US20090228568A1

US20090228568A1 - Multicasting Computer Bus Switch

Info

Publication number: US20090228568A1
Application number: US12/468,198
Authority: US
Inventors: Heath Stewart; Michael de la Garrigue; Chris Haywood; Thomas Reiner; Ken Wong
Original assignee: TOPSIDE RESEARCH LLC
Current assignee: Avago Technologies International Sales Pte Ltd
Priority date: 2004-01-05
Filing date: 2009-05-19
Publication date: 2009-09-10
Also published as: US20050147114A1; US7539190B2; US20160147693A1; US9984027B2

Abstract

There is disclosed apparatus and methods of multicasting in a shared address space. A shared memory address space may include two or more multicast portions. Each multicast portion may be associated with a respective end point and with at least one other multicast portion. Data units may be transmitted to at least some of the end points via memory-mapped I/O into the shared memory address space. When a destination address of a data unit is in a first multicast portion associated with a first end point, the data unit may be transmitted to the first end point, revised to specify a destination address in a second multicast portion associated with the first multicast portion, and transmitted to a second end point associated with the second multicast portion.

Description

RELATED APPLICATION INFORMATION

This patent is a continuation of application Ser. No. 10/778,857 filed Feb. 13, 2004, and entitled “Multicasting In A Shared Address Space”, which in turn claims priority from application Ser. No. 60/534,586 filed Jan. 5, 2004, and entitled “PCI Express Switch with Broadcast and/or Multicast Capability.”

NOTICE OF COPYRIGHTS AND TRADE DRESS

A portion of the disclosure of this patent document contains material which is subject to copyright protection. This patent document may show and/or describe matter which is or may become trade dress of the owner. The copyright and trade dress owner has no objection to the facsimile reproduction by any one of the patent disclosure as it appears in the Patent and Trademark Office patent files or records, but otherwise reserves all copyright and trade dress rights whatsoever.

BACKGROUND OF THE INVENTION

1. Field Of The Invention
The present invention relates to multicasting in a shared address space.
2. Description of the Related Art
The Peripheral Component Interconnect (“PCI”) standard was promulgated about ten years ago, and has since been updated a number of times. One update led to the PCI/X standard, and another, more recently, to PCI Express. The PCI standards are defined for chip-level interconnects, adapter cards and device drivers. The PCI standards are considered cost-effective, backwards compatible, scalable and forward-thinking.
PCI buses, whether they be PCI Express or previous PCI generations, provide an electrical, physical and logical interconnection for multiple peripheral components of microprocessor based systems. PCI Express systems differ substantially from their PCI and PCI/X predecessors in that all communication in the system is performed point-to-point. Unlike PCI/X systems in which two or more end points share the same electrical interface, PCI Express buses connect a maximum of two end points, one on each end of the bus. If a PCI Express bus must communicate with more than one end point, a switch, also known as a fan out device, is required to convert the single PCI Express source to multiple sources.
The communication protocol in a PCI Express system is identical to legacy PCI/X systems from the host software perspective. In all PCI systems, each end point is assigned one or more memory and 10 address ranges. Each end point is also assigned a bus/device/function number to uniquely identify it from other end points in the system. With these parameters set a system host can communicate with all end points in the system. In fact, all end points can communicate with all other end points within a system. However, communication in PCI Express is limited to two end points, a source and a destination, at a time.
The PCI Express standard specifies one limited form of broadcasting. That is, if the transaction is a TLP type Message (Msg) denoted by a Format and Type field of 0110011, the transaction is broadcast from the Root Complex to all end points. This broadcast is for system management and configuration and is not applicable to data transport transactions.

DESCRIPTION OF THE DRAWINGS

FIG. 1 is a block diagram of a switching environment.

FIG. 2 is a diagram of a shared address space.

FIG. 3 is a flow chart of a method of multicasting in a shared address space.

DETAILED DESCRIPTION OF THE INVENTION

Throughout this description, the embodiments and examples shown should be considered as exemplars, rather than limitations on the apparatus and methods of the present invention.

