On Wed, Jul 23, 2008 at 06:00:29PM +0100, Jonathan Cameron wrote: > Dear All, > > The need for an industrialio subsystem was discussed in > http://lkml.org/lkml/2008/5/20/135 The name is really bad, this sounds like something for doing large scale industrial process control. > Firstly thanks to all the people who have contributed to the discussion > of this in the past. > > In brief the intention is provide a kernel subsystem directed towards the > handling on sensors (and later related output devices) such as ADC's, > accelerometers and many others. We've already got an perfectly good hwmon framework, do we really need to do this again? > Key features of the subsystem include: > > * Provision of sysfs access for direct reading from devices (similar to hwmon > but without the buffering / update rate restrictions) > > * Provision of chrdevs through which events may be passed to userspace in a > similar fashion to the input subsystem. These events may be anything from > hardware thresholds set on the sensor itself to sw / hw ring buffer event > notifications (50% full etc). > > * Provision of access via chrdevs to hardware ring buffers on devices that > provide them. > > * Software ring buffer support to allow semi regular capture of data form the > device. Typically this will be driven from either datardy events, or from > a periodic timer interrupt (to this end a very simple wrapper for periodic > RTC's is included. This will move to more generic timer interfaces as and when > they become available. For now available rtc's must be registered with the > subsystem via the industrialio_register_ptimer function form within a board > init. > > * A set of sample drivers illustrating the main 'classes' of device. By classes > I really mean devices that are interfaced with in a similar way. > > The subsystem is now in a functional state with a small set of drivers: > > Max1363 (supports numerous Maxim i2c ADC's) (tested with max1363 and max1238 chips) > - Uses a periodic timer to provide ring buffer mode. > - All reads form these devices are scan modes so direct single element access > is not provided. > - Monitor mode on max1363 is not yet supported (need to do a bit debugging of > the board I have so as to be able to test this). > > ST LIS3L02DQ - SPI accelerometer. > - Uses a datardy interrupt to driver a software ring buffer. > - Most functionality of this device is supported. > > VTI SCA3000 (tested with an e05) > - Hardware ring buffer. > > More drivers in preparation. > > Next focus will be on cleaning up / implementing a more generic timer framework > and allowing the system to partly run if not all dependencies are met > (particularly availability of timers). > > An initial set of patches will be attached to this thread shortly. -- Ben Q: What's a light-year? A: One-third less calories than a regular year.