1.\" Copyright (c) 1996-1999 Whistle Communications, Inc. 2.\" All rights reserved. 3.\" 4.\" Subject to the following obligations and disclaimer of warranty, use and 5.\" redistribution of this software, in source or object code forms, with or 6.\" without modifications are expressly permitted by Whistle Communications; 7.\" provided, however, that: 8.\" 1. Any and all reproductions of the source or object code must include the 9.\" copyright notice above and the following disclaimer of warranties; and 10.\" 2. No rights are granted, in any manner or form, to use Whistle 11.\" Communications, Inc. trademarks, including the mark "WHISTLE 12.\" COMMUNICATIONS" on advertising, endorsements, or otherwise except as 13.\" such appears in the above copyright notice or in the software. 14.\" 15.\" THIS SOFTWARE IS BEING PROVIDED BY WHISTLE COMMUNICATIONS "AS IS", AND 16.\" TO THE MAXIMUM EXTENT PERMITTED BY LAW, WHISTLE COMMUNICATIONS MAKES NO 17.\" REPRESENTATIONS OR WARRANTIES, EXPRESS OR IMPLIED, REGARDING THIS SOFTWARE, 18.\" INCLUDING WITHOUT LIMITATION, ANY AND ALL IMPLIED WARRANTIES OF 19.\" MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, OR NON-INFRINGEMENT. 20.\" WHISTLE COMMUNICATIONS DOES NOT WARRANT, GUARANTEE, OR MAKE ANY 21.\" REPRESENTATIONS REGARDING THE USE OF, OR THE RESULTS OF THE USE OF THIS 22.\" SOFTWARE IN TERMS OF ITS CORRECTNESS, ACCURACY, RELIABILITY OR OTHERWISE. 23.\" IN NO EVENT SHALL WHISTLE COMMUNICATIONS BE LIABLE FOR ANY DAMAGES 24.\" RESULTING FROM OR ARISING OUT OF ANY USE OF THIS SOFTWARE, INCLUDING 25.\" WITHOUT LIMITATION, ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, 26.\" PUNITIVE, OR CONSEQUENTIAL DAMAGES, PROCUREMENT OF SUBSTITUTE GOODS OR 27.\" SERVICES, LOSS OF USE, DATA OR PROFITS, HOWEVER CAUSED AND UNDER ANY 28.\" THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 29.\" (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 30.\" THIS SOFTWARE, EVEN IF WHISTLE COMMUNICATIONS IS ADVISED OF THE POSSIBILITY 31.\" OF SUCH DAMAGE. 32.\" 33.\" Authors: Julian Elischer <julian@FreeBSD.org> 34.\" Archie Cobbs <archie@FreeBSD.org> 35.\" 36.\" $FreeBSD: src/share/man/man4/netgraph.4,v 1.39.2.1 2001/12/21 09:00:50 ru Exp $ 37.\" $DragonFly: src/share/man/man4/netgraph.4,v 1.12 2007/11/23 23:16:37 swildner Exp $ 38.\" $Whistle: netgraph.4,v 1.7 1999/01/28 23:54:52 julian Exp $ 39.\" 40.Dd January 19, 1999 41.Dt NETGRAPH 4 42.Os 43.Sh NAME 44.Nm netgraph 45.Nd graph based kernel networking subsystem 46.Sh DESCRIPTION 47The 48.Nm 49system provides a uniform and modular system for the implementation 50of kernel objects which perform various networking functions. The objects, 51known as 52.Em nodes , 53can be arranged into arbitrarily complicated graphs. Nodes have 54.Em hooks 55which are used to connect two nodes together, forming the edges in the graph. 56Nodes communicate along the edges to process data, implement protocols, etc. 57.Pp 58The aim of 59.Nm 60is to supplement rather than replace the existing kernel networking 61infrastructure. It provides: 62.Pp 63.Bl -bullet -compact -offset 2n 64.It 65A flexible way of combining protocol and link level drivers 66.It 67A modular way to implement new protocols 68.It 69A common framework for kernel entities to inter-communicate 70.It 71A reasonably fast, kernel-based implementation 72.El 73.Sh Nodes and Types 74The most fundamental concept in 75.Nm 76is that of a 77.Em node . 78All nodes implement a number of predefined methods which allow them 79to interact with other nodes in a well defined manner. 80.Pp 81Each node has a 82.Em type , 83which is a static property of the node determined at node creation time. 84A node's type is described by a unique 85.Tn ASCII 86type name. 87The type implies what the node does and how it may be connected 88to other nodes. 89.Pp 90In object-oriented language, types are classes and nodes are instances 91of their respective class. All node types are subclasses of the generic node 92type, and hence inherit certain common functionality and capabilities 93(e.g., the ability to have an 94.Tn ASCII 95name). 96.Pp 97Nodes may be assigned a globally unique 98.Tn ASCII 99name which can be 100used to refer to the node. 101The name must not contain the characters 102.Dq .\& 103or 104.Dq \&: 105and is limited to 106.Dv "NG_NODESIZ" 107characters (including NUL byte). 108.Pp 109Each node instance has a unique 110.Em ID number 111which is expressed as a 32-bit hex value. This value may be used to 112refer to a node when there is no 113.Tn ASCII 114name assigned to it. 115.Sh Hooks 116Nodes are connected to other nodes by connecting a pair of 117.Em hooks , 118one from each node. Data flows bidirectionally between nodes along 119connected pairs of hooks. A node may have as many hooks as it 120needs, and may assign whatever meaning it wants to a hook. 121.Pp 122Hooks have these properties: 123.Pp 124.Bl -bullet -compact -offset 2n 125.It 126A hook has an 127.Tn ASCII 128name which is unique among all hooks 129on that node (other hooks on other nodes may have the same name). 