[maemo-developers] N810 Resets

From: Eero Tamminen eero.tamminen at nokia.com
Date: Thu Dec 13 09:25:58 EET 2007
Hi,

ext Alex Iliadis wrote:
> Yeah, it's running as root, but it's a python script. After running top to 
> monitor it's memory usage it was ranging from 35-55MB.

How much memory it was using just before the device booted?
(By consuming all memory root process can make the device so slow
that HW watchdog reboots it.)


 > Let me know if you need any further details.

If running your script as normal user doesn't help
(e.g. from xterminal), You could install syslog and
give the information just before the device reboots.

I.e. after the reboot search for the last "Bootup reason:"
message from syslog and go up from that until you see where
the kernel starts and check what happens at that time).


	- Eero


> On Wednesday 12 December 2007 12:38:18 pm you wrote:
>> ext Alex Iliadis wrote:
>>> I suspected it was the watchdog which was doing software resets. Here
>>> is the output of the results:
>>>
>>> /proc/bootreason
>>> 32wd_to
>>>
>>> 32wd_to:
>>> 5
>> ...
>>
>>> After that I decided to disable the lifeguard resets in the device by
>>> using the flasher tool:
>>> flasher --set-rd-flags=no-lifeguard-reset -R
>> 32wd_to is the device HW watchdog reboot, not the SW watchdog one.
>>
>>> (side note: the device was in r&d mode in both runs)
>>>
>>> Unfortunately I ended up having the same exact result, the device
>>> rebooted while running my python script. I also enabled core files to
>>> see if it was an application crashing but that didn't turn out to be
>>> the case since there were no core's generated. Furthermore I installed
>>> sysklogd to see if there will be any time to save error messages prior
>>> the reset but nothing useful was in the logs. I would appreciate any
>>> pointers on debugging this issue.
>> Are you running your script as root?  And if yes, how much memory
>> it's using?  (root processes are protected from OOM-killing i.e.
>> like on any Linux device, root can mess the system)

More information about the maemo-developers mailing list