Re: blank filenames in /sys/modules, kernel 2.6.21

[Greg KH <greg@kroah.com>]

On Fri, Jun 08, 2007 at 02:44:12PM -0400, Robert P. J. Day wrote:
> On Fri, 8 Jun 2007, Greg KH wrote:
> 
> > On Fri, Jun 08, 2007 at 02:05:22PM -0400, Robert P. J. Day wrote:
> > > On Fri, 8 Jun 2007, Chuck Ebbert wrote:
> > >
> > > > We just found a file with this name in 2.6.21:
> > > >
> > > > Directory name: /sys/module/nousb/parameters
> > > > File name:      ""
> > > >
> > > > IOW the file has an empty name. Is this intentional?
> > > > It sure confuses updatedb...
> > > >
> > > > https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=242715
> > >
> > > that appears to be the result of:
> > >
> > > $ grep -r __module_param_call *
> > > drivers/usb/core/usb.c:__module_param_call("", nousb, param_set_bool, param_get_bool, &nousb, 0444);
> > > include/linux/moduleparam.h:#define __module_param_call(prefix, name, set, get, arg, perm)              \
> > > include/linux/moduleparam.h:    __module_param_call(MODULE_PARAM_PREFIX, name, set, get, arg, perm)
> > >
> > >   and from drivers/usb/core/usb.c:
> > >
> > > /* format to disable USB on kernel command line is: nousb */
> > > __module_param_call("", nousb, param_set_bool, param_get_bool, &nousb, 0444);
> > >
> > >   one has to wonder about the value of a macro that is invoked only
> > > once throughout the entire tree.
> >
> > Pete added that back in December of 2005 (git-blame rocks...).  So what
> > has changed recently to cause this to do different things in sysfs?
> >
> > Robert, what do you suggest we change this usage to look like?
> 
> um ... i'm fairly sure i'm totally the wrong person to be asked about
> this, but what does any of the above have to do with disabling USB on
> the kernel command line?  wouldn't you typically define a kernel
> command line option with something like:
> ...
> arch/i386/kernel/traps.c:__setup("kstack=", kstack_setup);
> arch/i386/kernel/traps.c:__setup("code_bytes=", code_bytes_setup);
> arch/i386/kernel/nmi.c:__setup("nmi_watchdog=", setup_nmi_watchdog);
> arch/i386/kernel/reboot.c:__setup("reboot=", reboot_setup);
> arch/i386/kernel/tsc.c:__setup("notsc", tsc_setup);
> ...
> 
> you know -- that sort of thing.  i mean, that's where *i'd* introduce
> a "nousb" option.

It's a module parameter as well as a kernel command line option, see
Pete's response.

thanks,

greg k-h