ARM NIC-301 User Manual

Download

AMBA® Network Interconnect (NIC-301)

Revision: r2p1

Technical Reference Manual

ARM DDI 0397G (ID031010)

AMBA Network Interconnect (NIC-301)

Technical Reference Manual

Release Information

The following changes have been made to this book.

Date Issue Confidentiality Change

05 May 2006 A Non-Confidential First release for r0p0.

19 September 2006 B Confidential First release for r1p0.

06 July 2007 C Confidential First release for r1p1.

02 October 2007 D Confidential First release for r1p2.

22 April 2008 E Confidential Second release for r1p2. Updates to the following:

• Decode register

• Read acceptance capability and write acceptance capability

• Write issuing capability

• Narrow to wide translation

• Physical data.

03 November 2009 F Non-Confidential First release for r2p0.

19 February 2010 G Non-Confidential First issue for r2p1.

Change history

Proprietary Notice

Words and logos marked with

or © are registered trademarks or trademarks of ARM in the EU and other countries, except as otherwise stated below in this proprietary notice. Other brands and names mentioned herein may be the trademarks of their respective owners.

Neither the whole nor any part of the information contained in, or the product described in, this document may be adapted or reproduced in any material form except with the prior written permission of the copyright holder.

The product described in this document is subject to continuous developments and improvements. All particulars of the product and its use contained in this document are given by ARM in good faith. However, all warranties implied or expressed, including but not limited to implied warranties of merchantability, or fitness for purpose, are excluded.

This document is intended only to assist the reader in the use of the product. ARM shall not be liable for any loss or damage arising from the use of any information in this document, or any error or omission in such information, or any incorrect use of the product.

Where the term ARM is used it means “ARM or any of its subsidiaries as appropriate”.

Confidentiality Status

This document is Non-Confidential. The right to use, copy and disclose this document may be subject to license restrictions in accordance with the terms of the agreement entered into by ARM and the party that ARM delivered this document to.

Unrestricted Access is an ARM internal classification.

Product Status

The information in this document is final, that is for a developed product.

Web Address

http://www.arm.com

AMBA Network Interconnect (NIC-301) Technical Reference Manual

Preface

About this book ............................................................................................................... vii

Feedback ......................................................................................................................... ix

Chapter 1 Introduction

1.1 About the AMBA Network Interconnect ........................................................................ 1-2

1.2 Key features .................................................................................................................. 1-3

1.3 Relationship between AMBA Network Interconnect and AMBA Designer .................... 1-4

1.4 Product revisions .......................................................................................................... 1-5

Chapter 2 Functional Description

2.1 About the functions ....................................................................................................... 2-2

2.2 Interfaces ...................................................................................................................... 2-3

2.3 Operation .................................................................................................................... 2-12

Chapter 3 Programmers Model

3.1 About this programmers model ..................................................................................... 3-2

3.2 Configuration programmers model ............................................................................... 3-3

Appendix A Revisions

Glossary

List of Tables

AMBA Network Interconnect (NIC-301) Technical Reference Manual

Change history ................................................................................................................................ ii

Table 2-1 Combination of configuration parameters ................................................................................... 2-4

Table 2-2 AXI burst type to AHB burst type mapping ................................................................................. 2-9

Table 2-3 Conversion of INCR bursts by the upsize function ................................................................... 2-13

Table 2-4 Conversion of WRAP bursts by the upsize function ................................................................. 2-13

Table 2-5 Conversion of INCR bursts by the downsize function ............................................................... 2-14

Table 2-6 Conversion of FIXED bursts by the downsize function ............................................................. 2-15

Table 2-7 How to change modes .............................................................................................................. 2-16

Table 3-1 Registers for each ASIB .............................................................................................................. 3-4

Table 3-2 Registers for each IB .................................................................................................................. 3-5

Table 3-3 Registers for each AMIB ............................................................................................................. 3-6

Table 3-4 Address region control registers ................................................................................................. 3-7

Table 3-5 Peripheral ID registers ................................................................................................................ 3-8

Table A-1 Differences between issue E and issue F ................................................................................... A-1

Table A-2 Differences between issue F and issue G .................................................................................. A-2

List of Figures

AMBA Network Interconnect (NIC-301) Technical Reference Manual

Figure 1-1 AMBA Network Interconnect top-level block diagram ................................................................. 1-2

Figure 2-1 No remap, remap set to 000 ..................................................................................................... 2-22

Figure 2-2 Remap set to 001 ...................................................................................................................... 2-22

Figure 2-3 Remap set to 010 ...................................................................................................................... 2-22

Figure 2-4 Remap set to 011 ...................................................................................................................... 2-23

Figure 2-5 Remap set to 101 ...................................................................................................................... 2-23

Figure 3-1 Address map of the programmers model .................................................................................... 3-3

Preface

This preface introduces the AMBA Network Interconnect (NIC-301). It contains the following

sections:

• About this book on page vii

• Feedback on page ix.

About this book

Product revision status

Intended audience

Using this book

Preface

This book is for AMBA Network Interconnect.

The rnpn identifier indicates the revision status of the product described in this book, where:

rn Identifies the major revision of the product.

pn Identifies the minor revision or modification status of the product.

This book is written for system designers, system integrators, and programmers who are designing or programming a System-on-Chip (SoC) that uses the AMBA Network Interconnect.

This book is organized into the following chapters:

Chapter 1 Introduction

Read this for a high-level view of the AMBA Network Interconnect and a description of its features.

Conventions

Chapter 2 Functional Description

Read this for a description of the major interfaces and components of the AMBA Network Interconnect. The chapter also describes how they operate.

Chapter 3 Programmers Model

Read this for a description the address map and registers of the AMBA Network Interconnect.

Appendix A Revisions

Read this for a description of the technical changes between released issues of this book.

Glossary Read this for definitions of terms used in this book.

Conventions that this book can use are described in:

• Typographical.

• Signals on page viii.

Typographical

The typographical conventions are:

italic Highlights important notes, introduces special terminology, denotes

internal cross-references, and citations.

bold Highlights interface elements, such as menu names. Denotes signal

names. Also used for terms in descriptive lists, where appropriate.

monospace

Denotes text that you can enter at the keyboard, such as commands, file

and program names, and source code.

Preface

Additional reading

monospace

Denotes a permitted abbreviation for a command or option. You can enter

the underlined text instead of the full command or option name.

monospace italic

Denotes arguments to monospace text where the argument is to be

replaced by a specific value.

monospace bold

Denotes language keywords when used outside example code.

< and > Enclose replaceable terms for assembler syntax where they appear in code

or code fragments. For example:

MRC p15, 0 <Rd>, <CRn>, <CRm>, <Opcode_2>

Signals

The signal conventions are:

Signal level The level of an asserted signal depends on whether the signal is

active-HIGH or active-LOW. Asserted means:

• HIGH for active-HIGH signals

• LOW for active-LOW signals.

Lower-case n At the start or end of a signal name denotes an active-LOW signal.

This section lists publications by ARM and by third parties.

See Infocenter, http://infocenter.arm.com, for access to ARM documentation.

ARM publications

This book contains information that is specific to this product. See the following documents for other relevant information:

• AMBA

Network Interconnect (NIC-301) Integration Manual (ARM DII 0157)

• AMBA Network Interconnect (NIC-301) Implementation Guide (ARM DII 0222)

• AMBA Designer (ADR-301) User Guide (ARM DUI 0333)

• AMBA Network Interconnect (NIC-301) Supplement to AMBA Designer (ADR-301) User Guide (ARM DSU 0003)

• AMBA AXI Protocol v1.0 Specification (ARM IHI 0022)

• AMBA 3 AHB-Lite Protocol v1.0 Specification (ARM IHI 0033)

• AMBA 3 APB Protocol v1.0 Specification (ARM IHI 0024)

• APB Rev E Specification (ARM IHI 0009).

Feedback

ARM welcomes feedback on this product and its documentation.

Feedback on this product

If you have any comments or suggestions about this product, contact your supplier and give:

• The product name.

• The product revision or version.

• An explanation with as much information as you can provide. Include symptoms and

Feedback on content

Preface

diagnostic procedures if appropriate.

If you have comments on content then send an e-mail to

errata@arm.com

. Give:

• the title

• the number, ARM DDI 0397G

• the page numbers to which your comments apply

• a concise explanation of your comments.

ARM also welcomes general suggestions for additions and improvements.

Chapter 1

Introduction

This chapter introduces the AMBA Network Interconnect. It contains the following sections:

• About the AMBA Network Interconnect on page 1-2

• Key features on page 1-3

• Relationship between AMBA Network Interconnect and AMBA Designer on page 1-4

• Product revisions on page 1-5.

