Network Working Group A. Terzis
Request for Comments: 2745 UCLA
Category: Standards Track B. Braden
ISI
S. Vincent
Cisco Systems
L. Zhang
UCLA
January 2000
RSVP Diagnostic Messages
Status of this Memo
This document specifies an Internet standards track protocol for the
Internet community, and requests discussion and suggestions for
improvements. Please refer to the current edition of the "Internet
Official Protocol Standards" (STD 1) for the standardization state
and status of this protocol. Distribution of this memo is unlimited.
Copyright Notice
Copyright (C) The Internet Society (2000). All Rights Reserved.
Abstract
This document specifies the RSVP diagnostic facility, which allows a
user to collect information about the RSVP state along a path. This
specification describes the functionality, diagnostic message
formats, and processing rules.
1. Introduction
In the basic RSVP protocol [RSVP], error messages are the only means
for an end host to receive feedback regarding a failure in setting up
either path state or reservation state. An error message carries
back only the information from the failed point, without any
information about the state at other hops before or after the
failure. In the absence of failures, a host receives no feedback
regarding the details of a reservation that has been put in place,
such as whether, or where, or how, its own reservation request is
being merged with that of others. Such missing information can be
highly desirable for debugging purposes, or for network resource
management in general.
This document specifies the RSVP diagnostic facility, which is
designed to fill this information gap. The diagnostic facility can
be used to collect and report RSVP state information along the path
from a receiver to a specific sender. It uses Diagnostic messages
that are independent of other RSVP control messages and produce no
side-effects; that is, they do not change any RSVP state at either
nodes or hosts. Similarly, they provide not an error report but
rather a collection of requested RSVP state information.
The RSVP diagnostic facility was designed with the following goals:
- To collect RSVP state information from every RSVP-capable hop
along a path defined by path state, either for an existing
reservation or before a reservation request is made. More
specifically, we want to be able to collect information about
flowspecs, refresh timer values, and reservation merging at each
hop along the path.
- To collect the IP hop count across each non-RSVP cloud.
- To avoid diagnostic packet implosion or eXPlosion.
The following is specifically identified as a non-goal:
- Checking the resource availability along a path. Such
functionality may be useful for future reservation requests, but
it would require modifications to existing admission control
modules that is beyond the scope of RSVP.
2. Overview
The diagnostic facility introduces two new RSVP message types:
Diagnostic Request (DREQ) and Diagnostic Reply (DREP). A DREQ
message can be originated by a client in a "requester" host, which
may or may not be a participant of the RSVP session to be diagnosed.
A client in the requester host invokes the RSVP diagnostic facility
by generating a DREQ packet and sending it towards the LAST-HOP node,
which should be on the RSVP path to be diagnosed. This DREQ packet
specifies the RSVP session and a sender host for that session.
Starting from the LAST-HOP, the DREQ packet collects information
hop-by-hop as it is forwarded towards the sender (see Figure 1),
until it reaches the ending node. Specifically, each RSVP-capable
hop adds to the DREQ message a response (DIAG_RESPONSE) object
containing local RSVP state for the specified RSVP session.
When the DREQ packet reaches the ending node, the message type is
changed to Diagnostic Reply (DREP) and the completed response is sent
to the original requester node. Partial responses may also be
returned before the DREQ packet reaches the ending node if an error
condition along the path, such as "no path state", prevents further
forwarding of the DREQ packet. To avoid packet implosion or
explosion, all diagnostic packets are forwarded via unicast only.
Thus, there are generally three nodes (hosts and/or routers) involved
in performing the diagnostic function: the requester node, the
starting node, and the ending node, as shown in Figure 1. It is
possible that the client invoking the diagnosis function may reside
directly on the starting node, in which case that the first two nodes
are the same. The starting node is named "LAST-HOP", meaning the
last-hop of the path segment to be diagnosed. The LAST-HOP node can
be either a receiver node or an intermediate node along the path.
The ending node is usually the specified sender host. However, the
client can limit the length of the path segment to be diagnosed by
specifying a hop-count limit in the DREQ message.
LAST-HOP Ending
Receiver node node Sender
__ __ __ __ __
--------- ------> --> ...--> --> ...---->
__ __ DREQ __ DREQ __ DREQ __
^ .
.
DREQ . DREP DREP
.
