[Xenomai-help] Writing a new driver, shoud use message pipes or shared memory?

[Xenomai-help] Writing a new driver, shoud use message pipes or shared memory?

[Carlos Eduardo de Brito Novaes]

Hello all!
Hope this question fits to the scope of the list.

I got a PCI acquisition board and wish to write a xenomai driver for it. I had 
did this before with an old ISA board and used memory heaps to exchange data 
from kernel (configuration and mainly acquisition routines) and the real time 
processing task.

Now, I would like to know if I can use only message pipes for this kind of 
application. I would like to use it so the entire proccess became more "unix 
like", using memory heaps I think that the internals of the driver are exposed 
and the any misfuncion can arise due to bad data passed from programmers. Is 
this right?

Can message pipes provide hard real time timings, about 1K/10K samples per 
second and about 100 bytes on each sample?

Thanks

Carlos Novaes.

Re: [Xenomai-help] Writing a new driver, shoud use message pipes or shared memory?

[Gilles Chanteperdrix]

On 06/30/2011 05:12 PM, Carlos Eduardo de Brito Novaes wrote:
> Hello all!
> Hope this question fits to the scope of the list.
> 
> I got a PCI acquisition board and wish to write a xenomai driver for it. I had 
> did this before with an old ISA board and used memory heaps to exchange data 
> from kernel (configuration and mainly acquisition routines) and the real time 
> processing task.
> 
> Now, I would like to know if I can use only message pipes for this kind of 
> application. I would like to use it so the entire proccess became more "unix 
> like", using memory heaps I think that the internals of the driver are exposed 
> and the any misfuncion can arise due to bad data passed from programmers. Is 
> this right?
> 
> Can message pipes provide hard real time timings, about 1K/10K samples per 
> second and about 100 bytes on each sample?

Hi,

you probably should use the RTDM skin to write a driver. So, use the
"read" call to get data, or if there is a big volume, use some mmaped
aread, using rtdm_mmap_to_user.

Regards.

-- 
                                                                Gilles.

Re: [Xenomai-help] Writing a new driver, Skin differences.

[Carlos Eduardo de Brito Novaes]

Hello,

I think I the question on using memory mapped are solved. Now I was wondering 
and cant find a big difference over the four skins, at least for a new project 
with no previous code written on the past. Is there any advantage in using the 
rtdm skin over the navive one?

Thanks so much.

Carlos Novaes


Em sex 01 jul 2011, às 04:57:27, Gilles Chanteperdrix escreveu:
> On 06/30/2011 05:12 PM, Carlos Eduardo de Brito Novaes wrote:
> > Hello all!
> > Hope this question fits to the scope of the list.
> > 
> > I got a PCI acquisition board and wish to write a xenomai driver for it.
> > I had did this before with an old ISA board and used memory heaps to
> > exchange data from kernel (configuration and mainly acquisition
> > routines) and the real time processing task.
> > 
> > Now, I would like to know if I can use only message pipes for this kind
> > of application. I would like to use it so the entire proccess became
> > more "unix like", using memory heaps I think that the internals of the
> > driver are exposed and the any misfuncion can arise due to bad data
> > passed from programmers. Is this right?
> > 
> > Can message pipes provide hard real time timings, about 1K/10K samples
> > per second and about 100 bytes on each sample?
> 
> Hi,
> 
> you probably should use the RTDM skin to write a driver. So, use the
> "read" call to get data, or if there is a big volume, use some mmaped
> aread, using rtdm_mmap_to_user.
> 
> Regards.

Re: [Xenomai-help] Writing a new driver, Skin differences.

[Gilles Chanteperdrix]

On 07/01/2011 07:23 PM, Carlos Eduardo de Brito Novaes wrote:
> Hello,
> 
> I think I the question on using memory mapped are solved. Now I was wondering 
> and cant find a big difference over the four skins, at least for a new project 
> with no previous code written on the past. Is there any advantage in using the 
> rtdm skin over the navive one?