1.1 About the AMBA Network Interconnect

Switch

Interface

Block (IB)

AMIB

ASIB

Distributed Global

Programmers

View (GPV)

Default

slave

AXI AXI

AMBA

master

interface

AXI

signals

AMBA Network Interconnect (NIC-301)

Routing omitted

for clarity

AMBA

master

interface

AMBA

slave

interface

AMBA

slave

interface

ASIB

AMBA

slave

interface

AXI

Switch

AXI

signals

AMIB

AMBA

master

interface

AXI

signals

The AMBA Network Interconnect is a highly configurable component that enables you to create a complete high performance, optimized AMBA-compliant network infrastructure. The possible configurations for the AMBA Network Interconnect can range from a single bridge component, for example an AHB to AXI protocol bridge, to a complex infrastructure that consists of up to 128 masters and 64 slaves of a combination of different AMBA protocols.

An AMBA Network Interconnect configuration can consist of multiple switches with many topology options. Figure 1-1 shows a top-level block diagram of the AMBA Network Interconnect that contains:

• multiple switches

• multiple AMBA Slave Interface Blocks (ASIBs)

• multiple AMBA Master Interface Blocks (AMIBs).

Introduction

Figure 1-1 AMBA Network Interconnect top-level block diagram

1.2 Key features

Introduction

The AMBA Network Interconnect is a highly configurable infrastructure component that supports:

• 1-128 AXI or AHB-Lite slave interfaces.

• 1-64 master interfaces that can be AXI, AHB-Lite, APB2, or APB3.

• Configuration of:

— an APB port to support 1-16 slaves

— an AXI port to support four region control bits.

• Single-cycle arbitration.

• Full pipelining to prevent master stalls.

• Programmable control for FIFO transaction release.

• Multiple switch networks.

• Complex topologies, including Network On Chip (NOC) loop-back connections between switches.

• Up to five cascaded switch networks between any master and slave interface pair.

• AXI or AHB-Lite masters and slaves with:

— an address width of 32-64 bits

— a data width of 32, 64, 128, or 256 bits.

• Non-contiguous APB slave address map for a single master interface.

• Independent widths of user-defined sideband signals for each channel.

• Global Programmers View (GPV) for the entire infrastructure that you can configure so that any master, or a discrete configuration slave interface, can access it. See Chapter 3 Programmers Model.

• Highly flexible timing closure options.

1.3 Relationship between AMBA Network Interconnect and AMBA Designer

AMBA Designer is a configuration tool that generates a specific implementation of an AMBA Network Interconnect. AMBA Designer drives the AMBA Network Interconnect generation engine to provide the following for a set of configuration parameters and implementation scripts:

• Verilog Register Transfer Level (RTL)

• testbench and stimulus synthesis scripts.

The documentation suites and implementation scripts for the AMBA Network Interconnect and AMBA Designer are designed to be used together to describe the principles of the AMBA Network Interconnect and the actual configuration options. There is no duplication between the two sets of documentation. The following sections describe the information that each documentation suite provides:

• AMBA Network Interconnect documentation

• AMBA Designer documentation.

1.3.1 AMBA Network Interconnect documentation

The AMBA Network Interconnect documentation consists of:

TRM The Technical Reference Manual (TRM) describes how to create the transfer

function and possible capabilities of the network component and how to dynamically change it using the programmers model.

Introduction

IG The Implementation Guide (IG) describes how to set up the network environment

and how to use it to run RTL simulations or implementation scripts.

IM The Integration Manual (IM) describes how to integrate a configured network

into a larger subsystem.

1.3.2 AMBA Designer documentation

The AMBA Designer (FD001) User Guide describes how to:

• Install AMBA Designer.

• Generate and verify RTL sub-systems of ARM IP.

• Stitch ARM components together. ARM components conform to the IP-XACT from the SPIRIT Consortium

The AMBA Network Interconnect (NIC-301) Supplement to AMBA Designer (ADR-301) User Guide describes how to produce a customized infrastructure.

™

standard

1.4 Product revisions

This section describes differences in functionality between product revisions of the AMBA Network Interconnect (NIC-301):

r0p0-r1p0 Contains the following differences in functionality:

r1p0-r1p1 Contains the following differences in functionality:

Introduction

• support for 128-bit data width on AXI and all AHB variant interfaces

• support for n-bit addressing on AXI and all AHB variant interfaces, where n is 32-64 bit inclusive

• the decode register is a slave interface property instead of a global property

• single-slave-per-ID Cyclic Dependency Avoidance Scheme (CDAS)

• ID register and configuration data

• use of updated synchronous bridges

• use of an AHB to AXI bridge optimized for accessing memory where appropriate.

• Separation of arbitration of AW and AR channels so that a slave can accept transactions from two different masters simultaneously.

• Changes to the way that the AMBA Network Interconnect handles LOCKed transactions.

• Configurable arbitration schemes, that can be:

— programmable Least Recently Granted (LRG), the scheme present in

r1p0

— a non-programmable Round-Robin (RR) scheme.

• You can select and configure arbitration schemes for each master interface.

• The APB programming interface enables you to program and interrogate the new, separate arbitration schemes.

• The AHB to AXI bridge is optimized for accessing memory is updated with performance enhancements, and to fix a defect.

• The way arbitration schemes are described has changed to enable you to select and configure arbitration schemes.

•The Quality of Service (QoS) tidemark value now represents the maximum permitted number of active transactions before the QoS mechanism is activated, instead of the minimum number of unused slots before the mechanism is activated. This means the combined acceptance capability attribute of a master interface is no longer required.

r1p1-r1p2 Contains the following differences in functionality:

• Updates to the example synthesis scripts.

• Addition of a programmable version of the fixed round-robin arbitration scheme.

• Shortening some long paths to improve synthesis performance.

• A change to the way register slices are instantiated. This makes it easier to use them to resolve timing issues during synthesis.

• The choice of CDAS is independent for reads and writes.

• Additional configuration option for the single-slave-per-ID CDAS that permits only a single data-active write transaction to be in progress.

r1p2-r2p0 Contains the following differences in functionality:

• Network of interconnects instead of a single interconnect

• Optimized translation latency replaces additive translation latency

• Single cycle arbitration instead of arbitration switching delay

• Enhanced buffering to reduce stalls

• Multiple outstanding downsizer transactions instead of limited outstanding downsizer transactions

• Upsizer packs data into the wide bus, instead of an expander, for data width changes

• Global dynamic QoS instead of local static QoS

• Internally programmable instead of externally programmed

• System-level address map replaces an address map for each interconnect

• Extended timing closure options replace limited timing closure options

• 256-bit maximum data width instead of 128-bit maximum data width

• Multi-region slave support replaces a single region per slave

• Write FIFO, transaction release control instead of fixed write transaction release point.

See Appendix A Revisions.

Introduction

r2p0-r2p1 Contains no differences in functionality.

Chapter 2

Functional Description

This chapter describes the functionality of the AMBA Network Interconnect. It contains the following sections:

• About the functions on page 2-2

• Interfaces on page 2-3

• Operation on page 2-12.

2.1 About the functions

You can consider the AMBA Network Interconnect to be built from functions that each have their own transfer function. A transfer function can:

• a domain crossing, for example:

— clock domain crossing

— data width crossing.

• used to create timing isolation, for optimizing critical network paths for latency.

Within a domain, a switch, or multiple switches, can exist to enable routing paths between any slave interface to any master interface.

The functions are configured into routing switches or Interface Blocks (IBs) and you can use AMBA Designer to create highly complex topologies using these modules. For more information, see the AMBA Designer (ADR-301) User Guide and AMBA Network Interconnect (NIC-301) Supplement to AMBA Designer (ADR-301) User Guide.

Functional Description

2.2 Interfaces

Note

2.2.1 Slave interfaces

Functional Description

This section describes the AMBA Network Interconnect interfaces and contains the following subsections:

• Slave interfaces

• Master interfaces on page 2-8.

The AMBA Network Interconnect supports the following slave interfaces:

• AXI slave interfaces

• AHB-Lite slave interfaces on page 2-4.

Any transaction that does not decode to a legal master interface destination, or programmers view register, receives a DECERR response. For an AHB master, the AXI DECERR is mapped back to an AHB ERROR.

The AXI DECERR error is mapped back to an AHB master ERROR if:

• you do not configure the early write response

• you configure INCR Promotion and Early Write Response and the transaction is non-cacheable

• the AHB burst is not broken.

