Getty flooding syslog after kernel upgrade with "cannot open as standard input"

Question

I just updated my kernel on an Ubuntu 9.10 system from 2.6.32.1 to 2.6.33.5, and everything seemed to go well, except now my system log is getting flooded with errors from getty:

Jul 25 11:01:54 kobol getty[2414]: /dev/tty4: cannot open as standard input: No such device
Jul 25 11:01:54 kobol getty[2415]: /dev/tty2: cannot open as standard input: No such device
Jul 25 11:01:54 kobol getty[2416]: /dev/tty6: cannot open as standard input: No such device
Jul 25 11:01:54 kobol getty[2417]: /dev/tty3: cannot open as standard input: No such device
Jul 25 11:01:54 kobol getty[2418]: /dev/tty5: cannot open as standard input: No such device
Jul 25 11:01:54 kobol init: tty4 main process (2409) terminated with status 1
Jul 25 11:01:54 kobol init: tty4 main process ended, respawning
Jul 25 11:01:54 kobol init: tty2 main process (2410) terminated with status 1
Jul 25 11:01:54 kobol init: tty2 main process ended, respawning
Jul 25 11:01:54 kobol init: tty6 main process (2411) terminated with status 1
Jul 25 11:01:54 kobol init: tty6 main process ended, respawning
Jul 25 11:01:54 kobol init: tty3 main process (2412) terminated with status 1
Jul 25 11:01:54 kobol init: tty3 main process ended, respawning
Jul 25 11:01:54 kobol init: tty5 main process (2413) terminated with status 1
Jul 25 11:01:54 kobol init: tty5 main process ended, respawning

Any idea what is causing this and how to get it to stop? I realize its not a critical issue, but it's still annoying that it's constantly happening (and filling up log files) and I don't know why.

UPDATE: This is the listing for my tty devices, the permissions of which seem to be reset on reboot

crw------- 1 root root 4, 0 2010-07-26 10:00 /dev/tty1
crw--w---- 1 root tty  4, 1 2010-07-26 10:00 /dev/tty2
crw--w---- 1 root tty  4, 2 2010-07-26 10:00 /dev/tty3
crw--w---- 1 root tty  4, 3 2010-07-26 10:00 /dev/tty4
crw--w---- 1 root tty  4, 4 2010-07-26 10:00 /dev/tty5
crw--w---- 1 root tty  4, 5 2010-07-26 10:00 /dev/tty6
crw--w---- 1 root tty  4, 6 2010-07-26 10:00 /dev/tty7
crw--w---- 1 root tty  4, 7 2010-07-26 10:00 /dev/tty8

And the init script for tty1:

# tty1 - getty
#
# This service maintains a getty on tty1 from the point the system is
# started until it is shut down again.

start on stopped rc RUNLEVEL=[2345]
stop on runlevel [!2345]

respawn
exec /sbin/getty -8 38400 tty1

And the (practically identical) init scripts for tty2-tty6:

# tty2 - getty
#
# This service maintains a getty on tty2 from the point the system is
# started until it is shut down again.

start on runlevel [23]
stop on runlevel [!23]

respawn
exec /sbin/getty -8 38400 tty2

UPDATE 2: Permissions for the init scripts follow...

-rw-r--r-- 1 root root 228 2009-12-10 11:00 /etc/init/tty1.conf
-rw-r--r-- 1 root root 213 2009-12-10 11:00 /etc/init/tty2.conf
-rw-r--r-- 1 root root 213 2009-12-10 11:00 /etc/init/tty3.conf
-rw-r--r-- 1 root root 213 2009-12-10 11:00 /etc/init/tty4.conf
-rw-r--r-- 1 root root 213 2009-12-10 11:00 /etc/init/tty5.conf
-rw-r--r-- 1 root root 213 2009-12-10 11:00 /etc/init/tty6.conf

A grep of running getty instances shows this:

$ ps auxww|grep [g]etty
root      2228  0.0  0.2   5992   604 ?        Ss+  10:00   0:00 /sbin/getty -8 38400 tty1
root      6352  0.8  0.2   5988   688 ?        Ss   11:06   0:00 /sbin/getty -8 38400 tty3
root      6355  0.5  0.2   5988   684 ?        Ss   11:06   0:00 /sbin/getty -8 38400 tty4
root      6356  0.5  0.2   5988   684 ?        Ss   11:06   0:00 /sbin/getty -8 38400 tty6
root      6357  0.8  0.2   5988   684 ?        Ss   11:06   0:00 /sbin/getty -8 38400 tty2
root      6358  0.6  0.2   5988   684 ?        Ss   11:06   0:00 /sbin/getty -8 38400 tty5

UPDATE 3: After downgrading to 2.6.32.1 the errors disappeared. I recently upgraded again to 2.6.34.1 and they showed up again. I'm superbly baffled now.

