|
|
|
|
|
|
|
|
|
|
|
|
| home | help | back | first | fref | pref | prev | next | nref | lref | last | post |
Date: Wed, 12 Jul 1995 18:18:28 -0400 From: Sam Hartman <hartmans@MIT.EDU> To: krbdev@MIT.EDU The purpose of this note is to seek comments on the external interfaces for libpty. Libpty is a library that will replace the PTY and UTMP handling for both rlogind and telnetd. We have decided not to maintain two different sets of PTY and UTMP handling code. Neither rlogind nor telnetd work perfectly, and they have different bugs; it would be easier to get things working if we were only faced with one implementation. Most of the code for libpty will come from Richard's changes to the rlogind pty handling. I've looked at several pty implementations including Telnet, Expect, a program called Pty, etc; the code in Rlogind is the clearest and takes the most advantage of Autoconf. I will be bringing in a few security fixes and bug fixes I've picked up from other packages. Unfortunately, the Rlogindcode doesn't support quite as much as Telnetd. In particular, it doesn't have the hooks for secure Unicos (Cray). The way this code is done for Telnetd is rather poor; there is a rather excessive amount of #ifdef CRAY, and the comments aren't sufficient to figure out what's going on without access to Unicos man pages. In other words, we will probably break Unicos enhanced security support. I've also already broken Convex because it was poorly done, broke seveural abstraction layers, and didn't leave enough information for me to understand how to do it right. In both these cases, I only remove as much code as necessary to implement what I'm implementing cleanly; it should be fairly easy for someone familiar with the operating systems in question to cleanly implement the code. Anyway, I propose the following as interfaces to be used by applications: (long is returned because that's what com_err takes; I don't want to use krb5_error_code as I don't want to depend on krb5.h.) long pty_getpty ( int *fd, char *slave) Find and initialize a clean master pty. This should open the pty as fd, and return the name of the slave. All processes should be cleared from the slave. The mode of the slave should be 0600 writable only by root. It should return 0 or an error code. long pty_open_slave (/*in */ char * slave, /* out*/ int *fd) Initialize the slave side by dissociating the current terminal and by setting process groups, etc. In addition, it will initialize the terminal flags (termios or old BSD) appropriately; the application may have to do additional customization, but this should sanitize things. long pty_update_utmp(struct utmp *ut, char *user, char *line, char *host) Update the utmp information or return an error. long pty_update_wtmpstruct utmp *ent) Dito. Also, in krlogind's wtmp code is logwtmp(). Apparently, there is a test to see if the system already has the function. I can't find any platforms that actually do. I would appreciate any comments on where the function originally comes from; I'm not sure if it should be pty_logwtmp, or logwtmp, and if pty_logwtmp if pty_logwtmp should just call logwtmp if defined, etc, etc. long pty_cleanup(char *slave, pid_t pid) Clean up after the slave application has exited. Close down the pty, HUPing processes associated with it. (pid is the pid of the slave process that may have died, slave is the name of the slave terminal.) Thanks for your comments, --Sam
|
|
|
|
|
|
|
|
|
|
|
|
| home | help | back | first | fref | pref | prev | next | nref | lref | last | post |