RFC 4172
iFCP - A Protocol for Internet Fibre Channel Storage Networking
4. The iFCP Network Model
The iFCP protocol enables the implementation of fibre channel fabric functionality on an IP network in which IP components and technology replace the fibre channel switching and routing infrastructure described in Section 3.2. The example of Figure 6 shows a fibre channel network with attached devices. Each device accesses the network through an N_PORT connected to an interface whose behavior is specified in [FC-FS] or [FC-AL2]. In this case, the N_PORT represents any of the variants described in Section 3.2. The interface to the fabric may be an L_PORT, F_PORT, or FL_PORT. Within the fibre channel device domain, addressable entities consist of other N_PORTs and fibre channel devices internal to the network that perform the fabric services defined in [FC-GS3].
Fibre Channel Network +--------+ +--------+ | FC | | FC | | Device | | Device | |........| FC |........| Fibre Channel | N_PORT |<......>| N_PORT | Device Domain +---+----+ Traffic+----+---+ ^ | | | +---+----+ +----+---+ | | Fabric | | Fabric | | | Port | | Port | | ==========+========+========+========+============== | FC Network & | | | Fabric Services | v | | Fibre Channel +--------------------------+ Network Domain Figure 6. A Fibre Channel Network Gateway Region Gateway Region +--------+ +--------+ +--------+ +--------+ | FC | | FC | | FC | | FC | | Device | | Device | | Device | | Device | Fibre |........| |........| FC |........| |........| Channel | N_PORT | | N_PORT |<.........>| N_PORT | | N_PORT | Device +---+----+ +---+----+ Traffic +----+---+ +----+---+ Domain | | | | ^ +---+----+ +---+----+ +----+---+ +----+---+ | | F_PORT | | F_PORT | | F_PORT | | F_PORT | | =+========+==+========+===========+========+==+========+========== | iFCP Layer |<--------->| iFCP Layer | | |....................| ^ |....................| | | iFCP Portal | | | iFCP Portal | v +--------+-----------+ | +----------+---------+ IP iFCP|Gateway Control iFCP|Gateway Network | Data | | | | | |<------Encapsulated Frames------->| | +------------------+ | | | | | +------+ IP Network +--------+ | | +------------------+ Figure 7. An iFCP Fabric Example
One example of an equivalent iFCP fabric is shown in Figure 7. The fabric consists of two gateway regions, each accessed by a single iFCP gateway. Each gateway contains two standards-compliant F_PORTs and an iFCP Portal for attachment to the IP network. Fibre channel devices in the region are those locally connected to the iFCP fabric through the gateway fabric ports. Looking into the fabric port, the gateway appears as a fibre channel switch element. At this interface, remote N_PORTs are presented as fabric-attached devices. Conversely, on the IP network side, the gateway presents each locally connected N_PORT as a logical fibre channel device. Extrapolating to the general case, each gateway region behaves like an autonomous system whose configuration is invisible to the IP network and other gateway regions. Consequently, in addition to the F_PORT shown in the example, a gateway implementation may transparently support the following fibre channel interfaces: Inter-Switch Link -- A fibre channel switch-to-switch interface used to access a region containing fibre channel switch elements. An implementation may support the E_PORT defined by [FC-SW2] or one of the proprietary interfaces provided by various fibre channel switch vendors. In this case, the gateway acts as a border switch connecting the gateway region to the IP network. FL_PORT -- An interface that provides fabric access for loop- attached fibre channel devices, as specified in [FC-FLA]. L_PORT -- An interface through which a gateway may emulate the fibre channel loop environment specified in [FC-AL2]. As discussed in appendix B, the gateway presents remotely accessed N_PORTS as loop-attached devices. The manner in which these interfaces are provided by a gateway is implementation specific and therefore beyond the scope of this document. Although each region is connected to the IP network through one gateway, a region may incorporate multiple gateways for added performance and fault tolerance if the following conditions are met: a) The gateways MUST coordinate the assignment of N_PORT IDs and aliases so that each N_PORT has one and only one address.
b) All iFCP traffic between a given remote and local N_PORT pair MUST
flow through the same iFCP session (see Section 5.2.1). However,
iFCP sessions to a given remotely attached N_PORT need not
traverse the same gateway.
Coordinating address assignments and managing the flow of traffic is
implementation specific and outside the scope of this specification.
4.1. iFCP Transport Services
N_PORT to N_PORT communications that traverse a TCP/IP network
require the intervention of the iFCP layer within the gateway. This
consists of the following operations:
a) Execution of the frame-addressing and -mapping functions described
in Section 4.4.
b) Encapsulation of fibre channel frames for injection into the
TCP/IP network and de-encapsulation of fibre channel frames
received from the TCP/IP network.
c) Establishment of an iFCP session in response to a PLOGI directed
to a remote device.
Section 4.4 discusses the iFCP frame-addressing mechanism and the way
that it is used to achieve communications transparency between
N_PORTs.
4.1.1. Fibre Channel Transport Services Supported by iFCP
An iFCP fabric supports Class 2 and Class 3 fibre channel transport
services, as specified in [FC-FS]. An iFCP fabric does not support
Class 4, Class 6, or Class 1 (dedicated connection) service. An
N_PORT discovers the classes of transport services supported by the
fabric during fabric login.
4.2. iFCP Device Discovery and Configuration Management
An iFCP implementation performs device discovery and iFCP fabric
management through the Internet Storage Name Service defined in
[ISNS]. Access to an iSNS server is required to perform the
following functions:
a) Emulate the services provided by the fibre channel name server
described in Section 3.3.1, including a mechanism for
asynchronously notifying an N_PORT of changes in the iFCP fabric
configuration.
b) Aggregate gateways into iFCP fabrics for interoperation.
c) Segment an iFCP fabric into fibre channel zones through the
definition and management of device discovery scopes, referred to
as 'discovery domains'.
d) Store and distribute security policies, as described in Section
10.2.9.
e) Implementation of the fibre channel broadcast mechanism.
