On Fri, Mar 26, 2021 at 09:10:48AM -0300, Jason Gunthorpe wrote:
> It is usually hidden and works like this:
>
> /* pci_register_driver() must be a macro so KBUILD_MODNAME can be expanded */
> #define pci_register_driver(driver) \
> __pci_register_driver(driver, THIS_MODULE, KBUILD_MODNAME)
>
> int __pci_register_driver(struct pci_driver *drv, struct module *owner,
> const char *mod_name)
> {
> drv->driver.owner = owner;
> drv->driver.mod_name = mod_name;
Indeed, there seem to be about two handful of instance of that.
> > I've not really seen that in anywere else, and the only user seems
> > to be module_add_driver for a rather odd case we shouldn't hit here.
>
> vfio_mdev could be compiled built in?
>
> AFAICT it handles the case where THIS_MODULE==NULL so we still need to
> create sysfs links to the built in module.
>
> If it is left NULL then a few sysfs files go missing for the built in
> mode but no harm done?
Yes, it seems to be needed for a few driver-specific files. So it looks
ok, even if rather unexpected.