130The name must not contain a 131.Dq .\& 132or a 133.Dq \&: 134and is 135limited to 136.Dv "NG_HOOKSIZ" 137characters (including NUL byte). 138.It 139A hook is always connected to another hook. That is, hooks are 140created at the time they are connected, and breaking an edge by 141removing either hook destroys both hooks. 142.El 143.Pp 144A node may decide to assign special meaning to some hooks. 145For example, connecting to the hook named 146.Dq debug 147might trigger 148the node to start sending debugging information to that hook. 149.Sh Data Flow 150Two types of information flow between nodes: data messages and 151control messages. Data messages are passed in mbuf chains along the edges 152in the graph, one edge at a time. The first mbuf in a chain must have the 153.Dv M_PKTHDR 154flag set. Each node decides how to handle data coming in on its hooks. 155.Pp 156Control messages are type-specific C structures sent from one node 157directly to some arbitrary other node. Control messages have a common 158header format, followed by type-specific data, and are binary structures 159for efficiency. However, node types also may support conversion of the 160type specific data between binary and 161.Tn ASCII 162for debugging and human interface purposes (see the 163.Dv NGM_ASCII2BINARY 164and 165.Dv NGM_BINARY2ASCII 166generic control messages below). Nodes are not required to support 167these conversions. 168.Pp 169There are two ways to address a control message. If 170there is a sequence of edges connecting the two nodes, the message 171may be 172.Dq source routed 173by specifying the corresponding sequence 174of hooks as the destination address for the message (relative 175addressing). Otherwise, the recipient node global 176.Tn ASCII 177name 178(or equivalent ID based name) is used as the destination address 179for the message (absolute addressing). The two types of addressing 180may be combined, by specifying an absolute start node and a sequence 181of hooks. 182.Pp 183Messages often represent commands that are followed by a reply message 184in the reverse direction. To facilitate this, the recipient of a 185control message is supplied with a 186.Dq return address 187that is suitable for addressing a reply. 188.Pp 189Each control message contains a 32 bit value called a 190.Em typecookie 191indicating the type of the message, i.e., how to interpret it. 192Typically each type defines a unique typecookie for the messages 193that it understands. However, a node may choose to recognize and 194implement more than one type of message. 195.Sh Netgraph is Functional 196In order to minimize latency, most 197.Nm 198operations are functional. 199That is, data and control messages are delivered by making function 200calls rather than by using queues and mailboxes. For example, if node 201A wishes to send a data mbuf to neighboring node B, it calls the 202generic 203.Nm 204data delivery function. This function in turn locates 205node B and calls B's 206.Dq receive data 207method. While this mode of operation 208results in good performance, it has a few implications for node 209developers: 210.Pp 211.Bl -bullet -compact -offset 2n 212.It 213Whenever a node delivers a data or control message, the node 214may need to allow for the possibility of receiving a returning 215message before the original delivery function call returns. 216.It 217Netgraph nodes and support routines generally run inside critical 218sections. 219However, some nodes may want to send data and control messages 220from a different priority level. Netgraph supplies queueing routines which 221utilize the NETISR system to move message delivery inside a critical 222section. 223Note that messages are always received from inside a critical section. 224.It 225It's possible for an infinite loop to occur if the graph contains cycles. 226.El 227.Pp 228So far, these issues have not proven problematical in practice. 229.Sh Interaction With Other Parts of the Kernel 230A node may have a hidden interaction with other components of the 231kernel outside of the 232.Nm 233subsystem, such as device hardware, 234kernel protocol stacks, etc. In fact, one of the benefits of 235.Nm 236is the ability to join disparate kernel networking entities together in a 237consistent communication framework. 238.Pp 239An example is the node type 240.Em socket 241which is both a netgraph node and a 242.Xr socket 2 243.Bx 244socket in the protocol family 245.Dv PF_NETGRAPH . 246Socket nodes allow user processes to participate in 247.Nm . 