Am 03.06.2007 22:28 schrieb Lars K.W. Gohlke: > Tilman Schmidt schrieb: >> Am 17.05.2007 08:15 schrieb huang ying: >>> I think the "serio" (through drivers/input/serio/serport.c) may be a >>> choice too, like that in linux/drivers/input/mouse/sermouse.c, which >>> is an example to program serial port in kernel space. >> Interesting. I wonder if that would have been a better choice for >> the Gigaset M101 driver. It seems even to have a probe mechanism >> so one could try to determine if the expected device is really >> connected to the port. > >> Is there any documentation on this interface? I find the source a >> bit hard to understand, sparsely commented as it is. > > how can I open ttyS1 with major=4 and minor=65? a) With "my" approach, ie. writing a line discipline, you open /dev/ttyS1 in a user space program and then push your line discipline onto it using ioctl(,TIOCSETD,). b) With the "serio" approach, if the probe mechanism is really what I think it is, you register your probe function and wait for it to be called for each active serial port, then do your thing to determine whether this is a/the port you want to use. > Does anybody have some code to read from it the first e.g. 2bytes? That's not how things work in the kernel. You don't "read the first bytes" at will. You have to provide means to process whatever data happens to arrive on the serial port, as it arrives. a) With a line discipline, you provide a callback function ".receive_buf" which will be called by the serial driver if some data has been received, and has to process that data then and there. You might just stow it away in a buffer of course, provided you are sure to get around to processing it later. b) With the "serio" interface, it seems the ".interrupt" function serves the same purpose - though, as I said, I am not really competent on that topic. HTH T. -- Tilman Schmidt E-Mail: tilman@imap.cc Bonn, Germany Diese Nachricht besteht zu 100% aus wiederverwerteten Bits. Ungeöffnet mindestens haltbar bis: (siehe Rückseite)