Functional Description
Block Write–Block Read Process Call
The block write-block read process call is a two-part message. The call begins with a
slave address and a write condition. After the command code the host issues a write
byte count (M) that describes how many more bytes will be written in the first part of
the message. If a master has 6 bytes to send, the byte count field will have the value 6
(0000 0110b), followed by the 6 bytes of data. The write byte count (M) cannot be 0.
The second part of the message is a block of read data beginning with a repeated start
condition followed by the slave address and a Read bit. The next byte is the read byte
count (N), which may differ from the write byte count (M). The read byte count (N)
cannot be 0.
The combined data payload must not exceed 32 bytes. The byte length restrictions of
this process call are summarized as follows:
• M ≥ 1 byte
• N ≥ 1 byte
• M + N ≤ 32 bytes
The read byte count does not include the PEC byte. The PEC is computed on the total
message beginning with the first slave address and using the normal PEC
computational rules. It is highly recommended that a PEC byte be used with the Block
Write-Block Read Process Call. Software must do a read to the command register
(offset 2h) to reset the 32 byte buffer pointer prior to reading the block data register.
Note that there is no STOP condition before the repeated START condition, and that a
NACK signifies the end of the read transfer.
Note:
E32B bit in the Auxiliary Control register must be set when using this protocol.
See section 5.5.8 of the System Management Bus (SMBus) Specification, Version 2.0
for the format of the protocol.
5.20.2
Bus Arbitration
Several masters may attempt to get on the bus at the same time by driving the
SMBDATA line low to signal a start condition. The ICH10 continuously monitors the
SMBDATA line. When the ICH10 is attempting to drive the bus to a 1 by letting go of
the SMBDATA line, and it samples SMBDATA low, then some other master is driving the
bus and the ICH10 will stop transferring data.
If the ICH10 sees that it has lost arbitration, the condition is called a collision. The
ICH10 will set the BUS_ERR bit in the Host Status Register, and if enabled, generate an
interrupt or SMI#. The processor is responsible for restarting the transaction.
When the ICH10 is a SMBus master, it drives the clock. When the ICH10 is sending
address or command as an SMBus master, or data bytes as a master on writes, it drives
data relative to the clock it is also driving. It will not start toggling the clock until the
start or stop condition meets proper setup and hold time. The ICH10 will also ensure
minimum time between SMBus transactions as a master.
Note:
The ICH10 supports the same arbitration protocol for both the SMBus and the System
Management (SMLINK) interfaces.
Datasheet
219
INTEL [ INTEL ]
