Intel 8255x 10/100 Mbps Ethernet
Controller Family
Open Source Software Developer Manual
January 2006
Revision 1.3
Download from Www.Somanuals.com. All Manuals Search And Download.
Contents
Contents
Introduction....................................................................................................................................1
Scope....................................................................................................................................1
Document Conventions ........................................................................................................2
1.2.1 Device References ..................................................................................................2
1.2.2 Numbering ...............................................................................................................2
1.2.3 Signal Name Representation...................................................................................2
1.2.4 Memory Alignment Terminology ..............................................................................2
Adapter and Controller Overview ................................................................................................5
Adapter Block Diagram.........................................................................................................5
Intel Fast Ethernet MAC Features ........................................................................................6
2.2.1 82557 Features........................................................................................................6
2.2.2 82558 Features........................................................................................................6
2.2.3 82559, 82550, 82551, and 82562 Features.............................................................7
Working with the Physical Layer...........................................................................................7
Power Management Interface.......................................................................................................9
Low Power Mode Requirements...........................................................................................9
Device Power States ............................................................................................................9
Power Management Registers .............................................................................................9
Link Operation ....................................................................................................................10
PCI Interface.................................................................................................................................11
PCI Configuration Space ....................................................................................................11
4.1.1 Vendor ID (Offset 0)...............................................................................................12
4.1.2 Device ID (Offset 2) ...............................................................................................12
4.1.3 Command Register (Offset 4)................................................................................12
4.1.4 Status Register (Offset 6) ......................................................................................12
4.1.5 Revision (Offset 8) .................................................................................................13
4.1.6 Class Code (Offset 9) ............................................................................................14
4.1.7 Cache Line Size (Offset C) ....................................................................................14
4.1.8 Latency Timer (Offset D) .......................................................................................14
4.1.9 Header Type (Offset E)..........................................................................................14
4.1.10 Built in Self Test (Offset F).....................................................................................15
4.1.11 Subsystem ID (Offset 2C)......................................................................................16
4.1.12 Subsystem Vendor ID (Offset 2E) .........................................................................16
4.1.13 Expansion ROM Base Address Register (Offset 30).............................................16
4.1.14 The Capabilities Pointer (Offset 34).......................................................................17
4.1.15 Interrupt Line (Offset 3C) .......................................................................................17
4.1.16 Interrupt Pin (Offset 3D).........................................................................................17
4.1.17 Max_Lat / Min_Gnt (Offset 3E) ..............................................................................18
4.1.18 Power Management PCI Configuration Registers .................................................18
PCI Command Usage.........................................................................................................21
4.2.1 Memory Write and Invalidate .................................................................................22
4.2.2 Read Align .............................................................................................................23
4.2.3 Odd Byte Alignment Support .................................................................................23
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
iii
Download from Www.Somanuals.com. All Manuals Search And Download.
Contents
EEPROM Interface.......................................................................................................................25
Host Software Interface ..............................................................................................................27
The Shared Memory Architecture.......................................................................................27
Initializing the LAN Controller .............................................................................................29
6.2.1 LAN Controller Addressing Format........................................................................29
Controlling the Device.........................................................................................................31
6.3.1 Control / Status Registers (CSR)...........................................................................31
6.3.2 System Control Block (SCB)..................................................................................33
6.3.3 PORT Interface......................................................................................................43
6.3.4 EEPROM Control Register ....................................................................................45
6.3.5 Management Data Interface Control Register .......................................................49
6.3.6 Receive Byte Count Register.................................................................................51
6.3.7 Early Receive Interrupt ..........................................................................................52
6.3.8 Flow Control Register ............................................................................................53
6.3.9 Power Management Driver Register......................................................................54
6.3.10 General Control Register.......................................................................................56
6.3.11 General Status Register ........................................................................................56
Shared Memory Structures.................................................................................................57
6.4.1 Action Commands and Operating Modes..............................................................57
6.4.2 Specific Action Commands....................................................................................59
6.4.3 Receive Operation.................................................................................................99
Command Unit and Receive Unit Operation.....................................................................105
6.5.1 Starting and Completing Control Commands ......................................................105
6.5.2 Generating and Acknowledging Interrupts...........................................................105
6.5.3 Command Unit Control ........................................................................................106
6.5.4 Receive Unit Control............................................................................................108
6.5.5 Updating SCB Status...........................................................................................110
Flow Control......................................................................................................................110
6.6.1 PHY Based Flow Control.....................................................................................111
6.6.2 Frame Based Flow Control..................................................................................111
6.6.3 Priority Aware Frame Based Flow Control...........................................................115
6.6.4 Half Duplex Flow Control.....................................................................................116
Collision Backoff Modification in Switched Environments.................................................116
Physical Layer Interface ...........................................................................................................117
Management Data Interface (MDI) ...................................................................................117
MDI Register Set ..............................................................................................................118
7.2.1 Control Register: Register 0 ................................................................................119
7.2.2 Status Register: Register 1..................................................................................120
7.2.3 Identification Registers: Registers 2 and 3 ..........................................................121
7.2.4 Auto-Negotiation Advertisement Register: Register 4 .........................................122
7.2.5 Auto-Negotiation Link Partner Ability Register: Register 5 ..................................122
7.2.6 Auto-Negotiation Expansion Register: Register 6 ...............................................123
Intel 82555 Specific Registers ..........................................................................................124
7.3.1 Status and Control Register: Register 16 ............................................................124
7.3.2 Special Control Register: Register 17..................................................................125
7.3.3 Clock Synthesis Test and Control Register: Register 18.....................................126
iv
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Contents
7.3.7 Receive Symbol Error Counter: Register 22........................................................127
7.3.11 Equalizer Control and Status Register: Register 26 ............................................128
7.3.12 Special Control Register: Register 27..................................................................129
Auto-Negotiation Functionality..........................................................................................130
7.4.1 Description...........................................................................................................130
7.4.2 Parallel Detection.................................................................................................131
Vendor-Specific PHY Programming .................................................................................132
7.5.1 Intel 82555 TX PHY.............................................................................................132
7.5.2 82558 and 82559 Embedded PHY Unit...............................................................132
Programming Recommendations............................................................................................135
8.1.1 8255x Initialization ...............................................................................................135
8.1.2 PHY Detection and Initialization ..........................................................................135
8.1.3 NOS Specific Initialization....................................................................................136
Transmit Processing.........................................................................................................136
Frame Reception ..............................................................................................................136
Interrupt Processing..........................................................................................................137
Appendices
Wake-up Functionality ..............................................................................................................139
82550 and 82551QM Specific Information...............................................................................155
Figures
82557 Network Interface Card Block Diagram .............................................................................5
Command Register.....................................................................................................................12
Command Register.....................................................................................................................13
Cache Line Size..........................................................................................................................14
Base Address Register for Memory Mapping.............................................................................15
Base Address Register for I/O Mapping .....................................................................................15
Expansion ROM Base Address Register....................................................................................17
8255x Memory Architecture........................................................................................................28
SCB Status Word........................................................................................................................34
10 SCB Command Word .................................................................................................................36
11 Self-Test Results Format............................................................................................................44
12 EEPROM Control Register .........................................................................................................46
13 EEPROM Read Timing Diagram ................................................................................................48
14 General Action Command Format..............................................................................................58
15 NOP Command Format..............................................................................................................59
16 Individual Address Setup Command Format..............................................................................60
17 Configure Command Format ......................................................................................................62
18 Multicast Setup Command Format .............................................................................................82
19 Transmit Command Format........................................................................................................83
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
v
Download from Www.Somanuals.com. All Manuals Search And Download.
Contents
20 Transmit Buffer Descriptor..........................................................................................................85
21 Load Microcode Command Format............................................................................................90
22 Dump Command Format............................................................................................................91
23 Diagnose Command Format ......................................................................................................97
24 Simplified Memory Structure ....................................................................................................100
25 Receive Frame Descriptor Format ...........................................................................................100
26 Management Frame Structure..................................................................................................118
27 Command Block Structure........................................................................................................146
Tables
PCI Configuration Space............................................................................................................11
Device and Revision ID ..............................................................................................................13
Base Address Register Summary ..............................................................................................16
Power Management Capabilities................................................................................................18
Power Management Control/Status Register .............................................................................20
Power Consumption / Dissipation Reporting..............................................................................21
Generated PCI Commands ........................................................................................................22
Reset Commands.......................................................................................................................29
Device Addressing Formats .......................................................................................................30
10 Alignment Requirements for 8255x Data Structures ..................................................................31
11 Control / Status Register ............................................................................................................32
12 System Control Block .................................................................................................................34
13 SCB Status Word Bits Descriptions............................................................................................35
14 SCB Command Word Bits Descriptions .....................................................................................37
15 SCB General Pointer for the CU Command...............................................................................39
16 SCB General Pointer for the RU Command...............................................................................40
17 Statistical Counters.....................................................................................................................40
18 Port Register Location................................................................................................................43
19 Port Selection Function ..............................................................................................................43
20 Dump Wake-up Data Structure ..................................................................................................45
21 EEPROM Control Register Locations.........................................................................................46
22 EEPROM Control Register Bits Definitions ................................................................................46
23 EEPROM Opcode Summary (64-register EEPROM).................................................................47
24 MDI Control Register Location ...................................................................................................49
25 Management Data Pins..............................................................................................................50
26 MDI Control Register Bits...........................................................................................................50
27 Receive Byte Count Register Location.......................................................................................52
28 Early Receive Interrupt Register Location..................................................................................52
29 Flow Control Registers Location.................................................................................................53
30 Flow Control Threshold Values ..................................................................................................54
31 Power Management Driver Register Location............................................................................55
32 Power Management Driver Register ..........................................................................................55
33 General Control Register Location .............................................................................................56
34 General Control Register............................................................................................................56
35 General Status Register Location...............................................................................................56
36 General Status Register .............................................................................................................57
37 Operation Codes ........................................................................................................................57
38 82557 Configuration Byte Map...................................................................................................62
39 82558 Configuration Byte Map...................................................................................................64
vi
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Contents
40 82559 Configuration Byte Map ...................................................................................................65
41 82557 Dual-Port FIFO Settings - Transmit .................................................................................66
42 82557 Dual-Port FIFO Settings - Receive ..................................................................................67
43 82558 and 82559 Dual-Port FIFO Settings - Transmit ...............................................................67
44 82558 and 82559 Dual-Port FIFO Settings - Receive ................................................................68
45 Extended Statistics Functionality................................................................................................72
46 Pre-amble Length .......................................................................................................................75
47 82558 B-step Configuration Block ARP Frame IP Address........................................................76
48 82558 B-step ARP Frame IP Address Mapping .........................................................................77
49 Full Duplex Functionality.............................................................................................................79
50 Dump Data Bytes (0-79).............................................................................................................92
51 Dump Data Dwords (20-148)......................................................................................................95
52 RFD Status Bit Descriptions .....................................................................................................101
53 Actual Count in Header RFD ....................................................................................................102
54 CU Control Commands: Actions at Acceptance Time..............................................................107
55 CU Activities Performed at the End of Execution .....................................................................107
56 RU Control Commands: Actions at Acceptance Time..............................................................108
57 Flow Control Frame Format......................................................................................................112
58 Flow Control Configuration Bits ................................................................................................115
59 MDI Register Set ......................................................................................................................118
60 82555 MDI Register Set ...........................................................................................................118
61 24-bit OUI Identification Number ..............................................................................................121
62 MDI Identification Registers 2 and 3: PHY ID Encoding...........................................................121
63 LED Switch Control...................................................................................................................130
64 Technology Ability Field Bit Assignments.................................................................................131
65 Technology Priority...................................................................................................................131
66 Fixed Wake-up Configuration Bits ............................................................................................143
67 82559 Port Commands.............................................................................................................150
68 Dump Data Structure................................................................................................................151
69 IPCB Structure..........................................................................................................................155
70 IP Activation Bits (Byte 13) .......................................................................................................155
71 IP Activation Bits (Byte 12) .......................................................................................................155
72 IPCB Fields...............................................................................................................................156
73 IPCB Structure Checksum Offload ...........................................................................................158
74 IPCB Structure Large Send ......................................................................................................162
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
vii
Download from Www.Somanuals.com. All Manuals Search And Download.
Contents
Revision History
Date
Revision
Description
•
•
Added Section 2.2.3.2., “82551ER Features.”
Modified the title of Appendix B.
January 2006
1.3
1.2
September 2005
•
•
Corrected minor typing errors.
•
•
September 2004
January 2003
1.1
1.0
Included descriptions for the Reset bit, Auto-Negotiation
•
Included a summary paragraph before the table of bits in
Initial release.
viii
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Introduction
1
®
This document is intended for use as a software technical reference manual for the Intel 10/100
Mbps Fast Ethernet controller family, which includes the 82557, 82558, 82559, 82550, and 82551,
as well as the 82562 Platform LAN Connect device. It also contains information for several PCI
®
™
®
™
LAN adapters based on these devices: Intel EtherExpress PRO/100+, Intel EtherExpress
®
™
®
™
PRO/100B Wake on LAN (WOL), Intel EtherExpress PRO/100B, and Intel EtherExpress
PRO/10+.
1.1
Scope
This manual is intended to be used as a technical reference for software and test engineers
®
developing device drivers or related software for adapters or systems using the Intel 82557,
®
82558, 82559, 82550, or 82551 Fast Ethernet controllers or the Intel 82562 Platform LAN
Connect (PLC) device. It contains reference information about the controllers as well as other
information that may be required by software developers (such as PHY information, EEPROM
contents, PCI scanning, etc.). Since this document uses many examples and contains sample code
fragments, it is assumed that the reader has a fundamental understanding of device driver
programming and a working knowledge of both C programming language and x86 assembler
programming language. Familiarity with at least one industry standard network operating system
(NOS) device driver interface (for example, Network Driver Interface Specification [NDIS] or
ODI) is also helpful.
®
The Intel 10/100 Mbps Fast Ethernet Controller Family includes the following devices in
successive order.
Device
Notes
82557
First generation Intel® 10/100 Mbps Fast Ethernet Controller (includes MAC unit only)
Second generation Intel® 10/100 Mbps Fast Ethernet Controller (includes MAC and an integrated
PHY unit)
82558
82559
82550
82551
Third generation Intel® 10/100 Mbps Fast Ethernet Controller (includes both a MAC and PHY unit)
Intel® 10/100 Mbps Fast Ethernet Controller (includes both MAC and PHY)
Intel® 10/100 Mbps Fast Ethernet Controller (includes both MAC and PHY)
In general, the Intel family of Fast Ethernet controllers are similar. All family members share the
same core hardware and software interface. The later generation components have a higher
integration and include support for miscellaneous features (for example, manageability). Since the
different generations of Fast Ethernet controllers are highly similar, this manual documents the
functionality of all devices and details the differences between the devices. It is intended to be used
as a tool to maintain and develop software for all devices in the Intel family of Fast Ethernet
controllers.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
1
Download from Www.Somanuals.com. All Manuals Search And Download.
Introduction
1.2
Document Conventions
1.2.1
Device References
This document encompasses information for all members of the Intel Fast Ethernet controllers:
82551, 82550, 82559, 82558, 82557 and the 82562.
Note: The 82562xx/ICHx combination are programmed like the 82559 using the same byte map and
settings.
The document convention, “8255x,” will be used to refer to all devices. In addition, there are
specific references to the 82557 throughout this manual that pertains to all 8255x devices. Device-
specific differences and exceptions will be documented.
1.2.2
Numbering
Decimal, binary, and hexadecimal numbers are used through the manual. They will be designated
as follows:
• Decimal numbers: Decimal numbers will not be followed by a suffix.
• Binary numbers: Binary numbers (base 2) will be followed by a “b” (for example, 01b).
• Hexadecimal numbers: Hexadecimal numbers (base 16) will be followed with the suffix “h”
(for example, 1Ch). Hexadecimal numbers may also be noted with a prefix of “0x” (for
example, 0x1c).
1.2.3
1.2.4
Signal Name Representation
Signals that are active in a low logic state when asserted are followed by the pound sign (#). For
example, FRAME# is asserted low by the master during a transaction. It is asserted low at the start
and duration of a transaction and de-asserted during the final data phase.
Signals that are not followed by a pound sign are active in a high logic state when asserted. For
example, the IDSEL signal is asserted high when the 82559 during PCI read and write transactions.
Memory Alignment Terminology
The 8255x data structures have special memory alignment requirements. This implies that the
starting physical address of a data structure must be aligned as specified. The following terms are
used for this purpose:
• Byte alignment: Byte alignment implies that the physical addresses can be odd or even.
Examples: 0FECBD9A1h or 02345ADC6h
• Word alignment: Word alignment implies that physical addresses must be aligned on even
boundaries. In other words, the last nibble of the address may only end in 0h, 2h, 4h, 6h, 8h,
Ah, Ch, or Eh.
Example: 0FECBD9A2h
2
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Introduction
• Dword alignment: Dword alignment implies that the physical addresses may only be aligned
on 4-byte boundaries. In other words, the last nibble of the address may only end in 0h, 4h, 8h,
or Ch.
Example: 0FECBD9A8h
• Paragraph alignment: Paragraph alignment implies that the physical addresses may only be
aligned on 16-byte boundaries. In other words, the last nibble must be a 0.
Example: 02345ADC0h
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
3
Download from Www.Somanuals.com. All Manuals Search And Download.
Introduction
4
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Adapter and Controller Overview 2
®
Adapters based on an Intel 8255x device support the ANSI/IEEE 802.3u standard for 100BASE-
TX (100 Mbps operation) and 10BASE-T (10 Mbps operation).
2.1
Adapter Block Diagram
The main components of Intel Fast Ethernet adapters are:
• A Fast Ethernet Media Access Controller (MAC), such as the 8255x, is the core component.
The MAC supports the Fast Ethernet ANSI/IEEE 802.3u standard.
• A Physical Layer (PHY) interface device is also required. The 82558, 82559, 82550, and
82551 components have an integrated PHY that supports 100BASE-TX and 10BASE-T.
Adapters based on the 82557 must include an appropriate PHY component for their design.
• A serial EEPROM is required to hold the adapter’s individual Ethernet node address and other
configuration information including fixed PCI configuration parameters.
The adapters are based on 100BASE-TX specifications. 100BASE-TX is a specific scheme
designed for use over 2 pairs of Category 5 unshielded twisted-pair cable. 100BASE-TX defines a
signaling scheme for 100 Mbps and provides compatibility with the existing 10 Mbps IEEE 802.3
10BASE-T signaling standard. Since only 2-wire pairs are used, TX technology allows full duplex
operation at 100 Mbps. The Intel 82555 is one possible TX solution.
®
The block diagram below illustrates an Intel PRO/100B adapter configuration based on the 82557
MAC with a TX or T4 PHY.
Figure 1. 82557 Network Interface Card Block Diagram
Filter
Module
100BASE-T4 or
100BASE-TX PHY
RJ-45
MII
Optional
Flash
Intel® 82557
EEPROM
PCI Local Bus
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
5
Download from Www.Somanuals.com. All Manuals Search And Download.
Adapter and Controller Overview
2.2
Intel Fast Ethernet MAC Features
2.2.1
82557 Features
• Glueless 32-bit, zero wait state PCI bus master interface compliant with PCI Specification,
Revision 2.1.
• 10 and 100 Mbps support in compliance with IEEE 802.3 10BASE-T and 802.3u 100BASE-
TX.
• Fast back-to-back transmit interframe spacing (IFS) of 960 ns in 100 Mbps networks and 9.6
µs in 10 Mbps networks.
• On-chip Control/Status Register (CSR) incorporating the System Control Block (SCB).
• Simple and flexible packet support with Dynamic transmit chaining.
• Packed Transmit Buffer Descriptors (TBDs).
• Early transmit complete indication.
• Simple receive packet support allows early receive interrupt support for concurrent processing
(in simplified mode).
• IEEE Media Independent Interface (MII) compliant PHY interface other MII compliant PHYs.
• Full and half duplex transmit and receive capability.
• Separate on-chip receive and transmit FIFOs.
• On-chip network management counters.
• EEPROM support.
• Optional Flash ROM support (256 Kbytes or 1 Mbyte).
2.2.2
82558 Features
For the most part, the 82558 is a superset of the 82557. In addition to incorporating the features of
the 82557, it also includes the following:
• Backward compatible to 82557 software.
• Integrated 100BASE-TX PHY.
• IEEE 802.3u auto-negotiation support in 10BASE-T, 100BASE-TX, full duplex and full
duplex flow control configurations.
• Auto-polarity correction for 10BASE-T.
• Optimized PCI interface with support for the memory write and invalidate PCI command.
• Automatic read of EEPROM (programmable ID).
• IEEE 802.3x flow control capable.
• PHY based flow control support when the internal 100BASE-TX PHY is used.
• Advanced Configuration and Power Interface (ACPI) Specification and PCI Power
Management Specification compliant.
• Remote power up support (for Magic Packet*).
6
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Adapter and Controller Overview
• Optional Flash support up to 64 Kbytes. (The 82557 is capable of larger Flash size support.)
2.2.3
82559, 82550, 82551, and 82562 Features
The 82559, 82550, and 82551 devices are supersets of the 82557 and 82558. However, the 82559
does not support PHY based flow control as the 82558 did. The new 82559 features are:
• Backward compatible to the 82557 and 82558 software.
• Low power 3.3 V device:
• Clockrun protocol support.
• System Management Bus (SMB) support.
• Wired for Management support (WfM).
• Expanded Wake on LAN capabilities.
• 128 Kbytes Flash size support. (The 82558 only supported a 64 Kbyte Flash.)
• Thin ball grid array (BGA) 15 x 15 mm package.
2.2.3.1
2.2.3.2
82559ER Features
The 82559ER is a member of the 82559 Fast Ethernet controllers. It is a subset of the 82559.
However, the 82559ER does not support:
• SMB.
• Wake on Magic Packet*.
82551ER Features
The 82551ER is a member of the 82559 Fast Ethernet controllers. It is a subset of the 82551.
However, the 82551ER does not support:
• SMB.
• Wake on Magic Packet*.
• Checksumming.
2.3
Working with the Physical Layer
The 82557 contains an IEEE MII compliant interface to a MII compliant PHY, allowing
connections to 10/100 Mbps networks. Software communicates to a MII compliant device through
the 82557 by using the its Management Data Interface (MDI) port.
The 82558, 82559, 82550 and 82551 contain an embedded PHY module. Although the PHY is
internal for these devices, software still communicates to the PHY unit through the MDI port.
®
For 10/100 Mbps connections, the 82557 can be used in conjunction with the Intel 82555. For 10
®
Mbps only connections, the 82557 can be interfaced to the Intel 82503 serial interface, while
maintaining software compatibility to 100 Mbps solutions. The 82558 and later devices do not
have a 10 Mbps only interface as the 82557. However, it is possible to interface these devices with
a 10 Mbps only MII device.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
7
Download from Www.Somanuals.com. All Manuals Search And Download.
Adapter and Controller Overview
8
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Power Management Interface
3
The 82557 has no power management support. The 82558 added support for the Advanced
Configuration and Power Interface (ACPI) Specification and limited support for Wake on LAN
(WOL). The 82558 B-step upgraded and expanded the WOL capability, while the 82559 expanded
and simplified the WOL functionality even more.
3.1
Low Power Mode Requirements
The 82558, 82559, 82550, and 82551 adhere to the emerging power management standards as
defined in:
• PCI Bus Power Management Interface Specification, Revision 1.0.
• Advanced Configuration and Power Interface Specification (ACPI), Rev 1.0; December 22,
1996.
• Device Class Power Management Reference Specification - Network Device Class, Revision
1.0.
These three specifications define how a PCI network device can be controlled in an OS Directed
Power Management (OSPM) environment. These devices all adhere to these specifications.
Additionally, they support bus isolation within the chip and Wake on LAN (WOL) capabilities.
3.2
3.3
Device Power States
Currently, operating systems only support the D0 and D3 power states. However, starting with the
82558, the Intel Fast Ethernet controller family supports all four power states as defined in the PCI
Power Management Specification. These power states are named D0, D1, D2 and D3. D0 is the
maximum powered state, and D3, the minimum powered state.
Power Management Registers
The 82558, 82559, 82550, and 82551 support power management registers:
• Power Management Capability Pointer (Cap_Ptr)
• Power Management Capabilities (PMC)
• Power Management Control/Status Register (PMCSR)
• Power Management Driver Register (PMDR)
The first three registers are located in PCI configuration space and are defined in the PCI Power
Management Specification. It is part of the device CSR, which is mapped into system memory and
I/O space.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
9
Download from Www.Somanuals.com. All Manuals Search And Download.
Power Management Interface
3.4
Link Operation
In the D0 state, the device maintains an active link. The 82558 B-step (refer to Table 2, “Device
and Revision ID” on page 13) and later devices also maintain an active link in the D3 state if PME
is enabled and the device has power. This implies:
• 10BASE-T Mode: The device expects a normal clock input on the X1 and X2 pins. It expects
to receive normal reception on the Receive Differential Positive and Receive Differential
Negative signals (RDP/RDN pair). The device will not transmit on the Transmit Differential
Positive and Transmit Differential Negative signals (TDP/TDN pair).
• 100BASE-TX Mode: The device expects a normal clock input on the X1 and X2 pins and to
receive normal reception on the RDP/RDN pair. It transmits a continuous idle stream on the
TDP/TDN pair, as required by the 100BASE-TX standard. The 82558 does not transmit
frames on the link.
• Auto-Negotiation: If the link fails while the device is in the D1 state, it performs the normal
auto-negotiation protocol in order to re-establish the link. For the 82558 B-step, if the link fails
in the D3 state and PME is enabled and the device has power, the device will attempt to use the
normal auto-negotiation protocol in order to re-establish the link. If the link fails on the 82559
in the D3 state and PME is enabled and the device has power, the 82559 will go into a deep
power down state, rather than trying to re-establish the link with the auto-negotiation protocol.
During the D3 power state, the 82558 A-step does not maintain an active link. The 82558 B-step
and later generation devices do not maintain a link in D3 if PME is disabled or if the device does
not have power.
10
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
PCI Interface
4
4.1
PCI Configuration Space
One of the most important functions for enabling superior configurability and ease of use is the
ability to relocate PCI devices in the address spaces. By default PCI devices support “Plug and
Play.” When the system is powered on, device independent software (usually the system BIOS)
determines present devices, builds an address map, and assigns non-conflicting resources to those
devices. The device independent software accomplishes this configuration task by writing to the
PCI configuration space of each individual PCI device.
The 8255x supports 16 Dwords of Type 0 Configuration Space Header, as defined in the PCI
Specification, Revision 2.1. The 82259 and 82558 also support a small section in the device
specific configuration space. The configuration space is depicted below. The registers that are not
identical between the devices are shaded.
Table 1. PCI Configuration Space
Byte Offset
(hexadecimal)
Byte 3
Byte 2
Byte 1
Vendor ID
Byte 0
0
4
Device ID
Status Register
Class Code (200000h)
BIST
Command Register
8
Revision ID
Cache Line Size
C
Header Type
Latency Timer
10
14
18
1C
20
24
28
2C
30
34
38
3C
DC
E0
CSR Memory Mapped Base Address Register
CSR I/O Mapped Base Address Register
Flash Memory Mapped Base Address Register
Reserved
Subsystem ID
Subsystem Vendor ID
Expansion ROM Base Address Register
Reserved
Reserved
Cap_Ptr
Max_Latency (FFh)
Power Management Capabilities
Reserved Data
Min_Grant (FFh)
Interrupt Pin (01h)
Next Item Pointer
Interrupt Line
Capability ID
Power Management CSR
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
11
Download from Www.Somanuals.com. All Manuals Search And Download.
PCI Interface
4.1.1
Vendor ID (Offset 0)
This field identifies the device manufacturer. For the 82557 B-step this field equals 8086h. For the
82557 C-Step, 82558, and 82559, this field is automatically loaded from the EEPROM at power on
or upon the assertion of PCI reset. If the EEPROM is not present or invalid, this value defaults to
8086h.
4.1.2
4.1.3
Device ID (Offset 2)
This field uniquely identifies the device. For the 82557 B-step this field is 1229h. For the 82557 C-
Step, 82558, and 82559, this field is automatically loaded from the EEPROM at power on or upon
the assertion of PCI reset. If the EEPROM is not present or invalid, this value defaults to 1229h for
the 82558 and 82559. The 82559ER does not load the Device ID from the EEPROM and will
always equal 1209h.
Command Register (Offset 4)
The Command Register provides control over the device’s ability to generate and respond to CPU
cycles. Its layout is shown below. The shaded bits are not used and are hard-wired to 0.
Figure 2. Command Register
15
10
9
0
Reserved
Command Bits
Initial
Value
Bits
Description
15:10
0
0
x
0
x
0
Reserved.
9
8
7
6
5
Fast back-to-back enable.
SERR# enable.
Wait cycle enable.
Parity error response
Palette snoop enable.
Memory write and invalidate (MWI) enable.
4
x
3
2
1
0
0
x
x
x
Special cycle monitoring.
Mastering enable.
Memory access enable.
I/O access enable.
4.1.4
Status Register (Offset 6)
The Status Register is used to record status information for PCI bus related events. Its layout is
shown below. The shaded bits are not used and are hard-wired to 0.
12
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
PCI Interface
Figure 3. Command Register
15
4
3
0
Status Bits
Reserved
Initial
Bits
Description
Value
15
14
13
12
11
10:9
8
x
x
Detected parity error.
Signaled system error.
Received master abort.
Received target abort.
Signaled target abort.
x
x
0
01
x
DEVSEL timing (indicates minimum timing).
Data parity reported.
7
1
Fast back-to-back capable.
UDF supported.
6
0
5
0
66 MHz capable.
1(82559
and
82558)
Capabilities list. This bit indicates whether the device implements a list of new capabilities
such as PCI Power Management. If it is set, the Cap_Ptr register in the PCI Configuration
Space points to the location of the first item in the Capabilities List.
4
NOTE: This bit is set to 1 for the 82559 and 82558 if it is not disabled by the EEPROM. It
is always equal to 0 for the 82557.
0
(82557)
3:0
0
Reserved.
4.1.5
Revision (Offset 8)
This register specifies a device specific revision identifier. For the 82557 C-Step, 82558, and
82559, this field may be automatically loaded from the EEPROM at power on or upon the assertion
of a PCI reset. The default revision register values for the various devices are:
Table 2. Device and Revision ID
PCI Revision
Supported
Intel Driver
Supported
Device
Revision ID
82557 A-Step
01h
2.0
2.0
2.1
2.1
2.1
2.1
2.1
2.2
Yes
82557 B-Step
82557 C-Step
82558 A-Step
82558 B-Step
82559 A-Step
82559 B-Step
82559 C-Step
02h
03h
04h
05h
06h
07h
08h
Yes
No
Yes
Yes
No
No
Yes
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
13
Download from Www.Somanuals.com. All Manuals Search And Download.
PCI Interface
Table 2. Device and Revision ID
PCI Revision
Supported
Intel Driver
Supported
Device
Revision ID
82559ER A-Step
82550
09h
2.2
Yes
0Ch, 0Dh, 0Eh 2.2
0Fh, 10h 2.2
Yes
Yes
82551
4.1.6
4.1.7
Class Code (Offset 9)
The class code, 020000h, identifies the device as an Ethernet adapter.
Cache Line Size (Offset C)
This register specifies the system cache line size in units of 32-bit words and can be read or written
to. The system BIOS or OS should initialize this register at power on or after a PCI reset.
The 82557 does not support Memory Write and Invalidate (MWI) and therefore returns 0 when this
register is read. The 82258 and 82559 support the MW I command and must support this register.
The 82558 and 82559 can only support cache line sizes of 8 and 16 Dwords. Any value other than
8 or 16 written to the register is ignored, and the device does not use the MWI command. If a value
other than 8 or 16 is written into the Cache Line Size (CLS) register, the device returns all zeroes
when the CLS register is read.
Figure 4. Cache Line Size
7
6
5
4
3
2
1
0
0
0
0
RW RW
0
0
0
Bit 3 is set to 1 only if the value 00001000b (8) is written to this register. Bit 4 is set to 1 only if the
value 00010000b (16) is written to this register. All other bits are read only and will return 0 on
read.
4.1.8
4.1.9
Latency Timer (Offset D)
This register specifies, in units of PCI bus clocks, the minimum time that a bus master can retain
ownership of the bus. This value is set by the PCI bus arbitrator based on the values in the
maximum latency (Max_Lat) and Maximum Grant (Max_Gnt) registers.
Header Type (Offset E)
This byte field identifies the layout of the second part of the predefined configuration space header
and if the device is a multi-function component. The 82557 and 82558 are both single function
devices and have this register hard-coded to 00h. For the 82559, the value of this register is
determined by a bit in the EEPROM. This register should read 00h for a standard Ethernet adapter,
00h.
14
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
PCI Interface
4.1.10
Built in Self Test (Offset F)
This optional register is used for control and status of Built in Self Test (BIST). This register is
hard-wired to 0 indicating that the devices do not support BIST.
Three base address registers are supported by the 8255x:
• CSR Memory Mapped Base Address Register (BAR 0 at offset 10)
• CSR I/O Mapped Base Address Register (BAR 1 at offset 14)
• Flash Memory Mapped Base Address Register (BAR 2 at offset 18)
Two request memory mapped resources, and the third, I/O mapping. Each register is 32 bits wide.
The least significant bit in each base address register determines whether it represents an I/O or
memory space. The figures below illustrate layouts for I/O and memory mapped base address
registers. After determining which resources will be used, the power-up software maps the I/O and
memory controllers into available locations and continues with the power up. To perform the
mapping in a device independent manner, the base registers are placed in the predefined header
portion of configuration space. Device drivers access this configuration space to determine the
mapping of a particular device.
Figure 5. Base Address Register for Memory Mapping
32
4
3
0
Base Address
Configuration Bits
Initial
Bits
Description
Value
31:4
3
x
x
Base Address.
Pre-fetchable.
00 = Locate address anywhere in 32-bit address space.
01 = Locate address below 1 MByte.
2:1
0
x
10 = Locate address anywhere in 64-bit address space.
11 = Reserved.
0
Memory space indicator.
NOTE: Bit 0 in all base registers is read-only and used to determine whether the register maps into memory or
I/O space. Base registers mapping to memory space must return a 0 in bit 0, and base registers
mapping to I/O space, a 1.
Figure 6. Base Address Register for I/O Mapping
32
2
1
0
Base Address
Reserved
1
NOTE: Base registers that map into I/O space are always 32 bits with bit 0 hard wired to a 1, bit 1 is reserved
and must return 0 on reads, and the other bits are used to map the device into I/O space.
The number of upper bits that a device actually implements depends on how much of the address
space the device responds to. A device that wants a 1 Mbyte memory address space would set the
most significant 12 bits of the base address register to be configurable, setting the other bits to 0.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
15
Download from Www.Somanuals.com. All Manuals Search And Download.
PCI Interface
The 8255x requires one BAR for I/O mapping and one BAR for memory mapping of these
registers anywhere within the 32-bit memory address space. The driver determines which BAR (I/
O or Memory) is used to access the Control/Status Registers. However, both are always requested
by the device.
One BAR is also required to map the accesses to an optional Flash memory. The 82557 implements
this register regardless of the presence or absence of a Flash chip on the adapter. The 82558 and
82559 only implement this register if a bit is set in the EEPROM. The size of the space requested
by this register is 1Mbyte, and it is always mapped anywhere in the 32-bit memory address space.
Note: Although the 82558 only supports up to 64 Kbytes of Flash memory and the 82559 only supports
128 Kbytes of Flash memory, 1 Mbyte of address space is still requested. Software should not
access Flash addresses above 64 Kbytes for the 82558 or 128 Kbytes for the 82559 because Flash
the base address registers.
Table 3. Base Address Register Summary
Register
Description
Location
Memory space for the device Control/Status Registers. The size of this space is 4 Kbytes
10h
and it is mapped anywhere in the 32-bit memory address space. It is marked as pre-
fetchable. Software should not assume that this memory will be granted below 1 Mbyte.
14h
I/O space for the device Control/Status Registers. The size of this space is 32 bytes.
Memory space for FLASH buffer accesses. The size of this space is 1Mbyte. It is mapped
anywhere in the 32-bit address space and is not pre-fetchable.
18h
1Ch - 27h
Reserved.
4.1.11
4.1.12
4.1.13
Subsystem ID (Offset 2C)
This register uniquely identifies the add-in adapter or subsystem where the PCI device resides. It
provides a mechanism to distinguish different adapters that use the same PCI controller. For the
82557 B-step this field equals 0000h. For the 82557 C-Step and later devices, this field is loaded
from the EEPROM at power on or upon the assertion of PCI reset. If the EEPROM is not present or
invalid, this value defaults to 0000h.
Subsystem Vendor ID (Offset 2E)
This register uniquely identifies the add-in adapter or subsystem where the PCI device resides. It
provides a mechanism to distinguish the vendor of a adapter from the vendor of the PCI controller
used on the adapter. For the 82557 B-step this field is 0000h. For the 82557 C-Step and later
devices, this field is automatically loaded from the EEPROM at power on or upon the assertion of
PCI reset. If the EEPROM is not present or invalid, this value defaults to 0000h.
Expansion ROM Base Address Register (Offset 30)
The 8255x provides an interface to a local Flash device (or EEPROM) which may be used as an
expansion ROM. A 32-bit Expansion ROM Base Address Register at offset 30h in the PCI
Configuration Space is defined to handle the address and size information for boot-time access to
16
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
PCI Interface
the Flash. The 82557 implements this register regardless of the presence or absence of a Flash
component on the adapter. For the 82558 and later Fast Ethernet controllers, this register is only
implemented if a bit is set in the EEPROM.
The register functions exactly like a 32-bit base address register except that the encoding (and
usage) of the bottom bits is different. The upper 21 bits correspond to the upper 21 bits of the
expansion ROM base address. The 8255x only allow an expansion ROM to be mapped on a 1
Mbyte boundary. Therefore, only the most significant 12 bits are configurable to indicate the 1
Mbyte size requirement (as with the Flash Memory Mapped BAR, the 82558 and 82559 request a 1
Mbyte mapping even though the maximum Flash size allowed with those devices is 65 Kbytes).
The format of the register is shown in the figure below.
Figure 7. Expansion ROM Base Address Register
32
20
19
1
0
Read / Write
Reserved (all bits set to 0)
En
Bit 0 in the register is used to control whether or not the device accepts accesses to its expansion
ROM. When this bit is reset, the devices expansion ROM address space is disabled. This bit is
programmed at initialization time by the system BIOS. The Memory Space bit in the Command
register has precedence over the Expansion ROM Base Address Enable bit. A device responds to
accesses to its expansion ROM only if both the Memory Space bit and the Expansion ROM Base
Address Enable bit are set to 1 (it is reset to 0 upon PCI reset).
4.1.14
The Capabilities Pointer (Offset 34)
This an 8-bit field that provides an offset in the device PCI Configuration Space for the location of
the first item in the Capabilities Linked List. The Power Management Interface documentation
specifies this linked list to provide access to all appropriate device information in the
implementation of the ACPI.
For the 82257, this register is hard-wired to 0 since it does not support power management.
For the 82558 this register is set to DCh if power management is enabled in the EEPROM. If power
management is disabled, then this register is set to 0.
For the 82559 and later Intel Fast Ethernet controllers, this register is hard-wired to DCh.
4.1.15
4.1.16
Interrupt Line (Offset 3C)
The Interrupt Line register is an 8-bit register used to communicate interrupt line routing
information. This register is configurable by the system BIOS or OS. POST software writes the
routing information into this register as it initializes and configures the system. The value in this
register specifies which system interrupt controller input the device interrupt pin is connected to.
Device drivers and operating systems use this information to determine priority and vector
information.
Interrupt Pin (Offset 3D)
The Interrupt Pin register specifies which interrupt pin the device (or device function) uses. This
register is always set to a 1, indicating that INTA# is used.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
17
Download from Www.Somanuals.com. All Manuals Search And Download.
PCI Interface
4.1.17
Max_Lat / Min_Gnt (Offset 3E)
These registers specify the device settings for Latency Timer values. For both registers, the value
specifies a period of time in units of ¼ microsecond. Min_Gnt is used to specify the burst length
period the device needs assuming a clock rate of 33 MHz. Max_Lat is used to specify how often
the device needs to gain access to the PCI bus. The values of these registers are 8h (2 µS) for
Min_Gnt and 18h (6 µS) for Max_Lat.
4.1.18
Power Management PCI Configuration Registers
4.1.18.1
Capability Identifier (Offset DC)
The Capability Identifier signals this item in the capability linked list as the PCI Power
Management registers. The PCI Power Management registers have been assigned the ID of 01h.
Since power management is not implemented in the 82557, this register is hard-coded to 0 for that
device. For the 82558 and later devices, this read only register returns 01h.
4.1.18.2
4.1.18.3
Next Item Pointer (Offset DD)
The Next Item Pointer register describes the location of the next item in the capability list. Since
power management is the last item in the list, this register is set to 0.
Power Management Capabilities (Offset DE)
The Power Management Capabilities (PMC) register is a 16-bit read-only register, which provides
information on the capabilities of the device related to power management. Since power
management is not implemented in the 82557, this register is hard-coded to 0 for that device. For
the 82558 and later devices, this register returns values according to the chart below.
Table 4. Power Management Capabilities
Bit
Default Value
R / W
Description
PME_Support. This five bit field indicates the power states in
which the device may assert PME#. A value of 0b for any bit
indicates that the function is not capable of asserting the PME#
signal while in that power state.
82558A: 00011
82558B, 82559:
no auxiliary power
- 01111
31:27
RO
The 82558 A-step supports wake-up from D0 and D1. The 82558
B-step and 82559 support wake-up from D0, D1, D2 and D3hot if no
auxiliary power is present and from all power states if auxiliary
power exists.
with auxiliary
power - 11111
D2_Support. If this bit is set, this function supports the D2 Power
Management State. All devices must support the D0 and D3 states.
The 82559 and later devices support the D2 Power Management
State.
26
25
1
1
RO
RO
D1_Support. If this bit is set, this function supports the D1 Power
Management State. The 82558 and later devices supports the D1
Power Management State.
FullClk. If this bit is set, this function requires a full speed clock at
all times when it is in the D0 state in order to perform its function. If
this bit is cleared, the function only requires a full speed PCI clock
while actually transferring data so dynamic clock control may be
used. The 82558 A-step requires a full speed clock at all times
when it is in the D0 state in order to perform its function.
