From mboxrd@z Thu Jan 1 00:00:00 1970 From: Guenter Roeck Date: Thu, 01 Nov 2012 20:54:04 +0000 Subject: Re: [lm-sensors] [PATCH 9/9] hwmon: (it87) Report thermal sensor type as Intel PECI if appropriate Message-Id: <20121101205404.GA2974@roeck-us.net> List-Id: References: <1351448401-13985-10-git-send-email-linux@roeck-us.net> In-Reply-To: <1351448401-13985-10-git-send-email-linux@roeck-us.net> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable To: lm-sensors@vger.kernel.org On Thu, Nov 01, 2012 at 07:12:08PM +0100, Jean Delvare wrote: > On Thu, 1 Nov 2012 10:17:07 -0700, Guenter Roeck wrote: > > On Thu, Nov 01, 2012 at 05:37:57PM +0100, Jean Delvare wrote: > > > On an IT8720F chip with temp3 configured for PECI: > > >=20 > > > temp1_offset: 0 > > > temp2_offset: 0 > > > temp3_offset: -128000 > > >=20 > > > I'm unsure how to interpret this -128000... This value is definitely > > > not neutral, doesn't sound very realistic. I would expect a positive > > > number for PECI. Maybe the offset register should be treated as > > > unsigned in PECI mode? Or maybe its value is irrelevant for the IT872= 0F > > > in PECI mode. > > >=20 > > > # grep . /sys/class/hwmon/hwmon2/device/temp3* > > > /sys/class/hwmon/hwmon2/device/temp3_alarm:0 > > > /sys/class/hwmon/hwmon2/device/temp3_beep:1 > > > /sys/class/hwmon/hwmon2/device/temp3_input:25000 > > > /sys/class/hwmon/hwmon2/device/temp3_max:70000 > > > /sys/class/hwmon/hwmon2/device/temp3_min:127000 > > > /sys/class/hwmon/hwmon2/device/temp3_offset:-128000 > > > /sys/class/hwmon/hwmon2/device/temp3_type:6 > > >=20 > > > This is with an it87 driver including all your recent patches. It you > > > have ideas or theories, I'm interested. > > >=20 > > Hi Jean, > >=20 > > I agree. The alternative (+128 degrees C) sounds a bit high, though. > > Maybe there is another register on IT8720F to store Tjmax ? Did you che= ck > > in the register map if there is a register value that happens to match = it ? >=20 > How would I know which value I am looking for? k10temp says this: > high =3D +70.0=B0C, crit =3D +90.0=B0C, hyst =3D +87.0=B0C > but the monitored value doesn't match it87's temp3. >=20 > > What happens if you change the offset register value to, say, 100 ? > > Does the reported temperature change as well ? >=20 > Yes it does. > temp3_offset =3D 120000 -> temp3_input =3D 14000 > temp3_offset =3D 125000 -> temp3_input =3D 19000 > temp3_offset =3D -128000 -> temp3_input =3D 22000 > temp3_offset =3D -126000 -> temp3_input =3D 23000 >=20 > This leads me to the conclusion that the offset registers are unsigned > in PECI mode. >=20 If so, then only for AMD, or it is is a chip dependent. On Intel (with IT8728F): temp3_offset =3D -20000: temp3_input =3D -96000 temp3_offset =3D -10000: temp3_input =3D -87000 temp3_offset =3D 0: temp3_input =3D -76000 temp3_offset =3D 10000: temp3_input =3D -67000 temp3_offset =3D 20000: temp3_input =3D -56000 temp3_offset =3D 105000: temp3_input =3D 28000 (matches Tjmax / coretemp) I don't know what AMD uses as equivalent of Tjmax - maybe they use a fixed value of 128=B0C. > > Another maybe dumb question - is this with Intel or AMD CPU ? Because > > if it is AMD, it would not be PECI but AMDSI (something to fix if someo= ne > > has a system w/ IT87xx and AMD CPU). >=20 > It is AMD indeed, that's another issue I wanted to report. Do you > already know how to fix it? >=20 Yes, I think so, at least I have an idea. Give me a couple of days. Guenter _______________________________________________ lm-sensors mailing list lm-sensors@lm-sensors.org http://lists.lm-sensors.org/mailman/listinfo/lm-sensors