Re: [Xenomai] xenomai-3.0-rc5 : binding named semaphores from external process

[Philippe Gerum <rpm@xenomai.org>]

On 07/29/2015 02:48 PM, Philippe Gerum wrote:
> On 07/29/2015 01:48 PM, Frederik Bayart wrote:
>>> On Wednesday, 29 July 2015, 12:12, Philippe Gerum <rpm@xenomai.org> wrote:
>>>
>>>
>>> On 07/28/2015 05:19 PM, Frederik Bayart wrote:
>>>>>> ./stest --session=foo 1
>>>>>>
>>>>>> The first process binds to the named semaphore which is not yet created. According to the description of rt_sem_bind,  if the object does not exist on entry, the caller may block until a semaphore of the given name is created.
>>>>>>
>>>>>> So based on this description, I would I expect that if the second process is started, the first process would bind and continue and blocks in the p operation. The same for the 2nd process. Is this correct ? At the moment, if I start the 2nd process, I get a segfault in rt_sem_create :
>>>>>>
>>>>>> [ 9758.887921] [Xenomai] switching main to secondary mode after exception #14 from user-space at 0x7f6ad8c4a43e (pid 9256)
>>>>>> [ 9758.887930] main[9256]: segfault at 7f903bf6e038 ip 00007f6ad8c4a43e sp 00007fffbd2c2eb8 error 6 in libcobalt.so.2.0.0[7f6ad8c3d000+1e000]
>>>>>>
>>>>>
>>>>> Looking at the copperplate code, yep, this definitely can't work. Ok,
>>>>> queued.
>>>>
>>>> I noticed also :
>>>>
>>>>  ./stest --session=foo 1
>>>>  ./stest --session=foo 0
>>>>
>>>> Both processes are waiting in rt_sem_p . The first process catches SIGINT, the second doesn't. If you press CTRL+C on the second process it stops but also the first process is falling through the rt_sem_p although no rt_sem_v is raised.
>>>
>>>
>>> Spurious wake up in the Cobalt kernel when aborting a sem wait operation
>>> due to a signal/unblock event, that is the reason why I could not see
>>> this from a Mercury setup. This is fixed in the -next branch.
>>>
>>>> At that moment no process is pending on rt_sem_p anymore. But if you do then a rt_sem_inquire, nwaiters is still 1.
>>>
>>> I don't observe this one. Assuming you enabled the registry, what does
>>> /var/run/xenomai/.../alchemy/semaphores/semtestsem tell you about the
>>> sema4 value?
>>>
>>
>> When I start my processes, I don't see this alchemy directory in my registry. Any suggestion why ?
>>
>> The output of 'xeno-config --info' is :
>>
>> Xenomai version: Xenomai/cobalt v3.0-rc5 -- 
>> Linux dev-x10sae 3.18.12-x86-64-xeno-3.0.rc5 #1 SMP PREEMPT Fri Jul 10 12:29:14 CEST 2015 x86_64 GNU/Linux
>> Kernel parameters: BOOT_IMAGE=/boot/vmlinuz-3.18.12-x86-64-xeno-3.0.rc5 root=UUID=fc8ecefa-fc73-487f-a045-cffa99c38a11 ro quiet console=tty0 console=ttyS0,115200n8
>> I-pipe release #1 detected
>> Cobalt core 3.0-rc5 detected
>> Compiler: gcc version 4.9.2 (Debian 4.9.2-10) 
>> Build args: --prefix=/usr --includedir=/usr/include/xenomai --mandir=/usr/share/man --with-testdir=/usr/lib/xenomai/testsuite --with-core=cobalt --enable-smp --enable-pshared --enable-registry --build x86_64-linux-gnu build_alias=x86_64-linux-gnu
>>
>> I have only this in the registery :
>>
>> find /var/run/xenomai
>>
>> /var/run/xenomai
>> /var/run/xenomai/root
>> /var/run/xenomai/root/foo
>> /var/run/xenomai/root/foo/6210
>> /var/run/xenomai/root/foo/6203
>> /var/run/xenomai/root/foo/system
>>
>>
>> In attachment my latest version of stest.c to be sure, there are only small modifications.
>>
>> Below the output of the first process
>>
>> sudo ./stest --session=foo 1
>> stest.c:82: __XENO_COMPAT__ not defined, create = 1
>> stest.c:121: enter CTRL+C to continue...
>> stest.c:43: binding sem semtestsem...
>> stest.c:51: calling rt_sem_p...
>> ========= Now CTRL+C on the second (other) process ======================
>> stest.c:60: rt_sem_p passed
> 
> You must mean interrupting the one enabling create mode, i.e.
> --session=foo 1. ^C on the other one will terminate it immediately since
> it does not trap this signal, leaving the first one hanging on
> rt_sem_p() as expected (with the latest fix applied).
> 
>> ========= Now CTRL+C on this (first) process to interrupt the wait loop =========
>> stest.c:32: signal(2)
>> stest.c:146: nwaiters = 1
>> stest.c:147: count = 0
>> stest.c:152: calling rt_sem_v 1...
>> stest.c:161: rt_sem_v called
>> stest.c:186: nwaiters = 0
>> stest.c:187: count = 0
>>
> 
> Looking at your code, I fail to see any issue with this trace output.
> What output would you expect instead?
> 

In absence of any particular signal disposition (e.g. blocking), SIGINT
can ne sent to the main() thread of your process, which then wakes up
while the one running rt_input() is still pending on rt_sem_p(). Hence
the non-zero waiter count when the main() thread inspects the sema4 data.

-- 
Philippe.