[gutsy] failure detecting DVI in MacMini since xserver-xorg-video-ati 6.7.192

[Alex Muntada]

Binary package hint: xserver-xorg-video-ati

On Sep 19th my MacMini PowerPC was upgraded from xserver-xorg-video-ati 1:6.6.193-1ubuntu1 to 1:6.7.192-4ubuntu1. After rebooting, gdm couldn't detect the Xserver running. Later, upgrading to 1:6.7.193-1ubuntu1 didn't make a difference.

I tried to dpkg-reconfigure xserver-xorg with no luck. Only downgrading to 1:6.6.193-1ubuntu1 gets rid of the problem.

My Dell 2005FPW LCD screen is connected to the MacMini through a DVI cable and my card is a "VGA compatible controller: ATI Technologies Inc RV280 [Radeon 9200] (rev 01)" per lspci.

[Alex Muntada (alex.muntada)]

Since gdm tries to start Xserver serveral times, some of them in failure mode, the Xorg.0.log files get overwritten too. To get the right Xorg.0.log with the error, just try to run a simple "X" as root.

[Alex Muntada (alex.muntada)]

This log shows the actual problem with the video driver, which can't see the LCD screen attached through DVI.

[Alex Muntada (alex.muntada)]

This log shows how failsafe mode cannot be started, but I just found that it's because I don't have xserver-xorg-video-vesa installed.

Anyway, the ATI driver issue isn't related to this.

[Marcus Comstedt (marcus-mc)]

The same thing happens to me, with a CRT monitor connected using a DVI-I->VGA adapter.

Since I didn't have a copy of 1:6.6.193-1ubuntu1 lying around, I tried downgrading to the
feisty version (1:6.6.3-2ubuntu6), and this worked as well.

I noticed that by setting the option MacModel to "ibook" or "powerbook" (there was no "macmini"
alternative listed in the manpage) I could get the server to start, but then my monitor didn't like
the signal.

[Alex Muntada (alex.muntada)]

Marcus, find the 1:6.6.193-1ubuntu1 version here:
http://ports.ubuntu.com/pool/main/x/xserver-xorg-video-ati/xserver-xorg-video-ati_6.6.193-1ubuntu1_powerpc.deb

[Tormod Volden (tormodvolden)]

It would be great if you could attach Xorg.0.log from using the 6.7.193-1ubuntu1 with the MacModel option.

[Alex Muntada (alex.muntada)]

On Sep 19th my MacMini PowerPC was upgraded from xserver-xorg-video-ati 1:6.6.193-1ubuntu1 to 1:6.7.192-4ubuntu1. After rebooting, gdm couldn't detect the Xserver running. Later, upgrading to 1:6.7.193-1ubuntu1 didn't make a difference.

I tried to dpkg-reconfigure xserver-xorg with no luck. Only downgrading to 1:6.6.193-1ubuntu1 gets rid of the problem.

My Dell 2005FPW LCD screen is connected to the MacMini through a DVI cable and my card is a "VGA compatible controller: ATI Technologies Inc RV280 [Radeon 9200] (rev 01)" per lspci.

[Alex Muntada (alex.muntada)]

Created an attachment (id=11684)
Xorg.0.log using latest git code

Just compiled and tried xf86-video-ati from git master (commit 5a6f74103f0ec0d451d0e2573442efe5922848af) and it doesn't work either. I'm attaching the Xorg.0.log.

[Alex Muntada (alex.muntada)]

For the record, I previously reported this bug in Ubuntu Gutsy:
https://bugs.launchpad.net/bugs/141372

[agd5f (agd5f)]

The mac mini needs a custom connector table like most other macs.
What connectors does it have? based on apple's web site, it looks like
it has just a DVI port that can drive DVI or VGA or tv-out (via DVI
adapters). I'm guessing it uses internal TMDS and the TV DAC. The
question is what ddc line it uses. Does adding the following option
in the device section of your config help?

Option "ConnectorTable" "X,1,0,3,0,1,2,6"

replace X with 1 or 2 or 3 or 4 or 5 until you get DDC working.

Once we sort this out, I'll add a MacModel option for the mini.

[Alex Muntada (alex.muntada)]

Mine has just a DVI port with a DVI-to-VGA adapter.

I tried the ConnectorTable option with 1..5 values without success. All values seemed to have the same result:

...
(EE) RADEON(0): No valid modes.
(II) UnloadModule: "ati"
(II) UnloadModule: "radeon"
(EE) Screen(s) found, but none have a usable configuration.

Fatal server error:
no screens found

However, for X=4 X didn't end with an error though the screen went to sleep mode (I guess that's because it was receiveing bad values).

I kept all 5 Xorg.0.log files; let me know which one would help or whether I should try different values in ConnectorTable.

Thanks a lot!

[Mark Brown (broonie)]

Having a similar problem with a Mac Mini I found that

   Option "ConnectorTable" "4,1,0,3,0,1,2,6"

caused DDC to kick into life, though I am still experiencing problems with this as discused in http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=443570 (which appear to match the problem that Alex is seeing).

