Digital PDFs

EK-KA660-TM-001

May 2000

221 pages

Original

0.8MB

Document:	KA660 CPU Module Technical Manual
Order Number:	EK-KA660-TM
Revision:	001
Pages:	221
Original Filename:

OCR Text

KA660 CPU Module Technical Manual
Order Number: EK-KA660-TM-001

Digital Equipment Corporation

The information in this document is subject to change without notice and should not be construed as a
commitment by Digital Equipment Corporation.
Digital Equipment Corporation assumes no responsibility for any errors that may appear in this document.
Any software described in this document is furnished under a license and may be used or copied only in
accordance with the terms of such license. No responsibility is assumed for the use or reliability of software
or equipment that is not supplied by Digital Equipment Corporation or its affiliated companies.
Restricted Rights: Use, duplication, or disclosure by the U.S. Government is subject to restrictions as set
forth in subparagraph (c)(1) (ii) of the Rights in Technical Data and Computer Software clause at DFARS
252.227-7013.
© Digital Equipment Corporation March 1990. All rights reserved.
Printed in U.S.A.
The Reader’s Comments form at the end of this document requests your critical evaluation to assist in
preparing future documentation.

The following are trademarks of Digital Equipment Corporation:
DEC
DEC/CMS
DEC/MMS
DECnet
DECsystem–10
DECSYSTEM–20
DECUS
DECwriter

DIBOL
EduSystem
IAS
MASSBUS
PDP
PDT
RSTS
RSX

UNIBUS
VAX
VAXcluster
VMS
VT

This document was prepared with VAX DOCUMENT, Version 1.2.

Preface
The KA660-AA CPU Module Technical Manual documents the functional, physical and environmental characteristics of both variations of the KA660 CPU module, and includes
information on the MS650 memory expansion modules. The KA660-AA is for a multiuser
environment. The KA660-BA is for a single user environment and does not support multiuser
VMS or ULTRIX operating system licenses.
This manual is intended for a design engineer or applications programmer who is familiar
with Digital’s extended LSI-11 bus (Q22-bus) and the VAX instruction set. This manual should
be used along with the VAX Architecture Reference Manual as a programmer’s reference to the
module.
The manual is divided into twelve chapters and eleven appendices.
Chapter 1, Overview, introduces the KA660 Subsystem including the KA660-AA CPU module,
the MS650 memory module, and the H3602 console module.
Chapter 2, Installation and Configuration, describes procedures for installing and configuring the CPU, the memory, and the console modules in the Q22-bus backplanes and system
enclosures.
Chapter 3, Central Processor provides information about the SOC central processor and
basic VAX architecture. This chapter lists all the internal processor registers used in the
KA660 processor design. Some information on error handling is given also.
Chapter 4, KA660 Cache Memory describes the organization of the KA660 cache. It shows
the format of cache entries provides information on cache address translation.
Chapter 5, KA660 Main Memory provides information about the organization of the main
memory including the registers associated with main memory error checking and status and
includes information on cycle access times.
Chapter 6, KA660 Console Serial Line describes the serial line interface and its associated
registers on the KA660.
Chapter 7, KA660 Clock and Timer Registers provides information about the VAX standard
TOY clock and timers.
Chapter 8, KA660 Boot and Diagnostic Facility describes the KA660 initialization process
and provides information about Boot and Diagnostic Registers and the EPROM memory
resident firmware.
Chapter 9, KA660 Q22-bus Interface describes the Q22-bus interface which is implemented
with the CQBIC. Information on Q22-bus address translation is provided along with descriptions of all Q22-bus interface registers. Some information on CQBIC error handling is
provided.

xxi

Preface
Chapter 10, KA660 Network Interface provides information on the SGEC chip and the
logic that supports the Ethernet network interface. This chapter provides an over of Ethernet
principles with descriptions of packet format and programming instructions. The transmission
and reception processes are described also.
Chapter 11, KA660 Mass Storage Interface describes how the Single Host Adapter Chip
(SHAC) provides a DSSI mass storage interface for the KA660. An overview is provided as
well as descriptions of all the registers associated with the DSSI interface.
Chapter 12, KA660 Firmware describes the functional firmware that is located in the
EPROMS. The services provided by the firmware are described.
Appendix A, Q22-bus Specification, describes the low-end member of Digital’s bus family. All
of Digital’s microcomputers, such as the MicroVAX I, MicroVAX II, MicroVAX 3500, MicroVAX
3600, MicroPDP–11, use the Q22-bus.
Appendix B, Specifications, describe the physical, electrical, and environmental characteristics of the KA660-AA CPU module.
Appendix C, Address Assignments, provides a map of VAX memory space.
Appendix D, VAX Instruction Set, is a list of the VAX instructions.
Appendix E, Machine State on Power-up, describes the state of the KA660 after a power-up
halt.
Appendix F, Maintenance Operation Protocol Support, describes the Maintenance
Operation Protocol (MOP)support features in the KA660 firmware.
Appendix G, ROM Partitioning, describes the ROM partitioning and subroutine entry points
that are public and guaranteed to be compatible over future versions of the KA660 firmware.
Appendix H, BBURAM Partitioning, describes how the KA660 firmware partitions the 1KB
of battery backed up RAM.
Appendix I, Data Structures, describes the global data structures that are used by the
KA660 firmware.
Appendix J, Error Messages list the firmware error messages detected by the KA660.
Appendix K , Glossary of Terms and Acronyms, provides a list of the acronyms and terms
used in this manual.
Conventions
The following table lists the conventions used in this manual.

xxii

Preface

Table 1: Conventions
Convention

Meaning

<x:y>

Represents a bit field, a set of lines, or signals, ranging from x through y. For
example, R0 <7:4> Indicates bits 7 through 4 in a general purpose register R0.

[x:y]

Represents a range of bits, from y through x.

Return

A label enclosed in a box represents a key (usually a control or a special character
key) on the keyboard (in this case, the carriage return key.)

Note

Contains general information.

Caution

Contains information to prevent damage to equipment.

Boldface small n indicates a variable.

{}

Represents a console command element

[]

Represents a console command element that is optional

...

Represents a list command elements

Related Documents
The following documents are related to the KA660 CPU.
•
•
•
•

Microcomputer Interfaces Handbook (EB-20175-20)
Microcomputers and Memories Handbook (EB-18451-20)
VAX Architecture Handbook (EB-19580-20)
VAX–11 Architecture Reference Manual (EK-VAXAR-RM)

You can order these documents from Digital Equipment Corporation, at the address given
below.
Digital Equipment Corporation
Accessories and Supplies Group
P.O Box CS2008
Nashua, NH 03061
Attention: Documentation Products

xxiii

Contents
PREFACE

xxi

CHAPTER 1 OVERVIEW

1–1

1.1

INTRODUCTION TO THE KA660 SUBSYSTEM

1–1

1.2

KA660 PROCESSOR MODULE

1–2

1.3

MS650 MEMORY MODULE

1–8

1.4

H3602 CONSOLE MODULE

1–10

CHAPTER 2 INSTALLATION AND CONFIGURATION

2–1

2.1

INTRODUCTION

2–1

2.2

INSTALLING THE KA660 AND MS650 MEMORY MODULES

2–1

2.3

MODULE CONFIGURATION AND NAMING

2–3

2.4

MASS STORAGE CONFIGURATION
2.4.1
Changing the Node Name
2.4.2
Changing the DSSI Unit Number
2.4.3
Access to RF-series Firmware in VMS Through DUP
2.4.3.1
Allocation Class • 2–7

2–4
2–4
2–5
2–6

2.5

DSSI CABLING, DEVICE IDENTITY, AND BUS TERMINATION

2–7

2.6

KA660 CONNECTORS

2–7

CHAPTER 3 CENTRAL PROCESSOR

3–1

3.1

PROCESSOR STATE

3–1

3.2

GENERAL PURPOSE REGISTERS (GPRS)

3–2

3.3

PROCESSOR STATUS LONGWORD (PSL)

3–3

3.4

INTERNAL PROCESSOR REGISTERS (IPRS)

3–5

3.5

PROCESS STRUCTURE

3–8

3.6

DATA TYPES

3–8

3.7

INSTRUCTION SET

3–8

3.8

MEMORY MANAGEMENT
3.8.1
Translation Buffer
3.8.2
Memory Management Control Registers

3–9
3–10
3–10

3.9

INTERRUPTS AND EXCEPTIONS
3.9.1
Interrupts
3.9.2
Exceptions

3–11
3–11
3–14

iii

Contents
3.9.3

3.9.4
3.9.5
3.9.6

3–16

3–21
3–23
3–24

3.10

SYSTEM IDENTIFICATION
3.10.1
System Identification Register
3.10.2
System Identification Extension Register (SIE) (20040004)

3–26
3–26
3–27

3.11

CPU REFERENCES
3.11.1
Request Instruction-Stream Read References
3.11.2
Demand Data-Stream Read References
3.11.3
Write References

3–28
3–29
3–29
3–29

KA660 CACHE MEMORY

4–1

4.1

CACHEABLE REFERENCES

4–1

4.2

CACHE ORGANIZATION
4.2.1
Cache Address Translation
4.2.2
Cache Data Block Allocation
4.2.3
Cache Behavior on Writes
4.2.4
Cache Control Register (CCR, IPR 37)
4.2.5
Bank Enable/Hit Miss Register (BEHR)
4.2.6
Memory System Error Register (MSER, IPR 39)
4.2.7
Cache Error Detection

4–2
4–4
4–6
4–6
4–6
4–10
4–11
4–13

KA660 MAIN MEMORY SYSTEM

5–1
5–1
5–3
5–3

CHAPTER 4

CHAPTER 5

5.0.1
5.0.2
5.0.3

Information Saved On A Machine Check Exception
3.9.3.1
Byte Count • 3–16
3.9.3.2
Machine Check Code Parameter • 3–16
3.9.3.3
Most Recent Virtual Address Parameter • 3–19
3.9.3.4
Internal State Information 1 Parameter • 3–19
3.9.3.5
Internal State Information 2 Parameter • 3–20
3.9.3.6
PC • 3–20
3.9.3.7
PSL • 3–20
System Control Block (SCB)
Hardware Detected Errors
The Hardware Halt Procedure

KA660 Timing
Main Memory Organization
Main Memory Addressing

5.1

MAIN MEMORY BEHAVIOR ON WRITES

5–4

5.2

MAIN MEMORY ERROR STATUS REGISTER (MEMCSR16)

5–4

5.3

MAIN MEMORY CONTROL AND DIAGNOSTIC STATUS REGISTER (MEMCSR17)

5–6

5.4

MAIN MEMORY ERROR DETECTION AND CORRECTION

5–10

Contents
CHAPTER 6

KA660 CONSOLE SERIAL LINE

6–1

6.1

CONSOLE REGISTERS
6.1.1
Console Receiver Control/Status Register - (IPR 32)
6.1.2
Console Receiver Data Buffer - (IPR 33)
6.1.3
Console Transmitter Control/Status Register - (IPR 34)
6.1.4
Console Transmitter Data Buffer - (IPR 35)

6–1
6–1
6–3
6–4
6–6

6.2

BREAK RESPONSE

6–7

6.3

BAUD RATE

6–7

6.4

CONSOLE INTERRUPT SPECIFICATIONS

6–8

KA660 CLOCK AND TIMER REGISTERS

7–1

7.1

TIME-OF-YEAR CLOCK (TOY) - EPR 27

7–1

7.2

INTERVAL TIMER (ICCS) - EPR 24

7–2

7.3

PROGRAMMABLE TIMERS
7.3.1
Timer Control Registers (TCR0 and TCR1)
7.3.2
Timer Interval Registers (TIR0 and TIR1)
7.3.3
Timer Next Interval Registers (TNIR0 and TNIR1)
7.3.4
Timer Interrupt Vector Registers (TIVR0 and TIVR1)

7–2
7–3
7–4
7–5
7–5

KA660 BOOT AND DIAGNOSTIC FACILITY

8–1

8.1

BOOT AND DIAGNOSTIC REGISTER (BDR)

8–2

8.2

DIAGNOSTIC LED REGISTER (DLEDR)

8–4

8.3

EPROM MEMORY
8.3.1
EPROM Address Space
8.3.2
KA660 Resident Firmware Operation
8.3.2.1
Power-Up Modes • 8–6

8–5
8–5
8–5

8.4

BATTERY BACKED-UP RAM

8–7

8.5

KA660 INITIALIZATION
8.5.1
Power-Up Initialization
8.5.2
I/O Bus Initialization
8.5.3
Processor Initialization
8.5.3.1
Configuring the Local I/O Page • 8–8
8.5.4
SSC Base Address Register (SSCBR)
8.5.5
BDR Address Decode Match Register (BDMTR)
8.5.6
BDR Address Decode Mask Register (BDMKR)
8.5.7
SSC Configuration Register (SSCCR)

8–7
8–7
8–7
8–8
8–8
8–9
8–9
8–10

CDAL BUS TIMEOUT CONTROL REGISTER (CBTCR)

8–12

CHAPTER 7

CHAPTER 8

8.6

Contents
CHAPTER 9

KA660 Q22-BUS INTERFACE

9–1

9.1

Q22-BUS TO MAIN MEMORY ADDRESS TRANSLATION
9.1.1
Q22-bus Map Registers (QMR)’s
9.1.2
Accessing the Q22-bus Map Registers
9.1.3
The Q22-bus Map Cache

9–2
9–4
9–5
9–6

9.2

CDAL TO Q22-BUS ADDRESS TRANSLATION

9–7

9.3

INTERPROCESSOR COMMUNICATIONS FACILITY
9.3.1
Interprocessor Communication Register (IPCR)
9.3.2
Interprocessor Doorbell Interrupts

9–8
9–8
9–10

9.4

Q22-BUS INTERRUPT HANDLING

9–10

9.5

CONFIGURING THE Q22-BUS MAP
9.5.1
Q22-bus Map Base Address Register (QBMBR)

9–10
9–10

9.6

SYSTEM CONFIGURATION REGISTER (SCR)

9–11

9.7

ERROR REPORTING REGISTERS
9.7.1
DMA System Error Register (DSER)
9.7.2
Q22-bus Error Address Register (QBEAR)
9.7.3
DMA Error Address Register (DEAR)

9–13
9–13
9–16
9–17

9.8

Q22-BUS INTERFACE ERROR HANDLING

9–18

KA660 NETWORK INTERFACE

10–1

10.1

ETHERNET OVERVIEW

10–1

10.2

NI STATION ADDRESS ROM (NISA ROM)

10–3

10.3

PROGRAMMING THE SGEC
10.3.1
Command and Status Registers
10.3.2
Host access to NICSRs
10.3.2.1
Physical NICSRs • 10–5
10.3.2.2
Virtual NICSRs • 10–5
10.3.2.2.1
NICSR write • 10–5
10.3.2.2.2
NICSR read • 10–5
10.3.3
Vector Address, IPL, Sync/Asynch (NICSR0)
10.3.4
Transmit Polling Demand (NICSR1)
10.3.5
Receive Polling Demand (NICSR2)
10.3.6
Descriptor List Addresses (NICSR3, NICSR4)
10.3.7
Status Register (NICSR5)
10.3.7.1
NICSR5 status report • 10–16
10.3.8
Command and Mode Register (NICSR6)
10.3.9
System Base Register (NICSR7)
10.3.10
Reserved register (NICSR8)
10.3.11
Watchdog Timers (NICSR9)
10.3.12
Revision Number and Missed Frame Count (NICSR10)
10.3.13
Boot Message (NICSR11, 12, 13)
10.3.14
Diagnostic Registers (NICSR14, 15)
10.3.14.1 Diagnostic Breakpoint Address Register (NICSR14) • 10–27
10.3.14.2 Monitor Command Register (NICSR15) • 10–28

10–3
10–4
10–4

CHAPTER 10

10–6
10–7
10–8
10–9
10–11
10–17
10–21
10–22
10–22
10–24
10–25
10–27

Contents
10.3.15
10.3.16

10.3.17

10.3.18

10.3.19

CHAPTER 11

Descriptors and buffers format
Receive Descriptors
10.3.16.1 RDES0 word • 10–30
10.3.16.2 RDES1 word • 10–32
10.3.16.3 RDES2 word • 10–33
10.3.16.4 RDES3 word • 10–34
10.3.16.5 Receive descriptor status validity • 10–34
Transmit descriptors
10.3.17.1 TDES0 word • 10–36
10.3.17.2 TDES1 word • 10–37
10.3.17.3 TDES2 word • 10–39
10.3.17.4 TDES3 word • 10–40
10.3.17.5 Transmit descriptor status validity • 10–40
Setup frame
10.3.18.1 First setup frame • 10–41
10.3.18.2 Subsequent setup frame • 10–41
10.3.18.3 Setup frame descriptor • 10–41
10.3.18.4 Perfect Filtering setup frame buffer • 10–43
10.3.18.5 Imperfect Filtering setup frame buffer • 10–45
SGEC operation
10.3.19.1 Hardware and Software Reset • 10–49
10.3.19.2 Interrupts • 10–50
10.3.19.3 Startup procedure • 10–51
10.3.19.4 Reception process • 10–52
10.3.19.5 Transmission process • 10–53
10.3.19.6 Loopback operations • 10–55
10.3.19.7 DNA CSMA/CD counters and events support • 10–56

10–29
10–30

10–35

10–40

10–49

KA660 MASS STORAGE INTERFACE

11–1

11.1

SHAC INTRODUCTION

11–1

11.2

CI-DSSI OVERVIEW

11–4

11.3

SHAC REGISTERS
11.3.1
CI Port Registers
11.3.1.1
Port Queue Block Base Register (PQBBR) • 11–6
11.3.1.2
Port Status Register (PSR) • 11–9
11.3.1.3
Port Error Status Register (PESR) • 11–12
11.3.1.4
Port Failing Address Register (PFAR) • 11–13
11.3.1.5
Port Parameter Register (PPR) • 11–14
11.3.1.6
Port Control Registers • 11–15
11.3.1.6.1
Port Command Queue 0 Control Register
(PCQ0CR) • 11–15
11.3.1.6.2
Port Command Queue 1 Control Register
(PCQ1CR) • 11–15
11.3.1.6.3
Port Command Queue 2 Control Register
(PCQ2CR) • 11–16
11.3.1.6.4
Port Command Queue 3 Control Register
(PCQ3CR) • 11–16
11.3.1.6.5
Port Datagram Free Queue Control Register
(PDFQCR) • 11–16
11.3.1.6.6
Port Message Free Queue Control Register
(PMFQCR) • 11–16

11–6
11–6

vii

Contents
11.3.1.6.7
11.3.1.6.8
11.3.1.6.9
11.3.1.6.10
11.3.1.6.11

11.3.2

Port Status Release Control Register (PSRCR) • 11–16
Port Enable Control Register (PECR) • 11–16
Port Disable Control Register (PDCR) • 11–17
Port Initialize Control Register (PICR) • 11–17
Port Maintenance Timer Control Register
(PMTCR) • 11–17
11.3.1.6.12
Port Maintenance Timer Expiration Control Register
(PMTECR) • 11–17
11.3.1.6.13 Port Maintenance Control and Status Register
(PMCSR) • 11–17
SHAC Specific Registers
11.3.2.1
SHAC Software Chip Reset Register (SSWCR) • 11–19
11.3.2.2
SHAC Shared Host Memory Address (SSHMA) • 11–20

CHAPTER 12 KA660 FIRMWARE

12–1

12.1

GENERAL DESCRIPTION

12–2

12.2

HALT CODE
12.2.1
Halt Entry - Saving Processor State
12.2.2
Halt Dispatch
12.2.2.1
External Halts • 12–6
12.2.3
Halt Exit - Restoring Processor state

12–4
12–4
12–5

12.3

12.4

viii

11–19

POWER-UP
12.3.1
Identifying the Console Device
12.3.1.1
Mode switch set to "Test" • 12–9
12.3.1.2
Mode switch set to "Query" • 12–9
12.3.1.3
Mode switch set to "Normal" • 12–10
12.3.2
LED codes
OPERATING SYSTEM BOOTSTRAP
12.4.1
Preparing for the Bootstrap
12.4.1.1
Boot Devices • 12–15
12.4.1.2
Boot Flags • 12–17
12.4.2
Primary Bootstrap, VMB
12.4.3
Device Dependent Bootstrap Procedures
12.4.3.1
Disk and Tape Bootstrap Procedure • 12–22
12.4.3.2
PROM Bootstrap Procedure • 12–23
12.4.3.3
Network Bootstrap Procedure • 12–24

12–7
12–8
12–8

12–11
12–13
12–13

12–18
12–22

12.5

OPERATING SYSTEM RESTART
12.5.1
Locating the RPB

12–26
12–26

12.6

CONSOLE SERVICE
12.6.1
Console Control Characters
12.6.2
Console Command Syntax
12.6.3
Console Command Keywords
12.6.4
Console Command Qualifiers
12.6.5
Console Numeric Expression Radix Specifiers
12.6.6
Command Address Specifiers
12.6.7
References to Processor Registers and Memory

12–28
12–28
12–30
12–30
12–32
12–32
12–32
12–35

Contents
12.7

CONSOLE COMMANDS
BOOT
CONFIGURE
CONTINUE
DEPOSIT
EXAMINE
FIND
HALT
HELP
INITIALIZE
MOVE
NEXT
REPEAT
SEARCH
SET
SHOW
START
TEST
UNJAM
X - BINARY LOAD AND UNLOAD
! - COMMENT

12–36
12–37
12–40
12–42
12–43
12–46
12–49
12–50
12–51
12–53
12–55
12–58
12–60
12–61
12–64
12–68
12–73
12–74
12–78
12–79
12–81

12.8

DIAGNOSTICS
12.8.1
Error reporting
12.8.2
Diagnostic Interdependancies
12.8.3
Areas not covered
12.8.4
Console Scripts

12–86
12–87
12–88
12–88
12–88

APPENDIX B ADDRESS ASSIGNMENTS

B–1

B.1

KA660 PHYSICAL ADDRESS SPACE

B–1

B.2

KA660 DETAILED PHYSICAL ADDRESS MAP

B–2

B.3

EXTERNAL, INTERNAL PROCESSOR REGISTERS

B–8

B.4

GLOBAL Q22-BUS PHYSICAL ADDRESS SPACE

B–9

APPENDIX C VAX INSTRUCTION SET

C–1

APPENDIX D MACHINE STATE ON POWER-UP

D–1

D.1

D.2

MAIN MEMORY LAYOUT AND STATE
D.1.1
Reserved Main Memory
D.1.1.1
PFN Bitmap • D–2
D.1.1.2
Scatter/Gather map • D–3
D.1.1.3
Firmware "Scratch Memory" • D–3
D.1.2
Contents of main memory

D–1
D–2

D–3

MEMORY CONTROLLER REGISTERS
D.2.1
On-chip Cache
D.2.2
Translation Buffer
D.2.3
Halt Protected Space

D–3
D–4
D–4
D–4

Contents
APPENDIX E MAINTENANCE OPERATIONS PROTOCOL (MOP) SUPPORT

E–1

E.1

NETWORK "LISTENING"

E–1

E.2

MOP COUNTERS

E–6

APPENDIX F ROM PARTITIONING
F.1

FIRMWARE EPROM LAYOUT
F.1.1
Call-back Entry Points
F.1.1.1
CP$GETCHAR_R4 • F–3
F.1.1.2
CP$MSG_OUT_NOLF_R4 • F–4
F.1.1.3
CP$READ_WTH_PRMPT_R4 • F–4
F.1.2
Boot Information Pointers

APPENDIX G BATTERY BACKED-UP RAM PARTITIONING
G.1

SSC RAM LAYOUT
G.1.1
Public Data structures
G.1.2
Console Program MailBox (CPMBX)
G.1.3
Firmware Stack
G.1.4
Diagnostic State
G.1.5
USER Area

APPENDIX H DATA STRUCTURES
H.0.1
H.0.2
H.0.3

APPENDIX I

Halt Dispatch State Machine
Restart Paramter Block
VMB Argument List

F–1
F–1
F–3

F–5
G–1
G–1
G–2
G–2
G–3
G–3
G–3
H–1
H–1
H–5
H–8

ERROR MESSAGES

I–1

I.1

MACHINE CHECK REGISTER DUMP

I–1

I.2

HALT CODE MESSAGES

I–1

I.3

VMB ERROR MESSAGES

I–3

I.4

CONSOLE ERROR MESSAGES

I–4

APPENDIX J GLOSSARY

J–1

INDEX
EXAMPLES

2–1

Changing a DSSI Node Name

2–5

2–2

Changing a DSSI Unit Number

2–6

10–1

Perfect filtering buffer

10–45

10–2

Imperfect filtering buffer

10–47

10–3

Imperfect filtering Setup frame buffer creation C program

10–48

Contents

FIGURES
1–1

KA660 Module in a System

1–2

The KA660 Processor Module

1–3

KA660 CPU Module Component Side

1–4

KA660 Processor Module Major Functional Blocks

1–5

1–4

MS650 Memory Module (16 MB)

1–9

1–4

MS650 Memory Module (8 MB)

1–10

1–7

H3602 Console Module

1–11

2–1

Backplane Slots

2–2

Processor and Memory Module Connection

2–3

3–1

GPR Format

3–2

PSL Format

3–3

Internal Processor Register (IPR) Format

3–5

3–4

Interrupt Priority Level Register (IPLR) - (IPR 1810 1216 )

3–13

3–5

Software Interrupt Request Register (SIRR) - (IPL 2010 1416 )

3–14

3–6

Software Interrupt Summary Register (SISR) - (IPL 2110 1516 )

3–14

3–7

The Processor Stack After a Machine Check Exception

3–16

3–8

System Control Block Base Register (SCBB)

3–21

3–9

Console Saved PC (SAVPC) - (IPR 4210 2A16 )

3–24

3–10

Console Saved PSL (SAVPSL) - (IPR 4310 2B16 )

3–24

3–11

System Identification Register (SID) - (IPR 6210 3E16 )

3–27

3–12

System Identification Extension Register (SIE)

3–28

4–1

Logical Organization of Cache

4–2

Cache Entry

4–2

4–3

Cache Tag Entry

4–3

4–4

Cache Data Entry

4–3

4–5

Cache Address Translation

4–5

4–6

Cache Control Register

4–7

Tag Diagnostic Write Data Format

4–9

4–8

Tag Diagnostic Read Data Format

4–9

Bank Enable/Hit Miss Register

4–10

Memory System Error Register (MSER, IPR 39)

4–12

5–1

Main Memory Control and Diagnostic Status Register (MEMCSR17)

5–7

6–1

Console Receiver Control/Status Register - (IPR 3210 2016 )

6–2

Console Receiver Data Buffer - (IPR 3310 2116 )

6–3

Console Transmitter Control/Status Register - (IPR 3410 2216 )

6–5

6–4

Console Transmitter Data Buffer - (IPR 3510 2316 )

6–7

7–1

Time-of-Year Clock (TOY) - (EPR 2710 1B16 )

7–1

7–2

Interval Timer (ICCS) - (EPR 2410 1816 )

7–2

7–3

Timer Control Registers (TCR0 and TCR1)

7–3

Contents
7–4

xii

Timer Interval Registers (TIR0 and TIR1)

7–5

Timer Next Interval Registers (TNIR0 and TNIR1)

7–5

7–6

Timer Interrupt Vector Registers (TIVR0 and TIVR1)

7–6

8–1

Boot and Diagnostic Register

8–2

Diagnostic LED Register (DLEDR)

8–4

8–3

SSC Base Address Register (SSCBR)

8–9

8–4

BDR Address Decode Match Register (BDMTR)

8–9

8–5

BDR Address Decode Mask Register (BDMKR)

8–10

8–6

SSC Configuration Register (SSCCR)

8–10

8–7

CP Bus Timeout Control Register (CBTCR)

8–13

9–1

Q22-bus Address Translation

9–3

9–2

Q22-bus Map Register Format

9–5

9–3

Q22-bus Map Cache Entry Format

9–7

9–4

Interprocessor Communication Register (IPCR)

9–8

9–5

Q22-bus Map Base Address Register (QBMBR)

9–11

9–6

System Configuration Register (SCR)

9–12

9–7

DMA System Error Register (DSER)

9–14

9–8

Q22-bus Error Address Register (QBEAR)

9–17

9–9

DMA Error Address Register (DBEAR)

9–17

10–1

Ethernet Packet Format

10–2

Vector Address, IPL, Sync/Asynch (NICSR0)

10–6

10–3

Polling Demand (NICSR1)

10–8

10–4

NICSR2 format

10–9

10–5

Descriptor list addresses format

10–10

10–6

NICSR5 bits

10–12

10–7

NICSR6 format

10–17

10–8

NICSR7 format

10–21

10–9

NICSR9 format

10–23

10–10

Revision Number and Missed Frame Count (VIRTUAL NICSR10)

10–25

10–11

Boot Message

10–26

10–12

NICSR14 format

10–27

10–13

NICSR15 format

10–28

10–14

Receive descriptor format

10–30

10–15

Transmit descriptor format

10–36

10–16

Setup frame descriptor format

10–42

10–17

Perfect Filtering setup frame buffer format

10–44

10–18

Imperfect Filtering setup frame format

10–46

11–1

Relationship of the DSSI to SCA and CI

11–3

11–2

Port Queue Block Base Register (PQBBR)

11–7

11–3

Port Queue Block Base Register (PQBBR) After RESET

11–7

11–4

Port Status Register (PSR)

11–9

11–5

Port Error Status Register (PESR)

11–12

Contents
11–6

Port Failing Address Register (PFAR)

11–13

11–7

Port Parameter Register (PPR)

11–14

11–8

Port Control Registers

11–15

11–9

Port Maintenance Control And Status Register (PMCSR)

11–18

11–10

SHAC Software Chip Reset (SSWCR)

11–20

11–11

SHAC Shared Host Memory Address (SSHMA)

11–20

12–1

KA660 Firmware Structural Components

12–3

12–2

Console Banner

12–8

12–3

Language Selection Menu

12–10

12–4

Normal Diagnostic Countdown

12–10

12–5

Abnormal Diagnostic Countdown

12–11

12–6

Console Prompt

12–11

12–7

Console Boot Display with No Default Boot Device

12–11

12–8

Memory Layout prior to VMB Entry

12–15

12–9

VMB Boot Flags (/R5:)

12–17

12–10

Successful Automatic Bootstrap

12–20

12–11

Memory Layout at VMB Exit

12–21

12–12

Boot Block Format

12–23

12–13

Locating the Restart Parameter Block

12–27

12–14

Diagnostic Register Dump

12–87

D–1

Memory Layout after Power-up Diagnostics

D–2

F–1

KA660 EPROM Layout

F–2

Boot Information Pointers

F–5

G–1

KA660 SSC BBU RAM Layout

G–1

G–2

NVR0 (20140400) : Console Program MailBoX (CPMBX)

G–2

G–3

NVR1 (20140401)

G–3

G–4

NVR2 (20140402)

G–3

TABLES
1

Conventions

xxiii

3–1

Special GPRs

3–2

Processor Status Longword Format

3–3

KA660 Internal Processor Registers

3–6

3–4

Interrupt Priority Levels

3–12

3–5

Exception Classes

3–15

3–6

Floating Point Error Machine Checks

3–17

3–7

Memory Management Error Machine Checks

3–17

3–8

Interrupt Error Machine Checks

3–18

3–9

Microcode Error Machine Checks

3–18

3–10

Read Error Machine Checks

3–18

3–11

Write Error Machine Checks

3–19

3–12

Internal State Information 1 Field Description

3–20

xiii

Contents
3–13

xiv

Internal State Information 2 Field Description

3–20

3–14

The System Control Block Format

3–21

3–15

CPU State After a HALT

3–25

3–16

HALT Codes

3–25

3–17

3–27

3–18

3–28

4–1

4–7

4–2

Cache Diagnostic Mode Addresses

4–8

4–3

Bank Enable/Hit Miss Register (BEHR)

4–10

4–4

Memory System Error Register (MSER, IPR 39)

4–12

5–1

KA660 Reference Timing

5–1
5–11

5–2

Uncorrectable Error Actions

6–1

Console Registers

6–1

6–2

Console Receiver Control/Status Register

6–2

6–3

Console Receiver Data Buffer

6–3

6–4

Console Transmitter Control/Status Buffer

6–6

6–5

Console Transmitter Data Buffer

6–7

6–6

Baud Rate Selection

6–8

7–1

Interval Timer Bit Descriptions

7–2

Timer Control Register Bit Descriptions

7–3

8–3

Diagnostic LED Register Bit Descriptions

8–4

Power-Up Modes

8–6

8–5

SSC Configuration Register Bit Descriptions

8–11

8–6

CP Bus Timeout Control Register Bit Descriptions

8–13

9–1

Q22-bus Map Register Addresses

9–4

9–2

Q22-bus Map Register Bit Description

9–5

9–3

Q22-bus Map Cache Entry Bit Description

9–7

9–4

Interprocessor Communication Register Bit Description

9–9

9–5

System Configuration Register Bit Description

9–12

9–6

DMA System Error Register Bit Description

9–15

10–1

Bit access modes

10–4

10–2

NICSR0 bits

10–7

10–3

NICSR0 access

10–7

10–4

NICSR1 bits

10–8

10–5

NICSR1 access

10–8

10–6

NICSR2 bits

10–9

10–7

NICSR2 access

10–9

10–8

Descriptor lists addresses bits

10–10

10–9

NICSR3 access

10–11

10–10

NICSR4 access

10–11

NICSR5 bits

10–12

10–13

NICSR5 access

10–16

Contents
10–14

NICSR6 bits

10–18

10–15

NICSR6 access

10–21

10–16

NICSR7 bits

10–22

10–17

NICSR7 access

10–22

10–18

NICSR9 bits

10–23

10–19

NICSR9 access

10–24

10–20

NICSR10 bits

10–25

10–21

NICSR10 access

10–25

10–22

NICSR11,12,13 bits

10–26

10–23

NICSR11,12,13 access

10–26

10–24

NICSR14 bits

10–27

10–25

NICSR14 access

10–27

10–26

NICSR15 bits

10–28

10–27

NICSR15 access

10–29

10–28

RDES0 bits

10–31

10–29

RDES1 bits

10–33

10–30

RDES2 bits

10–33

10–31

RDES3 bits

10–34

10–32

Receive descriptor status validity

10–35

10–33

TDES0 bits

10–36

10–34

TDES1 bits

10–38

10–35

TDES2 bits

10–39

10–36

TDES3 bits

10–40

10–37

Transmit descriptor status validity

10–40

10–38

Setup frame descriptor bits

10–39

10–42
10–50

10–40

Reception process state transitions

10–53

10–41

Transmission process state transitions

10–54

10–42

CSMA/CD counters

10–56

11–1

Port Queue Block Base Address Register (PQBBR)

11–7

11–2

Port Queue BLock Base Address Register Bits After RESET

11–8

11–3

Port Status Register Bit Descriptions

11–10

11–4

Port Error Status Register Bit Definitions

11–12

11–5

Port Parameter Register Bit Descriptions (PPR)

11–14

11–6

Port Maintenance Control and Status Register (PMCSR) Bits

11–18

12–1

Halt Action Summary

12–6

12–2

LED Codes

12–12

12–3

KA660 Supported Boot Devices

12–16

12–4

12–17

12–5

12–28

12–6

Command, Parameter, and Qualifier Keywords

12–30

12–7

Console Radix Specifiers

12–32

Contents
12–8

Console Symbolic Addresses

12–9

12–36

12–10

12–36

12–11

Console Command Summary

12–82

12–12

Console Qualifier Summary

12–84

12–13

Diagnostic Scripts

12–89

B–1

General Local Address Space Map

B–1

B–2

Detailed Local Address Space Map

B–2

B–3

External, Internal Processor Registers

B–9

B–4

Global Q22-bus Physical Address Map

B–9

E–1

KA660 Network Maintenance Operations Summary

E–2

Supported MOP Messages

E–3

Ethernet & IEEE 802.3 Packet Headers

E–5

E–4

MOP Multicast Addresses and Protocol Specifiers

E–5

MOP Counter Block

E–6

F–1

F–3

G–1

G–2

G–3

G–3
H–1

G–3
Firmware State Transition Table

H–2

Restart Parameter Block fields

H–5

H–3

VMB Argument List

H–8

I–1

HALT Messages

I–2

VMB Error Messages

I–3

Console Error Messages

J–1

xvi

12–33

I–4
J–1

Chapter 1
OVERVIEW

This chapter provides a brief description of the KA660 CPU/Memory Subsystem.

1.1 Introduction to The KA660 Subsystem
The KA660 processor module combines with the MS650 Memory module and the H3602
Console module to form the CPU/Memory subsystem for the VAX 4000-200 product. The
subsystem is available in two enclosures, the BA430 or the BA215. It uses the DSSI bus to
communicate with mass storage devices and the Q22-bus to communicate with I/O devices. A
single KA660 CPU module can support up to four MS650 memory modules.
Figure 1–1 is a block diagram of the CPU/Memory subsystem’s major functions.

OVERVIEW 1–1

Figure 1–1: KA660 Module in a System

Backplane Interconnect

Ethernet

H3602
Console
Module

Ribbon
Cable

DSSI

KA660
Processor
Module

Q22−bus

Serial Line
Memory
Interconnect

MS650
Memory
Module/s

The KA660 and the MS650 designs are implemented in standard quad-height sized modules.
Both modules mount in standard Q22-bus backplane slots which implement the Q22-bus in
the AB rows and the CD interconnect in the CD rows.
The KA660 Processor Module communicates with the memory modules across a memory
interconnect routed through a 50-pin ribbon cable and the CD interconnect on the backplane.
The DSSI connects through a 50-Pin ribbon cable located on top of the memory interconnect
cable. The backplane connector also connects the subsystem with the Q22-bus. There are no
jumpers or switches to configure on the processor module. The KA660 connects to the H3602
Console Module and the Ethernet Controller with a 40-pin ribbon cable. The Console Module
contains configuration switches, Ethernet and DSSI connectors, fuses, and an LED display.

1.2 KA660 Processor Module
The KA660 processor can be configured only as an arbiter on the Q22-bus. An arbiter is the
single entity responsible for controlling the Q22-bus. It must reside in the first backplane slot
where it arbitrates bus mastership and fields bus interrupt requests and any on-board interrupt requests. This processor module is designed for use in high-speed, real-time applications
and for multiuser, multitasking environments. There are two variants: the KA660-AA, runs
multiuser software; and the KA660-BA, runs single-user software.

1–2 OVERVIEW

Figure 1–2 is a photograph of the KA660 Processor Module.
Figure 1–2: The KA660 Processor Module

OVERVIEW 1–3

The major hardware components of the KA660 CPU module are listed below. The chip identification numbers are shown in Figure 1–3
System on a Chip (SOC) CPU

DC222

A Main Memory Controller (CMCTL)

DC557

Q22-bus Interface (CQBIC)

DC527

System Support Chip (SSC)

DC511

Second Generation Ethernet Controller (SGEC)

DC541

Single Host Adapter Chip to interface DSSI (SHAC)

DC542

Two Firmware EPROMs
A Boot and diagnostic facility
Console Connection
VAX compatible console port
Backplane Connection

Figure 1–3 shows the positions of the major chips on the KA660.
Figure 1–3: KA660 CPU Module Component Side

DSSI / MEMORY

CONSOLE
FIRMWARE
PROM
E19

FIRMWARE
PROM
E6

DC541
SGEC
DC542
SHAC

E18

DC511
SSC

E35

1–4 OVERVIEW

DC557
CMCTL

DC222
SOC

DC527
CQBIC

E32

E23

E11

The KA660 Processor Module is divided into several major functional subsystems as listed below
and shown in Figure 1–4.
The Central Processing Subsystem

The DSSI Subsystem

The Memory Control Subsystem

The System Support Subsystem

The Q22-bus Subsystem

The Ethernet Subsystem

Figure 1–4: KA660 Processor Module Major Functional Blocks

System Support
Subsystem

BCDAL<31:24> (NI−ROM DATA)
Transceiver
Subsystem

A<06:02> (NI−ROM ADDRESS)

CPU Subsystem

Address
Latch

Ethernet
Subsystem

CDAL <31:00>

Memory
Memory
Control
Subsystem

Q22-bus
Subsystem

DSSI
Subsystem

The rest of Section 1.1 describes the subsystems.
•

The Central Processing Subsystem

OVERVIEW 1–5

The Central Processing Subsystem features the System On a Chip (SOC) CPU and its
accompanying support logic. The SOC Chip is a unique design that contains several
system components on a single substrate contained in a 132-pin surface mount, CERQUAD
chip package. The SOC contains the Central Processing Unit (CPU), The Floating Point
Accelerator Unit (FPA), and 8 KB of cache to optimize system performance.
The central processor in the SOC supports the following MicroVAX instruction set with the
following string instructions:
CMPC3 (Compare Character 3 operand)
SKPC (Skip Character)

CMPC5 (Compare Character 5 operand)
SPANC (Span Character)

LOCC (Locate
Character)
SCANC (Scan
Character)

The following subset of the VAX data types are provided:
Byte

Word

Longword

Quadword

Character string

Variable-length bit field

Absolute queues

Self-relative queues

F-floating

G-floating

D-floating

Support for the remaining VAX data types can be provided through macrocode emulation.
The processor also supports full VAX memory management with demand paging and a
4GB virtual address space.
The Floating Point Accelerator Unit (FPU) in the SOC executes the VAX f_, d_, and g_
floating point instructions. It executes 61 floating point instructions and 2 longword-length
integer multiply instructions in the VAX base instruction group. It supports the MicroVAX
chip subset of the VAX floating point instruction set and data types.

