Hi Robert, I just checked around and a Debug Cable (34M8872 FTDI TTL-232R-3V3) is $42 on eBay. I don’t mind spending some time troubleshooting this but not willing to go out and invest in pieces and parts. I do have two Serial to USB cables to connect to some HAM Radio gear, but not the right header connectors.
I’ll upgrade to the latest 3.14 image too.
Hi Robert, I just checked around and a Debug Cable (34M8872 FTDI
TTL-232R-3V3) is $42 on eBay. I don't mind spending some time
troubleshooting this but not willing to go out and invest in pieces and
parts. I do have two Serial to USB cables to connect to some HAM Radio gear,
but not the right header connectors.
Wow, they go for a premium on ebay..
http://www.digikey.com/product-detail/en/TTL-232R-3V3/768-1015-ND/1836393
I'll upgrade to the latest 3.14 image too.
Try that newer 3.14 for a few days first.
Regards,
Robert, apt-get couldn’t find linux-image-3.14.17-ti-r19, will go check on your site and get install-me.sh
The repo wasn't added till mid july:
http://elinux.org/Beagleboard:BeagleBoneBlack_Debian#Debian_Image_Testing_Snapshots
in your case, you can also do:
cd /opt/scripts/tools/
git pull
sudo ./update_kernel.sh --ti-kernel
Regards,
Robert,
I think part of the reason ntp and dhcpclient aren’t getting network connections at boot is because they are set at S03 in init and wicd is set at S06 and is last to get going. It appears that eth0 is not coming up until wicd loads?
Also, the install-me.sh failed at the end with final zImage copy because it creates backups of everything and /boot (92Mb) runs out of space. I had to manually remove previous image files from /boot and /boot/uboot then run the script so it had room to install everything. Seems /boot partition is too small.
Robert,
I think part of the reason ntp and dhcpclient aren't getting network
connections at boot is because they are set at S03 in init and wicd is set
at S06 and is last to get going. It appears that eth0 is not coming up until
wicd loads?
Correct, wicd set's up eth0, that's how we got the 11-12 second bootup
time. Otherwise if eth0 is handled by /etc/network/interfaces bootup
could last 2 minutes for users who don't connect eth0. I should
atleast really move ntp from S03 to S06..
Also, the install-me.sh failed at the end with final zImage copy because it
creates backups of everything and /boot (92Mb) runs out of space. I had to
manually remove previous image files from /boot and /boot/uboot then run the
script so it had room to install everything. Seems /boot partition is too
small.
Yeah it runs out of space fast.
That's one of the reasons we moved all the boot files out of the 96Mb
boot partition..
You can see the high level details (1) what i did (first implemented
in the july testing image (2))
Basicly, all the kernel/modules/dtbs are now in the big ext4 partition...
1: Beagleboard:U-boot partitioning layout 2.0 - eLinux.org
2: Beagleboard:BeagleBoneBlack Debian - eLinux.org
Regards,
Robert,
I discovered the /etc/network/interfaces issue snooping around and just set eth0 to auto load there and my total boot time is now 7s instead of 12s with all the ntp and dhcpclient retries gone.
So I guess if using the install-me script, I’ll have to manually remove the previous kernel files in /boot and /boot/uboot before running it so there’s room. Does the .deb file install (apt-get install) also do backups? If so, it should fail as well. Looks like about 110Mb for /boot partition should be enough to hold current and backups of previous.
I’m getting a ton of these in syslog at boot:
kernel: [ xx.xxxxxx] unwind: Index not found bf0f4334
Also, I noticed that in a reboot it keeps the current time for all syslog boot messages, so I know it’s getting a full reset randomly because the time in syslog is from /etc/timestamp (which I update every 15 minutes so I can tell when it’s resetting).
