On Thursday, April 04, 2013 6:59 AM, Ian Abbott wrote: > This series of patches reorganizes the comedi core module somewhat. My > main aim was originally to implement proper reference counting for > comedi devices and get rid of various race conditions related to removal > of devices. I haven't got that far yet due to time constraints but I > have got as far as simplifying the data structures used by the core, > eliminating the `struct comedi_file_info` type, which seems like a > natural break point. I plan to come back to the reference counting > stuff later, but have real (payed) work to do in the meantime and I > don't want my existing patches to go stale. (Besides, this series of > patches is starting to get a bit long already!) Looks good. Nice job getting rid of the struct comedi_file_info. Dan Carpenter's suggestion of a WARN_ON in PATCH 1 is probably a good idea. A comment about the reference counts issue might be a good idea so it isn't forgotten. Question. Are all the BUG_ON checks of the minor actually needed? comedi_alloc_subdevice() and comedi_alloc_subdevice_minor() both validate the minor before setting s->minor. The BUG_ON checks in comedi_free_subdevice_minor() should never happen and the others shouldn't be possible either. Other than that... Reviewed-by: H Hartley Sweeten <hsweeten@xxxxxxxxxxxxxxxxxxx> _______________________________________________ devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxx http://driverdev.linuxdriverproject.org/mailman/listinfo/devel
