| TOC |
|
There have been many implementations and deployments of legacy syslog over TCP for many years. That protocol has evolved without being standardized and has proven to be quite interoperable in practice.
The aim of this specification is to document three things: how to transmit standardized syslog over TCP, how this has been done for legacy syslog, and how the new syslog architecture can interoperate with the legacy deployments.
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as “work in progress.”
This Internet-Draft will expire on October 25, 2010.
Copyright (c) 2010 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
1.
Introduction
2.
Conventions Used in This Document
3.
Message Transmission
3.1.
Session
3.2.
Session Initiation
3.3.
Message Transfer
3.3.1.
Octet-Counting
3.3.2.
Octet-Stuffing
3.4.
Retaining the Original Message
3.5.
Session Closure
4.
Security Considerations
4.1.
Sender Authentication and Message Forgery
4.2.
Message Observation
4.3.
Replaying
4.4.
Message Prioritization and Differentiation
4.5.
Denial of Service
4.6.
Reliability
5.
IANA Considerations
6.
Acknowledgments
7.
Notes to the RFC Editor and Change Log
8.
References
8.1.
Normative
8.2.
Informative
Appendix A.
Applicability to Legacy syslog
A.1.
Method Change
A.2.
Octet-Counting
A.3.
Octet-Stuffing
§
Authors' Addresses
| TOC |
The syslog protocol (Gerhards, R., “The Syslog Protocol,” March 2009.) [RFC5424] is a text-based protocol used to convey event information. Before that standard was produced, syslog messages were being transmitted over UDP. This was described in the INFORMATIONAL document [RFC3164] (Lonvick, C., “The BSD Syslog Protocol,” August 2001.). While there has been no documented standard for transporting syslog messages over TCP, it is widely used in practice and has proven to be quite interoperable among the implementations, with some minor issues in some configurations. While existing implementations interoperate quite well with each other, there are some differences in protocol handling. This document will describe the most commonly used approach and explain how to interoperate with them in a consistent way.
This specification applies to messages transmitted using the [RFC5424] (Gerhards, R., “The Syslog Protocol,” March 2009.) format. Diagram 1 shows how this relates to the other syslog message transport protocols. In this diagram three originators are seen, labeled A, B, and C, along with one collector. Originator A is using the TCP transport which is described in this document. Originator B is using the UDP transport which is described in [RFC5426] (Okmianski, A., “Transmission of Syslog Messages over UDP,” March 2009.). Originator C is using the TLS transport which is described in [RFC5425] (Miao, F., Ma, Y., and J. Salowey, “Transport Layer Security (TLS) Transport Mapping for Syslog,” March 2009.). The collector is shown with the capability to accept all three transports.
A discussion of how this may be applied to legacy syslog as described in [RFC3164] (Lonvick, C., “The BSD Syslog Protocol,” August 2001.) is contained below. Two format options have been observed with legacy syslog being transported over TCP. These are called octet-stuffing and octet-counting and are described in Section 3.3. This specification is written this way, with two format options, in an attempt to ensure that syslog transport receivers can receive and properly interpret messages sent from legacy syslog senders.
To ensure interoperability between syslog senders and receivers, the octet-counting method is REQUIRED in both syslog transport senders and receivers. To have a deployment that can handle the octet-stuffing methods, it is RECOMMENDED that syslog transport senders be able to receive these messages, but it is NOT RECOMMENDED that the octet-stuffing mechanism be used in either senders or receivers.
+---------------------+
| Originator A |
|---------------------|
| syslog application |
| |
|---------------------|
| syslog transport |
| TCP |
+---------------------+
v
|
/ +---------------------+
/ | Originator B |
/ |---------------------|
/ +----------------------+ | syslog application |
/ | Collector | | |
| |----------------------| |---------------------|
| | syslog application | | syslog transport |
| | | | UDP |
| |----------------------| |---------------------|
| | syslog transport | v
| | TCP | TLS | UDP | |
| +----------------------+ |
| ^ ^ ^ |
| | | | |
\ / | \ /
--------- | ------------------
|
|
| +---------------------+
| | Originator C |
| |---------------------|
| | syslog application |
| | |
| |---------------------|
| | syslog transport |
| | TLS |
| +---------------------+
| v
\ /
---------------
Diagram 1. Syslog Layers
There are several advantages to using TCP: flow control, error recovery, reliability, to name a few. One potential disadvantage is the buffering mechanism used by TCP. Ordinarily, TCP decides when enough data has been received from the application to form a segment for transmission. This may be adjusted through timers but still, some application data may wait in a buffer for a relatively long time. Syslog data is not normally time-sensitive but if this delay is a concern, the syslog transport sender may utilize the PUSH Flag to have the sending TCP immediately send all buffered data. [RFC0793] (Postel, J., “Transmission Control Protocol,” September 1981.)
