On 5/27/21 10:48 PM, Noah Goldstein wrote:
> On Mon, May 24, 2021 at 7:51 PM Pavel Begunkov <asml.silence@xxxxxxxxx> wrote:
>>
>> If for/while is simple enough it's more prefered over goto with labels
>> as the former one is more explicit and easier to read. So, replace a
>> trivial goto-based loop in io_wqe_worker() with a while.
>>
>> Signed-off-by: Pavel Begunkov <asml.silence@xxxxxxxxx>
>> ---
>> fs/io-wq.c | 9 +++++----
>> 1 file changed, 5 insertions(+), 4 deletions(-)
>>
>> diff --git a/fs/io-wq.c b/fs/io-wq.c
>> index a0e43d1b94af..712eb062f822 100644
>> --- a/fs/io-wq.c
>> +++ b/fs/io-wq.c
>> @@ -538,12 +538,13 @@ static int io_wqe_worker(void *data)
>> long ret;
>>
>> set_current_state(TASK_INTERRUPTIBLE);
>> -loop:
>> - raw_spin_lock_irq(&wqe->lock);
>> - if (io_wqe_run_queue(wqe)) {
>> + while (1) {
>> + raw_spin_lock_irq(&wqe->lock);
> Can acquiring the spinlock be hoisted from the loop?
lock();
while(1) { ... }
unlock();
If this, then no. If taken inside io_worker_handle_work(),
maybe, but not with this commit.
>> + if (!io_wqe_run_queue(wqe))
>> + break;
>> io_worker_handle_work(worker);
>> - goto loop;
>> }
>> +
>> __io_worker_idle(wqe, worker);
>> raw_spin_unlock_irq(&wqe->lock);
>> if (io_flush_signals())
>> --
>> 2.31.1
>>
--
Pavel Begunkov
