com6@uhnix1.uh.edu (J. Eric Townsend) (11/15/89)
Hi. We have an RT that is giving (and has always) problems with telneting into it. It says that all the network devices are in use, and closes telnet. Telneting *out* of the system is just hunky-dory. Telnet is refused even when noone is logged into the machine... :-( Any hints/suggestions? -- Skate UNIX. J. Eric Townsend com6@uhnix1.uh.edu | jet@karazm.math.uh.edU EastEnders Mailing list: eastender@flatline.UUCP
dyer@spdcc.COM (Steve Dyer) (11/16/89)
In article <16994@uhnix1.uh.edu> com6@uhnix1.uh.edu (J. Eric Townsend) writes: >Hi. We have an RT that is giving (and has always) problems with telneting >into it. It says that all the network devices are in use, and >closes telnet. Telneting *out* of the system is just hunky-dory. >Telnet is refused even when noone is logged into the machine... :-( Make more pty's by invoking the "devices" command. It sounds like you don't have any now. -- Steve Dyer dyer@ursa-major.spdcc.com aka {ima,harvard,rayssd,linus,m2c}!spdcc!dyer dyer@arktouros.mit.edu, dyer@hstbme.mit.edu
karish@forel.stanford.edu (Chuck Karish) (11/16/89)
In article <605@ursa-major.SPDCC.COM> dyer@ursa-major.spdcc.COM (Steve Dyer) wrote: |In article <16994@uhnix1.uh.edu> com6@uhnix1.uh.edu (J. Eric Townsend) writes: ||Hi. We have an RT that is giving (and has always) problems with telneting ||into it. It says that all the network devices are in use, and ||closes telnet. Telneting *out* of the system is just hunky-dory. ||Telnet is refused even when noone is logged into the machine... :-( | |Make more pty's by invoking the "devices" command. It sounds like you |don't have any now. Or else the ptys are present, but none are enabled. Use `devices' to turn the `ae' (auto-enable) option on for some of the ptys, so getty will listen for logins on them. Some un-enabled ptys should also be present, for outgoing telnet and ftp sessions. Chuck Karish karish@mindcraft.com (415) 323-9000 karish@forel.stanford.edu
dyer@spdcc.COM (Steve Dyer) (11/16/89)
In article <6680@portia.Stanford.EDU> karish@forel.stanford.edu (Chuck Karish) writes: >Or else the ptys are present, but none are enabled. Use `devices' to >turn the `ae' (auto-enable) option on for some of the ptys, so getty >will listen for logins on them. Some un-enabled ptys should also be >present, for outgoing telnet and ftp sessions. Are you telling me that AIX/RT doesn't use inetd but instead requires getty to be spawned and waiting on each pty?? Gad. At least AIX PS/2 gets THAT right. Why would telnet (or for that matter, ftp) require a pty on outgoing sessions? Is this just a speculation of yours or are you relating the awful truth? -- Steve Dyer dyer@ursa-major.spdcc.com aka {ima,harvard,rayssd,linus,m2c}!spdcc!dyer dyer@arktouros.mit.edu, dyer@hstbme.mit.edu
karish@forel.stanford.edu (Chuck Karish) (11/16/89)
In article <610@ursa-major.SPDCC.COM> dyer@ursa-major.spdcc.COM (Steve Dyer) wrote: >In article <6680@portia.Stanford.EDU> karish@forel.stanford.edu (Chuck Karish) writes: >>Or else the ptys are present, but none are enabled. Use `devices' to >>turn the `ae' (auto-enable) option on for some of the ptys, so getty >>will listen for logins on them. Some un-enabled ptys should also be >>present, for outgoing telnet and ftp sessions. > >Are you telling me that AIX/RT doesn't use inetd but instead requires >getty to be spawned and waiting on each pty?? Gad. At least AIX PS/2 >gets THAT right. I don't know precisely how this all works. AIX/RT doesn't run a getty for each pty. It seems to keep a pool of gettys available, and inetd associates them with ptys as requested. The documentation for setting up TCP/IP told me to make some ptys with ae set and some without, so I did it and it worked. It said that it used separate ptys for the different streams FTP uses (2.1.1 doc). AIX PS/2 also has two types of ptys, one with logins enabled and one not. It seems to set up ptys the same way the RT documentation told me to. The difference is that more of the configuration is pre-set in the configuration files used by `devices'. Chuck Karish karish@mindcraft.com (415) 323-9000 karish@forel.stanford.edu