__ DREP V V
Requester <------------------------------------
(client) ___
Figure 1
DREP packets can be unicast from the ending node back to the
requester either directly or hop-by-hop along the reverse of the path
taken by the DREQ message to the LAST-HOP, and thence to the
requester. The direct return is faster and more efficient, but the
hop-by-hop reverse-path route may be the only choice if the packets
have to cross firewalls. Hop-by-hop return is accomplished using an
optional ROUTE object, which is built incrementally to contain a list
of node addresses that the DREQ packet has passed through. The ROUTE
object is then used in reverse as a source route to forward the DREP
hop-by-hop back to the LAST-HOP node.
A DREQ message always consists of a single unfragmented IP datagram.
On the other hand, one DREQ message can generate multiple DREP
packets, each containing a fragment of the total DREQ message. When
the path consists of many hops, the total length of a DREP message
will exceed the MTU size before reaching the ending node; thus, the
message has to be fragmented. Relying on IP fragmentation and
reassembly, however, can be problematic, especially when DREP
messages are returned to the requester hop-by-hop, in which case
fragmentation/reassembly would have to be performed at every hop. To
avoid such excessive overhead, we let the requester define a default
path MTU size that is carried in every DREQ packet. If an
intermediate node finds that the default MTU size is bigger than the
MTU of the incoming interface, it reduces the default MTU size to the
MTU size of the incoming interface. If an intermediate node detects
that a DREQ packet size is larger than the default MTU size, it
returns to the requester (in either manner described above) a DREP
fragment containing accumulated responses. It then removes these
responses from the DREQ and continues to forward it. The requester
node can reassemble the resulting DREP fragments into a complete DREP
message.
When discussing diagnostic packet handling, this document uses
direction terminology that is consistent with the RSVP functional
specification [RSVP], relative to the direction of data packet flow.
Thus, a DREQ packet enters a node through an "outgoing interface" and
is forwarded towards the sender through an "incoming interface",
because DREQ packets travel in the reverse direction to the data
flow.
Notice that DREQ packets can be forwarded only after the RSVP path
state has been set up. If no path state exists, one may resort to
the traceroute or mtrace facility to examine whether the
unicast/multicast routing is working correctly.
3. Diagnostic Packet Format
Like other RSVP messages, DREQ and DREP messages consist of an RSVP
Common Header followed by a variable set of typed RSVP data objects.
The following sequence must be used:
+-----------------------------------+
RSVP Common Header
+-----------------------------------+
Session object
+-----------------------------------+
Next-Hop RSVP_HOP object
+-----------------------------------+
DIAGNOSTIC object
+-----------------------------------+
(optional) DIAG_SELECT object
+-----------------------------------+
(optional) ROUTE object
+-----------------------------------+
zero or more DIAG_RESPONSE objects
+-----------------------------------+
The session object identifies the RSVP session for which the state
information is being collected. We describe each of the other parts.
3.1. RSVP Message Common Header
The RSVP message common header is defined in [RSVP]. The following
specific exceptions and extensions are needed for DREP and DREQ.
Type field: define:
Type = 8: DREQ Diagnostic Request
Type = 9: DREP Diagnostic Reply
RSVP length:
If this is a DREP message and the MF flag in the DIAGNOSTIC object
(see below) is set, this field indicates the length of this single
DREP fragment rather than the total length of the complete DREP
reply message (which cannot generally be known in advance).
3.2. Next-Hop RSVP_HOP Object
This RSVP_HOP object carries the LIH of the interface through which
the DREQ should be received at the upstream node. This object is
updated hop-by hop. It is used for the same reasons that a RESV
message contains an RSVP_HOP object: to distinguish logical
interfaces and avoid problems caused by routing asymmetries and non-
RSVP clouds.
While the IP address is not really used during DREQ processing, for
consistency with the use of the RSVP_HOP object in other RSVP
messages, the IP address in the RSVP_HOP object to contain the
address of the interface through which the DREQ was sent.
3.3. DIAGNOSTIC Object
A DIAGNOSTIC object contains the common diagnostic control
information in both DREQ and DREP messages.
o IPv4 DIAGNOSTIC object: Class = 30, C-Type = 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Max-RSVP-hops RSVP-hop-count Reserved MF
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Request ID
+---------------+---------------+---------------+---------------+
Path MTU Fragment Offset
+---------------+---------------+---------------+---------------+
LAST-HOP Address
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
SENDER_TEMPLATE object
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Requester FILTER_SPEC object
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Here all IP addresses use the 4 byte IPv4 format, both explicitly in
the LAST-HOP Address and by using the IPv4 forms of the embedded
FILTER_SPEC and RSVP_HOP objects.
o IPv6 DIAGNOSTIC object: Class = 30, C-Type = 2
The format is the same, except all explicit and embedded IP addresses
are 16 byte IPv6 addresses.
