locking/percpu-rwsem: Fix a task_struct refcount
authorQian Cai <cai@lca.pw>
Mon, 30 Mar 2020 21:30:02 +0000 (17:30 -0400)
committerIngo Molnar <mingo@kernel.org>
Wed, 8 Apr 2020 10:05:06 +0000 (12:05 +0200)
The following commit:

  7f26482a872c ("locking/percpu-rwsem: Remove the embedded rwsem")

introduced task_struct memory leaks due to messing up the task_struct
refcount.

At the beginning of percpu_rwsem_wake_function(), it calls get_task_struct(),
but if the trylock failed, it will remain in the waitqueue. However, it
will run percpu_rwsem_wake_function() again with get_task_struct() to
increase the refcount but then only call put_task_struct() once the trylock
succeeded.

Fix it by adjusting percpu_rwsem_wake_function() a bit to guard against
when percpu_rwsem_wait() observing !private, terminating the wait and
doing a quick exit() while percpu_rwsem_wake_function() then doing
wake_up_process(p) as a use-after-free.

Fixes: 7f26482a872c ("locking/percpu-rwsem: Remove the embedded rwsem")
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Qian Cai <cai@lca.pw>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lkml.kernel.org/r/20200330213002.2374-1-cai@lca.pw
kernel/locking/percpu-rwsem.c

index a008a1b..8bbafe3 100644 (file)
@@ -118,14 +118,15 @@ static int percpu_rwsem_wake_function(struct wait_queue_entry *wq_entry,
                                      unsigned int mode, int wake_flags,
                                      void *key)
 {
-       struct task_struct *p = get_task_struct(wq_entry->private);
        bool reader = wq_entry->flags & WQ_FLAG_CUSTOM;
        struct percpu_rw_semaphore *sem = key;
+       struct task_struct *p;
 
        /* concurrent against percpu_down_write(), can get stolen */
        if (!__percpu_rwsem_trylock(sem, reader))
                return 1;
 
+       p = get_task_struct(wq_entry->private);
        list_del_init(&wq_entry->entry);
        smp_store_release(&wq_entry->private, NULL);