4.3. iFCP Fabric Properties
A collection of iFCP gateways may be configured for interoperation as
either a bounded or an unbounded iFCP fabric.
Gateways in a bounded iFCP fabric operate in address transparent
mode, as described in Section 4.5. In this mode, the scope of a
fibre channel N_PORT address is fabric-wide and is derived from
domain IDs issued by the iSNS server from a common pool. As
discussed in Section 4.3.2, the maximum number of domain IDs allowed
by the fibre channel limits the configuration of a bounded iFCP
fabric.
Gateways in an unbounded iFCP fabric operate in address translation
mode as described in Section 4.6. In this mode, the scope of an
N_PORT address is local to a gateway region. For fibre channel
traffic between regions, the translation of frame-embedded N_PORT
addresses is performed by the gateway. As discussed below, the
number of switch elements and gateways in an unbounded iFCP fabric
may exceed the limits of a conventional fibre channel fabric.
All iFCP gateways MUST support unbounded iFCP fabrics. Support for
bounded iFCP fabrics is OPTIONAL.
The decision to support bounded iFCP fabrics in a gateway
implementation depends on the address transparency, configuration
scalability, and fault tolerance considerations given in the
following sections.
4.3.1. Address Transparency
Although iFCP gateways in an unbounded fabric will convert N_PORT
addresses in the frame header and payload of standard link service
messages, a gateway cannot convert such addresses in the payload of
vendor- or user-specific fibre channel frame traffic.
Consequently, although both bounded and unbounded iFCP fabrics support standards-compliant FC-4 protocol implementations and link services used by mainstream fibre channel applications, a bounded iFCP fabric may also support vendor- or user-specific protocol and link service implementations that carry N_PORT IDs in the frame payload.4.3.2. Configuration Scalability
The scalability limits of a bounded fabric configuration are a consequence of the fibre channel address allocation policy discussed in Section 3.7.1. As noted, a bounded iFCP fabric using this address allocation scheme is limited to a combined total of 239 gateways and fibre channel switch elements. As the system expands, the network may grow to include many switch elements and gateways, each of which controls a small number of devices. In this case, the limitation in switch and gateway count may become a barrier to extending and fully integrating the storage network. Since N_PORT fibre channel addresses in an unbounded iFCP fabric are not fabric-wide, the limits imposed by fibre channel address allocation only apply within the gateway region. Across regions, the number of iFCP gateways, fibre channel devices, and switch elements that may be internetworked are not constrained by these limits. In exchange for improved scalability, however, implementations must consider the incremental overhead of address conversion, as well as the address transparency issues discussed in Section 4.3.1.4.3.3. Fault Tolerance
In a bounded iFCP fabric, address reassignment caused by a fault or reconfiguration, such as the addition of a new gateway region, may cascade to other regions, causing fabric-wide disruption as new N_PORT addresses are assigned. Furthermore, before a new gateway can be merged into the fabric, its iSNS server must be slaved to the iSNS server in the bounded fabric to centralize the issuance of domain IDs. In an unbounded iFCP fabric, coordinating the iSNS databases requires only that the iSNS servers exchange client attributes with one another. A bounded iFCP fabric also has an increased dependency on the availability of the iSNS server, which must act as the central address assignment authority. If connectivity with the server is lost, new DOMAIN_ID values cannot be automatically allocated as gateways and fibre channel switch elements are added.
4.4. The iFCP N_PORT Address Model
This section discusses iFCP extensions to the fibre channel addressing model of Section 3.7.1, which are required for the transparent routing of frames between locally and remotely attached N_PORTs. In the iFCP protocol, an N_PORT is represented by the following addresses: a) A 24-bit N_PORT ID. The fibre channel N_PORT address of a locally attached device. Depending on the gateway addressing mode, the scope is local either to a region or to a bounded iFCP fabric. In either mode, communications between N_PORTs in the same gateway region use the N_PORT ID. b) A 24-bit N_PORT alias. The fibre channel N_PORT address assigned by each gateway operating in address translation mode to identify a remotely attached N_PORT. Frame traffic is intercepted by an iFCP gateway and directed to a remotely attached N_PORT by means of the N_PORT alias. The address assigned by each gateway is unique within the scope of the gateway region. c) An N_PORT network address. A tuple consisting of the gateway IP address, TCP port number, and N_PORT ID. The N_PORT network address identifies the source and destination N_PORTs for fibre channel traffic on the IP network. To provide transparent communications between a remote and local N_PORT, a gateway MUST maintain an iFCP session descriptor (see Section 5.2.2.2) reflecting the association between the fibre channel address representing the remote N_PORT and the remote device's N_PORT network address. To establish this association, the iFCP gateway assigns and manages fibre channel N_PORT fabric addresses as described in the following paragraphs. In an iFCP fabric, the iFCP gateway performs the address assignment and frame routing functions of an FC switch element. Unlike an FC switch, however, an iFCP gateway must also direct frames to external devices attached to remote gateways on the IP network. In order to be transparent to FC devices, the gateway must deliver such frames using only the 24-bit destination address in the frame header. By exploiting its control of address allocation and access to frame traffic entering or leaving the gateway region, the gateway is able to achieve the necessary transparency.