AXI slave interfaces

An AXI slave interface supports the full AXI protocol.

Configuration options

You can configure the following properties:

• Address width of 32-64 bits.

• Data width of 32, 64, 128, or 256 bits.

• User sideband signal width of 0-32 bits.

• Data width upsize function, see Upsizing data width function on page 2-12.

• Data width downsize function, see Downsizing data width function on page 2-14.

• Frequency domain crossing of the following types:

—ASYNC

—SYNC 1:1

—SYNC 1:n

—SYNC n:1

—SYNC n:m.

• Security of the following types:

Secure All transactions originating from this slave interface are flagged as secure

transactions and can access both secure and non-secure components.

Functional Description

Note

Non-secure

All transactions originating from this slave interface are flagged as non-secure transactions and cannot access secure components.

Per access

The AxPROTx signal determines the security setting of each transaction, and the slaves that it can access.

• Support for the full AXI protocol.

You can achieve a gate count reduction and a performance increase if the attached master

does not create any AXI lock transactions.

• Write acceptance capability of 1-32 transactions.

• Read acceptance capability of 1-32 transactions.

• Buffering, see FIFO and clocking function on page 2-15.

• Timing isolation:

— from the external master

— from the infrastructure.

AHB-Lite slave interfaces

The AMBA Network Interconnect can support the full AHB-Lite protocol using either:

• an AHB-Lite slave interface

• an AHB-Lite mirror master interface.

The following configuration options can improve AHB-Lite to AXI performance, but cannot always be used robustly:

•

INCR promotion and Early Write Response

•

allow broken bursts

If you configure the interface as an AHB mirror master interface, you cannot configure

broken bursts

because the AHB-Lite protocol does not permit AHB-Lite masters to break

bursts.

Table 2-1 shows the four combinations for the configuration of

Response

INCR promotion and Early Write Response allow broken bursts

Configured Not configured Combination 1 on page 2-5

Not configured Configured Combination 2 on page 2-5

Configured Configured Combination 3 on page 2-5

and

allow broken bursts

and contains links to detailed descriptions for each option.

Table 2-1 Combination of configuration parameters

INCR promotion and Early Write

Description of combination

allow

Not configured Not configured Combination 4 on page 2-6

Combination 1

Functional Description

If you configure

bursts

then the network converts all:

INCR promotion and Early Write Response

and do not configure

allow broken

• AHB read fixed length bursts to AXI fixed length bursts.

• AHB write fixed length bursts with HPROT[3] asserted to AXI fixed length bursts:

— All AHB write data beats receive an automatic OKAY response from the bridge

irrespective of the B-channel AXI response. This means that if the network receives an error response, it does not feed it back to the master.

— The bridge can support up to five outstanding write accesses.

• Write fixed-length bursts with HPROT[3] negated to AXI fixed length bursts, and only the last AHB write data beat receives the AXI buffered response for the complete AHB transaction.

• AHB read INCR bursts with HPROT[3] asserted to AXI INCR4 bursts.

• Write INCR bursts with HPROT[3] asserted to AXI INCR4 bursts, and all AHB write data beats receive an automatic OKAY response from the bridge, irrespective of the B-channel AXI response. This means that if the network receives an error response, it does not feed it back to the master.

• Read INCR bursts with HPROT[3] negated to a series of AXI singles.

• Write INCR bursts with HPROT[3] negated to a series of AXI singles, and each AHB write beat is acknowledged with the AXI buffered write response.

Combination 2

If you configure

Response

, the network converts all:

allow broken bursts

and do not configure

INCR promotion and Early Write

• Read fixed length bursts with HPROT[3] asserted to AXI fixed length bursts.

• Read fixed length bursts with HPROT[3] negated to AXI singles.

• Write fixed length bursts with HPROT[3] asserted to AXI fixed length bursts, but only the last AHB write data beat receives the AXI buffered response for the whole AHB transaction. However, if the AHB burst is broken, then the network does not feed the AXI response back to the master.

• Write fixed length bursts with HPROT[3] negated to AXI singles, and each AHB write beat is acknowledged with the AXI buffered write response.

• Read INCR bursts to a series of AXI singles.

• Write INCR bursts to a series of AXI singles, and each AHB write beat is acknowledged with the AXI buffered write response.

Combination 3

If you configure

INCR promotion and Early Write Response

and configure

allow broken bursts

then the network converts all:

• Read fixed length bursts with HPROT[3] asserted to AXI fixed length bursts.

• Read fixed length bursts with HPROT[3] negated to AXI singles.

Functional Description

Note

• Write fixed length bursts with HPROT[3] asserted to AXI fixed length bursts:

— The bridge sends an automatic OKAY response to all the AHB write data beats,

disregarding the B-channel AXI response. Therefore, if the network generates an error response, it does not feed it back to the master.

— The bridge can support up to five outstanding write accesses because the RAW

hazard detection function supports up to four transactions. A fifth write is issued, but the AHB write response is not issued until a slot is freed in the RAW hazard monitor.

• Write fixed length bursts with HPROT[3] negated to AXI singles, and each AHB write beat is acknowledged with the AXI buffered write response.

• Read INCR bursts with HPROT[3] asserted speculatively to AXI INCR4 bursts.

• Write INCR bursts with HPROT[3] asserted speculatively to AXI INCR4 bursts, and all AHB write data beat receive an automatic OKAY response from the bridge irrespective of the B-channel AXI response. Therefore, if the network generates an error response, it does not feed it back to the master.

• Read INCR bursts with HPROT[3] negated to a series of AXI singles.

• Write INCR bursts with HPROT[3] negated to a series of AXI singles, and each AHB write beat is acknowledged with the AXI buffered write response.

Combination 4

If you do not configure

broken bursts

then the network converts all:

INCR promotion and Early Write Response

and do not configure

allow

• read fixed length bursts to AXI fixed length bursts

• write fixed length bursts to AXI fixed length bursts, and only the last AHB write data beat receives the AXI buffered response for the whole AHB transaction

• read INCR bursts to a series of AXI singles

• write INCR bursts to a series of AXI singles, and each AHB write beat is acknowledged with the AXI buffered write response.

If you select either the

INCR promotion and Early Write Response

allow broken bursts

configuration options, or both, then the following programmable function override bits also exist and you configure a GPV port:

rd_incr_override

Converts all AHB read transactions to a series of AXI singles.

wr_incr_override

Converts all AHB write transactions to a series of AXI singles.

See Chapter 3 Programmers Model.

Error response

If the AHB master cancels a burst when it receives an ERROR response, the bridge stalls the master until the network receives all the read data beats from the AXI domain. This is only possible with read transfers because AXI writes receive a response at the end of the burst only.

Functional Description

Note

When communicating with transfer-sensitive slave devices such as FIFOs, the master might not be aware of how many read data beats have been read.

Lock transactions

The only supported lock transactions are SWP-like locks. That is, a single locking read followed by a single unlocking write, with an undefined number of IDLE transactions in between.

If the network receives a non-SWP-like lock sequence, it is possible for a network path to be stalled, particularly if an odd number of lock transactions is issued. The stall is cancelled on the next transaction received that unlocks the stalled path.

If you configure lock support and a GPV, then a lock override function is also configured. You can program this option, named

lock_override

, to force no AXI lock transactions to be created.

See Chapter 3 Programmers Model.

Configuration options

You can configure the following AHB options:

• AHB slave or master mirror interface types.

• Address width of 32-64 bits.

• Data width of 32. 64, 128, or 256 bits.

• Data width upsize function that Upsizing data width function on page 2-12 describes.

• Data width downsize function that Downsizing data width function on page 2-14 describes.

• Frequency domain crossing of the following types:

—ASYNC

—SYNC 1:1

—SYNC 1:n

—SYNC n:1

—SYNC n:m.

• Security of the following types:

Secure All transactions originating from this slave interface are flagged as secure

transactions and can access both secure and non-secure components.

Non-secure

All transactions originating from this slave interface are flagged as non-secure transactions and cannot access secure components.

•

INCR promotion and Early Write Response

• Permit broken bursts using the

allow broken bursts

parameter.

• Support for the full AHB-Lite protocol with only SWAP-like locks.

You can reduce the gate count and increase the performance if the attached master does

not create any AHB lock transactions.

• Timing isolation:

Note

2.2.2 Master interfaces

The AMBA Network Interconnect supports the following master interfaces:

• AXI master interfaces

• AHB master interfaces on page 2-9

• APB master interfaces on page 2-10.

AXI master interfaces

The network supports the full AXI protocol using an AXI master interface.

