2002-07-16 17:13:36

by Richard B. Johnson

[permalink] [raw]
Subject: if_exist_pid()


Anybody know the 'correct' way of determining if a pid still
exists? I've been using "kill(pid, 0)" and, if it does not
return an error, it is supposed to exist. This is to release
a user-mode lock (semaphore) if the task that held the lock
crashed. Maybe there is a 'if_exist_pid(pid)' call somewhere?
Sending signal 0 to a pid sometimes returns 0, even if the pid
is long-gone and I don't want to read /proc to look for info.

Cheers,
Dick Johnson

Penguin : Linux version 2.4.18 on an i686 machine (797.90 BogoMips).

Windows-2000/Professional isn't.


2002-07-16 17:39:32

by Kasper Dupont

[permalink] [raw]
Subject: Re: if_exist_pid()

"Richard B. Johnson" wrote:
>
> Anybody know the 'correct' way of determining if a pid still
> exists? I've been using "kill(pid, 0)" and, if it does not
> return an error, it is supposed to exist.

That is correct.

> Sending signal 0 to a pid sometimes returns 0, even if the pid
> is long-gone

Really? That would mean there is a bug in kill(). Of course
you can get return value 0 if the pid has been recycled, but
otherwise it should not happen.

--
Kasper Dupont -- der bruger for meget tid p? usenet.
For sending spam use mailto:[email protected]
or mailto:[email protected]

2002-07-19 15:56:19

by Hubertus Franke

[permalink] [raw]
Subject: Re: if_exist_pid()

On Tuesday 16 July 2002 01:19 pm, Richard B. Johnson wrote:
> Anybody know the 'correct' way of determining if a pid still
> exists? I've been using "kill(pid, 0)" and, if it does not
> return an error, it is supposed to exist. This is to release
> a user-mode lock (semaphore) if the task that held the lock
> crashed. Maybe there is a 'if_exist_pid(pid)' call somewhere?
> Sending signal 0 to a pid sometimes returns 0, even if the pid
> is long-gone and I don't want to read /proc to look for info.
>
> Cheers,
> Dick Johnson
>
> Penguin : Linux version 2.4.18 on an i686 machine (797.90 BogoMips).
>
> Windows-2000/Professional isn't.
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/


You can take your clues from the get_pid algo and implementation.

The currently correct way to do this is to
scan all tasks and figure out whether either of pid, gid, tgid .. is using
your questioned pid number for small number of task this should be
trivial timewise, for large number of task its a different story

Bill Irwin is working on a patch that release pids upon their last usage
and such thing could then be recorded in a bitmap. Checking availability
would simply then mean checking a particular bit.
I have a patch lying around to move pid allocation to a bitmap.


--
-- Hubertus Franke ([email protected])