[solved in 3.5.0] Systemfreezes with 3.4.4

Manu311 · Tux's lil' helper Joined: 17 Nov 2010 Posts: 128

Hi,

I know that kernel is masked unstable, and I know I may have done something wrong configuring it, anyways I would like to fix this if it's on my side but I have no idea how, so here's my problem:

Recently I got new hardware (which isn't the cause of this, I remove it and it didn't solve anything). So I reconfigured my kernel and updated to the newest in portage (as I said, this may not be wise, but I got a safe kernel in the backhand) which is 3.4.4 and NOT masked stable (yet).

So here's my problem:
When I boot up my pc it starts without warnings or errors or anything. But within of 10 seconds after the login appeared (shell) the system completely hungs up. I've managed to tail -f /var/log/messages before that happened, nothing appeared there. Neither in any other log file in /var/log.
So this does not happen all of the time, just around 80% of my system boots - if it doesn't freeze everything is running fine. I've unplugged every single device and it still happens.
So is there anything known to cause this? Or a bug in the kernel? Or how do I get more informations out of that?
I think there are options inside the kernel config that could print more output, but I don't know which could be helpful in this case.

I know I could fix this with downgrading back to my old kernel, and I know it's likely to be my own fault, so no need to mention that, just help me with my questions if you're able to, thanks.

Logicien · Posted: Wed Jul 11, 2012 12:37 am Post subject:

Hello,
I have been interested by your message because I use the ~amd64 3.4.4 kernel with no freeze at all since the beginning. I suppose my kernel configuration is good, because all of the previous versions I used with this configuration haven't freeze the system.

I changed some things with time but now the only thing I can do now is to move the Radeon Dri from kernel to module and use an initramfs to load the appropriate firmwares files. Both, Radeon Dri and the firmwares are in the kernel at this time and KMS is working well.

I had to use the Vesafb framebuffer with Ati-drivers before, because I had a black screen with Radeon KMS. With more recents Linux kernel and firmwares versions, KMS became usable with Radeon Dri.

I just want to say that your problem can come from something else than the kernel configuration. Have you try to use the Magic SysRq keys to reboot or shutdown Gentoo when the freezes happen? Have you try to boot directly to X to see if the zreezes only arrived in text mode or in graphic mode to?

I give you my kernel configuration as reference: .config.
_________________
Paul

Logicien · Posted: Wed Jul 11, 2012 12:38 am Post subject:

Hello,
I have been interested by your message because I use the ~amd64 3.4.4 kernel with no freeze at all since the beginning. I suppose my kernel configuration is good, because all of the previous versions I used with this configuration haven't freeze the system.

I changed some things with time but now the only thing I can do now is to move the Radeon Dri from kernel to module and use an initramfs to load the appropriate firmwares files. Both, Radeon Dri and the firmwares are in the kernel at this time and KMS is working well.

I had to use the Vesafb framebuffer with Ati-drivers before, because I had a black screen with Radeon KMS. With more recents Linux kernel and firmwares versions, KMS became usable with Radeon Dri.

I just want to say that your problem can come from something else than the kernel configuration. Have you try to use the Magic SysRq keys to reboot or shutdown Gentoo when the freezes happen? Have you try to boot directly to X to see if the zreezes only arrived in text mode or in graphic mode to?

I give you my kernel configuration as reference: .config.
_________________
Paul

Manu311 · Tux's lil' helper Joined: 17 Nov 2010 Posts: 128

I've sshed into from my notebook which got disconnected when the pc freezed. I've managed to start X in time, anyways everything freezed (including the mouse).
Sure it COULD be something else then the kernel, but I haven'T changed anything else recently. Actually not even the kernel was reconfigured, I just used the same modules with the same options as far as possible. Nothing new, nothing less.
In addition I'm using the nvidia driver AND efi as Framebuffer. Since I'm not using any boot manager this is nessesessary. Anyways, I haven't changed any boot options and have used this method to boot since it was in an unstable kernel (I don't remember which it was) and it never had problems (after I've managed to configure it).

Manu311 · Tux's lil' helper Joined: 17 Nov 2010 Posts: 128

so there's news in my case.
First of all, when I last shutdown my pc, I still had mpd running - which is restarted at boot time (runlevel: default) and automatically starts playing the same song I've been playing when I've shutdown at the same position.
So I've recognized the systemfreeze is at 100% the same time (it's easy to hear since the last thing played before the freeze is looped infinitly). Also I've simply killed mpd via ssh which did not help a bit.
Sadly I'm currently having this issue 100% of the time, so I'll be required to boot my old, safe kernel.
I'm still without a clue, but maybe some1 else now got an idea.

//EDIT:
Alright, now I found some more messages in /var/log/message. And after my 20th (seriously) boot it booted so I got the diff - after my network connection the changes begin, while if it's booting 1 sec after network boot it telling me about my ssh-connection (since I sshed in to get the logs, I haven't expected a successful boot this time) and nothing beyond that.
While if it freezes the following line is repeated (starting 3 secs after network went online):

Manu311 · Tux's lil' helper Joined: 17 Nov 2010 Posts: 128

Well this seems to help as long as I don't have plugged in any usb3 device - meaning it still happens if the driver is needed. I haven't tested this much, anyways it seem as this ain't my fault, rather one in kernel development, I'll get a newer kernel soon to test if they fixed it recently.

ultraslinky · Posted: Fri Jul 13, 2012 6:50 pm Post subject:

Maybe your best bet is to try the release candidates (git-sources).
I have always had horrible experience with USB3 on linux, especially with older 3.2 kernels.

Also, you could try blacklisting the USB3 kernel module, or recompiling the kernel without it? If that solves it, you have narrowed down the problem to USB3

Manu311 · Tux's lil' helper Joined: 17 Nov 2010 Posts: 128

I know you're right - but as I said I already have ways to move around it - at least I tracked it down to one driver (thought I'm not 100% sure of this). I just want to know if I should report this or if it is already known. But I guess since it's an experimental driver I should assume it.

Anyways I'll update here if I get a solution or get closer to one. If some1 else got one, just tell me and I'll try (if not too much overkill).

Manu311 · Tux's lil' helper Joined: 17 Nov 2010 Posts: 128

I've blacklisted the usb3 modul but I've experienced the same freezes with the usb2 module. It's exactly the same output but it's xhci instead of ehci this time.
The only way I've figured out how I'm able to run my system without this freeze is with simply unplugging my (only) usb3 device and wait until the system booted up. Replugin never caused any freeze.

I'm currently recompiling the mainline-kernel and if the problem doesn't solve itself I'll report it to kernel.org bugzilla. I think that'll be the better place for this.

//EDIT:
Solved - 3.5.0 has this issue solved, I haven't searched for which change did it, but I've got 3 system boots now with 3.5.0 (I haven't changed even 1 config option) and no freeze.