[PATCH] preempt race in getppid
authorDavid Meybohm <dmeybohmlkml@bellsouth.net>
Mon, 22 Aug 2005 20:11:08 +0000 (13:11 -0700)
committerLinus Torvalds <torvalds@g5.osdl.org>
Tue, 23 Aug 2005 18:44:29 +0000 (11:44 -0700)
commit4c5640cb5f5a6fd780d99397eca028b575cb1206
tree3f5eb4b4390ea71031eb2c93261052e6698cbdc1
parent3f024c1a4bc8ef9a149879351ce8b3aa749e0c2f
[PATCH] preempt race in getppid

With CONFIG_PREEMPT && !CONFIG_SMP, it's possible for sys_getppid to
return a bogus value if the parent's task_struct gets reallocated after
current->group_leader->real_parent is read:

        asmlinkage long sys_getppid(void)
        {
                int pid;
                struct task_struct *me = current;
                struct task_struct *parent;

                parent = me->group_leader->real_parent;
RACE HERE =>    for (;;) {
                        pid = parent->tgid;
        #ifdef CONFIG_SMP
        {
                        struct task_struct *old = parent;

                        /*
                         * Make sure we read the pid before re-reading the
                         * parent pointer:
                         */
                        smp_rmb();
                        parent = me->group_leader->real_parent;
                        if (old != parent)
                                continue;
        }
        #endif
                        break;
                }
                return pid;
        }

If the process gets preempted at the indicated point, the parent process
can go ahead and call exit() and then get wait()'d on to reap its
task_struct. When the preempted process gets resumed, it will not do any
further checks of the parent pointer on !CONFIG_SMP: it will read the
bad pid and return.

So, the same algorithm used when SMP is enabled should be used when
preempt is enabled, which will recheck ->real_parent in this case.

Signed-off-by: David Meybohm <dmeybohmlkml@bellsouth.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
kernel/timer.c