Nouveau - Jun 2014 - [Bug 79820] New: Strange readout on fan1_input when pwm1_min < 33

https://bugs.freedesktop.org/show_bug.cgi?id=79820

          Priority: medium
            Bug ID: 79820
          Assignee: nouveau at lists.freedesktop.org
           Summary: Strange readout on fan1_input when pwm1_min < 33
        QA Contact: xorg-team at lists.x.org
          Severity: normal
    Classification: Unclassified
                OS: Linux (All)
          Reporter: freedesk.apriori at spamgourmet.com
          Hardware: x86-64 (AMD64)
            Status: NEW
           Version: unspecified
         Component: Driver/nouveau
           Product: xorg

Created attachment 100705
  --> https://bugs.freedesktop.org/attachment.cgi?id=100705&action=edit
jountalctl -b -k|grep nouveau

When running the drivers from kernel 3.15.0, pwm1_min seems to default to 35. I
noticed that this yields reported ~2k RPM in fan1_input for my card. Since the
fan runs at ~1k RPM in idle when running the NVIDIA blob, I tried lowering
pwm1_min and noticed odd results.

Lowering pwm1_min to 32 is fine and gives me a readout of ~1.8k RPM. However,
after that, I see weird output values:

31 ~ 4.5k
30 ~ 9k
29 ~ 5k
28 ~ 22k (sic)
27 ~ 75k (sic)
26 ~ 400k (sic)

At no time was my computer lifting up from the ground though, and even with the
reported 4.5k it seems that the fan did not actually speed up.

Additional background: It seems my fan is only supported as of Linux 3.15.0.
Before that, the fan ran at 100% and dmesg showed ?FAN control: none /
external?. On IRC I was informed that a new fan has been added to 3.15, and
indeed it seems to be recognized now (many thanks for that!).

I?m attaching my dmesg and my vbios in case it helps. Note though that
nvagetbios reported a warning when dumping the BIOS:
> No extraction method specified (using -s extraction_method). Defaulting to
PRAMIN.
> Attempt to extract the vbios from card 0 (nve6) using PRAMIN
> Invalid signature(0x55aa). You may want to try another retrieval method.
Let me know if there?s additional information I could supply.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20140609/1edacfec/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=79820

--- Comment #1 from freedesk.apriori at spamgourmet.com ---
Created attachment 100707
  --> https://bugs.freedesktop.org/attachment.cgi?id=100707&action=edit
vbios

Extracted by nvagetbios, which reported these warnings:
> No extraction method specified (using -s extraction_method). Defaulting to
PRAMIN.
> Attempt to extract the vbios from card 0 (nve6) using PRAMIN
> Invalid signature(0x55aa). You may want to try another retrieval method.
-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20140609/ce4305ad/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=79820

Ilia Mirkin <imirkin at alum.mit.edu> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|Strange readout on          |[NVE6] Strange readout on
                   |fan1_input when pwm1_min <  |fan1_input when pwm1_min
<
                   |33                          |33

--- Comment #2 from Ilia Mirkin <imirkin at alum.mit.edu> ---
With nouveau loaded, can you grab vbios from /sys/kernel/debug/dri/0/vbios.rom?
Or use -s prom with nvagetbios?

It's moderately likely that the reported fan rpm is just 2x what it should
be
-- did the fan sound louder with nouveau than with the blob? See
http://cgit.freedesktop.org/~darktama/nouveau/commit/?id=e3601d0869a18acc19e58f2d1216b810632d75f0
(which will be in 3.16-rc1).

Martin -- any idea about the weird RPMs reported for the lower values?

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20140609/ee3a6bce/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=79820

--- Comment #3 from freedesk.apriori at spamgourmet.com ---
The noise didn?t really noticeably increase. And I just checked the values
reported by nvidia-settings. When idle, these always seem to jump in increments
of 30 RPM (usually between 1050 and 1080). With nouveau, the output value jumps
in increments of 60 RPM, so you are probably right and the difference is only
nominal.

In any case, I?m attaching the BIOS file obtained from nvagetbios -s prom. No
warnings were reported this time.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20140609/c7abe7c7/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=79820

freedesk.apriori at spamgourmet.com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
 Attachment #100707|0                           |1
        is obsolete|                            |

--- Comment #4 from freedesk.apriori at spamgourmet.com ---
Created attachment 100758
  --> https://bugs.freedesktop.org/attachment.cgi?id=100758&action=edit
VBIOS

BIOS obtained by nvagetbios -s prom.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20140609/42b1f80f/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=79820

--- Comment #5 from Martin Peres <martin.peres at ensi-bourges.fr> ---
(In reply to comment #2)
> With nouveau loaded, can you grab vbios from
> /sys/kernel/debug/dri/0/vbios.rom? Or use -s prom with nvagetbios?
I just pushed a patch to nvagetbios to suggest to use -s prom when the
signature is invalid. Let's stop wasting time suggesting this.
> 
> It's moderately likely that the reported fan rpm is just 2x what it
should
> be -- did the fan sound louder with nouveau than with the blob? See
> http://cgit.freedesktop.org/~darktama/nouveau/commit/
> ?id=e3601d0869a18acc19e58f2d1216b810632d75f0 (which will be in 3.16-rc1).
Yes, this is likely the problem here.
> 
> Martin -- any idea about the weird RPMs reported for the lower values?
It probably is a bouncing problem of the Hall effect sensor. This one may not
have a proper hysteresis cycle which leads to an unstable signal during the
transition phases. The longer the transition takes, the more un-wanted
oscillation is introduced. It is possible that the fan manufacturer added a
low-pass filter on the fan that only works down to the minimum speed (35%) and
fails to clean up the signal after that.

In any case, this is not our bug unless nvidia is able to cope with it (which
is impossible to test with the coolbits since they don't allow a fan speed
lower than the minimum speed). I could edit the vbios to change the minimum
speed limit and ask you to re-upload it and check if nvidia yields the same
result. Would you be interested in doing that?

(In reply to comment #3)
> The noise didn?t really noticeably increase. And I just checked the values
> reported by nvidia-settings. When idle, these always seem to jump in
> increments of 30 RPM (usually between 1050 and 1080). With nouveau, the
> output value jumps in increments of 60 RPM, so you are probably right and
> the difference is only nominal.
Perfect! Thank you :)
> 
> In any case, I?m attaching the BIOS file obtained from nvagetbios -s prom.
> No warnings were reported this time.
Thanks, but I don't think it will be necessary...

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20140609/9b2b9f17/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=79820

Ilia Mirkin <imirkin at alum.mit.edu> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED

--- Comment #6 from Ilia Mirkin <imirkin at alum.mit.edu> ---
The reported RPM being 2x should be fixed in 3.16. Lowering the fan outside of
VBIOS-supplied values isn't explicitly supported, and the fact that
you're
getting junk RPM readings from that isn't unexpected.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20141120/1728eb62/attachment.html>