248Other nodes communicate with socket nodes using the usual methods, and the 249node hides the fact that it is also passing information to and from a 250cooperating user process. 251.Pp 252Another example is a device driver that presents 253a node interface to the hardware. 254.Sh Node Methods 255Nodes are notified of the following actions via function calls 256to the following node methods (all from inside critical sections) 257and may accept or reject that action (by returning the appropriate 258error code): 259.Bl -tag -width xxx 260.It Creation of a new node 261The constructor for the type is called. If creation of a new node is 262allowed, the constructor must call the generic node creation 263function (in object-oriented terms, the superclass constructor) 264and then allocate any special resources it needs. For nodes that 265correspond to hardware, this is typically done during the device 266attach routine. Often a global 267.Tn ASCII 268name corresponding to the 269device name is assigned here as well. 270.It Creation of a new hook 271The hook is created and tentatively 272linked to the node, and the node is told about the name that will be 273used to describe this hook. The node sets up any special data structures 274it needs, or may reject the connection, based on the name of the hook. 275.It Successful connection of two hooks 276After both ends have accepted their 277hooks, and the links have been made, the nodes get a chance to 278find out who their peer is across the link and can then decide to reject 279the connection. Tear-down is automatic. 280.It Destruction of a hook 281The node is notified of a broken connection. The node may consider some hooks 282to be critical to operation and others to be expendable: the disconnection 283of one hook may be an acceptable event while for another it 284may affect a total shutdown for the node. 285.It Shutdown of a node 286This method allows a node to clean up 287and to ensure that any actions that need to be performed 288at this time are taken. The method must call the generic (i.e., superclass) 289node destructor to get rid of the generic components of the node. 290Some nodes (usually associated with a piece of hardware) may be 291.Em persistent 292in that a shutdown breaks all edges and resets the node, 293but doesn't remove it, in which case the generic destructor is not called. 294.El 295.Sh Sending and Receiving Data 296Three other methods are also supported by all nodes: 297.Bl -tag -width xxx 298.It Receive data message 299An mbuf chain is passed to the node. 300The node is notified on which hook the data arrived, 301and can use this information in its processing decision. 302The node must must always 303.Fn m_freem 304the mbuf chain on completion or error, or pass it on to another node 305(or kernel module) which will then be responsible for freeing it. 306.Pp 307In addition to the mbuf chain itself there is also a pointer to a 308structure describing meta-data about the message 309(e.g. priority information). This pointer may be 310.Dv NULL 311if there is no additional information. The format for this information is 312described in 313.In netgraph/netgraph.h . 314The memory for meta-data must allocated via 315.Fn malloc 316with type 317.Dv M_NETGRAPH . 318As with the data itself, it is the receiver's responsibility to 319.Fn free 320the meta-data. If the mbuf chain is freed the meta-data must 321be freed at the same time. If the meta-data is freed but the 322real data on is passed on, then a 323.Dv NULL 324pointer must be substituted. 325.Pp 326The receiving node may decide to defer the data by queueing it in the 327.Nm 328NETISR system (see below). 329.Pp 330The structure and use of meta-data is still experimental, but is 331presently used in frame-relay to indicate that management packets 332should be queued for transmission 333at a higher priority than data packets. This is required for 334conformance with Frame Relay standards. 335.Pp 336.It Receive queued data message 337Usually this will be the same function as 338.Em Receive data message. 339This is the entry point called when a data message is being handed to 340the node after having been queued in the NETISR system. 341This allows a node to decide in the 342.Em Receive data message 343method that a message should be deferred and queued, 344and be sure that when it is processed from the queue, 345it will not be queued again. 346.It Receive control message 347This method is called when a control message is addressed to the node. 348A return address is always supplied, giving the address of the node 349that originated the message so a reply message can be sent anytime later. 350.Pp 351It is possible for a synchronous reply to be made, and in fact this 352is more common in practice. 