From owner-freebsd-current  Fri Aug 15 02:53:28 1997
Return-Path: <owner-freebsd-current>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.5/8.8.5) id CAA18583
          for current-outgoing; Fri, 15 Aug 1997 02:53:28 -0700 (PDT)
Received: from necropolis.org ([207.66.220.160])
          by hub.freebsd.org (8.8.5/8.8.5) with ESMTP id CAA18572
          for <current@freebsd.org>; Fri, 15 Aug 1997 02:53:23 -0700 (PDT)
Received: from localhost (edmond@localhost)
	by necropolis.org (8.8.7/8.8.5) with SMTP id CAA06370
	for <current@freebsd.org>; Fri, 15 Aug 1997 02:49:17 -0700 (PDT)
X-Authentication-Warning: necropolis.org: edmond owned process doing -bs
Date: Fri, 15 Aug 1997 02:49:15 -0700 (PDT)
From: "Andrew N. Edmond" <edmond@shaman.lycaeum.org>
X-Sender: edmond@necropolis.org
To: current@freebsd.org
Subject: 3.0: problems with libc_r and native pthreads
Message-ID: <Pine.BSF.3.96.970815023942.3581S-100000@necropolis.org>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-freebsd-current@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

I upgraded one of my single processor machines to FreeBSD 3.0 (august 11th
SNAP) to test out the new native pthreads, and have a question to ask (or
perhaps error to report).

I wrote a multithreaded program that's running a thread with listen() and
an accept() in a loop, that when hit reads about 512 bytes of data from
the socket, spawns a new thread and goes back to accept() wait for
another connection.

Now, with mit-pthreads, I can throw a good 500 hits per second into this
mechanism and it runs perfectly without any loss of data.  If instead I
use libc_r, I get this error message:

error reading stream message: Resource temporarily unavailable
error reading stream message: Resource temporarily unavailable
error reading stream message: Resource temporarily unavailable
error reading stream message: Resource temporarily unavailable
error reading stream message: Resource temporarily unavailable

This is generated from this code:

   listen(socketOpen, QUEUED_CONNECTS);
   do {
      ptrbuf = buf = (char *) malloc (PACKET_SIZE);
      bzero(buf, sizeof(buf));
      msgSock = accept(socketOpen, 0, 0);
      if (msgSock == -1) perror ("error on accept");
      else do {
         if ((nread = read(msgSock, ptrbuf, 256)) < 0) 
--->        perror("error reading stream message");
         ptrbuf += nread;
         } while (nread != 0);
      close(msgSock);
      // add to thread pool
      tpool_add_work(threadPool, myFunction, (void *)buf);
      } while (1);

I assume this is because there are not enough file descriptors available
for some reason in libc_r (but there is in mit-pthreads!).  I even set:

#define FD_SETSIZE 8192

In the source of my program before any other includes.

Is this a libc_r error or am I missing something?

Andy

    :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
\-/ ::::::::       Andrew N. Edmond  - finger for PGP key        :::::::::: \-/
/-\ ::::::                      ............                         :::::: /-\
\-/ :::     edmond@lycaeum.org     ::::::     an1@anon.nymserver.com    ::: \-/
/-\ :     Director of the Lycaeum    ::    the Nymserver Administrator    : /-\
\-/ :::      www.lycaeum.org       ::::::       www.nymserver.com       ::: \-/
    :::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::