The fields are as follows:
Max-RSVP-hops
An octet specifying the maximum number of RSVP hops over which
information will be collected. If an error condition in the
middle of the path prevents the DREQ packet from reaching the
specified ending node, the Max-RSVP-hops field may be used to
perform an expanding-length search to reach the point just before
the problem. If this value is 1, the starting node and the ending
node of the query will be the same. If it is zero, there is no
hop limit.
RSVP-hop-count
Records the number of RSVP hops that have been traversed so far.
If the starting and ending nodes are the same, this value will be
1 in the resulting DREP message.
Fragment Offset
Indicates where this DREP fragment belongs in the complete DREP
message, measured in octets. The first fragment has offset zero.
Fragment Offset is used also to determine if a DREQ message
containing zero DIAG_RESPONSE objects should be processed at an
RSVP capable node.
MF flag
Flag means "more fragments". It must be set to zero (0) in all
DREQ messages. It must be set to one (1) in all DREP packets that
carry partial results and are returned by intermediate nodes due
to the MTU limit. When the DREQ message is converted to a DREP
message in the ending node, the MF flag must remain zero.
Request ID
Identifies an individual DREQ message and the corresponding DREP
message (or all the fragments of the reply message).
One possible way to define the Request ID would use 16 bits to
specify the ID of the process making the query and 16 bits to
distinguish different queries from this process.
Path MTU
Specifies a default MTU size in octets for DREP and DREQ messages.
This value should not be smaller than the size of the "base" DREQ
packet. A "base" DREQ packet is one that contains a Common Header,
a Session object, a Next-Hop RSVP_HOP object, a DIAGNOSTIC object,
an empty ROUTE object and a single default DIAG_RESPONSE (see
below). The assumption made here is that a diagnostic packet of
this size can always be forwarded without IP fragmentation.
LAST-HOP Address
The IP address of the LAST-HOP node. The DREQ message starts
collecting information at this node and proceeds toward the
sender.
SENDER_TEMPLATE object
This IPv4/IPv6 SENDER_TEMPLATE object contains the IP address and
the port of a sender for the session being diagnosed. The DREQ
packet is forwarded hop-by-hop towards this address.
Requester FILTER_SPEC Object
This IPv4/IPv6 FILTER_SPEC object contains the IP address and the
port from which the request originated and to which the DREP
message(s) should be sent.
3.4. DIAG_SELECT Object
o DIAG_SELECT Class = 33, C-Type = 1.
A Diagnostic message may optionally contain a DIAG_SELECT object to
specify which specific RSVP objects should be reported in a
DIAG_RESPONSE object. In the absence of a DIAG_SELECT object, the
DIAG_RESPONSE object added by the node will contain a default set of
object types (see DIAG_RESPONSE object below).
The DIAG_SELECT object contains a list of [Class, C-type] pairs, in
the following format:
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
class C-Type class C-Type
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
// //
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
class C-Type class C-Type
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
When a DIAG_SELECT object is included in a DREQ message, each RSVP
node along the path will add a DIAG_RESPONSE object containing
response objects (see below) whose classes and C-Types match entries
in the DIAG_SELECT list (and are from matching path and reservation
state). A C-type octet of zero is a 'wildcard', matching any C-Type
associated with the associated class.
Depending on the type of objects requested, a node can find the
associated information in the path or reservation state stored for
the session described in the SESSION object. Specifically,
information for the RSVP_HOP,SENDER_TEMPLATE, SENDER_TSPEC, ADSPEC
objects can be extracted from the node's path state, while
information for the FLOWSPEC, FILTER_SPEC, CONFIRM, STYLE and SCOPE
objects can be found in the node's reservation state (if existent).
If the number of [Class, C-Type] pairs is odd, the last two octets of
the DIAG_SELECT object must be zero. A maximum DIAG_SELECT object is
one that contains the [Class, C-type] pairs for all the RSVP objects
that can be requested in a Diagnostic query.
3.5. ROUTE Object
A diagnostic message may contain a ROUTE object, which is used to
record the route of the DREQ message and as a source route for
returning the DREP message(s) hop-by-hop.
o IPv4 ROUTE object: Class = 31, C-Type = 1.
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
reserved R-pointer
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
+ RSVP Node List
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
This message signifies how the reply should be returned. If it does
not exist in the DREQ packet then DREP packets should be sent to the
requester directly. If it does exist, DREP packets must be returned
hop-by-hop along the reverse path to the LAST-HOP node and thence to
the requester node.
An empty ROUTE object is one that has an empty RSVP Node list and R-
pointer is equal to zero.
