futex: Fix a race condition between REQUEUE_PI and task death
authorBrian Silverman <bsilver16384@gmail.com>
Sun, 26 Oct 2014 00:20:37 +0000 (20:20 -0400)
committerThomas Gleixner <tglx@linutronix.de>
Sun, 26 Oct 2014 15:16:18 +0000 (16:16 +0100)
commit30a6b8031fe14031ab27c1fa3483cb9780e7f63c
treeda56baddfc24fef71ca5b6f123c947e784527d87
parent993b2ff221999066fcff231590593d0b98f45d32
futex: Fix a race condition between REQUEUE_PI and task death

free_pi_state and exit_pi_state_list both clean up futex_pi_state's.
exit_pi_state_list takes the hb lock first, and most callers of
free_pi_state do too. requeue_pi doesn't, which means free_pi_state
can free the pi_state out from under exit_pi_state_list. For example:

task A                            |  task B
exit_pi_state_list                |
  pi_state =                      |
      curr->pi_state_list->next   |
                                  |  futex_requeue(requeue_pi=1)
                                  |    // pi_state is the same as
                                  |    // the one in task A
                                  |    free_pi_state(pi_state)
                                  |      list_del_init(&pi_state->list)
                                  |      kfree(pi_state)
  list_del_init(&pi_state->list)  |

Move the free_pi_state calls in requeue_pi to before it drops the hb
locks which it's already holding.

[ tglx: Removed a pointless free_pi_state() call and the hb->lock held
   debugging. The latter comes via a seperate patch ]

Signed-off-by: Brian Silverman <bsilver16384@gmail.com>
Cc: austin.linux@gmail.com
Cc: darren@dvhart.com
Cc: peterz@infradead.org
Cc: stable@vger.kernel.org
Link: http://lkml.kernel.org/r/1414282837-23092-1-git-send-email-bsilver16384@gmail.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
kernel/futex.c