N_PORT addresses within a gateway region may be allocated in one of
two ways:
a) Address Translation Mode - A mode of N_PORT address assignment in
which the scope of an N_PORT fibre channel address is unique to
the gateway region. The address of a remote device is represented
in that gateway region by its gateway-assigned N_PORT alias.
b) Address Transparent Mode - A mode of N_PORT address assignment in
which the scope of an N_PORT fibre channel address is unique
across the set of gateway regions comprising a bounded iFCP
fabric.
In address transparent mode, gateways within a bounded fabric
cooperate in the assignment of addresses to locally attached N_PORTs.
Each gateway in control of a region is responsible for obtaining and
distributing unique domain IDs from the address assignment authority,
as described in Section 4.5.1. Consequently, within the scope of a
bounded fabric, the address of each N_PORT is unique. For that
reason, gateway-assigned aliases are not required for representing
remote N_PORTs.
All iFCP implementations MUST support operations in address
translation mode. Implementation of address transparent mode is
OPTIONAL but, of course, must be provided if bounded iFCP fabric
configurations are to be supported.
The mode of gateway operation is settable in an implementation-
specific manner. The implementation MUST NOT:
a) allow the mode to be changed after the gateway begins processing
fibre channel frame traffic,
b) permit operation in more than one mode at a time, or
c) establish an iFCP session with a gateway that is not in the same
mode.
4.5. Operation in Address Transparent Mode
The following considerations and requirements apply to this mode of
operation:
a) iFCP gateways in address transparent mode will not interoperate
with iFCP gateways that are not in address transparent mode.
b) When interoperating with locally attached fibre channel switch
elements, each iFCP gateway MUST assume control of DOMAIN_ID
assignments in accordance with the appropriate fibre channel
standard or vendor-specific protocol specification. As described
in Section 4.5.1, DOMAIN_ID values that are assigned to FC
switches internal to the gateway region must be issued by the iSNS
server.
c) When operating in address transparent Mode, fibre channel address
translation SHALL NOT take place.
When operating in address transparent mode, however, the gateway MUST
establish and maintain the context of each iFCP session in accordance
with Section 5.2.2.
4.5.1. Transparent Mode Domain ID Management
As described in Section 4.5, each gateway and fibre channel switch in
a bounded iFCP fabric has a unique domain ID. In a gateway region
containing fibre channel switch elements, each element obtains a
domain ID by querying the principal switch as described in [FC-SW2]
-- in this case, the iFCP gateway itself. The gateway, in turn,
obtains domain IDs on demand from the iSNS name server acting as the
central address allocation authority. In effect, the iSNS server
assumes the role of principal switch for the bounded fabric. In that
case, the iSNS database contains:
a) The definition for one or more bounded iFCP fabrics, and
b) For each bounded fabric, a worldwide-unique name identifying each
gateway in the fabric. A gateway in address transparent mode MUST
reside in one, and only one, bounded fabric.
As the Principal Switch within the gateway region, an iFCP gateway in
address transparent mode SHALL obtain domain IDs for use in the
gateway region by issuing the appropriate iSNS query, using its
worldwide name.
4.5.2. Incompatibility with Address Translation Mode
Except for the session control frames specified in Section 6, iFCP
gateways in address transparent mode SHALL NOT originate or accept
frames that do not have the TRP bit set to one in the iFCP flags
field of the encapsulation header (see Section 5.3.1). The iFCP
gateway SHALL immediately terminate all iFCP sessions with the iFCP
gateway from which it receives such frames.
4.6. Operation in Address Translation Mode
This section describes the process for managing the assignment of addresses within a gateway region that is part of an unbounded iFCP fabric, including the modification of FC frame addresses embedded in the frame header for frames sent and received from remotely attached N_PORTs. As described in Section 4.4, the scope of N_PORT addresses in this mode is local to the gateway region. A principal switch within the gateway region, possibly the iFCP gateway itself, oversees the assignment of such addresses, in accordance with the rules specified in [FC-FS] and [FC-FLA]. The assignment of N_PORT addresses to locally attached devices is controlled by the switch element to which the device is connected. The assignment of N_PORT addresses for remotely attached devices is controlled by the gateway by which the remote device is accessed. In this case, the gateway MUST assign a locally significant N_PORT alias to be used in place of the N_PORT ID assigned by the remote gateway. The N_PORT alias is assigned during device discovery, as described in Section 5.2.2.1. To perform address conversion and to enable the appropriate routing, the gateway MUST establish an iFCP session and generate the information required to map each N_PORT alias to the appropriate TCP/IP connection context and N_PORT ID of the remotely accessed N_PORT. These mappings are created and updated by means specified in Section 5.2.2.2. As described in that section, the required mapping information is represented by the iFCP session descriptor reproduced in Figure 8. +-----------------------+ |TCP Connection Context | +-----------------------+ | Local N_PORT ID | +-----------------------+ | Remote N_PORT ID | +-----------------------+ | Remote N_PORT Alias | +-----------------------+ Figure 8. iFCP Session Descriptor (from Section 5.2.2.2)
Except for frames comprising special link service messages (see Section 7.2), outbound frames are encapsulated and sent without modification. Address translation is deferred until receipt from the IP network, as specified in Section 4.6.1.4.6.1. Inbound Frame Address Translation
