Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!cs.utexas.edu!chinacat!uudell!pensoft!robin From: robin@pensoft.UUCP (Robin Wilson ) Newsgroups: comp.unix.aix Subject: Re: Sysadmin help needed ! Message-ID: <2664@pensoft.UUCP> Date: 8 Oct 90 02:22:11 GMT References: <1990Sep27.030149.13622@portia.Stanford.EDU> Reply-To: robin@pensoft.UUCP (Robin Wilson (Guest Account)) Organization: IBM NSD Level 2 AIX V.3 Support Lines: 84 In article <1990Sep27.030149.13622@portia.Stanford.EDU> glowell@portia.Stanford.EDU (gary lowell) writes: >1. Configuring serial ports. > The problem here is likely to be with the Telebit setup. Given the 2 suggested methods of setting up a tty for bi-directional use, there are a number of possible problems that can arise from mis-understanding what transpires in the login/terminal use attempt. 1) PDELAY (or just delay) -- can be used for bi-directional serial port usage (or any port for that matter... including /dev/hft/0 -- the console). In this mode, the port starts a getty process on the tty and tells getty to 'wait for a character to arrive on the input buffer'. When getty starts DTR is raised, and the modem should be able to go into 'autoanswer' mode. However, if the modem is set with DTR strapped high (AT&Dx -- hayes command set, or one of the 'S' registers on the Telebit enhanced command set) (forgive me if I am not precise about some of this, I am not at work right now where I have all of the docs in front of me for the TB+) it will already be in autoanswer mode, so having our getty raise DTR will accomplish nothing (assuming S0<=001). Also, if the 'incoming call response' is set to echo call activity to the DTE (in this case the RS/6000) (hayes command set 'Q' register -- TB extended command set '??' register (probably Q as well)), then incoming calls will echo "RING" everytime a call is received. This will cause getty to fire up a login process and issue a login herald. IF THE LOGIN HERALD BEGINS ECHOING TO THE MODEM BEFORE CARRIER IS ACHIVED, THE MODEM WILL GO ON HOOK because it thinks someone is attempting to issue a command locally. The solution is to set the command response to 'Q6' on the hayes command set, and the equivalent for the extended command set. 2) PSHARE (or just SHARE) -- is intended for bi-directional tty usage as well. However, the getty running on this process works a little differently than PDELAY. Getty on a shared port waits to see carrier come high on the tty it is watching (EIA standards define pin 8 of a DB-25 connector to be used for carrier). The telebit signals carrier by either turning on the fast or slow connection lights, depending on the connection speed. When getty sees carrier come high it issues a login process, and herald. If the TB has carrier strapped high (hayes command set &C0, extended command set S52=000 (I think, but you better check)) then getty will always see carrier, and immediately send a herald. You should set carrier signalling to follow true carrier (hayes - &C1, TB - ?? can't remember... just look in the quick setup guide). 3) PENABLE (ENABLED) -- is intended for all other terminal usage. getty starts up and issues a login process immediately, and the corresponding herald. You will also want to set your modem up with DTR to follow EIA standards (hayes - &D2, extended ??) and DSR to follow standards (hayes - &S1, extended ??). This will insure that the port can properly disconnect when the process running on it terminates. The tty will also probably need to have 'stty attributes at LOGIN' include 'hupcl,'. This will insure that the port disconnects after getty terminates on 3 incorrect logins. The game is different if you are using the tty for a slip connection, so I won't touch that here. see if this stuff helps any for now. > >2. User Logins. > I'm not real sure, but I believe that there is a security feature on the USERS side of SMIT that you will need to double check... I just can't think of it right now. It is best to double-check the "change" user screen on smit, and get to know each field of the user entry. Also, make sure that the users don't have a password of "*" in /etc/security/passwd. Just because "su" works doesn't mean that the account is useable as a login account (this is just a hunch, but by saying that the accounts work OK on the hft, do you mean you can "su" to them?? please don't take this the wrong way, I >3. Escape in csh and ksh > I don't know about "csh" (since I never use it), but for "ksh" to use ESC for things like command line editing, you must first define an EDITOR environment variable. This can be accomplished with: "export EDITOR=" or "set -o /usr/bin/vi". I imagine the "set -o" will also work on "csh". Robin D. Wilson.. Working for, but not officially representing to USENET IBM Level 2 Field Support.