Foster Snowhill <forst@xxxxxx> writes: > As the regular tethering implementation of iOS devices isn't compliant > with CDC NCM, it's not possible to use the `cdc_ncm` driver to handle > this functionality. Furthermore, while the logic required to properly > parse URBs with NCM-encapsulated frames is already part of said driver, > I haven't found a nice way to reuse the existing code without messing > with the `cdc_ncm` driver itself. Sorry for taking this up at such a late stage, and it might already been discussed, but I'm not convinced that you shouldn't "mess" with the `cdc_ncm` driver. We did a lot of that when prepping it for cdc_mbim reuse. Some of it, like supporting multiple NDPs, is completely useless for NCM. We still added code to cdc_ncm just to be able to share it with cdc_mbim. I think the generalized solutions still adds value to cdc_ncm by making the shared code targeted by more developers and users. And the huawei_cdc_ncm driver demonstrates that cdc_ncm can be reused for non-compliant vendor-specific protocols, so that's not a real problem. I don't really see the assymmetric TX/RX protocols as a big problem either. Reusing only the RX code from cdc_ncm should be fine. What I can understand is that you don't want to build on the current cdc_ncm code because you don't like the implementation and want to improve it. But this is also the main reason why I think code sharing would be good. The cdc_ncm code could certainly do with more developers looking at it and improving stuff. Like any other drivers, really. Yes, I know I'm saying this much too late for code that's ready for merging. And, of course, I don't expect you to start all over at this point. I just wanted to comment on that "messing with" because it's so wrong, and I remember feeling that way too. Messing with existing code is never a problem in Linux! Existing code and (internal) interfaces can always be changed to accommodate whatever needs you have. This is one of may ways the Linux development model wins over everything else. Bjørn
