This message generated a parse failure. Raw output follows here. Please use 'back' to navigate. From devnull@lkml.org Tue Apr 23 14:08:31 2024 Received: from spaans.ds9a.nl (adsl-xs4all.ds9a.nl [213.84.159.51]) by kylie.puddingonline.com (8.11.6/8.11.6) with SMTP id g8IDjZX08051 for ; Wed, 18 Sep 2002 15:45:35 +0200 Received: (qmail 10783 invoked from network); 18 Sep 2002 06:46:00 -0000 Received: from unknown (HELO spaans.ds9a.nl) (3ffe:8280:10:360:202:44ff:fe2a:a1dd) by mayo.ipv6.ds9a.nl with SMTP; 18 Sep 2002 06:46:00 -0000 Received: (qmail 18543 invoked by uid 1000); 17 Sep 2002 20:20:06 -0000 Received: (maildatabase); juh Received: (qmail 1942 invoked by alias); 26 Jun 2001 11:37:06 -0000 Received: (qmail 1939 invoked from network); 26 Jun 2001 11:37:05 -0000 Received: from vger.kernel.org (199.183.24.194) by spaans-smp.ds9a.tudelft.nl with SMTP; 26 Jun 2001 11:37:05 -0000 Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Tue, 26 Jun 2001 07:34:32 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Tue, 26 Jun 2001 07:34:23 -0400 Received: from plato.exmail.de ([194.97.6.205]:50069 "HELO plato.exmail.de") by vger.kernel.org with SMTP id ; Tue, 26 Jun 2001 07:34:12 -0400 Received: (qmail 8363 invoked from network); 26 Jun 2001 11:34:18 -0000 Received: from a1as03-p104.mch.tli.de (HELO oemcomputer) (195.252.162.104) by plato.exmail.de with SMTP; 26 Jun 2001 11:34:18 -0000 From: "Michael Kerrisk" To: linux-kernel@vger.kernel.org Date: Tue, 26 Jun 2001 13:35:58 +0200 Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7BIT Subject: New clone(2) and wait(2) man pages for review Message-Id: <3B388FBE.19842.16ED6EB@localhost> X-Mailer: Pegasus Mail for Win32 (v3.12c) Sender: linux-kernel-owner@vger.kernel.org Precedence: bulk X-Mailing-List: linux-kernel@vger.kernel.org Hello, I have just been upgrading the clone(2) and wait(2) man pages so that they correspond with (my reading of) kernel 2.4.x. Before these go into Andries' manual distribution, I would be happy if people who know better could review the changes. I have included the complete clone(2) man page and a section of the wait(2) man page below, with significant changes marked by "##". Changes to the clone(2) page include: * CLONE_PARENT option added, along with note of impact on signaling on child termination. * CLONE_THREAD options added, with a bief note on the purpose of thread groups. Additions to this description welcome. * Added CLONE_VFORK, and CLONE_TRACE A brief description of sys_clone differences has been included * Since clone() is prototyped as "clone()" rather than "__clone()" in , the name has been changed to the former. (Glibc provides a weak alias of "clone" for "__clone" ------------ For wait(2): Added __WCLONE, __WALL, and __WNOTHREAD Cheers Michael ----------------------------------------------- ## Complete revised clone(2) man page: CLONE(2) Linux Programmer's Manual CLONE(2) NAME clone - create a child process SYNOPSIS #include int clone(int (*fn) (void *arg), void *child_stack, int flags, void *arg) _syscall2(int, clone, int, flags, void *, child_stack); DESCRIPTION clone creates a new process in a similar manner to fork(2). clone is a library function layered on top of the underlying clone system call, hereinafter referred to as sys_clone. A description of sys_clone is given towards the end of this page. ## Changed: Unlike fork(2), these calls allow the child process to share parts of its execution context with the calling pro cess, such as the memory space, the table of file descrip tors, and the table of signal handlers. (Note that on this manual page, "calling process" normally corresponds to "parent process". But see the description of CLONE_PARENT below.) The main use of clone is to implement threads: multiple threads of control in a program that run concurrently in a shared memory space. When the child process is created with clone, it executes the function application fn(arg). (This differs from fork(2), where execution continues in the child from the point of the fork(2) call.) The fn argument is a pointer to a function that is called by the child process at the beginning of its execution. The arg argument is passed to the fn function. When the fn(arg) function application returns, the child process terminates. The integer returned by fn is the exit code for the child process. The child process may also terminate explicitely by calling exit(2) or after receiving a fatal signal. The child_stack argument specifies the location of the stack used by the child process. Since the child and calling process may share memory, it is not possible for the child