353This is done by setting a pointer (supplied as an extra function parameter) 354to point to the reply. 355Then when the control message delivery function returns, 356the caller can check if this pointer has been made non-NULL, 357and if so then it points to the reply message allocated via 358.Fn malloc 359and containing the synchronous response. In both directions, 360(request and response) it is up to the 361receiver of that message to 362.Fn free 363the control message buffer. All control messages and replies are 364allocated with 365.Fn malloc 366type 367.Dv M_NETGRAPH . 368.El 369.Pp 370Much use has been made of reference counts, so that nodes being 371free'd of all references are automatically freed, and this behaviour 372has been tested and debugged to present a consistent and trustworthy 373framework for the 374.Dq type module 375writer to use. 376.Sh Addressing 377The 378.Nm 379framework provides an unambiguous and simple to use method of specifically 380addressing any single node in the graph. The naming of a node is 381independent of its type, in that another node, or external component 382need not know anything about the node's type in order to address it so as 383to send it a generic message type. Node and hook names should be 384chosen so as to make addresses meaningful. 385.Pp 386Addresses are either absolute or relative. An absolute address begins 387with a node name, (or ID), followed by a colon, followed by a sequence of hook 388names separated by periods. This addresses the node reached by starting 389at the named node and following the specified sequence of hooks. 390A relative address includes only the sequence of hook names, implicitly 391starting hook traversal at the local node. 392.Pp 393There are a couple of special possibilities for the node name. 394The name 395.Dq .\& 396(referred to as 397.Dq \&.: ) 398always refers to the local node. 399Also, nodes that have no global name may be addressed by their ID numbers, 400by enclosing the hex representation of the ID number within square brackets. 401Here are some examples of valid netgraph addresses: 402.Bd -literal -offset 4n -compact 403 404 .: 405 foo: 406 .:hook1 407 foo:hook1.hook2 408 [f057cd80]:hook1 409.Ed 410.Pp 411Consider the following set of nodes might be created for a site with 412a single physical frame relay line having two active logical DLCI channels, 413with RFC 1490 frames on DLCI 16 and PPP frames over DLCI 20: 414.Pp 415.Bd -literal 416[type SYNC ] [type FRAME] [type RFC1490] 417[ "Frame1" ](uplink)<-->(data)[<un-named>](dlci16)<-->(mux)[<un-named> ] 418[ A ] [ B ](dlci20)<---+ [ C ] 419 | 420 | [ type PPP ] 421 +>(mux)[<un-named>] 422 [ D ] 423.Ed 424.Pp 425One could always send a control message to node C from anywhere 426by using the name 427.Em "Frame1:uplink.dlci16" . 428Similarly, 429.Em "Frame1:uplink.dlci20" 430could reliably be used to reach node D, and node A could refer 431to node B as 432.Em ".:uplink" , 433or simply 434.Em "uplink" . 435Conversely, B can refer to A as 436.Em "data" . 437The address 438.Em "mux.data" 439could be used by both nodes C and D to address a message to node A. 440.Pp 441Note that this is only for 442.Em control messages . 443Data messages are routed one hop at a time, by specifying the departing 444hook, with each node making the next routing decision. So when B 445receives a frame on hook 446.Em data 447it decodes the frame relay header to determine the DLCI, 448and then forwards the unwrapped frame to either C or D. 449.Pp 450A similar graph might be used to represent multi-link PPP running 451over an ISDN line: 452.Pp 453.Bd -literal 454[ type BRI ](B1)<--->(link1)[ type MPP ] 455[ "ISDN1" ](B2)<--->(link2)[ (no name) ] 456[ ](D) <-+ 457 | 458 +----------------+ 459 | 460 +->(switch)[ type Q.921 ](term1)<---->(datalink)[ type Q.931 ] 461 [ (no name) ] [ (no name) ] 462.Ed 463.Sh Netgraph Structures 464Interesting members of the node and hook structures are shown below: 465.Bd -literal 466struct ng_node { 467 char *name; /* Optional globally unique name */ 468 void *private; /* Node implementation private info */ 469 struct ng_type *type; /* The type of this node */ 470 int refs; /* Number of references to this struct */ 471 int numhooks; /* Number of connected hooks */ 472 hook_p hooks; /* Linked list of (connected) hooks */ 473}; 474typedef struct ng_node *node_p; 475 476struct ng_hook { 477 char *name; /* This node's name for this hook */ 478 void *private; /* Node implementation private info */ 479 int refs; /* Number of references to this struct */ 480 struct ng_node *node; /* The node this hook is attached to */ 481 struct ng_hook *peer; /* The other hook in this connected pair */ 482 struct ng_hook *next; /* Next in list of hooks for this node */ 483}; 484typedef struct ng_hook *hook_p; 485.Ed 486.Pp 487The maintenance of the name pointers, reference counts, and linked list 488of hooks for each node is handled automatically by the 489.Nm 490subsystem. 