locking/rwsem: Fix down_write_killable() for CONFIG_RWSEM_GENERIC_SPINLOCK=y
authorNiklas Cassel <niklas.cassel@axis.com>
Sat, 25 Feb 2017 00:17:53 +0000 (01:17 +0100)
committerIngo Molnar <mingo@kernel.org>
Thu, 16 Mar 2017 08:28:30 +0000 (09:28 +0100)
We hang if SIGKILL has been sent, but the task is stuck in down_read()
(after do_exit()), even though no task is doing down_write() on the
rwsem in question:

  INFO: task libupnp:21868 blocked for more than 120 seconds.
  libupnp         D    0 21868      1 0x08100008
  ...
  Call Trace:
  __schedule()
  schedule()
  __down_read()
  do_exit()
  do_group_exit()
  __wake_up_parent()

This bug has already been fixed for CONFIG_RWSEM_XCHGADD_ALGORITHM=y in
the following commit:

 04cafed7fc19 ("locking/rwsem: Fix down_write_killable()")

... however, this bug also exists for CONFIG_RWSEM_GENERIC_SPINLOCK=y.

Signed-off-by: Niklas Cassel <niklas.cassel@axis.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: <mhocko@suse.com>
Cc: <stable@vger.kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Niklas Cassel <niklass@axis.com>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: d47996082f52 ("locking/rwsem: Introduce basis for down_write_killable()")
Link: http://lkml.kernel.org/r/1487981873-12649-1-git-send-email-niklass@axis.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
kernel/locking/rwsem-spinlock.c

index 7bc24d477805d868b932aab7acc6997120931fc5..c65f7989f850d12508045896a2cb98d5b691c068 100644 (file)
@@ -213,10 +213,9 @@ int __sched __down_write_common(struct rw_semaphore *sem, int state)
                 */
                if (sem->count == 0)
                        break;
-               if (signal_pending_state(state, current)) {
-                       ret = -EINTR;
-                       goto out;
-               }
+               if (signal_pending_state(state, current))
+                       goto out_nolock;
+
                set_current_state(state);
                raw_spin_unlock_irqrestore(&sem->wait_lock, flags);
                schedule();
@@ -224,12 +223,19 @@ int __sched __down_write_common(struct rw_semaphore *sem, int state)
        }
        /* got the lock */
        sem->count = -1;
-out:
        list_del(&waiter.list);
 
        raw_spin_unlock_irqrestore(&sem->wait_lock, flags);
 
        return ret;
+
+out_nolock:
+       list_del(&waiter.list);
+       if (!list_empty(&sem->wait_list))
+               __rwsem_do_wake(sem, 1);
+       raw_spin_unlock_irqrestore(&sem->wait_lock, flags);
+
+       return -EINTR;
 }
 
 void __sched __down_write(struct rw_semaphore *sem)