Bug 292375

I expect, it doesn't fail, its just that the rules are created at the temporary location. That happens when /etc/ is not writable during network card initialization. These are persistent rules for the current hardware.

If the installer wants to use these rules, it should move the data to the real file in /etc, or remove all tmp files before continuing.

>> That happens when /etc/ is not writable during network card
initialization.

Steffen, is there any read-only mount when kernel modules for eth cards are loaded?

/etc is always writable, subdirectories might not.
Which are relevant for udev?

for udev is relevant /etc/udev/rules, but this _IS_ writable

Kay, how to solve that?

I can't reproduce this. In the "normal" system it works as expected. What do I need to do to run into it?

Yes, it works on normal system.

But during installation (begin of installation is enought).
Try to find /etc/udev/rules.d/70-* : no such file
Force trigger: for a in /sys/class/net/*/uevent; do echo -n add > $a; done : still no such file

try "/lib/udev/write_net_rules.rules all_interfaces" : same result

Problem is that there is /dev/.udev/tmp-rules--70-persistent-net.rules lock file (when I comment "lock" in write_net_rules.rules it works fine)

What I only want is from YaST copy /etc/udev/rules.d/70-net_persistent_names.rules to installed system to be sure that indexes in ethX device names will not be changed after reboot

Hmm, still no idea what's going wrong here.

Can't you just remove all previous (maybe inconsistent) state at that stage?
  rm -rf /dev/.udev/.lock-*
  rm /dev/.udev/tmp-rules*
  rm /etc/udev/rules.d/70-persistent-net.rules

And re-trigger all net events:
  udevtrigger --subsystem-match=net

Oh, and there is another bug in udev which may fail to create new rules. I'm checking-in a new package now.

Michal, i assign that to you, because it's your turn now to test kay's proposal.
And bnc-team-screening is totally wrong.

Please perform these steps from last comment at the end of stage 1 and copy 70-per*net.rules only then to the installed system if there is no such file in installed system. sysconfig.rpm:post converts old 30-net*rules to the new file in case of update.

ok, fixed (we will see ;-))

*** Bug 293366 has been marked as a duplicate of this bug. ***

*** Bug 307162 has been marked as a duplicate of this bug. ***

Okay, we found out the source of this failure:

Udev writes /dev/.udev/tmp-* files because it could not check if /etc/ is writeable due to a missing ln in our initrd. The /dev/.udev/.lock-* dir could not be removed because rmdir is also missing.

Steffen, can you please add ln and rmdir to our initrd?

Michal, Katarina: can you please remove all workarounds which removed the files in comment #7. Furthermore YaST should not call write_net_rules anymore! All you have to do is to copy /etc/udev/rules.d/70-persistent-net.rules and 70-persistend-cd.rules into the installed system at the end of stage1 and only if these files are *not* there.

added ln & rmdir

Thanks Steffen. Now it's time to fix it in YaST.

*** Bug 303862 has been marked as a duplicate of this bug. ***

ok, go ahead.

Ok, removed all the workarounds. 

But now, there must be *something* in inst-sys that creates udev rules (70-persistent-net.rules & friends) before YaST tries to copy them, since up to now they were created by above mentioned workarounds - removing related lockfiles and temporary rules, and calling udevtrigger afterwards.

We fixed this already and now udev creates these rules and save them to /etc/udev/rules.d/

Let's call this fixed until it's made a blocker again

*** Bug 326001 has been marked as a duplicate of this bug. ***

Tested with Build 1022 + modified inst-sys

/etc/udev/rules.d/70-persistent-net.rules is copied

but

/etc/udev/rules.d/70-persistent-cd.rules

isn't :(((

checking why...

Hmm, second time it went OK.
Both files were copied.