•

The Memory Control Subsystem
The Memory Control Subsystem contains the Memory Controller Chip (CMCTL) and its
associated termination logic. This subsystem provides an interface between the Data and
Address Lines (CDAL) lines from the KA660 CDAL bus and the Data and Address lines on
the MS650 MDATA bus.
The CMCTL chip contains approximately 25,000 transistors in a 132-pin CERQUAD
surface mount package. It supports up to 64 Kbytes of ECC memory, with a 450 ns cycle
time for longword transfers and a 720 ns cycle time for quadword transfers,
This memory resides on one to four MS650 memory modules, depending on the system configuration. The MS650 communicates with the KA660 through the memory interconnect
which uses the CD interconnect and a 50-pin ribbon cable.

•

The Q22-bus Subsystem
The Q22-bus subsystem contains the Q22-bus interface and asociated termination logic.
This subsystem provides an interface between the Q22-bus and the Central Processor’s
CDAL bus. The interface is implemented with the CQBIC. The CQBIC contains approximatley 40,870 transistors in a 132-pin CERQUAD surface mount package. The CQBIC is
a 32-bit to 16 bit adapter which provides physical memory address translation for direct
memory access (DMA) devices on the Q22-bus. It supports up to 16-word, block mode

1–6 OVERVIEW

transfers between a Q22-bus DMA device and main memory, and up to 2-word, block mode
transfers between the CPU and Q22-bus devices. The Q22-bus interface contains the
following:
• A 16-entry map cache for the 8192-entry, main memory-resident scatter-gather-map,
used for translating 22-bit Q22-bus addresses into 26-bit main memory addresses.
• Interrupt arbitration logic that recognizes Q22-bus interrupt requests BR7-BR4
• Q22-bus termination (240 )
•

The DSSI Subsystem
This subsystem provides an interface between the DSSI bus and the KA660 CDAL bus. It
contains the Single Host Adapter Chip (SHAC), the DSSI jumpers, 16MHZ Oscillator, and
associated termination and control logic. The SHAC is in a 164 pin CERQUAD package.
It facilitates scatter and gather mapping along with internal FIFO buffering.
The DSSI interface allows the DSSI bus on the KA660 to transmit packets of data to, and
receive packets from, up to seven other DSSI devices. These devices include the RF-series
Integrated Storage Elements (ISEs), a KFQSA module, a second KA660 module, or a
KA640 module.
The DSSI bus improves system performance because it has a higher transfer rate than
the Q22-bus and it relieves the Q22-bus of disk traffic. The DSSI bus has eight data lines,
one parity line, and eight control lines. Controllers are built into the ISEs, enabling many
functions to be handled without host or adapter intervention.

•

The System Support Subsystem
The system support subsystem handles the basic functions required to support the console
in a system environment. This subsystem contains the System Support Chip (SSC), the
Firmware ROMs, the Boot and Diagnostic Register, and the Station Address ROM.
The SSC chip is implemented in an 84 pin CERQUAD surface mount package. It provides
console and boot code support functions, operating system support functions, timers, and
the following features:
Word-wide ROM unpacking

1kB battery backed-up RAM

Halt-arbitration logic

Console serial line

Interval timer with 10ms
interrupts

VAX-standard time-of-year clock with battery backup

IORESET register

Programmable CDAL bus timeout

Two programmable timers

Resident firmware Read Only Memory is located on two chips, each 128 KByte by 8-bit
EPROMS. The firmware gains control when the CPU halts. This code contains programs
that provide the following services:
• Board initialization
• Power-up self-testing of the KA660 and MS650 modules

OVERVIEW 1–7

•

•
•

•

Emulation of a subset of the VAX standard console (auto or manual bootstrap, auto or
manual restart, and a simple command language for examining or altering the state of
the processor)
Booting from supported Q22-bus devices
Multilingual translation of key system messages

The Boot and Diagnostic Register (BDR) allows the firmware and the operating system to
read KA660 configuration bits. The station address ROM contains the network address of
the system. It is implemented in a 32-Byte by 8-bit ROM (6331).
The Ethernet Subsystem
The Ethernet subsystem handles communications between the CPU module and other
nodes on the Ethernet. It is implemented with the Second Generation Ethernet Controller
Chip (SGEC, DC541) on-board network interface. Used in connection with the H3602
console module, the SGEC allows the KA660 to connect to either a thinwire or standard
Ethernet. It supports the Ethernet Data Link Layer and the CP Bus Parity Protection.
The SGEC chip is in a 84 pin package. The chip facilitates scatter and gather mapping
along with dual internal FIFO buffering.

1.3 MS650 Memory Module
The MS650 memory module for the KA660 CPU is available in two variations. The MS650-BA
contains 16 Mbyte of memory and the MS650-BB contains 8 Mbytes. The memory is arranged
in 39-bit wide arrays implemented with 1 Mbyte, 120 ns, dynamic RAMs in surface mount
packages. Of the 39 bits, 32 bits are data and 7 bits are Error Checking and Correction
(ECC) Bits. The MS650 modules are single, quad-height, Q22-bus modules as shown in
<REFERENCE>(MS650-BA_PHOTO)and <REFERENCE>(MS650-BB_PHOTO).

1–8 OVERVIEW

Figure 1–4: MS650 Memory Module (16 MB)

OVERVIEW 1–9

Figure 1–4: MS650 Memory Module (8 MB)

1.4 H3602 Console Module
The H3602 console module (Figure 1–7) is a unique I/O panel that is used in BA213 and
BA215 enclosures, A one-piece ribbon cable on the H3602 plugs into J1 (system support
connector) on the KA660. The H3602 fits over backplane slots 1 and 2, covering both the
KA660 Processor module and the first of four possible MS650 memory modules. The H3602
allows the KA660 CPU module to interface to a serial line console device, a DSSI bus, and to
the Ethernet. Adhesive tags are included for the user to name the modules in the respective
slots.

1–10 OVERVIEW

Figure 1–7: H3602 Console Module

CPU Cover Panel

Break
Enable/
Disable
Switch

Standard
Ethernet
Connector

LED Display
Power-Up
Mode Switch
Modified
Modular Jack
Ethernet
Connector Switch
ThinWire Ethernet
Connector

MLO-005504

The outside H3602 console panel contains the following features:
•
•
•
•
•
•
•

Modified modular jack (MMU) SLU connector
Power-up mode switch
Hexadecimal LED display
Break enable switch
Standard/Thin wire Ethernet connectors
Standard/ThinWire Ethernet selector switch
Indicator LEDs

The console panel also has the following features inside:
•
•
•
•

Baud Rate rotary switch
Battery backup unit (BBU) for TOY clock
40-pin cable connector
List of baud rate switch settings

OVERVIEW 1–11

Chapter 2
Installation and Configuration

2.1 Introduction
This chapter describes how to install the KA660 in a system. It discusses the following topics.
•
•
•

Installing the KA660 and MS650 modules
Configuring the KA660
The KA660 connectors

2.2 Installing the KA660 and MS650 Memory Modules
The KA660 and MS650 (-BB or -BA only) modules must be installed in system enclosures having Q22/CD slots. These modules are not compatible with Q/Q backplane slots and therefore
should only be installed in Q22/CD backplane slots.
The KA660 CPU module and the MS650 memory modules must be installed in the five
rightmost backplane slots. The KA660 CPU module must be installed in slot 1 of the Q22/CD
backplane. MS650 memory modules must be installed in slots immediately adjacent to the
CPU module. Figure 2–1 shows the positions of the module slots in the backplane.

Installation and Configuration 2–1

Figure 2–1: Backplane Slots

MS650 no.4
MS650 no.3
MS650 no.2
MS650 no.1
KA660 CPU

A
Q22−bus
Blocks

CD
Interconnect
Blocks

B
C
D
12 11 10 9 8 7 6 5 4 3 2 1

Slot Number

Up to four MS650 memory modules can be installed, occupying slots 2,3,4 and 5 respectively.
A 50-pin ribbon cable is used to connect the KA660 processor module and the MS650 memory
module(s) , as shown in Figure 2–2.
The KA660 module is installed in backplane slot 1 and the memory modules are installed in
slots 2 through 5. Use the following procedure to install the KA660 and MS650 modules.
1. Install the KA660 CPU in slot 1 of the Q22-bus/CD backplane.
2. Install the MS650 memory module in slots 2, immediately adjacent to the KA660 CPU.
When installing additional memory use slots 3 through 5. Do not leave a gap between
memory modules.
3. Install a 50-pin ribbon cable between the KA660 CPU and the MS650 Memory Module/s.
(see Figure 2–2)

2–2 Installation and Configuration

Figure 2–2: Processor and Memory Module Connection

MS650 MS650 MS650 MS650
no.4 no.3 no.2 no.1

KA660 CPU

CPU/Memory
Interconnect
Cable (50-pin)

2.3 Module Configuration and Naming
Each module in a system must use a unique device address and interrupt vector. The device
address is also known as the control and status register (CSR) address. Most modules have
switches or jumpers for setting the CSR address and interrupt vector values. The value of a
floating address depends on what other modules are housed in the system.
Set CSR addresses and interrupt vectors for a module as follows:
1. Determine the correct values for the module with the CONFIGURE command at the
console I/O prompt (>>>). The CONFIG utility eliminates the need to boot the VMS
operating system to determine CSRs and interrupt vectors. Enter the CONFIGURE
command, then HELP for the list of supported devices:
>>> config
Enter device configuration, HELP, or EXIT
Device, Number? help
Devices:

Installation and Configuration 2–3

LPV11
RLV21
DMV11
RQC25
KXXXX-TAPE
CXB16
VSV21
IAV11A
IGQ11

KXJ11
TSV05
DELQA
KXXXX-DISK
KMV11
CXY08
IBQ01
IAV11B

DLV11J
RXV21
DEQNA
TQK50
IEQ11
VCB02
IDV11A
MIRA

DZQ11
DRV11W
RQDX3
TQK70
DHQ11
QDSS
IDV11B
ADQ32

DZV11
DRV11B
KDA50
TU81E
DHV11
DRV11J
IDV11C
DTC04

DFA01
DPV11
RRD50
RV20
CXA16
DRQ3B
IDV11D
DESQA

The LPV11–SA has two sets of CSR address and interrupt vectors. To determine the
correct values for an LPV11–SA, enter LPV11,2 at the DEVICE prompt for one LPV11–SA,
or enter LPV11,4 for two LPV11–SA modules.
2. See the KA660 CPU System Maintenance Manual for switch and CSR and interrupt vector
jumper settings for supported options.

2.4 Mass Storage Configuration
In a BA213 enclosure there is space for four mass storage devices, three integrated storage
element (ISE)s and one TK70 or else four ISEs. The ISEs, are part of the DIGITAL Storage
System Interconnect (DSSI) bus.
The DSSI bus is part of the backplane. The ISEs are of the RF-series, and they plug into the
backplane to become part of the bus. Each ISE must have its own unique DSSI node ID. The
ISE receives its node ID from a plug on the operator control panel (OCP) on the front panel.
The VMS operating system creates DSSI disk device names according to the following scheme:
nodename $ DIA unit number. For example, SUSAN$DIA3
You can use the device name for booting, as follows:
>>> BOOT SUSAN$DIA3

You can access local programs in the RF-series ISE through the MicroVAX Diagnostic Monitor
(MDM), or through the VMS operating system (version 5.0) and console I/O mode SET HOST
/DUP command. This command creates a virtual terminal connection to the storage device
and the designated local program using the Diagnostic and Utilities Protocol (DUP) standard
dialog. Section 2.4.3 describes the procedure for accessing DUP through the VMS operating
system.

2.4.1 Changing the Node Name
Each ISE has a node name that is maintained in EPROM on board the controller module.
This node name is determined in manufacturing from an algorithm based on the drive serial
number. You can change the node name of the DSSI device to something more meaningful by
following the procedure in Example 2–1. In the example, the node name for the ISE at DSSI
node address 1 is changed from R3YBNE to DATADISK.

2–4 Installation and Configuration

Example 2–1: Changing a DSSI Node Name
>>> sho dssi
DSSI Node 0 (MDC)
-DIA0 (RF71)
DSSI Node 1 (R3YBNE)
-DIA1 (RF71)

!The node name for this drive will be
!changed from R3YBNE to DATADISK.

DSSI Node 7 (*)
>>>
>>> set host/dup/dssi 1
Starting DUP server...
Copyright 1988 Digital Equipment Corporation
DRVEXR V1.0 D 5-NOV-1988 15:33:06
DRVTST V1.0 D 5-NOV-1988 15:33:06
HISTRY V1.0 D 5-NOV-1988 15:33:06
ERASE V1.0 D 5-NOV-1988 15:33:06
PARAMS V1.0 D 5-NOV-1988 15:33:06
DIRECT V1.0 D 5-NOV-1988 15:33:06
End of directory
Task Name? params
Copyright 1988 Digital Equipment Corporation
PARAMS> sho nodename
Parameter
Current
--------- ---------------NODENAME
R3YBNE

Default
---------------RF71

Type
-------String

Radix
----Ascii

PARAMS> set nodename datadisk
PARAMS> write

!This command writes the change
!to EPROM.
Changes require controller initialization, ok? [Y/(N)] y
Stopping DUP server...
>>> sho dssi
DSSI Node 0 (MDC)
-DIA0 (RF71)
DSSI Node 1 (DATADISK)
-DIA1 (RF71)

!The node name has changed from
!R3YBNE to DATADISK.

DSSI Node 7 (*)

2.4.2 Changing the DSSI Unit Number
By default, the ISE drive assigns the disk’s unit number to the same value as the DSSI node
address for that drive.
Example 2–2 shows how to change the unit number of a DSSI device. This example changes
the unit number for the RF71 drive at DSSI node address 2 from 1 to 50 (decimal). You must
change two parameters: UNITNUM and FORCEUNI. Changing these parameters overrides
the default, which assigns the unit number the same value as the node address.

Installation and Configuration 2–5

Example 2–2: Changing a DSSI Unit Number
>>> sho dssi
DSSI Node 0 (MDC)
-DIA0 (RF71)
DSSI Node 1 (R3QJNE)
-DIA1 (RF71)

!The unit number for this drive will be
!changed from 1 to 50 (DIA1 to DIA50).

DSSI Node 7 (*)
>>>
>>> set host/dup/dssi 1
Starting DUP server...
Copyright 1988 Digital Equipment Corporation
DRVEXR V1.0 D 5-NOV-1988 15:33:06
DRVTST V1.0 D 5-NOV-1988 15:33:06
HISTRY V1.0 D 5-NOV-1988 15:33:06
ERASE V1.0 D 5-NOV-1988 15:33:06
PARAMS V1.0 D 5-NOV-1988 15:33:06
DIRECT V1.0 D 5-NOV-1988 15:33:06
End of directory
Task Name? params
Copyright 1988 Digital Equipment Corporation
PARAMS> sho unitnum
Parameter
Current
--------- ---------------UNITNUM
0

Default
---------------0

Type
Radix
-------- ----Word
Dec
U

PARAMS> sho forceuni
Parameter
Current
--------- ---------------FORCEUNI
1

Default
Type
Radix
---------------- -------- ----1
Boolean
0/1
U

PARAMS> set unitnum 50
PARAMS> set forceuni 0
PARAMS> write

!This command writes the changes to EPROM.

PARAMS> ex
Exiting...
Task Name?
Stopping DUP server...
>>>
>>>sho dssi
DSSI Node 0 (MDC)
-DIA0 (RF71)
DSSI Node 1 (R3QJNE)
-DIA50 (RF71)

!The unit number has changed
!and the node ID remains at 1.

DSSI Node 7 (*)

2.4.3 Access to RF-series Firmware in VMS Through DUP
You can also access the RF-series ISE firmware utilities from the VMS operating system as
well as through the console commands.

2–6 Installation and Configuration

Access the ISE firmware through the VMS operating system to look up or to view parameter
settings, but not to change them. To change ISE parameter settings, enter the ISE firmware
through the console I/O mode SET HOST/DUP command.
Load the FYDRIVER using the following commands in SYSGEN:
$ MCR SYSGEN
SYSGEN> LOAD FYDRIVER/NOADAPTER
SYSGEN> CONNECT FYA0/NOADAPTER
SYSGEN> EXIT
$

You can then access the ISE firmware utilities using the following VMS command:
$ SET HOST/DUP/SERVER=MSCP$DUP/TASK=PARAMS nodename

2.4.3.1

Allocation Class

When a KA660 system containing ISEs is configured in a cluster, either as a boot node or a
satellite node, you must assign the allocation class in VMS SYSGEN and for the ISE matching
non-zero values. To change the allocation class of the ISE, use the following commands:
>>> SET HOST/DUP/DSSI <DSSI node number> PARAMS
Starting DUP server..
PARAMS> SET ALLCLASS <allocation class value>
PARAMS> WRITE
Changes require controller initialization, ok? [Y/N] Y
Stopping DUP server..
>>>

2.5 DSSI Cabling, Device Identity, and Bus Termination
The ISEs in one particular BA430 enclosure are connected to the system backplane, and
communicate internally over the backplane. There are no internal DSSI cables. Externally, a
50-pin ribbon cable connects the DSSI bus to other devices, either hosts or expanders.
All DSSI devices on the same bus must have unique identifiers.
The ID plug provides an identity for the DSSI bus.

2.6 KA660 Connectors
The KA660 uses two connectors, J1 and J2. J1 (system support connector) is the connector for
the 40-pin ribbon cable that goes to the console module. Users configure the KA660 through
the H3602 console module. Figure 1–3 in Chapter 1, OVERVIEW shows the location of the
connectors on the KA660 module. J2 is a dual connector. The upper half contains 50 pins for
the DSSI connection and the lower half contains 50 pins for the memory module connection.

Installation and Configuration 2–7

Chapter 3
Central Processor

This section provides summary information about the SOC/C (CPU) chip and the MicroVAX
architecture. It is not intended as a complete reference, but rather as an overview of the
user-visible features.
The central processor of the KA660 supports the MicroVAX Chip subset (plus six additional
string instructions) of the VAX instruction set and data types and full VAX memory management. It is implemented as part of the SOC/C chip.

3.1 Processor State
The processor state consists of that portion of the state of a process which is stored in processor
registers rather than in memory. The processor state is composed of sixteen General Purpose
Registers (GPR’s), the Processor Status Longword (PSL), and the Internal Processor Registers
(IPR’s).
Non-privileged software can access the GPR’s and the Processor Status Word (bits <15:00> of
the PSL). The IPR’s and bits <31:16> of the PSL can only be accessed by privileged software.
The IPR’s are explicitly accessible only by the Move To Processor Register (MTPR) and Move
From Processor Register (MFPR) instructions which can be executed only while running in
kernel mode.

Central Processor 3–1

3.2 General Purpose Registers (GPRs)
The KA660 implements 16 General Purpose Registers as implemented per the VAX
Architecture Reference Manual. These registers are used for temporary storage, accumulators, and base and index registers for addressing. These registers are denoted R0 - R15. The
bits of a register are numbered from right to left, <0> through <31>. Figure 3–1 shows the
General Purpose Register format. Table 3–1 describes the registers.

Figure 3–1: GPR Format
3
1

X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X
ESB90P0001

Table 3–1 lists certain registers that have been assigned special meaning by the VAX–11
architecture standard.
Table 3–1: Special GPRs
Register

Mnemonic

Description

R15

Program Counter

The PC contains the address of the
next instruction byte of the program.

R14

Stack Pointer

The SP contains the address of the top
of the processor defined stack.

R13

Frame Pointer

The VAX–11 procedure call convention
builds a data structure on the stack
called a stack frame. The FP contains
The address of the base of this data
structure.

R12

Argument Pointer

The VAX–11 procedure call convention
uses a data structure termed an argument list. The AP contains the address
of the base of this data structure.

Consult the VAX Architecture Reference Manual for more information on the operation and use
of these registers.

3–2 Central Processor

3.3 Processor Status Longword (PSL)
The KA660 Processor Status Longword (PSL) is implemented per The VAX Architecture
Reference Manual . The PSL is saved on the stack when an exception or interrupt occurs
and is saved in the Process Control Block (PCB) on a process context switch. Bits <15:00>
may be accessed by non-privileged software, while bits <31:16> may only be accessed by
privileged software. Processor initialization sets the PSL to 041F 000016 . Figure 3–2 shows
the Processor Status Longword format. Table 3–2 lists the bits and definitions.

Figure 3–2: PSL Format
3
1

3
0

C T
M P

2
9

2
8

MBZ

2
7

2
6

2
5

2
4

2
3

2
2

2
1

2
0

M
F
P I CUR PRV B
D S MOD MOD Z

1
6

1
5

IPL

MBZ

D F I
V U V

N Z

V C

ESB90P0002

Table 3–2: Processor Status Longword Format
PSL
Data Bit

Name

Definition

<31>

Compatibility Mode. This bit always reads as ZERO, loading a ONE into
this bit is a NOP.

<30>

Trace Pending

<29:28>

MBZ

Must Be written as Zero

<27>

FPD

First Part Done

<26>

Interrupt Stack

<25:24>

CUR

Current Mode

<23:22>

PRV

Previous Mode

<21>

MBZ

Must Be written as Zero

<20:16>

IPL

Interrupt Priority Level

<15:8>

MBZ

Must Be written as Zero

<7>

Decimal Overflow Trap Enable This read/write bit has no effect on KA660
hardware; it can be used by macrocode which emulates VAX decimal instructions.

<6>

Floating Underflow Fault Enable

Central Processor 3–3

Table 3–2 (Cont.): Processor Status Longword Format
PSL
Data Bit

Name

Definition

<5>

Integer Overflow Trap Enable

<4>

Trace Trap Enable

<3>

Negative Condition Code

<2>

Zero Condition Code

<1>

Overflow Condition Code

<0>

Carry Condition Code

NOTE
VAX Compatibility Mode instructions can be emulated by macrocode, but the
emulation software runs in native mode, so the CM bit is never set.

3–4 Central Processor

3.4 Internal Processor Registers (IPRs)
The privileged internal processor register space provides access to many types of CPU control
and status registers such as the memory management base registers, parts of the PSL, and the
multiple stack pointers. These registers are explicitly accessible only by the Move to Processor
Register (MTPR) and Move from Processor Register (MFPR) instructions which require kernel
privileges. The addresses of the KA660 internal processor registers are given in Table B–2.
Internal processor registers are longword size, as shown in Figure 3–3.
Figure 3–3: Internal Processor Register (IPR) Format

3
1

X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X
ESB90P0001

IPR Categories
Each IPR falls into one of the following categories:
(1) VAX standard IPRs implemented by KA660 in the SOC/C chip.
(2) VAX standard IPRs implemented by KA660 in the SSC chip.
(3) Unique KA660 IPRs implemented by all designs that use the SOC/C chip.
(4) Unique KA660 IPRs implemented by all designs that use the SSC chip.
(5) Not implemented, timed out by the CDAL Bus Timer (in the SSC chip) after 4µs. Read
as zero, nop on write.
(6) Access not allowed; accesses result in a reserved operand fault.
(7) Accessible, but not fully implemented, accesses yield unpredictable results.
Table 3–3 explains each IPR.

Central Processor 3–5

Table 3–3:

KA660 Internal Processor Registers

IPR Number

Implemented

Decimal

Hex

Mnemonic

Type

Scope

Where

Init?

Category

System Identification

SID

CPU

SOC/C

Translation Buffer Check

TBCHK

CPU

SOC/C

64-127

40-7F

Reserved

3.5 Process Structure
A process is a single thread of execution. The context of the current process is contained
in the Process Control Block (PCB) which is pointed to by the Process Control Block Base
Register (PCBB). ) The KA660 implements these structures as defined in the VAX Architecture
Reference Manual, which should be referenced for a description of the PCB and the PCBB.

3.6 Data Types
The central processor provides the following subset of the VAX data types:
Byte

Word

Longword

Quadword

Character string

Variable-length bit field

F-floating

G-floating

D-floating

Support for the remaining VAX data types can be provided via macrocode emulation.

3.7 Instruction Set
The KA660 CPU implements the following subset of the VAX instruction set types in microcode.
Integer arithmetic and logical

Address

Variable length bit field

Control

Procedure call

Miscellaneous

3–8 Central Processor

Operating system support

F_floating

G_floating

D_floating

Queue

Character string
•
•
•
•
•
•
•
•

MOVC3
MOVC5
CMPC3*
CMPC5*
LOCC*
SCANC*
SKPC*
SPANC*

* These instructions were in the microcode assisted category on the KA630-A (MicroVAX II) and therefore had
to be emulated.

The KA660 SOC CPU provides special microcode assistance to aid the macrocode emulation of
the following instruction groups:
•
•
•
•

Character string (except MOVC3, MOVC5, CMPC3*, CMPC5*, LOCC*, SCANC*, SKPC*,
SPANC*)
Decimal string
CRC
EDITPC

The following instruction groups are not implemented, but may be emulated by macrocode:
•
•

Octaword
Compatibility mode instructions

Appendix C lists the entire KA660 instruction set, indicating which instructions are implemented in the Floating Point Accelerator (FPA), and which instructions have microcode assists
to speed up macrocode emulation.

3.8 Memory Management
The KA660 implements full VAX Memory Management as defined in the VAX Architecture
Reference Manual . System space addresses are virtually mapped through single-level page
tables, and process space addresses are virtually mapped through two-level page tables. See
the VAX Architecture Reference Manual for descriptions of the virtual to physical address
translation process, and the format for VAX Page Table Entries (PTE’s).

Central Processor 3–9

3.8.1 Translation Buffer
To reduce the overhead associated with translating virtual addresses to physical addresses,
the KA660 employs a 28-entry, fully associative, translation buffer for caching VAX PTE’s in
modified form. Each entry can store a modified PTE for translating virtual addresses in either
the VAX Process Space, or VAX System Space. The translation buffer is flushed whenever the
following actions are performed:
•
•
•

Memory management is enabled or disabled (for example, by writes to IPR 56)
Any page table base or length registers are modified (for example, by writes to IPR’s 13:8)
IPR 57 (TBIA) or IPR 58 (TBIS) are written to.

Each entry is divided into two parts: a 23-bit Tag Register and a 32-bit PTE Register. The
Tag Register is used to store the Virtual Page Number (VPN) of the virtual page that the
corresponding PTE Register maps. The PTE register stores the 21-bit Page Frame Number
field, the PTE.V bit, the PTE.M bit and an 8-bit partially decoded representation of the 4-bit
VAX PTE PROT field, from the corresponding VAX PTE, as well as a Translation Buffer Valid
(TB.V) bit. The SOC CPU Design Spec can be referenced for details of the 8-bit PROT field.
During virtual-to-physical address translation, the contents of the 28 Tag Registers are compared with the Virtual Page Number Field (bits <31:9>) of the virtual address of the reference.
If there is a match with one of the Tag Registers then a translation buffer "hit" has occurred,
and the contents of the corresponding PTE register are used for the translation.
If there is no match, the translation buffer does not contain the necessary VAX PTE information to translate the address of the reference, and the PTE must be fetched from memory.
Upon fetching the PTE, the translation buffer is updated by replacing the entry that is selected
by the replacement pointer. Since this pointer is moved to the next sequential translation
buffer entry whenever it is pointing to an entry that is accessed, the replacement algorithm is
Not Last Used (NLU). This pointer is called the NLU pointer.

3.8.2 Memory Management Control Registers
There are four IPR’s that control the Memory Management Unit (MMU):
IPR 56 (MAPEN)
IPR 57 (TBIA)
IPR 58 (TBIS)
IPR 63 (TBCHK)
Memory management can be enabled/disabled via IPR 56 (MAPEN). Writing 0 to this register
with a MTPR instruction disables memory management and writing a 1 to this register with a
MTPR instruction enables memory management. Writes to this register flush the translation
buffer. To determine whether or not memory management is enabled, IPR 56 is read using the
MFPR instruction.
Translation buffer entries that map a particular virtual address can be invalidated by writing the virtual address to IPR 58 (TBIS) using the MTPR instruction. NOTE: Whenever
software changes a valid Page Table Entry for the system or current process region, or a System Page Table Entry that maps any part of the current process page

3–10 Central Processor

table, all process pages mapped by the Page Table Entry must be invalidated in the
translation buffer.
The entire translation buffer can be invalidated by writing a 0 to IPR 57 (TBIA) using the
MTPR instruction.
The translation buffer can be checked to see if it contains a valid translation for a particular
virtual page by writing a virtual address within that page to IPR 63 (TBCHK) using the
MTPR instruction. If the translation buffer contains a valid translation for the page, the
condition code V bit (bit<1> of the PSL) is set. NOTE: The TBIS, TBIA, and TBCHK IPRs
are write only. The operation of a MFPR instruction from any of these registers is
UNDEFINED.

3.9 Interrupts And Exceptions
Both interrupts and exceptions divert execution from the normal flow of control.
An interrupt is caused by some activity outside the current process and typically transfers
control outside the process (for example, an interrupt from an external hardware device). An
exception is caused by the execution of the current instruction and is typically handled by the
current process (for example, an arithmetic overflow).

3.9.1 Interrupts
Interrupts can be divided into two classes: non-maskable and maskable.
Non-maskable interrupts cause a halt via the hardware halt procedure. The hardware halt
procedure does the following:
•
•
•

Saves the PC, PSL, MAPEN<0> and a halt code in IPR’s
Raises the processor IPL to 1F
Passes control to the resident firmware

The firmware dispatches the interrupt to the appropriate service routine based on the halt
code and hardware event indicators. Non-maskable interrupts cannot be blocked by raising the
processor IPL, but can be blocked by running out of the Halt Protected Address Space (except
those non-maskable interrupts that generate a halt code of 3). Non-maskable interrupts with
a halt code of 3 cannot be blocked because this halt code is generated after a hardware reset).
Maskable interrupts cause the following:
•
•

•

The PC and PSL is saved
The processor IPL is raised to the priority level of the interrupt (except for Q22-bus, Mass
Storage and Network Interface interrupts where the processor IPL is set to 17 independent
of the level at which the interrupt was received)
The interrupt is dispatched to the appropriate service routine through the SCB.

Central Processor 3–11

The various interrupt conditions for the KA660 are listed in Table 3–4 along with their associated priority levels and SCB offsets.
Table 3–4: Interrupt Priority Levels
Priority Level

Interrupt Condition

SCB Offset

Non-maskable

BDCOK and BPOK negated then asserted on Q22-bus
(Power Up)

BDCOK negated then asserted while BPOK asserted on
Q22-bus (SCR<7> clear) .

BDCOK negated then asserted while BPOK asserted on
Q22-bus (SCR<7> set).

BINIT asserted on Q22-bus when configured as an auxiliary

BHALT asserted on Q22-bus

BREAK generated by the console device

Unused

BPOK negated on Q22-bus

CDAL Bus parity error

Q22-bus NXM on a write

Uncorrectable main memory errors

CDAL Bus timeout during DMA

Main memory NXM errors

1C:1B

Unused

Correctable main memory errors

19:18

Unused

BR7 L asserted

Q22-bus Vector plus
20016

Interval Timer Interrupt

BR6 L asserted

Q22-bus Vector plus
20016

BR5 L asserted

Q22-bus Vector plus
20016

Console Terminal

F8,F6

Programmable Timers

78,7C

SHAC Mass Storage Interface (DSSI port 1)(External)

104

SGEC Network Interface

10C

Interprocessor Doorbell

204

* —These conditions generate a hardware halt procedure with a halt code of 3 (hardware reset).
**—These conditions generate a hardware halt procedure with a halt code of 2 (external halt).

3–12 Central Processor

Table 3–4 (Cont.): Interrupt Priority Levels
Priority Level

Interrupt Condition

SCB Offset

BR4 L asserted

Q22-bus Vector plus
20016

13:10

Unused

0F:01

Software interrupt requests

84-BC

* —These conditions generate a hardware halt procedure with a halt code of 3 (hardware reset).
**—These conditions generate a hardware halt procedure with a halt code of 2 (external halt).

NOTE
Because the Q22-bus does not allow differentiation between the four bus grant
levels (for example, a level 7 device could respond to a level 4 bus grant), the KA660
CPU raises the IPL to 17 after responding to interrupts generated by the assertion
of either BR7 L, BR6 L, BR5 L, or BR4 L. The KA660 maintains the IPL at the
priority of the interrupt for all other interrupts.
The interrupt system is controlled by three IPR’s:
•
•
•

IPR 18, the Interrupt Priority Level Register (IPLR) (Figure 3–4), is used for loading the
processor priority field in the PSL (bits<20:16>).
IPR 20, the Software Interrupt Request Register (SIRR) (Figure 3–5), is used for creating
software interrupt requests.
IPR 21, the Software Interrupt Summary Register (SISR) (Figure 3–6), records pending
software interrupt requests at levels 1 through 15.

The format of these registers is presented in Figure 3–4, Figure 3–5, and Figure 3–6. Refer to
the VAX Architecture Reference Manual for more information on these registers.

Figure 3–4: Interrupt Priority Level Register (IPLR) - (IPR 1810 1216 )

3
1

5 4
Ignored, Returns 0

PSL<20:16>

:IPLR
ESB90P0003

Central Processor 3–13

Figure 3–5: Software Interrupt Request Register (SIRR) - (IPL 2010 1416 )

3
1

4 3
Ignored

Request

:SIRR

ESB90P0004

Figure 3–6: Software Interrupt Summary Register (SISR) - (IPL 2110 1516 )
3
1

1
6

1
5

0
Pending Software Interrupts
FEDCBA9

M
B
1 Z

:SISR

ESB90P0005

3.9.2 Exceptions
Exceptions can be divided into 3 types: trap, fault and abort.
A trap is an exception that occurs at the end of the instruction that caused the exception.
After an instruction traps, the PC saved on the stack is the address of the next instruction
that would have normally been executed and the instruction can be restarted.
A fault is an exception that occurs during an instruction. It leaves the registers and memory
in a consistent state, such that the elimination of the fault condition and restarting the
instruction will give correct results. After an instruction faults, the PC saved on the stack
points to the instruction that faulted.
An abort is an exception that occurs during an instruction, leaving the value of registers and
memory unpredictable, such that the instruction cannot necessarily be correctly restarted,
completed, simulated or undone. After an instruction aborts, the PC saved on the stack points
to the instruction that was aborted (which may or may not be the instruction that caused the
abort) and the instruction may or may not be restarted depending on the class of the exception
and the contents of the parameters that were saved.
Exceptions can be divided into six classes. A list of exceptions, grouped by class, is given in
the table below. The exceptions listed in Table 3–5 (except machine check) are described in
greater detail in the VAX Architecture Reference Manual. The machine check exception is
described in greater detail in section Section 3.9.3.2.

3–14 Central Processor

Table 3–5: Exception Classes
Exception Class

Type

SCB Offset

Trap

Arithmetic Exceptions
Integer overflow
Integer divide by zero

Trap

Subscript range

Trap

Floating overflow

Fault

Floating divide by zero

Fault

Floating underflow

Fault

Access control violation

Fault

Translation not valid

Fault

Memory Management Exceptions

Operand Reference Exceptions
Reserved addressing mode
Reserved operand

Instruction Execution Exceptions
Reserved/Privileged instruction

Fault

Emulated instruction

Fault

C8,CC

Change mode

Trap

40-4C

Breakpoint

Fault

Tracing Exception
Trace
System Failure Exceptions
Interrupt stack not valid

abort

Kernel stack not valid

abort

Machine check including the following:

abort

•
•
•
•
•
•
•
•

CDAL bus parity errors
Cache parity errors
Q22-bus NXM Errors
Q22-bus device parity Errors
Q22-bus NO GRANT Errors
CDAL Bus Timeout Errors
Main memory NXM Errors
Main Memory Uncorrectable Errors

* —Dispatched by resident firmware rather than through the SCB

Central Processor 3–15

3.9.3 Information Saved On A Machine Check Exception
In response to a machine check exception, the following information is pushed onto the stack
as shown in Figure 3–7:
•
•
•
•

Contents of the Processor Status Longword
Contents of the Program Counter
Four parameters
A Byte Count

Figure 3–7: The Processor Stack After a Machine Check Exception

3
1

0
Byte Count (00000010) hex
Machine Check Code
Most Recent Memory Address
Internal State Information 1
Internal State Information 2
PC
PSL
ESB90P0089

The diagram of the stack pointer is explained in the following paragraphs.
3.9.3.1

Byte Count

Byte Count<31:0> 0000001016 , 1610 . This value indicates the number of bytes of information
that follow on the stack (not including the PC and PSL).
3.9.3.2

Machine Check Code Parameter

Machine Check Code<31:0> A code value that indicates the type of machine check that occurred. A list of the possible machine check codes (in hex) and their associated causes follows:

3–16 Central Processor

Floating Point Errors - These codes indicate that the Floating Point Accelerator (FPA) detected an error while communicating with the CPU during the execution of a floating point
instruction. The most likely cause of these types of machine checks is a problem internal to
the SOC CPU chip. Machine checks due to floating point errors MAY BE RECOVERABLE, depending on the state of the VAX CAN’T RESTART flag (captured in Internal State Information
2 <15> and the FIRST PART DONE flag (captured in PSL <27>). If the FIRST PART DONE
flag is set, the error is recoverable. If the FIRST PART DONE FLAG is cleared, then the VAX
CAN’T RESTART flag must also be cleared for the error to be recoverable. Otherwise, the
error is unrecoverable and depending on the current mode, either the current process or the
operating system should be terminated. The information pushed on the stack by this type of
machine check is from the instruction that caused the machine check.
Table 3–6: Floating Point Error Machine Checks
Hex Code

Error Description

A protocol error was detected by the FPA while attempting to execute a floating point
instruction.

A reserved instruction was detected by the FPA while attempting to execute a floating
point instruction.

An illegal status code was returned by the FPA while attempting to execute a floating
point instruction. ?CPSTA<1:0>=10

An illegal status code was returned by the FPA while attempting to execute a floating
point instruction. ?CPSTA<1:0>=01

Memory Management Errors - These codes indicate that the microcode in the SOC CPU
chip detected an impossible situation while performing functions associated with memory
management. The most likely cause of this type of a machine check is a problem internal to
the SOC chip. Machine checks due to memory management errors are NON-RECOVERABLE.
Depending on the current mode, either the current process or the operating system should be
terminated. The state of the P0BR, P0LR, P1BR, P1LR, SBR and SLR should be logged.
Table 3–7: Memory Management Error Machine Checks
Hex Code

Error Description

The calculated virtual address for a process PTE was in the P0 space instead of the
System Space when the CPU attempted to access a process PTE after a translation
buffer "miss".

The calculated virtual address for a process PTE was in the P1 space instead of the
System Space when the CPU attempted to access a process PTE after a translation
buffer "miss".

The calculated virtual address for a process PTE was in the P0 space instead of the
System Space when the CPU attempted to access a process PTE to change the PTE<M>
bit before writing to a previously unmodified page.

The calculated virtual address for a process PTE was in the P1 space instead of the
System Space when the CPU attempted to access a process PTE to change the PTE<M>
bit before writing to a previously unmodified page.

Central Processor 3–17

Interrupt Errors - This code indicates that the interrupt controller in the SOC CPU requested a
hardware interrupt at an unused hardware IPL. The most likely cause of this type of a machine
check is a problem internal to the SOC CPU chip. Machine checks due to unused IPL errors are
NON-RECOVERABLE. A non-vectored interrupt generated by a serious error condition (memory error, power fail or processor halt) has probably been lost. The operating system should be
terminated.
Table 3–8: Interrupt Error Machine Checks

Hex Code

Error Description

A hardware interrupt was requested at an unused Interrupt Priority Level (IPL)

Microcode Errors - This code indicates that an impossible situation was detected by the
microcode during instruction execution. Note that most erroneous branches in the SOC CPU
microcode will cause random microinstructions to be executed. The most likely cause of this
type of machine check is a problem internal to the SOC CPU chip. Machine checks due
to microcode errors are NON-RECOVERABLE. Depending on the current mode, either the
current process or the operating system should be terminated.
Table 3–9: Microcode Error Machine Checks
Hex Code

Error Description

An impossible state was detected during a MOVC3 or MOVC5 instruction (not move
forward, move backward, or fill).

Read Errors - These codes indicate that an error was detected while the SOC CPU was attempting
to read from either the cache, main memory, or the Q22-bus. The most likely cause of this type of
machine check must be determined from the state of the MSER, DSER, MEMCSR16, MEAR and
SEAR. Machine checks due to read errors MAY BE RECOVERABLE, depending on the state of the
VAX CAN’T RESTART flag (captured in Internal State Information 2 <15> and the FIRST PART
DONE flag (captured in PSL <27>). If the FIRST PART DONE flag is set, the error is recoverable.
If the FIRST PART DONE FLAG is cleared, then the VAX CAN’T RESTART FLAG must also be
cleared for the error to be recoverable. Otherwise, the error is unrecoverable and depending on
the current mode, either the current process or the operating system should be terminated. The
information pushed on the stack by this type of machine check is from the instruction that caused
the machine check.
Table 3–10: Read Error Machine Checks
Hex Code

Error Description

An error occurred while reading an operand, a Process Page Table Entry during address
translation, or on any read generated as part of an interlocked instruction.

