Re: Recent linux-image update broke CPU fan

[Neal P. Murphy]

On Tue, 1 Dec 2015 21:12:29 -0500
"Neal P. Murphy"  wrote:

> On Sun, 29 Nov 2015 17:18:51 -0500
> "Neal P. Murphy"  wrote:
> 
> > On Sun, 29 Nov 2015 14:35:15 +0100
> > Sven Arvidsson  wrote:
> > 
> > > On Sun, 2015-11-29 at 02:16 -0500, Neal P. Murphy wrote:
> > > > I think the last linux-image update broke my CPU fan. OK, it didn't
> > > > actually break it. But since the last update, my CPU fan (stock with
> > > > vishera 8350 black) will NOT turn faster than about 3500 RPM.
> > > > 
> > > > I have been using fancontrol for months. It took weeks to get it set
> > > > correctly. At maximum speed, I've seen (and heard) the fan turn in
> > > > excess of 6500 RPM.
> > > > 
> > > > Until the recent kernel update, the CPU and case fans have ramped up
> > > > with increasing temp and ramped down with decreasing temp. The CPU
> > > > never exceeded about 60C when compiling linux using all 8 CPUs for 5
> > > > minutes. And that was with a mild overclock (4.4GHz, 1866 RAM at
> > > > 2133).
> > > > 
> > > > Now, the CPU fan will not exceed about 3700 RPM. That's barely fast
> > > > enough to keep 4 non-overclocked CPUs cool.
> > > > 
> > > > Has anyone else experienced this? Is it related to the kernel update?
> > > > Or do I have a power supply or motherboard failing (again)?
> > > 
> > > If this was on a stable system it seems odd, if it was unstable or
> > > testing it might just be one of the "charms" of running something that
> > > is in development.
> 
> It's still odd. But with no change to the OS, the fan speed has returned. So 
> I'm going to conclude that it is related to hardware and/or firmware (flaky 
> NVRAM, flaky power supply, flaky mainboard, ), that the kernel update was 
> not the cause.

Following up on this, it was almost certainly the power supply. Turns out the 
supplies I had on hand each had two 20A (or 22A) 12VDC rails. The 12VDC was 
dipping during periods of 'high load' (meaning the current draw exceeded the 
rail's ability to supply it). Modern 'high power' CPUs often need a lot more 
than that. I bought a Corsair RM750X supply with a single 625A 12VDC rail. 
Overkill, but it was only $60 after rebates. The Smoothwall Express build time 
remained at around 90 minutes (a little less for 32-bit, a little more for 
64-bit); Linux takes less than 8 minutes to build. The fans work correctly 
again.

All is well now. The CPU fan sometimes gets up to about 7000 RPM (when the room 
temp is high). The CPU stays under 60C (but will rise slightly higher than that 
if the room is over 80-85°F. The 12VDC is rock steady. It took me a while to 
retune /etc/fancontrol to keep the fans silent when the system is mostly 
quiescent.

Debian was not at fault.

N

Re: Recent linux-image update broke CPU fan

[Neal P. Murphy]

On Sun, 29 Nov 2015 17:18:51 -0500
"Neal P. Murphy"  wrote:

> On Sun, 29 Nov 2015 14:35:15 +0100
> Sven Arvidsson  wrote:
> 
> > On Sun, 2015-11-29 at 02:16 -0500, Neal P. Murphy wrote:
> > > I think the last linux-image update broke my CPU fan. OK, it didn't
> > > actually break it. But since the last update, my CPU fan (stock with
> > > vishera 8350 black) will NOT turn faster than about 3500 RPM.
> > > 
> > > I have been using fancontrol for months. It took weeks to get it set
> > > correctly. At maximum speed, I've seen (and heard) the fan turn in
> > > excess of 6500 RPM.
> > > 
> > > Until the recent kernel update, the CPU and case fans have ramped up
> > > with increasing temp and ramped down with decreasing temp. The CPU
> > > never exceeded about 60C when compiling linux using all 8 CPUs for 5
> > > minutes. And that was with a mild overclock (4.4GHz, 1866 RAM at
> > > 2133).
> > > 
> > > Now, the CPU fan will not exceed about 3700 RPM. That's barely fast
> > > enough to keep 4 non-overclocked CPUs cool.
> > > 
> > > Has anyone else experienced this? Is it related to the kernel update?
> > > Or do I have a power supply or motherboard failing (again)?
> > 
> > If this was on a stable system it seems odd, if it was unstable or
> > testing it might just be one of the "charms" of running something that
> > is in development.

