Re: some simple questions about sysfs

[Greg KH <greg@xxxxxxxxx>]

On Thu, Mar 01, 2007 at 02:56:15PM -0500, Robert P. J. Day wrote:

Simple?  Are you crazy?  This is one of the most complex, undocumented,
mess that we have in the kernel right now.  Everytime I dig down into
it, I try to get out as soon as possible and I forget what I learned...

>   i'm trying to clarify the interconnection between all of the sysfs
> structures, so a few questions:
> 
> 1) is it true that an entry exists at the top of the sysfs if and only
> if it is a subsystem?  there's no notion of a lower-level subsystem,
> is there?  that is, subsystems aren't defined recursively.

No, you can create an entry at the top if you use a kobject with no
parent.

And you can make a subsystem within another subsystem by setting the
parent like the pci_hotplug subsystem does it.

> 2) as i read it, a subsystem contains only the *default* kset of that
> subsystem.  so how can one (if it's possible), given a subsystem,
> generate a list of *all* of the subsystem's kset children?  (i'm
> assuming the "list" member structure in a kset is for the list of
> kobject children of that kset, right?)

Why would you want to do that?  But yes, I think you can just walk that
list.

> 3) as i read it, ksets are also not recursive, so the entire sysfs
> structure consists solely of subsystems, which contain ksets, which
> contain kobjects, and that's it, right?  ksets can't contain other
> ksets.

I think so.  Take a look at the lwn.net articles about this for a bit
more description of what is going on.

If you want to clean this whole mess up, I'm all for it and I'll buy you
a case of your favorite beer.  It really needs it as you are finding
out.

thanks,

greg k-h

--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx
Please read the FAQ at http://kernelnewbies.org/FAQ