[Serafean]

Hi,

I'm using xbmc as a media center and as an upnp streaming server. And I get random, albeit regular, crashes. I haven't managed to track down a behaviour that could trigger it, but when I leave xbmc idling, its less crash-prone...

log

It also generates an output in dmesg :

Code:

[329674.386900] xbmc.bin[7987] general protection ip:7f89ebe71a08 sp:7f89e66a0e40 error:0 in libnvidia-tls.so.290.06[7f89ebe71000+3000]
[350027.618103] xbmc.bin[8926]: segfault at 6e62742e ip 00007fe958e4ba08 sp 00007fe94f625e40 error 4 in libnvidia-tls.so.290.06[7fe958e4b000+3000]

I'd like to say that the second line was generated by this crash, but I'm not 100% sure.

running xbmc git 5b3b0b3fb481e4a4443c905a303f56af7bcfb304
nvidia drivers 290.06

Any ideas? any more steps I could do to help fixing this?

Thanks, serafean

[Cheule]

I've had mine segfault a few times too. This is quite serious, XBMC is attempting to overwrite memory allocated to another running program. As you have witnessed the result is a bit of a disaster.

[davilla]

I've had mine segfault a few times too. This is quite serious, XBMC is attempting to overwrite memory allocated to another running program. As you have witnessed the result is a bit of a disaster.

I doubt that, modern virtual memory prevents one program from even seeing another's memory space.

[Serafean]

I doubt that, modern virtual memory prevents one program from even seeing another's memory space.

Agreed, although :

A segmentation fault occurs when a program attempts to access a memory location that it is not allowed to access

So the operating system actually kills xbmc while it tries to acces memory that doesn't belong to it.

[jhsrennie]

Agreed, although : So the operating system actually kills xbmc while it tries to acces memory that doesn't belong to it.

Yes, as does every operating system since the days of Windows 95 (shudder!) and OSX 9.

To get any further with the crashing problem we need to see a debug log. See http://wiki.xbmc.org/index.php?title=Log_File for how to find the debug log. Remember to turn debug logging on from System settings, System, Debugging. Post the log on http://pastebin.com/ *not* in this forum.

JR

[Serafean]

I only have one small enough to be accepted by pastebin, and it's linked to in my original post. Taking xbmc_crashlog from ~
linking again : http://pastebin.com/hvAQWxHw

The rest is >1MB, I'll update when I get small enough log files.

[fetzerch]

I had the same issue on an archlinux with nvidia 285.05.09 and 290.06.
Got it solved by downgrading to 280.13 (and therefore xorg-server 1.10.4).
Seems to be a bug in the graphics driver...

[Serafean]

New traces : http://pastebin.com/AmK4gqrP
http://pastebin.com/ACRrswcr - in this one I used sed to remove all "avcodec_decode_video didn't consume the full packet." lines (btw what's the story with this?Is it also a bug?)

As far as the graphic drivers go, I have been experiencing this with all nvidia drivers for the past 6 months.I'll give nouveau a try, and we'll see. (I'm quite short on time ATM, so it won't be immediate).

Serafean

[Serafean]

Ok, it indeed seems that the nvidia drivers are to blame : running nouveau for a day now, and not one crash For some reason even compositing under kde is more useable under nouveau than the binary blob. The only downside is the unavailability of vdpau so I'm stuck with at most 720 videos (GLSL video rendering rocks!).

The "off by four" issue is still here

19:48:55 T:139967199700736 WARNING: Decode - avcodec_decode_video didn't consume the full packet. size: 35045, consumed: 22650
19:48:55 T:139967199700736 WARNING: Decode - avcodec_decode_video didn't consume the full packet. size: 10956, consumed: 20657
19:48:55 T:139967199700736 WARNING: Decode - avcodec_decode_video didn't consume the full packet. size: 25735, consumed: 26755
19:48:55 T:139967199700736 WARNING: Decode - avcodec_decode_video didn't consume the full packet. size: 9162, consumed: 35045

but it doesn't seem to be an issue.

Thanks for all the help,

Serafean

[toliman]

yeah, i've been waiting to see if the dailies have caught this problem for the last 2 weeks (can now delete 80mb of crash logs !), as i've had other things on my mind.

it's been crashing while idle (running the picture slideshow as screensaver) consistently every 3-4 hours, or after completing watching a TV show, or at the last 2 minutes of a show, or 20 minutes in, etc. i thought to revert to earlier builds from the 20th and the 19th, etc, but no dice.

hearing XBMC "reboot" through the speakers at 2am has been awesome the last few days, as has turning the speakers off, then forgetting to turn them back on

looks like it's time to revert back to 285.05, i didn't really notice i'd updated the nvidia kernel drivers from 2 weeks back, sigh.

[petie74]

Same on my side! I was trying to figure out were I went wrong but thought it was just the Eden builds Games. Im gonna try to revert Nvidia also. Cheers

[Zxurian]

same here. started happening with the most recent nightlies as of ~ 2 weeks ago.
However I am running the nVidia 205.05.09 drivers. Other people talk about rolling back to this version, so given that i'm already at this version and still getting issues, is there anything else that I could test for?

Happens randomly, sometimes stays idle/up for hours at a time, sometimes only a few minutes, sometimes I can watch an entire show, sometimes only 40m, etc...
crashlog after a reboot: http://pastebin.com/esStrf50

[Duduke]

+1 on that.

It started happening couple of weeks ago (i am using nighties for a while now).
in my case, I think it coincides with the 290.10 NVIDIA driver.

looking at the crash logs and "dmesg" I can see that it crashes on an NVIDIA driver module:
/usr/lib/nvidia-current/tls/libnvidia-tls.so.290.10

A couple of crash logs:
http://pastebin.com/4DZJei3Y

http://pastebin.com/d0z0xYX4

[Zxurian]

happened multiple times again tonight one time 15m into a tv show, the other after watching a 2 hour movie and me hitting the "stop" button.

I don't think it coincides with the 290.10 driver, as it was happening to me, and i've got 285.05.09, and it was running fine up until approximately 2-3 weeks ago.

i'd love to provide a crash log, but all of my crash logs have no information in the STACK TRACE area, after installing gdb.

i'd be willing to provide any other debug information if someone can tell me how to get it. (running xbmc live intalled, and following this guide that I've pieced together over time to fit my setup)

[Antipode]

I had the same problem as described above on my AsRock ION/NVIDIA running 11.10 with pre-Eden nightly builds. It appears that adding "nomodeset" to the GRUB_CMDLINE_LINUX_DEFAULT= directive in /etc/default/grub solved the issue. I'd be keen to hear if others find this solution to work as well.