man/man9/sysctl_add_oid.9

.\"
.\" Copyright (c) 2000, Andrzej Bialecki <abial@FreeBSD.org>
.\" 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.
.\" 3. The name of the author may not be used to endorse or promote products
.\"    derived from this software without specific prior written permission.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``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 AUTHOR OR CONTRIBUTORS 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: src/share/man/man9/sysctl_add_oid.9,v 1.3.2.10 2001/12/17 11:30:19 ru Exp $
.\" $DragonFly: src/share/man/man9/sysctl_add_oid.9,v 1.8 2008/05/01 23:36:43 swildner Exp $
.\"
.Dd June 22, 2006
.Dt SYSCTL_ADD_OID 9
.Os
.Sh NAME
.Nm sysctl_add_oid ,
.Nm sysctl_remove_oid
.Nd runtime sysctl tree manipulation
.Sh SYNOPSIS
.In sys/sysctl.h
.Ft struct sysctl_oid *
.Fo sysctl_add_oid
.Fa "struct sysctl_ctx_list *ctx"
.Fa "struct sysctl_oid_list *parent"
.Fa "int number"
.Fa "const char *name"
.Fa "int kind"
.Fa "void *arg1"
.Fa "int arg2"
.Fa "int (*handler) (SYSCTL_HANDLER_ARGS)"
.Fa "const char *format"
.Fa "const char *descr"
.Fc
.Ft int
.Fo sysctl_remove_oid
.Fa "struct sysctl_oid *oidp"
.Fa "int del"
.Fa "int recurse"
.Fc
.Ft struct sysctl_oid_list *
.Fo SYSCTL_CHILDREN
.Fa "struct sysctl_oid *oidp"
.Fc
.Ft struct sysctl_oid_list *
.Fo SYSCTL_STATIC_CHILDREN
.Fa "OID_NAME"
.Fc
.Ft struct sysctl_oid *
.Fo SYSCTL_ADD_OID
.Fa "struct sysctl_ctx_list *ctx"
.Fa "struct sysctl_oid_list *parent"
.Fa "int number"
.Fa "NAME"
.Fa "int kind"
.Fa "void *arg1"
.Fa "int arg2"
.Fa "int (*handler) (SYSCTL_HANDLER_ARGS)"
.Fa "const char *format"
.Fa "const char *descr"
.Fc
.Ft struct sysctl_oid *
.Fo SYSCTL_ADD_NODE
.Fa "struct sysctl_ctx_list *ctx"
.Fa "struct sysctl_oid_list *parent"
.Fa "int number"
.Fa "NAME"
.Fa "int access"
.Fa "int (*handler) (SYSCTL_HANDLER_ARGS)"
.Fa "const char *descr"
.Fc
.Ft struct sysctl_oid *
.Fo SYSCTL_ADD_STRING
.Fa "struct sysctl_ctx_list *ctx"
.Fa "struct sysctl_oid_list *parent"
.Fa "int number"
.Fa "NAME"
.Fa "int access"
.Fa "char *arg"
.Fa "0"
.Fa "const char *descr"
.Fc
.Ft struct sysctl_oid *
.Fo SYSCTL_ADD_INT
.Fa "struct sysctl_ctx_list *ctx"
.Fa "struct sysctl_oid_list *parent"
.Fa "int number"
.Fa "NAME"
.Fa "int access"
.Fa "int *arg"
.Fa "0"
.Fa "const char *descr"
.Fc
.Ft struct sysctl_oid *
.Fo SYSCTL_ADD_QUAD
.Fa "struct sysctl_ctx_list *ctx"
.Fa "struct sysctl_oid_list *parent"
.Fa "int number"
.Fa "NAME"
.Fa "int access"
.Fa "quad_t *arg"
.Fa "0"
.Fa "const char *descr"
.Fc
.Ft struct sysctl_oid *
.Fo SYSCTL_ADD_UQUAD
.Fa "struct sysctl_ctx_list *ctx"
.Fa "struct sysctl_oid_list *parent"
.Fa "int number"
.Fa "NAME"
.Fa "int access"
.Fa "u_quad_t *arg"
.Fa "0"
.Fa "const char *descr"
.Fc
.Ft struct sysctl_oid *
.Fo SYSCTL_ADD_UINT
.Fa "struct sysctl_ctx_list *ctx"
.Fa "struct sysctl_oid_list *parent"
.Fa "int number"
.Fa "NAME"