Configuration options

You can configure the following AXI options:

• Address width of 32-64 bits.

• Data width of 32, 64, 128, or 256 bits.

Functional Description

— from the external master

— from the network.

• Data width upsize function that Upsizing data width function on page 2-12 describes.

• User sideband signal width of 0-32 bits.

• Data width downsize function that Downsizing data width function on page 2-14 describes.

• Frequency domain crossing of type:

—ASYNC

—SYNC 1:1

—SYNC 1:n

—SYNC n:1

—SYNC n:m.

• Support for the full AXI protocol.

You can reduce the gate count and increase the performance if all attached master that can

access the master interface does not create any AXI lock transactions.

• Write issuing capability of 1-32 transactions.

• Read issuing capability of 1-32 transactions.

• Buffering that FIFO and clocking function on page 2-15 describes.

• Timing isolation:

— from the external slave

— from the network.

Functional Description

Note

AHB master interfaces

The network can support the full AHB-Lite master protocol and you can configure the network to provide an AHB-Lite mirrored slave protocol. Table 2-2 shows the mapping of AXI burst types to AHB burst types.

Table 2-2 AXI burst type to AHB burst type mapping

AxBURST AxLEN HBURST Notes

FIXED - SINGLE This is a series of singles and the number

depends on the AxLEN setting

INCR 1 SINGLE -

- 4 INCR4 -

- 8 INCR8 -

- 16 INCR16 -

- 2, 3, 5, 6, 7, 9, 10, 11, 12, 13, 14, 15 INCR Undefined length

WRAP 2 SINGLE Two transfers

-4 WRAP4-

-8 WRAP8-

-16 WRAP16-

Transactions from AHB slave interfaces are output as INCR transactions of an undefined length.

If the AHB protocol conversion function receives an unaligned address, or a write data beat without all the byte strobes set, the AMBA Network Interconnect detects it, and a programmable enable bit permits the network to create a DECERR response.

• If you set the

force_incr

programmable bit, see Table 3-3 on page 3-6, and a beat is received that has no write data strobes set, that write data beat is replaced with an IDLE beat.

• You can configure the inclusion of the programmable enable bit to create a reduced gate count implementation.

See Chapter 3 Programmers Model. The network still transmits the unaligned address transfer into the AHB domain, but it aligns the address by forcing the lower address bits of the transaction’s size to zeros.

The network breaks any transactions that cross a 1KB boundary into two AHB INCR bursts. You can configure a programmable option, named

force_incr

, see Table 3-3 on page 3-6, that

maps all transactions that are to be output to the AHB domain to be an undefined length INCR.

If the AXI burst is part of a locked sequence, the AHB-Lite translation keeps HMASTLOCK asserted across the boundary to ensure that the burst atomicity is not compromised. For write transactions, AHB responses are merged into a single AXI buffered response. The merged response is an AXI SLAVE ERROR if any of the AHB-Lite data beats have an AHB ERROR.

Functional Description

Note

Configuration options

You can configure the following options for the AHB interface:

• Address width of 32-64 bits.

• Data width of 32, 64, 128, or 256 bits.

• Data width upsize function that Upsizing data width function on page 2-12 describes.

• Data width downsize function that Downsizing data width function on page 2-14 describes.

• Frequency domain crossing of the following types:

—ASYNC

—SYNC 1:1

—SYNC 1:n

—SYNC n:1

—SYNC n:m.

• Security of the following types:

Secure Only secure transactions can access components attached to this master

interface.

Non-secure

Both secure and non-secure transactions can access components attached to this master interface.

Boot time secure

You can use software to configure whether it permits secure and non-secure transactions to access components attached to this master using the Secure and Non-secure options above.

• Support for the full AHB-Lite master protocol.

• Timing isolation:

— from the external slave

— from the network.

APB master interfaces

You can configure the APB interface to support a mixture of APB2 or APB3. The APB data width is always 32-bit, and it is therefore never necessary for the APB interface to require the upsize function. The APB interface can ignore AXI writes strobes. If the network receives a write transaction with all of the write strobes negated, then it does not perform the write.

APB SLVERR responses are converted to AXI SLVERR responses.

Any transaction that the network receives without all four WSTRBs asserted or negated still goes ahead. This means that erroneous data bytes might be written to the slave. The masters accessing the APB interface ensure that only WORD writes access the APB sub-system. The address and data widths are fixed as follows:

• address width of 32-bit

• data width of 32-bit.

Functional Description

Note

Although the AMBA Network Interconnect only outputs 32 address bits, you can configure the APB address of any peripheral to be anywhere in the address map.

Configuration options

You can configure the following options:

• data width downsize function that Downsizing data width function on page 2-14 describes

• frequency domain crossing for the majority of APB ports of the following types:

—ASYNC

—SYNC 1:1

—SYNC 1:n

—SYNC n:1

—SYNC n:m.

• buffering that FIFO and clocking function on page 2-15 describes

• 1-16 supported APB slaves

• configurable address region sizes

• non-contiguous address regions

• you can configure each APB slave for:

—APB2 or APB3

— asynchronous interface to the majority of APB ports.

• security of the following types:

— secure for each APB port

— non-secure for each APB port

— boot secure for all APB ports.

2.3 Operation

Note

This section describes how the AMBA Network Interconnect operates and contains the following subsections:

• Upsizing data width function

• Downsizing data width function on page 2-14

• FIFO and clocking function on page 2-15

• Arbitration on page 2-17

• Cyclic Dependency Avoidance Schemes (CDAS) on page 2-17

• Lock support on page 2-18

• TrustZone technology and security on page 2-19

• Remap on page 2-21.

2.3.1 Upsizing data width function

The upsizer function can expand the data width by the following ratios:

•1:2

•1:4

•1:8.

Functional Description

Upsizing only packs write data for write or read transactions that are cacheable. This section describes the packing rules for different burst types and acceptance capabilities, and the following definitions apply:

• an aligned input burst means that the address is aligned to the output data width word boundary, after the network aligns it to the size of the transfer

• an unaligned input burst means that the network does not align the address to the output data width word boundary, even after it aligns it to the size of the transfer

• if a transaction passes through, this means that the upsize function does not change the input transaction size and type.

• If the network splits input exclusive transactions into more than one output bus transaction, it removes the exclusive information from the multiple transactions it creates.

• If multiple responses from created transactions are combined into one response, then the order of priority is:

— DECERR is the highest priority

— SLVERR is the next highest priority

— OKAY is the lowest priority.

In the examples in this section, the input data width is 64-bit, and the output data width is 128-bit, unless otherwise stated. This section describes:

• INCR bursts on page 2-13

• WRAP bursts on page 2-13

• Fixed bursts on page 2-13

• Bypass merge on page 2-13

• Acceptance capability on page 2-14.

Functional Description

Note

INCR bursts

The network converts all input INCR bursts that complete within a single output data width into an INCR1 of the minimum SIZE possible, and it packs all INCR bursts into INCR bursts of the optimum size possible. Table 2-3 shows how the network converts INCR bursts when it upsizes them.

Table 2-3 Conversion of INCR bursts by the upsize function

INCR burst type Converted to

64-bit INCR1 Passes through unconverted

64-bit aligned INCR2 INCR1

8-bit aligned INCR8 INCR1, 128-bit

8-bit unaligned, byte address 1, 2, or 3, INCR5 INCR1, 128-bit

8-bit unaligned, byte address 4, 5, 6, or 7, INCR5 INCR2, 64-bit

64-bit unaligned INCR2 Passes through unconverted

64-bit aligned INCR4 INCR2

64-bit unaligned INCR4 Sparse INCR3

WRAP bursts

All WRAP bursts are either passed through unconverted as WRAP bursts, or converted to one or two INCR bursts of the output bus. Table 2-4 shows how the network converts WRAP bursts when it upsizes them from 64-bit to 128-bit, that is, a ratio of 1:2.

Table 2-4 Conversion of WRAP bursts by the upsize function

WRAP burst type Converted to

64-bit aligned WRAP2 INCR1

64-bit aligned WRAP4 WRAP2

64-bit unaligned WRAP4 Depending on the address:

• INCR2 + INCR1

• INCR1 + INCR2

The network converts input WRAP bursts with a total payload that is less than the output data width to a single INCR.

Fixed bursts

All FIXED bursts pass through unconverted.

Bypass merge

You can configure the upsizer function to have a programmable bit named

bypass_merge

is asserted, the network does not alter any transactions that could pass through

bypass_merge

. If

legally without alteration.

