[Xenomai-core] anonymous user-space tasks (again)

[Xenomai-core] anonymous user-space tasks (again)

[Stefan Kisdaroczi]

[-- Attachment #1: Type: text/plain, Size: 944 bytes --]

Hi,

I tested Philipp's fix "Allow creating anon user-space tasks" [1][2]
with Xenomai 2.1-RC2/2.6.15/x86 (2.0.3 should behave the same).

anon rt_task_shadow:
2.0.1: worked
2.1-RC2: works, but only once (EEXIST)

anon rt_task_spawn:
2.0.1: failed with ESRCH (-3)
2.1-RC2: works, but only once (EEXIST)

A small program only calling rt_task_shadow with NULL-name fails if i start it 
a second time. Error EEXIST (-17). Same for a program calling only 
rt_task_spawn with NULL-name.

There is nothing under /proc/xenomai/registry/*

Apropos registry:
In the trunk there is no file src/native/registry.c anymore. I do not really 
miss it, but the API documentation still says these calls can be made from 
user-space. Should I create a patch to update the docs? Or should it be 
there?

thank you
kisda

[1] http://svn.gna.org/viewcvs/xenomai?rev=405&view=rev
[2] http://svn.gna.org/viewcvs/xenomai?rev=406&view=rev

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [Xenomai-core] anonymous user-space tasks (again)

[Gilles Chanteperdrix]

Stefan Kisdaroczi wrote:
 > A small program only calling rt_task_shadow with NULL-name fails if i start it 
 > a second time. Error EEXIST (-17). Same for a program calling only 
 > rt_task_spawn with NULL-name.

I am able to observe this behaviour. I am on it.

 > Apropos registry:
 > In the trunk there is no file src/native/registry.c anymore. I do not really 
 > miss it, but the API documentation still says these calls can be made from 
 > user-space. Should I create a patch to update the docs? Or should it be 
 > there?

I will update the doc too. There are no registry syscalls in
ksrc/skins/native/syscall.c, and no skins/native/lib/registry.c in 2.0.x
branch either, so it looks intentional. Registry services are not
accessible from user-space. Philippe will be able to explain here.

Thanks for reporting.

-- 


					    Gilles Chanteperdrix.

Re: [Xenomai-core] anonymous user-space tasks (again)

[Gilles Chanteperdrix]

Gilles Chanteperdrix wrote:
 > Stefan Kisdaroczi wrote:
 >  > A small program only calling rt_task_shadow with NULL-name fails if i start it 
 >  > a second time. Error EEXIST (-17). Same for a program calling only 
 >  > rt_task_spawn with NULL-name.
 > 
 > I am able to observe this behaviour. I am on it.

It is fixed in svn trunk.

-- 


					    Gilles Chanteperdrix.

Re: [Xenomai-core] anonymous user-space tasks (again)

[Philippe Gerum]

Gilles Chanteperdrix wrote:
> Stefan Kisdaroczi wrote:
>  > A small program only calling rt_task_shadow with NULL-name fails if i start it 
>  > a second time. Error EEXIST (-17). Same for a program calling only 
>  > rt_task_spawn with NULL-name.
> 
> I am able to observe this behaviour. I am on it.
> 
>  > Apropos registry:
>  > In the trunk there is no file src/native/registry.c anymore. I do not really 
>  > miss it, but the API documentation still says these calls can be made from 
>  > user-space. Should I create a patch to update the docs? Or should it be 
>  > there?
> 
> I will update the doc too. There are no registry syscalls in
> ksrc/skins/native/syscall.c, and no skins/native/lib/registry.c in 2.0.x
> branch either, so it looks intentional. Registry services are not
> accessible from user-space. Philippe will be able to explain here.

Existing registry calls are intended for use in code implementing RT interfaces in 
kernel space, not by applications. This is why we do not have any user-space 
syscalls to reach them.

> 
> Thanks for reporting.
> 


-- 

Philippe.