Hello again folks: So it seems that there is some traction for bringing the comedi drivers into the mainline kernel. I thought I'd start to put together a list of items we may want to tackle. Let us know if you think these are good or bad ideas... * Cleanup and documentation for or removal of items listed under "4.7. Experimental functionality" in the comedilib documentation. In particular, analog triggering seems like a feature we should support. * Support for having a subdevice that can do both input and output commands * Documentation of the interface for "pulse-based signals". Do we have a well-baked API for counters, timers, encoders, PWM signals, etc? For what it's worth, I found it confusing last time I looked at it. * Removal of "slowly-varying input" filtering * A better mechanism for adding extensions than INSN_CONFIG * Use of plug-n-play/hotplug capabilities to remove need for comedi_config when possible * Improved driver portability: consistent use of u32, etc >From Frank Mori Hess: > First, the protocol for doing output commands needs to be changed (again). > The current scheme of doing: "command, load buffer, internal trigger" > doesn't translate to support external triggers. We need "command, load > buffer, arm". Then for start_src=TRIG_NOW the command would start > immeditately upon arming, analogously to how start_src=TRIG_INT works now. > For start_src=TRIG_EXT, the arm step would allow the analog output fifo to > be loaded or dma transfer initiated, and the board configured to start on > the external trigger input. Some cosmetic things: * Removal of typedefs; e.g. use of "struct comedi_subdevice" instead of "typedef struct comedi_subdevice_struct comedi_subdevice". * Removal of lsampl_t and sampl_t in favor of u32 and u16? This would only affect the kernel to comedilib interface -- comedilib should continue to use lsampl_t. * Removal of any legacy kernel function/struct/define names. (At first glance, the older-kernel support is done quite cleanly, so hopefully this is a no op). * Whitespace cleanup On 10/19/07, Bernd Porr <BerndPorr at f2s.com> wrote: > A while ago I had integrated comedi into the early 2.6.x kernels. > I might find the sed script somewhere which converted the > comedi drivers on the fly to fit into the kernel. Bernd, if you could find this script, it might be useful. -Tabish