Acceptance capability

You can configure the upsizer to support 1-32 read transactions and 1-32 write transactions. The issuing capability is a maximum of twice the acceptance capability.

2.3.2 Downsizing data width function

The downsize function reduces the data width by the following ratios:

•2:1

•4:1

•8:1.

The downsizer does not merge data narrower than the destination bus if the transaction is marked as non-cacheable.

This section describes the following:

• INCR bursts

• WRAP bursts

• FIXED bursts on page 2-15

• Bypass merge on page 2-15

• Acceptance capability on page 2-15.

Functional Description

INCR bursts

The AMBA Network Interconnect converts INCR bursts that fall within the maximum payload size of the output data bus to a single INCR. It converts INCR bursts that are greater than the maximum payload size of the output data bus to multiple INCR bursts. Table 2-5 shows how the network converts INCR bursts when it downsizes them.

Table 2-5 Conversion of INCR bursts by the downsize function

INCR burst type Converted to

Aligned INCR4 INCR8

Unaligned INCR4 INCR7

Aligned INCR9 INCR15 + INCR2

INCR bursts with a size that matches the output data width pass through unconverted.

The AMBA Network Interconnect packs INCR bursts with a SIZE smaller than the output data width to match the output width whenever possible, using the upsize transfer function. See Upsizing data width function on page 2-12.

WRAP bursts

The AMBA Network Interconnect always converts WRAP bursts to WRAP bursts of twice the length, up to the output data width maximum size of WRAP16, in which case, it treats the WRAP burst as two INCR bursts that can each map onto one or more INCR bursts.

Functional Description

Note

FIXED bursts

The AMBA Network Interconnect converts FIXED bursts to one or more INCR1 or INCRn bursts depending on the downsize ratio. Table 2-6 shows how the network converts FIXED bursts when it downsizes them.

Table 2-6 Conversion of FIXED bursts by the downsize function

FIXED burst type Converted to

FIXED1 INCR2

FIXED2 INCR2 + INCR2 + ...

The AMBA Network Interconnect optimizes unaligned fixed bursts. If an unaligned input fixed burst maps onto a single output beat, then the output is a fixed burst of the optimal size.

Bypass merge

You can configure the downsizer function to have a programmable bit named

bypass_merge

is asserted, the network does not perform any packing of beats to match the

bypass_merge

. If

optimum SIZE, up to the output data width SIZE.

An aligned input burst means that the address is aligned to the input data width word boundary after the network aligns it to the transfer size. An unaligned input burst means that the address is not aligned to the input data width word boundary, even after the network aligns it to the transfer size.

If a transaction passes through, this means that the downsize function does not change the input transaction size and type.

• If an exclusive transaction is split into multiple transactions at the output of the downsizer, the exclusive flag is removed and the master never receives an EXOKAY response. Response priority is the same as for the upsize function. See Upsizing data width function on page 2-12.

• In the following example, the input data width is 128-bit and the output data width is 64-bit unless otherwise stated.

Acceptance capability

You can configure the acceptance capability to 1-32 read transactions and 1-32 write transactions. The maximum issuing capability is (ratio x acceptance capability +1).

2.3.3 FIFO and clocking function

If you configure the network as a clock frequency crossing bridge, then a FIFO function is also configured.

You can configure the buffering for multiple outstanding transactions even if you are using a 1:1 clocking ratio.

Functional Description

Note

You can instantiate a FIFO on any channel. You can configure the FIFO to implement both buffering and clock domain crossing functionality. You can define the FIFO to be:

•SYNC 1:1

•1:n

•n:1

•ASYNC m:n

•SYNC m:n

You can dynamically select this from the ASYNC m:n option if you configure a GPV.

• programmable.

The network automatically determines that the width of the FIFO is the width of the payload. You can configure the depth of the FIFO to be 2-32.

All clock boundary crossings are implemented using a FIFO structure with circular read and write pointers, to give one cycle of latency. Data is clocked out when the pointers do not match on the read side.

For ASYNC, use a full synchronizer, that is, two or more cycles depending on metastability resolution, and the depth of the synchronizer, if it has changed.

For synchronous modes, use a single flop to synchronize each pointer, when the ratio is not 1, that is, both sides of m:n, and the slower side of m:1 and n:1.

For a 1:1 ratio, no extra synchronization is performed on either side.

Changing the synchronization when you select programmable mode

You can change the boundary type by modifying the synchronization that is applied to the two pointers as they pass between domains. This ensures that the data in the FIFO is stable and safe to use.

To change the clocks, the synchronization must remain correct at all times. Table 2-7 shows the actions you must take to convert from one mode to another.

Table 2-7 How to change modes

Original mode Required mode Action

ASYNC Any other mode Change the clocks then change the register.

Any mode ASYNC Change the register then change ASYNC. BRESP

from the GPV implies that the update is complete.

m:n 1:1 Change the clocks, then change the register.

1:1 m:n Change the register, then change the clocks.

For some changes, it is necessary to use a different setting, that is, you can only change safely from 1:n to 1:m by first programming the register to m:n, before the clock update.

2.3.4 Arbitration

Functional Description

Data release mechanism

When you configure a write data FIFO of at least 4, you can also set an additional write tidemark function, named

wr_tidemark

. This is a tidemark level that stalls the release of the transaction

until:

• The network receives the WLAST beat.

• The write FIFO becomes full.

• The number of occupied slots in the write data FIFO exceeds the write tidemark. See Chapter 3 Programmers Model.

You can program the arbitration algorithm for all arbitration nodes within the infrastructure.

At the entry point to the infrastructure, all transactions are allocated a local QoS that you can configure to be:

• static

• programmable

• received from the attached master, for AXI only.

The arbitration of the transaction throughout the infrastructure uses this QoS. See Chapter 3 Programmers Model.

At any arbitration node, a fixed priority exists for transactions with a different QoS. The highest value has the highest priority. If there are coincident transactions at an arbitration node with the same QoS that require arbitration, then the Network uses a Least Recently Used (LRU) algorithm.

2.3.5 Cyclic Dependency Avoidance Schemes (CDAS)

Because the AXI protocol permits re-ordering of transactions, it might be necessary for the AMBA Network Interconnect to enforce rules to prevent deadlock when routing multiple transactions concurrently to multiple slaves from a single point of divergence, that is, at a switch slave interface.

Each slave interface of a switch can have a different CDAS configured. The same CDAS scheme is configured for both read and write transactions, but they operate independently.

This section describes:

• Single slave

• Single slave per ID.

Single slave

This ensures that at a slave interface of a switch:

• all outstanding read transactions are to a single end destination

• all outstanding write transactions are to a single end destination.

If the slave interface receives a transaction to a different destination to the current destination for that transaction type, the network stalls the transactions until all the outstanding transactions of that type have been completed.

Single slave per ID

This ensures that at a slave interface of a switch:

• all outstanding read transactions with the same ID go the same destination

Functional Description

Note

• all outstanding write transactions with the same ID go the same destination.

When the slave interface receives a transaction:

• if it has an ID that does not match any outstanding transactions, it passes the CDAS

• if it has an ID that matches the ID of an outstanding transaction, and the destinations also match, it passes the CDAS

• if it has an ID that matches the ID of an outstanding transaction, and the destinations do not match, it fails the CDAS check and is stalled.

A stalled transaction remains stalled until one of the rules passes. For complex infrastructures, it might be necessary to configure an additional rule for write transactions.

This is only for the single-slave-per-ID CDAS configured slave interfaces.

The extended write rule ensures that if single-slave-per-ID rules are passed, then the network only issues a write transaction to a new destination if all the outstanding write transactions have had the last write data beat transmitted.

The AMBA Designer tool automatically detects when this is required. See Additional reading on page viii.

2.3.6 Lock support

You set support for locked transaction for masters and slaves at configuration time. The AMBA Network Interconnect infrastructure is configured for lock support into all switch master interfaces that are required to provide lock support for all the relevant masters and slaves in the system. See Lock transactions on page 2-7 for information on the scope of lock support for AHB-Lite slave interfaces.

At a switch master interface with lock support, logic exists that:

• Stalls a locked transaction after it is arbitrated.

If a co-incident transaction exists on the other address channel, it is not stalled unless it is

a lock transaction.

• Stalls the other address channel.

• Permits all the outstanding transactions to complete.

• Enables the locking transactions source read and write channels when there are no outstanding transactions.

• Enables all sources for arbitration, and normal operation continues when an unlocking transaction completes.

• When the network receives a locking transaction, if there is a co-incident lock transaction on the other address channel, then the read always takes priority, and the write address transaction is stalled.

