Bug 352867 - sax2: cannot configure NVIDIA dual GPU cards
Summary: sax2: cannot configure NVIDIA dual GPU cards
Status: RESOLVED DUPLICATE of bug 381644
: 352647 396543 403366 (view as bug list)
Alias: None
Product: openSUSE 10.3
Classification: openSUSE
Component: SaX2 (show other bugs)
Version: Final
Hardware: x86-64 openSUSE 10.3
: P2 - High : Enhancement (vote)
Target Milestone: ---
Assignee: Stefan Dirsch
QA Contact: E-mail List
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2008-01-10 11:58 UTC by Forgotten User Udskk56uJ3
Modified: 2008-06-24 20:07 UTC (History)
7 users (show)

See Also:
Found By: Third Party Developer/Partner
Services Priority:
Business Priority:
Blocker: ---
Marketing QA Status: ---
IT Deployment: ---


Attachments
Error no screen found for sax2 ..... (10.34 KB, text/x-log)
2008-01-10 11:58 UTC, Forgotten User Udskk56uJ3
Details
Error for configure (Ups lost card( with sax2 -r -m ... (896 bytes, text/x-log)
2008-01-10 12:00 UTC, Forgotten User Udskk56uJ3
Details
Configuration hangs (129.10 KB, text/x-log)
2008-01-10 12:01 UTC, Forgotten User Udskk56uJ3
Details
Working configuration, but only one screen created (4.00 KB, text/x-log)
2008-01-10 12:23 UTC, Forgotten User Udskk56uJ3
Details
includes SaX.log (43.33 KB, application/x-gzip-compressed)
2008-01-10 14:20 UTC, Forgotten User Udskk56uJ3
Details
Requested Data (27.16 KB, application/x-gzip-compressed)
2008-01-14 14:33 UTC, Forgotten User Udskk56uJ3
Details
requested xorg.conf ( #12 ) (2.75 KB, text/plain)
2008-01-17 09:25 UTC, Forgotten User Udskk56uJ3
Details

Note You need to log in before you can comment on or make changes to this bug.
Description Forgotten User Udskk56uJ3 2008-01-10 11:58:28 UTC
Created attachment 190047 [details]
Error no screen found for sax2 .....

All attempts to configure a NVIDIA Quadro NVS 440 card (dualhead) with sax2 failed. Configuration with nvidia-xconfig -a work, but more testing should be done. Single head configuration works, but wrong BusID is selected.
Comment 1 Forgotten User Udskk56uJ3 2008-01-10 12:00:17 UTC
Created attachment 190048 [details]
Error for configure (Ups lost card( with sax2 -r -m ...
Comment 2 Forgotten User Udskk56uJ3 2008-01-10 12:01:59 UTC
Created attachment 190049 [details]
Configuration hangs
Comment 3 Forgotten User Udskk56uJ3 2008-01-10 12:23:18 UTC
Created attachment 190050 [details]
Working configuration, but only one screen created

The configuration is basically okay, but ...
xorg.conf needs manual editing, because wrong card was selected.
BusID 8:0:0 was used, but it should be BusID 7:0:0
Comment 4 Marcus Schaefer 2008-01-10 12:49:52 UTC
as I said you have two cards one is at 7 the other at 8

sax2 -c 0   one screen
sax2 -c 1   the other screen

I'm sorry if both screens together doesn't work this is a driver
problem. I can't help you I'm sorry
Comment 5 Forgotten User Udskk56uJ3 2008-01-10 12:57:51 UTC
Hello Marcus,

Please pay attention it is one card with 2 gpus.
nvidia-xconfig works and have also a look at 352647,
under SLED10 SP1 Novell generates with the same driver a xorg.conf 
for dualhead ( which unfortunately does not work ).
Comment 6 Marcus Schaefer 2008-01-10 13:11:03 UTC
there are two devices with separate device ID's no matter if they are on
one card or not. SLED for sure doesn't use the same driver because the
kernel is different.