It's still odd. But with no change to the OS, the fan speed has returned. So 
I'm going to conclude that it is related to hardware and/or firmware (flaky 
NVRAM, flaky power supply, flaky mainboard, ), that the kernel update was 
not the cause.

Thanks,
N

Re: Recent linux-image update broke CPU fan

[Sven Arvidsson]

On Sun, 2015-11-29 at 17:18 -0500, Neal P. Murphy wrote:
> > If this was on a stable system it seems odd, if it was unstable or
> > testing it might just be one of the "charms" of running something
> > that
> > is in development.
> 
> Jessie: stable.

That is odd, but probably easier to track down on a stable system (if
it is a software problem)

> As of 28 November 2015, the latest update to linux-image-3.16.0-4
> -amd64 versus the previous update. I think the latest update is:
> linux-image-3.16.0-4-amd64_3.16.7-ckt11-1+deb8u6_amd64.deb
> and the previous is probably ...deb8u5 So how do I tell it to
> uninstall 'u6' and install 'u5'?

There's nothing in the changelog that indicates any changes when it
comes to fans, at least as far as I can tell.

You should be able to just grab the older package from
snapshot.debian.net and manually install (downgrade) it. 

http://snapshot.debian.org/binary/linux-image-3.16.0-4-amd64/

> I can live with "It's a coincidence." A lot of things can change; the
> obvious one was the recent kernel update, thus my query. If the
> latest changes/patches to 3.16 are well away from sensors and
> drivers, then the kernel update probably is not the cause and I
> should look elsewhere (like checking the capacitors on the M5A99FX
> mainboard; there are probably still some bad caps floating around.)

Yes, trying earlier kernels is probably a good idea, and possibly other
parts that might have changed, libsensor, lm-sensor? Or what it is that
fancontrol uses. 

-- 
Cheers,
Sven Arvidsson
http://www.whiz.se
PGP Key ID 6FAB5CD5





signature.asc
Description: This is a digitally signed message part

Re: Recent linux-image update broke CPU fan

[Sven Arvidsson]

On Mon, 2015-11-30 at 12:10 +, Lisi Reisz wrote:
> On Sunday 29 November 2015 22:18:51 Neal P. Murphy wrote:
> > As of 28 November 2015, the latest update to linux-image-3.16.0-4
> -amd64
> > versus the previous update. I think the latest update is:
> > linux-image-3.16.0-4-amd64_3.16.7-ckt11-1+deb8u6_amd64.deb
> > and the previous is probably ...deb8u5 So how do I tell it to
> uninstall
> > 'u6' and install 'u5'?
> 
> Surely you have not uninstalled the earlier one?  So just ask GRUB to
> change 
> the default to the older kernel.

That's an update to kernel package, so not a new kernel, are you sure
those can be installed in parallel?

-- 
Cheers,
Sven Arvidsson
http://www.whiz.se
PGP Key ID 6FAB5CD5





signature.asc
Description: This is a digitally signed message part

Re: Recent linux-image update broke CPU fan

[John Hasler]

Lisi writes:
> You've slightly lost me - when is a new kernel a new kernel and when
> is it an update?

I think it's a "new kernel" when the Debian maintainers package a new
upstream release.  It's an "update" when they repackage an existing
kernel to include some sort of change (to the kernel or its packaging)
that they made.  It's all the same to the package management system,
though.  To it if they have different version numbers they are different
kernels and you can have as many installed as you want.
-- 
John Hasler 
jhas...@newsguy.com
Elmwood, WI USA

Re: Recent linux-image update broke CPU fan

[Neal P. Murphy]

On Mon, 30 Nov 2015 18:31:46 +
Lisi Reisz  wrote:

> On Monday 30 November 2015 15:02:01 Sven Arvidsson wrote:
> > On Mon, 2015-11-30 at 12:10 +, Lisi Reisz wrote:
> > > On Sunday 29 November 2015 22:18:51 Neal P. Murphy wrote:
> > > > As of 28 November 2015, the latest update to linux-image-3.16.0-4
> > >
> > > -amd64
> > >
> > > > versus the previous update. I think the latest update is:
> > > > linux-image-3.16.0-4-amd64_3.16.7-ckt11-1+deb8u6_amd64.deb
> > > > and the previous is probably ...deb8u5 So how do I tell it to
> > >
> > > uninstall
> > >
> > > > 'u6' and install 'u5'?
> > >
> > > Surely you have not uninstalled the earlier one?  So just ask GRUB to
> > > change
> > > the default to the older kernel.
> >
> > That's an update to kernel package, so not a new kernel, are you sure
> > those can be installed in parallel?
> 
> You've slightly lost me - when is a new kernel a new kernel and when is it an 
> update?  I have:

I suspect that when it appears as a 'new package', it's a new kernel and will 
have a new entry in grub. That's usually when the kernel minor version is 
bumped. Or when one installs a kernel from backports, testing, or unstable.

