mibif_notify_f, mib_netsock, mib_if_set_dyn, mib_refresh_iflist,
mib_find_if, mib_find_if_sys, mib_find_if_name, mib_first_if,
mib_next_if, mib_register_newif, mib_unregister_newif, mib_fetch_ifmib,
mib_if_admin, mib_find_ifa, mib_first_ififa, mib_next_ififa,
mib_ifstack_create, mib_ifstack_delete, mib_find_rcvaddr,
mib_rcvaddr_create, mib_rcvaddr_delete, mibif_notify, mibif_unnotify --
mib-2 module for snmpd.
(begemotSnmpdModulePath."mibII" = /usr/local/lib/snmp_mibII.so)
#include <bsnmp/snmpmod.h>
#include <bsnmp/snmp_mibII.h>
typedef void
(*mibif_notify_f)(struct mibif *ifp, enum mibif_notify event,
void *uarg);
extern int mib_netsock;
void
mib_if_set_dyn(const char *ifname);
void
mib_refresh_iflist(void);
struct mibif *
mib_find_if(u_int ifindex);
struct mibif *
mib_find_if_sys(u_int sysindex);
struct mibif *
mib_find_if_name(const char *ifname);
struct mibif *
mib_first_if(void);
struct mibif *
mib_next_if(const struct mibif *ifp);
int
mib_register_newif(int (*func)(struct mibif *),
const struct lmodule *mod);
void
mib_unregister_newif(const struct lmodule *mod);
int
mib_fetch_ifmib(struct mibif *ifp);
int
mib_if_admin(struct mibif *ifp, int up);
struct mibifa *
mib_find_ifa(struct in_addr ipa);
struct mibifa *
mib_first_ififa(const struct mibif *ifp);
struct mibifa *
mib_next_ififa(struct mibifa *ifa);
int
mib_ifstack_create(const struct mibif *lower, const struct mibif *upper);
void
mib_ifstack_delete(const struct mibif *lower, const struct mibif *upper);
struct mibrcvaddr *
mib_find_rcvaddr(u_int ifindex, const u_char *addr, size_t addrlen);
struct mibrcvaddr *
mib_rcvaddr_create(struct mibif *ifp, const u_char *addr,
size_t addrlen);
void
mib_rcvaddr_delete(struct mibrcvaddr *addr);
void *
mibif_notify(struct mibif *ifp, const struct lmodule *mod,
mibif_notify_f func, void *uarg);
void
mibif_unnotify(void *reg);
The snmp_mibII module implements parts of the internet standard MIB-2.
Most of the relevant MIBs are implemented. Some of the tables are
restricted to be read-only instead of read-write. The exact current
implementation can be found in /usr/local/include/bsnmp/mibII_tree.def.
The module also exports a number of functions and global variables for
use by other modules, that need to handle network interfaces. This man
page describes these functions.
DIRECT NETWORK ACCESS [Toc] [Back]
The mibif_notify_f module opens a socket that is used to execute all network
related ioctl(2) functions. This socket is globally available under
the name mib_netsock.
NETWORK INTERFACES [Toc] [Back]
The mibif_notify_f module handles a list of all currently existing network
interfaces. It allows other modules to handle their own interface
lists with special information by providing a mechanism to register to
events that change the interface list (see below). The basic data structure
is the interface structure:
struct mibif {
TAILQ_ENTRY(mibif) link;
u_int flags;
u_int index; /* logical ifindex */
u_int sysindex;
char name[IFNAMSIZ];
char descr[256];
struct ifmibdata mib;
u_int32_t mibtick;
void *specmib;
size_t specmiblen;
u_char *physaddr;
u_int physaddrlen;
int has_connector;
int trap_enable;
u_int32_t counter_disc;
mibif_notify_f xnotify;
void *xnotify_data;
const struct lmodule *xnotify_mod;
};
The mibif_notify_f module tries to implement the semantic if ifIndex as
described in RFC-2863. This RFC states, that an interface indexes may not
be reused. That means, for example, if tun is a synthetic interface type
and the system creates the interface tun0, destroys this interfaces and
again creates a tun 0, then these interfaces must have different interface
indexes, because in fact they are different interfaces. If, on the
other hand, there is a hardware interface xl0 and this interface disappears,
because its driver is unloaded and appears again, because the
driver is loaded again, the interface index must stay the same.
mibif_notify_f implements this by differentiating between real and synthetic
(dynamic) interfaces. An interface type can be declared dynamic by
calling the function mib_if_set_dyn() with the name if the interface type
(for example "tun )." For real interfaces, the module keeps the mapping
between the interface name and its ifIndex in a special list, if the
interface is unloaded. For dynamic interfaces a new ifIndex is generated
each time the interface comes into existance. This means, that the interface
index as seen by SNMP is not the same index as used by the system.
The SNMP ifIndex is held in field index, the system's interface index is
sysindex.
A call to mib_refresh_iflist causes the entire interface list to be recreated.
The interface list can be traversed with the functions mib_first_if() and
mib_next_if(). Be sure not to change the interface list while traversing
the list with these two calls.
There are three functions to find an interface by name or index.
mib_find_if() finds an interface by searching for an SNMP ifIndex,
mib_find_if_sys() finds an interface by searching for a system interface
index and mib_find_if_name() finds an interface by looking for an interface
name. Each of the function returns NULL if the interface cannot be
found.
The function mib_fetch_ifmib() causes the interface MIB to be refreshed
from the kernel.
The function mib_if_admin() can be used to change the interface administrative
state to up (argument is 1) or down (argument is 0).
INTERFACE EVENTS [Toc] [Back]
A module can register itself to receive a notification when a new entry
is created in the interface list. This is done by calling
mib_register_newif(). A module can register only one function, a second
call to mib_register_newif() causes the registration to be overwritten.
The registration can be removed with a call to mib_unregister_newif().
If is unregistered automatically, when the registering module is
unloaded.
A module can also register to events on a specific interface. This is
done by calling mibif_notify(). This causes the given callback func to
be called with the interface pointer, a notification code and the user
argument uarg when any of the following events occur:
MIBIF_NOTIFY_DESTROY
The interface is destroyed.
This mechanism can be used to implement interface type specific MIB parts
in other modules. The registration can be removed with mib_unnotify()
which the return value from mib_notify. Any notification registration is
removed automatically when the interface is destroyed or the registering
module is unloaded. Note that only one module can register to any given
interface.
INTERFACE ADDRESSES [Toc] [Back]
The mibif_notify_f module handles a table of interface IP-addresses.
These addresses are held in a
struct mibifa {
TAILQ_ENTRY(mibifa) link;
struct in_addr inaddr;
struct in_addr inmask;
struct in_addr inbcast;
struct asn_oid index;
u_int ifindex;
u_int flags;
};
The (ordered) list of IP-addresses on a given interface can be traversed
by calling mib_first_ififa() and mib_next_ififa(). The list should not
be considered read-only.
INTERFACE RECEIVE ADDRESSES [Toc] [Back]
The internet MIB-2 contains a table of interface receive addresses. These
addresses are handled in:
struct mibrcvaddr {
TAILQ_ENTRY(mibrcvaddr) link;
struct asn_oid index;
u_int ifindex;
u_char addr[ASN_MAXOIDLEN];
size_t addrlen;
u_int flags;
};
enum {
MIBRCVADDR_VOLATILE = 0x00000001,
MIBRCVADDR_BCAST = 0x00000002,
MIBRCVADDR_HW = 0x00000004,
};
Note, that the assignment of MIBRCVADDR_BCAST is based on a list of known
interface types. The flags should be handled by modules inplementing
interface type specific MIBs.
A receive address can be created with mib_rcvaddr_create() and deleted
with mib_rcvaddr_delete(). This needs to be done only for addresses that
are not automatically handled by the system.
A receive address can be found with mib_find_rcvaddr().
INTERFACE STACK TABLE [Toc] [Back]
The mibif_notify_f module maintains also the interface stack table.
Because for complex stacks, there is no system supported generic way of
getting this information, interface type specific modules need to help
setting up stack entries. The mibif_notify_f module handles only the top
and bottom entries.
A table entry is created with mib_ifstack_create() and deleted with
mib_ifstack_delete(). Both functions need the pointers to the interfaces.
Entries are automatically deleted if any of the interfaces of the
entry is destroyed. The functions handle both the stack table and the
reverse stack table.
/usr/local/include/bsnmp/mibII_tree.def The description of the MIB tree
implemented by mibif_notify_f.
/usr/local/share/snmp/mibs The various internet MIBs.
snmpmod(3), gensnmptree(1)
This implementation conforms to the applicable IETF RFCs.
Hartmut Brandt <[email protected]>
FreeBSD 5.2.1 August 19, 2002 FreeBSD 5.2.1 [ Back ] |