(original) (raw)

Date: Mon, Jun 22, 2009 at 12:23
Subject: [Bug nptl/10311] New: clone(CLONE_VM) fails with pthread_getattr_np on i386
To: martinrb@google.com


I'm using clone() with flags CLONE_VM, but not CLONE_THREAD.

(background: I'm trying to solve the ancient overcommit failure

when spawning a small Unix process from a big process).



The act of calling clone appears to mess up the pthread library,

but only on i386, not on x86_64, using glibc version 2.7

(The bugzilla Version drop-down does not allow one to specify 2.7;

y'all should fix that)



Here's a shell transcript containing a program

that demonstrates the problem, and shows that

the problem does not occur when running in 64-bit mode

on 64-bit Linux. �(The problem also occurs when running in 32-bit mode

on 32-bit Linux).



A program like this would be a fine addition to the glibc test suite.



$ set -x; for flag in -m32 -m64; do gcc $flag -lpthread ./clone_bug.c &&

./a.out; done; cat clone_bug.c; uname -a; getconf GNU_LIBPTHREAD_VERSION;

getconf GNU_LIBC_VERSION

+zsh:1464> set -x

+zsh:1464> flag=-m32

+zsh:1464> gcc -m32 -lpthread ./clone_bug.c

+zsh:1464> ./a.out

count=2, pthread_getattr_np failed with errno = "No such process"

+zsh:1464> flag=-m64

+zsh:1464> gcc -m64 -lpthread ./clone_bug.c

+zsh:1464> ./a.out

+zsh:1464> cat clone_bug.c

#include <stdio.h>

#include <stdlib.h>

#include <stdarg.h>

#include <stddef.h>

#include <sys/types.h>

#include <wait.h>

#include <errno.h>

#include <unistd.h>

#include <pthread.h>

#include <syscall.h>

#include <sched.h>



static void

debugPrint(char *format, ...) {

�FILE *tty = fopen("/dev/tty", "w");

�va_list ap;

�va_start(ap, format);

�vfprintf(tty, format, ap);

�va_end(ap);

�fclose(tty);

}



static void debugPids(void) {

// � debugPrint("getpid()=%d gettid()=%d, syscall(getpid)=%d pthread_self=%d\n",

// � � � � � � �getpid(), syscall(SYS_gettid), syscall(SYS_getpid), pthread_self());

�static int count = 0;

�pthread_attr_t attr;

�int result;

�++count;

�if ((result = pthread_getattr_np(pthread_self(), &attr)) != 0)

� �debugPrint("count=%d, pthread_getattr_np failed with errno = \"%s\"\n",

� � � � � � � count, strerror(result));

}



static int childProcess(void *ignored) {

�_exit(0);

�// debugPrint("child\n");

�// execve("/bin/true", NULL, NULL);

�// perror("execve");

}



// I'm sure there's a better way to do this,

// but pthread_join ain't it - we can't trust it.

volatile int done = 0;



void* run(void *x) {

�const int stack_size = 1024 * 1024;

�void *clone_stack = malloc(2 * stack_size);

�int status;

�debugPids();

�int pid = clone(childProcess, clone_stack + stack_size,

� � � � � � � � �CLONE_VM | SIGCHLD, NULL);

�waitpid(pid, &status, 0);

�debugPids();

�done = 1;

�pthread_exit(0);

�return NULL;

}



int main(int argc, char *argv[]) {

�pthread_attr_t attr;

�pthread_t tid;



�pthread_attr_init(&attr);

�pthread_create(&tid, &attr, (void* (*)(void*)) run, NULL);

�// pthread_join(tid, NULL);

�while (! done)

� �;

}

+zsh:1464> uname -a

Linux spraggett.mtv.corp.google.com 2.6.24-gg23-generic #1 SMP Fri Jan 30

14:07:49 PST 2009 x86_64 GNU/Linux

+zsh:1464> getconf GNU_LIBPTHREAD_VERSION

NPTL 2.7

+zsh:1464> getconf GNU_LIBC_VERSION

glibc 2.7



--

� � � � � Summary: clone(CLONE_VM) fails with pthread_getattr_np on i386

� � � � � Product: glibc

� � � � � Version: 2.8

� � � � � �Status: NEW

� � � � �Severity: normal

� � � � �Priority: P2

� � � � Component: nptl

� � � �AssignedTo: drepper at redhat dot com

� � � �ReportedBy: martinrb at google dot com

� � � � � � � �CC: glibc-bugs at sources dot redhat dot com

�GCC host triplet: x86_64-unknown-linux-gnu





http://sourceware.org/bugzilla/show_bug.cgi?id=10311

Thanks, Michael

I'm hoping the following will placate sun studio cc:

diff --git a/src/solaris/native/java/lang/UNIXProcess\_md.c b/src/solaris/native/java/lang/UNIXProcess\_md.c
\--- a/src/solaris/native/java/lang/UNIXProcess\_md.c
+++ b/src/solaris/native/java/lang/UNIXProcess\_md.c
@@ -651,6 +651,7 @@
���� }
���� close(FAIL\_FILENO);
���� \_exit(-1);
+��� return 0;� /\* Suppress warning "no return value from function" \*/
�}

�
I'm also adding my manual test case BigFork.java.
It may be helpful while implementing the Solaris version of this feature.

webrev updated.

I need a Sun bug to commit these changes for Linux.� Please create one.

Synopsis: (process) Use clone(CLONE_VM), not fork, on Linux to avoid swap exhaustion


Description:
On Linux it is possible to use clone with CLONE_VM, but not CLONE_THREAD,
which is like fork() but much cheaper and avoids swap exhaustion due to momentary
overcommit of swap space.� One has to be very careful in this case to not mutate global

variables such as environ, but it's worth it.

Evaluation:
Make it so.

See also: 5049299

Once that is done, I will commit my changes.

Thanks,


Martin

On Thu, Jun 11, 2009 at 07:22, Michael McMahon <Michael.McMahon@sun.com> wrote:

Martin Buchholz wrote:

I broke down and finally created a "proper" webrev,
just like the good old days.

�http://cr.openjdk.java.net/\~martin/clone-exec/ <http://cr.openjdk.java.net/%7Emartin/clone-exec/>

I've run the regression tests on Solaris and Linux and they seem fine.
There is a compile warning on solaris at line 654: no return value from function.
Aside from that, I'm happy with the change now

\- Michael.