[PATCH 04/10] driver core: Don't return EPROBE_DEFER to userspace during sysfs bind

[Jason Gunthorpe]

On Mon, Jun 14, 2021 at 05:08:40PM +0200, Christoph Hellwig wrote:

> @@ -679,8 +666,6 @@ static int really_probe(struct device *dev, struct device_driver *drv)
>  		dev->pm_domain->dismiss(dev);
>  	pm_runtime_reinit(dev);
>  	dev_pm_set_driver_flags(dev, 0);
> -	if (probe_ret == -EPROBE_DEFER)
> -		driver_deferred_probe_add_trigger(dev, local_trigger_count);
>  done:

I like the new arrangement - however I'm looking at the ordering
relative to this:

>  	atomic_dec(&probe_count);
>  	wake_up_all(&probe_waitqueue);

And wondering if the idea is that driver_deferred_probe_add_trigger()
is supposed to be enclosed by the atomic, so that the
device_block_probing() / wait_for_device_probe() sequence is actually
a fence against queuing new work?

Which is suggesting that the other driver_deferred_probe_add_trigger()
at the top of really_probe is already ordered wrong?

Although, if that is the idea the wait_for_device_probe() doesn't look
entirely sequenced right..

It looks easy enough to fix by moving the probe_count up:

> +static int driver_probe_device(struct device_driver *drv, struct device *dev)
> +{
> +	int trigger_count = atomic_read(&deferred_trigger_count);
> +	int ret;
> +
> +	ret = __driver_probe_device(drv, dev);
> +	if (ret == -EPROBE_DEFER || ret == EPROBE_DEFER) {
> +		driver_deferred_probe_add(dev);
> +
> +		/*
> +		 * Did a trigger occur while probing? Need to re-trigger if yes
> +		 */
> +		if (trigger_count != atomic_read(&deferred_trigger_count) &&
> +		    !defer_all_probes)
> +			driver_deferred_probe_trigger();
> +	}

into here?

I didn't see a reason why it couldn't enclose the pm stuff too..

Jason