For inbound frames received from the IP network, the receiving gateway SHALL reference the session descriptor to fill in the D_ID field with the destination N_PORT ID and the S_ID field with the N_PORT alias it assigned. The translation process for inbound frames is shown in Figure 9. Network Format of Inbound Frame +--------------------------------------------+ iFCP | FC Encapsulation Header | Session +--------------------------------------------+ Descriptor | SOF Delimiter Word | | +========+===================================+ V | | D_ID Field | +--------+-----+ +--------+-----------------------------------+ | Lookup source| | | S_ID Field | | N_PORT Alias | +--------+-----------------------------------+ | and | | Control Information, Payload, | | destination | | and FC CRC | | N_PORT ID | | | +--------+-----+ | | | | | | +============================================+ | | EOF Delimiter Word | | +--------------------------------------------+ | | | Frame after Address Translation and De-encapsulation | +--------+-----------------------------------+ | | | Destination N_PORT ID |<-------------+ +--------+-----------------------------------+ | | | Source N_PORT Alias |<-------------+ +--------+-----------------------------------+ | | | Control information, Payload, | | and FC CRC | +--------------------------------------------+ Figure 9. Inbound Frame Address Translation
The receiving gateway SHALL consider the contents of the S_ID and D_ID fields to be undefined when received. After replacing these fields, the gateway MUST recalculate the FC CRC.4.6.2. Incompatibility with Address Transparent Mode
iFCP gateways in address translation mode SHALL NOT originate or accept frames that have the TRP bit set to one in the iFCP flags field of the encapsulation header. The iFCP gateway SHALL immediately abort all iFCP sessions with the iFCP gateway from which it receives frames such as those described in Section 5.2.3.5. iFCP Protocol
5.1. Overview
5.1.1. iFCP Transport Services
The main function of the iFCP protocol layer is to transport fibre channel frame images between locally and remotely attached N_PORTs. When transporting frames to a remote N_PORT, the iFCP layer encapsulates and routes the fibre channel frames comprising each fibre channel Information Unit via a predetermined TCP connection for transport across the IP network. When receiving fibre channel frame images from the IP network, the iFCP layer de-encapsulates and delivers each frame to the appropriate N_PORT. The iFCP layer processes the following types of traffic: a) FC-4 frame images associated with a fibre channel application protocol. b) FC-2 frames comprising fibre channel link service requests and responses. c) Fibre channel broadcast frames. d) iFCP control messages required to set up, manage, or terminate an iFCP session. For FC-4 N_PORT traffic and most FC-2 messages, the iFCP layer never interprets the contents of the frame payload. iFCP does interpret and process iFCP control messages and certain link service messages, as described in Section 5.1.2.
5.1.2. iFCP Support for Link Services
iFCP must intervene in the processing of those fibre channel link service messages that contain N_PORT addresses in the message payload or that require other special handling, such as an N_PORT login request (PLOGI). In the former case, an iFCP gateway operating in address translation mode MUST supplement the payload with additional information that enables the receiving gateway to convert such embedded N_PORT addresses to its frame of reference. For out bound fibre channel frames comprising such a link service, the iFCP layer creates the supplemental information based on frame content, modifies the frame payload, and then transmits the resulting fibre channel frame with supplemental data through the appropriate TCP connection. For incoming iFCP frames containing supplemented fibre channel link service frames, iFCP must interpret the frame, including any supplemental information, modify the frame content, and forward the resulting frame to the destination N_PORT for further processing. Section 7.1 describes the processing of these link service messages in detail.5.2. TCP Stream Transport of iFCP Frames
5.2.1. iFCP Session Model
An iFCP session consists of the pair of N_PORTs comprising the session endpoints joined by a single TCP/IP connection. No more than one iFCP session SHALL exist between a given pair of N_PORTs. An N_PORT is identified by its network address, consisting of: a) the N_PORT ID assigned by the gateway to which the N_PORT is locally attached, and b) the iFCP Portal address, consisting of its IP address and TCP port number. Because only one iFCP session may exist between a pair of N_PORTs, the iFCP session is uniquely identified by the network addresses of the session end points. TCP connections that may be used for iFCP sessions between pairs of iFCP portals are either "bound" or "unbound". An unbound connection
is a TCP connection that is not actively supporting an iFCP session. A gateway implementation MAY establish a pool of unbound connections to reduce the session setup time. Such pre-existing TCP connections between iFCP Portals remain unbound and uncommitted until allocated to an iFCP session through a CBIND message (see Section 6.1). When the iFCP layer creates an iFCP session, it may select an existing unbound TCP connection or establish a new TCP connection and send the CBIND message down that TCP connection. This allocates the TCP connection to that iFCP session.5.2.2. iFCP Session Management
This section describes the protocols and data structures required to establish and terminate an iFCP session.5.2.2.1. The Remote N_PORT Descriptor
In order to establish an iFCP session, an iFCP gateway MUST maintain information allowing it to locate a remotely attached N_PORT. For explanatory purposes, such information is assumed to reside in a descriptor with the format shown in Figure 10. +--------------------------------+ | N_PORT Worldwide Unique Name | +--------------------------------+ | iFCP Portal Address | +--------------------------------+ | N_PORT ID of Remote N_PORT | +--------------------------------+ | N_PORT Alias | +--------------------------------+ Figure 10. Remote N_PORT Descriptor Each descriptor aggregates the following information about a remotely attached N_PORT: N_PORT Worldwide Unique Name -- 64-bit N_PORT worldwide name as specified in [FC-FS]. A Remote N_PORT descriptor is uniquely identified by this parameter. iFCP Portal Address -- The IP address and TCP port number referenced when creation of the TCP connection associated with an iFCP session is requested. N_PORT ID -- N_PORT fibre channel address assigned to the remote device by the remote iFCP gateway.
N_PORT Alias -- N_PORT fibre channel address assigned to the
remote device by the 'local' iFCP gateway when it operates in
address translation mode.
An iFCP gateway SHALL have one and only one descriptor for each
remote N_PORT it accesses. If a descriptor does not exist, one SHALL
be created using the information returned by an iSNS name server
query. Such queries may result from:
a) a fibre channel Name Server request originated by a locally
attached N_PORT (see Sections 3.5 and 9.3), or
b) a CBIND request received from a remote fibre channel device (see
Section 5.2.2.2).