.Fa "int access"
.Fa "unsigned int *arg"
.Fa "0"
.Fa "const char *descr"
.Fc
.Ft struct sysctl_oid *
.Fo SYSCTL_ADD_LONG
.Fa "struct sysctl_ctx_list *ctx"
.Fa "struct sysctl_oid_list *parent"
.Fa "int number"
.Fa "NAME"
.Fa "int access"
.Fa "long *arg"
.Fa "0"
.Fa "const char *descr"
.Fc
.Ft struct sysctl_oid *
.Fo SYSCTL_ADD_ULONG
.Fa "struct sysctl_ctx_list *ctx"
.Fa "struct sysctl_oid_list *parent"
.Fa "int number"
.Fa "NAME"
.Fa "int access"
.Fa "unsigned long *arg"
.Fa "0"
.Fa "const char *descr"
.Fc
.Ft struct sysctl_oid *
.Fo SYSCTL_ADD_OPAQUE
.Fa "struct sysctl_ctx_list *ctx"
.Fa "struct sysctl_oid_list *parent"
.Fa "int number"
.Fa "NAME"
.Fa "int access"
.Fa "void *arg"
.Fa "size_t *len"
.Fa "const char *format"
.Fa "const char *descr"
.Fc
.Ft struct sysctl_oid *
.Fo SYSCTL_ADD_STRUCT
.Fa "struct sysctl_ctx_list *ctx"
.Fa "struct sysctl_oid_list *parent"
.Fa "int number"
.Fa "NAME"
.Fa "int access"
.Fa "struct TYPE *arg"
.Fa "TYPE"
.Fa "const char *descr"
.Fc
.Ft struct sysctl_oid *
.Fo SYSCTL_ADD_PROC
.Fa "struct sysctl_ctx_list *ctx"
.Fa "struct sysctl_oid_list *parent"
.Fa "int number"
.Fa "NAME"
.Fa "int access"
.Fa "0"
.Fa "0"
.Fa "int (*handler) (SYSCTL_HANDLER_ARGS)"
.Fa "const char *format"
.Fa "const char *descr"
.Fc
.Sh DESCRIPTION
These functions and macros provide an interface
for creating and deleting sysctl oids at runtime
(e.g.\& during lifetime of a module).
The alternative method,
based on linker sets (see
.In sys/linker_set.h
and
.\" XXX Manual pages should avoid referencing source files
.Pa src/sys/kern/kern_sysctl.c
for details), only allows creation and deletion
on module load and unload respectively.
.Pp
Dynamic oids of type
.Dv CTLTYPE_NODE
are reusable
so that several code sections can create and delete them,
but in reality they are allocated and freed
based on their reference count.
As a consequence,
it is possible for two or more code sections
to create partially overlapping trees that they both can use.
It is not possible to create overlapping leaves,
nor to create different child types with the same name and parent.
.Pp
Newly created oids are connected to their parent nodes.
In all these functions and macros
(with the exception of
.Fn sysctl_remove_oid ) ,
one of the required parameters is
.Fa parent ,
which points to the head of the parent's list of children.
.Pp
Most top level categories are created statically.
When connecting to existing static oids,
this pointer can be obtained with the
.Fn SYSCTL_STATIC_CHILDREN
macro, where the
.Fa OID_NAME
argument is name of the parent oid of type
.Dv CTLTYPE_NODE
(i.e., the name displayed by
.Xr sysctl 8 ,
preceded by underscore, and with all dots replaced with underscores).
.Pp
When connecting to an existing dynamic oid, this pointer
can be obtained with the
.Fn SYSCTL_CHILDREN
macro, where the
.Fa oidp
argument points to the parent oid of type
.Dv CTLTYPE_NODE .
