io-wq: kill hashed waitqueue before manager exits

If we race with shutting down the io-wq context and someone queueing
a hashed entry, then we can exit the manager with it armed. If it then
triggers after the manager has exited, we can have a use-after-free where
io_wqe_hash_wake() attempts to wake a now gone manager process.

Move the killing of the hashed write queue into the manager itself, so
that we know we've killed it before the task exits.

Fixes: e941894eae31 ("io-wq: make buffered file write hashed work map per-ctx")
Signed-off-by: Jens Axboe <axboe@kernel.dk>

diff --git a/fs/io-wq.c b/fs/io-wq.c
index d7cfe8fd282a20e0e47fb3bbf84c319eb8625da4..28868eb4cd099e91bc65eb77f6915863104f128e 100644 (file)
--- a/fs/io-wq.c
+++ b/fs/io-wq.c
@@ -726,6 +726,11 @@ static int io_wq_manager(void *data)
        if (atomic_read(&wq->worker_refs))
                wait_for_completion(&wq->worker_done);
 
+       spin_lock_irq(&wq->hash->wait.lock);
+       for_each_node(node)
+               list_del_init(&wq->wqes[node]->wait.entry);
+       spin_unlock_irq(&wq->hash->wait.lock);
+
        io_wq_cancel_pending(wq);
        complete(&wq->exited);
        do_exit(0);
@@ -1051,15 +1056,11 @@ static void io_wq_destroy(struct io_wq *wq)
        set_bit(IO_WQ_BIT_EXIT, &wq->state);
        io_wq_destroy_manager(wq);
 
-       spin_lock_irq(&wq->hash->wait.lock);
        for_each_node(node) {
                struct io_wqe *wqe = wq->wqes[node];
-
-               list_del_init(&wqe->wait.entry);
                WARN_ON_ONCE(!wq_list_empty(&wqe->work_list));
                kfree(wqe);
        }
-       spin_unlock_irq(&wq->hash->wait.lock);
        io_wq_put_hash(wq->hash);
        kfree(wq->wqes);
        kfree(wq);