Kenneth Aafløy writes: > On Wednesday 21 July 2004 00:34, Ralph Metzler wrote: > [snip] > > > > Alternative: get the i2c_adapter pointer via sub_device, > > > > lock it, (in case of linuxtv driver, change i2c adapter name), > > > > scan it for already present clients, init them, > > > > > > How the initialization is done? > > > > Via i2c client command calls. FE_REGISTER and FE_UNREGISTER call the > > dvb_(un)register_frontend functions. > > Currently, most, if not all frontends does not do that. They rather register > the frontend in i2c_attach_adapter.
Yes, sorry, it looks like this was done differently in the linuxtv drivers. > > > > somehow add a callback for sub_devices in attach_inform > > > > for future attachments, do the same for detachments (if one > > > > allows it at all), unlock i2c_adapter. > > > > > > We'll need both attach and detach I think. "rmmod $frontend" should > > > work, even if it is only for small driver development turnaround times. > > > > Yes, definitely better for debugging. > > As long as the frontend is unused by a *real* user it should be unloadable, > and that is also fixed in current CVS. Is it currently also possible to unload the dvb_adapter (und thus i2c_adapter) before the frontend? If FE_UNREGISTER is not really unregistering this would also lead to problems!? Ralph
