Re: [PATCH 2/8] HID: usbhid: Make hid_is_usb() non-inline

[Benjamin Tissoires]

On Thu, Dec 22, 2022 at 10:46 PM Thomas Weißschuh <thomas@xxxxxxxx> wrote:
>
>
> Dec 22, 2022 16:13:06 Benjamin Tissoires <benjamin.tissoires@xxxxxxxxxx>:
>
> > On Thu, Dec 22, 2022 at 6:16 AM Thomas Weißschuh <linux@xxxxxxxxxxxxxx> wrote:
> >>
> >> By making hid_is_usb() a non-inline function the lowlevel usbhid driver
> >> does not have to be exported anymore.
> >>
> >> Also mark the argument as const as it is not modified.
> >>
> >> Signed-off-by: Thomas Weißschuh <linux@xxxxxxxxxxxxxx>
> >> ---
> >> drivers/hid/usbhid/hid-core.c | 6 ++++++
> >> include/linux/hid.h           | 5 +----
> >> 2 files changed, 7 insertions(+), 4 deletions(-)
> >>
> >> diff --git a/drivers/hid/usbhid/hid-core.c b/drivers/hid/usbhid/hid-core.c
> >> index be4c731aaa65..54b0280d0073 100644
> >> --- a/drivers/hid/usbhid/hid-core.c
> >> +++ b/drivers/hid/usbhid/hid-core.c
> >> @@ -1334,6 +1334,12 @@ struct hid_ll_driver usb_hid_driver = {
> >> };
> >> EXPORT_SYMBOL_GPL(usb_hid_driver);
> >>
> >> +bool hid_is_usb(const struct hid_device *hdev)
> >> +{
> >> +       return hdev->ll_driver == &usb_hid_driver;
> >> +}
> >> +EXPORT_SYMBOL_GPL(hid_is_usb);
> >> +
> >> static int usbhid_probe(struct usb_interface *intf, const struct usb_device_id *id)
> >> {
> >>         struct usb_host_interface *interface = intf->cur_altsetting;
> >> diff --git a/include/linux/hid.h b/include/linux/hid.h
> >> index 8677ae38599e..e8400aa78522 100644
> >> --- a/include/linux/hid.h
> >> +++ b/include/linux/hid.h
> >> @@ -864,10 +864,7 @@ static inline bool hid_is_using_ll_driver(struct hid_device *hdev,
> >>         return hdev->ll_driver == driver;
> >> }
> >>
> >> -static inline bool hid_is_usb(struct hid_device *hdev)
> >> -{
> >> -       return hid_is_using_ll_driver(hdev, &usb_hid_driver);
> >> -}
> >> +extern bool hid_is_usb(const struct hid_device *hdev);
> >
> > The problem here is that CONFIG_USB_HID can be set to either m or n.
> > In the n case, you'll end up with an undefined symbol, in the m case,
> > it won't link too if CONFIG_HID is set to Y (and it'll be quite a mess
> > to call it if the module is not loaded yet).
>
> Shouldn't we already have the same problem with
> the symbol usb_hid_driver itself that is defined
> right next to the new hid_is_usb()?

Yeah, sorry, my bad. All of the callers of this function are modules
which depend on CONFIG_USB_HID in the Kconfig, so we should be good.
Sorry for the noise.

I shouldn't do reviews at 10pm :(

Cheers,
Benjamin

>
> Thomas
>
> >>
> >> #define        PM_HINT_FULLON  1<<5
> >> #define PM_HINT_NORMAL 1<<1
> >>
> >> --
> >> 2.39.0
> >>
>