[Barrelfish-users] RCK message buffers

[Simon Peter]

If I understand your test program correctly, then you're sending 
messages as quickly as possible on both ends of the connection.

You're getting a "RCK message buffer overflow", which means that you're 
overflowing the in-kernel message buffer used for notification, not the 
payload message buffer. Unfortunately, it's possible that that can 
happen (although it never did for me) and flow-control in this case is 
not properly implemented. The missing part is that the system has to 
remember to try again sending the *notification* (the message has 
already been sent).

Simon

On 07.05.2012 17:55, Zeus Gómez Marmolejo wrote:
> Hi all,
>
> I'm testing the message passing interface of the SCC and I see a lot of
> buffer overflows. The test program is very easy as it spawns 2
> instances, one in core 0, the other in core 1. Core 1 is continuously
> sending messages to core 0. It's using the continuation closure so there
> is a control flow, at least in the x86_64.
>
> But running it on the SCC I get this message:
>
> kernel 0: idctest: RCK message buffer overflow
>
> So I guess the control flow is not working properly... as the sending of
> messages is working much faster than the receiving ones...
>
> I attach the example,
>
> Thanks,
>
> --
> Zeus Gómez Marmolejo
> Barcelona Supercomputing Center
> PhD student
> http://www.bsc.es
>
>
>
>
> _______________________________________________
> Barrelfish-users mailing list
> Barrelfish-users at lists.inf.ethz.ch
> https://lists.inf.ethz.ch/mailman/listinfo/barrelfish-users