Re: [Xenomai-help] Floating Point Exceptions in RT-Tasks

[Philippe Gerum <rpm@xenomai.org>]

Wolfgang Grandegger wrote:
> Philippe Gerum wrote:
>> roderik.wildenburg@domain.hid wrote:
>>> On behalf of Thomas I would like to answer : 
>>>
>>>> Do you have CONFIG_XENO_OPT_DEBUG_NUCLEUS enabled in your 
>>> yes, we can aahhh have this option enabled.
>>>
>>>> kconfig, and if so, do
>>>> you get any message in your kernel log when running fetest?
>>> no, no message anywhere (/var/log/messages, console, dmesg).
>>> fetest runs and runs till I sent CTRL-C.
> 
> I just tried the program fetest on my TQM5200 with a fresh Xenomai 2.4.x
> and the linuxppc_2_4_devel tree and the excepetion does *not* get
> triggered. Actually I realize the following wired behavior:
>

I can reproduce this as well on the tqm now. This seems to be a nucleus issue 
wrt FPU management. More later.

> /fpu_test
> ^C
> ./fpu_test
> ^C
> Feb  3 17:16:12 tqm5200 kernel: Xenomai: Switching trivial to secondary
> mode after exception #1792 from user-space at 0x10001974 (pid 342)
> Floating point exception
> bash-2.05b# ./fpu_test
> ^C
> bash-2.05b# ./fpu_test
>  ^C
> bash-2.05b# ./fpu_test
> 
> After aborting the fpu_test program with ^C, I get the message above,
> but only after the second try.
> 
>> Gilles sent you a patch affecting include/asm-powerpc/bits/pod.h a few weeks
>> ago, that fixed an FPU issue. Are you running with this patch applied?
> 
> $ svn log include/asm-powerpc/bits/pod.h
> ------------------------------------------------------------------------
> r4600 | gch | 2009-01-31 19:46:37 +0100 (Sat, 31 Jan 2009) | 1 line
> 
> Is that the patch you have in mind?
> 
> Wolfgang.
>