Note

The NIC supports lock functionality for 32-bit data beat accesses. You can lock beats of other

Note

sizes, but if they are up-sized or down-sized, it is possible that leading write data are output from the sizing function for the unlocking transaction before all the locked transactions have completed.

2.3.7 TrustZone technology and security

This section applies if you are building a system based on the secure and non-secure capabilities that TrustZone technology provides. If the system does not require security using TrustZone technology, configure all master interfaces to be non-secure.

This section contains the following subsections:

• TrustZone scope

• Slave interface security on page 2-20

• Internal programmers view on page 2-20

• Security checking for master interfaces on page 2-20.

TrustZone scope

Functional Description

The security checks that TrustZone technology implements cover the scope of a configured network.

TrustZone is a brand name that represents aspects of implementing ARM security extensions.

For example, security checks that are not within the scope of the network are:

Physical attack

Physical attack on the device.

Non-TrustZone-aware masters being made secure

A master might require access to the Global Programmers View (GPV) and in this case, you can tie the security transaction indicator bits so that all accesses by that master are indicated as secure. This places that master permanently in the secure domain. However, depending on the other usage of that master, this might mean that the overall system is not as secure under all circumstances.

System implementation details

If you do not consider all the masters that have access to the GPV, this can produce security vulnerabilities. For example:

• If a non-secure state master can set QoS requirements effecting its non-secure transactions, then that non-secure state master can use this capability, in conjunction with traffic analysis, to determine the QoS and priority settings of a secure master. This can be a threat in particular implementations.

• A TrustZone-aware slave requires you to set the connecting network as non-secure so that the network does not filter the secure traffic and leaves the slave to determine the correct response. Consider the master that can make this non-secure configuration against and the master, or masters, that can program the TrustZone-aware slave.

Functional Description

Note

Topology issues

It might be possible to suffer timing attacks because of the topology configuration you chose. For example, if two cascaded switches exist with a shared AXI link between them, then continuous non-secure accesses to a non-secure slave might block secure transactions to a different secure slave.

Resets It might be possible to carry out a secure attack by resetting only parts of a data

path, whether it be a data path section in an individual clock domain within a network, or within a master or slave.

Slave interface security

At configuration time, each slave interface, whether it belongs to the AXI or AHB protocol, has the following options for setting the security assignment of all its transactions:

• input from the external master, for AXI masters only

• tied-off to always issue transactions as secure

• tied-off to always issue transactions as non-secure.

Internal programmers view

The programmers view is always secure access only. Any non-secure transaction intended to access a register, input to a configuration, returns a DECERR, and no register access is provided.

If you configure a dedicated configuration port to gain access to the GPV, then you must connect it to a secure master, or have a security check that is external to the AMBA Network Interconnect.

Security checking for master interfaces

You can configure each master interface to be:

Always secure

The master rejects non-secure transactions.

Always non-secure

The master accepts both secure and non-secure transactions.

Boot secure You can use software to configure whether it permits secure and non-secure

transactions to access components attached to this master using the Always secure and Always non-secure options above.

• If you change the security of a master interface, the change does not occur simultaneously for all the masters in the system because of the distributed nature of the GPV.

• Outstanding transactions, or active lock sequences, underway within the network at the time of the security update use the old security settings for their security check.

For an APB master interface, where multiple slaves exist on a single interface, each APB slave has its own security check.

If an incoming transaction is non-secure, either because the slave interface is configured to be non-secure, or the input security bit is set be non-secure, then if that transaction is intended for a master interface that is currently secure, then that transaction is returned with a DECERR, and the transaction is not transferred to the slave.

2.3.8 Remap

Note

Functional Description

All accesses must be secure to gain access to any programmers model register. Any non-secure accesses to the programmers model receive a DECERR response. See Chapter 3 Programmers Model.

Security registers are not updated if a pending transaction exists, or if a current ongoing lock sequence exists.

Registers in the programmers model control the remap functionality. See Table 3-4 on page 3-7 in Chapter 3 Programmers Model for more information.

You can define a number of remap states using eight bits of the remap register, and a bit in the remap register controls each remap state.

You can use each remap state to control the address decoding for one or more slave interfaces. If a slave interface is affected by two remap states that are both asserted, the remap state with the lowest remap bit number takes precedence.

You can configure each slave interface independently so that a remap state can perform different functions for different masters.

A remap state can:

• alias a memory region into two different address ranges

• move an address region

• remove an address region.

Because of the nature of the distributed register sub-system, the masters receive the updated remap bit states in sequence, and not simultaneously.

A slave interface does not update to the latest remap bit setting until:

• the address completion handshake accepts any transaction that is pending

• any current lock sequence completes.

The BRESP from a GPV after a remap update guarantees that the next transaction issued to each slave interface, or the first one after the completion of a locked sequence, uses the updated value.

Figure 2-1 on page 2-22 to Figure 2-5 on page 2-23 show examples of how different remap states interact with each other. Consider a configuration that uses three remap bits. Figure 2-1 on page 2-22 shows the memory map when remap is set to 000, representing no remap.

Functional Description

Note

Slave 0 region 0

Slave 0 region 1

Slave 1

Slave 2

Slave 0 region 0

Slave 3 region 1

Slave 0 region 0

Slave 0 region 1

Slave 1

Slave 2

Slave 3 region 0

Slave 3 region 1

Slave 0 region 0

Slave 0 region 1

Slave 2

Slave 1

Figure 2-1 No remap, remap set to 000

This has a default memory map that divides slave 0 and slave 3 into two separate regions. At power-up, slave 0 region 0 is aliased over slave 3 region 0. After power-up, the slave 0 region 0 alias can be removed as Figure 2-2 shows.

Figure 2-2 Remap set to 001

Alternatively, you can move slave 1 to the bottom of the address range by setting remap to 010 as Figure 2-3 shows.

Figure 2-3 Remap set to 010

Remap bit 0 still takes precedence if you set it as Figure 2-4 on page 2-23 shows.

Functional Description

Slave 0 region 0

Slave 0 region 1

Slave 2

Slave 3 region 0

Slave 1

Slave 0 region 0

Slave 0 region 1

Slave 2

Slave 3 region 0

Slave 3 region 1

Figure 2-4 Remap set to 011

In addition, you can remove memory regions entirely. Figure 2-5 shows that if you set remap to 101, Slave 1 is removed.

Figure 2-5 Remap set to 101

Chapter 3

Programmers Model

This chapter describes the programmers model.

It contains the following sections:

• About this programmers model on page 3-2

• Configuration programmers model on page 3-3.

3.1 About this programmers model

This chapter describes the architecture of the AMBA Network Interconnect AMBA infrastructure component. It describes the programmers interface and system characteristics.

Programmers Model

3.2 Configuration programmers model

Note

Slave interface m registers

Slave interface 2 registers

Slave interface 1 registers

Slave interface 0 registers

Master interface n registers

Master interface 2 registers

Master interface 1 registers

Master interface 0 registers

ID registers

Address control registers

Maximum m = 127

Maximum n = 63

0x000C2000

0x000C1000

0x00042000

0x00044000

0x00043000

0x00041000

0x00003000

0x00002000

0x00001000

0x00000000 Configurable base address

0x00004000

Internal interface 0 registers

Internal interface 1 registers

Internal interface 2 registers

Internal interface p registers

0x000C3000

0x000C4000

0x000FF000

Maximum p = 61

Address offset

from base

0x00100000

The AMBA Network Interconnect can contain configuration registers, partitioned into a number of individual 4KB blocks that you can program using the GPV. The base address of each GPV region is set at configuration time in AMBA Designer (ADR-301).

You can configure any slave interface to have access to all of the registers in the programmers view.

Any registers that a switch requires are implemented within the register block of the associated Interface Block (IB) register block. If no IB is attached, then you can configure an IB to specifically provide programmable registers.

Ensure that you access the Global Programmers View (GPV) using non-cacheable transactions.

3.2.1 Register block types

The following types of register block exist:

• one register block for each AMBA Network Interconnect configuration

• one register block for each IB, where the IB can be:

— AXI Slave Interface Block (ASIB), see Table 3-1 on page 3-4

— AXI Master Interface Block (AMIB), see Table 3-2 on page 3-5

— AXI internal network Interface Block (IB), see Table 3-3 on page 3-6.

Programmers Model

Figure 3-1 shows the address map of the programmers model. It contains one fixed base address, and all the other programmers model 4KB blocks are stacked.

