[Xenomai-help] Sleep in rtdm driver

[Xenomai-help] Sleep in rtdm driver

[robert165]

[-- Attachment #1: Type: text/plain, Size: 571 bytes --]

Hi everyone!

I'm writing a gpio driver in RTDM. I start it with the demo "Hard real time RTDM driver skeleton v1.1" of Jan Kiszka.
If I used rtdm_task_sleep in rt_ioctrl  function,  it return that "Xenomai: assertion failed at kernel/xenomai/skins/rtdm/drvlib.c:372 (!xnpod_unblockable_p())"  when running.
It is said that rtdm_task_sleep can be called from Kernel-based task. Why I can't use it in my driver. Optionally, I can do this in user-space by periodic calling the driver, but it seems slower than in kernel space.
 
Thank you very much!
 
Best regards
 
Luo
 

[-- Attachment #2: Type: text/html, Size: 802 bytes --]

Re: [Xenomai-help] Sleep in rtdm driver

[Jan Kiszka]

robert165 wrote:
> Hi everyone!
> 
> I'm writing a gpio driver in RTDM. I start it with the demo "Hard real time RTDM driver skeleton v1.1" of Jan Kiszka.
> If I used rtdm_task_sleep in rt_ioctrl  function,  it return that "Xenomai: assertion failed at kernel/xenomai/skins/rtdm/drvlib.c:372 (!xnpod_unblockable_p())"  when running.
> It is said that rtdm_task_sleep can be called from Kernel-based task. Why I can't use it in my driver. Optionally, I can do this in user-space by periodic calling the driver, but it seems slower than in kernel space.

The terms "kernel-based task" or "user space task" can be misleading.
This services requires a real-time task context. That can either be a
kernel-based *RT* task or a Xenomai shadow (aka *RT*) task in user space
that has invoked some driver service, and that driver can then call
rtdm_task_sleep on behalf of the user space task.

Blocking RTDM service *must not* be called by ordinary Linux kernel
threads or on behalf of ordinary (non-shadowed) Linux user space
threads. That's why the assertion failed for you.

HTH,
Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux

Re: [Xenomai-help] Sleep in rtdm driver

[Philippe Gerum]

On Wed, 2010-03-17 at 22:07 +0800, robert165 wrote:
> Hi everyone!
> 
> I'm writing a gpio driver in RTDM. I start it with the demo "Hard real
> time RTDM driver skeleton v1.1" of Jan Kiszka.
> If I used rtdm_task_sleep in rt_ioctrl  function,  it return that
> "Xenomai: assertion failed at kernel/xenomai/skins/rtdm/drvlib.c:372
> (!xnpod_unblockable_p())"  when running.
> It is said that rtdm_task_sleep can be called from Kernel-based task.
> Why I can't use it in my driver. Optionally, I can do this in
> user-space by periodic calling the driver, but it seems slower than in
> kernel space.

I'm puzzled: what does "slower" mean, regarding what operation?

>  
> Thank you very much!
>  
> Best regards
>  
> Luo
>  
> 
> 
> 
> _______________________________________________
> Xenomai-help mailing list
> Xenomai-help@domain.hid
> https://mail.gna.org/listinfo/xenomai-help


-- 
Philippe.

Re: [Xenomai-help] Sleep in rtdm driver

[robert165]

[-- Attachment #1: Type: text/plain, Size: 1458 bytes --]





在2010-03-17 23:05:33，"Jan Kiszka" jan.kiszka@siemens.com> 写道：
>robert165 wrote:
>> Hi everyone!
>> 
>> I'm writing a gpio driver in RTDM. I start it with the demo "Hard real time RTDM driver skeleton v1.1" of Jan Kiszka.
>> If I used rtdm_task_sleep in rt_ioctrl  function,  it return that "Xenomai: assertion failed at kernel/xenomai/skins/rtdm/drvlib.c:372 (!xnpod_unblockable_p())"  when running.
>> It is said that rtdm_task_sleep can be called from Kernel-based task. Why I can't use it in my driver. Optionally, I can do this in user-space by periodic calling the driver, but it seems slower than in kernel space.
>
>The terms "kernel-based task" or "user space task" can be misleading.
>This services requires a real-time task context. That can either be a
>kernel-based *RT* task or a Xenomai shadow (aka *RT*) task in user space
>that has invoked some driver service, and that driver can then call
>rtdm_task_sleep on behalf of the user space task.

