[lm-sensors] sensors-detect changes monitor settings on hp pavilion dv7

[lm-sensors] sensors-detect changes monitor settings on hp pavilion dv7

[Andrew Potter]

[-- Attachment #1.1: Type: text/plain, Size: 762 bytes --]

Hi, I wasn't quite sure from the information on lm-sensors.org how to open
an official ticket, but the details of my problem are here:

https://bbs.archlinux.org/viewtopic.php?pid=1170583

Basically, when I probe i915 gmbus panel (i2c-2) with sensors-detect, when
it probes address 0x4f my monitor settings (I think the sharpness mostly)
are changed drastically.  This is persistent over any operating system and
reboot, leading me to believe a) its a hardware issue, or b) nothing
usually touches whatever values were changed.  Any help would be greatly
appreciated, in the forum post I've also included some experimentation with
i2c-tools.  The problem originally occured on an Arch Linux system.  Let me
know if you need anything else!

Thanks,
Andrew Potter

[-- Attachment #1.2: Type: text/html, Size: 1481 bytes --]

[-- Attachment #2: Type: text/plain, Size: 153 bytes --]

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

Re: [lm-sensors] sensors-detect changes monitor settings on hp pavilion dv7

[Guenter Roeck]

On Thu, Oct 04, 2012 at 11:30:45AM -0400, Andrew Potter wrote:
> Hi, I wasn't quite sure from the information on lm-sensors.org how to open
> an official ticket, but the details of my problem are here:
> 
> https://bbs.archlinux.org/viewtopic.php?pid\x1170583
> 
> Basically, when I probe i915 gmbus panel (i2c-2) with sensors-detect, when
> it probes address 0x4f my monitor settings (I think the sharpness mostly)
> are changed drastically.  This is persistent over any operating system and
> reboot, leading me to believe a) its a hardware issue, or b) nothing
> usually touches whatever values were changed.  Any help would be greatly
> appreciated, in the forum post I've also included some experimentation with
> i2c-tools.  The problem originally occured on an Arch Linux system.  Let me
> know if you need anything else!
> 
Sensors-detect should tell you:
"Lastly, we can probe the I2C/SMBus adapters for connected hardware
monitoring devices. This is the most risky part, and while it works
reasonably well on most systems, it has been reported to cause trouble
on some systems."

Apparently it does cause trouble on your system. Problem is that some I2C
devcies interpret I2C read requests as command to do something ... in your case,
it looks like some device interprets a read on address 0x4f as request to change
some monitor settings.

Guenter

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

Re: [lm-sensors] sensors-detect changes monitor settings on hp pavilion dv7

[Jean Delvare]

Hi Andrew,

On Thu, 4 Oct 2012 11:30:45 -0400, Andrew Potter wrote:
> Hi, I wasn't quite sure from the information on lm-sensors.org how to open
> an official ticket, but the details of my problem are here:
> 
> https://bbs.archlinux.org/viewtopic.php?pid\x1170583
> 
> Basically, when I probe i915 gmbus panel (i2c-2) with sensors-detect, when
> it probes address 0x4f my monitor settings (I think the sharpness mostly)
> are changed drastically.  This is persistent over any operating system and
> reboot, leading me to believe a) its a hardware issue, or b) nothing
> usually touches whatever values were changed.  Any help would be greatly
> appreciated, in the forum post I've also included some experimentation with
> i2c-tools.  The problem originally occured on an Arch Linux system.  Let me
> know if you need anything else!

Thanks for reporting. We have had that kind of problem popping up every
other year or so, and take such reports very seriously. We've done our
best to make sensors-detect as safe as possible, but the fact is, the
I2C and SMBus protocols do not support device probing in general, so
the whole thing is a hack, and hacks can turn bad sometimes.

Please keep in mind that everything you will do to investigate and
solve this issue (including things I or others may suggest you do) may
make things even worse than they are today. So it's all up to you what
you want to do and when you prefer to stop and stick with what you have.

First thing I would like to know is the version of lm-sensors (or
sensors-detect) you have been running.

Second thing I would like to know is the exact revision of your laptop
HP Pavillon DV7. If possible, provide the dmidecode output. We want to
document the problem and prevent it from happening to other users with
the same hardware, using a blacklist if needed.

One thing I don't quite understand is how there can be two different
graphics chips (one Intel, one Nvidia) in a laptop. What sense does it
make, and which one are you using? If the Intel is an integrated thing
and the Nvidia chip was added and is supposed to be better, then I
would suppose that's what you're using, but then I do not understand
how messing up with the Intel I2C buses can have any adverse effect.

It might be a good idea to prevent scanning of I2C buses on graphics
cards by default, as we do for multimedia (TV) cards already. These
days drm drivers should take care of instantiating the hardware
monitoring devices present without user-space probing, so it's probably
doing more harm than good.

You said you ran i2cdump on the chip, did you save the output? I'd like
to take a look.

The only chance to get your system back to its original state by
software is to know what chips exactly are on the I2C bus of your Intel
graphics chip and in particular at I2C address 0x4a. Unfortunately this
information is usually very hard to get.

Another thing you should try is:
* Power down the machine.
* Remove the battery, unplug the AC adapter.
* Wait 1 minute at least, 10 minutes if you have time.
* Put the battery and AC adapter back in place.
* Power up the machine again.

With some luck this will restore the original state. It has worked for
others in the past.