Description of Systems

Referring now to FIG. 1, there is shown a block diagram of a switching environment 100. The switching environment includes a switch 110, a number of end points 120 a, 120 b, 120 c, 120 d. The switching environment 100 may be a point-to-point communications network.
The term “switch” as used herein means a system element that connects two or more ports to allow data units to be routed from one port to another, and the switch 110 is a switch. The switch 110 includes a number of ports 112 a, 112 b, 112 c, 112 d, which are logical interfaces between the switch 110 and the end points 120. The switch 110 further includes a buffer 115 and logic 117.
By data unit, it is meant a frame, cell, datagram, packet or other unit of information. In some embodiments, such as PCI, a data unit is unencapsulated. Data units may be stored in the buffer 115. By buffer, it is meant a dedicated or shared memory, a group or pipeline of registers, and/or other storage device or group of storage devices which can store data temporarily. The buffer 115 may operate at a speed commensurate with the communication speed of the switching environment 100. For example, it may be desirable to provide a dedicated memory for individual portions (as described below) and pipelined registers for multicast portions (as described below).
The logic 117 includes software and/or hardware for providing functionality and features described herein. The logic 117 may include one or more of: logic arrays, memories, analog circuits, digital circuits, software, firmware, and processors such as microprocessors, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), programmable logic devices (PLDs) and programmable logic arrays (PLAs). The hardware and firmware components of the logic 117 may include various specialized units, circuits, software and interfaces for providing the functionality and features described herein. The invention may be embodied in whole or in part in software which operates in the switch 110 and may be in the form of firmware, an application program, an applet (e.g., a Java applet), a browser plug-in, a COM object, a dynamic linked library (DLL), a script, one or more subroutines, or an operating system component or service. The hardware and software of the invention and its functions may be distributed such that some components are performed by the switch 110 and others by other devices.
The end points 120 a, 120 b, 120 c, 120 d are logical devices which connect to and communicate with the switch 110 respectively through the ports 112. The end points 120 a, 120 b, 120 c, 120 d may share an address space, such as a memory address space or an I/O address space. The term “address space” means the total range of addressable locations. If the shared address space is a memory address space, then data units are transmitted via memory mapped I/O to a destination address into the shared memory address space.
Referring now to FIG. 2, there is shown a diagram of a shared address space 200. The shared address space 200 shows contiguous ranges, but the address spaces associated with the end points 120 may be non-contiguous and the term “portions” is meant to refer to contiguous and non-contiguous spaces. Data units may be written into or communicated into an address portion. Address portions must therefore be large enough to accommodate at least one data unit. For at least these reasons, a single point or address within an address space cannot be a portion. An address portion must occupy at least two slots within the address space, and in most embodiments will have a sizable number of slots specified as a range. In a switch conforming to the PCI Express standard, it is expected that the address portions in a 32-bit shared memory address space or shared I/O address space will be at least as large as the largest expected transaction, and comparable to those shown in FIG. 2.
Within the shared address space 200, there is a gross address portion 210 a associated with end point A 120 a. Within the gross address portion 210 a, there is an individual portion 220 a, a multicast portion 230 a and a broadcast portion 240 a. Likewise, end point B 120 b may have a gross address portion 210 b with an individual portion 220 b, a multicast portion 230 b and a broadcast portion 240 b. Likewise, end point C 120 c may have a gross address portion 210 c with an individual portion 220 c, a multicast portion 230 c and a broadcast portion 240 c. A gross address portion, an individual portion, a multicast portion and a broadcast portion may be associated with end point D 120 d.
The address space 200 may be allocated so as to provide the end points 120 with unique gross address portions. The individual portions may be unique within the shared address space with respect to one another, as may be the multicast portions and the broadcast portions.
