[ECOS] FreeBSD-stack ping is unresponsive upon low priority thread spending 100% CPU

[Gary Thomas]

Øyvind Harboe wrote:
> I've got a thread running at low priority
> that is spending 100% of available CPU. The application is a
> telnet command line.
> 
> While this low priority threads is using 100% CPU, then ping
> is unresponsive. No packets are lost though.... when the low
> priority yields the CPU again, all ping packets are returned.
> 
> However, when no connections are currently open(no
> telnet session in progress), ping works fine even if the
> same low priority thread spends 100% of the CPU...
> 
> I would have said that preemptive multithreading was
> disabled somehow(scheduler lock count problems), except
> that I would have expected the idle thread never to yield
> the CPU to another thread then...
> 
> Below shows my app (pthread.00000800) running at
> priority 15, i.e. lower than all other threads except idle
> thread.
> 
>> threads
> {{Idle Thread} RUN 1 31 31}
> {{Network alarm support} RUN 2 6 6}
> {{Network support} SLEEP 3 7 7}
> {pthread.00000800 RUN 4 15 6}
> {{DHCP lease mgt} SLEEP 5 8 8}
> {{TFTP server} SLEEP 6 10 10}
> {{HTTPD Thread} SLEEP 7 16 16}
> {{uart thread} SLEEP 8 1 1}

Which of these fields is the thread priority?
It _kindof_ looks like your 'uart' thread has a priority
higher than any of the Network threads, which could explain
a lot.

Can you duplicate this [errant] behaviour in a [stand-alone] test?



-- 
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss