[lm-sensors] ATI graphics card temperature -> lm-sensors

[lm-sensors] ATI graphics card temperature -> lm-sensors

[Thomas]

Hallo,

I wonder how difficult it would be to incorporate temperature and
fanspeed of ATI graphics cards into lm-sensors.

There is the aticonfig tool which provides the data (and more, like
clockspeeds, ...):
aticonfig --adapter=0 --od-gettemperature
aticonfig --pplib-cmd "get fanspeed 0"

What would be the best way to do that?
I guess I could do if I knew where to start :)

-Thomas


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

Re: [lm-sensors] ATI graphics card temperature -> lm-sensors

[Henrique de Moraes Holschuh]

On Sun, 19 Oct 2008, Thomas wrote:
> What would be the best way to do that?

I suppose:

1. document what is needed to get the data from the chip
2. add it to the ATI DRM drivers.

-- 
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh

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

Re: [lm-sensors] ATI graphics card temperature -> lm-sensors

[Hans de Goede]

Thomas wrote:
> Hallo,
> 
> I wonder how difficult it would be to incorporate temperature and
> fanspeed of ATI graphics cards into lm-sensors.
> 
> There is the aticonfig tool which provides the data (and more, like
> clockspeeds, ...):
> aticonfig --adapter=0 --od-gettemperature
> aticonfig --pplib-cmd "get fanspeed 0"
> 
> What would be the best way to do that?
> I guess I could do if I knew where to start :)
> 

aticonfig is part of the proprietary driver, so lets forget about that path, 
besides that we really want to do this as standard in kernel hwmon drivers so 
that existing lm_sensors aware tools will just pick this info up as one 
additional sensor in the system.

I've actually been looking into this, and it should be quite doable. I've 
looked at how to do this for my radeon 9800pro, so your milage may very with 
other cards.

For my 9800 pro there is a standard (and already supported) fintek i2c hwmon IC 
(I don't remember the exact model) on the card. The big problem was that its on 
the radeon multimedia i2c channel, for which no in kernel i2c master ("adapter" 
in kernel / lm_sensor speak) driver exists, now this is fixable by just writing 
such a driver, but the problem is that the X server in some special cases (ATI 
all in wonder hardware) actually uses this i2c channel and thus pokes the 
register of the multimedia i2c controller off the radeon from userspace.

But in comes modesetting, and it comes to the rescue in this case, as 
modesetting completely removes the need and ability for userspace to directly 
poke radeon registers. This has broken all in wonder support, but I'm sure this 
will get fixed eventually and that is a price we will have to pay for progress.

In the hwmon case this is actually really good as now the way is free to add 
multimedia i2c support to the drm driver (and this is something which will be 
needed to fix the all in wonder cards) and once the i2c multimedia channel is 
supported as a normal kernel i2c adapter, lm_sensors should just work, atleast 
for my 9800.

I've discussed this with Dave Airlie during the Linux Plumbers Conference, and 
he agreed that adding support for the multimedia i2c channel to the drm driver 
is the way forward. I've added him to the CC in case he has anything to add.

If you want to look into adding support for hwmon to ati cards, writing 
multimedia i2c channel drm driver code would be the first step. I believe Dave 
Airlied is the person to talk to if you want to do that. I will gladly 
volunteer as a tester for this. (radeon 9800, radeon x1950 pro)

Regards,

Hans

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

Re: [lm-sensors] ATI graphics card temperature -> lm-sensors

[Ilyes Gouta]

Hi all,

Is there any documentation about the i2c multimedia channel? I'm also 
interested in adding such a feature to my system in order to monitor my 
ATI radeon's (3650HD) temperature/fan.

Regards,
Ilyes.

Hans de Goede wrote:
> Thomas wrote:
>> Hallo,
>>
>> I wonder how difficult it would be to incorporate temperature and
>> fanspeed of ATI graphics cards into lm-sensors.
>>
>> There is the aticonfig tool which provides the data (and more, like
>> clockspeeds, ...):
>> aticonfig --adapter=0 --od-gettemperature
>> aticonfig --pplib-cmd "get fanspeed 0"
>>
>> What would be the best way to do that?
>> I guess I could do if I knew where to start :)
>>
> 
> aticonfig is part of the proprietary driver, so lets forget about that path, 
> besides that we really want to do this as standard in kernel hwmon drivers so 
> that existing lm_sensors aware tools will just pick this info up as one 
> additional sensor in the system.
> 
> I've actually been looking into this, and it should be quite doable. I've 
> looked at how to do this for my radeon 9800pro, so your milage may very with 
> other cards.
> 
> For my 9800 pro there is a standard (and already supported) fintek i2c hwmon IC 
> (I don't remember the exact model) on the card. The big problem was that its on 
> the radeon multimedia i2c channel, for which no in kernel i2c master ("adapter" 
> in kernel / lm_sensor speak) driver exists, now this is fixable by just writing 
> such a driver, but the problem is that the X server in some special cases (ATI 
> all in wonder hardware) actually uses this i2c channel and thus pokes the 
> register of the multimedia i2c controller off the radeon from userspace.
> 
> But in comes modesetting, and it comes to the rescue in this case, as 
> modesetting completely removes the need and ability for userspace to directly 
> poke radeon registers. This has broken all in wonder support, but I'm sure this 
> will get fixed eventually and that is a price we will have to pay for progress.
> 
> In the hwmon case this is actually really good as now the way is free to add 
> multimedia i2c support to the drm driver (and this is something which will be 
> needed to fix the all in wonder cards) and once the i2c multimedia channel is 
> supported as a normal kernel i2c adapter, lm_sensors should just work, atleast 
> for my 9800.
> 
> I've discussed this with Dave Airlie during the Linux Plumbers Conference, and 
> he agreed that adding support for the multimedia i2c channel to the drm driver 
> is the way forward. I've added him to the CC in case he has anything to add.
> 
> If you want to look into adding support for hwmon to ati cards, writing 
> multimedia i2c channel drm driver code would be the first step. I believe Dave 
> Airlied is the person to talk to if you want to do that. I will gladly 
> volunteer as a tester for this. (radeon 9800, radeon x1950 pro)
> 
> Regards,
> 
> Hans
> 
> _______________________________________________
> lm-sensors mailing list
> lm-sensors@lm-sensors.org
> http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