491Typically a node's private info contains a back-pointer to the node or hook 492structure, which counts as a new reference that must be registered by 493incrementing 494.Dv "node->refs" . 495.Pp 496From a hook you can obtain the corresponding node, and from 497a node the list of all active hooks. 498.Pp 499Node types are described by these structures: 500.Bd -literal 501/** How to convert a control message from binary <-> ASCII */ 502struct ng_cmdlist { 503 u_int32_t cookie; /* typecookie */ 504 int cmd; /* command number */ 505 const char *name; /* command name */ 506 const struct ng_parse_type *mesgType; /* args if !NGF_RESP */ 507 const struct ng_parse_type *respType; /* args if NGF_RESP */ 508}; 509 510struct ng_type { 511 u_int32_t version; /* Must equal NG_VERSION */ 512 const char *name; /* Unique type name */ 513 514 /* Module event handler */ 515 modeventhand_t mod_event; /* Handle load/unload (optional) */ 516 517 /* Constructor */ 518 int (*constructor)(node_p *node); /* Create a new node */ 519 520 /** Methods using the node **/ 521 int (*rcvmsg)(node_p node, /* Receive control message */ 522 struct ng_mesg *msg, /* The message */ 523 const char *retaddr, /* Return address */ 524 struct ng_mesg **resp); /* Synchronous response */ 525 int (*shutdown)(node_p node); /* Shutdown this node */ 526 int (*newhook)(node_p node, /* create a new hook */ 527 hook_p hook, /* Pre-allocated struct */ 528 const char *name); /* Name for new hook */ 529 530 /** Methods using the hook **/ 531 int (*connect)(hook_p hook); /* Confirm new hook attachment */ 532 int (*rcvdata)(hook_p hook, /* Receive data on a hook */ 533 struct mbuf *m, /* The data in an mbuf */ 534 meta_p meta); /* Meta-data, if any */ 535 int (*disconnect)(hook_p hook); /* Notify disconnection of hook */ 536 537 /** How to convert control messages binary <-> ASCII */ 538 const struct ng_cmdlist *cmdlist; /* Optional; may be NULL */ 539}; 540.Ed 541.Pp 542Control messages have the following structure: 543.Bd -literal 544#define NG_CMDSTRSIZ 16 /* Max command string (including null) */ 545 546struct ng_mesg { 547 struct ng_msghdr { 548 u_char version; /* Must equal NG_VERSION */ 549 u_char spare; /* Pad to 2 bytes */ 550 u_short arglen; /* Length of cmd/resp data */ 551 u_long flags; /* Message status flags */ 552 u_long token; /* Reply should have the same token */ 553 u_long typecookie; /* Node type understanding this message */ 554 u_long cmd; /* Command identifier */ 555 u_char cmdstr[NG_CMDSTRSIZ]; /* Cmd string (for debug) */ 556 } header; 557 char data[0]; /* Start of cmd/resp data */ 558}; 559 560#define NG_VERSION 1 /* Netgraph version */ 561#define NGF_ORIG 0x0000 /* Command */ 562#define NGF_RESP 0x0001 /* Response */ 563.Ed 564.Pp 565Control messages have the fixed header shown above, followed by a 566variable length data section which depends on the type cookie 567and the command. Each field is explained below: 568.Bl -tag -width xxx 569.It Dv version 570Indicates the version of netgraph itself. The current version is 571.Dv NG_VERSION . 572.It Dv arglen 573This is the length of any extra arguments, which begin at 574.Dv data . 575.It Dv flags 576Indicates whether this is a command or a response control message. 577.It Dv token 578The 579.Dv token 580is a means by which a sender can match a reply message to the 581corresponding command message; the reply always has the same token. 582.Pp 583.It Dv typecookie 584The corresponding node type's unique 32-bit value. 585If a node doesn't recognize the type cookie it must reject the message 586by returning 587.Er EINVAL . 588.Pp 589Each type should have an include file that defines the commands, 590argument format, and cookie for its own messages. 591The typecookie 592insures that the same header file was included by both sender and 593receiver; when an incompatible change in the header file is made, 594the typecookie 595.Em must 596be changed. 597The de facto method for generating unique type cookies is to take the 598seconds from the epoch at the time the header file is written 599(i.e., the output of 600.Dv "date -u +'%s'" ) . 601.Pp 602There is a predefined typecookie 603.Dv NGM_GENERIC_COOKIE 604for the 605.Dq generic 606node type, and 607a corresponding set of generic messages which all nodes understand. 608The handling of these messages is automatic. 609.It Dv command 610The identifier for the message command. This is type specific, 611and is defined in the same header file as the typecookie. 612.It Dv cmdstr 613Room for a short human readable version of 614.Dq command 615(for debugging purposes only). 