Nouveau - Aug 2010 - [Bug 29455] New: Nouveau TV-out incorrectly detecting load on NV18

https://bugs.freedesktop.org/show_bug.cgi?id=29455

           Summary: Nouveau TV-out incorrectly detecting load on NV18
           Product: xorg
           Version: 7.5
          Platform: PowerPC
        OS/Version: Linux (All)
            Status: NEW
          Severity: normal
          Priority: medium
         Component: Driver/nouveau
        AssignedTo: nouveau at lists.freedesktop.org
        ReportedBy: rechner at vlado-do.de
         QAContact: xorg-team at lists.x.org


Nouveau is detecting and enabling TV-out, even though nothing is plugged in to
it.

This causes the non-X terminals to only offer reduced resolution in the upper
left corner of the display, and also has a few implications for X. Overall my
problem seems very similar to bug #26800 .

I'm afraid I can't help much in solving this problem, for my knowledge
of
kernel internals is close to non-existent, and anyway PowerPC is not supported
by the Nvidia binary driver.

I don't care much about the non-X terminals' resolution, but there is
one
annoying problem in X: Firefox (or Iceweasel in Debian) only expands its
bookmarks menu within the small area in the upper left corner of the screen.
Also the background in GDM and Gnome is show in reduced resolution in that
corner, in addition to a full screen version.

I can get rid of the problems in X by calling "xrandr --output TV-1
--off". A
permanent fix is the following simple /etc/X11/xorg.conf:

Section "Monitor" 
       Identifier      "TV-1"
       Option          "Ignore" "True" 
EndSection 

The problem occurs with Ubuntu 10.04 (Kernel 2.6.33, nouveau
0.15+git20100219+9b4118d-0ubuntu5) and with Debian sid (Kernel 2.6.35, nouveau
0.16+git20100518+4b8f1a0-1).

I keep getting messages like this in the Kernel log:

[ 6399.294705] [drm] nouveau 0000:00:10.0: Output TV-1 is running on CRTC 1
using output B
[ 6399.498411] [drm] nouveau 0000:00:10.0: Load detected on output B
[ 6399.730419] [drm] nouveau 0000:00:10.0: Load detected on output B
[ 6399.934411] [drm] nouveau 0000:00:10.0: Load detected on output B
[ 6400.126411] [drm] nouveau 0000:00:10.0: Load detected on output B

(from Debian)

[  124.508970] [drm] nouveau 0000:00:10.0: Load detected on output B
[  150.448932] [drm] nouveau 0000:00:10.0: Load detected on output B
[  232.721150] [drm] nouveau 0000:00:10.0: Load detected on output B
[  232.727489] [drm] nouveau 0000:00:10.0: Setting dpms mode 3 on TV encoder
(output 2)
[  243.165175] [drm] nouveau 0000:00:10.0: Load detected on output B

(from Ubuntu)

