Re: [PATCH 0/7] char/mei: Intel MEI Driver

[Arnd Bergmann <arnd@arndb.de>]

On Wednesday 23 March 2011, Greg KH wrote:
> On Wed, Mar 23, 2011 at 05:55:47PM +0200, Weil, Oren jer wrote:
> > >> >I suppose one of the important things to do here is integrate the
> > >> >watchdog part into the common watchdog infrastructure in
> > >> >drivers/watchdog. Maybe this driver can export a simple interface to
> > >> >another module that implements the watchdog device?
> > >>
> > >> Yes, this is one the ideas the Alan suggested, we are thinking about
> > >> this direction in the future.
> > >
> > >What's keeping you from doing this now?
> > 
> > We need to change the driver to expose functions to communicate with the FW
> > Client (feature) 
> > From other kernel models, the design for now is only from user space.

For a simple example, look at drivers/watchdog/riowd.c (most other
files in that directory do as well), just replace the riowd_writereg()
function in there with one that calls an exported function from your
hardware. This should really be trivial to do.
 
> So you need to add support to the watchdog infrastructure for this?  For
> some reason I thought you could kick the watchdog from within kernel
> code, did I get this wrong?

A watchdog must be triggered from user space, in Linux we do this by
writing to /dev/watchdog. If you did it from kernel space, the watchdog
would not detect the case where the system is half-broken and still
able to trigger the watchdog, but not do much beyond that.

	Arnd