1.. SPDX-License-Identifier: GPL-2.0
2
3==============================================
4Management Component Transport Protocol (MCTP)
5==============================================
6
7net/mctp/ contains protocol support for MCTP, as defined by DMTF standard
8DSP0236. Physical interface drivers ("bindings" in the specification) are
9provided in drivers/net/mctp/.
10
11The core code provides a socket-based interface to send and receive MCTP
12messages, through an AF_MCTP, SOCK_DGRAM socket.
13
14Structure: interfaces & networks
15================================
16
17The kernel models the local MCTP topology through two items: interfaces and
18networks.
19
20An interface (or "link") is an instance of an MCTP physical transport binding
21(as defined by DSP0236, section 3.2.47), likely connected to a specific hardware
22device. This is represented as a ``struct netdevice``.
23
24A network defines a unique address space for MCTP endpoints by endpoint-ID
25(described by DSP0236, section 3.2.31). A network has a user-visible identifier
26to allow references from userspace. Route definitions are specific to one
27network.
28
29Interfaces are associated with one network. A network may be associated with one
30or more interfaces.
31
32If multiple networks are present, each may contain endpoint IDs (EIDs) that are
33also present on other networks.
34
35Sockets API
36===========
37
38Protocol definitions
39--------------------
40
41MCTP uses ``AF_MCTP`` / ``PF_MCTP`` for the address- and protocol- families.
42Since MCTP is message-based, only ``SOCK_DGRAM`` sockets are supported.
43
44.. code-block:: C
45
46    int sd = socket(AF_MCTP, SOCK_DGRAM, 0);
47
48The only (current) value for the ``protocol`` argument is 0.
49
50As with all socket address families, source and destination addresses are
51specified with a ``sockaddr`` type, with a single-byte endpoint address:
52
53.. code-block:: C
54
55    typedef __u8		mctp_eid_t;
56
57    struct mctp_addr {
58            mctp_eid_t		s_addr;
59    };
60
61    struct sockaddr_mctp {
62            __kernel_sa_family_t smctp_family;
63            unsigned int         smctp_network;
64            struct mctp_addr     smctp_addr;
65            __u8                 smctp_type;
66            __u8                 smctp_tag;
67    };
68
69    #define MCTP_NET_ANY	0x0
70    #define MCTP_ADDR_ANY	0xff
71
72
73Syscall behaviour
74-----------------
75
76The following sections describe the MCTP-specific behaviours of the standard
77socket system calls. These behaviours have been chosen to map closely to the
78existing sockets APIs.
79
80``bind()`` : set local socket address
81^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
82
83Sockets that receive incoming request packets will bind to a local address,
84using the ``bind()`` syscall.
85
86.. code-block:: C
87
88    struct sockaddr_mctp addr;
89
90    addr.smctp_family = AF_MCTP;
91    addr.smctp_network = MCTP_NET_ANY;
92    addr.smctp_addr.s_addr = MCTP_ADDR_ANY;
93    addr.smctp_type = MCTP_TYPE_PLDM;
94    addr.smctp_tag = MCTP_TAG_OWNER;
95
96    int rc = bind(sd, (struct sockaddr *)&addr, sizeof(addr));
97
98This establishes the local address of the socket. Incoming MCTP messages that
99match the network, address, and message type will be received by this socket.
100The reference to 'incoming' is important here; a bound socket will only receive
101messages with the TO bit set, to indicate an incoming request message, rather
102than a response.
103
104The ``smctp_tag`` value will configure the tags accepted from the remote side of
105this socket. Given the above, the only valid value is ``MCTP_TAG_OWNER``, which
106will result in remotely "owned" tags being routed to this socket. Since
107``MCTP_TAG_OWNER`` is set, the 3 least-significant bits of ``smctp_tag`` are not
108used; callers must set them to zero.
109
110A ``smctp_network`` value of ``MCTP_NET_ANY`` will configure the socket to
111receive incoming packets from any locally-connected network. A specific network
112value will cause the socket to only receive incoming messages from that network.
113
114The ``smctp_addr`` field specifies a local address to bind to. A value of
115``MCTP_ADDR_ANY`` configures the socket to receive messages addressed to any
116local destination EID.