82558A: 1
82558B: 0
82559: 0
24
RO
18
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
PCI Interface
Table 4. Power Management Capabilities
Bit
Default Value
R / W
Description
23
0
RO
Reserved. This field is not used by the 8255x.
82558: 0
82559:
AUX_Current. If the device is connected to an auxiliary power
supply, the 82559 reports a “1” to indicate that it consume less than
250 mA from the 3.3 Vaux pin while in the D3cold state. This bit is a
reflection of bit 31.
no auxiliary power
- 0
22
21
RO
RO
with auxiliary
power - 1
DSI. The Device Specific Initialization bit indicates whether special
initialization of this function is required (beyond the standard PCI
configuration header) before the generic class device driver is able
to use it. Device specific initialization is required for the 82558 and
82559 after a D3 to D0 transition.
1
Auxiliary Power Source
82558A: 0
This bit is only meaningful if PMCSR bit 31 (D3cold supporting
PME) equals 1. When this bit also equals 1, it indicates that support
for PME# in D3cold requires an auxiliary power supply. The 82558
B-step and 82559 require auxiliary power for wake up from the
D3cold state. Therefore this bit is set to 1 if auxiliary power is
present.
82558B, 82559:
no auxiliary power
- 0
20
RO
with auxiliary
power - 1
PME Clock. When this bit is 1, it indicates that the PME#
generation logic requires its host PCI bus to maintain a free-running
PCI clock. When this bit is 0, it indicates that no host bus clock is
required for the function to generate PME#. The 82558 and later
generation devices do not require a clock to generate PME# and
return 0.
19
0
RO
RO
Version. This field specifies to software how to interpret the PMC
and PMCSR registers. A value of 001b indicates that the device
complies with the Revision 1.0 of the PCI Power Management
Interface Specification.
81:16
001
4.1.18.4
Power Management Control/Status (Offset E0)
The Power Management Control/Status Register (PMCSR) is used to determine and change the
current power state of the device. It also allows for the control of the power management interrupts
in a standard way. Since power management is not implemented in the 82557, this register is hard-
coded to 0 for that device. For the 82558 and later devices this register acts according to the chart
below.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
19
Download from Www.Somanuals.com. All Manuals Search And Download.
PCI Interface
Table 5. Power Management Control/Status Register
Bit
Value at Reset
R / W
Description
82558A: 0
PME Status. This bit is set upon a wake-up event from the link. It is
independent of the state of the PME_Enable bit. When software
writes 1 to this bit it is cleared and the device stops asserting PME#
(if enabled).
82558B, 82559:
Read
Clear
15
no auxiliary power
- 0
Sticky bit
Data Scale. The Data Scale is not supported on the 82558 and
always returns 0. For the 82559, it is a 2-bit read-only field
indicating the data register scaling factor. For the 82559, it equals
10b for registers 0 through 8 and 00b for registers 9 through 15.
82558: 00
14:13
12:9
RO
82559: 10 or 00
Data Select. This 4-bit field selects which data is reported through
the Data Register and Data Scale field. This register is only
supported on the 82559 and later generation devices.
0000
R/W
82558A = 0
82558B & 82559
= 0
Read
Clear
8
PME Enable. This bit enables the device to assert PME#.
Unknown
(0 if no auxiliary
power available)
7:5
4
000
Reserved.
Dynamic Data. The 82558 does not implement this register and
returns 0. The 82559 does not support the ability to monitor power
consumption dynamically.
0
RO
RO
3:2
00
Reserved.
Power State. This 2-bit field is used both to determine the current
power state of the 82258 or 82559 and to set the 82558 or 82559
into a new power state. The definition of the field values is given
below.
01:00
00
R/W
00b - D0
01b - D1
10b - D2
11b - D3
While wake-up events are not allowed in the D0 power state, hardware does not automatically
preclude this functionality. To ensure that wake-up events are not generated when in D0, software
must clear the PME Enable bit when putting the device into that state. To ensure that no spurious
wake-up events are generated by the function, the PME Status bit (in the PMDR register or the
PMCSR) must be specifically cleared (by writing a 1) when the PME Enable bit is set.
To support Wake on LAN mode (pre-boot wake), the PME Enable and PME Status bits are set with
known values after power-up reset. The ALTRST# pin should be connected to the device auxiliary
power good signal on the motherboard so that it will be active low on system power up. Assertion
of ALTRST# clears the PME Status bit and sets the PME Enable bit if the clock is active on the
CLK pin. Thus, if the Wake on LAN (WOL) bit in the EEPROM is set, the device will wake up the
system upon receiving of Magic Packet*.
20
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
PCI Interface
4.1.18.5
Ethernet Power Consumption Registers (Offset E2h)
The Data Register is an 8-bit read-only register providing a mechanism for the device to report
state dependent maximum power consumption and heat dissipation. The value reported in this
register depends on the value written to the Data Select field in the PMCSR register.
The power measurements defined in the register have a dynamic range of 0 to 2.55 W with 0.01 W
resolution according to the data scale.
Note: The required accuracy should be in the range of +20% and -10%. The 82557 and 82558 do not
implement this register. The 82559 and later Intel Fast Ethernet controllers do. The value reported
in this register is hard-coded in the 82559 silicon. The structure of the data register is presented
below:
Table 6. Power Consumption / Dissipation Reporting
Data Select
Data Scale
Data Reported
0
1
2
2
2
2
2
2
2
2
2
0
D0 Power Consumption = 58 (580 mW)
D1 Power Consumption = 40 (400 mW)
D2 Power Consumption = 40 (400 mW)
D3 Power Consumption = 40 (400 mW)
D0 Power Dissipated = 58 (580 mW)
D1 Power Dissipated = 40 (400 mW)
D2 Power Dissipated = 40 (400 mW)
D3 Power Dissipated = 40 (400 mW)
Common Function Power Dissipated = 00
Reserved 00 h
2
3
4
5
6
7
8
9-15
NOTE: The D1 and D2 power states are not currently supported by operating systems.
4.2
PCI Command Usage
The table below lists the PCI commands that the various Intel Fast Ethernet controllers can use.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
21
Download from Www.Somanuals.com. All Manuals Search And Download.
PCI Interface
Table 7. Generated PCI Commands
PCI Command
Name
Circumstance Used
TxCB “S” bit read.
0x6
MR
CB and RFD. Writing statistics counters or dump data
buffer to memory. Writing received packet data into
receive buffers.
0x7
MW
0xC
0xE
MRM
MRL
Reading transmit data buffers.
CB, TBD, and RFD.
MWI (82558 &
82559)
0xF
Writing received packet data into receive buffers.
The controllers do not generate I/O commands, Interrupt Acknowledge cycles, or Configuration
cycles. The controllers also do not support Dual Address Cycle (DAC). Targets (typically the
system bridge) must respond to all of the commands that the Ethernet controller generates.
4.2.1
Memory Write and Invalidate
The 82558, 82559, 82550, and 82551 have 4 internal DMA channels. Of these 4, the Receive DMA
channel is used to deposit packet data received from the link into system memory. The Receive
DMA channel uses both the Memory Write (MW) and the Memory Write and Invalidate (MWI)
commands. In order to use MWI the device must guarantee:
• A minimum transfer of one cache line.
• All byte-enable bits are active during each MWI access.
• The device may cross a cache line boundary only if it intends to transfer the entire next cache
line too.
In order to ensure the above conditions, the device may use the MWI command only if the
following conditions hold:
1. The cache line size written in the CLS register during PCI configuration is 8 or 16 Dwords.
2. The accessed address is cache line aligned.
3. The 82558 or 82559 has at least a cache line size of data byte in its Receive FIFO.
4. There is at least a cache line size of space left in the system memory buffer. In addition, the
device will use two configuration bits to enable and disable the use of MWI:
b. MWI Enable bit in the device Configure command (Section 6.4.2.3, “Configure (010b)”).
If any one of these conditions does not hold, the device uses the MW command. If a MWI cycle is
started and one of the conditions does not hold any more (for example, the data space in the
memory buffer is less than the CLS), then the device terminates the MWI cycle at the end of the
cache line. The next cycle is a MWI or MW cycle according to the conditions listed above.
If a MWI cycle is terminated by a Retry from the target, the device attempts to retry the access
using the MWI command. If a MWI cycle is terminated in the middle of a cache line by a
disconnect from the target (including Disconnect-C), the device issues a new cycle from the
disconnected point using the MW command. If the disconnect occurs on a cache line boundary, the
22
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
PCI Interface
device may start the next cycle using either MW or MWI according to the conditions listed above.
If the PCI latency timer or the 82558 (or later generation device) arbitration counter expires during
a MWI cycle, the device continues the cycle until the end of the cache line.
If the device started a MW cycle and reaches a cache line boundary, it either terminates the cycle or
continues according to the Term on CL configuration bit (Section 6.4.2.3, “Configure (010b)”). If
the Term on CL bit is set, the device terminates the MW cycle and attempts to start a new cycle.
The new cycle is a MWI cycle if all conditions are met. If the bit is not set, the device continues the
MW cycle across the cache line boundary if required.
4.2.2
Read Align
The Read Align feature is aimed to enhance performance in cache line oriented systems. Starting a
PCI transaction in these systems on a non-cache line aligned address may result in low
performance.
To solve this performance problem, the controller can be configured to terminate Transmit DMA
cycles on a cache line boundary, and start the next transaction on a cache line aligned address. This
feature is enabled when the Read Align Enable bit is set in device Configure command
(Section 6.4.2.3, “Configure (010b)”).
If this bit is set, the device operates as follows:
• When the device is close to running out of resources on the Transmit DMA (in other words,
the Transmit FIFO is almost full), it attempts to terminate the read transaction on the nearest
cache line boundary when possible.
• When the arbitration counters feature is enabled (maximum Transmit DMA byte count value is
set in configuration space), the device switches to other pending DMAs on cache line
boundary only.
4.2.3
Odd Byte Alignment Support
Various data structures have special memory alignment requirements. These alignment
requirements are detailed in Section 6.2.1, “LAN Controller Addressing Format”.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
23
Download from Www.Somanuals.com. All Manuals Search And Download.
PCI Interface
24
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
EEPROM Interface
5
The 8255x has a local memory interface that provides access to a serial EEPROM and optional
Flash device. All controllers implement these interfaces using multiplexed pins. Since the interface
uses multiplexed pins, it is not simultaneously available to software. Thus, software cannot read the
EEPROM at the same time as it is reading Flash memory. However, software can certainly read the
EEPROM and then read Flash memory or vice versa.
The Serial EEPROM stores configuration data (such as the Ethernet MAC address) for the 8255x.
The EEPROM is a serial in and serial out device. The 82557 and 82558 support a single size of
EEPROM that contains 64 registers of 16 bits per register. The 82559 and later generation devices
support either a 64 register EEPROM or a 256 register EEPROM.
Software may read or write to the EEPROM by accessing the EEPROM port in the 8255x.
All accesses, read or write, are preceded by a command instruction to the device. The command
instructions, begin with a logical 1 as a start bit, two opcode bits (indicating read, write, erase, etc.),
and n-bits of address. The address field is 6 bits for a 64-register EEPROM and 8 bits for a 256-
register EEPROM. The end of the address field is indicated by a “dummy 0” bit from the
EEPROM, which indicates the entire address field has been transferred to the device. A command
is issued by asserting the chip select signal and clocking the data into the EEPROM on its data
input pin relative to the serial clock input. The chip select signal is de-asserted after completion of
the EEPROM cycle (Command, Address and Data).
The 8255x performs an automatic read of several registers in the EEPROM following the de-
assertion of the PCI Reset signal. The controllers automatically read the EEPROM to properly set
several power-on default configurations. Since the 82559 and later devices are capable of
interfacing with different size EEPROMs (64 or 256 words), software determine the EEPROM size
first using the “dummy zero mechanism” before it accesses the EEPROM after a reset.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
25
Download from Www.Somanuals.com. All Manuals Search And Download.
EEPROM Interface
26
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
6
The 8255x LAN controllers establish a shared memory communication system with the host CPU.
Software controls the device by writing and reading data to and from this shared memory space.
All of the LAN controller functions (configuration, transmitting data, receiving data, etc.) that are
software manageable are controlled through this shared memory space.
Note: Although references are made to both simplified and flexible memory modes for transmit and
receive commands, only the simplified mode is supported. All bit settings and silicon
configurations only refer to the simplified memory mode.
6.1
The Shared Memory Architecture
The shared memory structure is divided into three parts: the Control/Status Registers (CSR), the
Command Block List (CBL), and the Receive Frame Area (RFA). The CSR physically resides on
the LAN controller and can be accessed by either I/O or memory cycles, while the rest of the
memory structures reside in system (host) memory. The first 8 bytes of the CSR is called the
System Control Block (SCB). The SCB serves as a central communication point for exchanging
control and status information between the host CPU and the 8255x. The host software controls the
state of the Command Unit (CU) and Receive Unit (RU) (for example, active, suspended or idle)
by writing commands to the SCB. The device posts the status of the CU and RU in the SCB Status
word and indicates status changes with an interrupt. The SCB also holds pointers to a linked list of
action commands called the CBL and a linked list of receive resources called the RFA. This type of
structure is shown in the figure below.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
27
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Figure 8. 8255x Memory Architecture
10/100 Mbps Device (8255x) Registers
System Control
Block (SCB)
Command Block List (System Memory)
Control Block
Control Block
Control Block
Receive Frame Area (System Memory)
Frame Descriptor
Frame Descriptor
Frame Descriptor
Buffer Descriptor
Buffer Descriptor
Buffer Descriptor
Receive Data
Buffer
Receive Data
Buffer
Receive Data
Buffer
The CBL consists of a linked list of individual action commands in structures called Command
Blocks (CBs). The CBs contain command parameters and status of the action commands. Action
commands are categorized as follows:
• Non-transmit (non-Tx) commands: This category includes commands such as no operation
(NOP), Configure, IA Setup, Multicast Setup, Dump and Diagnose.
• Transmit (Tx) command: This includes Transmit Command Blocks (TxCB).
The Receive Frame Area (RFA) consists of a list of Receive Frame Descriptors (RFDs) and a list of
user-prepared or NOS provided buffers. The receive architecture supports the simplified memory
model similar to the way it is supported by the transmit command. In the simplified memory
model, the data buffer immediately follows the RFD. The receive structures format and receive
The LAN controller also provides read and write access to an external EEPROM and the
Management Data Interface (MDI) registers. This is achieved through the EEPROM Control
Register and the MDI Control Register, respectively. These registers occupy offsets 0Ch through
14h of the CSR.
28
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
6.2
Initializing the LAN Controller
A hardware or software reset prepares the 8255x for normal operation. Since the PCI Specification
already provides automatic configuration of many critical parameters such as I/O, memory
mapping and interrupt assignment, the device is set to an operational default state after reset.
However, the device cannot transmit or receive frames until a Configure command is issued.
Different reset commands affect the controller in different ways as detailed by the table below.
Table 8. Reset Commands
Reset Operation
Effect on LAN Controller
Hardware reset. This occurs when the
Reset pin (RST#) is asserted. (This is
caused by turning the system on or by
pressing the system reset button.)
Resets all internal registers. A full initialization sequence
is needed to make the device operational.
Resets all internal registers, except the PCI configuration
registers. A full initialization sequence is needed to make
the device operational.
Software reset. (This is issued as Port
Reset command.)
Maintains PCI configuration, RU and CU base registers,
HDS size, error counters, configure, IA setup and
multicast setup command information. RU and CU are set
to the idle state. All other setup and configuration
information is lost.
Selective reset. (This is issued as Port
Selective Reset command.)
Resets all internal registers except for the PCI
configuration registers. A full initialization sequence is
needed to make the device operational. A selective reset
is issued internally before the command is executed. A
software reset is issued internally after the command is
completed.
Self test. (This is issued as a Port Self
Test command.)
The phrase “Software Reset” will be used throughout this manual to indicate a complete reset using
the Port Reset command, unless specified otherwise. Port commands are discussed in detail in
6.2.1
LAN Controller Addressing Format
The 8255x supports a 32-bit enhanced linear addressing mode and 32-bit segmented addressing
mode. The 8255x accommodates both types of addressing schemes with the enhanced linear
addressing mode. The controller always calculates a physical address by adding the appropriate 32-
bit BAR (CU base or RU base) to a 32-bit offset. This allows a linear addressing scheme to be used
by setting the base address to zero and using the full 32-bit offset registers to indicate the linear
address. A 32-bit segmented scheme can be used as well by programming the appropriate 32-bit
base address register and using the lower 16 bits of the 32-bit offset. This is illustrated in the table
below.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
29
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 9. Device Addressing Formats
Points to
Base Register
32-bit Offset Pointer
SCB General Pointer
Physical Address
Start of Command
Block List (CBL)
CU Base (32-bit)
RU Base (32-bit)
Base (32) + Offset (32)
Start of Receive Frame
Area (RFA)
SCB General Pointer
Base (32) + Offset (32)
Next Command Block
(CB)
CU Base (32-bit)
CU Base (32-bit)
RU Base (32-bit)
Link Address in CB
Base (32) + Offset (32)
Base (32) + Offset (32)
Base (32) + Offset (32)
Start of TBD Array
TBD Array Address in TxCB
Link Address in RFD
Next Receive Frame
Descriptor (RFD)
Offset (32)
Transmit Buffer #n Address in
TBD Array
TX Buffer
No Base Register
CU Base (32-bit)
No Base Register
(Physical address)
Dump Buffer (Dump
CB)
Buffer Address in CB
Port Address
Base (32) + Offset (32)
Offset (32)
Port Dump / Self-Test
Dump Counters
(Physical address)
Offset (32)
No Base Register
SCB General Pointer
(Physical address)
To support linear addressing, the device should be programmed as follows:
• Load a value of 00000000h into the CU base using the Load CU Base Address SCB command.
• Load a value of 00000000h into the RU base using the Load RU Base Address SCB command.
• Use the offset pointer values in the various data structures as absolute 32-bit linear addresses.
To support 32-bit segmented addressing, the device should be programmed as follows:
• Load the desired segment value into the CU base using the Load CU Base Address SCB
command.
• Load the desired segment value into the RU Base using the Load RU Base Address SCB
command.
• Use the offset pointer values in the various data structures as 16-bit offsets. Software must
ensure that the upper 16 bits of this offset equal 0000h as the device will add all 32 bits of the
base and offset values.
Note: The Load CU Base and the Load RU Base commands can only be executed when the CU and RU
are in the idle state. Issuing these commands when the CU or RU is not idle is prohibited.
As mentioned earlier, the 8255x data structures have special memory alignment requirements. The
table below lists these requirements. Most of the structures listed in the table will be discussed in
much greater detail in subsequent sections.
30
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 10. Alignment Requirements for 8255x Data Structures
Data Structure
Port Self-Test
Alignment Requirements
Paragraph aligned (16-byte)
Paragraph aligned (16-byte)
Port Dump
CSR and SCB
Address allocated by the BIOS. No other alignment requirements.
TxCB (buffer of TxCB in
simplified mode)
Word (even address) aligned (2-byte aligned). However, Dword (4-byte
aligned) structures are more efficient.
Word (even address) aligned (2-byte aligned). However, Dword (4-byte
aligned) structures are more efficient.
TBD
Transmit Buffer (flexible mode
only)
Byte aligned (address can be odd or even).
Word (even address) aligned (2-byte aligned). However, Dword (4-byte
aligned) structures are more efficient.
NOTE: In an MWI aware system, for best performance RFDs should be
allocated so that the RFD data area (if not zero) is cache line
aligned.
RFD (buffer of RFD in
simplified mode)
As the table above indicates, the 8255x have the same alignment restrictions with one exception:
The 82558, and 82559 have a limited capability to support odd byte aligned buffers.
6.3
Controlling the Device
Software issues control commands to the CU and RU through the SCB, which is part of the CSR.
The CPU instructs the device to activate, suspend, resume or idle the CU or RU by placing the
appropriate control command in the CU or RU control field. A CPU write access to the SCB causes
the device to read the SCB, including the Status word, Command word, CU and RU Control fields,
and the SCB General Pointer. Activating the CU causes the device to start executing the CBL.
When execution is completed the device updates the SCB with the CU status then interrupts the
CPU if it is configured. Activating the RU causes the device to access the RFA and go into the
ready state for frame reception. When a frame is received the RU updates the SCB with the RU
status and interrupts the CPU. It also automatically advances to the next free RFD in the RFA. This
interaction between the CPU and the device can continue until a software or selective reset is
issued to the device, at which point the initialization process must be executed again. The CPU can
also perform certain controller functions directly through a CPU port interface.
6.3.1
Control / Status Registers (CSR)
The Control/Status Registers make up the CSR space. The basic registers are the SCB Command
word, SCB Status word, SCB General Pointer, Port interface, EEPROM Control register, and MDI
Control register. Additionally, the 82558 and later devices also contain registers for flow control,
power management, etc. All of these registers are shown in the table below. Registers new to the
82558 are lightly shaded, and registers new to the 82559 (at offset 1Ch and beyond) are darkly
shaded. Accessing these higher offset areas in older devices has an unpredictable effect and may
cause errors.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
31
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 11. Control / Status Register
Upper Word
Lower Word
Offset
31
16 15
0
SCB Command Word
SCB Status Word
0h
SCB General Pointer
PORT
4h
8h
EEPROM Control Register
Reserved
Ch
MDI Control Register
RX DMA Byte Count
Flow Control Register
10h
14h
18h
1Ch
20h-2Ch
30h
34h
38h
3Ch
PMDR
Reserved
General Control
Reserved
General Status
Reserved
Function Event Register
Function Event Mask Register
Function Present State Register
Force Event Register
• SCB Command Word. This register is where software writes commands for the CU and RU.
• SCB Status Word. The device places the CU and RU status for the CPU to read in this word.
• SCB General Pointer. The SCB General Pointer points to various data structures in main
memory depending on the current SCB Command word.
• Port Interface. This special interface allows the CPU to reset the device and force it to dump
information to main memory or perform an internal self test.
• EEPROM Control Register. The EEPROM Control Register allows the CPU to read and write
to an external EEPROM.
• MDI Control Register. This register allows the CPU to read and write information from
Physical Layer components through the Management Data Interface.
• Early Receive Interrupt Rx Byte Count (RXBC) Register. This register allows the CPU to read
the current value in the Receive DMA byte count register. The Receive DMA byte count
register tracks the number of receive data bytes that have been placed into host memory.
• Flow Control Threshold Register. This register allows the driver to set the flow control
threshold value. (This register is not included in the 82557.)
• Flow Control Command Register. This register allows the driver to indicate flow control
commands to the 82558 and later devices.
• Power Management Driver Register (PMDR). This register indicates power management
events to the driver.
The CSR can be accessed as either an I/O mapped or memory mapped PCI slave.
Note: The PCI Configuration space Base Address Registers (BARs) automatically request memory space
for the CSR and I/O space for the CSR. Software may use either memory mapped or I/O mapped
mode or even use them interchangeably. In most environments, memory mapped mode is the
32
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
preferred method of accessing the CSR. Some bridges may not properly transfer data in memory
mapped mode and it may be necessary to have an I/O backup method if the memory method does
not work.
Note: All fields in the CSR are byte, word, or Dword addressable. Accesses to the CSR, especially the
SCB command and status word, should be limited to byte-wide operations to avoid and side
effects. For example, issuing a new command through the CU, only the lower byte of the CSR
command word should be accessed (byte 2 of the CSR). This will prevent any accidental
modification of the interrupt mask or software interrupt bits that occupy the upper byte of the
command word.
6.3.2
System Control Block (SCB)
The SCB plays a major role in communications between the CPU and the LAN controller.
Commands issued by the software and status reported by the device are placed in the SCB. The
SCB is part of the CSR and represents the first two Dwords of that structure.
Control commands are issued to the device by writing them into the SCB. This causes the device to
examine the command, clear the lower byte of the SCB command word (indicating command
acceptance), and perform the required action. Control commands perform the following types of
tasks:
• Operate the Command Unit (CU). The SCB controls the CU by specifying the address of the
Command Block List (CBL) and by starting or resuming execution of CBL commands.
• Operate the Receive Unit (RU). The SCB controls RU frame reception by specifying the
address of the Receive Frame Area (RFA) and by starting, resuming, or aborting frame
reception.
• Load the dump counters address.
• Command the device to dump or dump and reset its internal statistical counters.
In a similar manner, the CPU can send Interrupt Acknowledgments to the device by writing
them into the Interrupt Acknowledge byte (upper byte of the SCB Status word).
The device also indicates status to the CPU through bits in the SCB Status word such as CU
status and RU status.
• Indicate the cause of the current interrupt(s). Interrupts are caused by one or more of the
following events:
— The CU will interrupt the CPU when it completes an action command that has its interrupt
bit set (CX Interrupt).
— The CU will interrupt the CPU either when it leaves the active state (CNA Interrupt) or
when it enters the idle state (CI Interrupt), depending on how the device is configured.
— The CU will interrupt the CPU when it completes a transmit command with a bad status
(TNO Interrupt) if it is configured.
— The RU will interrupt the CPU when it receives either a complete frame or a predefined
part of it (FR Interrupt or ER Interrupt for the 82558 and 82559 devices).
— The RU is not ready (RNR Interrupt).
— A previously initiated read or write cycle to the MDI has been completed (MDI Interrupt).
— Software has requested an interrupt (SWI Interrupt).
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
33
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
— A flow control pause frame was received (FCP Interrupt). This does not apply to the
82557.
Note: TNO interrupts should be avoided. Protocol stacks automatically retry failed transmits. This
feature should only be enabled if software needs to know immediately about transmit failures.
Interrupt events can only be cleared by CPU acknowledgment. In other words, if the device has
asserted its interrupt pin, the only way to clear it is with a CPU Acknowledgment of that particular
interrupt bit in the SCB. Since multiple events could be active simultaneously, if some events are
not acknowledged by the ACK field, the interrupt signal will remain asserted. However, if a new
event occurs while an interrupt is set, it will not cause an additional interrupt.
The table below shows the SCB format. It is followed by a detailed description on the SCB bits and
their functions.
Table 12. System Control Block
31
16 15
0
Upper Word
Lower Word
Offset
SCB Command Word
SCB Status Word
Base + 00h
Base + 04h
SCB General Pointer
6.3.2.1
SCB Status Word
Figure 9. SCB Status Word
15
8
7
6
5
2
1
0
STAT / ACK
CUS
RUS
0
0
The SCB Status word is addressable as two bytes. The upper byte is called the STAT/ACK byte,
and the lower, the SCB Status byte. The SCB Status byte indicates the status of the CU and RU.
The STAT/ACK byte reports the device status as bits, which represent the causes of interrupts.
Writing to the STAT/ACK bits will acknowledge pending interrupts. As described below, there are
many different possible interrupt events. The LAN controller asserts the interrupt line to the CPU if
any of these interrupt events need to be serviced. More than one STAT/ACK bits may be set at the
same time. Writing 1 back to a STAT/ACK bit that was set will acknowledge that particular
interrupt bit. The device will de-assert its interrupt line only when all pending interrupt STAT bits
are acknowledged. All pending STAT bits do not need to be acknowledged in a single access, but it
is recommended if the interrupt service routine is likely to process all pending interrupts.
Note: The LAN controller latches interrupts internally. Interrupts are PCI compliant and level-triggered.
Setting a 1 in the interrupt acknowledge command for a non-pending interrupt does not cause any
34
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
malfunctions. It is simply ignored by the device. Also, any 0 bits in the interrupt acknowledge
command have no effect, whether the interrupt is pending or not.
Table 13. SCB Status Word Bits Descriptions
Bit
Symbol
Description
This bit indicates that the CU finished executing a command with its interrupt bit
set.
Bit 15
CX/TNO
The 82557 includes a TNO feature where the device can be configured to assert
this interrupt when a transmit command is completed with a status of not okay.
The TNO interrupt feature is not available in the 82558 or later devices.
This bit indicates that the RU has finished receiving a frame or the header portion
of a frame if the device is in header RFD mode.
Bit 14
Bit 13
FR
This bit indicates when the CU has left the active state or has entered the idle state.
There are 2 distinct states of the CU. When the device is configured to generate
CNA interrupt, the interrupt is activated when the CU leaves the active state and
enters either the idle or suspended state. When the device is configured to
generate CI interrupt, an interrupt will be generated only when the CU enters the
idle state.
CNA
This bit indicates when the RU leaves the ready state. The RU may leave the ready
state due to an RU Abort command or because there are no available resources or
if the RU filled an RFD with its suspend bit set.
Bit 12
Bit 11
RNR
This bit indicates when an MDI read or write cycle has completed. This interrupt
only occurs if it is enabled through the interrupt enable bit (bit 29) in the MDI
Control Register of the CSR.
MDI
SWI
This bit is used for software generated interrupts. In some cases, software may
need to generate an interrupt to re-enter the ISR.
Bit 10
Bit 9
Reserved This bit is reserved and should not be used.
This bit is used for flow control pause interrupt. It is present in the 82558 and later
devices.
Bit 8
FCP
This bit is not used on the 82557 and should be treated as a reserved bit.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
35
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 13. SCB Status Word Bits Descriptions
Bit
Symbol
Description
This field contains the CU status (2 bits). Valid values are for this field are:
00 Idle
Bits 7:6
CUS
01 Suspended
10 LPQ Active
11 HQP Active
This field contains the RU status (4 bits). Valid values are:
0000 Idle
0001 Suspended
0010 No resources
0011 Reserved
0100 Ready
0101 Reserved
0110 Reserved
Bits 5:2
RUS
0111 Reserved
1000 Reserved
1001 Reserved
1010 Reserved
1011 Reserved
1100 Reserved
1101 Reserved
1110 Reserved
1111 Reserved
Bits 1:0
Reserved These bits are reserved and should not be used.
Note: The SCB Status word is not updated immediately in response to SCB commands. For example, the
CU status will remain in the idle state for a period of time after the CU start command is issued.
Software should not rely exclusively on the state of the SCB Status word to determine when it is
appropriate to issue commands requiring the device to be in a specific state. Software may be
required to keep an internal state engine on the commands recently issued to the device to insure
that data read from the register is valid.
6.3.2.2
SCB Command Word
Figure 10. SCB Command Word
31
26
25
SI
24
23
20
19
18
16
Specific Interrupt Mask Bits
M
CU Command
0
RU Command
The SCB Command word is also addressable as two bytes. The upper byte is called the Interrupt
Control byte. The least significant byte is called the Command byte.
The Interrupt Control byte allows software to either force the generation of an interrupt or mask
device interrupts from occurring. The 82558 and later devices also allow individual interrupt
sources from within the device to be masked (this feature is not available in the 82557).
36
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
When software wants to issue an action command, it should write to the Command byte. The CUC
and RUC fields of the Command byte specify the actions to be performed by the 8255x. The
command is ready for acceptance by the device as soon as it is written into the CUC or RUC field.
The actual command execution may not start instantaneously and will depend on current receive
and transmit DMA activity. The Command byte is set by the CPU and cleared by the 8255x
indicating command acceptance.
Table 14. SCB Command Word Bits Descriptions
Bit
Symbol
Description
The mask bits range from bit 31 to 26. Writing a 1 to a mask bit disables the 8255x
(except the 82557) from generating an interrupt, or asserting the INTA# pin, due to
that corresponding event. The device may still generate interrupts due to other
interrupt events that are not masked. The corresponding bits and their masks are:
31 - CX Mask
30 - FR Mask
Specific
Bits 31:26
Interrupt
Mask Bits
29 - CNA Mask
28 - RNR Mask
27 - ER Mask
26 - FCP Mask
These bits are not present in the 82557 and should be treated as reserved.
This bit is used for the software generated interrupt. Writing a 1 to this bit causes
the device to generate an interrupt, and writing a 0 has no effect. Reads from this
bit always return a zero. The M bit (bit 24) has higher precedence than the SI bit.
Thus, if a 1 is simultaneously written to both, no interrupts occur.
Bit 25
Bit 24
SI
M
This bit is used as the interrupt mask bit. When this bit is set to 1, the device does
not assert its INTA# pin (PCI interrupt pin). The M bit has higher precedence than
bits 31 through 26 of this word and the SI bit (bit 25).
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
37
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 14. SCB Command Word Bits Descriptions
Bit
Symbol
Description
This field contains the CU Command. Valid values for this field are:
0000 NOP. The no operation command does not affect the current state of the
unit.
0001 CU Start. CU Start begins execution of the first command on the CBL. A
pointer to the first CB of the CBL should be placed in the SCB General Pointer
before issuing this command.
NOTE: The CU Start command should only be issued when the CU is in the
idle or suspended states (never when the CU is in the active state) and
all of the previously issued CBs have been processed and completed
by the CU. Sometimes, it is only possible to determine that all CBs are
completed by checking the complete bit in all previously issued
Command Blocks.
0010 CU Resume. The CU Resume command resumes CU operation by
executing the next command. If the CU is Idle, it ignores the CU Resume
command.
0100 Load Dump Counters Address. This command directs the device where to
write dump data when the Dump Statistical Counters or Dump and Reset
Statistical Counters command is used. It must be executed at least once before
the Dump Statistical Counters or Dump and Reset Statistical Counters
command is used. The address of the dump area must be placed in the general
pointer register.
Bits 23:20
CUC
0101 Dump Statistical Counters. This command directs the device to dump its
statistical counters to the area designated by the Load Dump Counters Address
command.
0110 Load CU Base. The internal CU Base Register is loaded with the value in
the SCB General Pointer.
0111 Dump and Reset Statistical Counters. This command directs the device to
first dump its statistical counters to the area designated by the Load Dump
Counters Address command and then to clear these counters.
1010 CU Static Resume. It resumes CU operation by executing the next
command. If the CU is idle, it will ignore the CU Resume command. This
command should be used only when the device CU is in the suspended state
and has no pending CU Resume commands. This command is only valid for the
82558 and later devices. It is not valid for the 82557.
Bit 19
Reserved This bit is reserved and should be set to 0.
38
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 14. SCB Command Word Bits Descriptions
Bit
Symbol
Description
This field contains the RU Command. Valid values are:
000 NOP. NOP is a no operation command and does not alter current state of
unit.
001 RU Start. RU Start enables the receive unit. The pointer to the RFA must be
placed in the SCB General Pointer before using this command. The device pre-
fetches the first RFD in preparation of receiving incoming frames that pass its
address filtering.
010 RU Resume. The RU Resume command resumes frame reception (only
when in suspended state).
011 Receive DMA Redirect. This command is only valid for the 82558 and later
devices. The buffers are indicated by an RBD chain, which is pointed to by an
offset stored in the general pointer register (in the RU base).
Bits 18:16
RUC
100 RU Abort. The RU Abort command immediately stops RU receive
operation.
101 Load Header Data Size (HDS). After a load HDS command is issued, the
device expects to only find header RFDs or to be used in Receive DMA mode
until it is reset. This value defines the size of the header portion of the RFDs or
receive buffers. The HDS value is defined by the lower 14 bits of the SCB
General Pointer; thus, bits 15 through 31 should always be set to zeros when
using this command. The value of HDS should be an even non-zero number.
110 Load RU Base. The internal RU Base Register is loaded with the value that
was placed in the SCB General Pointer just before this command was issued.
6.3.2.3
SCB General Pointer
The SCB General Pointer is a 32-bit entity, which points to various data structures depending on
the command in the CUC or RUC field. The two tables below indicate what the SCB pointer means
for the different commands.
Table 15. SCB General Pointer for the CU Command
RUC
Field
RU Command
SCB General Pointer
Added to
0
NOP
Don’t care
Pointer to first command block in the command
block list
1
2
3
CU Start
CU Base
CU Base
CU Resume
CU HPQ Start
Load Dump
Don’t care
Pointer to first command block in the HPQ
command block list
Absolute address written to by Dump Counters
Counters Address and Dump & Reset Counters commands
4
5
6
Dump Counters
Load CU Base
Don’t care
32-bit Base Register for CU data structures
Dump & Reset
Counters
7
Don’t care
10
11
CU Static Resume Don’t care
CU HPQ Resume Don’t care
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
39
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 16. SCB General Pointer for the RU Command
RUC
Field
RU Command
SCB General Pointer
Added to
0
1
2
3
4
5
6
7
NOP
Don’t care
RU Start
Pointer to first RFD in the Receive Frame Area RU Base
RU Resume
Reserved
RU Abort
Don’t care
Don’t care
Don’t care
Load HDS
Load RU Base
RBD Resume
Header Data Size (Upper 18 bits must be zero)
32-bit Base Register for RU data structures
Don’t care
6.3.2.4
Statistical Counters
The 8255x provides information for network management by providing on-chip statistical counters
that track a variety of events associated with both transmit and receive. The counters are updated
by the device when it completes the processing of a frame. For example, after the completion of
transmitting a frame on the link or when receiving a frame, the counter is updated. The Statistical
Counters are reported to the software on demand by issuing the Dump Statistical Counters
command or the Dump and Reset Statistical Counters command in the SCB CUC field. The
counters are internal to the device and are listed in the table below.
Table 17. Statistical Counters
Byte Offset
Device Statistic
Transmit good frames. This counter contains the number of frames
transmitted properly on the link. It is updated only after the actual
transmission on the link is completed and not when the frame was read from
memory as is done for the TxCB status.
0
Transmit maximum collisions (MAXCOL) errors. This counter contains
the number of frames that were not transmitted because they encountered
the configured maximum number of collisions. This counter should only
increment when the network is heavily saturated with traffic.
4
8
Transmit late collisions (LATECOL) errors. This counter contains the
number of frames that were not transmitted since they encountered a
collision outside of the normal collision window.
Transmit underrun errors. This counter contains the number of frames
that were either not transmitted or retransmitted due to a transmit DMA
underrun. If the device is configured to retransmit on underrun, this counter
may be updated multiple times for a single frame. Underruns occur due to a
lack of PCI bandwidth resulting in the internal transmit FIFO running dry
during the transmission of a frame.
12
16
Transmit lost carrier sense (CRS). This counter contains the number of
frames transmitted by the device despite the fact that it detected the de-
assertion of CRS during the transmission.
Transmit deferred. This counter contains the number of frames that were
deferred before transmission due to activity on the link.
20
24
Transmit single collision. This counter contains the number of transmitted
frames that encountered only one collision.
40
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 17. Statistical Counters
Byte Offset
Device Statistic
Transmit multiple collisions. This counter contains the number of
transmitted frames that encountered more than one collision.
28
Transmit total collisions. This counter contains the total number of
collisions that were encountered while attempting to transmit. This count
includes late collisions and collisions from frames that encountered
maximum collisions.
32
36
Receive good frames. This counter contains the number of frames that
were received properly from the link. It is updated only after the actual
reception from the link is completed and all data bytes are stored in
memory.
Receive CRC errors. This counter contains the number of aligned frames
discarded out to a CRC error. This counter is updated, if needed, regardless
of the RU state. If the RX_ER pin is asserted during a receive frame, this
counter is incremented (only once per receive frame). This counter is
counter is mutually exclusive to the alignment errors and short frames
counters.
40
44
48
Receive alignment errors. This counter contains the number of frames
that are both misaligned (in other words, CRS de-asserts on a non-octet
boundary) and contain a CRC error. The counter is updated, if needed,
regardless of the RU state. This counter is mutually exclusive to the CRC
errors and short frames counters.
Receive resource errors. This counter contains the number of good
frames discarded due to unavailable resources. Frames intended for a host
whose RU is in the no resources state fall into this category. If the device is
configured to save bad frames and the status of the received frame
indicates that it is a bad frame, this counter is not updated unless the RU is
in a no resources state.
Receive overrun errors. This counter contains the number of frames
known to be lost because the internal receive FIFO overflowed (also known
as receive overrun). This can occur if the device is unable to get the
necessary bandwidth on the PCI (system) bus. If the overflow condition
persists for more than one frame, the frames that follow the first can also be
lost. However, since a lost frame indicator does not exist, these lost frames
may not be counted. A frame that was counted as an overrun will not be
counted in other error counters (short frames, CRC errors, or alignment
errors).
52
Receive collision detect (CDT) errors. This counter contains the number
of frames that encountered collisions during frame reception. This counter is
always 0 on the 82559.
56
60
Receive short frame errors. This counter contains the number of received
frames that are shorter than the minimum frame length. It is mutually
exclusive to the CRC errors and alignment errors counters and has a higher
priority (in other words, a short frame will always increment only the short
frames counter).
Flow control transmit pause. This counter contains the number of flow
control frames transmitted by the device. The count includes both the XOFF
frames transmitted and XON frames (in other words, PAUSE(0))
transmitted.
64
68
Flow control receive pause. This counter contains the number of flow
control frames received by the device. It includes both the XOFF frames
received and XON frames (PAUSE(0)) received.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
41
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 17. Statistical Counters
Byte Offset
Device Statistic
Flow control receive unsupported. This counter contains the number of
MAC frames received by the device that are not flow control pause frames.
These frames are valid MAC frames with the predefined MAC type value
and a valid address; however, they contain an unsupported opcode. In
multimedia mode this counter tracks the pause low frames received. This
count includes both the XOFF_Low frames received and XON_Low frames
(PAUSE_Low(0)) received.
72
Transmit TCO frames. This counter is incremented when the 82559
transmits a packet initiated by the TCO controller (or ICH device). It should
be noted that any transmission of TCO packets also affects the normal
transmit counters.
76
78
Receive TCO frames. This counter is incremented when the 82559
receives a TCO packet. It should be noted that any reception of TCO
packets also affects the normal receive counters.
Applicable to all controllers.
Applicable only to 82558 and later generation controllers.
Applicable only to 82559 and later generation controllers.
As the above table indicates, the 8255x track of 16 different statistics. However, the 82558 also
maintains three additional statistics (lightly shaded in the above table) for a total of 19 counters. In
addition to the 19 statistics maintained by the 82558, the 82559 tracks two additional statistics and
six reserved statistics (indicated by darker shading in the above table).
The counters are initially set to zero by the device after reset. They cannot be preset to anything
other than zero. The device increments the counters by internally reading them, incrementing them,
and writing them back. This process is invisible to the CPU and PCI bus. In addition, the counters
adhere to the following rules:
• The counters are wrap around counters. After reaching 0FFFFFFFFh, the counters wrap
around to 0. There is no indication when the counters wrap around to 0. Software must track
this.
• The device updates the required counters for each frame. It is possible for more than one
counter to be updated as multiple errors can occur in a single frame.
• The counters are 32 bits wide and their behavior is fully compatible with the IEEE 802.1
standard. The device supports all mandatory and recommended statistics functions through the
status of the receive header and directly through these statistics counters.