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

--- Comment #1 from Francisco Jerez <currojerez at riseup.net> 2010-08-11
11:44:39 PDT ---
(In reply to comment #0)
> Nouveau is detecting and enabling TV-out, even though nothing is plugged in
to
> it.
> 
> This causes the non-X terminals to only offer reduced resolution in the
upper
> left corner of the display, and also has a few implications for X. Overall
my
> problem seems very similar to bug #26800 .
> 
> I'm afraid I can't help much in solving this problem, for my
knowledge of
> kernel internals is close to non-existent, and anyway PowerPC is not
supported
> by the Nvidia binary driver.
> 
> I don't care much about the non-X terminals' resolution, but there
is one
> annoying problem in X: Firefox (or Iceweasel in Debian) only expands its
> bookmarks menu within the small area in the upper left corner of the
screen.
> Also the background in GDM and Gnome is show in reduced resolution in that
> corner, in addition to a full screen version.
> 
> I can get rid of the problems in X by calling "xrandr --output TV-1
--off". A
> permanent fix is the following simple /etc/X11/xorg.conf:
> 
> Section "Monitor" 
>        Identifier      "TV-1"
>        Option          "Ignore" "True" 
> EndSection 
> 
> The problem occurs with Ubuntu 10.04 (Kernel 2.6.33, nouveau
> 0.15+git20100219+9b4118d-0ubuntu5) and with Debian sid (Kernel 2.6.35,
nouveau
> 0.16+git20100518+4b8f1a0-1).
> 
> I keep getting messages like this in the Kernel log:
> 
> [ 6399.294705] [drm] nouveau 0000:00:10.0: Output TV-1 is running on CRTC 1
> using output B
> [ 6399.498411] [drm] nouveau 0000:00:10.0: Load detected on output B
> [ 6399.730419] [drm] nouveau 0000:00:10.0: Load detected on output B
> [ 6399.934411] [drm] nouveau 0000:00:10.0: Load detected on output B
> [ 6400.126411] [drm] nouveau 0000:00:10.0: Load detected on output B
> 
> (from Debian)
> 
> [  124.508970] [drm] nouveau 0000:00:10.0: Load detected on output B
> [  150.448932] [drm] nouveau 0000:00:10.0: Load detected on output B
> [  232.721150] [drm] nouveau 0000:00:10.0: Load detected on output B
> [  232.727489] [drm] nouveau 0000:00:10.0: Setting dpms mode 3 on TV
encoder
> (output 2)
> [  243.165175] [drm] nouveau 0000:00:10.0: Load detected on output B
> 
> (from Ubuntu)
Ideally you'd take an mmiotrace from the nvidia blob to fix this. I
understand
that it's not an option in this case -- a video BIOS dump and a register
dump
*might* be enough though: [1] has instructions on dumping video BIOSes, and you
can use [2] after a fresh boot, before loading nouveau.ko to get a register
dump.
Full kernel logs may be useful too.

[1] http://nouveau.freedesktop.org/wiki/DumpingVideoBios
[2] http://cgit.freedesktop.org/~currojerez/tvdump/

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

--- Comment #2 from Vlado Plaga <rechner at vlado-do.de> 2010-08-12
04:21:06 PDT ---
Created an attachment (id=37809)
 --> (https://bugs.freedesktop.org/attachment.cgi?id=37809)
register dump

A register dump produced with "tvdump" on a Debian lenny system on my
iMac G4
(with NV18 graphics).

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

--- Comment #3 from Vlado Plaga <rechner at vlado-do.de> 2010-08-12
04:23:34 PDT ---
Created an attachment (id=37810)
 --> (https://bugs.freedesktop.org/attachment.cgi?id=37810)
video BIOS dump

This is a copy of "/proc/device-tree/pci at f0000000/NVDA,Parent at
10/NVDA,BMP".

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

--- Comment #4 from Vlado Plaga <rechner at vlado-do.de> 2010-08-12
04:29:26 PDT ---
Created an attachment (id=37811)
 --> (https://bugs.freedesktop.org/attachment.cgi?id=37811)
Full kernel log

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

--- Comment #5 from Vlado Plaga <rechner at vlado-do.de> 2010-08-12
04:34:18 PDT ---
Thanks for the quick reaction! It would be nice if I could contribute to fixing
this problem.

I made the register dump on the same machine, but with a different (and older)
kernel without support for nouveau. X was running with the nv driver. Is that
ok?

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

--- Comment #6 from Francisco Jerez <currojerez at riseup.net> 2010-08-12
05:28:14 PDT ---
Your DCB tables are a little bit unusual... Does your card really have a TV
output?

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

--- Comment #7 from Vlado Plaga <rechner at vlado-do.de> 2010-08-12
08:25:47 PDT ---
(In reply to comment #6)
> Your DCB tables are a little bit unusual... Does your card really have a TV
> output?
It's not a normal TV output, but the "mini-VGA video output" on my
iMac's back
allegedly can do s-video and composite output with the help of an adapter:
http://store.apple.com/us/product/M9109G/A?mco=MTY3ODQ5OTY