silver@s.cs.uiuc.edu (11/19/89)
Re: needing pts's enabled to telnet INTO an RT: We have pts0-3 autoenabled, and #'s 4-8 NOT autoenabled and starting 2 days ago we could have 8 telnets logged in at once (Okay, I only did 6, but close enough, huh?). The thing we changed 2 days ago was the number of processes able to run on the machine at a time. In /etc/master the limit was set at 60! Ugh! When we boot, all the daemons, etc. total ~40 processes before anyone is logged in! Noticing the warning limit is 350, I upped the total to 150 (after that I expect the machine to be VERY slow) and left the per-user process limit at 100. Is 'procs = 60' really the default for an RT? Perhaps this will help you use more pts's. May all you programs compile on the first attempt. Edward A. Silverstein silver@director.beckman.uiuc.edu
njs@scifi.UUCP (Nicholas J. Simicich) (11/20/89)
In article <610@ursa-major.SPDCC.COM> dyer@ursa-major.spdcc.COM (Steve Dyer) writes: >Are you telling me that AIX/RT doesn't use inetd but instead requires >getty to be spawned and waiting on each pty?? Gad. At least AIX PS/2 >gets THAT right. AIX/RT does use inetd. I didn't think that inetd had anything to do with running getty on these ports. Instead, I thought that was a function of telnetd. Essentially, what telnetd does step through the ptys using an ioctl to determine when there is a ptc which is being waited for by open in getty. When it finds one, it opens the pts side. An interesting side effect of this is that a user program can get a login session by doing the same thing. >Why would telnet (or for that matter, ftp) require a pty on outgoing >sessions? Is this just a speculation of yours or are you relating >the awful truth? As far as I know, they do not. -- Nick Simicich --- uunet!bywater!scifi!njs --- njs@ibm.com (Internet)
njs@scifi.UUCP (Nicholas J. Simicich) (11/20/89)
In article <6706@portia.Stanford.EDU> karish@forel.stanford.edu (Chuck Karish) writes: >I don't know precisely how this all works. AIX/RT doesn't run a getty >for each pty. It seems to keep a pool of gettys available, and inetd >associates them with ptys as requested. Not as far as I know. The gettys are actually started on the ptys which have ae enabled. I've written a program which depends on this behavior, and it seems to work. There is a pool of ptys, not a pool of gettys. >The documentation for >setting up TCP/IP told me to make some ptys with ae set and some >without, so I did it and it worked. It said that it used separate >ptys for the different streams FTP uses (2.1.1 doc). I somehow don't think that this is actually true. You can determine exactly what processes have a device open with for a in `fuser /dev/ptc0`;do ps -p $a;done for a in `fuser /dev/pts0`;do ps -p $a;done Thus, it would be possible to do a ftp and cycle through all of your ptys, and determine if ftp was actually using any of them. -- Nick Simicich --- uunet!bywater!scifi!njs --- njs@ibm.com (Internet)
gobran@ecs.umass.edu (11/22/89)
In article <213600008@s.cs.uiuc.edu>, silver@s.cs.uiuc.edu writes: > Re: needing pts's enabled to telnet INTO an RT: > > We have pts0-3 autoenabled, and #'s 4-8 NOT autoenabled and starting 2 > days ago we could have 8 telnets logged in at once (Okay, I only did 6, > but close enough, huh?). > > The thing we changed 2 days ago was the number of processes able to run > on the machine at a time. In /etc/master the limit was set at 60! Ugh! > > Is 'procs = 60' really the default for an RT? > > Edward A. Silverstein > silver@director.beckman.uiuc.edu If I remember correctly the X-windows manual for AIX/RT tells you to increase the values of 'procs' and 'kprocs' (and maybe some others) in /etc/master when you install X since you will be using a nimber of pty's with X. So I would assume that these changes may also be helpful when you have a large number of telnet sessions on a system. Dave Gobran UMass/Amherst gobran@ecs.umass.edu gobran@umaecs