Nouveau - Feb 2013 - [Bug 60704] New: [nouveau, git regression] - I2C PWM fan control broken on nv50 adt7475 on kernels 3.3.x+

https://bugs.freedesktop.org/show_bug.cgi?id=60704

          Priority: medium
            Bug ID: 60704
          Assignee: nouveau at lists.freedesktop.org
           Summary: [nouveau, git regression] - I2C PWM fan control broken
                    on nv50 adt7475 on kernels 3.3.x+
        QA Contact: xorg-team at lists.x.org
          Severity: major
    Classification: Unclassified
                OS: Linux (All)
          Reporter: m.meledandri at gmail.com
          Hardware: x86-64 (AMD64)
            Status: NEW
           Version: git
         Component: Driver/nouveau
           Product: xorg

Created attachment 74657
  --> https://bugs.freedesktop.org/attachment.cgi?id=74657&action=edit
ic2-bit-banging-nouveau-git-bisect.log

When resuming from suspend on affected kernels the lack of fan pwm control
causes the GPU fan to run at 100%. This is noisy, annoying and inefficient.

After performing a git bisect (attached), I've determined the commit that
broke
pwm fan control functionality for all kernels 3.3.x+ with adt7475 controller
(nv50 GPU in my case):

commit f553b79c03f0dbd52f6f03abe8233a2bef8cbd0d
Author: Ben Skeggs <bskeggs at redhat.com>
Date:   Wed Dec 21 18:09:12 2011 +1000

drm/nouveau/i2c: handle bit-banging ourselves

Along with the many kernels I built and tested during bisection, I've also
verified this broke sensors detecting on mainline builds of 3.3.8, 3.4.29,
3.7.3 and the current Nouveau git master.

I can verify proper PWM fan control/i2c enumeration on kernel 3.2.35 and
3.2.38. I've not tested earlier kernels than the 3.2 series, but git
bisection
went back to 3.2-rc1 and was working there.

I'm happy to provide any further testing to help expedite getting this
patched
for stable kernels. I'm particularly interested in the 3.4.x LTS series
which
is what led me to discover this issue on my system after I strayed from the
happy confines of the Debian 3.2.x LTS series kernel.

-- 
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/20130211/b5291f6d/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

--- Comment #1 from Marc Meledandri <m.meledandri at gmail.com> ---
Created attachment 74658
  --> https://bugs.freedesktop.org/attachment.cgi?id=74658&action=edit
properly-functioning-sensors-modules.log

-- 
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/20130211/165a17ce/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

--- Comment #2 from Marc Meledandri <m.meledandri at gmail.com> ---
Created attachment 74659
  --> https://bugs.freedesktop.org/attachment.cgi?id=74659&action=edit
failed-dmesg.log

-- 
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/20130211/1e35d407/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

--- Comment #3 from Marc Meledandri <m.meledandri at gmail.com> ---
Created attachment 74660
  --> https://bugs.freedesktop.org/attachment.cgi?id=74660&action=edit
video-bios-gtx-260-nv50.rom

-- 
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/20130211/7960ead3/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

Marc Meledandri <m.meledandri at gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #74660|text/plain                  |application/octet-stream
          mime type|                            |

-- 
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/20130211/102a7790/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

--- Comment #4 from Marc Meledandri <m.meledandri at gmail.com> ---
Created attachment 74661
  --> https://bugs.freedesktop.org/attachment.cgi?id=74661&action=edit
git-master-NAK-Bailout-dmesg.log

With kernels 3.4.29, 3.7.3 and the nouveau git master observing:

i2c i2c-4: sendbytes: NAK bailout.

messages in syslog.

-- 
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/20130212/b9b9dbe8/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

--- Comment #5 from Vincent Pelletier <plr.vincent at gmail.com> ---
I looked into this issue a year ago, reported the breakage, bisect result and
my attempts to fix in this thread:
http://lists.freedesktop.org/archives/nouveau/2012-January/009779.html

-- 
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/20130212/350742b3/attachment-0001.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

--- Comment #6 from Vincent Pelletier <plr.vincent at gmail.com> ---
The hash in my mail doesn't seem to exist anymore. I believe this is because
the branch was since rebased after a merge to linus' tree, and I'm quite
sure
it was the commit you bisected to.

-- 
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/20130212/c5bd84da/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

--- Comment #7 from Emil Velikov <emil.l.velikov at gmail.com> ---
Created attachment 74708
  --> https://bugs.freedesktop.org/attachment.cgi?id=74708&action=edit
Use 25kbps datarate for i2c transfers

The offending commit has been almost completely reverted later on

The only differences being
* timeout (5000 > 2200)
Data is based on the VESA spec
* bitrate (12,5kbps > 50kbps)
Bitrate of 50kbps is quite reasonable, considering the standards

There are a few possible solutions
A select correct delay depending on the i2c port type
B presence of adt7475(and other) i2c sensor(s) in the extdev vbios table
C use 25kbps for all transfers
D quirk it

Patch for option C is attached

-- 
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/20130212/9eeb282f/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

--- Comment #8 from Marc Meledandri <m.meledandri at gmail.com> ---
Thanks Emil! I confirm that your patch resolves the issue for my hardware.

I can now control my GPU fan with the PWM bits. This also solves the
suspend/resume case with automatic fan control enabled.

-- 
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/20130213/3e1cf6c5/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

--- Comment #9 from Marc Meledandri <m.meledandri at gmail.com> ---
Created attachment 74749
  --> https://bugs.freedesktop.org/attachment.cgi?id=74749&action=edit
i2c datarate transfer patch ported upstream

I built Emil's patch:  Use 25kbps datarate for i2c transfers on kernel
against
kernel 3.4.30.

