1. WHEA System Design And Implementation
4/20/2017 6:42 AM
WHEA System Design And Implementation
John Strange
Software Design Engineer
Microsoft Corporation
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

2. 4/20/2017 6:42 AM
Key Takeaways
Understand what it takes to implement a WHEA-enabled platform
Improve server reliability by implementing required WHEA features
Differentiate server products by extending WHEA capabilities
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

3. Agenda WHEA Overview Description of a WHEA-enabled platform
4/20/2017 6:42 AM
Agenda
WHEA Overview
Description of a WHEA-enabled platform
Key requirements of WHEA- enabled platform
Dell WHEA system implementation
Dell demo of WHEA-enabled platform
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

4. 4/20/2017 6:42 AM
WHEA Objective
To make Windows-based platforms more resilient in the face of hardware errors
Better root cause analysis
Better support for hardware error recovery
Error avoidance with health monitoring
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

5. WHEA Overview Operating System Support (OS)
4/20/2017 6:42 AM
WHEA Overview Operating System Support (OS)
Windows is much more agile with respect to hardware error sources
Platform describes error sources to Windows
Standardized hardware error record format
UEFI 2.1 Common Platform Error Record
Hardware error recovery support
Hardware error events
Event Tracing For Windows (ETW)
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

6. WHEA Overview Platform integration
4/20/2017 6:42 AM
WHEA Overview Platform integration
Platforms retain their existing investment in error handling features
Controls the level of integration with the OS
Leverages existing error handling and reporting features
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

7. WHEA Overview Platform implementation
4/20/2017 6:42 AM
WHEA Overview Platform implementation
Satisfying Windows Server 2008 logo requirements
Error record persistence
Error injection
WHEA _OSC method
BOOT Error Source (X86/X64 platforms only)
Extending WHEA feature set to add value
Add richer error data content (i.e. FRU info)
Participate in error recovery
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

8. WHEA Components WHEA-Enabled Applications User Kernel
WHEA-Enabled Management Applications
Provided by:
Microsoft
User
ISV
Kernel
OS Hardware Error Handlers
IHV
OEM
Platform Specific Hardware Error Driver
PSHED Plug-ins
Platform Hardware Error Handlers
WHEA ACPI Tables
WHEA ACPI Tables

9. WHEA-Enabled Platform
Feature/
Processor Architecture
x86
x64
Itanium
Error Source Enumeration
Optional: HEST or PSHED plug-in
Error Record Persistence
Required: ERST or PSHED plug-in
Required: ERST, PSHED plug-in, or UEFI 2.1 variables services
BOOT Error Source
Required
Optional
Error Injection
Required: EINJ or PSHED plug-in
Optional if PAL-based or MSR-based error injection is supported
Error Information Retrieval
Optional: PSHED plug-in
Error Source Control
Error Recovery
_OSC

10. Reporting Error Sources
4/20/2017 6:42 AM
Reporting Error Sources
The platform must report error sources to the Windows only for the following
To override default error source configuration
To report error sources Windows does not support by default
It needs firmware-first control of one or more error sources
It uses generic error source to inject errors
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

11. Error Source Defaults x86/x64 Machine check
4/20/2017 6:42 AM
Error Source Defaults x86/x64 Machine check
x86/x64 Machine Check Settings
IA32_MCG_CTL: 0xFFFFFFFFFFFFFFFF
IA32_MCi_CTL: 0xFFFFFFFFFFFFFFFF
OS respects settings in IA32_MC0_CTL
x86/x64 Corrected Machine Checks
Polling interval is 60 seconds
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

12. Error Source Defaults PCI Express AER
4/20/2017 6:42 AM
Error Source Defaults PCI Express AER
Endpoint Devices
Default Value
Device Control
0x0007
Uncorrectable Error Mask 0x
Uncorrectable Error Severity 0x
Correctable Error Mask 0x
Capabilities and Control 0x
Root Ports
Default Value
Root Error Command
0x0007
Bridges
Default Value
Secondary Uncorrectable Error Mask
0x000017A8
Secondary Uncorrectable Error Severity 0x
Secondary Capabilities and Control 0x
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

13. Error Record Persistence
4/20/2017 6:42 AM
Error Record Persistence
In Windows Server 2008
Windows writes error record only when system is to be bugchecked
Windows only requires space for one error record
Platform must implement persistence interface to get logo
Storage requirements
x64/x86 platforms require minimum of 1K
Itanium platforms require minimum of 128K
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

14. Error Record Persistence Platform implementation
4/20/2017 6:42 AM
Error Record Persistence Platform implementation
ACPI ERST Table
UEFI 2.1 Variable Services Error Record Extensions for EFI-based platforms
PSHED plug-in
This solution is generally discouraged
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

15. 4/20/2017 6:42 AM
Error Injection
Error injection interface allows hardware errors to be injected on a platform for the following purposes
Validation of OS/platform error handling flows
Validation of platform logo support for WHEA
Exercising hardware/firmware error flows for diagnostic purposes
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

16. Error Injection Platform implementation
4/20/2017 6:42 AM
Error Injection Platform implementation
Prefer true hardware error injection if possible
Enables system/component diagnostic
In cases where no true hardware injection is possible, generic error source can be used to simulate errors
Enables feature validation
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