Anyway if you don't mind call:

  init 3
  sax2

--> send the log file /var/log/SaX.log

thanks
Comment 7 Forgotten User Udskk56uJ3 2008-01-10 14:20:11 UTC
Created attachment 190079 [details]
includes SaX.log

Beside the requested /var/log/SaX.log also Xorg.0.log and Xorg.90.log are included.
Comment 8 Marcus Schaefer 2008-01-10 14:36:57 UTC
and what is the problem: you should see two displays one at 800x600
the other at 1024x768

(II) NVIDIA(0): Setting mode "800x600"
(--) NVIDIA(0): Connected display device(s) on Quadro NVS 440 at PCI:8:0:0:
(--) NVIDIA(0):     CRT-0

==> maybe no monitor connected or not detectable via DDC

(--) NVIDIA(1): Connected display device(s) on Quadro NVS 440 at PCI:7:0:0:
(--) NVIDIA(1):     FUS B17-2 (DFP-0)
(II) NVIDIA(1): Setting mode "1280x1024"

==> DDC ID FUSB172 (looks like a Fujitsu Siemens monitor)

looks good to me

???
Comment 9 Forgotten User Udskk56uJ3 2008-01-10 15:01:33 UTC
Sorry, the send log is not okay -
what I should have sent is an output for
sax2 -r

Unfortunately I am running now a test.
On Monday I will generate a new log.

Winfrid
Comment 10 Marcus Schaefer 2008-01-14 13:34:44 UTC
*** Bug 352647 has been marked as a duplicate of this bug. ***
Comment 11 Forgotten User Udskk56uJ3 2008-01-14 14:33:43 UTC
Created attachment 190413 [details]
Requested Data

I have now rerun the command.
Only sax2 -r -a works, in case of omitting the option -a
the display goes black and my only chance was to reboot the system.
With sax2 -r -a the order of screens is changed.

The behaviour is similar to 352647 but not always identical.
Comment 12 Marcus Schaefer 2008-01-14 15:02:06 UTC
Now I understand this is a card with two cpu's but they don't appear as
single cards on the bus. Well I think we never took care that this was
configured correctly. I also don't have such a card.

* Stefan you already mentioned that we can't configure dual cpu cards 
  Do we have such a card ?

* Winfrid you said nvidia-xconfig can configure the card correctly.
  So can you send the xorg.conf created by the tool ?

* Winfrid you said the wrong busID was selected but the only class 0x300
  device in the PCI stack is 08-00-0. The binary nvidia driver see two
  devices yes but the linux kernel doesn't or I overlooked it in the logs
  Stefan could you comment here ?

I don't think this is a bug but more a missing feature
Comment 13 Stefan Dirsch 2008-01-14 16:03:06 UTC
Marcus, unfortunately we don't have such a card. In the Xserver log I can see two devices with class 300.

(II) PCI: 07:00:0: chip 10de,014a card 10de,0358 rev a2 class 03,00,00 hdr 00
(II) PCI: 08:00:0: chip 10de,014a card 10de,0358 rev a2 class 03,00,00 hdr 00

SaX2 detects and configures two devices as well.

10-Jan 15:09:28 <I> [ Sysp: Server detection data
        Card0     =>  DomainId   : 0x0
        Card0     =>  BusId      : 0x7
        Card0     =>  SlotId     : 0x00
        Card0     =>  FuncId     : 0x0
        Card0     =>  Vendor     : NVidia
        Card0     =>  Device     : Quadro NVS 440
        Card0     =>  VID        : 0x10de
        Card0     =>  DID        : 0x014a
        Card0     =>  Module     : nv
        Card0     =>  BusType    : AGP
        Card0     =>  Detected   : 2
        Card0     =>  Flag       : DEFAULT
        Card0     =>  SUB-VID    : 0x10de
        Card0     =>  SUB-DID    : 0x0358
        Card0     =>  DrvProfile : Depth24,NVidia
        
