Re: envstat drivers on SUN Ultra-45?

[Julian Coleman <jdc%coris.org.uk@localhost>]

Hi,

An update on what I've found so far.

> The device at 0x4f does look like an LM76, and I can see:
> 
> [lm76temp2]
>   temperature:    68.000  160.000   -0.062   -0.062           degC
> 
> (-0.062 is register value 0xffff).

> I didn't try reading enough registers.  The chip at 0x2b also shows values
> for registers 0x10, 0x11, 0x12, 0x20, 0x21, 0x22, 0xb0 - 0xb4 and 0xb9.

I have Solaris on another disk, so I booted this.  This has discriptions
for the sensors via "env-monitor".  So, the chip at 0x2b is actually a
LM95221 which has a local and 2 remote temperatures.  From the readings,
I'm pretty sure that remote 1 is fireasic and remote 2 is lsi1064.  I'm
also pretty sure that the chip at 0x4f is an NXP LM75A - just a single
temperature, which I think is the PSU.  Other temperatures are measured
by the ADT7462 (remote 1 is cpu0, remote 2 is cpu1 and remote 3 is the
motherboard).  Solaris env-monitor also has a front panel temperature,
but I don't know where this comes from.

I have drivers for the LM95221 and NXP LM75A, but need to add some OFW
patches to modify the properties that we read from OFW (we do this for
a few machines already).  Otherwise, there is no OFW node for them to
match on.  Looking at autoconf.c, we add patches in two places, so it
would be good to remove set_hw_props(), which we only use for i2c
devices.

> [     5.504753] WARNING: preposterous TOD clock time
> [     5.554753] WARNING: using filesystem time
> [     5.611668] WARNING: CHECK AND RESET THE DATE!

I initially assumed that the bq4802 was compatible with the chip used
in older Ultras.  However, looking at the data sheet, it has a different
register layout.  I committed a driver for that earlier today, so the
next autobuild for current will have that.

Regards,

Julian

[*] I see quite a difference in CPU temperatures and fan speeds between
cpu0 and cpu1, which surprises me:
  cpu0 59 /  5273
  cpu1 75 / 21093

--