[Xenomai-help] Strange SIGXCPU signal at rt_task_trampoline

[Xenomai-help] Strange SIGXCPU signal at rt_task_trampoline

[M. Koehrer]

Hi all,

I have the strange effect, that I get a reproducable SIGXCPU signal in gdb 
when running my Xenomai (native skin) application.
It occurs in rt_task_trampoline() within the do..while loop.
The situation is that I am re-creating and starting a task that has been 
terminated before.
Unfortunately I was not able to create a very simple application that shows this
effect.
I am using xenomai (SVN 1882), Pentium 4, gcc 4.1.2
I have enclosed the output of the gdb session:

---------- GDB START ------------
[Thread debugging using libthread_db enabled]
[New Thread -1215682880 (LWP 3132)]
[New Thread -1215687504 (LWP 3135)]
[New Thread -1219687504 (LWP 3138)]
[New Thread -1219822672 (LWP 3160)]
[New Thread -1219957840 (LWP 3161)]
[New Thread -1220093008 (LWP 3162)]
[New Thread -1220228176 (LWP 3163)]
[New Thread -1220363344 (LWP 3164)]
[New Thread -1220498512 (LWP 3165)]
[New Thread -1220633680 (LWP 3166)]
[Thread -1220633680 (zombie) exited]
[Thread -1220363344 (zombie) exited]
[Thread -1220228176 (zombie) exited]
[Thread -1219957840 (zombie) exited]
[Thread -1219822672 (zombie) exited]
[Thread -1220093008 (zombie) exited]
[Thread -1220498512 (zombie) exited]
[New Thread -1219822672 (LWP 3167)]

Program received signal SIGXCPU, CPU time limit exceeded.
[Switching to Thread -1219822672 (LWP 3167)]
0xb78f1f5e in rt_task_trampoline () from /usr/xenomai/lib/libnative.so.0
(gdb) bt
#0  0xb78f1f5e in rt_task_trampoline () from /usr/xenomai/lib/libnative.so.0
#1  0xb78e1240 in start_thread () from /lib/tls/i686/cmov/libpthread.so.0
#2  0xb79bf32e in clone () from /lib/tls/i686/cmov/libc.so.6
(gdb)
---------- GDB END ----------------
What can this be?
Thanks for any help on that issue!

Regards

Mathias


-- 
Mathias Koehrer
mathias_koehrer@domain.hid


Viel oder wenig? Schnell oder langsam? Unbegrenzt surfen + telefonieren
ohne Zeit- und Volumenbegrenzung? DAS TOP ANGEBOT JETZT bei Arcor: günstig
und schnell mit DSL - das All-Inclusive-Paket für clevere Doppel-Sparer,
nur  44,85 €  inkl. DSL- und ISDN-Grundgebühr!
http://www.arcor.de/rd/emf-dsl-2

Re: [Xenomai-help] Strange SIGXCPU signal at rt_task_trampoline

[Philippe Gerum]

On Fri, 2006-11-24 at 14:48 +0100, M. Koehrer wrote:
> Hi all,
> 
> I have the strange effect, that I get a reproducable SIGXCPU signal in gdb 
> when running my Xenomai (native skin) application.
> It occurs in rt_task_trampoline() within the do..while loop.
> The situation is that I am re-creating and starting a task that has been 
> terminated before.
> Unfortunately I was not able to create a very simple application that shows this
> effect.
> I am using xenomai (SVN 1882), Pentium 4, gcc 4.1.2
> I have enclosed the output of the gdb session:
> 

Could you try this? TIA,

--- ksrc/nucleus/shadow.c	(revision 1890)
+++ ksrc/nucleus/shadow.c	(working copy)
@@ -733,7 +733,7 @@
 
 	xnstat_counter_inc(&thread->stat.ssw);	/* Account for secondary mode switch. */
 
-	if (notify && testbits(thread->status, XNTRAPSW))
+	if (notify && testbits(thread->status, XNTRAPSW | XNSTARTED) == XNTRAPSW | XNSTARTED)
 		/* Help debugging spurious relaxes. */
 		send_sig(SIGXCPU, current, 1);
 
-- 
Philippe.

Aw: Re: [Xenomai-help] Strange SIGXCPU signal at rt_task_trampoline

[M. Koehrer]

Hi Philippe,

unfortunately this patch did not help.
It is the same situation as before.