Robert,
I discovered the /etc/network/interfaces issue snooping around and just set
eth0 to auto load there and my total boot time is now 7s instead of 12s with
all the ntp and dhcpclient retries gone.So I guess if using the install-me script, I'll have to manually remove the
previous kernel files in /boot and /boot/uboot before running it so there's
room. Does the .deb file install (apt-get install) also do backups? If so,
it should fail as well. Looks like about 110Mb for /boot partition should be
enough to hold current and backups of previous.
With the new "apt-get install linux-image-xyz" partition setup
each "linux-image-xyz" has it's own set of:
/boot/vmlinuz-xyz
/boot/initrd.img-xyz
/boot/dtbs/xyz/*.dtb
You can switch between different kernels by updating (uname_r) in:
/boot/uEnv.txt
(all located in ext4, so it'll take a while to run out of space)
I'm getting a ton of these in syslog at boot:
kernel: [ xx.xxxxxx] unwind: Index not found bf0f4334
Yeap, unwind isn't setup yet.. The 3.14 is wip in progress, but it's
more reliable then v3.8.x for me..
Also, I noticed that in a reboot it keeps the current time for all syslog
boot messages, so I know it's getting a full reset randomly because the time
in syslog is from /etc/timestamp (which I update every 15 minutes so I can
tell when it's resetting).
Regards,
Robert,
I just made a note to myself to comment out the backup commands in your install-me script file (so that should leave enough space in uboot) before I run it.
With the direction we are heading, those install-me.sh's are becoming
obsolete.. 
Regards,
Makes sense, but if apt-get doesn’t work it’s a good fall-back for now.
I guess if I get more random resets, I’ll have to spring for that serial debug cable @ $23 shipped from DigiKey.
If you’ve got an MSP430 launchpad laying around, you could use that too ( for serial debug ). But you’ll have to make / use 3 jumpers, and baud rate is limited to 9600bps ( msp430 side ).
Do you mind saying how you'd use it for serial debug? does it have a
USB-serial passthrough burned in, or are you saying that it could be
programmed to bridge serial <-> USB?
It can be setup as a Serial pass through / Bridge. No programming required.
http://forum.beaglefu.com/topic/92-msp430-launchpad-v15-serial-debug/
Update: I installed 3.14.17-ti-r19 yesterday about 3pm and set eth0 to auto so it comes up earlier than wicp. Ran fine until 7:13am this morning then did it’s usual reset thing. Very frustrating. Guess I’ll have to spring for that cable.
Update: the BBB just reset again at 7:49am. That’s interesting, because it reset twice yesterday at almost the same times 7:35am and 8:13am and today at 7:13am and 7:49am. cron daily runs at 6:25am and I have no ctontab jobs at those times. Appears to be some sort of pattern due to the daily reset times being so close, but there is nothing is syslog except cron.hourly until the new boot sequence starts.
Update: I got a third reset yesterday at 11:53pm but no resets this morning between 7am and 8am. The only log file entry I could find that matched a reset was in wtmp: Sat Sep 13 07:13 - crash (-5369±11:- and there are a half dozen other ‘crash’ entries in the log as well.
Why don’t you share the crash log?
Another reset early this afternoon, so I have shutdown and removed the 5v2a cheapo PS and replaced it with a Garmin 5v1a PS I have that uses the mini USB instead of the banana plug. I want to eliminate the current PS as the problem. All my PS are in two power strips plugged into an APC Battery Backup so power is conditioned, filtered, and constant.
I’m spending waaayyyy too much time with this since my RasPi has been set up same way and running 24/7 since I plugged it in with zero issues.
Am I the only one with constant resets on a BBB?
I’m hesitant to drop $24 on a serial debug cable that I’ll use maybe two days to capture data that might or might not shed light on the issue.
william@arm:~$ uptime
14:38:02 up 4 days, 4:31, 1 user, load average: 0.04, 0.03, 0.05
And this is a small uptime. I’ve had it run for months at a time. Others yet have had longer uptimes.
One thing you have not mentioned, or at least i have not seen you mention is if you’re connecting anything on the IO pins. IF so, what.