The address portions (gross, individual, multicast and broadcast) may have various characteristics. The address portions may have respective sizes. The sizes may be fixed or variable. The address portions may be defined by a base address, as well as by a size or end address. The address portions may come to be associated with the end points 120 through an arbitrage process, through centralized assignment (e.g., by a host or the switch 110), otherwise or through a combination of these. The group portion, the individual portion, the multicast portion and the broadcast portion for a given end point 120 need not be contiguous. To avoid errors, it may be desirable if the individual portions, the multicast portions and the broadcast portions do not overlap.
Data units may be directed to one or more of the end points 120 by addressing. That is, a destination address is associated with and may be included in the data units. The destination address determines which end point 120 should receive a given data unit. Thus, data units addressed to the individual portion for a given end point 120 should be received only by that end point 120. Depending on the embodiment, the destination address may be the same as the base address or may be within the address portion.
Multicasting presents a somewhat more complex and flexible case than single-casting. To allow for multicasting to a group of selected end points, a multicast group is defined. Within the multicast group, the multicast portions of the selected end points are associated, and logic is provided which causes data units sent to the multicast portion of one end point in the multicast group to be sent to the multicast portions of the other end points in the multicast group. The data units addressed to the multicast portion for a given end point 120 should be received by all of the end points in the same multicast group. Alternatively, within a multicast group, one of the multicast portions may be selected as a “master” and the other multicast portions treated as ghosts. Accordingly, data units addressed to the master may be multicast to the group, but data units addressed to a ghost may be single-cast to the slave or treated as exceptions.
A given end point 120 may belong to multiple multicast groups and therefore have multiple multicast portions. For example, end point A 120 a may be in a first multicast group with end point B 120 b, in a second multicast group with end point C 120 c, and a third multicast group with end point B 120 b and end point C 120 c. In this example, end point A might have three multicast portions. The various multicast groups may also be grouped, to provide super-groupings. For example, there might be a first multicast group having end point A 120 a and end point B 120 b; a second multicast group having end point A 120 a and end point C 120 c; and a third multicast group having the first multicast group and the second multicast group, i.e., end point A 120 a, end point B 120 b and end point C 120 c.
It can be seen that single-casting and broadcasting are special cases of multicasting. In single-casting, the multicast group includes only one end point, and has only the one end point's individual portion. In contrast, in broadcasting, the multicast group includes all end points, and has the broadcast portions for all end points. In one alternative, there is a single broadcast portion, and logic is provided which causes data units which are sent to the broadcast portion to be sent to the individual portions of all end points 120.
Each multicast portion may be unique. Alternatively, there may be a single multicast portion for all of the end points 120 in a multicast group. An alternate way to support multicast would be to define multiple sub-portions within a master broadcast portion, each with its own vector defining which ports are to participate in the multicast transactions. Each sub-portion would define a multicast group and the associated vector would contain an enable bit for each port on the switch. If the enable bit for a port is set then the transaction is forwarded to that port. Any number of multicast portions could be defined by this mechanism.
The multicasting portions in a group may have nearly identical base addresses, and only differ from each other from a single or small number of bits or digits. The sizes of the individual portions for the various end points 120 may differ. In contrast, the multicasting portions in a multicast group may have substantially equal or equal sizes. Having such equal-sized multicast portions may ensure communication integrity and efficient use of the shared address space 300.
The end points 120 may be associated with respective ports 112. Through this association, a given end point 120 may send data units to and receive data units from its associated port 112. This association may be on a one-to-one basis. Because of these relationships, the ports 112 also have associations with the address portions of the end points 120. Thus, the ports 112 may be said to have address portions (including respective individual portions, multicast portions and broadcast portions) within the address space 200.

