freebsd/share/man/man9/kproc.9

.\" Copyright (c) 2000-2001
.\"	The Regents of the University of California.  All rights reserved.
.\"
.\" All rights reserved.
.\"
.\" Redistribution and use in source and binary forms, with or without
.\" modification, are permitted provided that the following conditions
.\" are met:
.\" 1. Redistributions of source code must retain the above copyright
.\"    notice, this list of conditions and the following disclaimer.
.\" 2. Redistributions in binary form must reproduce the above copyright
.\"    notice, this list of conditions and the following disclaimer in the
.\"    documentation and/or other materials provided with the distribution.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE DEVELOPERS ``AS IS'' AND ANY EXPRESS OR
.\" IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
.\" OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
.\" IN NO EVENT SHALL THE DEVELOPERS BE LIABLE FOR ANY DIRECT, INDIRECT,
.\" INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
.\" NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
.\" DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
.\" THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
.\" (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
.\" THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
.\"
.\" $FreeBSD$
.\"
.Dd October 19, 2007
.Dt KPROC 9
.Os
.Sh NAME
.Nm kproc_start ,
.Nm kproc_shutdown ,
.Nm kproc_create ,
.Nm kproc_exit ,
.Nm kproc_resume ,
.Nm kproc_suspend ,
.Nm kproc_suspend_check
.Nd "kernel processes"
.Sh SYNOPSIS
.In sys/kthread.h
.Ft void
.Fn kproc_start "const void *udata"
.Ft void
.Fn kproc_shutdown "void *arg" "int howto"
.Ft int
.Fo kproc_create
.Fa "void (*func)(void *)" "void *arg" "struct proc **newpp"
.Fa "int flags" "int pages"
.Fa "const char *fmt" ...
.Fc
.Ft void
.Fn kproc_exit "int ecode"
.Ft int
.Fn kproc_resume "struct proc *p"
.Ft int
.Fn kproc_suspend "struct proc *p" "int timo"
.Ft void
.Fn kproc_suspend_check "struct proc *p"
.Ft int
.Fo kproc_kthread_add
.Fa "void (*func)(void *)" "void *arg"
.Fa "struct proc **procptr" "struct thread **tdptr"
.Fa "int flags" "int pages" "char * procname" "const char *fmt" "..."
.Fc
.Sh DESCRIPTION
In
.Fx 8.0 ,
the
.Fn kthread* 9
family of functions was renamed to be the
.Fn kproc* 9
family of functions, as they were misnamed
and actually produced kernel processes.
A new family of
.Em different
.Fn kthread_* 9
functions was added to produce
.Em real
kernel
.Em threads .
See the
.Xr kthread 9
man page for more information on those calls.
Also note that the
.Fn kproc_kthread_add 9
function appears in both pages as its functionality is split.
.Pp
The function
.Fn kproc_start
is used to start
.Dq internal
daemons such as
.Nm bufdaemon , pagedaemon , vmdaemon ,
and the
.Nm syncer
and is intended
to be called from
.Xr SYSINIT 9 .
The
.Fa udata
argument is actually a pointer to a
.Vt "struct kproc_desc"
which describes the kernel process that should be created:
.Bd -literal -offset indent
struct kproc_desc {
	char		*arg0;
	void		(*func)(void);
	struct proc	**global_procpp;
};
.Ed
.Pp
The structure members are used by
.Fn kproc_start
as follows:
.Bl -tag -width ".Va global_procpp" -offset indent
.It Va arg0
String to be used for the name of the process.
This string will be copied into the
.Va p_comm
member of the new process'
.Vt "struct proc" .
.It Va func
The main function for this kernel process to run.
.It Va global_procpp
A pointer to a
.Vt "struct proc"
pointer that should be updated to point to the newly created process' process
structure.
If this variable is
.Dv NULL ,
then it is ignored.
.El
.Pp
The
.Fn kproc_create
function is used to create a kernel process.
The new process shares its address space with process 0, the
.Nm swapper
process,
and runs in kernel mode only.
The
.Fa func
argument specifies the function that the process should execute.
The
.Fa arg
argument is an arbitrary pointer that is passed in as the only argument to
.Fa func
when it is called by the new process.
The
.Fa newpp
pointer points to a
.Vt "struct proc"
pointer that is to be updated to point to the newly created process.
If this argument is
.Dv NULL ,
then it is ignored.
The
.Fa flags
argument specifies a set of flags as described in
.Xr rfork 2 .
The
.Fa pages
argument specifies the size of the new kernel process's stack in pages.
If 0 is used, the default kernel stack size is allocated.
The rest of the arguments form a
.Xr printf 9
argument list that is used to build the name of the new process and is stored
in the
.Va p_comm
member of the new process's
.Vt "struct proc" .
.Pp
The
.Fn kproc_exit
function is used to terminate kernel processes.
It should be called by the main function of the kernel process rather than
letting the main function return to its caller.
The
.Fa ecode
argument specifies the exit status of the process.
While exiting, the function
.Xr exit1 9
will initiate a call to
.Xr wakeup 9
on the process handle.
.Pp
The
.Fn kproc_resume ,
.Fn kproc_suspend ,
and
.Fn kproc_suspend_check
functions are used to suspend and resume a kernel process.
During the main loop of its execution, a kernel process that wishes to allow
itself to be suspended should call
.Fn kproc_suspend_check
passing in
.Va curproc
as the only argument.
This function checks to see if the kernel process has been asked to suspend.
If it has, it will
.Xr tsleep 9
until it is told to resume.
Once it has been told to resume it will return allowing execution of the
kernel process to continue.
The other two functions are used to notify a kernel process of a suspend or
resume request.
The
.Fa p
argument points to the
.Vt "struct proc"
of the kernel process to suspend or resume.
For
.Fn kproc_suspend ,
the
.Fa timo
argument specifies a timeout to wait for the kernel process to acknowledge the
suspend request and suspend itself.
.Pp
The
.Fn kproc_shutdown
function is meant to be registered as a shutdown event for kernel processes that
need to be suspended voluntarily during system shutdown so as not to interfere
with system shutdown activities.
The actual suspension of the kernel process is done with
.Fn kproc_suspend .
.Pp
The
.Fn kproc_kthread_add
function is much like the
.Fn kproc_create
function above except that if the kproc already exists,
then only a new thread (see
.Xr kthread 9 )
is created on the existing process.
The
.Fa func
argument specifies the function that the process should execute.
The
.Fa arg
argument is an arbitrary pointer that is passed in as the only argument to
.Fa func
when it is called by the new process.
The
.Fa procptr
pointer points to a
.Vt "struct proc"
pointer that is the location to be updated with the new proc pointer
if a new process is created, or if not
.Dv NULL ,
must contain the process pointer for the already existing process.
If this argument points to
.Dv NULL ,
then a new process is created and the field updated.
If not NULL, the
.Fa tdptr
pointer points to a
.Vt "struct thread"
pointer that is the location to be updated with the new thread pointer.
The
.Fa flags
argument specifies a set of flags as described in
.Xr rfork 2 .
The
.Fa pages
argument specifies the size of the new kernel thread's stack in pages.
If 0 is used, the default kernel stack size is allocated.
The procname argument is the name the new process should be given if it needs to be created.
It is
.Em NOT
a printf style format specifier but a simple string.
The rest of the arguments form a
.Xr printf 9
argument list that is used to build the name of the new thread and is stored
in the
.Va td_name
member of the new thread's
.Vt "struct thread" .
.Sh RETURN VALUES
The
.Fn kproc_create ,
.Fn kproc_resume ,
and
.Fn kproc_suspend
functions return zero on success and non-zero on failure.
.Sh EXAMPLES
This example demonstrates the use of a
.Vt "struct kproc_desc"
and the functions
.Fn kproc_start ,
.Fn kproc_shutdown ,
and
.Fn kproc_suspend_check
to run the
.Nm bufdaemon
process.
.Bd -literal -offset indent
static struct proc *bufdaemonproc;

