Problem : can't make pipe non-blocking on 2.5.X

Problem : can't make pipe non-blocking on 2.5.X

[Jean Tourrilhes]

	Hi,

	I've got the following problem : I can't make a pipe
non-blocking with 2.5.25.
	The various man pages don't mention anything
about it (actually, the "fifo" man page say that non-blocking mode
should succeed). And it seems to work fine on 2.4.20.
	Sorry if this has already been reported and fixed, but I
quicly searched the changelog and archive and found nothing.

	So, who should I complain to ?

------------- Setup ------------------------
	Kernel : 2.5.25 fails (but 2.4.20-pre2 works)
	Distrib : Debian 3.0
	libc : 2.2.5
------------- Test program -----------------
#include <unistd.h>
#include <fcntl.h>
#include <errno.h>
#include <stdio.h>

int main()
{
  int trigger_pipe[2];
  int err;
  int flags;

  pipe(trigger_pipe);

  err = fcntl(trigger_pipe[0], F_GETFL, &flags);
  fprintf(stderr, "GET FLAGS : %d - %X\n", err, flags);
  if(err >= 0)
    {
      flags |= O_NONBLOCK;
      err = fcntl(trigger_pipe[0], F_SETFL, flags);
      fprintf(stderr, "SET FLAGS : %d - %d\n", err, errno);
    }
  return(0);
}
------------- Output 2.5.25 ----------------
GET FLAGS : 0 - 40045F18
SET FLAGS : -1 - 22
------------- Output 2.4.20-pre2 -----------
GET FLAGS : 0 - 40043F18
SET FLAGS : 0 - 0
--------------------------------------------

	Have fun...

	Jean

Re: Problem : can't make pipe non-blocking on 2.5.X

[Richard B. Johnson]

On Wed, 14 Aug 2002, Jean Tourrilhes wrote:

> 
>   pipe(trigger_pipe);
> 
    if((flags = fcntl(trigger_pipe[0], F_GETFL)) != -1);
       flags &= ~O_NDELAY;
    fcntl(trigger_pipe[0], F_SETFL, flags);
 

Cheers,
Dick Johnson
Penguin : Linux version 2.4.19 on an i686 machine (797.90 BogoMips).
The US military has given us many words, FUBAR, SNAFU, now ENRON.
Yes, top management were graduates of West Point and Annapolis.

Re: Problem : can't make pipe non-blocking on 2.5.X

[Jean Tourrilhes]

On Wed, Aug 14, 2002 at 02:43:01PM -0400, Richard B. Johnson wrote:
> On Wed, 14 Aug 2002, Jean Tourrilhes wrote:
> 
> > 
> >   pipe(trigger_pipe);
> > 
>     if((flags = fcntl(trigger_pipe[0], F_GETFL)) != -1);
>        flags &= ~O_NDELAY;
>     fcntl(trigger_pipe[0], F_SETFL, flags);

	Same error. Please try again.

	Jean

Re: Problem : can't make pipe non-blocking on 2.5.X

[OGAWA Hirofumi]

Jean Tourrilhes <jt@bougret.hpl.hp.com> writes:

> 	Hi,
> 
> 	I've got the following problem : I can't make a pipe
> non-blocking with 2.5.25.
> 	The various man pages don't mention anything
> about it (actually, the "fifo" man page say that non-blocking mode
> should succeed). And it seems to work fine on 2.4.20.
> 	Sorry if this has already been reported and fixed, but I
> quicly searched the changelog and archive and found nothing.
> 
> 	So, who should I complain to ?
> 

[...]

>   err = fcntl(trigger_pipe[0], F_GETFL, &flags);

F_GETFL should be,
    flags = fcntl(trigger_pipe[0], F_GETFL, 0);

>   fprintf(stderr, "GET FLAGS : %d - %X\n", err, flags);
>   if(err >= 0)
>     {
>       flags |= O_NONBLOCK;
>       err = fcntl(trigger_pipe[0], F_SETFL, flags);
>       fprintf(stderr, "SET FLAGS : %d - %d\n", err, errno);
>     }
>   return(0);
> }
> ------------- Output 2.5.25 ----------------
> GET FLAGS : 0 - 40045F18
> SET FLAGS : -1 - 22
> ------------- Output 2.4.20-pre2 -----------
> GET FLAGS : 0 - 40043F18
> SET FLAGS : 0 - 0
> --------------------------------------------

Looks like effect of different implement of O_DIRECT(0x40000).
Thanks.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>

Re: Problem : can't make pipe non-blocking on 2.5.X

[Jean Tourrilhes]

On Thu, Aug 15, 2002 at 06:46:18AM +0900, OGAWA Hirofumi wrote:
> 
> F_GETFL should be,
>     flags = fcntl(trigger_pipe[0], F_GETFL, 0);

	Oups. I'll fix that. Thanks !

> > ------------- Output 2.5.25 ----------------
> > GET FLAGS : 0 - 40045F18
> > SET FLAGS : -1 - 22
> > ------------- Output 2.4.20-pre2 -----------
> > GET FLAGS : 0 - 40043F18
> > SET FLAGS : 0 - 0
> > --------------------------------------------
> 
> Looks like effect of different implement of O_DIRECT(0x40000).
> Thanks.

	Work in progress, I guess. Thanks...

	Jean