163 beta completely borked my installation with kernel panic

After the update (from 162 beta), booting resulted in a kernel panic. Relevant lines:


Run /init as init process
/bin/sh: symbol lookup error: /lib/libc.so.6: undefined symbol: _dl_fatal_printf, version GLIBC_PRIVATE
Kernel panic - not syncing: Attempted to kill init! exitcode=0x00007f00

…etc. The machine is a 64-bit Atom. I had to reinstall everything, but imaged the system, so I have access to all logs.

In the update log, there are a massive amount (2203 lines!) of

/usr/lib/dracut/modules.d/90kernel-modules/module-setup.sh: line 17: warning: command substitution: ignored null byte in input

These come after the line

*** Including module: kernel-modules ***

Hi,

thank you very much for your post! Glad to receive some testing feedback. :slight_smile:

I had to reinstall everything, but imaged the system, so I have access to all logs.

That’s a professional way of doing this. Great!

Could you please post or attach (probably works better) the update log here? The upgrade went smooth on my machine, so I am curious what went wrong in your case.

Cc: @arne_f

Thanks, and best regards,
Peter Müller

update-core-upgrade-163.zip (19.7 KB)

Certainly.

Strangely enough, the previous upgrade log (162) was also touched a minute or two before the 163 log was created, dunno if this is as it should be.

This is a problem with bash and dracut. (It’s not really a faulty but the warning is not suppressed.)

If you are on testing or next this is normal. (pakfire always reinstall the current version to make sure that you have all changes that maybee added after you has installed the version)

2 Likes

Hi,

skimming through the upgrade log, I am puzzled by this log message:

Could not find 'strip'. Not stripping the initramfs.

Also, the

cat: write error: Broken pipe

one does not sound good at all.

@arne_f: Do you have an idea regarding this? Can we make sure dracut was executed correctly?

Thanks, and best regards,
Peter Müller

1 Like

The strip is no problem because all our binaries are already stripped at build.
But the write error sounds like not enough disk space.

But this attempted to kill init looks bad. I think there is a missmatch between sysvinit and glibc.

FWIW, I did an upgrade to CU 163 on a test box and it seems to have worked okay other than the WUI never came back – after starting the upgrade from the WUI. Waiting and watching the console till things quiesced, I rebooted. It seems okay but I have done minimal testing as this box is downstream behind an edger outer hanging off my blue zone – damn little traffic there.
That said, it seems functional :grinning:

/boot was 128 MB. It’s not filled now, at least.
This system has always been on testing, and was once installed as CU136 if I am to believe the upgrade logs.

Hi,

during the last few days, I did some additional testing, and was unable to break any machines or VMs I had at hand.

So, whatever bug you were running into, it does not seem to be a deterministic one (I love those)… :expressionless:

Thanks, and best regards,
Peter Müller

1 Like

Got a lof of these warnings upgrading from 162 to 163. Will those be suppressed or better yet not happen at all on further updates?