Fix timeout in robust path

Hi Ingo,

I hit the following BUG when exercising the robust futex path

testpi-1/4920[CPU#0]: BUG in FREE_WAITER at kernel/rt.c:1368

 [<c011f180>] __WARN_ON+0x60/0x80 (8)
 [<c03f6581>] __down_mutex+0x601/0x844 (48)
 [<c013813a>] pi_setprio+0xa1/0x632 (104)
 [<c0127386>] lock_timer_base+0x19/0x33 (8)
 [<c03f884d>] _spin_lock_irqsave+0x1d/0x46 (12)
 [<c0127386>] lock_timer_base+0x19/0x33 (8)
 [<c0127386>] lock_timer_base+0x19/0x33 (16)
 [<c01273d8>] __mod_timer+0x38/0xdf (16)
 [<c013fb9b>] sub_preempt_count+0x1a/0x1e (12)
 [<c03f81e1>] __down_interruptible+0x922/0xaf7 (20)
 [<c01411f5>] futex_wait_robust+0x14c/0x216 (16)
 [<c01394e8>] process_timeout+0x0/0x9 (48)
 [<c01411f5>] futex_wait_robust+0x14c/0x216 (64)
 [<c013d042>] down_futex+0x7d/0xe2 (12)
 [<c01411f5>] futex_wait_robust+0x14c/0x216 (12)
 [<c013d066>] down_futex+0xa1/0xe2 (8)
 [<c01411f5>] futex_wait_robust+0x14c/0x216 (12)
 [<c01411f5>] futex_wait_robust+0x14c/0x216 (24)
 [<c01419de>] do_futex+0x92/0xf8 (72)
 [<c0141b3c>] sys_futex+0xf8/0x104 (40)
 [<c0103017>] sysenter_past_esp+0x54/0x75 (60)
---------------------------

| preempt count: 00000001 ]
| 1-level deep critical section nesting:
----------------------------------------
.. [<c013fb00>] .... add_preempt_count+0x1a/0x1e
.....[<00000000>] .. ( <= stext+0x3feffd68/0x8)

------------------------------
| showing all locks held by: | (testpi-1/4920 [f6326120, 59]):
------------------------------

When calling futex_wait_robust, we need to ensure that the timeout is reset to zero, incase userspace timeout is NULL. Please consider applying

-Dinakar

Index: linux-2.6.14/kernel/futex.c
===================================================================
@@ -1535,6 +1535,8 @@
 			return -EFAULT;
 		timeout = timespec_to_jiffies(&t) + 1;
 	}
+	if (op == FUTEX_WAIT_ROBUST && utime == NULL)
+		timeout = 0;
 	/*
 	 * requeue parameter in 'utime' if op == FUTEX_REQUEUE.
 	 */

thanks - i picked your fix up via David's patch-2.6.14-rt22-rf11.

Ingo