On Monday, April 06 2009, Chris Lumens said:
> > > For filesystems that we officially support, there is no change here. For
> > > those that require a cmdline option for support, module loading has now
> > > moved from within the loader to inside the __init__ methods for the formats
> > > themselves. The intention here is to avoid kernel errors in modules that
> > > the user never even wants to have involved.
> >
> > Why not just always have it so that we load them in the __init__ if
> > needed? Then we could drop even more duplication between the loader and
> > liveinst.sh (which needs this change too)
>
> So, extend what I've already done to all the other filesystems too?
Exactly
> > So that would let us lose msdos, fat and xfs from this list
>
> Assuming udev does what it is supposed to, we could also get rid of
> vfat, ext4, and ext2 from the much earlier load in loader.c, right?
Yep
> > > + raise FSError("Could not load %s kernel module: %s" % (self._module, e))
> > > +
> > > + if rc:
> > > + raise FSError("Could not load %s kernel module." % self._module)
> >
> > Should we raise an error here or just make it so the fs is unsupported,
> > much like we do if the filesystem utils aren't available? To also
> > handle my thought, we'd need to check if the fs was listed in
> > /proc/filesystems before doing the modprobe as well
>
> I'm fine with making the FS unsupported, as long as we also log an
> error.
>
> Why do we need to check /proc/filesystems? Assuming trying to double
> load the module doesn't cause explosions, I can't see what the harm here
> would be.
Then we don't have to worry about whether something is built-in or
modular. We can see if the filesystem is supported, if not try to load
the module, if that fails, then it's not supported
Jeremy
_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/anaconda-devel-list
