Intel nm70 847 celeron sandy bridge frequent crashes

[esk1]

I have extremely frequent crashes with both 2.99.3 and 2.99.4. This happens about once every 10 min of video playback, but also occurs on the menu system. I have found that for some reason 2.99.2 has them a lot less frequently (specifically the 32-bit intel build). However even with that build I still get 1 every 6-8 hours. I know other user's experienced similar issues and rolled back to 2.0 but there is no 5.1 support for this chip set until recently.

[esk1]

Issue continues with 2.99.5, rolling back to 2.99.2 so i can at least get through an entire movie.

[esk1]

Issue continues with 3.00

[fritsch]

I don't know how I can help there - as there is no log at all. Please pastebin a log with debug turned on and after playing some files - it does not matter, if it really appears - i just want to check some settings.

[esk1]

My bad, I thought a debug log wouldn't do any good without the problem occurring:

http://pastebin.com/u6bJRkfc

Let me know if you need any more or if there is anything else I can do to get this issue resolved.

[fritsch]

That was with 2.99.2 - the 3.0.0 is much worse?

Cause we bumped at least twice the Intel driver version since then - the libva-i965 driver was the same for both. The crash just restarts xbmc, so basically the log gets emptied and xbmc is restarted?

Try to forbid xbmc to restart by doing:

touch /var/lock/xbmc.disabled

after bootup and trigger the crash (with debugging off - as you don't get it with debugging).

Afterwards supply the logfile and also /var/log/messages

thank you

[esk1]

It got worse after 2.99.2 so I leave my box on that release right now. I did roll forward to 3.0.0 for you and run a movie, it only took about 10 min before it crashed (in non-debug mode)

log: http://pastebin.com/VXSaMdi9 varlog: http://pastebin.com/TYHmWZ1Z

[fritsch]

You are jumping forth and back in that file? It does not really look good, but the kernel does not complain. Will look tomorrow.

Btw. to rule out Audio issues, set it to 2.0 and choose Analog and there choose your hdmi device.

[esk1]

I've actually done that already and another user is running 2.0, I'm pretty sure that is not the problem.

I was skipping around a bunch in the file at the start, sometimes the menu items cause the crash to happen faster (hitting info, etc). Then I walked away and let it play for a few minutes before it crashed.

[fritsch]

As it does not happen with debugging turned on, which produces and overlay and therefore influences the rendering, please try to crash it with this advancedsettings.xml - which does not show the overlay:

<advancedsettings>
<loglevel>1</loglevel>
</advancedsettings>

[esk1]

That worked, reproducible in about 10 min like before on 3.0.0. I greatly exceeded pastebin's limits so I'm using google docs. Let me know if you have any issues getting the log, I threw the varlog stuff at the end of the file.

log: https://docs.google.com/file/d/0BzA-dTqmqvrwOGdLZ3N4dGhhV1U/edit

[fritsch]

Oki, please the same again with Skin Confluence.

You are controlling it remoteley with an App?

[esk1]

I use an IR mainly, but I have been using the app to monitor the box so when it crashes so I can use another input on the TV. I have run it without the app interaction and still gotten the crash.

Updated with confluence: https://docs.google.com/file/d/0BzA-dTqmqvrwTzRPT2RxWTBXbDA/edit?usp=sharing

[fritsch]

Okay, login via ssh mount /flash read/write edit the syslinux.cfg and add there to the kernel parameters:

i915.i915_enable_rc6=0

And reboot afterwards. This will disable the chips power management. I am really out of ideas now.

Also pastebin: dmesg | pastebinit after the boot

[esk1]

Can you expand on this? I'm not sure how to do the mount

[fritsch]

mount -o remount,rw /flash

nano /flash/extlinux.conf

write before the quiet:

i915.i915_enable_rc6=0

ctl o (save) ctl x (exit) sudo reboot

[esk1]

Okay thanks, I think I got that working but is there anything in the log that would confirm it?

[fritsch]

dmesg | pastebinit

will tell

[esk1]

http://pastebin.com/ZjuJZguH http://pastebin.com/xGugwdxG

edit: found it, second link

double edit: just crashed again, do you want the log or is the above good enough?

[fritsch]

It did not work: Enabling RC6 states: RC6 on, RC6p off, RC6pp off

Your command line (/flash/syslinux.cfg) is wrong: root=/dev/ram0 rdinit=/init boot=LABEL=System disk=LABEL=Storage quiet BOOT_IMAGE=/KERNEL

it must be: root=/dev/ram0 rdinit=/init boot=LABEL=System disk=LABEL=Storage i915.i915_enable_rc6=0 quiet BOOT_IMAGE=/KERNEL

Correctly output after it is working: Enabling RC6 states: RC6 off, RC6p off, RC6pp off

I am out for the weekend

2013/3/30 esk1 notifications@github.com

http://pastebin.com/ZjuJZguH

Here is the varlog stuff. It's still running but its only been like 5 min so I'm not sure if its going to crash or not.

— Reply to this email directly or view it on GitHubhttps://github.com/OpenELEC/OpenELEC.tv/issues/2008#issuecomment-15675649 .

               Key-ID:     0x1A995A9B
               keyserver: pgp.mit.edu

Fingerprint: 4606 DA19 EC2E 9A0B 0157 C81B DA07 CF63 1A99 5A9B

[esk1]

I didn't have a syslinux.cfg before this, so the new line is the only thing In there.

[fritsch]

There is another file with .cfg ending i always mix them :-) Am 30.03.2013 18:21 schrieb "esk1" notifications@github.com:

I didn't have a syslinux.cfg before this, so the new line is the only thing In there.

— Reply to this email directly or view it on GitHubhttps://github.com/OpenELEC/OpenELEC.tv/issues/2008#issuecomment-15678169 .

[esk1]

Okay I got the "Enabling RC6 states: RC6 off, RC6p off, RC6pp off" by editing the extlinux.conf and adding that line to turn it off.

Its been running for about 20 min now without an issue so I'm going to use it for a day or two and see what happens.

[esk1]

No crashes for 5 hours which is probably a record, thank you very much fritsch, this issue has plagued me for months but I could never get any suggestions.

[fritsch]

@esk1:

I think I have some good news here. OpenELEC current upgrades to 3.8.3 kernel which has an explicit fix for this RC6 problem. It was not only plaguing you, but a whole lot people out there on Fedora / Ubuntu and so on. The bad thing now is, that you loose approx 5W, that could be powersaved. When there is a new image with the newer kernel to test - I will ping you somehow.

@sraue: i915 RC6 powersaving causes issues on at least SandyBridge Celerons on 3.7.10 - it should be fixed with our new 3.8.x kernel - so when an image is ready, I think @esk1 is happy to test.

@esk1: For now please tell the people in the forum, that it - kind of works - so that they can benefit from our testing. Thanks again for all the logfiles. We have at least found something, that helps to not trigger it every some minutes :-) nice.

[esk1]

I would be more than willing to test it, I'm not too worried about the 5W since this chipset runs at so low power anyways. This also may explain why during the install of xbmcunbuntu (I was only trying to because I was sick of the OE restarts) my box also would crash.

Thanks again, let me know when the new kernel is out and I'll give it a go

[fritsch]

This is an upstream intel bug.