Bug#804139: bumblebee: after recent xorg + nvidia updates in sid, "Cannot access secondary GPU - error"

[Luca Boccassi]

On Wed, 2016-04-27 at 09:57 +0200, ø wrote:
> > Hi,
> > 
> > I actually found that it works for me with a rootless Xorg, but the
> > legacy wrapper breaks it.
> > 
> > Do you have xserver-xorg-legacy installed?
> > 
> > Kind regards,
> > Luca Boccassi
> 
> Yes I had. I installed it when after upgrading to rootless Xorg on the
> firt time optirun ceased to work.
> 
> Great news is after removing it, it works again! Oh my God! I've been
> months with this, I never thought on uninstalling legacy xorg again.

Great!

I suspect it's due to the different path. I'll try to investigate, and
eventually add a Breaks: xserver-xorg-legacy if it can't be done.

Kind regards,
Luca Boccassi


signature.asc
Description: This is a digitally signed message part

Bug#804139: bumblebee: after recent xorg + nvidia updates in sid, "Cannot access secondary GPU - error"

[ø]

> Hi,
> 
> I actually found that it works for me with a rootless Xorg, but the
> legacy wrapper breaks it.
> 
> Do you have xserver-xorg-legacy installed?
> 
> Kind regards,
> Luca Boccassi

Yes I had. I installed it when after upgrading to rootless Xorg on the
firt time optirun ceased to work.

Great news is after removing it, it works again! Oh my God! I've been
months with this, I never thought on uninstalling legacy xorg again.

Many many thanks.

signature.asc
Description: This is a digitally signed message part

Bug#804139: bumblebee: after recent xorg + nvidia updates in sid, "Cannot access secondary GPU - error"

[Luca Boccassi]

On Tue, 2016-04-26 at 22:01 +0200, ø wrote:
> > Can you please check if you are running Xorg as root?
> 
> I'm executing the Xorg process with my user, not root. That was what I
> tried to explain in my message, that since my debian changed to
> rootless Xorg, bumblebee stopped to work. Sorry to not be able to
> phrase it accurately.
> 
> > 
> > [1] https://github.com/Bumblebee-Project/Bumblebee/issues/704
> 
> I have tried the patch suggested on that page [2] on my debian
> packages, but I see no change in the result.
> 
> [2] https://paste.fedoraproject.org/305435/02152145/

Hi,

I actually found that it works for me with a rootless Xorg, but the
legacy wrapper breaks it.

Do you have xserver-xorg-legacy installed?

Kind regards,
Luca Boccassi


signature.asc
Description: This is a digitally signed message part

Bug#804139: bumblebee: after recent xorg + nvidia updates in sid, "Cannot access secondary GPU - error"

[ø]

> Can you please check if you are running Xorg as root?

I'm executing the Xorg process with my user, not root. That was what I
tried to explain in my message, that since my debian changed to
rootless Xorg, bumblebee stopped to work. Sorry to not be able to
phrase it accurately.

> 
> [1] https://github.com/Bumblebee-Project/Bumblebee/issues/704

I have tried the patch suggested on that page [2] on my debian
packages, but I see no change in the result.

[2] https://paste.fedoraproject.org/305435/02152145/

signature.asc
Description: This is a digitally signed message part

Bug#804139: bumblebee: after recent xorg + nvidia updates in sid, "Cannot access secondary GPU - error"

[Luca Boccassi]

On Sat, 2016-01-09 at 16:29 +, Nohus wrote:
> Luca Boccassi :
> > The lspci bash snippet is not finding the discrete card, I guess because
> > it was off when reportbug was run.
> > 
> > Could you please manually turn the card on, check manually which PCI id
> > it has, and then paste the output of:
> > 
> > lspci -vvnn -s 
> 
> Here you are:
> > Also I see you are using virtualgl, does it make any difference if you
> > use primus as a bridge instead?
> 
> Initially I tried primus, then I tried virtualgl when that didn't work. No 
> difference between them.
> 
> Thank you.

Can't really see anything out of the ordinary.