Solution 1

I had this same issue with a Rackspace-hosted server that I recently upgraded to Ubuntu 12.04 LTS. My /var/log/auth.log file was getting flooded with

May 13 12:50:40 foo getty[1796]: /dev/tty2: cannot open as standard input: No such device
May 13 12:50:51 foo getty[1798]: /dev/tty3: cannot open as standard input: No such device
May 13 12:50:51 foo getty[1797]: /dev/tty4: cannot open as standard input: No such device
May 13 12:50:52 foo getty[1800]: /dev/tty5: cannot open as standard input: No such device
May 13 12:50:52 foo getty[1799]: /dev/tty6: cannot open as standard input: No such device

All I did was remove (after backing up) the following files:

/etc/init/tty[2-6].conf

After I rebooted the problem was solved. No more noise in the log file and the Rackspace console still works, although with only one virtual terminal. (I never use virtual terminals on the console anyway.)

Solution 2

Look to see if those devices exist:

ls -l /dev/tty[2-6]

If they don't then do this:

cd /dev
MAKEDEV -n console

That command name is in all caps. The -n switch (for "no action") will show you the actions MAKEDEV will take. If you are satisfied with what it proposes, then change the -n to -v (for "verbose") and run it again.

Comments

[EvanK]

Good suggestion, but I checked & I have existing devices through tty8, all with root owner and group.

[Dennis Williamson]

@EvanK: Do they have appropriate major and minor device numbers?

[EvanK]

As far as I can tell. Major number is 4 for all of them, minor goes from 0 to 7 in order

[Dennis Williamson]

@EvanK: What do your /etc/init/tty[2-6].conf files look like? They should be similar to /etc/init/tty1.conf (except that it starts and stops on more runlevels). Are the permissions of /dev/tty[2-6] the same as tty1 ("crw-------" or "crw--w----")? Just for the heck of it, you might try deleting 2-6 and running MAKEDEV anyway.

[EvanK]

This is strange...I deleted 2-8 and couldnt recreate them, so I rebooted (which apparently recreated them). Now, tty1 is "crw------- 1 root root" while the others are "crw--w---- 1 root tty". I chowned & chmodded them to match tty1 & rebooted, and their permissions were reset to "crw--w---- 1 root tty" during the reboot.

[Dennis Williamson]

@EvanK: Weird. My 9.10 system has tty[1-6] all at "crw------- 1 root root". The ones that don't have corresponding /etc/init/tty?.conf (getty) are "crw--w---- 1 root tty". What does ps auxww|grep [g]etty show? What are the permissions/ownership (and dates) of /etc/init/tty?.conf?

[EvanK]

The init scripts all have the same permissions and same modified date (of presumably the date of the original install). Updated question w/ the details

[Dennis Williamson]

@EvanK: Those question marks in the output of ps tell me that something is running on those ttys. Take a look at each of these: ps auxww|grep [t]ty.|grep -v "/sbin/getty", lsof /dev/tty[1-8] and does sudo cat /dev/vcs2 show a login prompt or something else?

[EvanK]

The first two commands produce no output, and as for the third, /dev/vcs2 doesn't exist

[Dennis Williamson]

@EvanK: Do you have /dev/vcsa[1-8]?

[Dennis Williamson]

@EvanK: I'm stuck. Are you able to boot into the previous kernel and if so does this problem happen then?

[EvanK]

No, I have no vcs* in /dev/, those are video devices right? That would make sense then, this is a headless box with only console access. I will try reverting to the old kernel during off-peak hours and see if the problem persists. Thanks!

[Dennis Williamson]

@EvanK: No, the "v" stands for "virtual". They are the device entries for virtual console memory. They contain what's shown on the screen of tty[1-8]. See man vcs. I wonder if their absence could be the problem. The man page gives instructions for creating them.

[EvanK]

@Dennis: No manual entry for vcs, perhaps i need to install something, but apt-cache search vcs gives me no useful packages

[Dennis Williamson]

@EvanK: As an alternative, if you're near the server you can just Alt-2 to virtual console 2 and see if there's a login prompt (and try to log in to see if it works). Otherwise, I'm still stuck. No luck on the previous kernel?

[EvanK]

@Dennis, I missed my window yesterday, so I switched out the kernels just recently. I'm running .32.1 again and, though I still don't have any vcs devices, the flood of tty error messages has apparently stopped.

[Thorbjørn Lindeijer]

Yep, same here. Apparently the problem exist already since a while: forum.slicehost.com/index.php?p=/discussion/4974/…. Basically, when while upgrading it asks whether to put back these "removed" files, you shouldn't do it I guess. :-)