RSVP Node List
A list of RSVP node IPv4 addresses. The number of addresses in
this list can be computed from the object size.
R-pointer
Used in DREP messages only (see Section 4.2 for details), but it
is incremented as each hop adds its incoming interface address in
the ROUTE object.
o IPv6 ROUTE object: Class = 31, C-Type = 2
The same, except RSVP Node List contains IPv6 addresses.
In a DREQ message, RSVP Node List specifies all RSVP hops between the
LAST-HOP address specified in the DIAGNOSTIC object, and the last
RSVP node the DREQ message has visited. In a DREP message, RSVP Node
List specifies all RSVP hops between the LAST-HOP and the node that
returns this DREP message.
3.6. DIAG_RESPONSE Object
Each RSVP node attaches a DIAG_RESPONSE object to each DREQ message
it receives, before forwarding the message. The DIAG_RESPONSE object
contains the state to be reported for this node. It has a fixed-
format header and then a variable list of RSVP state objects, or
"response objects".
o IPv4 DIAG_RESPONSE object: Class = 32, C-Type = 1.
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
DREQ Arrival Time
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Incoming Interface Address
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Outgoing Interface Address
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Previous-RSVP-Hop Router Address
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
D-TTL MR-err K Timer value
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
(optional) TUNNEL object
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
// Response objects //
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
o IPv6 DIAG_RESPONSE object: Class = 32, C-Type = 2.
This object has the same format, except that all explicit and
embedded IP addresses are IPv6 addresses.
The fields are as follows:
DREQ Arrival Time
A 32-bit NTP timestamp specifying the time the DREQ message
arrived at this node. The 32-bit form of an NTP timestamp
consists of the middle 32 bits of the full 64-bit form, that is,
the low 16 bits of the integer part and the high 16 bits of the
fractional part.
Incoming Interface Address
Specifies the IP address of the interface on which messages from
the sender are expected to arrive, or 0 if unknown.
Outgoing Interface Address
Specifies the IP address of the interface through which the DREQ
message arrived and to which messages from the given sender and
for the specified session address flow, or 0 if unknown.
Previous-RSVP-Hop Router Address
Specifies the IP address from which this node receives RSVP PATH
messages for this source, or 0 if unknown. This is also the
interface to which the DREQ will be forwarded.
D-TTL
The number of IP hops this DREQ message traveled from the down-
stream RSVP node to the current node.
M flag
A single-bit flag which indicates whether the reservation
described by the response objects is merged with reservations from
other down-stream interfaces when being forwarded upstream.
R-error
A 3-bit field that indicates error conditions at a node. Currently
defined values are:
0x00: no error
0x01: No PATH state
0x02: packet too big
0x04: ROUTE object too big
K
The refresh timer multiple (defined in [RSVP]).
Timer value
The local refresh timer value in seconds.
The set of response objects to be included at the end of the
DIAG_RESPONSE object is determined by a DIAG_SELECT object, if one is
present. If no DIAG_SELECT object is present, the response objects
belong to the default list of classes:
SENDER_TSPEC object FILTER_SPEC object FLOWSPEC object
STYLE object
Any C-Type present in the local RSVP state will be used. These
response objects may be in any order but they must all be at the end
of the DIAG_RESPONSE object.
A default DIAG_RESPONSE object is one containing the default list of
classes described above.
3.7. TUNNEL Object
The optional TUNNEL object should be inserted when a DREQ message
arrives at an RSVP node that acts as a tunnel exit point.
The TUNNEL object provides the mapping between the end-to-end RSVP
session that is being diagnosed and the RSVP session over the tunnel.
This mapping information allows the diagnosis client to conduct
diagnosis over the involved tunnel session, by invoking a separate
Diagnostic query for the corresponding Tunnel Session and Tunnel
Sender. Keep in mind, however, that multiple end-to-end sessions may
all map to one pre-configured tunnel session that may have totally
different parameter settings.
The tunnel object is defined in the RSVP Tunnel Specification
[RSVPTUN].
4. Diagnostic Packet Forwarding Rules
4.1. DREQ Packet Forwarding
DREQ messages are forwarded hop-by-hop via unicast from the LAST-HOP
address to the Sender address, as specified in the DIAGNOSTIC object.
If an RSVP capable node, other than the LAST-HOP node, receives a
DREQ message that contains no DIAG_RESPONSE objects and has a zero
Fragment Offset, the node should forward the DREQ packet towards the
LAST-HOP without doing any of the processing mentioned below. The
reason is that such conditions apply only for nodes downstream of the
LAST-HOP where no information should be collected.
Processing begins when a DREQ message, DREQ_in, arrives at a node.