616.El 617.Pp 618Some modules may choose to implement messages from more than one 619of the header files and thus recognize more than one type cookie. 620.Sh Control Message ASCII Form 621Control messages are in binary format for efficiency. However, for 622debugging and human interface purposes, and if the node type supports 623it, control messages may be converted to and from an equivalent 624.Tn ASCII 625form. The 626.Tn ASCII 627form is similar to the binary form, with two exceptions: 628.Pp 629.Bl -tag -compact -width xxx 630.It o 631The 632.Dv cmdstr 633header field must contain the 634.Tn ASCII 635name of the command, corresponding to the 636.Dv cmd 637header field. 638.It o 639The 640.Dv args 641field contains a NUL-terminated 642.Tn ASCII 643string version of the message arguments. 644.El 645.Pp 646In general, the arguments field of a control message can be any 647arbitrary C data type. Netgraph includes parsing routines to support 648some pre-defined datatypes in 649.Tn ASCII 650with this simple syntax: 651.Pp 652.Bl -tag -compact -width xxx 653.It o 654Integer types are represented by base 8, 10, or 16 numbers. 655.It o 656Strings are enclosed in double quotes and respect the normal 657C language backslash escapes. 658.It o 659IP addresses have the obvious form. 660.It o 661Arrays are enclosed in square brackets, with the elements listed 662consecutively starting at index zero. An element may have an optional 663index and equals sign preceding it. Whenever an element 664does not have an explicit index, the index is implicitly the previous 665element's index plus one. 666.It o 667Structures are enclosed in curly braces, and each field is specified 668in the form 669.Dq fieldname=value . 670.It o 671Any array element or structure field whose value is equal to its 672.Dq default value 673may be omitted. For integer types, the default value 674is usually zero; for string types, the empty string. 675.It o 676Array elements and structure fields may be specified in any order. 677.El 678.Pp 679Each node type may define its own arbitrary types by providing 680the necessary routines to parse and unparse. 681.Tn ASCII 682forms defined 683for a specific node type are documented in the documentation for 684that node type. 685.Sh Generic Control Messages 686There are a number of standard predefined messages that will work 687for any node, as they are supported directly by the framework itself. 688These are defined in 689.In netgraph/ng_message.h 690along with the basic layout of messages and other similar information. 691.Bl -tag -width xxx 692.It Dv NGM_CONNECT 693Connect to another node, using the supplied hook names on either end. 694.It Dv NGM_MKPEER 695Construct a node of the given type and then connect to it using the 696supplied hook names. 697.It Dv NGM_SHUTDOWN 698The target node should disconnect from all its neighbours and shut down. 699Persistent nodes such as those representing physical hardware 700might not disappear from the node namespace, but only reset themselves. 701The node must disconnect all of its hooks. 702This may result in neighbors shutting themselves down, and possibly a 703cascading shutdown of the entire connected graph. 704.It Dv NGM_NAME 705Assign a name to a node. Nodes can exist without having a name, and this 706is the default for nodes created using the 707.Dv NGM_MKPEER 708method. Such nodes can only be addressed relatively or by their ID number. 709.It Dv NGM_RMHOOK 710Ask the node to break a hook connection to one of its neighbours. 711Both nodes will have their 712.Dq disconnect 713method invoked. 714Either node may elect to totally shut down as a result. 715.It Dv NGM_NODEINFO 716Asks the target node to describe itself. The four returned fields 717are the node name (if named), the node type, the node ID and the 718number of hooks attached. The ID is an internal number unique to that node. 719.It Dv NGM_LISTHOOKS 720This returns the information given by 721.Dv NGM_NODEINFO , 722but in addition 723includes an array of fields describing each link, and the description for 724the node at the far end of that link. 725.It Dv NGM_LISTNAMES 726This returns an array of node descriptions (as for 727.Dv NGM_NODEINFO ")" 728where each entry of the array describes a named node. 729All named nodes will be described. 730.It Dv NGM_LISTNODES 731This is the same as 732.Dv NGM_LISTNAMES 733except that all nodes are listed regardless of whether they have a name or not. 734.It Dv NGM_LISTTYPES 735This returns a list of all currently installed netgraph types. 736.It Dv NGM_TEXT_STATUS 737The node may return a text formatted status message. 