Software can access the counters by issuing a Dump Statistical Counters SCB command. This
provides a snapshot, in main memory, of the internal statistical counters. For the 82557, this dump
always consists of 16 statistics. For the 82558 and 82559, this dump may contain more statistics
depending on how the device is configured. It is recommended for software to use the following
sequence for maintaining its own statistics:
1. Allocate an array in host memory large enough to hold all of the statistics dumped plus one
additional Dword for status information (for example, 68 bytes for the 82557). This memory
space must be Dword aligned.
2. Load the absolute address of this location into the device using the Load Dump Counters
Address command.
42
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
3. Write zeros to the last Dword in this area. This can be done before or after step 2.
4. Write the Dump Statistical Counters or Dump and Reset Statistical Counters command into the
CUC field in the SCB.
5. Wait for the device to dump the content of the statistical counters into the allocated memory
area. The dump is followed by the device writing a completion status into the last Dword in
this area. Software should check this Dword before processing the counters. A value of A005h
indicates the Dump Statistical Counters command has completed. A value of A007h indicates
the Dump and Reset Statistical Counters command has completed.
There should be no interrupts from the device after the completion of this operation. Also, no
changes in the CU status or RU status fields should result after operation completion.
6.3.3
PORT Interface
Table 18. Port Register Location
Bits 31:16 (Upper Word)
Bits 15:0 (Lower Word)
Offset
SCB Command Word
SCB General Pointer
PORT
SCB Status Word
Base + 0h
Base + 4h
Base + 8h
The Port interface allows software to perform certain control functions on the device. Unlike action
commands, port commands do not require access to the SCB. To initiate a port command, software
should write the appropriate Dword (described below) to the Port register (offset 08h) in the CSR.
Port commands automatically generate an internal selective or complete software reset, depending
on the command. The Dword written as part of a Port command should include:
• 16-byte aligned address value on the AD31:AD4 data bus pins.
• Port function selection code on AD3:AD0
The port Dword may be written as a 32-bit entity, two 16-bit entities, or 4 8-bit entities. In the latter
case, the device accepts only the port command after the high byte (offset Bh) is written; therefore,
the high byte should be written last. Four different port commands are supported in the 82557 and
82558 devices. The 82559 and later generation controllers support an additional command, Dump
Wake-up.
Table 19. Port Selection Function
Function
Pointer Field (Bits 31:4)
Opcode (Bits 3:0)
Software Reset
Self-test
Don't care
0000
0001
0010
0011
0111
Self-test results pointer (16 byte alignment)
Don't care
Selective Reset
Dump
Dump area pointer (16 byte alignment)
Dump area pointer (16 byte alignment)
Dump Wake-up
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
43
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
6.3.3.1
PORT Software Reset
The Port Software Reset is synonymous with the software reset and is used to issue a complete
reset to the device. Software must wait for ten system clocks and five transmit clocks before
accessing the SCB registers again. (This may be a conservative 10 µs delay loop in software.) A
software reset clears the device CSR and the PCI master block internal registers. It also requires the
device to be completely re-initialized.
6.3.3.2
PORT Self-test
The controller self-test begins by issuing an internal selective reset and running a general internal
self-test of the device. The self-test function can be used to test the device micromachine
functionality, internal registers and internal ROM. After the self-test is completed, the results are
written to memory. The device provides the results of the self-test at the address specified by the
command checks the following blocks:
• ROM. The contents of the entire ROM are sequentially read into a Linear Feedback Shift
Register (LFSR). The LFSR compresses the data and produces a signature unique to one set of
data. The results of the LFSR are then compared to a known good ROM signature. The pass or
fail result and the LFSR contents are written into the address specified by the self-test port
command.
• Parallel Registers: The micromachine performs write and read operations to all internal
parallel registers and checks the contents for proper values. The pass or fail result is then
written into the address specified by the self-test port command.
• Diagnose: The micromachine issues an internal diagnose command to the serial subsystem.
The pass or fail result of the diagnose command is written into the address specified by the
self-test port command.
Figure 11. Self-Test Results Format
Odd Word
Even Word
31
16
15
0
CROM Content Signature
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 S
0 0
0 0
0 0
D 0
M R 0 0
where
S (bit 12) General Self Test result: 0 = pass, 1 = fail
D (bit 5) Diagnose result: 0 = pass, 1 = fail
M (bit 3) Register result: 0 = pass, 1 = fail
R (bit 2) ROM Content result: 0 = pass, 1 = fail
After completing the self-test and writing the results to memory, the device executes a full internal
reset and re-initializes to the default configuration.
44
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Note: The self-test does not generate an interrupt or similar indicator to the host CPU upon completion.
6.3.3.3
Port Selective Reset
The Port Selective Reset is useful when only the device needs to be reset and all configuration
parameters need to be maintained. The selective reset puts the CU and RU to the idle state but
maintains the current configuration parameters. The selective reset maintains RU and CU base,
HDS, error counters, configuration information, and individual and multicast addresses. As in a
Port software reset, software must wait for ten system clocks and five transmit clocks before
accessing the device (approximately 10 µs in software).
6.3.3.4
6.3.3.5
Port Dump
The Dump function writes dumped data to the specified location by the Dump Area Pointer. It is
useful for troubleshooting “No Response” problems. If the device is in a no response state, the Port
Dump operation can be executed to obtain internal device information without disturbing the rest
of the system. When the Port Dump command is completed, it writes a DWORD with the value
A006h at the end of the Dump space (Dword 149). The Dump command results format is discussed
PORT Dump Wake-up
The Port Dump Wake-Up command is only available on the 82559 and later generation controllers.
It is not available on the 82558 or 82557.
After a Port Dump Wake-up command, the 82559 writes the stored data of the wake-up packet to
the host memory starting at the address specified in the pointer field. The Dump Wake-up data
structure is shown below:
Table 20. Dump Wake-up Data Structure
Dword Offset
D31
D0
0
1
Reserved
Status Word (A000h)
Reserved
Wake-up Packet
Byte Count
2:n
The 82559 executes the following sequence after it receives a Port Dump Wake-up command:
1. Writes the byte count field at Dword 1. This field contains the actual number of bytes posted in
the host memory. A value of FFh indicates that the wake-up packet length exceeded the 120
bytes. In this case, only the first 120 bytes are posted.
2. Writes the Wake-up packet data starting at Dword 2.
3. Writes a status word composed of the Complete OK bits (equals A000h at Dword 0). Prior to
the Dump Wake up packet command, the driver should initialize the status word to 0. After the
Dump Wake-up packet command, it should poll the status word for a completion status.
6.3.4
EEPROM Control Register
The EEPROM control register is a 32-bit entity at offset 0Ch of the CSR space. They are used to
read from and enable writes to an external EEPROM component.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
45
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 21. EEPROM Control Register Locations
Upper Word (D31:D16)
Lower Word (D15:D0)
Offset
SCB Command Word
SCB General Pointer
PORT
SCB Status Word
Base + 0h
Base + 4h
Base + 8h
Base + Ch
EEPROM Control Register
Reserved
The serial EEPROM or equivalent integrated circuit (IC) stores configuration data for the
controller and the adapter. The EEPROM is a serial in and serial out device. Serial EEPROMs
range in size from 16 to 256 registers of 16 bits per register. All accesses, read or write, are
preceded by a command instruction to the device. The command instructions begin with a logical 1
as the start bit, two opcode bits (indicating read, write, erase, etc.), and n-bits of address. The
address field varies with the size of the EEPROM and is 6 bits for a 64 register EEPROM and 8 bits
for a 256 register device. The end of the address field is indicated by a dummy 0 bit from the
EEPROM, which indicates the entire address field has been transferred to the device. A command
is issued by asserting the chip select signal and clocking the data into the EEPROM on its data
input pin relative to the serial clock input. The chip select signal is de-asserted after the completion
of the EEPROM cycle (command, address and data).
6.3.4.1
CPU Accesses to the EEPROM
The EEPROM access port is shown below. This register is located at offset 0Eh in the device
Control register block. The CPU directly manipulates these bits to read to or write from the
EEPROM. There should be no other local bus activity at this time.
Figure 12. EEPROM Control Register
23
22
21
20
X
19
18
17
16
X
X
X
EEDO
EEDI
EECS
EESK
Table 22. EEPROM Control Register Bits Definitions
Bit
Symbol
Description
23:20
Reserved.
Serial Data Out. This bit contains the value read from the EEPROM when
performing a read operation on the EEPROM.
19
18
EEDO
EEDI
Serial Data In. The value of this bit is written to the EEPROM when performing
write operations.
Chip Select. Setting this bit to 1 enables the EEPROM. Setting the bit to 0 disables
the EEPROM. This bit must be set to 0 for a minimum of 1 µs between consecutive
instruction cycles.
17
16
EECS
EESK
Serial Clock. Setting this bit to 1 drives the serial clock line to the EEPROM high.
Setting this bit to 0 drives the serial clock line low. Toggling this bit high and then low
clocks data in or out of the EEPROM. The serial EEPROM specifies a minimum
clock period of 4 µs. The minimum period that the clock can be high or low is 1 µs. If
the clock is driven high for only 1 µs, then it must followed by a low period of 3 µs to
meet the minimum clock frequency specification.
46
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 23. EEPROM Opcode Summary (64-register EEPROM)
Start
Bit
Instruction
Opcode Address
Data
Comments
Read register A A A A A A
Read
1
1
1
1
1
1
1
10
01
11
00
00
00
00
A A A A A A
5
4
3
2
1
0
0
0
5
4
3
2
1
0
0
Write
A A A A A A
D1:D0
Write register A A A A A A
5
4
3
2
1
5
4
3
2
1
Erase
EWEN
EWDS
ERAL
WRAL
A A A A A A
Erase register A A A A A A
5
4
3
2
1
5
4
3
2
1
0
11xxxx
00xxxx
10xxxx
01xxxx
Erase/write enable
Erase/write disable
Erase all registers
Write all registers
D15:D0
6.3.4.2
Software Determination of EEPROM Size
To determine the size of the EEPROM, software may use the following steps.
Note: This algorithm will only work if the EEPROM drives a dummy zero to EEDO after receiving the
complete address field.
1. Activate the EEPROM by writing a 1 to the EECS bit.
2. Write the read opcode, including the start bit (110b), one bit at a time starting with the most
significant bit (1):
a. Write the opcode bit to the EEDI bit.
b. Write a 1 to EESK bit and wait the minimum SK high time.
c. Write a 0 to EESK bit and wait the minimum SK low time.
d. Repeat steps 2.a through 2.c for the next two opcode bits.
3. Write the address field, one bit at a time, keeping track of the number of bits shifted in, starting
with the most significant bit.
a. Write the address bit to the EEDI bit.
b. Write a 1 to the EESK bit and wait the minimum SK high time.
c. Write a 0 to the EESK bit and wait the minimum SK low time.
d. Read the EEDO bit, looking for the dummy 0 bit.
e. Repeat steps 3.a through 3.d until the EEDO bit equals 0. The number of loop iterations
performed is the number of bits in the address field.
4. Read a 16-bit word from the EEPROM one bit at a time, starting with the most significant bit,
to complete the transaction (but discard the output).
a. Write a 1 to the EESK bit then wait the minimum SK high time.
b. Read a data bit from the EEDO bit.
c. Write a 0 to the EESK bit then wait the minimum SK low time.
d. Repeat steps 4.a through 4.c an additional 15 times.
e. De-activate the EEPROM by writing a 0 to the EECS bit.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
47
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
6.3.4.3
Software Read Access from the EEPROM
To read from the EEPROM, software is required to perform the following steps. The example is a
read from address 02h (0000 0010b).
Note: Since the address field is written most significant bit first, software must know the address field
size prior to starting a read or write access.
1. Activate the EEPROM by writing a 1 to the EECS bit.
2. Write the read opcode, including the start bit (110b), one bit at a time, starting with the most
significant bit (1):
a. Write the opcode bit to the EEDI bit.
b. Write a 1 to EESK bit then wait the minimum SK high time.
c. Write a 0 to EESK bit then wait the minimum SK low time.
d. Repeat steps 2.a through 2.c for the next two opcode bits.
3. Write the address field, one bit at a time, starting with the most significant bit.
a. Write the address bit to the EEDI bit.
b. Write a 1 to EESK bit then wait the minimum SK high time.
c. Write a 0 to EESK bit then wait the minimum SK low time.
d. Read the EEDO bit (looking for the dummy 0 bit).
e. Repeat steps 3.a through 3.d until the EEDO bit equals 0, indicating that the address field
has been completely written.
4. Read a 16-bit word from the EEPROM, one bit at a time, starting with the most significant bit.
a. Write a 1 to the EESK bit then wait the minimum SK high time.
b. Read a data bit from the EEDO bit.
c. Write a 0 to the EESK bit then wait the minimum SK low time.
d. Repeat steps 4.a through 4.d an additional 15 times.
5. De-activate the EEPROM by writing a 0 to the EECS bit.
Figure 13. EEPROM Read Timing Diagram
EESK
EECS
A 5
A 4
A 2
A1
A 0
A 3
EEDI
READ OP code
D
D
15
0
EEDO
48
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
6.3.4.4
Software Write Access to the EEPROM
Write access to the EEPROM is similar to the read access outlined above, with the differences of a
write opcode and step 4:
1. Activate the EEPROM by writing a 1 to the EECS bit.
2. Write the read opcode, including the start bit (110b), one bit at a time, starting with the most
significant bit (1):
a. Write the opcode bit to the EEDI bit.
b. Write a 1 to EESK bit then wait the minimum SK high time.
c. Write a 0 to EESK bit then wait the minimum SK low time.
d. Repeat steps 2.a through 2.c for the next two opcode bits.
3. Write the address field, one bit at a time, starting with the most significant bit.
a. Write the address bit to the EEDI bit.
b. Write a 1 to EESK bit then wait the minimum SK high time.
c. Write a 0 to EESK bit then wait the minimum SK low time.
d. Read the EEDO bit (looking for the dummy 0 bit).
e. Repeat steps 3.a through 3.d until the EEDO bit equals 0, indicating that the address field
has been completely written.
4. Write a 16-bit word to the EEPROM, one bit at a time, starting with the most significant bit
(write a data bit to the EEDI bit):
a. Write a 1 to the EESK bit then wait the minimum SK high time.
b. Write a 0 to the EESK bit then wait the minimum SK low time.
c. Repeat steps 4.a through 4.c an additional 15 times.
5. De-activate the EEPROM by writing a 0 to the EECS bit.
6.3.5
Management Data Interface Control Register
The Management Data Interface (MDI) Control register is a 32-bit entity at offset 10h of the CSR.
The MDI Control register is used to read and write bits from the management data Interface. More
details regarding the MDI can be found in Section 7.1, “Management Data Interface (MDI)” and
Section 8.1.2, “PHY Detection and Initialization”.
Table 24. MDI Control Register Location
Upper Word (D31:D16)
Lower Word (D15:D0)
Offset
SCB Command Word
SCB Status Word
Base + 0h
Base + 4h
Base + 8h
Base + Ch
Base + 10h
SCB General Pointer
PORT
EEPROM Control Register
MDI Control Register
Reserved
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
49
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
The MII Management Interface allows software to have direct control over a MII compatible PHY
through a control register in the device. This allows the driver software to place the PHY in specific
modes such as full duplex, loopback, power down, etc., without the need for specific hardware pins
to select the desired mode. This register, called the MDI Control register, resides at offset 10h in
the Control register block. The CPU writes commands to this register and the Ethernet controller
reads or writes control and status parameters to the PHY device through a serial, bi-directional data
pin called Management Data Input/Output (MDIO). These serial data transfers are clocked by the
management data clock output from the LAN controller.
Table 25. Management Data Pins
Symbol
Type
Name and Function
Management Data Input/Output. MDIO is a bi-directional signal between the device
and an MII compatible PHY. It is used to transfer control information and status
between the device and the PHY. Control information is driven by the Ethernet
controller on the MDIO pin synchronously to MDC and sampled synchronously by the
PHY. Status information is driven synchronously by the PHY and sampled
synchronously by the LAN controller.
MDIO
In/Out
Management Data Clock. MDC provides the timing reference for transfer of control
information and status on the MDIO signal. The frequency of this clock is up to
2.5 MHz.
MDC
Out
6.3.5.1
MDI Control Register
The MDI register may be written as a 32-bit entity, two 16-bit entities, or four 8-bit entities. When
writing to the MDI register using word or byte access, the data is latched only on the write to the
most significant byte of the register, which is located at offset 13h. Thus, the high byte should be
written last.
Table 26. MDI Control Register Bits
Bits
Field
Description
31:30
29
Reserved Reserved. This field is reserved and returns 0.
Interrupt Enable. When this bit is set to 1 by software, it causes the device to assert
IE
an interrupt indicating the end of an MDI cycle.
Ready. set to 1 by the device at the end of MDI transaction (i.e., indicates a Read or
28
R
Write has been completed. It should be reset to 0 by software at the same time the
command is written.
Opcode. For an MDI write, the opcode equals 01b, and for MDI read, 10b. 00b and
11b are reserved and should not be used.
27:26
25:21
20:16
Opcode
PHYAdd
RegAdd
PHY Address.
PHY Register Address.
NOTE: This value equals 1 for Intel PRO/100B TX and T4 adapters.
Data. In a write command, software places the data bits here and the device shifts
them out to the PHY. In a read command the device reads these bits serially from the
PHY and software can read them from this location.
15:0
Data
6.3.5.2
MDI Write cycle
The sequence of events for a MDI write cycle is:
1. The CPU performs a PCI write cycle to the MDI register with:
a. Ready (bit 28) = 0
50
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
b. Interrupt Enable (bit 29) = 1 or 0
c. Opcode (bits 27:26) = 01b (write)
d. PHYAdd = the PHY address from the MDI register
e. RegAdd = the register address of the specific register to be accessed (0 through 31)
f. Data = data to be written to the specified PHY register
2. The LAN controller shifts the following sequence out of the MDIO pin:
<PREAMBLE><01><01><PHYADD><REGADD><10><DATA><IDLE>
3. The LAN controller asserts an interrupt indicating MDI is finished if the Interrupt Enable bit
was set.
4. The LAN Controller sets the Ready bit in the MDI register to indicate step 2 has been
completed.
5. The CPU may issue a new MDI command.
6.3.5.3
MDI Read cycle
The sequence of events for a MDI read cycle is:
1. The CPU performs a PCI write cycle to the MDI register with:
a. Ready (bit 28) = 0
b. Interrupt Enable (bit 29) = 1 or 0
c. Opcode (bits 27:26) = 10b (read)
d. PHYAdd = the PHY address from the MDI register
e. RegAdd = the register address of the specific register to be accessed (0 through 31)
2. The LAN controller shifts the following sequence out of the MDIO pin:
<PREAMBLE><01><10><PHYADD><REGADD><Z>
where Z = the LAN controller tri-stating the MDIO pin
3. The PHY shifts the following sequence out of the MDIO pin:
<0><DATA><IDLE>
4. The LAN controller discards the leading bit and places the following 16 data bits in the MDI
register.
5. The LAN Controller asserts an interrupt indicating MDI has completed if the Interrupt Enable
bit was set.
6. The LAN controller sets the Ready bit in the MDI register indicating the read is complete.
7. The CPU may read the data from the MDI register and issue a new MDI command.
6.3.6
Receive Byte Count Register
The early receive interrupt Receive Byte Count (RXBC) register is the 32-bit entity at offset 14h of
the CSR. This read only register reflects the value of the internal receive DMA byte count register.
Note: Unless the software uses a very complicated early receive interrupt scheme, which requires the use
of header RDFs, this register is of no value to software. Such a scheme could be used by software
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
51
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
to increase performance by decreasing NOS receive latencies. However, most software early
interrupt schemes would increase CPU utilization and software complexity. Thus, use of this
register is not recommended.
Table 27. Receive Byte Count Register Location
Upper Word (D31:D16)
Lower Word (D15:D0)
Offset
SCB Command Word
SCB Status Word
Base + 0h
Base + 4h
Base + 8h
Base + Ch
Base + 10h
Base + 14h
SCB General Pointer
PORT
EEPROM Control Register
MDI Control Register
Early Receive Interrupt Receive Byte Count Register
Reserved
Bits 13:11 of this register are reserved and should equal 0. Bits 10:3 contain the receive DMA byte
count. Bits 2:0 are hard wired to 0, giving an 8-byte granularity.
The RXBC register is first initialized to the size of the next receive data buffer. This data buffer
size could equal the HDS size (if header RFDs are used) or the RFD buffer size. When a frame is
received over the wire and passed to memory by the receive DMA, the register decrements until it
reaches zero. At this point the register is set to the size of the next receive data buffer (HDS or
RFD), and the receive DMA is restarted.
6.3.7
Early Receive Interrupt
Note: For operating systems with an increased interrupt latencies, the Early Receive Interrupt feature can
be used to mask some of the latency. However, for Linux or Unix operating systems, the Early
Receive Interrupt does not provide any benefit since these operating systems have little interrupt
latencies. Thus, there is essentially no use for this feature in Linux or Unix operating systems.
The Early Receive Interrupt register is an 8-bit field at offset 18h of the CSR. This register is not
present on the 82557. It is used for configuring the device to assert an additional receive interrupt
before the entire packet has been received and deposited into host memory.
Table 28. Early Receive Interrupt Register Location
Upper Word (D31:D16)
Lower Word (D15:D0)
Offset
SCB Command Word
SCB Status Word
Base + 0h
Base + 4h
Base + 8h
Base + Ch
Base + 10h
Base + 14h
Base + 18h
SCB General Pointer
PORT
EEPROM Control Register
MDI Control Register
Early Receive Interrupt Receive Byte Count Register
FC Xon/Xoff FC Threshold Early Rx Int
Reserved
PMDR
When operating with the Early Receive Interrupt scheme, the device generates an early interrupt
depending on the length of the frame. When a frame is received, the controller looks at the Type/
Length field (byte 13 and 14) of the received frame. If the Type/Length field contains a valid length
value (0 < Type/Length ≤ 1500), the device generates an early interrupt approximately X quad-
52
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
words before the end of the frame. If the Type/Length field contains a Type value, the device does
not generate an early interrupt, except in the case where the Type value is 8137h (IPX) or 0800h
(IP) and the device is configured to generate early interrupts on IPX or IP frames. In these two
cases, it is known that the actual frame length is defined in bytes 17 and 18.
The early receive interrupt value X, in 8 bytes resolution, is programmed into the Early Rx Int
register at address 18h in the device’s CSR. When this value is all zeros no early interrupt is
generated. The Early interrupt is indicated by the ER bit in the SCB. and the assertion of INTA#. X
should be determined by the driver as a function of the interrupt latency, PCI speed, etc.
The device also generates an interrupt at the end of the frame that will assure that no frame is
missed (in case of a race condition), but is in most cases ignored by software (the interrupt is either
already asserted or masked since the driver is in the Interrupt Handler).
The following list describes special cases for early receive interrupt assertion:
• If the programmed value is larger than the frame length, the device asserts the interrupt when it
is ready to post the length field into memory.
• Short and overrun frames that contain less than the length minus the programmed value do not
generate an early interrupt.
• The device does reclaim the RFD used by a frame that caused an early interrupt if this frame is
an error frame and the device is configured to discard bad frames. This implies that the
assertion of an ER interrupt does not guarantee that this frame will also generate an FR
interrupt (in other words, the driver should not poll for the end of frame if it is not set). If the
device is configured to SBF no RFD is reclaimed and the driver may safely assume that an FR
interrupt and RFD status will follow the ER interrupt.
• The ER interrupt mechanism operates only if the device does not discard the incoming frames.
Therefore, the device does not generate ER interrupts before the RU is started. The device also
may not assert the ER interrupt for frames that exceed the allocated buffer space and are being
discarded.
• When the ER interrupt mechanism is first activated, it may not generate an ER interrupt for the
first frame. An FR interrupt is generated if the RU is ready.
6.3.8
Flow Control Register
The flow control register is a 16-bit field at offset 18h (bits 23:8) of the CSR. This register does not
exits on the 82557. It reflects flow control status information and contains some control bits that
allow software to alter the flow control configuration parameters of the device.
Table 29. Flow Control Registers Location
Upper Word (D31:D16)
Lower Word (D15:D0)
Offset
SCB Command Word
SCB Status Word
SCB General Pointer
PORT
Base + 0h
Base + 4h
Base + 8h
Base + Ch
Base + 10h
Base + 14h
Base + 18h
EEPROM Control Register
MDI Control Register
Early Receive Interrupt Receive Byte Count Register
FC Xon/Xoff FC Threshold Early Rx Int
Reserved
PMDR
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
53
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
• Bits 23:21 - Reserved. These bits are reserved.
• Bit 20 - FC Paused Low. This read only bit is an indication of the device flow control state.
It is set by the device when it receives a pause low command with a value greater than zero and
cleared when the flow control timer reaches zero or a pause frame is received.
• Bit 19 - FC Paused. This read only bit is an indication of the device flow control state. It is
set by the device when it receives a pause command with a value greater than zero and cleared
when the flow control timer reaches zero.
• Bit 18 - FC Full. This read only bit indicates device flow control state. It is set by the device
when it sends a pause command regardless of its cause (either due to the FIFO reaching the
flow control threshold or due to the device fetching an RFD with its FCP bit set or due to
writing into the Xoff bit). The bit is cleared by the device when it exits the above mentioned
state.
• Bit 17 - Xoff. Writing 1 to this bit forces the Xoff request to 1. This causes the device to
behave as if the FIFO extender is full. The Xoff request is cleared by writing 1 to the Xon bit
(bit 16). Reading this bit returns 1 after it was set and 0 after the Xon bit was set. This bit
returns 1 after an Xoff request was generated through the RFD Xoff bit until the Xon bit is set.
• Bit 16 - Xon. Writing 1 to this bit resets the Xoff request to the device. The Xoff request can
become active through the RFD Xoff bit or if the driver writes 1 to the Xoff bit (bit 17).
Reading this bit returns 0.
• Bits 15:11 - Reserved. These bits are reserved.
• Bits 10:8 - FC Threshold. The 82558 or later generation controller is capable of generating a
flow control pause frame when its receive FIFO is almost full. This three-bit field determines
the number of bytes left in the receive FIFO when the pause frame is generated. The trade-off
occurs between a higher degree of data integrity (high flow control threshold value) or high
performance (low flow control threshold value).
Table 30. Flow Control Threshold Values
FC TH (free bytes
in receive FIFO)
FC TH Value
Comment
000
0.5 Kbyte
1 Kbyte
Fast system (recommended default).
001
010
011
100
101
110
111
1.25 Kbyte
1.5 Kbyte
1.75 Kbyte
2 Kbyte
2.25 Kbyte
2.5 Kbyte
Slow system.
6.3.9
Power Management Driver Register
The Power Management Driver Register (PMDR) provides an indication of power management
events. It is an 8-bit field located at offset 18h of the CSR. This register is only present in the 82558
and later generation controllers and is not valid on the 82557.
54
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 31. Power Management Driver Register Location
Upper Word (D31:D16)
Lower Word (D15:D0)
Offset
SCB Command Word
SCB Status Word
SCB General Pointer
PORT
Base + 0h
Base + 4h
Base + 8h
Base + Ch
Base + 10h
Base + 14h
Base + 18h
EEPROM Control Register
MDI Control Register
Early Receive Interrupt Receive Byte Count Register
FC Xon/Xoff FC Threshold Early Rx Int
Reserved
PMDR
The PMDR has evolved over time in the various Intel Fast Ethernet controllers. The PMDR bits for
the 82558 and 82559 are described below.
Note: Not all bits are meaningful in the different generations of devices.
Table 32. Power Management Driver Register
Bits
Operation
Default
Description
Valid for 82559 only.
Read/
Clear
31
0
Link Status Change Indication. The link status change bit indicates
change in the link status. Writing a 1 to this bit will clear it.
Valid for 82559 (not 82559ER) only.
Read/
Clear
Magic Packet. This bit is set when a Magic Packet is received regardless
of the Magic Packet Wake-up disable bit in the configuration command
and the PME enable bit in the PMCSR. Writing a 1 to this bit will clear it.
30
0
Valid for 82559 only.
Read/
Clear
Interesting Packet. This bit is set when an interesting packet is received.
The interesting packet is defined by the 82559 packet filters. Writing a 1 to
this bit will clear it.
29
28
27
0
0
0
Read Only
Read Write
Reserved.
Valid for 82558 B-step only.
TCO Ready. When this bit is set (by the driver), the TCO ready signal on
the TCO interface is active signaling the TCO controller that the 82558 is
idle and ready for a TCO cycle.
Valid for 82558 B-step and 82559 only.
26
25
Read Only
Read Only
0
0
Force TCO Indication.
Valid for 82558 B-step and 82559 only.
TCO Request. This bit is set to 1 when the 82559 is busy receiving
packets for or transmitting packets from the TCO controller.
Valid for the 82558 and 82559.
Read/
Clear
PME Status. This bit is reflects the PME status bit in the PMCSR. It is set
upon a wake-up event, independent of the PME enable bit. Writing a 1 to
this bit clears it. It also clears the PME status bit in the PMCSR and the
PME# signal. Writing a 0 has no effect on the 82558.
24
0
(No clear
on 82559)
For the 82559, PMDR is initialized at alternate reset only and not at PCI reset (unless a PCI reset
occurs with an alternate reset).
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
55
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
6.3.10
General Control Register
The General Control register provides control over some general purpose features in the 82559. It
is an 8-bit field at offset 1Ch of the CSR. This register is only present in the 82559 and later
generation controllers and is not valid for the 82558 or 82557.
Table 33. General Control Register Location
Upper Word (D31:D16)
Lower Word (D15:D0)
Offset
SCB Command Word
SCB Status Word
Base + 0h
Base + 4h
Base + 8h
Base + Ch
Base + 10h
Base + 14h
Base + 18h
Base + 1Ch
SCB General Pointer
PORT
EEPROM Control Register
MDI Control Register
Early Receive Interrupt Receive Byte Count Register
Reserved
PMDR
FC Xon/Xoff
Reserved
FC Threshold
General Status
Early Rx Int
General Control
Table 34. General Control Register
Default
PCI Reset
Bits
Operation
Description
Reserved.
7:2
R
0
Deep Power Down on Link Down. When this bit is 1, the 82559 may
enter a deep power down state (sub 7 mA) in the D2 and D3 power
states while the link is down. At this state, the 82559 does not maintain
link integrity. It is not supported for point to point connection of two end
stations.
1
0
R/W
R/W
0
0
Clockrun Disable. When this bit is 1, the 82559 always requests the PCI
CLK. This mode can be used to overcome potential receive overruns
caused by a very long system CLKRUN latency.
6.3.11
General Status Register
The General Status register provides some basic status information in the 82559. It is an 8-bit entity
at offset 1Dh of the CSR. This register is only present in the 82559 and is not valid for the 82558 or
82557.
Table 35. General Status Register Location
Upper Word (D31:D16)
Lower Word (D15:D0)
Offset
SCB Command Word
SCB Status Word
SCB General Pointer
PORT
Base + 0h
Base + 4h
Base + 8h
Base + Ch
Base + 10h
EEPROM Control Register
MDI Control Register
Reserved
56
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 35. General Status Register Location
Upper Word (D31:D16)
Lower Word (D15:D0)
Offset
Early Receive Interrupt Receive Byte Count Register
Base + 14h
Base + 18h
PMDR
FC Xon/Xoff
Reserved
FC Threshold
General Status
Early Rx Int
General Control
Base + 1Ch
Table 36. General Status Register
Bits
Operation
Default
Description
Reserved.
7:3
R
0
HDX / FDX. This bit indicates duplex mode: 0 = half duplex (HDX) and 1
= full duplex (FDX).
2
1
0
R
R
R
10 / 100 Mbps. This bit indicates the wire speed: 0 = 10 Mbps and 1 =
100 Mbps.
Link Status Indication. This bit indicates the status of the link: 0 = link
down and 1 = link up.
6.4
Shared Memory Structures
The 8255x shared memory structures consist of the Command Block List (CBL) and the Receive
Frame Area (RFA) and are controlled by the SCB portion of the CSR. The SCB is internal to the
device while the CBL and RFA reside in main system memory.
6.4.1
Action Commands and Operating Modes
In addition to SCB control commands, the device can be controlled with action commands. This
section lists all the action commands that can be a part of the CBL. Each command contains a
command field, status and control fields, a link to the next action command, and command specific
parameters. There are three basic types of action commands: device configuration and setup,
transmission, and diagnostics. Alignment requirements are detailed in Table 10, “Alignment
Table 37. Operation Codes
Opcode Name
Description
This command results in no action by the device other than the normal
command processing such as fetching the command and decoding the
command field.
000
001
NOP
Individual
Address Setup
This command is used to load the device unique address. The unique address is
contained in the parameter field of the command.
The configure command is used to load the device with its operating
parameters. Upon reset, the device initializes to the IEEE 802.3 based
parameters, with the exception of choosing either the PHY interface mode (for
example, MII). If the user wishes to use any other values, the configure
command is used.
010
011
Configure
This command allows the programmer to setup one or more multicast or
multiple individual addresses in the device. These addresses are located in the
parameter field of the command.
Multicast
Address Setup
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
57
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 37. Operation Codes
Opcode Name
Description
One transmit command is used to send a single frame. If more than one frame
needs to be sent, the host CPU can link multiple transmit commands together.
100
101
Transmit
This command downloads microcode to the device.
Load Microcode
Note: Documentation for microcode is beyond the scope of this manual.
This command causes the device to dump its internal registers into memory.
The registers included are those loaded by the configure and address setup
commands, plus status and other internal working registers.
110
111
Dump
The diagnose command puts the device CSMA/CD subsystem through a self-
test procedure and reports the result of the internal test.
Diagnose
6.4.1.1
General Action Command Format
The format common to all action commands and the algorithms for beginning and completing the
execution (also common to all action commands) is described below.
The general format of the Command Block (CB) includes the following fields:
Figure 14. General Action Command Format
Offset
00h
Command Word Bits 31:16
Status Word Bits 15:0
EL
Link Offset
Optional Address and Data Fields
S
I
0000000000
CMD
C
X
OK XXXXXXXXXXXXX
04h
08h
6.4.1.1.1
Beginning Execution
An action command can be started by either the CU start or CU resume control command.
Otherwise, it may follow a previous action command. However, the actual command start may be
delayed by RU activity.
The following sequence is performed by the CU at the beginning of execution of each action
command:
1. The device reads 4 Dwords of the CB in one continuous PCI burst (if possible).
2. The device analyzes the contents of the command word to determine the necessary action.
3. The device reads and analyzes the link offset of the next CB and saves it.
4. The device performs specific actions according to the action command specified in the current
command field.
If the commands are chained, the CU prefetches the next command block by accessing the address
specified in the link offset field of the current CB. The device reads, analyzes, and saves the
command word of the next CB (the following fields are saved for later use: EL, S, I and CMD).
6.4.1.1.2
Completing Execution
Command completion time is asynchronous to the beginning of the command. It is determined by
the command type, RU activity, CU control commands, bus latency, etc. The CU is always in the
active state at this time. The EL, S, and I bits determine the next actions.
58
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
The following sequence is performed by the CU at the completion of execution of an action
command:
1. The devices writes command specific status to the status word of the current CB (usually the C
and OK bits are written to). If the command is a transmit command, the C and OK bits are
updated when the last transmit buffer DMA has completed, not after the actual transfer of the
frame on the serial link. This allows the transmit buffer resources to be returned as soon as the
data is copied into the device internal transmit FIFO instead of waiting for actual transmission
on the wire. Transmit status is kept in the transmit statistical counters of the device.
2. If the I bit is set, the device sets a request for the CX interrupt.
3. If the EL bit is set, after completion of the command the CU becomes idle. If the S bit is set,
the CU becomes suspended. Otherwise, the CU requests the beginning of the next action
command. A transition from an active to suspended state also generates a CNA interrupt if the
device is configured to do so.
4. The device updates the status word in the SCB. (In step 1, the transmit command status is
actually set at the end of transmit DMA, not at the completion of the actual transmit
command.)
6.4.2
Specific Action Commands
6.4.2.1
NOP (000b)
This command results in no action by the device except for those performed in the normal
command processing. It is used to manipulate the CBL. The NOP command format is shown
below.
Figure 15. NOP Command Format
Offset
00h
Command Word Bits 31:16
Status Word Bits 15:0
EL
S
I
0000000000
000
C
X
OK XXXXXXXXXXXXX
04h
Link Address (A31:A0)
This is the 32-bit address of the next command block. It is added to the CU base to
obtain the actual address.
Link Address
EL (Bit 31)
If this bit is set to one, it indicates that this command block is the last one on the CBL.
The CU will go from the active to the idle state after the execution of the CB is finished.
This transition will always cause an interrupt with the CNA/CI bit set in the SCB.
If this bit is set to one, the CU will be suspended after the completion of this CB. A CNA
interrupt will be generated if the device is configured for this. The CU transitions from the
active to the suspended state after the execution of the CB.
S (Bit 30)
If the I bit is set to one, the device generates an interrupt after the execution of the CB is
finished. If I is not set to one, the CX interrupt will not be generated.
I (Bit 29)
Bits 28:19
These bits are reserved and should all be set to 0.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
59
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
CMD (Bits 18:16) This is the NOP command, which has a value of 000b.
This bit indicates the execution status of the command. Software should reset this bit
before issuing the command to the device. Following a command completion, the device
sets it to one.
C (Bit 15)
NOTE: The difference in the definition of the C bit for the transmit command
The OK bit indicates that the command was executed without error. If it equals one, no
error occurred (command executed OK). If the OK bit is zero and the C bit is set, then an
error occurred.
OK (Bit 13)
NOTE: The difference in the definition of the C bit for the transmit command
After reading the command and determining it is a NOP, the device CU performs the following
sequence:
1. Begins execution of the NOP action command.
2. Prepares the status word with C equal to 1 and OK equal to 1.
3. Completes the NOP action command.
6.4.2.2
Individual Address Setup (001b)
This command is used to load the device with the individual address. This address is used by the
device for inserting the source address during transmission and recognizing the destination address
during reception. After a full reset and prior to individual address setup command execution, the
device assumes the broadcast address (FF FF FF FF FF FFh) is the individual address in all
respects.
This address loaded into the device is used as the individual address match reference. It will also be
used as the source address of a transmitted frame (if the no source address insertion bit equals 0).
Figure 16. Individual Address Setup Command Format
Offset
00h
Command Word Bits 31:16
0000000000 001
Status Word Bits 15:0
EL
S
I
C
X
OK XXXXXXXXXXXXX
04h
08h
0Ch
Link Address (A31:A0)
IA 4th Byte, IA 3rd Byte
IA 2nd Byte, IA 1st Byte
IA 6th Byte, IA 5th Byte
This is the 32-bit address of the next command block. It is added to the CU base to
obtain the actual address.
Link Address
EL (Bit 31)
If this bit is set to one, it indicates that this command block is the last one on the CBL.
The CU will go from the active to the idle state after the execution of the CB is finished.
This transition will always cause an interrupt with the CNA/CI bit set in the SCB.
If this bit is set to one, the CU will be suspended after the completion of this CB. A CNA
interrupt will be generated if the device is configured for this. The CU transitions from the
active to the suspended state after the execution of the CB.
S (Bit 30)
If the I bit is set to one, the device generates an interrupt after the execution of the CB is
finished. If I is not set to one, the CX interrupt will not be generated.
I (Bit 29)
Bits 28:19
These bits are reserved and should all be set to 0.
60
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
CMD (Bits 18:16) This is the Individual Address Setup command, which has a value of 001b.
This bit indicates the execution status of the command. Software should reset this bit
before issuing the command to the device. Following a command completion, the device
sets it to one.
C (Bit 15)
NOTE: The difference in the definition of the C bit for the transmit command
The OK bit indicates that the command was executed without error. If it equals one, no
error occurred (command executed OK). If the OK bit is zero and the C bit is set, then an
error occurred.
OK (Bit 13)
NOTE: The difference in the definition of the C bit for the transmit command
The individual address of the node is 6 bytes long. IA byte 1 corresponds to the first byte
of the address that is transmitted over the wire. For example, if the node address is 00
AA 00 01 02 03h, the bytes will be programmed as follows:
Individual
Address
IA Byte 1 00h
IA Byte 2 AAh
IA Byte 3 00h
IA Byte 4 01h
IA Byte 5 02h
IA Byte 6 03h
The individual address is transferred by the transmit DMA through the transmit FIFO to the
execution machine in the CSMA/CD module. Therefore, it may take some time to execute. The
execution unit maintains a 48-bit individual address register used for source address insertion
during transmission and for destination address recognition during reception. A reset causes the
individual address register to be set to FF FF FF FF FF FFh.
After reading the command and determining whether it is an IA setup command, the device CU
performs the following sequence:
1. Begins execution of the IA setup action command.
2. Initiates the transmit DMA with the address of the first byte of the individual address and a
byte count of 6.
3. Waits for the transmit byte machine to complete the internal update of the individual address
register.
4. Completes the IA setup action command.
6.4.2.3
Configure (010b)
The configure command loads the device with its operating parameters. A maximum of 22
configuration bytes are supported. The first eight bytes are used by the CU, and the remaining
bytes are passed to the CSMA/CD unit through the transmit DMA. The minimum number of
configuration bytes is 8.
Parameters not configured automatically use default values. The only exception is the PHY
interface configuration bit. For 82557 based adapters, this bit must be set to either a zero or one
before the 82557 will transmit or receive frames. For 82558 and later generation controllers, this
bit must be set to 1 before the device will send and receive.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
61
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Figure 17. Configure Command Format
Offset
00h
Command Word Bits 31:16
0000000000 010
Status Word Bits 15:0
EL
S
I
C
X
OK XXXXXXXXXXXXX
04h
08h
0Ch
10h
14h
18h
1Ch
Link Address (A31:A0)
Byte 3
Byte 2
Byte 1
Byte 5
Byte 9
Byte 13
Byte 17
Byte 21
Byte 0
Byte 4
Byte 8
Byte 12
Byte 16
Byte 20
Byte 7
Byte 6
Byte 11
Byte 10
Byte 14
Byte 18
Byte 15
Byte 19
00 00 00 00
00 00 00 00
The individual bit fields of the configure command is another area where there are numerous
differences between the controllers (82557, 82558, 82559, etc.). Therefore, a complete
configuration map for each device will be presented below. Bit descriptions for the configuration
bits follow the configuration map.