The last thing I could think of, is that upstream doesn't seem to
support rootless Xorg yet, they are looking to add it to 4.0. [1]

Can you please check if you are running Xorg as root?

Kind regards,
Luca Boccassi

[1] https://github.com/Bumblebee-Project/Bumblebee/issues/704


signature.asc
Description: This is a digitally signed message part

Bug#804139: bumblebee: after recent xorg + nvidia updates in sid, "Cannot access secondary GPU - error"

[Nohus]

Luca Boccassi :
> The lspci bash snippet is not finding the discrete card, I guess because
> it was off when reportbug was run.
> 
> Could you please manually turn the card on, check manually which PCI id
> it has, and then paste the output of:
> 
> lspci -vvnn -s 

Here you are:

sudo lspci -vvnn -s 01:00.0
01:00.0 3D controller [0302]: NVIDIA Corporation GM107M [GeForce GTX 860M] 
[10de:1392] (rev a2)
Subsystem: Lenovo Device [17aa:3978]
Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- 
Stepping- SERR- FastB2B- DisINTx-
Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort- SERR- 
Capabilities: [600 v1] Vendor Specific Information: ID=0001 Rev=1 
Len=024 
Capabilities: [900 v1] #19

> Also I see you are using virtualgl, does it make any difference if you
> use primus as a bridge instead?

Initially I tried primus, then I tried virtualgl when that didn't work. No 
difference between them.

Thank you.

Bug#804139: bumblebee: after recent xorg + nvidia updates in sid, "Cannot access secondary GPU - error"

[Luca Boccassi]

On Mon, 2016-01-04 at 10:08 +0100, Debian stuff wrote:
> Happy new year, maintainer.
> 
>  
> > When the last upload was made to the repos I had no issues with the
> > latest drivers and the latest Bumblebee on Jessie. I wonder if it
> > might be due to an incompatibility with the newer version of Xorg in
> > Stretch/Sid?
> 
> Those are my suspicions too.

I upgraded the various Xorg packages on my system to the version in Sid
(still a Jessie install as at the moment I cannot upgrade my laptop to
Stretch, nor I can create a new partition), but I cannot reproduce this
problem with neither 340.96-3 nor 352.63-1.

In your reportbug output I can see these lines where we grep
in /etc/modprobe.d files:

/etc/modprobe.d/i915-kms.conf:# PABLO: creo que lo puse para el
bumblebee, pero ahora lo quito para intentar tirar directamente con
nvidia, que no meta el driver de intel y no me cree conflicto
/etc/modprobe.d/pablo-prohibir-intel.conf:# PABLO :evitar que los
drivers de intel entren en conflicto con los de nvidia

Do you have any particular custom blacklists/configurations for the
graphics modules?

Kind regards,
Luca Boccassi

Bug#804139: bumblebee: after recent xorg + nvidia updates in sid, "Cannot access secondary GPU - error"

[Luca Boccassi]

On Fri, 2016-01-01 at 21:17 +, Nohus wrote:
> Luca Boccassi :
> > Meanwhile, could you please send all the nvidia-related details of
> > your system by running:
> > 
> > reportbug --template nvidia-driver
> > 
> > and attaching the result?
> 
> Sure, here is the report: https://paste.nohus.eu/?g9YqEP

The lspci bash snippet is not finding the discrete card, I guess because
it was off when reportbug was run.

Could you please manually turn the card on, check manually which PCI id
it has, and then paste the output of:

lspci -vvnn -s 

Also I see you are using virtualgl, does it make any difference if you
use primus as a bridge instead?

Kind regards,
Luca Boccassi

Bug#804139: bumblebee: after recent xorg + nvidia updates in sid, "Cannot access secondary GPU - error"

[ø]

El vie, 08-01-2016 a las 00:48 +, Luca Boccassi escribió:
> In your reportbug output I can see these lines where we grep
> in /etc/modprobe.d files:
> 
> /etc/modprobe.d/i915-kms.conf:# PABLO: creo que lo puse para el
> bumblebee, pero ahora lo quito para intentar tirar directamente con
> nvidia, que no meta el driver de intel y no me cree conflicto
> /etc/modprobe.d/pablo-prohibir-intel.conf:# PABLO :evitar que los
> drivers de intel entren en conflicto con los de nvidia
> 
> Do you have any particular custom blacklists/configurations for the
> graphics modules?

