[Xenomai] xeno3_rc3_alchemy What is the effect of rt_task_delete on a RT_SEM?

[Xenomai] xeno3_rc3_alchemy What is the effect of rt_task_delete on a RT_SEM?

[Niels Wellens]

Hi,

While porting our code to xeno3 I noticed that one of our tasks blocked
on a semaphore during exit (code worked on xeno 2.6.4 Native, some small
changes where needed for Alchemy but nothing fundamental). It seems that
deletion of a task that is waiting on a RT_SEM doesn't decrement the
nwaiters field of the semaphore. Afterwards, when I raise the semaphore,
the nwaiters decrement without releasing an existing task. Are there
some changes here in Xeno3? I looked in the migration guide but the only
remark about semaphores is for posix.

Please find in attachment a small working example with some comments in
the exit part of the code.

Thanks in advance!
Kind regards,
Niels
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Makefile
Type: text/x-makefile
Size: 2140 bytes
Desc: not available
URL: <http://www.xenomai.org/pipermail/xenomai/attachments/20150313/83310ade/attachment.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: sem_test.c
Type: text/x-csrc
Size: 3443 bytes
Desc: not available
URL: <http://www.xenomai.org/pipermail/xenomai/attachments/20150313/83310ade/attachment.c>

Re: [Xenomai] xeno3_rc3_alchemy What is the effect of rt_task_delete on a RT_SEM?

[Jan Kiszka]

On 2015-03-13 16:57, Niels Wellens wrote:
> Hi,
> 
> While porting our code to xeno3 I noticed that one of our tasks blocked
> on a semaphore during exit (code worked on xeno 2.6.4 Native, some small
> changes where needed for Alchemy but nothing fundamental). It seems that
> deletion of a task that is waiting on a RT_SEM doesn't decrement the
> nwaiters field of the semaphore. Afterwards, when I raise the semaphore,
> the nwaiters decrement without releasing an existing task. Are there
> some changes here in Xeno3? I looked in the migration guide but the only
> remark about semaphores is for posix.
> 
> Please find in attachment a small working example with some comments in
> the exit part of the code.

I've pushed two related patches to
http://git.xenomai.org/xenomai-jki.git/log/?h=for-forge. The first one
fixes the semaphore value after interrupted waiting as you saw it, the
second one the reported nwaiters value if the semaphore is positive
(unrelated but caught by your test case as well).

Thanks for reporting and specifically for preparing the test case that well!

Jan

-- 
Siemens AG, Corporate Technology, CT RTC ITP SES-DE
Corporate Competence Center Embedded Linux

Re: [Xenomai] xeno3_rc3_alchemy What is the effect of rt_task_delete on a RT_SEM?

[Philippe Gerum]

On 03/13/2015 07:30 PM, Jan Kiszka wrote:
> On 2015-03-13 16:57, Niels Wellens wrote:
>> Hi,
>>
>> While porting our code to xeno3 I noticed that one of our tasks blocked
>> on a semaphore during exit (code worked on xeno 2.6.4 Native, some small
>> changes where needed for Alchemy but nothing fundamental). It seems that
>> deletion of a task that is waiting on a RT_SEM doesn't decrement the
>> nwaiters field of the semaphore. Afterwards, when I raise the semaphore,
>> the nwaiters decrement without releasing an existing task. Are there
>> some changes here in Xeno3? I looked in the migration guide but the only
>> remark about semaphores is for posix.
>>
>> Please find in attachment a small working example with some comments in
>> the exit part of the code.
> 
> I've pushed two related patches to
> http://git.xenomai.org/xenomai-jki.git/log/?h=for-forge. The first one
> fixes the semaphore value after interrupted waiting as you saw it, the
> second one the reported nwaiters value if the semaphore is positive
> (unrelated but caught by your test case as well).
> 
> Thanks for reporting and specifically for preparing the test case that well!
> 

Merged, thanks.

-- 
Philippe.