Re: [rtl] Weird side-effect with FIFOs

[kumsta christophe <christophe.kumsta@free.fr>]

Hi,
  something take my attention ... when you say your RT task 'print a message to a log file', are you sure you do not use kernel calls with non rt priority scheduling ??

04/01/2001 10:19:59, Laurent.Kersten@alcatel.be wrote:

>I discovered something strange with the FIFOs (RTAI-1.5). I wrote a small RT task that basically put a character to a FIFO, do a few thing (print a message in the log file 
and do a few computation). and then read the FIFO again. This is the only RT task in
>the system. So normally, you expect every read to succeed. However, I observed a strangebehaviour. If when my task is running, I lauch a "cat </dev/rtf10" process on 
the shell prompt. The command gives me characters !!!!!!!!. So, this means that the
>rtf_put function stop the thread and do some rescheduling and so allow the linux process to read the character ... And the consequence is .... a simple Linux user space 
program is able to preempt a real time task which is not the expected behaviour in a
>realtime system.

Kumsta christophe
Real-time image processing
RTAI/RTlinux (use the source luke!)
maf-roda
christophe.kumsta@free.fr