posix-timers: fix RLIMIT_CPU && setitimer(CPUCLOCK_PROF)
authorOleg Nesterov <oleg@redhat.com>
Fri, 27 Mar 2009 00:06:10 +0000 (01:06 +0100)
committerIngo Molnar <mingo@elte.hu>
Wed, 8 Apr 2009 15:51:39 +0000 (17:51 +0200)
update_rlimit_cpu() tries to optimize out set_process_cpu_timer() in case
when we already have CPUCLOCK_PROF timer which should expire first. But it
uses cputime_lt() instead of cputime_gt().

Test case:

int main(void)
{
struct itimerval it = {
.it_value = { .tv_sec = 1000 },
};

assert(!setitimer(ITIMER_PROF, &it, NULL));

struct rlimit rl = {
.rlim_cur = 1,
.rlim_max = 1,
};

assert(!setrlimit(RLIMIT_CPU, &rl));

for (;;)
;

return 0;
}

Without this patch, the task is not killed as RLIMIT_CPU demands.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Peter Lojkin <ia6432@inbox.ru>
Cc: Roland McGrath <roland@redhat.com>
Cc: stable@kernel.org
LKML-Reference: <20090327000610.GA10108@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
kernel/posix-cpu-timers.c

index 8e5d9a68b0222f0c028c42c059a39a736c636dba..bb53185d8c786800a185ae8c7192269e22ecda52 100644 (file)
@@ -18,7 +18,7 @@ void update_rlimit_cpu(unsigned long rlim_new)
 
        cputime = secs_to_cputime(rlim_new);
        if (cputime_eq(current->signal->it_prof_expires, cputime_zero) ||
-           cputime_lt(current->signal->it_prof_expires, cputime)) {
+           cputime_gt(current->signal->it_prof_expires, cputime)) {
                spin_lock_irq(&current->sighand->siglock);
                set_process_cpu_timer(current, CPUCLOCK_PROF, &cputime, NULL);
                spin_unlock_irq(&current->sighand->siglock);