An error occurred while reading a System Page Table Entry (SPTE), during address
translation, a Process Control Block (PCB) entry during a context switch, or a System
Control Block (SCB) entry while processing an interrupt.

3–18 Central Processor

Write Errors - These codes indicate that an error was detected while the SOC CPU was attempting
to write to either the cache, main memory, or the q22-bus. The most likely cause of this type of
machine check must be determined from the state of the MSER, DSER, MEMCSR16, MEAR, SEAR
and CBTCR. Machine checks due to write errors are NON-RECOVERABLE because the CPU is
capable of performing many read operations out of the cache before a write operation completes.
For this reason, the information that is pushed onto the stack by this type of machine check cannot
be guaranteed to be from the instruction that caused the machine check.
Table 3–11: Write Error Machine Checks
Hex Code

Error Description

An error occurred while writing an operand, or a Process Page Table Entry to change
the PTE<M> bit before writing a previously unmodified page.

An error occurred while writing a System Page Table Entry (SPTE) to change the
PTE<M> bit before writing a previously unmodified page, or a Process Control Block
(PCB) entry during a context switch or during the execution of instructions that modify
any stack pointers stored in the PCB.

3.9.3.3

Most Recent Virtual Address Parameter

Most Recent Virtual Address <31:0> captures the contents of the Virtual Address Pointer
Register at the time of the machine check. If a machine check other than a machine check
81 occurred on a read operation, this field represents the virtual address of the location that
was being read when the error occurred, plus four. If a machine check 81 occurred, this field
represents the physical address of the location that was being read when the error occurred,
plus four. If a machine check other than a machine check 83 occurred on a write operation,
this field represents the virtual address of a location that was being referenced either when
the error occurred, or sometime after the error occurred, plus four. If a machine check 83
occurred, this field represents the physical address of the location that was being referenced
either when the error occurred, or sometime after the error occurred, plus four. In other words,
if the machine check occurred on a write operation, the contents of this field cannot be used
for error recovery.
3.9.3.4

Internal State Information 1 Parameter

Internal State Information 1 is divided into five fields. The contents of these fields are described below.

Central Processor 3–19

Table 3–12: Internal State Information 1 Field Description
Bit Field

Description

<31:24>

This field captures the opcode of the instruction that was being read or executed at the
time of the machine check.

<23:20>

This field is read as <1111>.

<19:16>

This field captures the current contents of HSIR<3:0>.

<15:8>

This field captures the state of CCR<7:0> at the time of the machine check. See section
5.2.5 for interpretation of the contents of this register.

<7:0>

This field captures the state of MSER<7:0> at the time of the machine check. See section
5.2.6 for interpretation of the contents of this register.

3.9.3.5

Internal State Information 2 Parameter

Internal State Information 2 is divided into seven fields. The contents of these fields are
described below.
Table 3–13: Internal State Information 2 Field Description
Bit Field

Description

<31:24>

This field captures internal state of the SOC CPU chip at the time of the machine check.
This field contains the SC register <7:0>.

<23:22>

This field is read as <11>.

<21:16>

This field contains the State Flags <5:0>.

<15>

This field captures the state of the VAX CAN’T RESTART flag at the time of the machine
check.

<14:12>

This field is read as <111>.

<11:8>

These bits contain the Arithmetic Logic Unit (ALU) condition codes.

<7:0>

This field captures the offset between the virtual address of the start of the instruction
being executed at the time of the machine check (saved PC) and the virtual address of the
location being accessed (PC) at time of the machine check.

3.9.3.6

PC<31:0> Captures the virtual address of the start of the instruction being executed at the
time of the machine check.
3.9.3.7

PSL

PSL<31:0> Captures the contents of the PSL at the time of the machine check.

3–20 Central Processor

3.9.4 System Control Block (SCB)
The System Control Block (SCB) consists of two pages in main memory that contain the
vectors by which interrupts and exceptions are dispatched to the appropriate service routines.
The SCB is pointed to by IPR 17, the System Control Block Base Register (SCBB). The format
of the System Control Block Base Register is shown in Figure 3–8.
Figure 3–8: System Control Block Base Register (SCBB)

System Control Block
10C

System Control Block
Base Register
31

MBZ

9 8
Physical Longword Address of
System Control Block

0
MBZ

The description of the format is listed in Table 3–14.

Table 3–14: The System Control Block Format
SCB
Offset

Interrupt/Exception Name

Type

#
Params Notes

Unused

IRQ passive release on other
VAXes.

Machine Check

Abort

Parameters depend on error
type.

Kernel Stack Not Valid

Abort

Must be serviced on interrupt
stack

Power Fail

Interrupt

IPL is raised to 1E

Reserved/Privileged Instruction

Fault

Customer Reserved Instruction

Fault

XFC instruction

Reserved Operand

Fault/Abort

Not always recoverable

Reserved Addressing Mode

Fault

Central Processor 3–21

Table 3–14 (Cont.): The System Control Block Format
SCB
Offset

Interrupt/Exception Name

Type

#
Params Notes

Access Control Violation

Fault

Translation Not Valid

Fault

Trace Pending (TP)

Fault

Breakpoint Instruction

Fault

Unused

Compatibility mode in other
VAXes

Arithmetic

Trap/Fault

Parameter is type code

38-3C

Unused

CHMK

Trap

Parameter is sign-extended
operand word

CHME

Trap

Parameter is sign-extended
operand word

CHMS

Trap

Parameter is sign-extended
operand word

CHMU

Trap

Parameter is sign-extended
operand word

Unused

Parameters are virtual address, status code
2 Parameters are virtual
address, status code

Correctable main memory errors

Interrupt

58-5C

Unused

Memory Error

Interrupt

64-74

Unused

Programmable Timer 0

Interrupt

IPL is 14

Programmable Timer 1

Interrupt

IPL is 14

Unused

Software Level 1

Interrupt

Software Level 2

Interrupt

Ordinarily used for AST
delivery

Software Level 3

Interrupt

Ordinarily used for process
scheduling

90-BC

Software Levels 4-15

Interrupt

Interval Timer

Interrupt

Unused

Emulation Start

Fault

3–22 Central Processor

IPL is 1A (CRD_L)
IPL is 1D (MEMERR_L)

IPL is 16(INTTIM_L)
Same mode exception,FPD =
0; parameters are opcode, PC,
specifiers

Table 3–14 (Cont.): The System Control Block Format
SCB
Offset

Interrupt/Exception Name

Type

#
Params Notes

Emulation Continue

Fault

D0-DC

Unused

E0-EC

Unused

F0-F4

Unused

Console Receiver

Interrupt

IPL is 14

Console Transmitter

Interrupt

IPL is 14

104

Mass Storage Interface (DSSI
PORT)

Interrupt

IPL is 14

10C

Network Interface

Interrupt

IPL is 14

Same mode exception,FPD =
1: no parameters

3.9.5 Hardware Detected Errors
The KA660 is capable of detecting eleven types of error conditions during program execution:
1. CDAL Bus parity errors indicated by MSER<6> (on a read) or MEMCSR16<7> (on a write)
being set.
2. Cache Tag parity errors indicated by MSER<0> being set.
3. Cache Data parity errors indicated by MSER<1> being set.
4. Q22-bus NXM errors indicated by DSER<7> being set.
5. Q22-bus NO SACK errors (no indicator).
6. Q22-bus NO GRANT errors indicated by DSER<2> being set.
7. Q22-bus device parity errors indicated by DSER<5> being set.
8. CDAL-Bus timeout errors indicated by DSER<4>(only on DMA) being set.
9. Main Memory NXM errors indicated by DSER<0> (only on DMA) being set.
10. Main Memory correctable errors indicated by MEMCSR16<29> being set.
11. Main Memory uncorrectable errors indicated by MEMCSR16<31> and DSER<4> (only on
DMA) being set.
These errors will cause either a Machine Check Exception, a Memory Error Interrupt, or a
Corrected Read Data Interrupt depending on the severity of the error and the reference type
that caused the error.

Central Processor 3–23

3.9.6 The Hardware Halt Procedure
The Hardware Halt Procedure is the mechanism by which the hardware assists the firmware
in emulating a processor halt. The hardware Halt Procedure saves the current value of the
PC in IPR 42 (SAVPC), and the current value of the PSL, MAPEN<0>, a halt code and VALID
bit in IPR 43 (SAVPSL). The formats for (SAVPC) and (SAVPSL) are shown in Figure 3–9 and
Figure 3–10 respectively.

Figure 3–9: Console Saved PC (SAVPC) - (IPR 4210 2A16 )

3
1

0
Saved PC

(Read Only)

:SAVPC
ESB90P0008

Figure 3–10: Console Saved PSL (SAVPSL) - (IPR 4310 2B16 )

3
1

1 1 1 1
6 5 4 3
PSL<31:16>

8 7

HALT CODE

0
PSL<7:0>

:SAVPSL

MAPEN<0>
Valid Bit (Valid if ZERO)
ESB90P0009

The current stack pointer is saved in the appropriate internal register. The PSL is set to 041F
0000 16 (IPL=1F, kernel mode, using the interrupt stack) and the current stack pointer is
loaded from the interrupt stack pointer. Control is then passed to the resident firmware at
physical address 2004 0000 16 . Table 3–15 shows the state of the CPU.

3–24 Central Processor

Table 3–15: CPU State After a HALT
Register

New Contents

SAVPC

Saved PC

SAVPSL<31:16,7:0>

Saved PSL<31:16,7:0>

SAVPSL<15>

Saved MAPEN<0>

SAVPSL<14>

Valid PSL flag (unknown for halt code of 3)

SAVPSL<13:8>

Saved restart code

Current Interrupt Stack (IPR 4)

PSL

041F 0000 16

2004 0000 16

MAPEN

ICCS

0 (for a halt code of 3)

MSER

0 (for a halt code of 3)

CCR

0 (for a halt code of 3)

SISR

0 (for a halt code of 3)

ASTLVL

0 (for a halt code of 3)

All else

Undefined

The firmware uses the halt code in combination with hardware event indicators to dispatch
the interrupt or exception that caused the halt, to the appropriate firmware routine (either
console emulation, power-up, reboot, or restart). A list of halt codes and their event indicators
is given in Table 3–16. Complete halt code descriptions are given in Table I–1.
Table 3–16: HALT Codes
Halt Code

Interrupt Condition

Event Indicator

HALT Codes for UnMaskable Interrupts
2

External Halt (SOC/C HALT_L pin asserted)
BHALT asserted on the Q22-bus

DSER<15>

Remote Boot serviced by SGEC

NICSR5<7>

BDCOK negated and asserted on the Q22-bus while BPOK stays
asserted (Q22-bus REBOOT/RESTART) and SCR<7> is set

DSER<14>

BREAK generated by the console

RXDB<11>

Hardware Reset (SOC/C RESET pin asserted)
BDCOK and BPOK negated then asserted on the Q22-bus (PowerUp)

Central Processor 3–25

Table 3–16 (Cont.): HALT Codes
Halt Code

Interrupt Condition

BDCOK negated and asserted on the Q22-bus while BPOK stays
asserted (Q22-bus REBOOT/RESTART) and SCR<7> is clear.
HALT Codes for Exceptions
6

Event Indicator
-

HALT instruction executed in Kernel Mode

HALT Codes for Exceptions that occurred while Serving an Interrupt or Exception
4

Interrupt stack not valid during exception

Machine check during normal exception

SCB vector bits<1:0>= 11

SCB vector bits<1:0>= 10

CHMx executed while on interrupt stack

Access Control Violation (ACV) or Translation Not Valid (TNV)
during machine check exception

ACV or TNV during kernel stack not valid exception

Machine check during machine check exception

Machine check during kernel stack not valid exception

PSL<26:24>= 101 during interrupt or exception

PSL<26:24>= 110 during interrupt or exception

PSL<26:24>= 111 during interrupt or exception

PSL<26:24>= 101 during REI

PSL<26:24>= 110 during REI

PSL<26:24>= 111 during REI

3.10 System Identification
The firmware and operating system software references two registers to determine the processor on which they are running. The first, the System Identification register (SID), is an
internal processor register. The second, the System Identification Extension register (SIE), is
a firmware register located in the on board EPROM.

3.10.1

System Identification Register

The System Identification Register (SID), IPR 62, is a read-only register implemented in the
SOC CPU chip . This 32-bit, Read-Only register is used to identify the processor type and its
microcode revision level. The SID longword is read from IPR 62 using the MFPR instruction.
This longword value is processor specific. The format is shown in Figure 3–11. Bit definitions
are listed in Table 3–17.

3–26 Central Processor

Figure 3–11: System Identification Register (SID) - (IPR 6210 3E16 )

3
1

2 2
4 3
CPU_TYPE

8 7
RESERVED

MICROCODE REV.
ESB90P0010

Table 3–17:
Field

Name

Description

<31:24>

CPU_TYPE

CPU type is the processor specific identification code. This field
always reads as 2010 indicating the processor is implemented with
an SOC CPU chip.

<23:8>

RESERVED

Reserved for future use.

<7:0>

VERSION

Version of the microcode.

3.10.2

System Identification Extension Register (SIE) (20040004)

The System Identification Extension register is an extension of the SID register and is used to
further differentiate between hardware configurations. The SID register identifies which CPU
and microcode is executing, and the SIE register identifies what module and firmware revision
are present. Note, the fields in this register are dependent on SID<31:24>(CPU_TYPE).
By convention, all MicroVAX systems implement a longword at physical location 20040004 in
the firmware EPROM for the SIE register. This 32-bit Read-Only register is implemented in
the KA660 ROM. Figure 3–12 shows the format of this register. Table 3–18 lists the definitions
of the register bits.

Central Processor 3–27

Figure 3–12: System Identification Extension Register (SIE)

3
1

2 2
4 3
SYS_TYPE

1 1
6 5
REV LEVEL

8 7

SYS_SUB_TYPE

0
RESERVED
ESB90P0011

Table 3–18:
Field

Name

Description

31:24

SYS_TYPE

This field identifies the type of system for a specific processor.
01 : Q22-bus single processor system.

23:16

VERSION

This eight bit field contains two hexadecimal digits that reflect
the version of the resident firmware (EPROM). For example, if the
firmware version is V5.0, the banner will display V5.0 and this field
will encode as a 5016 .

15:8

SYS_SUB_
TYPE

This field reflects the particular system sub-type.

7:0

RESERVED

01 : KA650
02 : KA640
03 : KA655
04 : KA670
This field is reserved.

3.11 CPU References
All references by the CPU can be classified into one of three groups:
•
•
•

Request Instruction-Stream Read references
Demand Data-Stream Read references
Write references

3–28 Central Processor

3.11.1 Request Instruction-Stream Read References
The CPU has an instruction prefetcher for prefetching program instructions from either cache
or main memory. The prefetcher uses a 12-byte (3 longword) Instruction Prefetch Queue (IPQ).
Whenever there is an empty Longword in the IPQ, and the prefetcher is not halted due to an
error, the instruction prefetcher will generate an aligned longword Request Instruction-Stream
(I-Stream) read reference.

3.11.2 Demand Data-Stream Read References
Whenever data is immediately needed by the CPU to continue processing, a Demand DataStream (D-Stream) read reference is generated. More specifically, Demand D-Stream references are generated on the following references:
•
•
•
•

Operand
Page Table Entry (PTE)
System Control Block (SCB)
Process Control Block (PCB)

When interlocked instructions, such as Branch on Bit Set and Set Interlock (BBSSI) are
executed, a Demand D-Stream Read-Lock reference is generated.
Because the CPU does not impose any restrictions on data alignment (other than the aligned
operands of the ADAWI and interlocked queue instructions) and because memory can only
be accessed one aligned longword at a time, all data read references are translated into an
appropriate combination of masked and unmasked, aligned longword read references.
If the required data is a byte, a word within a longword or an aligned longword then a single,
aligned longword Demand D-Stream read reference is generated. If the required data is a
word that crosses a longword boundary, or an unaligned longword then two successive aligned
longword Demand D-Stream read references are generated. Data larger than a longword
is divided into a number of successive aligned longword Demand D-Stream reads, with no
optimization.

3.11.3 Write References
Whenever data is stored or moved, a write reference is generated. Because the CPU does not
impose any restrictions on data alignment (other than the aligned operands of the ADAWI
and interlocked queue instructions) and because memory can only be accessed one aligned
longword at a time, all data write references are translated into an appropriate combination of
masked and unmasked aligned longword write references.
If the required data is a byte, a word within a longword , or an aligned longword then a single,
aligned longword write reference is generated. If the required data is a word that crosses
a longword boundary, or an unaligned longword then two successive aligned longword write
references are generated. Data larger than a longword is divided into a number of successive
aligned longword writes.

Central Processor 3–29

Chapter 4
KA660 Cache Memory

To maximize CPU performance, the SOC CPU provides an 6 Kilo Byte, 6-way associative
write-through cache with an 8 byte block and fill size.

4.1 Cacheable References
Any reference that is stored by the cache is called a cacheable reference. The cache stores CPU
read references to the VAX Memory Space (bit <29> of the physical address equals (0)) only. It
does not store references to the VAX I/O space, or DMA references by the Q22-Bus Interface,
the DSSI interface, or the Ethernet interface.
Both I-Stream and D-Stream references are cached in the enabled banks of the cache.
Whenever the CPU generates a non-cacheable reference, a single longword reference of the
same type is generated on the CDAL Bus.
Whenever the CPU generates a cacheable reference that is stored in the cache, no reference is
generated on the CDAL Bus.
Whenever the CPU generates a cacheable READ reference that is not stored in the cache,
a quadword READ is generated on the CDAL Bus. If the CPU reference was a Request IStream Read, then the quadword transfer consists of two indivisible longword transfers. The
first transfer is a Request I-Stream Read (prefetch) and the second transfer is a Request IStream Read (fill). If the CPU reference was a Demand D-Stream Read, then the quadword
transfer consists of two indivisible longword transfers. The first is a Demand D-Stream Read
and the second is a Request D-Stream Read (fill).
If the CPU is retried on the second (fill) longword then the first longword is delivered to the
CPU but not cached. The request will not be retried if the need for the data is resolved (for
example, it was a prefetch but a branch was taken).

KA660 Cache Memory 4–1

4.2 Cache Organization
The cache is logically organized as 6, 1KB banks as shown below. It is a read-allocate, nowrite-allocate, and write-through cache. A single parity bit is generated, stored and checked
for each byte of data and each tag. The cache is enabled/disabled via a bit in the Cache
Control Register (CCR).
Other bits in these two registers allow the data and tags to be addressed directly, to check the
parity generating/checking logic and to provide the hit/miss status of each bank for the most
recent D-stream read or write cycle.

Figure 4–1: Logical Organization of Cache

Set 0
Row 1
Row 2
.
.
.
.
Row 128

128x21−bit
Tag Array

Set 4

Set 5

128x72−bit
Data Array
...

72 71

0
ESB90P0012

Each row within a set corresponds to a cache entry, and there are 128 entries in each set.
Each entry contains a 21-bit Tag Block and a 72-bit (eight-byte) Data Block. A cache entry is
logically organized as shown in Figure 4–2:
Figure 4–2: Cache Entry

9
3

77
21
Tag Block

0
Data Block
ESB90P0013

4–2

KA660 Cache Memory

Figure 4–3 shows the format of a cache tag entry.
Figure 4–3: Cache Tag Entry

20 19 18
P

Tag
Tag
Valid
Parity
ESB90P0014

The Parity bit stores odd parity over the Tag field only. The Valid bit is not included in the
parity calculation.
The Valid bit indicates whether or not the corresponding entry in the cache refers to a useable
data/address pair.
The Tag consists of bits <28:10> of the physical address.
Figure 4–4 shows the format of a cache data entry.
Figure 4–4: Cache Data Entry

63 56

55 48

47 40

39 32

31 24

23 16

Byte 7

Byte 6

Byte 5

Byte 4

Byte 3

Byte 2

Byte 1

Parity
Parity

Byte 0
Parity

Parity
ESB90P0015

Each Data entry consists of eight bytes of data, each with an associated parity bit. Odd parity is
generated for the odd data bytes and even parity is generated for the even data bytes.

KA660 Cache Memory 4–3

4.2.1 Cache Address Translation
Whenever the CPU requires an instruction or data, the contents of the cache is checked
to determine if the referenced location is stored there. The cache contents is checked by
translating the physical address as follows:
On non-cacheable references, the reference is never stored in the cache, so a cache "miss"
occurs and a single longword reference is generated on the CDAL Bus.
On cacheable references, the physical address must be translated to determine if the contents
of the referenced location is resident in the cache. The Cache Index Field, bits <9:3> of the
physical address, is used to select one of the 128 rows of the cache, with each row containing
a single entry from each set. The Label Field, bits <28:10> of the physical address, is then
compared to the Tag Block of the entry from all 6 sets in the selected row.
If a match occurs with the Tag Block of one of the set entries, and the valid bit within the
entry is set, the contents of the referenced location is contained in the cache and a cache "hit"
occurs. On a cache hit, the Set Match Signals generated by the compare operation select the
data block from the appropriate set. The Cache Displacement Field, bits <2:0> of the physical
address, is used to select the byte(s) within the block. No CDAL Bus transfers are initiated on
CPU references that "hit" the cache.
If no match occurs, then the contents of the referenced location is not contained in the cache
and a cache "miss" occurs. In this case, the data must be obtained from the main memory
controller, so a quadword transfer is initiated on the CDAL Bus.
Figure 4–5 shows how cache addresses are translated.

4–4

KA660 Cache Memory

Figure 4–5: Cache Address Translation
2 2
9 8

10 9

3 2

Label

0
Physical
Address

I/O Space

Cache Index
Cache Displacement

Valid Bit
|
V
21−
Bit
Tag

Valid Bit
|
V

Set 0
72−Bit
Data Block

_ _____ _______________
_ _____ _______________

...
|
|
<...>

21−
Bit
Tag

Set 5
72−Bit
Data Block

_ _____ _______________
_ _____ _______________

...
V

Set

0 Match?

1−5 Match?

...
V

V
Data
ESB90P0016

KA660 Cache Memory 4–5

4.2.2 Cache Data Block Allocation
Cacheable references that "miss" the cache, cause a quadword read to be initiated on the
CDAL bus. When the requested quadword is supplied by the main memory controller, the
requested longword is passed on to the CPU, and a data block is allocated in the cache to store
the entire quadword.
Because the cache is six-way associative, there are six data blocks (one in each set) that can be
allocated to a given quadword. The Cache Index field determines which row of 6 data blocks is
to be used while set selection is determined by a Not Most Recently Used (NMRU) algorithm.
The bank to be selected is pointed to by a three bit counter. The counter is set to 000 when
the cache is disabled (CCR ENABLE_CACHE=0); it is advanced:
•
•

Every time a block is allocated
Every time a read or write hits in the bank to which the counter is currently pointing.

The counter counts modulo 6 with no missing counts regardless of which banks are enabled. The contents of the counter is driven out on pins TEST<2:0> (test mode is in Observe
Miscellaneous State), which enables external logic to track which addresses are cached internally.

4.2.3 Cache Behavior on Writes
On CPU generated write references, the cache is "write through". All CPU write references
that "hit" the cache cause the contents of the referenced location in main memory to be updated as well as the copy in the cache.
When a DMA write references hits the cache, the cache entry containing the copy of the
referenced location is invalidated.

4.2.4 Cache Control Register (CCR, IPR 37)
The Cache Control Register (CCR), Internal Processor Register 37, controls the mode of operation of the cache, flushing the cache and is unique to CPU designs that use the SOC CPU chip.
Figure 4–6 shows the register format and Table 4–1 describes the bits.

4–6

KA660 Cache Memory

Figure 4–6: Cache Control Register

3
1
MBZ

W
W
P

E
N
A

F
L
U

D
I
G

ESB90P0017

Table 4–1:
Data Bit

Name

Description

Type

<31:5>

Unused

These bits always read as zeros. Writes have no
effect.

<4>

Unused

This bit is always read as a one. Writes have no
effect.

<3>

WWP

Write Wrong Parity. For diagnostic use. When
set, it inverts the data parity bits accessed from
the cache. This will cause a parity error when
they are compared to parity computed from the
accessed data, and is used to test the parity
generation/checking logic. When clear data parity
bits are not affected. This bit does not affect tag
parity bits. Cleared when RESET_L is asserted.

Read/Write

KA660 Cache Memory 4–7

Table 4–1 (Cont.):
Data Bit

Name

Description

Type

<2>

ENA

Enable Cache. It enables or disables normal operation of the cache. When set, both I-stream and
D-stream references are cached in the enabled
banks of the cache, and cache tag and data parity
errors are reported. When clear, all references
(read and write) result in a miss and no cache
parity is checked. When Enable Cache is set the
cache should be flushed by writing a one to the
Flush Cache bit. Enable Cache is cleared when
RESET_L is asserted.

Read/Write

NOTE
The cache may be operated
with both the Diagnostic bit
set and the Enable Cache bit
set.
<1>

FLU

Flush Cache. Always read as zero. Writing a one
to this bit clears all valid bits in the cache tag
array. Writing a zero has no effect.

Write Only

<0>

DIA

Diagnostic. When this bit is set, the cache entries
and BEHR register may be accessed via a region
in I/O space. When clear, references to the same
region of I/O space result in bus cycles.

Read/Write

Table 4–2: Cache Diagnostic Mode Addresses
Information

Address Range

Cache Tag

20150000 - 201503FF

Cache Data

20150400 - 201507FF

BEHR

20150800 - 20150FFF

Note the BEHR register may be accessed at multiple addresses.
When the Diagnostic bit is set, writes to cache data addresses will write to the cache data
longword indexed by bits <9:2> of the address. All banks which have the corresponding BEHR
ENABLE_BANK bit set will be written, and correct data parity will also be written. Byte and
word writes as well as longword are possible.
Reads from cache data addresses read the cache data longword addressed; if more than one
ENABLE_BANK bit is set then the highest priority bank enabled will return the data (bank 0
is highest priority, bank 7 is lowest).

4–8

KA660 Cache Memory

Writes to cache tag addresses write to the cache tag indexed by bits <9:3> of the address. All
banks which have the corresponding BEHR ENABLE_BANK bit set will be written. The write
data format is shown in Figure 4–7.
Figure 4–7: Tag Diagnostic Write Data Format

3
1

3
0

2
9

2
8

1
0 9
TAG

0
x

_ Valid Bit
Tag Parity
ESB90P0018

Reads from cache tag addresses read the cache tag addressed, plus the data parity from
the data longword addressed by bits <9:2> of the address as shown below. If more than one
ENABLE_BANK bit is set then the highest priority bank enabled will return the tag.
The read data format is shown in Figure 4–8.
Figure 4–8: Tag Diagnostic Read Data Format

3
1

3
0

2
9

2
8

Valid Bit
Tag Parity

1
0 9
TAG

4 3
x

Data Parity
of associated
Longword
ESB90P0019

KA660 Cache Memory 4–9

4.2.5 Bank Enable/Hit Miss Register (BEHR)
The BEHR register allows individual sets of the cache to be enabled/disabled and also provides bits
which indicate the hit/miss status for each set of the cache. The format is shown in Figure 4–9.

Figure 4–9: Bank Enable/Hit Miss Register

3
1

11
65

87
Bank
Hit/Miss

0
Bank
Enable

ESB90P0020

Table 4–3: Bank Enable/Hit Miss Register (BEHR)

4–10

Data Bit

Name

Description

Type

<15:8>

Bank Hit

These bits are provided for use in testing the
cache. They represent the hit/miss status of each
bank of the cache for the most recent D-stream
read or write cycle. A ’1’ indicates that there was
a hit in the corresponding bank, a ’0’ indicates a
miss. Bank Hit<7:0> are cleared when RESET_L
is asserted.

Read only

KA660 Cache Memory

Table 4–3 (Cont.): Bank Enable/Hit Miss Register (BEHR)
Data Bit

Name

Description

Type

<7:0>

Enable Bank

These bits are written by the power-up diagnostics. The diagnostics are responsible for
determining and enabling good cache banks. The
diagnostics search for a "good" cache bank and
enable it by If this filed contains a 1, this bank
is enabled. When set, these bits enable banks
<7:0> of the cache, respectively. Note that for
any bank to be enabled, CCR Enable Cache or
CCR Diagnostic must be set. When clear, the
corresponding banks are disabled. Enable Bank
<7:0> are cleared when RESET_L is asserted.

Read Only

Whenever a bank is either enabled or disabled,
software should also flush the cache by writing a
one to CCR Flush.
For any reference there should be a hit in
only one bank. If, due to a hardware malfunction (corrupted tag bit), the reference
hits in more than one bank, only the bank
with the higher priority drives the data.
The banks are numbered in decreasing
order of priority. Bank 0 has the highest
priority and bank 5 has the lowest priority. Normally, the corrupted tag would be
reported via the MSER

4.2.6 Memory System Error Register (MSER, IPR 39)
The Memory System Error Register (MSER), Internal Processor Register 39, reports information about DAL bus and cache errors. The two basic classes of errors reported are:
•
•

Errors that don’t immediately affect operation of the SOC CPU and will therefore post an
interrupt but not change instruction flow.
Errors that do affect operation and will cause a machine check (trap through SCB vector
4.)

The format is shown in Figure 4–10.

KA660 Cache Memory 4–11

Figure 4–10: Memory System Error Register (MSER, IPR 39)

3
1

7 6 5 4 3 2 1 0
B C D T T T I
E P P P P P N
R E E E 2 1 T

ZERO

ESB90P0021

Table 4–4: Memory System Error Register (MSER, IPR 39)
Data Bit

Name

<31:7>

Description

Type

Always read as zero

<6>

DPE

DAL Parity Error – This bit is set when a DAL
parity error is detected on either a demand
or request read cycle which receives a normal
termination response (RDY_L asserted, ERR_L
deasserted).

sticky

<5>

MCADPE

Machine Check Abort—DAL parity Error - This
bit is set whenever a machine check is caused by
a DAL parity Error. A DAL parity error will only
cause a machine check on a demand read cycle.

sticky

<4>

MCACPE

Machine Check Abort Cache Parity Error - This
bit is set whenever a machine check is caused
by a cache parity error (tag or data). A cache
parity error will only cause a machine check on a
demand read cycle that hits the cache.

sticky

<3:2>

Always read as zero

<1:0>

These bits are set independandtly to show the
scope of a cache parity error on either a demand
or request cycle, MSER<0> is set to indicate
that the cache parity error was caused by a tag
error; MSER<1> by a data error. Note that a
simultaneuos cache tag and data parity error will
only log the fact that a cache tag parity error
occured.

sticky

NOTE: MSER bits <6:4, 1:0> are "sticky" in the sence that once set, they will remain set until MSER
is explicitly cleared by writing the MSER (with a MTPR instruction) irrespective of the data,
Parity errors occuring while an error condition is posted in MSER can only set an additional bit
(For example MSER <6:4, 1:0> cannot be cleared through subsequent errors.)

4–12

KA660 Cache Memory

4.2.7 Cache Error Detection
Both the tag and data arrays in the cache are protected by parity. Each 8-bit byte of data and
the 19-bit tag is stored with an associated parity bit. The valid bit in the tag is not covered
by parity. Odd data bytes are stored with odd parity and even data bytes are stored with
even parity. The tag is stored with odd parity. The stored parity is valid only when the valid
bit associated with the cache entry is set. Tag and data parity (on the entire longword) are
checked on read references that hit the cache, while only tag parity is checked on CPU and
DMA write references that hit the cache.
The action taken following the detection of a cache parity error depends on the reference type.
The following list outlines the different reference types:
•

During a demand D-stream read reference, the entire cache is flushed, the CCR is cleared
(which disables the cache.) The cause of the error is logged in MSER<4,1:0> and a machine check abort is initiated.

•

During a request I-stream read reference, the entire cache is flushed (unless CCR<0> is
set), the cause of the error is logged in MSER<1:0>, the prefetch is halted, but no machine
check abort occurs, and the cache remains enabled.

•

During a masked or unmasked write reference, the entire cache is flushed (unless CCR<0>
is set), the cause of the error is logged in MSER<0> (only tag parity is checked on CPU
writes that hit the cache) there is no effect on CPU execution, and the cache remains
enabled.

•

During a DMA Write reference the cause of the error is logged in MSER<0> (only tag
parity is checked on DMA writes that hit the cache) there is no effect on CPU execution,
both caches remain enabled, and no invalidate operation occurs.

KA660 Cache Memory

4–13

Chapter 5
KA660 Main Memory System

The KA660-AA includes a main memory controller implemented via a single VLSI chip called
the CMCTL. The KA660 Main Memory Controller communicates with the MS650 memory
boards over the MS650 Memory Interconnect, which uses the CD interconnect for the address
and control lines and a 50-pin, ribbon cable for the data lines. It supports up to 4 (four) MS650
memory boards, for a maximum of 64MB of ECC memory.
NOTE
The KA660 supports only MS650_B* variations and does not support MS650-AA
variations.
The memory controller supports synchronous longword read references, and masked or unmasked synchronous write references generated by the CPU as well as synchronous, quadword
read references generated by cacheable CPU references that miss the cache.

5.0.1 KA660 Timing
The system clock for the VAX 4000-200 operates at 114.285 MHz creating a cycle time of 70
ns. Table 5–1 lists the reference times for CPU reads and writes, and Q22-bus interface reads
and writes. This table also provides information about error handling times.
Table 5–1: KA660 Reference Timing
CPU Read Reference
longword

280 ns

quadword

420 ns

first longword

280 ns

second longword

140 ns

KA660 Main Memory System 5–1

Table 5–1 (Cont.): KA660 Reference Timing
aborted reference

280 ns

longword (locked)

630 ns min

aborted reference

280 ns

retry (locked)

350 ns

CPU Write Reference
longword

140 ns

longword (masked)

350 ns

The controller also supports asynchronous longword and quadword DMA
read references and masked and unmasked asynchronous longword, quadword,
hexword, and octaword DMA write references from the Q22-bus Bus Interface.
Q22-bus Interface Read Reference
longword

420 ns

quadword

560 ns

first longword
second longword
longword (locked)

350 ns
210 ns
420 ns

Q22-bus Interface Write Reference
longword

280 ns

longword (masked)

420 ns

quadword

490 ns

first longword

2800 ns

second longword

210 ns

quadword (masked)

770 ns

first longword

280 ns

second longword

490 ns

hexword
first longword

280 ns

second longword

210 ns

third longword

210 ns

hexword (masked)

980 ns

first longword

280 ns

second longword

210 ns

third longword

4900 ns

octaword

5–2

700 ns

KA660 Main Memory System

910 ns

Table 5–1 (Cont.): KA660 Reference Timing
first longword

280 ns

second longword

210 ns

third longword

210 ns

fourth longword

2100 ns

octaword (masked)

1190 ns

first longword

280 ns

second longword

210 ns

third longword

210 ns

fourth longword

490 ns

The above timing assumes no exception conditions are encountered during the reference. Exception
conditions will add the following amount of time if they are encountered during a reference.
Error Handling
correctable error

70 ns

uncorrectable error

140 ns-read

uncorrectable error

70 ns-write

CDAL parity error

70 ns-write

refresh collision

280 ns

5.0.2 Main Memory Organization
Main memory is logically and physically divided into four boards which correspond to the
four possible MS650 memory expansion modules that can be attached to a KA660-AA. Each
memory board can contain zero (no memory module present) or four (MS650-BA present),
memory banks. Each bank contains 1,048,576 (1M) aligned longwords. Each aligned longword
is divided into four data bytes and is stored with seven ECC check bits, resulting in a memory
array width of 39 bits.

5.0.3 Main Memory Addressing
The KA660-AA Main Memory Controller is capable of controlling up to 16 banks of RAM,
each bank containing 4MB of storage. Each bank of main memory has a programmable base
address, determined by the state of bits <25:22> of the Main Memory Configuration Register
associated with the bank.
A 4MB bank is accessed when bit <29> of the physical address is equal to zero, indicating a
VAX Memory Space read/write reference, bits <28:26> of the physical address are equal to
zero, indicating a reference within the range of the main memory controller, and the bank
number of the bank matches bits <25:22> of the physical address. The remainder of the physical address (bits <21:2>) are used to determine the row and column of the desired longword
within the bank. The Byte Mask lines are ignored on read operations, but are used to select
the proper byte(s) within a longword during masked longword write references.

KA660 Main Memory System 5–3

5.1 Main Memory Behavior on Writes
On unmasked CPU write references, the main memory controller operates in "dump and run"
mode, terminating the CDAL Bus transaction after latching the data, but before checking
CDAL Bus parity, calculating the ECC check bits, and transferring the data to main memory.
On unmasked DMA write references by the Q22-bus Bus Interface, the data is latched, CDAL
Bus parity is NOT checked, the CDAL Bus transaction is terminated, the ECC check bits are
calculated, and the data is transferred to main memory.
On single masked CPU or DMA write references, CDAL Bus parity is checked (for CPU writes
only), the referenced longword is read from main memory, the ECC code checked, the check
bits recalculated to account for the new data byte(s), the CDAL transaction is terminated, and
the longword is rewritten.
On multiple transfer masked DMA writes, each longword write is acknowledged then the
CDAL transaction is terminated.

5.2 Main Memory Error Status Register (MEMCSR16)
The Main Memory Error Status Register, address 2008 014016 , is used to capture main memory error data. This register is unique to CPU designs that use the CMCTL memory controller
chip.

5–4

Data Bit

Name

Description

Type

<31>

RDS ERROR

When set, an uncorrectable ECC error occurred
during a memory read or masked write reference.
Cleared by writing a one to it. Writing a zero has
no effect. Undefined if MEMCSR16<7> (CDAL
BUS ERROR) is set. Cleared on power-up and
the negation of DCOK when SCR<7> is clear.

Read/Write to
clear

<30>

RDS HIGH
ERROR RATE

When set, an uncorrectable ECC error occurred
while the RDS ERROR LOG REQUEST bit was
set, indicating multiple uncorrectable memory
errors. Cleared by writing a one to it. Writing a
zero has no effect. Undefined if MEMCSR16<7>
(CDAL BUS ERROR) is set. Cleared on power-up
and the negation of DCOK when SCR<7> is clear.

Read/Write to
Clear.

<29>

CRD ERROR

When set, a correctable (single bit) error occurred
during a memory read or masked write reference.
Cleared by writing a one to it. Writing a zero has
no effect. Undefined if MEMCSR16<7> (CDAL
BUS ERROR) is set. Cleared by writing one,
on power-up and the negation of DCOK when
SCR<7> is clear.

Read/Write to
Clear.

KA660 Main Memory System

Data Bit

Name

Description

Type

<28:9>

PAGE
ADDRESS
OF ERROR

This field identifies the page (512 byte block)
containing the location that caused the memory
error. In the event of multiple memory errors,
the types of errors are prioritized and the page
address of the error with the highest priority
is captured. Errors with equal priority do not
overwrite previous contents. Writes have no
effect. Cleared on power-up and the negation of
DCOK when SCR<7> is clear. The types of error
conditions follow in order of priority:

Read Only.

•

CDAL Bus parity errors during a CPU write
reference, as logged by the CDAL BUS
ERROR bit.
Uncorrectable ECC errors during a CPU or
DMA read or masked write reference, as
logged by the RDS ERROR LOG bit.
Correctable ECC errors during a CPU or
DMA read or masked write reference, as
logged by CRD ERROR bit.

<8>

DMA ERROR

When set, an error occured during a DMA read
or write reference. Cleared by writing a one to it.
Writing a zero has no effect. Cleared on power-up
and the negation of DCOK when SCR<7> is clear.

Read/Write to
Clear.

<7>

CDAL BUS
ERROR

When set, a CDAL Bus parity error occurred
on a CPU write reference. Cleared by writing a
one to it. Writing a zero has no effect. Cleared
on power-up and the negation of DCOK when
SCR<7> is clear.

Read/Write to
Clear.

<6:0>

ERROR
SYNDROME

This field stores the error syndrome. A nonzero syndrome indicates a detectable error has
occured. A unique syndrome is generated for
each possible single bit (correctable) error. A
list of the these syndromes and their associated
single bit errors appears on the next page. Any
non-zero syndrome that is not contained on this
list indicates a multiple bit (uncorrectable) error
has occured. This field handles multiple errors in
the same manner as MEMCSR16<28:9>. Cleared
on power-up and the negation of DCOK when
SCR<7> is clear.

Read Only.

The following is a list of the syndromes and their associated single bit errors:

KA660 Main Memory System 5–5

SYNDROME<6:0>
=============
0000000

BIT POSITION IN ERROR
=====================
no error detected

1011000
0011100
0011010
1011110
0011111
1011011
1011101
0011001
1101000
0101100
0101010
1101110
0101111
1101011
1101101
0101001
1110000
0110100
0110010
1110110
0110111
1110011
1110101
0110001
0111000
1111100
1111010
0111110
1111111
0111011
0111101
1111001

0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

0000001
0000010
0000100
0001000
0010000
0100000
1000000

32
33
34
35
36
37
38

Data Bits (0-31 decimal)
|
|
V

Check Bits (32-38 decimal)
|
|
V

0000111

Result of incorrect check
bits written on detection of
a CDAL parity error.

All others

Multi-bit errors

5.3 Main Memory Control and Diagnostic Status Register (MEMCSR17)
The Main Memory Control and Diagnostic Status Register, address 2008 0144 16 , is used to
control the operating mode of the main memory controller as well as to store diagnostic status
information. This register is unique to CPU designs that use the CMCTL memory controller
chip. Figure 5–1 shows the format of this register.

5–6

KA660 Main Memory System

Figure 5–1: Main Memory Control and Diagnostic Status Register (MEMCSR17)

3
1

1 1 1 1 1 1
5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0
MBZ

ENABLE LOCK BIT
MAIN MEMORY CYCLE SELECT
CRD INTERRUPT ENABLE
FORCE REFRESH REQUEST
ERROR DETECT DISABLE
FAST DIAGNOSTIC TEST
CLEAR LOCK BIT
DIAGNOSTIC CHECK MODE
CHECK BITS
ESB90P0023

Data Bit

Name

Description

Type

<31:15>

Unused

This field reads as 0, must be written as 0.

<14>

ENABLE LOCK
BIT

When cleared, the main memory locking function
(reflected by the LOCK BIT in MEMCSR150<6>) is disabled. When set, the main memory
locking function (reflected by the LOCK BIT
in MEMCSR15-0<6>) is enabled. Writing this
bit has no effect on MEMCSR15-0<6>. This bit
should always be clear, because the KA660-AA
implements the main memory locking function
in the Q22-bus Interface chip (CQBIC). Cleared
on power-up and the negation of DCOK when
SCR<7> is clear.

Read/Write.

KA660 Main Memory System 5–7

Data Bit

Name

Description

Type

<13>

MAIN
MEMORY
CYCLE
SELECT

When set, longword reads and the first longword
in quadword reads occur in four CPU cycles
(320ns) and the second longword in a quadword
read occurs in two CPU cycles (160ns). When
cleared, longword reads and the first longword in
quadword reads occur in five CPU cycles (450ns)
and the second longword in a quadword read
occurs in three CPU cycles (240ns).

Read/Write

NOTE
With the KA660-AA this
bit must be cleared by the
firmware memory configuration routine thus using
the 5/3 timing. Cleared on
power-up and the negation of
DCOK when SCR<7> is clear.

5–8

<12>

CRD INTERRUPT
ENABLE

When cleared, single-bit errors are corrected by
the ECC logic, but no interrupt is generated.
When set, single-bit errors are corrected by the
ECC logic and they cause an interrupt to be
generated at IPL 1A with a vector of 5416 , .
This bit has no effect on the capturing of error
information in MEMCSR16, or on the reporting
of uncorrectable errors. Cleared on power-up and
the negation of DCOK when SCR<7> is clear.

Read/Write.

<11>

FORCE
REFRESH
REQUEST

When cleared, the refresh control logic operates
in normal mode (refresh every 9.1us for 80ns
cycles and 1 wait state). When set, one memory refresh operation occurs immediately after
the MEMCSR write reference that set this bit.
Setting this bit provides a mechanism for speeding up the testing of the refresh logic during
manufacturing test of the controller chip. This
bit is cleared by the memory controller upon
completion of the refresh operation. Cleared
on power-up and the negation of DCOK when
SCR<7> is clear.

Read/Write.

<10>

MEMORY
ERROR
DETECT
DISABLE

When set, error detection and correction (ECC)
is disabled, so all memory errors go undetected.
When cleared, error detection, correction, state
capture and reporting (via MEMCSR16) is enabled. Cleared on power-up and the negation of
DCOK when SCR<7> is clear.

Read/Write.

KA660 Main Memory System

Data Bit

Name

Description

Type

<9>

FAST DIAGNOSTIC This bit provides a mechanism for speeding up
TEST
the diagnostic testing of main memory. When set,
all main memory banks are read and written in
parallel. When cleared, this bit has no effect on
memory reads or writes. Cleared on power-up
and the negation of DCOK when SCR<7> is clear.

Read/Write.

NOTE
Due to excess power consumption do not use MOVC
instructions in fast diagnostic
test mode.
<8>

CLEAR LOCK
BIT

Writing a one to this bit clears MEMCSR15-0<6>
and "unlocks" main memory. Always read as
zero. This bit is used to unlock memory that
was locked by an interlocked instruction that
was aborted, due to an error condition, before it
could unlock memory. This bit should never be
needed, because the KA660-AA implements the
main memory locking function in the Q22-bus
Interface chip (CQBIC). Cleared on power-up and
the negation of DCOK when SCR<7> is clear.

Write Only.

<7>

DIAGNOSTIC
CHECK MODE

When set, the contents of MEMCSR17<6:0> are
written into the 7 ECC check bits of the location
(even if a CDAL parity error is detected) during
a memory write reference. When cleared, the
7 check bits calculated by the ECC generation
logic are loaded into the 7 ECC check bits of the
location during a write reference and a memory
read reference will load the state of the 7 ECC
check bits of the location that was read into
MEMCSR17<6:0>. Cleared on power-up and the
negation of DCOK when SCR<7> is clear.

Read/Write.

NOTE
DIAGNOSTIC CHECK
MODE is restricted to unmasked memory write references. No masked write
references are allowed when
DIAGNOSTIC CHECK
MODE is enabled.

KA660 Main Memory System 5–9

Data Bit

Name

Description

Type

<6:0>

CHECK BITS

When the DIAGNOSTIC CHECK MODE bit
is set, these bits are substituted for the check
bits that are generated by the ECC generation logic during a write reference. When the
DIAGNOSTIC CHECK MODE bit is cleared,
memory read references load the state of the 7
ECC check bits of the location that was read into
MEMCSR16<6:0>. Cleared on power-up and the
negation of DCOK when SCR<7> is clear.

Read/Write.

5.4 Main Memory Error Detection and Correction
The KA660-AA Main Memory Controller generates CDAL Bus parity on CPU read references,
and checks CDAL Bus parity on CPU write references.
The actions taken following the detection of a CDAL Bus parity error depend on the type of
write reference.
For unmasked CPU write references, incorrect check bits are written to main memory (potentially masking an as yet undetected memory error) along with the data and an interrupt
is generated at IPL 1D through vector 6016 , on the next cycle and MCSR16<7> is set. The
incorrect check bits are determined by calculating the seven "correct" check bits, and complementing the three least significant bits.
For masked CPU write references, incorrect check bits are written to main memory (potentially masking an as yet undetected memory error) along with the data, unless an uncorrectable error is detected during the read portion, MEMCSR16<7> is set, and a machine check
abort is initiated. If an uncorrectable error is detected on the read portion, no write operation
takes place. The incorrect check bits are determined by calculating the seven "correct" check
bits, and complementing the three least significant bits.
The memory controller protects main memory by using a 32-bit Modified Hamming code to
"encode" the 32-bit data longword with seven check Bits. This allows the controller to detect
and correct single-bit errors in the data field and detect single bit errors in the check bit field
and double-bit errors in the data field. The most likely causes of these errors are failures in
either the memory array or the 50-pin ribbon cable.
Upon detecting a correctable error on a read reference or the read portion of a masked write
reference, the data is corrected (if it is in the data field), before placing it on the CDAL Bus,
or back in main memory, an interrupt is generated at IPL 1A through vector 5416 , , bit <29>
of MEMCSR16 is set, bits <28:9> of MEMCSR16 are loaded with the address of the page
containing the location that caused the error, and bits <6:0> are loaded with the error syndrome which indicates which bit was in error. If the error was detected on a DMA reference,
MEMCSR16<8> is also set.
NOTE
The corrected data is not rewritten to main memory, so the single bit error will
remain there until rewritten by software.

5–10

KA660 Main Memory System

Upon detecting an uncorrectable error, the action depends on the type of reference being
performed. Table 5–2 lists the actions performed on uncorrectable errors.
Table 5–2: Uncorrectable Error Actions
Error

Action Performed

On a demand read reference
The affected row of the cache is invalidated.
Bit <31> of MEMCSR16 is set.
bits <28:9> of MEMCSR16 are loaded with the address of the
page containing the location that caused the error,
bits <6:0> are loaded with the error syndrome which indicates
that the error was uncorrectable.
A machine check abort is initiated.
If the read was a local-miss, global-hit read, or a read of the
Q22-bus Bus map, MEMCSR16<8> and DSER<4> are also
set, and DEAR<12:0> are loaded with the address of the page
containing the location that caused the error.
On a request read reference
The prefetch or fill cycle is aborted, but no machine check
occurs.
Bit <31> of MEMCSR16 is set.
Bits <28:9> of MEMCSR16 are loaded with the address of the
page containing the location that caused the error.
Bits <6:0> are loaded with the error syndrome which indicates
that the error was uncorrectable.
On the read portion of masked
write reference

Bit <31> of MEMCSR16 is set.
Bits <28:9> of MEMCSR16 are loaded with the address of the
page containing the location that caused the error.
Bits <6:0> are loaded with the error syndrome which indicates
that the error was uncorrectable.
A machine check abort is initiated.

KA660 Main Memory System 5–11

Table 5–2 (Cont.): Uncorrectable Error Actions
Error

Action Performed

On a DMA read reference
Bit <31> and bit <8> of MEMCSR16 are set, bits <28:9> of
MEMCSR16 are loaded with the address of the page containing
the location that caused the error.
Bits <6:0> are loaded with the error syndrome which indicates
that the error was uncorrectable,
DSER<4> is set,
DEAR<12:0> are loaded with the address of the page containing the location that caused the error.
BDAL<17:16> are asserted on the Q22-bus Bus along with the
data to notify the receiving device (unless it was a map read by
the Q22-bus bus interface during translation).
An interrupt is generated at IPL 1D through vector 6016 .
On a DMA masked write reference

5–12

KA660 Main Memory System

Bit<31> and bit <8> of MEMCSR16 are set.
Bits <28:9> of MEMCSR16 are loaded with the address of the
page containing the location that caused the error.
Bits <6:0> are loaded with the error syndrome which indicates
that the error was uncorrectable
DSER<4> is set. DEAR<12:0> are loaded with the address of
the page
containing the location that caused the error.
ICR<15> is set to notify the initiating device
An interrupt is generated at IPL 1D through vector 6016 , .

Chapter 6
KA660 Console Serial Line

The console serial line provides the KA660 processor with a full duplex, RS-423 EIA, serial
line interface, which is also RS-232C compatible. The only data format supported is 8-bit data
with no parity and one stop bit. The four Internal Processor Registers (IPR’s) that control the
operation of the console serial line are a superset of the VAX Console Serial Line Registers
described in the VAX Architecture Reference manual .

6.1 Console Registers
There are four registers associated with the Console Serial Line Unit. They are implemented
in the SSC chip and are accessed as IPR’s 32-35. Refer to Table 6–1.
Table 6–1: Console Registers
IPR Number

Mnemonic
RXCS

Dec

Hex

Console Receiver Control/Status

Console Receiver Data Buffer

RXDB

Console Transmit Control/Status

TXCS

Console Transmit Data Buffer

TXDB

6.1.1 Console Receiver Control/Status Register - (IPR 32)
The Console Receiver Control/Status Register (RXCS), Internal Processor Register 32, is used
to control and report the status of incoming data on the console serial line. The format is
shown in Figure 6–1. Table 6–2 lists the bit descriptions.

KA660 Console Serial Line 6–1

Figure 6–1: Console Receiver Control/Status Register - (IPR 3210 2016 )

3
1

8765
MBZ

0
MBZ

RX DONE
RX IE
ESB90P0024

Table 6–2: Console Receiver Control/Status Register

6–2

Data Bit

Name

Description

<31:8>

MBZ

These bits Read as ZEROs, Writes have no
effect.

<7>

RX DONE

Receiver Done. Read Only. Writes have no
effect. This bit is set when an entire character
has been received and is ready to be read from
the RXDB Register. This bit is automatically
cleared when the RXDB register is read. It is
also cleared on power-up and the negation of
DCOK.

<6>

RX IE

Receiver Interrupt Enable. Read/Write. When
set, this bit causes an interrupt to be requested at IPL 14 with an SCB offset of F8
If RX DONE is set. When cleared, interrupts
from the Console Receiver are disabled. This
bit is cleared on power-up and the negation of
DCOK when CSR<7> is clear .

<5:0>

Unused

These bits read as ZEROs. Writes have no
effect.

KA660 Console Serial Line

6.1.2 Console Receiver Data Buffer - (IPR 33)
The Console Receiver Data Buffer (RXDB), internal processor register 33, is used to buffer
incoming data on the serial line and capture error information. The format is shown in
Figure 6–2. Bit descriptions are listed in Table 6–3.
Figure 6–2: Console Receiver Data Buffer - (IPR 3310 2116 )

3
1

1 1 1 1 1 1 1
6 5 4 3 2 1 0
MBZ

8 7

MBZ

ERR
OVR ERR
FRM ERR
MBZ
RCV BRK
RECEIVED DATA BITS
ESB90P0025

Table 6–3: Console Receiver Data Buffer
Data Bit

Name

Description

<31:16>

MBZ

These bits always read as ZERO. Writes have
no effect.

<15>

ERR

Error. Read only. Writes have no effect. This
bit is set if RBUF <14> or <13> is set. It
is clear if these two bits are clear. This bit
cannot generate a program interrupt. Cleared
on power-up and the negation of DCOK.

<14>

OVR ERR

Overrun Error. Read only. Writes have no
effect. This bit is set if a previously received
character was not read before being overwritten by the present character. Cleared
by reading the RXDB, on power-up and the
negation of DCOK.

KA660 Console Serial Line 6–3

Table 6–3 (Cont.): Console Receiver Data Buffer
Data Bit

Name

Description

<13>

FRM ERR

Framing Error. Read only. Writes have no
effect. This bit is set if the present character
did not have a valid stop bit. Cleared by reading the RXDB, on power-up and the negation
of DCOK. Error conditions are updated
when the character is received and it remains present until the character is read,
at which point, the error bits are cleared.

<12>

MBZ

This bit always reads as ZERO. Writes have
no effect.

<11>

RCV BRK

Received Break. Read only. Writes have no
effect. This bit is set at the end of a received
character for which the serial data input
remained in the SPACE condition for 20 bit
times. Cleared by reading the RXDB, on
power-up and the negation of DCOK.

<10:8>

MBZ

These bits always read a as ZERO. Writes
have no effect.

<7:0>

Received Data Bits.

Read only. Writes have no effect. These bits
contain the last received character.

6.1.3 Console Transmitter Control/Status Register - (IPR 34)
The Console Transmitter Control/Status Register (TXCS), Internal Processor Register 34, is
used to control and report the status of outgoing data on the console serial line. The format is
shown in Figure 6–3. Bit descriptions are listed in Table 6–4.

6–4

KA660 Console Serial Line

Figure 6–3: Console Transmitter Control/Status Register - (IPR 3410 2216 )

3
1

8 7 6 5
MBZ

3 2 1 0

MBZ

TX RDY
TX IE
MAINT
MBZ
XMIT BRK
ESB90P0026

KA660 Console Serial Line 6–5

Table 6–4: Console Transmitter Control/Status Buffer
Data Bit

Name

Description

<31:8>

MBZ

These bits read as ZEROs. Writes have no
effect.

<7>

TX RDY

Transmitter Ready. Read only. Writes have
no effect. This bit is cleared when TXDB
is loaded and set when TXDB can receive
another character. This bit is set on power-up
and the negation of DCOK.

<6>

TX IE

Transmitter Interrupt Enable. Read/Write.
When set, this bit causes an interrupt to be
requested at IPL 14 with an SCB offset of FC
if TX RDY is set. When cleared, interrupts
from the Console Receiver are disabled. This
bit is cleared on power-up and the negation of
DCOK.

<5:3>

MBZ

Read as zeros. Writes have no effect.

<2>

MAINT

Maintenance.Read/Write. This bit is used
to facilitate a maintenance self-test. When
MAINT is set, the external serial output is set
to MARK and the serial output is used as the
serial input. This bit is cleared on power-up
and the negation of DCOK.

<1>

Unused

This bit read as ZERO. Writes have no effect.

<0>

XMIT BRK

Transmit Break. ead/Write. When this bit is
set, the serial output is forced to the SPACE
condition after the character in TXDB<7:0> is
sent. While XMIT BRK is set, the transmitter
will operate normally, but the output line
will remain low. Thus, software can transmit
dummy characters to time the break. This bit
is cleared on power-up.

6.1.4 Console Transmitter Data Buffer - (IPR 35)
The Console Transmitter Data Buffer (TXDB), Internal Processor Register 35, is used to buffer
outgoing data on the serial line. The format is shown in Figure 6–4. Table 6–5 lists the bit
descriptions.

6–6

KA660 Console Serial Line

Figure 6–4: Console Transmitter Data Buffer - (IPR 3510 2316 )
3
1

MBZ
Transmitted Data Bits
ESB90P0027

Table 6–5: Console Transmitter Data Buffer
Data Bit

Name

Description

<31:8>

MBZ

Read as 0. Writes have no effect.

<7:0>

Transmitted Data Bits

Write only. These bits are used to load the
character to be transmitted on the console
serial line.

6.2 Break Response
The console serial line unit recognizes a BREAK condition which consists of 20 consecutively
received SPACE bits. If the console detects a valid break condition, the RCV BRK bit is set
in the RXDB register. If the break was the result of 20 consecutively received SPACE bits,
the FRM ERR bit is also set. If halts are enabled the KA660will halt and transfer program
control to EPROM location 2004 000016 when the RCV BRK bit is set. RCV BRK is cleared by
reading RXDB. Another MARK followed by 20 consecutive SPACE bits must be received to set
RCV BRK again.

6.3 Baud rate
The receive and transmit baud rates are always identical and are controlled by the SSC
Configuration Register bits <14:12>.
The user selects the desired baud rate through the Baud Rate Select signals which are received from an external 8-position switch mounted on the console module (H3602) . The
KA660firmware reads this code from Boot and Diagnostic Register bits <6:4> and compliments
and then loads it into SSC Configuration Register bits <14:12>.
Table 6–6 shows the Baud Rate Selection, the corresponding code as read in the Boot and
Diagnostic Register bits <6:4>, and the INVERTED code that should be loaded into SSC
Configuration Register bits <14:12>:

KA660 Console Serial Line 6–7

Table 6–6: Baud Rate Selection
Baud Rate

BDR<6:4>

SSC<14:12>

300

111

000

600

110

001

1200

101

010

2400

100

011

4800

011

100

9600

010

101

19200

001

110

38400

000

111

6.4 Console Interrupt Specifications
The console serial line receiver and transmitter both generate interrupts at IPL 14. The
receiver interrupts with a vector of F816 , while the transmitter interrupts with a vector of
FC16 .

6–8

KA660 Console Serial Line

Chapter 7
KA660 Clock and Timer Registers

The KA660 clocks include Time Of Year Clock (TOY), a subset Interval Clock (subset ICCS),
as defined in the VAX Architecture Reference Manual and two additional programmable timers
modeled after the VAX Standard Interval Clock.

7.1 Time-of-Year Clock (TOY) - EPR 27
The KA660 Time-of-Year clock (TOY) forms an unsigned 32-bit binary counter that is driven
from a 100Hz oscillator, so that the least significant bit of the clock represents a resolution
of 10 milliseconds, with less than .0025% error. The register counts only when it contains
a non-zero value. This register is implemented in the SSC chip. The format is shown in
Figure 7–1.

Figure 7–1: Time-of-Year Clock (TOY) - (EPR 2710 1B16 )

3
1

0
Time of Year Since Setting
ESB90P0028

The Time Of Year Clock is maintained during power failure by battery backup circuitry which
interfaces, via the external connector, to a set of batteries which are mounted on the CPU
Console Module. The (TOY) will remain valid for greater than 162 hours when using the
NiCad battery pack (3 batteries in series) mounted on the H3602 cover panel .
The SSC Configuration Register contains a Battery Low (BLO) bit which, if set after initialization, the TOY is cleared, and will remain at zero until software writes a non-zero value into
it.

KA660 Clock and Timer Registers 7–1

NOTE
After writing a non-zero value into the TOY, software should clear the BLO bit by
writing a one to it.

7.2 Interval Timer (ICCS) - EPR 24
The KA660 Interval Timer (ICCS), Internal Processor Register 24, is implemented according to
the VAX Architecture Reference Manual. The Interval Clock Control/Status Register (ICCS), is
implemented as the standard subset of the Standard VAX ICCS in the CPU chip; while NICR
and ICR are not implemented. Figure 7–2 shows the format and Table 7–1 describes the bits.
Figure 7–2: Interval Timer (ICCS) - (EPR 2410 1816 )
3
1

765
MBZ

0
MBZ
IE
ESB90P0029

Table 7–1: Interval Timer Bit Descriptions
Data
Bit

Name

Description

<31:7>

MBZ

Read as zeros, must be written as zeros.

<6>

Interrupt Enable. Read/Write. This bit enables and disables the interval timer
interrupts. When the bit is set, an interval timer interrupt is requested every
10 msec with an error of less than .01%. When the bit is clear, interval timer
interrupts are disabled. This bit is cleared on power-up.

<5:0>

MBZ

Read as zeros, must be written as zeros.

Interval timer requests are posted at IPL 16 with a vector of C0: the interval timer is the highest
priority device at this IPL.

7.3 Programmable Timers
The KA660 features two programmable timers. Although they are modeled after the VAX
Standard Interval Clock, they are accessed as I/O space registers (rather than as Internal
Processor Registers) and a control bit has been added which stops the timer upon overflow.
If so enabled, the timers will interrupt at IPL 14 upon overflow. The interrupt vectors are
programmable and are set to 78 and 7C by the firmware.

7–2

KA660 Clock and Timer Registers

Each timer is composed of four registers:
a Timer n Control Register,
a Timer n Interval Register,
a Timer n Next Interval Register, and
a Timer n Interrupt Vector Register,
where n represents the timer number (0 or 1).

7.3.1 Timer Control Registers (TCR0 and TCR1)
The KA660 has two Timer Control Registers, one for controlling timer 0 (TCR0), and one for
controlling timer 1 (TCR1). TCR0 is accessible at address 2014 010016 , and TCR1 is accessible
at 2014 011016 . These registers are implemented in the SSC chip. Figure 7–3 shows the
format. Table 7–2 lists the bit descriptions.
Figure 7–3: Timer Control Registers (TCR0 and TCR1)
3 3
1 0

8 7 6 5 4 3 2 1 0
MBZ
ERR

RUN
MBZ
STP
MBZ
XFR
SGL
IE
INT
ESB90P0030

Table 7–2: Timer Control Register Bit Descriptions
Date Bit

Name

Description

<31>

ERR

Error. Read/Write to Clear. This bit is set whenever the Timer
Interval Register overflows and the INT bit is already set. Thus, the
ERR bit indicates a missed overflow. Writing a one to this bit clears
it. Cleared on power-up.

<30:8>

MBZ

Read as zeros, must be written as zeros.

<7>

INT

Read/Write to Clear. This bit is set whenever the Timer Interval
Register overflows. If IE is set when INT is set, an interrupt is
posted at IPL 14. Writing a one to this bit clears it. Cleared on
power-up.

KA660 Clock and Timer Registers 7–3

Table 7–2 (Cont.): Timer Control Register Bit Descriptions
Date Bit

Name

Description

<6>

Read/Write. When this bit is set, the timer will interrupt at IPL 14
when the INT bit is set. Cleared on power-up.

<5>

SGL

Read/Write. Setting this bit causes the Timer Interval Register to
be incremented by one if the RUN bit is cleared. If the RUN bit is
set, then writes to the SGL bit are ignored. This bit is always read
as zero.Cleared on power-up.

<4>

XFR

Read/Write. Setting this bit causes the Timer Next Interval Register
to be copied into the Timer Interval Register. This bit is always read
as zero. Cleared on power-up.

<3>

MBZ

Read as zeros, must be written as zeros.

<2>

STP

Read/Write. This bit determines whether the timer stops after an
overflow when the RUN bit is set. If the STP bit is set at overflow,
the RUN bit is cleared by the hardware at overflow and counting
stops. Cleared on power-up.

<1>

MBZ

Read as zeros, must be written as zeros.

<0>

RUN

Read/Write. When set, the Timer Interval Register is incremented
once every microsecond. The INT bit is set when the timer overflows. If the STP bit is set at overflow, the RUN bit is cleared by
the hardware at overflow and counting stops. When the RUN bit is
clear, the Timer Interval Register is not incremented automatically.
Cleared on power-up.

7.3.2 Timer Interval Registers (TIR0 and TIR1)
The KA660 has two Timer Interval Registers, one for timer 0 (TIR0), and one for timer 1
(TIR1). TIR0 is accessible at address 2014 010416 , and TIR1 is accessible at 2014 011416 .
The Timer Interval Register is a read only register containing the interval count. When the
RUN bit is 0, writing a 1 increments the register. When the RUN bit is 1, the register is
incremented once every microsecond. When the counter overflows, the INT bit is set, and an
interrupt is posted at IPL14 if the IE bit is set. Then, if the RUN and STP bits are both set,
the RUN bit is cleared and counting stops. Otherwise, the counter is reloaded. The maximum
delay that can be specified is approximately 1.2 hours. This register is cleared on power-up.
Figure 7–4 shows the format.

7–4

KA660 Clock and Timer Registers

Figure 7–4: Timer Interval Registers (TIR0 and TIR1)

3
1

0
Timer Interval Register
ESB90P0031

7.3.3 Timer Next Interval Registers (TNIR0 and TNIR1)
The KA660 has two Timer Next Interval Registers, one for timer zero (TNIR0), and one for
timer one (TNIR1). TNIR0 is accessible at address 2014 010816 , and TNIR1 is accessible
at 2014 011816 . These registers are implemented in the SSC chip. The format is shown in
Figure 7–5.
This Read/Write register contains the value which is written into the Timer Interval Register
after overflow, or in response to a one written to the XFR bit. This register is cleared on
power-up.

Figure 7–5: Timer Next Interval Registers (TNIR0 and TNIR1)

3
1

0
Timer Next Interval Register
ESB90P0032

7.3.4 Timer Interrupt Vector Registers (TIVR0 and TIVR1)
The KA660 has two Timer Interrupt Vector Registers, one for timer zero (TIVR0), and one for
timer one (TIVR1). TIVR0 is accessible at address 2014 010C16 , and TIVR1 is accessible at
2014 011C16 . These registers are implemented in the SSC chip and are set to 7816 and 7C16
respectively by the resident firmware. The format is shown in Figure 7–6.

KA660 Clock and Timer Registers 7–5

This Read/Write register contains the timer’s interrupt vector. Bits <31:10> and <1:0> are
read as zero and must be written as zero. When TCRn<6> (IE) and TCRn<7> (INT) transition
to one, an interrupt is posted at IPL 14. When a timer’s interrupt is acknowledged, the
content of the interrupt vector register is passed to the CPU, and the INT bit is cleared.
Interrupt requests can also be cleared by clearing either the IE or INT bit. This register is
cleared on power-up.

Figure 7–6: Timer Interrupt Vector Registers (TIVR0 and TIVR1)

3
1

10 9
MBZ

2 1 0
Interrupt Vector

MBZ
ESB90P0033

NOTE
Note that both timers interrupt at the same IPL (IPL 14) as the console serial line
unit. When multiple interrupts are pending, the console serial line has priority over
the timers, and timer 0 has priority over timer 1.

7–6

KA660 Clock and Timer Registers

Chapter 8
KA660 Boot and Diagnostic Facility

The KA660 Boot and Diagnostic Facility features two registers, 256K bytes of Erasable
Programmable Read Only Memory (EPROM) and 1KB of battery backed up RAM. The EPROM
and battery backed up RAM may be accessed with longword, word or byte references.
The 256K bytes of EPROM contain the resident firmware. If this EPROM is reprogrammed
for special applications, the new code must initialize and configure the board, provide HALT
and console emulation, as well as provide boot diagnostic functionality.

KA660 Boot and Diagnostic Facility 8–1

8.1 Boot and Diagnostic Register (BDR)
The Boot and Diagnostic Register (BDR) is a byte-wide register repeated in the VAX I/O page
at physical addresses 2008 4004 - 2008 407C16 . It is implemented uniquely on the KA660. It
can be accessed by KA660 software, but not by external Q22-Bus devices. The BDR allows the
Boot and Diagnostic EPROM programs to read various KA660 configuration bits.

Figure 8–1: Boot and Diagnostic Register

3
1

2 2 2
4 3 2

2 1 1 1 1 1
0 9 8 7 6 5

1 1 1
3 2 1

NI ROM

0
Read as ONES
RBE
DSSI ID<2:0>
BDG CD<1:0>
CPU CD<1:0>
BRS CD<2:0>
HLT ENB
ESB90P0034

8–2

Data Bit

Name

Description

Type

<31:24>

NI ROM

NI Station Address. This byte delivers the NI
Station address in the next 32 longwords.

Read only

<23>

HLT ENB

Halt Enable. This bit reflects the state of pin 35
(ENBHALT L) of the 40-pin connector. The assertion of this signal enables the halting of the CPU
upon detection of a console BREAK condition. On
a power-up, the KA660 resident firmware reads
the HLT ENB bit to decide whether to enter the
console emulation program (HLT ENB set) or to
boot the operating system (HLT ENB clear). On
the execution of of a HALT instruction while in
kernal mode, the KA660 resident firmware reads
the HLT ENB bit to decide whether to enter the
console emulation program (HLT ENB set) or to
restart the operating system (HLT ENB clear).

Read
Only.Writes
have no effect.

KA660 Boot and Diagnostic Facility

Data Bit

Name

Description

Type

<22:20>

BRS CD

Baud Rate Select. Writes have no effect.
These three bits originate from pins <20:28>
(BRS<2:0>) of the 40-pin connector. They reflect
the setting of the the baud rate select switch on
the CPU cover panel.

Read only

<19:18>

<17:16>

CPU CD

BDG CD

BDR<6:4>

Baud Rate

000

300

001

600

010

1200

011

2400

100

4800

101

9600

110

19200

111

38400

CPU Code. Writes have no effect. These two
bits always read as ZERO’s because the KA660
cannot be configured as an auxiliary CPU.
CPU CD <1:0>

Configuration

KA660-AA Arbiter

Boot and Diagnostic Code. This 2-bit code reflects the status of configuration and display
connector pins <37:36> (BDG CD<1:0>). The
KA660 EPROM programs use BDG CD <1:0> to
determine the power up mode as follows:
BDR<1:0>

Power-up Mode

Run

Language Inquiry

Test

Manufacturing

<15:13>

BDG DSSIID

DSSI ID. Writes have no effect. This 3-bit code
reflects the setting of the DSSI ID jumpers. This
code must be decoded and written to the SHAC
PPR register.

<12>

RBE

Remote Boot Enable.

<11:0>

Unused

Read as one’s

Read Only

Read only,
writes have
no effect.

read only

KA660 Boot and Diagnostic Facility 8–3

8.2 Diagnostic LED Register (DLEDR)
The Diagnostic LED Register (DLEDR), address 2014 002016 , is implemented in the SSC
chip and contains four Read/Write bits that control the external LED display. A zero in a bit
lights the corresponding LED; all four bits are cleared on power-up and the negation of DCOK
when SCR<7> is clear to provide a power-up lamp test. Figure 8–2 shows the register format.
Table 8–3 lists the bit descriptions.

Figure 8–2: Diagnostic LED Register (DLEDR)

3
1

4 3
MBZ

DSPL
ESB90P0035

Table 8–3: Diagnostic LED Register Bit Descriptions

8–4

Data Bit

Name

Description

<31:4>

MBZ

Read as zeros, must be written as zeros.

<3:0>

DSPL

Display. Read/Write. These four bits update an external LED display. Writing
a ZERO to a bit lights the corresponding LED. Writing a ONE to a bit turns
its LED off. The display bits are cleared (all LED’s are lit) on power-up and
the negation of DCOK.

KA660 Boot and Diagnostic Facility

8.3 EPROM Memory
The KA660 has 256KB of EPROM memory for storing code for board initialization, VAX
standard console emulation, board self tests, and boot code. The EPROM memory may be
accessed with byte, word and longword references. EPROM read accesses take 250 ns. The
EPROM is organized as a 128K x 8-bit array. CDAL Bus parity is neither checked nor generated on EPROM references. It should be noted that the EPROM size must be set in the SSC
Configuration Register before attempting to reference outside the first 8KB block of the Local
EPROM Space. (2004 0000 - 2004 1FFF16 )

8.3.1 EPROM Address Space
The entire 256KB Boot and Diagnostic EPROM can only be read in the 256KB Halt Protect
EPROM Space (2004 0000 - 2007 FFFF16 ).
Any I-Stream Read from the EPROM space places the KA660 in Halt Mode. The Q22-bus
SRUN signal is deasserted causing the front panel RUN light to extinguish and the CPU is
protected from further halts.
Writes and D-Stream Reads to any address space have no effect on Run Mode/Halt Mode
status. The KA660 logic that controls Halt Mode/Run Mode cannot detect I-stream read
references that "hit" the Cache; therefore Halt Mode/Run Mode is unaffected by these cache
hits.

8.3.2 KA660 Resident Firmware Operation
The KA660 CPU Module’s 256K bytes of EPROM contain the resident firmware, which can be
entered by transferring program control to location 2004 000016 .
Chapter 12 lists the various halt conditions which cause the KA660 to transfer program
control to location 2004 000016 .
When running, the resident firmware provides the services expected of a VAX–11 console
system. In particular, the following services are available:
•
•
•
•

Automatic restart or bootstrap following processor halts or initial power-up.
An interactive command language allowing the user to examine and alter the state of the
processor.
Diagnostic tests executed on power-up that check out the CPU, the memory system, and
the Q22-bus Map.
Support of video or hardcopy terminals as the console terminal.

Refer to the KA660 Console Program Specification for a complete description of these features.

KA660 Boot and Diagnostic Facility 8–5

8.3.2.1

Power-Up Modes

The Boot and Diagnostic EPROM programs use Boot and Diagnostic Code <1:0> (Refer to
Section 12.8 to determine the power-up modes shown in Table 8–4.
Table 8–4: Power-Up Modes
Code

Power-up Mode

Run (factory setting). If the console terminal supports the Multi-national Character Set (MCS),
the user will be prompted for a language if the time-of-year clock battery backup has failed, or if
the SSC RAM is corrupted or uninitialized (1st power-up). Full startup diagnostics are run.

Language Inquiry. If the console terminal supports MCS, the user will be prompted for language
on every power up and restart. Full startup diagnostics are run.

Test. UVROM programs run wrap-around serial line unit (SLU) tests.

Manufacturing. To provide for rapid startup during certain manufacturing test procedures, the
EPROM programs omit the power up memory diagnostics and set up the memory bit map on the
assumption that all available memory is functional.

8–6

KA660 Boot and Diagnostic Facility

8.4 Battery Backed-up RAM
The KA660 contains 1KB of battery backed-up static RAM found in the SSC,for use as a
console "scratchpad". The power for the RAM is provided on pins 38 (VBAT H) and 20, 18,
16-13, 10-8, 6-3 (GND) of the 40-pin connector. This RAM supports byte, word and longword
references. Read operations take 700ns to complete while write operations require 600ns. The
RAM is organized as a 256 X 32-bit (one longword) array. The array appears in a 1KB block of
the VAX I/O page at addresses 2014 0400- 2014 07FF16 . This array is not protected by parity,
and CDAL Bus parity is neither checked nor generated on reads or writes to this RAM.

8.5

KA660 Initialization
The VAX Architecture defines three kinds of hardware initialization:
•
•
•

Power-up initialization
I/O Bus initialization
Processor initialization

8.5.1 Power-Up Initialization
Power-up initialization is the result of the restoration of power and includes a hardware
reset, a processor initialization an I/O bus initialization, as well as the initialization of several
registers defined in the VAX Architecture Reference Manual.

Hardware Reset
A KA660 hardware reset occurs on power-up or the negation of DCOK. A hardware reset
causes the hardware halt procedure (see Section 3.9.6) to be initiated with a halt code of 03. It
also initializes some IPR’s and most I/O Page registers to a known state. Those IPR’s affected
by a hardware reset are noted in Section 3.4. The effect a hardware reset has on I/O Space
registers is documented in the description of the register.

8.5.2 I/O Bus Initialization
An I/O Bus Initialization occurs on power-up, the negation of DCOK, or as the result of a
MTPR to IPR 55 (IORESET) If the KA660 is an arbiter, an I/O Bus initialization clears the
IPCR and DSER, and causes the Q22-bus interface to acquire both the CDAL and Q22-buses,
then assert the Q22-bus BINIT signal.

KA660 Boot and Diagnostic Facility 8–7

I/O Bus Reset Register (IPR 55)
The I/O Bus Reset Register (IORESET), IPR 5510 is implemented in the SSC chip. If the
KA660 is configured as an arbiter, an MTPR of any value to IORESET causes an I/O bus
initialization.

8.5.3 Processor Initialization
A Processor Initialization occurs on power-up, the negation of DCOK, and after a halt caused
by an error condition.
In addition to initializing those registers defined in the VAX Architecture Reference Manual,
the KA660 firmware must also configure Main Memory, the Local I/O Page, and the Q22-bus
Map during a processor initialization.

8.5.3.1

Configuring the Local I/O Page

The following registers control the configuration of the local I/O Page. They are unique to CPU
designs that use the SSC system support chip and they must be configured by the firmware
during a processor initialization:
•
•
•
•
•

SSC Base Address Register
BDR Address Decode Match Register
BDR Address Decode Mask Register
SSC Configuration Register
CDAL Bus Timeout Register

8.5.4 SSC Base Address Register (SSCBR)
The SSC Base Address Register, address 2014 000016 , controls the base addresses of a 2KB
block of the Local I/O Space which includes the the following:
•
•
•
•
•
•

Battery backed-up RAM,
The registers for the programmable timers,
The BDR Address Decode Match and Mask Registers,
The Diagnostic LED Register,
The CDAL Bus Timeout Register
A set of diagnostic registers that allow several EPR’s to be accessed via I/O page addresses.

This read/write register is set to 2014 000016 on power-up and the negation of DCOK when
SCR <7> is clear. Bits SSCBR<31:30,10:0> are unused. They read as ZEROs, and must be
written as ZEROs. SSCBR<29> is read as ONE and must be written as ONE. This register
should also be set to 2014 000016 by firmware during processor initialization. The SSCBR has
the format shown in Figure 8–3.

8–8

KA660 Boot and Diagnostic Facility

Figure 8–3: SSC Base Address Register (SSCBR)

3 3 2 2
1 0 9 8
MBZ 1

1 1
1 0
BASE ADDRESS BITS

<28:11>

MBZ
ESB90P0036

8.5.5 BDR Address Decode Match Register (BDMTR)
The Local I/O Device Address Decode Match Register, address 2014 013016 , controls the base
address of the Boot and Diagnostic Register. This read/write register is cleared on power-up
and the negation of DCOK. BDMTR<31:30,1:0> are unused. They read as ZEROs, and must
be written as ZEROs. This register should be set to 2008 400016 by firmware during processor
initialization. The BDMTR has the format shown in Figure 8–4.
Figure 8–4: BDR Address Decode Match Register (BDMTR)

3 3 2
1 0 9
MBZ

2 1 0
BASE ADDRESS MATCH BITS<29:2>

MBZ
ESB90P0037

8.5.6 BDR Address Decode Mask Register (BDMKR)
The BDR Address Decode Mask Register, address 2014 013416 , controls the range of addresses
to which the BDR responds to. (An example is the number of copies of the BDR that appear in
the physical address space). This read/write register is cleared on power-up and the negation
of DCOK. Bits BDMKR<31:30,1:0> are unused. They read as ZEROs, and must be written as
ZEROs. This register should should be set to 0000 007F 16 (32 copies of the BDR) by firmware
during processor initialization . The BDMKR has the format shown in Figure 8–5.

KA660 Boot and Diagnostic Facility 8–9

Figure 8–5: BDR Address Decode Mask Register (BDMKR)

3 3 2
1 0 9

2 1 0

MBZ

BASE ADDRESS MASK BITS<29:2>

MBZ
ESB90P0038

NOTE
The KA660 uses only one of the SSC’s address strobes. The other strobe’s control
registers located at 2014 013016 and 2014 013416 are reserved and should not be
accessed as they could cause unpredictable behavior.

8.5.7 SSC Configuration Register (SSCCR)
The SSC Configuration Register, address 2014 001016 , controls the set-up parameters for the
console serial line, programmable timers, EPROM, TOY Clock and BDR. The format is shown
in Figure 8–6. Table 8–5 contains a list of the bit descriptions.
Figure 8–6: SSC Configuration Register (SSCCR)

3
1
B
L
O

3
0

2
8
MBZ

2
7

2
6

I M
V B
D Z

2
5

2
4

2
3

2
2

2
0

1
9

IPL
LVL
SEL

R
S
P

EPROM M
B
SIZE
Z
SEL

1
8

1
6

1
5

1
4

HALT
PROT
SPACE

C
T
P

CT
BAUD
SEL

1
2

1
1

7
MBZ

BDR
EN

MBZ

ESB90P0039

8–10

KA660 Boot and Diagnostic Facility

Table 8–5: SSC Configuration Register Bit Descriptions
Data Bit

Name

Description

<31>

BLO

Battery Low. Read/Write. If the battery voltage goes below threshold
while the module is powered down, this bit is set on power-up, after the
assertion of DCOK after the assertion of POK. Once set, this bit can only
be cleared by software writing it as ONE. If this bit is set, then the TOY
Clock will be cleared by power-up and and by the negation of DCOK.

