Hi! > > I now got my feet a little wet with hid-bpf regarding something else, and > > with that knowledge I would leave the long arrays in the beginning in the > > kernel code for the time being: > > > > sirius_16_ansii_kbl_mapping and sirius_16_iso_kbl_mapping are required > > during initialization so they have to exist in the kernel code anyway. > > > > report_descriptor will most likly not change even for future models and > > afaik having report_descriptors in kernel drivers is not unheard of. > > > > So the only things that could be meaningfully moved to a hid-bpf program > > are the sirius_16_*_kbl_mapping_pos_* arrays. But for these is have to give > > out some fallback value anyway for the case where a hid-bpf file is missing > > or fails to load. So why not use real world values from my test device for > > these values? > > > > As soon as there is a future device that can use the same driver with just > > these pos arrays different, then I would implement that change via a bpf > > program instead of a change to the kernel driver. > > > > Let me know if you too think this is a sensefull approach? > > > > > > Another question: Would this patch need to wait for a userspace > > implementation of lamp array before it can get accepted? > > It would be nice if you could test the LampArray implementation. But other than that > userspace can catch up later. > > Still, i am interested in the opinion of the LED maintainers > regarding the fake HID interface. Comments from previous review were not addressed. Most importantly, this is not a way to do kernel interface. We want reasonable interface that can be documented and modified as needed. We want to pass /dev/input to userspace, not raw HID. This is not ok. Best regards, Pavel -- People of Russia, stop Putin before his war on Ukraine escalates.
Attachment:
signature.asc
Description: PGP signature
