Red Hat Bugzilla – Bug 827847
An invalid entry in fstab locks up boot for 90 seconds
Last modified: 2012-06-04 07:03:33 EDT
Description of problem:
If there is an invalid entry in /etc/fstab that is set to be mounted on boot, then the entire boot process locks up completely for approximately 90 seconds before it displays the emergency mode prompt.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Add an invalid entry to /etc/fstab, e.g "/dev/sdz9 /mnt vfat defaults 1 1"
2. Save & reboot.
The system hangs on boot.
Either that the emergency prompt appears *immediately* after the mount fails, or that the system simply ignores that it fails and boots normally since this mount is NOT a dependency for *anything*. The 90 second delay where nothing can be done except waiting is incredibly annoying.
When dropping into recovery mode and rebooting using the "reboot" command, even that sometimes hangs.
(In reply to comment #0)
> Expected results:
> Either that the emergency prompt appears *immediately* after the mount
We do not know in advance if the block device does not exist at all or if it just takes a long time to discover by kernel+udev.
> or that the system simply ignores that it fails and boots normally
> since this mount is NOT a dependency for *anything*.
We cannot safely assume that the filesystem is not important.
There's the "nofail" parameter which you can use to express this property.
> When dropping into recovery mode and rebooting using the "reboot" command,
> even that sometimes hangs.
That sounds like a bug for which we'd like a dedicated bugreport with more details.
Is it possible to make Fedora print more verbose status messages during boot? I cannot remember this ever happening on older Fedoras, so when my system became unresponsive during boot, I was always hitting the reset button before it timed out. 90 seconds is a LONG time to wait without any indication that something is happening.
I suspect that with systemd-44-12.fc17 you would probably get boot a message like:
Expecting device /dev/sdz9...
but you may easily overlook it because of messages that are printed later. The boot is parallel and the last message is not necessarily the most important one.
To help with these kinds of issues, one day we'll probably start printing the list of running jobs if nothing is happening during boot for 15 seconds or so. Currently there is nothing like it built-in.