io-wq: fix bug of creating io-wokers unconditionally
authorHao Xu <haoxu@linux.alibaba.com>
Sun, 8 Aug 2021 13:54:33 +0000 (21:54 +0800)
committerJens Axboe <axboe@kernel.dk>
Tue, 10 Aug 2021 01:59:06 +0000 (19:59 -0600)
The former patch to add check between nr_workers and max_workers has a
bug, which will cause unconditionally creating io-workers. That's
because the result of the check doesn't affect the call of
create_io_worker(), fix it by bringing in a boolean value for it.

Fixes: 21698274da5b ("io-wq: fix lack of acct->nr_workers < acct->max_workers judgement")
Signed-off-by: Hao Xu <haoxu@linux.alibaba.com>
Link: https://lore.kernel.org/r/20210808135434.68667-2-haoxu@linux.alibaba.com
[axboe: drop hunk that isn't strictly needed]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
fs/io-wq.c

index 12fc193..ac8604a 100644 (file)
@@ -282,16 +282,24 @@ static void create_worker_cb(struct callback_head *cb)
        struct io_wq *wq;
        struct io_wqe *wqe;
        struct io_wqe_acct *acct;
+       bool do_create = false;
 
        cwd = container_of(cb, struct create_worker_data, work);
        wqe = cwd->wqe;
        wq = wqe->wq;
        acct = &wqe->acct[cwd->index];
        raw_spin_lock_irq(&wqe->lock);
-       if (acct->nr_workers < acct->max_workers)
+       if (acct->nr_workers < acct->max_workers) {
                acct->nr_workers++;
+               do_create = true;
+       }
        raw_spin_unlock_irq(&wqe->lock);
-       create_io_worker(wq, cwd->wqe, cwd->index);
+       if (do_create) {
+               create_io_worker(wq, cwd->wqe, cwd->index);
+       } else {
+               atomic_dec(&acct->nr_running);
+               io_worker_ref_put(wq);
+       }
        kfree(cwd);
 }