Hard freeze (ethernet related?)

Hello,

sorry if it already has been asked but I couldn’t find anything like this so I’m gonna post my issue.

I’m running ubuntu 13.04 with kernel 3.8.13-bone41 from emmc (powered by a 5V 2A power supply) and I am experiencing this:

unplugging (and/or plugging in back) the ethernet cable can result in the board freezing and the USR2 led stuck on.

I tried reading output from UART0 but there’s nothing useful.

How can I further debug this? Is it a known issue?

Thanks,
Riccardo

Can you try it using the latest Debian image?

http://beagleboard.org/latest-images

Gerald

Hello Gerald,

I downloaded debian, ran it from the uSD and I was installing a bunch
of stuff..after a while here's what I got:

[ 360.539306] INFO: task apt-get:1251 blocked for more than 60 seconds.
[ 360.546242] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs"
disables this message.
[ 360.554638] Kernel panic - not syncing: hung_task: blocked tasks
[ 360.561039] [<c0010443>] (unwind_backtrace+0x1/0x8a) from
[<c0455ced>] (panic+0x51/0x148)
[ 360.569697] [<c0455ced>] (panic+0x51/0x148) from [<c006770b>]
(watchdog+0x14f/0x194)
[ 360.577909] [<c006770b>] (watchdog+0x14f/0x194) from [<c003fb8f>]
(kthread+0x67/0x74)
[ 360.586206] [<c003fb8f>] (kthread+0x67/0x74) from [<c000c0dd>]
(ret_from_fork+0x11/0x34)
[ 360.594750] drm_kms_helper: panic occurred, switching back to text console

USR2 led stuck on

Ideas?

Thanks

Sounds like some sort of SW issue.

Gerald

That's the 3.8 mmc driver.. We've back-ported a few fixes from
v3.12/v3.13 to "fix" that issue, but for some devices it's still a
problem. If you aren't using any capes (or just a simple usart/spi
ones, we have another option for you, essentially running v3.13.x)

Regards,

Hello Robert,

well I just need usart1, usb and ethernet so the 3.13 option is fine for me.

Meanwhile I was just searching for similar issues and found this:

http://blog.machinekit.io/2013/10/hung-task-bug-in-xenomai-kernel.html
which points to
http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=7472bab236bdee1173412585591329e718f4d324

I am posting this since you replied so fast I didn't even took a look
at the patches you added to your kernels!

Do you think this is the same issue I posted in the original post?
Anyway, now I'll build a 3.13 deb.

P.S. thanks for your tools/scripts, the whole process works like a charm :slight_smile:

Thanks

Hello Robert,

well I just need usart1, usb and ethernet so the 3.13 option is fine for me.

Meanwhile I was just searching for similar issues and found this:

Machinekit Blog: Hung Task Bug in Xenomai Kernel
which points to
http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=7472bab236bdee1173412585591329e718f4d324

I am posting this since you replied so fast I didn't even took a look
at the patches you added to your kernels!

Yeap, that's the patch we backported..

https://github.com/RobertCNelson/linux-dev/blob/am33x-v3.8/patch.sh#L848

It helped, in some situations but something else between v3.8-v3.13 is
also needed..

Do you think this is the same issue I posted in the original post?
Anyway, now I'll build a 3.13 deb.

P.S. thanks for your tools/scripts, the whole process works like a charm :slight_smile:

So before you build, you can actually test..

cd /opt/scripts/tools/
git pull
sudo ./update_kernel.sh --beta-kernel

For capes, if you need "ttyO1" this the way v3.13.x is setup:

with: 2014-03-04

cd /boot/uboot/dtbs/
cp am335x-boneblack.dtb am335x-boneblack-bak.dtb
cp am335x-boneblack-ttyO1.dtb am335x-boneblack.dtb

with: 2014-03-19

Edit: /boot/uboot/uEnv.txt

set:
CAPE=ttyO1

Regards,

Thank you (awesome scripts!)

I'll try to test in the next days and let you know (even if it looks
like the issue is pretty much solved fortunately :))

Riccardo

Robert,

with the latest debian image (and 3.13) I can get it to crash with the
same procedure. Only this time no debug output in uart0 and USR0+USR2
leds stuck on.

I must say that i swapped wicd for network-manager. Do you think this
could be a cause? Any way I can help further debug this issue?

Thanks,
Riccardo

Update: looks like I obtain the bug in ubuntu raring with kernel 3.13
and wicd, so network-manager should not be an issue.

Another update: I think it could be dhcp-related, I can make it happen
also connecting an usb-wifi while ethernet is active and working..
really weird!

It's network-manager... (or any gtk2 app that seems to to decompress
something png related)

See:
http://bugs.elinux.org/issues/35

That's why i went with wicd..

TI's looking into it, so far we've tried to jtag debug it, but even
the jtag interface locks up..

Regards,

Ouch!

Is there some public bug I can subscribe to?

Thank you

Not really, It's one of the main subject's in our inside weekly
meeting. When it's fixed, there will be much rejoicing on this list.

Regards,

I'll follow the list then, thank you.

Riccardo