1. Create a new DIAG_RESPONSE object. Compute the IP hop count
from the previous RSVP hop. This is done by suBTracting the
value of the TTL value in the IP header from Send_TTL in the
RSVP common header. Save the result in the D-TTL field of the
DIAG_RESPONSE object.
2. Set the DREQ Arrival Time and the Outgoing Interface Address
in the DIAG_RESPONSE object. If this node is the LAST-HOP,
then the Out- going Interface Address field in the
DIAG_RESPONSE object contains the following value depending on
the session being diagnosed.
* If the session in question is a unicast session, then the
Out-going Interface Address field contains the address of
the interface LAST-HOP uses to send PATH messages and data
to the receiver specified by the session address.
* Otherwise, if it is a multicast session and there is at
least one receiver for this session, LAST_HOP should use the
address of one of local interfaces used to reach one of the
receivers.
* Otherwise Outgoing Interface Address should be zero.
3. Increment the RSVP-hop-count field in the DIAGNOSTIC message
object by one.
4. If no PATH state exists for the specified session, set R-error
= 0x01 (No PATH state) and goto step 7.
5. Set the rest of the fields in the DIAG_RESPONSE object. If
DREQ_in contains a DIAG_SELECT object, the response object
classes are those specified in the DIAG_SELECT; otherwise,
they are SENDER_TSPEC, STYLE, and FLOWSPEC objects. If no
reservation state exists for the specified RSVP session, the
DIAG_RESPONSE object will contain no FLOWSPEC, FILTER_SPEC or
STYLE object. If neither PATH nor reservation state exists for
the specified RSVP session, then no response objects will be
appended to the DIAG_RESPONSE object.
6. If RSVP-hop-count is less than Max-RSVP-hops and this node is
not the sender, then the DREQ is eligible for forwarding; set
the Path MTU to the min of the Path MTU and the MTU size of
the incoming interface for the sender being diagnosed.
7. If the size of DREQ_in plus the size of the new DIAG_RESPONSE
object plus the size of an IP address (if a ROUTE object
exists and R-error= 0) is larger than Path MTU, then the new
diagnostic message will be too large to be forwarded or
returned without fragmentation; set the "packet too big"
(0x02) error bit in DIAG_RESPONSE and goto Step SD1 in
Send_DREP (below).
8. If the "No PATH state" (0x01) error bit is set or if RSVP-
hop-count is equal to Max-RSVP-hops or if this node is the
sender, then the DREQ cannot be forwarded further; goto Step
10.
9. Forward the DREQ towards the sender, as follows. If a ROUTE
object exists, append the "Incoming Interface Address" to the
end of the ROUTE object and increment R-Pointer by one.
Update the Next-Hop RSVP_HOP object, append the new
DIAG_RESPONSE object to the list of DIAG_RESPONSE object, and
update the message length field in the RSVP common header
accordingly. Finally, recompute the checksum, forward DREQ_in
to the next hop towards the sender, and return.
10. Turn the DREQ into a DREP and return to the requester, as
follows. Append the DIAG_RESPONSE object to the end of
DREQ_in and update the packet length. If a ROUTE object is
present in the message, decrement the R-pointer and set target
address to the last address in the ROUTE object, otherwise set
target address to the requester address. Change the Type Field
in the Common header from DREQ to DREP. Finally, recompute
the checksum, send the DREP to the target address, and return.
Note that the MF bit must be off in this case.
Send_DREP:
This sequence is entered if the DREQ message augmented with the new
DIAG_RESPONSE object is too large to be forwarded towards the sender
or, if it is not eligible for forwarding, too large to be returned as
a DREP.
SD1. Make a copy of DREQ_in and change the message type field from
DREQ to DREP. Trim all DIAG_RESPONSE objects from DREQ_in and
adjust the Fragment Offset. The DREP message contains the
DIAG_RESPONSE objects accumulated by prior nodes.
SD2. Send the DREP message towards the requester, as follows. If a
ROUTE object is present in the DREP message, decrement the R-
pointer and set target address to the last address in the ROUTE
object, otherwise set target address to the requester address.
Set the MF bit, recompute the checksum and send the DREP message
back to the target address.
SD3. If the reduced size of DREQ_in plus the size of DIAG_RESPONSE
plus the size of an IP address (if a ROUTE object exists) is
smaller than or equal to Path MTU, then return to Step 8 of the
main DREQ processing sequence above.
SD4. If a ROUTE object exists, replace the ROUTE object in DREQ_in
with an empty ROUTE object and turn on the "ROUTE object too
big" (0x04) error bit in the DIAG_RESPONSE. In either case,
return to Step 8 of the main DREQ processing sequence above.
4.2. DREP Forwarding
When a ROUTE object is present, DREP messages are forwarded hop-by-
hop towards the requester, by reversing the route as listed in the
ROUTE object. Otherwise, DREP messages are sent directly to the
original requester.
When a node receives a DREP message, it simply decreases R-pointer by
one (address length), recomputes the checksum and forwards the
message to the address pointed to by R-pointer in the route list. If
a node, other than the LAST-HOP, receives a DREP packet where R-
pointer is equal to zero, it must send it directly to the requester.
When the LAST-HOP node receives a DREP message, it sends the message
to the requester.
4.3. MTU Selection and Adjustment
Because the DREQ message carries the allowed MTU size of previous
hops that the DREP messages will later traverse, this unique feature
allows easy semantic fragmentation as described above. Whenever the
DREQ message approaches the size of Path MTU, it can be trimmed
before being forwarded again.
When a requester sends a DREQ message, the Path MTU field in the
DIAGNOSTIC object can be set to a configured default value. It is
possible that the original Path MTU value is chosen larger than the
actual MTU value along some portion of the path being traced.
Therefore each intermediate RSVP node must check the MTU value when
processing a DREQ message. If the specified MTU value is larger than
the MTU of the incoming interface (that the DREQ message will be
forwarded to), the node changes the MTU value in the header to the
smaller value.
Whenever a DREQ message size becomes larger than the Path MTU value,
an intermediate RSVP node makes a copy of the message, converts it to
a DREP message to send back, and then trims off the partial results
from the DREQ message. If in this case also the DREQ cannot be
forwarded upstream due to a large ROUTE object, the "ROUTE object too
big" is set and the ROUTE object is trimmed. As a result of the ROUTE
object trimming, DREP(s) will come hop-by-hop up to this node and
will then immediately be forwarded to the requester address.
Even if the steps shown above are followed there are a few cases
where fragmentation at the IP layer will happen. For example, non-
RSVP hops with smaller MTUs may exist before LAST-HOP is reached, or
if the response is sent directly back to requester (as opposed to hop
by hop) the DREP may take a different route to the requester than the
DREQ took from the requester. Another case is when there exists a
link with MTU smaller than the minimum Path MTU value defined in
Section 3.3.
4.4. Errors
If an error condition prevents a DREP message from being forwarded
further, the message is simply dropped.
If an error condition, such as lack of PATH state, prevents a DREQ
message from being forwarded further, the node must change the
current message to DREP type and return it to the response address.
5. Problem Diagnosis by Using RSVP Diagnostic Facility
5.1. Across Firewalls
Firewalls may cause problems in diagnostic message forwarding. Let
us look at two different cases.
First, let us assume that the querier resides on a receiving host of
the session to be examined. In this case, firewalls should not
prevent the forwarding of the diagnostic messages in a hop-by-hop
manner, assuming that proper holes have been punched on the firewall
to allow hop-by-hop forwarding of other RSVP messages. The querier
may start by not including a ROUTE object, which can give a faster
response delivery and reduced overhead at intermediate nodes.
However if no response is received, the querier may resend the DREQ
message with a ROUTE object, specifying that a hop-by-hop reply
should be sent.
If the requester is a third party host and is separated from the
LAST-HOP address by a firewall (either the requester is behind a
firewall, or the LAST-HOP is a node behind a firewall, or both), at
this time we do not know any other solution but to change the LAST-
HOP to a node that is on the same side of the firewall as the
requester.
5.2. Examination of RSVP Timers
One can easily collect information about the current timer value at
each RSVP hop along the way. This will be very helpful in situations
when the reservation state goes up and down frequently, to find out
whether the state changes are due to improper setting of timer
values, or K values (when across lossy links), or frequent routing
changes.
5.3. Discovering Non-RSVP Clouds
The D-TTL field in each DIAG_RESPONSE object shows the number of
routing hops between adjacent RSVP nodes. Therefore any value
greater than one indicates a non-RSVP cloud in between. Together
with the arrival timestamps (assuming NTP works), this value can also
give some vague, though not necessarily accurate, indication of how
big that cloud might be. One might also find out all the
intermediate non-RSVP nodes by running either unicast or multicast
trace route.
5.4. Discovering Reservation Merges
The flowspec value in a DIAG_RESPONSE object specifies the amount of
resources being reserved for the data stream defined by the filter
spec in the same data block. When this value of adjacent
DIAG_RESPONSE objects differs, that is, a downstream node Rd has a
smaller value than its immediate upstream node Ru, it indicates a
merge of reservation with RSVP request(s) from other down stream
interface(s) at Rd. Further, in case of SE style reservation, one
can examine how the different SE scopes get merged at each hop.