When creating a descriptor in response to an incoming CBIND request,
the iFCP gateway SHALL perform an iSNS name server query using the
worldwide port name of the remote N_PORT in the SOURCE N_PORT NAME
field within the CBIND payload. The descriptor SHALL be filled in
using the query results.
After creating the descriptor, a gateway operating in address
translation mode SHALL create and add the 24-bit N_PORT alias.
5.2.2.1.1. Updating a Remote N_PORT Descriptor
A Remote N_PORT descriptor SHALL only be updated as the result of an
iSNS query to obtain information for the specified worldwide port
name or from information returned by an iSNS state change
notification. Following such an update, a new N_PORT alias SHALL NOT
be assigned.
Before such an update, the contents of a descriptor may have become
stale because of an event that invalidated or triggered a change in
the N_PORT network address of the remote device, such as a fabric
reconfiguration or the device's removal or replacement.
A collateral effect of such an event is that a fibre channel device
that has been added or whose N_PORT ID has changed will have no
active N_PORT logins. Consequently, FC-4 traffic directed to such an
N_PORT, because of a stale descriptor, will be rejected or discarded.
Once the originating N_PORT learns of the reconfiguration, usually
through the name server state change notification mechanism,
information returned in the notification or the subsequent name
server lookup needed to reestablish the iFCP session will
automatically purge such stale data from the gateway.
5.2.2.1.2. Deleting a Remote N_PORT Descriptor
Deleting a remote N_PORT descriptor is equivalent to freeing up the corresponding N_PORT alias for reuse. Consequently, the descriptor MUST NOT be deleted while there are any iFCP sessions that reference the remote N_PORT. Descriptors eligible for deletion should be removed based on a last in, first out policy.5.2.2.2. Creating an iFCP Session
An iFCP session may be in one of the following states: OPEN -- The session state in which fibre channel frame images may be sent and received. OPEN PENDING -- The session state after a gateway has issued a CBIND request but no response has yet been received. No fibre channel frames may be sent. The session may be initiated in response to a PLOGI ELS (see Section 7.3.1.7) or for any other implementation-specific reason. The gateway SHALL create the iFCP session as follows: a) Locate the remote N_PORT descriptor corresponding to the session end point. If the session is created in order to forward a fibre channel frame, then the session endpoint may be obtained by referencing the remote N_PORT alias contained in the frame header D_ID field. If no descriptor exists, an iFCP session SHALL NOT be created. b) Allocate a TCP connection to the gateway to which the remote N_PORT is locally attached. An implementation may use an existing connection in the Unbound state, or a new connection may be created and placed in the Unbound state. When a connection is created, the IP address and TCP Port number SHALL be obtained by referencing the remote N_PORT descriptor as specified in Section 5.2.2.1. c) If the TCP connection cannot be allocated or cannot be created due to limited resources, the gateway SHALL terminate session creation.
d) If the TCP connection is aborted for any reason before the iFCP
session enters the OPEN state, the gateway SHALL respond in
accordance with Section 5.2.3 and MAY terminate the attempt to
create a session or MAY try to establish the TCP connection again.
e) The gateway SHALL then issue a CBIND session control message (see
Section 6.1) and place the session in the OPEN PENDING state.
f) If a CBIND response is returned with a status other than "Success"
or "iFCP session already exists", the session SHALL be terminated,
and the TCP connection returned to the Unbound state.
g) A CBIND STATUS of "iFCP session already exists" indicates that the
remote gateway has concurrently initiated a CBIND request to
create an iFCP session between the same pair of N_PORTs. A
gateway receiving such a response SHALL terminate this attempt and
process the incoming CBIND request in accordance with Section
5.2.2.3.
h) In response to a CBIND STATUS of "Success", the gateway SHALL
place the session in the OPEN state.
Once the session is placed in the OPEN state, an iFCP session
descriptor SHALL be created, containing the information shown in
Figure 11:
+-----------------------+
|TCP Connection Context |
+-----------------------+
| Local N_PORT ID |
+-----------------------+
| Remote N_PORT ID |
+-----------------------+
| Remote N_PORT Alias |
+-----------------------+
Figure 11. iFCP Session Descriptor
TCP Connection Context -- Information required to identify the TCP
connection associated with the iFCP session.
Local N_PORT ID -- N_PORT ID of the locally attached fibre
channel device.
Remote N_PORT ID -- N_PORT ID assigned to the remote device by the
remote gateway.
Remote N_PORT Alias -- Alias assigned to the remote N_PORT by the
local gateway when it operates in address translation mode. If in
this mode, the gateway SHALL copy this parameter from the Remote
N_PORT descriptor. Otherwise, it is not filled in.
5.2.2.3. Responding to a CBIND Request
The gateway receiving a CBIND request SHALL respond as follows:
a) If the receiver has a duplicate iFCP session in the OPEN PENDING
state, then the receiving gateway SHALL compare the Source N_PORT
Name in the incoming CBIND payload with the Destination N_PORT
Name.
b) If the Source N_PORT Name is greater, the receiver SHALL issue a
CBIND response of "Success" and SHALL place the session in the
OPEN state.
c) If the Source N_PORT Name is less, the receiver shall issue a
CBIND RESPONSE of Failed - N_PORT session already exists. The
state of the receiver-initiated iFCP session SHALL BE unchanged.
d) If there is no duplicate iFCP session in the OPEN PENDING state,
the receiving gateway SHALL issue a CBIND response. If a status
of Success is returned, the receiving gateway SHALL create the
iFCP session and place it in the OPEN state. An iFCP session
descriptor SHALL be created as described in Section 5.2.2.2.
e) If a remote N_PORT descriptor does not exist, one SHALL be created
and filled in as described in Section 5.2.2.1.