Table 38. 82557 Configuration Byte Map
Byte
D7
D6
D5
D4
D3
D2
D1
D0
0
1
2
3
4
0
0
0
Byte Count
Transmit FIFO Limit
Receive FIFO Limit
Adaptive Interframe Spacing
Reserved (must be set to 0)
0
Receive DMA Maximum Byte Count
Transmit DMA Maximum Byte Count
Discard
DMBC
Enable
5
6
Save Bad
Frames
CI
TNO
Overrun
Receive
1
1
1
Late SCB
Interrupt
Interrupt
Discard
Short
7
0
0
0
0
0
Underrun Retry
Receive
8
9
0
0
0
0
0
0
0
0
0
0
1
0
0
1
503/MII
0
0
0
0
10
11
Loopback
0
Preamble Length
NSAI
0
0
0
0
Linear Priority
L PRI
MODE
12
Interframe Spacing
0
0
0
13
14
0
1
0
1
0
1
0
1
0
0
0
0
0
1
0
0
CRS and
CDT
Broadcast Promis-
15
1
0
0
1
0
Disable
cuous
16
17
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
62
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 38. 82557 Configuration Byte Map
Byte
D7
D6
D5
D4
D3
D2
D1
D0
Receive
CRC
18
1
1
1
1
0
Padding
Stripping
Transfer
FDX Pin
Enable
Force
FDX
19
20
21
0
1
0
0
1
0
0
1
0
1
1
0
1
0
0
1
1
Multiple
IA
0
0
Multicast
All
0
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
63
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 39. 82558 Configuration Byte Map
Byte
D7
D6
D5
D4
D3
D2
D1
D0
0
1
2
0
0
0
Byte Count
Transmit FIFO Limit
Receive FIFO Limit
Adaptive Interframe Spacing
Term
Read Al
Write on
Enable
CL
Type
Enable
MWI
Enable
3
0
0
0
0
0
4
5
Receive DMA Minimum Byte Count
Transmit DMA Maximum Byte Count
Extended
DMBC
Enable
Save Bad Discard
Frames
Ext. Stat.
Count
CI
Interrupt
6
Transmit
CB
0
1
0
Overruns
Discard
Short
Receive
Dynamic
TBD
2 Frames
in FIFO
7
8
9
0
0
0
0
0
0
0
Underrun Retry
CSMA
Disable
0
0
0
0
0
1
0
MC Match ARP
Wake-up
Enable
Link
Wake-up
Enable
VLAN
ARP (0)
Wake-up
Enable
10
11
12
13
14
Loopback
0
Preamble Length
NSAI
1
0
0
1
0
0
0
0
1
0
0
0
0
0
Interframe Spacing
00000000 IP_address_Low
11110010 IP_address_High
CRS and
Ignore
U/L
Wait After Broadcast Promis-
Win
15
1
0
1
CDT
Disable
cuous
16
17
FC Delay Least Significant Byte
FC Delay Most Significant Byte
Long
Receive
OK
Receive
CRC
Transfer
18
1
Priority FC Threshold
Padding
Stripping
IA
Receive
Reject FC FC
Receive
FC
Restop
Magic
Packet
Wake-up
Address
Match
Wake-up
Enable (0)
Automatic Force
Transmit
FC
19
FDX
FDX
Restart
Priority
FC
Location
Multiple
IA
20
21
0
0
1
0
1
1
1
1
0
1
1
Multicast
All
0
0
Note: The shaded bits in the table above have different meaning for the 82558 B-step.
64
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 40. 82559 Configuration Byte Map
Byte
D7
D6
D5
D4
D3
D2
D1
D0
0
1
2
0
0
0
Byte Count
Transmit FIFO Limit
Receive FIFO Limit
Adaptive Interframe Spacing
Term
Read Al
Write on
Enable
CL
Type
Enable
MWI
Enable
3
0
0
0
0
0
4
5
Receive DMA Minimum Byte Count
Transmit DMA Maximum Byte Count
DMBC
Enable
Save Bad Discard
Ext. Stat.
Count
Extended CI
TCO
Statistics
6
7
8
9
1
0
Frames
Overruns
TxCB
Interrupt
Discard
Short
Receive
Dynamic
TBD
2 Frames
in FIFO
0
0
0
0
Underrun Retry
CSMA
Disable
0
0
0
0
0
0
0
0
0
1
Link
Wake-up
Enable
TCP/UDP
Check-
sum
VLAN
TCO
0
10
11
12
13
14
Loopback
0
Pre-amble Length
NSAI
1
0
0
1
0
0
0
0
1
0
0
0
0
0
Interframe Spacing
(00000000)
(11110010)
CRS and
1
CRC16
(0)
Ignore
U/L
Wait After Broadcast Promis-
15
1
CDT
Win
Disable
cuous
16
17
FC Delay Least Significant Byte
FC Delay Most Significant Byte
Long
Receive
OK
Receive
CRC
Transfer
18
19
1
Priority FC Threshold
Padding
Stripping
Reserved
Receive
Receive
FC
Restop
Magic
Packet
Wake-up
Automatic Force
FDX
Transmit
FC
Reject FC FC
FDX
Restart
Priority
FC
Location
Multiple
IA
20
21
0
0
1
1
1
1
1
0
1
1
Multicast
All
0
0
0
6.4.2.3.1
Configuration Parameters
The interpretation of the fields from the configuration byte maps are:
• BYTE 0.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
65
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Bits 5:0 - Byte Count. The byte count indicates the number of Command Block bytes to be
configured (and is always included in the count). It allows changing some of the parameters by
specifying a byte count less than the maximum number of configuration bytes (22 bytes). The
first eight bytes are used by the CU, and the remaining bytes are passed to the CSMA/CD unit
through the transmit DMA. The value permitted is 8 bytes.
Default - none.
Recommended -16h.
Note: If a runtime algorithm for Adaptive IFS is implemented, it is recommended that software issue an 8
byte configure command. If any of the first 8 bytes needs to be re-configured and the last 14 bytes
do not need to be changed, then it is more appropriate to use an 8 byte configure command. This is
a more efficient way of re-configuring the device.
• BYTE 1.
— Bits 6:4 - Transmit FIFO Limit. The transmit FIFO limit specifies the number of bytes
located in the 64 byte dual-ported transmit FIFO at which the device requests the bus in
order to transfer data from system memory to its internal transmit FIFO. The transmit
FIFO is organized in 32-bit wide Dwords. The FIFO limit programming is showed in the
table below.
Default - 0.
Recommended - 0.
— Bits 3:0 - Receive FIFO Limit. The receive FIFO limit specifies the number of bytes
located in the dual-ported receive FIFO at which the device requests the bus in order to
transfer data from its internal receive FIFO to system memory. The dual-ported receive
FIFO is organized into 32-bit wide Dwords. For the 82557, the FIFO size is 64 bytes. For
the 82558 and 82559 the FIFO is 128 bytes. The FIFO limit programming is showed in
the table below.
Default - 8.
Recommended - The default value is fine. However, lower values will result in better PCI
efficiency, whereas higher values will result in lower latencies.
Table 41. 82557 Dual-Port FIFO Settings - Transmit
Configuration Value
Transmit FIFO Limit
Binary (Transmit Bits 6:4)
Dwords
Bytes
0
0
0
1
1
0
0
1
1
0
0
1
2
3
4
5
6
7
0 a
4
0
0
0
1
1
1
1
1
0
1
0
1
0
1
8
12
16
20
24
28
a.
This line represents the default values.
NOTE: The configuration values are from 0 through Fh. The table shows the values in binary (4 bits wide).
66
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 42. 82557 Dual-Port FIFO Settings - Receive
Configuration Value (Nibble Wide)
Receive FIFO Limit
Binary (Receive Bits 3:0)
Dwords
Bytes
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
0
0
0
0
1
1
1
1
0
0
0
0
1
1
1
1
0
0
1
1
0
0
1
1
0
0
1
1
0
0
1
1
0
1
0
1
0
1
0
1
0
1
0
1
0
1
0
1
16
15
14
13
12
11
10
9
64
60
56
52
48
44
40
36
32a
28
24
20
16
12
8
8
7
6
5
4
3
2
1
4
a.
This line represents the default values.
Table 43. 82558 and 82559 Dual-Port FIFO Settings - Transmit
Configuration Value
Transmit FIFO Limit
Binary (Transmit Bits 6:4)
Dwords
Bytes
0
0
0
1
1
0
0
1
1
0
0
1
2
3
4
5
6
7
0 a
4
0
0
0
1
1
1
1
1
0
1
0
1
0
1
8
12
16
20
24
28
a.
This line represents the default values.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
67
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 44. 82558 and 82559 Dual-Port FIFO Settings - Receive
Configuration Value (Nibble Wide)
Binary (Receive Bits 3:0)
Receive FIFO Limit
Dwords
Bytes
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
0
0
0
0
1
1
1
1
0
0
0
0
1
1
1
1
0
0
1
1
0
0
1
1
0
0
1
1
0
0
1
1
0
1
0
1
0
1
0
1
0
1
0
1
0
1
0
1
32
30
28
26
24
22
20
18
16
14
12
10
8
128
120
112
104
96
88
80
72
64a
56
48
40
32
6
24
4
16
2
8
a.
This line represents the default values.
• BYTE 2: Adaptive IFS. This byte indicates the minimum number of PCI clocks counted
between sending two transmit frames on the wire. The resolution of this counter is 8 PCI
clocks making the range from 0 to 2040 PCI clocks.
Default - 0.
Recommended - 0.
• BYTE 3.
— Bit 3 - Terminate Write on Cache Line. This bit is reserved on the 82557 and should be set
to 0.
However, when this bit is set on the 82558 or a later generation controller, the device
attempts to terminate its write accesses on cache lines. This may yield lower PCI
throughput in systems which are not extremely cache line oriented. This bit should
therefore be set only in systems that are extremely cache line oriented.
0 = Terminate Write on Cache Line disabled.
1 = Terminate Write on Cache Line enabled.
Default - 0 (Terminate Write on Cache Line disable).
Recommended - 0.
— Bit 2 - Read Alignment Enable. This bit is reserved on the 82557 and should be set to 0.
However, when this bit is set on the 82558 and later generation controllers, the device
attempts to align its read accesses to cache lines. This may yield lower PCI throughput in
systems that are not extremely cache line oriented. Thus, this bit should be set only in
68
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
systems that are extremely cache line oriented. More information of the read alignment
capability is detailed in Section 4.2.2, “Read Align”.
0 = Read Alignment disabled.
1 = Read Alignment enabled.
Default - 0 (Read Alignment disabled).
Recommended - 0.
— Bit 0 - MWI Enable. This bit is reserved on the 82557 and should be set to 0.
However, for the 82558 and later generation controllers, it enables the device to perform
Memory Write and Invalidate (MWI) cycles on the PCI bus. If both this bit and the MWI
enable bit in the PCI command register are both set, then the device attempts to perform
MWI cycles when writing data to system memory. If either this bit or the MWI enable bit
in the PCI command register are clear, the device will not perform MWI cycles. A more
detailed description of MWI can be found in Section 4.2.1, “Memory Write and
Invalidate”.
0 = MWI disabled. The device will not perform MWI cycles even if it is permitted by the
PCI command register.
1 = MWI enabled. The device will perform MWI cycles if it is permitted by the PCI
command register.
Default - 0 (MWI disabled).
Recommended - 1.
• BYTE 4.
Bits 6:0 - Receive DMA Maximum Byte Count. This byte indicates the maximum number of
receive DMA PCI transfers that will be completed before internal arbitration. The counter has
a 4 cycle resolution. This counter is useful in throttling back the receive DMA in order to let
other device DMA channels, such as the transmit DMA, CU DMA, or RU DMA, complete
PCI cycles. For instance, if the counter is set to 4, the receive DMA will only do a 16-cycle
PCI transfer if one of the other internal DMA channels also wants to initiate a transfer. If no
other internal DMA channels are requesting a transfer, the receive DMA may run an extended
PCI burst. In order for this counter to be enabled, the DMA maximum byte count enable bit
(byte 5, bit 7) must be set. If the enable bit is not set, the receive DMA will continue until it is
finished (no other DMA unit can pre-empt it).
Note:
If this counter is enabled and set to zero, then the receive DMA may be pre-empted
almost immediately.
Default - 0.
Recommended - 0.
• BYTE 5.
— Bit 7 - DMA Maximum Byte Count Enable. Bit 7 enables the receive and transmit DMA
maximum byte count enable counters. These counters are only valid when this bit is set to
1. This bit enables both the receive and transmit DMA maximum byte counters.
Default - 0.
Recommended - 0.
— Bit 6:0 - Transmit DMA Maximum Byte Count. This byte indicates the maximum number
of transmit DMA PCI cycles that will be completed after internal arbitration. The counter
has a 4 cycle resolution. It is useful in throttling back the transmit DMA in order to let
other DMA channels, such as the receive DMA, CU DMA, or RU DMA complete PCI
cycles. For instance, if the counter is set to 4, the transmit DMA will only do a 16-cycle
PCI transfer if one of the other internal DMA channels also wants to initiate a transfer. If
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
69
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
no other internal DMA channels are requesting a transfer, the transmit DMA may perform
an extended PCI burst. In order for this counter to be enabled, the DMA maximum byte
count enable bit (byte 5, bit 7) must be set. If the enable bit is not set, the transmit DMA
will continue until it is done (no other DMA unit can pre-empt it).
Note:
If this counter is enabled and set to zero, then the transmit DMA can be pre-empted
almost immediately.
Default - 0.
Recommended - 0.
• BYTE 6.
— Bit 7 - Save Bad Frames.
This bit determines whether erroneous frames (CRC error, alignment error, etc.) are to be
discarded or saved. Erroneous frames are those where the OK bit equals 0 in the frame
descriptor status field. All frames are saved regardless of their status.
When the device is configured to save bad frames, the Receive Frame Descriptor (RFD) is
not re-used for the next frame. When bad frames are not saved, these structures are re-
used and no information is left in memory.
Note:
The statistical counters are still updated upon receiving bad frames regardless of
the state of this bit.
0 = Received bad frames are not saved in memory.
1 = Received bad frames are saved in memory.
Default - 0 (do not save bad frames).
Recommended - 0 (1 for promiscuous mode).
— Bit 6 - Discard Overrun Receive Frames. This bit determines whether Receive Overrun
frames are to be discarded or saved. When activated (set to 0) the device may internally
discard frames that were Overrun. When not activated (set to 1) the device will pass these
frames to memory and only then reclaim the memory space or not according to the SBF
configuration. If this bit is cleared (set to 0), Overrun frames will be discarded regardless
of the setting of SBF. Note that Overrun frames will not always be discarded even if this
bit is activated. If a frame has started to be transferred to memory before the overrun is
detected the frame will be passed to memory regardless of the configuration.
0 = Discard overrun frames.
1 = Pass overrun frames to memory.
Default - 0 (do not pass overrun frames to memory).
Recommended - 0.
— Bit 5 - Extended Statistical Counter.
This bit is reserved on the 82557, and should be set to 1. For the 82558 or 82559, it
determines the number of statistical counters that are dumped by the device when the
Dump Statistical Counters or Dump and Reset Statistical Counters command is issued. If
it is set to 1, the device dumps the 82557 compatible 16 counters into 68 bytes of memory.
If the bit is 0, the device dumps the full 19 counters into 80 bytes of memory.
0 = Extended Statistical Counters.
1 = Standard Statistical Counters.
Default - 1 (standard statistical counters).
Recommended - 1.
70
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
— Bit 4 - Extended Transmit CB (TxCB). This bit is reserved on the 82557 and should be set
to 1. However, for the 82558 or 82559, it determines the type of TxCB that is to be used
by the device.
If this bit is 1, the device reads the standard 4 Dword TxCB. When this bit equals 0, the
device reads 8 Dwords for all CBs and processes the TxCBs as Extended TxCBs as
0 = Extended TxCB.
1 = Standard TxCB.
Default - 1 (Standard TxCB).
Recommended - 1 for compatibility reasons. If performance is the main criteria, it is
recommended that this bit equal 0.
— Bit 3 - CI Interrupt = CU Idle Interrupt. This bit determines whether the device generates
an interrupt when the CU leaves the Active state (CNA interrupt) or when the CU enters
the Idle state (CI interrupt). If CNA interrupt is enabled, the device will generate an
interrupt when the CU goes from the Active to a non-active state (Idle or Suspended).
Interrupts are generated whenever the device sees an EL or S bit in a CB that causes it to
go into the Idle or Suspended state respectively on completion of the command. The CI
interrupt will generate interrupts only on a transition from an Active to the Idle state. If the
CI mode is enabled, interrupts can be generated in dynamic chaining (suspend/resume) by
setting the I-bit on individual CBs.
0 = CNA Interrupt. An interrupt is generated when the CU goes from active to idle or
suspended state.
1 = CI Interrupt. An interrupt is generated when the CU goes from the active to the idle
state.
Default - 0 (CNA interrupt).
Recommended - 0, depending on the implementation of the transmit code.
— Bit 2. This bit is only used on the 82557 and 82559. However, it has a completely different
meaning for both devices. For the 82557, it is the TNO Interrupt = Transmit Not OK
Interrupt (82557 only), and for the 82559, the TCO Statistical Counter.
For the 82557, this bit determines whether or not the device generates an interrupt when a
transmission ends with a bad status. If it is configured to TNO Interrupt, the device
generates an interrupt by setting the CX interrupt bit in the SCB register and asserting the
INTA# signal. This interrupt is related to the completion of actual transmission on the link
and cannot be correlated to a specific transmit CB status. The status of the bad
transmission is reflected only in a statistical manner through the statistical counters.
Note:
When it is configured to TNO Interrupt, the 82557 still generates a CX interrupt if
it encounters a transmit CB with its I bit set.
0 = CX Interrupt only.
1 = TNO Interrupt enabled.
For the 82559, setting this bit to 1 causes the device to provide TCO statistical counters.
In this case, the statistical counters are 24-Dword long structures with the last 4 Dwords.
The effect of the TCO statistics bit together with the extended statistical counters bit is
shown in the table below:
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
71
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 45. Extended Statistics Functionality
TCO Statistical
Counters
Extended Statistical
Statistics Counters Functionality
Counters
0
0
1
1
1
0
1
0
82557 compatible (16 counters / 16 Dwords)
82558 compatible (19 counters / 19 Dwords)
82559 mode (21 counters / 24 Dwords)
Reserved
Default - 0.
Recommended - 0.
— Bit 0 - Late SCB = Late SCB Update. This bit is reserved on the 82558 and 82559 and
should be set to 0 on those devices.
This bit only has meaning on the 82557 and determines when the device updates the SCB
in relation to the completion of an action command. When it equals zero, the device
updates the SCB status if there is an interrupt to report after completing the action
command and before the next action command is started. If it is set to one, the device
delays the updating of the SCB until after the next command on the CBL is started.
0 = Inactive (update SCB after command completion).
1 = Active (update SCB after next command is started).
Default - 0.
Recommended - 0.
• BYTE 7.
— Bit 6 - Two Frames in FIFO. This bit is reserved on the 82557 and should be set to 0. It is
a valid bit for the 82558 and 82559 devices.
When this bit is set on the 82558 or 82559, the device limits the number of transmit
frames in its FIFO to no more than two. This bit is expected to be used only when the
device is used in multimedia mode.
0 = Two frames in FIFO disabled.
1 = Two frames in FIFO enabled.
Default - 0 (disabled).
Recommended - 0.
Bit 7 - Dynamic TBD. This bit is reserved on the 82557 and should be set to 0. However,
when this bit is set on the 82558 or 82559, the device checks the validity of the transmit
buffer pointer in the TBD and the EL bit in the TBD. When it is clear (0), the device
assumes that the transmit buffer pointer in the TBD is always valid and the last TBD is
indicated by the TBD count field in the transmit CB. When this configuration bit is set,
the driver should set the TBD count field in the transmit CB to FFh.
0 = Dynamic TBD disabled.
1 = Dynamic TBD enabled.
Default - 0 (disabled).
Recommended - 0 (unless reducing transmit latency is large concern).
— Bits 2:1 - Underrun Retry. This field specifies the number of transmission retries after an
underrun has occurred.
72
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
0 (00) = No re-transmission. If a transmitted frame encounters an underrun it will not be
re-transmitted and the status indicating that the transmission failed will be returned and
counted in the transmit underrun counter.
1 (01) = One re-transmission. If a transmitted frame encounters an underrun, it will be re-
transmitted after the whole frame is stored in the FIFO.
2 (10) = Two re-transmissions. If a transmitted frame encounters an underrun it will be re-
transmitted when there are 512 bytes in the FIFO. If the transmission encounters another
underrun, the frame will be transmitted once again when the whole frame is stored in the
FIFO.
3 (11) = Three re-transmissions. If a transmitted frame encounters an underrun, it will be
re-transmitted when there are 512 bytes in the FIFO. If the transmission encounters
another underrun, the frame will be transmitted once again when there are 1024 bytes in
the FIFO. If the third attempt also encounters an underrun, the device will transmit it
again when the whole frame is stored in the FIFO.
Default - 0 (no retransmission).
Recommended - 1.
— Bit 0 - Discard Short Receive Frames.
This bit determines whether short frames (shorter than 64 bytes) are to be discarded or
saved. When it is set to 1, the device internally discards frames shorter than 64 bytes on
the link regardless of the SBF setting. When it equals 0, the device passes these frames to
memory and reclaims the memory space according to the SBF configuration. (Depending
on how the device is configured, it may not reclaim memory space.)
0 = Pass short frames to memory.
1 = Discard short frames.
Default - 0 (pass short frames to memory).
Recommended - 1 (0 in promiscuous mode).
The discard short frames feature should we used with caution when it is combined with
header receive interrupts. If the discard short frames feature is used, no data is passed to
memory before 64 bytes are received. Therefore, even if the HDS is set to less than 64, the
device will not pass a bad receive status if a short frame is encountered. However, a
problem may occur if the discard short frames feature is not used and HDS is set to less
than 64 bytes. In this case, the device may report a bad receive status if a short frame is
encountered since it does not reclaim an RFD that has had its HDS field filled.
• BYTE 8.
— 82557: Bit 0 - 503/MII. This bit is reserved on the 82558 and 82559 and should be set to 1
on those devices. It is valid on the 82557. It is used to select the link interface mode of the
82557. If set to 503 mode (0), the 82557 transfers data to and from the link assuming 10
Mbps operation as done when operating with the 82503 or an equivalent serial interface.
If set to MII mode (1), the 82557 transfers data to and from the link nibble-wide,
assuming MII compatible operation.
0 = 503 mode.
1 = MII mode.
Default - none.
Recommended - For the 82557, the recommended value depends on the PHY detection.
For the 82558 and 82559, the recommended value is 1.
— 82558/82559: Bit 7 - CSMA Disable. This bit is reserved on the 82557 and should be set
to 0. It is valid on the 82558 or 82559 and used to disable the link operation of the device.
If it is set to 1, the device will not receive data to or from the link. If it is set to 0, the
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
73
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
device transfers data to and from the link. Software should always set this bit to 0 when it
is issuing a configure command with more than 8 bytes.
0 = Enable.
1 = Disable.
Default - 0.
Recommended - 0.
• BYTE 9.
— Bit 7 - Multicast Match Wake Enable.
This bit is available only in the 82558 B-step. It should be set to 0 on 82557, 82558 A-
step, and 82559 devices.
This bit enables assertion of the power management event signal (PME#) upon reception
of packets that pass the multicast address filtering. The PME# signal is further gated by
the PME enable bit in the PMCSR. Although this bit is not present in the 82559, this
functionality is present through the extended wake-up packet command.
0 = Disabled.
1 = Enabled.
Default - 0 (disabled).
Recommended - 0.
— Bit 6 - ARP Wake-up Enable.
This bit is available only in the 82558 B-step. It should be set to 0 on the 82557, 82558 A-
step, and 82559 devices.
This bit enables assertion of the power management event signal (PME#) upon reception
of ARP frames (as defined above). The PME# signal is further gated by the PME enable
bit in the PMCSR. Although this bit is not present in the 82559, this functionality is
present through the extended wake-up packet command.
0 = Disabled.
1 = Enabled.
Default - 0 (disabled).
Recommended - 0.
— Bit 5 - Link Status Change Wake Enable. This bit is available only in the 82558 B-step
and the 82559. It should be set to 0 on the 82557 and 82558 A-step devices.
This bit enables assertion of PME# upon a link status change event. The PME# signal is
further gated by the PME enable bit in the PMCSR.
0 = Disabled.
1 = Enabled.
Default - 0 (disabled).
Recommended - 0.
— Bit 4 - VLAN ARP (82558 B-step) or VLAN TCO (82559). This bit is available only in
the 82558 B-step and 82559. It should be set to 0 on 82557 and 82558 A-step devices.
VLAN ARP: For the 82558 B-step, this bit enables wake-up upon reception of ARP
frames with a dynamic presence of a VLAN header. This bit takes affect only if the ARP
wake-up enable bit is also set. This same functionality has been moved on the 82559 to
the extended wake-up packet command.
0 = VLAN header not supported.
1 = Dynamic VLAN header supported.
74
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Default - 0 (off).
Recommended - 0.
VLAN TCO: On the 82559, this bit activates VLAN capability filtering of received TCO
packets at nominal D0 state. When this bit is clear, the 82559 implements receive TCO in
D0 for non-tagged TCO packets only. If this bit is set, the 82559 looks for both tagged and
non-tagged TCO packets. When the 82559 is in the power down state or the Force TCO
state, the 82559 looks for the VLAN type for recognition of tagged versus non-tagged
packets. In all other states, the 82559 does not look for the VLAN type for qualification.
0 = Only TCO packets without VLAN headers are supported.
1 = TCO packets with or without VLAN headers are supported.
Default - 0 (off).
Recommended - 0.
— Bit 0 - TCP/UDP Checksum. This bit is reserved on the 82557 and 82558, and should be
set to 0 on those devices. This bit was added for the 82559. When this bit is set to ‘1, the
82559 provides a checksum word of incoming packets, excluding MAC header and CRC.
A detailed description of the checksum calculation and memory structure can be found in
0 = Disabled.
1 = Enabled.
Default - 0.
Recommended - 0 (unless the NOS supports TCP/UDP checksum offload).
• BYTE 10.
— Bits 7:6 - Loopback. This bit defines the type of loopback.
00 = Normal operation (no loopback).
01 - Internal loopback.
10 - Reserved.
11 - External loopback (loopback pin active).
Default - 00.
Recommended - 00.
— Bits 5:4 - Pre-amble Length. This bit selects the length of the pre-amble, not including the
SFD, according to table below.
Table 46. Pre-amble Length
D5
D4
Preamble Length
1 byte
0
0
1
1
0
1
0
1
3 bytes
7 bytes
15 bytes
Default setting - 10b (7 bytes).
Recommended - 10b.
— Bit 3 - No Source Address Insertion. This bit determines the source of the source address.
0 = SA insertion (SA comes from internal device IA).
1 = No SA insertion (SA comes from memory).
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
75
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Default - 1.
Recommended - Depends on the NOS and driver environment.
• BYTE 11.
— Bits 2:0 - Linear Priority. These bits are reserved on the 82558 and 82559 and should be
set to 000b on those devices.
For the 82557, these bits correspond to he number of slot times that the device will wait
after the IFS or after backoff before enabling transmission. A higher number reduces the
priority. Stations with this value set to 0, the highest priority, conform to the IEEE 802.3
backoff algorithm.
Default - 000 (normal CSMA/CD operation).
Recommended - 000.
• BYTE 12.
— Bits 7:4 - Interframe Spacing. This field specifies the period (in multiples of 16 bit times)
that the device must defer after the later of the following two events:
• The last bit has been transmitted.
• Carrier sense becomes inactive.
Default - 96 (6 in the register).
Recommended - 6.
— Bit 0 - Linear Priority Mode. This bit is reserved on the 82558 and 82559 and should be
set to 1 for those devices. For the 82557, it determines the way the linear priority
mechanism works.
0 = Wait after transmit only. The device defers for IFS + N * slot time after the
transmission of the frame only.
1 = Wait transmit or receive. The device defers for IFS + N * slot time after the
transmission or reception of a frame.
N = Linear priority number.
Default - 0.
Recommended - 0.
• BYTE 13 and BYTE 14.
Bits 7:0 (byte 13 and byte14) - IP Address. This field is available only in the 82558 B-step.
Byte 13 should be set to 0h, and byte 14 should be set to F2h for the 82557, 82558 A-step, and
82559 devices.
For the 82558 B-step, this field holds the 16 least significant bits of the IP address used for
ARP frame filtering. For example, to configure the filter for ARP frames with an IP address of
012h 034h 056h 078h, the following values are written to the configuration block:
Table 47. 82558 B-step Configuration Block ARP Frame IP Address
Configuration Block
Offset
Configuration
Parameter Name
Example Value
13
14
IP Address Low
IP Address High
078h
056h
The ARP filter compares the value stored in offset 13 of the configuration block to the byte at
offset 41 in an ARP frame without a VLAN header and to byte 45 in ARP frames with a
VLAN header.
76
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Similarly, the value at offset 14 of the configuration block is compared to the byte at offset 40
in ARP frames without a VLAN header and to byte 44 in ARP frames with a VLAN header.
The 16-bit value of the IP address in the configuration block is in non-canonical format, while
the IP address of an ARP frame is stored in canonical format. Using the same IP address in the
example above (012h 034h 056h 078h), the ARP filter performs the following comparison:
Table 48. 82558 B-step ARP Frame IP Address Mapping
Configuration Block
Offset 13
Offset 14
Block Offset
IP Address Low
IP Address High
Example Value
Frame Offset
078h
40
056h
41
Incoming Frame
Although this field is not present in the 82559, its functionality is present in the extended
wake-up packet command.
Default - 00h, F2h (for backward compatibility).
• BYTE 15.
— Bit 7 - CRS or CDT. When this bit is set, the device will interpret an active CDT during
transmission as an active carrier.
0 = CRS only.
1 = CRS or CDT.
Default - 1.
Recommended - 1 for 82557/82503 based designs, 0 for 82557/MII based designs, 0 for
82558 or 82559 based designs.
— Bit 5 - CRC16.
This bit selects the 16-bit or 32-bit CRC engine. When it is set to 1, the 82559 operates
with a 16-bit CRC generator. Clearing this bit selects the 32-bit CRC engine. (Ethernet
operates with a 32-bit CRC.)
0 = 32-bit CRC.
1 = 16-bit CRC.
Default - 0.
Recommended - 0 (must be 0).
— Bit 4 - Ignore U/L. This bit is reserved on the 82557 and should be set to 0. When this bit
is set on the 82558 or 82559, the device ignores the U/L bit when checking for IA match
on received frames.
0 = Consider U/L bit.
1 = Ignore U/L bit.
Default - 0 (Consider U/L bit).
Recommended - 0.
— Bit 2 - Wait After Win. This bit is reserved on the 82557 and should be set to 0. For the
82558 or 82559, it activates the modified backoff algorithm, Wait After Win (Section 6.7,
0 = Wait After Win disabled.
1 = Wait After Win enabled.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
77
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Default - 0 (disabled).
Recommended - 0.
— Bit 1 - Broadcast Disable. When this bit is set, it disables the device from receiving any
frames with a broadcast address (address of all 1s). Promiscuous mode setting overrides
broadcast disable.
Default - 0 (off).
Recommended - 0.
— Bit 0 - Promiscuous Mode. When this bit is set, it causes the device to receive all frames
regardless of their destination address.
Default - 0 (off).
Recommended - 0 (1 if promiscuous mode will be enabled).
• BYTE 16.
Bits 7:0 - FC Delay Least Significant Byte. This byte is reserved on the 82557 and should be
set to 00h.
For the 82558 or 82559, this byte corresponds to the least significant byte of the flow control
delay field. This delay is used as the time parameter for the assembly of transmitted flow
control frames. The value is defined in slot time (512 bit time) resolution.
Default - 0 (82557 compatible).
Recommended - 0.
• BYTE 17.
Bit 7:0 - FC Delay Most Significant Byte. This byte is reserved on the 82557 and should be set
to 40h.
For the 82558 or 82559, this byte corresponds to the most significant byte of the flow control
delay field. This delay is used as the time parameter for the assembly of transmitted flow
control frames.
Default: 01000000 (82557 compatible).
Recommended: 0.
• BYTE 18.
— Bit 3 - Long Receive OK. This bit is reserved on the 82557 and should be set to 0.
When this bit is set on the 82558 or 82559, the device considers received frames that have
a data field longer than 1500 bytes as good frames. The frames are still flagged as long in
the RFD status word but the OK bit is set. Software can pass the frame to the NOS if long
frames are supported.
Default - 0 (disabled).
Recommended - 0 (unless the device is used in a VLAN environment).
— Bit 2 - Receive CRC Transfer. When this feature is enabled, the CSMA/CD block
transfers the CRC to host memory. If the CRC is not transferred to memory it is stripped.
The report of CRC and alignment error is reported immediately. Setting this bit disables
the stripping enable bit. Thus, if the frame is padded (the length is less than the byte
count), the frame will be transferred to memory as a whole, without stripping, even if
stripping is enabled.
Default - 0 (disabled).
Recommended - 0.
— Bit 1 - Padding Enable. If this bit is set to 1, the device enables the padding mechanism. If
the byte count of a transmitted frame is less than the minimum frame length, a padding
78
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
byte (7Eh) will be transmitted to pad (in other words, fill) the minimum frame length. The
CRC will include the padded bytes. If padding is disabled, no padding bytes will be added
even if the frame is a short frame.
Default - 1 (enabled).
Recommended - 1.
— Bit 0 - Stripping Enable. If this bit is set to 1, the device enables the stripping mechanism.
If the byte count of a received frame is lower than the actual length received, every byte
beyond the specified length will be stripped from the frame except the CRC. If it is set to
0, no stripping will be performed.
Stripping is performed only on frames that have the Length/Type field set to Length (0 <
value ≤ 1500).
Default - 0 (disabled).
Recommended - 1. However, it should be avoided if the minimum packet length cannot be
safely assumed.
• BYTE 18.
Bits 6:4 - Priority Flow Control Threshold. These bits are reserved on the 82557 and should be
set to 111.
For the 82558 or 82559, this three-bit field defines the threshold at which the device
differentiates between Pause and Pause Low FC frames (Section 6.6.3.1, “Priority Flow
Control Operation”). Every FC frame with “priority field” greater than “Priority FC
Threshold” is considered Pause_Low. Setting this configuration field to any value other than
the default 111 activates the Priority FC mode.
Default - 111 (disabled).
Recommended - 111 (unless the priority flow control threshold mechanism is implemented).
• BYTE 19.
— Bit 7 - Full Duplex Pin Enable for the 82557 and 82558 A-step devices. This bit is
reserved in the 82558 B-step, 82559, 82550, and 81551 and should be set to 1b for these
devices.
When this bit is set, the device examines the FDX# pin to determine if it should operate in
full duplex or half duplex mode. If the force full duplex bit (bit 6) is set to one, then this
bit has no meaning and the device will not examine the level of the FDX# pin. This is
described in the table below.
Default - 0 (off) for the 82557 and 82558 A -step devices.
Table 49. Full Duplex Functionality
FDX PIN ENABLE
(bit 7)
FORCE FDX
(bit 6)
Device Operating
Mode
State of FDX#
0
1
0
1
0
1
0
1
0
0
1
1
0
0
1
1
0
0
0
0
1
1
1
1
Half Duplex
Full Duplex
Full Duplex
Full Duplex
Half Duplex
Half Duplex
Full Duplex
Full Duplex
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
79
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
— Bit 6 - Force Full Duplex. This bit forces the device to operate in full duplex mode.
Transmit and receive execution can be active simultaneously. CRS is only a receive
activity indicator. Minimum reception spacing between back to back frames is two bytes.
Default - 0 - off.
Recommended - 0 (1 if the user specifies a valid override).
— Bit 5 - Reject FC (address filtering of full duplex transmit flow control frames). This bit is
reserved on the 82557, and should be set to 0.
When this bit is set on the 82558 or 82559, received flow control frames will not be
passed to memory, regardless of any address mechanism they might pass. This bit has no
effect on the action taken upon reception of such a frame.
Default - 0 (82557 compatible).
Recommended - 0.
— Bit 4 - Full Duplex Restart Flow Control. This bit is reserved on the 82557 and should be
set to 0.
When this bit is set on the 82558 or 82559, it enables transmissions of flow control frames
to the peer station in order to stop its transmissions. The sending of such a frame is
triggered by the high threshold parameter, as set in the flow control threshold register
the configured flow control delay value in the time field. When the receive FIFO is empty,
another flow control frame is sent with the value 0 in the time field.
Default - 0 (82557 compatible).
Recommended - 0.
— Bit 3 - Full Duplex Restop Flow Control. This bit is reserved on the 82557 and should be
set to 0.
When this bit is set on the 82558 or 82559, it enables transmissions of flow control frames
to the peer station in order to stop its transmissions. The sending of such a frame is
triggered by the high threshold parameter, as set in the flow control threshold register
the configured flow control delay value in the time field. When this delay expires, the
device checks the receive FIFO state. If the FIFO is not empty, another flow control frame
is sent.
Default - 0 (82557 compatible).
Recommended - 0.
— Bit 2 - Full Duplex Transmit Flow Control Disable. This bit is reserved on the 82557 and
should be set to 0.
When this bit is 0 on the 82558 or 82559, it enables the transmit flow to be paused by
incoming flow control commands. Flow control commands come from the link as special
flow control frames with a time parameter. In this mode, upon reception of such a frame,
the device pauses transmissions according to the time parameter.
Default - 0 (82557 compatible).
Recommended - 0.
— Bit 1 - Magic Packet Wake-up disable. This bit is reserved on the 82557 and 8259ER and
should be set to 0 for those devices.
When this bit is set on the 82558 or 82559, it disables the assertion of a special wake-up
signal upon reception of a Magic Packet* frame (which is a frame with certain predefined
fields). This bit takes effect only if the wake enable bit is set in the PMCSR.
Default: 0 - on (82557 compatible).
80
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Recommended - 0.
— Bit 0 - Address Wake-up (82558 A-step); IA Match Wake Enable (82558 B-step). This bit
is reserved on the 82557 and 82559 and should be set to 0 on those devices.
When this bit is set on the 82558 A-step, it enables assertion of the INTA# signal as a
special wake-up signal upon reception of a frame that passes any of device address
filtering mechanisms (according to the configuration of broadcast, promiscuous, IA,
multicast all or multiple IA). This bit takes effect only if the wake enable bit is set in the
PMCSR.
For the 82558 B-Step, this bit has a similar but slightly different function. On the 82558
B-step, it enables the assertion of the PME# signal upon reception of packets that pass the
individual address filtering. The PME# signal is further gated by the PME enable bit in the
PMCSR.
Default - 0 (off; 82557 compatible).
Recommended - 0.
• BYTE 20.
— Bit 6 - Multiple IA. When this bit is set, it enables the device to receive multiple IA
frames using the HASH mechanism. If it is disabled, HASH will only be used for
multicast frames (odd address number).
Default - 0 (disabled).
Recommended - 0.
— Bit 5: Priority FC Location. This bit is reserved on the 82557 and should be set to 01.
For the 82558 and 82559, this bit determines the location of the priority field in the flow
control frame. When it equals 0, the priority field is in byte #19 (after the time filed).
When it is 1, the priority field is in byte #31 (12 bytes later).
0 = Priority field in byte #19.
1 = Priority field in byte #31.
Default - 1.
• BYTE 21.
— Bit 3 - Multicast All. This bit enables the device to receive all frames with a multicast
address (1 in the least significant byte - odd address).
Default - 0 (disabled).
Recommended - 0.
The first 8 bytes of the configuration are kept by the CU, and the remainder are transferred by the
transmit DMA to the execution machine. When a configuration command is received, the CU
performs the following sequence:
1. Begins execution of the configuration action command.
2. Reads the first eight configure bytes and saves their content.
3. Writes the configure command to the transmit FIFO.
4. Initiates the transmit DMA to transfer the remainder of the configure bytes, up to the specified
byte count, to the execution machine.
5. Waits for the execution machine to complete its internal update of configuration registers.
6. Prepares the status word with C = 1 and OK = 1.
7. Completes the configuration action command.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
81
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
In the case of a port selective reset, the execution machine maintains configuration registers for the
device. In the case of a port software reset or a hardware reset, the device reverts to the default
values.
6.4.2.4
Multicast Setup (011b)
The multicast setup command is used for loading multicast IDs into the device for filtering
purposes. As previously noted, the filtering done on the multicast IDs is not perfect and some
unwanted frames may be accepted. This command resets the current filter and reloads it with the
specified multicast IDs. The format of the multicast addresses setup command is shown below.
Figure 18. Multicast Setup Command Format
Offset
00h
Command Word Bits 31:16
0000000000 011
Status Word Bits 15:0
EL
S
I
C
X
X
X
OK XXXXXXXXXXXXX
04h
08h
0Ch
Link Address (A31:A0)
2nd Byte
1st Byte
Multicast Count
Multicast Address List
Nth Byte
This is the 32-bit address of the next command block. It is added to the CU base to
obtain the actual address.
Link Address
EL (Bit 31)
If this bit is set to one, it indicates that this command block is the last one on the CBL.
The CU will go from the active to the idle state after the execution of the CB is finished.
This transition will always cause an interrupt with the CNA/CI bit set in the SCB.
If this bit is set to one, the CU will be suspended after the completion of this CB. A CNA
interrupt will be generated if the device is configured for this. The CU transitions from the
active to the suspended state after the execution of the CB.
S (Bit 30)
If the I bit is set to one, the device generates an interrupt after the execution of the CB is
finished. If I is not set to one, the CX interrupt will not be generated.
I (Bit 29)
Bits 28:19
These bits are reserved and should all be set to 0.
CMD (Bits 18:16) This is the multicast setup command, which has a value of 011b.
This bit indicates the execution status of the command. Software should reset this bit
before issuing the command to the device. Following a command completion, the device
sets it to one.
C (Bit 15)
NOTE: The difference in the definition of the C bit for the transmit command
The OK bit indicates that the command was executed without error. If it equals one, no
error occurred (command executed OK). If the OK bit is zero and the C bit is set, then an
error occurred.
OK (Bit 13)
NOTE: The difference in the definition of the C bit for the transmit command
This 14-bit field indicates the number of bytes in the multicast list field. The multicast
count must be a multiple of 6 bytes; otherwise, the device reduces the multicast count to
the nearest multiple of 6. If the multicast count equals 0, it resets the hash table, which is
equivalent to disabling the multicast filtering mechanism.
Multicast Count
Multicast List
This field contains a list of multicast addresses or multiple IAs to be accepted by the
device. The least significant bit of the most significant byte of each multicast address
must equal 1.
82
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
The transmit DMA transfers the list of multicast addresses from memory to the execution machine
through the transmit FIFO. The CU performs the following sequence:
1. Begins execution of the multicast setup action command.
2. Reads the multicast count field and saves it internally.
3. Initiates the transmit DMA with the multicast list address and byte count according to the
multicast count field.
4. Waits for the transmit byte machine to complete the internal hash table update.
5. Completes the multicast setup action command.
The receive byte machine maintains a 64-bit hash table used for checking multicast addresses
during reception. After the execution machine reads a multicast setup command, it clears the hash
table and reads the bytes in groups of 6. Each group is hashed using CRC logic, and the bit in the
hash table that bits 2 through 7 of the CRC register point to is set to one. A group that is not
complete has no effect on the hash table. The execution machine notifies the CU after completion.
An incoming frame is accepted if it has a destination address with the significant bit in the most
significant byte equal to 1 and after hashing points to a bit in the hash table whose value is one. The
hash function is selecting bits 2 through 7 of the transmit CRC register. A software reset causes the
hash table to become all zeros.
6.4.2.5
Transmit (100b)
Transmit commands can use either the simplified or flexible memory structure. The simplified
structure expects the transmit data to reside entirely in the memory space immediately after the
transmit command block (TCB). The flexible transmit structure allows multiple data buffers to be
accessed through a transmit buffer descriptor (TBD) array. Both models require the use of one
transmit command block per frame transmitted.
The 82558 introduced several new enhancements to the design of the software and hardware
interface for transmits. Both the 82558 and the 82559 allow software to use either the original
82557 compatible TCB format or the new extended TCB format. For the 82558 and 82559 devices,
The format of the 82557 TCB (original TCB format) is illustrated in the figure below. There were a
few additional capabilities added in the 82558 and 82559 that can be utilized through this
command block interface. These new capabilities are highlighted.
Figure 19. Transmit Command Format
Offset
00h
Command Word Bits 31:16
CID 000 NC
Status Word Bits 15:0
EL
S
I
SF
100
C
X
0
OK
U
XXXXXXXXXXXX
04h
08h
Link Address (A31:A0)
Transmit Buffer Descriptor Array Address
TBD Number
Transmit Threshold
EOF
Transmit Command Block Byte Count
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
83
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
This is the 32-bit address of the next command block. It is added to the CU base to
obtain the actual address.
Link Address
EL (Bit 31)
If this bit is set to one, it indicates that this command block is the last one on the CBL.
The CU will go from the active to the idle state after the execution of the CB is finished.
This transition will always cause an interrupt with the CNA/CI bit set in the SCB.
If this bit is set to one, the CU will be suspended after the completion of this CB. A CNA
interrupt will be generated if the device is configured for this. The CU transitions from the
active to the suspended state after the execution of the CB.
S (Bit 30)
If the I bit is set to one, the device generates an interrupt after the execution of the CB is
finished. If I is not set to one, the CX interrupt will not be generated.
I (Bit 29)
The CNA Interrupt Delay field is only present on 82558 and later generation controllers.
(It is not a valid field for the 82557, unless special microcode is downloaded to this
device.) The CID indicates the length of time CNA interrupts are delayed by the device.
CID (Bits 28:24)
Bits 23:21
NC
These bits are reserved and should all be set to 0.
0: CRC and Source Address are inserted by the controller. If the “No Source Address
Insertion” (NSAI) bit is set by the configure command, then only the CRC is inserted by
the controller. Normally, this bit should be set because it is desirable to have the device
compute and insert the CRC automatically.
1: CRC and Source Address are not inserted by the controller and are assumed to come
from memory.
This bit indicates whether the device is operating in simplified or flexible mode.
SF
0 = Simplified Mode. All transmit data is in the TCB, and the TBD array address field
must equal all 1s.
1 = Flexible Mode. Data is in the TCB (optional) and in a linked list of the TBDs.
CMD (Bits 18:16) This is the transmit command, which has a value of 100b.
The C bit indicates that the transmit DMA has completed processing the last byte of data
associated with the TCB. This is not the actual completion of the transmit command as
the C bit indicates in other action commands. The actual completion of a transmit
command occurs when the frame is actually sent out on the wire. At the end of actual
transmission, no further status is posted in the TCB, but the transmit statistical counters
are updated.
C (Bit 15)
The OK bit indicates that the command was executed without error. If it equals 1, no error
occurred (command executed OK). If the OK bit is zero and the C bit is set, then an error
occurred.
OK (Bit 13)
U (Bit 12)
Bits 11:0
NOTE: For the transmit command, the OK bit is always set when the C bit is set.
The U bit indicates that one or more underruns were encountered by this or previously
transmitted frames since the last TCB status update. Since there is no mechanism for
indicating underruns during or at the end of frame transmission, this bit is set in addition
to the transmit underruns statistical counter for software management purposes.
These bits must be set to all zeros.
In flexible mode, this is a 32-bit address pointing to the first TBD in a contiguous list of
TBDs called the TBD array. A TBD is two Dwords, a transmit buffer pointer and buffer
size data. In simplified mode this field should be set by software to a null pointer
(0FFFFFFFFh).
TBD Array
Address
In flexible mode, this represents the number of transmit buffers in the contiguous TBD
array. It should have a one to one correspondence of TBDs and buffers in the array. If the
device finds the TBD number equal to 0, it assumes the TBD array address is a null
pointer and the EOF bit is set. The 82558 and 82559 have a special dynamic TBD mode
that the 82557 does not have. If the dynamic TBD mode is enabled (in the configure
command), software should write a value of FFh into this field. Software should also
mark each TBD as valid or invalid. In the 82557, the TBD number is the only indication
that the TBD is the last associated with a particular transmit frame.
TBD Number
84
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
The transmit threshold defines the number of bytes that should be present in the
Transmit
controller's transmit FIFO before it starts transmitting the frame. The value is internally
multiplied by 8 to give a granularity of 8 bytes. For example, a value of 1 means the
82557 will start transmitting only when it has 8 bytes in its transmit FIFO. The transmit
threshold should be within a range of 1 to 0E0h. (The value 0FFh should not be used.)
Threshold
The EOF bit indicates if the whole frame is in the transmit command block. For
consistency, it should be set by software, although it is not checked in simplified or
flexible mode.
EOF
For either simplified or flexible mode, the controller is able to transmit data from memory
immediately contiguous to the TCB itself. The amount of data to be read from this space
is determined by the 14-bit TCB byte count. This counter indicates the number of bytes
that will be transmitted from the transmit command block, starting with the third byte after
the TCB count field (address N + 10h). The TCB count field can be any number of bytes
up to a maximum of 2600, which allows the user to transmit a frame with a header having
an odd number of bytes. In simplified mode, the TCB byte count indicates the total
number of bytes to be transmitted and should not equal zero. In flexible mode, if the TCB
byte count equals 0, then all data is taken from the buffers pointed to by the TBD array.
TCB Byte Count
The 82558 and 82559 also offer a more advance transmit command block. When they are
configured to use extended TCBs, the device reads an 8-Dword TCB from host memory into its
internal registers instead of the standard 4-Dword TCB. The new TCB structure is composed of the
4 standard TCB Dwords followed by 2 TBDs or 2 Dwords each.
The fields in the first 4 Dwords are identical to the first 4 Dwords of the standard TCB except for
the TBD array address, which points to the third TBD rather than the first one. In other words, if
the frame consists of more than two transmit buffers, the rest of the TBDs, from the third one
onwards, are placed in a standard TBD array, which is pointed to by the TBD array address field.
The TBD number field indicates the total number of TBDs including the two TBDs located in the
latter 4 Dwords of the extended TCB. If a transmitted frame consists of less than two TBDs, the
driver can set the size field of the second (or both) TBD to zero or set the EL bit on the first TBD.
The advantage of the extended TCB is that it enables the device to read the TCB and the first two
TBDs in one 8-Dword PCI burst. This eliminates one PCI read and its associated latency and
enables both the TCB and its immediate data field to be cache line aligned.
An extended TCB is assumed to be flexible. The two TBDs that are part of the extended TCB may
use the EL bit, but it is required that the transmit buffer pointers in the two TBDs are always valid
(in other words, not equal to 0).
The transmit buffer descriptor (TBD) array is a contiguous structure of TBDs. A TBD is defined as
a transmit buffer address and a transmit buffer size. The format of the TBD array is shown below.
Figure 20. Transmit Buffer Descriptor
Odd Word (Bits 31:16)
Even Word (Bits 15:0)
Transmit Buffer #0 Address
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
Transmit Buffer #1 Address
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
Transmit Buffer #N Address
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0
EL
EL
EL
0
0
0
Size (Actual Count)
4
8
Size (Actual Count)
Size (Actual Count)
C
N*8
N*8+4
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
85
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
This is the starting address of the memory area that contains the data to be sent. It is an
absolute 32-bit address. It does not add the CU base value to determine the physical
address.
Transmit Buffer
#N
The EL bit is not used by the 82557 and is only valid for 82558 and later generation
devices. When it is set, the TBD is the last TBD associated with this transmit frame.
EL (End of List)
Size
(Actual Count)
This 14-bit quantity specifies the number of bytes that hold information for the current
buffer. It is set by the CPU before transmission.
6.4.2.5.1
Dynamic TBD Mode
Note: Dynamic TBD mode only exists in the 82558 and 82559 devices. It is not a valid mode for the
82557.
The 82557 requires all TBDs to be setup by the driver before the device is issued the CU start or
CU resume command. However, in environments where virtual addresses must be translated to
physical addresses, TBD setup is a very time consuming process. The 82558 and 82559 support a
new configuration mode called “dynamic TBD” mode, which activates two new features in the
TBD structure. (Details regarding configuration of this mode are in Section 6.4.2.3, “Configure
(010b)”.) Each TBD, which still has two 32-bit Dwords as defined in the 82557, has the following
two features defined.
• NV - Not valid pointer. When the device is configured to dynamic TBD, it checks the
transmit buffer pointer in the TBD. If it equals all zeros, it is considered to be an invalid
pointer. The device discards the TBD and attempts to read it again as soon as possible. When
this pointer is valid, the TBD is valid and the device can use the transmit buffer.
• EL - End of list bit. When this bit is set, the current TBD is the last TBD associated with this
transmit frame. The EL bit does not have to be set in the last TBD as indicated by the TBD
number field in the TCB. If the device reaches the last TBD in the array as indicated in the
TBD number field, it terminates the transmission regardless of the EL bit status. However, if
the device detects a TBD with a valid pointer and its EL bit set, it terminates the frame even if
it did not reach the number of TBDs indicated in the TBD number field. If dynamic TBD
configuration is currently in use, the driver should set the value of the TBD number field in the
TCB to FFh.
These two features enable the driver to spontaneously add TBDs after issuing the CU resume
command. The goal of automatically adding transmit buffer descriptors is to reduce overall latency
by achieving more parallelism between the driver and the device. This scheme allows the driver to
issue the CU resume command after filling in the first TBD (or even before that) and, while the
device is processing the transmit command block, the first TBD, and first transmit buffer, to
continue setting up the TBD array.
The driver programs FFh in the TBD number field of the transmit command block. The driver
prepares the first TBD in the TBD array with a valid pointer, and it is considered valid. Afterwards,
the driver issues the command to hardware.
The device flow is:
1. Fetch any immediate data from TCB.
2. Fetch the first and second TBDs.
3. Fetch the first transmit buffer since the pointer is valid in the first TBD.
4. Begin transmission (depending on the transmit threshold value).
86
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
5. Fetch data if the transmit buffer pointer is zero (invalid) in the second TBD or poll the TBD.
6. Finish the transmission if the EL bit is set.
6.4.2.5.2
Transmit Command Operation
The execution of a transmit command causes frame transmission. If the frame experiences
collisions, the device automatically attempts to re-transmit the frame up to 15 times. If it still
experiences collisions after 16 tries, the device increments the maximum collisions counter. The
following sequence outlines a general transmit command operation for the flexible memory
structure (TCBs and TBDs).
1. Place the transmit command opcode (100b) in the command word.
2. Place the destination address and length field in the appropriate transmit structure.
The TBD array address should point to the first TBD in the array. When the simplified
memory structure is used, the TBD array address is not used.
3. Configure the transmit buffer address and size (actual count) for each buffer. The last buffer in
the TBD array is determined by the TBD number field in the TCB.
The flow of events for transmitting a single frame using a flexible TCB is:
1. The CPU creates a TCB and TBD array in system memory. The transmit buffer address
pointers in the TBDs point to valid data buffers in host memory.
2. The CPU writes a CU start command (or CU resume if the CU is suspended) into the SCB.
The write event causes the device to read the CUC field, and the device notices that it should
start the CU.
3. The device processes the SCB, reads the SCB general pointer, and clears the SCB command
word.
4. The device reads the first TCB in the CBL and the first TBD from the TBD array.
5. If the TCB size field does not equal zero, the TCB holds data to be transmitted and the device
reads this data first.
6. The controller reads the first transmit data buffer from host memory at the address provided in
the transmit buffer #0 address field of the transmit buffer array.
7. After the transmit threshold bytes are read (either from one or multiple transmit buffers), the
controller begins frame transmission to the PHY interface.
8. If there are multiple TBDs, the controller reads the next TBD from the TBD array.
9. After the first buffer has been completely read, the device starts reading the transmit data from
the next buffer.
10. After the last buffer is completely read, the device sets the C bit in the TCB, enabling the
driver to re-use reuse the TCB, TBDs, and transmit buffers. The controller posts the underrun
bit in the TCB if an underrun occurred since the last TCB status was reported.
11. The device completes the frame transmission to the serial interface (for the 82557, either MII
or 82503).
12. The controller updates its internal transmit status counters.
The transmit command differs from other action commands. Generally, the action commands have
parameters in one memory block. However, the transmit command may have parts of the
parameters scattered in a linked list of buffers. The CU spontaneously pre-fetches the buffers in the
list.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
87
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
While the CU pre-fetches the address and byte count of one buffer, the transmit DMA is
transferring the previous buffer to the transmit byte machine. Completion of a buffer transfer by the
transmit DMA triggers the CU to initiate the transmit DMA for the next buffer (if it is already pre-
fetched) and to start the pre-fetch of the next buffer. The buffer pre-fetch cycle is terminated when
the transmit DMA reads the last buffer (indicated by the TBD number) and transfers it to the
transmit FIFO.
Internally, the controller CU performs the following sequence during transmission:
1. Begins execution of the transmit action command.
2. Reads and saves the TBD array address.
3. If the TCB size field is greater than zero, the device performs as follows:
a. If the TBD array address is not equal to all ones, the CU performs a pre-fetch and transfer
cycle, initiates the transmit DMA to the address of the first byte of the destination address
field in the CB and to the byte count of the last specified data byte in the command block.
b. If the TBD array address equals all ones, after completing DMA of the command block
the CU writes the end of command byte to the transmit FIFO.
4. If the TCB size field in the command block is zero, it runs a buffer pre-fetch and transfer cycle
and forces one dummy DMA completion.
5. The CU waits for completion of the transmit command. This includes only the transfer of the
whole frame to the transmit FIFO subsystem, not the frame transmission by the CSMA/CD
unit. At this point, the device posts the C bit (to 1) in the TCB. The CPU can reclaim the TCB
and associated data structures.
6. If transmission completed with a collision (but did not exceed the maximum collisions),
regardless of errors, the subsystem generates a re-transmit command and sends the data bytes
again from the FIFO. This causes re-transmission of the frame without any additional PCI bus
access.
7. If the transmit DMA encountered an underrun due to a lack of PCI bus bandwidth, it appends a
jam pattern to the end of the partially transmitted frame. Frames that are aborted during
transmission are jammed. Such an interruption of transmission can be caused by several
different events. Jamming will not start before completion of pre-amble transmission (before
the first byte of the destination address is sent). Collisions detected during transmission of the
last 11 bits of the frame will not result in jamming.
8. The device CU completes the transmit action command.
The device may report completion of a transmit command before the actual transmission on the
link has completed. Software can reuse the resources to prepare a new transmit command. When
the frame is eventually transmitted on the link, the CSMA/CD sub-system will return the status of
the transmission to the 82557 micro-machine, but the TxCB Status WILL NOT be updated in host
memory. The CU will update the internal Tx counters according to the Tx status
6.4.2.5.3
Framing Operation
The transmit byte machine maintains the following registers for construction of frames: pre-amble
pattern, SFD field, source address, CRC generator, and jam patterns.
After the transmit byte machine reads the transmit command from the transmit FIFO, a frame is
constructed and transferred to the transmit bit machine for bit and nibble transmission. The
transmit byte machine performs the following sequence:
1. Pre-amble bytes are transferred according to pre-amble length configuration parameter.
88
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
2. The SFD field is transferred.
3. Start CRC calculation.
4. Read and transfer the 6 destination address bytes from the transmit FIFO.
5. If the no source address insertion configuration parameter is zero, the individual address
should be transferred as the source address. Otherwise, the source address should be read and
transferred from the transmit FIFO. If the no source address insertion is 1 and there are less
than address length bytes in the transmit FIFO, a DMA underrun is forced.
6. All remaining bytes from the transmit FIFO are read and transferred. These are the length and
data fields.
7. The CRC is transferred.
8. If the device is configured to enable padding, the flag bytes (07Eh) are transferred
automatically so that a valid frame (64 bytes including CRC) is transferred onto the link.
If a collision or underrun occurred during transmission, the transmit byte machine completes the
transfer of the pre-amble and transfers 4 bytes of the jam pattern. If a collision occurred, the retry
counter is incremented. Jamming will not start before completing pre-amble transmission.
If a collision is detected during transmission of the last 11 bits in the frame, it does not result in
jamming. If the collision is detected during transmission of the last bit or later, the collision is not
reported and re-transmission does not occur. This can happen for an invalid frame shorter in length
than the slot time.
Note: A DMA underrun cannot logically occur during the pre-amble because the serial subsystem
generates its own pre-amble.
6.4.2.5.4
Delayed CNA Interrupts
The 82558 and later generation controllers have the ability to delay the CNA interrupt for a
predefined length of time, called the CNA interrupt delay (CID). If the CID is set to a non-zero
value, the device does not assert the interrupt immediately when entering a non-active state.
Instead, it initializes an internal counter with the CID parameter. The interrupt is asserted only
when the counter expires. If a CU resume or CU start command is issued while the counter is
counting, the interrupt will not be asserted. This opens a window for the device driver to set a new
command without the overhead of an additional interrupt service routine (ISR).
The device delays the interrupt, regardless of whether it is configured for CI interrupts or CNA
interrupts. However, the controller does not delay the updating of the CU status field. Therefore, if
the CID is greater than zero, it posts the CU status field (without the CNA bit) before it posts the
CNA bit and asserts the INTA# signal. (This feature is primarily targeted to NDIS systems but can
be beneficial for other systems as well.)
The CID parameter is set on a frame by frame basis, and its value is read by the device from the
TCB. Since the internal counter is automatically initialized to the CID value from the current TCB,
the existing value of the counter (set by the previous TCB) is overwritten, causing the counter to
reset even if it has not yet reached zero. This allows a rolling delay, where a number of back to
back TCBs can be given to the controller while only generating one interrupt at the end of the
chain.
The purpose of the delay is to avoid issuing this interrupt if it is not required. It is assumed that the
interrupt is not required in the following cases:
• The device was issued another action command and the CU returns to the active state.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
89
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
• The device received a frame and generated a receive interrupt.
If neither of these events occurred, the controller generates a CNA interrupt when the CID time
interval has elapsed. The actual delay experienced may be longer than the CID value that was
loaded. The CID is given in a granularity of approximately 256 PCI clocks and the maximum value
is 8192 clocks (which corresponds to 8 to 256 µs in a 33 MHz system).
The delayed CNA interrupt flow is outlined below.
1. The delayed CNA interrupt is issued in the suspend or idle state. In other words, if the device
is in the suspend or idle state, raising the interrupt would be delayed by specified time in the
CID field of each command header.
2. The end of receive processing cancels the pending delayed CNA interrupt. It also causes the
CNA interrupt to be set simultaneously with the frame interrupt, regardless of the internal
counter value. This is based on the theory that any pending transmit cleanup would be done in
the context of a receive interrupt.
3. Resume and start commands cancel pending delayed CNA interrupts. This allows only the last
TCB of a chain to be interrupted (the rolling delay).
4. The CX interrupt (caused by the I bit) is not affected in any way by this mode or delay
parameter. It may be that regardless of anything else, we may want to interrupt on, say, every
third TX in a chain to return resources to the protocol. This would be accomplished by setting
the I bit in the TxCB. There would be no delay associated with an I-bit interrupt. Note that if I
and S bits are set in a TxCB and the CID field is set to a non-zero value, the CX & CNA
interrupts will not occur together.
5. The delay specification is a 5-bit field and ranges between 8 and 256 µs, in 8 µs resolution.
The actual delay will only be within a certain percentage of the value specified (but never less
than the specified delay). The inaccuracy percentage is typically in the range of 10 to 20%.
However, in a few extreme conditions (for example, a lot of bad frames received), the delay
may be more than 20% above the specified delay.
The CNA interrupt delay (CID) field in the TCB is located in bits 28:24 of the first Dword of the
TCB.
6.4.2.6
Load Microcode (101b)
Note: Documentation for microcode is beyond the scope of this manual.
The load microcode command downloads a 64 Dword microcode patch to the device’s internal
microcode.
The microcode that operates on one device (for example, the 82557), will not operate on another
device (the 82558 or 82559). The load microcode command format is shown below:
Figure 21. Load Microcode Command Format
Offset
00h
Command Word Bits 31:16
0000000000 101
Status Word Bits 15:0
EL
S
I
C
X
OK XXXXXXXXXXXXX
04h
08h
Link Address (A31:A0)
First Microcode Dword
260h
64th Microcode Dword
90
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
This is the 32-bit address of the next command block. It is added to the CU base to
obtain the actual address.
Link Address
EL (Bit 31)
If this bit is set to one, it indicates that this command block is the last one on the CBL.
The CU will go from the active to the idle state after the execution of the CB is finished.
This transition will always cause an interrupt with the CNA/CI bit set in the SCB.
If this bit is set to one, the CU will be suspended after the completion of this CB. A CNA
interrupt will be generated if the device is configured for this. The CU transitions from the
active to the suspended state after the execution of the CB.
S (Bit 30)
If the I bit is set to one, the device generates an interrupt after the execution of the CB is
finished. If I is not set to one, the CX interrupt will not be generated.
I (Bit 29)
Bits 28:19
These bits are reserved and should all be set to 0.
CMD (Bits 18:16) This is the load microcode command, which has a value of 101b.
This bit indicates the execution status of the command. Software should reset this bit
before issuing the command to the device. Following a command completion, the device
sets it to one.
C (Bit 15)
NOTE: The difference in the definition of the C bit for the transmit command
The OK bit indicates that the command was executed without error. If it equals one, no
error occurred (command executed OK). If the OK bit is zero and the C bit is set, then an
error occurred.
NOTE: The difference in the definition of the C bit for the transmit command
OK (Bit 13)
Microcode Data
This field contains the 64 Dwords of microcode data downloaded to the device. This data
patches the device’s hard-coded microcode, which allows the behavior of the device to
be altered or adapted.
The load microcode command instructs the device to download microcode data from host memory
into its internal microcode RAM. The microcode data is organized as a 64-Dword memory block
that is appended to a standard command block header. The device starts execution of downloaded
microcode immediately following the successful completion of the load microcode command. The
device continues executing the downloaded microcode until the device is reset through its
hardware or software reset mechanisms.
®
Note: Documentation for developing new microcode patches for the Intel Fast Ethernet controllers is
beyond the scope of this manual.
6.4.2.7
Dump (110b)
This command causes the contents of various device registers to be placed in a memory area
specified by the user. It is supplied as a self diagnostic tool and provides registers of interest to the
user. The format of the dump command is shown below.
Figure 22. Dump Command Format
Offset
00h
Command Word Bits 31:16
0000000000 110
Status Word Bits 15:0
EL
S
I
C
X
OK XXXXXXXXXXXXX
04h
08h
Link Address (A31:A0)
Buffer Address
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
91
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
This is the 32-bit address of the next command block. It is added to the CU base to
obtain the actual address.
Link Address
EL (Bit 31)
If this bit is set to one, it indicates that this command block is the last one on the CBL.
The CU will go from the active to the idle state after the execution of the CB is finished.
This transition will always cause an interrupt with the CNA/CI bit set in the SCB.
If this bit is set to one, the CU will be suspended after the completion of this CB. A CNA
interrupt will be generated if the device is configured for this. The CU transitions from the
active to the suspended state after the execution of the CB.
S (Bit 30)
If the I bit is set to one, the device generates an interrupt after the execution of the CB is
finished. If I is not set to one, the CX interrupt will not be generated.
I (Bit 29)
Bits 28:19
These bits are reserved and should all be set to 0.
CMD (Bits 18:16) This is the dump command, which has a value of 110b.
This bit indicates the execution status of the command. Software should reset this bit
before issuing the command to the device. Following a command completion, the device
sets it to one.
C (Bit 15)
NOTE: The difference in the definition of the C bit for the transmit command
The OK bit indicates that the command was executed without error. If it equals one, no
error occurred (command executed OK). If the OK bit is zero and the C bit is set, then an
error occurred.
OK (Bit 13)
NOTE: The difference in the definition of the C bit for the transmit command
This field is a 32-bit offset to the dump area address. The size of the dump area is 596
bytes.
Buffer Pointer
Configuration parameters and contents of other registers are transferred from the CSMA/CD unit
through the status FIFO by the Command Unit to memory. The CU performs the following
sequence:
1. Starts the dump action command.
2. Writes the dump command byte to the transit FIFO.
3. Waits for the dump marker to return from the CSMA/CD module.
4. Dumps the FEXT and CSMA/CD registers content through the status FIFO.
5. Dumps the parallel registers.
6. Prepares the status word with C equal to 1 and the OK bit equal to 1.
7. Completes the action command.
Table 50. Dump Data Bytes (0-79)
Byte
D7
D6
D5
D4
D3
D2
D1
D0
0
1
2
3
4
FEXT RCV_WR Base Address Register (low)
FEXT RCV_WR Base Address Register (high)
FEXT RCV_WR Current Address Register (low)
FEXT RCV_WR Current Address Register (high)
FEXT RCV_RD Current Address Register (low)
92
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 50. Dump Data Bytes (0-79)
Byte
D7
D6
D5
D4
D3
D2
D1
D0
5
6
7
8
9
FEXT RCV_RD Current Address Register (high)
FEXT RCV_RD Base Address Register (low)
FEXT RCV_RD Base Address Register (high)
FEXT EXEC_WR Current Address Counter (low)
FEXT EXEC_WR Current Address Counter (high)
FEXT EXEC_WR Base Address Register (low)
FEXT EXEC_WR Base Address Register (high)
FEXT EXEC_RD Current Address Counter (low)
FEXT EXEC_RD Current Address Counter (high)
FEXT EXEC_RD Base Address Register (low)
FEXT EXEC_RD Base Address Register (high)
FEXT RCV_WR Byte Counter (low)
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
FEXT RCV_WR Byte Counter (high)
FEXT EXEC_WR Byte Counter (low)
FEXT EXEC_WR Byte Counter (high)
FEXT EXEC_WR Initial Threshold Register (low)
FEXT EXEC_WR Initial Threshold Register (high)
FEXT EXEC_WR Current Threshold Register (low)
FEXT EXEC_WR Current Threshold Register (high)
Configure Byte 8
Configure Byte 9
Configure Byte 10
Configure Byte 11
Configure Byte 12
Configure Byte 13
Configure Byte 14
Configure Byte 15
Configure Byte 16
Configure Byte 17
Configure Byte 18
Configure Byte 19
Configure Byte 20
Configure Byte 21
Reserved
Individual Address Register 1
Individual Address Register 2
Individual Address Register 3
Individual Address Register 4
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
93
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 50. Dump Data Bytes (0-79)
Byte
43
D7
D6
D5
D4
D3
D2
D1
D0
Individual Address Register 5
Individual Address Register 6
Transmit Status (low byte)
Transmit Status (high byte)
Transmit CRC 0
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
Transmit CRC 1
Transmit CRC 2
Transmit CRC 3
Receive CRC 0
Receive CRC 1
Receive CRC 2
Receive CRC 3
Temporary Memory 0
Temporary Memory 1
Temporary Memory 2
Temporary Memory 3
Temporary Memory 4
Temporary Memory 5
Receive Status (low byte)
Receive Status (high byte)
Hash Register 0
Hash Register 1
Hash Register 2
Hash Register 3
Hash Register 4
Hash Register 5
Hash Register 6
Hash Register 7
X
X
X
1
X
X
X
1
X
X
X
1
X
X
X
1
X
X
X
1
X
X
X
1
X
X
X
X
X
X
X
X
Receive Length (high)
Receive Length (low)
77 – 79 Reserved
94
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 51. Dump Data Dwords (20-148)
Dword
Byte 3
Byte 2
Byte 1
20 – 29
30
Reserved
Micromachine (MM) Register File 39
Micromachine (MM) Register File 38
Micromachine (MM) Register File 37
Micromachine (MM) Register File 36
Micromachine (MM) Register File 35
Micromachine (MM) Register File 34
Micromachine (MM) Register File 33
Micromachine (MM) Register File 32
Receive Short Frame Errors
31
32
33
34
35
36
37
38
39
Receive CDT Errors Counter
Receive Overrun Errors Counter
Receive Resource Errors Counter
Receive Alignment Errors Counter
Receive CRC Errors Counter
Receive Good Frames Counter
Transmit Total Collisions Counter
Transmit Multiple Collisions Counter
Transmit Single Collisions Counter
Transmit Deferred Counter
40
41
42
43
44
45
46
47
48
49
Transmit Lost CRS Counter
50
Transmit Underrun Errors Counter
Transmit Late Collisions Errors Counter
Transmit Maximum Collisions Errors Counter
Transmit Good Frames Counter
Reserved
51
52
53
54 – 64
65
Reserved
Execution Status Port
66
SCB Port Register
67
SCB Pointer register
SCB Interrupt SCB CUC and
SCB CUS and
RUS Byte
68
69
SCB Interrupt Byte
SCB Interrupt Byte
Mask Byte
RUC Byte
SCB Interrupt SCB CUC and
SCB CUS and
RUS Byte
Mask Byte
Reserved
RUC Byte
70 – 81
82
DMA Arbitration Registers
Reserved
83 – 85
86
Micromachine Register File 31
Micromachine Register File 30
87
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
95
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 51. Dump Data Dwords (20-148)
Dword
88
Byte 3
Byte 2
Byte 1
Micromachine Register File 29
Micromachine Register File 28
Micromachine Register File 27
Micromachine Register File 26
Micromachine Register File 25
Micromachine Register File 24
Micromachine Register File 23
Micromachine Register File 22
Micromachine Register File 21
Micromachine Register File 20
Micromachine Register File 19
Micromachine Register File 18
Micromachine Register File 17
Micromachine Register File 16
Micromachine Register File 15
Micromachine Register File 14
Micromachine Register File 13
Micromachine Register File 12
Micromachine Register File 11
Micromachine Register File 10
Micromachine Register File 9
Micromachine Register File 8
Micromachine Register File 7
Micromachine Register File 6
Micromachine Register File 5
Micromachine Register File 4
Micromachine Register File 3
Micromachine Register File 2
Micromachine Register File 1
Micromachine Register File 0
Micromachine LFSR
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
Micromachine Bit Flag Array 6
Micromachine Bit Flag Array 5
Micromachine Bit Flag Array 4
Micromachine Bit Flag Array 3
Micromachine Bit Flag Array 2
Micromachine Bit Flag Array 1
Micromachine Bit Flag Array 0
96
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 51. Dump Data Dwords (20-148)
Dword
126
Byte 3
Byte 2
Byte 1
Micromachine Input Port 3
Micromachine Input Port 2
Micromachine Input Port 1
Micromachine Input Port 0
Micromachine ALU
127
128
129
130
131
132
133
134
135
136
137
138
139
140
Micromachine Temporary B Register
Micromachine Temporary A Register Rotated Right
Micromachine Temporary A Register
Transmit DMA Byte Count Register
Micromachine Input Port Address Register
Transmit DMA Address Register
Micromachine Output Port Register
Receive DMA Byte Count Register
Micromachine Output Port Address Register
Receive DMA Address Register
141 – 142 Reserved
143
144
145
146
147
148
DIU Control Register
Micromachine ALU Shifted by Byte
Reserved
BIU Control Register
Reserved
Micromachine Status Register
6.4.2.8
Diagnose (111b)
The diagnose command triggers an internal self-test procedure that checks the internal device
hardware. Its format is illustrated below.
Figure 23. Diagnose Command Format
Offset
00h
04h
Command Word Bits 31:16
0000000000 111
Link Address (A31:A0)
Status Word Bits 15:0
XXXXXXXXX
XX
EL
S
I
C
X
OK
X
F
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
97
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
This is the 32-bit address of the next command block. It is added to the CU base to
obtain the actual address.
Link Address
EL (Bit 31)
If this bit is set to one, it indicates that this command block is the last one on the CBL.
The CU will go from the active to the idle state after the execution of the CB is finished.
This transition will always cause an interrupt with the CNA/CI bit set in the SCB.
If this bit is set to one, the CU will be suspended after the completion of this CB. A CNA
interrupt will be generated if the device is configured for this. The CU transitions from the
active to the suspended state after the execution of the CB.
S (Bit 30)
If the I bit is set to one, the device generates an interrupt after the execution of the CB is
finished. If I is not set to one, the CX interrupt will not be generated.
I (Bit 29)
Bits 28:19
These bits are reserved and should all be set to 0.
CMD (Bits 18:16) This is the diagnose command, which has a value of 111b.
This bit indicates the execution status of the command. Software should reset this bit
before issuing the command to the device. Following a command completion, the device
sets it to one.
NOTE: The difference in the definition of the C bit for the transmit command
C (Bit 15)
OK (Bit 13)
F (Bit 11)
The OK bit indicates that the command was executed without error. If it equals one, no
error occurred (command executed OK). If the OK bit is zero and the C bit is set, then an
error occurred.
NOTE: The difference in the definition of the C bit for the transmit command
This bit indicates the results of the self-test procedure. A 0 indicates a pass, and a 1,
failure.
The diagnose command checks the following internal device circuitry:
• Exponential backoff random number generator (linear feedback shift register).
• Exponential backoff time-out counter.
• Slot time period counter.
• Collision number counter.
• Exponential backoff shift register.
• Exponential backoff mask logic.
This procedure checks the operation of the backoff block, which resides in the serial side and is not
easily controlled. The CU triggers the self-test procedure of the serial subsystem. It performs the
following sequence:
1. Begins execution of the diagnose action command.
2. Waits for command completion.
3. Prepares the status word with the C bit equal to 1, OK equal to 1, and F equal to 0 if the
diagnose succeeded. Otherwise, the status word has the C bit equal to 1, OK equal to 0, and F
equal to 1.
4. Completes action command.
The CSMA/CD module performs the self test procedure in two phases: phase 1 tests the counters
and phase 2 tests the trigger logic.
98
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
During Phase 1, the linear feedback shift register (LFSR), exponential backoff time-out, slot time,
and collision counters are checked. The test is performed in the following manner:
1. All counters and shift registers are reset simultaneously.
2. The unit starts counting and shifting the registers.
3. The exponential backoff shift register reaches all ones.
4. The unit checks the exponential backoff shift register for all ones when the LFSR content is all
ones in its 10 least significant bits.
5. The unit stops counting when the LFSR (30 bits) reaches a specific state, and the exponential
backoff counter (10 bits) wraps from all ones to all zeroes. Simultaneously, the slot time
counter switches from 01111111111 to 10000000000, and the collision counter (4 bits) wraps
from all ones to all zeroes.
6. Phase 1 is successful if the 10 least significant bits (when applicable) of all four counters are
all zero.
During Phase 2, the test is performed in the following steps:
1. The exponential backoff shift register, LFSR, and all counters are reset.
2. The exponential backoff logic is temporarily configured accordingly:
a. SLOT-TIME = 3h
b. LIN-PRIO = 6h
c. EXP-PRIO = 3h
d. BOF-MET = 0h
3. Transmission and collisions are emulated internally.
4. If the most significant bit of exponential backoff shift register is 0, then step 3 is repeated.
If Step 4 is successful, then a passed status is returned; otherwise, a failed status is returned.
6.4.3
Receive Operation
6.4.3.1
Receive Frame Area
The 8255x supports the concept of a receive frame area (RFA). The RFA is the list of free receive
resources and consists of Receive Frame Descriptors (RFDs). The RFDs contain data buffers
capable of holding maximum Ethernet size packets immediately following the RFD header. This
constitutes the simplified memory model. Each receive frame is described by one RFD.
6.4.3.1.1
Simplified RFA Structure
In the simplified RFA structure, the data portion of the received frame (including the Ethernet
header) is part of the RFD and is located in contiguous memory immediately after the size field in
the RFD. The simplified memory structure is shown in the figure below.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
99
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Figure 24. Simplified Memory Structure
SCB
RECEIVE FRAME AREA
RFD
RFD
RFD
RFD
Sequential
Data Buffer
Sequential
Data Buffer
Sequential
Data Buffer
Sequential
Data Buffer
6.4.3.1.2
Receive Frame Descriptor Format
Figure 25. Receive Frame Descriptor Format
Offset
Command Word Bits 31:16
000000000 SF
Status Word Bits 15:0
00h
04h
08h
0Ch
EL
S
H
000
C
0
F
OK
Status Bits
Link Address (A31:A0)
Reserved
0
0
Size
EOF
Actual Count
EL (Bit 31)
S (Bit 30)
H (Bit 20)
The EL bit indicates that this RFD is the last one in the RFA.
The S bit suspends the RU after receiving the frame.
The H bit indicates if the current RFD is a header RFD. If it equals 1, the current RFD is
a header RFD, and if it is 0, it is not a header RFD.
NOTE: If a load HDS command was not previously issued, the device disregards this
bit.
SF (Bit 19)
C (Bit 15)
The SF bit equals 0 for simplified mode.
This bit indicates the completion of frame reception. It is set by the device.
The OK bit indicates whether the frame was received without any errors and stored in
memory. If the last frame was received with sufficient memory space, the OK bit will be
set, even if it was the last RFD in the RFA with the EL bit set. After receiving the frame,
the device enters the no resource condition, generates an RNR interrupt, and starts
discarding frames until the RU is restarted with sufficient resources.
OK (Bit 13)
Status Bits
(Bits 12:0)
This field contains the results of the receive operation:
100
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
The link address is a 32-bit offset to the next RFD. It is added to the RU base. The link
address of the last frame can be used to form a cyclical link to the first RFD.
Link Address
Size
This field is used in the simplified mode and represents the data buffer size. In the
header RFD, the size field identifies the data buffer size excluding the header area. The
size value should be an even number.
This bit is set by the device when it has completed placing data in the data area. Before a
new RFD can be included in the RFA, the EOF bit must be cleared by software.
EOF
This bit is set by the device when it updates the actual count field. Before a new RFD can
be included in the RFA, the F bit must be cleared by software.
F
Actual Count
The number of bytes written into the data area.
Table 52. RFD Status Bit Descriptions
Status
Bit
Description
Bit 12
Bit 11
This bit is reserved.
CRC error in an aligned frame. This bit may be set only in the save bad frames mode.
Alignment error (CRC error in misaligned frame). This bit identifies the number of bits in the frame
that were not an octet multiple. This bit may be set only in save bad frames mode.
Bit 10
Ran out of buffer space; no resources. This bit indicates that the incoming frame was larger than
the possible receive data area for that frame. In simplified mode, this means that the size of the
RFD was not large enough to accommodate the entire frame.
NOTE: This status bit is not related to the RFD EL and S bits. Status bit 9 reflects the status of the
current frame. The EL and S bits control the RU machine status after the reception of the
current frame.
Bit 9
Bit 8
Bit 7
Bit 6
Bit 5
DMA overrun failure to acquire the system bus. This bit may be set only in save bad frames mode.
Frame too short. It implies that the length of the received frame was less than 64 bytes including
the CRC. This bit may be set only in save bad frames mode.
This bit is reserved.
Type/Length. If this bit is set, it indicates that the received frame was a type frame (the value of the
Ethernet header Type/Length field was either 0 or greater than 1500 decimal).
Receive Error. This bit is set if the RX_ER pin was asserted at least once during frame reception. It
is set by the PHY when it detects a character error. The actual detection of character error depends
on the PHY technology and the algorithm implemented. The device increments the device CRC
error counter when this bit is set. This bit may be set only in save bad frames mode.
Bit 4
Bit 3
This bit is reserved.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
101
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 52. RFD Status Bit Descriptions
Status
Bit
Description
No address match. If this bit is set to 1, the destination address of the received frame does not
match the individual address, multicast address, or the broadcast address in the filter. For
example, this bit will be set when the device is in promiscuous mode and the destination address
does not match any of the other address filtering mechanisms.
Bit 2
Bit 1
IA match bit. When this bit equals 0, it implies that the destination address of the received frame
matches the individual address. When it is 1, the destination address of the received frame does
not match the individual address. For example, a multicast or broadcast address will set this bit to
a 1.
Receive collision (82557 and 82558 only). When this bit is set, it indicates a collision was detected
during a reception.
TCO indication (82559 and later generation controllers). For the 82559 and later generation
controllers, this bit no longer reflects collision detection. Instead, it indicates that the device is
processing a TCO packet. In an environment with a TCO controller where receive to TCO is
enabled, the controller introduces a new behavior. TCO packets are first posted to the RFA before
they are transferred back to the TCO controller. In most cases this process is completely
transparent to the software driver since the RFD that contains the TCO packet is reclaimed.
However, if header RFDs are being used and the TCO packet size is larger than the header size,
then the RFD is not reclaimed. In this case, a bad status indication will be posted to memory. The
bad status indication should be used by the software driver to discard the packet.
Bit 0
Table 53. Actual Count in Header RFD
F
EOF
Actual Count
0
0
1
1
0
1
0
1
Invalid
Invalid
HDS size
Total byte count
6.4.3.2
6.4.3.3
Initial Receive Frame Area Structure
To enable the device to receive frames, software must setup the following structure:
1. The SCB general pointer in the SCB should point to the first RFD on the list.
2. The link offset of each RFD in the list should point to the next RFD.
3. The EL bit in the last RFD should be set.
Operation of Frame Reception
The serial subsystem of the device selects the frames destined for the station according to the
destination address of the frames passing on the link. A frame is selected if it is at least 6 bytes long
and its address matches either the individual address, multicast address, multiple IA, or broadcast
address (promiscuous mode). It transfers the selected frames with their status to the receive FIFO.
