[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Problem with C NodeOS ( Moab) (fwd)
> Date: Fri, 7 Jul 2000 19:38:09 -0400 (EDT)
> From: Srinivasan Venkatraman <svenk1@sheffield.cslab.uky.edu>
> To: janos-discuss@fast.cs.utah.edu
> Subject: Problem with C NodeOS ( Moab) (fwd)
>
> I forgot to mention that the platform is SunOS 5.6 sparc
> SUNW,SPARCstation-20.
>
> ---------- Forwarded message ----------
> Date: Fri, 7 Jul 2000 19:31:58 -0400 (EDT)
> From: Srinivasan Venkatraman <svenk1@hero.cslab.uky.edu>
> To: janos-discuss@flux.cs.utah.edu
> Subject: Problem with C NodeOS ( Moab)
>
>
> Hi,
>
> I am trying to develop an EE for the C implementation of NodeOS from your
> group. I am facing few problems in the thread code of the implementation.
> For testing purposes I wrote a very small program that creates a socket
> inchannel with an an_chanrecv_t function to get the size of the packet
> received. The problem happens once the function returns after printing the
> size of the packet received. I looked through the code and debugged using
> gdb and found the sequence of calls to be : ( This is after the
> an_chan_recv_t function returns)
> ani_thread_loop ( line 305) --> pthread_cleanup_pop(1) ( which calls
> ani_thread_done ) ---> ani_chan_thread_exit ---->
> threadpool_thread_schedule
>
This certainly sounds like it should work!
> Now is chan->tpool and thread->pool the same for all the flows. Ideally
> they should be but when I tested with my application, repeatedly the
> addresses of these two were different and there was SIGSEGV occuring
> always in threadpool_lock() call in threadpool_thread_schedule().
There is one threadpool per flow. So for all channels and threads within
a flow, those pointers should be the same. They would be different for
channels (or threads) in different flows.
> I am not sure how this can happen but my test program does nothing other
> than creating an socket inchan.
> I would be glad if you can fix/identify the problem. Incase you need my
> test program I'll be grad to send it.
>
Yes, if you can send me the test program, that will help. How are you
generating the traffic that your EE is seeing? Are you flooding it with
packets or just sending a modest number (it should work either way, I just
need to be sure I can reproduce your situation).
The Solaris support hasn't been that well tested unfortunately. We don't
use the socket support much either right now. So there may be some bit-rot
there.
Thanks for trying out Moab, hopefully I will get the problem fixed quickly.
[ Janos ] [ OSKit ] [ Network Testbed ] [ Flick ] [ Fluke ]
Flux Research Group / Department of Computer Science / University of Utah