On Sun, 2009-04-05 at 06:14 -0300, Mauro Carvalho Chehab wrote: > > IMO, doing all those tricks to support an out-of-tree driver is the wrong > approach. This is just postponing a more serious discussion about what should > be done in kernel, in order to better support IR's. The "tricks" were to not break the current user experience by saddling them with a kernel module that they may not want. (If the tricks are needed at all - I'm will test later today.) I agree that better in kernel infrastructure needs to be in place for IR. LIRC's kernel space infrastructure module, lirc_dev, probably isn't a bad model for support of IR devices. The individual LIRC modules for supporting specific sets of devices are the ones that have problems to varying degrees. > In the case of lirc, the userspace part has already an event interface. If the > drivers are doing the right thing with their IR part, lirc can just use the > event interface for all drivers. This seems to be the proper approach. Input event interfaces alone neglect IR blasters. > >From what I got from Andy and Mike's comments is that the real issue is that > the IR kernel code is incomplete, broken or bad designed. So, several users and > userspace apps don't rely on the kernel code but, instead, use lirc as an > alternative. There is at least one other motivation: LIRC also handles a number of other hardware interfaces that are not I2C: serial ports (/dev/ttySX), parallel port, USB, etc. I happen to use the lirc_mceusb2 module for my Phillips Home IR receiver/blaster (I'm not sure if the blaster works under linux.) > That's said, I propose a different approach: > > 1) Add some entry at feature-removal-schedule.txt posting a date to end support > for out-of-tree I2C IR modules; > > 2) Start discussing with lirc people (and input/event maintainers if needed) > about what is needed to properly support the required functionalities for a > better lirc usage; > > 3) Propose a few API additions in order to support those functionalities; > 4) apply IR patches on kernel to support the missing functionalities; The scope of a complete kernel IR infrastructure goes a bit beyond I2C bus devices that are only input devices. What's the scope of what you want to tackle here? I certainly don't want to reinvent something that's going to look just like the LIRC driver model: http://www.lirc.org/html/technical.html Which already has an infrastructure for IR driver module writers: http://www.lirc.org/html/technical.html#lirc_dev Do we just convert lirc_dev, lirc_i2c, and lirc_zilog to a cleaned up set of in kernel modules? lirc_i2c can certainly be broken up into several modules: 1 per supported device. Should these create an input device as well to maintain compatability with apps expecting an ir-kbd-i2c like function? Or do we split up ir-kbd-i2c into per device modules and in addition to the input event interface, have it register with the lirc_dev module? Do we leverage LIRC's lirc_dev infrastructure module at all? (I think it would be a waste of time not to do so.) Regards, Andy > 5) remove the support for out-of-tree i2c IR modules. > > Cheers, > Mauro > -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