I have the option modeset=1 for driver i915, because some tutorial
years ago told me to in order to bumblebee to work.

I wrote the second file, pablo-prohibir-intel.conf when I tried to avoid intel 
drivers to load, for testing purposes, but it's currently commented, so it 
isn't affecting anything more.

I don't think I have any other custom configuration for modules, apart from the 
modeset=1 that has always worked.

Many thanks for your efforts.

signature.asc
Description: This is a digitally signed message part

Bug#804139: bumblebee: after recent xorg + nvidia updates in sid, "Cannot access secondary GPU - error"

[Nohus]

Luca Boccassi :
> Meanwhile, could you please send all the nvidia-related details of
> your system by running:
> 
> reportbug --template nvidia-driver
> 
> and attaching the result?

Sure, here is the report: https://paste.nohus.eu/?g9YqEP

> Also, to narrow the scope, when the card is ON are you able to
> manually load the nvidia kernel modules with modprobe?

Yes, after turning the card on with "tee /proc/acpi/bbswitch <<< ON", I can 
load "nvidia" with modprobe successfully.

Bug#804139: bumblebee: after recent xorg + nvidia updates in sid, "Cannot access secondary GPU - error"

[Luca Boccassi]

On 31 December 2015 at 17:02, Nohus  wrote:
> I am affected by the same bug since I recently upgraded from Debian Jessie to 
> Stretch. Bumblebee worked previously. Now I get this problem, both on nouveau 
> and on non free nvidia drivers.
>
> Here are my logs:
>

Hi,

When the last upload was made to the repos I had no issues with the
latest drivers and the latest Bumblebee on Jessie. I wonder if it
might be due to an incompatibility with the newer version of Xorg in
Stretch/Sid?

I am away without my optimus laptop so I cannot try and
reproduce/debug right now. I will be able to investigate later next
week.

Meanwhile, could you please send all the nvidia-related details of
your system by running:

reportbug --template nvidia-driver

and attaching the result? (as a side note, it's probably worth
considering shipping the same bugscript from nvidia-driver in the
bumblebee packages too)

Also, to narrow the scope, when the card is ON are you able to
manually load the nvidia kernel modules with modprobe?

Thanks!

Kind regards,
Luca Boccassi

Bug#804139: bumblebee: after recent xorg + nvidia updates in sid, "Cannot access secondary GPU - error"

[Nohus]

I am affected by the same bug since I recently upgraded from Debian Jessie to 
Stretch. Bumblebee worked previously. Now I get this problem, both on nouveau 
and on non free nvidia drivers.

Here are my logs:

[nohus] nohus:(~)❯❯❯ optirun -vv --debug glxgears
[ 3527.143580] [DEBUG]Reading file: /etc/bumblebee/bumblebee.conf
[ 3527.143825] [DEBUG]optirun version 3.2.1 starting...
[ 3527.143837] [DEBUG]Active configuration:
[ 3527.143844] [DEBUG] bumblebeed config file: /etc/bumblebee/bumblebee.conf
[ 3527.143851] [DEBUG] X display: :8
[ 3527.143858] [DEBUG] LD_LIBRARY_PATH: 
[ 3527.143865] [DEBUG] Socket path: /var/run/bumblebee.socket
[ 3527.143872] [DEBUG] Accel/display bridge: auto
[ 3527.143878] [DEBUG] VGL Compression: proxy
[ 3527.143886] [DEBUG] VGLrun extra options: 
[ 3527.143893] [DEBUG] Primus LD Path: 
/usr/lib/x86_64-linux-gnu/primus:/usr/lib/i386-linux-gnu/primus:/usr/lib/primus:/usr/lib32/primus
[ 3527.594436] [DEBUG]Using auto-detected bridge virtualgl
[ 3528.051511] [INFO]Response: No - error: [XORG] (EE) 

[ 3528.051551] [ERROR]Cannot access secondary GPU - error: [XORG] (EE) 

[ 3528.051558] [DEBUG]Socket closed.
[ 3528.051573] [ERROR]Aborting because fallback start is disabled.
[ 3528.051579] [DEBUG]Killing all remaining processes.

