X-Recipient: archive-cygwin AT delorie DOT com
X-SWARE-Spam-Status: No, hits=-2.4 required=5.0 	tests=AWL,BAYES_00,SPF_PASS
X-Spam-Check-By: sourceware.org
Message-ID: <4A24FD89.8090105@gmail.com>
Date: Tue, 02 Jun 2009 11:23:05 +0100
From: Dave Korn <dave DOT korn DOT cygwin AT googlemail DOT com>
User-Agent: Thunderbird 2.0.0.17 (Windows/20080914)
MIME-Version: 1.0
To: cygwin AT cygwin DOT com
Subject: Re: [1.7] pthread_create block
References: <9bfd07db0906020210h57d4c12h62f47a6ba3a5730e AT mail DOT gmail DOT com>
In-Reply-To: <9bfd07db0906020210h57d4c12h62f47a6ba3a5730e@mail.gmail.com>
Content-Type: multipart/mixed;  boundary="------------030804020300020409010901"
Mailing-List: contact cygwin-help AT cygwin DOT com; run by ezmlm
List-Id: <cygwin.cygwin.com>
List-Subscribe: <mailto:cygwin-subscribe AT cygwin DOT com>
List-Archive: <http://sourceware.org/ml/cygwin/>
List-Post: <mailto:cygwin AT cygwin DOT com>
List-Help: <mailto:cygwin-help AT cygwin DOT com>, <http://sourceware.org/ml/#faqs>
Sender: cygwin-owner AT cygwin DOT com
Mail-Followup-To: cygwin AT cygwin DOT com
Delivered-To: mailing list cygwin AT cygwin DOT com

--------------030804020300020409010901
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit

Thomas Stalder wrote:

> Sometimes pthread_create block and never return.
> 
> I have made a simple program (test.c) to reproduce the problem.
> 
>> gcc test.c -lpthread -o test
>> ./test
> thread id=1
> thread id=2
> thread id=3
> thread id=4
> ........
> thread id=3736
> 
> 
> Sometimes pthread_create block after creating arround 10 threads,
> sometimes after creating arround 10000 threads or more.

  You may have run into the pthread interlocking bug that I'm tracking down in
this thread on the cygwin-dev list:

    http://www.cygwin.com/ml/cygwin-developers/2009-05/threads.html#00084

  If you'd like to try building the cygwin DLL yourself from the CVS sources
with the attached patch applied, you may find it solves your problem; at least
for me it does.  I've been running your test program while writing this email
and it's got as far as this:

thread id=1260546
thread id=1260547

... without a hiccup.  Actually, now it's got as far as

thread id=2405875

by the time I diff'd the patch for you.  (If you'd prefer, I could send you a
binary of the DLL built with the patch, off-list; let me know).

    cheers,
      DaveK

--------------030804020300020409010901
Content-Type: text/x-c;
 name="asm-interlock.diff"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline;
 filename="asm-interlock.diff"

Index: winsup/cygwin/winbase.h
===================================================================
RCS file: /cvs/src/src/winsup/cygwin/winbase.h,v
retrieving revision 1.14
diff -p -u -r1.14 winbase.h
--- winsup/cygwin/winbase.h	12 Jul 2008 18:09:17 -0000	1.14
+++ winsup/cygwin/winbase.h	2 Jun 2009 10:06:17 -0000
@@ -38,21 +38,21 @@ ilockdecr (volatile long *m)
 extern __inline__ long
 ilockexch (volatile long *t, long v)
 {
-  register int __res;
+  register long __res __asm__ ("%eax") = *t;
   __asm__ __volatile__ ("\n\
-1:	lock	cmpxchgl %3,(%1)\n\
+1:	lock	cmpxchgl %2,%1\n\
 	jne 1b\n\
- 	": "=a" (__res), "=q" (t): "1" (t), "q" (v), "0" (*t): "cc");
+ 	": "+a" (__res), "=m" (*t): "q" (v), "m" (*t) : "memory", "cc");
   return __res;
 }
 
 extern __inline__ long
 ilockcmpexch (volatile long *t, long v, long c)
 {
-  register int __res;
+  register long __res __asm ("%eax") = c;
   __asm__ __volatile__ ("\n\
-	lock cmpxchgl %3,(%1)\n\
-	": "=a" (__res), "=q" (t) : "1" (t), "q" (v), "0" (c): "cc");
+	lock cmpxchgl %2,%1\n\
+	": "+a" (__res), "=m" (*t) : "q" (v), "m" (*t) : "memory", "cc");
   return __res;
 }
 


--------------030804020300020409010901
Content-Type: text/plain; charset=us-ascii

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/
--------------030804020300020409010901--