Whats the latest on the 30 second login bug for 5.06

[Anthony Hall]

Hi,

I have seen the sco ta and also recommendations re deleting ttys-t. Anyone
had a positive result on preventing the 30 second login problem for 5.06

Anthony Hall.

[David H. Funte]

I applied the ta info, and the login delay has disappeared. It's been a
while, so I don't remember what I did..But it worked. I'm hoping that they
come out with a RS506a soon.

Anthony Hall <ha...@promedicus.com.au> wrote in message
news:FuLn6.36652$Xx3.2...@news1.eburwd1.vic.optushome.com.au...

[Darryl Krasman]

"David H. Funte" wrote:
>
> I applied the ta info, and the login delay has disappeared. It's been a
> while, so I don't remember what I did..But it worked. I'm hoping that they
> come out with a RS506a soon.

I have only seen these symptoms once - mind you I have only one 5.0.6
machine out there at the moment. I have had other problems with 5.0.6
that I have posted here but it seems that once "-s20" was removed from
the modem's getty line and the client rebooted that things have settled
down. Coincidence? Most likely - but I don't have time and no one is
paying me to pursue it further....
--
Darryl
Ideal Computer Group Inc.

[Stuart J. Browne]

> I have seen the sco ta and also recommendations re deleting ttys-t.
Anyone
> had a positive result on preventing the 30 second login problem for 5.06

SCO's TA reccomends running the 'tcbck' every roughly 5 seconds to try fix
the securities database. However, upon testing this on one machine I've had
this on, it did not fix the problem. However, removing the file
'/etc/auth/system/ttys-t' did fix the problem.

Possibly the best action is a combination of both (tcbck to check redibility
of database, then "[ ! -s /etc/auth/system/ttys-t ] && rm
/etc/auth/system/ttys-t" if the file is there but empty (thus not going to
work)).

bkx

[Bela Lubkin]

Anthony Hall wrote:

> > I have seen the sco ta and also recommendations re deleting ttys-t. Anyone
> > had a positive result on preventing the 30 second login problem for 5.06

David H. Funte replied:

> I applied the ta info, and the login delay has disappeared. It's been a
> while, so I don't remember what I did..But it worked. I'm hoping that they
> come out with a RS506a soon.

This problem is indeed fixed in rs506a, which is supposed to be
appearing on ftp.sco.com "soon". I've lost track of the exact date it's
going to be there.

The actual bug is in telnetd; it can be killed by a stray self-generated
signal at a critical moment. This bug was introduced during 5.0.6
development. One workaround (until you can get rs506a) is to use the
5.0.5 version of /etc/telnetd. I wouldn't use the workaround from the
TA if I could avoid it, it's pretty heavy-handed...

A little more detail, for the curious:

5.0.5 telnetd was OK (in fact, unchanged from 5.0.0 through rs505a).

5.0.6 telnetd included several minor fixes:

- try to clean up utmp/utmpx records even if trouble with SecureWare
files

- log session exit status to utmp[x]

- on exit, shutdown the socket before all the file manipulation
(telnet client exits much more quickly)

- fix syslog messages in "ripped-out" security mode

- fix option negotiation problem w/HP-UX telnet client

Unfortunately a bit of signal handling was missed out such that a
SIGCHLD could come in at just the wrong moment and kill telnetd between
creating and deleting ttys-t.

rs506a telnetd fixes:

- SIGCHLD race condition causing immortal ttys-t

- use sane ut_id for /dev/ttyp1000..9999 ("%d" rather than "p%d",
which got truncated to 4 chars, thus overwriting other entries)

>Bela<