.Pp
The
.Fn sysctl_add_oid
function creates raw oids of any type.
If the oid is successfully created,
the function returns a pointer to it;
otherwise it returns
.Dv NULL .
Many of the arguments for
.Fn sysctl_add_oid
are common to the macros.
The arguments are as follows:
.Bl -tag -width handler
.It Fa ctx
A pointer to an optional sysctl context, or
.Dv NULL .
See
.Xr sysctl_ctx_init 9
for details.
Programmers are strongly advised to use contexts
to organize the dynamic oids which they create,
unless special creation and deletion sequences are required.
If
.Fa ctx
is not
.Dv NULL ,
the newly created oid will be added to this context
as its first entry.
.It Fa parent
A pointer to a
.Li struct sysctl_oid_list ,
which is the head of the parent's list of children.
.It Fa number
The oid number that will be assigned to this oid.
In almost all cases this should be set to
.Dv OID_AUTO ,
which will result in the assignment of the next available oid number.
.It Fa name
The name of the oid.
The newly created oid will contain a copy of the name.
.It Fa kind
The kind of oid,
specified as a bit mask of the type and access values defined in the
.In sys/sysctl.h
header file.
Oids created dynamically always have the
.Dv CTLFLAG_DYN
flag set.
Access flags specify whether this oid is read-only or read-write,
and whether it may be modified by all users
or by the superuser only.
.It Fa arg1
A pointer to any data that the oid should reference, or
.Dv NULL .
.It Fa arg2
The size of
.Fa arg1 ,
or 0 if
.Fa arg1
is
.Dv NULL .
.It Fa handler
A pointer to the function
that is responsible for handling read and write requests
to this oid.
There are several standard handlers
that support operations on nodes,
integers, strings and opaque objects.
It is possible also to define new handlers using the
.Fn SYSCTL_ADD_PROC
macro.
.It Fa format
A pointer to a string
which specifies the format of the oid symbolically.
This format is used as a hint by
.Xr sysctl 8
to apply proper data formatting for display purposes.
Currently used format names are:
.Dq N
for node,
.Dq A
for
.Li "char *" ,
.Dq I
for
.Li "int" ,
.Dq IU
for
.Li "unsigned int" ,
.Dq L
for
.Li "long" ,
.Dq LU
for
.Li "unsigned long"
and
.Dq S,TYPE
for
.Li "struct TYPE"
structures.
.It Fa descr
A pointer to a textual description of the oid.
.El
.Pp
The
.Fn sysctl_remove_oid
function removes a dynamically created oid from the tree,
optionally freeing its resources.
It takes the following arguments:
.Bl -tag -width recurse
.It Fa oidp
A pointer to the dynamic oid to be removed.
If the oid is not dynamic, or the pointer is
.Dv NULL ,
the function returns
.Er EINVAL .
.It Fa del
If non-zero,
.Fn sysctl_remove_oid
will try to free the oid's resources
when the reference count of the oid becomes zero.
However, if
.Fa del
is set to 0,
the routine will only deregister the oid from the tree,
without freeing its resources.
This behaviour is useful when the caller expects to rollback
(possibly partially failed)
deletion of many oids later.
.It Fa recurse
If non-zero, attempt to remove the node and all its children.
If
.Fa recurse
is set to 0,
any attempt to remove a node that contains any children
will result in a
.Er ENOTEMPTY
error.
.Em WARNING : "use recursive deletion with extreme caution" !
Normally it should not be needed if contexts are used.
Contexts take care of tracking inter-dependencies
between users of the tree.
However, in some extreme cases it might be necessary
to remove part of the subtree no matter how it was created,
in order to free some other resources.
Be aware, though, that this may result in a system
.Xr panic 9
if other code sections continue to use removed subtrees.
.El
.Pp
.\" XXX sheldonh finished up to here
Again, in most cases the programmer should use contexts,
as described in
.Xr sysctl_ctx_init 9 ,
to keep track of created oids,
and to delete them later in orderly fashion.
.Pp
There is a set of macros defined
that helps to create oids of given type.
