Intel C2518 FH8065501516710 Data Sheet

Product codes
FH8065501516710
Page of 746
Intel
®
 Atom™ Processor C2000 Product Family for Microserver
September 2014
Datasheet, Vol. 2 of 3
Order Number: 330061-002US
239
Volume 2—PCI Express Root Ports (RP)—C2000 Product Family
PCI Express RAS Features
12.9
PCI Express RAS Features
The PCI Express Base Specification, Revision 2.1 defines a standard set of error 
reporting mechanisms. The root ports supports all of them including Error Poisoning 
and Advanced Error Reporting (AER).
The root ports also support:
• Link-Level Cyclical Redundancy Code (LCRC) and Retry Management as described 
in Section 3.5. Data Integrity, of the specification.
• Dynamic Link Width (DLW) reduction on link failure. See Section 4.2.4. Link 
Initialization and Training, of the specification.
12.9.1
Error Detecting, Reporting and Logging
The PCI Express Root Ports provide error detection and logging as specified in the 
aforementioned specification.
Error messages from the PCI Express Functions or Devices in the hierarchy associated 
with the root port are detected. The root port, if enabled to do so, reports an interrupt 
to the CPU. The three classes of errors are:
• Fatal Error (ERR_FATAL Error Message)
Uncorrectable error conditions which render the particular link and related 
hardware unreliable. For fatal errors, a reset of the components on the link is 
required to return to reliable operation.
• Non-Fatal Error (ERR_NONFATAL Error Message)
Uncorrectable errors which cause a particular transaction to be unreliable but 
the link is otherwise fully functional. Isolating non-fatal from fatal errors 
provides requester/receiver logic in a device or system management software 
the opportunity to recover from the error without resetting the components on 
the link and disturbing other transactions in progress. Devices not associated 
with the transaction in error are not impacted by the error.
• Correctable Error (ERR_COR Error Message)
Correctable errors include those error conditions where the hardware recovers 
without any loss of information. The hardware corrects these errors and 
software intervention is not required.
Each error class generates an interrupt to the CPU if the corresponding error class 
Reporting Enable is set in the Root Port Root Error Command register. This register is 
located in the configuration space of the root port in its PCI Express Advanced Error 
Reporting (AER) Extended Capability Structure.
The root port itself as a PCI Express Device also generates the three classes of error 
messages to the root complex if the particular error class is enabled in the Root Port 
Device Command register in its PCI Express Capability Registers in Configuration 
Space.
Another way to enable error reporting by the root port is the Root Port PCI Command 
register contains SERR# Enable (SEE). When set, this bit enables reporting of non-fatal 
and fatal errors detected by root port to the root complex. This bit also controls 
transmission by the root port of ERR_NONFATAL and ERR_FATAL error messages 
forwarded from the downstream interface. ERR_COR messages are not affected by this 
bit.
Devices and Functions in the downstream hierarchy that support the AER Capability, 
and the root ports themselves as devices, also have an Uncorrectable Error Mask 
register and a Correctable Error Mask register allows each error condition to be masked 
independently.