Figure 3-1 Address map of the programmers model

Programmers Model

Note

The type defines the number of register blocks in a single AMBA Network Interconnect configuration. Table 3-1, Table 3-2 on page 3-5, and Table 3-3 on page 3-6 show the register block sub-types for each of the main types.

Table 3-4 on page 3-7 shows the address region control registers and Table 3-5 on page 3-8 shows the peripheral ID registers.

In Table 3-1 to Table 3-5 on page 3-8, reserved means:

• read as zeros

• writes are ignored.

AHB only means that this register is interpreted as reserved if the interface is not AHB.

Table 3-1 shows the registers that exist for each ASIB.

Table 3-1 Registers for each ASIB

Address offset

0x000

0x004

0x008

0x00C

0x020

0x024

0x028

Typ e W i dth

Reset value

Name Description

-- - - Reserved.

RW 3 4

sync_mode

This register is only present if you configure the block as a programmable FIFO. You can configure the register bits as follows:

0 sync 1:1.

1 sync n:1.

2 sync 1:n.

3 sync m:n.

4 async.

5 reserved.

6 reserved.

7 reserved.

RW 1 0

fn_mod2

Bypass merge. This register is only present if upsizing or downsizing, see

Upsizing data width function on page 2-12, Downsizing data width function on page 2-14, and Bypass merge on page 2-13.

RW 3 0

fn_mod_ahb

This register is valid for AHB interfaces only. You can configure the register bits as follows:

rd_incr_override.

wr_incr_override

lock_override.

See Lock transactions on page 2-7 for information on overriding locks. See Combination 4 on page 2-6 for information on

rd_incr_override

wr_incr_override

and

0x02C

0x03C

0x040

-- - - Reserved.

RW 4

wr_tidemark

Valid only with a FIFO for the WFIFO channel, and if not an AHB slave interface. See FIFO and clocking function on page 2-15 for information on

wr_tidemark

Programmers Model

Table 3-1 Registers for each ASIB (continued)

Address offset

0x044

0x0FC

0x100

0x104

0x108

Typ e W i dth

-- - - Reserved.

RW 4 0

RW 2 0

Reset value

Name Description

read_qos

write_qos

fn_mod_iss

Read channel QoS value.

Write channel quality value.

Issuing functionality modification register. This register sets the block issuing capability to one outstanding transaction. You can configure the register bits as follows:

0 Read issuing,

1 Write issuing,

0x10C

0xFFC

a. The reset value is initialized to the tidemark value that you set in the configuration GUI in AMBA Designer (ADR-301).

-- - - Reserved.

read_iss_override

write_iss_override

Table 3-2 shows the registers that exist for each IB.

Table 3-2 Registers for each IB

Address offset

0x000

Typ e W i dth

-- - - Reserved.

Reset value

Name Description

0x004

0x008

0x00C

0x020

0x024

-- - - Reserved.

RW 2 0

fn_mod_bm_iss

Bus matrix issuing functionality modification register. This register is only present if the block is connected directly to a switch.

This register sets the issuing capability of the preceding switch arbitration scheme to 1. You can configure the register bits as follows:

0 Read issuing,

1 Write issuing,

-- - - Reserved.

RW 3 4

Sync_mode

This register is only available if you have a FIFO for all channels. You can configure the register bits for the following clock domain boundaries:

0 sync 1:1.

1 sync n:1.

2 sync 1:n.

3 sync m:n.

4 async.

5 reserved.

6 reserved.

7 reserved.

RW 1 0

fn_mod2

Bypass merge. This register is only present if upsizing or downsizing. See Upsizing data width function on page 2-12, Downsizing data width function on page 2-14, and Bypass merge on page 2-13.

read_iss_override

write_iss_override

0x028

0x03C

0x040

-- - - Reserved.

RW 4

wr_tidemark

Value, only with a FIFO for the WFIFO channel.

Programmers Model

Table 3-2 Registers for each IB (continued)

Address offset

0x044

0x100

0x104

0x108

Typ e W i dth

-- - - Reserved.

RW 2 0

Reset value

Name Description

fn_mod_iss

Issuing functionality modification register.

Issuing override, sets block issuing capability to one transaction and you can configure the bits as follows:

0 Read issuing,

1 Write issuing,

0x10C

0x110

a. The reset value is initialized to the tidemark value that you set in the configuration GUI in AMBA Designer (ADR-301).

-- - - Reserved

read_iss_override

write_iss_override

Table 3-3 shows the registers that exist for each AMIB.

Table 3-3 Registers for each AMIB

Address offset

Type W idt h

Reset value

Name Description

0x000

0x004

0x008

0x00C

0x020

0x024

---- Reserved.

RW 2 0

fn_mod_bm_iss

Bus matrix issuing functionality modification register. This register is only present if the block is connected directly to a switch.

This register sets the issuing capability of the preceding switch arbitration scheme to 1. You can configure the register bits as follows:

0 Read issuing,

1 Write issuing,

---- Reserved.

RW 3 4

Sync_mode

This register is only available if you have a FIFO for all channels. You can configure the register bits to create different clock domain boundaries as follows:

0 sync 1:1.

1 sync n:1.

2 sync 1:n.

3 sync m:n.

4 async.

5 reserved.

6 reserved.

7 reserved.

RW 1 0

fn_mod2

Bypass merge. This register is only present if upsizing or downsizing. See

Upsizing data width function on page 2-12 and Downsizing data width function on page 2-14.

read_iss_override

write_iss_override

0x028

---- Reserved.

Programmers Model

Note

Table 3-3 Registers for each AMIB (continued)

Address offset

0x040

0x044

Type W idt h

RW 4

RW 2 -

Reset value

Name Description

wr_tidemark

ahb_cntl

Reserved.

This register is available for AHB only. You can configure the register bits as follows:

decerr_en

force_incr

See AHB master interfaces on page 2-9.

0x100

0x104

0x108

---- Reserved.

RW 2 0

fn_mod_iss

Issuing functionality modification register. This register is only available if you are upsizing or downsizing, or you have a FIFO for any of the channels. This register sets the block issuing capability to be forced to one transaction. You can configure the register bits as follows:

0 Read issuing,

1 Write issuing,

a. The reset value is initialized to the tidemark value that you set in the configuration GUI in AMBA Designer (ADR-301).

read_iss_override

write_iss_override

3.2.2 Register blocks

This section contains the following subsections:

• Address region control

• Peripheral ID registers on page 3-8.

Address offset

0x0

0x4

0xC

0x10

Address region control

Table 3-4 shows the address region control registers.

Type W idt h

WO 8 0x00

WO - - - Reserved.

WO 1 - 16 -

Reset value

Name Description

Remap

security0

Remap register. Up to eight global remap states are available.

Slave 0 security setting. This consists of one bit for non-virtual slaves, and up to 16 bits for virtual or APB master interfaces, and you can configure the register bits as follows:

0 Secure.

1 Non-secure.

For virtual or APB master interfaces with 16 security setting bits, each bit position maps onto the region number. For example, the security1[5] bit is the security setting for the address region for master interface node number 1, region 5.

Table 3-4 Address region control registers

Programmers Model

Note

Table 3-4 Address region control registers (continued)

Address offset

0x14

0x18

0x10C

0x110

0xFFF

Type W idt h

WO 1 - 16 -

Reset value

Name Description

security1

Slave 1 security setting. This consists of one bit for non- virtual slaves, and up to 16 bits for virtual or APB master interfaces, and you can configure the register bits as follows:

0 Secure.

1 Non-secure.

WO 1 - 16 -

security<n>

Slave n security setting. It contain one bit for non-virtual slaves, and up to 16 bits for APB master interfaces.

RO - - - Reserved.

A configuration can contain a maximum of 64 security registers, that is, 1 < n < 64. Therefore, if the configuration contains 64 master interfaces, then register security 63 is registers are write-only because they are global accesses on the GPV.

Peripheral ID registers

0x10C

. These

Address offset

0x0

0xFCC

0xFD0

0xFD4

0xFD8

0xFDC

0xFE0

0xFE4

0xFE8

0xFEC

0xFF0

If you configure any registers in the programmers view, peripheral ID registers are always created. This provides a low gate count option for identification. Table 3-5 shows the peripheral ID registers.