In particular, if a receiver sends a DREQ message before sending its
own reservation, it can discover (1) how many RSVP hops there are
along the path between the specified sender and itself, (2) how many
of the hops already have some reservation by other receivers, and (3)
possibly a rough prediction of how its reservation request might get
merged with other existing ones.
5.5. Error Diagnosis
In addition to examining the state of a working reservation, RSVP
diagnostic messages are more likely to be invoked when things are not
working correctly. For example, a receiver has reserved an adequate
pipe for a specified incoming data stream, yet the observed delay or
loss ratio is much higher than expected. In this case the receiver
can use the diagnostic facility to examine the reservation state at
each RSVP hop along the way to find out whether the RSVP state is set
up correctly, whether there is any black-hole along the way that
caused RSVP message losses, or whether there are non-RSVP clouds, and
where they are, that may have caused the performance problem.
5.6. Crossing "Legacy" RSVP Routers
Since this diagnosis facility was developed and added to RSVP after a
number of RSVP implementations were in place, it is possible, or even
likely, that when performing RSVP diagnosis, one may encounter one or
more RSVP-capable nodes that do not understand diagnostic messages
and drop them. When this happens, the invoking client will get no
response from its requests.
One way to by-pass such "legacy" RSVP nodes is to perform RSVP
diagnosis repeatedly, guided by information from traceroute, or
mtrace in case of multicast. When an RSVP diagnostic query times out
(see next section), one may first use traceroute to get the list of
nodes along the path, and then gradually increase the value of Max-
RSVP-hops field in the DREQ message, starting from a low value until
one no longer receives a response. One can then try RSVP diagnosis
again by starting with the first node (which is further upstream
towards the sender) after the unresponding one.
There are two problem with the method mentioned above in the case of
unicast sessions. Both problems are related to the fact that
traceroute information provides the path from the requester to the
sender. The first problem is that the LAST-HOP may not be on the path
from the requester to the sender. In this case we can get information
only from the portion of the path from the LAST-HOP to the sender
which intersects with the path from the requester to the sender. If
routers that are not on the intersection of the two paths don't have
PATH state for the session being diagnosed then they will reply with
R-error=0x01. The requester can overcome this problem by sending a
DREQ to every router on the path (from itself to the sender) until it
reaches the first router that belongs to the path from the sender to
the LAST-HOP.
The second problem is that traceroute provides the path from the
requester to the sender which, due to routing asymmetries, may be
different than the path traffic from the sender to the LAST-HOP uses.
There is (at least) one case where this asymmetry will cause the
diagnosis to fail. We present this case below.
Downstream Path Sender
__ __ __ __
Receiver +------ <------ <-- ...--- -----
__ __ / __ __ __ __
--....--X _/ ^
__ __ \ Router B
Black \ __
Hole +-----> ---->---+
__ Upstream Path
Router A
Figure 2
Here the first hop upstream of the black hole is different on the
upstream path and the downstream path. Traceroute will indicate
router A as the previous hop (instead of router B which is the right
one). Sending a DREQ to router A will result in A responding with R-
error 0x01 (No PATH State). If the two paths converge again then the
requester can use the solution proposed above to get any (partial)
information from the rest of the path.
We don't have, for the moment, any complete solutions for the
problematic scenarios described here.
6. Comments on Diagnostic Client Implementation.
Following the design principle that nodes in the network should not
hold more than necessary state, RSVP nodes are responsible only for
forwarding Diagnostic messages and filling DIAG_RESPONSE objects.
Additional diagnostic functionality should be carried out by the
diagnostic clients. Furthermore, if the diagnostic function is
invoked from a third-party host, we should not require that host be
running an RSVP daemon to perform the function. Below we sketch out
the basic functions that a diagnostic client daemon should carry out.
1. Take input from the user about the session to be diagnosed, the
last-hop and the sender address, the Max-RSVP-hops, and
possibly the DIAG_SELECT list, create a DREQ message and send
to the LAST-HOP RSVP node using raw IP message with protocol
number 46 (RSVP). If the user specified that the response
should be sent hop-by-hop include an empty ROUTE object to the
DREQ message sent. Set the Path_MTU to the smaller of the user
request and the MTU of the link through which the DREQ will be
sent.
The port of the UDP socket on which the Diagnostic Client is
listening for replies should be included in the Requester
FILTER_SPEC object.