5.2.2.4. Monitoring iFCP Connectivity
During extended periods of inactivity, an iFCP session may be
terminated due to a hardware failure within the gateway or through
loss of TCP/IP connectivity. The latter may occur when the session
traverses a stateful intermediate device, such as a NA(P)T box or
firewall, that detects and purges connections it believes are unused.
To test session liveness, expedite the detection of connectivity
failures, and avoid spontaneous connection termination, an iFCP
gateway may maintain a low level of session activity and monitor the
session by requesting that the remote gateway periodically transmit
the LTEST message described in Section 6.3. All iFCP gateways SHALL
support liveness testing as described in this specification.
A gateway requests the LTEST heartbeat by specifying a non-zero value for the LIVENESS TEST INTERVAL in the CBIND request or response message as described in Section 6.1. If both gateways seek to monitor liveness, each must set the LIVENESS TEST INTERVAL in the CBIND request or response. Upon receiving such a request, the gateway providing the heartbeat SHALL transmit LTEST messages at the specified interval. The first message SHALL be sent as soon as the iFCP session enters the OPEN state. LTEST messages SHALL NOT be sent when the iFCP session is not in the OPEN state. An iFCP session SHALL be terminated as described in Section 5.2.3 if: a) the contents of the LTEST message are incorrect, or b) an LTEST message is not received within twice the specified interval or the iFCP session has been quiescent for longer than twice the specified interval. The gateway to receive the LTEST message SHALL measure the interval for the first expected LTEST message from when the session is placed in the OPEN state. Thereafter, the interval SHALL be measured relative to the last LTEST message received. To maximize liveness test coverage, LTEST messages SHOULD flow through all the gateway components used to enter and retrieve fibre channel frames from the IP network, including the mechanisms for encapsulating and de-encapsulating fibre channel frames. In addition to monitoring a session, information in the LTEST message encapsulation header may also be used to compute an estimate of network propagation delay, as described in Section 8.2.1. However, the propagation delay limit SHALL NOT be enforced for LTEST traffic.5.2.2.5. Use of TCP Features and Settings
This section describes ground rules for the use of TCP features in an iFCP session. The core TCP protocol is defined in [RFC793]. TCP implementation requirements and guidelines are specified in [RFC1122].
+-----------+------------+--------------+------------+------------+ | Feature | Applicable | RFC | Peer-Wise | Requirement| | | RFCs | Status | Agreement | Level | | | | | Required? | | +===========+============+==============+============+============+ | Keep Alive| [RFC1122] | None | No | Should not | | |(discussion)| | | use | +-----------+------------+--------------+------------+------------+ | Tiny | [RFC896] | Standard | No | Should not | | Segment | | | | use | | Avoidance | | | | | | (Nagle) | | | | | +-----------+------------+--------------+------------+------------+ | Window | [RFC1323] | Proposed | No | Should use | | Scale | | Standard | | | +-----------+------------+--------------+------------+------------+ | Wrapped | [RFC1323] | Proposed | No | SHOULD use | | Sequence | | Standard | | | | Protection| | | | | | (PAWS) | | | | | +-----------+------------+--------------+------------+------------+ Table 1. Usage of Optional TCP Features The following sections describe these options in greater detail.5.2.2.5.1. Keep Alive
Keep Alive speeds the detection and cleanup of dysfunctional TCP connections by sending traffic when a connection would otherwise be idle. The issues are discussed in [RFC1122]. In order to test the device more comprehensively, fibre channel applications, such as storage, may implement an equivalent keep alive function at the FC-4 level. Alternatively, periodic liveness test messages may be issued as described in Section 5.2.2.4. Because of these more comprehensive end-to-end mechanisms and the considerations described in [RFC1122], keep alive at the transport layer should not be implemented.5.2.2.5.2. 'Tiny' Segment Avoidance (Nagle)
The Nagle algorithm described in [RFC896] is designed to avoid the overhead of small segments by delaying transmission in order to agglomerate transfer requests into a large segment. In iFCP, such small transfers often contain I/O requests. The transmission delay of the Nagle algorithm may decrease I/O throughput. Therefore, the Nagle algorithm should not be used.
5.2.2.5.3. Window Scale
Window scaling, as specified in [RFC1323], allows full use of links with large bandwidth - delay products and should be supported by an iFCP implementation.5.2.2.5.4. Wrapped Sequence Protection (PAWS)
TCP segments are identified with 32-bit sequence numbers. In networks with large bandwidth - delay products, it is possible for more than one TCP segment with the same sequence number to be in flight. In iFCP, receipt of such a sequence out of order may cause out-of-order frame delivery or data corruption. Consequently, this feature SHOULD be supported as described in [RFC1323].
