linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Matt Chapman <matthewc@cse.unsw.edu.au>
To: linux-kernel@vger.kernel.org
Subject: SMP race: child gets SIGSTOP as parent exits
Date: Tue, 11 Sep 2001 17:19:08 +1000	[thread overview]
Message-ID: <20010911171908.A26199@cse.unsw.edu.au> (raw)

This was tested on 2.4.7 but I can't see any relevant changes in
the code or the changelog.  In arch/i386/kernel/signal.c:

                if (ka->sa.sa_handler == SIG_DFL) {
...
                        case SIGSTOP:
                                current->state = TASK_STOPPED;
                                current->exit_code = signr;
                                if (!(current->p_pptr->sig->action[SIGCHLD-1].sa
.sa_flags & SA_NOCLDSTOP))
                                        notify_parent(current, SIGCHLD);

The problem is that if the parent is in the process of exiting,
do_exit calls exit_sighand which sets tsk->sig = NULL _before_
the children get reparented in exit_notify.  So the above code
oopses dereferencing current->p_pptr->sig.

This simple program demonstrates the bug (when run in a while
true loop for a second or two):

#include <stdio.h>
#include <stdlib.h>
#include <signal.h>

int main()
{
        switch (fork())
        {
                case -1:
                        perror("fork");
                        exit(1);

                case 0: /* child */
                        kill(getpid(), SIGSTOP);
                        break;

                default: /* parent */
                        exit(0);
        }
}

The obvious (not necessarily best) fix is to check that
current->p_ptr->sig is not NULL.

Cheers,
Matt


diff -ru linux-2.4.10-pre8/arch/i386/kernel/signal.c linux-2.4.10-pre8-fix/arch/i386/kernel/signal.c
--- linux-2.4.10-pre8/arch/i386/kernel/signal.c	Tue Sep 11 17:07:30 2001
+++ linux-2.4.10-pre8-fix/arch/i386/kernel/signal.c	Tue Sep 11 17:12:27 2001
@@ -651,6 +651,7 @@
 		}
 
 		if (ka->sa.sa_handler == SIG_DFL) {
+			struct signal_struct *p_sig;
 			int exit_code = signr;
 
 			/* Init gets no signals it doesn't want.  */
@@ -669,7 +670,8 @@
 			case SIGSTOP:
 				current->state = TASK_STOPPED;
 				current->exit_code = signr;
-				if (!(current->p_pptr->sig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDSTOP))
+				p_sig = current->p_pptr->sig;
+				if ((p_sig != NULL) && !(p_sig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDSTOP))
 					notify_parent(current, SIGCHLD);
 				schedule();
 				continue;

                 reply	other threads:[~2001-09-11  7:19 UTC|newest]

Thread overview: [no followups] expand[flat|nested]  mbox.gz  Atom feed

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20010911171908.A26199@cse.unsw.edu.au \
    --to=matthewc@cse.unsw.edu.au \
    --cc=linux-kernel@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).