greg@cityzoo.acs.umbc.edu (Greg Sylvain,Lib 007a,3929,) (07/05/90)
Hello again,
I recently upgraded to 7.0 and I'm still having some problems. First
off, I should tell you it's a 7.0 tape that I got from a friend of mine who got
her tape back in Feburary. (Hp still hasn't shipped ours, and I couldn't wait
until it came in the mail. Too much other stuff that hinged on 7.0 was waiting
for it.) My friend said that the 3 patches had allready been integrated into
the tape. And in fact the HP S/W person was able to confirm that.
But I still have the problem where /bin/login won't let anyboy loggin.
Sometimes it will let you remsh over, but that's rare. And it turns out that
if you delete /etc/utmp and then try it, you can get in. Also, the check sum
is way of (my sum of /bin/login is :50122 200 /bin/login) what the normal sum
should be, something like 60000 or so. The person I got the tape from said she
had a similar problem after she installed, but it `went away'. Does any one
have a clue ?
Thanks alot,
greg
(i'm running 7.0 on a hp9000s300 machine)
Greg Sylvain
Academic Computing Services
Systems Programmer
UUCP: ...!{uunet}!umbc5!greg
Internet (Arpa) : greg@umbc5.umbc.edu
BITNET : GREGS@UMBCtomg@hpcvlx.cv.hp.com (Thomas J. Gilg) (07/06/90)
> But I still have the problem where /bin/login won't let anyboy loggin. > Sometimes it will let you remsh over, but that's rare. And it turns out that > if you delete /etc/utmp and then try it, you can get in. ................... The format of utmp has changed for 7.0, and pre 7.0 programs that modify utmp/wtmp/btmp need to be updated. Some programs to watch out for include: /bin/login /bin/su /etc/getty /usr/bin/X11/hpterm /usr/bin/X11/xterm If you do a fairly complete update, you should take care of the old utmp programs. Thomas Gilg tomg@cv.hp.com
rodean@hpfcdc.HP.COM (Bruce Rodean) (07/06/90)
In article <3569@umbc3.UMBC.EDU> greg@cityzoo.acs.umbc.edu (Greg Sylvain,Lib 007a,3929,) writes: > But I still have the problem where /bin/login won't let anyboy loggin. >Sometimes it will let you remsh over, but that's rare. And it turns out that >if you delete /etc/utmp and then try it, you can get in. Also, the check sum >is way of (my sum of /bin/login is :50122 200 /bin/login) what the normal sum >should be, something like 60000 or so. The person I got the tape from said she >had a similar problem after she installed, but it `went away'. Does any one >have a clue ? You most likely have a unsupported command that writes to /etc/utmp. At 7.0 the structure and/or size of /etc/utmp changed. When /etc/utmp is corrupted in this manner, you will see messages like ''No utmp entry. You must exec "login" from the lowest level "sh"''. One common offender I have heard about is telnetd. Whatever it is, it needs to be recompiled. Hope this helps. Bruce Rodean rodean@hpfcla.FC.HP.COM This does not represent any official position of Hewlett-Packard Co. I alone am responsible for its content. No guarantees given or implied.
raf@hpfcdc.HP.COM (Rick Ferreri) (07/10/90)
> Hello again, > > I recently upgraded to 7.0 and I'm still having some problems. First > off, I should tell you it's a 7.0 tape that I got from a friend of mine who got > her tape back in Feburary. (Hp still hasn't shipped ours, and I couldn't wait > until it came in the mail. Too much other stuff that hinged on 7.0 was waiting > for it.) My friend said that the 3 patches had allready been integrated into > the tape. And in fact the HP S/W person was able to confirm that. > > But I still have the problem where /bin/login won't let anyboy loggin. > Sometimes it will let you remsh over, but that's rare. And it turns out that > if you delete /etc/utmp and then try it, you can get in. Also, the check sum > is way of (my sum of /bin/login is :50122 200 /bin/login) what the normal sum > should be, something like 60000 or so. The person I got the tape from said she > had a similar problem after she installed, but it `went away'. Does any one > have a clue ? > > Thanks alot, > greg > > (i'm running 7.0 on a hp9000s300 machine) > Greg Sylvain > Academic Computing Services > Systems Programmer > > UUCP: ...!{uunet}!umbc5!greg > Internet (Arpa) : greg@umbc5.umbc.edu > BITNET : GREGS@UMBC > ---------- > The /etc/utmp file structure changed in 7.0. The 7.0 ReleaseNotes file should have had the following statement, however, it was not shipped with the on-line version. <utmp.h>: --------- Due to changes made for POSIX conformance, a new type has been introduced for 7.0. This new type is pid_t. Prior to 7.0, pid's were of type short, but for 7.0 the kernel will be treating pid's as type long. Although pid's will now be treated as longs, the value of MAXPID will not be changed in user space in the near future. Since the type pid_t is now being introduced, it was felt that this is the best time (7.0) to introduce this change to the utmp structure defined in <utmp.h>. Changing the ut_pid field of struct utmp from a short to a long (pid_t) will break object compatibility of any program which attempts to read from or write to the /etc/utmp, /etc/wtmp or /etc/btmp files. In addition to changing ut_pid to a long from a short, it was also decided to add a couple of new fields to the utmp structure since we are already breaking object compatibility. The ut_host[] field is a 16 byte array of type char (as in BSD) used to maintain the name of the remote host from which a user is logged into the local host. Also, the ut_addr field has been added. This field contains the internet address. Note that the 7.0 commands do not support these fields. Support for ut_host and ut_addr is planned for a future release of HP-UX. **If there is an application left on a 7.0 system which writes entries into the /etc/utmp file in the *old* format, you will see the following message when trying to log into the system (after that application has been executed): 'No utmp entry. You must exec "login" from lowest level "sh"' This message means that the /etc/utmp file has been corrupted. It is quite likely that the /etc/wtmp file has been corrupted too. To remedy this situation, you must "cat /dev/null > /etc/wtmp" as root. This will clear out the /etc/wtmp file. Then reboot your system (this will clear out the /etc/utmp file). After this has been done, please remove the offending executable. You may recompile the source code on the 7.0 system and then re-install it. Note that applications which manipulate tty's and pty's often create utmp entries. Please note that the corruption of the utmp file may occur because of a user application which writes data into the /etc/utmp file in the old format. If this is the case, then the simpelest solution is to recompile that source code on your 7.0 system. Another possibility is that you haven't updated the entire system. Is it possible that you still have 6.5 versions of /etc/rlogind, /etc/telnetd, /usr/bin/X11/hpterm or /usr/bin/X11/xterm? An older version of these or a similar program will cause this problem too. I hope this information helps. Thanks, Rick Ferreri Hewlett-Packard Co. Ft. Collins, Co UUCP: hplabs!hpfcla!raf ARPA: raf%hpfcla@hplabs.hp.com Phone: (303) 229-6346