5.2.3. Terminating iFCP Sessions
iFCP sessions SHALL be terminated in response to one of the events in Table 2: +-------------------------------------------+---------------------+ | Event | iFCP Sessions | | | to Terminate | +===========================================+=====================+ | PLOGI terminated with LS_RJT response | Peer N_PORT | +-------------------------------------------+---------------------+ | State change notification indicating | All iFCP Sessions | | N_PORT removal or reconfiguration. | from the | | | reconfigured N_PORT | +-------------------------------------------+---------------------+ | LOGO ACC response from peer N_PORT | Peer N_PORT | +-------------------------------------------+---------------------+ | ACC response to LOGO ELS sent to F_PORT | All iFCP sessions | | server (D_ID = 0xFF-FF-FE) (fabric | from the originating| | logout) | N_PORT | +-------------------------------------------+---------------------+ | Implicit N_PORT LOGO as defined in | All iFCP sessions | | [FC-FS] | from the N_PORT | | | logged out | +-------------------------------------------+---------------------+ | LTEST Message Error (see Section 5.2.2.4) | Peer N_PORT | +-------------------------------------------+---------------------+ | Non fatal encapsulation error as | Peer N_PORT | | specified in Section 5.3.3 | | +-------------------------------------------+---------------------+ | Failure of the TCP connection associated | Peer N_PORT | | with the iFCP session | | +-------------------------------------------+---------------------+ | Receipt of an UNBIND session control | Peer N_PORT | | message | | +-------------------------------------------+---------------------+ | Gateway enters the Unsynchronized state | All iFCP sessions | | (see Section 8.2.1) | | +-------------------------------------------+---------------------+ | Gateway detects incorrect address mode | All iFCP sessions | | to peer gateway(see Section 4.6.2) | with peer gateway | +-------------------------------------------+---------------------+ Table 2. Session Termination Events
If a session is being terminated due to an incorrect address mode
with the peer gateway, the TCP connection SHALL be aborted by means
of a connection reset (RST) without performing an UNBIND. Otherwise,
if the TCP connection is still open following the event, the gateway
SHALL shut down the connection as follows:
a) Stop sending fibre channel frames over the TCP connection.
b) Discard all incoming traffic, except for an UNBIND session control
message.
c) If an UNBIND message is received at any time, return a response in
accordance with Section 6.2.
d) If session termination was not triggered by an UNBIND message,
issue the UNBIND session control message, as described in Section
6.2.
e) If the UNBIND message completes with a status of Success, the TCP
connection MAY remain open at the discretion of either gateway and
may be kept in a pool of unbound connections in order to speed up
the creation of a new iFCP session.
If the UNBIND fails for any reason, the TCP connection MUST be
terminated. In this case, the connection SHOULD be aborted with a
connection reset (RST).
For each terminated session, the session descriptor SHALL be deleted.
If a session was terminated by an event other than an implicit LOGO
or a LOGO ACC response, the gateway shall issue a LOGO to the locally
attached N_PORT on behalf of the remote N_PORT.
To recover resources, either gateway may spontaneously close an
unbound TCP connection at any time. If a gateway terminates a
connection with a TCP close operation, the peer gateway MUST respond
by executing a TCP close.
5.3. Fibre Channel Frame Encapsulation
This section describes the iFCP encapsulation of fibre channel
frames. The encapsulation complies with the common encapsulation
format defined in [ENCAP], portions of which are included here for
convenience.
The format of an encapsulated frame is shown below: +--------------------+ | Header | +--------------------+-----+ | SOF | f | +--------------------+ F r | | FC frame content | C a | +--------------------+ m | | EOF | e | +--------------------+-----+ Figure 12. Encapsulation Format The encapsulation consists of a 7-word header, an SOF delimiter word, the FC frame (including the fibre channel CRC), and an EOF delimiter word. The header and delimiter formats are described in the following sections.5.3.1. Encapsulation Header Format
W|------------------------------Bit------------------------------| o| | r| 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 3 3| d|0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1| +---------------+---------------+---------------+---------------+ 0| Protocol# | Version | -Protocol# | -Version | +---------------+---------------+---------------+---------------+ 1| Reserved (must be zero) | +---------------+---------------+---------------+---------------+ 2| LS_COMMAND_ACC| iFCP Flags | SOF | EOF | +-----------+---+---------------+-----------+---+---------------+ 3| Flags | Frame Length | -Flags | -Frame Length | +-----------+-------------------+-----------+-------------------+ 4| Time Stamp [integer] | +---------------------------------------------------------------+ 5| Time Stamp [fraction] | +---------------------------------------------------------------+ 6| CRC | +---------------------------------------------------------------+ Figure 13. Encapsulation Header Format Common Encapsulation Fields: Protocol# IANA-assigned protocol number identifying the protocol using the encapsulation. For iFCP, the value assigned by [ENCAP] is 2.
Version Encapsulation version, as specified in [ENCAP]. -Protocol# Ones complement of the Protocol#. -Version Ones complement of the version. Flags Encapsulation flags (see 5.3.1.1). Frame Length Contains the length of the entire FC Encapsulated frame, including the FC Encapsulation Header and the FC frame (including SOF and EOF words) in units of 32-bit words. -Flags Ones complement of the Flags field. -Frame Length Ones complement of the Frame Length field. Time Stamp [integer] Integer component of the frame time stamp, as specified in [ENCAP]. Time Stamp Fractional component of the time stamp, [fraction] as specified in [ENCAP]. CRC Header CRC. MUST be valid for iFCP. The time stamp fields are used to enforce the limit on the lifetime of a fibre channel frame as described in Section 8.2.1. iFCP-Specific Fields: LS_COMMAND_ACC For a special link service ACC response to be processed by iFCP, the LS_COMMAND_ACC field SHALL contain a copy of bits 0 through 7 of the LS_COMMAND to which the ACC applies. Otherwise, the LS_COMMAND_ACC field SHALL be set to zero. iFCP Flags iFCP-specific flags (see below). SOF Copy of the SOF delimiter encoding (see Section 5.3.2). EOF Copy of the EOF delimiter encoding (see Section 5.3.2).
