On Sat, Jan 28, 2012 at 03:34:00PM -0500, Jean Delvare wrote:
> On Sat, 28 Jan 2012 09:04:28 -0800, Guenter Roeck wrote:
> > Logic was that I only need to check if the fans are connected to 90..92 if they are
> > not connected to 3..5. However, it looks like 90..92 take precedence, so
> > I think I have to check if the fans are connected to 90..92 first, and only
> > check if they are connected to 3..5 if they are not connected to 90..92.
> >
> > Something like:
> >
> > fan3pin = 0;
> > fan4pin = 0;
> > fan5pin = 0;
> > if (superio_inb(sio_data->sioreg, 0x27) & 0x80) {
> > superio_select(sio_data->sioreg, W83627EHF_LD_HWM);
> > regval = superio_inb(sio_data->sioreg, SIO_REG_ENABLE);
> > fan3pin = !!(regval & 0x80);
> > fan4pin = !!(regval & 0x40);
> > fan5pin = !!(regval & 0x20);
> > }
> > if (!fan3pin)
> > fan3pin = !(superio_inb(sio_data->sioreg, 0x24) & 0x40);
> > if (!fan4pin)
> > fan4pin = !!(superio_inb(sio_data->sioreg, 0x1C) & 0x01);
> > if (!fan5pin)
> > fan5pin = !!(superio_inb(sio_data->sioreg, 0x1C) & 0x02);
> >
> > Does that make sense ?
>
> Yes, this should work in practice. However some improper chip
> configurations would not be dealt with properly by the code above. For
> example bits 5, 6 and 7 of configuration register SIO_REG_ENABLE set but
> bit 7 of configuration register 0x27 not set makes no sense, but if the
> chip is so configured then I believe the fans cannot be monitored,
> regardless of the bit values in configuration registers 0x24 and 0x1C.
> If you want to handle this case, you have to check every fan
> separately, like:
>
> gpok = superio_inb(sio_data->sioreg, 0x27) & 0x80;
> regval = superio_inb(sio_data->sioreg, SIO_REG_ENABLE);
>
> if (regval & 0x80)
> fan3pin = gpok;
> else
> fan3pin = !(superio_inb(sio_data->sioreg, 0x24) & 0x40);
>
> if (regval & 0x40)
> fan4pin = gpok;
> else
> fan4pin = !!(superio_inb(sio_data->sioreg, 0x1C) & 0x01);
> [etc]
>
> It's up to you if you want to do that for absolute correctness
> (assuming I read the datasheet properly...) or prefer to stick to your
> approach which should be equivalent for properly configured chips.
>
I might as well (try to) do it right, so I'll use your approach.
I'll send an updated patch in a minute.
Guenter
_______________________________________________
lm-sensors mailing list
lm-sensors@xxxxxxxxxxxxxx
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
