Re: [PATCH] hwmon/lm70: adding support for NS LM74 chip

[Guenter Roeck <linux@roeck-us.net>]

On Tue, Aug 28, 2012 at 02:09:00PM +0200, leroy christophe wrote:
> 
> Le 26/08/2012 19:47, Guenter Roeck a écrit :
> >On Thu, Aug 23, 2012 at 11:13:17AM -0700, Guenter Roeck wrote:
> >>On Thu, Aug 23, 2012 at 05:32:12PM +0200, Christophe Leroy wrote:
> >>>Hello,
> >>>
> >>Hi Christophe,
> >>
> >Hi again,
> >
> >[ ... ]
> Hi Guenter,
> 
> >>>-	/* 3-wire link (shared SI/SO) for LM70 */
> >>>-	if (chip == LM70_CHIP_LM70 && !(spi->mode & SPI_3WIRE))
> >>>+	/* 3-wire link (shared SI/SO) for LM70 and LM74 */
> >>>+	if ((chip == LM70_CHIP_LM70 || chip == LM70_CHIP_LM74)
> >>>+			&& !(spi->mode & SPI_3WIRE))
> >>Ok for now, but I'll have to check this. The driver does not really write
> >>anything to the chip, so it should be irrelevant which mode the SPI master
> >>controller supports (MOSI does not have to be connected). Besides, from the
> >>chip specification it looks like it is possible to connect the chip to a
> >>standard SPI interface by adding a 10k resistor between MOSI and MISO.
> >>
> >After looking into the above, the proper change or fix is really to remove the
> >above check for SPI_3WIRE entirely. The driver works perfectly fine if a LM70 is
> >connected to a controller which does not explicitly support SPI_3WIRE; all that
> >needs to be done is to connect MOSI through a resistor or not at all. This is,
> >however, a board design decision, not a SPI master chip limitation.
> Ok, then I'll prepare a patch for this.
> >
> >With that, we don't need to explicitly add any LM74 specific code to the driver,
> >since the TMP121 code does the trick. All we need to do is to update the
> >documentation, adding a note that LM74 is supported as well, and that it can be
> >selected with 'tmp121'.
> Yes, could be the solution, but I'm however a bit puzzled with doing
> this. The driver is originaly for NS LM70 component. Later on an
> extension has been added for TMP121 which is from a different
> family, even if compatible. Now we are talking about the LM74 which
> is in the same family as the original LM70. Therefore I would expect
> to be able to handle the LM74 directly from this driver, not through
> a tricky compatibility with the 'tmp121'. I would expect to handle
> it the other way round, ie consider the 'tmp121' as compatible with
> the LM74.
> One of the reasons for that is that the driver reports the component
> ID through /sys/class/hwmon/hwmonX/name
> 
> As the 'tmp121' is already in the driver, it is not a good idea to
> replace it by LM74, however, I would prefer having the LM74
> explicitly addressable by the driver, in additional to the 'tmp121'
> 
> I'm preparing a patch to include support for both LM71 and LM74 in
> the LM70 driver, in order to have a driver handling the complete
> family. Do you see a problem for that ? It would look strange that
> the LM70 driver handles LM70 and LM71, but LM74 through the
> compatible 'tmp121'
> 
Ok, I'll accept that.

> Nb, refer to your first answer, implementation for 'tmp122/124' is
> another story because those have a programmable precision, not a
> fixed one.

That isn't really a reason - we could just assume the highest precision, since
the chip returns 0 in the lower bits if it is not configured for it. But then
we can as well wait until someone asks for it, which will probably never happen.
The more intersting aspect of the tmp122/124 is that it has limit registers,
and supporting those would be more interesting anyway.

Thanks,
Guenter