The receive DMA unit transfers the frame from the receive FIFO to host memory under control of
the receive unit. If discard short frames is enabled, any receive frame shorter than 64 bytes on the
link (including padding and CRC) will be completely discarded. Although these frames are
discarded, they are still counted in the short frame counter.
102
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
For every frame, the RU configures a RFD in memory. The loading of each buffer is done by the
receive DMA in parallel with pre-fetching the next buffer by the RU. After completing frame
reception, the RU closes the last RFD and configures the structure for receiving the next frame.
6.4.3.3.1
Configuring the Next RFD
The RU performs the following sequence to set up a RFD.
1. Reads the 4 Dwords of the current RFD and saves the EL and S bits internally.
2. Analyzes the link offset of the current RFD and saves it as the address of the next RFD.
3. Initiates a receive DMA if the size of the data field in the RFD is greater than zero. The receive
DMA is initiated with the address of the first byte of the destination address to the byte count
specified by the RFD.
4. Forces a receive DMA completion if the size of the data field in the RFD is zero.
5. Goes to the buffer pre-fetch and transfer cycle.
6.4.3.3.2
Close Frame
When the RU reads an end of frame from the receive FIFO (indicating the end of a received or
discarded frame), it performs the following sequence to close the frame:
1. Reads the status from the receive FIFO and saves it internally.
2. Skips to the completion of reception sequence if the RU is not in the ready state.
3. Reclaims the RFD if the frame status indicates that there is an error (including short frame)
and the save bad frame configuration parameter is zero. In addition to reclaiming the RFD, the
device backs up the current RFD to the next RFD and jumps to step 8. If a header RFD is being
used and the header field was filled, the RFD is not reclaimed.
4. Writes C equal to 1, OK equal to 0, status bit 9 equal to 1, and the remaining bits from the
CSMA/CD module to the status word of the current RFD if the RU ran out of resources during
reception of this frame.
5. Writes C equal to 1 and the remaining bits from the CSMA/CD module to the status word if
the RU did not run out of resources.
6. Requests FR interrupt.
7. Creates a new RFD if the S bit was not set and the RU did not run out of resources.
8. Goes to completion of reception of frame.
6.4.3.3.3
Completion of Reception
Reception completion occurs when the RU encounters an end of frame regardless of its state. The
procedure is determined by the following: EL bit, S bit, RU start request, and RU resources. The
following sequence is performed by the RU at the completion of reception:
1. If the RU ran out of buffers or frame descriptors during current frame reception, then the state
changes to no resources. The RU requests an RNR interrupt, sets the internal S bit, and starts
discarding frames.
2. If the S bit of current frame is set but the RU start request is not and the RU is in the ready
state, an RNR interrupt request is initiated and the state is changed to suspended.
3. If the RU start request bit is set but the S bit is not, then the RU state is changed to ready and a
new RFD is created.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
103
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
4. If the RU start request and S bits are set, a new RFD is created and the state is changed to
suspended.
5. If the RU state is not ready, frames should be discarded.
6. If the RU is in the ready state or has just exited the ready state, the following steps are
performed:
a. Update the SCB status word according to the new state and clear internal interrupt flags.
b. Activate the hardware interrupt signal.
6.4.3.4
No Buffer Performance Improvements (82558 and 82559)
During normal receive operation, the 82558 and 82559 treat received packets in the same manner
as the 82557. The data structures and driver-hardware interface remain unchanged. However, some
internal modifications were made for performance improvement.
In a no resources situation, the 82558 and 82559 operate differently than the 82557. In this
scenario, the 82557 starts discarding frames when it reaches a no resources or suspended state. The
82558 and 82559 start storing all incoming frames in the receive FIFO in these states. When the
FIFO is full, the next frame is marked as an overrun frame while the following frames are lost. This
enables the network device driver to react to the no resources situation and successfully receive the
following two or more frames.
6.4.3.4.1
TCP/UDP Checksum Support (82559 only)
The 82559 provides a checksum word in the receive structure if it is configured to (Section 6.4.2.3,
“Configure (010b)”). The 82557 and 82558 do not have this capability.
The checksum word is calculated on the incoming packet excluding the MAC header (first 14 bytes
of the packet) and Ethernet CRC. The checksum word is always appended at the end of the data
posted to the receive buffer(s) with the least significant byte first. If the 82559 is configured to post
the Ethernet CRC into memory, then the checksum word follows the CRC. The byte count field in
the receive memory structure(s) includes the checksum word. If software enables this feature,
software must subtract 2 bytes from the reported length of the packet to determine the actual packet
length.
If the incoming frame is a TCP or UDP packet, the device driver can accelerate the checksum word
calculation using this capability in the 82559. The driver should subtract from the 82559 checksum
word non-relevant fields within the packet, negate the result (1’s complement), and compare it to
the packet’s TCP/UDP checksum word.
The 82559 calculates the checksum word as a sum of the incoming words of the received packet
(not including the MAC header and CRC bytes). After the summation of each two words, the carry
is added to the least significant bit of the result. If the packet has an odd byte count, the last byte is
padded with 8 zeroes as its most significant bits. The following equation and example demonstrate
this:
Example 1. 82559 Checksum Calculation
Assume the following incoming Packet: SA DA Type B B B B B B … B
B
B
CRC
0
1
2
3
4
5
2N-2 2N-1 2N
Checksum = { B B + B B + C + B B + C + …+ B
B
+ C
+ 00B + C
},
N-1
1
0
3
2
0
5
4
1
2N-1 2N-2
N-2
2N
where C , C , … C
are the carry out results of the intermediate sum operations.
0
1
N-1
104
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Example 2. Numerical Calculation
Assume the following incoming packet: SA DA Type F1 F5 54 79 E7 9E F5 CRC
S = F5F1+7954 = 6F45, C = 1
0
0
S = 6F45 + 9EE7 + 1 = 0E2D,
C = 1
1
1
S = 0E2C + 00F5 + 1 = 0F23,
C = 0
2
2
Check Sum = S = 0F21+ 0 = 0F23
3
When all data bytes are written to the memory, the device writes the actual count. The device
writes the frame status to the RFD status word. The device asserts an interrupt to indicate the end of
receive processing. The driver can poll in memory for the frame status and mask the last interrupt.
6.5
Command Unit and Receive Unit Operation
6.5.1
Starting and Completing Control Commands
Software can issue control commands by writing to the RUC and CUC fields of the SCB command
word. The SCB CU and RU command fields are two fields in the lower byte of the SCB command
word, called the SCB command byte. Since the 8255x clears the SCB command byte when the
control command is accepted:
• Software must wait for this byte to be cleared before the next control command can be issued.
• CU and RU control commands must never be issued together in the same SCB write cycle.
The 8255x does not necessarily accept the control commands immediately after they are written to
the SCB since it may be engaged in higher priority tasks. For example, the device may be pre-
fetching new buffers, handling buffer switches, or finishing frame reception. When the device is
becomes available, it performs the start of control command sequence described below.
1. Reads the SCB command byte.
2. Reads the SCB general pointer.
3. The 8255x issues an internal request to the RU to perform an RU command acceptance
sequence if the RUC field is not zero.
4. The acceptance sequence for a CU command is performed immediately if the CUC field is not
zero.
After the CU and RU have completed the acceptance sequence, the 8255x updates the SCB status
according to the internal CU status, RU status, and interrupt requests.
6.5.2
Generating and Acknowledging Interrupts
When the CPU is interrupted by the 8255x, it should acknowledge the interrupt by setting the
corresponding acknowledge bits in the SCB interrupt acknowledge byte. When the CPU writes to
this byte, the corresponding interrupt bits are immediately cleared. If all interrupt bits are cleared,
the 8255x clears its INTA# line. If the interrupt service routine is likely to process all pending
interrupts, then all the bits can be acknowledged in one PCI write cycle.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
105
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
6.5.3
Command Unit Control
The CU is the 8255x logical unit that executes action commands from the command block list
(CBL). This section describes how software controls the execution of action commands.
Specifically, start, stop, suspend, or resume of the CU are discussed. The CU can be modeled as a
logical machine that exists in one of the following states at any given time:
• Idle. The CU is currently not executing an action command and is not associated with a CB in
the CBL. This is the initial state. It is also the state reached after the CU finishes executing a
CBL where the last CB had an EL bit set. A CU start command must be issued to begin
execution on a new CBL.
• Suspended. The CU is not executing a CB but has read a next link pointer in the last CB that
it executed before it suspended execution. A CU resume command forces the 8255x to
continue execution from the CB at the next link address.
• Active. The CU is currently executing an action command.
Software affects CU operation in two ways: issuing a CU control command or setting bits in the
command word of the action command. This causes the CU to do one of the following:
• Start executing a list of action commands.
• Resume execution of a list of action commands.
• Stop execution after completing an action command. Usually, this is the last command block
in the list.
• Suspend execution after completing an action command (if the S bit was set).
• Issue interrupts after completing action commands.
There are two important points of time in the execution of commands:
• Acceptance Time. This is the time following a write to the SCB command byte. It is when
the CU reads the control command, clears it, and begins execution.
• Completion of Execution. This is when the CU completes executing a command.
The CU uses two internal flags to remember requests from acceptance time that are to be acted on
at completion of execution: CU_START_REQUEST and CU_RESUME_REQUEST.
At command acceptance time, after the 8255x has finished higher priority tasks and detected a new
command, the device reads and analyzes the SCB command byte. (Higher priority tasks can
include: receive end of frame processing, receive or transmit buffer pre-fetching, completion of
transmit DMA, completion of action commands, and dumping counters.) The CUC field indicates
one of the following commands: CU_START or CU_RESUME.
6.5.3.1
CU Start Command
When the CU detects the CU Start (CU_START) command, it begins executing the first action
command in the list. It is prohibited from issuing a CU_START if the CU is in the active state. This
means that software must ensure that the CU is idle or suspended before issuing the CU_START
command. Upon acceptance of a CU_START command, the CU initiates the following sequence:
1. Reads the CB offset from the SCB (general pointer register) and saves it as a pointer to the first
CB in the CBL.
2. Becomes active and starts execution from the beginning of the CBL if the CU is not in the
active state.
106
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
6.5.3.2
CU Resume Command
The CU Resume (CU_RESUME) command resumes CU operation. The 8255x completes the
following sequence:
1. If the CU is in the suspended state it goes to the active state and requests the beginning of the
next CB. Since the 8255x remembers the pointer to the next action command, it does not re-
read the link pointer in the previous CB. However, the 8255x re-reads the S bit in the previous
CB to determine if the driver has modified it while the device was suspended. If the S bit is
cleared, it proceeds to execute the new CB. If the S bit is still set in the previous CB, the CU
goes back to the suspended state.
2. If the CU is in the active state, it verifies the validity of the S bits in the current and next action
commands. If the S bit is cleared in the current CB, it proceeds to the next CB in the list after
execution of the current CB is completed.
3. If the CU is in the idle state, it ignores the CU_RESUME command.
6.5.3.3
CU Control Commands Response
The start and resume CU control (CUC) commands can be issued through the SCB CUC command
field. The command block offset pointer in the general pointer register of the SCB points to the first
executable action command. These action commands have a field for the end of list (EL) and
suspend (S) bits. The EL bit indicates that the current action command is the last on the command
list. The presence of the S bit indicates that the user wants the CU to enter the suspended state after
The mechanism used to suspend the CU is the S bit in the action command block. Suspending the
CU through the action command block results in suspension after a specific command is executed.
The CU can be re-activated by issuing a resume command.
Table 54. CU Control Commands: Actions at Acceptance Time
CU Start
CU Resume
Present State
Next State
Active
Action
Next State
Idle
Action
Start processing CB
pointed to by SCB
general pointer.
Idle
None.
Start processing CB
pointed to by SCB
general pointer.
Start processing next CB
if S bit is clear on current
CB.
Suspended
Active
Active
Active
Active
Re-check S bit on current
CB. If not set, process
next CB.
Prohibited
Not applicable.
Table 55. CU Activities Performed at the End of Execution
EL Bit
S Bit
Next State
Action
0
0
1
1
0
1
0
1
Active
Start processing CB
Suspended CNA interrupt
Idle
Idle
CNA/CI interrupt
CNA/CI interrupt
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
107
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
6.5.4
Receive Unit Control
The receive unit (RU) is the logical unit that receives frames and stores them in memory. It uses
free buffers and descriptors prepared by the CPU. This section describes how the CPU controls
frame reception (starts, stops, suspends, and resumes the RU). Reception can also be halted due to
a no resource condition.
The RU is modeled as a logical machine that takes one of the following states at any given time.
Software can determine the current RU status by reading the SCB status word in the CSR (bits 5:2).
• Idle (0000). The RU has no memory resources and is discarding incoming frames. This is the
initial RU state after reset.
• No Resources Due to No More RFDs (0010). The RU has no memory resources due to a
lack of RFDs and is discarding incoming frames. This state differs from the idle state in that
the RU accumulates statistics on the number of frames it has to discard. The 8255x enters this
state after it processes an RFD that its EL bit set.
• Suspended (0001). The RU discards all incoming frames even though free memory resources
exist to store incoming frames. The 8255x enters this state after it processes an RFD with its S
bit set.
• Ready (0100). The RU has free memory resources and is ready to store incoming frames.
Table 56. RU Control Commands: Actions at Acceptance Time
RU Start
RU Resume
RU Abort
Present
State
Next
State
Next
State
Next
State
Action
Action
Action
Set Up
RFD
Idle
Ready
Idle
None
Idle
None
No
Resources
due to no
RFDs
Set Up
RFD
Ready
Prohibited N/A
Idle
Idle
Idle
Idle
Idle
Idle
None
None
Set Up
RFD
Set Up
RFD
Suspended Ready
Ready
Start
Discard -
RNR
Ready
Prohibited N/A
Prohibited N/A
Interrupt
Request
Start
Idle
No
Idle
No
Idle
None
None
None
None
Resources Resource Request
due to No
RFDs
Prohibited N/A
due to No Start
RFDs
Request
Start
Request
Resume
Suspended Suspend
Suspend
Abort
DMA -
Start
Discard -
RNR
Ready
Prohibited N/A
Prohibited N/A
Idle
Interrupt
108
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Frames arrive at the device independent of the state of the RU. When a frame is arriving, the 8255x
is referred to as actively receiving, even when the RU is not in the ready state and the frame is
being discarded.
Software can affect RU operation in three ways: by issuing an RU control command, by setting bits
in the RFD. In general, software can cause the RU to do the following:
• Start frame reception.
• Resume reception if the RU is in the suspended state.
• Abort reception immediately and return to the idle state.
• Stop reception after a specified RFD is filled (frame received). This RFD is referred to by the
8255x as the last RFD on the list.
• Suspend reception after a specified RFD is filled (frame received).
The RU issues an interrupt after every received frame. The 8255x may issue two interrupts for one
frame if the RU is using header RFDs. There are two important points of time in frame reception:
• Acceptance time. This is the time after an RU command is issued by software (written to the
SCB command byte). It is when the RU reads the control command, takes initial action, and
clears the SCB command byte.
• Completion of Reception. This is the time the RU finishes receiving or discarding an
incoming frame.
The RU uses an internal RU start request flag to remember RU start requests from acceptance time
that are to be acted on at completion of reception. The RU starts analyzing the command at
acceptance time after the CU has passed the control command to the RU and finished accepting its
own control command. The SCB RU control field can take one of the following values:
RU_START, RU_ABORT, or LOAD_HDS.
6.5.4.1
RU Start Command
For the RU start (RU_START) command, the CPU activates the RU for frame reception. At
acceptance time, the RU may or may not be actively receiving a frame. The RU performs the
following when it is actively receiving a frame when an RU_START command is accepted:
1. Reads the RFA Offset word from the SCB and saves it internally as the pointer to the next
RFD.
2. Sets the internal RU_START_REQUEST flag. When the current frame has been received or
discarded the RU goes to the Ready state and sets up the next RFD.
When an RU_START command is accepted, the RU performs the following sequence when NOT
actively receiving a frame.
Clears the internal RU_START_REQUEST flag.
Reads the RFA Offset word from the SCB and saves it internally as the pointer to the next RFD.
If the RU is not in the READY state and the DMA did not transfer any data to the current RFD, it
does the following. It stops discarding, goes to the READY state, gives up the pre-fetched current
buffers, and sets up a new RFD. Setting up a new RFD uses the pointer to the next RFD to prepare
to receive the next frame.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
109
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
6.5.4.2
RU Resume Command
The RU Resume (RU_RESUME) command resumes frame reception. The RU performs the
following tasks:
1. The RU goes to the ready state and configures a new RFD if the RU is in the suspended state
and not actively discarding a frame.
2. The RU sets the RU resume request flag if the RU is in the suspended state and the device is
discarding a frame.
3. The RU ignores the command if it is not in the suspended state.
6.5.4.3
RU Abort Command
The RU abort (RU_ABORT) command immediately stops frame reception and enters the idle state.
1. The RU requests an RNR interrupt if the RU is in the ready state.
2. The RU stops all DMA activity and starts discarding incoming data.
3. The device transitions the RU to the idle state.
6.5.5
Updating SCB Status
The device updates the SCB status word when any of the following events occur.
1. When a control command is accepted. (The 8255x updates the SCB status and clears the SCB
command word to indicate that acceptance has completed.)
2. When the CU or RU changes state.
3. After receiving a frame.
4. When a command with its I bit set is completed.
Note: The device may first clear the SCB command byte and return at a later time to update new status in
the SCB status byte.
6.6
Flow Control
The 82557 does not have any support for flow control. However, the 82558 and later generation
controllers support frame based (IEEE) flow control (FC). In addition, the 82558 also supports a
proprietary PHY based mechanism for flow control (known as Bay flow control).
Flow control is implemented as a means of reducing the possibility of receive buffer overflows,
which can result dropped received packets, and allows local control of network congestion levels.
This can be accomplished by sending an indication to a transmitting station of an almost full
receive buffer condition at a receiving station.
110
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
6.6.1
PHY Based Flow Control
The 82558 supports the PHY based flow control scheme known as the “Bay Flow Control”
scheme. This scheme is supported only when the 82558 is operating using its internal PHY TX
unit. It is not supported when an external PHY is used through the MII. The Bay Flow Control
scheme is activated by setting bit 15 in MII register 16 in the 82558 PHY TX unit.
The 82558 generates flow control pause indications according to the number of bytes in its receive
FIFO. This pause indication tells the 82558’s link partner to pause (or delay) future transmissions
to allow the 82558 some time to empty its receive FIFO. The threshold at which the 82558 triggers
the pause indication is determined by the flow control threshold register. PHY based flow control
should not be used at the same time as frame based flow control.
Since frame based flow control is an IEEE 802.3x standard, it is recommended that software use
the frame based FC method instead of PHY based FC.
6.6.2
Frame Based Flow Control
The 82558 and later generation devices support the frame based flow control scheme as specified
in the IEEE 802.3x standard. Frame based flow control operates independently of the PHY. Thus,
this method is supported with either the 8255x (except the 82557) internal PHY or an external
PHY.
These devices also support a priority aware flow control scheme, which is a non-standard variation
Frame based flow control should be negotiated with a device’s link partner. This is accomplished
through the N-Way auto-negotiation algorithm. To advertise frame based FC capability, the device
should have their pause operation for full duplex (FDX) links bit set in the auto-negotiation
advertisement register (MDI register 4, bit 10 [Section 7.2.4, “Auto-Negotiation Advertisement
Register: Register 4”]). A device should enable the pause functionality only if both the local device
and the link partner advertise this capability and the negotiated link utilizes a FDX technology
(regardless of data rate).
The 82558 defaults to advertising that it is not capable of FDX FC. The 82559 defaults to
advertising that it is capable of FDX FC. To enable this feature on the 82558, software needs to
configure the corresponding bit in the ability advertisement register and force re-negotiation.
6.6.2.1
Protocol Description
IEEE 802.3x flow control (FC) is defined as point to point flow control. This means that it is
relevant only for two devices connected by a dedicated link. This form of flow control is
implemented in full duplex only. It includes a set of commands (and indications) for stopping and
re-starting frame transportation between the two connected devices. The commands are encoded
into special frames.
Flow control frames are the minimum Ethernet frame length (64 bytes). The flow control frame
includes the following fields:
Special DA (6 bytes), SA (6 bytes), Type (2 bytes), Command (2 bytes), Parameters (2 bytes),
Pad (42 bytes), CRC (4 bytes).
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
111
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 57. Flow Control Frame Format
S
one byte
Pre-amble
SFD
up to 6 bytes
one byte
Destination
Address
6 bytes
Source Address 6 bytes
Type/Length
2 bytes
2 bytes
MAC Control
Opcode
MAC Control
Parameters
FCS
T
4 bytes
one byte
A flow control frame is identified by a special type field (bytes transmitted left to right): 88 to 80.
The reception of a FC frame can be done either through the regular individual address filtering
mechanism or by a special multicast address detection mechanism. Since the transmitting MAC
may not know the individual address of the other station, it may not be able to send FC frames to
the individual address of the other node. A special multicast destination address is defined for use
in FC frames. Special address filtering is used to receive frames addressed to that multicast
address. The address is (bytes transmitted left to right): 01-80-C2-00-00-01.
6.6.2.2
Pause Operation
When a station’s receive buffer reaches a high watermark (running out of storage space for
incoming packets), it may send its link partner a pause frame (also known as an XOFF frame).
When the station receives the pause request, it stops transmitting for a number of slot times as
specified by the time parameter in the pause frame. The time parameter is a 2 byte field.
The reaction time for a transmitting station (the congester) is 512 bit times from the end of the
received pause frame. If the congester is transmitting when the time expires, it should complete the
transmitted frame. The congester should not start transmitting a frame after the 512-bit time period.
Transmission is restarted upon expiration of the pause timer or upon reception of a pause request
with a time parameter of zero (also known as an XON frame).
6.6.2.3
Flow Control Functionality
Note: The 82557 does not support flow control functionality.
112
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Flow control is set by three configuration bits: one for transmit and two for receive flow control.
Way registers to determine whether to turn it on. Flow control cannot be activated by auto-
negotiation alone.
The 8255x (except the 82557) provides status information to software regarding flow control. If the
device is currently paused from sending frames or if it is sending a pause frame, it sets an
indication in the flow control register. The device also has statistical counters that count the number
6.6.2.3.1
Transmit Flow Control
In FC mode, the controller receives FC frames independently of the state of the receive FIFO. This
allows other stations to put pressure on the device. The received FC frame is passed to the FIFO
only if it is configured for this (according to the reject FC bit, Section 6.4.2.3, “Configure (010b)”)
and only if there is room for it. The reject FC bit causes the device to reject frames sent to the
special FC multicast address. However, if the FC frame was destined to the device individual
address, the frame is passed to the FIFO regardless of the reject FC bit.
After a pause request is received, a 16 bit count down register is loaded with the time parameter
value that was in the pause frame. The counter decrements each slot time and stops when it reaches
zero. Upon receiving another pause frame, the device reloads the counter with the new time
parameter.
If the value of the counter is not zero, transmission of future frames is inhibited. This does not
interfere with the current transmission. The currently transmitted frame is completed and further
transmits are paused. This does not prevent the device from transmitting FC frames itself if it is
required to.
6.6.2.3.2
Receive Flow Control
In FC mode, the controller detects potential overrun conditions and sends a pause frame to the
other node. The transmission of the pause frame occurs after the current transmit is completed
(there is no immediate termination of transmissions).
The time parameter for the pause frame assembly is based on a pre-configured value that comes
from the configure command.
The prediction of overruns occurs through detection of a threshold crossing in the receive FIFO, as
threshold value, a pause command is sent (with the pre-configured time value). Once a predicted
overrun is detected, the device remains in the congested state until the receive FIFO is absolutely
empty (no bytes in the receive FIFO).
The device has two additional configurable modes that affect transmit flow control. These modes
are “ReStop” and “ReStart” and are detailed below:
• ReStop. The controller has sent a pause command. Just before pause time (sent by the device)
expires, if the receive FIFO is not empty, the device sends another pause command. The
additional pause is sent so that it is assured to reach the congester before the pause time
expires.
• ReStart. The controller has sent a pause command. When the receive FIFO is empty, an XON
(PAUSE(0)) command is sent. The same mechanism used for sending pause commands is used
for sending pause (time equals 0) commands. In this mode, the device sends a pause command
and waits for the receive FIFO to empty. When it is empty, the device sends an XON command
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
113
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
immediately. If the pause time (from the device) has expired, the controller starts monitoring
the FIFO again to determine if a new pause command should be sent.
The two modes, ReStop and ReStart, can be used together. The controller sends a pause command.
If the receive FIFO is empty, the device sends an XON command. If the time-out interval is
exceeded, it sends another pause command.
In the ReStop mode, after the 8255x (excluding the 82557) transmits a pause command, the device
may be required to extend the pause time by sending subsequent pause frames. The device is
required to send subsequent pause commands far enough in advance to insure that it reaches the
congester before its own timer expires. To achieve this, the device needs to attempt to transmit the
pause command more than 27 slot times before the timer expires. These 27 slot times account for
the time period required to transmit the pause command (1 slot time), the propagation delay (1 slot
time), the congester’s reaction time (1 slot time), and a possible delay in the transmission of the
pause command due to the transmission of a data frame (up to 24 slot times). To guarantee this time
delta, the device driver needs to set the FC delay configuration value with its least significant 5 bits
exceeding 27. The controller ignores these 5 bits in its internal counting. In order to guarantee that
the time delta is sufficient to prevent transmission from the congester, it is recommended to set the
least significant bits to 1Fh (31 decimal).
If the pre-configured value is less than 32, then:
• In ReStop mode (or both modes active), the device sends pause commands continuously until
its receive FIFO is empty.
• In ReStart mode, the device sends one pause command without any following XON
commands.
Two examples are presented below:
Example 3. Configured “FC Delay” (011Fh)
— Send a pause command with 011Fh (287 decimal) slot time parameter.
— Load counter with 100h (256 decimal) value.
— ReStop Mode: After 256 slot times, if the FIFO is not empty, send another pause
command.
— ReStart Mode: When the FIFO is empty (presumably before 256 slot times), send an XON
command.
Example 4. Configured “FC Delay” (0017h)
— Send a PAUSE command with 17h (23 decimal) slot time parameter.
— Do not load the counter.
— ReStop Mode (not recommended): If the FIFO is not empty, send another pause command
immediately (this will saturate the link).
— ReStart Mode: Nothing further should be done.
For convenience, flow control related configuration bits (Section 6.4.2.3, “Configure (010b)”
contains more detail) are described in the table below:
114
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
Table 58. Flow Control Configuration Bits
Configuration Byte
Name
Description
IEEE frame based transmit flow control.
Default
Map Location
Transmit FC Byte 19, bit 2
0 (on)
Receive FC
Byte 19, bit 4
ReStart
IEEE frame based receive flow control - ReStart mode. 0 (off)
IEEE frame based receive flow control - ReStop mode. 0 (off)
Receive FC
Byte 19, bit 3
ReStop
When this bit is set, FC frames will not be passed to
0 (off)
Reject FC
FC Delay
Byte 19, bit 5
the receive FIFO like regular frames.
This is the slot time delay number used for the time
Byte 16, bits 7:0 (LSB)
Byte 17, bits 7:0 (MSB)
parameter in the assembly of pause frames (for
pausing the other node transmissions).
4000h
6.6.3
Priority Aware Frame Based Flow Control
The 82558 and later generation controllers have the ability to respond to priority aware frame
based flow control frames. Their operation relates to multiple queues.
6.6.3.1
Priority Flow Control Operation
The 82558 and later generation controllers can receive two types of flow control frames:
• Pause. The normal IEEE pause frames that stop all transmission (as discussed above).
• Pause Low. This is a new low priority pause frame that stops only the low priority queue
(LPQ).
When the device receives a pause frame, it stops all transmission at the CSMA level as defined in
IEEE draft standard 802.3x.
When the controller receives a pause low frame, it stops transmitting the LPQ at the micromachine
level. When a pause low flow control frame is received, the device continues to transmit all frames
in its transmit FIFO. When the device is used with priority aware FC it is recommended that the
two frames in FIFO configuration bit is set so that the number of frames is no more than two.
Note: The 82558 and 82559 do not transmit pause low frames.
When the device receives a flow control frame (either a pause frame or a pause low frame), this
frame overrides any flow control frame previously received. If the device was paused due to a
pause frame and it receives a pause low frame, it starts transmitting high priority frames (and any
low priority packets that were in the transmit FIFO). If the device was paused low due to a pause
low frame and it receives a pause frame, it stops transmitting high priority frames immediately.
The pause and pause low frames do not affect the device’s CU state (as reflected in the CUS field
in the SCB).
6.6.3.2
Flow Control Frame Format
The pause and pause low flow control frames share the same frame format. Table 57 on page 112
illustrates a flow control frame.
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
115
Download from Www.Somanuals.com. All Manuals Search And Download.
Host Software Interface
The priority field is the field that differentiates between pause and pause low frames. Only the three
least significant bits in this byte are considered. These three bits are compared to the FC priority
threshold configuration field:
• If the received value is equal to or less than the configured value, the frame is a pause frame.
• If the received value is greater than the configured value, the frame is a pause low frame.
This scheme enables the link partner to support up to 8 different priority queues. If a queue is full,
the link partner should send a FC frame with the number of the queue in the priority field. The
device is guaranteed to stop transmitting to that queue and lower priority queues.
Some examples are listed below:
• Assume the configured FC priority threshold value is 4 (100b).
Any FC frame with the values 000b through 100b in the priority field will cause the device to
pause (stop all transmissions). Any FC frame with the values 101b through 111b in the priority
field will cause the device to pause low (stop only transmissions in the LPQ).
• Assume the configured FC priority threshold value is 0 (000b).
Any FC frame with the values 000b in the Priority field will cause the device to pause (stop all
transmissions). Any FC frame with the values 001b through 111b in the Priority field will
cause the device to pause low (stop only transmissions in the LPQ).
• Assume the configured FC priority threshold value is 7 (111b).
Any FC frame with any value in the priority field will cause the device to pause (stop all
transmissions). This is the IEEE compliant mode.
6.6.4
Half Duplex Flow Control
The 82558 and 82559 support frame based flow control frames in both full duplex and half duplex
switched environments. It is not intended for flow control to be enabled in a shared media
environment. A flow control frame may encounter a collision in half duplex and will be
retransmitted after the backoff time. Wait After Win (described in Section 6.7, “Collision Backoff
Modification in Switched Environments”) guarantees that the flow control frame will be
transmitted within a bounded delay.
6.7
Collision Backoff Modification in Switched
Environments
The 82558 and 82559 support a modification of the CSMA/CD backoff algorithm named “Wait
After Win” (WAW). When WAW is enabled, the device extends the interframe spacing gap to one
slot time after the successful retransmission of a frame that previously encountered a collision.
WAW is activated by a separate WAW configuration bit.
The WAW feature, combined with the “Half Slot Time BackOff” feature in the switch, guarantees
that any collision that occurring on a dedicated half duplex link will be resolved after the first
collision.
116
Intel 8255x 10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Physical Layer Interface
7
Intel Fast Ethernet adapters all have a physical layer (PHY) component that interfaces the network
adapter to the wire. The MAC component of the adapter interfaces to the PHY component via the
IEEE Media Independent Interface (MII). Sometimes it is necessary for software to access these
PHY registers to properly configure the PHY. The Management Data Interface (MDI) in the 82557,
82558 and 82559 allows communication across the MII bus to registers on these PHYs.
For 100Mbps applications, the 82557 contains an IEEE MII compliant interface to the Intel 82555
physical layer device (or other MII compliant PHYs) which allow connections to both 10Mbps and
100Mbps networks. The 82558 and 82559 contain an embedded 82555 module. Software still
communicates with the embedded 82558 through the MDI port. The MDI register sets for several
Physical Layer Interface.
PRO/100B adapters use both Intel (and third-party) PHYs that support 10BASE-T, 100BASE-TX,
and/or 100BASE-T4 physical layers and are capable of auto-negotiation. The PHY module on the
PRO/100B adapter is a separate discreet component from the 82557. There are different versions
and generations of the PRO/100B that use different PHYs. Because certain vendor-specific
programming hooks may be required to fully support various PHYs, software should determine at
runtime which specific PHY is on the PRO/100B adapter being driven (Section 8.1.2, “PHY
Detection and Initialization” contains more details).
PRO/100+ adapters use the 82558’s embedded 82555, which supports auto-negotiation, 10BASE-
T, and 100BASE-TX. The 82559 also contains an embedded 82555. Although the PHY is
embedded in the 82558 and 82559, software still accesses the PHY via the MDI interface in the
manner that software uses on 82557 based adapters.
This section includes information on MDI, the 82555 MDI register sets, and auto-negotiation (N-
Way) functionality. It also includes information for items specific to working with the 82555 TX
PHY as well as the 82558 and 82559 embedded PHYs.
7.1
Management Data Interface (MDI)
The 82553, 82555, and other MII compliant devices provide status and accept management
information via the Management Data Interface (MDI). This is accomplished via read and write
operations to various registers according to the IEEE 802.3u MII specification. A read or write of a
particular register is called a management frame, which is sent serially over the MDIO pin
synchronous to MDC. Read and Write cycles are from the perspective of the controller. Therefore,
the controller would always drive the Start, Opcode, PHY Address and Register Address on to the
MDIO pin. For a write, the controller would also drive the transition bits and the data. For a read,
the PHY drives the transition bits and data onto the MDIO pin. The controller should drive address
and data on the falling edge of MDC and the PHY latches that data on the rising edge of MDC. In
an application where only one PHY is present, the PHY uses a default PHY address of 00001b. The
management frame structure is as follows:
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
117
Download from Www.Somanuals.com. All Manuals Search And Download.
Physical Layer Interface
Figure 26. Management Frame Structure
Start
Opcode
PHY Address Register Address Transition
Data
READ
< 01 > < 10 >
< 01 > < 01 >
< AAAAA >
< AAAAA >
< RRRRR >
< RRRRR >
< Z0 >
< 10 >
16 Bits
16 Bits
WRITE
This structure allows a controller, or other management hardware, to query the PHY for the status
of the link or configure the PHY to one of many modes. The next section discusses the MDI
registers.
7.2
MDI Register Set
The generic MDI register set is defined as follows:
Table 59. MDI Register Set
Register Address
Register Name and Function
00000
Control Register (MDI Standard Register)
Status Register (MDI Standard Register)
PHY Identification Register (Word 1)
PHY Identification Register (Word 2)
Auto-Negotiation Advertisement Register
Auto-Negotiation Link Partner Ability Register
Auto-Negotiation Expansion Register
Reserved
00001
00010
00011
00100
00101
00110
00111-01111
®
The Intel 82555-specific (and thus 82558 and 82559 specific as well) MDI registers are listed in
the table below. (These registers also apply to the 82558 and 82559.)
Table 60. 82555 MDI Register Set
Register Address
Register Name and Function
10000
10001
10010
10011
10100
10101
10110
10111
11000
11001
11010
11011
Status and Control
Special Control
Clock Synthesis Test and Control
100BASE-TX Receive False Carrier Counter
100BASE-TX Receive Disconnect Counter
100BASE-TX Receive Error Frame Counter
Receive Symbol Error Counter
100BASE-TX Receive Premature End of Frame Error Counter
10BASET Receive End of Frame Error Counter
10BASE-T Transmit Jabber Detect Counter
Equalizer Control and Status
Special Control
118
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Physical Layer Interface
The individual registers are defined in the following subsections using the following conventions:
R: Read
W: Write
RO: Read only
SC: Self clearing
Note: The default values listed for the 82555 registers also apply to the registers in the embedded PHY
modules of the 82558 and 82559.
7.2.1
Control Register: Register 0
The Control register provides control over the PHY functions.
Bit
Name
R / W
Description
Default
The Reset bit causes the device to reset to its
default state. Configuration settings that are issued
with the Reset bit will not be accepted, and the
device is set to the default state when the reset is
completed. This bit should set by itself. For write
operations to other bits of the Control register,
software should ensure the reset bit is not set.
Resetting the PHY also re-starts Auto-Negotiation.
R/W
SC
15
Reset
0
1 = PHY reset
0 = normal operation
1 = loopback mode
0 = normal operation
14
13
Loopback
10/100
R/W
R/W
0
1
1 = 100 Mbps
0 = 10 Mbps
The Auto-Negotiation settings take precedence
over forced speed and duplex settings. In other
words, if both speed and duplex are set, Auto-
Negotiation will override these settings. To
successfully enable forced speed and duplex, Auto-
Negotiation must be disabled through this bit and
the Restart Auto Configuration bit (bit 9) must equal
0.
12
Auto-Negotiation Enable
R/W
1
1 = Enable Auto-Negotiation
0 = Disable Auto-Negotiation
1 = power down
11
10
Power Down
Isolate
R/W
R/W
0
0
0 = normal operation
1 = electrically isolate PHY from MII
0 = normal operation
The Auto-Negotiation settings take precedence
over forced speed and duplex settings. In other
words, if both speed and duplex are set, Auto-
Negotiation will override these settings. To
successfully enable forced speed and duplex, Auto-
Negotiation must be disabled through this bit and
the Auto-Negotiation Enable bit (bit 12) must equal
0.
R/W
SC
Restart Auto
Configuration
9
0
1= Restart the auto negotiation
0 = Normal operation
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
119
Download from Www.Somanuals.com. All Manuals Search And Download.
Physical Layer Interface
Bit
Name
R / W
Description
Default
1= Full duplex
0 = Half duplex
8
Full/Half
R/W
0
This bit is valid only when the device in is in
loopback mode. When it is used for transmitting
broadcast frames, it increases the collision count
statistic. After the transmission collision test is
finished, the PHY should be reset to ensure it
returns to normal mode operation. (Using this bit
outside of loopback mode is inappropriate and may
cause unpredictable behavior.)
7
Collision Test Enable
Reserved
R/W
R/W
0
0
1 = enable the collision CDT test during loopback
Written as 0, don't care on read
6:0
7.2.2
Status Register: Register 1
For maximum accuracy of link status, the Auto-Negatiation Complete bit (bit 5), should be polled
at a continuous interval of at least 300 milliseconds. After the Auto-Negotiation process has
completed, the sticky bit will be set to 1. When the Auto-Negotiation Complete bit equals 1, the
Link Status bit is updated for accuracy. Failure to delay the recommended time or using the Link
Status bit before a valid resolution of the Auto-Negotiation Complete bit can cause inaccurate
results.
Bit
Name
R / W
Description
Default
1= T4 capable
15
T4 capable
RO
0
0 = Not T4 capable
1= TX full duplex capable
14
13
12
TX full duplex capable
TX half duplex capable
RO
RO
RO
1
1
1
0 = Not TX full duplex capable
1 = TX half duplex capable
0 = Not TX half duplex capable
1 = 10BASE-T full duplex capable
10BASE-T full duplex
capable
0 = Not 10BASE-T full duplex capable
1 = 10BASE-T half duplex capable
10BASE-T half duplex
capable
11
RO
RO
RO
1
0 = Not 10BASE-T half duplex capable
10:6
5
Reserved
These bits are reserved and written as 0.
0
1 = Auto-negotiation complete
Auto-negotiation
complete
Sticky bit
0 = Auto-negotiation not complete
1 = Remote fault detected
4
3
Remote fault detect
RO
RO
0
1
0 = Remote fault not detected
1 = Auto-negotiation enabled
0 = Auto-negotiation disabled
Auto-negotiation ability
120
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Physical Layer Interface
Bit
Name
R / W
Description
Default
1 = Valid link
2
Link Status (LINKFLT)
RO
Sticky bit
0 = Invalid link
1 = Jabber condition detected
1
0
Jabber Detect (JABFLT)
Extended Capabilities
RO
RO
Sticky bit
1
0 = No jabber condition detected
NOTE: This bit is only valid in 10 Mbps mode.
1= Extended register capabilities
0 = Basic register set capabilities only
7.2.3
Identification Registers: Registers 2 and 3
The 32-bit ID register provides a mechanism for software to determine which PHY is present. The
contents of these registers differ depending on the PHY. There are three values encoded in registers
2 and 3 that uniquely identify the PHY device:
• The OUI of the PHY manufacture.
• The model number of the PHY.
• The revision number of the PHY.
The OUI is a 24-bit manufacturer identification number. Only the least significant 22 bits of the
OUI are stored in the registers. The two most significant bits, bits 23:22, are fixed at 0. Bits 21:6 of
the OUI are stored in register 2 while the remaining bits of the OUI, bits 5:0, are stored in bits
15:10 of register 3. (The Intel manufacturer OUI number is “00AA00h.”)
The model number is a 6-bit value assigned by the manufacturer identifying the PHY model. It is
stored in bits 9:4 of register 3. (The Intel PHY model number is “010101.”)
The revision number is a 4-bit value assigned by the manufacturer identifying the PHY revision
number. The Intel devices use revision numbers “0000” through “0100.” The Intel 82558 has a
revision ID number of 0000b, and the 82559, 0100b.
The map below shows how the these three numbers (OUI, model and revision numbers) are
mapped into the MDI registers.
Table 61. 24-bit OUI Identification Number
OUI
21:6
Manufacturer
OUI
Model
Revision
23:22
5:0
Intel
00AA00h
00AA00h
00b
00b
00 0000 1010 1010 00b
00 0000 1010 1010 00b
00 0000b
00 0000b
Intel 82555
010101b
0000b
The PHY identification register values are as follows:
Table 62. MDI Identification Registers 2 and 3: PHY ID Encoding
Intel 82555
(Revision 4)
Intel 82555
(Revision 1)
Intel 82555
(Revision 0)
MDI Register
Register 2
Register 3
02A8h
0154h
02A8h
0151h
02A8h
0150h
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
121
Download from Www.Somanuals.com. All Manuals Search And Download.
Physical Layer Interface
7.2.4
Auto-Negotiation Advertisement Register: Register 4
This register contains the advertisement ability of the PHY. It is used by software to determine the
highest common denominator technology after the auto-negotiation process has finished. Any
changes to this register prior to auto-negotiation must be followed by setting the Renegotiate bit in
the Command register.
Bit
Name
R / W
Description
Default
Next page is not supported and this bit should be
set to 0b.
15
Next Page
R/W
0
This bit position in the transmitted code word is
used to indicate an acknowledge. It is set only by
the auto-negotiation logic after receiving three
consecutive and matching code words from the link
partner.
