Intel
®
81341 and 81342 I/O Processors
December 2007
Developer’s Manual
Order Number: 315037-002US
109
Address Translation Unit (PCI-X)—Intel
®
81341 and 81342
2.7.3.9
Inbound Configuration Write Request
As a target, the ATU may encounter this error when operating in the Conventional or
PCI-X modes.
2.7.3.9.1 Conventional PCI Mode
To allow for correct data parity calculations for delayed write transactions, the ATU
delays the assertion of
STOP#
(signalling a Retry) until
PAR
is driven by the master. A
parity error during a delayed write transaction (inbound configuration write cycle) can
occur in any of the following parts of the transactions:
• During the initial Delayed Write Request cycle on the PCI bus when the ATU latches
the address/command and data for delayed delivery to the internal configuration
register.
• During the Delayed Write Completion cycle on the PCI bus when the ATU delivers
the status of the operation back to the original master.
The 81341 and 81342 ATU PCI interface has the following responses to a delayed write
parity error for inbound transactions during Delayed Write Request cycles with the
given constraints:
• When the Parity Error Response bit in the ATUCMD is set, the ATU asserts
TRDY#
(disconnects with data) and two clock cycles later asserts
PERR#
notifying the
initiator of the parity error. The delayed write cycle is not enqueued and forwarded
to the internal bus.
• When the Parity Error Response bit in the ATUCMD is cleared, the ATU retries the
transaction by asserting
STOP#
and enqueues the Delayed Write Request cycle to
be forwarded to the internal bus.
PERR#
is not asserted.
• The Detected Parity Error bit in the ATUSR is set. When the ATU sets this bit,
additional actions are taken:
— When the ATU Detected Parity Error Interrupt Mask bit in the ATUIMR is clear,
set the Detected Parity Error bit in the ATUISR. When set, no action.
• For PCI-X Mode 2, update the
“ECC Control and Status Register - ECCCSR” on
, the
“ECC First Address Register - ECCFAR” on page 203
, the
Second Address Register - ECCSAR” on page 204
for the transaction.
For the original write transaction to be completed, the initiator retries the transaction
on the PCI bus and the ATU returns the status from the internal bus, completing the
transaction.
For the Delayed Write Completion transaction on the PCI bus where a data parity error
occurs and therefore does not agree with the status being returned from the internal
bus (i.e. status being returned is normal completion) the ATU performs the following
actions with the given constraints:
• When the Parity Error Response Bit is set in the ATUCMD, the ATU asserts
TRDY#
(disconnects with data) and two clocks later asserts
PERR#
. The Delayed
Completion cycle in the IDWQ remains since the data of retried command did not
match the data within the queue.
• The Detected Parity Error bit in the ATUSR is set. When the ATU sets this bit,
additional actions are taken:
— When the ATU Detected Parity Error Interrupt Mask bit in the ATUIMR is clear,
set the Detected Parity Error bit in the ATUISR. When set, no action.