[SOLVED] Mga2 hangs on boot

[mittfh]

[EDIT: Resolved via reformat and fresh install.]

Yesterday I attempted upgrading from Mageia 1 to 2 (via the new release notification - so installing while still logged on)

The download / installation of packages went OK (apart from a minor hitch relating to some notification libraries attempting to install before their dependencies - which then installed themselves successfully later on), but when I rebooted the problems began.

Grub loaded OK, followed by a progress bar at the bottom of the screen (well, a three-in-one progress bar - a dark blue stippled bar started first, then as that progressed, a lighter blue bar started progressing on top of it, followed by a white bar), puzzlingly though advancing towards text at the bottom right of the screen saying "Mageia 1". After several minutes of watching that, it dumped me in "Emergency mode". Obligingly I pressed Ctrl + D to try continuing, but it hung on the second bubble.

Reboot, try again - same problem (well, I had to verify it wasn't just a temporary glitch that would resolve itself!).

Reboot, this time hit F3 in Grub, choose Default and remove "splash" from the boot command so I can see what (if anything) falls over en-route.

After the reassuring "Welcome to Mageia 2" message, I spot a bunch of lines about failing to load cryptography. After disconnecting my external USB HDD (encrypted) and commenting out its line in /etc/fstab (thinking that if it hasn't got to worry about trying to mount an encrypted drive, it might have more success), I try again.

Almost immediately:

Code: Select all

udevd[204]: Cannot find module by name="r8169"

Various modules then load OK, but there is a mention of failing to load cryptography (possibly due to a dependency - the line flashed by too fast for me to scribble it down). Also failing to start:

Code: Select all

Failed to start CPU frequency scaling daemon
See 'systemctl status cpufreq.service' for details

Failed to start LSB: cpudynd
See 'systemctl status cpudyn.service' for details

Failed to start LSB: TOMOYO Linux MAC logging daemon
See 'systemctl status tomoyo-auditd.service' for details

Eventually:

Code: Select all

Started LSB: Bring up/down networking [ OK ]
Starting LSB: Wait for the hotplugged network to be up...

Whereupon the system hangs.

Try rebooting, this time selecting "safe settings" from the kernel options dialog and removing splash again, and it follows much the same process, except it hangs slightly earlier:

Code: Select all

Started network manager [ OK ]
Starting LSB: Bring up/down networking...

So anyone any ideas on how I can get my system working again - if need be I can buy a Linux magazine and boot from one of the live distros they inevitably include to be able to access my HDD / the 'net.

[EDIT] A quick google reveals that r8169 is the onboard network adaptor. But even so, if it can't load the module, why can't it just display a [ FAIL ] then continue booting to a command prompt? If I can somehow get it to boot to a (standalone) command prompt, then I can use a live distro to download the module (onto a USB stick if necessary), then boot up Mga2, compile the module against the 3.3 kernel and hopefully get 'net access within Mga2. Not to mention researching more into the other bits that fell over / urpmi-ing them / getting to a GUI desktop...

[joeshmoe]

Maybe you could start in single user mode. Quick search suggests appending kernel parameter '1' or 'single'.

[mittfh]

I'll try that when I get home - if it works, I'll hopefully be able to dive into the log files, find out exactly what fell over at boot, then use a Live distro to research further / download stuff onto the HDD (I assume since they come with scripts to allow themselves to be installed onto your HDD, they can access your HDD without installing themselves on it!)

Fingers crossed... it'll be about six or more hours before I can find out as I'm currently at work (gazing out at the mini heatwave - for once the office is cooler than outside!), but before I leave I'll check back to see if anyone else has replied with additional useful hints / tips...

[joeshmoe]

Best research that kernel parameter before you go home. I don't know that it works or is the same across distributions.

Maybe you can run draktools in single user mode. Or try 'init 3' to start networking and see what happens.

Yes, a live medium can access your harddisk non-destructively. If you use a Mageia live medium, you can use the partioning tool in drakconf to sort out and mount your partitions. Otherwise, as a fallback, 'fdisk -l' and mount them manually.

Good luck!

[arelatensis]

I can repeat word by word the first post of Mittfh. But my may problem is that system hangs by detecting, and attempting to mount windows partitions, so:

systemctl status media win_c. mount
Dependency failed. aborted start of /media/win_c
Dependency failed. aborted start of setup links in /boot for running kernel
Dependency failed. aborted start of Mark the need to relabel after reboot
Dependency failed. aborted start of Relabel all filesystems if necessary

[mittfh]

Argh! Runlevel 1 doesn't auto-load networking - so far, so good. It at least means I can access my system.

However...

I've reinstalled the r8169 driver, and indeed can get the network up and running from runlevel 1 via:

Code: Select all

ifconfig eth0 up
ifconfig eth0 192.168.1.64
route add default gw 192.168.1.254


I can then load Lynx and browse away - albeit with the usual limitations of Lynx. So therefore there's nothing wrong with the module, and /etc/sysconfig/network_scripts/eth0 is still intact and retains the settings from mga1.

However, higher runlevels still won't work, with udevd still being unable to find the module...

Code: Select all

udevd[200]: Could not find module by name="r8169"

This is the very first line of text to appear in the boot process (once grub clears), before the "Welcome to Mageia 2" message. To attempt to compensate, I've tried adding r8169 to /etc/modprobe.preload (no effect) and modprobe.conf (again, no effect).

What now?! How do I persuade udevd to see r8169? If there's an easier way than backing up everything, downloading the Mga2 iso and reinstalling, I'd like to know...

[joeshmoe]

Just noticed I forgot it's systemd now. So what I said about init was misleading or plain wrong; sorry. Hope someone can help you.

[arelatensis]

What now?! How do I persuade udevd to see r8169? If there's an easier way than backing up everything, downloading the Mga2 iso and reinstalling, I'd like to know...

I made a clean installation also, but the problems with devices and partitions remain. The cause is systemctl, that is part of systemd - "systemctl: used to introspect and control the state of the systemd system and service manager". Systemctl calls replace their old workflow in SysVinit. So, it seems, that the cause of the troubles laid in systems itself.

[mittfh]

Well, kernel-firmware-nonfree fixed the "Could not find module by name" but had no effect on anything else.
Even adding all the tainted / nonfree repositories (courtesy of urpmi.addmeida --distrib --interactive / uprmi --replacefiles --auto-update --auto) didn't work.

Nor an upgrade install from the DVD iso.

So it looks as though I'm left with the nuclear option: take backups of /home, /opt and /usr/local, nuke the partitions and do a fresh install.

[mittfh]

24 hours later, and I resorted to the nuclear option: back up /home, /usr/local and /opt, nuke the partitions, and reinstall.

I've now got a working system again, albeit with a few niggles it's probably best to launch new threads for...