A few notes on testing

Matthew Burgess matthew at linuxfromscratch.org
Tue Jul 13 14:49:07 PDT 2004


On Tue, 13 Jul 2004 13:28:48 -0700
"Kevin P. Fleming" <kpfleming at linuxfromscratch.org> wrote:

> Matthew Burgess wrote:
> 
> > But I thought that /sys (as the mount point for sysfs) will only be
> > populated by the information exported via the device drivers.
> 
> Device drivers, bus drivers, filesystems, device "class" drivers, etc.
> 
> Remember that the PCI bus code is itself a "driver", and exports its 
> information into sysfs as well, as does the USB bus driver(s) and the 
> IEEE-1394 driver(s).
> 
> Even with no actual _device_ drivers loaded, all available PCI devices
> 
> are listed in /sys/bus/pci with their PCI enumeration information. The
> 
> hotplug scripts can then use this to determine which modules to load,
> if any are needed.

Ah OK then.  So as long as [bus] support is compiled in directly (i.e.
not as a module) there shouldn't be a problem.  I wonder whether then,
along with /proc/bus/* being deprecated the options to compile bus
drivers in as modules will also be removed.  This at least would stop
people getting into the chicken-and-egg problem I think.  Why one would
even want/need to compile such stuff in as a module is beyond me, but
then I'm particularly module-averse anyway.

Thanks a lot for the explanations,

Matt.



More information about the lfs-dev mailing list