Nouveau - May 2013 - [Bug 64904] New: [BISECTED] nouveau, nv50: External VGA not detected anymore

https://bugs.freedesktop.org/show_bug.cgi?id=64904

          Priority: medium
            Bug ID: 64904
          Assignee: nouveau at lists.freedesktop.org
           Summary: [BISECTED] nouveau, nv50: External VGA not detected
                    anymore
        QA Contact: xorg-team at lists.x.org
          Severity: normal
    Classification: Unclassified
                OS: Linux (All)
          Reporter: wippbox at gmx.net
          Hardware: x86 (IA32)
            Status: NEW
           Version: unspecified
         Component: Driver/nouveau
           Product: xorg

I have a lenove thinkpad T61p with a NVIDIA Corporation G84GLM [Quadro FX
570M].
When connecting my external VGA monitor my monitor is no more detected
properly:
> cat /sys/class/drm/card0-VGA-1/status
gives now "disconnected"

I bisected the kernel sources from
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
and filed a bug report at https://bugzilla.kernel.org
(https://bugzilla.kernel.org/show_bug.cgi?id=58021).

Last good commit is
ef22c8bb7b3fac45919b7fde412d36d1a8367d51
drm/nv50/disp: call into core to handle dac/sor power state changes

First bad commit is
7ebb38b556485449bfaa506a196439f6a6fd6ebd
drm/nv50/disp: call into core for dac load detection

The responsible function is
nv50_dac_detect in drivers/gpu/drm/nouveau/nv50_dac.c (moved to
drivers/gpu/drm/nouveau/nv50_display.c in more recent kernels).

Also tried a 3.9.1 kernel. The problem still remains.

However, replacing nv50_dac_detect with the corresponding version from the last
good commit fixed the problem (tried with a 3.8.10 and a 3.8.13 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/20130523/9008090c/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=64904

--- Comment #1 from Gerhard Br?unlich <wippbox at gmx.net> ---
I didn't mention: As a consequence the monitor stays blank.

-- 
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/20130523/4c48a9ad/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=64904

--- Comment #2 from Emil Velikov <emil.l.velikov at gmail.com> ---
Hi Gerhard

There are three distinct changes introduced with the bad commit

* the timeout has been decreased - 45 > 9.5 ms
* the DPMS_MODE was explicitly set before + after the DAC detection, but not
any more
* the NV50_PDISPLAY_DAC_CLK_CTRL1 was set at the beginning, whereas the new
version is not touching the register

So one at a time, can you increase the delay back to 45ms

-- 
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/20130524/e74deb71/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=64904

--- Comment #3 from Gerhard Br?unlich <wippbox at gmx.net> ---
(In reply to comment #2)
> So one at a time, can you increase the delay back to 45ms
Hi Emil
If you get you right, that means that I should put

mdelay(45);

back into the function nv50_dac_detect.
Is that correct?

-- 
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/20130524/5bd3c90d/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=64904

--- Comment #4 from Emil Velikov <emil.l.velikov at gmail.com> ---
Created attachment 79775
  --> https://bugs.freedesktop.org/attachment.cgi?id=79775&action=edit
Increase the delay during dac sense

Take a look at the patch to clear any confusion. It should apply cleanly on top
of the 3.9 branch

-- 
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/20130524/e62ee248/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=64904

--- Comment #5 from Gerhard Br?unlich <wippbox at gmx.net> ---
(In reply to comment #4)
> Created attachment 79775 [details] [review]
> Increase the delay during dac sense
> 
> Take a look at the patch to clear any confusion. It should apply cleanly on
> top of the 3.9 branch
OK thank you! I don't have my thinkpad with me at the moment. But I will
test
the patch tomorrow.

-- 
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/20130525/7fc85390/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=64904

--- Comment #6 from Gerhard Br?unlich <wippbox at gmx.net> ---
(In reply to comment #2)
> So one at a time, can you increase the delay back to 45ms
OK, done: this doesn't fix the problem!

Also tried to set NV50_PDISPLAY_DAC_CLK_CTRL1 (see patch). Didn't fix it
too.

