DragonFly On-Line Manual Pages

Search: Section:  


NG_PPPOE(4)           DragonFly Kernel Interfaces Manual           NG_PPPOE(4)

NAME

ng_pppoe -- RFC 2516 PPPOE protocol netgraph node type

SYNOPSIS

#include <net/ethernet.h> #include <netgraph/pppoe/ng_pppoe.h>

DESCRIPTION

The pppoe node type performs the PPPoE protocol. It is used in conjunction with the netgraph(4) extensions to the Ethernet framework to divert and inject Ethernet packets to and from a PPP agent (which is not specified). The NGM_PPPOE_GET_STATUS control message can be used at any time to query the current status of the PPPOE module. The only statistics presently available are the total packet counts for input and output. This node does not yet support the NGM_TEXT_STATUS control message.

HOOKS

This node type supports the following hooks: ethernet The hook that should normally be connected to an Ethernet node. debug Presently no use. [unspecified] Any other name is assumed to be a session hook that will be connected to a PPP client agent, or a ppp server agent.

CONTROL MESSAGES

This node type supports the generic control messages, plus the following: NGM_PPPOE_GET_STATUS This command returns status information in a struct ngpppoestat: struct ngpppoestat { u_int packets_in; /* packets in from ethernet */ u_int packets_out; /* packets out towards ethernet */ }; NGM_TEXT_STATUS This generic message returns is a human-readable version of the node status. (not yet) NGM_PPPOE_CONNECT Tell a nominated newly created hook that its session should enter the state machine in a manner to become a client. It must be newly created and a service name can be given as an argument. It is legal to specify a zero length service name. This is common on some DSL setups. A session request packet will be broadcast on the Ethernet. This command uses the ngpppoe_init_data structure shown below. NGM_PPPOE_LISTEN Tell a nominated newly created hook that its session should enter the state machine in a manner to become a server listener. The argument given is the name of the service to listen on behalf of. A zero length service length will match all requests for service. A matching service request packet will be passed unmodified back to the process responsible for starting the service. It can then examine it and pass it on to the session that is started to answer the request. This command uses the ngpppoe_init_data structure shown below. NGM_PPPOE_OFFER Tell a nominated newly created hook that its session should enter the state machine in a manner to become a server. The argument given is the name of the service to offer. A zero length service is legal. The State machine will progress to a state where it will await a request packet to be forwarded to it from the startup server, which in turn probably received it from a LISTEN mode hook ( see above). This is so that information that is required for the session that is embedded in the original session request packet, is made available to the state machine that eventually answers the request. When the Session request packet is received, the session negotiation will proceed. This command uses the ngpppoe_init_data structure shown below. The three commands above use a common data structure: struct ngpppoe_init_data { char hook[NG_HOOKSIZ]; /* hook to monitor on */ u_int16_t data_len; /* service name length */ char data[0]; /* init data goes here */ }; NGM_PPPOE_SUCCESS This command is sent to the node that started this session with one of the above messages, and reports a state change. This message reports successful Session negotiation. It uses the structure shown below, and reports back the hook name corresponding to the successful session. NGM_NGM_PPPOE_FAIL This command is sent to the node that started this session with one of the above messages, and reports a state change. This message reports failed Session negotiation. It uses the structure shown below, and reports back the hook name corresponding to the failed session. The hook will probably have been removed immediately after sending this message NGM_NGM_PPPOE_CLOSE This command is sent to the node that started this session with one of the above messages, and reports a state change. This message reports a request to close a session. It uses the structure shown below, and reports back the hook name corresponding to the closed session. The hook will probably have been removed immediately after sending this message. At present this message is not yet used and a 'failed' message will be received at closure instead. The three commands above use a common data structure: struct ngpppoe_sts { char hook[NG_HOOKSIZ]; /* hook associated with event session */ };

SHUTDOWN

This node shuts down upon receipt of a NGM_SHUTDOWN control message, when all session have been disconnected or when the ethernet hook is disconnected.

