Re: [RFC v2 1/1] HID: User-space I/O driver support for HID subsystem

[Nicolas Pouillon <npouillon@freebox.fr>]

David,

----- Original Message -----

> >> + =C2=A0 =C2=A0 strncpy(hid->name, ev->u.create.name, 128);
> >> + =C2=A0 =C2=A0 hid->name[127] =3D 0;
> >> + =C2=A0 =C2=A0 hid->ll_driver =3D &uhid_hid_driver;
> >> + =C2=A0 =C2=A0 hid->hid_get_raw_report =3D uhid_hid_get_raw;
> >> + =C2=A0 =C2=A0 hid->hid_output_raw_report =3D uhid_hid_output_raw;
> >> + =C2=A0 =C2=A0 hid->bus =3D ev->u.create.bus;
> >> + =C2=A0 =C2=A0 hid->vendor =3D ev->u.create.vendor;
> >> + =C2=A0 =C2=A0 hid->product =3D ev->u.create.product;
> >> + =C2=A0 =C2=A0 hid->version =3D ev->u.create.version;
> >> + =C2=A0 =C2=A0 hid->country =3D ev->u.create.country;
> >> + =C2=A0 =C2=A0 hid->phys[0] =3D 0;
> >> + =C2=A0 =C2=A0 hid->uniq[0] =3D 0;
> >> + =C2=A0 =C2=A0 hid->driver_data =3D uhid;
> >> + =C2=A0 =C2=A0 hid->dev.parent =3D uhid->parent;
> >
> > Here we have to make sure that we have all required information provide=
d
> > by userspace. Anything else that HID might require to work better. Or
> > that BR/EDR or LE provides additionally over USB.
>=20
> Beside phys and uniq I have copied all information that HIDP and
> USBHID use. I haven't found any other field that could be of interest
> here. We can also always add UHID_CREATE2 with additional fields to
> allow further additions (or use a "size" field as you suggested
> below).

HID report descriptor may reference Physical Descriptor (HID 6.2.3) and
String Descriptor entries (USB 9.6.8, HID E.11).  Physical Descriptor
blob could be another __user buffer + size in uhid_create_req.  String
descriptor should probably be an event for querying strings.

hidraw interface also offers access to Feature Reports, adding events
to send/receive and query (kernel querying uhid client) feature reports
could be of some interest.

Cheers,
--=20
Nicolas Pouillon