Re: [PATCH] usb: ulpi: don't register drivers if bus doesn't exist

[Alan Stern]

On Thu, 28 May 2015, Felipe Balbi wrote:

> On Thu, May 28, 2015 at 11:09:38AM +0530, Sudip Mukherjee wrote:
> > On Wed, May 27, 2015 at 08:21:16AM -0700, Greg KH wrote:
> > > On Wed, May 27, 2015 at 11:16:34AM -0400, Alan Stern wrote:
> > > > On Wed, 27 May 2015, Heikki Krogerus wrote:
> > > 
> > > Maybe we need to test for this in the driver core, not allowing drivers
> > > for busses that are not registered, that might solve the main problem
> > > here.  I'll try to look at it tonight.
> > may i suggest something like this ?
> > buildtest with allmodconfig and allyesconfig on x86_64.
> > built and booted on x86.
> > 
> > 
> > diff --git a/drivers/base/bus.c b/drivers/base/bus.c
> > index 5005924..95cefa0 100644
> > --- a/drivers/base/bus.c
> > +++ b/drivers/base/bus.c
> > @@ -943,6 +943,7 @@ int bus_register(struct bus_type *bus)
> >  	if (retval)
> >  		goto bus_groups_fail;
> >  
> > +	bus->registered = true;
> 
> once set, it's never cleared.

It's worse than that...

> >  	pr_debug("bus: '%s': registered\n", bus->name);
> >  	return 0;
> >  
> > diff --git a/drivers/base/driver.c b/drivers/base/driver.c
> > index 4eabfe2..1acae5b 100644
> > --- a/drivers/base/driver.c
> > +++ b/drivers/base/driver.c
> > @@ -150,6 +150,11 @@ int driver_register(struct device_driver *drv)
> >  	int ret;
> >  	struct device_driver *other;
> >  
> > +	if (!drv->bus->registered) {
> > +		pr_err("Driver %s registration failed. bus not yet registered\n",
> > +		       drv->name);
> > +		return -ENODEV;
> > +	}
> >  	BUG_ON(!drv->bus->p);
> >  
> >  	if ((drv->bus->probe && drv->probe) ||
> > diff --git a/include/linux/device.h b/include/linux/device.h
> > index 00ac57c..8fe4745 100644
> > --- a/include/linux/device.h
> > +++ b/include/linux/device.h
> > @@ -126,6 +126,7 @@ struct bus_type {
> >  	const struct dev_pm_ops *pm;
> >  
> >  	const struct iommu_ops *iommu_ops;
> > +	bool registered;	/* DON'T TOUCH THIS */
> 
> I would rather use an atomic_t

What reason is there to use an atomic_t?  The value is never going to 
be changed by two threads at the same time.

More importantly, clearing the flag races with checking it.  If 
somebody tries to register a driver at the same time as the bus is 
unregistered, the result is undefined.

Of course, the same problem exists when a device is added to a bus at 
the same time as the bus is unregistered.

Alan Stern

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/