Kernel page, once again

Kevin P. Fleming kpfleming at linuxfromscratch.org
Tue Jun 15 21:47:58 PDT 2004


Jeremy Utley wrote:

> Because, unless I'm mistaken, proper functioning of Hotplug requires 
> that these devices *BE* modules, so they can be loaded when the device 
> is connected, and unloaded when the device is removed.

This is not at all true. I use plenty of hotplug devices here, and none 
of my kernels use modules. If the module is loaded, it will be bound to 
the device when the device is plugged in, and unbound when the device is 
unplugged. If the driver requires special parameters when it is loaded, 
and has not been upgraded to accept those parameters via the kernel 
command line, that is a problem. Also, if the driver is not shipped with 
the kernel, again that is a problem. However, these are not the majority 
of the drivers people are going to be dealing with, especially for 
hotpluggable hardware.

Module unloading is not even available in the 2.6 kernels unless you 
used the "advanced" method to turn it on in the kernel configuration, 
and even then it's not intended to be used in this way any longer, as 
best I can tell. The benefits of unloading these driver modules are less 
certain anyway, given that they do not consume a large amount of memory 
to begin with.

I'm not saying that using kernel modules has no place: it certainly 
does. However, it's less useful than it used to be, and the kernel 
developers are pushing people away from using them for everyday 
operation with "known" hardware, even if that hardware is plugged in and 
removed during normal operation.



More information about the lfs-dev mailing list