1.. SPDX-License-Identifier: GPL-2.0 2 3================ 4Kernel Connector 5================ 6 7Kernel connector - new netlink based userspace <-> kernel space easy 8to use communication module. 9 10The Connector driver makes it easy to connect various agents using a 11netlink based network. One must register a callback and an identifier. 12When the driver receives a special netlink message with the appropriate 13identifier, the appropriate callback will be called. 14 15From the userspace point of view it's quite straightforward: 16 17 - socket(); 18 - bind(); 19 - send(); 20 - recv(); 21 22But if kernelspace wants to use the full power of such connections, the 23driver writer must create special sockets, must know about struct sk_buff 24handling, etc... The Connector driver allows any kernelspace agents to use 25netlink based networking for inter-process communication in a significantly 26easier way:: 27 28 int cn_add_callback(struct cb_id *id, char *name, void (*callback) (struct cn_msg *, struct netlink_skb_parms *)); 29 void cn_netlink_send_multi(struct cn_msg *msg, u16 len, u32 portid, u32 __group, int gfp_mask); 30 void cn_netlink_send(struct cn_msg *msg, u32 portid, u32 __group, int gfp_mask); 31 32 struct cb_id 33 { 34 __u32 idx; 35 __u32 val; 36 }; 37 38idx and val are unique identifiers which must be registered in the 39connector.h header for in-kernel usage. `void (*callback) (void *)` is a 40callback function which will be called when a message with above idx.val 41is received by the connector core. The argument for that function must 42be dereferenced to `struct cn_msg *`:: 43 44 struct cn_msg 45 { 46 struct cb_id id; 47 48 __u32 seq; 49 __u32 ack; 50 51 __u32 len; /* Length of the following data */ 52 __u8 data[0]; 53 }; 54 55Connector interfaces 56==================== 57 58 .. kernel-doc:: include/linux/connector.h 59 60 Note: 61 When registering new callback user, connector core assigns 62 netlink group to the user which is equal to its id.idx. 63 64Protocol description 65==================== 66 67The current framework offers a transport layer with fixed headers. The 68recommended protocol which uses such a header is as following: 69 70msg->seq and msg->ack are used to determine message genealogy. When 71someone sends a message, they use a locally unique sequence and random 72acknowledge number. The sequence number may be copied into 73nlmsghdr->nlmsg_seq too. 74 75The sequence number is incremented with each message sent. 76 77If you expect a reply to the message, then the sequence number in the 78received message MUST be the same as in the original message, and the 79acknowledge number MUST be the same + 1. 80 81If we receive a message and its sequence number is not equal to one we 82are expecting, then it is a new message. If we receive a message and 83its sequence number is the same as one we are expecting, but its 84acknowledge is not equal to the sequence number in the original 85message + 1, then it is a new message. 86 87Obviously, the protocol header contains the above id. 88 89The connector allows event notification in the following form: kernel 90driver or userspace process can ask connector to notify it when 91selected ids will be turned on or off (registered or unregistered its 92callback). It is done by sending a special command to the connector 93driver (it also registers itself with id={-1, -1}). 94 95As example of this usage can be found in the cn_test.c module which 96uses the connector to request notification and to send messages. 97 98Reliability 99=========== 100 101Netlink itself is not a reliable protocol. That means that messages can 102be lost due to memory pressure or process' receiving queue overflowed, 103so caller is warned that it must be prepared. That is why the struct 104cn_msg [main connector's message header] contains u32 seq and u32 ack 105fields. 106 107Userspace usage 108=============== 109 1102.6.14 has a new netlink socket implementation, which by default does not 111allow people to send data to netlink groups other than 1. 112So, if you wish to use a netlink socket (for example using connector) 113with a different group number, the userspace application must subscribe to 114that group first. It can be achieved by the following pseudocode:: 115 116 s = socket(PF_NETLINK, SOCK_DGRAM, NETLINK_CONNECTOR); 117 118 l_local.nl_family = AF_NETLINK; 119 l_local.nl_groups = 12345; 120 l_local.nl_pid = 0; 121 122 if (bind(s, (struct sockaddr *)&l_local, sizeof(struct sockaddr_nl)) == -1) { 123 perror("bind"); 124 close(s); 125 return -1; 126 } 127 128 { 129 int on = l_local.nl_groups; 130 setsockopt(s, 270, 1, &on, sizeof(on)); 131 } 132 133Where 270 above is SOL_NETLINK, and 1 is a NETLINK_ADD_MEMBERSHIP socket 134option. To drop a multicast subscription, one should call the above socket 135option with the NETLINK_DROP_MEMBERSHIP parameter which is defined as 0. 136 1372.6.14 netlink code only allows to select a group which is less or equal to 138the maximum group number, which is used at netlink_kernel_create() time. 139In case of connector it is CN_NETLINK_USERS + 0xf, so if you want to use 140group number 12345, you must increment CN_NETLINK_USERS to that number. 141Additional 0xf numbers are allocated to be used by non-in-kernel users. 142 143Due to this limitation, group 0xffffffff does not work now, so one can 144not use add/remove connector's group notifications, but as far as I know, 145only cn_test.c test module used it. 146 147Some work in netlink area is still being done, so things can be changed in 1482.6.15 timeframe, if it will happen, documentation will be updated for that 149kernel. 150 151Code samples 152============ 153 154Sample code for a connector test module and user space can be found 155in samples/connector/. To build this code, enable CONFIG_CONNECTOR 156and CONFIG_SAMPLES. 157