cachepc-linux

Fork of AMDESE/linux with modifications for CachePC side-channel attack
git clone https://git.sinitax.com/sinitax/cachepc-linux
Log | Files | Refs | README | LICENSE | sfeed.txt

dccp.rst (9830B)


      1.. SPDX-License-Identifier: GPL-2.0
      2
      3=============
      4DCCP protocol
      5=============
      6
      7
      8.. Contents
      9   - Introduction
     10   - Missing features
     11   - Socket options
     12   - Sysctl variables
     13   - IOCTLs
     14   - Other tunables
     15   - Notes
     16
     17
     18Introduction
     19============
     20Datagram Congestion Control Protocol (DCCP) is an unreliable, connection
     21oriented protocol designed to solve issues present in UDP and TCP, particularly
     22for real-time and multimedia (streaming) traffic.
     23It divides into a base protocol (RFC 4340) and pluggable congestion control
     24modules called CCIDs. Like pluggable TCP congestion control, at least one CCID
     25needs to be enabled in order for the protocol to function properly. In the Linux
     26implementation, this is the TCP-like CCID2 (RFC 4341). Additional CCIDs, such as
     27the TCP-friendly CCID3 (RFC 4342), are optional.
     28For a brief introduction to CCIDs and suggestions for choosing a CCID to match
     29given applications, see section 10 of RFC 4340.
     30
     31It has a base protocol and pluggable congestion control IDs (CCIDs).
     32
     33DCCP is a Proposed Standard (RFC 2026), and the homepage for DCCP as a protocol
     34is at http://www.ietf.org/html.charters/dccp-charter.html
     35
     36
     37Missing features
     38================
     39The Linux DCCP implementation does not currently support all the features that are
     40specified in RFCs 4340...42.
     41
     42The known bugs are at:
     43
     44	http://www.linuxfoundation.org/collaborate/workgroups/networking/todo#DCCP
     45
     46For more up-to-date versions of the DCCP implementation, please consider using
     47the experimental DCCP test tree; instructions for checking this out are on:
     48http://www.linuxfoundation.org/collaborate/workgroups/networking/dccp_testing#Experimental_DCCP_source_tree
     49
     50
     51Socket options
     52==============
     53DCCP_SOCKOPT_QPOLICY_ID sets the dequeuing policy for outgoing packets. It takes
     54a policy ID as argument and can only be set before the connection (i.e. changes
     55during an established connection are not supported). Currently, two policies are
     56defined: the "simple" policy (DCCPQ_POLICY_SIMPLE), which does nothing special,
     57and a priority-based variant (DCCPQ_POLICY_PRIO). The latter allows to pass an
     58u32 priority value as ancillary data to sendmsg(), where higher numbers indicate
     59a higher packet priority (similar to SO_PRIORITY). This ancillary data needs to
     60be formatted using a cmsg(3) message header filled in as follows::
     61
     62	cmsg->cmsg_level = SOL_DCCP;
     63	cmsg->cmsg_type	 = DCCP_SCM_PRIORITY;
     64	cmsg->cmsg_len	 = CMSG_LEN(sizeof(uint32_t));	/* or CMSG_LEN(4) */
     65
     66DCCP_SOCKOPT_QPOLICY_TXQLEN sets the maximum length of the output queue. A zero
     67value is always interpreted as unbounded queue length. If different from zero,
     68the interpretation of this parameter depends on the current dequeuing policy
     69(see above): the "simple" policy will enforce a fixed queue size by returning
     70EAGAIN, whereas the "prio" policy enforces a fixed queue length by dropping the
     71lowest-priority packet first. The default value for this parameter is
     72initialised from /proc/sys/net/dccp/default/tx_qlen.
     73
     74DCCP_SOCKOPT_SERVICE sets the service. The specification mandates use of
     75service codes (RFC 4340, sec. 8.1.2); if this socket option is not set,
     76the socket will fall back to 0 (which means that no meaningful service code
     77is present). On active sockets this is set before connect(); specifying more
     78than one code has no effect (all subsequent service codes are ignored). The
     79case is different for passive sockets, where multiple service codes (up to 32)
     80can be set before calling bind().
     81
     82DCCP_SOCKOPT_GET_CUR_MPS is read-only and retrieves the current maximum packet
     83size (application payload size) in bytes, see RFC 4340, section 14.
     84
     85DCCP_SOCKOPT_AVAILABLE_CCIDS is also read-only and returns the list of CCIDs
     86supported by the endpoint. The option value is an array of type uint8_t whose
     87size is passed as option length. The minimum array size is 4 elements, the
     88value returned in the optlen argument always reflects the true number of
     89built-in CCIDs.
     90
     91DCCP_SOCKOPT_CCID is write-only and sets both the TX and RX CCIDs at the same
     92time, combining the operation of the next two socket options. This option is
     93preferable over the latter two, since often applications will use the same
     94type of CCID for both directions; and mixed use of CCIDs is not currently well
     95understood. This socket option takes as argument at least one uint8_t value, or
     96an array of uint8_t values, which must match available CCIDS (see above). CCIDs
     97must be registered on the socket before calling connect() or listen().
     98
     99DCCP_SOCKOPT_TX_CCID is read/write. It returns the current CCID (if set) or sets
    100the preference list for the TX CCID, using the same format as DCCP_SOCKOPT_CCID.
    101Please note that the getsockopt argument type here is ``int``, not uint8_t.
    102
    103DCCP_SOCKOPT_RX_CCID is analogous to DCCP_SOCKOPT_TX_CCID, but for the RX CCID.
    104
    105DCCP_SOCKOPT_SERVER_TIMEWAIT enables the server (listening socket) to hold