Description of Methods

Referring now to FIG. 3, there is shown a flow chart of a method of multicasting in a shared address space. The switch 110 may receive a data unit, e.g., through port D 120 d (step 305). The logic 117 causes the received data unit to be stored in the buffer 115 (step 310). The data unit may be stored in whole or in part in the buffer 115. For example, in streaming applications, it may be desirable to store a header in the buffer but switch the payload directly from the ingress port to the egress port in a cut-through manner. The logic 117 also determines the destination address of the data unit and selects the port 120 associated with the destination address (step 315). Step 315 may be performed, for example, using a lookup table, or through hard wiring addresses to ports. Next, the logic 117 forwards the data unit for transmission out the selected port 120 (step 320).
If the destination address is in the individual portion associated with one of the ports (step 325), then the logic 117 causes or allows the data unit to be removed from the buffer 115 (step 395).
If the destination address is in the multicast portion associated with one of the ports (step 325), then step 395 is deferred. Instead, the data unit is forwarded for transmission out the other ports in the same group as the multicast portion encompassing the destination address. This may be achieved by replacing the destination address of the data unit with that of another (e.g., the next) multicast portion in the same group (step 330), and then forwarding the data unit for transmission out the port associated with the (revised) destination address (step 335). If there is more than one port in the multicast group (step 340), steps 330 and 335 may be continued until the data unit has been forwarded for transmission out all of the ports in the group. Then, the data unit may be removed from the buffer 115 (step 395).
The replacing step 330 may be performed in a number of ways. For example, the destination address may be revised by drawing addresses from a table of multicast portions. Alternatively, the multicast portions in a multicast group may differ from one another according to a rule, and the rule used to determine the next destination address. For example, as shown in FIG. 2, the multicast portions may be contiguous blocks of 0×10000000 spaced apart by 0×40000000.
Broadcasting may be handled similarly to multicasting. Thus, if a data unit has a destination address in the broadcast portion for a port, then the data unit is forwarded for transmission out the port, the destination address is revised as in step 330 and the data unit is forwarded as in step 335. This may be continued until the data unit has been forwarded for transmission out all of the ports.
The use of shared memory space as described may be considered as providing “real” ports which are associated with the individual portions, and “virtual” ports which are associated with the multicast portions and broadcast portions. The virtual ports may be mapped to the real ports. Thus, data units may be multicast simply by selecting an appropriate address, and neither the format of the data units nor the content of the data units need be changed to accommodate multicasting. Intelligence in the switch recognizes that an address is a multicast address, and replicates and re-maps the address of the data units to the other ports in the multicast group.
Although broadcast has been treated as a special case of multicast, the converse is also possible. According to one alternative, broadcast support is enabled and ports outside of the multicast group are disabled. This could be done ahead of each multicast data unit. For example, to send a data unit from end point D 120 d to both end point A 120 a and end point C 120 c, end point D 110 d could send an instruction to the switch 110 to enable broadcast, but disable port B 112 b. End point D 120 d would then send the data unit which the switch 110 would route to port A 112 a and port C 112 c.
With regard to FIG. 2, additional and fewer steps may be taken, and the steps as shown may be combined or further refined to achieve the methods described herein.
There may be anticipated and unanticipated conditions in which one or more of the ports 112 is removed or otherwise becomes unavailable, either in a controlled or uncontrolled manner. To maintain desirable data flow, the logic 117 may include a capability to resolve these types of port exceptions. If a port becomes unavailable, for example, the logic 117 may ignore or discard those data units addressed to the individual portion, the multicast portion and/or the broadcast portion for that port. The logic 117 may multicast portion or a broadcast portion for an unavailable port, the logic may skip the unavailable port and continue the multicast or broadcast to other ports. Alternatively, the logic 117 may discontinue the multicast or broadcast altogether. The logic 117 may report the port exceptions and its response to the source of the data units and/or to other destinations.
The invention may be used to advantage in PCI Express switches and devices. For example, PCI Express-compliant video graphics systems and communications data backplanes may benefit from the invention. It is believed that the invention is compatible with the PCI Express memory write request transaction. The invention may be compatible with other PCI Express transaction types and other standards.
The PCI Express standard provides for confirmation messages in some situations, which the standard refers to as non-posted transactions. The system and methods described herein are compatible with both posted and non-posted transactions, though it may be desirable to consolidate or otherwise dispose of confirmation messages responsive to multicast and broadcast data units.
Although exemplary embodiments of the present invention have been shown and described, it will be apparent to those having ordinary skill in the art that a number of changes, modifications, or alterations to the invention as described herein may be made, none of which depart from the spirit of the present invention. All such changes, modifications and alterations should therefore be seen as within the scope of the present invention.

Claims

1. A method of multicasting data units in a shared memory address space, comprising:

associating two or more multicast portions of the shared memory address space with respective end points, wherein each multicast portion is associated with at least one other multicast portion

transmitting data units to at least some of the end points via memory-mapped I/O into the shared memory address space

wherein, when a destination address of a data unit is in a first multicast portion associated with a first end point, the method further comprises:

transmitting the data unit to the first end point

revising the data unit to specify a destination address in a second multicast portion associated with the first multicast portion

transmitting the data unit to a second end point associated with the second multicast portion.

2. The method of multicasting data units of claim 1, wherein the method further comprises:

associating a plurality of individual portions of the shared memory address space with respective end points

when the destination address of a data unit is in an individual portion, transmitting the data unit to the respective end point.

3. The method of multicasting data units of claim 2, wherein the method further comprises:

associating each end point with a respective broadcast portion of the shared memory address space

when the destination address of a data unit is in a first broadcast portion, the method further comprises

(a) transmitting the data unit to an end point associated with the first broadcast portion

(b) revising the data unit to specify a destination address in another broadcast portion

(c) transmitting the data unit to an end point associated with the another broadcast portion

(d) repeating (b) and (c) until the data unit has been transmitted to every end point.

4. The method of multicasting data units of claim 3, wherein

each end point is connected to a respective port of a switch

data units are transmitted to the end points via the respective ports.

5. The method of multicasting data units of claim 3, wherein each of the multicast portions, the individual portions, and the broadcast portions is at least as large as a largest expected data unit.

