Upload
tyrone-cox
View
216
Download
1
Embed Size (px)
Citation preview
SBP/2 over IEEE-1394and future 1394 specifications
Contact Info:
Christopher Lee
University of New Hampshire InterOperability Lab - 1394 Consortium
UNH InterOperability Lab - [email protected]
2
What this presentation is…• A general technical overview of the Serial Bus Protocol 2• Questions that this presentation attempts to answer
– What is SBP/2?– Why is it needed? – What problems does it solve?– How does it work?– How is it implemented?
UNH InterOperability Lab - [email protected]
3
What is SBP/2?
• The Serial Bus Protocol 2 was created by the T10 SCSI group. The group noticed the trend of the computer industry moving to faster, smaller, cheaper. Standard SCSI interfaces were bulky, expensive, and difficult to manage.
• Termination was a large problem, as was the “fat” cables used by SCSI technology.
UNH InterOperability Lab - [email protected]
4
What is SBP/2? (cont)
• A parallel bus like SCSI is great for single peer to peer connections, but serial buses are better for peer to peer connections with multiple devices
• SBP/2 did not start originally for 1394, but modifications have made 1394 implementation anything but perfect
UNH InterOperability Lab - [email protected]
5
Why SBP/2?
• 1394 needed a protocol for mass data storage. There are no defined command sets or protocols for 1394 so a generic data transfer command set was designed similar to SCSI.
• SBP/2 is to allow different devices to exchange data and commands between each other.
UNH InterOperability Lab - [email protected]
6
Why SBP/2? (cont)
• SBP/2 can be used for: – hard drives, – magneto optical (MO) drives, – digital cameras, – CD-ROMs, – DVD-ROMs, – mass storage devices, – any device that can implement SCSI commands
UNH InterOperability Lab - [email protected]
7
What problems does it solve?• Ability of devices to identify themselves without
prior knowledge
• To transfer data and commands abstractly, (encapsulated SCSI commands)
• Generic packet set of 1394 makes implementation of different protocols easier
• Enables devices to form large command task sets without transferring set beforehand
• Some security features via password login
UNH InterOperability Lab - [email protected]
8
How does SBP/2 work?• Each capability of a node is designated with
a logical unit number (LUN)– nodes may have up to eight (8) LUNs
• The commands for the target are kept in system memory of the initiator– the target uses a read request block to fetch the
command (ORB)– the target uses a field in the ORB to determine
the direction of information flow
UNH InterOperability Lab - [email protected]
9
Command Implementation
• All SBP/2 commands use the standard asynchronous packet transfers of read/write quadlet request/response, and read/write block request/response
UNH InterOperability Lab - [email protected]
10
Initiators and targets...
• A transfer involves only one initiator, but can involve more than one target.
• All commands and data are encapsulated within a packet known as an operation request block (ORB)
UNH InterOperability Lab - [email protected]
11
Address Pointers and Page Tables
– SBP/2 standard address pointer
» Last two bits not used, offset is higher
– When a data block is fragmented, a page table may be utilized
UNH InterOperability Lab - [email protected]
12
ORBs• There are several different types of ORBs
– Management ORBs• responsible for administration commands, such as
login, logout, password management, and target lists
– Dummy ORB• blank ORBs used as a placeholder in a linked list
– Command Block ORB• actual command ORBs, encapsulated SCSI
commands
UNH InterOperability Lab - [email protected]
13
ORBs (cont)
• All ORBs have the following fields in common– Notify bit
• advises the target whether or not completion notification is necessary
– rq_fmt field• specifies what type of ORB it is
UNH InterOperability Lab - [email protected]
14
ORB Format
• Four ORBtype dependentfields
• Notify bit ->
• Last portionof field can -> vary in size
UNH InterOperability Lab - [email protected]
15
ORB Linked Lists
• One of the features of SBP/2 is the ability for ORBs to be formed into a linked list– Each ORB has a field that gives the address of
the next ORB• A target may fetch several ORBs and depending
upon the node’s capabilities execute the commands in order or it may optimize the commands to make operation more efficient
UNH InterOperability Lab - [email protected]
17
Management ORBs
• Cannot be linked together
• Fields are function dependent• login
• query logins
• reconnect
• set password
• logout
• abort task
• target reset
UNH InterOperability Lab - [email protected]
18
Management ORB Format
– Notify set to 1 ->
– Status block ->
UNH InterOperability Lab - [email protected]
19
Dummy ORBs• Serve as placeholders in a command ORB
linked list– can be used to initialize a target fetch agent
Next_ORB pointer ->
UNH InterOperability Lab - [email protected]
20
Command ORBs• Used to encapsulate data transfer or device
control commands
• All command ORBs have a next_ORB field– This field permits ORBs to be put in a linked list
• Contains a field indicating the size of the data block to be transferred
• Direction bit– Used to determine whether it needs to read or write
UNH InterOperability Lab - [email protected]
21
Command Block ORB Format
Next_ORB pointer ->
Contains address of data buffer for command in -> command_block
Encapsulated SCSI command ->
UNH InterOperability Lab - [email protected]
22
ORB Implementation• ORBs (commands) are stored on the initiator in
system memory
• Each node utilizes a doorbell register– The doorbell register is used to indicate to the target that
the initiator has added ORBs to a suspended linked list• any value written to the doorbell indicates the doorbell has been
“rung”
• Each node has a next_ORB register– An ORBs address is written to this register prior to
“ringing the doorbell”
UNH InterOperability Lab - [email protected]
23
Login
• An initator must login to a target before using a node’s resources– A password verification must take place– Using a login procedure, addressing issues in a
dynamic environment are solved– Discovery of a node’s capabilities take place
during login
UNH InterOperability Lab - [email protected]
24
Login ORB Format
» Password ->
» Address for response block ->
» Notify, reconnect, and LUN ->
» Password and response length ->
» Status block address ->
UNH InterOperability Lab - [email protected]
25
Status Blocks• A login ORB specifies the address in
system memory that the target will use to store the status of command execution and/or completion
The status response is dependent on the command ->
UNH InterOperability Lab - [email protected]
26
Passwords• Each node has a set of two passwords
– A master password • cannot be changed
• stored in static memory
• can only be read when already logged in
– Secondary password• can be changed
• should be noted on the end product
UNH InterOperability Lab - [email protected]
27
Change Password Management ORB Format
• Changes secondary password– Uses login ID for verification
New password ->
Login_id of currently logged in initiator ->
Status block address ->
UNH InterOperability Lab - [email protected]
28
Reconnect
• The login ORB specifies a reconnect time– If disconnected, the initiator attempts to
reconnect within the time-out period with a reconnect ORB
• The reconnect ORB specifies the address of a status block that the reconnecting target writes to
UNH InterOperability Lab - [email protected]
29
Reconnect ORB Format
<- Login_id that node received from login process
^ Status block address ^
UNH InterOperability Lab - [email protected]
30
Logout• When an initator no longer requires access
to a target’s resources, it shall signal a logout request
• Upon a successful logout, all resources are released
Status block address ->
UNH InterOperability Lab - [email protected]
31
End Notes
• Current Microsoft implementation of SBP/2– Uses a null password for login– Hard drives stay logged in– Printers log out after a print job– Notify is set at all times