-- 
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/20130526/0bb7e327/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=64904

--- Comment #7 from Gerhard Br?unlich <wippbox at gmx.net> ---
Created attachment 79820
  --> https://bugs.freedesktop.org/attachment.cgi?id=79820&action=edit
set NV50_PDISPLAY_DAC_CLK_CTRL1

-- 
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/20130526/b59c4f0a/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=64904

--- Comment #8 from Emil Velikov <emil.l.velikov at gmail.com> ---
Created attachment 79852
  --> https://bugs.freedesktop.org/attachment.cgi?id=79852&action=edit
mangle with dpms mode, before/after dac sense

This is the final piece.

If it does not help, then the three changes could be more closely related than
I expected. Either way give it a try, it should apply cleanly on top of linus
3.9

-- 
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/20130527/1c05abdc/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=64904

--- Comment #9 from Gerhard Br?unlich <wippbox at gmx.net> ---
(In reply to comment #8)
> Created attachment 79852 [details] [review]
> mangle with dpms mode, before/after dac sense
> 
> This is the final piece.
> 
> If it does not help, then the three changes could be more closely related
> than I expected. Either way give it a try, it should apply cleanly on top
of
> linus 3.9
Didn't fix it.
Don't know if the following is useful:
In the statement (nv50_dac_detect)
    if (ret || load != 7)
        return connector_status_disconnected;
I get ret=0 and load!=7 (but don't know the value of load). So load!=7 is
responsible that connector_status_disconnected is returned.

-- 
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/20130527/c9009352/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=64904

--- Comment #10 from Maarten Lankhorst <m.b.lankhorst at gmail.com> ---
Strictly speaking it should probably be (load & 7) == 7, but I don't
think
it'll fix it on its own, you might want to dump the value of load to be
sure,
with all those patches.

-- 
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/20130527/06cbcf0b/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=64904

--- Comment #11 from Gerhard Br?unlich <wippbox at gmx.net> ---
(In reply to comment #10)
> Strictly speaking it should probably be (load & 7) == 7, but I
don't think
> it'll fix it on its own, you might want to dump the value of load to be
> sure, with all those patches.
OK, I get load=0.

-- 
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/20130528/0bbcfdd9/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=64904

--- Comment #12 from Joshua Baergen <joshuabaergen at gmail.com> ---
I can confirm this issue on a ThinkPad W530, NVIDIA Corporation GK107GLM
[Quadro K1000M] (rev a1).  I noticed the issue at work when I tried to connect
to VGA extender which in turn was hooked up to a projector.  However,
connecting directly to a VGA monitor worked fine.

I took Gerhard's suggestion and patched my 2.8.13 Gentoo kernel with the
nv50_dac_detect from ef22c8bb7b3fac45919b7fde412d36d1a8367d51 and I was then
able to connect to the projector.

FWIW, at least one other thing I noticed in the patch is that in the working
code, there's the following sequence:

1) Write (NV50_PDISPLAY_DAC_LOAD_CTRL_ACTIVE | load_pattern) to
DAC_LOAD_CTRL(i).
2) Read DAC_LOAD_CTRL(i).
3) Write 0 to DAC_LOAD_CTRL(i).

In the new code, it's:

1) Write (NV50_PDISPLAY_DAC_LOAD_CTRL_ACTIVE | load_pattern) to
DAC_LOAD_CTRL(i).
2) Write NV50_PDISPLAY_DAC_LOAD_CTRL_DONE to DAC_LOAD_CTRL(i).
3) Read DAC_LOAD_CTRL(i).

Being completely unfamiliar with the hardware, I'm not sure if this is
significant, and unfortunately, I don't have much (if any) opportunity to
experiment.

Josh

-- 
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/20130530/0c49856c/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=64904

Emil Velikov <emil.l.velikov at gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED

--- Comment #13 from Emil Velikov <emil.l.velikov at gmail.com> ---
(In reply to comment #12)
> ...
> 2) Write NV50_PDISPLAY_DAC_LOAD_CTRL_DONE to DAC_LOAD_CTRL(i).
Nice catch Josh, although this write is not the one causing the issue.


Just got a VGA cable and I can reproduce the issue.

Would need to do some testing if the fix does not break the behaviour on any
other cards

-- 
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/20130602/52e06341/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=64904

Emil Velikov <emil.l.velikov at gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|nouveau at lists.freedesktop.o |emil.l.velikov at gmail.com
                   |rg                          |

-- 
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/20130602/5206f153/attachment.html>