The machine is:

machine : PowerMac10,2
motherboard : PowerMac10,2 MacRISC3 Power Macintosh
detected as : 287 (Unknown Intrepid-based)

[Mark Brown (broonie)]

(I should probably add that the information read via DDC does look accurate for my monitor.)

[agd5f (agd5f)]

(In reply to comment #5)
> Having a similar problem with a Mac Mini I found that
>
> Option "ConnectorTable" "4,1,0,3,0,1,2,6"
>
> caused DDC to kick into life, though I am still experiencing problems with this
> as discused in http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=443570 (which
> appear to match the problem that Alex is seeing).

perhaps it's using external TMDS, try:

Option "ConnectorTable" "4,1,1,3,0,1,2,6"

[Alex Muntada (alex.muntada)]

This one worked wonders, though the cursor issue from bug #11796 is shown.

[agd5f (agd5f)]

(In reply to comment #8)
> This one worked wonders, though the cursor issue from bug #11796 is shown.
>

So to confirm, Option "ConnectorTable" "4,1,1,3,0,1,2,6" (external TMDS) fixes the display for you?

[agd5f (agd5f)]

(In reply to comment #8)
> This one worked wonders, though the cursor issue from bug #11796 is shown.
>

Also is this with DVI or VGA connection?

[Alex Muntada (alex.muntada)]

Created an attachment (id=11686)
working Xorg.0.log

The attached Xorg.0.log confirms that Option "ConnectorTable" "4,1,1,3,0,1,2,6" (external TMDS) fixes the DVI-to-DVI display on my MacMini.

[agd5f (agd5f)]

Excellent. Fixed upstream: 81114af1cdddc0b10d076f2e38c7a00c1223cc48

With ati git master
Option "MacModel" "mini"
should do the right thing automatically.

[Alex Muntada (alex.muntada)]

None of the 3 MacModel values worked for my MacMini :-/

BTW, I'm evaluating whether https://bugs.freedesktop.org/show_bug.cgi?id=9955 is the right upstream bug to link to. Though it's closed right now, I'll probably reopen it to report the failures in MacMini.

[Alex Muntada (alex.muntada)]

Just recompiled latest git master to confirm that Option "MacModel" "mini" works fine with my DVI-to-DVI display.

Thanks a lot!

[Alex Muntada (alex.muntada)]

Reported a new bug upstream, since freedesktop/9955 doesn't seem to be the same.

[Alex Muntada (alex.muntada)]

Upstream confirmed this bug and provided a work-around: add this line to "Device" section in xorg.conf

Option "ConnectorTable" "4,1,1,3,0,1,2,6"

[Alex Muntada (alex.muntada)]

Upstream just committed a patch to add this new option to get MacMini displays working properly. See for details:
https://bugs.freedesktop.org/show_bug.cgi?id=12525#c12

I also reported this change to debian bug, so the experimental xserver-xorg-video-ati version could be updated accordingly and include it in Gutsy.

BTW, I know that confirming oneself's bug report isn't fair enough, but I think in this case it's okay.

[Alex Muntada (alex.muntada)]

By new option above, I meant: Option "MacModel" "mini".

[Tormod Volden (tormodvolden)]

Alex, thanks for reporting and getting it fixed upstream!

[Marcus Comstedt (marcus-mc)]

The ConnectorTable workaround works here as well. Thanks Alex for the heads up.

I hope the MacModel option gets inserted automatically by the Gutsy upgrader/installer, having to tinker manually with xorg.conf just to get X working would be a big step backwards from previous releases. Working right out of the box is one of the things i appreciate about Ubuntu...

[Bret Towe (magnade)]

the ConnectorTable option works for me also
but the monitor goes to sleep and I need to switch to a console and back to wake it
this is a ppc mac mini using a dvi monitor

[Tormod Volden (tormodvolden)]

Fixed in 6.7.194, uploaded to be included post-beta.

Test packages can be downloaded from http://ppa.launchpad.net/tormodvolden/ubuntu/pool/main/x/xserver-xorg-video-ati/

[Michael R. Head (burner)]

I upgraded to gutsy today and it hosed X on my Mac Mini. I was happy to see the 6.7.194 package appear, but it has been of little help on my PPC-based Mac Mini with a Dell 2000FP DVI-connected monitor.

I attempted a "sudo dpkg-reconfigure xserver-xorg", which appears to correctly detect the monitor and radeon. It doesn't add the option "MacModel" "mini" line, but even after manually adding this line myself, I can't get X to display on my monitor.

X now starts, but it doesn't output a signal that my monitor can handle, so the monitor goes into power save mode.

[Michael R. Head (burner)]

Interestingly, if I connect the same monitor via the Apple DVI-VGA adapter using the monitor's D-SUB input, X _does_ start up OK (even using the same ModeLine, AFAICT)

[Tormod Volden (tormodvolden)]

Michael, can you please attach Xorg.0.log from both cases? You can also mention this in the upstream bug.