117
118The ``smctp_type`` field specifies which message types to receive. Only the
119lower 7 bits of the type is matched on incoming messages (ie., the
120most-significant IC bit is not part of the match). This results in the socket
121receiving packets with and without a message integrity check footer.
122
123``sendto()``, ``sendmsg()``, ``send()`` : transmit an MCTP message
124^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
125
126An MCTP message is transmitted using one of the ``sendto()``, ``sendmsg()`` or
127``send()`` syscalls. Using ``sendto()`` as the primary example:
128
129.. code-block:: C
130
131    struct sockaddr_mctp addr;
132    char buf[14];
133    ssize_t len;
134
135    /* set message destination */
136    addr.smctp_family = AF_MCTP;
137    addr.smctp_network = 0;
138    addr.smctp_addr.s_addr = 8;
139    addr.smctp_tag = MCTP_TAG_OWNER;
140    addr.smctp_type = MCTP_TYPE_ECHO;
141
142    /* arbitrary message to send, with message-type header */
143    buf[0] = MCTP_TYPE_ECHO;
144    memcpy(buf + 1, "hello, world!", sizeof(buf) - 1);
145
146    len = sendto(sd, buf, sizeof(buf), 0,
147                    (struct sockaddr_mctp *)&addr, sizeof(addr));
148
149The network and address fields of ``addr`` define the remote address to send to.
150If ``smctp_tag`` has the ``MCTP_TAG_OWNER``, the kernel will ignore any bits set
151in ``MCTP_TAG_VALUE``, and generate a tag value suitable for the destination
152EID. If ``MCTP_TAG_OWNER`` is not set, the message will be sent with the tag
153value as specified. If a tag value cannot be allocated, the system call will
154report an errno of ``EAGAIN``.
155
156The application must provide the message type byte as the first byte of the
157message buffer passed to ``sendto()``. If a message integrity check is to be
158included in the transmitted message, it must also be provided in the message
159buffer, and the most-significant bit of the message type byte must be 1.
160
161The ``sendmsg()`` system call allows a more compact argument interface, and the
162message buffer to be specified as a scatter-gather list. At present no ancillary
163message types (used for the ``msg_control`` data passed to ``sendmsg()``) are
164defined.
165
166Transmitting a message on an unconnected socket with ``MCTP_TAG_OWNER``
167specified will cause an allocation of a tag, if no valid tag is already
168allocated for that destination. The (destination-eid,tag) tuple acts as an
169implicit local socket address, to allow the socket to receive responses to this
170outgoing message. If any previous allocation has been performed (to for a
171different remote EID), that allocation is lost.
172
173Sockets will only receive responses to requests they have sent (with TO=1) and
174may only respond (with TO=0) to requests they have received.
175
176``recvfrom()``, ``recvmsg()``, ``recv()`` : receive an MCTP message
177^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
178
179An MCTP message can be received by an application using one of the
180``recvfrom()``, ``recvmsg()``, or ``recv()`` system calls. Using ``recvfrom()``
181as the primary example:
182
183.. code-block:: C
184
185    struct sockaddr_mctp addr;
186    socklen_t addrlen;
187    char buf[14];
188    ssize_t len;
189
190    addrlen = sizeof(addr);
191
192    len = recvfrom(sd, buf, sizeof(buf), 0,
193                    (struct sockaddr_mctp *)&addr, &addrlen);
194
195    /* We can expect addr to describe an MCTP address */
196    assert(addrlen >= sizeof(buf));
197    assert(addr.smctp_family == AF_MCTP);
198
199    printf("received %zd bytes from remote EID %d\n", rc, addr.smctp_addr);
200
201The address argument to ``recvfrom`` and ``recvmsg`` is populated with the
202remote address of the incoming message, including tag value (this will be needed
203in order to reply to the message).
204
205The first byte of the message buffer will contain the message type byte. If an
206integrity check follows the message, it will be included in the received buffer.
207
208The ``recv()`` system call behaves in a similar way, but does not provide a
209remote address to the application. Therefore, these are only useful if the
210remote address is already known, or the message does not require a reply.