Regards

Mathias
> > I have the strange effect, that I get a reproducable SIGXCPU signal in gdb
> 
> > when running my Xenomai (native skin) application.
> > It occurs in rt_task_trampoline() within the do..while loop.
> > The situation is that I am re-creating and starting a task that has been 
> > terminated before.
> > Unfortunately I was not able to create a very simple application that
> shows this
> > effect.
> > I am using xenomai (SVN 1882), Pentium 4, gcc 4.1.2
> > I have enclosed the output of the gdb session:
> > 
> 
> Could you try this? TIA,


-- 
Mathias Koehrer
mathias_koehrer@domain.hid


Viel oder wenig? Schnell oder langsam? Unbegrenzt surfen + telefonieren
ohne Zeit- und Volumenbegrenzung? DAS TOP ANGEBOT JETZT bei Arcor: günstig
und schnell mit DSL - das All-Inclusive-Paket für clevere Doppel-Sparer,
nur  44,85 €  inkl. DSL- und ISDN-Grundgebühr!
http://www.arcor.de/rd/emf-dsl-2

Re: Aw: Re: [Xenomai-help] Strange SIGXCPU signal at rt_task_trampoline

[Philippe Gerum]

On Fri, 2006-11-24 at 15:30 +0100, M. Koehrer wrote:
> Hi Philippe,
> 
> unfortunately this patch did not help.
> It is the same situation as before.
> 

Mm, ok, thanks for trying this out. Is T_WARNSW passed to
rt_task_create() in your case?

-- 
Philippe.

Aw: Re: Aw: Re: [Xenomai-help] Strange SIGXCPU signal at

[M. Koehrer]

HI Philippe, 

> > unfortunately this patch did not help.
> > It is the same situation as before.
> > 
> 
> Mm, ok, thanks for trying this out. Is T_WARNSW passed to
> rt_task_create() in your case?
No, I pass T_FPU | T_CPU(0)

Well, I have decided to restructure the application to create the tasks
only once during lifetime (which has some other benefits as well).
That means, that this issue hopefully will no longer hurt...

Thanks a lot

Mathias

-- 
Mathias Koehrer
mathias_koehrer@domain.hid


Viel oder wenig? Schnell oder langsam? Unbegrenzt surfen + telefonieren
ohne Zeit- und Volumenbegrenzung? DAS TOP ANGEBOT JETZT bei Arcor: günstig
und schnell mit DSL - das All-Inclusive-Paket für clevere Doppel-Sparer,
nur  44,85 €  inkl. DSL- und ISDN-Grundgebühr!
http://www.arcor.de/rd/emf-dsl-2

Re: Aw: Re: Aw: Re: [Xenomai-help] Strange SIGXCPU signal at

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 683 bytes --]

M. Koehrer wrote:
> HI Philippe, 
> 
>>> unfortunately this patch did not help.
>>> It is the same situation as before.
>>>
>> Mm, ok, thanks for trying this out. Is T_WARNSW passed to
>> rt_task_create() in your case?
> No, I pass T_FPU | T_CPU(0)
> 
> Well, I have decided to restructure the application to create the tasks
> only once during lifetime (which has some other benefits as well).
> That means, that this issue hopefully will no longer hurt...
> 

As I just read over the related code in the nucleus: SIGXCPU is also
raised if the process's memory isn't locked. Did you invoke mlockall? Or
do we see a false positive cause be that part here?

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]

Re: Aw: Re: Aw: Re: [Xenomai-help] Strange SIGXCPU signal at

[Philippe Gerum]

On Fri, 2006-11-24 at 15:58 +0100, M. Koehrer wrote:
> HI Philippe, 
> 
> > > unfortunately this patch did not help.
> > > It is the same situation as before.
> > > 
> > 
> > Mm, ok, thanks for trying this out. Is T_WARNSW passed to
> > rt_task_create() in your case?
> No, I pass T_FPU | T_CPU(0)
> 
> Well, I have decided to restructure the application to create the tasks
> only once during lifetime (which has some other benefits as well).
> That means, that this issue hopefully will no longer hurt...

Ok; still, I'd like to figure out what's been going on since this might
hide a problem.

Do you invoke rt_task_set_mode() from any context before the issue
starts showing up?

> 
> Thanks a lot
> 
> Mathias
> 
-- 
Philippe.