Possible 2.6.24-rc7 issue w/respect to pthreads

Possible 2.6.24-rc7 issue w/respect to pthreads

[tom]

[-- Attachment #1: Type: text/plain, Size: 1195 bytes --]

To Whom It May Concern,

After I patched my 2.6.23 kernel to 2.6.24-rc7 this morning, I noticed
some odd behavior with respect to POSIX threads in a test program I had
written (originally to test epoll.)

The behavior is as follows:

1.  main() creates a new thread of execution with pthread_create
2.  thread_func() immediately calls pthread_detach(), which is supposed to
ensure that thread resources are cleaned up when the thread terminates.
3.  The spawned thread sleeps and then prints a message "got here"
4.  The main thread calls pthread_join().  According to the POSIX
documentation, this should suspend execution until the spawned thread has
terminated.

What I'm seeing is that the main thread terminates immediately.  If I
comment out the call to pthread_detach(), the program runs normally.  If I
boot with my 2.6.23 kernel, the program runs normally.

I'm not sure if this is some oversight on my part but similar programs
operate correctly on my Debian etch installation with 2.6.18 as well as
with the newer 23 kernel, but not with the patch.

If you have any questions, don't hesitate to email me; my address is
provided below.

See Attached (bug.c)

Tom R. Dial
tom@kavaga.com

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: bug.c --]
[-- Type: text/x-csrc; name="bug.c", Size: 1540 bytes --]

//
// bug.c
//
// Author: Tom R. Dial  <tom@kavaga.com>
//
// This simple program demonstrates what I believe to be a problem
// with the 2.6.24-rc7 kernel.  thread_func() starts by calling
// pthread_detach(), which according to POSIX (as I read it) is
// *supposed* to ensure that thread resources are cleaned up when
// the thread terminates.  This seemed to work in 2.6.23, but when
// I applied the patch for rc7, I noticed that main() seems to
// terminate immediately.  The sample program sleeps and prints
// a message from the spawned thread.  The call to pthread_join()
// in main should ensure that this message gets printed.  If the
// call to pthread_detach() is commented out, the program functions
// as expected and prints the "got here" message.  If the call
// to pthread_detach() is made, the program exits immediately
// without printing the error message.
//
// gcc -Wall -lpthread -o bug bug.c

#include <errno.h>
#include <pthread.h>
#include <stdio.h>
#include <unistd.h>

void* thread_func(void* param)
{
   int ptd = 0;
   ptd = pthread_detach( pthread_self() );
   //printf("pthread_detach() returned %d, errno=%d\n", ptd, errno );
   sleep( 10 );
   printf("got here\n");
   return 0;
}

int main(int argc, char** argv)
{
   pthread_t thread = 0;
   int status = 0;
   
   status = pthread_create( &thread, 0, thread_func, 0 );
   if (status < 0)
   {
      printf( "pthread_create() returned %d, errno=%d\n", status, errno );
      return 0;            
   }
   
   pthread_join( thread, 0 );
   
   return 0;
}

Re: Possible 2.6.24-rc7 issue w/respect to pthreads

[Marc Kleine-Budde]

[-- Attachment #1: Type: text/plain, Size: 1070 bytes --]

tom@kavaga.com wrote:
> To Whom It May Concern,
> 
> After I patched my 2.6.23 kernel to 2.6.24-rc7 this morning, I noticed
> some odd behavior with respect to POSIX threads in a test program I had
> written (originally to test epoll.)
> 
> The behavior is as follows:
> 
> 1.  main() creates a new thread of execution with pthread_create
> 2.  thread_func() immediately calls pthread_detach(), which is supposed to
> ensure that thread resources are cleaned up when the thread terminates.
> 3.  The spawned thread sleeps and then prints a message "got here"
> 4.  The main thread calls pthread_join().  According to the POSIX
> documentation, this should suspend execution until the spawned thread has
> terminated.

Quoting the man page of pthread_detach():

After pthread_detach completes, subsequent attempts to perform
pthread_join on th will fail.

regard - Marc
-- 
 Marc Kleine-Budde              Phone: +49-231-2826-924
 Pengutronix - Linux Solutions for Science and Industry
 Vertretung West/Dortmund     http://www.pengutronix.de


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 252 bytes --]

Re: Possible 2.6.24-rc7 issue w/respect to pthreads

[Jakub Jelinek]

On Wed, Jan 09, 2008 at 02:35:32AM -0800, tom@kavaga.com wrote:
> After I patched my 2.6.23 kernel to 2.6.24-rc7 this morning, I noticed
> some odd behavior with respect to POSIX threads in a test program I had
> written (originally to test epoll.)
> 
> The behavior is as follows:
> 
> 1.  main() creates a new thread of execution with pthread_create
> 2.  thread_func() immediately calls pthread_detach(), which is supposed to
> ensure that thread resources are cleaned up when the thread terminates.
> 3.  The spawned thread sleeps and then prints a message "got here"
> 4.  The main thread calls pthread_join().  According to the POSIX
> documentation, this should suspend execution until the spawned thread has
> terminated.

Your testcase is buggy.  Detached threads aren't joinable, you can't call
pthread_join on them.

	Jakub

Re: Possible 2.6.24-rc7 issue w/respect to pthreads

[Eric Dumazet]

On Wed, 9 Jan 2008 02:35:32 -0800 (PST)
tom@kavaga.com wrote:

> To Whom It May Concern,
> 
> After I patched my 2.6.23 kernel to 2.6.24-rc7 this morning, I noticed
> some odd behavior with respect to POSIX threads in a test program I had
> written (originally to test epoll.)
> 
> The behavior is as follows:
> 
> 1.  main() creates a new thread of execution with pthread_create
> 2.  thread_func() immediately calls pthread_detach(), which is supposed to
> ensure that thread resources are cleaned up when the thread terminates.
> 3.  The spawned thread sleeps and then prints a message "got here"
> 4.  The main thread calls pthread_join().  According to the POSIX
> documentation, this should suspend execution until the spawned thread has
> terminated.
> 
> What I'm seeing is that the main thread terminates immediately.  If I
> comment out the call to pthread_detach(), the program runs normally.  If I
> boot with my 2.6.23 kernel, the program runs normally.
> 
> I'm not sure if this is some oversight on my part but similar programs
> operate correctly on my Debian etch installation with 2.6.18 as well as
> with the newer 23 kernel, but not with the patch.
> 
> If you have any questions, don't hesitate to email me; my address is
> provided below.
> 
> See Attached (bug.c)
> 
> Tom R. Dial
> tom@kavaga.com

Hello Tom

Your program is buggy and depends on scheduling policies.

To make it fails on 2.6.23 (or say 2.6.9) too, just add a sleep(1) before pthread_join() call

Once a thread calls pthread_detach(pthread_t this_id), it's not possible anymore to
pthread_join(this_id, &ret) or effect is undefined.

(quoting POSIX : "The results of multiple simultaneous calls to pthread_join() specifying the same target thread are undefined" )