process to execute in the same stack as the calling process. The calling process must therefore set up memory space (in its data segment or heap) for the child stack and pass a pointer to this space to clone. Stacks grow downwards on all processors that run Linux (except the HP PA processors), so child_stack usually points to the topmost address of the memory space set up for the child stack. The low byte of flags contains the number of the signal sent to the parent when the child dies. If this signal is specified as anything other SIGCHLD, then the parent pro cess must specify the __WALL or __WCLONE options when waiting for the child with wait(2). If no signal is spec ified, then the parent process is not signaled when the child terminates. flags may also be bitwise-or'ed with one or several of the following constants, in order to specify what is shared between the calling process and the child process: ## New: CLONE_PARENT (Linux 2.4 onwards) If CLONE_PARENT is set, then the parent of the new child (as returned by getp pid(2)) will be the same as that of the calling process. If CLONE_PARENT is not set, then (as with fork(2)) the child's parent is the calling process. Note that it is the parent process, as returned by getppid(2), which will be signaled when the child terminates. ## For CLONE_THREAD and CLONE_PARENT, I have just put "Linux 2.4 onwards". If someone can supply a 2.3.x version number, I will add that. CLONE_FS If CLONE_FS is set, the caller and the child pro cesses share the same file system information. This includes the root of the file system, the cur rent working directory, and the umask. Any call to chroot(2), chdir(2), or umask(2) performed by the callng process or the child process also takes effect in the other process. If CLONE_FS is not set, the child process works on a copy of the file system information of the call ing process at the time of clonecall. Calls to chroot(2), chdir(2), umask(2) performed later by one of the processes does not affect the other. CLONE_FILES If CLONE_FILES is set, the calling process and the child processes share the same file descriptor table. File descriptors always refer to the same files in the calling process and in the child pro cess. Any file descriptor created by the calling process or by the child process is also valid in the other process. Similarly, if one of the processes closes a file descriptor, or changes its associated flags, the other process is also affected. If CLONE_FILES is not set, the child process inher its a copy of all file descriptors opened in the calling process at the time of clone. Operations on file descriptors performed later by either the calling process or the child process do not affect the other. CLONE_SIGHAND If CLONE_SIGHAND is set, the calling process and the child processes share the same table of signal handlers. If the calling process or child process calls sigaction(2) to change the behavior associ ated with a signal, the behavior is also changed in the other process as well. However, the calling process and child processes still have distinct signal masks and sets of pending signals. So, one of them may block or unblock some signals using sigprocmask(2) without affecting the other process. If CLONE_SIGHAND is not set, the child process inherits a copy of the signal handlers of the call ing process at the time clone is called. Calls to sigaction(2) performed later by one of the pro cesses have no effect on the other process. ## New: CLONE_PTRACE If CLONE_PTRACE is specified, and the calling pro cess is being traced, then trace the child also (see ptrace(2)). ## New: CLONE_VFORK If CLONE_VFORK s set, the execution of the calling process is suspended until the child releases its virtual memory resources via a call to execve(2) or _exit(2) (as with vfork(2)). If CLONE_VFORK is not set then both the calling process and the child are schedulable after the call, and an application should not rely on execu tion occurring in any particular order. CLONE_VM If CLONE_VM is set, the calling process and the child processes run in the same memory space. In particular, memory writes performed by the calling process or by the child process are also visible in the other process. Moreover, any memory mapping or unmapping performed with mmap(2) or munmap(2) by the child or calling process also affects the other process. If CLONE_VM is not set, the child process runs in a separate copy of the memory space of the calling process at the time of clone. Memory writes or file mappings/unmappings performed by one of the processes do not affect the other, as with