Re: [PATCH 1/8] rmnet: Add skeleton
Denis Kenzior <[email protected]>
| Newsgroups | dev.linux.lists.ofono |
|---|---|
| Message-ID | <[email protected]> |
Hi Marcel, > > does this make really sense? Putting this into drivers/qmimodem/rmnet.c is not possible? > I switched most of the atom drivers to use the OFONO_ATOM_DRIVER_BUILTIN macro. The drivers that performed static initialization during 'plugin_init' using OFONO_PLUGIN_DEFINE went *poof*, including qmimodem. Only individual atom drivers remain. There is one straggler, namely drivers/stemodem/stemodem.c. That one does something somewhat similar in drivers/stemodem/caif_rtnl.c. Similarly, for modem drivers, OFONO_MODEM_DRIVER_BUILTIN was introduced. This leaves only a few users of OFONO_PLUGIN_DEFINE. Longer term I'd like to get rid of the whole plugin concept. However, it still leaves the open question of what to do with code that requires plugin_init / plugin_exit style static initialization. Easiest way is to convert such users to core modules, like I do with rmnet here, but I'm open to suggestions. > We might actually need a generic abstraction for del_interface / add_interface and allow for drivers to register their implementation for it, but that is a different story. Yes, true. This will surely come up with MBIM. On another note, I was thinking of eventually moving transport implementations into src/ or transport/ to allow easier mixing, e.g. running QMI over MBIM, or AT commands over QMI, etc. Regards, -Denis