[Alex Muntada (alex.muntada)]

Michael, try different values for ConnectorTable option as described in:
https://bugs.freedesktop.org/show_bug.cgi?id=12525

Maybe your MacMini's hardware is a bit different and it needs different ConnectorTable values too. If that's the case, we should try to find those differences and file a new bug both in LP and upstream.

[Marcus Comstedt (marcus-mc)]

With version 1:6.7.194-1ubuntu1, the X server starts and correctly detects my monitor, however, modes
over 1280x1024 doesn't seem to work. The picture gets all out of sync, and after switching to a different
virtual console and back the monitor doesn't even want to display it.

This is with the MacModel "mini" option. Any suggestions on ConnectorTable variants to try?
The two listed in the freedesktop.org bug don't give any better results.

[Marcus Comstedt (marcus-mc)]

After some wardialing, I found that a ConnectorTable of "4,1,1,3,0,1,2,1" gives satisfactory
results on my system (analog CRT connected through VGA cable + DVI dongle).

[Michael R. Head (burner)]

Marcus, have you tried using an Xv app such as totem? I found that when I used your setting, videos wouldn't display on the overlay (totem would just show a blank black rectangle).

[Tormod Volden (tormodvolden)]

Marcus, great! Can you comment on this upstream, together with exact hardware details please?

[Marcus Comstedt (marcus-mc)]

Michael, you're right. Video overlays don't work with that setting.
I found another one with which the overlays do work: "4,1,1,1,0,1,2,1"
However, this setting has a different problem: When the X server
starts, the monitor goes into power save mode. To get picture, I have
to switch to a different virtual console, and then back again. But after
that everything works (including video overlay and GL).

Isn't there some documentation somewhere what the numbers mean?
Just guessing randomly doesn't seem like the best way to arrive at
a combination where everything works...

[Luke Yelavich (themuso)]

The ConnectorTable "4,1,1,1,0,1,2,1" works for me, but as already stated, this causes no signal
to be sent to the monitor, until one switches away and back to X.

This is on a Mac Mini G4, 1.42Ghz, Radeon 9200 with 32MB video RAM.

[Luke Yelavich (themuso)]

Forgot to add that I am using the DVI -> VGA adapter, connected to a KVM that doesn't support
DDC probing. My monitor resolutions and sync ranges are correctly configured in xorg.conf
however, yet with the connectortable option mentioned in my previous comment, I only get 640x480
resolution. This is on a ChiMei CMV 22" widescreen LCD monitor, 1680x1050 max resolution.

[Michael R. Head (burner)]

I wish there were a way to revert to the behavior in feisty. How is it that drivers regress like this?

[Tormod Volden (tormodvolden)]

You can always install the latest 6.6.3 version instead. The drivers are generally progressing, but with major rework being done, there might be some regressions for some configurations. 6.8 is not even fully released yet so we can expect some improvements before then. However, the 6.8 pre-releases landed in Ubuntu because they already fix more than they break. And what's here broken gets repaired at quite high rate, like in this original bug report.

[Tormod Volden (tormodvolden)]

FYI, the xserver-xorg-video-ati_6.7.196+git20071202-0ubuntu0tormod from https://wiki.ubuntu.com/XorgOnTheEdge has the latest upstream fixes for mac mini autodetection. Testing and feedback would be good.

[Luke Yelavich (themuso)]

I am using hardy on my mini now, ati driver version 6.7.196-2, and still no joy with getting xorg to work. Attaching my log, and xorg.conf. I have
not yet tried the connectortable option.

[Luke Yelavich (themuso)]

xorg.conf

[Luke Yelavich (themuso)]

Ok, the original connectortable option suggested in this bug still doesn't work for me. Rebuilding git version from PPA for hardy, and will see if that yields any better results.

[Luke Yelavich (themuso)]

After noticing a newer git snapshot release in PPA, I built, and attempted to use that, however, still no joy.

Here is my card info from lspci.

0000:00:10.0 VGA compatible controller [0300]: ATI Technologies Inc RV280 [Radeon 9200] [1002:5962] (rev 01) (prog-if 00 [VGA controller])
 Subsystem: ATI Technologies Inc RV280 [Radeon 9200] [1002:5962]
 Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B-
 Status: Cap+ 66MHz+ UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR-
 Latency: 255 (2000ns min), Cache Line Size: 32 bytes
 Interrupt: pin A routed to IRQ 48
 Region 0: Memory at 98000000 (32-bit, prefetchable) [size=128M]
 Region 1: I/O ports at 0400 [size=256]
 Region 2: Memory at 90000000 (32-bit, non-prefetchable) [size=64K]
 Expansion ROM at f1000000 [size=128K]
 Capabilities: <access denied>

[Tormod Volden (tormodvolden)]

Luke, if it still doesn't work for you, it means that you have another bug than the one reported here. Please file a new bug, and if you can, please file a new bug upstream, where you attach a log from running the latest version.