6. A switch for multicasting in a shared memory address space, the switch comprising

a buffer to receive data units transmitted via memory mapped I/O to addresses in the shared memory address space

a plurality of ports associated with a corresponding plurality of end points, at least two of the plurality ports associated with a respective multicast address portion in the shared memory address space, wherein each of the multicast address portions is associated with at least one other multicast address portion

logic to

cause data units in the buffer having a destination address in a first multicast address portion associated with a first port to be forwarded for transmission out the first port without being removed from the buffer and then replace the destination address with an address in a second multicast address portion associated with the first multicast address portion

cause data units in the buffer having a destination address in the second multicast address portion to be forwarded for transmission out the second port.

7. The switch for multicasting in a shared memory address space of claim 6, wherein

each of the plurality of ports is associated with a respective individual address portion in the shared memory address space

the logic further to cause data units in the buffer having a destination address in an individual address portion to be forwarded for transmission out the respective port.

8. The switch for multicasting in a shared memory address space of claim 7, wherein

each of the plurality of ports is associated with a respective broadcast address portion in the shared memory address space

the logic further to cause data units in the buffer having a destination address in a broadcast address portion to be forwarded for transmission out all of the plurality of ports.

9. The switch for multicasting in a shared memory address space of claim 8, wherein each of the multicast address portions, the individual address portions, and the broadcast address portion is at least as large as a largest expected data unit.

10. A switch for multicasting in a shared memory address space, the switch comprising

a plurality of ports associated with a corresponding plurality of end points, at least some of the ports associated with one or more multicast groups, wherein at least two ports are associated with each multicast group and each multicast group is associated with a respective multicast address portion of the shared memory address space

logic to cause data units in the buffer having a destination address in a first multicast address portion to be forwarded for transmission out all ports associated with a first multicast group associated with the first multicast address portion.

11. The switch for multicasting in a shared memory address space of claim 10, wherein

12. The switch for multicasting in a shared memory address space of claim 11, wherein

each of the plurality of ports is associated with a broadcast address portion in the shared memory address space

the logic further to cause data units in the buffer having a destination address in the broadcast address portion to be forwarded for transmission out all of the plurality of ports.

13. The switch for multicasting in a shared memory address space of claim 12, wherein each of the multicast address portions, the individual address portions, and the broadcast address portion is at least as large as a largest expected data unit.

14. A method of multicasting in a shared memory address space using a switch having a plurality of ports associated with a corresponding plurality of end points, the method comprising:

associating a plurality of multicast portions of the shared memory address space with respective ports; wherein each multicast portion is associated with at least one other multicast portion

receiving data units transmitted via memory-mapped I/O to destination addresses in the shared memory address space

wherein, when a destination address of a data unit is in a first multicast portion associated with a first port, the method further comprises:

forwarding the data unit for transmission out the first port

forwarding the data unit for transmission out a second port associated with the second multicast portion.

15. The method of multicasting data units of claim 14, wherein the method further comprises:

associating a plurality of individual portions of the shared memory address space with respective ports

when the destination address of a data unit is in an individual portion, forwarding the data unit for transmission out a respective port.

16. The method of multicasting data units of claim 15, wherein the method further comprises:

associating each port with a respective broadcast portion of the shared memory address space

(a) forwarding the data unit for transmission out a first port associated with the first broadcast portion

(c) forwarding the data unit for transmission out another port associated with the another broadcast portion

(d) repeating (b) and (c) until the data unit has been forwarded for transmission out all of the plurality of ports.

17. The method of multicasting data units of claim 16, wherein each of the multicast portions, the individual portions, and the broadcast portions is at least as large as a largest expected data unit.

18. A method of multicasting in a shared memory address space using a switch having a plurality of ports associated with a corresponding plurality of end points, the method comprising:

associating at least one multicast portion of the shared memory address space with two or more ports

receiving data units addressed to destination addresses in the shared memory address space, wherein

when a destination address of a data unit is in an individual portion, forwarding the data unit for transmission out the respective port

when a destination address of a data unit is in a multicast portion, forwarding the data unit for transmission out each of the ports associated with multicast portion

when a destination address of a data unit is in a broadcast portion of the shared memory address space, forwarding the data unit for transmission out all of the plurality of ports.

19. The method of multicasting data units of claim 16, wherein each of the multicast portions, the individual portions, and the broadcast portion is at least as large as a largest expected data unit.