211
212Like the send calls, sockets will only receive responses to requests they have
213sent (TO=1) and may only respond (TO=0) to requests they have received.
214
215``ioctl(SIOCMCTPALLOCTAG)`` and ``ioctl(SIOCMCTPDROPTAG)``
216^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
217
218These tags give applications more control over MCTP message tags, by allocating
219(and dropping) tag values explicitly, rather than the kernel automatically
220allocating a per-message tag at ``sendmsg()`` time.
221
222In general, you will only need to use these ioctls if your MCTP protocol does
223not fit the usual request/response model. For example, if you need to persist
224tags across multiple requests, or a request may generate more than one response.
225In these cases, the ioctls allow you to decouple the tag allocation (and
226release) from individual message send and receive operations.
227
228Both ioctls are passed a pointer to a ``struct mctp_ioc_tag_ctl``:
229
230.. code-block:: C
231
232    struct mctp_ioc_tag_ctl {
233        mctp_eid_t      peer_addr;
234        __u8		tag;
235        __u16   	flags;
236    };
237
238``SIOCMCTPALLOCTAG`` allocates a tag for a specific peer, which an application
239can use in future ``sendmsg()`` calls. The application populates the
240``peer_addr`` member with the remote EID. Other fields must be zero.
241
242On return, the ``tag`` member will be populated with the allocated tag value.
243The allocated tag will have the following tag bits set:
244
245 - ``MCTP_TAG_OWNER``: it only makes sense to allocate tags if you're the tag
246   owner
247
248 - ``MCTP_TAG_PREALLOC``: to indicate to ``sendmsg()`` that this is a
249   preallocated tag.
250
251 - ... and the actual tag value, within the least-significant three bits
252   (``MCTP_TAG_MASK``). Note that zero is a valid tag value.
253
254The tag value should be used as-is for the ``smctp_tag`` member of ``struct
255sockaddr_mctp``.
256
257``SIOCMCTPDROPTAG`` releases a tag that has been previously allocated by a
258``SIOCMCTPALLOCTAG`` ioctl. The ``peer_addr`` must be the same as used for the
259allocation, and the ``tag`` value must match exactly the tag returned from the
260allocation (including the ``MCTP_TAG_OWNER`` and ``MCTP_TAG_PREALLOC`` bits).
261The ``flags`` field must be zero.
262
263Kernel internals
264================
265
266There are a few possible packet flows in the MCTP stack:
267
2681. local TX to remote endpoint, message <= MTU::
269
270	sendmsg()
271	 -> mctp_local_output()
272	    : route lookup
273	    -> rt->output() (== mctp_route_output)
274	       -> dev_queue_xmit()
275
2762. local TX to remote endpoint, message > MTU::
277
278	sendmsg()
279	-> mctp_local_output()
280	    -> mctp_do_fragment_route()
281	       : creates packet-sized skbs. For each new skb:
282	       -> rt->output() (== mctp_route_output)
283	          -> dev_queue_xmit()
284
2853. remote TX to local endpoint, single-packet message::
286
287	mctp_pkttype_receive()
288	: route lookup
289	-> rt->output() (== mctp_route_input)
290	   : sk_key lookup
291	   -> sock_queue_rcv_skb()
292
2934. remote TX to local endpoint, multiple-packet message::
294
295	mctp_pkttype_receive()
296	: route lookup
297	-> rt->output() (== mctp_route_input)
298	   : sk_key lookup
299	   : stores skb in struct sk_key->reasm_head
300
301	mctp_pkttype_receive()
302	: route lookup
303	-> rt->output() (== mctp_route_input)
304	   : sk_key lookup
305	   : finds existing reassembly in sk_key->reasm_head
306	   : appends new fragment
307	   -> sock_queue_rcv_skb()
308
309Key refcounts
310-------------
311
312 * keys are refed by:
313
314   - a skb: during route output, stored in ``skb->cb``.
315
316   - netns and sock lists.
317
318 * keys can be associated with a device, in which case they hold a
319   reference to the dev (set through ``key->dev``, counted through
320   ``dev->key_count``). Multiple keys can reference the device.
321