Bug#296894: initrd-tools: function module_exist doesn't recognize modules in subdirs

[Cajus Pollmeier <pollmeier@gonicus.de>]

Am 01.03.2005 um 04:23 schrieb Horms:

> On Fri, Feb 25, 2005 at 02:30:16PM +0100, Cajus Pollmeier wrote:
> > Package: initrd-tools
> > Version: 0.1.77
> > Severity: normal
> >
> > While playing around with virtual scsi drivers, I noticed that modules
> > that are not directly located inside drivers/scsi where not found by
> > the script.
> >
> > The attached patch solves this problem.
>
> > --- mkinitrd	2005-01-23 19:37:41.000000000 +0100
> > +++ mkinitrd.patched	2005-02-25 14:10:50.919976624 +0100
> > @@ -413,7 +413,7 @@
> >  }
> >
> >  module_exists() {
> > -	[ -f "$MODULEDIR/kernel/$1.$o" ]
> > +	[ -n "$(find "$MODULEDIR/kernel/${1%%/*}" -name "${1##*/}.$o")" ]
> >  }
> >
> >  print_module() {
>
> That seems fair enough to me, my only reservation is speed.
> Did you notice a significant slowdown when using this?
> Also, could you be a litte more specific about which modules
> are not being found, and where they are coming from?

The find is taking place in a kernel module subdirectory and machines 
are fast enough so that I didn't recognize any changes ;-) The module 
I'm talking about is (even in vanilla kernels >= 2.6.9) ibmvscsic, a 
virtual scsi driver on IBM power4/5 systems. Stumbled here while 
adapting the debian installer to work on this hardware.

Cheers,
Cajus