[ldv-project] [PATCH] usb/core/devio.c: Don't use GFP_KERNEL while we cannot reset a storage device

[Alan Stern]

On Tue, 12 Mar 2013, Ming Lei wrote:

> > In general it isn't, no.  But usbfs uses the lock to prevent races with
> > driver_disconnect() -- it is invalid to submit URBs after the
> > disconnect routine has returned.
> 
> If so, we may introduce another lock to avoid the race.
> 
> So I think we may figure out to decrease the device lock
> scope in devio.c, and most of ioctl commands might not require it
> at all, then the problem addressed by the patch can be fixed too.

That might work.  A mutex could be added to the dev_state structure.  
The mutex would be locked whenever an URB was being submitted and
during driver_disconnect, and perhaps a few other places too, but not
when memory was being allocated.  The device itself would remain 
unlocked most of the time.

Alan Stern