Bug#918322: initramfs-tools: kernel fails to boot with "Gave up waiting for root file system device"

[Michael Biebl]

Hi Laurence!

Am 06.01.19 um 19:29 schrieb Laurence Abbott:
> I'm pretty sure I came across the same issue yesterday: I ran a
> dist-upgrade on an X99-based system (the first for several months) and
> the system was left unbootable. My root system is on raid and booting
> failed with mdadm errors about it not being able to find any of the
> raid arrays listed in /etc/mdadm/mdadm.conf.
> 
> From the initramfs shell (having eventually found a PS/2 keyboard!), I
> could see no disks or partitions under /dev. I even tried booting off
> a (recent - downloaded to rescue this system!) testing installation
> DVD and had the same problem: it got so far and then lost the DVD
> drive! As before, a shell showed no disk or partition devices under
> /dev.
> 
> After a lot of panic, I eventually worked out that a simple "modprobe
> ahci" at the initramfs prompt and exiting it allowed the system to
> boot properly.
> 
> The problem (at least for me) is that the ahci module isn't getting
> loaded automatically on boot. Load it by hand and it boots normally. I
> suspect this would work for others with a similar setup.

Yeah, this seems to be the problem. The auto-loading of modules is no
longer properly triggered by udev.

> I don't know enough about the minutiae of the boot process to debug
> this much more but I assume that it is udev that should be pulling in
> any needed modules based on the hardware. As part of the upgrade
> (looking back now), udev was upgraded from 239-13 to 240-2.
> 
> I did think about trying to get the initrd to force loading the ahci
> module but that's never been needed in the past!

If you look into /usr/share/initramfs-tools/scripts/init-top/udev the
following command are/should be responsible for triggering the modules
auto-loading:

udevadm trigger --type=subsystems --action=add
udevadm trigger --type=devices --action=add

For some reason, this no longer works properly under certain conditions.

When you are dropped into a rescue shell, can you execute the above
commands (again). Do they trigger the auto-load of the modules or not?

Once we know that, we can further debug this.

Btw, if you are online via IRC, we might debug this more interactively.
If so, please come by #debian-systemd on oftc. My nick is mbiebl.

Regards,
Michael





-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <http://alioth-lists.debian.net/pipermail/pkg-systemd-maintainers/attachments/20190106/2215f630/attachment.sig>