possible problems with kernel threading code?

possible problems with kernel threading code?

[Chris Friesen]

I'm running the following code under 2.4.18-2.4.20, with the standard 
thread library, and sometimes I get an error of "No such process".  How 
can this possibly happen?  The pthread_setschedparam() call is running 
in the thread that it is trying to operate on, but somehow it can't find 
itself?

I also have managed to trigger reboots with this code, and I'm not sure how.

Anyone have any ideas?  Is this a kernel problem?  I can trigger it on 
ppc hardware under kernel versions 2.4.18-2.4.20 and a modified 2.4.22. 
(I haven't tried 2.4.21)  My x86 with a 2.4.18-19.8.0 redhat kernel 
doesn't show the problem.

Thanks,

Chris



#include <stdio.h>
#include <sys/types.h>
#include <sys/wait.h>
#include <unistd.h>
#include <pthread.h>
#include <stdlib.h>
#include <string.h>

pthread_t t1;

void *func(void *p)
{
   pid_t pid;
   int status;

   struct sched_param schedParam;
   int         policy = SCHED_RR;
   int         priority = 40;

  schedParam.sched_priority = priority;
  int schedRc = pthread_setschedparam(t1, policy, &schedParam);
  if (schedRc){
       fprintf(stderr, "pthread_setschedparam error:%s, priority:%d,
               policy:%d\n",strerror(schedRc), priority, policy);
   }
   pthread_exit(EXIT_SUCCESS);
   return 0;
}


int main(int argc, char **argv)
{
   pthread_create(&t1, 0, func,(void*)1);
   pthread_join(t1,0);
   return 0;
}


-- 
Chris Friesen                    | MailStop: 043/33/F10
Nortel Networks                  | work: (613) 765-0557
3500 Carling Avenue              | fax:  (613) 765-2986
Nepean, ON K2H 8E9 Canada        | email: cfriesen@nortelnetworks.com

Re: possible problems with kernel threading code?

[Ulrich Drepper]

Not that this is of any interest for this list but what the heck.

The problem is your code.

>  int schedRc = pthread_setschedparam(t1, policy, &schedParam);

There is no guarantee that t1 is filled before you it here in the newly
created thread.  Only when pthread_create() returns is the thread handle
guaranteed to be written.

-- 
➧ Ulrich Drepper ➧ Red Hat, Inc. ➧ 444 Castro St ➧ Mountain View, CA ❖

Re: possible problems with kernel threading code?

[Chris Friesen]

Ulrich Drepper wrote:
> Not that this is of any interest for this list but what the heck.

I thought that the kernel hangs might be of interest.

> The problem is your code.

>> int schedRc = pthread_setschedparam(t1, policy, &schedParam);
> 
> There is no guarantee that t1 is filled before you it here in the newly
> created thread.  Only when pthread_create() returns is the thread handle
> guaranteed to be written.

Ah, that explains the error.  I'm still trying to track down the kernel 
hangs though.

Chris


-- 
Chris Friesen                    | MailStop: 043/33/F10
Nortel Networks                  | work: (613) 765-0557
3500 Carling Avenue              | fax:  (613) 765-2986
Nepean, ON K2H 8E9 Canada        | email: cfriesen@nortelnetworks.com