Hi Andreas, On Mon, 28 Apr 2014, Andreas Färber wrote:
I'm attaching my original message for your convenience. The device is there at ssh login time, and yet systemctl-restart'ing the service at that time still times out, not finding it. So it seems that systemd is caching things or not getting file system change notifications? How do these dev-foo.device thingies get created, and can I inspect them through some command or in some log? What exactly does YaST do differently when I ask to start the service there?
Hm, I'm not really sure about the difference when activating the service via YaST or systemctl. The getty services file is generated at system boot by the systemd-getty-generator. You should find the service file in /usr/lib/systemd/system/ It contains a parameter configured as: TTYPath=/dev/%I %I is the unescaped instance name, e.g. the string between the "@" character %and the suffix of the unit name. Maybe something is wrong with parsing this name. Still, this should be the same behaviour from within YaST: it would be good if I could look directly on such a system. Is it possible to get remote access to a machine where you can see this? Can you open a bug report?
I've seen a similar timeout issue once I assigned a mount point /boot to a partition residing on an SD card (/dev/mmcblk0p1), dropping me into the dracut rescue shell after timeout. Not using an initrd, this was at a point where /dev/mmcblk0p2 was being used as root already. So there must be a mismatch between /dev devtmpfs and systemd's view of it.
Device node creation is normally done via udev inside the initrd. Not sure what happens if you don't have one. When the udev device generation runs after system root is mounted, you might have a race between the mount and the udev services. Still, not sure wheter this is the same issue. Did the mount work in the running system? Regards Thomas