But security updates do not count as new pkgs. Thus the kernel is merely 
'replaced'. In fact, the previous updates (1-5) are no longer cached and I 
can't reinstall them.

In the case of stock Jessie, the kernel pkg is (pretty much) 
linux-image-3.16.0-4-amd64; that part does not change for security updates.

N

Re: Recent linux-image update broke CPU fan

[Lisi Reisz]

On Monday 30 November 2015 15:02:01 Sven Arvidsson wrote:
> On Mon, 2015-11-30 at 12:10 +, Lisi Reisz wrote:
> > On Sunday 29 November 2015 22:18:51 Neal P. Murphy wrote:
> > > As of 28 November 2015, the latest update to linux-image-3.16.0-4
> >
> > -amd64
> >
> > > versus the previous update. I think the latest update is:
> > > linux-image-3.16.0-4-amd64_3.16.7-ckt11-1+deb8u6_amd64.deb
> > > and the previous is probably ...deb8u5 So how do I tell it to
> >
> > uninstall
> >
> > > 'u6' and install 'u5'?
> >
> > Surely you have not uninstalled the earlier one?  So just ask GRUB to
> > change
> > the default to the older kernel.
>
> That's an update to kernel package, so not a new kernel, are you sure
> those can be installed in parallel?

You've slightly lost me - when is a new kernel a new kernel and when is it an 
update?  I have:

Linux 3.16-0.bpo.3-amd64
and
Linux 3.16-0.bpo.4-amd64

I would have said that the second was an update of the first, and not a new 
kernel, but I am open to correction.

I had others but I have purged them.

I have other kernels, but only those two 3.16-0.bpo.x

Lisi

Re: Recent linux-image update broke CPU fan

[Lisi Reisz]

On Sunday 29 November 2015 22:18:51 Neal P. Murphy wrote:
> As of 28 November 2015, the latest update to linux-image-3.16.0-4-amd64
> versus the previous update. I think the latest update is:
> linux-image-3.16.0-4-amd64_3.16.7-ckt11-1+deb8u6_amd64.deb
> and the previous is probably ...deb8u5 So how do I tell it to uninstall
> 'u6' and install 'u5'?

Surely you have not uninstalled the earlier one?  So just ask GRUB to change 
the default to the older kernel.

I could have told you how on GRUB1.  Someone else will have to tell you on 
GRUB2 - but you can choose your kernel manually at boot time as soon as the 
GRUB splash screen appears.

Lisi

Re: Recent linux-image update broke CPU fan

[Neal P. Murphy]

On Sun, 29 Nov 2015 14:35:15 +0100
Sven Arvidsson  wrote:

> On Sun, 2015-11-29 at 02:16 -0500, Neal P. Murphy wrote:
> > I think the last linux-image update broke my CPU fan. OK, it didn't
> > actually break it. But since the last update, my CPU fan (stock with
> > vishera 8350 black) will NOT turn faster than about 3500 RPM.
> > 
> > I have been using fancontrol for months. It took weeks to get it set
> > correctly. At maximum speed, I've seen (and heard) the fan turn in
> > excess of 6500 RPM.
> > 
> > Until the recent kernel update, the CPU and case fans have ramped up
> > with increasing temp and ramped down with decreasing temp. The CPU
> > never exceeded about 60C when compiling linux using all 8 CPUs for 5
> > minutes. And that was with a mild overclock (4.4GHz, 1866 RAM at
> > 2133).
> > 
> > Now, the CPU fan will not exceed about 3700 RPM. That's barely fast
> > enough to keep 4 non-overclocked CPUs cool.
> > 
> > Has anyone else experienced this? Is it related to the kernel update?
> > Or do I have a power supply or motherboard failing (again)?
> 
> If this was on a stable system it seems odd, if it was unstable or
> testing it might just be one of the "charms" of running something that
> is in development.

Jessie: stable.

> 
> You need to figure out what package is responsible for the change, boot
> the previous kernel for example. Then figure out if it's a bug or just
> changed behaviour.

As of 28 November 2015, the latest update to linux-image-3.16.0-4-amd64 versus 
the previous update. I think the latest update is:
linux-image-3.16.0-4-amd64_3.16.7-ckt11-1+deb8u6_amd64.deb
and the previous is probably ...deb8u5 So how do I tell it to uninstall 
'u6' and install 'u5'?