In journalctl when running optirun:

Dec 31 15:53:14 nohus bumblebeed[719]: [ 4023.682085] [ERROR][XORG] (EE)
Dec 31 15:53:14 nohus bumblebeed[719]: [ 4023.682109] [ERROR][XORG] (EE)
Dec 31 15:53:14 nohus bumblebeed[719]: [ 4023.682116] [ERROR][XORG] (EE)
Dec 31 15:53:14 nohus bumblebeed[719]: [ 4023.682121] [ERROR][XORG] (EE)
Dec 31 15:53:14 nohus bumblebeed[719]: [ 4023.682127] [ERROR][XORG] (EE)
Dec 31 15:53:14 nohus bumblebeed[719]: [ 4023.682133] [ERROR]X did not start 
properly

Also, the GPU stays ON as reported by /proc/acpi/bbswitch after trying to run 
optirun, until turned OFF manually with "rmmod nouveau; tee /proc/acpi/bbswitch 
<<< OFF"

Bug#804139: bumblebee: after recent xorg + nvidia updates in sid, "Cannot access secondary GPU - error"

[Debian stuff]

Finishing December, I'm still suffering the same problem since that
update. Maybe it has something to do with the new Xorg being run on
userspace?

I add my information because it contains logs from bumblebee:

    $ sudo systemctl status bumblebeed
● bumblebeed.service - Bumblebee C Daemon
   Loaded: loaded (/lib/systemd/system/bumblebeed.service; enabled;
vendor preset: enabled)
   Active: active (running) since mié 2015-12-30 09:30:34 CET;
20min ago
 Main PID: 875 (bumblebeed)
   CGroup: /system.slice/bumblebeed.service
   └─875 /usr/sbin/bumblebeed

dic 30 09:30:34 arepa systemd[1]: Started Bumblebee C Daemon.
dic 30 09:31:47 arepa bumblebeed[875]: [   96.854001]
[INFO]/usr/sbin/bumblebeed 3.2.1 started

$ optirun -vv glxgears
[ 1316.366496] [DEBUG]Reading file: /etc/bumblebee/bumblebee.conf
[ 1316.367218] [DEBUG]optirun version 3.2.1 starting...
[ 1316.367276] [DEBUG]Active configuration:
[ 1316.367297] [DEBUG] bumblebeed config file:
/etc/bumblebee/bumblebee.conf
[ 1316.367316] [DEBUG] X display: :8
[ 1316.367335] [DEBUG] LD_LIBRARY_PATH: /usr/lib/x86_64-linux-
gnu/nvidia:/usr/lib/i386-linux-gnu/nvidia:/usr/lib/nvidia
[ 1316.367354] [DEBUG] Socket path: /var/run/bumblebee.socket
[ 1316.367372] [DEBUG] Accel/display bridge: auto
[ 1316.367390] [DEBUG] VGL Compression: proxy
[ 1316.367408] [DEBUG] VGLrun extra options:
[ 1316.367426] [DEBUG] Primus LD Path: /usr/lib/x86_64-linux-
gnu/primus:/usr/lib/i386-linux-
gnu/primus:/usr/lib/primus:/usr/lib32/primus
[ 1316.367564] [DEBUG]Using auto-detected bridge primus
[ 1316.374559] [INFO]Response: No - error: [XORG] (EE)

[ 1316.374600] [ERROR]Cannot access secondary GPU - error: [XORG]
(EE)

[ 1316.374620] [DEBUG]Socket closed.
[ 1316.374649] [ERROR]Aborting because fallback start is disabled.
[ 1316.374672] [DEBUG]Killing all remaining processes.

