On Fri, Dec 03, 2021 at 04:23:00PM -0800, Eric Biggers wrote:
> @@ -1680,20 +1690,24 @@ static int aio_poll_wake(struct wait_queue_entry *wait, unsigned mode, int sync,
> if (mask && !(mask & req->events))
> return 0;
>
> - list_del_init(&req->wait.entry);
> -
> + /*
> + * Complete the iocb inline if possible. This requires that two
> + * conditions be met:
> + * 1. The event mask must have been passed. If a regular wakeup was
> + * done instead, then mask == 0 and we have to call vfs_poll() to
> + * get the events, so inline completion isn't possible.
> + * 2. ctx_lock must not be busy. We have to use trylock because we
> + * already hold the waitqueue lock, so this inverts the normal
> + * locking order. Use irqsave/irqrestore because not all
> + * filesystems (e.g. fuse) call this function with IRQs disabled,
> + * yet IRQs have to be disabled before ctx_lock is obtained.
> + */
> if (mask && spin_trylock_irqsave(&iocb->ki_ctx->ctx_lock, flags)) {
> struct kioctx *ctx = iocb->ki_ctx;
>
> - /*
> - * Try to complete the iocb inline if we can. Use
> - * irqsave/irqrestore because not all filesystems (e.g. fuse)
> - * call this function with IRQs disabled and because IRQs
> - * have to be disabled before ctx_lock is obtained.
> - */
> + list_del_init(&req->wait.entry);
> list_del(&iocb->ki_list);
> iocb->ki_res.res = mangle_poll(mask);
> - req->done = true;
> if (iocb->ki_eventfd && eventfd_signal_allowed()) {
> iocb = NULL;
> INIT_WORK(&req->work, aio_poll_put_work);
> @@ -1703,7 +1717,16 @@ static int aio_poll_wake(struct wait_queue_entry *wait, unsigned mode, int sync,
> if (iocb)
> iocb_put(iocb);
> } else {
I think I missed something here. Now that the request is left on the waitqueue,
there needs to be a third condition for completing the iocb inline: the
completion work must not have already been scheduled.
- Eric