It is still RECOMMENDED to use the TLS transport (Gerhards, R., “The Syslog Protocol,” March 2009.) [RFC5424] to convey syslog messages. This specification is provided to ensure interoperability for transporting syslog over TCP.
| TOC |
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119 (Bradner, S., “Key words for use in RFCs to Indicate Requirement Levels,” March 1997.) [RFC2119].
The terminology defined in Section 3 of [RFC5424] (Gerhards, R., “The Syslog Protocol,” March 2009.) is used throughout this specification. The reader should be familiar with that to follow this discussion.
This document also references devices that use syslog transported over UDP as described in [RFC3164] (Lonvick, C., “The BSD Syslog Protocol,” August 2001.). Devices that continue to use that will be described as legacy syslog senders and receivers in this document.
| TOC |
As described in [RFC5424] (Gerhards, R., “The Syslog Protocol,” March 2009.), syslog is simplex in nature. Traditional implementations of syslog over TCP do not use any backchannel mechanism to convey information to the transport sender, and consequently do not use any application-level acknowledgement for syslog receiver to sender signaling. Reliability and flow control are provided by the capabilities of TCP.
| TOC |
A syslog over TCP session is a TCP connection between a client and a server. The syslog transport sender is the host that sends the original SYN. The syslog transport receiver is the device that receives the original SYN and responds with a SYN+ACK. After initiation, messages are sent from the transport sender to the transport receiver. No application-level data is transmitted from the transport receiver to the transport sender. The roles of transport sender and receiver are fixed once the session is established, and they can not be reversed during the session. However, there can be multiple sessions between two TCP hosts, and for each session the role of transport sender and transport receiver can be different based upon which device initiates the session.
It is valid (but rare) for no messages to be exchanged during a TCP session.
If an error occurs, the peer detecting the error will gracefully close the TCP session, but has no means to notify its remote peer about the state of the peer syslog application.
| TOC |
The peer that intends to act as a syslog transport receiver listens to TCP port <TBD>. The peer that intends to act as the transport sender initiates a TCP session to the syslog transport receiver as specified in [RFC0793] (Postel, J., “Transmission Control Protocol,” September 1981.).
| TOC |
During the message transfer phase, the syslog transport sender sends a stream of messages to the transport receiver. Either of the peers may initiate session closure at any time as specified in Section 3.5 of [RFC0793] (Postel, J., “Transmission Control Protocol,” September 1981.). In practice, this is often seen after a prolonged time of inactivity.
Syslog messages are sent in sequence within a TCP transport stream. One message is encapsulated inside a frame. Syslog transport senders MUST support the octet-counting method and they MAY support the octet-stuffing method. Syslog transport receivers are REQUIRED to support the octet-counting method and are RECOMMENDED to support the octet-stuffing method to promote interoperability with legacy devices that may only use that framing method. Transport senders do not send any notice about the format they use to the transport receiver. However, the format itself enables the transport receiver to detect which framing is used. The syslog transport sender MUST NOT change the format after it has sent the first message. If the format needs to be changed, the TCP session must be concluded and a new session established.
All syslog messages MUST be sent as TCP "data" as per Transmission Control Protocol (Postel, J., “Transmission Control Protocol,” September 1981.) [RFC0793]. The syslog message stream has the following ABNF (Crocker, D. and P. Overell, “Augmented BNF for Syntax Specifications: ABNF,” January 2008.) [RFC5234] definition:
TCP-DATA = *SYSLOG-FRAME
SYSLOG-FRAME = MSG-LEN SP SYSLOG-MSG ; Octet-counting method
SYSLOG-FRAME =/ SYSLOG-MSG TRAILER ; Octet-stuffing method
MSG-LEN = NONZERO-DIGIT *DIGIT
SP = %d32
NONZERO-DIGIT = %d49-57
DIGIT = %d48 / NONZERO-DIGIT
TRAILER = LF | APP-DEFINED
LF = %d10
APP-DEFINED = 1*2OCTET
SYSLOG-MSG is defined in the syslog protocol [RFC5424]
| TOC |
This mode is somewhat similar to the framing used in [RFC5425] (Miao, F., Ma, Y., and J. Salowey, “Transport Layer Security (TLS) Transport Mapping for Syslog,” March 2009.). Here the message length, in octets, is specified as HEADER, followed by SYSLOG-MSG and no trailer.
MSG-LEN is the octet count of the SYSLOG-MSG in the SYSLOG-FRAME. A transport receiver MUST use the message length to delimit a syslog message. There is no upper limit for a message length per se. However, in order to establish a baseline for interoperability, this specification requires that a transport receiver MUST be able to process messages with a length up to and including 2048 octets. Transport receivers SHOULD be able to process messages with lengths up to and including 8192 octets.
