>This is the pertinent part of the "w" output:
> 5:25pm up 5 days, 6:43, 9 users, load average: 8.92, 8.06, 6.89
>someone ttyp5 Thu 9am 2days -
>I should add that "ps" shows no ttyp5 or someone processes.
>I tried "chown root /dev/ttyp5; chgrp root /dev/ttyp5" (they did the changes)
>But, the same "w" output still... What's going on here?
I have seen this too. In one case I know of, it happens when a
login shell gets orphaned in a 'peculiar' way:
# the numbers indicate different shells. A (#) indicated login shell
$(1) screen (screen is not suid so it doesn't make an /etc/utmp entry and
therefore only the login shell that called screen is recorded)
$2 exit (kill screen)
$(1) now back to original login shell and output of 'w' shows that '-'
process which has no pid associated with it but it does have an
The only way that I know of to kill this process is to reboot the machine.
Normally, init would inherit such processes and their pid would be 1 but
in this case I can't tell what is going on. Anyone else know exactly what
is happening? I don't think the second login shell is ever recorded
properly in /etc/utmp since you called login (which does write to utmp)
from a shell that was not recorded. Obviously, what was known about it
was written (like username, tty, etc) but no location. Why can't I
reconnect to it though and why doesn't it show up on a ps?
From the Lab of the MaD ScIenTiST: