| 1 |
<pre>Network Working Group J. Mogul |
| 2 |
Request For Comments: 1063 C. Kent |
| 3 |
DEC |
| 4 |
C. Partridge |
| 5 |
BBN |
| 6 |
K. McCloghrie |
| 7 |
TWG |
| 8 |
July 1988 |
| 9 |
|
| 10 |
|
| 11 |
<span class="h1">IP MTU Discovery Options</span> |
| 12 |
|
| 13 |
STATUS OF THIS MEMO |
| 14 |
|
| 15 |
A pair of IP options that can be used to learn the minimum MTU of a |
| 16 |
path through an internet is described, along with its possible uses. |
| 17 |
This is a proposal for an Experimental protocol. Distribution of |
| 18 |
this memo is unlimited. |
| 19 |
|
| 20 |
INTRODUCTION |
| 21 |
|
| 22 |
Although the Internet Protocol allows gateways to fragment packets |
| 23 |
that are too large to forward, fragmentation is not always desirable. |
| 24 |
It can lead to poor performance or even total communication failure |
| 25 |
in circumstances that are surprisingly common. (For a thorough |
| 26 |
discussion of this issue, see [<a href="#ref-1" title=""Fragmentation Considered Harmful"">1</a>]). |
| 27 |
|
| 28 |
A datagram will be fragmented if it is larger than the Maximum |
| 29 |
Transmission Unit (MTU) of some network along the path it follows. |
| 30 |
In order to avoid fragmentation, a host sending an IP datagram must |
| 31 |
ensure that the datagram is no larger than the Minimum MTU (MINMTU) |
| 32 |
over the entire path. |
| 33 |
|
| 34 |
It has long been recognized that the methods for discovering the |
| 35 |
MINMTU of an IP internetwork path are inadequate. The methods |
| 36 |
currently available fall into two categories: (1) choosing small MTUs |
| 37 |
to avoid fragmentation or (2) using additional probe packets to |
| 38 |
discover when fragmentation will occur. Both methods have problems. |
| 39 |
|
| 40 |
Choosing MTUs requires a balance between network utilization (which |
| 41 |
requires the use of the largest possible datagram) and fragmentation |
| 42 |
avoidance (which in the absence of knowledge about the network path |
| 43 |
encourages the use of small, and thus too many, datagrams). Any |
| 44 |
choice for the MTU size, without information from the network, is |
| 45 |
likely to either fail to properly utilize the network or fail to |
| 46 |
avoid fragmentation. |
| 47 |
|
| 48 |
Probe packets have the problem of burdening the network with |
| 49 |
|
| 50 |
|
| 51 |
|
| 52 |
<span class="grey">Mogul, Kent, Partridge, & McCloghrie [Page 1]</span></pre> |
| 53 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-2" ></span> |
| 54 |
<span class="grey"><a href="./rfc1063">RFC 1063</a> IP MTU Discovery Options July 1988</span> |
| 55 |
|
| 56 |
|
| 57 |
unnecessary packets. And because network paths often change during |
| 58 |
the lifetime of a TCP connection, probe packets will have to be sent |
| 59 |
on a regular basis to detect any changes in the effective MINMTU. |
| 60 |
|
| 61 |
Implementors sometimes mistake the TCP MSS option as a mechanism for |
| 62 |
learning the network MINMTU. In fact, the MSS option is only a |
| 63 |
mechanism for learning about buffering capabilities at the two TCP |
| 64 |
peers. Separate provisions must be made to learn the IP MINMTU. |
| 65 |
|
| 66 |
In this memo, we propose two new IP options that, when used in |
| 67 |
conjunction will permit two peers to determine the MINMTU of the |
| 68 |
paths between them. In this scheme, one option is used to determine |
| 69 |
the lowest MTU in a path; the second option is used to convey this |
| 70 |
MTU back to the sender (possibly in the IP datagram containing the |
| 71 |
transport acknowledgement to the datagram which contained the MTU |
| 72 |
discovery option). |
| 73 |
|
| 74 |
OPTION FORMATS |
| 75 |
|
| 76 |
Probe MTU Option (Number 11) |
| 77 |
|
| 78 |
Format |
| 79 |
|
| 80 |
+--------+--------+--------+--------+ |
| 81 |
|00001011|00000100| 2 octet value | |
| 82 |
+--------+--------+--------+--------+ |
| 83 |
|
| 84 |
Definition |
| 85 |
|
| 86 |
This option always contains the lowest MTU of all the networks |
| 87 |
that have been traversed so far by the datagram. |
| 88 |
|
| 89 |
A host that sends this option must initialize the value field to |
| 90 |
be the MTU of the directly-connected network. If the host is |
| 91 |
multi-homed, this should be for the first-hop network. |
| 92 |
|
| 93 |
Each gateway that receives a datagram containing this option must |
| 94 |
compare the MTU field with the MTUs of the inbound and outbound |
| 95 |
links for the datagram. If either MTU is lower than the value in |
| 96 |
the MTU field of the option, the option value should be set to the |
| 97 |
lower MTU. (Note that gateways conforming to <a href="./rfc1009">RFC-1009</a> may not |
| 98 |
know either the inbound interface or the outbound interface at the |
| 99 |
time that IP options are processed. Accordingly, support for this |
| 100 |
option may require major gateway software changes). |
| 101 |
|
| 102 |
Any host receiving a datagram containing this option should |
| 103 |
confirm that value of the MTU field of the option is less than or |
| 104 |
equal to that of the inbound link, and if necessary, reduce the |
| 105 |
|
| 106 |
|
| 107 |
|
| 108 |
<span class="grey">Mogul, Kent, Partridge, & McCloghrie [Page 2]</span></pre> |
| 109 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-3" ></span> |
| 110 |
<span class="grey"><a href="./rfc1063">RFC 1063</a> IP MTU Discovery Options July 1988</span> |
| 111 |
|
| 112 |
|
| 113 |
MTU field value, before processing the option. |
| 114 |
|
| 115 |
If the receiving host is not able to accept datagrams as large as |
| 116 |
specified by the value of the MTU field of the option, then it |
| 117 |
should reduce the MTU field to the size of the largest datagram it |
| 118 |
can accept. |
| 119 |
|
| 120 |
Reply MTU Option (Number 12) |
| 121 |
|
| 122 |
Format |
| 123 |
|
| 124 |
+--------+--------+--------+--------+ |
| 125 |
|00001100|00000100| 2 octet value | |
| 126 |
+--------+--------+--------+--------+ |
| 127 |
|
| 128 |
Definition |
| 129 |
|
| 130 |
This option is used to return the value learned from a Probe MTU |
| 131 |
option to the sender of the Probe MTU option. |
| 132 |
|
| 133 |
RELATION TO TCP MSS |
| 134 |
|
| 135 |
Note that there are two superficially similar problems in choosing |
| 136 |
the size of a datagram. First, there is the restriction [<a href="#ref-2" title=""Internet Protocol"">2</a>] that a |
| 137 |
host not send a datagram larger than 576 octets unless it has |
| 138 |
assurance that the destination is prepared to accept a larger |
| 139 |
datagram. Second, the sending host should not send a datagram larger |
| 140 |
than MINMTU, in order to avoid fragmentation. The datagram size |
| 141 |
should normally be the minimum of these two lower bounds. |
| 142 |
|
| 143 |
In the past, the TCP MSS option [<a href="#ref-3" title=""Transmission Control Protocol"">3</a>] has been used to avoid sending |
| 144 |
packets larger than the destination can accept. Unfortunately, this |
| 145 |
is not the most general mechanism; it is not available to other |
| 146 |
transport layers, and it cannot determine the MINMTU (because |
| 147 |
gateways do not parse TCP options). |
| 148 |
|
| 149 |
Because the MINMTU returned by a probe cannot be larger than the |
| 150 |
maximum datagram size that the destination can accept, this IP option |
| 151 |
could, in theory, supplant the use of the TCP MSS option, providing |
| 152 |
an economy of mechanism. (Note however, that some researchers |
| 153 |
believe that the value of the TCP MSS is distinct from the path's |
| 154 |
MINMTU. The MSS is the upper limit of the data size that the peer |
| 155 |
will accept, while the MINMTU represents a statement about the data |
| 156 |
size supported by the path). |
| 157 |
|
| 158 |
Note that a failure to observe the MINMTU restriction is not normally |
| 159 |
fatal; fragmentation will occur, but this is supposed to work. A |
| 160 |
failure to observe the TCP MSS option, however, could be fatal |
| 161 |
|
| 162 |
|
| 163 |
|
| 164 |
<span class="grey">Mogul, Kent, Partridge, & McCloghrie [Page 3]</span></pre> |
| 165 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-4" ></span> |
| 166 |
<span class="grey"><a href="./rfc1063">RFC 1063</a> IP MTU Discovery Options July 1988</span> |
| 167 |
|
| 168 |
|
| 169 |
because it might lead to datagrams that can never be accepted by the |
| 170 |
destination. Therefore, unless and until the Probe MTU option is |
| 171 |
universally implemented, at least by hosts, the TCP MSS option must |
| 172 |
be used as well. |
| 173 |
|
| 174 |
IMPLEMENTATION APPROACHES |
| 175 |
|
| 176 |
Who Sends the Option |
| 177 |
|
| 178 |
There are at least two ways to implement the MTU discovery scheme. |
| 179 |
One method makes the transport layer responsible for MTU |
| 180 |
discovery; the other method makes the IP layer responsible for MTU |
| 181 |
discovery. A host system should support one of the two schemes. |
| 182 |
|
| 183 |
Transport Discovery |
| 184 |
|
| 185 |
In the transport case, the transport layer can include the Probe |
| 186 |
MTU option in an outbound datagram. When a datagram containing |
| 187 |
the Probe MTU option is received, the option must be passed up to |
| 188 |
the receiving transport layer, which should then acknowledge the |
| 189 |
Probe with a Reply MTU option in the next return datagram. Note |
| 190 |
that because the options are placed on unreliable datagrams, the |
| 191 |
original sender will have to resend Probes (possibly once per |
| 192 |
window of data) until it receives a Reply option. Also note that |
| 193 |
the Reply MTU option may be returned on an IP datagram for a |
| 194 |
different transport protocol from which it was sent (e.g., TCP |
| 195 |
generated the probe but the Reply was received on a UDP datagram). |
| 196 |
|
| 197 |
IP Discovery |
| 198 |
|
| 199 |
A better scheme is to put MTU discovery into the IP layer, using |
| 200 |
control mechanisms in the routing cache. Whenever an IP datagram |
| 201 |
is sent, the IP layer checks in the routing cache to see if a |
| 202 |
Probe or Reply MTU option needs to be inserted in the datagram. |
| 203 |
Whenever a datagram containing either option is received, the |
| 204 |
information in those options is placed in the routing cache. |
| 205 |
|
| 206 |
The basic working of the protocol is somewhat complex. We trace |
| 207 |
it here through one round-trip. Implementors should realize that |
| 208 |
there may be cases where both options are contained in one |
| 209 |
datagram. For the purposes of this exposition, the sender of the |
| 210 |
probe is called the Probe-Sender and the receiver, Probe-Receiver. |
| 211 |
|
| 212 |
When the IP layer is asked to send a Probe MTU option (see the |
| 213 |
section below on when to probe), it makes some record in the |
| 214 |
routing cache that indicates the next IP datagram to Probe- |
| 215 |
Receiver should contain the Probe MTU option. |
| 216 |
|
| 217 |
|
| 218 |
|
| 219 |
|
| 220 |
<span class="grey">Mogul, Kent, Partridge, & McCloghrie [Page 4]</span></pre> |
| 221 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-5" ></span> |
| 222 |
<span class="grey"><a href="./rfc1063">RFC 1063</a> IP MTU Discovery Options July 1988</span> |
| 223 |
|
| 224 |
|
| 225 |
When the next IP datagram to Probe-Receiver is sent, the Probe MTU |
| 226 |
option is inserted. The IP layer in Probe-Sender should continue |
| 227 |
to send an occasional Probe MTU in subsequent datagrams until a |
| 228 |
Reply MTU option is received. It is strongly recommended that the |
| 229 |
Probe MTU not be sent in all datagrams but only at such a rate |
| 230 |
that, on average, one Probe MTU will be sent per round-trip |
| 231 |
interval. (Another way of saying this is that we would hope that |
| 232 |
only one datagram in a transport protocol window worth of data has |
| 233 |
the Probe MTU option set). This mechanism might be implemented by |
| 234 |
sending every Nth packet, or, in those implementations where the |
| 235 |
round-trip time estimate to the destination is cached with the |
| 236 |
route, once every estimated RTT. |
| 237 |
|
| 238 |
When a Probe MTU option is received by Probe-Receiver, the |
| 239 |
receiving IP should place the value of this option in the next |
| 240 |
datagram it sends back to Probe-Sender. The value is then |
| 241 |
discarded. In other words, each Probe MTU option causes the Reply |
| 242 |
MTU option to be placed in one return datagram. |
| 243 |
|
| 244 |
When Probe-Sender receives the Reply MTU option, it should check |
| 245 |
the value of the option against the current MINMTU estimate in the |
| 246 |
routing cache. If the option value is lower, it becomes the new |
| 247 |
MINMTU estimate. If the option value is higher, Probe-Sender |
| 248 |
should be more conservative about changing the MINMTU estimate. |
| 249 |
If a route is flapping, the MINMTU may change frequently. In such |
| 250 |
situations, keeping the smallest MINMTU of various routes in use |
| 251 |
is preferred. As a result, a higher MINMTU estimate should only |
| 252 |
be accepted after a lower estimate has been permitted to "age" a |
| 253 |
bit. In other words, if the probe value is higher than the |
| 254 |
estimated MINMTU, only update the estimate if the estimate is |
| 255 |
several seconds old or more. Finally, whenever the Probe-Sender |
| 256 |
receives a Reply MTU option, it should stop retransmitting probes |
| 257 |
to Probe-Receiver. |
| 258 |
|
| 259 |
A few additional issues complicate this discussion. |
| 260 |
|
| 261 |
One problem is setting the default MINMTU when no Reply MTU |
| 262 |
options have been received. We recommend the use of the minimum |
| 263 |
of the supported IP datagram size (576 octets) and the connected |
| 264 |
network MTU for destinations not on the local connected network, |
| 265 |
and the connected network MTU for hosts on the connected network. |
| 266 |
|
| 267 |
The MINMTU information, while kept by the Internet layer, is in |
| 268 |
fact, only of interest to the transport and higher layers. |
| 269 |
Accordingly, the Internet layer must keep the transport layer |
| 270 |
informed of the current value of the estimated MINMTU. |
| 271 |
Furthermore, minimal transport protocols, such as UDP, must be |
| 272 |
prepared to pass this information up to the transport protocol |
| 273 |
|
| 274 |
|
| 275 |
|
| 276 |
<span class="grey">Mogul, Kent, Partridge, & McCloghrie [Page 5]</span></pre> |
| 277 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-6" ></span> |
| 278 |
<span class="grey"><a href="./rfc1063">RFC 1063</a> IP MTU Discovery Options July 1988</span> |
| 279 |
|
| 280 |
|
| 281 |
user. |
| 282 |
|
| 283 |
It is expected that there will be a transition period during which |
| 284 |
some hosts support this option and some do not. As a result, |
| 285 |
hosts should stop sending Probe MTU options and refuse to send any |
| 286 |
further options if it does not receive either a Probe MTU option |
| 287 |
or Reply MTU option from the remote system after a certain number |
| 288 |
of Probe MTU options have been sent. In short, if Probe-Sender |
| 289 |
has sent several probes but has gotten no indication that Probe- |
| 290 |
Receiver supports MTU probing, then Probe-Sender should assume |
| 291 |
that Probe-Receiver does not support probes. (Obviously, if |
| 292 |
Probe-Sender later receives a probe option from Probe-Receiver, it |
| 293 |
should revise its opinion.) |
| 294 |
|
| 295 |
Implementations should not assume that routes to the same |
| 296 |
destination that have a different TOS have the same estimated |
| 297 |
MINMTU. We recommend that the MTU be probed separately for each |
| 298 |
TOS. |
| 299 |
|
| 300 |
Respecting the TCP MSS |
| 301 |
|
| 302 |
One issue concerning TCP MSS is that it is usually negotiated |
| 303 |
assuming an IP header that contains no options. If the transport |
| 304 |
layer is sending maximum size segments, it may not leave space for |
| 305 |
IP to fit the options into the datagram. Thus, insertion of the |
| 306 |
Probe MTU or Reply MTU option may violate the MSS restriction. |
| 307 |
Because, unlike other IP options, the MTU options can be inserted |
| 308 |
without the knowledge of the transport layer, the implementor must |
| 309 |
carefully consider the implications of adding options to an IP |
| 310 |
datagram. |
| 311 |
|
| 312 |
One approach is to reserve 4 bytes from the MINMTU reported to the |
| 313 |
transport layer; this will allow the IP layer to insert at least |
| 314 |
one MTU option in every datagram (it can compare the size of the |
| 315 |
outgoing datagram with the MINMTU stored in the route cache to see |
| 316 |
how much room there actually is). This is simple to implement, |
| 317 |
but does waste a little bandwidth in the normal case. |
| 318 |
|
| 319 |
Another approach is to provide a means for the IP layer to notify |
| 320 |
the transport layer that space must be reserved for sending an |
| 321 |
option; the transport layer would then make a forthcoming segment |
| 322 |
somewhat smaller than usual. |
| 323 |
|
| 324 |
When a Probe Can Be Sent |
| 325 |
|
| 326 |
A system that receives a Probe MTU option should always respond |
| 327 |
with a Reply MTU option, unless the probe was sent to an IP or LAN |
| 328 |
broadcast address. |
| 329 |
|
| 330 |
|
| 331 |
|
| 332 |
<span class="grey">Mogul, Kent, Partridge, & McCloghrie [Page 6]</span></pre> |
| 333 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-7" ></span> |
| 334 |
<span class="grey"><a href="./rfc1063">RFC 1063</a> IP MTU Discovery Options July 1988</span> |
| 335 |
|
| 336 |
|
| 337 |
A Probe MTU option should be sent in any of the following |
| 338 |
situations: |
| 339 |
|
| 340 |
(1) The MINMTU for the path is not yet known; |
| 341 |
|
| 342 |
(2) A received datagram suffers a fragmentation re-assembly |
| 343 |
timeout. (This is a strong hint the path has changed; |
| 344 |
send a probe to the datagram's source); |
| 345 |
|
| 346 |
(3) An ICMP Time Exceeded/Fragmentation Reassembly Timeout is |
| 347 |
received (this is the only message we will get that |
| 348 |
indicates fragmentation occurred along the network path); |
| 349 |
|
| 350 |
(4) The transport layer requests it. |
| 351 |
|
| 352 |
Implementations may also wish to periodically probe a path, even |
| 353 |
if there is no indication that fragmentation is occurring. This |
| 354 |
practice is perfectly reasonable; if fragmentation and reassembly |
| 355 |
is working perfectly, the sender may never get any indication that |
| 356 |
the path MINMTU has changed unless a probe is sent. We recommend, |
| 357 |
however, that implementations send such periodic probes sparingly. |
| 358 |
Once every few minutes, or once every few hundred datagrams is |
| 359 |
probably sufficient. |
| 360 |
|
| 361 |
There are also some scenarios in which the Probe MTU should not be |
| 362 |
sent, even though there may be some indication of an MINMTU |
| 363 |
change: |
| 364 |
|
| 365 |
(1) Probes should not be sent in response to the receipt of |
| 366 |
a probe option. Although the fact that the remote peer |
| 367 |
is probing indicates that the MINMTU may have changed, |
| 368 |
sending a probe in response to a probe causes a continuous |
| 369 |
exchange of probe options. |
| 370 |
|
| 371 |
(2) Probes must not be sent in response to fragmented |
| 372 |
datagrams except when the fragmentation reassembly |
| 373 |
of the datagram fails. The problem in this case is |
| 374 |
that the receiver has no mechanism for informing the remote |
| 375 |
peer that fragmentation has occurred, unless fragmentation |
| 376 |
reassembly fails (in which case an ICMP message is sent). |
| 377 |
Thus, a peer may use the wrong MTU for some time before |
| 378 |
discovering a problem. If we probe on fragmented |
| 379 |
datagrams, we may probe, unnecessarily, for some time |
| 380 |
until the remote peer corrects its MTU. |
| 381 |
|
| 382 |
(3) For compatibility with hosts that do not implement the |
| 383 |
option, no Probe MTU Option should be sent more than |
| 384 |
ten times without receiving a Reply MTU Option or a |
| 385 |
|
| 386 |
|
| 387 |
|
| 388 |
<span class="grey">Mogul, Kent, Partridge, & McCloghrie [Page 7]</span></pre> |
| 389 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-8" ></span> |
| 390 |
<span class="grey"><a href="./rfc1063">RFC 1063</a> IP MTU Discovery Options July 1988</span> |
| 391 |
|
| 392 |
|
| 393 |
Probe MTU Option from the remote peer. Peers which |
| 394 |
ignore probes and do not send probes must be treated |
| 395 |
as not supporting probes. |
| 396 |
|
| 397 |
(4) Probes should not be sent to an IP or LAN broadcast |
| 398 |
address. |
| 399 |
|
| 400 |
(5) We recommend that Probe MTUs not be sent to other hosts |
| 401 |
on the directly-connected network, but that this feature |
| 402 |
be configurable. There are situations (for example, when |
| 403 |
Proxy ARP is in use) where it may be difficult to determine |
| 404 |
which systems are on the directly-connected network. In |
| 405 |
this case, probing may make sense. |
| 406 |
|
| 407 |
SAMPLE IMPLEMENTATION SKETCH |
| 408 |
|
| 409 |
We present here a somewhat more concrete description of how an IP- |
| 410 |
layer implementation of MTU probing might be designed. |
| 411 |
|
| 412 |
First, the routing cache entries are enhanced to store seven |
| 413 |
additional values: |
| 414 |
|
| 415 |
MINMTU: The current MINMTU of the path. |
| 416 |
|
| 417 |
ProbeRetry: A timestamp indicating when the next probe |
| 418 |
should be sent. |
| 419 |
|
| 420 |
LastDecreased: A timestamp showing when the MTU was |
| 421 |
last decreased. |
| 422 |
|
| 423 |
ProbeReply: A bit indicating a Reply MTU option should be |
| 424 |
sent. |
| 425 |
|
| 426 |
ReplyMTU: The value to go in the Reply MTU option. |
| 427 |
|
| 428 |
SupportsProbes: A bit indicating that the remote peer |
| 429 |
can deal with probes (always defaults to |
| 430 |
1=true). |
| 431 |
|
| 432 |
ConsecutiveProbes: The number of probes sent without |
| 433 |
the receipt of a Probe MTU or Reply |
| 434 |
MTU option. |
| 435 |
|
| 436 |
There are also several configuration parameters; these should be |
| 437 |
configurable by appropriate network management software; the values |
| 438 |
we suggest are "reasonable": |
| 439 |
|
| 440 |
Default_MINMTU: The default value for the MINMTU field of the |
| 441 |
|
| 442 |
|
| 443 |
|
| 444 |
<span class="grey">Mogul, Kent, Partridge, & McCloghrie [Page 8]</span></pre> |
| 445 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-9" ></span> |
| 446 |
<span class="grey"><a href="./rfc1063">RFC 1063</a> IP MTU Discovery Options July 1988</span> |
| 447 |
|
| 448 |
|
| 449 |
routing cache entry, to be used when the real |
| 450 |
MINMTU is unknown. Recommended value: 576. |
| 451 |
|
| 452 |
Max_ConsecutiveProbs: The maximum number of probes to send |
| 453 |
before assuming that the destination does |
| 454 |
not support the probe option. |
| 455 |
Recommended value: 10. |
| 456 |
|
| 457 |
ProbeRetryTime: The time (in seconds) to wait before retrying |
| 458 |
an unanswered probe. Recommended value: |
| 459 |
60 seconds, or 2*RTT if the the RTT is available |
| 460 |
to the IP layer. |
| 461 |
|
| 462 |
ReprobeInterval: The time to wait before sending a probe after |
| 463 |
receiving a successful Reply MTU, in order to |
| 464 |
detect increases in the route's MINMTU. |
| 465 |
Recommended value: 5 times the ProbeRetryTime. |
| 466 |
|
| 467 |
IncreaseInterval: The time to wait before increasing the MINMTU |
| 468 |
after the value has been decreased, to prevent |
| 469 |
flapping. Recommended value: same as |
| 470 |
ProbeRetryTime. |
| 471 |
|
| 472 |
When a new route is entered into the routing cache, the initial |
| 473 |
values should be set as follows: |
| 474 |
|
| 475 |
MINMTU = Default_MINMTU |
| 476 |
|
| 477 |
ProbeRetry = Current Time |
| 478 |
|
| 479 |
LastDecreased = Current Time - IncreaseInterval |
| 480 |
|
| 481 |
ProbeReply = false |
| 482 |
|
| 483 |
SupportsProbes = true |
| 484 |
|
| 485 |
ConsecutiveProbes = 0 |
| 486 |
|
| 487 |
This initialization is done before attempting to send the first |
| 488 |
packet along this route, so that the first packet will contain a |
| 489 |
Probe MTU option. |
| 490 |
|
| 491 |
Whenever the IP layer sends a datagram on this route it checks the |
| 492 |
SupportsProbes bit to see if the remote system supports probing. If |
| 493 |
the SupportsProbes bit is set, and the timestamp in ProbeRetry is |
| 494 |
less than or equal to the current time, a Probe option should be sent |
| 495 |
in the datagram, and the ProbeRetry field incremented by |
| 496 |
ProbeRetryTime. |
| 497 |
|
| 498 |
|
| 499 |
|
| 500 |
<span class="grey">Mogul, Kent, Partridge, & McCloghrie [Page 9]</span></pre> |
| 501 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-10" ></span> |
| 502 |
<span class="grey"><a href="./rfc1063">RFC 1063</a> IP MTU Discovery Options July 1988</span> |
| 503 |
|
| 504 |
|
| 505 |
Whether or not the Probe MTU option is sent in a datagram, if the |
| 506 |
ProbeReply bit is set, then a Reply MTU option with the value of the |
| 507 |
ReplyMTU field is placed in the outbound datagram. The ProbeReply |
| 508 |
bit is then cleared. |
| 509 |
|
| 510 |
Every time a Probe option is sent, the ConsecutiveProbes value should |
| 511 |
be incremented. If this value reaches Max_ConsecutiveProbes, the |
| 512 |
SupportsProbe bit should be cleared. |
| 513 |
|
| 514 |
When an IP datagram containing the Probe MTU option is received, the |
| 515 |
receiving IP sets the ReplyMTU to the Probe MTU option value and sets |
| 516 |
the ProbeReply bit in its outbound route to the source of the |
| 517 |
datagram. The SupportsProbe bit is set, and the ConsecutiveProbes |
| 518 |
value is reset to 0. |
| 519 |
|
| 520 |
If an IP datagram containing the Reply MTU option is received, the IP |
| 521 |
layer must locate the routing cache entry corresponding to the source |
| 522 |
of the Reply MTU option; if no such entry exists, a new one (with |
| 523 |
default values) should be created. The SupportsProbe bit is set, and |
| 524 |
the ConsecutiveProbes value is reset to 0. The ProbeRetry field is |
| 525 |
set to the current time plus ReprobeInterval. |
| 526 |
|
| 527 |
Four cases are possible when a Reply MTU option is received: |
| 528 |
|
| 529 |
