[Xenomai-help] very simple example for rtdm

[Xenomai-help] very simple example for rtdm

[trem]

Hi

I've written a very simple example for rtdm. And I would be pleased to
get a feedback about it. You can found it here :

http://zarb.org/users/svn/trem/trem/kernel/xenomai/rtdm/

regards,
trem

Re: [Xenomai-help] very simple example for rtdm

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 1474 bytes --]

trem wrote:
> Hi
> 
> I've written a very simple example for rtdm. And I would be pleased to
> get a feedback about it. You can found it here :
> 
> http://zarb.org/users/svn/trem/trem/kernel/xenomai/rtdm/
> 

Looks good and (almost) clean. A few remarks and some suggestion:

 o Use kernel coding style for the driver (linux/scripts/Lindent +
   manual tweaking)

 o Put your driver under RTDM_CLASS_EXPERIMENTAL, TESTING is for
   benchmarks and regression test drivers. And you don't need
   rttesting.h.


 o As RTDM is mainly about "RT", why not switching to .read_rt and
   .write_rt? That means of course creating a shadow thread in
   read_simple_rtdm.

 o Once you are real-time with read/write, you could demonstrate
   synchronisation. One FAQ is "How do I wait with my application on
   something inside a driver?" You could do rtdm_sem_up on write and
   rtdm_sem_[timed]down on read.

Alternatively:

 o What about creating a small series of RTDM examples, consisting of
   two demos so far: the first one a reduced version of you current
   code, just showing device registration and handler invocation, the
   second one about read/write and synchronisation. And if you feel like
   doing more, one could certainly continue this series with further
   aspects...

Are you interested that we merge your example(s) into the Xenomai repos
(examples/rtdm/driver-api)? I would welcome this and support your effort!

Jan


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

Re: [Xenomai-help] very simple example for rtdm

[trem]

Jan Kiszka wrote:
> trem wrote:
>> Hi
>>
>> I've written a very simple example for rtdm. And I would be pleased to
>> get a feedback about it. You can found it here :
>>
>> http://zarb.org/users/svn/trem/trem/kernel/xenomai/rtdm/
>>
> 
> Looks good and (almost) clean. A few remarks and some suggestion:
> 
>  o Use kernel coding style for the driver (linux/scripts/Lindent +
>    manual tweaking)

I don't know this kernel coding style for driver. I only know kernel
coding (Documentation/CodingStyle). But I'll apply this coding style as
soon as I found them.

>  o Put your driver under RTDM_CLASS_EXPERIMENTAL, TESTING is for
>    benchmarks and regression test drivers. And you don't need
>    rttesting.h.

yes, done.

>  o As RTDM is mainly about "RT", why not switching to .read_rt and
>    .write_rt? That means of course creating a shadow thread in
>    read_simple_rtdm.

I supose that it needs to add a rt task in read_simple_rtdm,
and all read/write will be done in this task. My goal was to
do a as simple as possible example, but this idea add some
code. Is this added code really usefull for a simple example ?

>  o Once you are real-time with read/write, you could demonstrate
>    synchronisation. One FAQ is "How do I wait with my application on
>    something inside a driver?" You could do rtdm_sem_up on write and
>    rtdm_sem_[timed]down on read.

Same answer that former. I prefer to keep a really simple example.

> Alternatively:
> 
>  o What about creating a small series of RTDM examples, consisting of
>    two demos so far: the first one a reduced version of you current
>    code, just showing device registration and handler invocation, the
>    second one about read/write and synchronisation. And if you feel like
>    doing more, one could certainly continue this series with further
>    aspects...

Could you stop to read my mind ? ;)
I like this idea, and I'm ready to work on a series of example.

I propose to write an example per concept (synchronisation, task, clock,
....). Of course, each example should have a lot of comment that explain
everything. With the idea, that the series could be used as a tutorial.

> Are you interested that we merge your example(s) into the Xenomai repos
> (examples/rtdm/driver-api)? I would welcome this and support your effort!

I think it's the logical behaviour of the previous question (and
answer). So if you think that this example is good enough, I'll be
pleased to see it added in the repository.