EXAMPLES

The following code uses libnetgraph to set up a ng_pppoe node and connect it to both a socket node and an Ethernet node. It can handle the case of when a ng_pppoe node is already attached to the Ethernet. It then starts a client session. #include <stdio.h> #include <stdlib.h> #include <string.h> #include <ctype.h> #include <unistd.h> #include <sysexits.h> #include <errno.h> #include <err.h> #include <sys/types.h> #include <sys/socket.h> #include <sys/select.h> #include <net/ethernet.h> #include <netgraph.h> #include <netgraph/ng_ether.h> #include <netgraph/ng_pppoe.h> #include <netgraph/ng_socket.h> static int setup(char *ethername, char *service, char *sessname, int *dfd, int *cfd); int main() { int fd1, fd2; setup("xl0", NULL, "fred", &fd1, &fd2); sleep (30); } static int setup(char *ethername, char *service, char *sessname, int *dfd, int *cfd) { struct ngm_connect ngc; /* connect */ struct ngm_mkpeer mkp; /* mkpeer */ /******** nodeinfo stuff **********/ u_char rbuf[2 * 1024]; struct ng_mesg *const resp = (struct ng_mesg *) rbuf; struct hooklist *const hlist = (struct hooklist *) resp->data; struct nodeinfo *const ninfo = &hlist->nodeinfo; int ch, no_hooks = 0; struct linkinfo *link; struct nodeinfo *peer; /****message to connect pppoe session*****/ struct { struct ngpppoe_init_data idata; char service[100]; } message; /********tracking our little graph ********/ char path[100]; char source_ID[NG_NODESIZ]; char pppoe_node_name[100]; int k; /* * Create the data and control sockets */ if (NgMkSockNode(NULL, cfd, dfd) < 0) { return (errno); } /* * find the ether node of the name requested by asking it for * its inquiry information. */ if (strlen(ethername) > 16) return (EINVAL); sprintf(path, "%s:", ethername); if (NgSendMsg(*cfd, path, NGM_GENERIC_COOKIE, NGM_LISTHOOKS, NULL, 0) < 0) { return (errno); } /* * the command was accepted so it exists. Await the reply (It's * almost certainly already waiting). */ if (NgRecvMsg(*cfd, resp, sizeof(rbuf), NULL) < 0) { return (errno); } /** * The following is available about the node: * ninfo->name (string) * ninfo->type (string) * ninfo->id (u_int32_t) * ninfo->hooks (u_int32_t) (count of hooks) * check it is the correct type. and get its ID for use * with mkpeer later. */ if (strncmp(ninfo->type, NG_ETHER_NODE_TYPE, strlen(NG_ETHER_NODE_TYPE)) != 0) { return (EPROTOTYPE); } sprintf(source_ID, "[%08x]:", ninfo->id); /* * look for a hook already attached. */ for (k = 0; k < ninfo->hooks; k++) { /** * The following are available about each hook. * link->ourhook (string) * link->peerhook (string) * peer->name (string) * peer->type (string) * peer->id (u_int32_t) * peer->hooks (u_int32_t) */ link = &hlist->link[k]; peer = &hlist->link[k].nodeinfo; /* Ignore debug hooks */ if (strcmp("debug", link->ourhook) == 0) continue; /* If the orphans hook is attached, use that */ if (strcmp(NG_ETHER_HOOK_ORPHAN, link->ourhook) == 0) { break; } /* the other option is the 'divert' hook */ if (strcmp("NG_ETHER_HOOK_DIVERT", link->ourhook) == 0) { break; } } /* * See if we found a hook there. */ if (k < ninfo->hooks) { if (strcmp(peer->type, NG_PPPOE_NODE_TYPE) == 0) { /* * If it's a type pppoe, we skip making one * ourself, but we continue, using * the existing one. */ sprintf(pppoe_node_name, "[%08x]:", peer->id); } else { /* * There is already someone hogging the data, * return an error. Some day we'll try * daisy-chaining.. */ return (EBUSY); } } else { /* * Try make a node of type pppoe against node "ID" * On hook NG_ETHER_HOOK_ORPHAN. */ snprintf(mkp.type, sizeof(mkp.type), "%s", NG_PPPOE_NODE_TYPE); snprintf(mkp.ourhook, sizeof(mkp.ourhook), "%s", NG_ETHER_HOOK_ORPHAN); snprintf(mkp.peerhook, sizeof(mkp.peerhook), "%s", NG_PPPOE_HOOK_ETHERNET); /* Send message */ if (NgSendMsg(*cfd, source_ID, NGM_GENERIC_COOKIE, NGM_MKPEER, &mkp, sizeof(mkp)) < 0) { return (errno); } /* * Work out a name for the new node. */ sprintf(pppoe_node_name, "%s:%s", source_ID, NG_ETHER_HOOK_ORPHAN); } /* * We now have a pppoe node attached to the ethernet * card. The Ethernet is addressed as ethername: The pppoe * node is addressed as pppoe_node_name: attach to it. * Connect socket node to specified node Use the same hook * name on both ends of the link. */ snprintf(ngc.path, sizeof(ngc.path), "%s", pppoe_node_name); snprintf(ngc.ourhook, sizeof(ngc.ourhook), "%s", sessname); snprintf(ngc.peerhook, sizeof(ngc.peerhook), "%s", sessname); if (NgSendMsg(*cfd, ".:", NGM_GENERIC_COOKIE, NGM_CONNECT, &ngc, sizeof(ngc)) < 0) { return (errno); } /* * Send it a message telling it to start up. */ bzero(&message, sizeof(message)); snprintf(message.idata.hook, sizeof(message.idata.hook), "%s", sessname); if (service == NULL) { message.idata.data_len = 0; } else { snprintf(message.idata.data, sizeof(message.idata.data), "%s", service); message.idata.data_len = strlen(service); } /* Tell session/hook to start up as a client */ if (NgSendMsg(*cfd, ngc.path, NGM_PPPOE_COOKIE, NGM_PPPOE_CONNECT, &message.idata, sizeof(message.idata) + message.idata.data_len) < 0) { return (errno); } return (0); }

SEE ALSO

netgraph(3), netgraph(4), ng_ppp(4), ng_socket(4), ngctl(8) L. Mamakos, K. Lidl, J. Evarts, D. Carrel, D. Simone, and R. Wheeler, A Method for transmitting PPP over Ethernet (PPPoE), RFC 2516.

HISTORY

The ng_pppoe node type was implemented in FreeBSD 4.0.

AUTHORS

Julian Elischer <julian@FreeBSD.org> DragonFly 5.3 October 28, 1999 DragonFly 5.3

Search: Section: