Cache Implementation

Collaboration diagram for Cache Implementation:

Data Structures
struct	nl_msgtype
	Message type to cache action association. More...
struct	nl_af_group
	Address family to netlink group association. More...
struct	nl_parser_param
struct	nl_cache_ops
	Cache Operations. More...

Macros
#define	NL_ACT_MAX   (__NL_ACT_MAX - 1)
#define	END_OF_MSGTYPES_LIST   { -1, -1, NULL }
#define	END_OF_GROUP_LIST   AF_UNSPEC, 0

Enumerations
enum	{   NL_ACT_UNSPEC,   NL_ACT_NEW,   NL_ACT_DEL,   NL_ACT_GET,   NL_ACT_SET,   NL_ACT_CHANGE,   __NL_ACT_MAX }

Detailed Description

1) Cache Definition

struct nl_cache_ops my_cache_ops = {
    .co_name        = "route/link",
    .co_protocol        = NETLINK_ROUTE,
    .co_hdrsize     = sizeof(struct ifinfomsg),
    .co_obj_ops     = &my_obj_ops,
};

2)

// The simplest way to fill a cache is by providing a request-update
// function which must trigger a complete dump on the kernel-side of
// whatever the cache covers.
static int my_request_update(struct nl_cache *cache,
                struct nl_sock *socket)
{
    // In this example, we request a full dump of the interface table
    return nl_rtgen_request(socket, RTM_GETLINK, AF_UNSPEC, NLM_F_DUMP);
}

// The resulting netlink messages sent back will be fed into a message
// parser one at a time. The message parser has to extract all relevant
// information from the message and create an object reflecting the
// contents of the message and pass it on to the parser callback function
// provide which will add the object to the cache.
static int my_msg_parser(struct nl_cache_ops *ops, struct sockaddr_nl *who,
                struct nlmsghdr *nlh, struct nl_parser_param *pp)
{
    struct my_obj *obj;

    obj = my_obj_alloc();
    obj->ce_msgtype = nlh->nlmsg_type;

    // Parse the netlink message and continue creating the object.

    err = pp->pp_cb((struct nl_object *) obj, pp);
    if (err < 0)
        goto errout;
}

struct nl_cache_ops my_cache_ops = {
    ...
    .co_request_update  = my_request_update,
    .co_msg_parser      = my_msg_parser,
};

3) Notification based Updates

// Caches can be kept up-to-date based on notifications if the kernel
// sends out notifications whenever an object is added/removed/changed.
//
// It is trivial to support this, first a list of groups needs to be
// defined which are required to join in order to receive all necessary
// notifications. The groups are separated by address family to support
// the common situation where a separate group is used for each address
// family. If there is only one group, simply specify AF_UNSPEC.
static struct nl_af_group addr_groups[] = {
    { AF_INET,  RTNLGRP_IPV4_IFADDR },
    { AF_INET6, RTNLGRP_IPV6_IFADDR },
    { END_OF_GROUP_LIST },
};

// In order for the caching system to know the meaning of each message
// type it requires a table which maps each supported message type to
// a cache action, e.g. RTM_NEWADDR means address has been added or
// updated, RTM_DELADDR means address has been removed.
static struct nl_cache_ops rtnl_addr_ops = {
    ...
    .co_msgtypes        = {
                    { RTM_NEWADDR, NL_ACT_NEW, "new" },
                    { RTM_DELADDR, NL_ACT_DEL, "del" },
                    { RTM_GETADDR, NL_ACT_GET, "get" },
                    END_OF_MSGTYPES_LIST,
                },
    .co_groups      = addr_groups,
};

// It is now possible to keep the cache up-to-date using the cache manager.

Macro Definition Documentation

#define END_OF_GROUP_LIST   AF_UNSPEC, 0

#define END_OF_MSGTYPES_LIST   { -1, -1, NULL }

#define NL_ACT_MAX   (__NL_ACT_MAX - 1)

Enumeration Type Documentation

anonymous enum

Enumerator
NL_ACT_UNSPEC
NL_ACT_NEW
NL_ACT_DEL
NL_ACT_GET
NL_ACT_SET
NL_ACT_CHANGE
__NL_ACT_MAX