[PATCH 7/8] staging: unisys: stop device registration before visorbus registration

Greg KH gregkh at linuxfoundation.org
Thu Sep 3 00:38:55 UTC 2015


On Wed, Aug 26, 2015 at 07:15:26AM -0400, Jes Sorensen wrote:
> Sudip Mukherjee <sudipm.mukherjee at gmail.com> writes:
> > On Tue, Aug 25, 2015 at 08:33:55AM -0400, Jes Sorensen wrote:
> >> Sudip Mukherjee <sudipm.mukherjee at gmail.com> writes:
> >> > On Tue, Aug 18, 2015 at 03:14:02PM -0400, Benjamin Romer wrote:
> >> >> In cases where visorbus is compiled directly into the kernel, if
> >> >> visorbus registration fails for any reason, it is still possible for
> >> >> other drivers to call visorbus_register_visor_driver(), which could
> >> >> cause an oops. Prevent this by returning an error code when the bus
> >> >> hasn't been registered.
> >> >> 
> >> >> Signed-off-by: Benjamin Romer <benjamin.romer at unisys.com>
> >> >> ---
> >> >>  drivers/staging/unisys/visorbus/visorbus_main.c | 3 +++
> >> >>  1 file changed, 3 insertions(+)
> >> >> 
> >> >> diff --git a/drivers/staging/unisys/visorbus/visorbus_main.c b/drivers/staging/unisys/visorbus/visorbus_main.c
> >> >> index 7905ea9..ad2b1ac 100644
> >> >> --- a/drivers/staging/unisys/visorbus/visorbus_main.c
> >> >> +++ b/drivers/staging/unisys/visorbus/visorbus_main.c
> >> >> @@ -863,6 +863,9 @@ int visorbus_register_visor_driver(struct visor_driver *drv)
> >> >>  {
> >> >>  	int rc = 0;
> >> >>  
> >> >> +	if (!visorbus_type.p)
> >> >> +		return -ENODEV; /*can't register on a nonexistent bus*/
> >> >> +
> >> > IIRC, Greg once told that we should not be working with the internal
> >> > data structures of struct bus_type.
> >> 
> >> If you looked at the code you would have noticed this is in fact the bus
> >> driver, and visorbus_type is defined in this file. I guess we could tell
> >> visorbus_main.c to not touch visorbus_type by deleting the file
> >> completely .....
> > Yes, this is the bus driver, struct bus_type. Maybe the following might
> > be a better approach where the struct bus_type internals are not touched
> > yet it will server the purpose.
> 
> Replacing it with a global variable? Ehm sorry, but no thanks!
> 
> NACK

It should be static to the bus code.  Don't go poking at any .p
structures in the driver core, that's not ok either, I can't take this
as-is.  You should know if you can properly register a device/driver or
not.

thanks,

greg k-h


More information about the devel mailing list