On Thu, Dec 08, 2022 at 11:02:36AM +0100, Greg Kroah-Hartman wrote: > On Mon, Dec 05, 2022 at 02:24:03PM +1300, Daniel Beer wrote: > > On Sun, Dec 04, 2022 at 10:39:21AM +0100, Greg Kroah-Hartman wrote: > > > > Thanks for reviewing. This device is quite strange -- it presents itself > > > > as a USB HID, but it provides both an I2C master and a UART. The > > > > existing driver supports only the I2C functionality currently. > > > > > A whole new major for just a single tty port? Please no, use dynamic > > > > > majors if you have to, or better yet, tie into the usb-serial > > > > > implementation (this is a USB device, right?) and then you don't have to > > > > > mess with this at all. > > > > > > > > As far as I understand it, I don't think usb-serial is usable, due to > > > > the fact that this is already an HID driver. > > > > > > That should not be a restriction at all. You are adding a tty device to > > > this driver, no reason you can't interact with usb-serial instead. That > > > way you share the correct userspace tty name and major/minor numbers and > > > all userspace tools should "just work" as they know that name and how to > > > interact with it already. > > > > > > Try doing that instead of your own "raw" tty device please. > > > > Maybe I've misunderstood something. The reason I thought usb-serial was > > unusable in this instance was that I couldn't see a way to create a port > > except via usb-serial's own probe function (otherwise, the API looked > > fine). > > > > I don't know whether I'm looking at a serial or an I2C interface until > > after it's already been probed by HID core, I have a struct hid_device > > and I've asked what type of interface it is via an HID feature report. > > This can't be determined otherwise, because strapping pins affect the > > presentation of interfaces. > > > > At that point, I (currently) call uart_add_one_port. I might have missed > > it, but I didn't see anything analogous in the usb-serial API. Am I > > going about this the wrong way? > > I thought that this could be done, but I might be wrong. Johan, any > ideas? It seems to me like this should be implemented as a new tty driver as neither USB-serial or serial (core) is a good fit for such a HID device. So this appears to be right approach in general: https://lore.kernel.org/all/20221207220617.116082-1-contact@xxxxxxxxxxxxxxxxxx/ Johan