static struct kproc_desc buf_kp = {
	"bufdaemon",
	buf_daemon,
	&bufdaemonproc
};
SYSINIT(bufdaemon, SI_SUB_KTHREAD_BUF, SI_ORDER_FIRST, kproc_start,
    &buf_kp)

static void
buf_daemon()
{
	...
	/*
	 * This process needs to be suspended prior to shutdown sync.
	 */
	EVENTHANDLER_REGISTER(shutdown_pre_sync, kproc_shutdown,
	    bufdaemonproc, SHUTDOWN_PRI_LAST);
	...
	for (;;) {
		kproc_suspend_check(bufdaemonproc);
		...
	}
}
.Ed
.Sh ERRORS
The
.Fn kproc_resume
and
.Fn kproc_suspend
functions will fail if:
.Bl -tag -width Er
.It Bq Er EINVAL
The
.Fa p
argument does not reference a kernel process.
.El
.Pp
The
.Fn kproc_create
function will fail if:
.Bl -tag -width Er
.It Bq Er EAGAIN
The system-imposed limit on the total
number of processes under execution would be exceeded.
The limit is given by the
.Xr sysctl 3
MIB variable
.Dv KERN_MAXPROC .
.It Bq Er EINVAL
The
.Dv RFCFDG
flag was specified in the
.Fa flags
parameter.
.El
.Sh SEE ALSO
.Xr rfork 2 ,
.Xr exit1 9 ,
.Xr kthread 9 ,
.Xr SYSINIT 9 ,
.Xr wakeup 9
.Sh HISTORY
The
.Fn kproc_start
function first appeared in
.Fx 2.2 .
The
.Fn kproc_shutdown ,
.Fn kproc_create ,
.Fn kproc_exit ,
.Fn kproc_resume ,
.Fn kproc_suspend ,
and
.Fn kproc_suspend_check
functions were introduced in
.Fx 4.0 .
Prior to
.Fx 5.0 ,
the
.Fn kproc_shutdown ,
.Fn kproc_resume ,
.Fn kproc_suspend ,
and
.Fn kproc_suspend_check
functions were named
.Fn shutdown_kproc ,
.Fn resume_kproc ,
.Fn shutdown_kproc ,
and
.Fn kproc_suspend_loop ,
respectively.
Originally they had the names
.Fn kthread_*
but were changed to
.Fn kproc_*
when real kthreads became available.