To test this fix upstream on kernels 3.7.7 and 3.8-rc7, I've attached the
same
fix ported to the new file/directory structure as the code has been rebased and
is no longer in the same file or location.

I hope I documented properly to credit the original author!

-- 
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/20130213/1bed4cd8/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

Marc Meledandri <m.meledandri at gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #74749|0                           |1
           is patch|                            |

-- 
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/20130213/6ae82348/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

--- Comment #10 from Martin Peres <martin.peres at ensi-bourges.fr> ---
Marc: Has this patch been applied yet?

-- 
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/20130615/c48f1488/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

--- Comment #11 from Marc Meledandri <m.meledandri at gmail.com> ---
Martin: Applied, as in accepted to the upstream kernel tree? No.

I continue building each 3.4.x kernel with this patch though. It still patches
cleanly and fixes my issue.

For the latest stable kernels, it was applying cleanly to 3.8.x kernels, but
I've not yet built or tested any from 3.9.x series.

-- 
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/20130615/c85a745e/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

--- Comment #12 from Martin Peres <martin.peres at ensi-bourges.fr> ---
@Marc: Ok, thanks for the precision.

@Darktama: what do we do with this problem? I know that merging this would
break other stuff so what can we do?

-- 
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/20130626/812ae5e9/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

--- Comment #13 from Marc Meledandri <m.meledandri at gmail.com> ---
FWIW, on kernel series 3.10.x, adjusting the i2c datarate still provides a
working solution on my nv50 hardware. Sensors detect properly and fan control
is enabled. The necessary patch has changed again:

--- a/drivers/gpu/drm/nouveau/core/subdev/i2cbase.c    2013-07-25
18:16:45.000000000 -0400
+++ b/drivers/gpu/drm/nouveau/core/subdev/i2cbase.c.patched    2013-07-29
19:26:09.393689862 -0400
@@ -122,7 +122,7 @@
         if (!bit)
             return -ENOMEM;

-        bit->udelay = 10;
+        bit->udelay = 20;
         bit->timeout = usecs_to_jiffies(2200);
         bit->data = port;
         bit->pre_xfer = nouveau_i2c_pre_xfer;

-- 
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/20130729/fc2bb591/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

--- Comment #14 from Marc Meledandri <m.meledandri at gmail.com> ---
Disregard previous diff. This rather:

--- a/drivers/gpu/drm/nouveau/core/subdev/i2c/base.c    2013-07-25
18:16:45.000000000 -0400
+++ b/drivers/gpu/drm/nouveau/core/subdev/i2c/base.c    2013-07-29
19:26:09.393689862 -0400
@@ -122,7 +122,7 @@
         if (!bit)
             return -ENOMEM;

-        bit->udelay = 10;
+        bit->udelay = 20;
         bit->timeout = usecs_to_jiffies(2200);
         bit->data = port;
         bit->pre_xfer = nouveau_i2c_pre_xfer;

-- 
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/20130729/92881c76/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

--- Comment #15 from Emil Velikov <emil.l.velikov at gmail.com> ---
To make things even better adt7473 seems to load fine in some cases :)

User has the original nv50, running a 3.9.x kernel.

[  DEVICE][0000:01:00.0] BOOT0  : 0x450300a2
[  DEVICE][0000:01:00.0] Chipset: G80 (NV50)
...
adt7475 4-002e: ADT7473 device, revision 0
adt7475 4-002e: Optional features: fan4
[  PTHERM][0000:01:00.0] Found an adt7473 at address 0x2e (controlled by
lm_sensors, temp offset +0 C)
[     I2C][0000:01:00.0] detected monitoring device: adt7473

https://bugs.freedesktop.org/show_bug.cgi?id=67382

-- 
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/20130731/4a6969f4/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

--- Comment #16 from Marc Meledandri <m.meledandri at gmail.com> ---
That's good news. I've been so used to the required patch to build the
3.4.x
LTS kernels I didn't bother to build a vanilla 3.10.x.

I'll try a vanilla build tonight as I got a nouveau PGRAPH crash/hard lockup
last night on 3.10.4 with the bit->udelay = 20.

-- 
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/20130731/6e2e780f/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

--- Comment #17 from Marc Meledandri <m.meledandri at gmail.com> ---
Unfortunately, no adt7473 is detected for my gpu hardware (NVA0 (GT200)) with
vanilla kernel 3.10.4 build. Still seeing:

[  2.083495] i2c i2c-4: sendbytes: NAK bailout.
[  2.086618] i2c i2c-4: sendbytes: NAK bailout.

I'll still need to patch, which is working for me. We must be closing in on
another LTS kernel becoming maintained, and I was hoping a fix would have made
it into the stable tree.

I'll file a separate bug regarding the pgraph crasher.

-- 
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/20130731/f1242f34/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

--- Comment #18 from Emil Velikov <emil.l.velikov at gmail.com> ---
Created attachment 84814
  --> https://bugs.freedesktop.org/attachment.cgi?id=84814&action=edit
Another try

Please give the attached patch a try. It should handle the case in a more
appropriate manner.

Cheers
Emil

-- 
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/20130828/256425dc/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

--- Comment #19 from Martin Peres <martin.peres at ensi-bourges.fr> ---
Created attachment 87907
  --> https://bugs.freedesktop.org/attachment.cgi?id=87907&action=edit
This patch should work

This patch should work, at least it does for me. Sorry it took me so long to
actually do something about it.

-- 
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/20131021/c3bc5d58/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=60704

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

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

--- Comment #20 from Ilia Mirkin <imirkin at alum.mit.edu> ---
Martin's patch should be in 3.13. Marking this as fixed, since there
hasn't
been any response from the reporter for several months.

-- 
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/20140123/29ff8f9b/attachment-0001.html>