u.u_limit structure on UNIXPC

[Lenny Tropiano]

In article <2725 at umn-d-ub.D.UMN.EDU> rhealey at ub.d.umn.edu (Rob Healey) writes:
|>
|>	How does one set/modify the limits in the u.u_limit array for
|>	the user under 3.5x? See /usr/include/sys/vlimit.h for details.
|>	IS there any way you can set these so your core files, CPU time,
|>	etc is actually limited? I didn't see the SIGNALS involved in
|>	the limits defined anywhere. Just VERY curious if this is just
|>	another dead end brought to you by Convergent and the mortal
|>	remains of CTIX within the UNIX PC.
|>			

Well you tweeked my curiosity, so I had to try it for myself.  I started
by researching those files, <sys/vlimit.h>, <sys/user.h>, etc..  Then
I wrote a program to get the address of my user structure, then lseek in
/dev/kmem for that address and read in struct user.   I dumped out the
values for u->u_limit[0..6] to see what they were...

	u->u_limit[LIM_NORAISE] = 0
	u->u_limit[LIM_CPU] 	= INFINITY (2147483647)
	u->u_limit[LIM_FSIZE] 	= INFINITY
	u->u_limit[LIM_DATA] 	= 2621440
	u->u_limit[LIM_STACK] 	= 524288
	u->u_limit[LIM_CORE] 	= INFINITY
	u->u_limit[LIM_MAXRSS] 	= INFINITY

For what I can see the stuff like CPU time limits, core file size limits,
etc, are BSD left-overs from the Convergent port.  Yes, as Rob pointed
out there is no signals to handle the limit overflows.  On a BSD system
there would be a signal like SIGXCPU that would be for exceeding the CPU
time limit.   The LIM_FSIZE correlates to the ulimit(2) system call,
and the ulimit built-in for /bin/sh and /bin/ksh.   I tested my program
by doing a:  ulimit 1000 and then re-running it and I saw that:

	u->u_limit[LIM_FSIZE]	= 1000

So what I did was modify the program to write the user structure back out
to kmem with some values changed.  I changed the LIM_CPU to 10 (10 cpu secs)
and LIM_CORE to 2 (two blocks).  I then wrote a simple program that just
looped forever and ever.  It accrued minutes of cpu time, and didn't stop.
I also created a program that malloc'd as much memory as it could (sbrk(2))
and then I did a "kill -4 pid", which created a BIG core file!  Definately
not two blocks ...

Now what's interesting, and I haven't really tested anything with this
yet, is the values that aren't INFINITY;  LIM_DATA, LIM_STACK.

LIM_DATA is at a two megabyte limit (which sound correct), and
LIM_STACK is at a .5MB limit.   Since it's late already tonight, and I'm
dead tired, I'll play with this more tomorrow.  But I have a feeling
that the u_limit[] array in the user block isn't used for anything but
the file size limit.   For more information on BSD only stuff, look in
the manual under ksh(1), and the built-in ulimit.  I did, BTW try
doing a ulimit -c ##, and it said bad options(s) -- as suspected!

Hope that helps...   It would be nice if those BSDisms were there! 

-Lenny
-- 
Lenny Tropiano             ICUS Software Systems         [w] +1 (516) 589-7930
lenny at icus.islp.ny.us      Telex; 154232428 ICUS         [h] +1 (516) 968-8576
{ames,pacbell,decuac,hombre,talcott,sbcs}!icus!lenny     attmail!icus!lenny
        ICUS Software Systems -- PO Box 1; Islip Terrace, NY  11752