Re: "usb_wwan: error case of resume" (16871dcac) is buggy

[Oliver Neukum <oneukum@suse.de>]

Am Montag, 21. März 2011, 01:36:45 schrieb Jiri Kosina:
> Hi,
> 
> the commit in subject make the kernel with CONFIG_PM_RUNTIME unset fail 
> during compilation, as struct dev_pm_info doesn't have whole bunch of 
> members in such case.
> 
> The commit in question adds this code:
> 
> 	/* we have to throw away the rest */
> 	do {
> 		unbusy_queued_urb(urb, portdata);
> 		//extremely dirty
> 		atomic_dec(&port->serial->interface->dev.power.usage_count);
> 	} while ((urb = usb_get_from_anchor(&portdata->delayed)));
> 
> The 'extermely dirty' comment makes me a bit nervous whether the patch 
> below is correct or some more thinking would be necessary.

I've since posted a clean patch, although your fix looks correct to me.
The problem was that I was unsure about the locking during resume().
I rather wrote a layering violation than risk a deadlock in an error path
hard to test. But I did not feel well doing it.

	Regards
		Oliver