17. WHEA _OSC Method New \_SB _OSC method
4/20/2017 6:42 AM
WHEA _OSC Method
New \_SB _OSC method
GUID {ed855e0c-6c90-47bf-a62a- 26de0fc5ad5c}
Notifies platform that Windows implements WHEA so platform can perform any necessary configuration
If platform does not implement \_SB _OSC or if the platform returns “Unrecognized UUID”, Windows does not configure WHEA support for the platform
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

18. Boot Error Source For fatal errors that cannot be processed by the OS
4/20/2017 6:42 AM
Boot Error Source
For fatal errors that cannot be processed by the OS
Firmware-initiated reset
BMC-initiated reset
Sync-flood reset
Platform describes the error to Windows using the BOOT error source
ACPI BERT tables describes the platform’s BOOT error source to Windows
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

19. Firmware First
Platform can indicate that error sources should be handled first by firmware
Via error source enumeration interface
Some error sources cannot do firmware- first (i.e. machine check exception)
Generally, an error source reported as firmware-first is configured by the platform to generate an SMI

20. Firmware First Enumerating error sources
The error source for which platform wants firmware-first control is marked as FIRMWARE_FIRST
A paired generic error source must be enumerated
This error source is how the platform will signal errors from the firmware-first source to the OS

21. Firmware First Error handling flow
Platform gains control when error occurs (SMI)
Platform processes and possibly logs the error
Platform may void errors in some cases
Platform fills in error status block with information describing the error
Platform is responsible for clearing HW error status

22. Firmware First Error handling flow
Platform signals the error to Windows using the notification mechanism it reported when it enumerated the error source
This means platform generates an NMI, interrupt, or allows Windows to poll, etc
Signaling mechanism depends on type of error (i.e. corrected/uncorrected)
Windows clears bits in block status to signal that it has processed the error

23. Dell WHEA System Design
4/20/2017 6:42 AM
Dell WHEA System Design
Mukund Khatri
Server Strategist
Dell Inc.
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION. 23

24. 4/20/2017 6:42 AM
Overview
Close collaboration between Dell and Microsoft on WHEA feature design over last couple of years
Design enhancements
Prototype efforts
WHEA architecture holds great promise for future server designs
Enables OS participation in error handling flows
Flexibility to retain full value in existing error handling infrastructure
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION. 24

25. WHEA: Dell Implementation
4/20/2017 6:42 AM
WHEA: Dell Implementation
Dell Implementation incorporates support for
Error Enumeration and Control
Error record persistence
WHEA _OSC method
BOOT Error Source
Error injection
Implementation uses Firmware First Mode
Complementary to OS-first mode
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION. 25

26. Firmware First Mode Implementation considerations
4/20/2017 6:42 AM
Firmware First Mode Implementation considerations
Ability at platform firmware level to override defaults in OS without PSHED plug-ins
Silicon errata management
Updates to interface specifications
Control over level of integration with OS
Extend WHEA feature set to add value
Add richer error data content (ex: FRU info)
Retain existing investments in error handling infrastructures
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION. 26

27. Platform Hardware Event Flow Firmware First Mode
New for ecosystem consumption
Errors handled by OS
Platform Errors
ETW
Data
Errors handled by
Platform Firmware
Service Processor and
Management Consoles
Error Flow with:
No WHEA
Existing error management paradigm still retained
Richer error records and ETW available for consumption
New with WHEA

28. Dell PowerEdge™ System
4/20/2017 6:42 AM
demo
Dell PowerEdge™ System
Mukund Khatri
Server Strategist
Dell Inc.
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

29. Demo: Dell PowerEdge™ Server
4/20/2017 6:42 AM
Demo: Dell PowerEdge™ Server
Injection of PCI-Express uncorrectable error
Error captured and processed by platform firmware
Firmware creates and uploads GES data packet and triggers NMI to OS
WHEA error record stored in persistent storage
System bug-checks and subsequently reboots
OS retrieves WHEA error record on next boot
Event viewer reports the event along with error record
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION. 29

30. PCI-Express Error Record Output

31. 4/20/2017 6:42 AM
Dell: Key Takeaways
New Dell servers will include full support for WHEA
We intend to build on WHEA architecture to add end customer value in future Dell servers
Dell and Microsoft partnering on WHEA architecture and implementation
© 2006 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION. 31

32. Call To Action WHEA-enable your server platforms now
Work with Microsoft to get BIOS reference implementations
Validate WHEA support
Run Logo Tests to validate WHEA implementation
Fully implement and validate Advanced Error Reporting capability in PCI-express devices

33. Additional Resources Related Sessions WHEA Feedback: Specifications
SVR-T464 WHEA Platform Implementation
SVR-C460 WHEA PSHED Plug-in
SVR-T325 Dynamic Partition: Windows Server
WHEA Feedback:
WHEA introduction:
Specifications
WHEA Platform Design Guide
UEFI 2.1 Specification
microsoft.com

34. © 2007 Microsoft Corporation. All rights reserved
© 2007 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.