[Xenomai] Improper closure of rt pipe

[Xenomai] Improper closure of rt pipe

[Makarand Pradhan]

Hi All,

While working with rt pipes I noticed that:

 > When a process that is writing to the real time pipe is killed, I am 
not able to write to the pipe again after restarting the process.
 > When the pipe is closed properly with an fclose, then everything works.
 > The only way to recover the rt pipe was to reboot the system.

Would highly appreciate if you can comment on this behaviour.
Is it the expected behaviour?
Is there a way to recover the rt pipe without rebooting? Anything that 
can be done, on the other side of the pipe after detecting closure? I 
have tried deleting and recreating the pipe with no luck. I am running 
in user space so cannot flush the pipe.

Am attaching code that can be used to reproduce this issue.

I am running linux 3.0.0 on MPC8360.

Thanks and Rgds,
Makarand.


Real time task creating the pipe and reading the pipe:

rt_pipe.c
#include <xenomai/asm-generic/bits/current.h>
#include <stdio.h>
#include <sys/mman.h>
#include <native/task.h>
#include <native/mutex.h>
#include <native/timer.h>
#include <native/pipe.h>
#include <rtdk.h>
#include <stdlib.h>
#include <unistd.h>

RT_PIPE rt_pipe_1;
#define PIPE_NAME "pn"
#define PIPE_MINOR_NO 4

void task1_func(void *arg)
{

     while(true)
     {
         int bytes_read = 0;
         char buff[1024];

         bytes_read = rt_pipe_read(&rt_pipe_1, buff, 1024, TM_INFINITE);
         rt_printf("Read: %i\n", bytes_read);

         if (0 == bytes_read)
         {
             int rc = 0;
             // Delete and recreate.
             rc = rt_pipe_delete(&rt_pipe_1);
             rt_printf("Pipe delete: %i\n", rc);
             rc = rt_pipe_create(&rt_pipe_1, PIPE_NAME, PIPE_MINOR_NO, 0);
             rt_printf("Pipe creation: %i\n", rc);
         }
         rt_task_sleep(33333333LL);
     }
}

int main(int argc, char **argv)
{
     RT_TASK task0, task1;
     int rc = 0;

     mlockall(MCL_CURRENT|MCL_FUTURE);

     /* Perform auto-init of rt_print buffers if the task doesn't do so */
     rt_print_auto_init(1);

     rt_task_shadow(&task0, "Task 0", 10, 0);

     rc = rt_pipe_create(&rt_pipe_1, PIPE_NAME, PIPE_MINOR_NO, 0);
     rt_printf("Pipe creation: %i\n", rc);

     printf("Spawning: task1\n");
     rt_task_spawn(&task1, "TASK1", 10, 0, 0, task1_func, NULL);

     while (1) {
         rt_task_sleep(5000000LL);
     }
}



Linux thread that can be killed to cause the problem:

cat pipe_client.c
#include <xenomai/asm-generic/bits/current.h>
#include <stdio.h>
#include <sys/mman.h>
#include <native/task.h>
#include <native/mutex.h>
#include <native/timer.h>
#include <native/pipe.h>
#include <rtdk.h>
#include <stdlib.h>
#include <unistd.h>

#define PIPE_NAME "/proc/xenomai/registry/native/pipes/pn"
#define PIPE_MINOR_NO 4

int main(int argc, char **argv)
{

     FILE *fp;

     fp = fopen(PIPE_NAME, "r+");

     while (1==1)
     {

         printf("fp: %p\n", fp);

         fwrite("Test", 1, 4, fp);

         sleep(5);

     }

     fclose(fp);

     return 0;

}

-- 
___________________________________________________________________________
NOTICE OF CONFIDENTIALITY:
This e-mail and any attachments may contain confidential and privileged information.  If you are
not the intended recipient, please notify the sender immediately by return e-mail and delete this
e-mail and any copies.  Any dissemination or use of this information by a person other than the
intended recipient is unauthorized and may be illegal.
_____________________________________________________________________

  

Re: [Xenomai] Improper closure of rt pipe

[Gilles Chanteperdrix]

Makarand Pradhan wrote:
> Hi All,
>
> While working with rt pipes I noticed that:
>
>  > When a process that is writing to the real time pipe is killed, I am
> not able to write to the pipe again after restarting the process.
>  > When the pipe is closed properly with an fclose, then everything works.
>  > The only way to recover the rt pipe was to reboot the system.
>
> Would highly appreciate if you can comment on this behaviour.
> Is it the expected behaviour?
> Is there a way to recover the rt pipe without rebooting? Anything that
> can be done, on the other side of the pipe after detecting closure? I
> have tried deleting and recreating the pipe with no luck. I am running
> in user space so cannot flush the pipe.
>
> Am attaching code that can be used to reproduce this issue.
>
> I am running linux 3.0.0 on MPC8360.

Since it looks like a Xenomai bug, the really important information is the
version of Xenomai you are using. Anyway, this looks like a bug that was
reported a long time ago and fixed recently. Could you try the current HEAD
of the "master" branch in xenomai 2.6 git?

git://git.xenomai.org/xenomai-2.6.git

-- 
                    Gilles.

Re: [Xenomai] Improper closure of rt pipe

[Makarand Pradhan]

On 31/10/12 06:50 PM, Gilles Chanteperdrix wrote:
> Makarand Pradhan wrote:
>> Hi All,
>>
>> While working with rt pipes I noticed that:
>>
>>   >  When a process that is writing to the real time pipe is killed, I am
>> not able to write to the pipe again after restarting the process.
>>   >  When the pipe is closed properly with an fclose, then everything works.
>>   >  The only way to recover the rt pipe was to reboot the system.
>>
>> Would highly appreciate if you can comment on this behaviour.
>> Is it the expected behaviour?
>> Is there a way to recover the rt pipe without rebooting? Anything that
>> can be done, on the other side of the pipe after detecting closure? I
>> have tried deleting and recreating the pipe with no luck. I am running
>> in user space so cannot flush the pipe.
>>
>> Am attaching code that can be used to reproduce this issue.
>>
>> I am running linux 3.0.0 on MPC8360.
> Since it looks like a Xenomai bug, the really important information is the
> version of Xenomai you are using. Anyway, this looks like a bug that was
> reported a long time ago and fixed recently. Could you try the current HEAD
> of the "master" branch in xenomai 2.6 git?
>
> git://git.xenomai.org/xenomai-2.6.git
>

We are running xenomai-2.6 (stable). Will try to go to the HEAD.

Rgds,
Makarand.

-- 
___________________________________________________________________________
NOTICE OF CONFIDENTIALITY:
This e-mail and any attachments may contain confidential and privileged information.  If you are
not the intended recipient, please notify the sender immediately by return e-mail and delete this
e-mail and any copies.  Any dissemination or use of this information by a person other than the
intended recipient is unauthorized and may be illegal.
_____________________________________________________________________