fork(2). CLONE_PID If CLONE_PID is set, the child process is created with the same process ID as the calling process. If CLONE_PID is not set, the child process pos sesses a unique process ID, distinct from that of the calling process. This flag can only be specified by the system boot process (PID 0). ## New: CLONE_THREAD If CLONE_THREAD is set, the child is placed in the same thread group as the calling process. If CLONE_THREAD is not set, then the child is placed in its own (new) thread group, whose ID is the same as the process ID. (Thread groups are feature added in Linux 2.4 to support the POSIX threads notion of a set of threads sharing a single PID. In Linux 2.4, calls to getpid(2) return the thread group ID of the caller.) ## A more expanded discussion of thread groups anyone? (Its's not clear to me how much status they enjoy at the moment. Glibc 2.2.3 linuxthreads don't seem to use them.) ## New: The sys_clone system call corresponds more closely to fork(2) in that execution in the child continues from the point of the call. Thus sys_clone only requires the flags and child_stack arguments, which have the same meaning as for clone. (Note that the order of these arguments dif fers from clone.) Another difference for sys_clone is that the child_stack argument may be zero, in which case copy-on-write seman tics ensure that the child gets separate copies of stack pages when either process modifies the stack. In this case, for correct operation, the CLONE_VM option should not be specified. RETURN VALUE On success, the PID of the child process is returned in the caller's thread of execution. On failure, a -1 will be returned in the caller's context, no child process will be created, and errno will be set appropriately. ERRORS EAGAIN Too many processes are already running. ENOMEM Cannot allocate sufficient memory to allocate a task structure for the child, or to copy those parts of the caller's context that need to be copied. EINVAL Returned by clone when a zero value is specified for child_stack. EPERM CLONE_PID was specified by a process with a non- zero PID. BUGS As of version 2.1.97 of the kernel, the CLONE_PID flag should not be used, since other parts of the kernel and most system software still assume that process IDs are unique. There is no entry for clone in libc version 5. libc 6 (a.k.a. glibc 2) provides clone as described in this man ual page. CONFORMING TO The clone and sys_clone calls are Linux-specific and should not be used in programs intended to be portable. For programming threaded applications (multiple threads of control in the same memory space), it is better to use a library implementing the POSIX 1003.1c thread API, such as the LinuxThreads library (included in glibc2). See pthread_create(3thr). This manual page corresponds to kernels 2.0.x, 2.1.x, 2.2.x, 2.4.x, and to glibc 2.0.x and 2.1.x. SEE ALSO fork(2), pthread_create(3thr), wait(2) ---------------------------- ## The following text is added to wait(2): The following Linux-specific options are for use with children created using clone(2). __WCLONE which means wait for "clone" children only. If omitted then wait for "non-clone" children only. (A "clone" child is one which delivers no signal, or a signal other than SIGCHLD to its parent upon termination.) This option is ignored if __WALL is also specified. __WALL (Linux 2.4 onwards) which means wait for all chil dren, regardless of type ("clone" or "non-clone"). __WNOTHREAD (Linux 2.4 onwards) which means do not wait for other children in the same thread group. ## If someone can tell me the version of 2.3.x which included the last two, I will include it. ## I strongly suspect that the following text from wait(2) needs to reworded for 2.4.x, and would appreciate any suggestions: In the Linux kernel, a kernel-scheduled thread is not a distinct construct from a process. Instead, a thread is simply a process that is created using the Linux-unique clone(2) system call; other routines such as the portable pthread_create(3) call are implemented using clone(2). Thus, if two threads A and B are siblings, then thread A cannot wait on any processes forked by thread B or its descendents, because an uncle cannot wait on his nephews. In some other Unix-like systems, where multiple threads are implemented as belonging to a single process, thread A can wait on any processes forked by sibling thread B; you will have to rewrite any code that makes this assumption for it to work on Linux. __________________________________________ Michael Kerrisk mailto: mtk16@ext.canterbury.ac.nz - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/