Re: [Xenomai-core] [Adeos-main] New I-pipe patch for ARM.

[Philippe Gerum <philippe.gerum@domain.hid>]

Bosko Radivojevic wrote:
> OK here are a bit more details, output from the module (rt_task_create
> and rt_task_start lines are printing the return value). It looks like
> the same problem as Johan reported a few hours ago. Am I right?
>

Yes, you may not call kmalloc() on behalf of a real-time thread in kernel space.
If you run on top of rtdm, then rtdm_malloc/free() will work. Otherwise, you may
pull memory directly from the Xenomai heap using xnmalloc()/xnfree(). Make sure
to set CONFIG_XENO_OPT_SYS_HEAPSZ accordingly.
If you happen to use the native skin, then you may create your own private heap
using rt_heap_create() as well (from module_init context).

> rt_task_create(): 0
> 
> I-pipe: Detected illicit call from domain 'Xenomai'
>         into a service reserved for domain 'Linux' and below.
> 
> [<c001e0b0>] (dump_stack+0x0/0x14) from [<c00575a8>] (ipipe_check_context+0x80/0
> x9c)
> [<c0057528>] (ipipe_check_context+0x0/0x9c) from [<c00a7af0>] (__kmalloc+0x60/0x
> 108)
>  r4:c3d2fc00
> [<c00a7a90>] (__kmalloc+0x0/0x108) from [<bf000724>] (ebustxrx+0x30/0x148 [macb2
> ])
>  r7:00000010 r6:c3d7fb40 r5:bf0015cc r4:c3d2fc00
> [<bf0006f4>] (ebustxrx+0x0/0x148 [macb2]) from [<bf003b7c>] (bus_fixed_msg_cycle
> +0x1c/0x3c [rt_bus_driver])
> [<bf003b60>] (bus_fixed_msg_cycle+0x0/0x3c [rt_bus_driver]) from [<bf004f60>] (R
> ead_inputs+0x28/0x164 [rt_bus_driver])
> [<bf004f38>] (Read_inputs+0x0/0x164 [rt_bus_driver]) from [<bf0038d0>] (do_upg+0
> x44/0x250 [rt_bus_driver])
> [<bf00388c>] (do_upg+0x0/0x250 [rt_bus_driver]) from [<c005f964>] (xnarch_thread
> _trampoline+0x28/0x34)
>  r7:00000000 r6:00000000 r5:bf00b508 r4:bf00b508
> [<c005f93c>] (xnarch_thread_trampoline+0x0/0x34) from [<c005ee14>] (xnpod_schedu
> le+0x470/0x760)
>  r4:c005f93c
> 
> rt_task_start(): 0
> 
> 
> On Fri, Mar 14, 2008 at 12:07 PM, Bosko Radivojevic
> <bosko.radivojevic@domain.hid> wrote:
>> Sorry, my fingers are faster than my mind. The same happens when I
>>  remove rt_task_set_mode call. In the meantime, I realized that
>>  T_PRIMARY is invalid for kernel based tasks. So... maybe this is a
>>  bug? :)
>>
>>  On Fri, Mar 14, 2008 at 12:04 PM, Bosko Radivojevic
>>
>> <bosko.radivojevic@domain.hid> wrote:
>>
>>
>>> Hi,
>>  >
>>  >  now, I have a problem with our application. It is using native skin,
>>  >  when I start it I get:
>>  >
>>  >  kernel BUG at kernel/ipipe/core.c:322!
>>  >  Xenomai: suspending kernel thread bf00b4e8 ('do_upg') at 0xc001db14
>>  >  after exception #0
>>  >
>>  >  It looks to me that this happens after rt_task_set_mode(0, T_PRIMARY,
>>  >  NULL) call in our module. Maybe something changed from 2.3.x Xenomai
>>  >  to 2.4.x?
>>  >
>>  >  On Wed, Mar 12, 2008 at 5:23 PM, Gilles Chanteperdrix
>>  >
>>  > <gilles.chanteperdrix@xenomai.org> wrote:
>>  >
>>  >
>>  > > Hi,
>>  >  >
>>  >  >  as you may have already noticed, the ARM I-pipe patch for linux 2.6.24
>>  >  >  is available on Adeos download site and in Xenomai svn.
>>  >  >
>>  >  >  This patch incorporates modifications of all boards code to adapt to
>>  >  >  the ipipe_tickdev system which couples timer requests with Linux
>>  >  >  clockevent/clocksource framework.
>>  >  >
>>  >  >  As usual, it is tested only on AT91RM9200 and only compile-tested for
>>  >  >  other boards. So, it would be nice if people could test this new patch
>>  >  >  on their board.
>>  >  >
>>  >  >  Thanks and regards.
>>  >  >
>>  >  >  --
>>  >  >   Gilles Chanteperdrix
>>  >  >
>>  >  >  _______________________________________________
>>  >  >  Adeos-main mailing list
>>  >  >  Adeos-main@domain.hid
>>  >  >  https://mail.gna.org/listinfo/adeos-main
>>  >  >
>>  >
>>
> 
> _______________________________________________
> Adeos-main mailing list
> Adeos-main@domain.hid
> https://mail.gna.org/listinfo/adeos-main
> 


-- 
Philippe.