-- 
Jean Delvare

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

Re: [lm-sensors] sensors-detect changes monitor settings on hp pavilion dv7

[Guenter Roeck]

On Thu, Oct 04, 2012 at 06:30:11PM +0200, Jean Delvare wrote:
> Hi Andrew,
> 
> On Thu, 4 Oct 2012 11:30:45 -0400, Andrew Potter wrote:
> > Hi, I wasn't quite sure from the information on lm-sensors.org how to open
> > an official ticket, but the details of my problem are here:
> > 
> > https://bbs.archlinux.org/viewtopic.php?pid\x1170583
> > 
> > Basically, when I probe i915 gmbus panel (i2c-2) with sensors-detect, when
> > it probes address 0x4f my monitor settings (I think the sharpness mostly)
> > are changed drastically.  This is persistent over any operating system and
> > reboot, leading me to believe a) its a hardware issue, or b) nothing
> > usually touches whatever values were changed.  Any help would be greatly
> > appreciated, in the forum post I've also included some experimentation with
> > i2c-tools.  The problem originally occured on an Arch Linux system.  Let me
> > know if you need anything else!
> 
> Thanks for reporting. We have had that kind of problem popping up every
> other year or so, and take such reports very seriously. We've done our
> best to make sensors-detect as safe as possible, but the fact is, the
> I2C and SMBus protocols do not support device probing in general, so
> the whole thing is a hack, and hacks can turn bad sometimes.
> 
> Please keep in mind that everything you will do to investigate and
> solve this issue (including things I or others may suggest you do) may
> make things even worse than they are today. So it's all up to you what
> you want to do and when you prefer to stop and stick with what you have.
> 
> First thing I would like to know is the version of lm-sensors (or
> sensors-detect) you have been running.
> 
> Second thing I would like to know is the exact revision of your laptop
> HP Pavillon DV7. If possible, provide the dmidecode output. We want to
> document the problem and prevent it from happening to other users with
> the same hardware, using a blacklist if needed.
> 
> One thing I don't quite understand is how there can be two different
> graphics chips (one Intel, one Nvidia) in a laptop. What sense does it
> make, and which one are you using? If the Intel is an integrated thing
> and the Nvidia chip was added and is supposed to be better, then I
> would suppose that's what you're using, but then I do not understand
> how messing up with the Intel I2C buses can have any adverse effect.
> 

I think I know the answer to that one: the laptop uses the Intel CPU graphics
for most applications to consserve power, but can switch to the Nvidia chip
for high-demand graphics applications such as games.

Guenter

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

Re: [lm-sensors] sensors-detect changes monitor settings on hp pavilion dv7

[Jean Delvare]

On Thu, 4 Oct 2012 18:30:11 +0200, Jean Delvare wrote:
> Hi Andrew,
> 
> On Thu, 4 Oct 2012 11:30:45 -0400, Andrew Potter wrote:
> > Hi, I wasn't quite sure from the information on lm-sensors.org how to open
> > an official ticket, but the details of my problem are here:
> > 
> > https://bbs.archlinux.org/viewtopic.php?pid\x1170583
> > 
> > Basically, when I probe i915 gmbus panel (i2c-2) with sensors-detect, when
> > it probes address 0x4f my monitor settings (I think the sharpness mostly)
> > are changed drastically.  This is persistent over any operating system and
> > reboot, leading me to believe a) its a hardware issue, or b) nothing
> > usually touches whatever values were changed.  Any help would be greatly
> > appreciated, in the forum post I've also included some experimentation with
> > i2c-tools.  The problem originally occured on an Arch Linux system.  Let me
> > know if you need anything else!
> 
> Thanks for reporting. We have had that kind of problem popping up every
> other year or so, and take such reports very seriously. We've done our
> best to make sensors-detect as safe as possible, but the fact is, the
> I2C and SMBus protocols do not support device probing in general, so
> the whole thing is a hack, and hacks can turn bad sometimes.
> 
> Please keep in mind that everything you will do to investigate and
> solve this issue (including things I or others may suggest you do) may
> make things even worse than they are today. So it's all up to you what
> you want to do and when you prefer to stop and stick with what you have.
> 
> First thing I would like to know is the version of lm-sensors (or
> sensors-detect) you have been running.
> 
> Second thing I would like to know is the exact revision of your laptop
> HP Pavillon DV7. If possible, provide the dmidecode output. We want to
> document the problem and prevent it from happening to other users with
> the same hardware, using a blacklist if needed.
> 
> One thing I don't quite understand is how there can be two different
> graphics chips (one Intel, one Nvidia) in a laptop. What sense does it
> make, and which one are you using? If the Intel is an integrated thing
> and the Nvidia chip was added and is supposed to be better, then I
> would suppose that's what you're using, but then I do not understand
> how messing up with the Intel I2C buses can have any adverse effect.
> 
> It might be a good idea to prevent scanning of I2C buses on graphics
> cards by default, as we do for multimedia (TV) cards already. These
> days drm drivers should take care of instantiating the hardware
> monitoring devices present without user-space probing, so it's probably
> doing more harm than good.
> 
> You said you ran i2cdump on the chip, did you save the output? I'd like
> to take a look.
> 
> The only chance to get your system back to its original state by
> software is to know what chips exactly are on the I2C bus of your Intel
> graphics chip and in particular at I2C address 0x4a. Unfortunately this
> information is usually very hard to get.
> 
> Another thing you should try is:
> * Power down the machine.
> * Remove the battery, unplug the AC adapter.
> * Wait 1 minute at least, 10 minutes if you have time.
> * Put the battery and AC adapter back in place.
> * Power up the machine again.
> 
> With some luck this will restore the original state. It has worked for
> others in the past.

Andrew, were you able to solve the problem somehow, using my suggestion
or otherwise?

-- 
Jean Delvare

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