    106timewait state when closing the connection (RFC 4340, 8.3). The usual case is
    107that the closing server sends a CloseReq, whereupon the client holds timewait
    108state. When this boolean socket option is on, the server sends a Close instead
    109and will enter TIMEWAIT. This option must be set after accept() returns.
    110
    111DCCP_SOCKOPT_SEND_CSCOV and DCCP_SOCKOPT_RECV_CSCOV are used for setting the
    112partial checksum coverage (RFC 4340, sec. 9.2). The default is that checksums
    113always cover the entire packet and that only fully covered application data is
    114accepted by the receiver. Hence, when using this feature on the sender, it must
    115be enabled at the receiver, too with suitable choice of CsCov.
    116
    117DCCP_SOCKOPT_SEND_CSCOV sets the sender checksum coverage. Values in the
    118	range 0..15 are acceptable. The default setting is 0 (full coverage),
    119	values between 1..15 indicate partial coverage.
    120
    121DCCP_SOCKOPT_RECV_CSCOV is for the receiver and has a different meaning: it
    122	sets a threshold, where again values 0..15 are acceptable. The default
    123	of 0 means that all packets with a partial coverage will be discarded.
    124	Values in the range 1..15 indicate that packets with minimally such a
    125	coverage value are also acceptable. The higher the number, the more
    126	restrictive this setting (see [RFC 4340, sec. 9.2.1]). Partial coverage
    127	settings are inherited to the child socket after accept().
    128
    129The following two options apply to CCID 3 exclusively and are getsockopt()-only.
    130In either case, a TFRC info struct (defined in <linux/tfrc.h>) is returned.
    131
    132DCCP_SOCKOPT_CCID_RX_INFO
    133	Returns a ``struct tfrc_rx_info`` in optval; the buffer for optval and
    134	optlen must be set to at least sizeof(struct tfrc_rx_info).
    135
    136DCCP_SOCKOPT_CCID_TX_INFO
    137	Returns a ``struct tfrc_tx_info`` in optval; the buffer for optval and
    138	optlen must be set to at least sizeof(struct tfrc_tx_info).
    139
    140On unidirectional connections it is useful to close the unused half-connection
    141via shutdown (SHUT_WR or SHUT_RD): this will reduce per-packet processing costs.
    142
    143
    144Sysctl variables
    145================
    146Several DCCP default parameters can be managed by the following sysctls
    147(sysctl net.dccp.default or /proc/sys/net/dccp/default):
    148
    149request_retries
    150	The number of active connection initiation retries (the number of
    151	Requests minus one) before timing out. In addition, it also governs
    152	the behaviour of the other, passive side: this variable also sets
    153	the number of times DCCP repeats sending a Response when the initial
    154	handshake does not progress from RESPOND to OPEN (i.e. when no Ack
    155	is received after the initial Request).  This value should be greater
    156	than 0, suggested is less than 10. Analogue of tcp_syn_retries.
    157
    158retries1
    159	How often a DCCP Response is retransmitted until the listening DCCP
    160	side considers its connecting peer dead. Analogue of tcp_retries1.
    161
    162retries2
    163	The number of times a general DCCP packet is retransmitted. This has
    164	importance for retransmitted acknowledgments and feature negotiation,
    165	data packets are never retransmitted. Analogue of tcp_retries2.
    166
    167tx_ccid = 2
    168	Default CCID for the sender-receiver half-connection. Depending on the
    169	choice of CCID, the Send Ack Vector feature is enabled automatically.
    170
    171rx_ccid = 2
    172	Default CCID for the receiver-sender half-connection; see tx_ccid.
    173
    174seq_window = 100
    175	The initial sequence window (sec. 7.5.2) of the sender. This influences
    176	the local ackno validity and the remote seqno validity windows (7.5.1).
    177	Values in the range Wmin = 32 (RFC 4340, 7.5.2) up to 2^32-1 can be set.
    178
    179tx_qlen = 5
    180	The size of the transmit buffer in packets. A value of 0 corresponds
    181	to an unbounded transmit buffer.
    182
    183sync_ratelimit = 125 ms
    184	The timeout between subsequent DCCP-Sync packets sent in response to
    185	sequence-invalid packets on the same socket (RFC 4340, 7.5.4). The unit
    186	of this parameter is milliseconds; a value of 0 disables rate-limiting.
    187
    188
    189IOCTLS
    190======
    191FIONREAD
    192	Works as in udp(7): returns in the ``int`` argument pointer the size of
    193	the next pending datagram in bytes, or 0 when no datagram is pending.
    194
    195SIOCOUTQ
    196	Returns the number of unsent data bytes in the socket send queue as ``int``
    197	into the buffer specified by the argument pointer.
    198
    199Other tunables
    200==============
    201Per-route rto_min support
    202	CCID-2 supports the RTAX_RTO_MIN per-route setting for the minimum value
    203	of the RTO timer. This setting can be modified via the 'rto_min' option
    204	of iproute2; for example::
    205
    206		> ip route change 10.0.0.0/24   rto_min 250j dev wlan0
    207		> ip route add    10.0.0.254/32 rto_min 800j dev wlan0
    208		> ip route show dev wlan0
    209
    210	CCID-3 also supports the rto_min setting: it is used to define the lower
    211	bound for the expiry of the nofeedback timer. This can be useful on LANs
    212	with very low RTTs (e.g., loopback, Gbit ethernet).
    213
    214
    215Notes
    216=====
    217DCCP does not travel through NAT successfully at present on many boxes. This is
    218because the checksum covers the pseudo-header as per TCP and UDP. Linux NAT
    219support for DCCP has been added.