From owner-freebsd-current  Fri May  7 13:37: 2 1999
Delivered-To: freebsd-current@freebsd.org
Received: from apollo.backplane.com (apollo.backplane.com [209.157.86.2])
	by hub.freebsd.org (Postfix) with ESMTP id 43F6514CB9
	for <current@FreeBSD.ORG>; Fri,  7 May 1999 13:36:53 -0700 (PDT)
	(envelope-from dillon@apollo.backplane.com)
Received: (from dillon@localhost)
	by apollo.backplane.com (8.9.3/8.9.1) id NAA47016;
	Fri, 7 May 1999 13:36:51 -0700 (PDT)
	(envelope-from dillon)
Date: Fri, 7 May 1999 13:36:51 -0700 (PDT)
From: Matthew Dillon <dillon@apollo.backplane.com>
Message-Id: <199905072036.NAA47016@apollo.backplane.com>
To: Doug Rabson <dfr@nlsystems.com>
Cc: current@FreeBSD.ORG
Subject: Re: Debugging FreeBSD user threads with gdb
References:  <Pine.BSF.4.05.9905072123330.10400-100000@herring.nlsystems.com>
Sender: owner-freebsd-current@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.ORG

:> :--- uthread_create.c	1999/03/23 05:07:55	1.12
:> :+++ uthread_create.c	1999/05/06 15:27:33
:> :@@ -42,6 +42,8 @@
:> : #include "pthread_private.h"
:> : #include "libc_private.h"
:> : 
:> :+static int next_tid = 1;
:> :+
:> : int
:> : pthread_create(pthread_t * thread, const pthread_attr_t * attr,
:> : 	       void *(*start_routine) (void *), void *arg)
:> :@@ -87,6 +89,7 @@
:> : 		} else {
:> : 			/* Initialise the thread structure: */
:> : 			memset(new_thread, 0, sizeof(struct pthread));
:> :+			new_thread->tid = next_tid++;
:> :
:> :Doug Rabson				Mail:  dfr@nlsystems.com
:> :Nonlinear Systems Ltd.			Phone: +44 181 442 9037
:> 
:>     Hmmm.   tid is only an int and some programs which create and destroy
:>     threads a lot are almost certainly going to overflow it.  4 billion
:>     is not hard to reach.  This can result in duplicate tid's.
:
:Didn't think of that.  It gets a bit ugly inside gdb since gdb needs a
:single int to encode the pid and thread id.  I'm currently allowing 20
:bits for the pid and 11 for the tid. Any ideas?
:
:--
:Doug Rabson				Mail:  dfr@nlsystems.com
:Nonlinear Systems Ltd.			Phone: +44 181 442 9037

    Hmmm.  11 bits is only 2048 threads.  You are going to have problems
    no matter what.

    You could have a pool of tid structures.  A simple freelist.  Pseudo
    code below.  This would fix the unique-tid problem but it would not fix
    the 2048 thread limit with gdb.

    #define TIDINCR	32

    struct tid {
	struct tid *t_next;
	int t_tid;
    };

    int maxtid = 1;
    struct tid *tidpool;

    struct tid *
    alloctid(void)
    {
	struct tid *tid;

	while ((tid = tidpool) == NULL) {
		int i;

		tidpool = malloc(sizeof(struct tid) * TIDINCR);
		if (tidpool == NULL) {
			... memory allocation failed ...
		}
		for (i = 0; i < TIDINCR; ++i) {
		    tidpool[i].t_tid = maxtid + i;
		    tidpool[i].t_next = &tidpool[i+1];
		}
		tidpool[i-1].t_next = NULL;
		maxtid += TIDINCR;
	}
	tidpool = tid->t_next;
	return(tid);
    }

    freetid(struct tid *tid)
    {
	tid->t_next = tidpool;
	tidpool = tid;
    }

					-Matt
					Matthew Dillon 
					<dillon@backplane.com>



To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message