A transport receiver MUST assume that octet-counting framing is used if a syslog frame starts with a digit.
| TOC |
In octet-stuffing mode, there is no header, but a trailer is appended after SYSLOG-MSG.
A transport receiver MUST accept the USASCII LF character as a TRAILER. It MAY be configurable to accept other characters. A discussion of this is below.
A transport receiver MUST assume that octet-stuffing framing is used if a syslog frame starts with the USASCII character "<" (%d60).
The octet-stuffing method is NOT RECOMMENDED.
| TOC |
In both of these methods, a modification is made to the original message. This is a temporary transformation performed by the transport sender. According to Section 5 of [RFC5425] (Miao, F., Ma, Y., and J. Salowey, “Transport Layer Security (TLS) Transport Mapping for Syslog,” March 2009.), this temporary transformation MUST be reversed by the transport protocol at the transport receiver so that the relay or collector will see an exact copy of the message generated by the originator or relay.
In the octet-counting method, a count and a space character are prepended to the messages. This is very much like the method described in [RFC5426] (Okmianski, A., “Transmission of Syslog Messages over UDP,” March 2009.). The count and space character MUST be removed by the transport receiver after it has validated that the count is correct.
Similarly, the syslog transport receiver MUST discard the TRAILER as it accepts the packet in the octet-stuffing method.
| TOC |
The SYSLOG session is closed when one of the peers decides to do so. It then initiates a local TCP session closure. It does not notify its remote peer of its intension to close the session, nor does it accept any messages that are still in transit.
| TOC |
Using this specification on an unsecured network is NOT RECOMMENDED. Several syslog security considerations are discussed in [RFC5424] (Gerhards, R., “The Syslog Protocol,” March 2009.) This section focuses on security considerations specific to the syslog transport over TCP. Some of the security issues raised in this section can be mitigated through the use of TLS as defined in [RFC5425] (Miao, F., Ma, Y., and J. Salowey, “Transport Layer Security (TLS) Transport Mapping for Syslog,” March 2009.)
| TOC |
This transport mapping does not provide for strong transport sender authentication. The receiver of the syslog message will not be able to ascertain that the message was indeed sent from the reported sender, or whether the packet was sent from another device. This can also lead to a case of mistaken identity if an inappropriately configured machine sends syslog messages representing itself as another machine.
This transport mapping does not provide protection against syslog message forgery. An attacker can transmit syslog messages (either from the machine from which the messages are purportedly sent or from any other machine) to a receiver.
In one case, an attacker can hide the true nature of an attack amidst many other messages. As an example, an attacker can start generating forged messages indicating a problem on some machine. This can get the attention of the system administrators, who will spend their time investigating the alleged problem. During this time, the attacker could be able to compromise a different machine or a different process on the same machine.
Additionally, an attacker can generate false syslog messages to give untrue indications of the status of systems. As an example, an attacker can stop a critical process on a machine, which could generate a notification of exit. The attacker can subsequently generate a forged notification that the process had been restarted. The system administrators could accept that misinformation and not verify that the process had indeed not been restarted.
| TOC |
This transport mapping does not provide confidentiality of the messages in transit. If syslog messages are in clear text, this is how they will be transferred. In most cases, passing clear-text, human-readable messages is a benefit to the administrators. Unfortunately, an attacker could also be able to observe the human- readable contents of syslog messages. The attacker could then use the knowledge gained from these messages to compromise a machine. It is RECOMMENDED that no sensitive information be transmitted via this transport mapping or that transmission of such information be restricted to properly secured networks.
| TOC |
Message forgery and observation can be combined into a replay attack. An attacker could record a set of messages that indicate normal activity of a machine. At a later time, an attacker could remove that machine from the network and replay the syslog messages with new time stamps. The administrators could find nothing unusual in the received messages, and their receipt would falsely indicate normal activity of the machine.
| TOC |
This transport mapping does not mandate prioritization of syslog messages either on the wire or when processed on the receiving host based on their severity. Unless some prioritization is implemented by sender, receiver, and/or network, the security implication of such behavior is that the syslog receiver or network devices could get overwhelmed with low-severity messages and be forced to discard potentially high-severity messages.
| TOC |
An attacker could overwhelm a receiver by sending more messages to it than could be handled by the infrastructure or the device itself. Implementers SHOULD attempt to provide features that minimize this threat, such as optionally restricting reception of messages to a set of known source IP addresses.
| TOC |
It should be noted that the syslog transport specified in this document does not use application-layer acknowledgments. TCP uses retransmissions to provide protection against some forms of data loss. However, if the TCP connection is broken for some reason (or closed by the transport receiver), the syslog transport sender cannot always know what messages were successfully delivered to the syslog application at the other end.
| TOC |
IANA is requested to provide a TCP port for this protocol.
