Logging method for high execution frequency paths of Linux kernel

Abstract

Understanding Operating System behavior is very critical for any embedded designer to make informed design decision. We present a new logging method which can capture the high granular details of the kernel activity. It reduces the logging latency by 95-97% & logging memory usage by 70% compared to conventional “printk”. We utilize the string literal pool of the Linux kernel to reconstruct the log offline, and store only the parameter values passed to a printk function, instead of current method putting the log as string after printk formatting.