> Jan

regards,
trem

Re: [Xenomai-help] very simple example for rtdm

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 3416 bytes --]

trem wrote:
> Jan Kiszka wrote:
>> trem wrote:
>>   
>>> Hi
>>>
>>> I've written a very simple example for rtdm. And I would be pleased to
>>> get a feedback about it. You can found it here :
>>>
>>> http://zarb.org/users/svn/trem/trem/kernel/xenomai/rtdm/
>>>
>>>     
>>
>> Looks good and (almost) clean. A few remarks and some suggestion:
>>
>>  o Use kernel coding style for the driver (linux/scripts/Lindent +
>>    manual tweaking)
>>
>>   
> I don't know this kernel coding style for driver. I only know kernel coding
> (Documentation/CodingStyle). But I'll apply this coding style as soon as
> I found them.

That script should actually implement the documented style.

>>  o Put your driver under RTDM_CLASS_EXPERIMENTAL, TESTING is for
>>    benchmarks and regression test drivers. And you don't need
>>    rttesting.h.
>>
>>   
> yes, done.
>>  o As RTDM is mainly about "RT", why not switching to .read_rt and
>>    .write_rt? That means of course creating a shadow thread in
>>    read_simple_rtdm.
>>
>>   
> I supose that it needs to add a rt task in read_simple_rtdm,
> and all read/write will be done in this task. My goal was to
> do a as simple as possible example, but this idea add some
> code. Is this added code really usefull for a simple example ?

Well, the complication consists of a rt_task_shadow() in
read_simple_rtdm's main. But as I said later, you may also come to this
topic incrementally.

> 
>>  o Once you are real-time with read/write, you could demonstrate
>>    synchronisation. One FAQ is "How do I wait with my application on
>>    something inside a driver?" You could do rtdm_sem_up on write and
>>    rtdm_sem_[timed]down on read.
>>
>>   
> Same answer that former. I prefer to keep a really simple example.
>> Alternatively:
>>
>>  o What about creating a small series of RTDM examples, consisting of
>>    two demos so far: the first one a reduced version of you current
>>    code, just showing device registration and handler invocation, the
>>    second one about read/write and synchronisation. And if you feel like
>>    doing more, one could certainly continue this series with further
>>    aspects...
>>
>>   
> Could you stop to read my mind ? ;)

Oops, sorry, couldn't resist.

> I like this idea, and I'm ready to work on a series of example.
> 
> I propose to write an example per concept (synchronisation, task, clock, ....).
> Of course, each example should have a lot of comment that explain everything.
> With the idea, that the series could be used as a tutorial.

Perfect! That's what I had in mind as well.

> 
>> Are you interested that we merge your example(s) into the Xenomai repos
>> (examples/rtdm/driver-api)? I would welcome this and support your effort!
>>
>>   
> I think it's the logical behaviour of the previous question (and answer).
> So if you think that this example is good enough, I'll be pleased to see
> it added in the repository.

Then please post your demo in form of a patch against Xenomai trunk. You
could pick a regular naming scheme for driver and userspace sources
(e.g. "tut01-sceleton-drv.c" and "tut01-sceleton-app.c"), put them into
the driver-api folder, and extend the Makefile. See the wiki on how to
generate patches [1].

TiA for your effort!
Jan

[1] http://www.xenomai.org/index.php/Xenomai:Contributing_Patches


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 249 bytes --]

Re: [Xenomai-help] very simple example for rtdm

[Gilles Chanteperdrix]

Jan Kiszka wrote:
 > trem wrote:
 > > Jan Kiszka wrote:
 > >> trem wrote:
 > >>  o As RTDM is mainly about "RT", why not switching to .read_rt and
 > >>    .write_rt? That means of course creating a shadow thread in
 > >>    read_simple_rtdm.
 > >>
 > >>   
 > > I supose that it needs to add a rt task in read_simple_rtdm,
 > > and all read/write will be done in this task. My goal was to
 > > do a as simple as possible example, but this idea add some
 > > code. Is this added code really usefull for a simple example ?
 > 
 > Well, the complication consists of a rt_task_shadow() in
 > read_simple_rtdm's main. But as I said later, you may also come to this
 > topic incrementally.

... or you could use the POSIX skin, in which case the shadowing of the
main thread is automatically done by the skin.

-- 


					    Gilles Chanteperdrix.