$ sudo systemctl status bumblebeed
● bumblebeed.service - Bumblebee C Daemon
   Loaded: loaded (/lib/systemd/system/bumblebeed.service; enabled;
vendor preset: enabled)
   Active: active (running) since mié 2015-12-30 09:30:34 CET;
22min ago
 Main PID: 875 (bumblebeed)
   CGroup: /system.slice/bumblebeed.service
   └─875 /usr/sbin/bumblebeed

dic 30 09:52:07 arepa bumblebeed[875]: [ 1316.374416] [ERROR][XORG]
(EE)
dic 30 09:52:07 arepa bumblebeed[875]: [ 1316.374445] [ERROR][XORG]
(EE)
dic 30 09:52:07 arepa bumblebeed[875]: [ 1316.374452] [ERROR][XORG]
(EE)
dic 30 09:52:07 arepa bumblebeed[875]: [ 1316.374456] [ERROR][XORG]
(EE)
dic 30 09:52:07 arepa bumblebeed[875]: [ 1316.374461] [ERROR][XORG]
(EE)
dic 30 09:52:07 arepa bumblebeed[875]: [ 1316.374465] [ERROR]X did
not start properly

Bug#804139: bumblebee: after recent xorg + nvidia updates in sid, "Cannot access secondary GPU - error"

[Reiner Herrmann]

I recently installed bumblebee (with nouveau) and have the same problem:

> $ optirun -vv glxgears
> [13350.435409] [DEBUG]Reading file: /etc/bumblebee/bumblebee.conf
> [13350.435715] [DEBUG]optirun version 3.2.1 starting...
> [13350.435724] [DEBUG]Active configuration:
> [13350.435729] [DEBUG] bumblebeed config file:
> /etc/bumblebee/bumblebee.conf
> [13350.435733] [DEBUG] X display: :8
> [13350.435737] [DEBUG] LD_LIBRARY_PATH: 
> [13350.435742] [DEBUG] Socket path: /var/run/bumblebee.socket
> [13350.435746] [DEBUG] Accel/display bridge: auto
> [13350.435750] [DEBUG] VGL Compression: proxy
> [13350.435754] [DEBUG] VGLrun extra options: 
> [13350.435758] [DEBUG] Primus LD Path:
> /usr/lib/x86_64-linux-gnu/primus:/usr/lib/i386-linux-gnu/primus:/usr/lib/primus:/usr/lib32/primus
> [13350.435791] [DEBUG]Using auto-detected bridge primus
> [13350.438641] [INFO]Response: No - error: [XORG] (EE) 
> 
> [13350.438660] [ERROR]Cannot access secondary GPU - error: [XORG] (EE) 
> 
> [13350.438668] [DEBUG]Socket closed.
> [13350.438681] [ERROR]Aborting because fallback start is disabled.
> [13350.438689] [DEBUG]Killing all remaining processes.



signature.asc
Description: Digital signature

Bug#804139: bumblebee: after recent xorg + nvidia updates in sid, "Cannot access secondary GPU - error"

[Giacomo Mulas]

Package: bumblebee
Version: 3.2.1-10
Severity: important

Dear Maintainer,

I routinely run my laptop on sid (I need to compile scientific software
using recent versions of just too many libraries), usually with no problems,
or simple ones that a relatively old hand at debian system administration
can solve easily.
However, recently I was hit by nasty bug that resulted from a combination of
package updates, so I don't know exactly which package is to blame. I am
filing the bug raport on bumbleee because I was able to work around the rest
of the problems, but maybe it's not bumblebee's fault. If you can determine
that some other specific package is to blame, do feel free to move the
bugreport to that package.

In short: after some recent upgrade of xorg and nvidia, glx support was
completely broken on my system. It stopped working both on my primary Intel
graphic card and on the nvidia optimus. This caused a mess, since I used
gnome and gdm3, and I was left with no working display manager, and no easy
way to enable network connections via network-manager.
However, after some manual configuration, I got the network on, and was able
to run lightdm and the mate desktop (thanks heaven there are still some
desktop environments that don't depend on glx!).

Thinking that the glx alternatives may have been messed up by the upgrade, I
removed completely all nvidia packages from my system, and also
glx-alternatives itself.  This meant also removing all nvidia dependencies,
including optimus.  This fixed glx on my primary display.  So now I had
again working gdm3 and gnome.  Then I went on to reinstall bumblebee and
nvidia packages.  The install itself went smoothly, with no reported errors. 
The nvidia-current kernel module got properly compiled and installed by
dkms, everything looked fine.  And glx on my primary display was _not_
broken again by installing nvidia packages, glx-alternatives, etc.

