On Fri, Jan 14, 2022 at 02:55:49PM -0600, Stephen Boyd wrote: > Quoting Dmitry Torokhov (2022-01-14 11:59:50) > > On Fri, Jan 14, 2022 at 01:25:12PM -0600, Stephen Boyd wrote: > > > > > > My understanding is that 'vivaldi' is mostly a keyboard layout and > > > 'hammer' is a detachable keyboard. We want to prevent the hid-vivaldi > > > driver from probing this particular device because the hid-vivaldi > > > driver doesn't know about detachable keyboards. Hammer devices also > > > support 360 degree wraparound so we know that the keyboard has been put > > > behind the screen or that it's being used to stand up the device on a > > > table. > > > > > > Given all that, I'm still confused. If we make the hid-google-hammer > > > driver probe this device and the keyboard layout is vivaldi then we'd > > > want the part of the vivaldi driver that exposes the > > > function_row_physmap through sysfs. Otherwise userspace won't know how > > > to handle the function row properly. I think we need the device to stack > > > two drivers here. Does that happen with HID? > > > > As far as I know HID does not easily allow "stacking" drivers like that. > > Ok. > > > > > Probably the easiest way would be to export vivaldi_feature_mapping() > > and the show method for the physical row map and call them from the > > hammer driver. > > > > I worry about builtin vs. modular drivers so probably ought to make some Just make hid-hammer depend on hid-vivaldi, hid-vivaldi is tiny otherwise. > hid-vivaldi-common.c file that has the physmap code and then have both > drivers call that mini-library. The 'vivaldi_data' structure would need > to be figured out too. The hammer driver stores 'hammer_kbd_leds' in the > hid_get_drvdata() whereas the vivaldi driver stores 'vivalid_data' so we > can't simply call the show method for the sysfs attribute without some > minor surgery. Yeah, we'll need to convert them into helpers into which callers can pass buffers. Thanks. -- Dmitry
