Based on kernel version 2.6.26. Page generated on 2008-07-16 21:12 EST.
1 /*****************************************/ 2 Kernel Connector. 3 /*****************************************/ 4 5 Kernel connector - new netlink based userspace <-> kernel space easy 6 to use communication module. 7 8 Connector driver adds possibility to connect various agents using 9 netlink based network. One must register callback and 10 identifier. When driver receives special netlink message with 11 appropriate identifier, appropriate callback will be called. 12 13 From the userspace point of view it's quite straightforward: 14 15 socket(); 16 bind(); 17 send(); 18 recv(); 19 20 But if kernelspace want to use full power of such connections, driver 21 writer must create special sockets, must know about struct sk_buff 22 handling... Connector allows any kernelspace agents to use netlink 23 based networking for inter-process communication in a significantly 24 easier way: 25 26 int cn_add_callback(struct cb_id *id, char *name, void (*callback) (void *)); 27 void cn_netlink_send(struct cn_msg *msg, u32 __group, int gfp_mask); 28 29 struct cb_id 30 { 31 __u32 idx; 32 __u32 val; 33 }; 34 35 idx and val are unique identifiers which must be registered in 36 connector.h for in-kernel usage. void (*callback) (void *) - is a 37 callback function which will be called when message with above idx.val 38 will be received by connector core. Argument for that function must 39 be dereferenced to struct cn_msg *. 40 41 struct 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 /*****************************************/ 53 Connector interfaces. 54 /*****************************************/ 55 56 int cn_add_callback(struct cb_id *id, char *name, void (*callback) (void *)); 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) (void *) - connector's callback. 64 Argument must be dereferenced to struct cn_msg *. 65 66 void cn_del_callback(struct cb_id *id); 67 68 Unregisters new callback with connector core. 69 70 struct cb_id *id - unique connector's user identifier. 71 72 int cn_netlink_send(struct cn_msg *msg, u32 __groups, int gfp_mask); 73 74 Sends message to the specified groups. It can be safely called from 75 softirq context, but may silently fail under strong memory pressure. 76 If there are no listeners for given group -ESRCH can be returned. 77 78 struct cn_msg * - message header(with attached data). 79 u32 __group - destination group. 80 If __group is zero, then appropriate group will 81 be searched through all registered connector users, 82 and message will be delivered to the group which was 83 created for user with the same ID as in msg. 84 If __group is not zero, then message will be delivered 85 to the specified group. 86 int gfp_mask - GFP mask. 87 88 Note: When registering new callback user, connector core assigns 89 netlink group to the user which is equal to it's id.idx. 90 91 /*****************************************/ 92 Protocol description. 93 /*****************************************/ 94 95 Current offers transport layer with fixed header. Recommended 96 protocol which uses such header is following: 97 98 msg->seq and msg->ack are used to determine message genealogy. When 99 someone sends message it puts there locally unique sequence and random 100 acknowledge numbers. Sequence number may be copied into 101 nlmsghdr->nlmsg_seq too. 102 103 Sequence number is incremented with each message to be sent. 104 105 If we expect reply to our message, then sequence number in received 106 message MUST be the same as in original message, and acknowledge 107 number MUST be the same + 1. 108 109 If we receive message and it's sequence number is not equal to one we 110 are expecting, then it is new message. If we receive message and it's 111 sequence number is the same as one we are expecting, but it's 112 acknowledge is not equal acknowledge number in original message + 1, 113 then it is new message. 114 115 Obviously, protocol header contains above id. 116 117 connector allows event notification in the following form: kernel 118 driver or userspace process can ask connector to notify it when 119 selected id's will be turned on or off(registered or unregistered it's 120 callback). It is done by sending special command to connector 121 driver(it also registers itself with id={-1, -1}). 122 123 As example of usage Documentation/connector now contains cn_test.c - 124 testing module which uses connector to request notification and to 125 send messages. 126 127 /*****************************************/ 128 Reliability. 129 /*****************************************/ 130 131 Netlink itself is not reliable protocol, that means that messages can 132 be lost due to memory pressure or process' receiving queue overflowed, 133 so caller is warned must be prepared. That is why struct cn_msg [main 134 connector's message header] contains u32 seq and u32 ack fields. 135 136 /*****************************************/ 137 Userspace usage. 138 /*****************************************/ 139 2.6.14 has a new netlink socket implementation, which by default does not 140 allow to send data to netlink groups other than 1. 141 So, if to use netlink socket (for example using connector) 142 with different group number userspace application must subscribe to 143 that group. It can be achieved by following pseudocode: 144 145 s = socket(PF_NETLINK, SOCK_DGRAM, NETLINK_CONNECTOR); 146 147 l_local.nl_family = AF_NETLINK; 148 l_local.nl_groups = 12345; 149 l_local.nl_pid = 0; 150 151 if (bind(s, (struct sockaddr *)&l_local, sizeof(struct sockaddr_nl)) == -1) { 152 perror("bind"); 153 close(s); 154 return -1; 155 } 156 157 { 158 int on = l_local.nl_groups; 159 setsockopt(s, 270, 1, &on, sizeof(on)); 160 } 161 162 Where 270 above is SOL_NETLINK, and 1 is a NETLINK_ADD_MEMBERSHIP socket 163 option. To drop multicast subscription one should call above socket option 164 with NETLINK_DROP_MEMBERSHIP parameter which is defined as 0. 165 166 2.6.14 netlink code only allows to select a group which is less or equal to 167 the maximum group number, which is used at netlink_kernel_create() time. 168 In case of connector it is CN_NETLINK_USERS + 0xf, so if you want to use 169 group number 12345, you must increment CN_NETLINK_USERS to that number. 170 Additional 0xf numbers are allocated to be used by non-in-kernel users. 171 172 Due to this limitation, group 0xffffffff does not work now, so one can 173 not use add/remove connector's group notifications, but as far as I know, 174 only cn_test.c test module used it. 175 176 Some work in netlink area is still being done, so things can be changed in 177 2.6.15 timeframe, if it will happen, documentation will be updated for that 178 kernel.