        Card1     =>  DomainId   : 0x0
        Card1     =>  BusId      : 0x8
        Card1     =>  SlotId     : 0x00
        Card1     =>  FuncId     : 0x0
        Card1     =>  Vendor     : NVidia
        Card1     =>  Device     : Quadro NVS 440
        Card1     =>  VID        : 0x10de
        Card1     =>  DID        : 0x014a
        Card1     =>  Module     : nv
        Card1     =>  BusType    : AGP
        Card1     =>  Detected   : 2
        Card1     =>  Flag       : DEFAULT
        Card1     =>  SUB-VID    : 0x10de
        Card1     =>  SUB-DID    : 0x0358
        Card1     =>  DrvProfile : Depth24,NVidia


        Section "Device"
          BoardName    "Quadro NVS 440"
          BusID        "7:0:0"
          Driver       "nvidia"
          Identifier   "Device[0]"
          VendorName   "NVidia"
        EndSection
        
        Section "Device"
          BoardName    "Quadro NVS 440"
          BusID        "8:0:0"
          Driver       "nvidia"
          Identifier   "Device[1]"
          VendorName   "NVidia"
        EndSection

--> comment #7.
Comment 14 Marcus Schaefer 2008-01-14 16:07:44 UTC
but not comment #11

I'm maximal confused
Comment 15 Stefan Dirsch 2008-01-14 16:19:20 UTC
Right. In comment #11 SaX2 only detects one device. The Xserver PCI scan detects two devices though. Don't know why.
Comment 16 Stefan Dirsch 2008-01-15 04:43:22 UTC
Would it be possible that FSC sends us such a card short-term? This would make things much more easier to investigate. Location is Nürnberg, Germany.
Comment 18 Forgotten User Udskk56uJ3 2008-01-17 09:25:14 UTC
Created attachment 190808 [details]
requested xorg.conf (  #12 )

Hello Stefan,

Don't be confused the xorg.conf was generated on 
a Celsius V840 running SLED10 SP1 64bit.

If NVidia does not contact you until Jan. 23 -
you should give me a note.

Winfrid
Comment 19 Marcus Schaefer 2008-01-17 10:22:37 UTC
I can't see something special in the file. As I said in my last comment I'm
confused why one sax log find two devices while another sax log on the same
machine doesn't. I guess this one here can only be solved if we have such
a card. Stefan could you drop me a note if you got the card ?

thanks
Comment 20 Stefan Dirsch 2008-01-17 10:47:35 UTC
I agree. I'll let you know when/if the card arrives in the next two weeks.
Comment 21 Marcus Schaefer 2008-01-25 13:34:23 UTC
==> set to later
Comment 22 Stefan Dirsch 2008-02-08 10:26:27 UTC
The card just arrived.
Comment 23 Stefan Dirsch 2008-04-02 13:42:25 UTC
Finally I got the promised 2 NVS 440 (PCIE + PCIX16) from NVIDIA.
Comment 24 Stefan Dirsch 2008-04-11 20:21:52 UTC
Since I finally have the hardware for testing now I plan to look at this issue
in the next week. For this it would be good to know what the requirements/expectations are from your side for SaX2 for creating the configuration for NVS 440 cards. This should include monitor setup

- how many monitors connected to the graphics card
- connected to which GPU/DMS-59 connector
- connected via VGA or DVI-D

Possibly we need a special profile for this card.
Comment 25 Forgotten User Udskk56uJ3 2008-04-14 12:11:11 UTC
Hello Stefan,

I informed Rainer to provide you the requested information.

Winfrid
Comment 26 Forgotten User Udskk56uJ3 2008-04-16 07:21:36 UTC
Hello Stefan,

Maybe it was not really clear from the inital issue -
my minimum expectation was, that a graphical installation with
one display works - the situation at the time when I opened the incident, 
was -
all steps of the graphical installation worked until the display is 
configured, after this the display is "black". 

All other requirements should be discussed with Rainer König.

Winfrid
Comment 27 Stefan Dirsch 2008-04-16 07:28:19 UTC
Thanks for clarification. But even for this minimal singlehead configuration I need to know how the hardware setup looks like.
Comment 28 Daniel Rahn 2008-04-16 07:38:40 UTC
I might be wrong here, but the minimal requirement is to have a working single-head display on the first connector. 

Getting dual-head right comes on top of that.

