Hi again,
Sorry for a late reply - I apprecieate your help. I had other duties
(and then I had a flu), to do so I didin't have time to look into this
and reply before.
But, I believe I've had no luck so far. Maybe the additional sensor is
not handled trough the nct6791d after all? Details inline:
2015-04-12 18:40 GMT+03:00 Guenter Roeck <linux@xxxxxxxxxxxx>:
On 04/12/2015 06:37 AM, Ville Aakko wrote:
2015-04-11 5:12 GMT+03:00 Guenter Roeck <linux@xxxxxxxxxxxx>:
[SNIP]
Additional temperature registers:
Bank 0:
0x27 SMIOVT1
The temperature source for this register is set in bank 6, register 0x21.
0x73, 0x74 Monitor Temperature Register 1 (0x73=fraction,
0x74=Integer)
0x75, 0x76 Reg 2
0x77, 0x78 Reg 3
0x79, 0x7a Reg 4
0x7b, 0x7c Reg 5
However, if I've understod your information correctly, SMIOVT1 is
always shoving 0x24 or 0x25, and SMIOVT2 seems to be 0x33. The select
registers are set to 01 and 02, (and reg 0x0, bank 0xa to 0x03), which
does not make sense? They should be set to 0x73 to 0x7c to provide
anything meaningful? See below:
I hope I did not say that ;-). The source select registers always
have the same definition, and that definition is further below.
0x01 is SYSTIN, 0x02 is CPUTIN.
Yes, I was confused about the bank in question since I haven't fiddled
with this kind of things before =). But I think I understand now what
you meant.
Should I try to write to registers (0x0 in bank 1, 2, 3, 8 or 9 to
anything ranging from 0x0 to 0x20 and the corresponding set registers)
with isaset? I'm not sure exactly what I should input, and the isaset
man page does not actually encourage tinkering if one is not totally
sure what he/she is doing =). How to select bank with isaset? I'll be
responsible of course if I will blow up my computer with isaset, but
I'd still like to know if that is the right thing to do before I try
it.
You could try that with bank 8. The PCH temperatures as well as BYTE_TEMP
are also reported in bank 4, registers 0x00 to 0x1a (two registers per
temperature). Those all report 0 per your log above.
The only unreported temperatures are therefore the temperatures
reported through the SMBUs master sources. So you could set register
0x00 of bank 8 to 0x08 .. 0x0f and see if that makes a difference
for the temperature reported at 0x79 / 0x7a.
The bank register is at register 0x4e in every page. You would set
the bank first with isaset, followed by the actual isaset command.
sudo isaset -y 0x295 0x296 0x4e 8
sudo isaset -y 0x295 0x296 0x0 0x08 [0x09, 0x0a, ... , 0x0f]
sudo isadump -y 0x295 0x296 0
If 0x79 / 0x7a reports anything but 0, we may have found your sensor.
You might want to unload the nct6775 driver before doing those tests.
I did this. At first I need to load (after a reboot) the nct6775
driver - otherwise, all the registers would report all 0xff's when I
tried to dump them with isadump. So something is initialized by the
driver, is this normal? But after modprobe I removed the driver and
tried the commands. I even did it again after a fresh reboot, in case
the chip was in some odd state because I could have mad a mistake /
after I had been running lm_sensors initialization script (I disabled
the script / service to make sure nothing has been done to the
registers).
In 0x79 I usually had 0xfd, but I had seen anything from 0xfe to 0xff
without any apparent pattern. 0x7a is always 00.
I also dumped bank 8 to check 0x00 that it was written corretly - and
it was. I also wrote the value for PCH and SYSTIN in the select
register, since I know what PCH and SYSTIN should approximately show,
and the right values seemed to appear in bank 0 / 0x79-0x7a.
So, any other ideas?
If you are running out of ideas, I was thinking about contacting ASUS
support, maybe they'd like to share some information on this (where
the additional sensor is actually handled and if they'd like to
disclose any information). Hearing / reading about their reputation
earlier on Linux support, I'd doubt that would help, but I guess it
couldn't hurt either to ask.
(I can provide the dumps from isadump if you wish, but I didn't notice
anything peculiar in them)
I am wondering - when you do all your tests in the BIOS, do you
see the temperature (changes) reported there correctly ?
Yes. I tried it (in real time while watching the figure) in room
temperature (21-22°C), glass of (fresh / really cold and full of ice)
water (0°C) and my graphics card heat sink (44°C, though I don't have
a reference measurement for the heat sink).
Thanks for your help!
--
--
Ville Aakko - ville.aakko@xxxxxxxxx