14
Acknowledge
RO
0
This bit indicates remote fault in the local station. It
may be set by management to indicate the remote
fault condition of the partner.
13
Remote Fault
R/W
0
12:11 Reserved
RO
RO
These bits are reserved.
000
82558: 0
82559: 1
This bit indicates that Pause is supported by the
local node.
10
Pause
This field indicates that 100BASE-T4 is supported
by the local node.
9
100BASE-T4a
R/W
R/W
R/W
R/W
R/W
R/W
0
100BASE-TX Full
Duplexa
This indicates that 100BASE-TX full duplex is
supported by the local node.
8
1
This bit indicates that 100BASE-TX is supported by
the local node.
7
100BASE-TXa
10BASE-T Full Duplexa
10BASE-Ta
1
This bit indicates that 10BASE-T full duplex is
supported by the local node.
6
1
This bit indicates that 10BASE-T is supported by
the local node.
5
1
This field identifies the protocol supported. IEEE
802.3 is indicated by a selector value 00001b.
4:0
Selector Field
00001
a.
During normal operation, the driver (management agent) does not need to change this register value. If a certain ability is not
advertised, the respective bit in the Technology Ability field must be cleared. A bit that is not supported by the device must
not be set within this field. Otherwise, the auto-negotiation protocol will be violated.
7.2.5
Auto-Negotiation Link Partner Ability Register: Register 5
This register holds the link code word captured from the link partner’s PHY (in other words, the
device at the other end of the link segment). Its value is valid only when the Auto-Negotiation
Complete bit is set in the Status register.
Bit
Name
R / W
Description
Default
15
Next Page
RO
Next page is not supported.
0
This bit indicates if a device has successfully
received the link code word from its link partner.
14
13
Acknowledge
Remote Fault
RO
RO
0
0
This bit identifies if the remote fault bit is set in the
received code word.
122
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Physical Layer Interface
Bit
Name
R / W
Description
Default
12:11 Technology Ability Field
RO
These bits are reserved.
000
This bit indicates if Pause is supported by the link
partner.
10
9
Pause
RO
RO
RO
RO
RO
RO
RO
0
This bit indicates if 100BASE-T4 is supported by
the link partner.
100BASE-T4a
0
100BASE-TX Full
Duplexa
This bit indicates if 100Base-TX full duplex is
supported by the local node.
8
0
This bit indicates if 100BASE-TX is supported by
the link partner
7
100BASE-TXa
10BASE-T Full Duplexa
10BASE-Ta
0
This bit indicates it 10BASE-T full duplex is
supported by the link partner.
6
0
This bit indicates if 10BASE-T is supported by the
link partner.
5
0
This field contains the Selector value from the link
partner.
4:0
Selector Field
00000
a.
During normal operation, the driver (management agent) does not need to change this register value. If a certain ability is not
advertised, the respective bit in the Technology Ability field must be cleared. A bit that is not supported by the device must
not be set within this field. Otherwise, the auto-negotiation protocol will be violated.
NOTE: The Auto-Negotiation Link Partner Ability Register is read only.
7.2.6
Auto-Negotiation Expansion Register: Register 6
Register 6 contains supplemental information used by the auto-negotiation process.
Bit
Name
R / W
Description
Default
15:5
Reserved
RO
These bits are reserved and should be set to 0.
0
1 = More than one of the 10BASE-T, 100BASE-TX,
or 100BASE-T4 PMAs detects a valid link.
4
Parallel Detection Fault
RO
0
0 = One or zero of the 10BASE-T, 100BASE-TX, or
100BASE-T4 PMAs detects a valid link.
1 = Link partner supports Next Page.
Link Partner Next Page
Able
3
2
RO
RO
0
0
0 = Link partner does not support Next Page.
1 = Local device supports Next Page.
Next Page Able
Page Received
0 = Local device does not support Next Page.
1 = Three identical and consecutive link code words
have been received.
1
0
RO
RO
0
0
0 = Three identical and consecutive link code words
have not been received.
1 = Link partner supports auto-negotiation.
Link Partner Auto-
Negotiation Able
0 = Link partner does not support auto-negotiation.
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
123
Download from Www.Somanuals.com. All Manuals Search And Download.
Physical Layer Interface
7.3
Intel 82555 Specific Registers
Note: The Intel MAC/PHY silicon devices (82558, 82559, 82550, and 82551) use the 82555 as the base
for their integrated PHY units. Therefore, the information contained in this section and the
following subsections apply to the all 8255x Fast Ethernet controllers except the 82557.
7.3.1
Status and Control Register: Register 16
Bit
Name
R / W
Description
Default
1 = Flow control enabled.
0 = Flow control disabled.
NOTE: This bit should always equal 0 for the
82559.
15
Flow Control
RW
0
This bit enable T4 when auto-negotiation is
disabled.
14
13
T4 Enable
RW
RW
0
1 = Enable T4 technology
0 = Disable T4 technology
This bit controls the RX100 CRS disconnect
function in repeater mode.
NOTE: This bit should always equal 0 for the
82559.
0 = DTE
1 = Rptr
CRS Disconnect Control
Reserved
12
11
This bit is reserved.
0
0
RCV De-Serializer In-
Sync Indication
This bit is used as the 100BASE-TX RCV De-
Serializer In Sync Indication
RO
RO
This bit provides 100BASE-T Power Down
Indication.
10
9
100 Power Down
10 Power Down
0
0
1 = Power down
0 = Normal operation
This bit provides 10BASE-T Power Down
Indication.
RO
RO
1 = Power down
0 = Normal operation
This bit indicates 10BASE-T polarity.
1 = Reverse polarity
8
Polarity
Reserved
T4
0
0
0
0 = Normal polarity
7:3
2
These bits are reserved.
This bit is a result of the auto-negotiation process.
1 = 100BASE-T4.
RO
RO
RO
0 = No 100BASE-T4
This bit is a result of the auto-negotiation process.
1
0
Speed
Duplex
1 = 100 Mbps
0 = 10 Mbps
0
0
This bit is a result of the auto-negotiation process.
1 = Full duplex
0 = Half duplex
124
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Physical Layer Interface
7.3.2
Special Control Register: Register 17
Bit
Name
R / W
Description
Default
1 = Bypass Scrambler
0 = Normal operation
15
Scrambler Bypass
RW
0
1 = Bypass 4-bit to 5-bit
0 = Normal operation
14
13
12
4/5 Bypass
RW
0
0
0
1 = Force H pattern
0 = Normal operation
Force Transmit H Pattern RW
Force 34 Transmit Pattern RW
1 = Force 34 pattern
0 = Normal operation
1 = 100BASE-TX good link indication, forcing to
ASD output
11
Good Link
RW
0
0 = Normal operation
1 = Tri-state MDI interface
0 = Normal operation
9
8
MDI Tristate
RW
RW
RW
0
0
1 = Disable Dynamic Power Down.
0 = Normal operation
Dynamic Power Down
Disable
1 = Auto-negotiation loopback
Auto-Negotiation
Loopback
7
6
5
0
0
0
0= Auto negotiation normal Mode.
Reserved
This bit is reserved.
1 = Bypass filter
Filter Bypass
RW
RW
0 = Normal operation
1 = Disable auto-polarity
0 = Normal operation
4
Auto-Polarity Disable
0
1 = Disable 10BASE-T squelch test
0 = Normal squelch operation
3
2
1
Squelch Test Disable
Extended Squelch
Link Integrity Disable
RW
RW
RW
0
0
0
This bit indicates extended squelch control
1 = Disable link integrity operation
0 = Normal link integrity operation
1 = Disable jabber function
0 = Normal jabber operation.
0
Jabber Function disable
RW
0
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
125
Download from Www.Somanuals.com. All Manuals Search And Download.
Physical Layer Interface
7.3.3
Clock Synthesis Test and Control Register: Register 18
Bit
Name
R / W
Description
Default
Clock Synthesizer Shift command. One shot
signaling YS10ACLK domain.
RW
SC
15
Clock Timing
Can be active only when bit 14 is ‘0
Clock Synthesizer load command. One shot
signaling YS10ACLK domain.
RW
SC
14
Clock Timing
Can be active only when bit 15 is ‘0
Break Down Timer
Enable
Logic 1 enables manipulate Break Down counter
with phya1, phya4 and test high.
13
12
RW
Equalizer Probe Mode
Enable
Logic 1 enables the Equalizer output through the
Speed LED.
0
0
1 = Enable 10BASE-T dig outputs through the
LEDs
NOTE: This function is only present on the 82559.
10BASE-T Probe Mode
Enable
11
RW
RO
10:8
4:0
Reserved
These bits are reserved.
0
PHY Address
This field contains the PHY address.
00001
7.3.4
7.3.5
100BASE-TX Receive False Carrier Counter: Register 19
Bit
Name
R / W
Description
Default
This register contains a 16-bit counter for false
carrier events. A false carrier event occurs when a
frame that does not start with “JK” is detected.
When the counter is full, additional false carrier
events are not counted. This counter is self-clearing
on read.
RO
SC
15:0
Receive False Carrier
0
100Base-TX Receive Disconnect Counter: Register 20
Bit
Name
R / W
Description
Default
This register contains a 16-bit counter for
disconnect events. The counter is incremented for
each frame detected in repeater mode that does
not start with a “JK.” When the counter is full,
additional disconnect events are not counted. This
counter is self-clearing on read.
RO
SC
15:0
Disconnect Event
0
126
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Physical Layer Interface
7.3.6
100BASE-TX Receive Error Frame Counter: Register 21
Bit
Name
R / W
Description
Default
This register contains a 16-bit counter for receive
error frames. It is incremented for frames with a
receive error condition (frames containing a symbol
error or frames with a premature end of frame).
When the counter is full, additional error frames are
not counted. This counter is self-clearing on read.
RO
SC
15:0
Receive Error Frame
0
7.3.7
7.3.8
7.3.9
7.3.10
Receive Symbol Error Counter: Register 22
Bit
Name
R / W
Description
Default
This register contains a 16-bit counter and
increments for each symbol error. The counter stop
counting additional symbol errors when it is full.
This counter is self-clearing on read.
RO
SC
15:0
Symbol Error
0
100BASE-TX Receive EOF Error Counter: Register 23
Bit
Name
R / W
Description
Default
This register contains a 16-bit counter and
RO
SC
increments for each premature end of frame event.
It stops counting additional premature end of frame
events when it is full. It is self-clearing on read.
15:0
Premature End of Frame
0
10BASE-T Receive EOF Error Counter: Register 24
Bit
Name
R / W
Description
Default
This register is a 16-bit counter that increments for
each end of frame error event. The counter stops
counting additional errors when it is full. It is self-
clearing on read.
RO
SC
15:0
End of Frame
0
10BASE-T Transmit Jabber Detect Counter: Register 25
Bit
Name
R / W
Description
Default
This register is a 16-bit counter that increments for
each jabber detection event. The counter stops
counting additional events when it is full. It is self-
clearing on read.
RO
SC
15:0
Jabber Detect
0
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
127
Download from Www.Somanuals.com. All Manuals Search And Download.
Physical Layer Interface
7.3.11
Equalizer Control and Status Register: Register 26
This register is used to control and monitor the operation of the 8255x PHY module equalizer
(excluding the 82557 since it does not have an integrated PHY unit). Bits 15:13 specify the
command, and bits 12:0 contain the data field for the command.
Bit
Name
R / W
Description
Default
Equalizer Control and
Status
Bits 15:13 contain the opcode command while bits
12:0 hold the command data.
15:0
RW
Opcode Command (bits 15:13)
Command Data (bits 12:0)
000 NOP
[12] Set zero command. Set value of bits 3:0.
[11:10] FSM high threshold transitions:
00 = FM: 2.19 ms; SM: 2.01 ms (0.5 ms - 2.03 ms)
01 = 2.19 ms
10 = FM:2.19; SM:2.03 (0.5 ms - disabled)
11 = Disabled
[9:8] FSM low threshold transitions:
00 = FM: 1.83 ms; SM: 1.99 ms (0.5 ms - 1.97 ms)
01 = 1.83 ms
001 Write to ASD configuration register
0
10 = FM: 1.83 ms; SM: 1.97 ms (0.5 ms - disabled)
11 = 1 ms
[7] Signal squelch force enable.
[6] Squelch signal forcing value.
[5] Reserved.
[4] Enable/disable zero forcing.
[3:0] Coded zero 0 through 15.
[12:11] Reserved.
[10:9] TMD100 transition ration bits/LPF ratio:
00 = 0.5 / 6
01 = 0.5 / 5
10 = 0.25 / 6
11 = 0 / 6
[8:7] Signal detect 5-bit counter setting value:
00 = 10h
010 Write to ASD configuration register
1
01 = 18h
10 = 1Ch
11 = 1Fh
[6] Set signal detect counter command.
[5] Reserved.
[4] Disable lock adaptation mechanism.
[3:1] Reserved.
[0] Force test mode and activate LFSR register.
128
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Physical Layer Interface
Opcode Command (bits 15:13)
Command Data (bits 12:0)
[9] Breakdown ASD counters.
[8] Selects signal detections or transitions.
[7:6] Slow mode adaptation time configure:
00 = 67 ms (default)
01 = 0.5 ms
10 = 16.8 ms
011 Write to ASD configuration register
2
11 = 134 ms
[5] Signal detect force enable.
[4] Signal detect force value.
[3:2] Reserved.
[1] Disable signal detect high threshold value.
[0] Disable signal detect low threshold value.
[15:14] Reserved.
[13:11] ASD command/address register.
[10:9] TMD100 transition ratio bits.
[8:7] ASD state machine state.
[6] Lock adaptation signal.
[5] Signal detect.
100 Read status register
[4] Squelch signal.
[3:0] Coded zero.
Read cycle:
[15:0] Reflects the jitter register bits.
Write cycle:
101 Read jitter register
[1:0] Selects the register lines reflected by read.
00 = Bits 15:0
01 = Bits 21:6
1x = Bits 26:11
Read cycle:
[15:0] Clock counter value.
Write cycle:
110 Read clock register
111 Reserved
[0] Selects the window reflected by read.
0 = 15:0
1 = 23:8
7.3.12
Special Control Register: Register 27
Bit
Name
R / W
Description
Default
Bits 15:3 are reserved, and bits 2:0 are used for the
LED switch control.
15:0
Special Control Register
RW
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
129
Download from Www.Somanuals.com. All Manuals Search And Download.
Physical Layer Interface
Table 63. LED Switch Control
Bits
ACTLED# Pin
2:0
LILED# Pin
000
001
010
011
100
101
110
111
Activity
Speed
Speed
Activity
Off
Link
Collision
Link
Collision
Off
Off
On
On
Off
On
On
7.4
Auto-Negotiation Functionality
The PHY units of the 8255x devices (excluding the 82557) all support auto-negotiation (N-Way).
Auto-negotiation is an automatic configuration scheme designed to manage interoperability in
heterogeneous LAN environments. It allows two stations with “N” different modes of
communication to establish one common mode of operation. Upon power-up, auto-negotiation
establishes a link using one or more devices capable of auto-negotiating. An hub that supports
auto-negotiation can detect and automatically configure its ports to take maximum advantage of
common modes of operation without user intervention or prior knowledge by either station. The
possible common modes of operation are: 100BASE-TX, 100BASE-TX full duplex, 10BASE-T,
and 10BASE-T full duplex.
7.4.1
Description
Auto-Negotiation selects the fastest operating mode (in other words, the highest common
technology denominator) available to hardware at both ends of the cable. A PHY’s capability is
encoded by bursts of link pulses called Fast Link Pulses (FLPs). Connection is established by FLP
exchange and handshake during initialization time. Once the link is established by this handshake,
the native link pulse scheme resumes (for example, 10BASE-T or 100BASE-TX). If only one end
of the twisted pair is auto-negotiation capable, then the FLP exchange fails and another scheme
called parallel detection is used to determine the link settings. A PHY reset, or management
renegotiate command (through the MDI interface), restarts the auto-negotiation process. To enable
auto-negotiation, bit 12 of the MDI Control Register must be set. If the PHY cannot perform auto-
negotiation, this bit is set to 0.
Since only one technology can be used at a time (after every renegotiate command), a prioritization
scheme must be used to ensure that the highest common denominator ability is chosen. Table 64
lists the technology ability field bit assignments. Each bit in this table is set according to the PHY
130
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Physical Layer Interface
Table 64. Technology Ability Field Bit Assignments
Bit Setting Technology
0
1
2
3
4
5
6
7
10BASE-T
10BASE-T Full Duplex
100BASE-TX
100BASE-TX Full Duplex
100BASE-T4
Reserved
Reserved
Reserved
Table 65. Technology Priority
Priority
Technology
1
2
3
4
5
100BASE-TX Full Duplex
100BASE-T4
100BASE-TX
10BASE-T Full Duplex
10BASE-T
To detect the correct technology, the two register fields are ANDed together to obtain the highest
common denominator. This value is used to map into a priority resolution table used by the MAC
driver to select the appropriate technology. The auto-negotiation process occurs in the following
steps:
1. Receive 3 consecutive, matching code words.
2. Set acknowledge bit in transmit code word.
3. Receive 3 consecutive, matching code words with acknowledge bit set.
4. Transmit 6 to 8 more code words with acknowledge bit set.
5. Use priority table to determine operating mode.
6. FLP received from link partner is recorded in MDI register.
7.4.2
Parallel Detection
The key to auto-negotiation’s interoperation with installed legacy LANs is the Parallel Detection
function. Parallel Detection can be used to determine what the line speed is if the link partner does
not support an N-Way (the FLP exchange is not supported) repeater or switch. Parallel Detection
works by passing the signals present on the receiver to the 100BASE-TX and 100BASE-T4 link
monitor functions. If one link monitor function indicates a valid link, then it connects that
technology to the media.
The 82555 PHY and the 8255x (excluding the 82557) PHY modules support N-Way and Parallel
Detection (in the event that their link partner does not respond to FLPs).
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
131
Download from Www.Somanuals.com. All Manuals Search And Download.
Physical Layer Interface
7.5
Vendor-Specific PHY Programming
®
The Intel PRO/100B adapters are designed to support Intel and third-party PHYs using TX and
T4 PHYs. The PHYs will be capable of auto-negotiation, but certain vendor specific programming
hooks may be required to fully support these PHYs. These issues are addressed in this section.
7.5.1
Intel 82555 TX PHY
The 82555 is a 100BASE-TX PHY that supports all standard MDI registers described in
Section 7.2, “MDI Register Set” of this document.
The best method for software to detect the speed and duplex of operation for the 82555, assuming
it successfully completed auto-negotiation, is by reading bits 1 and 0 of MDI register 16 (10h).
7.5.2
82558 and 82559 Embedded PHY Unit
The embedded PHY used in the 82558 and 82559 supports all standard MDI registers described in
Section 7.2, “MDI Register Set” of this document. In addition, the embedded PHY module also
supports some registers and bits specific to the 82555 which are detailed above.
The best method for software to detect the speed and duplex operation of the 82558 or 82559,
assuming it successfully completed auto-negotiation, is by reading bits 1 and 0 of MDI register 16
(10h).
132
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Physical Layer Interface
7.5.2.1
PHY Stand Alone (PHYSA) Mode
Only the 82558 supports a special mode where its PHY unit can be used with an external controller
through an MII-like interface. This mode is not fully MII compliant and should be used with care.
The PHY Stand Alone (PHYSA) mode is enabled if the PHYSA bit (EEPROM word 0Ah, bit D4)
is set in the EEPROM. If this bit is set, the 82558 regards the FLD3 pin as a PHYSA input pin.
When this pin is asserted (high), the 82558 enters PHYSA mode in which:
• The 82558 MII is enabled in PHY mode (TXD are inputs, RXD are outputs, etc.).
• The 82558 operates through this MII as if it is an 82555 in DTE mode.
• The 82558 CSMA unit is disabled (has no clock).
• The PCI unit is functional but is expected to be non-operational (isolated).
The MII-like interface is intended for use with one on-board controller and may not meet the MII
timing specification. No additional PHY device (or MII connector) should be connected to this MII
bus.
The CLK signal must be provided for at least 6000 clocks after Reset or Alternate Reset is de-
asserted. The Isolate signal is to be de-asserted and the Clock signal active for at least 4 clocks if a
Reset signal or Alternate Reset signal is to be propagated into the 82558.
The 82558 is not automatically reset upon entering PHYSA mode. In PHYSA mode the Alternate
Reset pin affects only the PCI and CSMA units. It does not propagate into the PHY unit. The PCI
and CSMA should be reset by asserting the Alternate Reset pin.
The settings used by the PHY prior to entering PHYSA mode may not be supported by the external
controller (for example, full duplex flow control).
In PHY Stand Alone mode, a typical event sequence is as follows:
• RST# or ALTRST# is asserted.
• ISOLATE# should be asserted after at least 4 CLK clocks.
• The PCI bus may be stopped 4 CLKs after ISOLATE# is asserted (excluding CLK).
• RST# or ALTRST# is de-asserted, but CLK is still running.
• The 82558 reads the EEPROM (6000 CLK clocks).
• The 82558 enters PHYSA enable state.
• PHYSA pin is asserted.
• ALTRST# should be asserted after at least 4 PHY clocks.
To exit the PHY Stand Alone mode, the following sequence of events occurs:
• PHYSA pin is de-asserted.
• The PCI bus is activated and CLK is valid.
• ISOLATE# is de-asserted.
• RST# or ALTRST# or a software reset is recommended.
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
133
Download from Www.Somanuals.com. All Manuals Search And Download.
Physical Layer Interface
134
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Programming Recommendations 8
8.1
Adapter Initialization
The initialization code can be broadly split in the following code modules:
• 8255x initialization
• PHY detection and initialization
• NOS specific initialization
The sample source code provides specific coding examples. A list of programming issues to
consider is listed in the following subsections by code module.
8.1.1
8255x Initialization
• The device is internally fully reset on power up. However, the adapter and the 8255x are not
fully reset after a warm reboot. Thus, it is recommended that the software issues a Port
Software Reset command before accessing the device.
Note:
All Port commands cause a complete internal reset requiring software to re-
initialize the device.
• The interrupt mask bit (SCB Command word, Mask bit) is set to 0 after a complete reset,
implying that interrupts are enabled. It is software responsibility to ensure that this bit is set to
disable interrupts after a Port command is issued and before polled mode commands are
executed.
• If promiscuous mode is enabled, the following Configure command bits are also affected:
— SAVE BAD FRAMES. This feature should be enabled.
— DISCARD SHORT FRAMES. This feature should be disabled.
— PADDING. This feature should be disabled.
— STRIPPING. This feature should be disabled.
• It is recommended that all non-transmit action commands are executed in polled mode.
Software should maintain a single list of transmit CBs and modify a CB to a non-transmit CB
when required. The 8255x supports dynamic CU Resume commands on non-transmit
commands; therefore, no checks are required before such a command is issued.
8.1.2
PHY Detection and Initialization
®
It is recommended that drivers support the Intel 82503 and 82555 and the National
Semiconductor* DP83840 devices as well as the integrated PHY units of the 82558, 82559, 82550,
and 82551. In addition, a vendor specific routine should be executed immediately after PHY
detection to initialize specific registers in other third party PHYs (for example, the DP83840).
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
135
Download from Www.Somanuals.com. All Manuals Search And Download.
Programming Recommendations
8.1.3
NOS Specific Initialization
Software should be written so that NOS specific interface routines call lower level driver routines.
This will enable code re-use.
8.2
Transmit Processing
Frame transmission is the most critical part of driver software, especially in the server
environment. The 8255x supports dynamic chaining that allows frames to be chained to the active
transmit queue, even as the device processes frames in the queue. A simplified pseudo code flow
follows:
Get free TxCB
Set the S bit in the TxCB;
Set up other fields in the TxCB;
Reset S bit in the previous TxCB;
Wait for previous command to be accepted by the 8255x (SCB command = 0);
CU_RESUME
For dynamic chaining:
1. TxCBs should be initialized in a static circularly linked list at initialization time. Subsequently,
software should not change these links. The 8255x remembers the next CB link if it suspends.
When a new CU Resume is issued, it follows the link to the next CB without accessing host
memory to read the next link.
2. The device can be configured to generate interrupts in one of 2 ways:
a. If the device is configured to generate CNA interrupts, an interrupt is generated when the
CU becomes inactive as a result of entering the suspended state. If a CU Resume
command is issued, when the CU actively transmits the current last TxCB the 8255x re-
reads the current last TxCB S bit. If the S bit is still set, it ignores the CU Resume
command. If the bit is reset, it proceeds to execute the next TxCB.
b. If interrupts are not desired each time the CU suspends, the 8255x can be configured for
CI interrupts. In this case, an interrupt is generated only when the CU encounters a TxCB
with the I bit set.
8.3
Frame Reception
The 8255x supports early receive interrupts only in the simplified memory model. It should be
noted that:
• The 8255x generates an interrupt when a frame is received from the wire. If early receive
interrupts are enabled, the first interrupt is generated after HDS bytes are deposited in host
memory and the second (final) interrupt is generated when the frame is fully received in
memory. Software cannot selectively disable these interrupts.
• An RNR interrupt is generated on the first frame that the device fails to transfer to host
memory due to a condition of unavailable resources.
136
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Programming Recommendations
8.4
Interrupt Processing
The 8255x supports latched level triggered interrupts. Interrupts can be shared in the system if the
software and NOS support this mechanism. The SCB Command and Status words provide the
necessary interface for interrupt management. The Mask bit in the SCB Command word should be
set to 1 to disable interrupts at the adapter. Writing this bit to 0 re-enables adapter interrupts. The
device supports interrupts from multiple sources as defined by the interrupt status bits in the SCB.
If more than one interrupt source is active, the device interrupt will stay asserted until all bits are
acknowledged.
Another useful interrupt source is the Software Interrupt (SWI) bit. Setting this bit in the SCB
Command word generates an interrupt. Setting the corresponding bit in the SCB Status word de-
asserts the interrupt line. Software that need to generate interrupt requests can use this feature.
However, if the Mask bit and SWI bit are set at the same time, the Mask bit takes precedence. The
SWI bit will cause an interrupt request to wait. As soon as the Mask bit is de-asserted, the interrupt
will propagate to the CPU.
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
137
Download from Www.Somanuals.com. All Manuals Search And Download.
Programming Recommendations
138
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Wake-up Functionality
A
Note: This appendix applies only to the 82558 and subsequent devices.
Wake-up functionality was first introduced with the 82558 A-step. This component is capable of
being brought out of a power managed stated by programming it to wake on reception of a packet
addressed to it, a multicast or broadcast packet, any received packet, or a Magic Packet*. These
wake-up capabilities are enabled through the Configuration Command bits.
This functionality was carried forward to the 82558 B-step and all subsequent Fast Ethernet
devices. However, the wake on reception of a broadcast packet was changed to wake on reception
of broadcasted ARP or VLAN ARP. The Configuration Command bits used to enable these wake-
up capabilities were also modified from the 82558 A-step. In addition, the 82558 B-step and all
subsequent silicon also support the ability to wake on link status change.
If the device is enabled, it notifies the system of a wake-up packet event by asserting the PME# pin.
The PME enable bit is located in the PMCSR. It is system responsibility to ensure that this bit is set
only when the device is in the D1, D2, or D3 states (as required by the PCI Power Management
Specification). After a wake-up event, the PME status bits in the PMCSR and CSR are set
regardless of the value of the PME enable bit.
There are two types of wake-up events: reception of a wake-up packet and a link status event. The
detection mechanism for wake-up packets (packet filtering) is further categorized by fixed packet
filtering implemented by the hardware in the CSMA unit and flexible packet filtering implemented
by the loadable microcode.
A.1
Wake-up Capability
The device uses three different mechanisms to wake the system up:
1. Fixed packet wake up. This is implemented in hardware using the clock driven on the X1 pin.
2. Flexible packet filtering. This is implemented in the loadable microcode using the PCI CLK
pin as clock. (The discussion of microcode is beyond the scope of this document.)
3. Link status change event.
The device is able to detect fixed packet and link status wake-up events to the system if power is
supplied to the device. This may be the main PC power supply or an auxiliary power supply that is
active whenever the system is plugged in.
The ability to use loadable microcode for flexible frame filtering requires a valid clock on the PCI
CLK pin. This may be the CLK signal from the PCI bus or other valid clock signal connected to the
device clock pin while the system is in low power mode.
The following table describes wake-up capabilities in different combinations of power supply and
clock:
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
139
Download from Www.Somanuals.com. All Manuals Search And Download.
Wake-up Functionality
External
Clock
Circuit
Auxiliary
Power
Fixed Filtering
Wake Upa
Flexible Filtering
Wake Up
Link Status
Change Wake Up
PME Support
in PMC
No
No
No
Yes
No
D1, D2, D3h
D1
D1, D2, D3h
01111b
11111b
01111b
11111b
D1, D2, D3h, D3c
D1, D2, D3h
D1
D1, D2, D3h, D3c
D1, D2, D3h
Yes
Yes
D1,D2,D3h
D1, D2, D3h, D3c
Yes
D1,D2, D3h, D3c
D1,D2, D3h, D3c
a.
If the PME enable bit is set in the D0 state, the device will assert PME# for every wake-up event. It is BIOS
responsibility to disable PME# in D0.
A.2
Low Power Modes
The device wake-up capabilities require internal PHY and CSMA blocks to be fully active. When
the controller is set into the D2 or D3 power state and wake up is disabled, the internal PHY and
CSMA units are set into low power modes. In this mode, the device loses the link connection. In
D1, the PHY and CSMA units are always active.
Note: Placing the device into low power mode (D2 or D3 power state) with the PME enable bit cleared
immediately turns off the internal PHY unit. Setting the PME enable bit while the controller is in a
low power state will power up the internal PHY and establish link connection. If the wake up on
link status change feature is enabled, the 82558 (or later generation controller) will wake up the
system. To prevent the above scenario, the PME enable bit should be active before the device is set
into low power mode.
A.3
Power Management Context After Reset
The controller has two hardware reset signals that are both active low:
• RST#. This is the PCI bus reset, which is active on power up of the PCI power source.
• ALTRST#. This is the power on reset (POR) and is active on power up of the auxiliary power
source. It is connected to the system auxiliary power good signal.
The PCI Power Management Specification requires that the PME status and PME enable bits in the
PMCSR will not be cleared by PCI RST# if the function supports wake up from the D3
state. If
cold
the device samples link status (LISTAT) at 0 after a hardware reset, it assumes that an auxiliary
power source is used and wake up from D3
clear these bits.
is possible. In this case, assertion of RST# does not
cold
In order to support Wake on LAN (WOL) mode (preboot wake up), the controller should be able to
generate a wake up event without any software configuration, including PCI power management
configuration. In this scenario, the default configuration of the power management bits should
enable wake up after a hardware reset. The auxiliary power support indication is used to set these
bits after reset.
140
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Wake-up Functionality
A.3.1
A.3.2
Auxiliary Power Support
The LISTAT signal should be 0 after a hardware reset. For WOL mode, the default value after
power up reset (ALTRST# is asserted) of the PME enable and status bits are:
PME_Enable = 1 (wake up is enabled)
PME_status = 0 (no wake-up event)
Since the OS/BIOS expect unknown values in these bits after power up, they will clear both bits
using PCI configuration command. If WOL mode is enabled, the device will wake up the system
when it receives a Magic Packet* after power up without any need for software interaction.
Auxiliary Power Non-support
If the system does not support and auxiliary power source, LISTAT is floating after a hardware
reset. In this case wake up from D3
is not supported. Both the PME enable and status bits
cold
should equal 0 after any hardware reset. To enable the WOL functionality, the PME enable bit is
bypassed if WOL mode is enabled and auxiliary power indication is not active (same as 82558A).
The table below summarizes the PME enable and status bits default values and properties in
different configurations.
Auxiliary Power
(indicated by
LISTAT)
PME Enable
Default
(ALTRST#)
PME Status
Default
(ALTRST#)
PME Enable
bypassed on
wake-up event?
WOL bit
(EEPROM)
RST# clears
bits to 0?
No
No
0
1
X
0
0
1
0
0
0
Yes
Yes
No
No
Yes
No
Yes
NOTE: When the WOL bit in the EEPROM is set and auxiliary power is not supported, the PME enable bit is
always ignored regardless of the wake-up event type.
A.4
Fixed Packet Filtering
Fixed packet filtering includes the wake up capabilities implemented in the hardware with a pre-
defined fixed pattern. It includes Address Match packets, Magic Packets, and ARP packets. The
device filters all incoming frames, monitoring for one of the packet types mentioned, regardless of
the power state. Each fixed filter can be enabled using the configuration command. When the
corresponding filter bit is enabled, the device asserts PME# for frames passing the filter with a
correct CRC value. This filtering mechanism is active when power is supplied to the device with or
without an active clock on the PCI bus.
Any packet that passes address filtering is transferred to the internal micromachine. The
micromachine may store or further analyze the packet if the following conditions are met:
• Loadable microcode for analyzing and storing the packet in the micromachine resources is
active.
• The clock signal is active on the PCI CLK pin. This may be the PCI bus clock or any other
clock generated by an external clock switching circuitry.
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
141
Download from Www.Somanuals.com. All Manuals Search And Download.
Wake-up Functionality
A.4.1
A.4.2
Magic Packet*
The 82558 and later generation controllers (except the 82559ER) are capable of generating a wake-
up event upon reception of a Magic Packet. This feature is enabled by setting a bit in the
Configuration command.
Address Matching
The controller may be configured to wake up the system on any packet that passes the Individual
Address filtering or the multicast address filtering. The wake up enable control is done by the
individual address match wake enable bit and the multicast match wake enable bit of the
configuration structure.
Note: The 82558 B-step has the following modifications:
• Address match wake up depends on two configuration bits:
— Bit 0 in Byte 19 of the configuration structure changed its role from Address Wake Enable
to IA Match Wake Enable.
— Bit 7 in Byte 9 of the configuration structure is MC Match Wake Enable.
• Regardless of the power management event configuration, packets that pass address filtering
are transferred to the micromachine for either storage or further processing if the clock signal
is active on the PCI CLK pin.
• The 82558 B-step will not wake up the system on a broadcast address match (except for ARP
frames and frames that match the flexible filtering definition).
A.4.3
ARP Packet Filtering
Address Resolution Protocol (ARP) is used for MAC address resolution of a machine. This
protocol generally precedes any IP transaction. The controller can wake up the system when an
ARP frame is received if configured by the ARP Wake Enable (ARP_Wake_En) configuration bit.
Note: The broadcast disable configuration bit should equal 0 (broadcast enabled).
The 82558 and 82559 devices are capable of handling 802.1q VLAN headers if they are configured
by the VLAN ARP configuration bit. If this bit is set to 1, the devices can handle ARP frames with
or without a 4-byte VLAN header. If it is set to 0 (default), only ARP frames without the VLAN
header pass the filter. Multiple IP addresses are not supported by the fixed ARP filter. However,
they can be handled through flexible filtering.
The controller needs to be configured with the 16 least significant bits of the IP address
(IP_Address configuration word) if ARP filtering is enabled.
The ARP frame format is illustrated below.
Offset
0-5
Hexadecimal Pattern
Description
Broadcast destination address
FF, ... FF
142
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Wake-up Functionality
Offset
Hexadecimal Pattern
0806
Description
12
21
38
Protocol type (0806 = ARP)
ARP opcode (01 = request)
01
X1, ... X4
A sequence of 4 bytes, which is the requested IP address
Ethernet Type II frame
Ethernet Type II frame with VLAN type
msb
lsb
FF
msb
FF
lsb
FF
Destination
Address
Destination
Address
FF
FF
08
FF
FF
FF
FF
FF
FF
FF
Source
Address
Source
Address
Frame
Type
VLAN
Header
06
01
12
14
Frame
Type
Data
06
01
08
18
22
26
30
34
38
42
46
Data
18
22
26
30
34
38
42
46
IP address x4
x3
x2
x1
IP address x4
x3
x2
x1
The controller only filters the shaded fields in the frame format above. Only the two low bytes of
the IP address are compared. The controller does not check for VLAN type (any frame with type
different than 0806 is treated as a VLAN frame if VLAN ARP is enabled).
A.4.4
Configuration Bits for Fixed Wake-up filters
The following configuration bits were added to the 82558 B-step to support fixed wake-up
filtering:
Table 66. Fixed Wake-up Configuration Bits
Configuration Bit
Description
This bit enables a wake-up event upon reception of ARP frames with a dynamic
presence of a VLAN header. This bit takes effect only if the ARP enable bit is
also set.
VLAN ARP
Link Status Change Wake
Enable
This bit enables assertion of the PME# signal upon a link status change event.
The PME# signal is further gated by the PME enable bit in the PMCSR.
This bit enables assertion of the PME# signal upon reception of ARP frames.
The PME# signal is further gated by the PME enable bit in the PMCSR.
ARP Wake Enable
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
143
Download from Www.Somanuals.com. All Manuals Search And Download.
Wake-up Functionality
Table 66. Fixed Wake-up Configuration Bits
Configuration Bit
Description
This bit enables assertion of the PME# signal upon reception of packets that
pass through the multicast address filtering. The PME# signal is further gated by
the PME enable bit in the PMCSR.
Multicast Match Wake
Enable
These bits are used for ARP frame filtering. These are the least significant bits
that make the low byte of the IP address, which is located in byte number 40 in a
packet without a VLAN header and byte number 44 in a packet with a VLAN
header.
IP Address Low Byte
These bits are used for ARP frame filtering. They make the high byte of the IP
address, which is located in byte number 41 in a packet without a VLAN header
and byte number 45 in a packet with a VLAN header.
IP Address High Byte
IA Match Wake Enable
This bit enables the assertion of the PME# signal upon reception of packets that
pass through the Individual Address filtering. The PME# signal is further gated
by the PME enable bit in the PMCSR.
Magic Packet Wake-up
Disable
This bit disables wake up upon reception of a Magic Packet frame.
A.5
A.6
Link Status Event
The controller may be configured to wake up the system on link disconnect and connect events.
The link status wake-up enable bit was added to the configuration command for the 82558 and later
generation devices. If this bit is set, the device generates a power management event when it
detects a link disconnect or connect. A link status event is indicated through the Power
Management Driver Register (PMDR).
Note: In the 82558 B-step, there is no indication to differentiate this event from any other wake-up event.
Frame storage is also not associated with it.
Flexible Packet Filtering
Programmable packet filtering is targeted to support wake-up packets that are not supported by the
hardware filters. Known wake-up packets are listed below.
Wake-up packets supported by the hardware filters are:
• ARP packets at EII packet with single IP address filtering
• Directed packet
• Multicast Address
• Magic Packet
• Link event
Wake-up packets not supported by hardware filters are:
• Directed IP packet
• Net BIOS queries
• 802.2 ARP
144
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Wake-up Functionality
• Multiple IP address recognition
A.6.1
Flexible Filtering Terminology
Filter. A filter is a set of a signature and segments generated for a specific frame format. Each
filter defines one frame that causes the controller to wake the system. The Intel Fast Ethernet
controllers support 3 different filters.
Segment. A segment defines a continuous sequence of bytes in a frame that should be filtered by
the controller. Any byte in a frame, which is not included in any segment, is ignored during
filtering. A segment is defined by 3 parameters:
• Location of first Dword in the segment
• Location of last Dword in the segment
• Byte masks for first and last Dwords
The Intel controllers support a variable number of segments per filter.
Mask. A mask defines which bytes in each segment should be filtered by the device. Since each
segment holds a continuous sequence of bytes, the mask information is required only for the first
and last Dwords of the segment.
Signature. A special type of CRC is calculated for each filter. There is one 32-bit signature for all
segments of each filter. For each frame, the controller compares the filter signature with the one
calculated on all segments and wakes the system if the signatures are equal.
A.6.2
Flexible Filtering Limitations
The flexible filtering feature in the device requires an active clock at the CLK input, which can be
the PCI clock while the system bus is at B0 or B1. An active clock should be provided externally
while the host bus is in the B2 or B3 states. The controller provides a 25 MHz output clock that can
be used to feed the clock during these states. External clock switching circuitry is required to
ensure a smooth transition between the PCI and the 25 MHz clock. For more details refer to the
device data sheet.
Flexible filtering and frame storing are disabled while the Force TCO pin is asserted. In this case,
the device wakes the system when it receives a packet that matches one of the fixed packet filters
or when a link status change event occurs. However, no frames will be stored.
Only the first 124 bytes of each frame may be filtered and stored. The frame length is not limited.
The segment limitations are:
• The maximum total number of segments in all filters is 10.
• The first filter may hold 1 to 4 segments.
nd
• The 2 filter may hold 0 to 6 segments.
rd
• The 3 filter may hold 0 to 2 segments.
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
145
Download from Www.Somanuals.com. All Manuals Search And Download.
Wake-up Functionality
A.6.3
Wake-up Packet Storage
The device uses its internal registers for packet storage during power down mode. Only the first
124 bytes of a frame may be filtered and stored by the device. The residual section of the packet
beyond the first 124 bytes is discarded by the device and cannot be used by the software driver
after the system is awake.
Note: Any packet passing the fixed packet filtering mechanism (such as ARP) can be stored in the
temporary storage area of the micromachine and will be available to the driver when the system is
awake. The same limitation for the flexible packet filter storage holds to the hardware filter
storage. However, in cases where the frame is inherently lost (for example, in an overrun) or the
flexible packet filter is not active (either not loaded or temporarily disabled as in TCO mode), the
controller may wake up the system without storing the packet.
A.7
82559 and Later Generation Device Implementation
The 82559 uses different methods for loading programmable wake-up packet filters (both fixed and
flexible filters) that are more elegant sophisticated than the mechanisms used by the 82558. The
following subsections detail wake-up packet filtering for the 82559.
All of the following subsections refer to the 82559 and 82559ER. However, the 82559ER does not
support Magic Packet.
A.7.1
Load Programmable Filter Command Structures
The load programmable filter command of the 82559 is not backward compatible to the 82558.
The 82559 has no explicit limitation on the number of programmable filters, but the length of the
command structure is limited to 18 Dwords. The command structure is presented here.
Figure 27. Command Block Structure
31
16 15
0
CB Command Word
CB Status Word
Next CB Pointer = FFFF FFFFh
Programmable Filters Data (up to 16 Dwords total)
• Command Word. The end of list (EL) and start (S) bits are contained in this word. Either of
these must be active (1). The load programmable filters command should be the last command
issued to the 82559 by the driver. Therefore, the Next CB Pointer field should be a null pointer
(FF,…Fh), even if the S bit is used. The CB opcode is 0008 0000h.
