From owner-freebsd-hackers  Sun Jan  3 11:40:27 1999
Return-Path: <owner-freebsd-hackers@FreeBSD.ORG>
Received: (from majordom@localhost)
          by hub.freebsd.org (8.8.8/8.8.8) id LAA29285
          for freebsd-hackers-outgoing; Sun, 3 Jan 1999 11:40:27 -0800 (PST)
          (envelope-from owner-freebsd-hackers@FreeBSD.ORG)
Received: from apollo.backplane.com (apollo.backplane.com [209.157.86.2])
          by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id LAA29277
          for <freebsd-hackers@FreeBSD.ORG>; Sun, 3 Jan 1999 11:40:25 -0800 (PST)
          (envelope-from dillon@apollo.backplane.com)
Received: (from dillon@localhost)
	by apollo.backplane.com (8.9.1/8.9.1) id LAA51723;
	Sun, 3 Jan 1999 11:39:37 -0800 (PST)
	(envelope-from dillon)
Date: Sun, 3 Jan 1999 11:39:37 -0800 (PST)
From: Matthew Dillon <dillon@apollo.backplane.com>
Message-Id: <199901031939.LAA51723@apollo.backplane.com>
To: Chia-liang Kao <clkao@CirX.ORG>
Cc: freebsd-hackers@FreeBSD.ORG
Subject: Re: setjmp/longjmp corrupts stack?
Sender: owner-freebsd-hackers@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.ORG


:Hi,
:
:I have a little program attached below causing SIGSEGV.  But the
:program works out (partially, see below) dramatically if the function
:being called in main() (haha()) changes to hehe().
:
:In my trace record, the stack corrupted right after longjmp to j2.
:But if I change the haha() in main() to hehe(), although the result is
:as expected, the stack is somewhat corrupted too. Like the following:
:...

    What you are doing is totally illegal.  You cannot push down the
    stack, setjmp, return from that context, then longjmp() back down
    into it.

    The only legal way to use setjmp()/longjmp() is for the context that
    you setjmp() to remain *valid* from the point setjmp() is called to the
    point longjmp() is called to get back to it.

    In the code below you setjmp j1 in main, then call haha which 
    setjmp's j2 and longjmp's back to j1 (which is ok), but that means
    you've popped the haha() context.  Then your main function longjmp's back
    to j2, inside haha(), which is now totally illegal because that context
    was popped when you longjmp'd out of it the first time.

						-Matt


:Regards, 
:CLK
:
:======================
:#include <setjmp.h>
:#include <stdio.h>
:
:jmp_buf j1, j2;
:
:void
:haha()
:{
:	int r;
:	static int cnt;
:	/* ... */
:	printf("send\n");
:	if(!(r =setjmp(j2))) {
:		/* go back */
:		longjmp(j1, ++cnt);
:	}
:	/* resume */
:	printf("resume\n");
:	return;
:}
:
:void
:hehe()
:{
:	haha();
:}
:
:int
:main()
:{
:	int r;
:	if((r = setjmp(j1))) {
:		printf("jmp %d\n", r);
:		if(r == 1)
:			longjmp(j2, 1);
:		else
:			exit(0);
:	}
:	printf("main\n");
:	haha();
:	printf("after longjmp\n");
:	return 0;
:}
:
:To Unsubscribe: send mail to majordomo@FreeBSD.org
:with "unsubscribe freebsd-hackers" in the body of the message
:

    Matthew Dillon  Engineering, HiWay Technologies, Inc. & BEST Internet 
                    Communications & God knows what else.
    <dillon@backplane.com> (Please include original email in any response)    

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message