However, bumblebee still does not work, so I have no accelerated graphics.
If I try e.g. optirun -vv glxinfo, I get:

gmulas@capitanata:~$ optirun -vvv glxinfo
[54329.188696] [DEBUG]Reading file: /etc/bumblebee/bumblebee.conf
[54329.189560] [DEBUG]optirun version 3.2.1 starting...
[54329.189614] [DEBUG]Active configuration:
[54329.189633] [DEBUG] bumblebeed config file: /etc/bumblebee/bumblebee.conf
[54329.189652] [DEBUG] X display: :8
[54329.189670] [DEBUG] LD_LIBRARY_PATH:
/usr/lib/x86_64-linux-gnu/nvidia:/usr/lib/i386-linux-gnu/nvidia:/usr/lib/nvidia
[54329.189689] [DEBUG] Socket path: /var/run/bumblebee.socket
[54329.189708] [DEBUG] Accel/display bridge: auto
[54329.189727] [DEBUG] VGL Compression: proxy
[54329.189745] [DEBUG] VGLrun extra options: 
[54329.189762] [DEBUG] Primus LD Path:
/usr/lib/x86_64-linux-gnu/primus:/usr/lib/i386-linux-gnu/primus:/usr/lib/primus:/usr/lib32/primus
[54329.189861] [DEBUG]Using auto-detected bridge primus
[54329.236079] [INFO]Response: No - error: [XORG] (EE) 

[54329.236092] [ERROR]Cannot access secondary GPU - error: [XORG] (EE) 

[54329.236095] [DEBUG]Socket closed.
[54329.236106] [ERROR]Aborting because fallback start is disabled.
[54329.236111] [DEBUG]Killing all remaining processes.

Nothing useful is dumped to kern.log, and in syslog I get only

Nov  5 11:20:13 capitanata bumblebeed[16456]: [54358.478063] 
[INFO]/usr/sbin/bumblebeed 3.2.1 started
Nov  5 09:46:49 capitanata kernel: bbswitch: disabling discrete graphics
Nov  5 09:46:49 capitanata kernel: ACPI Warning: \x5c_SB_.PCI0.PEG0.PEGP._DSM: 
Argument #4 type mismatch - Found [Buffer], ACPI requires [Package] 
(20150619/nsarguments-95)
Nov  5 09:46:49 capitanata kernel: pci :01:00.0: Refused to change power 
state, currently in D0

which may be related to another reported bug, maybe?

Please let me know if there is something I can do to provide more info to
nail the problem. Also, if you can suggest some way to make at least cuda
and opencl work, with or without bumblebee, and with or without graphics
acceleration, that would be most useful (I test software on my laptop and
now I can't do this with those that need GPU acceleration).

Thanks in advance, bye
Giacomo Mulas



-- System Information:
Debian Release: stretch/sid
  APT prefers unstable
  APT policy: (401, 'unstable'), (10, 'experimental')
Architecture: amd64 (x86_64)
Foreign Architectures: i386

Kernel: Linux 4.2.5-jak (SMP w/4 CPU cores)
Locale: LANG=it_IT.UTF-8, LC_CTYPE=it_IT.UTF-8 (charmap=UTF-8) (ignored: LC_ALL 
set to it_IT.UTF-8)
Shell: /bin/sh linked to /bin/dash
Init: systemd (via /run/systemd/system)

Versions of packages bumblebee depends on:
ii  bbswitch-dkms0.8-3
ii  bbswitch-source  0.8-3
ii  dpkg 1.18.3
ii  init-system-helpers  1.24
ii  libbsd0  0.7.0-2
ii  libc62.19-22
ii  libglib2.0-0 2.46.1-2
ii  libx11-6 2:1.6.3-1

Versions of packages bumblebee recommends:
ii  primus  0~20150328-1

Versions of packages bumblebee suggests:
ii  bumblebee-nvidia  3.2.1-10

-- no