Profiling and Tracing

Profiling helps find code bottlenecks. Trace tools provide the ability to extract and interpret details about code execution.

Sreekrishnan Venkateswaran

Monday, August 15th, 2005

Share This:

Last month’s column showed you how to use debuggers to debug kernel code. This month, let’s look a bunch of other tools you should have in your Linux kernel debugging repertoire.

Kernel Profiling using OProfile

A profiler identifies those portions of code that burn more CPU cycles. Profilers come in different flavors. The OProfile kernel profiler uses hardware assist to gather profile data, while the gprof application profiler relies on compiler assist to collect profile information.

The Right Tool for the Job

Linux supports profilers, trace tools, serviceability mechanisms, and code exercisers. You can use any or a combination of these tools to suit your debugging needs.

The OProfile kernel profiler helps sense the presence of code bottlenecks.

The Linux Trace Toolkit (LTT) extracts execution traces that are particularly useful for postmortem analysis.

Linux Kernel Crash Dump (LKCD) and kprobes aid serviceability if your kernel crashes at a customer site where you cannot use a debugger.

Exercisers, such as the Linux Test Project (LTP) measure the reliability and robustness of your kernel port.

Configuration options under Kernel Hacking turn on debug code that help detect problems like memory poisoning and spinlock errors.

User Mode Linux (UML) lets you debug some sections of the kernel using a user-mode debugger without oops ing the kernel.

OProfile is included in the 2.6 kernel. (If you’re using a 2.4 kernel, you will need to patch your sources with…

Please log in to view this content.

Not Yet a Member?

Hands-on Content
White Papers
Community Features
And more.

Already a Member?
Log in!

Forgotten your password?
Forgotten your username?