738The status information is determined entirely by the node type. 739It is the only "generic" message 740that requires any support within the node itself and as such the node may 741elect to not support this message. The text response must be less than 742.Dv NG_TEXTRESPONSE 743bytes in length (presently 1024). This can be used to return general 744status information in human readable form. 745.It Dv NGM_BINARY2ASCII 746This message converts a binary control message to its 747.Tn ASCII 748form. 749The entire control message to be converted is contained within the 750arguments field of the 751.Dv NGM_BINARY2ASCII 752message itself. If successful, the reply will contain the same control 753message in 754.Tn ASCII 755form. 756A node will typically only know how to translate messages that it 757itself understands, so the target node of the 758.Dv NGM_BINARY2ASCII 759is often the same node that would actually receive that message. 760.It Dv NGM_ASCII2BINARY 761The opposite of 762.Dv NGM_BINARY2ASCII . 763The entire control message to be converted, in 764.Tn ASCII 765form, is contained 766in the arguments section of the 767.Dv NGM_ASCII2BINARY 768and need only have the 769.Dv flags , 770.Dv cmdstr , 771and 772.Dv arglen 773header fields filled in, plus the NUL-terminated string version of 774the arguments in the arguments field. If successful, the reply 775contains the binary version of the control message. 776.El 777.Sh Metadata 778Data moving through the 779.Nm 780system can be accompanied by meta-data that describes some 781aspect of that data. The form of the meta-data is a fixed header, 782which contains enough information for most uses, and can optionally 783be supplemented by trailing 784.Em option 785structures, which contain a 786.Em cookie 787(see the section on control messages), an identifier, a length and optional 788data. If a node does not recognize the cookie associated with an option, 789it should ignore that option. 790.Pp 791Meta data might include such things as priority, discard eligibility, 792or special processing requirements. It might also mark a packet for 793debug status, etc. The use of meta-data is still experimental. 794.Sh INITIALIZATION 795The base 796.Nm 797code may either be statically compiled 798into the kernel or else loaded dynamically as a KLD via 799.Xr kldload 8 . 800In the former case, include 801.Pp 802.D1 Cd options NETGRAPH 803.Pp 804in your kernel configuration file. You may also include selected 805node types in the kernel compilation, for example: 806.Bd -unfilled -offset indent 807.Cd options NETGRAPH 808.Cd options NETGRAPH_SOCKET 809.Cd options NETGRAPH_ECHO 810.Ed 811.Pp 812Once the 813.Nm 814subsystem is loaded, individual node types may be loaded at any time 815as KLD modules via 816.Xr kldload 8 . 817Moreover, 818.Nm 819knows how to automatically do this; when a request to create a new 820node of unknown type 821.Em type 822is made, 823.Nm 824will attempt to load the KLD module 825.Pa ng_type.ko . 826.Pp 827Types can also be installed at boot time, as certain device drivers 828may want to export each instance of the device as a netgraph node. 829.Pp 830In general, new types can be installed at any time from within the 831kernel by calling 832.Fn ng_newtype , 833supplying a pointer to the type's 834.Dv struct ng_type 835structure. 836.Pp 837The 838.Fn NETGRAPH_INIT 839macro automates this process by using a linker set. 840.Sh EXISTING NODE TYPES 841Several node types currently exist. Each is fully documented 842in its own man page: 843.Bl -tag -width xxx 844.It SOCKET 845The socket type implements two new sockets in the new protocol domain 846.Dv PF_NETGRAPH . 847The new sockets protocols are 848.Dv NG_DATA 849and 850.Dv NG_CONTROL , 851both of type 852.Dv SOCK_DGRAM . 853Typically one of each is associated with a socket node. 854When both sockets have closed, the node will shut down. The 855.Dv NG_DATA 856socket is used for sending and receiving data, while the 857.Dv NG_CONTROL 858socket is used for sending and receiving control messages. 859Data and control messages are passed using the 860.Xr sendto 2 861and 862.Xr recvfrom 2 863calls, using a 864.Dv struct sockaddr_ng 865socket address. 866.Pp 867.It HOLE 868Responds only to generic messages and is a 869.Dq black hole 870for data, Useful for testing. Always accepts new hooks. 871.Pp 872.It ECHO 873Responds only to generic messages and always echoes data back through the 874hook from which it arrived. Returns any non generic messages as their 875own response. Useful for testing. Always accepts new hooks. 876.Pp 877.It TEE 878This node is useful for 879.Dq snooping . 880It has 4 hooks: 881.Dv left , 882.Dv right , 883.Dv left2right , 884and 885.Dv right2left . 886Data entering from the right is passed to the left and duplicated on 887.Dv right2left , 888and data entering from the left is passed to the right and 889duplicated on 890.Dv left2right . 