Mo Libden [[email protected]] wrote:
>
> now this is intriguing.
> AFAIK, classical vfork was invented in earlier BSD to avoid expensive
> duplication of a parent process in case all the child does is launch of
> other executable. SysV solved it with CoW, BSD came up with vfork.
>
> Now, how come OpenBSD has vfork which does not work as the classic
> vfork, instead we can use rfork, which does what vfork is supposed to do?
rfork is the spork of forks. vfork, not so much.
vfork(2)
vfork() was originally used to create new processes without fully copying
the address space of the old process, which is horrendously inefficient
in a paged environment. It was useful when the purpose of fork(2) would
have been to create a new system context for an execve(2). Since fork(2)
is now efficient, even in the above case, the need for vfork() has
diminished. vfork() differs from fork(2) in that the parent is suspended
until the child makes a call to execve(2) or an exit (either by a call to
_exit(2) or abnormally). In addition, fork handlers established using
pthread_atfork(3) are not called when a multithreaded program calls
vfork().
rfork(2)
The fork functions (fork(2), vfork(2), and rfork()) create new processes.
The new process (child process) is an exact copy of the calling process
(parent process), except as outlined in the fork(2) manual page. rfork()
is used to manipulate the resources of the parent process and the child
process.
Operations currently supported include whether to copy or share the file
descriptor table between the two processes, whether to share the address
space, and whether the parent should wait(2) for the child process to
_exit(2). rfork() takes a single argument, flags, which controls which
of these resources should be manipulated. They are defined in the header
file <sys/param.h> and are the logical OR of one or more of the
following:
RFFDG Copy the parent's file descriptor table. If this flag is
unset, the parent and child will share the parent's file
descriptor table. Descriptors will remain in existence until
they are closed by all child processes using the table copies
as well as by the parent process. May not be used in
conjunction with RFCFDG.
RFPROC Create a new process. The current implementation requires this
flag to always be set.
RFMEM Force sharing of the entire address space between the parent
and child processes. The child will then inherit all the
shared segments the parent process owns. Subsequent forks by
the parent will then propagate the shared data and BSS segments
among children.
RFNOWAIT Child processes will have their resources reaped immediately
and implicitly when they terminate instead of turning into
zombies, so the parent process may not call wait(2) to collect
their exit statuses and have their resources released
explicitly.
RFCFDG Zero the child's file descriptor table (i.e. start with a blank
file descriptor table). May not be used in conjunction with
RFFDG.
RFTHREAD Create a kernel thread in the current process instead of a
separate process. Must be combined with RFMEM. Automatically
enables RFNOWAIT. The kern.rthreads sysctl must be enabled for
this to succeed.
fork(2) can be implemented as a call to rfork() using "RFFDG|RFPROC", but
isn't for backwards compatibility. If a process has file descriptor
table sharing active, setuid or setgid programs will not execve(2) with
extra privileges.
