On Wed, 19 Dec 2018, Thomas Gleixner wrote:
> On 2018-12-18 10:31, Thomas Gleixner wrote:
> > On Wed, 12 Dec 2018, Peter Zijlstra wrote:
> > > On Mon, Dec 10, 2018 at 06:43:51PM +0100, Thomas Gleixner wrote:
> > > @@ -806,6 +806,8 @@ void __noreturn do_exit(long code)
> > > * task into the wait for ever nirwana as well.
> > > */
> > > tsk->flags |= PF_EXITPIDONE;
> > > + smp_mb();
> > > + wake_up_bit(&tsk->flags, 3 /* PF_EXITPIDONE */);
> >
> > Using ilog2(PF_EXITPIDONE) spares that horrible inline comment and more
> > importantly selects the right bit. 0x04 is bit 2 ....
>
> Plus wake_up_bit() and wait_on_bit() want an unsigned long, but tsk->flags is
> unsigned int....
>
> Moar staring....
Aside of that calling wake_on_bit() unconditionally can be slow if the
waitqueue in the hash bucket is not empty.
So while cooking up an alternative solution I found yet another exit race:
CPU0 CPU1
sys_futex() sys_exit()
futex_lock_pi() do_exit()
No waiters:
*uaddr == 0x00000PID;
Set waiter bit
*uaddr = 0x80000PID;
attach_to_pi_owner()
tsk = get_task(PID); exit_signals(tsk)
if (!(tsk->flags & PF_EXITING))
... tsk->flags |= PF_EXITING;
mm_release(tsk)
exit_robust_list(tsk)
Set owner died and clear PID
*uaddr = 0xC0000000;
if
(unlikely(!list_empty(&tsk->pi_state_list)))
list_add(&pi_state->list,
&tsk->pi_state_list);
I put that all on hold until Jan 7.
If somebody is really bored, here is the WIP patch series which addresses
the live lock mess: https://tglx.de/~tglx/patches.tar
Thanks,
tglx
