DPU Function Detection and BIOS Exclusion to Prevent Host Boot Loops
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In virtualized environments, issues with data processing unit (DPU) functions can cause the host device operating system to crash or fail to operate, leading to reboot loops, with existing techniques lacking the ability to identify and hot-remove problematic functions dynamically.
Innovation Solution
A platform with a detection engine identifies and hot-removes problematic functions by collecting crash dump logs, using a baseboard management controller (BMC) to communicate with the BIOS, and storing problematic function details in non-volatile memory to exclude them from the booting process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If DPU functions are offloaded to accelerate system performance, then network communication speed and host device performance are improved, but system reliability deteriorates due to potential DPU function failures causing host OS crashes
Solution Approach 1:
The system performs preliminary identification of problematic DPU functions during the boot process by collecting crash dump logs and analyzing them to determine which functions caused previous failures. The BIOS then proactively excludes these identified problematic functions from configuration before the host OS attempts to initialize them, preventing crashes before they occur while still allowing safe DPU functions to operate for accelerated performance
Solution Approach 2:
The BIOS acts as an intermediary layer between the DPU functions and the host operating system. It intercepts the function configuration process, filters out problematic functions based on crash dump analysis, and only passes safe functions to the host OS. This intermediary mechanism enables the system to maintain high productivity by allowing most DPU functions to operate while blocking only the specific problematic ones that would cause reliability issues
2Adaptability or versatility
If traditional reboot loops occur after DPU function failures, then system self-correction is attempted, but loss of time occurs due to repeated boot failures and inability to operate
Solution Approach 1:
Instead of waiting for repeated boot failures to occur, the system performs the corrective action in advance by identifying problematic functions during the initial boot process and excluding them from configuration. This preliminary correction prevents the time loss associated with multiple reboot loops, as the system configures only safe functions from the start, achieving adaptability without the penalty of repeated failures
Solution Approach 2:
The system implements feedback by collecting crash dump logs that provide information about which DPU functions caused failures. This feedback loop allows the BIOS to learn from previous failures and automatically adjust the function configuration by excluding problematic functions, enabling rapid adaptation without manual intervention or repeated reboot cycles
Data Source
AI summary
A method comprises executing a booting operation comprising booting of an operating system of a host device and booting of an operating system of a data processing unit running on the host device, and detecting a failure of at least the booting of the operating system of the data processing unit. Execution of the booting operation is paused in response to the detecting, and data corresponding to the failure is collected. At least one function associated with the data processing unit that is contributing to the failure is identified based on the collected data. A basic input/output system of the host device is provided with access to identifying information for the at least one function. The booting operation is re-executed, wherein the basic input/output system excludes the at least one function from being configured by the operating system of the host device based on the identifying information.