The iFCP flags word has the following format: |------------------------Bit----------------------------| | | | 8 9 10 11 12 13 14 15 | +------+------+------+------+------+------+------+------+ | Reserved | SES | TRP | SPC | +------+------+------+------+------+------+------+------+ Figure 14. iFCP Flags Word iFCP Flags: SES 1 = Session control frame (TRP and SPC MUST be 0) TRP 1 = Address transparent mode enabled 0 = Address translation mode enabled SPC 1 = Frame is part of a link service message requiring special processing by iFCP prior to forwarding to the destination N_PORT.5.3.1.1. Common Encapsulation Flags
The iFCP usage of the common encapsulation flags defined in [ENCAP] is shown in Figure 15: |------------------------Bit--------------------------| | | | 0 1 2 3 4 5 | +--------------------------------------------+--------+ | Reserved | CRCV | +--------------------------------------------+--------+ Figure 15. iFCP Common Encapsulation Flags For iFCP, the CRC field MUST be valid, and CRCV MUST be set to one.
5.3.2. SOF and EOF Delimiter Fields
The format of the delimiter fields is shown below. W|------------------------------Bit------------------------------| o| | r| 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 3 3| d|0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1| +---------------+---------------+---------------+---------------+ 0| SOF | SOF | -SOF | -SOF | +---------------+---------------+---------------+---------------+ 1| | +----- FC frame content -----+ | | +---------------+---------------+---------------+---------------+ n| EOF | EOF | -EOF | -EOF | +---------------+---------------+---------------+---------------+ Figure 16. FC Frame Encapsulation Format SOF (bits 0-7 and bits 8-15 in word 0): iFCP uses the following subset of the SOF fields specified in [ENCAP]. For convenience, these are reproduced in Table 3. The authoritative encodings should be obtained from [ENCAP]. +-------+----------+ | FC | | | SOF | SOF Code | +-------+----------+ | SOFi2 | 0x2D | | SOFn2 | 0x35 | | SOFi3 | 0x2E | | SOFn3 | 0x36 | +-------+----------+ Table 3. Translation of FC SOF Values to SOF Field Contents -SOF (bits 16-23 and 24-31 in word 0): The -SOF fields contain the ones complement the value in the SOF fields. EOF (bits 0-7 and 8-15 in word n): iFCP uses the following subset of EOF fields specified in [ENCAP]. For convenience, these are reproduced in Table 4. The authoritative encodings should be obtained from [ENCAP].
+-------+----------+
| FC | |
| EOF | EOF Code |
+-------+----------+
| EOFn | 0x41 |
| EOFt | 0x42 |
+-------+----------+
Table 4. Translation of FC EOF Values to EOF Field Contents
-EOF (bits 16-23 and 24-31 in word n): The -EOF fields contain the
ones complement the value in the EOF fields.
iFCP implementations SHALL place a copy of the SOF and EOF delimiter
codes in the appropriate header fields.
5.3.3. Frame Encapsulation
A fibre channel Frame to be encapsulated MUST first be validated as
described in [FC-FS]. Any frames received from a locally attached
fibre channel device that do not pass the validity tests in [FC-FS]
SHALL be discarded by the gateway.
If the frame is a PLOGI ELS, the creation of an iFCP session, as
described in Section 7.3.1.7, may precede encapsulation. Once the
session has been created, frame encapsulation SHALL proceed as
follows.
The S_ID and D_ID fields in the frame header SHALL be referenced to
look up the iFCP session descriptor (see Section 5.2.2.2). If no
iFCP session descriptor exists, the frame SHALL be discarded.
Frame types submitted for encapsulation and forwarding on the IP
network SHALL have one of the SOF delimiters in Table 3 and an EOF
delimiter from Table 4. Other valid frame types MUST be processed
internally by the gateway as specified in the appropriate fibre
channel specification.
If operating in address translation mode and processing a special
link service message requiring the inclusion of supplemental data,
the gateway SHALL format the frame payload and add the supplemental
information specified in Section 7.1. The gateway SHALL then
calculate a new FC CRC on the reformatted frame.
Otherwise, the frame contents SHALL NOT be modified and the gateway
MAY encapsulate and transmit the frame image without recalculating
the FC CRC.
The frame originator MUST then create and fill in the header and the SOF and EOF delimiter words, as specified in Sections 5.3.1 and 5.3.2.5.3.4. Frame De-encapsulation
The receiving gateway SHALL perform de-encapsulation as follows: Upon receiving the encapsulated frame, the gateway SHALL check the header CRC. If the header CRC is valid, the receiving gateway SHALL check the iFCP flags field. If one of the error conditions in Table 5 is detected, the gateway SHALL handle the error as specified in Section 5.2.3. +------------------------------+-------------------------+ | Condition | Error Type | +==============================+=========================+ | Header CRC Invalid | Encapsulation error | +------------------------------+-------------------------+ | SES = 1, TRP or SPC not 0 | Encapsulation error | +------------------------------+-------------------------+ | SES = 0, TRP set incorrectly | Incorrect address mode | +------------------------------+-------------------------+ Table 5. Encapsulation Header Errors The receiving gateway SHALL then verify the frame propagation delay as described in Section 8.2.1. If the propagation delay is too long, the frame SHALL be discarded. Otherwise, the gateway SHALL check the SOF and EOF in the encapsulation header. A frame SHALL be discarded if it has an SOF code that is not in Table 3 or an EOF code that is not in Table 4. The gateway SHALL then de-encapsulate the frame as follows: a) Check the FC CRC and discard the frame if the CRC is invalid. b) If operating in address translation mode, replace the S_ID field with the N_PORT alias of the frame originator, and the D_ID with the N_PORT ID, of the frame recipient. Both parameters SHALL be obtained from the iFCP session descriptor. c) If processing a special link service message, replace the frame with a copy whose payload has been modified as specified in Section 7.1.
The de-encapsulated frame SHALL then be forwarded to the N_PORT specified in the D_ID field. If the frame contents have been modified by the receiving gateway, a new FC CRC SHALL be calculated.
(page 47 continued on part 3)