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

Re: [lm-sensors] ATI graphics card temperature -> lm-sensors

[Dave Airlie]

On Sun, 2008-10-19 at 19:24 +0200, Hans de Goede wrote:
> Thomas wrote:
> > Hallo,
> > 
> > I wonder how difficult it would be to incorporate temperature and
> > fanspeed of ATI graphics cards into lm-sensors.
> > 
> > There is the aticonfig tool which provides the data (and more, like
> > clockspeeds, ...):
> > aticonfig --adapter=0 --od-gettemperature
> > aticonfig --pplib-cmd "get fanspeed 0"
> > 
> > What would be the best way to do that?
> > I guess I could do if I knew where to start :)
> > 
> 
> aticonfig is part of the proprietary driver, so lets forget about that path, 
> besides that we really want to do this as standard in kernel hwmon drivers so 
> that existing lm_sensors aware tools will just pick this info up as one 
> additional sensor in the system.
> 
> I've actually been looking into this, and it should be quite doable. I've 
> looked at how to do this for my radeon 9800pro, so your milage may very with 
> other cards.
> 
> For my 9800 pro there is a standard (and already supported) fintek i2c hwmon IC 
> (I don't remember the exact model) on the card. The big problem was that its on 
> the radeon multimedia i2c channel, for which no in kernel i2c master ("adapter" 
> in kernel / lm_sensor speak) driver exists, now this is fixable by just writing 
> such a driver, but the problem is that the X server in some special cases (ATI 
> all in wonder hardware) actually uses this i2c channel and thus pokes the 
> register of the multimedia i2c controller off the radeon from userspace.
> 
> But in comes modesetting, and it comes to the rescue in this case, as 
> modesetting completely removes the need and ability for userspace to directly 
> poke radeon registers. This has broken all in wonder support, but I'm sure this 
> will get fixed eventually and that is a price we will have to pay for progress.
> 
> In the hwmon case this is actually really good as now the way is free to add 
> multimedia i2c support to the drm driver (and this is something which will be 
> needed to fix the all in wonder cards) and once the i2c multimedia channel is 
> supported as a normal kernel i2c adapter, lm_sensors should just work, atleast 
> for my 9800.
> 
> I've discussed this with Dave Airlie during the Linux Plumbers Conference, and 
> he agreed that adding support for the multimedia i2c channel to the drm driver 
> is the way forward. I've added him to the CC in case he has anything to add.
> 
> If you want to look into adding support for hwmon to ati cards, writing 
> multimedia i2c channel drm driver code would be the first step. I believe Dave 
> Airlied is the person to talk to if you want to do that. I will gladly 
> volunteer as a tester for this. (radeon 9800, radeon x1950 pro)

Matthew Garrett has already started to look into doing this, hopefully
this info helps him out.

Dave.


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

Re: [lm-sensors] ATI graphics card temperature -> lm-sensors

[Matthew Garrett]

On Mon, Oct 20, 2008 at 06:33:34AM +1000, Dave Airlie wrote:
> On Sun, 2008-10-19 at 19:24 +0200, Hans de Goede wrote:
> > I've discussed this with Dave Airlie during the Linux Plumbers Conference, and 
> > he agreed that adding support for the multimedia i2c channel to the drm driver 
> > is the way forward. I've added him to the CC in case he has anything to add.
> > 
> > If you want to look into adding support for hwmon to ati cards, writing 
> > multimedia i2c channel drm driver code would be the first step. I believe Dave 
> > Airlied is the person to talk to if you want to do that. I will gladly 
> > volunteer as a tester for this. (radeon 9800, radeon x1950 pro)
> 
> Matthew Garrett has already started to look into doing this, hopefully
> this info helps him out.

I've been looking into this. I'm testing with an X1900, but 
sensors-detect is currently not picking anything up on my multimedia 
bus. To be fair, neither is X, so I suspect that the interface on the 
r500s may not be identical to the one on the older cards. Or, 
alternatively, my X1900 simply doesn't have any I2C monitoring hardware. 
I should have time to poke this more this week.

-- 
Matthew Garrett | mjg59@srcf.ucam.org

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