2. Set a retransmission timer, waiting for the reply (one or more
DREP messages). Listen to the specified UDP port for responses
from the LAST-HOP RSVP node.
The LAST-HOP RSVP node, upon receiving DREP messages, sends
them to the Diagnostic Client as UDP packets, using the port
supplied in the Requester FILTER_SPEC object.
3. Upon receiving a DREP message to an outstanding diagnostic
request, the client should clear the retransmission timer,
check to see if the reply contains the complete result of the
requested diagnosis. If so, it should pass the result up to
the invoking entity immediately.
4. Reassemble DREP fragments. If the first reply to an
outstanding diagnostic request contains only a fragment of the
expected result, the client should set up a reassembly timer in
a way similar to IP packet reassembly timer. If the timer goes
off before all fragments arrive, the client should pass the
partial result to the invoking entity.
5. Use retransmission and reassembly timers to gracefully handle
packet losses and reply fragment scenarios.
In the absence of response to the first diagnostic request, a
client should retransmit the request a few times. If all the
retransmissions also fail, the client should invoke traceroute
or mtrace to obtain the list of hops along the path segment to
be diagnosed, and then perform an iteration of diagnosis with
increasing hop count as suggested in Section 5.6 in order to
cross RSVP-capable but diagnosis-incapable nodes.
6. If all the above efforts fail, the client must notify the
invoking entity.
7. Security Considerations
RSVP Diagnostics, as any other diagnostic tool, can be a security
threat since it can reveal possibly sensitive RSVP state information
to unwanted third parties.
We feel that the threat is minimal, since as explained in the
Introduction Diagnostics messages produce no side-effects and
therefore they cannot change RSVP state in the nodes. In this respect
RSVP Diagnostics is less a security threat than other diagnostic
tools and protocols such as SNMP.
Furthermore, processing of Diagnostic messages can be disabled if it
is felt that is a security threat.
8. Acknowledgments
The idea of developing a diagnostic facility for RSVP was first
suggested by Mark Handley of ACIRI. Many thanks to Lee Breslau of
AT&T Labs and John Krawczyk of Nortel Networks for their valuable
comments on the first draft of this memo. Lee Breslau, Bob Braden,
and John Krawczyk contributed further comments after March 1996 IETF.
Steven Berson provided valuable comments on various drafts of the
memo. Tim Gleeson contributed an extensive list of editorial
comments. We would also like to acknowledge Intel for providing a
research grant as a partial support for this work. Subramaniam
Vincent did most of this work while a graduate research assistant at
the USC Information Sciences Institute (ISI).
9. References
[RSVP] Braden, R., Zhang, L., Berson, S., Herzog, S. and S. Jamin,
"Resource ReserVation Protocol -- Version 1 Functional
Specification", RFC2205, September 1997.
[RSVPTUN] Terzis, A., Krawczyk, J., Wroclawski, J. and L. Zhang,
"RSVP Operation Over IP Tunnels", RFC2746, January 2000.
10. Authors' Addresses
Andreas Terzis
UCLA
4677 Boelter Hall
Los Angeles, CA 90095
Phone: 310-267-2190
EMail: terzis@cs.ucla.edu
Bob Braden
USC Information Sciences Institute
4676 Admiralty Way
Marina del Rey, CA 90292
Phone: 310 822-1511
EMail:
braden@isi.edu
Subramaniam Vincent
Cisco Systems
275, E Tasman Drive, MS SJC04/2/1
San Jose, CA 95134
Phone: 408 525 3474
EMail: svincent@cisco.com
Lixia Zhang
UCLA
4531G Boelter Hall
Los Angeles, CA 90095
Phone: 310-825-2695
EMail: lixia@cs.ucla.edu
10. Full Copyright Statement
Copyright (C) The Internet Society (2000). All Rights Reserved.
This document and translations of it may be copied and furnished to
others, and derivative works that comment on or otherwise explain it
or assist in its implementation may be prepared, copied, published
and distributed, in whole or in part, without restriction of any
kind, provided that the above copyright notice and this paragraph are
included on all such copies and derivative works. However, this
document itself may not be modified in any way, such as by removing
the copyright notice or references to the Internet Society or other
Internet organizations, except as needed for the purpose of
developing Internet standards in which case the procedures for
copyrights defined in the Internet Standards process must be
followed, or as required to translate it into languages other than
English.
The limited permissions granted above are perpetual and will not be
revoked by the Internet Society or its successors or assigns.
This document and the information contained herein is provided on an
"AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING
TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING
BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION
HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF
MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
Acknowledgement
Funding for the RFCEditor function is currently provided by the
Internet Society.
