Re: [BK PATCH] devfs cleanups for 2.5.29

[Greg KH]

On Tue, Jul 30, 2002 at 05:12:07PM -0600, Richard Gooch wrote:
> Greg KH writes:
> > Hi,
> > 
> > When devfs came alone, it created devfs_[un]register_chrdev and
> > devfs_[un]register_blkdev, which required that all drivers be changed to
> > be compatible with devfs. This change has been bothering a lot of people
> > for quite some time :)
> > 
> > These two small changesets (patches to follow this email) fix that
> > problem by removing these functions, and having the original
> > [un]register_chrdev and [un]register_blkdev ask devfs if the operation
> > should be performed _if_ devfs is currently compiled into the kernel.
> > No functionality is changed, but the kernel code base is reduced, and we
> > are back to a common API.
> 
> Your patch misses the reason why I created those functions: some
> drivers had to always register with the major table. With your
> "fixups", those drivers will break when "devfs=only" is passed in. If
> you first fix the drivers so that they work without an entry in the
> major table, then your patch is safe to apply.

Ah, then this "feature" should be written down somewhere.  Which drivers
does this happen for?  And why penalize _all_ of the kernel drivers for
only the few that need this?

thanks,

greg k-h