• Status Word. The status word contains the complete (C) and OK bits. After completing the
load programmable filters command, the 82559 posts a complete OK word equal to A000h.
• Next CB Pointer. The Next CB Pointer should be set by the software driver to FF,…Fh (null
pointer).
• Programmable Filters Data Structure. Programmable filters may have one of two types:
flexible filters or pre-defined filters. Any receive packet is filtered first by the MAC address
filtering and then by the filters defined below:
146
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Wake-up Functionality
— Pre-Defined Filters.
The 82559 contains pre-defined filters for both Ethernet Type II and 802.2 snap. The
82559 distinguishes between these two types according to the MAC Type/Length field.
The 82559 also handles VLAN tagging per packet filter as described below. The Magic
Packet filter is a unique pre-defined filter, enabled by the configuration command. All
other pre-defined filters are controlled by the load programmable filter command.
Filter Type
Reserved (0)
Word Match
Filter Type:
31
30
29
28
27
26
25
24
EL
FIX=1
VLAN
NBH
ARP
IA-Type
IA-Match
TCO
EL
The EL bit indicates if this filter is the last active one.
If this bit is set 1, it indicates a pre-defined filter.
FIX
When this bit is active (1), the 82559 skips a 4-byte field of the VLAN type tagging (byte
offsets 12 through 15). The VLAN bit affects only the filters defined in the same byte.
Other filters that may possibly be affected are the ARP and IA-Type filters.
VLAN
The neighborhood filter looks for a word match defining a 2-byte field within the MAC
destination address to be compared. These two bytes compose a hash value of the
node IP address located in offset 02h and 03h of the MAC address. The active fields of
the NBH filter are shown below:
NBH
Byte Offset
Pattern
Meaning
0
Odd value
Word match
LSB of the MAC destination address.
Hash value of the destination IP address.
3, 2
The ARP filter is used to find a word match with the lower 16 bits of the IP address.
Typically, it is ideal for IP filtering. ARP filtering is based on the active fields shown
below. The byte offset of the active fields in 802.2 snap type are shifted by 8 bytes
ahead. If the filter VLAN flag is set, it is shifted by 4 bytes. For example, the opcode byte
is at offset 21 in EII with no VLAN and at offset 33 in 802.2 snap with active VLAN.
ARP
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
147
Download from Www.Somanuals.com. All Manuals Search And Download.
Wake-up Functionality
EII Byte Offset
13, 12
Pattern
06, 08
Meaning
ARP type = 0806h.
21
01
ARP opcode (01 = request).
Two LSB of the destination IP address.
41, 40
Word match
The IA with Type match bit monitors for the destination MAC address of the incoming
packet identical to the 6 bytes of the 82559 Individual Address, and the MAC Type/
Length field must be identical to the filter word match fields. The byte offset of the MAC
Type/Length field in 802.2 snap type is shifted by 8 bytes ahead and shifted by an
additional 4 bytes if the filter VLAN flag is set.
IA-Type
The destination MAC address of the incoming packet is identical to the 6 bytes of the
82559 Individual Address.
IA-Match
TCO
The 82559 supports a TCO packet filter. The 82559 recognizes TCO packets for power
management events at Ethernet Type II with or without VLAN tagging according to the
following fields.
Byte Offset
5, 4, ..., 0
13, 12
14
Pattern
IA Address
00, 08
45
Meaning
MAC destination address.
IP type = 0800h.
IPv4, 4 Dword length.
Protocol: either UDP or TCP.
TCO Port number 026Fh (0623 decimal).
23
11/04
37, 36
02, 6F
NOTE: If receive to a TCO controller is enabled, then this bit has no affect on PME.
There are two groups of pre-defined filters:
— The NBH, ARP and IA-Type pre-defined filters each use the word match field.
— The IA-Match and TCO pre-defined filters do not use the word match field.
A single programming word may contain both a pre-defined filter from group 1 and any
of the pre-defined filters in group 2. For example, a single programming word may be
used to activate: ARP IA-Match and TCO with or without VLAN tagging (depending on
the VLAN bit). The examples below describe 82559 wake-up programming for:
— ARP without VLAN, IP address equals 8F.B9.3F.3Dh
— Individual IP with VLAN tagging
— Wake on TCO packet without VLAN (due to unavailable hardware management on
the SMB)
— Neighborhood name query (which is described in more detail under the Flexible
Filter bullet)
Example 5. 82559 Wake-up Programming
CB Command / Status Word 8008 0000
The EL bit is active.
This is the pointer to the next CB structure (no affect if
the EL bit is active).
Next CB PTR
yyyy yyyy
Pre-defined Filter 1
Pre-defined Filter 2
Flexible Filter
4900 3F3D
6400 0800
ARP and TCO share the same programming filter.
IA-Type with VLAN tagging with IP type (0800h)
98xx xxxx
zzzz zzzz
zzzz zzzz
zzzz zzzz
This is the flexible filter for neighborhood name query.
The CRC word equals xxxxxxh. The filter mask is
composed of 3 Dwords. If the filter resources are a
concern, the driver may use a more efficient mask for
this filter.
148
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Wake-up Functionality
• Flexible Filter
Filter type
CRC word
Filter mask (up to 4 Dwords: DW0, DW1, DW2, DW3)
Filter type:
31
EL
30
29
28
27
26
25
24
FIX=0
MLEN
Reserved (0)
EL
The end of list bit indicates if this filter is the last active one.
Clearing this bit to 0 indicates a flexible filter.
FIX
This field defines the mask length. Each bit in the mask defines a byte in the packet. For
example, byte 0 in DW0 starts with byte 0 of the packet, and bit 31 of DW2 defines byte 95
of the packet.
MLEN
000
011
Single Dword mask
3 Dword mask
001
111
2 Dword mask
4 Dword mask
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
149
Download from Www.Somanuals.com. All Manuals Search And Download.
Wake-up Functionality
A.7.2
CRC Word calculation of a Flexible Filter
COEFFICIENTS = 0x04C11DB7;
Signature = 0;
// CRC Polynomial Coefficients
// Initialize the CRC Signature
for(n=i=0; n<128 & n<FrameLength; ++n)
// Search the frame, using only bytes
that are part of the flexible filter
pattern mask.
{
if(Byte n of the Frame is in the
Flexible Filter)
{
// Shift factor to move the frame byte to
bit position 0, 8 or 16 of the 32-bit
Dword.
ShiftBy = (i modulo 3) * 8;
// Bit position at 24 of the Dword is not
used, because the signature is only 24
bits.
if(Signature AND 0x80000000)
{
Signature = (Signature << 1)
// If the most significant bit is 1, XOR
in the coefficients, previous signature
and frame byte shifted over to the
proper position.
XOR
(FrameByte << ShiftBy)
XOR
COEFFICIENTS;
}
// Otherwise, just XOR the previous
signature and frame byte shifted over to
the proper position.
else
{
Signature = (Signature << 1) XOR
(FrameByte << ShiftBy);
}
// Include the Dword slot for the next
frame byte that is in the pattern mask.
++i;
}
}
output bits 0-23 of Signature
A.7.3
Port Dump Wake Up Packet
up Packet command:
Table 67. 82559 Port Commands
Pointer Field
Opcode Field
Function
31
4
3
0
Software Reset
Self Test
Don't care
0000
0001
0010
0011
0111
Self test results pointer (16-byte alignment)
Don't care
Selective Reset
Dump
Dump area pointer (16-byte alignment)
Dump area pointer (16-byte alignment)
Dump Wake-up
Following the Dump Wake-up command, the 82559 writes the stored data of the wake-up packet to
the host memory, starting at the address specified in the pointer field. The dump data structure is
shown below.
150
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Wake-up Functionality
Table 68. Dump Data Structure
Offset
D31
D0
0
1
Reserved
Status Word (A000h)
Byte Count
Reserved
Wake-up Packet
2:N
The sequence of events after a Dump Wake-up command that the 82559 performs are:
1. Write the byte count field at Dword 1. This field contains the actual number of bytes posted in
the host memory. A value of FFh indicates that the Wake-up packet length exceeded the 120
bytes. In this case only the first 120 bytes are posted.
2. Write the wake-up packet data starting at Dword 2.
3. Write a status word composed of the Complete OK bits equal to A000h at Dword 0.
Prior to the Dump Wake-up packet command, the driver should first initialize the status word to 0.
After the Dump Wake-up packet command, it should pole the Status word for the Complete bit.
Note: The port dump wake-up packet causes an internal selective reset to the 82559.
Note: The interesting packet bit in the PMDR is set together with the PME status bit when an interesting
packet is received.
Note: The 82559 uses the statistic counters resources to store the wake-up packet. The software driver
should assume that the statistic counters are infected at power down state. Therefore, it should
issue a Dump Reset Statistic Counters command before it resumes nominal operation.
Note: Magic Packets are exceptional to all other wake-up packets. The Magic Packets may cause a power
management event and set an indication bit in the PMDR. However, it is not stored by the 82559
for system use when it is awake. The 82559ER does not support Magic Packet wake up.
A.7.4
Power Management Software Flow
The 82559 adheres to the PCI Power Management Specification supporting all three power states
D0 through D3. This section describes the required flow of events for software to set the 82559 into
the power down states.
A.7.4.1
Power Down without Wake-up Capabilities
The 82559 supports a very low power state if wake-up capabilities are not required. The OS should
follow these steps:
1. Clear the PME enable bit in the PMCSR to the PME disable state.
2. Set the 82559 to the D3 power state (by OS).
Note:
Step 2 may be executed together with step 1 (same cycle) but not before. If step 2
is executed before step 1, the 82559 might assert PME# if wake on link status
change is enabled. At this state, the 82559 enters the deep power down state, where
the PHY is turned off. At this state, the 82559 consume less than 7 mA if the PCI
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
151
Download from Www.Somanuals.com. All Manuals Search And Download.
Wake-up Functionality
CLK is inactive and 10 mA if it is active. The deep power down state due to PME
disable is enabled in the EEPROM.
A.7.4.2
Power Down with Wake-up Capabilities
The 82559 provides wake-up capabilities at all power states. At the D0 state, the 82559 provides a
wake on link status change capability, while wake on interesting packets is operating at the other
power states: D1, D2, D3 , D3
.
hot
cold
Note: The 82559 requires a presence of a system power supply to support wake-up capabilities in the
D3 state (bit 4 of the PMC register). The OS should follow the following steps listed:
cold
1. Set the RU into the idle state.
2. Issue a configure command to the 82559 setting the wake-up on Magic Packet (except for the
82559ER as it does not support Magic Packet) and wake on link status change (which is
supported on 82559ER) as required. If the 82559 has been configured at initialization time as
required for the power down state, then this step is redundant.
3. Load the programmable filter if wake on interesting packet is required. This command must be
issued at each entrance to the power down state. It must be the last command issued to the
82559 by the driver before it is set by the OS to the power down state as shown in the next
step(s).
4. Wait for the command to be completely executed and the CU is at the idle state.
5. Set the PME enable bit in the PMCSR to the active state (by OS).
6. Set the 82559 to the selected Dx (x > 0) power state (by OS).
Note: Step 6 may be executed together with step 5 (same cycle) but not before. Otherwise, the 82559 will
issue a redundant power management event. If the deep power down due to PME disable is enabled
and the 82559 is set into the D3 state while PME is not enabled, the PHY unit will be turned off.
After the PME enable bit is set to the active state, the PHY is turned on. After link is established,
the 82559 issues a power management event due to link status change if wake on link status change
is enabled.
A.7.4.3
Wake-up Sequence
To initiate the wake-up sequence, the software (OS and driver) are expected to follow the steps
below:
1. OS sets the 82559 to the D0 state.
2. OS is expected to read and clear the PME bit in the PMCSR. If the PME bit is not cleared, the
82559 will hold the PME# signal active.
3. If the device is set to D0 from D3, initialization of the base address registers are required.
4. The driver should read and clear the PME indication bits in the PMDR. The PME indication
bits are: link status change bit, Magic Packet bit, and the interesting packet bit. The PME
status bit in the PMDR may already be cleared by the OS through the PMCSR.
5. The driver should either issue a selective reset command or dump wake-up packet command,
depending whether the interesting packet bit is set in the PMDR. If the interesting packet bit is
active, the driver should read the wake-up packet using the port dump wake-up packet
command.
152
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
Wake-up Functionality
6. The device statistic counters are corrupted during power down state. Therefore, the driver
should clear the statistic counters by first issuing load dump counters address and then a dump
and reset statistic counters.
7. If the 82559 was at the D3 power state, a full initialization sequence is required. Otherwise, the
driver should initialize the CU and RU base addresses.
8. The driver should issue an RU start command pointing to a valid RFA.
9. The driver is now ready to issue transmit commands and receive packets.
A.7.4.4
Dummy Wake-up Sequence
After the OS requests to set the 82559 in the power down state, the driver aborts the receive
activity. Then, the driver loads the required programmable filters to wake-up the system. At this
point, the 82559 is capable of waking up the system if the power management event is enabled.
The driver informs the OS that it may place the 82559 in the power down state. The OS may
proceed with the power down sequence, or it might resume nominal operation. If the 82559 driver
is requested to resume nominal operation, it should follow the same flow as if the 82559 was at the
power down state.
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
153
Download from Www.Somanuals.com. All Manuals Search And Download.
Wake-up Functionality
154
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
82550 and 82551QM Specific
Information
B
®
This appendix applies to the Intel 82550 and 82551QM devices.
B.1
IPCB
The IP command block (IPCB) is new and used to activate the new offloading features of the
82550 and 82551. The value of the command field for IPCB is 9h. The relevant aspects of the
IPCB for each feature is described in the following subsections. This section summarizes the most
useful combinations of the IPCB fields.
Table 69. IPCB Structure
Odd Word (D31:D16)
Even Word (D15:D0)
Offset
E
L
N
C
O
K
S
I
CID (5 bits)
000
CMD = 1001
C
X
U
XXXX XXXX XXXX (12 bits)
0h
Link Address (32 bits)
4h
8h
TBD Array Address (32 bits)
E
Maximum TCP Payload
O
F
0
IPCB Byte Count (14 bits)
Ch
TBD Number
Transmit Threshold
IP Activation (12 bits)
TCP Header Offset (8 bits)
Activation (20 bits)
VLAN (16 bits)
10h
14h
IP Header Offset (8 bits)
Transmit Buffer #0 Address (32 bits)
Reserved (32 bits)
18h
1Ch
Transmit Buffer #0 Size (14 bits)
Reserved (14 bits)
E
0
L
Total TCP Payload (16 bits)
NOTE: Source address insertion is not allowed when IPCB is used. Thus, the no source address insertion
(NSAI) bit of the configure command must be 1.
Table 70. IP Activation Bits (Byte 13)
15
14
13
12
11
10
9
8
Scheduled
Send
Insert
VLAN
Hardware
Parse
0
0
0
Reserved
Reserved
Table 71. IP Activation Bits (Byte 12)
7
6
5
4
3
2
1
0
TCP/UDP
Number
TCP/UDP
Checksum
IP
Large Send
X
X
X
X
Checksum
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
155
Download from Www.Somanuals.com. All Manuals Search And Download.
82550 and 82551QM Specific Information
The location and definition of the IPCB fields are summarized in the following table.
Table 72. IPCB Fields
Field Name
Byte
Bit(s) Function
Description
This field specifies the cumulative number of TCP payload
bytes requested for transmission as part of this IPCB
instance. This field must be valid and will be used when the
Hardware Parsing bit is clear. The Total TCP/UDP Payload
field is ignored when the Large Send bit is clear.
Total TCP/UDP
Payload
1Fh:
1Eh
Parameter,
Status
15:0
This field defines the offset in bytes from first frame byte
retrieved in memory to the first byte of the TCP or UDP
header. This field is ignored when the Hardware Parsing bit is
set. When the IP Checksum and TCP/UDP Checksum bits
are set and the Hardware Parsing is clear, the TCP Header
Offset parameter must be greater than IP Header Offset
parameter and IP header length combined, as derived from
the data.
TCP Header
Offset
17h
16h
7:0
Parameter
This field defines the offset in bytes from first frame byte
retrieved in memory to the first byte of the IP header. The 4
VLAN bytes are not included in the count only if it is inserted
by hardware. This field is ignored when the Hardware Parsing
bit is set.
IP Header Offset
VLAN
7:0
Parameter
Parameter
If the Insert VLAN bit is set, the controller inserts the VLAN
type (8100h, also known as QTAG protocol type). This 16-bit
VLAN ID number and the user priority (32 bits total) are
inserted in the transmitted frame.
15h:
14h
15:0
Reserved
13h
13h
7:5
4
Reserved
Mode
These bits are reserved and should be set to 000.
When this bit is set, the controller transmits the associated
frame using the scheduling mechanism.
Scheduled Send
If this bit is set, the controller inserts the VLAN type and tag
as defined in the VLAN bits description.
Insert VLAN
13h
1
0
Mode
Mode
When this bit is clear, hardware parsing of the transmitted
packet is disabled. If this bit equals 0 (clear), the IP Header
Offset, TCP Header Offset and total TCP/UDP payload must
be specified to enable checksum operation.
Hardware Parsing 13h
When this bit is set, it indicates that the associated packet
should be transmitted using the Large Send mechanism.
Buffer(s) chained to the IPCB contain data larger than what
may be sent in a single Ethernet frame. The 82550 and
82551 break large packets into smaller ones and transmit
them using several Ethernet frames.
Large Send
12h
7
6
Mode
This bit is used to differentiate between TCP and UDP
packets. When it is set, it indicates a TCP frame; when clear,
a UDP frame. This parameter is relevant and required if the
Hardware Parsing bit is clear and the TCP/UDP Checksum is
set.
TCP/UDP Number 12h
Parameter
TCP/UDP
12h
This bit indicates that the TCP/UDP header checksum should
be performed in hardware.
5
4
Mode
Mode
Checksum
This bit indicates that the IP header checksum should be
performed in hardware.
IP Checksum
12h
156
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
82550 and 82551QM Specific Information
Table 72. IPCB Fields
If the Scheduled Send bit is set, the transmission of the frame
associated with the IPCB is delayed until the internal
scheduling counter of the controller reaches the scheduling
value.
12h:
10h
Scheduling
19:0
Parameter
Bits 16:0 are the time stamp and bits 19:17 may be used to
specify time window.
This field specifies the maximum number of bytes to be
transmitted in the TCP payload of an Ethernet frame that is
part of the Large Send. All frames use this size except the last
one. This field is ignored when the Large Send bit is clear.
Maximum TCP
Payload
Fh:Eh 15:0
Parameter
Parameter
In units of 8 bytes, this field defines the number of bytes that
should be present in the Transmit FIFO before transmission
starts. It is relevant only if the checksum requires the whole
frame to be in Transmit FIFO. The value of this parameter
should be in the range of 1h to E0h.
Transmit
Threshold
Eh
7:0
Note: Using software parsing is only allowed with legal TCP/IP or UDP/IP packets. When software
parsing is used, IP and TCP offsets in the IPCB must point to the appropriate headers and the total
TCP/UDP payload should be specified. For all other datagrams, hardware parsing must be used.
Note: VLAN tagging (either in memory or by hardware) and SNAP headers are allowed when software
parsing is used as long as the above guideline is met. IP and TCP header offsets are calculated
according to memory data structures. For example, if VLAN tag is inserted by hardware, offsets do
not change.
The IPCB byte count specifies the number of bytes of transmit data in the immediate data area
(from offset 20h and on).
B.1.1
Maximum TCP Payload / TBD Number and Transmit
Threshold
When the Large Send mode is selected, this field is interpreted as the Maximum TCP Payload.
Otherwise, this 16-bit field is interpreted as the TBD number and Transmit Threshold. The TBD
number specifies the total number of TBDs associated with this IPCB. The Transmit Threshold is
Note: A mode bit puts the 82550 and 82551 into a specific mode. When the mode bit is clear, all
parameters that relate to that mode are ignored by the device hardware. Mode bits should obey
consistency rules since not all combinations are allowed. The result of a non-supported
combination is unpredictable.
Note: It is driver responsibility to guarantee that the total length of a frame is not larger than the allowed
MTU size for that connection (which is dependent upon the IP address pair). This is relevant in all
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
157
Download from Www.Somanuals.com. All Manuals Search And Download.
82550 and 82551QM Specific Information
modes of operation. For instance, the driver must guarantee proper values for the Maximum TCP
Payload in Large Send mode and VLAN length inclusion.
Note: IP fragmentation is not supported by the 82550. Therefore, the driver should not request any
offload features for an IP fragment.
B.2
82550 Checksum Operation
The 82550 hardware supports an IP header (including options), UDP and TCP checksum
computation. The 82550 computes the checksum using two’s complement addition, and therefore
performs end around carry. In other words, it adds the carry generated from bit 15 to the least
significant bit in checksum computation.
Note: The 82550 is capable of computing one level of IP header and one TCP/UDP header and payload.
In case of multiple IP headers, the driver should compute all except but the outer most IP header
checksum.
Note: Per RFC768, if the computed UDP checksum is zero, it is transmitted as all ones. A UDP
checksum transmitted as all zeros means that the transmitter generated no UDP checksum.
B.2.1
Driver Interface
Software should compute the partial checksum of the pseudo header and store it in the checksum
field of the TCP or UDP header. The partial checksum is the 16-bit one’s complement sum of the 6
words in the pseudo header. It is not the one’s complement of the one’s complement sum. If this is
not done by the stack, the driver must be aware of the IP header offset and base offset and the
length of the TCP segment to compute and store the pseudo header checksum.
The IPCB structure used for the checksum offload is presented below. The fields relevant to
checksum operation are shaded.
Table 73. IPCB Structure Checksum Offload
Odd Word (D31:D16)
Even Word (D15:D0)
Offset
E
L
N
C
O
K
S
I
CID (5 bits)
000
CMD (4 bits)
C
X
U
XXXX XXXX XXXX (12 bits)
0h
Link Address (32 bits)
4h
8h
TBD Array Address (32 bits)
E
Maximum TCP Payload
O
F
0
IPCB Byte Count (14 bits)
Ch
TBD Number
Transmit Threshold
IP Activation (12 bits)
TCP Header Offset (8 bits)
Scheduling (20 bits)
10h
14h
IP Header Offset (8 bits)
VLAN (16 bits)
Transmit Buffer #0 Address (32 bits)
Reserved (32 bits)
18h
1Ch
Transmit Buffer #0 Size (14 bits)
Reserved (14 bits)
E
0
L
Total TCP Payload (16 bits)
158
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
82550 and 82551QM Specific Information
B.2.2
IPCB Field Assignment
The mode bits in the IP Activation field control the checksum operation of the transmit command.
• IPv4 Checksum (1bit). When this bit is set to 1, the device is forced to perform checksum
operation using the IPv4 header and option fields (when present) prior to transmission. This
mode can be used with any frame and when the TCP/UDP checksum, Large Send, VLAN,
Hardware Parsing, or Scheduling modes are used.
Note:
The stack can command the driver to compute a checksum or not on a frame by
frame basis (or specific to the IP or TCP/UDP fields). As a result, the driver may
command the hardware not to perform the checksum operation for an IP header.
This would be the case when a frame is not a native TCP/UDP/IPv4 frame.
• TCP/UDP Checksum (1bit). When this bit is set to 1, the controller is forced to perform TCP
checksum operations (including option fields when present) or UDP checksum operations,
based on the TCP/UDP parameter defined prior to transmission. This mode can be used with
any frame and when IPv4 checksum, Large Send, VLAN, Hardware Parsing, or Scheduling.
Note:
The stack can command the driver to compute the checksum or not on a frame by
frame basis (or specific to IP or TCP/UDP fields). As a result, the driver may
command the hardware not to perform the checksum operation for a TCP/UDP
frame. This would be the case where the frame is not a native TCP/UDP/IPv4
frame.
• Hardware Parsing (1 bit). When this bit is set to 1, the controller is forced to use its internal
transmit parser (autonomous mode is set). The internal transmit parser generates the IP header
offset and TCP/UDP header offset for internal consumption by the device blocks. In
autonomous mode, the IP header offset, TCP/UDP number and TCP/UDP header offset
parameters are ignored. When the Hardware Parsing bit is clear (semi-autonomous mode), the
controller uses the IP header offset and TCP/UDP header offset parameters to locate the IPv4
header and TCP/UDP header offsets in the transmit data stream. These are fetched from host
memory. Hardware Parsing mode can be used when the IPv4 checksum, TCP/UDP checksum,
VLAN, Scheduling, or Large Send bit is set. The Hardware Parsing bit must be cleared when
the Security bit is set.
B.2.2.1
Parameters
• TCP/UDP Number (1bit). This parameter is used to denote the frame type for the checksum
operation when the TCP/UDP checksum mode is set. When this bit is set, the TCP checksum
is performed, and when it is clear, UDP checksum is performed. This parameter is relevant if
the Hardware Parsing bit is clear and the TCP/UDP checksum is set. It is also used to select the
location of the checksum field.
• IP Header Offset (8 bits). This parameter denotes the offset from the first byte of the
destination address (DA) field, which is read by the controller from host memory, to the first
byte of the IPv4 header. The 82550 reads this parameter when the Hardware Parsing bit is
clear and the IPv4 checksum set.
Note:
VLAN mode will not impact this number since the 4 VLAN bytes are not included
in the count since they are inserted by hardware.
• TCP/UDP Header Offset (8 bits). This parameter denotes the offset from the first byte of the
destination address (DA) field, which is read by the network device from host memory, to the
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
159
Download from Www.Somanuals.com. All Manuals Search And Download.
82550 and 82551QM Specific Information
first byte of the TCP or UDP header. The controller reads this parameter when Hardware
Parsing is clear and TCP/UDP checksum is set.
Note:
Note:
If TCP/UDP headset offset is specified, then the IP header offset must also be
specified.
VLAN mode does not impact this number since the 4 VLAN bytes are not
included in the count if they are inserted by hardware.
• Transmit Threshold.
B.2.2.2
Status
• If IPv4 checksum mode is set in the IPCB, but the IPv4 header was not found or had a length
field smaller than 5, the 82550 may corrupt the frame.
• If TCP/UDP checksum mode is set, but a TCP/UDP header was not found or the length of the
TCP/UDP packet was smaller than the offset of the checksum, the 82550 may corrupt the
frame.
B.2.2.3
Configuration
IPCB usage is enabled only if the 82550 is in the Extended TxCB mode.
B.2.3
Data Flow
The flow of events for a single transmit frame with checksum offloading is described below.
1. The 82550 enters checksum mode if IPv4 checksum or TCP/UDP checksum is set.
2. If Hardware Parsing is clear, the 82550 uses the IP header offset and TCP/UDP header offset
from IPCB. Otherwise, the 82550 parses the frame to locate these offsets.
3. The 82550 computes the checksum over the IP header while it is being access from memory
using the IP header length field from the frame (regardless of Hardware Parsing). The
checksum is stored in the appropriate location.
4. The 82550 computes the checksum over the TCP/UDP header and payload according to the
TCP/UDP header offset, the IP header total length, and the TCP or UDP bit (according to the
IP header protocol field if Hardware Parsing is set or according to the TCP/UDP bit if
Hardware Parsing is clear). The checksum is stored in the appropriate location.
5. At this point, the frame is ready for transmission.
The TCP/UDP checksum includes the pseudo header partial checksum that should be stored by the
driver in the checksum field. In Large Send mode, the driver should store the partial checksum of
the pseudo header of the first packet (which includes the header and payload length [or TCP
segment size] of the first frame). The controller generates the correct payload length and
corresponding pseudo header checksum for each frame transmitted within the Large Send.
160
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
82550 and 82551QM Specific Information
Note: The partial checksum required by the 82550 is not the partial checksum passed by the Microsoft*
IP stack per Microsoft offloading specification, v0.106.
For TCP/UDP checksum computation, the 82550 requires that the whole frame is copied into its
internal FIFO. The result is stored in the proper header fields and only then can the frame be
cleared for transmission or further processing. The 82550 pipelines transmission of frame N-1 and
the checksum computation of frame N such that back-to-back frames on the transmit queue will
flow at maximal DMA rate and checksum operation and transmission are at wire speed. This
minimizes the performance hit to “first frames events,” where the 82550 transmit FIFO was empty
prior to the transmission command.
B.2.4
Tunneling Support
The 82550 and 82551 provide one level of IPv4 header and TCP/UDP checksum computation. In
IP tunnel mode, the driver is expected to compute all IP headers checksum except one that may be
handled by the device. As these frame formats are not supported in autonomous mode, the driver is
expected to provide the semi-autonomous mode parameters to the 82550.
B.3
Large Send
B.3.1
Rationale
Large Send (also known as TCP Segmentation Offload [TSO]) allows the stack to transfer a frame
(larger than the MTU for that media) to the driver. (The maximum frame size for 10/100 Mbps
Ethernet is 1518 bytes.) For TCP/IP, a typical frame size may be the transmit window (the default
is 8760 bytes, which is approximately 6 full size frames). The 82550 supports Large Send of
frames up to 64 Kbytes. Performance increases would result from:
• Stack computes only one header per block.
• Stack does not segment the block to fit the MTU size. This decreases the overhead of multiple
virtual address mapping and linked list management.
The 82550 loads the prototype header from host memory only once (and stores it in the device) to
reduce PCI overhead.
The Large Send operation is supported for TCP only (not for UDP). Since servers are typically
more involved in transmission than client systems, the Large Send feature primarily benefits these
server systems. The estimated reduction in CPU utilization is 20%.
B.3.2
Driver interface
The IPCB structure used for Large Send operation is presented below. The fields relevant for the
Large Send feature are shaded. The driver must ensure that the TCP payload of a Large Send frame
is greater than the maximum TCP payload. In other words, there must be more than one frame in a
Large Send.
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
161
Download from Www.Somanuals.com. All Manuals Search And Download.
82550 and 82551QM Specific Information
Table 74. IPCB Structure Large Send
Odd Word (D31:D16)
Even Word (D15:D0)
Offset
E
L
N
C
O
K
S
I
CID (5 bits)
000
CMD = 1001
C
X
U
XXXX XXXX XXXX (12 bits)
0h
Link Address (32 bits)
4h
8h
TBD Array Address (32 bits)
E
Maximum TCP Payload
IP Activation (12 bits)
TCP Header Offset (8 bits)
O
F
0
IPCB Byte Count (14 bits)
Ch
Scheduling (20 bits)
10h
14h
IP Header Offset (8 bits)
VLAN (16 bits)
Transmit Buffer #0 Address (32 bits)
Reserved
18h
1Ch
E
0
L
Total TCP Payload (16 bits)
Transmit Buffer #0 Size (14 bits)
Note: To use Large Send, the 82550 should be configured to use dynamic Transmit Buffer Descriptors
(TBDs). The driver should ensure transmit buffers associated with the Large Send IPCB contain
enough bytes (headers size and total TCP payload). An end of list (EL) bit should be placed in the
last TBD.
B.3.3
IPCB Large Send
The mode bits in the IP Activation field, control the Large Send operation of the transmit
command. When the Large Send bit is set to 1, the 82550 is forced into Large Send mode for the
processing of this IPCB. The data bytes attached are subject to segmentation by the 82550 and
multiple frames of MTU size (except for the last frame less than or equal to the MTU size) will be
transmitted.
B.3.3.1
Parameters
• Maximum TCP Payload. This parameter specifies the maximum number of bytes that can be
transmitted in the TCP payload portion of an Ethernet frame that is part of this Large Send. All
frames in this Large Send, except the last one, will have exactly the maximum TCP payload
bytes of TCP payload. The last frame TCP payload size will be the remainder (in other words,
it will be less than or equal to the maximum TCP payload bytes).
Note:
The driver must guarantee that the maximum TCP payload value will not cause the
complete frame (including VLAN) to be larger than the allowed MTU for that
connection based on the IP address pair used.
Note:
The IEEE 802.3ac committee plans to extend the 10/100 Mbps Ethernet frame size
to account for VLAN fields. This should not impact the TCP/IP MTU.
• Total TCP Payload. This parameter defines the total number of TCP payload bytes (header not
included) requested for transmission as part of this Large Send. This parameter should not be
used as the sole indication for Large Send activation. For example, 1500 bytes can be a Large
Send on an MTU of 576 bytes.
162
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
82550 and 82551QM Specific Information
Note:
Note:
If the IPv4 checksum and TCP/UDP checksum are clear (checksum offload is not
requested), frames will be transmitted without computing and replacing the
checksum fields content. Therefore, the driver should set both bits so that the
generated frames carry a proper checksum.
When the Large Send feature is used, the 82550 should be configured to pad
outgoing packets to 64 bytes minimum size.
For Large Send packets, the driver should prepare all header fields so that they can be transmitted
in the first Large Send frame. For frame length related fields, the modifications may be required
before IPCB is submitted for transmission. Relevant header fields and their required operation
value for the prototype header are listed below.
• SNAP length (if SNAP is used). This should be adjusted for first frame transmission.
• IP total length. This should be adjusted for first frame transmission.
• TCP pseudo header partial sum. This should be adjusted according to first frame length.
• TCP push flag. This cannot be adjusted for the first frame. The driver leaves this flag
unchanged. If it is set, the 82550 turns it off in all frames except the last one transmitted.
• TCP FIN flag. This cannot be adjusted for the first frame. The driver leaves this flag
unchanged. If it is set, the 82550 turns it off in all frames except the last one transmitted.
B.3.4
Data Flow
The 82550 enters Large Send mode when the Large Send bit is set.
B.3.4.1
Headers Parsing
When the 82550 detects a Large Send IPCB, it parses its L2, IP and TCP headers to obtain header
offsets and additional headers information. For this purpose, the 82550 performs a PCI read DMA
from the beginning of the first active transmit buffer (either the IPCB immediate area or area
pointed to by a TBD). The size of that DMA is minimum. All headers (L2, IP and TCP) must be
contained in the first active transmit buffer.
The size of the on-chip prototype header memory is 80 bytes. The driver should ensure that the
header of Large Send packets fit in this buffer. This limits the size of the options that can be
included in the IP and TCP headers.
Note: Large Send IPCB must be used with the Hardware Parsing bit set.
Note: When Large Send is used, the 82550 should be configured to pad outgoing packets to a minimum
size of 64 bytes minimum.
B.3.4.2
First Frame Processing
• L2 header remains unchanged.
• IP header is processed:
— IP total length is unchanged.
— MF bit equals 0. This is expected but not checked by hardware.
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
163
Download from Www.Somanuals.com. All Manuals Search And Download.
82550 and 82551QM Specific Information
— Fragment offset equals 0. This is expected but not checked by hardware.
— IP options are not altered by hardware if they are present.
— IP header checksum is calculated by the checksum hardware if the IP checksum bit is set
in the IPCB.
• TCP header is processed:
— Sequence number is not changed.
— URG, RST, and SYN flags are expected to be clear but are not checked by hardware.
— ACK flag value will be transmitted as in the prototype header.
— Urgent pointer equals 0. This is expected but not checked by hardware.
— If the FIN flag equals 1, then it is cleared in the first frame.
— If the PSH flag equals 1, then it is cleared in the first frame.
— TCP options are not altered by hardware if they are present. TCP options include time
stamp.
— TCP checksum is calculated by checksum hardware if the TCP checksum bit is set in the
IPCB.
B.3.4.3
Intermediate Frames Processing
The number of bytes left for transmission after the first frame is calculated by subtracting the N *
Maximum TCP Payload from the Total TCP/UDP Payload. In other words,
(N * MAX TCP PAYLOAD) – (TOTAL TCP/UDP PAYLOAD)
If the next frame is not the last frame, the number of bytes transmitted is updated and the frame
processing rules below should be followed.
• L2 header should be left intact as in the first frame.
• IP header is processed:
— IP identification is incremented from last value.
— IP options are not altered by hardware if they are present.
— IP header checksum is calculated by checksum hardware if the IP checksum bit is set in
the IPCB.
• TCP header is processed:
— Sequence number is updated for outgoing packets. The previous frame size (TCP payload
32
only) is added to the previous frame sequence number (modulo 2 ).
— All flags are left intact as in the first frame.
— Urgent pointer is left intact as in the first frame.
— TCP options are not altered by hardware if they are present
— Pseudo header checksum remains unchanged.
— TCP checksum is calculated by checksum hardware if the TCP checksum bit is set in the
IPCB.
164
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
82550 and 82551QM Specific Information
When the headers are finished, they are subject to checksum for processing. The rest of the
transmission process is similar to transmission of a small send. This process repeats itself until the
last frame is about to be transmitted.
B.3.4.4
Handling the Last Frame
The number of bytes left for transmission after the first frame is calculated by subtracting the N *
Maximum TCP Payload from the Total TCP/UDP Payload. In other words:
(N * MAX TCP PAYLOAD) – (TOTAL TCP/UDP PAYLOAD)
To obtain the last frame TCP payload size, the above calculation is performed. The last frame TCP
payload size is less than or equal to the maximum TCP payload. The last frame is processed using
the rules listed below.
• L2 header is processed. If SNAP is used, the SNAP length field is calculated using the IP total
length computations for the IP total length.
SNAP length field =
(SNAP length of first packet) – (maximum TCP payload - TCP
payload of last packet)
• IP header is processed:
— IP total length equals the IP total length of the first packet.
(MAX TCP PAYLOAD) – (TCP payload of last packet)
— IP identification is incremented from the last value.
— IP options are not altered by hardware if they are present.
— IP header checksum is calculated by checksum hardware if the IP checksum bit is set in
the IPCB.
• TCP header is processed:
— Sequence number is updated for outgoing packets. The previous frame size (TCP payload
32
only) is added to the previous frame sequence number (modulo 2 ).
— ACK, URG, RST and SYN flag values are transmitted as in the prototype header.
— If the FIN flag equals 1 in the prototype header, it should also be set in this last frame.
— If the PSH flag equals 1 in the prototype header, it should also be set it in this last frame.
— Urgent pointer is left intact as in the first frame.
— Pseudo header checksum is fixed with the TCP length as above.
— TCP options are not altered by hardware if they are present.
— TCP checksum is calculated by checksum hardware if the TCP checksum bit is set in the
IPCB.
The 82550 reports the transmit status in the IPCB upon completion for all data in the Large Send.
This enables the whole Large Send buffer to be reclaimed. In the case of a transmission failure (for
example, the maximum collision was exceeded), the Large Send session is not terminated and the
transmit operation proceeds with the next Large Send packet.
Note: The stack is responsible for maintaining the TCP sequence number and IP identification integrity.
Note: The 82550 is configured to use dynamic TBDs for Large Send since TBDs are read from host
memory in pairs and the number of TBDs is not known in advance. When it is configured in this
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
165
Download from Www.Somanuals.com. All Manuals Search And Download.
82550 and 82551QM Specific Information
way, it may read an extra TBD (or 8 bytes) after the last valid TBD. The driver is responsible for
allocating enough memory for the TBD array. Otherwise, it needs to ensure that an extra 8 byte
read access from the PCI will not affect anything.
B.3.4.5
Performance Considerations
For better performance, the maximum TCP payload should be a multiple of 4. It is recommended
to use the largest possible value for this parameter.
B.3.5
Features Co-existence
B.3.5.1
Large Send and Checksum
Large Send and checksum are orthogonal.
B.3.5.2
B.3.5.3
Large Send and Software Parsing
Large Send cannot be used with software parsing.
Large Send and Scheduling Assist
All frames of one Large Send instance are subject to the same scheduling policy. Specifically, if the
prototype has a time value for transmission, all subsequent frames are considered ready for
transmission after that time and may be transmitted back-to-back.
B.4
RCV Checksum Processing
The 82550 and 82551 compute and verify the IP header checksum and the TCP/UDP checksum on
frames successfully parsed. It indicates a checksum match or mismatch in the RFD. An incoming
UDP frame with a checksum field of 0 is treated as a checksum match. If the 82550 or 82551
cannot decrypt a frame, the checksum can only be verified to the first IP header only (relevant to
types IV, V, and VI).
B.4.1
Data Flow
In checksum modes, the device checks incoming frames for format correctness. If the check fails, a
receive parser failure occurs and the frame is passed as is. The incoming frame is subject to all
MAC layer checks (for example, CRC, minimum length, address filtering). If the value of the word
following the DA and SA fields is less than 5DCh (1500), the controller assumes it is an Ethernet
v2 frame. If the word value is greater than or equal to 5DCh (1500), the 82550 and 82551 compare
in the incoming frame format to SNAP.
In the 82559 compatibility mode, incoming frames are not verified for their format. The checksum
is calculated on all non-MAC frame bytes. Supported frame formats are Ethernet v2 and SNAP.
166
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
Download from Www.Somanuals.com. All Manuals Search And Download.
82550 and 82551QM Specific Information
B.4.1.1
B.4.1.2
Frame Types
• Ethernet v2. If the Ethernet v2 type field equals 0800h, the first byte of IP header is expected
right after the optional VLAN field.
• SNAP. The 82550 and 82551 skip the DSAP and SSAP fields. The Control field is expected to
have a value of 03h. If a match occurs, the first byte of IP header is expected right after the
optional VLAN field.
Verification Types
• IPv4 header checks. For IPv4 frames, no fragmentation occurs. The protocol field should
equal:
— 06 for TCP
— 17 for UDP
— 04 for IP
— 50 for ESP
— 51 for AH
Other protocols will not have a checksum.
• TCP header checks. The packet length should be greater than or equal to the TCP header.
• UDP header checks. The packet length should be greater than or equal to the UDP header.
• Checksum. The receive checksum machine computes the IP header checksum, the tunneled IP
header checksum (if one is present), the TCP/UDP pseudo header checksum, and the TCP/
UDP checksum. If a mismatch occurs, the frame is transferred to memory with an error
indication according to the configuration.
B.4.2
82559 Compatibility
In 82559 compatibility mode, the 82550 and 82551 compute the checksum of the whole incoming
packet excluding the MAC header and CRC. The checksum word is appended at the end of the
packet in the receive buffer in host memory. If the CRC is posted to host memory, the checksum
word follows the CRC. The byte count field in the receive memory structures includes the
checksum word.
B.5
VLAN Tagging
When the 82550 or 82551 receives a frame with a VLAN tag, it sets bit T in the RFD status word
(bytes 1h:0h) and copies the tag (the two bytes after the constant 8100h) to the VLAN field in the
RFD (bytes 11h:10h). If VLAN stripping is enabled, the device strips the four VLAN bytes, and
the received packet is copied to host memory without these four bytes.
10/100 Mbps Ethernet Controller Family Open Source Software Developer Manual
167
Download from Www.Somanuals.com. All Manuals Search And Download.
|