The RTDM is meant for writing drivers in kernel-space and continuity
will be ensured in Xenomai 3.x, whereas the other kernel-space skins
will no longer be available in Xenomai 3.x. Writing drivers in
kernel-space with either the native or the posix skin is deprecated.

-- 
                                                                Gilles.

Re: [Xenomai-help] Writing a new driver, Skin differences.

[Luis Rossi]

[-- Attachment #1: Type: text/plain, Size: 1412 bytes --]

Sorry to get inside in the middle ofconversation... what you mean writing
driver in kernel-space is deprecated. Is there any other way to make a
driver? I belive a kernel-driver should be the most RT possible no?

On Sat, Jul 2, 2011 at 5:13 PM, Gilles Chanteperdrix <
gilles.chanteperdrix@xenomai.org> wrote:

> On 07/01/2011 07:23 PM, Carlos Eduardo de Brito Novaes wrote:
> > Hello,
> >
> > I think I the question on using memory mapped are solved. Now I was
> wondering
> > and cant find a big difference over the four skins, at least for a new
> project
> > with no previous code written on the past. Is there any advantage in
> using the
> > rtdm skin over the navive one?
>
> The RTDM is meant for writing drivers in kernel-space and continuity
> will be ensured in Xenomai 3.x, whereas the other kernel-space skins
> will no longer be available in Xenomai 3.x. Writing drivers in
> kernel-space with either the native or the posix skin is deprecated.
>
> --
>                                                                 Gilles.
>
> _______________________________________________
> Xenomai-help mailing list
> Xenomai-help@domain.hid
> https://mail.gna.org/listinfo/xenomai-help
>



-- 
Luís Filipe Rossi
Electrical Engineer
Biomechatronics Lab. / Grupo de Sensores Integráveis e Sistemas
Escola Politécnica
Universidade de São Paulo
Cel. +55 (11) 7662-9234

[-- Attachment #2: Type: text/html, Size: 2563 bytes --]

Re: [Xenomai-help] Writing a new driver, Skin differences.

[Gilles Chanteperdrix]

On 07/02/2011 10:38 PM, Luis Rossi wrote:
> Sorry to get inside in the middle ofconversation... what you mean writing
> driver in kernel-space is deprecated. Is there any other way to make a
> driver? I belive a kernel-driver should be the most RT possible no?

Read again: writing a driver in kernel-space with the posix or native
skin is deprecated. Not with the RTDM skin.

-- 
                                                                Gilles.

Re: [Xenomai-help] Writing a new driver, Skin differences.

[Carlos Eduardo de Brito Novaes]

Thank you very much, I got it now. I had found out those points on xenomai 3.x 
and rtdm, so I must be a little more carefull on reading before asking. Sorry 
for that.

I am starting working on it, at first I think that using rt_dev_read shoud do 
the work, its easy to set up and there are examples over the net, but I cant 
find any (online) example on using rtdm_mmap_to_user and also could not figure 
out the parameters to get it working. Can you point me any other resource or 
example on using it?

The rtdm skin sens to be well documented but its very poor on examples and 
detailed description, at least I cant find any resource as informative as those 
related to the native api.


Thank you,

Carlos Novaes

Em sáb 02 jul 2011, às 17:13:55, Gilles Chanteperdrix escreveu:
> On 07/01/2011 07:23 PM, Carlos Eduardo de Brito Novaes wrote:
> > Hello,
> > 
> > I think I the question on using memory mapped are solved. Now I was
> > wondering and cant find a big difference over the four skins, at least
> > for a new project with no previous code written on the past. Is there
> > any advantage in using the rtdm skin over the navive one?
> 
> The RTDM is meant for writing drivers in kernel-space and continuity
> will be ensured in Xenomai 3.x, whereas the other kernel-space skins
> will no longer be available in Xenomai 3.x. Writing drivers in
> kernel-space with either the native or the posix skin is deprecated.