paws@sphinx.UChicago.UUCP (Randy Smith) (06/11/85)
I'm having problems with using the ioctl system call. I eventually
got a sample of the problem I'm having down to the following C program.
-----------------------------------------------------
#include <sys/ioctl.h>
#include <sgtty.h>
#include <stdio.h>
#include <sys/wait.h>
FILE *debug;
main()
{
long me, mepg;
int imepg;
union wait status;
debug = fopen("debug.file","w");
if (!fork()) {
me = getpid();
imepg = (int) mepg = getpgrp((int)me);
fprintf(debug,"Me: %d, Pg: %d\n",me,mepg);
fflush(debug);
fprintf(debug,"Return from setpgrp: %d\n",setpgrp(me,me));
fflush(debug);
fprintf(debug,"Return from ioctl: %d\n",ioctl(1,TIOCSPGRP,&imepg));
fflush(debug);
fprintf(debug,"I made it (thought you'd like to know)\n");
fflush(debug);
exit(0);
}
fprintf(debug,"Pid from wait: %d\n",wait(&status));
fprintf(debug,"Status code returned: %d\n",status);
}
------------------------------------------------------
(forgive the sloppy last printf).
When I copile and run this code, it hangs in the ioctl. My debug
file shows up (after I ^C) looking like:
--------------------------------------------------------
Me: 5567, Pg: 5566
Return from setpgrp: 0
--------------------------------------
If I do a ps (from another terminal) while the thing is hanging, I get:
F UID PID PPID CP PRI NI ADDR SZ RSS WCHAN STAT TTY TIME COMMAND
8201 134 4098 1 0 15 0 ff8b8800 58 14 0 I i05 0:13 -tcsh
408021 134 5217 4098 0 25 0 ff8c4800 202 8 T i05 0:04 emacs
8001 134 5354 4098 0 5 0 ff918000 28 6 b698c I i05 0:00 testio
8001 134 5355 5354 0 4 0 ff954000 36 6 T i05 0:00 testio
8201 134 5373 1 0 15 0 ff8ed800 56 50 0 S i13 0:11 -tcsh
8001 134 5462 5373 34 33 0 ff926800 70 38 R i13 0:00 ps glx
8001 134 5463 5373 4 1 0 ff93c800 30 10 3f944 S i13 0:00 tee ex
where testio is the name of the program, and tee and emacs should be ignored.
Why is this call hanging on me? I'm on a Pyramid 90x, but I got the same
(ie. confusing enough that I couldn't tell the difference) results on a
VAX 750. Suggestions from the net?
-- Randy Smith
P.S. If the parent process doesn't wait for the child, the child EXITS
during the ioctl.
UUcp: . . . ihnp4!gargoyle!sphinx!paws
Arpa: dvlp.dilvish%uchicago.mailnet@mit-multics.arpa
CSnet: randy@uchicago.csnetpaws@sphinx.UChicago.UUCP (06/12/85)
Not a day after I posted my message I figured out what was going on.
(Oops) The ioctl is done when the process is no longer in the process
group of the terminal (after the setpgrp) and unix apparently doesn't
like ioctl's preformed from such processes. Switching the order of the
ioctl and the setpgrp got rid of the hang.
-- Randy Smith
--
UUcp: . . . ihnp4!gargoyle!sphinx!paws
Arpa: dvlp.dilvish%uchicago.mailnet@mit-multics.arpa
CSnet: randy@uchicago.csnetguy@sun.uucp (Guy Harris) (06/13/85)
> I'm having problems with using the ioctl system call.
The problem is not with the "ioctl" system call but with the specific
"ioctl" call you're using, namely TIOCSPGRP. If a process attempts to do a
TRIOCSPGRP on its control terminal and it isn't in the same process group as
the terminal, the process gets a SIGTTOU signal and stops (which is why the
state field of the "ps" reports T).
Moral: do the TIOCSPGRP first and then do the "setpgrp" on yourself second.
Guy Harrischris@umcp-cs.UUCP (Chris Torek) (06/16/85)
> Moral: do the TIOCSPGRP first and then do the "setpgrp" on yourself second.
And make darned sure you've held SIGINT and the like!
--
In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 4251)
UUCP: seismo!umcp-cs!chris
CSNet: chris@umcp-cs ARPA: chris@maryland