BeagleBoneBlack freezes when compiling

GoogleGroups
1

All,

I have my BeagleBoneBlack with Ubunutu 12.04 armhf running kernel 3.8.13-bone28 and I’m using the board for a project to port a gnuradio project to ARM. I’m able to get all the external dependencies for my code installed/compiled but when compiling my code on the board it always freezes when it gets to our SWIG stuff

I have a 4GB swap partition and it’s being used, the kernel never throws an OutOfMemory error, the board just locks up both via ssh and locally via HDMI. It does respond to pings but that’s it.

I was running 3.8.13-bone20, so I updated to -bone28. There are 3.12.0 kernel builds I could try.

Any ideas?

-Brandon

2

Any chance is this build script publicly available?

As far as stress testing on my end i just build the gcc-4.8 branch on
the boards..

Usually takes a good two days, of 100% load..

http://rcn-ee.homeip.net:81/dl/farm/gcc/archive/20131014-10:18-gitec5e98e-gcc-4_8-branch-am335x-boneblack-512mb-0/

But, they usually survive without a hitch..

Regards,

3

Any chance is this build script publicly available?

No, this is a work project. I’ve tried both our autobuild script and building manually with verbose on and the same problem.

It gets to:

Building CXX object swig/gr_xxxx/CMakeFiles/_gr_xxx_swig.dir/gr_xxx_swigPYTHON_wrap.cxx.o

and after a few minutes the board locks up.

I’ve tried a few tuneables, and well as set the CPU limit for gcc and swig to 85% max. This seemed to take a little longer for the board to freeze but it never gets past that one swig object.

It’s frozen in the past when compiling gnuradio but a reboot and running make again eventually works.

-Brandon

4

With what? The on-board bandgap sensor is junk on the am335x
according to what i've read on l-o.

Plus i have a fan to keep the "stack" cool... :wink:

https://plus.google.com/106813818225399872098/posts/HvzNr6vGNnY

As there are currently 12 arm systems setup, plus sata drives, plus
network power relays, dual atx power supplies, etc..

Regards,

5

Yeah, lm-sensors does not work.

I have some more powerful ARM hardware on the way, but I wanted to start this project with a free board we had lying around. If I find a fix for the BBB I’ll post it here.

-Brandon

6

Or:
http://hipstercircuits.com/dallas-one-wire-temperature-reading-on-beaglebone-black-with-dto/

Regards,

7

With what? The on-board bandgap sensor is junk on the am335x

I didn't know the bandgap was junk. What does that mean?
FWIW, I use a handheld that has a laser pointer.

See:
http://www.mail-archive.com/linaro-dev@lists.linaro.org/msg16298.html

Eduardo Valentin is the thermal bandgap maintainer..

Regards,

8

>>
>> With what? The on-board bandgap sensor is junk on the am335x
>
> I didn't know the bandgap was junk. What does that mean?
> FWIW, I use a handheld that has a laser pointer.

See:
Re: Beaglebone Black Core temperature

Self-censorship is the new plague over the world it seems....

"variability in sensor accuracy" means the sensor is broken
Now I wonder when they will fix it or if it's possible to fix.
The bandgap in the TI AM335x is broken, there, I said it.

Well it works with [OMAP4430/OMAP4460/OMAP4470/OMAP543[0|2] and DRA7
devices.](1) so just something messed up on AM335x devices..

Eduardo Valentin is the thermal bandgap maintainer..

Or not, since there's no upstream code. At least I always equate maintainer
with someone that maintains code in the mainline kernel.

https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/MAINTAINERS#n8323

and see kernel code for (1)...

Regards,