Re: [Q] i2c-taos-evm bus driver

[Dmitry Torokhov]

On Thu, Feb 11, 2010 at 12:12:51PM -0500, H Hartley Sweeten wrote:
> On Tuesday, February 09, 2010 1:43 PM, Jean Delvare wrote:
> > Hi Hartley,
> 
> Hello.
> 
> > On Tue, 9 Feb 2010 11:38:37 -0500, H Hartley Sweeten wrote:
> >> On Tuesday, February 09, 2010 1:17 AM, Jean Delvare wrote:
> >>> No, it works reasonably fine. I was using it myself no later than one
> >>> month ago. Out of curiosity, with which exact evaluation module do you
> >>> plan to use it?
> >> 
> >> I'm not actually...
> >> 
> >> I am trying to figure out how a serio driver actually gets "hooked" to a
> >> serial port.  The i2c-taos-evm driver looked simple enough to play with
> >> in order to figure it out.  I was trying to follow the Documentation in
> >> order to use it and ran into the issue below.
> >
> > OK... but please keep in mind that you won't be able to use this driver
> > without supported hardware. The driver will not bind to the serial port
> > if it doesn't detect a supported device.
> 
> That's what I am trying to figure out.
> 
> Now that I know the baud rate and port setting from your patch to inputattach.c
> I was just going to dummy something up to one of my serial ports.  I really
> don't need the i2c-taos-evm driver to "work" I just want to figure out how
> the serio connection is made.
> 
> It still seems strange to me that a user space application is needed to connect
> two kernel drivers together.  Actually three if you include the serio "bridge".
> I assume the actual connection between the drivers is something like this:

It is much easier to initialize device from userspace instead of putting
all this code into the kernel, it simplifies interrupt handler and the
overlall driver structure. None of these are hotpluggable so there is no
issue with running a program at startup.

BTW, several drivers do do autopropbe (like sunkbd) but most rely on
inputattach to prepare port and device.

> 
> 	some_tty_driver <--> serio <--> i2c-taos-evm
> 

Yep, this is correct.

> >> (...)
> >> Do you know if there is a way to make the connection in kernel?
> 
> > I suppose this is possible from a pure code perspective, but I can't
> > see why one would want to do that. The kernel has no idea what is
> > connected to the serial ports: devices on the serial port can't be
> > reliably nor safely probed. This is why we rely on user-space to
> > declare which device is connected to which port, based on the
> > assumption that the user knows what he or she is doing.
> 
> My intention is to do this in an embedded system.  The serial port
> I am using will always be connected to the target device.  I am just
> Trying to figure out a way to make the kernel driver connection.

Just have your driver bind to SERIO_ANY and have it query the device.
You still need to create serio of proper type (with input attach or see
drivers/serial/sunzilog.c for example of creating dedicated serio port).

> 
> >> Also, do you know if there is any information on the serio stuff
> >> available anywhere?  Other than just reading the kernel source I
> >> have not been able to locate anything.
> >
> > I seem to remember I hit the exact same problem back when I wrote
> > the i2c-taos-evm driver. I ended up reading the source code of other
> > serial device drivers and used them as an example. It was enough to
> > get things to work, but I won't claim I understood all of serio...
> > just enough to do what I needed back then.
> >
> > Maybe Dmitry Torokhov (Cc'd) will be more helpful than me.
> 
> Hopefully... There seems to be a real lack of information available.
> 

Neither Vojtech nor I had time to document serio. Patches are always
welcome of course ;)

-- 
Dmitry
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/