> 
> I'm not sure how fancontrol works. What exactly broke? Do you not get
> readings from the temperature sensors, or is it not possible to control
> the fan speed? 

Exactly: the CPU fan no longer exceeds about 3700 RPM. It is capable of 
exceeding 6000 RPM. Around 4500-5000 RPM is needed to keep the CPU from 
exceeding about 60C.

Fancontrol uses the interfaces in /sys/class/hwmon (in my case, the IT87 in 
.../hwmon/hwmon0/pwm1*). Fancontrol writes "0" to "255" to pwm[1-3]. "0" is the 
slowest allowed RPM; at "0", the CPU fan slows to around 1300 RPM and the case 
fans stop. "255" is the fastest RPM; this should have the CPU fan screaming 
painfully loud while the case fans top out at around 1300 RPM. But the CPU fan 
now refuses to go faster than around 3700 RPM, even with clocking restored to 
stock timings.

The sensors work and report sane RPM and temp readings; they are generally 
where I expect them to be with respect to CPU load. But the CPU fan just won't 
go any faster.

Thus my query: did the latest linux-image update do anything with sensors 
and/or sensor drivers? That is, mayhap the IT87 driver changed to scale the 
pwm1 input? (No, it doesn't make sense, but strange things happen with software 
now and again.)

I can live with "It's a coincidence." A lot of things can change; the obvious 
one was the recent kernel update, thus my query. If the latest changes/patches 
to 3.16 are well away from sensors and drivers, then the kernel update probably 
is not the cause and I should look elsewhere (like checking the capacitors on 
the M5A99FX mainboard; there are probably still some bad caps floating around.)

Thanks,
Neal

Re: Recent linux-image update broke CPU fan

[Sven Arvidsson]

On Sun, 2015-11-29 at 02:16 -0500, Neal P. Murphy wrote:
> I think the last linux-image update broke my CPU fan. OK, it didn't
> actually break it. But since the last update, my CPU fan (stock with
> vishera 8350 black) will NOT turn faster than about 3500 RPM.
> 
> I have been using fancontrol for months. It took weeks to get it set
> correctly. At maximum speed, I've seen (and heard) the fan turn in
> excess of 6500 RPM.
> 
> Until the recent kernel update, the CPU and case fans have ramped up
> with increasing temp and ramped down with decreasing temp. The CPU
> never exceeded about 60C when compiling linux using all 8 CPUs for 5
> minutes. And that was with a mild overclock (4.4GHz, 1866 RAM at
> 2133).
> 
> Now, the CPU fan will not exceed about 3700 RPM. That's barely fast
> enough to keep 4 non-overclocked CPUs cool.
> 
> Has anyone else experienced this? Is it related to the kernel update?
> Or do I have a power supply or motherboard failing (again)?

If this was on a stable system it seems odd, if it was unstable or
testing it might just be one of the "charms" of running something that
is in development.

You need to figure out what package is responsible for the change, boot
the previous kernel for example. Then figure out if it's a bug or just
changed behaviour.

I'm not sure how fancontrol works. What exactly broke? Do you not get
readings from the temperature sensors, or is it not possible to control
the fan speed? 

-- 
Cheers,
Sven Arvidsson
http://www.whiz.se
PGP Key ID 6FAB5CD5





signature.asc
Description: This is a digitally signed message part

Re: Recent linux-image update broke CPU fan

[Pascal Hambourg]

Neal P. Murphy a écrit :
> I think the last linux-image update broke my CPU fan.

It may help to mention which are the previous and current versions.

Recent linux-image update broke CPU fan

[Neal P. Murphy]

I think the last linux-image update broke my CPU fan. OK, it didn't actually 
break it. But since the last update, my CPU fan (stock with vishera 8350 black) 
will NOT turn faster than about 3500 RPM.

I have been using fancontrol for months. It took weeks to get it set correctly. 
At maximum speed, I've seen (and heard) the fan turn in excess of 6500 RPM.

Until the recent kernel update, the CPU and case fans have ramped up with 
increasing temp and ramped down with decreasing temp. The CPU never exceeded 
about 60C when compiling linux using all 8 CPUs for 5 minutes. And that was 
with a mild overclock (4.4GHz, 1866 RAM at 2133).

Now, the CPU fan will not exceed about 3700 RPM. That's barely fast enough to 
keep 4 non-overclocked CPUs cool.

Has anyone else experienced this? Is it related to the kernel update? Or do I 
have a power supply or motherboard failing (again)?

Thanks,
Neal