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

--- Comment #8 from Francisco Jerez <currojerez at riseup.net> 2010-08-12
14:04:58 PDT ---
Created an attachment (id=37826)
 View: https://bugs.freedesktop.org/attachment.cgi?id=37826
 Review: https://bugs.freedesktop.org/review?bug=29455&attachment=37826

nouveau_imac_g4_tvout.patch

(In reply to comment #7)
> (In reply to comment #6)
> > Your DCB tables are a little bit unusual... Does your card really have
a TV
> > output?
> 
> It's not a normal TV output, but the "mini-VGA video output"
on my iMac's back
> allegedly can do s-video and composite output with the help of an adapter:
> http://store.apple.com/us/product/M9109G/A?mco=MTY3ODQ5OTY
Ah, that explains a few things. TV load detection probably fails because in VGA
mode the TV encoder lines aren't even electrically connected to the mini-VGA
port.

Most likely there's a GPIO-controlled switch you have to poke to make TV-out
work: the attached patch is just a wild guess, if it has no effect you could
try to play a bit with the "gpio->line" parameter (values in 7-9
and 2-4 should
be safe).

BTW, do you happen to have access to one of these adapters?

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

--- Comment #9 from Vlado Plaga <rechner at vlado-do.de> 2010-08-13
04:03:54 PDT ---
(In reply to comment #8)
> Most likely there's a GPIO-controlled switch you have to poke to make
TV-out
> work: the attached patch is just a wild guess, if it has no effect you
could
> try to play a bit with the "gpio->line" parameter (values in
7-9 and 2-4 should
> be safe).
I applied the patch (manually) to the 2.6.35 Kernel from Debian (experimental),
but it had no effect. I don't feel comfortable fiddling around with the
values
for "gpio->line" without having the slightest idea what this would
or should
cause. The value in your patch was "7", so there would still be 5
possibilities
left to try?
> BTW, do you happen to have access to one of these adapters?
No, I don't have on (yet). But I'm now trying to buy one second-hand on
eBay. I
don't feel like spending ? 19 on the adapter from Apple, but apart from
testing
for "nouveau" I might need the adapter later in case my iMac spends
his last
years at my mum's place, as a combined internet machine and DVD player. If I
should get the adapter (i.e. no-one else gives a higher bid in the next week)
it would still take a few more days until it could be here. But the coming week
I'll be on holidays anyway.

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

--- Comment #10 from Francisco Jerez <currojerez at riseup.net> 2010-08-13
05:51:45 PDT ---
Created an attachment (id=37847)
 View: https://bugs.freedesktop.org/attachment.cgi?id=37847
 Review: https://bugs.freedesktop.org/review?bug=29455&attachment=37847

nouveau_dac_detect_debug.patch

(In reply to comment #9)
> (In reply to comment #8)
> 
> > Most likely there's a GPIO-controlled switch you have to poke to
make TV-out
> > work: the attached patch is just a wild guess, if it has no effect you
could
> > try to play a bit with the "gpio->line" parameter (values
in 7-9 and 2-4 should
> > be safe).
> 
> I applied the patch (manually) to the 2.6.35 Kernel from Debian
(experimental),
> but it had no effect. I don't feel comfortable fiddling around with the
values
> for "gpio->line" without having the slightest idea what this
would or should
> cause. The value in your patch was "7", so there would still be 5
possibilities
> left to try?
> 
Yeah, the attached patch adds some debugging output that may be useful for you
to see the effects of your changes, "output=2 sample=YXXXXXXX" with
Y!=0 means
load is incorrectly being detected in your tv output.

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

--- Comment #11 from Vlado Plaga <rechner at vlado-do.de> 2010-08-13
07:47:56 PDT ---
> Yeah, the attached patch adds some debugging output that may be useful for
you
> to see the effects of your changes, "output=2 sample=YXXXXXXX"
with Y!=0 means
> load is incorrectly being detected in your tv output.
Debugging output still does not help me understand much... but nevertheless I
wanted to give it a try. 