<30:28>

MBZ

Read as ZEROs, must be written as ZEROs.

<27>

IVD

Interrupt Vector Disable. Read/Write. When set, the console serial line
and programmable timers will not respond to interrupt acknowledge
cycles. Cleared on power-up and, by the negation of DCOK, and by a
processor initialization.

<26>

MBZ

Read as ZEROs, must be written as ZEROs.

<25:24>

IPL_LVL_
SEL

IPL Level Select Read/Write. These bits are used to specify the IPL
level of interrupt acknowledge cycle that the console serial line and programmable timers respond to. These bits must be cleared (programmed
to 00 (binary)) for the console serial line and programmable timers to
respond to interrupt acknowledge cycles that they generated (IPL 14).
These bits are cleared on power-up, by the negation of DCOK and by a
processor initialization.

<23>

RSP

ROM Speed. Read/Write. This bit is used to select the EPROM access
time. This bit must be set for the KA660 EPROMs to run at maximum
speed. This bit is cleared on power-up and by the negation of DCOK. It
must be set to ONE by a processor initialization.

<22:20>

ROM_
SIZE_SEL

EPROM Address Space Size Select. Read/Write. These bits control the
size of the range of addresses to which the EPROM responds. These
bits must be set to 101 (binary) because the KA660 contains 256KB
of EPROM, yielding an address range of 256KB (2004 0000 - 2007
FFFF16 ). These bits are cleared on power-up and by the negation of
DCOK, yielding an address range of 8KB (2004 0000 - 2004 1FFF16 ).
These bits must be set to the proper value by a processor initialization.

<18:16>

HALT
PROT
SPACE

EPROM Halt Protect Address Space Size Select. Read/Write. These bits
control the size of the Halt Mode address range. These bits must be set
to 101 (binary) because the KA660 contains 256KB of EPROM, yielding
a Halt Mode address range of 256KB (2004 0000 - 2007 FFFF16 ). These
bits are cleared on power-up and by the negation of DCOK yielding a
Halt Mode address range of 8KB (2004 0000 - 2004 1FFF16 ). These bits
must be set to the proper value by a processor initialization. Note, any
instruction fetch from the EPROM put the KA660 in HALT Protect Mode.

<15>

CTP

Control P Enable. Read/Write. When this bit is set, a CTRL/P typed at
the console causes the CPU to be halted, if halts are enabled (BDR<7>
set). When this bit is cleared, a BREAK typed at the console causes the
CPU to be halted, if halts are enabled (BDR<7> set). This bit is cleared
on power-up and by the negation of DCOK.

KA660 Boot and Diagnostic Facility 8–11

Table 8–5 (Cont.): SSC Configuration Register Bit Descriptions
Data Bit

Name

Description

<14:12>

CT BAUD
SELECT

Console Terminal Baud Rate Select. Read/Write. These bits are used
to select the baud rate of the Console Terminal Serial Line. They are
cleared on power-up and by the negation of DCOK. They should be
loaded from compliment of BDR<6:4> by the processor initialization code.
The bit encodings correspond to selected baud rates as shown in the
following table.
SSCCR<14:12>

Baud Rate

000

300

001

600

010

1200

011

2400

100

4800

101

9600

110

19200

111

38400

<11:7>

MBZ

Read as ZERO, must be written as ZERO.

<6:4>

BDR EN

Read/Write. These bits are used to enable the Boot and Diagnostic
Register. They are cleared on power-up and by the negation of DCOK.
These bits must be set to 111 (binary) by a processor initialization to
enable the BDR.

<3:0>

MBZ

Read as ZERO, must be written as ZERO.

NOTE
The SSC baud clock runs about 1.7% fast which is within the SRM mandated
accuracy. This is due to the accuracy of the crystal oscillator.

8.6 CDAL Bus Timeout Control Register (CBTCR)
The CDAL Bus Timeout Register, address 2014 0020 16 , controls the amount of time allowed
to elapse before a CDAL Bus cycle is aborted. The effect of this timer is blocked by the KA660
logic on all Q22-Bus references, because the Q22-Bus interface has its own timers for Q22-Bus
references. This timer prevents unanswered CDAL Bus cycles (other than those that go to
the Q22-Bus interface) from hanging the system any longer than the timeout interval. Even
though the effect of the timer is blocked on all Q22-bus references, bits<31:30> will
still be set on Q22-bus references that take longer than the programmmed value
(4us), so these bits are not useful as error indicators. Figure 8–7 shows the format.
Table 8–6 lists the bit descriptions.

8–12

KA660 Boot and Diagnostic Facility

Figure 8–7: CP Bus Timeout Control Register (CBTCR)

3 3
1 0

2 2
4 3
MBZ

0
BUS TIMEOUT INTERVAL

RWT
BTO
ESB90P0040

Table 8–6: CP Bus Timeout Control Register Bit Descriptions
Data Bit

Name

Description

<31>

BTO

CP Bus Timeout. Read/Write to Clear. This bit is set when the BUS
TIMEOUT INTERVAL set in bits <23:0> has expired during any CP
Bus cycle. This bit is cleared by writing a ONE, on power-up and
the negation of DCOK.

<30>

RWT

CP Bus Read/Write Timeout. Read/Write to Clear. This bit is set
when the BUS TIMEOUT INTERVAL set in bits <23:0> has expired
during a CPU or DMA read or write cycle on the CP Bus. This bit
is cleared by writing a one, on power-up and the negation of DCOK.

<29:22>

MBZ

Read as ZEROs, must be written as ZEROs.

<23:0>

BUS TIMEOUT
INTERVAL

Read/Write. These bits are used to program the desired timeout
period. The available range of 1 to FFFFFF16 corresponds to a
selectable timeout range of 1µs to 16.77 seconds in 1µs increments.
Writing a zero to this field disables the bus timeout function. The
BTO bit is used to signify that a bus timeout has occurred. This
field is cleared on power-up and the negation of DCOK. This register
should be loaded with 0000 400016 on a processor initialization for a
timeout value of 15 milliseconds.

KA660 Boot and Diagnostic Facility 8–13

Chapter 9
KA660 Q22-bus Interface

The KA660 includes a Q22-bus interface implemented via a single VLSI chip called the
CQBIC. It contains a CDAL Bus to Q22-bus interface, that supports the following:
•

•
•

•

A programmable mapping function (scatter-gather map) for translating 22-bit, Q22-bus
addresses into 29-bit CDAL addresses that allows any page in the Q22-bus memory space
to be mapped to any page in main memory.
A direct mapping function for translating 29-bit CDAL addresses in the local Q22-bus
address space and local Q22-bus I/O page into 22-bit, Q22-bus addresses.
Masked and unmasked longword reads and writes from the CPU to the Q22-bus Memory
and I/O Space and the Q22-bus interface registers. Longword reads and writes of the local
Q22-bus memory space are buffered and translated into two-word, block mode, transfers
on the Q22-bus. Longword reads and writes of the local Q22-bus I/O space are buffered
and translated into two, single-word transfers on the Q22-bus.
Up to sixteen-word, block mode, writes from the Q22-bus to main memory. These words
are buffered then transferred to main memory using two asynchronous DMA octaword
transfers. For block mode writes of less than sixteen words, the words are buffered and
transferred to main memory using the most efficient combination of octaword, quadword,
and longword asynchronous DMA transfers. The maximum write bandwidth for block
mode references is 3.3 MB/sec. Block mode reads of main memory from the Q22-bus cause
the Q22-bus interface to perform an asynchronous DMA quadword read of main memory
and buffer all four words, so that on block mode reads, the next three words of the block
mode read can be delivered without any additional CDAL cycles. The maximum read
bandwidth for Q22-bus block mode references is 2.4 MB/sec. Q22-bus Burst Mode DMA
transfers result in single-word reads and writes of Main Memory.
Transfers from the CPU to the local Q22-bus memory space, that result in the Q22-bus
Map translating the address back into main memory (Local-Miss, Global-Hit transactions).

The Q22-bus interface contains several registers for Q22-bus control and configuration, interprocessor communication, and error reporting.
The interface also contains Q22-bus interrupt arbitration logic that recognizes Q22-bus interrupt requests BR7-BR4 and translates them into CPU interrupts at levels 17-14.
The Q22-bus interface detects Q22-bus "NO SACK" timeouts, Q22-bus Interrupt Acknowledge
timeouts, Q22-bus non-existent memory timeouts, main memory errors on DMA accesses from
the Q22-bus and Q22-bus device parity errors.

KA660 Q22-bus Interface 9–1

9.1 Q22-bus to Main Memory Address Translation
On DMA references to main memory, the 22-bit, Q22-bus address must be translated into a
29-bit main memory address (Figure 9–1.) This translation process is performed by the Q22bus interface by using the Q22-bus Map. This map contains 8192 mapping registers, (one for
each page in the Q22-bus Memory Space), each of which can map a page (512 bytes) of the
Q22-bus Memory address space into any of the 1024K pages in main memory. Since Local I/O
Space addresses cannot be mapped to Q22-bus pages, the Local I/O Page is unaccessible to
devices on the Q22-bus. Figure 9–1 shows how Q22-bus addresses are translated into main
memory addresses.

9–2

KA660 Q22-bus Interface

Figure 9–1: Q22-bus Address Translation

2
1

9 8

Q22−Bus Address

CACHE TAGS

−−− −−−
/
\ V
\___/

MATCH

0
MAPPING CACHE
|
|
V
3
1

Mapping Register
1
9
2
8

9 8

Physical Address of Main Memory
ESB90P0042

At power up time, the Q22-bus Map Registers, including the valid bits, are undefined.
External access to main memory is disabled so long as the Interprocessor Communication
Register LM EAE bit is cleared. The Q22-bus Interface monitors each Q22-bus cycle and
responds if the following three conditions are met:
1. The Interprocessor Communication Register LM EAE bit is set.
2. The Valid bit of the selected mapping register is set.

KA660 Q22-bus Interface 9–3

3. During read operations, the mapping register must map into existent main memory, or a
Q22-bus timeout occurs. (During write operations, the Q22-bus Interface returns Q22-bus
BRPLY before checking for existent local memory; the response depends only on conditions
1 and 2 above).

NOTE
In the case of local-miss, global-hit, the state of the LM EAE bit is ignored.
If the map cache does not contain the needed Q22-bus Map Register, then the Q22-bus interface will perform an asynchronous DMA read of the Q22-bus Map register before proceeding
with the Q22-bus Bus DMA transfer.

9.1.1 Q22-bus Map Registers (QMR)’s
The Q22-bus Map contains 8192 registers that control the mapping of Q22-bus addresses into
main memory. Each register maps a page of the Q22-bus Memory Space into a page of main
memory. These registers are implemented in a 32KB block of main memory, but are accessed
through the CQBIC chip via a block of addresses in the I/O Page.
The Local I/O Space address of each register was chosen so that register address bits <14:2>
are identical to Q22-bus address bits <21:9> of the Q22-bus page which the register maps.
Table 9–1 lists the register addresses.
Table 9–1: Q22-bus Map Register Addresses

Q22-bus Addresses
Mapped
(Hex)
----------------00 0000 - 00 01FF
00 0200 - 00 03FF
00 0400 - 00 05FF
00 0600 - 00 07FF
00 0800 - 00 09FF
00 0A00 - 00 0BFF
00 0C00 - 00 0DFF
00 0E00 - 00 0FFF

Q22-bus Addresses
Mapped (Octal)
----------------------00 000 000 - 00 000 777
00 001 000 - 00 001 777
00 002 000 - 00 002 777
00 003 000 - 00 003 777
00 004 000 - 00 004 777
00 005 000 - 00 005 777
00 006 000 - 00 006 777
00 007 000 - 00 007 777

.
.
.

2008 FFF0
2008 FFF4
2008 FFF8
2008 FFFC

3F F800 - 3F F9FF
3F FA00 - 3F FBFF
3F FC00 - 3F FDFF
3F FA00 - 3F FFFF

17 774 000 - 17 774 777
17 775 000 - 17 775 777
17 776 000 - 17 776 777
17 776 000 - 17 777 777

The Q22-bus Map Registers (QMR’s) have the format shown in Figure 9–2.

9–4

KA660 Q22-bus Interface

Figure 9–2: Q22-bus Map Register Format

3 3
1 0

2 1
0 9

MBZ

0
A28 − A9
ESB90P0043

Table 9–2 describes the bits in the Q22-bus Map Register.
Table 9–2: Q22-bus Map Register Bit Description
Data Bit

Name

Description

<31>

Valid. Read/Write. When a Q22-bus Map Register is selected
by bits <21:9> of the Q22-bus address, the Valid bit determines
whether mapping is enabled for that Q22-bus page. If the Valid
bit is set, the mapping is enabled, and Q22-bus addresses within
the page controlled by the register are mapped into the main
memory page determined by bits <28:9>. If the Valid bit is clear,
the mapping register is disabled, and the Q22-bus interface
does not respond to addresses within that page. This bit is
UNDEFINED on power-up and the negation of DCOK.

<30:20>

Unused

These bits always read as zero and must be written as zero.

<19:0>

A28-A9

Address Bits <28:9> Read/Write. When a Q22-bus Map Register
is selected by a Q22-bus address, and if that register’s Valid bit
is set, then these 20 bits are used as main memory address bits.
Q22-bus address bits <8:0> are used as main memory address
bits <8:0>. These bits are UNDEFINED on power-up and the
negation of DCOK.

9.1.2 Accessing the Q22-bus Map Registers
Although the CPU accesses the Q22-bus Map Registers with aligned, longword references to
the Local I/O Page (addresses 2008 8000 - 2008 FFFC 16 ), the map actually resides in a 32KB
block of main memory. The starting address of this block is controlled by the contents of the
Q22-bus Map Base Register. The Q22-bus interface also contains a 16-entry, fully associative,
Q22-bus Map Cache to reduce the number of main memory accesses required for address
translation.
NOTE
The system software must protect the pages of memory that contain the Q22-bus
Map from direct accesses that will corrupt the map or cause the entries in the
Q22-bus Map Cache to become stale. Either of these conditions will result in the
incorrect operation of the mapping function.

KA660 Q22-bus Interface 9–5

When the CPU accesses the Q22-bus Map through the Local I/O Page addresses, the Q22-bus
interface reads or writes the map in main memory. The Q22-bus Bus interface does not have
to gain Q22-bus mastership when accessing the Q22-bus Map. Because these addresses are in
the local I/O space, they are not accessible from the Q22-bus.
On a Q22-bus Map read by the CPU, the Q22-bus interface decodes the Local I/O Space
address (2008 8000 - 2008 FFFC16 . If the register is in the Q22-bus Map Cache, the Q22-bus
interface will internally resolve any conflicts between CPU and Q22-bus transactions (if both
are attempting to access the Q22-bus Map Cache entries at the same time), then return the
data. If the map register is not in the map cache, the Q22-bus interface will force the CPU to
retry, acquire the CDAL bus and perform an asynchronous DMA read of the map register. On
completion of the read, the CPU is provided with the data when its read operation is retried.
A map read by the CPU does not cause the register that was read to be stored in the map
cache.
On a Q22-bus Map write by the CPU, the Q22-bus interface latches the data, then on the completion of the CPU write, acquires the CDAL bus and performs an asynchronous DMA write to
the map register. If the map register is in the Q22-bus Map Cache, then the CAMValid bit for
that entry will be cleared to prevent the entry from becoming stale. A Q22-bus Map write by
the CPU does not update any cached copies of the Q22-bus Map Register.

9.1.3 The Q22-bus Map Cache
To speed up the process of translating Q22-bus address to Main Memory addresses, the
Q22-bus interface utilizes a fully associative, sixteen entry, Q22-bus Map Cache which is
implemented in the CQBIC chip.
If a DMA transfer ends on a page boundary, the Q22-bus interface will prefetch the mapping
register required to translate the next page and load it into the cache, before starting a new
DMA transfer. This allows Q22-bus block mode DMA transfers that cross page boundaries
to proceed without delay. The replacement algorithm for updating the Q22-bus Map Cache is
FIFO.
The cached copy of the Q22-bus Map Register is used for the address translation process.
If the required map entry for a Q22-bus address (as determined by bits <21:9> of the Q22bus address), is not in the map cache, then the Q22-bus interface uses the contents of the
Map Base Register to access main memory and retrieve the required entry. After obtaining
the entry from main memory, the valid bit is checked. If it is set, the entry is stored in the
cache and the Q22-bus cycle continues. Figure 9–3 shows the format. Table 9–3 contains a
description of the Q22-bus Map Cache Entry bits.

9–6

KA660 Q22-bus Interface

Figure 9–3: Q22-bus Map Cache Entry Format

3
3

3
2

2 1
0 9
Q22−BUS ADR <21:9>

0
A28 − A9
ESB90P0044

Table 9–3: Q22-bus Map Cache Entry Bit Description

9.2

Data Bit

Name

Description

<33>

CAMValid

When a mapping register is selected by a Q22-bus address, the
CAMValid bit determines whether the cached copy of the mapping
register for that address is valid. If the CAMValid bit is set, the
mapping register is enabled, and addresses within that page can be
mapped. If the CAMValid bit is clear, the Q22-bus interface must
read the map in local memory to determine if the mapping register
is enabled. This bit is cleared on power-up, the negation of DCOK,
by setting the QMCIA (Q22-bus Map Cache Invalidate All) bit in
the Interprocessor Communication Register, on writes to IPR 55
(IORESET), by a write to the Q22-bus Map Base Register, or by
writing to the QMR that is being cached.

<32:20>

QBUS ADR

These bits contain the Q22-bus address bits <21:9> of the page
that this entry maps. This is the content addressable field of the
16 entry cache for determining if the map register for a particular
Q22-bus address is in the map cache. These bits are undefined on
power-up.

<19:0>

Address bits
A28-A9

When a mapping register is selected by a Q22-bus address, and if
that register’s CAMValid bit is set, then these 20 bits are used as
main memory address bits 28 through 9. Q22-bus address bits 8
through 0 are used as local memory address bits 8 through 0. These
bits are undefined on power-up.

CDAL to Q22-bus Address Translation
CDAL bus addresses within the CDAL Translation) Local Q22-bus I/O Space, addresses 2000
0000 - 2000 1FFF16 , are translated into Q22-bus I/O Space addresses by using bits <12:0> of
the CDAL bus address as bits <12:0> of the Q22-bus address and asserting BBS7. Q22-bus
address bits <21:13> are driven as zeros.
CDAL bus addresses within the Local Q22-bus Memory Space, addresses 3000 0000 - 303F
FFFF 16 , are translated into Q22-bus Memory Space addresses by using bits <21:0> of the
CDAL bus address as bits <21:0> of the Q22-bus address.

KA660 Q22-bus Interface 9–7

9.3 Interprocessor Communications Facility
The KA660 can only be configured as a Q22-bus arbiter.
The KA660 Interprocessor Communication Facility allows other processors on the system to
request program interrupts from the KA660 without using the Q22-bus interrupt request
lines. It also controls external access to local memory (via the Q22-bus map).

9.3.1 Interprocessor Communication Register (IPCR)
The Interprocessor Communication Register is a 16-bit register which resides in the Q22bus I/O page address space and can be accessed by any device which can become Q22-bus
master (including the KA660 itself). The IPCR is implemented in the CQBIC chip and is
byte accessible, meaning that a write byte instruction can write to either the low or high byte
without affecting the other byte. The I/O page address of the IPCR is constant with the KA660
because it only supports arbiter mode and not auxiliary mode. The hex 32-bit address is 2000
1F40 and the octal 22-bit address is 17 777 500. Figure 9–4 shows the format. Table 9–4
describes the bits.

Figure 9–4: Interprocessor
Register (IPCR)
1 1 1 1 1 Communication
1
5

1
MBZ

7
M
B
Z

MBZ

DMA QME
QMCIA
AUX HLT
DBI IE
LM EAE
DBI RQ
ESB90P0045

9–8

KA660 Q22-bus Interface

Table 9–4: Interprocessor Communication Register Bit Description
Data Bit

Name

Description

<15>

DMA QME

DMA Q22-bus Address Space Memory error. Read/Write to clear.
This bit indicates that an error occurred when a Q22-bus device was
attempting to read main memory. It is set if DMA System Error
Register bit DSER<4> (MAIN MEMORY ERROR) is set, or the
CDAL timer expires. The MAIN MEMORY ERROR bit indicates
that an uncorrectable error occurred when an external device (or
CPU) was accessing the KA660 local memory. The CDAL timer
expiring indicates that the memory controller did not respond when
the Q22-bus interface initiated a DMA transfer. This bit is cleared
by writing a one to it,on power-up, by the negation of DCOK, by
writes to IPR 55 (IORESET), and whenever DSER<4> is cleared.

<14>

QMCIA

Q22-bus Map Cache Invalidate All. Write Only. Writing a one to
this bit clears the CAMValid bits in the cached copy of the MAP.
This bit always reads as zero. Writing a zero has no effect.

<13:09>

Unused

Read as zeros. Must be written as zeros.

<8>

AUX HLT

Auxiliary Halt. Read Only. When this bit is set it has no effect on
the operation of the on-board CPU. This bit is cleared on power-up,
by the negation of DCOK, by writes to IPR 55 (IORESET). Note:
This bit should never be set because the processor does not
support auxiliary mode.

<7>

Unused

Read as zero. Must be written as zero.

<6>

DBI IE

Doorbell Interrupt Enable. Read/Write when the KA660 is Q22-bus
master. Read Only when another device is Q22-bus master. When
set, this bit enables interprocessor doorbell interrupt requests via
IPCR<0>. Cleared on power-up, by the negation of DCOK, and
writes to IPR 55 (IORESET).

<5>

LM EAE

Local Memory External Access Enable. Read/Write when the KA660
is Q22-bus master. Read Only when another device is Q22-bus
master. When set, this bit enables external access to local memory
(via the Q22-bus map). Cleared on power-up and by the negation of
DCOK.

<4:1>

Unused

Read as zeros. Must be written as zeros.

<0>

DBI RQ

Doorbell Interrupt Request.Read/Write. If IPCR<6> (DBI IE) is set,
setting this bit generates a doorbell interrupt request. If IPCR<6>
is clear, setting this bit has no effect. Clearing this bit has no effect.
DBI RQ is cleared when the CPU grants the doorbell interrupt
request. DBI RQ is held clear whenever DBI IE is clear. This bit is
cleared on power-up and the negation of DCOK.

KA660 Q22-bus Interface

9–9

9.3.2 Interprocessor Doorbell Interrupts
If the Interprocessor Communication Register DBI IE bit is set, any Q22-bus master can
request an interprocessor doorbell interrupt by writing a one into IPCR bit <0>.
The interrupt vector is 20416 and the Interrupt priority is 1416 . This IPL is the same as
BR4 on the Q22-bus. The interprocessor doorbell is the third highest priority IPL 14 device,
directly after the console serial line unit and the programmable timers.
NOTE
Following an interprocessor doorbell interrupt, the KA660 CPU sets the IPL to 14.
The IPL is set to 17 for external Q22-bus BR4 interrupts.

9.4 Q22-bus Interrupt Handling
The KA660 responds to interrupt requests BR7-4 with the standard Q22-bus interrupt acknowledge protocol (DIN followed by IAK). The console serial line unit, the programmable
timers, and the interprocessor doorbell request interrupt at IPL 14 and have priority over
all Q22-bus BR4 interrupt requests. After responding to any interrupt request BR7-4, the
CPU sets the processor priority to IPL 17. All BR7-4 interrupt requests are disabled unless
software lowers the interrupt priority level.
Interrupt requests from the KA660 interval timer are handled directly by the CPU. Interval
timer interrupt requests have a higher priority than BR6 interrupt requests. After responding
to an interval timer interrupt request, the CPU sets the processor priority to IPL 16. Thus,
BR7 interrupt requests remain enabled.

9.5 Configuring the Q22-bus Map
The KA660 implements the Q22-bus Map in an 8K longword (32KB) block of main memory.
This Map must be configured by the KA660 firmware during a processor initialization by
writing the base address of the uppermost 32KB block of good main memory into the Q22-bus
Map Base Register. The base of this map must be located on a 32KB boundary.
NOTE
This 32KB block of main memory must be protected by the system software. The
only access to the map should be through Local I/O page addresses 2008 8000 - 2008
FFFC 16 .

9.5.1 Q22-bus Map Base Address Register (QBMBR)
The Q22-bus Map Base Address Register, address 2008 0010 16 controls the main memory
location of the 32KB block of Q22-bus Map Registers. This Read/Write register is accessible by
the CPU on a longword boundary only. Bits <31:29,14:0> are unused and should be written as
zero and will return zero when read. Figure 9–5 shows the format.
A write to the Map Base Register will flush the Q22-bus Map Cache by clearing the CAMValid
bits in all the entries.

9–10

KA660 Q22-bus Interface

The contents of this register are undefined on power-up and the negation of DCOK when
SCR<7> is clear and is not affected by BINIT being asserted on the Q22-bus.
Figure 9–5: Q22-bus Map Base Address Register (QBMBR)

3 2 2
1 9 8
MBZ

1 1
5 4
MAP BASE

0
MBZ
ESB90P0046

9.6 System Configuration Register (SCR)
The System Configuration Register, address 2008 0000 16 , contains the processor number
which determines the address of the IPCR register, a BHALT enable bit, a power OK flag
and an AUX flag. Figure 9–6 shows the format. Table 9–5 describes the bits in the System
Configuration Register.
The System Configuration Register (SCR) is longword, word, and byte accessible. Programmable
option fields are cleared on power-up and by the negation of DCOK when SCR<7> is clear.

KA660 Q22-bus Interface 9–11

Figure 9–6: System Configuration Register (SCR)

3
1

1 1 1 1 1 1
5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0
MBZ

MBZ

POK
BHALT ENB
AUX
ACTION ON DCOK NEGATION
DOORBELL OFFSET SELECT
MUST BE ZERO
ESB90P0047

Table 9–5: System Configuration Register Bit Description

9–12

Data Bit

Name

Description

<31:16>

Unused

Read as zero. Must be written as zero.

<15>

POK

Power OK. Read Only. Writes have no effect. This bit is set if the
Q22-bus BPOK signal is asserted and clear if it is negated. This bit
is cleared on power-up and by the negation of DCOK.

<14>

BHALT EN

BHALT Enable. Read/Write. This bit controls the effect the Q22-bus
BHALT signal has on the CPU. When set, asserting the Q22-bus
BHALT signal will halt the CPU and assert DSER<15>. When
cleared, The Q22-bus BHALT signal will have no effect. This bit is
cleared on power-up and by the negation of DCOK.

<13:11>

Unused

Read as zero. Must be written as zero.

<10>

AUX

Auxiliary. Read Only. Writes have no effect. This bit defines
Auxiliary and Arbiter mode of operation of the KA660 When read as
a zero, Arbiter mode is selected, and when read as a one, Auxiliary
mode is selected. Because the KA660 can only be configured as an
arbiter this bit should always read as zero.

<9:8>

Unused

Read as zero. Must be written as zero.

<7>

ACTION
ON DCOK
NEGATION

Read/Write. When cleared, the Q22-bus interface asserts
SYSRESET (causing a hardware reset of the board and control
to be passed to the resident firmware via the hardware halt procedure with a halt code of 3) when DCOK is negated on the Q22-bus.
When set, the Q22-bus interface asserts HALCYON (causing control
to be passed to the resident firmware via the hardware halt procedure with a halt code of 2) when DCOK is negated on the Q22-bus.
Cleared on power-up and the negation of DCOK.

<6:4>

Unused

Read as zero. Must be written as zero.

KA660 Q22-bus Interface

Table 9–5 (Cont.): System Configuration Register Bit Description
Data Bit

Name

Description

<3:1>

RESERVED

Reserved for use by Digital.

<0>

Unused

Read as zero. Must be written as zero.

9.7 Error Reporting Registers
There are three registers associated with Q22-bus interface error reporting:
•
•
•

The DMA System Error Register (DSER)
The Q22-bus Error Address Register (QBEAR)
The DMA Error Address Register (DEAR)

These registers are located in the local VAX I/O address space and can only be accessed by the
local processor. The DSER is implemented in the CQBIC chip and it logs main memory errors
on DMA transfers, Q22-bus parity errors, Q22-bus non-existent memory errors, and Q22-bus
no-Grant The QBEAR contains the address of the page in Q22-bus space which caused a
parity error during an access by the local processor. The DEAR contains the address of the
page in local memory which caused a memory error during an access by an external device or
the processor during a local-miss global-hit transaction. An access by the local processor which
the Q22-bus interface maps into main memory will provide error status to the processor when
the processor does a RETRY for a READ local miss-global hit, or by an interrupt in the case of
a local-miss global-hit write.

9.7.1 DMA System Error Register (DSER)
The DSER (address 2008 000416 )is a longword, word, or byte accessible Read/Write register
available to the local processor. The bits in this register are cleared to zero on power-up, by
the negation of DCOK when SCR <7> is clear, and by writes to IPR 55 (IORESET). All bits
are set to one to record the occurrence of an event. They are cleared by writing a one, writing
zeros has no effect.
The format of the DMA System Error Register is shown in Figure 9–7. Table 9–6 describes
the bits in the System Error Register.

KA660 Q22-bus Interface 9–13

Figure 9–7: DMA System Error Register (DSER)

3
1

1 1 1 1
6 5 4 3
MBZ

8 7 6 5 4 3 2 1 0
MBZ

Q22−BUS BHALT DETECTED
Q22−BUS DCOK NEGATION DETECTED
MASTER DMA NXM
MUST BE ZERO
Q22−BUS PE
MAIN MEMORY ERROR
LOST ERROR BIT
NO GRANT
MUST BE ZERO
DMA NXM
ESB90P0048

9–14

KA660 Q22-bus Interface

Table 9–6: DMA System Error Register Bit Description
Data Bit

Name

Description

<31:16>

Unused

Read as zero. Must be written as zero.

<15>

Q22-BUS BHALT DETECTED

Read/Write to clear. This bit is set when the Q22bus interface detects that the Q22-bus BHALT line
was asserted and SCR<14> (BHALT ENABLE)
is set. Cleared by writing a one, writes to IPR
55 (IORESET), on power-up and the negation of
DCOK.

<14>

Q22-BUS DCOK NEGATION
DETECTED

Read/Write to clear. This bit is set when the Q22bus interface detects the negation of DCOK on
the Q22-bus and SCR<7> (ACTION ON DCOK
NEGATION) is set. Cleared by writing a one,
writes to IPR 55 (IORESET), on power-up and the
negation of DCOK.

<13:8>

Unused

Read as zero. Must be written as zero.

<7>

MASTER DMA NXM

Read/Write to clear. This bit is set when the CPU
performs a demand Q22-bus read cycle or write cycle that does not reply after 10us. During interrupt
acknowledge cycles, or request read cycles, this bit
is not set. Cleared by writing a one, on power-up,
by the negation of DCOK and by writes to IPR 55
(IORESET).

<6>

Unused

Read as zero. Must be written as zero.

<5>

Q22-bus PARITY ERROR

Read/Write to clear. This bit is set when the CPU
performs a Q22-bus Demand read cycle which
returns a parity error. During interrupt acknowledge cycles, or request read cycles this bit is not
set. Cleared by writing a one, on power-up, by
the negation of DCOK and by writes to IPR 55
(IORESET).

<4>

MAIN MEMORY ERROR

Read/Write to clear. This bit is set if an external
Q22-bus device or local miss global hit receives a
memory error while reading local memory. The
IPCR<15> reports the memory error to the external
Q22-bus device. Cleared by writing a one, on
power-up, by the negation of DCOK and by writes
to IPR 55 (IORESET).

KA660 Q22-bus Interface 9–15

Table 9–6 (Cont.): DMA System Error Register Bit Description
Data Bit

Name

Description

<3>

LOST ERROR

Read/Write to clear. This bit indicates that an error
address has been lost because of DSER<7,5,4,0>
having been previously set and a subsequent error
of either type occurs which would have normally
captured an address and set either DSER<7,5,4,0>
flag. Cleared by writing a one, on power-up, by
the negation of DCOK and by writes to IPR 55
(IORESET).

<2>

NO GRANT TIMEOUT

Read/Write to clear. This bit is set if the Q22-bus
does not return a bus grant within 10ms of the bus
request from a CPU demand read cycle, or write
cycle. During interrupt acknowledge or request
read cycles this bit is not set. Cleared by writing a
one, on power-up, by the negation of DCOK and by
writes to IPR 55 (IORESET).

<1>

Unused

Read as zero. Must be written as zero.

<0>

DMA NXM

Read/Write to clear. This bit is set on a DMA
transfer to a non-existent main memory location.
This includes local-miss global-hit cycles and map
accesses to non-existent memory. Cleared by writing a one, on power-up, by the negation of DCOK
when SCR<7> is clear, and by writes to IPR 55
(IORESET).

9.7.2 Q22-bus Error Address Register (QBEAR)
The Q22-bus Error Address Register, address 2008 0008 16 , is a Read Only, longword accessible, register which is implemented in the CQBIC chip. Its contents are valid only if DSER<5>
(Q22-BUS PARITY ERROR) is set, or if DSER<7> (MASTER DMA NXM) is set.
Reading this register when DSER<5> and DSER<7> are clear will return undefined results.
Additional Q22-bus parity errors that could have set DSER<5> or Q22-bus timeout errors that
could have caused DSER<7> to set, will cause DSER<3> to set.
The QBEAR contains the address of the page in Q22-bus space which caused a parity error
during an access by the on-board CPU which set DSER<5> or a master timeout which set
DSER<7>.
Q22-bus address bits <21:9> are loaded into QBEAR bits <12:0>. QBEAR bits <31:13> always
read as zeros.

9–16

KA660 Q22-bus Interface

Figure 9–8: Q22-bus Error Address Register (QBEAR)

3
1

11
32
MBZ

0
Q22−bus
Address Bits <21:9>
ESB90P0049

Figure 9–9: DMA Error Address Register (DBEAR)

3
1

2 1
0 9
MBZ

0
MAPPED Q22−BUS
Address Bits <28:9>
ESB90P0050

NOTE
This is a Read Only register, if a write is attempted a hard error (IPL 1D) will be
generated.

9.7.3 DMA Error Address Register (DEAR)
The DMA Error Address Register address 2008 000C 16 is a read only, longword accessible, register which is implemented in the CQBIC chip. It contains valid information only
when DSER<4> (MAIN MEMORY ERROR) is set or when DSER<0> (DMA NXM) is set .
Reading this register when DSER<4> and DSER <0> are clear will return UNDEFINED data.
Figure 9–9 shows the format.
The DBEAR contains the map translated address of the page in local memory which caused
a memory error or non existent memory error during an access by an external device or the
Q22-bus interface for the CPU during a local-miss global-hit transaction or Q22-bus Map
access.
The contents of this register are latched when DSER<4> or DSER<0 > are set. Additional
main memory errors or non-existent memory errors have no effect on the DBEAR until software clears DSER<4> and DSER<0> .
Mapped Q22-bus address bits <28:9> are loaded into DBEAR bits <19:0>. DBEAR bits
<31:20> always read as zeros.

KA660 Q22-bus Interface 9–17

NOTE
This is a Read Only register, if a write is attempted a hard error (IPL 1D) will be
generated.

9.8 Q22-bus Interface Error Handling
The Q22-bus interface does not generate or check CDAL parity.
The Q22-bus interface checks all CPU references to Q22-bus Memory and I/O spaces to insure
that nothing but masked and un-masked longword accesses are attempted. Any other type of
reference will cause a machine check abort to be initiated.
The Q22-bus interface maintains several timers to prevent incomplete accesses from hanging
the system indefinitely. They include: a 10µs non-existent memory timer for accesses to the
Q22-bus Memory and I/O Spaces, a 10µs "NO SACK" timer for acknowledgment of Q22-bus
DMA grants, and a 10ms "NO GRANT" timer for acquiring the Q22-bus.
If there is a non existent memory (NXM) error (10µs timeout) while accessing the Q22-bus
on a demand read reference, associated row in the cache is invalidated , bit DSER<7> is set,
the address of the Q22-bus page being accessed is captured in QBEAR<12:0>, and a machine
check abort is initiated.
If there is a NXM error on a prefetch read, or an interrupt acknowledge vector read, then the
prefetch or interrupt acknowledge reference is aborted but no information is captured and no
machine check occurs.
If there is a NXM error on a masked write reference, then DSER<7> is set, the address of the
Q22-bus page being accessed is captured in QBEAR<12:0>, and an interrupt is generated at
IPL 1D through vector 6016 .
If the Q22-bus interface does not receive an acknowledgment within 10µs after it has granted
the Q22-bus, the grant is withdrawn, no errors are reported, and the Q22-bus interface waits
500ns to clear the Q22-bus grant daisy chain before beginning arbitration again.
If the Q22-bus interface tries to obtain Q22-bus mastership on a CPU demand read reference,
and does not obtain it within 10ms, associated row in the cache is invalidated DSER<2> is set,
and a machine check abort is initiated.
The Q22-bus interface also monitors Q22-bus signals BDAL<17:16> while reading information
over the Q22-bus so that parity errors detected by the device being read from are recognized.
If a parity error is detected by another Q22-bus device on a CPU demand read reference to
Q22-bus Memory or I/O Space, then associated row in the cache is invalidated , DSER<5>
is set, the address of the Q22-bus page being accessed is captured in QBEAR<12:0>, and a
machine check abort is initiated.
If a parity error is detected by another Q22-bus device on a prefetch request read by the
CPU, the prefetch is aborted, associated row in the cache is invalidated , DSER<5> is set,
the address of the Q22-bus page being accessed is captured in QBEAR<12:0>, but no machine
check is generated.

9–18

KA660 Q22-bus Interface

The Q22-bus interface also monitors the backplane BPOK signal to detect power failures. If
BPOK is negated on the Q22-bus, a power fail trap is generated, and the CPU traps through
vector 0C16 . The state of the Q22-bus BPOK signal can be read from SCR<15>. The Q22- Bus
interface continues to operate after generating the powerfail trap, until DCOK is negated.

KA660 Q22-bus Interface 9–19

Chapter 10
KA660 Network Interface

The KA660 includes a network interface that is implemented via the Second Generation
Ethernet Controller chip (SGEC). When used in conjunction with the H3602 cover panel, this
interface allows the KA660 to be connected to either a thinwire or standard Ethernet network.
It supports the Ethernet Data Link Layer as specified in the VAX Architecture Reference
Manual. The SGEC also supports CP Bus Parity Protection.

10.1 Ethernet Overview
Ethernet is a serial bus that can support up to 1,024 nodes with a maximum separation of
2.8 kilometers (1.7 miles). Data is passed over the Ethernet in Manchester encoded format
at a rate of 10 million bits per second in variable-length packets. Each packet has the format
shown in Figure 10–1.

KA660 Network Interface 10–1

Figure 10–1: Ethernet Packet Format

6 bytes

Destination Address

6 bytes

Source Address

2 bytes

Type

46..1500 bytes
~
~

4 bytes

Data
.
.
.
.
.

~
~

CRC Check Code

ESB90P0051

The minimum size of a packet is 64 bytes, which implies a minimum data length of 46 bytes.
Packets shorter than this are called runt packets and are treated as erroneous when received
by the network controller.
All nodes on the Ethernet have equal priority. The technique used to control access to the
bus is Carrier Sense, Multiple Access, with Collision Detection (CSMA/CD). To access the bus,
devices must first wait for the bus to clear (no carrier sensed). Once the bus is clear, all devices
that want to access the bus have equal priority (multi-access), so they all attempt to transmit.
After starting transmission, devices must monitor the bus for collisions (collision detection). If
no collision is detected, the device may continue with transmission. If a collision is detected,
then the device waits for a random amount of time and repeats the access sequence.
Ethernet allows point to point communication between two devices, as well as simultaneous
communication between multiple devices. To support these two modes of communication, there
are two types of network addresses, physical and multicast. These two types of addresses are
both 48 bits (6 bytes) long and are described below.
Physical address: The unique address associated with a particular station on the Ethernet,
which should be distinct from the physical address of any other station on any other Ethernet.
Multicast address: A multi-destination address associated with one or more stations on a given
Ethernet (sometimes called a logical address). There are two kinds of multicast addresses:

10–2

KA660 Network Interface

Multicast-group address: An address associated by higher-level convention with a group of
logically related stations.
Broadcast address: A predefined multicast address which denotes the set of all the stations on
the Ethernet.)
Bit 0 (the least significant bit of the first byte) of an address denotes the type: it is 0 for
physical addresses and 1 for multicast addresses. In either case the remaining 47 bits form
the address value. A value of 48 ones is always treated as the broadcast address.
The hardware address of the KA660 module is determined at the time of manufacture and is
stored in the Network Interface Station Address ROM. Because every device that is intended
to connect to an Ethernet network must have a unique physical address, the bit pattern
blasted into the Network Interface Station Address ROM must be unique for each KA660 .
The Multicast Addresses to which the KA660 will respond are determined by the Multicast
Address Filter Mask in the Network Interface Initialization Block.