After that port has been assigned, this section will be revised to list that port.
| TOC |
The authors wish to thank David Harrington, Tom Petch, and all other people who commented on various versions of this proposal.
| TOC |
These are notes to the RFC editor. Please delete this section after the notes have been followed.
Please replace the instances of <TBD> the port number assigned by IANA.
Version -04 was submitted in April, 2010 to clean up some items.
Version -03 was submitted in April, 2010 based upon further review comments from Tom Petch.
Version -02 was submitted in March, 2010 based upon review comments from Tom Petch.
Version -01 was submitted based upon review comments from David Harrington.
Version -00 was created in November, 2009.
| TOC |
| TOC |
| [RFC0793] | Postel, J., “Transmission Control Protocol,” STD 7, RFC 793, September 1981 (TXT). |
| [RFC2119] | Bradner, S., “Key words for use in RFCs to Indicate Requirement Levels,” BCP 14, RFC 2119, March 1997 (TXT, HTML, XML). |
| [RFC5234] | Crocker, D. and P. Overell, “Augmented BNF for Syntax Specifications: ABNF,” STD 68, RFC 5234, January 2008 (TXT). |
| [RFC5424] | Gerhards, R., “The Syslog Protocol,” RFC 5424, March 2009 (TXT). |
| [RFC5425] | Miao, F., Ma, Y., and J. Salowey, “Transport Layer Security (TLS) Transport Mapping for Syslog,” RFC 5425, March 2009 (TXT). |
| [RFC5426] | Okmianski, A., “Transmission of Syslog Messages over UDP,” RFC 5426, March 2009 (TXT). |
| TOC |
| [RFC3164] | Lonvick, C., “The BSD Syslog Protocol,” RFC 3164, August 2001 (TXT). |
| TOC |
This is an informative appendix provided to promote interoperability within the various observed implementations. Even though this specification does not cover legacy syslog messages, the language used here will be consistent with [RFC2119] (Bradner, S., “Key words for use in RFCs to Indicate Requirement Levels,” March 1997.) to be clear in this and to show how the new syslog architecture will interoperate with the legacy implementations.
Syslog over TCP has been around for a number of years. Just like legacy syslog over UDP, several different implementations exist. The older method of octet-stuffing has problems so is NOT RECOMMENDED, but SHOULD be implemented to ensure interoperability with older clients or servers that may only use this method. The newer method of octet-counting is reliable and, as is consistent with this specification, SHOULD be implemented. When implementers do implement both methods, it is RECOMMENDED that the default method be octet-counting.
| TOC |
It has been observed in legacy implementations that the framing may change on a frame-by-frame basis. This behavior is NOT RECOMMENDED. However, for interoperability, a transport receiver wishing to interoperate with these legacy systems SHOULD be prepared to accept different framing for each frame received.
| TOC |
This framing allows for the transmission of all characters inside SYSLOG-MSG. Along these lines, some transport senders have been seen to use this framing to stack multiple messages within a single TCP frame by using a USASCII LF (%d10) character to separate them. As an example, two messages may be placed within a single frame with a LF character between them. This behavior is NOT RECOMMENDED although it has been observed. Since it cannot be known in advance if the legacy transport sender is going to send multiple messages within single frames, it MUST be assumed that they WILL NOT, and that each frame using this method contains only one syslog message. Implementers MAY attempt to address this by looking for TRAILER characters within each frame to try to separate multiple messages. However, this may lead to more problems than it resolves.
A transport receiver MUST assume that octet-counting framing is used if a syslog frame starts with a digit.
| TOC |
The problem with octet-stuffing framing comes from the use of [RFC3164] (Lonvick, C., “The BSD Syslog Protocol,” August 2001.) messages. In that, the traditional trailer character is not escaped within SYSLOG-MSG which causes problems for the receiver. For example, a message in the style of [RFC3164] (Lonvick, C., “The BSD Syslog Protocol,” August 2001.) containing one or more LF characters may be misinterpreted as multiple messages by the transport receiver. There is no method to avoid this problem with the octet-stuffing framing.
In this legacy implementation, the TRAILER consists of a single character and most often is the USASCII LF (%d10) character. However, other characters have also been seen occasionally, with USACII NUL (%d00) being a prominent example. Some devices also emit a two-character TRAILER, which is usually CR and LF.
Transport senders MUST support the option to use the USASCII LF character. Transport receivers MUST also support this. Transport senders and receivers MAY also support other characters.
| TOC |
| Rainer Gerhards | |
| Adiscon GmbH | |
| Mozartstrasse 21 | |
| Grossrinderfeld, BW 97950 | |
| Germany | |
| Email: | rgerhards@adiscon.com |
| Chris Lonvick | |
| Cisco Systems, Inc | |
| 12515 Research Blvd. | |
| Austin, TX 78759 | |
| USA | |
| Email: | clonvick@cisco.com |