The first problem with the patch was: I don't have the file
"nv10_gpio.c". Is
that a problem? Ok, I guess I should be using latest nouveau git instead of a
stable kernel anyway... so I tried the out-of-tree build (alternative), but
there was a new problem:

In file included from
/usr/src/nouveau/master/drivers/gpu/drm/nouveau/nouveau_bios.c:31:
include/linux/io-mapping.h:25:23: error: asm/iomap.h: No such file or directory
make[3]: *** [/usr/src/nouveau/master/drivers/gpu/drm/nouveau/nouveau_bios.o]
Error 1
make[2]: *** [/usr/src/nouveau/master/drivers/gpu/drm/nouveau] Error 2
make[1]: *** [_module_/usr/src/nouveau/master/drivers/gpu/drm] Error 2
make[1]: Leaving directory `/usr/src/linux-source-2.6.35'
make: *** [modules] Error 2

I suppose it is an architecture-related problem. In kernel 2.6.35
nouveau_bios.c does not include "linux/io-mapping.h" yet, so
"asm/iomap.h" is
not required either. The later file is in
"linux-source-2.6.35/arch/x86/include/asm/iomap.h" on my system, whis
is
probably not good for a PPC. There is also a
"linux-source-2.6.35/include/asm-generic/iomap.h", which might be an
alternative.

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

--- Comment #12 from Francisco Jerez <currojerez at riseup.net> 2010-08-13
08:23:18 PDT ---
(In reply to comment #11)
> > Yeah, the attached patch adds some debugging output that may be useful
for you
> > to see the effects of your changes, "output=2
sample=YXXXXXXX" with Y!=0 means
> > load is incorrectly being detected in your tv output.
> 
> Debugging output still does not help me understand much... but nevertheless
I
> wanted to give it a try. 
> 
> The first problem with the patch was: I don't have the file
"nv10_gpio.c". Is
> that a problem? Ok, I guess I should be using latest nouveau git instead of
a
> stable kernel anyway... so I tried the out-of-tree build (alternative), but
> there was a new problem:
> 
> In file included from
> /usr/src/nouveau/master/drivers/gpu/drm/nouveau/nouveau_bios.c:31:
> include/linux/io-mapping.h:25:23: error: asm/iomap.h: No such file or
directory
> make[3]: ***
[/usr/src/nouveau/master/drivers/gpu/drm/nouveau/nouveau_bios.o]
> Error 1
> make[2]: *** [/usr/src/nouveau/master/drivers/gpu/drm/nouveau] Error 2
> make[1]: *** [_module_/usr/src/nouveau/master/drivers/gpu/drm] Error 2
> make[1]: Leaving directory `/usr/src/linux-source-2.6.35'
> make: *** [modules] Error 2
> 
> I suppose it is an architecture-related problem. In kernel 2.6.35
> nouveau_bios.c does not include "linux/io-mapping.h" yet, so
"asm/iomap.h" is
> not required either. The later file is in
> "linux-source-2.6.35/arch/x86/include/asm/iomap.h" on my system,
whis is
> probably not good for a PPC. There is also a
> "linux-source-2.6.35/include/asm-generic/iomap.h", which might be
an
> alternative.
That's already fixed in Linus' kernel tree, try cherry-picking
31ce4bfdfd10bf5db9bf85c92bbe0cf2edbdcad8.

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

Vlado Plaga <rechner at vlado-do.de> changed:

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

--- Comment #13 from Vlado Plaga <rechner at vlado-do.de> 2010-08-14
13:09:21 PDT ---
Great! With "gpio->line = 4" a TV is no longer incorrectly
detected. Of course
I had to try the four other values before, and no longer expected any
difference... but the more positive was my surprise when the last attempt
worked.

I used Linus' git tree first, where both patches could be applied. But there
were still many errors in the kernel log, so I went back to kernel 2.6.35,
where the  nouveau_imac_g4_tvout.patch (with "gpio-> = 4") also has
the desired
effect.

Thank you, Francisco Jerez! Keep up the good work with nouveau!

P.S.: If you're interested, I can still upload kernel logs with the new
errors
from 2.6.35+, and with no errors with a patched 2.6.35.