10.2 NI Station Address ROM (NISA ROM)
The Network Interface includes a byte-wide, 32-byte, socketted ROM called the Network
Interface Station Address ROM. One byte of this ROM appears in the second byte of each of
32 consecutive longwords in the address range 2008 4000 - 2008 407C16 . Bytes one, three and
four of each longword are defined in the Boot Diagnostic Register section 9.1. The second byte
of the first six longwords contain the 48-bit Network Physical Address (NPA) of the KA660 .
The low-order byte in the remaining 26 longwords are used for testing. This address range is
Read Only. Writes to this address range will result in a CP Bus Timeout and a machine check.

10.3 Programming the SGEC
The operation of the SGEC is controlled by a program in host memory called the port driver.
The SGEC and the port driver communicate through two data structures: Network Interface
Command and Status Registers (NICSRs) located in the SGEC and mapped in the host
I/O address space, and through descriptor lists and data buffers, , collectively called Host
Communication Area , in host memory.
The NICSRs are used for initialization, global pointers, commands and global errors reporting,
while the host memory resident structures handle the actions and statuses related to buffer
management.
The SGEC can be viewed as two independent, concurrently executing processes: Reception
and Transmission. After the SGEC completes its Initialization sequence, these two processes
alternate between three states: STOPPED, RUNNING or SUSPENDED. State transitions
occur as a result of port driver commands (writing to a NICSR) or various external events
occurrences. Some of the port driver commands require the referenced process to be in a
specific state.
A simple programming sequence of the chip may be summarized as:
1. After power on (or reset), verifying the self test completed successfully.
2. Writing NICSRs to set major parameters such as System Base Register, Interrupt Vector,
Address Filtering mode and so on.

KA660 Network Interface 10–3

3. Creating the transmit and receive lists in memory and writing the NICSRs to identify
them to the SGEC.
4. Placing a setup frame in the transmit list, to load the internal reception address filtering
table.
5. Starting the Reception and Transmission processes placing them in the RUNNING state.
6. Waiting for SGEC interrupts. NICSR5 contains all the global interrupt status bits.
7. Remedying the suspension cause, if either of the Reception or Transmission processes
enter the SUSPENDED state.
8. Issuing a Tx Poll Demand command, to return the Transmission process to the RUNNING
state. In addition to remedy the Reception process suspension cause, a Rx Poll Demand
could be issued to return the Reception process to the RUNNING state.
If the Rx poll demand is not issued, the Reception process will return to the RUNNING
state when the SGEC receives the next recognized incoming frame.
The following sections contain detailed programming and state transitions information.

10.3.1

Command and Status Registers

The SGEC contains 16 command and status registers which may be accessed by the host.

10.3.2

Host access to NICSRs

The SGEC’s NICSRs are located in VAX I/O address space.
The NICSRs must be longword aligned and can only be accessed using longword instructions. The address of NICSRx is the base address plus 4x bytes. For example, if the base
address is 2000 8000, then the address of NICSR2 is 2000 8008. In the following paragraphs,
NICSRs bits are specified with several access modes. The different access modes for bits are
as follows:
Table 10–1: Bit access modes
Bit marked

Meaning

Reserved for future expansion - Ignored on Write, Read as ‘‘0’’

Reserved for future expansion - Ignored on Write, Read as ‘‘1’’

Read only, ignored on Write

R/W

Read or Write

Write only, unpredictable on Read

R/W1

Read, or Clear by writing a ‘‘1’’. Writing with a ‘‘0’’ has no effect.

In order to save chip real estate, yet not tie up the host bus for extended periods of time, the
16 NICSRs are subdivided into two groups:
1. Physical NICSRs - 0 through 7, 15.
2. Virtual NICSRs - 8 through 14.

10–4

KA660 Network Interface

The group the NICSR is part of, determines the way the host will access it.

10.3.2.1

Physical NICSRs

These registers are physically present in the chip. Host access to these NICSRs is by a single
instruction (For example, MOVL). There is no host perceivable delay and the instruction
completes immediately. Most commonly used SGEC features are contained in the physical
NICSRs.
10.3.2.2

Virtual NICSRs

These registers are not physically present in the SGEC and are incarnated by the on-chip
processor. Accesses to SGEC functions implied by these registers may take up to 20 µseconds.
So as not to tie up the host bus, virtual NICSR access requires several steps by the host.
NICSR5<DN> is used to synchronize access to the virtual NICSRs: after the first virtual
NICSR access, the SGEC de-asserts NICSR5<DN> until it will complete the action.
NOTE
Accessing the virtual NICSRs, without polling first on the NICSR5<DN> reassertion
will cause unpredictable results.
10.3.2.2.1

NICSR write

To write to a virtual NICSR the host takes the following actions:
1. Issue a write NICSR instruction. Instruction completes immediately, but the data is not
yet copied by the SGEC.
2. Wait for NICSR5<DN>. No SGEC virtual NICSR may be accessed before NICSR5<DN>
asserts.
10.3.2.2.2

NICSR read

To read a virtual NICSR the host takes the following actions:
1. Issue a read NICSR instruction. Instruction completes immediately, but no valid data is
sent to the host.
2. Wait for NICSR5<DN>. No SGEC virtual NICSR may be accessed before NICSR5<DN>
asserts.
3. Reissue a read NICSR instruction, to the same NICSR as in step 1. The host receives valid
data.

KA660 Network Interface 10–5

10.3.3

Vector Address, IPL, Sync/Asynch (NICSR0)

Because the SGEC may generate an interrupt, on parity errors, during host writes to NICSR’s,
this register must be the first one written by the host. The format is shown in Figure 10–2
and the bit description is given in Table 10–2.
Figure 10–2: Vector Address, IPL, Sync/Asynch (NICSR0)

3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1
1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0
IP

Must Be One

− Interrupt Vector

1 1

SA
I/O Address: 2000 8000
(16)
Longword Read/Write Access
ESB90P0052

NOTE
A parity error during NICSR0 host write may cause a host system crash due to an
erroneous Interrupt Vector. To protect against such an eventuality, NICSR0 must be
written as follows while the IPL - to which the SGEC is assigned - is disabled:
1.
2.
3.
4.

10–6

Write NICSR0.
Read NICSR0.
Compare value read to value written. If values mismatch, repeat from step 1.
Read NICSR5 and examine NICSR5<ME> for pending parity interrupt. Should
an interrupt be pending, write NICSR5 to clear it.

KA660 Network Interface

Table 10–2: NICSR0 bits
Bit

Name

Access

Description

<31:30>

R/W

Interrupt Priority - is the VAX interrupt priority
level that the SGEC will respond to.
IP

IPL (hex)

Although the SGEC has only one interrupt
request pin, that pin might be wired to any of
the four IRQ pins on the host. The value in IP
should be 1416 for the KA690.
<29>

R/W

Sync/Asynch - This bit determines the SGEC operating mode when it is the bus master. When
set, the SGEC will operate as a synchronous
device and when clear, the SGEC will operate
as an asynchronous device.

<15:00>

R/W

Interrupt Vector - During an Interrupt
Acknowledge cycle for an SGEC interrupt,
this is the value that the SGEC will drive on
the host bus CDAL<31:0> pins (CDAL pins
<1:0> and <31:16> are set to ‘‘0’’). Bits <1:0>
are ignored when NICSR0 is written, and set to
‘‘1’’ when read.

Table 10–3: NICSR0 access
Value after RESET:

1FFF0003 hex

Read access rules:

None

Write access rules:

The IPL to which the SGEC is assigned - must be
DISABLED

10.3.4

Transmit Polling Demand (NICSR1)

The Polling Demand NICSR (NICSR1) is used by the port driver to tell the SGEC that it put a
packet on the transmit or receive list. The format is shown in Figure 10–3 and the bit description
is in Table 10–4.

KA660 Network Interface 10–7

Figure 10–3: Polling Demand (NICSR1)

3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1
1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0
Must Be One
PD
I/O Address: 2000

8004

(16)
Longword Write Only Access
ESB90P0053

Table 10–4: NICSR1 bits
Bit

Name

<31:01>

MBZ

<00>

Access

Description
Must be one. This field is reserved for future
expansion. Write as ONE.

Tx Polling Demand - Checks the transmit list
for frames to be transmitted.
The PD value is meaningless.

Table 10–5: NICSR1 access
Value after RESET:

not applicable

Read access rules:

None

Write access rules:

Tx process SUSPENDED

10.3.5

10–8

Receive Polling Demand (NICSR2)

KA660 Network Interface

Figure 10–4: NICSR2 format

3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1
1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0
Must Be One
PD
I/O Address: 2000

8008
(16)
ESB90P0054

Table 10–6: NICSR2 bits
Bit

Name

<31:01>

MBO

<00>

Access

Description
Must be one. This field is reserved for future
expansion. Write as ONE.

Rx Polling Demand - Checks the receive list
for receive descriptors to be acquired. The PD
value is meaningless.

Table 10–7: NICSR2 access
Value after RESET:

not applicable

Read access rules:

None

Write access rules:

Rx process SUSPENDED

10.3.6

Descriptor List Addresses (NICSR3, NICSR4)

The two descriptor list address registers are identical in function, one being used for the
transmit buffer descriptors and one being used for the receive buffer descriptors. In both
cases, the registers are used to point the SGEC to the start of the appropriate buffer descriptor
list.
The descriptor lists reside in VAX physical memory space and must be longword aligned.
NOTE
For best performance, it is recommended that the descriptor lists be OCTAWORD
aligned.

KA660 Network Interface 10–9

TRANSMIT LIST
If the Transmit descriptor list is built as a ring (the chain descriptor points at the
first descriptor of the list), the ring must contain, at least, TWO descriptors in
addition to the chain descriptor.
Initially, these registers must be written before the respective Start command is given
(see Section 10.3.8), else the respective process will remain in the STOPPED state. New
list head addresses are only acceptable while the respective process is in the STOPPED or
SUSPENDED states. Addresses written while the respective process is in the RUNNING
state, are ignored and discarded.
If the host attempts to read any of these registers before ever writing to them, the SGEC
responds with unpredictable values.
Figure 10–5: Descriptor list addresses format

3 3
1 0
MBZ

2 1 0
Start of Receive List − RBA

I/O Address: 2000

MBZ

NICSR3

800C
(16)

Longword Read/Write Access

3 3
1 0
MBZ

2 1 0
Start of Transmit List − TBA

MBZ

NICSR4

8010

I/O Address: 2000

(16)
Longword Read/Write Access
ESB90P0055

Table 10–8: Descriptor lists addresses bits
Bit

Name

<31:30>

MBZ

10–10

KA660 Network Interface

Access

Description
Must be ZERO. Ignored on writes, read as
ZERO

Table 10–8 (Cont.): Descriptor lists addresses bits
Bit

Name

Access

Description

<29:00>

RBA or TBA

R/W

Address of the start of the receive list (NICSR3)
or transmit list (NICSR4).This is a 30-bit VAX
physical address.

NOTE
The descriptor lists must be longword aligned.
Table 10–9: NICSR3 access
Value after RESET:

unpredictable

Read access rules:

None

Write access rules:

Rx process STOPPED or SUSPENDED

Table 10–10: NICSR4 access
Value after RESET:

unpredictable

Read access rules:

None

Write access rules:

Tx process STOPPED or SUSPENDED

After either of NICSR3 or NICSR4 are written, the new address is readable from the written
NICSR. However, if the SGEC status did not match the related write access rules, the new
address does not take effect and the written information is lost, EVEN if the SGEC matches
later the right condition.

10.3.7

Status Register (NICSR5)

This register contains all the status bits the SGEC reports to the host. Figure 10–6 shows the
register format and Table 10–11 describes the register bits.

KA660 Network Interface 10–11

Figure 10–6: NICSR5 bits
3 3 2
1 0 9

2 2 2 2 2 2
6 5 4 3 2 1
SS

1 1 1 1 1
9 8 7 6 5

MBO

7 6 5 4 3 2 1 0
MUST BE ONE
IS
TI
RI
RU
ME
RW
TW
BO
DN
SF
ID

I/O Address: 2000

8014
(16)

Longword Access with:
Bits <31:16> Read Only
Read/Write One to Clear
Bits <16:0>
ESB90P0056

Table 10–11: NICSR5 bits
Bit

Name

Access

Description

<31>

Initialization Done - When set, indicates the
SGEC has completed the Initialization (reset
and self test) sequences, and is ready for further
commands. When clear, indicates the SGEC
is performing the Initialization sequence and
ignoring all commands. After the Initialization
sequence completes, the Transmission and
Reception processes are in the STOPPED state.

<30>

Self test Failed - When set, indicates the SGEC
self test has failed. The self test completion
code bits indicate the failure type.

10–12

KA660 Network Interface

Table 10–11 (Cont.): NICSR5 bits
Bit

Name

Access

Description

<29:26>

Self test Status - The self test completion code
according to the following table. Only valid if
SF is set.
Value

Meaning

0001

ROM error

0010

RAM error

0011

Address filter RAM error

0100

Transmit FIFO error

0101

Receive FIFO error

0110

Self_test loopback error

INFO
Self test takes 25ms to
complete after Hardware or
Software RESET.
<25:24>

Transmission process state - Indicates the
current state of the Transmission process, as
follows:
Value

Meaning

STOPPED

RUNNING

SUSPENDED

Section 10.3.19.5 explains the Transmission
process operation and state transitions.
<23:22>

Reception process State - Indicates the current
state of the Reception process, as follows:
Value

Meaning

STOPPED

RUNNING

SUSPENDED

Section 10.3.19.4 explains the Reception process
operation and state transitions.

KA660 Network Interface 10–13

Table 10–11 (Cont.): NICSR5 bits
Bit

Name

Access

Description

<18:17>

Operating Mode - These bits indicate the current SGEC operating mode as in the following
table:
Value

Meaning

Normal operating mode.

Internal Loopback - Indicates
the SGEC is disengaged from
the Ethernet wire. Frames
from the transmit list are
looped back to the receive list,
subject to address filtering.
Section 10.3.19.6 explains this
mode of operation.

External Loopback - Indicates
the SGEC is working in full
duplex mode. Frames from the
transmit list are transmitted
on the Ethernet wire and also
looped back to the receive list,
subject to address filtering.
Section 10.3.19.6 explains this
mode of operation.

Reserved for Diagnostics

<16>

Done - When set, indicates the SGEC has
completed a requested virtual NICSR access.
After a reset, this bit is set.

<15:8>

MBO

<7>

R/W1

Boot_Message - When set, indicates that the
SGEC has detected a boot_message on the
serial line and has set the external pin BOOT_
L.

<6>

R/W1

Transmit Watchdog Timer interrupt - When
set, indicates the transmit watchdog timer has
timed out, indicating the SGEC transmitter
was babbling. The Transmission process is
aborted and placed in the STOPPED state.
(Also reported into the Tx descriptor status
TDES0<TO> flag).

10–14

KA660 Network Interface

Must Be One . This field is reserved. Writes are
ignored, read as ONE.

Table 10–11 (Cont.): NICSR5 bits
Bit

Name

Access

Description

<5>

R/W1

Receive Watchdog Timer interrupt - When set,
indicates the Receive Watchdog Timer has timed
out, indicating that some other node is babbling
on the network. Current frame reception is
aborted and RDES0<LE> and RDES0<LS>
will be set. Bit NICSR5<RI> will also set. The
Reception process remains in the RUNNING
state.

<4>

R/W1

Memory Error - Is set when any of the followings occur:
•

•

SGEC is the CP-BUS Master and the
ERR_L pin is asserted by external logic
(generally indicative of a memory problem).
Parity error detected on an host to SGEC
NICSR write or SGEC read from memory.

When a Memory Error is set, the Reception and
Transmission processes are aborted and placed
in the STOPPED state.

NOTE
At this point, it is mandatory that the port driver
issue a Reset command and
rewrite all NICSRs.
<3>

R/W1

Receive buffer Unavailable - When set, indicates
that the next descriptor on the receive list is
owned by the host and could not be acquired
by the SGEC. The Reception process is placed
in the SUSPENDED state. Section 10.3.19.4
explains the Reception process state transitions.
Once set by the SGEC, this bit will not be set
again until the SGEC encounters a descriptor it
can not acquire. To resume processing receive
descriptors, the host must flip the ownership
bit of the descriptor and can issue the Rx Poll
Demand command. If no Rx poll demand is
issued, the Reception process resumes when the
next recognized incoming frame is received.

<2>

R/W1

Receive Interrupt - When set, indicates that
a frame has been placed on the receive list.
Frame specific status information was posted in
the descriptor. The Reception process remains
in the RUNNING state.

KA660 Network Interface 10–15

Table 10–11 (Cont.): NICSR5 bits
Bit

Name

Access

Description

<1>

R/W1

Transmit Interrupt - When set, indicates one of
the following:
•

•

<0>

R/W1

Either all the frames in the transmit list
have been transmitted (next descriptor
owned by the host), or a frame transmission
was aborted due to a locally induced error.
The port driver must scan down the list of
descriptors to determine the exact cause.
The Transmission process is placed in the
SUSPENDED state. Section 10.3.19.5
explains the Transmission process state
transitions. To resume processing transmit
descriptors, the port driver must issue the
Poll Demand command.
A frame transmission completed, and
TDES1<IC> was set. The Transmission
process remains in the RUNNING state,
unless the next descriptor is owned by the
host or the frame transmission aborted
due to an error. In the latter cases, the
Transmission process is placed in the
SUSPENDED state.

Interrupt Summary - The logical ‘‘OR’’ of
NICSR5 bits 1 through 6.

Table 10–13: NICSR5 access
Value after RESET:

0039FF00 hex

Read access rules:

None

Write access rules:

NICSR5<07:01> bits cleared by 1, others bits not writeable

10.3.7.1

NICSR5 status report

The Status register NICSR5 is split into two words:
- the high word which contains the global status of the SGEC, as the initialization status, the
DMA and operation mode and the Receive and Transmit process states.
-the low word which contains the status related to the Receive and Transmit frames.
Any change of the NICSR5 bits <ID>, <SF>, <OM> or <DN> - which is always the result of a
host command - is reported without an interrupt.
Any process state change initiated by a host command NICSR6<ST> or NICSR6<SR>, is
reported without an interrupt.

10–16

KA660 Network Interface

In the above two cases, the driver must poll on NICSR5 to get the acknowledge of its command (For example, polling on <ID, SF> after Reset or polling on <TS> after a START_TX
command).
Any process state change initiated by the SGEC activity is immediately followed by at least
one of the NICSR5<6:1> interrupts and the interrupt_summary NICSR5<IS>.
The SGEC 16 bit internal processor updates the 32 bits NICSR5 register in two phases: the
high word is modified first, then the low word is written, which generates an interrupt to the
host. In this case, the driver must scan first the NICSR5 low word to get the interrupt status,
then the NICSR high word to get the related process state. (For example, <TI> interrupt with
<TS> = SUSPENDED reports an end of transmission due to a Tx descriptor unavailable.)
If the host polls on the process state change, it may detect a change without interrupt, due to
the small time window separating the NICSR5 high word and low word updates.
Maximum time window is 4*Tcycles of the host clock.

10.3.8

Command and Mode Register (NICSR6)

This register is used to establish operating modes and for port driver commands.
Figure 10–7: NICSR6 format

3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1
1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0
R I r
E E

I/O Address: 2000

Must
be ONE

B S r
E E

Must
be ONE

S S OM D F r
T R
C C

r P
B

8018
(16)

Longword Read/Write Access
r = reserved
ESB90P0057

KA660 Network Interface 10–17

Table 10–14: NICSR6 bits
Bit

Name

Access

Description

<31>

R/W

Reset command - Upon being set, the SGEC will abort all processes
and start the reset sequence. After completing the reset and self test
sequence, the SGEC will set bit NICSR5<ID>. Clearing this bit has no
effect.

NOTE
The NICSR6<RE> value is unpredictable on read
after HARDWARE reset.
<30>

<29>

<28:25>

R/W

Interrupt Enable mode - When set, setting of NICSR5 bits 1 through 6
will cause an interrupt to be generated.

R/W

Burst Limit mode - Specifies the maximum number of longwords to be
transferred in a single DMA burst on the host bus.

reserved

When NICSR6<SE> is cleared, permissible values are 1,2,4,8 . When
SE is set, the only permissible values are 1 and 4: a value of 2 or 8 is
respectively forced to 1 or 4.
After Initialization, the burst limit is set to 1.
<24:21>

MBO

<20>

R/W

Boot_message Enable mode - When set, enables the boot_message
recognition. When the SGEC recognizes an incoming boot message on
the serial line, NICSR5<BO> is set and the external pin BOOT_L is
asserted for a duration of 6*Tcycles (of the host clock).

<19>

R/W

Single_cycle Enable mode - When set, the SGEC transfers only a single
longword or an octaword in a single DMA burst on the host bus.

R/W

Start/Stop Transmission command - When set, the Transmission process is placed the RUNNING state, the SGEC checks the transmit list
at the current position for a frame to transmit - the address set by
NICSR4 or the position retained when the Tx process was previously
stopped -. If it does not find a frame to transmit, the Transmission process enters the SUSPENDED state. The Start Transmission command
is honored only when the Transmission process is in the STOPPED
state. The first time this command is issued, an additional requirement
is that NICSR4 have already been written to, else the Transmission
process will remain in the STOPPED state.

<18:12>

MBO

<11>

This field is reserved Writes are ignored, read as one.

Must Be One. This field is reserved. Writes are ignored, read as ONE.

When cleared the Transmission process is placed in the STOPPED
state after completing transmission of the current frame. The next
descriptor position in the transmit list is saved, and becomes the
current position after transmission is restarted.
The Stop Transmission command is honored only when the Transmission
process is in the RUNNING or SUSPENDED states.
Refer to Section 10.3.19.5 for more information.

10–18

KA660 Network Interface

Table 10–14 (Cont.): NICSR6 bits
Bit

Name

Access

Description

<10>

R/W

Start/Stop Reception command - When set, the Reception process is
placed in the RUNNING state, the SGEC attempts to acquire a descriptor from the receive list and process incoming frames. Descriptor
acquisition is attempted from the current position in the list - the address set by NICSR3 or the position retained when the Rx process was
previously stopped -. If no descriptor can be acquired, the Reception
process enters the SUSPENDED state.
The Start Reception command is honored only when the Reception
process is in the STOPPED state. The first time this command is
issued, an additional requirement is that NICSR3 have already been
written to, else the Reception process will remain in the STOPPED state.
When cleared, the Reception process is placed in the STOPPED state,
after completing reception of the current frame. The next descriptor
position in the receive list is saved, and becomes the current position
after reception is restarted. The Stop Reception command is honored
only when the Reception process is in the RUNNING or SUSPENDED
states.
Refer to Section 10.3.19.4 for more information.

<9:8>

<7>

R/W

Operating Mode - These bits determine the SGEC main operating
mode.
Value

Meaning

Normal operating mode.

Internal Loopback - The SGEC will loopback buffers
from the transmit list. The data will be passed from
the transmit logic back to the receive logic. The receive
logic will treat the looped frame as it would any other
frame, and subject it to the address filtering and
validity check process.

External Loopback - The SGEC transmits normally
and in addition, will enable its receive logic to its own
transmissions. The receive logic will treat the looped
frame as it would any other frame, and subject it to the
address filtering and validity check process.

Reserved for Diagnostic

Disable data Chaining mode - When set, no data chaining will occur in
reception; frames, longer than the current receive buffer, will be truncated. RDES0<FS,LS> will always be set. The frame length returned
in RDES0<FL> will be the true length of the non-truncated frame while
RDES0<BO> will indicate that the frame has been truncated due to
buffer overflow.
When clear, frames too long for the current receive buffer, will be
transferred to the next buffer(s) in the receive list.

KA660 Network Interface 10–19

Table 10–14 (Cont.): NICSR6 bits
Bit

Name

Access

Description

<6>

R/W

Force Collision mode - This bit allows the collision logic to be tested.
The chip must be in internal loopback mode for FC to be valid. If FC
is set, a collision will be forced during the next transmission attempt.
This will result in 16 transmission attempts with Excessive Collision
reported in the transmit descriptor.

R/W

Pass Bad Frames mode - When this bit is set, the SGEC will pass
frames that have been damaged by collisions or are too short due
to premature reception termination. Both events should have occurred within the collision window (64 bytes), else other errors will be
reported.

<5:4>

MBO

<3>

Must Be ONE. This field is reserved. Writes are ignored, read as ONE.

When clear, these frames will be discarded and never show up in the
host receive buffers.

NOTE
Pass Bad Frames is subject the address filtering
mode. For example, to monitor the network, this
mode must be set together with the promiscuous
address filtering mode.
<2:1>

<0>

10–20

R/W

MBO

KA660 Network Interface

Address Filtering mode - These bits define the way incoming frames
will be address filtered:
Value

Meaning

Normal - Incoming frames will be filtered according
to the values of the <HP> and <IF> bits of the setup
frame descriptor.

Promiscuous - All incoming frames will be passed to
the host, regardless of the <HP> bit value.

All Multicast - All incoming frames with Multicast address destinations will be passed to the host. Incoming
frames with physical address destinations will be
filtered according to the <HP> bit value.

Unused - Reserved.

Must Be ONE. This field is reserved. Writes are ignored, read as ONE.

Table 10–15: NICSR6 access
Value after RESET:

83E0F000 hex or 03E0F000 hex

Read access rules:

None

Write access rules:
* <RE, IE, BE>

Unconditional

* <BL, SE, OM>

Rx and Tx processes STOPPED

* <FC>

Rx and Tx processes STOPPED, Internal_Loopback
mode

* <DC, PB, AF>

Rx STOPPED

* Start_Receive <SR>=1

Rx STOPPED and NICSR3 Initialized

* Start_Transmit <ST>=1

Tx STOPPED and NICSR4 Initialized

* Stop_Receive <SR>=0

Rx RUNNING or SUSPENDED

* Stop_Transmit <ST>=0

Tx RUNNING or SUSPENDED

After NICSR6 is written, the new value is readable from NICSR6. However, if the SGEC
status does not match the related write access rules, the new mode setting and command do
not take effect and the written information is lost, EVEN if the SGEC matches later the
right condition.

10.3.9

System Base Register (NICSR7)

This NICSR contains the physical starting address of the VAX System Page Table. This
register must be loaded by host software before any address translation occurs so that memory
will not be corrupted.
Figure 10–8: NICSR7 format

3 3 2
1 0 9

2 1 0

MBZ
I/O Address: 2000

System Base Address

MBZ

801C
(16)

Longword Read/Write Access
ESB90P0058

KA660 Network Interface 10–21

Table 10–16: NICSR7 bits
Bit

Name

<31:30>

MBZ

<29:00>

Access

Description
Must Be ZERO. Read as zero. Writes are
ignored.

R/W

System Base address - The physical starting
address of the VAX System Page Table. Not
used if VA (Virtual Addressing) is cleared in all
descriptors.
This register should be loaded only once
after a reset. Subsequent modifications of
this register at any other time may cause
unpredictable results.

Table 10–17: NICSR7 access
Value after RESET:

unpredictable

Read access rules:

None

Write access rules:

Writing once after initialization

10.3.10 Reserved register (NICSR8)
This entire register is reserved.

10.3.11 Watchdog Timers (NICSR9)
The SGEC has two timers that restrict the length of time in which the chip can receive or
transmit.

10–22

KA660 Network Interface

Figure 10–9: NICSR9 format

3
1

1 1
6 5

RECEIVE TIME−OUT − RT

TRANSMIT TIME−OUT − TT

I/O Address: 2000 8024
(16)
Longword Read/Write Access
ESB90P0059

Table 10–18: NICSR9 bits
Bit

Name

Access

Description

<31:16>

R/W

RECEIVE WATCHDOG TIME-OUT - The
Receive Watchdog Timer protects the host CPU
against babbling transmitters on the network.
If the receiver stays on for  cycles of the
serial clock, the SGEC will cut off reception and
set the NICSR5<RW> bit. If the timer is set
to zero, it will never time-out. The value of RT
is an unsigned integer. With a 10 Mhz serial
clock, this provides a range of 72µs to 100ms.
The default value is 1250 corresponding to 2ms.
The Rx watchdog timer is programmed only
while the Reception process is in the STOPPED
state.

NOTE
A Rx watchdog value between 1 and 44 is forced
to the minimum time_out
value of 45 (72µs).

KA660 Network Interface 10–23

Table 10–18 (Cont.): NICSR9 bits
Bit

Name

Access

Description

<15:00>

R/W

TRANSMIT WATCHDOG TIME-OUT - The
Transmit Watchdog Timer protects the network
against babbling SGEC transmissions, on top
of any such circuitry present in tranceivers.
If the transmitter stays on for      cycles
of the serial clock, the SGEC will cut off the
transmitter and set the NICSR5<TW> bit. If
the timer is set to zero, it will never time-out.
The value of TT is an unsigned integer. With
a 10 Mhz serial clock, this provides a range
of 72µs to 100ms. The default value is 1250
corresponding to 2ms.
The Tx watchdog timer is programmed only
while the Transmission process is in the
STOPPED state.

NOTE
A Tx watchdog value between 1 and 44 is forced
to the minimum time_out
value of 45 (72µs).

Table 10–19: NICSR9 access
Value after RESET:

00000000 hex

Read access rules:

None

Write access rules:
* Rx Watchdog timer

Rx process STOPPED

* Tx Watchdog timer

Tx process STOPPED

These watchdog timers are enabled by default. These timers will assume the default values
after hardware or software resets.

10.3.12 Revision Number and Missed Frame Count (NICSR10)
This register contains a missed frame counter and SGEC identification information.

10–24

KA660 Network Interface

Figure 10–10: Revision Number and Missed Frame Count (VIRTUAL NICSR10)

3
1

2 1 1 1 1 1
0 9 8 7 6 5
MBZ

MFC

I/O Address: 2000 802C
(16)
Longword Read Only Access
ESB90P0060

Table 10–20: NICSR10 bits
Bit

Name

Access

Description

<31:21>

MBZ

<20:16>

Chip Revision Number - This stores the revision
number for this particular SGEC.

<15:00>

MFC

Missed Frame Count - Counter for the number
of frames that were discarded and lost because
host receive buffers were unavailable. The
counter is cleared when read by the host.

Must BE ZERO. Read as ZERO. Writes are
ignored.

Table 10–21: NICSR10 access
Value after RESET:

00030000 hex

Read access rules:

Missed_frame counter cleared by read

Write access rules:

Not applicable

10.3.13 Boot Message (NICSR11, 12, 13)
These registers contain the Boot message VERIFICATION and PROCESSOR fields. The
format is shown in Figure 10–11 and the bit descriptions are in Table 10–22.

KA660 Network Interface 10–25

Figure 10–11: Boot Message
3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1
1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0
NICSR11
20000802C
16

VERIFICATION VRF <31:00>
3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1
1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0

NICSR12
20008030
16

VERIFICATION VRF <63:32>
3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1
1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0
0|0|0|0|0|0|0|0|0|0|0|0|0

0|0|0

0|0|0|0|0|0|0|0| PROCESSOR PRC

NICSR13
20008034
16

Longword Read/Write Access
ESB90P0061

Table 10–22: NICSR11,12,13 bits
Bit

Name

Access

Description

NICSR11 <31:00>

VRF<31:00>

R/W

Boot message VERIFICATION field
<31:00>

NICSR12 <31:00>

VRF<63:32>

R/W

Boot message VERIFICATION field
<63:32>

NICSR13 <07:00>

PRC

R/W

Boot message PROCESSOR field -

NOTE
The least significant bit of the Verification Field (VRF<0>) corresponds to the first
incoming bit of the verification field in the serial boot message.
Table 10–23: NICSR11,12,13 access
Value after RESET:

00000000 hex for each of NICSR11,NICSR12,NICSR13

Read access rules:

None

Write access rules:

Boot message DISABLED (<NICSR6<BE> = 0)

10–26

KA660 Network Interface

10.3.14 Diagnostic Registers (NICSR14, 15)
These registers are reserved for diagnostic features.
10.3.14.1

Diagnostic Breakpoint Address Register (NICSR14)

This register is virtual CSR. It contains the breakpoint address that will cause the internal
CPU to jump to a patch address. Figure 10–12 shows the format of the register. Table 10–24
lists the bits and descriptions. This register can be loaded only in diagnostic mode (NICSR6
<OM>=<11>).
Figure 10–12: NICSR14 format

3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1
1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0
B
E

CODE RESTART ADDRESS
(CRA)

BREAKPOINT ADDRESS
(BPA)
ESB90P0062

Table 10–24: NICSR14 bits
Bit

Name

Type

Description

<31>

R/W

When set, Breakpoint enabled.

<30:16>

CRA

R/W

Code Restart Address - The first address in the
internal RAM where the internal processor will
jump to after a breakpoint occurred.

<15:0>

BPA

R/W

Breakpoint Address - The internal processor
address at which the program will halt and
jump to the RAM loaded code.

NOTE
This register in conjunction with the diagnostic descriptors allows software patches.

Table 10–25: NICSR14 access
Value after RESET:

0000000016

Read access rules:

None

Write access rules:

DIAGNOSTIC mode

Violation:

Addressing NICSR14 while NICSR5<DN> is deasserted

KA660 Network Interface 10–27

10.3.14.2

Monitor Command Register (NICSR15)

This register is a physical CSR. It contains the bits which will select the internal test block
operation mode. Figure 10–13 shows the format of the register. Table 10–26 lists the bits and
descriptions.
Figure 10–13: NICSR15 format

3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1
1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0
MBZ

S QAD B
S
T

ADDRESS/DATA

ESB90P0063

Table 10–26: NICSR15 bits
Bit

Name

Type

Description

<31:16>

ADDR/DATA

R/W

- Before the "Examine" cycle, it points to the
location to be read. Three cycles after the
assertion of <ST>, it contains the READ data.

<15>

Start Read - When set, starts the "Examine"
cycle: the data addressed by CSR<31:16> is
fetched and stored into the same register field.
Reset by hardware at the end of operation.

<14:13>

QAD

Quad selects bits- These bits define the specific
four bits of the internal Data_bus or Address_
bus which are monitored on the external test
pins BM_L/TEST<3:0>. Meaningful only in test
mode (TSM=1).
The 2 bit code is interpreted as follows.

10–28

KA660 Network Interface

QAD

data

address

<03:00>

<07:04>

<11:08>

<15:12>

0, IOP_WR_L,<13:12>

Table 10–26 (Cont.): NICSR15 bits
Bit

Name

Type

Description

<12>

Bus select- When reset the internal Data_bus
is monitored on the external test pins BM_L
/TEST<3:0>. When set, the monitoring is applied on the internal Address_bus. Meaningful
only in test mode (TSM=1).

<11:0>

MBZ

Must be zero.

Table 10–27: NICSR15 access
Value after RESET:

00000FFF16

Read access rules:

None

Write access rules:

Reserved for DEBUGGING

Violation:

Setting <ST> with "random" SGEC internal address

10.3.15 Descriptors and buffers format
The SGEC transfers frame data to and from receive and transmit buffers in host memory.
These buffers are pointed to by descriptors which are also resident in host memory.
There are two descriptor lists: one for receive and one for transmit. The starting address of
each list is written into NICSRs 3 and 4 respectively. A descriptor list is a forward-linked
(either implicitly or explicitly) list of descriptors, the last of which may point back to the
first entry, thus creating a ring structure. Explicit chaining of descriptors, through setting
xDES1<CA> is called Descriptor Chaining. The descriptor lists reside in VAX physical memory
address space.
NOTE
The SGEC first reads the descriptors, ignoring all unused bits regardless of their
state. The only word the SGEC writes back, is the first word (xDES0) of each
descriptor. Unused bits in xDES0 will be written as ‘‘0’’. Unused bits in xDES1 xDES3 may be used by the port driver and the SGEC will never disturb them.
A data buffer can contain an entire frame or part of a frame, but it cannot contain more than
a single frame. Buffers contain only data; buffer status is contained in the descriptor. The
term Data Chaining is used to refer to frames spanning multiple data buffers. Data Chaining
can be enabled or disabled, in reception, through NICSR6<DC>. Data buffers reside in VAX
memory space, either physical or virtual.
NOTE
The virtual to physical address translation is based on the assumption that PTEs
are locked in the host memory the time the SGEC owns the related buffer.

KA660 Network Interface 10–29

NOTE
For best performance in virtual addressing mode, PPTE vectors must not cross a
page of the PPTE table.

10.3.16 Receive Descriptors
The receive descriptor format is shown in Figure 10–14, and described in the following paragraphs.
Figure 10–14: Receive descriptor format

3
1

3
0

2
9

O
W
C
A

2
8

2
7

2
6

2
5

2
4

2
3

2
2

Frame Length
V
A

2
1

2
0

1
9

1
8

1
7

1
6

1
5

1
4

1
3

E
S

L
E

V
T

1
2

1
1

1
0

R
F

B
O

F
S

L
S

T
L

C
S

F
T

T
N

D
B

C
E

O
F

RDES0
RDES1

u
RDES2
Buffer Size

Page Offset
u

RDES3

BUFFER SVAPTE/Physical Address

0 − SGEC writes as ZERO
u − Ignored by the SGEC on read, never written
ESB90P0064

10.3.16.1

RDES0 word

RDES0 word contains received frame status, length and descriptor ownership information.

10–30

KA660 Network Interface

Table 10–28: RDES0 bits
Bit

Name

Description

<31>

Own bit - When set, indicates the descriptor is owned by the SGEC.
When cleared, indicates the descriptor is owned by the host. The
SGEC clears this bit upon completing processing of the descriptor
and its associated buffer.

<30:16>

Frame Length - The length in bytes of the received frame.
Meaningless if RDES0<LE> is set.

<15>

Error Summary - The logical ‘‘OR’’ of RDES0 bits OF,CE,TN,CS,TL,LE,RF.

<14>

Length Error - When set, indicates a frame truncation caused by one
of the following:
•
•

<13:12>

The frame segment does not fit within the current buffer and the
SGEC does not own the next descriptor. The frame is truncated.
The Receive Watchdog timer expired. NICSR5<RW> is also set.

Data Type - Indicates the type of frame the buffer contains, according
to the following table:
Value

Meaning

Serial received frame

Internally looped back frame

Externally looped back frame , Serial received frame

<11>

Runt Frame - When set, indicates this frame was damaged by
a collision or premature termination before the collision window
had passed. Runt frames will only be passed on to the host if
(NICSR6<PB>) is set. Meaningless if RDES0<OF> is set.

<10>

Buffer overflow - When set, indicates that the frame has been
truncated due to a buffer too small to fit the frame size. This bit may
only be set if Data Chaining is disabled (NICSR6<DC> = 1).

<09>

First Segment - When set, indicates this buffer contains the first
segment of a frame.

<08>

Last Segment - When set, indicates this buffer contains the last
segment of a frame and status information is valid.

<07>

Frame Too Long - When set, indicates the frame length exceeds the
maximum Ethernet specified size of 1518 bytes.

NOTE
Frame Too Long is only a frame length indication and does not cause any frame truncation.
<06>

Collision Seen - When set, indicates the frame was damaged by a
collision that occurred after the 64 bytes following the SFD.

KA660 Network Interface 10–31

Table 10–28 (Cont.): RDES0 bits
Bit

Name

Description

<05>

Frame Type - When set, indicates the frame is an Ethernet type
frame (Frame Length_Field > 1500). When clear, indicates the frame
is an IEEE 802.3 type frame. Meaningless for Runt frames < 14
bytes.

<04>

Zero. SGEC writes as ZERO.

<03>

Translation Not Valid - When set, indicates that a translation error
occurred when the SGEC was translating a VAX virtual buffer
address. It will only set if RDES1<VA> was set. The Reception
process remains in the RUNNING state and attempts to acquire the
next descriptor.

<02>

Dribbling Bits - When set, indicates the frame contained a noninteger multiple of eight bits. This error will be reported only if
the number of dribbling bits in the last byte is greater than two.
Meaningless if RDES0<CS> or RDES0<RF> are set.
The CRC check is performed independent of this error, however,
only whole bytes are run through the CRC logic. Consequently,
received frames with up to six dribbling bits will have this bit
set, but if <CE> (or another error indicator) is not set, these
frames should be considered valid:
CE

Error

None

CRC error

Alignment error

<01>

CRC Error - When set, indicates that a CRC error has occurred on
the received frame.

<00>

Overflow - When set, indicates received data in this descriptor’s
buffer was corrupted due to internal FIFO overflow. This will generally occur if SGEC DMA requests are not granted before the internal
receive FIFO fills up.

10.3.16.2

10–32

RDES1 word

KA660 Network Interface

Table 10–29: RDES1 bits
Bit

Name

Descriptor

<31>

Chain Address - When set, RDES3 is interpreted as another descriptor’s VAX physical address. This allows the SGEC to process
multiple, non-contiguous descriptor lists and explicitly "chain" the
lists. Note that contiguous descriptors are implicitly chained.
In contrast to what is done for a Rx buffer descriptor, the SGEC
clears neither the ownership bit RDES0<OW> nor one of the other
bits of RDES0 of the chain descriptor after processing.
To protect against infinite loop, a chain descriptor pointing back
itself, is seen as owned by the host, regardless of the ownership bit
state.

