lkml.org 
[lkml]   [2011]   [Dec]   [26]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [BUG] TASK_DEAD task is able to be woken up in special condition
> On 12/23, Peter Zijlstra wrote:
> >
> > On Thu, 2011-12-22 at 09:42 +0900, Yasunori Goto wrote:
> > >
> > > ----------------------------------+-----------------------------
> > > |
> > > CPU A | CPU B
> > > ----------------------------------+-----------------------------
> > > TASK A calls exit()....
> > >
> > > do_exit()
> > >
> > > exit_mm()
> > > down_read(mm->mmap_sem);
> > >
> > > rwsem_down_failed_common()
> > >
> > > set TASK_UNINTERRUPTIBLE
> > > set waiter.task <= task A
> > > list_add to sem->wait_list
> > > :
> > > raw_spin_unlock_irq()
> > > (I/O interruption occured)
> > >
> > > __rwsem_do_wake(mmap_sem)
> > >
> > > list_del(&waiter->list);
> > > waiter->task = NULL
> > > wake_up_process(task A)
> > > try_to_wake_up()
> > > (task is still
> > > TASK_UNINTERRUPTIBLE)
> > > p->on_rq is still 1.)
> > >
> > > ttwu_do_wakeup()
> > > (*A)
> > > :
> > > (I/O interruption handler finished)
> > >
> > > if (!waiter.task)
> > > schedule() is not called
> > > due to waiter.task is NULL.
> > >
> > > tsk->state = TASK_RUNNING
> > >
> > > :
> > > check_preempt_curr();
> > > :
> > > task->state = TASK_DEAD
> > > (*B)
> > > <--- set TASK_RUNNING (*C)
> > >
> > >
> > >
> > > schedule()
> > > (exit task is running again)
> > > BUG_ON() is called!
> > > --------------------------------------------------------
> >
> > <snip>
> >
> > > This is very bad senario.
> > > But, I suppose this phenomenon is able to occur on a guest system of
> > > virtual machine too.
> > >
> > > Please fix it.
> > >
> > > I suppose task->pi_lock should be held when task->state is changed to
> > > TASK_DEAD like the following patch (not tested yet).
> > > Because try_to_wake_up() hold it before checking task state.
> >
> > I don't think this can actually happen, note the raw_spin_unlock_wait()
> > in do_exit() long before setting TASK_DEAD, that should synchronize
> > against the in-progress wakeup and ensure its finished and has set
> > TASK_RUNNING.
>
> How? raw_spin_unlock_wait() is calles before exit_mm(), then the exiting
> task plays with its ->state.
>
> IIRC, this was already discussed a bit. Say, try_to_wake_up(TASK_INTERRUPTIBLE)
> can wakeup a TASK_UNINTERRUPTIBLE task if it temporary sets INTERRUPTIBLE but
> doesn't call schedule() in this state.

Oleg-san,

Could you point the discussion?
I don't understand yet how it occurred...


>
> May be ttwu_do_wakeup() should do cmpxchg(old_state, RUNNING) ?
>
> Oleg.
>

--
Yasunori Goto




\
 
 \ /
  Last update: 2011-12-26 09:27    [W:0.122 / U:0.312 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site