(1) The Reply MTU option value is less than the current |
| 530 |
MINMTU: the MINMTU field is set to the new value, and |
| 531 |
the LastDecreased field is set to the current time. |
| 532 |
|
| 533 |
(2) The Reply MTU option value is greater than the |
| 534 |
current MINMTU and the LastDecreased field plus |
| 535 |
IncreaseInterval is less than the current time: set the |
| 536 |
ProbeRetry field to LastDecreased plus IncreaseInterval, |
| 537 |
but do not change MINMTU. |
| 538 |
|
| 539 |
(3) The Reply MTU option value is greater than the |
| 540 |
current MINMTU and the LastDecreased field plus |
| 541 |
IncreaseInterval is greater than the current time: set |
| 542 |
the MINMTU field to the new value. |
| 543 |
|
| 544 |
(4) The Reply MTU option value is equal to the current |
| 545 |
MINMTU: do nothing more. |
| 546 |
|
| 547 |
Whenever the MTU field is changed, the transport layer should be |
| 548 |
notified, either by an upcall or by a change in a shared variable |
| 549 |
(which may be accessed from the transport layer by a downcall). |
| 550 |
|
| 551 |
If a fragmentation reassembly timeout occurs, if an ICMP Time |
| 552 |
Exceeded/Fragmentation Reassembly Timeout is received, or if the IP |
| 553 |
|
| 554 |
|
| 555 |
|
| 556 |
<span class="grey">Mogul, Kent, Partridge, & McCloghrie [Page 10]</span></pre> |
| 557 |
<hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-11" ></span> |
| 558 |
<span class="grey"><a href="./rfc1063">RFC 1063</a> IP MTU Discovery Options July 1988</span> |
| 559 |
|
| 560 |
|
| 561 |
layer is asked to send a probe by a higher layer, the ProbeRetry |
| 562 |
field for the appropriate routing cache entry is set to the current |
| 563 |
time. This will cause a Probe option to be sent with the next |
| 564 |
datagram (unless the SupportsProbe bit is turned off). |
| 565 |
|
| 566 |
MANAGEMENT PARAMETERS |
| 567 |
|
| 568 |
We suggest that the following parameters be made available to local |
| 569 |
applications and remote network management systems: |
| 570 |
|
| 571 |
(1) The number of probe retries to be made before determining |
| 572 |
a system is down. The value of 10 is certain to be wrong |
| 573 |
in some situations. |
| 574 |
|
| 575 |
(2) The frequency with which probes are sent. Systems may |
| 576 |
find that more or less frequent probing is more cost |
| 577 |
effective. |
| 578 |
|
| 579 |
(3) The default MINMTU used to initialize routes. |
| 580 |
|
| 581 |
(4) Applications should have the ability to force a probe |
| 582 |
on a particular route. There are cases where a probe |
| 583 |
needs to be sent but the sender doesn't know it. An |
| 584 |
operator must be able to cause a probe in such situations. |
| 585 |
Furthermore, it may be useful for applications to "ping" |
| 586 |
for the MTU. |
| 587 |
|
| 588 |
REFERENCES |
| 589 |
|
| 590 |
[<a id="ref-1">1</a>] Kent, C. and J. Mogul, "Fragmentation Considered |
| 591 |
Harmful", Proc. ACM SIGCOMM '87, Stowe, VT, August 1987. |
| 592 |
|
| 593 |
[<a id="ref-2">2</a>] Postel, J., Ed., "Internet Protocol", <a href="./rfc791">RFC-791</a>, |
| 594 |
USC/Information Sciences Institute, Marina del Rey, CA, |
| 595 |
September 1981. |
| 596 |
|
| 597 |
[<a id="ref-3">3</a>] Postel, J., Ed., "Transmission Control Protocol", <a href="./rfc793">RFC-793</a>, |
| 598 |
USC/Information Sciences Institute, Marina del Rey, CA, |
| 599 |
September 1981. |
| 600 |
|
| 601 |
[<a id="ref-4">4</a>] Postel, J., "The TCP Maximum Segment Size and Related Topics", |
| 602 |
<a href="./rfc879">RFC-879</a>, USC/Information Sciences Institute, Marina del Rey, |
| 603 |
CA, November 1983. |
| 604 |
|
| 605 |
|
| 606 |
|
| 607 |
|
| 608 |
|
| 609 |
|
| 610 |
|
| 611 |
|
| 612 |
Mogul, Kent, Partridge, & McCloghrie [Page 11] |
| 613 |
</pre> |