<30>

Virtual Addressing - When set, RDES3 is interpreted as a virtual
address. The type of virtual address translation is determined by the
RDES1<VT> bit. The SGEC uses RDES3 and RDES2<Page Offset>
to perform a VAX virtual address translation process to obtain the
physical address of the buffer. When clear, RDES3 is interpreted as
the actual physical address of the buffer:
VA

Addressing mode

Physical

Virtual - SVAPTE

Virtual - PAPTE

<29>

Virtual Type - In case of virtual addressing (RDES1<VA> = 1),
indicates the type of virtual address translation. When set, the
buffer address RDES3 is interpreted as a SVAPTE (System Virtual
Address of the Page Table Entry). When clear, the buffer address is
interpreted as a PAPTE (Physical Address of the Page Table Entry).
Meaningful only if RDES1<VA> is set.

<28:0>

Unused. Ignored by the SGEC on reads. Never written.

10.3.16.3

RDES2 word

Table 10–30: RDES2 bits
Bit

Name

Descriptor

<31>

Unused. Ignored by the SGEC on reads. Never written.

<30:16>

Buffer Size - The size, in bytes, of the data buffer.

NOTE
Receive buffers size must be an EVEN number
of bytes.

KA660 Network Interface 10–33

Table 10–30 (Cont.): RDES2 bits
Bit

Name

Descriptor

<15:9>

Unused. Ignored by the SGEC on reads. Never written.

<08:00>

Page Offset - The byte offset of the buffer within the page. Only
meaningful if RDES1<VA> is set.

NOTE
Receive buffers must be word aligned.

10.3.16.4

RDES3 word

Table 10–31: RDES3 bits
Bit

Name

Descriptor

<31:00>

SV/PV/PA

SVAPTE/PAPTE/Physical Address - When RDES1<VA> is set,
RDES3 is interpreted as the address of the Page Table Entry and
used in the virtual address translation process. The type of the
address System Virtual address (SVAPTE) or Physical Address
(PAPTE) is determined by RDES1<VT>. When RDES1<VA> is
clear, RDES3 is interpreted as the physical address of the buffer.
When RDES1<CA> is set, RDES3 is interpreted as the VAX physical
address of another descriptor.

NOTE
Receive buffers must be word aligned.

10.3.16.5

Receive descriptor status validity

The following table summarizes the validity of the Receive descriptor status bits regarding the
Reception completion status:

10–34

KA660 Network Interface

Table 10–32: Receive descriptor status validity
Reception

Rx Status report

status

(ES,LE,BO,DT,FS,LS,FL,TN,OF)

Overflow

Collision after 512 bits

Runt frame

Runt frame < 14 bytes

Watchdog timeout

V - Valid
X - Meaningless

10.3.17 Transmit descriptors
The transmit descriptor format is shown in Figure 10–15, and described in the following
paragraphs.

KA660 Network Interface 10–35

Figure 10–15: Transmit descriptor format

3
1

3
0

2
9

2
8

2
7

2
6

O
W
C
A

2
5

2
4

2
3

2
2

TDR
V
A

A
C

F
S

I
C

L
S

2
1

2
0

1
9

1
8

1
7

1
6

1
5

1
4

1
3

1
2

1
1

1
0

E
S

T
O

L
E

L
O

N
C

L
C

E
C

H
F

Coll.
Count

T
N

U
F

D
E

TDES0
TDES1

V
T

u
TDES2

Buffer Size

Page Offset
TDES3

BUFFER SVAPTE/Physical Address
0 − SGEC writes as "0"
u − Ignored by the SGEC on read, never written
ESB90P0065

10.3.17.1

TDES0 word

TDES0 word contains transmitted frame status and descriptor ownership information.
Table 10–33: TDES0 bits
Bit

Name

Description

<31>

Own bit - When set, indicates the descriptor is owned by the SGEC. When cleared,
indicates the descriptor is owned by the host. The SGEC clears this bit upon completing
processing of the descriptor and its associated buffer.

<29:16>

TDR

Time Domain Reflectometer - This is a count of bit time and is useful for locating a fault
on the cable using the velocity of propagation on the cable. Only valid if TDES0<EC>
is also set. Two Excessive Collisions in a row and with the same or similar (within 20)
TDR values indicate a possible cable open.

<15>

Error Summary - The logical ‘‘OR’’ of UF, TN, EC, LC, NC, LO, LE and TO.

<14>

Transmit Watchdog Timeout - When set, indicates the transmit watchdog timer has
timed out, indicating the SGEC transmitter was babbling. The interrupt NICSR5<TW>
is set and the Transmission process is aborted and placed in the STOPPED state.

<13>

MBZ

Must be ZERO. SGEC writes as ZERO

10–36

KA660 Network Interface

Table 10–33 (Cont.): TDES0 bits
Bit

Name

Description

<12>

Length Error - When set, indicates one of the following:
•
•
•

Descriptor unavailable (owned by the host) in the middle of data chained descriptors.
Zero length buffer in the middle of data chained descriptors.
Setup or Diagnostic descriptors (Data type TDES1<DT> <> 0) in the middle of data
chained descriptors.
• Incorrect order of first_segment TDES1<FS> and last_segment TDES1<LS> descriptors in the descriptor list.
The Transmission process enters the SUSPENDED state and sets NICSR5<TI>.
<11>

Loss of Carrier - When set, indicates loss of carrier during transmission (possible short
circuit in the Ethernet cable).

<10>

No Carrier - When set, indicates the carrier signal from the transceiver was not present
during transmission (possible problem in the transceiver or transceiver cable).

<09>

Late Collision - When set, indicates frame transmission was aborted due to a late
collision. Meaningless if TDES0<UF>.

<08>

Excessive Collisions - When set, indicates that the transmission was aborted because 16
successive collisions occurred while attempting to transmit the current frame.

<07>

Heartbeat Fail - When set, indicates Heartbeat Collision Check failure (the transceiver
failed to return a collision pulse as a check after the transmission. Some tranceivers
do not generate heartbeat, and so will always have this bit set. If the transceiver does
support it, it indicates transceiver failure.) Meaningless if TDES0<UF>.

<06:03>

Collision Count - A four bit counter indicating the number of collisions that occurred
before the transmission attempt succeeded or failed. Meaningless when TDES0<EC> is
also set.

<02>

Translation Not Valid - When set, indicates that a translation error occurred when the
SGEC was translating a VAX virtual buffer address. It may only set if TDES1<VA> was
set. The Transmission process enters the SUSPENDED state and sets NICSR5<TI>.

<01>

Underflow Error - When set, indicates that the transmitter has truncated a message due
to data late from memory. UF indicates that the SGEC encountered an empty transmit
FIFO while in the midst of transmitting a frame. The Transmission process enters the
SUSPENDED state and sets NICSR5<TI>.

<00>

Deferred - When set, indicates that the SGEC had to defer while trying to transmit a
frame. This condition occurs if the channel is busy when the SGEC is ready to transmit.

10.3.17.2

TDES1 word

Meaningless in internal loopback mode (NICSR5<OM>=1).

KA660 Network Interface 10–37

Table 10–34: TDES1 bits
Bit

Name

Descriptor

<31>

Chain Address - When set, TDES3 is interpreted as another descriptor’s VAX physical address. This allows the SGEC to process
multiple, non-contiguous descriptor lists and explicitly "chain" the
lists. Note that contiguous descriptors are implicitly chained.
In contrast to what is done for a Rx buffer descriptor, the SGEC
clears neither the ownership bit TDES0<OW> or one of the other
bits of TDES0 of the chain descriptor after processing.
To protect against infinite loop, a chain descriptor pointing back to
itself, is seen as owned by the host, regardless of the ownership bit
state.

<30>

<29:28>

<27>

Virtual Addressing - When set, TDES3 is interpreted as a virtual
address. The type of virtual address translation is determined by the
TDES1<VT> bit. The SGEC uses TDES3 and TDES2<Page Offset>
to perform a VAX virtual address translation process to obtain the
physical address of the buffer. When clear, TDES3 is interpreted as
the actual physical address of the buffer:
VA

Addressing mode

Physical

Virtual - SVAPTE

Virtual - PAPTE

Data Type - Indicates the type of data the buffer contains, according
to the following table:
Value

Meaning

Normal transmit frame data

Setup frame - Explained in Section 10.3.18.

Diagnostic frame

Add CRC disable - When set, the SGEC will not append the CRC to
the end of the transmitted frame. To take effect, this bit must be set
in the descriptor where FS is set.

NOTE
If the transmitted frame is shorter than 64
bytes, the SGEC will add the padding field and
the CRC regardless the <AC> flag.
<26>

First Segment - When set, indicates the buffer contains the first
segment of a frame.

<25>

Last Segment - When set, indicates the buffer contains the last
segment of a frame.

10–38

KA660 Network Interface

Table 10–34 (Cont.): TDES1 bits
Bit

Name

Descriptor

<24>

Interrupt on Completion - When set, the SGEC will set NICSR5<TI>
after this frame has been transmitted. To take effect, this bit must
be set in the descriptor where LS is set.

<23>

Virtual Type - In case of virtual addressing (TDES1<VA> = 1),
indicates the type of virtual address translation. When set, the
buffer address TDES3 is interpreted as a SVAPTE (System Virtual
Address of the Page Table Entry). When clear, the buffer address is
interpreted as a PAPTE (Physical Address of the Page Table Entry).
Meaningful only if TDES1<VA> is set.

<22:0>

Unused. Ignored by the SGEC on reads. Never written.

10.3.17.3

TDES2 word

Table 10–35: TDES2 bits
Bit

Name

Descriptor

<31>

Unused. Ignored by the SGEC on reads. Never written.

<30:16>

Buffer Size - The size, in bytes, of the data buffer. If this field is
0, the SGEC will skip over this buffer and ignore it. The frame
size is the sum of all BS fields of the frame segments (between and
including the descriptors having TDES1<FS> and TDES1<LS> set.)

NOTE
If the port driver wishes to suppress transmission of a frame, this field must be set to
0 in all descriptors comprising the frame and
prior to the SGEC acquiring them. If this rule
is not adhered to, corrupted frames might be
transmitted.
<08:00>

Page Offset - The byte offset of the buffer within the page. Only
meaningful if TDES1<VA> is set.

NOTE
Transmit buffers may start on arbitrary byte
boundaries.

KA660 Network Interface 10–39

10.3.17.4

TDES3 word

Table 10–36: TDES3 bits
Bit

Name

Descriptor

<31:00>

SV/PV/PA

SVAPTE/PAPTE/Physical Address - When TDES1<VA> is set, TDES3
is interpreted as the address of the Page Table Entry and used in
the virtual address translation process. The type of the address
System Virtual address (SVAPTE) or Physical Address (PAPTE) is
determined by TDES1<VT>. When TDES1<VA> is clear, TDES3 is
interpreted as the physical address of the buffer. When TDES1<CA>
is set, TDES3 is interpreted as the VAX physical address of another
descriptor.

NOTE
Transmit buffers may start on arbitrary byte
boundaries.

10.3.17.5

Transmit descriptor status validity

Table 10–37 summarizes the validity of the Transmit descriptor status bits regarding the
Transmission completion status:
Table 10–37: Transmit descriptor status validity
Transmission

Tx Status report

status

(ES,TO,LE,TN,UF,DE)

Underflow

Excessive collisions

Watchdog timeout

Internal Loopback

V - Valid
X - Meaningless

10.3.18 Setup frame
A setup frame defines SGEC Ethernet destination addresses. These addresses will be used
to filter all incoming frames. The setup frame is never transmitted over the Ethernet, nor
looped back to the receive list. While the setup frame is being processed, the receiver logic will
temporarily disengage from the Ethernet wire. The setup frame size is always 128 bytes and
must be wholly contained in a single transmit buffer. There are two types of setup frames:
1. Perfect Filtering addresses (16) list

10–40

KA660 Network Interface

2. Imperfect Filtering hash bucket (512) heads + one physical address.
10.3.18.1

First setup frame

A setup frame must be queued (placed in the transmit list with SGEC ownership) to
the SGEC before the Reception process is started, except for when the SGEC operates in
promiscuous reception mode.
NOTE
The self test completes with the SGEC Address filtering table fully set to "0". A
Reception process started without loading a setup frame will reject all the incoming
frames except those with a destination physical address = 000000h .
10.3.18.2

Subsequent setup frame

Subsequent setup frames may be queued to the SGEC regardless of the Reception process
state. The only requirement for the setup frame to be processed, is that the Transmission
process be in the RUNNING state. The setup frame will be processed after all preceding
frames have been transmitted and after the current frame reception, if any, is completed.
The setup frame does not affect the Reception process state but during the setup frame processing, the SGEC is disengaged from the Ethernet wire.
10.3.18.3

Setup frame descriptor

The setup frame descriptor format is shown in Figure 10–16, and described in the following
paragraphs.

KA660 Network Interface 10–41

Figure 10–16: Setup frame descriptor format

3
1

3
0

2
9

2
8

2
7

2
6

2
5

2
4

O
W

2
3

2
2

2
1

2
0

1
9

MBZ
DT

I
F

H
P

1
8

1
7

1
6

1
5

1
4

1
3

E
S

S
E

I
C

1
2

1
1

1
0

0
SDES0

MBZ
SDES1
u
SDES2

Buffer Size

u
u

SDES3

Setup Buffer Physical Address

0 − SGEC writes as "0"
u − Ignored by the SGEC on read, never written
ESB90P0066

Table 10–38: Setup frame descriptor bits
Word

Bit

Name

Description

SDES0

<13>

Setup Error - When set, indicates the setup
frame buffer size in not 128 bytes.

<15>

Error Summary - Set when SE is set.

<31>

Own bit - When set, indicates the descriptor is
owned by the SGEC. When cleared, indicates
the descriptor is owned by the host. The SGEC
clears this bit upon completing processing of the
descriptor and its associated buffer.

<24>

Interrupt on Completion - When set, the SGEC
will set NICSR5<TI> after this setup frame has
been processed.

SDES1

10–42

KA660 Network Interface

Table 10–38 (Cont.): Setup frame descriptor bits
Word

Bit

Name

Description

<25>

Hash/Perfect filtering mode - When set, the
SGEC will interpret the setup frame as a hash
table, and do an imperfect address filtering.
The imperfect mode is useful when there are
more than 16 Multicast addresses to listen to.
When clear, the SGEC will do a perfect address filter of incoming frames according to the
addresses specified in the setup frame.

<26>

Inverse filtering - When set, the SGEC will do
an inverse filtering: the SGEC will receive the
incoming frames with destination address not
matching the perfect addresses and will reject
the frames with destination address matching
one of the perfect addresses.
Meaningful only for Perfect_filtering (SDES1<HP>=0),
while Promiscuous and All_Multicast modes are
not selected (NICSR6<AF>=0).

<29:28>

Data Type - Must be 2 to indicate setup frame.

SDES2

<30:16>

Buffer Size - Must be 128.

SDES3

<29:1>

Physical Address - Physical address of setup
buffer.

NOTE
Setup buffer must be word
aligned.

10.3.18.4

Perfect Filtering setup frame buffer

This section describes how the SGEC interprets a setup frame buffer when SDES1<HP> is
clear.
The SGEC can store 16 - full 48 bits Ethernet - destination addresses. It will compare the
addresses of any incoming frame to these, and regarding the status of Inverse_Filtering flag
SDES1<IF>, will reject the following:
•
•

Those which do not match, if SDES1<IF> = 0
Those which match, if SDES1<IF> = 1

The setup frame must always supply all 16 addresses. Any mix of physical and Multicast
addresses can be used. Unused addresses should be duplicates of one of the valid addresses.
The addresses are formatted as shown in Figure 10–17.

KA660 Network Interface 10–43

Figure 10–17: Perfect Filtering setup frame buffer format

16 15

bit

Bytes <3:0>
PERFECT ADDRESS_00
<7:4> xxxxxxxxxxxxxxx|
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
PERFECT ADDRESS_01
xxxxxxxxxxxxxxx|
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
PERFECT ADDRESS_02
xxxxxxxxxxxxxxx|
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
PERFECT ADDRESS_03
xxxxxxxxxxxxxxx|
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
PERFECT ADDRESS_04
xxxxxxxxxxxxxxx|
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
PERFECT ADDRESS_05

~
~

.
.
.

Physical/Multicast bit

~
~

−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
PERFECT ADDRESS_13
xxxxxxxxxxxxxxx|
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
PERFECT ADDRESS_14
xxxxxxxxxxxxxxx|
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
PERFECT ADDRESS_15
<123:120>
<127:124> xxxxxxxxxxxxxxx|

xxxxxx = don’t care
ESB90P0067

10–44

KA660 Network Interface

The low-order bit of the low-order bytes is the address’s Multicast bit.
Example 10–1 illustrates a Perfect Filtering Setup buffer (fragment).
Example 10–1: Perfect filtering buffer
1

Ethernet addresses to be filtered:
A8-09-65-12-34-76
09-BC-87-DE-03-15
.
.
.
Setup frame buffer fragment:
126509A8
00007634
DE87BC09
00001503
.
.
.

1
2

10.3.18.5

Two Ethernet addresses written according to the DEC STD 134 specification for address
display.
Those two addresses as they would appear in the buffer.

Imperfect Filtering setup frame buffer

This section describes how the SGEC interprets a setup frame buffer when SDES1<HP> is set.
The SGEC can store 512 bits, serving as hash bucket heads, and one physical 48 bit Ethernet
address. Incoming frames with Multicast destination addresses will be subjected to the imperfect filtering. Frames with physical destination addresses will be checked against the single
physical address.
For any incoming frame with a Multicast destination address, the SGEC applies the standard
Ethernet CRC function to the first six bytes containing the destination address, then uses the
most significant nine bits of the result, as a bit index into the table. If the indexed bit is set,
the frame is accepted. If it is cleared, the frame is rejected.
This filtering mode is called imperfect, because Multicast frames not addressed to this station
may slip through, but it will still cut down the number of frames the host will be presented
with.
The format for the hash table and the physical address is shown in Figure 10–18.

KA660 Network Interface 10–45

Figure 10–18: Imperfect Filtering setup frame format

31
bytes <3:0>
<7:4>

<63:60>
<67:64>
<71:68>
<75:72>

<127:120>

16 15

HASH_FILTER_00
HASH_FILTER_01
.
.
.
HASH_FILTER_14
HASH_FILTER_15
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
PHYSICAL ADDRESS
xxxxxxxxxxxxxxxx|
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

bit

Physical/Multicast bit

xxxxxx = don’t care
ESB90P0068

Bits are sequentially numbered from right to left and down the table. For example, if
CRC(destination address)<8:0> = 33, the SGEC will examine bit #1 in the second longword.
Example 10–2 illustrates an Imperfect Filtering Setup frame buffer.

10–46

KA660 Network Interface

Example 10–2: Imperfect filtering buffer
1

Ethernet addresses to be filtered:
25-00-25-00-27-00
A3-C5-62-3F-25-87
D9-C2-C0-99-0B-82
7D-48-4D-FD-CC-0A
E7-C1-96-36-89-DD
61-CC-28-55-D3-C7
6B-46-0A-55-2D-7E

A8-12-34-35-76-08

Setup frame buffer:
00000000
10000000
00000000
00000000
00000000
40000000
00000080
00100000
00000000
10000000
00000000
00000000
00000000
00010000
00000000
00400000
4
353412A8
00000876

1
2
3
4

Ethernet Multicast addresses written according to the DEC STD 134 specification for
address display.
An Ethernet physical address.
The first part of an Imperfect filter Setup frame buffer with set bits for the 1 Multicast
addresses.
The second part of the buffer with the 2 physical address.

KA660 Network Interface 10–47

Example 10–3 shows a C program to compute the hash bucket heads and create the resultant Setup
frame buffer.
Example 10–3: Imperfect filtering Setup frame buffer creation C program
#include <stdio>
unsigned int imperfect_setup_frame[128/4],
/* bytes
*/
address[2],
crc[33];
/* CRC residue vector
*/

/* The setup buffer - 128

main()
{
int i, hash;
/*
*/
/* This program accepts 48 bits Ethernet addresses and builds a Setup frame */
/* buffer for imperfect filtering.
*/
/*
*/
/* Addresses must be entered in hexadecimal. The multicast bit is the least */
/* significant bit of the least significant digit of the first 32 bits.
*/
/* Non-multicast addresses are ignored.
*/
/*
*/
/* Input is terminated by keying CTRL/Z after which the program prints out */
/* the buffer.
*/
/*
*/
main_loop:
/* Prompt user for the Ethernet address
*/
printf("\n\n Enter the first 32 bits (HEX) - ");
if (scanf("%x", &address[0]) == EOF)
{
printf("\n\n Imperfect Setup buffer printout\n");
for (i=0; i < 128/4; i++)
printf("%08X\n", imperfect_setup_frame[i]);
exit(1);
}
printf("\n Enter the remaining 16 bits (HEX) - ");
scanf("%x",&address[1]);
/* Ignore non multicast addresses
if ((address[0] & 1) == 0)
goto main_loop;

/* Compute the hash function
*/
hash = address_crc(address[0],address[1]);
/* Set the appropriate bit in the Setup buffer
imperfect_setup_frame[hash/32] =
imperfect_setup_frame[hash/32] | 1 << hash%32;
goto main_loop;
}

Example 10–3 (continued on next page)

10–48

KA660 Network Interface

Example 10–3 (Cont.): Imperfect filtering Setup frame buffer creation C program
int address_crc( unsigned int lsb32 , unsigned int msb16)
{
int j,hash = 0;
/* Set CRC to all 1’s
for (j=0;
crc[j] = 1;

j < 33;

*/
j++)

/* Compute the address CRC by running the CRC 48 steps

for (j=0; j < 32; j++)
nextstate(lsb32 & 1<<j ? 1 : 0);
for (j=0; j < 16; j++)
nextstate(msb16 & 1<<j ? 1 : 0);
/* Extract 9 most significant bits from the CRC residue

for (j=24; j < 33; j++)
hash = hash<<1 | crc[j];
return hash;
}
nextstate(dat)
int dat;
{
int i,mean;
mean = crc[32] ^ dat;
for(i=32;i>=2;i--) crc[i]=crc[i-1];
crc[27] = crc[27] ^ mean;
crc[24] = crc[24] ^ mean;
crc[23] = crc[23] ^ mean;
crc[17] = crc[17] ^ mean;
crc[13] = crc[13] ^ mean;
crc[12] = crc[12] ^ mean;
crc[11] = crc[11] ^ mean;
crc[9] = crc[9] ^ mean;
crc[8] = crc[8] ^ mean;
crc[6] = crc[6] ^ mean;
crc[5] = crc[5] ^ mean;
crc[3] = crc[3] ^ mean;
crc[2] = crc[2] ^ mean;
crc[1] = mean;

10.3.19 SGEC operation
10.3.19.1

Hardware and Software Reset

The SGEC responds to two types of reset commands: a hardware reset through the RESET_L
pin, and a software reset command triggered by setting NICSR6<RE>. In both cases, the
SGEC aborts all ongoing processing and starts the Reset sequence. The SGEC restarts and
reinitializes all internal states and registers. No internal states are retained, no descriptors are
owned and all the host visible registers are set to ‘‘0’’, except where otherwise noted.

KA660 Network Interface 10–49

NOTE
The SGEC does not explicitly disown any owned descriptor; so descriptors OWNED
bits might be left in a state indicating SGEC ownership.
Table 10–39 indicates the NICSR fields which are not set to ‘‘0’’ after reset:
Table 10–39:
Field

Value

NICSR3

Unpredictable

NICSR4

Unpredictable

NICSR5<DN>

NICSR6<BL>

NICSR6<RE>

Unpredictable after HARDWARE reset
1 after SOFTWARE reset

NICSR7

Unpredictable

NICSR9

RT = TT = 1250

After the reset sequence completes, the SGEC executes the self test procedure to do basic
checking.
If the self test completes successfully, the SGEC initializes the SGEC, and sets the
Initialization Done flag NICSR5<ID>.
At the first failure detected in one of the basic tests executed in the self_test routine, the test
is aborted and the self_test failure NICSR5<SF> is set together with the self_test error status
NICSR5<SS>which indicates the failure reason.
INFO
The self test takes 25ms to complete after Hardware or Software RESET.
If the initialization completes successfully, the SGEC is ready to accept further host commands. Both the Reception and Transmission processes are placed in the STOPPED state.
Successive reset commands (either hardware or software) may be issued. The only restriction
is that SGEC NICSRs should not be accessed during a 1µsecond period following the reset.
Access during this period will result in a CP-BUS timeout error. Access to SGEC NICSRs
during the self test are permitted; however, only NICSR5 reads should be performed.
10.3.19.2

Interrupts

Interrupts are generated as a result of various events. NICSR5 contains all the status bits
which may cause an interrupt, provided NICSR6<IE> is set. The port driver must clear the
interrupt bits (by writing a ‘‘1’’ to the bit position), to enable further interrupts from the same
source.

10–50

KA660 Network Interface

Interrupts are not queued, and if the interrupting event reoccurs before the port driver has responded to it, no additional interrupts will be generated. For example, NICSR5<RI> indicates
one or more frames were delivered to host memory. The port driver should scan all descriptors,
from its last recorded position up to the first SGEC owned one.
An interrupt will only be generated once for simultaneous, multiple interrupting events. It
is the port driver responsibility to scan NICSR5 for the interrupt cause(s). The interrupt
will not be regenerated, unless a new interrupting event occurs after the host acknowledged
the previous one, and provided the port driver cleared the appropriate NICSR5 bit(s). For
example, NICSR5<TI> and NICSR5<RI> may both set, the host acknowledges the interrupt
and the port driver begins executing by reading NICSR5. Now NICSR5<RU> sets. The
port driver writes back its copy of NICSR5, clearing NICSR5<TI> and NICSR5<RI>. After
the host IPL is lowered below the SGEC level, another interrupt will be delivered with the
NICSR5<RU> bit set.
Should the port driver clear all NICSR5 set interrupt bits before the interrupt has been
acknowledged, the interrupt will be suppressed.
10.3.19.3

Startup procedure

A sequence of checks and commands must be performed by the port driver to prepare the
SGEC for operation.
1. Wait for the SGEC to complete its Initialization sequence by polling on NICSR5<ID> and
NICSR5<SF> (refer to Section 10.3.7 for details).
2. Examine NICSR5<SF> to find out whether the SGEC passed its self test. If it did not, it
should be replaced (refer to Section 10.3.7 for details).
3. Write NICSR0 to establish system configuration dependent parameters (refer to
Section 10.3.3 for details).
4. If the port driver intends to use VAX virtual addresses, NICSR7 must be written to identify the System Page Table to the SGEC (refer to Section 10.3.9 for details).
5. If the port driver wishes to change the default settings of the watchdog timers, it must
write to NICSR9 (refer to Section 10.3.11 for details).
6. The port driver must create the transmit and receive descriptor lists, then write to
NICSR3 and NICSR4 to provide the SGEC with the starting address of each list. The first
descriptor on the transmit list will usually contain a setup frame (refer to Section 10.3.6
for details).
7. Write NICSR6 to set global operating parameters and start the Transmission and
Reception processes. The Reception and Transmission processes enter the RUNNING
state and attempt to acquire descriptors from the respective descriptor lists and begin processing incoming and outgoing frames (refer to Section 10.3.8 for details). The Reception
and Transmission processes are independent of each other and can be started and stopped
separately.
CAUTION
If address filtering (either perfect or imperfect) is desired, the Reception process
should only be started after the Setup frame has been processed.

KA660 Network Interface 10–51

8. The port driver now waits for any SGEC interrupts. If either the Reception or Transmission
processes were SUSPENDED, the port driver must issue the Poll Demand command after
it has rectified the suspension cause.
10.3.19.4

Reception process

While in the RUNNING state, the Reception process polls the receive descriptor list, attempting to acquire free descriptors. Incoming frames are processed and placed in acquired
descriptors’ data buffers, while status information is written to the descriptor RDES0 words.
The SGEC always tries to acquire an extra descriptor in anticipation of incoming frames.
Descriptor acquisition is attempted under the following conditions:
•
•
•
•

Immediately after being placed in the RUNNING state through setting of NICSR6<SR>.
The SGEC begins writing frame data to a data buffer pointed to by the current descriptor.
The last acquired descriptor chained (RDES1<CA> = 1 ) to another descriptor.
A virtual translation error was encountered RDES0<TN> while the SGEC was translating
the buffer base address of the acquired descriptor .

As incoming frames arrive, the SGEC strips the preamble bits and stores the frame data in
the receive FIFO. Concurrently, it performs address filtering according to NICSR6 fields AF,
HP and its internal filtering table. If the frame fails the address filtering, it is ignored and
purged from the FIFO. Frames which are shorter than 64 bytes, due to collision or premature
termination are also ignored and purged from the FIFO, unless NICSR6<PB> is set.
After 64 bytes have been received, the SGEC begins transferring the frame data to the buffer
pointed to by the current descriptor. If Data Chaining is enabled (NICSR6<DC> clear), the
SGEC will write frame data overflowing the current data buffer into successive buffer(s). The
SGEC sets the RDES0<FS> and RDES0<LS> in the first and last descriptors, respectively, to
delimit the frame. Descriptors are released (RDES0<OW> bit cleared) as their data buffers fill
up or the last segment of a frame has been transferred to a buffer.
The SGEC sets RDES0<LS> and the RDES0 status bits in the last descriptor it releases for a
frame. After the last descriptor of a frame is released, the SGEC sets NICSR5<RI>.
This process is repeated until the SGEC encounters a descriptor flagged as owned by the host.
After filling up all previously acquired buffers, the Reception sets NICSR5<RU> and enters
the SUSPENDED state. The position in the receive list is retained.
Any incoming frames while in this state will cause the SGEC to fetch the current descriptor
in the host memory. If the descriptor is now owned by the SGEC, the Reception re-enters the
RUNNING state and starts the frame reception.
If the descriptor is still owned by the host, the SGEC increments the Missed Frames Counter
(NICSR10<MFC>) and discards the frame.
Table 10–40 summarizes the Reception process state transitions and resulting actions:

10–52

KA660 Network Interface

Table 10–40: Reception process state transitions
From state

Event

To state

Action

STOPPED

Start Reception command

RUNNING

Receive polling begins from
last list position or from the
the list head if this is the first
Start command issued, or if the
receive descriptor list address
(NICSR3) was modified by the
port driver.

RUNNING

SGEC attempts acquisition of a
descriptor owned by the host

SUSPENDED

NICSR5<RU> is set when the
last acquired descriptor buffer
is consumed. Position in list
retained.

RUNNING

Stop Reception command

STOPPED

Reception process is STOPPED
after the current frame, if any,
is completely transferred to
data buffer(s). Position in list
retained.

RUNNING

Memory or host bus parity error
encountered

STOPPED

Reception is cut off and
NICSR5<ME> is set.

RUNNING

Reset command

STOPPED

Reception is cut off.

SUSPENDED

Rx Poll demand or Incoming
frame and available descriptor

RUNNING

Receive polling resumes from
last list position or from the list
head if NICSR3 was modified
by the port driver.

SUSPENDED

Stop Reception command

STOPPED

None.

SUSPENDED

Reset command

STOPPED

None.

10.3.19.5

Transmission process

While in the RUNNING state, the Transmission process polls the transmit descriptor list
for any frames to transmit. Frames are built and transmitted on the Ethernet wire. Upon
completing frame transmission (or giving up), status information is written to the TDES0
words. Once polling starts, it continues (in sequential or descriptor chained order) until the
SGEC encounters a descriptor flagged as owned by the host, or an error condition. At this
point, the Transmission process is placed in the SUSPENDED state and NICSR5<TI> is set.
NICSR5<TI> will also be set after completing transmission of a frame which has TDES1<IC>
set in its last descriptor. In this case, the Transmission process remains in the RUNNING
state.
Frames may be data chained and span several buffers. Frames must be delimited by
TDES1<FS> and TDES1<LS> in the first and last descriptors, respectively, containing the
frame. While in the RUNNING state, as the Transmission process starts, it first expects a
descriptor with TDES1<FS> set. Frame data transfer from the host buffer to the internal
FIFO is initiated. Concurrently, if the current frame had TDES1<LS> clear, the Transmission
process attempts to acquire the next descriptor, expecting TDES1<FS> and TDES1<LS> to
be clear indicating an intermediary buffer, or TDES1<LS> to be set, indicating the end of the

KA660 Network Interface 10–53

frame. After the last buffer of the frame has been transmitted, the SGEC writes back final
status information to the TDES0 word of the descriptor having TDES1<LS> set, optionally
sets NICSR5<TI> if TDES1<IC> was set, and repeats the process with the next descriptor(s).
Actual frame transmission begins after at least 72 bytes have been transferred to the internal
FIFO, or a full frame is contained in the FIFO. Descriptors are released (TDES0<OW> bit
cleared) as soon as the SGEC is through processing a descriptor.
Transmit polling suspends under the following conditions:
•
•
•

the SGEC reaches a descriptor with TDES0<OW> clear. To resume, the port driver must
give descriptor ownership to the SGEC and issue a Poll Demand command.
The TDES1<FS> and TDES1<LS> are incorrectly paired or out of order. TDES0<LE> will
be set.
A frame transmission is given up due to a locally induced error. The appropriate TDES0
bit is set.

The Transmission process enters the SUSPENDED state and sets NICSR5<TI>. Status
information is written to the TDES0 word of the descriptor causing the suspension. The
position in the transmit list, in all of the above cases, is retained. The retained position is that
of the descriptor following the last descriptor closed (set to host ownership) by the SGEC.
NOTE
The SGEC does not automatically poll the Tx descriptor list and the port driver
must explicitly issue a Tx Poll Demand command after rectifying the suspension
cause.
The following table summarizes the Transmission process state transitions:
Table 10–41: Transmission process state transitions
From state

Event

To state

Action

STOPPED

Start Transmission command

RUNNING

Transmit polling begins
from the last list position
or from the head of
the list if this is the
first Start command
issued, or if the transmit
descriptor list address
(NICSR4) was modified
by the port driver.

RUNNING

SGEC attempts acquisition of a descriptor owned by the host

SUSPENDED

NICSR5<TI> is set.
Position in list retained.

RUNNING

Out of order delimiting flag (TDES0<FS>
or TDES0<LS>) encountered.

SUSPENDED

TDES0<LE> and
NICSR5<TI> are set.
Position in list retained.

10–54

KA660 Network Interface

Table 10–41 (Cont.): Transmission process state transitions
From state

Event

To state

Action

RUNNING

Frame transmission aborts due to a
locally induced error.

SUSPENDED

Appropriate TDES0
and NICSR5<TI> bits
are set. Position in list
retained.

RUNNING

Stop Transmission command

STOPPED

Transmission process
is STOPPED after the
current frame, if any, is
transmitted. Position in
list retained.

RUNNING

Transmit watchdog expires

STOPPED

Transmission is cut off
and NICSR5<TW> ,
TDES0<TO> are set.
Position in list retained.

RUNNING

Memory or host bus parity error encountered

STOPPED

Transmission is cut off
and NICSR5<ME> is
set.

RUNNING

Reset command

STOPPED

Transmission is cut off.

SUSPENDED

Tx Poll Demand command

RUNNING

Transmit polling resumes from last list
position or from the list
head if NICSR4 was
modified by the port
driver.

SUSPENDED

Stop Transmission command

STOPPED

None.

SUSPENDED

Reset command

STOPPED

None.

10.3.19.6

Loopback operations

The SGEC supports two loopback modes:
•

•

Internal loopback
This mode is generally used to verify correct operations of the SGEC internal logic. While
in this mode, the SGEC will take frames from the transmit list and loop them back,
internally, to the receive list. The SGEC is disengaged from the Ethernet wire while in
this mode.
External loopback
This mode is generally used to verify correct operations up to the Ethernet cable. While
in this mode, the SGEC will take frames from the transmit list and transmit them on
the Ethernet wire. Concurrently, the SGEC listens to the line which carries its own
transmissions and places incoming frames in the receive list.
NOTE
Caution should be exercised in this mode as transmitted frames are placed on
the Ethernet wire. Furthermore, the SGEC does not check the origin of any

KA660 Network Interface 10–55

incoming frames, consequently , frames not necessarily originating from the
SGEC might make it to the receive buffers.
In either of these modes, all the address filtering and validity checking rules apply. The port
driver needs to take the following actions:
1. Place the Reception and Transmission processes in the STOPPED state. The port driver
must wait for any previously scheduled frame activity to cease. This is done by polling the
TS and RS fields in NICSR5.
2. Prepare appropriate transmit and receive descriptor lists in host memory. These may
follow the existing lists at the point of suspension, or may be new lists which will have to
be identified to the SGEC by appropriately writing NICSR3 and NICSR4.
3. Write to NICSR6<OM> according to the desired loopback mode and place the Transmission
and Reception processes in the RUNNING state through Start commands.
4. Respond and process any SGEC interrupts, as in normal processing.
To restore normal operations, the port driver must execute above step #1, then write the OM
field in NICSR6 with ‘‘00’’.
10.3.19.7

DNA CSMA/CD counters and events support

This section describes the SGEC features that support the port driver in implementing and
reporting the specified counters and events 4 .
Table 10–42: CSMA/CD counters
Counter

SGEC feature

Time since counter creation

Supported by the host driver.

Bytes received

Port driver must add up the RDES0<FL> fields of all
successfully received frames.

Bytes sent

Port driver must add up the TDES2<BS> fields of all
successfully transmitted buffers.

Frames received

Port driver must count the successfully received frames
in the receive descriptor list.

Frames sent

Port driver must count the successfully transmitted
frames in the transmit descriptor list.

Multicast bytes received

Port driver must add up the RDES0<FL> fields of all
successfully received frames with Multicast address
destinations.

Multicast frames received

Port driver must count the successfully received frames
with Multicast address destinations.

Frames sent, initially deferred

Port driver must count the successfully transmitted
frames with TDES0<DE> set.

As specified in the DNA Maintenance Operations (MOP) Functional Specification, Version T.4.0.0, 28 January 1988.

10–56

KA660 Network Interface

Table 10–42 (Cont.): CSMA/CD counters
Counter

SGEC feature

Frames sent, single collision

Port driver must count the successfully transmitted
frames with TDES0<CC> equal to 1.

Frames sent, multiple collisions

Port driver must count the successfully transmitted
frames with TDES0<CC> greater than 1.

Send failure- Excessive collisions

Port driver must count the transmit descriptors having
TDES0<EC> set.

Send failure- Carrier check failed

Port driver must count the transmit descriptors having
TDES0<LC> set.

Send failure- Short circuit

Two successive transmit descriptors with the No_carrier
flag TDES0<NC> set, indicates a short circuit.

Send failure- Open circuit

Two successive transmit descriptors with the Excessive_
collisions flag TDES0<EC> set with the same Time
domain reflectometer value TDES0<TDR>, indicate an
open circuit.

Send failure- Remote Failure to Defer

Flagged as a late collision TDES0<LC> in the transmit
descriptors.

Receive failure- Block Check Error

Port driver must count the receive descriptors having
RDES0<CE> set with RDES0<DB> cleared.

Receive failure- Framing Error

Port driver must count the receive descriptors having
both RDES0<CE> and RDES0<DB> set.

Receive failure- Frame too long

Port driver must count the receive descriptors having
RDES0<TL> set.

Unrecognized frame destination

Not applicable.

Data overrun

Port driver must count the receive descriptors having
RDES0<OF> set.

System buffer unavailable

Reported in the Missed_frame counter NICSR10<MFC>
(refer to Table 10–20).

User buffer unavailable

Not applicable.

Collision detect check failed

Port driver must count the transmit descriptors having
TDES0<HF> set.

CSMA/CD specified events can be reported by the port driver based on the above table. The
Initialization Failed event is reported through NICSR5<SF>.

KA660 Network Interface 10–57

Chapter 11
KA660 Mass Storage Interface

The KA660 contains a DSSI Bus Interface which is implemented with the Single Host Adapter
Chip (SHAC). This interface allows the KA660 to transmit packets of data to, and receive
packets of data from, up to seven other DSSI devices (typically RF type disk drives and TF
type streaming tape drives). It should also be noted that the SHAC supports CP Bus Parity
Protection.

11.1 SHAC Introduction
SHAC (Single Host Adapter Chip) is a single-chip, VLSI version of an SCA port that uses a
DSSI bus as the physical interconnect. Another SCA realization, CI, has defined a port-driver
/port interface which has been used to connect VAXs in clusters. DSSI has adopted the same
interface, so the same VMS port driver will be able to drive either a CI-port or SHAC. The
SHAC can be used to connect a host to any other device that can communicate through the
CI-DSSI protocol. In particular, it provides a solution to the following:
•
•

The problem of interfacing a group of mass-storage device controllers (MSDCs) to a VAX.
The problem of interfacing several VAXs to a common group of MSDCs and, if higher level
protocols support this option, to one another.