.Pp
They are as follows:
.Bl -tag -width SYSCTL_ADD_STRINGXX
.It Fn SYSCTL_ADD_OID
creates a raw oid.
This macro is functionally equivalent to the
.Fn sysctl_add_oid
function.
.It Fn SYSCTL_ADD_NODE
creates an oid of type
.Dv CTLTYPE_NODE ,
to which child oids may be added.
.It Fn SYSCTL_ADD_STRING
creates an oid that handles a zero-terminated character string.
.It Fn SYSCTL_ADD_INT
creates an oid that handles an
.Li int
variable.
.It Fn SYSCTL_ADD_QUAD
creates an oid that handles a 64-bit
.Li int
variable.
.It Fn SYSCTL_ADD_UQUAD
creates an oid that handles a 64-bit
.Li unsigned int
variable.
.It Fn SYSCTL_ADD_UINT
creates an oid that handles an
.Li unsigned int
variable.
.It Fn SYSCTL_ADD_LONG
creates an oid that handles a
.Li long
variable.
.It Fn SYSCTL_ADD_ULONG
creates an oid that handles an
.Li unsigned long
variable.
.It Fn SYSCTL_ADD_OPAQUE
creates an oid that handles any chunk of opaque data
of the size specified by the
.Fa len
argument,
which is a pointer to a
.Li "size_t *" .
.It Fn SYSCTL_ADD_STRUCT
creates an oid that handles a
.Li "struct TYPE"
structure.
The
.Fa format
parameter will be set to
.Dq S,TYPE
to provide proper hints to the
.Xr sysctl 8
utility.
.It Fn SYSCTL_ADD_PROC
creates an oid with the specified
.Fa handler
function.
The handler is responsible for handling read and write requests
to the oid.
This oid type is especially useful
if the kernel data is not easily accessible,
or needs to be processed before exporting.
.El
.Sh EXAMPLES
The following is an example of
how to create a new top-level category
and how to hook up another subtree to an existing static node.
This example does not use contexts,
which results in tedious management of all intermediate oids,
as they need to be freed later on:
.Bd -literal
#include <sys/sysctl.h>
 ...
/* Need to preserve pointers to newly created subtrees, to be able
 * to free them later.
 */
struct sysctl_oid *root1, *root2, *oidp;
int a_int;
char *string = "dynamic sysctl";
 ...

root1 = SYSCTL_ADD_NODE( NULL, SYSCTL_STATIC_CHILDREN(/* tree top */),
	OID_AUTO, "newtree", CTLFLAG_RW, 0, "new top level tree");
oidp = SYSCTL_ADD_INT( NULL, SYSCTL_CHILDREN(root1),
	OID_AUTO, "newint", CTLFLAG_RW, &a_int, 0, "new int leaf");
 ...
root2 = SYSCTL_ADD_NODE( NULL, SYSCTL_STATIC_CHILDREN(_debug),
	OID_AUTO, "newtree", CTLFLAG_RW, 0, "new tree under debug");
oidp = SYSCTL_ADD_STRING( NULL, SYSCTL_CHILDREN(root2),
	OID_AUTO, "newstring", CTLFLAG_RD, string, 0, "new string leaf");
.Ed
.Pp
This example creates the following subtrees:
.Bd -literal -offset indent
debug.newtree.newstring
newtree.newint
.Ed
.Pp
.Em "Care should be taken to free all oids once they are no longer needed!"
.Sh SEE ALSO
.Xr sysctl 8 ,
.Xr sysctl 9 ,
.Xr sysctl_ctx_free 9 ,
.Xr sysctl_ctx_init 9
.Sh HISTORY
These functions first appeared in
.Fx 4.2 .
.Sh AUTHORS
.An Andrzej Bialecki Aq abial@FreeBSD.org
.Sh BUGS
Sharing nodes between many code sections
causes interdependencies that sometimes may lock the resources.
For example,
if module A hooks up a subtree to an oid created by module B,
module B will be unable to delete that oid.
These issues are handled properly by sysctl contexts.
.Pp
Many operations on the tree involve traversing linked lists.
For this reason, oid creation and removal is relatively costly.