kmod-11 rmmod --wait timeout

[twistedfall]

Kmod-11 deprecated the '--wait' argument to rmmod. There is now a intentional 10s pause when this switch is specified and the switch will go away in the future (together with the kernel-side support for it). Here is the LKML announcement: https://lkml.org/lkml/2012/11/8/1

Bumblebee uses rmmod with '--wait' switch to unload drivers after game session is done and thus timeouts in trying to do so:

Nov 08 14:19:40 razor bumblebeed[439]: rmmod: ERROR: 'Wait' behavior is targeted for removal from kernel.
Nov 08 14:19:40 razor bumblebeed[439]: We will now sleep for 10s, and then continue.
Nov 08 14:19:50 razor bumblebeed[439]: [15915.308601] [ERROR]Unloading nvidia driver timed out.

Although the driver unloads successfully after that.

I'm not sure what is the proposed solution on synchronous driver unloading, but it looks like something needs to be done on Bumblebee side to stay compatible with future kernels.

[Lekensteyn]

It was added in commit bbd3939, possibly because X does not die quick enough for removing the module. Looks like we have to use polling again?

[amonakov]

No, just remove the module without waiting and install udev rules to power on/off the nvidia card upon loading/unloading nvidia module via bbswitch.

[twistedfall]

Hm, right, as a side effect nvidia card stays turned on, I didn't notice it before.

[vixus0]

This issue has started recently affecting me on Arch, though I have no idea what changed.
So now the only way to get the nvidia card switched off on startup is to blacklist the nvidia module.

However when I run things with optirun, the card fails to switch off after optirun completes and I have to do so manually.

[tetris4]

I was affected by this as well on Chakra, the card would not turn off. The patch mentioned in the AUR comments seems to have fixed it for me.

[ArchangeGabriel]

This is going to be removed widely, with insctructions on adding the udev rule.

[antiaircraft]

Until the udev solution is implemented, would raising the Bumblebee daemon's internal timeout to more than 10s get rid of the issue where it fails to power down the nvidia card?

[ArchangeGabriel]

@Bumblebee-Project Could someone take care of fixing this issue? I'm not sure about what we need to do exactly.

[vixus0]

The issues I'm having currently:

• On boot, when the bbswitch module loads (with load_state=0) it deactivates the card but the card is reactivated again. In dmesg, I see three messages saying the card has been deactivated by bbswitch.
• On closing the laptop lid, the nvidia card is switched on, causing heat buildup. This started happening after a recent kernel upgrade to 3.7.4

[ArchangeGabriel]

@amonakov and @Lekensteyn, what is your opinion on this ? udev rule (if so, could one of you provide it ?), timer in daemon side ?

[gdevenyi]

Something has changed recently and the 3 second delay is not enough causing a timeout:
bumblebeed[1414]: Unloading nvidia driver timed out.

I can rmmod and disable the card manually after noticing it's still on

ii bumblebee-nvidia 3.2.1-4+xedgerssaucy1 amd64 NVIDIA Optimus support using the proprietary NVIDIA driver
ii nvidia-331 331.20-0ubuntu1xedgerssaucy1 amd64 NVIDIA binary Xorg driver, kernel module and VDPAU library
ii nvidia-settings-331 331.20-0ubuntu1xedgerssaucy1 amd64 Tool for configuring the NVIDIA graphics driver
ii bumblebee 3.2.1-4+xedgerssaucy1 amd64 NVIDIA Optimus support for Linux
ii bbswitch-dkms 0.7-2 amd64 Interface for toggling the power on nVidia Optimus video cards

Linux delta 3.12.2-pf-uksm #26 SMP PREEMPT Thu Dec 12 16:10:59 EST 2013 x86_64 x86_64 x86_64 GNU/Linux