Where two or more VAXs connect to a group of MSDCs (or to one another) through DSSI, each
has a SHAC or another DSSI port. When a group of MSDCs connect to the DSSI bus, the
controllers provide both the bus interface and the intelligent control required to respond to the
CI commands received over the DSSI.
On the 1-byte wide DSSI bus both the MSDCs and the several VAXs communicate at highspeed, with a 4 to 5 MB/s burst transfer rate. The SHAC handles the problem of providing
effective, efficient and reliable interfacing between this DSSI bus and the SOC CPU , having
direct host memory access (DMA) over the host’s 32-bit wide, 16 MB/s CP bus. All communications between those connected to the DSSI will follow the CI protocol with the DSSI protocols
providing handshaking in the transactions.
Structural parameters limit the number of possible combinations that can be realized with
DSSI and SHAC.
•

A single DSSI bus has room for 8 nodes which may be partitioned among host adapters
(for example, SHACs) and MSDCs.

KA660 Mass Storage Interface 11–1

•
•

Up to 4 SHACs can be installed on a single host bus.
Because there must be a host, there can be up to 7 MSDCs on a single DSSI.

The SHAC provides a small amount of buffering (1.2KB) on chip to improve bus utilization
on both sides, but the SHAC is designed to pass data through from one bus to the other as
rapidly as the two busses permit. DMA services to and from the main memory reside in the
SHAC, which responds to requests for transfers between the host and the remote nodes.
The SHAC is operated by an on-chip RISC that obtains its code and internal data from on-chip
RAM and ROM. The RAM will be loaded from main memory both during initialization and
as circumstances require during normal run time. With this capability, it can read in new
code and data from the main memory and thus adapt its behavior to new circumstances. This
will permit inexpensive upgrades of SHACs after they are installed in the field. Furthermore,
it will allow the SHAC to store infrequently accessed code in main memory, providing more
capability than could be included in on-chip ROM.
The overall communication architecture under which the SHAC works is Digital’s Systems
Communications Architecture (SCA). In this general architecture, four layers are defined, as
shown in Figure 11–1. The architecture can be realized in a variety of ways. Two particularizations of the lowest two levels in the diagram are CI (Computer Interconnect) and DSSI
(Digital Storage System Interconnect). They share the same lowest host layer (CI Port driver)
but have distinctly different physical interconnects. The layers between the Port Driver and
the DSSI bus itself can be realized at both board and chip level and products at both levels
are in design within Digital. The SHAC is a chip-level product which connects the host-bus to
the DSSI bus, controlled by the SOC CPU through a CI port driver and accepts and delivers
CI-defined packets over the DSSI bus. Layers above the port driver are invisible to SHAC.

11–2

KA660 Mass Storage Interface

Figure 11–1: Relationship of the DSSI to SCA and CI

SCA
3. I/O Applications
(SYSAP)
2. System Communications
(SCS)
1. Port/Port Driver
(PDP)

CI
1b.

DSSI
CI Port Driver

1a. CI Port
0. Physical Interconnect
(PI)

0b. CI Data Link
0a. CI bus

1a. DSSI Port
0b. DSSI Data Link

S
H
A
C

0a. DSSI bus
ESB90P0079

The port driver maintains a set of 7 queues in its system space. Four of these contain commands for the SHAC to execute. The priority of the command is determined by the queue
it is on; order is determined by the position in the queue. Another queue contains all of the
responses for the host (from the SHAC or the remote nodes). Finally, there are two queues of
"empty envelopes" for the host and the SHAC to use to stuff with commands and responses
and then to queue them on the other queues.
These "envelopes" are simply standard-sized "queuable" blocks of host memory. All commands
and responses are copied into one of these standard-sized blocks. Included in the header
on each block are a pair of queue pointers (for a doubly linked queue) and various standard
identifiers which specify what is contained in the block and how much of the block represents
the actual command or response. To be visible, a block must be on a queue, where pointers
from other elements or the queue header show its presence. Once a block is removed from a
queue, it is visible only to the entity which removed it.
The SHAC’s principal task is in accepting and delivering "mail" to other nodes. Externally
(For example, on DSSI) the SHAC deals only in standard CI formats. Internally, the SHAC
deals with the envelopes just described and with blocks of data. Because DSSI deals with
bytes and the CP bus deals in longwords, the SHAC must frequently do byte alignment tasks
during transcription.
The SHAC deals with the port driver in the virtual-address mode, unloading from the SOC
CPU the obligation to do virtual-to-physical address translation and to be aware of page crossings in virtually-contiguous blocks of information. The SHAC supports full virtual address
translation including the use of global I/O pages (to a depth of 1).
The rest of this SHAC overview section describes a typical set of steps that the SHAC goes
through in serving its role as the CI Port, with "mail" in both directions.

KA660 Mass Storage Interface 11–3

11.2 CI-DSSI Overview
At start-up, the host provides the SHAC with a number of pointers to internal host structures,
one of which, the Port Queue Block (PQB), contains pointers and data on all of the queues
that the host maintains for CI. The SHAC uses this data to carry on its normal business in
the following way.
If traffic is not coming in on the DSSI bus, the SHAC goes to the highest command queue
which has something enqueued. Choices are CMDQ0..CMDQ3, with 3 being most urgent. It
dequeues an element from the queue and examines its header to see what it must do with
the queue entry. It could be a command for the SHAC or an item to be delivered to one of the
nodes on the DSSI. A command might be an order to deliver a block of data to a remote node.
An item to be delivered would be either a datagram or a message.
A datagram is a "one-sided" communication—that is, one which will be sent without any
assurance of either receipt or reply. An obvious application for such a communication is a
request for the party at the other node to identify itself. If the host does not know if anything
at all is out there, it must transmit its request without expectation. For this or any similar
purpose, it employs a datagram. Datagrams are all of lengths guaranteed to fit in a datagram
envelope.
A message is a "two-sided" communication used when a virtual circuit (an established formal
relationship) between members of the bus exist. Once such a virtual circuit is established,
the host(s) understand how to make requests of the other side. Such a request could be an
order for a data transfer in either direction. The message itself (move data) is contained in
a command (deliver this message to ...). Messages are all of lengths guaranteed to fit in a
message envelope. Messages are always delivered sequentially to a given node—that is, in the
order in which they were enqueued on a particular queue. While the SHAC supports retries
if a message fails to get through, once the retry limit is reached without successful delivery,
SHAC returns the command to the host, marking it as undeliverable, and then breaks the
virtual circuit to that node.
A full transaction might go something like this:
1. The host queues a message for node 3 (for example, a disk controller) to copy a block of
16 KB from host memory, starting at location X and to be stored in location Y on disk.
The queues are doubly-linked, so at the top of every envelope there is a forward link
FLINK and a backward link BLINK. Enqueuing involves putting link values into the
new element’s FLINK and BLINK and making the previous last-element’s FLINK and the
queue header’s BLINK point to the new element.
2. When this message gets to the head of the queue, the SHAC dequeues it 1 , reads the
header and finds that it should "dial up" node 3. To do this, the SHAC goes through
the DSSI protocols, contending for the DSSI bus and then, if successful in getting bus,
specifying node 3 as the target. These steps are called arbitration and selection.

11–4

Note that the SHAC ends up holding the only pointer to the dequeued block of memory that constitutes the queue
element. The port driver no longer "knows" where it is.

KA660 Mass Storage Interface

3. Node 3 responds by asking for the DSSI-command (command-out phase). In this phase, the
SHAC tells node 3 how many bytes are coming and repeats the identification information
to confirm a proper selection. Node 3 then tells the SHAC to switch to the data-out phase.
The SHAC sends a pair of CI header bytes to identify what type of message this is, and
then proceeds to transmit the actual message read from the message block in host memory.
The step-by-step details of the transfer are handled by hardware in the SHAC which
permits simultaneous, buffered reading and writing on the two busses to which the SHAC
is connected. Upon proper completion of the transmission, node 3 responds with a 1-byte
acknowledgment of success (parity and check-sum proper and no other errors).
4. The SHAC is still holding the only pointer to the message block in host memory. It returns
this to the host in one of two ways. If the host has requested a "return receipt", the SHAC
puts the block on the Response Queue RSPQ to indicate proper delivery. This is where
the port-driver software in the host will look for responses.
Alternatively, the SHAC simply puts it back on the MFREEQ which holds the standard
envelopes for messages. At this point the single message has been delivered and the
message envelope is back in circulation.
5. After whatever delay node 3 needed to process the message, it contends for the bus and
upon winning it, selects the SHAC as its target. It then sends a standard CI message as
above telling the SHAC to transmit the required data. In general, the SHAC does not do
this immediately, because it is obliged to handle traffic according to position in the queue
and according to queue priority. Instead, it takes an empty envelope from MFREEQ,
writes into it the message it is receiving and puts it on the proper CMDQ as specified in
the message it just received.
6. When that message gets to the head of its queue, the SHAC dequeues it once more, carries
out its command (in transmissions of 4 KB whenever possible—a 4 KB transmission takes
about 1 msec on the DSSI), possibly interleaving other transmissions of higher priority to
this node or any priority to other nodes, until the last byte is sent. Once the SHAC has
completed this operation, it returns the message block to the MFREEQ.
7. Node 3 has put its data on the disk and must report to the host the successful completion
of the transaction. Again it contends for the bus and upon winning specifies the SHAC as
its target. Then it sends a message to the port-driver through the SHAC confirming the
successful transaction. The SHAC dequeues another free envelope and writes this message
into that block. Then it queues it on the host’s RSPQ. Except for higher level responses in
the host, that concludes a whole transaction.
The enqueue/dequeue operations represent a considerable fraction of the effort in delivering
a message or datagram. To minimize this effort, the SHAC caches a small number of the
envelopes (that is, it hangs onto the pointers to the memory blocks) as they become free in
its normal activity. It only fetches an envelope from the free queues when its own supply has
disappeared, and it only returns them to the free queues when it has a full supply (4 of a type).
By this and other attention to effort reduction and traffic conservation, the SHAC attempts to
optimize its rate of doing useful work.

KA660 Mass Storage Interface 11–5

11.3 SHAC Registers
The CPU communicates directly with the SHAC chip through a set of device registers in each
the SHAC. These registers occupy a one-page (512-byte) region in I/O address-space, aligned
on a page boundary.
All of the registers are longword registers. They may be accessed only through longword
operations.
In addition to the access restrictions listed for specific registers, no register other than SHAC
Software Chip Reset (SSWCR) may be read or written while certain chip intialization functions
are being executed. The results of such an access during the 100 milliseconds following a
reset (power-up or a write to SSWCR), or during the 50 microseconds following a MIN-bit
(PMCSR<0>) reset are UNPREDICTABLE.
The registers can be divided into two categories:
•
•

The CI Port registers defined in the CI Port Architecture specification;
The SHAC specific registers.

11.3.1 CI Port Registers

11.3.1.1

Port Queue Block Base Register (PQBBR)

SHAC I/O Address: 2000 424816
This Port Queue Block Base Register (PQBBR) contains the uppermost bits of the physical
address of the base of the Port Queue Block (PQB). After a RESET the PQBBR is loaded by
the SHAC with configuration information. This information remains in the PQBBR until the
PQBBR is written with the address of the Port Queue Block. Figure 11–2 shows the format.
Table 11–1 lists the bit descriptions.
PQBBR is writeable only when the port is in the disabled or disabled/maintenance state and
readable anytime (except during chip intialization).

11–6

KA660 Mass Storage Interface

Figure 11–2: Port Queue Block Base Register (PQBBR)

3
1

2 2
1 0

MBZ

PQB Base <29:9>

Longword Read/Write Access.
ESB90P0069

Table 11–1: Port Queue Block Base Address Register (PQBBR)
Data Bit

Name

Description

<31:21>

MBZ

Read as ZERO, must be Written as ZERO.

<20:0>

PQB Base
<29:9>

This field contains the uppermost bits of the physical address of
the base of the Port Queue Block (PQB). Note, the PQB must be
page-aligned, so the remaining bits of the address are assumed to be
ZERO.

Following chip reset, PQBBR contains the configuration shown in Figure 11–3. The bit descriptions are listed in Table 11–2.
Figure 11–3: Port Queue Block Base Register (PQBBR) After RESET

3
1

22
43
HW Ver.

11
65
FW Ver.

87
SHM Ver.

0
Maint ID
ESB90P0070

KA660 Mass Storage Interface 11–7

Table 11–2: Port Queue BLock Base Address Register Bits After RESET

11–8

Data Bit

Name

Description

<31:24>

HW Ver.

Hardware Version. The Hardware version of the SHAC which is
greater than ZERO.

<23:16>

FW Ver.

Firmware Version. The Firmware version of the SHAC which is
greater than ZERO.

<15:8>

SHW Ver.

Shared Host Memory Version. The Shared Host Memory version of
the SHAC which is ZERO until the Shared Host Memory Data Area
has been read in; thereafter, greater than ZERO.

<7:0>

Maint ID

CI Port Maintenance ID. The CI Port Maintenance ID which should
always be 22 16 .

KA660 Mass Storage Interface

11.3.1.2

Port Status Register (PSR)

SHAC I/O Address: 2000 424C16
The Port Status Register (PSR) contains a status report. If interrupts are enabled, for example
(PMCSR<2>) set, the port interrupts the SOC CPU each time that it writes to this register.
Once an interrupt is requested by the port, the value of PSR is fixed and is not changed until
the SOC CPU releases it by writing the Port Status Release Control Register (PSRCR). The
Port Status Register format is shown in Figure 11–4 and the bit descriptions are in Table 11–3.
PSR is read only and may be read anytime by the port driver, except during chip initialization.
Its value following a write to it is UNPREDICTABLE.

Figure 11–4: Port Status Register (PSR)

3 3
1 0

2 2 2 1 1 1 1 1
2 1 0 9 8 7 6 5

8 7 6 5 4 3 2 1 0
MBZ
RQA
MFQE
PDC
PIC
DSE
MSE
MTE
MISC
SHME
SMPE
ISN
DE
QDE
II
ME

Longword Read Only Access.
ESB90P0071

KA660 Mass Storage Interface 11–9

Table 11–3: Port Status Register Bit Descriptions
Data Bit

Name

Description

<31>

MTE

Maintenance Error. When set, the port has detected an implementation
specific error (or hardware status condition). The source of the error may
be more accurately determined from the other bits in the upper word of this
register (PSR) and the contents of other registers. Also when set the port
is in the uninitialized state (port is non-functional). Maintenance Errors
normally indicate a severe SHAC hardware or software failure.

<30:22>

MBZ

Read as ZERO, writes have no effect.

<21>

Illegal Interrupt. When set, this bit indicates a SHAC internal error, detected when the SHAC’s microprocessor received an interrupt from a
invalid source. This causes ME (PSR<31>) to set and the port to enter
the uninitialized state (port is non-functional).

<20>

QDE

QUIP Detected Error. When set this bit indicates a SHAC internal error
detected when the SHAC’s microprocessor (QUIP) was given an invalid
instruction. This causes ME (PSR<31>) to set and the port to enter the
uninitialized state (port is non-functional).

<19>

Diagnostic Error. When set an error was detected while the SHAC was
running its internal self-test. The causes ME (PSR<31>) to set and the port
to enter the uninitialized state (port is non-functional).

<18>

ISN

Illegal Segment Number. When set this indicates a SHAC internal error in
which it attempted to load a non-existent External Segment from the SHAC
Shared Host Memory. The causes ME (PSR<31>) to set and the port to enter
the uninitialized state (port is non-functional).

<17>

SMPE

Slave Mode Parity Error. This bit is set by the occurrence of a parity
error during a SOC CPU access of a SHAC device register. The causes
ME (PSR<31>) to set and the port to enter the uninitialized state (port is
non-functional).

<16>

SHME

Share Host Memory Error. This bit is set by the occurrence of an error
involving the SHAC shared Host Memory. The causes ME (PSR<31>) to set
and the port to enter the uninitialized state (port is non-functional).

<15:8>

MBZ

Read as ZERO, writes have on effect.

<7>

MISC

Miscellaneous. When set this bit indicates that the port microcode has
detected one of the miscellaneous errors and the port is about to enter the
disabled/maintenance state. The actual error code is stored in the Port
Error Status Register.

<6>

Maintenance Timer Expiration. When set the maintenance timer has
expired. The port is in the uninitialized/maintenance state.

<5>

MSE

Memory System Error. When set the port has encountered an uncorrectable
data or non-existent memory error in referencing memory. Port is in the
disabled or disabled/maintenance state. See the Port Failing Address
Register (PFAR) for further information.

11–10

KA660 Mass Storage Interface

Table 11–3 (Cont.): Port Status Register Bit Descriptions
Data Bit

Name

Description

<4>

DSE

Data Structure Error. When set, the port has encountered an error in a port
data structure (for example, queue entry, PQB, BDT or page table). Port is
in the disabled or disabled/maintenance state. See the Port Error Status
Register (PESR) and the Port Failing Address Register (PFAR) for further
information. Note that errors in queue structures leave the queues locked.

<3>

PIC

Port Initialization Complete. When set, the port has completed internal
initialization. The port is in the disabled or disabled/maintenance state.

<2>

PDC

Port Disable Complete. When set, the port is in the disabled or disabled
/maintenance state.

<1>

MFQE

Message Free Queue Empty. When set, the port attempted to remove an
entry from the Message Free Queue (MFREEQ) and found it empty. Port
processing of commands continues and, thus, the MFREEQ may not be
empty at the time the port driver gets control.

<0>

RQA

Response Queue Available. When set this bit indicates port has inserted an
entry on an empty Response Queue.

KA660 Mass Storage Interface 11–11

11.3.1.3

Port Error Status Register (PESR)

SHAC I/O Address: 2000 425016
The Port Error Status Register (PESR) indicates the type of error which resulted in a DSE
(PSR<4>) or an MISC (PSR<7>) error. Figure 11–5 shows the format. Table 11–4 lists the bit
descriptions.
PESR is read only by the SOC CPU and valid only after either a DSE or MISC error, or after
certain ME (PSR<31>) and DE (PSR<19>) errors. Its value at any other time, or following a
write to it, is UNPREDICTABLE.
Figure 11–5: Port Error Status Register (PESR)
3
1

1 1
6 5

MEC

DEC

Longword Read Only Access
ESB90P0072

Table 11–4: Port Error Status Register Bit Definitions
Data Bit

Name

Description

<31:16>

MEC

Miscellaneous Error Code. This code comprises two fields: bits
<31:24> define the the module within the SHAC code where the
error occurred, and bits <23:16> contain the specific error that
occurred. These codes are implementation specific.

<15:0>

DEC

Data Structure Error Code.

11–12

KA660 Mass Storage Interface

11.3.1.4

Port Failing Address Register (PFAR)

SHAC I/O Address: 2000 425416
The format for the Port Failing Address Register is shown in Figure 11–6.
After an DSE, MSE, and ME or DE error (as indicated by PSR), or after a response with
Buffer Memory System Error status, the Port Failing Address Register (PFAR) contains the
memory address at which the failure occurred. The address may be the exact failing address,
an address in the same page as the exact failing address or, in the case of DSE, an address in
some part of the data structure. For DSE, PFAR contains a virtual address or offset, while for
MSE interrupts and buffer memory system errors the PFAR contains a physical address. For
ME, the interpretation of the address is error-dependent.
Because the port continues command execution and packet processing after Buffer Memory
System Errors, the PFAR is overwritten if subsequent errors occur. For DSE, MSE, and
ME errors the PFAR is effectively fixed because the port enters the disabled, disabled
/maintenance, or uninitialized state.
PFAR is read only by the SOC CPU and readable after a DSE, MSE, or ME or DE errors or
after a response with Buffer Memory System Error status. Its value at any other time, or
following a write to it, is UNPREDICTABLE.
Figure 11–6: Port Failing Address Register (PFAR)

3
1

0
Failing Address

Longword Read Only Access
ESB90P0073

KA660 Mass Storage Interface 11–13

11.3.1.5

Port Parameter Register (PPR)

SHAC I/O Address: 2000 425816
The Port Parameter Register (PPR) contains port implementation parameters and the
port number. The value of the PPR is set by the port during initialization and valid after a PIC (PSR <3>) interrupt. Its value at any other time, or following a write to it, is
UNPREDICTABLE. PPR is read only by the SOC CPU . The Port Parameter Register format
is shown in Figure 11–7. The bit descriptions are listed in Table 11–5.
Figure 11–7: Port Parameter Register (PPR)

3
1

2 2
9 8

CSZ

1 1 1
6 5 4
IBUF_LEN

8 7
ISDI

0
PORT_NO

Longword Read Only Access
ESB90P0074

Table 11–5: Port Parameter Register Bit Descriptions (PPR)
Data Bit

Name

Description

<31:29>

CSZ

Cluster Size. For SHAC, this value always is ZERO, indicating
a maximum of 16 ports on the DSSI bus. (Note that the DSSI
architecture only allows up to 8 ports on the bus, but 16 is the
smallest size defined for the CSZ field.)

<28:16>

IBUF_LEN

Internal Buffer Length. This field indicates the size of internal
buffers available for message and data transfers. Maximum data
packet = IBUF_LEN - 16 bytes. Maximum message or datagram
length = IBUF_LEN. For SHAC, the value is 4112 101016 .

<15>

MBZ

Read as ZERO, writes have an UNPREDICTABLE effect.

<14:8>

ISDI

Implementation Specific Diagnostic Information. The bits in
this field contain information about the local adapter’s link layer
configuration. For SHAC, the definitions of these bits are Read as
Zero.

<7:0>

Port_NO

Port Number. This is the same as the SHAC’s DSSI ID.

11–14

KA660 Mass Storage Interface

11.3.1.6

Port Control Registers

The port control registers are 32-bit registers which are write-only by the SOC CPU . To
invoke the function provided by any of the control registers, the SOC CPU writes a ONE to
the register.
The result of writing any other value to any of these registers is UNPREDICTABLE. The value
read from any of them is also UNPREDICTABLE. The format for the Port Control Registers is
shown in Figure 11–8.

Figure 11–8: Port Control Registers

3
1

1 0
MBZ
MBO
Longword Write Only Access
ESB90P0075

11.3.1.6.1

Port Command Queue 0 Control Register (PCQ0CR)

SHAC I/O Address: 2000 428016
When the port driver inserts an entry in an empty CMDQ0, the port driver writes PCQ0CR to
initiate port execution of the Command Queue. PCQ0CR can be written only when the port
is in the enabled or enabled/maintenance state. Writing to PCQ0CR when the port is in any
other state has no effect. The SHAC I/O Address is 2000 428016

11.3.1.6.2

Port Command Queue 1 Control Register (PCQ1CR)

SHAC I/O Address: 2000 428416
Same as PCQ0CR except refers to CMDQ1. The SHAC I/O Address is 2000 428416

KA660 Mass Storage Interface 11–15

11.3.1.6.3

Port Command Queue 2 Control Register (PCQ2CR)

SHAC I/O Address: 2000 428816
Same as PCQ0CR except refers to CMDQ2. The SHAC I/O Address is 2000 428816

11.3.1.6.4

Port Command Queue 3 Control Register (PCQ3CR)

SHAC I/O Address: 2000 428C16
Same as PCQ0CR except refers to CMDQ3. The SHAC I/O Address is 2000 428C16 .

11.3.1.6.5

Port Datagram Free Queue Control Register (PDFQCR)

SHAC I/O Address: 2000 429016
When the port driver inserts an entry on the DFREEQ and the latter was previously empty,
the port driver writes PDFQCR to indicate the availability of DFREEQ entries. PDFQCR
can be written only if the port is in the enabled or enabled/maintenance State. Writing to
PDFQCR when the port is in any other state has no effect. The SHAC I/O Address is 2000
429016

11.3.1.6.6

Port Message Free Queue Control Register (PMFQCR)

SHAC I/O Address: 2000 429416
Same as PDFQCR except refers to MFREEQ. The SHAC I/O Address is 2000 429416

11.3.1.6.7

Port Status Release Control Register (PSRCR)

SHAC I/O Address: 2000 429816
After the port driver has received an interrupt and read the PSR, it returns the PSR to the
port by writing PSRCR. The SHAC I/O Address is 2000 429816

11.3.1.6.8

Port Enable Control Register (PECR)

SHAC I/O Address: 2000 429C16
The port driver enables the port by writing PECR. PECR is ignored if the port is in the
uninitialized , uninitialized/maintenance , enabled , or enabled/maintenance state. The
SHAC I/O Address is 2000 429C16

11–16

KA660 Mass Storage Interface

11.3.1.6.9

Port Disable Control Register (PDCR)

SHAC I/O Address: 2000 42A016
The port driver disables the port by writing PDCR. When the port is disabled, the port sets
PDC (PSR <2>) and if interrupts are enabled requests an interrupt. PDCR is ignored if the
port is in the uninitialized, uninitialized/maintenance, disabled, or disabled/maintenance
state. The SHAC I/O Address is 2000 42A016

11.3.1.6.10

Port Initialize Control Register (PICR)

SHAC I/O Address: 2000 42A416
The port driver initializes the port by writing PICR. When the initialization is complete the
port sets PDC (PSR <2>) and requests an interrupt if interrupts are enabled. As part of the
initialization, the maintenance timer is set to expire in 100 seconds. The SHAC I/O Address is
2000 42A416

11.3.1.6.11

Port Maintenance Timer Control Register (PMTCR)

SHAC I/O Address: 2000 42A816
The port driver forces the maintenance timer to reset its expiration time by writing the
PMTCR. If the PMTCR is not written again before the expiration time, the port will enter
the uninitialized/maintenance state setting MTE (PSR <6>) and request an interrupt if
interrupts are enabled. PMTCR is ignored if the maintenance timer is not running. The
SHAC I/O Address is 2000 42A816

11.3.1.6.12

Port Maintenance Timer Expiration Control Register (PMTECR)

SHAC I/O Address: 2000 42AC16
The port driver forces a Maintenance-Timer-Expiration Interrupt by writing the PMTECR.
This register may be written only when the port is in the enabled, enabled/maintenance,
disabled, and disabled/maintenance states and only while the Maintenance Timer is not
disabled. The SHAC I/O Address is 2000 42AC16

11.3.1.6.13

Port Maintenance Control and Status Register (PMCSR)

SHAC I/O Address: 2000 425C16
The Port Maintenance Control and Status Register (PMCSR) is used for maintenance level
control and status reporting. The CI Port specification defines all but the 2 least significant
bits. The format is shown in Figure 11–9 and the bit descriptions are listed in Table 11–6

KA660 Mass Storage Interface 11–17

The bits can be divided into the following categories:
•

Status bits - Which are set by the port to report various conditions. They are cleared by
maintenance initialization or clearing the condition in another register. PMCSR does not
include any status bits at this time.
Function control bits are read/write by the port driver only. They are clear on a RESET.
These bits are of two classes:
1. Init: this type of bit invokes a function (for example, initialization) by setting it. It
always reads as zero, except while the function is active.
2. Enable/disable: this type of bit causes an activity or state to exist while the bit is set.
Clearing the bit stops the activity or changes the state. The bit always reads the most
recently written value. The bit is never changed by the port.

•

Figure 11–9: Port Maintenance Control And Status Register (PMCSR)
3
1

5 4 3 2 1 0
RESERVED
MIN
MTD
IE
SIMP
HAC

Longword Read/Write Access
ESB90P0076

Table 11–6: Port Maintenance Control and Status Register (PMCSR) Bits
Data Bit

Name

Description

<31:5>

RESERVED

This bits are reserved. They should not be written; reads return
UNPREDICTABLE results.

<4>

HAC

Host Access Feature. This bit Must Be ZERO, except for diagnostic
purposes. This is an enable/disable class control bit.

<3>

SIMP

Simple SHAC Mode. Must Be ZERO, except for diagnostic purposes.
This is an enable/disable class control bit.

<2>

Interrupt Enable. When set, interrupts from the port to the SOC
CPU are enabled. Power-up state is clear (interrupts disabled). This
is an enable/disable class control bit.

11–18

KA660 Mass Storage Interface

Table 11–6 (Cont.): Port Maintenance Control and Status Register (PMCSR) Bits
Data Bit

Name

Description

<1>

MTD

Maintenance Timer Disable. Read/Write by SOC CPU . If set, the
maintenance timer is turned off. Timer is set to the initial value
and suspended. If clear, timer functions normally. Power-up state is
clear (timer enabled). This is an enable/disable class control bit.

<0>

MIN

Maintenance Init. Writing a ONE to this bit resets the port. Upon
completion, the port is in the uninitialized state and MIN is clear.
Writing a ZERO to this bit has no effect. It always reads as ZERO,
except while the reset function is active.
Although Maintenance Init resets the port, it is not equivalent to a
write to the SHAC Software Chip Reset register. In particular, the
SHAC Shared Host Memory Address is not reset by Maintenance
Init.

11.3.2 SHAC Specific Registers
These registers, which are not defined in the CI Port Architecture, are used for additional
maintenance level control.

11.3.2.1

SHAC Software Chip Reset Register (SSWCR)

SHAC I/O Address: 2000 423016
When the SOC CPU writes FFFF FFFF16 to the SHAC Software Chip Reset register (SSWCR),
a chip reset is performed. The result is equivalent to that of the hardware chip reset that
occurs following system power-up. On completion, all device registers are reset to their powerup state, and the port is in the uninitialized state. The format is shown in Figure 11–10.
SSWCR is write only by the SOC CPU and may be written to at any time. Its value when
read is UNPREDICTABLE. The result if other than FFFF FFFF16 is written to SSWCR is
UNDEFINED.

KA660 Mass Storage Interface 11–19

Figure 11–10: SHAC Software Chip Reset (SSWCR)

3
1

0
MUST BE ONE

Longword Write Only Access
ESB90P0077

11.3.2.2

SHAC Shared Host Memory Address (SSHMA)

SHAC I/O Address: 2000 424416
The format for the SHAC Shared Host Memory Address is shown in Figure 11–11.

Figure 11–11: SHAC Shared Host Memory Address (SSHMA)

3 3 2
1 0 9
MBZ

4 3
SSHMA<29:4>

0
MBZ

Longword Read/Write Access
ESB90P0078

Following chip reset, the SOC CPU writes into the SHAC Shared Memory Address register (SSHMA) the physical address of the Shared Host Memory Header. The area must be
octaword aligned and contiguous in physical memory.
SSHMA is read/write by the SOC CPU , but may be written only when the port is in the
uninitialized state. Writing when the port is in any other state can produce unpredictable
results.

11–20

KA660 Mass Storage Interface

Chapter 12
KA660 Firmware

This document describes the KA660 functional firmware. The firmware is VAX–11 code which
resides in EPROM on the KA660 module. Typically KA660 firmware gains control whenever
the onboard CPU "halts", or more precisely, performs a "processor restart" operation. However,
portions of the firmware can also be invoked by applications through a public subroutine
linkage.
When the KA660 firmware is running, it provides services expected of a standard VAX console
subsystem. In particular, the following services are available:
•
•
•

Automatic restart or bootstrap of customer application images at power-up, on reset, or
conditionally after processor halts.
Diagnostic tests executed both at power-up and by request, which verify the correct operation of the CPU and memory modules.
Operator interface providing complete examination or modification of the processor state.

Throughout this document, "firmware" is a generic term describing all program code located in
the KA660 EPROM. Sometimes it is referred to as either the "boot ROM", "diagnostics ROM",
or "console ROM", depending on context. Each major element of the firmware is referred to by
other terms, for instance, the boot program as "VMB" or "primary bootstrap", the ROM based
diagnostic program as the "diagnostic" or "self-test", and the operator interface as the "console"
or "console program".
Certain terminology and conventions are used throughout this document. With one exception,
numbers unless otherwise indicated or implied are decimal. Eight digit numbers throughout
this document are hexadecimal longwords, typically representing VAX 32 bit addresses or
data. Where there is ambiguity, the radix is explicitly stated. For instance, 72 is assumed to
be decimal and for clarity can be written as 72 (dec). However, alternate representations for
72 are 1001000 (bin) for binary, 110 (oct) for octal, or 48 (hex) for hexadecimal. On the other
hand, 20040000 is the hexadecimal address or the base of the firmware EPROM.
Ranges of integers are expressed as a pair of numbers separated by a colon and are always
inclusive. For example, 7:4 specifies the range of integers from 7 to 4, namely 7, 6, 5, and 4.
A bit field or position within a register or data structure follows the structure name and is
enclosed in angle brackets. The associated field name (if defined) typically follows the field
definition and appears in parenthesis. For instance, PSL<20:16> (IPL) represents the five bit
field for the Interrupt Priority Level in the Processor Status Longword.

KA660 Firmware 12–1

12.1 General Description
The KA660 firmware provides the following services:
•
•
•
•
•
•

Diagnostics that test all components on the board and verify the module is working correctly.
Automatic/manual bootstrap of an operating system following processor halts.
Automatic/manual restart of an operating system following processor halts.
An interactive command language that allows the user to examine and alter the state of
the processor.
Support of various terminals and devices as the system console.
Multilanguage support for displaying critical system messages and handling LK201 country specific keyboards.

The remainder of this section describes in detail the functions and external characteristics of
the KA660 firmware.

12–2 KA660 Firmware

The KA660 firmware is comprised of several major functional blocks of code. The halt entry
code is invoked following system halts, resets, or severe errors. This code is responsible
for saving the machine state and transfering control to the halt dispatcher code. The halt
dispatcher code determines the nature of the halt, then transfers control to the appropriate
subcode. The halt exit code is invoked whenever a transition is desired from a halted state to
the running state. This code performs a restoration of the saved context prior to the transition.
Figure 12–1 illustrates this and these functions are discussed in detail in Section 12.2.
Figure 12–1: KA660 Firmware Structural Components

Halt
Entry
Code
ROM Based
Diagnostics

Halt
Dispatch
Code

System
Bootstrap
System
Restart
Console
Service

Halt
Exit
Code

The ROM based diagnostics consist of an initial power-up test and a series of functional
component diagnostics invoked by a diagnostic executive. These functions are described in
Section 12.3 on power-up and in Section 12.8 on diagnostics.
Depending on the nature of the halt and the hardware context, the firmware attempts either
an operating system restart (discussed in Section 12.5), a bootstrap operation (described in
Section 12.4), or transitions to console I/O mode (covered in Section 12.6).

KA660 Firmware 12–3

12.2 Halt Code
The main purpose of the halt code is to save the state of the machine on halt entry, invoke the
dispatcher, and restore the state of the machine on exit to program I/O mode. It is comprised
of Halt Entry, Halt Dispatch, and Halt Exit codes.

12.2.1

Halt Entry - Saving Processor State

The entry code, residing at physical address 20040000, is executed whenever the KA660 halts.
The value that the program counter contained when the processor was halted is saved in IPR
42 (PR$_SAVPC). On a power-up, the PR$_SAVPC register value is undefined.
The processor will halt for a variety of reasons. The reason for the halt is stored in PR$_
SAVPSL<13:8>(RESTART_CODE), IPR 43. A complete list of the halt reasons and the associated console messages can be found in Table I–1 in Appendix I.
After a halt, the firmware first saves the current LED code then writes an "E" to the diagnostic
LEDs. This action occurs within several instructions after the firmware has been invoked. The
intent of saving the LED code is to let the user know that at least some instructions have been
successfully executed.
The KA660 firmware unconditionally saves the contents of the following registers on any halt:
•
•
•
•

R0 through R15, the general purpose registers.
PR$_SAVPSL, the saved PSL register.
PR$_SCBB, the system control block base register.
DLEDR, the diagnostic LED register.
NOTE
The SSC programmable timer registers are not saved. In some cases, such as
bootstrap, the timers are used by the firmware and previous "time" context is lost.

Several registers are unconditionally set to pre-determined values by the firmware on any
halt, processor init, or bootstrap. This action insures that the firmware itself can run and
protects the board from physical damage.
The following is a list of registers that fall into this category.
•
•
•
•

The SSC configuration register (SSCCR)
The SSC address match and mask registers (ADxMCH & ADxMSK)
The CDAL bus timeout control register (CBTCR)
The SSC timer interrupt vector registers (TIVRx)

Whenever the Halt Entry Code is invoked, the firmware sets the console serial line baud rate
based on the value read from the BDR and extends the halt protection from 8KB to 256KB to
include the all of the EPROM.

12–4 KA660 Firmware

12.2.2

Halt Dispatch

The action taken by the firmware on a halt is dependent primarily on the following information:
•
•
•
•

The state of BREAK enable switch, BDR<23>(HALT_ENABLE).
The state of the console program mailbox, CPMBX<1:0>(HALT_ACTION).
The user defined halt action (SET HALT).
The halt code, PR$_SAVPSL<13:8>(RESTART_CODE).

In general, the BREAK enable switch governs whether or not a BREAK condition from the
console serial line is recognized by the KA660. This swith also determines the default action
taken on a power-up or other internal halt condition. By default, if BREAKs are enabled, the
firmware invokes the console emulation code. If BREAKs are disabled, the firmware attempts
a recovery operation.
However, the console program mailbox, CPMBX<1:0>(HALT_ACTION). (refer to Figure G–2)
is used by operating systems to override the BREAK enable switch and instruct the firmware
to invoke the console service, attempt to restart the operating system, or reboot the system
following a halt, regardless of the setting of the BREAK enable switch.
The user defined halt action invoked by using the SET HALT console command (refer to the
description of the SET command in Section 12.7, Console commands is an alternative way
to specify a default halt action. This feature allows users to specify auto-booting on powerups, even when BREAKs are enabled. For HALT instructions and error halt conditions, it is
similar in function to the console program mailbox but has lower precedence and is only used
when the console program mailbox is 0. This provides the user with a mechanism by which to
specify what action should be taken, in the event that the operating system or user application
does not set the console program mailbox.
The halt (or restart) code is automatically deposited in PR$_SAVPSL<13:8>(RESTART_CODE)
on any halt condition. This field indicates the cause of the halt and for the purpose of dispatching collapses into three categories.
02: External halts.
03: Reset/power-up.
xx: (All other values) HALT instruction and all error halts.

Table 12–1 summarizes the action taken on all halt conditons, except external halts which are
described in Section 12.2.2.1. The actual halt dispatch state machine is described in detail in
Section H.0.1 of Appendix H.

KA660 Firmware 12–5

Table 12–1: Halt Action Summary
Halt
Code=
3

Break
Enable
Switch

User
Defined
Halt
Action

Console
Program
Mailbox

Action(s)

0,1,3

diagnostics, console

2,4

diagnostics, if success boot, if either fail
console

console

restart, if this fails boot, if that fails
console

restart, if it fails console

boot, if it fails console

console

restart, if this fails boot, if that fails
console

restart, if it fails console

boot, if it fails console

console

"T" TRUE—indicates a Reset or Power-up condition.
"F" FALSE—indicates a HALT instruction or error halt condition.
"x" DON’T CARE—indicates that the condition is "don’t care".

Because the KA660 does not support battery backed up main memory, an operating system
restart operation is not attempted on a power-up.
12.2.2.1

External Halts

Several conditions can trigger an external halt and different actions are taken depending on
the condition.
An external halt can be caused by one of the following conditions.
1. A BREAK condition on the system console serial line, if the BREAK enable switch is
set to "enabled". In this case BDR<23>(HALT_ENABLE) = 1 and the console code is
invoked. Control-P may be established as the "BREAK" condition by using the
SET CONTROLP ENABLE console command.
2. The assertion of the BHALT line on the Q22-bus, causes an external halt if the
SCR<14>(BHALT_ENABLE) bit in the CQBIC is set. As a result, the console code is
invoked.
3. The negation of DCOK on the Q22-bus, if the SCR<7>(DCOK_ACTION) bit is set causes
an external halt. (by default this bit is clear). As a result, the console code is invoked.

12–6 KA660 Firmware

4. Recognition of a valid MOP BOOT message by an appropriately initialized SGEC, if the
REMOTE_BOOT_ENABLE jumper is in place (BDR<12>(REMOTE_BOOT_ENABLE) =
1). As a result, a bootstrap is attempted and if that fails, the console is entered.
NOTE
The firmware does not initialize the SGEC for this operation. The operating
system must set up the SGEC to support this feature.
NOTE
The switch labeled "RESTART" negates DCOK. The DCOK bit may also be negated
by the DEQNA sanity timer, or any other Q22-bus module that chooses to implement
the Q22-bus restart/reboot protocol. Because the SCR<7>(DCOK_ACTION) bit is
cleared on power-up, the default consequence to deasserting DCOK is to generate
a processor restart. Hence, pushing the "RESTART" button typically initiates a
power-up sequence and destroys system state.

12.2.3

Halt Exit - Restoring Processor state

When the firmware exits, it uses the currently defined saved context. This context is initially
determined by what was saved when the firmware code was invoked. However this context
may be modified by console commands, or automatic operations such as an automatic bootstrap
on power-up.
When restoring the context, the firmware will flush the CPU internal cache if enabled, and
invalidate all translation buffer entries via the internal processor register PR$_TBIA, IPR 57.
In restoring the context, the console pushes the user’s PSL and PC onto the user’s interrupt
stack, then executes a return from exception or interrupt instruction (REI) from that stack.
This implies that the user’s Interrupt Stack Pointer (ISP) is valid before the firmware can exit.
This is done automatically on a bootstrap. However, it is suggested that the stack pointer
(SP) be set to a valid memory location before issuing the START or CONTINUE command.
Furthermore, the user should validate the System Control Block Base Register (SCBB or PR$_
SCBB) prior to executing a NEXT command, because the firmware uses the trace trap vector
for this function. At power-up, the user ISP is set to 200 (hex) and the SYstem Control Block
Base Register is undefined.

KA660 Firmware 12–7