Re: device_detach() never clears devclasses

This looks good to me.  I think I should have had !=3D here anyway. :)

Warner

On Feb 27, 2012, at 10:08 AM, John Baldwin wrote:

> I haven't reproduced this, I just have a hunch based on code-reading =
while=20
> fixing another bug.  Ah, looking in SVN history it looks like this was =
broken=20
> in 129711 when device_attach() was split out of =
device_probe_and_attach(). =20
> Specifically, if DEVICE_ATTACH() fails then this:
>=20
> 		/* Unset the class; set in device_probe_child */
> 		if (dev->devclass =3D=3D NULL)
> 			(void)device_set_devclass(dev, NULL);
>=20
> is wrong.  dev->devclass is always non-NULL at this point.  However, =
we should=20
> be clearing the devclass for any device that doesn't have a fixed =
device=20
> class.  And in fact that is what device_detach() does:
>=20
> 	if (!(dev->flags & DF_FIXEDCLASS))
> 		devclass_delete_device(dev->devclass, dev);
>=20
> The patch below fixes this:
>=20
> Index: subr_bus.c
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> --- subr_bus.c	(revision 232218)
> +++ subr_bus.c	(working copy)
> @@ -2732,9 +2732,8 @@ device_attach(device_t dev)
> 	if ((error =3D DEVICE_ATTACH(dev)) !=3D 0) {
> 		printf("device_attach: %s%d attach returned %d\n",
> 		    dev->driver->name, dev->unit, error);
> -		/* Unset the class; set in device_probe_child */
> -		if (dev->devclass =3D=3D NULL)
> -			(void)device_set_devclass(dev, NULL);
> +		if (!(dev->flags & DF_FIXEDCLASS))
> +			devclass_delete_device(dev->devclass, dev);
> 		(void)device_set_driver(dev, NULL);
> 		device_sysctl_fini(dev);
> 		dev->state =3D DS_NOTPRESENT;
>=20
>=20
> --=20
> John Baldwin
>=20
>=20

_______________________________________________
[email protected] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-new-bus
To unsubscribe, send any mail to "[email protected]"