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

--- Comment #14 from Francisco Jerez <currojerez at riseup.net> 2010-08-14
14:43:50 PDT ---
(In reply to comment #13)
> Great! With "gpio->line = 4" a TV is no longer incorrectly
detected. Of course
> I had to try the four other values before, and no longer expected any
> difference... but the more positive was my surprise when the last attempt
> worked.
> 
Good, I also need the output from "lspci -vvnn" to add a quirk for
your card.
> I used Linus' git tree first, where both patches could be applied. But
there
> were still many errors in the kernel log, so I went back to kernel 2.6.35,
> where the  nouveau_imac_g4_tvout.patch (with "gpio-> = 4")
also has the desired
> effect.
> 
What kind of errors do you get with Linus' master?

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

--- Comment #15 from Vlado Plaga <rechner at vlado-do.de> 2010-08-15
02:41:22 PDT ---
(In reply to comment #14)
> I also need the output from "lspci -vvnn" to add a quirk for your
card.
I'll add a small attachment, so that no new line breaks are introduced.
> > I used Linus' git tree first, where both patches could be applied.
But there
> > were still many errors in the kernel log, so I went back to kernel
2.6.35,
> > where the  nouveau_imac_g4_tvout.patch (with "gpio-> =
4") also has the desired
> > effect.
> > 
> What kind of errors do you get with Linus' master?
I'm not sure these errors are caused by nouveau, but it seem so. The system
was
working, but KDE also gave notification of kernel failures. I'll add a
kern.log. I cut a huge chunk in the middle, because the file was 5,7 MBytes
originally.

The last boot recorded was with the modified working 2.6.35 kernel:
Aug 14 21:40:07 platypus kernel: [    0.000000] Linux version 2.6.35 (2.6.35)
(root at platypus) (gcc version 4.4.5 20100728 (prerelease) (Debian 4.4.4-8) )
#1
Sat Aug 14 20:18:29 CEST 2010

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

--- Comment #16 from Vlado Plaga <rechner at vlado-do.de> 2010-08-15
02:43:12 PDT ---
Created an attachment (id=37880)
 --> (https://bugs.freedesktop.org/attachment.cgi?id=37880)
Kernel log (truncated)

Kernel log including DRM errors with Kernel 2.6.35+ (git)

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

--- Comment #17 from Vlado Plaga <rechner at vlado-do.de> 2010-08-15
02:43:55 PDT ---
Created an attachment (id=37881)
 --> (https://bugs.freedesktop.org/attachment.cgi?id=37881)
Output of "lspci -vvnn"

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

Vlado Plaga <rechner at vlado-do.de> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #37880|text/x-log                  |text/plain
          mime type|                            |

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

Vlado Plaga <rechner at vlado-do.de> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #37881|application/octet-stream    |text/plain
          mime type|                            |

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.

https://bugs.freedesktop.org/show_bug.cgi?id=29455

--- Comment #18 from Francisco Jerez <currojerez at riseup.net> 2010-08-15
06:36:08 PDT ---
(In reply to comment #15)
> (In reply to comment #14)
> > I also need the output from "lspci -vvnn" to add a quirk for
your card.
> 
> I'll add a small attachment, so that no new line breaks are introduced.
> 
Thanks, this should be fixed in master now. There are still a few assumptions
we have to fix to get the Apple VGA-to-TV-out adapter working. Feel free to
open a new bug if you get one and we'll sort it out.
> > > I used Linus' git tree first, where both patches could be
applied. But there
> > > were still many errors in the kernel log, so I went back to
kernel 2.6.35,
> > > where the  nouveau_imac_g4_tvout.patch (with "gpio-> =
4") also has the desired
> > > effect.
> > > 
> > What kind of errors do you get with Linus' master?
> 
> I'm not sure these errors are caused by nouveau, but it seem so. The
system was
> working, but KDE also gave notification of kernel failures. I'll add a
> kern.log. I cut a huge chunk in the middle, because the file was 5,7 MBytes
> originally.
> 
The EDID errors you get are definitely nouveau's fault and they're
already
fixed in nouveau/master. OTOH the "scheduling while atomic" errors
don't look
nouveau-related.

-- 
Configure bugmail: https://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.