Thanks for your demo, I started my code with it.
Is that means the demo_ioctrl_rt  is not a kernel based 'RT' task?  But why using rtdm_event_timewait in demo_read_rt is OK? 
Thanks for help.
>Blocking RTDM service *must not* be called by ordinary Linux kernel
>threads or on behalf of ordinary (non-shadowed) Linux user space
>threads. That's why the assertion failed for you.
>
>HTH,
>Jan
>
>-- 
>Siemens AG, Corporate Technology, CT T DE IT 1
>Corporate Competence Center Embedded Linux

[-- Attachment #2: Type: text/html, Size: 2569 bytes --]

Re: [Xenomai-help] Sleep in rtdm driver

[robert165]

[-- Attachment #1: Type: text/plain, Size: 931 bytes --]

Hi,



在2010-03-18 00:19:47，"Philippe Gerum" <rpm@xenomai.org> 写道：
>On Wed, 2010-03-17 at 22:07 +0800, robert165 wrote:
>> Hi everyone!
>> 
>> I'm writing a gpio driver in RTDM. I start it with the demo "Hard real
>> time RTDM driver skeleton v1.1" of Jan Kiszka.
>> If I used rtdm_task_sleep in rt_ioctrl  function,  it return that
>> "Xenomai: assertion failed at kernel/xenomai/skins/rtdm/drvlib.c:372
>> (!xnpod_unblockable_p())"  when running.
>> It is said that rtdm_task_sleep can be called from Kernel-based task.
>> Why I can't use it in my driver. Optionally, I can do this in
>> user-space by periodic calling the driver, but it seems slower than in
>> kernel space.
>
>I'm puzzled: what does "slower" mean, regarding what operation?
I am sorry, I means is it more effective if doing things in kernel space.  I think when periodic calling driver in user-space, it may spent more in context switching.
Thanks a lot.

[-- Attachment #2: Type: text/html, Size: 1719 bytes --]

Re: [Xenomai-help] Sleep in rtdm driver

[Jan Kiszka]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=x-gbk, Size: 1272 bytes --]

robert165 wrote:
> Hi,
> 
> 
> 
> ÔÚ2010-03-18 00:19:47£¬"Philippe Gerum" <rpm@xenomai.org> дµÀ£º
>> On Wed, 2010-03-17 at 22:07 +0800, robert165 wrote:
>>> Hi everyone!
>>>
>>> I'm writing a gpio driver in RTDM. I start it with the demo "Hard real
>>> time RTDM driver skeleton v1.1" of Jan Kiszka.
>>> If I used rtdm_task_sleep in rt_ioctrl  function,  it return that
>>> "Xenomai: assertion failed at kernel/xenomai/skins/rtdm/drvlib.c:372
>>> (!xnpod_unblockable_p())"  when running.
>>> It is said that rtdm_task_sleep can be called from Kernel-based task.
>>> Why I can't use it in my driver. Optionally, I can do this in
>>> user-space by periodic calling the driver, but it seems slower than in
>>> kernel space.
>> I'm puzzled: what does "slower" mean, regarding what operation?
> I am sorry, I means is it more effective if doing things in kernel space.  I think when periodic calling driver in user-space, it may spent more in context switching.

If you really depend on this minor difference, you are already designing
with a too small safety margin.

Have you measured the differences and compared them to your requirements?

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux

Re: [Xenomai-help] Sleep in rtdm driver

[robert165]

[-- Attachment #1: Type: text/plain, Size: 1395 bytes --]





在2010-03-18 23:18:55，"Jan Kiszka" <jan.kiszka@siemens.com> 写道：
>robert165 wrote:
>> Hi,
>> 
>> 
>> 
>> 在2010-03-18 00:19:47，"Philippe Gerum" <rpm@xenomai.org> 写道：
>>> On Wed, 2010-03-17 at 22:07 +0800, robert165 wrote:
>>>> Hi everyone!
>>>>
>>>> I'm writing a gpio driver in RTDM. I start it with the demo "Hard real
>>>> time RTDM driver skeleton v1.1" of Jan Kiszka.
>>>> If I used rtdm_task_sleep in rt_ioctrl  function,  it return that
>>>> "Xenomai: assertion failed at kernel/xenomai/skins/rtdm/drvlib.c:372
>>>> (!xnpod_unblockable_p())"  when running.
>>>> It is said that rtdm_task_sleep can be called from Kernel-based task.
>>>> Why I can't use it in my driver. Optionally, I can do this in
>>>> user-space by periodic calling the driver, but it seems slower than in
>>>> kernel space.
>>> I'm puzzled: what does "slower" mean, regarding what operation?
>> I am sorry, I means is it more effective if doing things in kernel space.  I think when periodic calling driver in user-space, it may spent more in context switching.
>
>If you really depend on this minor difference, you are already designing
>with a too small safety margin.
>
>Have you measured the differences and compared them to your requirements?
No, the difference is not so important. I just want to kown the the reason why I got the error message that you have explained to me clearly. 
Thanks for help!

[-- Attachment #2: Type: text/html, Size: 2568 bytes --]