Don't you agree?
Comment 29 Stefan Dirsch 2008-04-16 09:59:05 UTC
Yes, but SaX2/driver behaviour might be completely different when connecting more than one monitor to the graphics card. Therefore I still would prefer to know the detailed hardware setup.
Comment 31 Stefan Dirsch 2008-04-18 14:19:54 UTC
I'll give it a try using openSUSE 11.0 Beta1 with a single-head configuration - monitor connected to the first connected via DVI-D.
Comment 32 Stefan Dirsch 2008-04-18 14:56:12 UTC
# sax2 -p
Chip: 0  is -> VESA Framebuffer Graphics        04:00:0 0x10de 0x014a AGP fbdev
Chip: 1  is -> VESA Framebuffer Graphics        05:00:0 0x10de 0x014a AGP fbdev

# sax2 -r
SaX: initializing please wait...
SaX: your current configuration will not be read in

SaX: no X-Server is running
SaX: will start own server if needed
SaX: ups lost card during probing... abort
SaX: something went wrong while X was called with -probeonly
SaX: try to call 'sax2 -p' and select a single device ?
abort

/var/log/SaX.log
[...]
18-Apr 16:51:14 <I> [ Sysp: XStuff detection data
        SaX: ups lost card during probing... abort
]

Using 

  sax2 -r -c 0 m 0=nv

works fine, but I'm not sure how to achieve this by a profile. Marcus, any hints? Machine is f29.
Comment 33 Stefan Dirsch 2008-04-18 15:05:42 UTC
xorg-x11-driver-video.changes
-------------------------------------------------------------------
Fri Apr 18 17:05:41 CEST 2008 - sndirsch@suse.de

- added NVIDIA Quadro NVS 440 Device ID to sax2/hwinfo data
[...]
Comment 34 Stefan Dirsch 2008-04-18 17:24:13 UTC
It looks very similar when using the nvidia driver:

# sax2 -p
Chip: 0  is -> NVIDIA Quadro NVS 440           04:00:0 0x10de 0x014a AGP nvidia
Chip: 1  is -> NVIDIA Quadro NVS 440           05:00:0 0x10de 0x014a AGP nvidia

# sax2 -r
SaX: initializing please wait...
SaX: your current configuration will not be read in

SaX: no X-Server is running
SaX: will start own server if needed
SaX: ups lost card during probing... abort
SaX: something went wrong while X was called with -probeonly
SaX: try to call 'sax2 -p' and select a single device ?
abort

/var/log/SaX.log
18-Apr 19:19:41 <I> [ Sysp: XStuff detection data
        SaX: ups lost card during probing... abort
]

The workaround would be the same:

  sax2 -r -c 0

So here we would need a special profile as well.

Comment 35 Marcus Schaefer 2008-04-21 09:20:08 UTC
I'm sorry I don't have time to take care for multi card support.
Especially in all of these cases I end up with X-Server problems
which are really annoying and all I found is what I wrote in the
past what has worked before doesn't work anymore. It would be really
really really the best thing if Xorg would support and maintain
their own configuration tool as soon as possible
Comment 37 Stefan Dirsch 2008-04-21 09:42:01 UTC
There has been a misunderstanding. I'll do some more investigation on this issue.
Comment 39 Stefan Dirsch 2008-05-08 09:32:42 UTC
Implementing enhancement Bug #381644 would resolve this issue.
Comment 40 Stefan Dirsch 2008-06-03 05:56:04 UTC
*** Bug 396543 has been marked as a duplicate of this bug. ***
Comment 41 Stefan Dirsch 2008-06-17 12:38:41 UTC
Same problem with a GeForce 9800 GX2.
Comment 42 Stefan Dirsch 2008-06-20 19:22:38 UTC
*** Bug 402164 has been marked as a duplicate of this bug. ***
Comment 43 Stefan Dirsch 2008-06-24 18:31:25 UTC
*** Bug 403366 has been marked as a duplicate of this bug. ***
Comment 44 Stefan Dirsch 2008-06-24 20:07:01 UTC

*** This bug has been marked as a duplicate of bug 381644 ***