Deal with busy devices

Author: IvanVoras FreeBSD

Reviewer: name contact BSD flavour

Concept

Understand what can cause a process to hang, how to detect related processes, and how to fix the situation.

Introduction

Occasionally there are situations where a userland process can completely hang due to kernel being stuck in the I/O path. The most common way this could happen is by using mount(8) and umount(8) tools on unreliable and removable media (like USB and CD drives) and network devices (like smbfs). In the general case, there's no reliable way to fix this situation except a reboot, but sometimes the system utilities can kill the stuck process and allow most of the system to continue running.

The first step is to identify the process that's stuck. In most cases, this is obvious since it's the process the user has just started or is currently interfacing with. But even in this case, it's useful to find out more information about the process for future reference and to help debug the problem. A good starting point is the ps(1) utility. The command:

ps axl

will display a table containing (a)ll processes, even those without an attached terminal (x), together with (l)ock channels. The last information is the same as the "STATE" field shown by the top(8) utility.

There are no fixed examples of which states inicate a "hung" process, but any debugging information submitted to developers should include the output of the above command for it, especially the MWCHAN field.

TODO: maybe cross-reference with View and configure system hardware?

fstat: identify active files

umount: unmount file systems

lsof: list open files

Examples

Practice Exercises

More information

ps(1), fstat(1), kill(1), umount(8) and the third-party lsof utility