digital signature

Reply via email to