Type W idt h

RO - - - Reserved

RO 8

Reset value

0x04

0x00

0x01

0xB3

0x4B

0x00

0x0D

Name Description

Peripheral ID4 4KB count, JEP106 continuation code

Peripheral ID5 Reserved

Peripheral ID6 Reserved

Peripheral ID7 Reserved

Peripheral ID0 Part Number [7:0]

Peripheral ID1 JEP106[3:0], part number [11:8]

Peripheral ID2 Revision, JEP106 code flag, JEP106[6:4]

Peripheral ID3 You can set this using the AMBA Designer Graphical User Interface

(GUI)

Component ID0 Preamble

Table 3-5 Peripheral ID registers

Programmers Model

Note

Table 3-5 Peripheral ID registers (continued)

Address offset

0xFF4

0xFF8

0xFFC

Type W idt h

RO 8

Reset value

0xF0

0x05

0xB1

Name Description

Component ID1 Generic IP component class, preamble

Component ID2 Preamble

Component ID3 Preamble

It is possible for AMIB registers to exist in regions that are not contiguous, for example, you could have a register named AMIB_0 in the region in the region

0xA000

0x4000

, and have a register named AMIB_1

Appendix A

Revisions

This appendix describes the technical changes between released issues of this book.

Table A-1 Differences between issue E and issue F

Change Location Affects

Network of interconnects instead of a single interconnect Throughout the document r2p0

Multiple outstanding downsizer transactions instead of limited outstanding downsizer transactions

Upsizer packs data into the wide bus, instead of an expander, for data width changes Throughout the document r2p0

Global dynamic QoS instead of local static QoS Throughout the document r2p0

Internally programmable instead of externally programmed Throughout the document r2p0

System-level address map replaces an address map for each interconnect Throughout the document r2p0

Extended timing closure options replace limited timing closure options Throughout the document r2p0

256-bit maximum data width instead of 128-bit maximum data width Throughout the document r2p0

Multi-region slave support replaces a single region per slave Throughout the document r2p0

Write FIFO, transaction release control instead of fixed write transaction release point Throughout the document r2p0

Throughout the document r2p0

Revisions

Table A-2 Differences between issue F and issue G

Change Location Affects

No technical changes - -

Glossary

This glossary describes some of the terms used in technical documents from ARM.

Advanced eXtensible Interface (AXI)

A bus protocol that supports separate address/control and data phases, unaligned data transfers using byte strobes, burst-based transactions with only start address issued, separate read and write data channels to enable low-cost DMA, ability to issue multiple outstanding addresses, out-of-order transaction completion, and easy addition of register stages to provide timing closure.

The AXI protocol also includes optional extensions to cover signaling for low-power operation.

AXI is targeted at high performance, high clock frequency system designs and includes a number of features that make it very suitable for high speed sub-micron interconnect.

Advanced High-performance Bus (AHB)

A bus protocol with a fixed pipeline between address/control and data phases. It only supports a subset of the functionality provided by the AMBA AXI protocol. The full AMBA AHB protocol specification includes a number of features that are not commonly required for master and slave IP developments and ARM recommends only a subset of the protocol is usually used. This subset is defined as the AMBA AHB-Lite protocol.

See also Advanced Microcontroller Bus Architecture and AHB-Lite.

Advanced Microcontroller Bus Architecture (AMBA)

A family of protocol specifications that describe a strategy for the interconnect. AMBA is the ARM open standard for on-chip buses. It is an on-chip bus specification that describes a strategy for the interconnection and management of functional blocks that make up a System-on-Chip (SoC). It aids in the development of embedded processors with one or more CPUs or signal processors and multiple peripherals. AMBA complements a reusable design methodology by defining a common backbone for SoC modules.

Glossary

Note

AXI

interconnect

Write address channel (AW)

Write data channel (W)

Write response channel (B)

Read address channel (AR)

Read data channel (R)

Write address channel (AW)

Write data channel (W)

Write response channel (B)

Read address channel (AR)

Read data channel (R)

AXI slave

interface

AXI master

interface

AXI

master

AXI

slave

AXI master

interface

AXI slave

interface

Advanced Peripheral Bus (APB)

A simpler bus protocol than AXI and AHB. It is designed for use with ancillary or general-purpose peripherals such as timers, interrupt controllers, UARTs, and I/O ports. Connection to the main system bus is through a system-to-peripheral bus bridge that helps to reduce system power consumption.

AHB See Advanced High-performance Bus.

AHB-Lite A subset of the full AMBA AHB protocol specification. It provides all of the basic functions

required by the majority of AMBA AHB slave and master designs, particularly when used with a multi-layer AMBA interconnect. In most cases, the extra facilities provided by a full AMBA AHB interface are implemented more efficiently by using an AMBA AXI protocol interface.

Aligned A data item stored at an address that is divisible by the number of bytes that defines the data size

is said to be aligned. Aligned words and halfwords have addresses that are divisible by four and two respectively. The terms word-aligned and halfword-aligned therefore stipulate addresses that are divisible by four and two respectively.

AMBA See Advanced Microcontroller Bus Architecture.

APB See Advanced Peripheral Bus.

AXI See Advanced eXtensible Interface.

AXI channel order and interfaces

The block diagram shows:

• the order in which AXI channel signals are described

• the MI and SI conventions for AXI components.

AXI terminology The following AXI terms are general. They apply to both masters and slaves:

Active read transaction

A transaction for which the read address has transferred, but the last read data has not yet transferred.

Active transfer

A transfer for which the xVALID handshake has asserted, but for which xREADY has not yet asserted.

The letter x in the signal name denotes an AXI channel as follows:

AW Write address channel.

W Write data channel.

B Write response channel.

AR Read address channel.

R Read data channel.

Glossary

Active write transaction

A transaction for which the write address or leading write data has transferred, but the write response has not yet transferred.

Completed transfer

A transfer for which the xVALID/xREADY handshake is complete.

Payload The non-handshake signals in a transfer.

Transaction An entire burst of transfers, comprising an address, one or more data transfers and

a response transfer (writes only).

Transmit An initiator driving the payload and asserting the relevant xVALID signal.

Transfer A single exchange of information. That is, with one xVALID/xREADY

handshake.

The following AXI terms are MI attributes. To obtain optimum performance, they must be specified for all components with an AXI MI:

Combined issuing capability

The maximum number of active transactions that a master interface can generate. It is specified for master interfaces that use combined storage for active write and read transactions. If not specified then it is assumed to be equal to the sum of the write and read issuing capabilities.

Read ID capability

The maximum number of different ARID values that an MI can generate for all active read transactions at any one time.

Read ID width

The number of bits in the ARID bus.

Read issuing capability

The maximum number of active read transactions that an MI can generate.

Write ID capability

The maximum number of different AW I D values that an MI can generate for all active write transactions at any one time.

Write ID width

The number of bits in the AW I D and WID buses.

Write interleave capability

The number of active write transactions for which the MI is capable of transmitting data. This is counted from the earliest transaction.

Write issuing capability

The maximum number of active write transactions that an MI can generate.

Glossary

The following AXI terms are SI attributes. To obtain optimum performance, they must be specified for all components with an AXI SI:

Combined acceptance capability

The maximum number of active transactions that a slave interface can accept. It is specified for slave interfaces that use combined storage for active write and read transactions. If not specified then it is assumed to be equal to the sum of the write and read acceptance capabilities.

Read acceptance capability

The maximum number of active read transactions that an SI can accept.

Read data reordering depth

The number of active read transactions for which an SI can transmit data. This is counted from the earliest transaction.

Write acceptance capability

The maximum number of active write transactions that an SI can accept.

Write interleave depth

The number of active write transactions for which the SI can receive data. This is counted from the earliest transaction.

Beat Alternative word for an individual transfer within a burst. For example, an INCR4 burst

comprises four beats.

ARM NIC-301 User Manual

Specifications and Main Features

Frequently Asked Questions

User Manual

Contents

List of Tables

List of Figures

Preface

About this book

Product revision status

Intended audience

Using this book

Conventions

Additional reading

Feedback

Feedback on this product

Feedback on content

Introduction

1.1 About the AMBA Network Interconnect

1.2 Key features

1.3 Relationship between AMBA Network Interconnect and AMBA Designer

1.3.1 AMBA Network Interconnect documentation

1.3.2 AMBA Designer documentation

1.4 Product revisions

Functional Description

2.1 About the functions

2.2 Interfaces

2.2.1 Slave interfaces

2.2.2 Master interfaces

2.3 Operation

2.3.1 Upsizing data width function

2.3.2 Downsizing data width function

2.3.3 FIFO and clocking function

2.3.4 Arbitration

2.3.5 Cyclic Dependency Avoidance Schemes (CDAS)

2.3.6 Lock support

2.3.7 TrustZone technology and security

2.3.8 Remap

Programmers Model

3.1 About this programmers model

3.2 Configuration programmers model

3.2.1 Register block types

3.2.2 Register blocks

Revisions

Glossary