Root/Documentation/connector/connector.txt

1/*****************************************/
2Kernel Connector.
3/*****************************************/
4
5Kernel connector - new netlink based userspace <-> kernel space easy
6to use communication module.
7
8The Connector driver makes it easy to connect various agents using a
9netlink based network. One must register a callback and an identifier.
10When the driver receives a special netlink message with the appropriate
11identifier, the appropriate callback will be called.
12
13From the userspace point of view it's quite straightforward:
14
15    socket();
16    bind();
17    send();
18    recv();
19
20But if kernelspace wants to use the full power of such connections, the
21driver writer must create special sockets, must know about struct sk_buff
22handling, etc... The Connector driver allows any kernelspace agents to use
23netlink based networking for inter-process communication in a significantly
24easier way:
25
26int cn_add_callback(struct cb_id *id, char *name, void (*callback) (struct cn_msg *, struct netlink_skb_parms *));
27void cn_netlink_send(struct cn_msg *msg, u32 __group, int gfp_mask);
28
29struct cb_id
30{
31    __u32 idx;
32    __u32 val;
33};
34
35idx and val are unique identifiers which must be registered in the
36connector.h header for in-kernel usage. void (*callback) (void *) is a
37callback function which will be called when a message with above idx.val
38is received by the connector core. The argument for that function must
39be dereferenced to struct cn_msg *.
40
41struct cn_msg
42{
43    struct cb_id id;
44
45    __u32 seq;
46    __u32 ack;
47
48    __u32 len; /* Length of the following data */
49    __u8 data[0];
50};
51
52/*****************************************/
53Connector interfaces.
54/*****************************************/
55
56int cn_add_callback(struct cb_id *id, char *name, void (*callback) (struct cn_msg *, struct netlink_skb_parms *));
57
58 Registers new callback with connector core.
59
60 struct cb_id *id - unique connector's user identifier.
61                  It must be registered in connector.h for legal in-kernel users.
62 char *name - connector's callback symbolic name.
63 void (*callback) (struct cn..) - connector's callback.
64                  cn_msg and the sender's credentials
65
66
67void cn_del_callback(struct cb_id *id);
68
69 Unregisters new callback with connector core.
70
71 struct cb_id *id - unique connector's user identifier.
72
73
74int cn_netlink_send(struct cn_msg *msg, u32 __groups, int gfp_mask);
75
76 Sends message to the specified groups. It can be safely called from
77 softirq context, but may silently fail under strong memory pressure.
78 If there are no listeners for given group -ESRCH can be returned.
79
80 struct cn_msg * - message header(with attached data).
81 u32 __group - destination group.
82                  If __group is zero, then appropriate group will
83                  be searched through all registered connector users,
84                  and message will be delivered to the group which was
85                  created for user with the same ID as in msg.
86                  If __group is not zero, then message will be delivered
87                  to the specified group.
88 int gfp_mask - GFP mask.
89
90 Note: When registering new callback user, connector core assigns
91 netlink group to the user which is equal to it's id.idx.
92
93/*****************************************/
94Protocol description.
95/*****************************************/
96
97The current framework offers a transport layer with fixed headers. The
98recommended protocol which uses such a header is as following:
99
100msg->seq and msg->ack are used to determine message genealogy. When
101someone sends a message, they use a locally unique sequence and random
102acknowledge number. The sequence number may be copied into
103nlmsghdr->nlmsg_seq too.
104
105The sequence number is incremented with each message sent.
106
107If you expect a reply to the message, then the sequence number in the
108received message MUST be the same as in the original message, and the
109acknowledge number MUST be the same + 1.
110
111If we receive a message and its sequence number is not equal to one we
112are expecting, then it is a new message. If we receive a message and
113its sequence number is the same as one we are expecting, but its
114acknowledge is not equal to the acknowledge number in the original
115message + 1, then it is a new message.
116
117Obviously, the protocol header contains the above id.
118
119The connector allows event notification in the following form: kernel
120driver or userspace process can ask connector to notify it when
121selected ids will be turned on or off (registered or unregistered its
122callback). It is done by sending a special command to the connector
123driver (it also registers itself with id={-1, -1}).
124
125As example of this usage can be found in the cn_test.c module which
126uses the connector to request notification and to send messages.
127
128/*****************************************/
129Reliability.
130/*****************************************/
131
132Netlink itself is not a reliable protocol. That means that messages can
133be lost due to memory pressure or process' receiving queue overflowed,
134so caller is warned that it must be prepared. That is why the struct
135cn_msg [main connector's message header] contains u32 seq and u32 ack
136fields.
137
138/*****************************************/
139Userspace usage.
140/*****************************************/
141
1422.6.14 has a new netlink socket implementation, which by default does not
143allow people to send data to netlink groups other than 1.
144So, if you wish to use a netlink socket (for example using connector)
145with a different group number, the userspace application must subscribe to
146that group first. It can be achieved by the following pseudocode:
147
148s = socket(PF_NETLINK, SOCK_DGRAM, NETLINK_CONNECTOR);
149
150l_local.nl_family = AF_NETLINK;
151l_local.nl_groups = 12345;
152l_local.nl_pid = 0;
153
154if (bind(s, (struct sockaddr *)&l_local, sizeof(struct sockaddr_nl)) == -1) {
155    perror("bind");
156    close(s);
157    return -1;
158}
159
160{
161    int on = l_local.nl_groups;
162    setsockopt(s, 270, 1, &on, sizeof(on));
163}
164
165Where 270 above is SOL_NETLINK, and 1 is a NETLINK_ADD_MEMBERSHIP socket
166option. To drop a multicast subscription, one should call the above socket
167option with the NETLINK_DROP_MEMBERSHIP parameter which is defined as 0.
168
1692.6.14 netlink code only allows to select a group which is less or equal to
170the maximum group number, which is used at netlink_kernel_create() time.
171In case of connector it is CN_NETLINK_USERS + 0xf, so if you want to use
172group number 12345, you must increment CN_NETLINK_USERS to that number.
173Additional 0xf numbers are allocated to be used by non-in-kernel users.
174
175Due to this limitation, group 0xffffffff does not work now, so one can
176not use add/remove connector's group notifications, but as far as I know,
177only cn_test.c test module used it.
178
179Some work in netlink area is still being done, so things can be changed in
1802.6.15 timeframe, if it will happen, documentation will be updated for that
181kernel.
182

Archive Download this file



interactive