Re: [lm-sensors] [PATCH 4/7] hwmon: (pmbus/ltc2978) Use detected chip ID to select supported functio

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

On Wed, Feb 27, 2013 at 10:35:48AM +0100, Jean Delvare wrote:
> On Fri, 22 Feb 2013 18:19:32 -0800, Guenter Roeck wrote:
> > We read the chip ID from the chip, use it to determine if the chip ID provided
> > to the driver is correct, and report it if wrong. We should also use the
> > correct chip ID to select supported functionality.
> > 
> > Signed-off-by: Guenter Roeck <linux@roeck-us.net>
> > ---
> > Candidate for -stable.
> > 
> >  drivers/hwmon/pmbus/ltc2978.c |    2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/drivers/hwmon/pmbus/ltc2978.c b/drivers/hwmon/pmbus/ltc2978.c
> > index eec294a..14f96ae 100644
> > --- a/drivers/hwmon/pmbus/ltc2978.c
> > +++ b/drivers/hwmon/pmbus/ltc2978.c
> > @@ -326,7 +326,7 @@ static int ltc2978_probe(struct i2c_client *client,
> >  	data->temp_min = 0x7bff;
> >  	data->temp_max = 0x7c00;
> >  
> > -	switch (id->driver_data) {
> > +	switch (data->id) {
> >  	case ltc2978:
> >  		info->read_word_data = ltc2978_read_word_data;
> >  		info->pages = 8;
> 
> Because it makes things consistent:
> 
> Acked-by: Jean Delvare <khali@linux-fr.org>
> 
> That being said, probe() isn't supposed to do device detection in the
> first place. If you want devices to be auto-detected, there is the
> detect() function for that. Right now there is no way for the user to
> instantiate a device if it doesn't have exactly the expected ID. This
> could happen though, if using a compatible part from a different
> manufacturer, or a recent part with a new ID, for example.
> 
Yes. Problem with PMBus devices, though, is that most of them can reside on any
I2C address, so a detect function would have to cover too many i2c addresses
and is not really feasible.

If there is a new part, people should use the generic PMBus driver,
not any of the device specific drivers. I have not seen an instance where the
device specific driver could be used for another chip or chip variant without
a change in the driver. Call it over-cautious, but one of the concerns I have
is that a wrongly addressed or programmed PMBus chip can easily make a board
unusable or even physically destroy it. I managed to do both several times
myself - last time just a couple of days ago with a LM25066, which went up in
smoke after writing a wrong value into one of its registers. I want to have
as many safeguards as I can in place to prevent that from happening.

Thanks,
Guenter

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors