Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/5/84; site mordor.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!gamma!epsilon!zeta!sabre!petrus!bellcore!decvax!genrad!panda!talcott!harvard!seismo!ut-sally!mordor!jdb
From: jdb@mordor.UUCP (John Bruner)
Newsgroups: net.bugs.4bsd
Subject: rlogin drops output when stop/start characters change
Message-ID: <3216@mordor.UUCP>
Date: Fri, 30-Aug-85 14:47:06 EDT
Article-I.D.: mordor.3216
Posted: Fri Aug 30 14:47:06 1985
Date-Received: Sun, 1-Sep-85 12:44:28 EDT
Distribution: net
Organization: S-1 Project, LLNL
Lines: 163

Index:	ucb/rlogin.c 4.2 Fix

Description:
	"rlogin" throws away output whenever the stop/start
	characters are changed to/from ^S/^Q.  This is most
	noticeable when using a shell which provides its
	own input line editing (e.g. "tcsh"), and often
	causes the last portion of the output from the
	previous program to be lost.

Repeat-By:
	Use "rlogin" to connect to a machine which has such
	a shell and enable that shell's line editor.  Then
	run some commands which produce quite a bit of
	output (e.g. "cat /etc/passwd").  The problem is
	timing dependent, but it will probably show up after
	a couple of attempts.

Fix:
	"rlogind", the remote daemon for "rlogin", operates
	the remote pseudo-terminal in packet mode.  Each time
	that the stop/start characters are changed to/from
	^S/^Q it receives a TIOCPKT_DOSTOP or TIOCPKT_NOSTOP
	packet from the master side of the pty.  It sends this
	as an out-of-band message to the local "rlogin".
	"rlogind" also sends an out-of-band message if it
	receives a TIOCPKT_FLUSHWRITE packet from the master
	side of the pty.  (I haven't look at the code myself,
	but I'm told that the kernel never produces this
	packet type.)

	The local "rlogin" process always flushes pending
	output when it receives an out-of-band message.
	After flushing output it examines the message to
	determine if it was TIOCPKT_[DN]OSTOP; if so, then
	it sets or resets the stop/start characters on the
	local tty.  Because it unconditionally flushes output,
	a TIOCPKT_[DN]OSTOP message causes output to be lost.

	"rlogin" cannot simply recv() the out-of-band message
	because it uses the position of this message among
	the in-band data to determine how much output must
	be flushed.  (It does this with the undocumented
	SIOCATMARK ioctl.)  In order to handle this properly,
	it must "peek" at the message first to determine
	its contents, flush output if necessary, and then
	recv() the message and take any other appropriate
	actions.

	Guy Harris at Sun suggested that I include code to
	handle TIOCKPT_FLUSHREAD messages for completeness.
	The following includes that change as well:

***************
*** 20,25
  #include <sys/types.h>
  #include <sys/socket.h>
  #include <sys/wait.h>
  
  #include <netinet/in.h>
  

--- 20,26 -----
  #include <sys/types.h>
  #include <sys/socket.h>
  #include <sys/wait.h>
+ #include <sys/file.h>
  
  #include <netinet/in.h>
  
***************
*** 362,368
  
  oob()
  {
! 	int out = 1+1, atmark;
  	char waste[BUFSIZ], mark;
  
  	ioctl(1, TIOCFLUSH, (char *)&out);

--- 363,369 -----
  
  oob()
  {
! 	int out = FWRITE, in = FREAD, atmark;
  	char waste[BUFSIZ], mark;
  
  	recv(rem, &mark, 1, MSG_OOB|MSG_PEEK);
***************
*** 365,375
  	int out = 1+1, atmark;
  	char waste[BUFSIZ], mark;
  
! 	ioctl(1, TIOCFLUSH, (char *)&out);
! 	for (;;) {
! 		if (ioctl(rem, SIOCATMARK, &atmark) < 0) {
! 			perror("ioctl");
! 			break;
  		}
  		if (atmark)
  			break;

--- 366,383 -----
  	int out = FWRITE, in = FREAD, atmark;
  	char waste[BUFSIZ], mark;
  
! 	recv(rem, &mark, 1, MSG_OOB|MSG_PEEK);
! 	if (mark & TIOCPKT_FLUSHWRITE) {
! 		ioctl(1, TIOCFLUSH, (char *)&out);
! 		for (;;) {
! 			if (ioctl(rem, SIOCATMARK, &atmark) < 0) {
! 				perror("ioctl");
! 				break;
! 			}
! 			if (atmark)
! 				break;
! 			if (read(rem, waste, sizeof (waste)) <= 0)
! 				break;
  		}
  	}
  	recv(rem, &mark, 1, MSG_OOB);
***************
*** 371,380
  			perror("ioctl");
  			break;
  		}
- 		if (atmark)
- 			break;
- 		if (read(rem, waste, sizeof (waste)) <= 0)
- 			break;
  	}
  	recv(rem, &mark, 1, MSG_OOB);
  	if (mark & TIOCPKT_NOSTOP) {

--- 379,384 -----
  			if (read(rem, waste, sizeof (waste)) <= 0)
  				break;
  		}
  	}
  	recv(rem, &mark, 1, MSG_OOB);
  	if (mark & TIOCPKT_FLUSHREAD)
***************
*** 377,382
  			break;
  	}
  	recv(rem, &mark, 1, MSG_OOB);
  	if (mark & TIOCPKT_NOSTOP) {
  		notc.t_stopc = -1;
  		notc.t_startc = -1;

--- 381,388 -----
  		}
  	}
  	recv(rem, &mark, 1, MSG_OOB);
+ 	if (mark & TIOCPKT_FLUSHREAD)
+ 		ioctl(0, TIOCFLUSH, (char *)&in);
  	if (mark & TIOCPKT_NOSTOP) {
  		notc.t_stopc = -1;
  		notc.t_startc = -1;
-- 
  John Bruner (S-1 Project, Lawrence Livermore National Laboratory)
  MILNET: jdb@mordor [jdb@s1-c.ARPA]	(415) 422-0758
  UUCP: ...!ucbvax!dual!mordor!jdb 	...!seismo!mordor!jdb