Hi Gilles

I attached the kernel configuration.

Actually I sent whatever error message I saw in the console. However
next couple of times I saw different error messages.

One time I saw --

WARNING: at
/home/travis/projects/customers/pivotal_systems/workspace/linux-2.6.28-x
enomai-pivotal/kernel/sched.c:4382 add_preempt_count+0x84
/0x98()
Unable to handle kernel NULL pointer dereference at virtual address
0000000c


Another time ---

Xenomai: fatal: zombie thread ROOT (c03a0d10) would not die...
 CPU  PID    PRI      TIMEOUT  STAT      NAME
>  0  0       -1      0        00400080  ROOT
   0  971      0      0        00200380  sleep_test
   0  972     90      0        00200188  sleep_test
Master time base: clock=91270892


thanks
Abhijit

-----Original Message-----
From: Gilles Chanteperdrix [mailto:[email protected]] 
Sent: Wednesday, June 16, 2010 11:51 AM
To: Abhijit Majumdar
Cc: Philippe Gerum; [email protected]; Steve Deiters
Subject: Re: [Xenomai-help] rt_task_sleep causing crashes on PowerPC

Abhijit Majumdar wrote:
> Hi Philippe
> 
> I have attached a standalone test program which crashes my system. It
> uses poxis skin to create a simple task which sleeps for 10 micro sec
> and wakes up and continues to do so until the user kills the process.
> 
> Interesting things note -- 
> 
>       1. If I uncomment the printf statement at the beginning of the
> while loop the program does not crash any more.
>       2. OR if I use rt_task_sleep instead of clock_nanosleep it does
> not crash.
> 
> I am using xenomai 2.5.2 and the arm platform has kernel 2.6.28.

That is a not vague. In any case, 10us on an ARM is much too short. As I
think I already told someone at pivotal systems, ARM latencies are
larger than 50 us, and most of the time larger than 100us. So, asking
for a 10us sleep is preposterous.

However, if you want your report to be useful:
- please try the latest version (that would be Xenomai 2.5.3)
- send us the full log of the serial console, not the part you think is
relevant (the error comes here while xnpod_schedule is trying to print a
message, the interesting part is not the failure of add_preempt_count,
but probably more the error message which xnpod_schedule wants to
print).
- sens us your kernel configuration.

-- 
                                            Gilles.

Attachment: config-2.6.28.10-xenomai_20100610
Description: config-2.6.28.10-xenomai_20100610

_______________________________________________
Xenomai-help mailing list
[email protected]
https://mail.gna.org/listinfo/xenomai-help

Reply via email to