891Data entering from 892.Dv left2right 893is sent to the right and data from 894.Dv right2left 895to left. 896.Pp 897.It RFC1490 MUX 898Encapsulates/de-encapsulates frames encoded according to RFC 1490. 899Has a hook for the encapsulated packets 900.Pq Dq downstream 901and one hook 902for each protocol (i.e., IP, PPP, etc.). 903.Pp 904.It FRAME RELAY MUX 905Encapsulates/de-encapsulates Frame Relay frames. 906Has a hook for the encapsulated packets 907.Pq Dq downstream 908and one hook 909for each DLCI. 910.Pp 911.It FRAME RELAY LMI 912Automatically handles frame relay 913.Dq LMI 914(link management interface) operations and packets. 915Automatically probes and detects which of several LMI standards 916is in use at the exchange. 917.Pp 918.It TTY 919This node is also a line discipline. It simply converts between mbuf 920frames and sequential serial data, allowing a tty to appear as a netgraph 921node. It has a programmable 922.Dq hotkey 923character. 924.Pp 925.It ASYNC 926This node encapsulates and de-encapsulates asynchronous frames 927according to RFC 1662. This is used in conjunction with the TTY node 928type for supporting PPP links over asynchronous serial lines. 929.Pp 930.It INTERFACE 931This node is also a system networking interface. It has hooks representing 932each protocol family (IP, AppleTalk, IPX, etc.) and appears in the output of 933.Xr ifconfig 8 . 934The interfaces are named 935.Em ng0 , 936.Em ng1 , 937etc. 938.El 939.Sh NOTES 940Whether a named node exists can be checked by trying to send a control message 941to it (e.g., 942.Dv NGM_NODEINFO ) . 943If it does not exist, 944.Er ENOENT 945will be returned. 946.Pp 947All data messages are mbuf chains with the M_PKTHDR flag set. 948.Pp 949Nodes are responsible for freeing what they allocate. 950There are three exceptions: 951.Bl -tag -width xxxx 952.It 1 953Mbufs sent across a data link are never to be freed by the sender. 954.It 2 955Any meta-data information traveling with the data has the same restriction. 956It might be freed by any node the data passes through, and a 957.Dv NULL 958passed onwards, but the caller will never free it. 959Two macros 960.Fn NG_FREE_META "meta" 961and 962.Fn NG_FREE_DATA "m" "meta" 963should be used if possible to free data and meta data (see 964.In netgraph/netgraph.h ) . 965.It 3 966Messages sent using 967.Fn ng_send_msg 968are freed by the callee. As in the case above, the addresses 969associated with the message are freed by whatever allocated them so the 970recipient should copy them if it wants to keep that information. 971.El 972.Sh FILES 973.Bl -tag -width xxxxx -compact 974.It In netgraph/netgraph.h 975Definitions for use solely within the kernel by 976.Nm 977nodes. 978.It In netgraph/ng_message.h 979Definitions needed by any file that needs to deal with 980.Nm 981messages. 982.It In netgraph/socket/ng_socket.h 983Definitions needed to use 984.Nm 985socket type nodes. 986.It In netgraph/{type}/ng_{type}.h 987Definitions needed to use 988.Nm 989{type} 990nodes, including the type cookie definition. 991.It Pa /modules/netgraph.ko 992Netgraph subsystem loadable KLD module. 993.It Pa /modules/ng_{type}.ko 994Loadable KLD module for node type {type}. 995.El 996.Sh USER MODE SUPPORT 997There is a library for supporting user-mode programs that wish 998to interact with the netgraph system. See 999.Xr netgraph 3 1000for details. 1001.Pp 1002Two user-mode support programs, 1003.Xr ngctl 8 1004and 1005.Xr nghook 8 , 1006are available to assist manual configuration and debugging. 1007.Pp 1008There are a few useful techniques for debugging new node types. 1009First, implementing new node types in user-mode first 1010makes debugging easier. 1011The 1012.Em tee 1013node type is also useful for debugging, especially in conjunction with 1014.Xr ngctl 8 1015and 1016.Xr nghook 8 . 1017.Sh SEE ALSO 1018.Xr socket 2 , 1019.Xr netgraph 3 , 1020.Xr ng_async 4 , 1021.Xr ng_bpf 4 , 1022.Xr ng_bridge 4 , 1023.Xr ng_cisco 4 , 1024.Xr ng_echo 4 , 1025.Xr ng_eiface 4 , 1026.Xr ng_etf 4 , 1027.Xr ng_ether 4 , 1028.Xr ng_frame_relay 4 , 1029.Xr ng_hole 4 , 1030.Xr ng_iface 4 , 1031.Xr ng_ksocket 4 , 1032.Xr ng_l2tp 4 , 1033.Xr ng_lmi 4 , 1034.Xr ng_mppc 4 , 1035.Xr ng_one2many 4 , 1036.Xr ng_ppp 4 , 1037.Xr ng_pppoe 4 , 1038.Xr ng_rfc1490 4 , 1039.Xr ng_socket 4 , 1040.Xr ng_tee 4 , 1041.Xr ng_tty 4 , 1042.Xr ng_UI 4 , 1043.Xr ng_vjc 4 , 1044.Xr ngctl 8 , 1045.Xr nghook 8 1046.Sh HISTORY 1047The 1048.Nm 1049system was designed and first implemented at Whistle Communications, Inc.\& 1050in a version of 1051.Fx 2.2 1052customized for the Whistle InterJet. 1053It first made its debut in the main tree in 1054.Fx 3.4 . 1055.Sh AUTHORS 1056.An -nosplit 1057.An Julian Elischer Aq julian@FreeBSD.org , 1058with contributions by 1059.An Archie Cobbs Aq archie@FreeBSD.org . 1060